diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -3,254900 +3,246703 @@
   "best_model_checkpoint": null,
   "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 36411,
+  "global_step": 35240,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 2.746422784323419e-05,
+      "epoch": 2.837684449489217e-05,
       "grad_norm": 0.0,
-      "learning_rate": 1.8298261665141813e-08,
-      "loss": 1.5936,
+      "learning_rate": 1.890359168241966e-08,
+      "loss": 1.577,
       "step": 1
     },
     {
-      "epoch": 5.492845568646838e-05,
+      "epoch": 5.675368898978434e-05,
       "grad_norm": 0.0,
-      "learning_rate": 3.659652333028363e-08,
-      "loss": 1.5551,
+      "learning_rate": 3.780718336483932e-08,
+      "loss": 1.5748,
       "step": 2
     },
     {
-      "epoch": 8.239268352970256e-05,
+      "epoch": 8.51305334846765e-05,
       "grad_norm": 0.0,
-      "learning_rate": 5.489478499542544e-08,
-      "loss": 1.6017,
+      "learning_rate": 5.671077504725899e-08,
+      "loss": 1.6437,
       "step": 3
     },
     {
-      "epoch": 0.00010985691137293675,
+      "epoch": 0.00011350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 7.319304666056725e-08,
-      "loss": 1.595,
+      "learning_rate": 7.561436672967865e-08,
+      "loss": 1.5787,
       "step": 4
     },
     {
-      "epoch": 0.00013732113921617094,
+      "epoch": 0.00014188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.149130832570908e-08,
-      "loss": 1.5455,
+      "learning_rate": 9.45179584120983e-08,
+      "loss": 1.6143,
       "step": 5
     },
     {
-      "epoch": 0.00016478536705940512,
+      "epoch": 0.000170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.0978956999085087e-07,
-      "loss": 1.6211,
+      "learning_rate": 1.1342155009451797e-07,
+      "loss": 1.6899,
       "step": 6
     },
     {
-      "epoch": 0.00019224959490263933,
+      "epoch": 0.0001986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.2808783165599268e-07,
-      "loss": 1.5092,
+      "learning_rate": 1.3232514177693763e-07,
+      "loss": 1.597,
       "step": 7
     },
     {
-      "epoch": 0.0002197138227458735,
+      "epoch": 0.00022701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.463860933211345e-07,
-      "loss": 1.5684,
+      "learning_rate": 1.512287334593573e-07,
+      "loss": 1.6447,
       "step": 8
     },
     {
-      "epoch": 0.0002471780505891077,
+      "epoch": 0.00025539160045402954,
       "grad_norm": 0.0,
-      "learning_rate": 1.646843549862763e-07,
-      "loss": 1.6221,
+      "learning_rate": 1.7013232514177695e-07,
+      "loss": 1.6228,
       "step": 9
     },
     {
-      "epoch": 0.00027464227843234187,
+      "epoch": 0.00028376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.8298261665141815e-07,
-      "loss": 1.3969,
+      "learning_rate": 1.890359168241966e-07,
+      "loss": 1.4889,
       "step": 10
     },
     {
-      "epoch": 0.00030210650627557605,
+      "epoch": 0.00031214528944381386,
       "grad_norm": 0.0,
-      "learning_rate": 2.0128087831655995e-07,
-      "loss": 1.563,
+      "learning_rate": 2.079395085066163e-07,
+      "loss": 1.5586,
       "step": 11
     },
     {
-      "epoch": 0.00032957073411881023,
+      "epoch": 0.000340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.1957913998170175e-07,
-      "loss": 1.6204,
+      "learning_rate": 2.2684310018903595e-07,
+      "loss": 1.5734,
       "step": 12
     },
     {
-      "epoch": 0.0003570349619620444,
+      "epoch": 0.0003688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 2.3787740164684357e-07,
-      "loss": 1.6775,
+      "learning_rate": 2.457466918714556e-07,
+      "loss": 1.5389,
       "step": 13
     },
     {
-      "epoch": 0.00038449918980527865,
+      "epoch": 0.0003972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 2.5617566331198537e-07,
-      "loss": 1.7063,
+      "learning_rate": 2.6465028355387527e-07,
+      "loss": 1.635,
       "step": 14
     },
     {
-      "epoch": 0.00041196341764851283,
+      "epoch": 0.0004256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 2.744739249771272e-07,
-      "loss": 1.6238,
+      "learning_rate": 2.835538752362949e-07,
+      "loss": 1.5775,
       "step": 15
     },
     {
-      "epoch": 0.000439427645491747,
+      "epoch": 0.0004540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 2.92772186642269e-07,
-      "loss": 1.6914,
+      "learning_rate": 3.024574669187146e-07,
+      "loss": 1.5248,
       "step": 16
     },
     {
-      "epoch": 0.0004668918733349812,
+      "epoch": 0.00048240635641316683,
       "grad_norm": 0.0,
-      "learning_rate": 3.1107044830741084e-07,
-      "loss": 1.5363,
+      "learning_rate": 3.2136105860113424e-07,
+      "loss": 1.5082,
       "step": 17
     },
     {
-      "epoch": 0.0004943561011782154,
+      "epoch": 0.0005107832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 3.293687099725526e-07,
-      "loss": 1.5259,
+      "learning_rate": 3.402646502835539e-07,
+      "loss": 1.3573,
       "step": 18
     },
     {
-      "epoch": 0.0005218203290214496,
+      "epoch": 0.0005391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 3.4766697163769443e-07,
-      "loss": 1.6575,
+      "learning_rate": 3.591682419659736e-07,
+      "loss": 1.5769,
       "step": 19
     },
     {
-      "epoch": 0.0005492845568646837,
+      "epoch": 0.0005675368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 3.659652333028363e-07,
-      "loss": 1.4421,
+      "learning_rate": 3.780718336483932e-07,
+      "loss": 1.4963,
       "step": 20
     },
     {
-      "epoch": 0.000576748784707918,
+      "epoch": 0.0005959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 3.842634949679781e-07,
-      "loss": 1.5054,
+      "learning_rate": 3.9697542533081287e-07,
+      "loss": 1.6018,
       "step": 21
     },
     {
-      "epoch": 0.0006042130125511521,
+      "epoch": 0.0006242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 4.025617566331199e-07,
-      "loss": 1.3625,
+      "learning_rate": 4.158790170132326e-07,
+      "loss": 1.5967,
       "step": 22
     },
     {
-      "epoch": 0.0006316772403943863,
+      "epoch": 0.0006526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 4.2086001829826167e-07,
-      "loss": 1.4928,
+      "learning_rate": 4.347826086956522e-07,
+      "loss": 1.4023,
       "step": 23
     },
     {
-      "epoch": 0.0006591414682376205,
+      "epoch": 0.000681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 4.391582799634035e-07,
-      "loss": 1.4324,
+      "learning_rate": 4.536862003780719e-07,
+      "loss": 1.4445,
       "step": 24
     },
     {
-      "epoch": 0.0006866056960808547,
+      "epoch": 0.0007094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.5745654162854537e-07,
-      "loss": 1.3921,
+      "learning_rate": 4.7258979206049156e-07,
+      "loss": 1.3446,
       "step": 25
     },
     {
-      "epoch": 0.0007140699239240888,
+      "epoch": 0.0007377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 4.7575480329368714e-07,
-      "loss": 1.502,
+      "learning_rate": 4.914933837429112e-07,
+      "loss": 1.4777,
       "step": 26
     },
     {
-      "epoch": 0.0007415341517673231,
+      "epoch": 0.0007661748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 4.94053064958829e-07,
-      "loss": 1.4207,
+      "learning_rate": 5.103969754253309e-07,
+      "loss": 1.3499,
       "step": 27
     },
     {
-      "epoch": 0.0007689983796105573,
+      "epoch": 0.0007945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.123513266239707e-07,
-      "loss": 1.3965,
+      "learning_rate": 5.293005671077505e-07,
+      "loss": 1.2587,
       "step": 28
     },
     {
-      "epoch": 0.0007964626074537914,
+      "epoch": 0.0008229284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 5.306495882891126e-07,
-      "loss": 1.4606,
+      "learning_rate": 5.482041587901702e-07,
+      "loss": 1.3668,
       "step": 29
     },
     {
-      "epoch": 0.0008239268352970257,
+      "epoch": 0.000851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 5.489478499542544e-07,
-      "loss": 1.4561,
+      "learning_rate": 5.671077504725898e-07,
+      "loss": 1.4567,
       "step": 30
     },
     {
-      "epoch": 0.0008513910631402598,
+      "epoch": 0.0008796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 5.672461116193962e-07,
-      "loss": 1.5658,
+      "learning_rate": 5.860113421550095e-07,
+      "loss": 1.4492,
       "step": 31
     },
     {
-      "epoch": 0.000878855290983494,
+      "epoch": 0.0009080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 5.85544373284538e-07,
-      "loss": 1.3806,
+      "learning_rate": 6.049149338374292e-07,
+      "loss": 1.3674,
       "step": 32
     },
     {
-      "epoch": 0.0009063195188267282,
+      "epoch": 0.0009364358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 6.038426349496798e-07,
-      "loss": 1.4355,
+      "learning_rate": 6.238185255198488e-07,
+      "loss": 1.3464,
       "step": 33
     },
     {
-      "epoch": 0.0009337837466699624,
+      "epoch": 0.0009648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 6.221408966148217e-07,
-      "loss": 1.3265,
+      "learning_rate": 6.427221172022685e-07,
+      "loss": 1.2973,
       "step": 34
     },
     {
-      "epoch": 0.0009612479745131965,
+      "epoch": 0.0009931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 6.404391582799636e-07,
-      "loss": 1.4443,
+      "learning_rate": 6.61625708884688e-07,
+      "loss": 1.4635,
       "step": 35
     },
     {
-      "epoch": 0.0009887122023564308,
+      "epoch": 0.0010215664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 6.587374199451052e-07,
-      "loss": 1.3655,
+      "learning_rate": 6.805293005671078e-07,
+      "loss": 1.2485,
       "step": 36
     },
     {
-      "epoch": 0.0010161764301996649,
+      "epoch": 0.0010499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 6.770356816102471e-07,
-      "loss": 1.4266,
+      "learning_rate": 6.994328922495275e-07,
+      "loss": 1.357,
       "step": 37
     },
     {
-      "epoch": 0.0010436406580428992,
+      "epoch": 0.0010783200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 6.953339432753889e-07,
-      "loss": 1.3354,
+      "learning_rate": 7.183364839319472e-07,
+      "loss": 1.4143,
       "step": 38
     },
     {
-      "epoch": 0.0010711048858861334,
+      "epoch": 0.0011066969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 7.136322049405307e-07,
-      "loss": 1.3412,
+      "learning_rate": 7.372400756143668e-07,
+      "loss": 1.305,
       "step": 39
     },
     {
-      "epoch": 0.0010985691137293675,
+      "epoch": 0.0011350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 7.319304666056726e-07,
-      "loss": 1.3121,
+      "learning_rate": 7.561436672967864e-07,
+      "loss": 1.2737,
       "step": 40
     },
     {
-      "epoch": 0.0011260333415726016,
+      "epoch": 0.001163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 7.502287282708143e-07,
-      "loss": 1.2714,
+      "learning_rate": 7.750472589792062e-07,
+      "loss": 1.2682,
       "step": 41
     },
     {
-      "epoch": 0.001153497569415836,
+      "epoch": 0.001191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 7.685269899359562e-07,
-      "loss": 1.3658,
+      "learning_rate": 7.939508506616257e-07,
+      "loss": 1.3037,
       "step": 42
     },
     {
-      "epoch": 0.00118096179725907,
+      "epoch": 0.0012202043132803631,
       "grad_norm": 0.0,
-      "learning_rate": 7.868252516010979e-07,
-      "loss": 1.321,
+      "learning_rate": 8.128544423440454e-07,
+      "loss": 1.3223,
       "step": 43
     },
     {
-      "epoch": 0.0012084260251023042,
+      "epoch": 0.0012485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 8.051235132662398e-07,
-      "loss": 1.279,
+      "learning_rate": 8.317580340264652e-07,
+      "loss": 1.2573,
       "step": 44
     },
     {
-      "epoch": 0.0012358902529455383,
+      "epoch": 0.0012769580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 8.234217749313817e-07,
-      "loss": 1.4259,
+      "learning_rate": 8.506616257088847e-07,
+      "loss": 1.3235,
       "step": 45
     },
     {
-      "epoch": 0.0012633544807887727,
+      "epoch": 0.0013053348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 8.417200365965233e-07,
-      "loss": 1.2783,
+      "learning_rate": 8.695652173913044e-07,
+      "loss": 1.0813,
       "step": 46
     },
     {
-      "epoch": 0.0012908187086320068,
+      "epoch": 0.001333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 8.600182982616652e-07,
-      "loss": 1.2484,
+      "learning_rate": 8.884688090737241e-07,
+      "loss": 1.4175,
       "step": 47
     },
     {
-      "epoch": 0.001318282936475241,
+      "epoch": 0.001362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 8.78316559926807e-07,
-      "loss": 1.4684,
+      "learning_rate": 9.073724007561438e-07,
+      "loss": 1.1378,
       "step": 48
     },
     {
-      "epoch": 0.0013457471643184753,
+      "epoch": 0.0013904653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 8.966148215919489e-07,
-      "loss": 1.343,
+      "learning_rate": 9.262759924385634e-07,
+      "loss": 1.3513,
       "step": 49
     },
     {
-      "epoch": 0.0013732113921617094,
+      "epoch": 0.0014188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.149130832570907e-07,
-      "loss": 1.3803,
+      "learning_rate": 9.451795841209831e-07,
+      "loss": 1.1996,
       "step": 50
     },
     {
-      "epoch": 0.0014006756200049435,
+      "epoch": 0.0014472190692395007,
       "grad_norm": 0.0,
-      "learning_rate": 9.332113449222324e-07,
-      "loss": 1.2894,
+      "learning_rate": 9.640831758034028e-07,
+      "loss": 1.1925,
       "step": 51
     },
     {
-      "epoch": 0.0014281398478481777,
+      "epoch": 0.0014755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 9.515096065873743e-07,
-      "loss": 1.3667,
+      "learning_rate": 9.829867674858224e-07,
+      "loss": 1.158,
       "step": 52
     },
     {
-      "epoch": 0.001455604075691412,
+      "epoch": 0.0015039727582292848,
       "grad_norm": 0.0,
-      "learning_rate": 9.698078682525162e-07,
-      "loss": 1.3089,
+      "learning_rate": 1.001890359168242e-06,
+      "loss": 1.2653,
       "step": 53
     },
     {
-      "epoch": 0.0014830683035346461,
+      "epoch": 0.0015323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 9.88106129917658e-07,
-      "loss": 1.3933,
+      "learning_rate": 1.0207939508506617e-06,
+      "loss": 1.3096,
       "step": 54
     },
     {
-      "epoch": 0.0015105325313778803,
+      "epoch": 0.0015607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.0064043915827997e-06,
-      "loss": 1.3556,
+      "learning_rate": 1.0396975425330814e-06,
+      "loss": 1.3491,
       "step": 55
     },
     {
-      "epoch": 0.0015379967592211146,
+      "epoch": 0.0015891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.0247026532479415e-06,
-      "loss": 1.3142,
+      "learning_rate": 1.058601134215501e-06,
+      "loss": 1.2775,
       "step": 56
     },
     {
-      "epoch": 0.0015654609870643487,
+      "epoch": 0.0016174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.0430009149130832e-06,
-      "loss": 1.2067,
+      "learning_rate": 1.0775047258979207e-06,
+      "loss": 1.3036,
       "step": 57
     },
     {
-      "epoch": 0.0015929252149075829,
+      "epoch": 0.0016458569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.0612991765782252e-06,
-      "loss": 1.2626,
+      "learning_rate": 1.0964083175803404e-06,
+      "loss": 1.2415,
       "step": 58
     },
     {
-      "epoch": 0.001620389442750817,
+      "epoch": 0.001674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.079597438243367e-06,
-      "loss": 1.2023,
+      "learning_rate": 1.11531190926276e-06,
+      "loss": 1.2456,
       "step": 59
     },
     {
-      "epoch": 0.0016478536705940513,
+      "epoch": 0.00170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.0978956999085088e-06,
-      "loss": 1.303,
+      "learning_rate": 1.1342155009451797e-06,
+      "loss": 1.2328,
       "step": 60
     },
     {
-      "epoch": 0.0016753178984372855,
+      "epoch": 0.0017309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.1161939615736505e-06,
-      "loss": 1.2014,
+      "learning_rate": 1.1531190926275994e-06,
+      "loss": 1.1286,
       "step": 61
     },
     {
-      "epoch": 0.0017027821262805196,
+      "epoch": 0.0017593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.1344922232387923e-06,
-      "loss": 1.2105,
+      "learning_rate": 1.172022684310019e-06,
+      "loss": 1.3732,
       "step": 62
     },
     {
-      "epoch": 0.0017302463541237537,
+      "epoch": 0.0017877412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.1527904849039343e-06,
-      "loss": 1.2161,
+      "learning_rate": 1.1909262759924387e-06,
+      "loss": 1.1947,
       "step": 63
     },
     {
-      "epoch": 0.001757710581966988,
+      "epoch": 0.0018161180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.171088746569076e-06,
-      "loss": 1.28,
+      "learning_rate": 1.2098298676748583e-06,
+      "loss": 1.161,
       "step": 64
     },
     {
-      "epoch": 0.0017851748098102222,
+      "epoch": 0.0018444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.1893870082342178e-06,
-      "loss": 1.2353,
+      "learning_rate": 1.228733459357278e-06,
+      "loss": 1.1613,
       "step": 65
     },
     {
-      "epoch": 0.0018126390376534563,
+      "epoch": 0.0018728717366628832,
       "grad_norm": 0.0,
-      "learning_rate": 1.2076852698993596e-06,
-      "loss": 1.2277,
+      "learning_rate": 1.2476370510396976e-06,
+      "loss": 1.1978,
       "step": 66
     },
     {
-      "epoch": 0.0018401032654966907,
+      "epoch": 0.0019012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.2259835315645016e-06,
-      "loss": 1.3584,
+      "learning_rate": 1.266540642722117e-06,
+      "loss": 1.2115,
       "step": 67
     },
     {
-      "epoch": 0.0018675674933399248,
+      "epoch": 0.0019296254256526673,
       "grad_norm": 0.0,
-      "learning_rate": 1.2442817932296434e-06,
-      "loss": 1.3751,
+      "learning_rate": 1.285444234404537e-06,
+      "loss": 1.1695,
       "step": 68
     },
     {
-      "epoch": 0.001895031721183159,
+      "epoch": 0.0019580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.2625800548947851e-06,
-      "loss": 1.3365,
+      "learning_rate": 1.3043478260869566e-06,
+      "loss": 1.3129,
       "step": 69
     },
     {
-      "epoch": 0.001922495949026393,
+      "epoch": 0.0019863791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.280878316559927e-06,
-      "loss": 1.2231,
+      "learning_rate": 1.323251417769376e-06,
+      "loss": 1.1815,
       "step": 70
     },
     {
-      "epoch": 0.0019499601768696274,
+      "epoch": 0.002014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.2991765782250687e-06,
-      "loss": 1.2771,
+      "learning_rate": 1.342155009451796e-06,
+      "loss": 1.2826,
       "step": 71
     },
     {
-      "epoch": 0.0019774244047128615,
+      "epoch": 0.0020431328036322363,
       "grad_norm": 0.0,
-      "learning_rate": 1.3174748398902104e-06,
-      "loss": 1.2168,
+      "learning_rate": 1.3610586011342156e-06,
+      "loss": 1.2458,
       "step": 72
     },
     {
-      "epoch": 0.002004888632556096,
+      "epoch": 0.0020715096481271284,
       "grad_norm": 0.0,
-      "learning_rate": 1.3357731015553524e-06,
-      "loss": 1.2792,
+      "learning_rate": 1.3799621928166355e-06,
+      "loss": 1.0872,
       "step": 73
     },
     {
-      "epoch": 0.0020323528603993298,
+      "epoch": 0.0020998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.3540713632204942e-06,
-      "loss": 1.2772,
+      "learning_rate": 1.398865784499055e-06,
+      "loss": 1.2012,
       "step": 74
     },
     {
-      "epoch": 0.002059817088242564,
+      "epoch": 0.0021282633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.3723696248856362e-06,
-      "loss": 1.1987,
+      "learning_rate": 1.4177693761814746e-06,
+      "loss": 1.1517,
       "step": 75
     },
     {
-      "epoch": 0.0020872813160857985,
+      "epoch": 0.0021566401816118046,
       "grad_norm": 0.0,
-      "learning_rate": 1.3906678865507777e-06,
-      "loss": 1.2405,
+      "learning_rate": 1.4366729678638944e-06,
+      "loss": 1.1891,
       "step": 76
     },
     {
-      "epoch": 0.0021147455439290324,
+      "epoch": 0.002185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.4089661482159195e-06,
-      "loss": 1.2092,
+      "learning_rate": 1.4555765595463139e-06,
+      "loss": 1.226,
       "step": 77
     },
     {
-      "epoch": 0.0021422097717722667,
+      "epoch": 0.002213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.4272644098810615e-06,
-      "loss": 1.2432,
+      "learning_rate": 1.4744801512287335e-06,
+      "loss": 1.1941,
       "step": 78
     },
     {
-      "epoch": 0.0021696739996155006,
+      "epoch": 0.0022417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.4455626715462032e-06,
-      "loss": 1.2573,
+      "learning_rate": 1.4933837429111534e-06,
+      "loss": 1.2591,
       "step": 79
     },
     {
-      "epoch": 0.002197138227458735,
+      "epoch": 0.0022701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.4638609332113452e-06,
-      "loss": 1.3247,
+      "learning_rate": 1.5122873345935729e-06,
+      "loss": 1.301,
       "step": 80
     },
     {
-      "epoch": 0.0022246024553019693,
+      "epoch": 0.0022985244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.4821591948764868e-06,
-      "loss": 1.2956,
+      "learning_rate": 1.5311909262759925e-06,
+      "loss": 1.174,
       "step": 81
     },
     {
-      "epoch": 0.0022520666831452032,
+      "epoch": 0.002326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.5004574565416286e-06,
-      "loss": 1.2766,
+      "learning_rate": 1.5500945179584124e-06,
+      "loss": 1.1715,
       "step": 82
     },
     {
-      "epoch": 0.0022795309109884376,
+      "epoch": 0.00235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5187557182067705e-06,
-      "loss": 1.2224,
+      "learning_rate": 1.5689981096408318e-06,
+      "loss": 1.13,
       "step": 83
     },
     {
-      "epoch": 0.002306995138831672,
+      "epoch": 0.002383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.5370539798719123e-06,
-      "loss": 1.0414,
+      "learning_rate": 1.5879017013232515e-06,
+      "loss": 1.2639,
       "step": 84
     },
     {
-      "epoch": 0.002334459366674906,
+      "epoch": 0.002412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.5553522415370543e-06,
-      "loss": 1.2047,
+      "learning_rate": 1.6068052930056714e-06,
+      "loss": 1.2665,
       "step": 85
     },
     {
-      "epoch": 0.00236192359451814,
+      "epoch": 0.0024404086265607263,
       "grad_norm": 0.0,
-      "learning_rate": 1.5736505032021959e-06,
-      "loss": 1.2291,
+      "learning_rate": 1.6257088846880908e-06,
+      "loss": 1.2371,
       "step": 86
     },
     {
-      "epoch": 0.0023893878223613745,
+      "epoch": 0.002468785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.5919487648673376e-06,
-      "loss": 1.2036,
+      "learning_rate": 1.6446124763705105e-06,
+      "loss": 1.1932,
       "step": 87
     },
     {
-      "epoch": 0.0024168520502046084,
+      "epoch": 0.002497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.6102470265324796e-06,
-      "loss": 1.1955,
+      "learning_rate": 1.6635160680529303e-06,
+      "loss": 1.1747,
       "step": 88
     },
     {
-      "epoch": 0.0024443162780478428,
+      "epoch": 0.002525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.6285452881976214e-06,
-      "loss": 1.2741,
+      "learning_rate": 1.6824196597353498e-06,
+      "loss": 1.2115,
       "step": 89
     },
     {
-      "epoch": 0.0024717805058910767,
+      "epoch": 0.002553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.6468435498627634e-06,
-      "loss": 1.2816,
+      "learning_rate": 1.7013232514177694e-06,
+      "loss": 1.1421,
       "step": 90
     },
     {
-      "epoch": 0.002499244733734311,
+      "epoch": 0.002582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.665141811527905e-06,
-      "loss": 1.3589,
+      "learning_rate": 1.7202268431001893e-06,
+      "loss": 1.0763,
       "step": 91
     },
     {
-      "epoch": 0.0025267089615775454,
+      "epoch": 0.0026106696935300796,
       "grad_norm": 0.0,
-      "learning_rate": 1.6834400731930467e-06,
-      "loss": 1.2254,
+      "learning_rate": 1.7391304347826088e-06,
+      "loss": 1.237,
       "step": 92
     },
     {
-      "epoch": 0.0025541731894207793,
+      "epoch": 0.0026390465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.7017383348581887e-06,
-      "loss": 1.2515,
+      "learning_rate": 1.7580340264650286e-06,
+      "loss": 1.2073,
       "step": 93
     },
     {
-      "epoch": 0.0025816374172640136,
+      "epoch": 0.002667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.7200365965233304e-06,
-      "loss": 1.2297,
+      "learning_rate": 1.7769376181474483e-06,
+      "loss": 1.1453,
       "step": 94
     },
     {
-      "epoch": 0.002609101645107248,
+      "epoch": 0.002695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.7383348581884724e-06,
-      "loss": 1.1483,
+      "learning_rate": 1.7958412098298677e-06,
+      "loss": 1.1752,
       "step": 95
     },
     {
-      "epoch": 0.002636565872950482,
+      "epoch": 0.002724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.756633119853614e-06,
-      "loss": 1.2375,
+      "learning_rate": 1.8147448015122876e-06,
+      "loss": 1.2231,
       "step": 96
     },
     {
-      "epoch": 0.002664030100793716,
+      "epoch": 0.0027525539160045405,
       "grad_norm": 0.0,
-      "learning_rate": 1.7749313815187558e-06,
-      "loss": 1.2148,
+      "learning_rate": 1.8336483931947073e-06,
+      "loss": 1.0829,
       "step": 97
     },
     {
-      "epoch": 0.0026914943286369506,
+      "epoch": 0.0027809307604994326,
       "grad_norm": 0.0,
-      "learning_rate": 1.7932296431838977e-06,
-      "loss": 1.1724,
+      "learning_rate": 1.8525519848771267e-06,
+      "loss": 1.2483,
       "step": 98
     },
     {
-      "epoch": 0.0027189585564801845,
+      "epoch": 0.0028093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8115279048490395e-06,
-      "loss": 1.1733,
+      "learning_rate": 1.8714555765595466e-06,
+      "loss": 1.1451,
       "step": 99
     },
     {
-      "epoch": 0.002746422784323419,
+      "epoch": 0.0028376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.8298261665141815e-06,
-      "loss": 1.3197,
+      "learning_rate": 1.8903591682419662e-06,
+      "loss": 1.1886,
       "step": 100
     },
     {
-      "epoch": 0.0027738870121666527,
+      "epoch": 0.002866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.848124428179323e-06,
-      "loss": 1.1848,
+      "learning_rate": 1.909262759924386e-06,
+      "loss": 1.0508,
       "step": 101
     },
     {
-      "epoch": 0.002801351240009887,
+      "epoch": 0.0028944381384790013,
       "grad_norm": 0.0,
-      "learning_rate": 1.8664226898444648e-06,
-      "loss": 1.1934,
+      "learning_rate": 1.9281663516068055e-06,
+      "loss": 1.0985,
       "step": 102
     },
     {
-      "epoch": 0.0028288154678531214,
+      "epoch": 0.0029228149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.8847209515096068e-06,
-      "loss": 1.2139,
+      "learning_rate": 1.947069943289225e-06,
+      "loss": 1.2544,
       "step": 103
     },
     {
-      "epoch": 0.0028562796956963553,
+      "epoch": 0.0029511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.9030192131747486e-06,
-      "loss": 1.2678,
+      "learning_rate": 1.965973534971645e-06,
+      "loss": 1.2699,
       "step": 104
     },
     {
-      "epoch": 0.0028837439235395897,
+      "epoch": 0.0029795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9213174748398906e-06,
-      "loss": 1.2308,
+      "learning_rate": 1.9848771266540645e-06,
+      "loss": 1.261,
       "step": 105
     },
     {
-      "epoch": 0.002911208151382824,
+      "epoch": 0.0030079455164585696,
       "grad_norm": 0.0,
-      "learning_rate": 1.9396157365050323e-06,
-      "loss": 1.2008,
+      "learning_rate": 2.003780718336484e-06,
+      "loss": 1.1757,
       "step": 106
     },
     {
-      "epoch": 0.002938672379226058,
+      "epoch": 0.003036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.957913998170174e-06,
-      "loss": 1.101,
+      "learning_rate": 2.022684310018904e-06,
+      "loss": 1.0599,
       "step": 107
     },
     {
-      "epoch": 0.0029661366070692923,
+      "epoch": 0.0030646992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 1.976212259835316e-06,
-      "loss": 1.2411,
+      "learning_rate": 2.0415879017013235e-06,
+      "loss": 1.1872,
       "step": 108
     },
     {
-      "epoch": 0.0029936008349125266,
+      "epoch": 0.0030930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945105215004576e-06,
-      "loss": 1.2816,
+      "learning_rate": 2.060491493383743e-06,
+      "loss": 1.1823,
       "step": 109
     },
     {
-      "epoch": 0.0030210650627557605,
+      "epoch": 0.0031214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 2.0128087831655994e-06,
-      "loss": 1.1897,
+      "learning_rate": 2.079395085066163e-06,
+      "loss": 1.1038,
       "step": 110
     },
     {
-      "epoch": 0.003048529290598995,
+      "epoch": 0.0031498297389330305,
       "grad_norm": 0.0,
-      "learning_rate": 2.031107044830741e-06,
-      "loss": 1.2439,
+      "learning_rate": 2.0982986767485825e-06,
+      "loss": 1.1498,
       "step": 111
     },
     {
-      "epoch": 0.003075993518442229,
+      "epoch": 0.003178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 2.049405306495883e-06,
-      "loss": 1.1527,
+      "learning_rate": 2.117202268431002e-06,
+      "loss": 1.2322,
       "step": 112
     },
     {
-      "epoch": 0.003103457746285463,
+      "epoch": 0.003206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 2.067703568161025e-06,
-      "loss": 1.2227,
+      "learning_rate": 2.1361058601134218e-06,
+      "loss": 1.0839,
       "step": 113
     },
     {
-      "epoch": 0.0031309219741286975,
+      "epoch": 0.003234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 2.0860018298261665e-06,
-      "loss": 1.2294,
+      "learning_rate": 2.1550094517958414e-06,
+      "loss": 1.0966,
       "step": 114
     },
     {
-      "epoch": 0.0031583862019719314,
+      "epoch": 0.0032633371169125992,
       "grad_norm": 0.0,
-      "learning_rate": 2.1043000914913083e-06,
-      "loss": 1.1948,
+      "learning_rate": 2.173913043478261e-06,
+      "loss": 1.1529,
       "step": 115
     },
     {
-      "epoch": 0.0031858504298151657,
+      "epoch": 0.0032917139614074913,
       "grad_norm": 0.0,
-      "learning_rate": 2.1225983531564504e-06,
-      "loss": 1.0785,
+      "learning_rate": 2.1928166351606808e-06,
+      "loss": 1.1382,
       "step": 116
     },
     {
-      "epoch": 0.0032133146576584,
+      "epoch": 0.003320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 2.1408966148215922e-06,
-      "loss": 1.1322,
+      "learning_rate": 2.2117202268431004e-06,
+      "loss": 1.0868,
       "step": 117
     },
     {
-      "epoch": 0.003240778885501634,
+      "epoch": 0.003348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 2.159194876486734e-06,
-      "loss": 1.2626,
+      "learning_rate": 2.23062381852552e-06,
+      "loss": 1.1593,
       "step": 118
     },
     {
-      "epoch": 0.0032682431133448683,
+      "epoch": 0.003376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.1774931381518758e-06,
-      "loss": 1.2365,
+      "learning_rate": 2.2495274102079397e-06,
+      "loss": 1.0725,
       "step": 119
     },
     {
-      "epoch": 0.0032957073411881027,
+      "epoch": 0.00340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.1957913998170175e-06,
-      "loss": 1.1459,
+      "learning_rate": 2.2684310018903594e-06,
+      "loss": 1.0975,
       "step": 120
     },
     {
-      "epoch": 0.0033231715690313366,
+      "epoch": 0.003433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 2.2140896614821593e-06,
-      "loss": 1.1844,
+      "learning_rate": 2.287334593572779e-06,
+      "loss": 1.0643,
       "step": 121
     },
     {
-      "epoch": 0.003350635796874571,
+      "epoch": 0.0034619750283768447,
       "grad_norm": 0.0,
-      "learning_rate": 2.232387923147301e-06,
-      "loss": 1.3112,
+      "learning_rate": 2.3062381852551987e-06,
+      "loss": 1.2552,
       "step": 122
     },
     {
-      "epoch": 0.0033781000247178053,
+      "epoch": 0.0034903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 2.2506861848124433e-06,
-      "loss": 1.1517,
+      "learning_rate": 2.3251417769376184e-06,
+      "loss": 1.2235,
       "step": 123
     },
     {
-      "epoch": 0.003405564252561039,
+      "epoch": 0.003518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 2.2689844464775846e-06,
-      "loss": 1.2363,
+      "learning_rate": 2.344045368620038e-06,
+      "loss": 1.3259,
       "step": 124
     },
     {
-      "epoch": 0.0034330284804042735,
+      "epoch": 0.003547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.2872827081427264e-06,
-      "loss": 1.242,
+      "learning_rate": 2.3629489603024577e-06,
+      "loss": 1.1089,
       "step": 125
     },
     {
-      "epoch": 0.0034604927082475074,
+      "epoch": 0.003575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 2.3055809698078686e-06,
-      "loss": 1.2004,
+      "learning_rate": 2.3818525519848773e-06,
+      "loss": 1.092,
       "step": 126
     },
     {
-      "epoch": 0.0034879569360907418,
+      "epoch": 0.0036038592508513055,
       "grad_norm": 0.0,
-      "learning_rate": 2.3238792314730103e-06,
-      "loss": 1.2158,
+      "learning_rate": 2.400756143667297e-06,
+      "loss": 1.2257,
       "step": 127
     },
     {
-      "epoch": 0.003515421163933976,
+      "epoch": 0.0036322360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 2.342177493138152e-06,
-      "loss": 1.1562,
+      "learning_rate": 2.4196597353497167e-06,
+      "loss": 1.1103,
       "step": 128
     },
     {
-      "epoch": 0.00354288539177721,
+      "epoch": 0.0036606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 2.360475754803294e-06,
-      "loss": 1.2055,
+      "learning_rate": 2.4385633270321363e-06,
+      "loss": 1.0021,
       "step": 129
     },
     {
-      "epoch": 0.0035703496196204444,
+      "epoch": 0.0036889897843359817,
       "grad_norm": 0.0,
-      "learning_rate": 2.3787740164684357e-06,
-      "loss": 1.1396,
+      "learning_rate": 2.457466918714556e-06,
+      "loss": 1.2441,
       "step": 130
     },
     {
-      "epoch": 0.0035978138474636787,
+      "epoch": 0.003717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.3970722781335774e-06,
-      "loss": 1.179,
+      "learning_rate": 2.4763705103969756e-06,
+      "loss": 1.1216,
       "step": 131
     },
     {
-      "epoch": 0.0036252780753069126,
+      "epoch": 0.0037457434733257663,
       "grad_norm": 0.0,
-      "learning_rate": 2.415370539798719e-06,
-      "loss": 1.2169,
+      "learning_rate": 2.4952741020793953e-06,
+      "loss": 1.09,
       "step": 132
     },
     {
-      "epoch": 0.003652742303150147,
+      "epoch": 0.0037741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 2.4336688014638614e-06,
-      "loss": 1.1897,
+      "learning_rate": 2.514177693761815e-06,
+      "loss": 1.2769,
       "step": 133
     },
     {
-      "epoch": 0.0036802065309933813,
+      "epoch": 0.0038024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 2.451967063129003e-06,
-      "loss": 1.1252,
+      "learning_rate": 2.533081285444234e-06,
+      "loss": 1.1239,
       "step": 134
     },
     {
-      "epoch": 0.0037076707588366152,
+      "epoch": 0.0038308740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 2.4702653247941445e-06,
-      "loss": 1.0972,
+      "learning_rate": 2.5519848771266543e-06,
+      "loss": 1.1168,
       "step": 135
     },
     {
-      "epoch": 0.0037351349866798496,
+      "epoch": 0.0038592508513053347,
       "grad_norm": 0.0,
-      "learning_rate": 2.4885635864592867e-06,
-      "loss": 1.1776,
+      "learning_rate": 2.570888468809074e-06,
+      "loss": 1.1944,
       "step": 136
     },
     {
-      "epoch": 0.0037625992145230835,
+      "epoch": 0.003887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.5068618481244285e-06,
-      "loss": 1.2098,
+      "learning_rate": 2.589792060491493e-06,
+      "loss": 1.181,
       "step": 137
     },
     {
-      "epoch": 0.003790063442366318,
+      "epoch": 0.003916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 2.5251601097895702e-06,
-      "loss": 1.1565,
+      "learning_rate": 2.6086956521739132e-06,
+      "loss": 1.1676,
       "step": 138
     },
     {
-      "epoch": 0.003817527670209552,
+      "epoch": 0.003944381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 2.5434583714547116e-06,
-      "loss": 1.1798,
+      "learning_rate": 2.627599243856333e-06,
+      "loss": 1.1534,
       "step": 139
     },
     {
-      "epoch": 0.003844991898052786,
+      "epoch": 0.003972758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 2.561756633119854e-06,
-      "loss": 1.2996,
+      "learning_rate": 2.646502835538752e-06,
+      "loss": 1.1686,
       "step": 140
     },
     {
-      "epoch": 0.0038724561258960204,
+      "epoch": 0.004001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 2.5800548947849956e-06,
-      "loss": 1.1483,
+      "learning_rate": 2.6654064272211722e-06,
+      "loss": 1.1894,
       "step": 141
     },
     {
-      "epoch": 0.0038999203537392548,
+      "epoch": 0.004029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 2.5983531564501373e-06,
-      "loss": 1.2255,
+      "learning_rate": 2.684310018903592e-06,
+      "loss": 1.1708,
       "step": 142
     },
     {
-      "epoch": 0.003927384581582489,
+      "epoch": 0.00405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.616651418115279e-06,
-      "loss": 1.2207,
+      "learning_rate": 2.703213610586012e-06,
+      "loss": 1.2236,
       "step": 143
     },
     {
-      "epoch": 0.003954848809425723,
+      "epoch": 0.004086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 2.634949679780421e-06,
-      "loss": 1.2468,
+      "learning_rate": 2.722117202268431e-06,
+      "loss": 1.084,
       "step": 144
     },
     {
-      "epoch": 0.003982313037268957,
+      "epoch": 0.004114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 2.653247941445563e-06,
-      "loss": 1.1663,
+      "learning_rate": 2.741020793950851e-06,
+      "loss": 1.0896,
       "step": 145
     },
     {
-      "epoch": 0.004009777265112192,
+      "epoch": 0.004143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 2.671546203110705e-06,
-      "loss": 1.1552,
+      "learning_rate": 2.759924385633271e-06,
+      "loss": 1.2076,
       "step": 146
     },
     {
-      "epoch": 0.004037241492955425,
+      "epoch": 0.004171396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 2.6898444647758466e-06,
-      "loss": 1.2036,
+      "learning_rate": 2.77882797731569e-06,
+      "loss": 1.0709,
       "step": 147
     },
     {
-      "epoch": 0.0040647057207986595,
+      "epoch": 0.004199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 2.7081427264409884e-06,
-      "loss": 1.1792,
+      "learning_rate": 2.79773156899811e-06,
+      "loss": 1.2851,
       "step": 148
     },
     {
-      "epoch": 0.004092169948641894,
+      "epoch": 0.0042281498297389334,
       "grad_norm": 0.0,
-      "learning_rate": 2.7264409881061297e-06,
-      "loss": 1.21,
+      "learning_rate": 2.81663516068053e-06,
+      "loss": 1.1168,
       "step": 149
     },
     {
-      "epoch": 0.004119634176485128,
+      "epoch": 0.004256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 2.7447392497712723e-06,
-      "loss": 1.1571,
+      "learning_rate": 2.835538752362949e-06,
+      "loss": 1.1165,
       "step": 150
     },
     {
-      "epoch": 0.0041470984043283626,
+      "epoch": 0.004284903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 2.7630375114364137e-06,
-      "loss": 1.1296,
+      "learning_rate": 2.854442344045369e-06,
+      "loss": 1.2343,
       "step": 151
     },
     {
-      "epoch": 0.004174562632171597,
+      "epoch": 0.004313280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 2.7813357731015555e-06,
-      "loss": 1.1715,
+      "learning_rate": 2.873345935727789e-06,
+      "loss": 1.1205,
       "step": 152
     },
     {
-      "epoch": 0.00420202686001483,
+      "epoch": 0.004341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 2.7996340347666972e-06,
-      "loss": 1.1208,
+      "learning_rate": 2.892249527410208e-06,
+      "loss": 1.1872,
       "step": 153
     },
     {
-      "epoch": 0.004229491087858065,
+      "epoch": 0.004370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 2.817932296431839e-06,
-      "loss": 1.0857,
+      "learning_rate": 2.9111531190926278e-06,
+      "loss": 1.0697,
       "step": 154
     },
     {
-      "epoch": 0.004256955315701299,
+      "epoch": 0.004398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.836230558096981e-06,
-      "loss": 1.1745,
+      "learning_rate": 2.930056710775048e-06,
+      "loss": 1.1793,
       "step": 155
     },
     {
-      "epoch": 0.004284419543544533,
+      "epoch": 0.004426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 2.854528819762123e-06,
-      "loss": 1.3831,
+      "learning_rate": 2.948960302457467e-06,
+      "loss": 1.1258,
       "step": 156
     },
     {
-      "epoch": 0.004311883771387768,
+      "epoch": 0.00445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 2.8728270814272647e-06,
-      "loss": 1.1889,
+      "learning_rate": 2.9678638941398867e-06,
+      "loss": 1.1127,
       "step": 157
     },
     {
-      "epoch": 0.004339347999231001,
+      "epoch": 0.004483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 2.8911253430924065e-06,
-      "loss": 1.1417,
+      "learning_rate": 2.986767485822307e-06,
+      "loss": 1.1584,
       "step": 158
     },
     {
-      "epoch": 0.004366812227074236,
+      "epoch": 0.004511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 2.909423604757548e-06,
-      "loss": 1.2154,
+      "learning_rate": 3.005671077504726e-06,
+      "loss": 1.0905,
       "step": 159
     },
     {
-      "epoch": 0.00439427645491747,
+      "epoch": 0.004540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 2.9277218664226905e-06,
-      "loss": 1.2657,
+      "learning_rate": 3.0245746691871457e-06,
+      "loss": 1.2025,
       "step": 160
     },
     {
-      "epoch": 0.004421740682760704,
+      "epoch": 0.004568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 2.946020128087832e-06,
-      "loss": 1.1462,
+      "learning_rate": 3.043478260869566e-06,
+      "loss": 1.2253,
       "step": 161
     },
     {
-      "epoch": 0.004449204910603939,
+      "epoch": 0.004597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 2.9643183897529736e-06,
-      "loss": 1.2156,
+      "learning_rate": 3.062381852551985e-06,
+      "loss": 1.2033,
       "step": 162
     },
     {
-      "epoch": 0.004476669138447173,
+      "epoch": 0.004625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 2.9826166514181154e-06,
-      "loss": 1.1558,
+      "learning_rate": 3.0812854442344047e-06,
+      "loss": 1.1696,
       "step": 163
     },
     {
-      "epoch": 0.0045041333662904064,
+      "epoch": 0.004653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 3.000914913083257e-06,
-      "loss": 1.1626,
+      "learning_rate": 3.1001890359168248e-06,
+      "loss": 1.1219,
       "step": 164
     },
     {
-      "epoch": 0.004531597594133641,
+      "epoch": 0.004682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 3.0192131747483993e-06,
-      "loss": 1.1611,
+      "learning_rate": 3.119092627599244e-06,
+      "loss": 1.0379,
       "step": 165
     },
     {
-      "epoch": 0.004559061821976875,
+      "epoch": 0.0047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.037511436413541e-06,
-      "loss": 1.1149,
+      "learning_rate": 3.1379962192816637e-06,
+      "loss": 1.1075,
       "step": 166
     },
     {
-      "epoch": 0.0045865260498201095,
+      "epoch": 0.004738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 3.055809698078683e-06,
-      "loss": 1.1115,
+      "learning_rate": 3.1568998109640837e-06,
+      "loss": 1.0141,
       "step": 167
     },
     {
-      "epoch": 0.004613990277663344,
+      "epoch": 0.004767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 3.0741079597438246e-06,
-      "loss": 1.1837,
+      "learning_rate": 3.175803402646503e-06,
+      "loss": 1.169,
       "step": 168
     },
     {
-      "epoch": 0.004641454505506577,
+      "epoch": 0.004795686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 3.092406221408966e-06,
-      "loss": 1.2615,
+      "learning_rate": 3.1947069943289226e-06,
+      "loss": 1.0287,
       "step": 169
     },
     {
-      "epoch": 0.004668918733349812,
+      "epoch": 0.004824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 3.1107044830741086e-06,
-      "loss": 1.0778,
+      "learning_rate": 3.2136105860113427e-06,
+      "loss": 1.2131,
       "step": 170
     },
     {
-      "epoch": 0.004696382961193046,
+      "epoch": 0.004852440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 3.12900274473925e-06,
-      "loss": 1.0592,
+      "learning_rate": 3.232514177693762e-06,
+      "loss": 1.093,
       "step": 171
     },
     {
-      "epoch": 0.00472384718903628,
+      "epoch": 0.004880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 3.1473010064043917e-06,
-      "loss": 1.1002,
+      "learning_rate": 3.2514177693761816e-06,
+      "loss": 1.0597,
       "step": 172
     },
     {
-      "epoch": 0.004751311416879515,
+      "epoch": 0.004909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 3.1655992680695335e-06,
-      "loss": 1.2027,
+      "learning_rate": 3.2703213610586017e-06,
+      "loss": 1.137,
       "step": 173
     },
     {
-      "epoch": 0.004778775644722749,
+      "epoch": 0.004937570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 3.1838975297346752e-06,
-      "loss": 1.177,
+      "learning_rate": 3.289224952741021e-06,
+      "loss": 1.1445,
       "step": 174
     },
     {
-      "epoch": 0.0048062398725659825,
+      "epoch": 0.004965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 3.2021957913998174e-06,
-      "loss": 1.139,
+      "learning_rate": 3.3081285444234406e-06,
+      "loss": 1.2112,
       "step": 175
     },
     {
-      "epoch": 0.004833704100409217,
+      "epoch": 0.004994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 3.2204940530649592e-06,
-      "loss": 1.1938,
+      "learning_rate": 3.3270321361058607e-06,
+      "loss": 1.2593,
       "step": 176
     },
     {
-      "epoch": 0.004861168328252451,
+      "epoch": 0.005022701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 3.238792314730101e-06,
-      "loss": 1.1224,
+      "learning_rate": 3.34593572778828e-06,
+      "loss": 1.1005,
       "step": 177
     },
     {
-      "epoch": 0.0048886325560956855,
+      "epoch": 0.005051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 3.2570905763952428e-06,
-      "loss": 1.1071,
+      "learning_rate": 3.3648393194706996e-06,
+      "loss": 1.1233,
       "step": 178
     },
     {
-      "epoch": 0.00491609678393892,
+      "epoch": 0.0050794551645856984,
       "grad_norm": 0.0,
-      "learning_rate": 3.2753888380603845e-06,
-      "loss": 1.1241,
+      "learning_rate": 3.3837429111531196e-06,
+      "loss": 1.2104,
       "step": 179
     },
     {
-      "epoch": 0.004943561011782153,
+      "epoch": 0.00510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 3.2936870997255267e-06,
-      "loss": 1.2053,
+      "learning_rate": 3.402646502835539e-06,
+      "loss": 1.1046,
       "step": 180
     },
     {
-      "epoch": 0.004971025239625388,
+      "epoch": 0.005136208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 3.3119853613906685e-06,
-      "loss": 1.1533,
+      "learning_rate": 3.4215500945179585e-06,
+      "loss": 1.0977,
       "step": 181
     },
     {
-      "epoch": 0.004998489467468622,
+      "epoch": 0.005164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 3.33028362305581e-06,
-      "loss": 1.1522,
+      "learning_rate": 3.4404536862003786e-06,
+      "loss": 1.1314,
       "step": 182
     },
     {
-      "epoch": 0.005025953695311856,
+      "epoch": 0.005192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 3.3485818847209516e-06,
-      "loss": 1.1573,
+      "learning_rate": 3.459357277882798e-06,
+      "loss": 1.2799,
       "step": 183
     },
     {
-      "epoch": 0.005053417923155091,
+      "epoch": 0.005221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 3.3668801463860934e-06,
-      "loss": 1.0895,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 1.1058,
       "step": 184
     },
     {
-      "epoch": 0.005080882150998325,
+      "epoch": 0.005249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 3.3851784080512356e-06,
-      "loss": 1.0415,
+      "learning_rate": 3.4971644612476376e-06,
+      "loss": 1.1795,
       "step": 185
     },
     {
-      "epoch": 0.0051083463788415585,
+      "epoch": 0.0052780930760499434,
       "grad_norm": 0.0,
-      "learning_rate": 3.4034766697163773e-06,
-      "loss": 1.1145,
+      "learning_rate": 3.5160680529300572e-06,
+      "loss": 1.0912,
       "step": 186
     },
     {
-      "epoch": 0.005135810606684793,
+      "epoch": 0.005306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 3.421774931381519e-06,
-      "loss": 1.1654,
+      "learning_rate": 3.5349716446124765e-06,
+      "loss": 1.1478,
       "step": 187
     },
     {
-      "epoch": 0.005163274834528027,
+      "epoch": 0.005334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 3.440073193046661e-06,
-      "loss": 1.2862,
+      "learning_rate": 3.5538752362948966e-06,
+      "loss": 1.0885,
       "step": 188
     },
     {
-      "epoch": 0.005190739062371262,
+      "epoch": 0.00536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 3.4583714547118027e-06,
-      "loss": 1.1634,
+      "learning_rate": 3.5727788279773162e-06,
+      "loss": 1.027,
       "step": 189
     },
     {
-      "epoch": 0.005218203290214496,
+      "epoch": 0.005391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 3.476669716376945e-06,
-      "loss": 1.1613,
+      "learning_rate": 3.5916824196597355e-06,
+      "loss": 1.0976,
       "step": 190
     },
     {
-      "epoch": 0.005245667518057729,
+      "epoch": 0.005419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 3.4949679780420866e-06,
-      "loss": 1.1861,
+      "learning_rate": 3.6105860113421555e-06,
+      "loss": 1.2166,
       "step": 191
     },
     {
-      "epoch": 0.005273131745900964,
+      "epoch": 0.005448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 3.513266239707228e-06,
-      "loss": 1.2162,
+      "learning_rate": 3.629489603024575e-06,
+      "loss": 1.1633,
       "step": 192
     },
     {
-      "epoch": 0.005300595973744198,
+      "epoch": 0.0054767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 3.5315645013723697e-06,
-      "loss": 1.1166,
+      "learning_rate": 3.6483931947069944e-06,
+      "loss": 1.1325,
       "step": 193
     },
     {
-      "epoch": 0.005328060201587432,
+      "epoch": 0.005505107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 3.5498627630375115e-06,
-      "loss": 1.0579,
+      "learning_rate": 3.6672967863894145e-06,
+      "loss": 1.1855,
       "step": 194
     },
     {
-      "epoch": 0.005355524429430667,
+      "epoch": 0.005533484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 3.5681610247026537e-06,
-      "loss": 1.1235,
+      "learning_rate": 3.686200378071834e-06,
+      "loss": 1.0189,
       "step": 195
     },
     {
-      "epoch": 0.005382988657273901,
+      "epoch": 0.005561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 3.5864592863677955e-06,
-      "loss": 1.1003,
+      "learning_rate": 3.7051039697542534e-06,
+      "loss": 1.1936,
       "step": 196
     },
     {
-      "epoch": 0.005410452885117135,
+      "epoch": 0.005590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 3.6047575480329372e-06,
-      "loss": 1.1634,
+      "learning_rate": 3.7240075614366735e-06,
+      "loss": 1.0137,
       "step": 197
     },
     {
-      "epoch": 0.005437917112960369,
+      "epoch": 0.005618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 3.623055809698079e-06,
-      "loss": 1.2423,
+      "learning_rate": 3.742911153119093e-06,
+      "loss": 1.1463,
       "step": 198
     },
     {
-      "epoch": 0.005465381340803603,
+      "epoch": 0.005646992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 3.6413540713632208e-06,
-      "loss": 1.1833,
+      "learning_rate": 3.7618147448015124e-06,
+      "loss": 1.1313,
       "step": 199
     },
     {
-      "epoch": 0.005492845568646838,
+      "epoch": 0.0056753688989784334,
       "grad_norm": 0.0,
-      "learning_rate": 3.659652333028363e-06,
-      "loss": 1.1599,
+      "learning_rate": 3.7807183364839325e-06,
+      "loss": 1.2205,
       "step": 200
     },
     {
-      "epoch": 0.005520309796490072,
+      "epoch": 0.005703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 3.6779505946935047e-06,
-      "loss": 1.1855,
+      "learning_rate": 3.799621928166352e-06,
+      "loss": 1.1478,
       "step": 201
     },
     {
-      "epoch": 0.0055477740243333054,
+      "epoch": 0.005732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 3.696248856358646e-06,
-      "loss": 1.1403,
+      "learning_rate": 3.818525519848772e-06,
+      "loss": 1.0767,
       "step": 202
     },
     {
-      "epoch": 0.00557523825217654,
+      "epoch": 0.00576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 3.714547118023788e-06,
-      "loss": 1.1366,
+      "learning_rate": 3.8374291115311914e-06,
+      "loss": 0.9917,
       "step": 203
     },
     {
-      "epoch": 0.005602702480019774,
+      "epoch": 0.005788876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 3.7328453796889296e-06,
-      "loss": 1.1817,
+      "learning_rate": 3.856332703213611e-06,
+      "loss": 1.151,
       "step": 204
     },
     {
-      "epoch": 0.0056301667078630085,
+      "epoch": 0.005817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 3.751143641354072e-06,
-      "loss": 1.2004,
+      "learning_rate": 3.875236294896031e-06,
+      "loss": 1.1743,
       "step": 205
     },
     {
-      "epoch": 0.005657630935706243,
+      "epoch": 0.005845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 3.7694419030192136e-06,
-      "loss": 1.12,
+      "learning_rate": 3.89413988657845e-06,
+      "loss": 1.1749,
       "step": 206
     },
     {
-      "epoch": 0.005685095163549477,
+      "epoch": 0.0058740068104426784,
       "grad_norm": 0.0,
-      "learning_rate": 3.7877401646843554e-06,
-      "loss": 1.0684,
+      "learning_rate": 3.91304347826087e-06,
+      "loss": 1.1226,
       "step": 207
     },
     {
-      "epoch": 0.005712559391392711,
+      "epoch": 0.005902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 3.806038426349497e-06,
-      "loss": 1.2816,
+      "learning_rate": 3.93194706994329e-06,
+      "loss": 1.0245,
       "step": 208
     },
     {
-      "epoch": 0.005740023619235945,
+      "epoch": 0.0059307604994324635,
       "grad_norm": 0.0,
-      "learning_rate": 3.8243366880146385e-06,
-      "loss": 1.1233,
+      "learning_rate": 3.950850661625709e-06,
+      "loss": 1.0116,
       "step": 209
     },
     {
-      "epoch": 0.005767487847079179,
+      "epoch": 0.005959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 3.842634949679781e-06,
-      "loss": 1.186,
+      "learning_rate": 3.969754253308129e-06,
+      "loss": 1.0735,
       "step": 210
     },
     {
-      "epoch": 0.005794952074922414,
+      "epoch": 0.005987514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 3.860933211344923e-06,
-      "loss": 1.1672,
+      "learning_rate": 3.988657844990549e-06,
+      "loss": 1.1202,
       "step": 211
     },
     {
-      "epoch": 0.005822416302765648,
+      "epoch": 0.006015891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 3.879231473010065e-06,
-      "loss": 1.2286,
+      "learning_rate": 4.007561436672968e-06,
+      "loss": 1.0577,
       "step": 212
     },
     {
-      "epoch": 0.005849880530608882,
+      "epoch": 0.006044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 3.897529734675206e-06,
-      "loss": 1.1808,
+      "learning_rate": 4.026465028355388e-06,
+      "loss": 1.0854,
       "step": 213
     },
     {
-      "epoch": 0.005877344758452116,
+      "epoch": 0.006072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 3.915827996340348e-06,
-      "loss": 1.1252,
+      "learning_rate": 4.045368620037808e-06,
+      "loss": 1.087,
       "step": 214
     },
     {
-      "epoch": 0.00590480898629535,
+      "epoch": 0.006101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 3.934126258005489e-06,
-      "loss": 1.1938,
+      "learning_rate": 4.064272211720227e-06,
+      "loss": 1.0675,
       "step": 215
     },
     {
-      "epoch": 0.0059322732141385845,
+      "epoch": 0.0061293984108967085,
       "grad_norm": 0.0,
-      "learning_rate": 3.952424519670632e-06,
-      "loss": 1.1943,
+      "learning_rate": 4.083175803402647e-06,
+      "loss": 1.0993,
       "step": 216
     },
     {
-      "epoch": 0.005959737441981819,
+      "epoch": 0.0061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 3.9707227813357735e-06,
-      "loss": 1.1385,
+      "learning_rate": 4.102079395085067e-06,
+      "loss": 1.1538,
       "step": 217
     },
     {
-      "epoch": 0.005987201669825053,
+      "epoch": 0.006186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 3.989021043000915e-06,
-      "loss": 1.1805,
+      "learning_rate": 4.120982986767486e-06,
+      "loss": 1.0317,
       "step": 218
     },
     {
-      "epoch": 0.006014665897668287,
+      "epoch": 0.006214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 4.007319304666057e-06,
-      "loss": 1.1771,
+      "learning_rate": 4.139886578449906e-06,
+      "loss": 1.1135,
       "step": 219
     },
     {
-      "epoch": 0.006042130125511521,
+      "epoch": 0.006242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 4.025617566331199e-06,
-      "loss": 1.0819,
+      "learning_rate": 4.158790170132326e-06,
+      "loss": 1.108,
       "step": 220
     },
     {
-      "epoch": 0.006069594353354755,
+      "epoch": 0.006271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 4.0439158279963406e-06,
-      "loss": 1.2445,
+      "learning_rate": 4.177693761814745e-06,
+      "loss": 1.1342,
       "step": 221
     },
     {
-      "epoch": 0.00609705858119799,
+      "epoch": 0.006299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 4.062214089661482e-06,
-      "loss": 1.2114,
+      "learning_rate": 4.196597353497165e-06,
+      "loss": 1.1353,
       "step": 222
     },
     {
-      "epoch": 0.006124522809041224,
+      "epoch": 0.0063280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 4.080512351326624e-06,
-      "loss": 1.0655,
+      "learning_rate": 4.215500945179585e-06,
+      "loss": 1.0895,
       "step": 223
     },
     {
-      "epoch": 0.006151987036884458,
+      "epoch": 0.006356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 4.098810612991766e-06,
-      "loss": 1.0993,
+      "learning_rate": 4.234404536862004e-06,
+      "loss": 1.1093,
       "step": 224
     },
     {
-      "epoch": 0.006179451264727692,
+      "epoch": 0.006384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 4.117108874656908e-06,
-      "loss": 1.1742,
+      "learning_rate": 4.253308128544424e-06,
+      "loss": 1.0825,
       "step": 225
     },
     {
-      "epoch": 0.006206915492570926,
+      "epoch": 0.00641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.13540713632205e-06,
-      "loss": 1.0986,
+      "learning_rate": 4.2722117202268436e-06,
+      "loss": 1.1407,
       "step": 226
     },
     {
-      "epoch": 0.006234379720414161,
+      "epoch": 0.006441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 4.153705397987192e-06,
-      "loss": 1.1749,
+      "learning_rate": 4.291115311909263e-06,
+      "loss": 1.1452,
       "step": 227
     },
     {
-      "epoch": 0.006261843948257395,
+      "epoch": 0.006469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 4.172003659652333e-06,
-      "loss": 1.1092,
+      "learning_rate": 4.310018903591683e-06,
+      "loss": 1.1328,
       "step": 228
     },
     {
-      "epoch": 0.006289308176100629,
+      "epoch": 0.006498297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 4.190301921317475e-06,
-      "loss": 1.1416,
+      "learning_rate": 4.3289224952741025e-06,
+      "loss": 1.0636,
       "step": 229
     },
     {
-      "epoch": 0.006316772403943863,
+      "epoch": 0.0065266742338251985,
       "grad_norm": 0.0,
-      "learning_rate": 4.2086001829826165e-06,
-      "loss": 1.1833,
+      "learning_rate": 4.347826086956522e-06,
+      "loss": 1.1859,
       "step": 230
     },
     {
-      "epoch": 0.006344236631787097,
+      "epoch": 0.006555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 4.226898444647759e-06,
-      "loss": 1.1022,
+      "learning_rate": 4.366729678638942e-06,
+      "loss": 1.122,
       "step": 231
     },
     {
-      "epoch": 0.0063717008596303314,
+      "epoch": 0.006583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.245196706312901e-06,
-      "loss": 1.1034,
+      "learning_rate": 4.3856332703213615e-06,
+      "loss": 1.0177,
       "step": 232
     },
     {
-      "epoch": 0.006399165087473566,
+      "epoch": 0.006611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 4.263494967978043e-06,
-      "loss": 1.1933,
+      "learning_rate": 4.404536862003781e-06,
+      "loss": 1.0872,
       "step": 233
     },
     {
-      "epoch": 0.0064266293153168,
+      "epoch": 0.006640181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 4.2817932296431844e-06,
-      "loss": 1.1765,
+      "learning_rate": 4.423440453686201e-06,
+      "loss": 1.056,
       "step": 234
     },
     {
-      "epoch": 0.0064540935431600345,
+      "epoch": 0.006668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 4.300091491308326e-06,
-      "loss": 1.1736,
+      "learning_rate": 4.4423440453686205e-06,
+      "loss": 1.1278,
       "step": 235
     },
     {
-      "epoch": 0.006481557771003268,
+      "epoch": 0.006696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 4.318389752973468e-06,
-      "loss": 1.1617,
+      "learning_rate": 4.46124763705104e-06,
+      "loss": 1.0988,
       "step": 236
     },
     {
-      "epoch": 0.006509021998846502,
+      "epoch": 0.0067253121452894435,
       "grad_norm": 0.0,
-      "learning_rate": 4.33668801463861e-06,
-      "loss": 1.1404,
+      "learning_rate": 4.48015122873346e-06,
+      "loss": 0.9534,
       "step": 237
     },
     {
-      "epoch": 0.006536486226689737,
+      "epoch": 0.006753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 4.3549862763037515e-06,
-      "loss": 1.1012,
+      "learning_rate": 4.4990548204158795e-06,
+      "loss": 1.065,
       "step": 238
     },
     {
-      "epoch": 0.006563950454532971,
+      "epoch": 0.0067820658342792285,
       "grad_norm": 0.0,
-      "learning_rate": 4.373284537968893e-06,
-      "loss": 1.1377,
+      "learning_rate": 4.517958412098299e-06,
+      "loss": 1.0014,
       "step": 239
     },
     {
-      "epoch": 0.006591414682376205,
+      "epoch": 0.00681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 4.391582799634035e-06,
-      "loss": 1.1825,
+      "learning_rate": 4.536862003780719e-06,
+      "loss": 1.1644,
       "step": 240
     },
     {
-      "epoch": 0.006618878910219439,
+      "epoch": 0.006838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 4.409881061299177e-06,
-      "loss": 1.1163,
+      "learning_rate": 4.5557655954631384e-06,
+      "loss": 1.1082,
       "step": 241
     },
     {
-      "epoch": 0.006646343138062673,
+      "epoch": 0.006867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 4.428179322964319e-06,
-      "loss": 1.1724,
+      "learning_rate": 4.574669187145558e-06,
+      "loss": 1.15,
       "step": 242
     },
     {
-      "epoch": 0.0066738073659059075,
+      "epoch": 0.006895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 4.44647758462946e-06,
-      "loss": 1.1242,
+      "learning_rate": 4.593572778827978e-06,
+      "loss": 1.1852,
       "step": 243
     },
     {
-      "epoch": 0.006701271593749142,
+      "epoch": 0.006923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 4.464775846294602e-06,
-      "loss": 1.1963,
+      "learning_rate": 4.612476370510397e-06,
+      "loss": 1.0718,
       "step": 244
     },
     {
-      "epoch": 0.006728735821592376,
+      "epoch": 0.006952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 4.483074107959744e-06,
-      "loss": 1.2042,
+      "learning_rate": 4.631379962192817e-06,
+      "loss": 1.0342,
       "step": 245
     },
     {
-      "epoch": 0.0067562000494356105,
+      "epoch": 0.0069807037457434735,
       "grad_norm": 0.0,
-      "learning_rate": 4.5013723696248865e-06,
-      "loss": 1.1319,
+      "learning_rate": 4.650283553875237e-06,
+      "loss": 1.1017,
       "step": 246
     },
     {
-      "epoch": 0.006783664277278844,
+      "epoch": 0.007009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 4.519670631290028e-06,
-      "loss": 1.1329,
+      "learning_rate": 4.669187145557656e-06,
+      "loss": 1.0341,
       "step": 247
     },
     {
-      "epoch": 0.006811128505122078,
+      "epoch": 0.007037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 4.537968892955169e-06,
-      "loss": 1.145,
+      "learning_rate": 4.688090737240076e-06,
+      "loss": 1.1276,
       "step": 248
     },
     {
-      "epoch": 0.006838592732965313,
+      "epoch": 0.00706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 4.556267154620311e-06,
-      "loss": 1.072,
+      "learning_rate": 4.706994328922496e-06,
+      "loss": 1.0712,
       "step": 249
     },
     {
-      "epoch": 0.006866056960808547,
+      "epoch": 0.007094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.574565416285453e-06,
-      "loss": 1.1544,
+      "learning_rate": 4.725897920604915e-06,
+      "loss": 1.1603,
       "step": 250
     },
     {
-      "epoch": 0.006893521188651781,
+      "epoch": 0.007122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 4.592863677950595e-06,
-      "loss": 1.1995,
+      "learning_rate": 4.744801512287335e-06,
+      "loss": 1.0171,
       "step": 251
     },
     {
-      "epoch": 0.006920985416495015,
+      "epoch": 0.007150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 4.611161939615737e-06,
-      "loss": 1.1395,
+      "learning_rate": 4.763705103969755e-06,
+      "loss": 1.1587,
       "step": 252
     },
     {
-      "epoch": 0.006948449644338249,
+      "epoch": 0.0071793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 4.629460201280879e-06,
-      "loss": 1.0445,
+      "learning_rate": 4.782608695652174e-06,
+      "loss": 1.0914,
       "step": 253
     },
     {
-      "epoch": 0.0069759138721814835,
+      "epoch": 0.007207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 4.647758462946021e-06,
-      "loss": 1.1965,
+      "learning_rate": 4.801512287334594e-06,
+      "loss": 1.1912,
       "step": 254
     },
     {
-      "epoch": 0.007003378100024718,
+      "epoch": 0.007236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 4.6660567246111625e-06,
-      "loss": 1.103,
+      "learning_rate": 4.820415879017014e-06,
+      "loss": 1.0142,
       "step": 255
     },
     {
-      "epoch": 0.007030842327867952,
+      "epoch": 0.007264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 4.684354986276304e-06,
-      "loss": 1.1757,
+      "learning_rate": 4.839319470699433e-06,
+      "loss": 1.0599,
       "step": 256
     },
     {
-      "epoch": 0.007058306555711187,
+      "epoch": 0.007292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 4.702653247941446e-06,
-      "loss": 1.2317,
+      "learning_rate": 4.858223062381853e-06,
+      "loss": 1.136,
       "step": 257
     },
     {
-      "epoch": 0.00708577078355442,
+      "epoch": 0.007321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 4.720951509606588e-06,
-      "loss": 1.1733,
+      "learning_rate": 4.877126654064273e-06,
+      "loss": 1.0036,
       "step": 258
     },
     {
-      "epoch": 0.007113235011397654,
+      "epoch": 0.007349602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 4.7392497712717295e-06,
-      "loss": 1.1459,
+      "learning_rate": 4.896030245746692e-06,
+      "loss": 1.0686,
       "step": 259
     },
     {
-      "epoch": 0.007140699239240889,
+      "epoch": 0.0073779795686719635,
       "grad_norm": 0.0,
-      "learning_rate": 4.757548032936871e-06,
-      "loss": 1.1281,
+      "learning_rate": 4.914933837429112e-06,
+      "loss": 1.1184,
       "step": 260
     },
     {
-      "epoch": 0.007168163467084123,
+      "epoch": 0.007406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 4.775846294602013e-06,
-      "loss": 1.0634,
+      "learning_rate": 4.933837429111532e-06,
+      "loss": 1.0451,
       "step": 261
     },
     {
-      "epoch": 0.007195627694927357,
+      "epoch": 0.007434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 4.794144556267155e-06,
-      "loss": 1.0823,
+      "learning_rate": 4.952741020793951e-06,
+      "loss": 1.1512,
       "step": 262
     },
     {
-      "epoch": 0.007223091922770591,
+      "epoch": 0.00746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 4.812442817932297e-06,
-      "loss": 1.0497,
+      "learning_rate": 4.971644612476371e-06,
+      "loss": 1.0852,
       "step": 263
     },
     {
-      "epoch": 0.007250556150613825,
+      "epoch": 0.007491486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 4.830741079597438e-06,
-      "loss": 1.1079,
+      "learning_rate": 4.990548204158791e-06,
+      "loss": 1.1717,
       "step": 264
     },
     {
-      "epoch": 0.00727802037845706,
+      "epoch": 0.007519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 4.84903934126258e-06,
-      "loss": 1.1727,
+      "learning_rate": 5.00945179584121e-06,
+      "loss": 1.066,
       "step": 265
     },
     {
-      "epoch": 0.007305484606300294,
+      "epoch": 0.007548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 4.867337602927723e-06,
-      "loss": 1.0744,
+      "learning_rate": 5.02835538752363e-06,
+      "loss": 1.0351,
       "step": 266
     },
     {
-      "epoch": 0.007332948834143528,
+      "epoch": 0.0075766174801362085,
       "grad_norm": 0.0,
-      "learning_rate": 4.8856358645928646e-06,
-      "loss": 1.0855,
+      "learning_rate": 5.04725897920605e-06,
+      "loss": 1.0977,
       "step": 267
     },
     {
-      "epoch": 0.007360413061986763,
+      "epoch": 0.007604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.903934126258006e-06,
-      "loss": 1.1646,
+      "learning_rate": 5.066162570888468e-06,
+      "loss": 1.1569,
       "step": 268
     },
     {
-      "epoch": 0.007387877289829996,
+      "epoch": 0.0076333711691259935,
       "grad_norm": 0.0,
-      "learning_rate": 4.922232387923147e-06,
-      "loss": 1.1924,
+      "learning_rate": 5.085066162570889e-06,
+      "loss": 1.0821,
       "step": 269
     },
     {
-      "epoch": 0.0074153415176732304,
+      "epoch": 0.007661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 4.940530649588289e-06,
-      "loss": 1.0686,
+      "learning_rate": 5.1039697542533085e-06,
+      "loss": 1.0617,
       "step": 270
     },
     {
-      "epoch": 0.007442805745516465,
+      "epoch": 0.007690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 4.958828911253432e-06,
-      "loss": 1.0571,
+      "learning_rate": 5.122873345935728e-06,
+      "loss": 1.1791,
       "step": 271
     },
     {
-      "epoch": 0.007470269973359699,
+      "epoch": 0.007718501702610669,
       "grad_norm": 0.0,
-      "learning_rate": 4.977127172918573e-06,
-      "loss": 1.1096,
+      "learning_rate": 5.141776937618148e-06,
+      "loss": 1.049,
       "step": 272
     },
     {
-      "epoch": 0.0074977342012029335,
+      "epoch": 0.007746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 4.995425434583715e-06,
-      "loss": 1.1412,
+      "learning_rate": 5.160680529300568e-06,
+      "loss": 1.0999,
       "step": 273
     },
     {
-      "epoch": 0.007525198429046167,
+      "epoch": 0.007775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 5.013723696248857e-06,
-      "loss": 1.1759,
+      "learning_rate": 5.179584120982986e-06,
+      "loss": 1.1115,
       "step": 274
     },
     {
-      "epoch": 0.007552662656889401,
+      "epoch": 0.007803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 5.032021957913999e-06,
-      "loss": 1.1808,
+      "learning_rate": 5.198487712665407e-06,
+      "loss": 1.1638,
       "step": 275
     },
     {
-      "epoch": 0.007580126884732636,
+      "epoch": 0.007832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 5.0503202195791405e-06,
-      "loss": 1.0981,
+      "learning_rate": 5.2173913043478265e-06,
+      "loss": 1.0799,
       "step": 276
     },
     {
-      "epoch": 0.00760759111257587,
+      "epoch": 0.007860385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 5.068618481244281e-06,
-      "loss": 1.1174,
+      "learning_rate": 5.236294896030246e-06,
+      "loss": 1.0193,
       "step": 277
     },
     {
-      "epoch": 0.007635055340419104,
+      "epoch": 0.007888762769580024,
       "grad_norm": 0.0,
-      "learning_rate": 5.086916742909423e-06,
-      "loss": 1.1232,
+      "learning_rate": 5.255198487712666e-06,
+      "loss": 1.1257,
       "step": 278
     },
     {
-      "epoch": 0.007662519568262339,
+      "epoch": 0.007917139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 5.105215004574567e-06,
-      "loss": 1.1825,
+      "learning_rate": 5.274102079395086e-06,
+      "loss": 1.1393,
       "step": 279
     },
     {
-      "epoch": 0.007689983796105572,
+      "epoch": 0.007945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.123513266239708e-06,
-      "loss": 1.1703,
+      "learning_rate": 5.293005671077504e-06,
+      "loss": 1.0826,
       "step": 280
     },
     {
-      "epoch": 0.0077174480239488065,
+      "epoch": 0.0079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.141811527904849e-06,
-      "loss": 1.1295,
+      "learning_rate": 5.311909262759925e-06,
+      "loss": 1.0242,
       "step": 281
     },
     {
-      "epoch": 0.007744912251792041,
+      "epoch": 0.008002270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 5.160109789569991e-06,
-      "loss": 1.2668,
+      "learning_rate": 5.3308128544423444e-06,
+      "loss": 1.0492,
       "step": 282
     },
     {
-      "epoch": 0.007772376479635275,
+      "epoch": 0.008030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 5.178408051235133e-06,
-      "loss": 1.1536,
+      "learning_rate": 5.349716446124764e-06,
+      "loss": 1.0583,
       "step": 283
     },
     {
-      "epoch": 0.0077998407074785095,
+      "epoch": 0.008059023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 5.196706312900275e-06,
-      "loss": 1.169,
+      "learning_rate": 5.368620037807184e-06,
+      "loss": 1.0333,
       "step": 284
     },
     {
-      "epoch": 0.007827304935321744,
+      "epoch": 0.008087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 5.215004574565416e-06,
-      "loss": 1.2267,
+      "learning_rate": 5.387523629489604e-06,
+      "loss": 1.1055,
       "step": 285
     },
     {
-      "epoch": 0.007854769163164977,
+      "epoch": 0.00811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.233302836230558e-06,
-      "loss": 1.1351,
+      "learning_rate": 5.406427221172024e-06,
+      "loss": 1.1657,
       "step": 286
     },
     {
-      "epoch": 0.007882233391008213,
+      "epoch": 0.008144154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 5.2516010978957e-06,
-      "loss": 1.1339,
+      "learning_rate": 5.425330812854443e-06,
+      "loss": 1.0406,
       "step": 287
     },
     {
-      "epoch": 0.007909697618851446,
+      "epoch": 0.008172531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 5.269899359560842e-06,
-      "loss": 1.0414,
+      "learning_rate": 5.444234404536862e-06,
+      "loss": 1.2001,
       "step": 288
     },
     {
-      "epoch": 0.00793716184669468,
+      "epoch": 0.008200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 5.288197621225984e-06,
-      "loss": 1.0904,
+      "learning_rate": 5.463137996219282e-06,
+      "loss": 1.1033,
       "step": 289
     },
     {
-      "epoch": 0.007964626074537915,
+      "epoch": 0.008229284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 5.306495882891126e-06,
-      "loss": 1.1382,
+      "learning_rate": 5.482041587901702e-06,
+      "loss": 1.126,
       "step": 290
     },
     {
-      "epoch": 0.007992090302381148,
+      "epoch": 0.008257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 5.324794144556268e-06,
-      "loss": 1.2112,
+      "learning_rate": 5.500945179584122e-06,
+      "loss": 1.1374,
       "step": 291
     },
     {
-      "epoch": 0.008019554530224383,
+      "epoch": 0.008286038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 5.34309240622141e-06,
-      "loss": 1.0925,
+      "learning_rate": 5.519848771266542e-06,
+      "loss": 1.0359,
       "step": 292
     },
     {
-      "epoch": 0.008047018758067617,
+      "epoch": 0.008314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 5.3613906678865514e-06,
-      "loss": 1.1296,
+      "learning_rate": 5.538752362948961e-06,
+      "loss": 1.1772,
       "step": 293
     },
     {
-      "epoch": 0.00807448298591085,
+      "epoch": 0.008342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 5.379688929551693e-06,
-      "loss": 1.0937,
+      "learning_rate": 5.55765595463138e-06,
+      "loss": 1.09,
       "step": 294
     },
     {
-      "epoch": 0.008101947213754086,
+      "epoch": 0.00837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 5.397987191216835e-06,
-      "loss": 1.0829,
+      "learning_rate": 5.5765595463138e-06,
+      "loss": 1.0872,
       "step": 295
     },
     {
-      "epoch": 0.008129411441597319,
+      "epoch": 0.008399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 5.416285452881977e-06,
-      "loss": 1.1942,
+      "learning_rate": 5.59546313799622e-06,
+      "loss": 1.107,
       "step": 296
     },
     {
-      "epoch": 0.008156875669440554,
+      "epoch": 0.008427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 5.434583714547118e-06,
-      "loss": 1.1516,
+      "learning_rate": 5.614366729678639e-06,
+      "loss": 1.0838,
       "step": 297
     },
     {
-      "epoch": 0.008184339897283788,
+      "epoch": 0.008456299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 5.4528819762122594e-06,
-      "loss": 1.1746,
+      "learning_rate": 5.63327032136106e-06,
+      "loss": 1.0784,
       "step": 298
     },
     {
-      "epoch": 0.008211804125127021,
+      "epoch": 0.008484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 5.471180237877403e-06,
-      "loss": 1.0529,
+      "learning_rate": 5.652173913043479e-06,
+      "loss": 1.0838,
       "step": 299
     },
     {
-      "epoch": 0.008239268352970256,
+      "epoch": 0.00851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 5.489478499542545e-06,
-      "loss": 1.1154,
+      "learning_rate": 5.671077504725898e-06,
+      "loss": 1.0481,
       "step": 300
     },
     {
-      "epoch": 0.00826673258081349,
+      "epoch": 0.008541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 5.507776761207686e-06,
-      "loss": 1.0875,
+      "learning_rate": 5.689981096408318e-06,
+      "loss": 1.0033,
       "step": 301
     },
     {
-      "epoch": 0.008294196808656725,
+      "epoch": 0.008569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 5.526075022872827e-06,
-      "loss": 1.1081,
+      "learning_rate": 5.708884688090738e-06,
+      "loss": 0.9697,
       "step": 302
     },
     {
-      "epoch": 0.008321661036499959,
+      "epoch": 0.008598183881952328,
       "grad_norm": 0.0,
-      "learning_rate": 5.544373284537969e-06,
-      "loss": 1.1856,
+      "learning_rate": 5.727788279773157e-06,
+      "loss": 1.1099,
       "step": 303
     },
     {
-      "epoch": 0.008349125264343194,
+      "epoch": 0.008626560726447218,
       "grad_norm": 0.0,
-      "learning_rate": 5.562671546203111e-06,
-      "loss": 1.1754,
+      "learning_rate": 5.746691871455578e-06,
+      "loss": 1.0736,
       "step": 304
     },
     {
-      "epoch": 0.008376589492186427,
+      "epoch": 0.008654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 5.580969807868253e-06,
-      "loss": 1.1003,
+      "learning_rate": 5.7655954631379966e-06,
+      "loss": 1.1558,
       "step": 305
     },
     {
-      "epoch": 0.00840405372002966,
+      "epoch": 0.008683314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 5.5992680695333944e-06,
-      "loss": 1.0606,
+      "learning_rate": 5.784499054820416e-06,
+      "loss": 1.1485,
       "step": 306
     },
     {
-      "epoch": 0.008431517947872896,
+      "epoch": 0.008711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 5.617566331198536e-06,
-      "loss": 1.0421,
+      "learning_rate": 5.803402646502836e-06,
+      "loss": 1.0186,
       "step": 307
     },
     {
-      "epoch": 0.00845898217571613,
+      "epoch": 0.008740068104426789,
       "grad_norm": 0.0,
-      "learning_rate": 5.635864592863678e-06,
-      "loss": 1.1219,
+      "learning_rate": 5.8223062381852555e-06,
+      "loss": 1.083,
       "step": 308
     },
     {
-      "epoch": 0.008486446403559365,
+      "epoch": 0.00876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 5.654162854528821e-06,
-      "loss": 1.1296,
+      "learning_rate": 5.841209829867675e-06,
+      "loss": 1.0646,
       "step": 309
     },
     {
-      "epoch": 0.008513910631402598,
+      "epoch": 0.008796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 5.672461116193962e-06,
-      "loss": 1.0713,
+      "learning_rate": 5.860113421550096e-06,
+      "loss": 1.1739,
       "step": 310
     },
     {
-      "epoch": 0.008541374859245832,
+      "epoch": 0.008825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 5.690759377859104e-06,
-      "loss": 1.0714,
+      "learning_rate": 5.8790170132325145e-06,
+      "loss": 1.0208,
       "step": 311
     },
     {
-      "epoch": 0.008568839087089067,
+      "epoch": 0.008853575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 5.709057639524246e-06,
-      "loss": 1.0625,
+      "learning_rate": 5.897920604914934e-06,
+      "loss": 1.1558,
       "step": 312
     },
     {
-      "epoch": 0.0085963033149323,
+      "epoch": 0.00888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 5.727355901189388e-06,
-      "loss": 1.1181,
+      "learning_rate": 5.916824196597354e-06,
+      "loss": 1.0154,
       "step": 313
     },
     {
-      "epoch": 0.008623767542775536,
+      "epoch": 0.00891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 5.7456541628545295e-06,
-      "loss": 1.2025,
+      "learning_rate": 5.9357277882797735e-06,
+      "loss": 1.1487,
       "step": 314
     },
     {
-      "epoch": 0.008651231770618769,
+      "epoch": 0.008938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 5.763952424519671e-06,
-      "loss": 1.1424,
+      "learning_rate": 5.954631379962193e-06,
+      "loss": 1.1968,
       "step": 315
     },
     {
-      "epoch": 0.008678695998462002,
+      "epoch": 0.008967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 5.782250686184813e-06,
-      "loss": 1.1202,
+      "learning_rate": 5.973534971644614e-06,
+      "loss": 1.0932,
       "step": 316
     },
     {
-      "epoch": 0.008706160226305238,
+      "epoch": 0.008995459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 5.800548947849955e-06,
-      "loss": 1.1774,
+      "learning_rate": 5.9924385633270325e-06,
+      "loss": 1.0035,
       "step": 317
     },
     {
-      "epoch": 0.008733624454148471,
+      "epoch": 0.00902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 5.818847209515096e-06,
-      "loss": 1.1004,
+      "learning_rate": 6.011342155009452e-06,
+      "loss": 1.1041,
       "step": 318
     },
     {
-      "epoch": 0.008761088681991706,
+      "epoch": 0.009052213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 5.837145471180239e-06,
-      "loss": 1.0948,
+      "learning_rate": 6.030245746691872e-06,
+      "loss": 1.0483,
       "step": 319
     },
     {
-      "epoch": 0.00878855290983494,
+      "epoch": 0.009080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 5.855443732845381e-06,
-      "loss": 1.1798,
+      "learning_rate": 6.0491493383742914e-06,
+      "loss": 1.109,
       "step": 320
     },
     {
-      "epoch": 0.008816017137678173,
+      "epoch": 0.009108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 5.873741994510523e-06,
-      "loss": 1.1943,
+      "learning_rate": 6.068052930056711e-06,
+      "loss": 1.1079,
       "step": 321
     },
     {
-      "epoch": 0.008843481365521409,
+      "epoch": 0.009137343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 5.892040256175664e-06,
-      "loss": 1.1416,
+      "learning_rate": 6.086956521739132e-06,
+      "loss": 1.0806,
       "step": 322
     },
     {
-      "epoch": 0.008870945593364642,
+      "epoch": 0.009165720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 5.910338517840805e-06,
-      "loss": 1.1185,
+      "learning_rate": 6.10586011342155e-06,
+      "loss": 1.0574,
       "step": 323
     },
     {
-      "epoch": 0.008898409821207877,
+      "epoch": 0.009194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 5.928636779505947e-06,
-      "loss": 1.1128,
+      "learning_rate": 6.12476370510397e-06,
+      "loss": 1.0582,
       "step": 324
     },
     {
-      "epoch": 0.00892587404905111,
+      "epoch": 0.009222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 5.946935041171089e-06,
-      "loss": 1.088,
+      "learning_rate": 6.14366729678639e-06,
+      "loss": 1.0743,
       "step": 325
     },
     {
-      "epoch": 0.008953338276894346,
+      "epoch": 0.009250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 5.965233302836231e-06,
-      "loss": 1.2142,
+      "learning_rate": 6.162570888468809e-06,
+      "loss": 1.0721,
       "step": 326
     },
     {
-      "epoch": 0.00898080250473758,
+      "epoch": 0.00927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 5.9835315645013725e-06,
-      "loss": 1.1305,
+      "learning_rate": 6.181474480151229e-06,
+      "loss": 1.0522,
       "step": 327
     },
     {
-      "epoch": 0.009008266732580813,
+      "epoch": 0.009307604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 6.001829826166514e-06,
-      "loss": 1.0944,
+      "learning_rate": 6.2003780718336495e-06,
+      "loss": 1.1191,
       "step": 328
     },
     {
-      "epoch": 0.009035730960424048,
+      "epoch": 0.009335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 6.020128087831657e-06,
-      "loss": 1.1139,
+      "learning_rate": 6.219281663516069e-06,
+      "loss": 1.1987,
       "step": 329
     },
     {
-      "epoch": 0.009063195188267282,
+      "epoch": 0.009364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 6.038426349496799e-06,
-      "loss": 1.0875,
+      "learning_rate": 6.238185255198488e-06,
+      "loss": 1.1167,
       "step": 330
     },
     {
-      "epoch": 0.009090659416110517,
+      "epoch": 0.009392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 6.05672461116194e-06,
-      "loss": 1.1429,
+      "learning_rate": 6.257088846880908e-06,
+      "loss": 1.0319,
       "step": 331
     },
     {
-      "epoch": 0.00911812364395375,
+      "epoch": 0.0094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.075022872827082e-06,
-      "loss": 1.0479,
+      "learning_rate": 6.275992438563327e-06,
+      "loss": 1.0634,
       "step": 332
     },
     {
-      "epoch": 0.009145587871796984,
+      "epoch": 0.009449489216799093,
       "grad_norm": 0.0,
-      "learning_rate": 6.093321134492224e-06,
-      "loss": 1.02,
+      "learning_rate": 6.294896030245747e-06,
+      "loss": 1.1193,
       "step": 333
     },
     {
-      "epoch": 0.009173052099640219,
+      "epoch": 0.009477866061293983,
       "grad_norm": 0.0,
-      "learning_rate": 6.111619396157366e-06,
-      "loss": 1.1081,
+      "learning_rate": 6.3137996219281675e-06,
+      "loss": 1.0862,
       "step": 334
     },
     {
-      "epoch": 0.009200516327483452,
+      "epoch": 0.009506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 6.1299176578225075e-06,
-      "loss": 1.0781,
+      "learning_rate": 6.332703213610587e-06,
+      "loss": 1.1658,
       "step": 335
     },
     {
-      "epoch": 0.009227980555326688,
+      "epoch": 0.009534619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 6.148215919487649e-06,
-      "loss": 1.0494,
+      "learning_rate": 6.351606805293006e-06,
+      "loss": 1.189,
       "step": 336
     },
     {
-      "epoch": 0.009255444783169921,
+      "epoch": 0.009562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 6.166514181152791e-06,
-      "loss": 1.159,
+      "learning_rate": 6.370510396975426e-06,
+      "loss": 1.0882,
       "step": 337
     },
     {
-      "epoch": 0.009282909011013155,
+      "epoch": 0.009591373439273554,
       "grad_norm": 0.0,
-      "learning_rate": 6.184812442817932e-06,
-      "loss": 1.1321,
+      "learning_rate": 6.389413988657845e-06,
+      "loss": 1.1045,
       "step": 338
     },
     {
-      "epoch": 0.00931037323885639,
+      "epoch": 0.009619750283768444,
       "grad_norm": 0.0,
-      "learning_rate": 6.203110704483074e-06,
-      "loss": 1.1869,
+      "learning_rate": 6.408317580340265e-06,
+      "loss": 1.1162,
       "step": 339
     },
     {
-      "epoch": 0.009337837466699623,
+      "epoch": 0.009648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 6.221408966148217e-06,
-      "loss": 1.1788,
+      "learning_rate": 6.4272211720226854e-06,
+      "loss": 1.0182,
       "step": 340
     },
     {
-      "epoch": 0.009365301694542858,
+      "epoch": 0.00967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.239707227813359e-06,
-      "loss": 1.2104,
+      "learning_rate": 6.446124763705105e-06,
+      "loss": 1.0938,
       "step": 341
     },
     {
-      "epoch": 0.009392765922386092,
+      "epoch": 0.009704880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 6.2580054894785e-06,
-      "loss": 1.1486,
+      "learning_rate": 6.465028355387524e-06,
+      "loss": 1.1387,
       "step": 342
     },
     {
-      "epoch": 0.009420230150229327,
+      "epoch": 0.009733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 6.276303751143642e-06,
-      "loss": 1.1144,
+      "learning_rate": 6.4839319470699436e-06,
+      "loss": 1.0679,
       "step": 343
     },
     {
-      "epoch": 0.00944769437807256,
+      "epoch": 0.009761634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 6.294602012808783e-06,
-      "loss": 1.1456,
+      "learning_rate": 6.502835538752363e-06,
+      "loss": 1.0858,
       "step": 344
     },
     {
-      "epoch": 0.009475158605915794,
+      "epoch": 0.009790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 6.312900274473925e-06,
-      "loss": 1.0535,
+      "learning_rate": 6.521739130434783e-06,
+      "loss": 1.0499,
       "step": 345
     },
     {
-      "epoch": 0.00950262283375903,
+      "epoch": 0.00981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 6.331198536139067e-06,
-      "loss": 1.0894,
+      "learning_rate": 6.540642722117203e-06,
+      "loss": 0.9648,
       "step": 346
     },
     {
-      "epoch": 0.009530087061602263,
+      "epoch": 0.009846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 6.349496797804209e-06,
-      "loss": 1.0774,
+      "learning_rate": 6.559546313799623e-06,
+      "loss": 1.1292,
       "step": 347
     },
     {
-      "epoch": 0.009557551289445498,
+      "epoch": 0.009875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 6.3677950594693505e-06,
-      "loss": 1.1392,
+      "learning_rate": 6.578449905482042e-06,
+      "loss": 1.0131,
       "step": 348
     },
     {
-      "epoch": 0.009585015517288731,
+      "epoch": 0.009903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 6.386093321134492e-06,
-      "loss": 1.0812,
+      "learning_rate": 6.5973534971644615e-06,
+      "loss": 1.1511,
       "step": 349
     },
     {
-      "epoch": 0.009612479745131965,
+      "epoch": 0.009931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 6.404391582799635e-06,
-      "loss": 1.0724,
+      "learning_rate": 6.616257088846881e-06,
+      "loss": 1.1293,
       "step": 350
     },
     {
-      "epoch": 0.0096399439729752,
+      "epoch": 0.009960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 6.422689844464777e-06,
-      "loss": 1.1321,
+      "learning_rate": 6.635160680529301e-06,
+      "loss": 1.1228,
       "step": 351
     },
     {
-      "epoch": 0.009667408200818434,
+      "epoch": 0.009988649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 6.4409881061299184e-06,
-      "loss": 1.192,
+      "learning_rate": 6.654064272211721e-06,
+      "loss": 1.0753,
       "step": 352
     },
     {
-      "epoch": 0.009694872428661669,
+      "epoch": 0.010017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 6.45928636779506e-06,
-      "loss": 1.156,
+      "learning_rate": 6.672967863894141e-06,
+      "loss": 1.2051,
       "step": 353
     },
     {
-      "epoch": 0.009722336656504902,
+      "epoch": 0.010045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 6.477584629460202e-06,
-      "loss": 1.1307,
+      "learning_rate": 6.69187145557656e-06,
+      "loss": 0.9972,
       "step": 354
     },
     {
-      "epoch": 0.009749800884348136,
+      "epoch": 0.01007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 6.495882891125344e-06,
-      "loss": 1.1056,
+      "learning_rate": 6.7107750472589795e-06,
+      "loss": 1.0515,
       "step": 355
     },
     {
-      "epoch": 0.009777265112191371,
+      "epoch": 0.010102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 6.5141811527904855e-06,
-      "loss": 1.1079,
+      "learning_rate": 6.729678638941399e-06,
+      "loss": 1.1081,
       "step": 356
     },
     {
-      "epoch": 0.009804729340034605,
+      "epoch": 0.010130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 6.532479414455627e-06,
-      "loss": 1.1653,
+      "learning_rate": 6.748582230623819e-06,
+      "loss": 1.1013,
       "step": 357
     },
     {
-      "epoch": 0.00983219356787784,
+      "epoch": 0.010158910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 6.550777676120769e-06,
-      "loss": 1.1686,
+      "learning_rate": 6.767485822306239e-06,
+      "loss": 0.9883,
       "step": 358
     },
     {
-      "epoch": 0.009859657795721073,
+      "epoch": 0.010187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 6.56907593778591e-06,
-      "loss": 1.3115,
+      "learning_rate": 6.786389413988659e-06,
+      "loss": 1.0848,
       "step": 359
     },
     {
-      "epoch": 0.009887122023564307,
+      "epoch": 0.01021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 6.5873741994510534e-06,
-      "loss": 1.0395,
+      "learning_rate": 6.805293005671078e-06,
+      "loss": 1.1311,
       "step": 360
     },
     {
-      "epoch": 0.009914586251407542,
+      "epoch": 0.010244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 6.605672461116195e-06,
-      "loss": 1.261,
+      "learning_rate": 6.824196597353497e-06,
+      "loss": 1.182,
       "step": 361
     },
     {
-      "epoch": 0.009942050479250775,
+      "epoch": 0.010272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 6.623970722781337e-06,
-      "loss": 1.0651,
+      "learning_rate": 6.843100189035917e-06,
+      "loss": 1.115,
       "step": 362
     },
     {
-      "epoch": 0.00996951470709401,
+      "epoch": 0.010300794551645858,
       "grad_norm": 0.0,
-      "learning_rate": 6.642268984446478e-06,
-      "loss": 1.0784,
+      "learning_rate": 6.862003780718337e-06,
+      "loss": 1.1503,
       "step": 363
     },
     {
-      "epoch": 0.009996978934937244,
+      "epoch": 0.010329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 6.66056724611162e-06,
-      "loss": 1.1669,
+      "learning_rate": 6.880907372400757e-06,
+      "loss": 1.1635,
       "step": 364
     },
     {
-      "epoch": 0.01002444316278048,
+      "epoch": 0.010357548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 6.6788655077767614e-06,
-      "loss": 1.1372,
+      "learning_rate": 6.899810964083177e-06,
+      "loss": 1.1221,
       "step": 365
     },
     {
-      "epoch": 0.010051907390623713,
+      "epoch": 0.010385925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 6.697163769441903e-06,
-      "loss": 1.1578,
+      "learning_rate": 6.918714555765596e-06,
+      "loss": 1.083,
       "step": 366
     },
     {
-      "epoch": 0.010079371618466946,
+      "epoch": 0.010414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 6.715462031107045e-06,
-      "loss": 1.1105,
+      "learning_rate": 6.937618147448015e-06,
+      "loss": 0.9808,
       "step": 367
     },
     {
-      "epoch": 0.010106835846310181,
+      "epoch": 0.010442678774120319,
       "grad_norm": 0.0,
-      "learning_rate": 6.733760292772187e-06,
-      "loss": 1.0366,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 1.0874,
       "step": 368
     },
     {
-      "epoch": 0.010134300074153415,
+      "epoch": 0.01047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.7520585544373285e-06,
-      "loss": 1.2154,
+      "learning_rate": 6.975425330812855e-06,
+      "loss": 1.0268,
       "step": 369
     },
     {
-      "epoch": 0.01016176430199665,
+      "epoch": 0.010499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 6.770356816102471e-06,
-      "loss": 1.1219,
+      "learning_rate": 6.994328922495275e-06,
+      "loss": 1.0733,
       "step": 370
     },
     {
-      "epoch": 0.010189228529839884,
+      "epoch": 0.010527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 6.788655077767613e-06,
-      "loss": 1.1031,
+      "learning_rate": 7.013232514177695e-06,
+      "loss": 1.1044,
       "step": 371
     },
     {
-      "epoch": 0.010216692757683117,
+      "epoch": 0.010556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 6.806953339432755e-06,
-      "loss": 1.0647,
+      "learning_rate": 7.0321361058601145e-06,
+      "loss": 1.1287,
       "step": 372
     },
     {
-      "epoch": 0.010244156985526352,
+      "epoch": 0.01058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 6.8252516010978965e-06,
-      "loss": 1.1259,
+      "learning_rate": 7.051039697542533e-06,
+      "loss": 1.0375,
       "step": 373
     },
     {
-      "epoch": 0.010271621213369586,
+      "epoch": 0.01061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 6.843549862763038e-06,
-      "loss": 1.0842,
+      "learning_rate": 7.069943289224953e-06,
+      "loss": 1.1217,
       "step": 374
     },
     {
-      "epoch": 0.010299085441212821,
+      "epoch": 0.010641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.86184812442818e-06,
-      "loss": 1.1506,
+      "learning_rate": 7.088846880907373e-06,
+      "loss": 1.1165,
       "step": 375
     },
     {
-      "epoch": 0.010326549669056054,
+      "epoch": 0.010669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 6.880146386093322e-06,
-      "loss": 1.0303,
+      "learning_rate": 7.107750472589793e-06,
+      "loss": 1.0785,
       "step": 376
     },
     {
-      "epoch": 0.010354013896899288,
+      "epoch": 0.010698070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 6.8984446477584635e-06,
-      "loss": 1.1472,
+      "learning_rate": 7.126654064272213e-06,
+      "loss": 0.9635,
       "step": 377
     },
     {
-      "epoch": 0.010381478124742523,
+      "epoch": 0.01072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 6.916742909423605e-06,
-      "loss": 1.0455,
+      "learning_rate": 7.1455576559546324e-06,
+      "loss": 1.1927,
       "step": 378
     },
     {
-      "epoch": 0.010408942352585757,
+      "epoch": 0.010754824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 6.935041171088746e-06,
-      "loss": 1.0761,
+      "learning_rate": 7.164461247637051e-06,
+      "loss": 1.0847,
       "step": 379
     },
     {
-      "epoch": 0.010436406580428992,
+      "epoch": 0.010783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 6.95333943275389e-06,
-      "loss": 1.16,
+      "learning_rate": 7.183364839319471e-06,
+      "loss": 1.101,
       "step": 380
     },
     {
-      "epoch": 0.010463870808272225,
+      "epoch": 0.010811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.9716376944190315e-06,
-      "loss": 1.1306,
+      "learning_rate": 7.2022684310018906e-06,
+      "loss": 1.1551,
       "step": 381
     },
     {
-      "epoch": 0.010491335036115459,
+      "epoch": 0.010839954597048809,
       "grad_norm": 0.0,
-      "learning_rate": 6.989935956084173e-06,
-      "loss": 1.1476,
+      "learning_rate": 7.221172022684311e-06,
+      "loss": 1.1528,
       "step": 382
     },
     {
-      "epoch": 0.010518799263958694,
+      "epoch": 0.010868331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 7.008234217749314e-06,
-      "loss": 1.1554,
+      "learning_rate": 7.240075614366731e-06,
+      "loss": 1.1289,
       "step": 383
     },
     {
-      "epoch": 0.010546263491801927,
+      "epoch": 0.010896708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 7.026532479414456e-06,
-      "loss": 1.2032,
+      "learning_rate": 7.25897920604915e-06,
+      "loss": 1.0753,
       "step": 384
     },
     {
-      "epoch": 0.010573727719645163,
+      "epoch": 0.010925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 7.044830741079598e-06,
-      "loss": 1.2022,
+      "learning_rate": 7.277882797731569e-06,
+      "loss": 1.02,
       "step": 385
     },
     {
-      "epoch": 0.010601191947488396,
+      "epoch": 0.010953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 7.0631290027447395e-06,
-      "loss": 1.134,
+      "learning_rate": 7.296786389413989e-06,
+      "loss": 1.0767,
       "step": 386
     },
     {
-      "epoch": 0.010628656175331631,
+      "epoch": 0.01098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 7.081427264409881e-06,
-      "loss": 1.1182,
+      "learning_rate": 7.3156899810964085e-06,
+      "loss": 1.037,
       "step": 387
     },
     {
-      "epoch": 0.010656120403174865,
+      "epoch": 0.011010215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 7.099725526075023e-06,
-      "loss": 1.0266,
+      "learning_rate": 7.334593572778829e-06,
+      "loss": 1.2678,
       "step": 388
     },
     {
-      "epoch": 0.010683584631018098,
+      "epoch": 0.011038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 7.118023787740165e-06,
-      "loss": 1.1895,
+      "learning_rate": 7.353497164461249e-06,
+      "loss": 1.0829,
       "step": 389
     },
     {
-      "epoch": 0.010711048858861334,
+      "epoch": 0.011066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 7.136322049405307e-06,
-      "loss": 1.2248,
+      "learning_rate": 7.372400756143668e-06,
+      "loss": 1.1423,
       "step": 390
     },
     {
-      "epoch": 0.010738513086704567,
+      "epoch": 0.011095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 7.154620311070449e-06,
-      "loss": 1.1049,
+      "learning_rate": 7.391304347826087e-06,
+      "loss": 1.0005,
       "step": 391
     },
     {
-      "epoch": 0.010765977314547802,
+      "epoch": 0.01112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 7.172918572735591e-06,
-      "loss": 1.0527,
+      "learning_rate": 7.410207939508507e-06,
+      "loss": 1.0382,
       "step": 392
     },
     {
-      "epoch": 0.010793441542391036,
+      "epoch": 0.011152099886492623,
       "grad_norm": 0.0,
-      "learning_rate": 7.191216834400733e-06,
-      "loss": 1.2294,
+      "learning_rate": 7.4291115311909265e-06,
+      "loss": 1.0858,
       "step": 393
     },
     {
-      "epoch": 0.01082090577023427,
+      "epoch": 0.011180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 7.2095150960658745e-06,
-      "loss": 1.09,
+      "learning_rate": 7.448015122873347e-06,
+      "loss": 1.0151,
       "step": 394
     },
     {
-      "epoch": 0.010848369998077504,
+      "epoch": 0.011208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 7.227813357731016e-06,
-      "loss": 1.2212,
+      "learning_rate": 7.466918714555767e-06,
+      "loss": 1.1294,
       "step": 395
     },
     {
-      "epoch": 0.010875834225920738,
+      "epoch": 0.011237230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 7.246111619396158e-06,
-      "loss": 1.0251,
+      "learning_rate": 7.485822306238186e-06,
+      "loss": 1.1019,
       "step": 396
     },
     {
-      "epoch": 0.010903298453763973,
+      "epoch": 0.011265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 7.2644098810613e-06,
-      "loss": 1.0366,
+      "learning_rate": 7.504725897920605e-06,
+      "loss": 1.0646,
       "step": 397
     },
     {
-      "epoch": 0.010930762681607207,
+      "epoch": 0.011293984108967084,
       "grad_norm": 0.0,
-      "learning_rate": 7.2827081427264416e-06,
-      "loss": 1.1177,
+      "learning_rate": 7.523629489603025e-06,
+      "loss": 1.1392,
       "step": 398
     },
     {
-      "epoch": 0.01095822690945044,
+      "epoch": 0.011322360953461974,
       "grad_norm": 0.0,
-      "learning_rate": 7.301006404391583e-06,
-      "loss": 1.1044,
+      "learning_rate": 7.542533081285444e-06,
+      "loss": 1.01,
       "step": 399
     },
     {
-      "epoch": 0.010985691137293675,
+      "epoch": 0.011350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 7.319304666056726e-06,
-      "loss": 1.1979,
+      "learning_rate": 7.561436672967865e-06,
+      "loss": 1.1423,
       "step": 400
     },
     {
-      "epoch": 0.011013155365136909,
+      "epoch": 0.01137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 7.337602927721868e-06,
-      "loss": 0.9887,
+      "learning_rate": 7.580340264650285e-06,
+      "loss": 0.9815,
       "step": 401
     },
     {
-      "epoch": 0.011040619592980144,
+      "epoch": 0.011407491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 7.3559011893870095e-06,
-      "loss": 1.1738,
+      "learning_rate": 7.599243856332704e-06,
+      "loss": 1.1646,
       "step": 402
     },
     {
-      "epoch": 0.011068083820823377,
+      "epoch": 0.011435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 7.374199451052151e-06,
-      "loss": 1.2031,
+      "learning_rate": 7.618147448015123e-06,
+      "loss": 1.092,
       "step": 403
     },
     {
-      "epoch": 0.011095548048666611,
+      "epoch": 0.011464245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 7.392497712717292e-06,
-      "loss": 1.1228,
+      "learning_rate": 7.637051039697544e-06,
+      "loss": 1.1157,
       "step": 404
     },
     {
-      "epoch": 0.011123012276509846,
+      "epoch": 0.011492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 7.410795974382434e-06,
-      "loss": 1.0339,
+      "learning_rate": 7.655954631379963e-06,
+      "loss": 1.081,
       "step": 405
     },
     {
-      "epoch": 0.01115047650435308,
+      "epoch": 0.01152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 7.429094236047576e-06,
-      "loss": 1.0858,
+      "learning_rate": 7.674858223062383e-06,
+      "loss": 1.0903,
       "step": 406
     },
     {
-      "epoch": 0.011177940732196315,
+      "epoch": 0.011549375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 7.4473924977127175e-06,
-      "loss": 1.0754,
+      "learning_rate": 7.693761814744803e-06,
+      "loss": 1.1231,
       "step": 407
     },
     {
-      "epoch": 0.011205404960039548,
+      "epoch": 0.011577752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 7.465690759377859e-06,
-      "loss": 1.1747,
+      "learning_rate": 7.712665406427222e-06,
+      "loss": 1.0812,
       "step": 408
     },
     {
-      "epoch": 0.011232869187882783,
+      "epoch": 0.011606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 7.483989021043001e-06,
-      "loss": 1.1652,
+      "learning_rate": 7.731568998109642e-06,
+      "loss": 1.1128,
       "step": 409
     },
     {
-      "epoch": 0.011260333415726017,
+      "epoch": 0.011634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 7.502287282708144e-06,
-      "loss": 1.1842,
+      "learning_rate": 7.750472589792062e-06,
+      "loss": 1.0083,
       "step": 410
     },
     {
-      "epoch": 0.01128779764356925,
+      "epoch": 0.011662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 7.520585544373285e-06,
-      "loss": 1.0657,
+      "learning_rate": 7.769376181474481e-06,
+      "loss": 1.0358,
       "step": 411
     },
     {
-      "epoch": 0.011315261871412486,
+      "epoch": 0.011691259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 7.538883806038427e-06,
-      "loss": 1.1514,
+      "learning_rate": 7.7882797731569e-06,
+      "loss": 1.1496,
       "step": 412
     },
     {
-      "epoch": 0.011342726099255719,
+      "epoch": 0.011719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 7.557182067703569e-06,
-      "loss": 1.0731,
+      "learning_rate": 7.80718336483932e-06,
+      "loss": 1.1403,
       "step": 413
     },
     {
-      "epoch": 0.011370190327098954,
+      "epoch": 0.011748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 7.575480329368711e-06,
-      "loss": 1.1253,
+      "learning_rate": 7.82608695652174e-06,
+      "loss": 1.0926,
       "step": 414
     },
     {
-      "epoch": 0.011397654554942188,
+      "epoch": 0.01177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 7.5937785910338525e-06,
-      "loss": 1.1044,
+      "learning_rate": 7.84499054820416e-06,
+      "loss": 1.1217,
       "step": 415
     },
     {
-      "epoch": 0.011425118782785421,
+      "epoch": 0.011804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 7.612076852698994e-06,
-      "loss": 1.1577,
+      "learning_rate": 7.86389413988658e-06,
+      "loss": 1.0541,
       "step": 416
     },
     {
-      "epoch": 0.011452583010628656,
+      "epoch": 0.011833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 7.630375114364136e-06,
-      "loss": 1.0702,
+      "learning_rate": 7.882797731568999e-06,
+      "loss": 1.1663,
       "step": 417
     },
     {
-      "epoch": 0.01148004723847189,
+      "epoch": 0.011861520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 7.648673376029277e-06,
-      "loss": 1.093,
+      "learning_rate": 7.901701323251419e-06,
+      "loss": 1.1205,
       "step": 418
     },
     {
-      "epoch": 0.011507511466315125,
+      "epoch": 0.011889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 7.66697163769442e-06,
-      "loss": 1.0846,
+      "learning_rate": 7.920604914933838e-06,
+      "loss": 1.0402,
       "step": 419
     },
     {
-      "epoch": 0.011534975694158359,
+      "epoch": 0.01191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 7.685269899359562e-06,
-      "loss": 1.0654,
+      "learning_rate": 7.939508506616258e-06,
+      "loss": 0.9959,
       "step": 420
     },
     {
-      "epoch": 0.011562439922001592,
+      "epoch": 0.011946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 7.703568161024703e-06,
-      "loss": 1.067,
+      "learning_rate": 7.958412098298678e-06,
+      "loss": 0.9762,
       "step": 421
     },
     {
-      "epoch": 0.011589904149844827,
+      "epoch": 0.011975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 7.721866422689846e-06,
-      "loss": 1.2257,
+      "learning_rate": 7.977315689981097e-06,
+      "loss": 1.0894,
       "step": 422
     },
     {
-      "epoch": 0.01161736837768806,
+      "epoch": 0.012003405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 7.740164684354987e-06,
-      "loss": 1.0793,
+      "learning_rate": 7.996219281663517e-06,
+      "loss": 1.1104,
       "step": 423
     },
     {
-      "epoch": 0.011644832605531296,
+      "epoch": 0.012031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 7.75846294602013e-06,
-      "loss": 1.2566,
+      "learning_rate": 8.015122873345937e-06,
+      "loss": 1.1134,
       "step": 424
     },
     {
-      "epoch": 0.01167229683337453,
+      "epoch": 0.012060158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 7.77676120768527e-06,
-      "loss": 1.0791,
+      "learning_rate": 8.034026465028356e-06,
+      "loss": 1.1134,
       "step": 425
     },
     {
-      "epoch": 0.011699761061217765,
+      "epoch": 0.012088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 7.795059469350413e-06,
-      "loss": 1.0902,
+      "learning_rate": 8.052930056710776e-06,
+      "loss": 1.1705,
       "step": 426
     },
     {
-      "epoch": 0.011727225289060998,
+      "epoch": 0.012116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 7.813357731015554e-06,
-      "loss": 1.12,
+      "learning_rate": 8.071833648393196e-06,
+      "loss": 1.0048,
       "step": 427
     },
     {
-      "epoch": 0.011754689516904232,
+      "epoch": 0.012145289443813849,
       "grad_norm": 0.0,
-      "learning_rate": 7.831655992680696e-06,
-      "loss": 1.1301,
+      "learning_rate": 8.090737240075615e-06,
+      "loss": 1.0726,
       "step": 428
     },
     {
-      "epoch": 0.011782153744747467,
+      "epoch": 0.01217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.849954254345837e-06,
-      "loss": 1.0629,
+      "learning_rate": 8.109640831758035e-06,
+      "loss": 1.1926,
       "step": 429
     },
     {
-      "epoch": 0.0118096179725907,
+      "epoch": 0.012202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 7.868252516010978e-06,
-      "loss": 1.1594,
+      "learning_rate": 8.128544423440455e-06,
+      "loss": 1.0779,
       "step": 430
     },
     {
-      "epoch": 0.011837082200433936,
+      "epoch": 0.012230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 7.886550777676123e-06,
-      "loss": 1.1732,
+      "learning_rate": 8.147448015122874e-06,
+      "loss": 1.1087,
       "step": 431
     },
     {
-      "epoch": 0.011864546428277169,
+      "epoch": 0.012258796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 7.904849039341263e-06,
-      "loss": 1.0484,
+      "learning_rate": 8.166351606805294e-06,
+      "loss": 1.1183,
       "step": 432
     },
     {
-      "epoch": 0.011892010656120403,
+      "epoch": 0.01228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 7.923147301006406e-06,
-      "loss": 1.0276,
+      "learning_rate": 8.185255198487714e-06,
+      "loss": 1.2776,
       "step": 433
     },
     {
-      "epoch": 0.011919474883963638,
+      "epoch": 0.0123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 7.941445562671547e-06,
-      "loss": 1.0932,
+      "learning_rate": 8.204158790170133e-06,
+      "loss": 0.9655,
       "step": 434
     },
     {
-      "epoch": 0.011946939111806871,
+      "epoch": 0.012343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 7.959743824336688e-06,
-      "loss": 1.2897,
+      "learning_rate": 8.223062381852553e-06,
+      "loss": 1.0998,
       "step": 435
     },
     {
-      "epoch": 0.011974403339650106,
+      "epoch": 0.012372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 7.97804208600183e-06,
-      "loss": 1.1136,
+      "learning_rate": 8.241965973534973e-06,
+      "loss": 1.1248,
       "step": 436
     },
     {
-      "epoch": 0.01200186756749334,
+      "epoch": 0.012400681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 7.996340347666971e-06,
-      "loss": 1.0561,
+      "learning_rate": 8.260869565217392e-06,
+      "loss": 1.1497,
       "step": 437
     },
     {
-      "epoch": 0.012029331795336573,
+      "epoch": 0.01242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 8.014638609332114e-06,
-      "loss": 1.2742,
+      "learning_rate": 8.279773156899812e-06,
+      "loss": 1.0568,
       "step": 438
     },
     {
-      "epoch": 0.012056796023179809,
+      "epoch": 0.012457434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 8.032936870997255e-06,
-      "loss": 1.1256,
+      "learning_rate": 8.298676748582232e-06,
+      "loss": 1.0396,
       "step": 439
     },
     {
-      "epoch": 0.012084260251023042,
+      "epoch": 0.012485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 8.051235132662398e-06,
-      "loss": 1.1754,
+      "learning_rate": 8.317580340264651e-06,
+      "loss": 1.096,
       "step": 440
     },
     {
-      "epoch": 0.012111724478866277,
+      "epoch": 0.012514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 8.06953339432754e-06,
-      "loss": 1.1093,
+      "learning_rate": 8.336483931947071e-06,
+      "loss": 1.109,
       "step": 441
     },
     {
-      "epoch": 0.01213918870670951,
+      "epoch": 0.012542565266742339,
       "grad_norm": 0.0,
-      "learning_rate": 8.087831655992681e-06,
-      "loss": 1.1612,
+      "learning_rate": 8.35538752362949e-06,
+      "loss": 1.0496,
       "step": 442
     },
     {
-      "epoch": 0.012166652934552744,
+      "epoch": 0.012570942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 8.106129917657824e-06,
-      "loss": 1.1297,
+      "learning_rate": 8.37429111531191e-06,
+      "loss": 1.0344,
       "step": 443
     },
     {
-      "epoch": 0.01219411716239598,
+      "epoch": 0.012599318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 8.124428179322965e-06,
-      "loss": 1.2053,
+      "learning_rate": 8.39319470699433e-06,
+      "loss": 0.9294,
       "step": 444
     },
     {
-      "epoch": 0.012221581390239213,
+      "epoch": 0.012627695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 8.142726440988107e-06,
-      "loss": 1.166,
+      "learning_rate": 8.41209829867675e-06,
+      "loss": 1.0494,
       "step": 445
     },
     {
-      "epoch": 0.012249045618082448,
+      "epoch": 0.012656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 8.161024702653248e-06,
-      "loss": 1.135,
+      "learning_rate": 8.43100189035917e-06,
+      "loss": 0.9863,
       "step": 446
     },
     {
-      "epoch": 0.012276509845925682,
+      "epoch": 0.0126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 8.179322964318391e-06,
-      "loss": 1.0687,
+      "learning_rate": 8.449905482041589e-06,
+      "loss": 1.136,
       "step": 447
     },
     {
-      "epoch": 0.012303974073768917,
+      "epoch": 0.012712826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 8.197621225983532e-06,
-      "loss": 0.9941,
+      "learning_rate": 8.468809073724009e-06,
+      "loss": 1.1558,
       "step": 448
     },
     {
-      "epoch": 0.01233143830161215,
+      "epoch": 0.012741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 8.215919487648674e-06,
-      "loss": 1.0218,
+      "learning_rate": 8.487712665406428e-06,
+      "loss": 1.1218,
       "step": 449
     },
     {
-      "epoch": 0.012358902529455384,
+      "epoch": 0.012769580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 8.234217749313815e-06,
-      "loss": 1.0146,
+      "learning_rate": 8.506616257088848e-06,
+      "loss": 1.202,
       "step": 450
     },
     {
-      "epoch": 0.012386366757298619,
+      "epoch": 0.012797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 8.252516010978958e-06,
-      "loss": 1.0087,
+      "learning_rate": 8.525519848771267e-06,
+      "loss": 1.0699,
       "step": 451
     },
     {
-      "epoch": 0.012413830985141852,
+      "epoch": 0.01282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 8.2708142726441e-06,
-      "loss": 1.1233,
+      "learning_rate": 8.544423440453687e-06,
+      "loss": 0.9576,
       "step": 452
     },
     {
-      "epoch": 0.012441295212985088,
+      "epoch": 0.012854710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 8.289112534309241e-06,
-      "loss": 1.0727,
+      "learning_rate": 8.563327032136107e-06,
+      "loss": 1.1714,
       "step": 453
     },
     {
-      "epoch": 0.012468759440828321,
+      "epoch": 0.012883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 8.307410795974384e-06,
-      "loss": 1.0584,
+      "learning_rate": 8.582230623818526e-06,
+      "loss": 1.0603,
       "step": 454
     },
     {
-      "epoch": 0.012496223668671555,
+      "epoch": 0.012911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 8.325709057639525e-06,
-      "loss": 1.0406,
+      "learning_rate": 8.601134215500946e-06,
+      "loss": 1.1562,
       "step": 455
     },
     {
-      "epoch": 0.01252368789651479,
+      "epoch": 0.012939841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 8.344007319304666e-06,
-      "loss": 1.1141,
+      "learning_rate": 8.620037807183366e-06,
+      "loss": 1.1459,
       "step": 456
     },
     {
-      "epoch": 0.012551152124358023,
+      "epoch": 0.012968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 8.362305580969809e-06,
-      "loss": 1.1128,
+      "learning_rate": 8.638941398865785e-06,
+      "loss": 1.029,
       "step": 457
     },
     {
-      "epoch": 0.012578616352201259,
+      "epoch": 0.012996594778660614,
       "grad_norm": 0.0,
-      "learning_rate": 8.38060384263495e-06,
-      "loss": 1.0663,
+      "learning_rate": 8.657844990548205e-06,
+      "loss": 1.1277,
       "step": 458
     },
     {
-      "epoch": 0.012606080580044492,
+      "epoch": 0.013024971623155504,
       "grad_norm": 0.0,
-      "learning_rate": 8.398902104300092e-06,
-      "loss": 1.0608,
+      "learning_rate": 8.676748582230625e-06,
+      "loss": 1.1028,
       "step": 459
     },
     {
-      "epoch": 0.012633544807887725,
+      "epoch": 0.013053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 8.417200365965233e-06,
-      "loss": 1.0624,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 1.0356,
       "step": 460
     },
     {
-      "epoch": 0.01266100903573096,
+      "epoch": 0.01308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 8.435498627630376e-06,
-      "loss": 1.1125,
+      "learning_rate": 8.714555765595464e-06,
+      "loss": 1.1004,
       "step": 461
     },
     {
-      "epoch": 0.012688473263574194,
+      "epoch": 0.013110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 8.453796889295518e-06,
-      "loss": 1.125,
+      "learning_rate": 8.733459357277884e-06,
+      "loss": 1.1775,
       "step": 462
     },
     {
-      "epoch": 0.01271593749141743,
+      "epoch": 0.013138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 8.47209515096066e-06,
-      "loss": 1.1604,
+      "learning_rate": 8.752362948960303e-06,
+      "loss": 0.9222,
       "step": 463
     },
     {
-      "epoch": 0.012743401719260663,
+      "epoch": 0.013166855845629965,
       "grad_norm": 0.0,
-      "learning_rate": 8.490393412625802e-06,
-      "loss": 1.2139,
+      "learning_rate": 8.771266540642723e-06,
+      "loss": 1.1071,
       "step": 464
     },
     {
-      "epoch": 0.012770865947103896,
+      "epoch": 0.013195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 8.508691674290943e-06,
-      "loss": 1.184,
+      "learning_rate": 8.790170132325143e-06,
+      "loss": 1.0911,
       "step": 465
     },
     {
-      "epoch": 0.012798330174947132,
+      "epoch": 0.01322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 8.526989935956085e-06,
-      "loss": 1.1196,
+      "learning_rate": 8.809073724007562e-06,
+      "loss": 1.1357,
       "step": 466
     },
     {
-      "epoch": 0.012825794402790365,
+      "epoch": 0.013251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 8.545288197621226e-06,
-      "loss": 1.1046,
+      "learning_rate": 8.827977315689982e-06,
+      "loss": 1.1003,
       "step": 467
     },
     {
-      "epoch": 0.0128532586306336,
+      "epoch": 0.013280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 8.563586459286369e-06,
-      "loss": 1.0825,
+      "learning_rate": 8.846880907372402e-06,
+      "loss": 1.1507,
       "step": 468
     },
     {
-      "epoch": 0.012880722858476834,
+      "epoch": 0.013308740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 8.58188472095151e-06,
-      "loss": 1.1666,
+      "learning_rate": 8.865784499054821e-06,
+      "loss": 1.1572,
       "step": 469
     },
     {
-      "epoch": 0.012908187086320069,
+      "epoch": 0.013337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 8.600182982616652e-06,
-      "loss": 1.0309,
+      "learning_rate": 8.884688090737241e-06,
+      "loss": 1.0898,
       "step": 470
     },
     {
-      "epoch": 0.012935651314163302,
+      "epoch": 0.013365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 8.618481244281795e-06,
-      "loss": 1.107,
+      "learning_rate": 8.90359168241966e-06,
+      "loss": 1.0504,
       "step": 471
     },
     {
-      "epoch": 0.012963115542006536,
+      "epoch": 0.013393870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 8.636779505946936e-06,
-      "loss": 1.1352,
+      "learning_rate": 8.92249527410208e-06,
+      "loss": 1.0406,
       "step": 472
     },
     {
-      "epoch": 0.012990579769849771,
+      "epoch": 0.013422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 8.655077767612079e-06,
-      "loss": 1.1299,
+      "learning_rate": 8.9413988657845e-06,
+      "loss": 0.9498,
       "step": 473
     },
     {
-      "epoch": 0.013018043997693005,
+      "epoch": 0.013450624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 8.67337602927722e-06,
-      "loss": 1.1337,
+      "learning_rate": 8.96030245746692e-06,
+      "loss": 1.0526,
       "step": 474
     },
     {
-      "epoch": 0.01304550822553624,
+      "epoch": 0.01347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 8.69167429094236e-06,
-      "loss": 1.1982,
+      "learning_rate": 8.97920604914934e-06,
+      "loss": 1.0418,
       "step": 475
     },
     {
-      "epoch": 0.013072972453379473,
+      "epoch": 0.013507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 8.709972552607503e-06,
-      "loss": 1.112,
+      "learning_rate": 8.998109640831759e-06,
+      "loss": 1.1428,
       "step": 476
     },
     {
-      "epoch": 0.013100436681222707,
+      "epoch": 0.013535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 8.728270814272644e-06,
-      "loss": 1.0905,
+      "learning_rate": 9.017013232514179e-06,
+      "loss": 1.0874,
       "step": 477
     },
     {
-      "epoch": 0.013127900909065942,
+      "epoch": 0.013564131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 8.746569075937787e-06,
-      "loss": 1.1716,
+      "learning_rate": 9.035916824196598e-06,
+      "loss": 1.1464,
       "step": 478
     },
     {
-      "epoch": 0.013155365136909175,
+      "epoch": 0.013592508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 8.764867337602928e-06,
-      "loss": 1.1954,
+      "learning_rate": 9.054820415879018e-06,
+      "loss": 1.1772,
       "step": 479
     },
     {
-      "epoch": 0.01318282936475241,
+      "epoch": 0.01362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 8.78316559926807e-06,
-      "loss": 1.114,
+      "learning_rate": 9.073724007561438e-06,
+      "loss": 1.1157,
       "step": 480
     },
     {
-      "epoch": 0.013210293592595644,
+      "epoch": 0.013649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 8.801463860933213e-06,
-      "loss": 1.1398,
+      "learning_rate": 9.092627599243857e-06,
+      "loss": 1.0257,
       "step": 481
     },
     {
-      "epoch": 0.013237757820438878,
+      "epoch": 0.013677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 8.819762122598354e-06,
-      "loss": 1.1361,
+      "learning_rate": 9.111531190926277e-06,
+      "loss": 1.2037,
       "step": 482
     },
     {
-      "epoch": 0.013265222048282113,
+      "epoch": 0.013706015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 8.838060384263496e-06,
-      "loss": 1.0065,
+      "learning_rate": 9.130434782608697e-06,
+      "loss": 1.0027,
       "step": 483
     },
     {
-      "epoch": 0.013292686276125346,
+      "epoch": 0.013734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 8.856358645928637e-06,
-      "loss": 1.2231,
+      "learning_rate": 9.149338374291116e-06,
+      "loss": 1.0844,
       "step": 484
     },
     {
-      "epoch": 0.013320150503968581,
+      "epoch": 0.013762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 8.87465690759378e-06,
-      "loss": 1.1344,
+      "learning_rate": 9.168241965973536e-06,
+      "loss": 1.0396,
       "step": 485
     },
     {
-      "epoch": 0.013347614731811815,
+      "epoch": 0.013791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 8.89295516925892e-06,
-      "loss": 0.9965,
+      "learning_rate": 9.187145557655956e-06,
+      "loss": 1.1398,
       "step": 486
     },
     {
-      "epoch": 0.013375078959655048,
+      "epoch": 0.013819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 8.911253430924063e-06,
-      "loss": 1.1442,
+      "learning_rate": 9.206049149338375e-06,
+      "loss": 1.08,
       "step": 487
     },
     {
-      "epoch": 0.013402543187498284,
+      "epoch": 0.013847900113507379,
       "grad_norm": 0.0,
-      "learning_rate": 8.929551692589204e-06,
-      "loss": 1.1366,
+      "learning_rate": 9.224952741020795e-06,
+      "loss": 1.0933,
       "step": 488
     },
     {
-      "epoch": 0.013430007415341517,
+      "epoch": 0.01387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 8.947849954254347e-06,
-      "loss": 1.2277,
+      "learning_rate": 9.243856332703214e-06,
+      "loss": 1.1354,
       "step": 489
     },
     {
-      "epoch": 0.013457471643184752,
+      "epoch": 0.013904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 8.966148215919488e-06,
-      "loss": 1.0955,
+      "learning_rate": 9.262759924385634e-06,
+      "loss": 0.9667,
       "step": 490
     },
     {
-      "epoch": 0.013484935871027986,
+      "epoch": 0.013933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 8.98444647758463e-06,
-      "loss": 1.0126,
+      "learning_rate": 9.281663516068054e-06,
+      "loss": 0.9978,
       "step": 491
     },
     {
-      "epoch": 0.013512400098871221,
+      "epoch": 0.013961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 9.002744739249773e-06,
-      "loss": 1.1127,
+      "learning_rate": 9.300567107750473e-06,
+      "loss": 1.0295,
       "step": 492
     },
     {
-      "epoch": 0.013539864326714455,
+      "epoch": 0.01398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 9.021043000914914e-06,
-      "loss": 1.1368,
+      "learning_rate": 9.319470699432893e-06,
+      "loss": 1.045,
       "step": 493
     },
     {
-      "epoch": 0.013567328554557688,
+      "epoch": 0.01401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 9.039341262580057e-06,
-      "loss": 1.0303,
+      "learning_rate": 9.338374291115313e-06,
+      "loss": 1.1342,
       "step": 494
     },
     {
-      "epoch": 0.013594792782400923,
+      "epoch": 0.014046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 9.057639524245198e-06,
-      "loss": 1.0304,
+      "learning_rate": 9.357277882797732e-06,
+      "loss": 1.0482,
       "step": 495
     },
     {
-      "epoch": 0.013622257010244157,
+      "epoch": 0.014074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 9.075937785910338e-06,
-      "loss": 1.1729,
+      "learning_rate": 9.376181474480152e-06,
+      "loss": 1.0491,
       "step": 496
     },
     {
-      "epoch": 0.013649721238087392,
+      "epoch": 0.014103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 9.094236047575481e-06,
-      "loss": 1.1046,
+      "learning_rate": 9.395085066162572e-06,
+      "loss": 1.0694,
       "step": 497
     },
     {
-      "epoch": 0.013677185465930625,
+      "epoch": 0.0141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.112534309240622e-06,
-      "loss": 1.0938,
+      "learning_rate": 9.413988657844991e-06,
+      "loss": 1.1411,
       "step": 498
     },
     {
-      "epoch": 0.013704649693773859,
+      "epoch": 0.014160045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 9.130832570905765e-06,
-      "loss": 1.1035,
+      "learning_rate": 9.432892249527411e-06,
+      "loss": 1.0182,
       "step": 499
     },
     {
-      "epoch": 0.013732113921617094,
+      "epoch": 0.014188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.149130832570906e-06,
-      "loss": 1.1687,
+      "learning_rate": 9.45179584120983e-06,
+      "loss": 1.0368,
       "step": 500
     },
     {
-      "epoch": 0.013759578149460328,
+      "epoch": 0.014216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 9.167429094236048e-06,
-      "loss": 0.9672,
+      "learning_rate": 9.47069943289225e-06,
+      "loss": 1.0369,
       "step": 501
     },
     {
-      "epoch": 0.013787042377303563,
+      "epoch": 0.014245175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 9.18572735590119e-06,
-      "loss": 1.0415,
+      "learning_rate": 9.48960302457467e-06,
+      "loss": 1.0325,
       "step": 502
     },
     {
-      "epoch": 0.013814506605146796,
+      "epoch": 0.014273552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 9.204025617566332e-06,
-      "loss": 1.0899,
+      "learning_rate": 9.50850661625709e-06,
+      "loss": 1.0409,
       "step": 503
     },
     {
-      "epoch": 0.01384197083299003,
+      "epoch": 0.014301929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 9.222323879231474e-06,
-      "loss": 1.1692,
+      "learning_rate": 9.52741020793951e-06,
+      "loss": 1.0643,
       "step": 504
     },
     {
-      "epoch": 0.013869435060833265,
+      "epoch": 0.014330306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 9.240622140896615e-06,
-      "loss": 1.0685,
+      "learning_rate": 9.546313799621929e-06,
+      "loss": 1.1072,
       "step": 505
     },
     {
-      "epoch": 0.013896899288676498,
+      "epoch": 0.014358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.258920402561758e-06,
-      "loss": 1.1523,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 1.0961,
       "step": 506
     },
     {
-      "epoch": 0.013924363516519734,
+      "epoch": 0.01438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 9.277218664226899e-06,
-      "loss": 1.2092,
+      "learning_rate": 9.584120982986768e-06,
+      "loss": 1.0848,
       "step": 507
     },
     {
-      "epoch": 0.013951827744362967,
+      "epoch": 0.014415437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 9.295516925892041e-06,
-      "loss": 1.1728,
+      "learning_rate": 9.603024574669188e-06,
+      "loss": 1.1181,
       "step": 508
     },
     {
-      "epoch": 0.013979291972206202,
+      "epoch": 0.014443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 9.313815187557182e-06,
-      "loss": 1.1227,
+      "learning_rate": 9.621928166351608e-06,
+      "loss": 1.0671,
       "step": 509
     },
     {
-      "epoch": 0.014006756200049436,
+      "epoch": 0.014472190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 9.332113449222325e-06,
-      "loss": 1.132,
+      "learning_rate": 9.640831758034027e-06,
+      "loss": 1.0238,
       "step": 510
     },
     {
-      "epoch": 0.01403422042789267,
+      "epoch": 0.014500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 9.350411710887466e-06,
-      "loss": 1.1444,
+      "learning_rate": 9.659735349716447e-06,
+      "loss": 1.0671,
       "step": 511
     },
     {
-      "epoch": 0.014061684655735904,
+      "epoch": 0.01452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 9.368709972552608e-06,
-      "loss": 1.0914,
+      "learning_rate": 9.678638941398867e-06,
+      "loss": 1.0871,
       "step": 512
     },
     {
-      "epoch": 0.014089148883579138,
+      "epoch": 0.014557321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 9.387008234217751e-06,
-      "loss": 1.1141,
+      "learning_rate": 9.697542533081286e-06,
+      "loss": 1.0355,
       "step": 513
     },
     {
-      "epoch": 0.014116613111422373,
+      "epoch": 0.014585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 9.405306495882892e-06,
-      "loss": 1.179,
+      "learning_rate": 9.716446124763706e-06,
+      "loss": 1.0803,
       "step": 514
     },
     {
-      "epoch": 0.014144077339265607,
+      "epoch": 0.014614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 9.423604757548035e-06,
-      "loss": 1.1824,
+      "learning_rate": 9.735349716446126e-06,
+      "loss": 1.0284,
       "step": 515
     },
     {
-      "epoch": 0.01417154156710884,
+      "epoch": 0.014642451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 9.441903019213176e-06,
-      "loss": 1.072,
+      "learning_rate": 9.754253308128545e-06,
+      "loss": 1.0406,
       "step": 516
     },
     {
-      "epoch": 0.014199005794952075,
+      "epoch": 0.014670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 9.460201280878316e-06,
-      "loss": 1.0786,
+      "learning_rate": 9.773156899810965e-06,
+      "loss": 1.053,
       "step": 517
     },
     {
-      "epoch": 0.014226470022795309,
+      "epoch": 0.014699205448354144,
       "grad_norm": 0.0,
-      "learning_rate": 9.478499542543459e-06,
-      "loss": 1.1263,
+      "learning_rate": 9.792060491493385e-06,
+      "loss": 1.1042,
       "step": 518
     },
     {
-      "epoch": 0.014253934250638544,
+      "epoch": 0.014727582292849034,
       "grad_norm": 0.0,
-      "learning_rate": 9.4967978042086e-06,
-      "loss": 1.1624,
+      "learning_rate": 9.810964083175804e-06,
+      "loss": 1.0751,
       "step": 519
     },
     {
-      "epoch": 0.014281398478481777,
+      "epoch": 0.014755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 9.515096065873743e-06,
-      "loss": 1.0445,
+      "learning_rate": 9.829867674858224e-06,
+      "loss": 0.9958,
       "step": 520
     },
     {
-      "epoch": 0.014308862706325011,
+      "epoch": 0.01478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 9.533394327538884e-06,
-      "loss": 1.0809,
+      "learning_rate": 9.848771266540644e-06,
+      "loss": 1.1558,
       "step": 521
     },
     {
-      "epoch": 0.014336326934168246,
+      "epoch": 0.014812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 9.551692589204026e-06,
-      "loss": 1.0471,
+      "learning_rate": 9.867674858223063e-06,
+      "loss": 0.9994,
       "step": 522
     },
     {
-      "epoch": 0.01436379116201148,
+      "epoch": 0.014841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 9.569990850869169e-06,
-      "loss": 0.957,
+      "learning_rate": 9.886578449905483e-06,
+      "loss": 1.0357,
       "step": 523
     },
     {
-      "epoch": 0.014391255389854715,
+      "epoch": 0.014869466515323495,
       "grad_norm": 0.0,
-      "learning_rate": 9.58828911253431e-06,
-      "loss": 1.1519,
+      "learning_rate": 9.905482041587903e-06,
+      "loss": 1.0869,
       "step": 524
     },
     {
-      "epoch": 0.014418719617697948,
+      "epoch": 0.014897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 9.606587374199452e-06,
-      "loss": 1.1639,
+      "learning_rate": 9.924385633270322e-06,
+      "loss": 1.0451,
       "step": 525
     },
     {
-      "epoch": 0.014446183845541182,
+      "epoch": 0.01492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 9.624885635864593e-06,
-      "loss": 1.1105,
+      "learning_rate": 9.943289224952742e-06,
+      "loss": 1.1299,
       "step": 526
     },
     {
-      "epoch": 0.014473648073384417,
+      "epoch": 0.014954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 9.643183897529736e-06,
-      "loss": 1.1614,
+      "learning_rate": 9.962192816635162e-06,
+      "loss": 1.2252,
       "step": 527
     },
     {
-      "epoch": 0.01450111230122765,
+      "epoch": 0.014982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 9.661482159194877e-06,
-      "loss": 1.0829,
+      "learning_rate": 9.981096408317581e-06,
+      "loss": 1.0944,
       "step": 528
     },
     {
-      "epoch": 0.014528576529070886,
+      "epoch": 0.015011350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 9.67978042086002e-06,
-      "loss": 1.0879,
+      "learning_rate": 1e-05,
+      "loss": 1.1097,
       "step": 529
     },
     {
-      "epoch": 0.01455604075691412,
+      "epoch": 0.015039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 9.69807868252516e-06,
-      "loss": 1.1412,
+      "learning_rate": 1.001890359168242e-05,
+      "loss": 1.1234,
       "step": 530
     },
     {
-      "epoch": 0.014583504984757354,
+      "epoch": 0.015068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 9.716376944190303e-06,
-      "loss": 1.0612,
+      "learning_rate": 1.003780718336484e-05,
+      "loss": 1.121,
       "step": 531
     },
     {
-      "epoch": 0.014610969212600588,
+      "epoch": 0.015096481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 9.734675205855446e-06,
-      "loss": 1.0485,
+      "learning_rate": 1.005671077504726e-05,
+      "loss": 1.1518,
       "step": 532
     },
     {
-      "epoch": 0.014638433440443821,
+      "epoch": 0.015124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 9.752973467520586e-06,
-      "loss": 1.2066,
+      "learning_rate": 1.007561436672968e-05,
+      "loss": 1.0138,
       "step": 533
     },
     {
-      "epoch": 0.014665897668287057,
+      "epoch": 0.015153234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 9.771271729185729e-06,
-      "loss": 1.2123,
+      "learning_rate": 1.00945179584121e-05,
+      "loss": 1.1559,
       "step": 534
     },
     {
-      "epoch": 0.01469336189613029,
+      "epoch": 0.01518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 9.78956999085087e-06,
-      "loss": 1.1908,
+      "learning_rate": 1.011342155009452e-05,
+      "loss": 1.0013,
       "step": 535
     },
     {
-      "epoch": 0.014720826123973525,
+      "epoch": 0.015209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 9.807868252516013e-06,
-      "loss": 1.151,
+      "learning_rate": 1.0132325141776937e-05,
+      "loss": 1.1763,
       "step": 536
     },
     {
-      "epoch": 0.014748290351816759,
+      "epoch": 0.015238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 9.826166514181154e-06,
-      "loss": 1.0673,
+      "learning_rate": 1.0151228733459358e-05,
+      "loss": 0.9456,
       "step": 537
     },
     {
-      "epoch": 0.014775754579659992,
+      "epoch": 0.015266742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 9.844464775846294e-06,
-      "loss": 1.1224,
+      "learning_rate": 1.0170132325141778e-05,
+      "loss": 1.0938,
       "step": 538
     },
     {
-      "epoch": 0.014803218807503227,
+      "epoch": 0.015295119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 9.862763037511437e-06,
-      "loss": 1.0863,
+      "learning_rate": 1.0189035916824197e-05,
+      "loss": 1.0457,
       "step": 539
     },
     {
-      "epoch": 0.014830683035346461,
+      "epoch": 0.01532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 9.881061299176578e-06,
-      "loss": 1.1022,
+      "learning_rate": 1.0207939508506617e-05,
+      "loss": 0.976,
       "step": 540
     },
     {
-      "epoch": 0.014858147263189696,
+      "epoch": 0.015351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 9.89935956084172e-06,
-      "loss": 1.0009,
+      "learning_rate": 1.0226843100189037e-05,
+      "loss": 1.2176,
       "step": 541
     },
     {
-      "epoch": 0.01488561149103293,
+      "epoch": 0.015380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 9.917657822506863e-06,
-      "loss": 1.1115,
+      "learning_rate": 1.0245746691871456e-05,
+      "loss": 1.0466,
       "step": 542
     },
     {
-      "epoch": 0.014913075718876163,
+      "epoch": 0.015408626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 9.935956084172004e-06,
-      "loss": 1.1578,
+      "learning_rate": 1.0264650283553876e-05,
+      "loss": 1.0492,
       "step": 543
     },
     {
-      "epoch": 0.014940539946719398,
+      "epoch": 0.015437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 9.954254345837147e-06,
-      "loss": 1.1396,
+      "learning_rate": 1.0283553875236296e-05,
+      "loss": 1.1672,
       "step": 544
     },
     {
-      "epoch": 0.014968004174562632,
+      "epoch": 0.015465380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 9.972552607502288e-06,
-      "loss": 1.1135,
+      "learning_rate": 1.0302457466918715e-05,
+      "loss": 1.0318,
       "step": 545
     },
     {
-      "epoch": 0.014995468402405867,
+      "epoch": 0.015493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 9.99085086916743e-06,
-      "loss": 1.0934,
+      "learning_rate": 1.0321361058601137e-05,
+      "loss": 1.1222,
       "step": 546
     },
     {
-      "epoch": 0.0150229326302491,
+      "epoch": 0.015522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.0009149130832571e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.0340264650283556e-05,
+      "loss": 1.1519,
       "step": 547
     },
     {
-      "epoch": 0.015050396858092334,
+      "epoch": 0.015550510783200909,
       "grad_norm": 0.0,
-      "learning_rate": 1.0027447392497714e-05,
-      "loss": 1.1591,
+      "learning_rate": 1.0359168241965973e-05,
+      "loss": 1.0659,
       "step": 548
     },
     {
-      "epoch": 0.015077861085935569,
+      "epoch": 0.0155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.0045745654162855e-05,
-      "loss": 1.1989,
+      "learning_rate": 1.0378071833648394e-05,
+      "loss": 1.1028,
       "step": 549
     },
     {
-      "epoch": 0.015105325313778803,
+      "epoch": 0.015607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.0064043915827997e-05,
-      "loss": 1.1375,
+      "learning_rate": 1.0396975425330814e-05,
+      "loss": 1.1011,
       "step": 550
     },
     {
-      "epoch": 0.015132789541622038,
+      "epoch": 0.015635641316685586,
       "grad_norm": 0.0,
-      "learning_rate": 1.0082342177493138e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.0415879017013233e-05,
+      "loss": 1.0847,
       "step": 551
     },
     {
-      "epoch": 0.015160253769465271,
+      "epoch": 0.015664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.0100640439158281e-05,
-      "loss": 1.2038,
+      "learning_rate": 1.0434782608695653e-05,
+      "loss": 1.1486,
       "step": 552
     },
     {
-      "epoch": 0.015187717997308506,
+      "epoch": 0.015692395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 1.0118938700823422e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.0453686200378073e-05,
+      "loss": 1.1779,
       "step": 553
     },
     {
-      "epoch": 0.01521518222515174,
+      "epoch": 0.015720771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 1.0137236962488563e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.0472589792060492e-05,
+      "loss": 1.076,
       "step": 554
     },
     {
-      "epoch": 0.015242646452994973,
+      "epoch": 0.015749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.0155535224153705e-05,
-      "loss": 1.1065,
+      "learning_rate": 1.0491493383742912e-05,
+      "loss": 1.0219,
       "step": 555
     },
     {
-      "epoch": 0.015270110680838209,
+      "epoch": 0.015777525539160047,
       "grad_norm": 0.0,
-      "learning_rate": 1.0173833485818846e-05,
-      "loss": 1.1435,
+      "learning_rate": 1.0510396975425332e-05,
+      "loss": 1.0999,
       "step": 556
     },
     {
-      "epoch": 0.015297574908681442,
+      "epoch": 0.015805902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.019213174748399e-05,
-      "loss": 1.2084,
+      "learning_rate": 1.0529300567107751e-05,
+      "loss": 1.2273,
       "step": 557
     },
     {
-      "epoch": 0.015325039136524677,
+      "epoch": 0.01583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.0210430009149133e-05,
-      "loss": 1.205,
+      "learning_rate": 1.0548204158790173e-05,
+      "loss": 1.0422,
       "step": 558
     },
     {
-      "epoch": 0.01535250336436791,
+      "epoch": 0.015862656072644723,
       "grad_norm": 0.0,
-      "learning_rate": 1.0228728270814274e-05,
-      "loss": 1.1865,
+      "learning_rate": 1.0567107750472592e-05,
+      "loss": 1.0613,
       "step": 559
     },
     {
-      "epoch": 0.015379967592211144,
+      "epoch": 0.015891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.0247026532479417e-05,
-      "loss": 1.1472,
+      "learning_rate": 1.0586011342155009e-05,
+      "loss": 0.9979,
       "step": 560
     },
     {
-      "epoch": 0.01540743182005438,
+      "epoch": 0.015919409761634508,
       "grad_norm": 0.0,
-      "learning_rate": 1.0265324794144558e-05,
-      "loss": 1.0334,
+      "learning_rate": 1.060491493383743e-05,
+      "loss": 1.0312,
       "step": 561
     },
     {
-      "epoch": 0.015434896047897613,
+      "epoch": 0.0159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.0283623055809699e-05,
-      "loss": 1.2318,
+      "learning_rate": 1.062381852551985e-05,
+      "loss": 1.112,
       "step": 562
     },
     {
-      "epoch": 0.015462360275740848,
+      "epoch": 0.01597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.0301921317474841e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.064272211720227e-05,
+      "loss": 0.9374,
       "step": 563
     },
     {
-      "epoch": 0.015489824503584082,
+      "epoch": 0.016004540295119184,
       "grad_norm": 0.0,
-      "learning_rate": 1.0320219579139982e-05,
-      "loss": 1.0425,
+      "learning_rate": 1.0661625708884689e-05,
+      "loss": 1.089,
       "step": 564
     },
     {
-      "epoch": 0.015517288731427315,
+      "epoch": 0.016032917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 1.0338517840805125e-05,
-      "loss": 1.1456,
+      "learning_rate": 1.0680529300567109e-05,
+      "loss": 1.1803,
       "step": 565
     },
     {
-      "epoch": 0.01554475295927055,
+      "epoch": 0.01606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.0356816102470266e-05,
-      "loss": 1.1198,
+      "learning_rate": 1.0699432892249528e-05,
+      "loss": 0.989,
       "step": 566
     },
     {
-      "epoch": 0.015572217187113784,
+      "epoch": 0.01608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.0375114364135408e-05,
-      "loss": 1.064,
+      "learning_rate": 1.0718336483931948e-05,
+      "loss": 1.1544,
       "step": 567
     },
     {
-      "epoch": 0.015599681414957019,
+      "epoch": 0.01611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.039341262580055e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.0737240075614367e-05,
+      "loss": 1.0525,
       "step": 568
     },
     {
-      "epoch": 0.015627145642800253,
+      "epoch": 0.016146424517593645,
       "grad_norm": 0.0,
-      "learning_rate": 1.0411710887465692e-05,
-      "loss": 1.1238,
+      "learning_rate": 1.0756143667296787e-05,
+      "loss": 1.1099,
       "step": 569
     },
     {
-      "epoch": 0.015654609870643488,
+      "epoch": 0.016174801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 1.0430009149130833e-05,
-      "loss": 1.1685,
+      "learning_rate": 1.0775047258979208e-05,
+      "loss": 1.0101,
       "step": 570
     },
     {
-      "epoch": 0.01568207409848672,
+      "epoch": 0.01620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.0448307410795975e-05,
-      "loss": 1.1176,
+      "learning_rate": 1.0793950850661628e-05,
+      "loss": 1.0596,
       "step": 571
     },
     {
-      "epoch": 0.015709538326329955,
+      "epoch": 0.01623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.0466605672461116e-05,
-      "loss": 1.1141,
+      "learning_rate": 1.0812854442344048e-05,
+      "loss": 1.0419,
       "step": 572
     },
     {
-      "epoch": 0.01573700255417319,
+      "epoch": 0.01625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.0484903934126259e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.0831758034026466e-05,
+      "loss": 1.2113,
       "step": 573
     },
     {
-      "epoch": 0.015764466782016425,
+      "epoch": 0.016288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.05032021957914e-05,
-      "loss": 1.1573,
+      "learning_rate": 1.0850661625708885e-05,
+      "loss": 0.9956,
       "step": 574
     },
     {
-      "epoch": 0.015791931009859657,
+      "epoch": 0.016316685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.0521500457456541e-05,
-      "loss": 1.0947,
+      "learning_rate": 1.0869565217391305e-05,
+      "loss": 1.054,
       "step": 575
     },
     {
-      "epoch": 0.015819395237702892,
+      "epoch": 0.01634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.0539798719121683e-05,
-      "loss": 1.1157,
+      "learning_rate": 1.0888468809073725e-05,
+      "loss": 1.1067,
       "step": 576
     },
     {
-      "epoch": 0.015846859465546127,
+      "epoch": 0.01637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.0558096980786828e-05,
-      "loss": 1.1666,
+      "learning_rate": 1.0907372400756144e-05,
+      "loss": 1.0895,
       "step": 577
     },
     {
-      "epoch": 0.01587432369338936,
+      "epoch": 0.016401816118047672,
       "grad_norm": 0.0,
-      "learning_rate": 1.0576395242451969e-05,
-      "loss": 1.1404,
+      "learning_rate": 1.0926275992438564e-05,
+      "loss": 1.011,
       "step": 578
     },
     {
-      "epoch": 0.015901787921232594,
+      "epoch": 0.016430192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.0594693504117111e-05,
-      "loss": 1.075,
+      "learning_rate": 1.0945179584120984e-05,
+      "loss": 1.0725,
       "step": 579
     },
     {
-      "epoch": 0.01592925214907583,
+      "epoch": 0.016458569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.0612991765782252e-05,
-      "loss": 1.0893,
+      "learning_rate": 1.0964083175803403e-05,
+      "loss": 1.1512,
       "step": 580
     },
     {
-      "epoch": 0.01595671637691906,
+      "epoch": 0.01648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.0631290027447395e-05,
-      "loss": 1.225,
+      "learning_rate": 1.0982986767485823e-05,
+      "loss": 1.0124,
       "step": 581
     },
     {
-      "epoch": 0.015984180604762296,
+      "epoch": 0.016515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.0649588289112536e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.1001890359168244e-05,
+      "loss": 1.0335,
       "step": 582
     },
     {
-      "epoch": 0.01601164483260553,
+      "epoch": 0.016543700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.0667886550777677e-05,
-      "loss": 1.0982,
+      "learning_rate": 1.1020793950850664e-05,
+      "loss": 1.1806,
       "step": 583
     },
     {
-      "epoch": 0.016039109060448767,
+      "epoch": 0.016572077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.068618481244282e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.1039697542533084e-05,
+      "loss": 1.1337,
       "step": 584
     },
     {
-      "epoch": 0.016066573288292,
+      "epoch": 0.016600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.070448307410796e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.1058601134215502e-05,
+      "loss": 1.1096,
       "step": 585
     },
     {
-      "epoch": 0.016094037516135234,
+      "epoch": 0.016628830874006812,
       "grad_norm": 0.0,
-      "learning_rate": 1.0722781335773103e-05,
-      "loss": 1.0495,
+      "learning_rate": 1.1077504725897921e-05,
+      "loss": 0.9867,
       "step": 586
     },
     {
-      "epoch": 0.01612150174397847,
+      "epoch": 0.016657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.0741079597438244e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.1096408317580341e-05,
+      "loss": 1.1279,
       "step": 587
     },
     {
-      "epoch": 0.0161489659718217,
+      "epoch": 0.016685584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.0759377859103386e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.111531190926276e-05,
+      "loss": 1.1595,
       "step": 588
     },
     {
-      "epoch": 0.016176430199664936,
+      "epoch": 0.016713961407491488,
       "grad_norm": 0.0,
-      "learning_rate": 1.0777676120768527e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.113421550094518e-05,
+      "loss": 1.1272,
       "step": 589
     },
     {
-      "epoch": 0.01620389442750817,
+      "epoch": 0.01674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.079597438243367e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.11531190926276e-05,
+      "loss": 0.9953,
       "step": 590
     },
     {
-      "epoch": 0.016231358655351406,
+      "epoch": 0.016770715096481273,
       "grad_norm": 0.0,
-      "learning_rate": 1.0814272644098811e-05,
-      "loss": 1.1422,
+      "learning_rate": 1.117202268431002e-05,
+      "loss": 1.0692,
       "step": 591
     },
     {
-      "epoch": 0.016258822883194638,
+      "epoch": 0.016799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.0832570905763953e-05,
-      "loss": 1.0802,
+      "learning_rate": 1.119092627599244e-05,
+      "loss": 1.0696,
       "step": 592
     },
     {
-      "epoch": 0.016286287111037873,
+      "epoch": 0.016827468785471054,
       "grad_norm": 0.0,
-      "learning_rate": 1.0850869167429094e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.1209829867674859e-05,
+      "loss": 1.0686,
       "step": 593
     },
     {
-      "epoch": 0.01631375133888111,
+      "epoch": 0.01685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.0869167429094235e-05,
-      "loss": 1.0819,
+      "learning_rate": 1.1228733459357279e-05,
+      "loss": 0.9491,
       "step": 594
     },
     {
-      "epoch": 0.01634121556672434,
+      "epoch": 0.01688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.0887465690759378e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.12476370510397e-05,
+      "loss": 1.0397,
       "step": 595
     },
     {
-      "epoch": 0.016368679794567575,
+      "epoch": 0.016912599318955734,
       "grad_norm": 0.0,
-      "learning_rate": 1.0905763952424519e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.126654064272212e-05,
+      "loss": 1.1132,
       "step": 596
     },
     {
-      "epoch": 0.01639614402241081,
+      "epoch": 0.016940976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.0924062214089661e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.1285444234404538e-05,
+      "loss": 1.0736,
       "step": 597
     },
     {
-      "epoch": 0.016423608250254042,
+      "epoch": 0.016969353007945515,
       "grad_norm": 0.0,
-      "learning_rate": 1.0942360475754806e-05,
-      "loss": 1.1365,
+      "learning_rate": 1.1304347826086957e-05,
+      "loss": 1.1422,
       "step": 598
     },
     {
-      "epoch": 0.016451072478097278,
+      "epoch": 0.01699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.0960658737419947e-05,
-      "loss": 1.032,
+      "learning_rate": 1.1323251417769377e-05,
+      "loss": 1.1134,
       "step": 599
     },
     {
-      "epoch": 0.016478536705940513,
+      "epoch": 0.0170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.097895699908509e-05,
-      "loss": 1.2051,
+      "learning_rate": 1.1342155009451797e-05,
+      "loss": 1.1223,
       "step": 600
     },
     {
-      "epoch": 0.016506000933783748,
+      "epoch": 0.017054483541430195,
       "grad_norm": 0.0,
-      "learning_rate": 1.099725526075023e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.1361058601134216e-05,
+      "loss": 1.0376,
       "step": 601
     },
     {
-      "epoch": 0.01653346516162698,
+      "epoch": 0.017082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.1015553522415371e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.1379962192816636e-05,
+      "loss": 1.114,
       "step": 602
     },
     {
-      "epoch": 0.016560929389470215,
+      "epoch": 0.017111237230419976,
       "grad_norm": 0.0,
-      "learning_rate": 1.1033851784080514e-05,
-      "loss": 1.122,
+      "learning_rate": 1.1398865784499056e-05,
+      "loss": 1.0723,
       "step": 603
     },
     {
-      "epoch": 0.01658839361731345,
+      "epoch": 0.01713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.1052150045745655e-05,
-      "loss": 1.157,
+      "learning_rate": 1.1417769376181475e-05,
+      "loss": 1.1404,
       "step": 604
     },
     {
-      "epoch": 0.016615857845156682,
+      "epoch": 0.01716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.1070448307410797e-05,
-      "loss": 1.1319,
+      "learning_rate": 1.1436672967863895e-05,
+      "loss": 0.9975,
       "step": 605
     },
     {
-      "epoch": 0.016643322072999917,
+      "epoch": 0.017196367763904655,
       "grad_norm": 0.0,
-      "learning_rate": 1.1088746569075938e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.1455576559546314e-05,
+      "loss": 1.0676,
       "step": 606
     },
     {
-      "epoch": 0.016670786300843152,
+      "epoch": 0.017224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.1107044830741081e-05,
-      "loss": 1.1201,
+      "learning_rate": 1.1474480151228736e-05,
+      "loss": 1.1318,
       "step": 607
     },
     {
-      "epoch": 0.016698250528686388,
+      "epoch": 0.017253121452894437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1125343092406222e-05,
-      "loss": 1.2271,
+      "learning_rate": 1.1493383742911156e-05,
+      "loss": 1.0956,
       "step": 608
     },
     {
-      "epoch": 0.01672571475652962,
+      "epoch": 0.01728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.1143641354071364e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.1512287334593572e-05,
+      "loss": 1.1684,
       "step": 609
     },
     {
-      "epoch": 0.016753178984372855,
+      "epoch": 0.017309875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.1161939615736505e-05,
-      "loss": 1.069,
+      "learning_rate": 1.1531190926275993e-05,
+      "loss": 1.1311,
       "step": 610
     },
     {
-      "epoch": 0.01678064321221609,
+      "epoch": 0.017338251986379116,
       "grad_norm": 0.0,
-      "learning_rate": 1.1180237877401648e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.1550094517958413e-05,
+      "loss": 1.0422,
       "step": 611
     },
     {
-      "epoch": 0.01680810744005932,
+      "epoch": 0.017366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.1198536139066789e-05,
-      "loss": 1.177,
+      "learning_rate": 1.1568998109640832e-05,
+      "loss": 1.1046,
       "step": 612
     },
     {
-      "epoch": 0.016835571667902557,
+      "epoch": 0.017395005675368898,
       "grad_norm": 0.0,
-      "learning_rate": 1.1216834400731932e-05,
-      "loss": 1.1188,
+      "learning_rate": 1.1587901701323252e-05,
+      "loss": 1.007,
       "step": 613
     },
     {
-      "epoch": 0.016863035895745792,
+      "epoch": 0.017423382519863792,
       "grad_norm": 0.0,
-      "learning_rate": 1.1235132662397072e-05,
-      "loss": 1.2456,
+      "learning_rate": 1.1606805293005672e-05,
+      "loss": 1.0799,
       "step": 614
     },
     {
-      "epoch": 0.016890500123589024,
+      "epoch": 0.017451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.1253430924062213e-05,
-      "loss": 1.1498,
+      "learning_rate": 1.1625708884688091e-05,
+      "loss": 1.0965,
       "step": 615
     },
     {
-      "epoch": 0.01691796435143226,
+      "epoch": 0.017480136208853577,
       "grad_norm": 0.0,
-      "learning_rate": 1.1271729185727356e-05,
-      "loss": 1.1175,
+      "learning_rate": 1.1644612476370511e-05,
+      "loss": 1.0836,
       "step": 616
     },
     {
-      "epoch": 0.016945428579275494,
+      "epoch": 0.017508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.1290027447392497e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.166351606805293e-05,
+      "loss": 0.9422,
       "step": 617
     },
     {
-      "epoch": 0.01697289280711873,
+      "epoch": 0.01753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1308325709057641e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.168241965973535e-05,
+      "loss": 1.0988,
       "step": 618
     },
     {
-      "epoch": 0.01700035703496196,
+      "epoch": 0.017565266742338253,
       "grad_norm": 0.0,
-      "learning_rate": 1.1326623970722784e-05,
-      "loss": 1.2339,
+      "learning_rate": 1.1701323251417772e-05,
+      "loss": 1.127,
       "step": 619
     },
     {
-      "epoch": 0.017027821262805196,
+      "epoch": 0.017593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.1344922232387925e-05,
-      "loss": 1.1278,
+      "learning_rate": 1.1720226843100191e-05,
+      "loss": 1.0518,
       "step": 620
     },
     {
-      "epoch": 0.01705528549064843,
+      "epoch": 0.017622020431328038,
       "grad_norm": 0.0,
-      "learning_rate": 1.1363220494053067e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.1739130434782611e-05,
+      "loss": 1.0561,
       "step": 621
     },
     {
-      "epoch": 0.017082749718491663,
+      "epoch": 0.01765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.1381518755718208e-05,
-      "loss": 1.0669,
+      "learning_rate": 1.1758034026465029e-05,
+      "loss": 1.0829,
       "step": 622
     },
     {
-      "epoch": 0.0171102139463349,
+      "epoch": 0.01767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.139981701738335e-05,
-      "loss": 1.1265,
+      "learning_rate": 1.1776937618147449e-05,
+      "loss": 1.0552,
       "step": 623
     },
     {
-      "epoch": 0.017137678174178134,
+      "epoch": 0.017707150964812714,
       "grad_norm": 0.0,
-      "learning_rate": 1.1418115279048492e-05,
-      "loss": 1.016,
+      "learning_rate": 1.1795841209829868e-05,
+      "loss": 1.1144,
       "step": 624
     },
     {
-      "epoch": 0.01716514240202137,
+      "epoch": 0.017735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1436413540713633e-05,
-      "loss": 1.2626,
+      "learning_rate": 1.1814744801512288e-05,
+      "loss": 1.1992,
       "step": 625
     },
     {
-      "epoch": 0.0171926066298646,
+      "epoch": 0.0177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.1454711802378775e-05,
-      "loss": 1.0905,
+      "learning_rate": 1.1833648393194708e-05,
+      "loss": 1.0677,
       "step": 626
     },
     {
-      "epoch": 0.017220070857707836,
+      "epoch": 0.01779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.1473010064043916e-05,
-      "loss": 1.1247,
+      "learning_rate": 1.1852551984877127e-05,
+      "loss": 1.1686,
       "step": 627
     },
     {
-      "epoch": 0.01724753508555107,
+      "epoch": 0.01782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.1491308325709059e-05,
-      "loss": 1.1783,
+      "learning_rate": 1.1871455576559547e-05,
+      "loss": 1.049,
       "step": 628
     },
     {
-      "epoch": 0.017274999313394303,
+      "epoch": 0.017849035187287175,
       "grad_norm": 0.0,
-      "learning_rate": 1.15096065873742e-05,
-      "loss": 1.1636,
+      "learning_rate": 1.1890359168241967e-05,
+      "loss": 1.1206,
       "step": 629
     },
     {
-      "epoch": 0.017302463541237538,
+      "epoch": 0.017877412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.1527904849039342e-05,
-      "loss": 1.1716,
+      "learning_rate": 1.1909262759924386e-05,
+      "loss": 1.023,
       "step": 630
     },
     {
-      "epoch": 0.017329927769080773,
+      "epoch": 0.01790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.1546203110704483e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.1928166351606808e-05,
+      "loss": 0.9014,
       "step": 631
     },
     {
-      "epoch": 0.017357391996924005,
+      "epoch": 0.01793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.1564501372369626e-05,
-      "loss": 1.1255,
+      "learning_rate": 1.1947069943289227e-05,
+      "loss": 1.0382,
       "step": 632
     },
     {
-      "epoch": 0.01738485622476724,
+      "epoch": 0.01796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.1582799634034767e-05,
-      "loss": 1.1931,
+      "learning_rate": 1.1965973534971647e-05,
+      "loss": 1.0891,
       "step": 633
     },
     {
-      "epoch": 0.017412320452610475,
+      "epoch": 0.017990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.160109789569991e-05,
-      "loss": 1.1804,
+      "learning_rate": 1.1984877126654065e-05,
+      "loss": 1.0846,
       "step": 634
     },
     {
-      "epoch": 0.01743978468045371,
+      "epoch": 0.018019296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.161939615736505e-05,
-      "loss": 1.232,
+      "learning_rate": 1.2003780718336485e-05,
+      "loss": 1.05,
       "step": 635
     },
     {
-      "epoch": 0.017467248908296942,
+      "epoch": 0.01804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.1637694419030191e-05,
-      "loss": 1.1425,
+      "learning_rate": 1.2022684310018904e-05,
+      "loss": 1.0553,
       "step": 636
     },
     {
-      "epoch": 0.017494713136140178,
+      "epoch": 0.01807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.1655992680695334e-05,
-      "loss": 1.0942,
+      "learning_rate": 1.2041587901701324e-05,
+      "loss": 1.0789,
       "step": 637
     },
     {
-      "epoch": 0.017522177363983413,
+      "epoch": 0.018104426787741202,
       "grad_norm": 0.0,
-      "learning_rate": 1.1674290942360478e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.2060491493383744e-05,
+      "loss": 1.1219,
       "step": 638
     },
     {
-      "epoch": 0.017549641591826644,
+      "epoch": 0.018132803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.169258920402562e-05,
-      "loss": 1.1532,
+      "learning_rate": 1.2079395085066163e-05,
+      "loss": 1.1429,
       "step": 639
     },
     {
-      "epoch": 0.01757710581966988,
+      "epoch": 0.018161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.1710887465690762e-05,
-      "loss": 1.1199,
+      "learning_rate": 1.2098298676748583e-05,
+      "loss": 1.2378,
       "step": 640
     },
     {
-      "epoch": 0.017604570047513115,
+      "epoch": 0.01818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.1729185727355903e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.2117202268431003e-05,
+      "loss": 1.1236,
       "step": 641
     },
     {
-      "epoch": 0.017632034275356347,
+      "epoch": 0.018217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.1747483989021045e-05,
-      "loss": 1.1314,
+      "learning_rate": 1.2136105860113422e-05,
+      "loss": 1.2139,
       "step": 642
     },
     {
-      "epoch": 0.017659498503199582,
+      "epoch": 0.018246311010215663,
       "grad_norm": 0.0,
-      "learning_rate": 1.1765782250686186e-05,
-      "loss": 1.1037,
+      "learning_rate": 1.2155009451795844e-05,
+      "loss": 1.191,
       "step": 643
     },
     {
-      "epoch": 0.017686962731042817,
+      "epoch": 0.018274687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.1784080512351327e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.2173913043478263e-05,
+      "loss": 1.1034,
       "step": 644
     },
     {
-      "epoch": 0.017714426958886052,
+      "epoch": 0.018303064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.180237877401647e-05,
-      "loss": 1.1409,
+      "learning_rate": 1.2192816635160683e-05,
+      "loss": 1.0853,
       "step": 645
     },
     {
-      "epoch": 0.017741891186729284,
+      "epoch": 0.018331441543700342,
       "grad_norm": 0.0,
-      "learning_rate": 1.182067703568161e-05,
-      "loss": 1.1707,
+      "learning_rate": 1.22117202268431e-05,
+      "loss": 1.124,
       "step": 646
     },
     {
-      "epoch": 0.01776935541457252,
+      "epoch": 0.018359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.1838975297346753e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.223062381852552e-05,
+      "loss": 1.0785,
       "step": 647
     },
     {
-      "epoch": 0.017796819642415754,
+      "epoch": 0.018388195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.1857273559011894e-05,
-      "loss": 1.1811,
+      "learning_rate": 1.224952741020794e-05,
+      "loss": 1.0229,
       "step": 648
     },
     {
-      "epoch": 0.017824283870258986,
+      "epoch": 0.018416572077185018,
       "grad_norm": 0.0,
-      "learning_rate": 1.1875571820677037e-05,
-      "loss": 1.1361,
+      "learning_rate": 1.226843100189036e-05,
+      "loss": 1.0643,
       "step": 649
     },
     {
-      "epoch": 0.01785174809810222,
+      "epoch": 0.01844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.1893870082342178e-05,
-      "loss": 1.131,
+      "learning_rate": 1.228733459357278e-05,
+      "loss": 1.2037,
       "step": 650
     },
     {
-      "epoch": 0.017879212325945457,
+      "epoch": 0.018473325766174803,
       "grad_norm": 0.0,
-      "learning_rate": 1.191216834400732e-05,
-      "loss": 1.0926,
+      "learning_rate": 1.2306238185255199e-05,
+      "loss": 0.9857,
       "step": 651
     },
     {
-      "epoch": 0.017906676553788692,
+      "epoch": 0.018501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.1930466605672461e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.2325141776937619e-05,
+      "loss": 0.9998,
       "step": 652
     },
     {
-      "epoch": 0.017934140781631924,
+      "epoch": 0.018530079455164584,
       "grad_norm": 0.0,
-      "learning_rate": 1.1948764867337604e-05,
-      "loss": 1.063,
+      "learning_rate": 1.2344045368620038e-05,
+      "loss": 1.1326,
       "step": 653
     },
     {
-      "epoch": 0.01796160500947516,
+      "epoch": 0.01855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.1967063129002745e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.2362948960302458e-05,
+      "loss": 1.0415,
       "step": 654
     },
     {
-      "epoch": 0.017989069237318394,
+      "epoch": 0.01858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1985361390667888e-05,
-      "loss": 1.1188,
+      "learning_rate": 1.238185255198488e-05,
+      "loss": 1.176,
       "step": 655
     },
     {
-      "epoch": 0.018016533465161626,
+      "epoch": 0.018615209988649264,
       "grad_norm": 0.0,
-      "learning_rate": 1.2003659652333028e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.2400756143667299e-05,
+      "loss": 1.1318,
       "step": 656
     },
     {
-      "epoch": 0.01804399769300486,
+      "epoch": 0.018643586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.202195791399817e-05,
-      "loss": 1.2876,
+      "learning_rate": 1.2419659735349719e-05,
+      "loss": 1.1105,
       "step": 657
     },
     {
-      "epoch": 0.018071461920848096,
+      "epoch": 0.018671963677639045,
       "grad_norm": 0.0,
-      "learning_rate": 1.2040256175663314e-05,
-      "loss": 1.0763,
+      "learning_rate": 1.2438563327032138e-05,
+      "loss": 1.1023,
       "step": 658
     },
     {
-      "epoch": 0.018098926148691328,
+      "epoch": 0.01870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.2058554437328456e-05,
-      "loss": 1.1041,
+      "learning_rate": 1.2457466918714556e-05,
+      "loss": 1.2043,
       "step": 659
     },
     {
-      "epoch": 0.018126390376534563,
+      "epoch": 0.01872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.2076852698993597e-05,
-      "loss": 1.1812,
+      "learning_rate": 1.2476370510396976e-05,
+      "loss": 1.1423,
       "step": 660
     },
     {
-      "epoch": 0.0181538546043778,
+      "epoch": 0.018757094211123725,
       "grad_norm": 0.0,
-      "learning_rate": 1.209515096065874e-05,
-      "loss": 1.1761,
+      "learning_rate": 1.2495274102079396e-05,
+      "loss": 1.1603,
       "step": 661
     },
     {
-      "epoch": 0.018181318832221034,
+      "epoch": 0.018785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.211344922232388e-05,
-      "loss": 1.1653,
+      "learning_rate": 1.2514177693761815e-05,
+      "loss": 1.1532,
       "step": 662
     },
     {
-      "epoch": 0.018208783060064265,
+      "epoch": 0.018813847900113506,
       "grad_norm": 0.0,
-      "learning_rate": 1.2131747483989023e-05,
-      "loss": 1.1499,
+      "learning_rate": 1.2533081285444235e-05,
+      "loss": 1.083,
       "step": 663
     },
     {
-      "epoch": 0.0182362472879075,
+      "epoch": 0.0188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2150045745654164e-05,
-      "loss": 1.1956,
+      "learning_rate": 1.2551984877126655e-05,
+      "loss": 1.116,
       "step": 664
     },
     {
-      "epoch": 0.018263711515750736,
+      "epoch": 0.01887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.2168344007319305e-05,
-      "loss": 1.1268,
+      "learning_rate": 1.2570888468809074e-05,
+      "loss": 1.0292,
       "step": 665
     },
     {
-      "epoch": 0.018291175743593967,
+      "epoch": 0.018898978433598185,
       "grad_norm": 0.0,
-      "learning_rate": 1.2186642268984448e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.2589792060491494e-05,
+      "loss": 1.0607,
       "step": 666
     },
     {
-      "epoch": 0.018318639971437203,
+      "epoch": 0.018927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.2204940530649589e-05,
-      "loss": 1.1267,
+      "learning_rate": 1.2608695652173915e-05,
+      "loss": 1.0036,
       "step": 667
     },
     {
-      "epoch": 0.018346104199280438,
+      "epoch": 0.018955732122587967,
       "grad_norm": 0.0,
-      "learning_rate": 1.2223238792314731e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.2627599243856335e-05,
+      "loss": 1.1885,
       "step": 668
     },
     {
-      "epoch": 0.018373568427123673,
+      "epoch": 0.01898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.2241537053979872e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.2646502835538755e-05,
+      "loss": 1.215,
       "step": 669
     },
     {
-      "epoch": 0.018401032654966905,
+      "epoch": 0.019012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.2259835315645015e-05,
-      "loss": 1.0554,
+      "learning_rate": 1.2665406427221174e-05,
+      "loss": 1.1202,
       "step": 670
     },
     {
-      "epoch": 0.01842849688281014,
+      "epoch": 0.019040862656072646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2278133577310156e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.2684310018903592e-05,
+      "loss": 1.1649,
       "step": 671
     },
     {
-      "epoch": 0.018455961110653375,
+      "epoch": 0.019069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.2296431838975299e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.2703213610586012e-05,
+      "loss": 0.9935,
       "step": 672
     },
     {
-      "epoch": 0.018483425338496607,
+      "epoch": 0.019097616345062428,
       "grad_norm": 0.0,
-      "learning_rate": 1.231473010064044e-05,
-      "loss": 1.1957,
+      "learning_rate": 1.2722117202268432e-05,
+      "loss": 1.1705,
       "step": 673
     },
     {
-      "epoch": 0.018510889566339842,
+      "epoch": 0.019125993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.2333028362305582e-05,
-      "loss": 1.17,
+      "learning_rate": 1.2741020793950851e-05,
+      "loss": 1.0602,
       "step": 674
     },
     {
-      "epoch": 0.018538353794183077,
+      "epoch": 0.019154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.2351326623970723e-05,
-      "loss": 1.2136,
+      "learning_rate": 1.2759924385633271e-05,
+      "loss": 0.9731,
       "step": 675
     },
     {
-      "epoch": 0.01856581802202631,
+      "epoch": 0.019182746878547107,
       "grad_norm": 0.0,
-      "learning_rate": 1.2369624885635864e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.277882797731569e-05,
+      "loss": 1.0178,
       "step": 676
     },
     {
-      "epoch": 0.018593282249869544,
+      "epoch": 0.019211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.2387923147301007e-05,
-      "loss": 1.1052,
+      "learning_rate": 1.279773156899811e-05,
+      "loss": 1.0421,
       "step": 677
     },
     {
-      "epoch": 0.01862074647771278,
+      "epoch": 0.01923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.2406221408966147e-05,
-      "loss": 1.1858,
+      "learning_rate": 1.281663516068053e-05,
+      "loss": 1.1722,
       "step": 678
     },
     {
-      "epoch": 0.018648210705556015,
+      "epoch": 0.019267877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.2424519670631292e-05,
-      "loss": 1.0313,
+      "learning_rate": 1.2835538752362951e-05,
+      "loss": 1.0453,
       "step": 679
     },
     {
-      "epoch": 0.018675674933399247,
+      "epoch": 0.019296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.2442817932296434e-05,
-      "loss": 1.1434,
+      "learning_rate": 1.2854442344045371e-05,
+      "loss": 1.0212,
       "step": 680
     },
     {
-      "epoch": 0.018703139161242482,
+      "epoch": 0.019324631101021568,
       "grad_norm": 0.0,
-      "learning_rate": 1.2461116193961575e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.287334593572779e-05,
+      "loss": 1.072,
       "step": 681
     },
     {
-      "epoch": 0.018730603389085717,
+      "epoch": 0.01935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2479414455626718e-05,
-      "loss": 1.1349,
+      "learning_rate": 1.289224952741021e-05,
+      "loss": 1.0455,
       "step": 682
     },
     {
-      "epoch": 0.01875806761692895,
+      "epoch": 0.01938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.2497712717291859e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.2911153119092628e-05,
+      "loss": 1.1078,
       "step": 683
     },
     {
-      "epoch": 0.018785531844772184,
+      "epoch": 0.019409761634506244,
       "grad_norm": 0.0,
-      "learning_rate": 1.2516010978957e-05,
-      "loss": 1.0849,
+      "learning_rate": 1.2930056710775048e-05,
+      "loss": 0.9597,
       "step": 684
     },
     {
-      "epoch": 0.01881299607261542,
+      "epoch": 0.019438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.2534309240622142e-05,
-      "loss": 1.123,
+      "learning_rate": 1.2948960302457467e-05,
+      "loss": 0.9636,
       "step": 685
     },
     {
-      "epoch": 0.018840460300458654,
+      "epoch": 0.01946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.2552607502287283e-05,
-      "loss": 1.1606,
+      "learning_rate": 1.2967863894139887e-05,
+      "loss": 1.0811,
       "step": 686
     },
     {
-      "epoch": 0.018867924528301886,
+      "epoch": 0.01949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.2570905763952426e-05,
-      "loss": 1.152,
+      "learning_rate": 1.2986767485822307e-05,
+      "loss": 1.0591,
       "step": 687
     },
     {
-      "epoch": 0.01889538875614512,
+      "epoch": 0.01952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.2589204025617567e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.3005671077504726e-05,
+      "loss": 1.0716,
       "step": 688
     },
     {
-      "epoch": 0.018922852983988356,
+      "epoch": 0.019551645856980705,
       "grad_norm": 0.0,
-      "learning_rate": 1.260750228728271e-05,
-      "loss": 1.1698,
+      "learning_rate": 1.3024574669187146e-05,
+      "loss": 0.9895,
       "step": 689
     },
     {
-      "epoch": 0.018950317211831588,
+      "epoch": 0.019580022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.262580054894785e-05,
-      "loss": 1.0955,
+      "learning_rate": 1.3043478260869566e-05,
+      "loss": 1.0777,
       "step": 690
     },
     {
-      "epoch": 0.018977781439674823,
+      "epoch": 0.01960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.2644098810612993e-05,
-      "loss": 1.1964,
+      "learning_rate": 1.3062381852551987e-05,
+      "loss": 1.1207,
       "step": 691
     },
     {
-      "epoch": 0.01900524566751806,
+      "epoch": 0.01963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.2662397072278134e-05,
-      "loss": 1.0596,
+      "learning_rate": 1.3081285444234407e-05,
+      "loss": 1.0518,
       "step": 692
     },
     {
-      "epoch": 0.01903270989536129,
+      "epoch": 0.01966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.2680695333943277e-05,
-      "loss": 1.215,
+      "learning_rate": 1.3100189035916826e-05,
+      "loss": 1.1628,
       "step": 693
     },
     {
-      "epoch": 0.019060174123204526,
+      "epoch": 0.019693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.2698993595608417e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.3119092627599246e-05,
+      "loss": 0.9706,
       "step": 694
     },
     {
-      "epoch": 0.01908763835104776,
+      "epoch": 0.019721906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.271729185727356e-05,
-      "loss": 1.1125,
+      "learning_rate": 1.3137996219281666e-05,
+      "loss": 1.1305,
       "step": 695
     },
     {
-      "epoch": 0.019115102578890996,
+      "epoch": 0.01975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.2735590118938701e-05,
-      "loss": 1.1608,
+      "learning_rate": 1.3156899810964084e-05,
+      "loss": 1.1601,
       "step": 696
     },
     {
-      "epoch": 0.019142566806734228,
+      "epoch": 0.01977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.2753888380603842e-05,
-      "loss": 1.1635,
+      "learning_rate": 1.3175803402646503e-05,
+      "loss": 1.0982,
       "step": 697
     },
     {
-      "epoch": 0.019170031034577463,
+      "epoch": 0.019807037457434732,
       "grad_norm": 0.0,
-      "learning_rate": 1.2772186642268985e-05,
-      "loss": 1.2012,
+      "learning_rate": 1.3194706994328923e-05,
+      "loss": 1.1296,
       "step": 698
     },
     {
-      "epoch": 0.019197495262420698,
+      "epoch": 0.019835414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.2790484903934129e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.3213610586011343e-05,
+      "loss": 1.0998,
       "step": 699
     },
     {
-      "epoch": 0.01922495949026393,
+      "epoch": 0.019863791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.280878316559927e-05,
-      "loss": 1.0368,
+      "learning_rate": 1.3232514177693762e-05,
+      "loss": 1.0375,
       "step": 700
     },
     {
-      "epoch": 0.019252423718107165,
+      "epoch": 0.01989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.2827081427264412e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.3251417769376182e-05,
+      "loss": 1.0855,
       "step": 701
     },
     {
-      "epoch": 0.0192798879459504,
+      "epoch": 0.019920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.2845379688929553e-05,
-      "loss": 1.222,
+      "learning_rate": 1.3270321361058602e-05,
+      "loss": 1.0593,
       "step": 702
     },
     {
-      "epoch": 0.019307352173793632,
+      "epoch": 0.019948921679909193,
       "grad_norm": 0.0,
-      "learning_rate": 1.2863677950594696e-05,
-      "loss": 1.1362,
+      "learning_rate": 1.3289224952741023e-05,
+      "loss": 1.1246,
       "step": 703
     },
     {
-      "epoch": 0.019334816401636867,
+      "epoch": 0.019977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.2881976212259837e-05,
-      "loss": 1.1844,
+      "learning_rate": 1.3308128544423443e-05,
+      "loss": 1.1074,
       "step": 704
     },
     {
-      "epoch": 0.019362280629480103,
+      "epoch": 0.020005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.2900274473924978e-05,
-      "loss": 1.1476,
+      "learning_rate": 1.3327032136105862e-05,
+      "loss": 1.1425,
       "step": 705
     },
     {
-      "epoch": 0.019389744857323338,
+      "epoch": 0.020034052213393872,
       "grad_norm": 0.0,
-      "learning_rate": 1.291857273559012e-05,
-      "loss": 1.1225,
+      "learning_rate": 1.3345935727788282e-05,
+      "loss": 1.0214,
       "step": 706
     },
     {
-      "epoch": 0.01941720908516657,
+      "epoch": 0.020062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.2936870997255261e-05,
-      "loss": 1.1557,
+      "learning_rate": 1.3364839319470702e-05,
+      "loss": 0.9668,
       "step": 707
     },
     {
-      "epoch": 0.019444673313009805,
+      "epoch": 0.020090805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.2955169258920404e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.338374291115312e-05,
+      "loss": 1.0623,
       "step": 708
     },
     {
-      "epoch": 0.01947213754085304,
+      "epoch": 0.020119182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 1.2973467520585545e-05,
-      "loss": 1.2031,
+      "learning_rate": 1.340264650283554e-05,
+      "loss": 1.1541,
       "step": 709
     },
     {
-      "epoch": 0.01949960176869627,
+      "epoch": 0.02014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.2991765782250687e-05,
-      "loss": 1.1807,
+      "learning_rate": 1.3421550094517959e-05,
+      "loss": 1.0668,
       "step": 710
     },
     {
-      "epoch": 0.019527065996539507,
+      "epoch": 0.020175936435868333,
       "grad_norm": 0.0,
-      "learning_rate": 1.3010064043915828e-05,
-      "loss": 1.1949,
+      "learning_rate": 1.3440453686200379e-05,
+      "loss": 1.133,
       "step": 711
     },
     {
-      "epoch": 0.019554530224382742,
+      "epoch": 0.020204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.3028362305580971e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.3459357277882798e-05,
+      "loss": 1.0334,
       "step": 712
     },
     {
-      "epoch": 0.019581994452225977,
+      "epoch": 0.020232690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.3046660567246112e-05,
-      "loss": 1.1528,
+      "learning_rate": 1.3478260869565218e-05,
+      "loss": 1.1704,
       "step": 713
     },
     {
-      "epoch": 0.01960945868006921,
+      "epoch": 0.02026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.3064958828911255e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.3497164461247638e-05,
+      "loss": 1.0205,
       "step": 714
     },
     {
-      "epoch": 0.019636922907912444,
+      "epoch": 0.0202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3083257090576395e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.3516068052930059e-05,
+      "loss": 1.1558,
       "step": 715
     },
     {
-      "epoch": 0.01966438713575568,
+      "epoch": 0.020317820658342794,
       "grad_norm": 0.0,
-      "learning_rate": 1.3101555352241538e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.3534971644612479e-05,
+      "loss": 1.1239,
       "step": 716
     },
     {
-      "epoch": 0.01969185136359891,
+      "epoch": 0.020346197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.3119853613906679e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.3553875236294898e-05,
+      "loss": 1.0981,
       "step": 717
     },
     {
-      "epoch": 0.019719315591442146,
+      "epoch": 0.020374574347332575,
       "grad_norm": 0.0,
-      "learning_rate": 1.313815187557182e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.3572778827977318e-05,
+      "loss": 1.1115,
       "step": 718
     },
     {
-      "epoch": 0.01974677981928538,
+      "epoch": 0.02040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.3156450137236964e-05,
-      "loss": 1.2114,
+      "learning_rate": 1.3591682419659738e-05,
+      "loss": 1.102,
       "step": 719
     },
     {
-      "epoch": 0.019774244047128613,
+      "epoch": 0.02043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.3174748398902107e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.3610586011342156e-05,
+      "loss": 1.0573,
       "step": 720
     },
     {
-      "epoch": 0.01980170827497185,
+      "epoch": 0.020459704880817255,
       "grad_norm": 0.0,
-      "learning_rate": 1.3193046660567248e-05,
-      "loss": 1.0706,
+      "learning_rate": 1.3629489603024575e-05,
+      "loss": 1.1352,
       "step": 721
     },
     {
-      "epoch": 0.019829172502815084,
+      "epoch": 0.020488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.321134492223239e-05,
-      "loss": 1.1349,
+      "learning_rate": 1.3648393194706995e-05,
+      "loss": 1.0493,
       "step": 722
     },
     {
-      "epoch": 0.01985663673065832,
+      "epoch": 0.020516458569807036,
       "grad_norm": 0.0,
-      "learning_rate": 1.3229643183897531e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.3667296786389414e-05,
+      "loss": 1.1145,
       "step": 723
     },
     {
-      "epoch": 0.01988410095850155,
+      "epoch": 0.02054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.3247941445562674e-05,
-      "loss": 1.0858,
+      "learning_rate": 1.3686200378071834e-05,
+      "loss": 0.9449,
       "step": 724
     },
     {
-      "epoch": 0.019911565186344786,
+      "epoch": 0.02057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.3266239707227815e-05,
-      "loss": 1.0993,
+      "learning_rate": 1.3705103969754254e-05,
+      "loss": 1.0547,
       "step": 725
     },
     {
-      "epoch": 0.01993902941418802,
+      "epoch": 0.020601589103291715,
       "grad_norm": 0.0,
-      "learning_rate": 1.3284537968892956e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.3724007561436673e-05,
+      "loss": 1.1419,
       "step": 726
     },
     {
-      "epoch": 0.019966493642031253,
+      "epoch": 0.020629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.3302836230558098e-05,
-      "loss": 1.1674,
+      "learning_rate": 1.3742911153119093e-05,
+      "loss": 1.104,
       "step": 727
     },
     {
-      "epoch": 0.019993957869874488,
+      "epoch": 0.020658342792281497,
       "grad_norm": 0.0,
-      "learning_rate": 1.332113449222324e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.3761814744801514e-05,
+      "loss": 1.0639,
       "step": 728
     },
     {
-      "epoch": 0.020021422097717723,
+      "epoch": 0.02068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.3339432753888382e-05,
-      "loss": 1.1778,
+      "learning_rate": 1.3780718336483934e-05,
+      "loss": 1.1784,
       "step": 729
     },
     {
-      "epoch": 0.02004888632556096,
+      "epoch": 0.020715096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 1.3357731015553523e-05,
-      "loss": 1.169,
+      "learning_rate": 1.3799621928166354e-05,
+      "loss": 1.1049,
       "step": 730
     },
     {
-      "epoch": 0.02007635055340419,
+      "epoch": 0.020743473325766176,
       "grad_norm": 0.0,
-      "learning_rate": 1.3376029277218666e-05,
-      "loss": 1.1769,
+      "learning_rate": 1.3818525519848773e-05,
+      "loss": 1.1318,
       "step": 731
     },
     {
-      "epoch": 0.020103814781247425,
+      "epoch": 0.020771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.3394327538883806e-05,
-      "loss": 1.0728,
+      "learning_rate": 1.3837429111531191e-05,
+      "loss": 1.0363,
       "step": 732
     },
     {
-      "epoch": 0.02013127900909066,
+      "epoch": 0.020800227014755958,
       "grad_norm": 0.0,
-      "learning_rate": 1.3412625800548949e-05,
-      "loss": 1.1609,
+      "learning_rate": 1.3856332703213611e-05,
+      "loss": 1.0326,
       "step": 733
     },
     {
-      "epoch": 0.020158743236933892,
+      "epoch": 0.020828603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.343092406221409e-05,
-      "loss": 1.0721,
+      "learning_rate": 1.387523629489603e-05,
+      "loss": 1.2156,
       "step": 734
     },
     {
-      "epoch": 0.020186207464777128,
+      "epoch": 0.020856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.3449222323879233e-05,
-      "loss": 1.1615,
+      "learning_rate": 1.389413988657845e-05,
+      "loss": 1.0814,
       "step": 735
     },
     {
-      "epoch": 0.020213671692620363,
+      "epoch": 0.020885357548240637,
       "grad_norm": 0.0,
-      "learning_rate": 1.3467520585544374e-05,
-      "loss": 1.1064,
+      "learning_rate": 1.391304347826087e-05,
+      "loss": 1.0659,
       "step": 736
     },
     {
-      "epoch": 0.020241135920463595,
+      "epoch": 0.020913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.3485818847209516e-05,
-      "loss": 1.1558,
+      "learning_rate": 1.393194706994329e-05,
+      "loss": 0.9708,
       "step": 737
     },
     {
-      "epoch": 0.02026860014830683,
+      "epoch": 0.02094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.3504117108874657e-05,
-      "loss": 1.0939,
+      "learning_rate": 1.395085066162571e-05,
+      "loss": 1.0581,
       "step": 738
     },
     {
-      "epoch": 0.020296064376150065,
+      "epoch": 0.020970488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 1.3522415370539801e-05,
-      "loss": 1.1903,
+      "learning_rate": 1.3969754253308129e-05,
+      "loss": 1.0936,
       "step": 739
     },
     {
-      "epoch": 0.0203235286039933,
+      "epoch": 0.020998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.3540713632204942e-05,
-      "loss": 1.1313,
+      "learning_rate": 1.398865784499055e-05,
+      "loss": 1.0244,
       "step": 740
     },
     {
-      "epoch": 0.020350992831836532,
+      "epoch": 0.021027241770715098,
       "grad_norm": 0.0,
-      "learning_rate": 1.3559011893870085e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.400756143667297e-05,
+      "loss": 1.1244,
       "step": 741
     },
     {
-      "epoch": 0.020378457059679767,
+      "epoch": 0.02105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.3577310155535226e-05,
-      "loss": 1.1392,
+      "learning_rate": 1.402646502835539e-05,
+      "loss": 1.1279,
       "step": 742
     },
     {
-      "epoch": 0.020405921287523002,
+      "epoch": 0.02108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.3595608417200368e-05,
-      "loss": 1.2002,
+      "learning_rate": 1.404536862003781e-05,
+      "loss": 1.0732,
       "step": 743
     },
     {
-      "epoch": 0.020433385515366234,
+      "epoch": 0.021112372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.361390667886551e-05,
-      "loss": 1.0886,
+      "learning_rate": 1.4064272211720229e-05,
+      "loss": 1.1562,
       "step": 744
     },
     {
-      "epoch": 0.02046084974320947,
+      "epoch": 0.021140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.363220494053065e-05,
-      "loss": 1.1925,
+      "learning_rate": 1.4083175803402647e-05,
+      "loss": 1.0708,
       "step": 745
     },
     {
-      "epoch": 0.020488313971052705,
+      "epoch": 0.02116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.3650503202195793e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.4102079395085067e-05,
+      "loss": 1.0504,
       "step": 746
     },
     {
-      "epoch": 0.020515778198895936,
+      "epoch": 0.02119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.3668801463860934e-05,
-      "loss": 1.1697,
+      "learning_rate": 1.4120982986767486e-05,
+      "loss": 1.059,
       "step": 747
     },
     {
-      "epoch": 0.02054324242673917,
+      "epoch": 0.02122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.3687099725526076e-05,
-      "loss": 1.108,
+      "learning_rate": 1.4139886578449906e-05,
+      "loss": 1.0971,
       "step": 748
     },
     {
-      "epoch": 0.020570706654582407,
+      "epoch": 0.021254256526674235,
       "grad_norm": 0.0,
-      "learning_rate": 1.3705397987191217e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.4158790170132326e-05,
+      "loss": 1.094,
       "step": 749
     },
     {
-      "epoch": 0.020598170882425642,
+      "epoch": 0.021282633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.372369624885636e-05,
-      "loss": 1.0597,
+      "learning_rate": 1.4177693761814745e-05,
+      "loss": 1.0623,
       "step": 750
     },
     {
-      "epoch": 0.020625635110268874,
+      "epoch": 0.02131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.3741994510521501e-05,
-      "loss": 1.1559,
+      "learning_rate": 1.4196597353497165e-05,
+      "loss": 1.1278,
       "step": 751
     },
     {
-      "epoch": 0.02065309933811211,
+      "epoch": 0.02133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.3760292772186644e-05,
-      "loss": 1.0769,
+      "learning_rate": 1.4215500945179586e-05,
+      "loss": 0.9839,
       "step": 752
     },
     {
-      "epoch": 0.020680563565955344,
+      "epoch": 0.0213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.3778591033851784e-05,
-      "loss": 1.2125,
+      "learning_rate": 1.4234404536862006e-05,
+      "loss": 1.1566,
       "step": 753
     },
     {
-      "epoch": 0.020708027793798576,
+      "epoch": 0.021396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.3796889295516927e-05,
-      "loss": 1.083,
+      "learning_rate": 1.4253308128544426e-05,
+      "loss": 1.2277,
       "step": 754
     },
     {
-      "epoch": 0.02073549202164181,
+      "epoch": 0.021424517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.3815187557182068e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.4272211720226845e-05,
+      "loss": 1.0525,
       "step": 755
     },
     {
-      "epoch": 0.020762956249485046,
+      "epoch": 0.02145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.383348581884721e-05,
-      "loss": 1.1125,
+      "learning_rate": 1.4291115311909265e-05,
+      "loss": 0.9971,
       "step": 756
     },
     {
-      "epoch": 0.02079042047732828,
+      "epoch": 0.02148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.3851784080512352e-05,
-      "loss": 1.1639,
+      "learning_rate": 1.4310018903591683e-05,
+      "loss": 1.0791,
       "step": 757
     },
     {
-      "epoch": 0.020817884705171513,
+      "epoch": 0.021509648127128262,
       "grad_norm": 0.0,
-      "learning_rate": 1.3870082342177492e-05,
-      "loss": 1.1848,
+      "learning_rate": 1.4328922495274103e-05,
+      "loss": 1.0686,
       "step": 758
     },
     {
-      "epoch": 0.02084534893301475,
+      "epoch": 0.021538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.3888380603842635e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.4347826086956522e-05,
+      "loss": 1.0901,
       "step": 759
     },
     {
-      "epoch": 0.020872813160857984,
+      "epoch": 0.021566401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.390667886550778e-05,
-      "loss": 1.1577,
+      "learning_rate": 1.4366729678638942e-05,
+      "loss": 1.0773,
       "step": 760
     },
     {
-      "epoch": 0.020900277388701215,
+      "epoch": 0.02159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.392497712717292e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.4385633270321361e-05,
+      "loss": 1.1132,
       "step": 761
     },
     {
-      "epoch": 0.02092774161654445,
+      "epoch": 0.021623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.3943275388838063e-05,
-      "loss": 1.1098,
+      "learning_rate": 1.4404536862003781e-05,
+      "loss": 1.0583,
       "step": 762
     },
     {
-      "epoch": 0.020955205844387686,
+      "epoch": 0.021651532349602723,
       "grad_norm": 0.0,
-      "learning_rate": 1.3961573650503204e-05,
-      "loss": 1.0726,
+      "learning_rate": 1.44234404536862e-05,
+      "loss": 1.2404,
       "step": 763
     },
     {
-      "epoch": 0.020982670072230918,
+      "epoch": 0.021679909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 1.3979871912168346e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.4442344045368622e-05,
+      "loss": 1.0255,
       "step": 764
     },
     {
-      "epoch": 0.021010134300074153,
+      "epoch": 0.021708286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.3998170173833487e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.4461247637051042e-05,
+      "loss": 1.065,
       "step": 765
     },
     {
-      "epoch": 0.021037598527917388,
+      "epoch": 0.021736662883087402,
       "grad_norm": 0.0,
-      "learning_rate": 1.4016468435498628e-05,
-      "loss": 1.0778,
+      "learning_rate": 1.4480151228733461e-05,
+      "loss": 1.1645,
       "step": 766
     },
     {
-      "epoch": 0.021065062755760623,
+      "epoch": 0.021765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.4034766697163771e-05,
-      "loss": 1.1353,
+      "learning_rate": 1.4499054820415881e-05,
+      "loss": 1.0922,
       "step": 767
     },
     {
-      "epoch": 0.021092526983603855,
+      "epoch": 0.021793416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.4053064958828912e-05,
-      "loss": 1.0894,
+      "learning_rate": 1.45179584120983e-05,
+      "loss": 1.1274,
       "step": 768
     },
     {
-      "epoch": 0.02111999121144709,
+      "epoch": 0.021821793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.4071363220494054e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.4536862003780719e-05,
+      "loss": 1.0566,
       "step": 769
     },
     {
-      "epoch": 0.021147455439290325,
+      "epoch": 0.02185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.4089661482159195e-05,
-      "loss": 1.1389,
+      "learning_rate": 1.4555765595463138e-05,
+      "loss": 1.0489,
       "step": 770
     },
     {
-      "epoch": 0.021174919667133557,
+      "epoch": 0.021878547105561863,
       "grad_norm": 0.0,
-      "learning_rate": 1.4107959743824338e-05,
-      "loss": 1.1629,
+      "learning_rate": 1.4574669187145558e-05,
+      "loss": 1.1254,
       "step": 771
     },
     {
-      "epoch": 0.021202383894976792,
+      "epoch": 0.021906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.4126258005489479e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.4593572778827978e-05,
+      "loss": 1.0549,
       "step": 772
     },
     {
-      "epoch": 0.021229848122820028,
+      "epoch": 0.021935300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4144556267154622e-05,
-      "loss": 1.1712,
+      "learning_rate": 1.4612476370510397e-05,
+      "loss": 1.0265,
       "step": 773
     },
     {
-      "epoch": 0.021257312350663263,
+      "epoch": 0.02196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.4162854528819762e-05,
-      "loss": 1.1244,
+      "learning_rate": 1.4631379962192817e-05,
+      "loss": 1.0092,
       "step": 774
     },
     {
-      "epoch": 0.021284776578506494,
+      "epoch": 0.02199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.4181152790484905e-05,
-      "loss": 1.0885,
+      "learning_rate": 1.4650283553875237e-05,
+      "loss": 1.1013,
       "step": 775
     },
     {
-      "epoch": 0.02131224080634973,
+      "epoch": 0.022020431328036324,
       "grad_norm": 0.0,
-      "learning_rate": 1.4199451052150046e-05,
-      "loss": 1.1843,
+      "learning_rate": 1.4669187145557658e-05,
+      "loss": 1.0743,
       "step": 776
     },
     {
-      "epoch": 0.021339705034192965,
+      "epoch": 0.022048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.4217749313815189e-05,
-      "loss": 1.1932,
+      "learning_rate": 1.4688090737240078e-05,
+      "loss": 1.0157,
       "step": 777
     },
     {
-      "epoch": 0.021367169262036197,
+      "epoch": 0.022077185017026105,
       "grad_norm": 0.0,
-      "learning_rate": 1.423604757548033e-05,
-      "loss": 1.1577,
+      "learning_rate": 1.4706994328922497e-05,
+      "loss": 1.1353,
       "step": 778
     },
     {
-      "epoch": 0.021394633489879432,
+      "epoch": 0.022105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.425434583714547e-05,
-      "loss": 1.0853,
+      "learning_rate": 1.4725897920604917e-05,
+      "loss": 1.12,
       "step": 779
     },
     {
-      "epoch": 0.021422097717722667,
+      "epoch": 0.02213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.4272644098810615e-05,
-      "loss": 1.095,
+      "learning_rate": 1.4744801512287337e-05,
+      "loss": 1.2262,
       "step": 780
     },
     {
-      "epoch": 0.0214495619455659,
+      "epoch": 0.022162315550510785,
       "grad_norm": 0.0,
-      "learning_rate": 1.4290942360475757e-05,
-      "loss": 1.1033,
+      "learning_rate": 1.4763705103969756e-05,
+      "loss": 1.1503,
       "step": 781
     },
     {
-      "epoch": 0.021477026173409134,
+      "epoch": 0.022190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.4309240622140898e-05,
-      "loss": 1.2321,
+      "learning_rate": 1.4782608695652174e-05,
+      "loss": 1.2505,
       "step": 782
     },
     {
-      "epoch": 0.02150449040125237,
+      "epoch": 0.022219069239500566,
       "grad_norm": 0.0,
-      "learning_rate": 1.4327538883806041e-05,
-      "loss": 1.1648,
+      "learning_rate": 1.4801512287334594e-05,
+      "loss": 1.0793,
       "step": 783
     },
     {
-      "epoch": 0.021531954629095604,
+      "epoch": 0.02224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4345837145471182e-05,
-      "loss": 1.119,
+      "learning_rate": 1.4820415879017014e-05,
+      "loss": 1.1154,
       "step": 784
     },
     {
-      "epoch": 0.021559418856938836,
+      "epoch": 0.02227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.4364135407136324e-05,
-      "loss": 1.1653,
+      "learning_rate": 1.4839319470699433e-05,
+      "loss": 1.0829,
       "step": 785
     },
     {
-      "epoch": 0.02158688308478207,
+      "epoch": 0.022304199772985246,
       "grad_norm": 0.0,
-      "learning_rate": 1.4382433668801465e-05,
-      "loss": 1.1479,
+      "learning_rate": 1.4858223062381853e-05,
+      "loss": 1.0953,
       "step": 786
     },
     {
-      "epoch": 0.021614347312625307,
+      "epoch": 0.022332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.4400731930466606e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.4877126654064273e-05,
+      "loss": 1.1858,
       "step": 787
     },
     {
-      "epoch": 0.02164181154046854,
+      "epoch": 0.022360953461975027,
       "grad_norm": 0.0,
-      "learning_rate": 1.4419030192131749e-05,
-      "loss": 1.1046,
+      "learning_rate": 1.4896030245746694e-05,
+      "loss": 1.084,
       "step": 788
     },
     {
-      "epoch": 0.021669275768311774,
+      "epoch": 0.02238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.443732845379689e-05,
-      "loss": 1.0969,
+      "learning_rate": 1.4914933837429114e-05,
+      "loss": 1.1877,
       "step": 789
     },
     {
-      "epoch": 0.02169673999615501,
+      "epoch": 0.022417707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 1.4455626715462032e-05,
-      "loss": 1.0845,
+      "learning_rate": 1.4933837429111533e-05,
+      "loss": 1.1587,
       "step": 790
     },
     {
-      "epoch": 0.021724204223998244,
+      "epoch": 0.022446083995459706,
       "grad_norm": 0.0,
-      "learning_rate": 1.4473924977127173e-05,
-      "loss": 1.1139,
+      "learning_rate": 1.4952741020793953e-05,
+      "loss": 0.9821,
       "step": 791
     },
     {
-      "epoch": 0.021751668451841476,
+      "epoch": 0.022474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.4492223238792316e-05,
-      "loss": 1.1273,
+      "learning_rate": 1.4971644612476373e-05,
+      "loss": 1.1376,
       "step": 792
     },
     {
-      "epoch": 0.02177913267968471,
+      "epoch": 0.022502837684449488,
       "grad_norm": 0.0,
-      "learning_rate": 1.4510521500457457e-05,
-      "loss": 1.231,
+      "learning_rate": 1.4990548204158792e-05,
+      "loss": 1.0818,
       "step": 793
     },
     {
-      "epoch": 0.021806596907527946,
+      "epoch": 0.022531214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.45288197621226e-05,
-      "loss": 1.1609,
+      "learning_rate": 1.500945179584121e-05,
+      "loss": 1.0822,
       "step": 794
     },
     {
-      "epoch": 0.021834061135371178,
+      "epoch": 0.022559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.454711802378774e-05,
-      "loss": 1.1396,
+      "learning_rate": 1.502835538752363e-05,
+      "loss": 1.1562,
       "step": 795
     },
     {
-      "epoch": 0.021861525363214413,
+      "epoch": 0.022587968217934167,
       "grad_norm": 0.0,
-      "learning_rate": 1.4565416285452883e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.504725897920605e-05,
+      "loss": 1.1499,
       "step": 796
     },
     {
-      "epoch": 0.02188898959105765,
+      "epoch": 0.022616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.4583714547118024e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.506616257088847e-05,
+      "loss": 1.1045,
       "step": 797
     },
     {
-      "epoch": 0.02191645381890088,
+      "epoch": 0.02264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.4602012808783167e-05,
-      "loss": 1.1535,
+      "learning_rate": 1.5085066162570889e-05,
+      "loss": 1.0016,
       "step": 798
     },
     {
-      "epoch": 0.021943918046744115,
+      "epoch": 0.022673098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.4620311070448308e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.5103969754253308e-05,
+      "loss": 1.0579,
       "step": 799
     },
     {
-      "epoch": 0.02197138227458735,
+      "epoch": 0.022701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.4638609332113452e-05,
-      "loss": 1.1784,
+      "learning_rate": 1.512287334593573e-05,
+      "loss": 1.1352,
       "step": 800
     },
     {
-      "epoch": 0.021998846502430586,
+      "epoch": 0.022729852440408628,
       "grad_norm": 0.0,
-      "learning_rate": 1.4656907593778593e-05,
-      "loss": 1.1599,
+      "learning_rate": 1.514177693761815e-05,
+      "loss": 1.1414,
       "step": 801
     },
     {
-      "epoch": 0.022026310730273817,
+      "epoch": 0.02275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.4675205855443735e-05,
-      "loss": 1.1361,
+      "learning_rate": 1.516068052930057e-05,
+      "loss": 1.1441,
       "step": 802
     },
     {
-      "epoch": 0.022053774958117053,
+      "epoch": 0.02278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.4693504117108876e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.5179584120982989e-05,
+      "loss": 1.1287,
       "step": 803
     },
     {
-      "epoch": 0.022081239185960288,
+      "epoch": 0.022814982973893304,
       "grad_norm": 0.0,
-      "learning_rate": 1.4711802378774019e-05,
-      "loss": 1.1938,
+      "learning_rate": 1.5198487712665408e-05,
+      "loss": 1.0449,
       "step": 804
     },
     {
-      "epoch": 0.02210870341380352,
+      "epoch": 0.022843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.473010064043916e-05,
-      "loss": 1.1417,
+      "learning_rate": 1.5217391304347828e-05,
+      "loss": 1.1455,
       "step": 805
     },
     {
-      "epoch": 0.022136167641646755,
+      "epoch": 0.02287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.4748398902104303e-05,
-      "loss": 1.142,
+      "learning_rate": 1.5236294896030246e-05,
+      "loss": 1.1479,
       "step": 806
     },
     {
-      "epoch": 0.02216363186948999,
+      "epoch": 0.02290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.4766697163769443e-05,
-      "loss": 1.1666,
+      "learning_rate": 1.5255198487712666e-05,
+      "loss": 1.0531,
       "step": 807
     },
     {
-      "epoch": 0.022191096097333222,
+      "epoch": 0.02292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.4784995425434584e-05,
-      "loss": 1.0358,
+      "learning_rate": 1.5274102079395087e-05,
+      "loss": 1.025,
       "step": 808
     },
     {
-      "epoch": 0.022218560325176457,
+      "epoch": 0.022956867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.4803293687099727e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.5293005671077507e-05,
+      "loss": 1.1536,
       "step": 809
     },
     {
-      "epoch": 0.022246024553019692,
+      "epoch": 0.022985244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.4821591948764868e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.5311909262759926e-05,
+      "loss": 1.1478,
       "step": 810
     },
     {
-      "epoch": 0.022273488780862927,
+      "epoch": 0.02301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.483989021043001e-05,
-      "loss": 1.1716,
+      "learning_rate": 1.5330812854442346e-05,
+      "loss": 1.0617,
       "step": 811
     },
     {
-      "epoch": 0.02230095300870616,
+      "epoch": 0.02304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.4858188472095151e-05,
-      "loss": 1.184,
+      "learning_rate": 1.5349716446124766e-05,
+      "loss": 1.0875,
       "step": 812
     },
     {
-      "epoch": 0.022328417236549394,
+      "epoch": 0.02307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.4876486733760294e-05,
-      "loss": 1.1531,
+      "learning_rate": 1.5368620037807185e-05,
+      "loss": 0.9966,
       "step": 813
     },
     {
-      "epoch": 0.02235588146439263,
+      "epoch": 0.023098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.4894784995425435e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.5387523629489605e-05,
+      "loss": 1.0558,
       "step": 814
     },
     {
-      "epoch": 0.02238334569223586,
+      "epoch": 0.023127128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 1.4913083257090578e-05,
-      "loss": 1.1466,
+      "learning_rate": 1.5406427221172025e-05,
+      "loss": 1.1823,
       "step": 815
     },
     {
-      "epoch": 0.022410809920079097,
+      "epoch": 0.02315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.4931381518755719e-05,
-      "loss": 1.2089,
+      "learning_rate": 1.5425330812854444e-05,
+      "loss": 1.1734,
       "step": 816
     },
     {
-      "epoch": 0.022438274147922332,
+      "epoch": 0.0231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.4949679780420861e-05,
-      "loss": 1.1654,
+      "learning_rate": 1.5444234404536864e-05,
+      "loss": 1.0455,
       "step": 817
     },
     {
-      "epoch": 0.022465738375765567,
+      "epoch": 0.023212258796821792,
       "grad_norm": 0.0,
-      "learning_rate": 1.4967978042086002e-05,
-      "loss": 1.0409,
+      "learning_rate": 1.5463137996219284e-05,
+      "loss": 1.0964,
       "step": 818
     },
     {
-      "epoch": 0.0224932026036088,
+      "epoch": 0.023240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.4986276303751143e-05,
-      "loss": 1.1489,
+      "learning_rate": 1.54820415879017e-05,
+      "loss": 1.06,
       "step": 819
     },
     {
-      "epoch": 0.022520666831452034,
+      "epoch": 0.023269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.5004574565416287e-05,
-      "loss": 1.1263,
+      "learning_rate": 1.5500945179584123e-05,
+      "loss": 1.1407,
       "step": 820
     },
     {
-      "epoch": 0.02254813105929527,
+      "epoch": 0.02329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.502287282708143e-05,
-      "loss": 1.1682,
+      "learning_rate": 1.5519848771266543e-05,
+      "loss": 1.0937,
       "step": 821
     },
     {
-      "epoch": 0.0225755952871385,
+      "epoch": 0.023325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.504117108874657e-05,
-      "loss": 1.2006,
+      "learning_rate": 1.5538752362948962e-05,
+      "loss": 1.0789,
       "step": 822
     },
     {
-      "epoch": 0.022603059514981736,
+      "epoch": 0.023354143019296253,
       "grad_norm": 0.0,
-      "learning_rate": 1.5059469350411713e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.5557655954631382e-05,
+      "loss": 1.0791,
       "step": 823
     },
     {
-      "epoch": 0.02263052374282497,
+      "epoch": 0.023382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.5077767612076854e-05,
-      "loss": 1.1192,
+      "learning_rate": 1.55765595463138e-05,
+      "loss": 1.2011,
       "step": 824
     },
     {
-      "epoch": 0.022657987970668203,
+      "epoch": 0.023410896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 1.5096065873741997e-05,
-      "loss": 1.0819,
+      "learning_rate": 1.559546313799622e-05,
+      "loss": 1.0826,
       "step": 825
     },
     {
-      "epoch": 0.022685452198511438,
+      "epoch": 0.023439273552780932,
       "grad_norm": 0.0,
-      "learning_rate": 1.5114364135407138e-05,
-      "loss": 1.1081,
+      "learning_rate": 1.561436672967864e-05,
+      "loss": 1.0533,
       "step": 826
     },
     {
-      "epoch": 0.022712916426354673,
+      "epoch": 0.023467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.5132662397072279e-05,
-      "loss": 1.1073,
+      "learning_rate": 1.563327032136106e-05,
+      "loss": 1.2472,
       "step": 827
     },
     {
-      "epoch": 0.02274038065419791,
+      "epoch": 0.023496027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5150960658737421e-05,
-      "loss": 1.1503,
+      "learning_rate": 1.565217391304348e-05,
+      "loss": 1.0664,
       "step": 828
     },
     {
-      "epoch": 0.02276784488204114,
+      "epoch": 0.023524404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 1.5169258920402562e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.56710775047259e-05,
+      "loss": 1.0267,
       "step": 829
     },
     {
-      "epoch": 0.022795309109884376,
+      "epoch": 0.0235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5187557182067705e-05,
-      "loss": 1.2308,
+      "learning_rate": 1.568998109640832e-05,
+      "loss": 1.0615,
       "step": 830
     },
     {
-      "epoch": 0.02282277333772761,
+      "epoch": 0.023581157775255393,
       "grad_norm": 0.0,
-      "learning_rate": 1.5205855443732846e-05,
-      "loss": 1.1288,
+      "learning_rate": 1.5708884688090736e-05,
+      "loss": 1.0719,
       "step": 831
     },
     {
-      "epoch": 0.022850237565570843,
+      "epoch": 0.023609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.5224153705397989e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.572778827977316e-05,
+      "loss": 1.081,
       "step": 832
     },
     {
-      "epoch": 0.022877701793414078,
+      "epoch": 0.023637911464245175,
       "grad_norm": 0.0,
-      "learning_rate": 1.524245196706313e-05,
-      "loss": 1.1626,
+      "learning_rate": 1.574669187145558e-05,
+      "loss": 1.1225,
       "step": 833
     },
     {
-      "epoch": 0.022905166021257313,
+      "epoch": 0.02366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.5260750228728272e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.5765595463137998e-05,
+      "loss": 1.0608,
       "step": 834
     },
     {
-      "epoch": 0.022932630249100548,
+      "epoch": 0.02369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.5279048490393413e-05,
-      "loss": 1.1753,
+      "learning_rate": 1.5784499054820418e-05,
+      "loss": 1.0762,
       "step": 835
     },
     {
-      "epoch": 0.02296009447694378,
+      "epoch": 0.023723041997729854,
       "grad_norm": 0.0,
-      "learning_rate": 1.5297346752058554e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.5803402646502838e-05,
+      "loss": 1.108,
       "step": 836
     },
     {
-      "epoch": 0.022987558704787015,
+      "epoch": 0.023751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.5315645013723698e-05,
-      "loss": 1.155,
+      "learning_rate": 1.5822306238185257e-05,
+      "loss": 1.1208,
       "step": 837
     },
     {
-      "epoch": 0.02301502293263025,
+      "epoch": 0.023779795686719635,
       "grad_norm": 0.0,
-      "learning_rate": 1.533394327538884e-05,
-      "loss": 1.1371,
+      "learning_rate": 1.5841209829867677e-05,
+      "loss": 1.1525,
       "step": 838
     },
     {
-      "epoch": 0.023042487160473482,
+      "epoch": 0.02380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.535224153705398e-05,
-      "loss": 1.1992,
+      "learning_rate": 1.5860113421550097e-05,
+      "loss": 0.9816,
       "step": 839
     },
     {
-      "epoch": 0.023069951388316717,
+      "epoch": 0.02383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.5370539798719124e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.5879017013232516e-05,
+      "loss": 1.1038,
       "step": 840
     },
     {
-      "epoch": 0.023097415616159953,
+      "epoch": 0.023864926220204315,
       "grad_norm": 0.0,
-      "learning_rate": 1.5388838060384265e-05,
-      "loss": 1.019,
+      "learning_rate": 1.5897920604914936e-05,
+      "loss": 1.1779,
       "step": 841
     },
     {
-      "epoch": 0.023124879844003184,
+      "epoch": 0.023893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.5407136322049406e-05,
-      "loss": 1.0985,
+      "learning_rate": 1.5916824196597355e-05,
+      "loss": 1.1103,
       "step": 842
     },
     {
-      "epoch": 0.02315234407184642,
+      "epoch": 0.023921679909194096,
       "grad_norm": 0.0,
-      "learning_rate": 1.542543458371455e-05,
-      "loss": 1.1273,
+      "learning_rate": 1.5935727788279772e-05,
+      "loss": 1.1468,
       "step": 843
     },
     {
-      "epoch": 0.023179808299689655,
+      "epoch": 0.02395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.544373284537969e-05,
-      "loss": 1.0858,
+      "learning_rate": 1.5954631379962195e-05,
+      "loss": 1.1675,
       "step": 844
     },
     {
-      "epoch": 0.02320727252753289,
+      "epoch": 0.02397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.5462031107044832e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.5973534971644614e-05,
+      "loss": 1.1182,
       "step": 845
     },
     {
-      "epoch": 0.02323473675537612,
+      "epoch": 0.024006810442678776,
       "grad_norm": 0.0,
-      "learning_rate": 1.5480329368709973e-05,
-      "loss": 1.2077,
+      "learning_rate": 1.5992438563327034e-05,
+      "loss": 1.0066,
       "step": 846
     },
     {
-      "epoch": 0.023262200983219357,
+      "epoch": 0.024035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.5498627630375118e-05,
-      "loss": 1.1911,
+      "learning_rate": 1.6011342155009454e-05,
+      "loss": 1.1363,
       "step": 847
     },
     {
-      "epoch": 0.023289665211062592,
+      "epoch": 0.024063564131668557,
       "grad_norm": 0.0,
-      "learning_rate": 1.551692589204026e-05,
-      "loss": 1.1707,
+      "learning_rate": 1.6030245746691873e-05,
+      "loss": 1.1112,
       "step": 848
     },
     {
-      "epoch": 0.023317129438905824,
+      "epoch": 0.02409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.55352241537054e-05,
-      "loss": 1.1658,
+      "learning_rate": 1.6049149338374293e-05,
+      "loss": 1.0151,
       "step": 849
     },
     {
-      "epoch": 0.02334459366674906,
+      "epoch": 0.024120317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 1.555352241537054e-05,
-      "loss": 1.1459,
+      "learning_rate": 1.6068052930056713e-05,
+      "loss": 1.1466,
       "step": 850
     },
     {
-      "epoch": 0.023372057894592294,
+      "epoch": 0.024148694665153236,
       "grad_norm": 0.0,
-      "learning_rate": 1.557182067703568e-05,
-      "loss": 1.1191,
+      "learning_rate": 1.6086956521739132e-05,
+      "loss": 1.0952,
       "step": 851
     },
     {
-      "epoch": 0.02339952212243553,
+      "epoch": 0.024177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.5590118938700826e-05,
-      "loss": 1.1304,
+      "learning_rate": 1.6105860113421552e-05,
+      "loss": 1.1207,
       "step": 852
     },
     {
-      "epoch": 0.02342698635027876,
+      "epoch": 0.024205448354143018,
       "grad_norm": 0.0,
-      "learning_rate": 1.5608417200365967e-05,
-      "loss": 1.2095,
+      "learning_rate": 1.6124763705103972e-05,
+      "loss": 1.1338,
       "step": 853
     },
     {
-      "epoch": 0.023454450578121996,
+      "epoch": 0.024233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.5626715462031107e-05,
-      "loss": 1.0498,
+      "learning_rate": 1.614366729678639e-05,
+      "loss": 1.0369,
       "step": 854
     },
     {
-      "epoch": 0.02348191480596523,
+      "epoch": 0.024262202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.564501372369625e-05,
-      "loss": 1.2312,
+      "learning_rate": 1.616257088846881e-05,
+      "loss": 1.051,
       "step": 855
     },
     {
-      "epoch": 0.023509379033808463,
+      "epoch": 0.024290578887627697,
       "grad_norm": 0.0,
-      "learning_rate": 1.5663311985361393e-05,
-      "loss": 1.0791,
+      "learning_rate": 1.618147448015123e-05,
+      "loss": 1.0521,
       "step": 856
     },
     {
-      "epoch": 0.0235368432616517,
+      "epoch": 0.024318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.5681610247026534e-05,
-      "loss": 1.1495,
+      "learning_rate": 1.620037807183365e-05,
+      "loss": 1.051,
       "step": 857
     },
     {
-      "epoch": 0.023564307489494934,
+      "epoch": 0.02434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.5699908508691675e-05,
-      "loss": 1.1576,
+      "learning_rate": 1.621928166351607e-05,
+      "loss": 1.1226,
       "step": 858
     },
     {
-      "epoch": 0.023591771717338166,
+      "epoch": 0.024375709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 1.5718206770356816e-05,
-      "loss": 1.0969,
+      "learning_rate": 1.623818525519849e-05,
+      "loss": 1.0548,
       "step": 859
     },
     {
-      "epoch": 0.0236192359451814,
+      "epoch": 0.024404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.5736505032021956e-05,
-      "loss": 1.1946,
+      "learning_rate": 1.625708884688091e-05,
+      "loss": 1.0794,
       "step": 860
     },
     {
-      "epoch": 0.023646700173024636,
+      "epoch": 0.024432463110102158,
       "grad_norm": 0.0,
-      "learning_rate": 1.57548032936871e-05,
-      "loss": 1.3373,
+      "learning_rate": 1.627599243856333e-05,
+      "loss": 1.0048,
       "step": 861
     },
     {
-      "epoch": 0.02367416440086787,
+      "epoch": 0.02446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.5773101555352245e-05,
-      "loss": 1.1577,
+      "learning_rate": 1.629489603024575e-05,
+      "loss": 0.9879,
       "step": 862
     },
     {
-      "epoch": 0.023701628628711103,
+      "epoch": 0.02448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.5791399817017386e-05,
-      "loss": 1.101,
+      "learning_rate": 1.631379962192817e-05,
+      "loss": 1.1754,
       "step": 863
     },
     {
-      "epoch": 0.023729092856554338,
+      "epoch": 0.024517593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.5809698078682527e-05,
-      "loss": 1.1054,
+      "learning_rate": 1.6332703213610588e-05,
+      "loss": 1.141,
       "step": 864
     },
     {
-      "epoch": 0.023756557084397573,
+      "epoch": 0.024545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.5827996340347668e-05,
-      "loss": 1.1486,
+      "learning_rate": 1.6351606805293008e-05,
+      "loss": 1.0932,
       "step": 865
     },
     {
-      "epoch": 0.023784021312240805,
+      "epoch": 0.02457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.5846294602012812e-05,
-      "loss": 1.165,
+      "learning_rate": 1.6370510396975427e-05,
+      "loss": 1.0335,
       "step": 866
     },
     {
-      "epoch": 0.02381148554008404,
+      "epoch": 0.02460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.5864592863677953e-05,
-      "loss": 1.2039,
+      "learning_rate": 1.6389413988657847e-05,
+      "loss": 1.0093,
       "step": 867
     },
     {
-      "epoch": 0.023838949767927275,
+      "epoch": 0.0246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.5882891125343094e-05,
-      "loss": 1.1987,
+      "learning_rate": 1.6408317580340267e-05,
+      "loss": 1.0605,
       "step": 868
     },
     {
-      "epoch": 0.023866413995770507,
+      "epoch": 0.024659477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.5901189387008235e-05,
-      "loss": 1.1188,
+      "learning_rate": 1.6427221172022686e-05,
+      "loss": 1.1274,
       "step": 869
     },
     {
-      "epoch": 0.023893878223613742,
+      "epoch": 0.024687854710556185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5919487648673376e-05,
-      "loss": 1.1124,
+      "learning_rate": 1.6446124763705106e-05,
+      "loss": 1.1568,
       "step": 870
     },
     {
-      "epoch": 0.023921342451456978,
+      "epoch": 0.02471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.593778591033852e-05,
-      "loss": 1.0472,
+      "learning_rate": 1.6465028355387526e-05,
+      "loss": 1.0097,
       "step": 871
     },
     {
-      "epoch": 0.023948806679300213,
+      "epoch": 0.02474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.595608417200366e-05,
-      "loss": 1.1035,
+      "learning_rate": 1.6483931947069945e-05,
+      "loss": 1.114,
       "step": 872
     },
     {
-      "epoch": 0.023976270907143445,
+      "epoch": 0.02477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.5974382433668802e-05,
-      "loss": 1.1387,
+      "learning_rate": 1.6502835538752365e-05,
+      "loss": 1.0433,
       "step": 873
     },
     {
-      "epoch": 0.02400373513498668,
+      "epoch": 0.024801362088535756,
       "grad_norm": 0.0,
-      "learning_rate": 1.5992680695333943e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.6521739130434785e-05,
+      "loss": 1.1086,
       "step": 874
     },
     {
-      "epoch": 0.024031199362829915,
+      "epoch": 0.024829738933030646,
       "grad_norm": 0.0,
-      "learning_rate": 1.6010978956999087e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.6540642722117204e-05,
+      "loss": 1.1088,
       "step": 875
     },
     {
-      "epoch": 0.024058663590673147,
+      "epoch": 0.02485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.6029277218664228e-05,
-      "loss": 1.1874,
+      "learning_rate": 1.6559546313799624e-05,
+      "loss": 1.1254,
       "step": 876
     },
     {
-      "epoch": 0.024086127818516382,
+      "epoch": 0.02488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.604757548032937e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.6578449905482044e-05,
+      "loss": 1.0885,
       "step": 877
     },
     {
-      "epoch": 0.024113592046359617,
+      "epoch": 0.024914869466515322,
       "grad_norm": 0.0,
-      "learning_rate": 1.606587374199451e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.6597353497164463e-05,
+      "loss": 1.0163,
       "step": 878
     },
     {
-      "epoch": 0.024141056274202852,
+      "epoch": 0.024943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.6084172003659654e-05,
-      "loss": 1.0971,
+      "learning_rate": 1.6616257088846883e-05,
+      "loss": 1.1274,
       "step": 879
     },
     {
-      "epoch": 0.024168520502046084,
+      "epoch": 0.024971623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.6102470265324795e-05,
-      "loss": 1.215,
+      "learning_rate": 1.6635160680529302e-05,
+      "loss": 1.0764,
       "step": 880
     },
     {
-      "epoch": 0.02419598472988932,
+      "epoch": 0.025,
       "grad_norm": 0.0,
-      "learning_rate": 1.612076852698994e-05,
-      "loss": 1.1262,
+      "learning_rate": 1.6654064272211722e-05,
+      "loss": 1.1566,
       "step": 881
     },
     {
-      "epoch": 0.024223448957732555,
+      "epoch": 0.025028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.613906678865508e-05,
-      "loss": 1.121,
+      "learning_rate": 1.6672967863894142e-05,
+      "loss": 1.0825,
       "step": 882
     },
     {
-      "epoch": 0.024250913185575786,
+      "epoch": 0.025056753688989783,
       "grad_norm": 0.0,
-      "learning_rate": 1.615736505032022e-05,
-      "loss": 1.0925,
+      "learning_rate": 1.669187145557656e-05,
+      "loss": 1.1797,
       "step": 883
     },
     {
-      "epoch": 0.02427837741341902,
+      "epoch": 0.025085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.6175663311985362e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.671077504725898e-05,
+      "loss": 1.1177,
       "step": 884
     },
     {
-      "epoch": 0.024305841641262257,
+      "epoch": 0.025113507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 1.6193961573650507e-05,
-      "loss": 1.1211,
+      "learning_rate": 1.67296786389414e-05,
+      "loss": 1.0751,
       "step": 885
     },
     {
-      "epoch": 0.02433330586910549,
+      "epoch": 0.025141884222474462,
       "grad_norm": 0.0,
-      "learning_rate": 1.6212259835315648e-05,
-      "loss": 1.1066,
+      "learning_rate": 1.674858223062382e-05,
+      "loss": 1.0567,
       "step": 886
     },
     {
-      "epoch": 0.024360770096948724,
+      "epoch": 0.025170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.623055809698079e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.676748582230624e-05,
+      "loss": 1.1445,
       "step": 887
     },
     {
-      "epoch": 0.02438823432479196,
+      "epoch": 0.025198637911464244,
       "grad_norm": 0.0,
-      "learning_rate": 1.624885635864593e-05,
-      "loss": 1.1171,
+      "learning_rate": 1.678638941398866e-05,
+      "loss": 1.1209,
       "step": 888
     },
     {
-      "epoch": 0.024415698552635194,
+      "epoch": 0.025227014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 1.626715462031107e-05,
-      "loss": 1.1105,
+      "learning_rate": 1.680529300567108e-05,
+      "loss": 1.0367,
       "step": 889
     },
     {
-      "epoch": 0.024443162780478426,
+      "epoch": 0.02525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.6285452881976215e-05,
-      "loss": 1.1872,
+      "learning_rate": 1.68241965973535e-05,
+      "loss": 1.1688,
       "step": 890
     },
     {
-      "epoch": 0.02447062700832166,
+      "epoch": 0.025283768444948923,
       "grad_norm": 0.0,
-      "learning_rate": 1.6303751143641356e-05,
-      "loss": 1.148,
+      "learning_rate": 1.684310018903592e-05,
+      "loss": 1.0974,
       "step": 891
     },
     {
-      "epoch": 0.024498091236164896,
+      "epoch": 0.025312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.6322049405306496e-05,
-      "loss": 1.0407,
+      "learning_rate": 1.686200378071834e-05,
+      "loss": 1.176,
       "step": 892
     },
     {
-      "epoch": 0.024525555464008128,
+      "epoch": 0.025340522133938705,
       "grad_norm": 0.0,
-      "learning_rate": 1.6340347666971637e-05,
-      "loss": 1.0945,
+      "learning_rate": 1.6880907372400758e-05,
+      "loss": 1.079,
       "step": 893
     },
     {
-      "epoch": 0.024553019691851363,
+      "epoch": 0.0253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.6358645928636782e-05,
-      "loss": 1.1857,
+      "learning_rate": 1.6899810964083178e-05,
+      "loss": 1.1097,
       "step": 894
     },
     {
-      "epoch": 0.0245804839196946,
+      "epoch": 0.02539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.6376944190301923e-05,
-      "loss": 1.1092,
+      "learning_rate": 1.6918714555765597e-05,
+      "loss": 1.0207,
       "step": 895
     },
     {
-      "epoch": 0.024607948147537834,
+      "epoch": 0.025425652667423384,
       "grad_norm": 0.0,
-      "learning_rate": 1.6395242451967064e-05,
-      "loss": 1.1022,
+      "learning_rate": 1.6937618147448017e-05,
+      "loss": 1.1157,
       "step": 896
     },
     {
-      "epoch": 0.024635412375381065,
+      "epoch": 0.025454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.6413540713632204e-05,
-      "loss": 1.1363,
+      "learning_rate": 1.6956521739130437e-05,
+      "loss": 1.1028,
       "step": 897
     },
     {
-      "epoch": 0.0246628766032243,
+      "epoch": 0.025482406356413165,
       "grad_norm": 0.0,
-      "learning_rate": 1.643183897529735e-05,
-      "loss": 1.171,
+      "learning_rate": 1.6975425330812856e-05,
+      "loss": 1.1507,
       "step": 898
     },
     {
-      "epoch": 0.024690340831067536,
+      "epoch": 0.02551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.645013723696249e-05,
-      "loss": 1.1235,
+      "learning_rate": 1.6994328922495276e-05,
+      "loss": 1.0549,
       "step": 899
     },
     {
-      "epoch": 0.024717805058910768,
+      "epoch": 0.02553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.646843549862763e-05,
-      "loss": 1.1695,
+      "learning_rate": 1.7013232514177696e-05,
+      "loss": 1.1238,
       "step": 900
     },
     {
-      "epoch": 0.024745269286754003,
+      "epoch": 0.025567536889897845,
       "grad_norm": 0.0,
-      "learning_rate": 1.6486733760292775e-05,
-      "loss": 1.1162,
+      "learning_rate": 1.7032136105860115e-05,
+      "loss": 1.177,
       "step": 901
     },
     {
-      "epoch": 0.024772733514597238,
+      "epoch": 0.025595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.6505032021957916e-05,
-      "loss": 1.1658,
+      "learning_rate": 1.7051039697542535e-05,
+      "loss": 1.1144,
       "step": 902
     },
     {
-      "epoch": 0.02480019774244047,
+      "epoch": 0.025624290578887626,
       "grad_norm": 0.0,
-      "learning_rate": 1.6523330283623057e-05,
-      "loss": 1.1012,
+      "learning_rate": 1.7069943289224955e-05,
+      "loss": 1.1054,
       "step": 903
     },
     {
-      "epoch": 0.024827661970283705,
+      "epoch": 0.02565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.65416285452882e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.7088846880907374e-05,
+      "loss": 1.0129,
       "step": 904
     },
     {
-      "epoch": 0.02485512619812694,
+      "epoch": 0.02568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.6559926806953342e-05,
-      "loss": 1.1323,
+      "learning_rate": 1.7107750472589794e-05,
+      "loss": 1.1375,
       "step": 905
     },
     {
-      "epoch": 0.024882590425970175,
+      "epoch": 0.025709421112372306,
       "grad_norm": 0.0,
-      "learning_rate": 1.6578225068618483e-05,
-      "loss": 1.1507,
+      "learning_rate": 1.7126654064272214e-05,
+      "loss": 1.189,
       "step": 906
     },
     {
-      "epoch": 0.024910054653813407,
+      "epoch": 0.025737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.6596523330283624e-05,
-      "loss": 1.2062,
+      "learning_rate": 1.7145557655954633e-05,
+      "loss": 1.0265,
       "step": 907
     },
     {
-      "epoch": 0.024937518881656642,
+      "epoch": 0.025766174801362087,
       "grad_norm": 0.0,
-      "learning_rate": 1.6614821591948768e-05,
-      "loss": 1.1064,
+      "learning_rate": 1.7164461247637053e-05,
+      "loss": 1.1796,
       "step": 908
     },
     {
-      "epoch": 0.024964983109499878,
+      "epoch": 0.02579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.663311985361391e-05,
-      "loss": 1.1493,
+      "learning_rate": 1.7183364839319473e-05,
+      "loss": 1.0213,
       "step": 909
     },
     {
-      "epoch": 0.02499244733734311,
+      "epoch": 0.025822928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 1.665141811527905e-05,
-      "loss": 1.1245,
+      "learning_rate": 1.7202268431001892e-05,
+      "loss": 1.0744,
       "step": 910
     },
     {
-      "epoch": 0.025019911565186344,
+      "epoch": 0.025851305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.666971637694419e-05,
-      "loss": 1.1414,
+      "learning_rate": 1.7221172022684312e-05,
+      "loss": 1.0767,
       "step": 911
     },
     {
-      "epoch": 0.02504737579302958,
+      "epoch": 0.025879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.6688014638609332e-05,
-      "loss": 1.1469,
+      "learning_rate": 1.724007561436673e-05,
+      "loss": 1.087,
       "step": 912
     },
     {
-      "epoch": 0.02507484002087281,
+      "epoch": 0.025908059023836548,
       "grad_norm": 0.0,
-      "learning_rate": 1.6706312900274476e-05,
-      "loss": 1.086,
+      "learning_rate": 1.725897920604915e-05,
+      "loss": 1.1349,
       "step": 913
     },
     {
-      "epoch": 0.025102304248716047,
+      "epoch": 0.025936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.6724611161939617e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.727788279773157e-05,
+      "loss": 1.0972,
       "step": 914
     },
     {
-      "epoch": 0.025129768476559282,
+      "epoch": 0.025964812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.6742909423604758e-05,
-      "loss": 1.0671,
+      "learning_rate": 1.729678638941399e-05,
+      "loss": 1.1146,
       "step": 915
     },
     {
-      "epoch": 0.025157232704402517,
+      "epoch": 0.025993189557321227,
       "grad_norm": 0.0,
-      "learning_rate": 1.67612076852699e-05,
-      "loss": 1.2155,
+      "learning_rate": 1.731568998109641e-05,
+      "loss": 0.9762,
       "step": 916
     },
     {
-      "epoch": 0.02518469693224575,
+      "epoch": 0.026021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.6779505946935043e-05,
-      "loss": 1.0885,
+      "learning_rate": 1.733459357277883e-05,
+      "loss": 1.1717,
       "step": 917
     },
     {
-      "epoch": 0.025212161160088984,
+      "epoch": 0.02604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.6797804208600184e-05,
-      "loss": 1.1504,
+      "learning_rate": 1.735349716446125e-05,
+      "loss": 1.0688,
       "step": 918
     },
     {
-      "epoch": 0.02523962538793222,
+      "epoch": 0.026078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.6816102470265325e-05,
-      "loss": 1.0737,
+      "learning_rate": 1.737240075614367e-05,
+      "loss": 1.0612,
       "step": 919
     },
     {
-      "epoch": 0.02526708961577545,
+      "epoch": 0.026106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.6834400731930466e-05,
-      "loss": 1.2266,
+      "learning_rate": 1.739130434782609e-05,
+      "loss": 1.0578,
       "step": 920
     },
     {
-      "epoch": 0.025294553843618686,
+      "epoch": 0.026135073779795688,
       "grad_norm": 0.0,
-      "learning_rate": 1.685269899359561e-05,
-      "loss": 1.1118,
+      "learning_rate": 1.741020793950851e-05,
+      "loss": 0.9924,
       "step": 921
     },
     {
-      "epoch": 0.02532201807146192,
+      "epoch": 0.02616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.687099725526075e-05,
-      "loss": 1.1451,
+      "learning_rate": 1.7429111531190928e-05,
+      "loss": 1.1282,
       "step": 922
     },
     {
-      "epoch": 0.025349482299305157,
+      "epoch": 0.02619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.6889295516925896e-05,
-      "loss": 1.1151,
+      "learning_rate": 1.7448015122873348e-05,
+      "loss": 1.0682,
       "step": 923
     },
     {
-      "epoch": 0.02537694652714839,
+      "epoch": 0.026220204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.6907593778591037e-05,
-      "loss": 1.0895,
+      "learning_rate": 1.7466918714555767e-05,
+      "loss": 1.1993,
       "step": 924
     },
     {
-      "epoch": 0.025404410754991624,
+      "epoch": 0.026248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.6925892040256177e-05,
-      "loss": 1.1313,
+      "learning_rate": 1.7485822306238187e-05,
+      "loss": 1.0298,
       "step": 925
     },
     {
-      "epoch": 0.02543187498283486,
+      "epoch": 0.02627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.694419030192132e-05,
-      "loss": 1.1009,
+      "learning_rate": 1.7504725897920607e-05,
+      "loss": 1.0383,
       "step": 926
     },
     {
-      "epoch": 0.02545933921067809,
+      "epoch": 0.02630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.6962488563586463e-05,
-      "loss": 1.1416,
+      "learning_rate": 1.7523629489603026e-05,
+      "loss": 1.0659,
       "step": 927
     },
     {
-      "epoch": 0.025486803438521326,
+      "epoch": 0.02633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.6980786825251604e-05,
-      "loss": 1.1048,
+      "learning_rate": 1.7542533081285446e-05,
+      "loss": 1.0561,
       "step": 928
     },
     {
-      "epoch": 0.02551426766636456,
+      "epoch": 0.026362088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.6999085086916745e-05,
-      "loss": 1.097,
+      "learning_rate": 1.7561436672967866e-05,
+      "loss": 1.004,
       "step": 929
     },
     {
-      "epoch": 0.025541731894207793,
+      "epoch": 0.026390465380249715,
       "grad_norm": 0.0,
-      "learning_rate": 1.7017383348581885e-05,
-      "loss": 1.1074,
+      "learning_rate": 1.7580340264650285e-05,
+      "loss": 1.0173,
       "step": 930
     },
     {
-      "epoch": 0.025569196122051028,
+      "epoch": 0.02641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.7035681610247026e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.7599243856332705e-05,
+      "loss": 1.0912,
       "step": 931
     },
     {
-      "epoch": 0.025596660349894263,
+      "epoch": 0.0264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.705397987191217e-05,
-      "loss": 1.2278,
+      "learning_rate": 1.7618147448015125e-05,
+      "loss": 0.9771,
       "step": 932
     },
     {
-      "epoch": 0.0256241245777375,
+      "epoch": 0.02647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.707227813357731e-05,
-      "loss": 1.1425,
+      "learning_rate": 1.7637051039697544e-05,
+      "loss": 0.9924,
       "step": 933
     },
     {
-      "epoch": 0.02565158880558073,
+      "epoch": 0.026503972758229286,
       "grad_norm": 0.0,
-      "learning_rate": 1.7090576395242453e-05,
-      "loss": 1.165,
+      "learning_rate": 1.7655954631379964e-05,
+      "loss": 1.0828,
       "step": 934
     },
     {
-      "epoch": 0.025679053033423965,
+      "epoch": 0.026532349602724176,
       "grad_norm": 0.0,
-      "learning_rate": 1.7108874656907593e-05,
-      "loss": 1.1035,
+      "learning_rate": 1.7674858223062384e-05,
+      "loss": 1.1503,
       "step": 935
     },
     {
-      "epoch": 0.0257065172612672,
+      "epoch": 0.02656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.7127172918572738e-05,
-      "loss": 1.1823,
+      "learning_rate": 1.7693761814744803e-05,
+      "loss": 1.1491,
       "step": 936
     },
     {
-      "epoch": 0.025733981489110432,
+      "epoch": 0.02658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.714547118023788e-05,
-      "loss": 1.212,
+      "learning_rate": 1.7712665406427223e-05,
+      "loss": 1.1957,
       "step": 937
     },
     {
-      "epoch": 0.025761445716953667,
+      "epoch": 0.026617480136208852,
       "grad_norm": 0.0,
-      "learning_rate": 1.716376944190302e-05,
-      "loss": 1.1843,
+      "learning_rate": 1.7731568998109643e-05,
+      "loss": 1.0767,
       "step": 938
     },
     {
-      "epoch": 0.025788909944796903,
+      "epoch": 0.026645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.718206770356816e-05,
-      "loss": 1.2084,
+      "learning_rate": 1.7750472589792062e-05,
+      "loss": 1.1028,
       "step": 939
     },
     {
-      "epoch": 0.025816374172640138,
+      "epoch": 0.026674233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.7200365965233305e-05,
-      "loss": 1.1781,
+      "learning_rate": 1.7769376181474482e-05,
+      "loss": 1.0693,
       "step": 940
     },
     {
-      "epoch": 0.02584383840048337,
+      "epoch": 0.02670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.7218664226898446e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.77882797731569e-05,
+      "loss": 1.0373,
       "step": 941
     },
     {
-      "epoch": 0.025871302628326605,
+      "epoch": 0.026730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.723696248856359e-05,
-      "loss": 1.1548,
+      "learning_rate": 1.780718336483932e-05,
+      "loss": 1.2275,
       "step": 942
     },
     {
-      "epoch": 0.02589876685616984,
+      "epoch": 0.026759364358683313,
       "grad_norm": 0.0,
-      "learning_rate": 1.725526075022873e-05,
-      "loss": 1.1386,
+      "learning_rate": 1.782608695652174e-05,
+      "loss": 1.1487,
       "step": 943
     },
     {
-      "epoch": 0.025926231084013072,
+      "epoch": 0.026787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.7273559011893872e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.784499054820416e-05,
+      "loss": 1.2285,
       "step": 944
     },
     {
-      "epoch": 0.025953695311856307,
+      "epoch": 0.026816118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.7291857273559013e-05,
-      "loss": 1.1835,
+      "learning_rate": 1.786389413988658e-05,
+      "loss": 1.0129,
       "step": 945
     },
     {
-      "epoch": 0.025981159539699542,
+      "epoch": 0.026844494892167992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7310155535224157e-05,
-      "loss": 1.1381,
+      "learning_rate": 1.7882797731569e-05,
+      "loss": 1.2041,
       "step": 946
     },
     {
-      "epoch": 0.026008623767542774,
+      "epoch": 0.026872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.7328453796889298e-05,
-      "loss": 1.0842,
+      "learning_rate": 1.790170132325142e-05,
+      "loss": 1.0586,
       "step": 947
     },
     {
-      "epoch": 0.02603608799538601,
+      "epoch": 0.026901248581157774,
       "grad_norm": 0.0,
-      "learning_rate": 1.734675205855444e-05,
-      "loss": 1.1196,
+      "learning_rate": 1.792060491493384e-05,
+      "loss": 1.142,
       "step": 948
     },
     {
-      "epoch": 0.026063552223229244,
+      "epoch": 0.026929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 1.736505032021958e-05,
-      "loss": 1.1144,
+      "learning_rate": 1.793950850661626e-05,
+      "loss": 1.1145,
       "step": 949
     },
     {
-      "epoch": 0.02609101645107248,
+      "epoch": 0.02695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.738334858188472e-05,
-      "loss": 1.1276,
+      "learning_rate": 1.795841209829868e-05,
+      "loss": 1.0354,
       "step": 950
     },
     {
-      "epoch": 0.02611848067891571,
+      "epoch": 0.026986379114642453,
       "grad_norm": 0.0,
-      "learning_rate": 1.7401646843549865e-05,
-      "loss": 1.1225,
+      "learning_rate": 1.7977315689981098e-05,
+      "loss": 1.1082,
       "step": 951
     },
     {
-      "epoch": 0.026145944906758947,
+      "epoch": 0.027014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7419945105215006e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.7996219281663518e-05,
+      "loss": 1.1336,
       "step": 952
     },
     {
-      "epoch": 0.026173409134602182,
+      "epoch": 0.027043132803632235,
       "grad_norm": 0.0,
-      "learning_rate": 1.7438243366880147e-05,
-      "loss": 1.0299,
+      "learning_rate": 1.8015122873345938e-05,
+      "loss": 1.1254,
       "step": 953
     },
     {
-      "epoch": 0.026200873362445413,
+      "epoch": 0.02707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.7456541628545288e-05,
-      "loss": 1.1442,
+      "learning_rate": 1.8034026465028357e-05,
+      "loss": 1.106,
       "step": 954
     },
     {
-      "epoch": 0.02622833759028865,
+      "epoch": 0.02709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.7474839890210432e-05,
-      "loss": 1.222,
+      "learning_rate": 1.8052930056710777e-05,
+      "loss": 1.0362,
       "step": 955
     },
     {
-      "epoch": 0.026255801818131884,
+      "epoch": 0.027128263337116914,
       "grad_norm": 0.0,
-      "learning_rate": 1.7493138151875573e-05,
-      "loss": 1.101,
+      "learning_rate": 1.8071833648393197e-05,
+      "loss": 1.1838,
       "step": 956
     },
     {
-      "epoch": 0.02628326604597512,
+      "epoch": 0.027156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.7511436413540714e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.8090737240075616e-05,
+      "loss": 1.1169,
       "step": 957
     },
     {
-      "epoch": 0.02631073027381835,
+      "epoch": 0.027185017026106695,
       "grad_norm": 0.0,
-      "learning_rate": 1.7529734675205855e-05,
-      "loss": 1.1752,
+      "learning_rate": 1.8109640831758036e-05,
+      "loss": 1.0865,
       "step": 958
     },
     {
-      "epoch": 0.026338194501661586,
+      "epoch": 0.02721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.7548032936871e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.8128544423440455e-05,
+      "loss": 1.2031,
       "step": 959
     },
     {
-      "epoch": 0.02636565872950482,
+      "epoch": 0.02724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.756633119853614e-05,
-      "loss": 1.0997,
+      "learning_rate": 1.8147448015122875e-05,
+      "loss": 1.0189,
       "step": 960
     },
     {
-      "epoch": 0.026393122957348053,
+      "epoch": 0.027270147559591375,
       "grad_norm": 0.0,
-      "learning_rate": 1.758462946020128e-05,
-      "loss": 1.23,
+      "learning_rate": 1.8166351606805295e-05,
+      "loss": 1.0428,
       "step": 961
     },
     {
-      "epoch": 0.026420587185191288,
+      "epoch": 0.027298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.7602927721866425e-05,
-      "loss": 1.0622,
+      "learning_rate": 1.8185255198487714e-05,
+      "loss": 1.1097,
       "step": 962
     },
     {
-      "epoch": 0.026448051413034523,
+      "epoch": 0.027326901248581156,
       "grad_norm": 0.0,
-      "learning_rate": 1.7621225983531566e-05,
-      "loss": 1.1304,
+      "learning_rate": 1.8204158790170134e-05,
+      "loss": 1.0812,
       "step": 963
     },
     {
-      "epoch": 0.026475515640877755,
+      "epoch": 0.02735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.7639524245196707e-05,
-      "loss": 1.1748,
+      "learning_rate": 1.8223062381852554e-05,
+      "loss": 1.1042,
       "step": 964
     },
     {
-      "epoch": 0.02650297986872099,
+      "epoch": 0.02738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.765782250686185e-05,
-      "loss": 1.1423,
+      "learning_rate": 1.8241965973534973e-05,
+      "loss": 1.2125,
       "step": 965
     },
     {
-      "epoch": 0.026530444096564226,
+      "epoch": 0.027412031782065836,
       "grad_norm": 0.0,
-      "learning_rate": 1.7676120768526993e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.8260869565217393e-05,
+      "loss": 1.1216,
       "step": 966
     },
     {
-      "epoch": 0.02655790832440746,
+      "epoch": 0.027440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.7694419030192133e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.8279773156899813e-05,
+      "loss": 1.1035,
       "step": 967
     },
     {
-      "epoch": 0.026585372552250693,
+      "epoch": 0.027468785471055617,
       "grad_norm": 0.0,
-      "learning_rate": 1.7712717291857274e-05,
-      "loss": 1.1864,
+      "learning_rate": 1.8298676748582232e-05,
+      "loss": 1.1222,
       "step": 968
     },
     {
-      "epoch": 0.026612836780093928,
+      "epoch": 0.02749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.773101555352242e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.8317580340264652e-05,
+      "loss": 1.1665,
       "step": 969
     },
     {
-      "epoch": 0.026640301007937163,
+      "epoch": 0.027525539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.774931381518756e-05,
-      "loss": 1.1076,
+      "learning_rate": 1.8336483931947072e-05,
+      "loss": 1.2081,
       "step": 970
     },
     {
-      "epoch": 0.026667765235780395,
+      "epoch": 0.027553916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.77676120768527e-05,
-      "loss": 1.1355,
+      "learning_rate": 1.835538752362949e-05,
+      "loss": 1.1211,
       "step": 971
     },
     {
-      "epoch": 0.02669522946362363,
+      "epoch": 0.027582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.778591033851784e-05,
-      "loss": 1.1296,
+      "learning_rate": 1.837429111531191e-05,
+      "loss": 1.1106,
       "step": 972
     },
     {
-      "epoch": 0.026722693691466865,
+      "epoch": 0.027610669693530078,
       "grad_norm": 0.0,
-      "learning_rate": 1.7804208600182982e-05,
-      "loss": 1.1343,
+      "learning_rate": 1.839319470699433e-05,
+      "loss": 1.1011,
       "step": 973
     },
     {
-      "epoch": 0.026750157919310097,
+      "epoch": 0.027639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.7822506861848127e-05,
-      "loss": 1.1607,
+      "learning_rate": 1.841209829867675e-05,
+      "loss": 1.162,
       "step": 974
     },
     {
-      "epoch": 0.026777622147153332,
+      "epoch": 0.027667423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.7840805123513268e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.843100189035917e-05,
+      "loss": 1.0762,
       "step": 975
     },
     {
-      "epoch": 0.026805086374996567,
+      "epoch": 0.027695800227014757,
       "grad_norm": 0.0,
-      "learning_rate": 1.785910338517841e-05,
-      "loss": 1.098,
+      "learning_rate": 1.844990548204159e-05,
+      "loss": 1.0404,
       "step": 976
     },
     {
-      "epoch": 0.026832550602839803,
+      "epoch": 0.027724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.787740164684355e-05,
-      "loss": 1.1765,
+      "learning_rate": 1.846880907372401e-05,
+      "loss": 1.0825,
       "step": 977
     },
     {
-      "epoch": 0.026860014830683034,
+      "epoch": 0.02775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7895699908508694e-05,
-      "loss": 1.17,
+      "learning_rate": 1.848771266540643e-05,
+      "loss": 1.077,
       "step": 978
     },
     {
-      "epoch": 0.02688747905852627,
+      "epoch": 0.027780930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 1.7913998170173835e-05,
-      "loss": 1.1207,
+      "learning_rate": 1.850661625708885e-05,
+      "loss": 1.1219,
       "step": 979
     },
     {
-      "epoch": 0.026914943286369505,
+      "epoch": 0.027809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.7932296431838976e-05,
-      "loss": 1.079,
+      "learning_rate": 1.8525519848771268e-05,
+      "loss": 1.14,
       "step": 980
     },
     {
-      "epoch": 0.026942407514212736,
+      "epoch": 0.027837684449489218,
       "grad_norm": 0.0,
-      "learning_rate": 1.7950594693504117e-05,
-      "loss": 1.1613,
+      "learning_rate": 1.8544423440453688e-05,
+      "loss": 1.1385,
       "step": 981
     },
     {
-      "epoch": 0.02696987174205597,
+      "epoch": 0.02786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.796889295516926e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.8563327032136108e-05,
+      "loss": 1.1255,
       "step": 982
     },
     {
-      "epoch": 0.026997335969899207,
+      "epoch": 0.027894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7987191216834402e-05,
-      "loss": 1.1061,
+      "learning_rate": 1.8582230623818527e-05,
+      "loss": 1.1345,
       "step": 983
     },
     {
-      "epoch": 0.027024800197742442,
+      "epoch": 0.027922814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.8005489478499546e-05,
-      "loss": 1.2192,
+      "learning_rate": 1.8601134215500947e-05,
+      "loss": 1.1709,
       "step": 984
     },
     {
-      "epoch": 0.027052264425585674,
+      "epoch": 0.027951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.8023787740164687e-05,
-      "loss": 1.1007,
+      "learning_rate": 1.8620037807183367e-05,
+      "loss": 1.0066,
       "step": 985
     },
     {
-      "epoch": 0.02707972865342891,
+      "epoch": 0.02797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.8042086001829828e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.8638941398865786e-05,
+      "loss": 1.1105,
       "step": 986
     },
     {
-      "epoch": 0.027107192881272144,
+      "epoch": 0.02800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.806038426349497e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.8657844990548206e-05,
+      "loss": 1.1282,
       "step": 987
     },
     {
-      "epoch": 0.027134657109115376,
+      "epoch": 0.02803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.8078682525160113e-05,
-      "loss": 1.1283,
+      "learning_rate": 1.8676748582230626e-05,
+      "loss": 1.0744,
       "step": 988
     },
     {
-      "epoch": 0.02716212133695861,
+      "epoch": 0.028064699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 1.8096980786825254e-05,
-      "loss": 1.1747,
+      "learning_rate": 1.8695652173913045e-05,
+      "loss": 1.0988,
       "step": 989
     },
     {
-      "epoch": 0.027189585564801846,
+      "epoch": 0.028093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8115279048490395e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.8714555765595465e-05,
+      "loss": 1.0661,
       "step": 990
     },
     {
-      "epoch": 0.027217049792645078,
+      "epoch": 0.02812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.8133577310155536e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.8733459357277885e-05,
+      "loss": 1.1689,
       "step": 991
     },
     {
-      "epoch": 0.027244514020488313,
+      "epoch": 0.02814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8151875571820677e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.8752362948960304e-05,
+      "loss": 1.1183,
       "step": 992
     },
     {
-      "epoch": 0.02727197824833155,
+      "epoch": 0.02817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.817017383348582e-05,
-      "loss": 1.0864,
+      "learning_rate": 1.8771266540642724e-05,
+      "loss": 0.959,
       "step": 993
     },
     {
-      "epoch": 0.027299442476174784,
+      "epoch": 0.028206583427922816,
       "grad_norm": 0.0,
-      "learning_rate": 1.8188472095150962e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.8790170132325144e-05,
+      "loss": 1.0887,
       "step": 994
     },
     {
-      "epoch": 0.027326906704018016,
+      "epoch": 0.028234960272417706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8206770356816103e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.8809073724007563e-05,
+      "loss": 1.1091,
       "step": 995
     },
     {
-      "epoch": 0.02735437093186125,
+      "epoch": 0.0282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.8225068618481244e-05,
-      "loss": 1.1177,
+      "learning_rate": 1.8827977315689983e-05,
+      "loss": 1.0734,
       "step": 996
     },
     {
-      "epoch": 0.027381835159704486,
+      "epoch": 0.02829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.8243366880146388e-05,
-      "loss": 1.1354,
+      "learning_rate": 1.8846880907372402e-05,
+      "loss": 1.1206,
       "step": 997
     },
     {
-      "epoch": 0.027409299387547718,
+      "epoch": 0.028320090805902382,
       "grad_norm": 0.0,
-      "learning_rate": 1.826166514181153e-05,
-      "loss": 1.137,
+      "learning_rate": 1.8865784499054822e-05,
+      "loss": 1.2136,
       "step": 998
     },
     {
-      "epoch": 0.027436763615390953,
+      "epoch": 0.028348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.827996340347667e-05,
-      "loss": 1.1914,
+      "learning_rate": 1.8884688090737242e-05,
+      "loss": 1.08,
       "step": 999
     },
     {
-      "epoch": 0.027464227843234188,
+      "epoch": 0.028376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.829826166514181e-05,
-      "loss": 1.1512,
+      "learning_rate": 1.890359168241966e-05,
+      "loss": 1.0858,
       "step": 1000
     },
     {
-      "epoch": 0.027491692071077423,
+      "epoch": 0.02840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8316559926806955e-05,
-      "loss": 1.0778,
+      "learning_rate": 1.892249527410208e-05,
+      "loss": 1.1063,
       "step": 1001
     },
     {
-      "epoch": 0.027519156298920655,
+      "epoch": 0.028433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.8334858188472096e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.89413988657845e-05,
+      "loss": 1.0768,
       "step": 1002
     },
     {
-      "epoch": 0.02754662052676389,
+      "epoch": 0.028461975028376843,
       "grad_norm": 0.0,
-      "learning_rate": 1.835315645013724e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.896030245746692e-05,
+      "loss": 1.159,
       "step": 1003
     },
     {
-      "epoch": 0.027574084754607125,
+      "epoch": 0.028490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.837145471180238e-05,
-      "loss": 1.1362,
+      "learning_rate": 1.897920604914934e-05,
+      "loss": 1.2452,
       "step": 1004
     },
     {
-      "epoch": 0.027601548982450357,
+      "epoch": 0.028518728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.8389752973467522e-05,
-      "loss": 1.1913,
+      "learning_rate": 1.899810964083176e-05,
+      "loss": 1.1211,
       "step": 1005
     },
     {
-      "epoch": 0.027629013210293592,
+      "epoch": 0.028547105561861522,
       "grad_norm": 0.0,
-      "learning_rate": 1.8408051235132663e-05,
-      "loss": 1.123,
+      "learning_rate": 1.901701323251418e-05,
+      "loss": 1.0608,
       "step": 1006
     },
     {
-      "epoch": 0.027656477438136828,
+      "epoch": 0.028575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.8426349496797808e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.90359168241966e-05,
+      "loss": 1.0038,
       "step": 1007
     },
     {
-      "epoch": 0.02768394166598006,
+      "epoch": 0.028603859250851304,
       "grad_norm": 0.0,
-      "learning_rate": 1.844464775846295e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.905482041587902e-05,
+      "loss": 1.1209,
       "step": 1008
     },
     {
-      "epoch": 0.027711405893823295,
+      "epoch": 0.028632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.846294602012809e-05,
-      "loss": 1.0929,
+      "learning_rate": 1.907372400756144e-05,
+      "loss": 1.1105,
       "step": 1009
     },
     {
-      "epoch": 0.02773887012166653,
+      "epoch": 0.02866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.848124428179323e-05,
-      "loss": 1.112,
+      "learning_rate": 1.9092627599243858e-05,
+      "loss": 1.0688,
       "step": 1010
     },
     {
-      "epoch": 0.027766334349509765,
+      "epoch": 0.028688989784335983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8499542543458375e-05,
-      "loss": 1.1685,
+      "learning_rate": 1.9111531190926278e-05,
+      "loss": 1.1584,
       "step": 1011
     },
     {
-      "epoch": 0.027793798577352997,
+      "epoch": 0.028717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.8517840805123516e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.9130434782608697e-05,
+      "loss": 0.9859,
       "step": 1012
     },
     {
-      "epoch": 0.027821262805196232,
+      "epoch": 0.028745743473325765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8536139066788657e-05,
-      "loss": 1.1862,
+      "learning_rate": 1.9149338374291117e-05,
+      "loss": 1.0729,
       "step": 1013
     },
     {
-      "epoch": 0.027848727033039467,
+      "epoch": 0.02877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8554437328453798e-05,
-      "loss": 1.2511,
+      "learning_rate": 1.9168241965973537e-05,
+      "loss": 1.1955,
       "step": 1014
     },
     {
-      "epoch": 0.0278761912608827,
+      "epoch": 0.02880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.857273559011894e-05,
-      "loss": 1.0842,
+      "learning_rate": 1.9187145557655956e-05,
+      "loss": 1.0184,
       "step": 1015
     },
     {
-      "epoch": 0.027903655488725934,
+      "epoch": 0.028830874006810444,
       "grad_norm": 0.0,
-      "learning_rate": 1.8591033851784083e-05,
-      "loss": 1.0946,
+      "learning_rate": 1.9206049149338376e-05,
+      "loss": 1.0976,
       "step": 1016
     },
     {
-      "epoch": 0.02793111971656917,
+      "epoch": 0.028859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.8609332113449224e-05,
-      "loss": 1.168,
+      "learning_rate": 1.9224952741020796e-05,
+      "loss": 0.9903,
       "step": 1017
     },
     {
-      "epoch": 0.027958583944412405,
+      "epoch": 0.028887627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.8627630375114365e-05,
-      "loss": 1.1642,
+      "learning_rate": 1.9243856332703215e-05,
+      "loss": 1.0305,
       "step": 1018
     },
     {
-      "epoch": 0.027986048172255636,
+      "epoch": 0.02891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.8645928636779506e-05,
-      "loss": 1.1227,
+      "learning_rate": 1.9262759924385635e-05,
+      "loss": 1.0703,
       "step": 1019
     },
     {
-      "epoch": 0.02801351240009887,
+      "epoch": 0.02894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.866422689844465e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.9281663516068055e-05,
+      "loss": 1.1443,
       "step": 1020
     },
     {
-      "epoch": 0.028040976627942107,
+      "epoch": 0.028972758229284905,
       "grad_norm": 0.0,
-      "learning_rate": 1.868252516010979e-05,
-      "loss": 1.1178,
+      "learning_rate": 1.9300567107750474e-05,
+      "loss": 1.2634,
       "step": 1021
     },
     {
-      "epoch": 0.02806844085578534,
+      "epoch": 0.029001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.8700823421774932e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.9319470699432894e-05,
+      "loss": 1.1668,
       "step": 1022
     },
     {
-      "epoch": 0.028095905083628574,
+      "epoch": 0.029029511918274686,
       "grad_norm": 0.0,
-      "learning_rate": 1.8719121683440076e-05,
-      "loss": 1.099,
+      "learning_rate": 1.9338374291115314e-05,
+      "loss": 1.1374,
       "step": 1023
     },
     {
-      "epoch": 0.02812336931147181,
+      "epoch": 0.02905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.8737419945105217e-05,
-      "loss": 1.1772,
+      "learning_rate": 1.9357277882797733e-05,
+      "loss": 1.2186,
       "step": 1024
     },
     {
-      "epoch": 0.02815083353931504,
+      "epoch": 0.02908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.8755718206770358e-05,
-      "loss": 1.14,
+      "learning_rate": 1.9376181474480153e-05,
+      "loss": 1.0564,
       "step": 1025
     },
     {
-      "epoch": 0.028178297767158276,
+      "epoch": 0.029114642451759366,
       "grad_norm": 0.0,
-      "learning_rate": 1.8774016468435502e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.9395085066162573e-05,
+      "loss": 1.171,
       "step": 1026
     },
     {
-      "epoch": 0.02820576199500151,
+      "epoch": 0.029143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.8792314730100643e-05,
-      "loss": 1.1272,
+      "learning_rate": 1.9413988657844992e-05,
+      "loss": 1.0893,
       "step": 1027
     },
     {
-      "epoch": 0.028233226222844746,
+      "epoch": 0.029171396140749147,
       "grad_norm": 0.0,
-      "learning_rate": 1.8810612991765784e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.9432892249527412e-05,
+      "loss": 1.1413,
       "step": 1028
     },
     {
-      "epoch": 0.028260690450687978,
+      "epoch": 0.02919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.8828911253430925e-05,
-      "loss": 1.1314,
+      "learning_rate": 1.945179584120983e-05,
+      "loss": 1.1973,
       "step": 1029
     },
     {
-      "epoch": 0.028288154678531213,
+      "epoch": 0.029228149829738932,
       "grad_norm": 0.0,
-      "learning_rate": 1.884720951509607e-05,
-      "loss": 1.1808,
+      "learning_rate": 1.947069943289225e-05,
+      "loss": 1.0253,
       "step": 1030
     },
     {
-      "epoch": 0.02831561890637445,
+      "epoch": 0.029256526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.886550777676121e-05,
-      "loss": 1.1776,
+      "learning_rate": 1.948960302457467e-05,
+      "loss": 1.1682,
       "step": 1031
     },
     {
-      "epoch": 0.02834308313421768,
+      "epoch": 0.029284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.888380603842635e-05,
-      "loss": 1.1423,
+      "learning_rate": 1.950850661625709e-05,
+      "loss": 1.1101,
       "step": 1032
     },
     {
-      "epoch": 0.028370547362060915,
+      "epoch": 0.029313280363223608,
       "grad_norm": 0.0,
-      "learning_rate": 1.8902104300091492e-05,
-      "loss": 1.0587,
+      "learning_rate": 1.952741020793951e-05,
+      "loss": 1.0394,
       "step": 1033
     },
     {
-      "epoch": 0.02839801158990415,
+      "epoch": 0.029341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.8920402561756633e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.954631379962193e-05,
+      "loss": 1.1735,
       "step": 1034
     },
     {
-      "epoch": 0.028425475817747382,
+      "epoch": 0.029370034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.8938700823421777e-05,
-      "loss": 1.077,
+      "learning_rate": 1.956521739130435e-05,
+      "loss": 1.1027,
       "step": 1035
     },
     {
-      "epoch": 0.028452940045590618,
+      "epoch": 0.029398410896708287,
       "grad_norm": 0.0,
-      "learning_rate": 1.8956999085086918e-05,
-      "loss": 1.1567,
+      "learning_rate": 1.958412098298677e-05,
+      "loss": 1.103,
       "step": 1036
     },
     {
-      "epoch": 0.028480404273433853,
+      "epoch": 0.029426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.897529734675206e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.960302457466919e-05,
+      "loss": 1.0724,
       "step": 1037
     },
     {
-      "epoch": 0.028507868501277088,
+      "epoch": 0.02945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.89935956084172e-05,
-      "loss": 1.2169,
+      "learning_rate": 1.962192816635161e-05,
+      "loss": 1.0233,
       "step": 1038
     },
     {
-      "epoch": 0.02853533272912032,
+      "epoch": 0.029483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.9011893870082344e-05,
-      "loss": 1.1341,
+      "learning_rate": 1.9640831758034028e-05,
+      "loss": 1.1242,
       "step": 1039
     },
     {
-      "epoch": 0.028562796956963555,
+      "epoch": 0.029511918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9030192131747485e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.9659735349716448e-05,
+      "loss": 1.0401,
       "step": 1040
     },
     {
-      "epoch": 0.02859026118480679,
+      "epoch": 0.029540295119182748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9048490393412626e-05,
-      "loss": 1.1748,
+      "learning_rate": 1.9678638941398867e-05,
+      "loss": 1.0078,
       "step": 1041
     },
     {
-      "epoch": 0.028617725412650022,
+      "epoch": 0.02956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9066788655077767e-05,
-      "loss": 1.1563,
+      "learning_rate": 1.9697542533081287e-05,
+      "loss": 1.0826,
       "step": 1042
     },
     {
-      "epoch": 0.028645189640493257,
+      "epoch": 0.02959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.908508691674291e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.9716446124763707e-05,
+      "loss": 1.1392,
       "step": 1043
     },
     {
-      "epoch": 0.028672653868336492,
+      "epoch": 0.029625425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.9103385178408052e-05,
-      "loss": 1.1147,
+      "learning_rate": 1.9735349716446126e-05,
+      "loss": 0.906,
       "step": 1044
     },
     {
-      "epoch": 0.028700118096179728,
+      "epoch": 0.029653802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.9121683440073197e-05,
-      "loss": 1.1955,
+      "learning_rate": 1.9754253308128546e-05,
+      "loss": 1.1176,
       "step": 1045
     },
     {
-      "epoch": 0.02872758232402296,
+      "epoch": 0.02968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.9139981701738338e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.9773156899810966e-05,
+      "loss": 1.1241,
       "step": 1046
     },
     {
-      "epoch": 0.028755046551866194,
+      "epoch": 0.0297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.915827996340348e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.9792060491493385e-05,
+      "loss": 1.1982,
       "step": 1047
     },
     {
-      "epoch": 0.02878251077970943,
+      "epoch": 0.02973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.917657822506862e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9810964083175805e-05,
+      "loss": 1.0876,
       "step": 1048
     },
     {
-      "epoch": 0.02880997500755266,
+      "epoch": 0.029767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.9194876486733764e-05,
-      "loss": 1.1683,
+      "learning_rate": 1.9829867674858225e-05,
+      "loss": 1.1379,
       "step": 1049
     },
     {
-      "epoch": 0.028837439235395897,
+      "epoch": 0.029795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9213174748398905e-05,
-      "loss": 1.0625,
+      "learning_rate": 1.9848771266540644e-05,
+      "loss": 1.0798,
       "step": 1050
     },
     {
-      "epoch": 0.028864903463239132,
+      "epoch": 0.02982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9231473010064046e-05,
-      "loss": 1.1384,
+      "learning_rate": 1.9867674858223064e-05,
+      "loss": 1.1047,
       "step": 1051
     },
     {
-      "epoch": 0.028892367691082364,
+      "epoch": 0.02985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9249771271729187e-05,
-      "loss": 1.2318,
+      "learning_rate": 1.9886578449905484e-05,
+      "loss": 1.0774,
       "step": 1052
     },
     {
-      "epoch": 0.0289198319189256,
+      "epoch": 0.02988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.9268069533394327e-05,
-      "loss": 1.1513,
+      "learning_rate": 1.9905482041587903e-05,
+      "loss": 1.1715,
       "step": 1053
     },
     {
-      "epoch": 0.028947296146768834,
+      "epoch": 0.029909194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9286367795059472e-05,
-      "loss": 1.1262,
+      "learning_rate": 1.9924385633270323e-05,
+      "loss": 1.1429,
       "step": 1054
     },
     {
-      "epoch": 0.02897476037461207,
+      "epoch": 0.029937570942111236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9304666056724613e-05,
-      "loss": 1.0817,
+      "learning_rate": 1.9943289224952743e-05,
+      "loss": 1.1236,
       "step": 1055
     },
     {
-      "epoch": 0.0290022246024553,
+      "epoch": 0.02996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9322964318389754e-05,
-      "loss": 1.223,
+      "learning_rate": 1.9962192816635162e-05,
+      "loss": 1.0279,
       "step": 1056
     },
     {
-      "epoch": 0.029029688830298536,
+      "epoch": 0.02999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.9341262580054895e-05,
-      "loss": 1.059,
+      "learning_rate": 1.9981096408317582e-05,
+      "loss": 1.1338,
       "step": 1057
     },
     {
-      "epoch": 0.02905715305814177,
+      "epoch": 0.030022701475595912,
       "grad_norm": 0.0,
-      "learning_rate": 1.935956084172004e-05,
-      "loss": 1.2723,
+      "learning_rate": 2e-05,
+      "loss": 1.3414,
       "step": 1058
     },
     {
-      "epoch": 0.029084617285985003,
+      "epoch": 0.030051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.937785910338518e-05,
-      "loss": 1.1831,
+      "learning_rate": 1.9999999957764777e-05,
+      "loss": 1.0999,
       "step": 1059
     },
     {
-      "epoch": 0.02911208151382824,
+      "epoch": 0.030079455164585697,
       "grad_norm": 0.0,
-      "learning_rate": 1.939615736505032e-05,
-      "loss": 1.1887,
+      "learning_rate": 1.9999999831059104e-05,
+      "loss": 1.0542,
       "step": 1060
     },
     {
-      "epoch": 0.029139545741671474,
+      "epoch": 0.03010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.941445562671546e-05,
-      "loss": 1.1568,
+      "learning_rate": 1.999999961988298e-05,
+      "loss": 1.1465,
       "step": 1061
     },
     {
-      "epoch": 0.02916700996951471,
+      "epoch": 0.030136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.9432753888380606e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.999999932423641e-05,
+      "loss": 1.0553,
       "step": 1062
     },
     {
-      "epoch": 0.02919447419735794,
+      "epoch": 0.030164585698070373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9451052150045747e-05,
-      "loss": 1.198,
+      "learning_rate": 1.9999998944119402e-05,
+      "loss": 1.0979,
       "step": 1063
     },
     {
-      "epoch": 0.029221938425201176,
+      "epoch": 0.030192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.946935041171089e-05,
-      "loss": 1.1735,
+      "learning_rate": 1.9999998479531948e-05,
+      "loss": 1.1504,
       "step": 1064
     },
     {
-      "epoch": 0.02924940265304441,
+      "epoch": 0.030221339387060158,
       "grad_norm": 0.0,
-      "learning_rate": 1.9487648673376032e-05,
-      "loss": 1.103,
+      "learning_rate": 1.9999997930474058e-05,
+      "loss": 1.1111,
       "step": 1065
     },
     {
-      "epoch": 0.029276866880887643,
+      "epoch": 0.030249716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 1.9505946935041173e-05,
-      "loss": 1.1534,
+      "learning_rate": 1.9999997296945736e-05,
+      "loss": 1.1091,
       "step": 1066
     },
     {
-      "epoch": 0.029304331108730878,
+      "epoch": 0.030278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.9524245196706314e-05,
-      "loss": 1.0709,
+      "learning_rate": 1.9999996578946986e-05,
+      "loss": 0.903,
       "step": 1067
     },
     {
-      "epoch": 0.029331795336574113,
+      "epoch": 0.030306469920544834,
       "grad_norm": 0.0,
-      "learning_rate": 1.9542543458371458e-05,
-      "loss": 1.1202,
+      "learning_rate": 1.999999577647782e-05,
+      "loss": 1.084,
       "step": 1068
     },
     {
-      "epoch": 0.029359259564417345,
+      "epoch": 0.030334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.95608417200366e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.9999994889538238e-05,
+      "loss": 1.1923,
       "step": 1069
     },
     {
-      "epoch": 0.02938672379226058,
+      "epoch": 0.03036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.957913998170174e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.999999391812825e-05,
+      "loss": 1.0813,
       "step": 1070
     },
     {
-      "epoch": 0.029414188020103815,
+      "epoch": 0.030391600454029513,
       "grad_norm": 0.0,
-      "learning_rate": 1.959743824336688e-05,
-      "loss": 1.1808,
+      "learning_rate": 1.999999286224786e-05,
+      "loss": 1.0852,
       "step": 1071
     },
     {
-      "epoch": 0.02944165224794705,
+      "epoch": 0.030419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9615736505032025e-05,
-      "loss": 1.0806,
+      "learning_rate": 1.9999991721897084e-05,
+      "loss": 0.9843,
       "step": 1072
     },
     {
-      "epoch": 0.029469116475790282,
+      "epoch": 0.030448354143019295,
       "grad_norm": 0.0,
-      "learning_rate": 1.9634034766697166e-05,
-      "loss": 1.1195,
+      "learning_rate": 1.9999990497075926e-05,
+      "loss": 1.2191,
       "step": 1073
     },
     {
-      "epoch": 0.029496580703633517,
+      "epoch": 0.03047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.9652333028362307e-05,
-      "loss": 1.136,
+      "learning_rate": 1.9999989187784404e-05,
+      "loss": 1.1279,
       "step": 1074
     },
     {
-      "epoch": 0.029524044931476753,
+      "epoch": 0.03050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.9670631290027448e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.9999987794022518e-05,
+      "loss": 1.0517,
       "step": 1075
     },
     {
-      "epoch": 0.029551509159319984,
+      "epoch": 0.030533484676503974,
       "grad_norm": 0.0,
-      "learning_rate": 1.968892955169259e-05,
-      "loss": 1.0842,
+      "learning_rate": 1.9999986315790288e-05,
+      "loss": 1.0317,
       "step": 1076
     },
     {
-      "epoch": 0.02957897338716322,
+      "epoch": 0.030561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9707227813357733e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.9999984753087725e-05,
+      "loss": 1.1552,
       "step": 1077
     },
     {
-      "epoch": 0.029606437615006455,
+      "epoch": 0.030590238365493756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9725526075022874e-05,
-      "loss": 1.1433,
+      "learning_rate": 1.999998310591484e-05,
+      "loss": 1.0956,
       "step": 1078
     },
     {
-      "epoch": 0.029633901842849687,
+      "epoch": 0.03061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9743824336688015e-05,
-      "loss": 1.1236,
+      "learning_rate": 1.9999981374271648e-05,
+      "loss": 1.1677,
       "step": 1079
     },
     {
-      "epoch": 0.029661366070692922,
+      "epoch": 0.03064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9762122598353156e-05,
-      "loss": 1.1397,
+      "learning_rate": 1.9999979558158165e-05,
+      "loss": 1.0638,
       "step": 1080
     },
     {
-      "epoch": 0.029688830298536157,
+      "epoch": 0.030675368898978435,
       "grad_norm": 0.0,
-      "learning_rate": 1.97804208600183e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.9999977657574403e-05,
+      "loss": 1.1348,
       "step": 1081
     },
     {
-      "epoch": 0.029716294526379392,
+      "epoch": 0.030703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.979871912168344e-05,
-      "loss": 1.1269,
+      "learning_rate": 1.9999975672520385e-05,
+      "loss": 1.1708,
       "step": 1082
     },
     {
-      "epoch": 0.029743758754222624,
+      "epoch": 0.030732122587968216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9817017383348586e-05,
-      "loss": 1.0535,
+      "learning_rate": 1.9999973602996116e-05,
+      "loss": 1.0725,
       "step": 1083
     },
     {
-      "epoch": 0.02977122298206586,
+      "epoch": 0.03076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835315645013727e-05,
-      "loss": 1.1855,
+      "learning_rate": 1.9999971449001627e-05,
+      "loss": 1.0288,
       "step": 1084
     },
     {
-      "epoch": 0.029798687209909094,
+      "epoch": 0.030788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9853613906678867e-05,
-      "loss": 1.1591,
+      "learning_rate": 1.9999969210536924e-05,
+      "loss": 1.145,
       "step": 1085
     },
     {
-      "epoch": 0.029826151437752326,
+      "epoch": 0.030817253121452896,
       "grad_norm": 0.0,
-      "learning_rate": 1.987191216834401e-05,
-      "loss": 1.0692,
+      "learning_rate": 1.9999966887602036e-05,
+      "loss": 1.239,
       "step": 1086
     },
     {
-      "epoch": 0.02985361566559556,
+      "epoch": 0.030845629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890210430009153e-05,
-      "loss": 1.1105,
+      "learning_rate": 1.9999964480196977e-05,
+      "loss": 1.0939,
       "step": 1087
     },
     {
-      "epoch": 0.029881079893438797,
+      "epoch": 0.030874006810442677,
       "grad_norm": 0.0,
-      "learning_rate": 1.9908508691674294e-05,
-      "loss": 1.2206,
+      "learning_rate": 1.9999961988321765e-05,
+      "loss": 1.1566,
       "step": 1088
     },
     {
-      "epoch": 0.029908544121282032,
+      "epoch": 0.03090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926806953339435e-05,
-      "loss": 1.2359,
+      "learning_rate": 1.999995941197643e-05,
+      "loss": 1.1091,
       "step": 1089
     },
     {
-      "epoch": 0.029936008349125263,
+      "epoch": 0.030930760499432462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945105215004575e-05,
-      "loss": 1.01,
+      "learning_rate": 1.9999956751160986e-05,
+      "loss": 1.1646,
       "step": 1090
     },
     {
-      "epoch": 0.0299634725769685,
+      "epoch": 0.030959137343927357,
       "grad_norm": 0.0,
-      "learning_rate": 1.996340347666972e-05,
-      "loss": 1.1895,
+      "learning_rate": 1.9999954005875457e-05,
+      "loss": 1.1113,
       "step": 1091
     },
     {
-      "epoch": 0.029990936804811734,
+      "epoch": 0.030987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.998170173833486e-05,
-      "loss": 1.1586,
+      "learning_rate": 1.999995117611987e-05,
+      "loss": 1.1141,
       "step": 1092
     },
     {
-      "epoch": 0.030018401032654966,
+      "epoch": 0.031015891032917138,
       "grad_norm": 0.0,
-      "learning_rate": 2e-05,
-      "loss": 1.1984,
+      "learning_rate": 1.999994826189424e-05,
+      "loss": 1.1011,
       "step": 1093
     },
     {
-      "epoch": 0.0300458652604982,
+      "epoch": 0.031044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999999960438064e-05,
-      "loss": 1.0796,
+      "learning_rate": 1.99999452631986e-05,
+      "loss": 1.1009,
       "step": 1094
     },
     {
-      "epoch": 0.030073329488341436,
+      "epoch": 0.031072644721906923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999999841752255e-05,
-      "loss": 1.1995,
+      "learning_rate": 1.9999942180032973e-05,
+      "loss": 1.1047,
       "step": 1095
     },
     {
-      "epoch": 0.030100793716184668,
+      "epoch": 0.031101021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999999643942574e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.999993901239739e-05,
+      "loss": 1.0235,
       "step": 1096
     },
     {
-      "epoch": 0.030128257944027903,
+      "epoch": 0.031129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999999367009017e-05,
-      "loss": 1.2153,
+      "learning_rate": 1.9999935760291868e-05,
+      "loss": 0.9928,
       "step": 1097
     },
     {
-      "epoch": 0.030155722171871138,
+      "epoch": 0.0311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.99999990109516e-05,
-      "loss": 1.1879,
+      "learning_rate": 1.9999932423716437e-05,
+      "loss": 1.189,
       "step": 1098
     },
     {
-      "epoch": 0.030183186399714373,
+      "epoch": 0.031186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999857577031e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.999992900267113e-05,
+      "loss": 1.012,
       "step": 1099
     },
     {
-      "epoch": 0.030210650627557605,
+      "epoch": 0.031214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999806146516e-05,
-      "loss": 1.1471,
+      "learning_rate": 1.9999925497155973e-05,
+      "loss": 1.0475,
       "step": 1100
     },
     {
-      "epoch": 0.03023811485540084,
+      "epoch": 0.031242905788876278,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999746803615e-05,
-      "loss": 1.1373,
+      "learning_rate": 1.9999921907170994e-05,
+      "loss": 1.0188,
       "step": 1101
     },
     {
-      "epoch": 0.030265579083244076,
+      "epoch": 0.03127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999679548329e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.9999918232716228e-05,
+      "loss": 0.9334,
       "step": 1102
     },
     {
-      "epoch": 0.030293043311087307,
+      "epoch": 0.03129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999604380658e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.9999914473791704e-05,
+      "loss": 1.1162,
       "step": 1103
     },
     {
-      "epoch": 0.030320507538930543,
+      "epoch": 0.031328036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999995213006026e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.999991063039745e-05,
+      "loss": 1.1131,
       "step": 1104
     },
     {
-      "epoch": 0.030347971766773778,
+      "epoch": 0.03135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999994303081638e-05,
-      "loss": 1.2142,
+      "learning_rate": 1.99999067025335e-05,
+      "loss": 0.9573,
       "step": 1105
     },
     {
-      "epoch": 0.030375435994617013,
+      "epoch": 0.031384790011350736,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999993314033422e-05,
-      "loss": 1.1517,
+      "learning_rate": 1.999990269019989e-05,
+      "loss": 1.2264,
       "step": 1106
     },
     {
-      "epoch": 0.030402900222460245,
+      "epoch": 0.03141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999992245861382e-05,
-      "loss": 1.2003,
+      "learning_rate": 1.9999898593396655e-05,
+      "loss": 1.0248,
       "step": 1107
     },
     {
-      "epoch": 0.03043036445030348,
+      "epoch": 0.031441543700340524,
       "grad_norm": 0.0,
-      "learning_rate": 1.999999109856553e-05,
-      "loss": 1.1503,
+      "learning_rate": 1.9999894412123825e-05,
+      "loss": 1.1758,
       "step": 1108
     },
     {
-      "epoch": 0.030457828678146715,
+      "epoch": 0.03146992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.999998987214588e-05,
-      "loss": 1.1548,
+      "learning_rate": 1.999989014638144e-05,
+      "loss": 1.159,
       "step": 1109
     },
     {
-      "epoch": 0.030485292905989947,
+      "epoch": 0.031498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999988566602435e-05,
-      "loss": 1.0836,
+      "learning_rate": 1.999988579616953e-05,
+      "loss": 1.1033,
       "step": 1110
     },
     {
-      "epoch": 0.030512757133833182,
+      "epoch": 0.0315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999987181935202e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.9999881361488135e-05,
+      "loss": 1.1097,
       "step": 1111
     },
     {
-      "epoch": 0.030540221361676417,
+      "epoch": 0.031555051078320094,
       "grad_norm": 0.0,
-      "learning_rate": 1.99999857181442e-05,
-      "loss": 1.1333,
+      "learning_rate": 1.9999876842337294e-05,
+      "loss": 1.1604,
       "step": 1112
     },
     {
-      "epoch": 0.03056768558951965,
+      "epoch": 0.03158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999984175229437e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.9999872238717044e-05,
+      "loss": 1.2469,
       "step": 1113
     },
     {
-      "epoch": 0.030595149817362884,
+      "epoch": 0.031611804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.999998255319093e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.9999867550627426e-05,
+      "loss": 1.0451,
       "step": 1114
     },
     {
-      "epoch": 0.03062261404520612,
+      "epoch": 0.03164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999980852028683e-05,
-      "loss": 1.173,
+      "learning_rate": 1.9999862778068474e-05,
+      "loss": 1.1688,
       "step": 1115
     },
     {
-      "epoch": 0.030650078273049355,
+      "epoch": 0.03166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999979071742712e-05,
-      "loss": 1.1844,
+      "learning_rate": 1.9999857921040232e-05,
+      "loss": 0.9243,
       "step": 1116
     },
     {
-      "epoch": 0.030677542500892586,
+      "epoch": 0.03169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999977212333035e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.999985297954274e-05,
+      "loss": 1.1182,
       "step": 1117
     },
     {
-      "epoch": 0.03070500672873582,
+      "epoch": 0.031725312145289446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999975273799666e-05,
-      "loss": 1.1294,
+      "learning_rate": 1.9999847953576038e-05,
+      "loss": 1.1164,
       "step": 1118
     },
     {
-      "epoch": 0.030732470956579057,
+      "epoch": 0.03175368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 1.999997325614262e-05,
-      "loss": 1.1804,
+      "learning_rate": 1.999984284314018e-05,
+      "loss": 1.1098,
       "step": 1119
     },
     {
-      "epoch": 0.03075993518442229,
+      "epoch": 0.03178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999971159361906e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.999983764823519e-05,
+      "loss": 1.0537,
       "step": 1120
     },
     {
-      "epoch": 0.030787399412265524,
+      "epoch": 0.03181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999968983457552e-05,
-      "loss": 1.0946,
+      "learning_rate": 1.9999832368861126e-05,
+      "loss": 1.1184,
       "step": 1121
     },
     {
-      "epoch": 0.03081486364010876,
+      "epoch": 0.031838819523269016,
       "grad_norm": 0.0,
-      "learning_rate": 1.999996672842957e-05,
-      "loss": 1.1655,
+      "learning_rate": 1.9999827005018028e-05,
+      "loss": 1.1411,
       "step": 1122
     },
     {
-      "epoch": 0.030842327867951994,
+      "epoch": 0.0318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999964394277974e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.9999821556705942e-05,
+      "loss": 1.2577,
       "step": 1123
     },
     {
-      "epoch": 0.030869792095795226,
+      "epoch": 0.0318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.999996198100279e-05,
-      "loss": 1.1916,
+      "learning_rate": 1.9999816023924915e-05,
+      "loss": 1.0672,
       "step": 1124
     },
     {
-      "epoch": 0.03089725632363846,
+      "epoch": 0.03192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.999995948860403e-05,
-      "loss": 1.1862,
+      "learning_rate": 1.999981040667499e-05,
+      "loss": 1.0743,
       "step": 1125
     },
     {
-      "epoch": 0.030924720551481696,
+      "epoch": 0.03195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999956917081716e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.999980470495622e-05,
+      "loss": 1.0168,
       "step": 1126
     },
     {
-      "epoch": 0.030952184779324928,
+      "epoch": 0.03198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999954266435873e-05,
-      "loss": 1.0988,
+      "learning_rate": 1.9999798918768648e-05,
+      "loss": 1.1617,
       "step": 1127
     },
     {
-      "epoch": 0.030979649007168163,
+      "epoch": 0.03200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999951536666517e-05,
-      "loss": 1.1725,
+      "learning_rate": 1.9999793048112327e-05,
+      "loss": 1.1187,
       "step": 1128
     },
     {
-      "epoch": 0.0310071132350114,
+      "epoch": 0.032037457434733255,
       "grad_norm": 0.0,
-      "learning_rate": 1.999994872777367e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.9999787092987303e-05,
+      "loss": 1.1457,
       "step": 1129
     },
     {
-      "epoch": 0.03103457746285463,
+      "epoch": 0.03206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999945839757356e-05,
-      "loss": 1.1193,
+      "learning_rate": 1.9999781053393626e-05,
+      "loss": 1.1346,
       "step": 1130
     },
     {
-      "epoch": 0.031062041690697866,
+      "epoch": 0.03209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999942872617594e-05,
-      "loss": 1.1906,
+      "learning_rate": 1.999977492933135e-05,
+      "loss": 1.0232,
       "step": 1131
     },
     {
-      "epoch": 0.0310895059185411,
+      "epoch": 0.03212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999939826354415e-05,
-      "loss": 1.1713,
+      "learning_rate": 1.999976872080053e-05,
+      "loss": 1.1667,
       "step": 1132
     },
     {
-      "epoch": 0.031116970146384336,
+      "epoch": 0.032150964812712825,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999936700967835e-05,
-      "loss": 1.1259,
+      "learning_rate": 1.9999762427801207e-05,
+      "loss": 1.0385,
       "step": 1133
     },
     {
-      "epoch": 0.031144434374227568,
+      "epoch": 0.03217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999933496457884e-05,
-      "loss": 1.0923,
+      "learning_rate": 1.9999756050333446e-05,
+      "loss": 1.114,
       "step": 1134
     },
     {
-      "epoch": 0.031171898602070803,
+      "epoch": 0.03220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999930212824587e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.9999749588397293e-05,
+      "loss": 1.0872,
       "step": 1135
     },
     {
-      "epoch": 0.031199362829914038,
+      "epoch": 0.0322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999926850067967e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.999974304199281e-05,
+      "loss": 1.103,
       "step": 1136
     },
     {
-      "epoch": 0.03122682705775727,
+      "epoch": 0.032264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.999992340818805e-05,
-      "loss": 1.156,
+      "learning_rate": 1.9999736411120044e-05,
+      "loss": 1.1421,
       "step": 1137
     },
     {
-      "epoch": 0.031254291285600505,
+      "epoch": 0.03229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999919887184867e-05,
-      "loss": 1.1791,
+      "learning_rate": 1.9999729695779057e-05,
+      "loss": 1.052,
       "step": 1138
     },
     {
-      "epoch": 0.03128175551344374,
+      "epoch": 0.032321225879682176,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999916287058444e-05,
-      "loss": 1.1865,
+      "learning_rate": 1.9999722895969904e-05,
+      "loss": 0.9908,
       "step": 1139
     },
     {
-      "epoch": 0.031309219741286975,
+      "epoch": 0.03234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.999991260780881e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.9999716011692644e-05,
+      "loss": 1.133,
       "step": 1140
     },
     {
-      "epoch": 0.03133668396913021,
+      "epoch": 0.032377979568671965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999908849435993e-05,
-      "loss": 1.1447,
+      "learning_rate": 1.9999709042947327e-05,
+      "loss": 1.1512,
       "step": 1141
     },
     {
-      "epoch": 0.03136414819697344,
+      "epoch": 0.03240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999905011940026e-05,
-      "loss": 1.0643,
+      "learning_rate": 1.9999701989734025e-05,
+      "loss": 1.1963,
       "step": 1142
     },
     {
-      "epoch": 0.031391612424816674,
+      "epoch": 0.032434733257661746,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999901095320935e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.9999694852052788e-05,
+      "loss": 1.2105,
       "step": 1143
     },
     {
-      "epoch": 0.03141907665265991,
+      "epoch": 0.03246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999897099578755e-05,
-      "loss": 1.1535,
+      "learning_rate": 1.999968762990368e-05,
+      "loss": 1.1259,
       "step": 1144
     },
     {
-      "epoch": 0.031446540880503145,
+      "epoch": 0.032491486946651535,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999893024713513e-05,
-      "loss": 1.1708,
+      "learning_rate": 1.999968032328676e-05,
+      "loss": 1.099,
       "step": 1145
     },
     {
-      "epoch": 0.03147400510834638,
+      "epoch": 0.03251986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999888870725247e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.999967293220209e-05,
+      "loss": 1.1671,
       "step": 1146
     },
     {
-      "epoch": 0.031501469336189615,
+      "epoch": 0.032548240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999884637613983e-05,
-      "loss": 1.1041,
+      "learning_rate": 1.999966545664973e-05,
+      "loss": 1.1042,
       "step": 1147
     },
     {
-      "epoch": 0.03152893356403285,
+      "epoch": 0.03257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999880325379758e-05,
-      "loss": 1.1139,
+      "learning_rate": 1.9999657896629752e-05,
+      "loss": 1.1061,
       "step": 1148
     },
     {
-      "epoch": 0.03155639779187608,
+      "epoch": 0.0326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999875934022607e-05,
-      "loss": 1.084,
+      "learning_rate": 1.9999650252142214e-05,
+      "loss": 1.082,
       "step": 1149
     },
     {
-      "epoch": 0.031583862019719314,
+      "epoch": 0.03263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999871463542567e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.9999642523187178e-05,
+      "loss": 1.0002,
       "step": 1150
     },
     {
-      "epoch": 0.03161132624756255,
+      "epoch": 0.03266174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999866913939668e-05,
-      "loss": 1.1598,
+      "learning_rate": 1.9999634709764716e-05,
+      "loss": 1.0108,
       "step": 1151
     },
     {
-      "epoch": 0.031638790475405784,
+      "epoch": 0.03269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999862285213948e-05,
-      "loss": 1.1288,
+      "learning_rate": 1.9999626811874885e-05,
+      "loss": 1.125,
       "step": 1152
     },
     {
-      "epoch": 0.03166625470324902,
+      "epoch": 0.03271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999857577365443e-05,
-      "loss": 1.1953,
+      "learning_rate": 1.9999618829517763e-05,
+      "loss": 1.1801,
       "step": 1153
     },
     {
-      "epoch": 0.031693718931092255,
+      "epoch": 0.03274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999852790394195e-05,
-      "loss": 1.1607,
+      "learning_rate": 1.9999610762693407e-05,
+      "loss": 1.1085,
       "step": 1154
     },
     {
-      "epoch": 0.03172118315893549,
+      "epoch": 0.03277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999847924300238e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.999960261140189e-05,
+      "loss": 1.1584,
       "step": 1155
     },
     {
-      "epoch": 0.03174864738677872,
+      "epoch": 0.032803632236095344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999842979083608e-05,
-      "loss": 1.1599,
+      "learning_rate": 1.9999594375643277e-05,
+      "loss": 1.0705,
       "step": 1156
     },
     {
-      "epoch": 0.03177611161462195,
+      "epoch": 0.03283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.999983795474435e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.9999586055417645e-05,
+      "loss": 1.075,
       "step": 1157
     },
     {
-      "epoch": 0.03180357584246519,
+      "epoch": 0.03286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999832851282493e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.999957765072506e-05,
+      "loss": 1.0788,
       "step": 1158
     },
     {
-      "epoch": 0.031831040070308424,
+      "epoch": 0.03288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999827668698094e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.9999569161565595e-05,
+      "loss": 1.0356,
       "step": 1159
     },
     {
-      "epoch": 0.03185850429815166,
+      "epoch": 0.032917139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 1.999982240699118e-05,
-      "loss": 1.1384,
+      "learning_rate": 1.9999560587939316e-05,
+      "loss": 1.0715,
       "step": 1160
     },
     {
-      "epoch": 0.031885968525994894,
+      "epoch": 0.03294551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.99998170661618e-05,
-      "loss": 1.2329,
+      "learning_rate": 1.9999551929846298e-05,
+      "loss": 0.994,
       "step": 1161
     },
     {
-      "epoch": 0.03191343275383812,
+      "epoch": 0.0329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999811646209992e-05,
-      "loss": 1.0784,
+      "learning_rate": 1.999954318728662e-05,
+      "loss": 1.0629,
       "step": 1162
     },
     {
-      "epoch": 0.03194089698168136,
+      "epoch": 0.03300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.99998061471358e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.9999534360260348e-05,
+      "loss": 1.2315,
       "step": 1163
     },
     {
-      "epoch": 0.03196836120952459,
+      "epoch": 0.033030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.999980056893927e-05,
-      "loss": 1.1954,
+      "learning_rate": 1.9999525448767558e-05,
+      "loss": 1.1747,
       "step": 1164
     },
     {
-      "epoch": 0.03199582543736783,
+      "epoch": 0.03305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999794911620444e-05,
-      "loss": 1.0894,
+      "learning_rate": 1.999951645280833e-05,
+      "loss": 1.0809,
       "step": 1165
     },
     {
-      "epoch": 0.03202328966521106,
+      "epoch": 0.033087400681044266,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999789175179364e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.9999507372382738e-05,
+      "loss": 0.9746,
       "step": 1166
     },
     {
-      "epoch": 0.0320507538930543,
+      "epoch": 0.03311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999783359616082e-05,
-      "loss": 1.0692,
+      "learning_rate": 1.9999498207490856e-05,
+      "loss": 1.1134,
       "step": 1167
     },
     {
-      "epoch": 0.032078218120897534,
+      "epoch": 0.033144154370034054,
       "grad_norm": 0.0,
-      "learning_rate": 1.999977746493064e-05,
-      "loss": 1.1208,
+      "learning_rate": 1.9999488958132764e-05,
+      "loss": 1.1071,
       "step": 1168
     },
     {
-      "epoch": 0.03210568234874076,
+      "epoch": 0.03317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999771491123084e-05,
-      "loss": 1.1653,
+      "learning_rate": 1.9999479624308537e-05,
+      "loss": 1.1554,
       "step": 1169
     },
     {
-      "epoch": 0.032133146576584,
+      "epoch": 0.033200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999765438193463e-05,
-      "loss": 1.1486,
+      "learning_rate": 1.999947020601826e-05,
+      "loss": 1.064,
       "step": 1170
     },
     {
-      "epoch": 0.03216061080442723,
+      "epoch": 0.03322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.999975930614183e-05,
-      "loss": 1.0872,
+      "learning_rate": 1.9999460703262004e-05,
+      "loss": 1.2153,
       "step": 1171
     },
     {
-      "epoch": 0.03218807503227047,
+      "epoch": 0.033257661748013624,
       "grad_norm": 0.0,
-      "learning_rate": 1.999975309496822e-05,
-      "loss": 1.1159,
+      "learning_rate": 1.9999451116039858e-05,
+      "loss": 1.0988,
       "step": 1172
     },
     {
-      "epoch": 0.0322155392601137,
+      "epoch": 0.03328603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999746804672694e-05,
-      "loss": 1.2299,
+      "learning_rate": 1.9999441444351898e-05,
+      "loss": 1.1158,
       "step": 1173
     },
     {
-      "epoch": 0.03224300348795694,
+      "epoch": 0.033314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999740435255294e-05,
-      "loss": 1.1532,
+      "learning_rate": 1.9999431688198205e-05,
+      "loss": 1.0896,
       "step": 1174
     },
     {
-      "epoch": 0.03227046771580017,
+      "epoch": 0.0333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999733986716078e-05,
-      "loss": 1.1402,
+      "learning_rate": 1.9999421847578864e-05,
+      "loss": 1.0946,
       "step": 1175
     },
     {
-      "epoch": 0.0322979319436434,
+      "epoch": 0.03337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.999972745905509e-05,
-      "loss": 1.1595,
+      "learning_rate": 1.9999411922493958e-05,
+      "loss": 1.16,
       "step": 1176
     },
     {
-      "epoch": 0.03232539617148664,
+      "epoch": 0.03339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999720852272387e-05,
-      "loss": 1.1029,
+      "learning_rate": 1.999940191294357e-05,
+      "loss": 1.1924,
       "step": 1177
     },
     {
-      "epoch": 0.03235286039932987,
+      "epoch": 0.033427922814982976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999714166368017e-05,
+      "learning_rate": 1.9999391818927783e-05,
       "loss": 1.0836,
       "step": 1178
     },
     {
-      "epoch": 0.03238032462717311,
+      "epoch": 0.03345629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999707401342035e-05,
-      "loss": 1.1709,
+      "learning_rate": 1.9999381640446682e-05,
+      "loss": 1.1586,
       "step": 1179
     },
     {
-      "epoch": 0.03240778885501634,
+      "epoch": 0.03348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.99997005571945e-05,
-      "loss": 1.2031,
+      "learning_rate": 1.9999371377500356e-05,
+      "loss": 1.194,
       "step": 1180
     },
     {
-      "epoch": 0.03243525308285958,
+      "epoch": 0.03351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999693633925453e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.9999361030088894e-05,
+      "loss": 1.136,
       "step": 1181
     },
     {
-      "epoch": 0.03246271731070281,
+      "epoch": 0.033541430192962546,
       "grad_norm": 0.0,
-      "learning_rate": 1.999968663153496e-05,
-      "loss": 1.1217,
+      "learning_rate": 1.9999350598212377e-05,
+      "loss": 1.0001,
       "step": 1182
     },
     {
-      "epoch": 0.03249018153854604,
+      "epoch": 0.03356980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999679550023074e-05,
-      "loss": 1.118,
+      "learning_rate": 1.9999340081870894e-05,
+      "loss": 1.054,
       "step": 1183
     },
     {
-      "epoch": 0.032517645766389276,
+      "epoch": 0.03359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.999967238938985e-05,
-      "loss": 1.129,
+      "learning_rate": 1.999932948106454e-05,
+      "loss": 1.1672,
       "step": 1184
     },
     {
-      "epoch": 0.03254510999423251,
+      "epoch": 0.03362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999665149635345e-05,
-      "loss": 1.1692,
+      "learning_rate": 1.99993187957934e-05,
+      "loss": 1.0803,
       "step": 1185
     },
     {
-      "epoch": 0.03257257422207575,
+      "epoch": 0.03365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.999965783075961e-05,
-      "loss": 1.1975,
+      "learning_rate": 1.999930802605756e-05,
+      "loss": 1.1684,
       "step": 1186
     },
     {
-      "epoch": 0.03260003844991898,
+      "epoch": 0.033683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999650432762717e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.9999297171857122e-05,
+      "loss": 1.1723,
       "step": 1187
     },
     {
-      "epoch": 0.03262750267776222,
+      "epoch": 0.0337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.999964295564471e-05,
-      "loss": 1.1607,
+      "learning_rate": 1.9999286233192167e-05,
+      "loss": 1.1069,
       "step": 1188
     },
     {
-      "epoch": 0.03265496690560545,
+      "epoch": 0.033740068104426785,
       "grad_norm": 0.0,
-      "learning_rate": 1.999963539940566e-05,
-      "loss": 1.1918,
+      "learning_rate": 1.999927521006279e-05,
+      "loss": 1.0953,
       "step": 1189
     },
     {
-      "epoch": 0.03268243113344868,
+      "epoch": 0.03376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999627764045617e-05,
-      "loss": 1.1183,
+      "learning_rate": 1.9999264102469094e-05,
+      "loss": 1.1483,
       "step": 1190
     },
     {
-      "epoch": 0.032709895361291916,
+      "epoch": 0.03379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999620049564647e-05,
-      "loss": 1.0587,
+      "learning_rate": 1.9999252910411156e-05,
+      "loss": 1.1584,
       "step": 1191
     },
     {
-      "epoch": 0.03273735958913515,
+      "epoch": 0.03382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.999961225596281e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.9999241633889085e-05,
+      "loss": 1.2137,
       "step": 1192
     },
     {
-      "epoch": 0.032764823816978386,
+      "epoch": 0.033853575482406355,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999604383240164e-05,
-      "loss": 1.1361,
+      "learning_rate": 1.999923027290297e-05,
+      "loss": 1.0601,
       "step": 1193
     },
     {
-      "epoch": 0.03279228804482162,
+      "epoch": 0.03388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999596431396777e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.9999218827452902e-05,
+      "loss": 1.066,
       "step": 1194
     },
     {
-      "epoch": 0.03281975227266486,
+      "epoch": 0.03391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.999958840043271e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.9999207297538987e-05,
+      "loss": 1.0919,
       "step": 1195
     },
     {
-      "epoch": 0.032847216500508085,
+      "epoch": 0.03393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999580290348027e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.9999195683161318e-05,
+      "loss": 1.0858,
       "step": 1196
     },
     {
-      "epoch": 0.03287468072835132,
+      "epoch": 0.033967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.999957210114279e-05,
-      "loss": 1.1931,
+      "learning_rate": 1.9999183984319993e-05,
+      "loss": 1.142,
       "step": 1197
     },
     {
-      "epoch": 0.032902144956194555,
+      "epoch": 0.03399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999563832817067e-05,
-      "loss": 1.2068,
+      "learning_rate": 1.999917220101511e-05,
+      "loss": 1.0593,
       "step": 1198
     },
     {
-      "epoch": 0.03292960918403779,
+      "epoch": 0.034023836549375706,
       "grad_norm": 0.0,
-      "learning_rate": 1.999955548537092e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.9999160333246774e-05,
+      "loss": 1.1547,
       "step": 1199
     },
     {
-      "epoch": 0.032957073411881026,
+      "epoch": 0.0340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999547058804414e-05,
-      "loss": 1.1747,
+      "learning_rate": 1.999914838101508e-05,
+      "loss": 1.1496,
       "step": 1200
     },
     {
-      "epoch": 0.03298453763972426,
+      "epoch": 0.034080590238365495,
       "grad_norm": 0.0,
-      "learning_rate": 1.999953855311762e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.9999136344320128e-05,
+      "loss": 1.0646,
       "step": 1201
     },
     {
-      "epoch": 0.033012001867567496,
+      "epoch": 0.03410896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999529968310606e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.9999124223162023e-05,
+      "loss": 1.0381,
       "step": 1202
     },
     {
-      "epoch": 0.033039466095410724,
+      "epoch": 0.034137343927355276,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999521304383432e-05,
-      "loss": 1.1143,
+      "learning_rate": 1.9999112017540866e-05,
+      "loss": 1.1329,
       "step": 1203
     },
     {
-      "epoch": 0.03306693032325396,
+      "epoch": 0.03416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999512561336178e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.9999099727456757e-05,
+      "loss": 1.2656,
       "step": 1204
     },
     {
-      "epoch": 0.033094394551097195,
+      "epoch": 0.034194097616345065,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999503739168904e-05,
-      "loss": 1.2193,
+      "learning_rate": 1.9999087352909807e-05,
+      "loss": 1.1876,
       "step": 1205
     },
     {
-      "epoch": 0.03312185877894043,
+      "epoch": 0.03422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999494837881682e-05,
-      "loss": 1.2335,
+      "learning_rate": 1.9999074893900116e-05,
+      "loss": 1.0812,
       "step": 1206
     },
     {
-      "epoch": 0.033149323006783665,
+      "epoch": 0.034250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999485857474586e-05,
-      "loss": 1.1244,
+      "learning_rate": 1.9999062350427792e-05,
+      "loss": 1.1744,
       "step": 1207
     },
     {
-      "epoch": 0.0331767872346269,
+      "epoch": 0.03427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999476797947684e-05,
-      "loss": 1.1911,
+      "learning_rate": 1.9999049722492934e-05,
+      "loss": 1.1959,
       "step": 1208
     },
     {
-      "epoch": 0.033204251462470136,
+      "epoch": 0.03430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999467659301048e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.9999037010095658e-05,
+      "loss": 1.0103,
       "step": 1209
     },
     {
-      "epoch": 0.033231715690313364,
+      "epoch": 0.03433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.999945844153475e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.9999024213236064e-05,
+      "loss": 1.1115,
       "step": 1210
     },
     {
-      "epoch": 0.0332591799181566,
+      "epoch": 0.03436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999449144648864e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.9999011331914264e-05,
+      "loss": 1.0889,
       "step": 1211
     },
     {
-      "epoch": 0.033286644145999834,
+      "epoch": 0.03439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999439768643464e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.9998998366130368e-05,
+      "loss": 1.2025,
       "step": 1212
     },
     {
-      "epoch": 0.03331410837384307,
+      "epoch": 0.0344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.999943031351862e-05,
-      "loss": 1.3047,
+      "learning_rate": 1.999898531588448e-05,
+      "loss": 1.0353,
       "step": 1213
     },
     {
-      "epoch": 0.033341572601686305,
+      "epoch": 0.03444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999420779274413e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.9998972181176715e-05,
+      "loss": 1.0947,
       "step": 1214
     },
     {
-      "epoch": 0.03336903682952954,
+      "epoch": 0.03447786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999411165910912e-05,
-      "loss": 1.1502,
+      "learning_rate": 1.999895896200718e-05,
+      "loss": 1.2683,
       "step": 1215
     },
     {
-      "epoch": 0.033396501057372775,
+      "epoch": 0.034506242905788874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999401473428198e-05,
-      "loss": 1.192,
+      "learning_rate": 1.9998945658375995e-05,
+      "loss": 1.0974,
       "step": 1216
     },
     {
-      "epoch": 0.033423965285216004,
+      "epoch": 0.03453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999391701826346e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.9998932270283264e-05,
+      "loss": 1.1192,
       "step": 1217
     },
     {
-      "epoch": 0.03345142951305924,
+      "epoch": 0.03456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999381851105436e-05,
-      "loss": 1.0672,
+      "learning_rate": 1.9998918797729103e-05,
+      "loss": 1.1417,
       "step": 1218
     },
     {
-      "epoch": 0.033478893740902474,
+      "epoch": 0.03459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.999937192126554e-05,
-      "loss": 1.1297,
+      "learning_rate": 1.999890524071362e-05,
+      "loss": 1.1412,
       "step": 1219
     },
     {
-      "epoch": 0.03350635796874571,
+      "epoch": 0.034619750283768444,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999361912306743e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.999889159923694e-05,
+      "loss": 1.1199,
       "step": 1220
     },
     {
-      "epoch": 0.033533822196588944,
+      "epoch": 0.03464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.999935182422912e-05,
-      "loss": 1.1202,
+      "learning_rate": 1.999887787329917e-05,
+      "loss": 1.1506,
       "step": 1221
     },
     {
-      "epoch": 0.03356128642443218,
+      "epoch": 0.03467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.999934165703275e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.9998864062900434e-05,
+      "loss": 1.1284,
       "step": 1222
     },
     {
-      "epoch": 0.03358875065227541,
+      "epoch": 0.03470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999331410717716e-05,
-      "loss": 1.2971,
+      "learning_rate": 1.999885016804084e-05,
+      "loss": 1.1406,
       "step": 1223
     },
     {
-      "epoch": 0.03361621488011864,
+      "epoch": 0.034733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999321085284103e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.999883618872051e-05,
+      "loss": 1.0665,
       "step": 1224
     },
     {
-      "epoch": 0.03364367910796188,
+      "epoch": 0.03476163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.999931068073198e-05,
-      "loss": 1.191,
+      "learning_rate": 1.9998822124939565e-05,
+      "loss": 1.1335,
       "step": 1225
     },
     {
-      "epoch": 0.033671143335805113,
+      "epoch": 0.034790011350737796,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999300197061443e-05,
-      "loss": 1.1277,
+      "learning_rate": 1.9998807976698114e-05,
+      "loss": 1.172,
       "step": 1226
     },
     {
-      "epoch": 0.03369860756364835,
+      "epoch": 0.03481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.999928963427257e-05,
-      "loss": 1.1207,
+      "learning_rate": 1.9998793743996285e-05,
+      "loss": 1.1661,
       "step": 1227
     },
     {
-      "epoch": 0.033726071791491584,
+      "epoch": 0.034846765039727584,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999278992365444e-05,
-      "loss": 1.0595,
+      "learning_rate": 1.9998779426834195e-05,
+      "loss": 1.0439,
       "step": 1228
     },
     {
-      "epoch": 0.03375353601933482,
+      "epoch": 0.03487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999268271340146e-05,
-      "loss": 1.1524,
+      "learning_rate": 1.999876502521197e-05,
+      "loss": 1.0295,
       "step": 1229
     },
     {
-      "epoch": 0.03378100024717805,
+      "epoch": 0.034903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999257471196767e-05,
-      "loss": 1.204,
+      "learning_rate": 1.999875053912972e-05,
+      "loss": 1.1784,
       "step": 1230
     },
     {
-      "epoch": 0.03380846447502128,
+      "epoch": 0.03493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.999924659193539e-05,
-      "loss": 1.1821,
+      "learning_rate": 1.999873596858758e-05,
+      "loss": 1.1077,
       "step": 1231
     },
     {
-      "epoch": 0.03383592870286452,
+      "epoch": 0.034960272417707154,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999235633556096e-05,
-      "loss": 1.1135,
+      "learning_rate": 1.9998721313585668e-05,
+      "loss": 0.9946,
       "step": 1232
     },
     {
-      "epoch": 0.03386339293070775,
+      "epoch": 0.03498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.999922459605898e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.9998706574124104e-05,
+      "loss": 1.1368,
       "step": 1233
     },
     {
-      "epoch": 0.03389085715855099,
+      "epoch": 0.035017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999213479444127e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.9998691750203018e-05,
+      "loss": 1.0347,
       "step": 1234
     },
     {
-      "epoch": 0.03391832138639422,
+      "epoch": 0.03504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999202283711624e-05,
-      "loss": 1.1027,
+      "learning_rate": 1.9998676841822532e-05,
+      "loss": 1.0566,
       "step": 1235
     },
     {
-      "epoch": 0.03394578561423746,
+      "epoch": 0.03507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999191008861556e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.9998661848982775e-05,
+      "loss": 1.1119,
       "step": 1236
     },
     {
-      "epoch": 0.03397324984208069,
+      "epoch": 0.03510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999179654894015e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.9998646771683876e-05,
+      "loss": 1.1189,
       "step": 1237
     },
     {
-      "epoch": 0.03400071406992392,
+      "epoch": 0.035130533484676506,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999168221809095e-05,
-      "loss": 1.1138,
+      "learning_rate": 1.999863160992595e-05,
+      "loss": 1.1071,
       "step": 1238
     },
     {
-      "epoch": 0.03402817829776716,
+      "epoch": 0.03515891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 1.999915670960688e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.9998616363709136e-05,
+      "loss": 1.0993,
       "step": 1239
     },
     {
-      "epoch": 0.03405564252561039,
+      "epoch": 0.03518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999145118287463e-05,
-      "loss": 1.0612,
+      "learning_rate": 1.999860103303356e-05,
+      "loss": 1.0732,
       "step": 1240
     },
     {
-      "epoch": 0.03408310675345363,
+      "epoch": 0.03521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999133447850937e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.999858561789936e-05,
+      "loss": 1.0735,
       "step": 1241
     },
     {
-      "epoch": 0.03411057098129686,
+      "epoch": 0.035244040862656076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999121698297398e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.9998570118306647e-05,
+      "loss": 1.0306,
       "step": 1242
     },
     {
-      "epoch": 0.0341380352091401,
+      "epoch": 0.03527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.999910986962693e-05,
-      "loss": 1.0667,
+      "learning_rate": 1.9998554534255565e-05,
+      "loss": 1.1354,
       "step": 1243
     },
     {
-      "epoch": 0.034165499436983326,
+      "epoch": 0.03530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999097961839632e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.9998538865746245e-05,
+      "loss": 1.135,
       "step": 1244
     },
     {
-      "epoch": 0.03419296366482656,
+      "epoch": 0.03532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999085974935602e-05,
-      "loss": 1.2332,
+      "learning_rate": 1.999852311277882e-05,
+      "loss": 1.2171,
       "step": 1245
     },
     {
-      "epoch": 0.0342204278926698,
+      "epoch": 0.03535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999073908914927e-05,
-      "loss": 1.1873,
+      "learning_rate": 1.9998507275353417e-05,
+      "loss": 1.1097,
       "step": 1246
     },
     {
-      "epoch": 0.03424789212051303,
+      "epoch": 0.03538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999061763777706e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.9998491353470176e-05,
+      "loss": 1.0301,
       "step": 1247
     },
     {
-      "epoch": 0.03427535634835627,
+      "epoch": 0.03541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999049539524033e-05,
-      "loss": 1.1198,
+      "learning_rate": 1.9998475347129226e-05,
+      "loss": 1.1252,
       "step": 1248
     },
     {
-      "epoch": 0.0343028205761995,
+      "epoch": 0.035442678774120315,
       "grad_norm": 0.0,
-      "learning_rate": 1.999903723615401e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.999845925633071e-05,
+      "loss": 1.0934,
       "step": 1249
     },
     {
-      "epoch": 0.03433028480404274,
+      "epoch": 0.03547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999024853667728e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.9998443081074754e-05,
+      "loss": 1.1041,
       "step": 1250
     },
     {
-      "epoch": 0.034357749031885966,
+      "epoch": 0.0354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.9999012392065292e-05,
-      "loss": 1.1088,
+      "learning_rate": 1.9998426821361502e-05,
+      "loss": 1.196,
       "step": 1251
     },
     {
-      "epoch": 0.0343852132597292,
+      "epoch": 0.035527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998999851346795e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.9998410477191087e-05,
+      "loss": 1.1141,
       "step": 1252
     },
     {
-      "epoch": 0.034412677487572436,
+      "epoch": 0.035556186152099885,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998987231512337e-05,
-      "loss": 0.9549,
+      "learning_rate": 1.9998394048563653e-05,
+      "loss": 1.0136,
       "step": 1253
     },
     {
-      "epoch": 0.03444014171541567,
+      "epoch": 0.03558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.999897453256202e-05,
-      "loss": 1.1564,
+      "learning_rate": 1.9998377535479334e-05,
+      "loss": 1.0909,
       "step": 1254
     },
     {
-      "epoch": 0.03446760594325891,
+      "epoch": 0.03561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998961754495942e-05,
-      "loss": 1.1839,
+      "learning_rate": 1.9998360937938274e-05,
+      "loss": 1.1531,
       "step": 1255
     },
     {
-      "epoch": 0.03449507017110214,
+      "epoch": 0.03564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998948897314206e-05,
-      "loss": 1.0466,
+      "learning_rate": 1.9998344255940602e-05,
+      "loss": 1.0243,
       "step": 1256
     },
     {
-      "epoch": 0.03452253439894537,
+      "epoch": 0.035669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998935961016916e-05,
-      "loss": 1.0887,
+      "learning_rate": 1.9998327489486474e-05,
+      "loss": 1.1557,
       "step": 1257
     },
     {
-      "epoch": 0.034549998626788606,
+      "epoch": 0.03569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998922945604166e-05,
-      "loss": 1.2325,
+      "learning_rate": 1.999831063857602e-05,
+      "loss": 1.0771,
       "step": 1258
     },
     {
-      "epoch": 0.03457746285463184,
+      "epoch": 0.035726447219069236,
       "grad_norm": 0.0,
-      "learning_rate": 1.999890985107607e-05,
-      "loss": 1.1257,
+      "learning_rate": 1.9998293703209388e-05,
+      "loss": 1.0991,
       "step": 1259
     },
     {
-      "epoch": 0.034604927082475076,
+      "epoch": 0.03575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.999889667743272e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.9998276683386716e-05,
+      "loss": 1.0704,
       "step": 1260
     },
     {
-      "epoch": 0.03463239131031831,
+      "epoch": 0.035783200908059025,
       "grad_norm": 0.0,
-      "learning_rate": 1.999888342467423e-05,
-      "loss": 1.1575,
+      "learning_rate": 1.9998259579108155e-05,
+      "loss": 1.1071,
       "step": 1261
     },
     {
-      "epoch": 0.034659855538161546,
+      "epoch": 0.03581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998870092800703e-05,
-      "loss": 0.9541,
+      "learning_rate": 1.9998242390373846e-05,
+      "loss": 1.0708,
       "step": 1262
     },
     {
-      "epoch": 0.03468731976600478,
+      "epoch": 0.035839954597048806,
       "grad_norm": 0.0,
-      "learning_rate": 1.999885668181224e-05,
-      "loss": 1.1764,
+      "learning_rate": 1.999822511718393e-05,
+      "loss": 0.9999,
       "step": 1263
     },
     {
-      "epoch": 0.03471478399384801,
+      "epoch": 0.0358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.999884319170895e-05,
-      "loss": 1.212,
+      "learning_rate": 1.999820775953856e-05,
+      "loss": 1.0459,
       "step": 1264
     },
     {
-      "epoch": 0.034742248221691245,
+      "epoch": 0.035896708286038595,
       "grad_norm": 0.0,
-      "learning_rate": 1.999882962249094e-05,
-      "loss": 1.2043,
+      "learning_rate": 1.9998190317437883e-05,
+      "loss": 1.1022,
       "step": 1265
     },
     {
-      "epoch": 0.03476971244953448,
+      "epoch": 0.03592508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998815974158315e-05,
-      "loss": 1.1593,
+      "learning_rate": 1.999817279088204e-05,
+      "loss": 1.1146,
       "step": 1266
     },
     {
-      "epoch": 0.034797176677377716,
+      "epoch": 0.03595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998802246711188e-05,
-      "loss": 1.2079,
+      "learning_rate": 1.999815517987118e-05,
+      "loss": 1.176,
       "step": 1267
     },
     {
-      "epoch": 0.03482464090522095,
+      "epoch": 0.03598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998788440149663e-05,
-      "loss": 1.1838,
+      "learning_rate": 1.9998137484405456e-05,
+      "loss": 1.1653,
       "step": 1268
     },
     {
-      "epoch": 0.034852105133064186,
+      "epoch": 0.03601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.999877455447385e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.9998119704485016e-05,
+      "loss": 0.9491,
       "step": 1269
     },
     {
-      "epoch": 0.03487956936090742,
+      "epoch": 0.03603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.999876058968386e-05,
-      "loss": 1.1293,
+      "learning_rate": 1.999810184011001e-05,
+      "loss": 1.1568,
       "step": 1270
     },
     {
-      "epoch": 0.03490703358875065,
+      "epoch": 0.03606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998746545779802e-05,
-      "loss": 1.1675,
+      "learning_rate": 1.9998083891280584e-05,
+      "loss": 1.1309,
       "step": 1271
     },
     {
-      "epoch": 0.034934497816593885,
+      "epoch": 0.03609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998732422761787e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.99980658579969e-05,
+      "loss": 1.0787,
       "step": 1272
     },
     {
-      "epoch": 0.03496196204443712,
+      "epoch": 0.03612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998718220629933e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.99980477402591e-05,
+      "loss": 1.1589,
       "step": 1273
     },
     {
-      "epoch": 0.034989426272280355,
+      "epoch": 0.03615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998703939384347e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.999802953806734e-05,
+      "loss": 1.2008,
       "step": 1274
     },
     {
-      "epoch": 0.03501689050012359,
+      "epoch": 0.03618047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998689579025138e-05,
-      "loss": 1.1948,
+      "learning_rate": 1.9998011251421775e-05,
+      "loss": 1.0872,
       "step": 1275
     },
     {
-      "epoch": 0.035044354727966825,
+      "epoch": 0.036208853575482404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998675139552426e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.9997992880322564e-05,
+      "loss": 1.1526,
       "step": 1276
     },
     {
-      "epoch": 0.03507181895581006,
+      "epoch": 0.0362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.999866062096632e-05,
-      "loss": 1.1771,
+      "learning_rate": 1.9997974424769855e-05,
+      "loss": 1.0873,
       "step": 1277
     },
     {
-      "epoch": 0.03509928318365329,
+      "epoch": 0.03626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998646023266942e-05,
-      "loss": 1.1096,
+      "learning_rate": 1.9997955884763804e-05,
+      "loss": 1.1399,
       "step": 1278
     },
     {
-      "epoch": 0.035126747411496524,
+      "epoch": 0.03629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998631346454402e-05,
-      "loss": 1.088,
+      "learning_rate": 1.9997937260304577e-05,
+      "loss": 1.0861,
       "step": 1279
     },
     {
-      "epoch": 0.03515421163933976,
+      "epoch": 0.036322360953461974,
       "grad_norm": 0.0,
-      "learning_rate": 1.999861659052882e-05,
-      "loss": 1.1638,
+      "learning_rate": 1.999791855139232e-05,
+      "loss": 1.1161,
       "step": 1280
     },
     {
-      "epoch": 0.035181675867182995,
+      "epoch": 0.03635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998601755490307e-05,
-      "loss": 1.2166,
+      "learning_rate": 1.9997899758027194e-05,
+      "loss": 1.0652,
       "step": 1281
     },
     {
-      "epoch": 0.03520914009502623,
+      "epoch": 0.03637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.999858684133899e-05,
-      "loss": 1.1382,
+      "learning_rate": 1.999788088020936e-05,
+      "loss": 1.0107,
       "step": 1282
     },
     {
-      "epoch": 0.035236604322869465,
+      "epoch": 0.03640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998571848074973e-05,
-      "loss": 1.1396,
+      "learning_rate": 1.999786191793898e-05,
+      "loss": 1.0565,
       "step": 1283
     },
     {
-      "epoch": 0.03526406855071269,
+      "epoch": 0.036435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998556775698385e-05,
-      "loss": 1.1525,
+      "learning_rate": 1.9997842871216207e-05,
+      "loss": 1.0727,
       "step": 1284
     },
     {
-      "epoch": 0.03529153277855593,
+      "epoch": 0.03646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998541624209344e-05,
-      "loss": 1.0815,
+      "learning_rate": 1.999782374004121e-05,
+      "loss": 1.1263,
       "step": 1285
     },
     {
-      "epoch": 0.035318997006399164,
+      "epoch": 0.036492622020431326,
       "grad_norm": 0.0,
-      "learning_rate": 1.999852639360797e-05,
-      "loss": 1.1761,
+      "learning_rate": 1.9997804524414145e-05,
+      "loss": 1.1734,
       "step": 1286
     },
     {
-      "epoch": 0.0353464612342424,
+      "epoch": 0.03652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.999851108389438e-05,
-      "loss": 1.181,
+      "learning_rate": 1.9997785224335174e-05,
+      "loss": 1.0879,
       "step": 1287
     },
     {
-      "epoch": 0.035373925462085634,
+      "epoch": 0.036549375709421114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998495695068695e-05,
-      "loss": 1.232,
+      "learning_rate": 1.9997765839804464e-05,
+      "loss": 1.1596,
       "step": 1288
     },
     {
-      "epoch": 0.03540138968992887,
+      "epoch": 0.036577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.999848022713104e-05,
-      "loss": 1.0709,
+      "learning_rate": 1.9997746370822174e-05,
+      "loss": 1.2098,
       "step": 1289
     },
     {
-      "epoch": 0.035428853917772105,
+      "epoch": 0.036606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.999846468008154e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.9997726817388476e-05,
+      "loss": 1.1445,
       "step": 1290
     },
     {
-      "epoch": 0.03545631814561533,
+      "epoch": 0.03663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998449053920314e-05,
-      "loss": 1.1989,
+      "learning_rate": 1.9997707179503526e-05,
+      "loss": 1.015,
       "step": 1291
     },
     {
-      "epoch": 0.03548378237345857,
+      "epoch": 0.036662883087400684,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998433348647484e-05,
-      "loss": 1.1813,
+      "learning_rate": 1.9997687457167495e-05,
+      "loss": 1.0678,
       "step": 1292
     },
     {
-      "epoch": 0.0355112466013018,
+      "epoch": 0.03669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998417564263176e-05,
-      "loss": 1.1766,
+      "learning_rate": 1.9997667650380547e-05,
+      "loss": 1.0296,
       "step": 1293
     },
     {
-      "epoch": 0.03553871082914504,
+      "epoch": 0.036719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998401700767516e-05,
-      "loss": 1.2605,
+      "learning_rate": 1.9997647759142852e-05,
+      "loss": 1.0941,
       "step": 1294
     },
     {
-      "epoch": 0.035566175056988274,
+      "epoch": 0.03674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998385758160628e-05,
-      "loss": 1.0739,
+      "learning_rate": 1.9997627783454577e-05,
+      "loss": 1.0073,
       "step": 1295
     },
     {
-      "epoch": 0.03559363928483151,
+      "epoch": 0.03677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.999836973644264e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.999760772331589e-05,
+      "loss": 1.1444,
       "step": 1296
     },
     {
-      "epoch": 0.035621103512674744,
+      "epoch": 0.03680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998353635613676e-05,
-      "loss": 1.1479,
+      "learning_rate": 1.999758757872696e-05,
+      "loss": 1.0406,
       "step": 1297
     },
     {
-      "epoch": 0.03564856774051797,
+      "epoch": 0.036833144154370036,
       "grad_norm": 0.0,
-      "learning_rate": 1.999833745567387e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.9997567349687963e-05,
+      "loss": 1.0349,
       "step": 1298
     },
     {
-      "epoch": 0.03567603196836121,
+      "epoch": 0.03686152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 1.999832119662334e-05,
-      "loss": 1.1045,
+      "learning_rate": 1.9997547036199062e-05,
+      "loss": 1.1228,
       "step": 1299
     },
     {
-      "epoch": 0.03570349619620444,
+      "epoch": 0.03688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.999830485846222e-05,
-      "loss": 1.1288,
+      "learning_rate": 1.999752663826043e-05,
+      "loss": 1.0758,
       "step": 1300
     },
     {
-      "epoch": 0.03573096042404768,
+      "epoch": 0.03691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.999828844119064e-05,
-      "loss": 1.1273,
+      "learning_rate": 1.9997506155872246e-05,
+      "loss": 1.0991,
       "step": 1301
     },
     {
-      "epoch": 0.03575842465189091,
+      "epoch": 0.036946651532349606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998271944808727e-05,
-      "loss": 1.2073,
+      "learning_rate": 1.9997485589034676e-05,
+      "loss": 1.1101,
       "step": 1302
     },
     {
-      "epoch": 0.03578588887973415,
+      "epoch": 0.03697502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998255369316616e-05,
-      "loss": 1.2317,
+      "learning_rate": 1.9997464937747893e-05,
+      "loss": 1.0897,
       "step": 1303
     },
     {
-      "epoch": 0.035813353107577384,
+      "epoch": 0.03700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998238714714436e-05,
-      "loss": 1.141,
+      "learning_rate": 1.999744420201208e-05,
+      "loss": 1.1467,
       "step": 1304
     },
     {
-      "epoch": 0.03584081733542061,
+      "epoch": 0.03703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998221981002316e-05,
-      "loss": 1.1154,
+      "learning_rate": 1.99974233818274e-05,
+      "loss": 1.096,
       "step": 1305
     },
     {
-      "epoch": 0.03586828156326385,
+      "epoch": 0.03706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998205168180392e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.999740247719404e-05,
+      "loss": 1.0614,
       "step": 1306
     },
     {
-      "epoch": 0.03589574579110708,
+      "epoch": 0.03708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998188276248796e-05,
-      "loss": 1.1052,
+      "learning_rate": 1.9997381488112166e-05,
+      "loss": 1.1507,
       "step": 1307
     },
     {
-      "epoch": 0.03592321001895032,
+      "epoch": 0.03711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.999817130520766e-05,
-      "loss": 1.1245,
+      "learning_rate": 1.9997360414581967e-05,
+      "loss": 1.1402,
       "step": 1308
     },
     {
-      "epoch": 0.03595067424679355,
+      "epoch": 0.037145289443813845,
       "grad_norm": 0.0,
-      "learning_rate": 1.999815425505712e-05,
-      "loss": 1.1702,
+      "learning_rate": 1.9997339256603613e-05,
+      "loss": 1.1121,
       "step": 1309
     },
     {
-      "epoch": 0.03597813847463679,
+      "epoch": 0.03717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998137125797314e-05,
-      "loss": 1.1147,
+      "learning_rate": 1.9997318014177283e-05,
+      "loss": 1.1083,
       "step": 1310
     },
     {
-      "epoch": 0.03600560270248002,
+      "epoch": 0.03720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.999811991742837e-05,
-      "loss": 1.1758,
+      "learning_rate": 1.9997296687303162e-05,
+      "loss": 1.0427,
       "step": 1311
     },
     {
-      "epoch": 0.03603306693032325,
+      "epoch": 0.03723041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.999810262995043e-05,
-      "loss": 1.1235,
+      "learning_rate": 1.9997275275981428e-05,
+      "loss": 1.0449,
       "step": 1312
     },
     {
-      "epoch": 0.03606053115816649,
+      "epoch": 0.037258796821793415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998085263363627e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.9997253780212253e-05,
+      "loss": 1.0616,
       "step": 1313
     },
     {
-      "epoch": 0.03608799538600972,
+      "epoch": 0.03728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998067817668104e-05,
-      "loss": 1.1237,
+      "learning_rate": 1.9997232199995832e-05,
+      "loss": 1.1141,
       "step": 1314
     },
     {
-      "epoch": 0.03611545961385296,
+      "epoch": 0.0373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998050292863994e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.9997210535332337e-05,
+      "loss": 1.0897,
       "step": 1315
     },
     {
-      "epoch": 0.03614292384169619,
+      "epoch": 0.03734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998032688951437e-05,
-      "loss": 1.1107,
+      "learning_rate": 1.9997188786221958e-05,
+      "loss": 1.0572,
       "step": 1316
     },
     {
-      "epoch": 0.03617038806953943,
+      "epoch": 0.037372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9998015005930574e-05,
-      "loss": 1.105,
+      "learning_rate": 1.9997166952664877e-05,
+      "loss": 1.141,
       "step": 1317
     },
     {
-      "epoch": 0.036197852297382656,
+      "epoch": 0.03740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.999799724380154e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.9997145034661274e-05,
+      "loss": 1.1018,
       "step": 1318
     },
     {
-      "epoch": 0.03622531652522589,
+      "epoch": 0.037429057888762766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997979402564483e-05,
-      "loss": 1.2092,
+      "learning_rate": 1.9997123032211343e-05,
+      "loss": 0.9648,
       "step": 1319
     },
     {
-      "epoch": 0.036252780753069126,
+      "epoch": 0.03745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.999796148221954e-05,
-      "loss": 1.064,
+      "learning_rate": 1.9997100945315256e-05,
+      "loss": 1.0932,
       "step": 1320
     },
     {
-      "epoch": 0.03628024498091236,
+      "epoch": 0.037485811577752555,
       "grad_norm": 0.0,
-      "learning_rate": 1.999794348276685e-05,
-      "loss": 1.0927,
+      "learning_rate": 1.9997078773973217e-05,
+      "loss": 1.152,
       "step": 1321
     },
     {
-      "epoch": 0.0363077092087556,
+      "epoch": 0.03751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.999792540420656e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.9997056518185397e-05,
+      "loss": 1.1836,
       "step": 1322
     },
     {
-      "epoch": 0.03633517343659883,
+      "epoch": 0.037542565266742337,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997907246538812e-05,
-      "loss": 1.157,
+      "learning_rate": 1.9997034177951995e-05,
+      "loss": 1.0556,
       "step": 1323
     },
     {
-      "epoch": 0.03636263766444207,
+      "epoch": 0.03757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997889009763746e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.9997011753273195e-05,
+      "loss": 1.04,
       "step": 1324
     },
     {
-      "epoch": 0.036390101892285295,
+      "epoch": 0.037599318955732125,
       "grad_norm": 0.0,
-      "learning_rate": 1.999787069388151e-05,
-      "loss": 1.1633,
+      "learning_rate": 1.999698924414919e-05,
+      "loss": 1.0475,
       "step": 1325
     },
     {
-      "epoch": 0.03641756612012853,
+      "epoch": 0.03762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997852298892254e-05,
-      "loss": 1.097,
+      "learning_rate": 1.999696665058016e-05,
+      "loss": 1.0991,
       "step": 1326
     },
     {
-      "epoch": 0.036445030347971766,
+      "epoch": 0.03765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.999783382479611e-05,
-      "loss": 1.1261,
+      "learning_rate": 1.999694397256631e-05,
+      "loss": 1.1665,
       "step": 1327
     },
     {
-      "epoch": 0.036472494575815,
+      "epoch": 0.0376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997815271593235e-05,
-      "loss": 1.1551,
+      "learning_rate": 1.9996921210107823e-05,
+      "loss": 1.1594,
       "step": 1328
     },
     {
-      "epoch": 0.036499958803658236,
+      "epoch": 0.03771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997796639283774e-05,
-      "loss": 1.1303,
+      "learning_rate": 1.999689836320489e-05,
+      "loss": 1.1004,
       "step": 1329
     },
     {
-      "epoch": 0.03652742303150147,
+      "epoch": 0.03774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997777927867874e-05,
-      "loss": 1.1335,
+      "learning_rate": 1.9996875431857715e-05,
+      "loss": 1.1137,
       "step": 1330
     },
     {
-      "epoch": 0.03655488725934471,
+      "epoch": 0.03776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.999775913734568e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.9996852416066477e-05,
+      "loss": 1.1244,
       "step": 1331
     },
     {
-      "epoch": 0.036582351487187935,
+      "epoch": 0.03779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997740267717343e-05,
-      "loss": 1.084,
+      "learning_rate": 1.999682931583138e-05,
+      "loss": 1.1667,
       "step": 1332
     },
     {
-      "epoch": 0.03660981571503117,
+      "epoch": 0.03782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997721318983014e-05,
-      "loss": 1.1143,
+      "learning_rate": 1.9996806131152614e-05,
+      "loss": 1.0464,
       "step": 1333
     },
     {
-      "epoch": 0.036637279942874405,
+      "epoch": 0.03785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.999770229114284e-05,
-      "loss": 1.1449,
+      "learning_rate": 1.999678286203038e-05,
+      "loss": 1.0611,
       "step": 1334
     },
     {
-      "epoch": 0.03666474417071764,
+      "epoch": 0.03788308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.999768318419697e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.999675950846487e-05,
+      "loss": 1.0367,
       "step": 1335
     },
     {
-      "epoch": 0.036692208398560876,
+      "epoch": 0.037911464245175934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997663998145562e-05,
-      "loss": 1.1199,
+      "learning_rate": 1.999673607045628e-05,
+      "loss": 1.2136,
       "step": 1336
     },
     {
-      "epoch": 0.03671967262640411,
+      "epoch": 0.03793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.999764473298876e-05,
-      "loss": 1.2533,
+      "learning_rate": 1.999671254800482e-05,
+      "loss": 1.1363,
       "step": 1337
     },
     {
-      "epoch": 0.036747136854247346,
+      "epoch": 0.03796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997625388726725e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9996688941110672e-05,
+      "loss": 1.0368,
       "step": 1338
     },
     {
-      "epoch": 0.036774601082090574,
+      "epoch": 0.03799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.99976059653596e-05,
-      "loss": 1.1164,
+      "learning_rate": 1.9996665249774047e-05,
+      "loss": 1.1336,
       "step": 1339
     },
     {
-      "epoch": 0.03680206530993381,
+      "epoch": 0.038024971623155504,
       "grad_norm": 0.0,
-      "learning_rate": 1.999758646288755e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.9996641473995138e-05,
+      "loss": 1.0932,
       "step": 1340
     },
     {
-      "epoch": 0.036829529537777045,
+      "epoch": 0.0380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.999756688131072e-05,
-      "loss": 1.0451,
+      "learning_rate": 1.9996617613774152e-05,
+      "loss": 1.0707,
       "step": 1341
     },
     {
-      "epoch": 0.03685699376562028,
+      "epoch": 0.03808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.999754722062927e-05,
-      "loss": 1.1865,
+      "learning_rate": 1.9996593669111286e-05,
+      "loss": 1.0793,
       "step": 1342
     },
     {
-      "epoch": 0.036884457993463515,
+      "epoch": 0.03811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.999752748084335e-05,
-      "loss": 1.1534,
+      "learning_rate": 1.9996569640006744e-05,
+      "loss": 1.0394,
       "step": 1343
     },
     {
-      "epoch": 0.03691192222130675,
+      "epoch": 0.038138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.999750766195312e-05,
-      "loss": 1.2191,
+      "learning_rate": 1.9996545526460732e-05,
+      "loss": 1.1033,
       "step": 1344
     },
     {
-      "epoch": 0.03693938644914998,
+      "epoch": 0.03816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997487763958738e-05,
-      "loss": 1.1968,
+      "learning_rate": 1.9996521328473446e-05,
+      "loss": 1.1003,
       "step": 1345
     },
     {
-      "epoch": 0.036966850676993214,
+      "epoch": 0.038195232690124856,
       "grad_norm": 0.0,
-      "learning_rate": 1.999746778686036e-05,
-      "loss": 1.1425,
+      "learning_rate": 1.9996497046045096e-05,
+      "loss": 1.1318,
       "step": 1346
     },
     {
-      "epoch": 0.03699431490483645,
+      "epoch": 0.03822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997447730658147e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.9996472679175887e-05,
+      "loss": 1.0308,
       "step": 1347
     },
     {
-      "epoch": 0.037021779132679684,
+      "epoch": 0.038251986379114644,
       "grad_norm": 0.0,
-      "learning_rate": 1.999742759535225e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.9996448227866024e-05,
+      "loss": 1.1315,
       "step": 1348
     },
     {
-      "epoch": 0.03704924336052292,
+      "epoch": 0.03828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997407380942838e-05,
-      "loss": 1.1582,
+      "learning_rate": 1.9996423692115714e-05,
+      "loss": 1.0494,
       "step": 1349
     },
     {
-      "epoch": 0.037076707588366155,
+      "epoch": 0.038308740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997387087430064e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.999639907192516e-05,
+      "loss": 1.0369,
       "step": 1350
     },
     {
-      "epoch": 0.03710417181620939,
+      "epoch": 0.03833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.999736671481409e-05,
-      "loss": 1.04,
+      "learning_rate": 1.999637436729458e-05,
+      "loss": 1.0959,
       "step": 1351
     },
     {
-      "epoch": 0.03713163604405262,
+      "epoch": 0.038365493757094214,
       "grad_norm": 0.0,
-      "learning_rate": 1.999734626309508e-05,
-      "loss": 1.1211,
+      "learning_rate": 1.999634957822417e-05,
+      "loss": 1.2135,
       "step": 1352
     },
     {
-      "epoch": 0.037159100271895854,
+      "epoch": 0.0383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.999732573227319e-05,
-      "loss": 1.105,
+      "learning_rate": 1.999632470471415e-05,
+      "loss": 1.1397,
       "step": 1353
     },
     {
-      "epoch": 0.03718656449973909,
+      "epoch": 0.038422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.999730512234859e-05,
-      "loss": 1.0988,
+      "learning_rate": 1.9996299746764725e-05,
+      "loss": 1.1454,
       "step": 1354
     },
     {
-      "epoch": 0.037214028727582324,
+      "epoch": 0.03845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997284433321438e-05,
-      "loss": 0.9931,
+      "learning_rate": 1.9996274704376105e-05,
+      "loss": 1.0349,
       "step": 1355
     },
     {
-      "epoch": 0.03724149295542556,
+      "epoch": 0.03847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.99972636651919e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.9996249577548504e-05,
+      "loss": 1.0767,
       "step": 1356
     },
     {
-      "epoch": 0.037268957183268794,
+      "epoch": 0.03850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997242817960138e-05,
-      "loss": 1.075,
+      "learning_rate": 1.9996224366282135e-05,
+      "loss": 1.0404,
       "step": 1357
     },
     {
-      "epoch": 0.03729642141111203,
+      "epoch": 0.038535754824063566,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997221891626318e-05,
-      "loss": 1.107,
+      "learning_rate": 1.999619907057721e-05,
+      "loss": 1.032,
       "step": 1358
     },
     {
-      "epoch": 0.03732388563895526,
+      "epoch": 0.03856413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997200886190605e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.999617369043394e-05,
+      "loss": 1.1699,
       "step": 1359
     },
     {
-      "epoch": 0.03735134986679849,
+      "epoch": 0.03859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.999717980165317e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.999614822585254e-05,
+      "loss": 1.2491,
       "step": 1360
     },
     {
-      "epoch": 0.03737881409464173,
+      "epoch": 0.03862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.999715863801417e-05,
-      "loss": 1.1614,
+      "learning_rate": 1.9996122676833227e-05,
+      "loss": 1.2008,
       "step": 1361
     },
     {
-      "epoch": 0.037406278322484963,
+      "epoch": 0.038649262202043136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997137395273783e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.999609704337622e-05,
+      "loss": 1.071,
       "step": 1362
     },
     {
-      "epoch": 0.0374337425503282,
+      "epoch": 0.03867763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997116073432172e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.9996071325481728e-05,
+      "loss": 1.0821,
       "step": 1363
     },
     {
-      "epoch": 0.037461206778171434,
+      "epoch": 0.03870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997094672489505e-05,
-      "loss": 1.1676,
+      "learning_rate": 1.9996045523149977e-05,
+      "loss": 1.2499,
       "step": 1364
     },
     {
-      "epoch": 0.03748867100601467,
+      "epoch": 0.03873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.999707319244595e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.9996019636381178e-05,
+      "loss": 1.0806,
       "step": 1365
     },
     {
-      "epoch": 0.0375161352338579,
+      "epoch": 0.0387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997051633301684e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.9995993665175547e-05,
+      "loss": 1.0557,
       "step": 1366
     },
     {
-      "epoch": 0.03754359946170113,
+      "epoch": 0.03879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 1.999702999505687e-05,
-      "loss": 1.1933,
+      "learning_rate": 1.9995967609533314e-05,
+      "loss": 1.1001,
       "step": 1367
     },
     {
-      "epoch": 0.03757106368954437,
+      "epoch": 0.03881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.9997008277711685e-05,
-      "loss": 1.0134,
+      "learning_rate": 1.999594146945469e-05,
+      "loss": 1.0077,
       "step": 1368
     },
     {
-      "epoch": 0.0375985279173876,
+      "epoch": 0.038847900113507375,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996986481266292e-05,
-      "loss": 1.0686,
+      "learning_rate": 1.99959152449399e-05,
+      "loss": 1.115,
       "step": 1369
     },
     {
-      "epoch": 0.03762599214523084,
+      "epoch": 0.03887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996964605720877e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.9995888935989164e-05,
+      "loss": 1.108,
       "step": 1370
     },
     {
-      "epoch": 0.03765345637307407,
+      "epoch": 0.03890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.99969426510756e-05,
-      "loss": 1.1151,
+      "learning_rate": 1.9995862542602708e-05,
+      "loss": 1.1609,
       "step": 1371
     },
     {
-      "epoch": 0.03768092060091731,
+      "epoch": 0.03893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996920617330646e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.999583606478075e-05,
+      "loss": 1.0772,
       "step": 1372
     },
     {
-      "epoch": 0.03770838482876054,
+      "epoch": 0.038961407491486945,
       "grad_norm": 0.0,
-      "learning_rate": 1.999689850448618e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9995809502523513e-05,
+      "loss": 1.2271,
       "step": 1373
     },
     {
-      "epoch": 0.03773584905660377,
+      "epoch": 0.03898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.999687631254238e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.9995782855831224e-05,
+      "loss": 0.9818,
       "step": 1374
     },
     {
-      "epoch": 0.03776331328444701,
+      "epoch": 0.03901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996854041499423e-05,
-      "loss": 1.1837,
+      "learning_rate": 1.999575612470411e-05,
+      "loss": 1.0999,
       "step": 1375
     },
     {
-      "epoch": 0.03779077751229024,
+      "epoch": 0.03904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996831691357483e-05,
-      "loss": 1.1181,
+      "learning_rate": 1.9995729309142396e-05,
+      "loss": 1.2104,
       "step": 1376
     },
     {
-      "epoch": 0.03781824174013348,
+      "epoch": 0.039074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.999680926211674e-05,
-      "loss": 1.2766,
+      "learning_rate": 1.9995702409146303e-05,
+      "loss": 1.0274,
       "step": 1377
     },
     {
-      "epoch": 0.03784570596797671,
+      "epoch": 0.03910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996786753777367e-05,
-      "loss": 1.1303,
+      "learning_rate": 1.9995675424716066e-05,
+      "loss": 1.1594,
       "step": 1378
     },
     {
-      "epoch": 0.03787317019581994,
+      "epoch": 0.039131668558456296,
       "grad_norm": 0.0,
-      "learning_rate": 1.999676416633955e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.9995648355851907e-05,
+      "loss": 1.0819,
       "step": 1379
     },
     {
-      "epoch": 0.037900634423663176,
+      "epoch": 0.03916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.999674149980346e-05,
-      "loss": 1.172,
+      "learning_rate": 1.9995621202554058e-05,
+      "loss": 1.0876,
       "step": 1380
     },
     {
-      "epoch": 0.03792809865150641,
+      "epoch": 0.039188422247446085,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996718754169277e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.9995593964822748e-05,
+      "loss": 1.1374,
       "step": 1381
     },
     {
-      "epoch": 0.03795556287934965,
+      "epoch": 0.03921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996695929437185e-05,
-      "loss": 1.0748,
+      "learning_rate": 1.9995566642658208e-05,
+      "loss": 1.0517,
       "step": 1382
     },
     {
-      "epoch": 0.03798302710719288,
+      "epoch": 0.03924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996673025607362e-05,
-      "loss": 1.0915,
+      "learning_rate": 1.9995539236060664e-05,
+      "loss": 1.057,
       "step": 1383
     },
     {
-      "epoch": 0.03801049133503612,
+      "epoch": 0.03927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.999665004267999e-05,
-      "loss": 1.1509,
+      "learning_rate": 1.9995511745030352e-05,
+      "loss": 1.1052,
       "step": 1384
     },
     {
-      "epoch": 0.03803795556287935,
+      "epoch": 0.039301929625425655,
       "grad_norm": 0.0,
-      "learning_rate": 1.999662698065525e-05,
-      "loss": 1.0988,
+      "learning_rate": 1.9995484169567503e-05,
+      "loss": 1.0792,
       "step": 1385
     },
     {
-      "epoch": 0.03806541979072258,
+      "epoch": 0.03933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996603839533327e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.999545650967235e-05,
+      "loss": 1.1333,
       "step": 1386
     },
     {
-      "epoch": 0.038092884018565816,
+      "epoch": 0.03935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.99965806193144e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.9995428765345125e-05,
+      "loss": 1.124,
       "step": 1387
     },
     {
-      "epoch": 0.03812034824640905,
+      "epoch": 0.03938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996557319998656e-05,
-      "loss": 1.1345,
+      "learning_rate": 1.9995400936586063e-05,
+      "loss": 1.1782,
       "step": 1388
     },
     {
-      "epoch": 0.038147812474252286,
+      "epoch": 0.03941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996533941586277e-05,
-      "loss": 1.1661,
+      "learning_rate": 1.9995373023395404e-05,
+      "loss": 1.0876,
       "step": 1389
     },
     {
-      "epoch": 0.03817527670209552,
+      "epoch": 0.03944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996510484077447e-05,
-      "loss": 1.1832,
+      "learning_rate": 1.9995345025773377e-05,
+      "loss": 1.1339,
       "step": 1390
     },
     {
-      "epoch": 0.03820274092993876,
+      "epoch": 0.03947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996486947472358e-05,
-      "loss": 1.08,
+      "learning_rate": 1.9995316943720225e-05,
+      "loss": 1.0193,
       "step": 1391
     },
     {
-      "epoch": 0.03823020515778199,
+      "epoch": 0.0395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.999646333177119e-05,
-      "loss": 1.0601,
+      "learning_rate": 1.9995288777236177e-05,
+      "loss": 1.097,
       "step": 1392
     },
     {
-      "epoch": 0.03825766938562522,
+      "epoch": 0.03952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996439636974134e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.9995260526321474e-05,
+      "loss": 1.158,
       "step": 1393
     },
     {
-      "epoch": 0.038285133613468456,
+      "epoch": 0.03955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.999641586308137e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.999523219097636e-05,
+      "loss": 1.1865,
       "step": 1394
     },
     {
-      "epoch": 0.03831259784131169,
+      "epoch": 0.03958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996392010093094e-05,
-      "loss": 1.0173,
+      "learning_rate": 1.9995203771201072e-05,
+      "loss": 1.1821,
       "step": 1395
     },
     {
-      "epoch": 0.038340062069154926,
+      "epoch": 0.039614074914869464,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996368078009495e-05,
-      "loss": 1.0705,
+      "learning_rate": 1.9995175266995847e-05,
+      "loss": 1.1078,
       "step": 1396
     },
     {
-      "epoch": 0.03836752629699816,
+      "epoch": 0.03964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996344066830758e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.9995146678360925e-05,
+      "loss": 1.2579,
       "step": 1397
     },
     {
-      "epoch": 0.038394990524841396,
+      "epoch": 0.03967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996319976557074e-05,
-      "loss": 1.095,
+      "learning_rate": 1.999511800529655e-05,
+      "loss": 1.14,
       "step": 1398
     },
     {
-      "epoch": 0.03842245475268463,
+      "epoch": 0.03969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996295807188636e-05,
-      "loss": 1.1134,
+      "learning_rate": 1.9995089247802966e-05,
+      "loss": 1.0497,
       "step": 1399
     },
     {
-      "epoch": 0.03844991898052786,
+      "epoch": 0.039727582292849034,
       "grad_norm": 0.0,
-      "learning_rate": 1.999627155872563e-05,
-      "loss": 1.2131,
+      "learning_rate": 1.9995060405880412e-05,
+      "loss": 1.0768,
       "step": 1400
     },
     {
-      "epoch": 0.038477383208371095,
+      "epoch": 0.03975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996247231168253e-05,
-      "loss": 1.1008,
+      "learning_rate": 1.999503147952913e-05,
+      "loss": 1.0794,
       "step": 1401
     },
     {
-      "epoch": 0.03850484743621433,
+      "epoch": 0.03978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996222824516694e-05,
-      "loss": 1.0741,
+      "learning_rate": 1.9995002468749374e-05,
+      "loss": 1.0789,
       "step": 1402
     },
     {
-      "epoch": 0.038532311664057566,
+      "epoch": 0.03981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.999619833877115e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.999497337354138e-05,
+      "loss": 1.0711,
       "step": 1403
     },
     {
-      "epoch": 0.0385597758919008,
+      "epoch": 0.039841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996173773931815e-05,
-      "loss": 1.0794,
+      "learning_rate": 1.9994944193905396e-05,
+      "loss": 1.0861,
       "step": 1404
     },
     {
-      "epoch": 0.038587240119744036,
+      "epoch": 0.0398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996149129998877e-05,
-      "loss": 1.109,
+      "learning_rate": 1.9994914929841672e-05,
+      "loss": 1.121,
       "step": 1405
     },
     {
-      "epoch": 0.038614704347587264,
+      "epoch": 0.039897843359818386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996124406972533e-05,
-      "loss": 1.2077,
+      "learning_rate": 1.9994885581350448e-05,
+      "loss": 1.1877,
       "step": 1406
     },
     {
-      "epoch": 0.0386421685754305,
+      "epoch": 0.03992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996099604852984e-05,
-      "loss": 1.09,
+      "learning_rate": 1.999485614843198e-05,
+      "loss": 1.24,
       "step": 1407
     },
     {
-      "epoch": 0.038669632803273735,
+      "epoch": 0.039954597048808174,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996074723640425e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.9994826631086514e-05,
+      "loss": 1.0684,
       "step": 1408
     },
     {
-      "epoch": 0.03869709703111697,
+      "epoch": 0.03998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996049763335047e-05,
-      "loss": 1.2659,
+      "learning_rate": 1.999479702931429e-05,
+      "loss": 1.0558,
       "step": 1409
     },
     {
-      "epoch": 0.038724561258960205,
+      "epoch": 0.040011350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 1.9996024723937053e-05,
-      "loss": 1.1536,
+      "learning_rate": 1.9994767343115574e-05,
+      "loss": 1.2299,
       "step": 1410
     },
     {
-      "epoch": 0.03875202548680344,
+      "epoch": 0.04003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.999599960544664e-05,
-      "loss": 1.1744,
+      "learning_rate": 1.9994737572490607e-05,
+      "loss": 1.0783,
       "step": 1411
     },
     {
-      "epoch": 0.038779489714646675,
+      "epoch": 0.040068104426787744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995974407864e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.9994707717439643e-05,
+      "loss": 1.0994,
       "step": 1412
     },
     {
-      "epoch": 0.038806953942489904,
+      "epoch": 0.04009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995949131189344e-05,
-      "loss": 1.1258,
+      "learning_rate": 1.9994677777962934e-05,
+      "loss": 1.1059,
       "step": 1413
     },
     {
-      "epoch": 0.03883441817033314,
+      "epoch": 0.040124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995923775422867e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.999464775406073e-05,
+      "loss": 1.0369,
       "step": 1414
     },
     {
-      "epoch": 0.038861882398176374,
+      "epoch": 0.04015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.999589834056477e-05,
-      "loss": 1.0884,
+      "learning_rate": 1.999461764573329e-05,
+      "loss": 1.0925,
       "step": 1415
     },
     {
-      "epoch": 0.03888934662601961,
+      "epoch": 0.04018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995872826615248e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.999458745298086e-05,
+      "loss": 1.0956,
       "step": 1416
     },
     {
-      "epoch": 0.038916810853862845,
+      "epoch": 0.0402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.999584723357451e-05,
-      "loss": 1.1547,
+      "learning_rate": 1.9994557175803707e-05,
+      "loss": 1.1477,
       "step": 1417
     },
     {
-      "epoch": 0.03894427508170608,
+      "epoch": 0.040238365493757096,
       "grad_norm": 0.0,
-      "learning_rate": 1.999582156144276e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.9994526814202076e-05,
+      "loss": 1.1672,
       "step": 1418
     },
     {
-      "epoch": 0.038971739309549315,
+      "epoch": 0.04026674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995795810220193e-05,
-      "loss": 1.1379,
+      "learning_rate": 1.9994496368176232e-05,
+      "loss": 1.0673,
       "step": 1419
     },
     {
-      "epoch": 0.03899920353739254,
+      "epoch": 0.04029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.999576997990702e-05,
-      "loss": 1.1479,
+      "learning_rate": 1.999446583772642e-05,
+      "loss": 1.0695,
       "step": 1420
     },
     {
-      "epoch": 0.03902666776523578,
+      "epoch": 0.04032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.999574407050344e-05,
-      "loss": 1.1394,
+      "learning_rate": 1.9994435222852913e-05,
+      "loss": 0.9402,
       "step": 1421
     },
     {
-      "epoch": 0.039054131993079014,
+      "epoch": 0.040351872871736666,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995718082009667e-05,
-      "loss": 1.1728,
+      "learning_rate": 1.999440452355596e-05,
+      "loss": 1.131,
       "step": 1422
     },
     {
-      "epoch": 0.03908159622092225,
+      "epoch": 0.04038024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995692014425897e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.9994373739835824e-05,
+      "loss": 1.1447,
       "step": 1423
     },
     {
-      "epoch": 0.039109060448765484,
+      "epoch": 0.04040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995665867752342e-05,
-      "loss": 1.0525,
+      "learning_rate": 1.9994342871692764e-05,
+      "loss": 0.9223,
       "step": 1424
     },
     {
-      "epoch": 0.03913652467660872,
+      "epoch": 0.04043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995639641989204e-05,
-      "loss": 1.1196,
+      "learning_rate": 1.999431191912704e-05,
+      "loss": 0.9412,
       "step": 1425
     },
     {
-      "epoch": 0.039163988904451955,
+      "epoch": 0.04046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.999561333713669e-05,
-      "loss": 1.2214,
+      "learning_rate": 1.9994280882138913e-05,
+      "loss": 1.1995,
       "step": 1426
     },
     {
-      "epoch": 0.03919145313229518,
+      "epoch": 0.04049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.999558695319502e-05,
-      "loss": 1.1845,
+      "learning_rate": 1.9994249760728645e-05,
+      "loss": 1.1981,
       "step": 1427
     },
     {
-      "epoch": 0.03921891736013842,
+      "epoch": 0.04052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995560490164386e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.9994218554896503e-05,
+      "loss": 1.0422,
       "step": 1428
     },
     {
-      "epoch": 0.03924638158798165,
+      "epoch": 0.040550510783200905,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995533948045013e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.9994187264642746e-05,
+      "loss": 1.0927,
       "step": 1429
     },
     {
-      "epoch": 0.03927384581582489,
+      "epoch": 0.0405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.99955073268371e-05,
-      "loss": 1.044,
+      "learning_rate": 1.999415588996764e-05,
+      "loss": 1.0881,
       "step": 1430
     },
     {
-      "epoch": 0.039301310043668124,
+      "epoch": 0.04060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.999548062654086e-05,
-      "loss": 1.1995,
+      "learning_rate": 1.9994124430871448e-05,
+      "loss": 1.0957,
       "step": 1431
     },
     {
-      "epoch": 0.03932877427151136,
+      "epoch": 0.04063564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995453847156506e-05,
-      "loss": 1.158,
+      "learning_rate": 1.999409288735444e-05,
+      "loss": 1.0766,
       "step": 1432
     },
     {
-      "epoch": 0.03935623849935459,
+      "epoch": 0.040664018161180475,
       "grad_norm": 0.0,
-      "learning_rate": 1.999542698868425e-05,
-      "loss": 1.066,
+      "learning_rate": 1.999406125941688e-05,
+      "loss": 1.085,
       "step": 1433
     },
     {
-      "epoch": 0.03938370272719782,
+      "epoch": 0.04069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.99954000511243e-05,
-      "loss": 1.1367,
+      "learning_rate": 1.9994029547059036e-05,
+      "loss": 1.0342,
       "step": 1434
     },
     {
-      "epoch": 0.03941116695504106,
+      "epoch": 0.04072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995373034476878e-05,
-      "loss": 1.2473,
+      "learning_rate": 1.9993997750281172e-05,
+      "loss": 0.9865,
       "step": 1435
     },
     {
-      "epoch": 0.03943863118288429,
+      "epoch": 0.04074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.999534593874219e-05,
-      "loss": 1.131,
+      "learning_rate": 1.9993965869083562e-05,
+      "loss": 1.1448,
       "step": 1436
     },
     {
-      "epoch": 0.03946609541072753,
+      "epoch": 0.040777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995318763920456e-05,
-      "loss": 1.049,
+      "learning_rate": 1.9993933903466472e-05,
+      "loss": 1.0605,
       "step": 1437
     },
     {
-      "epoch": 0.03949355963857076,
+      "epoch": 0.04080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995291510011884e-05,
-      "loss": 1.1998,
+      "learning_rate": 1.9993901853430175e-05,
+      "loss": 1.0706,
       "step": 1438
     },
     {
-      "epoch": 0.039521023866414,
+      "epoch": 0.040834279228149827,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995264177016696e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.999386971897494e-05,
+      "loss": 0.9794,
       "step": 1439
     },
     {
-      "epoch": 0.03954848809425723,
+      "epoch": 0.04086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995236764935105e-05,
-      "loss": 1.031,
+      "learning_rate": 1.9993837500101035e-05,
+      "loss": 1.1321,
       "step": 1440
     },
     {
-      "epoch": 0.03957595232210046,
+      "epoch": 0.040891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.999520927376733e-05,
-      "loss": 1.0489,
+      "learning_rate": 1.999380519680874e-05,
+      "loss": 0.9838,
       "step": 1441
     },
     {
-      "epoch": 0.0396034165499437,
+      "epoch": 0.04091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995181703513588e-05,
-      "loss": 1.1566,
+      "learning_rate": 1.999377280909832e-05,
+      "loss": 1.0408,
       "step": 1442
     },
     {
-      "epoch": 0.03963088077778693,
+      "epoch": 0.0409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995154054174095e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.999374033697005e-05,
+      "loss": 1.0914,
       "step": 1443
     },
     {
-      "epoch": 0.03965834500563017,
+      "epoch": 0.04097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995126325749073e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.9993707780424208e-05,
+      "loss": 1.0272,
       "step": 1444
     },
     {
-      "epoch": 0.0396858092334734,
+      "epoch": 0.041004540295119185,
       "grad_norm": 0.0,
-      "learning_rate": 1.999509851823874e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.999367513946107e-05,
+      "loss": 1.0509,
       "step": 1445
     },
     {
-      "epoch": 0.03971327346131664,
+      "epoch": 0.04103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995070631643313e-05,
-      "loss": 1.08,
+      "learning_rate": 1.9993642414080905e-05,
+      "loss": 1.0273,
       "step": 1446
     },
     {
-      "epoch": 0.039740737689159866,
+      "epoch": 0.04106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995042665963018e-05,
-      "loss": 1.0806,
+      "learning_rate": 1.9993609604283993e-05,
+      "loss": 1.2256,
       "step": 1447
     },
     {
-      "epoch": 0.0397682019170031,
+      "epoch": 0.04108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9995014621198072e-05,
-      "loss": 1.1971,
+      "learning_rate": 1.9993576710070615e-05,
+      "loss": 1.1097,
       "step": 1448
     },
     {
-      "epoch": 0.03979566614484634,
+      "epoch": 0.04111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.99949864973487e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.9993543731441046e-05,
+      "loss": 1.1451,
       "step": 1449
     },
     {
-      "epoch": 0.03982313037268957,
+      "epoch": 0.04114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994958294415123e-05,
-      "loss": 1.1105,
+      "learning_rate": 1.999351066839556e-05,
+      "loss": 1.079,
       "step": 1450
     },
     {
-      "epoch": 0.03985059460053281,
+      "epoch": 0.04117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994930012397563e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.9993477520934442e-05,
+      "loss": 1.1017,
       "step": 1451
     },
     {
-      "epoch": 0.03987805882837604,
+      "epoch": 0.04120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994901651296247e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.999344428905797e-05,
+      "loss": 1.0412,
       "step": 1452
     },
     {
-      "epoch": 0.03990552305621928,
+      "epoch": 0.04123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994873211111395e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.999341097276642e-05,
+      "loss": 1.1742,
       "step": 1453
     },
     {
-      "epoch": 0.039932987284062506,
+      "epoch": 0.04125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994844691843237e-05,
-      "loss": 1.148,
+      "learning_rate": 1.9993377572060084e-05,
+      "loss": 1.1304,
       "step": 1454
     },
     {
-      "epoch": 0.03996045151190574,
+      "epoch": 0.04128830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994816093491994e-05,
-      "loss": 1.0907,
+      "learning_rate": 1.999334408693924e-05,
+      "loss": 1.1185,
       "step": 1455
     },
     {
-      "epoch": 0.039987915739748976,
+      "epoch": 0.041316685584562994,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994787416057896e-05,
-      "loss": 1.112,
+      "learning_rate": 1.9993310517404165e-05,
+      "loss": 1.0686,
       "step": 1456
     },
     {
-      "epoch": 0.04001537996759221,
+      "epoch": 0.04134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.999475865954117e-05,
-      "loss": 1.1577,
+      "learning_rate": 1.999327686345515e-05,
+      "loss": 1.1513,
       "step": 1457
     },
     {
-      "epoch": 0.04004284419543545,
+      "epoch": 0.04137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.999472982394204e-05,
-      "loss": 1.1306,
+      "learning_rate": 1.9993243125092473e-05,
+      "loss": 1.0223,
       "step": 1458
     },
     {
-      "epoch": 0.04007030842327868,
+      "epoch": 0.04140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994700909260735e-05,
-      "loss": 1.1169,
+      "learning_rate": 1.9993209302316425e-05,
+      "loss": 1.1275,
       "step": 1459
     },
     {
-      "epoch": 0.04009777265112192,
+      "epoch": 0.041430192962542564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994671915497486e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9993175395127285e-05,
+      "loss": 1.278,
       "step": 1460
     },
     {
-      "epoch": 0.040125236878965145,
+      "epoch": 0.04145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.999464284265252e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.9993141403525346e-05,
+      "loss": 1.0834,
       "step": 1461
     },
     {
-      "epoch": 0.04015270110680838,
+      "epoch": 0.04148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.999461369072607e-05,
-      "loss": 1.1478,
+      "learning_rate": 1.9993107327510892e-05,
+      "loss": 1.0311,
       "step": 1462
     },
     {
-      "epoch": 0.040180165334651616,
+      "epoch": 0.04151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994584459718363e-05,
-      "loss": 1.1831,
+      "learning_rate": 1.999307316708421e-05,
+      "loss": 1.1417,
       "step": 1463
     },
     {
-      "epoch": 0.04020762956249485,
+      "epoch": 0.041543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994555149629634e-05,
-      "loss": 1.1418,
+      "learning_rate": 1.999303892224559e-05,
+      "loss": 1.0139,
       "step": 1464
     },
     {
-      "epoch": 0.040235093790338086,
+      "epoch": 0.04157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994525760460113e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.9993004592995323e-05,
+      "loss": 1.0518,
       "step": 1465
     },
     {
-      "epoch": 0.04026255801818132,
+      "epoch": 0.041600454029511916,
       "grad_norm": 0.0,
-      "learning_rate": 1.999449629221003e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.9992970179333695e-05,
+      "loss": 0.9776,
       "step": 1466
     },
     {
-      "epoch": 0.04029002224602455,
+      "epoch": 0.04162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994466744879625e-05,
-      "loss": 1.0169,
+      "learning_rate": 1.9992935681261e-05,
+      "loss": 1.1852,
       "step": 1467
     },
     {
-      "epoch": 0.040317486473867785,
+      "epoch": 0.041657207718501704,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994437118469126e-05,
-      "loss": 1.0698,
+      "learning_rate": 1.9992901098777525e-05,
+      "loss": 1.0026,
       "step": 1468
     },
     {
-      "epoch": 0.04034495070171102,
+      "epoch": 0.04168558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.999440741297877e-05,
-      "loss": 1.1407,
+      "learning_rate": 1.999286643188357e-05,
+      "loss": 1.0977,
       "step": 1469
     },
     {
-      "epoch": 0.040372414929554255,
+      "epoch": 0.041713961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 1.999437762840879e-05,
-      "loss": 1.1255,
+      "learning_rate": 1.999283168057942e-05,
+      "loss": 1.0851,
       "step": 1470
     },
     {
-      "epoch": 0.04039987915739749,
+      "epoch": 0.04174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994347764759422e-05,
-      "loss": 1.1318,
+      "learning_rate": 1.9992796844865375e-05,
+      "loss": 1.0076,
       "step": 1471
     },
     {
-      "epoch": 0.040427343385240726,
+      "epoch": 0.041770715096481274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994317822030905e-05,
-      "loss": 1.144,
+      "learning_rate": 1.9992761924741718e-05,
+      "loss": 1.1319,
       "step": 1472
     },
     {
-      "epoch": 0.04045480761308396,
+      "epoch": 0.04179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.999428780022347e-05,
-      "loss": 1.0975,
+      "learning_rate": 1.9992726920208757e-05,
+      "loss": 1.1639,
       "step": 1473
     },
     {
-      "epoch": 0.04048227184092719,
+      "epoch": 0.041827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.999425769933736e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.9992691831266785e-05,
+      "loss": 1.0253,
       "step": 1474
     },
     {
-      "epoch": 0.040509736068770424,
+      "epoch": 0.04185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994227519372817e-05,
-      "loss": 1.1429,
+      "learning_rate": 1.9992656657916093e-05,
+      "loss": 1.0916,
       "step": 1475
     },
     {
-      "epoch": 0.04053720029661366,
+      "epoch": 0.04188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994197260330068e-05,
-      "loss": 1.1343,
+      "learning_rate": 1.9992621400156982e-05,
+      "loss": 1.1104,
       "step": 1476
     },
     {
-      "epoch": 0.040564664524456895,
+      "epoch": 0.04191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.999416692220936e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.999258605798975e-05,
+      "loss": 1.0718,
       "step": 1477
     },
     {
-      "epoch": 0.04059212875230013,
+      "epoch": 0.041940976163450626,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994136505010934e-05,
-      "loss": 1.1547,
+      "learning_rate": 1.999255063141469e-05,
+      "loss": 1.0734,
       "step": 1478
     },
     {
-      "epoch": 0.040619592980143365,
+      "epoch": 0.04196935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 1.999410600873503e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.9992515120432106e-05,
+      "loss": 1.1507,
       "step": 1479
     },
     {
-      "epoch": 0.0406470572079866,
+      "epoch": 0.04199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994075433381884e-05,
-      "loss": 1.1655,
+      "learning_rate": 1.9992479525042305e-05,
+      "loss": 1.1052,
       "step": 1480
     },
     {
-      "epoch": 0.04067452143582983,
+      "epoch": 0.0420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9994044778951745e-05,
-      "loss": 1.124,
+      "learning_rate": 1.9992443845245572e-05,
+      "loss": 1.136,
       "step": 1481
     },
     {
-      "epoch": 0.040701985663673064,
+      "epoch": 0.042054483541430196,
       "grad_norm": 0.0,
-      "learning_rate": 1.999401404544485e-05,
-      "loss": 1.0949,
+      "learning_rate": 1.999240808104222e-05,
+      "loss": 1.1523,
       "step": 1482
     },
     {
-      "epoch": 0.0407294498915163,
+      "epoch": 0.04208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993983232861445e-05,
-      "loss": 1.2471,
+      "learning_rate": 1.9992372232432548e-05,
+      "loss": 1.054,
       "step": 1483
     },
     {
-      "epoch": 0.040756914119359534,
+      "epoch": 0.04211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993952341201775e-05,
-      "loss": 1.1098,
+      "learning_rate": 1.9992336299416857e-05,
+      "loss": 1.0933,
       "step": 1484
     },
     {
-      "epoch": 0.04078437834720277,
+      "epoch": 0.04213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993921370466078e-05,
-      "loss": 1.1487,
+      "learning_rate": 1.9992300281995454e-05,
+      "loss": 1.0848,
       "step": 1485
     },
     {
-      "epoch": 0.040811842575046005,
+      "epoch": 0.04216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.999389032065461e-05,
-      "loss": 1.1359,
+      "learning_rate": 1.999226418016864e-05,
+      "loss": 1.1591,
       "step": 1486
     },
     {
-      "epoch": 0.04083930680288924,
+      "epoch": 0.04219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.999385919176761e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9992227993936723e-05,
+      "loss": 1.1549,
       "step": 1487
     },
     {
-      "epoch": 0.04086677103073247,
+      "epoch": 0.04222474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993827983805318e-05,
-      "loss": 1.0532,
+      "learning_rate": 1.9992191723300004e-05,
+      "loss": 1.1093,
       "step": 1488
     },
     {
-      "epoch": 0.040894235258575704,
+      "epoch": 0.042253121452894435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993796696767992e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.9992155368258794e-05,
+      "loss": 1.2156,
       "step": 1489
     },
     {
-      "epoch": 0.04092169948641894,
+      "epoch": 0.04228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993765330655878e-05,
-      "loss": 1.1079,
+      "learning_rate": 1.99921189288134e-05,
+      "loss": 1.121,
       "step": 1490
     },
     {
-      "epoch": 0.040949163714262174,
+      "epoch": 0.04230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993733885469217e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.9992082404964126e-05,
+      "loss": 1.282,
       "step": 1491
     },
     {
-      "epoch": 0.04097662794210541,
+      "epoch": 0.04233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993702361208266e-05,
-      "loss": 1.0337,
+      "learning_rate": 1.9992045796711283e-05,
+      "loss": 1.1324,
       "step": 1492
     },
     {
-      "epoch": 0.041004092169948644,
+      "epoch": 0.042366628830874005,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993670757873273e-05,
-      "loss": 1.1542,
+      "learning_rate": 1.9992009104055178e-05,
+      "loss": 1.1132,
       "step": 1493
     },
     {
-      "epoch": 0.04103155639779187,
+      "epoch": 0.0423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.999363907546448e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.999197232699613e-05,
+      "loss": 1.0533,
       "step": 1494
     },
     {
-      "epoch": 0.04105902062563511,
+      "epoch": 0.04242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993607313982144e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.9991935465534434e-05,
+      "loss": 1.1573,
       "step": 1495
     },
     {
-      "epoch": 0.04108648485347834,
+      "epoch": 0.04245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993575473426517e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.9991898519670412e-05,
+      "loss": 1.1808,
       "step": 1496
     },
     {
-      "epoch": 0.04111394908132158,
+      "epoch": 0.042480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.999354355379785e-05,
-      "loss": 1.0218,
+      "learning_rate": 1.999186148940438e-05,
+      "loss": 1.0348,
       "step": 1497
     },
     {
-      "epoch": 0.041141413309164813,
+      "epoch": 0.04250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993511555096395e-05,
-      "loss": 1.1301,
+      "learning_rate": 1.999182437473664e-05,
+      "loss": 1.1705,
       "step": 1498
     },
     {
-      "epoch": 0.04116887753700805,
+      "epoch": 0.04253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993479477322405e-05,
-      "loss": 1.123,
+      "learning_rate": 1.999178717566751e-05,
+      "loss": 1.0907,
       "step": 1499
     },
     {
-      "epoch": 0.041196341764851284,
+      "epoch": 0.04256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993447320476135e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9991749892197304e-05,
+      "loss": 1.1386,
       "step": 1500
     },
     {
-      "epoch": 0.04122380599269451,
+      "epoch": 0.042593643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 1.999341508455784e-05,
-      "loss": 1.1098,
+      "learning_rate": 1.999171252432634e-05,
+      "loss": 1.1598,
       "step": 1501
     },
     {
-      "epoch": 0.04125127022053775,
+      "epoch": 0.04262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993382769567774e-05,
-      "loss": 1.2339,
+      "learning_rate": 1.9991675072054933e-05,
+      "loss": 0.9728,
       "step": 1502
     },
     {
-      "epoch": 0.04127873444838098,
+      "epoch": 0.04265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.999335037550619e-05,
-      "loss": 1.1478,
+      "learning_rate": 1.999163753538339e-05,
+      "loss": 1.0206,
       "step": 1503
     },
     {
-      "epoch": 0.04130619867622422,
+      "epoch": 0.04267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993317902373347e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.9991599914312043e-05,
+      "loss": 1.1821,
       "step": 1504
     },
     {
-      "epoch": 0.04133366290406745,
+      "epoch": 0.042707150964812715,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993285350169505e-05,
-      "loss": 1.0732,
+      "learning_rate": 1.99915622088412e-05,
+      "loss": 1.1884,
       "step": 1505
     },
     {
-      "epoch": 0.04136112713191069,
+      "epoch": 0.0427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993252718894913e-05,
-      "loss": 1.1031,
+      "learning_rate": 1.9991524418971182e-05,
+      "loss": 1.1673,
       "step": 1506
     },
     {
-      "epoch": 0.04138859135975392,
+      "epoch": 0.0427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.999322000854984e-05,
-      "loss": 0.9936,
+      "learning_rate": 1.9991486544702306e-05,
+      "loss": 1.1766,
       "step": 1507
     },
     {
-      "epoch": 0.04141605558759715,
+      "epoch": 0.04279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993187219134537e-05,
-      "loss": 1.1103,
+      "learning_rate": 1.9991448586034897e-05,
+      "loss": 1.1561,
       "step": 1508
     },
     {
-      "epoch": 0.04144351981544039,
+      "epoch": 0.04282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993154350649263e-05,
-      "loss": 1.1171,
+      "learning_rate": 1.999141054296927e-05,
+      "loss": 0.9215,
       "step": 1509
     },
     {
-      "epoch": 0.04147098404328362,
+      "epoch": 0.04284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993121403094283e-05,
-      "loss": 1.1772,
+      "learning_rate": 1.999137241550575e-05,
+      "loss": 1.0902,
       "step": 1510
     },
     {
-      "epoch": 0.04149844827112686,
+      "epoch": 0.04287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.999308837646986e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.999133420364466e-05,
+      "loss": 1.1257,
       "step": 1511
     },
     {
-      "epoch": 0.04152591249897009,
+      "epoch": 0.04290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993055270776244e-05,
-      "loss": 1.1703,
+      "learning_rate": 1.9991295907386317e-05,
+      "loss": 1.0868,
       "step": 1512
     },
     {
-      "epoch": 0.04155337672681333,
+      "epoch": 0.04293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.9993022086013706e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.9991257526731052e-05,
+      "loss": 1.0951,
       "step": 1513
     },
     {
-      "epoch": 0.04158084095465656,
+      "epoch": 0.04296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992988822182505e-05,
-      "loss": 1.1143,
+      "learning_rate": 1.9991219061679182e-05,
+      "loss": 1.0966,
       "step": 1514
     },
     {
-      "epoch": 0.04160830518249979,
+      "epoch": 0.04299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992955479282907e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.999118051223104e-05,
+      "loss": 1.1363,
       "step": 1515
     },
     {
-      "epoch": 0.041635769410343026,
+      "epoch": 0.043019296254256524,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992922057315175e-05,
-      "loss": 1.0713,
+      "learning_rate": 1.9991141878386944e-05,
+      "loss": 0.9785,
       "step": 1516
     },
     {
-      "epoch": 0.04166323363818626,
+      "epoch": 0.04304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.999288855627957e-05,
-      "loss": 1.1756,
+      "learning_rate": 1.9991103160147226e-05,
+      "loss": 1.2314,
       "step": 1517
     },
     {
-      "epoch": 0.0416906978660295,
+      "epoch": 0.04307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992854976176365e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.9991064357512208e-05,
+      "loss": 1.1276,
       "step": 1518
     },
     {
-      "epoch": 0.04171816209387273,
+      "epoch": 0.0431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992821317005818e-05,
-      "loss": 1.1765,
+      "learning_rate": 1.999102547048222e-05,
+      "loss": 1.0728,
       "step": 1519
     },
     {
-      "epoch": 0.04174562632171597,
+      "epoch": 0.043132803632236094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992787578768194e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.9990986499057593e-05,
+      "loss": 1.1172,
       "step": 1520
     },
     {
-      "epoch": 0.0417730905495592,
+      "epoch": 0.04316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.999275376146377e-05,
-      "loss": 1.1964,
+      "learning_rate": 1.9990947443238647e-05,
+      "loss": 1.0302,
       "step": 1521
     },
     {
-      "epoch": 0.04180055477740243,
+      "epoch": 0.04318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992719865092806e-05,
-      "loss": 1.1903,
+      "learning_rate": 1.999090830302573e-05,
+      "loss": 0.9705,
       "step": 1522
     },
     {
-      "epoch": 0.041828019005245666,
+      "epoch": 0.04321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.999268588965557e-05,
-      "loss": 1.1965,
+      "learning_rate": 1.9990869078419154e-05,
+      "loss": 1.0174,
       "step": 1523
     },
     {
-      "epoch": 0.0418554832330889,
+      "epoch": 0.043246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992651835152333e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.999082976941926e-05,
+      "loss": 1.0326,
       "step": 1524
     },
     {
-      "epoch": 0.041882947460932136,
+      "epoch": 0.04327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992617701583362e-05,
-      "loss": 1.1324,
+      "learning_rate": 1.9990790376026377e-05,
+      "loss": 1.0992,
       "step": 1525
     },
     {
-      "epoch": 0.04191041168877537,
+      "epoch": 0.043303064699205446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992583488948928e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.9990750898240837e-05,
+      "loss": 1.0937,
       "step": 1526
     },
     {
-      "epoch": 0.04193787591661861,
+      "epoch": 0.04333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992549197249307e-05,
-      "loss": 1.1664,
+      "learning_rate": 1.9990711336062976e-05,
+      "loss": 1.0822,
       "step": 1527
     },
     {
-      "epoch": 0.041965340144461835,
+      "epoch": 0.043359818388195234,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992514826484765e-05,
-      "loss": 1.1328,
+      "learning_rate": 1.9990671689493132e-05,
+      "loss": 1.0337,
       "step": 1528
     },
     {
-      "epoch": 0.04199280437230507,
+      "epoch": 0.04338819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992480376655573e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.9990631958531627e-05,
+      "loss": 1.1381,
       "step": 1529
     },
     {
-      "epoch": 0.042020268600148306,
+      "epoch": 0.043416572077185016,
       "grad_norm": 0.0,
-      "learning_rate": 1.999244584776201e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.999059214317881e-05,
+      "loss": 1.1149,
       "step": 1530
     },
     {
-      "epoch": 0.04204773282799154,
+      "epoch": 0.04344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.999241123980434e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.999055224343501e-05,
+      "loss": 1.0766,
       "step": 1531
     },
     {
-      "epoch": 0.042075197055834776,
+      "epoch": 0.043473325766174804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992376552782842e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.9990512259300567e-05,
+      "loss": 1.0404,
       "step": 1532
     },
     {
-      "epoch": 0.04210266128367801,
+      "epoch": 0.04350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.999234178669779e-05,
-      "loss": 1.0731,
+      "learning_rate": 1.999047219077582e-05,
+      "loss": 1.0703,
       "step": 1533
     },
     {
-      "epoch": 0.042130125511521246,
+      "epoch": 0.043530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992306941549466e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.9990432037861104e-05,
+      "loss": 1.0418,
       "step": 1534
     },
     {
-      "epoch": 0.042157589739364475,
+      "epoch": 0.04355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.999227201733813e-05,
-      "loss": 1.1249,
+      "learning_rate": 1.9990391800556757e-05,
+      "loss": 1.193,
       "step": 1535
     },
     {
-      "epoch": 0.04218505396720771,
+      "epoch": 0.04358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.999223701406407e-05,
-      "loss": 1.1137,
+      "learning_rate": 1.9990351478863124e-05,
+      "loss": 1.0692,
       "step": 1536
     },
     {
-      "epoch": 0.042212518195050945,
+      "epoch": 0.04361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.999220193172756e-05,
-      "loss": 1.2112,
+      "learning_rate": 1.999031107278054e-05,
+      "loss": 1.039,
       "step": 1537
     },
     {
-      "epoch": 0.04223998242289418,
+      "epoch": 0.043643586833144156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992166770328878e-05,
-      "loss": 1.1382,
+      "learning_rate": 1.9990270582309353e-05,
+      "loss": 1.1281,
       "step": 1538
     },
     {
-      "epoch": 0.042267446650737416,
+      "epoch": 0.04367196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.99921315298683e-05,
-      "loss": 1.1875,
+      "learning_rate": 1.9990230007449898e-05,
+      "loss": 1.0673,
       "step": 1539
     },
     {
-      "epoch": 0.04229491087858065,
+      "epoch": 0.04370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.999209621034611e-05,
-      "loss": 1.1365,
+      "learning_rate": 1.9990189348202526e-05,
+      "loss": 1.1174,
       "step": 1540
     },
     {
-      "epoch": 0.042322375106423886,
+      "epoch": 0.04372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992060811762585e-05,
-      "loss": 1.0805,
+      "learning_rate": 1.999014860456757e-05,
+      "loss": 0.9782,
       "step": 1541
     },
     {
-      "epoch": 0.042349839334267114,
+      "epoch": 0.043757094211123726,
       "grad_norm": 0.0,
-      "learning_rate": 1.9992025334118e-05,
-      "loss": 1.0887,
+      "learning_rate": 1.999010777654538e-05,
+      "loss": 1.1076,
       "step": 1542
     },
     {
-      "epoch": 0.04237730356211035,
+      "epoch": 0.04378547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991989777412644e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.9990066864136305e-05,
+      "loss": 1.0124,
       "step": 1543
     },
     {
-      "epoch": 0.042404767789953585,
+      "epoch": 0.04381384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991954141646794e-05,
-      "loss": 1.1426,
+      "learning_rate": 1.9990025867340683e-05,
+      "loss": 1.1193,
       "step": 1544
     },
     {
-      "epoch": 0.04243223201779682,
+      "epoch": 0.0438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.999191842682073e-05,
-      "loss": 1.1859,
+      "learning_rate": 1.9989984786158864e-05,
+      "loss": 1.0733,
       "step": 1545
     },
     {
-      "epoch": 0.042459696245640055,
+      "epoch": 0.04387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.999188263293474e-05,
-      "loss": 1.1261,
+      "learning_rate": 1.9989943620591195e-05,
+      "loss": 1.0951,
       "step": 1546
     },
     {
-      "epoch": 0.04248716047348329,
+      "epoch": 0.04389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.99918467599891e-05,
-      "loss": 1.127,
+      "learning_rate": 1.998990237063802e-05,
+      "loss": 1.0034,
       "step": 1547
     },
     {
-      "epoch": 0.042514624701326525,
+      "epoch": 0.04392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.99918108079841e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.9989861036299695e-05,
+      "loss": 1.1717,
       "step": 1548
     },
     {
-      "epoch": 0.042542088929169754,
+      "epoch": 0.043955732122587965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991774776920025e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.9989819617576562e-05,
+      "loss": 1.0789,
       "step": 1549
     },
     {
-      "epoch": 0.04256955315701299,
+      "epoch": 0.04398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.999173866679715e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9989778114468974e-05,
+      "loss": 1.1419,
       "step": 1550
     },
     {
-      "epoch": 0.042597017384856224,
+      "epoch": 0.04401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991702477615776e-05,
-      "loss": 1.019,
+      "learning_rate": 1.998973652697728e-05,
+      "loss": 1.0998,
       "step": 1551
     },
     {
-      "epoch": 0.04262448161269946,
+      "epoch": 0.04404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.999166620937618e-05,
-      "loss": 1.1271,
+      "learning_rate": 1.9989694855101837e-05,
+      "loss": 1.1829,
       "step": 1552
     },
     {
-      "epoch": 0.042651945840542695,
+      "epoch": 0.044069239500567535,
       "grad_norm": 0.0,
-      "learning_rate": 1.999162986207865e-05,
-      "loss": 1.072,
+      "learning_rate": 1.998965309884299e-05,
+      "loss": 1.1104,
       "step": 1553
     },
     {
-      "epoch": 0.04267941006838593,
+      "epoch": 0.04409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991593435723473e-05,
-      "loss": 1.17,
+      "learning_rate": 1.998961125820109e-05,
+      "loss": 1.1509,
       "step": 1554
     },
     {
-      "epoch": 0.04270687429622916,
+      "epoch": 0.044125993189557323,
       "grad_norm": 0.0,
-      "learning_rate": 1.999155693031094e-05,
-      "loss": 1.0809,
+      "learning_rate": 1.99895693331765e-05,
+      "loss": 1.1085,
       "step": 1555
     },
     {
-      "epoch": 0.04273433852407239,
+      "epoch": 0.04415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991520345841334e-05,
-      "loss": 1.2201,
+      "learning_rate": 1.9989527323769565e-05,
+      "loss": 1.1931,
       "step": 1556
     },
     {
-      "epoch": 0.04276180275191563,
+      "epoch": 0.044182746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991483682314953e-05,
-      "loss": 1.0883,
+      "learning_rate": 1.9989485229980648e-05,
+      "loss": 1.1035,
       "step": 1557
     },
     {
-      "epoch": 0.042789266979758864,
+      "epoch": 0.044211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991446939732083e-05,
-      "loss": 1.1515,
+      "learning_rate": 1.9989443051810098e-05,
+      "loss": 1.0297,
       "step": 1558
     },
     {
-      "epoch": 0.0428167312076021,
+      "epoch": 0.04423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991410118093012e-05,
-      "loss": 1.1153,
+      "learning_rate": 1.9989400789258272e-05,
+      "loss": 1.0768,
       "step": 1559
     },
     {
-      "epoch": 0.042844195435445334,
+      "epoch": 0.04426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991373217398032e-05,
-      "loss": 1.1674,
+      "learning_rate": 1.9989358442325532e-05,
+      "loss": 1.0315,
       "step": 1560
     },
     {
-      "epoch": 0.04287165966328857,
+      "epoch": 0.044296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.999133623764744e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.998931601101223e-05,
+      "loss": 1.1033,
       "step": 1561
     },
     {
-      "epoch": 0.0428991238911318,
+      "epoch": 0.04432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991299178841527e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.9989273495318726e-05,
+      "loss": 1.1325,
       "step": 1562
     },
     {
-      "epoch": 0.04292658811897503,
+      "epoch": 0.04435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.999126204098058e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.9989230895245382e-05,
+      "loss": 1.201,
       "step": 1563
     },
     {
-      "epoch": 0.04295405234681827,
+      "epoch": 0.04438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991224824064902e-05,
-      "loss": 1.1348,
+      "learning_rate": 1.998918821079255e-05,
+      "loss": 1.1086,
       "step": 1564
     },
     {
-      "epoch": 0.0429815165746615,
+      "epoch": 0.044409761634506245,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991187528094784e-05,
-      "loss": 1.1924,
+      "learning_rate": 1.9989145441960604e-05,
+      "loss": 1.1896,
       "step": 1565
     },
     {
-      "epoch": 0.04300898080250474,
+      "epoch": 0.04443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991150153070515e-05,
-      "loss": 1.1496,
+      "learning_rate": 1.998910258874989e-05,
+      "loss": 1.0022,
       "step": 1566
     },
     {
-      "epoch": 0.043036445030347974,
+      "epoch": 0.04446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991112698992397e-05,
-      "loss": 1.1633,
+      "learning_rate": 1.9989059651160786e-05,
+      "loss": 1.1567,
       "step": 1567
     },
     {
-      "epoch": 0.04306390925819121,
+      "epoch": 0.04449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991075165860727e-05,
-      "loss": 1.2357,
+      "learning_rate": 1.998901662919364e-05,
+      "loss": 0.9951,
       "step": 1568
     },
     {
-      "epoch": 0.04309137348603444,
+      "epoch": 0.04452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.9991037553675802e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.998897352284882e-05,
+      "loss": 1.1557,
       "step": 1569
     },
     {
-      "epoch": 0.04311883771387767,
+      "epoch": 0.0445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.999099986243791e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.9988930332126694e-05,
+      "loss": 1.0872,
       "step": 1570
     },
     {
-      "epoch": 0.04314630194172091,
+      "epoch": 0.0445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990962092147364e-05,
-      "loss": 1.1349,
+      "learning_rate": 1.9988887057027627e-05,
+      "loss": 1.1135,
       "step": 1571
     },
     {
-      "epoch": 0.04317376616956414,
+      "epoch": 0.04460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990924242804452e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.998884369755198e-05,
+      "loss": 1.1525,
       "step": 1572
     },
     {
-      "epoch": 0.04320123039740738,
+      "epoch": 0.04463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990886314409482e-05,
-      "loss": 1.186,
+      "learning_rate": 1.9988800253700118e-05,
+      "loss": 1.0186,
       "step": 1573
     },
     {
-      "epoch": 0.04322869462525061,
+      "epoch": 0.04466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990848306962744e-05,
-      "loss": 1.1252,
+      "learning_rate": 1.9988756725472417e-05,
+      "loss": 0.9871,
       "step": 1574
     },
     {
-      "epoch": 0.04325615885309385,
+      "epoch": 0.04469353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.999081022046455e-05,
-      "loss": 1.144,
+      "learning_rate": 1.9988713112869234e-05,
+      "loss": 1.1436,
       "step": 1575
     },
     {
-      "epoch": 0.04328362308093708,
+      "epoch": 0.044721906923950054,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990772054915187e-05,
-      "loss": 1.1851,
+      "learning_rate": 1.9988669415890943e-05,
+      "loss": 1.025,
       "step": 1576
     },
     {
-      "epoch": 0.04331108730878031,
+      "epoch": 0.04475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990733810314972e-05,
-      "loss": 1.1704,
+      "learning_rate": 1.9988625634537916e-05,
+      "loss": 1.1335,
       "step": 1577
     },
     {
-      "epoch": 0.04333855153662355,
+      "epoch": 0.04477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.99906954866642e-05,
-      "loss": 1.1309,
+      "learning_rate": 1.9988581768810517e-05,
+      "loss": 1.155,
       "step": 1578
     },
     {
-      "epoch": 0.04336601576446678,
+      "epoch": 0.04480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990657083963175e-05,
-      "loss": 1.1971,
+      "learning_rate": 1.998853781870912e-05,
+      "loss": 1.1158,
       "step": 1579
     },
     {
-      "epoch": 0.04339347999231002,
+      "epoch": 0.044835414301929624,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990618602212203e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.998849378423409e-05,
+      "loss": 1.155,
       "step": 1580
     },
     {
-      "epoch": 0.04342094422015325,
+      "epoch": 0.04486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990580041411583e-05,
-      "loss": 1.189,
+      "learning_rate": 1.998844966538581e-05,
+      "loss": 1.0762,
       "step": 1581
     },
     {
-      "epoch": 0.04344840844799649,
+      "epoch": 0.04489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990541401561624e-05,
-      "loss": 1.1293,
+      "learning_rate": 1.9988405462164642e-05,
+      "loss": 0.9684,
       "step": 1582
     },
     {
-      "epoch": 0.043475872675839716,
+      "epoch": 0.0449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990502682662636e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.998836117457097e-05,
+      "loss": 1.0786,
       "step": 1583
     },
     {
-      "epoch": 0.04350333690368295,
+      "epoch": 0.044948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990463884714915e-05,
-      "loss": 1.1084,
+      "learning_rate": 1.9988316802605158e-05,
+      "loss": 1.0676,
       "step": 1584
     },
     {
-      "epoch": 0.04353080113152619,
+      "epoch": 0.04497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990425007718777e-05,
-      "loss": 1.2344,
+      "learning_rate": 1.9988272346267587e-05,
+      "loss": 1.086,
       "step": 1585
     },
     {
-      "epoch": 0.04355826535936942,
+      "epoch": 0.045005675368898976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990386051674525e-05,
-      "loss": 1.0804,
+      "learning_rate": 1.998822780555863e-05,
+      "loss": 1.0784,
       "step": 1586
     },
     {
-      "epoch": 0.04358572958721266,
+      "epoch": 0.04503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990347016582468e-05,
-      "loss": 1.1159,
+      "learning_rate": 1.9988183180478658e-05,
+      "loss": 1.0782,
       "step": 1587
     },
     {
-      "epoch": 0.04361319381505589,
+      "epoch": 0.045062429057888764,
       "grad_norm": 0.0,
-      "learning_rate": 1.999030790244292e-05,
-      "loss": 1.1407,
+      "learning_rate": 1.998813847102806e-05,
+      "loss": 1.1075,
       "step": 1588
     },
     {
-      "epoch": 0.04364065804289912,
+      "epoch": 0.04509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.999026870925618e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.998809367720721e-05,
+      "loss": 1.1141,
       "step": 1589
     },
     {
-      "epoch": 0.043668122270742356,
+      "epoch": 0.045119182746878546,
       "grad_norm": 0.0,
-      "learning_rate": 1.999022943702257e-05,
-      "loss": 1.0571,
+      "learning_rate": 1.9988048799016476e-05,
+      "loss": 1.1046,
       "step": 1590
     },
     {
-      "epoch": 0.04369558649858559,
+      "epoch": 0.04514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.999019008574239e-05,
-      "loss": 1.1315,
+      "learning_rate": 1.9988003836456247e-05,
+      "loss": 1.0073,
       "step": 1591
     },
     {
-      "epoch": 0.043723050726428826,
+      "epoch": 0.045175936435868334,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990150655415958e-05,
-      "loss": 1.2513,
+      "learning_rate": 1.9987958789526902e-05,
+      "loss": 1.0355,
       "step": 1592
     },
     {
-      "epoch": 0.04375051495427206,
+      "epoch": 0.04520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990111146043585e-05,
-      "loss": 1.0316,
+      "learning_rate": 1.998791365822882e-05,
+      "loss": 1.0837,
       "step": 1593
     },
     {
-      "epoch": 0.0437779791821153,
+      "epoch": 0.045232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.999007155762558e-05,
-      "loss": 1.1522,
+      "learning_rate": 1.9987868442562383e-05,
+      "loss": 1.0924,
       "step": 1594
     },
     {
-      "epoch": 0.04380544340995853,
+      "epoch": 0.04526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.9990031890162264e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.9987823142527972e-05,
+      "loss": 1.0542,
       "step": 1595
     },
     {
-      "epoch": 0.04383290763780176,
+      "epoch": 0.0452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989992143653943e-05,
-      "loss": 1.1215,
+      "learning_rate": 1.998777775812597e-05,
+      "loss": 1.0951,
       "step": 1596
     },
     {
-      "epoch": 0.043860371865644995,
+      "epoch": 0.04531782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989952318100935e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.9987732289356757e-05,
+      "loss": 1.1334,
       "step": 1597
     },
     {
-      "epoch": 0.04388783609348823,
+      "epoch": 0.045346197502837686,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989912413503554e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.9987686736220726e-05,
+      "loss": 1.1186,
       "step": 1598
     },
     {
-      "epoch": 0.043915300321331466,
+      "epoch": 0.04537457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989872429862116e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.9987641098718255e-05,
+      "loss": 1.0501,
       "step": 1599
     },
     {
-      "epoch": 0.0439427645491747,
+      "epoch": 0.04540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989832367176938e-05,
-      "loss": 1.1783,
+      "learning_rate": 1.998759537684973e-05,
+      "loss": 1.0245,
       "step": 1600
     },
     {
-      "epoch": 0.043970228777017936,
+      "epoch": 0.04543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989792225448337e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.9987549570615534e-05,
+      "loss": 1.0498,
       "step": 1601
     },
     {
-      "epoch": 0.04399769300486117,
+      "epoch": 0.045459704880817256,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989752004676626e-05,
-      "loss": 1.073,
+      "learning_rate": 1.998750368001606e-05,
+      "loss": 1.088,
       "step": 1602
     },
     {
-      "epoch": 0.0440251572327044,
+      "epoch": 0.04548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989711704862132e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.9987457705051696e-05,
+      "loss": 1.0878,
       "step": 1603
     },
     {
-      "epoch": 0.044052621460547635,
+      "epoch": 0.04551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.998967132600517e-05,
-      "loss": 1.069,
+      "learning_rate": 1.9987411645722826e-05,
+      "loss": 1.1068,
       "step": 1604
     },
     {
-      "epoch": 0.04408008568839087,
+      "epoch": 0.04554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989630868106058e-05,
-      "loss": 1.0896,
+      "learning_rate": 1.998736550202984e-05,
+      "loss": 1.0068,
       "step": 1605
     },
     {
-      "epoch": 0.044107549916234105,
+      "epoch": 0.04557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989590331165117e-05,
-      "loss": 1.0698,
+      "learning_rate": 1.998731927397313e-05,
+      "loss": 1.1708,
       "step": 1606
     },
     {
-      "epoch": 0.04413501414407734,
+      "epoch": 0.04560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989549715182667e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.9987272961553087e-05,
+      "loss": 1.1256,
       "step": 1607
     },
     {
-      "epoch": 0.044162478371920576,
+      "epoch": 0.04562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.998950902015903e-05,
-      "loss": 1.1206,
+      "learning_rate": 1.9987226564770095e-05,
+      "loss": 1.0287,
       "step": 1608
     },
     {
-      "epoch": 0.04418994259976381,
+      "epoch": 0.045658342792281495,
       "grad_norm": 0.0,
-      "learning_rate": 1.998946824609453e-05,
-      "loss": 1.1543,
+      "learning_rate": 1.9987180083624557e-05,
+      "loss": 1.114,
       "step": 1609
     },
     {
-      "epoch": 0.04421740682760704,
+      "epoch": 0.04568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989427392989483e-05,
-      "loss": 1.1306,
+      "learning_rate": 1.9987133518116858e-05,
+      "loss": 1.0789,
       "step": 1610
     },
     {
-      "epoch": 0.044244871055450274,
+      "epoch": 0.04571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.998938646084422e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.998708686824739e-05,
+      "loss": 1.0861,
       "step": 1611
     },
     {
-      "epoch": 0.04427233528329351,
+      "epoch": 0.04574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989345449659063e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.998704013401655e-05,
+      "loss": 1.1345,
       "step": 1612
     },
     {
-      "epoch": 0.044299799511136745,
+      "epoch": 0.045771850170261065,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989304359434334e-05,
-      "loss": 1.1586,
+      "learning_rate": 1.998699331542474e-05,
+      "loss": 1.0867,
       "step": 1613
     },
     {
-      "epoch": 0.04432726373897998,
+      "epoch": 0.04580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.998926319017036e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.9986946412472342e-05,
+      "loss": 1.192,
       "step": 1614
     },
     {
-      "epoch": 0.044354727966823215,
+      "epoch": 0.045828603859250854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989221941867463e-05,
-      "loss": 1.1363,
+      "learning_rate": 1.9986899425159762e-05,
+      "loss": 0.9879,
       "step": 1615
     },
     {
-      "epoch": 0.044382192194666444,
+      "epoch": 0.04585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989180614525975e-05,
-      "loss": 1.077,
+      "learning_rate": 1.998685235348739e-05,
+      "loss": 1.1135,
       "step": 1616
     },
     {
-      "epoch": 0.04440965642250968,
+      "epoch": 0.045885357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989139208146222e-05,
-      "loss": 1.1896,
+      "learning_rate": 1.9986805197455628e-05,
+      "loss": 1.1932,
       "step": 1617
     },
     {
-      "epoch": 0.044437120650352914,
+      "epoch": 0.04591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989097722728528e-05,
-      "loss": 1.0579,
+      "learning_rate": 1.998675795706488e-05,
+      "loss": 1.0075,
       "step": 1618
     },
     {
-      "epoch": 0.04446458487819615,
+      "epoch": 0.04594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9989056158273227e-05,
-      "loss": 1.2067,
+      "learning_rate": 1.9986710632315533e-05,
+      "loss": 1.1204,
       "step": 1619
     },
     {
-      "epoch": 0.044492049106039384,
+      "epoch": 0.04597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.998901451478064e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.9986663223207993e-05,
+      "loss": 1.056,
       "step": 1620
     },
     {
-      "epoch": 0.04451951333388262,
+      "epoch": 0.045998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988972792251103e-05,
-      "loss": 1.0942,
+      "learning_rate": 1.998661572974266e-05,
+      "loss": 1.0387,
       "step": 1621
     },
     {
-      "epoch": 0.044546977561725855,
+      "epoch": 0.0460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988930990684944e-05,
-      "loss": 1.1451,
+      "learning_rate": 1.9986568151919935e-05,
+      "loss": 1.1643,
       "step": 1622
     },
     {
-      "epoch": 0.04457444178956908,
+      "epoch": 0.04605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.998888911008249e-05,
-      "loss": 1.1305,
+      "learning_rate": 1.998652048974022e-05,
+      "loss": 1.0826,
       "step": 1623
     },
     {
-      "epoch": 0.04460190601741232,
+      "epoch": 0.04608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988847150444083e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.998647274320392e-05,
+      "loss": 1.0145,
       "step": 1624
     },
     {
-      "epoch": 0.044629370245255554,
+      "epoch": 0.046112372304199775,
       "grad_norm": 0.0,
-      "learning_rate": 1.998880511177004e-05,
-      "loss": 1.024,
+      "learning_rate": 1.998642491231143e-05,
+      "loss": 1.2176,
       "step": 1625
     },
     {
-      "epoch": 0.04465683447309879,
+      "epoch": 0.04614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988762994060706e-05,
-      "loss": 1.092,
+      "learning_rate": 1.9986376997063163e-05,
+      "loss": 1.0624,
       "step": 1626
     },
     {
-      "epoch": 0.044684298700942024,
+      "epoch": 0.04616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.998872079731641e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.9986328997459523e-05,
+      "loss": 1.0152,
       "step": 1627
     },
     {
-      "epoch": 0.04471176292878526,
+      "epoch": 0.04619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988678521537485e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.9986280913500907e-05,
+      "loss": 1.0501,
       "step": 1628
     },
     {
-      "epoch": 0.044739227156628494,
+      "epoch": 0.04622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988636166724265e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.998623274518773e-05,
+      "loss": 1.091,
       "step": 1629
     },
     {
-      "epoch": 0.04476669138447172,
+      "epoch": 0.04625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988593732877088e-05,
-      "loss": 1.1602,
+      "learning_rate": 1.99861844925204e-05,
+      "loss": 1.1566,
       "step": 1630
     },
     {
-      "epoch": 0.04479415561231496,
+      "epoch": 0.04628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.998855121999629e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.9986136155499318e-05,
+      "loss": 1.0364,
       "step": 1631
     },
     {
-      "epoch": 0.04482161984015819,
+      "epoch": 0.04631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988508628082203e-05,
-      "loss": 1.1185,
+      "learning_rate": 1.9986087734124895e-05,
+      "loss": 1.132,
       "step": 1632
     },
     {
-      "epoch": 0.04484908406800143,
+      "epoch": 0.04633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988465957135166e-05,
-      "loss": 1.1825,
+      "learning_rate": 1.9986039228397536e-05,
+      "loss": 1.0508,
       "step": 1633
     },
     {
-      "epoch": 0.044876548295844663,
+      "epoch": 0.0463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.998842320715552e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.9985990638317663e-05,
+      "loss": 1.1039,
       "step": 1634
     },
     {
-      "epoch": 0.0449040125236879,
+      "epoch": 0.0463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.99883803781436e-05,
-      "loss": 1.1181,
+      "learning_rate": 1.9985941963885672e-05,
+      "loss": 1.0591,
       "step": 1635
     },
     {
-      "epoch": 0.044931476751531134,
+      "epoch": 0.046424517593643584,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988337470099743e-05,
-      "loss": 1.1384,
+      "learning_rate": 1.998589320510198e-05,
+      "loss": 1.0598,
       "step": 1636
     },
     {
-      "epoch": 0.04495894097937436,
+      "epoch": 0.04645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988294483024292e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.9985844361967005e-05,
+      "loss": 1.0894,
       "step": 1637
     },
     {
-      "epoch": 0.0449864052072176,
+      "epoch": 0.04648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988251416917585e-05,
-      "loss": 1.124,
+      "learning_rate": 1.9985795434481146e-05,
+      "loss": 1.0312,
       "step": 1638
     },
     {
-      "epoch": 0.04501386943506083,
+      "epoch": 0.04650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.998820827177997e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.9985746422644833e-05,
+      "loss": 1.1414,
       "step": 1639
     },
     {
-      "epoch": 0.04504133366290407,
+      "epoch": 0.046538024971623154,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988165047611773e-05,
-      "loss": 1.1958,
+      "learning_rate": 1.998569732645846e-05,
+      "loss": 1.2029,
       "step": 1640
     },
     {
-      "epoch": 0.0450687978907473,
+      "epoch": 0.04656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.998812174441335e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.9985648145922464e-05,
+      "loss": 0.9942,
       "step": 1641
     },
     {
-      "epoch": 0.04509626211859054,
+      "epoch": 0.04659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9988078362185036e-05,
-      "loss": 1.0861,
+      "learning_rate": 1.9985598881037243e-05,
+      "loss": 1.138,
       "step": 1642
     },
     {
-      "epoch": 0.04512372634643377,
+      "epoch": 0.04662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.998803490092718e-05,
-      "loss": 1.1608,
+      "learning_rate": 1.998554953180322e-05,
+      "loss": 1.0823,
       "step": 1643
     },
     {
-      "epoch": 0.045151190574277,
+      "epoch": 0.046651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.998799136064012e-05,
-      "loss": 1.2289,
+      "learning_rate": 1.9985500098220813e-05,
+      "loss": 0.9982,
       "step": 1644
     },
     {
-      "epoch": 0.04517865480212024,
+      "epoch": 0.04667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.998794774132421e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.9985450580290434e-05,
+      "loss": 1.1318,
       "step": 1645
     },
     {
-      "epoch": 0.04520611902996347,
+      "epoch": 0.046708286038592506,
       "grad_norm": 0.0,
-      "learning_rate": 1.998790404297978e-05,
-      "loss": 1.1471,
+      "learning_rate": 1.9985400978012508e-05,
+      "loss": 1.1391,
       "step": 1646
     },
     {
-      "epoch": 0.04523358325780671,
+      "epoch": 0.0467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987860265607187e-05,
-      "loss": 1.1214,
+      "learning_rate": 1.998535129138745e-05,
+      "loss": 1.0183,
       "step": 1647
     },
     {
-      "epoch": 0.04526104748564994,
+      "epoch": 0.046765039727582294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987816409206776e-05,
-      "loss": 1.2086,
+      "learning_rate": 1.998530152041568e-05,
+      "loss": 1.1011,
       "step": 1648
     },
     {
-      "epoch": 0.04528851171349318,
+      "epoch": 0.04679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.998777247377889e-05,
-      "loss": 1.1169,
+      "learning_rate": 1.998525166509762e-05,
+      "loss": 1.0677,
       "step": 1649
     },
     {
-      "epoch": 0.045315975941336406,
+      "epoch": 0.046821793416572076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987728459323882e-05,
-      "loss": 1.1726,
+      "learning_rate": 1.9985201725433693e-05,
+      "loss": 1.018,
       "step": 1650
     },
     {
-      "epoch": 0.04534344016917964,
+      "epoch": 0.04685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987684365842096e-05,
-      "loss": 1.0911,
+      "learning_rate": 1.9985151701424317e-05,
+      "loss": 1.1255,
       "step": 1651
     },
     {
-      "epoch": 0.045370904397022876,
+      "epoch": 0.046878547105561864,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987640193333882e-05,
-      "loss": 1.1061,
+      "learning_rate": 1.9985101593069914e-05,
+      "loss": 1.0742,
       "step": 1652
     },
     {
-      "epoch": 0.04539836862486611,
+      "epoch": 0.04690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987595941799594e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.998505140037091e-05,
+      "loss": 1.0602,
       "step": 1653
     },
     {
-      "epoch": 0.04542583285270935,
+      "epoch": 0.046935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987551611239575e-05,
-      "loss": 1.2385,
+      "learning_rate": 1.9985001123327728e-05,
+      "loss": 1.0327,
       "step": 1654
     },
     {
-      "epoch": 0.04545329708055258,
+      "epoch": 0.04696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987507201654178e-05,
-      "loss": 1.2244,
+      "learning_rate": 1.9984950761940792e-05,
+      "loss": 1.1322,
       "step": 1655
     },
     {
-      "epoch": 0.04548076130839582,
+      "epoch": 0.04699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987462713043757e-05,
-      "loss": 1.1581,
+      "learning_rate": 1.998490031621053e-05,
+      "loss": 1.1163,
       "step": 1656
     },
     {
-      "epoch": 0.045508225536239046,
+      "epoch": 0.04702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987418145408662e-05,
-      "loss": 1.0394,
+      "learning_rate": 1.9984849786137363e-05,
+      "loss": 1.1312,
       "step": 1657
     },
     {
-      "epoch": 0.04553568976408228,
+      "epoch": 0.047048808172531216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987373498749246e-05,
-      "loss": 1.1108,
+      "learning_rate": 1.9984799171721723e-05,
+      "loss": 1.1323,
       "step": 1658
     },
     {
-      "epoch": 0.045563153991925516,
+      "epoch": 0.0470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.998732877306586e-05,
-      "loss": 1.1799,
+      "learning_rate": 1.9984748472964034e-05,
+      "loss": 0.9733,
       "step": 1659
     },
     {
-      "epoch": 0.04559061821976875,
+      "epoch": 0.047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987283968358864e-05,
-      "loss": 1.1287,
+      "learning_rate": 1.998469768986473e-05,
+      "loss": 0.97,
       "step": 1660
     },
     {
-      "epoch": 0.045618082447611986,
+      "epoch": 0.04713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987239084628607e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.998464682242423e-05,
+      "loss": 1.1909,
       "step": 1661
     },
     {
-      "epoch": 0.04564554667545522,
+      "epoch": 0.047162315550510786,
       "grad_norm": 0.0,
-      "learning_rate": 1.998719412187544e-05,
-      "loss": 1.1774,
+      "learning_rate": 1.9984595870642972e-05,
+      "loss": 1.1464,
       "step": 1662
     },
     {
-      "epoch": 0.04567301090329846,
+      "epoch": 0.04719069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987149080099733e-05,
-      "loss": 1.005,
+      "learning_rate": 1.9984544834521385e-05,
+      "loss": 1.0805,
       "step": 1663
     },
     {
-      "epoch": 0.045700475131141685,
+      "epoch": 0.04721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.998710395930183e-05,
-      "loss": 1.1598,
+      "learning_rate": 1.9984493714059897e-05,
+      "loss": 1.0629,
       "step": 1664
     },
     {
-      "epoch": 0.04572793935898492,
+      "epoch": 0.04724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987058759482093e-05,
-      "loss": 1.2667,
+      "learning_rate": 1.9984442509258942e-05,
+      "loss": 1.0893,
       "step": 1665
     },
     {
-      "epoch": 0.045755403586828156,
+      "epoch": 0.04727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987013480640876e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.9984391220118953e-05,
+      "loss": 1.0031,
       "step": 1666
     },
     {
-      "epoch": 0.04578286781467139,
+      "epoch": 0.04730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.998696812277854e-05,
-      "loss": 1.1747,
+      "learning_rate": 1.9984339846640362e-05,
+      "loss": 1.0901,
       "step": 1667
     },
     {
-      "epoch": 0.045810332042514626,
+      "epoch": 0.04733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986922685895446e-05,
-      "loss": 1.0144,
+      "learning_rate": 1.9984288388823608e-05,
+      "loss": 1.0601,
       "step": 1668
     },
     {
-      "epoch": 0.04583779627035786,
+      "epoch": 0.047360953461975025,
       "grad_norm": 0.0,
-      "learning_rate": 1.998687716999195e-05,
-      "loss": 1.1742,
+      "learning_rate": 1.9984236846669115e-05,
+      "loss": 1.1797,
       "step": 1669
     },
     {
-      "epoch": 0.045865260498201096,
+      "epoch": 0.04738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986831575068413e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.9984185220177327e-05,
+      "loss": 1.1362,
       "step": 1670
     },
     {
-      "epoch": 0.045892724726044325,
+      "epoch": 0.047417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986785901125195e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.9984133509348678e-05,
+      "loss": 1.0934,
       "step": 1671
     },
     {
-      "epoch": 0.04592018895388756,
+      "epoch": 0.04744608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 1.998674014816266e-05,
-      "loss": 1.2347,
+      "learning_rate": 1.9984081714183604e-05,
+      "loss": 1.0096,
       "step": 1672
     },
     {
-      "epoch": 0.045947653181730795,
+      "epoch": 0.047474460839954595,
       "grad_norm": 0.0,
-      "learning_rate": 1.998669431618117e-05,
-      "loss": 1.1135,
+      "learning_rate": 1.9984029834682543e-05,
+      "loss": 0.9875,
       "step": 1673
     },
     {
-      "epoch": 0.04597511740957403,
+      "epoch": 0.04750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986648405181084e-05,
-      "loss": 1.1283,
+      "learning_rate": 1.998397787084593e-05,
+      "loss": 1.0656,
       "step": 1674
     },
     {
-      "epoch": 0.046002581637417266,
+      "epoch": 0.047531214528944384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986602415162765e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.998392582267421e-05,
+      "loss": 1.1882,
       "step": 1675
     },
     {
-      "epoch": 0.0460300458652605,
+      "epoch": 0.04755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986556346126586e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.9983873690167822e-05,
+      "loss": 1.0566,
       "step": 1676
     },
     {
-      "epoch": 0.04605751009310373,
+      "epoch": 0.047587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986510198072898e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.9983821473327202e-05,
+      "loss": 1.1548,
       "step": 1677
     },
     {
-      "epoch": 0.046084974320946964,
+      "epoch": 0.04761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.998646397100208e-05,
-      "loss": 1.2004,
+      "learning_rate": 1.9983769172152794e-05,
+      "loss": 1.1414,
       "step": 1678
     },
     {
-      "epoch": 0.0461124385487902,
+      "epoch": 0.04764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986417664914482e-05,
-      "loss": 1.1533,
+      "learning_rate": 1.9983716786645035e-05,
+      "loss": 1.0154,
       "step": 1679
     },
     {
-      "epoch": 0.046139902776633435,
+      "epoch": 0.04767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986371279810483e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9983664316804378e-05,
+      "loss": 1.0992,
       "step": 1680
     },
     {
-      "epoch": 0.04616736700447667,
+      "epoch": 0.047701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.998632481569045e-05,
-      "loss": 1.148,
+      "learning_rate": 1.9983611762631254e-05,
+      "loss": 1.0854,
       "step": 1681
     },
     {
-      "epoch": 0.046194831232319905,
+      "epoch": 0.04772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.998627827255474e-05,
-      "loss": 0.9512,
+      "learning_rate": 1.9983559124126116e-05,
+      "loss": 1.099,
       "step": 1682
     },
     {
-      "epoch": 0.04622229546016314,
+      "epoch": 0.04775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986231650403732e-05,
-      "loss": 1.1142,
+      "learning_rate": 1.9983506401289403e-05,
+      "loss": 1.0267,
       "step": 1683
     },
     {
-      "epoch": 0.04624975968800637,
+      "epoch": 0.04778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.998618494923779e-05,
-      "loss": 1.1177,
+      "learning_rate": 1.9983453594121565e-05,
+      "loss": 1.0981,
       "step": 1684
     },
     {
-      "epoch": 0.046277223915849604,
+      "epoch": 0.047814982973893305,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986138169057282e-05,
-      "loss": 1.0459,
+      "learning_rate": 1.9983400702623045e-05,
+      "loss": 0.9428,
       "step": 1685
     },
     {
-      "epoch": 0.04630468814369284,
+      "epoch": 0.04784335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986091309862586e-05,
-      "loss": 1.117,
+      "learning_rate": 1.998334772679429e-05,
+      "loss": 1.1086,
       "step": 1686
     },
     {
-      "epoch": 0.046332152371536074,
+      "epoch": 0.04787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.9986044371654064e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.9983294666635745e-05,
+      "loss": 1.1893,
       "step": 1687
     },
     {
-      "epoch": 0.04635961659937931,
+      "epoch": 0.04790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.998599735443209e-05,
-      "loss": 1.1556,
+      "learning_rate": 1.9983241522147864e-05,
+      "loss": 1.1312,
       "step": 1688
     },
     {
-      "epoch": 0.046387080827222545,
+      "epoch": 0.04792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985950258197036e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.9983188293331097e-05,
+      "loss": 1.1901,
       "step": 1689
     },
     {
-      "epoch": 0.04641454505506578,
+      "epoch": 0.04795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985903082949282e-05,
-      "loss": 1.1682,
+      "learning_rate": 1.9983134980185884e-05,
+      "loss": 1.0831,
       "step": 1690
     },
     {
-      "epoch": 0.04644200928290901,
+      "epoch": 0.04798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985855828689186e-05,
-      "loss": 1.0351,
+      "learning_rate": 1.9983081582712684e-05,
+      "loss": 1.0482,
       "step": 1691
     },
     {
-      "epoch": 0.04646947351075224,
+      "epoch": 0.04801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985808495417136e-05,
-      "loss": 1.1722,
+      "learning_rate": 1.9983028100911946e-05,
+      "loss": 1.0884,
       "step": 1692
     },
     {
-      "epoch": 0.04649693773859548,
+      "epoch": 0.04804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.99857610831335e-05,
-      "loss": 1.0854,
+      "learning_rate": 1.9982974534784118e-05,
+      "loss": 1.0925,
       "step": 1693
     },
     {
-      "epoch": 0.046524401966438714,
+      "epoch": 0.04807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.998571359183865e-05,
-      "loss": 1.1688,
+      "learning_rate": 1.9982920884329654e-05,
+      "loss": 0.9237,
       "step": 1694
     },
     {
-      "epoch": 0.04655186619428195,
+      "epoch": 0.04809875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.998566602153297e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.998286714954901e-05,
+      "loss": 1.0304,
       "step": 1695
     },
     {
-      "epoch": 0.046579330422125184,
+      "epoch": 0.048127128263337114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985618372216834e-05,
-      "loss": 1.123,
+      "learning_rate": 1.9982813330442644e-05,
+      "loss": 1.1534,
       "step": 1696
     },
     {
-      "epoch": 0.04660679464996842,
+      "epoch": 0.04815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985570643890613e-05,
-      "loss": 1.1326,
+      "learning_rate": 1.9982759427011e-05,
+      "loss": 1.1363,
       "step": 1697
     },
     {
-      "epoch": 0.04663425887781165,
+      "epoch": 0.0481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985522836554693e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.9982705439254533e-05,
+      "loss": 1.0915,
       "step": 1698
     },
     {
-      "epoch": 0.04666172310565488,
+      "epoch": 0.04821225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985474950209447e-05,
-      "loss": 1.1126,
+      "learning_rate": 1.9982651367173712e-05,
+      "loss": 0.9594,
       "step": 1699
     },
     {
-      "epoch": 0.04668918733349812,
+      "epoch": 0.048240635641316684,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985426984855255e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.9982597210768986e-05,
+      "loss": 1.0999,
       "step": 1700
     },
     {
-      "epoch": 0.04671665156134135,
+      "epoch": 0.04826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985378940492496e-05,
-      "loss": 1.1101,
+      "learning_rate": 1.998254297004081e-05,
+      "loss": 1.045,
       "step": 1701
     },
     {
-      "epoch": 0.04674411578918459,
+      "epoch": 0.04829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.998533081712155e-05,
-      "loss": 1.155,
+      "learning_rate": 1.9982488644989647e-05,
+      "loss": 1.028,
       "step": 1702
     },
     {
-      "epoch": 0.046771580017027824,
+      "epoch": 0.04832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985282614742802e-05,
-      "loss": 1.145,
+      "learning_rate": 1.9982434235615953e-05,
+      "loss": 1.1301,
       "step": 1703
     },
     {
-      "epoch": 0.04679904424487106,
+      "epoch": 0.048354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985234333356626e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.9982379741920187e-05,
+      "loss": 1.1125,
       "step": 1704
     },
     {
-      "epoch": 0.04682650847271429,
+      "epoch": 0.04838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.998518597296341e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.9982325163902812e-05,
+      "loss": 1.014,
       "step": 1705
     },
     {
-      "epoch": 0.04685397270055752,
+      "epoch": 0.048410896708286036,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985137533563534e-05,
-      "loss": 1.1435,
+      "learning_rate": 1.9982270501564286e-05,
+      "loss": 1.1206,
       "step": 1706
     },
     {
-      "epoch": 0.04688143692840076,
+      "epoch": 0.04843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9985089015157383e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.998221575490507e-05,
+      "loss": 1.0394,
       "step": 1707
     },
     {
-      "epoch": 0.04690890115624399,
+      "epoch": 0.048467650397275824,
       "grad_norm": 0.0,
-      "learning_rate": 1.998504041774534e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.9982160923925633e-05,
+      "loss": 1.0919,
       "step": 1708
     },
     {
-      "epoch": 0.04693636538408723,
+      "epoch": 0.04849602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984991741327787e-05,
-      "loss": 1.1042,
+      "learning_rate": 1.9982106008626434e-05,
+      "loss": 1.0678,
       "step": 1709
     },
     {
-      "epoch": 0.04696382961193046,
+      "epoch": 0.048524404086265606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984942985905114e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.9982051009007937e-05,
+      "loss": 1.0738,
       "step": 1710
     },
     {
-      "epoch": 0.04699129383977369,
+      "epoch": 0.0485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984894151477707e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.9981995925070607e-05,
+      "loss": 1.1398,
       "step": 1711
     },
     {
-      "epoch": 0.04701875806761693,
+      "epoch": 0.048581157775255394,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984845238045944e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.9981940756814904e-05,
+      "loss": 1.1462,
       "step": 1712
     },
     {
-      "epoch": 0.04704622229546016,
+      "epoch": 0.04860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.998479624561022e-05,
-      "loss": 1.1714,
+      "learning_rate": 1.99818855042413e-05,
+      "loss": 1.154,
       "step": 1713
     },
     {
-      "epoch": 0.0470736865233034,
+      "epoch": 0.048637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.998474717417092e-05,
-      "loss": 1.1518,
+      "learning_rate": 1.9981830167350262e-05,
+      "loss": 1.1026,
       "step": 1714
     },
     {
-      "epoch": 0.04710115075114663,
+      "epoch": 0.04866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984698023728435e-05,
-      "loss": 1.2137,
+      "learning_rate": 1.9981774746142254e-05,
+      "loss": 1.2369,
       "step": 1715
     },
     {
-      "epoch": 0.04712861497898987,
+      "epoch": 0.04869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.998464879428315e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.998171924061775e-05,
+      "loss": 1.0884,
       "step": 1716
     },
     {
-      "epoch": 0.0471560792068331,
+      "epoch": 0.04872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984599485835455e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.9981663650777213e-05,
+      "loss": 1.0955,
       "step": 1717
     },
     {
-      "epoch": 0.04718354343467633,
+      "epoch": 0.048751418842224746,
       "grad_norm": 0.0,
-      "learning_rate": 1.998455009838574e-05,
-      "loss": 1.0828,
+      "learning_rate": 1.9981607976621115e-05,
+      "loss": 1.0678,
       "step": 1718
     },
     {
-      "epoch": 0.047211007662519566,
+      "epoch": 0.04877979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.99845006319344e-05,
-      "loss": 1.09,
+      "learning_rate": 1.9981552218149924e-05,
+      "loss": 1.0985,
       "step": 1719
     },
     {
-      "epoch": 0.0472384718903628,
+      "epoch": 0.04880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.998445108648182e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.9981496375364115e-05,
+      "loss": 1.047,
       "step": 1720
     },
     {
-      "epoch": 0.04726593611820604,
+      "epoch": 0.04883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.99844014620284e-05,
-      "loss": 1.0556,
+      "learning_rate": 1.9981440448264155e-05,
+      "loss": 1.0356,
       "step": 1721
     },
     {
-      "epoch": 0.04729340034604927,
+      "epoch": 0.048864926220204316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984351758574526e-05,
-      "loss": 1.1493,
+      "learning_rate": 1.9981384436850518e-05,
+      "loss": 0.9744,
       "step": 1722
     },
     {
-      "epoch": 0.04732086457389251,
+      "epoch": 0.0488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984301976120593e-05,
-      "loss": 1.0899,
+      "learning_rate": 1.998132834112368e-05,
+      "loss": 1.1263,
       "step": 1723
     },
     {
-      "epoch": 0.04734832880173574,
+      "epoch": 0.0489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984252114666997e-05,
-      "loss": 1.1938,
+      "learning_rate": 1.9981272161084113e-05,
+      "loss": 0.9855,
       "step": 1724
     },
     {
-      "epoch": 0.04737579302957897,
+      "epoch": 0.04895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984202174214128e-05,
-      "loss": 1.1201,
+      "learning_rate": 1.9981215896732292e-05,
+      "loss": 1.1915,
       "step": 1725
     },
     {
-      "epoch": 0.047403257257422206,
+      "epoch": 0.04897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984152154762385e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.998115954806869e-05,
+      "loss": 1.1288,
       "step": 1726
     },
     {
-      "epoch": 0.04743072148526544,
+      "epoch": 0.04900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.998410205631216e-05,
-      "loss": 1.1157,
+      "learning_rate": 1.9981103115093786e-05,
+      "loss": 1.1508,
       "step": 1727
     },
     {
-      "epoch": 0.047458185713108676,
+      "epoch": 0.04903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984051878863856e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.9981046597808054e-05,
+      "loss": 1.0825,
       "step": 1728
     },
     {
-      "epoch": 0.04748564994095191,
+      "epoch": 0.049063564131668555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9984001622417865e-05,
-      "loss": 1.1666,
+      "learning_rate": 1.9980989996211974e-05,
+      "loss": 0.9093,
       "step": 1729
     },
     {
-      "epoch": 0.04751311416879515,
+      "epoch": 0.04909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983951286974586e-05,
-      "loss": 1.1549,
+      "learning_rate": 1.998093331030602e-05,
+      "loss": 1.1195,
       "step": 1730
     },
     {
-      "epoch": 0.04754057839663838,
+      "epoch": 0.049120317820658344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983900872534416e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.9980876540090675e-05,
+      "loss": 1.2018,
       "step": 1731
     },
     {
-      "epoch": 0.04756804262448161,
+      "epoch": 0.04914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983850379097754e-05,
-      "loss": 1.2313,
+      "learning_rate": 1.9980819685566416e-05,
+      "loss": 1.0881,
       "step": 1732
     },
     {
-      "epoch": 0.047595506852324845,
+      "epoch": 0.049177071509648125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983799806665e-05,
-      "loss": 1.0303,
+      "learning_rate": 1.998076274673373e-05,
+      "loss": 1.0943,
       "step": 1733
     },
     {
-      "epoch": 0.04762297108016808,
+      "epoch": 0.04920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.998374915523656e-05,
-      "loss": 1.1075,
+      "learning_rate": 1.9980705723593088e-05,
+      "loss": 1.0335,
       "step": 1734
     },
     {
-      "epoch": 0.047650435308011316,
+      "epoch": 0.049233825198637914,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983698424812825e-05,
-      "loss": 1.1476,
+      "learning_rate": 1.9980648616144974e-05,
+      "loss": 1.1083,
       "step": 1735
     },
     {
-      "epoch": 0.04767789953585455,
+      "epoch": 0.0492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.99836476153942e-05,
-      "loss": 1.1154,
+      "learning_rate": 1.9980591424389878e-05,
+      "loss": 1.0764,
       "step": 1736
     },
     {
-      "epoch": 0.047705363763697786,
+      "epoch": 0.049290578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.998359672698109e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.9980534148328275e-05,
+      "loss": 1.0922,
       "step": 1737
     },
     {
-      "epoch": 0.047732827991541014,
+      "epoch": 0.04931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983545759573894e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.998047678796065e-05,
+      "loss": 1.1416,
       "step": 1738
     },
     {
-      "epoch": 0.04776029221938425,
+      "epoch": 0.04934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983494713173017e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.998041934328749e-05,
+      "loss": 1.0356,
       "step": 1739
     },
     {
-      "epoch": 0.047787756447227485,
+      "epoch": 0.04937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983443587778863e-05,
-      "loss": 1.1851,
+      "learning_rate": 1.9980361814309282e-05,
+      "loss": 1.016,
       "step": 1740
     },
     {
-      "epoch": 0.04781522067507072,
+      "epoch": 0.049404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.998339238339184e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.9980304201026506e-05,
+      "loss": 1.0849,
       "step": 1741
     },
     {
-      "epoch": 0.047842684902913955,
+      "epoch": 0.04943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983341100012342e-05,
-      "loss": 1.0895,
+      "learning_rate": 1.998024650343965e-05,
+      "loss": 1.0447,
       "step": 1742
     },
     {
-      "epoch": 0.04787014913075719,
+      "epoch": 0.04946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983289737640786e-05,
-      "loss": 1.0855,
+      "learning_rate": 1.9980188721549207e-05,
+      "loss": 1.1741,
       "step": 1743
     },
     {
-      "epoch": 0.047897613358600426,
+      "epoch": 0.04948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983238296277576e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.9980130855355657e-05,
+      "loss": 1.1127,
       "step": 1744
     },
     {
-      "epoch": 0.047925077586443654,
+      "epoch": 0.049517593643586835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983186775923112e-05,
-      "loss": 1.1456,
+      "learning_rate": 1.9980072904859496e-05,
+      "loss": 1.0802,
       "step": 1745
     },
     {
-      "epoch": 0.04795254181428689,
+      "epoch": 0.04954597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983135176577812e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.9980014870061212e-05,
+      "loss": 1.1255,
       "step": 1746
     },
     {
-      "epoch": 0.047980006042130124,
+      "epoch": 0.04957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983083498242078e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.9979956750961293e-05,
+      "loss": 1.0358,
       "step": 1747
     },
     {
-      "epoch": 0.04800747026997336,
+      "epoch": 0.04960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9983031740916317e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.9979898547560227e-05,
+      "loss": 0.9149,
       "step": 1748
     },
     {
-      "epoch": 0.048034934497816595,
+      "epoch": 0.0496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982979904600946e-05,
-      "loss": 1.0824,
+      "learning_rate": 1.997984025985851e-05,
+      "loss": 1.0429,
       "step": 1749
     },
     {
-      "epoch": 0.04806239872565983,
+      "epoch": 0.04965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.998292798929637e-05,
-      "loss": 1.1148,
+      "learning_rate": 1.9979781887856638e-05,
+      "loss": 1.1014,
       "step": 1750
     },
     {
-      "epoch": 0.048089862953503065,
+      "epoch": 0.04968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982875995003003e-05,
-      "loss": 1.1648,
+      "learning_rate": 1.9979723431555098e-05,
+      "loss": 1.1538,
       "step": 1751
     },
     {
-      "epoch": 0.048117327181346294,
+      "epoch": 0.04971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982823921721247e-05,
-      "loss": 1.1505,
+      "learning_rate": 1.9979664890954383e-05,
+      "loss": 1.0909,
       "step": 1752
     },
     {
-      "epoch": 0.04814479140918953,
+      "epoch": 0.04974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982771769451528e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.997960626605499e-05,
+      "loss": 1.0879,
       "step": 1753
     },
     {
-      "epoch": 0.048172255637032764,
+      "epoch": 0.04977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982719538194248e-05,
-      "loss": 1.1726,
+      "learning_rate": 1.997954755685742e-05,
+      "loss": 0.9962,
       "step": 1754
     },
     {
-      "epoch": 0.048199719864876,
+      "epoch": 0.04980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982667227949828e-05,
-      "loss": 1.0727,
+      "learning_rate": 1.997948876336216e-05,
+      "loss": 1.0319,
       "step": 1755
     },
     {
-      "epoch": 0.048227184092719234,
+      "epoch": 0.049829738933030644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982614838718674e-05,
-      "loss": 1.1664,
+      "learning_rate": 1.9979429885569707e-05,
+      "loss": 1.0741,
       "step": 1756
     },
     {
-      "epoch": 0.04825464832056247,
+      "epoch": 0.04985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982562370501207e-05,
-      "loss": 0.9953,
+      "learning_rate": 1.9979370923480563e-05,
+      "loss": 1.1923,
       "step": 1757
     },
     {
-      "epoch": 0.048282112548405705,
+      "epoch": 0.04988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.998250982329784e-05,
-      "loss": 1.2021,
+      "learning_rate": 1.9979311877095228e-05,
+      "loss": 1.171,
       "step": 1758
     },
     {
-      "epoch": 0.04830957677624893,
+      "epoch": 0.04991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982457197108985e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.997925274641419e-05,
+      "loss": 1.1318,
       "step": 1759
     },
     {
-      "epoch": 0.04833704100409217,
+      "epoch": 0.049943246311010214,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982404491935067e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.9979193531437962e-05,
+      "loss": 1.1082,
       "step": 1760
     },
     {
-      "epoch": 0.048364505231935404,
+      "epoch": 0.04997162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982351707776492e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.997913423216703e-05,
+      "loss": 1.1227,
       "step": 1761
     },
     {
-      "epoch": 0.04839196945977864,
+      "epoch": 0.05,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982298844633687e-05,
-      "loss": 1.0868,
+      "learning_rate": 1.997907484860191e-05,
+      "loss": 1.1214,
       "step": 1762
     },
     {
-      "epoch": 0.048419433687621874,
+      "epoch": 0.05002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982245902507066e-05,
-      "loss": 1.0891,
+      "learning_rate": 1.9979015380743093e-05,
+      "loss": 1.1306,
       "step": 1763
     },
     {
-      "epoch": 0.04844689791546511,
+      "epoch": 0.050056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982192881397048e-05,
-      "loss": 1.1758,
+      "learning_rate": 1.9978955828591085e-05,
+      "loss": 1.1901,
       "step": 1764
     },
     {
-      "epoch": 0.04847436214330834,
+      "epoch": 0.05008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982139781304054e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.997889619214639e-05,
+      "loss": 0.9865,
       "step": 1765
     },
     {
-      "epoch": 0.04850182637115157,
+      "epoch": 0.050113507377979566,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982086602228504e-05,
-      "loss": 1.1748,
+      "learning_rate": 1.9978836471409506e-05,
+      "loss": 1.1825,
       "step": 1766
     },
     {
-      "epoch": 0.04852929059899481,
+      "epoch": 0.05014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982033344170818e-05,
-      "loss": 1.1643,
+      "learning_rate": 1.9978776666380945e-05,
+      "loss": 0.9525,
       "step": 1767
     },
     {
-      "epoch": 0.04855675482683804,
+      "epoch": 0.050170261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981980007131416e-05,
-      "loss": 1.1262,
+      "learning_rate": 1.997871677706121e-05,
+      "loss": 0.9607,
       "step": 1768
     },
     {
-      "epoch": 0.04858421905468128,
+      "epoch": 0.05019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981926591110725e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9978656803450804e-05,
+      "loss": 0.9821,
       "step": 1769
     },
     {
-      "epoch": 0.048611683282524513,
+      "epoch": 0.050227014755959136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981873096109158e-05,
-      "loss": 1.1094,
+      "learning_rate": 1.9978596745550236e-05,
+      "loss": 1.1352,
       "step": 1770
     },
     {
-      "epoch": 0.04863914751036775,
+      "epoch": 0.05025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.998181952212715e-05,
-      "loss": 1.1259,
+      "learning_rate": 1.997853660336001e-05,
+      "loss": 1.1225,
       "step": 1771
     },
     {
-      "epoch": 0.04866661173821098,
+      "epoch": 0.050283768444948924,
       "grad_norm": 0.0,
-      "learning_rate": 1.998176586916512e-05,
-      "loss": 1.1649,
+      "learning_rate": 1.997847637688064e-05,
+      "loss": 1.0872,
       "step": 1772
     },
     {
-      "epoch": 0.04869407596605421,
+      "epoch": 0.05031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981712137223486e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.9978416066112628e-05,
+      "loss": 1.1976,
       "step": 1773
     },
     {
-      "epoch": 0.04872154019389745,
+      "epoch": 0.050340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981658326302682e-05,
-      "loss": 1.1734,
+      "learning_rate": 1.997835567105649e-05,
+      "loss": 1.1021,
       "step": 1774
     },
     {
-      "epoch": 0.04874900442174068,
+      "epoch": 0.0503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981604436403134e-05,
-      "loss": 1.2246,
+      "learning_rate": 1.9978295191712736e-05,
+      "loss": 1.115,
       "step": 1775
     },
     {
-      "epoch": 0.04877646864958392,
+      "epoch": 0.05039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.998155046752526e-05,
-      "loss": 1.134,
+      "learning_rate": 1.997823462808187e-05,
+      "loss": 1.0061,
       "step": 1776
     },
     {
-      "epoch": 0.04880393287742715,
+      "epoch": 0.05042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981496419669495e-05,
-      "loss": 1.21,
+      "learning_rate": 1.997817398016441e-05,
+      "loss": 1.1981,
       "step": 1777
     },
     {
-      "epoch": 0.04883139710527039,
+      "epoch": 0.050454029511918276,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981442292836264e-05,
-      "loss": 1.1331,
+      "learning_rate": 1.9978113247960862e-05,
+      "loss": 1.1353,
       "step": 1778
     },
     {
-      "epoch": 0.048858861333113617,
+      "epoch": 0.05048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981388087025993e-05,
-      "loss": 1.186,
+      "learning_rate": 1.9978052431471742e-05,
+      "loss": 1.1035,
       "step": 1779
     },
     {
-      "epoch": 0.04888632556095685,
+      "epoch": 0.05051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981333802239116e-05,
-      "loss": 1.1276,
+      "learning_rate": 1.9977991530697566e-05,
+      "loss": 1.0378,
       "step": 1780
     },
     {
-      "epoch": 0.04891378978880009,
+      "epoch": 0.05053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.998127943847606e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.997793054563885e-05,
+      "loss": 1.0845,
       "step": 1781
     },
     {
-      "epoch": 0.04894125401664332,
+      "epoch": 0.050567536889897846,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981224995737252e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.9977869476296105e-05,
+      "loss": 0.9818,
       "step": 1782
     },
     {
-      "epoch": 0.04896871824448656,
+      "epoch": 0.05059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981170474023126e-05,
-      "loss": 1.1602,
+      "learning_rate": 1.9977808322669848e-05,
+      "loss": 1.0648,
       "step": 1783
     },
     {
-      "epoch": 0.04899618247232979,
+      "epoch": 0.05062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.998111587333411e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.9977747084760594e-05,
+      "loss": 1.0585,
       "step": 1784
     },
     {
-      "epoch": 0.04902364670017303,
+      "epoch": 0.05065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981061193670642e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.997768576256886e-05,
+      "loss": 1.0356,
       "step": 1785
     },
     {
-      "epoch": 0.049051110928016256,
+      "epoch": 0.05068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.9981006435033154e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.997762435609517e-05,
+      "loss": 1.0701,
       "step": 1786
     },
     {
-      "epoch": 0.04907857515585949,
+      "epoch": 0.050709421112372303,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980951597422075e-05,
-      "loss": 1.093,
+      "learning_rate": 1.997756286534004e-05,
+      "loss": 1.143,
       "step": 1787
     },
     {
-      "epoch": 0.049106039383702726,
+      "epoch": 0.0507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.998089668083784e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9977501290303984e-05,
+      "loss": 1.0034,
       "step": 1788
     },
     {
-      "epoch": 0.04913350361154596,
+      "epoch": 0.05076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980841685280885e-05,
-      "loss": 1.1698,
+      "learning_rate": 1.9977439630987527e-05,
+      "loss": 1.172,
       "step": 1789
     },
     {
-      "epoch": 0.0491609678393892,
+      "epoch": 0.05079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980786610751645e-05,
-      "loss": 1.1804,
+      "learning_rate": 1.997737788739119e-05,
+      "loss": 1.0299,
       "step": 1790
     },
     {
-      "epoch": 0.04918843206723243,
+      "epoch": 0.050822928490351874,
       "grad_norm": 0.0,
-      "learning_rate": 1.998073145725055e-05,
-      "loss": 1.1175,
+      "learning_rate": 1.997731605951549e-05,
+      "loss": 1.0405,
       "step": 1791
     },
     {
-      "epoch": 0.04921589629507567,
+      "epoch": 0.05085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.998067622477805e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.9977254147360958e-05,
+      "loss": 1.1224,
       "step": 1792
     },
     {
-      "epoch": 0.049243360522918896,
+      "epoch": 0.050879682179341655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980620913334564e-05,
-      "loss": 1.1707,
+      "learning_rate": 1.997719215092811e-05,
+      "loss": 1.1185,
       "step": 1793
     },
     {
-      "epoch": 0.04927082475076213,
+      "epoch": 0.05090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980565522920548e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.997713007021747e-05,
+      "loss": 1.0739,
       "step": 1794
     },
     {
-      "epoch": 0.049298288978605366,
+      "epoch": 0.050936435868331444,
       "grad_norm": 0.0,
-      "learning_rate": 1.998051005353643e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.997706790522956e-05,
+      "loss": 1.133,
       "step": 1795
     },
     {
-      "epoch": 0.0493257532064486,
+      "epoch": 0.05096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980454505182644e-05,
-      "loss": 1.1835,
+      "learning_rate": 1.9977005655964916e-05,
+      "loss": 1.1054,
       "step": 1796
     },
     {
-      "epoch": 0.049353217434291836,
+      "epoch": 0.050993189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 1.998039887785964e-05,
-      "loss": 1.0141,
+      "learning_rate": 1.9976943322424053e-05,
+      "loss": 1.0506,
       "step": 1797
     },
     {
-      "epoch": 0.04938068166213507,
+      "epoch": 0.05102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980343171567855e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.99768809046075e-05,
+      "loss": 1.0236,
       "step": 1798
     },
     {
-      "epoch": 0.0494081458899783,
+      "epoch": 0.051049943246311014,
       "grad_norm": 0.0,
-      "learning_rate": 1.998028738630773e-05,
-      "loss": 1.23,
+      "learning_rate": 1.997681840251579e-05,
+      "loss": 1.1409,
       "step": 1799
     },
     {
-      "epoch": 0.049435610117821535,
+      "epoch": 0.0510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.99802315220797e-05,
-      "loss": 1.1166,
+      "learning_rate": 1.9976755816149443e-05,
+      "loss": 1.1895,
       "step": 1800
     },
     {
-      "epoch": 0.04946307434566477,
+      "epoch": 0.051106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980175578884218e-05,
-      "loss": 1.1113,
+      "learning_rate": 1.9976693145508994e-05,
+      "loss": 1.0565,
       "step": 1801
     },
     {
-      "epoch": 0.049490538573508006,
+      "epoch": 0.05113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.998011955672172e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.9976630390594966e-05,
+      "loss": 1.0639,
       "step": 1802
     },
     {
-      "epoch": 0.04951800280135124,
+      "epoch": 0.05116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980063455592647e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.9976567551407896e-05,
+      "loss": 1.1689,
       "step": 1803
     },
     {
-      "epoch": 0.049545467029194476,
+      "epoch": 0.05119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.9980007275497443e-05,
-      "loss": 1.2063,
+      "learning_rate": 1.9976504627948308e-05,
+      "loss": 1.0856,
       "step": 1804
     },
     {
-      "epoch": 0.04957293125703771,
+      "epoch": 0.051220204313280365,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979951016436564e-05,
-      "loss": 1.1231,
+      "learning_rate": 1.997644162021674e-05,
+      "loss": 0.9677,
       "step": 1805
     },
     {
-      "epoch": 0.04960039548488094,
+      "epoch": 0.05124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.997989467841044e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.997637852821372e-05,
+      "loss": 0.9794,
       "step": 1806
     },
     {
-      "epoch": 0.049627859712724175,
+      "epoch": 0.05127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.997983826141953e-05,
-      "loss": 1.127,
+      "learning_rate": 1.9976315351939782e-05,
+      "loss": 1.1109,
       "step": 1807
     },
     {
-      "epoch": 0.04965532394056741,
+      "epoch": 0.05130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979781765464267e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.997625209139546e-05,
+      "loss": 1.097,
       "step": 1808
     },
     {
-      "epoch": 0.049682788168410645,
+      "epoch": 0.051333711691259935,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979725190545106e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.9976188746581288e-05,
+      "loss": 1.2153,
       "step": 1809
     },
     {
-      "epoch": 0.04971025239625388,
+      "epoch": 0.05136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979668536662498e-05,
-      "loss": 1.036,
+      "learning_rate": 1.99761253174978e-05,
+      "loss": 1.0952,
       "step": 1810
     },
     {
-      "epoch": 0.049737716624097116,
+      "epoch": 0.05139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979611803816883e-05,
-      "loss": 1.1056,
+      "learning_rate": 1.9976061804145535e-05,
+      "loss": 1.0575,
       "step": 1811
     },
     {
-      "epoch": 0.04976518085194035,
+      "epoch": 0.05141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979554992008716e-05,
-      "loss": 1.1059,
+      "learning_rate": 1.9975998206525024e-05,
+      "loss": 1.1812,
       "step": 1812
     },
     {
-      "epoch": 0.04979264507978358,
+      "epoch": 0.0514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.997949810123844e-05,
-      "loss": 1.0726,
+      "learning_rate": 1.997593452463681e-05,
+      "loss": 1.1626,
       "step": 1813
     },
     {
-      "epoch": 0.049820109307626814,
+      "epoch": 0.05147559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979441131506513e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.997587075848143e-05,
+      "loss": 1.0648,
       "step": 1814
     },
     {
-      "epoch": 0.04984757353547005,
+      "epoch": 0.05150397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 1.997938408281338e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.997580690805942e-05,
+      "loss": 0.9951,
       "step": 1815
     },
     {
-      "epoch": 0.049875037763313285,
+      "epoch": 0.051532349602724174,
       "grad_norm": 0.0,
-      "learning_rate": 1.99793269551595e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.997574297337132e-05,
+      "loss": 1.0155,
       "step": 1816
     },
     {
-      "epoch": 0.04990250199115652,
+      "epoch": 0.05156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979269748545315e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.9975678954417672e-05,
+      "loss": 1.0792,
       "step": 1817
     },
     {
-      "epoch": 0.049929966218999755,
+      "epoch": 0.05158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.997921246297128e-05,
-      "loss": 1.1407,
+      "learning_rate": 1.9975614851199015e-05,
+      "loss": 1.056,
       "step": 1818
     },
     {
-      "epoch": 0.04995743044684299,
+      "epoch": 0.05161748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979155098437853e-05,
-      "loss": 1.0754,
+      "learning_rate": 1.997555066371589e-05,
+      "loss": 1.0553,
       "step": 1819
     },
     {
-      "epoch": 0.04998489467468622,
+      "epoch": 0.051645856980703744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979097654945487e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.997548639196884e-05,
+      "loss": 1.2516,
       "step": 1820
     },
     {
-      "epoch": 0.050012358902529454,
+      "epoch": 0.05167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.997904013249463e-05,
-      "loss": 1.0934,
+      "learning_rate": 1.997542203595841e-05,
+      "loss": 1.1085,
       "step": 1821
     },
     {
-      "epoch": 0.05003982313037269,
+      "epoch": 0.05170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.997898253108575e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.997535759568514e-05,
+      "loss": 1.1097,
       "step": 1822
     },
     {
-      "epoch": 0.050067287358215924,
+      "epoch": 0.05173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978924850719287e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9975293071149577e-05,
+      "loss": 0.9525,
       "step": 1823
     },
     {
-      "epoch": 0.05009475158605916,
+      "epoch": 0.051759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978867091395708e-05,
-      "loss": 1.1384,
+      "learning_rate": 1.9975228462352265e-05,
+      "loss": 1.0793,
       "step": 1824
     },
     {
-      "epoch": 0.050122215813902395,
+      "epoch": 0.05178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978809253115468e-05,
-      "loss": 1.2151,
+      "learning_rate": 1.9975163769293748e-05,
+      "loss": 1.0842,
       "step": 1825
     },
     {
-      "epoch": 0.05014968004174562,
+      "epoch": 0.051816118047673096,
       "grad_norm": 0.0,
-      "learning_rate": 1.997875133587902e-05,
-      "loss": 1.0601,
+      "learning_rate": 1.9975098991974575e-05,
+      "loss": 1.1093,
       "step": 1826
     },
     {
-      "epoch": 0.05017714426958886,
+      "epoch": 0.05184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978693339686832e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.9975034130395295e-05,
+      "loss": 1.0504,
       "step": 1827
     },
     {
-      "epoch": 0.05020460849743209,
+      "epoch": 0.051872871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978635264539355e-05,
-      "loss": 1.1347,
+      "learning_rate": 1.997496918455645e-05,
+      "loss": 1.1671,
       "step": 1828
     },
     {
-      "epoch": 0.05023207272527533,
+      "epoch": 0.05190124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978577110437046e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.997490415445859e-05,
+      "loss": 1.0347,
       "step": 1829
     },
     {
-      "epoch": 0.050259536953118564,
+      "epoch": 0.051929625425652666,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978518877380374e-05,
-      "loss": 1.071,
+      "learning_rate": 1.997483904010227e-05,
+      "loss": 1.0821,
       "step": 1830
     },
     {
-      "epoch": 0.0502870011809618,
+      "epoch": 0.05195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978460565369792e-05,
-      "loss": 1.2056,
+      "learning_rate": 1.9974773841488036e-05,
+      "loss": 1.0982,
       "step": 1831
     },
     {
-      "epoch": 0.050314465408805034,
+      "epoch": 0.051986379114642454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978402174405768e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.997470855861644e-05,
+      "loss": 1.0218,
       "step": 1832
     },
     {
-      "epoch": 0.05034192963664826,
+      "epoch": 0.05201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.997834370448876e-05,
-      "loss": 1.1219,
+      "learning_rate": 1.997464319148803e-05,
+      "loss": 1.0032,
       "step": 1833
     },
     {
-      "epoch": 0.0503693938644915,
+      "epoch": 0.052043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.997828515561923e-05,
-      "loss": 1.1836,
+      "learning_rate": 1.9974577740103363e-05,
+      "loss": 0.9853,
       "step": 1834
     },
     {
-      "epoch": 0.05039685809233473,
+      "epoch": 0.05207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978226527797643e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.9974512204462985e-05,
+      "loss": 1.0292,
       "step": 1835
     },
     {
-      "epoch": 0.05042432232017797,
+      "epoch": 0.05209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978167821024463e-05,
-      "loss": 1.111,
+      "learning_rate": 1.9974446584567458e-05,
+      "loss": 1.1791,
       "step": 1836
     },
     {
-      "epoch": 0.0504517865480212,
+      "epoch": 0.05212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978109035300158e-05,
-      "loss": 1.261,
+      "learning_rate": 1.997438088041733e-05,
+      "loss": 1.0132,
       "step": 1837
     },
     {
-      "epoch": 0.05047925077586444,
+      "epoch": 0.052156640181611806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9978050170625183e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.9974315092013162e-05,
+      "loss": 1.2145,
       "step": 1838
     },
     {
-      "epoch": 0.050506715003707674,
+      "epoch": 0.0521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.997799122700001e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.9974249219355502e-05,
+      "loss": 1.1343,
       "step": 1839
     },
     {
-      "epoch": 0.0505341792315509,
+      "epoch": 0.05221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.997793220442511e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.997418326244491e-05,
+      "loss": 0.9472,
       "step": 1840
     },
     {
-      "epoch": 0.05056164345939414,
+      "epoch": 0.05224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977873102900944e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.997411722128195e-05,
+      "loss": 1.0993,
       "step": 1841
     },
     {
-      "epoch": 0.05058910768723737,
+      "epoch": 0.052270147559591376,
       "grad_norm": 0.0,
-      "learning_rate": 1.997781392242798e-05,
-      "loss": 1.2308,
+      "learning_rate": 1.9974051095867167e-05,
+      "loss": 1.1379,
       "step": 1842
     },
     {
-      "epoch": 0.05061657191508061,
+      "epoch": 0.05229852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977754663006686e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.9973984886201128e-05,
+      "loss": 1.0552,
       "step": 1843
     },
     {
-      "epoch": 0.05064403614292384,
+      "epoch": 0.05232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.997769532463753e-05,
-      "loss": 1.1591,
+      "learning_rate": 1.997391859228439e-05,
+      "loss": 1.1167,
       "step": 1844
     },
     {
-      "epoch": 0.05067150037076708,
+      "epoch": 0.05235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.997763590732099e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.9973852214117512e-05,
+      "loss": 1.0447,
       "step": 1845
     },
     {
-      "epoch": 0.05069896459861031,
+      "epoch": 0.05238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977576411057528e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9973785751701057e-05,
+      "loss": 0.9611,
       "step": 1846
     },
     {
-      "epoch": 0.05072642882645354,
+      "epoch": 0.052412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977516835847613e-05,
-      "loss": 1.1255,
+      "learning_rate": 1.9973719205035587e-05,
+      "loss": 1.073,
       "step": 1847
     },
     {
-      "epoch": 0.05075389305429678,
+      "epoch": 0.05244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.997745718169172e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.9973652574121663e-05,
+      "loss": 1.1274,
       "step": 1848
     },
     {
-      "epoch": 0.05078135728214001,
+      "epoch": 0.05246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977397448590324e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.9973585858959846e-05,
+      "loss": 1.1375,
       "step": 1849
     },
     {
-      "epoch": 0.05080882150998325,
+      "epoch": 0.05249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977337636543895e-05,
-      "loss": 1.1605,
+      "learning_rate": 1.99735190595507e-05,
+      "loss": 1.1826,
       "step": 1850
     },
     {
-      "epoch": 0.05083628573782648,
+      "epoch": 0.052525539160045404,
       "grad_norm": 0.0,
-      "learning_rate": 1.99772777455529e-05,
-      "loss": 1.1621,
+      "learning_rate": 1.997345217589479e-05,
+      "loss": 1.0445,
       "step": 1851
     },
     {
-      "epoch": 0.05086374996566972,
+      "epoch": 0.0525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977217775617826e-05,
-      "loss": 1.1233,
+      "learning_rate": 1.9973385207992684e-05,
+      "loss": 1.1722,
       "step": 1852
     },
     {
-      "epoch": 0.05089121419351295,
+      "epoch": 0.052582292849035185,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977157726739135e-05,
-      "loss": 1.1412,
+      "learning_rate": 1.997331815584494e-05,
+      "loss": 0.9827,
       "step": 1853
     },
     {
-      "epoch": 0.05091867842135618,
+      "epoch": 0.05261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.997709759891731e-05,
-      "loss": 1.1341,
+      "learning_rate": 1.9973251019452132e-05,
+      "loss": 1.1126,
       "step": 1854
     },
     {
-      "epoch": 0.050946142649199416,
+      "epoch": 0.052639046538024974,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977037392152822e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.9973183798814828e-05,
+      "loss": 1.1521,
       "step": 1855
     },
     {
-      "epoch": 0.05097360687704265,
+      "epoch": 0.05266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.997697710644615e-05,
-      "loss": 1.2134,
+      "learning_rate": 1.9973116493933586e-05,
+      "loss": 1.0545,
       "step": 1856
     },
     {
-      "epoch": 0.05100107110488589,
+      "epoch": 0.052695800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976916741797773e-05,
-      "loss": 1.0567,
+      "learning_rate": 1.9973049104808985e-05,
+      "loss": 1.1232,
       "step": 1857
     },
     {
-      "epoch": 0.05102853533272912,
+      "epoch": 0.05272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976856298208162e-05,
-      "loss": 1.189,
+      "learning_rate": 1.997298163144159e-05,
+      "loss": 1.0005,
       "step": 1858
     },
     {
-      "epoch": 0.05105599956057236,
+      "epoch": 0.052752553916004544,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976795775677802e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.9972914073831967e-05,
+      "loss": 1.2048,
       "step": 1859
     },
     {
-      "epoch": 0.051083463788415585,
+      "epoch": 0.05278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.997673517420717e-05,
-      "loss": 1.1181,
+      "learning_rate": 1.9972846431980693e-05,
+      "loss": 1.0792,
       "step": 1860
     },
     {
-      "epoch": 0.05111092801625882,
+      "epoch": 0.052809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976674493796742e-05,
-      "loss": 1.1156,
+      "learning_rate": 1.9972778705888337e-05,
+      "loss": 1.0759,
       "step": 1861
     },
     {
-      "epoch": 0.051138392244102056,
+      "epoch": 0.05283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976613734447e-05,
-      "loss": 1.0669,
+      "learning_rate": 1.997271089555547e-05,
+      "loss": 1.0875,
       "step": 1862
     },
     {
-      "epoch": 0.05116585647194529,
+      "epoch": 0.05286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.997655289615843e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.9972643000982667e-05,
+      "loss": 0.9955,
       "step": 1863
     },
     {
-      "epoch": 0.051193320699788526,
+      "epoch": 0.052894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976491978931505e-05,
-      "loss": 1.0598,
+      "learning_rate": 1.99725750221705e-05,
+      "loss": 1.1052,
       "step": 1864
     },
     {
-      "epoch": 0.05122078492763176,
+      "epoch": 0.052922814982973895,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976430982766714e-05,
-      "loss": 1.1857,
+      "learning_rate": 1.997250695911954e-05,
+      "loss": 1.132,
       "step": 1865
     },
     {
-      "epoch": 0.051248249155475,
+      "epoch": 0.05295119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976369907664534e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.997243881183037e-05,
+      "loss": 1.0418,
       "step": 1866
     },
     {
-      "epoch": 0.051275713383318225,
+      "epoch": 0.05297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976308753625453e-05,
-      "loss": 1.127,
+      "learning_rate": 1.9972370580303556e-05,
+      "loss": 1.018,
       "step": 1867
     },
     {
-      "epoch": 0.05130317761116146,
+      "epoch": 0.05300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.997624752064995e-05,
-      "loss": 1.2144,
+      "learning_rate": 1.9972302264539686e-05,
+      "loss": 1.1015,
       "step": 1868
     },
     {
-      "epoch": 0.051330641839004695,
+      "epoch": 0.053036322360953465,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976186208738512e-05,
-      "loss": 1.051,
+      "learning_rate": 1.9972233864539325e-05,
+      "loss": 1.0466,
       "step": 1869
     },
     {
-      "epoch": 0.05135810606684793,
+      "epoch": 0.05306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976124817891628e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.9972165380303058e-05,
+      "loss": 1.1041,
       "step": 1870
     },
     {
-      "epoch": 0.051385570294691166,
+      "epoch": 0.05309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9976063348109778e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.997209681183146e-05,
+      "loss": 1.0406,
       "step": 1871
     },
     {
-      "epoch": 0.0514130345225344,
+      "epoch": 0.05312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.997600179939345e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.9972028159125115e-05,
+      "loss": 1.1823,
       "step": 1872
     },
     {
-      "epoch": 0.051440498750377636,
+      "epoch": 0.05314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975940171743132e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.99719594221846e-05,
+      "loss": 1.0578,
       "step": 1873
     },
     {
-      "epoch": 0.051467962978220864,
+      "epoch": 0.05317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.997587846515931e-05,
-      "loss": 1.1398,
+      "learning_rate": 1.9971890601010497e-05,
+      "loss": 1.0512,
       "step": 1874
     },
     {
-      "epoch": 0.0514954272060641,
+      "epoch": 0.05320658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975816679642476e-05,
-      "loss": 1.0529,
+      "learning_rate": 1.9971821695603383e-05,
+      "loss": 1.0874,
       "step": 1875
     },
     {
-      "epoch": 0.051522891433907335,
+      "epoch": 0.053234960272417704,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975754815193115e-05,
-      "loss": 1.1368,
+      "learning_rate": 1.997175270596384e-05,
+      "loss": 1.0391,
       "step": 1876
     },
     {
-      "epoch": 0.05155035566175057,
+      "epoch": 0.0532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975692871811714e-05,
-      "loss": 1.1238,
+      "learning_rate": 1.997168363209246e-05,
+      "loss": 0.9897,
       "step": 1877
     },
     {
-      "epoch": 0.051577819889593805,
+      "epoch": 0.05329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.997563084949877e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.9971614473989817e-05,
+      "loss": 1.1074,
       "step": 1878
     },
     {
-      "epoch": 0.05160528411743704,
+      "epoch": 0.05332009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 1.997556874825477e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9971545231656498e-05,
+      "loss": 1.1343,
       "step": 1879
     },
     {
-      "epoch": 0.051632748345280276,
+      "epoch": 0.053348467650397274,
       "grad_norm": 0.0,
-      "learning_rate": 1.997550656808021e-05,
-      "loss": 1.1836,
+      "learning_rate": 1.9971475905093087e-05,
+      "loss": 1.1169,
       "step": 1880
     },
     {
-      "epoch": 0.051660212573123504,
+      "epoch": 0.05337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975444308975575e-05,
-      "loss": 1.1198,
+      "learning_rate": 1.9971406494300174e-05,
+      "loss": 1.1955,
       "step": 1881
     },
     {
-      "epoch": 0.05168767680096674,
+      "epoch": 0.05340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975381970941355e-05,
-      "loss": 0.988,
+      "learning_rate": 1.997133699927834e-05,
+      "loss": 1.0346,
       "step": 1882
     },
     {
-      "epoch": 0.051715141028809974,
+      "epoch": 0.05343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975319553978054e-05,
-      "loss": 1.101,
+      "learning_rate": 1.9971267420028175e-05,
+      "loss": 1.2126,
       "step": 1883
     },
     {
-      "epoch": 0.05174260525665321,
+      "epoch": 0.053461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975257058086163e-05,
-      "loss": 1.1357,
+      "learning_rate": 1.9971197756550263e-05,
+      "loss": 1.1011,
       "step": 1884
     },
     {
-      "epoch": 0.051770069484496445,
+      "epoch": 0.05349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.997519448326617e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.99711280088452e-05,
+      "loss": 1.0734,
       "step": 1885
     },
     {
-      "epoch": 0.05179753371233968,
+      "epoch": 0.053518728717366626,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975131829518575e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.9971058176913566e-05,
+      "loss": 1.0799,
       "step": 1886
     },
     {
-      "epoch": 0.05182499794018291,
+      "epoch": 0.05354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9975069096843872e-05,
-      "loss": 1.1661,
+      "learning_rate": 1.997098826075596e-05,
+      "loss": 1.1577,
       "step": 1887
     },
     {
-      "epoch": 0.051852462168026144,
+      "epoch": 0.053575482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.997500628524256e-05,
-      "loss": 1.1891,
+      "learning_rate": 1.9970918260372962e-05,
+      "loss": 1.0904,
       "step": 1888
     },
     {
-      "epoch": 0.05187992639586938,
+      "epoch": 0.05360385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974943394715133e-05,
-      "loss": 1.1174,
+      "learning_rate": 1.9970848175765174e-05,
+      "loss": 1.0442,
       "step": 1889
     },
     {
-      "epoch": 0.051907390623712614,
+      "epoch": 0.053632236095346196,
       "grad_norm": 0.0,
-      "learning_rate": 1.997488042526209e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.9970778006933183e-05,
+      "loss": 1.1646,
       "step": 1890
     },
     {
-      "epoch": 0.05193485485155585,
+      "epoch": 0.05366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974817376883932e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.997070775387758e-05,
+      "loss": 1.078,
       "step": 1891
     },
     {
-      "epoch": 0.051962319079399084,
+      "epoch": 0.053688989784335985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974754249581153e-05,
-      "loss": 1.0108,
+      "learning_rate": 1.997063741659896e-05,
+      "loss": 1.0853,
       "step": 1892
     },
     {
-      "epoch": 0.05198978330724232,
+      "epoch": 0.05371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974691043354256e-05,
-      "loss": 1.0352,
+      "learning_rate": 1.997056699509792e-05,
+      "loss": 1.1241,
       "step": 1893
     },
     {
-      "epoch": 0.05201724753508555,
+      "epoch": 0.053745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.997462775820374e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.997049648937505e-05,
+      "loss": 1.0231,
       "step": 1894
     },
     {
-      "epoch": 0.05204471176292878,
+      "epoch": 0.05377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974564394130104e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.997042589943095e-05,
+      "loss": 1.06,
       "step": 1895
     },
     {
-      "epoch": 0.05207217599077202,
+      "epoch": 0.05380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974500951133845e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.997035522526621e-05,
+      "loss": 1.1386,
       "step": 1896
     },
     {
-      "epoch": 0.052099640218615254,
+      "epoch": 0.05383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974437429215478e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.9970284466881437e-05,
+      "loss": 1.0753,
       "step": 1897
     },
     {
-      "epoch": 0.05212710444645849,
+      "epoch": 0.053859250851305336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974373828375497e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.997021362427722e-05,
+      "loss": 1.0108,
       "step": 1898
     },
     {
-      "epoch": 0.052154568674301724,
+      "epoch": 0.05388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974310148614408e-05,
-      "loss": 1.1655,
+      "learning_rate": 1.997014269745416e-05,
+      "loss": 1.1526,
       "step": 1899
     },
     {
-      "epoch": 0.05218203290214496,
+      "epoch": 0.05391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.997424638993271e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.997007168641286e-05,
+      "loss": 1.1998,
       "step": 1900
     },
     {
-      "epoch": 0.05220949712998819,
+      "epoch": 0.05394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974182552330913e-05,
-      "loss": 1.1318,
+      "learning_rate": 1.9970000591153913e-05,
+      "loss": 1.172,
       "step": 1901
     },
     {
-      "epoch": 0.05223696135783142,
+      "epoch": 0.053972758229284906,
       "grad_norm": 0.0,
-      "learning_rate": 1.997411863580952e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.996992941167792e-05,
+      "loss": 0.9428,
       "step": 1902
     },
     {
-      "epoch": 0.05226442558567466,
+      "epoch": 0.054001135073779793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9974054640369034e-05,
-      "loss": 1.1635,
+      "learning_rate": 1.996985814798549e-05,
+      "loss": 1.1326,
       "step": 1903
     },
     {
-      "epoch": 0.05229188981351789,
+      "epoch": 0.05402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973990566009965e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.9969786800077217e-05,
+      "loss": 1.0875,
       "step": 1904
     },
     {
-      "epoch": 0.05231935404136113,
+      "epoch": 0.05405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973926412732822e-05,
-      "loss": 1.1581,
+      "learning_rate": 1.996971536795371e-05,
+      "loss": 1.1203,
       "step": 1905
     },
     {
-      "epoch": 0.052346818269204363,
+      "epoch": 0.05408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973862180538108e-05,
-      "loss": 1.2034,
+      "learning_rate": 1.996964385161557e-05,
+      "loss": 1.087,
       "step": 1906
     },
     {
-      "epoch": 0.0523742824970476,
+      "epoch": 0.054114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973797869426334e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.9969572251063397e-05,
+      "loss": 1.1569,
       "step": 1907
     },
     {
-      "epoch": 0.05240174672489083,
+      "epoch": 0.05414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973733479398007e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.9969500566297798e-05,
+      "loss": 0.9488,
       "step": 1908
     },
     {
-      "epoch": 0.05242921095273406,
+      "epoch": 0.05417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973669010453637e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.996942879731938e-05,
+      "loss": 1.0688,
       "step": 1909
     },
     {
-      "epoch": 0.0524566751805773,
+      "epoch": 0.05419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.997360446259373e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.996935694412875e-05,
+      "loss": 1.0853,
       "step": 1910
     },
     {
-      "epoch": 0.05248413940842053,
+      "epoch": 0.054228149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973539835818808e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.9969285006726513e-05,
+      "loss": 1.1064,
       "step": 1911
     },
     {
-      "epoch": 0.05251160363626377,
+      "epoch": 0.05425652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973475130129373e-05,
-      "loss": 1.1231,
+      "learning_rate": 1.9969212985113278e-05,
+      "loss": 1.0593,
       "step": 1912
     },
     {
-      "epoch": 0.052539067864107,
+      "epoch": 0.054284903518728715,
       "grad_norm": 0.0,
-      "learning_rate": 1.997341034552594e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.9969140879289654e-05,
+      "loss": 1.0373,
       "step": 1913
     },
     {
-      "epoch": 0.05256653209195024,
+      "epoch": 0.05431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973345482009022e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.9969068689256246e-05,
+      "loss": 1.1328,
       "step": 1914
     },
     {
-      "epoch": 0.052593996319793467,
+      "epoch": 0.054341657207718504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973280539579126e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.996899641501367e-05,
+      "loss": 0.9223,
       "step": 1915
     },
     {
-      "epoch": 0.0526214605476367,
+      "epoch": 0.05437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973215518236775e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.996892405656253e-05,
+      "loss": 1.0717,
       "step": 1916
     },
     {
-      "epoch": 0.05264892477547994,
+      "epoch": 0.054398410896708285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973150417982482e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.9968851613903443e-05,
+      "loss": 1.0536,
       "step": 1917
     },
     {
-      "epoch": 0.05267638900332317,
+      "epoch": 0.05442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.9973085238816754e-05,
-      "loss": 1.075,
+      "learning_rate": 1.9968779087037016e-05,
+      "loss": 1.0187,
       "step": 1918
     },
     {
-      "epoch": 0.05270385323116641,
+      "epoch": 0.054455164585698074,
       "grad_norm": 0.0,
-      "learning_rate": 1.997301998074012e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.9968706475963863e-05,
+      "loss": 1.0691,
       "step": 1919
     },
     {
-      "epoch": 0.05273131745900964,
+      "epoch": 0.05448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.997295464375308e-05,
-      "loss": 1.1008,
+      "learning_rate": 1.99686337806846e-05,
+      "loss": 0.923,
       "step": 1920
     },
     {
-      "epoch": 0.05275878168685287,
+      "epoch": 0.054511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972889227856165e-05,
-      "loss": 1.1131,
+      "learning_rate": 1.9968561001199843e-05,
+      "loss": 1.0846,
       "step": 1921
     },
     {
-      "epoch": 0.052786245914696106,
+      "epoch": 0.05454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972823733049888e-05,
-      "loss": 1.0971,
+      "learning_rate": 1.99684881375102e-05,
+      "loss": 1.1037,
       "step": 1922
     },
     {
-      "epoch": 0.05281371014253934,
+      "epoch": 0.05456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972758159334763e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.996841518961629e-05,
+      "loss": 0.9933,
       "step": 1923
     },
     {
-      "epoch": 0.052841174370382576,
+      "epoch": 0.05459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972692506711312e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.9968342157518724e-05,
+      "loss": 1.0575,
       "step": 1924
     },
     {
-      "epoch": 0.05286863859822581,
+      "epoch": 0.054625425652667425,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972626775180057e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.996826904121813e-05,
+      "loss": 1.1027,
       "step": 1925
     },
     {
-      "epoch": 0.05289610282606905,
+      "epoch": 0.05465380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972560964741517e-05,
-      "loss": 1.1733,
+      "learning_rate": 1.9968195840715118e-05,
+      "loss": 1.0408,
       "step": 1926
     },
     {
-      "epoch": 0.05292356705391228,
+      "epoch": 0.05468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.997249507539621e-05,
-      "loss": 1.0798,
+      "learning_rate": 1.9968122556010305e-05,
+      "loss": 1.0275,
       "step": 1927
     },
     {
-      "epoch": 0.05295103128175551,
+      "epoch": 0.0547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972429107144658e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.9968049187104316e-05,
+      "loss": 1.0276,
       "step": 1928
     },
     {
-      "epoch": 0.052978495509598746,
+      "epoch": 0.054738933030646995,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972363059987386e-05,
-      "loss": 1.053,
+      "learning_rate": 1.9967975733997763e-05,
+      "loss": 1.1469,
       "step": 1929
     },
     {
-      "epoch": 0.05300595973744198,
+      "epoch": 0.05476730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972296933924915e-05,
-      "loss": 1.1227,
+      "learning_rate": 1.9967902196691274e-05,
+      "loss": 0.9847,
       "step": 1930
     },
     {
-      "epoch": 0.053033423965285216,
+      "epoch": 0.05479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972230728957768e-05,
-      "loss": 1.1154,
+      "learning_rate": 1.9967828575185467e-05,
+      "loss": 1.1055,
       "step": 1931
     },
     {
-      "epoch": 0.05306088819312845,
+      "epoch": 0.05482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972164445086462e-05,
-      "loss": 1.1192,
+      "learning_rate": 1.9967754869480963e-05,
+      "loss": 1.226,
       "step": 1932
     },
     {
-      "epoch": 0.053088352420971686,
+      "epoch": 0.05485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972098082311534e-05,
-      "loss": 1.2255,
+      "learning_rate": 1.9967681079578386e-05,
+      "loss": 1.0755,
       "step": 1933
     },
     {
-      "epoch": 0.05311581664881492,
+      "epoch": 0.05488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.9972031640633504e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.996760720547836e-05,
+      "loss": 1.1952,
       "step": 1934
     },
     {
-      "epoch": 0.05314328087665815,
+      "epoch": 0.05490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971965120052896e-05,
-      "loss": 1.076,
+      "learning_rate": 1.996753324718151e-05,
+      "loss": 1.0931,
       "step": 1935
     },
     {
-      "epoch": 0.053170745104501385,
+      "epoch": 0.054937570942111234,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971898520570233e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9967459204688452e-05,
+      "loss": 1.0486,
       "step": 1936
     },
     {
-      "epoch": 0.05319820933234462,
+      "epoch": 0.05496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.997183184218605e-05,
-      "loss": 1.1373,
+      "learning_rate": 1.9967385077999822e-05,
+      "loss": 1.0103,
       "step": 1937
     },
     {
-      "epoch": 0.053225673560187856,
+      "epoch": 0.05499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971765084900872e-05,
-      "loss": 1.0603,
+      "learning_rate": 1.996731086711624e-05,
+      "loss": 1.0655,
       "step": 1938
     },
     {
-      "epoch": 0.05325313778803109,
+      "epoch": 0.05502270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971698248715225e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.9967236572038337e-05,
+      "loss": 1.0971,
       "step": 1939
     },
     {
-      "epoch": 0.053280602015874326,
+      "epoch": 0.055051078320090804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971631333629637e-05,
-      "loss": 1.0856,
+      "learning_rate": 1.996716219276674e-05,
+      "loss": 1.09,
       "step": 1940
     },
     {
-      "epoch": 0.05330806624371756,
+      "epoch": 0.0550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.997156433964464e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.996708772930207e-05,
+      "loss": 1.0802,
       "step": 1941
     },
     {
-      "epoch": 0.05333553047156079,
+      "epoch": 0.05510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971497266760764e-05,
-      "loss": 1.1489,
+      "learning_rate": 1.996701318164497e-05,
+      "loss": 1.0537,
       "step": 1942
     },
     {
-      "epoch": 0.053362994699404025,
+      "epoch": 0.05513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971430114978538e-05,
-      "loss": 1.2977,
+      "learning_rate": 1.9966938549796056e-05,
+      "loss": 1.0274,
       "step": 1943
     },
     {
-      "epoch": 0.05339045892724726,
+      "epoch": 0.055164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971362884298496e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.9966863833755966e-05,
+      "loss": 1.0435,
       "step": 1944
     },
     {
-      "epoch": 0.053417923155090495,
+      "epoch": 0.05519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971295574721168e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.996678903352533e-05,
+      "loss": 0.9958,
       "step": 1945
     },
     {
-      "epoch": 0.05344538738293373,
+      "epoch": 0.055221339387060156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971228186247088e-05,
-      "loss": 1.1924,
+      "learning_rate": 1.9966714149104776e-05,
+      "loss": 1.03,
       "step": 1946
     },
     {
-      "epoch": 0.053472851610776966,
+      "epoch": 0.05524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971160718876784e-05,
-      "loss": 1.1474,
+      "learning_rate": 1.996663918049494e-05,
+      "loss": 1.0184,
       "step": 1947
     },
     {
-      "epoch": 0.053500315838620194,
+      "epoch": 0.055278093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971093172610797e-05,
-      "loss": 0.9936,
+      "learning_rate": 1.996656412769646e-05,
+      "loss": 1.1021,
       "step": 1948
     },
     {
-      "epoch": 0.05352778006646343,
+      "epoch": 0.05530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9971025547449656e-05,
-      "loss": 1.1666,
+      "learning_rate": 1.996648899070996e-05,
+      "loss": 1.0828,
       "step": 1949
     },
     {
-      "epoch": 0.053555244294306664,
+      "epoch": 0.055334846765039726,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970957843393902e-05,
-      "loss": 1.0975,
+      "learning_rate": 1.9966413769536078e-05,
+      "loss": 1.1503,
       "step": 1950
     },
     {
-      "epoch": 0.0535827085221499,
+      "epoch": 0.05536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970890060444063e-05,
-      "loss": 1.1689,
+      "learning_rate": 1.9966338464175457e-05,
+      "loss": 1.028,
       "step": 1951
     },
     {
-      "epoch": 0.053610172749993135,
+      "epoch": 0.055391600454029515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970822198600683e-05,
-      "loss": 1.1205,
+      "learning_rate": 1.9966263074628723e-05,
+      "loss": 1.2029,
       "step": 1952
     },
     {
-      "epoch": 0.05363763697783637,
+      "epoch": 0.0554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970754257864296e-05,
-      "loss": 0.9596,
+      "learning_rate": 1.9966187600896516e-05,
+      "loss": 1.0735,
       "step": 1953
     },
     {
-      "epoch": 0.053665101205679605,
+      "epoch": 0.055448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970686238235437e-05,
-      "loss": 1.1124,
+      "learning_rate": 1.9966112042979476e-05,
+      "loss": 1.1306,
       "step": 1954
     },
     {
-      "epoch": 0.05369256543352283,
+      "epoch": 0.05547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970618139714646e-05,
-      "loss": 1.1366,
+      "learning_rate": 1.9966036400878237e-05,
+      "loss": 0.967,
       "step": 1955
     },
     {
-      "epoch": 0.05372002966136607,
+      "epoch": 0.05550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970549962302465e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9965960674593443e-05,
+      "loss": 1.1264,
       "step": 1956
     },
     {
-      "epoch": 0.053747493889209304,
+      "epoch": 0.05553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970481705999427e-05,
-      "loss": 1.031,
+      "learning_rate": 1.996588486412573e-05,
+      "loss": 1.1111,
       "step": 1957
     },
     {
-      "epoch": 0.05377495811705254,
+      "epoch": 0.055561861520998866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970413370806074e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.996580896947574e-05,
+      "loss": 1.175,
       "step": 1958
     },
     {
-      "epoch": 0.053802422344895774,
+      "epoch": 0.05559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970344956722953e-05,
-      "loss": 1.1679,
+      "learning_rate": 1.9965732990644116e-05,
+      "loss": 1.0022,
       "step": 1959
     },
     {
-      "epoch": 0.05382988657273901,
+      "epoch": 0.05561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970276463750598e-05,
-      "loss": 1.1783,
+      "learning_rate": 1.9965656927631497e-05,
+      "loss": 1.1198,
       "step": 1960
     },
     {
-      "epoch": 0.053857350800582245,
+      "epoch": 0.05564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970207891889552e-05,
-      "loss": 1.1729,
+      "learning_rate": 1.9965580780438522e-05,
+      "loss": 1.0251,
       "step": 1961
     },
     {
-      "epoch": 0.05388481502842547,
+      "epoch": 0.055675368898978436,
       "grad_norm": 0.0,
-      "learning_rate": 1.997013924114036e-05,
-      "loss": 1.1409,
+      "learning_rate": 1.9965504549065843e-05,
+      "loss": 1.1548,
       "step": 1962
     },
     {
-      "epoch": 0.05391227925626871,
+      "epoch": 0.055703745743473324,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970070511503564e-05,
-      "loss": 1.132,
+      "learning_rate": 1.9965428233514092e-05,
+      "loss": 1.085,
       "step": 1963
     },
     {
-      "epoch": 0.05393974348411194,
+      "epoch": 0.05573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.9970001702979714e-05,
-      "loss": 1.082,
+      "learning_rate": 1.996535183378393e-05,
+      "loss": 1.0967,
       "step": 1964
     },
     {
-      "epoch": 0.05396720771195518,
+      "epoch": 0.05576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.996993281556934e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.9965275349875987e-05,
+      "loss": 1.0532,
       "step": 1965
     },
     {
-      "epoch": 0.053994671939798414,
+      "epoch": 0.055788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969863849273003e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.9965198781790917e-05,
+      "loss": 0.9056,
       "step": 1966
     },
     {
-      "epoch": 0.05402213616764165,
+      "epoch": 0.055817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969794804091236e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.9965122129529362e-05,
+      "loss": 1.115,
       "step": 1967
     },
     {
-      "epoch": 0.054049600395484884,
+      "epoch": 0.05584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.996972568002459e-05,
-      "loss": 1.1183,
+      "learning_rate": 1.9965045393091974e-05,
+      "loss": 1.0776,
       "step": 1968
     },
     {
-      "epoch": 0.05407706462332811,
+      "epoch": 0.05587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969656477073617e-05,
-      "loss": 1.0965,
+      "learning_rate": 1.9964968572479403e-05,
+      "loss": 1.0346,
       "step": 1969
     },
     {
-      "epoch": 0.05410452885117135,
+      "epoch": 0.05590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.996958719523886e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.9964891667692292e-05,
+      "loss": 1.1388,
       "step": 1970
     },
     {
-      "epoch": 0.05413199307901458,
+      "epoch": 0.055930760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.996951783452086e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.996481467873129e-05,
+      "loss": 1.0622,
       "step": 1971
     },
     {
-      "epoch": 0.05415945730685782,
+      "epoch": 0.05595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.996944839492018e-05,
-      "loss": 1.11,
+      "learning_rate": 1.9964737605597054e-05,
+      "loss": 1.0449,
       "step": 1972
     },
     {
-      "epoch": 0.05418692153470105,
+      "epoch": 0.055987514188422245,
       "grad_norm": 0.0,
-      "learning_rate": 1.996937887643736e-05,
-      "loss": 1.0423,
+      "learning_rate": 1.9964660448290232e-05,
+      "loss": 1.1296,
       "step": 1973
     },
     {
-      "epoch": 0.05421438576254429,
+      "epoch": 0.05601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969309279072955e-05,
-      "loss": 1.1295,
+      "learning_rate": 1.9964583206811475e-05,
+      "loss": 1.059,
       "step": 1974
     },
     {
-      "epoch": 0.054241849990387524,
+      "epoch": 0.056044267877412034,
       "grad_norm": 0.0,
-      "learning_rate": 1.996923960282751e-05,
-      "loss": 1.1369,
+      "learning_rate": 1.9964505881161435e-05,
+      "loss": 1.0673,
       "step": 1975
     },
     {
-      "epoch": 0.05426931421823075,
+      "epoch": 0.05607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969169847701582e-05,
-      "loss": 1.1417,
+      "learning_rate": 1.9964428471340765e-05,
+      "loss": 1.0173,
       "step": 1976
     },
     {
-      "epoch": 0.05429677844607399,
+      "epoch": 0.056101021566401815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969100013695718e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.9964350977350123e-05,
+      "loss": 1.1125,
       "step": 1977
     },
     {
-      "epoch": 0.05432424267391722,
+      "epoch": 0.05612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.9969030100810477e-05,
-      "loss": 1.2327,
+      "learning_rate": 1.9964273399190156e-05,
+      "loss": 1.0588,
       "step": 1978
     },
     {
-      "epoch": 0.05435170690176046,
+      "epoch": 0.056157775255391604,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968960109046404e-05,
-      "loss": 1.1285,
+      "learning_rate": 1.9964195736861524e-05,
+      "loss": 1.0306,
       "step": 1979
     },
     {
-      "epoch": 0.05437917112960369,
+      "epoch": 0.05618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968890038404058e-05,
-      "loss": 1.0437,
+      "learning_rate": 1.9964117990364885e-05,
+      "loss": 1.0672,
       "step": 1980
     },
     {
-      "epoch": 0.05440663535744693,
+      "epoch": 0.056214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968819888883994e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.9964040159700896e-05,
+      "loss": 1.1533,
       "step": 1981
     },
     {
-      "epoch": 0.054434099585290156,
+      "epoch": 0.05624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968749660486763e-05,
-      "loss": 1.093,
+      "learning_rate": 1.9963962244870205e-05,
+      "loss": 0.9707,
       "step": 1982
     },
     {
-      "epoch": 0.05446156381313339,
+      "epoch": 0.05627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968679353212925e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.996388424587348e-05,
+      "loss": 1.1298,
       "step": 1983
     },
     {
-      "epoch": 0.05448902804097663,
+      "epoch": 0.05629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968608967063034e-05,
-      "loss": 1.1489,
+      "learning_rate": 1.996380616271138e-05,
+      "loss": 1.0885,
       "step": 1984
     },
     {
-      "epoch": 0.05451649226881986,
+      "epoch": 0.056328036322360955,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968538502037652e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.9963727995384558e-05,
+      "loss": 1.0892,
       "step": 1985
     },
     {
-      "epoch": 0.0545439564966631,
+      "epoch": 0.05635641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968467958137324e-05,
-      "loss": 1.1199,
+      "learning_rate": 1.996364974389368e-05,
+      "loss": 1.1026,
       "step": 1986
     },
     {
-      "epoch": 0.05457142072450633,
+      "epoch": 0.05638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968397335362623e-05,
-      "loss": 1.154,
+      "learning_rate": 1.9963571408239405e-05,
+      "loss": 1.1049,
       "step": 1987
     },
     {
-      "epoch": 0.05459888495234957,
+      "epoch": 0.05641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968326633714094e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.996349298842239e-05,
+      "loss": 1.1036,
       "step": 1988
     },
     {
-      "epoch": 0.054626349180192796,
+      "epoch": 0.056441543700340525,
       "grad_norm": 0.0,
-      "learning_rate": 1.996825585319231e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.9963414484443304e-05,
+      "loss": 1.0108,
       "step": 1989
     },
     {
-      "epoch": 0.05465381340803603,
+      "epoch": 0.05646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.996818499379782e-05,
-      "loss": 1.1892,
+      "learning_rate": 1.996333589630281e-05,
+      "loss": 1.0296,
       "step": 1990
     },
     {
-      "epoch": 0.054681277635879266,
+      "epoch": 0.05649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.996811405553119e-05,
-      "loss": 1.0601,
+      "learning_rate": 1.9963257224001568e-05,
+      "loss": 1.049,
       "step": 1991
     },
     {
-      "epoch": 0.0547087418637225,
+      "epoch": 0.0565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9968043038392978e-05,
-      "loss": 1.1167,
+      "learning_rate": 1.9963178467540244e-05,
+      "loss": 1.1864,
       "step": 1992
     },
     {
-      "epoch": 0.05473620609156574,
+      "epoch": 0.05655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.996797194238375e-05,
-      "loss": 1.1521,
+      "learning_rate": 1.9963099626919506e-05,
+      "loss": 0.9172,
       "step": 1993
     },
     {
-      "epoch": 0.05476367031940897,
+      "epoch": 0.05658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.996790076750407e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9963020702140016e-05,
+      "loss": 1.0971,
       "step": 1994
     },
     {
-      "epoch": 0.05479113454725221,
+      "epoch": 0.05661180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967829513754497e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.996294169320244e-05,
+      "loss": 1.08,
       "step": 1995
     },
     {
-      "epoch": 0.054818598775095435,
+      "epoch": 0.056640181611804764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967758181135594e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.9962862600107452e-05,
+      "loss": 1.1657,
       "step": 1996
     },
     {
-      "epoch": 0.05484606300293867,
+      "epoch": 0.05666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967686769647926e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.9962783422855715e-05,
+      "loss": 1.0987,
       "step": 1997
     },
     {
-      "epoch": 0.054873527230781906,
+      "epoch": 0.05669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.996761527929206e-05,
-      "loss": 1.1552,
+      "learning_rate": 1.9962704161447894e-05,
+      "loss": 1.0039,
       "step": 1998
     },
     {
-      "epoch": 0.05490099145862514,
+      "epoch": 0.05672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967543710068562e-05,
-      "loss": 1.1494,
+      "learning_rate": 1.996262481588467e-05,
+      "loss": 0.9565,
       "step": 1999
     },
     {
-      "epoch": 0.054928455686468376,
+      "epoch": 0.056753688989784334,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967472061978e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.9962545386166702e-05,
+      "loss": 1.076,
       "step": 2000
     },
     {
-      "epoch": 0.05495591991431161,
+      "epoch": 0.05678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967400335020932e-05,
-      "loss": 1.1067,
+      "learning_rate": 1.9962465872294663e-05,
+      "loss": 1.055,
       "step": 2001
     },
     {
-      "epoch": 0.05498338414215485,
+      "epoch": 0.05681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967328529197933e-05,
-      "loss": 1.1536,
+      "learning_rate": 1.996238627426923e-05,
+      "loss": 1.0233,
       "step": 2002
     },
     {
-      "epoch": 0.055010848369998075,
+      "epoch": 0.05683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967256644509573e-05,
-      "loss": 1.1129,
+      "learning_rate": 1.9962306592091073e-05,
+      "loss": 1.0026,
       "step": 2003
     },
     {
-      "epoch": 0.05503831259784131,
+      "epoch": 0.056867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.9967184680956414e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.9962226825760865e-05,
+      "loss": 1.0051,
       "step": 2004
     },
     {
-      "epoch": 0.055065776825684545,
+      "epoch": 0.0568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.996711263853903e-05,
-      "loss": 1.1088,
+      "learning_rate": 1.9962146975279274e-05,
+      "loss": 1.0939,
       "step": 2005
     },
     {
-      "epoch": 0.05509324105352778,
+      "epoch": 0.056923950056753686,
       "grad_norm": 0.0,
-      "learning_rate": 1.996704051725799e-05,
-      "loss": 1.1873,
+      "learning_rate": 1.9962067040646983e-05,
+      "loss": 0.9321,
       "step": 2006
     },
     {
-      "epoch": 0.055120705281371016,
+      "epoch": 0.05695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966968317113865e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.9961987021864666e-05,
+      "loss": 1.1409,
       "step": 2007
     },
     {
-      "epoch": 0.05514816950921425,
+      "epoch": 0.056980703745743475,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966896038107226e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.9961906918932996e-05,
+      "loss": 1.0884,
       "step": 2008
     },
     {
-      "epoch": 0.05517563373705748,
+      "epoch": 0.05700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966823680238644e-05,
-      "loss": 1.036,
+      "learning_rate": 1.996182673185265e-05,
+      "loss": 1.0526,
       "step": 2009
     },
     {
-      "epoch": 0.055203097964900714,
+      "epoch": 0.057037457434733256,
       "grad_norm": 0.0,
-      "learning_rate": 1.996675124350869e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.9961746460624304e-05,
+      "loss": 0.9333,
       "step": 2010
     },
     {
-      "epoch": 0.05523056219274395,
+      "epoch": 0.05706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966678727917943e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.9961666105248637e-05,
+      "loss": 1.0242,
       "step": 2011
     },
     {
-      "epoch": 0.055258026420587185,
+      "epoch": 0.057094211123723045,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966606133466974e-05,
-      "loss": 1.0867,
+      "learning_rate": 1.9961585665726332e-05,
+      "loss": 1.1112,
       "step": 2012
     },
     {
-      "epoch": 0.05528549064843042,
+      "epoch": 0.05712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.996653346015635e-05,
-      "loss": 1.1513,
+      "learning_rate": 1.9961505142058063e-05,
+      "loss": 1.1519,
       "step": 2013
     },
     {
-      "epoch": 0.055312954876273655,
+      "epoch": 0.057150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966460707986658e-05,
-      "loss": 1.1193,
+      "learning_rate": 1.9961424534244514e-05,
+      "loss": 1.1085,
       "step": 2014
     },
     {
-      "epoch": 0.05534041910411689,
+      "epoch": 0.05717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.996638787695847e-05,
-      "loss": 1.0476,
+      "learning_rate": 1.9961343842286363e-05,
+      "loss": 1.09,
       "step": 2015
     },
     {
-      "epoch": 0.05536788333196012,
+      "epoch": 0.05720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966314967072356e-05,
-      "loss": 1.1219,
+      "learning_rate": 1.9961263066184292e-05,
+      "loss": 1.099,
       "step": 2016
     },
     {
-      "epoch": 0.055395347559803354,
+      "epoch": 0.0572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.99662419783289e-05,
-      "loss": 1.1198,
+      "learning_rate": 1.9961182205938984e-05,
+      "loss": 1.1071,
       "step": 2017
     },
     {
-      "epoch": 0.05542281178764659,
+      "epoch": 0.057264472190692396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966168910728677e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.9961101261551127e-05,
+      "loss": 1.1096,
       "step": 2018
     },
     {
-      "epoch": 0.055450276015489824,
+      "epoch": 0.05729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966095764272263e-05,
-      "loss": 1.1472,
+      "learning_rate": 1.9961020233021395e-05,
+      "loss": 1.0361,
       "step": 2019
     },
     {
-      "epoch": 0.05547774024333306,
+      "epoch": 0.05732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.9966022538960242e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.9960939120350477e-05,
+      "loss": 0.9811,
       "step": 2020
     },
     {
-      "epoch": 0.055505204471176295,
+      "epoch": 0.05734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965949234793188e-05,
-      "loss": 1.0855,
+      "learning_rate": 1.996085792353906e-05,
+      "loss": 1.1366,
       "step": 2021
     },
     {
-      "epoch": 0.05553266869901953,
+      "epoch": 0.057377979568671966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965875851771683e-05,
-      "loss": 1.0442,
+      "learning_rate": 1.9960776642587828e-05,
+      "loss": 1.0561,
       "step": 2022
     },
     {
-      "epoch": 0.05556013292686276,
+      "epoch": 0.057406356413166854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965802389896306e-05,
-      "loss": 1.1287,
+      "learning_rate": 1.996069527749747e-05,
+      "loss": 1.1013,
       "step": 2023
     },
     {
-      "epoch": 0.055587597154705994,
+      "epoch": 0.05743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965728849167644e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.996061382826867e-05,
+      "loss": 1.0924,
       "step": 2024
     },
     {
-      "epoch": 0.05561506138254923,
+      "epoch": 0.05746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965655229586272e-05,
-      "loss": 1.11,
+      "learning_rate": 1.9960532294902114e-05,
+      "loss": 1.047,
       "step": 2025
     },
     {
-      "epoch": 0.055642525610392464,
+      "epoch": 0.05749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965581531152778e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.9960450677398495e-05,
+      "loss": 1.1148,
       "step": 2026
     },
     {
-      "epoch": 0.0556699898382357,
+      "epoch": 0.057519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965507753867742e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.9960368975758505e-05,
+      "loss": 1.0682,
       "step": 2027
     },
     {
-      "epoch": 0.055697454066078934,
+      "epoch": 0.05754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.996543389773175e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.9960287189982826e-05,
+      "loss": 0.9956,
       "step": 2028
     },
     {
-      "epoch": 0.05572491829392217,
+      "epoch": 0.05757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965359962745384e-05,
-      "loss": 1.1725,
+      "learning_rate": 1.9960205320072158e-05,
+      "loss": 0.9899,
       "step": 2029
     },
     {
-      "epoch": 0.0557523825217654,
+      "epoch": 0.0576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965285948909224e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.9960123366027187e-05,
+      "loss": 1.0283,
       "step": 2030
     },
     {
-      "epoch": 0.05577984674960863,
+      "epoch": 0.057633371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965211856223868e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.9960041327848602e-05,
+      "loss": 1.1094,
       "step": 2031
     },
     {
-      "epoch": 0.05580731097745187,
+      "epoch": 0.05766174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.9965137684689892e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.9959959205537105e-05,
+      "loss": 1.0514,
       "step": 2032
     },
     {
-      "epoch": 0.055834775205295104,
+      "epoch": 0.057690124858115775,
       "grad_norm": 0.0,
-      "learning_rate": 1.996506343430789e-05,
-      "loss": 1.1438,
+      "learning_rate": 1.9959876999093382e-05,
+      "loss": 1.0329,
       "step": 2033
     },
     {
-      "epoch": 0.05586223943313834,
+      "epoch": 0.05771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.996498910507844e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.995979470851813e-05,
+      "loss": 1.1328,
       "step": 2034
     },
     {
-      "epoch": 0.055889703660981574,
+      "epoch": 0.057746878547105564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964914697002142e-05,
-      "loss": 1.1197,
+      "learning_rate": 1.9959712333812047e-05,
+      "loss": 0.9457,
       "step": 2035
     },
     {
-      "epoch": 0.05591716788882481,
+      "epoch": 0.05777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.996484021007958e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.9959629874975823e-05,
+      "loss": 1.0764,
       "step": 2036
     },
     {
-      "epoch": 0.05594463211666804,
+      "epoch": 0.057803632236095345,
       "grad_norm": 0.0,
-      "learning_rate": 1.996476564431134e-05,
-      "loss": 1.123,
+      "learning_rate": 1.995954733201016e-05,
+      "loss": 1.009,
       "step": 2037
     },
     {
-      "epoch": 0.05597209634451127,
+      "epoch": 0.05783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964690999698013e-05,
-      "loss": 1.1594,
+      "learning_rate": 1.9959464704915753e-05,
+      "loss": 1.0139,
       "step": 2038
     },
     {
-      "epoch": 0.05599956057235451,
+      "epoch": 0.057860385925085134,
       "grad_norm": 0.0,
-      "learning_rate": 1.996461627624019e-05,
-      "loss": 1.1131,
+      "learning_rate": 1.99593819936933e-05,
+      "loss": 1.0575,
       "step": 2039
     },
     {
-      "epoch": 0.05602702480019774,
+      "epoch": 0.05788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964541473938465e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.99592991983435e-05,
+      "loss": 1.1348,
       "step": 2040
     },
     {
-      "epoch": 0.05605448902804098,
+      "epoch": 0.057917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964466592793425e-05,
-      "loss": 1.152,
+      "learning_rate": 1.995921631886705e-05,
+      "loss": 1.0089,
       "step": 2041
     },
     {
-      "epoch": 0.056081953255884213,
+      "epoch": 0.05794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.996439163280567e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9959133355264653e-05,
+      "loss": 1.1407,
       "step": 2042
     },
     {
-      "epoch": 0.05610941748372744,
+      "epoch": 0.0579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964316593975788e-05,
-      "loss": 1.1235,
+      "learning_rate": 1.9959050307537006e-05,
+      "loss": 1.0033,
       "step": 2043
     },
     {
-      "epoch": 0.05613688171157068,
+      "epoch": 0.05800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.996424147630437e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.9958967175684818e-05,
+      "loss": 1.0869,
       "step": 2044
     },
     {
-      "epoch": 0.05616434593941391,
+      "epoch": 0.058030646992054485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964166279792016e-05,
-      "loss": 1.0699,
+      "learning_rate": 1.9958883959708783e-05,
+      "loss": 1.0931,
       "step": 2045
     },
     {
-      "epoch": 0.05619181016725715,
+      "epoch": 0.05805902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.996409100443932e-05,
-      "loss": 1.1033,
+      "learning_rate": 1.995880065960961e-05,
+      "loss": 1.1234,
       "step": 2046
     },
     {
-      "epoch": 0.05621927439510038,
+      "epoch": 0.05808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.9964015650246873e-05,
-      "loss": 1.0896,
+      "learning_rate": 1.9958717275388e-05,
+      "loss": 1.0978,
       "step": 2047
     },
     {
-      "epoch": 0.05624673862294362,
+      "epoch": 0.05811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963940217215273e-05,
-      "loss": 1.1648,
+      "learning_rate": 1.9958633807044657e-05,
+      "loss": 1.062,
       "step": 2048
     },
     {
-      "epoch": 0.05627420285078685,
+      "epoch": 0.058144154370034055,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963864705345123e-05,
-      "loss": 1.0358,
+      "learning_rate": 1.9958550254580284e-05,
+      "loss": 1.0973,
       "step": 2049
     },
     {
-      "epoch": 0.05630166707863008,
+      "epoch": 0.05817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963789114637016e-05,
-      "loss": 1.0991,
+      "learning_rate": 1.9958466617995593e-05,
+      "loss": 1.1095,
       "step": 2050
     },
     {
-      "epoch": 0.056329131306473317,
+      "epoch": 0.05820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963713445091547e-05,
-      "loss": 1.1445,
+      "learning_rate": 1.9958382897291287e-05,
+      "loss": 1.1229,
       "step": 2051
     },
     {
-      "epoch": 0.05635659553431655,
+      "epoch": 0.05822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963637696709317e-05,
-      "loss": 1.1793,
+      "learning_rate": 1.995829909246807e-05,
+      "loss": 1.0529,
       "step": 2052
     },
     {
-      "epoch": 0.05638405976215979,
+      "epoch": 0.05825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963561869490928e-05,
-      "loss": 1.1923,
+      "learning_rate": 1.9958215203526655e-05,
+      "loss": 1.0524,
       "step": 2053
     },
     {
-      "epoch": 0.05641152399000302,
+      "epoch": 0.05828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.996348596343698e-05,
-      "loss": 1.1155,
+      "learning_rate": 1.9958131230467747e-05,
+      "loss": 0.9191,
       "step": 2054
     },
     {
-      "epoch": 0.05643898821784626,
+      "epoch": 0.05831441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963409978548067e-05,
-      "loss": 1.1576,
+      "learning_rate": 1.9958047173292058e-05,
+      "loss": 1.1382,
       "step": 2055
     },
     {
-      "epoch": 0.05646645244568949,
+      "epoch": 0.058342792281498294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963333914824798e-05,
-      "loss": 1.2032,
+      "learning_rate": 1.9957963032000298e-05,
+      "loss": 1.0986,
       "step": 2056
     },
     {
-      "epoch": 0.05649391667353272,
+      "epoch": 0.05837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.996325777226777e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.9957878806593176e-05,
+      "loss": 1.0134,
       "step": 2057
     },
     {
-      "epoch": 0.056521380901375956,
+      "epoch": 0.05839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963181550877588e-05,
-      "loss": 1.1315,
+      "learning_rate": 1.9957794497071403e-05,
+      "loss": 1.0594,
       "step": 2058
     },
     {
-      "epoch": 0.05654884512921919,
+      "epoch": 0.05842792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 1.9963105250654853e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.9957710103435694e-05,
+      "loss": 1.0177,
       "step": 2059
     },
     {
-      "epoch": 0.056576309357062426,
+      "epoch": 0.058456299659477864,
       "grad_norm": 0.0,
-      "learning_rate": 1.996302887160017e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.9957625625686757e-05,
+      "loss": 1.1846,
       "step": 2060
     },
     {
-      "epoch": 0.05660377358490566,
+      "epoch": 0.05848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962952413714145e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.9957541063825314e-05,
+      "loss": 1.1044,
       "step": 2061
     },
     {
-      "epoch": 0.0566312378127489,
+      "epoch": 0.05851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.996287587699738e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.995745641785207e-05,
+      "loss": 1.0801,
       "step": 2062
     },
     {
-      "epoch": 0.05665870204059213,
+      "epoch": 0.05854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962799261450484e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.995737168776774e-05,
+      "loss": 1.0537,
       "step": 2063
     },
     {
-      "epoch": 0.05668616626843536,
+      "epoch": 0.058569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.996272256707406e-05,
-      "loss": 1.059,
+      "learning_rate": 1.995728687357305e-05,
+      "loss": 1.1534,
       "step": 2064
     },
     {
-      "epoch": 0.056713630496278596,
+      "epoch": 0.05859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.996264579386871e-05,
-      "loss": 1.1174,
+      "learning_rate": 1.995720197526871e-05,
+      "loss": 1.0161,
       "step": 2065
     },
     {
-      "epoch": 0.05674109472412183,
+      "epoch": 0.058626560726447216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962568941835053e-05,
-      "loss": 1.1178,
+      "learning_rate": 1.9957116992855435e-05,
+      "loss": 1.0196,
       "step": 2066
     },
     {
-      "epoch": 0.056768558951965066,
+      "epoch": 0.05865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962492010973694e-05,
-      "loss": 1.1415,
+      "learning_rate": 1.9957031926333944e-05,
+      "loss": 1.1523,
       "step": 2067
     },
     {
-      "epoch": 0.0567960231798083,
+      "epoch": 0.058683314415437005,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962415001285235e-05,
-      "loss": 1.0,
+      "learning_rate": 1.995694677570496e-05,
+      "loss": 1.0767,
       "step": 2068
     },
     {
-      "epoch": 0.056823487407651536,
+      "epoch": 0.0587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.996233791277029e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9956861540969198e-05,
+      "loss": 1.1447,
       "step": 2069
     },
     {
-      "epoch": 0.056850951635494765,
+      "epoch": 0.058740068104426786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962260745429466e-05,
-      "loss": 1.097,
+      "learning_rate": 1.9956776222127376e-05,
+      "loss": 0.9709,
       "step": 2070
     },
     {
-      "epoch": 0.056878415863338,
+      "epoch": 0.05876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.996218349926338e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.995669081918022e-05,
+      "loss": 1.0032,
       "step": 2071
     },
     {
-      "epoch": 0.056905880091181235,
+      "epoch": 0.058796821793416575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9962106174272637e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.995660533212845e-05,
+      "loss": 1.0803,
       "step": 2072
     },
     {
-      "epoch": 0.05693334431902447,
+      "epoch": 0.05882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.996202877045785e-05,
-      "loss": 1.0867,
+      "learning_rate": 1.9956519760972787e-05,
+      "loss": 1.0918,
       "step": 2073
     },
     {
-      "epoch": 0.056960808546867706,
+      "epoch": 0.058853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961951287819636e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.9956434105713954e-05,
+      "loss": 1.0907,
       "step": 2074
     },
     {
-      "epoch": 0.05698827277471094,
+      "epoch": 0.05888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961873726358605e-05,
-      "loss": 1.0876,
+      "learning_rate": 1.9956348366352675e-05,
+      "loss": 1.0871,
       "step": 2075
     },
     {
-      "epoch": 0.057015737002554176,
+      "epoch": 0.05891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961796086075368e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.9956262542889674e-05,
+      "loss": 1.0328,
       "step": 2076
     },
     {
-      "epoch": 0.057043201230397404,
+      "epoch": 0.05893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.996171836697054e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.9956176635325676e-05,
+      "loss": 1.0143,
       "step": 2077
     },
     {
-      "epoch": 0.05707066545824064,
+      "epoch": 0.058967082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 1.996164056904474e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.99560906436614e-05,
+      "loss": 0.9993,
       "step": 2078
     },
     {
-      "epoch": 0.057098129686083875,
+      "epoch": 0.05899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.996156269229858e-05,
-      "loss": 1.1027,
+      "learning_rate": 1.9956004567897587e-05,
+      "loss": 1.0403,
       "step": 2079
     },
     {
-      "epoch": 0.05712559391392711,
+      "epoch": 0.05902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961484736732677e-05,
-      "loss": 1.1761,
+      "learning_rate": 1.995591840803495e-05,
+      "loss": 0.9684,
       "step": 2080
     },
     {
-      "epoch": 0.057153058141770345,
+      "epoch": 0.0590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961406702347646e-05,
-      "loss": 1.1399,
+      "learning_rate": 1.9955832164074226e-05,
+      "loss": 1.08,
       "step": 2081
     },
     {
-      "epoch": 0.05718052236961358,
+      "epoch": 0.059080590238365496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961328589144112e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.995574583601614e-05,
+      "loss": 1.0548,
       "step": 2082
     },
     {
-      "epoch": 0.057207986597456816,
+      "epoch": 0.059108967082860384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961250397122684e-05,
-      "loss": 1.1214,
+      "learning_rate": 1.9955659423861423e-05,
+      "loss": 0.9439,
       "step": 2083
     },
     {
-      "epoch": 0.057235450825300044,
+      "epoch": 0.05913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961172126283984e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.9955572927610798e-05,
+      "loss": 1.0931,
       "step": 2084
     },
     {
-      "epoch": 0.05726291505314328,
+      "epoch": 0.05916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.996109377662863e-05,
-      "loss": 1.1302,
+      "learning_rate": 1.9955486347265006e-05,
+      "loss": 1.0887,
       "step": 2085
     },
     {
-      "epoch": 0.057290379280986514,
+      "epoch": 0.05919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961015348157245e-05,
-      "loss": 1.1644,
+      "learning_rate": 1.995539968282477e-05,
+      "loss": 1.0127,
       "step": 2086
     },
     {
-      "epoch": 0.05731784350882975,
+      "epoch": 0.059222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.996093684087045e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.9955312934290827e-05,
+      "loss": 1.1313,
       "step": 2087
     },
     {
-      "epoch": 0.057345307736672985,
+      "epoch": 0.05925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9960858254768865e-05,
-      "loss": 1.0677,
+      "learning_rate": 1.9955226101663906e-05,
+      "loss": 1.0292,
       "step": 2088
     },
     {
-      "epoch": 0.05737277196451622,
+      "epoch": 0.05927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.996077958985311e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.9955139184944746e-05,
+      "loss": 1.1541,
       "step": 2089
     },
     {
-      "epoch": 0.057400236192359455,
+      "epoch": 0.05930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.9960700846123803e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9955052184134077e-05,
+      "loss": 1.0131,
       "step": 2090
     },
     {
-      "epoch": 0.05742770042020268,
+      "epoch": 0.059335981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.996062202358158e-05,
-      "loss": 1.092,
+      "learning_rate": 1.9954965099232636e-05,
+      "loss": 1.1861,
       "step": 2091
     },
     {
-      "epoch": 0.05745516464804592,
+      "epoch": 0.05936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.996054312222705e-05,
-      "loss": 1.1054,
+      "learning_rate": 1.9954877930241152e-05,
+      "loss": 1.0425,
       "step": 2092
     },
     {
-      "epoch": 0.057482628875889154,
+      "epoch": 0.059392735527809305,
       "grad_norm": 0.0,
-      "learning_rate": 1.996046414206085e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.9954790677160368e-05,
+      "loss": 1.1117,
       "step": 2093
     },
     {
-      "epoch": 0.05751009310373239,
+      "epoch": 0.0594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.99603850830836e-05,
-      "loss": 1.111,
+      "learning_rate": 1.9954703339991024e-05,
+      "loss": 1.0251,
       "step": 2094
     },
     {
-      "epoch": 0.057537557331575624,
+      "epoch": 0.059449489216799094,
       "grad_norm": 0.0,
-      "learning_rate": 1.996030594529592e-05,
-      "loss": 1.1931,
+      "learning_rate": 1.995461591873385e-05,
+      "loss": 1.1614,
       "step": 2095
     },
     {
-      "epoch": 0.05756502155941886,
+      "epoch": 0.05947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.9960226728698447e-05,
-      "loss": 1.1327,
+      "learning_rate": 1.995452841338959e-05,
+      "loss": 1.1038,
       "step": 2096
     },
     {
-      "epoch": 0.05759248578726209,
+      "epoch": 0.059506242905788875,
       "grad_norm": 0.0,
-      "learning_rate": 1.99601474332918e-05,
-      "loss": 1.1391,
+      "learning_rate": 1.995444082395898e-05,
+      "loss": 1.0569,
       "step": 2097
     },
     {
-      "epoch": 0.05761995001510532,
+      "epoch": 0.05953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9960068059076606e-05,
-      "loss": 1.0922,
+      "learning_rate": 1.995435315044276e-05,
+      "loss": 1.1608,
       "step": 2098
     },
     {
-      "epoch": 0.05764741424294856,
+      "epoch": 0.059562996594778664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959988606053496e-05,
-      "loss": 1.1633,
+      "learning_rate": 1.995426539284167e-05,
+      "loss": 1.0686,
       "step": 2099
     },
     {
-      "epoch": 0.05767487847079179,
+      "epoch": 0.05959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959909074223098e-05,
-      "loss": 1.194,
+      "learning_rate": 1.9954177551156454e-05,
+      "loss": 1.1168,
       "step": 2100
     },
     {
-      "epoch": 0.05770234269863503,
+      "epoch": 0.059619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959829463586042e-05,
-      "loss": 1.171,
+      "learning_rate": 1.9954089625387856e-05,
+      "loss": 1.0912,
       "step": 2101
     },
     {
-      "epoch": 0.057729806926478264,
+      "epoch": 0.05964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.995974977414296e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.9954001615536612e-05,
+      "loss": 1.1555,
       "step": 2102
     },
     {
-      "epoch": 0.0577572711543215,
+      "epoch": 0.05967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959670005894474e-05,
-      "loss": 1.1126,
+      "learning_rate": 1.995391352160347e-05,
+      "loss": 1.0819,
       "step": 2103
     },
     {
-      "epoch": 0.05778473538216473,
+      "epoch": 0.05970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959590158841223e-05,
-      "loss": 1.1894,
+      "learning_rate": 1.9953825343589175e-05,
+      "loss": 1.012,
       "step": 2104
     },
     {
-      "epoch": 0.05781219961000796,
+      "epoch": 0.059733257661748015,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959510232983837e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.9953737081494466e-05,
+      "loss": 1.0555,
       "step": 2105
     },
     {
-      "epoch": 0.0578396638378512,
+      "epoch": 0.0597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959430228322948e-05,
-      "loss": 1.1527,
+      "learning_rate": 1.9953648735320096e-05,
+      "loss": 1.1178,
       "step": 2106
     },
     {
-      "epoch": 0.05786712806569443,
+      "epoch": 0.0597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959350144859192e-05,
-      "loss": 1.1478,
+      "learning_rate": 1.9953560305066806e-05,
+      "loss": 1.0199,
       "step": 2107
     },
     {
-      "epoch": 0.05789459229353767,
+      "epoch": 0.05981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959269982593197e-05,
-      "loss": 1.0979,
+      "learning_rate": 1.9953471790735345e-05,
+      "loss": 1.0753,
       "step": 2108
     },
     {
-      "epoch": 0.0579220565213809,
+      "epoch": 0.059846765039727585,
       "grad_norm": 0.0,
-      "learning_rate": 1.99591897415256e-05,
-      "loss": 1.166,
+      "learning_rate": 1.995338319232646e-05,
+      "loss": 1.0781,
       "step": 2109
     },
     {
-      "epoch": 0.05794952074922414,
+      "epoch": 0.05987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9959109421657038e-05,
-      "loss": 1.1562,
+      "learning_rate": 1.9953294509840896e-05,
+      "loss": 1.1143,
       "step": 2110
     },
     {
-      "epoch": 0.05797698497706737,
+      "epoch": 0.05990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.995902902298814e-05,
-      "loss": 1.1491,
+      "learning_rate": 1.995320574327941e-05,
+      "loss": 1.014,
       "step": 2111
     },
     {
-      "epoch": 0.0580044492049106,
+      "epoch": 0.05993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.995894854551955e-05,
-      "loss": 1.1004,
+      "learning_rate": 1.995311689264275e-05,
+      "loss": 1.11,
       "step": 2112
     },
     {
-      "epoch": 0.05803191343275384,
+      "epoch": 0.05996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958867989251903e-05,
-      "loss": 1.08,
+      "learning_rate": 1.9953027957931658e-05,
+      "loss": 1.0388,
       "step": 2113
     },
     {
-      "epoch": 0.05805937766059707,
+      "epoch": 0.05998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.995878735418583e-05,
-      "loss": 1.126,
+      "learning_rate": 1.9952938939146897e-05,
+      "loss": 1.0557,
       "step": 2114
     },
     {
-      "epoch": 0.05808684188844031,
+      "epoch": 0.06001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.995870664032198e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.995284983628921e-05,
+      "loss": 1.1618,
       "step": 2115
     },
     {
-      "epoch": 0.05811430611628354,
+      "epoch": 0.060045402951191824,
       "grad_norm": 0.0,
-      "learning_rate": 1.995862584766098e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.9952760649359354e-05,
+      "loss": 1.1568,
       "step": 2116
     },
     {
-      "epoch": 0.05814177034412678,
+      "epoch": 0.06007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958544976203475e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.9952671378358085e-05,
+      "loss": 1.1805,
       "step": 2117
     },
     {
-      "epoch": 0.058169234571970006,
+      "epoch": 0.06010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958464025950105e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.9952582023286145e-05,
+      "loss": 1.1037,
       "step": 2118
     },
     {
-      "epoch": 0.05819669879981324,
+      "epoch": 0.06013053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.995838299690151e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9952492584144302e-05,
+      "loss": 0.9972,
       "step": 2119
     },
     {
-      "epoch": 0.05822416302765648,
+      "epoch": 0.060158910329171394,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958301889058336e-05,
-      "loss": 1.0826,
+      "learning_rate": 1.995240306093331e-05,
+      "loss": 1.0449,
       "step": 2120
     },
     {
-      "epoch": 0.05825162725549971,
+      "epoch": 0.06018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958220702421217e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.995231345365392e-05,
+      "loss": 1.0334,
       "step": 2121
     },
     {
-      "epoch": 0.05827909148334295,
+      "epoch": 0.06021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9958139436990794e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.995222376230689e-05,
+      "loss": 1.1099,
       "step": 2122
     },
     {
-      "epoch": 0.05830655571118618,
+      "epoch": 0.06024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.995805809276772e-05,
-      "loss": 1.1191,
+      "learning_rate": 1.9952133986892976e-05,
+      "loss": 1.0894,
       "step": 2123
     },
     {
-      "epoch": 0.05833401993902942,
+      "epoch": 0.060272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957976669752633e-05,
-      "loss": 1.081,
+      "learning_rate": 1.995204412741294e-05,
+      "loss": 1.009,
       "step": 2124
     },
     {
-      "epoch": 0.058361484166872646,
+      "epoch": 0.06030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957895167946173e-05,
-      "loss": 1.102,
+      "learning_rate": 1.995195418386754e-05,
+      "loss": 1.1465,
       "step": 2125
     },
     {
-      "epoch": 0.05838894839471588,
+      "epoch": 0.060329171396140746,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957813587348993e-05,
-      "loss": 1.1228,
+      "learning_rate": 1.9951864156257537e-05,
+      "loss": 1.0267,
       "step": 2126
     },
     {
-      "epoch": 0.058416412622559116,
+      "epoch": 0.06035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957731927961733e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.995177404458369e-05,
+      "loss": 0.9958,
       "step": 2127
     },
     {
-      "epoch": 0.05844387685040235,
+      "epoch": 0.060385925085130535,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957650189785037e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.9951683848846764e-05,
+      "loss": 1.0534,
       "step": 2128
     },
     {
-      "epoch": 0.05847134107824559,
+      "epoch": 0.06041430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957568372819562e-05,
-      "loss": 1.1725,
+      "learning_rate": 1.9951593569047513e-05,
+      "loss": 1.0239,
       "step": 2129
     },
     {
-      "epoch": 0.05849880530608882,
+      "epoch": 0.060442678774120316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957486477065946e-05,
-      "loss": 1.0687,
+      "learning_rate": 1.9951503205186703e-05,
+      "loss": 1.1071,
       "step": 2130
     },
     {
-      "epoch": 0.05852626953393205,
+      "epoch": 0.06047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.995740450252484e-05,
-      "loss": 1.1572,
+      "learning_rate": 1.99514127572651e-05,
+      "loss": 1.0476,
       "step": 2131
     },
     {
-      "epoch": 0.058553733761775285,
+      "epoch": 0.060499432463110105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957322449196893e-05,
-      "loss": 1.0721,
+      "learning_rate": 1.9951322225283466e-05,
+      "loss": 0.9771,
       "step": 2132
     },
     {
-      "epoch": 0.05858119798961852,
+      "epoch": 0.06052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957240317082754e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.9951231609242563e-05,
+      "loss": 1.0548,
       "step": 2133
     },
     {
-      "epoch": 0.058608662217461756,
+      "epoch": 0.060556186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.995715810618307e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.995114090914316e-05,
+      "loss": 0.9857,
       "step": 2134
     },
     {
-      "epoch": 0.05863612644530499,
+      "epoch": 0.06058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.9957075816498498e-05,
-      "loss": 1.1148,
+      "learning_rate": 1.9951050124986023e-05,
+      "loss": 0.9822,
       "step": 2135
     },
     {
-      "epoch": 0.058663590673148226,
+      "epoch": 0.06061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.995699344802968e-05,
-      "loss": 1.1845,
+      "learning_rate": 1.995095925677192e-05,
+      "loss": 1.1638,
       "step": 2136
     },
     {
-      "epoch": 0.05869105490099146,
+      "epoch": 0.06064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956911000777273e-05,
-      "loss": 1.1538,
+      "learning_rate": 1.9950868304501617e-05,
+      "loss": 1.035,
       "step": 2137
     },
     {
-      "epoch": 0.05871851912883469,
+      "epoch": 0.060669693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956828474741935e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.995077726817588e-05,
+      "loss": 1.077,
       "step": 2138
     },
     {
-      "epoch": 0.058745983356677925,
+      "epoch": 0.06069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.995674586992431e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.9950686147795483e-05,
+      "loss": 1.1177,
       "step": 2139
     },
     {
-      "epoch": 0.05877344758452116,
+      "epoch": 0.06072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956663186325053e-05,
-      "loss": 1.1552,
+      "learning_rate": 1.995059494336119e-05,
+      "loss": 1.0376,
       "step": 2140
     },
     {
-      "epoch": 0.058800911812364395,
+      "epoch": 0.06075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956580423944825e-05,
-      "loss": 1.1414,
+      "learning_rate": 1.9950503654873775e-05,
+      "loss": 1.1316,
       "step": 2141
     },
     {
-      "epoch": 0.05882837604020763,
+      "epoch": 0.060783200908059026,
       "grad_norm": 0.0,
-      "learning_rate": 1.995649758278427e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.9950412282334005e-05,
+      "loss": 1.0558,
       "step": 2142
     },
     {
-      "epoch": 0.058855840268050866,
+      "epoch": 0.060811577752553914,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956414662844053e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.9950320825742658e-05,
+      "loss": 0.9715,
       "step": 2143
     },
     {
-      "epoch": 0.0588833044958941,
+      "epoch": 0.06083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956331664124824e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.9950229285100506e-05,
+      "loss": 1.1074,
       "step": 2144
     },
     {
-      "epoch": 0.05891076872373733,
+      "epoch": 0.0608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956248586627245e-05,
-      "loss": 1.2621,
+      "learning_rate": 1.9950137660408318e-05,
+      "loss": 1.1356,
       "step": 2145
     },
     {
-      "epoch": 0.058938232951580564,
+      "epoch": 0.06089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.995616543035197e-05,
-      "loss": 1.1174,
+      "learning_rate": 1.995004595166687e-05,
+      "loss": 1.1609,
       "step": 2146
     },
     {
-      "epoch": 0.0589656971794238,
+      "epoch": 0.060925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9956082195299656e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.9949954158876938e-05,
+      "loss": 1.1341,
       "step": 2147
     },
     {
-      "epoch": 0.058993161407267035,
+      "epoch": 0.06095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955998881470964e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.994986228203929e-05,
+      "loss": 1.0507,
       "step": 2148
     },
     {
-      "epoch": 0.05902062563511027,
+      "epoch": 0.06098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.995591548886655e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.9949770321154715e-05,
+      "loss": 1.0827,
       "step": 2149
     },
     {
-      "epoch": 0.059048089862953505,
+      "epoch": 0.06101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.995583201748708e-05,
-      "loss": 1.1311,
+      "learning_rate": 1.994967827622398e-05,
+      "loss": 1.0328,
       "step": 2150
     },
     {
-      "epoch": 0.05907555409079674,
+      "epoch": 0.061038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955748467333207e-05,
-      "loss": 1.1152,
+      "learning_rate": 1.9949586147247867e-05,
+      "loss": 1.1063,
       "step": 2151
     },
     {
-      "epoch": 0.05910301831863997,
+      "epoch": 0.06106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.99556648384056e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.9949493934227153e-05,
+      "loss": 1.0448,
       "step": 2152
     },
     {
-      "epoch": 0.059130482546483204,
+      "epoch": 0.061095346197502835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955581130704913e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.994940163716261e-05,
+      "loss": 0.9604,
       "step": 2153
     },
     {
-      "epoch": 0.05915794677432644,
+      "epoch": 0.06112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955497344231812e-05,
-      "loss": 1.1131,
+      "learning_rate": 1.9949309256055028e-05,
+      "loss": 1.0205,
       "step": 2154
     },
     {
-      "epoch": 0.059185411002169674,
+      "epoch": 0.061152099886492624,
       "grad_norm": 0.0,
-      "learning_rate": 1.995541347898696e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.9949216790905186e-05,
+      "loss": 0.9872,
       "step": 2155
     },
     {
-      "epoch": 0.05921287523001291,
+      "epoch": 0.06118047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955329534971018e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.994912424171386e-05,
+      "loss": 1.0592,
       "step": 2156
     },
     {
-      "epoch": 0.059240339457856145,
+      "epoch": 0.061208853575482405,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955245512184656e-05,
-      "loss": 1.1523,
+      "learning_rate": 1.994903160848184e-05,
+      "loss": 1.2107,
       "step": 2157
     },
     {
-      "epoch": 0.05926780368569937,
+      "epoch": 0.0612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955161410628534e-05,
-      "loss": 1.0733,
+      "learning_rate": 1.994893889120989e-05,
+      "loss": 1.0016,
       "step": 2158
     },
     {
-      "epoch": 0.05929526791354261,
+      "epoch": 0.061265607264472194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9955077230303317e-05,
-      "loss": 1.1375,
+      "learning_rate": 1.9948846089898816e-05,
+      "loss": 1.1381,
       "step": 2159
     },
     {
-      "epoch": 0.059322732141385844,
+      "epoch": 0.06129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954992971209674e-05,
-      "loss": 1.1806,
+      "learning_rate": 1.9948753204549393e-05,
+      "loss": 1.0269,
       "step": 2160
     },
     {
-      "epoch": 0.05935019636922908,
+      "epoch": 0.061322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954908633348272e-05,
-      "loss": 0.9215,
+      "learning_rate": 1.99486602351624e-05,
+      "loss": 1.0975,
       "step": 2161
     },
     {
-      "epoch": 0.059377660597072314,
+      "epoch": 0.06135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954824216719776e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.9948567181738627e-05,
+      "loss": 1.0718,
       "step": 2162
     },
     {
-      "epoch": 0.05940512482491555,
+      "epoch": 0.06137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954739721324853e-05,
-      "loss": 1.1408,
+      "learning_rate": 1.994847404427886e-05,
+      "loss": 1.0911,
       "step": 2163
     },
     {
-      "epoch": 0.059432589052758784,
+      "epoch": 0.06140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954655147164176e-05,
-      "loss": 1.075,
+      "learning_rate": 1.9948380822783887e-05,
+      "loss": 1.0464,
       "step": 2164
     },
     {
-      "epoch": 0.05946005328060201,
+      "epoch": 0.061435868331441545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954570494238407e-05,
-      "loss": 1.1329,
+      "learning_rate": 1.994828751725449e-05,
+      "loss": 0.9786,
       "step": 2165
     },
     {
-      "epoch": 0.05948751750844525,
+      "epoch": 0.06146424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954485762548223e-05,
-      "loss": 1.1376,
+      "learning_rate": 1.9948194127691467e-05,
+      "loss": 1.0974,
       "step": 2166
     },
     {
-      "epoch": 0.05951498173628848,
+      "epoch": 0.06149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954400952094293e-05,
-      "loss": 1.069,
+      "learning_rate": 1.9948100654095596e-05,
+      "loss": 1.0373,
       "step": 2167
     },
     {
-      "epoch": 0.05954244596413172,
+      "epoch": 0.06152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954316062877285e-05,
-      "loss": 1.1691,
+      "learning_rate": 1.9948007096467673e-05,
+      "loss": 1.0705,
       "step": 2168
     },
     {
-      "epoch": 0.059569910191974954,
+      "epoch": 0.061549375709421116,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954231094897873e-05,
-      "loss": 1.0702,
+      "learning_rate": 1.994791345480849e-05,
+      "loss": 1.0131,
       "step": 2169
     },
     {
-      "epoch": 0.05959737441981819,
+      "epoch": 0.061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9954146048156725e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.9947819729118833e-05,
+      "loss": 1.1085,
       "step": 2170
     },
     {
-      "epoch": 0.059624838647661424,
+      "epoch": 0.0616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.995406092265452e-05,
-      "loss": 1.1378,
+      "learning_rate": 1.9947725919399496e-05,
+      "loss": 0.997,
       "step": 2171
     },
     {
-      "epoch": 0.05965230287550465,
+      "epoch": 0.06163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.995397571839193e-05,
-      "loss": 1.1063,
+      "learning_rate": 1.9947632025651267e-05,
+      "loss": 1.0521,
       "step": 2172
     },
     {
-      "epoch": 0.05967976710334789,
+      "epoch": 0.06166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.995389043536963e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.9947538047874948e-05,
+      "loss": 1.0536,
       "step": 2173
     },
     {
-      "epoch": 0.05970723133119112,
+      "epoch": 0.06169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953805073588294e-05,
-      "loss": 1.1424,
+      "learning_rate": 1.9947443986071328e-05,
+      "loss": 0.9862,
       "step": 2174
     },
     {
-      "epoch": 0.05973469555903436,
+      "epoch": 0.06171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953719633048593e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.99473498402412e-05,
+      "loss": 1.2814,
       "step": 2175
     },
     {
-      "epoch": 0.05976215978687759,
+      "epoch": 0.061748013620885354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953634113751207e-05,
-      "loss": 1.192,
+      "learning_rate": 1.9947255610385357e-05,
+      "loss": 1.0329,
       "step": 2176
     },
     {
-      "epoch": 0.05978962401472083,
+      "epoch": 0.06177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.995354851569681e-05,
-      "loss": 1.0372,
+      "learning_rate": 1.9947161296504603e-05,
+      "loss": 1.0471,
       "step": 2177
     },
     {
-      "epoch": 0.059817088242564063,
+      "epoch": 0.06180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953462838886083e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.9947066898599728e-05,
+      "loss": 0.9783,
       "step": 2178
     },
     {
-      "epoch": 0.05984455247040729,
+      "epoch": 0.06183314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 1.995337708331971e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.9946972416671533e-05,
+      "loss": 1.149,
       "step": 2179
     },
     {
-      "epoch": 0.05987201669825053,
+      "epoch": 0.061861520998864925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953291248998348e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.9946877850720818e-05,
+      "loss": 1.0159,
       "step": 2180
     },
     {
-      "epoch": 0.05989948092609376,
+      "epoch": 0.06188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.99532053359227e-05,
-      "loss": 1.1505,
+      "learning_rate": 1.9946783200748374e-05,
+      "loss": 1.0466,
       "step": 2181
     },
     {
-      "epoch": 0.059926945153937,
+      "epoch": 0.06191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953119344093436e-05,
-      "loss": 1.0631,
+      "learning_rate": 1.9946688466755006e-05,
+      "loss": 1.1643,
       "step": 2182
     },
     {
-      "epoch": 0.05995440938178023,
+      "epoch": 0.0619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9953033273511232e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.9946593648741515e-05,
+      "loss": 1.1557,
       "step": 2183
     },
     {
-      "epoch": 0.05998187360962347,
+      "epoch": 0.061975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9952947124176774e-05,
-      "loss": 1.1206,
+      "learning_rate": 1.99464987467087e-05,
+      "loss": 1.0425,
       "step": 2184
     },
     {
-      "epoch": 0.0600093378374667,
+      "epoch": 0.06200340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.9952860896090748e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.994640376065736e-05,
+      "loss": 1.1028,
       "step": 2185
     },
     {
-      "epoch": 0.06003680206530993,
+      "epoch": 0.062031782065834276,
       "grad_norm": 0.0,
-      "learning_rate": 1.9952774589253823e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.9946308690588304e-05,
+      "loss": 1.1209,
       "step": 2186
     },
     {
-      "epoch": 0.060064266293153167,
+      "epoch": 0.06206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.9952688203666697e-05,
-      "loss": 1.2063,
+      "learning_rate": 1.994621353650233e-05,
+      "loss": 1.0532,
       "step": 2187
     },
     {
-      "epoch": 0.0600917305209964,
+      "epoch": 0.062088535754824065,
       "grad_norm": 0.0,
-      "learning_rate": 1.995260173933004e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.9946118298400242e-05,
+      "loss": 0.9855,
       "step": 2188
     },
     {
-      "epoch": 0.06011919474883964,
+      "epoch": 0.06211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.9952515196244547e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.994602297628285e-05,
+      "loss": 1.149,
       "step": 2189
     },
     {
-      "epoch": 0.06014665897668287,
+      "epoch": 0.062145289443813846,
       "grad_norm": 0.0,
-      "learning_rate": 1.99524285744109e-05,
-      "loss": 1.0919,
+      "learning_rate": 1.994592757015095e-05,
+      "loss": 1.2413,
       "step": 2190
     },
     {
-      "epoch": 0.06017412320452611,
+      "epoch": 0.06217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.995234187382978e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.9945832080005354e-05,
+      "loss": 1.1023,
       "step": 2191
     },
     {
-      "epoch": 0.060201587432369336,
+      "epoch": 0.062202043132803635,
       "grad_norm": 0.0,
-      "learning_rate": 1.995225509450188e-05,
-      "loss": 1.1963,
+      "learning_rate": 1.9945736505846866e-05,
+      "loss": 1.0183,
       "step": 2192
     },
     {
-      "epoch": 0.06022905166021257,
+      "epoch": 0.06223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.995216823642788e-05,
-      "loss": 1.098,
+      "learning_rate": 1.9945640847676296e-05,
+      "loss": 1.0215,
       "step": 2193
     },
     {
-      "epoch": 0.060256515888055806,
+      "epoch": 0.062258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.995208129960847e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.9945545105494454e-05,
+      "loss": 0.9984,
       "step": 2194
     },
     {
-      "epoch": 0.06028398011589904,
+      "epoch": 0.06228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951994284044335e-05,
-      "loss": 1.0462,
+      "learning_rate": 1.994544927930214e-05,
+      "loss": 1.199,
       "step": 2195
     },
     {
-      "epoch": 0.060311444343742276,
+      "epoch": 0.0623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951907189736167e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.9945353369100172e-05,
+      "loss": 1.0598,
       "step": 2196
     },
     {
-      "epoch": 0.06033890857158551,
+      "epoch": 0.06234392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951820016684657e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.994525737488936e-05,
+      "loss": 1.0331,
       "step": 2197
     },
     {
-      "epoch": 0.06036637279942875,
+      "epoch": 0.062372304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 1.995173276489049e-05,
-      "loss": 1.1301,
+      "learning_rate": 1.9945161296670505e-05,
+      "loss": 1.0437,
       "step": 2198
     },
     {
-      "epoch": 0.060393837027271975,
+      "epoch": 0.06240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.995164543435436e-05,
-      "loss": 1.1986,
+      "learning_rate": 1.994506513444443e-05,
+      "loss": 0.9817,
       "step": 2199
     },
     {
-      "epoch": 0.06042130125511521,
+      "epoch": 0.06242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951558025076955e-05,
-      "loss": 1.0254,
+      "learning_rate": 1.994496888821194e-05,
+      "loss": 1.0573,
       "step": 2200
     },
     {
-      "epoch": 0.060448765482958446,
+      "epoch": 0.06245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951470537058972e-05,
-      "loss": 1.1257,
+      "learning_rate": 1.994487255797385e-05,
+      "loss": 1.1385,
       "step": 2201
     },
     {
-      "epoch": 0.06047622971080168,
+      "epoch": 0.062485811577752556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951382970301094e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.9944776143730978e-05,
+      "loss": 1.1289,
       "step": 2202
     },
     {
-      "epoch": 0.060503693938644916,
+      "epoch": 0.06251418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.995129532480402e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.9944679645484136e-05,
+      "loss": 1.2634,
       "step": 2203
     },
     {
-      "epoch": 0.06053115816648815,
+      "epoch": 0.06254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951207600568442e-05,
-      "loss": 1.1957,
+      "learning_rate": 1.9944583063234132e-05,
+      "loss": 1.0071,
       "step": 2204
     },
     {
-      "epoch": 0.060558622394331386,
+      "epoch": 0.06257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951119797595058e-05,
-      "loss": 1.1597,
+      "learning_rate": 1.994448639698179e-05,
+      "loss": 1.1777,
       "step": 2205
     },
     {
-      "epoch": 0.060586086622174615,
+      "epoch": 0.06259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.9951031915884556e-05,
-      "loss": 1.1558,
+      "learning_rate": 1.994438964672793e-05,
+      "loss": 1.2089,
       "step": 2206
     },
     {
-      "epoch": 0.06061355085001785,
+      "epoch": 0.06262769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 1.995094395543764e-05,
-      "loss": 1.0923,
+      "learning_rate": 1.994429281247336e-05,
+      "loss": 0.9887,
       "step": 2207
     },
     {
-      "epoch": 0.060641015077861085,
+      "epoch": 0.06265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950855916254996e-05,
-      "loss": 1.1626,
+      "learning_rate": 1.9944195894218897e-05,
+      "loss": 1.0681,
       "step": 2208
     },
     {
-      "epoch": 0.06066847930570432,
+      "epoch": 0.0626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950767798337327e-05,
-      "loss": 1.0943,
+      "learning_rate": 1.994409889196537e-05,
+      "loss": 0.9717,
       "step": 2209
     },
     {
-      "epoch": 0.060695943533547556,
+      "epoch": 0.0627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.995067960168533e-05,
-      "loss": 1.0827,
+      "learning_rate": 1.994400180571359e-05,
+      "loss": 1.0571,
       "step": 2210
     },
     {
-      "epoch": 0.06072340776139079,
+      "epoch": 0.06274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950591326299697e-05,
-      "loss": 1.1484,
+      "learning_rate": 1.9943904635464383e-05,
+      "loss": 1.0356,
       "step": 2211
     },
     {
-      "epoch": 0.060750871989234026,
+      "epoch": 0.06276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950502972181137e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.994380738121856e-05,
+      "loss": 1.0749,
       "step": 2212
     },
     {
-      "epoch": 0.060778336217077254,
+      "epoch": 0.06279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.995041453933034e-05,
-      "loss": 1.0041,
+      "learning_rate": 1.9943710042976955e-05,
+      "loss": 1.1425,
       "step": 2213
     },
     {
-      "epoch": 0.06080580044492049,
+      "epoch": 0.06282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950326027748005e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9943612620740383e-05,
+      "loss": 1.1008,
       "step": 2214
     },
     {
-      "epoch": 0.060833264672763725,
+      "epoch": 0.06285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.995023743743484e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.9943515114509667e-05,
+      "loss": 0.9453,
       "step": 2215
     },
     {
-      "epoch": 0.06086072890060696,
+      "epoch": 0.06288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950148768391544e-05,
-      "loss": 1.069,
+      "learning_rate": 1.994341752428563e-05,
+      "loss": 1.0688,
       "step": 2216
     },
     {
-      "epoch": 0.060888193128450195,
+      "epoch": 0.06291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950060020618816e-05,
-      "loss": 1.1979,
+      "learning_rate": 1.9943319850069104e-05,
+      "loss": 1.0365,
       "step": 2217
     },
     {
-      "epoch": 0.06091565735629343,
+      "epoch": 0.06293984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949971194117354e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.9943222091860902e-05,
+      "loss": 1.0635,
       "step": 2218
     },
     {
-      "epoch": 0.06094312158413666,
+      "epoch": 0.06296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.994988228888787e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.994312424966186e-05,
+      "loss": 0.9835,
       "step": 2219
     },
     {
-      "epoch": 0.060970585811979894,
+      "epoch": 0.06299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949793304931062e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.99430263234728e-05,
+      "loss": 1.0384,
       "step": 2220
     },
     {
-      "epoch": 0.06099805003982313,
+      "epoch": 0.0630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949704242247636e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9942928313294548e-05,
+      "loss": 1.0839,
       "step": 2221
     },
     {
-      "epoch": 0.061025514267666364,
+      "epoch": 0.0630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949615100838294e-05,
-      "loss": 1.1431,
+      "learning_rate": 1.9942830219127935e-05,
+      "loss": 0.986,
       "step": 2222
     },
     {
-      "epoch": 0.0610529784955096,
+      "epoch": 0.06308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949525880703745e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.9942732040973788e-05,
+      "loss": 0.9996,
       "step": 2223
     },
     {
-      "epoch": 0.061080442723352835,
+      "epoch": 0.06311010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949436581844693e-05,
-      "loss": 1.019,
+      "learning_rate": 1.9942633778832936e-05,
+      "loss": 1.0829,
       "step": 2224
     },
     {
-      "epoch": 0.06110790695119607,
+      "epoch": 0.06313847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949347204261844e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.9942535432706212e-05,
+      "loss": 1.0043,
       "step": 2225
     },
     {
-      "epoch": 0.0611353711790393,
+      "epoch": 0.06316685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949257747955905e-05,
-      "loss": 1.0856,
+      "learning_rate": 1.994243700259444e-05,
+      "loss": 1.1107,
       "step": 2226
     },
     {
-      "epoch": 0.06116283540688253,
+      "epoch": 0.06319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949168212927587e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.9942338488498457e-05,
+      "loss": 0.9619,
       "step": 2227
     },
     {
-      "epoch": 0.06119029963472577,
+      "epoch": 0.06322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9949078599177593e-05,
-      "loss": 1.2577,
+      "learning_rate": 1.9942239890419097e-05,
+      "loss": 1.0009,
       "step": 2228
     },
     {
-      "epoch": 0.061217763862569004,
+      "epoch": 0.06325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948988906706637e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.9942141208357185e-05,
+      "loss": 1.0313,
       "step": 2229
     },
     {
-      "epoch": 0.06124522809041224,
+      "epoch": 0.06328036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948899135515428e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.994204244231356e-05,
+      "loss": 1.0413,
       "step": 2230
     },
     {
-      "epoch": 0.061272692318255474,
+      "epoch": 0.06330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948809285604676e-05,
-      "loss": 1.0949,
+      "learning_rate": 1.994194359228906e-05,
+      "loss": 1.0188,
       "step": 2231
     },
     {
-      "epoch": 0.06130015654609871,
+      "epoch": 0.06333711691259931,
       "grad_norm": 0.0,
-      "learning_rate": 1.994871935697509e-05,
-      "loss": 1.1158,
+      "learning_rate": 1.9941844658284513e-05,
+      "loss": 1.0146,
       "step": 2232
     },
     {
-      "epoch": 0.06132762077394194,
+      "epoch": 0.06336549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 1.994862934962738e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.9941745640300756e-05,
+      "loss": 1.1964,
       "step": 2233
     },
     {
-      "epoch": 0.06135508500178517,
+      "epoch": 0.0633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948539263562264e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.994164653833863e-05,
+      "loss": 0.8035,
       "step": 2234
     },
     {
-      "epoch": 0.06138254922962841,
+      "epoch": 0.06342224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 1.994844909878045e-05,
-      "loss": 1.0556,
+      "learning_rate": 1.994154735239896e-05,
+      "loss": 1.0314,
       "step": 2235
     },
     {
-      "epoch": 0.06141001345747164,
+      "epoch": 0.06345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948358855282654e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.99414480824826e-05,
+      "loss": 1.0891,
       "step": 2236
     },
     {
-      "epoch": 0.06143747768531488,
+      "epoch": 0.06347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948268533069586e-05,
-      "loss": 1.1239,
+      "learning_rate": 1.9941348728590385e-05,
+      "loss": 0.9944,
       "step": 2237
     },
     {
-      "epoch": 0.061464941913158114,
+      "epoch": 0.06350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948178132141968e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.9941249290723143e-05,
+      "loss": 1.1032,
       "step": 2238
     },
     {
-      "epoch": 0.06149240614100135,
+      "epoch": 0.06353575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 1.9948087652500512e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.9941149768881725e-05,
+      "loss": 1.0668,
       "step": 2239
     },
     {
-      "epoch": 0.06151987036884458,
+      "epoch": 0.06356413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.994799709414593e-05,
-      "loss": 1.1675,
+      "learning_rate": 1.9941050163066962e-05,
+      "loss": 0.9964,
       "step": 2240
     },
     {
-      "epoch": 0.06154733459668781,
+      "epoch": 0.06359250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947906457078944e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.994095047327971e-05,
+      "loss": 1.0717,
       "step": 2241
     },
     {
-      "epoch": 0.06157479882453105,
+      "epoch": 0.06362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947815741300264e-05,
-      "loss": 1.0801,
+      "learning_rate": 1.9940850699520798e-05,
+      "loss": 0.9899,
       "step": 2242
     },
     {
-      "epoch": 0.06160226305237428,
+      "epoch": 0.06364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947724946810617e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.9940750841791074e-05,
+      "loss": 1.0942,
       "step": 2243
     },
     {
-      "epoch": 0.06162972728021752,
+      "epoch": 0.06367763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947634073610715e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.994065090009138e-05,
+      "loss": 0.9886,
       "step": 2244
     },
     {
-      "epoch": 0.06165719150806075,
+      "epoch": 0.06370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.994754312170128e-05,
-      "loss": 1.0658,
+      "learning_rate": 1.994055087442256e-05,
+      "loss": 1.0993,
       "step": 2245
     },
     {
-      "epoch": 0.06168465573590399,
+      "epoch": 0.0637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.994745209108303e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.9940450764785465e-05,
+      "loss": 1.1165,
       "step": 2246
     },
     {
-      "epoch": 0.06171211996374722,
+      "epoch": 0.06376276958002271,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947360981756684e-05,
-      "loss": 1.1582,
+      "learning_rate": 1.9940350571180933e-05,
+      "loss": 1.0253,
       "step": 2247
     },
     {
-      "epoch": 0.06173958419159045,
+      "epoch": 0.0637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947269793722967e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.994025029360981e-05,
+      "loss": 0.9656,
       "step": 2248
     },
     {
-      "epoch": 0.06176704841943369,
+      "epoch": 0.06381952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947178526982596e-05,
-      "loss": 1.0559,
+      "learning_rate": 1.9940149932072947e-05,
+      "loss": 1.0098,
       "step": 2249
     },
     {
-      "epoch": 0.06179451264727692,
+      "epoch": 0.06384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.9947087181536296e-05,
-      "loss": 1.1351,
+      "learning_rate": 1.9940049486571195e-05,
+      "loss": 1.0304,
       "step": 2250
     },
     {
-      "epoch": 0.06182197687512016,
+      "epoch": 0.06387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.994699575738479e-05,
-      "loss": 1.1606,
+      "learning_rate": 1.9939948957105394e-05,
+      "loss": 1.0224,
       "step": 2251
     },
     {
-      "epoch": 0.06184944110296339,
+      "epoch": 0.06390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.99469042545288e-05,
-      "loss": 1.1398,
+      "learning_rate": 1.9939848343676398e-05,
+      "loss": 1.1219,
       "step": 2252
     },
     {
-      "epoch": 0.06187690533080662,
+      "epoch": 0.06393303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946812672969052e-05,
-      "loss": 1.0501,
+      "learning_rate": 1.9939747646285058e-05,
+      "loss": 1.0747,
       "step": 2253
     },
     {
-      "epoch": 0.061904369558649856,
+      "epoch": 0.06396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946721012706264e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.9939646864932218e-05,
+      "loss": 0.9686,
       "step": 2254
     },
     {
-      "epoch": 0.06193183378649309,
+      "epoch": 0.06398978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946629273741173e-05,
-      "loss": 1.1141,
+      "learning_rate": 1.9939545999618737e-05,
+      "loss": 0.9386,
       "step": 2255
     },
     {
-      "epoch": 0.06195929801433633,
+      "epoch": 0.06401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946537456074493e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.9939445050345464e-05,
+      "loss": 1.0728,
       "step": 2256
     },
     {
-      "epoch": 0.06198676224217956,
+      "epoch": 0.06404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.994644555970696e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.9939344017113255e-05,
+      "loss": 1.1245,
       "step": 2257
     },
     {
-      "epoch": 0.0620142264700228,
+      "epoch": 0.06407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.994635358463929e-05,
-      "loss": 1.0086,
+      "learning_rate": 1.9939242899922954e-05,
+      "loss": 1.0152,
       "step": 2258
     },
     {
-      "epoch": 0.06204169069786603,
+      "epoch": 0.06410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946261530872224e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.9939141698775424e-05,
+      "loss": 1.0815,
       "step": 2259
     },
     {
-      "epoch": 0.06206915492570926,
+      "epoch": 0.0641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946169398406484e-05,
-      "loss": 1.1381,
+      "learning_rate": 1.993904041367152e-05,
+      "loss": 0.9763,
       "step": 2260
     },
     {
-      "epoch": 0.062096619153552496,
+      "epoch": 0.06416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.9946077187242795e-05,
-      "loss": 1.0488,
+      "learning_rate": 1.9938939044612097e-05,
+      "loss": 1.1401,
       "step": 2261
     },
     {
-      "epoch": 0.06212408338139573,
+      "epoch": 0.06418842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945984897381895e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.9938837591598003e-05,
+      "loss": 0.9922,
       "step": 2262
     },
     {
-      "epoch": 0.062151547609238966,
+      "epoch": 0.06421679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945892528824505e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.9938736054630103e-05,
+      "loss": 1.0805,
       "step": 2263
     },
     {
-      "epoch": 0.0621790118370822,
+      "epoch": 0.06424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945800081571365e-05,
-      "loss": 1.1377,
+      "learning_rate": 1.9938634433709254e-05,
+      "loss": 1.0129,
       "step": 2264
     },
     {
-      "epoch": 0.06220647606492544,
+      "epoch": 0.06427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.99457075556232e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.993853272883631e-05,
+      "loss": 0.929,
       "step": 2265
     },
     {
-      "epoch": 0.06223394029276867,
+      "epoch": 0.06430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945614950980745e-05,
-      "loss": 1.0127,
+      "learning_rate": 1.9938430940012138e-05,
+      "loss": 1.0608,
       "step": 2266
     },
     {
-      "epoch": 0.0622614045206119,
+      "epoch": 0.06433030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945522267644733e-05,
-      "loss": 1.1635,
+      "learning_rate": 1.993832906723759e-05,
+      "loss": 1.0079,
       "step": 2267
     },
     {
-      "epoch": 0.062288868748455135,
+      "epoch": 0.06435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945429505615894e-05,
-      "loss": 1.0783,
+      "learning_rate": 1.993822711051353e-05,
+      "loss": 1.0424,
       "step": 2268
     },
     {
-      "epoch": 0.06231633297629837,
+      "epoch": 0.06438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945336664894963e-05,
-      "loss": 1.2086,
+      "learning_rate": 1.9938125069840817e-05,
+      "loss": 1.0535,
       "step": 2269
     },
     {
-      "epoch": 0.062343797204141606,
+      "epoch": 0.06441543700340523,
       "grad_norm": 0.0,
-      "learning_rate": 1.994524374548268e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.993802294522032e-05,
+      "loss": 1.1017,
       "step": 2270
     },
     {
-      "epoch": 0.06237126143198484,
+      "epoch": 0.06444381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945150747379775e-05,
-      "loss": 1.1136,
+      "learning_rate": 1.993792073665289e-05,
+      "loss": 1.1283,
       "step": 2271
     },
     {
-      "epoch": 0.062398725659828076,
+      "epoch": 0.064472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.994505767058698e-05,
-      "loss": 1.2339,
+      "learning_rate": 1.99378184441394e-05,
+      "loss": 0.9802,
       "step": 2272
     },
     {
-      "epoch": 0.06242618988767131,
+      "epoch": 0.0645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944964515105043e-05,
-      "loss": 1.0697,
+      "learning_rate": 1.9937716067680712e-05,
+      "loss": 1.149,
       "step": 2273
     },
     {
-      "epoch": 0.06245365411551454,
+      "epoch": 0.06452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944871280934693e-05,
-      "loss": 1.1164,
+      "learning_rate": 1.993761360727769e-05,
+      "loss": 1.1481,
       "step": 2274
     },
     {
-      "epoch": 0.062481118343357775,
+      "epoch": 0.06455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944777968076666e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.9937511062931197e-05,
+      "loss": 1.2144,
       "step": 2275
     },
     {
-      "epoch": 0.06250858257120101,
+      "epoch": 0.06458569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944684576531707e-05,
-      "loss": 1.135,
+      "learning_rate": 1.99374084346421e-05,
+      "loss": 1.0313,
       "step": 2276
     },
     {
-      "epoch": 0.06253604679904425,
+      "epoch": 0.06461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.994459110630055e-05,
-      "loss": 1.102,
+      "learning_rate": 1.993730572241127e-05,
+      "loss": 1.0728,
       "step": 2277
     },
     {
-      "epoch": 0.06256351102688748,
+      "epoch": 0.06464245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944497557383933e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.993720292623957e-05,
+      "loss": 1.1033,
       "step": 2278
     },
     {
-      "epoch": 0.06259097525473072,
+      "epoch": 0.06467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944403929782604e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.9937100046127872e-05,
+      "loss": 0.957,
       "step": 2279
     },
     {
-      "epoch": 0.06261843948257395,
+      "epoch": 0.06469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944310223497296e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.9936997082077043e-05,
+      "loss": 1.1226,
       "step": 2280
     },
     {
-      "epoch": 0.06264590371041719,
+      "epoch": 0.06472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944216438528755e-05,
-      "loss": 1.1446,
+      "learning_rate": 1.9936894034087952e-05,
+      "loss": 1.1047,
       "step": 2281
     },
     {
-      "epoch": 0.06267336793826042,
+      "epoch": 0.06475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.994412257487772e-05,
-      "loss": 1.1321,
+      "learning_rate": 1.993679090216147e-05,
+      "loss": 0.9836,
       "step": 2282
     },
     {
-      "epoch": 0.06270083216610366,
+      "epoch": 0.06478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.9944028632544937e-05,
-      "loss": 1.1515,
+      "learning_rate": 1.993668768629847e-05,
+      "loss": 1.1187,
       "step": 2283
     },
     {
-      "epoch": 0.06272829639394688,
+      "epoch": 0.06481271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943934611531147e-05,
-      "loss": 1.1316,
+      "learning_rate": 1.9936584386499824e-05,
+      "loss": 1.076,
       "step": 2284
     },
     {
-      "epoch": 0.06275576062179011,
+      "epoch": 0.0648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.994384051183709e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.99364810027664e-05,
+      "loss": 1.0547,
       "step": 2285
     },
     {
-      "epoch": 0.06278322484963335,
+      "epoch": 0.06486946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 1.994374633346352e-05,
-      "loss": 1.1636,
+      "learning_rate": 1.9936377535099073e-05,
+      "loss": 1.1398,
       "step": 2286
     },
     {
-      "epoch": 0.06281068907747658,
+      "epoch": 0.0648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.994365207641118e-05,
-      "loss": 1.2123,
+      "learning_rate": 1.993627398349872e-05,
+      "loss": 1.0319,
       "step": 2287
     },
     {
-      "epoch": 0.06283815330531982,
+      "epoch": 0.06492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943557740680807e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.9936170347966215e-05,
+      "loss": 1.0098,
       "step": 2288
     },
     {
-      "epoch": 0.06286561753316305,
+      "epoch": 0.06495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943463326273156e-05,
-      "loss": 1.1424,
+      "learning_rate": 1.9936066628502432e-05,
+      "loss": 1.0332,
       "step": 2289
     },
     {
-      "epoch": 0.06289308176100629,
+      "epoch": 0.06498297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943368833188968e-05,
-      "loss": 1.0976,
+      "learning_rate": 1.9935962825108248e-05,
+      "loss": 1.058,
       "step": 2290
     },
     {
-      "epoch": 0.06292054598884952,
+      "epoch": 0.06501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943274261429e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.9935858937784537e-05,
+      "loss": 1.0574,
       "step": 2291
     },
     {
-      "epoch": 0.06294801021669276,
+      "epoch": 0.06503972758229284,
       "grad_norm": 0.0,
-      "learning_rate": 1.994317961099399e-05,
-      "loss": 1.0625,
+      "learning_rate": 1.993575496653218e-05,
+      "loss": 1.0221,
       "step": 2292
     },
     {
-      "epoch": 0.062975474444536,
+      "epoch": 0.06506810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943084881884696e-05,
-      "loss": 1.1772,
+      "learning_rate": 1.9935650911352055e-05,
+      "loss": 1.0027,
       "step": 2293
     },
     {
-      "epoch": 0.06300293867237923,
+      "epoch": 0.06509648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.994299007410186e-05,
-      "loss": 1.1043,
+      "learning_rate": 1.9935546772245042e-05,
+      "loss": 1.1157,
       "step": 2294
     },
     {
-      "epoch": 0.06303040290022247,
+      "epoch": 0.06512485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942895187646232e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.9935442549212016e-05,
+      "loss": 1.0892,
       "step": 2295
     },
     {
-      "epoch": 0.0630578671280657,
+      "epoch": 0.06515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.994280022251857e-05,
-      "loss": 1.1748,
+      "learning_rate": 1.993533824225386e-05,
+      "loss": 1.0455,
       "step": 2296
     },
     {
-      "epoch": 0.06308533135590892,
+      "epoch": 0.06518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.994270517871962e-05,
-      "loss": 1.006,
+      "learning_rate": 1.9935233851371456e-05,
+      "loss": 1.0944,
       "step": 2297
     },
     {
-      "epoch": 0.06311279558375216,
+      "epoch": 0.0652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942610056250135e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9935129376565686e-05,
+      "loss": 1.0828,
       "step": 2298
     },
     {
-      "epoch": 0.06314025981159539,
+      "epoch": 0.0652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942514855110866e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.9935024817837433e-05,
+      "loss": 1.1168,
       "step": 2299
     },
     {
-      "epoch": 0.06316772403943863,
+      "epoch": 0.06526674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942419575302568e-05,
-      "loss": 1.0933,
+      "learning_rate": 1.993492017518757e-05,
+      "loss": 1.05,
       "step": 2300
     },
     {
-      "epoch": 0.06319518826728186,
+      "epoch": 0.06529511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942324216826e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.9934815448617e-05,
+      "loss": 1.1275,
       "step": 2301
     },
     {
-      "epoch": 0.0632226524951251,
+      "epoch": 0.06532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942228779681905e-05,
-      "loss": 1.1106,
+      "learning_rate": 1.9934710638126594e-05,
+      "loss": 1.0964,
       "step": 2302
     },
     {
-      "epoch": 0.06325011672296833,
+      "epoch": 0.06535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942133263871048e-05,
-      "loss": 1.1952,
+      "learning_rate": 1.993460574371724e-05,
+      "loss": 1.0863,
       "step": 2303
     },
     {
-      "epoch": 0.06327758095081157,
+      "epoch": 0.06538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9942037669394185e-05,
-      "loss": 1.0564,
+      "learning_rate": 1.9934500765389824e-05,
+      "loss": 1.0931,
       "step": 2304
     },
     {
-      "epoch": 0.0633050451786548,
+      "epoch": 0.06540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.994194199625206e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.9934395703145234e-05,
+      "loss": 1.1202,
       "step": 2305
     },
     {
-      "epoch": 0.06333250940649804,
+      "epoch": 0.06543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.9941846244445447e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.9934290556984356e-05,
+      "loss": 1.01,
       "step": 2306
     },
     {
-      "epoch": 0.06335997363434127,
+      "epoch": 0.06546538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.994175041397509e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.9934185326908082e-05,
+      "loss": 1.0324,
       "step": 2307
     },
     {
-      "epoch": 0.06338743786218451,
+      "epoch": 0.06549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.9941654504841757e-05,
-      "loss": 1.029,
+      "learning_rate": 1.9934080012917295e-05,
+      "loss": 1.1113,
       "step": 2308
     },
     {
-      "epoch": 0.06341490209002774,
+      "epoch": 0.06552213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 1.99415585170462e-05,
-      "loss": 1.0599,
+      "learning_rate": 1.9933974615012888e-05,
+      "loss": 1.0459,
       "step": 2309
     },
     {
-      "epoch": 0.06344236631787098,
+      "epoch": 0.06555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.9941462450589183e-05,
-      "loss": 1.1113,
+      "learning_rate": 1.9933869133195752e-05,
+      "loss": 1.0397,
       "step": 2310
     },
     {
-      "epoch": 0.0634698305457142,
+      "epoch": 0.0655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.994136630547146e-05,
-      "loss": 1.1484,
+      "learning_rate": 1.9933763567466776e-05,
+      "loss": 0.9309,
       "step": 2311
     },
     {
-      "epoch": 0.06349729477355744,
+      "epoch": 0.06560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.99412700816938e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.993365791782685e-05,
+      "loss": 1.063,
       "step": 2312
     },
     {
-      "epoch": 0.06352475900140067,
+      "epoch": 0.06563564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 1.994117377925696e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.9933552184276873e-05,
+      "loss": 0.9634,
       "step": 2313
     },
     {
-      "epoch": 0.0635522232292439,
+      "epoch": 0.06566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.99410773981617e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.9933446366817735e-05,
+      "loss": 1.1072,
       "step": 2314
     },
     {
-      "epoch": 0.06357968745708714,
+      "epoch": 0.06569239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940980938408784e-05,
-      "loss": 1.079,
+      "learning_rate": 1.9933340465450328e-05,
+      "loss": 1.0513,
       "step": 2315
     },
     {
-      "epoch": 0.06360715168493038,
+      "epoch": 0.06572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.994088439999898e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.9933234480175545e-05,
+      "loss": 1.0477,
       "step": 2316
     },
     {
-      "epoch": 0.06363461591277361,
+      "epoch": 0.06574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940787782933043e-05,
-      "loss": 1.1654,
+      "learning_rate": 1.9933128410994288e-05,
+      "loss": 1.0329,
       "step": 2317
     },
     {
-      "epoch": 0.06366208014061685,
+      "epoch": 0.06577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940691087211745e-05,
-      "loss": 1.126,
+      "learning_rate": 1.9933022257907445e-05,
+      "loss": 1.062,
       "step": 2318
     },
     {
-      "epoch": 0.06368954436846008,
+      "epoch": 0.06580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.994059431283585e-05,
-      "loss": 1.028,
+      "learning_rate": 1.9932916020915917e-05,
+      "loss": 1.0421,
       "step": 2319
     },
     {
-      "epoch": 0.06371700859630332,
+      "epoch": 0.06583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.994049745980612e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.9932809700020602e-05,
+      "loss": 1.0038,
       "step": 2320
     },
     {
-      "epoch": 0.06374447282414655,
+      "epoch": 0.06586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940400528123325e-05,
-      "loss": 0.9574,
+      "learning_rate": 1.9932703295222398e-05,
+      "loss": 1.0353,
       "step": 2321
     },
     {
-      "epoch": 0.06377193705198979,
+      "epoch": 0.06589103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 1.994030351778823e-05,
-      "loss": 1.0622,
+      "learning_rate": 1.99325968065222e-05,
+      "loss": 1.081,
       "step": 2322
     },
     {
-      "epoch": 0.06379940127983302,
+      "epoch": 0.0659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940206428801602e-05,
-      "loss": 1.1244,
+      "learning_rate": 1.993249023392091e-05,
+      "loss": 0.9538,
       "step": 2323
     },
     {
-      "epoch": 0.06382686550767624,
+      "epoch": 0.0659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940109261164212e-05,
-      "loss": 1.1912,
+      "learning_rate": 1.9932383577419432e-05,
+      "loss": 1.1124,
       "step": 2324
     },
     {
-      "epoch": 0.06385432973551948,
+      "epoch": 0.06597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.9940012014876828e-05,
-      "loss": 1.002,
+      "learning_rate": 1.993227683701866e-05,
+      "loss": 1.1124,
       "step": 2325
     },
     {
-      "epoch": 0.06388179396336272,
+      "epoch": 0.06600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939914689940215e-05,
-      "loss": 0.9616,
+      "learning_rate": 1.9932170012719504e-05,
+      "loss": 1.0739,
       "step": 2326
     },
     {
-      "epoch": 0.06390925819120595,
+      "epoch": 0.06603291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939817286355146e-05,
-      "loss": 1.1567,
+      "learning_rate": 1.9932063104522857e-05,
+      "loss": 1.0601,
       "step": 2327
     },
     {
-      "epoch": 0.06393672241904919,
+      "epoch": 0.06606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939719804122394e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.9931956112429625e-05,
+      "loss": 1.0764,
       "step": 2328
     },
     {
-      "epoch": 0.06396418664689242,
+      "epoch": 0.06608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939622243242727e-05,
-      "loss": 1.0188,
+      "learning_rate": 1.9931849036440715e-05,
+      "loss": 1.0587,
       "step": 2329
     },
     {
-      "epoch": 0.06399165087473566,
+      "epoch": 0.06611804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939524603716926e-05,
-      "loss": 1.0934,
+      "learning_rate": 1.9931741876557033e-05,
+      "loss": 1.0231,
       "step": 2330
     },
     {
-      "epoch": 0.06401911510257889,
+      "epoch": 0.06614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.993942688554575e-05,
-      "loss": 1.0827,
+      "learning_rate": 1.9931634632779477e-05,
+      "loss": 1.1528,
       "step": 2331
     },
     {
-      "epoch": 0.06404657933042213,
+      "epoch": 0.06617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.993932908872998e-05,
-      "loss": 1.163,
+      "learning_rate": 1.9931527305108956e-05,
+      "loss": 1.0458,
       "step": 2332
     },
     {
-      "epoch": 0.06407404355826536,
+      "epoch": 0.06620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939231213270386e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.993141989354638e-05,
+      "loss": 0.9999,
       "step": 2333
     },
     {
-      "epoch": 0.0641015077861086,
+      "epoch": 0.06623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939133259167745e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.9931312398092654e-05,
+      "loss": 1.1028,
       "step": 2334
     },
     {
-      "epoch": 0.06412897201395183,
+      "epoch": 0.06625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939035226422833e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.9931204818748685e-05,
+      "loss": 1.0627,
       "step": 2335
     },
     {
-      "epoch": 0.06415643624179507,
+      "epoch": 0.06628830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938937115036423e-05,
-      "loss": 1.1102,
+      "learning_rate": 1.9931097155515383e-05,
+      "loss": 1.0614,
       "step": 2336
     },
     {
-      "epoch": 0.0641839004696383,
+      "epoch": 0.066316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938838925009293e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.993098940839366e-05,
+      "loss": 0.9848,
       "step": 2337
     },
     {
-      "epoch": 0.06421136469748152,
+      "epoch": 0.06634506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 1.993874065634222e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.9930881577384417e-05,
+      "loss": 1.0674,
       "step": 2338
     },
     {
-      "epoch": 0.06423882892532476,
+      "epoch": 0.06637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.993864230903598e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.9930773662488573e-05,
+      "loss": 1.0789,
       "step": 2339
     },
     {
-      "epoch": 0.064266293153168,
+      "epoch": 0.06640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938543883091353e-05,
-      "loss": 1.081,
+      "learning_rate": 1.9930665663707035e-05,
+      "loss": 0.9802,
       "step": 2340
     },
     {
-      "epoch": 0.06429375738101123,
+      "epoch": 0.06643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.993844537850912e-05,
-      "loss": 1.1403,
+      "learning_rate": 1.993055758104072e-05,
+      "loss": 1.1653,
       "step": 2341
     },
     {
-      "epoch": 0.06432122160885446,
+      "epoch": 0.06645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938346795290054e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.9930449414490544e-05,
+      "loss": 1.1218,
       "step": 2342
     },
     {
-      "epoch": 0.0643486858366977,
+      "epoch": 0.06648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938248133434936e-05,
-      "loss": 1.1189,
+      "learning_rate": 1.993034116405741e-05,
+      "loss": 1.1315,
       "step": 2343
     },
     {
-      "epoch": 0.06437615006454094,
+      "epoch": 0.06651532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938149392944553e-05,
-      "loss": 1.0372,
+      "learning_rate": 1.993023282974224e-05,
+      "loss": 1.0801,
       "step": 2344
     },
     {
-      "epoch": 0.06440361429238417,
+      "epoch": 0.06654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.9938050573819683e-05,
-      "loss": 1.021,
+      "learning_rate": 1.9930124411545943e-05,
+      "loss": 1.0185,
       "step": 2345
     },
     {
-      "epoch": 0.0644310785202274,
+      "epoch": 0.06657207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937951676061106e-05,
-      "loss": 1.1242,
+      "learning_rate": 1.993001590946944e-05,
+      "loss": 1.0932,
       "step": 2346
     },
     {
-      "epoch": 0.06445854274807064,
+      "epoch": 0.06660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937852699669604e-05,
-      "loss": 1.0678,
+      "learning_rate": 1.992990732351365e-05,
+      "loss": 1.0713,
       "step": 2347
     },
     {
-      "epoch": 0.06448600697591388,
+      "epoch": 0.06662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937753644645963e-05,
-      "loss": 0.9059,
+      "learning_rate": 1.992979865367948e-05,
+      "loss": 1.1374,
       "step": 2348
     },
     {
-      "epoch": 0.06451347120375711,
+      "epoch": 0.0666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937654510990967e-05,
-      "loss": 1.1128,
+      "learning_rate": 1.9929689899967858e-05,
+      "loss": 1.0532,
       "step": 2349
     },
     {
-      "epoch": 0.06454093543160035,
+      "epoch": 0.0666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937555298705397e-05,
-      "loss": 1.1426,
+      "learning_rate": 1.99295810623797e-05,
+      "loss": 1.0518,
       "step": 2350
     },
     {
-      "epoch": 0.06456839965944358,
+      "epoch": 0.06671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.993745600779004e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.992947214091592e-05,
+      "loss": 0.9565,
       "step": 2351
     },
     {
-      "epoch": 0.0645958638872868,
+      "epoch": 0.06674233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.993735663824568e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.992936313557745e-05,
+      "loss": 1.0167,
       "step": 2352
     },
     {
-      "epoch": 0.06462332811513004,
+      "epoch": 0.06677071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937257190073108e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.9929254046365198e-05,
+      "loss": 1.1486,
       "step": 2353
     },
     {
-      "epoch": 0.06465079234297327,
+      "epoch": 0.06679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.993715766327311e-05,
-      "loss": 1.1843,
+      "learning_rate": 1.9929144873280092e-05,
+      "loss": 0.9929,
       "step": 2354
     },
     {
-      "epoch": 0.06467825657081651,
+      "epoch": 0.06682746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9937058057846466e-05,
-      "loss": 1.1218,
+      "learning_rate": 1.9929035616323054e-05,
+      "loss": 1.0189,
       "step": 2355
     },
     {
-      "epoch": 0.06470572079865974,
+      "epoch": 0.06685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936958373793973e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.9928926275495006e-05,
+      "loss": 1.1032,
       "step": 2356
     },
     {
-      "epoch": 0.06473318502650298,
+      "epoch": 0.06688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936858611116413e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.9928816850796865e-05,
+      "loss": 1.0807,
       "step": 2357
     },
     {
-      "epoch": 0.06476064925434621,
+      "epoch": 0.06691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936758769814585e-05,
-      "loss": 1.1058,
+      "learning_rate": 1.9928707342229568e-05,
+      "loss": 1.0085,
       "step": 2358
     },
     {
-      "epoch": 0.06478811348218945,
+      "epoch": 0.06694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936658849889264e-05,
-      "loss": 1.1648,
+      "learning_rate": 1.9928597749794035e-05,
+      "loss": 1.1438,
       "step": 2359
     },
     {
-      "epoch": 0.06481557771003268,
+      "epoch": 0.06696935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936558851341255e-05,
-      "loss": 1.157,
+      "learning_rate": 1.992848807349119e-05,
+      "loss": 1.1235,
       "step": 2360
     },
     {
-      "epoch": 0.06484304193787592,
+      "epoch": 0.0669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936458774171342e-05,
-      "loss": 1.1735,
+      "learning_rate": 1.9928378313321953e-05,
+      "loss": 1.0916,
       "step": 2361
     },
     {
-      "epoch": 0.06487050616571916,
+      "epoch": 0.0670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936358618380316e-05,
-      "loss": 1.1368,
+      "learning_rate": 1.9928268469287264e-05,
+      "loss": 0.962,
       "step": 2362
     },
     {
-      "epoch": 0.06489797039356239,
+      "epoch": 0.06705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936258383968972e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.9928158541388046e-05,
+      "loss": 1.0493,
       "step": 2363
     },
     {
-      "epoch": 0.06492543462140563,
+      "epoch": 0.06708286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936158070938104e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.992804852962522e-05,
+      "loss": 1.0789,
       "step": 2364
     },
     {
-      "epoch": 0.06495289884924885,
+      "epoch": 0.06711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.9936057679288504e-05,
-      "loss": 1.1659,
+      "learning_rate": 1.992793843399973e-05,
+      "loss": 1.1963,
       "step": 2365
     },
     {
-      "epoch": 0.06498036307709208,
+      "epoch": 0.06713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935957209020968e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.9927828254512495e-05,
+      "loss": 0.9551,
       "step": 2366
     },
     {
-      "epoch": 0.06500782730493532,
+      "epoch": 0.06716799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 1.993585666013629e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.9927717991164445e-05,
+      "loss": 1.0491,
       "step": 2367
     },
     {
-      "epoch": 0.06503529153277855,
+      "epoch": 0.06719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935756032635264e-05,
-      "loss": 1.0899,
+      "learning_rate": 1.9927607643956518e-05,
+      "loss": 1.0743,
       "step": 2368
     },
     {
-      "epoch": 0.06506275576062179,
+      "epoch": 0.06722474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.993565532651869e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.9927497212889643e-05,
+      "loss": 1.092,
       "step": 2369
     },
     {
-      "epoch": 0.06509021998846502,
+      "epoch": 0.06725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935554541787356e-05,
-      "loss": 1.0955,
+      "learning_rate": 1.992738669796475e-05,
+      "loss": 1.0544,
       "step": 2370
     },
     {
-      "epoch": 0.06511768421630826,
+      "epoch": 0.06728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935453678442068e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.9927276099182777e-05,
+      "loss": 1.0327,
       "step": 2371
     },
     {
-      "epoch": 0.0651451484441515,
+      "epoch": 0.06730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935352736483626e-05,
-      "loss": 1.2277,
+      "learning_rate": 1.9927165416544656e-05,
+      "loss": 1.0555,
       "step": 2372
     },
     {
-      "epoch": 0.06517261267199473,
+      "epoch": 0.06733825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.993525171591282e-05,
-      "loss": 1.0802,
+      "learning_rate": 1.9927054650051324e-05,
+      "loss": 1.0949,
       "step": 2373
     },
     {
-      "epoch": 0.06520007689983796,
+      "epoch": 0.067366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935150616730457e-05,
-      "loss": 1.2573,
+      "learning_rate": 1.9926943799703716e-05,
+      "loss": 1.1088,
       "step": 2374
     },
     {
-      "epoch": 0.0652275411276812,
+      "epoch": 0.0673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9935049438937334e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.9926832865502766e-05,
+      "loss": 1.0378,
       "step": 2375
     },
     {
-      "epoch": 0.06525500535552443,
+      "epoch": 0.0674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.993494818253425e-05,
-      "loss": 1.213,
+      "learning_rate": 1.992672184744941e-05,
+      "loss": 1.0398,
       "step": 2376
     },
     {
-      "epoch": 0.06528246958336767,
+      "epoch": 0.06745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.993484684752201e-05,
-      "loss": 1.1002,
+      "learning_rate": 1.992661074554459e-05,
+      "loss": 1.0589,
       "step": 2377
     },
     {
-      "epoch": 0.0653099338112109,
+      "epoch": 0.06748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.993474543390141e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.9926499559789245e-05,
+      "loss": 1.1829,
       "step": 2378
     },
     {
-      "epoch": 0.06533739803905413,
+      "epoch": 0.06750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9934643941673258e-05,
-      "loss": 1.0889,
+      "learning_rate": 1.992638829018431e-05,
+      "loss": 0.9743,
       "step": 2379
     },
     {
-      "epoch": 0.06536486226689736,
+      "epoch": 0.06753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.993454237083835e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.9926276936730728e-05,
+      "loss": 0.9507,
       "step": 2380
     },
     {
-      "epoch": 0.0653923264947406,
+      "epoch": 0.06756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.99344407213975e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.9926165499429437e-05,
+      "loss": 1.0154,
       "step": 2381
     },
     {
-      "epoch": 0.06541979072258383,
+      "epoch": 0.06759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.9934338993351506e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.9926053978281384e-05,
+      "loss": 0.99,
       "step": 2382
     },
     {
-      "epoch": 0.06544725495042707,
+      "epoch": 0.06762202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.9934237186701174e-05,
-      "loss": 1.1797,
+      "learning_rate": 1.99259423732875e-05,
+      "loss": 1.0809,
       "step": 2383
     },
     {
-      "epoch": 0.0654747191782703,
+      "epoch": 0.06765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9934135301447308e-05,
-      "loss": 1.1411,
+      "learning_rate": 1.992583068444874e-05,
+      "loss": 1.042,
       "step": 2384
     },
     {
-      "epoch": 0.06550218340611354,
+      "epoch": 0.06767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.9934033337590713e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.992571891176604e-05,
+      "loss": 1.0907,
       "step": 2385
     },
     {
-      "epoch": 0.06552964763395677,
+      "epoch": 0.06770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933931295132202e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.9925607055240345e-05,
+      "loss": 0.9768,
       "step": 2386
     },
     {
-      "epoch": 0.06555711186180001,
+      "epoch": 0.06773552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933829174072577e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.9925495114872603e-05,
+      "loss": 1.0216,
       "step": 2387
     },
     {
-      "epoch": 0.06558457608964324,
+      "epoch": 0.0677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933726974412648e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.9925383090663758e-05,
+      "loss": 1.0032,
       "step": 2388
     },
     {
-      "epoch": 0.06561204031748648,
+      "epoch": 0.06779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933624696153222e-05,
-      "loss": 1.0832,
+      "learning_rate": 1.9925270982614755e-05,
+      "loss": 1.0734,
       "step": 2389
     },
     {
-      "epoch": 0.06563950454532971,
+      "epoch": 0.06782065834279229,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933522339295112e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.992515879072654e-05,
+      "loss": 0.986,
       "step": 2390
     },
     {
-      "epoch": 0.06566696877317295,
+      "epoch": 0.06784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.993341990383912e-05,
-      "loss": 1.1531,
+      "learning_rate": 1.9925046515000063e-05,
+      "loss": 1.0822,
       "step": 2391
     },
     {
-      "epoch": 0.06569443300101617,
+      "epoch": 0.06787741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933317389786067e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.9924934155436274e-05,
+      "loss": 1.1248,
       "step": 2392
     },
     {
-      "epoch": 0.0657218972288594,
+      "epoch": 0.06790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.993321479713676e-05,
-      "loss": 1.2432,
+      "learning_rate": 1.9924821712036116e-05,
+      "loss": 1.0187,
       "step": 2393
     },
     {
-      "epoch": 0.06574936145670264,
+      "epoch": 0.06793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.9933112125892003e-05,
-      "loss": 1.168,
+      "learning_rate": 1.9924709184800546e-05,
+      "loss": 1.1668,
       "step": 2394
     },
     {
-      "epoch": 0.06577682568454588,
+      "epoch": 0.06796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.993300937605262e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.9924596573730514e-05,
+      "loss": 1.0547,
       "step": 2395
     },
     {
-      "epoch": 0.06580428991238911,
+      "epoch": 0.06799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.993290654761942e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.9924483878826965e-05,
+      "loss": 1.017,
       "step": 2396
     },
     {
-      "epoch": 0.06583175414023235,
+      "epoch": 0.06801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932803640593215e-05,
-      "loss": 1.1154,
+      "learning_rate": 1.9924371100090857e-05,
+      "loss": 1.0293,
       "step": 2397
     },
     {
-      "epoch": 0.06585921836807558,
+      "epoch": 0.06804767309875141,
       "grad_norm": 0.0,
-      "learning_rate": 1.993270065497482e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.992425823752314e-05,
+      "loss": 1.1043,
       "step": 2398
     },
     {
-      "epoch": 0.06588668259591882,
+      "epoch": 0.06807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.993259759076505e-05,
-      "loss": 0.989,
+      "learning_rate": 1.9924145291124764e-05,
+      "loss": 0.9837,
       "step": 2399
     },
     {
-      "epoch": 0.06591414682376205,
+      "epoch": 0.0681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932494447964717e-05,
-      "loss": 1.1305,
+      "learning_rate": 1.992403226089669e-05,
+      "loss": 1.0342,
       "step": 2400
     },
     {
-      "epoch": 0.06594161105160529,
+      "epoch": 0.06813280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932391226574643e-05,
-      "loss": 1.1019,
+      "learning_rate": 1.9923919146839866e-05,
+      "loss": 1.0648,
       "step": 2401
     },
     {
-      "epoch": 0.06596907527944852,
+      "epoch": 0.06816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932287926595643e-05,
-      "loss": 1.0761,
+      "learning_rate": 1.9923805948955254e-05,
+      "loss": 1.025,
       "step": 2402
     },
     {
-      "epoch": 0.06599653950729176,
+      "epoch": 0.06818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932184548028536e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.9923692667243806e-05,
+      "loss": 1.1037,
       "step": 2403
     },
     {
-      "epoch": 0.06602400373513499,
+      "epoch": 0.06821793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.9932081090874133e-05,
-      "loss": 1.138,
+      "learning_rate": 1.992357930170648e-05,
+      "loss": 1.1838,
       "step": 2404
     },
     {
-      "epoch": 0.06605146796297823,
+      "epoch": 0.06824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931977555133256e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.992346585234423e-05,
+      "loss": 1.0368,
       "step": 2405
     },
     {
-      "epoch": 0.06607893219082145,
+      "epoch": 0.06827468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.993187394080673e-05,
-      "loss": 1.0493,
+      "learning_rate": 1.9923352319158025e-05,
+      "loss": 1.1508,
       "step": 2406
     },
     {
-      "epoch": 0.06610639641866468,
+      "epoch": 0.06830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931770247895365e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9923238702148812e-05,
+      "loss": 0.9807,
       "step": 2407
     },
     {
-      "epoch": 0.06613386064650792,
+      "epoch": 0.06833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.993166647639999e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.9923125001317557e-05,
+      "loss": 0.9876,
       "step": 2408
     },
     {
-      "epoch": 0.06616132487435115,
+      "epoch": 0.06835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931562626321423e-05,
-      "loss": 1.039,
+      "learning_rate": 1.992301121666522e-05,
+      "loss": 1.085,
       "step": 2409
     },
     {
-      "epoch": 0.06618878910219439,
+      "epoch": 0.06838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931458697660485e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.9922897348192758e-05,
+      "loss": 1.0325,
       "step": 2410
     },
     {
-      "epoch": 0.06621625333003762,
+      "epoch": 0.06841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931354690418e-05,
-      "loss": 1.0914,
+      "learning_rate": 1.9922783395901138e-05,
+      "loss": 1.0321,
       "step": 2411
     },
     {
-      "epoch": 0.06624371755788086,
+      "epoch": 0.0684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.9931250604594784e-05,
-      "loss": 1.0769,
+      "learning_rate": 1.9922669359791322e-05,
+      "loss": 1.156,
       "step": 2412
     },
     {
-      "epoch": 0.0662711817857241,
+      "epoch": 0.0684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.993114644019167e-05,
-      "loss": 1.0647,
+      "learning_rate": 1.992255523986427e-05,
+      "loss": 1.05,
       "step": 2413
     },
     {
-      "epoch": 0.06629864601356733,
+      "epoch": 0.06850170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.993104219720948e-05,
-      "loss": 1.1429,
+      "learning_rate": 1.992244103612095e-05,
+      "loss": 1.1171,
       "step": 2414
     },
     {
-      "epoch": 0.06632611024141057,
+      "epoch": 0.06853007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930937875649034e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.9922326748562323e-05,
+      "loss": 1.1129,
       "step": 2415
     },
     {
-      "epoch": 0.0663535744692538,
+      "epoch": 0.06855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930833475511163e-05,
-      "loss": 1.0407,
+      "learning_rate": 1.9922212377189355e-05,
+      "loss": 0.9969,
       "step": 2416
     },
     {
-      "epoch": 0.06638103869709704,
+      "epoch": 0.06858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.993072899679669e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.9922097922003016e-05,
+      "loss": 1.0946,
       "step": 2417
     },
     {
-      "epoch": 0.06640850292494027,
+      "epoch": 0.06861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930624439506444e-05,
-      "loss": 1.1651,
+      "learning_rate": 1.9921983383004267e-05,
+      "loss": 1.0554,
       "step": 2418
     },
     {
-      "epoch": 0.06643596715278349,
+      "epoch": 0.06864358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930519803641245e-05,
-      "loss": 1.0961,
+      "learning_rate": 1.9921868760194075e-05,
+      "loss": 1.127,
       "step": 2419
     },
     {
-      "epoch": 0.06646343138062673,
+      "epoch": 0.06867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930415089201933e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.9921754053573418e-05,
+      "loss": 1.051,
       "step": 2420
     },
     {
-      "epoch": 0.06649089560846996,
+      "epoch": 0.06870034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930310296189328e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.9921639263143256e-05,
+      "loss": 1.0953,
       "step": 2421
     },
     {
-      "epoch": 0.0665183598363132,
+      "epoch": 0.06872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930205424604262e-05,
-      "loss": 1.0921,
+      "learning_rate": 1.992152438890456e-05,
+      "loss": 1.0477,
       "step": 2422
     },
     {
-      "epoch": 0.06654582406415643,
+      "epoch": 0.06875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.9930100474447563e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.9921409430858304e-05,
+      "loss": 1.0334,
       "step": 2423
     },
     {
-      "epoch": 0.06657328829199967,
+      "epoch": 0.06878547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.992999544572006e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.9921294389005453e-05,
+      "loss": 1.0635,
       "step": 2424
     },
     {
-      "epoch": 0.0666007525198429,
+      "epoch": 0.06881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.992989033842259e-05,
-      "loss": 1.0639,
+      "learning_rate": 1.9921179263346986e-05,
+      "loss": 0.9706,
       "step": 2425
     },
     {
-      "epoch": 0.06662821674768614,
+      "epoch": 0.0688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929785152555983e-05,
-      "loss": 1.1142,
+      "learning_rate": 1.992106405388387e-05,
+      "loss": 1.2063,
       "step": 2426
     },
     {
-      "epoch": 0.06665568097552937,
+      "epoch": 0.0688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929679888121065e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.992094876061708e-05,
+      "loss": 1.0378,
       "step": 2427
     },
     {
-      "epoch": 0.06668314520337261,
+      "epoch": 0.06889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929574545118674e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.9920833383547592e-05,
+      "loss": 1.084,
       "step": 2428
     },
     {
-      "epoch": 0.06671060943121584,
+      "epoch": 0.06892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929469123549647e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9920717922676375e-05,
+      "loss": 1.0638,
       "step": 2429
     },
     {
-      "epoch": 0.06673807365905908,
+      "epoch": 0.06895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929363623414807e-05,
-      "loss": 1.1111,
+      "learning_rate": 1.9920602378004407e-05,
+      "loss": 1.0687,
       "step": 2430
     },
     {
-      "epoch": 0.06676553788690232,
+      "epoch": 0.06898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929258044715002e-05,
-      "loss": 1.1847,
+      "learning_rate": 1.9920486749532667e-05,
+      "loss": 1.0461,
       "step": 2431
     },
     {
-      "epoch": 0.06679300211474555,
+      "epoch": 0.06901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929152387451062e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.9920371037262132e-05,
+      "loss": 1.0089,
       "step": 2432
     },
     {
-      "epoch": 0.06682046634258877,
+      "epoch": 0.06904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929046651623816e-05,
-      "loss": 1.1072,
+      "learning_rate": 1.9920255241193773e-05,
+      "loss": 1.1061,
       "step": 2433
     },
     {
-      "epoch": 0.06684793057043201,
+      "epoch": 0.06906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.992894083723411e-05,
-      "loss": 1.1661,
+      "learning_rate": 1.9920139361328574e-05,
+      "loss": 1.1046,
       "step": 2434
     },
     {
-      "epoch": 0.06687539479827524,
+      "epoch": 0.06909761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928834944282776e-05,
-      "loss": 1.019,
+      "learning_rate": 1.992002339766751e-05,
+      "loss": 1.0121,
       "step": 2435
     },
     {
-      "epoch": 0.06690285902611848,
+      "epoch": 0.06912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928728972770656e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.9919907350211564e-05,
+      "loss": 1.1346,
       "step": 2436
     },
     {
-      "epoch": 0.06693032325396171,
+      "epoch": 0.06915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928622922698588e-05,
-      "loss": 1.0965,
+      "learning_rate": 1.9919791218961714e-05,
+      "loss": 1.1152,
       "step": 2437
     },
     {
-      "epoch": 0.06695778748180495,
+      "epoch": 0.0691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.992851679406741e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.9919675003918944e-05,
+      "loss": 0.9783,
       "step": 2438
     },
     {
-      "epoch": 0.06698525170964818,
+      "epoch": 0.069211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928410586877956e-05,
-      "loss": 1.1793,
+      "learning_rate": 1.9919558705084232e-05,
+      "loss": 1.0765,
       "step": 2439
     },
     {
-      "epoch": 0.06701271593749142,
+      "epoch": 0.06923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928304301131073e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.991944232245856e-05,
+      "loss": 1.1439,
       "step": 2440
     },
     {
-      "epoch": 0.06704018016533465,
+      "epoch": 0.06926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.99281979368276e-05,
-      "loss": 1.0542,
+      "learning_rate": 1.9919325856042914e-05,
+      "loss": 1.0473,
       "step": 2441
     },
     {
-      "epoch": 0.06706764439317789,
+      "epoch": 0.06929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 1.9928091493968383e-05,
-      "loss": 1.0798,
+      "learning_rate": 1.9919209305838273e-05,
+      "loss": 0.983,
       "step": 2442
     },
     {
-      "epoch": 0.06709510862102112,
+      "epoch": 0.06932463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9927984972554257e-05,
-      "loss": 1.065,
+      "learning_rate": 1.991909267184563e-05,
+      "loss": 1.0829,
       "step": 2443
     },
     {
-      "epoch": 0.06712257284886436,
+      "epoch": 0.06935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.992787837258607e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.9918975954065964e-05,
+      "loss": 1.1662,
       "step": 2444
     },
     {
-      "epoch": 0.0671500370767076,
+      "epoch": 0.06938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.9927771694064662e-05,
-      "loss": 1.1745,
+      "learning_rate": 1.9918859152500264e-05,
+      "loss": 1.1079,
       "step": 2445
     },
     {
-      "epoch": 0.06717750130455082,
+      "epoch": 0.06940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.992766493699088e-05,
-      "loss": 1.1372,
+      "learning_rate": 1.991874226714951e-05,
+      "loss": 1.1272,
       "step": 2446
     },
     {
-      "epoch": 0.06720496553239405,
+      "epoch": 0.06943813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9927558101365565e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.9918625298014696e-05,
+      "loss": 1.0447,
       "step": 2447
     },
     {
-      "epoch": 0.06723242976023729,
+      "epoch": 0.06946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.992745118718957e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.991850824509681e-05,
+      "loss": 1.1203,
       "step": 2448
     },
     {
-      "epoch": 0.06725989398808052,
+      "epoch": 0.06949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.992734419446373e-05,
-      "loss": 1.0883,
+      "learning_rate": 1.991839110839684e-05,
+      "loss": 1.1371,
       "step": 2449
     },
     {
-      "epoch": 0.06728735821592376,
+      "epoch": 0.06952326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 1.99272371231889e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.9918273887915773e-05,
+      "loss": 1.1733,
       "step": 2450
     },
     {
-      "epoch": 0.06731482244376699,
+      "epoch": 0.0695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9927129973365923e-05,
-      "loss": 0.9654,
+      "learning_rate": 1.99181565836546e-05,
+      "loss": 1.0855,
       "step": 2451
     },
     {
-      "epoch": 0.06734228667161023,
+      "epoch": 0.06958002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 1.992702274499565e-05,
-      "loss": 1.1434,
+      "learning_rate": 1.9918039195614315e-05,
+      "loss": 1.0376,
       "step": 2452
     },
     {
-      "epoch": 0.06736975089945346,
+      "epoch": 0.06960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.992691543807893e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.9917921723795903e-05,
+      "loss": 0.9678,
       "step": 2453
     },
     {
-      "epoch": 0.0673972151272967,
+      "epoch": 0.06963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926808052616605e-05,
-      "loss": 1.1125,
+      "learning_rate": 1.9917804168200362e-05,
+      "loss": 1.1274,
       "step": 2454
     },
     {
-      "epoch": 0.06742467935513993,
+      "epoch": 0.06966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.992670058860953e-05,
-      "loss": 0.959,
+      "learning_rate": 1.9917686528828682e-05,
+      "loss": 1.0511,
       "step": 2455
     },
     {
-      "epoch": 0.06745214358298317,
+      "epoch": 0.06969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926593046058556e-05,
-      "loss": 1.1812,
+      "learning_rate": 1.9917568805681862e-05,
+      "loss": 1.0992,
       "step": 2456
     },
     {
-      "epoch": 0.0674796078108264,
+      "epoch": 0.06972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926485424964533e-05,
-      "loss": 1.0667,
+      "learning_rate": 1.991745099876089e-05,
+      "loss": 1.0132,
       "step": 2457
     },
     {
-      "epoch": 0.06750707203866964,
+      "epoch": 0.06975028376844494,
       "grad_norm": 0.0,
-      "learning_rate": 1.992637772532831e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9917333108066764e-05,
+      "loss": 1.0661,
       "step": 2458
     },
     {
-      "epoch": 0.06753453626651287,
+      "epoch": 0.06977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926269947150745e-05,
-      "loss": 1.1208,
+      "learning_rate": 1.9917215133600478e-05,
+      "loss": 1.0488,
       "step": 2459
     },
     {
-      "epoch": 0.0675620004943561,
+      "epoch": 0.06980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926162090432685e-05,
-      "loss": 1.149,
+      "learning_rate": 1.9917097075363033e-05,
+      "loss": 1.0998,
       "step": 2460
     },
     {
-      "epoch": 0.06758946472219933,
+      "epoch": 0.06983541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926054155174986e-05,
-      "loss": 1.2006,
+      "learning_rate": 1.991697893335542e-05,
+      "loss": 1.1353,
       "step": 2461
     },
     {
-      "epoch": 0.06761692895004257,
+      "epoch": 0.06986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.99259461413785e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.9916860707578642e-05,
+      "loss": 0.9968,
       "step": 2462
     },
     {
-      "epoch": 0.0676443931778858,
+      "epoch": 0.06989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925838049044084e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.9916742398033694e-05,
+      "loss": 1.0874,
       "step": 2463
     },
     {
-      "epoch": 0.06767185740572904,
+      "epoch": 0.06992054483541431,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925729878172593e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.991662400472158e-05,
+      "loss": 0.994,
       "step": 2464
     },
     {
-      "epoch": 0.06769932163357227,
+      "epoch": 0.0699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.992562162876488e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.9916505527643296e-05,
+      "loss": 1.0849,
       "step": 2465
     },
     {
-      "epoch": 0.0677267858614155,
+      "epoch": 0.06997729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925513300821806e-05,
-      "loss": 1.018,
+      "learning_rate": 1.991638696679984e-05,
+      "loss": 1.0941,
       "step": 2466
     },
     {
-      "epoch": 0.06775425008925874,
+      "epoch": 0.07000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925404894344228e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.9916268322192224e-05,
+      "loss": 1.0577,
       "step": 2467
     },
     {
-      "epoch": 0.06778171431710198,
+      "epoch": 0.07003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925296409332998e-05,
-      "loss": 1.2128,
+      "learning_rate": 1.991614959382144e-05,
+      "loss": 1.0192,
       "step": 2468
     },
     {
-      "epoch": 0.06780917854494521,
+      "epoch": 0.07006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.992518784578898e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.9916030781688495e-05,
+      "loss": 1.158,
       "step": 2469
     },
     {
-      "epoch": 0.06783664277278845,
+      "epoch": 0.07009080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.9925079203713032e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.9915911885794395e-05,
+      "loss": 1.1432,
       "step": 2470
     },
     {
-      "epoch": 0.06786410700063168,
+      "epoch": 0.07011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.992497048310601e-05,
-      "loss": 1.1494,
+      "learning_rate": 1.9915792906140138e-05,
+      "loss": 1.0707,
       "step": 2471
     },
     {
-      "epoch": 0.06789157122847492,
+      "epoch": 0.07014755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.992486168396878e-05,
-      "loss": 1.0495,
+      "learning_rate": 1.9915673842726733e-05,
+      "loss": 1.1093,
       "step": 2472
     },
     {
-      "epoch": 0.06791903545631815,
+      "epoch": 0.07017593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.99247528063022e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.9915554695555186e-05,
+      "loss": 0.9374,
       "step": 2473
     },
     {
-      "epoch": 0.06794649968416137,
+      "epoch": 0.07020431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.9924643850107128e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.9915435464626504e-05,
+      "loss": 1.0326,
       "step": 2474
     },
     {
-      "epoch": 0.06797396391200461,
+      "epoch": 0.07023269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.992453481538443e-05,
-      "loss": 1.1065,
+      "learning_rate": 1.9915316149941694e-05,
+      "loss": 1.0177,
       "step": 2475
     },
     {
-      "epoch": 0.06800142813984784,
+      "epoch": 0.07026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.9924425702134972e-05,
-      "loss": 1.2001,
+      "learning_rate": 1.9915196751501757e-05,
+      "loss": 1.1109,
       "step": 2476
     },
     {
-      "epoch": 0.06802889236769108,
+      "epoch": 0.0702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.992431651035961e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.991507726930771e-05,
+      "loss": 1.0547,
       "step": 2477
     },
     {
-      "epoch": 0.06805635659553431,
+      "epoch": 0.07031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.992420724005921e-05,
-      "loss": 1.1703,
+      "learning_rate": 1.9914957703360564e-05,
+      "loss": 1.0151,
       "step": 2478
     },
     {
-      "epoch": 0.06808382082337755,
+      "epoch": 0.07034619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 1.9924097891234643e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.991483805366132e-05,
+      "loss": 1.0739,
       "step": 2479
     },
     {
-      "epoch": 0.06811128505122079,
+      "epoch": 0.07037457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923988463886767e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.9914718320210994e-05,
+      "loss": 1.0057,
       "step": 2480
     },
     {
-      "epoch": 0.06813874927906402,
+      "epoch": 0.07040295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.992387895801645e-05,
-      "loss": 1.1234,
+      "learning_rate": 1.99145985030106e-05,
+      "loss": 1.1395,
       "step": 2481
     },
     {
-      "epoch": 0.06816621350690726,
+      "epoch": 0.07043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923769373624555e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.991447860206114e-05,
+      "loss": 1.065,
       "step": 2482
     },
     {
-      "epoch": 0.06819367773475049,
+      "epoch": 0.07045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923659710711954e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.991435861736364e-05,
+      "loss": 1.0942,
       "step": 2483
     },
     {
-      "epoch": 0.06822114196259373,
+      "epoch": 0.07048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923549969279518e-05,
-      "loss": 1.114,
+      "learning_rate": 1.9914238548919106e-05,
+      "loss": 1.0609,
       "step": 2484
     },
     {
-      "epoch": 0.06824860619043696,
+      "epoch": 0.07051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.992344014932811e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.9914118396728547e-05,
+      "loss": 1.0809,
       "step": 2485
     },
     {
-      "epoch": 0.0682760704182802,
+      "epoch": 0.07054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923330250858594e-05,
-      "loss": 1.1561,
+      "learning_rate": 1.991399816079299e-05,
+      "loss": 0.9764,
       "step": 2486
     },
     {
-      "epoch": 0.06830353464612342,
+      "epoch": 0.07057321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923220273871848e-05,
-      "loss": 1.0612,
+      "learning_rate": 1.9913877841113444e-05,
+      "loss": 1.2016,
       "step": 2487
     },
     {
-      "epoch": 0.06833099887396665,
+      "epoch": 0.07060158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923110218368743e-05,
-      "loss": 1.0841,
+      "learning_rate": 1.9913757437690928e-05,
+      "loss": 0.9975,
       "step": 2488
     },
     {
-      "epoch": 0.06835846310180989,
+      "epoch": 0.0706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923000084350142e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.9913636950526455e-05,
+      "loss": 1.099,
       "step": 2489
     },
     {
-      "epoch": 0.06838592732965312,
+      "epoch": 0.0706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.992288987181692e-05,
-      "loss": 1.0625,
+      "learning_rate": 1.9913516379621046e-05,
+      "loss": 1.1277,
       "step": 2490
     },
     {
-      "epoch": 0.06841339155749636,
+      "epoch": 0.07068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.992277958076995e-05,
-      "loss": 1.1077,
+      "learning_rate": 1.9913395724975718e-05,
+      "loss": 1.0233,
       "step": 2491
     },
     {
-      "epoch": 0.0684408557853396,
+      "epoch": 0.07071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922669211210105e-05,
-      "loss": 1.058,
+      "learning_rate": 1.9913274986591488e-05,
+      "loss": 1.0486,
       "step": 2492
     },
     {
-      "epoch": 0.06846832001318283,
+      "epoch": 0.07074347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922558763138257e-05,
-      "loss": 0.9676,
+      "learning_rate": 1.9913154164469382e-05,
+      "loss": 1.0503,
       "step": 2493
     },
     {
-      "epoch": 0.06849578424102606,
+      "epoch": 0.07077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922448236555284e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.991303325861042e-05,
+      "loss": 1.0247,
       "step": 2494
     },
     {
-      "epoch": 0.0685232484688693,
+      "epoch": 0.07080022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.992233763146205e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.991291226901561e-05,
+      "loss": 0.9715,
       "step": 2495
     },
     {
-      "epoch": 0.06855071269671253,
+      "epoch": 0.07082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922226947859445e-05,
-      "loss": 1.0873,
+      "learning_rate": 1.9912791195685992e-05,
+      "loss": 1.0267,
       "step": 2496
     },
     {
-      "epoch": 0.06857817692455577,
+      "epoch": 0.07085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922116185748332e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.9912670038622583e-05,
+      "loss": 1.006,
       "step": 2497
     },
     {
-      "epoch": 0.068605641152399,
+      "epoch": 0.07088535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 1.9922005345129592e-05,
-      "loss": 1.0898,
+      "learning_rate": 1.9912548797826402e-05,
+      "loss": 1.0259,
       "step": 2498
     },
     {
-      "epoch": 0.06863310538024224,
+      "epoch": 0.07091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921894426004104e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.9912427473298473e-05,
+      "loss": 1.0486,
       "step": 2499
     },
     {
-      "epoch": 0.06866056960808548,
+      "epoch": 0.07094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921783428372744e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.9912306065039827e-05,
+      "loss": 1.0762,
       "step": 2500
     },
     {
-      "epoch": 0.0686880338359287,
+      "epoch": 0.0709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.992167235223639e-05,
-      "loss": 1.1513,
+      "learning_rate": 1.9912184573051486e-05,
+      "loss": 1.073,
       "step": 2501
     },
     {
-      "epoch": 0.06871549806377193,
+      "epoch": 0.0709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921561197595923e-05,
-      "loss": 1.051,
+      "learning_rate": 1.9912062997334476e-05,
+      "loss": 1.0759,
       "step": 2502
     },
     {
-      "epoch": 0.06874296229161517,
+      "epoch": 0.0710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921449964452218e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.9911941337889825e-05,
+      "loss": 1.0722,
       "step": 2503
     },
     {
-      "epoch": 0.0687704265194584,
+      "epoch": 0.07105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921338652806157e-05,
-      "loss": 1.0597,
+      "learning_rate": 1.9911819594718556e-05,
+      "loss": 0.9524,
       "step": 2504
     },
     {
-      "epoch": 0.06879789074730164,
+      "epoch": 0.07108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921227262658624e-05,
-      "loss": 1.1329,
+      "learning_rate": 1.9911697767821705e-05,
+      "loss": 1.0877,
       "step": 2505
     },
     {
-      "epoch": 0.06882535497514487,
+      "epoch": 0.07111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921115794010496e-05,
-      "loss": 1.0934,
+      "learning_rate": 1.9911575857200296e-05,
+      "loss": 1.1547,
       "step": 2506
     },
     {
-      "epoch": 0.06885281920298811,
+      "epoch": 0.07114074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.9921004246862656e-05,
-      "loss": 1.1306,
+      "learning_rate": 1.9911453862855362e-05,
+      "loss": 1.1913,
       "step": 2507
     },
     {
-      "epoch": 0.06888028343083134,
+      "epoch": 0.07116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.992089262121599e-05,
-      "loss": 1.0812,
+      "learning_rate": 1.991133178478793e-05,
+      "loss": 1.039,
       "step": 2508
     },
     {
-      "epoch": 0.06890774765867458,
+      "epoch": 0.07119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9920780917071377e-05,
-      "loss": 1.1402,
+      "learning_rate": 1.991120962299903e-05,
+      "loss": 1.0932,
       "step": 2509
     },
     {
-      "epoch": 0.06893521188651781,
+      "epoch": 0.07122587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 1.99206691344297e-05,
-      "loss": 1.1331,
+      "learning_rate": 1.99110873774897e-05,
+      "loss": 1.1093,
       "step": 2510
     },
     {
-      "epoch": 0.06896267611436105,
+      "epoch": 0.07125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.992055727329185e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.991096504826097e-05,
+      "loss": 1.0519,
       "step": 2511
     },
     {
-      "epoch": 0.06899014034220428,
+      "epoch": 0.07128263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 1.9920445333658705e-05,
-      "loss": 1.2794,
+      "learning_rate": 1.9910842635313875e-05,
+      "loss": 1.1132,
       "step": 2512
     },
     {
-      "epoch": 0.06901760457004752,
+      "epoch": 0.07131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.9920333315531154e-05,
-      "loss": 1.059,
+      "learning_rate": 1.991072013864944e-05,
+      "loss": 1.0199,
       "step": 2513
     },
     {
-      "epoch": 0.06904506879789074,
+      "epoch": 0.07133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9920221218910085e-05,
-      "loss": 1.1665,
+      "learning_rate": 1.9910597558268713e-05,
+      "loss": 0.9458,
       "step": 2514
     },
     {
-      "epoch": 0.06907253302573398,
+      "epoch": 0.0713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9920109043796385e-05,
-      "loss": 1.2139,
+      "learning_rate": 1.9910474894172717e-05,
+      "loss": 1.0148,
       "step": 2515
     },
     {
-      "epoch": 0.06909999725357721,
+      "epoch": 0.0713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9919996790190933e-05,
-      "loss": 1.098,
+      "learning_rate": 1.99103521463625e-05,
+      "loss": 1.1014,
       "step": 2516
     },
     {
-      "epoch": 0.06912746148142045,
+      "epoch": 0.07142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.991988445809463e-05,
-      "loss": 1.1144,
+      "learning_rate": 1.9910229314839088e-05,
+      "loss": 1.0468,
       "step": 2517
     },
     {
-      "epoch": 0.06915492570926368,
+      "epoch": 0.07145289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9919772047508355e-05,
-      "loss": 1.1187,
+      "learning_rate": 1.9910106399603525e-05,
+      "loss": 0.9307,
       "step": 2518
     },
     {
-      "epoch": 0.06918238993710692,
+      "epoch": 0.07148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.9919659558433002e-05,
-      "loss": 1.0134,
+      "learning_rate": 1.9909983400656848e-05,
+      "loss": 1.104,
       "step": 2519
     },
     {
-      "epoch": 0.06920985416495015,
+      "epoch": 0.07150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.991954699086946e-05,
-      "loss": 1.1634,
+      "learning_rate": 1.9909860318000097e-05,
+      "loss": 1.0524,
       "step": 2520
     },
     {
-      "epoch": 0.06923731839279339,
+      "epoch": 0.07153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.991943434481862e-05,
-      "loss": 1.1604,
+      "learning_rate": 1.990973715163431e-05,
+      "loss": 1.0929,
       "step": 2521
     },
     {
-      "epoch": 0.06926478262063662,
+      "epoch": 0.07156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.9919321620281372e-05,
-      "loss": 1.0041,
+      "learning_rate": 1.990961390156053e-05,
+      "loss": 1.1543,
       "step": 2522
     },
     {
-      "epoch": 0.06929224684847986,
+      "epoch": 0.07159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.991920881725861e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.990949056777979e-05,
+      "loss": 1.1684,
       "step": 2523
     },
     {
-      "epoch": 0.06931971107632309,
+      "epoch": 0.07162315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 1.9919095935751226e-05,
-      "loss": 1.044,
+      "learning_rate": 1.990936715029314e-05,
+      "loss": 1.0423,
       "step": 2524
     },
     {
-      "epoch": 0.06934717530416633,
+      "epoch": 0.07165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.991898297576011e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.9909243649101625e-05,
+      "loss": 1.0888,
       "step": 2525
     },
     {
-      "epoch": 0.06937463953200956,
+      "epoch": 0.07167990919409761,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918869937286163e-05,
-      "loss": 1.1415,
+      "learning_rate": 1.9909120064206277e-05,
+      "loss": 1.1314,
       "step": 2526
     },
     {
-      "epoch": 0.0694021037598528,
+      "epoch": 0.07170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.991875682033027e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.990899639560815e-05,
+      "loss": 0.9606,
       "step": 2527
     },
     {
-      "epoch": 0.06942956798769602,
+      "epoch": 0.0717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918643624893336e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.9908872643308284e-05,
+      "loss": 1.0649,
       "step": 2528
     },
     {
-      "epoch": 0.06945703221553925,
+      "epoch": 0.07176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918530350976248e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.9908748807307725e-05,
+      "loss": 0.9916,
       "step": 2529
     },
     {
-      "epoch": 0.06948449644338249,
+      "epoch": 0.07179341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918416998579907e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.990862488760752e-05,
+      "loss": 0.9976,
       "step": 2530
     },
     {
-      "epoch": 0.06951196067122573,
+      "epoch": 0.07182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918303567705205e-05,
-      "loss": 0.8809,
+      "learning_rate": 1.9908500884208716e-05,
+      "loss": 1.0401,
       "step": 2531
     },
     {
-      "epoch": 0.06953942489906896,
+      "epoch": 0.07185017026106696,
       "grad_norm": 0.0,
-      "learning_rate": 1.991819005835305e-05,
-      "loss": 1.1584,
+      "learning_rate": 1.9908376797112357e-05,
+      "loss": 1.0511,
       "step": 2532
     },
     {
-      "epoch": 0.0695668891269122,
+      "epoch": 0.07187854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918076470524328e-05,
-      "loss": 1.0686,
+      "learning_rate": 1.9908252626319494e-05,
+      "loss": 0.9583,
       "step": 2533
     },
     {
-      "epoch": 0.06959435335475543,
+      "epoch": 0.07190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917962804219946e-05,
-      "loss": 1.2073,
+      "learning_rate": 1.9908128371831177e-05,
+      "loss": 1.1569,
       "step": 2534
     },
     {
-      "epoch": 0.06962181758259867,
+      "epoch": 0.07193530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.99178490594408e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.9908004033648452e-05,
+      "loss": 0.9966,
       "step": 2535
     },
     {
-      "epoch": 0.0696492818104419,
+      "epoch": 0.07196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917735236187788e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.9907879611772377e-05,
+      "loss": 1.1358,
       "step": 2536
     },
     {
-      "epoch": 0.06967674603828514,
+      "epoch": 0.07199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917621334461813e-05,
-      "loss": 1.1707,
+      "learning_rate": 1.9907755106203992e-05,
+      "loss": 1.1798,
       "step": 2537
     },
     {
-      "epoch": 0.06970421026612837,
+      "epoch": 0.07202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917507354263778e-05,
-      "loss": 1.0225,
+      "learning_rate": 1.9907630516944358e-05,
+      "loss": 1.1101,
       "step": 2538
     },
     {
-      "epoch": 0.06973167449397161,
+      "epoch": 0.07204880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917393295594584e-05,
-      "loss": 1.127,
+      "learning_rate": 1.9907505843994524e-05,
+      "loss": 1.1682,
       "step": 2539
     },
     {
-      "epoch": 0.06975913872181484,
+      "epoch": 0.0720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.991727915845513e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.990738108735554e-05,
+      "loss": 1.0648,
       "step": 2540
     },
     {
-      "epoch": 0.06978660294965806,
+      "epoch": 0.07210556186152099,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917164942846324e-05,
-      "loss": 1.107,
+      "learning_rate": 1.9907256247028463e-05,
+      "loss": 1.0364,
       "step": 2541
     },
     {
-      "epoch": 0.0698140671775013,
+      "epoch": 0.0721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.9917050648769066e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.9907131323014347e-05,
+      "loss": 1.0339,
       "step": 2542
     },
     {
-      "epoch": 0.06984153140534453,
+      "epoch": 0.07216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.9916936276224262e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.990700631531425e-05,
+      "loss": 1.0399,
       "step": 2543
     },
     {
-      "epoch": 0.06986899563318777,
+      "epoch": 0.07219069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.991682182521282e-05,
-      "loss": 1.034,
+      "learning_rate": 1.990688122392922e-05,
+      "loss": 1.0823,
       "step": 2544
     },
     {
-      "epoch": 0.069896459861031,
+      "epoch": 0.07221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.991670729573564e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.9906756048860323e-05,
+      "loss": 1.077,
       "step": 2545
     },
     {
-      "epoch": 0.06992392408887424,
+      "epoch": 0.07224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.991659268779363e-05,
-      "loss": 1.1647,
+      "learning_rate": 1.9906630790108612e-05,
+      "loss": 1.011,
       "step": 2546
     },
     {
-      "epoch": 0.06995138831671747,
+      "epoch": 0.07227582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.99164780013877e-05,
-      "loss": 1.166,
+      "learning_rate": 1.9906505447675144e-05,
+      "loss": 1.1248,
       "step": 2547
     },
     {
-      "epoch": 0.06997885254456071,
+      "epoch": 0.07230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.991636323651875e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.990638002156098e-05,
+      "loss": 0.9417,
       "step": 2548
     },
     {
-      "epoch": 0.07000631677240395,
+      "epoch": 0.07233257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 1.9916248393187695e-05,
-      "loss": 1.099,
+      "learning_rate": 1.990625451176718e-05,
+      "loss": 1.0848,
       "step": 2549
     },
     {
-      "epoch": 0.07003378100024718,
+      "epoch": 0.07236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.991613347139544e-05,
-      "loss": 1.1089,
+      "learning_rate": 1.99061289182948e-05,
+      "loss": 1.0529,
       "step": 2550
     },
     {
-      "epoch": 0.07006124522809042,
+      "epoch": 0.07238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9916018471142896e-05,
-      "loss": 1.053,
+      "learning_rate": 1.9906003241144907e-05,
+      "loss": 0.9844,
       "step": 2551
     },
     {
-      "epoch": 0.07008870945593365,
+      "epoch": 0.07241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915903392430976e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.9905877480318556e-05,
+      "loss": 1.0559,
       "step": 2552
     },
     {
-      "epoch": 0.07011617368377689,
+      "epoch": 0.07244608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915788235260586e-05,
-      "loss": 1.1153,
+      "learning_rate": 1.9905751635816816e-05,
+      "loss": 1.1498,
       "step": 2553
     },
     {
-      "epoch": 0.07014363791162012,
+      "epoch": 0.0724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915672999632638e-05,
-      "loss": 1.0791,
+      "learning_rate": 1.990562570764074e-05,
+      "loss": 1.0977,
       "step": 2554
     },
     {
-      "epoch": 0.07017110213946334,
+      "epoch": 0.07250283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915557685548042e-05,
-      "loss": 1.1215,
+      "learning_rate": 1.9905499695791405e-05,
+      "loss": 1.0185,
       "step": 2555
     },
     {
-      "epoch": 0.07019856636730658,
+      "epoch": 0.07253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915442293007715e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.9905373600269868e-05,
+      "loss": 1.0683,
       "step": 2556
     },
     {
-      "epoch": 0.07022603059514981,
+      "epoch": 0.07255959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.991532682201257e-05,
-      "loss": 1.0671,
+      "learning_rate": 1.990524742107719e-05,
+      "loss": 0.9836,
       "step": 2557
     },
     {
-      "epoch": 0.07025349482299305,
+      "epoch": 0.07258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.9915211272563516e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.990512115821445e-05,
+      "loss": 0.9332,
       "step": 2558
     },
     {
-      "epoch": 0.07028095905083628,
+      "epoch": 0.07261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.991509564466147e-05,
-      "loss": 1.04,
+      "learning_rate": 1.9904994811682702e-05,
+      "loss": 1.1359,
       "step": 2559
     },
     {
-      "epoch": 0.07030842327867952,
+      "epoch": 0.07264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9914979938307345e-05,
-      "loss": 1.1252,
+      "learning_rate": 1.9904868381483015e-05,
+      "loss": 1.075,
       "step": 2560
     },
     {
-      "epoch": 0.07033588750652275,
+      "epoch": 0.07267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.991486415350206e-05,
-      "loss": 1.1205,
+      "learning_rate": 1.9904741867616463e-05,
+      "loss": 1.0912,
       "step": 2561
     },
     {
-      "epoch": 0.07036335173436599,
+      "epoch": 0.07270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9914748290246528e-05,
-      "loss": 1.0973,
+      "learning_rate": 1.9904615270084113e-05,
+      "loss": 1.0843,
       "step": 2562
     },
     {
-      "epoch": 0.07039081596220922,
+      "epoch": 0.07272985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 1.991463234854167e-05,
-      "loss": 1.2068,
+      "learning_rate": 1.990448858888703e-05,
+      "loss": 1.0455,
       "step": 2563
     },
     {
-      "epoch": 0.07041828019005246,
+      "epoch": 0.07275822928490353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9914516328388396e-05,
-      "loss": 1.1735,
+      "learning_rate": 1.9904361824026284e-05,
+      "loss": 1.0604,
       "step": 2564
     },
     {
-      "epoch": 0.0704457444178957,
+      "epoch": 0.07278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.9914400229787628e-05,
-      "loss": 1.0961,
+      "learning_rate": 1.990423497550295e-05,
+      "loss": 0.9955,
       "step": 2565
     },
     {
-      "epoch": 0.07047320864573893,
+      "epoch": 0.0728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9914284052740286e-05,
-      "loss": 1.1719,
+      "learning_rate": 1.99041080433181e-05,
+      "loss": 1.08,
       "step": 2566
     },
     {
-      "epoch": 0.07050067287358217,
+      "epoch": 0.0728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.991416779724729e-05,
-      "loss": 1.1396,
+      "learning_rate": 1.99039810274728e-05,
+      "loss": 1.1174,
       "step": 2567
     },
     {
-      "epoch": 0.07052813710142539,
+      "epoch": 0.07287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.991405146330956e-05,
-      "loss": 1.0206,
+      "learning_rate": 1.9903853927968134e-05,
+      "loss": 1.0415,
       "step": 2568
     },
     {
-      "epoch": 0.07055560132926862,
+      "epoch": 0.07290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.991393505092801e-05,
-      "loss": 1.1346,
+      "learning_rate": 1.9903726744805163e-05,
+      "loss": 0.994,
       "step": 2569
     },
     {
-      "epoch": 0.07058306555711186,
+      "epoch": 0.07292849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913818560103567e-05,
-      "loss": 1.117,
+      "learning_rate": 1.990359947798497e-05,
+      "loss": 1.0046,
       "step": 2570
     },
     {
-      "epoch": 0.07061052978495509,
+      "epoch": 0.07295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913701990837147e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.9903472127508623e-05,
+      "loss": 1.1033,
       "step": 2571
     },
     {
-      "epoch": 0.07063799401279833,
+      "epoch": 0.07298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913585343129683e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.9903344693377203e-05,
+      "loss": 1.1935,
       "step": 2572
     },
     {
-      "epoch": 0.07066545824064156,
+      "epoch": 0.07301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.991346861698209e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.9903217175591786e-05,
+      "loss": 1.1702,
       "step": 2573
     },
     {
-      "epoch": 0.0706929224684848,
+      "epoch": 0.07304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913351812395294e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.9903089574153452e-05,
+      "loss": 1.1328,
       "step": 2574
     },
     {
-      "epoch": 0.07072038669632803,
+      "epoch": 0.07307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913234929370215e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.990296188906327e-05,
+      "loss": 1.019,
       "step": 2575
     },
     {
-      "epoch": 0.07074785092417127,
+      "epoch": 0.07309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.9913117967907785e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.9902834120322326e-05,
+      "loss": 1.0583,
       "step": 2576
     },
     {
-      "epoch": 0.0707753151520145,
+      "epoch": 0.07312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.991300092800892e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.9902706267931698e-05,
+      "loss": 1.0551,
       "step": 2577
     },
     {
-      "epoch": 0.07080277937985774,
+      "epoch": 0.073155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.991288380967456e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.9902578331892464e-05,
+      "loss": 1.1555,
       "step": 2578
     },
     {
-      "epoch": 0.07083024360770097,
+      "epoch": 0.0731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912766612905618e-05,
-      "loss": 1.0596,
+      "learning_rate": 1.9902450312205707e-05,
+      "loss": 1.2054,
       "step": 2579
     },
     {
-      "epoch": 0.07085770783554421,
+      "epoch": 0.07321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912649337703027e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.99023222088725e-05,
+      "loss": 1.0443,
       "step": 2580
     },
     {
-      "epoch": 0.07088517206338744,
+      "epoch": 0.07324063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912531984067717e-05,
-      "loss": 1.1275,
+      "learning_rate": 1.9902194021893943e-05,
+      "loss": 1.0368,
       "step": 2581
     },
     {
-      "epoch": 0.07091263629123067,
+      "epoch": 0.07326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912414552000615e-05,
-      "loss": 1.1045,
+      "learning_rate": 1.9902065751271103e-05,
+      "loss": 1.032,
       "step": 2582
     },
     {
-      "epoch": 0.0709401005190739,
+      "epoch": 0.07329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912297041502647e-05,
-      "loss": 1.0473,
+      "learning_rate": 1.9901937397005065e-05,
+      "loss": 1.1155,
       "step": 2583
     },
     {
-      "epoch": 0.07096756474691714,
+      "epoch": 0.07332576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912179452574743e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.990180895909692e-05,
+      "loss": 1.0077,
       "step": 2584
     },
     {
-      "epoch": 0.07099502897476037,
+      "epoch": 0.07335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.9912061785217837e-05,
-      "loss": 1.0869,
+      "learning_rate": 1.990168043754775e-05,
+      "loss": 1.0623,
       "step": 2585
     },
     {
-      "epoch": 0.0710224932026036,
+      "epoch": 0.07338251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.991194403943286e-05,
-      "loss": 1.1362,
+      "learning_rate": 1.9901551832358642e-05,
+      "loss": 1.2307,
       "step": 2586
     },
     {
-      "epoch": 0.07104995743044684,
+      "epoch": 0.07341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.991182621522074e-05,
-      "loss": 1.0498,
+      "learning_rate": 1.9901423143530677e-05,
+      "loss": 1.023,
       "step": 2587
     },
     {
-      "epoch": 0.07107742165829008,
+      "epoch": 0.07343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9911708312582415e-05,
-      "loss": 1.0917,
+      "learning_rate": 1.9901294371064947e-05,
+      "loss": 1.1315,
       "step": 2588
     },
     {
-      "epoch": 0.07110488588613331,
+      "epoch": 0.07346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.9911590331518813e-05,
-      "loss": 1.1228,
+      "learning_rate": 1.9901165514962536e-05,
+      "loss": 1.0136,
       "step": 2589
     },
     {
-      "epoch": 0.07113235011397655,
+      "epoch": 0.07349602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 1.991147227203087e-05,
-      "loss": 1.0336,
+      "learning_rate": 1.990103657522454e-05,
+      "loss": 1.1786,
       "step": 2590
     },
     {
-      "epoch": 0.07115981434181978,
+      "epoch": 0.07352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.9911354134119517e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.990090755185204e-05,
+      "loss": 1.255,
       "step": 2591
     },
     {
-      "epoch": 0.07118727856966302,
+      "epoch": 0.0735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.991123591778569e-05,
-      "loss": 1.0904,
+      "learning_rate": 1.9900778444846126e-05,
+      "loss": 1.1486,
       "step": 2592
     },
     {
-      "epoch": 0.07121474279750625,
+      "epoch": 0.0735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.9911117623030324e-05,
-      "loss": 1.0218,
+      "learning_rate": 1.9900649254207894e-05,
+      "loss": 0.9801,
       "step": 2593
     },
     {
-      "epoch": 0.07124220702534949,
+      "epoch": 0.07360953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.991099924985436e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.9900519979938436e-05,
+      "loss": 1.0435,
       "step": 2594
     },
     {
-      "epoch": 0.07126967125319271,
+      "epoch": 0.07363791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.991088079825873e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.9900390622038834e-05,
+      "loss": 1.0346,
       "step": 2595
     },
     {
-      "epoch": 0.07129713548103594,
+      "epoch": 0.07366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910762268244372e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.9900261180510192e-05,
+      "loss": 1.082,
       "step": 2596
     },
     {
-      "epoch": 0.07132459970887918,
+      "epoch": 0.07369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910643659812225e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.9900131655353597e-05,
+      "loss": 1.1472,
       "step": 2597
     },
     {
-      "epoch": 0.07135206393672242,
+      "epoch": 0.07372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910524972963224e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.9900002046570147e-05,
+      "loss": 1.0288,
       "step": 2598
     },
     {
-      "epoch": 0.07137952816456565,
+      "epoch": 0.07375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.991040620769831e-05,
-      "loss": 1.0962,
+      "learning_rate": 1.989987235416093e-05,
+      "loss": 1.1273,
       "step": 2599
     },
     {
-      "epoch": 0.07140699239240889,
+      "epoch": 0.07377979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910287364018422e-05,
-      "loss": 1.1166,
+      "learning_rate": 1.989974257812705e-05,
+      "loss": 1.0779,
       "step": 2600
     },
     {
-      "epoch": 0.07143445662025212,
+      "epoch": 0.07380817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910168441924506e-05,
-      "loss": 1.0919,
+      "learning_rate": 1.9899612718469603e-05,
+      "loss": 0.9949,
       "step": 2601
     },
     {
-      "epoch": 0.07146192084809536,
+      "epoch": 0.07383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9910049441417497e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.9899482775189675e-05,
+      "loss": 1.0545,
       "step": 2602
     },
     {
-      "epoch": 0.07148938507593859,
+      "epoch": 0.07386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909930362498337e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.9899352748288377e-05,
+      "loss": 1.1253,
       "step": 2603
     },
     {
-      "epoch": 0.07151684930378183,
+      "epoch": 0.07389330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 1.990981120516797e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.9899222637766798e-05,
+      "loss": 1.02,
       "step": 2604
     },
     {
-      "epoch": 0.07154431353162506,
+      "epoch": 0.0739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909691969427337e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.989909244362604e-05,
+      "loss": 0.9231,
       "step": 2605
     },
     {
-      "epoch": 0.0715717777594683,
+      "epoch": 0.07395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909572655277383e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.9898962165867204e-05,
+      "loss": 1.017,
       "step": 2606
     },
     {
-      "epoch": 0.07159924198731153,
+      "epoch": 0.07397843359818389,
       "grad_norm": 0.0,
-      "learning_rate": 1.990945326271905e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.989883180449139e-05,
+      "loss": 1.0338,
       "step": 2607
     },
     {
-      "epoch": 0.07162670621515477,
+      "epoch": 0.07400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909333791753286e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.9898701359499698e-05,
+      "loss": 1.0941,
       "step": 2608
     },
     {
-      "epoch": 0.07165417044299799,
+      "epoch": 0.07403518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909214242381037e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.989857083089323e-05,
+      "loss": 1.0181,
       "step": 2609
     },
     {
-      "epoch": 0.07168163467084122,
+      "epoch": 0.07406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.9909094614603243e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.989844021867309e-05,
+      "loss": 0.9684,
       "step": 2610
     },
     {
-      "epoch": 0.07170909889868446,
+      "epoch": 0.07409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.990897490842085e-05,
-      "loss": 1.1094,
+      "learning_rate": 1.989830952284038e-05,
+      "loss": 1.0087,
       "step": 2611
     },
     {
-      "epoch": 0.0717365631265277,
+      "epoch": 0.07412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.9908855123834816e-05,
-      "loss": 1.0692,
+      "learning_rate": 1.9898178743396208e-05,
+      "loss": 1.1486,
       "step": 2612
     },
     {
-      "epoch": 0.07176402735437093,
+      "epoch": 0.07414869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.990873526084608e-05,
-      "loss": 1.0664,
+      "learning_rate": 1.9898047880341674e-05,
+      "loss": 1.1004,
       "step": 2613
     },
     {
-      "epoch": 0.07179149158221416,
+      "epoch": 0.07417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.990861531945559e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.989791693367788e-05,
+      "loss": 1.1404,
       "step": 2614
     },
     {
-      "epoch": 0.0718189558100574,
+      "epoch": 0.07420544835414301,
       "grad_norm": 0.0,
-      "learning_rate": 1.9908495299664303e-05,
-      "loss": 1.0802,
+      "learning_rate": 1.989778590340594e-05,
+      "loss": 1.1234,
       "step": 2615
     },
     {
-      "epoch": 0.07184642003790064,
+      "epoch": 0.07423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.990837520147316e-05,
-      "loss": 1.1152,
+      "learning_rate": 1.9897654789526957e-05,
+      "loss": 1.0663,
       "step": 2616
     },
     {
-      "epoch": 0.07187388426574387,
+      "epoch": 0.0742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.990825502488311e-05,
-      "loss": 1.1052,
+      "learning_rate": 1.989752359204204e-05,
+      "loss": 1.0382,
       "step": 2617
     },
     {
-      "epoch": 0.0719013484935871,
+      "epoch": 0.07429057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 1.9908134769895117e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.9897392310952295e-05,
+      "loss": 1.0348,
       "step": 2618
     },
     {
-      "epoch": 0.07192881272143034,
+      "epoch": 0.07431895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.9908014436510116e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.9897260946258833e-05,
+      "loss": 1.2169,
       "step": 2619
     },
     {
-      "epoch": 0.07195627694927358,
+      "epoch": 0.07434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907894024729073e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.9897129497962764e-05,
+      "loss": 1.002,
       "step": 2620
     },
     {
-      "epoch": 0.07198374117711681,
+      "epoch": 0.07437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907773534552927e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.9896997966065194e-05,
+      "loss": 1.0812,
       "step": 2621
     },
     {
-      "epoch": 0.07201120540496005,
+      "epoch": 0.07440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907652965982646e-05,
-      "loss": 1.1716,
+      "learning_rate": 1.989686635056724e-05,
+      "loss": 1.0351,
       "step": 2622
     },
     {
-      "epoch": 0.07203866963280327,
+      "epoch": 0.07443246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.990753231901917e-05,
-      "loss": 1.1015,
+      "learning_rate": 1.9896734651470013e-05,
+      "loss": 1.0816,
       "step": 2623
     },
     {
-      "epoch": 0.0720661338606465,
+      "epoch": 0.07446083995459706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907411593663467e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.9896602868774617e-05,
+      "loss": 1.0122,
       "step": 2624
     },
     {
-      "epoch": 0.07209359808848974,
+      "epoch": 0.07448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907290789916483e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.9896471002482175e-05,
+      "loss": 1.0958,
       "step": 2625
     },
     {
-      "epoch": 0.07212106231633297,
+      "epoch": 0.07451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907169907779175e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.98963390525938e-05,
+      "loss": 1.162,
       "step": 2626
     },
     {
-      "epoch": 0.07214852654417621,
+      "epoch": 0.07454597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907048947252502e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.98962070191106e-05,
+      "loss": 1.0746,
       "step": 2627
     },
     {
-      "epoch": 0.07217599077201944,
+      "epoch": 0.07457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906927908337422e-05,
-      "loss": 1.1592,
+      "learning_rate": 1.98960749020337e-05,
+      "loss": 1.1483,
       "step": 2628
     },
     {
-      "epoch": 0.07220345499986268,
+      "epoch": 0.0746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906806791034888e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.9895942701364207e-05,
+      "loss": 1.0732,
       "step": 2629
     },
     {
-      "epoch": 0.07223091922770591,
+      "epoch": 0.0746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.990668559534586e-05,
-      "loss": 1.1626,
+      "learning_rate": 1.989581041710324e-05,
+      "loss": 1.0281,
       "step": 2630
     },
     {
-      "epoch": 0.07225838345554915,
+      "epoch": 0.0746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906564321271304e-05,
-      "loss": 1.1046,
+      "learning_rate": 1.989567804925192e-05,
+      "loss": 1.1464,
       "step": 2631
     },
     {
-      "epoch": 0.07228584768339238,
+      "epoch": 0.07468785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906442968812173e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.989554559781136e-05,
+      "loss": 0.9626,
       "step": 2632
     },
     {
-      "epoch": 0.07231331191123562,
+      "epoch": 0.07471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906321537969422e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.9895413062782683e-05,
+      "loss": 0.9904,
       "step": 2633
     },
     {
-      "epoch": 0.07234077613907886,
+      "epoch": 0.07474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.990620002874402e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.9895280444167006e-05,
+      "loss": 1.0397,
       "step": 2634
     },
     {
-      "epoch": 0.07236824036692209,
+      "epoch": 0.07477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.9906078441136926e-05,
-      "loss": 1.1724,
+      "learning_rate": 1.9895147741965453e-05,
+      "loss": 1.1033,
       "step": 2635
     },
     {
-      "epoch": 0.07239570459476531,
+      "epoch": 0.07480136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905956775149105e-05,
-      "loss": 1.1942,
+      "learning_rate": 1.9895014956179143e-05,
+      "loss": 1.0931,
       "step": 2636
     },
     {
-      "epoch": 0.07242316882260855,
+      "epoch": 0.07482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905835030781514e-05,
-      "loss": 1.03,
+      "learning_rate": 1.989488208680919e-05,
+      "loss": 0.9296,
       "step": 2637
     },
     {
-      "epoch": 0.07245063305045178,
+      "epoch": 0.07485811577752553,
       "grad_norm": 0.0,
-      "learning_rate": 1.990571320803512e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.989474913385673e-05,
+      "loss": 1.0757,
       "step": 2638
     },
     {
-      "epoch": 0.07247809727829502,
+      "epoch": 0.07488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905591306910886e-05,
-      "loss": 1.1079,
+      "learning_rate": 1.9894616097322877e-05,
+      "loss": 0.9433,
       "step": 2639
     },
     {
-      "epoch": 0.07250556150613825,
+      "epoch": 0.07491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905469327409777e-05,
-      "loss": 1.139,
+      "learning_rate": 1.9894482977208757e-05,
+      "loss": 0.9733,
       "step": 2640
     },
     {
-      "epoch": 0.07253302573398149,
+      "epoch": 0.07494324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905347269532758e-05,
-      "loss": 1.1578,
+      "learning_rate": 1.9894349773515493e-05,
+      "loss": 1.0098,
       "step": 2641
     },
     {
-      "epoch": 0.07256048996182472,
+      "epoch": 0.07497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905225133280792e-05,
-      "loss": 1.0741,
+      "learning_rate": 1.9894216486244214e-05,
+      "loss": 0.9911,
       "step": 2642
     },
     {
-      "epoch": 0.07258795418966796,
+      "epoch": 0.075,
       "grad_norm": 0.0,
-      "learning_rate": 1.9905102918654847e-05,
-      "loss": 1.1162,
+      "learning_rate": 1.9894083115396042e-05,
+      "loss": 1.1122,
       "step": 2643
     },
     {
-      "epoch": 0.0726154184175112,
+      "epoch": 0.0750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904980625655894e-05,
-      "loss": 1.0639,
+      "learning_rate": 1.9893949660972105e-05,
+      "loss": 1.0543,
       "step": 2644
     },
     {
-      "epoch": 0.07264288264535443,
+      "epoch": 0.07505675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904858254284896e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9893816122973534e-05,
+      "loss": 1.0416,
       "step": 2645
     },
     {
-      "epoch": 0.07267034687319766,
+      "epoch": 0.07508513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904735804542825e-05,
-      "loss": 1.0701,
+      "learning_rate": 1.9893682501401448e-05,
+      "loss": 1.0443,
       "step": 2646
     },
     {
-      "epoch": 0.0726978111010409,
+      "epoch": 0.07511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904613276430647e-05,
-      "loss": 1.0188,
+      "learning_rate": 1.9893548796256984e-05,
+      "loss": 1.0253,
       "step": 2647
     },
     {
-      "epoch": 0.07272527532888413,
+      "epoch": 0.07514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.990449066994933e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.989341500754127e-05,
+      "loss": 1.0598,
       "step": 2648
     },
     {
-      "epoch": 0.07275273955672737,
+      "epoch": 0.07517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904367985099848e-05,
-      "loss": 1.0373,
+      "learning_rate": 1.989328113525543e-05,
+      "loss": 1.0975,
       "step": 2649
     },
     {
-      "epoch": 0.07278020378457059,
+      "epoch": 0.07519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.990424522188317e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.9893147179400602e-05,
+      "loss": 1.106,
       "step": 2650
     },
     {
-      "epoch": 0.07280766801241383,
+      "epoch": 0.07522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.9904122380300265e-05,
-      "loss": 1.2004,
+      "learning_rate": 1.9893013139977917e-05,
+      "loss": 1.0759,
       "step": 2651
     },
     {
-      "epoch": 0.07283513224025706,
+      "epoch": 0.07525539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.990399946035211e-05,
-      "loss": 1.0863,
+      "learning_rate": 1.9892879016988506e-05,
+      "loss": 1.0382,
       "step": 2652
     },
     {
-      "epoch": 0.0728625964681003,
+      "epoch": 0.07528376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.990387646203967e-05,
-      "loss": 1.0552,
+      "learning_rate": 1.9892744810433498e-05,
+      "loss": 1.0516,
       "step": 2653
     },
     {
-      "epoch": 0.07289006069594353,
+      "epoch": 0.07531214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.990375338536393e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.989261052031403e-05,
+      "loss": 0.9838,
       "step": 2654
     },
     {
-      "epoch": 0.07291752492378677,
+      "epoch": 0.0753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.9903630230325852e-05,
-      "loss": 1.0873,
+      "learning_rate": 1.989247614663124e-05,
+      "loss": 1.0884,
       "step": 2655
     },
     {
-      "epoch": 0.07294498915163,
+      "epoch": 0.0753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9903506996926415e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.9892341689386253e-05,
+      "loss": 1.0046,
       "step": 2656
     },
     {
-      "epoch": 0.07297245337947324,
+      "epoch": 0.07539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.9903383685166598e-05,
-      "loss": 1.0798,
+      "learning_rate": 1.9892207148580216e-05,
+      "loss": 0.9816,
       "step": 2657
     },
     {
-      "epoch": 0.07299991760731647,
+      "epoch": 0.07542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.990326029504737e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.989207252421426e-05,
+      "loss": 1.0579,
       "step": 2658
     },
     {
-      "epoch": 0.07302738183515971,
+      "epoch": 0.07545402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 1.990313682656971e-05,
-      "loss": 1.017,
+      "learning_rate": 1.989193781628952e-05,
+      "loss": 1.0722,
       "step": 2659
     },
     {
-      "epoch": 0.07305484606300294,
+      "epoch": 0.07548240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9903013279734597e-05,
-      "loss": 1.0985,
+      "learning_rate": 1.989180302480714e-05,
+      "loss": 0.921,
       "step": 2660
     },
     {
-      "epoch": 0.07308231029084618,
+      "epoch": 0.07551078320090807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9902889654543008e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.9891668149768253e-05,
+      "loss": 1.052,
       "step": 2661
     },
     {
-      "epoch": 0.07310977451868941,
+      "epoch": 0.07553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.990276595099592e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.9891533191174002e-05,
+      "loss": 1.0237,
       "step": 2662
     },
     {
-      "epoch": 0.07313723874653263,
+      "epoch": 0.07556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.990264216909431e-05,
-      "loss": 1.025,
+      "learning_rate": 1.9891398149025522e-05,
+      "loss": 1.0414,
       "step": 2663
     },
     {
-      "epoch": 0.07316470297437587,
+      "epoch": 0.07559591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9902518308839157e-05,
-      "loss": 1.2172,
+      "learning_rate": 1.9891263023323963e-05,
+      "loss": 1.165,
       "step": 2664
     },
     {
-      "epoch": 0.0731921672022191,
+      "epoch": 0.07562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9902394370231448e-05,
-      "loss": 1.0619,
+      "learning_rate": 1.9891127814070454e-05,
+      "loss": 1.0303,
       "step": 2665
     },
     {
-      "epoch": 0.07321963143006234,
+      "epoch": 0.07565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.990227035327216e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.9890992521266147e-05,
+      "loss": 0.9995,
       "step": 2666
     },
     {
-      "epoch": 0.07324709565790558,
+      "epoch": 0.07568104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.990214625796227e-05,
-      "loss": 1.1074,
+      "learning_rate": 1.989085714491218e-05,
+      "loss": 1.0732,
       "step": 2667
     },
     {
-      "epoch": 0.07327455988574881,
+      "epoch": 0.0757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.990202208430276e-05,
-      "loss": 1.1417,
+      "learning_rate": 1.9890721685009703e-05,
+      "loss": 1.0743,
       "step": 2668
     },
     {
-      "epoch": 0.07330202411359205,
+      "epoch": 0.07573779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 1.990189783229462e-05,
-      "loss": 1.0985,
+      "learning_rate": 1.989058614155985e-05,
+      "loss": 1.0353,
       "step": 2669
     },
     {
-      "epoch": 0.07332948834143528,
+      "epoch": 0.0757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9901773501938825e-05,
-      "loss": 1.0984,
+      "learning_rate": 1.9890450514563776e-05,
+      "loss": 1.0167,
       "step": 2670
     },
     {
-      "epoch": 0.07335695256927852,
+      "epoch": 0.07579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9901649093236366e-05,
-      "loss": 1.1187,
+      "learning_rate": 1.989031480402262e-05,
+      "loss": 1.1784,
       "step": 2671
     },
     {
-      "epoch": 0.07338441679712175,
+      "epoch": 0.07582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.9901524606188224e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.9890179009937527e-05,
+      "loss": 1.0027,
       "step": 2672
     },
     {
-      "epoch": 0.07341188102496499,
+      "epoch": 0.07585130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.9901400040795386e-05,
-      "loss": 1.1503,
+      "learning_rate": 1.989004313230965e-05,
+      "loss": 1.0757,
       "step": 2673
     },
     {
-      "epoch": 0.07343934525280822,
+      "epoch": 0.07587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.990127539705883e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.9889907171140134e-05,
+      "loss": 1.0609,
       "step": 2674
     },
     {
-      "epoch": 0.07346680948065146,
+      "epoch": 0.07590805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.9901150674979548e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.9889771126430128e-05,
+      "loss": 0.9979,
       "step": 2675
     },
     {
-      "epoch": 0.07349427370849469,
+      "epoch": 0.07593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.990102587455853e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.988963499818078e-05,
+      "loss": 0.9901,
       "step": 2676
     },
     {
-      "epoch": 0.07352173793633791,
+      "epoch": 0.07596481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.9900900995796755e-05,
-      "loss": 1.023,
+      "learning_rate": 1.988949878639324e-05,
+      "loss": 1.045,
       "step": 2677
     },
     {
-      "epoch": 0.07354920216418115,
+      "epoch": 0.07599318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 1.990077603869522e-05,
-      "loss": 1.1438,
+      "learning_rate": 1.9889362491068658e-05,
+      "loss": 1.0239,
       "step": 2678
     },
     {
-      "epoch": 0.07357666639202438,
+      "epoch": 0.07602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.990065100325491e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.988922611220819e-05,
+      "loss": 1.011,
       "step": 2679
     },
     {
-      "epoch": 0.07360413061986762,
+      "epoch": 0.07604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.9900525889476815e-05,
-      "loss": 1.0694,
+      "learning_rate": 1.988908964981298e-05,
+      "loss": 1.1345,
       "step": 2680
     },
     {
-      "epoch": 0.07363159484771085,
+      "epoch": 0.07607832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 1.990040069736192e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.988895310388419e-05,
+      "loss": 1.1384,
       "step": 2681
     },
     {
-      "epoch": 0.07365905907555409,
+      "epoch": 0.0761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9900275426911224e-05,
-      "loss": 1.2064,
+      "learning_rate": 1.9888816474422966e-05,
+      "loss": 1.1786,
       "step": 2682
     },
     {
-      "epoch": 0.07368652330339732,
+      "epoch": 0.07613507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 1.9900150078125714e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.9888679761430463e-05,
+      "loss": 1.0778,
       "step": 2683
     },
     {
-      "epoch": 0.07371398753124056,
+      "epoch": 0.07616345062429059,
       "grad_norm": 0.0,
-      "learning_rate": 1.9900024651006377e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.988854296490784e-05,
+      "loss": 1.0721,
       "step": 2684
     },
     {
-      "epoch": 0.0737414517590838,
+      "epoch": 0.07619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899899145554212e-05,
-      "loss": 1.1365,
+      "learning_rate": 1.9888406084856253e-05,
+      "loss": 1.0942,
       "step": 2685
     },
     {
-      "epoch": 0.07376891598692703,
+      "epoch": 0.07622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.989977356177021e-05,
-      "loss": 1.1346,
+      "learning_rate": 1.9888269121276852e-05,
+      "loss": 1.0947,
       "step": 2686
     },
     {
-      "epoch": 0.07379638021477027,
+      "epoch": 0.07624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899647899655367e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.9888132074170797e-05,
+      "loss": 1.0805,
       "step": 2687
     },
     {
-      "epoch": 0.0738238444426135,
+      "epoch": 0.07627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899522159210673e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.9887994943539247e-05,
+      "loss": 1.0779,
       "step": 2688
     },
     {
-      "epoch": 0.07385130867045674,
+      "epoch": 0.07630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899396340437124e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.9887857729383356e-05,
+      "loss": 1.1391,
       "step": 2689
     },
     {
-      "epoch": 0.07387877289829996,
+      "epoch": 0.07633371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.989927044333572e-05,
-      "loss": 1.1094,
+      "learning_rate": 1.9887720431704288e-05,
+      "loss": 1.0648,
       "step": 2690
     },
     {
-      "epoch": 0.07390623712614319,
+      "epoch": 0.07636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899144467907452e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9887583050503204e-05,
+      "loss": 1.0063,
       "step": 2691
     },
     {
-      "epoch": 0.07393370135398643,
+      "epoch": 0.07639046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 1.9899018414153317e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.988744558578126e-05,
+      "loss": 1.0508,
       "step": 2692
     },
     {
-      "epoch": 0.07396116558182966,
+      "epoch": 0.07641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.9898892282074316e-05,
-      "loss": 1.0011,
+      "learning_rate": 1.9887308037539616e-05,
+      "loss": 0.9489,
       "step": 2693
     },
     {
-      "epoch": 0.0739886298096729,
+      "epoch": 0.0764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9898766071671442e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.988717040577944e-05,
+      "loss": 1.2203,
       "step": 2694
     },
     {
-      "epoch": 0.07401609403751613,
+      "epoch": 0.07647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.98986397829457e-05,
-      "loss": 1.101,
+      "learning_rate": 1.9887032690501887e-05,
+      "loss": 1.0726,
       "step": 2695
     },
     {
-      "epoch": 0.07404355826535937,
+      "epoch": 0.07650397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 1.989851341589808e-05,
-      "loss": 1.1004,
+      "learning_rate": 1.988689489170813e-05,
+      "loss": 1.0831,
       "step": 2696
     },
     {
-      "epoch": 0.0740710224932026,
+      "epoch": 0.07653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.9898386970529593e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.988675700939932e-05,
+      "loss": 0.9997,
       "step": 2697
     },
     {
-      "epoch": 0.07409848672104584,
+      "epoch": 0.07656072644721906,
       "grad_norm": 0.0,
-      "learning_rate": 1.9898260446841233e-05,
-      "loss": 1.1217,
+      "learning_rate": 1.9886619043576636e-05,
+      "loss": 1.0041,
       "step": 2698
     },
     {
-      "epoch": 0.07412595094888907,
+      "epoch": 0.07658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9898133844834e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.9886480994241235e-05,
+      "loss": 0.9855,
       "step": 2699
     },
     {
-      "epoch": 0.07415341517673231,
+      "epoch": 0.07661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.98980071645089e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.988634286139428e-05,
+      "loss": 1.0165,
       "step": 2700
     },
     {
-      "epoch": 0.07418087940457554,
+      "epoch": 0.07664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.9897880405866933e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.988620464503695e-05,
+      "loss": 1.0619,
       "step": 2701
     },
     {
-      "epoch": 0.07420834363241878,
+      "epoch": 0.07667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.9897753568909102e-05,
-      "loss": 1.1337,
+      "learning_rate": 1.9886066345170398e-05,
+      "loss": 1.1176,
       "step": 2702
     },
     {
-      "epoch": 0.07423580786026202,
+      "epoch": 0.07670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.989762665363641e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.9885927961795804e-05,
+      "loss": 1.1003,
       "step": 2703
     },
     {
-      "epoch": 0.07426327208810524,
+      "epoch": 0.07673098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.9897499660049863e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.988578949491433e-05,
+      "loss": 1.0561,
       "step": 2704
     },
     {
-      "epoch": 0.07429073631594847,
+      "epoch": 0.07675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 1.9897372588150463e-05,
-      "loss": 1.1437,
+      "learning_rate": 1.9885650944527146e-05,
+      "loss": 0.9846,
       "step": 2705
     },
     {
-      "epoch": 0.07431820054379171,
+      "epoch": 0.0767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.989724543793922e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.988551231063543e-05,
+      "loss": 0.9136,
       "step": 2706
     },
     {
-      "epoch": 0.07434566477163494,
+      "epoch": 0.0768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.989711820941713e-05,
-      "loss": 1.0598,
+      "learning_rate": 1.9885373593240342e-05,
+      "loss": 1.045,
       "step": 2707
     },
     {
-      "epoch": 0.07437312899947818,
+      "epoch": 0.07684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.9896990902585216e-05,
-      "loss": 1.1427,
+      "learning_rate": 1.988523479234306e-05,
+      "loss": 0.9653,
       "step": 2708
     },
     {
-      "epoch": 0.07440059322732141,
+      "epoch": 0.07687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.9896863517444473e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.9885095907944754e-05,
+      "loss": 1.0573,
       "step": 2709
     },
     {
-      "epoch": 0.07442805745516465,
+      "epoch": 0.07690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.989673605399591e-05,
-      "loss": 1.0492,
+      "learning_rate": 1.98849569400466e-05,
+      "loss": 1.0053,
       "step": 2710
     },
     {
-      "epoch": 0.07445552168300788,
+      "epoch": 0.07692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.9896608512240543e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.988481788864977e-05,
+      "loss": 1.0949,
       "step": 2711
     },
     {
-      "epoch": 0.07448298591085112,
+      "epoch": 0.07695800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.989648089217937e-05,
-      "loss": 1.0473,
+      "learning_rate": 1.988467875375544e-05,
+      "loss": 1.0005,
       "step": 2712
     },
     {
-      "epoch": 0.07451045013869435,
+      "epoch": 0.07698637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.989635319381341e-05,
-      "loss": 1.1303,
+      "learning_rate": 1.988453953536478e-05,
+      "loss": 1.06,
       "step": 2713
     },
     {
-      "epoch": 0.07453791436653759,
+      "epoch": 0.07701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.9896225417143668e-05,
-      "loss": 1.0895,
+      "learning_rate": 1.9884400233478976e-05,
+      "loss": 1.024,
       "step": 2714
     },
     {
-      "epoch": 0.07456537859438082,
+      "epoch": 0.07704313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 1.989609756217116e-05,
-      "loss": 1.1898,
+      "learning_rate": 1.98842608480992e-05,
+      "loss": 1.037,
       "step": 2715
     },
     {
-      "epoch": 0.07459284282222406,
+      "epoch": 0.07707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.989596962889689e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.988412137922662e-05,
+      "loss": 1.1903,
       "step": 2716
     },
     {
-      "epoch": 0.07462030705006728,
+      "epoch": 0.07709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.989584161732188e-05,
-      "loss": 1.12,
+      "learning_rate": 1.988398182686243e-05,
+      "loss": 1.0625,
       "step": 2717
     },
     {
-      "epoch": 0.07464777127791052,
+      "epoch": 0.0771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.9895713527447137e-05,
-      "loss": 1.1631,
+      "learning_rate": 1.9883842191007796e-05,
+      "loss": 1.069,
       "step": 2718
     },
     {
-      "epoch": 0.07467523550575375,
+      "epoch": 0.07715664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 1.9895585359273672e-05,
-      "loss": 1.1077,
+      "learning_rate": 1.9883702471663906e-05,
+      "loss": 0.9618,
       "step": 2719
     },
     {
-      "epoch": 0.07470269973359699,
+      "epoch": 0.0771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.9895457112802507e-05,
-      "loss": 1.1702,
+      "learning_rate": 1.9883562668831936e-05,
+      "loss": 1.033,
       "step": 2720
     },
     {
-      "epoch": 0.07473016396144022,
+      "epoch": 0.0772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.989532878803465e-05,
-      "loss": 1.2433,
+      "learning_rate": 1.988342278251307e-05,
+      "loss": 1.1261,
       "step": 2721
     },
     {
-      "epoch": 0.07475762818928346,
+      "epoch": 0.07724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9895200384971117e-05,
-      "loss": 1.1076,
+      "learning_rate": 1.9883282812708483e-05,
+      "loss": 1.0018,
       "step": 2722
     },
     {
-      "epoch": 0.07478509241712669,
+      "epoch": 0.07727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.9895071903612925e-05,
-      "loss": 1.1314,
+      "learning_rate": 1.9883142759419364e-05,
+      "loss": 1.1218,
       "step": 2723
     },
     {
-      "epoch": 0.07481255664496993,
+      "epoch": 0.07729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.9894943343961092e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.9883002622646894e-05,
+      "loss": 1.0644,
       "step": 2724
     },
     {
-      "epoch": 0.07484002087281316,
+      "epoch": 0.07732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.9894814706016635e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.9882862402392256e-05,
+      "loss": 1.0234,
       "step": 2725
     },
     {
-      "epoch": 0.0748674851006564,
+      "epoch": 0.07735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9894685989780574e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.9882722098656637e-05,
+      "loss": 0.9679,
       "step": 2726
     },
     {
-      "epoch": 0.07489494932849963,
+      "epoch": 0.07738365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 1.9894557195253922e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.9882581711441217e-05,
+      "loss": 1.1383,
       "step": 2727
     },
     {
-      "epoch": 0.07492241355634287,
+      "epoch": 0.07741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.98944283224377e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.9882441240747184e-05,
+      "loss": 1.1344,
       "step": 2728
     },
     {
-      "epoch": 0.0749498777841861,
+      "epoch": 0.07744040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 1.989429937133293e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.988230068657573e-05,
+      "loss": 0.9771,
       "step": 2729
     },
     {
-      "epoch": 0.07497734201202934,
+      "epoch": 0.07746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.989417034194063e-05,
-      "loss": 0.9747,
+      "learning_rate": 1.9882160048928036e-05,
+      "loss": 1.0359,
       "step": 2730
     },
     {
-      "epoch": 0.07500480623987256,
+      "epoch": 0.07749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.989404123426182e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.9882019327805294e-05,
+      "loss": 1.0734,
       "step": 2731
     },
     {
-      "epoch": 0.0750322704677158,
+      "epoch": 0.0775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893912048297523e-05,
-      "loss": 1.0334,
+      "learning_rate": 1.9881878523208687e-05,
+      "loss": 1.1139,
       "step": 2732
     },
     {
-      "epoch": 0.07505973469555903,
+      "epoch": 0.0775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893782784048766e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.988173763513941e-05,
+      "loss": 1.0589,
       "step": 2733
     },
     {
-      "epoch": 0.07508719892340227,
+      "epoch": 0.07758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.989365344151656e-05,
-      "loss": 1.0459,
+      "learning_rate": 1.988159666359865e-05,
+      "loss": 1.0533,
       "step": 2734
     },
     {
-      "epoch": 0.0751146631512455,
+      "epoch": 0.07761066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893524020701943e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.98814556085876e-05,
+      "loss": 1.0115,
       "step": 2735
     },
     {
-      "epoch": 0.07514212737908874,
+      "epoch": 0.07763904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 1.989339452160593e-05,
-      "loss": 1.1158,
+      "learning_rate": 1.988131447010745e-05,
+      "loss": 1.1069,
       "step": 2736
     },
     {
-      "epoch": 0.07516959160693197,
+      "epoch": 0.07766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893264944229544e-05,
-      "loss": 1.0409,
+      "learning_rate": 1.988117324815939e-05,
+      "loss": 1.1563,
       "step": 2737
     },
     {
-      "epoch": 0.0751970558347752,
+      "epoch": 0.07769580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893135288573812e-05,
-      "loss": 1.1351,
+      "learning_rate": 1.9881031942744617e-05,
+      "loss": 1.0208,
       "step": 2738
     },
     {
-      "epoch": 0.07522452006261844,
+      "epoch": 0.07772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9893005554639764e-05,
-      "loss": 1.085,
+      "learning_rate": 1.9880890553864325e-05,
+      "loss": 1.0632,
       "step": 2739
     },
     {
-      "epoch": 0.07525198429046168,
+      "epoch": 0.07775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9892875742428426e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.9880749081519703e-05,
+      "loss": 1.061,
       "step": 2740
     },
     {
-      "epoch": 0.07527944851830491,
+      "epoch": 0.07778093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.989274585194082e-05,
-      "loss": 1.1304,
+      "learning_rate": 1.9880607525711954e-05,
+      "loss": 1.0445,
       "step": 2741
     },
     {
-      "epoch": 0.07530691274614815,
+      "epoch": 0.07780930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 1.989261588317798e-05,
-      "loss": 1.0138,
+      "learning_rate": 1.9880465886442264e-05,
+      "loss": 1.1056,
       "step": 2742
     },
     {
-      "epoch": 0.07533437697399138,
+      "epoch": 0.07783768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 1.9892485836140926e-05,
-      "loss": 1.1705,
+      "learning_rate": 1.9880324163711834e-05,
+      "loss": 1.0189,
       "step": 2743
     },
     {
-      "epoch": 0.07536184120183462,
+      "epoch": 0.07786606129398412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9892355710830696e-05,
-      "loss": 1.1455,
+      "learning_rate": 1.9880182357521867e-05,
+      "loss": 1.1119,
       "step": 2744
     },
     {
-      "epoch": 0.07538930542967784,
+      "epoch": 0.077894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9892225507248314e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.988004046787355e-05,
+      "loss": 1.049,
       "step": 2745
     },
     {
-      "epoch": 0.07541676965752107,
+      "epoch": 0.07792281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.9892095225394815e-05,
-      "loss": 1.0998,
+      "learning_rate": 1.9879898494768093e-05,
+      "loss": 1.0715,
       "step": 2746
     },
     {
-      "epoch": 0.07544423388536431,
+      "epoch": 0.07795119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.9891964865271224e-05,
-      "loss": 1.0906,
+      "learning_rate": 1.9879756438206687e-05,
+      "loss": 0.9899,
       "step": 2747
     },
     {
-      "epoch": 0.07547169811320754,
+      "epoch": 0.07797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.9891834426878572e-05,
-      "loss": 1.1346,
+      "learning_rate": 1.987961429819053e-05,
+      "loss": 1.0576,
       "step": 2748
     },
     {
-      "epoch": 0.07549916234105078,
+      "epoch": 0.07800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.9891703910217898e-05,
-      "loss": 1.148,
+      "learning_rate": 1.987947207472083e-05,
+      "loss": 0.8771,
       "step": 2749
     },
     {
-      "epoch": 0.07552662656889401,
+      "epoch": 0.07803632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.989157331529023e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.987932976779879e-05,
+      "loss": 0.9766,
       "step": 2750
     },
     {
-      "epoch": 0.07555409079673725,
+      "epoch": 0.07806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9891442642096604e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.9879187377425598e-05,
+      "loss": 1.0651,
       "step": 2751
     },
     {
-      "epoch": 0.07558155502458049,
+      "epoch": 0.07809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.989131189063805e-05,
-      "loss": 1.049,
+      "learning_rate": 1.987904490360247e-05,
+      "loss": 1.0876,
       "step": 2752
     },
     {
-      "epoch": 0.07560901925242372,
+      "epoch": 0.07812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.989118106091561e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.987890234633061e-05,
+      "loss": 1.1172,
       "step": 2753
     },
     {
-      "epoch": 0.07563648348026696,
+      "epoch": 0.07814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.9891050152930308e-05,
-      "loss": 1.1428,
+      "learning_rate": 1.9878759705611213e-05,
+      "loss": 1.0282,
       "step": 2754
     },
     {
-      "epoch": 0.07566394770811019,
+      "epoch": 0.07817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.989091916668319e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.987861698144549e-05,
+      "loss": 0.926,
       "step": 2755
     },
     {
-      "epoch": 0.07569141193595343,
+      "epoch": 0.07820658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890788102175284e-05,
-      "loss": 1.1256,
+      "learning_rate": 1.9878474173834642e-05,
+      "loss": 0.9254,
       "step": 2756
     },
     {
-      "epoch": 0.07571887616379666,
+      "epoch": 0.0782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890656959407635e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.9878331282779883e-05,
+      "loss": 0.9851,
       "step": 2757
     },
     {
-      "epoch": 0.07574634039163988,
+      "epoch": 0.07826333711691259,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890525738381274e-05,
-      "loss": 1.1396,
+      "learning_rate": 1.9878188308282414e-05,
+      "loss": 1.0843,
       "step": 2758
     },
     {
-      "epoch": 0.07577380461948312,
+      "epoch": 0.0782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890394439097244e-05,
-      "loss": 1.0466,
+      "learning_rate": 1.9878045250343445e-05,
+      "loss": 1.064,
       "step": 2759
     },
     {
-      "epoch": 0.07580126884732635,
+      "epoch": 0.07832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890263061556578e-05,
-      "loss": 1.1848,
+      "learning_rate": 1.9877902108964182e-05,
+      "loss": 1.0652,
       "step": 2760
     },
     {
-      "epoch": 0.07582873307516959,
+      "epoch": 0.07834846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890131605760324e-05,
-      "loss": 1.1213,
+      "learning_rate": 1.9877758884145838e-05,
+      "loss": 1.0373,
       "step": 2761
     },
     {
-      "epoch": 0.07585619730301282,
+      "epoch": 0.07837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.989000007170952e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.9877615575889622e-05,
+      "loss": 1.0829,
       "step": 2762
     },
     {
-      "epoch": 0.07588366153085606,
+      "epoch": 0.07840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.98898684594052e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.987747218419674e-05,
+      "loss": 1.1512,
       "step": 2763
     },
     {
-      "epoch": 0.0759111257586993,
+      "epoch": 0.07843359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.988973676884841e-05,
-      "loss": 1.174,
+      "learning_rate": 1.9877328709068407e-05,
+      "loss": 1.0511,
       "step": 2764
     },
     {
-      "epoch": 0.07593858998654253,
+      "epoch": 0.07846197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.988960500004019e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.9877185150505834e-05,
+      "loss": 0.9342,
       "step": 2765
     },
     {
-      "epoch": 0.07596605421438576,
+      "epoch": 0.07849035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 1.988947315298159e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.9877041508510238e-05,
+      "loss": 1.1913,
       "step": 2766
     },
     {
-      "epoch": 0.075993518442229,
+      "epoch": 0.07851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9889341227673642e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.9876897783082825e-05,
+      "loss": 1.1114,
       "step": 2767
     },
     {
-      "epoch": 0.07602098267007223,
+      "epoch": 0.07854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9889209224117397e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.987675397422481e-05,
+      "loss": 1.0608,
       "step": 2768
     },
     {
-      "epoch": 0.07604844689791547,
+      "epoch": 0.07857548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.9889077142313898e-05,
-      "loss": 1.0559,
+      "learning_rate": 1.9876610081937416e-05,
+      "loss": 1.0657,
       "step": 2769
     },
     {
-      "epoch": 0.0760759111257587,
+      "epoch": 0.07860385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888944982264192e-05,
-      "loss": 1.084,
+      "learning_rate": 1.9876466106221847e-05,
+      "loss": 1.2003,
       "step": 2770
     },
     {
-      "epoch": 0.07610337535360194,
+      "epoch": 0.0786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.988881274396932e-05,
-      "loss": 1.0812,
+      "learning_rate": 1.9876322047079326e-05,
+      "loss": 1.1002,
       "step": 2771
     },
     {
-      "epoch": 0.07613083958144516,
+      "epoch": 0.07866061293984108,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888680427430335e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.987617790451107e-05,
+      "loss": 1.0434,
       "step": 2772
     },
     {
-      "epoch": 0.0761583038092884,
+      "epoch": 0.07868898978433599,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888548032648278e-05,
-      "loss": 1.0336,
+      "learning_rate": 1.9876033678518295e-05,
+      "loss": 1.0399,
       "step": 2773
     },
     {
-      "epoch": 0.07618576803713163,
+      "epoch": 0.07871736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888415559624197e-05,
-      "loss": 1.124,
+      "learning_rate": 1.987588936910222e-05,
+      "loss": 0.976,
       "step": 2774
     },
     {
-      "epoch": 0.07621323226497487,
+      "epoch": 0.07874574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888283008359142e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.9875744976264063e-05,
+      "loss": 1.1287,
       "step": 2775
     },
     {
-      "epoch": 0.0762406964928181,
+      "epoch": 0.07877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.988815037885416e-05,
-      "loss": 1.038,
+      "learning_rate": 1.9875600500005046e-05,
+      "loss": 1.0782,
       "step": 2776
     },
     {
-      "epoch": 0.07626816072066134,
+      "epoch": 0.07880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9888017671110306e-05,
-      "loss": 1.2093,
+      "learning_rate": 1.9875455940326386e-05,
+      "loss": 1.0809,
       "step": 2777
     },
     {
-      "epoch": 0.07629562494850457,
+      "epoch": 0.07883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887884885128623e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.987531129722931e-05,
+      "loss": 1.1218,
       "step": 2778
     },
     {
-      "epoch": 0.07632308917634781,
+      "epoch": 0.07885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887752020910165e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.987516657071503e-05,
+      "loss": 1.126,
       "step": 2779
     },
     {
-      "epoch": 0.07635055340419104,
+      "epoch": 0.07888762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887619078455984e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.9875021760784776e-05,
+      "loss": 1.1321,
       "step": 2780
     },
     {
-      "epoch": 0.07637801763203428,
+      "epoch": 0.07891600454029513,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887486057767133e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.9874876867439766e-05,
+      "loss": 1.0215,
       "step": 2781
     },
     {
-      "epoch": 0.07640548185987751,
+      "epoch": 0.07894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887352958844658e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.9874731890681232e-05,
+      "loss": 0.9843,
       "step": 2782
     },
     {
-      "epoch": 0.07643294608772075,
+      "epoch": 0.0789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887219781689614e-05,
-      "loss": 1.001,
+      "learning_rate": 1.9874586830510393e-05,
+      "loss": 0.9521,
       "step": 2783
     },
     {
-      "epoch": 0.07646041031556398,
+      "epoch": 0.0790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9887086526303064e-05,
-      "loss": 1.2277,
+      "learning_rate": 1.9874441686928473e-05,
+      "loss": 1.0736,
       "step": 2784
     },
     {
-      "epoch": 0.0764878745434072,
+      "epoch": 0.07902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.9886953192686053e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.98742964599367e-05,
+      "loss": 1.0786,
       "step": 2785
     },
     {
-      "epoch": 0.07651533877125044,
+      "epoch": 0.07905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9886819780839636e-05,
-      "loss": 1.1455,
+      "learning_rate": 1.9874151149536304e-05,
+      "loss": 1.0254,
       "step": 2786
     },
     {
-      "epoch": 0.07654280299909368,
+      "epoch": 0.07908626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 1.988668629076487e-05,
-      "loss": 1.078,
+      "learning_rate": 1.9874005755728505e-05,
+      "loss": 1.0779,
       "step": 2787
     },
     {
-      "epoch": 0.07657026722693691,
+      "epoch": 0.07911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.9886552722462816e-05,
-      "loss": 1.037,
+      "learning_rate": 1.987386027851454e-05,
+      "loss": 1.029,
       "step": 2788
     },
     {
-      "epoch": 0.07659773145478015,
+      "epoch": 0.07914301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.988641907593453e-05,
-      "loss": 1.1508,
+      "learning_rate": 1.987371471789563e-05,
+      "loss": 1.1727,
       "step": 2789
     },
     {
-      "epoch": 0.07662519568262338,
+      "epoch": 0.07917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.988628535118106e-05,
-      "loss": 1.168,
+      "learning_rate": 1.987356907387301e-05,
+      "loss": 1.0892,
       "step": 2790
     },
     {
-      "epoch": 0.07665265991046662,
+      "epoch": 0.07919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9886151548203476e-05,
-      "loss": 1.2169,
+      "learning_rate": 1.9873423346447905e-05,
+      "loss": 0.9356,
       "step": 2791
     },
     {
-      "epoch": 0.07668012413830985,
+      "epoch": 0.07922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.988601766700283e-05,
-      "loss": 1.1651,
+      "learning_rate": 1.9873277535621557e-05,
+      "loss": 1.0252,
       "step": 2792
     },
     {
-      "epoch": 0.07670758836615309,
+      "epoch": 0.07925652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.9885883707580183e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.9873131641395183e-05,
+      "loss": 1.003,
       "step": 2793
     },
     {
-      "epoch": 0.07673505259399632,
+      "epoch": 0.07928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.9885749669936592e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9872985663770024e-05,
+      "loss": 1.067,
       "step": 2794
     },
     {
-      "epoch": 0.07676251682183956,
+      "epoch": 0.0793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.988561555407312e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.9872839602747308e-05,
+      "loss": 1.0594,
       "step": 2795
     },
     {
-      "epoch": 0.07678998104968279,
+      "epoch": 0.0793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.988548135999083e-05,
-      "loss": 1.0726,
+      "learning_rate": 1.987269345832828e-05,
+      "loss": 1.046,
       "step": 2796
     },
     {
-      "epoch": 0.07681744527752603,
+      "epoch": 0.07937003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 1.9885347087690783e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.987254723051416e-05,
+      "loss": 1.0573,
       "step": 2797
     },
     {
-      "epoch": 0.07684490950536926,
+      "epoch": 0.07939841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 1.988521273717404e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.987240091930619e-05,
+      "loss": 0.9778,
       "step": 2798
     },
     {
-      "epoch": 0.07687237373321248,
+      "epoch": 0.07942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.9885078308441668e-05,
-      "loss": 1.1298,
+      "learning_rate": 1.9872254524705608e-05,
+      "loss": 1.087,
       "step": 2799
     },
     {
-      "epoch": 0.07689983796105572,
+      "epoch": 0.07945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9884943801494722e-05,
-      "loss": 1.1115,
+      "learning_rate": 1.9872108046713644e-05,
+      "loss": 1.0157,
       "step": 2800
     },
     {
-      "epoch": 0.07692730218889895,
+      "epoch": 0.07948354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.9884809216334276e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.9871961485331544e-05,
+      "loss": 1.0942,
       "step": 2801
     },
     {
-      "epoch": 0.07695476641674219,
+      "epoch": 0.07951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.988467455296139e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.987181484056054e-05,
+      "loss": 1.0822,
       "step": 2802
     },
     {
-      "epoch": 0.07698223064458543,
+      "epoch": 0.07954029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9884539811377125e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.9871668112401872e-05,
+      "loss": 0.9392,
       "step": 2803
     },
     {
-      "epoch": 0.07700969487242866,
+      "epoch": 0.07956867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9884404991582555e-05,
-      "loss": 1.0614,
+      "learning_rate": 1.987152130085678e-05,
+      "loss": 1.0742,
       "step": 2804
     },
     {
-      "epoch": 0.0770371591002719,
+      "epoch": 0.07959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.9884270093578743e-05,
-      "loss": 1.1203,
+      "learning_rate": 1.9871374405926506e-05,
+      "loss": 1.0646,
       "step": 2805
     },
     {
-      "epoch": 0.07706462332811513,
+      "epoch": 0.07962542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.988413511736676e-05,
-      "loss": 1.1566,
+      "learning_rate": 1.9871227427612285e-05,
+      "loss": 0.9622,
       "step": 2806
     },
     {
-      "epoch": 0.07709208755595837,
+      "epoch": 0.07965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.988400006294767e-05,
-      "loss": 1.1501,
+      "learning_rate": 1.9871080365915366e-05,
+      "loss": 0.9899,
       "step": 2807
     },
     {
-      "epoch": 0.0771195517838016,
+      "epoch": 0.07968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.988386493032254e-05,
-      "loss": 1.0793,
+      "learning_rate": 1.9870933220836983e-05,
+      "loss": 1.0186,
       "step": 2808
     },
     {
-      "epoch": 0.07714701601164484,
+      "epoch": 0.0797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9883729719492445e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.987078599237839e-05,
+      "loss": 1.0703,
       "step": 2809
     },
     {
-      "epoch": 0.07717448023948807,
+      "epoch": 0.079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.988359443045845e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.987063868054082e-05,
+      "loss": 1.1122,
       "step": 2810
     },
     {
-      "epoch": 0.07720194446733131,
+      "epoch": 0.07976730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.988345906322163e-05,
-      "loss": 1.1086,
+      "learning_rate": 1.9870491285325516e-05,
+      "loss": 1.0978,
       "step": 2811
     },
     {
-      "epoch": 0.07722940869517453,
+      "epoch": 0.07979568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 1.9883323617783053e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9870343806733735e-05,
+      "loss": 1.02,
       "step": 2812
     },
     {
-      "epoch": 0.07725687292301776,
+      "epoch": 0.07982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.988318809414379e-05,
-      "loss": 1.1568,
+      "learning_rate": 1.9870196244766715e-05,
+      "loss": 1.1178,
       "step": 2813
     },
     {
-      "epoch": 0.077284337150861,
+      "epoch": 0.07985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9883052492304915e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.9870048599425704e-05,
+      "loss": 1.0224,
       "step": 2814
     },
     {
-      "epoch": 0.07731180137870423,
+      "epoch": 0.07988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.98829168122675e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.9869900870711947e-05,
+      "loss": 1.0956,
       "step": 2815
     },
     {
-      "epoch": 0.07733926560654747,
+      "epoch": 0.07990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.9882781054032615e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.9869753058626696e-05,
+      "loss": 0.9984,
       "step": 2816
     },
     {
-      "epoch": 0.0773667298343907,
+      "epoch": 0.07993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.9882645217601343e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.9869605163171194e-05,
+      "loss": 1.0709,
       "step": 2817
     },
     {
-      "epoch": 0.07739419406223394,
+      "epoch": 0.07996594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 1.9882509302974754e-05,
-      "loss": 1.1915,
+      "learning_rate": 1.9869457184346698e-05,
+      "loss": 1.1053,
       "step": 2818
     },
     {
-      "epoch": 0.07742165829007717,
+      "epoch": 0.07999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.988237331015392e-05,
-      "loss": 1.019,
+      "learning_rate": 1.9869309122154447e-05,
+      "loss": 1.038,
       "step": 2819
     },
     {
-      "epoch": 0.07744912251792041,
+      "epoch": 0.08002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.9882237239139924e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.9869160976595703e-05,
+      "loss": 1.0919,
       "step": 2820
     },
     {
-      "epoch": 0.07747658674576365,
+      "epoch": 0.08005107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 1.988210108993384e-05,
-      "loss": 1.0666,
+      "learning_rate": 1.9869012747671715e-05,
+      "loss": 1.0854,
       "step": 2821
     },
     {
-      "epoch": 0.07750405097360688,
+      "epoch": 0.0800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.988196486253674e-05,
-      "loss": 1.1533,
+      "learning_rate": 1.9868864435383724e-05,
+      "loss": 1.124,
       "step": 2822
     },
     {
-      "epoch": 0.07753151520145012,
+      "epoch": 0.08010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881828556949706e-05,
-      "loss": 1.1993,
+      "learning_rate": 1.9868716039733e-05,
+      "loss": 1.0162,
       "step": 2823
     },
     {
-      "epoch": 0.07755897942929335,
+      "epoch": 0.08013620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 1.988169217317382e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.9868567560720783e-05,
+      "loss": 1.0468,
       "step": 2824
     },
     {
-      "epoch": 0.07758644365713659,
+      "epoch": 0.08016458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881555711210153e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.9868418998348336e-05,
+      "loss": 1.0735,
       "step": 2825
     },
     {
-      "epoch": 0.07761390788497981,
+      "epoch": 0.08019296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881419171059796e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.986827035261691e-05,
+      "loss": 1.0823,
       "step": 2826
     },
     {
-      "epoch": 0.07764137211282304,
+      "epoch": 0.08022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881282552723818e-05,
-      "loss": 1.1937,
+      "learning_rate": 1.9868121623527757e-05,
+      "loss": 1.0967,
       "step": 2827
     },
     {
-      "epoch": 0.07766883634066628,
+      "epoch": 0.08024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881145856203307e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.986797281108214e-05,
+      "loss": 1.0806,
       "step": 2828
     },
     {
-      "epoch": 0.07769630056850951,
+      "epoch": 0.08027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881009081499338e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9867823915281312e-05,
+      "loss": 1.0942,
       "step": 2829
     },
     {
-      "epoch": 0.07772376479635275,
+      "epoch": 0.08030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880872228613005e-05,
-      "loss": 1.1,
+      "learning_rate": 1.986767493612653e-05,
+      "loss": 0.9627,
       "step": 2830
     },
     {
-      "epoch": 0.07775122902419598,
+      "epoch": 0.08033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880735297545378e-05,
-      "loss": 1.199,
+      "learning_rate": 1.9867525873619056e-05,
+      "loss": 1.1732,
       "step": 2831
     },
     {
-      "epoch": 0.07777869325203922,
+      "epoch": 0.08036322360953461,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880598288297545e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.9867376727760147e-05,
+      "loss": 1.0535,
       "step": 2832
     },
     {
-      "epoch": 0.07780615747988245,
+      "epoch": 0.08039160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880461200870594e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.9867227498551066e-05,
+      "loss": 1.0287,
       "step": 2833
     },
     {
-      "epoch": 0.07783362170772569,
+      "epoch": 0.0804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.988032403526561e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.9867078185993068e-05,
+      "loss": 0.9642,
       "step": 2834
     },
     {
-      "epoch": 0.07786108593556892,
+      "epoch": 0.08044835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880186791483668e-05,
-      "loss": 1.0537,
+      "learning_rate": 1.9866928790087418e-05,
+      "loss": 1.0695,
       "step": 2835
     },
     {
-      "epoch": 0.07788855016341216,
+      "epoch": 0.08047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.9880049469525863e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.9866779310835372e-05,
+      "loss": 1.0102,
       "step": 2836
     },
     {
-      "epoch": 0.0779160143912554,
+      "epoch": 0.08050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.987991206939328e-05,
-      "loss": 1.0925,
+      "learning_rate": 1.9866629748238206e-05,
+      "loss": 1.0191,
       "step": 2837
     },
     {
-      "epoch": 0.07794347861909863,
+      "epoch": 0.08053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.9879774591087005e-05,
-      "loss": 1.2143,
+      "learning_rate": 1.986648010229717e-05,
+      "loss": 0.9096,
       "step": 2838
     },
     {
-      "epoch": 0.07797094284694185,
+      "epoch": 0.08056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.987963703460813e-05,
-      "loss": 1.1472,
+      "learning_rate": 1.9866330373013532e-05,
+      "loss": 1.1255,
       "step": 2839
     },
     {
-      "epoch": 0.07799840707478509,
+      "epoch": 0.08059023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.9879499399957735e-05,
-      "loss": 1.0752,
+      "learning_rate": 1.986618056038856e-05,
+      "loss": 0.9983,
       "step": 2840
     },
     {
-      "epoch": 0.07802587130262832,
+      "epoch": 0.08061861520998866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9879361687136916e-05,
-      "loss": 1.109,
+      "learning_rate": 1.9866030664423518e-05,
+      "loss": 1.1166,
       "step": 2841
     },
     {
-      "epoch": 0.07805333553047156,
+      "epoch": 0.08064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.987922389614676e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.9865880685119673e-05,
+      "loss": 1.0067,
       "step": 2842
     },
     {
-      "epoch": 0.07808079975831479,
+      "epoch": 0.08067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.987908602698836e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.9865730622478286e-05,
+      "loss": 1.0742,
       "step": 2843
     },
     {
-      "epoch": 0.07810826398615803,
+      "epoch": 0.08070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.98789480796628e-05,
-      "loss": 1.098,
+      "learning_rate": 1.986558047650063e-05,
+      "loss": 1.0395,
       "step": 2844
     },
     {
-      "epoch": 0.07813572821400126,
+      "epoch": 0.08073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.987881005417118e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.986543024718797e-05,
+      "loss": 1.0714,
       "step": 2845
     },
     {
-      "epoch": 0.0781631924418445,
+      "epoch": 0.0807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.9878671950514587e-05,
-      "loss": 1.0921,
+      "learning_rate": 1.9865279934541585e-05,
+      "loss": 1.0587,
       "step": 2846
     },
     {
-      "epoch": 0.07819065666968773,
+      "epoch": 0.08078887627695801,
       "grad_norm": 0.0,
-      "learning_rate": 1.9878533768694112e-05,
-      "loss": 1.0904,
+      "learning_rate": 1.986512953856273e-05,
+      "loss": 1.0413,
       "step": 2847
     },
     {
-      "epoch": 0.07821812089753097,
+      "epoch": 0.0808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.9878395508710853e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.9864979059252685e-05,
+      "loss": 1.0122,
       "step": 2848
     },
     {
-      "epoch": 0.0782455851253742,
+      "epoch": 0.08084562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 1.9878257170565908e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.9864828496612717e-05,
+      "loss": 1.051,
       "step": 2849
     },
     {
-      "epoch": 0.07827304935321744,
+      "epoch": 0.08087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.9878118754260357e-05,
-      "loss": 1.0783,
+      "learning_rate": 1.98646778506441e-05,
+      "loss": 1.1117,
       "step": 2850
     },
     {
-      "epoch": 0.07830051358106067,
+      "epoch": 0.08090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.987798025979531e-05,
-      "loss": 1.118,
+      "learning_rate": 1.9864527121348108e-05,
+      "loss": 1.0403,
       "step": 2851
     },
     {
-      "epoch": 0.07832797780890391,
+      "epoch": 0.08093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.9877841687171853e-05,
-      "loss": 1.002,
+      "learning_rate": 1.9864376308726006e-05,
+      "loss": 1.1978,
       "step": 2852
     },
     {
-      "epoch": 0.07835544203674713,
+      "epoch": 0.08095913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.987770303639109e-05,
-      "loss": 1.101,
+      "learning_rate": 1.986422541277908e-05,
+      "loss": 1.0442,
       "step": 2853
     },
     {
-      "epoch": 0.07838290626459037,
+      "epoch": 0.08098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.987756430745411e-05,
-      "loss": 1.0948,
+      "learning_rate": 1.9864074433508597e-05,
+      "loss": 1.0206,
       "step": 2854
     },
     {
-      "epoch": 0.0784103704924336,
+      "epoch": 0.08101589103291713,
       "grad_norm": 0.0,
-      "learning_rate": 1.987742550036202e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.9863923370915834e-05,
+      "loss": 1.1049,
       "step": 2855
     },
     {
-      "epoch": 0.07843783472027684,
+      "epoch": 0.08104426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.9877286615115908e-05,
-      "loss": 1.1031,
+      "learning_rate": 1.9863772225002066e-05,
+      "loss": 1.1604,
       "step": 2856
     },
     {
-      "epoch": 0.07846529894812007,
+      "epoch": 0.08107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.987714765171688e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.986362099576857e-05,
+      "loss": 1.0176,
       "step": 2857
     },
     {
-      "epoch": 0.0784927631759633,
+      "epoch": 0.08110102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 1.9877008610166036e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.9863469683216622e-05,
+      "loss": 1.1833,
       "step": 2858
     },
     {
-      "epoch": 0.07852022740380654,
+      "epoch": 0.08112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.9876869490464468e-05,
-      "loss": 1.0579,
+      "learning_rate": 1.9863318287347507e-05,
+      "loss": 1.0378,
       "step": 2859
     },
     {
-      "epoch": 0.07854769163164978,
+      "epoch": 0.0811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9876730292613292e-05,
-      "loss": 1.086,
+      "learning_rate": 1.9863166808162497e-05,
+      "loss": 1.1114,
       "step": 2860
     },
     {
-      "epoch": 0.07857515585949301,
+      "epoch": 0.0811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9876591016613593e-05,
-      "loss": 1.0965,
+      "learning_rate": 1.9863015245662872e-05,
+      "loss": 1.1239,
       "step": 2861
     },
     {
-      "epoch": 0.07860262008733625,
+      "epoch": 0.08121452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 1.9876451662466482e-05,
-      "loss": 1.1389,
+      "learning_rate": 1.9862863599849917e-05,
+      "loss": 0.9142,
       "step": 2862
     },
     {
-      "epoch": 0.07863008431517948,
+      "epoch": 0.08124290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 1.987631223017306e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.9862711870724908e-05,
+      "loss": 1.05,
       "step": 2863
     },
     {
-      "epoch": 0.07865754854302272,
+      "epoch": 0.08127128263337118,
       "grad_norm": 0.0,
-      "learning_rate": 1.987617271973443e-05,
-      "loss": 0.996,
+      "learning_rate": 1.9862560058289127e-05,
+      "loss": 1.168,
       "step": 2864
     },
     {
-      "epoch": 0.07868501277086595,
+      "epoch": 0.08129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9876033131151696e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.986240816254386e-05,
+      "loss": 0.9996,
       "step": 2865
     },
     {
-      "epoch": 0.07871247699870917,
+      "epoch": 0.08132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.987589346442596e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.9862256183490387e-05,
+      "loss": 0.9531,
       "step": 2866
     },
     {
-      "epoch": 0.07873994122655241,
+      "epoch": 0.08135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875753719558333e-05,
-      "loss": 1.0923,
+      "learning_rate": 1.9862104121129992e-05,
+      "loss": 1.0161,
       "step": 2867
     },
     {
-      "epoch": 0.07876740545439564,
+      "epoch": 0.08138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875613896549916e-05,
-      "loss": 1.062,
+      "learning_rate": 1.9861951975463962e-05,
+      "loss": 0.9814,
       "step": 2868
     },
     {
-      "epoch": 0.07879486968223888,
+      "epoch": 0.08141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875473995401812e-05,
-      "loss": 1.1072,
+      "learning_rate": 1.986179974649358e-05,
+      "loss": 1.0613,
       "step": 2869
     },
     {
-      "epoch": 0.07882233391008212,
+      "epoch": 0.08144154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875334016115136e-05,
-      "loss": 1.1907,
+      "learning_rate": 1.986164743422013e-05,
+      "loss": 1.0678,
       "step": 2870
     },
     {
-      "epoch": 0.07884979813792535,
+      "epoch": 0.08146992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875193958690992e-05,
-      "loss": 1.1942,
+      "learning_rate": 1.9861495038644905e-05,
+      "loss": 1.0635,
       "step": 2871
     },
     {
-      "epoch": 0.07887726236576859,
+      "epoch": 0.0814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.987505382313049e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.9861342559769184e-05,
+      "loss": 1.0859,
       "step": 2872
     },
     {
-      "epoch": 0.07890472659361182,
+      "epoch": 0.0815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9874913609434734e-05,
-      "loss": 1.0992,
+      "learning_rate": 1.986118999759426e-05,
+      "loss": 0.9062,
       "step": 2873
     },
     {
-      "epoch": 0.07893219082145506,
+      "epoch": 0.08155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.9874773317604837e-05,
-      "loss": 1.077,
+      "learning_rate": 1.986103735212142e-05,
+      "loss": 0.9502,
       "step": 2874
     },
     {
-      "epoch": 0.07895965504929829,
+      "epoch": 0.08158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.987463294764191e-05,
-      "loss": 1.1237,
+      "learning_rate": 1.9860884623351957e-05,
+      "loss": 1.0077,
       "step": 2875
     },
     {
-      "epoch": 0.07898711927714153,
+      "epoch": 0.08161180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.987449249954706e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.9860731811287157e-05,
+      "loss": 1.03,
       "step": 2876
     },
     {
-      "epoch": 0.07901458350498476,
+      "epoch": 0.08164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.9874351973321402e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.986057891592831e-05,
+      "loss": 1.0106,
       "step": 2877
     },
     {
-      "epoch": 0.079042047732828,
+      "epoch": 0.08166855845629965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9874211368966044e-05,
-      "loss": 1.0668,
+      "learning_rate": 1.9860425937276714e-05,
+      "loss": 1.0676,
       "step": 2878
     },
     {
-      "epoch": 0.07906951196067123,
+      "epoch": 0.08169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.9874070686482102e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.9860272875333653e-05,
+      "loss": 1.0466,
       "step": 2879
     },
     {
-      "epoch": 0.07909697618851445,
+      "epoch": 0.08172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873929925870687e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.986011973010043e-05,
+      "loss": 1.0519,
       "step": 2880
     },
     {
-      "epoch": 0.07912444041635769,
+      "epoch": 0.08175368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873789087132917e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.9859966501578325e-05,
+      "loss": 1.0861,
       "step": 2881
     },
     {
-      "epoch": 0.07915190464420092,
+      "epoch": 0.08178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.98736481702699e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.9859813189768643e-05,
+      "loss": 1.0951,
       "step": 2882
     },
     {
-      "epoch": 0.07917936887204416,
+      "epoch": 0.08181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873507175282753e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.9859659794672677e-05,
+      "loss": 1.0648,
       "step": 2883
     },
     {
-      "epoch": 0.0792068330998874,
+      "epoch": 0.08183881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873366102172596e-05,
-      "loss": 1.109,
+      "learning_rate": 1.9859506316291716e-05,
+      "loss": 1.1274,
       "step": 2884
     },
     {
-      "epoch": 0.07923429732773063,
+      "epoch": 0.0818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873224950940536e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.9859352754627067e-05,
+      "loss": 1.0288,
       "step": 2885
     },
     {
-      "epoch": 0.07926176155557386,
+      "epoch": 0.0818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.98730837215877e-05,
-      "loss": 1.1549,
+      "learning_rate": 1.985919910968002e-05,
+      "loss": 1.1359,
       "step": 2886
     },
     {
-      "epoch": 0.0792892257834171,
+      "epoch": 0.0819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9872942414115203e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.985904538145188e-05,
+      "loss": 1.089,
       "step": 2887
     },
     {
-      "epoch": 0.07931669001126034,
+      "epoch": 0.08195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9872801028524156e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.9858891569943936e-05,
+      "loss": 1.0921,
       "step": 2888
     },
     {
-      "epoch": 0.07934415423910357,
+      "epoch": 0.08198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.987265956481569e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.985873767515749e-05,
+      "loss": 1.0232,
       "step": 2889
     },
     {
-      "epoch": 0.0793716184669468,
+      "epoch": 0.08200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.987251802299091e-05,
-      "loss": 1.081,
+      "learning_rate": 1.9858583697093845e-05,
+      "loss": 0.9455,
       "step": 2890
     },
     {
-      "epoch": 0.07939908269479004,
+      "epoch": 0.08203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.987237640305095e-05,
-      "loss": 1.1575,
+      "learning_rate": 1.9858429635754303e-05,
+      "loss": 1.0395,
       "step": 2891
     },
     {
-      "epoch": 0.07942654692263328,
+      "epoch": 0.08206583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 1.987223470499692e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.9858275491140156e-05,
+      "loss": 1.0055,
       "step": 2892
     },
     {
-      "epoch": 0.07945401115047651,
+      "epoch": 0.08209421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 1.9872092928829944e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.985812126325272e-05,
+      "loss": 1.1758,
       "step": 2893
     },
     {
-      "epoch": 0.07948147537831973,
+      "epoch": 0.08212258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871951074551142e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.9857966952093285e-05,
+      "loss": 0.9824,
       "step": 2894
     },
     {
-      "epoch": 0.07950893960616297,
+      "epoch": 0.08215096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871809142161646e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.9857812557663166e-05,
+      "loss": 1.1331,
       "step": 2895
     },
     {
-      "epoch": 0.0795364038340062,
+      "epoch": 0.08217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871667131662565e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.9857658079963657e-05,
+      "loss": 1.0859,
       "step": 2896
     },
     {
-      "epoch": 0.07956386806184944,
+      "epoch": 0.08220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871525043055033e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.985750351899607e-05,
+      "loss": 1.012,
       "step": 2897
     },
     {
-      "epoch": 0.07959133228969267,
+      "epoch": 0.0822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871382876340174e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.9857348874761704e-05,
+      "loss": 1.1076,
       "step": 2898
     },
     {
-      "epoch": 0.07961879651753591,
+      "epoch": 0.0822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9871240631519105e-05,
-      "loss": 1.126,
+      "learning_rate": 1.985719414726187e-05,
+      "loss": 1.0446,
       "step": 2899
     },
     {
-      "epoch": 0.07964626074537914,
+      "epoch": 0.08229284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 1.987109830859296e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.9857039336497875e-05,
+      "loss": 1.0458,
       "step": 2900
     },
     {
-      "epoch": 0.07967372497322238,
+      "epoch": 0.08232122587968219,
       "grad_norm": 0.0,
-      "learning_rate": 1.987095590756286e-05,
-      "loss": 1.1557,
+      "learning_rate": 1.9856884442471024e-05,
+      "loss": 0.9798,
       "step": 2901
     },
     {
-      "epoch": 0.07970118920106561,
+      "epoch": 0.08234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.987081342842993e-05,
-      "loss": 1.1095,
+      "learning_rate": 1.985672946518263e-05,
+      "loss": 1.017,
       "step": 2902
     },
     {
-      "epoch": 0.07972865342890885,
+      "epoch": 0.08237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.9870670871195305e-05,
-      "loss": 1.1149,
+      "learning_rate": 1.9856574404633996e-05,
+      "loss": 1.0648,
       "step": 2903
     },
     {
-      "epoch": 0.07975611765675208,
+      "epoch": 0.08240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.987052823586011e-05,
-      "loss": 1.0254,
+      "learning_rate": 1.985641926082644e-05,
+      "loss": 0.9646,
       "step": 2904
     },
     {
-      "epoch": 0.07978358188459532,
+      "epoch": 0.08243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9870385522425467e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.9856264033761263e-05,
+      "loss": 1.0797,
       "step": 2905
     },
     {
-      "epoch": 0.07981104611243856,
+      "epoch": 0.08246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9870242730892516e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.985610872343978e-05,
+      "loss": 0.9978,
       "step": 2906
     },
     {
-      "epoch": 0.07983851034028178,
+      "epoch": 0.08249148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 1.9870099861262378e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.9855953329863302e-05,
+      "loss": 1.0783,
       "step": 2907
     },
     {
-      "epoch": 0.07986597456812501,
+      "epoch": 0.08251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869956913536185e-05,
-      "loss": 1.1228,
+      "learning_rate": 1.9855797853033148e-05,
+      "loss": 0.985,
       "step": 2908
     },
     {
-      "epoch": 0.07989343879596825,
+      "epoch": 0.08254824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869813887715072e-05,
-      "loss": 1.1432,
+      "learning_rate": 1.9855642292950622e-05,
+      "loss": 1.1302,
       "step": 2909
     },
     {
-      "epoch": 0.07992090302381148,
+      "epoch": 0.08257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869670783800168e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.9855486649617047e-05,
+      "loss": 1.0178,
       "step": 2910
     },
     {
-      "epoch": 0.07994836725165472,
+      "epoch": 0.0826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869527601792603e-05,
-      "loss": 1.1714,
+      "learning_rate": 1.9855330923033732e-05,
+      "loss": 1.0968,
       "step": 2911
     },
     {
-      "epoch": 0.07997583147949795,
+      "epoch": 0.08263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869384341693517e-05,
-      "loss": 1.0643,
+      "learning_rate": 1.9855175113201994e-05,
+      "loss": 1.1755,
       "step": 2912
     },
     {
-      "epoch": 0.08000329570734119,
+      "epoch": 0.08266174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869241003504035e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.9855019220123145e-05,
+      "loss": 1.0068,
       "step": 2913
     },
     {
-      "epoch": 0.08003075993518442,
+      "epoch": 0.08269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.9869097587225297e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.985486324379851e-05,
+      "loss": 1.0953,
       "step": 2914
     },
     {
-      "epoch": 0.08005822416302766,
+      "epoch": 0.08271850170261066,
       "grad_norm": 0.0,
-      "learning_rate": 1.9868954092858436e-05,
-      "loss": 1.0849,
+      "learning_rate": 1.98547071842294e-05,
+      "loss": 1.017,
       "step": 2915
     },
     {
-      "epoch": 0.0800856883908709,
+      "epoch": 0.08274687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9868810520404587e-05,
-      "loss": 0.9665,
+      "learning_rate": 1.9854551041417137e-05,
+      "loss": 1.0985,
       "step": 2916
     },
     {
-      "epoch": 0.08011315261871413,
+      "epoch": 0.08277525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.986866686986489e-05,
-      "loss": 1.1229,
+      "learning_rate": 1.9854394815363037e-05,
+      "loss": 1.1315,
       "step": 2917
     },
     {
-      "epoch": 0.08014061684655736,
+      "epoch": 0.08280363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.986852314124047e-05,
-      "loss": 1.135,
+      "learning_rate": 1.985423850606842e-05,
+      "loss": 0.993,
       "step": 2918
     },
     {
-      "epoch": 0.0801680810744006,
+      "epoch": 0.08283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9868379334532478e-05,
-      "loss": 1.1136,
+      "learning_rate": 1.985408211353461e-05,
+      "loss": 1.1005,
       "step": 2919
     },
     {
-      "epoch": 0.08019554530224383,
+      "epoch": 0.08286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.9868235449742043e-05,
-      "loss": 1.2059,
+      "learning_rate": 1.9853925637762925e-05,
+      "loss": 1.1917,
       "step": 2920
     },
     {
-      "epoch": 0.08022300953008706,
+      "epoch": 0.08288876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 1.986809148687031e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.9853769078754685e-05,
+      "loss": 1.106,
       "step": 2921
     },
     {
-      "epoch": 0.08025047375793029,
+      "epoch": 0.08291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867947445918413e-05,
-      "loss": 1.1019,
+      "learning_rate": 1.9853612436511215e-05,
+      "loss": 1.0167,
       "step": 2922
     },
     {
-      "epoch": 0.08027793798577353,
+      "epoch": 0.0829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867803326887495e-05,
-      "loss": 1.2917,
+      "learning_rate": 1.985345571103384e-05,
+      "loss": 1.0828,
       "step": 2923
     },
     {
-      "epoch": 0.08030540221361676,
+      "epoch": 0.0829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867659129778692e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.985329890232388e-05,
+      "loss": 1.0757,
       "step": 2924
     },
     {
-      "epoch": 0.08033286644146,
+      "epoch": 0.08300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867514854593147e-05,
-      "loss": 1.0951,
+      "learning_rate": 1.9853142010382662e-05,
+      "loss": 1.2318,
       "step": 2925
     },
     {
-      "epoch": 0.08036033066930323,
+      "epoch": 0.08303064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867370501332005e-05,
-      "loss": 1.0839,
+      "learning_rate": 1.985298503521151e-05,
+      "loss": 1.0027,
       "step": 2926
     },
     {
-      "epoch": 0.08038779489714647,
+      "epoch": 0.08305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9867226069996404e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.9852827976811752e-05,
+      "loss": 1.116,
       "step": 2927
     },
     {
-      "epoch": 0.0804152591249897,
+      "epoch": 0.08308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.986708156058749e-05,
-      "loss": 1.1259,
+      "learning_rate": 1.985267083518471e-05,
+      "loss": 1.0129,
       "step": 2928
     },
     {
-      "epoch": 0.08044272335283294,
+      "epoch": 0.08311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866936973106398e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.9852513610331717e-05,
+      "loss": 1.1226,
       "step": 2929
     },
     {
-      "epoch": 0.08047018758067617,
+      "epoch": 0.08314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866792307554284e-05,
-      "loss": 1.0851,
+      "learning_rate": 1.9852356302254098e-05,
+      "loss": 1.0136,
       "step": 2930
     },
     {
-      "epoch": 0.08049765180851941,
+      "epoch": 0.08317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.986664756393229e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.985219891095318e-05,
+      "loss": 1.1228,
       "step": 2931
     },
     {
-      "epoch": 0.08052511603636264,
+      "epoch": 0.08320090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866502742241554e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.9852041436430302e-05,
+      "loss": 1.1567,
       "step": 2932
     },
     {
-      "epoch": 0.08055258026420588,
+      "epoch": 0.08322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866357842483227e-05,
-      "loss": 1.0827,
+      "learning_rate": 1.985188387868678e-05,
+      "loss": 1.0062,
       "step": 2933
     },
     {
-      "epoch": 0.0805800444920491,
+      "epoch": 0.08325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866212864658455e-05,
-      "loss": 1.1302,
+      "learning_rate": 1.9851726237723953e-05,
+      "loss": 1.0765,
       "step": 2934
     },
     {
-      "epoch": 0.08060750871989233,
+      "epoch": 0.08328603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.9866067808768385e-05,
-      "loss": 0.9087,
+      "learning_rate": 1.9851568513543154e-05,
+      "loss": 1.0142,
       "step": 2935
     },
     {
-      "epoch": 0.08063497294773557,
+      "epoch": 0.08331441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.9865922674814167e-05,
-      "loss": 1.1342,
+      "learning_rate": 1.985141070614571e-05,
+      "loss": 1.1237,
       "step": 2936
     },
     {
-      "epoch": 0.0806624371755788,
+      "epoch": 0.0833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.986577746279694e-05,
-      "loss": 1.1509,
+      "learning_rate": 1.985125281553296e-05,
+      "loss": 1.1594,
       "step": 2937
     },
     {
-      "epoch": 0.08068990140342204,
+      "epoch": 0.08337116912599318,
       "grad_norm": 0.0,
-      "learning_rate": 1.9865632172717867e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.9851094841706232e-05,
+      "loss": 1.1067,
       "step": 2938
     },
     {
-      "epoch": 0.08071736563126528,
+      "epoch": 0.08339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.986548680457809e-05,
-      "loss": 1.0613,
+      "learning_rate": 1.9850936784666865e-05,
+      "loss": 1.0757,
       "step": 2939
     },
     {
-      "epoch": 0.08074482985910851,
+      "epoch": 0.08342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.986534135837876e-05,
-      "loss": 1.0925,
+      "learning_rate": 1.9850778644416194e-05,
+      "loss": 1.0395,
       "step": 2940
     },
     {
-      "epoch": 0.08077229408695175,
+      "epoch": 0.08345629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.9865195834121028e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.9850620420955548e-05,
+      "loss": 1.0716,
       "step": 2941
     },
     {
-      "epoch": 0.08079975831479498,
+      "epoch": 0.08348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.986505023180604e-05,
-      "loss": 1.1906,
+      "learning_rate": 1.9850462114286275e-05,
+      "loss": 1.0537,
       "step": 2942
     },
     {
-      "epoch": 0.08082722254263822,
+      "epoch": 0.08351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9864904551434957e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.98503037244097e-05,
+      "loss": 1.0148,
       "step": 2943
     },
     {
-      "epoch": 0.08085468677048145,
+      "epoch": 0.08354143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 1.986475879300893e-05,
-      "loss": 1.1277,
+      "learning_rate": 1.985014525132717e-05,
+      "loss": 1.0786,
       "step": 2944
     },
     {
-      "epoch": 0.08088215099832469,
+      "epoch": 0.08356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9864612956529105e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.9849986695040017e-05,
+      "loss": 1.0854,
       "step": 2945
     },
     {
-      "epoch": 0.08090961522616792,
+      "epoch": 0.08359818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 1.9864467041996645e-05,
-      "loss": 1.1689,
+      "learning_rate": 1.9849828055549584e-05,
+      "loss": 1.0581,
       "step": 2946
     },
     {
-      "epoch": 0.08093707945401116,
+      "epoch": 0.08362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.98643210494127e-05,
-      "loss": 1.0785,
+      "learning_rate": 1.9849669332857212e-05,
+      "loss": 1.055,
       "step": 2947
     },
     {
-      "epoch": 0.08096454368185438,
+      "epoch": 0.08365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.9864174978778427e-05,
-      "loss": 1.1572,
+      "learning_rate": 1.9849510526964243e-05,
+      "loss": 1.0903,
       "step": 2948
     },
     {
-      "epoch": 0.08099200790969761,
+      "epoch": 0.083683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9864028830094977e-05,
-      "loss": 1.1059,
+      "learning_rate": 1.984935163787201e-05,
+      "loss": 1.0459,
       "step": 2949
     },
     {
-      "epoch": 0.08101947213754085,
+      "epoch": 0.0837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.9863882603363512e-05,
-      "loss": 1.0686,
+      "learning_rate": 1.9849192665581865e-05,
+      "loss": 1.0738,
       "step": 2950
     },
     {
-      "epoch": 0.08104693636538408,
+      "epoch": 0.08374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.9863736298585188e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.9849033610095144e-05,
+      "loss": 0.9703,
       "step": 2951
     },
     {
-      "epoch": 0.08107440059322732,
+      "epoch": 0.08376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.986358991576116e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.9848874471413196e-05,
+      "loss": 1.0571,
       "step": 2952
     },
     {
-      "epoch": 0.08110186482107055,
+      "epoch": 0.08379682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 1.986344345489259e-05,
-      "loss": 1.1392,
+      "learning_rate": 1.9848715249537363e-05,
+      "loss": 0.9756,
       "step": 2953
     },
     {
-      "epoch": 0.08112932904891379,
+      "epoch": 0.08382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.9863296915980632e-05,
-      "loss": 1.0935,
+      "learning_rate": 1.984855594446899e-05,
+      "loss": 1.1504,
       "step": 2954
     },
     {
-      "epoch": 0.08115679327675702,
+      "epoch": 0.08385357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.986315029902645e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.984839655620942e-05,
+      "loss": 1.0977,
       "step": 2955
     },
     {
-      "epoch": 0.08118425750460026,
+      "epoch": 0.08388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.98630036040312e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.9848237084760004e-05,
+      "loss": 1.1891,
       "step": 2956
     },
     {
-      "epoch": 0.0812117217324435,
+      "epoch": 0.08391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.9862856830996048e-05,
-      "loss": 1.0632,
+      "learning_rate": 1.9848077530122083e-05,
+      "loss": 0.9978,
       "step": 2957
     },
     {
-      "epoch": 0.08123918596028673,
+      "epoch": 0.08393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.986270997992215e-05,
-      "loss": 1.089,
+      "learning_rate": 1.984791789229701e-05,
+      "loss": 1.004,
       "step": 2958
     },
     {
-      "epoch": 0.08126665018812997,
+      "epoch": 0.08396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.9862563050810668e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.984775817128613e-05,
+      "loss": 1.1451,
       "step": 2959
     },
     {
-      "epoch": 0.0812941144159732,
+      "epoch": 0.08399545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 1.986241604366277e-05,
-      "loss": 1.1848,
+      "learning_rate": 1.98475983670908e-05,
+      "loss": 1.0538,
       "step": 2960
     },
     {
-      "epoch": 0.08132157864381642,
+      "epoch": 0.08402383654937572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9862268958479617e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.984743847971236e-05,
+      "loss": 1.0885,
       "step": 2961
     },
     {
-      "epoch": 0.08134904287165966,
+      "epoch": 0.0840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9862121795262366e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.984727850915216e-05,
+      "loss": 0.9894,
       "step": 2962
     },
     {
-      "epoch": 0.08137650709950289,
+      "epoch": 0.08408059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861974554012192e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.984711845541156e-05,
+      "loss": 1.1683,
       "step": 2963
     },
     {
-      "epoch": 0.08140397132734613,
+      "epoch": 0.08410896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861827234730256e-05,
-      "loss": 1.1326,
+      "learning_rate": 1.984695831849191e-05,
+      "loss": 1.1648,
       "step": 2964
     },
     {
-      "epoch": 0.08143143555518936,
+      "epoch": 0.08413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861679837417723e-05,
-      "loss": 1.1201,
+      "learning_rate": 1.9846798098394558e-05,
+      "loss": 1.0745,
       "step": 2965
     },
     {
-      "epoch": 0.0814588997830326,
+      "epoch": 0.08416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861532362075757e-05,
-      "loss": 1.056,
+      "learning_rate": 1.984663779512086e-05,
+      "loss": 1.0828,
       "step": 2966
     },
     {
-      "epoch": 0.08148636401087583,
+      "epoch": 0.08419409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861384808705525e-05,
-      "loss": 1.1111,
+      "learning_rate": 1.984647740867217e-05,
+      "loss": 1.0204,
       "step": 2967
     },
     {
-      "epoch": 0.08151382823871907,
+      "epoch": 0.08422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.98612371773082e-05,
-      "loss": 1.0401,
+      "learning_rate": 1.9846316939049847e-05,
+      "loss": 1.1292,
       "step": 2968
     },
     {
-      "epoch": 0.0815412924665623,
+      "epoch": 0.08425085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9861089467884946e-05,
-      "loss": 1.1599,
+      "learning_rate": 1.9846156386255236e-05,
+      "loss": 1.1005,
       "step": 2969
     },
     {
-      "epoch": 0.08156875669440554,
+      "epoch": 0.08427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.986094168043693e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.98459957502897e-05,
+      "loss": 1.0501,
       "step": 2970
     },
     {
-      "epoch": 0.08159622092224877,
+      "epoch": 0.08430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.9860793814965326e-05,
-      "loss": 1.0246,
+      "learning_rate": 1.9845835031154598e-05,
+      "loss": 1.1031,
       "step": 2971
     },
     {
-      "epoch": 0.08162368515009201,
+      "epoch": 0.08433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.98606458714713e-05,
-      "loss": 1.1413,
+      "learning_rate": 1.984567422885128e-05,
+      "loss": 0.962,
       "step": 2972
     },
     {
-      "epoch": 0.08165114937793524,
+      "epoch": 0.08436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9860497849956023e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.9845513343381116e-05,
+      "loss": 1.0873,
       "step": 2973
     },
     {
-      "epoch": 0.08167861360577848,
+      "epoch": 0.0843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9860349750420672e-05,
-      "loss": 1.1096,
+      "learning_rate": 1.9845352374745455e-05,
+      "loss": 1.0949,
       "step": 2974
     },
     {
-      "epoch": 0.0817060778336217,
+      "epoch": 0.0844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.986020157286641e-05,
-      "loss": 1.1659,
+      "learning_rate": 1.984519132294566e-05,
+      "loss": 1.1229,
       "step": 2975
     },
     {
-      "epoch": 0.08173354206146494,
+      "epoch": 0.0844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.9860053317294417e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.984503018798309e-05,
+      "loss": 1.0181,
       "step": 2976
     },
     {
-      "epoch": 0.08176100628930817,
+      "epoch": 0.08447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.985990498370586e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.984486896985911e-05,
+      "loss": 0.9611,
       "step": 2977
     },
     {
-      "epoch": 0.08178847051715141,
+      "epoch": 0.08450624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 1.9859756572101915e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.9844707668575078e-05,
+      "loss": 0.9576,
       "step": 2978
     },
     {
-      "epoch": 0.08181593474499464,
+      "epoch": 0.08453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9859608082483757e-05,
-      "loss": 1.1703,
+      "learning_rate": 1.984454628413236e-05,
+      "loss": 1.1555,
       "step": 2979
     },
     {
-      "epoch": 0.08184339897283788,
+      "epoch": 0.08456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9859459514852565e-05,
-      "loss": 1.0878,
+      "learning_rate": 1.9844384816532313e-05,
+      "loss": 1.1057,
       "step": 2980
     },
     {
-      "epoch": 0.08187086320068111,
+      "epoch": 0.08459137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 1.985931086920951e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.9844223265776305e-05,
+      "loss": 1.1341,
       "step": 2981
     },
     {
-      "epoch": 0.08189832742852435,
+      "epoch": 0.08461975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.9859162145555762e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.9844061631865703e-05,
+      "loss": 1.041,
       "step": 2982
     },
     {
-      "epoch": 0.08192579165636758,
+      "epoch": 0.08464812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.9859013343892506e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.9843899914801867e-05,
+      "loss": 0.9773,
       "step": 2983
     },
     {
-      "epoch": 0.08195325588421082,
+      "epoch": 0.08467650397275824,
       "grad_norm": 0.0,
-      "learning_rate": 1.985886446422092e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.984373811458617e-05,
+      "loss": 1.0862,
       "step": 2984
     },
     {
-      "epoch": 0.08198072011205405,
+      "epoch": 0.08470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.985871550654218e-05,
-      "loss": 1.2103,
+      "learning_rate": 1.9843576231219973e-05,
+      "loss": 1.0594,
       "step": 2985
     },
     {
-      "epoch": 0.08200818433989729,
+      "epoch": 0.08473325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.985856647085746e-05,
-      "loss": 1.004,
+      "learning_rate": 1.9843414264704646e-05,
+      "loss": 1.0238,
       "step": 2986
     },
     {
-      "epoch": 0.08203564856774052,
+      "epoch": 0.08476163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.9858417357167945e-05,
-      "loss": 1.031,
+      "learning_rate": 1.9843252215041552e-05,
+      "loss": 1.0253,
       "step": 2987
     },
     {
-      "epoch": 0.08206311279558375,
+      "epoch": 0.0847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9858268165474813e-05,
-      "loss": 1.192,
+      "learning_rate": 1.984309008223207e-05,
+      "loss": 1.0424,
       "step": 2988
     },
     {
-      "epoch": 0.08209057702342698,
+      "epoch": 0.08481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9858118895779248e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.984292786627756e-05,
+      "loss": 0.9973,
       "step": 2989
     },
     {
-      "epoch": 0.08211804125127022,
+      "epoch": 0.08484676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 1.985796954808242e-05,
-      "loss": 1.1872,
+      "learning_rate": 1.9842765567179394e-05,
+      "loss": 1.0931,
       "step": 2990
     },
     {
-      "epoch": 0.08214550547911345,
+      "epoch": 0.08487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9857820122385523e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.9842603184938948e-05,
+      "loss": 1.0396,
       "step": 2991
     },
     {
-      "epoch": 0.08217296970695669,
+      "epoch": 0.08490351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.9857670618689737e-05,
-      "loss": 1.1118,
+      "learning_rate": 1.984244071955759e-05,
+      "loss": 1.0175,
       "step": 2992
     },
     {
-      "epoch": 0.08220043393479992,
+      "epoch": 0.08493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.985752103699624e-05,
-      "loss": 1.2046,
+      "learning_rate": 1.9842278171036692e-05,
+      "loss": 1.1586,
       "step": 2993
     },
     {
-      "epoch": 0.08222789816264316,
+      "epoch": 0.08496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.9857371377306216e-05,
-      "loss": 1.1524,
+      "learning_rate": 1.984211553937763e-05,
+      "loss": 1.1019,
       "step": 2994
     },
     {
-      "epoch": 0.08225536239048639,
+      "epoch": 0.08498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.9857221639620854e-05,
-      "loss": 1.1603,
+      "learning_rate": 1.9841952824581773e-05,
+      "loss": 1.2157,
       "step": 2995
     },
     {
-      "epoch": 0.08228282661832963,
+      "epoch": 0.08501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.9857071823941336e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.98417900266505e-05,
+      "loss": 1.0819,
       "step": 2996
     },
     {
-      "epoch": 0.08231029084617286,
+      "epoch": 0.08504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856921930268846e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.9841627145585184e-05,
+      "loss": 1.1494,
       "step": 2997
     },
     {
-      "epoch": 0.0823377550740161,
+      "epoch": 0.08507377979568671,
       "grad_norm": 0.0,
-      "learning_rate": 1.985677195860457e-05,
-      "loss": 1.114,
+      "learning_rate": 1.9841464181387197e-05,
+      "loss": 1.0197,
       "step": 2998
     },
     {
-      "epoch": 0.08236521930185933,
+      "epoch": 0.08510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856621908949704e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.9841301134057925e-05,
+      "loss": 1.0233,
       "step": 2999
     },
     {
-      "epoch": 0.08239268352970257,
+      "epoch": 0.0851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856471781305423e-05,
-      "loss": 1.0548,
+      "learning_rate": 1.9841138003598737e-05,
+      "loss": 1.0634,
       "step": 3000
     },
     {
-      "epoch": 0.0824201477575458,
+      "epoch": 0.0851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856321575672918e-05,
-      "loss": 1.1765,
+      "learning_rate": 1.9840974790011015e-05,
+      "loss": 1.0414,
       "step": 3001
     },
     {
-      "epoch": 0.08244761198538902,
+      "epoch": 0.08518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856171292053378e-05,
-      "loss": 1.1649,
+      "learning_rate": 1.9840811493296134e-05,
+      "loss": 1.0423,
       "step": 3002
     },
     {
-      "epoch": 0.08247507621323226,
+      "epoch": 0.08521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9856020930447992e-05,
-      "loss": 1.0675,
+      "learning_rate": 1.9840648113455478e-05,
+      "loss": 1.1871,
       "step": 3003
     },
     {
-      "epoch": 0.0825025404410755,
+      "epoch": 0.08524404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 1.9855870490857953e-05,
-      "loss": 1.1388,
+      "learning_rate": 1.9840484650490424e-05,
+      "loss": 1.0109,
       "step": 3004
     },
     {
-      "epoch": 0.08253000466891873,
+      "epoch": 0.08527241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 1.9855719973284447e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.984032110440236e-05,
+      "loss": 0.9941,
       "step": 3005
     },
     {
-      "epoch": 0.08255746889676197,
+      "epoch": 0.08530079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.985556937772867e-05,
-      "loss": 1.0867,
+      "learning_rate": 1.9840157475192655e-05,
+      "loss": 1.1481,
       "step": 3006
     },
     {
-      "epoch": 0.0825849331246052,
+      "epoch": 0.08532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.9855418704191807e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.9839993762862698e-05,
+      "loss": 0.9224,
       "step": 3007
     },
     {
-      "epoch": 0.08261239735244844,
+      "epoch": 0.08535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9855267952675055e-05,
-      "loss": 1.0957,
+      "learning_rate": 1.983982996741387e-05,
+      "loss": 1.0468,
       "step": 3008
     },
     {
-      "epoch": 0.08263986158029167,
+      "epoch": 0.08538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.9855117123179606e-05,
-      "loss": 1.0897,
+      "learning_rate": 1.983966608884756e-05,
+      "loss": 1.1698,
       "step": 3009
     },
     {
-      "epoch": 0.0826673258081349,
+      "epoch": 0.08541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.9854966215706653e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.9839502127165145e-05,
+      "loss": 1.0002,
       "step": 3010
     },
     {
-      "epoch": 0.08269479003597814,
+      "epoch": 0.08544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.985481523025739e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.9839338082368017e-05,
+      "loss": 1.0838,
       "step": 3011
     },
     {
-      "epoch": 0.08272225426382138,
+      "epoch": 0.0854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9854664166833005e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.9839173954457553e-05,
+      "loss": 1.0893,
       "step": 3012
     },
     {
-      "epoch": 0.08274971849166461,
+      "epoch": 0.0854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.985451302543471e-05,
-      "loss": 1.0728,
+      "learning_rate": 1.983900974343515e-05,
+      "loss": 1.0081,
       "step": 3013
     },
     {
-      "epoch": 0.08277718271950785,
+      "epoch": 0.085527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9854361806063686e-05,
-      "loss": 1.014,
+      "learning_rate": 1.9838845449302183e-05,
+      "loss": 1.0851,
       "step": 3014
     },
     {
-      "epoch": 0.08280464694735108,
+      "epoch": 0.08555618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 1.985421050872113e-05,
-      "loss": 1.2236,
+      "learning_rate": 1.983868107206005e-05,
+      "loss": 0.9441,
       "step": 3015
     },
     {
-      "epoch": 0.0828321111751943,
+      "epoch": 0.08558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.985405913340825e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.9838516611710136e-05,
+      "loss": 1.0245,
       "step": 3016
     },
     {
-      "epoch": 0.08285957540303754,
+      "epoch": 0.08561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.9853907680126236e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.983835206825383e-05,
+      "loss": 0.9027,
       "step": 3017
     },
     {
-      "epoch": 0.08288703963088077,
+      "epoch": 0.08564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.9853756148876285e-05,
-      "loss": 1.1655,
+      "learning_rate": 1.9838187441692517e-05,
+      "loss": 1.0741,
       "step": 3018
     },
     {
-      "epoch": 0.08291450385872401,
+      "epoch": 0.08566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.98536045396596e-05,
-      "loss": 1.1362,
+      "learning_rate": 1.9838022732027597e-05,
+      "loss": 1.0633,
       "step": 3019
     },
     {
-      "epoch": 0.08294196808656724,
+      "epoch": 0.08569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.9853452852477378e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.9837857939260457e-05,
+      "loss": 1.0803,
       "step": 3020
     },
     {
-      "epoch": 0.08296943231441048,
+      "epoch": 0.08572644721906925,
       "grad_norm": 0.0,
-      "learning_rate": 1.985330108733082e-05,
-      "loss": 1.0969,
+      "learning_rate": 1.9837693063392487e-05,
+      "loss": 1.0415,
       "step": 3021
     },
     {
-      "epoch": 0.08299689654225371,
+      "epoch": 0.08575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.9853149244221126e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.9837528104425082e-05,
+      "loss": 1.0489,
       "step": 3022
     },
     {
-      "epoch": 0.08302436077009695,
+      "epoch": 0.08578320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.98529973231495e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.9837363062359632e-05,
+      "loss": 1.1132,
       "step": 3023
     },
     {
-      "epoch": 0.08305182499794019,
+      "epoch": 0.08581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9852845324117144e-05,
-      "loss": 1.0889,
+      "learning_rate": 1.983719793719754e-05,
+      "loss": 0.9943,
       "step": 3024
     },
     {
-      "epoch": 0.08307928922578342,
+      "epoch": 0.08583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.9852693247125255e-05,
-      "loss": 1.1539,
+      "learning_rate": 1.9837032728940194e-05,
+      "loss": 0.9725,
       "step": 3025
     },
     {
-      "epoch": 0.08310675345362666,
+      "epoch": 0.0858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9852541092175042e-05,
-      "loss": 1.1089,
+      "learning_rate": 1.9836867437588987e-05,
+      "loss": 1.1106,
       "step": 3026
     },
     {
-      "epoch": 0.08313421768146989,
+      "epoch": 0.0858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9852388859267712e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.983670206314532e-05,
+      "loss": 1.0229,
       "step": 3027
     },
     {
-      "epoch": 0.08316168190931313,
+      "epoch": 0.08592508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.985223654840446e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.9836536605610587e-05,
+      "loss": 1.0389,
       "step": 3028
     },
     {
-      "epoch": 0.08318914613715635,
+      "epoch": 0.08595346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 1.9852084159586495e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.983637106498619e-05,
+      "loss": 1.1334,
       "step": 3029
     },
     {
-      "epoch": 0.08321661036499958,
+      "epoch": 0.08598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9851931692815028e-05,
-      "loss": 1.1009,
+      "learning_rate": 1.983620544127352e-05,
+      "loss": 1.1211,
       "step": 3030
     },
     {
-      "epoch": 0.08324407459284282,
+      "epoch": 0.08601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.985177914809126e-05,
-      "loss": 1.1598,
+      "learning_rate": 1.9836039734473983e-05,
+      "loss": 1.0032,
       "step": 3031
     },
     {
-      "epoch": 0.08327153882068605,
+      "epoch": 0.08603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.98516265254164e-05,
-      "loss": 1.031,
+      "learning_rate": 1.9835873944588978e-05,
+      "loss": 1.0383,
       "step": 3032
     },
     {
-      "epoch": 0.08329900304852929,
+      "epoch": 0.08606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.985147382479165e-05,
-      "loss": 1.0952,
+      "learning_rate": 1.98357080716199e-05,
+      "loss": 0.9944,
       "step": 3033
     },
     {
-      "epoch": 0.08332646727637252,
+      "epoch": 0.08609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.9851321046218226e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.9835542115568157e-05,
+      "loss": 1.0566,
       "step": 3034
     },
     {
-      "epoch": 0.08335393150421576,
+      "epoch": 0.08612372304199772,
       "grad_norm": 0.0,
-      "learning_rate": 1.9851168189697335e-05,
-      "loss": 1.1344,
+      "learning_rate": 1.9835376076435146e-05,
+      "loss": 1.0078,
       "step": 3035
     },
     {
-      "epoch": 0.083381395732059,
+      "epoch": 0.08615209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.9851015255230185e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.983520995422227e-05,
+      "loss": 0.9987,
       "step": 3036
     },
     {
-      "epoch": 0.08340885995990223,
+      "epoch": 0.08618047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.9850862242817986e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9835043748930936e-05,
+      "loss": 1.0574,
       "step": 3037
     },
     {
-      "epoch": 0.08343632418774546,
+      "epoch": 0.0862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.985070915246195e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.9834877460562546e-05,
+      "loss": 1.0175,
       "step": 3038
     },
     {
-      "epoch": 0.0834637884155887,
+      "epoch": 0.0862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.985055598416328e-05,
-      "loss": 1.1079,
+      "learning_rate": 1.98347110891185e-05,
+      "loss": 1.0207,
       "step": 3039
     },
     {
-      "epoch": 0.08349125264343193,
+      "epoch": 0.08626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.9850402737923206e-05,
-      "loss": 1.0456,
+      "learning_rate": 1.983454463460021e-05,
+      "loss": 1.0212,
       "step": 3040
     },
     {
-      "epoch": 0.08351871687127517,
+      "epoch": 0.08629398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 1.9850249413742926e-05,
-      "loss": 1.0442,
+      "learning_rate": 1.983437809700908e-05,
+      "loss": 1.0435,
       "step": 3041
     },
     {
-      "epoch": 0.0835461810991184,
+      "epoch": 0.08632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9850096011623658e-05,
-      "loss": 1.1168,
+      "learning_rate": 1.9834211476346516e-05,
+      "loss": 1.0912,
       "step": 3042
     },
     {
-      "epoch": 0.08357364532696163,
+      "epoch": 0.08635073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.984994253156661e-05,
-      "loss": 1.1582,
+      "learning_rate": 1.9834044772613924e-05,
+      "loss": 1.0043,
       "step": 3043
     },
     {
-      "epoch": 0.08360110955480486,
+      "epoch": 0.08637911464245177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9849788973573007e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.9833877985812716e-05,
+      "loss": 1.0407,
       "step": 3044
     },
     {
-      "epoch": 0.0836285737826481,
+      "epoch": 0.08640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.9849635337644053e-05,
-      "loss": 1.2164,
+      "learning_rate": 1.9833711115944295e-05,
+      "loss": 1.0564,
       "step": 3045
     },
     {
-      "epoch": 0.08365603801049133,
+      "epoch": 0.08643586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.984948162378097e-05,
-      "loss": 0.9941,
+      "learning_rate": 1.9833544163010074e-05,
+      "loss": 1.0326,
       "step": 3046
     },
     {
-      "epoch": 0.08368350223833457,
+      "epoch": 0.08646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9849327831984974e-05,
-      "loss": 1.1453,
+      "learning_rate": 1.9833377127011468e-05,
+      "loss": 1.0674,
       "step": 3047
     },
     {
-      "epoch": 0.0837109664661778,
+      "epoch": 0.08649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.984917396225728e-05,
-      "loss": 1.0413,
+      "learning_rate": 1.9833210007949878e-05,
+      "loss": 1.0298,
       "step": 3048
     },
     {
-      "epoch": 0.08373843069402104,
+      "epoch": 0.08652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9849020014599108e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.9833042805826725e-05,
+      "loss": 1.0465,
       "step": 3049
     },
     {
-      "epoch": 0.08376589492186427,
+      "epoch": 0.08654937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 1.9848865989011678e-05,
-      "loss": 1.1133,
+      "learning_rate": 1.9832875520643417e-05,
+      "loss": 0.9803,
       "step": 3050
     },
     {
-      "epoch": 0.08379335914970751,
+      "epoch": 0.086577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.98487118854962e-05,
-      "loss": 1.1216,
+      "learning_rate": 1.9832708152401366e-05,
+      "loss": 1.0693,
       "step": 3051
     },
     {
-      "epoch": 0.08382082337755074,
+      "epoch": 0.08660612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 1.9848557704053898e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.9832540701101987e-05,
+      "loss": 1.0912,
       "step": 3052
     },
     {
-      "epoch": 0.08384828760539398,
+      "epoch": 0.08663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.9848403444685998e-05,
-      "loss": 1.1614,
+      "learning_rate": 1.9832373166746695e-05,
+      "loss": 1.05,
       "step": 3053
     },
     {
-      "epoch": 0.08387575183323721,
+      "epoch": 0.08666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.984824910739371e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.983220554933691e-05,
+      "loss": 1.1015,
       "step": 3054
     },
     {
-      "epoch": 0.08390321606108045,
+      "epoch": 0.08669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9848094692178263e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.9832037848874038e-05,
+      "loss": 1.0365,
       "step": 3055
     },
     {
-      "epoch": 0.08393068028892367,
+      "epoch": 0.08671963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.9847940199040874e-05,
-      "loss": 1.1191,
+      "learning_rate": 1.98318700653595e-05,
+      "loss": 0.9373,
       "step": 3056
     },
     {
-      "epoch": 0.0839581445167669,
+      "epoch": 0.08674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.984778562798277e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9831702198794713e-05,
+      "loss": 1.1041,
       "step": 3057
     },
     {
-      "epoch": 0.08398560874461014,
+      "epoch": 0.08677639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9847630979005168e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.9831534249181095e-05,
+      "loss": 1.1612,
       "step": 3058
     },
     {
-      "epoch": 0.08401307297245338,
+      "epoch": 0.08680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9847476252109294e-05,
-      "loss": 1.1287,
+      "learning_rate": 1.9831366216520068e-05,
+      "loss": 1.1271,
       "step": 3059
     },
     {
-      "epoch": 0.08404053720029661,
+      "epoch": 0.08683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.9847321447296378e-05,
-      "loss": 1.1118,
+      "learning_rate": 1.9831198100813047e-05,
+      "loss": 1.0277,
       "step": 3060
     },
     {
-      "epoch": 0.08406800142813985,
+      "epoch": 0.08686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.9847166564567638e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.9831029902061455e-05,
+      "loss": 1.1796,
       "step": 3061
     },
     {
-      "epoch": 0.08409546565598308,
+      "epoch": 0.08688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.98470116039243e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.983086162026671e-05,
+      "loss": 1.0175,
       "step": 3062
     },
     {
-      "epoch": 0.08412292988382632,
+      "epoch": 0.08691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.9846856565367595e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9830693255430236e-05,
+      "loss": 1.1006,
       "step": 3063
     },
     {
-      "epoch": 0.08415039411166955,
+      "epoch": 0.08694665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 1.9846701448898748e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.9830524807553455e-05,
+      "loss": 1.0496,
       "step": 3064
     },
     {
-      "epoch": 0.08417785833951279,
+      "epoch": 0.0869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.984654625451898e-05,
-      "loss": 1.0386,
+      "learning_rate": 1.9830356276637788e-05,
+      "loss": 1.0273,
       "step": 3065
     },
     {
-      "epoch": 0.08420532256735602,
+      "epoch": 0.08700340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9846390982229526e-05,
-      "loss": 1.0816,
+      "learning_rate": 1.983018766268466e-05,
+      "loss": 1.0015,
       "step": 3066
     },
     {
-      "epoch": 0.08423278679519926,
+      "epoch": 0.08703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.9846235632031613e-05,
-      "loss": 1.0908,
+      "learning_rate": 1.9830018965695494e-05,
+      "loss": 1.0775,
       "step": 3067
     },
     {
-      "epoch": 0.08426025102304249,
+      "epoch": 0.08706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.984608020392647e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.9829850185671717e-05,
+      "loss": 1.0767,
       "step": 3068
     },
     {
-      "epoch": 0.08428771525088573,
+      "epoch": 0.08708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.9845924697915323e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.9829681322614754e-05,
+      "loss": 1.0396,
       "step": 3069
     },
     {
-      "epoch": 0.08431517947872895,
+      "epoch": 0.08711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.984576911399941e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.9829512376526027e-05,
+      "loss": 1.0515,
       "step": 3070
     },
     {
-      "epoch": 0.08434264370657218,
+      "epoch": 0.08714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.9845613452179954e-05,
-      "loss": 1.041,
+      "learning_rate": 1.982934334740697e-05,
+      "loss": 1.0743,
       "step": 3071
     },
     {
-      "epoch": 0.08437010793441542,
+      "epoch": 0.08717366628830873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9845457712458196e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.982917423525901e-05,
+      "loss": 1.099,
       "step": 3072
     },
     {
-      "epoch": 0.08439757216225865,
+      "epoch": 0.08720204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.9845301894835357e-05,
-      "loss": 1.1264,
+      "learning_rate": 1.982900504008357e-05,
+      "loss": 0.9913,
       "step": 3073
     },
     {
-      "epoch": 0.08442503639010189,
+      "epoch": 0.08723041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.9845145999312677e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.9828835761882085e-05,
+      "loss": 1.0457,
       "step": 3074
     },
     {
-      "epoch": 0.08445250061794513,
+      "epoch": 0.08725879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 1.984499002589139e-05,
-      "loss": 1.182,
+      "learning_rate": 1.9828666400655982e-05,
+      "loss": 0.9859,
       "step": 3075
     },
     {
-      "epoch": 0.08447996484578836,
+      "epoch": 0.08728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.984483397457273e-05,
-      "loss": 1.1343,
+      "learning_rate": 1.982849695640669e-05,
+      "loss": 1.1194,
       "step": 3076
     },
     {
-      "epoch": 0.0845074290736316,
+      "epoch": 0.0873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.984467784535793e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.9828327429135645e-05,
+      "loss": 1.02,
       "step": 3077
     },
     {
-      "epoch": 0.08453489330147483,
+      "epoch": 0.08734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.984452163824822e-05,
-      "loss": 1.1573,
+      "learning_rate": 1.9828157818844277e-05,
+      "loss": 1.064,
       "step": 3078
     },
     {
-      "epoch": 0.08456235752931807,
+      "epoch": 0.08737230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.9844365353244845e-05,
-      "loss": 1.1104,
+      "learning_rate": 1.9827988125534014e-05,
+      "loss": 1.0123,
       "step": 3079
     },
     {
-      "epoch": 0.0845898217571613,
+      "epoch": 0.08740068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 1.984420899034904e-05,
-      "loss": 1.1182,
+      "learning_rate": 1.9827818349206296e-05,
+      "loss": 1.0573,
       "step": 3080
     },
     {
-      "epoch": 0.08461728598500454,
+      "epoch": 0.08742905788876278,
       "grad_norm": 0.0,
-      "learning_rate": 1.9844052549562038e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.9827648489862553e-05,
+      "loss": 1.0428,
       "step": 3081
     },
     {
-      "epoch": 0.08464475021284777,
+      "epoch": 0.08745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.984389603088508e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.9827478547504224e-05,
+      "loss": 1.0088,
       "step": 3082
     },
     {
-      "epoch": 0.084672214440691,
+      "epoch": 0.08748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.98437394343194e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.982730852213274e-05,
+      "loss": 0.9504,
       "step": 3083
     },
     {
-      "epoch": 0.08469967866853423,
+      "epoch": 0.08751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.9843582759866242e-05,
-      "loss": 1.1285,
+      "learning_rate": 1.982713841374954e-05,
+      "loss": 1.1702,
       "step": 3084
     },
     {
-      "epoch": 0.08472714289637746,
+      "epoch": 0.08754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.9843426007526845e-05,
-      "loss": 1.0561,
+      "learning_rate": 1.982696822235606e-05,
+      "loss": 1.0767,
       "step": 3085
     },
     {
-      "epoch": 0.0847546071242207,
+      "epoch": 0.08757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.984326917730245e-05,
-      "loss": 1.1252,
+      "learning_rate": 1.9826797947953738e-05,
+      "loss": 1.0703,
       "step": 3086
     },
     {
-      "epoch": 0.08478207135206393,
+      "epoch": 0.08759931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 1.9843112269194297e-05,
-      "loss": 1.1526,
+      "learning_rate": 1.9826627590544013e-05,
+      "loss": 1.0164,
       "step": 3087
     },
     {
-      "epoch": 0.08480953557990717,
+      "epoch": 0.08762769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842955283203623e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.982645715012832e-05,
+      "loss": 1.0264,
       "step": 3088
     },
     {
-      "epoch": 0.0848369998077504,
+      "epoch": 0.0876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842798219331674e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.9826286626708106e-05,
+      "loss": 1.0847,
       "step": 3089
     },
     {
-      "epoch": 0.08486446403559364,
+      "epoch": 0.0876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842641077579694e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.9826116020284803e-05,
+      "loss": 0.9575,
       "step": 3090
     },
     {
-      "epoch": 0.08489192826343687,
+      "epoch": 0.08771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842483857948927e-05,
-      "loss": 1.1477,
+      "learning_rate": 1.9825945330859857e-05,
+      "loss": 0.9956,
       "step": 3091
     },
     {
-      "epoch": 0.08491939249128011,
+      "epoch": 0.08774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842326560440616e-05,
-      "loss": 1.1476,
+      "learning_rate": 1.9825774558434712e-05,
+      "loss": 1.0322,
       "step": 3092
     },
     {
-      "epoch": 0.08494685671912335,
+      "epoch": 0.08776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.9842169185056e-05,
-      "loss": 1.1813,
+      "learning_rate": 1.9825603703010804e-05,
+      "loss": 1.1379,
       "step": 3093
     },
     {
-      "epoch": 0.08497432094696658,
+      "epoch": 0.08779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.984201173179633e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.982543276458958e-05,
+      "loss": 1.2012,
       "step": 3094
     },
     {
-      "epoch": 0.08500178517480982,
+      "epoch": 0.08782633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841854200662852e-05,
-      "loss": 1.1733,
+      "learning_rate": 1.9825261743172486e-05,
+      "loss": 1.0625,
       "step": 3095
     },
     {
-      "epoch": 0.08502924940265305,
+      "epoch": 0.08785471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841696591656812e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.9825090638760963e-05,
+      "loss": 0.9854,
       "step": 3096
     },
     {
-      "epoch": 0.08505671363049627,
+      "epoch": 0.08788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841538904779453e-05,
-      "loss": 1.0259,
+      "learning_rate": 1.982491945135646e-05,
+      "loss": 1.0372,
       "step": 3097
     },
     {
-      "epoch": 0.08508417785833951,
+      "epoch": 0.08791146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841381140032033e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.9824748180960415e-05,
+      "loss": 1.0804,
       "step": 3098
     },
     {
-      "epoch": 0.08511164208618274,
+      "epoch": 0.08793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841223297415787e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.9824576827574287e-05,
+      "loss": 1.0005,
       "step": 3099
     },
     {
-      "epoch": 0.08513910631402598,
+      "epoch": 0.08796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9841065376931972e-05,
-      "loss": 1.1147,
+      "learning_rate": 1.9824405391199514e-05,
+      "loss": 0.8981,
       "step": 3100
     },
     {
-      "epoch": 0.08516657054186921,
+      "epoch": 0.08799659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840907378581834e-05,
-      "loss": 1.028,
+      "learning_rate": 1.9824233871837547e-05,
+      "loss": 0.9965,
       "step": 3101
     },
     {
-      "epoch": 0.08519403476971245,
+      "epoch": 0.0880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840749302366627e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.9824062269489836e-05,
+      "loss": 1.0287,
       "step": 3102
     },
     {
-      "epoch": 0.08522149899755568,
+      "epoch": 0.0880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840591148287598e-05,
-      "loss": 1.1617,
+      "learning_rate": 1.9823890584157828e-05,
+      "loss": 0.9384,
       "step": 3103
     },
     {
-      "epoch": 0.08524896322539892,
+      "epoch": 0.0880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840432916346e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.9823718815842975e-05,
+      "loss": 1.0393,
       "step": 3104
     },
     {
-      "epoch": 0.08527642745324215,
+      "epoch": 0.08811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840274606543088e-05,
-      "loss": 1.1258,
+      "learning_rate": 1.9823546964546727e-05,
+      "loss": 1.0001,
       "step": 3105
     },
     {
-      "epoch": 0.08530389168108539,
+      "epoch": 0.08813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.9840116218880107e-05,
-      "loss": 1.1186,
+      "learning_rate": 1.9823375030270537e-05,
+      "loss": 1.0969,
       "step": 3106
     },
     {
-      "epoch": 0.08533135590892862,
+      "epoch": 0.08816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.983995775335832e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.9823203013015856e-05,
+      "loss": 0.9513,
       "step": 3107
     },
     {
-      "epoch": 0.08535882013677186,
+      "epoch": 0.08819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.9839799209978973e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.982303091278414e-05,
+      "loss": 1.0456,
       "step": 3108
     },
     {
-      "epoch": 0.0853862843646151,
+      "epoch": 0.08822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.983964058874332e-05,
-      "loss": 1.1203,
+      "learning_rate": 1.9822858729576838e-05,
+      "loss": 1.0796,
       "step": 3109
     },
     {
-      "epoch": 0.08541374859245832,
+      "epoch": 0.08825198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 1.9839481889652623e-05,
-      "loss": 1.0788,
+      "learning_rate": 1.9822686463395406e-05,
+      "loss": 1.0314,
       "step": 3110
     },
     {
-      "epoch": 0.08544121282030155,
+      "epoch": 0.08828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9839323112708133e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.9822514114241302e-05,
+      "loss": 1.0544,
       "step": 3111
     },
     {
-      "epoch": 0.08546867704814479,
+      "epoch": 0.08830874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9839164257911107e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.982234168211598e-05,
+      "loss": 0.9525,
       "step": 3112
     },
     {
-      "epoch": 0.08549614127598802,
+      "epoch": 0.08833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.9839005325262805e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.9822169167020894e-05,
+      "loss": 1.0435,
       "step": 3113
     },
     {
-      "epoch": 0.08552360550383126,
+      "epoch": 0.08836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.9838846314764477e-05,
-      "loss": 1.1197,
+      "learning_rate": 1.9821996568957506e-05,
+      "loss": 1.115,
       "step": 3114
     },
     {
-      "epoch": 0.08555106973167449,
+      "epoch": 0.0883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9838687226417387e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.9821823887927264e-05,
+      "loss": 1.0748,
       "step": 3115
     },
     {
-      "epoch": 0.08557853395951773,
+      "epoch": 0.088422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.983852806022279e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.9821651123931643e-05,
+      "loss": 0.9722,
       "step": 3116
     },
     {
-      "epoch": 0.08560599818736096,
+      "epoch": 0.08845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.9838368816181953e-05,
-      "loss": 1.1573,
+      "learning_rate": 1.982147827697209e-05,
+      "loss": 1.0799,
       "step": 3117
     },
     {
-      "epoch": 0.0856334624152042,
+      "epoch": 0.08847900113507377,
       "grad_norm": 0.0,
-      "learning_rate": 1.983820949429613e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.982130534705007e-05,
+      "loss": 1.0217,
       "step": 3118
     },
     {
-      "epoch": 0.08566092664304743,
+      "epoch": 0.08850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.983805009456658e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.982113233416704e-05,
+      "loss": 1.0721,
       "step": 3119
     },
     {
-      "epoch": 0.08568839087089067,
+      "epoch": 0.08853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837890616994565e-05,
-      "loss": 1.069,
+      "learning_rate": 1.9820959238324463e-05,
+      "loss": 1.0549,
       "step": 3120
     },
     {
-      "epoch": 0.0857158550987339,
+      "epoch": 0.08856413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837731061581355e-05,
-      "loss": 1.0486,
+      "learning_rate": 1.9820786059523804e-05,
+      "loss": 0.9958,
       "step": 3121
     },
     {
-      "epoch": 0.08574331932657714,
+      "epoch": 0.08859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837571428328202e-05,
-      "loss": 1.1029,
+      "learning_rate": 1.9820612797766527e-05,
+      "loss": 1.2394,
       "step": 3122
     },
     {
-      "epoch": 0.08577078355442037,
+      "epoch": 0.08862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837411717236374e-05,
-      "loss": 1.0631,
+      "learning_rate": 1.9820439453054085e-05,
+      "loss": 1.0591,
       "step": 3123
     },
     {
-      "epoch": 0.0857982477822636,
+      "epoch": 0.08864926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837251928307134e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.9820266025387953e-05,
+      "loss": 0.9767,
       "step": 3124
     },
     {
-      "epoch": 0.08582571201010683,
+      "epoch": 0.08867763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837092061541746e-05,
-      "loss": 0.9597,
+      "learning_rate": 1.9820092514769595e-05,
+      "loss": 0.9785,
       "step": 3125
     },
     {
-      "epoch": 0.08585317623795007,
+      "epoch": 0.08870601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 1.9836932116941474e-05,
-      "loss": 1.077,
+      "learning_rate": 1.981991892120047e-05,
+      "loss": 1.1031,
       "step": 3126
     },
     {
-      "epoch": 0.0858806404657933,
+      "epoch": 0.08873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.9836772094507582e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.981974524468205e-05,
+      "loss": 1.0479,
       "step": 3127
     },
     {
-      "epoch": 0.08590810469363654,
+      "epoch": 0.0887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9836611994241343e-05,
-      "loss": 1.1129,
+      "learning_rate": 1.98195714852158e-05,
+      "loss": 1.0386,
       "step": 3128
     },
     {
-      "epoch": 0.08593556892147977,
+      "epoch": 0.08879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 1.983645181614402e-05,
-      "loss": 1.1175,
+      "learning_rate": 1.9819397642803187e-05,
+      "loss": 1.1217,
       "step": 3129
     },
     {
-      "epoch": 0.085963033149323,
+      "epoch": 0.08881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.983629156021688e-05,
-      "loss": 1.1403,
+      "learning_rate": 1.9819223717445686e-05,
+      "loss": 1.1127,
       "step": 3130
     },
     {
-      "epoch": 0.08599049737716624,
+      "epoch": 0.08884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.9836131226461187e-05,
-      "loss": 1.1586,
+      "learning_rate": 1.9819049709144754e-05,
+      "loss": 1.1587,
       "step": 3131
     },
     {
-      "epoch": 0.08601796160500948,
+      "epoch": 0.08887627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.983597081487822e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.9818875617901874e-05,
+      "loss": 0.9863,
       "step": 3132
     },
     {
-      "epoch": 0.08604542583285271,
+      "epoch": 0.08890465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835810325469237e-05,
-      "loss": 1.1964,
+      "learning_rate": 1.9818701443718504e-05,
+      "loss": 1.0681,
       "step": 3133
     },
     {
-      "epoch": 0.08607289006069595,
+      "epoch": 0.08893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835649758235513e-05,
-      "loss": 1.2438,
+      "learning_rate": 1.9818527186596128e-05,
+      "loss": 1.0673,
       "step": 3134
     },
     {
-      "epoch": 0.08610035428853918,
+      "epoch": 0.08896140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.983548911317832e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.9818352846536205e-05,
+      "loss": 1.0964,
       "step": 3135
     },
     {
-      "epoch": 0.08612781851638242,
+      "epoch": 0.08898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835328390298928e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.9818178423540217e-05,
+      "loss": 1.0964,
       "step": 3136
     },
     {
-      "epoch": 0.08615528274422565,
+      "epoch": 0.08901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835167589598604e-05,
-      "loss": 1.1102,
+      "learning_rate": 1.9818003917609637e-05,
+      "loss": 0.9451,
       "step": 3137
     },
     {
-      "epoch": 0.08618274697206887,
+      "epoch": 0.08904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.9835006711078625e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.9817829328745936e-05,
+      "loss": 1.0577,
       "step": 3138
     },
     {
-      "epoch": 0.08621021119991211,
+      "epoch": 0.08907491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834845754740266e-05,
-      "loss": 0.995,
+      "learning_rate": 1.9817654656950584e-05,
+      "loss": 1.0256,
       "step": 3139
     },
     {
-      "epoch": 0.08623767542775534,
+      "epoch": 0.0891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834684720584797e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.981747990222507e-05,
+      "loss": 1.0408,
       "step": 3140
     },
     {
-      "epoch": 0.08626513965559858,
+      "epoch": 0.0891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834523608613492e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9817305064570854e-05,
+      "loss": 1.149,
       "step": 3141
     },
     {
-      "epoch": 0.08629260388344182,
+      "epoch": 0.0891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834362418827625e-05,
-      "loss": 1.1121,
+      "learning_rate": 1.9817130143989424e-05,
+      "loss": 0.9412,
       "step": 3142
     },
     {
-      "epoch": 0.08632006811128505,
+      "epoch": 0.08918842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834201151228476e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.9816955140482258e-05,
+      "loss": 1.0599,
       "step": 3143
     },
     {
-      "epoch": 0.08634753233912829,
+      "epoch": 0.08921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9834039805817317e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.9816780054050824e-05,
+      "loss": 0.9804,
       "step": 3144
     },
     {
-      "epoch": 0.08637499656697152,
+      "epoch": 0.08924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9833878382595427e-05,
-      "loss": 1.1015,
+      "learning_rate": 1.981660488469661e-05,
+      "loss": 1.0039,
       "step": 3145
     },
     {
-      "epoch": 0.08640246079481476,
+      "epoch": 0.08927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.983371688156408e-05,
-      "loss": 1.081,
+      "learning_rate": 1.9816429632421095e-05,
+      "loss": 0.9846,
       "step": 3146
     },
     {
-      "epoch": 0.08642992502265799,
+      "epoch": 0.08930192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.9833555302724557e-05,
-      "loss": 1.0929,
+      "learning_rate": 1.9816254297225758e-05,
+      "loss": 1.2161,
       "step": 3147
     },
     {
-      "epoch": 0.08645738925050123,
+      "epoch": 0.08933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.9833393646078134e-05,
-      "loss": 1.1589,
+      "learning_rate": 1.981607887911208e-05,
+      "loss": 1.0774,
       "step": 3148
     },
     {
-      "epoch": 0.08648485347834446,
+      "epoch": 0.08935868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.9833231911626094e-05,
-      "loss": 1.1369,
+      "learning_rate": 1.9815903378081538e-05,
+      "loss": 0.9139,
       "step": 3149
     },
     {
-      "epoch": 0.0865123177061877,
+      "epoch": 0.08938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.983307009936971e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.9815727794135622e-05,
+      "loss": 0.9292,
       "step": 3150
     },
     {
-      "epoch": 0.08653978193403092,
+      "epoch": 0.08941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.983290820931027e-05,
-      "loss": 1.111,
+      "learning_rate": 1.9815552127275814e-05,
+      "loss": 1.116,
       "step": 3151
     },
     {
-      "epoch": 0.08656724616187415,
+      "epoch": 0.08944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.983274624144905e-05,
-      "loss": 1.1809,
+      "learning_rate": 1.9815376377503592e-05,
+      "loss": 0.9881,
       "step": 3152
     },
     {
-      "epoch": 0.08659471038971739,
+      "epoch": 0.08947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.9832584195787332e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.9815200544820444e-05,
+      "loss": 1.1057,
       "step": 3153
     },
     {
-      "epoch": 0.08662217461756062,
+      "epoch": 0.0895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9832422072326402e-05,
-      "loss": 1.1465,
+      "learning_rate": 1.981502462922786e-05,
+      "loss": 1.0333,
       "step": 3154
     },
     {
-      "epoch": 0.08664963884540386,
+      "epoch": 0.08952894438138478,
       "grad_norm": 0.0,
-      "learning_rate": 1.9832259871067535e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.9814848630727323e-05,
+      "loss": 1.1231,
       "step": 3155
     },
     {
-      "epoch": 0.0866771030732471,
+      "epoch": 0.08955732122587969,
       "grad_norm": 0.0,
-      "learning_rate": 1.9832097592012023e-05,
-      "loss": 1.1333,
+      "learning_rate": 1.981467254932031e-05,
+      "loss": 0.989,
       "step": 3156
     },
     {
-      "epoch": 0.08670456730109033,
+      "epoch": 0.08958569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.9831935235161145e-05,
-      "loss": 1.1211,
+      "learning_rate": 1.9814496385008323e-05,
+      "loss": 0.971,
       "step": 3157
     },
     {
-      "epoch": 0.08673203152893356,
+      "epoch": 0.08961407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.983177280051619e-05,
-      "loss": 1.0683,
+      "learning_rate": 1.9814320137792846e-05,
+      "loss": 1.0616,
       "step": 3158
     },
     {
-      "epoch": 0.0867594957567768,
+      "epoch": 0.08964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.9831610288078435e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.981414380767536e-05,
+      "loss": 0.9978,
       "step": 3159
     },
     {
-      "epoch": 0.08678695998462004,
+      "epoch": 0.08967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9831447697849174e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.9813967394657363e-05,
+      "loss": 1.0588,
       "step": 3160
     },
     {
-      "epoch": 0.08681442421246327,
+      "epoch": 0.08969920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.983128502982969e-05,
-      "loss": 1.0669,
+      "learning_rate": 1.981379089874034e-05,
+      "loss": 0.9516,
       "step": 3161
     },
     {
-      "epoch": 0.0868418884403065,
+      "epoch": 0.08972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.9831122284021268e-05,
-      "loss": 1.1565,
+      "learning_rate": 1.9813614319925785e-05,
+      "loss": 1.0363,
       "step": 3162
     },
     {
-      "epoch": 0.08686935266814974,
+      "epoch": 0.08975595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9830959460425204e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.9813437658215192e-05,
+      "loss": 0.9661,
       "step": 3163
     },
     {
-      "epoch": 0.08689681689599298,
+      "epoch": 0.08978433598183883,
       "grad_norm": 0.0,
-      "learning_rate": 1.9830796559042778e-05,
-      "loss": 1.0622,
+      "learning_rate": 1.9813260913610048e-05,
+      "loss": 1.0234,
       "step": 3164
     },
     {
-      "epoch": 0.0869242811238362,
+      "epoch": 0.08981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.983063357987528e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.9813084086111847e-05,
+      "loss": 0.998,
       "step": 3165
     },
     {
-      "epoch": 0.08695174535167943,
+      "epoch": 0.0898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.9830470522924e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.9812907175722085e-05,
+      "loss": 0.9939,
       "step": 3166
     },
     {
-      "epoch": 0.08697920957952267,
+      "epoch": 0.0898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9830307388190234e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.9812730182442253e-05,
+      "loss": 1.0566,
       "step": 3167
     },
     {
-      "epoch": 0.0870066738073659,
+      "epoch": 0.08989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.9830144175675262e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.9812553106273848e-05,
+      "loss": 1.0298,
       "step": 3168
     },
     {
-      "epoch": 0.08703413803520914,
+      "epoch": 0.08992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9829980885380385e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.9812375947218366e-05,
+      "loss": 1.013,
       "step": 3169
     },
     {
-      "epoch": 0.08706160226305237,
+      "epoch": 0.08995459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 1.982981751730689e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.9812198705277306e-05,
+      "loss": 1.0811,
       "step": 3170
     },
     {
-      "epoch": 0.08708906649089561,
+      "epoch": 0.08998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.9829654071456074e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.981202138045216e-05,
+      "loss": 1.0925,
       "step": 3171
     },
     {
-      "epoch": 0.08711653071873884,
+      "epoch": 0.09001135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.9829490547829224e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.981184397274443e-05,
+      "loss": 0.9519,
       "step": 3172
     },
     {
-      "epoch": 0.08714399494658208,
+      "epoch": 0.09003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.982932694642764e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.9811666482155608e-05,
+      "loss": 0.9943,
       "step": 3173
     },
     {
-      "epoch": 0.08717145917442531,
+      "epoch": 0.09006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.9829163267252612e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.9811488908687204e-05,
+      "loss": 1.1011,
       "step": 3174
     },
     {
-      "epoch": 0.08719892340226855,
+      "epoch": 0.09009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.982899951030544e-05,
-      "loss": 1.1716,
+      "learning_rate": 1.981131125234071e-05,
+      "loss": 1.041,
       "step": 3175
     },
     {
-      "epoch": 0.08722638763011178,
+      "epoch": 0.09012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.9828835675587412e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.981113351311763e-05,
+      "loss": 1.013,
       "step": 3176
     },
     {
-      "epoch": 0.08725385185795502,
+      "epoch": 0.09015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.982867176309983e-05,
-      "loss": 1.0859,
+      "learning_rate": 1.981095569101946e-05,
+      "loss": 0.9621,
       "step": 3177
     },
     {
-      "epoch": 0.08728131608579824,
+      "epoch": 0.0901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.9828507772843992e-05,
-      "loss": 1.0571,
+      "learning_rate": 1.981077778604771e-05,
+      "loss": 1.1559,
       "step": 3178
     },
     {
-      "epoch": 0.08730878031364148,
+      "epoch": 0.0902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.982834370482119e-05,
-      "loss": 1.1443,
+      "learning_rate": 1.981059979820388e-05,
+      "loss": 1.0059,
       "step": 3179
     },
     {
-      "epoch": 0.08733624454148471,
+      "epoch": 0.09023836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 1.9828179559032726e-05,
-      "loss": 1.1377,
+      "learning_rate": 1.981042172748947e-05,
+      "loss": 0.959,
       "step": 3180
     },
     {
-      "epoch": 0.08736370876932795,
+      "epoch": 0.09026674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.9828015335479903e-05,
-      "loss": 1.1432,
+      "learning_rate": 1.9810243573905987e-05,
+      "loss": 0.9784,
       "step": 3181
     },
     {
-      "epoch": 0.08739117299717118,
+      "epoch": 0.09029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.982785103416401e-05,
-      "loss": 1.101,
+      "learning_rate": 1.9810065337454934e-05,
+      "loss": 1.1113,
       "step": 3182
     },
     {
-      "epoch": 0.08741863722501442,
+      "epoch": 0.09032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9827686655086355e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.9809887018137824e-05,
+      "loss": 0.9549,
       "step": 3183
     },
     {
-      "epoch": 0.08744610145285765,
+      "epoch": 0.09035187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.9827522198248234e-05,
-      "loss": 1.1096,
+      "learning_rate": 1.9809708615956152e-05,
+      "loss": 1.056,
       "step": 3184
     },
     {
-      "epoch": 0.08747356568070089,
+      "epoch": 0.09038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9827357663650956e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.980953013091143e-05,
+      "loss": 1.1133,
       "step": 3185
     },
     {
-      "epoch": 0.08750102990854412,
+      "epoch": 0.09040862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9827193051295814e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.9809351563005168e-05,
+      "loss": 1.1522,
       "step": 3186
     },
     {
-      "epoch": 0.08752849413638736,
+      "epoch": 0.09043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.9827028361184114e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.9809172912238872e-05,
+      "loss": 1.0492,
       "step": 3187
     },
     {
-      "epoch": 0.0875559583642306,
+      "epoch": 0.09046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.9826863593317157e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.9808994178614054e-05,
+      "loss": 1.1964,
       "step": 3188
     },
     {
-      "epoch": 0.08758342259207383,
+      "epoch": 0.09049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.982669874769625e-05,
-      "loss": 1.0827,
+      "learning_rate": 1.9808815362132216e-05,
+      "loss": 1.0738,
       "step": 3189
     },
     {
-      "epoch": 0.08761088681991706,
+      "epoch": 0.09052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.9826533824322697e-05,
-      "loss": 1.1164,
+      "learning_rate": 1.980863646279488e-05,
+      "loss": 1.028,
       "step": 3190
     },
     {
-      "epoch": 0.0876383510477603,
+      "epoch": 0.09055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.9826368823197802e-05,
-      "loss": 1.0861,
+      "learning_rate": 1.9808457480603547e-05,
+      "loss": 1.0062,
       "step": 3191
     },
     {
-      "epoch": 0.08766581527560352,
+      "epoch": 0.0905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.982620374432287e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.9808278415559732e-05,
+      "loss": 1.0396,
       "step": 3192
     },
     {
-      "epoch": 0.08769327950344676,
+      "epoch": 0.0906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.9826038587699208e-05,
-      "loss": 1.1062,
+      "learning_rate": 1.980809926766495e-05,
+      "loss": 1.0601,
       "step": 3193
     },
     {
-      "epoch": 0.08772074373128999,
+      "epoch": 0.09063564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.982587335332812e-05,
-      "loss": 1.02,
+      "learning_rate": 1.980792003692071e-05,
+      "loss": 0.9476,
       "step": 3194
     },
     {
-      "epoch": 0.08774820795913323,
+      "epoch": 0.09066401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.982570804121092e-05,
-      "loss": 1.0449,
+      "learning_rate": 1.980774072332853e-05,
+      "loss": 1.0639,
       "step": 3195
     },
     {
-      "epoch": 0.08777567218697646,
+      "epoch": 0.09069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.982554265134891e-05,
-      "loss": 1.1821,
+      "learning_rate": 1.9807561326889926e-05,
+      "loss": 1.1099,
       "step": 3196
     },
     {
-      "epoch": 0.0878031364148197,
+      "epoch": 0.09072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.98253771837434e-05,
-      "loss": 1.1758,
+      "learning_rate": 1.980738184760641e-05,
+      "loss": 1.0213,
       "step": 3197
     },
     {
-      "epoch": 0.08783060064266293,
+      "epoch": 0.09074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.98252116383957e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.9807202285479493e-05,
+      "loss": 1.0658,
       "step": 3198
     },
     {
-      "epoch": 0.08785806487050617,
+      "epoch": 0.09077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.9825046015307115e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.9807022640510704e-05,
+      "loss": 1.1291,
       "step": 3199
     },
     {
-      "epoch": 0.0878855290983494,
+      "epoch": 0.09080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.982488031447897e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.980684291270155e-05,
+      "loss": 1.0424,
       "step": 3200
     },
     {
-      "epoch": 0.08791299332619264,
+      "epoch": 0.09083427922814984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9824714535912557e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.9806663102053555e-05,
+      "loss": 1.0102,
       "step": 3201
     },
     {
-      "epoch": 0.08794045755403587,
+      "epoch": 0.09086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9824548679609202e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.9806483208568234e-05,
+      "loss": 1.0331,
       "step": 3202
     },
     {
-      "epoch": 0.08796792178187911,
+      "epoch": 0.09089103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.982438274557021e-05,
-      "loss": 1.0613,
+      "learning_rate": 1.9806303232247112e-05,
+      "loss": 0.9719,
       "step": 3203
     },
     {
-      "epoch": 0.08799538600972234,
+      "epoch": 0.09091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.98242167337969e-05,
-      "loss": 1.0906,
+      "learning_rate": 1.9806123173091704e-05,
+      "loss": 1.0991,
       "step": 3204
     },
     {
-      "epoch": 0.08802285023756556,
+      "epoch": 0.0909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.982405064429058e-05,
-      "loss": 1.0612,
+      "learning_rate": 1.980594303110353e-05,
+      "loss": 1.0121,
       "step": 3205
     },
     {
-      "epoch": 0.0880503144654088,
+      "epoch": 0.09097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.982388447705256e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.980576280628412e-05,
+      "loss": 0.9644,
       "step": 3206
     },
     {
-      "epoch": 0.08807777869325203,
+      "epoch": 0.09100454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 1.9823718232084165e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.9805582498634987e-05,
+      "loss": 1.0246,
       "step": 3207
     },
     {
-      "epoch": 0.08810524292109527,
+      "epoch": 0.09103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.9823551909386705e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.9805402108157658e-05,
+      "loss": 0.9504,
       "step": 3208
     },
     {
-      "epoch": 0.0881327071489385,
+      "epoch": 0.09106129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.98233855089615e-05,
-      "loss": 1.1655,
+      "learning_rate": 1.9805221634853656e-05,
+      "loss": 1.1105,
       "step": 3209
     },
     {
-      "epoch": 0.08816017137678174,
+      "epoch": 0.09108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.982321903080986e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.980504107872451e-05,
+      "loss": 1.0808,
       "step": 3210
     },
     {
-      "epoch": 0.08818763560462498,
+      "epoch": 0.09111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.982305247493311e-05,
-      "loss": 1.0936,
+      "learning_rate": 1.9804860439771735e-05,
+      "loss": 1.1361,
       "step": 3211
     },
     {
-      "epoch": 0.08821509983246821,
+      "epoch": 0.09114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822885841332563e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.9804679717996866e-05,
+      "loss": 1.1819,
       "step": 3212
     },
     {
-      "epoch": 0.08824256406031145,
+      "epoch": 0.09117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822719130009536e-05,
-      "loss": 1.0689,
+      "learning_rate": 1.980449891340143e-05,
+      "loss": 1.0068,
       "step": 3213
     },
     {
-      "epoch": 0.08827002828815468,
+      "epoch": 0.09120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822552340965348e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.9804318025986946e-05,
+      "loss": 1.012,
       "step": 3214
     },
     {
-      "epoch": 0.08829749251599792,
+      "epoch": 0.09123155505107831,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822385474201327e-05,
-      "loss": 1.1177,
+      "learning_rate": 1.980413705575495e-05,
+      "loss": 1.0943,
       "step": 3215
     },
     {
-      "epoch": 0.08832495674384115,
+      "epoch": 0.09125993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822218529718783e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.9803956002706967e-05,
+      "loss": 1.0037,
       "step": 3216
     },
     {
-      "epoch": 0.08835242097168439,
+      "epoch": 0.0912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.982205150751904e-05,
-      "loss": 0.9358,
+      "learning_rate": 1.9803774866844527e-05,
+      "loss": 1.0651,
       "step": 3217
     },
     {
-      "epoch": 0.08837988519952762,
+      "epoch": 0.09131668558456299,
       "grad_norm": 0.0,
-      "learning_rate": 1.9821884407603425e-05,
-      "loss": 1.1482,
+      "learning_rate": 1.980359364816916e-05,
+      "loss": 1.0483,
       "step": 3218
     },
     {
-      "epoch": 0.08840734942737084,
+      "epoch": 0.09134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.9821717229973254e-05,
-      "loss": 1.1301,
+      "learning_rate": 1.9803412346682393e-05,
+      "loss": 1.0516,
       "step": 3219
     },
     {
-      "epoch": 0.08843481365521408,
+      "epoch": 0.09137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.9821549974629852e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.9803230962385766e-05,
+      "loss": 1.0642,
       "step": 3220
     },
     {
-      "epoch": 0.08846227788305731,
+      "epoch": 0.09140181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 1.9821382641574542e-05,
-      "loss": 1.1271,
+      "learning_rate": 1.9803049495280805e-05,
+      "loss": 1.0385,
       "step": 3221
     },
     {
-      "epoch": 0.08848974211090055,
+      "epoch": 0.09143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.982121523080865e-05,
-      "loss": 1.0931,
+      "learning_rate": 1.9802867945369044e-05,
+      "loss": 1.1652,
       "step": 3222
     },
     {
-      "epoch": 0.08851720633874378,
+      "epoch": 0.09145856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.9821047742333492e-05,
-      "loss": 1.1083,
+      "learning_rate": 1.980268631265202e-05,
+      "loss": 1.0547,
       "step": 3223
     },
     {
-      "epoch": 0.08854467056658702,
+      "epoch": 0.09148694665153236,
       "grad_norm": 0.0,
-      "learning_rate": 1.982088017615041e-05,
-      "loss": 1.1457,
+      "learning_rate": 1.9802504597131258e-05,
+      "loss": 1.058,
       "step": 3224
     },
     {
-      "epoch": 0.08857213479443025,
+      "epoch": 0.09151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9820712532260712e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.98023227988083e-05,
+      "loss": 1.0549,
       "step": 3225
     },
     {
-      "epoch": 0.08859959902227349,
+      "epoch": 0.09154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.9820544810665735e-05,
-      "loss": 1.1475,
+      "learning_rate": 1.980214091768468e-05,
+      "loss": 1.0141,
       "step": 3226
     },
     {
-      "epoch": 0.08862706325011672,
+      "epoch": 0.09157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.98203770113668e-05,
-      "loss": 1.111,
+      "learning_rate": 1.980195895376194e-05,
+      "loss": 1.0279,
       "step": 3227
     },
     {
-      "epoch": 0.08865452747795996,
+      "epoch": 0.09160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.982020913436524e-05,
-      "loss": 1.1215,
+      "learning_rate": 1.980177690704161e-05,
+      "loss": 0.989,
       "step": 3228
     },
     {
-      "epoch": 0.0886819917058032,
+      "epoch": 0.0916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.9820041179662385e-05,
-      "loss": 1.0313,
+      "learning_rate": 1.980159477752523e-05,
+      "loss": 1.0735,
       "step": 3229
     },
     {
-      "epoch": 0.08870945593364643,
+      "epoch": 0.09165720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 1.9819873147259557e-05,
-      "loss": 1.1664,
+      "learning_rate": 1.9801412565214338e-05,
+      "loss": 0.9756,
       "step": 3230
     },
     {
-      "epoch": 0.08873692016148967,
+      "epoch": 0.0916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.981970503715809e-05,
-      "loss": 1.1617,
+      "learning_rate": 1.9801230270110474e-05,
+      "loss": 1.0358,
       "step": 3231
     },
     {
-      "epoch": 0.08876438438933289,
+      "epoch": 0.09171396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 1.981953684935931e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.9801047892215176e-05,
+      "loss": 1.0862,
       "step": 3232
     },
     {
-      "epoch": 0.08879184861717612,
+      "epoch": 0.09174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.9819368583864555e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.9800865431529988e-05,
+      "loss": 1.0848,
       "step": 3233
     },
     {
-      "epoch": 0.08881931284501936,
+      "epoch": 0.09177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.981920024067515e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.9800682888056446e-05,
+      "loss": 0.9367,
       "step": 3234
     },
     {
-      "epoch": 0.08884677707286259,
+      "epoch": 0.09179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.9819031819792428e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.9800500261796096e-05,
+      "loss": 1.074,
       "step": 3235
     },
     {
-      "epoch": 0.08887424130070583,
+      "epoch": 0.09182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818863321217726e-05,
-      "loss": 1.074,
+      "learning_rate": 1.980031755275048e-05,
+      "loss": 1.1093,
       "step": 3236
     },
     {
-      "epoch": 0.08890170552854906,
+      "epoch": 0.09185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818694744952373e-05,
-      "loss": 1.2264,
+      "learning_rate": 1.9800134760921145e-05,
+      "loss": 1.0247,
       "step": 3237
     },
     {
-      "epoch": 0.0889291697563923,
+      "epoch": 0.09188422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818526090997703e-05,
-      "loss": 1.0728,
+      "learning_rate": 1.9799951886309624e-05,
+      "loss": 1.0486,
       "step": 3238
     },
     {
-      "epoch": 0.08895663398423553,
+      "epoch": 0.09191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818357359355048e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.9799768928917474e-05,
+      "loss": 1.036,
       "step": 3239
     },
     {
-      "epoch": 0.08898409821207877,
+      "epoch": 0.09194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818188550025754e-05,
-      "loss": 1.0157,
+      "learning_rate": 1.9799585888746235e-05,
+      "loss": 1.2516,
       "step": 3240
     },
     {
-      "epoch": 0.089011562439922,
+      "epoch": 0.09196935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.9818019663011144e-05,
-      "loss": 1.1211,
+      "learning_rate": 1.979940276579745e-05,
+      "loss": 1.0283,
       "step": 3241
     },
     {
-      "epoch": 0.08903902666776524,
+      "epoch": 0.09199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.9817850698312558e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9799219560072673e-05,
+      "loss": 1.0818,
       "step": 3242
     },
     {
-      "epoch": 0.08906649089560847,
+      "epoch": 0.0920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9817681655931333e-05,
-      "loss": 1.1442,
+      "learning_rate": 1.9799036271573447e-05,
+      "loss": 1.1303,
       "step": 3243
     },
     {
-      "epoch": 0.08909395512345171,
+      "epoch": 0.0920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.981751253586881e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.979885290030132e-05,
+      "loss": 1.0066,
       "step": 3244
     },
     {
-      "epoch": 0.08912141935129494,
+      "epoch": 0.09208286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 1.981734333812633e-05,
-      "loss": 1.1537,
+      "learning_rate": 1.9798669446257844e-05,
+      "loss": 0.9967,
       "step": 3245
     },
     {
-      "epoch": 0.08914888357913817,
+      "epoch": 0.09211123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 1.981717406270522e-05,
-      "loss": 1.108,
+      "learning_rate": 1.9798485909444563e-05,
+      "loss": 1.0948,
       "step": 3246
     },
     {
-      "epoch": 0.0891763478069814,
+      "epoch": 0.09213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.981700470960683e-05,
-      "loss": 1.1076,
+      "learning_rate": 1.9798302289863035e-05,
+      "loss": 1.0991,
       "step": 3247
     },
     {
-      "epoch": 0.08920381203482464,
+      "epoch": 0.09216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.981683527883249e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.9798118587514804e-05,
+      "loss": 1.0377,
       "step": 3248
     },
     {
-      "epoch": 0.08923127626266787,
+      "epoch": 0.09219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.981666577038355e-05,
-      "loss": 0.9941,
+      "learning_rate": 1.9797934802401425e-05,
+      "loss": 1.0662,
       "step": 3249
     },
     {
-      "epoch": 0.08925874049051111,
+      "epoch": 0.09222474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.981649618426135e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.9797750934524453e-05,
+      "loss": 0.9948,
       "step": 3250
     },
     {
-      "epoch": 0.08928620471835434,
+      "epoch": 0.09225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.9816326520467228e-05,
-      "loss": 1.1688,
+      "learning_rate": 1.9797566983885437e-05,
+      "loss": 1.0279,
       "step": 3251
     },
     {
-      "epoch": 0.08931366894619758,
+      "epoch": 0.09228149829738932,
       "grad_norm": 0.0,
-      "learning_rate": 1.9816156779002532e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.979738295048593e-05,
+      "loss": 1.028,
       "step": 3252
     },
     {
-      "epoch": 0.08934113317404081,
+      "epoch": 0.09230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9815986959868594e-05,
-      "loss": 1.0062,
+      "learning_rate": 1.9797198834327492e-05,
+      "loss": 1.099,
       "step": 3253
     },
     {
-      "epoch": 0.08936859740188405,
+      "epoch": 0.09233825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.981581706306677e-05,
-      "loss": 1.0706,
+      "learning_rate": 1.9797014635411674e-05,
+      "loss": 1.0753,
       "step": 3254
     },
     {
-      "epoch": 0.08939606162972728,
+      "epoch": 0.092366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9815647088598395e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.9796830353740036e-05,
+      "loss": 1.086,
       "step": 3255
     },
     {
-      "epoch": 0.08942352585757052,
+      "epoch": 0.0923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.981547703646482e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.9796645989314126e-05,
+      "loss": 0.9627,
       "step": 3256
     },
     {
-      "epoch": 0.08945099008541375,
+      "epoch": 0.09242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.981530690666739e-05,
-      "loss": 1.0946,
+      "learning_rate": 1.979646154213551e-05,
+      "loss": 1.0598,
       "step": 3257
     },
     {
-      "epoch": 0.08947845431325699,
+      "epoch": 0.09245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.9815136699207447e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.9796277012205744e-05,
+      "loss": 1.0856,
       "step": 3258
     },
     {
-      "epoch": 0.08950591854110021,
+      "epoch": 0.09248013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814966414086344e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.9796092399526383e-05,
+      "loss": 1.0626,
       "step": 3259
     },
     {
-      "epoch": 0.08953338276894345,
+      "epoch": 0.09250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814796051305424e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.979590770409899e-05,
+      "loss": 0.9382,
       "step": 3260
     },
     {
-      "epoch": 0.08956084699678668,
+      "epoch": 0.09253688989784337,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814625610866033e-05,
-      "loss": 1.0876,
+      "learning_rate": 1.9795722925925126e-05,
+      "loss": 0.9819,
       "step": 3261
     },
     {
-      "epoch": 0.08958831122462992,
+      "epoch": 0.09256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814455092769523e-05,
-      "loss": 1.0865,
+      "learning_rate": 1.9795538065006348e-05,
+      "loss": 0.9449,
       "step": 3262
     },
     {
-      "epoch": 0.08961577545247315,
+      "epoch": 0.09259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814284497017243e-05,
-      "loss": 1.0394,
+      "learning_rate": 1.979535312134422e-05,
+      "loss": 1.0308,
       "step": 3263
     },
     {
-      "epoch": 0.08964323968031639,
+      "epoch": 0.09262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9814113823610544e-05,
-      "loss": 1.0671,
+      "learning_rate": 1.9795168094940303e-05,
+      "loss": 1.077,
       "step": 3264
     },
     {
-      "epoch": 0.08967070390815962,
+      "epoch": 0.09265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9813943072550775e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.979498298579616e-05,
+      "loss": 1.1018,
       "step": 3265
     },
     {
-      "epoch": 0.08969816813600286,
+      "epoch": 0.09267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.981377224383928e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.979479779391336e-05,
+      "loss": 1.1416,
       "step": 3266
     },
     {
-      "epoch": 0.08972563236384609,
+      "epoch": 0.09270715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 1.9813601337477427e-05,
-      "loss": 1.0667,
+      "learning_rate": 1.9794612519293462e-05,
+      "loss": 1.0382,
       "step": 3267
     },
     {
-      "epoch": 0.08975309659168933,
+      "epoch": 0.0927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9813430353466555e-05,
-      "loss": 1.153,
+      "learning_rate": 1.9794427161938027e-05,
+      "loss": 1.0165,
       "step": 3268
     },
     {
-      "epoch": 0.08978056081953256,
+      "epoch": 0.09276390465380249,
       "grad_norm": 0.0,
-      "learning_rate": 1.9813259291808017e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.979424172184863e-05,
+      "loss": 0.9869,
       "step": 3269
     },
     {
-      "epoch": 0.0898080250473758,
+      "epoch": 0.0927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.9813088152503175e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.979405619902683e-05,
+      "loss": 0.9706,
       "step": 3270
     },
     {
-      "epoch": 0.08983548927521903,
+      "epoch": 0.09282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.9812916935553377e-05,
-      "loss": 1.0375,
+      "learning_rate": 1.9793870593474202e-05,
+      "loss": 1.0291,
       "step": 3271
     },
     {
-      "epoch": 0.08986295350306227,
+      "epoch": 0.09284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.981274564095998e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.9793684905192302e-05,
+      "loss": 0.9843,
       "step": 3272
     },
     {
-      "epoch": 0.08989041773090549,
+      "epoch": 0.09287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.981257426872434e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.979349913418271e-05,
+      "loss": 0.9858,
       "step": 3273
     },
     {
-      "epoch": 0.08991788195874872,
+      "epoch": 0.09290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.9812402818847806e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.979331328044699e-05,
+      "loss": 1.2099,
       "step": 3274
     },
     {
-      "epoch": 0.08994534618659196,
+      "epoch": 0.09293416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.9812231291331742e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.979312734398671e-05,
+      "loss": 1.172,
       "step": 3275
     },
     {
-      "epoch": 0.0899728104144352,
+      "epoch": 0.09296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9812059686177502e-05,
-      "loss": 1.023,
+      "learning_rate": 1.9792941324803444e-05,
+      "loss": 0.9655,
       "step": 3276
     },
     {
-      "epoch": 0.09000027464227843,
+      "epoch": 0.09299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.9811888003386448e-05,
-      "loss": 1.0824,
+      "learning_rate": 1.979275522289876e-05,
+      "loss": 1.0391,
       "step": 3277
     },
     {
-      "epoch": 0.09002773887012167,
+      "epoch": 0.09301929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9811716242959935e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.979256903827424e-05,
+      "loss": 1.1383,
       "step": 3278
     },
     {
-      "epoch": 0.0900552030979649,
+      "epoch": 0.09304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.981154440489932e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.9792382770931443e-05,
+      "loss": 1.0054,
       "step": 3279
     },
     {
-      "epoch": 0.09008266732580814,
+      "epoch": 0.09307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.9811372489205967e-05,
-      "loss": 1.1681,
+      "learning_rate": 1.979219642087195e-05,
+      "loss": 1.0399,
       "step": 3280
     },
     {
-      "epoch": 0.09011013155365137,
+      "epoch": 0.09310442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 1.9811200495881234e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.9792009988097334e-05,
+      "loss": 1.1169,
       "step": 3281
     },
     {
-      "epoch": 0.0901375957814946,
+      "epoch": 0.0931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9811028424926482e-05,
-      "loss": 1.1281,
+      "learning_rate": 1.9791823472609167e-05,
+      "loss": 1.0315,
       "step": 3282
     },
     {
-      "epoch": 0.09016506000933784,
+      "epoch": 0.09316118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9810856276343073e-05,
-      "loss": 1.0986,
+      "learning_rate": 1.9791636874409032e-05,
+      "loss": 1.0413,
       "step": 3283
     },
     {
-      "epoch": 0.09019252423718108,
+      "epoch": 0.09318955732122589,
       "grad_norm": 0.0,
-      "learning_rate": 1.981068405013237e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.97914501934985e-05,
+      "loss": 1.0582,
       "step": 3284
     },
     {
-      "epoch": 0.09021998846502431,
+      "epoch": 0.09321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.9810511746295732e-05,
-      "loss": 1.1156,
+      "learning_rate": 1.979126342987914e-05,
+      "loss": 1.0698,
       "step": 3285
     },
     {
-      "epoch": 0.09024745269286755,
+      "epoch": 0.09324631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.9810339364834525e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.9791076583552543e-05,
+      "loss": 1.1348,
       "step": 3286
     },
     {
-      "epoch": 0.09027491692071077,
+      "epoch": 0.09327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9810166905750114e-05,
-      "loss": 1.0817,
+      "learning_rate": 1.9790889654520282e-05,
+      "loss": 1.1465,
       "step": 3287
     },
     {
-      "epoch": 0.090302381148554,
+      "epoch": 0.09330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.980999436904386e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.9790702642783938e-05,
+      "loss": 1.1039,
       "step": 3288
     },
     {
-      "epoch": 0.09032984537639724,
+      "epoch": 0.09333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9809821754717132e-05,
-      "loss": 1.0173,
+      "learning_rate": 1.9790515548345085e-05,
+      "loss": 1.0408,
       "step": 3289
     },
     {
-      "epoch": 0.09035730960424047,
+      "epoch": 0.09335981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.9809649062771296e-05,
-      "loss": 1.0134,
+      "learning_rate": 1.979032837120531e-05,
+      "loss": 1.1668,
       "step": 3290
     },
     {
-      "epoch": 0.09038477383208371,
+      "epoch": 0.09338819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.9809476293207717e-05,
-      "loss": 1.1054,
+      "learning_rate": 1.979014111136619e-05,
+      "loss": 1.1219,
       "step": 3291
     },
     {
-      "epoch": 0.09041223805992694,
+      "epoch": 0.09341657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 1.980930344602776e-05,
-      "loss": 1.1785,
+      "learning_rate": 1.9789953768829306e-05,
+      "loss": 1.1169,
       "step": 3292
     },
     {
-      "epoch": 0.09043970228777018,
+      "epoch": 0.09344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.9809130521232795e-05,
-      "loss": 1.1816,
+      "learning_rate": 1.9789766343596244e-05,
+      "loss": 1.0452,
       "step": 3293
     },
     {
-      "epoch": 0.09046716651561341,
+      "epoch": 0.0934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9808957518824186e-05,
-      "loss": 1.1009,
+      "learning_rate": 1.978957883566859e-05,
+      "loss": 0.9305,
       "step": 3294
     },
     {
-      "epoch": 0.09049463074345665,
+      "epoch": 0.09350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.980878443880331e-05,
-      "loss": 1.1542,
+      "learning_rate": 1.9789391245047915e-05,
+      "loss": 1.0405,
       "step": 3295
     },
     {
-      "epoch": 0.09052209497129989,
+      "epoch": 0.09353007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.9808611281171532e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.9789203571735816e-05,
+      "loss": 1.0965,
       "step": 3296
     },
     {
-      "epoch": 0.09054955919914312,
+      "epoch": 0.09355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.9808438045930218e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.9789015815733878e-05,
+      "loss": 1.0203,
       "step": 3297
     },
     {
-      "epoch": 0.09057702342698636,
+      "epoch": 0.09358683314415436,
       "grad_norm": 0.0,
-      "learning_rate": 1.9808264733080747e-05,
-      "loss": 0.897,
+      "learning_rate": 1.978882797704368e-05,
+      "loss": 1.0753,
       "step": 3298
     },
     {
-      "epoch": 0.09060448765482959,
+      "epoch": 0.09361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.980809134262448e-05,
-      "loss": 1.0322,
+      "learning_rate": 1.978864005566681e-05,
+      "loss": 1.1902,
       "step": 3299
     },
     {
-      "epoch": 0.09063195188267281,
+      "epoch": 0.09364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9807917874562803e-05,
-      "loss": 1.1071,
+      "learning_rate": 1.978845205160486e-05,
+      "loss": 1.1082,
       "step": 3300
     },
     {
-      "epoch": 0.09065941611051605,
+      "epoch": 0.09367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.9807744328897076e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.9788263964859415e-05,
+      "loss": 1.0112,
       "step": 3301
     },
     {
-      "epoch": 0.09068688033835928,
+      "epoch": 0.09370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.980757070562868e-05,
-      "loss": 1.1699,
+      "learning_rate": 1.9788075795432065e-05,
+      "loss": 0.9495,
       "step": 3302
     },
     {
-      "epoch": 0.09071434456620252,
+      "epoch": 0.09372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.980739700475898e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.9787887543324397e-05,
+      "loss": 1.0324,
       "step": 3303
     },
     {
-      "epoch": 0.09074180879404575,
+      "epoch": 0.09375709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9807223226289363e-05,
-      "loss": 1.0686,
+      "learning_rate": 1.9787699208538e-05,
+      "loss": 1.1414,
       "step": 3304
     },
     {
-      "epoch": 0.09076927302188899,
+      "epoch": 0.09378547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.980704937022119e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.9787510791074475e-05,
+      "loss": 1.1165,
       "step": 3305
     },
     {
-      "epoch": 0.09079673724973222,
+      "epoch": 0.0938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806875436555848e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.9787322290935403e-05,
+      "loss": 1.0354,
       "step": 3306
     },
     {
-      "epoch": 0.09082420147757546,
+      "epoch": 0.0938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806701425294705e-05,
-      "loss": 1.0579,
+      "learning_rate": 1.978713370812238e-05,
+      "loss": 1.1451,
       "step": 3307
     },
     {
-      "epoch": 0.0908516657054187,
+      "epoch": 0.09387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806527336439147e-05,
-      "loss": 1.0942,
+      "learning_rate": 1.9786945042636998e-05,
+      "loss": 1.0158,
       "step": 3308
     },
     {
-      "epoch": 0.09087912993326193,
+      "epoch": 0.09389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806353169990543e-05,
-      "loss": 1.1126,
+      "learning_rate": 1.9786756294480855e-05,
+      "loss": 1.0743,
       "step": 3309
     },
     {
-      "epoch": 0.09090659416110516,
+      "epoch": 0.09392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806178925950272e-05,
-      "loss": 1.1514,
+      "learning_rate": 1.9786567463655536e-05,
+      "loss": 1.0896,
       "step": 3310
     },
     {
-      "epoch": 0.0909340583889484,
+      "epoch": 0.09395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9806004604319716e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.9786378550162645e-05,
+      "loss": 1.0375,
       "step": 3311
     },
     {
-      "epoch": 0.09096152261679163,
+      "epoch": 0.09398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.9805830205100254e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.9786189554003773e-05,
+      "loss": 1.0253,
       "step": 3312
     },
     {
-      "epoch": 0.09098898684463487,
+      "epoch": 0.09401248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9805655728293266e-05,
-      "loss": 1.1402,
+      "learning_rate": 1.9786000475180516e-05,
+      "loss": 1.0197,
       "step": 3313
     },
     {
-      "epoch": 0.09101645107247809,
+      "epoch": 0.09404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.980548117390013e-05,
-      "loss": 1.0378,
+      "learning_rate": 1.9785811313694477e-05,
+      "loss": 1.0563,
       "step": 3314
     },
     {
-      "epoch": 0.09104391530032133,
+      "epoch": 0.09406923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 1.9805306541922228e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.978562206954725e-05,
+      "loss": 1.0118,
       "step": 3315
     },
     {
-      "epoch": 0.09107137952816456,
+      "epoch": 0.09409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9805131832360943e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.978543274274043e-05,
+      "loss": 1.083,
       "step": 3316
     },
     {
-      "epoch": 0.0910988437560078,
+      "epoch": 0.09412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9804957045217658e-05,
-      "loss": 1.1909,
+      "learning_rate": 1.9785243333275622e-05,
+      "loss": 0.9782,
       "step": 3317
     },
     {
-      "epoch": 0.09112630798385103,
+      "epoch": 0.0941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9804782180493754e-05,
-      "loss": 1.0108,
+      "learning_rate": 1.9785053841154426e-05,
+      "loss": 1.0338,
       "step": 3318
     },
     {
-      "epoch": 0.09115377221169427,
+      "epoch": 0.09418274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 1.9804607238190617e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9784864266378434e-05,
+      "loss": 1.0688,
       "step": 3319
     },
     {
-      "epoch": 0.0911812364395375,
+      "epoch": 0.094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9804432218309627e-05,
-      "loss": 1.0914,
+      "learning_rate": 1.9784674608949258e-05,
+      "loss": 0.9292,
       "step": 3320
     },
     {
-      "epoch": 0.09120870066738074,
+      "epoch": 0.0942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.980425712085217e-05,
-      "loss": 1.1237,
+      "learning_rate": 1.9784484868868494e-05,
+      "loss": 0.9317,
       "step": 3321
     },
     {
-      "epoch": 0.09123616489522397,
+      "epoch": 0.09426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.980408194581964e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.978429504613775e-05,
+      "loss": 1.0899,
       "step": 3322
     },
     {
-      "epoch": 0.09126362912306721,
+      "epoch": 0.09429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.980390669321341e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.9784105140758623e-05,
+      "loss": 1.1006,
       "step": 3323
     },
     {
-      "epoch": 0.09129109335091044,
+      "epoch": 0.09432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.980373136303487e-05,
-      "loss": 1.1343,
+      "learning_rate": 1.978391515273272e-05,
+      "loss": 1.2171,
       "step": 3324
     },
     {
-      "epoch": 0.09131855757875368,
+      "epoch": 0.09435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.9803555955285416e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.9783725082061646e-05,
+      "loss": 1.0463,
       "step": 3325
     },
     {
-      "epoch": 0.09134602180659691,
+      "epoch": 0.09438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.980338046996643e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.9783534928747006e-05,
+      "loss": 0.958,
       "step": 3326
     },
     {
-      "epoch": 0.09137348603444014,
+      "epoch": 0.09440976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.9803204907079294e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.9783344692790407e-05,
+      "loss": 0.9923,
       "step": 3327
     },
     {
-      "epoch": 0.09140095026228337,
+      "epoch": 0.09443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9803029266625405e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.9783154374193455e-05,
+      "loss": 1.1509,
       "step": 3328
     },
     {
-      "epoch": 0.0914284144901266,
+      "epoch": 0.09446651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 1.9802853548606155e-05,
-      "loss": 1.1197,
+      "learning_rate": 1.978296397295776e-05,
+      "loss": 1.0559,
       "step": 3329
     },
     {
-      "epoch": 0.09145587871796984,
+      "epoch": 0.09449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.980267775302293e-05,
-      "loss": 1.0793,
+      "learning_rate": 1.9782773489084927e-05,
+      "loss": 0.9695,
       "step": 3330
     },
     {
-      "epoch": 0.09148334294581308,
+      "epoch": 0.09452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.980250187987712e-05,
-      "loss": 1.1887,
+      "learning_rate": 1.9782582922576567e-05,
+      "loss": 1.0674,
       "step": 3331
     },
     {
-      "epoch": 0.09151080717365631,
+      "epoch": 0.0945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9802325929170115e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.978239227343429e-05,
+      "loss": 1.019,
       "step": 3332
     },
     {
-      "epoch": 0.09153827140149955,
+      "epoch": 0.0945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9802149900903313e-05,
-      "loss": 1.0202,
+      "learning_rate": 1.9782201541659705e-05,
+      "loss": 1.0071,
       "step": 3333
     },
     {
-      "epoch": 0.09156573562934278,
+      "epoch": 0.09460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801973795078105e-05,
-      "loss": 1.1938,
+      "learning_rate": 1.9782010727254427e-05,
+      "loss": 1.0502,
       "step": 3334
     },
     {
-      "epoch": 0.09159319985718602,
+      "epoch": 0.09463677639046537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801797611695884e-05,
-      "loss": 1.1318,
+      "learning_rate": 1.9781819830220058e-05,
+      "loss": 1.0178,
       "step": 3335
     },
     {
-      "epoch": 0.09162066408502925,
+      "epoch": 0.09466515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 1.980162135075804e-05,
-      "loss": 1.0275,
+      "learning_rate": 1.9781628850558224e-05,
+      "loss": 1.0819,
       "step": 3336
     },
     {
-      "epoch": 0.09164812831287249,
+      "epoch": 0.09469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801445012265975e-05,
-      "loss": 1.2257,
+      "learning_rate": 1.978143778827053e-05,
+      "loss": 1.0248,
       "step": 3337
     },
     {
-      "epoch": 0.09167559254071572,
+      "epoch": 0.09472190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801268596221076e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.978124664335859e-05,
+      "loss": 0.953,
       "step": 3338
     },
     {
-      "epoch": 0.09170305676855896,
+      "epoch": 0.09475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801092102624747e-05,
-      "loss": 1.121,
+      "learning_rate": 1.9781055415824015e-05,
+      "loss": 1.014,
       "step": 3339
     },
     {
-      "epoch": 0.09173052099640219,
+      "epoch": 0.09477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9800915531478378e-05,
-      "loss": 1.1469,
+      "learning_rate": 1.978086410566843e-05,
+      "loss": 1.0883,
       "step": 3340
     },
     {
-      "epoch": 0.09175798522424541,
+      "epoch": 0.09480703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 1.980073888278337e-05,
-      "loss": 1.147,
+      "learning_rate": 1.9780672712893447e-05,
+      "loss": 1.0946,
       "step": 3341
     },
     {
-      "epoch": 0.09178544945208865,
+      "epoch": 0.09483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.9800562156541122e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.9780481237500682e-05,
+      "loss": 1.2756,
       "step": 3342
     },
     {
-      "epoch": 0.09181291367993188,
+      "epoch": 0.09486379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 1.9800385352753027e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.9780289679491752e-05,
+      "loss": 0.902,
       "step": 3343
     },
     {
-      "epoch": 0.09184037790777512,
+      "epoch": 0.09489216799091942,
       "grad_norm": 0.0,
-      "learning_rate": 1.980020847142049e-05,
-      "loss": 1.0332,
+      "learning_rate": 1.978009803886827e-05,
+      "loss": 1.1265,
       "step": 3344
     },
     {
-      "epoch": 0.09186784213561835,
+      "epoch": 0.0949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9800031512544903e-05,
-      "loss": 1.1068,
+      "learning_rate": 1.977990631563187e-05,
+      "loss": 0.9805,
       "step": 3345
     },
     {
-      "epoch": 0.09189530636346159,
+      "epoch": 0.09494892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 1.9799854476127673e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.9779714509784155e-05,
+      "loss": 0.927,
       "step": 3346
     },
     {
-      "epoch": 0.09192277059130483,
+      "epoch": 0.09497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.9799677362170194e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.9779522621326754e-05,
+      "loss": 1.0476,
       "step": 3347
     },
     {
-      "epoch": 0.09195023481914806,
+      "epoch": 0.09500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.9799500170673875e-05,
-      "loss": 1.0113,
+      "learning_rate": 1.9779330650261282e-05,
+      "loss": 1.0313,
       "step": 3348
     },
     {
-      "epoch": 0.0919776990469913,
+      "epoch": 0.09503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.9799322901640113e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.977913859658937e-05,
+      "loss": 1.016,
       "step": 3349
     },
     {
-      "epoch": 0.09200516327483453,
+      "epoch": 0.09506242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9799145555070317e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.9778946460312632e-05,
+      "loss": 1.0275,
       "step": 3350
     },
     {
-      "epoch": 0.09203262750267777,
+      "epoch": 0.09509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.979896813096588e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.9778754241432696e-05,
+      "loss": 1.0253,
       "step": 3351
     },
     {
-      "epoch": 0.092060091730521,
+      "epoch": 0.09511918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.9798790629328217e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.9778561939951178e-05,
+      "loss": 1.0728,
       "step": 3352
     },
     {
-      "epoch": 0.09208755595836424,
+      "epoch": 0.09514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9798613050158722e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.9778369555869714e-05,
+      "loss": 1.138,
       "step": 3353
     },
     {
-      "epoch": 0.09211502018620746,
+      "epoch": 0.09517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.9798435393458807e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.9778177089189917e-05,
+      "loss": 1.1155,
       "step": 3354
     },
     {
-      "epoch": 0.0921424844140507,
+      "epoch": 0.09520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.9798257659229875e-05,
-      "loss": 1.1478,
+      "learning_rate": 1.9777984539913423e-05,
+      "loss": 1.1815,
       "step": 3355
     },
     {
-      "epoch": 0.09216994864189393,
+      "epoch": 0.09523269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.9798079847473335e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.977779190804185e-05,
+      "loss": 1.1064,
       "step": 3356
     },
     {
-      "epoch": 0.09219741286973716,
+      "epoch": 0.095261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9797901958190587e-05,
-      "loss": 1.1338,
+      "learning_rate": 1.9777599193576833e-05,
+      "loss": 1.129,
       "step": 3357
     },
     {
-      "epoch": 0.0922248770975804,
+      "epoch": 0.0952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.979772399138305e-05,
-      "loss": 0.952,
+      "learning_rate": 1.9777406396519997e-05,
+      "loss": 1.0394,
       "step": 3358
     },
     {
-      "epoch": 0.09225234132542363,
+      "epoch": 0.0953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.979754594705212e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.9777213516872965e-05,
+      "loss": 1.1003,
       "step": 3359
     },
     {
-      "epoch": 0.09227980555326687,
+      "epoch": 0.09534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.9797367825199214e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.9777020554637376e-05,
+      "loss": 1.0303,
       "step": 3360
     },
     {
-      "epoch": 0.0923072697811101,
+      "epoch": 0.09537457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.979718962582574e-05,
-      "loss": 1.1758,
+      "learning_rate": 1.977682750981485e-05,
+      "loss": 0.9908,
       "step": 3361
     },
     {
-      "epoch": 0.09233473400895334,
+      "epoch": 0.09540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9797011348933105e-05,
-      "loss": 1.0937,
+      "learning_rate": 1.9776634382407027e-05,
+      "loss": 1.0547,
       "step": 3362
     },
     {
-      "epoch": 0.09236219823679657,
+      "epoch": 0.09543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.979683299452272e-05,
-      "loss": 0.9698,
+      "learning_rate": 1.977644117241553e-05,
+      "loss": 1.0215,
       "step": 3363
     },
     {
-      "epoch": 0.09238966246463981,
+      "epoch": 0.09545970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.9796654562595998e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.9776247879841997e-05,
+      "loss": 1.0409,
       "step": 3364
     },
     {
-      "epoch": 0.09241712669248305,
+      "epoch": 0.09548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.9796476053154347e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.9776054504688057e-05,
+      "loss": 1.0162,
       "step": 3365
     },
     {
-      "epoch": 0.09244459092032628,
+      "epoch": 0.09551645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.979629746619919e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.9775861046955347e-05,
+      "loss": 1.1134,
       "step": 3366
     },
     {
-      "epoch": 0.09247205514816952,
+      "epoch": 0.09554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.979611880173193e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.9775667506645497e-05,
+      "loss": 1.2161,
       "step": 3367
     },
     {
-      "epoch": 0.09249951937601274,
+      "epoch": 0.09557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9795940059753977e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.9775473883760147e-05,
+      "loss": 1.0745,
       "step": 3368
     },
     {
-      "epoch": 0.09252698360385597,
+      "epoch": 0.09560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.979576124026676e-05,
-      "loss": 1.1482,
+      "learning_rate": 1.977528017830093e-05,
+      "loss": 1.039,
       "step": 3369
     },
     {
-      "epoch": 0.09255444783169921,
+      "epoch": 0.09562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.979558234327168e-05,
-      "loss": 1.0771,
+      "learning_rate": 1.9775086390269476e-05,
+      "loss": 1.0899,
       "step": 3370
     },
     {
-      "epoch": 0.09258191205954244,
+      "epoch": 0.0956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9795403368770164e-05,
-      "loss": 1.1435,
+      "learning_rate": 1.9774892519667433e-05,
+      "loss": 1.045,
       "step": 3371
     },
     {
-      "epoch": 0.09260937628738568,
+      "epoch": 0.09568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.9795224316763616e-05,
-      "loss": 1.1344,
+      "learning_rate": 1.9774698566496435e-05,
+      "loss": 1.0668,
       "step": 3372
     },
     {
-      "epoch": 0.09263684051522891,
+      "epoch": 0.09571509648127129,
       "grad_norm": 0.0,
-      "learning_rate": 1.9795045187253464e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.9774504530758115e-05,
+      "loss": 1.0602,
       "step": 3373
     },
     {
-      "epoch": 0.09266430474307215,
+      "epoch": 0.09574347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9794865980241116e-05,
-      "loss": 1.0134,
+      "learning_rate": 1.9774310412454117e-05,
+      "loss": 1.0526,
       "step": 3374
     },
     {
-      "epoch": 0.09269176897091538,
+      "epoch": 0.09577185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.9794686695727994e-05,
-      "loss": 1.1521,
+      "learning_rate": 1.977411621158608e-05,
+      "loss": 1.0491,
       "step": 3375
     },
     {
-      "epoch": 0.09271923319875862,
+      "epoch": 0.09580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.979450733371552e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.9773921928155645e-05,
+      "loss": 0.9114,
       "step": 3376
     },
     {
-      "epoch": 0.09274669742660185,
+      "epoch": 0.09582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.9794327894205108e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.977372756216445e-05,
+      "loss": 1.1138,
       "step": 3377
     },
     {
-      "epoch": 0.09277416165444509,
+      "epoch": 0.09585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.979414837719818e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.9773533113614142e-05,
+      "loss": 1.0186,
       "step": 3378
     },
     {
-      "epoch": 0.09280162588228832,
+      "epoch": 0.09588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.979396878269616e-05,
-      "loss": 1.1727,
+      "learning_rate": 1.9773338582506357e-05,
+      "loss": 1.0389,
       "step": 3379
     },
     {
-      "epoch": 0.09282909011013156,
+      "epoch": 0.09591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.9793789110700457e-05,
-      "loss": 1.1138,
+      "learning_rate": 1.9773143968842743e-05,
+      "loss": 0.9442,
       "step": 3380
     },
     {
-      "epoch": 0.09285655433797478,
+      "epoch": 0.09594211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.9793609361212505e-05,
-      "loss": 1.1433,
+      "learning_rate": 1.9772949272624943e-05,
+      "loss": 0.8875,
       "step": 3381
     },
     {
-      "epoch": 0.09288401856581802,
+      "epoch": 0.09597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.9793429534233724e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.97727544938546e-05,
+      "loss": 0.9076,
       "step": 3382
     },
     {
-      "epoch": 0.09291148279366125,
+      "epoch": 0.0959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9793249629765533e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.9772559632533362e-05,
+      "loss": 1.0511,
       "step": 3383
     },
     {
-      "epoch": 0.09293894702150449,
+      "epoch": 0.0960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.979306964780936e-05,
-      "loss": 1.1976,
+      "learning_rate": 1.9772364688662874e-05,
+      "loss": 0.9625,
       "step": 3384
     },
     {
-      "epoch": 0.09296641124934772,
+      "epoch": 0.09605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.9792889588366624e-05,
-      "loss": 1.1393,
+      "learning_rate": 1.9772169662244777e-05,
+      "loss": 1.1696,
       "step": 3385
     },
     {
-      "epoch": 0.09299387547719096,
+      "epoch": 0.09608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9792709451438755e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9771974553280727e-05,
+      "loss": 1.1187,
       "step": 3386
     },
     {
-      "epoch": 0.09302133970503419,
+      "epoch": 0.09611237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 1.979252923702717e-05,
-      "loss": 1.1343,
+      "learning_rate": 1.9771779361772368e-05,
+      "loss": 0.9477,
       "step": 3387
     },
     {
-      "epoch": 0.09304880393287743,
+      "epoch": 0.09614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.9792348945133303e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.9771584087721353e-05,
+      "loss": 1.0238,
       "step": 3388
     },
     {
-      "epoch": 0.09307626816072066,
+      "epoch": 0.09616912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.979216857575858e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.9771388731129325e-05,
+      "loss": 1.0061,
       "step": 3389
     },
     {
-      "epoch": 0.0931037323885639,
+      "epoch": 0.09619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791988128904425e-05,
-      "loss": 1.042,
+      "learning_rate": 1.9771193291997938e-05,
+      "loss": 1.0017,
       "step": 3390
     },
     {
-      "epoch": 0.09313119661640713,
+      "epoch": 0.09622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791807604572265e-05,
-      "loss": 1.1206,
+      "learning_rate": 1.977099777032884e-05,
+      "loss": 1.1331,
       "step": 3391
     },
     {
-      "epoch": 0.09315866084425037,
+      "epoch": 0.09625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791627002763532e-05,
-      "loss": 1.0603,
+      "learning_rate": 1.977080216612369e-05,
+      "loss": 0.9851,
       "step": 3392
     },
     {
-      "epoch": 0.0931861250720936,
+      "epoch": 0.09628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791446323479654e-05,
-      "loss": 1.1697,
+      "learning_rate": 1.977060647938413e-05,
+      "loss": 1.1373,
       "step": 3393
     },
     {
-      "epoch": 0.09321358929993684,
+      "epoch": 0.09631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791265566722056e-05,
-      "loss": 1.1342,
+      "learning_rate": 1.9770410710111817e-05,
+      "loss": 1.0918,
       "step": 3394
     },
     {
-      "epoch": 0.09324105352778006,
+      "epoch": 0.0963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.9791084732492175e-05,
-      "loss": 0.971,
+      "learning_rate": 1.9770214858308407e-05,
+      "loss": 1.0576,
       "step": 3395
     },
     {
-      "epoch": 0.0932685177556233,
+      "epoch": 0.0963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.979090382079144e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9770018923975555e-05,
+      "loss": 1.137,
       "step": 3396
     },
     {
-      "epoch": 0.09329598198346653,
+      "epoch": 0.09639614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 1.979072283162128e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.976982290711491e-05,
+      "loss": 1.11,
       "step": 3397
     },
     {
-      "epoch": 0.09332344621130977,
+      "epoch": 0.09642451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 1.9790541764983127e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.9769626807728134e-05,
+      "loss": 1.0186,
       "step": 3398
     },
     {
-      "epoch": 0.093350910439153,
+      "epoch": 0.09645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9790360620878417e-05,
-      "loss": 1.1346,
+      "learning_rate": 1.9769430625816882e-05,
+      "loss": 1.1018,
       "step": 3399
     },
     {
-      "epoch": 0.09337837466699624,
+      "epoch": 0.09648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.979017939930858e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.976923436138281e-05,
+      "loss": 1.0255,
       "step": 3400
     },
     {
-      "epoch": 0.09340583889483947,
+      "epoch": 0.09650964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.9789998100275053e-05,
-      "loss": 1.1,
+      "learning_rate": 1.9769038014427574e-05,
+      "loss": 1.067,
       "step": 3401
     },
     {
-      "epoch": 0.0934333031226827,
+      "epoch": 0.09653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.978981672377927e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.9768841584952837e-05,
+      "loss": 0.9745,
       "step": 3402
     },
     {
-      "epoch": 0.09346076735052594,
+      "epoch": 0.09656640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.978963526982266e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.9768645072960256e-05,
+      "loss": 1.1588,
       "step": 3403
     },
     {
-      "epoch": 0.09348823157836918,
+      "epoch": 0.09659477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.9789453738406666e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.9768448478451493e-05,
+      "loss": 1.1032,
       "step": 3404
     },
     {
-      "epoch": 0.09351569580621241,
+      "epoch": 0.09662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.978927212953272e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.9768251801428203e-05,
+      "loss": 1.006,
       "step": 3405
     },
     {
-      "epoch": 0.09354316003405565,
+      "epoch": 0.09665153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.9789090443202267e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.976805504189205e-05,
+      "loss": 1.0493,
       "step": 3406
     },
     {
-      "epoch": 0.09357062426189888,
+      "epoch": 0.09667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.9788908679416733e-05,
-      "loss": 1.0876,
+      "learning_rate": 1.9767858199844697e-05,
+      "loss": 1.0494,
       "step": 3407
     },
     {
-      "epoch": 0.09359808848974212,
+      "epoch": 0.09670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.9788726838177564e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.976766127528781e-05,
+      "loss": 0.9875,
       "step": 3408
     },
     {
-      "epoch": 0.09362555271758534,
+      "epoch": 0.0967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9788544919486193e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.9767464268223047e-05,
+      "loss": 1.0834,
       "step": 3409
     },
     {
-      "epoch": 0.09365301694542857,
+      "epoch": 0.0967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.9788362923344068e-05,
-      "loss": 1.1495,
+      "learning_rate": 1.9767267178652075e-05,
+      "loss": 1.0919,
       "step": 3410
     },
     {
-      "epoch": 0.09368048117327181,
+      "epoch": 0.09679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.978818084975262e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.9767070006576557e-05,
+      "loss": 1.1204,
       "step": 3411
     },
     {
-      "epoch": 0.09370794540111504,
+      "epoch": 0.09682179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 1.9787998698713296e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.9766872751998163e-05,
+      "loss": 1.0404,
       "step": 3412
     },
     {
-      "epoch": 0.09373540962895828,
+      "epoch": 0.09685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9787816470227533e-05,
-      "loss": 1.0598,
+      "learning_rate": 1.9766675414918554e-05,
+      "loss": 1.0547,
       "step": 3413
     },
     {
-      "epoch": 0.09376287385680152,
+      "epoch": 0.09687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9787634164296775e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.9766477995339395e-05,
+      "loss": 0.905,
       "step": 3414
     },
     {
-      "epoch": 0.09379033808464475,
+      "epoch": 0.09690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9787451780922465e-05,
-      "loss": 1.1175,
+      "learning_rate": 1.9766280493262362e-05,
+      "loss": 0.9047,
       "step": 3415
     },
     {
-      "epoch": 0.09381780231248799,
+      "epoch": 0.09693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.9787269320106048e-05,
-      "loss": 1.0326,
+      "learning_rate": 1.9766082908689117e-05,
+      "loss": 1.0473,
       "step": 3416
     },
     {
-      "epoch": 0.09384526654033122,
+      "epoch": 0.09696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.978708678184896e-05,
-      "loss": 1.0943,
+      "learning_rate": 1.976588524162133e-05,
+      "loss": 0.9252,
       "step": 3417
     },
     {
-      "epoch": 0.09387273076817446,
+      "epoch": 0.09699205448354142,
       "grad_norm": 0.0,
-      "learning_rate": 1.978690416615265e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.9765687492060674e-05,
+      "loss": 0.9709,
       "step": 3418
     },
     {
-      "epoch": 0.09390019499601769,
+      "epoch": 0.09702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.9786721473018565e-05,
-      "loss": 1.1119,
+      "learning_rate": 1.9765489660008813e-05,
+      "loss": 1.0221,
       "step": 3419
     },
     {
-      "epoch": 0.09392765922386093,
+      "epoch": 0.09704880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.9786538702448152e-05,
-      "loss": 1.1682,
+      "learning_rate": 1.9765291745467423e-05,
+      "loss": 1.0105,
       "step": 3420
     },
     {
-      "epoch": 0.09395512345170416,
+      "epoch": 0.09707718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.978635585444285e-05,
-      "loss": 1.1348,
+      "learning_rate": 1.9765093748438176e-05,
+      "loss": 1.0939,
       "step": 3421
     },
     {
-      "epoch": 0.09398258767954738,
+      "epoch": 0.097105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.978617292900411e-05,
-      "loss": 1.1401,
+      "learning_rate": 1.9764895668922743e-05,
+      "loss": 0.9964,
       "step": 3422
     },
     {
-      "epoch": 0.09401005190739062,
+      "epoch": 0.09713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785989926133376e-05,
-      "loss": 1.0832,
+      "learning_rate": 1.9764697506922797e-05,
+      "loss": 1.0641,
       "step": 3423
     },
     {
-      "epoch": 0.09403751613523385,
+      "epoch": 0.09716231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785806845832105e-05,
-      "loss": 1.1465,
+      "learning_rate": 1.9764499262440014e-05,
+      "loss": 1.1073,
       "step": 3424
     },
     {
-      "epoch": 0.09406498036307709,
+      "epoch": 0.09719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785623688101735e-05,
-      "loss": 1.127,
+      "learning_rate": 1.9764300935476065e-05,
+      "loss": 1.0741,
       "step": 3425
     },
     {
-      "epoch": 0.09409244459092032,
+      "epoch": 0.09721906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785440452943724e-05,
-      "loss": 1.1149,
+      "learning_rate": 1.9764102526032625e-05,
+      "loss": 0.8891,
       "step": 3426
     },
     {
-      "epoch": 0.09411990881876356,
+      "epoch": 0.09724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785257140359513e-05,
-      "loss": 0.9699,
+      "learning_rate": 1.9763904034111375e-05,
+      "loss": 1.0419,
       "step": 3427
     },
     {
-      "epoch": 0.0941473730466068,
+      "epoch": 0.09727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.978507375035056e-05,
-      "loss": 1.05,
+      "learning_rate": 1.9763705459713985e-05,
+      "loss": 1.0368,
       "step": 3428
     },
     {
-      "epoch": 0.09417483727445003,
+      "epoch": 0.09730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.9784890282918315e-05,
-      "loss": 1.062,
+      "learning_rate": 1.9763506802842142e-05,
+      "loss": 0.9238,
       "step": 3429
     },
     {
-      "epoch": 0.09420230150229326,
+      "epoch": 0.09733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.9784706738064222e-05,
-      "loss": 1.1112,
+      "learning_rate": 1.976330806349751e-05,
+      "loss": 0.9627,
       "step": 3430
     },
     {
-      "epoch": 0.0942297657301365,
+      "epoch": 0.09736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.9784523115789742e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.9763109241681785e-05,
+      "loss": 1.0133,
       "step": 3431
     },
     {
-      "epoch": 0.09425722995797974,
+      "epoch": 0.09738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9784339416096326e-05,
-      "loss": 1.1103,
+      "learning_rate": 1.976291033739663e-05,
+      "loss": 1.0398,
       "step": 3432
     },
     {
-      "epoch": 0.09428469418582297,
+      "epoch": 0.09741770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 1.9784155638985426e-05,
-      "loss": 1.07,
+      "learning_rate": 1.976271135064374e-05,
+      "loss": 1.067,
       "step": 3433
     },
     {
-      "epoch": 0.0943121584136662,
+      "epoch": 0.0974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.9783971784458496e-05,
-      "loss": 1.1101,
+      "learning_rate": 1.9762512281424778e-05,
+      "loss": 1.0749,
       "step": 3434
     },
     {
-      "epoch": 0.09433962264150944,
+      "epoch": 0.09747446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 1.978378785251699e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.9762313129741444e-05,
+      "loss": 1.0556,
       "step": 3435
     },
     {
-      "epoch": 0.09436708686935266,
+      "epoch": 0.09750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.9783603843162365e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.976211389559541e-05,
+      "loss": 0.9664,
       "step": 3436
     },
     {
-      "epoch": 0.0943945510971959,
+      "epoch": 0.09753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.978341975639608e-05,
-      "loss": 1.1,
+      "learning_rate": 1.9761914578988358e-05,
+      "loss": 0.9712,
       "step": 3437
     },
     {
-      "epoch": 0.09442201532503913,
+      "epoch": 0.09755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.978323559221959e-05,
-      "loss": 1.0771,
+      "learning_rate": 1.976171517992198e-05,
+      "loss": 1.053,
       "step": 3438
     },
     {
-      "epoch": 0.09444947955288237,
+      "epoch": 0.09758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.978305135063434e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.9761515698397952e-05,
+      "loss": 1.0278,
       "step": 3439
     },
     {
-      "epoch": 0.0944769437807256,
+      "epoch": 0.09761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.9782867031641806e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.9761316134417963e-05,
+      "loss": 1.0391,
       "step": 3440
     },
     {
-      "epoch": 0.09450440800856884,
+      "epoch": 0.09764472190692396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9782682635243438e-05,
-      "loss": 1.03,
+      "learning_rate": 1.9761116487983697e-05,
+      "loss": 1.0753,
       "step": 3441
     },
     {
-      "epoch": 0.09453187223641207,
+      "epoch": 0.09767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.9782498161440692e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.9760916759096846e-05,
+      "loss": 1.1611,
       "step": 3442
     },
     {
-      "epoch": 0.09455933646425531,
+      "epoch": 0.09770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9782313610235035e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.9760716947759086e-05,
+      "loss": 0.9018,
       "step": 3443
     },
     {
-      "epoch": 0.09458680069209854,
+      "epoch": 0.09772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.978212898162792e-05,
-      "loss": 1.1037,
+      "learning_rate": 1.9760517053972114e-05,
+      "loss": 1.0439,
       "step": 3444
     },
     {
-      "epoch": 0.09461426491994178,
+      "epoch": 0.09775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.9781944275620816e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.9760317077737615e-05,
+      "loss": 1.0625,
       "step": 3445
     },
     {
-      "epoch": 0.09464172914778501,
+      "epoch": 0.0977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9781759492215174e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.976011701905728e-05,
+      "loss": 0.9885,
       "step": 3446
     },
     {
-      "epoch": 0.09466919337562825,
+      "epoch": 0.09781498297389331,
       "grad_norm": 0.0,
-      "learning_rate": 1.9781574631412467e-05,
-      "loss": 1.015,
+      "learning_rate": 1.9759916877932795e-05,
+      "loss": 1.0078,
       "step": 3447
     },
     {
-      "epoch": 0.09469665760347148,
+      "epoch": 0.0978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.978138969321415e-05,
-      "loss": 1.1207,
+      "learning_rate": 1.9759716654365854e-05,
+      "loss": 1.0072,
       "step": 3448
     },
     {
-      "epoch": 0.0947241218313147,
+      "epoch": 0.09787173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 1.978120467762169e-05,
-      "loss": 1.0786,
+      "learning_rate": 1.975951634835815e-05,
+      "loss": 1.1088,
       "step": 3449
     },
     {
-      "epoch": 0.09475158605915794,
+      "epoch": 0.09790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9781019584636548e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.975931595991137e-05,
+      "loss": 1.1278,
       "step": 3450
     },
     {
-      "epoch": 0.09477905028700118,
+      "epoch": 0.09792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.978083441426019e-05,
-      "loss": 1.1702,
+      "learning_rate": 1.975911548902721e-05,
+      "loss": 1.1053,
       "step": 3451
     },
     {
-      "epoch": 0.09480651451484441,
+      "epoch": 0.09795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.9780649166494083e-05,
-      "loss": 1.1559,
+      "learning_rate": 1.9758914935707366e-05,
+      "loss": 1.0845,
       "step": 3452
     },
     {
-      "epoch": 0.09483397874268765,
+      "epoch": 0.09798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.978046384133969e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.9758714299953526e-05,
+      "loss": 1.0907,
       "step": 3453
     },
     {
-      "epoch": 0.09486144297053088,
+      "epoch": 0.09801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.978027843879848e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.9758513581767388e-05,
+      "loss": 1.0812,
       "step": 3454
     },
     {
-      "epoch": 0.09488890719837412,
+      "epoch": 0.09804199772985243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9780092958871917e-05,
-      "loss": 1.1513,
+      "learning_rate": 1.975831278115065e-05,
+      "loss": 0.8931,
       "step": 3455
     },
     {
-      "epoch": 0.09491637142621735,
+      "epoch": 0.09807037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 1.977990740156147e-05,
-      "loss": 0.979,
+      "learning_rate": 1.9758111898105002e-05,
+      "loss": 0.9612,
       "step": 3456
     },
     {
-      "epoch": 0.09494383565406059,
+      "epoch": 0.09809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.9779721766868608e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.9757910932632142e-05,
+      "loss": 1.1064,
       "step": 3457
     },
     {
-      "epoch": 0.09497129988190382,
+      "epoch": 0.09812712826333711,
       "grad_norm": 0.0,
-      "learning_rate": 1.97795360547948e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.9757709884733774e-05,
+      "loss": 0.992,
       "step": 3458
     },
     {
-      "epoch": 0.09499876410974706,
+      "epoch": 0.09815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.9779350265341513e-05,
-      "loss": 1.1188,
+      "learning_rate": 1.975750875441159e-05,
+      "loss": 1.1125,
       "step": 3459
     },
     {
-      "epoch": 0.0950262283375903,
+      "epoch": 0.0981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9779164398510218e-05,
-      "loss": 1.0754,
+      "learning_rate": 1.975730754166729e-05,
+      "loss": 1.1681,
       "step": 3460
     },
     {
-      "epoch": 0.09505369256543353,
+      "epoch": 0.0982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.977897845430239e-05,
-      "loss": 1.0701,
+      "learning_rate": 1.9757106246502577e-05,
+      "loss": 0.9614,
       "step": 3461
     },
     {
-      "epoch": 0.09508115679327676,
+      "epoch": 0.09824063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.9778792432719493e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.9756904868919145e-05,
+      "loss": 1.0956,
       "step": 3462
     },
     {
-      "epoch": 0.09510862102111999,
+      "epoch": 0.09826901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 1.9778606333763003e-05,
-      "loss": 1.0832,
+      "learning_rate": 1.9756703408918706e-05,
+      "loss": 1.1205,
       "step": 3463
     },
     {
-      "epoch": 0.09513608524896322,
+      "epoch": 0.09829738933030648,
       "grad_norm": 0.0,
-      "learning_rate": 1.977842015743439e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.9756501866502947e-05,
+      "loss": 1.1531,
       "step": 3464
     },
     {
-      "epoch": 0.09516354947680646,
+      "epoch": 0.09832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9778233903735132e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.9756300241673583e-05,
+      "loss": 1.0279,
       "step": 3465
     },
     {
-      "epoch": 0.09519101370464969,
+      "epoch": 0.09835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.9778047572666695e-05,
-      "loss": 1.0815,
+      "learning_rate": 1.9756098534432314e-05,
+      "loss": 0.9157,
       "step": 3466
     },
     {
-      "epoch": 0.09521847793249293,
+      "epoch": 0.09838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.9777861164230563e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.9755896744780835e-05,
+      "loss": 1.0191,
       "step": 3467
     },
     {
-      "epoch": 0.09524594216033616,
+      "epoch": 0.09841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.9777674678428204e-05,
-      "loss": 1.0893,
+      "learning_rate": 1.9755694872720863e-05,
+      "loss": 1.0406,
       "step": 3468
     },
     {
-      "epoch": 0.0952734063881794,
+      "epoch": 0.09843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9777488115261098e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.97554929182541e-05,
+      "loss": 1.0674,
       "step": 3469
     },
     {
-      "epoch": 0.09530087061602263,
+      "epoch": 0.09846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 1.9777301474730715e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.9755290881382244e-05,
+      "loss": 1.0881,
       "step": 3470
     },
     {
-      "epoch": 0.09532833484386587,
+      "epoch": 0.09849602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.9777114756838537e-05,
-      "loss": 1.1869,
+      "learning_rate": 1.975508876210701e-05,
+      "loss": 1.0353,
       "step": 3471
     },
     {
-      "epoch": 0.0953557990717091,
+      "epoch": 0.0985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9776927961586036e-05,
-      "loss": 1.0937,
+      "learning_rate": 1.97548865604301e-05,
+      "loss": 1.111,
       "step": 3472
     },
     {
-      "epoch": 0.09538326329955234,
+      "epoch": 0.0985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9776741088974698e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.9754684276353226e-05,
+      "loss": 1.0345,
       "step": 3473
     },
     {
-      "epoch": 0.09541072752739557,
+      "epoch": 0.09858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.9776554139005998e-05,
-      "loss": 1.1164,
+      "learning_rate": 1.97544819098781e-05,
+      "loss": 1.0751,
       "step": 3474
     },
     {
-      "epoch": 0.09543819175523881,
+      "epoch": 0.09860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.977636711168141e-05,
-      "loss": 1.143,
+      "learning_rate": 1.9754279461006424e-05,
+      "loss": 1.0522,
       "step": 3475
     },
     {
-      "epoch": 0.09546565598308203,
+      "epoch": 0.09863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.9776180007002422e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.9754076929739906e-05,
+      "loss": 1.1185,
       "step": 3476
     },
     {
-      "epoch": 0.09549312021092526,
+      "epoch": 0.09866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.977599282497051e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.9753874316080268e-05,
+      "loss": 1.031,
       "step": 3477
     },
     {
-      "epoch": 0.0955205844387685,
+      "epoch": 0.09869466515323495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9775805565587157e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.9753671620029213e-05,
+      "loss": 0.9507,
       "step": 3478
     },
     {
-      "epoch": 0.09554804866661173,
+      "epoch": 0.09872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9775618228853842e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.9753468841588456e-05,
+      "loss": 1.0375,
       "step": 3479
     },
     {
-      "epoch": 0.09557551289445497,
+      "epoch": 0.09875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.977543081477205e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.9753265980759707e-05,
+      "loss": 1.019,
       "step": 3480
     },
     {
-      "epoch": 0.0956029771222982,
+      "epoch": 0.09877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.977524332334326e-05,
-      "loss": 1.0116,
+      "learning_rate": 1.9753063037544686e-05,
+      "loss": 1.1734,
       "step": 3481
     },
     {
-      "epoch": 0.09563044135014144,
+      "epoch": 0.09880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9775055754568964e-05,
-      "loss": 1.073,
+      "learning_rate": 1.9752860011945098e-05,
+      "loss": 1.1365,
       "step": 3482
     },
     {
-      "epoch": 0.09565790557798468,
+      "epoch": 0.09883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9774868108450634e-05,
-      "loss": 1.1019,
+      "learning_rate": 1.9752656903962666e-05,
+      "loss": 1.034,
       "step": 3483
     },
     {
-      "epoch": 0.09568536980582791,
+      "epoch": 0.09886492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 1.9774680384989767e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.97524537135991e-05,
+      "loss": 0.9448,
       "step": 3484
     },
     {
-      "epoch": 0.09571283403367115,
+      "epoch": 0.0988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.977449258418784e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.9752250440856123e-05,
+      "loss": 1.0301,
       "step": 3485
     },
     {
-      "epoch": 0.09574029826151438,
+      "epoch": 0.0989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9774304706046338e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.9752047085735447e-05,
+      "loss": 1.0322,
       "step": 3486
     },
     {
-      "epoch": 0.09576776248935762,
+      "epoch": 0.098950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9774116750566755e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.975184364823879e-05,
+      "loss": 1.0555,
       "step": 3487
     },
     {
-      "epoch": 0.09579522671720085,
+      "epoch": 0.09897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9773928717750573e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.9751640128367872e-05,
+      "loss": 1.071,
       "step": 3488
     },
     {
-      "epoch": 0.09582269094504409,
+      "epoch": 0.09900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.977374060759928e-05,
-      "loss": 1.1026,
+      "learning_rate": 1.9751436526124412e-05,
+      "loss": 1.1312,
       "step": 3489
     },
     {
-      "epoch": 0.09585015517288731,
+      "epoch": 0.09903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.9773552420114364e-05,
-      "loss": 1.1807,
+      "learning_rate": 1.975123284151013e-05,
+      "loss": 1.0601,
       "step": 3490
     },
     {
-      "epoch": 0.09587761940073054,
+      "epoch": 0.09906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.977336415529732e-05,
-      "loss": 1.0843,
+      "learning_rate": 1.9751029074526743e-05,
+      "loss": 1.0803,
       "step": 3491
     },
     {
-      "epoch": 0.09590508362857378,
+      "epoch": 0.09909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.977317581314963e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.9750825225175976e-05,
+      "loss": 1.0916,
       "step": 3492
     },
     {
-      "epoch": 0.09593254785641701,
+      "epoch": 0.09912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.977298739367279e-05,
-      "loss": 1.1597,
+      "learning_rate": 1.975062129345955e-05,
+      "loss": 1.1354,
       "step": 3493
     },
     {
-      "epoch": 0.09596001208426025,
+      "epoch": 0.09914869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 1.9772798896868283e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.975041727937919e-05,
+      "loss": 0.7789,
       "step": 3494
     },
     {
-      "epoch": 0.09598747631210348,
+      "epoch": 0.09917707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 1.977261032273761e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.9750213182936614e-05,
+      "loss": 0.9893,
       "step": 3495
     },
     {
-      "epoch": 0.09601494053994672,
+      "epoch": 0.09920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.9772421671282255e-05,
-      "loss": 1.0358,
+      "learning_rate": 1.975000900413355e-05,
+      "loss": 0.9857,
       "step": 3496
     },
     {
-      "epoch": 0.09604240476778995,
+      "epoch": 0.09923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.9772232942503716e-05,
-      "loss": 0.9941,
+      "learning_rate": 1.974980474297172e-05,
+      "loss": 1.0864,
       "step": 3497
     },
     {
-      "epoch": 0.09606986899563319,
+      "epoch": 0.0992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.977204413640349e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.9749600399452852e-05,
+      "loss": 0.9478,
       "step": 3498
     },
     {
-      "epoch": 0.09609733322347642,
+      "epoch": 0.0992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9771855252983056e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.9749395973578674e-05,
+      "loss": 1.1232,
       "step": 3499
     },
     {
-      "epoch": 0.09612479745131966,
+      "epoch": 0.09931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.9771666292243924e-05,
-      "loss": 1.1071,
+      "learning_rate": 1.974919146535091e-05,
+      "loss": 1.0239,
       "step": 3500
     },
     {
-      "epoch": 0.0961522616791629,
+      "epoch": 0.09934733257661749,
       "grad_norm": 0.0,
-      "learning_rate": 1.977147725418758e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.9748986874771283e-05,
+      "loss": 1.0718,
       "step": 3501
     },
     {
-      "epoch": 0.09617972590700613,
+      "epoch": 0.09937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.9771288138815526e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.9748782201841528e-05,
+      "loss": 1.0713,
       "step": 3502
     },
     {
-      "epoch": 0.09620719013484935,
+      "epoch": 0.09940408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.9771098946129252e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.974857744656337e-05,
+      "loss": 0.9443,
       "step": 3503
     },
     {
-      "epoch": 0.09623465436269259,
+      "epoch": 0.09943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.977090967613026e-05,
-      "loss": 0.931,
+      "learning_rate": 1.974837260893854e-05,
+      "loss": 1.0094,
       "step": 3504
     },
     {
-      "epoch": 0.09626211859053582,
+      "epoch": 0.09946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.9770720328820042e-05,
-      "loss": 1.1152,
+      "learning_rate": 1.974816768896877e-05,
+      "loss": 1.0887,
       "step": 3505
     },
     {
-      "epoch": 0.09628958281837906,
+      "epoch": 0.09948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9770530904200104e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.974796268665579e-05,
+      "loss": 1.0668,
       "step": 3506
     },
     {
-      "epoch": 0.09631704704622229,
+      "epoch": 0.09951759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 1.9770341402271944e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.974775760200133e-05,
+      "loss": 1.004,
       "step": 3507
     },
     {
-      "epoch": 0.09634451127406553,
+      "epoch": 0.09954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.977015182303705e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.9747552435007125e-05,
+      "loss": 0.9909,
       "step": 3508
     },
     {
-      "epoch": 0.09637197550190876,
+      "epoch": 0.09957434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 1.9769962166496938e-05,
-      "loss": 1.0326,
+      "learning_rate": 1.9747347185674902e-05,
+      "loss": 0.975,
       "step": 3509
     },
     {
-      "epoch": 0.096399439729752,
+      "epoch": 0.09960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9769772432653095e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.9747141854006405e-05,
+      "loss": 1.0869,
       "step": 3510
     },
     {
-      "epoch": 0.09642690395759523,
+      "epoch": 0.0996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.976958262150703e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.9746936440003357e-05,
+      "loss": 0.972,
       "step": 3511
     },
     {
-      "epoch": 0.09645436818543847,
+      "epoch": 0.09965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.9769392733060245e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.97467309436675e-05,
+      "loss": 1.1942,
       "step": 3512
     },
     {
-      "epoch": 0.0964818324132817,
+      "epoch": 0.09968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.9769202767314234e-05,
-      "loss": 1.1101,
+      "learning_rate": 1.9746525365000574e-05,
+      "loss": 1.0796,
       "step": 3513
     },
     {
-      "epoch": 0.09650929664112494,
+      "epoch": 0.09971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.9769012724270513e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.9746319704004305e-05,
+      "loss": 1.0573,
       "step": 3514
     },
     {
-      "epoch": 0.09653676086896817,
+      "epoch": 0.09974460839954596,
       "grad_norm": 0.0,
-      "learning_rate": 1.9768822603930578e-05,
-      "loss": 1.1548,
+      "learning_rate": 1.9746113960680435e-05,
+      "loss": 1.1512,
       "step": 3515
     },
     {
-      "epoch": 0.09656422509681141,
+      "epoch": 0.09977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.976863240629593e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.9745908135030705e-05,
+      "loss": 1.0711,
       "step": 3516
     },
     {
-      "epoch": 0.09659168932465463,
+      "epoch": 0.09980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9768442131368085e-05,
-      "loss": 1.1659,
+      "learning_rate": 1.974570222705685e-05,
+      "loss": 1.0122,
       "step": 3517
     },
     {
-      "epoch": 0.09661915355249787,
+      "epoch": 0.09982973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.976825177914854e-05,
-      "loss": 1.1169,
+      "learning_rate": 1.9745496236760608e-05,
+      "loss": 1.0543,
       "step": 3518
     },
     {
-      "epoch": 0.0966466177803411,
+      "epoch": 0.09985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.97680613496388e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.974529016414372e-05,
+      "loss": 1.1036,
       "step": 3519
     },
     {
-      "epoch": 0.09667408200818434,
+      "epoch": 0.09988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.9767870842840377e-05,
-      "loss": 1.1332,
+      "learning_rate": 1.9745084009207934e-05,
+      "loss": 1.0399,
       "step": 3520
     },
     {
-      "epoch": 0.09670154623602757,
+      "epoch": 0.09991486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 1.9767680258754778e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.974487777195498e-05,
+      "loss": 1.0736,
       "step": 3521
     },
     {
-      "epoch": 0.09672901046387081,
+      "epoch": 0.09994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 1.9767489597383505e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.974467145238661e-05,
+      "loss": 1.0114,
       "step": 3522
     },
     {
-      "epoch": 0.09675647469171404,
+      "epoch": 0.0999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.9767298858728074e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.9744465050504558e-05,
+      "loss": 1.247,
       "step": 3523
     },
     {
-      "epoch": 0.09678393891955728,
+      "epoch": 0.1,
       "grad_norm": 0.0,
-      "learning_rate": 1.9767108042789988e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.9744258566310574e-05,
+      "loss": 0.9433,
       "step": 3524
     },
     {
-      "epoch": 0.09681140314740051,
+      "epoch": 0.10002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.976691714957076e-05,
-      "loss": 1.015,
+      "learning_rate": 1.9744051999806397e-05,
+      "loss": 0.9902,
       "step": 3525
     },
     {
-      "epoch": 0.09683886737524375,
+      "epoch": 0.10005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.9766726179071906e-05,
-      "loss": 1.1306,
+      "learning_rate": 1.974384535099378e-05,
+      "loss": 1.0911,
       "step": 3526
     },
     {
-      "epoch": 0.09686633160308698,
+      "epoch": 0.10008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.9766535131294925e-05,
-      "loss": 1.0398,
+      "learning_rate": 1.9743638619874458e-05,
+      "loss": 1.06,
       "step": 3527
     },
     {
-      "epoch": 0.09689379583093022,
+      "epoch": 0.10011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.976634400624134e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.9743431806450185e-05,
+      "loss": 1.0044,
       "step": 3528
     },
     {
-      "epoch": 0.09692126005877345,
+      "epoch": 0.10014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9766152803912654e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.9743224910722706e-05,
+      "loss": 1.1276,
       "step": 3529
     },
     {
-      "epoch": 0.09694872428661667,
+      "epoch": 0.10017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.9765961524310386e-05,
-      "loss": 1.1737,
+      "learning_rate": 1.9743017932693766e-05,
+      "loss": 1.101,
       "step": 3530
     },
     {
-      "epoch": 0.09697618851445991,
+      "epoch": 0.10019863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.9765770167436043e-05,
-      "loss": 1.1099,
+      "learning_rate": 1.974281087236512e-05,
+      "loss": 1.0162,
       "step": 3531
     },
     {
-      "epoch": 0.09700365274230315,
+      "epoch": 0.10022701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.9765578733291146e-05,
-      "loss": 1.1089,
+      "learning_rate": 1.9742603729738506e-05,
+      "loss": 1.086,
       "step": 3532
     },
     {
-      "epoch": 0.09703111697014638,
+      "epoch": 0.10025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.976538722187721e-05,
-      "loss": 1.076,
+      "learning_rate": 1.9742396504815687e-05,
+      "loss": 0.9487,
       "step": 3533
     },
     {
-      "epoch": 0.09705858119798962,
+      "epoch": 0.10028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.9765195633195744e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.97421891975984e-05,
+      "loss": 1.1104,
       "step": 3534
     },
     {
-      "epoch": 0.09708604542583285,
+      "epoch": 0.10031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.976500396724827e-05,
-      "loss": 1.0853,
+      "learning_rate": 1.974198180808841e-05,
+      "loss": 1.0964,
       "step": 3535
     },
     {
-      "epoch": 0.09711350965367609,
+      "epoch": 0.10034052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.97648122240363e-05,
-      "loss": 1.148,
+      "learning_rate": 1.9741774336287456e-05,
+      "loss": 1.0555,
       "step": 3536
     },
     {
-      "epoch": 0.09714097388151932,
+      "epoch": 0.1003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9764620403561355e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.97415667821973e-05,
+      "loss": 1.0655,
       "step": 3537
     },
     {
-      "epoch": 0.09716843810936256,
+      "epoch": 0.1003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9764428505824952e-05,
-      "loss": 1.021,
+      "learning_rate": 1.9741359145819688e-05,
+      "loss": 0.9799,
       "step": 3538
     },
     {
-      "epoch": 0.09719590233720579,
+      "epoch": 0.10042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.9764236530828604e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.974115142715638e-05,
+      "loss": 1.1064,
       "step": 3539
     },
     {
-      "epoch": 0.09722336656504903,
+      "epoch": 0.10045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.976404447857384e-05,
-      "loss": 1.1755,
+      "learning_rate": 1.9740943626209127e-05,
+      "loss": 1.0018,
       "step": 3540
     },
     {
-      "epoch": 0.09725083079289226,
+      "epoch": 0.10048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.976385234906217e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.9740735742979688e-05,
+      "loss": 1.0405,
       "step": 3541
     },
     {
-      "epoch": 0.0972782950207355,
+      "epoch": 0.10051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9763660142295117e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.974052777746981e-05,
+      "loss": 0.932,
       "step": 3542
     },
     {
-      "epoch": 0.09730575924857873,
+      "epoch": 0.10053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.9763467858274207e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.9740319729681262e-05,
+      "loss": 1.0559,
       "step": 3543
     },
     {
-      "epoch": 0.09733322347642195,
+      "epoch": 0.10056753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.9763275497000952e-05,
-      "loss": 0.993,
+      "learning_rate": 1.9740111599615793e-05,
+      "loss": 0.9965,
       "step": 3544
     },
     {
-      "epoch": 0.09736068770426519,
+      "epoch": 0.10059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9763083058476882e-05,
-      "loss": 1.038,
+      "learning_rate": 1.9739903387275165e-05,
+      "loss": 0.9683,
       "step": 3545
     },
     {
-      "epoch": 0.09738815193210842,
+      "epoch": 0.10062429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 1.976289054270352e-05,
-      "loss": 1.1549,
+      "learning_rate": 1.9739695092661132e-05,
+      "loss": 1.0754,
       "step": 3546
     },
     {
-      "epoch": 0.09741561615995166,
+      "epoch": 0.10065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.976269794968238e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.9739486715775463e-05,
+      "loss": 1.0304,
       "step": 3547
     },
     {
-      "epoch": 0.0974430803877949,
+      "epoch": 0.10068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.9762505279414996e-05,
-      "loss": 1.13,
+      "learning_rate": 1.9739278256619906e-05,
+      "loss": 1.0106,
       "step": 3548
     },
     {
-      "epoch": 0.09747054461563813,
+      "epoch": 0.1007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.9762312531902886e-05,
-      "loss": 1.0715,
+      "learning_rate": 1.9739069715196227e-05,
+      "loss": 1.1338,
       "step": 3549
     },
     {
-      "epoch": 0.09749800884348137,
+      "epoch": 0.1007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.976211970714758e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.973886109150619e-05,
+      "loss": 0.9789,
       "step": 3550
     },
     {
-      "epoch": 0.0975254730713246,
+      "epoch": 0.10076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.9761926805150603e-05,
-      "loss": 1.085,
+      "learning_rate": 1.973865238555156e-05,
+      "loss": 1.0524,
       "step": 3551
     },
     {
-      "epoch": 0.09755293729916784,
+      "epoch": 0.10079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.976173382591348e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.973844359733409e-05,
+      "loss": 1.0058,
       "step": 3552
     },
     {
-      "epoch": 0.09758040152701107,
+      "epoch": 0.10082292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 1.9761540769437736e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.9738234726855553e-05,
+      "loss": 0.9272,
       "step": 3553
     },
     {
-      "epoch": 0.0976078657548543,
+      "epoch": 0.10085130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.97613476357249e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.9738025774117707e-05,
+      "loss": 1.0956,
       "step": 3554
     },
     {
-      "epoch": 0.09763532998269754,
+      "epoch": 0.10087968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.97611544247765e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.973781673912232e-05,
+      "loss": 1.0414,
       "step": 3555
     },
     {
-      "epoch": 0.09766279421054078,
+      "epoch": 0.10090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.976096113659406e-05,
-      "loss": 1.1451,
+      "learning_rate": 1.9737607621871157e-05,
+      "loss": 1.1408,
       "step": 3556
     },
     {
-      "epoch": 0.09769025843838401,
+      "epoch": 0.10093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.9760767771179123e-05,
-      "loss": 1.105,
+      "learning_rate": 1.9737398422365986e-05,
+      "loss": 1.0978,
       "step": 3557
     },
     {
-      "epoch": 0.09771772266622723,
+      "epoch": 0.10096481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.9760574328533205e-05,
-      "loss": 1.1168,
+      "learning_rate": 1.973718914060857e-05,
+      "loss": 1.0461,
       "step": 3558
     },
     {
-      "epoch": 0.09774518689407047,
+      "epoch": 0.10099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.9760380808657845e-05,
-      "loss": 1.1434,
+      "learning_rate": 1.973697977660068e-05,
+      "loss": 0.9981,
       "step": 3559
     },
     {
-      "epoch": 0.0977726511219137,
+      "epoch": 0.10102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.976018721155457e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.9736770330344084e-05,
+      "loss": 0.958,
       "step": 3560
     },
     {
-      "epoch": 0.09780011534975694,
+      "epoch": 0.10104994324631102,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759993537224912e-05,
-      "loss": 1.0949,
+      "learning_rate": 1.9736560801840555e-05,
+      "loss": 1.1153,
       "step": 3561
     },
     {
-      "epoch": 0.09782757957760017,
+      "epoch": 0.1010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759799785670405e-05,
-      "loss": 1.0982,
+      "learning_rate": 1.973635119109186e-05,
+      "loss": 1.0241,
       "step": 3562
     },
     {
-      "epoch": 0.09785504380544341,
+      "epoch": 0.10110669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759605956892585e-05,
-      "loss": 1.1389,
+      "learning_rate": 1.9736141498099765e-05,
+      "loss": 0.9738,
       "step": 3563
     },
     {
-      "epoch": 0.09788250803328664,
+      "epoch": 0.10113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759412050892977e-05,
-      "loss": 1.1405,
+      "learning_rate": 1.9735931722866045e-05,
+      "loss": 1.1551,
       "step": 3564
     },
     {
-      "epoch": 0.09790997226112988,
+      "epoch": 0.10116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759218067673126e-05,
-      "loss": 1.1156,
+      "learning_rate": 1.9735721865392473e-05,
+      "loss": 1.1132,
       "step": 3565
     },
     {
-      "epoch": 0.09793743648897311,
+      "epoch": 0.10119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.9759024007234557e-05,
-      "loss": 1.1799,
+      "learning_rate": 1.973551192568082e-05,
+      "loss": 1.0854,
       "step": 3566
     },
     {
-      "epoch": 0.09796490071681635,
+      "epoch": 0.10122020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.975882986957881e-05,
-      "loss": 1.1184,
+      "learning_rate": 1.9735301903732862e-05,
+      "loss": 1.0569,
       "step": 3567
     },
     {
-      "epoch": 0.09799236494465959,
+      "epoch": 0.10124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.975863565470742e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.973509179955037e-05,
+      "loss": 1.0623,
       "step": 3568
     },
     {
-      "epoch": 0.09801982917250282,
+      "epoch": 0.10127695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 1.975844136262193e-05,
-      "loss": 1.199,
+      "learning_rate": 1.973488161313512e-05,
+      "loss": 1.1047,
       "step": 3569
     },
     {
-      "epoch": 0.09804729340034606,
+      "epoch": 0.10130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9758246993323868e-05,
-      "loss": 1.0816,
+      "learning_rate": 1.9734671344488886e-05,
+      "loss": 1.137,
       "step": 3570
     },
     {
-      "epoch": 0.09807475762818928,
+      "epoch": 0.10133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.9758052546814776e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.973446099361345e-05,
+      "loss": 0.9609,
       "step": 3571
     },
     {
-      "epoch": 0.09810222185603251,
+      "epoch": 0.10136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.975785802309619e-05,
-      "loss": 1.1487,
+      "learning_rate": 1.973425056051058e-05,
+      "loss": 1.1196,
       "step": 3572
     },
     {
-      "epoch": 0.09812968608387575,
+      "epoch": 0.10139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.975766342216966e-05,
-      "loss": 1.139,
+      "learning_rate": 1.973404004518206e-05,
+      "loss": 1.0605,
       "step": 3573
     },
     {
-      "epoch": 0.09815715031171898,
+      "epoch": 0.10141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.9757468744036706e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.9733829447629665e-05,
+      "loss": 1.1011,
       "step": 3574
     },
     {
-      "epoch": 0.09818461453956222,
+      "epoch": 0.1014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9757273988698883e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.9733618767855173e-05,
+      "loss": 1.0525,
       "step": 3575
     },
     {
-      "epoch": 0.09821207876740545,
+      "epoch": 0.1014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9757079156157734e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.9733408005860372e-05,
+      "loss": 1.044,
       "step": 3576
     },
     {
-      "epoch": 0.09823954299524869,
+      "epoch": 0.10150397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 1.975688424641479e-05,
-      "loss": 1.1417,
+      "learning_rate": 1.973319716164703e-05,
+      "loss": 1.0375,
       "step": 3577
     },
     {
-      "epoch": 0.09826700722309192,
+      "epoch": 0.10153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.97566892594716e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.9732986235216937e-05,
+      "loss": 1.0272,
       "step": 3578
     },
     {
-      "epoch": 0.09829447145093516,
+      "epoch": 0.10156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.9756494195329704e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.973277522657187e-05,
+      "loss": 1.1305,
       "step": 3579
     },
     {
-      "epoch": 0.0983219356787784,
+      "epoch": 0.10158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.975629905399065e-05,
-      "loss": 1.075,
+      "learning_rate": 1.9732564135713614e-05,
+      "loss": 1.0924,
       "step": 3580
     },
     {
-      "epoch": 0.09834939990662163,
+      "epoch": 0.10161748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.9756103835455976e-05,
-      "loss": 1.0613,
+      "learning_rate": 1.9732352962643952e-05,
+      "loss": 1.1166,
       "step": 3581
     },
     {
-      "epoch": 0.09837686413446486,
+      "epoch": 0.10164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.975590853972723e-05,
-      "loss": 1.1255,
+      "learning_rate": 1.9732141707364667e-05,
+      "loss": 1.1133,
       "step": 3582
     },
     {
-      "epoch": 0.0984043283623081,
+      "epoch": 0.10167423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.9755713166805958e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.973193036987754e-05,
+      "loss": 1.0737,
       "step": 3583
     },
     {
-      "epoch": 0.09843179259015133,
+      "epoch": 0.10170261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.97555177166937e-05,
-      "loss": 1.1789,
+      "learning_rate": 1.9731718950184368e-05,
+      "loss": 1.0994,
       "step": 3584
     },
     {
-      "epoch": 0.09845925681799456,
+      "epoch": 0.10173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.975532218939201e-05,
-      "loss": 1.0987,
+      "learning_rate": 1.973150744828692e-05,
+      "loss": 1.0409,
       "step": 3585
     },
     {
-      "epoch": 0.09848672104583779,
+      "epoch": 0.10175936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 1.975512658490243e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.9731295864186996e-05,
+      "loss": 0.9987,
       "step": 3586
     },
     {
-      "epoch": 0.09851418527368103,
+      "epoch": 0.10178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.9754930903226514e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.9731084197886376e-05,
+      "loss": 0.9942,
       "step": 3587
     },
     {
-      "epoch": 0.09854164950152426,
+      "epoch": 0.1018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9754735144365803e-05,
-      "loss": 1.1091,
+      "learning_rate": 1.973087244938685e-05,
+      "loss": 1.1995,
       "step": 3588
     },
     {
-      "epoch": 0.0985691137293675,
+      "epoch": 0.10184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.9754539308321848e-05,
-      "loss": 1.1248,
+      "learning_rate": 1.973066061869021e-05,
+      "loss": 1.1969,
       "step": 3589
     },
     {
-      "epoch": 0.09859657795721073,
+      "epoch": 0.10187287173666289,
       "grad_norm": 0.0,
-      "learning_rate": 1.97543433950962e-05,
-      "loss": 1.0714,
+      "learning_rate": 1.973044870579824e-05,
+      "loss": 1.0117,
       "step": 3590
     },
     {
-      "epoch": 0.09862404218505397,
+      "epoch": 0.10190124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.975414740469041e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.9730236710712733e-05,
+      "loss": 1.0494,
       "step": 3591
     },
     {
-      "epoch": 0.0986515064128972,
+      "epoch": 0.10192962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 1.9753951337106025e-05,
-      "loss": 1.001,
+      "learning_rate": 1.9730024633435478e-05,
+      "loss": 1.1045,
       "step": 3592
     },
     {
-      "epoch": 0.09867897064074044,
+      "epoch": 0.10195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.97537551923446e-05,
-      "loss": 1.1239,
+      "learning_rate": 1.972981247396827e-05,
+      "loss": 1.0735,
       "step": 3593
     },
     {
-      "epoch": 0.09870643486858367,
+      "epoch": 0.10198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.9753558970407686e-05,
-      "loss": 1.0687,
+      "learning_rate": 1.9729600232312896e-05,
+      "loss": 1.0153,
       "step": 3594
     },
     {
-      "epoch": 0.09873389909642691,
+      "epoch": 0.10201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.9753362671296835e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.9729387908471153e-05,
+      "loss": 1.0352,
       "step": 3595
     },
     {
-      "epoch": 0.09876136332427014,
+      "epoch": 0.10204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.9753166295013603e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.9729175502444832e-05,
+      "loss": 1.0937,
       "step": 3596
     },
     {
-      "epoch": 0.09878882755211338,
+      "epoch": 0.10207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.9752969841559536e-05,
-      "loss": 1.1724,
+      "learning_rate": 1.972896301423573e-05,
+      "loss": 1.1312,
       "step": 3597
     },
     {
-      "epoch": 0.0988162917799566,
+      "epoch": 0.10209988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 1.9752773310936198e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.972875044384564e-05,
+      "loss": 1.0407,
       "step": 3598
     },
     {
-      "epoch": 0.09884375600779984,
+      "epoch": 0.10212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.975257670314514e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.9728537791276353e-05,
+      "loss": 1.0355,
       "step": 3599
     },
     {
-      "epoch": 0.09887122023564307,
+      "epoch": 0.1021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9752380018187914e-05,
-      "loss": 1.1315,
+      "learning_rate": 1.9728325056529675e-05,
+      "loss": 0.9993,
       "step": 3600
     },
     {
-      "epoch": 0.0988986844634863,
+      "epoch": 0.1021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.975218325606608e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.9728112239607394e-05,
+      "loss": 1.0873,
       "step": 3601
     },
     {
-      "epoch": 0.09892614869132954,
+      "epoch": 0.10221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.9751986416781198e-05,
-      "loss": 1.1548,
+      "learning_rate": 1.9727899340511312e-05,
+      "loss": 1.143,
       "step": 3602
     },
     {
-      "epoch": 0.09895361291917278,
+      "epoch": 0.10224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.975178950033482e-05,
-      "loss": 1.1205,
+      "learning_rate": 1.9727686359243233e-05,
+      "loss": 0.9604,
       "step": 3603
     },
     {
-      "epoch": 0.09898107714701601,
+      "epoch": 0.10227014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 1.9751592506728503e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.9727473295804945e-05,
+      "loss": 1.0585,
       "step": 3604
     },
     {
-      "epoch": 0.09900854137485925,
+      "epoch": 0.10229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.975139543596381e-05,
-      "loss": 1.1406,
+      "learning_rate": 1.9727260150198254e-05,
+      "loss": 0.9715,
       "step": 3605
     },
     {
-      "epoch": 0.09903600560270248,
+      "epoch": 0.10232690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.97511982880423e-05,
-      "loss": 1.0512,
+      "learning_rate": 1.9727046922424956e-05,
+      "loss": 0.9969,
       "step": 3606
     },
     {
-      "epoch": 0.09906346983054572,
+      "epoch": 0.10235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9751001062965533e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.9726833612486864e-05,
+      "loss": 1.1148,
       "step": 3607
     },
     {
-      "epoch": 0.09909093405838895,
+      "epoch": 0.10238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9750803760735068e-05,
-      "loss": 1.0763,
+      "learning_rate": 1.9726620220385762e-05,
+      "loss": 1.1417,
       "step": 3608
     },
     {
-      "epoch": 0.09911839828623219,
+      "epoch": 0.10241203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.9750606381352468e-05,
-      "loss": 1.032,
+      "learning_rate": 1.9726406746123464e-05,
+      "loss": 1.0886,
       "step": 3609
     },
     {
-      "epoch": 0.09914586251407542,
+      "epoch": 0.10244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.9750408924819288e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.9726193189701775e-05,
+      "loss": 1.1069,
       "step": 3610
     },
     {
-      "epoch": 0.09917332674191866,
+      "epoch": 0.10246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.97502113911371e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.972597955112249e-05,
+      "loss": 1.0205,
       "step": 3611
     },
     {
-      "epoch": 0.09920079096976188,
+      "epoch": 0.1024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.975001378030746e-05,
-      "loss": 1.0957,
+      "learning_rate": 1.972576583038742e-05,
+      "loss": 1.0976,
       "step": 3612
     },
     {
-      "epoch": 0.09922825519760511,
+      "epoch": 0.1025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.974981609233194e-05,
-      "loss": 1.1617,
+      "learning_rate": 1.972555202749837e-05,
+      "loss": 0.9735,
       "step": 3613
     },
     {
-      "epoch": 0.09925571942544835,
+      "epoch": 0.1025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.974961832721209e-05,
-      "loss": 1.0202,
+      "learning_rate": 1.9725338142457145e-05,
+      "loss": 1.0559,
       "step": 3614
     },
     {
-      "epoch": 0.09928318365329158,
+      "epoch": 0.10258229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 1.974942048494949e-05,
-      "loss": 1.0929,
+      "learning_rate": 1.972512417526555e-05,
+      "loss": 1.0803,
       "step": 3615
     },
     {
-      "epoch": 0.09931064788113482,
+      "epoch": 0.10261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9749222565545695e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.9724910125925395e-05,
+      "loss": 1.0328,
       "step": 3616
     },
     {
-      "epoch": 0.09933811210897805,
+      "epoch": 0.10263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.9749024569002276e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9724695994438487e-05,
+      "loss": 1.0627,
       "step": 3617
     },
     {
-      "epoch": 0.09936557633682129,
+      "epoch": 0.10266742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 1.9748826495320796e-05,
-      "loss": 1.1501,
+      "learning_rate": 1.9724481780806635e-05,
+      "loss": 1.055,
       "step": 3618
     },
     {
-      "epoch": 0.09939304056466453,
+      "epoch": 0.10269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.9748628344502827e-05,
-      "loss": 1.0259,
+      "learning_rate": 1.972426748503165e-05,
+      "loss": 0.9743,
       "step": 3619
     },
     {
-      "epoch": 0.09942050479250776,
+      "epoch": 0.10272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9748430116549932e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.9724053107115337e-05,
+      "loss": 0.9743,
       "step": 3620
     },
     {
-      "epoch": 0.099447969020351,
+      "epoch": 0.10275255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 1.9748231811463686e-05,
-      "loss": 1.1882,
+      "learning_rate": 1.972383864705951e-05,
+      "loss": 1.0058,
       "step": 3621
     },
     {
-      "epoch": 0.09947543324819423,
+      "epoch": 0.10278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.974803342924565e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.972362410486598e-05,
+      "loss": 1.0513,
       "step": 3622
     },
     {
-      "epoch": 0.09950289747603747,
+      "epoch": 0.10280930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.9747834969897398e-05,
-      "loss": 1.0864,
+      "learning_rate": 1.9723409480536566e-05,
+      "loss": 1.0341,
       "step": 3623
     },
     {
-      "epoch": 0.0995303617038807,
+      "epoch": 0.10283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.9747636433420497e-05,
-      "loss": 1.2733,
+      "learning_rate": 1.9723194774073068e-05,
+      "loss": 1.0746,
       "step": 3624
     },
     {
-      "epoch": 0.09955782593172392,
+      "epoch": 0.10286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.9747437819816526e-05,
-      "loss": 0.93,
+      "learning_rate": 1.9722979985477315e-05,
+      "loss": 1.0501,
       "step": 3625
     },
     {
-      "epoch": 0.09958529015956716,
+      "epoch": 0.102894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9747239129087048e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9722765114751105e-05,
+      "loss": 1.0183,
       "step": 3626
     },
     {
-      "epoch": 0.0996127543874104,
+      "epoch": 0.1029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.974704036123364e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.9722550161896263e-05,
+      "loss": 1.0383,
       "step": 3627
     },
     {
-      "epoch": 0.09964021861525363,
+      "epoch": 0.10295119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.974684151625787e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.9722335126914607e-05,
+      "loss": 1.0457,
       "step": 3628
     },
     {
-      "epoch": 0.09966768284309686,
+      "epoch": 0.10297956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 1.9746642594161317e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.9722120009807947e-05,
+      "loss": 1.0242,
       "step": 3629
     },
     {
-      "epoch": 0.0996951470709401,
+      "epoch": 0.10300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.9746443594945553e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.97219048105781e-05,
+      "loss": 0.9679,
       "step": 3630
     },
     {
-      "epoch": 0.09972261129878333,
+      "epoch": 0.10303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9746244518612152e-05,
-      "loss": 1.1086,
+      "learning_rate": 1.9721689529226885e-05,
+      "loss": 1.0583,
       "step": 3631
     },
     {
-      "epoch": 0.09975007552662657,
+      "epoch": 0.10306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9746045365162687e-05,
-      "loss": 1.1094,
+      "learning_rate": 1.9721474165756124e-05,
+      "loss": 1.0414,
       "step": 3632
     },
     {
-      "epoch": 0.0997775397544698,
+      "epoch": 0.10309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9745846134598735e-05,
-      "loss": 1.246,
+      "learning_rate": 1.9721258720167634e-05,
+      "loss": 1.0932,
       "step": 3633
     },
     {
-      "epoch": 0.09980500398231304,
+      "epoch": 0.10312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.9745646826921876e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.9721043192463234e-05,
+      "loss": 0.9842,
       "step": 3634
     },
     {
-      "epoch": 0.09983246821015627,
+      "epoch": 0.10314982973893302,
       "grad_norm": 0.0,
-      "learning_rate": 1.9745447442133683e-05,
-      "loss": 1.0524,
+      "learning_rate": 1.9720827582644745e-05,
+      "loss": 1.0837,
       "step": 3635
     },
     {
-      "epoch": 0.09985993243799951,
+      "epoch": 0.10317820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9745247980235736e-05,
-      "loss": 1.0716,
+      "learning_rate": 1.9720611890713988e-05,
+      "loss": 0.9911,
       "step": 3636
     },
     {
-      "epoch": 0.09988739666584275,
+      "epoch": 0.10320658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.974504844122961e-05,
-      "loss": 1.0739,
+      "learning_rate": 1.9720396116672786e-05,
+      "loss": 0.9745,
       "step": 3637
     },
     {
-      "epoch": 0.09991486089368598,
+      "epoch": 0.10323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.9744848825116887e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.9720180260522964e-05,
+      "loss": 1.149,
       "step": 3638
     },
     {
-      "epoch": 0.0999423251215292,
+      "epoch": 0.1032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9744649131899145e-05,
-      "loss": 1.0169,
+      "learning_rate": 1.9719964322266335e-05,
+      "loss": 1.1184,
       "step": 3639
     },
     {
-      "epoch": 0.09996978934937244,
+      "epoch": 0.10329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.9744449361577963e-05,
-      "loss": 0.9448,
+      "learning_rate": 1.9719748301904736e-05,
+      "loss": 1.0411,
       "step": 3640
     },
     {
-      "epoch": 0.09999725357721567,
+      "epoch": 0.10332009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 1.9744249514154925e-05,
-      "loss": 1.0425,
+      "learning_rate": 1.9719532199439986e-05,
+      "loss": 1.1326,
       "step": 3641
     },
     {
-      "epoch": 0.10002471780505891,
+      "epoch": 0.10334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.9744049589631608e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.971931601487391e-05,
+      "loss": 1.0947,
       "step": 3642
     },
     {
-      "epoch": 0.10005218203290214,
+      "epoch": 0.10337684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 1.97438495880096e-05,
-      "loss": 1.1885,
+      "learning_rate": 1.9719099748208332e-05,
+      "loss": 1.0917,
       "step": 3643
     },
     {
-      "epoch": 0.10007964626074538,
+      "epoch": 0.10340522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 1.9743649509290478e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.9718883399445087e-05,
+      "loss": 0.9757,
       "step": 3644
     },
     {
-      "epoch": 0.10010711048858861,
+      "epoch": 0.10343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.9743449353475826e-05,
-      "loss": 1.0815,
+      "learning_rate": 1.9718666968585992e-05,
+      "loss": 1.0706,
       "step": 3645
     },
     {
-      "epoch": 0.10013457471643185,
+      "epoch": 0.10346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.974324912056723e-05,
-      "loss": 1.1072,
+      "learning_rate": 1.9718450455632884e-05,
+      "loss": 1.0762,
       "step": 3646
     },
     {
-      "epoch": 0.10016203894427508,
+      "epoch": 0.10349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.974304881056627e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.9718233860587586e-05,
+      "loss": 1.0131,
       "step": 3647
     },
     {
-      "epoch": 0.10018950317211832,
+      "epoch": 0.10351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9742848423474537e-05,
-      "loss": 1.08,
+      "learning_rate": 1.971801718345193e-05,
+      "loss": 1.1465,
       "step": 3648
     },
     {
-      "epoch": 0.10021696739996155,
+      "epoch": 0.10354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.974264795929361e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.9717800424227744e-05,
+      "loss": 0.9613,
       "step": 3649
     },
     {
-      "epoch": 0.10024443162780479,
+      "epoch": 0.10357548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.974244741802508e-05,
-      "loss": 1.1405,
+      "learning_rate": 1.9717583582916862e-05,
+      "loss": 1.0522,
       "step": 3650
     },
     {
-      "epoch": 0.10027189585564802,
+      "epoch": 0.1036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.974224679967053e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.971736665952112e-05,
+      "loss": 0.9089,
       "step": 3651
     },
     {
-      "epoch": 0.10029936008349125,
+      "epoch": 0.10363223609534619,
       "grad_norm": 0.0,
-      "learning_rate": 1.9742046104231552e-05,
-      "loss": 1.0856,
+      "learning_rate": 1.9717149654042337e-05,
+      "loss": 0.994,
       "step": 3652
     },
     {
-      "epoch": 0.10032682431133448,
+      "epoch": 0.10366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.974184533170973e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.971693256648236e-05,
+      "loss": 1.1401,
       "step": 3653
     },
     {
-      "epoch": 0.10035428853917772,
+      "epoch": 0.10368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.9741644482106655e-05,
-      "loss": 1.1505,
+      "learning_rate": 1.971671539684301e-05,
+      "loss": 1.0274,
       "step": 3654
     },
     {
-      "epoch": 0.10038175276702095,
+      "epoch": 0.10371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.9741443555423916e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.9716498145126134e-05,
+      "loss": 1.0338,
       "step": 3655
     },
     {
-      "epoch": 0.10040921699486419,
+      "epoch": 0.10374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.9741242551663104e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.9716280811333563e-05,
+      "loss": 0.9206,
       "step": 3656
     },
     {
-      "epoch": 0.10043668122270742,
+      "epoch": 0.10377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.9741041470825807e-05,
-      "loss": 1.0341,
+      "learning_rate": 1.9716063395467128e-05,
+      "loss": 1.1053,
       "step": 3657
     },
     {
-      "epoch": 0.10046414545055066,
+      "epoch": 0.10380249716231556,
       "grad_norm": 0.0,
-      "learning_rate": 1.974084031291361e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.9715845897528666e-05,
+      "loss": 0.9131,
       "step": 3658
     },
     {
-      "epoch": 0.10049160967839389,
+      "epoch": 0.10383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.9740639077928117e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.9715628317520022e-05,
+      "loss": 1.0494,
       "step": 3659
     },
     {
-      "epoch": 0.10051907390623713,
+      "epoch": 0.10385925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.9740437765870914e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.9715410655443027e-05,
+      "loss": 1.0027,
       "step": 3660
     },
     {
-      "epoch": 0.10054653813408036,
+      "epoch": 0.10388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.9740236376743592e-05,
-      "loss": 0.9521,
+      "learning_rate": 1.971519291129952e-05,
+      "loss": 0.998,
       "step": 3661
     },
     {
-      "epoch": 0.1005740023619236,
+      "epoch": 0.10391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.974003491054775e-05,
-      "loss": 1.0794,
+      "learning_rate": 1.9714975085091343e-05,
+      "loss": 1.1465,
       "step": 3662
     },
     {
-      "epoch": 0.10060146658976683,
+      "epoch": 0.10394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.9739833367284976e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.9714757176820336e-05,
+      "loss": 1.0056,
       "step": 3663
     },
     {
-      "epoch": 0.10062893081761007,
+      "epoch": 0.10397275822928491,
       "grad_norm": 0.0,
-      "learning_rate": 1.973963174695687e-05,
-      "loss": 1.1185,
+      "learning_rate": 1.9714539186488336e-05,
+      "loss": 0.9564,
       "step": 3664
     },
     {
-      "epoch": 0.1006563950454533,
+      "epoch": 0.1040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9739430049565022e-05,
-      "loss": 1.0935,
+      "learning_rate": 1.9714321114097187e-05,
+      "loss": 1.0884,
       "step": 3665
     },
     {
-      "epoch": 0.10068385927329652,
+      "epoch": 0.10402951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 1.973922827511103e-05,
-      "loss": 0.9282,
+      "learning_rate": 1.9714102959648732e-05,
+      "loss": 1.0174,
       "step": 3666
     },
     {
-      "epoch": 0.10071132350113976,
+      "epoch": 0.10405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9739026423596498e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.971388472314481e-05,
+      "loss": 0.9662,
       "step": 3667
     },
     {
-      "epoch": 0.100738787728983,
+      "epoch": 0.10408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.9738824495023012e-05,
-      "loss": 1.0856,
+      "learning_rate": 1.9713666404587274e-05,
+      "loss": 1.0676,
       "step": 3668
     },
     {
-      "epoch": 0.10076625195682623,
+      "epoch": 0.10411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.9738622489392175e-05,
-      "loss": 1.2155,
+      "learning_rate": 1.9713448003977956e-05,
+      "loss": 1.0526,
       "step": 3669
     },
     {
-      "epoch": 0.10079371618466947,
+      "epoch": 0.10414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.9738420406705582e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.9713229521318705e-05,
+      "loss": 1.0454,
       "step": 3670
     },
     {
-      "epoch": 0.1008211804125127,
+      "epoch": 0.10417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.973821824696484e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.971301095661137e-05,
+      "loss": 0.9981,
       "step": 3671
     },
     {
-      "epoch": 0.10084864464035594,
+      "epoch": 0.10419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9738016010171538e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.9712792309857795e-05,
+      "loss": 1.0312,
       "step": 3672
     },
     {
-      "epoch": 0.10087610886819917,
+      "epoch": 0.10422814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.9737813696327278e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.9712573581059826e-05,
+      "loss": 1.1616,
       "step": 3673
     },
     {
-      "epoch": 0.1009035730960424,
+      "epoch": 0.10425652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.973761130543367e-05,
-      "loss": 1.0587,
+      "learning_rate": 1.971235477021931e-05,
+      "loss": 1.0757,
       "step": 3674
     },
     {
-      "epoch": 0.10093103732388564,
+      "epoch": 0.10428490351872871,
       "grad_norm": 0.0,
-      "learning_rate": 1.9737408837492307e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.97121358773381e-05,
+      "loss": 1.0664,
       "step": 3675
     },
     {
-      "epoch": 0.10095850155172888,
+      "epoch": 0.10431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9737206292504794e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.9711916902418037e-05,
+      "loss": 1.0455,
       "step": 3676
     },
     {
-      "epoch": 0.10098596577957211,
+      "epoch": 0.1043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.9737003670472737e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.9711697845460978e-05,
+      "loss": 0.973,
       "step": 3677
     },
     {
-      "epoch": 0.10101343000741535,
+      "epoch": 0.1043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.973680097139773e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.971147870646877e-05,
+      "loss": 0.9734,
       "step": 3678
     },
     {
-      "epoch": 0.10104089423525858,
+      "epoch": 0.10439841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.9736598195281384e-05,
-      "loss": 1.0783,
+      "learning_rate": 1.9711259485443266e-05,
+      "loss": 0.962,
       "step": 3679
     },
     {
-      "epoch": 0.1010683584631018,
+      "epoch": 0.10442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.9736395342125297e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.9711040182386317e-05,
+      "loss": 1.0018,
       "step": 3680
     },
     {
-      "epoch": 0.10109582269094504,
+      "epoch": 0.10445516458569808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9736192411931086e-05,
-      "loss": 1.1953,
+      "learning_rate": 1.9710820797299773e-05,
+      "loss": 1.1083,
       "step": 3681
     },
     {
-      "epoch": 0.10112328691878827,
+      "epoch": 0.10448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.9735989404700343e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.971060133018549e-05,
+      "loss": 0.9137,
       "step": 3682
     },
     {
-      "epoch": 0.10115075114663151,
+      "epoch": 0.10451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.9735786320434685e-05,
-      "loss": 1.123,
+      "learning_rate": 1.971038178104532e-05,
+      "loss": 1.1232,
       "step": 3683
     },
     {
-      "epoch": 0.10117821537447474,
+      "epoch": 0.10454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.9735583159135713e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.971016214988112e-05,
+      "loss": 0.9523,
       "step": 3684
     },
     {
-      "epoch": 0.10120567960231798,
+      "epoch": 0.10456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9735379920805034e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.9709942436694745e-05,
+      "loss": 1.0433,
       "step": 3685
     },
     {
-      "epoch": 0.10123314383016122,
+      "epoch": 0.10459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.9735176605444258e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.970972264148805e-05,
+      "loss": 1.0799,
       "step": 3686
     },
     {
-      "epoch": 0.10126060805800445,
+      "epoch": 0.10462542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 1.9734973213054996e-05,
-      "loss": 1.0986,
+      "learning_rate": 1.970950276426289e-05,
+      "loss": 0.9716,
       "step": 3687
     },
     {
-      "epoch": 0.10128807228584769,
+      "epoch": 0.10465380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.9734769743638852e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.970928280502112e-05,
+      "loss": 0.9983,
       "step": 3688
     },
     {
-      "epoch": 0.10131553651369092,
+      "epoch": 0.1046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.973456619719744e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.9709062763764604e-05,
+      "loss": 1.0112,
       "step": 3689
     },
     {
-      "epoch": 0.10134300074153416,
+      "epoch": 0.1047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.973436257373237e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.97088426404952e-05,
+      "loss": 1.0959,
       "step": 3690
     },
     {
-      "epoch": 0.10137046496937739,
+      "epoch": 0.10473893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.9734158873245253e-05,
-      "loss": 1.1392,
+      "learning_rate": 1.9708622435214768e-05,
+      "loss": 1.1113,
       "step": 3691
     },
     {
-      "epoch": 0.10139792919722063,
+      "epoch": 0.10476730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.97339550957377e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.9708402147925164e-05,
+      "loss": 1.0699,
       "step": 3692
     },
     {
-      "epoch": 0.10142539342506385,
+      "epoch": 0.10479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.9733751241211324e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.9708181778628247e-05,
+      "loss": 0.9507,
       "step": 3693
     },
     {
-      "epoch": 0.10145285765290708,
+      "epoch": 0.10482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9733547309667736e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.9707961327325886e-05,
+      "loss": 1.1241,
       "step": 3694
     },
     {
-      "epoch": 0.10148032188075032,
+      "epoch": 0.10485244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9733343301108555e-05,
-      "loss": 1.1102,
+      "learning_rate": 1.9707740794019937e-05,
+      "loss": 1.0333,
       "step": 3695
     },
     {
-      "epoch": 0.10150778610859355,
+      "epoch": 0.10488081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.973313921553539e-05,
-      "loss": 1.022,
+      "learning_rate": 1.9707520178712268e-05,
+      "loss": 1.0883,
       "step": 3696
     },
     {
-      "epoch": 0.10153525033643679,
+      "epoch": 0.10490919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.9732935052949857e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.970729948140474e-05,
+      "loss": 1.1978,
       "step": 3697
     },
     {
-      "epoch": 0.10156271456428002,
+      "epoch": 0.10493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.973273081335357e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.9707078702099213e-05,
+      "loss": 0.9967,
       "step": 3698
     },
     {
-      "epoch": 0.10159017879212326,
+      "epoch": 0.10496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9732526496748146e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.9706857840797557e-05,
+      "loss": 0.965,
       "step": 3699
     },
     {
-      "epoch": 0.1016176430199665,
+      "epoch": 0.10499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.9732322103135204e-05,
-      "loss": 1.122,
+      "learning_rate": 1.9706636897501637e-05,
+      "loss": 1.0683,
       "step": 3700
     },
     {
-      "epoch": 0.10164510724780973,
+      "epoch": 0.10502270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 1.973211763251636e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.970641587221332e-05,
+      "loss": 0.934,
       "step": 3701
     },
     {
-      "epoch": 0.10167257147565296,
+      "epoch": 0.10505107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 1.9731913084893232e-05,
-      "loss": 1.1284,
+      "learning_rate": 1.970619476493447e-05,
+      "loss": 1.1224,
       "step": 3702
     },
     {
-      "epoch": 0.1017000357034962,
+      "epoch": 0.1050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.9731708460267436e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.9705973575666956e-05,
+      "loss": 1.1058,
       "step": 3703
     },
     {
-      "epoch": 0.10172749993133944,
+      "epoch": 0.1051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9731503758640594e-05,
-      "loss": 1.1461,
+      "learning_rate": 1.970575230441265e-05,
+      "loss": 0.9666,
       "step": 3704
     },
     {
-      "epoch": 0.10175496415918267,
+      "epoch": 0.10513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.9731298980014325e-05,
-      "loss": 1.1265,
+      "learning_rate": 1.9705530951173414e-05,
+      "loss": 1.1479,
       "step": 3705
     },
     {
-      "epoch": 0.1017824283870259,
+      "epoch": 0.10516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.973109412439025e-05,
-      "loss": 1.1096,
+      "learning_rate": 1.9705309515951123e-05,
+      "loss": 1.0548,
       "step": 3706
     },
     {
-      "epoch": 0.10180989261486913,
+      "epoch": 0.10519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.9730889191769985e-05,
-      "loss": 0.989,
+      "learning_rate": 1.970508799874765e-05,
+      "loss": 0.9705,
       "step": 3707
     },
     {
-      "epoch": 0.10183735684271236,
+      "epoch": 0.10522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.9730684182155157e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.970486639956486e-05,
+      "loss": 1.031,
       "step": 3708
     },
     {
-      "epoch": 0.1018648210705556,
+      "epoch": 0.10524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.973047909554739e-05,
-      "loss": 1.145,
+      "learning_rate": 1.9704644718404626e-05,
+      "loss": 0.9939,
       "step": 3709
     },
     {
-      "epoch": 0.10189228529839883,
+      "epoch": 0.10527809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.9730273931948298e-05,
-      "loss": 1.1227,
+      "learning_rate": 1.970442295526882e-05,
+      "loss": 1.0391,
       "step": 3710
     },
     {
-      "epoch": 0.10191974952624207,
+      "epoch": 0.10530646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.973006869135951e-05,
-      "loss": 1.1874,
+      "learning_rate": 1.970420111015932e-05,
+      "loss": 0.9757,
       "step": 3711
     },
     {
-      "epoch": 0.1019472137540853,
+      "epoch": 0.10533484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 1.9729863373782654e-05,
-      "loss": 1.0632,
+      "learning_rate": 1.9703979183078e-05,
+      "loss": 1.0735,
       "step": 3712
     },
     {
-      "epoch": 0.10197467798192854,
+      "epoch": 0.10536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9729657979219348e-05,
-      "loss": 1.037,
+      "learning_rate": 1.9703757174026726e-05,
+      "loss": 1.0389,
       "step": 3713
     },
     {
-      "epoch": 0.10200214220977177,
+      "epoch": 0.10539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.9729452507671216e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.970353508300738e-05,
+      "loss": 1.0256,
       "step": 3714
     },
     {
-      "epoch": 0.10202960643761501,
+      "epoch": 0.1054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9729246959139892e-05,
-      "loss": 1.2003,
+      "learning_rate": 1.970331291002184e-05,
+      "loss": 1.0171,
       "step": 3715
     },
     {
-      "epoch": 0.10205707066545824,
+      "epoch": 0.1054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9729041333626995e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.9703090655071978e-05,
+      "loss": 1.1111,
       "step": 3716
     },
     {
-      "epoch": 0.10208453489330148,
+      "epoch": 0.10547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.972883563113415e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.9702868318159673e-05,
+      "loss": 1.1331,
       "step": 3717
     },
     {
-      "epoch": 0.10211199912114471,
+      "epoch": 0.10550510783200909,
       "grad_norm": 0.0,
-      "learning_rate": 1.9728629851662994e-05,
-      "loss": 1.108,
+      "learning_rate": 1.97026458992868e-05,
+      "loss": 0.9752,
       "step": 3718
     },
     {
-      "epoch": 0.10213946334898795,
+      "epoch": 0.10553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.9728423995215153e-05,
-      "loss": 1.1487,
+      "learning_rate": 1.9702423398455245e-05,
+      "loss": 1.1581,
       "step": 3719
     },
     {
-      "epoch": 0.10216692757683117,
+      "epoch": 0.10556186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.9728218061792247e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.970220081566688e-05,
+      "loss": 0.992,
       "step": 3720
     },
     {
-      "epoch": 0.1021943918046744,
+      "epoch": 0.10559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.9728012051395914e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.9701978150923594e-05,
+      "loss": 1.0598,
       "step": 3721
     },
     {
-      "epoch": 0.10222185603251764,
+      "epoch": 0.10561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.972780596402778e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.970175540422726e-05,
+      "loss": 1.1051,
       "step": 3722
     },
     {
-      "epoch": 0.10224932026036088,
+      "epoch": 0.10564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9727599799689475e-05,
-      "loss": 1.074,
+      "learning_rate": 1.9701532575579757e-05,
+      "loss": 1.0385,
       "step": 3723
     },
     {
-      "epoch": 0.10227678448820411,
+      "epoch": 0.10567536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.9727393558382636e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.970130966498298e-05,
+      "loss": 0.9106,
       "step": 3724
     },
     {
-      "epoch": 0.10230424871604735,
+      "epoch": 0.10570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.972718724010889e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.97010866724388e-05,
+      "loss": 1.1364,
       "step": 3725
     },
     {
-      "epoch": 0.10233171294389058,
+      "epoch": 0.10573212258796821,
       "grad_norm": 0.0,
-      "learning_rate": 1.972698084486987e-05,
-      "loss": 1.0088,
+      "learning_rate": 1.9700863597949104e-05,
+      "loss": 1.0276,
       "step": 3726
     },
     {
-      "epoch": 0.10235917717173382,
+      "epoch": 0.10576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9726774372667213e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.970064044151578e-05,
+      "loss": 1.0707,
       "step": 3727
     },
     {
-      "epoch": 0.10238664139957705,
+      "epoch": 0.105788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9726567823502546e-05,
-      "loss": 1.1208,
+      "learning_rate": 1.970041720314071e-05,
+      "loss": 1.0216,
       "step": 3728
     },
     {
-      "epoch": 0.10241410562742029,
+      "epoch": 0.10581725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 1.9726361197377506e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.9700193882825777e-05,
+      "loss": 1.0145,
       "step": 3729
     },
     {
-      "epoch": 0.10244156985526352,
+      "epoch": 0.10584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.972615449429373e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.969997048057287e-05,
+      "loss": 1.0427,
       "step": 3730
     },
     {
-      "epoch": 0.10246903408310676,
+      "epoch": 0.10587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.972594771425285e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.969974699638388e-05,
+      "loss": 1.0398,
       "step": 3731
     },
     {
-      "epoch": 0.10249649831095,
+      "epoch": 0.10590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.972574085725651e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.969952343026069e-05,
+      "loss": 1.0881,
       "step": 3732
     },
     {
-      "epoch": 0.10252396253879323,
+      "epoch": 0.10593076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9725533923306336e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9699299782205186e-05,
+      "loss": 1.1079,
       "step": 3733
     },
     {
-      "epoch": 0.10255142676663645,
+      "epoch": 0.10595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.972532691240397e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.9699076052219263e-05,
+      "loss": 1.0644,
       "step": 3734
     },
     {
-      "epoch": 0.10257889099447969,
+      "epoch": 0.10598751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.9725119824551054e-05,
-      "loss": 1.1241,
+      "learning_rate": 1.969885224030481e-05,
+      "loss": 1.011,
       "step": 3735
     },
     {
-      "epoch": 0.10260635522232292,
+      "epoch": 0.10601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.972491265974922e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.9698628346463714e-05,
+      "loss": 1.0438,
       "step": 3736
     },
     {
-      "epoch": 0.10263381945016616,
+      "epoch": 0.10604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.972470541800011e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.969840437069787e-05,
+      "loss": 1.02,
       "step": 3737
     },
     {
-      "epoch": 0.10266128367800939,
+      "epoch": 0.10607264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 1.9724498099305362e-05,
-      "loss": 1.1336,
+      "learning_rate": 1.9698180313009166e-05,
+      "loss": 1.0275,
       "step": 3738
     },
     {
-      "epoch": 0.10268874790585263,
+      "epoch": 0.10610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.9724290703666623e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.96979561733995e-05,
+      "loss": 1.1321,
       "step": 3739
     },
     {
-      "epoch": 0.10271621213369586,
+      "epoch": 0.1061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.9724083231085526e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.969773195187076e-05,
+      "loss": 1.0251,
       "step": 3740
     },
     {
-      "epoch": 0.1027436763615391,
+      "epoch": 0.1061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9723875681563717e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.9697507648424844e-05,
+      "loss": 1.1226,
       "step": 3741
     },
     {
-      "epoch": 0.10277114058938233,
+      "epoch": 0.1061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9723668055102835e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.9697283263063645e-05,
+      "loss": 1.0585,
       "step": 3742
     },
     {
-      "epoch": 0.10279860481722557,
+      "epoch": 0.10621452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.9723460351704525e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.9697058795789056e-05,
+      "loss": 1.0884,
       "step": 3743
     },
     {
-      "epoch": 0.1028260690450688,
+      "epoch": 0.10624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.9723252571370432e-05,
-      "loss": 1.049,
+      "learning_rate": 1.9696834246602977e-05,
+      "loss": 1.0474,
       "step": 3744
     },
     {
-      "epoch": 0.10285353327291204,
+      "epoch": 0.10627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.9723044714102196e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.9696609615507304e-05,
+      "loss": 1.0028,
       "step": 3745
     },
     {
-      "epoch": 0.10288099750075527,
+      "epoch": 0.10629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9722836779901465e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.9696384902503932e-05,
+      "loss": 0.9473,
       "step": 3746
     },
     {
-      "epoch": 0.1029084617285985,
+      "epoch": 0.10632803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.972262876876988e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.969616010759476e-05,
+      "loss": 1.0796,
       "step": 3747
     },
     {
-      "epoch": 0.10293592595644173,
+      "epoch": 0.10635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.9722420680709094e-05,
-      "loss": 1.093,
+      "learning_rate": 1.969593523078169e-05,
+      "loss": 1.0498,
       "step": 3748
     },
     {
-      "epoch": 0.10296339018428496,
+      "epoch": 0.10638479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 1.9722212515720752e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.9695710272066623e-05,
+      "loss": 1.0355,
       "step": 3749
     },
     {
-      "epoch": 0.1029908544121282,
+      "epoch": 0.10641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9722004273806496e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.9695485231451448e-05,
+      "loss": 1.045,
       "step": 3750
     },
     {
-      "epoch": 0.10301831863997143,
+      "epoch": 0.10644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.9721795954967975e-05,
-      "loss": 1.1388,
+      "learning_rate": 1.9695260108938078e-05,
+      "loss": 0.9935,
       "step": 3751
     },
     {
-      "epoch": 0.10304578286781467,
+      "epoch": 0.10646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.9721587559206837e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.969503490452841e-05,
+      "loss": 0.965,
       "step": 3752
     },
     {
-      "epoch": 0.1030732470956579,
+      "epoch": 0.10649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.9721379086524734e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.969480961822434e-05,
+      "loss": 0.9568,
       "step": 3753
     },
     {
-      "epoch": 0.10310071132350114,
+      "epoch": 0.1065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9721170536923314e-05,
-      "loss": 1.1133,
+      "learning_rate": 1.9694584250027784e-05,
+      "loss": 1.1124,
       "step": 3754
     },
     {
-      "epoch": 0.10312817555134438,
+      "epoch": 0.10655505107832008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9720961910404232e-05,
-      "loss": 1.1515,
+      "learning_rate": 1.9694358799940637e-05,
+      "loss": 1.0518,
       "step": 3755
     },
     {
-      "epoch": 0.10315563977918761,
+      "epoch": 0.10658342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 1.9720753206969127e-05,
-      "loss": 0.978,
+      "learning_rate": 1.9694133267964804e-05,
+      "loss": 1.0455,
       "step": 3756
     },
     {
-      "epoch": 0.10318310400703085,
+      "epoch": 0.10661180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.972054442661966e-05,
-      "loss": 1.108,
+      "learning_rate": 1.9693907654102197e-05,
+      "loss": 1.0206,
       "step": 3757
     },
     {
-      "epoch": 0.10321056823487408,
+      "epoch": 0.10664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.972033556935748e-05,
-      "loss": 1.0735,
+      "learning_rate": 1.969368195835471e-05,
+      "loss": 1.0918,
       "step": 3758
     },
     {
-      "epoch": 0.10323803246271732,
+      "epoch": 0.10666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.972012663518424e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.969345618072426e-05,
+      "loss": 1.1094,
       "step": 3759
     },
     {
-      "epoch": 0.10326549669056055,
+      "epoch": 0.10669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.971991762410159e-05,
-      "loss": 1.1334,
+      "learning_rate": 1.9693230321212748e-05,
+      "loss": 1.0937,
       "step": 3760
     },
     {
-      "epoch": 0.10329296091840377,
+      "epoch": 0.10672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.971970853611119e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.9693004379822083e-05,
+      "loss": 1.0017,
       "step": 3761
     },
     {
-      "epoch": 0.10332042514624701,
+      "epoch": 0.10675368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.971949937121469e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.9692778356554174e-05,
+      "loss": 0.9722,
       "step": 3762
     },
     {
-      "epoch": 0.10334788937409024,
+      "epoch": 0.10678206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 1.9719290129413745e-05,
-      "loss": 1.1665,
+      "learning_rate": 1.969255225141093e-05,
+      "loss": 1.0777,
       "step": 3763
     },
     {
-      "epoch": 0.10337535360193348,
+      "epoch": 0.10681044267877413,
       "grad_norm": 0.0,
-      "learning_rate": 1.971908081071001e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.9692326064394266e-05,
+      "loss": 1.0916,
       "step": 3764
     },
     {
-      "epoch": 0.10340281782977671,
+      "epoch": 0.10683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.9718871415105148e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.9692099795506084e-05,
+      "loss": 1.1364,
       "step": 3765
     },
     {
-      "epoch": 0.10343028205761995,
+      "epoch": 0.1068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.971866194260081e-05,
-      "loss": 1.046,
+      "learning_rate": 1.96918734447483e-05,
+      "loss": 1.0923,
       "step": 3766
     },
     {
-      "epoch": 0.10345774628546318,
+      "epoch": 0.1068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.971845239319865e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.9691647012122825e-05,
+      "loss": 1.0422,
       "step": 3767
     },
     {
-      "epoch": 0.10348521051330642,
+      "epoch": 0.10692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.9718242766900332e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.9691420497631576e-05,
+      "loss": 1.0167,
       "step": 3768
     },
     {
-      "epoch": 0.10351267474114965,
+      "epoch": 0.10695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.971803306370751e-05,
-      "loss": 1.1491,
+      "learning_rate": 1.9691193901276456e-05,
+      "loss": 0.9993,
       "step": 3769
     },
     {
-      "epoch": 0.10354013896899289,
+      "epoch": 0.10698070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 1.9717823283621853e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.9690967223059386e-05,
+      "loss": 0.9253,
       "step": 3770
     },
     {
-      "epoch": 0.10356760319683612,
+      "epoch": 0.10700908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.9717613426645008e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9690740462982284e-05,
+      "loss": 1.2001,
       "step": 3771
     },
     {
-      "epoch": 0.10359506742467936,
+      "epoch": 0.10703745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9717403492778643e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9690513621047064e-05,
+      "loss": 0.9581,
       "step": 3772
     },
     {
-      "epoch": 0.1036225316525226,
+      "epoch": 0.10706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.971719348202442e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.9690286697255633e-05,
+      "loss": 1.0869,
       "step": 3773
     },
     {
-      "epoch": 0.10364999588036582,
+      "epoch": 0.10709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.9716983394383993e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.969005969160992e-05,
+      "loss": 1.0121,
       "step": 3774
     },
     {
-      "epoch": 0.10367746010820905,
+      "epoch": 0.10712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9716773229859033e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.9689832604111836e-05,
+      "loss": 1.059,
       "step": 3775
     },
     {
-      "epoch": 0.10370492433605229,
+      "epoch": 0.10715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.97165629884512e-05,
-      "loss": 1.0975,
+      "learning_rate": 1.9689605434763298e-05,
+      "loss": 1.0981,
       "step": 3776
     },
     {
-      "epoch": 0.10373238856389552,
+      "epoch": 0.10717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.971635267016215e-05,
-      "loss": 1.0454,
+      "learning_rate": 1.968937818356623e-05,
+      "loss": 1.0545,
       "step": 3777
     },
     {
-      "epoch": 0.10375985279173876,
+      "epoch": 0.10720771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 1.971614227499356e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.9689150850522548e-05,
+      "loss": 1.0156,
       "step": 3778
     },
     {
-      "epoch": 0.10378731701958199,
+      "epoch": 0.1072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9715931802947088e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.9688923435634174e-05,
+      "loss": 1.0841,
       "step": 3779
     },
     {
-      "epoch": 0.10381478124742523,
+      "epoch": 0.10726447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 1.9715721254024397e-05,
-      "loss": 1.0994,
+      "learning_rate": 1.9688695938903025e-05,
+      "loss": 0.9929,
       "step": 3780
     },
     {
-      "epoch": 0.10384224547526846,
+      "epoch": 0.1072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.971551062822716e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.9688468360331026e-05,
+      "loss": 1.0351,
       "step": 3781
     },
     {
-      "epoch": 0.1038697097031117,
+      "epoch": 0.10732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.971529992555704e-05,
-      "loss": 1.1814,
+      "learning_rate": 1.9688240699920104e-05,
+      "loss": 0.9815,
       "step": 3782
     },
     {
-      "epoch": 0.10389717393095493,
+      "epoch": 0.10734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.97150891460157e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.9688012957672172e-05,
+      "loss": 1.2264,
       "step": 3783
     },
     {
-      "epoch": 0.10392463815879817,
+      "epoch": 0.10737797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.9714878289604813e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.9687785133589165e-05,
+      "loss": 0.9223,
       "step": 3784
     },
     {
-      "epoch": 0.1039521023866414,
+      "epoch": 0.10740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.971466735632605e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.9687557227672997e-05,
+      "loss": 1.038,
       "step": 3785
     },
     {
-      "epoch": 0.10397956661448464,
+      "epoch": 0.10743473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.9714456346181067e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.9687329239925595e-05,
+      "loss": 1.0379,
       "step": 3786
     },
     {
-      "epoch": 0.10400703084232787,
+      "epoch": 0.10746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9714245259171548e-05,
-      "loss": 1.085,
+      "learning_rate": 1.9687101170348893e-05,
+      "loss": 1.0182,
       "step": 3787
     },
     {
-      "epoch": 0.1040344950701711,
+      "epoch": 0.10749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.9714034095299156e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.9686873018944812e-05,
+      "loss": 1.0819,
       "step": 3788
     },
     {
-      "epoch": 0.10406195929801433,
+      "epoch": 0.10751986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.971382285456556e-05,
-      "loss": 1.1244,
+      "learning_rate": 1.9686644785715276e-05,
+      "loss": 0.982,
       "step": 3789
     },
     {
-      "epoch": 0.10408942352585757,
+      "epoch": 0.10754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.9713611536972445e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.9686416470662218e-05,
+      "loss": 0.9872,
       "step": 3790
     },
     {
-      "epoch": 0.1041168877537008,
+      "epoch": 0.10757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9713400142521462e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.9686188073787563e-05,
+      "loss": 1.0275,
       "step": 3791
     },
     {
-      "epoch": 0.10414435198154404,
+      "epoch": 0.1076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.97131886712143e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.968595959509324e-05,
+      "loss": 0.9137,
       "step": 3792
     },
     {
-      "epoch": 0.10417181620938727,
+      "epoch": 0.107633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9712977123052626e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.968573103458119e-05,
+      "loss": 1.0807,
       "step": 3793
     },
     {
-      "epoch": 0.10419928043723051,
+      "epoch": 0.10766174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 1.9712765498038117e-05,
-      "loss": 1.0171,
+      "learning_rate": 1.9685502392253326e-05,
+      "loss": 1.0134,
       "step": 3794
     },
     {
-      "epoch": 0.10422674466507374,
+      "epoch": 0.10769012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.9712553796172437e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.9685273668111588e-05,
+      "loss": 0.9907,
       "step": 3795
     },
     {
-      "epoch": 0.10425420889291698,
+      "epoch": 0.10771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.9712342017457277e-05,
-      "loss": 1.003,
+      "learning_rate": 1.9685044862157912e-05,
+      "loss": 1.0128,
       "step": 3796
     },
     {
-      "epoch": 0.10428167312076021,
+      "epoch": 0.10774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.97121301618943e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.9684815974394223e-05,
+      "loss": 1.005,
       "step": 3797
     },
     {
-      "epoch": 0.10430913734860345,
+      "epoch": 0.10777525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.971191822948519e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.9684587004822462e-05,
+      "loss": 1.0106,
       "step": 3798
     },
     {
-      "epoch": 0.10433660157644668,
+      "epoch": 0.10780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.971170622023162e-05,
-      "loss": 1.0806,
+      "learning_rate": 1.9684357953444558e-05,
+      "loss": 1.0547,
       "step": 3799
     },
     {
-      "epoch": 0.10436406580428992,
+      "epoch": 0.10783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9711494134135266e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.9684128820262444e-05,
+      "loss": 1.0667,
       "step": 3800
     },
     {
-      "epoch": 0.10439153003213314,
+      "epoch": 0.10786038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9711281971197813e-05,
-      "loss": 1.1126,
+      "learning_rate": 1.9683899605278062e-05,
+      "loss": 0.9875,
       "step": 3801
     },
     {
-      "epoch": 0.10441899425997637,
+      "epoch": 0.10788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.9711069731420936e-05,
-      "loss": 1.1569,
+      "learning_rate": 1.968367030849334e-05,
+      "loss": 1.012,
       "step": 3802
     },
     {
-      "epoch": 0.10444645848781961,
+      "epoch": 0.10791713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.971085741480631e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.9683440929910223e-05,
+      "loss": 1.0896,
       "step": 3803
     },
     {
-      "epoch": 0.10447392271566285,
+      "epoch": 0.10794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.971064502135562e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.9683211469530646e-05,
+      "loss": 0.9877,
       "step": 3804
     },
     {
-      "epoch": 0.10450138694350608,
+      "epoch": 0.1079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9710432551070543e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.9682981927356545e-05,
+      "loss": 1.0458,
       "step": 3805
     },
     {
-      "epoch": 0.10452885117134932,
+      "epoch": 0.10800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.9710220003952762e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.9682752303389864e-05,
+      "loss": 0.9866,
       "step": 3806
     },
     {
-      "epoch": 0.10455631539919255,
+      "epoch": 0.10803064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 1.9710007380003963e-05,
-      "loss": 1.1182,
+      "learning_rate": 1.9682522597632533e-05,
+      "loss": 1.011,
       "step": 3807
     },
     {
-      "epoch": 0.10458377962703579,
+      "epoch": 0.10805902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.970979467922582e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.9682292810086503e-05,
+      "loss": 1.0513,
       "step": 3808
     },
     {
-      "epoch": 0.10461124385487902,
+      "epoch": 0.10808740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 1.9709581901620023e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.968206294075371e-05,
+      "loss": 0.9928,
       "step": 3809
     },
     {
-      "epoch": 0.10463870808272226,
+      "epoch": 0.10811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9709369047188253e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.9681832989636093e-05,
+      "loss": 0.9918,
       "step": 3810
     },
     {
-      "epoch": 0.10466617231056549,
+      "epoch": 0.10814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.970915611593219e-05,
-      "loss": 1.0897,
+      "learning_rate": 1.96816029567356e-05,
+      "loss": 1.0676,
       "step": 3811
     },
     {
-      "epoch": 0.10469363653840873,
+      "epoch": 0.10817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.9708943107853526e-05,
-      "loss": 1.005,
+      "learning_rate": 1.968137284205417e-05,
+      "loss": 0.9949,
       "step": 3812
     },
     {
-      "epoch": 0.10472110076625196,
+      "epoch": 0.10820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9708730022953944e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.968114264559375e-05,
+      "loss": 0.9839,
       "step": 3813
     },
     {
-      "epoch": 0.1047485649940952,
+      "epoch": 0.10822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9708516861235128e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.968091236735628e-05,
+      "loss": 1.1088,
       "step": 3814
     },
     {
-      "epoch": 0.10477602922193842,
+      "epoch": 0.10825766174801361,
       "grad_norm": 0.0,
-      "learning_rate": 1.970830362269877e-05,
-      "loss": 1.1086,
+      "learning_rate": 1.968068200734371e-05,
+      "loss": 1.0597,
       "step": 3815
     },
     {
-      "epoch": 0.10480349344978165,
+      "epoch": 0.10828603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.9708090307346547e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.9680451565557985e-05,
+      "loss": 1.0405,
       "step": 3816
     },
     {
-      "epoch": 0.10483095767762489,
+      "epoch": 0.1083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9707876915180155e-05,
-      "loss": 1.126,
+      "learning_rate": 1.968022104200105e-05,
+      "loss": 1.0875,
       "step": 3817
     },
     {
-      "epoch": 0.10485842190546812,
+      "epoch": 0.1083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.970766344620128e-05,
-      "loss": 1.0715,
+      "learning_rate": 1.9679990436674853e-05,
+      "loss": 1.0832,
       "step": 3818
     },
     {
-      "epoch": 0.10488588613331136,
+      "epoch": 0.10837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.970744990041161e-05,
-      "loss": 1.0583,
+      "learning_rate": 1.967975974958134e-05,
+      "loss": 1.0407,
       "step": 3819
     },
     {
-      "epoch": 0.1049133503611546,
+      "epoch": 0.10839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.970723627781284e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.967952898072246e-05,
+      "loss": 1.0824,
       "step": 3820
     },
     {
-      "epoch": 0.10494081458899783,
+      "epoch": 0.10842792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 1.9707022578406653e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9679298130100165e-05,
+      "loss": 1.0775,
       "step": 3821
     },
     {
-      "epoch": 0.10496827881684107,
+      "epoch": 0.10845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.9706808802194744e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.9679067197716408e-05,
+      "loss": 1.0666,
       "step": 3822
     },
     {
-      "epoch": 0.1049957430446843,
+      "epoch": 0.10848467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.9706594949178804e-05,
-      "loss": 1.1425,
+      "learning_rate": 1.967883618357313e-05,
+      "loss": 0.9252,
       "step": 3823
     },
     {
-      "epoch": 0.10502320727252754,
+      "epoch": 0.10851305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9706381019360527e-05,
-      "loss": 1.0553,
+      "learning_rate": 1.967860508767229e-05,
+      "loss": 1.0374,
       "step": 3824
     },
     {
-      "epoch": 0.10505067150037077,
+      "epoch": 0.10854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.9706167012741602e-05,
-      "loss": 1.0947,
+      "learning_rate": 1.9678373910015834e-05,
+      "loss": 1.0803,
       "step": 3825
     },
     {
-      "epoch": 0.105078135728214,
+      "epoch": 0.10856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.9705952929323722e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.9678142650605724e-05,
+      "loss": 0.9329,
       "step": 3826
     },
     {
-      "epoch": 0.10510559995605724,
+      "epoch": 0.10859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.9705738769108587e-05,
-      "loss": 1.0778,
+      "learning_rate": 1.9677911309443907e-05,
+      "loss": 1.0905,
       "step": 3827
     },
     {
-      "epoch": 0.10513306418390048,
+      "epoch": 0.10862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.9705524532097883e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.9677679886532335e-05,
+      "loss": 1.084,
       "step": 3828
     },
     {
-      "epoch": 0.1051605284117437,
+      "epoch": 0.1086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.9705310218293313e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.9677448381872966e-05,
+      "loss": 1.0524,
       "step": 3829
     },
     {
-      "epoch": 0.10518799263958693,
+      "epoch": 0.10868331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 1.9705095827696566e-05,
-      "loss": 1.1475,
+      "learning_rate": 1.967721679546776e-05,
+      "loss": 1.1514,
       "step": 3830
     },
     {
-      "epoch": 0.10521545686743017,
+      "epoch": 0.1087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.970488136030934e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.9676985127318665e-05,
+      "loss": 1.1299,
       "step": 3831
     },
     {
-      "epoch": 0.1052429210952734,
+      "epoch": 0.10874006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 1.9704666816133336e-05,
-      "loss": 1.0527,
+      "learning_rate": 1.967675337742764e-05,
+      "loss": 1.0794,
       "step": 3832
     },
     {
-      "epoch": 0.10527038532311664,
+      "epoch": 0.10876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9704452195170248e-05,
-      "loss": 1.0512,
+      "learning_rate": 1.9676521545796648e-05,
+      "loss": 1.0464,
       "step": 3833
     },
     {
-      "epoch": 0.10529784955095987,
+      "epoch": 0.10879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.9704237497421777e-05,
-      "loss": 1.0982,
+      "learning_rate": 1.9676289632427643e-05,
+      "loss": 1.0044,
       "step": 3834
     },
     {
-      "epoch": 0.10532531377880311,
+      "epoch": 0.10882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.9704022722889616e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.9676057637322584e-05,
+      "loss": 1.1542,
       "step": 3835
     },
     {
-      "epoch": 0.10535277800664634,
+      "epoch": 0.10885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.9703807871575475e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.967582556048343e-05,
+      "loss": 1.0635,
       "step": 3836
     },
     {
-      "epoch": 0.10538024223448958,
+      "epoch": 0.10888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9703592943481042e-05,
-      "loss": 0.947,
+      "learning_rate": 1.9675593401912142e-05,
+      "loss": 1.0304,
       "step": 3837
     },
     {
-      "epoch": 0.10540770646233281,
+      "epoch": 0.10891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.9703377938608023e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.9675361161610686e-05,
+      "loss": 1.0347,
       "step": 3838
     },
     {
-      "epoch": 0.10543517069017605,
+      "epoch": 0.10893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.9703162856958118e-05,
-      "loss": 1.1363,
+      "learning_rate": 1.9675128839581013e-05,
+      "loss": 0.9639,
       "step": 3839
     },
     {
-      "epoch": 0.10546263491801929,
+      "epoch": 0.10896708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 1.970294769853303e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.96748964358251e-05,
+      "loss": 0.8531,
       "step": 3840
     },
     {
-      "epoch": 0.10549009914586252,
+      "epoch": 0.10899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.9702732463334464e-05,
-      "loss": 1.1338,
+      "learning_rate": 1.9674663950344894e-05,
+      "loss": 1.1161,
       "step": 3841
     },
     {
-      "epoch": 0.10551756337370574,
+      "epoch": 0.10902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.9702517151364116e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.967443138314237e-05,
+      "loss": 1.1257,
       "step": 3842
     },
     {
-      "epoch": 0.10554502760154898,
+      "epoch": 0.1090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.97023017626237e-05,
-      "loss": 1.1122,
+      "learning_rate": 1.967419873421949e-05,
+      "loss": 1.0345,
       "step": 3843
     },
     {
-      "epoch": 0.10557249182939221,
+      "epoch": 0.1090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.970208629711491e-05,
-      "loss": 1.1455,
+      "learning_rate": 1.9673966003578215e-05,
+      "loss": 0.9323,
       "step": 3844
     },
     {
-      "epoch": 0.10559995605723545,
+      "epoch": 0.10910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.9701870754839455e-05,
-      "loss": 1.1144,
+      "learning_rate": 1.9673733191220515e-05,
+      "loss": 1.0724,
       "step": 3845
     },
     {
-      "epoch": 0.10562742028507868,
+      "epoch": 0.10913734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.970165513579904e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.9673500297148357e-05,
+      "loss": 1.081,
       "step": 3846
     },
     {
-      "epoch": 0.10565488451292192,
+      "epoch": 0.10916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9701439439995375e-05,
-      "loss": 1.0375,
+      "learning_rate": 1.9673267321363707e-05,
+      "loss": 1.1114,
       "step": 3847
     },
     {
-      "epoch": 0.10568234874076515,
+      "epoch": 0.10919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.970122366743016e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.9673034263868533e-05,
+      "loss": 1.0005,
       "step": 3848
     },
     {
-      "epoch": 0.10570981296860839,
+      "epoch": 0.10922247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.970100781810511e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.9672801124664807e-05,
+      "loss": 1.1236,
       "step": 3849
     },
     {
-      "epoch": 0.10573727719645162,
+      "epoch": 0.10925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9700791892021925e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.967256790375449e-05,
+      "loss": 1.1028,
       "step": 3850
     },
     {
-      "epoch": 0.10576474142429486,
+      "epoch": 0.10927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.9700575889182317e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.967233460113956e-05,
+      "loss": 1.0385,
       "step": 3851
     },
     {
-      "epoch": 0.1057922056521381,
+      "epoch": 0.10930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.9700359809587996e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.9672101216821983e-05,
+      "loss": 1.0285,
       "step": 3852
     },
     {
-      "epoch": 0.10581966987998133,
+      "epoch": 0.10933598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9700143653240668e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.9671867750803737e-05,
+      "loss": 1.0535,
       "step": 3853
     },
     {
-      "epoch": 0.10584713410782456,
+      "epoch": 0.10936435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 1.9699927420142045e-05,
-      "loss": 1.0527,
+      "learning_rate": 1.9671634203086784e-05,
+      "loss": 1.1032,
       "step": 3854
     },
     {
-      "epoch": 0.1058745983356678,
+      "epoch": 0.1093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.969971111029384e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.9671400573673104e-05,
+      "loss": 1.0309,
       "step": 3855
     },
     {
-      "epoch": 0.10590206256351102,
+      "epoch": 0.1094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9699494723697768e-05,
-      "loss": 1.1027,
+      "learning_rate": 1.9671166862564668e-05,
+      "loss": 1.0415,
       "step": 3856
     },
     {
-      "epoch": 0.10592952679135426,
+      "epoch": 0.10944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.9699278260355532e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.967093306976345e-05,
+      "loss": 0.9711,
       "step": 3857
     },
     {
-      "epoch": 0.10595699101919749,
+      "epoch": 0.10947786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 1.969906172026885e-05,
-      "loss": 1.1101,
+      "learning_rate": 1.967069919527143e-05,
+      "loss": 0.9654,
       "step": 3858
     },
     {
-      "epoch": 0.10598445524704073,
+      "epoch": 0.10950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.9698845103439438e-05,
-      "loss": 1.0575,
+      "learning_rate": 1.9670465239090576e-05,
+      "loss": 1.0178,
       "step": 3859
     },
     {
-      "epoch": 0.10601191947488396,
+      "epoch": 0.10953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9698628409869002e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.9670231201222866e-05,
+      "loss": 0.9866,
       "step": 3860
     },
     {
-      "epoch": 0.1060393837027272,
+      "epoch": 0.10956299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.969841163955926e-05,
-      "loss": 1.0144,
+      "learning_rate": 1.966999708167028e-05,
+      "loss": 1.0566,
       "step": 3861
     },
     {
-      "epoch": 0.10606684793057043,
+      "epoch": 0.10959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.9698194792511934e-05,
-      "loss": 1.1444,
+      "learning_rate": 1.9669762880434796e-05,
+      "loss": 1.0374,
       "step": 3862
     },
     {
-      "epoch": 0.10609431215841367,
+      "epoch": 0.10961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.969797786872873e-05,
-      "loss": 1.1517,
+      "learning_rate": 1.9669528597518388e-05,
+      "loss": 1.0616,
       "step": 3863
     },
     {
-      "epoch": 0.1061217763862569,
+      "epoch": 0.10964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.969776086821137e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.9669294232923034e-05,
+      "loss": 1.055,
       "step": 3864
     },
     {
-      "epoch": 0.10614924061410014,
+      "epoch": 0.10967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.969754379096157e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.9669059786650722e-05,
+      "loss": 1.0702,
       "step": 3865
     },
     {
-      "epoch": 0.10617670484194337,
+      "epoch": 0.10970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9697326636981044e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.9668825258703424e-05,
+      "loss": 1.0611,
       "step": 3866
     },
     {
-      "epoch": 0.10620416906978661,
+      "epoch": 0.10973325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 1.9697109406271516e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.966859064908313e-05,
+      "loss": 1.1299,
       "step": 3867
     },
     {
-      "epoch": 0.10623163329762984,
+      "epoch": 0.1097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.96968920988347e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.966835595779181e-05,
+      "loss": 1.097,
       "step": 3868
     },
     {
-      "epoch": 0.10625909752547306,
+      "epoch": 0.10979001135073779,
       "grad_norm": 0.0,
-      "learning_rate": 1.969667471467232e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.9668121184831456e-05,
+      "loss": 1.092,
       "step": 3869
     },
     {
-      "epoch": 0.1062865617533163,
+      "epoch": 0.1098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9696457253786093e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.9667886330204045e-05,
+      "loss": 1.0119,
       "step": 3870
     },
     {
-      "epoch": 0.10631402598115954,
+      "epoch": 0.10984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.9696239716177737e-05,
-      "loss": 0.9145,
+      "learning_rate": 1.9667651393911565e-05,
+      "loss": 0.998,
       "step": 3871
     },
     {
-      "epoch": 0.10634149020900277,
+      "epoch": 0.10987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9696022101848982e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.9667416375956e-05,
+      "loss": 1.0754,
       "step": 3872
     },
     {
-      "epoch": 0.106368954436846,
+      "epoch": 0.10990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.969580441080154e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.966718127633933e-05,
+      "loss": 1.1221,
       "step": 3873
     },
     {
-      "epoch": 0.10639641866468924,
+      "epoch": 0.10993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.969558664303714e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.9666946095063553e-05,
+      "loss": 1.0263,
       "step": 3874
     },
     {
-      "epoch": 0.10642388289253248,
+      "epoch": 0.10996027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.96953687985575e-05,
-      "loss": 1.0891,
+      "learning_rate": 1.966671083213064e-05,
+      "loss": 1.0847,
       "step": 3875
     },
     {
-      "epoch": 0.10645134712037571,
+      "epoch": 0.10998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9695150877364347e-05,
-      "loss": 1.1002,
+      "learning_rate": 1.966647548754259e-05,
+      "loss": 1.1119,
       "step": 3876
     },
     {
-      "epoch": 0.10647881134821895,
+      "epoch": 0.11001702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 1.969493287945941e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.9666240061301388e-05,
+      "loss": 1.0319,
       "step": 3877
     },
     {
-      "epoch": 0.10650627557606218,
+      "epoch": 0.11004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9694714804844402e-05,
-      "loss": 1.1756,
+      "learning_rate": 1.966600455340902e-05,
+      "loss": 0.9365,
       "step": 3878
     },
     {
-      "epoch": 0.10653373980390542,
+      "epoch": 0.11007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9694496653521055e-05,
-      "loss": 1.0885,
+      "learning_rate": 1.9665768963867474e-05,
+      "loss": 1.1236,
       "step": 3879
     },
     {
-      "epoch": 0.10656120403174865,
+      "epoch": 0.11010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.96942784254911e-05,
-      "loss": 1.1929,
+      "learning_rate": 1.966553329267875e-05,
+      "loss": 0.9409,
       "step": 3880
     },
     {
-      "epoch": 0.10658866825959189,
+      "epoch": 0.11013053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.969406012075626e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.9665297539844826e-05,
+      "loss": 1.067,
       "step": 3881
     },
     {
-      "epoch": 0.10661613248743512,
+      "epoch": 0.1101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9693841739318254e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.96650617053677e-05,
+      "loss": 1.0491,
       "step": 3882
     },
     {
-      "epoch": 0.10664359671527834,
+      "epoch": 0.11018728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.969362328117882e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.9664825789249366e-05,
+      "loss": 0.9978,
       "step": 3883
     },
     {
-      "epoch": 0.10667106094312158,
+      "epoch": 0.11021566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 1.9693404746339688e-05,
-      "loss": 1.1542,
+      "learning_rate": 1.9664589791491812e-05,
+      "loss": 1.0654,
       "step": 3884
     },
     {
-      "epoch": 0.10669852517096481,
+      "epoch": 0.11024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.9693186134802578e-05,
-      "loss": 1.1101,
+      "learning_rate": 1.9664353712097033e-05,
+      "loss": 0.9473,
       "step": 3885
     },
     {
-      "epoch": 0.10672598939880805,
+      "epoch": 0.11027241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.9692967446569223e-05,
-      "loss": 1.1702,
+      "learning_rate": 1.9664117551067025e-05,
+      "loss": 1.0544,
       "step": 3886
     },
     {
-      "epoch": 0.10675345362665128,
+      "epoch": 0.11030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.969274868164136e-05,
-      "loss": 0.991,
+      "learning_rate": 1.9663881308403786e-05,
+      "loss": 0.9942,
       "step": 3887
     },
     {
-      "epoch": 0.10678091785449452,
+      "epoch": 0.11032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.969252984002071e-05,
-      "loss": 1.0263,
+      "learning_rate": 1.9663644984109302e-05,
+      "loss": 1.1034,
       "step": 3888
     },
     {
-      "epoch": 0.10680838208233775,
+      "epoch": 0.11035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9692310921709014e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.9663408578185575e-05,
+      "loss": 1.0482,
       "step": 3889
     },
     {
-      "epoch": 0.10683584631018099,
+      "epoch": 0.11038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.9692091926707996e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.96631720906346e-05,
+      "loss": 1.0196,
       "step": 3890
     },
     {
-      "epoch": 0.10686331053802423,
+      "epoch": 0.11041430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.9691872855019394e-05,
-      "loss": 1.108,
+      "learning_rate": 1.966293552145838e-05,
+      "loss": 1.0534,
       "step": 3891
     },
     {
-      "epoch": 0.10689077476586746,
+      "epoch": 0.11044267877412031,
       "grad_norm": 0.0,
-      "learning_rate": 1.9691653706644942e-05,
-      "loss": 1.1018,
+      "learning_rate": 1.9662698870658904e-05,
+      "loss": 1.1731,
       "step": 3892
     },
     {
-      "epoch": 0.1069182389937107,
+      "epoch": 0.11047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9691434481586368e-05,
-      "loss": 1.1287,
+      "learning_rate": 1.966246213823818e-05,
+      "loss": 1.0567,
       "step": 3893
     },
     {
-      "epoch": 0.10694570322155393,
+      "epoch": 0.1104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.969121517984541e-05,
-      "loss": 1.0053,
+      "learning_rate": 1.96622253241982e-05,
+      "loss": 1.0325,
       "step": 3894
     },
     {
-      "epoch": 0.10697316744939717,
+      "epoch": 0.11052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.9690995801423806e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.966198842854097e-05,
+      "loss": 1.0061,
       "step": 3895
     },
     {
-      "epoch": 0.10700063167724039,
+      "epoch": 0.11055618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.9690776346323287e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.9661751451268493e-05,
+      "loss": 1.1212,
       "step": 3896
     },
     {
-      "epoch": 0.10702809590508362,
+      "epoch": 0.11058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.9690556814545595e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9661514392382764e-05,
+      "loss": 1.04,
       "step": 3897
     },
     {
-      "epoch": 0.10705556013292686,
+      "epoch": 0.11061293984108968,
       "grad_norm": 0.0,
-      "learning_rate": 1.969033720609246e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.966127725188579e-05,
+      "loss": 1.0472,
       "step": 3898
     },
     {
-      "epoch": 0.1070830243607701,
+      "epoch": 0.11064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.9690117520965625e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.966104002977957e-05,
+      "loss": 1.1032,
       "step": 3899
     },
     {
-      "epoch": 0.10711048858861333,
+      "epoch": 0.11066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.9689897759166826e-05,
-      "loss": 1.1407,
+      "learning_rate": 1.9660802726066114e-05,
+      "loss": 1.1075,
       "step": 3900
     },
     {
-      "epoch": 0.10713795281645656,
+      "epoch": 0.11069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9689677920697803e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.9660565340747423e-05,
+      "loss": 0.9967,
       "step": 3901
     },
     {
-      "epoch": 0.1071654170442998,
+      "epoch": 0.11072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9689458005560297e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.9660327873825503e-05,
+      "loss": 1.0114,
       "step": 3902
     },
     {
-      "epoch": 0.10719288127214303,
+      "epoch": 0.11075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.9689238013756044e-05,
-      "loss": 1.1002,
+      "learning_rate": 1.9660090325302357e-05,
+      "loss": 1.0151,
       "step": 3903
     },
     {
-      "epoch": 0.10722034549998627,
+      "epoch": 0.11078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.9689017945286784e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.9659852695179994e-05,
+      "loss": 1.1564,
       "step": 3904
     },
     {
-      "epoch": 0.1072478097278295,
+      "epoch": 0.11081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9688797800154263e-05,
-      "loss": 1.048,
+      "learning_rate": 1.965961498346042e-05,
+      "loss": 1.0523,
       "step": 3905
     },
     {
-      "epoch": 0.10727527395567274,
+      "epoch": 0.1108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9688577578360222e-05,
-      "loss": 1.07,
+      "learning_rate": 1.9659377190145646e-05,
+      "loss": 1.0898,
       "step": 3906
     },
     {
-      "epoch": 0.10730273818351597,
+      "epoch": 0.1108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9688357279906403e-05,
-      "loss": 1.1135,
+      "learning_rate": 1.9659139315237677e-05,
+      "loss": 1.1128,
       "step": 3907
     },
     {
-      "epoch": 0.10733020241135921,
+      "epoch": 0.11089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.9688136904794547e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.9658901358738526e-05,
+      "loss": 1.0497,
       "step": 3908
     },
     {
-      "epoch": 0.10735766663920245,
+      "epoch": 0.11092508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 1.9687916453026396e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.9658663320650198e-05,
+      "loss": 1.0126,
       "step": 3909
     },
     {
-      "epoch": 0.10738513086704567,
+      "epoch": 0.11095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.96876959246037e-05,
-      "loss": 1.1256,
+      "learning_rate": 1.9658425200974708e-05,
+      "loss": 1.0396,
       "step": 3910
     },
     {
-      "epoch": 0.1074125950948889,
+      "epoch": 0.11098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9687475319528203e-05,
-      "loss": 0.993,
+      "learning_rate": 1.9658186999714067e-05,
+      "loss": 0.9486,
       "step": 3911
     },
     {
-      "epoch": 0.10744005932273214,
+      "epoch": 0.11101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.968725463780164e-05,
-      "loss": 1.1848,
+      "learning_rate": 1.9657948716870287e-05,
+      "loss": 0.9679,
       "step": 3912
     },
     {
-      "epoch": 0.10746752355057537,
+      "epoch": 0.11103859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.9687033879425778e-05,
-      "loss": 1.1044,
+      "learning_rate": 1.965771035244538e-05,
+      "loss": 1.0301,
       "step": 3913
     },
     {
-      "epoch": 0.10749498777841861,
+      "epoch": 0.11106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.9686813044402343e-05,
-      "loss": 1.0898,
+      "learning_rate": 1.9657471906441358e-05,
+      "loss": 1.129,
       "step": 3914
     },
     {
-      "epoch": 0.10752245200626184,
+      "epoch": 0.11109534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.9686592132733096e-05,
-      "loss": 1.0481,
+      "learning_rate": 1.9657233378860235e-05,
+      "loss": 1.1477,
       "step": 3915
     },
     {
-      "epoch": 0.10754991623410508,
+      "epoch": 0.11112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9686371144419777e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.9656994769704032e-05,
+      "loss": 1.1165,
       "step": 3916
     },
     {
-      "epoch": 0.10757738046194831,
+      "epoch": 0.11115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.9686150079464136e-05,
-      "loss": 1.0419,
+      "learning_rate": 1.9656756078974758e-05,
+      "loss": 0.9768,
       "step": 3917
     },
     {
-      "epoch": 0.10760484468979155,
+      "epoch": 0.11118047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 1.9685928937867928e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.965651730667443e-05,
+      "loss": 1.0361,
       "step": 3918
     },
     {
-      "epoch": 0.10763230891763478,
+      "epoch": 0.11120885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 1.9685707719632896e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.9656278452805067e-05,
+      "loss": 1.0682,
       "step": 3919
     },
     {
-      "epoch": 0.10765977314547802,
+      "epoch": 0.1112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9685486424760794e-05,
-      "loss": 1.0818,
+      "learning_rate": 1.9656039517368686e-05,
+      "loss": 1.0405,
       "step": 3920
     },
     {
-      "epoch": 0.10768723737332125,
+      "epoch": 0.1112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.9685265053253368e-05,
-      "loss": 1.0555,
+      "learning_rate": 1.9655800500367304e-05,
+      "loss": 1.0437,
       "step": 3921
     },
     {
-      "epoch": 0.10771470160116449,
+      "epoch": 0.11129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.9685043605112375e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.9655561401802945e-05,
+      "loss": 1.0963,
       "step": 3922
     },
     {
-      "epoch": 0.10774216582900771,
+      "epoch": 0.11132236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 1.9684822080339566e-05,
-      "loss": 1.0025,
+      "learning_rate": 1.965532222167762e-05,
+      "loss": 1.0634,
       "step": 3923
     },
     {
-      "epoch": 0.10776963005685095,
+      "epoch": 0.11135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.9684600478936692e-05,
-      "loss": 1.0666,
+      "learning_rate": 1.9655082959993358e-05,
+      "loss": 0.9977,
       "step": 3924
     },
     {
-      "epoch": 0.10779709428469418,
+      "epoch": 0.11137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.968437880090551e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.965484361675217e-05,
+      "loss": 1.1841,
       "step": 3925
     },
     {
-      "epoch": 0.10782455851253742,
+      "epoch": 0.11140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.9684157046247768e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.965460419195609e-05,
+      "loss": 1.1149,
       "step": 3926
     },
     {
-      "epoch": 0.10785202274038065,
+      "epoch": 0.11143586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.9683935214965228e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.9654364685607134e-05,
+      "loss": 1.0471,
       "step": 3927
     },
     {
-      "epoch": 0.10787948696822389,
+      "epoch": 0.11146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.968371330705964e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.9654125097707324e-05,
+      "loss": 1.13,
       "step": 3928
     },
     {
-      "epoch": 0.10790695119606712,
+      "epoch": 0.11149262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 1.968349132253276e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.9653885428258686e-05,
+      "loss": 0.9367,
       "step": 3929
     },
     {
-      "epoch": 0.10793441542391036,
+      "epoch": 0.11152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9683269261386345e-05,
-      "loss": 1.1566,
+      "learning_rate": 1.9653645677263246e-05,
+      "loss": 1.0815,
       "step": 3930
     },
     {
-      "epoch": 0.10796187965175359,
+      "epoch": 0.11154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.9683047123622154e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.965340584472302e-05,
+      "loss": 1.1346,
       "step": 3931
     },
     {
-      "epoch": 0.10798934387959683,
+      "epoch": 0.111577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9682824909241943e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.9653165930640047e-05,
+      "loss": 1.0823,
       "step": 3932
     },
     {
-      "epoch": 0.10801680810744006,
+      "epoch": 0.1116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.968260261824747e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9652925935016344e-05,
+      "loss": 1.0073,
       "step": 3933
     },
     {
-      "epoch": 0.1080442723352833,
+      "epoch": 0.11163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.9682380250640497e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.9652685857853947e-05,
+      "loss": 1.0493,
       "step": 3934
     },
     {
-      "epoch": 0.10807173656312653,
+      "epoch": 0.11166288308740067,
       "grad_norm": 0.0,
-      "learning_rate": 1.9682157806422778e-05,
-      "loss": 0.9084,
+      "learning_rate": 1.9652445699154874e-05,
+      "loss": 1.0998,
       "step": 3935
     },
     {
-      "epoch": 0.10809920079096977,
+      "epoch": 0.11169125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 1.9681935285596075e-05,
-      "loss": 1.1388,
+      "learning_rate": 1.965220545892116e-05,
+      "loss": 1.0231,
       "step": 3936
     },
     {
-      "epoch": 0.10812666501881299,
+      "epoch": 0.11171963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.9681712688162155e-05,
-      "loss": 1.0451,
+      "learning_rate": 1.9651965137154833e-05,
+      "loss": 0.953,
       "step": 3937
     },
     {
-      "epoch": 0.10815412924665622,
+      "epoch": 0.11174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.968149001412277e-05,
-      "loss": 1.1379,
+      "learning_rate": 1.965172473385792e-05,
+      "loss": 1.0634,
       "step": 3938
     },
     {
-      "epoch": 0.10818159347449946,
+      "epoch": 0.11177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9681267263479685e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.9651484249032456e-05,
+      "loss": 1.0216,
       "step": 3939
     },
     {
-      "epoch": 0.1082090577023427,
+      "epoch": 0.11180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9681044436234664e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.965124368268047e-05,
+      "loss": 0.9788,
       "step": 3940
     },
     {
-      "epoch": 0.10823652193018593,
+      "epoch": 0.11183314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 1.9680821532389468e-05,
-      "loss": 1.1284,
+      "learning_rate": 1.9651003034803998e-05,
+      "loss": 1.0262,
       "step": 3941
     },
     {
-      "epoch": 0.10826398615802917,
+      "epoch": 0.11186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.9680598551945866e-05,
-      "loss": 1.1155,
+      "learning_rate": 1.9650762305405065e-05,
+      "loss": 1.0673,
       "step": 3942
     },
     {
-      "epoch": 0.1082914503858724,
+      "epoch": 0.11188989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 1.9680375494905617e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.965052149448571e-05,
+      "loss": 0.9691,
       "step": 3943
     },
     {
-      "epoch": 0.10831891461371564,
+      "epoch": 0.11191827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9680152361270486e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.9650280602047965e-05,
+      "loss": 0.9687,
       "step": 3944
     },
     {
-      "epoch": 0.10834637884155887,
+      "epoch": 0.1119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679929151042238e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.965003962809387e-05,
+      "loss": 1.0421,
       "step": 3945
     },
     {
-      "epoch": 0.1083738430694021,
+      "epoch": 0.11197502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679705864222645e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.9649798572625457e-05,
+      "loss": 1.1017,
       "step": 3946
     },
     {
-      "epoch": 0.10840130729724534,
+      "epoch": 0.11200340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679482500813466e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.9649557435644757e-05,
+      "loss": 1.0643,
       "step": 3947
     },
     {
-      "epoch": 0.10842877152508858,
+      "epoch": 0.11203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679259060816475e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.9649316217153812e-05,
+      "loss": 1.0882,
       "step": 3948
     },
     {
-      "epoch": 0.10845623575293181,
+      "epoch": 0.11206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679035544233436e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.964907491715466e-05,
+      "loss": 0.9493,
       "step": 3949
     },
     {
-      "epoch": 0.10848369998077505,
+      "epoch": 0.11208853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.967881195106612e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.9648833535649342e-05,
+      "loss": 0.9595,
       "step": 3950
     },
     {
-      "epoch": 0.10851116420861827,
+      "epoch": 0.11211691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.9678588281316293e-05,
-      "loss": 1.108,
+      "learning_rate": 1.964859207263989e-05,
+      "loss": 1.0338,
       "step": 3951
     },
     {
-      "epoch": 0.1085386284364615,
+      "epoch": 0.11214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9678364534985724e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.9648350528128346e-05,
+      "loss": 1.0324,
       "step": 3952
     },
     {
-      "epoch": 0.10856609266430474,
+      "epoch": 0.11217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9678140712076187e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.9648108902116754e-05,
+      "loss": 1.0728,
       "step": 3953
     },
     {
-      "epoch": 0.10859355689214797,
+      "epoch": 0.11220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.9677916812589454e-05,
-      "loss": 1.0047,
+      "learning_rate": 1.964786719460715e-05,
+      "loss": 1.0371,
       "step": 3954
     },
     {
-      "epoch": 0.10862102111999121,
+      "epoch": 0.11223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.9677692836527294e-05,
-      "loss": 1.137,
+      "learning_rate": 1.964762540560158e-05,
+      "loss": 1.0735,
       "step": 3955
     },
     {
-      "epoch": 0.10864848534783444,
+      "epoch": 0.11225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.9677468783891478e-05,
-      "loss": 1.0413,
+      "learning_rate": 1.9647383535102085e-05,
+      "loss": 1.0755,
       "step": 3956
     },
     {
-      "epoch": 0.10867594957567768,
+      "epoch": 0.1122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.967724465468378e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.9647141583110704e-05,
+      "loss": 0.9772,
       "step": 3957
     },
     {
-      "epoch": 0.10870341380352092,
+      "epoch": 0.11231555051078321,
       "grad_norm": 0.0,
-      "learning_rate": 1.9677020448905974e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.964689954962949e-05,
+      "loss": 0.9538,
       "step": 3958
     },
     {
-      "epoch": 0.10873087803136415,
+      "epoch": 0.1123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.9676796166559834e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.9646657434660478e-05,
+      "loss": 1.0361,
       "step": 3959
     },
     {
-      "epoch": 0.10875834225920739,
+      "epoch": 0.11237230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 1.9676571807647136e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.964641523820572e-05,
+      "loss": 0.989,
       "step": 3960
     },
     {
-      "epoch": 0.10878580648705062,
+      "epoch": 0.11240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.9676347372169653e-05,
-      "loss": 1.0462,
+      "learning_rate": 1.9646172960267254e-05,
+      "loss": 0.9778,
       "step": 3961
     },
     {
-      "epoch": 0.10881327071489386,
+      "epoch": 0.11242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9676122860129157e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.9645930600847134e-05,
+      "loss": 1.1116,
       "step": 3962
     },
     {
-      "epoch": 0.10884073494273709,
+      "epoch": 0.11245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9675898271527433e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.96456881599474e-05,
+      "loss": 1.0504,
       "step": 3963
     },
     {
-      "epoch": 0.10886819917058031,
+      "epoch": 0.11248581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.9675673606366253e-05,
-      "loss": 1.112,
+      "learning_rate": 1.9645445637570112e-05,
+      "loss": 1.0173,
       "step": 3964
     },
     {
-      "epoch": 0.10889566339842355,
+      "epoch": 0.11251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.9675448864647397e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.964520303371731e-05,
+      "loss": 1.1243,
       "step": 3965
     },
     {
-      "epoch": 0.10892312762626678,
+      "epoch": 0.11254256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.9675224046372638e-05,
-      "loss": 1.1265,
+      "learning_rate": 1.964496034839104e-05,
+      "loss": 1.0309,
       "step": 3966
     },
     {
-      "epoch": 0.10895059185411002,
+      "epoch": 0.11257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.967499915154376e-05,
-      "loss": 1.0581,
+      "learning_rate": 1.964471758159336e-05,
+      "loss": 1.0316,
       "step": 3967
     },
     {
-      "epoch": 0.10897805608195325,
+      "epoch": 0.11259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.9674774180162545e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.9644474733326316e-05,
+      "loss": 0.9843,
       "step": 3968
     },
     {
-      "epoch": 0.10900552030979649,
+      "epoch": 0.11262769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.9674549132230764e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.9644231803591964e-05,
+      "loss": 1.0639,
       "step": 3969
     },
     {
-      "epoch": 0.10903298453763972,
+      "epoch": 0.11265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.9674324007750204e-05,
-      "loss": 1.1026,
+      "learning_rate": 1.9643988792392347e-05,
+      "loss": 1.0083,
       "step": 3970
     },
     {
-      "epoch": 0.10906044876548296,
+      "epoch": 0.1126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9674098806722646e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.9643745699729523e-05,
+      "loss": 1.0775,
       "step": 3971
     },
     {
-      "epoch": 0.1090879129933262,
+      "epoch": 0.11271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.967387352914987e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.9643502525605548e-05,
+      "loss": 1.0143,
       "step": 3972
     },
     {
-      "epoch": 0.10911537722116943,
+      "epoch": 0.11274120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 1.967364817503366e-05,
-      "loss": 1.0697,
+      "learning_rate": 1.9643259270022472e-05,
+      "loss": 1.0037,
       "step": 3973
     },
     {
-      "epoch": 0.10914284144901266,
+      "epoch": 0.11276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.9673422744375794e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.9643015932982355e-05,
+      "loss": 1.1753,
       "step": 3974
     },
     {
-      "epoch": 0.1091703056768559,
+      "epoch": 0.11279795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 1.9673197237178067e-05,
-      "loss": 1.0771,
+      "learning_rate": 1.9642772514487244e-05,
+      "loss": 1.1593,
       "step": 3975
     },
     {
-      "epoch": 0.10919776990469914,
+      "epoch": 0.11282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9672971653442253e-05,
-      "loss": 1.0788,
+      "learning_rate": 1.96425290145392e-05,
+      "loss": 1.0141,
       "step": 3976
     },
     {
-      "epoch": 0.10922523413254237,
+      "epoch": 0.11285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.9672745993170138e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.9642285433140285e-05,
+      "loss": 0.9432,
       "step": 3977
     },
     {
-      "epoch": 0.10925269836038559,
+      "epoch": 0.11288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9672520256363513e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.964204177029255e-05,
+      "loss": 1.0115,
       "step": 3978
     },
     {
-      "epoch": 0.10928016258822883,
+      "epoch": 0.11291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.967229444302416e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.9641798025998057e-05,
+      "loss": 1.08,
       "step": 3979
     },
     {
-      "epoch": 0.10930762681607206,
+      "epoch": 0.11293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.967206855315387e-05,
-      "loss": 1.0498,
+      "learning_rate": 1.9641554200258856e-05,
+      "loss": 1.0492,
       "step": 3980
     },
     {
-      "epoch": 0.1093350910439153,
+      "epoch": 0.11296821793416573,
       "grad_norm": 0.0,
-      "learning_rate": 1.9671842586754422e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.9641310293077017e-05,
+      "loss": 0.9889,
       "step": 3981
     },
     {
-      "epoch": 0.10936255527175853,
+      "epoch": 0.11299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.967161654382761e-05,
-      "loss": 1.074,
+      "learning_rate": 1.9641066304454596e-05,
+      "loss": 0.8692,
       "step": 3982
     },
     {
-      "epoch": 0.10939001949960177,
+      "epoch": 0.1130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9671390424375225e-05,
-      "loss": 1.119,
+      "learning_rate": 1.9640822234393654e-05,
+      "loss": 0.9804,
       "step": 3983
     },
     {
-      "epoch": 0.109417483727445,
+      "epoch": 0.1130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.967116422839905e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.9640578082896254e-05,
+      "loss": 0.9531,
       "step": 3984
     },
     {
-      "epoch": 0.10944494795528824,
+      "epoch": 0.11308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.967093795590088e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.9640333849964456e-05,
+      "loss": 0.9972,
       "step": 3985
     },
     {
-      "epoch": 0.10947241218313147,
+      "epoch": 0.11311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.9670711606882496e-05,
-      "loss": 1.115,
+      "learning_rate": 1.9640089535600327e-05,
+      "loss": 1.0086,
       "step": 3986
     },
     {
-      "epoch": 0.10949987641097471,
+      "epoch": 0.11313847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 1.96704851813457e-05,
-      "loss": 1.0476,
+      "learning_rate": 1.963984513980593e-05,
+      "loss": 1.0038,
       "step": 3987
     },
     {
-      "epoch": 0.10952734063881794,
+      "epoch": 0.11316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.967025867929228e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.9639600662583325e-05,
+      "loss": 1.1813,
       "step": 3988
     },
     {
-      "epoch": 0.10955480486666118,
+      "epoch": 0.11319523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.967003210072403e-05,
-      "loss": 1.0824,
+      "learning_rate": 1.9639356103934578e-05,
+      "loss": 1.0617,
       "step": 3989
     },
     {
-      "epoch": 0.10958226909450441,
+      "epoch": 0.11322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9669805445642732e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.963911146386176e-05,
+      "loss": 1.0091,
       "step": 3990
     },
     {
-      "epoch": 0.10960973332234764,
+      "epoch": 0.11325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.9669578714050196e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.9638866742366932e-05,
+      "loss": 1.0696,
       "step": 3991
     },
     {
-      "epoch": 0.10963719755019087,
+      "epoch": 0.11328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9669351905948203e-05,
-      "loss": 1.022,
+      "learning_rate": 1.9638621939452166e-05,
+      "loss": 1.0374,
       "step": 3992
     },
     {
-      "epoch": 0.1096646617780341,
+      "epoch": 0.11330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.9669125021338552e-05,
-      "loss": 1.175,
+      "learning_rate": 1.9638377055119522e-05,
+      "loss": 1.0124,
       "step": 3993
     },
     {
-      "epoch": 0.10969212600587734,
+      "epoch": 0.11333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.966889806022304e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.9638132089371078e-05,
+      "loss": 1.0608,
       "step": 3994
     },
     {
-      "epoch": 0.10971959023372058,
+      "epoch": 0.1133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.966867102260346e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.96378870422089e-05,
+      "loss": 0.9642,
       "step": 3995
     },
     {
-      "epoch": 0.10974705446156381,
+      "epoch": 0.1133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.966844390848161e-05,
-      "loss": 1.0955,
+      "learning_rate": 1.9637641913635055e-05,
+      "loss": 1.0489,
       "step": 3996
     },
     {
-      "epoch": 0.10977451868940705,
+      "epoch": 0.11342224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 1.9668216717859287e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.9637396703651616e-05,
+      "loss": 0.9766,
       "step": 3997
     },
     {
-      "epoch": 0.10980198291725028,
+      "epoch": 0.1134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.966798945073829e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.9637151412260654e-05,
+      "loss": 1.0166,
       "step": 3998
     },
     {
-      "epoch": 0.10982944714509352,
+      "epoch": 0.11347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9667762107120412e-05,
-      "loss": 1.1376,
+      "learning_rate": 1.963690603946424e-05,
+      "loss": 0.9995,
       "step": 3999
     },
     {
-      "epoch": 0.10985691137293675,
+      "epoch": 0.11350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.966753468700746e-05,
-      "loss": 1.1115,
+      "learning_rate": 1.9636660585264447e-05,
+      "loss": 1.1224,
       "step": 4000
     },
     {
-      "epoch": 0.10988437560077999,
+      "epoch": 0.11353575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 1.9667307190401222e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.963641504966335e-05,
+      "loss": 0.9715,
       "step": 4001
     },
     {
-      "epoch": 0.10991183982862322,
+      "epoch": 0.11356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.966707961730351e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.9636169432663022e-05,
+      "loss": 1.0498,
       "step": 4002
     },
     {
-      "epoch": 0.10993930405646646,
+      "epoch": 0.11359250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.9666851967716117e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.9635923734265536e-05,
+      "loss": 0.9585,
       "step": 4003
     },
     {
-      "epoch": 0.1099667682843097,
+      "epoch": 0.11362088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.966662424164085e-05,
-      "loss": 1.1468,
+      "learning_rate": 1.963567795447297e-05,
+      "loss": 0.9693,
       "step": 4004
     },
     {
-      "epoch": 0.10999423251215291,
+      "epoch": 0.11364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.9666396439079502e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.96354320932874e-05,
+      "loss": 1.1066,
       "step": 4005
     },
     {
-      "epoch": 0.11002169673999615,
+      "epoch": 0.11367763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.9666168560033886e-05,
-      "loss": 0.985,
+      "learning_rate": 1.9635186150710898e-05,
+      "loss": 0.9887,
       "step": 4006
     },
     {
-      "epoch": 0.11004916096783939,
+      "epoch": 0.11370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.9665940604505795e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.9634940126745548e-05,
+      "loss": 1.0702,
       "step": 4007
     },
     {
-      "epoch": 0.11007662519568262,
+      "epoch": 0.11373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.9665712572497043e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.9634694021393427e-05,
+      "loss": 1.0962,
       "step": 4008
     },
     {
-      "epoch": 0.11010408942352586,
+      "epoch": 0.1137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9665484464009425e-05,
-      "loss": 1.1363,
+      "learning_rate": 1.9634447834656615e-05,
+      "loss": 0.9379,
       "step": 4009
     },
     {
-      "epoch": 0.11013155365136909,
+      "epoch": 0.1137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.966525627904475e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.9634201566537182e-05,
+      "loss": 0.9187,
       "step": 4010
     },
     {
-      "epoch": 0.11015901787921233,
+      "epoch": 0.11381952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 1.9665028017604826e-05,
-      "loss": 1.1528,
+      "learning_rate": 1.963395521703722e-05,
+      "loss": 0.98,
       "step": 4011
     },
     {
-      "epoch": 0.11018648210705556,
+      "epoch": 0.11384790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 1.9664799679691454e-05,
-      "loss": 1.1782,
+      "learning_rate": 1.9633708786158803e-05,
+      "loss": 1.0034,
       "step": 4012
     },
     {
-      "epoch": 0.1102139463348988,
+      "epoch": 0.11387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9664571265306446e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.9633462273904018e-05,
+      "loss": 1.0912,
       "step": 4013
     },
     {
-      "epoch": 0.11024141056274203,
+      "epoch": 0.11390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.9664342774451604e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9633215680274943e-05,
+      "loss": 1.1396,
       "step": 4014
     },
     {
-      "epoch": 0.11026887479058527,
+      "epoch": 0.11393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.966411420712874e-05,
-      "loss": 1.1437,
+      "learning_rate": 1.963296900527366e-05,
+      "loss": 0.9824,
       "step": 4015
     },
     {
-      "epoch": 0.1102963390184285,
+      "epoch": 0.11396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.966388556333966e-05,
-      "loss": 1.0047,
+      "learning_rate": 1.9632722248902257e-05,
+      "loss": 1.0099,
       "step": 4016
     },
     {
-      "epoch": 0.11032380324627174,
+      "epoch": 0.11398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.9663656843086176e-05,
-      "loss": 1.035,
+      "learning_rate": 1.9632475411162814e-05,
+      "loss": 0.9951,
       "step": 4017
     },
     {
-      "epoch": 0.11035126747411496,
+      "epoch": 0.11401816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9663428046370092e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.963222849205742e-05,
+      "loss": 1.1091,
       "step": 4018
     },
     {
-      "epoch": 0.1103787317019582,
+      "epoch": 0.11404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.9663199173193222e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.963198149158816e-05,
+      "loss": 1.0598,
       "step": 4019
     },
     {
-      "epoch": 0.11040619592980143,
+      "epoch": 0.11407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.966297022355738e-05,
-      "loss": 1.123,
+      "learning_rate": 1.9631734409757115e-05,
+      "loss": 1.0353,
       "step": 4020
     },
     {
-      "epoch": 0.11043366015764466,
+      "epoch": 0.11410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.9662741197464376e-05,
-      "loss": 1.0524,
+      "learning_rate": 1.963148724656638e-05,
+      "loss": 1.0403,
       "step": 4021
     },
     {
-      "epoch": 0.1104611243854879,
+      "epoch": 0.1141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9662512094916018e-05,
-      "loss": 1.0561,
+      "learning_rate": 1.9631240002018038e-05,
+      "loss": 1.0114,
       "step": 4022
     },
     {
-      "epoch": 0.11048858861333113,
+      "epoch": 0.11416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.966228291591412e-05,
-      "loss": 1.0936,
+      "learning_rate": 1.963099267611418e-05,
+      "loss": 1.1235,
       "step": 4023
     },
     {
-      "epoch": 0.11051605284117437,
+      "epoch": 0.11418842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 1.9662053660460497e-05,
-      "loss": 1.1189,
+      "learning_rate": 1.9630745268856892e-05,
+      "loss": 1.0745,
       "step": 4024
     },
     {
-      "epoch": 0.1105435170690176,
+      "epoch": 0.11421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9661824328556966e-05,
-      "loss": 1.0713,
+      "learning_rate": 1.9630497780248268e-05,
+      "loss": 0.9385,
       "step": 4025
     },
     {
-      "epoch": 0.11057098129686084,
+      "epoch": 0.11424517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.9661594920205337e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.9630250210290397e-05,
+      "loss": 1.0494,
       "step": 4026
     },
     {
-      "epoch": 0.11059844552470408,
+      "epoch": 0.11427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9661365435407426e-05,
-      "loss": 1.135,
+      "learning_rate": 1.9630002558985366e-05,
+      "loss": 0.9693,
       "step": 4027
     },
     {
-      "epoch": 0.11062590975254731,
+      "epoch": 0.11430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.966113587416505e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.962975482633527e-05,
+      "loss": 1.023,
       "step": 4028
     },
     {
-      "epoch": 0.11065337398039055,
+      "epoch": 0.11433030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.966090623648002e-05,
-      "loss": 1.081,
+      "learning_rate": 1.9629507012342207e-05,
+      "loss": 1.0156,
       "step": 4029
     },
     {
-      "epoch": 0.11068083820823378,
+      "epoch": 0.11435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.966067652235416e-05,
-      "loss": 1.0923,
+      "learning_rate": 1.9629259117008264e-05,
+      "loss": 1.0849,
       "step": 4030
     },
     {
-      "epoch": 0.11070830243607702,
+      "epoch": 0.11438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9660446731789288e-05,
-      "loss": 1.059,
+      "learning_rate": 1.9629011140335536e-05,
+      "loss": 1.103,
       "step": 4031
     },
     {
-      "epoch": 0.11073576666392024,
+      "epoch": 0.11441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9660216864787218e-05,
-      "loss": 1.049,
+      "learning_rate": 1.9628763082326113e-05,
+      "loss": 1.0548,
       "step": 4032
     },
     {
-      "epoch": 0.11076323089176347,
+      "epoch": 0.11444381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 1.9659986921349768e-05,
-      "loss": 1.0037,
+      "learning_rate": 1.96285149429821e-05,
+      "loss": 1.0284,
       "step": 4033
     },
     {
-      "epoch": 0.11079069511960671,
+      "epoch": 0.114472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9659756901478762e-05,
-      "loss": 1.0869,
+      "learning_rate": 1.9628266722305587e-05,
+      "loss": 0.9993,
       "step": 4034
     },
     {
-      "epoch": 0.11081815934744994,
+      "epoch": 0.11450056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 1.9659526805176016e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.9628018420298675e-05,
+      "loss": 1.0354,
       "step": 4035
     },
     {
-      "epoch": 0.11084562357529318,
+      "epoch": 0.11452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9659296632443353e-05,
-      "loss": 1.1505,
+      "learning_rate": 1.9627770036963457e-05,
+      "loss": 0.9359,
       "step": 4036
     },
     {
-      "epoch": 0.11087308780313641,
+      "epoch": 0.11455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.965906638328259e-05,
-      "loss": 1.0423,
+      "learning_rate": 1.9627521572302035e-05,
+      "loss": 0.9583,
       "step": 4037
     },
     {
-      "epoch": 0.11090055203097965,
+      "epoch": 0.11458569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.965883605769556e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.96272730263165e-05,
+      "loss": 0.9571,
       "step": 4038
     },
     {
-      "epoch": 0.11092801625882288,
+      "epoch": 0.11461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.9658605655684073e-05,
-      "loss": 1.1527,
+      "learning_rate": 1.962702439900896e-05,
+      "loss": 0.8769,
       "step": 4039
     },
     {
-      "epoch": 0.11095548048666612,
+      "epoch": 0.11464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.9658375177249955e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.9626775690381513e-05,
+      "loss": 1.0424,
       "step": 4040
     },
     {
-      "epoch": 0.11098294471450935,
+      "epoch": 0.11467082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 1.9658144622395033e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.9626526900436257e-05,
+      "loss": 1.0158,
       "step": 4041
     },
     {
-      "epoch": 0.11101040894235259,
+      "epoch": 0.11469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.965791399112113e-05,
-      "loss": 1.101,
+      "learning_rate": 1.9626278029175296e-05,
+      "loss": 0.9816,
       "step": 4042
     },
     {
-      "epoch": 0.11103787317019582,
+      "epoch": 0.11472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.9657683283430072e-05,
-      "loss": 1.0714,
+      "learning_rate": 1.9626029076600734e-05,
+      "loss": 1.0655,
       "step": 4043
     },
     {
-      "epoch": 0.11106533739803906,
+      "epoch": 0.11475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.965745249932368e-05,
-      "loss": 1.184,
+      "learning_rate": 1.9625780042714672e-05,
+      "loss": 1.082,
       "step": 4044
     },
     {
-      "epoch": 0.11109280162588228,
+      "epoch": 0.11478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.965722163880378e-05,
-      "loss": 1.2,
+      "learning_rate": 1.9625530927519206e-05,
+      "loss": 1.1532,
       "step": 4045
     },
     {
-      "epoch": 0.11112026585372552,
+      "epoch": 0.11481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.965699070187221e-05,
-      "loss": 1.1837,
+      "learning_rate": 1.9625281731016455e-05,
+      "loss": 0.8811,
       "step": 4046
     },
     {
-      "epoch": 0.11114773008156875,
+      "epoch": 0.11484108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 1.9656759688530785e-05,
-      "loss": 1.0495,
+      "learning_rate": 1.9625032453208516e-05,
+      "loss": 0.9869,
       "step": 4047
     },
     {
-      "epoch": 0.11117519430941199,
+      "epoch": 0.1148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9656528598781333e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.962478309409749e-05,
+      "loss": 1.0527,
       "step": 4048
     },
     {
-      "epoch": 0.11120265853725522,
+      "epoch": 0.11489784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 1.965629743262569e-05,
-      "loss": 1.2052,
+      "learning_rate": 1.9624533653685495e-05,
+      "loss": 0.9981,
       "step": 4049
     },
     {
-      "epoch": 0.11123012276509846,
+      "epoch": 0.11492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9656066190065682e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.9624284131974625e-05,
+      "loss": 1.1624,
       "step": 4050
     },
     {
-      "epoch": 0.11125758699294169,
+      "epoch": 0.11495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.9655834871103135e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.9624034528967e-05,
+      "loss": 1.0918,
       "step": 4051
     },
     {
-      "epoch": 0.11128505122078493,
+      "epoch": 0.11498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.9655603475739884e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.962378484466472e-05,
+      "loss": 1.0639,
       "step": 4052
     },
     {
-      "epoch": 0.11131251544862816,
+      "epoch": 0.11501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.9655372003977756e-05,
-      "loss": 0.964,
+      "learning_rate": 1.9623535079069897e-05,
+      "loss": 1.0518,
       "step": 4053
     },
     {
-      "epoch": 0.1113399796764714,
+      "epoch": 0.11503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9655140455818587e-05,
-      "loss": 1.033,
+      "learning_rate": 1.962328523218464e-05,
+      "loss": 0.9777,
       "step": 4054
     },
     {
-      "epoch": 0.11136744390431463,
+      "epoch": 0.11506810442678773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9654908831264205e-05,
-      "loss": 1.1672,
+      "learning_rate": 1.9623035304011062e-05,
+      "loss": 1.0412,
       "step": 4055
     },
     {
-      "epoch": 0.11139490813215787,
+      "epoch": 0.11509648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.9654677130316443e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.9622785294551273e-05,
+      "loss": 1.0355,
       "step": 4056
     },
     {
-      "epoch": 0.1114223723600011,
+      "epoch": 0.11512485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.965444535297714e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.9622535203807382e-05,
+      "loss": 0.9867,
       "step": 4057
     },
     {
-      "epoch": 0.11144983658784434,
+      "epoch": 0.11515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.9654213499248125e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.9622285031781505e-05,
+      "loss": 1.1063,
       "step": 4058
     },
     {
-      "epoch": 0.11147730081568756,
+      "epoch": 0.11518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9653981569131232e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.9622034778475753e-05,
+      "loss": 0.9658,
       "step": 4059
     },
     {
-      "epoch": 0.1115047650435308,
+      "epoch": 0.1152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9653749562628295e-05,
-      "loss": 1.1007,
+      "learning_rate": 1.962178444389224e-05,
+      "loss": 1.01,
       "step": 4060
     },
     {
-      "epoch": 0.11153222927137403,
+      "epoch": 0.1152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.9653517479741152e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.962153402803308e-05,
+      "loss": 1.1501,
       "step": 4061
     },
     {
-      "epoch": 0.11155969349921727,
+      "epoch": 0.11526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.9653285320471642e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.962128353090039e-05,
+      "loss": 1.0704,
       "step": 4062
     },
     {
-      "epoch": 0.1115871577270605,
+      "epoch": 0.11529511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 1.96530530848216e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.962103295249629e-05,
+      "loss": 1.2034,
       "step": 4063
     },
     {
-      "epoch": 0.11161462195490374,
+      "epoch": 0.11532349602724178,
       "grad_norm": 0.0,
-      "learning_rate": 1.965282077279286e-05,
-      "loss": 1.131,
+      "learning_rate": 1.962078229282289e-05,
+      "loss": 1.1685,
       "step": 4064
     },
     {
-      "epoch": 0.11164208618274697,
+      "epoch": 0.11535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.965258838438726e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.9620531551882305e-05,
+      "loss": 0.8994,
       "step": 4065
     },
     {
-      "epoch": 0.11166955041059021,
+      "epoch": 0.11538024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.965235591960665e-05,
-      "loss": 1.077,
+      "learning_rate": 1.9620280729676662e-05,
+      "loss": 1.0763,
       "step": 4066
     },
     {
-      "epoch": 0.11169701463843344,
+      "epoch": 0.11540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.9652123378452854e-05,
-      "loss": 0.9802,
+      "learning_rate": 1.9620029826208074e-05,
+      "loss": 0.9896,
       "step": 4067
     },
     {
-      "epoch": 0.11172447886627668,
+      "epoch": 0.11543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.965189076092772e-05,
-      "loss": 1.0647,
+      "learning_rate": 1.9619778841478662e-05,
+      "loss": 1.03,
       "step": 4068
     },
     {
-      "epoch": 0.11175194309411991,
+      "epoch": 0.11546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.9651658067033092e-05,
-      "loss": 0.9992,
+      "learning_rate": 1.9619527775490543e-05,
+      "loss": 1.1281,
       "step": 4069
     },
     {
-      "epoch": 0.11177940732196315,
+      "epoch": 0.11549375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 1.96514252967708e-05,
-      "loss": 1.0568,
+      "learning_rate": 1.9619276628245846e-05,
+      "loss": 1.0923,
       "step": 4070
     },
     {
-      "epoch": 0.11180687154980638,
+      "epoch": 0.11552213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 1.9651192450142697e-05,
-      "loss": 0.8107,
+      "learning_rate": 1.961902539974668e-05,
+      "loss": 1.2062,
       "step": 4071
     },
     {
-      "epoch": 0.11183433577764962,
+      "epoch": 0.1155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.965095952715062e-05,
-      "loss": 1.13,
+      "learning_rate": 1.961877408999518e-05,
+      "loss": 1.0784,
       "step": 4072
     },
     {
-      "epoch": 0.11186180000549284,
+      "epoch": 0.1155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9650726527796413e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.961852269899346e-05,
+      "loss": 1.0352,
       "step": 4073
     },
     {
-      "epoch": 0.11188926423333607,
+      "epoch": 0.11560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.965049345208192e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.9618271226743644e-05,
+      "loss": 1.0471,
       "step": 4074
     },
     {
-      "epoch": 0.11191672846117931,
+      "epoch": 0.11563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.9650260300008985e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.9618019673247864e-05,
+      "loss": 1.0102,
       "step": 4075
     },
     {
-      "epoch": 0.11194419268902255,
+      "epoch": 0.11566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.9650027071579448e-05,
-      "loss": 1.0488,
+      "learning_rate": 1.961776803850824e-05,
+      "loss": 1.0044,
       "step": 4076
     },
     {
-      "epoch": 0.11197165691686578,
+      "epoch": 0.11569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9649793766795163e-05,
-      "loss": 1.2349,
+      "learning_rate": 1.961751632252689e-05,
+      "loss": 1.0907,
       "step": 4077
     },
     {
-      "epoch": 0.11199912114470902,
+      "epoch": 0.11572077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.964956038565797e-05,
-      "loss": 1.1037,
+      "learning_rate": 1.9617264525305955e-05,
+      "loss": 0.9081,
       "step": 4078
     },
     {
-      "epoch": 0.11202658537255225,
+      "epoch": 0.11574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9649326928169718e-05,
-      "loss": 1.1285,
+      "learning_rate": 1.9617012646847547e-05,
+      "loss": 1.0083,
       "step": 4079
     },
     {
-      "epoch": 0.11205404960039549,
+      "epoch": 0.11577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.9649093394332254e-05,
-      "loss": 1.1023,
+      "learning_rate": 1.9616760687153807e-05,
+      "loss": 0.9765,
       "step": 4080
     },
     {
-      "epoch": 0.11208151382823872,
+      "epoch": 0.11580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.9648859784147424e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.9616508646226853e-05,
+      "loss": 1.0609,
       "step": 4081
     },
     {
-      "epoch": 0.11210897805608196,
+      "epoch": 0.11583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.9648626097617076e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.9616256524068822e-05,
+      "loss": 0.9904,
       "step": 4082
     },
     {
-      "epoch": 0.11213644228392519,
+      "epoch": 0.11586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9648392334743065e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.9616004320681836e-05,
+      "loss": 1.0531,
       "step": 4083
     },
     {
-      "epoch": 0.11216390651176843,
+      "epoch": 0.11589103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 1.9648158495527235e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.961575203606803e-05,
+      "loss": 1.0703,
       "step": 4084
     },
     {
-      "epoch": 0.11219137073961166,
+      "epoch": 0.1159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.9647924579971436e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9615499670229537e-05,
+      "loss": 1.0513,
       "step": 4085
     },
     {
-      "epoch": 0.11221883496745488,
+      "epoch": 0.1159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.964769058807752e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.9615247223168482e-05,
+      "loss": 0.9217,
       "step": 4086
     },
     {
-      "epoch": 0.11224629919529812,
+      "epoch": 0.1159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9647456519847343e-05,
-      "loss": 1.034,
+      "learning_rate": 1.9614994694887003e-05,
+      "loss": 1.1132,
       "step": 4087
     },
     {
-      "epoch": 0.11227376342314135,
+      "epoch": 0.11600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.964722237528275e-05,
-      "loss": 1.1548,
+      "learning_rate": 1.9614742085387232e-05,
+      "loss": 1.0988,
       "step": 4088
     },
     {
-      "epoch": 0.11230122765098459,
+      "epoch": 0.11603291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.96469881543856e-05,
-      "loss": 1.0853,
+      "learning_rate": 1.9614489394671305e-05,
+      "loss": 1.011,
       "step": 4089
     },
     {
-      "epoch": 0.11232869187882782,
+      "epoch": 0.11606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.964675385715774e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.961423662274135e-05,
+      "loss": 1.0355,
       "step": 4090
     },
     {
-      "epoch": 0.11235615610667106,
+      "epoch": 0.11608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9646519483601024e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.961398376959951e-05,
+      "loss": 1.0412,
       "step": 4091
     },
     {
-      "epoch": 0.1123836203345143,
+      "epoch": 0.11611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.9646285033717313e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.9613730835247914e-05,
+      "loss": 1.0043,
       "step": 4092
     },
     {
-      "epoch": 0.11241108456235753,
+      "epoch": 0.11614642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.9646050507508455e-05,
-      "loss": 1.0866,
+      "learning_rate": 1.9613477819688703e-05,
+      "loss": 1.1013,
       "step": 4093
     },
     {
-      "epoch": 0.11243854879020077,
+      "epoch": 0.11617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.964581590497631e-05,
-      "loss": 1.1158,
+      "learning_rate": 1.961322472292401e-05,
+      "loss": 0.9677,
       "step": 4094
     },
     {
-      "epoch": 0.112466013018044,
+      "epoch": 0.11620317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 1.9645581226122735e-05,
-      "loss": 1.1271,
+      "learning_rate": 1.9612971544955978e-05,
+      "loss": 1.0467,
       "step": 4095
     },
     {
-      "epoch": 0.11249347724588724,
+      "epoch": 0.11623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9645346470949586e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.9612718285786743e-05,
+      "loss": 1.069,
       "step": 4096
     },
     {
-      "epoch": 0.11252094147373047,
+      "epoch": 0.11625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.9645111639458717e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.9612464945418444e-05,
+      "loss": 1.0786,
       "step": 4097
     },
     {
-      "epoch": 0.1125484057015737,
+      "epoch": 0.11628830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.964487673165199e-05,
-      "loss": 1.0007,
+      "learning_rate": 1.961221152385322e-05,
+      "loss": 1.1109,
       "step": 4098
     },
     {
-      "epoch": 0.11257586992941694,
+      "epoch": 0.116316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.964464174753126e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.9611958021093214e-05,
+      "loss": 1.0944,
       "step": 4099
     },
     {
-      "epoch": 0.11260333415726016,
+      "epoch": 0.11634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.964440668709839e-05,
-      "loss": 1.1582,
+      "learning_rate": 1.961170443714057e-05,
+      "loss": 1.07,
       "step": 4100
     },
     {
-      "epoch": 0.1126307983851034,
+      "epoch": 0.11637343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 1.9644171550355237e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.9611450771997422e-05,
+      "loss": 1.1453,
       "step": 4101
     },
     {
-      "epoch": 0.11265826261294663,
+      "epoch": 0.11640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.9643936337303666e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.9611197025665916e-05,
+      "loss": 1.0072,
       "step": 4102
     },
     {
-      "epoch": 0.11268572684078987,
+      "epoch": 0.11643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.9643701047945533e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.9610943198148203e-05,
+      "loss": 1.0079,
       "step": 4103
     },
     {
-      "epoch": 0.1127131910686331,
+      "epoch": 0.11645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.96434656822827e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.961068928944642e-05,
+      "loss": 1.0444,
       "step": 4104
     },
     {
-      "epoch": 0.11274065529647634,
+      "epoch": 0.11648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9643230240317034e-05,
-      "loss": 1.0655,
+      "learning_rate": 1.9610435299562706e-05,
+      "loss": 0.9949,
       "step": 4105
     },
     {
-      "epoch": 0.11276811952431957,
+      "epoch": 0.11651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.964299472205039e-05,
-      "loss": 1.0532,
+      "learning_rate": 1.961018122849922e-05,
+      "loss": 1.0096,
       "step": 4106
     },
     {
-      "epoch": 0.11279558375216281,
+      "epoch": 0.11654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.9642759127484644e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.96099270762581e-05,
+      "loss": 0.8753,
       "step": 4107
     },
     {
-      "epoch": 0.11282304798000604,
+      "epoch": 0.11657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.964252345662165e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.960967284284149e-05,
+      "loss": 1.0131,
       "step": 4108
     },
     {
-      "epoch": 0.11285051220784928,
+      "epoch": 0.11660045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 1.964228770946327e-05,
-      "loss": 1.1289,
+      "learning_rate": 1.9609418528251543e-05,
+      "loss": 1.0196,
       "step": 4109
     },
     {
-      "epoch": 0.11287797643569251,
+      "epoch": 0.11662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.9642051886011383e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.9609164132490407e-05,
+      "loss": 0.9666,
       "step": 4110
     },
     {
-      "epoch": 0.11290544066353575,
+      "epoch": 0.1166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.964181598626784e-05,
-      "loss": 1.1257,
+      "learning_rate": 1.9608909655560227e-05,
+      "loss": 0.9917,
       "step": 4111
     },
     {
-      "epoch": 0.11293290489137899,
+      "epoch": 0.11668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.9641580010234517e-05,
-      "loss": 1.0462,
+      "learning_rate": 1.9608655097463156e-05,
+      "loss": 1.0291,
       "step": 4112
     },
     {
-      "epoch": 0.1129603691192222,
+      "epoch": 0.11671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.9641343957913276e-05,
-      "loss": 1.1195,
+      "learning_rate": 1.9608400458201348e-05,
+      "loss": 1.0295,
       "step": 4113
     },
     {
-      "epoch": 0.11298783334706544,
+      "epoch": 0.11674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.964110782930599e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.9608145737776945e-05,
+      "loss": 1.0032,
       "step": 4114
     },
     {
-      "epoch": 0.11301529757490868,
+      "epoch": 0.11677071509648126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9640871624414523e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.9607890936192105e-05,
+      "loss": 1.1292,
       "step": 4115
     },
     {
-      "epoch": 0.11304276180275191,
+      "epoch": 0.11679909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9640635343240744e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.9607636053448975e-05,
+      "loss": 1.0858,
       "step": 4116
     },
     {
-      "epoch": 0.11307022603059515,
+      "epoch": 0.11682746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9640398985786523e-05,
-      "loss": 1.1125,
+      "learning_rate": 1.9607381089549715e-05,
+      "loss": 1.0394,
       "step": 4117
     },
     {
-      "epoch": 0.11309769025843838,
+      "epoch": 0.11685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.964016255205373e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.9607126044496476e-05,
+      "loss": 1.0952,
       "step": 4118
     },
     {
-      "epoch": 0.11312515448628162,
+      "epoch": 0.11688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.963992604204424e-05,
-      "loss": 1.013,
+      "learning_rate": 1.9606870918291405e-05,
+      "loss": 1.0227,
       "step": 4119
     },
     {
-      "epoch": 0.11315261871412485,
+      "epoch": 0.11691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.9639689455759922e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.9606615710936668e-05,
+      "loss": 0.9898,
       "step": 4120
     },
     {
-      "epoch": 0.11318008294196809,
+      "epoch": 0.11694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 1.9639452793202644e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.960636042243442e-05,
+      "loss": 0.9616,
       "step": 4121
     },
     {
-      "epoch": 0.11320754716981132,
+      "epoch": 0.11696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.963921605437428e-05,
-      "loss": 1.1228,
+      "learning_rate": 1.9606105052786808e-05,
+      "loss": 1.1262,
       "step": 4122
     },
     {
-      "epoch": 0.11323501139765456,
+      "epoch": 0.1169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9638979239276706e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.9605849601995994e-05,
+      "loss": 1.1374,
       "step": 4123
     },
     {
-      "epoch": 0.1132624756254978,
+      "epoch": 0.1170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.96387423479118e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.960559407006414e-05,
+      "loss": 1.0926,
       "step": 4124
     },
     {
-      "epoch": 0.11328993985334103,
+      "epoch": 0.1170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.9638505380281424e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.96053384569934e-05,
+      "loss": 1.0398,
       "step": 4125
     },
     {
-      "epoch": 0.11331740408118426,
+      "epoch": 0.11708286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.9638268336387463e-05,
-      "loss": 0.8133,
+      "learning_rate": 1.9605082762785933e-05,
+      "loss": 1.0651,
       "step": 4126
     },
     {
-      "epoch": 0.11334486830902749,
+      "epoch": 0.11711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.963803121623179e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.9604826987443905e-05,
+      "loss": 1.0579,
       "step": 4127
     },
     {
-      "epoch": 0.11337233253687072,
+      "epoch": 0.11713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.963779401981628e-05,
-      "loss": 1.0812,
+      "learning_rate": 1.9604571130969467e-05,
+      "loss": 0.9307,
       "step": 4128
     },
     {
-      "epoch": 0.11339979676471396,
+      "epoch": 0.11716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.963755674714281e-05,
-      "loss": 1.0546,
+      "learning_rate": 1.9604315193364787e-05,
+      "loss": 1.1095,
       "step": 4129
     },
     {
-      "epoch": 0.11342726099255719,
+      "epoch": 0.11719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.9637319398213256e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.9604059174632023e-05,
+      "loss": 0.9648,
       "step": 4130
     },
     {
-      "epoch": 0.11345472522040043,
+      "epoch": 0.11722474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.9637081973029498e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.960380307477334e-05,
+      "loss": 1.0556,
       "step": 4131
     },
     {
-      "epoch": 0.11348218944824366,
+      "epoch": 0.11725312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 1.9636844471593415e-05,
-      "loss": 1.0923,
+      "learning_rate": 1.9603546893790903e-05,
+      "loss": 1.1156,
       "step": 4132
     },
     {
-      "epoch": 0.1135096536760869,
+      "epoch": 0.11728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9636606893906887e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.960329063168687e-05,
+      "loss": 1.1877,
       "step": 4133
     },
     {
-      "epoch": 0.11353711790393013,
+      "epoch": 0.11730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.9636369239971792e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.960303428846341e-05,
+      "loss": 1.1138,
       "step": 4134
     },
     {
-      "epoch": 0.11356458213177337,
+      "epoch": 0.11733825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.963613150979001e-05,
-      "loss": 1.1474,
+      "learning_rate": 1.960277786412269e-05,
+      "loss": 1.0115,
       "step": 4135
     },
     {
-      "epoch": 0.1135920463596166,
+      "epoch": 0.11736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.9635893703363424e-05,
-      "loss": 1.037,
+      "learning_rate": 1.960252135866687e-05,
+      "loss": 1.1026,
       "step": 4136
     },
     {
-      "epoch": 0.11361951058745984,
+      "epoch": 0.1173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.963565582069391e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.9602264772098123e-05,
+      "loss": 1.122,
       "step": 4137
     },
     {
-      "epoch": 0.11364697481530307,
+      "epoch": 0.1174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.963541786178336e-05,
-      "loss": 1.0493,
+      "learning_rate": 1.9602008104418615e-05,
+      "loss": 1.1437,
       "step": 4138
     },
     {
-      "epoch": 0.11367443904314631,
+      "epoch": 0.11745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.963517982663365e-05,
-      "loss": 1.0336,
+      "learning_rate": 1.9601751355630512e-05,
+      "loss": 1.1245,
       "step": 4139
     },
     {
-      "epoch": 0.11370190327098953,
+      "epoch": 0.11748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.9634941715246665e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.960149452573598e-05,
+      "loss": 0.9928,
       "step": 4140
     },
     {
-      "epoch": 0.11372936749883276,
+      "epoch": 0.11750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.963470352762429e-05,
-      "loss": 1.1066,
+      "learning_rate": 1.9601237614737196e-05,
+      "loss": 1.0892,
       "step": 4141
     },
     {
-      "epoch": 0.113756831726676,
+      "epoch": 0.11753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9634465263768405e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.9600980622636328e-05,
+      "loss": 1.0927,
       "step": 4142
     },
     {
-      "epoch": 0.11378429595451924,
+      "epoch": 0.11756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.96342269236809e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.960072354943554e-05,
+      "loss": 1.075,
       "step": 4143
     },
     {
-      "epoch": 0.11381176018236247,
+      "epoch": 0.11759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.963398850736366e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.9600466395137014e-05,
+      "loss": 1.0562,
       "step": 4144
     },
     {
-      "epoch": 0.1138392244102057,
+      "epoch": 0.11762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9633750014818572e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.9600209159742913e-05,
+      "loss": 0.9906,
       "step": 4145
     },
     {
-      "epoch": 0.11386668863804894,
+      "epoch": 0.11765039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 1.963351144604752e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.9599951843255415e-05,
+      "loss": 1.0276,
       "step": 4146
     },
     {
-      "epoch": 0.11389415286589218,
+      "epoch": 0.11767877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.9633272801052398e-05,
-      "loss": 1.0755,
+      "learning_rate": 1.959969444567669e-05,
+      "loss": 1.0477,
       "step": 4147
     },
     {
-      "epoch": 0.11392161709373541,
+      "epoch": 0.11770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.9633034079835088e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.959943696700892e-05,
+      "loss": 0.9979,
       "step": 4148
     },
     {
-      "epoch": 0.11394908132157865,
+      "epoch": 0.1177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.963279528239748e-05,
-      "loss": 0.9777,
+      "learning_rate": 1.959917940725427e-05,
+      "loss": 1.175,
       "step": 4149
     },
     {
-      "epoch": 0.11397654554942188,
+      "epoch": 0.1177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9632556408741467e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.9598921766414923e-05,
+      "loss": 0.9969,
       "step": 4150
     },
     {
-      "epoch": 0.11400400977726512,
+      "epoch": 0.11779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.9632317458868933e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.959866404449305e-05,
+      "loss": 1.0733,
       "step": 4151
     },
     {
-      "epoch": 0.11403147400510835,
+      "epoch": 0.11782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.9632078432781775e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.959840624149083e-05,
+      "loss": 1.1105,
       "step": 4152
     },
     {
-      "epoch": 0.11405893823295159,
+      "epoch": 0.11784903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 1.9631839330481875e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.959814835741044e-05,
+      "loss": 1.0872,
       "step": 4153
     },
     {
-      "epoch": 0.11408640246079481,
+      "epoch": 0.11787741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9631600151971138e-05,
-      "loss": 1.1388,
+      "learning_rate": 1.959789039225406e-05,
+      "loss": 1.1017,
       "step": 4154
     },
     {
-      "epoch": 0.11411386668863804,
+      "epoch": 0.11790578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.9631360897251447e-05,
-      "loss": 1.0352,
+      "learning_rate": 1.9597632346023873e-05,
+      "loss": 1.0811,
       "step": 4155
     },
     {
-      "epoch": 0.11414133091648128,
+      "epoch": 0.11793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.96311215663247e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.959737421872205e-05,
+      "loss": 1.1561,
       "step": 4156
     },
     {
-      "epoch": 0.11416879514432451,
+      "epoch": 0.11796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9630882159192784e-05,
-      "loss": 0.9808,
+      "learning_rate": 1.9597116010350773e-05,
+      "loss": 1.0168,
       "step": 4157
     },
     {
-      "epoch": 0.11419625937216775,
+      "epoch": 0.11799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.96306426758576e-05,
-      "loss": 1.0951,
+      "learning_rate": 1.959685772091223e-05,
+      "loss": 1.0173,
       "step": 4158
     },
     {
-      "epoch": 0.11422372360001098,
+      "epoch": 0.11801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.963040311632104e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.9596599350408594e-05,
+      "loss": 1.0641,
       "step": 4159
     },
     {
-      "epoch": 0.11425118782785422,
+      "epoch": 0.11804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.9630163480585003e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.959634089884206e-05,
+      "loss": 0.9816,
       "step": 4160
     },
     {
-      "epoch": 0.11427865205569745,
+      "epoch": 0.11807604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 1.962992376865138e-05,
-      "loss": 1.1714,
+      "learning_rate": 1.9596082366214797e-05,
+      "loss": 0.9308,
       "step": 4161
     },
     {
-      "epoch": 0.11430611628354069,
+      "epoch": 0.1181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.962968398052207e-05,
-      "loss": 1.1273,
+      "learning_rate": 1.9595823752528993e-05,
+      "loss": 1.0388,
       "step": 4162
     },
     {
-      "epoch": 0.11433358051138393,
+      "epoch": 0.11813280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 1.962944411619897e-05,
-      "loss": 1.1327,
+      "learning_rate": 1.9595565057786838e-05,
+      "loss": 1.0391,
       "step": 4163
     },
     {
-      "epoch": 0.11436104473922716,
+      "epoch": 0.11816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.9629204175683977e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9595306281990514e-05,
+      "loss": 1.0997,
       "step": 4164
     },
     {
-      "epoch": 0.1143885089670704,
+      "epoch": 0.11818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.9628964158978993e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.9595047425142204e-05,
+      "loss": 1.1123,
       "step": 4165
     },
     {
-      "epoch": 0.11441597319491363,
+      "epoch": 0.11821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.9628724066085913e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.95947884872441e-05,
+      "loss": 0.985,
       "step": 4166
     },
     {
-      "epoch": 0.11444343742275685,
+      "epoch": 0.11824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.962848389700664e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.9594529468298383e-05,
+      "loss": 1.028,
       "step": 4167
     },
     {
-      "epoch": 0.11447090165060009,
+      "epoch": 0.11827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9628243651743074e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.9594270368307247e-05,
+      "loss": 1.0864,
       "step": 4168
     },
     {
-      "epoch": 0.11449836587844332,
+      "epoch": 0.11830306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.962800333029711e-05,
-      "loss": 1.0904,
+      "learning_rate": 1.9594011187272875e-05,
+      "loss": 1.0398,
       "step": 4169
     },
     {
-      "epoch": 0.11452583010628656,
+      "epoch": 0.11833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9627762932670656e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.959375192519746e-05,
+      "loss": 0.9711,
       "step": 4170
     },
     {
-      "epoch": 0.1145532943341298,
+      "epoch": 0.11835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.962752245886561e-05,
-      "loss": 1.2206,
+      "learning_rate": 1.9593492582083196e-05,
+      "loss": 1.0677,
       "step": 4171
     },
     {
-      "epoch": 0.11458075856197303,
+      "epoch": 0.11838819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.962728190888388e-05,
-      "loss": 1.1446,
+      "learning_rate": 1.9593233157932265e-05,
+      "loss": 1.0674,
       "step": 4172
     },
     {
-      "epoch": 0.11460822278981626,
+      "epoch": 0.11841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.9627041282727366e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.9592973652746862e-05,
+      "loss": 1.1241,
       "step": 4173
     },
     {
-      "epoch": 0.1146356870176595,
+      "epoch": 0.11844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.9626800580397973e-05,
-      "loss": 0.9842,
+      "learning_rate": 1.959271406652918e-05,
+      "loss": 1.0658,
       "step": 4174
     },
     {
-      "epoch": 0.11466315124550273,
+      "epoch": 0.1184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9626559801897605e-05,
-      "loss": 1.019,
+      "learning_rate": 1.9592454399281414e-05,
+      "loss": 1.0431,
       "step": 4175
     },
     {
-      "epoch": 0.11469061547334597,
+      "epoch": 0.1185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9626318947228164e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.959219465100575e-05,
+      "loss": 1.1276,
       "step": 4176
     },
     {
-      "epoch": 0.1147180797011892,
+      "epoch": 0.11853007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 1.962607801639156e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.959193482170439e-05,
+      "loss": 1.0502,
       "step": 4177
     },
     {
-      "epoch": 0.11474554392903244,
+      "epoch": 0.11855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.9625837009389698e-05,
-      "loss": 1.0828,
+      "learning_rate": 1.9591674911379523e-05,
+      "loss": 1.0696,
       "step": 4178
     },
     {
-      "epoch": 0.11477300815687567,
+      "epoch": 0.11858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9625595926224487e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.9591414920033345e-05,
+      "loss": 0.8493,
       "step": 4179
     },
     {
-      "epoch": 0.11480047238471891,
+      "epoch": 0.11861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.9625354766897828e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.9591154847668057e-05,
+      "loss": 1.0138,
       "step": 4180
     },
     {
-      "epoch": 0.11482793661256213,
+      "epoch": 0.11864358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.9625113531411637e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.9590894694285852e-05,
+      "loss": 1.0174,
       "step": 4181
     },
     {
-      "epoch": 0.11485540084040537,
+      "epoch": 0.11867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.962487221976782e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.9590634459888927e-05,
+      "loss": 1.0338,
       "step": 4182
     },
     {
-      "epoch": 0.1148828650682486,
+      "epoch": 0.11870034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.9624630831968284e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.9590374144479484e-05,
+      "loss": 1.0905,
       "step": 4183
     },
     {
-      "epoch": 0.11491032929609184,
+      "epoch": 0.11872871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.962438936801494e-05,
-      "loss": 1.1465,
+      "learning_rate": 1.959011374805972e-05,
+      "loss": 1.0157,
       "step": 4184
     },
     {
-      "epoch": 0.11493779352393507,
+      "epoch": 0.11875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.9624147827909697e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.9589853270631833e-05,
+      "loss": 1.0246,
       "step": 4185
     },
     {
-      "epoch": 0.11496525775177831,
+      "epoch": 0.11878547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.9623906211654472e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.958959271219802e-05,
+      "loss": 1.0075,
       "step": 4186
     },
     {
-      "epoch": 0.11499272197962154,
+      "epoch": 0.11881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.9623664519251174e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.9589332072760493e-05,
+      "loss": 1.0144,
       "step": 4187
     },
     {
-      "epoch": 0.11502018620746478,
+      "epoch": 0.1188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9623422750701712e-05,
-      "loss": 1.0527,
+      "learning_rate": 1.9589071352321443e-05,
+      "loss": 1.0042,
       "step": 4188
     },
     {
-      "epoch": 0.11504765043530801,
+      "epoch": 0.11887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.9623180906008e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.958881055088308e-05,
+      "loss": 1.0262,
       "step": 4189
     },
     {
-      "epoch": 0.11507511466315125,
+      "epoch": 0.11889897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 1.9622938985171958e-05,
-      "loss": 1.1078,
+      "learning_rate": 1.9588549668447595e-05,
+      "loss": 0.9699,
       "step": 4190
     },
     {
-      "epoch": 0.11510257889099448,
+      "epoch": 0.11892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.962269698819549e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.9588288705017205e-05,
+      "loss": 0.9053,
       "step": 4191
     },
     {
-      "epoch": 0.11513004311883772,
+      "epoch": 0.11895573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 1.962245491508052e-05,
-      "loss": 1.0793,
+      "learning_rate": 1.9588027660594114e-05,
+      "loss": 0.9716,
       "step": 4192
     },
     {
-      "epoch": 0.11515750734668095,
+      "epoch": 0.11898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.962221276582896e-05,
-      "loss": 1.129,
+      "learning_rate": 1.9587766535180518e-05,
+      "loss": 1.1319,
       "step": 4193
     },
     {
-      "epoch": 0.11518497157452418,
+      "epoch": 0.11901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.9621970540442722e-05,
-      "loss": 1.0001,
+      "learning_rate": 1.9587505328778626e-05,
+      "loss": 1.0848,
       "step": 4194
     },
     {
-      "epoch": 0.11521243580236741,
+      "epoch": 0.11904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.9621728238923724e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.9587244041390646e-05,
+      "loss": 0.9986,
       "step": 4195
     },
     {
-      "epoch": 0.11523990003021065,
+      "epoch": 0.11906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.962148586127389e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.9586982673018785e-05,
+      "loss": 1.1578,
       "step": 4196
     },
     {
-      "epoch": 0.11526736425805388,
+      "epoch": 0.11909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.962124340749513e-05,
-      "loss": 1.1446,
+      "learning_rate": 1.958672122366525e-05,
+      "loss": 1.0323,
       "step": 4197
     },
     {
-      "epoch": 0.11529482848589712,
+      "epoch": 0.11912599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.9621000877589367e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.958645969333225e-05,
+      "loss": 1.0347,
       "step": 4198
     },
     {
-      "epoch": 0.11532229271374035,
+      "epoch": 0.11915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.9620758271558518e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.9586198082021995e-05,
+      "loss": 1.1457,
       "step": 4199
     },
     {
-      "epoch": 0.11534975694158359,
+      "epoch": 0.1191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.9620515589404502e-05,
-      "loss": 1.022,
+      "learning_rate": 1.9585936389736695e-05,
+      "loss": 1.1045,
       "step": 4200
     },
     {
-      "epoch": 0.11537722116942682,
+      "epoch": 0.119211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.962027283112924e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.9585674616478558e-05,
+      "loss": 1.189,
       "step": 4201
     },
     {
-      "epoch": 0.11540468539727006,
+      "epoch": 0.11923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9620029996734655e-05,
-      "loss": 0.968,
+      "learning_rate": 1.9585412762249798e-05,
+      "loss": 1.181,
       "step": 4202
     },
     {
-      "epoch": 0.11543214962511329,
+      "epoch": 0.11926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.961978708622266e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.9585150827052622e-05,
+      "loss": 0.964,
       "step": 4203
     },
     {
-      "epoch": 0.11545961385295653,
+      "epoch": 0.11929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 1.961954409959519e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.958488881088925e-05,
+      "loss": 1.0262,
       "step": 4204
     },
     {
-      "epoch": 0.11548707808079976,
+      "epoch": 0.11932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.9619301036854156e-05,
-      "loss": 1.1552,
+      "learning_rate": 1.958462671376189e-05,
+      "loss": 1.0238,
       "step": 4205
     },
     {
-      "epoch": 0.115514542308643,
+      "epoch": 0.11935300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.961905789800149e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.9584364535672753e-05,
+      "loss": 0.9445,
       "step": 4206
     },
     {
-      "epoch": 0.11554200653648623,
+      "epoch": 0.11938138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 1.961881468303911e-05,
-      "loss": 1.1629,
+      "learning_rate": 1.958410227662406e-05,
+      "loss": 1.08,
       "step": 4207
     },
     {
-      "epoch": 0.11556947076432945,
+      "epoch": 0.11940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.9618571391968944e-05,
-      "loss": 1.2286,
+      "learning_rate": 1.958383993661803e-05,
+      "loss": 1.084,
       "step": 4208
     },
     {
-      "epoch": 0.11559693499217269,
+      "epoch": 0.11943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.9618328024792913e-05,
-      "loss": 1.1514,
+      "learning_rate": 1.9583577515656867e-05,
+      "loss": 0.9503,
       "step": 4209
     },
     {
-      "epoch": 0.11562439922001592,
+      "epoch": 0.11946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9618084581512944e-05,
-      "loss": 1.1476,
+      "learning_rate": 1.9583315013742798e-05,
+      "loss": 1.1407,
       "step": 4210
     },
     {
-      "epoch": 0.11565186344785916,
+      "epoch": 0.11949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.9617841062130967e-05,
-      "loss": 1.0701,
+      "learning_rate": 1.9583052430878035e-05,
+      "loss": 1.1064,
       "step": 4211
     },
     {
-      "epoch": 0.1156793276757024,
+      "epoch": 0.1195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9617597466648904e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.9582789767064795e-05,
+      "loss": 1.0126,
       "step": 4212
     },
     {
-      "epoch": 0.11570679190354563,
+      "epoch": 0.1195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9617353795068687e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.9582527022305303e-05,
+      "loss": 1.0219,
       "step": 4213
     },
     {
-      "epoch": 0.11573425613138887,
+      "epoch": 0.1195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9617110047392235e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.958226419660177e-05,
+      "loss": 1.0579,
       "step": 4214
     },
     {
-      "epoch": 0.1157617203592321,
+      "epoch": 0.11960839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 1.9616866223621488e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.9582001289956423e-05,
+      "loss": 0.9217,
       "step": 4215
     },
     {
-      "epoch": 0.11578918458707534,
+      "epoch": 0.11963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.961662232375837e-05,
-      "loss": 1.0961,
+      "learning_rate": 1.958173830237148e-05,
+      "loss": 1.0652,
       "step": 4216
     },
     {
-      "epoch": 0.11581664881491857,
+      "epoch": 0.11966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.961637834780481e-05,
-      "loss": 1.1777,
+      "learning_rate": 1.9581475233849165e-05,
+      "loss": 1.0363,
       "step": 4217
     },
     {
-      "epoch": 0.1158441130427618,
+      "epoch": 0.11969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.961613429576274e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.9581212084391693e-05,
+      "loss": 1.0451,
       "step": 4218
     },
     {
-      "epoch": 0.11587157727060504,
+      "epoch": 0.11972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.961589016763409e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.9580948854001293e-05,
+      "loss": 1.0936,
       "step": 4219
     },
     {
-      "epoch": 0.11589904149844828,
+      "epoch": 0.11975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9615645963420796e-05,
-      "loss": 1.049,
+      "learning_rate": 1.9580685542680192e-05,
+      "loss": 0.8717,
       "step": 4220
     },
     {
-      "epoch": 0.11592650572629151,
+      "epoch": 0.11977866061293985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9615401683124782e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.958042215043061e-05,
+      "loss": 0.972,
       "step": 4221
     },
     {
-      "epoch": 0.11595396995413473,
+      "epoch": 0.11980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9615157326747988e-05,
-      "loss": 1.0243,
+      "learning_rate": 1.9580158677254762e-05,
+      "loss": 1.0518,
       "step": 4222
     },
     {
-      "epoch": 0.11598143418197797,
+      "epoch": 0.11983541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.961491289429234e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.957989512315489e-05,
+      "loss": 0.9891,
       "step": 4223
     },
     {
-      "epoch": 0.1160088984098212,
+      "epoch": 0.11986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.961466838575978e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.957963148813321e-05,
+      "loss": 1.0074,
       "step": 4224
     },
     {
-      "epoch": 0.11603636263766444,
+      "epoch": 0.11989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.961442380115224e-05,
-      "loss": 1.1268,
+      "learning_rate": 1.9579367772191956e-05,
+      "loss": 1.0855,
       "step": 4225
     },
     {
-      "epoch": 0.11606382686550767,
+      "epoch": 0.1199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9614179140471655e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.957910397533335e-05,
+      "loss": 1.0192,
       "step": 4226
     },
     {
-      "epoch": 0.11609129109335091,
+      "epoch": 0.1199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.9613934403719957e-05,
-      "loss": 1.235,
+      "learning_rate": 1.9578840097559624e-05,
+      "loss": 1.0047,
       "step": 4227
     },
     {
-      "epoch": 0.11611875532119414,
+      "epoch": 0.11997729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.9613689590899086e-05,
-      "loss": 1.0378,
+      "learning_rate": 1.9578576138873003e-05,
+      "loss": 1.1127,
       "step": 4228
     },
     {
-      "epoch": 0.11614621954903738,
+      "epoch": 0.12000567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.9613444702010984e-05,
-      "loss": 1.117,
+      "learning_rate": 1.957831209927572e-05,
+      "loss": 1.0288,
       "step": 4229
     },
     {
-      "epoch": 0.11617368377688062,
+      "epoch": 0.12003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.961319973705758e-05,
-      "loss": 1.0429,
+      "learning_rate": 1.9578047978770003e-05,
+      "loss": 1.0513,
       "step": 4230
     },
     {
-      "epoch": 0.11620114800472385,
+      "epoch": 0.12006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.9612954696040818e-05,
-      "loss": 1.1887,
+      "learning_rate": 1.9577783777358085e-05,
+      "loss": 1.0339,
       "step": 4231
     },
     {
-      "epoch": 0.11622861223256709,
+      "epoch": 0.12009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.9612709578962633e-05,
-      "loss": 1.1287,
+      "learning_rate": 1.9577519495042194e-05,
+      "loss": 1.1266,
       "step": 4232
     },
     {
-      "epoch": 0.11625607646041032,
+      "epoch": 0.12011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.9612464385824968e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.9577255131824568e-05,
+      "loss": 1.0908,
       "step": 4233
     },
     {
-      "epoch": 0.11628354068825356,
+      "epoch": 0.12014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9612219116629756e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.9576990687707437e-05,
+      "loss": 1.055,
       "step": 4234
     },
     {
-      "epoch": 0.11631100491609678,
+      "epoch": 0.12017593643586832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9611973771378947e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.9576726162693036e-05,
+      "loss": 1.103,
       "step": 4235
     },
     {
-      "epoch": 0.11633846914394001,
+      "epoch": 0.12020431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 1.961172835007448e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.9576461556783597e-05,
+      "loss": 1.0001,
       "step": 4236
     },
     {
-      "epoch": 0.11636593337178325,
+      "epoch": 0.12023269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.961148285271829e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.957619686998136e-05,
+      "loss": 0.929,
       "step": 4237
     },
     {
-      "epoch": 0.11639339759962648,
+      "epoch": 0.12026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.961123727931233e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.9575932102288553e-05,
+      "loss": 1.0137,
       "step": 4238
     },
     {
-      "epoch": 0.11642086182746972,
+      "epoch": 0.1202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9610991629858532e-05,
-      "loss": 1.1245,
+      "learning_rate": 1.957566725370742e-05,
+      "loss": 1.0313,
       "step": 4239
     },
     {
-      "epoch": 0.11644832605531295,
+      "epoch": 0.12031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.961074590435885e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.9575402324240195e-05,
+      "loss": 1.0002,
       "step": 4240
     },
     {
-      "epoch": 0.11647579028315619,
+      "epoch": 0.12034619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 1.9610500102815222e-05,
-      "loss": 1.0965,
+      "learning_rate": 1.9575137313889115e-05,
+      "loss": 1.032,
       "step": 4241
     },
     {
-      "epoch": 0.11650325451099942,
+      "epoch": 0.12037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.9610254225229596e-05,
-      "loss": 0.988,
+      "learning_rate": 1.957487222265642e-05,
+      "loss": 1.0192,
       "step": 4242
     },
     {
-      "epoch": 0.11653071873884266,
+      "epoch": 0.12040295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.9610008271603912e-05,
-      "loss": 1.0518,
+      "learning_rate": 1.9574607050544346e-05,
+      "loss": 1.1024,
       "step": 4243
     },
     {
-      "epoch": 0.1165581829666859,
+      "epoch": 0.12043132803632237,
       "grad_norm": 0.0,
-      "learning_rate": 1.9609762241940124e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.9574341797555143e-05,
+      "loss": 0.9782,
       "step": 4244
     },
     {
-      "epoch": 0.11658564719452913,
+      "epoch": 0.12045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.960951613624017e-05,
-      "loss": 1.0291,
+      "learning_rate": 1.957407646369104e-05,
+      "loss": 1.1007,
       "step": 4245
     },
     {
-      "epoch": 0.11661311142237236,
+      "epoch": 0.12048808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.960926995450601e-05,
-      "loss": 1.1143,
+      "learning_rate": 1.957381104895428e-05,
+      "loss": 1.028,
       "step": 4246
     },
     {
-      "epoch": 0.1166405756502156,
+      "epoch": 0.12051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.9609023696739576e-05,
-      "loss": 1.1617,
+      "learning_rate": 1.9573545553347114e-05,
+      "loss": 0.9325,
       "step": 4247
     },
     {
-      "epoch": 0.11666803987805884,
+      "epoch": 0.12054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9608777362942828e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.9573279976871772e-05,
+      "loss": 1.0269,
       "step": 4248
     },
     {
-      "epoch": 0.11669550410590206,
+      "epoch": 0.12057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.960853095311771e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.957301431953051e-05,
+      "loss": 1.0361,
       "step": 4249
     },
     {
-      "epoch": 0.11672296833374529,
+      "epoch": 0.12060158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 1.9608284467266175e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.9572748581325564e-05,
+      "loss": 1.1081,
       "step": 4250
     },
     {
-      "epoch": 0.11675043256158853,
+      "epoch": 0.1206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.960803790539017e-05,
-      "loss": 1.1982,
+      "learning_rate": 1.9572482762259176e-05,
+      "loss": 1.0004,
       "step": 4251
     },
     {
-      "epoch": 0.11677789678943176,
+      "epoch": 0.12065834279228149,
       "grad_norm": 0.0,
-      "learning_rate": 1.9607791267491647e-05,
-      "loss": 1.146,
+      "learning_rate": 1.9572216862333597e-05,
+      "loss": 1.0369,
       "step": 4252
     },
     {
-      "epoch": 0.116805361017275,
+      "epoch": 0.1206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9607544553572557e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.9571950881551073e-05,
+      "loss": 1.1804,
       "step": 4253
     },
     {
-      "epoch": 0.11683282524511823,
+      "epoch": 0.12071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.9607297763634855e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.9571684819913853e-05,
+      "loss": 1.0,
       "step": 4254
     },
     {
-      "epoch": 0.11686028947296147,
+      "epoch": 0.12074347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.960705089768049e-05,
-      "loss": 0.9759,
+      "learning_rate": 1.9571418677424177e-05,
+      "loss": 0.9698,
       "step": 4255
     },
     {
-      "epoch": 0.1168877537008047,
+      "epoch": 0.12077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.9606803955711416e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.95711524540843e-05,
+      "loss": 1.0908,
       "step": 4256
     },
     {
-      "epoch": 0.11691521792864794,
+      "epoch": 0.12080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.9606556937729592e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.9570886149896467e-05,
+      "loss": 0.9986,
       "step": 4257
     },
     {
-      "epoch": 0.11694268215649117,
+      "epoch": 0.12082860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 1.9606309843736966e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.9570619764862925e-05,
+      "loss": 1.0992,
       "step": 4258
     },
     {
-      "epoch": 0.11697014638433441,
+      "epoch": 0.12085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.9606062673735493e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.957035329898593e-05,
+      "loss": 1.0654,
       "step": 4259
     },
     {
-      "epoch": 0.11699761061217764,
+      "epoch": 0.12088535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 1.9605815427727134e-05,
-      "loss": 0.98,
+      "learning_rate": 1.9570086752267734e-05,
+      "loss": 0.9616,
       "step": 4260
     },
     {
-      "epoch": 0.11702507484002088,
+      "epoch": 0.12091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.960556810571384e-05,
-      "loss": 1.117,
+      "learning_rate": 1.956982012471058e-05,
+      "loss": 1.0689,
       "step": 4261
     },
     {
-      "epoch": 0.1170525390678641,
+      "epoch": 0.12094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9605320707697574e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.9569553416316725e-05,
+      "loss": 0.9921,
       "step": 4262
     },
     {
-      "epoch": 0.11708000329570734,
+      "epoch": 0.12097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.960507323368029e-05,
-      "loss": 1.0273,
+      "learning_rate": 1.9569286627088426e-05,
+      "loss": 1.0349,
       "step": 4263
     },
     {
-      "epoch": 0.11710746752355057,
+      "epoch": 0.12099886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 1.9604825683663938e-05,
-      "loss": 1.1623,
+      "learning_rate": 1.956901975702793e-05,
+      "loss": 1.0918,
       "step": 4264
     },
     {
-      "epoch": 0.1171349317513938,
+      "epoch": 0.1210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.960457805765049e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.9568752806137495e-05,
+      "loss": 0.9783,
       "step": 4265
     },
     {
-      "epoch": 0.11716239597923704,
+      "epoch": 0.12105561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 1.96043303556419e-05,
-      "loss": 1.0636,
+      "learning_rate": 1.9568485774419377e-05,
+      "loss": 1.0912,
       "step": 4266
     },
     {
-      "epoch": 0.11718986020708028,
+      "epoch": 0.12108399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 1.9604082577640127e-05,
-      "loss": 1.0818,
+      "learning_rate": 1.9568218661875827e-05,
+      "loss": 0.9618,
       "step": 4267
     },
     {
-      "epoch": 0.11721732443492351,
+      "epoch": 0.12111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.9603834723647135e-05,
-      "loss": 1.081,
+      "learning_rate": 1.9567951468509104e-05,
+      "loss": 1.1501,
       "step": 4268
     },
     {
-      "epoch": 0.11724478866276675,
+      "epoch": 0.12114074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.9603586793664882e-05,
-      "loss": 1.0595,
+      "learning_rate": 1.9567684194321462e-05,
+      "loss": 0.95,
       "step": 4269
     },
     {
-      "epoch": 0.11727225289060998,
+      "epoch": 0.12116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.9603338787695328e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.9567416839315168e-05,
+      "loss": 0.9897,
       "step": 4270
     },
     {
-      "epoch": 0.11729971711845322,
+      "epoch": 0.12119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9603090705740436e-05,
-      "loss": 1.0714,
+      "learning_rate": 1.9567149403492467e-05,
+      "loss": 1.0158,
       "step": 4271
     },
     {
-      "epoch": 0.11732718134629645,
+      "epoch": 0.12122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9602842547802177e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.956688188685563e-05,
+      "loss": 1.0293,
       "step": 4272
     },
     {
-      "epoch": 0.11735464557413969,
+      "epoch": 0.12125425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.9602594313882503e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.956661428940691e-05,
+      "loss": 1.1032,
       "step": 4273
     },
     {
-      "epoch": 0.11738210980198292,
+      "epoch": 0.12128263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 1.9602346003983382e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.956634661114857e-05,
+      "loss": 1.0709,
       "step": 4274
     },
     {
-      "epoch": 0.11740957402982616,
+      "epoch": 0.12131101021566401,
       "grad_norm": 0.0,
-      "learning_rate": 1.9602097618106784e-05,
-      "loss": 1.085,
+      "learning_rate": 1.956607885208287e-05,
+      "loss": 1.0605,
       "step": 4275
     },
     {
-      "epoch": 0.11743703825766938,
+      "epoch": 0.12133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9601849156254665e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.956581101221207e-05,
+      "loss": 1.1083,
       "step": 4276
     },
     {
-      "epoch": 0.11746450248551261,
+      "epoch": 0.1213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9601600618429003e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.9565543091538436e-05,
+      "loss": 1.1079,
       "step": 4277
     },
     {
-      "epoch": 0.11749196671335585,
+      "epoch": 0.1213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9601352004631753e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.956527509006423e-05,
+      "loss": 1.0455,
       "step": 4278
     },
     {
-      "epoch": 0.11751943094119909,
+      "epoch": 0.12142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.9601103314864887e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.9565007007791713e-05,
+      "loss": 1.029,
       "step": 4279
     },
     {
-      "epoch": 0.11754689516904232,
+      "epoch": 0.12145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9600854549130374e-05,
-      "loss": 0.9469,
+      "learning_rate": 1.9564738844723153e-05,
+      "loss": 1.0315,
       "step": 4280
     },
     {
-      "epoch": 0.11757435939688556,
+      "epoch": 0.12148127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 1.9600605707430178e-05,
-      "loss": 1.1528,
+      "learning_rate": 1.9564470600860816e-05,
+      "loss": 1.0883,
       "step": 4281
     },
     {
-      "epoch": 0.11760182362472879,
+      "epoch": 0.12150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.9600356789766272e-05,
-      "loss": 1.1157,
+      "learning_rate": 1.956420227620696e-05,
+      "loss": 0.927,
       "step": 4282
     },
     {
-      "epoch": 0.11762928785257203,
+      "epoch": 0.12153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.9600107796140627e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.9563933870763864e-05,
+      "loss": 1.0752,
       "step": 4283
     },
     {
-      "epoch": 0.11765675208041526,
+      "epoch": 0.12156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.9599858726555206e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.9563665384533786e-05,
+      "loss": 1.2196,
       "step": 4284
     },
     {
-      "epoch": 0.1176842163082585,
+      "epoch": 0.12159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9599609581011987e-05,
-      "loss": 1.0273,
+      "learning_rate": 1.9563396817518994e-05,
+      "loss": 1.0078,
       "step": 4285
     },
     {
-      "epoch": 0.11771168053610173,
+      "epoch": 0.12162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.9599360359512938e-05,
-      "loss": 1.1454,
+      "learning_rate": 1.9563128169721763e-05,
+      "loss": 0.9704,
       "step": 4286
     },
     {
-      "epoch": 0.11773914476394497,
+      "epoch": 0.12165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.959911106206003e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.9562859441144357e-05,
+      "loss": 0.9243,
       "step": 4287
     },
     {
-      "epoch": 0.1177666089917882,
+      "epoch": 0.12167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.9598861688655238e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.9562590631789045e-05,
+      "loss": 1.108,
       "step": 4288
     },
     {
-      "epoch": 0.11779407321963142,
+      "epoch": 0.1217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9598612239300535e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.9562321741658106e-05,
+      "loss": 0.9806,
       "step": 4289
     },
     {
-      "epoch": 0.11782153744747466,
+      "epoch": 0.1217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9598362713997893e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.9562052770753804e-05,
+      "loss": 1.063,
       "step": 4290
     },
     {
-      "epoch": 0.1178490016753179,
+      "epoch": 0.12176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.9598113112749288e-05,
-      "loss": 1.0481,
+      "learning_rate": 1.956178371907841e-05,
+      "loss": 1.0398,
       "step": 4291
     },
     {
-      "epoch": 0.11787646590316113,
+      "epoch": 0.12179341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.959786343555669e-05,
-      "loss": 1.0739,
+      "learning_rate": 1.95615145866342e-05,
+      "loss": 1.0381,
       "step": 4292
     },
     {
-      "epoch": 0.11790393013100436,
+      "epoch": 0.12182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.9597613682422085e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.9561245373423445e-05,
+      "loss": 0.9782,
       "step": 4293
     },
     {
-      "epoch": 0.1179313943588476,
+      "epoch": 0.12185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9597363853347437e-05,
-      "loss": 1.0068,
+      "learning_rate": 1.956097607944842e-05,
+      "loss": 1.0728,
       "step": 4294
     },
     {
-      "epoch": 0.11795885858669083,
+      "epoch": 0.12187854710556185,
       "grad_norm": 0.0,
-      "learning_rate": 1.959711394833473e-05,
-      "loss": 1.1833,
+      "learning_rate": 1.9560706704711402e-05,
+      "loss": 1.006,
       "step": 4295
     },
     {
-      "epoch": 0.11798632281453407,
+      "epoch": 0.12190692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.9596863967385942e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.9560437249214665e-05,
+      "loss": 1.0505,
       "step": 4296
     },
     {
-      "epoch": 0.1180137870423773,
+      "epoch": 0.12193530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.9596613910503047e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.9560167712960486e-05,
+      "loss": 1.0984,
       "step": 4297
     },
     {
-      "epoch": 0.11804125127022054,
+      "epoch": 0.12196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.9596363777688027e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.9559898095951137e-05,
+      "loss": 0.9584,
       "step": 4298
     },
     {
-      "epoch": 0.11806871549806378,
+      "epoch": 0.12199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.959611356894286e-05,
-      "loss": 0.956,
+      "learning_rate": 1.9559628398188903e-05,
+      "loss": 1.0835,
       "step": 4299
     },
     {
-      "epoch": 0.11809617972590701,
+      "epoch": 0.12202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.9595863284269524e-05,
-      "loss": 1.1436,
+      "learning_rate": 1.9559358619676053e-05,
+      "loss": 1.0414,
       "step": 4300
     },
     {
-      "epoch": 0.11812364395375025,
+      "epoch": 0.12204880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 1.959561292367e-05,
-      "loss": 1.0817,
+      "learning_rate": 1.9559088760414875e-05,
+      "loss": 1.0281,
       "step": 4301
     },
     {
-      "epoch": 0.11815110818159348,
+      "epoch": 0.12207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.959536248714627e-05,
-      "loss": 1.1323,
+      "learning_rate": 1.9558818820407645e-05,
+      "loss": 1.0573,
       "step": 4302
     },
     {
-      "epoch": 0.1181785724094367,
+      "epoch": 0.122105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9595111974700316e-05,
-      "loss": 1.0986,
+      "learning_rate": 1.955854879965664e-05,
+      "loss": 1.0529,
       "step": 4303
     },
     {
-      "epoch": 0.11820603663727994,
+      "epoch": 0.1221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.9594861386334117e-05,
-      "loss": 1.0341,
+      "learning_rate": 1.9558278698164147e-05,
+      "loss": 1.144,
       "step": 4304
     },
     {
-      "epoch": 0.11823350086512317,
+      "epoch": 0.12216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.959461072204966e-05,
-      "loss": 1.0083,
+      "learning_rate": 1.955800851593244e-05,
+      "loss": 0.9736,
       "step": 4305
     },
     {
-      "epoch": 0.11826096509296641,
+      "epoch": 0.12219069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.9594359981848928e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.955773825296381e-05,
+      "loss": 1.0709,
       "step": 4306
     },
     {
-      "epoch": 0.11828842932080964,
+      "epoch": 0.12221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.95941091657339e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.955746790926053e-05,
+      "loss": 1.1087,
       "step": 4307
     },
     {
-      "epoch": 0.11831589354865288,
+      "epoch": 0.12224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9593858273706565e-05,
-      "loss": 1.1346,
+      "learning_rate": 1.9557197484824893e-05,
+      "loss": 1.0559,
       "step": 4308
     },
     {
-      "epoch": 0.11834335777649611,
+      "epoch": 0.12227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.9593607305768906e-05,
-      "loss": 1.059,
+      "learning_rate": 1.9556926979659178e-05,
+      "loss": 0.9951,
       "step": 4309
     },
     {
-      "epoch": 0.11837082200433935,
+      "epoch": 0.12230419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 1.9593356261922913e-05,
-      "loss": 1.0947,
+      "learning_rate": 1.9556656393765673e-05,
+      "loss": 1.0044,
       "step": 4310
     },
     {
-      "epoch": 0.11839828623218258,
+      "epoch": 0.12233257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 1.9593105142170564e-05,
-      "loss": 1.1511,
+      "learning_rate": 1.955638572714666e-05,
+      "loss": 0.9946,
       "step": 4311
     },
     {
-      "epoch": 0.11842575046002582,
+      "epoch": 0.12236095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 1.9592853946513855e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.9556114979804426e-05,
+      "loss": 1.1227,
       "step": 4312
     },
     {
-      "epoch": 0.11845321468786905,
+      "epoch": 0.12238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.959260267495477e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.955584415174126e-05,
+      "loss": 1.0947,
       "step": 4313
     },
     {
-      "epoch": 0.11848067891571229,
+      "epoch": 0.12241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.9592351327495296e-05,
-      "loss": 1.1388,
+      "learning_rate": 1.955557324295945e-05,
+      "loss": 1.034,
       "step": 4314
     },
     {
-      "epoch": 0.11850814314355552,
+      "epoch": 0.1224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.9592099904137424e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.955530225346128e-05,
+      "loss": 1.0192,
       "step": 4315
     },
     {
-      "epoch": 0.11853560737139875,
+      "epoch": 0.1224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9591848404883137e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.9555031183249046e-05,
+      "loss": 1.1046,
       "step": 4316
     },
     {
-      "epoch": 0.11856307159924198,
+      "epoch": 0.12250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.9591596829734433e-05,
-      "loss": 1.0418,
+      "learning_rate": 1.9554760032325032e-05,
+      "loss": 1.1022,
       "step": 4317
     },
     {
-      "epoch": 0.11859053582708522,
+      "epoch": 0.12253121452894439,
       "grad_norm": 0.0,
-      "learning_rate": 1.95913451786933e-05,
-      "loss": 1.0942,
+      "learning_rate": 1.9554488800691532e-05,
+      "loss": 1.1385,
       "step": 4318
     },
     {
-      "epoch": 0.11861800005492845,
+      "epoch": 0.12255959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 1.959109345176173e-05,
-      "loss": 1.0948,
+      "learning_rate": 1.955421748835084e-05,
+      "loss": 0.9344,
       "step": 4319
     },
     {
-      "epoch": 0.11864546428277169,
+      "epoch": 0.12258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.959084164894171e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.9553946095305235e-05,
+      "loss": 1.0627,
       "step": 4320
     },
     {
-      "epoch": 0.11867292851061492,
+      "epoch": 0.12261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.9590589770235236e-05,
-      "loss": 0.9283,
+      "learning_rate": 1.9553674621557023e-05,
+      "loss": 0.9851,
       "step": 4321
     },
     {
-      "epoch": 0.11870039273845816,
+      "epoch": 0.12264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9590337815644304e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.955340306710849e-05,
+      "loss": 1.0997,
       "step": 4322
     },
     {
-      "epoch": 0.11872785696630139,
+      "epoch": 0.12267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.9590085785170902e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.955313143196193e-05,
+      "loss": 1.0444,
       "step": 4323
     },
     {
-      "epoch": 0.11875532119414463,
+      "epoch": 0.12270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9589833678817026e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.955285971611964e-05,
+      "loss": 1.0107,
       "step": 4324
     },
     {
-      "epoch": 0.11878278542198786,
+      "epoch": 0.12272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.958958149658467e-05,
-      "loss": 0.9089,
+      "learning_rate": 1.9552587919583915e-05,
+      "loss": 1.0318,
       "step": 4325
     },
     {
-      "epoch": 0.1188102496498311,
+      "epoch": 0.12275822928490351,
       "grad_norm": 0.0,
-      "learning_rate": 1.9589329238475833e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.9552316042357054e-05,
+      "loss": 0.9616,
       "step": 4326
     },
     {
-      "epoch": 0.11883771387767433,
+      "epoch": 0.12278660612939842,
       "grad_norm": 0.0,
-      "learning_rate": 1.9589076904492508e-05,
-      "loss": 1.0654,
+      "learning_rate": 1.9552044084441347e-05,
+      "loss": 1.0266,
       "step": 4327
     },
     {
-      "epoch": 0.11886517810551757,
+      "epoch": 0.1228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.958882449463669e-05,
-      "loss": 1.0698,
+      "learning_rate": 1.9551772045839094e-05,
+      "loss": 1.0522,
       "step": 4328
     },
     {
-      "epoch": 0.1188926423333608,
+      "epoch": 0.12284335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 1.958857200891038e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.9551499926552595e-05,
+      "loss": 1.1049,
       "step": 4329
     },
     {
-      "epoch": 0.11892010656120403,
+      "epoch": 0.12287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.9588319447315577e-05,
-      "loss": 0.984,
+      "learning_rate": 1.9551227726584145e-05,
+      "loss": 1.0912,
       "step": 4330
     },
     {
-      "epoch": 0.11894757078904726,
+      "epoch": 0.12290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9588066809854272e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.9550955445936047e-05,
+      "loss": 1.0573,
       "step": 4331
     },
     {
-      "epoch": 0.1189750350168905,
+      "epoch": 0.12292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.958781409652847e-05,
-      "loss": 1.0525,
+      "learning_rate": 1.9550683084610598e-05,
+      "loss": 1.0483,
       "step": 4332
     },
     {
-      "epoch": 0.11900249924473373,
+      "epoch": 0.12295686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 1.958756130734017e-05,
-      "loss": 1.0524,
+      "learning_rate": 1.95504106426101e-05,
+      "loss": 1.0861,
       "step": 4333
     },
     {
-      "epoch": 0.11902996347257697,
+      "epoch": 0.12298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.9587308442291372e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.955013811993685e-05,
+      "loss": 1.1733,
       "step": 4334
     },
     {
-      "epoch": 0.1190574277004202,
+      "epoch": 0.12301362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.958705550138407e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.9549865516593158e-05,
+      "loss": 1.0495,
       "step": 4335
     },
     {
-      "epoch": 0.11908489192826344,
+      "epoch": 0.12304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9586802484620278e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.9549592832581323e-05,
+      "loss": 1.0343,
       "step": 4336
     },
     {
-      "epoch": 0.11911235615610667,
+      "epoch": 0.12307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.9586549392001986e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.9549320067903644e-05,
+      "loss": 1.0581,
       "step": 4337
     },
     {
-      "epoch": 0.11913982038394991,
+      "epoch": 0.12309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.9586296223531204e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.9549047222562434e-05,
+      "loss": 1.1263,
       "step": 4338
     },
     {
-      "epoch": 0.11916728461179314,
+      "epoch": 0.12312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.958604297920993e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.954877429655999e-05,
+      "loss": 1.1585,
       "step": 4339
     },
     {
-      "epoch": 0.11919474883963638,
+      "epoch": 0.123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9585789659040176e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.954850128989862e-05,
+      "loss": 0.9437,
       "step": 4340
     },
     {
-      "epoch": 0.11922221306747961,
+      "epoch": 0.1231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.958553626302394e-05,
-      "loss": 1.0951,
+      "learning_rate": 1.9548228202580635e-05,
+      "loss": 1.1284,
       "step": 4341
     },
     {
-      "epoch": 0.11924967729532285,
+      "epoch": 0.1232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.9585282791163224e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.9547955034608333e-05,
+      "loss": 1.0905,
       "step": 4342
     },
     {
-      "epoch": 0.11927714152316608,
+      "epoch": 0.12324063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 1.958502924346004e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9547681785984024e-05,
+      "loss": 1.0514,
       "step": 4343
     },
     {
-      "epoch": 0.1193046057510093,
+      "epoch": 0.12326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.9584775619916395e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.954740845671002e-05,
+      "loss": 1.0458,
       "step": 4344
     },
     {
-      "epoch": 0.11933206997885254,
+      "epoch": 0.12329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9584521920534288e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.9547135046788627e-05,
+      "loss": 1.1177,
       "step": 4345
     },
     {
-      "epoch": 0.11935953420669577,
+      "epoch": 0.12332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9584268145315733e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.954686155622216e-05,
+      "loss": 1.0179,
       "step": 4346
     },
     {
-      "epoch": 0.11938699843453901,
+      "epoch": 0.12335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.958401429426273e-05,
-      "loss": 1.158,
+      "learning_rate": 1.9546587985012918e-05,
+      "loss": 1.0141,
       "step": 4347
     },
     {
-      "epoch": 0.11941446266238225,
+      "epoch": 0.12338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.9583760367377303e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.954631433316322e-05,
+      "loss": 1.1134,
       "step": 4348
     },
     {
-      "epoch": 0.11944192689022548,
+      "epoch": 0.12341089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9583506364661446e-05,
-      "loss": 1.1097,
+      "learning_rate": 1.9546040600675375e-05,
+      "loss": 1.138,
       "step": 4349
     },
     {
-      "epoch": 0.11946939111806872,
+      "epoch": 0.12343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9583252286117177e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.9545766787551697e-05,
+      "loss": 1.0737,
       "step": 4350
     },
     {
-      "epoch": 0.11949685534591195,
+      "epoch": 0.12346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.95829981317465e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.95454928937945e-05,
+      "loss": 1.1171,
       "step": 4351
     },
     {
-      "epoch": 0.11952431957375519,
+      "epoch": 0.12349602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.9582743901551432e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.954521891940609e-05,
+      "loss": 1.0168,
       "step": 4352
     },
     {
-      "epoch": 0.11955178380159842,
+      "epoch": 0.12352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.9582489595533982e-05,
-      "loss": 1.0619,
+      "learning_rate": 1.954494486438879e-05,
+      "loss": 1.0589,
       "step": 4353
     },
     {
-      "epoch": 0.11957924802944166,
+      "epoch": 0.1235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9582235213696163e-05,
-      "loss": 1.1126,
+      "learning_rate": 1.9544670728744912e-05,
+      "loss": 1.0789,
       "step": 4354
     },
     {
-      "epoch": 0.11960671225728489,
+      "epoch": 0.12358115777525538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9581980756039988e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.954439651247677e-05,
+      "loss": 1.0273,
       "step": 4355
     },
     {
-      "epoch": 0.11963417648512813,
+      "epoch": 0.12360953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 1.9581726222567465e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.9544122215586682e-05,
+      "loss": 1.0322,
       "step": 4356
     },
     {
-      "epoch": 0.11966164071297135,
+      "epoch": 0.12363791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.9581471613280615e-05,
-      "loss": 1.1187,
+      "learning_rate": 1.9543847838076966e-05,
+      "loss": 1.0545,
       "step": 4357
     },
     {
-      "epoch": 0.11968910494081458,
+      "epoch": 0.12366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.9581216928181453e-05,
-      "loss": 1.093,
+      "learning_rate": 1.9543573379949937e-05,
+      "loss": 1.044,
       "step": 4358
     },
     {
-      "epoch": 0.11971656916865782,
+      "epoch": 0.12369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.958096216727199e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.9543298841207913e-05,
+      "loss": 1.0353,
       "step": 4359
     },
     {
-      "epoch": 0.11974403339650105,
+      "epoch": 0.12372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.958070733055424e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.9543024221853218e-05,
+      "loss": 1.0447,
       "step": 4360
     },
     {
-      "epoch": 0.11977149762434429,
+      "epoch": 0.12375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.9580452418030226e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9542749521888165e-05,
+      "loss": 1.0099,
       "step": 4361
     },
     {
-      "epoch": 0.11979896185218752,
+      "epoch": 0.12377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.958019742970196e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.954247474131508e-05,
+      "loss": 0.9728,
       "step": 4362
     },
     {
-      "epoch": 0.11982642608003076,
+      "epoch": 0.12380817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.9579942365571458e-05,
-      "loss": 1.0963,
+      "learning_rate": 1.9542199880136284e-05,
+      "loss": 1.1204,
       "step": 4363
     },
     {
-      "epoch": 0.119853890307874,
+      "epoch": 0.12383654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 1.9579687225640744e-05,
-      "loss": 1.102,
+      "learning_rate": 1.9541924938354096e-05,
+      "loss": 1.0256,
       "step": 4364
     },
     {
-      "epoch": 0.11988135453571723,
+      "epoch": 0.12386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.9579432009911836e-05,
-      "loss": 1.0836,
+      "learning_rate": 1.9541649915970835e-05,
+      "loss": 1.0178,
       "step": 4365
     },
     {
-      "epoch": 0.11990881876356047,
+      "epoch": 0.1238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9579176718386747e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.954137481298883e-05,
+      "loss": 1.091,
       "step": 4366
     },
     {
-      "epoch": 0.1199362829914037,
+      "epoch": 0.1239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9578921351067503e-05,
-      "loss": 1.068,
+      "learning_rate": 1.9541099629410405e-05,
+      "loss": 1.0563,
       "step": 4367
     },
     {
-      "epoch": 0.11996374721924694,
+      "epoch": 0.12395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.957866590795612e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.954082436523788e-05,
+      "loss": 0.9957,
       "step": 4368
     },
     {
-      "epoch": 0.11999121144709017,
+      "epoch": 0.12397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9578410389054626e-05,
-      "loss": 1.0975,
+      "learning_rate": 1.9540549020473584e-05,
+      "loss": 1.0125,
       "step": 4369
     },
     {
-      "epoch": 0.1200186756749334,
+      "epoch": 0.12400681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 1.9578154794365037e-05,
-      "loss": 1.0298,
+      "learning_rate": 1.9540273595119842e-05,
+      "loss": 1.1509,
       "step": 4370
     },
     {
-      "epoch": 0.12004613990277663,
+      "epoch": 0.12403518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.957789912388938e-05,
-      "loss": 1.0987,
+      "learning_rate": 1.953999808917898e-05,
+      "loss": 1.0643,
       "step": 4371
     },
     {
-      "epoch": 0.12007360413061986,
+      "epoch": 0.12406356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 1.957764337762967e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.9539722502653324e-05,
+      "loss": 1.0358,
       "step": 4372
     },
     {
-      "epoch": 0.1201010683584631,
+      "epoch": 0.12409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.957738755558794e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.9539446835545202e-05,
+      "loss": 1.0133,
       "step": 4373
     },
     {
-      "epoch": 0.12012853258630633,
+      "epoch": 0.12412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.9577131657766207e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.953917108785695e-05,
+      "loss": 1.0349,
       "step": 4374
     },
     {
-      "epoch": 0.12015599681414957,
+      "epoch": 0.12414869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 1.95768756841665e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.9538895259590885e-05,
+      "loss": 1.0717,
       "step": 4375
     },
     {
-      "epoch": 0.1201834610419928,
+      "epoch": 0.12417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.957661963479084e-05,
-      "loss": 1.1504,
+      "learning_rate": 1.9538619350749345e-05,
+      "loss": 1.1575,
       "step": 4376
     },
     {
-      "epoch": 0.12021092526983604,
+      "epoch": 0.12420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.957636350964126e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.9538343361334657e-05,
+      "loss": 1.0524,
       "step": 4377
     },
     {
-      "epoch": 0.12023838949767927,
+      "epoch": 0.12423382519863792,
       "grad_norm": 0.0,
-      "learning_rate": 1.9576107308719782e-05,
-      "loss": 1.1426,
+      "learning_rate": 1.9538067291349154e-05,
+      "loss": 0.9786,
       "step": 4378
     },
     {
-      "epoch": 0.12026585372552251,
+      "epoch": 0.1242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.957585103202843e-05,
-      "loss": 1.0631,
+      "learning_rate": 1.953779114079517e-05,
+      "loss": 1.0057,
       "step": 4379
     },
     {
-      "epoch": 0.12029331795336574,
+      "epoch": 0.12429057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 1.9575594679569238e-05,
-      "loss": 1.093,
+      "learning_rate": 1.9537514909675034e-05,
+      "loss": 1.0663,
       "step": 4380
     },
     {
-      "epoch": 0.12032078218120898,
+      "epoch": 0.1243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.957533825134423e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.9537238597991082e-05,
+      "loss": 1.1024,
       "step": 4381
     },
     {
-      "epoch": 0.12034824640905221,
+      "epoch": 0.12434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9575081747355438e-05,
-      "loss": 1.153,
+      "learning_rate": 1.9536962205745647e-05,
+      "loss": 1.029,
       "step": 4382
     },
     {
-      "epoch": 0.12037571063689545,
+      "epoch": 0.12437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.957482516760489e-05,
-      "loss": 1.1398,
+      "learning_rate": 1.953668573294106e-05,
+      "loss": 1.104,
       "step": 4383
     },
     {
-      "epoch": 0.12040317486473867,
+      "epoch": 0.12440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.9574568512094615e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.953640917957966e-05,
+      "loss": 1.1447,
       "step": 4384
     },
     {
-      "epoch": 0.1204306390925819,
+      "epoch": 0.12443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9574311780826643e-05,
-      "loss": 1.0594,
+      "learning_rate": 1.953613254566379e-05,
+      "loss": 0.9972,
       "step": 4385
     },
     {
-      "epoch": 0.12045810332042514,
+      "epoch": 0.12446083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.9574054973803012e-05,
-      "loss": 1.028,
+      "learning_rate": 1.9535855831195776e-05,
+      "loss": 1.043,
       "step": 4386
     },
     {
-      "epoch": 0.12048556754826838,
+      "epoch": 0.12448921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 1.957379809102575e-05,
-      "loss": 1.1104,
+      "learning_rate": 1.9535579036177955e-05,
+      "loss": 1.0994,
       "step": 4387
     },
     {
-      "epoch": 0.12051303177611161,
+      "epoch": 0.12451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.9573541132496885e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.9535302160612674e-05,
+      "loss": 1.0052,
       "step": 4388
     },
     {
-      "epoch": 0.12054049600395485,
+      "epoch": 0.12454597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 1.9573284098218458e-05,
-      "loss": 1.0799,
+      "learning_rate": 1.9535025204502265e-05,
+      "loss": 1.0405,
       "step": 4389
     },
     {
-      "epoch": 0.12056796023179808,
+      "epoch": 0.12457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.9573026988192497e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.9534748167849068e-05,
+      "loss": 1.122,
       "step": 4390
     },
     {
-      "epoch": 0.12059542445964132,
+      "epoch": 0.12460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9572769802421035e-05,
-      "loss": 0.9864,
+      "learning_rate": 1.9534471050655428e-05,
+      "loss": 0.9954,
       "step": 4391
     },
     {
-      "epoch": 0.12062288868748455,
+      "epoch": 0.1246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9572512540906118e-05,
-      "loss": 1.1547,
+      "learning_rate": 1.9534193852923683e-05,
+      "loss": 0.931,
       "step": 4392
     },
     {
-      "epoch": 0.12065035291532779,
+      "epoch": 0.1246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9572255203649766e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.953391657465617e-05,
+      "loss": 1.14,
       "step": 4393
     },
     {
-      "epoch": 0.12067781714317102,
+      "epoch": 0.12468785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9571997790654028e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.953363921585524e-05,
+      "loss": 1.0948,
       "step": 4394
     },
     {
-      "epoch": 0.12070528137101426,
+      "epoch": 0.12471623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.957174030192093e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.9533361776523226e-05,
+      "loss": 1.0461,
       "step": 4395
     },
     {
-      "epoch": 0.1207327455988575,
+      "epoch": 0.12474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.957148273745252e-05,
-      "loss": 1.0709,
+      "learning_rate": 1.953308425666248e-05,
+      "loss": 1.0425,
       "step": 4396
     },
     {
-      "epoch": 0.12076020982670073,
+      "epoch": 0.12477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.957122509725083e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.953280665627534e-05,
+      "loss": 1.1243,
       "step": 4397
     },
     {
-      "epoch": 0.12078767405454395,
+      "epoch": 0.12480136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.95709673813179e-05,
-      "loss": 1.0914,
+      "learning_rate": 1.9532528975364154e-05,
+      "loss": 1.056,
       "step": 4398
     },
     {
-      "epoch": 0.12081513828238719,
+      "epoch": 0.12482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.957070958965577e-05,
-      "loss": 1.1111,
+      "learning_rate": 1.953225121393127e-05,
+      "loss": 1.0702,
       "step": 4399
     },
     {
-      "epoch": 0.12084260251023042,
+      "epoch": 0.12485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.9570451722266472e-05,
-      "loss": 1.1145,
+      "learning_rate": 1.953197337197903e-05,
+      "loss": 1.0561,
       "step": 4400
     },
     {
-      "epoch": 0.12087006673807366,
+      "epoch": 0.12488649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 1.9570193779152057e-05,
-      "loss": 1.0711,
+      "learning_rate": 1.953169544950978e-05,
+      "loss": 1.0368,
       "step": 4401
     },
     {
-      "epoch": 0.12089753096591689,
+      "epoch": 0.12491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.956993576031456e-05,
-      "loss": 1.024,
+      "learning_rate": 1.9531417446525873e-05,
+      "loss": 0.8981,
       "step": 4402
     },
     {
-      "epoch": 0.12092499519376013,
+      "epoch": 0.12494324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.956967766575603e-05,
-      "loss": 1.07,
+      "learning_rate": 1.953113936302965e-05,
+      "loss": 1.1171,
       "step": 4403
     },
     {
-      "epoch": 0.12095245942160336,
+      "epoch": 0.12497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.9569419495478497e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.9530861199023472e-05,
+      "loss": 1.0072,
       "step": 4404
     },
     {
-      "epoch": 0.1209799236494466,
+      "epoch": 0.125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9569161249484013e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.9530582954509675e-05,
+      "loss": 0.9988,
       "step": 4405
     },
     {
-      "epoch": 0.12100738787728983,
+      "epoch": 0.1250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.9568902927774615e-05,
-      "loss": 0.996,
+      "learning_rate": 1.953030462949062e-05,
+      "loss": 0.9577,
       "step": 4406
     },
     {
-      "epoch": 0.12103485210513307,
+      "epoch": 0.12505675368898977,
       "grad_norm": 0.0,
-      "learning_rate": 1.9568644530352354e-05,
-      "loss": 1.0911,
+      "learning_rate": 1.953002622396865e-05,
+      "loss": 1.0273,
       "step": 4407
     },
     {
-      "epoch": 0.1210623163329763,
+      "epoch": 0.1250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.956838605721927e-05,
-      "loss": 1.1286,
+      "learning_rate": 1.9529747737946117e-05,
+      "loss": 1.1085,
       "step": 4408
     },
     {
-      "epoch": 0.12108978056081954,
+      "epoch": 0.12511350737797958,
       "grad_norm": 0.0,
-      "learning_rate": 1.956812750837741e-05,
-      "loss": 1.1232,
+      "learning_rate": 1.9529469171425383e-05,
+      "loss": 1.1099,
       "step": 4409
     },
     {
-      "epoch": 0.12111724478866277,
+      "epoch": 0.12514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.956786888382882e-05,
-      "loss": 1.0878,
+      "learning_rate": 1.952919052440879e-05,
+      "loss": 1.0776,
       "step": 4410
     },
     {
-      "epoch": 0.121144709016506,
+      "epoch": 0.12517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.956761018357554e-05,
-      "loss": 1.0303,
+      "learning_rate": 1.95289117968987e-05,
+      "loss": 1.0391,
       "step": 4411
     },
     {
-      "epoch": 0.12117217324434923,
+      "epoch": 0.12519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.9567351407619625e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.952863298889746e-05,
+      "loss": 1.0793,
       "step": 4412
     },
     {
-      "epoch": 0.12119963747219246,
+      "epoch": 0.12522701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.956709255596312e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.9528354100407433e-05,
+      "loss": 0.8961,
       "step": 4413
     },
     {
-      "epoch": 0.1212271017000357,
+      "epoch": 0.12525539160045404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9566833628608074e-05,
-      "loss": 1.1711,
+      "learning_rate": 1.9528075131430967e-05,
+      "loss": 1.1005,
       "step": 4414
     },
     {
-      "epoch": 0.12125456592787894,
+      "epoch": 0.12528376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.956657462555653e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.9527796081970423e-05,
+      "loss": 1.135,
       "step": 4415
     },
     {
-      "epoch": 0.12128203015572217,
+      "epoch": 0.12531214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.9566315546810546e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.9527516952028156e-05,
+      "loss": 1.0354,
       "step": 4416
     },
     {
-      "epoch": 0.1213094943835654,
+      "epoch": 0.1253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.956605639237217e-05,
-      "loss": 1.1166,
+      "learning_rate": 1.9527237741606527e-05,
+      "loss": 1.0938,
       "step": 4417
     },
     {
-      "epoch": 0.12133695861140864,
+      "epoch": 0.1253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9565797162243445e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.952695845070789e-05,
+      "loss": 1.0822,
       "step": 4418
     },
     {
-      "epoch": 0.12136442283925188,
+      "epoch": 0.12539727582292848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9565537856426426e-05,
-      "loss": 1.08,
+      "learning_rate": 1.952667907933461e-05,
+      "loss": 1.0072,
       "step": 4419
     },
     {
-      "epoch": 0.12139188706709511,
+      "epoch": 0.1254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.956527847492317e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.952639962748904e-05,
+      "loss": 0.9518,
       "step": 4420
     },
     {
-      "epoch": 0.12141935129493835,
+      "epoch": 0.12545402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 1.9565019017735727e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.9526120095173548e-05,
+      "loss": 1.1264,
       "step": 4421
     },
     {
-      "epoch": 0.12144681552278158,
+      "epoch": 0.12548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.9564759484866146e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.952584048239049e-05,
+      "loss": 1.0065,
       "step": 4422
     },
     {
-      "epoch": 0.12147427975062482,
+      "epoch": 0.12551078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 1.956449987631648e-05,
-      "loss": 1.1853,
+      "learning_rate": 1.952556078914223e-05,
+      "loss": 1.1197,
       "step": 4423
     },
     {
-      "epoch": 0.12150174397846805,
+      "epoch": 0.12553916004540294,
       "grad_norm": 0.0,
-      "learning_rate": 1.956424019208879e-05,
-      "loss": 1.0947,
+      "learning_rate": 1.9525281015431127e-05,
+      "loss": 1.0114,
       "step": 4424
     },
     {
-      "epoch": 0.12152920820631127,
+      "epoch": 0.12556753688989786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9563980432185127e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.952500116125955e-05,
+      "loss": 1.0599,
       "step": 4425
     },
     {
-      "epoch": 0.12155667243415451,
+      "epoch": 0.12559591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9563720596607546e-05,
-      "loss": 1.037,
+      "learning_rate": 1.9524721226629856e-05,
+      "loss": 1.0851,
       "step": 4426
     },
     {
-      "epoch": 0.12158413666199774,
+      "epoch": 0.12562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9563460685358102e-05,
-      "loss": 1.0646,
+      "learning_rate": 1.9524441211544416e-05,
+      "loss": 0.949,
       "step": 4427
     },
     {
-      "epoch": 0.12161160088984098,
+      "epoch": 0.12565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9563200698438854e-05,
-      "loss": 1.134,
+      "learning_rate": 1.9524161116005596e-05,
+      "loss": 0.9689,
       "step": 4428
     },
     {
-      "epoch": 0.12163906511768421,
+      "epoch": 0.1256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.9562940635851856e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.9523880940015755e-05,
+      "loss": 1.1024,
       "step": 4429
     },
     {
-      "epoch": 0.12166652934552745,
+      "epoch": 0.1257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.9562680497599168e-05,
-      "loss": 1.0148,
+      "learning_rate": 1.9523600683577263e-05,
+      "loss": 1.0253,
       "step": 4430
     },
     {
-      "epoch": 0.12169399357337068,
+      "epoch": 0.1257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9562420283682846e-05,
-      "loss": 1.1674,
+      "learning_rate": 1.952332034669249e-05,
+      "loss": 0.9807,
       "step": 4431
     },
     {
-      "epoch": 0.12172145780121392,
+      "epoch": 0.1257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.9562159994104953e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.95230399293638e-05,
+      "loss": 1.0428,
       "step": 4432
     },
     {
-      "epoch": 0.12174892202905715,
+      "epoch": 0.12579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9561899628867544e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.9522759431593566e-05,
+      "loss": 0.999,
       "step": 4433
     },
     {
-      "epoch": 0.12177638625690039,
+      "epoch": 0.12582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.9561639187972684e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.9522478853384154e-05,
+      "loss": 1.0584,
       "step": 4434
     },
     {
-      "epoch": 0.12180385048474363,
+      "epoch": 0.12585130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.9561378671422426e-05,
-      "loss": 1.0493,
+      "learning_rate": 1.9522198194737935e-05,
+      "loss": 1.0218,
       "step": 4435
     },
     {
-      "epoch": 0.12183131471258686,
+      "epoch": 0.12587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.9561118079218842e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.952191745565728e-05,
+      "loss": 1.0748,
       "step": 4436
     },
     {
-      "epoch": 0.1218587789404301,
+      "epoch": 0.12590805902383656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9560857411363984e-05,
-      "loss": 0.981,
+      "learning_rate": 1.952163663614456e-05,
+      "loss": 1.0215,
       "step": 4437
     },
     {
-      "epoch": 0.12188624316827332,
+      "epoch": 0.12593643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 1.956059666785992e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.9521355736202148e-05,
+      "loss": 1.0691,
       "step": 4438
     },
     {
-      "epoch": 0.12191370739611655,
+      "epoch": 0.12596481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.9560335848708713e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.9521074755832416e-05,
+      "loss": 1.0283,
       "step": 4439
     },
     {
-      "epoch": 0.12194117162395979,
+      "epoch": 0.12599318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 1.9560074953912423e-05,
-      "loss": 0.975,
+      "learning_rate": 1.952079369503774e-05,
+      "loss": 1.0756,
       "step": 4440
     },
     {
-      "epoch": 0.12196863585180302,
+      "epoch": 0.1260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.9559813983473117e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9520512553820485e-05,
+      "loss": 1.1709,
       "step": 4441
     },
     {
-      "epoch": 0.12199610007964626,
+      "epoch": 0.126049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9559552937392864e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.9520231332183037e-05,
+      "loss": 1.1193,
       "step": 4442
     },
     {
-      "epoch": 0.1220235643074895,
+      "epoch": 0.1260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.955929181567372e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.9519950030127765e-05,
+      "loss": 0.8883,
       "step": 4443
     },
     {
-      "epoch": 0.12205102853533273,
+      "epoch": 0.1261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9559030618317757e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.951966864765705e-05,
+      "loss": 0.9745,
       "step": 4444
     },
     {
-      "epoch": 0.12207849276317596,
+      "epoch": 0.1261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.9558769345327044e-05,
-      "loss": 1.1807,
+      "learning_rate": 1.9519387184773264e-05,
+      "loss": 0.9839,
       "step": 4445
     },
     {
-      "epoch": 0.1221059569910192,
+      "epoch": 0.12616345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 1.955850799670364e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.9519105641478788e-05,
+      "loss": 1.0721,
       "step": 4446
     },
     {
-      "epoch": 0.12213342121886243,
+      "epoch": 0.12619182746878546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9558246572449624e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.9518824017775995e-05,
+      "loss": 0.9609,
       "step": 4447
     },
     {
-      "epoch": 0.12216088544670567,
+      "epoch": 0.12622020431328038,
       "grad_norm": 0.0,
-      "learning_rate": 1.9557985072567055e-05,
-      "loss": 1.0423,
+      "learning_rate": 1.9518542313667272e-05,
+      "loss": 1.0299,
       "step": 4448
     },
     {
-      "epoch": 0.1221883496745489,
+      "epoch": 0.12624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.955772349705801e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.9518260529154994e-05,
+      "loss": 1.1074,
       "step": 4449
     },
     {
-      "epoch": 0.12221581390239214,
+      "epoch": 0.12627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.9557461845924547e-05,
-      "loss": 1.0544,
+      "learning_rate": 1.9517978664241538e-05,
+      "loss": 1.0185,
       "step": 4450
     },
     {
-      "epoch": 0.12224327813023537,
+      "epoch": 0.12630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.955720011916875e-05,
-      "loss": 1.0551,
+      "learning_rate": 1.951769671892929e-05,
+      "loss": 1.0052,
       "step": 4451
     },
     {
-      "epoch": 0.1222707423580786,
+      "epoch": 0.12633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.955693831679268e-05,
-      "loss": 1.1351,
+      "learning_rate": 1.951741469322063e-05,
+      "loss": 0.9365,
       "step": 4452
     },
     {
-      "epoch": 0.12229820658592183,
+      "epoch": 0.1263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.955667643879841e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.951713258711794e-05,
+      "loss": 1.0705,
       "step": 4453
     },
     {
-      "epoch": 0.12232567081376507,
+      "epoch": 0.12639046538024973,
       "grad_norm": 0.0,
-      "learning_rate": 1.955641448518802e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.9516850400623605e-05,
+      "loss": 0.995,
       "step": 4454
     },
     {
-      "epoch": 0.1223531350416083,
+      "epoch": 0.12641884222474462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9556152455963575e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.9516568133740005e-05,
+      "loss": 1.105,
       "step": 4455
     },
     {
-      "epoch": 0.12238059926945154,
+      "epoch": 0.1264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.955589035112715e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.9516285786469526e-05,
+      "loss": 0.8887,
       "step": 4456
     },
     {
-      "epoch": 0.12240806349729477,
+      "epoch": 0.1264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.9555628170680818e-05,
-      "loss": 1.037,
+      "learning_rate": 1.9516003358814553e-05,
+      "loss": 1.1491,
       "step": 4457
     },
     {
-      "epoch": 0.12243552772513801,
+      "epoch": 0.12650397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 1.9555365914626658e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.9515720850777475e-05,
+      "loss": 0.9475,
       "step": 4458
     },
     {
-      "epoch": 0.12246299195298124,
+      "epoch": 0.12653234960272416,
       "grad_norm": 0.0,
-      "learning_rate": 1.9555103582966733e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.9515438262360673e-05,
+      "loss": 0.9865,
       "step": 4459
     },
     {
-      "epoch": 0.12249045618082448,
+      "epoch": 0.12656072644721908,
       "grad_norm": 0.0,
-      "learning_rate": 1.9554841175703134e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.9515155593566536e-05,
+      "loss": 1.0779,
       "step": 4460
     },
     {
-      "epoch": 0.12251792040866771,
+      "epoch": 0.12658910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 1.955457869283793e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.951487284439745e-05,
+      "loss": 0.9311,
       "step": 4461
     },
     {
-      "epoch": 0.12254538463651095,
+      "epoch": 0.12661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.9554316134373196e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.951459001485581e-05,
+      "loss": 1.0508,
       "step": 4462
     },
     {
-      "epoch": 0.12257284886435418,
+      "epoch": 0.12664585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9554053500311013e-05,
-      "loss": 1.0429,
+      "learning_rate": 1.9514307104943997e-05,
+      "loss": 0.8262,
       "step": 4463
     },
     {
-      "epoch": 0.12260031309219742,
+      "epoch": 0.12667423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.9553790790653457e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.9514024114664404e-05,
+      "loss": 1.0121,
       "step": 4464
     },
     {
-      "epoch": 0.12262777732004064,
+      "epoch": 0.12670261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9553528005402608e-05,
-      "loss": 1.0127,
+      "learning_rate": 1.9513741044019423e-05,
+      "loss": 1.1328,
       "step": 4465
     },
     {
-      "epoch": 0.12265524154788388,
+      "epoch": 0.12673098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.9553265144560544e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.9513457893011445e-05,
+      "loss": 1.031,
       "step": 4466
     },
     {
-      "epoch": 0.12268270577572711,
+      "epoch": 0.12675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 1.9553002208129347e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.9513174661642857e-05,
+      "loss": 1.0106,
       "step": 4467
     },
     {
-      "epoch": 0.12271017000357035,
+      "epoch": 0.1267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.9552739196111095e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.9512891349916056e-05,
+      "loss": 0.8683,
       "step": 4468
     },
     {
-      "epoch": 0.12273763423141358,
+      "epoch": 0.1268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9552476108507872e-05,
-      "loss": 1.0279,
+      "learning_rate": 1.9512607957833438e-05,
+      "loss": 1.0208,
       "step": 4469
     },
     {
-      "epoch": 0.12276509845925682,
+      "epoch": 0.12684449489216798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9552212945321754e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.951232448539739e-05,
+      "loss": 1.1266,
       "step": 4470
     },
     {
-      "epoch": 0.12279256268710005,
+      "epoch": 0.1268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9551949706554832e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.9512040932610307e-05,
+      "loss": 1.0003,
       "step": 4471
     },
     {
-      "epoch": 0.12282002691494329,
+      "epoch": 0.12690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.955168639220918e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.951175729947459e-05,
+      "loss": 0.9303,
       "step": 4472
     },
     {
-      "epoch": 0.12284749114278652,
+      "epoch": 0.12692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.9551423002286887e-05,
-      "loss": 1.0598,
+      "learning_rate": 1.951147358599263e-05,
+      "loss": 1.1216,
       "step": 4473
     },
     {
-      "epoch": 0.12287495537062976,
+      "epoch": 0.12695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9551159536790035e-05,
-      "loss": 1.1426,
+      "learning_rate": 1.9511189792166826e-05,
+      "loss": 0.9809,
       "step": 4474
     },
     {
-      "epoch": 0.12290241959847299,
+      "epoch": 0.12698637911464244,
       "grad_norm": 0.0,
-      "learning_rate": 1.955089599572071e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.9510905917999573e-05,
+      "loss": 0.973,
       "step": 4475
     },
     {
-      "epoch": 0.12292988382631623,
+      "epoch": 0.12701475595913733,
       "grad_norm": 0.0,
-      "learning_rate": 1.9550632379080993e-05,
-      "loss": 1.1335,
+      "learning_rate": 1.9510621963493272e-05,
+      "loss": 0.9947,
       "step": 4476
     },
     {
-      "epoch": 0.12295734805415946,
+      "epoch": 0.12704313280363225,
       "grad_norm": 0.0,
-      "learning_rate": 1.9550368686872977e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.9510337928650318e-05,
+      "loss": 1.0152,
       "step": 4477
     },
     {
-      "epoch": 0.1229848122820027,
+      "epoch": 0.12707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.955010491909874e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.9510053813473115e-05,
+      "loss": 1.1266,
       "step": 4478
     },
     {
-      "epoch": 0.12301227650984592,
+      "epoch": 0.12709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9549841075760378e-05,
-      "loss": 1.0202,
+      "learning_rate": 1.9509769617964057e-05,
+      "loss": 1.0918,
       "step": 4479
     },
     {
-      "epoch": 0.12303974073768915,
+      "epoch": 0.1271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.9549577156859972e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.9509485342125548e-05,
+      "loss": 1.0941,
       "step": 4480
     },
     {
-      "epoch": 0.12306720496553239,
+      "epoch": 0.1271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9549313162399613e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.950920098595999e-05,
+      "loss": 0.9063,
       "step": 4481
     },
     {
-      "epoch": 0.12309466919337562,
+      "epoch": 0.12718501702610668,
       "grad_norm": 0.0,
-      "learning_rate": 1.9549049092381387e-05,
-      "loss": 1.0883,
+      "learning_rate": 1.9508916549469784e-05,
+      "loss": 1.0302,
       "step": 4482
     },
     {
-      "epoch": 0.12312213342121886,
+      "epoch": 0.1272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.9548784946807385e-05,
-      "loss": 1.02,
+      "learning_rate": 1.9508632032657327e-05,
+      "loss": 1.1306,
       "step": 4483
     },
     {
-      "epoch": 0.1231495976490621,
+      "epoch": 0.12724177071509649,
       "grad_norm": 0.0,
-      "learning_rate": 1.95485207256797e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.9508347435525035e-05,
+      "loss": 1.0192,
       "step": 4484
     },
     {
-      "epoch": 0.12317706187690533,
+      "epoch": 0.12727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.9548256429000423e-05,
-      "loss": 1.013,
+      "learning_rate": 1.9508062758075304e-05,
+      "loss": 0.9943,
       "step": 4485
     },
     {
-      "epoch": 0.12320452610474857,
+      "epoch": 0.12729852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.954799205677164e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.9507778000310535e-05,
+      "loss": 1.0387,
       "step": 4486
     },
     {
-      "epoch": 0.1232319903325918,
+      "epoch": 0.12732690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.9547727608995445e-05,
-      "loss": 0.886,
+      "learning_rate": 1.950749316223314e-05,
+      "loss": 1.0338,
       "step": 4487
     },
     {
-      "epoch": 0.12325945456043504,
+      "epoch": 0.12735527809307606,
       "grad_norm": 0.0,
-      "learning_rate": 1.954746308567393e-05,
-      "loss": 1.0419,
+      "learning_rate": 1.9507208243845522e-05,
+      "loss": 0.9158,
       "step": 4488
     },
     {
-      "epoch": 0.12328691878827827,
+      "epoch": 0.12738365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 1.954719848680919e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.9506923245150092e-05,
+      "loss": 0.9373,
       "step": 4489
     },
     {
-      "epoch": 0.1233143830161215,
+      "epoch": 0.12741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.9546933812403318e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.950663816614925e-05,
+      "loss": 1.188,
       "step": 4490
     },
     {
-      "epoch": 0.12334184724396474,
+      "epoch": 0.12744040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 1.954666906245841e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.950635300684541e-05,
+      "loss": 0.9934,
       "step": 4491
     },
     {
-      "epoch": 0.12336931147180798,
+      "epoch": 0.1274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9546404236976553e-05,
-      "loss": 1.0299,
+      "learning_rate": 1.950606776724098e-05,
+      "loss": 0.9985,
       "step": 4492
     },
     {
-      "epoch": 0.1233967756996512,
+      "epoch": 0.1274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.954613933595985e-05,
-      "loss": 1.1652,
+      "learning_rate": 1.9505782447338367e-05,
+      "loss": 1.0064,
       "step": 4493
     },
     {
-      "epoch": 0.12342423992749443,
+      "epoch": 0.12752553916004541,
       "grad_norm": 0.0,
-      "learning_rate": 1.9545874359410392e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.9505497047139982e-05,
+      "loss": 1.1032,
       "step": 4494
     },
     {
-      "epoch": 0.12345170415533767,
+      "epoch": 0.1275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.954560930733028e-05,
-      "loss": 1.0502,
+      "learning_rate": 1.9505211566648238e-05,
+      "loss": 1.0414,
       "step": 4495
     },
     {
-      "epoch": 0.1234791683831809,
+      "epoch": 0.1275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.9545344179721612e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.9504926005865542e-05,
+      "loss": 0.9835,
       "step": 4496
     },
     {
-      "epoch": 0.12350663261102414,
+      "epoch": 0.12761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.954507897658648e-05,
-      "loss": 0.9389,
+      "learning_rate": 1.950464036479431e-05,
+      "loss": 1.0311,
       "step": 4497
     },
     {
-      "epoch": 0.12353409683886737,
+      "epoch": 0.12763904653802496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9544813697926988e-05,
-      "loss": 1.1989,
+      "learning_rate": 1.950435464343695e-05,
+      "loss": 1.0563,
       "step": 4498
     },
     {
-      "epoch": 0.12356156106671061,
+      "epoch": 0.12766742338251985,
       "grad_norm": 0.0,
-      "learning_rate": 1.9544548343745232e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.9504068841795886e-05,
+      "loss": 1.0758,
       "step": 4499
     },
     {
-      "epoch": 0.12358902529455384,
+      "epoch": 0.12769580022701477,
       "grad_norm": 0.0,
-      "learning_rate": 1.9544282914043312e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.950378295987352e-05,
+      "loss": 0.9931,
       "step": 4500
     },
     {
-      "epoch": 0.12361648952239708,
+      "epoch": 0.12772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.954401740882333e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.9503496997672275e-05,
+      "loss": 1.0852,
       "step": 4501
     },
     {
-      "epoch": 0.12364395375024032,
+      "epoch": 0.12775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9543751828087382e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.950321095519456e-05,
+      "loss": 1.1292,
       "step": 4502
     },
     {
-      "epoch": 0.12367141797808355,
+      "epoch": 0.12778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.9543486171837576e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.9502924832442796e-05,
+      "loss": 1.0229,
       "step": 4503
     },
     {
-      "epoch": 0.12369888220592679,
+      "epoch": 0.12780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.954322044007601e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.95026386294194e-05,
+      "loss": 1.1027,
       "step": 4504
     },
     {
-      "epoch": 0.12372634643377002,
+      "epoch": 0.12783768444948923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9542954632804786e-05,
-      "loss": 1.0746,
+      "learning_rate": 1.950235234612679e-05,
+      "loss": 1.1139,
       "step": 4505
     },
     {
-      "epoch": 0.12375381066161324,
+      "epoch": 0.12786606129398412,
       "grad_norm": 0.0,
-      "learning_rate": 1.954268875002601e-05,
-      "loss": 1.1584,
+      "learning_rate": 1.9502065982567382e-05,
+      "loss": 1.0578,
       "step": 4506
     },
     {
-      "epoch": 0.12378127488945648,
+      "epoch": 0.127894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9542422791741782e-05,
-      "loss": 1.1458,
+      "learning_rate": 1.9501779538743593e-05,
+      "loss": 1.0697,
       "step": 4507
     },
     {
-      "epoch": 0.12380873911729971,
+      "epoch": 0.1279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.954215675795421e-05,
-      "loss": 1.1152,
+      "learning_rate": 1.9501493014657848e-05,
+      "loss": 1.0737,
       "step": 4508
     },
     {
-      "epoch": 0.12383620334514295,
+      "epoch": 0.12795119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 1.95418906486654e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.950120641031256e-05,
+      "loss": 1.1315,
       "step": 4509
     },
     {
-      "epoch": 0.12386366757298618,
+      "epoch": 0.12797956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 1.954162446387745e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.950091972571016e-05,
+      "loss": 1.0094,
       "step": 4510
     },
     {
-      "epoch": 0.12389113180082942,
+      "epoch": 0.12800794551645858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9541358203592474e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.950063296085306e-05,
+      "loss": 1.1406,
       "step": 4511
     },
     {
-      "epoch": 0.12391859602867265,
+      "epoch": 0.12803632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.954109186781257e-05,
-      "loss": 1.1133,
+      "learning_rate": 1.9500346115743687e-05,
+      "loss": 0.895,
       "step": 4512
     },
     {
-      "epoch": 0.12394606025651589,
+      "epoch": 0.12806469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.954082545653986e-05,
-      "loss": 1.1044,
+      "learning_rate": 1.9500059190384463e-05,
+      "loss": 0.9243,
       "step": 4513
     },
     {
-      "epoch": 0.12397352448435912,
+      "epoch": 0.12809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.9540558969776436e-05,
-      "loss": 1.0362,
+      "learning_rate": 1.9499772184777814e-05,
+      "loss": 0.9912,
       "step": 4514
     },
     {
-      "epoch": 0.12400098871220236,
+      "epoch": 0.12812145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 1.954029240752442e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.949948509892616e-05,
+      "loss": 1.0574,
       "step": 4515
     },
     {
-      "epoch": 0.1240284529400456,
+      "epoch": 0.12814982973893302,
       "grad_norm": 0.0,
-      "learning_rate": 1.9540025769785912e-05,
-      "loss": 1.0914,
+      "learning_rate": 1.949919793283193e-05,
+      "loss": 0.9457,
       "step": 4516
     },
     {
-      "epoch": 0.12405591716788883,
+      "epoch": 0.12817820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9539759056563024e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.949891068649755e-05,
+      "loss": 0.9875,
       "step": 4517
     },
     {
-      "epoch": 0.12408338139573206,
+      "epoch": 0.12820658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.953949226785787e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.949862335992544e-05,
+      "loss": 1.0781,
       "step": 4518
     },
     {
-      "epoch": 0.1241108456235753,
+      "epoch": 0.1282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9539225403672554e-05,
-      "loss": 1.004,
+      "learning_rate": 1.9498335953118032e-05,
+      "loss": 1.1153,
       "step": 4519
     },
     {
-      "epoch": 0.12413830985141852,
+      "epoch": 0.1282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9538958464009194e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.9498048466077755e-05,
+      "loss": 0.9688,
       "step": 4520
     },
     {
-      "epoch": 0.12416577407926176,
+      "epoch": 0.12829171396140748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9538691448869896e-05,
-      "loss": 1.058,
+      "learning_rate": 1.9497760898807035e-05,
+      "loss": 1.0112,
       "step": 4521
     },
     {
-      "epoch": 0.12419323830710499,
+      "epoch": 0.12832009080590237,
       "grad_norm": 0.0,
-      "learning_rate": 1.953842435825678e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.9497473251308303e-05,
+      "loss": 0.9839,
       "step": 4522
     },
     {
-      "epoch": 0.12422070253494823,
+      "epoch": 0.12834846765039729,
       "grad_norm": 0.0,
-      "learning_rate": 1.953815719217196e-05,
-      "loss": 1.0184,
+      "learning_rate": 1.9497185523583986e-05,
+      "loss": 1.0644,
       "step": 4523
     },
     {
-      "epoch": 0.12424816676279146,
+      "epoch": 0.12837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.953788995061754e-05,
-      "loss": 1.0784,
+      "learning_rate": 1.9496897715636516e-05,
+      "loss": 1.0759,
       "step": 4524
     },
     {
-      "epoch": 0.1242756309906347,
+      "epoch": 0.12840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9537622633595638e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.9496609827468323e-05,
+      "loss": 1.0115,
       "step": 4525
     },
     {
-      "epoch": 0.12430309521847793,
+      "epoch": 0.12843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.9537355241108373e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.9496321859081844e-05,
+      "loss": 1.051,
       "step": 4526
     },
     {
-      "epoch": 0.12433055944632117,
+      "epoch": 0.12846197502837683,
       "grad_norm": 0.0,
-      "learning_rate": 1.9537087773157858e-05,
-      "loss": 1.1466,
+      "learning_rate": 1.9496033810479506e-05,
+      "loss": 1.0855,
       "step": 4527
     },
     {
-      "epoch": 0.1243580236741644,
+      "epoch": 0.12849035187287175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9536820229746215e-05,
-      "loss": 0.9385,
+      "learning_rate": 1.9495745681663737e-05,
+      "loss": 0.989,
       "step": 4528
     },
     {
-      "epoch": 0.12438548790200764,
+      "epoch": 0.12851872871736664,
       "grad_norm": 0.0,
-      "learning_rate": 1.953655261087555e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.9495457472636984e-05,
+      "loss": 0.9203,
       "step": 4529
     },
     {
-      "epoch": 0.12441295212985087,
+      "epoch": 0.12854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9536284916547993e-05,
-      "loss": 1.106,
+      "learning_rate": 1.9495169183401676e-05,
+      "loss": 0.9284,
       "step": 4530
     },
     {
-      "epoch": 0.12444041635769411,
+      "epoch": 0.1285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.953601714676565e-05,
+      "learning_rate": 1.9494880813960245e-05,
       "loss": 1.1149,
       "step": 4531
     },
     {
-      "epoch": 0.12446788058553734,
+      "epoch": 0.1286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.953574930153065e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.9494592364315125e-05,
+      "loss": 1.0198,
       "step": 4532
     },
     {
-      "epoch": 0.12449534481338057,
+      "epoch": 0.1286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9535481380845106e-05,
-      "loss": 1.0246,
+      "learning_rate": 1.949430383446876e-05,
+      "loss": 1.079,
       "step": 4533
     },
     {
-      "epoch": 0.1245228090412238,
+      "epoch": 0.1286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.9535213384711138e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.9494015224423583e-05,
+      "loss": 1.0387,
       "step": 4534
     },
     {
-      "epoch": 0.12455027326906704,
+      "epoch": 0.128688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.953494531313087e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.9493726534182037e-05,
+      "loss": 0.9759,
       "step": 4535
     },
     {
-      "epoch": 0.12457773749691027,
+      "epoch": 0.12871736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 1.953467716610642e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.949343776374655e-05,
+      "loss": 0.9684,
       "step": 4536
     },
     {
-      "epoch": 0.1246052017247535,
+      "epoch": 0.12874574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.953440894363991e-05,
-      "loss": 1.1163,
+      "learning_rate": 1.9493148913119572e-05,
+      "loss": 1.0074,
       "step": 4537
     },
     {
-      "epoch": 0.12463266595259674,
+      "epoch": 0.12877412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.9534140645733466e-05,
-      "loss": 1.046,
+      "learning_rate": 1.9492859982303535e-05,
+      "loss": 1.131,
       "step": 4538
     },
     {
-      "epoch": 0.12466013018043998,
+      "epoch": 0.12880249716231554,
       "grad_norm": 0.0,
-      "learning_rate": 1.9533872272389204e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.9492570971300885e-05,
+      "loss": 1.0976,
       "step": 4539
     },
     {
-      "epoch": 0.12468759440828321,
+      "epoch": 0.12883087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.9533603823609258e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.949228188011406e-05,
+      "loss": 1.0475,
       "step": 4540
     },
     {
-      "epoch": 0.12471505863612645,
+      "epoch": 0.12885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.953333529939574e-05,
-      "loss": 1.1705,
+      "learning_rate": 1.9491992708745502e-05,
+      "loss": 1.075,
       "step": 4541
     },
     {
-      "epoch": 0.12474252286396968,
+      "epoch": 0.12888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.9533066699750782e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9491703457197658e-05,
+      "loss": 0.9607,
       "step": 4542
     },
     {
-      "epoch": 0.12476998709181292,
+      "epoch": 0.12891600454029511,
       "grad_norm": 0.0,
-      "learning_rate": 1.9532798024676507e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.949141412547297e-05,
+      "loss": 1.0831,
       "step": 4543
     },
     {
-      "epoch": 0.12479745131965615,
+      "epoch": 0.12894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9532529274175043e-05,
-      "loss": 1.0898,
+      "learning_rate": 1.9491124713573873e-05,
+      "loss": 1.0429,
       "step": 4544
     },
     {
-      "epoch": 0.12482491554749939,
+      "epoch": 0.12897275822928492,
       "grad_norm": 0.0,
-      "learning_rate": 1.953226044824851e-05,
-      "loss": 1.022,
+      "learning_rate": 1.9490835221502825e-05,
+      "loss": 1.0658,
       "step": 4545
     },
     {
-      "epoch": 0.12485237977534262,
+      "epoch": 0.1290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9531991546899048e-05,
-      "loss": 1.053,
+      "learning_rate": 1.9490545649262262e-05,
+      "loss": 0.9835,
       "step": 4546
     },
     {
-      "epoch": 0.12487984400318584,
+      "epoch": 0.1290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9531722570128768e-05,
-      "loss": 1.0083,
+      "learning_rate": 1.9490255996854633e-05,
+      "loss": 1.0013,
       "step": 4547
     },
     {
-      "epoch": 0.12490730823102908,
+      "epoch": 0.12905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9531453517939813e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.948996626428239e-05,
+      "loss": 1.0755,
       "step": 4548
     },
     {
-      "epoch": 0.12493477245887231,
+      "epoch": 0.12908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.9531184390334303e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.948967645154797e-05,
+      "loss": 1.1118,
       "step": 4549
     },
     {
-      "epoch": 0.12496223668671555,
+      "epoch": 0.12911464245175935,
       "grad_norm": 0.0,
-      "learning_rate": 1.953091518731437e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.948938655865383e-05,
+      "loss": 1.0358,
       "step": 4550
     },
     {
-      "epoch": 0.12498970091455879,
+      "epoch": 0.12914301929625427,
       "grad_norm": 0.0,
-      "learning_rate": 1.9530645908882143e-05,
-      "loss": 1.005,
+      "learning_rate": 1.9489096585602413e-05,
+      "loss": 1.0435,
       "step": 4551
     },
     {
-      "epoch": 0.12501716514240202,
+      "epoch": 0.12917139614074916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9530376555039758e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.948880653239617e-05,
+      "loss": 1.057,
       "step": 4552
     },
     {
-      "epoch": 0.12504462937024527,
+      "epoch": 0.12919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.953010712578934e-05,
-      "loss": 1.0535,
+      "learning_rate": 1.9488516399037552e-05,
+      "loss": 1.0685,
       "step": 4553
     },
     {
-      "epoch": 0.1250720935980885,
+      "epoch": 0.12922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.952983762113302e-05,
-      "loss": 1.04,
+      "learning_rate": 1.948822618552901e-05,
+      "loss": 1.0414,
       "step": 4554
     },
     {
-      "epoch": 0.1250995578259317,
+      "epoch": 0.12925652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.952956804107294e-05,
-      "loss": 1.049,
+      "learning_rate": 1.9487935891872997e-05,
+      "loss": 0.9803,
       "step": 4555
     },
     {
-      "epoch": 0.12512702205377496,
+      "epoch": 0.1292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.9529298385611218e-05,
-      "loss": 1.0915,
+      "learning_rate": 1.948764551807196e-05,
+      "loss": 1.0276,
       "step": 4556
     },
     {
-      "epoch": 0.12515448628161818,
+      "epoch": 0.12931328036322362,
       "grad_norm": 0.0,
-      "learning_rate": 1.952902865475e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.9487355064128356e-05,
+      "loss": 1.0033,
       "step": 4557
     },
     {
-      "epoch": 0.12518195050946143,
+      "epoch": 0.1293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.9528758848491416e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.9487064530044636e-05,
+      "loss": 1.0996,
       "step": 4558
     },
     {
-      "epoch": 0.12520941473730465,
+      "epoch": 0.1293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.95284889668376e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.9486773915823258e-05,
+      "loss": 1.0008,
       "step": 4559
     },
     {
-      "epoch": 0.1252368789651479,
+      "epoch": 0.12939841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 1.952821900979069e-05,
-      "loss": 1.11,
+      "learning_rate": 1.948648322146667e-05,
+      "loss": 1.0949,
       "step": 4560
     },
     {
-      "epoch": 0.12526434319299112,
+      "epoch": 0.12942678774120317,
       "grad_norm": 0.0,
-      "learning_rate": 1.9527948977352816e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.9486192446977336e-05,
+      "loss": 0.9406,
       "step": 4561
     },
     {
-      "epoch": 0.12529180742083437,
+      "epoch": 0.12945516458569806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9527678869526127e-05,
-      "loss": 1.1233,
+      "learning_rate": 1.9485901592357708e-05,
+      "loss": 0.9581,
       "step": 4562
     },
     {
-      "epoch": 0.1253192716486776,
+      "epoch": 0.12948354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.9527408686312746e-05,
-      "loss": 1.1183,
+      "learning_rate": 1.948561065761024e-05,
+      "loss": 1.0338,
       "step": 4563
     },
     {
-      "epoch": 0.12534673587652084,
+      "epoch": 0.12951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9527138427714818e-05,
-      "loss": 1.088,
+      "learning_rate": 1.9485319642737396e-05,
+      "loss": 0.9832,
       "step": 4564
     },
     {
-      "epoch": 0.12537420010436406,
+      "epoch": 0.12954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.952686809373448e-05,
-      "loss": 1.1312,
+      "learning_rate": 1.9485028547741628e-05,
+      "loss": 1.1264,
       "step": 4565
     },
     {
-      "epoch": 0.1254016643322073,
+      "epoch": 0.12956867196367763,
       "grad_norm": 0.0,
-      "learning_rate": 1.952659768437387e-05,
-      "loss": 1.028,
+      "learning_rate": 1.9484737372625394e-05,
+      "loss": 1.0071,
       "step": 4566
     },
     {
-      "epoch": 0.12542912856005053,
+      "epoch": 0.12959704880817252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9526327199635132e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.9484446117391162e-05,
+      "loss": 1.1017,
       "step": 4567
     },
     {
-      "epoch": 0.12545659278789376,
+      "epoch": 0.12962542565266744,
       "grad_norm": 0.0,
-      "learning_rate": 1.95260566395204e-05,
-      "loss": 1.0809,
+      "learning_rate": 1.948415478204139e-05,
+      "loss": 1.0107,
       "step": 4568
     },
     {
-      "epoch": 0.125484057015737,
+      "epoch": 0.12965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.952578600403182e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.948386336657853e-05,
+      "loss": 1.0441,
       "step": 4569
     },
     {
-      "epoch": 0.12551152124358023,
+      "epoch": 0.1296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.952551529317153e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.9483571871005052e-05,
+      "loss": 1.0261,
       "step": 4570
     },
     {
-      "epoch": 0.12553898547142348,
+      "epoch": 0.1297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9525244506941675e-05,
-      "loss": 1.1351,
+      "learning_rate": 1.9483280295323416e-05,
+      "loss": 1.1154,
       "step": 4571
     },
     {
-      "epoch": 0.1255664496992667,
+      "epoch": 0.12973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.9524973645344394e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9482988639536086e-05,
+      "loss": 0.9446,
       "step": 4572
     },
     {
-      "epoch": 0.12559391392710995,
+      "epoch": 0.12976730987514187,
       "grad_norm": 0.0,
-      "learning_rate": 1.9524702708381835e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.9482696903645524e-05,
+      "loss": 1.069,
       "step": 4573
     },
     {
-      "epoch": 0.12562137815495317,
+      "epoch": 0.1297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.9524431696056135e-05,
-      "loss": 1.0907,
+      "learning_rate": 1.9482405087654194e-05,
+      "loss": 1.0801,
       "step": 4574
     },
     {
-      "epoch": 0.12564884238279642,
+      "epoch": 0.12982406356413168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9524160608369443e-05,
-      "loss": 1.0235,
+      "learning_rate": 1.948211319156456e-05,
+      "loss": 1.0996,
       "step": 4575
     },
     {
-      "epoch": 0.12567630661063964,
+      "epoch": 0.12985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9523889445323904e-05,
-      "loss": 1.047,
+      "learning_rate": 1.9481821215379097e-05,
+      "loss": 1.003,
       "step": 4576
     },
     {
-      "epoch": 0.1257037708384829,
+      "epoch": 0.12988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.9523618206921658e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.9481529159100258e-05,
+      "loss": 1.0534,
       "step": 4577
     },
     {
-      "epoch": 0.1257312350663261,
+      "epoch": 0.12990919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.952334689316486e-05,
-      "loss": 1.0963,
+      "learning_rate": 1.9481237022730516e-05,
+      "loss": 0.9887,
       "step": 4578
     },
     {
-      "epoch": 0.12575869929416936,
+      "epoch": 0.12993757094211122,
       "grad_norm": 0.0,
-      "learning_rate": 1.9523075504055654e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.9480944806272345e-05,
+      "loss": 1.0877,
       "step": 4579
     },
     {
-      "epoch": 0.12578616352201258,
+      "epoch": 0.12996594778660614,
       "grad_norm": 0.0,
-      "learning_rate": 1.952280403959618e-05,
-      "loss": 0.9007,
+      "learning_rate": 1.94806525097282e-05,
+      "loss": 1.0644,
       "step": 4580
     },
     {
-      "epoch": 0.1258136277498558,
+      "epoch": 0.12999432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.9522532499788598e-05,
-      "loss": 1.1012,
+      "learning_rate": 1.948036013310056e-05,
+      "loss": 1.0763,
       "step": 4581
     },
     {
-      "epoch": 0.12584109197769905,
+      "epoch": 0.13002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.9522260884635045e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.948006767639189e-05,
+      "loss": 1.0391,
       "step": 4582
     },
     {
-      "epoch": 0.12586855620554227,
+      "epoch": 0.1300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.9521989194137682e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.947977513960466e-05,
+      "loss": 0.891,
       "step": 4583
     },
     {
-      "epoch": 0.12589602043338552,
+      "epoch": 0.1300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.9521717428298642e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.9479482522741348e-05,
+      "loss": 0.9867,
       "step": 4584
     },
     {
-      "epoch": 0.12592348466122874,
+      "epoch": 0.1301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9521445587120093e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.9479189825804423e-05,
+      "loss": 0.9827,
       "step": 4585
     },
     {
-      "epoch": 0.125950948889072,
+      "epoch": 0.1301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9521173670604175e-05,
-      "loss": 1.0832,
+      "learning_rate": 1.947889704879635e-05,
+      "loss": 1.0865,
       "step": 4586
     },
     {
-      "epoch": 0.1259784131169152,
+      "epoch": 0.13016458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 1.9520901678753045e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.9478604191719607e-05,
+      "loss": 1.0073,
       "step": 4587
     },
     {
-      "epoch": 0.12600587734475846,
+      "epoch": 0.13019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.952062961156885e-05,
-      "loss": 1.064,
+      "learning_rate": 1.9478311254576673e-05,
+      "loss": 1.0308,
       "step": 4588
     },
     {
-      "epoch": 0.12603334157260168,
+      "epoch": 0.13022133938706015,
       "grad_norm": 0.0,
-      "learning_rate": 1.952035746905375e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.9478018237370014e-05,
+      "loss": 0.9808,
       "step": 4589
     },
     {
-      "epoch": 0.12606080580044493,
+      "epoch": 0.13024971623155504,
       "grad_norm": 0.0,
-      "learning_rate": 1.952008525120989e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.947772514010211e-05,
+      "loss": 1.0699,
       "step": 4590
     },
     {
-      "epoch": 0.12608827002828815,
+      "epoch": 0.13027809307604996,
       "grad_norm": 0.0,
-      "learning_rate": 1.951981295803943e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.9477431962775438e-05,
+      "loss": 1.0444,
       "step": 4591
     },
     {
-      "epoch": 0.1261157342561314,
+      "epoch": 0.13030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.951954058954452e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.9477138705392468e-05,
+      "loss": 0.9916,
       "step": 4592
     },
     {
-      "epoch": 0.12614319848397462,
+      "epoch": 0.13033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.951926814572732e-05,
-      "loss": 1.0755,
+      "learning_rate": 1.9476845367955686e-05,
+      "loss": 1.0069,
       "step": 4593
     },
     {
-      "epoch": 0.12617066271181784,
+      "epoch": 0.13036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.951899562658998e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.947655195046756e-05,
+      "loss": 1.0457,
       "step": 4594
     },
     {
-      "epoch": 0.1261981269396611,
+      "epoch": 0.1303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.9518723032134664e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.9476258452930577e-05,
+      "loss": 1.0172,
       "step": 4595
     },
     {
-      "epoch": 0.12622559116750431,
+      "epoch": 0.1304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.951845036236352e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.947596487534721e-05,
+      "loss": 0.9276,
       "step": 4596
     },
     {
-      "epoch": 0.12625305539534756,
+      "epoch": 0.1304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.951817761727871e-05,
-      "loss": 0.8425,
+      "learning_rate": 1.9475671217719947e-05,
+      "loss": 1.0588,
       "step": 4597
     },
     {
-      "epoch": 0.12628051962319078,
+      "epoch": 0.1304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.9517904796882392e-05,
-      "loss": 1.0822,
+      "learning_rate": 1.947537748005126e-05,
+      "loss": 1.0237,
       "step": 4598
     },
     {
-      "epoch": 0.12630798385103403,
+      "epoch": 0.13050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.9517631901176725e-05,
-      "loss": 1.073,
+      "learning_rate": 1.9475083662343633e-05,
+      "loss": 1.1265,
       "step": 4599
     },
     {
-      "epoch": 0.12633544807887725,
+      "epoch": 0.13053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.9517358930163867e-05,
-      "loss": 1.1445,
+      "learning_rate": 1.947478976459955e-05,
+      "loss": 1.1464,
       "step": 4600
     },
     {
-      "epoch": 0.1263629123067205,
+      "epoch": 0.13056186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.9517085883845977e-05,
-      "loss": 1.1178,
+      "learning_rate": 1.9474495786821493e-05,
+      "loss": 1.0089,
       "step": 4601
     },
     {
-      "epoch": 0.12639037653456373,
+      "epoch": 0.13059023836549374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9516812762225222e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.947420172901194e-05,
+      "loss": 0.917,
       "step": 4602
     },
     {
-      "epoch": 0.12641784076240697,
+      "epoch": 0.13061861520998866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9516539565303753e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.947390759117338e-05,
+      "loss": 1.0709,
       "step": 4603
     },
     {
-      "epoch": 0.1264453049902502,
+      "epoch": 0.13064699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 1.951626629308374e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.94736133733083e-05,
+      "loss": 1.141,
       "step": 4604
     },
     {
-      "epoch": 0.12647276921809344,
+      "epoch": 0.13067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.951599294556734e-05,
-      "loss": 1.1392,
+      "learning_rate": 1.947331907541918e-05,
+      "loss": 1.0096,
       "step": 4605
     },
     {
-      "epoch": 0.12650023344593667,
+      "epoch": 0.13070374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 1.9515719522756716e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.947302469750851e-05,
+      "loss": 1.0785,
       "step": 4606
     },
     {
-      "epoch": 0.12652769767377992,
+      "epoch": 0.1307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9515446024654038e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.947273023957877e-05,
+      "loss": 1.0078,
       "step": 4607
     },
     {
-      "epoch": 0.12655516190162314,
+      "epoch": 0.13076049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 1.951517245126146e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.9472435701632457e-05,
+      "loss": 1.008,
       "step": 4608
     },
     {
-      "epoch": 0.12658262612946636,
+      "epoch": 0.130788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9514898802581156e-05,
-      "loss": 1.1416,
+      "learning_rate": 1.947214108367205e-05,
+      "loss": 1.1033,
       "step": 4609
     },
     {
-      "epoch": 0.1266100903573096,
+      "epoch": 0.1308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.9514625078615282e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.9471846385700042e-05,
+      "loss": 0.9841,
       "step": 4610
     },
     {
-      "epoch": 0.12663755458515283,
+      "epoch": 0.13084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.951435127936601e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.947155160771892e-05,
+      "loss": 0.9753,
       "step": 4611
     },
     {
-      "epoch": 0.12666501881299608,
+      "epoch": 0.13087400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.9514077404835508e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.9471256749731177e-05,
+      "loss": 1.0661,
       "step": 4612
     },
     {
-      "epoch": 0.1266924830408393,
+      "epoch": 0.13090238365493756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9513803455025938e-05,
-      "loss": 1.0171,
+      "learning_rate": 1.9470961811739304e-05,
+      "loss": 1.1617,
       "step": 4613
     },
     {
-      "epoch": 0.12671994726868255,
+      "epoch": 0.13093076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9513529429939472e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9470666793745793e-05,
+      "loss": 0.9553,
       "step": 4614
     },
     {
-      "epoch": 0.12674741149652577,
+      "epoch": 0.13095913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.9513255329578275e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.947037169575313e-05,
+      "loss": 1.0348,
       "step": 4615
     },
     {
-      "epoch": 0.12677487572436902,
+      "epoch": 0.13098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.9512981153944517e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.9470076517763813e-05,
+      "loss": 0.9897,
       "step": 4616
     },
     {
-      "epoch": 0.12680233995221224,
+      "epoch": 0.13101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9512706903040365e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.9469781259780332e-05,
+      "loss": 1.0145,
       "step": 4617
     },
     {
-      "epoch": 0.1268298041800555,
+      "epoch": 0.13104426787741202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9512432576867995e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.9469485921805183e-05,
+      "loss": 1.0746,
       "step": 4618
     },
     {
-      "epoch": 0.1268572684078987,
+      "epoch": 0.1310726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.951215817542957e-05,
-      "loss": 0.989,
+      "learning_rate": 1.9469190503840868e-05,
+      "loss": 1.1158,
       "step": 4619
     },
     {
-      "epoch": 0.12688473263574196,
+      "epoch": 0.13110102156640183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9511883698727265e-05,
-      "loss": 1.1228,
+      "learning_rate": 1.9468895005889866e-05,
+      "loss": 1.0975,
       "step": 4620
     },
     {
-      "epoch": 0.12691219686358518,
+      "epoch": 0.1311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.9511609146763253e-05,
-      "loss": 1.106,
+      "learning_rate": 1.9468599427954685e-05,
+      "loss": 0.8598,
       "step": 4621
     },
     {
-      "epoch": 0.1269396610914284,
+      "epoch": 0.1311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9511334519539703e-05,
-      "loss": 1.1261,
+      "learning_rate": 1.946830377003782e-05,
+      "loss": 1.0443,
       "step": 4622
     },
     {
-      "epoch": 0.12696712531927165,
+      "epoch": 0.1311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9511059817058794e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.946800803214177e-05,
+      "loss": 1.0811,
       "step": 4623
     },
     {
-      "epoch": 0.12699458954711487,
+      "epoch": 0.13121452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.9510785039322693e-05,
-      "loss": 1.1104,
+      "learning_rate": 1.9467712214269028e-05,
+      "loss": 1.1818,
       "step": 4624
     },
     {
-      "epoch": 0.12702205377495812,
+      "epoch": 0.1312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9510510186333577e-05,
-      "loss": 1.0001,
+      "learning_rate": 1.9467416316422093e-05,
+      "loss": 1.0259,
       "step": 4625
     },
     {
-      "epoch": 0.12704951800280134,
+      "epoch": 0.13127128263337118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9510235258093623e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.946712033860347e-05,
+      "loss": 1.03,
       "step": 4626
     },
     {
-      "epoch": 0.1270769822306446,
+      "epoch": 0.13129965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.9509960254604997e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9466824280815654e-05,
+      "loss": 1.2318,
       "step": 4627
     },
     {
-      "epoch": 0.1271044464584878,
+      "epoch": 0.13132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.9509685175869884e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.9466528143061148e-05,
+      "loss": 1.0018,
       "step": 4628
     },
     {
-      "epoch": 0.12713191068633106,
+      "epoch": 0.13135641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 1.9509410021890463e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.9466231925342453e-05,
+      "loss": 1.0174,
       "step": 4629
     },
     {
-      "epoch": 0.12715937491417428,
+      "epoch": 0.13138479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 1.95091347926689e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.9465935627662075e-05,
+      "loss": 1.0393,
       "step": 4630
     },
     {
-      "epoch": 0.12718683914201753,
+      "epoch": 0.13141316685584564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9508859488207385e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.946563925002251e-05,
+      "loss": 0.9847,
       "step": 4631
     },
     {
-      "epoch": 0.12721430336986075,
+      "epoch": 0.13144154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 1.950858410850809e-05,
-      "loss": 1.1502,
+      "learning_rate": 1.9465342792426264e-05,
+      "loss": 0.9718,
       "step": 4632
     },
     {
-      "epoch": 0.127241767597704,
+      "epoch": 0.13146992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 1.950830865357319e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.9465046254875842e-05,
+      "loss": 0.9832,
       "step": 4633
     },
     {
-      "epoch": 0.12726923182554722,
+      "epoch": 0.1314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.950803312340487e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.946474963737375e-05,
+      "loss": 0.931,
       "step": 4634
     },
     {
-      "epoch": 0.12729669605339045,
+      "epoch": 0.1315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.9507757518005312e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.9464452939922492e-05,
+      "loss": 1.0666,
       "step": 4635
     },
     {
-      "epoch": 0.1273241602812337,
+      "epoch": 0.13155505107832008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9507481837376687e-05,
-      "loss": 1.1672,
+      "learning_rate": 1.9464156162524578e-05,
+      "loss": 1.0729,
       "step": 4636
     },
     {
-      "epoch": 0.12735162450907692,
+      "epoch": 0.131583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9507206081521187e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.9463859305182508e-05,
+      "loss": 1.0114,
       "step": 4637
     },
     {
-      "epoch": 0.12737908873692017,
+      "epoch": 0.13161180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.950693025044099e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.9463562367898793e-05,
+      "loss": 1.027,
       "step": 4638
     },
     {
-      "epoch": 0.1274065529647634,
+      "epoch": 0.13164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.9506654344138277e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.946326535067594e-05,
+      "loss": 1.017,
       "step": 4639
     },
     {
-      "epoch": 0.12743401719260664,
+      "epoch": 0.13166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9506378362615235e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.9462968253516462e-05,
+      "loss": 1.023,
       "step": 4640
     },
     {
-      "epoch": 0.12746148142044986,
+      "epoch": 0.13169693530079454,
       "grad_norm": 0.0,
-      "learning_rate": 1.950610230587404e-05,
-      "loss": 1.1364,
+      "learning_rate": 1.9462671076422863e-05,
+      "loss": 0.9801,
       "step": 4641
     },
     {
-      "epoch": 0.1274889456482931,
+      "epoch": 0.13172531214528943,
       "grad_norm": 0.0,
-      "learning_rate": 1.9505826173916887e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.946237381939766e-05,
+      "loss": 0.967,
       "step": 4642
     },
     {
-      "epoch": 0.12751640987613633,
+      "epoch": 0.13175368898978435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9505549966745952e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.9462076482443355e-05,
+      "loss": 0.9576,
       "step": 4643
     },
     {
-      "epoch": 0.12754387410397958,
+      "epoch": 0.13178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9505273684363425e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.9461779065562467e-05,
+      "loss": 1.0659,
       "step": 4644
     },
     {
-      "epoch": 0.1275713383318228,
+      "epoch": 0.13181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.950499732677149e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.946148156875751e-05,
+      "loss": 1.0311,
       "step": 4645
     },
     {
-      "epoch": 0.12759880255966605,
+      "epoch": 0.131838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9504720893972333e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9461183992030984e-05,
+      "loss": 1.0096,
       "step": 4646
     },
     {
-      "epoch": 0.12762626678750927,
+      "epoch": 0.1318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.9504444385968143e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.946088633538542e-05,
+      "loss": 1.0939,
       "step": 4647
     },
     {
-      "epoch": 0.1276537310153525,
+      "epoch": 0.1318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.950416780276111e-05,
-      "loss": 1.013,
+      "learning_rate": 1.9460588598823317e-05,
+      "loss": 0.9298,
       "step": 4648
     },
     {
-      "epoch": 0.12768119524319574,
+      "epoch": 0.1319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9503891144353415e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.94602907823472e-05,
+      "loss": 1.0662,
       "step": 4649
     },
     {
-      "epoch": 0.12770865947103896,
+      "epoch": 0.13195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9503614410747254e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.945999288595958e-05,
+      "loss": 1.0439,
       "step": 4650
     },
     {
-      "epoch": 0.1277361236988822,
+      "epoch": 0.13198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.9503337601944812e-05,
-      "loss": 1.0722,
+      "learning_rate": 1.945969490966298e-05,
+      "loss": 1.0263,
       "step": 4651
     },
     {
-      "epoch": 0.12776358792672543,
+      "epoch": 0.13200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.9503060717948287e-05,
-      "loss": 1.1396,
+      "learning_rate": 1.9459396853459906e-05,
+      "loss": 1.1933,
       "step": 4652
     },
     {
-      "epoch": 0.12779105215456868,
+      "epoch": 0.13203745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9502783758759862e-05,
-      "loss": 1.0341,
+      "learning_rate": 1.9459098717352883e-05,
+      "loss": 1.0708,
       "step": 4653
     },
     {
-      "epoch": 0.1278185163824119,
+      "epoch": 0.13206583427922816,
       "grad_norm": 0.0,
-      "learning_rate": 1.950250672438173e-05,
-      "loss": 1.1138,
+      "learning_rate": 1.9458800501344426e-05,
+      "loss": 0.9729,
       "step": 4654
     },
     {
-      "epoch": 0.12784598061025515,
+      "epoch": 0.13209421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 1.9502229614816087e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.9458502205437063e-05,
+      "loss": 1.0594,
       "step": 4655
     },
     {
-      "epoch": 0.12787344483809837,
+      "epoch": 0.13212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.950195243006512e-05,
-      "loss": 1.0735,
+      "learning_rate": 1.94582038296333e-05,
+      "loss": 0.9234,
       "step": 4656
     },
     {
-      "epoch": 0.12790090906594162,
+      "epoch": 0.13215096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 1.950167517013102e-05,
-      "loss": 1.1747,
+      "learning_rate": 1.9457905373935667e-05,
+      "loss": 0.9799,
       "step": 4657
     },
     {
-      "epoch": 0.12792837329378484,
+      "epoch": 0.1321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.9501397835015994e-05,
-      "loss": 1.1625,
+      "learning_rate": 1.945760683834668e-05,
+      "loss": 1.0086,
       "step": 4658
     },
     {
-      "epoch": 0.1279558375216281,
+      "epoch": 0.1322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.9501120424722225e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.9457308222868866e-05,
+      "loss": 1.0815,
       "step": 4659
     },
     {
-      "epoch": 0.1279833017494713,
+      "epoch": 0.1322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.950084293925191e-05,
-      "loss": 1.083,
+      "learning_rate": 1.945700952750474e-05,
+      "loss": 0.979,
       "step": 4660
     },
     {
-      "epoch": 0.12801076597731456,
+      "epoch": 0.1322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9500565378607246e-05,
-      "loss": 1.0868,
+      "learning_rate": 1.9456710752256835e-05,
+      "loss": 0.9948,
       "step": 4661
     },
     {
-      "epoch": 0.12803823020515778,
+      "epoch": 0.1322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9500287742790423e-05,
-      "loss": 1.1353,
+      "learning_rate": 1.9456411897127667e-05,
+      "loss": 1.0494,
       "step": 4662
     },
     {
-      "epoch": 0.128065694433001,
+      "epoch": 0.13232122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 1.950001003180365e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.9456112962119762e-05,
+      "loss": 1.0773,
       "step": 4663
     },
     {
-      "epoch": 0.12809315866084425,
+      "epoch": 0.13234960272417706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9499732245649116e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.9455813947235644e-05,
+      "loss": 1.0629,
       "step": 4664
     },
     {
-      "epoch": 0.12812062288868747,
+      "epoch": 0.13237797956867198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9499454384329022e-05,
-      "loss": 1.1037,
+      "learning_rate": 1.9455514852477843e-05,
+      "loss": 1.0214,
       "step": 4665
     },
     {
-      "epoch": 0.12814808711653072,
+      "epoch": 0.13240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.9499176447845563e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.945521567784888e-05,
+      "loss": 1.0297,
       "step": 4666
     },
     {
-      "epoch": 0.12817555134437394,
+      "epoch": 0.13243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9498898436200943e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9454916423351288e-05,
+      "loss": 0.9964,
       "step": 4667
     },
     {
-      "epoch": 0.1282030155722172,
+      "epoch": 0.13246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9498620349397355e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.9454617088987592e-05,
+      "loss": 1.0146,
       "step": 4668
     },
     {
-      "epoch": 0.12823047980006042,
+      "epoch": 0.13249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.9498342187437008e-05,
-      "loss": 0.9768,
+      "learning_rate": 1.9454317674760318e-05,
+      "loss": 0.9777,
       "step": 4669
     },
     {
-      "epoch": 0.12825794402790366,
+      "epoch": 0.13251986379114641,
       "grad_norm": 0.0,
-      "learning_rate": 1.9498063950322094e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.9454018180672002e-05,
+      "loss": 1.0457,
       "step": 4670
     },
     {
-      "epoch": 0.12828540825574689,
+      "epoch": 0.13254824063564133,
       "grad_norm": 0.0,
-      "learning_rate": 1.949778563805482e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.9453718606725166e-05,
+      "loss": 1.0579,
       "step": 4671
     },
     {
-      "epoch": 0.12831287248359013,
+      "epoch": 0.13257661748013622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9497507250637388e-05,
-      "loss": 1.1123,
+      "learning_rate": 1.9453418952922344e-05,
+      "loss": 1.0236,
       "step": 4672
     },
     {
-      "epoch": 0.12834033671143336,
+      "epoch": 0.1326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9497228788071997e-05,
-      "loss": 1.1128,
+      "learning_rate": 1.9453119219266066e-05,
+      "loss": 1.038,
       "step": 4673
     },
     {
-      "epoch": 0.1283678009392766,
+      "epoch": 0.132633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9496950250360855e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.9452819405758868e-05,
+      "loss": 1.0503,
       "step": 4674
     },
     {
-      "epoch": 0.12839526516711983,
+      "epoch": 0.13266174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 1.9496671637506164e-05,
-      "loss": 1.0007,
+      "learning_rate": 1.9452519512403277e-05,
+      "loss": 0.9994,
       "step": 4675
     },
     {
-      "epoch": 0.12842272939496305,
+      "epoch": 0.13269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.949639294951013e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.945221953920183e-05,
+      "loss": 1.0273,
       "step": 4676
     },
     {
-      "epoch": 0.1284501936228063,
+      "epoch": 0.13271850170261068,
       "grad_norm": 0.0,
-      "learning_rate": 1.9496114186374956e-05,
-      "loss": 1.167,
+      "learning_rate": 1.9451919486157062e-05,
+      "loss": 1.0464,
       "step": 4677
     },
     {
-      "epoch": 0.12847765785064952,
+      "epoch": 0.13274687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9495835348102846e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.9451619353271503e-05,
+      "loss": 1.108,
       "step": 4678
     },
     {
-      "epoch": 0.12850512207849277,
+      "epoch": 0.13277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.9495556434696007e-05,
-      "loss": 1.1242,
+      "learning_rate": 1.9451319140547692e-05,
+      "loss": 1.0107,
       "step": 4679
     },
     {
-      "epoch": 0.128532586306336,
+      "epoch": 0.13280363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.949527744615665e-05,
-      "loss": 1.1421,
+      "learning_rate": 1.945101884798816e-05,
+      "loss": 1.0576,
       "step": 4680
     },
     {
-      "epoch": 0.12856005053417924,
+      "epoch": 0.13283200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 1.949499838248698e-05,
-      "loss": 1.0876,
+      "learning_rate": 1.945071847559545e-05,
+      "loss": 0.9905,
       "step": 4681
     },
     {
-      "epoch": 0.12858751476202246,
+      "epoch": 0.13286038592508512,
       "grad_norm": 0.0,
-      "learning_rate": 1.9494719243689204e-05,
-      "loss": 1.063,
+      "learning_rate": 1.9450418023372094e-05,
+      "loss": 1.13,
       "step": 4682
     },
     {
-      "epoch": 0.1286149789898657,
+      "epoch": 0.13288876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 1.949444002976553e-05,
-      "loss": 0.9571,
+      "learning_rate": 1.9450117491320633e-05,
+      "loss": 0.9693,
       "step": 4683
     },
     {
-      "epoch": 0.12864244321770893,
+      "epoch": 0.13291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.949416074071817e-05,
-      "loss": 1.0455,
+      "learning_rate": 1.9449816879443604e-05,
+      "loss": 0.9962,
       "step": 4684
     },
     {
-      "epoch": 0.12866990744555218,
+      "epoch": 0.1329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.949388137654933e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.9449516187743546e-05,
+      "loss": 1.0366,
       "step": 4685
     },
     {
-      "epoch": 0.1286973716733954,
+      "epoch": 0.1329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.949360193726123e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.9449215416223003e-05,
+      "loss": 0.9914,
       "step": 4686
     },
     {
-      "epoch": 0.12872483590123865,
+      "epoch": 0.13300227014755958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9493322422856068e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.9448914564884515e-05,
+      "loss": 1.0449,
       "step": 4687
     },
     {
-      "epoch": 0.12875230012908187,
+      "epoch": 0.1330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9493042833336065e-05,
-      "loss": 1.1745,
+      "learning_rate": 1.9448613633730614e-05,
+      "loss": 1.0851,
       "step": 4688
     },
     {
-      "epoch": 0.1287797643569251,
+      "epoch": 0.13305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9492763168703426e-05,
-      "loss": 1.1251,
+      "learning_rate": 1.9448312622763852e-05,
+      "loss": 0.9816,
       "step": 4689
     },
     {
-      "epoch": 0.12880722858476834,
+      "epoch": 0.13308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.949248342896037e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.9448011531986772e-05,
+      "loss": 1.0301,
       "step": 4690
     },
     {
-      "epoch": 0.12883469281261156,
+      "epoch": 0.13311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.949220361410911e-05,
-      "loss": 0.906,
+      "learning_rate": 1.944771036140191e-05,
+      "loss": 1.0715,
       "step": 4691
     },
     {
-      "epoch": 0.1288621570404548,
+      "epoch": 0.13314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.9491923724151857e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.9447409111011814e-05,
+      "loss": 1.0583,
       "step": 4692
     },
     {
-      "epoch": 0.12888962126829803,
+      "epoch": 0.13317253121452893,
       "grad_norm": 0.0,
-      "learning_rate": 1.9491643759090825e-05,
-      "loss": 1.1602,
+      "learning_rate": 1.9447107780819028e-05,
+      "loss": 1.0811,
       "step": 4693
     },
     {
-      "epoch": 0.12891708549614128,
+      "epoch": 0.13320090805902385,
       "grad_norm": 0.0,
-      "learning_rate": 1.949136371892823e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.94468063708261e-05,
+      "loss": 1.0612,
       "step": 4694
     },
     {
-      "epoch": 0.1289445497239845,
+      "epoch": 0.13322928490351874,
       "grad_norm": 0.0,
-      "learning_rate": 1.949108360366629e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.944650488103557e-05,
+      "loss": 1.0114,
       "step": 4695
     },
     {
-      "epoch": 0.12897201395182775,
+      "epoch": 0.13325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.949080341330722e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.944620331144999e-05,
+      "loss": 1.0122,
       "step": 4696
     },
     {
-      "epoch": 0.12899947817967097,
+      "epoch": 0.1332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.949052314785324e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.9445901662071908e-05,
+      "loss": 1.158,
       "step": 4697
     },
     {
-      "epoch": 0.12902694240751422,
+      "epoch": 0.1333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.949024280730656e-05,
-      "loss": 1.0861,
+      "learning_rate": 1.944559993290387e-05,
+      "loss": 1.0326,
       "step": 4698
     },
     {
-      "epoch": 0.12905440663535744,
+      "epoch": 0.13334279228149828,
       "grad_norm": 0.0,
-      "learning_rate": 1.9489962391669404e-05,
-      "loss": 1.14,
+      "learning_rate": 1.9445298123948425e-05,
+      "loss": 0.9444,
       "step": 4699
     },
     {
-      "epoch": 0.1290818708632007,
+      "epoch": 0.1333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.9489681900943988e-05,
-      "loss": 1.1365,
+      "learning_rate": 1.944499623520812e-05,
+      "loss": 1.1225,
       "step": 4700
     },
     {
-      "epoch": 0.12910933509104391,
+      "epoch": 0.1333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.9489401335132536e-05,
-      "loss": 0.9545,
+      "learning_rate": 1.944469426668551e-05,
+      "loss": 1.0154,
       "step": 4701
     },
     {
-      "epoch": 0.12913679931888716,
+      "epoch": 0.13342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.9489120694237264e-05,
-      "loss": 1.0472,
+      "learning_rate": 1.944439221838314e-05,
+      "loss": 1.0539,
       "step": 4702
     },
     {
-      "epoch": 0.12916426354673038,
+      "epoch": 0.13345629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9488839978260393e-05,
-      "loss": 1.0679,
+      "learning_rate": 1.9444090090303567e-05,
+      "loss": 0.9347,
       "step": 4703
     },
     {
-      "epoch": 0.1291917277745736,
+      "epoch": 0.13348467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.9488559187204147e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.944378788244934e-05,
+      "loss": 1.0651,
       "step": 4704
     },
     {
-      "epoch": 0.12921919200241685,
+      "epoch": 0.13351305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.948827832107074e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.9443485594823012e-05,
+      "loss": 0.9453,
       "step": 4705
     },
     {
-      "epoch": 0.12924665623026008,
+      "epoch": 0.13354143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 1.9487997379862404e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.9443183227427137e-05,
+      "loss": 1.0273,
       "step": 4706
     },
     {
-      "epoch": 0.12927412045810333,
+      "epoch": 0.13356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9487716363581354e-05,
-      "loss": 1.0796,
+      "learning_rate": 1.9442880780264266e-05,
+      "loss": 1.0954,
       "step": 4707
     },
     {
-      "epoch": 0.12930158468594655,
+      "epoch": 0.13359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.9487435272229818e-05,
-      "loss": 1.2001,
+      "learning_rate": 1.944257825333696e-05,
+      "loss": 0.9572,
       "step": 4708
     },
     {
-      "epoch": 0.1293290489137898,
+      "epoch": 0.1336265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9487154105810017e-05,
-      "loss": 1.1113,
+      "learning_rate": 1.9442275646647768e-05,
+      "loss": 1.0363,
       "step": 4709
     },
     {
-      "epoch": 0.12935651314163302,
+      "epoch": 0.1336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.948687286432418e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.944197296019925e-05,
+      "loss": 0.9055,
       "step": 4710
     },
     {
-      "epoch": 0.12938397736947627,
+      "epoch": 0.13368331441543702,
       "grad_norm": 0.0,
-      "learning_rate": 1.9486591547774533e-05,
-      "loss": 1.032,
+      "learning_rate": 1.9441670193993962e-05,
+      "loss": 1.0806,
       "step": 4711
     },
     {
-      "epoch": 0.1294114415973195,
+      "epoch": 0.1337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.9486310156163294e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.9441367348034464e-05,
+      "loss": 1.01,
       "step": 4712
     },
     {
-      "epoch": 0.12943890582516274,
+      "epoch": 0.1337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.94860286894927e-05,
-      "loss": 1.1608,
+      "learning_rate": 1.944106442232331e-05,
+      "loss": 1.076,
       "step": 4713
     },
     {
-      "epoch": 0.12946637005300596,
+      "epoch": 0.13376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9485747147764966e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.944076141686306e-05,
+      "loss": 1.0958,
       "step": 4714
     },
     {
-      "epoch": 0.1294938342808492,
+      "epoch": 0.13379682179341656,
       "grad_norm": 0.0,
-      "learning_rate": 1.948546553098233e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.9440458331656272e-05,
+      "loss": 0.938,
       "step": 4715
     },
     {
-      "epoch": 0.12952129850869243,
+      "epoch": 0.13382519863791145,
       "grad_norm": 0.0,
-      "learning_rate": 1.9485183839147018e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.944015516670551e-05,
+      "loss": 1.0461,
       "step": 4716
     },
     {
-      "epoch": 0.12954876273653565,
+      "epoch": 0.13385357548240637,
       "grad_norm": 0.0,
-      "learning_rate": 1.9484902072261254e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.9439851922013333e-05,
+      "loss": 1.0711,
       "step": 4717
     },
     {
-      "epoch": 0.1295762269643789,
+      "epoch": 0.13388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.948462023032727e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.9439548597582304e-05,
+      "loss": 0.9865,
       "step": 4718
     },
     {
-      "epoch": 0.12960369119222212,
+      "epoch": 0.13391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.94843383133473e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.9439245193414984e-05,
+      "loss": 0.9833,
       "step": 4719
     },
     {
-      "epoch": 0.12963115542006537,
+      "epoch": 0.13393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.948405632132357e-05,
-      "loss": 1.08,
+      "learning_rate": 1.9438941709513933e-05,
+      "loss": 1.0398,
       "step": 4720
     },
     {
-      "epoch": 0.1296586196479086,
+      "epoch": 0.13396708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 1.9483774254258312e-05,
-      "loss": 1.1443,
+      "learning_rate": 1.9438638145881718e-05,
+      "loss": 1.052,
       "step": 4721
     },
     {
-      "epoch": 0.12968608387575184,
+      "epoch": 0.1339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9483492112153758e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.94383345025209e-05,
+      "loss": 1.0258,
       "step": 4722
     },
     {
-      "epoch": 0.12971354810359506,
+      "epoch": 0.13402383654937572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9483209895012144e-05,
-      "loss": 1.1463,
+      "learning_rate": 1.943803077943405e-05,
+      "loss": 0.9958,
       "step": 4723
     },
     {
-      "epoch": 0.1297410123314383,
+      "epoch": 0.1340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.9482927602835696e-05,
-      "loss": 0.972,
+      "learning_rate": 1.9437726976623726e-05,
+      "loss": 1.0387,
       "step": 4724
     },
     {
-      "epoch": 0.12976847655928153,
+      "epoch": 0.1340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9482645235626655e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.9437423094092503e-05,
+      "loss": 0.998,
       "step": 4725
     },
     {
-      "epoch": 0.12979594078712478,
+      "epoch": 0.13410896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 1.9482362793387248e-05,
-      "loss": 1.1895,
+      "learning_rate": 1.9437119131842937e-05,
+      "loss": 1.1107,
       "step": 4726
     },
     {
-      "epoch": 0.129823405014968,
+      "epoch": 0.13413734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.9482080276119713e-05,
-      "loss": 1.0493,
+      "learning_rate": 1.9436815089877607e-05,
+      "loss": 1.0124,
       "step": 4727
     },
     {
-      "epoch": 0.12985086924281125,
+      "epoch": 0.13416572077185018,
       "grad_norm": 0.0,
-      "learning_rate": 1.948179768382629e-05,
-      "loss": 1.125,
+      "learning_rate": 1.9436510968199072e-05,
+      "loss": 1.0897,
       "step": 4728
     },
     {
-      "epoch": 0.12987833347065447,
+      "epoch": 0.13419409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 1.9481515016509208e-05,
-      "loss": 0.9853,
+      "learning_rate": 1.9436206766809906e-05,
+      "loss": 1.1339,
       "step": 4729
     },
     {
-      "epoch": 0.1299057976984977,
+      "epoch": 0.13422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.9481232274170708e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.9435902485712677e-05,
+      "loss": 1.0872,
       "step": 4730
     },
     {
-      "epoch": 0.12993326192634094,
+      "epoch": 0.13425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9480949456813026e-05,
-      "loss": 1.0947,
+      "learning_rate": 1.9435598124909955e-05,
+      "loss": 1.0993,
       "step": 4731
     },
     {
-      "epoch": 0.12996072615418416,
+      "epoch": 0.13427922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 1.9480666564438397e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.943529368440431e-05,
+      "loss": 1.0969,
       "step": 4732
     },
     {
-      "epoch": 0.1299881903820274,
+      "epoch": 0.13430760499432462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9480383597049063e-05,
-      "loss": 1.1047,
+      "learning_rate": 1.943498916419832e-05,
+      "loss": 1.0626,
       "step": 4733
     },
     {
-      "epoch": 0.13001565460987063,
+      "epoch": 0.13433598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 1.948010055464726e-05,
-      "loss": 1.047,
+      "learning_rate": 1.9434684564294547e-05,
+      "loss": 1.0873,
       "step": 4734
     },
     {
-      "epoch": 0.13004311883771388,
+      "epoch": 0.13436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9479817437235232e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.9434379884695573e-05,
+      "loss": 1.0288,
       "step": 4735
     },
     {
-      "epoch": 0.1300705830655571,
+      "epoch": 0.1343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9479534244815214e-05,
-      "loss": 1.0007,
+      "learning_rate": 1.9434075125403966e-05,
+      "loss": 1.1635,
       "step": 4736
     },
     {
-      "epoch": 0.13009804729340035,
+      "epoch": 0.1344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.947925097738945e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.9433770286422305e-05,
+      "loss": 1.0417,
       "step": 4737
     },
     {
-      "epoch": 0.13012551152124358,
+      "epoch": 0.13444948921679908,
       "grad_norm": 0.0,
-      "learning_rate": 1.947896763496018e-05,
-      "loss": 1.0783,
+      "learning_rate": 1.9433465367753158e-05,
+      "loss": 0.8746,
       "step": 4738
     },
     {
-      "epoch": 0.13015297574908682,
+      "epoch": 0.13447786606129397,
       "grad_norm": 0.0,
-      "learning_rate": 1.9478684217529645e-05,
-      "loss": 1.009,
+      "learning_rate": 1.9433160369399108e-05,
+      "loss": 0.9302,
       "step": 4739
     },
     {
-      "epoch": 0.13018043997693005,
+      "epoch": 0.1345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.947840072510009e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.943285529136273e-05,
+      "loss": 1.0304,
       "step": 4740
     },
     {
-      "epoch": 0.1302079042047733,
+      "epoch": 0.13453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.947811715767376e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.9432550133646594e-05,
+      "loss": 1.0749,
       "step": 4741
     },
     {
-      "epoch": 0.13023536843261652,
+      "epoch": 0.13456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.947783351525289e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.9432244896253287e-05,
+      "loss": 0.9208,
       "step": 4742
     },
     {
-      "epoch": 0.13026283266045974,
+      "epoch": 0.13459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.9477549797839735e-05,
-      "loss": 1.0479,
+      "learning_rate": 1.9431939579185384e-05,
+      "loss": 1.0308,
       "step": 4743
     },
     {
-      "epoch": 0.130290296888303,
+      "epoch": 0.13461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.947726600543653e-05,
-      "loss": 1.0275,
+      "learning_rate": 1.943163418244546e-05,
+      "loss": 0.9863,
       "step": 4744
     },
     {
-      "epoch": 0.1303177611161462,
+      "epoch": 0.13464812712826335,
       "grad_norm": 0.0,
-      "learning_rate": 1.9476982138045527e-05,
-      "loss": 1.0954,
+      "learning_rate": 1.9431328706036102e-05,
+      "loss": 1.1182,
       "step": 4745
     },
     {
-      "epoch": 0.13034522534398946,
+      "epoch": 0.13467650397275824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9476698195668972e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.9431023149959882e-05,
+      "loss": 0.9562,
       "step": 4746
     },
     {
-      "epoch": 0.13037268957183268,
+      "epoch": 0.13470488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 1.947641417830911e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.9430717514219387e-05,
+      "loss": 1.0922,
       "step": 4747
     },
     {
-      "epoch": 0.13040015379967593,
+      "epoch": 0.134733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9476130085968186e-05,
-      "loss": 1.1018,
+      "learning_rate": 1.94304117988172e-05,
+      "loss": 1.116,
       "step": 4748
     },
     {
-      "epoch": 0.13042761802751915,
+      "epoch": 0.1347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.9475845918648452e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.94301060037559e-05,
+      "loss": 0.9255,
       "step": 4749
     },
     {
-      "epoch": 0.1304550822553624,
+      "epoch": 0.1347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9475561676352153e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.9429800129038068e-05,
+      "loss": 0.9765,
       "step": 4750
     },
     {
-      "epoch": 0.13048254648320562,
+      "epoch": 0.1348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9475277359081542e-05,
-      "loss": 1.1068,
+      "learning_rate": 1.942949417466629e-05,
+      "loss": 1.1655,
       "step": 4751
     },
     {
-      "epoch": 0.13051001071104887,
+      "epoch": 0.1348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.9474992966838863e-05,
-      "loss": 1.106,
+      "learning_rate": 1.9429188140643154e-05,
+      "loss": 1.0732,
       "step": 4752
     },
     {
-      "epoch": 0.1305374749388921,
+      "epoch": 0.13487514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 1.9474708499626372e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.942888202697124e-05,
+      "loss": 0.9914,
       "step": 4753
     },
     {
-      "epoch": 0.13056493916673534,
+      "epoch": 0.13490351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.9474423957446315e-05,
-      "loss": 1.1395,
+      "learning_rate": 1.9428575833653137e-05,
+      "loss": 0.9616,
       "step": 4754
     },
     {
-      "epoch": 0.13059240339457856,
+      "epoch": 0.13493189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 1.9474139340300945e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.942826956069143e-05,
+      "loss": 1.0185,
       "step": 4755
     },
     {
-      "epoch": 0.1306198676224218,
+      "epoch": 0.13496027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.947385464819252e-05,
-      "loss": 1.0411,
+      "learning_rate": 1.9427963208088704e-05,
+      "loss": 0.8133,
       "step": 4756
     },
     {
-      "epoch": 0.13064733185026503,
+      "epoch": 0.13498864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9473569881123284e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.9427656775847553e-05,
+      "loss": 1.0311,
       "step": 4757
     },
     {
-      "epoch": 0.13067479607810825,
+      "epoch": 0.13501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.947328503909549e-05,
-      "loss": 1.0912,
+      "learning_rate": 1.9427350263970557e-05,
+      "loss": 0.9747,
       "step": 4758
     },
     {
-      "epoch": 0.1307022603059515,
+      "epoch": 0.13504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9473000122111395e-05,
-      "loss": 1.0554,
+      "learning_rate": 1.942704367246031e-05,
+      "loss": 1.0197,
       "step": 4759
     },
     {
-      "epoch": 0.13072972453379472,
+      "epoch": 0.13507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.947271513017326e-05,
-      "loss": 1.0969,
+      "learning_rate": 1.9426737001319402e-05,
+      "loss": 0.9315,
       "step": 4760
     },
     {
-      "epoch": 0.13075718876163797,
+      "epoch": 0.1351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.947243006328333e-05,
-      "loss": 1.0869,
+      "learning_rate": 1.9426430250550424e-05,
+      "loss": 0.9854,
       "step": 4761
     },
     {
-      "epoch": 0.1307846529894812,
+      "epoch": 0.1351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9472144921443863e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.9426123420155965e-05,
+      "loss": 1.0383,
       "step": 4762
     },
     {
-      "epoch": 0.13081211721732444,
+      "epoch": 0.1351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9471859704657117e-05,
-      "loss": 1.1698,
+      "learning_rate": 1.942581651013862e-05,
+      "loss": 0.9684,
       "step": 4763
     },
     {
-      "epoch": 0.13083958144516766,
+      "epoch": 0.1351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.947157441292535e-05,
-      "loss": 1.0937,
+      "learning_rate": 1.9425509520500977e-05,
+      "loss": 1.1064,
       "step": 4764
     },
     {
-      "epoch": 0.1308670456730109,
+      "epoch": 0.13521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9471289046250815e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.942520245124563e-05,
+      "loss": 1.009,
       "step": 4765
     },
     {
-      "epoch": 0.13089450990085413,
+      "epoch": 0.13524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.9471003604635772e-05,
-      "loss": 1.029,
+      "learning_rate": 1.9424895302375177e-05,
+      "loss": 0.9044,
       "step": 4766
     },
     {
-      "epoch": 0.13092197412869738,
+      "epoch": 0.13527241770715095,
       "grad_norm": 0.0,
-      "learning_rate": 1.9470718088082477e-05,
-      "loss": 1.061,
+      "learning_rate": 1.9424588073892207e-05,
+      "loss": 1.0396,
       "step": 4767
     },
     {
-      "epoch": 0.1309494383565406,
+      "epoch": 0.13530079455164587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9470432496593198e-05,
-      "loss": 1.1245,
+      "learning_rate": 1.942428076579932e-05,
+      "loss": 1.0208,
       "step": 4768
     },
     {
-      "epoch": 0.13097690258438385,
+      "epoch": 0.13532917139614076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9470146830170186e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.942397337809911e-05,
+      "loss": 0.9939,
       "step": 4769
     },
     {
-      "epoch": 0.13100436681222707,
+      "epoch": 0.13535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9469861088815705e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.9423665910794175e-05,
+      "loss": 1.0417,
       "step": 4770
     },
     {
-      "epoch": 0.1310318310400703,
+      "epoch": 0.13538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.9469575272532013e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.9423358363887105e-05,
+      "loss": 0.9853,
       "step": 4771
     },
     {
-      "epoch": 0.13105929526791354,
+      "epoch": 0.13541430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.946928938132137e-05,
-      "loss": 1.0998,
+      "learning_rate": 1.9423050737380507e-05,
+      "loss": 0.9926,
       "step": 4772
     },
     {
-      "epoch": 0.13108675949575677,
+      "epoch": 0.1354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.946900341518605e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.9422743031276977e-05,
+      "loss": 1.0235,
       "step": 4773
     },
     {
-      "epoch": 0.13111422372360002,
+      "epoch": 0.13547105561861522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9468717374128302e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.942243524557911e-05,
+      "loss": 1.024,
       "step": 4774
     },
     {
-      "epoch": 0.13114168795144324,
+      "epoch": 0.1354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.9468431258150394e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.9422127380289508e-05,
+      "loss": 1.0119,
       "step": 4775
     },
     {
-      "epoch": 0.13116915217928649,
+      "epoch": 0.135527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9468145067254593e-05,
-      "loss": 1.0802,
+      "learning_rate": 1.9421819435410778e-05,
+      "loss": 0.9802,
       "step": 4776
     },
     {
-      "epoch": 0.1311966164071297,
+      "epoch": 0.13555618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 1.9467858801443156e-05,
-      "loss": 1.0326,
+      "learning_rate": 1.942151141094551e-05,
+      "loss": 1.0577,
       "step": 4777
     },
     {
-      "epoch": 0.13122408063497296,
+      "epoch": 0.13558456299659477,
       "grad_norm": 0.0,
-      "learning_rate": 1.9467572460718357e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.942120330689631e-05,
+      "loss": 1.043,
       "step": 4778
     },
     {
-      "epoch": 0.13125154486281618,
+      "epoch": 0.13561293984108966,
       "grad_norm": 0.0,
-      "learning_rate": 1.9467286045082455e-05,
-      "loss": 0.9513,
+      "learning_rate": 1.942089512326579e-05,
+      "loss": 1.13,
       "step": 4779
     },
     {
-      "epoch": 0.13127900909065943,
+      "epoch": 0.13564131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.9466999554537718e-05,
-      "loss": 1.1329,
+      "learning_rate": 1.942058686005654e-05,
+      "loss": 0.9643,
       "step": 4780
     },
     {
-      "epoch": 0.13130647331850265,
+      "epoch": 0.13566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.9466712989086414e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.942027851727117e-05,
+      "loss": 0.95,
       "step": 4781
     },
     {
-      "epoch": 0.1313339375463459,
+      "epoch": 0.13569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.946642634873081e-05,
-      "loss": 1.1138,
+      "learning_rate": 1.9419970094912283e-05,
+      "loss": 1.0923,
       "step": 4782
     },
     {
-      "epoch": 0.13136140177418912,
+      "epoch": 0.13572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9466139633473173e-05,
-      "loss": 1.0715,
+      "learning_rate": 1.9419661592982487e-05,
+      "loss": 1.1141,
       "step": 4783
     },
     {
-      "epoch": 0.13138886600203234,
+      "epoch": 0.13575482406356412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9465852843315772e-05,
-      "loss": 0.985,
+      "learning_rate": 1.941935301148439e-05,
+      "loss": 1.061,
       "step": 4784
     },
     {
-      "epoch": 0.1314163302298756,
+      "epoch": 0.13578320090805904,
       "grad_norm": 0.0,
-      "learning_rate": 1.946556597826088e-05,
-      "loss": 1.1209,
+      "learning_rate": 1.9419044350420585e-05,
+      "loss": 0.9884,
       "step": 4785
     },
     {
-      "epoch": 0.1314437944577188,
+      "epoch": 0.13581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.946527903831076e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.9418735609793696e-05,
+      "loss": 1.0316,
       "step": 4786
     },
     {
-      "epoch": 0.13147125868556206,
+      "epoch": 0.1358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9464992023467687e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.941842678960632e-05,
+      "loss": 0.8896,
       "step": 4787
     },
     {
-      "epoch": 0.13149872291340528,
+      "epoch": 0.1358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.946470493373393e-05,
-      "loss": 1.0817,
+      "learning_rate": 1.9418117889861074e-05,
+      "loss": 1.1373,
       "step": 4788
     },
     {
-      "epoch": 0.13152618714124853,
+      "epoch": 0.1358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.9464417769111767e-05,
-      "loss": 1.1335,
+      "learning_rate": 1.941780891056056e-05,
+      "loss": 1.0101,
       "step": 4789
     },
     {
-      "epoch": 0.13155365136909175,
+      "epoch": 0.13592508513053347,
       "grad_norm": 0.0,
-      "learning_rate": 1.9464130529603458e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.9417499851707392e-05,
+      "loss": 0.9325,
       "step": 4790
     },
     {
-      "epoch": 0.131581115596935,
+      "epoch": 0.1359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.946384321521128e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.941719071330418e-05,
+      "loss": 1.0112,
       "step": 4791
     },
     {
-      "epoch": 0.13160857982477822,
+      "epoch": 0.13598183881952328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9463555825937515e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.9416881495353533e-05,
+      "loss": 0.9467,
       "step": 4792
     },
     {
-      "epoch": 0.13163604405262147,
+      "epoch": 0.13601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.9463268361784427e-05,
-      "loss": 1.1577,
+      "learning_rate": 1.9416572197858063e-05,
+      "loss": 1.1531,
       "step": 4793
     },
     {
-      "epoch": 0.1316635082804647,
+      "epoch": 0.13603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.9462980822754295e-05,
-      "loss": 0.8712,
+      "learning_rate": 1.9416262820820386e-05,
+      "loss": 0.9598,
       "step": 4794
     },
     {
-      "epoch": 0.13169097250830794,
+      "epoch": 0.13606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.9462693208849394e-05,
-      "loss": 1.0572,
+      "learning_rate": 1.9415953364243117e-05,
+      "loss": 1.0335,
       "step": 4795
     },
     {
-      "epoch": 0.13171843673615116,
+      "epoch": 0.13609534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.9462405520071998e-05,
-      "loss": 1.1319,
+      "learning_rate": 1.941564382812886e-05,
+      "loss": 1.0066,
       "step": 4796
     },
     {
-      "epoch": 0.13174590096399438,
+      "epoch": 0.13612372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.946211775642438e-05,
-      "loss": 1.1273,
+      "learning_rate": 1.941533421248024e-05,
+      "loss": 1.0315,
       "step": 4797
     },
     {
-      "epoch": 0.13177336519183763,
+      "epoch": 0.13615209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.946182991790883e-05,
-      "loss": 1.0062,
+      "learning_rate": 1.9415024517299872e-05,
+      "loss": 1.0112,
       "step": 4798
     },
     {
-      "epoch": 0.13180082941968085,
+      "epoch": 0.13618047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 1.9461542004527606e-05,
-      "loss": 1.0706,
+      "learning_rate": 1.9414714742590363e-05,
+      "loss": 1.044,
       "step": 4799
     },
     {
-      "epoch": 0.1318282936475241,
+      "epoch": 0.1362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9461254016283006e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.9414404888354338e-05,
+      "loss": 0.8898,
       "step": 4800
     },
     {
-      "epoch": 0.13185575787536732,
+      "epoch": 0.1362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9460965953177292e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.9414094954594413e-05,
+      "loss": 1.001,
       "step": 4801
     },
     {
-      "epoch": 0.13188322210321057,
+      "epoch": 0.13626560726447218,
       "grad_norm": 0.0,
-      "learning_rate": 1.9460677815212754e-05,
-      "loss": 1.1448,
+      "learning_rate": 1.9413784941313203e-05,
+      "loss": 1.0451,
       "step": 4802
     },
     {
-      "epoch": 0.1319106863310538,
+      "epoch": 0.1362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.9460389602391666e-05,
-      "loss": 0.988,
+      "learning_rate": 1.9413474848513326e-05,
+      "loss": 1.0649,
       "step": 4803
     },
     {
-      "epoch": 0.13193815055889704,
+      "epoch": 0.13632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9460101314716314e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.941316467619741e-05,
+      "loss": 1.078,
       "step": 4804
     },
     {
-      "epoch": 0.13196561478674027,
+      "epoch": 0.13635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.9459812952188976e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.9412854424368063e-05,
+      "loss": 1.0681,
       "step": 4805
     },
     {
-      "epoch": 0.13199307901458351,
+      "epoch": 0.13637911464245175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9459524514811932e-05,
-      "loss": 1.1328,
+      "learning_rate": 1.9412544093027915e-05,
+      "loss": 1.0967,
       "step": 4806
     },
     {
-      "epoch": 0.13202054324242674,
+      "epoch": 0.13640749148694664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9459236002587466e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.941223368217958e-05,
+      "loss": 0.9809,
       "step": 4807
     },
     {
-      "epoch": 0.13204800747026998,
+      "epoch": 0.13643586833144156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9458947415517863e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.9411923191825687e-05,
+      "loss": 0.8885,
       "step": 4808
     },
     {
-      "epoch": 0.1320754716981132,
+      "epoch": 0.13646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9458658753605402e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.9411612621968855e-05,
+      "loss": 0.985,
       "step": 4809
     },
     {
-      "epoch": 0.13210293592595646,
+      "epoch": 0.13649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.945837001685237e-05,
-      "loss": 1.0582,
+      "learning_rate": 1.941130197261171e-05,
+      "loss": 0.9578,
       "step": 4810
     },
     {
-      "epoch": 0.13213040015379968,
+      "epoch": 0.13652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9458081205261048e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.941099124375687e-05,
+      "loss": 0.9689,
       "step": 4811
     },
     {
-      "epoch": 0.1321578643816429,
+      "epoch": 0.1365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.9457792318833725e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.941068043540697e-05,
+      "loss": 1.0014,
       "step": 4812
     },
     {
-      "epoch": 0.13218532860948615,
+      "epoch": 0.136577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9457503357572688e-05,
-      "loss": 1.0352,
+      "learning_rate": 1.9410369547564623e-05,
+      "loss": 1.0307,
       "step": 4813
     },
     {
-      "epoch": 0.13221279283732937,
+      "epoch": 0.1366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.9457214321480217e-05,
-      "loss": 1.019,
+      "learning_rate": 1.9410058580232464e-05,
+      "loss": 1.1096,
       "step": 4814
     },
     {
-      "epoch": 0.13224025706517262,
+      "epoch": 0.1366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.9456925210558603e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.940974753341312e-05,
+      "loss": 1.1026,
       "step": 4815
     },
     {
-      "epoch": 0.13226772129301584,
+      "epoch": 0.13666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.9456636024810135e-05,
-      "loss": 1.1736,
+      "learning_rate": 1.940943640710921e-05,
+      "loss": 0.992,
       "step": 4816
     },
     {
-      "epoch": 0.1322951855208591,
+      "epoch": 0.13669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.94563467642371e-05,
-      "loss": 1.1121,
+      "learning_rate": 1.9409125201323372e-05,
+      "loss": 1.0599,
       "step": 4817
     },
     {
-      "epoch": 0.1323226497487023,
+      "epoch": 0.13671963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.9456057428841787e-05,
-      "loss": 1.1137,
+      "learning_rate": 1.940881391605823e-05,
+      "loss": 1.1032,
       "step": 4818
     },
     {
-      "epoch": 0.13235011397654556,
+      "epoch": 0.13674801362088534,
       "grad_norm": 0.0,
-      "learning_rate": 1.9455768018626485e-05,
-      "loss": 1.136,
+      "learning_rate": 1.9408502551316413e-05,
+      "loss": 1.0869,
       "step": 4819
     },
     {
-      "epoch": 0.13237757820438878,
+      "epoch": 0.13677639046538026,
       "grad_norm": 0.0,
-      "learning_rate": 1.945547853359348e-05,
-      "loss": 1.0813,
+      "learning_rate": 1.9408191107100554e-05,
+      "loss": 1.0371,
       "step": 4820
     },
     {
-      "epoch": 0.13240504243223203,
+      "epoch": 0.13680476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9455188973745068e-05,
-      "loss": 1.0784,
+      "learning_rate": 1.940787958341328e-05,
+      "loss": 1.0125,
       "step": 4821
     },
     {
-      "epoch": 0.13243250666007525,
+      "epoch": 0.13683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.945489933908354e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.940756798025722e-05,
+      "loss": 1.1152,
       "step": 4822
     },
     {
-      "epoch": 0.1324599708879185,
+      "epoch": 0.13686152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 1.9454609629611187e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.9407256297635017e-05,
+      "loss": 0.9961,
       "step": 4823
     },
     {
-      "epoch": 0.13248743511576172,
+      "epoch": 0.1368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.9454319845330296e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.9406944535549296e-05,
+      "loss": 0.9787,
       "step": 4824
     },
     {
-      "epoch": 0.13251489934360494,
+      "epoch": 0.13691827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9454029986243164e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.940663269400269e-05,
+      "loss": 0.9865,
       "step": 4825
     },
     {
-      "epoch": 0.1325423635714482,
+      "epoch": 0.1369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.945374005235209e-05,
-      "loss": 1.0368,
+      "learning_rate": 1.940632077299783e-05,
+      "loss": 0.9302,
       "step": 4826
     },
     {
-      "epoch": 0.1325698277992914,
+      "epoch": 0.1369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.945345004365936e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.9406008772537364e-05,
+      "loss": 0.9564,
       "step": 4827
     },
     {
-      "epoch": 0.13259729202713466,
+      "epoch": 0.13700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.945315996016727e-05,
-      "loss": 0.984,
+      "learning_rate": 1.9405696692623915e-05,
+      "loss": 1.0092,
       "step": 4828
     },
     {
-      "epoch": 0.13262475625497788,
+      "epoch": 0.13703178206583427,
       "grad_norm": 0.0,
-      "learning_rate": 1.945286980187812e-05,
-      "loss": 1.137,
+      "learning_rate": 1.9405384533260124e-05,
+      "loss": 0.9861,
       "step": 4829
     },
     {
-      "epoch": 0.13265222048282113,
+      "epoch": 0.13706015891032916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9452579568794198e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.9405072294448627e-05,
+      "loss": 1.0842,
       "step": 4830
     },
     {
-      "epoch": 0.13267968471066435,
+      "epoch": 0.13708853575482408,
       "grad_norm": 0.0,
-      "learning_rate": 1.9452289260917806e-05,
-      "loss": 1.0086,
+      "learning_rate": 1.9404759976192064e-05,
+      "loss": 1.0547,
       "step": 4831
     },
     {
-      "epoch": 0.1327071489385076,
+      "epoch": 0.13711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.9451998878251238e-05,
-      "loss": 1.1174,
+      "learning_rate": 1.9404447578493063e-05,
+      "loss": 1.1016,
       "step": 4832
     },
     {
-      "epoch": 0.13273461316635082,
+      "epoch": 0.13714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.9451708420796797e-05,
-      "loss": 0.9853,
+      "learning_rate": 1.9404135101354277e-05,
+      "loss": 0.9909,
       "step": 4833
     },
     {
-      "epoch": 0.13276207739419407,
+      "epoch": 0.13717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9451417888556777e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.940382254477834e-05,
+      "loss": 1.1479,
       "step": 4834
     },
     {
-      "epoch": 0.1327895416220373,
+      "epoch": 0.13720204313280362,
       "grad_norm": 0.0,
-      "learning_rate": 1.945112728153348e-05,
-      "loss": 1.018,
+      "learning_rate": 1.940350990876789e-05,
+      "loss": 0.9308,
       "step": 4835
     },
     {
-      "epoch": 0.13281700584988054,
+      "epoch": 0.1372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.94508365997292e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.9403197193325566e-05,
+      "loss": 1.0094,
       "step": 4836
     },
     {
-      "epoch": 0.13284447007772376,
+      "epoch": 0.13725879682179343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9450545843146238e-05,
-      "loss": 1.1871,
+      "learning_rate": 1.9402884398454015e-05,
+      "loss": 1.0241,
       "step": 4837
     },
     {
-      "epoch": 0.13287193430556699,
+      "epoch": 0.13728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.94502550117869e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.9402571524155877e-05,
+      "loss": 1.018,
       "step": 4838
     },
     {
-      "epoch": 0.13289939853341023,
+      "epoch": 0.1373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9449964105653482e-05,
-      "loss": 1.088,
+      "learning_rate": 1.9402258570433794e-05,
+      "loss": 1.0446,
       "step": 4839
     },
     {
-      "epoch": 0.13292686276125346,
+      "epoch": 0.1373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.9449673124748286e-05,
-      "loss": 0.92,
+      "learning_rate": 1.940194553729041e-05,
+      "loss": 0.9459,
       "step": 4840
     },
     {
-      "epoch": 0.1329543269890967,
+      "epoch": 0.13737230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 1.944938206907362e-05,
-      "loss": 1.0689,
+      "learning_rate": 1.940163242472837e-05,
+      "loss": 0.9656,
       "step": 4841
     },
     {
-      "epoch": 0.13298179121693993,
+      "epoch": 0.13740068104426786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9449090938631777e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.9401319232750317e-05,
+      "loss": 0.9288,
       "step": 4842
     },
     {
-      "epoch": 0.13300925544478318,
+      "epoch": 0.13742905788876278,
       "grad_norm": 0.0,
-      "learning_rate": 1.9448799733425073e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.9401005961358898e-05,
+      "loss": 1.0375,
       "step": 4843
     },
     {
-      "epoch": 0.1330367196726264,
+      "epoch": 0.13745743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 1.9448508453455802e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.940069261055676e-05,
+      "loss": 1.1296,
       "step": 4844
     },
     {
-      "epoch": 0.13306418390046965,
+      "epoch": 0.13748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.9448217098726272e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.940037918034655e-05,
+      "loss": 1.0042,
       "step": 4845
     },
     {
-      "epoch": 0.13309164812831287,
+      "epoch": 0.13751418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9447925669238786e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.940006567073091e-05,
+      "loss": 1.0621,
       "step": 4846
     },
     {
-      "epoch": 0.13311911235615612,
+      "epoch": 0.13754256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.9447634164995658e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.9399752081712498e-05,
+      "loss": 1.0551,
       "step": 4847
     },
     {
-      "epoch": 0.13314657658399934,
+      "epoch": 0.13757094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9447342585999186e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.9399438413293955e-05,
+      "loss": 0.9393,
       "step": 4848
     },
     {
-      "epoch": 0.1331740408118426,
+      "epoch": 0.13759931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 1.9447050932251677e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.9399124665477934e-05,
+      "loss": 1.0757,
       "step": 4849
     },
     {
-      "epoch": 0.1332015050396858,
+      "epoch": 0.13762769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 1.9446759203755445e-05,
-      "loss": 1.0897,
+      "learning_rate": 1.9398810838267087e-05,
+      "loss": 1.0601,
       "step": 4850
     },
     {
-      "epoch": 0.13322896926752906,
+      "epoch": 0.1376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.9446467400512794e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9398496931664058e-05,
+      "loss": 1.0182,
       "step": 4851
     },
     {
-      "epoch": 0.13325643349537228,
+      "epoch": 0.1376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9446175522526035e-05,
-      "loss": 1.0927,
+      "learning_rate": 1.9398182945671507e-05,
+      "loss": 1.1577,
       "step": 4852
     },
     {
-      "epoch": 0.1332838977232155,
+      "epoch": 0.13771282633371168,
       "grad_norm": 0.0,
-      "learning_rate": 1.944588356979747e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.9397868880292077e-05,
+      "loss": 1.0511,
       "step": 4853
     },
     {
-      "epoch": 0.13331136195105875,
+      "epoch": 0.1377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.9445591542329424e-05,
-      "loss": 1.0906,
+      "learning_rate": 1.939755473552843e-05,
+      "loss": 1.0151,
       "step": 4854
     },
     {
-      "epoch": 0.13333882617890197,
+      "epoch": 0.13776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.9445299440124193e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.9397240511383213e-05,
+      "loss": 0.9753,
       "step": 4855
     },
     {
-      "epoch": 0.13336629040674522,
+      "epoch": 0.13779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.9445007263184097e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.9396926207859085e-05,
+      "loss": 1.0366,
       "step": 4856
     },
     {
-      "epoch": 0.13339375463458844,
+      "epoch": 0.13782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9444715011511443e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9396611824958696e-05,
+      "loss": 1.0108,
       "step": 4857
     },
     {
-      "epoch": 0.1334212188624317,
+      "epoch": 0.13785471055618614,
       "grad_norm": 0.0,
-      "learning_rate": 1.944442268510855e-05,
-      "loss": 1.1974,
+      "learning_rate": 1.939629736268471e-05,
+      "loss": 0.978,
       "step": 4858
     },
     {
-      "epoch": 0.1334486830902749,
+      "epoch": 0.13788308740068103,
       "grad_norm": 0.0,
-      "learning_rate": 1.944413028397772e-05,
-      "loss": 1.008,
+      "learning_rate": 1.9395982821039772e-05,
+      "loss": 1.0273,
       "step": 4859
     },
     {
-      "epoch": 0.13347614731811816,
+      "epoch": 0.13791146424517595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9443837808121276e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.939566820002655e-05,
+      "loss": 1.1221,
       "step": 4860
     },
     {
-      "epoch": 0.13350361154596138,
+      "epoch": 0.13793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.944354525754153e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.9395353499647692e-05,
+      "loss": 1.0351,
       "step": 4861
     },
     {
-      "epoch": 0.13353107577380463,
+      "epoch": 0.13796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9443252632240794e-05,
-      "loss": 1.1989,
+      "learning_rate": 1.9395038719905862e-05,
+      "loss": 1.0343,
       "step": 4862
     },
     {
-      "epoch": 0.13355854000164785,
+      "epoch": 0.1379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9442959932221383e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.939472386080372e-05,
+      "loss": 1.11,
       "step": 4863
     },
     {
-      "epoch": 0.1335860042294911,
+      "epoch": 0.1380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9442667157485616e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.939440892234392e-05,
+      "loss": 1.0529,
       "step": 4864
     },
     {
-      "epoch": 0.13361346845733432,
+      "epoch": 0.1380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.944237430803581e-05,
-      "loss": 1.1148,
+      "learning_rate": 1.939409390452913e-05,
+      "loss": 1.1329,
       "step": 4865
     },
     {
-      "epoch": 0.13364093268517754,
+      "epoch": 0.1380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9442081383874284e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.9393778807362e-05,
+      "loss": 0.9935,
       "step": 4866
     },
     {
-      "epoch": 0.1336683969130208,
+      "epoch": 0.13811010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 1.944178838500335e-05,
-      "loss": 1.1222,
+      "learning_rate": 1.93934636308452e-05,
+      "loss": 1.025,
       "step": 4867
     },
     {
-      "epoch": 0.13369586114086401,
+      "epoch": 0.13813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.944149531142533e-05,
-      "loss": 1.083,
+      "learning_rate": 1.9393148374981395e-05,
+      "loss": 1.0944,
       "step": 4868
     },
     {
-      "epoch": 0.13372332536870726,
+      "epoch": 0.13816685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.9441202163142542e-05,
-      "loss": 1.0036,
+      "learning_rate": 1.939283303977324e-05,
+      "loss": 0.903,
       "step": 4869
     },
     {
-      "epoch": 0.13375078959655048,
+      "epoch": 0.13819523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9440908940157302e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.93925176252234e-05,
+      "loss": 1.1352,
       "step": 4870
     },
     {
-      "epoch": 0.13377825382439373,
+      "epoch": 0.13822360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9440615642471935e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.9392202131334545e-05,
+      "loss": 1.0302,
       "step": 4871
     },
     {
-      "epoch": 0.13380571805223695,
+      "epoch": 0.13825198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 1.944032227008876e-05,
-      "loss": 1.1162,
+      "learning_rate": 1.9391886558109333e-05,
+      "loss": 1.0933,
       "step": 4872
     },
     {
-      "epoch": 0.1338331822800802,
+      "epoch": 0.13828036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.94400288230101e-05,
-      "loss": 1.0425,
+      "learning_rate": 1.939157090555044e-05,
+      "loss": 1.1042,
       "step": 4873
     },
     {
-      "epoch": 0.13386064650792343,
+      "epoch": 0.13830874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9439735301238274e-05,
-      "loss": 0.9439,
+      "learning_rate": 1.9391255173660516e-05,
+      "loss": 1.0828,
       "step": 4874
     },
     {
-      "epoch": 0.13388811073576667,
+      "epoch": 0.1383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.9439441704775605e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.939093936244224e-05,
+      "loss": 0.9267,
       "step": 4875
     },
     {
-      "epoch": 0.1339155749636099,
+      "epoch": 0.1383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9439148033624416e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.939062347189828e-05,
+      "loss": 1.0731,
       "step": 4876
     },
     {
-      "epoch": 0.13394303919145314,
+      "epoch": 0.13839387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 1.9438854287787032e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.9390307502031304e-05,
+      "loss": 1.1428,
       "step": 4877
     },
     {
-      "epoch": 0.13397050341929637,
+      "epoch": 0.138422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.943856046726578e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.9389991452843974e-05,
+      "loss": 1.0053,
       "step": 4878
     },
     {
-      "epoch": 0.1339979676471396,
+      "epoch": 0.1384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.9438266572062976e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.9389675324338965e-05,
+      "loss": 1.05,
       "step": 4879
     },
     {
-      "epoch": 0.13402543187498284,
+      "epoch": 0.13847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9437972602180956e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.9389359116518945e-05,
+      "loss": 1.0345,
       "step": 4880
     },
     {
-      "epoch": 0.13405289610282606,
+      "epoch": 0.13850737797956866,
       "grad_norm": 0.0,
-      "learning_rate": 1.943767855762204e-05,
-      "loss": 1.1875,
+      "learning_rate": 1.938904282938659e-05,
+      "loss": 0.9975,
       "step": 4881
     },
     {
-      "epoch": 0.1340803603306693,
+      "epoch": 0.13853575482406355,
       "grad_norm": 0.0,
-      "learning_rate": 1.943738443838855e-05,
-      "loss": 1.0916,
+      "learning_rate": 1.9388726462944564e-05,
+      "loss": 1.0972,
       "step": 4882
     },
     {
-      "epoch": 0.13410782455851253,
+      "epoch": 0.13856413166855847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9437090244482823e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.9388410017195544e-05,
+      "loss": 0.9411,
       "step": 4883
     },
     {
-      "epoch": 0.13413528878635578,
+      "epoch": 0.13859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.9436795975907184e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.9388093492142205e-05,
+      "loss": 1.0124,
       "step": 4884
     },
     {
-      "epoch": 0.134162753014199,
+      "epoch": 0.13862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9436501632663957e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.9387776887787218e-05,
+      "loss": 1.0184,
       "step": 4885
     },
     {
-      "epoch": 0.13419021724204225,
+      "epoch": 0.13864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.9436207214755474e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.9387460204133254e-05,
+      "loss": 1.0117,
       "step": 4886
     },
     {
-      "epoch": 0.13421768146988547,
+      "epoch": 0.13867763904653801,
       "grad_norm": 0.0,
-      "learning_rate": 1.9435912722184066e-05,
-      "loss": 1.1744,
+      "learning_rate": 1.9387143441183e-05,
+      "loss": 1.0162,
       "step": 4887
     },
     {
-      "epoch": 0.13424514569772872,
+      "epoch": 0.13870601589103293,
       "grad_norm": 0.0,
-      "learning_rate": 1.943561815495206e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.9386826598939114e-05,
+      "loss": 1.0798,
       "step": 4888
     },
     {
-      "epoch": 0.13427260992557194,
+      "epoch": 0.13873439273552782,
       "grad_norm": 0.0,
-      "learning_rate": 1.9435323513061788e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9386509677404286e-05,
+      "loss": 1.0046,
       "step": 4889
     },
     {
-      "epoch": 0.1343000741534152,
+      "epoch": 0.1387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.943502879651558e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.938619267658119e-05,
+      "loss": 1.0789,
       "step": 4890
     },
     {
-      "epoch": 0.1343275383812584,
+      "epoch": 0.1387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.9434734005315774e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.9385875596472502e-05,
+      "loss": 1.0068,
       "step": 4891
     },
     {
-      "epoch": 0.13435500260910163,
+      "epoch": 0.13881952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 1.9434439139464695e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.9385558437080898e-05,
+      "loss": 1.0023,
       "step": 4892
     },
     {
-      "epoch": 0.13438246683694488,
+      "epoch": 0.13884790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 1.9434144198964676e-05,
-      "loss": 1.0489,
+      "learning_rate": 1.938524119840906e-05,
+      "loss": 0.9032,
       "step": 4893
     },
     {
-      "epoch": 0.1344099310647881,
+      "epoch": 0.13887627695800228,
       "grad_norm": 0.0,
-      "learning_rate": 1.943384918381806e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.938492388045967e-05,
+      "loss": 1.0584,
       "step": 4894
     },
     {
-      "epoch": 0.13443739529263135,
+      "epoch": 0.13890465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.9433554094027173e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.9384606483235407e-05,
+      "loss": 0.9827,
       "step": 4895
     },
     {
-      "epoch": 0.13446485952047457,
+      "epoch": 0.13893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9433258929594347e-05,
-      "loss": 0.9376,
+      "learning_rate": 1.9384289006738947e-05,
+      "loss": 1.096,
       "step": 4896
     },
     {
-      "epoch": 0.13449232374831782,
+      "epoch": 0.13896140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.9432963690521927e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.938397145097298e-05,
+      "loss": 0.9829,
       "step": 4897
     },
     {
-      "epoch": 0.13451978797616104,
+      "epoch": 0.13898978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9432668376812243e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.9383653815940184e-05,
+      "loss": 1.1701,
       "step": 4898
     },
     {
-      "epoch": 0.1345472522040043,
+      "epoch": 0.13901816118047672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9432372988467633e-05,
-      "loss": 1.0818,
+      "learning_rate": 1.9383336101643242e-05,
+      "loss": 1.0979,
       "step": 4899
     },
     {
-      "epoch": 0.1345747164318475,
+      "epoch": 0.13904653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 1.9432077525490435e-05,
-      "loss": 1.0987,
+      "learning_rate": 1.9383018308084836e-05,
+      "loss": 1.056,
       "step": 4900
     },
     {
-      "epoch": 0.13460218065969076,
+      "epoch": 0.13907491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9431781987882984e-05,
-      "loss": 1.125,
+      "learning_rate": 1.9382700435267653e-05,
+      "loss": 1.0877,
       "step": 4901
     },
     {
-      "epoch": 0.13462964488753398,
+      "epoch": 0.1391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.943148637564762e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.938238248319438e-05,
+      "loss": 0.9977,
       "step": 4902
     },
     {
-      "epoch": 0.13465710911537723,
+      "epoch": 0.1391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9431190688786685e-05,
-      "loss": 1.1249,
+      "learning_rate": 1.93820644518677e-05,
+      "loss": 1.0111,
       "step": 4903
     },
     {
-      "epoch": 0.13468457334322045,
+      "epoch": 0.13916004540295118,
       "grad_norm": 0.0,
-      "learning_rate": 1.9430894927302513e-05,
-      "loss": 1.1654,
+      "learning_rate": 1.93817463412903e-05,
+      "loss": 0.9663,
       "step": 4904
     },
     {
-      "epoch": 0.1347120375710637,
+      "epoch": 0.1391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.943059909119745e-05,
-      "loss": 1.0954,
+      "learning_rate": 1.9381428151464863e-05,
+      "loss": 1.098,
       "step": 4905
     },
     {
-      "epoch": 0.13473950179890692,
+      "epoch": 0.13921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9430303180473835e-05,
-      "loss": 0.9806,
+      "learning_rate": 1.9381109882394084e-05,
+      "loss": 1.0654,
       "step": 4906
     },
     {
-      "epoch": 0.13476696602675015,
+      "epoch": 0.13924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9430007195134007e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.9380791534080648e-05,
+      "loss": 0.8942,
       "step": 4907
     },
     {
-      "epoch": 0.1347944302545934,
+      "epoch": 0.13927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.9429711135180307e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.9380473106527242e-05,
+      "loss": 1.0557,
       "step": 4908
     },
     {
-      "epoch": 0.13482189448243662,
+      "epoch": 0.13930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.9429415000615083e-05,
-      "loss": 1.1097,
+      "learning_rate": 1.938015459973656e-05,
+      "loss": 1.0863,
       "step": 4909
     },
     {
-      "epoch": 0.13484935871027987,
+      "epoch": 0.13933030646992053,
       "grad_norm": 0.0,
-      "learning_rate": 1.942911879144067e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.937983601371129e-05,
+      "loss": 1.1867,
       "step": 4910
     },
     {
-      "epoch": 0.1348768229381231,
+      "epoch": 0.13935868331441545,
       "grad_norm": 0.0,
-      "learning_rate": 1.942882250765942e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.9379517348454128e-05,
+      "loss": 1.0483,
       "step": 4911
     },
     {
-      "epoch": 0.13490428716596634,
+      "epoch": 0.13938706015891034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9428526149273673e-05,
-      "loss": 1.1712,
+      "learning_rate": 1.9379198603967756e-05,
+      "loss": 1.0158,
       "step": 4912
     },
     {
-      "epoch": 0.13493175139380956,
+      "epoch": 0.13941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.942822971628578e-05,
-      "loss": 1.0529,
+      "learning_rate": 1.937887978025487e-05,
+      "loss": 1.0973,
       "step": 4913
     },
     {
-      "epoch": 0.1349592156216528,
+      "epoch": 0.1394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.9427933208698074e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9378560877318168e-05,
+      "loss": 1.0049,
       "step": 4914
     },
     {
-      "epoch": 0.13498667984949603,
+      "epoch": 0.139472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9427636626512912e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.9378241895160342e-05,
+      "loss": 1.1872,
       "step": 4915
     },
     {
-      "epoch": 0.13501414407733928,
+      "epoch": 0.13950056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 1.9427339969732635e-05,
-      "loss": 1.0014,
+      "learning_rate": 1.9377922833784085e-05,
+      "loss": 1.0562,
       "step": 4916
     },
     {
-      "epoch": 0.1350416083051825,
+      "epoch": 0.1395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9427043238359594e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.937760369319209e-05,
+      "loss": 1.0096,
       "step": 4917
     },
     {
-      "epoch": 0.13506907253302575,
+      "epoch": 0.1395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9426746432396134e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.9377284473387054e-05,
+      "loss": 0.9842,
       "step": 4918
     },
     {
-      "epoch": 0.13509653676086897,
+      "epoch": 0.13958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.9426449551844606e-05,
-      "loss": 1.1585,
+      "learning_rate": 1.9376965174371678e-05,
+      "loss": 0.9394,
       "step": 4919
     },
     {
-      "epoch": 0.1351240009887122,
+      "epoch": 0.13961407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.9426152596707358e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.9376645796148656e-05,
+      "loss": 1.0061,
       "step": 4920
     },
     {
-      "epoch": 0.13515146521655544,
+      "epoch": 0.13964245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9425855566986738e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.937632633872068e-05,
+      "loss": 1.0774,
       "step": 4921
     },
     {
-      "epoch": 0.13517892944439866,
+      "epoch": 0.13967082860385924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9425558462685097e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.937600680209046e-05,
+      "loss": 1.011,
       "step": 4922
     },
     {
-      "epoch": 0.1352063936722419,
+      "epoch": 0.13969920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9425261283804788e-05,
-      "loss": 1.005,
+      "learning_rate": 1.9375687186260684e-05,
+      "loss": 0.8471,
       "step": 4923
     },
     {
-      "epoch": 0.13523385790008513,
+      "epoch": 0.13972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.942496403034816e-05,
-      "loss": 1.0174,
+      "learning_rate": 1.9375367491234058e-05,
+      "loss": 0.8737,
       "step": 4924
     },
     {
-      "epoch": 0.13526132212792838,
+      "epoch": 0.13975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.9424666702317566e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.9375047717013283e-05,
+      "loss": 1.1132,
       "step": 4925
     },
     {
-      "epoch": 0.1352887863557716,
+      "epoch": 0.13978433598183881,
       "grad_norm": 0.0,
-      "learning_rate": 1.942436929971536e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.937472786360106e-05,
+      "loss": 1.0482,
       "step": 4926
     },
     {
-      "epoch": 0.13531625058361485,
+      "epoch": 0.1398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.942407182254389e-05,
-      "loss": 1.05,
+      "learning_rate": 1.9374407931000087e-05,
+      "loss": 1.036,
       "step": 4927
     },
     {
-      "epoch": 0.13534371481145807,
+      "epoch": 0.13984108967082862,
       "grad_norm": 0.0,
-      "learning_rate": 1.9423774270805514e-05,
-      "loss": 1.1177,
+      "learning_rate": 1.937408791921307e-05,
+      "loss": 0.9308,
       "step": 4928
     },
     {
-      "epoch": 0.13537117903930132,
+      "epoch": 0.1398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9423476644502588e-05,
-      "loss": 0.948,
+      "learning_rate": 1.9373767828242708e-05,
+      "loss": 1.0672,
       "step": 4929
     },
     {
-      "epoch": 0.13539864326714454,
+      "epoch": 0.1398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.9423178943637465e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.9373447658091713e-05,
+      "loss": 0.9675,
       "step": 4930
     },
     {
-      "epoch": 0.1354261074949878,
+      "epoch": 0.13992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9422881168212498e-05,
-      "loss": 0.953,
+      "learning_rate": 1.937312740876278e-05,
+      "loss": 0.9927,
       "step": 4931
     },
     {
-      "epoch": 0.135453571722831,
+      "epoch": 0.13995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.942258331823004e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.937280708025862e-05,
+      "loss": 0.9878,
       "step": 4932
     },
     {
-      "epoch": 0.13548103595067423,
+      "epoch": 0.13998297389330305,
       "grad_norm": 0.0,
-      "learning_rate": 1.9422285393692458e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.9372486672581937e-05,
+      "loss": 1.0555,
       "step": 4933
     },
     {
-      "epoch": 0.13550850017851748,
+      "epoch": 0.14001135073779797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9421987394602105e-05,
-      "loss": 0.8815,
+      "learning_rate": 1.9372166185735437e-05,
+      "loss": 1.087,
       "step": 4934
     },
     {
-      "epoch": 0.1355359644063607,
+      "epoch": 0.14003972758229286,
       "grad_norm": 0.0,
-      "learning_rate": 1.942168932096134e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.937184561972183e-05,
+      "loss": 0.9567,
       "step": 4935
     },
     {
-      "epoch": 0.13556342863420395,
+      "epoch": 0.14006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.9421391172772518e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.937152497454382e-05,
+      "loss": 0.963,
       "step": 4936
     },
     {
-      "epoch": 0.13559089286204717,
+      "epoch": 0.14009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.9421092950037996e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.9371204250204116e-05,
+      "loss": 0.942,
       "step": 4937
     },
     {
-      "epoch": 0.13561835708989042,
+      "epoch": 0.14012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.942079465276014e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.937088344670543e-05,
+      "loss": 1.0681,
       "step": 4938
     },
     {
-      "epoch": 0.13564582131773364,
+      "epoch": 0.1401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9420496280941307e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.937056256405047e-05,
+      "loss": 0.9935,
       "step": 4939
     },
     {
-      "epoch": 0.1356732855455769,
+      "epoch": 0.14018161180476732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9420197834583856e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.9370241602241952e-05,
+      "loss": 1.006,
       "step": 4940
     },
     {
-      "epoch": 0.13570074977342012,
+      "epoch": 0.1402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9419899313690152e-05,
-      "loss": 1.0719,
+      "learning_rate": 1.9369920561282575e-05,
+      "loss": 0.9484,
       "step": 4941
     },
     {
-      "epoch": 0.13572821400126336,
+      "epoch": 0.1402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.9419600718262556e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.936959944117506e-05,
+      "loss": 0.9738,
       "step": 4942
     },
     {
-      "epoch": 0.13575567822910659,
+      "epoch": 0.14026674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9419302048303434e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.9369278241922114e-05,
+      "loss": 0.9468,
       "step": 4943
     },
     {
-      "epoch": 0.13578314245694983,
+      "epoch": 0.14029511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 1.941900330381514e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.9368956963526456e-05,
+      "loss": 1.0227,
       "step": 4944
     },
     {
-      "epoch": 0.13581060668479306,
+      "epoch": 0.14032349602724178,
       "grad_norm": 0.0,
-      "learning_rate": 1.9418704484800044e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.93686356059908e-05,
+      "loss": 1.0968,
       "step": 4945
     },
     {
-      "epoch": 0.1358380709126363,
+      "epoch": 0.14035187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.941840559126051e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.9368314169317858e-05,
+      "loss": 1.046,
       "step": 4946
     },
     {
-      "epoch": 0.13586553514047953,
+      "epoch": 0.14038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9418106623198903e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.9367992653510344e-05,
+      "loss": 0.945,
       "step": 4947
     },
     {
-      "epoch": 0.13589299936832275,
+      "epoch": 0.14040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.941780758061759e-05,
-      "loss": 1.1376,
+      "learning_rate": 1.936767105857097e-05,
+      "loss": 1.0165,
       "step": 4948
     },
     {
-      "epoch": 0.135920463596166,
+      "epoch": 0.14043700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.9417508463518936e-05,
-      "loss": 1.1031,
+      "learning_rate": 1.9367349384502462e-05,
+      "loss": 0.9402,
       "step": 4949
     },
     {
-      "epoch": 0.13594792782400922,
+      "epoch": 0.14046538024971622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9417209271905303e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.9367027631307533e-05,
+      "loss": 1.0114,
       "step": 4950
     },
     {
-      "epoch": 0.13597539205185247,
+      "epoch": 0.14049375709421114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9416910005779066e-05,
-      "loss": 1.1143,
+      "learning_rate": 1.9366705798988902e-05,
+      "loss": 1.0457,
       "step": 4951
     },
     {
-      "epoch": 0.1360028562796957,
+      "epoch": 0.14052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.941661066514259e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.9366383887549283e-05,
+      "loss": 1.0184,
       "step": 4952
     },
     {
-      "epoch": 0.13603032050753894,
+      "epoch": 0.1405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.941631124999824e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.93660618969914e-05,
+      "loss": 1.1269,
       "step": 4953
     },
     {
-      "epoch": 0.13605778473538216,
+      "epoch": 0.1405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9416011760348387e-05,
-      "loss": 1.101,
+      "learning_rate": 1.9365739827317967e-05,
+      "loss": 1.1476,
       "step": 4954
     },
     {
-      "epoch": 0.1360852489632254,
+      "epoch": 0.14060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.9415712196195403e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.9365417678531713e-05,
+      "loss": 0.978,
       "step": 4955
     },
     {
-      "epoch": 0.13611271319106863,
+      "epoch": 0.14063564131668557,
       "grad_norm": 0.0,
-      "learning_rate": 1.9415412557541657e-05,
-      "loss": 1.1429,
+      "learning_rate": 1.9365095450635353e-05,
+      "loss": 1.0804,
       "step": 4956
     },
     {
-      "epoch": 0.13614017741891188,
+      "epoch": 0.1406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.941511284438952e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.936477314363161e-05,
+      "loss": 1.0419,
       "step": 4957
     },
     {
-      "epoch": 0.1361676416467551,
+      "epoch": 0.14069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9414813056741363e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.936445075752321e-05,
+      "loss": 0.9638,
       "step": 4958
     },
     {
-      "epoch": 0.13619510587459835,
+      "epoch": 0.14072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.941451319459956e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.9364128292312872e-05,
+      "loss": 0.9981,
       "step": 4959
     },
     {
-      "epoch": 0.13622257010244157,
+      "epoch": 0.14074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9414213257966476e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.9363805748003323e-05,
+      "loss": 1.0549,
       "step": 4960
     },
     {
-      "epoch": 0.1362500343302848,
+      "epoch": 0.14077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.9413913246844496e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.936348312459728e-05,
+      "loss": 0.9509,
       "step": 4961
     },
     {
-      "epoch": 0.13627749855812804,
+      "epoch": 0.14080590238365492,
       "grad_norm": 0.0,
-      "learning_rate": 1.9413613161235983e-05,
-      "loss": 0.993,
+      "learning_rate": 1.936316042209748e-05,
+      "loss": 1.0727,
       "step": 4962
     },
     {
-      "epoch": 0.13630496278597126,
+      "epoch": 0.14083427922814984,
       "grad_norm": 0.0,
-      "learning_rate": 1.941331300114332e-05,
-      "loss": 1.1208,
+      "learning_rate": 1.936283764050664e-05,
+      "loss": 0.9037,
       "step": 4963
     },
     {
-      "epoch": 0.1363324270138145,
+      "epoch": 0.14086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9413012766568874e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.9362514779827494e-05,
+      "loss": 1.0487,
       "step": 4964
     },
     {
-      "epoch": 0.13635989124165773,
+      "epoch": 0.1408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.941271245751503e-05,
-      "loss": 1.036,
+      "learning_rate": 1.936219184006276e-05,
+      "loss": 0.9592,
       "step": 4965
     },
     {
-      "epoch": 0.13638735546950098,
+      "epoch": 0.1409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.9412412073984154e-05,
-      "loss": 1.1038,
+      "learning_rate": 1.9361868821215174e-05,
+      "loss": 0.9466,
       "step": 4966
     },
     {
-      "epoch": 0.1364148196973442,
+      "epoch": 0.1409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9412111615978633e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.9361545723287462e-05,
+      "loss": 1.0604,
       "step": 4967
     },
     {
-      "epoch": 0.13644228392518745,
+      "epoch": 0.1409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9411811083500836e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.9361222546282346e-05,
+      "loss": 1.0846,
       "step": 4968
     },
     {
-      "epoch": 0.13646974815303067,
+      "epoch": 0.1410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.9411510476553143e-05,
-      "loss": 1.0406,
+      "learning_rate": 1.9360899290202568e-05,
+      "loss": 0.8837,
       "step": 4969
     },
     {
-      "epoch": 0.13649721238087392,
+      "epoch": 0.14103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.941120979513793e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.936057595505085e-05,
+      "loss": 0.9858,
       "step": 4970
     },
     {
-      "epoch": 0.13652467660871714,
+      "epoch": 0.14106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.9410909039257587e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.936025254082993e-05,
+      "loss": 0.9212,
       "step": 4971
     },
     {
-      "epoch": 0.1365521408365604,
+      "epoch": 0.14108967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 1.941060820891448e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.9359929047542533e-05,
+      "loss": 1.0105,
       "step": 4972
     },
     {
-      "epoch": 0.13657960506440361,
+      "epoch": 0.14111804767309874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9410307304111003e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.9359605475191396e-05,
+      "loss": 1.0915,
       "step": 4973
     },
     {
-      "epoch": 0.13660706929224684,
+      "epoch": 0.14114642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.9410006324849524e-05,
-      "loss": 1.0603,
+      "learning_rate": 1.935928182377925e-05,
+      "loss": 1.0461,
       "step": 4974
     },
     {
-      "epoch": 0.13663453352009008,
+      "epoch": 0.14117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.940970527113243e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.9358958093308828e-05,
+      "loss": 1.0335,
       "step": 4975
     },
     {
-      "epoch": 0.1366619977479333,
+      "epoch": 0.14120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9409404142962104e-05,
-      "loss": 0.977,
+      "learning_rate": 1.935863428378287e-05,
+      "loss": 1.0894,
       "step": 4976
     },
     {
-      "epoch": 0.13668946197577655,
+      "epoch": 0.14123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.940910294034093e-05,
-      "loss": 1.1447,
+      "learning_rate": 1.9358310395204104e-05,
+      "loss": 0.9928,
       "step": 4977
     },
     {
-      "epoch": 0.13671692620361978,
+      "epoch": 0.1412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9408801663271283e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.935798642757527e-05,
+      "loss": 0.9822,
       "step": 4978
     },
     {
-      "epoch": 0.13674439043146303,
+      "epoch": 0.1412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.9408500311755557e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.9357662380899107e-05,
+      "loss": 0.9913,
       "step": 4979
     },
     {
-      "epoch": 0.13677185465930625,
+      "epoch": 0.141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.940819888579613e-05,
-      "loss": 0.8552,
+      "learning_rate": 1.9357338255178348e-05,
+      "loss": 0.9787,
       "step": 4980
     },
     {
-      "epoch": 0.1367993188871495,
+      "epoch": 0.1413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.9407897385395393e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.935701405041573e-05,
+      "loss": 1.0289,
       "step": 4981
     },
     {
-      "epoch": 0.13682678311499272,
+      "epoch": 0.14137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.940759581055572e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.9356689766613993e-05,
+      "loss": 1.0381,
       "step": 4982
     },
     {
-      "epoch": 0.13685424734283597,
+      "epoch": 0.14140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.940729416127951e-05,
-      "loss": 1.1321,
+      "learning_rate": 1.9356365403775877e-05,
+      "loss": 1.0742,
       "step": 4983
     },
     {
-      "epoch": 0.1368817115706792,
+      "epoch": 0.14143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.9406992437569144e-05,
-      "loss": 1.0921,
+      "learning_rate": 1.935604096190413e-05,
+      "loss": 0.9475,
       "step": 4984
     },
     {
-      "epoch": 0.13690917579852244,
+      "epoch": 0.14145856980703747,
       "grad_norm": 0.0,
-      "learning_rate": 1.940669063942701e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.9355716441001475e-05,
+      "loss": 1.0412,
       "step": 4985
     },
     {
-      "epoch": 0.13693664002636566,
+      "epoch": 0.14148694665153236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9406388766855493e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.9355391841070665e-05,
+      "loss": 1.0302,
       "step": 4986
     },
     {
-      "epoch": 0.13696410425420888,
+      "epoch": 0.14151532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 1.9406086819856988e-05,
-      "loss": 1.0679,
+      "learning_rate": 1.935506716211444e-05,
+      "loss": 1.0808,
       "step": 4987
     },
     {
-      "epoch": 0.13699156848205213,
+      "epoch": 0.14154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.9405784798433882e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.935474240413554e-05,
+      "loss": 1.0117,
       "step": 4988
     },
     {
-      "epoch": 0.13701903270989535,
+      "epoch": 0.14157207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 1.940548270258856e-05,
-      "loss": 1.1936,
+      "learning_rate": 1.9354417567136716e-05,
+      "loss": 1.0195,
       "step": 4989
     },
     {
-      "epoch": 0.1370464969377386,
+      "epoch": 0.1416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.9405180532323416e-05,
-      "loss": 0.9853,
+      "learning_rate": 1.93540926511207e-05,
+      "loss": 0.9507,
       "step": 4990
     },
     {
-      "epoch": 0.13707396116558182,
+      "epoch": 0.14162883087400682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9404878287640843e-05,
-      "loss": 1.0,
+      "learning_rate": 1.935376765609025e-05,
+      "loss": 1.0184,
       "step": 4991
     },
     {
-      "epoch": 0.13710142539342507,
+      "epoch": 0.1416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9404575968543224e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.93534425820481e-05,
+      "loss": 1.1026,
       "step": 4992
     },
     {
-      "epoch": 0.1371288896212683,
+      "epoch": 0.1416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.940427357503296e-05,
-      "loss": 1.0661,
+      "learning_rate": 1.9353117428997e-05,
+      "loss": 1.037,
       "step": 4993
     },
     {
-      "epoch": 0.13715635384911154,
+      "epoch": 0.14171396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 1.9403971107112442e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.9352792196939697e-05,
+      "loss": 1.0868,
       "step": 4994
     },
     {
-      "epoch": 0.13718381807695476,
+      "epoch": 0.14174233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.940366856478406e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.9352466885878937e-05,
+      "loss": 1.0258,
       "step": 4995
     },
     {
-      "epoch": 0.137211282304798,
+      "epoch": 0.14177071509648126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9403365948050214e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.935214149581747e-05,
+      "loss": 0.9849,
       "step": 4996
     },
     {
-      "epoch": 0.13723874653264123,
+      "epoch": 0.14179909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9403063256913288e-05,
-      "loss": 0.8758,
+      "learning_rate": 1.9351816026758042e-05,
+      "loss": 1.0383,
       "step": 4997
     },
     {
-      "epoch": 0.13726621076048448,
+      "epoch": 0.14182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.940276049137569e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.9351490478703403e-05,
+      "loss": 0.8959,
       "step": 4998
     },
     {
-      "epoch": 0.1372936749883277,
+      "epoch": 0.14185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9402457651439804e-05,
-      "loss": 1.1429,
+      "learning_rate": 1.9351164851656304e-05,
+      "loss": 1.0773,
       "step": 4999
     },
     {
-      "epoch": 0.13732113921617095,
+      "epoch": 0.14188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.940215473710803e-05,
-      "loss": 1.0448,
+      "learning_rate": 1.9350839145619496e-05,
+      "loss": 1.0483,
       "step": 5000
     },
     {
-      "epoch": 0.13734860344401417,
+      "epoch": 0.14191259931895572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9401851748382764e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.935051336059573e-05,
+      "loss": 0.996,
       "step": 5001
     },
     {
-      "epoch": 0.1373760676718574,
+      "epoch": 0.1419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.940154868526641e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.9350187496587757e-05,
+      "loss": 0.9728,
       "step": 5002
     },
     {
-      "epoch": 0.13740353189970064,
+      "epoch": 0.14196935300794553,
       "grad_norm": 0.0,
-      "learning_rate": 1.940124554776136e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.934986155359833e-05,
+      "loss": 0.932,
       "step": 5003
     },
     {
-      "epoch": 0.13743099612754386,
+      "epoch": 0.1419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.940094233587001e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.93495355316302e-05,
+      "loss": 1.0408,
       "step": 5004
     },
     {
-      "epoch": 0.1374584603553871,
+      "epoch": 0.1420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9400639049594767e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.9349209430686125e-05,
+      "loss": 1.0301,
       "step": 5005
     },
     {
-      "epoch": 0.13748592458323033,
+      "epoch": 0.1420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.9400335688938026e-05,
-      "loss": 1.0719,
+      "learning_rate": 1.9348883250768858e-05,
+      "loss": 1.0464,
       "step": 5006
     },
     {
-      "epoch": 0.13751338881107358,
+      "epoch": 0.14208286038592507,
       "grad_norm": 0.0,
-      "learning_rate": 1.9400032253902186e-05,
-      "loss": 1.0056,
+      "learning_rate": 1.9348556991881154e-05,
+      "loss": 1.1004,
       "step": 5007
     },
     {
-      "epoch": 0.1375408530389168,
+      "epoch": 0.14211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.939972874448965e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.934823065402577e-05,
+      "loss": 0.9401,
       "step": 5008
     },
     {
-      "epoch": 0.13756831726676005,
+      "epoch": 0.14213961407491488,
       "grad_norm": 0.0,
-      "learning_rate": 1.939942516070282e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.934790423720546e-05,
+      "loss": 0.9591,
       "step": 5009
     },
     {
-      "epoch": 0.13759578149460328,
+      "epoch": 0.14216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9399121502544095e-05,
-      "loss": 1.0143,
+      "learning_rate": 1.9347577741422983e-05,
+      "loss": 0.9517,
       "step": 5010
     },
     {
-      "epoch": 0.13762324572244652,
+      "epoch": 0.14219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.939881777001588e-05,
-      "loss": 0.8642,
+      "learning_rate": 1.9347251166681094e-05,
+      "loss": 0.99,
       "step": 5011
     },
     {
-      "epoch": 0.13765070995028975,
+      "epoch": 0.14222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.939851396312058e-05,
-      "loss": 1.1205,
+      "learning_rate": 1.9346924512982555e-05,
+      "loss": 1.1144,
       "step": 5012
     },
     {
-      "epoch": 0.137678174178133,
+      "epoch": 0.14225312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 1.9398210081860595e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.9346597780330126e-05,
+      "loss": 1.0928,
       "step": 5013
     },
     {
-      "epoch": 0.13770563840597622,
+      "epoch": 0.14228149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9397906126238334e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.9346270968726567e-05,
+      "loss": 0.9828,
       "step": 5014
     },
     {
-      "epoch": 0.13773310263381944,
+      "epoch": 0.14230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9397602096256195e-05,
-      "loss": 1.0632,
+      "learning_rate": 1.9345944078174637e-05,
+      "loss": 0.9921,
       "step": 5015
     },
     {
-      "epoch": 0.1377605668616627,
+      "epoch": 0.14233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.939729799191659e-05,
-      "loss": 1.047,
+      "learning_rate": 1.934561710867709e-05,
+      "loss": 1.0625,
       "step": 5016
     },
     {
-      "epoch": 0.1377880310895059,
+      "epoch": 0.142366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9396993813221928e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.9345290060236706e-05,
+      "loss": 0.8986,
       "step": 5017
     },
     {
-      "epoch": 0.13781549531734916,
+      "epoch": 0.1423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9396689560174604e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.9344962932856228e-05,
+      "loss": 1.0826,
       "step": 5018
     },
     {
-      "epoch": 0.13784295954519238,
+      "epoch": 0.14242338251986378,
       "grad_norm": 0.0,
-      "learning_rate": 1.9396385232777036e-05,
-      "loss": 1.1006,
+      "learning_rate": 1.934463572653843e-05,
+      "loss": 0.9622,
       "step": 5019
     },
     {
-      "epoch": 0.13787042377303563,
+      "epoch": 0.1424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9396080831031625e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.934430844128608e-05,
+      "loss": 1.0456,
       "step": 5020
     },
     {
-      "epoch": 0.13789788800087885,
+      "epoch": 0.14248013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 1.9395776354940787e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.934398107710193e-05,
+      "loss": 1.0391,
       "step": 5021
     },
     {
-      "epoch": 0.1379253522287221,
+      "epoch": 0.14250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9395471804506924e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.9343653633988752e-05,
+      "loss": 1.0834,
       "step": 5022
     },
     {
-      "epoch": 0.13795281645656532,
+      "epoch": 0.14253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.939516717973245e-05,
-      "loss": 1.0988,
+      "learning_rate": 1.9343326111949315e-05,
+      "loss": 0.9883,
       "step": 5023
     },
     {
-      "epoch": 0.13798028068440857,
+      "epoch": 0.14256526674233824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9394862480619775e-05,
-      "loss": 1.0552,
+      "learning_rate": 1.934299851098638e-05,
+      "loss": 1.0815,
       "step": 5024
     },
     {
-      "epoch": 0.1380077449122518,
+      "epoch": 0.14259364358683316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9394557707171305e-05,
-      "loss": 1.0927,
+      "learning_rate": 1.9342670831102716e-05,
+      "loss": 0.9861,
       "step": 5025
     },
     {
-      "epoch": 0.13803520914009504,
+      "epoch": 0.14262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9394252859389454e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.9342343072301096e-05,
+      "loss": 1.0551,
       "step": 5026
     },
     {
-      "epoch": 0.13806267336793826,
+      "epoch": 0.14265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9393947937276643e-05,
-      "loss": 1.01,
+      "learning_rate": 1.9342015234584277e-05,
+      "loss": 0.9713,
       "step": 5027
     },
     {
-      "epoch": 0.13809013759578148,
+      "epoch": 0.14267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.939364294083527e-05,
-      "loss": 1.0595,
+      "learning_rate": 1.934168731795504e-05,
+      "loss": 0.9762,
       "step": 5028
     },
     {
-      "epoch": 0.13811760182362473,
+      "epoch": 0.1427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.9393337870067758e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.934135932241615e-05,
+      "loss": 1.0309,
       "step": 5029
     },
     {
-      "epoch": 0.13814506605146795,
+      "epoch": 0.1427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.939303272497652e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.9341031247970377e-05,
+      "loss": 1.0364,
       "step": 5030
     },
     {
-      "epoch": 0.1381725302793112,
+      "epoch": 0.1427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9392727505563966e-05,
-      "loss": 1.0296,
+      "learning_rate": 1.9340703094620493e-05,
+      "loss": 0.9403,
       "step": 5031
     },
     {
-      "epoch": 0.13819999450715442,
+      "epoch": 0.1427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.9392422211832516e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.9340374862369273e-05,
+      "loss": 0.9427,
       "step": 5032
     },
     {
-      "epoch": 0.13822745873499767,
+      "epoch": 0.14282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.939211684378458e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.934004655121948e-05,
+      "loss": 1.1016,
       "step": 5033
     },
     {
-      "epoch": 0.1382549229628409,
+      "epoch": 0.14284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.939181140142258e-05,
-      "loss": 1.2201,
+      "learning_rate": 1.93397181611739e-05,
+      "loss": 0.9761,
       "step": 5034
     },
     {
-      "epoch": 0.13828238719068414,
+      "epoch": 0.14287741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9391505884748928e-05,
-      "loss": 1.1107,
+      "learning_rate": 1.93393896922353e-05,
+      "loss": 1.0286,
       "step": 5035
     },
     {
-      "epoch": 0.13830985141852736,
+      "epoch": 0.14290578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.9391200293766047e-05,
-      "loss": 0.972,
+      "learning_rate": 1.9339061144406455e-05,
+      "loss": 0.9762,
       "step": 5036
     },
     {
-      "epoch": 0.1383373156463706,
+      "epoch": 0.14293416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9390894628476347e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.9338732517690138e-05,
+      "loss": 1.02,
       "step": 5037
     },
     {
-      "epoch": 0.13836477987421383,
+      "epoch": 0.14296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9390588888882254e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.933840381208913e-05,
+      "loss": 1.128,
       "step": 5038
     },
     {
-      "epoch": 0.13839224410205708,
+      "epoch": 0.14299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.9390283074986183e-05,
-      "loss": 0.9652,
+      "learning_rate": 1.933807502760621e-05,
+      "loss": 1.0542,
       "step": 5039
     },
     {
-      "epoch": 0.1384197083299003,
+      "epoch": 0.14301929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9389977186790558e-05,
-      "loss": 1.1281,
+      "learning_rate": 1.933774616424414e-05,
+      "loss": 1.0672,
       "step": 5040
     },
     {
-      "epoch": 0.13844717255774353,
+      "epoch": 0.1430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.938967122429779e-05,
-      "loss": 1.1887,
+      "learning_rate": 1.9337417222005715e-05,
+      "loss": 1.0054,
       "step": 5041
     },
     {
-      "epoch": 0.13847463678558677,
+      "epoch": 0.1430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.938936518751031e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.9337088200893706e-05,
+      "loss": 1.0143,
       "step": 5042
     },
     {
-      "epoch": 0.13850210101343,
+      "epoch": 0.1431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9389059076430536e-05,
-      "loss": 1.175,
+      "learning_rate": 1.9336759100910893e-05,
+      "loss": 1.0265,
       "step": 5043
     },
     {
-      "epoch": 0.13852956524127324,
+      "epoch": 0.1431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9388752891060888e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.9336429922060055e-05,
+      "loss": 1.0917,
       "step": 5044
     },
     {
-      "epoch": 0.13855702946911647,
+      "epoch": 0.143161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.938844663140379e-05,
-      "loss": 1.1286,
+      "learning_rate": 1.9336100664343976e-05,
+      "loss": 1.0579,
       "step": 5045
     },
     {
-      "epoch": 0.13858449369695972,
+      "epoch": 0.14318955732122587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9388140297461665e-05,
-      "loss": 1.1108,
+      "learning_rate": 1.9335771327765433e-05,
+      "loss": 1.0947,
       "step": 5046
     },
     {
-      "epoch": 0.13861195792480294,
+      "epoch": 0.14321793416572076,
       "grad_norm": 0.0,
-      "learning_rate": 1.938783388923694e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.933544191232721e-05,
+      "loss": 0.9694,
       "step": 5047
     },
     {
-      "epoch": 0.13863942215264619,
+      "epoch": 0.14324631101021568,
       "grad_norm": 0.0,
-      "learning_rate": 1.9387527406732035e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.933511241803209e-05,
+      "loss": 0.9744,
       "step": 5048
     },
     {
-      "epoch": 0.1386668863804894,
+      "epoch": 0.14327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9387220849949378e-05,
-      "loss": 1.1241,
+      "learning_rate": 1.9334782844882856e-05,
+      "loss": 0.9749,
       "step": 5049
     },
     {
-      "epoch": 0.13869435060833266,
+      "epoch": 0.14330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9386914218891393e-05,
-      "loss": 1.1083,
+      "learning_rate": 1.933445319288229e-05,
+      "loss": 0.98,
       "step": 5050
     },
     {
-      "epoch": 0.13872181483617588,
+      "epoch": 0.14333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9386607513560505e-05,
-      "loss": 1.2162,
+      "learning_rate": 1.9334123462033184e-05,
+      "loss": 1.0035,
       "step": 5051
     },
     {
-      "epoch": 0.13874927906401913,
+      "epoch": 0.14335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.9386300733959146e-05,
-      "loss": 1.1519,
+      "learning_rate": 1.9333793652338313e-05,
+      "loss": 1.0417,
       "step": 5052
     },
     {
-      "epoch": 0.13877674329186235,
+      "epoch": 0.1433881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.9385993880089734e-05,
-      "loss": 0.9806,
+      "learning_rate": 1.9333463763800467e-05,
+      "loss": 1.0192,
       "step": 5053
     },
     {
-      "epoch": 0.1388042075197056,
+      "epoch": 0.14341657207718503,
       "grad_norm": 0.0,
-      "learning_rate": 1.9385686951954706e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.9333133796422437e-05,
+      "loss": 1.0464,
       "step": 5054
     },
     {
-      "epoch": 0.13883167174754882,
+      "epoch": 0.14344494892167992,
       "grad_norm": 0.0,
-      "learning_rate": 1.938537994955649e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.9332803750207002e-05,
+      "loss": 0.9957,
       "step": 5055
     },
     {
-      "epoch": 0.13885913597539204,
+      "epoch": 0.1434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9385072872897507e-05,
-      "loss": 1.0845,
+      "learning_rate": 1.9332473625156957e-05,
+      "loss": 1.0678,
       "step": 5056
     },
     {
-      "epoch": 0.1388866002032353,
+      "epoch": 0.1435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9384765721980193e-05,
-      "loss": 1.1386,
+      "learning_rate": 1.9332143421275084e-05,
+      "loss": 0.9483,
       "step": 5057
     },
     {
-      "epoch": 0.1389140644310785,
+      "epoch": 0.14353007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 1.938445849680698e-05,
-      "loss": 1.1503,
+      "learning_rate": 1.9331813138564177e-05,
+      "loss": 0.9503,
       "step": 5058
     },
     {
-      "epoch": 0.13894152865892176,
+      "epoch": 0.14355845629965946,
       "grad_norm": 0.0,
-      "learning_rate": 1.9384151197380296e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.933148277702703e-05,
+      "loss": 1.0764,
       "step": 5059
     },
     {
-      "epoch": 0.13896899288676498,
+      "epoch": 0.14358683314415438,
       "grad_norm": 0.0,
-      "learning_rate": 1.938384382370257e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.9331152336666422e-05,
+      "loss": 0.9857,
       "step": 5060
     },
     {
-      "epoch": 0.13899645711460823,
+      "epoch": 0.14361520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 1.938353637577624e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.9330821817485154e-05,
+      "loss": 0.9607,
       "step": 5061
     },
     {
-      "epoch": 0.13902392134245145,
+      "epoch": 0.14364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9383228853603732e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.9330491219486013e-05,
+      "loss": 0.9927,
       "step": 5062
     },
     {
-      "epoch": 0.1390513855702947,
+      "epoch": 0.14367196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.9382921257187485e-05,
-      "loss": 1.1368,
+      "learning_rate": 1.9330160542671794e-05,
+      "loss": 1.1264,
       "step": 5063
     },
     {
-      "epoch": 0.13907884979813792,
+      "epoch": 0.14370034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.938261358652993e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.932982978704529e-05,
+      "loss": 1.0888,
       "step": 5064
     },
     {
-      "epoch": 0.13910631402598117,
+      "epoch": 0.14372871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.93823058416335e-05,
-      "loss": 1.1027,
+      "learning_rate": 1.9329498952609295e-05,
+      "loss": 0.9701,
       "step": 5065
     },
     {
-      "epoch": 0.1391337782538244,
+      "epoch": 0.14375709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9381998022500634e-05,
-      "loss": 1.035,
+      "learning_rate": 1.9329168039366602e-05,
+      "loss": 0.9874,
       "step": 5066
     },
     {
-      "epoch": 0.13916124248166764,
+      "epoch": 0.14378547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.9381690129133765e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.932883704732001e-05,
+      "loss": 1.0935,
       "step": 5067
     },
     {
-      "epoch": 0.13918870670951086,
+      "epoch": 0.1438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.938138216153533e-05,
-      "loss": 1.054,
+      "learning_rate": 1.9328505976472307e-05,
+      "loss": 1.1889,
       "step": 5068
     },
     {
-      "epoch": 0.13921617093735408,
+      "epoch": 0.1438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9381074119707763e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.9328174826826297e-05,
+      "loss": 0.9474,
       "step": 5069
     },
     {
-      "epoch": 0.13924363516519733,
+      "epoch": 0.14387060158910328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9380766003653507e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.9327843598384775e-05,
+      "loss": 1.0393,
       "step": 5070
     },
     {
-      "epoch": 0.13927109939304055,
+      "epoch": 0.1438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.938045781337499e-05,
-      "loss": 1.114,
+      "learning_rate": 1.9327512291150537e-05,
+      "loss": 1.028,
       "step": 5071
     },
     {
-      "epoch": 0.1392985636208838,
+      "epoch": 0.14392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.9380149548874665e-05,
-      "loss": 1.0806,
+      "learning_rate": 1.9327180905126386e-05,
+      "loss": 1.0393,
       "step": 5072
     },
     {
-      "epoch": 0.13932602784872702,
+      "epoch": 0.14395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.937984121015496e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.932684944031512e-05,
+      "loss": 1.1185,
       "step": 5073
     },
     {
-      "epoch": 0.13935349207657027,
+      "epoch": 0.14398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.937953279721832e-05,
-      "loss": 1.053,
+      "learning_rate": 1.9326517896719533e-05,
+      "loss": 0.9554,
       "step": 5074
     },
     {
-      "epoch": 0.1393809563044135,
+      "epoch": 0.14401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.937922431006718e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.932618627434243e-05,
+      "loss": 1.021,
       "step": 5075
     },
     {
-      "epoch": 0.13940842053225674,
+      "epoch": 0.14404086265607263,
       "grad_norm": 0.0,
-      "learning_rate": 1.9378915748703985e-05,
-      "loss": 1.0899,
+      "learning_rate": 1.9325854573186618e-05,
+      "loss": 0.8838,
       "step": 5076
     },
     {
-      "epoch": 0.13943588476009997,
+      "epoch": 0.14406923950056755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9378607113131178e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.932552279325489e-05,
+      "loss": 1.054,
       "step": 5077
     },
     {
-      "epoch": 0.13946334898794321,
+      "epoch": 0.14409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9378298403351192e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.932519093455005e-05,
+      "loss": 0.9839,
       "step": 5078
     },
     {
-      "epoch": 0.13949081321578644,
+      "epoch": 0.14412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9377989619366483e-05,
-      "loss": 1.102,
+      "learning_rate": 1.9324858997074904e-05,
+      "loss": 1.0461,
       "step": 5079
     },
     {
-      "epoch": 0.13951827744362968,
+      "epoch": 0.1441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9377680761179485e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.9324526980832256e-05,
+      "loss": 0.9269,
       "step": 5080
     },
     {
-      "epoch": 0.1395457416714729,
+      "epoch": 0.1441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.9377371828792644e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.932419488582491e-05,
+      "loss": 1.0455,
       "step": 5081
     },
     {
-      "epoch": 0.13957320589931613,
+      "epoch": 0.14421112372304198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9377062822208404e-05,
-      "loss": 1.055,
+      "learning_rate": 1.9323862712055668e-05,
+      "loss": 1.0242,
       "step": 5082
     },
     {
-      "epoch": 0.13960067012715938,
+      "epoch": 0.1442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.937675374142921e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.9323530459527342e-05,
+      "loss": 1.0342,
       "step": 5083
     },
     {
-      "epoch": 0.1396281343550026,
+      "epoch": 0.1442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.9376444586457515e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.932319812824273e-05,
+      "loss": 0.9563,
       "step": 5084
     },
     {
-      "epoch": 0.13965559858284585,
+      "epoch": 0.14429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.937613535729575e-05,
-      "loss": 1.1298,
+      "learning_rate": 1.932286571820465e-05,
+      "loss": 0.9883,
       "step": 5085
     },
     {
-      "epoch": 0.13968306281068907,
+      "epoch": 0.14432463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9375826053946376e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.93225332294159e-05,
+      "loss": 0.9598,
       "step": 5086
     },
     {
-      "epoch": 0.13971052703853232,
+      "epoch": 0.14435300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.937551667641183e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.9322200661879296e-05,
+      "loss": 0.964,
       "step": 5087
     },
     {
-      "epoch": 0.13973799126637554,
+      "epoch": 0.14438138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9375207224694564e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.9321868015597642e-05,
+      "loss": 0.9982,
       "step": 5088
     },
     {
-      "epoch": 0.1397654554942188,
+      "epoch": 0.14440976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.937489769879703e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.9321535290573748e-05,
+      "loss": 1.0737,
       "step": 5089
     },
     {
-      "epoch": 0.139792919722062,
+      "epoch": 0.14443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9374588098721676e-05,
-      "loss": 0.893,
+      "learning_rate": 1.9321202486810427e-05,
+      "loss": 0.9344,
       "step": 5090
     },
     {
-      "epoch": 0.13982038394990526,
+      "epoch": 0.14446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9374278424470946e-05,
-      "loss": 1.1495,
+      "learning_rate": 1.9320869604310495e-05,
+      "loss": 0.8775,
       "step": 5091
     },
     {
-      "epoch": 0.13984784817774848,
+      "epoch": 0.1444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.9373968676047292e-05,
-      "loss": 1.082,
+      "learning_rate": 1.9320536643076756e-05,
+      "loss": 0.9296,
       "step": 5092
     },
     {
-      "epoch": 0.13987531240559173,
+      "epoch": 0.1445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.937365885345317e-05,
-      "loss": 1.0378,
+      "learning_rate": 1.932020360311202e-05,
+      "loss": 1.1159,
       "step": 5093
     },
     {
-      "epoch": 0.13990277663343495,
+      "epoch": 0.14455164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 1.9373348956691026e-05,
-      "loss": 0.9831,
+      "learning_rate": 1.931987048441911e-05,
+      "loss": 1.0958,
       "step": 5094
     },
     {
-      "epoch": 0.1399302408612782,
+      "epoch": 0.1445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9373038985763317e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.9319537287000832e-05,
+      "loss": 0.9436,
       "step": 5095
     },
     {
-      "epoch": 0.13995770508912142,
+      "epoch": 0.1446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.937272894067249e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.9319204010860007e-05,
+      "loss": 1.0235,
       "step": 5096
     },
     {
-      "epoch": 0.13998516931696464,
+      "epoch": 0.14463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.9372418821421002e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.9318870655999445e-05,
+      "loss": 1.1387,
       "step": 5097
     },
     {
-      "epoch": 0.1400126335448079,
+      "epoch": 0.14466515323496026,
       "grad_norm": 0.0,
-      "learning_rate": 1.9372108628011307e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.9318537222421964e-05,
+      "loss": 1.0562,
       "step": 5098
     },
     {
-      "epoch": 0.1400400977726511,
+      "epoch": 0.14469353007945515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9371798360445856e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.931820371013038e-05,
+      "loss": 0.9832,
       "step": 5099
     },
     {
-      "epoch": 0.14006756200049436,
+      "epoch": 0.14472190692395007,
       "grad_norm": 0.0,
-      "learning_rate": 1.937148801872711e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.9317870119127508e-05,
+      "loss": 1.0508,
       "step": 5100
     },
     {
-      "epoch": 0.14009502622833758,
+      "epoch": 0.14475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9371177602857515e-05,
-      "loss": 1.0708,
+      "learning_rate": 1.931753644941617e-05,
+      "loss": 1.0927,
       "step": 5101
     },
     {
-      "epoch": 0.14012249045618083,
+      "epoch": 0.14477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9370867112839535e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.9317202700999186e-05,
+      "loss": 0.9924,
       "step": 5102
     },
     {
-      "epoch": 0.14014995468402405,
+      "epoch": 0.14480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9370556548675624e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.9316868873879372e-05,
+      "loss": 0.9238,
       "step": 5103
     },
     {
-      "epoch": 0.1401774189118673,
+      "epoch": 0.14483541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.9370245910368243e-05,
-      "loss": 1.1772,
+      "learning_rate": 1.9316534968059542e-05,
+      "loss": 1.0301,
       "step": 5104
     },
     {
-      "epoch": 0.14020488313971052,
+      "epoch": 0.14486379114642453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9369935197919844e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.9316200983542525e-05,
+      "loss": 1.0003,
       "step": 5105
     },
     {
-      "epoch": 0.14023234736755377,
+      "epoch": 0.14489216799091942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9369624411332888e-05,
-      "loss": 0.8866,
+      "learning_rate": 1.9315866920331143e-05,
+      "loss": 0.9599,
       "step": 5106
     },
     {
-      "epoch": 0.140259811595397,
+      "epoch": 0.1449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9369313550609834e-05,
-      "loss": 1.028,
+      "learning_rate": 1.931553277842821e-05,
+      "loss": 0.975,
       "step": 5107
     },
     {
-      "epoch": 0.14028727582324024,
+      "epoch": 0.1449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.9369002615753145e-05,
-      "loss": 1.0157,
+      "learning_rate": 1.9315198557836555e-05,
+      "loss": 0.9938,
       "step": 5108
     },
     {
-      "epoch": 0.14031474005108346,
+      "epoch": 0.14497729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 1.9368691606765274e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.9314864258558998e-05,
+      "loss": 0.9626,
       "step": 5109
     },
     {
-      "epoch": 0.14034220427892669,
+      "epoch": 0.14500567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.9368380523648688e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.931452988059836e-05,
+      "loss": 0.9448,
       "step": 5110
     },
     {
-      "epoch": 0.14036966850676993,
+      "epoch": 0.14503405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9368069366405847e-05,
-      "loss": 0.992,
+      "learning_rate": 1.9314195423957475e-05,
+      "loss": 1.0648,
       "step": 5111
     },
     {
-      "epoch": 0.14039713273461316,
+      "epoch": 0.14506242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9367758135039213e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.9313860888639164e-05,
+      "loss": 0.9891,
       "step": 5112
     },
     {
-      "epoch": 0.1404245969624564,
+      "epoch": 0.14509080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.9367446829551244e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.9313526274646247e-05,
+      "loss": 1.0041,
       "step": 5113
     },
     {
-      "epoch": 0.14045206119029963,
+      "epoch": 0.14511918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.936713544994441e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.9313191581981554e-05,
+      "loss": 1.0038,
       "step": 5114
     },
     {
-      "epoch": 0.14047952541814288,
+      "epoch": 0.14514755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.936682399622117e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.9312856810647918e-05,
+      "loss": 1.0034,
       "step": 5115
     },
     {
-      "epoch": 0.1405069896459861,
+      "epoch": 0.14517593643586832,
       "grad_norm": 0.0,
-      "learning_rate": 1.936651246838399e-05,
-      "loss": 1.0362,
+      "learning_rate": 1.931252196064816e-05,
+      "loss": 0.9598,
       "step": 5116
     },
     {
-      "epoch": 0.14053445387382935,
+      "epoch": 0.14520431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 1.936620086643534e-05,
-      "loss": 1.0329,
+      "learning_rate": 1.9312187031985105e-05,
+      "loss": 1.0912,
       "step": 5117
     },
     {
-      "epoch": 0.14056191810167257,
+      "epoch": 0.14523269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.9365889190377672e-05,
-      "loss": 1.1109,
+      "learning_rate": 1.9311852024661594e-05,
+      "loss": 1.1048,
       "step": 5118
     },
     {
-      "epoch": 0.14058938232951582,
+      "epoch": 0.145261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.936557744021347e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.931151693868045e-05,
+      "loss": 0.8706,
       "step": 5119
     },
     {
-      "epoch": 0.14061684655735904,
+      "epoch": 0.1452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9365265615945185e-05,
-      "loss": 0.9546,
+      "learning_rate": 1.93111817740445e-05,
+      "loss": 1.0604,
       "step": 5120
     },
     {
-      "epoch": 0.1406443107852023,
+      "epoch": 0.14531782065834278,
       "grad_norm": 0.0,
-      "learning_rate": 1.936495371757529e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.931084653075658e-05,
+      "loss": 0.9897,
       "step": 5121
     },
     {
-      "epoch": 0.1406717750130455,
+      "epoch": 0.14534619750283767,
       "grad_norm": 0.0,
-      "learning_rate": 1.936464174510626e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.931051120881952e-05,
+      "loss": 0.9938,
       "step": 5122
     },
     {
-      "epoch": 0.14069923924088873,
+      "epoch": 0.14537457434733259,
       "grad_norm": 0.0,
-      "learning_rate": 1.9364329698540553e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.9310175808236157e-05,
+      "loss": 0.9562,
       "step": 5123
     },
     {
-      "epoch": 0.14072670346873198,
+      "epoch": 0.14540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.936401757788064e-05,
-      "loss": 0.9818,
+      "learning_rate": 1.930984032900932e-05,
+      "loss": 0.9933,
       "step": 5124
     },
     {
-      "epoch": 0.1407541676965752,
+      "epoch": 0.14543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9363705383129e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.9309504771141844e-05,
+      "loss": 1.0159,
       "step": 5125
     },
     {
-      "epoch": 0.14078163192441845,
+      "epoch": 0.14545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.9363393114288088e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.9309169134636558e-05,
+      "loss": 0.9857,
       "step": 5126
     },
     {
-      "epoch": 0.14080909615226167,
+      "epoch": 0.14548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.9363080771360385e-05,
-      "loss": 1.1971,
+      "learning_rate": 1.9308833419496307e-05,
+      "loss": 0.8425,
       "step": 5127
     },
     {
-      "epoch": 0.14083656038010492,
+      "epoch": 0.14551645856980705,
       "grad_norm": 0.0,
-      "learning_rate": 1.9362768354348362e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.9308497625723918e-05,
+      "loss": 1.0358,
       "step": 5128
     },
     {
-      "epoch": 0.14086402460794814,
+      "epoch": 0.14554483541430194,
       "grad_norm": 0.0,
-      "learning_rate": 1.936245586325449e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.9308161753322232e-05,
+      "loss": 1.0035,
       "step": 5129
     },
     {
-      "epoch": 0.1408914888357914,
+      "epoch": 0.14557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9362143298081238e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.9307825802294083e-05,
+      "loss": 1.0593,
       "step": 5130
     },
     {
-      "epoch": 0.1409189530636346,
+      "epoch": 0.1456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.936183065883108e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.9307489772642317e-05,
+      "loss": 0.9749,
       "step": 5131
     },
     {
-      "epoch": 0.14094641729147786,
+      "epoch": 0.1456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.936151794550649e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.9307153664369762e-05,
+      "loss": 1.007,
       "step": 5132
     },
     {
-      "epoch": 0.14097388151932108,
+      "epoch": 0.1456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.936120515810995e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.930681747747926e-05,
+      "loss": 0.9774,
       "step": 5133
     },
     {
-      "epoch": 0.14100134574716433,
+      "epoch": 0.1456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9360892296643926e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.9306481211973658e-05,
+      "loss": 1.0429,
       "step": 5134
     },
     {
-      "epoch": 0.14102880997500755,
+      "epoch": 0.1457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.9360579361110896e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.930614486785579e-05,
+      "loss": 1.0511,
       "step": 5135
     },
     {
-      "epoch": 0.14105627420285077,
+      "epoch": 0.14574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9360266351513335e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.9305808445128495e-05,
+      "loss": 1.0343,
       "step": 5136
     },
     {
-      "epoch": 0.14108373843069402,
+      "epoch": 0.14577185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.935995326785372e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.9305471943794617e-05,
+      "loss": 1.0117,
       "step": 5137
     },
     {
-      "epoch": 0.14111120265853724,
+      "epoch": 0.14580022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9359640110134534e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.9305135363857e-05,
+      "loss": 1.0472,
       "step": 5138
     },
     {
-      "epoch": 0.1411386668863805,
+      "epoch": 0.14582860385925084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9359326878358246e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.9304798705318487e-05,
+      "loss": 0.9442,
       "step": 5139
     },
     {
-      "epoch": 0.14116613111422371,
+      "epoch": 0.14585698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9359013572527337e-05,
-      "loss": 1.08,
+      "learning_rate": 1.9304461968181923e-05,
+      "loss": 1.0387,
       "step": 5140
     },
     {
-      "epoch": 0.14119359534206696,
+      "epoch": 0.14588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.935870019264429e-05,
-      "loss": 1.0272,
+      "learning_rate": 1.9304125152450147e-05,
+      "loss": 0.9883,
       "step": 5141
     },
     {
-      "epoch": 0.14122105956991018,
+      "epoch": 0.14591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.935838673871158e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.930378825812601e-05,
+      "loss": 1.0001,
       "step": 5142
     },
     {
-      "epoch": 0.14124852379775343,
+      "epoch": 0.14594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.935807321073169e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.9303451285212356e-05,
+      "loss": 1.1179,
       "step": 5143
     },
     {
-      "epoch": 0.14127598802559665,
+      "epoch": 0.1459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.93577596087071e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.930311423371203e-05,
+      "loss": 0.9242,
       "step": 5144
     },
     {
-      "epoch": 0.1413034522534399,
+      "epoch": 0.14599886492622022,
       "grad_norm": 0.0,
-      "learning_rate": 1.9357445932640292e-05,
-      "loss": 1.0413,
+      "learning_rate": 1.9302777103627876e-05,
+      "loss": 1.0788,
       "step": 5145
     },
     {
-      "epoch": 0.14133091648128313,
+      "epoch": 0.1460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.9357132182533746e-05,
-      "loss": 1.072,
+      "learning_rate": 1.930243989496275e-05,
+      "loss": 1.0625,
       "step": 5146
     },
     {
-      "epoch": 0.14135838070912637,
+      "epoch": 0.14605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9356818358389943e-05,
-      "loss": 1.0215,
+      "learning_rate": 1.9302102607719494e-05,
+      "loss": 1.0856,
       "step": 5147
     },
     {
-      "epoch": 0.1413858449369696,
+      "epoch": 0.14608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.9356504460211375e-05,
-      "loss": 0.9266,
+      "learning_rate": 1.930176524190096e-05,
+      "loss": 1.0537,
       "step": 5148
     },
     {
-      "epoch": 0.14141330916481284,
+      "epoch": 0.14611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.935619048800051e-05,
-      "loss": 1.0878,
+      "learning_rate": 1.930142779751e-05,
+      "loss": 1.0195,
       "step": 5149
     },
     {
-      "epoch": 0.14144077339265607,
+      "epoch": 0.14614074914869465,
       "grad_norm": 0.0,
-      "learning_rate": 1.9355876441759852e-05,
-      "loss": 0.962,
+      "learning_rate": 1.9301090274549453e-05,
+      "loss": 1.0358,
       "step": 5150
     },
     {
-      "epoch": 0.1414682376204993,
+      "epoch": 0.14616912599318957,
       "grad_norm": 0.0,
-      "learning_rate": 1.9355562321491868e-05,
-      "loss": 1.0962,
+      "learning_rate": 1.9300752673022187e-05,
+      "loss": 0.9857,
       "step": 5151
     },
     {
-      "epoch": 0.14149570184834254,
+      "epoch": 0.14619750283768446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9355248127199055e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.930041499293104e-05,
+      "loss": 1.0426,
       "step": 5152
     },
     {
-      "epoch": 0.14152316607618576,
+      "epoch": 0.14622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9354933858883894e-05,
-      "loss": 1.0646,
+      "learning_rate": 1.930007723427887e-05,
+      "loss": 1.001,
       "step": 5153
     },
     {
-      "epoch": 0.141550630304029,
+      "epoch": 0.14625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.935461951654887e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.9299739397068535e-05,
+      "loss": 1.0639,
       "step": 5154
     },
     {
-      "epoch": 0.14157809453187223,
+      "epoch": 0.14628263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 1.9354305100196475e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.9299401481302874e-05,
+      "loss": 1.0866,
       "step": 5155
     },
     {
-      "epoch": 0.14160555875971548,
+      "epoch": 0.146311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.9353990609829198e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.9299063486984756e-05,
+      "loss": 1.0325,
       "step": 5156
     },
     {
-      "epoch": 0.1416330229875587,
+      "epoch": 0.14633938706015892,
       "grad_norm": 0.0,
-      "learning_rate": 1.9353676045449516e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.929872541411703e-05,
+      "loss": 1.1312,
       "step": 5157
     },
     {
-      "epoch": 0.14166048721540195,
+      "epoch": 0.1463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.935336140705993e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.9298387262702555e-05,
+      "loss": 0.9934,
       "step": 5158
     },
     {
-      "epoch": 0.14168795144324517,
+      "epoch": 0.1463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9353046694662926e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.929804903274418e-05,
+      "loss": 1.014,
       "step": 5159
     },
     {
-      "epoch": 0.14171541567108842,
+      "epoch": 0.14642451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 1.9352731908260995e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.9297710724244768e-05,
+      "loss": 1.0228,
       "step": 5160
     },
     {
-      "epoch": 0.14174287989893164,
+      "epoch": 0.14645289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 1.9352417047856624e-05,
-      "loss": 1.0062,
+      "learning_rate": 1.929737233720718e-05,
+      "loss": 1.1325,
       "step": 5161
     },
     {
-      "epoch": 0.1417703441267749,
+      "epoch": 0.14648127128263336,
       "grad_norm": 0.0,
-      "learning_rate": 1.9352102113452304e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.9297033871634266e-05,
+      "loss": 1.0166,
       "step": 5162
     },
     {
-      "epoch": 0.1417978083546181,
+      "epoch": 0.14650964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.935178710505053e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.9296695327528888e-05,
+      "loss": 1.0214,
       "step": 5163
     },
     {
-      "epoch": 0.14182527258246133,
+      "epoch": 0.14653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9351472022653797e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.929635670489391e-05,
+      "loss": 1.0709,
       "step": 5164
     },
     {
-      "epoch": 0.14185273681030458,
+      "epoch": 0.14656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.9351156866264595e-05,
-      "loss": 0.9089,
+      "learning_rate": 1.9296018003732185e-05,
+      "loss": 1.0356,
       "step": 5165
     },
     {
-      "epoch": 0.1418802010381478,
+      "epoch": 0.14659477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9350841635885418e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.9295679224046582e-05,
+      "loss": 0.9978,
       "step": 5166
     },
     {
-      "epoch": 0.14190766526599105,
+      "epoch": 0.14662315550510782,
       "grad_norm": 0.0,
-      "learning_rate": 1.9350526331518756e-05,
-      "loss": 1.0303,
+      "learning_rate": 1.9295340365839955e-05,
+      "loss": 1.0266,
       "step": 5167
     },
     {
-      "epoch": 0.14193512949383427,
+      "epoch": 0.14665153234960274,
       "grad_norm": 0.0,
-      "learning_rate": 1.935021095316711e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.9295001429115175e-05,
+      "loss": 0.9261,
       "step": 5168
     },
     {
-      "epoch": 0.14196259372167752,
+      "epoch": 0.14667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.9349895500832976e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.9294662413875097e-05,
+      "loss": 0.9827,
       "step": 5169
     },
     {
-      "epoch": 0.14199005794952074,
+      "epoch": 0.1467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.9349579974518844e-05,
-      "loss": 1.0639,
+      "learning_rate": 1.9294323320122586e-05,
+      "loss": 1.0001,
       "step": 5170
     },
     {
-      "epoch": 0.142017522177364,
+      "epoch": 0.1467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9349264374227212e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.929398414786051e-05,
+      "loss": 0.9579,
       "step": 5171
     },
     {
-      "epoch": 0.1420449864052072,
+      "epoch": 0.14676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.9348948699960582e-05,
-      "loss": 1.0603,
+      "learning_rate": 1.9293644897091735e-05,
+      "loss": 1.1112,
       "step": 5172
     },
     {
-      "epoch": 0.14207245063305046,
+      "epoch": 0.14679341657207717,
       "grad_norm": 0.0,
-      "learning_rate": 1.9348632951721446e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.9293305567819118e-05,
+      "loss": 0.9832,
       "step": 5173
     },
     {
-      "epoch": 0.14209991486089368,
+      "epoch": 0.1468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.9348317129512306e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.9292966160045537e-05,
+      "loss": 1.1111,
       "step": 5174
     },
     {
-      "epoch": 0.14212737908873693,
+      "epoch": 0.14685017026106698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9348001233335657e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.9292626673773845e-05,
+      "loss": 0.9146,
       "step": 5175
     },
     {
-      "epoch": 0.14215484331658015,
+      "epoch": 0.14687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9347685263194008e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.9292287109006926e-05,
+      "loss": 1.0672,
       "step": 5176
     },
     {
-      "epoch": 0.14218230754442338,
+      "epoch": 0.14690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9347369219089845e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.929194746574764e-05,
+      "loss": 1.1445,
       "step": 5177
     },
     {
-      "epoch": 0.14220977177226662,
+      "epoch": 0.14693530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.934705310102568e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.929160774399885e-05,
+      "loss": 1.1132,
       "step": 5178
     },
     {
-      "epoch": 0.14223723600010985,
+      "epoch": 0.14696367763904652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9346736909004006e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.9291267943763435e-05,
+      "loss": 1.051,
       "step": 5179
     },
     {
-      "epoch": 0.1422647002279531,
+      "epoch": 0.14699205448354144,
       "grad_norm": 0.0,
-      "learning_rate": 1.934642064302733e-05,
-      "loss": 1.099,
+      "learning_rate": 1.929092806504426e-05,
+      "loss": 1.0509,
       "step": 5180
     },
     {
-      "epoch": 0.14229216445579632,
+      "epoch": 0.14702043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 1.934610430309816e-05,
-      "loss": 1.0553,
+      "learning_rate": 1.9290588107844203e-05,
+      "loss": 1.0723,
       "step": 5181
     },
     {
-      "epoch": 0.14231962868363957,
+      "epoch": 0.14704880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.9345787889218984e-05,
-      "loss": 0.9303,
+      "learning_rate": 1.9290248072166125e-05,
+      "loss": 1.0098,
       "step": 5182
     },
     {
-      "epoch": 0.1423470929114828,
+      "epoch": 0.1470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.934547140139232e-05,
-      "loss": 1.1046,
+      "learning_rate": 1.9289907958012907e-05,
+      "loss": 1.1193,
       "step": 5183
     },
     {
-      "epoch": 0.14237455713932604,
+      "epoch": 0.147105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9345154839620663e-05,
-      "loss": 1.0794,
+      "learning_rate": 1.9289567765387417e-05,
+      "loss": 1.1517,
       "step": 5184
     },
     {
-      "epoch": 0.14240202136716926,
+      "epoch": 0.1471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.934483820390652e-05,
-      "loss": 1.005,
+      "learning_rate": 1.928922749429253e-05,
+      "loss": 1.0365,
       "step": 5185
     },
     {
-      "epoch": 0.1424294855950125,
+      "epoch": 0.1471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.9344521494252403e-05,
-      "loss": 1.1174,
+      "learning_rate": 1.9288887144731126e-05,
+      "loss": 1.0754,
       "step": 5186
     },
     {
-      "epoch": 0.14245694982285573,
+      "epoch": 0.14719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.9344204710660807e-05,
-      "loss": 1.052,
+      "learning_rate": 1.928854671670607e-05,
+      "loss": 1.0329,
       "step": 5187
     },
     {
-      "epoch": 0.14248441405069898,
+      "epoch": 0.14721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.9343887853134245e-05,
-      "loss": 1.122,
+      "learning_rate": 1.928820621022024e-05,
+      "loss": 1.097,
       "step": 5188
     },
     {
-      "epoch": 0.1425118782785422,
+      "epoch": 0.14724744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9343570921675222e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.928786562527652e-05,
+      "loss": 0.9505,
       "step": 5189
     },
     {
-      "epoch": 0.14253934250638542,
+      "epoch": 0.14727582292849034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9343253916286246e-05,
-      "loss": 1.072,
+      "learning_rate": 1.928752496187778e-05,
+      "loss": 1.0678,
       "step": 5190
     },
     {
-      "epoch": 0.14256680673422867,
+      "epoch": 0.14730419977298526,
       "grad_norm": 0.0,
-      "learning_rate": 1.934293683696983e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.9287184220026898e-05,
+      "loss": 1.1303,
       "step": 5191
     },
     {
-      "epoch": 0.1425942709620719,
+      "epoch": 0.14733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.9342619683728473e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.9286843399726754e-05,
+      "loss": 1.0931,
       "step": 5192
     },
     {
-      "epoch": 0.14262173518991514,
+      "epoch": 0.14736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.9342302456564696e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9286502500980226e-05,
+      "loss": 1.0196,
       "step": 5193
     },
     {
-      "epoch": 0.14264919941775836,
+      "epoch": 0.14738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9341985155480996e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.9286161523790197e-05,
+      "loss": 1.0154,
       "step": 5194
     },
     {
-      "epoch": 0.1426766636456016,
+      "epoch": 0.1474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9341667780479896e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.928582046815954e-05,
+      "loss": 1.0045,
       "step": 5195
     },
     {
-      "epoch": 0.14270412787344483,
+      "epoch": 0.1474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.9341350331563902e-05,
-      "loss": 1.1,
+      "learning_rate": 1.9285479334091142e-05,
+      "loss": 0.9854,
       "step": 5196
     },
     {
-      "epoch": 0.14273159210128808,
+      "epoch": 0.1474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.9341032808735522e-05,
-      "loss": 0.9996,
+      "learning_rate": 1.928513812158788e-05,
+      "loss": 1.0277,
       "step": 5197
     },
     {
-      "epoch": 0.1427590563291313,
+      "epoch": 0.1475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9340715211997273e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.9284796830652642e-05,
+      "loss": 0.9273,
       "step": 5198
     },
     {
-      "epoch": 0.14278652055697455,
+      "epoch": 0.14753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.934039754135167e-05,
-      "loss": 1.0799,
+      "learning_rate": 1.9284455461288307e-05,
+      "loss": 1.1288,
       "step": 5199
     },
     {
-      "epoch": 0.14281398478481777,
+      "epoch": 0.14755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.934007979680122e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.928411401349776e-05,
+      "loss": 0.9738,
       "step": 5200
     },
     {
-      "epoch": 0.14284144901266102,
+      "epoch": 0.14758796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.9339761978348442e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.9283772487283885e-05,
+      "loss": 1.0229,
       "step": 5201
     },
     {
-      "epoch": 0.14286891324050424,
+      "epoch": 0.14761634506242904,
       "grad_norm": 0.0,
-      "learning_rate": 1.933944408599585e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.9283430882649562e-05,
+      "loss": 0.9583,
       "step": 5202
     },
     {
-      "epoch": 0.1428963774683475,
+      "epoch": 0.14764472190692396,
       "grad_norm": 0.0,
-      "learning_rate": 1.933912611974596e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.9283089199597688e-05,
+      "loss": 0.9956,
       "step": 5203
     },
     {
-      "epoch": 0.1429238416961907,
+      "epoch": 0.14767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.9338808079601287e-05,
-      "loss": 1.1337,
+      "learning_rate": 1.928274743813114e-05,
+      "loss": 1.0528,
       "step": 5204
     },
     {
-      "epoch": 0.14295130592403393,
+      "epoch": 0.14770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9338489965564344e-05,
-      "loss": 1.1956,
+      "learning_rate": 1.92824055982528e-05,
+      "loss": 1.0279,
       "step": 5205
     },
     {
-      "epoch": 0.14297877015187718,
+      "epoch": 0.14772985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 1.933817177763765e-05,
-      "loss": 1.0619,
+      "learning_rate": 1.9282063679965573e-05,
+      "loss": 1.0109,
       "step": 5206
     },
     {
-      "epoch": 0.1430062343797204,
+      "epoch": 0.1477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.9337853515823727e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.928172168327233e-05,
+      "loss": 1.072,
       "step": 5207
     },
     {
-      "epoch": 0.14303369860756365,
+      "epoch": 0.14778660612939842,
       "grad_norm": 0.0,
-      "learning_rate": 1.9337535180125088e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.928137960817597e-05,
+      "loss": 0.9688,
       "step": 5208
     },
     {
-      "epoch": 0.14306116283540687,
+      "epoch": 0.1478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9337216770544255e-05,
-      "loss": 1.1825,
+      "learning_rate": 1.9281037454679377e-05,
+      "loss": 0.9887,
       "step": 5209
     },
     {
-      "epoch": 0.14308862706325012,
+      "epoch": 0.1478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.9336898287083744e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.9280695222785443e-05,
+      "loss": 1.055,
       "step": 5210
     },
     {
-      "epoch": 0.14311609129109334,
+      "epoch": 0.14787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.9336579729746077e-05,
-      "loss": 1.0028,
+      "learning_rate": 1.9280352912497062e-05,
+      "loss": 0.9918,
       "step": 5211
     },
     {
-      "epoch": 0.1431435555189366,
+      "epoch": 0.14790011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9336261098533774e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.928001052381712e-05,
+      "loss": 1.0181,
       "step": 5212
     },
     {
-      "epoch": 0.14317101974677982,
+      "epoch": 0.14792849035187286,
       "grad_norm": 0.0,
-      "learning_rate": 1.933594239344936e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.9279668056748514e-05,
+      "loss": 1.0592,
       "step": 5213
     },
     {
-      "epoch": 0.14319848397462306,
+      "epoch": 0.14795686719636778,
       "grad_norm": 0.0,
-      "learning_rate": 1.933562361449535e-05,
-      "loss": 1.0411,
+      "learning_rate": 1.9279325511294137e-05,
+      "loss": 1.1319,
       "step": 5214
     },
     {
-      "epoch": 0.14322594820246629,
+      "epoch": 0.14798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.933530476167427e-05,
-      "loss": 0.986,
+      "learning_rate": 1.9278982887456877e-05,
+      "loss": 0.9644,
       "step": 5215
     },
     {
-      "epoch": 0.14325341243030953,
+      "epoch": 0.14801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9334985834988646e-05,
-      "loss": 1.0885,
+      "learning_rate": 1.927864018523963e-05,
+      "loss": 1.1096,
       "step": 5216
     },
     {
-      "epoch": 0.14328087665815276,
+      "epoch": 0.14804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9334666834440996e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.9278297404645294e-05,
+      "loss": 0.9535,
       "step": 5217
     },
     {
-      "epoch": 0.14330834088599598,
+      "epoch": 0.14807037457434732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9334347760033845e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.9277954545676763e-05,
+      "loss": 1.0555,
       "step": 5218
     },
     {
-      "epoch": 0.14333580511383923,
+      "epoch": 0.1480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.933402861176972e-05,
-      "loss": 1.0473,
+      "learning_rate": 1.9277611608336935e-05,
+      "loss": 0.9209,
       "step": 5219
     },
     {
-      "epoch": 0.14336326934168245,
+      "epoch": 0.14812712826333713,
       "grad_norm": 0.0,
-      "learning_rate": 1.9333709389651147e-05,
-      "loss": 0.979,
+      "learning_rate": 1.92772685926287e-05,
+      "loss": 1.0107,
       "step": 5220
     },
     {
-      "epoch": 0.1433907335695257,
+      "epoch": 0.14815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.9333390093680648e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.9276925498554963e-05,
+      "loss": 1.0768,
       "step": 5221
     },
     {
-      "epoch": 0.14341819779736892,
+      "epoch": 0.1481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9333070723860747e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.927658232611862e-05,
+      "loss": 0.9861,
       "step": 5222
     },
     {
-      "epoch": 0.14344566202521217,
+      "epoch": 0.1482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.933275128019398e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.9276239075322568e-05,
+      "loss": 1.018,
       "step": 5223
     },
     {
-      "epoch": 0.1434731262530554,
+      "epoch": 0.14824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 1.9332431762682868e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.927589574616971e-05,
+      "loss": 0.9665,
       "step": 5224
     },
     {
-      "epoch": 0.14350059048089864,
+      "epoch": 0.1482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.933211217132994e-05,
-      "loss": 1.1664,
+      "learning_rate": 1.927555233866294e-05,
+      "loss": 0.9908,
       "step": 5225
     },
     {
-      "epoch": 0.14352805470874186,
+      "epoch": 0.14829738933030648,
       "grad_norm": 0.0,
-      "learning_rate": 1.933179250613773e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.927520885280516e-05,
+      "loss": 1.0481,
       "step": 5226
     },
     {
-      "epoch": 0.1435555189365851,
+      "epoch": 0.14832576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.9331472767108758e-05,
-      "loss": 0.8824,
+      "learning_rate": 1.9274865288599282e-05,
+      "loss": 1.0257,
       "step": 5227
     },
     {
-      "epoch": 0.14358298316442833,
+      "epoch": 0.14835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.933115295424556e-05,
-      "loss": 1.1028,
+      "learning_rate": 1.927452164604819e-05,
+      "loss": 0.9236,
       "step": 5228
     },
     {
-      "epoch": 0.14361044739227158,
+      "epoch": 0.14838251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.9330833067550666e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.9274177925154806e-05,
+      "loss": 0.9915,
       "step": 5229
     },
     {
-      "epoch": 0.1436379116201148,
+      "epoch": 0.14841089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9330513107026607e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.9273834125922017e-05,
+      "loss": 1.0778,
       "step": 5230
     },
     {
-      "epoch": 0.14366537584795802,
+      "epoch": 0.14843927355278094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9330193072675912e-05,
-      "loss": 1.1626,
+      "learning_rate": 1.927349024835274e-05,
+      "loss": 0.9812,
       "step": 5231
     },
     {
-      "epoch": 0.14369284007580127,
+      "epoch": 0.14846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 1.932987296450112e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.927314629244987e-05,
+      "loss": 1.0294,
       "step": 5232
     },
     {
-      "epoch": 0.1437203043036445,
+      "epoch": 0.14849602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 1.9329552782504755e-05,
-      "loss": 1.1156,
+      "learning_rate": 1.927280225821632e-05,
+      "loss": 0.9571,
       "step": 5233
     },
     {
-      "epoch": 0.14374776853148774,
+      "epoch": 0.1485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9329232526689354e-05,
-      "loss": 1.0732,
+      "learning_rate": 1.927245814565499e-05,
+      "loss": 0.9497,
       "step": 5234
     },
     {
-      "epoch": 0.14377523275933096,
+      "epoch": 0.1485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.932891219705745e-05,
-      "loss": 1.0945,
+      "learning_rate": 1.9272113954768786e-05,
+      "loss": 1.0724,
       "step": 5235
     },
     {
-      "epoch": 0.1438026969871742,
+      "epoch": 0.14858115777525538,
       "grad_norm": 0.0,
-      "learning_rate": 1.932859179361158e-05,
-      "loss": 0.954,
+      "learning_rate": 1.9271769685560623e-05,
+      "loss": 1.058,
       "step": 5236
     },
     {
-      "epoch": 0.14383016121501743,
+      "epoch": 0.1486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.9328271316354283e-05,
-      "loss": 1.0859,
+      "learning_rate": 1.927142533803341e-05,
+      "loss": 1.1655,
       "step": 5237
     },
     {
-      "epoch": 0.14385762544286068,
+      "epoch": 0.14863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.932795076528809e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.927108091219004e-05,
+      "loss": 0.9962,
       "step": 5238
     },
     {
-      "epoch": 0.1438850896707039,
+      "epoch": 0.14866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.932763014041553e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.927073640803344e-05,
+      "loss": 1.0688,
       "step": 5239
     },
     {
-      "epoch": 0.14391255389854715,
+      "epoch": 0.14869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.932730944173915e-05,
-      "loss": 1.0859,
+      "learning_rate": 1.9270391825566506e-05,
+      "loss": 0.9183,
       "step": 5240
     },
     {
-      "epoch": 0.14394001812639037,
+      "epoch": 0.14872304199772984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9326988669261488e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.9270047164792163e-05,
+      "loss": 1.0336,
       "step": 5241
     },
     {
-      "epoch": 0.14396748235423362,
+      "epoch": 0.14875141884222473,
       "grad_norm": 0.0,
-      "learning_rate": 1.9326667822985078e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.926970242571331e-05,
+      "loss": 0.9821,
       "step": 5242
     },
     {
-      "epoch": 0.14399494658207684,
+      "epoch": 0.14877979568671965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9326346902912454e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.9269357608332867e-05,
+      "loss": 0.9,
       "step": 5243
     },
     {
-      "epoch": 0.1440224108099201,
+      "epoch": 0.14880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9326025909046167e-05,
-      "loss": 0.9881,
+      "learning_rate": 1.926901271265374e-05,
+      "loss": 0.8723,
       "step": 5244
     },
     {
-      "epoch": 0.14404987503776331,
+      "epoch": 0.14883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9325704841388744e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.9268667738678846e-05,
+      "loss": 1.0503,
       "step": 5245
     },
     {
-      "epoch": 0.14407733926560654,
+      "epoch": 0.1488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.9325383699942738e-05,
-      "loss": 1.0736,
+      "learning_rate": 1.92683226864111e-05,
+      "loss": 1.0022,
       "step": 5246
     },
     {
-      "epoch": 0.14410480349344978,
+      "epoch": 0.1488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9325062484710682e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.9267977555853418e-05,
+      "loss": 0.9291,
       "step": 5247
     },
     {
-      "epoch": 0.144132267721293,
+      "epoch": 0.1489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9324741195695118e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.926763234700871e-05,
+      "loss": 1.1389,
       "step": 5248
     },
     {
-      "epoch": 0.14415973194913625,
+      "epoch": 0.148950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9324419832898593e-05,
-      "loss": 1.0631,
+      "learning_rate": 1.92672870598799e-05,
+      "loss": 0.9551,
       "step": 5249
     },
     {
-      "epoch": 0.14418719617697948,
+      "epoch": 0.14897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.932409839632364e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.9266941694469895e-05,
+      "loss": 0.9963,
       "step": 5250
     },
     {
-      "epoch": 0.14421466040482273,
+      "epoch": 0.14900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9323776885972814e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.9266596250781616e-05,
+      "loss": 0.974,
       "step": 5251
     },
     {
-      "epoch": 0.14424212463266595,
+      "epoch": 0.14903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.9323455301848653e-05,
-      "loss": 0.995,
+      "learning_rate": 1.9266250728817985e-05,
+      "loss": 0.9374,
       "step": 5252
     },
     {
-      "epoch": 0.1442695888605092,
+      "epoch": 0.14906356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 1.9323133643953703e-05,
-      "loss": 1.0513,
+      "learning_rate": 1.9265905128581917e-05,
+      "loss": 1.0415,
       "step": 5253
     },
     {
-      "epoch": 0.14429705308835242,
+      "epoch": 0.14909194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9322811912290504e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.926555945007633e-05,
+      "loss": 1.0622,
       "step": 5254
     },
     {
-      "epoch": 0.14432451731619567,
+      "epoch": 0.14912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9322490106861608e-05,
-      "loss": 1.0324,
+      "learning_rate": 1.9265213693304147e-05,
+      "loss": 0.9156,
       "step": 5255
     },
     {
-      "epoch": 0.1443519815440389,
+      "epoch": 0.14914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.932216822766956e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.926486785826829e-05,
+      "loss": 0.9388,
       "step": 5256
     },
     {
-      "epoch": 0.14437944577188214,
+      "epoch": 0.14917707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 1.9321846274716906e-05,
-      "loss": 1.059,
+      "learning_rate": 1.9264521944971673e-05,
+      "loss": 1.1249,
       "step": 5257
     },
     {
-      "epoch": 0.14440690999972536,
+      "epoch": 0.149205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9321524248006192e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.9264175953417222e-05,
+      "loss": 1.0027,
       "step": 5258
     },
     {
-      "epoch": 0.14443437422756858,
+      "epoch": 0.1492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.932120214753997e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.9263829883607867e-05,
+      "loss": 0.9301,
       "step": 5259
     },
     {
-      "epoch": 0.14446183845541183,
+      "epoch": 0.1492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.932087997332078e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.9263483735546518e-05,
+      "loss": 1.0985,
       "step": 5260
     },
     {
-      "epoch": 0.14448930268325505,
+      "epoch": 0.1492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9320557725351184e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.926313750923611e-05,
+      "loss": 1.1348,
       "step": 5261
     },
     {
-      "epoch": 0.1445167669110983,
+      "epoch": 0.1493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.932023540363372e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.926279120467956e-05,
+      "loss": 1.0101,
       "step": 5262
     },
     {
-      "epoch": 0.14454423113894152,
+      "epoch": 0.14934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9319913008170944e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.9262444821879796e-05,
+      "loss": 1.0244,
       "step": 5263
     },
     {
-      "epoch": 0.14457169536678477,
+      "epoch": 0.14937570942111236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9319590538965407e-05,
-      "loss": 0.9099,
+      "learning_rate": 1.9262098360839747e-05,
+      "loss": 1.0991,
       "step": 5264
     },
     {
-      "epoch": 0.144599159594628,
+      "epoch": 0.14940408626560728,
       "grad_norm": 0.0,
-      "learning_rate": 1.931926799601966e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.9261751821562336e-05,
+      "loss": 1.0307,
       "step": 5265
     },
     {
-      "epoch": 0.14462662382247124,
+      "epoch": 0.14943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9318945379336257e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.926140520405049e-05,
+      "loss": 1.0335,
       "step": 5266
     },
     {
-      "epoch": 0.14465408805031446,
+      "epoch": 0.14946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.9318622688917743e-05,
-      "loss": 1.0373,
+      "learning_rate": 1.9261058508307137e-05,
+      "loss": 0.9919,
       "step": 5267
     },
     {
-      "epoch": 0.1446815522781577,
+      "epoch": 0.14948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.931829992476668e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.926071173433521e-05,
+      "loss": 0.9973,
       "step": 5268
     },
     {
-      "epoch": 0.14470901650600093,
+      "epoch": 0.14951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.931797708688562e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.926036488213763e-05,
+      "loss": 1.0249,
       "step": 5269
     },
     {
-      "epoch": 0.14473648073384418,
+      "epoch": 0.14954597048808171,
       "grad_norm": 0.0,
-      "learning_rate": 1.931765417527711e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.9260017951717334e-05,
+      "loss": 0.9984,
       "step": 5270
     },
     {
-      "epoch": 0.1447639449616874,
+      "epoch": 0.14957434733257663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9317331189943717e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.925967094307725e-05,
+      "loss": 0.889,
       "step": 5271
     },
     {
-      "epoch": 0.14479140918953062,
+      "epoch": 0.14960272417707152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9317008130887986e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.925932385622031e-05,
+      "loss": 1.1302,
       "step": 5272
     },
     {
-      "epoch": 0.14481887341737387,
+      "epoch": 0.1496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.931668499811248e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.9258976691149446e-05,
+      "loss": 1.0418,
       "step": 5273
     },
     {
-      "epoch": 0.1448463376452171,
+      "epoch": 0.1496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9316361791619754e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.9258629447867588e-05,
+      "loss": 1.0199,
       "step": 5274
     },
     {
-      "epoch": 0.14487380187306034,
+      "epoch": 0.14968785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 1.9316038511412364e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.925828212637767e-05,
+      "loss": 1.0109,
       "step": 5275
     },
     {
-      "epoch": 0.14490126610090356,
+      "epoch": 0.14971623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.9315715157492864e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.9257934726682627e-05,
+      "loss": 1.0767,
       "step": 5276
     },
     {
-      "epoch": 0.1449287303287468,
+      "epoch": 0.14974460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 1.9315391729863822e-05,
-      "loss": 1.0858,
+      "learning_rate": 1.9257587248785396e-05,
+      "loss": 0.9549,
       "step": 5277
     },
     {
-      "epoch": 0.14495619455659003,
+      "epoch": 0.14977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.931506822852779e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.9257239692688907e-05,
+      "loss": 0.9442,
       "step": 5278
     },
     {
-      "epoch": 0.14498365878443328,
+      "epoch": 0.14980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.9314744653487335e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.9256892058396098e-05,
+      "loss": 1.0399,
       "step": 5279
     },
     {
-      "epoch": 0.1450111230122765,
+      "epoch": 0.14982973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.9314421004745007e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.9256544345909904e-05,
+      "loss": 1.0143,
       "step": 5280
     },
     {
-      "epoch": 0.14503858724011975,
+      "epoch": 0.14985811577752553,
       "grad_norm": 0.0,
-      "learning_rate": 1.9314097282303373e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.9256196555233268e-05,
+      "loss": 0.8882,
       "step": 5281
     },
     {
-      "epoch": 0.14506605146796298,
+      "epoch": 0.14988649262202042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9313773486164993e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.925584868636912e-05,
+      "loss": 0.9648,
       "step": 5282
     },
     {
-      "epoch": 0.14509351569580622,
+      "epoch": 0.14991486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 1.9313449616332432e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.9255500739320405e-05,
+      "loss": 1.1407,
       "step": 5283
     },
     {
-      "epoch": 0.14512097992364945,
+      "epoch": 0.14994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 1.931312567280825e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.9255152714090056e-05,
+      "loss": 0.9287,
       "step": 5284
     },
     {
-      "epoch": 0.14514844415149267,
+      "epoch": 0.1499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.9312801655595006e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.925480461068102e-05,
+      "loss": 0.9946,
       "step": 5285
     },
     {
-      "epoch": 0.14517590837933592,
+      "epoch": 0.15,
       "grad_norm": 0.0,
-      "learning_rate": 1.931247756469527e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.925445642909623e-05,
+      "loss": 0.9367,
       "step": 5286
     },
     {
-      "epoch": 0.14520337260717914,
+      "epoch": 0.15002837684449488,
       "grad_norm": 0.0,
-      "learning_rate": 1.93121534001116e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.925410816933863e-05,
+      "loss": 1.0344,
       "step": 5287
     },
     {
-      "epoch": 0.1452308368350224,
+      "epoch": 0.1500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.9311829161846573e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.9253759831411166e-05,
+      "loss": 1.0266,
       "step": 5288
     },
     {
-      "epoch": 0.1452583010628656,
+      "epoch": 0.15008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.9311504849902743e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.9253411415316772e-05,
+      "loss": 0.9715,
       "step": 5289
     },
     {
-      "epoch": 0.14528576529070886,
+      "epoch": 0.15011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.931118046428268e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.9253062921058402e-05,
+      "loss": 0.9268,
       "step": 5290
     },
     {
-      "epoch": 0.14531322951855208,
+      "epoch": 0.15014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.931085600498895e-05,
-      "loss": 0.96,
+      "learning_rate": 1.925271434863899e-05,
+      "loss": 1.0503,
       "step": 5291
     },
     {
-      "epoch": 0.14534069374639533,
+      "epoch": 0.15017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.931053147202412e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.925236569806148e-05,
+      "loss": 0.9456,
       "step": 5292
     },
     {
-      "epoch": 0.14536815797423855,
+      "epoch": 0.15019863791146423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9310206865390757e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.9252016969328826e-05,
+      "loss": 1.0093,
       "step": 5293
     },
     {
-      "epoch": 0.1453956222020818,
+      "epoch": 0.15022701475595915,
       "grad_norm": 0.0,
-      "learning_rate": 1.9309882185091434e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.925166816244397e-05,
+      "loss": 1.0473,
       "step": 5294
     },
     {
-      "epoch": 0.14542308642992502,
+      "epoch": 0.15025539160045404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9309557431128716e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.9251319277409855e-05,
+      "loss": 1.1249,
       "step": 5295
     },
     {
-      "epoch": 0.14545055065776827,
+      "epoch": 0.15028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.9309232603505173e-05,
-      "loss": 0.9777,
+      "learning_rate": 1.925097031422943e-05,
+      "loss": 0.8892,
       "step": 5296
     },
     {
-      "epoch": 0.1454780148856115,
+      "epoch": 0.1503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.9308907702223375e-05,
-      "loss": 1.035,
+      "learning_rate": 1.9250621272905643e-05,
+      "loss": 1.162,
       "step": 5297
     },
     {
-      "epoch": 0.14550547911345474,
+      "epoch": 0.1503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.9308582727285895e-05,
-      "loss": 1.0548,
+      "learning_rate": 1.925027215344144e-05,
+      "loss": 0.8555,
       "step": 5298
     },
     {
-      "epoch": 0.14553294334129796,
+      "epoch": 0.15036889897843358,
       "grad_norm": 0.0,
-      "learning_rate": 1.93082576786953e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.9249922955839774e-05,
+      "loss": 1.0005,
       "step": 5299
     },
     {
-      "epoch": 0.14556040756914118,
+      "epoch": 0.1503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9307932556454167e-05,
-      "loss": 1.1354,
+      "learning_rate": 1.9249573680103596e-05,
+      "loss": 0.9896,
       "step": 5300
     },
     {
-      "epoch": 0.14558787179698443,
+      "epoch": 0.1504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.9307607360565067e-05,
-      "loss": 1.082,
+      "learning_rate": 1.9249224326235852e-05,
+      "loss": 1.113,
       "step": 5301
     },
     {
-      "epoch": 0.14561533602482765,
+      "epoch": 0.15045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.930728209103057e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.924887489423949e-05,
+      "loss": 1.0276,
       "step": 5302
     },
     {
-      "epoch": 0.1456428002526709,
+      "epoch": 0.15048240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9306956747853253e-05,
-      "loss": 1.0544,
+      "learning_rate": 1.924852538411747e-05,
+      "loss": 1.1002,
       "step": 5303
     },
     {
-      "epoch": 0.14567026448051412,
+      "epoch": 0.15051078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 1.930663133103569e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.9248175795872738e-05,
+      "loss": 1.0797,
       "step": 5304
     },
     {
-      "epoch": 0.14569772870835737,
+      "epoch": 0.15053916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.930630584058045e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.9247826129508254e-05,
+      "loss": 0.9502,
       "step": 5305
     },
     {
-      "epoch": 0.1457251929362006,
+      "epoch": 0.15056753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.9305980276490117e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.924747638502696e-05,
+      "loss": 0.94,
       "step": 5306
     },
     {
-      "epoch": 0.14575265716404384,
+      "epoch": 0.15059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9305654638767262e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.9247126562431824e-05,
+      "loss": 1.0094,
       "step": 5307
     },
     {
-      "epoch": 0.14578012139188706,
+      "epoch": 0.15062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9305328927414466e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.924677666172579e-05,
+      "loss": 1.0161,
       "step": 5308
     },
     {
-      "epoch": 0.1458075856197303,
+      "epoch": 0.1506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.93050031424343e-05,
-      "loss": 0.8736,
+      "learning_rate": 1.924642668291182e-05,
+      "loss": 0.9973,
       "step": 5309
     },
     {
-      "epoch": 0.14583504984757353,
+      "epoch": 0.1506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.930467728382934e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.9246076625992865e-05,
+      "loss": 1.0193,
       "step": 5310
     },
     {
-      "epoch": 0.14586251407541678,
+      "epoch": 0.15070942111237232,
       "grad_norm": 0.0,
-      "learning_rate": 1.930435135160218e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.924572649097189e-05,
+      "loss": 1.0251,
       "step": 5311
     },
     {
-      "epoch": 0.14588997830326,
+      "epoch": 0.1507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.930402534575538e-05,
-      "loss": 1.0252,
+      "learning_rate": 1.9245376277851846e-05,
+      "loss": 1.0231,
       "step": 5312
     },
     {
-      "epoch": 0.14591744253110323,
+      "epoch": 0.1507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.930369926629153e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.924502598663569e-05,
+      "loss": 1.0141,
       "step": 5313
     },
     {
-      "epoch": 0.14594490675894647,
+      "epoch": 0.15079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9303373113213206e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.9244675617326388e-05,
+      "loss": 1.1013,
       "step": 5314
     },
     {
-      "epoch": 0.1459723709867897,
+      "epoch": 0.15082292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.9303046886522993e-05,
-      "loss": 1.02,
+      "learning_rate": 1.924432516992689e-05,
+      "loss": 0.8914,
       "step": 5315
     },
     {
-      "epoch": 0.14599983521463294,
+      "epoch": 0.15085130533484675,
       "grad_norm": 0.0,
-      "learning_rate": 1.9302720586223468e-05,
-      "loss": 1.1231,
+      "learning_rate": 1.924397464444017e-05,
+      "loss": 0.9453,
       "step": 5316
     },
     {
-      "epoch": 0.14602729944247617,
+      "epoch": 0.15087968217934167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9302394212317214e-05,
-      "loss": 0.9,
+      "learning_rate": 1.9243624040869173e-05,
+      "loss": 1.0838,
       "step": 5317
     },
     {
-      "epoch": 0.14605476367031942,
+      "epoch": 0.15090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9302067764806813e-05,
-      "loss": 1.1188,
+      "learning_rate": 1.9243273359216873e-05,
+      "loss": 1.0783,
       "step": 5318
     },
     {
-      "epoch": 0.14608222789816264,
+      "epoch": 0.15093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.930174124369485e-05,
-      "loss": 1.0174,
+      "learning_rate": 1.9242922599486225e-05,
+      "loss": 1.0113,
       "step": 5319
     },
     {
-      "epoch": 0.14610969212600589,
+      "epoch": 0.15096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.9301414648983904e-05,
-      "loss": 1.1331,
+      "learning_rate": 1.9242571761680193e-05,
+      "loss": 0.9737,
       "step": 5320
     },
     {
-      "epoch": 0.1461371563538491,
+      "epoch": 0.15099318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 1.9301087980676567e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.9242220845801746e-05,
+      "loss": 1.0413,
       "step": 5321
     },
     {
-      "epoch": 0.14616462058169236,
+      "epoch": 0.15102156640181613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9300761238775413e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.924186985185384e-05,
+      "loss": 1.0203,
       "step": 5322
     },
     {
-      "epoch": 0.14619208480953558,
+      "epoch": 0.15104994324631102,
       "grad_norm": 0.0,
-      "learning_rate": 1.930043442328304e-05,
-      "loss": 1.1598,
+      "learning_rate": 1.9241518779839444e-05,
+      "loss": 1.0415,
       "step": 5323
     },
     {
-      "epoch": 0.14621954903737883,
+      "epoch": 0.1510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.930010753420202e-05,
-      "loss": 1.036,
+      "learning_rate": 1.9241167629761527e-05,
+      "loss": 1.0372,
       "step": 5324
     },
     {
-      "epoch": 0.14624701326522205,
+      "epoch": 0.1511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.9299780571534952e-05,
-      "loss": 1.014,
+      "learning_rate": 1.924081640162305e-05,
+      "loss": 0.9978,
       "step": 5325
     },
     {
-      "epoch": 0.14627447749306527,
+      "epoch": 0.15113507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 1.9299453535284418e-05,
-      "loss": 0.994,
+      "learning_rate": 1.924046509542698e-05,
+      "loss": 1.069,
       "step": 5326
     },
     {
-      "epoch": 0.14630194172090852,
+      "epoch": 0.15116345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 1.9299126425453005e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.924011371117629e-05,
+      "loss": 0.9997,
       "step": 5327
     },
     {
-      "epoch": 0.14632940594875174,
+      "epoch": 0.15119182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 1.9298799242043296e-05,
-      "loss": 1.176,
+      "learning_rate": 1.9239762248873938e-05,
+      "loss": 0.9886,
       "step": 5328
     },
     {
-      "epoch": 0.146356870176595,
+      "epoch": 0.15122020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.9298471985057893e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.9239410708522903e-05,
+      "loss": 1.004,
       "step": 5329
     },
     {
-      "epoch": 0.1463843344044382,
+      "epoch": 0.15124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.9298144654499375e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.923905909012615e-05,
+      "loss": 1.1161,
       "step": 5330
     },
     {
-      "epoch": 0.14641179863228146,
+      "epoch": 0.15127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.9297817250370335e-05,
-      "loss": 1.1383,
+      "learning_rate": 1.9238707393686648e-05,
+      "loss": 1.0331,
       "step": 5331
     },
     {
-      "epoch": 0.14643926286012468,
+      "epoch": 0.15130533484676503,
       "grad_norm": 0.0,
-      "learning_rate": 1.929748977267336e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.9238355619207372e-05,
+      "loss": 0.9795,
       "step": 5332
     },
     {
-      "epoch": 0.14646672708796793,
+      "epoch": 0.15133371169125992,
       "grad_norm": 0.0,
-      "learning_rate": 1.9297162221411046e-05,
-      "loss": 0.9571,
+      "learning_rate": 1.923800376669129e-05,
+      "loss": 1.0485,
       "step": 5333
     },
     {
-      "epoch": 0.14649419131581115,
+      "epoch": 0.15136208853575484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9296834596585987e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.9237651836141374e-05,
+      "loss": 1.1508,
       "step": 5334
     },
     {
-      "epoch": 0.1465216555436544,
+      "epoch": 0.15139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.9296506898200768e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.9237299827560595e-05,
+      "loss": 0.9393,
       "step": 5335
     },
     {
-      "epoch": 0.14654911977149762,
+      "epoch": 0.1514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9296179126257985e-05,
-      "loss": 1.1176,
+      "learning_rate": 1.923694774095193e-05,
+      "loss": 1.0027,
       "step": 5336
     },
     {
-      "epoch": 0.14657658399934087,
+      "epoch": 0.1514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9295851280760233e-05,
-      "loss": 1.1108,
+      "learning_rate": 1.923659557631836e-05,
+      "loss": 1.0564,
       "step": 5337
     },
     {
-      "epoch": 0.1466040482271841,
+      "epoch": 0.15147559591373438,
       "grad_norm": 0.0,
-      "learning_rate": 1.9295523361710104e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9236243333662844e-05,
+      "loss": 0.9406,
       "step": 5338
     },
     {
-      "epoch": 0.14663151245502734,
+      "epoch": 0.15150397275822927,
       "grad_norm": 0.0,
-      "learning_rate": 1.9295195369110193e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.9235891012988365e-05,
+      "loss": 1.0219,
       "step": 5339
     },
     {
-      "epoch": 0.14665897668287056,
+      "epoch": 0.1515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.9294867302963096e-05,
-      "loss": 1.0595,
+      "learning_rate": 1.9235538614297905e-05,
+      "loss": 1.0407,
       "step": 5340
     },
     {
-      "epoch": 0.14668644091071378,
+      "epoch": 0.15156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.929453916327141e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.923518613759443e-05,
+      "loss": 1.0289,
       "step": 5341
     },
     {
-      "epoch": 0.14671390513855703,
+      "epoch": 0.15158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9294210950037727e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.9234833582880923e-05,
+      "loss": 1.0149,
       "step": 5342
     },
     {
-      "epoch": 0.14674136936640025,
+      "epoch": 0.15161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.929388266326465e-05,
-      "loss": 1.1328,
+      "learning_rate": 1.9234480950160366e-05,
+      "loss": 1.1766,
       "step": 5343
     },
     {
-      "epoch": 0.1467688335942435,
+      "epoch": 0.15164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.9293554302954773e-05,
-      "loss": 1.0692,
+      "learning_rate": 1.923412823943573e-05,
+      "loss": 0.8758,
       "step": 5344
     },
     {
-      "epoch": 0.14679629782208672,
+      "epoch": 0.15167423382519865,
       "grad_norm": 0.0,
-      "learning_rate": 1.9293225869110694e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.9233775450709996e-05,
+      "loss": 1.1714,
       "step": 5345
     },
     {
-      "epoch": 0.14682376204992997,
+      "epoch": 0.15170261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9292897361735007e-05,
+      "learning_rate": 1.923342258398615e-05,
       "loss": 1.0573,
       "step": 5346
     },
     {
-      "epoch": 0.1468512262777732,
+      "epoch": 0.15173098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.9292568780830324e-05,
-      "loss": 1.1345,
+      "learning_rate": 1.9233069639267165e-05,
+      "loss": 1.0227,
       "step": 5347
     },
     {
-      "epoch": 0.14687869050561644,
+      "epoch": 0.1517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.929224012639923e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.9232716616556028e-05,
+      "loss": 0.9568,
       "step": 5348
     },
     {
-      "epoch": 0.14690615473345967,
+      "epoch": 0.1517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.929191139844434e-05,
-      "loss": 1.0809,
+      "learning_rate": 1.9232363515855717e-05,
+      "loss": 1.0904,
       "step": 5349
     },
     {
-      "epoch": 0.14693361896130291,
+      "epoch": 0.1518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9291582596968243e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.9232010337169216e-05,
+      "loss": 1.0116,
       "step": 5350
     },
     {
-      "epoch": 0.14696108318914614,
+      "epoch": 0.151844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9291253721973545e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.9231657080499507e-05,
+      "loss": 1.0298,
       "step": 5351
     },
     {
-      "epoch": 0.14698854741698938,
+      "epoch": 0.1518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9290924773462848e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.9231303745849578e-05,
+      "loss": 1.1021,
       "step": 5352
     },
     {
-      "epoch": 0.1470160116448326,
+      "epoch": 0.15190124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.9290595751438758e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.9230950333222415e-05,
+      "loss": 0.88,
       "step": 5353
     },
     {
-      "epoch": 0.14704347587267583,
+      "epoch": 0.15192962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 1.9290266655903874e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.9230596842620994e-05,
+      "loss": 0.9983,
       "step": 5354
     },
     {
-      "epoch": 0.14707094010051908,
+      "epoch": 0.15195800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.92899374868608e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.9230243274048306e-05,
+      "loss": 0.9238,
       "step": 5355
     },
     {
-      "epoch": 0.1470984043283623,
+      "epoch": 0.15198637911464244,
       "grad_norm": 0.0,
-      "learning_rate": 1.9289608244312147e-05,
-      "loss": 1.1071,
+      "learning_rate": 1.9229889627507344e-05,
+      "loss": 0.9879,
       "step": 5356
     },
     {
-      "epoch": 0.14712586855620555,
+      "epoch": 0.15201475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.928927892826051e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.9229535903001085e-05,
+      "loss": 1.0186,
       "step": 5357
     },
     {
-      "epoch": 0.14715333278404877,
+      "epoch": 0.15204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.9288949538708504e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.9229182100532524e-05,
+      "loss": 1.0183,
       "step": 5358
     },
     {
-      "epoch": 0.14718079701189202,
+      "epoch": 0.15207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.928862007565873e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.9228828220104647e-05,
+      "loss": 0.861,
       "step": 5359
     },
     {
-      "epoch": 0.14720826123973524,
+      "epoch": 0.15209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9288290539113793e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.922847426172044e-05,
+      "loss": 1.1565,
       "step": 5360
     },
     {
-      "epoch": 0.1472357254675785,
+      "epoch": 0.1521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.9287960929076304e-05,
-      "loss": 1.1629,
+      "learning_rate": 1.9228120225382895e-05,
+      "loss": 1.0591,
       "step": 5361
     },
     {
-      "epoch": 0.1472631896954217,
+      "epoch": 0.15215664018161182,
       "grad_norm": 0.0,
-      "learning_rate": 1.928763124554887e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.9227766111095008e-05,
+      "loss": 1.0219,
       "step": 5362
     },
     {
-      "epoch": 0.14729065392326496,
+      "epoch": 0.1521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.9287301488534106e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.9227411918859764e-05,
+      "loss": 1.0776,
       "step": 5363
     },
     {
-      "epoch": 0.14731811815110818,
+      "epoch": 0.1522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.928697165803461e-05,
-      "loss": 1.1619,
+      "learning_rate": 1.9227057648680155e-05,
+      "loss": 1.0651,
       "step": 5364
     },
     {
-      "epoch": 0.14734558237895143,
+      "epoch": 0.15224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9286641754052997e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.922670330055918e-05,
+      "loss": 0.9659,
       "step": 5365
     },
     {
-      "epoch": 0.14737304660679465,
+      "epoch": 0.15227014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.9286311776591877e-05,
-      "loss": 0.927,
+      "learning_rate": 1.9226348874499824e-05,
+      "loss": 1.0714,
       "step": 5366
     },
     {
-      "epoch": 0.14740051083463787,
+      "epoch": 0.15229852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.928598172565386e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.9225994370505085e-05,
+      "loss": 1.1549,
       "step": 5367
     },
     {
-      "epoch": 0.14742797506248112,
+      "epoch": 0.15232690124858117,
       "grad_norm": 0.0,
-      "learning_rate": 1.928565160124156e-05,
-      "loss": 1.0921,
+      "learning_rate": 1.9225639788577958e-05,
+      "loss": 0.8909,
       "step": 5368
     },
     {
-      "epoch": 0.14745543929032434,
+      "epoch": 0.15235527809307606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9285321403357588e-05,
-      "loss": 1.0869,
+      "learning_rate": 1.9225285128721435e-05,
+      "loss": 1.0487,
       "step": 5369
     },
     {
-      "epoch": 0.1474829035181676,
+      "epoch": 0.15238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9284991132004557e-05,
-      "loss": 1.1411,
+      "learning_rate": 1.922493039093851e-05,
+      "loss": 0.9223,
       "step": 5370
     },
     {
-      "epoch": 0.1475103677460108,
+      "epoch": 0.15241203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.9284660787185078e-05,
-      "loss": 1.096,
+      "learning_rate": 1.922457557523219e-05,
+      "loss": 0.9903,
       "step": 5371
     },
     {
-      "epoch": 0.14753783197385406,
+      "epoch": 0.15244040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 1.9284330368901764e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.9224220681605464e-05,
+      "loss": 0.9133,
       "step": 5372
     },
     {
-      "epoch": 0.14756529620169728,
+      "epoch": 0.1524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9283999877157238e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.9223865710061328e-05,
+      "loss": 0.8601,
       "step": 5373
     },
     {
-      "epoch": 0.14759276042954053,
+      "epoch": 0.15249716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 1.9283669311954105e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.9223510660602785e-05,
+      "loss": 0.9028,
       "step": 5374
     },
     {
-      "epoch": 0.14762022465738375,
+      "epoch": 0.1525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9283338673294983e-05,
-      "loss": 1.0413,
+      "learning_rate": 1.9223155533232837e-05,
+      "loss": 0.9727,
       "step": 5375
     },
     {
-      "epoch": 0.147647688885227,
+      "epoch": 0.1525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.928300796118249e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.9222800327954476e-05,
+      "loss": 1.0124,
       "step": 5376
     },
     {
-      "epoch": 0.14767515311307022,
+      "epoch": 0.15258229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 1.9282677175619242e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.9222445044770706e-05,
+      "loss": 0.8979,
       "step": 5377
     },
     {
-      "epoch": 0.14770261734091347,
+      "epoch": 0.15261066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.9282346316607857e-05,
-      "loss": 0.9295,
+      "learning_rate": 1.9222089683684527e-05,
+      "loss": 1.0922,
       "step": 5378
     },
     {
-      "epoch": 0.1477300815687567,
+      "epoch": 0.15263904653802496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9282015384150953e-05,
-      "loss": 1.0552,
+      "learning_rate": 1.9221734244698944e-05,
+      "loss": 1.073,
       "step": 5379
     },
     {
-      "epoch": 0.14775754579659992,
+      "epoch": 0.15266742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 1.928168437825115e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.9221378727816958e-05,
+      "loss": 0.9775,
       "step": 5380
     },
     {
-      "epoch": 0.14778501002444316,
+      "epoch": 0.15269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.9281353298911062e-05,
-      "loss": 1.083,
+      "learning_rate": 1.922102313304157e-05,
+      "loss": 0.9754,
       "step": 5381
     },
     {
-      "epoch": 0.14781247425228639,
+      "epoch": 0.15272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.928102214613331e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.9220667460375787e-05,
+      "loss": 1.148,
       "step": 5382
     },
     {
-      "epoch": 0.14783993848012963,
+      "epoch": 0.15275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.9280690919920518e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.922031170982261e-05,
+      "loss": 1.0168,
       "step": 5383
     },
     {
-      "epoch": 0.14786740270797286,
+      "epoch": 0.15278093076049942,
       "grad_norm": 0.0,
-      "learning_rate": 1.9280359620275302e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.9219955881385042e-05,
+      "loss": 1.209,
       "step": 5384
     },
     {
-      "epoch": 0.1478948669358161,
+      "epoch": 0.15280930760499434,
       "grad_norm": 0.0,
-      "learning_rate": 1.9280028247200293e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.9219599975066096e-05,
+      "loss": 0.9443,
       "step": 5385
     },
     {
-      "epoch": 0.14792233116365933,
+      "epoch": 0.15283768444948923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9279696800698097e-05,
-      "loss": 1.0218,
+      "learning_rate": 1.9219243990868776e-05,
+      "loss": 0.9756,
       "step": 5386
     },
     {
-      "epoch": 0.14794979539150258,
+      "epoch": 0.1528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.927936528077135e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.9218887928796083e-05,
+      "loss": 1.0079,
       "step": 5387
     },
     {
-      "epoch": 0.1479772596193458,
+      "epoch": 0.152894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9279033687422666e-05,
-      "loss": 1.054,
+      "learning_rate": 1.9218531788851034e-05,
+      "loss": 1.0302,
       "step": 5388
     },
     {
-      "epoch": 0.14800472384718905,
+      "epoch": 0.1529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.9278702020654676e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.921817557103663e-05,
+      "loss": 1.0624,
       "step": 5389
     },
     {
-      "epoch": 0.14803218807503227,
+      "epoch": 0.15295119182746877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9278370280470003e-05,
-      "loss": 1.1361,
+      "learning_rate": 1.9217819275355882e-05,
+      "loss": 1.0759,
       "step": 5390
     },
     {
-      "epoch": 0.14805965230287552,
+      "epoch": 0.1529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.9278038466871267e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.92174629018118e-05,
+      "loss": 1.0445,
       "step": 5391
     },
     {
-      "epoch": 0.14808711653071874,
+      "epoch": 0.15300794551645858,
       "grad_norm": 0.0,
-      "learning_rate": 1.92777065798611e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.9217106450407397e-05,
+      "loss": 0.9506,
       "step": 5392
     },
     {
-      "epoch": 0.148114580758562,
+      "epoch": 0.15303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9277374619442125e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.921674992114568e-05,
+      "loss": 0.9984,
       "step": 5393
     },
     {
-      "epoch": 0.1481420449864052,
+      "epoch": 0.15306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9277042585616964e-05,
-      "loss": 1.047,
+      "learning_rate": 1.921639331402966e-05,
+      "loss": 1.1188,
       "step": 5394
     },
     {
-      "epoch": 0.14816950921424843,
+      "epoch": 0.15309307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.9276710478388253e-05,
-      "loss": 1.0819,
+      "learning_rate": 1.9216036629062356e-05,
+      "loss": 1.0276,
       "step": 5395
     },
     {
-      "epoch": 0.14819697344209168,
+      "epoch": 0.15312145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 1.9276378297758614e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.9215679866246774e-05,
+      "loss": 1.0539,
       "step": 5396
     },
     {
-      "epoch": 0.1482244376699349,
+      "epoch": 0.15314982973893304,
       "grad_norm": 0.0,
-      "learning_rate": 1.9276046043730674e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.921532302558593e-05,
+      "loss": 1.1873,
       "step": 5397
     },
     {
-      "epoch": 0.14825190189777815,
+      "epoch": 0.15317820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9275713716307067e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.9214966107082835e-05,
+      "loss": 0.9886,
       "step": 5398
     },
     {
-      "epoch": 0.14827936612562137,
+      "epoch": 0.15320658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.9275381315490418e-05,
-      "loss": 1.108,
+      "learning_rate": 1.9214609110740514e-05,
+      "loss": 1.1812,
       "step": 5399
     },
     {
-      "epoch": 0.14830683035346462,
+      "epoch": 0.1532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9275048841283358e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.9214252036561973e-05,
+      "loss": 1.0403,
       "step": 5400
     },
     {
-      "epoch": 0.14833429458130784,
+      "epoch": 0.1532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9274716293688526e-05,
-      "loss": 0.9062,
+      "learning_rate": 1.9213894884550225e-05,
+      "loss": 0.9954,
       "step": 5401
     },
     {
-      "epoch": 0.1483617588091511,
+      "epoch": 0.1532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.927438367270854e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.92135376547083e-05,
+      "loss": 0.956,
       "step": 5402
     },
     {
-      "epoch": 0.1483892230369943,
+      "epoch": 0.1533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9274050978346038e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.9213180347039203e-05,
+      "loss": 1.1311,
       "step": 5403
     },
     {
-      "epoch": 0.14841668726483756,
+      "epoch": 0.15334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.927371821060365e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.921282296154596e-05,
+      "loss": 1.0209,
       "step": 5404
     },
     {
-      "epoch": 0.14844415149268078,
+      "epoch": 0.15337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.9273385369484018e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.9212465498231587e-05,
+      "loss": 1.102,
       "step": 5405
     },
     {
-      "epoch": 0.14847161572052403,
+      "epoch": 0.15340522133938705,
       "grad_norm": 0.0,
-      "learning_rate": 1.9273052454989764e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.9212107957099103e-05,
+      "loss": 1.0712,
       "step": 5406
     },
     {
-      "epoch": 0.14849907994836725,
+      "epoch": 0.15343359818388194,
       "grad_norm": 0.0,
-      "learning_rate": 1.927271946712353e-05,
-      "loss": 1.1123,
+      "learning_rate": 1.9211750338151532e-05,
+      "loss": 0.9357,
       "step": 5407
     },
     {
-      "epoch": 0.14852654417621047,
+      "epoch": 0.15346197502837686,
       "grad_norm": 0.0,
-      "learning_rate": 1.9272386405887946e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.921139264139189e-05,
+      "loss": 1.0224,
       "step": 5408
     },
     {
-      "epoch": 0.14855400840405372,
+      "epoch": 0.15349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.927205327128565e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.92110348668232e-05,
+      "loss": 0.9309,
       "step": 5409
     },
     {
-      "epoch": 0.14858147263189694,
+      "epoch": 0.15351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9271720063319272e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.9210677014448484e-05,
+      "loss": 0.8807,
       "step": 5410
     },
     {
-      "epoch": 0.1486089368597402,
+      "epoch": 0.15354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9271386781991457e-05,
-      "loss": 0.953,
+      "learning_rate": 1.9210319084270766e-05,
+      "loss": 0.9591,
       "step": 5411
     },
     {
-      "epoch": 0.14863640108758341,
+      "epoch": 0.1535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.927105342730484e-05,
-      "loss": 1.1719,
+      "learning_rate": 1.9209961076293068e-05,
+      "loss": 0.9703,
       "step": 5412
     },
     {
-      "epoch": 0.14866386531542666,
+      "epoch": 0.1536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.9270719999262054e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.9209602990518413e-05,
+      "loss": 0.9623,
       "step": 5413
     },
     {
-      "epoch": 0.14869132954326988,
+      "epoch": 0.1536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9270386497865742e-05,
-      "loss": 1.1079,
+      "learning_rate": 1.9209244826949833e-05,
+      "loss": 0.9738,
       "step": 5414
     },
     {
-      "epoch": 0.14871879377111313,
+      "epoch": 0.1536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.927005292311854e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.9208886585590343e-05,
+      "loss": 1.1146,
       "step": 5415
     },
     {
-      "epoch": 0.14874625799895635,
+      "epoch": 0.15368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.926971927502309e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.9208528266442976e-05,
+      "loss": 1.142,
       "step": 5416
     },
     {
-      "epoch": 0.1487737222267996,
+      "epoch": 0.15371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.926938555358203e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.9208169869510754e-05,
+      "loss": 0.9124,
       "step": 5417
     },
     {
-      "epoch": 0.14880118645464283,
+      "epoch": 0.15374574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.9269051758797997e-05,
-      "loss": 0.9495,
+      "learning_rate": 1.920781139479671e-05,
+      "loss": 0.925,
       "step": 5418
     },
     {
-      "epoch": 0.14882865068248607,
+      "epoch": 0.15377412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.9268717890673644e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.920745284230387e-05,
+      "loss": 0.9803,
       "step": 5419
     },
     {
-      "epoch": 0.1488561149103293,
+      "epoch": 0.15380249716231556,
       "grad_norm": 0.0,
-      "learning_rate": 1.92683839492116e-05,
-      "loss": 1.0721,
+      "learning_rate": 1.920709421203526e-05,
+      "loss": 1.0526,
       "step": 5420
     },
     {
-      "epoch": 0.14888357913817252,
+      "epoch": 0.15383087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.926804993441451e-05,
-      "loss": 1.0344,
+      "learning_rate": 1.9206735503993914e-05,
+      "loss": 1.0437,
       "step": 5421
     },
     {
-      "epoch": 0.14891104336601577,
+      "epoch": 0.15385925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.9267715846285023e-05,
-      "loss": 1.1204,
+      "learning_rate": 1.9206376718182855e-05,
+      "loss": 1.0349,
       "step": 5422
     },
     {
-      "epoch": 0.148938507593859,
+      "epoch": 0.15388762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 1.9267381684825775e-05,
-      "loss": 1.1257,
+      "learning_rate": 1.9206017854605122e-05,
+      "loss": 1.0032,
       "step": 5423
     },
     {
-      "epoch": 0.14896597182170224,
+      "epoch": 0.1539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.9267047450039412e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.920565891326374e-05,
+      "loss": 0.9675,
       "step": 5424
     },
     {
-      "epoch": 0.14899343604954546,
+      "epoch": 0.15394438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 1.9266713141928585e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.9205299894161743e-05,
+      "loss": 1.0228,
       "step": 5425
     },
     {
-      "epoch": 0.1490209002773887,
+      "epoch": 0.1539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.926637876049593e-05,
-      "loss": 1.0091,
+      "learning_rate": 1.9204940797302165e-05,
+      "loss": 1.0701,
       "step": 5426
     },
     {
-      "epoch": 0.14904836450523193,
+      "epoch": 0.1540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.92660443057441e-05,
-      "loss": 1.0934,
+      "learning_rate": 1.920458162268804e-05,
+      "loss": 1.043,
       "step": 5427
     },
     {
-      "epoch": 0.14907582873307518,
+      "epoch": 0.1540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9265709777675733e-05,
-      "loss": 0.943,
+      "learning_rate": 1.92042223703224e-05,
+      "loss": 1.0027,
       "step": 5428
     },
     {
-      "epoch": 0.1491032929609184,
+      "epoch": 0.15405788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 1.9265375176293488e-05,
-      "loss": 1.0668,
+      "learning_rate": 1.9203863040208274e-05,
+      "loss": 1.0861,
       "step": 5429
     },
     {
-      "epoch": 0.14913075718876165,
+      "epoch": 0.15408626560726446,
       "grad_norm": 0.0,
-      "learning_rate": 1.9265040501599998e-05,
-      "loss": 1.1304,
+      "learning_rate": 1.920350363234871e-05,
+      "loss": 1.0389,
       "step": 5430
     },
     {
-      "epoch": 0.14915822141660487,
+      "epoch": 0.15411464245175938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9264705753597924e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.9203144146746736e-05,
+      "loss": 1.0598,
       "step": 5431
     },
     {
-      "epoch": 0.14918568564444812,
+      "epoch": 0.15414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.9264370932289906e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.9202784583405386e-05,
+      "loss": 0.9086,
       "step": 5432
     },
     {
-      "epoch": 0.14921314987229134,
+      "epoch": 0.15417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.92640360376786e-05,
-      "loss": 1.064,
+      "learning_rate": 1.9202424942327705e-05,
+      "loss": 1.0026,
       "step": 5433
     },
     {
-      "epoch": 0.14924061410013456,
+      "epoch": 0.15419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.926370106976665e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.9202065223516722e-05,
+      "loss": 0.9148,
       "step": 5434
     },
     {
-      "epoch": 0.1492680783279778,
+      "epoch": 0.15422814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.926336602855671e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.9201705426975485e-05,
+      "loss": 0.9375,
       "step": 5435
     },
     {
-      "epoch": 0.14929554255582103,
+      "epoch": 0.1542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.9263030914051424e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.9201345552707028e-05,
+      "loss": 1.0808,
       "step": 5436
     },
     {
-      "epoch": 0.14932300678366428,
+      "epoch": 0.15428490351872873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9262695726253456e-05,
-      "loss": 1.0804,
+      "learning_rate": 1.920098560071439e-05,
+      "loss": 1.0565,
       "step": 5437
     },
     {
-      "epoch": 0.1493504710115075,
+      "epoch": 0.15431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.926236046516545e-05,
-      "loss": 1.0568,
+      "learning_rate": 1.9200625571000616e-05,
+      "loss": 1.0612,
       "step": 5438
     },
     {
-      "epoch": 0.14937793523935075,
+      "epoch": 0.1543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.926202513079006e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.9200265463568742e-05,
+      "loss": 0.9051,
       "step": 5439
     },
     {
-      "epoch": 0.14940539946719397,
+      "epoch": 0.1543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.926168972312994e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.919990527842181e-05,
+      "loss": 1.1047,
       "step": 5440
     },
     {
-      "epoch": 0.14943286369503722,
+      "epoch": 0.15439841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 1.9261354242187742e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.9199545015562866e-05,
+      "loss": 0.9906,
       "step": 5441
     },
     {
-      "epoch": 0.14946032792288044,
+      "epoch": 0.1544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.9261018687966123e-05,
-      "loss": 1.0952,
+      "learning_rate": 1.9199184674994952e-05,
+      "loss": 0.9565,
       "step": 5442
     },
     {
-      "epoch": 0.1494877921507237,
+      "epoch": 0.15445516458569808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9260683060467734e-05,
-      "loss": 1.1097,
+      "learning_rate": 1.9198824256721113e-05,
+      "loss": 1.0612,
       "step": 5443
     },
     {
-      "epoch": 0.1495152563785669,
+      "epoch": 0.15448354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.9260347359695236e-05,
-      "loss": 1.0544,
+      "learning_rate": 1.919846376074439e-05,
+      "loss": 1.0489,
       "step": 5444
     },
     {
-      "epoch": 0.14954272060641016,
+      "epoch": 0.15451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.926001158565128e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.919810318706783e-05,
+      "loss": 1.0202,
       "step": 5445
     },
     {
-      "epoch": 0.14957018483425338,
+      "epoch": 0.15454029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.9259675738338528e-05,
-      "loss": 1.129,
+      "learning_rate": 1.919774253569448e-05,
+      "loss": 0.9666,
       "step": 5446
     },
     {
-      "epoch": 0.14959764906209663,
+      "epoch": 0.15456867196367763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9259339817759635e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.9197381806627383e-05,
+      "loss": 1.0822,
       "step": 5447
     },
     {
-      "epoch": 0.14962511328993985,
+      "epoch": 0.15459704880817254,
       "grad_norm": 0.0,
-      "learning_rate": 1.9259003823917253e-05,
-      "loss": 1.0936,
+      "learning_rate": 1.9197020999869594e-05,
+      "loss": 1.0344,
       "step": 5448
     },
     {
-      "epoch": 0.14965257751778308,
+      "epoch": 0.15462542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 1.9258667756814054e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.919666011542415e-05,
+      "loss": 1.0183,
       "step": 5449
     },
     {
-      "epoch": 0.14968004174562632,
+      "epoch": 0.15465380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.9258331616452686e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.9196299153294107e-05,
+      "loss": 0.9817,
       "step": 5450
     },
     {
-      "epoch": 0.14970750597346955,
+      "epoch": 0.1546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9257995402835812e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.919593811348251e-05,
+      "loss": 1.0802,
       "step": 5451
     },
     {
-      "epoch": 0.1497349702013128,
+      "epoch": 0.1547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9257659115966093e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.919557699599241e-05,
+      "loss": 1.0194,
       "step": 5452
     },
     {
-      "epoch": 0.14976243442915602,
+      "epoch": 0.15473893303064698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9257322755846188e-05,
-      "loss": 0.9597,
+      "learning_rate": 1.919521580082686e-05,
+      "loss": 0.9781,
       "step": 5453
     },
     {
-      "epoch": 0.14978989865699927,
+      "epoch": 0.1547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.9256986322478755e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.919485452798891e-05,
+      "loss": 0.9692,
       "step": 5454
     },
     {
-      "epoch": 0.1498173628848425,
+      "epoch": 0.15479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.9256649815866467e-05,
-      "loss": 1.0839,
+      "learning_rate": 1.9194493177481607e-05,
+      "loss": 0.9359,
       "step": 5455
     },
     {
-      "epoch": 0.14984482711268574,
+      "epoch": 0.15482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9256313236011978e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.9194131749308006e-05,
+      "loss": 1.0516,
       "step": 5456
     },
     {
-      "epoch": 0.14987229134052896,
+      "epoch": 0.15485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.925597658291795e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.9193770243471164e-05,
+      "loss": 1.0618,
       "step": 5457
     },
     {
-      "epoch": 0.1498997555683722,
+      "epoch": 0.15488081725312144,
       "grad_norm": 0.0,
-      "learning_rate": 1.9255639856587054e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.919340865997413e-05,
+      "loss": 0.8846,
       "step": 5458
     },
     {
-      "epoch": 0.14992721979621543,
+      "epoch": 0.15490919409761633,
       "grad_norm": 0.0,
-      "learning_rate": 1.925530305702195e-05,
-      "loss": 1.0668,
+      "learning_rate": 1.919304699881996e-05,
+      "loss": 0.9635,
       "step": 5459
     },
     {
-      "epoch": 0.14995468402405868,
+      "epoch": 0.15493757094211125,
       "grad_norm": 0.0,
-      "learning_rate": 1.92549661842253e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.919268526001171e-05,
+      "loss": 1.0362,
       "step": 5460
     },
     {
-      "epoch": 0.1499821482519019,
+      "epoch": 0.15496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.9254629238199776e-05,
-      "loss": 1.0171,
+      "learning_rate": 1.919232344355243e-05,
+      "loss": 1.0284,
       "step": 5461
     },
     {
-      "epoch": 0.15000961247974512,
+      "epoch": 0.15499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.925429221894804e-05,
-      "loss": 1.018,
+      "learning_rate": 1.9191961549445186e-05,
+      "loss": 1.0541,
       "step": 5462
     },
     {
-      "epoch": 0.15003707670758837,
+      "epoch": 0.1550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.925395512647276e-05,
-      "loss": 1.0119,
+      "learning_rate": 1.9191599577693026e-05,
+      "loss": 1.0839,
       "step": 5463
     },
     {
-      "epoch": 0.1500645409354316,
+      "epoch": 0.1550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.92536179607766e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.9191237528299014e-05,
+      "loss": 1.0335,
       "step": 5464
     },
     {
-      "epoch": 0.15009200516327484,
+      "epoch": 0.1550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.925328072186223e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.9190875401266203e-05,
+      "loss": 0.9605,
       "step": 5465
     },
     {
-      "epoch": 0.15011946939111806,
+      "epoch": 0.1551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9252943409732318e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.9190513196597656e-05,
+      "loss": 1.1068,
       "step": 5466
     },
     {
-      "epoch": 0.1501469336189613,
+      "epoch": 0.15513620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 1.9252606024389537e-05,
-      "loss": 1.0164,
+      "learning_rate": 1.919015091429643e-05,
+      "loss": 1.0993,
       "step": 5467
     },
     {
-      "epoch": 0.15017439784680453,
+      "epoch": 0.15516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.925226856583655e-05,
-      "loss": 1.1785,
+      "learning_rate": 1.9189788554365586e-05,
+      "loss": 1.0101,
       "step": 5468
     },
     {
-      "epoch": 0.15020186207464778,
+      "epoch": 0.15519296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.925193103407603e-05,
-      "loss": 1.1003,
+      "learning_rate": 1.9189426116808185e-05,
+      "loss": 1.0135,
       "step": 5469
     },
     {
-      "epoch": 0.150229326302491,
+      "epoch": 0.15522133938706015,
       "grad_norm": 0.0,
-      "learning_rate": 1.925159342911065e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.918906360162729e-05,
+      "loss": 1.0291,
       "step": 5470
     },
     {
-      "epoch": 0.15025679053033425,
+      "epoch": 0.15524971623155506,
       "grad_norm": 0.0,
-      "learning_rate": 1.9251255750943077e-05,
-      "loss": 1.1137,
+      "learning_rate": 1.9188701008825962e-05,
+      "loss": 0.9921,
       "step": 5471
     },
     {
-      "epoch": 0.15028425475817747,
+      "epoch": 0.15527809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.9250917999575987e-05,
-      "loss": 0.997,
+      "learning_rate": 1.918833833840726e-05,
+      "loss": 1.0966,
       "step": 5472
     },
     {
-      "epoch": 0.15031171898602072,
+      "epoch": 0.15530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.9250580175012047e-05,
-      "loss": 1.1223,
+      "learning_rate": 1.9187975590374254e-05,
+      "loss": 1.043,
       "step": 5473
     },
     {
-      "epoch": 0.15033918321386394,
+      "epoch": 0.15533484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 1.9250242277253934e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.9187612764730004e-05,
+      "loss": 0.9752,
       "step": 5474
     },
     {
-      "epoch": 0.15036664744170716,
+      "epoch": 0.1553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9249904306304323e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.9187249861477575e-05,
+      "loss": 1.038,
       "step": 5475
     },
     {
-      "epoch": 0.1503941116695504,
+      "epoch": 0.1553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.9249566262165886e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.9186886880620036e-05,
+      "loss": 1.1429,
       "step": 5476
     },
     {
-      "epoch": 0.15042157589739363,
+      "epoch": 0.15541997729852441,
       "grad_norm": 0.0,
-      "learning_rate": 1.9249228144841298e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.918652382216045e-05,
+      "loss": 0.9842,
       "step": 5477
     },
     {
-      "epoch": 0.15044904012523688,
+      "epoch": 0.1554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9248889954333232e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.918616068610188e-05,
+      "loss": 1.0947,
       "step": 5478
     },
     {
-      "epoch": 0.1504765043530801,
+      "epoch": 0.1554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.9248551690644366e-05,
-      "loss": 1.0548,
+      "learning_rate": 1.9185797472447402e-05,
+      "loss": 0.9774,
       "step": 5479
     },
     {
-      "epoch": 0.15050396858092335,
+      "epoch": 0.15550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.924821335377738e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.918543418120008e-05,
+      "loss": 0.9594,
       "step": 5480
     },
     {
-      "epoch": 0.15053143280876657,
+      "epoch": 0.15553348467650396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9247874943734946e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.9185070812362982e-05,
+      "loss": 1.138,
       "step": 5481
     },
     {
-      "epoch": 0.15055889703660982,
+      "epoch": 0.15556186152099888,
       "grad_norm": 0.0,
-      "learning_rate": 1.924753646051974e-05,
-      "loss": 1.0889,
+      "learning_rate": 1.9184707365939177e-05,
+      "loss": 0.9925,
       "step": 5482
     },
     {
-      "epoch": 0.15058636126445304,
+      "epoch": 0.15559023836549377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9247197904134446e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9184343841931735e-05,
+      "loss": 0.9534,
       "step": 5483
     },
     {
-      "epoch": 0.1506138254922963,
+      "epoch": 0.15561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.924685927458174e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.9183980240343732e-05,
+      "loss": 1.1293,
       "step": 5484
     },
     {
-      "epoch": 0.15064128972013952,
+      "epoch": 0.15564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9246520571864304e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.9183616561178227e-05,
+      "loss": 0.9725,
       "step": 5485
     },
     {
-      "epoch": 0.15066875394798276,
+      "epoch": 0.15567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.924618179598481e-05,
-      "loss": 0.986,
+      "learning_rate": 1.918325280443831e-05,
+      "loss": 1.1056,
       "step": 5486
     },
     {
-      "epoch": 0.15069621817582599,
+      "epoch": 0.15570374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 1.9245842946945948e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.9182888970127037e-05,
+      "loss": 0.9981,
       "step": 5487
     },
     {
-      "epoch": 0.15072368240366923,
+      "epoch": 0.15573212258796823,
       "grad_norm": 0.0,
-      "learning_rate": 1.9245504024750392e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.918252505824749e-05,
+      "loss": 1.023,
       "step": 5488
     },
     {
-      "epoch": 0.15075114663151246,
+      "epoch": 0.15576049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9245165029400833e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.9182161068802742e-05,
+      "loss": 0.975,
       "step": 5489
     },
     {
-      "epoch": 0.15077861085935568,
+      "epoch": 0.155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.924482596089994e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.9181797001795864e-05,
+      "loss": 1.0256,
       "step": 5490
     },
     {
-      "epoch": 0.15080607508719893,
+      "epoch": 0.1558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.9244486819250406e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.9181432857229936e-05,
+      "loss": 1.0709,
       "step": 5491
     },
     {
-      "epoch": 0.15083353931504215,
+      "epoch": 0.15584562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 1.924414760445491e-05,
-      "loss": 1.0933,
+      "learning_rate": 1.9181068635108032e-05,
+      "loss": 1.0067,
       "step": 5492
     },
     {
-      "epoch": 0.1508610035428854,
+      "epoch": 0.15587400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.924380831651614e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.9180704335433228e-05,
+      "loss": 1.0697,
       "step": 5493
     },
     {
-      "epoch": 0.15088846777072862,
+      "epoch": 0.15590238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 1.9243468955436775e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.9180339958208603e-05,
+      "loss": 1.0497,
       "step": 5494
     },
     {
-      "epoch": 0.15091593199857187,
+      "epoch": 0.15593076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9243129521219506e-05,
-      "loss": 1.143,
+      "learning_rate": 1.9179975503437235e-05,
+      "loss": 0.9691,
       "step": 5495
     },
     {
-      "epoch": 0.1509433962264151,
+      "epoch": 0.15595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.9242790013867014e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.91796109711222e-05,
+      "loss": 1.0974,
       "step": 5496
     },
     {
-      "epoch": 0.15097086045425834,
+      "epoch": 0.15598751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.9242450433381986e-05,
-      "loss": 1.1166,
+      "learning_rate": 1.9179246361266576e-05,
+      "loss": 0.9753,
       "step": 5497
     },
     {
-      "epoch": 0.15099832468210156,
+      "epoch": 0.15601589103291713,
       "grad_norm": 0.0,
-      "learning_rate": 1.924211077976711e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.9178881673873448e-05,
+      "loss": 1.0832,
       "step": 5498
     },
     {
-      "epoch": 0.1510257889099448,
+      "epoch": 0.15604426787741202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9241771053025076e-05,
-      "loss": 1.0108,
+      "learning_rate": 1.917851690894589e-05,
+      "loss": 1.0217,
       "step": 5499
     },
     {
-      "epoch": 0.15105325313778803,
+      "epoch": 0.15607264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 1.9241431253158568e-05,
-      "loss": 1.1354,
+      "learning_rate": 1.9178152066486988e-05,
+      "loss": 0.9703,
       "step": 5500
     },
     {
-      "epoch": 0.15108071736563128,
+      "epoch": 0.15610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.9241091380170277e-05,
-      "loss": 1.1072,
+      "learning_rate": 1.9177787146499826e-05,
+      "loss": 1.0653,
       "step": 5501
     },
     {
-      "epoch": 0.1511081815934745,
+      "epoch": 0.1561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.924075143406289e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.9177422148987482e-05,
+      "loss": 1.0728,
       "step": 5502
     },
     {
-      "epoch": 0.15113564582131772,
+      "epoch": 0.1561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.92404114148391e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.917705707395304e-05,
+      "loss": 1.0308,
       "step": 5503
     },
     {
-      "epoch": 0.15116311004916097,
+      "epoch": 0.15618615209988648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9240071322501594e-05,
-      "loss": 1.1456,
+      "learning_rate": 1.9176691921399585e-05,
+      "loss": 1.0987,
       "step": 5504
     },
     {
-      "epoch": 0.1511905742770042,
+      "epoch": 0.1562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.9239731157053064e-05,
-      "loss": 1.1091,
+      "learning_rate": 1.91763266913302e-05,
+      "loss": 0.9064,
       "step": 5505
     },
     {
-      "epoch": 0.15121803850484744,
+      "epoch": 0.15624290578887629,
       "grad_norm": 0.0,
-      "learning_rate": 1.9239390918496202e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.9175961383747973e-05,
+      "loss": 1.1,
       "step": 5506
     },
     {
-      "epoch": 0.15124550273269066,
+      "epoch": 0.15627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.92390506068337e-05,
-      "loss": 1.001,
+      "learning_rate": 1.9175595998655988e-05,
+      "loss": 0.9667,
       "step": 5507
     },
     {
-      "epoch": 0.1512729669605339,
+      "epoch": 0.15629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9238710222068254e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.917523053605733e-05,
+      "loss": 1.1573,
       "step": 5508
     },
     {
-      "epoch": 0.15130043118837713,
+      "epoch": 0.15632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.923836976420255e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.9174864995955085e-05,
+      "loss": 1.2079,
       "step": 5509
     },
     {
-      "epoch": 0.15132789541622038,
+      "epoch": 0.15635641316685583,
       "grad_norm": 0.0,
-      "learning_rate": 1.9238029233239288e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.9174499378352346e-05,
+      "loss": 1.061,
       "step": 5510
     },
     {
-      "epoch": 0.1513553596440636,
+      "epoch": 0.15638479001135075,
       "grad_norm": 0.0,
-      "learning_rate": 1.923768862918116e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.9174133683252195e-05,
+      "loss": 0.9802,
       "step": 5511
     },
     {
-      "epoch": 0.15138282387190685,
+      "epoch": 0.15641316685584564,
       "grad_norm": 0.0,
-      "learning_rate": 1.923734795203086e-05,
-      "loss": 1.022,
+      "learning_rate": 1.9173767910657724e-05,
+      "loss": 0.9934,
       "step": 5512
     },
     {
-      "epoch": 0.15141028809975007,
+      "epoch": 0.15644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.9237007201791086e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.9173402060572028e-05,
+      "loss": 0.9729,
       "step": 5513
     },
     {
-      "epoch": 0.15143775232759332,
+      "epoch": 0.1564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.923666637846453e-05,
-      "loss": 0.9954,
+      "learning_rate": 1.9173036132998192e-05,
+      "loss": 1.0411,
       "step": 5514
     },
     {
-      "epoch": 0.15146521655543654,
+      "epoch": 0.1564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.9236325482053897e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.9172670127939304e-05,
+      "loss": 0.9336,
       "step": 5515
     },
     {
-      "epoch": 0.15149268078327977,
+      "epoch": 0.15652667423382519,
       "grad_norm": 0.0,
-      "learning_rate": 1.9235984512561872e-05,
-      "loss": 1.0811,
+      "learning_rate": 1.917230404539846e-05,
+      "loss": 1.005,
       "step": 5516
     },
     {
-      "epoch": 0.15152014501112301,
+      "epoch": 0.1565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.9235643469991166e-05,
-      "loss": 1.0299,
+      "learning_rate": 1.9171937885378752e-05,
+      "loss": 0.9917,
       "step": 5517
     },
     {
-      "epoch": 0.15154760923896624,
+      "epoch": 0.156583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.923530235434447e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.9171571647883272e-05,
+      "loss": 1.0356,
       "step": 5518
     },
     {
-      "epoch": 0.15157507346680948,
+      "epoch": 0.15661180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.9234961165624484e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9171205332915113e-05,
+      "loss": 0.9081,
       "step": 5519
     },
     {
-      "epoch": 0.1516025376946527,
+      "epoch": 0.15664018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 1.9234619903833905e-05,
-      "loss": 1.082,
+      "learning_rate": 1.917083894047737e-05,
+      "loss": 0.9413,
       "step": 5520
     },
     {
-      "epoch": 0.15163000192249595,
+      "epoch": 0.15666855845629965,
       "grad_norm": 0.0,
-      "learning_rate": 1.9234278568975442e-05,
-      "loss": 1.0286,
+      "learning_rate": 1.917047247057314e-05,
+      "loss": 1.0912,
       "step": 5521
     },
     {
-      "epoch": 0.15165746615033918,
+      "epoch": 0.15669693530079457,
       "grad_norm": 0.0,
-      "learning_rate": 1.923393716105179e-05,
-      "loss": 0.9604,
+      "learning_rate": 1.9170105923205517e-05,
+      "loss": 1.1143,
       "step": 5522
     },
     {
-      "epoch": 0.15168493037818243,
+      "epoch": 0.15672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.923359568006564e-05,
-      "loss": 1.0922,
+      "learning_rate": 1.91697392983776e-05,
+      "loss": 1.0658,
       "step": 5523
     },
     {
-      "epoch": 0.15171239460602565,
+      "epoch": 0.15675368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.9233254126019717e-05,
-      "loss": 0.975,
+      "learning_rate": 1.9169372596092477e-05,
+      "loss": 1.201,
       "step": 5524
     },
     {
-      "epoch": 0.1517398588338689,
+      "epoch": 0.15678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9232912498916703e-05,
-      "loss": 1.1182,
+      "learning_rate": 1.9169005816353254e-05,
+      "loss": 0.9645,
       "step": 5525
     },
     {
-      "epoch": 0.15176732306171212,
+      "epoch": 0.15681044267877411,
       "grad_norm": 0.0,
-      "learning_rate": 1.923257079875931e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.916863895916303e-05,
+      "loss": 0.8652,
       "step": 5526
     },
     {
-      "epoch": 0.15179478728955537,
+      "epoch": 0.156838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9232229025550242e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.9168272024524895e-05,
+      "loss": 0.9798,
       "step": 5527
     },
     {
-      "epoch": 0.1518222515173986,
+      "epoch": 0.15686719636776392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9231887179292198e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.9167905012441955e-05,
+      "loss": 1.027,
       "step": 5528
     },
     {
-      "epoch": 0.1518497157452418,
+      "epoch": 0.1568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.9231545259987893e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.916753792291731e-05,
+      "loss": 1.0089,
       "step": 5529
     },
     {
-      "epoch": 0.15187717997308506,
+      "epoch": 0.1569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.923120326764002e-05,
-      "loss": 1.0849,
+      "learning_rate": 1.9167170755954062e-05,
+      "loss": 0.9726,
       "step": 5530
     },
     {
-      "epoch": 0.15190464420092828,
+      "epoch": 0.15695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9230861202251297e-05,
-      "loss": 1.069,
+      "learning_rate": 1.916680351155531e-05,
+      "loss": 0.9689,
       "step": 5531
     },
     {
-      "epoch": 0.15193210842877153,
+      "epoch": 0.15698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.9230519063824417e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.9166436189724154e-05,
+      "loss": 1.0567,
       "step": 5532
     },
     {
-      "epoch": 0.15195957265661475,
+      "epoch": 0.15700908059023835,
       "grad_norm": 0.0,
-      "learning_rate": 1.9230176852362098e-05,
-      "loss": 1.076,
+      "learning_rate": 1.91660687904637e-05,
+      "loss": 0.984,
       "step": 5533
     },
     {
-      "epoch": 0.151987036884458,
+      "epoch": 0.15703745743473327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9229834567867046e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.9165701313777055e-05,
+      "loss": 1.0547,
       "step": 5534
     },
     {
-      "epoch": 0.15201450111230122,
+      "epoch": 0.15706583427922816,
       "grad_norm": 0.0,
-      "learning_rate": 1.9229492210341968e-05,
-      "loss": 1.0134,
+      "learning_rate": 1.9165333759667314e-05,
+      "loss": 0.9479,
       "step": 5535
     },
     {
-      "epoch": 0.15204196534014447,
+      "epoch": 0.15709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.9229149779789572e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.916496612813759e-05,
+      "loss": 1.0095,
       "step": 5536
     },
     {
-      "epoch": 0.1520694295679877,
+      "epoch": 0.15712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9228807276212566e-05,
-      "loss": 1.074,
+      "learning_rate": 1.9164598419190982e-05,
+      "loss": 1.0941,
       "step": 5537
     },
     {
-      "epoch": 0.15209689379583094,
+      "epoch": 0.15715096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 1.922846469961366e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.9164230632830604e-05,
+      "loss": 1.0235,
       "step": 5538
     },
     {
-      "epoch": 0.15212435802367416,
+      "epoch": 0.1571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.922812204999557e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.9163862769059554e-05,
+      "loss": 0.9923,
       "step": 5539
     },
     {
-      "epoch": 0.1521518222515174,
+      "epoch": 0.15720771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 1.9227779327361004e-05,
-      "loss": 1.0796,
+      "learning_rate": 1.9163494827880943e-05,
+      "loss": 0.9975,
       "step": 5540
     },
     {
-      "epoch": 0.15217928647936063,
+      "epoch": 0.1572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9227436531712674e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.916312680929788e-05,
+      "loss": 0.9266,
       "step": 5541
     },
     {
-      "epoch": 0.15220675070720388,
+      "epoch": 0.1572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.922709366305329e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.9162758713313473e-05,
+      "loss": 1.0372,
       "step": 5542
     },
     {
-      "epoch": 0.1522342149350471,
+      "epoch": 0.15729284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 1.9226750721385564e-05,
-      "loss": 1.1113,
+      "learning_rate": 1.916239053993083e-05,
+      "loss": 1.0197,
       "step": 5543
     },
     {
-      "epoch": 0.15226167916289032,
+      "epoch": 0.15732122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 1.9226407706712214e-05,
-      "loss": 0.9193,
+      "learning_rate": 1.9162022289153068e-05,
+      "loss": 0.9893,
       "step": 5544
     },
     {
-      "epoch": 0.15228914339073357,
+      "epoch": 0.15734960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 1.9226064619035954e-05,
-      "loss": 1.0679,
+      "learning_rate": 1.9161653960983285e-05,
+      "loss": 1.0234,
       "step": 5545
     },
     {
-      "epoch": 0.1523166076185768,
+      "epoch": 0.15737797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.9225721458359498e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.9161285555424604e-05,
+      "loss": 0.891,
       "step": 5546
     },
     {
-      "epoch": 0.15234407184642004,
+      "epoch": 0.15740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.922537822468556e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.916091707248013e-05,
+      "loss": 1.0404,
       "step": 5547
     },
     {
-      "epoch": 0.15237153607426326,
+      "epoch": 0.15743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.9225034918016856e-05,
-      "loss": 1.0968,
+      "learning_rate": 1.9160548512152975e-05,
+      "loss": 1.0096,
       "step": 5548
     },
     {
-      "epoch": 0.1523990003021065,
+      "epoch": 0.15746311010215663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9224691538356097e-05,
-      "loss": 1.1115,
+      "learning_rate": 1.916017987444626e-05,
+      "loss": 1.0446,
       "step": 5549
     },
     {
-      "epoch": 0.15242646452994973,
+      "epoch": 0.15749148694665152,
       "grad_norm": 0.0,
-      "learning_rate": 1.9224348085706012e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.9159811159363092e-05,
+      "loss": 1.0182,
       "step": 5550
     },
     {
-      "epoch": 0.15245392875779298,
+      "epoch": 0.15751986379114644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9224004560069306e-05,
-      "loss": 1.0858,
+      "learning_rate": 1.9159442366906585e-05,
+      "loss": 0.9917,
       "step": 5551
     },
     {
-      "epoch": 0.1524813929856362,
+      "epoch": 0.15754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.922366096144871e-05,
-      "loss": 0.8678,
+      "learning_rate": 1.9159073497079856e-05,
+      "loss": 0.8899,
       "step": 5552
     },
     {
-      "epoch": 0.15250885721347945,
+      "epoch": 0.1575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.922331728984693e-05,
-      "loss": 1.1356,
+      "learning_rate": 1.9158704549886025e-05,
+      "loss": 1.0366,
       "step": 5553
     },
     {
-      "epoch": 0.15253632144132268,
+      "epoch": 0.1576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.922297354526669e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.9158335525328206e-05,
+      "loss": 0.9791,
       "step": 5554
     },
     {
-      "epoch": 0.15256378566916592,
+      "epoch": 0.15763337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.9222629727710712e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.915796642340951e-05,
+      "loss": 1.0125,
       "step": 5555
     },
     {
-      "epoch": 0.15259124989700915,
+      "epoch": 0.15766174801362087,
       "grad_norm": 0.0,
-      "learning_rate": 1.9222285837181716e-05,
-      "loss": 1.0948,
+      "learning_rate": 1.9157597244133065e-05,
+      "loss": 1.0732,
       "step": 5556
     },
     {
-      "epoch": 0.15261871412485237,
+      "epoch": 0.1576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.9221941873682422e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.915722798750198e-05,
+      "loss": 1.0345,
       "step": 5557
     },
     {
-      "epoch": 0.15264617835269562,
+      "epoch": 0.15771850170261068,
       "grad_norm": 0.0,
-      "learning_rate": 1.922159783721555e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.915685865351938e-05,
+      "loss": 1.1241,
       "step": 5558
     },
     {
-      "epoch": 0.15267364258053884,
+      "epoch": 0.15774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9221253727783827e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.9156489242188385e-05,
+      "loss": 1.038,
       "step": 5559
     },
     {
-      "epoch": 0.1527011068083821,
+      "epoch": 0.15777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.922090954538997e-05,
-      "loss": 1.0813,
+      "learning_rate": 1.9156119753512114e-05,
+      "loss": 1.1698,
       "step": 5560
     },
     {
-      "epoch": 0.1527285710362253,
+      "epoch": 0.15780363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.9220565290036707e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.9155750187493684e-05,
+      "loss": 0.9265,
       "step": 5561
     },
     {
-      "epoch": 0.15275603526406856,
+      "epoch": 0.15783200908059025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9220220961726755e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.9155380544136223e-05,
+      "loss": 0.98,
       "step": 5562
     },
     {
-      "epoch": 0.15278349949191178,
+      "epoch": 0.15786038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9219876560462846e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.9155010823442852e-05,
+      "loss": 0.9771,
       "step": 5563
     },
     {
-      "epoch": 0.15281096371975503,
+      "epoch": 0.15788876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 1.9219532086247702e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.9154641025416694e-05,
+      "loss": 0.9859,
       "step": 5564
     },
     {
-      "epoch": 0.15283842794759825,
+      "epoch": 0.15791713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.921918753908405e-05,
-      "loss": 1.1475,
+      "learning_rate": 1.915427115006087e-05,
+      "loss": 0.9267,
       "step": 5565
     },
     {
-      "epoch": 0.1528658921754415,
+      "epoch": 0.1579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.9218842918974618e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.9153901197378507e-05,
+      "loss": 1.0977,
       "step": 5566
     },
     {
-      "epoch": 0.15289335640328472,
+      "epoch": 0.1579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9218498225922126e-05,
-      "loss": 1.1074,
+      "learning_rate": 1.9153531167372725e-05,
+      "loss": 1.0752,
       "step": 5567
     },
     {
-      "epoch": 0.15292082063112797,
+      "epoch": 0.1580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.9218153459929305e-05,
-      "loss": 1.021,
+      "learning_rate": 1.915316106004666e-05,
+      "loss": 0.9445,
       "step": 5568
     },
     {
-      "epoch": 0.1529482848589712,
+      "epoch": 0.1580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9217808620998884e-05,
-      "loss": 1.1077,
+      "learning_rate": 1.9152790875403428e-05,
+      "loss": 1.1137,
       "step": 5569
     },
     {
-      "epoch": 0.1529757490868144,
+      "epoch": 0.15805902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.921746370913359e-05,
-      "loss": 1.0405,
+      "learning_rate": 1.915242061344616e-05,
+      "loss": 1.0955,
       "step": 5570
     },
     {
-      "epoch": 0.15300321331465766,
+      "epoch": 0.15808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.921711872433615e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.9152050274177988e-05,
+      "loss": 1.0126,
       "step": 5571
     },
     {
-      "epoch": 0.15303067754250088,
+      "epoch": 0.15811577752553915,
       "grad_norm": 0.0,
-      "learning_rate": 1.92167736666093e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.915167985760203e-05,
+      "loss": 1.0206,
       "step": 5572
     },
     {
-      "epoch": 0.15305814177034413,
+      "epoch": 0.15814415437003404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9216428535955765e-05,
-      "loss": 1.0664,
+      "learning_rate": 1.915130936372142e-05,
+      "loss": 0.9755,
       "step": 5573
     },
     {
-      "epoch": 0.15308560599818735,
+      "epoch": 0.15817253121452896,
       "grad_norm": 0.0,
-      "learning_rate": 1.921608333237828e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.9150938792539297e-05,
+      "loss": 1.0434,
       "step": 5574
     },
     {
-      "epoch": 0.1531130702260306,
+      "epoch": 0.15820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.921573805587957e-05,
-      "loss": 1.1026,
+      "learning_rate": 1.9150568144058774e-05,
+      "loss": 0.9305,
       "step": 5575
     },
     {
-      "epoch": 0.15314053445387382,
+      "epoch": 0.15822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9215392706462372e-05,
-      "loss": 1.0926,
+      "learning_rate": 1.9150197418282993e-05,
+      "loss": 1.1519,
       "step": 5576
     },
     {
-      "epoch": 0.15316799868171707,
+      "epoch": 0.15825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.9215047284129416e-05,
-      "loss": 1.1181,
+      "learning_rate": 1.9149826615215085e-05,
+      "loss": 1.0011,
       "step": 5577
     },
     {
-      "epoch": 0.1531954629095603,
+      "epoch": 0.1582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.921470178888344e-05,
-      "loss": 1.1403,
+      "learning_rate": 1.914945573485818e-05,
+      "loss": 0.9897,
       "step": 5578
     },
     {
-      "epoch": 0.15322292713740354,
+      "epoch": 0.1583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9214356220727174e-05,
-      "loss": 1.076,
+      "learning_rate": 1.9149084777215412e-05,
+      "loss": 0.8718,
       "step": 5579
     },
     {
-      "epoch": 0.15325039136524676,
+      "epoch": 0.1583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.921401057966335e-05,
-      "loss": 1.1213,
+      "learning_rate": 1.9148713742289914e-05,
+      "loss": 1.0128,
       "step": 5580
     },
     {
-      "epoch": 0.15327785559309,
+      "epoch": 0.1583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.9213664865694703e-05,
-      "loss": 1.058,
+      "learning_rate": 1.9148342630084817e-05,
+      "loss": 0.9705,
       "step": 5581
     },
     {
-      "epoch": 0.15330531982093323,
+      "epoch": 0.15839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9213319078823974e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.9147971440603257e-05,
+      "loss": 0.9387,
       "step": 5582
     },
     {
-      "epoch": 0.15333278404877645,
+      "epoch": 0.15842792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.921297321905389e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.9147600173848375e-05,
+      "loss": 1.0447,
       "step": 5583
     },
     {
-      "epoch": 0.1533602482766197,
+      "epoch": 0.15845629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9212627286387203e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.91472288298233e-05,
+      "loss": 0.9703,
       "step": 5584
     },
     {
-      "epoch": 0.15338771250446293,
+      "epoch": 0.15848467650397277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9212281280826632e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.9146857408531174e-05,
+      "loss": 1.0214,
       "step": 5585
     },
     {
-      "epoch": 0.15341517673230617,
+      "epoch": 0.15851305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9211935202374925e-05,
-      "loss": 1.0778,
+      "learning_rate": 1.914648590997513e-05,
+      "loss": 0.9598,
       "step": 5586
     },
     {
-      "epoch": 0.1534426409601494,
+      "epoch": 0.15854143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 1.921158905103482e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.9146114334158314e-05,
+      "loss": 1.0342,
       "step": 5587
     },
     {
-      "epoch": 0.15347010518799264,
+      "epoch": 0.15856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.9211242826809055e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.914574268108385e-05,
+      "loss": 1.0658,
       "step": 5588
     },
     {
-      "epoch": 0.15349756941583587,
+      "epoch": 0.15859818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 1.9210896529700363e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.9145370950754896e-05,
+      "loss": 1.0171,
       "step": 5589
     },
     {
-      "epoch": 0.15352503364367912,
+      "epoch": 0.1586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9210550159711493e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.9144999143174576e-05,
+      "loss": 1.0626,
       "step": 5590
     },
     {
-      "epoch": 0.15355249787152234,
+      "epoch": 0.15865493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 1.9210203716845185e-05,
-      "loss": 1.0341,
+      "learning_rate": 1.9144627258346042e-05,
+      "loss": 1.0273,
       "step": 5591
     },
     {
-      "epoch": 0.15357996209936559,
+      "epoch": 0.158683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9209857201104175e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.9144255296272426e-05,
+      "loss": 1.0193,
       "step": 5592
     },
     {
-      "epoch": 0.1536074263272088,
+      "epoch": 0.1587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.920951061249121e-05,
-      "loss": 1.058,
+      "learning_rate": 1.9143883256956875e-05,
+      "loss": 1.0134,
       "step": 5593
     },
     {
-      "epoch": 0.15363489055505206,
+      "epoch": 0.15874006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 1.9209163951009024e-05,
-      "loss": 1.0915,
+      "learning_rate": 1.9143511140402532e-05,
+      "loss": 1.1464,
       "step": 5594
     },
     {
-      "epoch": 0.15366235478289528,
+      "epoch": 0.15876844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.920881721666037e-05,
-      "loss": 1.0583,
+      "learning_rate": 1.914313894661254e-05,
+      "loss": 1.0806,
       "step": 5595
     },
     {
-      "epoch": 0.15368981901073853,
+      "epoch": 0.15879682179341656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9208470409447987e-05,
-      "loss": 1.048,
+      "learning_rate": 1.9142766675590043e-05,
+      "loss": 0.8908,
       "step": 5596
     },
     {
-      "epoch": 0.15371728323858175,
+      "epoch": 0.15882519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.9208123529374616e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.914239432733818e-05,
+      "loss": 0.9206,
       "step": 5597
     },
     {
-      "epoch": 0.15374474746642497,
+      "epoch": 0.15885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.9207776576443007e-05,
-      "loss": 0.965,
+      "learning_rate": 1.9142021901860107e-05,
+      "loss": 1.031,
       "step": 5598
     },
     {
-      "epoch": 0.15377221169426822,
+      "epoch": 0.15888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.9207429550655903e-05,
-      "loss": 1.0456,
+      "learning_rate": 1.914164939915896e-05,
+      "loss": 1.0123,
       "step": 5599
     },
     {
-      "epoch": 0.15379967592211144,
+      "epoch": 0.15891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.9207082452016048e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.9141276819237893e-05,
+      "loss": 0.9437,
       "step": 5600
     },
     {
-      "epoch": 0.1538271401499547,
+      "epoch": 0.15893870601589102,
       "grad_norm": 0.0,
-      "learning_rate": 1.9206735280526195e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.914090416210005e-05,
+      "loss": 1.082,
       "step": 5601
     },
     {
-      "epoch": 0.1538546043777979,
+      "epoch": 0.15896708286038594,
       "grad_norm": 0.0,
-      "learning_rate": 1.920638803618908e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.9140531427748574e-05,
+      "loss": 1.12,
       "step": 5602
     },
     {
-      "epoch": 0.15388206860564116,
+      "epoch": 0.15899545970488083,
       "grad_norm": 0.0,
-      "learning_rate": 1.920604071900746e-05,
-      "loss": 1.0761,
+      "learning_rate": 1.9140158616186625e-05,
+      "loss": 1.0705,
       "step": 5603
     },
     {
-      "epoch": 0.15390953283348438,
+      "epoch": 0.1590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.920569332898408e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.9139785727417338e-05,
+      "loss": 1.0289,
       "step": 5604
     },
     {
-      "epoch": 0.15393699706132763,
+      "epoch": 0.1590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.920534586612169e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.9139412761443874e-05,
+      "loss": 0.9597,
       "step": 5605
     },
     {
-      "epoch": 0.15396446128917085,
+      "epoch": 0.1590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.9204998330423034e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.9139039718269378e-05,
+      "loss": 0.9722,
       "step": 5606
     },
     {
-      "epoch": 0.1539919255170141,
+      "epoch": 0.15910896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 1.9204650721890866e-05,
-      "loss": 1.0486,
+      "learning_rate": 1.9138666597897004e-05,
+      "loss": 1.0728,
       "step": 5607
     },
     {
-      "epoch": 0.15401938974485732,
+      "epoch": 0.1591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.9204303040527935e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.9138293400329902e-05,
+      "loss": 1.0102,
       "step": 5608
     },
     {
-      "epoch": 0.15404685397270057,
+      "epoch": 0.15916572077185018,
       "grad_norm": 0.0,
-      "learning_rate": 1.9203955286336996e-05,
-      "loss": 1.1115,
+      "learning_rate": 1.9137920125571225e-05,
+      "loss": 0.9041,
       "step": 5609
     },
     {
-      "epoch": 0.1540743182005438,
+      "epoch": 0.15919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.9203607459320793e-05,
-      "loss": 1.0454,
+      "learning_rate": 1.9137546773624126e-05,
+      "loss": 1.0009,
       "step": 5610
     },
     {
-      "epoch": 0.154101782428387,
+      "epoch": 0.15922247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.9203259559482088e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.913717334449176e-05,
+      "loss": 1.0451,
       "step": 5611
     },
     {
-      "epoch": 0.15412924665623026,
+      "epoch": 0.15925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.920291158682363e-05,
-      "loss": 1.0778,
+      "learning_rate": 1.913679983817728e-05,
+      "loss": 0.9232,
       "step": 5612
     },
     {
-      "epoch": 0.15415671088407348,
+      "epoch": 0.15927922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 1.9202563541348162e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.9136426254683837e-05,
+      "loss": 0.9788,
       "step": 5613
     },
     {
-      "epoch": 0.15418417511191673,
+      "epoch": 0.15930760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.920221542305845e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.9136052594014594e-05,
+      "loss": 1.0347,
       "step": 5614
     },
     {
-      "epoch": 0.15421163933975995,
+      "epoch": 0.15933598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9201867231957248e-05,
-      "loss": 1.0451,
+      "learning_rate": 1.9135678856172704e-05,
+      "loss": 0.994,
       "step": 5615
     },
     {
-      "epoch": 0.1542391035676032,
+      "epoch": 0.15936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.9201518968047306e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.9135305041161322e-05,
+      "loss": 0.9903,
       "step": 5616
     },
     {
-      "epoch": 0.15426656779544642,
+      "epoch": 0.1593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.9201170631331382e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.913493114898361e-05,
+      "loss": 1.0737,
       "step": 5617
     },
     {
-      "epoch": 0.15429403202328967,
+      "epoch": 0.1594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.920082222181223e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.9134557179642722e-05,
+      "loss": 1.0553,
       "step": 5618
     },
     {
-      "epoch": 0.1543214962511329,
+      "epoch": 0.15944948921679908,
       "grad_norm": 0.0,
-      "learning_rate": 1.920047373949261e-05,
-      "loss": 1.0951,
+      "learning_rate": 1.913418313314182e-05,
+      "loss": 1.0131,
       "step": 5619
     },
     {
-      "epoch": 0.15434896047897614,
+      "epoch": 0.159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.920012518437528e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.9133809009484063e-05,
+      "loss": 1.0977,
       "step": 5620
     },
     {
-      "epoch": 0.15437642470681937,
+      "epoch": 0.15950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.9199776556462994e-05,
-      "loss": 1.0845,
+      "learning_rate": 1.913343480867261e-05,
+      "loss": 0.8984,
       "step": 5621
     },
     {
-      "epoch": 0.15440388893466261,
+      "epoch": 0.15953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.919942785575851e-05,
-      "loss": 1.1295,
+      "learning_rate": 1.913306053071062e-05,
+      "loss": 1.0982,
       "step": 5622
     },
     {
-      "epoch": 0.15443135316250584,
+      "epoch": 0.15956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9199079082264592e-05,
-      "loss": 1.0141,
+      "learning_rate": 1.913268617560126e-05,
+      "loss": 0.929,
       "step": 5623
     },
     {
-      "epoch": 0.15445881739034906,
+      "epoch": 0.15959137343927354,
       "grad_norm": 0.0,
-      "learning_rate": 1.9198730235983998e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.913231174334769e-05,
+      "loss": 0.9746,
       "step": 5624
     },
     {
-      "epoch": 0.1544862816181923,
+      "epoch": 0.15961975028376846,
       "grad_norm": 0.0,
-      "learning_rate": 1.9198381316919485e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.9131937233953072e-05,
+      "loss": 1.0604,
       "step": 5625
     },
     {
-      "epoch": 0.15451374584603553,
+      "epoch": 0.15964812712826335,
       "grad_norm": 0.0,
-      "learning_rate": 1.9198032325073816e-05,
-      "loss": 1.1157,
+      "learning_rate": 1.9131562647420568e-05,
+      "loss": 0.9603,
       "step": 5626
     },
     {
-      "epoch": 0.15454121007387878,
+      "epoch": 0.15967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.919768326044975e-05,
-      "loss": 1.047,
+      "learning_rate": 1.9131187983753344e-05,
+      "loss": 1.0482,
       "step": 5627
     },
     {
-      "epoch": 0.154568674301722,
+      "epoch": 0.15970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9197334123050055e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.9130813242954564e-05,
+      "loss": 0.9981,
       "step": 5628
     },
     {
-      "epoch": 0.15459613852956525,
+      "epoch": 0.159733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9196984912877486e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.9130438425027397e-05,
+      "loss": 1.0447,
       "step": 5629
     },
     {
-      "epoch": 0.15462360275740847,
+      "epoch": 0.1597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.9196635629934816e-05,
-      "loss": 1.1445,
+      "learning_rate": 1.9130063529975004e-05,
+      "loss": 0.909,
       "step": 5630
     },
     {
-      "epoch": 0.15465106698525172,
+      "epoch": 0.1597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.91962862742248e-05,
-      "loss": 1.0026,
+      "learning_rate": 1.9129688557800555e-05,
+      "loss": 0.9181,
       "step": 5631
     },
     {
-      "epoch": 0.15467853121309494,
+      "epoch": 0.1598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9195936845750203e-05,
-      "loss": 1.0052,
+      "learning_rate": 1.9129313508507213e-05,
+      "loss": 0.9089,
       "step": 5632
     },
     {
-      "epoch": 0.1547059954409382,
+      "epoch": 0.15984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.919558734451379e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.9128938382098155e-05,
+      "loss": 1.0451,
       "step": 5633
     },
     {
-      "epoch": 0.1547334596687814,
+      "epoch": 0.15987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.919523777051833e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.912856317857654e-05,
+      "loss": 1.0278,
       "step": 5634
     },
     {
-      "epoch": 0.15476092389662466,
+      "epoch": 0.15990351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.919488812376659e-05,
-      "loss": 1.1432,
+      "learning_rate": 1.9128187897945545e-05,
+      "loss": 0.9689,
       "step": 5635
     },
     {
-      "epoch": 0.15478838812446788,
+      "epoch": 0.15993189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 1.9194538404261326e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.9127812540208333e-05,
+      "loss": 1.1801,
       "step": 5636
     },
     {
-      "epoch": 0.15481585235231113,
+      "epoch": 0.15996027241770716,
       "grad_norm": 0.0,
-      "learning_rate": 1.9194188612005317e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.912743710536808e-05,
+      "loss": 1.0443,
       "step": 5637
     },
     {
-      "epoch": 0.15484331658015435,
+      "epoch": 0.15998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9193838747001327e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.9127061593427954e-05,
+      "loss": 0.9183,
       "step": 5638
     },
     {
-      "epoch": 0.15487078080799757,
+      "epoch": 0.16001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.9193488809252122e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.912668600439113e-05,
+      "loss": 0.9731,
       "step": 5639
     },
     {
-      "epoch": 0.15489824503584082,
+      "epoch": 0.16004540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 1.919313879876047e-05,
-      "loss": 1.1447,
+      "learning_rate": 1.912631033826078e-05,
+      "loss": 1.0636,
       "step": 5640
     },
     {
-      "epoch": 0.15492570926368404,
+      "epoch": 0.1600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.919278871552915e-05,
-      "loss": 1.1677,
+      "learning_rate": 1.9125934595040072e-05,
+      "loss": 1.0396,
       "step": 5641
     },
     {
-      "epoch": 0.1549531734915273,
+      "epoch": 0.16010215664018163,
       "grad_norm": 0.0,
-      "learning_rate": 1.919243855956092e-05,
-      "loss": 1.0448,
+      "learning_rate": 1.912555877473219e-05,
+      "loss": 0.9003,
       "step": 5642
     },
     {
-      "epoch": 0.1549806377193705,
+      "epoch": 0.1601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.9192088330858553e-05,
-      "loss": 1.093,
+      "learning_rate": 1.91251828773403e-05,
+      "loss": 1.0492,
       "step": 5643
     },
     {
-      "epoch": 0.15500810194721376,
+      "epoch": 0.1601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.9191738029424825e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.9124806902867577e-05,
+      "loss": 0.9931,
       "step": 5644
     },
     {
-      "epoch": 0.15503556617505698,
+      "epoch": 0.1601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.9191387655262504e-05,
-      "loss": 1.138,
+      "learning_rate": 1.9124430851317204e-05,
+      "loss": 1.0289,
       "step": 5645
     },
     {
-      "epoch": 0.15506303040290023,
+      "epoch": 0.16021566401816117,
       "grad_norm": 0.0,
-      "learning_rate": 1.9191037208374365e-05,
-      "loss": 0.9954,
+      "learning_rate": 1.9124054722692352e-05,
+      "loss": 1.0815,
       "step": 5646
     },
     {
-      "epoch": 0.15509049463074345,
+      "epoch": 0.16024404086265606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9190686688763178e-05,
-      "loss": 1.08,
+      "learning_rate": 1.91236785169962e-05,
+      "loss": 1.0264,
       "step": 5647
     },
     {
-      "epoch": 0.1551179588585867,
+      "epoch": 0.16027241770715098,
       "grad_norm": 0.0,
-      "learning_rate": 1.919033609643172e-05,
-      "loss": 1.1107,
+      "learning_rate": 1.9123302234231924e-05,
+      "loss": 0.9789,
       "step": 5648
     },
     {
-      "epoch": 0.15514542308642992,
+      "epoch": 0.16030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.918998543138276e-05,
-      "loss": 1.023,
+      "learning_rate": 1.9122925874402707e-05,
+      "loss": 0.9834,
       "step": 5649
     },
     {
-      "epoch": 0.15517288731427317,
+      "epoch": 0.16032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.9189634693619078e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.912254943751172e-05,
+      "loss": 1.0338,
       "step": 5650
     },
     {
-      "epoch": 0.1552003515421164,
+      "epoch": 0.16035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9189283883143445e-05,
-      "loss": 1.1427,
+      "learning_rate": 1.9122172923562152e-05,
+      "loss": 1.0242,
       "step": 5651
     },
     {
-      "epoch": 0.15522781576995962,
+      "epoch": 0.16038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.918893299995864e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.912179633255718e-05,
+      "loss": 0.8877,
       "step": 5652
     },
     {
-      "epoch": 0.15525527999780286,
+      "epoch": 0.1604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.9188582044067433e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.912141966449998e-05,
+      "loss": 1.0226,
       "step": 5653
     },
     {
-      "epoch": 0.15528274422564609,
+      "epoch": 0.16044267877412033,
       "grad_norm": 0.0,
-      "learning_rate": 1.918823101547261e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.9121042919393742e-05,
+      "loss": 1.0276,
       "step": 5654
     },
     {
-      "epoch": 0.15531020845348933,
+      "epoch": 0.16047105561861522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9187879914176943e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.9120666097241644e-05,
+      "loss": 1.019,
       "step": 5655
     },
     {
-      "epoch": 0.15533767268133256,
+      "epoch": 0.1604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.9187528740183213e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.9120289198046868e-05,
+      "loss": 1.1377,
       "step": 5656
     },
     {
-      "epoch": 0.1553651369091758,
+      "epoch": 0.160527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9187177493494194e-05,
-      "loss": 1.0915,
+      "learning_rate": 1.91199122218126e-05,
+      "loss": 0.9869,
       "step": 5657
     },
     {
-      "epoch": 0.15539260113701903,
+      "epoch": 0.16055618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 1.918682617411267e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.9119535168542024e-05,
+      "loss": 0.9179,
       "step": 5658
     },
     {
-      "epoch": 0.15542006536486228,
+      "epoch": 0.16058456299659477,
       "grad_norm": 0.0,
-      "learning_rate": 1.918647478204142e-05,
-      "loss": 0.948,
+      "learning_rate": 1.911915803823832e-05,
+      "loss": 1.0453,
       "step": 5659
     },
     {
-      "epoch": 0.1554475295927055,
+      "epoch": 0.16061293984108968,
       "grad_norm": 0.0,
-      "learning_rate": 1.9186123317283216e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.911878083090468e-05,
+      "loss": 0.8893,
       "step": 5660
     },
     {
-      "epoch": 0.15547499382054875,
+      "epoch": 0.16064131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.918577177984085e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.9118403546544294e-05,
+      "loss": 1.1146,
       "step": 5661
     },
     {
-      "epoch": 0.15550245804839197,
+      "epoch": 0.16066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.91854201697171e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.911802618516034e-05,
+      "loss": 1.1563,
       "step": 5662
     },
     {
-      "epoch": 0.15552992227623522,
+      "epoch": 0.16069807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.9185068486914748e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.911764874675601e-05,
+      "loss": 0.9517,
       "step": 5663
     },
     {
-      "epoch": 0.15555738650407844,
+      "epoch": 0.16072644721906923,
       "grad_norm": 0.0,
-      "learning_rate": 1.9184716731436575e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.9117271231334488e-05,
+      "loss": 1.066,
       "step": 5664
     },
     {
-      "epoch": 0.15558485073192166,
+      "epoch": 0.16075482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.9184364903285368e-05,
-      "loss": 1.0864,
+      "learning_rate": 1.9116893638898965e-05,
+      "loss": 0.9962,
       "step": 5665
     },
     {
-      "epoch": 0.1556123149597649,
+      "epoch": 0.16078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.9184013002463904e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9116515969452637e-05,
+      "loss": 1.0047,
       "step": 5666
     },
     {
-      "epoch": 0.15563977918760813,
+      "epoch": 0.16081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.9183661028974975e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.9116138222998685e-05,
+      "loss": 1.0599,
       "step": 5667
     },
     {
-      "epoch": 0.15566724341545138,
+      "epoch": 0.1608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.918330898282136e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.9115760399540303e-05,
+      "loss": 0.9026,
       "step": 5668
     },
     {
-      "epoch": 0.1556947076432946,
+      "epoch": 0.1608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.918295686400585e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.9115382499080687e-05,
+      "loss": 0.987,
       "step": 5669
     },
     {
-      "epoch": 0.15572217187113785,
+      "epoch": 0.16089670828603858,
       "grad_norm": 0.0,
-      "learning_rate": 1.9182604672531223e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.911500452162302e-05,
+      "loss": 0.9966,
       "step": 5670
     },
     {
-      "epoch": 0.15574963609898107,
+      "epoch": 0.1609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.9182252408400276e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.9114626467170505e-05,
+      "loss": 1.0113,
       "step": 5671
     },
     {
-      "epoch": 0.15577710032682432,
+      "epoch": 0.16095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.918190007161579e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.911424833572633e-05,
+      "loss": 1.053,
       "step": 5672
     },
     {
-      "epoch": 0.15580456455466754,
+      "epoch": 0.16098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9181547662180554e-05,
-      "loss": 1.036,
+      "learning_rate": 1.9113870127293688e-05,
+      "loss": 0.9873,
       "step": 5673
     },
     {
-      "epoch": 0.1558320287825108,
+      "epoch": 0.16101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.9181195180097353e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.911349184187577e-05,
+      "loss": 0.992,
       "step": 5674
     },
     {
-      "epoch": 0.155859493010354,
+      "epoch": 0.16103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.918084262536898e-05,
-      "loss": 0.8196,
+      "learning_rate": 1.9113113479475784e-05,
+      "loss": 0.9713,
       "step": 5675
     },
     {
-      "epoch": 0.15588695723819726,
+      "epoch": 0.16106696935300793,
       "grad_norm": 0.0,
-      "learning_rate": 1.9180489997998227e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.911273504009692e-05,
+      "loss": 0.9906,
       "step": 5676
     },
     {
-      "epoch": 0.15591442146604048,
+      "epoch": 0.16109534619750285,
       "grad_norm": 0.0,
-      "learning_rate": 1.918013729798788e-05,
-      "loss": 1.1799,
+      "learning_rate": 1.911235652374237e-05,
+      "loss": 0.9621,
       "step": 5677
     },
     {
-      "epoch": 0.1559418856938837,
+      "epoch": 0.16112372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.917978452534073e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.9111977930415334e-05,
+      "loss": 0.9903,
       "step": 5678
     },
     {
-      "epoch": 0.15596934992172695,
+      "epoch": 0.16115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.917943168005957e-05,
-      "loss": 1.2133,
+      "learning_rate": 1.9111599260119013e-05,
+      "loss": 0.9901,
       "step": 5679
     },
     {
-      "epoch": 0.15599681414957017,
+      "epoch": 0.1611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.917907876214719e-05,
-      "loss": 1.0248,
+      "learning_rate": 1.9111220512856602e-05,
+      "loss": 1.0062,
       "step": 5680
     },
     {
-      "epoch": 0.15602427837741342,
+      "epoch": 0.1612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.9178725771606383e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.9110841688631303e-05,
+      "loss": 0.9371,
       "step": 5681
     },
     {
-      "epoch": 0.15605174260525664,
+      "epoch": 0.1612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9178372708439942e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.9110462787446313e-05,
+      "loss": 0.9251,
       "step": 5682
     },
     {
-      "epoch": 0.1560792068330999,
+      "epoch": 0.1612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.917801957265066e-05,
-      "loss": 1.0708,
+      "learning_rate": 1.9110083809304835e-05,
+      "loss": 0.9934,
       "step": 5683
     },
     {
-      "epoch": 0.15610667106094311,
+      "epoch": 0.1612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.9177666364241333e-05,
-      "loss": 1.0884,
+      "learning_rate": 1.910970475421007e-05,
+      "loss": 1.0958,
       "step": 5684
     },
     {
-      "epoch": 0.15613413528878636,
+      "epoch": 0.16132236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 1.9177313083214755e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.910932562216522e-05,
+      "loss": 1.0046,
       "step": 5685
     },
     {
-      "epoch": 0.15616159951662958,
+      "epoch": 0.16135073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.917695972957372e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.9108946413173485e-05,
+      "loss": 1.0488,
       "step": 5686
     },
     {
-      "epoch": 0.15618906374447283,
+      "epoch": 0.16137911464245175,
       "grad_norm": 0.0,
-      "learning_rate": 1.917660630332103e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.9108567127238072e-05,
+      "loss": 0.9313,
       "step": 5687
     },
     {
-      "epoch": 0.15621652797231605,
+      "epoch": 0.16140749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 1.917625280445947e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.9108187764362185e-05,
+      "loss": 0.9604,
       "step": 5688
     },
     {
-      "epoch": 0.1562439922001593,
+      "epoch": 0.16143586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.9175899232991843e-05,
-      "loss": 1.0816,
+      "learning_rate": 1.910780832454902e-05,
+      "loss": 1.0008,
       "step": 5689
     },
     {
-      "epoch": 0.15627145642800253,
+      "epoch": 0.16146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.917554558892095e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.9107428807801796e-05,
+      "loss": 0.9893,
       "step": 5690
     },
     {
-      "epoch": 0.15629892065584577,
+      "epoch": 0.16149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.9175191872249583e-05,
-      "loss": 1.0626,
+      "learning_rate": 1.9107049214123704e-05,
+      "loss": 1.0449,
       "step": 5691
     },
     {
-      "epoch": 0.156326384883689,
+      "epoch": 0.1615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.9174838082980547e-05,
-      "loss": 1.1123,
+      "learning_rate": 1.9106669543517965e-05,
+      "loss": 0.9731,
       "step": 5692
     },
     {
-      "epoch": 0.15635384911153222,
+      "epoch": 0.1615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.9174484221116637e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.9106289795987774e-05,
+      "loss": 0.9041,
       "step": 5693
     },
     {
-      "epoch": 0.15638131333937547,
+      "epoch": 0.16157775255391602,
       "grad_norm": 0.0,
-      "learning_rate": 1.9174130286660654e-05,
-      "loss": 1.1108,
+      "learning_rate": 1.9105909971536346e-05,
+      "loss": 1.1672,
       "step": 5694
     },
     {
-      "epoch": 0.1564087775672187,
+      "epoch": 0.1616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.91737762796154e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.910553007016689e-05,
+      "loss": 1.0721,
       "step": 5695
     },
     {
-      "epoch": 0.15643624179506194,
+      "epoch": 0.1616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.9173422199983668e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.9105150091882608e-05,
+      "loss": 0.9245,
       "step": 5696
     },
     {
-      "epoch": 0.15646370602290516,
+      "epoch": 0.16166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.917306804776827e-05,
-      "loss": 1.0613,
+      "learning_rate": 1.9104770036686716e-05,
+      "loss": 0.9045,
       "step": 5697
     },
     {
-      "epoch": 0.1564911702507484,
+      "epoch": 0.16169125993189556,
       "grad_norm": 0.0,
-      "learning_rate": 1.9172713822972007e-05,
-      "loss": 1.0697,
+      "learning_rate": 1.910438990458242e-05,
+      "loss": 0.9825,
       "step": 5698
     },
     {
-      "epoch": 0.15651863447859163,
+      "epoch": 0.16171963677639045,
       "grad_norm": 0.0,
-      "learning_rate": 1.9172359525597677e-05,
-      "loss": 1.0863,
+      "learning_rate": 1.9104009695572933e-05,
+      "loss": 0.9807,
       "step": 5699
     },
     {
-      "epoch": 0.15654609870643488,
+      "epoch": 0.16174801362088537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9172005155648087e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.9103629409661468e-05,
+      "loss": 1.157,
       "step": 5700
     },
     {
-      "epoch": 0.1565735629342781,
+      "epoch": 0.16177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.917165071312604e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.9103249046851237e-05,
+      "loss": 1.0272,
       "step": 5701
     },
     {
-      "epoch": 0.15660102716212135,
+      "epoch": 0.16180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9171296198034334e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.910286860714545e-05,
+      "loss": 1.0145,
       "step": 5702
     },
     {
-      "epoch": 0.15662849138996457,
+      "epoch": 0.16183314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.9170941610375782e-05,
-      "loss": 1.0848,
+      "learning_rate": 1.9102488090547323e-05,
+      "loss": 1.0505,
       "step": 5703
     },
     {
-      "epoch": 0.15665595561780782,
+      "epoch": 0.16186152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 1.9170586950153192e-05,
-      "loss": 0.892,
+      "learning_rate": 1.9102107497060072e-05,
+      "loss": 0.9296,
       "step": 5704
     },
     {
-      "epoch": 0.15668341984565104,
+      "epoch": 0.16188989784335983,
       "grad_norm": 0.0,
-      "learning_rate": 1.917023221736936e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.910172682668691e-05,
+      "loss": 1.0712,
       "step": 5705
     },
     {
-      "epoch": 0.15671088407349426,
+      "epoch": 0.16191827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9169877412027103e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.9101346079431052e-05,
+      "loss": 1.1414,
       "step": 5706
     },
     {
-      "epoch": 0.1567383483013375,
+      "epoch": 0.1619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.9169522534129222e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.910096525529571e-05,
+      "loss": 1.0324,
       "step": 5707
     },
     {
-      "epoch": 0.15676581252918073,
+      "epoch": 0.1619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9169167583678524e-05,
-      "loss": 1.061,
+      "learning_rate": 1.910058435428411e-05,
+      "loss": 0.9553,
       "step": 5708
     },
     {
-      "epoch": 0.15679327675702398,
+      "epoch": 0.16200340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 1.9168812560677825e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.9100203376399465e-05,
+      "loss": 0.9121,
       "step": 5709
     },
     {
-      "epoch": 0.1568207409848672,
+      "epoch": 0.16203178206583427,
       "grad_norm": 0.0,
-      "learning_rate": 1.916845746512993e-05,
-      "loss": 1.109,
+      "learning_rate": 1.9099822321644993e-05,
+      "loss": 0.9136,
       "step": 5710
     },
     {
-      "epoch": 0.15684820521271045,
+      "epoch": 0.16206015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 1.9168102297037642e-05,
-      "loss": 1.1002,
+      "learning_rate": 1.909944119002391e-05,
+      "loss": 1.0497,
       "step": 5711
     },
     {
-      "epoch": 0.15687566944055367,
+      "epoch": 0.16208853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.916774705640378e-05,
-      "loss": 1.125,
+      "learning_rate": 1.9099059981539437e-05,
+      "loss": 0.9849,
       "step": 5712
     },
     {
-      "epoch": 0.15690313366839692,
+      "epoch": 0.16211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.9167391743231154e-05,
-      "loss": 1.1112,
+      "learning_rate": 1.90986786961948e-05,
+      "loss": 1.0691,
       "step": 5713
     },
     {
-      "epoch": 0.15693059789624014,
+      "epoch": 0.16214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.916703635752257e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.9098297333993214e-05,
+      "loss": 0.9991,
       "step": 5714
     },
     {
-      "epoch": 0.1569580621240834,
+      "epoch": 0.16217366628830873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9166680899280846e-05,
-      "loss": 1.0926,
+      "learning_rate": 1.9097915894937902e-05,
+      "loss": 1.1378,
       "step": 5715
     },
     {
-      "epoch": 0.1569855263519266,
+      "epoch": 0.16220204313280362,
       "grad_norm": 0.0,
-      "learning_rate": 1.9166325368508794e-05,
-      "loss": 1.1075,
+      "learning_rate": 1.9097534379032083e-05,
+      "loss": 0.9415,
       "step": 5716
     },
     {
-      "epoch": 0.15701299057976986,
+      "epoch": 0.16223041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.9165969765209223e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.9097152786278983e-05,
+      "loss": 0.9821,
       "step": 5717
     },
     {
-      "epoch": 0.15704045480761308,
+      "epoch": 0.16225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.9165614089384944e-05,
-      "loss": 0.9,
+      "learning_rate": 1.9096771116681825e-05,
+      "loss": 1.0465,
       "step": 5718
     },
     {
-      "epoch": 0.1570679190354563,
+      "epoch": 0.1622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.916525834103878e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.9096389370243833e-05,
+      "loss": 0.9514,
       "step": 5719
     },
     {
-      "epoch": 0.15709538326329955,
+      "epoch": 0.1623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.9164902520173543e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.909600754696823e-05,
+      "loss": 1.0088,
       "step": 5720
     },
     {
-      "epoch": 0.15712284749114278,
+      "epoch": 0.16234392735527808,
       "grad_norm": 0.0,
-      "learning_rate": 1.9164546626792047e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.909562564685824e-05,
+      "loss": 0.9846,
       "step": 5721
     },
     {
-      "epoch": 0.15715031171898602,
+      "epoch": 0.162372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.9164190660897106e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.9095243669917095e-05,
+      "loss": 0.9573,
       "step": 5722
     },
     {
-      "epoch": 0.15717777594682925,
+      "epoch": 0.16240068104426789,
       "grad_norm": 0.0,
-      "learning_rate": 1.916383462249154e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.9094861616148018e-05,
+      "loss": 1.2346,
       "step": 5723
     },
     {
-      "epoch": 0.1572052401746725,
+      "epoch": 0.16242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9163478511578163e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9094479485554235e-05,
+      "loss": 0.9609,
       "step": 5724
     },
     {
-      "epoch": 0.15723270440251572,
+      "epoch": 0.16245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9163122328159795e-05,
-      "loss": 1.0556,
+      "learning_rate": 1.9094097278138975e-05,
+      "loss": 0.8745,
       "step": 5725
     },
     {
-      "epoch": 0.15726016863035897,
+      "epoch": 0.16248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.9162766072239255e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.9093714993905467e-05,
+      "loss": 0.9808,
       "step": 5726
     },
     {
-      "epoch": 0.1572876328582022,
+      "epoch": 0.16251418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.9162409743819363e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.9093332632856936e-05,
+      "loss": 0.9641,
       "step": 5727
     },
     {
-      "epoch": 0.15731509708604544,
+      "epoch": 0.16254256526674235,
       "grad_norm": 0.0,
-      "learning_rate": 1.916205334290293e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.909295019499662e-05,
+      "loss": 0.9659,
       "step": 5728
     },
     {
-      "epoch": 0.15734256131388866,
+      "epoch": 0.16257094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9161696869492784e-05,
-      "loss": 1.0741,
+      "learning_rate": 1.909256768032774e-05,
+      "loss": 1.0023,
       "step": 5729
     },
     {
-      "epoch": 0.1573700255417319,
+      "epoch": 0.16259931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 1.9161340323591745e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.9092185088853535e-05,
+      "loss": 1.0073,
       "step": 5730
     },
     {
-      "epoch": 0.15739748976957513,
+      "epoch": 0.162627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9160983705202632e-05,
-      "loss": 1.0853,
+      "learning_rate": 1.9091802420577237e-05,
+      "loss": 0.9997,
       "step": 5731
     },
     {
-      "epoch": 0.15742495399741835,
+      "epoch": 0.1626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.916062701432827e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.909141967550207e-05,
+      "loss": 0.8861,
       "step": 5732
     },
     {
-      "epoch": 0.1574524182252616,
+      "epoch": 0.1626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.9160270250971473e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.909103685363128e-05,
+      "loss": 1.0823,
       "step": 5733
     },
     {
-      "epoch": 0.15747988245310482,
+      "epoch": 0.1627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.915991341513507e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.9090653954968084e-05,
+      "loss": 1.1039,
       "step": 5734
     },
     {
-      "epoch": 0.15750734668094807,
+      "epoch": 0.1627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.9159556506821888e-05,
-      "loss": 1.0231,
+      "learning_rate": 1.9090270979515728e-05,
+      "loss": 1.0264,
       "step": 5735
     },
     {
-      "epoch": 0.1575348109087913,
+      "epoch": 0.16276958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.9159199526034743e-05,
-      "loss": 1.0887,
+      "learning_rate": 1.9089887927277446e-05,
+      "loss": 1.0197,
       "step": 5736
     },
     {
-      "epoch": 0.15756227513663454,
+      "epoch": 0.16279795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 1.9158842472776466e-05,
-      "loss": 1.1121,
+      "learning_rate": 1.9089504798256472e-05,
+      "loss": 0.8589,
       "step": 5737
     },
     {
-      "epoch": 0.15758973936447776,
+      "epoch": 0.16282633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.915848534704988e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.9089121592456044e-05,
+      "loss": 0.9785,
       "step": 5738
     },
     {
-      "epoch": 0.157617203592321,
+      "epoch": 0.16285471055618614,
       "grad_norm": 0.0,
-      "learning_rate": 1.915812814885781e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.9088738309879394e-05,
+      "loss": 0.8833,
       "step": 5739
     },
     {
-      "epoch": 0.15764466782016423,
+      "epoch": 0.16288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.9157770878203078e-05,
-      "loss": 1.0532,
+      "learning_rate": 1.9088354950529764e-05,
+      "loss": 1.0081,
       "step": 5740
     },
     {
-      "epoch": 0.15767213204800748,
+      "epoch": 0.16291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.915741353508852e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.908797151441039e-05,
+      "loss": 1.073,
       "step": 5741
     },
     {
-      "epoch": 0.1576995962758507,
+      "epoch": 0.16293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.9157056119516958e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.9087588001524514e-05,
+      "loss": 0.9679,
       "step": 5742
     },
     {
-      "epoch": 0.15772706050369395,
+      "epoch": 0.16296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.915669863149122e-05,
-      "loss": 1.114,
+      "learning_rate": 1.9087204411875375e-05,
+      "loss": 0.9732,
       "step": 5743
     },
     {
-      "epoch": 0.15775452473153717,
+      "epoch": 0.1629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.9156341071014138e-05,
-      "loss": 1.0603,
+      "learning_rate": 1.908682074546621e-05,
+      "loss": 1.0574,
       "step": 5744
     },
     {
-      "epoch": 0.15778198895938042,
+      "epoch": 0.16302497162315552,
       "grad_norm": 0.0,
-      "learning_rate": 1.9155983438088538e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.9086437002300262e-05,
+      "loss": 0.9574,
       "step": 5745
     },
     {
-      "epoch": 0.15780945318722364,
+      "epoch": 0.1630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.9155625732717247e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.908605318238077e-05,
+      "loss": 0.945,
       "step": 5746
     },
     {
-      "epoch": 0.15783691741506686,
+      "epoch": 0.1630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.91552679549031e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.908566928571098e-05,
+      "loss": 1.0203,
       "step": 5747
     },
     {
-      "epoch": 0.1578643816429101,
+      "epoch": 0.16311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.915491010464893e-05,
-      "loss": 0.996,
+      "learning_rate": 1.9085285312294137e-05,
+      "loss": 1.1006,
       "step": 5748
     },
     {
-      "epoch": 0.15789184587075333,
+      "epoch": 0.16313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.915455218195756e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.9084901262133475e-05,
+      "loss": 0.9881,
       "step": 5749
     },
     {
-      "epoch": 0.15791931009859658,
+      "epoch": 0.16316685584562995,
       "grad_norm": 0.0,
-      "learning_rate": 1.915419418683183e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.9084517135232245e-05,
+      "loss": 1.0513,
       "step": 5750
     },
     {
-      "epoch": 0.1579467743264398,
+      "epoch": 0.16319523269012487,
       "grad_norm": 0.0,
-      "learning_rate": 1.915383611927457e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.908413293159369e-05,
+      "loss": 1.0079,
       "step": 5751
     },
     {
-      "epoch": 0.15797423855428305,
+      "epoch": 0.16322360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 1.915347797928861e-05,
-      "loss": 0.8563,
+      "learning_rate": 1.9083748651221057e-05,
+      "loss": 1.0126,
       "step": 5752
     },
     {
-      "epoch": 0.15800170278212627,
+      "epoch": 0.16325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.915311976687679e-05,
-      "loss": 1.1345,
+      "learning_rate": 1.908336429411759e-05,
+      "loss": 1.0098,
       "step": 5753
     },
     {
-      "epoch": 0.15802916700996952,
+      "epoch": 0.16328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9152761482041938e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.9082979860286535e-05,
+      "loss": 1.0681,
       "step": 5754
     },
     {
-      "epoch": 0.15805663123781274,
+      "epoch": 0.16330874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 1.915240312478689e-05,
-      "loss": 1.014,
+      "learning_rate": 1.9082595349731144e-05,
+      "loss": 0.9638,
       "step": 5755
     },
     {
-      "epoch": 0.158084095465656,
+      "epoch": 0.1633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.9152044695114487e-05,
-      "loss": 0.9344,
+      "learning_rate": 1.908221076245466e-05,
+      "loss": 1.0506,
       "step": 5756
     },
     {
-      "epoch": 0.15811155969349922,
+      "epoch": 0.16336549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 1.9151686193027563e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.9081826098460335e-05,
+      "loss": 1.0455,
       "step": 5757
     },
     {
-      "epoch": 0.15813902392134246,
+      "epoch": 0.1633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.9151327618528948e-05,
-      "loss": 1.1443,
+      "learning_rate": 1.9081441357751415e-05,
+      "loss": 0.9822,
       "step": 5758
     },
     {
-      "epoch": 0.15816648814918569,
+      "epoch": 0.163422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.9150968971621488e-05,
-      "loss": 1.1422,
+      "learning_rate": 1.9081056540331152e-05,
+      "loss": 0.9767,
       "step": 5759
     },
     {
-      "epoch": 0.1581939523770289,
+      "epoch": 0.16345062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 1.915061025230801e-05,
-      "loss": 0.963,
+      "learning_rate": 1.9080671646202797e-05,
+      "loss": 1.0299,
       "step": 5760
     },
     {
-      "epoch": 0.15822141660487216,
+      "epoch": 0.16347900113507377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9150251460591368e-05,
-      "loss": 1.066,
+      "learning_rate": 1.90802866753696e-05,
+      "loss": 1.0127,
       "step": 5761
     },
     {
-      "epoch": 0.15824888083271538,
+      "epoch": 0.16350737797956869,
       "grad_norm": 0.0,
-      "learning_rate": 1.9149892596474387e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.9079901627834813e-05,
+      "loss": 0.9851,
       "step": 5762
     },
     {
-      "epoch": 0.15827634506055863,
+      "epoch": 0.16353575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 1.914953365995991e-05,
-      "loss": 1.0867,
+      "learning_rate": 1.9079516503601687e-05,
+      "loss": 1.0888,
       "step": 5763
     },
     {
-      "epoch": 0.15830380928840185,
+      "epoch": 0.16356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.9149174651050782e-05,
-      "loss": 1.0375,
+      "learning_rate": 1.9079131302673478e-05,
+      "loss": 0.8767,
       "step": 5764
     },
     {
-      "epoch": 0.1583312735162451,
+      "epoch": 0.16359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.914881556974984e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.9078746025053442e-05,
+      "loss": 1.0595,
       "step": 5765
     },
     {
-      "epoch": 0.15835873774408832,
+      "epoch": 0.16362088535754823,
       "grad_norm": 0.0,
-      "learning_rate": 1.9148456416059925e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.9078360670744826e-05,
+      "loss": 1.0094,
       "step": 5766
     },
     {
-      "epoch": 0.15838620197193157,
+      "epoch": 0.16364926220204312,
       "grad_norm": 0.0,
-      "learning_rate": 1.9148097189983875e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.907797523975089e-05,
+      "loss": 0.8962,
       "step": 5767
     },
     {
-      "epoch": 0.1584136661997748,
+      "epoch": 0.16367763904653804,
       "grad_norm": 0.0,
-      "learning_rate": 1.914773789152454e-05,
-      "loss": 1.1195,
+      "learning_rate": 1.907758973207489e-05,
+      "loss": 1.0996,
       "step": 5768
     },
     {
-      "epoch": 0.15844113042761804,
+      "epoch": 0.16370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.914737852068476e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.9077204147720085e-05,
+      "loss": 1.1318,
       "step": 5769
     },
     {
-      "epoch": 0.15846859465546126,
+      "epoch": 0.1637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.9147019077467377e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.9076818486689725e-05,
+      "loss": 0.9662,
       "step": 5770
     },
     {
-      "epoch": 0.1584960588833045,
+      "epoch": 0.1637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.9146659561875235e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.907643274898707e-05,
+      "loss": 1.0413,
       "step": 5771
     },
     {
-      "epoch": 0.15852352311114773,
+      "epoch": 0.1637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.914629997391118e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.9076046934615382e-05,
+      "loss": 1.0273,
       "step": 5772
     },
     {
-      "epoch": 0.15855098733899095,
+      "epoch": 0.16381952326901247,
       "grad_norm": 0.0,
-      "learning_rate": 1.9145940313578057e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.9075661043577915e-05,
+      "loss": 1.0425,
       "step": 5773
     },
     {
-      "epoch": 0.1585784515668342,
+      "epoch": 0.1638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.9145580580878712e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.907527507587793e-05,
+      "loss": 1.0277,
       "step": 5774
     },
     {
-      "epoch": 0.15860591579467742,
+      "epoch": 0.16387627695800228,
       "grad_norm": 0.0,
-      "learning_rate": 1.9145220775815988e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.907488903151869e-05,
+      "loss": 1.0171,
       "step": 5775
     },
     {
-      "epoch": 0.15863338002252067,
+      "epoch": 0.16390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.9144860898392737e-05,
-      "loss": 1.0883,
+      "learning_rate": 1.907450291050346e-05,
+      "loss": 1.0803,
       "step": 5776
     },
     {
-      "epoch": 0.1586608442503639,
+      "epoch": 0.16393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.9144500948611804e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.907411671283549e-05,
+      "loss": 1.0982,
       "step": 5777
     },
     {
-      "epoch": 0.15868830847820714,
+      "epoch": 0.16396140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.9144140926476037e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.907373043851805e-05,
+      "loss": 1.0479,
       "step": 5778
     },
     {
-      "epoch": 0.15871577270605036,
+      "epoch": 0.16398978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 1.9143780831988287e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.90733440875544e-05,
+      "loss": 0.9821,
       "step": 5779
     },
     {
-      "epoch": 0.1587432369338936,
+      "epoch": 0.16401816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9143420665151397e-05,
-      "loss": 1.014,
+      "learning_rate": 1.9072957659947805e-05,
+      "loss": 0.9402,
       "step": 5780
     },
     {
-      "epoch": 0.15877070116173683,
+      "epoch": 0.16404653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 1.9143060425968224e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.907257115570153e-05,
+      "loss": 0.9699,
       "step": 5781
     },
     {
-      "epoch": 0.15879816538958008,
+      "epoch": 0.16407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.9142700114441614e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.9072184574818838e-05,
+      "loss": 0.9464,
       "step": 5782
     },
     {
-      "epoch": 0.1588256296174233,
+      "epoch": 0.1641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.9142339730574416e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.9071797917302995e-05,
+      "loss": 0.987,
       "step": 5783
     },
     {
-      "epoch": 0.15885309384526655,
+      "epoch": 0.1641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.914197927436949e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.907141118315727e-05,
+      "loss": 0.9969,
       "step": 5784
     },
     {
-      "epoch": 0.15888055807310977,
+      "epoch": 0.1641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.914161874582968e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.9071024372384923e-05,
+      "loss": 1.0508,
       "step": 5785
     },
     {
-      "epoch": 0.15890802230095302,
+      "epoch": 0.1641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.914125814495784e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.9070637484989224e-05,
+      "loss": 1.0612,
       "step": 5786
     },
     {
-      "epoch": 0.15893548652879624,
+      "epoch": 0.16421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.914089747175683e-05,
-      "loss": 1.0839,
+      "learning_rate": 1.9070250520973444e-05,
+      "loss": 1.0124,
       "step": 5787
     },
     {
-      "epoch": 0.15896295075663947,
+      "epoch": 0.16424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9140536726229494e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.9069863480340852e-05,
+      "loss": 1.0582,
       "step": 5788
     },
     {
-      "epoch": 0.15899041498448271,
+      "epoch": 0.16427355278093075,
       "grad_norm": 0.0,
-      "learning_rate": 1.914017590837869e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.9069476363094715e-05,
+      "loss": 0.9652,
       "step": 5789
     },
     {
-      "epoch": 0.15901787921232594,
+      "epoch": 0.16430192962542564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9139815018207275e-05,
-      "loss": 1.0914,
+      "learning_rate": 1.90690891692383e-05,
+      "loss": 1.0672,
       "step": 5790
     },
     {
-      "epoch": 0.15904534344016918,
+      "epoch": 0.16433030646992056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9139454055718105e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.9068701898774885e-05,
+      "loss": 0.998,
       "step": 5791
     },
     {
-      "epoch": 0.1590728076680124,
+      "epoch": 0.16435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.913909302091403e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.9068314551707738e-05,
+      "loss": 1.0033,
       "step": 5792
     },
     {
-      "epoch": 0.15910027189585565,
+      "epoch": 0.16438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9138731913797915e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.9067927128040127e-05,
+      "loss": 1.0796,
       "step": 5793
     },
     {
-      "epoch": 0.15912773612369888,
+      "epoch": 0.16441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.9138370734372613e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.906753962777533e-05,
+      "loss": 1.0453,
       "step": 5794
     },
     {
-      "epoch": 0.15915520035154213,
+      "epoch": 0.1644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.913800948264098e-05,
-      "loss": 1.0848,
+      "learning_rate": 1.906715205091662e-05,
+      "loss": 0.9471,
       "step": 5795
     },
     {
-      "epoch": 0.15918266457938535,
+      "epoch": 0.164472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.913764815860588e-05,
-      "loss": 1.1593,
+      "learning_rate": 1.9066764397467266e-05,
+      "loss": 1.1074,
       "step": 5796
     },
     {
-      "epoch": 0.1592101288072286,
+      "epoch": 0.1645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9137286762270163e-05,
-      "loss": 0.9698,
+      "learning_rate": 1.9066376667430546e-05,
+      "loss": 0.9034,
       "step": 5797
     },
     {
-      "epoch": 0.15923759303507182,
+      "epoch": 0.1645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.9136925293636695e-05,
-      "loss": 1.0569,
+      "learning_rate": 1.9065988860809735e-05,
+      "loss": 1.0042,
       "step": 5798
     },
     {
-      "epoch": 0.15926505726291507,
+      "epoch": 0.16455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.9136563752708335e-05,
-      "loss": 1.0401,
+      "learning_rate": 1.9065600977608107e-05,
+      "loss": 1.0441,
       "step": 5799
     },
     {
-      "epoch": 0.1592925214907583,
+      "epoch": 0.16458569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.9136202139487943e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.906521301782894e-05,
+      "loss": 1.1101,
       "step": 5800
     },
     {
-      "epoch": 0.1593199857186015,
+      "epoch": 0.16461407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.9135840453978384e-05,
-      "loss": 1.0862,
+      "learning_rate": 1.9064824981475514e-05,
+      "loss": 1.0155,
       "step": 5801
     },
     {
-      "epoch": 0.15934744994644476,
+      "epoch": 0.16464245175936437,
       "grad_norm": 0.0,
-      "learning_rate": 1.913547869618251e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.9064436868551103e-05,
+      "loss": 1.0239,
       "step": 5802
     },
     {
-      "epoch": 0.15937491417428798,
+      "epoch": 0.16467082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 1.9135116866103196e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.9064048679058986e-05,
+      "loss": 1.1002,
       "step": 5803
     },
     {
-      "epoch": 0.15940237840213123,
+      "epoch": 0.16469920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9134754963743296e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.9063660413002442e-05,
+      "loss": 1.0409,
       "step": 5804
     },
     {
-      "epoch": 0.15942984262997445,
+      "epoch": 0.16472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.9134392989105678e-05,
-      "loss": 1.049,
+      "learning_rate": 1.906327207038475e-05,
+      "loss": 0.9577,
       "step": 5805
     },
     {
-      "epoch": 0.1594573068578177,
+      "epoch": 0.16475595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 1.91340309421932e-05,
-      "loss": 1.1564,
+      "learning_rate": 1.9062883651209193e-05,
+      "loss": 0.9881,
       "step": 5806
     },
     {
-      "epoch": 0.15948477108566092,
+      "epoch": 0.1647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9133668823008737e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.9062495155479053e-05,
+      "loss": 1.0004,
       "step": 5807
     },
     {
-      "epoch": 0.15951223531350417,
+      "epoch": 0.16481271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 1.9133306631555143e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.9062106583197605e-05,
+      "loss": 1.0702,
       "step": 5808
     },
     {
-      "epoch": 0.1595396995413474,
+      "epoch": 0.1648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.913294436783529e-05,
-      "loss": 1.1187,
+      "learning_rate": 1.9061717934368137e-05,
+      "loss": 0.926,
       "step": 5809
     },
     {
-      "epoch": 0.15956716376919064,
+      "epoch": 0.1648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.913258203185205e-05,
-      "loss": 1.1111,
+      "learning_rate": 1.9061329208993927e-05,
+      "loss": 0.9864,
       "step": 5810
     },
     {
-      "epoch": 0.15959462799703386,
+      "epoch": 0.16489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.9132219623608273e-05,
-      "loss": 0.9656,
+      "learning_rate": 1.9060940407078266e-05,
+      "loss": 1.0761,
       "step": 5811
     },
     {
-      "epoch": 0.1596220922248771,
+      "epoch": 0.16492622020431327,
       "grad_norm": 0.0,
-      "learning_rate": 1.9131857143106845e-05,
-      "loss": 1.0472,
+      "learning_rate": 1.906055152862443e-05,
+      "loss": 1.1371,
       "step": 5812
     },
     {
-      "epoch": 0.15964955645272033,
+      "epoch": 0.16495459704880816,
       "grad_norm": 0.0,
-      "learning_rate": 1.9131494590350622e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.9060162573635713e-05,
+      "loss": 1.0498,
       "step": 5813
     },
     {
-      "epoch": 0.15967702068056355,
+      "epoch": 0.16498297389330308,
       "grad_norm": 0.0,
-      "learning_rate": 1.9131131965342476e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.9059773542115394e-05,
+      "loss": 1.0021,
       "step": 5814
     },
     {
-      "epoch": 0.1597044849084068,
+      "epoch": 0.16501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.913076926808528e-05,
-      "loss": 0.9925,
+      "learning_rate": 1.905938443406676e-05,
+      "loss": 1.0931,
       "step": 5815
     },
     {
-      "epoch": 0.15973194913625002,
+      "epoch": 0.16503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9130406498581898e-05,
-      "loss": 1.097,
+      "learning_rate": 1.90589952494931e-05,
+      "loss": 1.0201,
       "step": 5816
     },
     {
-      "epoch": 0.15975941336409327,
+      "epoch": 0.16506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.9130043656835205e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.9058605988397692e-05,
+      "loss": 1.0536,
       "step": 5817
     },
     {
-      "epoch": 0.1597868775919365,
+      "epoch": 0.16509648127128262,
       "grad_norm": 0.0,
-      "learning_rate": 1.9129680742848067e-05,
-      "loss": 0.909,
+      "learning_rate": 1.905821665078384e-05,
+      "loss": 0.9903,
       "step": 5818
     },
     {
-      "epoch": 0.15981434181977974,
+      "epoch": 0.1651248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.912931775662336e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.905782723665482e-05,
+      "loss": 0.997,
       "step": 5819
     },
     {
-      "epoch": 0.15984180604762296,
+      "epoch": 0.16515323496027243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9128954698163956e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.905743774601393e-05,
+      "loss": 0.9667,
       "step": 5820
     },
     {
-      "epoch": 0.1598692702754662,
+      "epoch": 0.16518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.9128591567472725e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.9057048178864456e-05,
+      "loss": 1.128,
       "step": 5821
     },
     {
-      "epoch": 0.15989673450330943,
+      "epoch": 0.1652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.9128228364552543e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.905665853520969e-05,
+      "loss": 1.1763,
       "step": 5822
     },
     {
-      "epoch": 0.15992419873115268,
+      "epoch": 0.1652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.912786508940628e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.905626881505292e-05,
+      "loss": 1.1056,
       "step": 5823
     },
     {
-      "epoch": 0.1599516629589959,
+      "epoch": 0.16526674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 1.9127501742036812e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.9055879018397438e-05,
+      "loss": 0.9867,
       "step": 5824
     },
     {
-      "epoch": 0.15997912718683915,
+      "epoch": 0.1652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.9127138322447018e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.905548914524654e-05,
+      "loss": 1.0419,
       "step": 5825
     },
     {
-      "epoch": 0.16000659141468238,
+      "epoch": 0.16532349602724178,
       "grad_norm": 0.0,
-      "learning_rate": 1.9126774830639766e-05,
-      "loss": 0.996,
+      "learning_rate": 1.905509919560352e-05,
+      "loss": 0.9425,
       "step": 5826
     },
     {
-      "epoch": 0.1600340556425256,
+      "epoch": 0.16535187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.912641126661794e-05,
-      "loss": 1.1247,
+      "learning_rate": 1.905470916947167e-05,
+      "loss": 1.0196,
       "step": 5827
     },
     {
-      "epoch": 0.16006151987036885,
+      "epoch": 0.16538024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.912604763038441e-05,
-      "loss": 1.036,
+      "learning_rate": 1.9054319066854285e-05,
+      "loss": 1.1003,
       "step": 5828
     },
     {
-      "epoch": 0.16008898409821207,
+      "epoch": 0.16540862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 1.9125683921942057e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.9053928887754656e-05,
+      "loss": 0.9435,
       "step": 5829
     },
     {
-      "epoch": 0.16011644832605532,
+      "epoch": 0.16543700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.912532014129376e-05,
-      "loss": 1.1985,
+      "learning_rate": 1.9053538632176088e-05,
+      "loss": 1.0128,
       "step": 5830
     },
     {
-      "epoch": 0.16014391255389854,
+      "epoch": 0.16546538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.912495628844239e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.905314830012187e-05,
+      "loss": 0.9286,
       "step": 5831
     },
     {
-      "epoch": 0.1601713767817418,
+      "epoch": 0.16549375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 1.9124592363390834e-05,
-      "loss": 1.007,
+      "learning_rate": 1.90527578915953e-05,
+      "loss": 1.0312,
       "step": 5832
     },
     {
-      "epoch": 0.160198841009585,
+      "epoch": 0.16552213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.912422836614197e-05,
-      "loss": 1.1132,
+      "learning_rate": 1.9052367406599678e-05,
+      "loss": 1.1654,
       "step": 5833
     },
     {
-      "epoch": 0.16022630523742826,
+      "epoch": 0.1655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.9123864296698677e-05,
-      "loss": 1.1214,
+      "learning_rate": 1.90519768451383e-05,
+      "loss": 1.0048,
       "step": 5834
     },
     {
-      "epoch": 0.16025376946527148,
+      "epoch": 0.1655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9123500155063836e-05,
-      "loss": 1.037,
+      "learning_rate": 1.9051586207214468e-05,
+      "loss": 1.0502,
       "step": 5835
     },
     {
-      "epoch": 0.16028123369311473,
+      "epoch": 0.16560726447219068,
       "grad_norm": 0.0,
-      "learning_rate": 1.9123135941240323e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.9051195492831485e-05,
+      "loss": 0.9796,
       "step": 5836
     },
     {
-      "epoch": 0.16030869792095795,
+      "epoch": 0.1656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.9122771655231027e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.905080470199264e-05,
+      "loss": 0.9851,
       "step": 5837
     },
     {
-      "epoch": 0.1603361621488012,
+      "epoch": 0.16566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.912240729703883e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.9050413834701244e-05,
+      "loss": 0.9369,
       "step": 5838
     },
     {
-      "epoch": 0.16036362637664442,
+      "epoch": 0.16569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9122042866666608e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.9050022890960597e-05,
+      "loss": 1.103,
       "step": 5839
     },
     {
-      "epoch": 0.16039109060448767,
+      "epoch": 0.16572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.9121678364117255e-05,
-      "loss": 1.0466,
+      "learning_rate": 1.9049631870773996e-05,
+      "loss": 1.0114,
       "step": 5840
     },
     {
-      "epoch": 0.1604185548323309,
+      "epoch": 0.16574914869466514,
       "grad_norm": 0.0,
-      "learning_rate": 1.9121313789393644e-05,
-      "loss": 1.0119,
+      "learning_rate": 1.904924077414475e-05,
+      "loss": 1.0319,
       "step": 5841
     },
     {
-      "epoch": 0.1604460190601741,
+      "epoch": 0.16577752553916006,
       "grad_norm": 0.0,
-      "learning_rate": 1.912094914249867e-05,
-      "loss": 1.0191,
+      "learning_rate": 1.904884960107616e-05,
+      "loss": 0.9857,
       "step": 5842
     },
     {
-      "epoch": 0.16047348328801736,
+      "epoch": 0.16580590238365495,
       "grad_norm": 0.0,
-      "learning_rate": 1.912058442343521e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.904845835157153e-05,
+      "loss": 1.0626,
       "step": 5843
     },
     {
-      "epoch": 0.16050094751586058,
+      "epoch": 0.16583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.9120219632206153e-05,
-      "loss": 1.0922,
+      "learning_rate": 1.9048067025634166e-05,
+      "loss": 0.9158,
       "step": 5844
     },
     {
-      "epoch": 0.16052841174370383,
+      "epoch": 0.16586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9119854768814388e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.9047675623267374e-05,
+      "loss": 1.0538,
       "step": 5845
     },
     {
-      "epoch": 0.16055587597154705,
+      "epoch": 0.1658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.9119489833262796e-05,
-      "loss": 0.9942,
+      "learning_rate": 1.904728414447446e-05,
+      "loss": 1.0393,
       "step": 5846
     },
     {
-      "epoch": 0.1605833401993903,
+      "epoch": 0.1659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.911912482555427e-05,
-      "loss": 0.8902,
+      "learning_rate": 1.9046892589258728e-05,
+      "loss": 1.1502,
       "step": 5847
     },
     {
-      "epoch": 0.16061080442723352,
+      "epoch": 0.1659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.9118759745691698e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.9046500957623485e-05,
+      "loss": 0.946,
       "step": 5848
     },
     {
-      "epoch": 0.16063826865507677,
+      "epoch": 0.1659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.9118394593677963e-05,
-      "loss": 1.118,
+      "learning_rate": 1.9046109249572046e-05,
+      "loss": 1.0047,
       "step": 5849
     },
     {
-      "epoch": 0.16066573288292,
+      "epoch": 0.16600454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 1.9118029369515956e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.9045717465107715e-05,
+      "loss": 1.0874,
       "step": 5850
     },
     {
-      "epoch": 0.16069319711076324,
+      "epoch": 0.16603291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.9117664073208576e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.90453256042338e-05,
+      "loss": 1.101,
       "step": 5851
     },
     {
-      "epoch": 0.16072066133860646,
+      "epoch": 0.16606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.91172987047587e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.9044933666953615e-05,
+      "loss": 1.1093,
       "step": 5852
     },
     {
-      "epoch": 0.1607481255664497,
+      "epoch": 0.16608967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 1.9116933264169227e-05,
-      "loss": 0.973,
+      "learning_rate": 1.9044541653270467e-05,
+      "loss": 1.0895,
       "step": 5853
     },
     {
-      "epoch": 0.16077558979429293,
+      "epoch": 0.16611804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.9116567751443045e-05,
-      "loss": 1.086,
+      "learning_rate": 1.904414956318767e-05,
+      "loss": 0.8993,
       "step": 5854
     },
     {
-      "epoch": 0.16080305402213615,
+      "epoch": 0.16614642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.911620216658305e-05,
-      "loss": 1.1246,
+      "learning_rate": 1.9043757396708536e-05,
+      "loss": 1.0568,
       "step": 5855
     },
     {
-      "epoch": 0.1608305182499794,
+      "epoch": 0.16617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.911583650959213e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.9043365153836378e-05,
+      "loss": 0.9394,
       "step": 5856
     },
     {
-      "epoch": 0.16085798247782263,
+      "epoch": 0.16620317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 1.911547078047318e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.9042972834574506e-05,
+      "loss": 1.0204,
       "step": 5857
     },
     {
-      "epoch": 0.16088544670566587,
+      "epoch": 0.1662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.91151049792291e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.9042580438926234e-05,
+      "loss": 1.0286,
       "step": 5858
     },
     {
-      "epoch": 0.1609129109335091,
+      "epoch": 0.1662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9114739105862773e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.9042187966894882e-05,
+      "loss": 1.0114,
       "step": 5859
     },
     {
-      "epoch": 0.16094037516135234,
+      "epoch": 0.1662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.9114373160377097e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.904179541848376e-05,
+      "loss": 1.0518,
       "step": 5860
     },
     {
-      "epoch": 0.16096783938919557,
+      "epoch": 0.166316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.9114007142774975e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.9041402793696186e-05,
+      "loss": 1.0491,
       "step": 5861
     },
     {
-      "epoch": 0.16099530361703882,
+      "epoch": 0.1663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.9113641053059297e-05,
-      "loss": 1.0896,
+      "learning_rate": 1.9041010092535477e-05,
+      "loss": 0.9589,
       "step": 5862
     },
     {
-      "epoch": 0.16102276784488204,
+      "epoch": 0.16637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.911327489123296e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.9040617315004945e-05,
+      "loss": 1.0249,
       "step": 5863
     },
     {
-      "epoch": 0.16105023207272529,
+      "epoch": 0.16640181611804766,
       "grad_norm": 0.0,
-      "learning_rate": 1.9112908657298865e-05,
-      "loss": 1.0177,
+      "learning_rate": 1.9040224461107918e-05,
+      "loss": 0.9607,
       "step": 5864
     },
     {
-      "epoch": 0.1610776963005685,
+      "epoch": 0.16643019296254258,
       "grad_norm": 0.0,
-      "learning_rate": 1.9112542351259903e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9039831530847706e-05,
+      "loss": 0.9552,
       "step": 5865
     },
     {
-      "epoch": 0.16110516052841176,
+      "epoch": 0.16645856980703747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9112175973118977e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.903943852422763e-05,
+      "loss": 1.0603,
       "step": 5866
     },
     {
-      "epoch": 0.16113262475625498,
+      "epoch": 0.16648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.9111809522878985e-05,
-      "loss": 1.0047,
+      "learning_rate": 1.903904544125101e-05,
+      "loss": 1.0385,
       "step": 5867
     },
     {
-      "epoch": 0.1611600889840982,
+      "epoch": 0.16651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9111443000542825e-05,
-      "loss": 1.0577,
+      "learning_rate": 1.903865228192117e-05,
+      "loss": 1.0562,
       "step": 5868
     },
     {
-      "epoch": 0.16118755321194145,
+      "epoch": 0.16654370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.91110764061134e-05,
-      "loss": 1.0394,
+      "learning_rate": 1.9038259046241424e-05,
+      "loss": 0.9934,
       "step": 5869
     },
     {
-      "epoch": 0.16121501743978467,
+      "epoch": 0.16657207718501701,
       "grad_norm": 0.0,
-      "learning_rate": 1.911070973959361e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.90378657342151e-05,
+      "loss": 0.9786,
       "step": 5870
     },
     {
-      "epoch": 0.16124248166762792,
+      "epoch": 0.16660045402951193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9110343000986353e-05,
-      "loss": 1.1059,
+      "learning_rate": 1.903747234584552e-05,
+      "loss": 1.0144,
       "step": 5871
     },
     {
-      "epoch": 0.16126994589547114,
+      "epoch": 0.16662883087400682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9109976190294533e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.9037078881136e-05,
+      "loss": 1.1326,
       "step": 5872
     },
     {
-      "epoch": 0.1612974101233144,
+      "epoch": 0.1666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.9109609307521057e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.903668534008987e-05,
+      "loss": 0.8886,
       "step": 5873
     },
     {
-      "epoch": 0.1613248743511576,
+      "epoch": 0.1666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.910924235266882e-05,
-      "loss": 0.99,
+      "learning_rate": 1.9036291722710454e-05,
+      "loss": 1.1085,
       "step": 5874
     },
     {
-      "epoch": 0.16135233857900086,
+      "epoch": 0.16671396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 1.9108875325740727e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.9035898029001076e-05,
+      "loss": 1.0781,
       "step": 5875
     },
     {
-      "epoch": 0.16137980280684408,
+      "epoch": 0.16674233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.9108508226739692e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.903550425896506e-05,
+      "loss": 0.9869,
       "step": 5876
     },
     {
-      "epoch": 0.16140726703468733,
+      "epoch": 0.16677071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 1.9108141055668604e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.903511041260573e-05,
+      "loss": 0.9988,
       "step": 5877
     },
     {
-      "epoch": 0.16143473126253055,
+      "epoch": 0.16679909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9107773812530378e-05,
-      "loss": 1.0531,
+      "learning_rate": 1.903471648992642e-05,
+      "loss": 0.9881,
       "step": 5878
     },
     {
-      "epoch": 0.1614621954903738,
+      "epoch": 0.16682746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.910740649732792e-05,
-      "loss": 1.0991,
+      "learning_rate": 1.9034322490930456e-05,
+      "loss": 0.9726,
       "step": 5879
     },
     {
-      "epoch": 0.16148965971821702,
+      "epoch": 0.16685584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.910703911006413e-05,
-      "loss": 1.013,
+      "learning_rate": 1.903392841562116e-05,
+      "loss": 1.0274,
       "step": 5880
     },
     {
-      "epoch": 0.16151712394606027,
+      "epoch": 0.16688422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 1.910667165074192e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.9033534264001867e-05,
+      "loss": 1.0284,
       "step": 5881
     },
     {
-      "epoch": 0.1615445881739035,
+      "epoch": 0.16691259931895575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9106304119364196e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.90331400360759e-05,
+      "loss": 0.9948,
       "step": 5882
     },
     {
-      "epoch": 0.1615720524017467,
+      "epoch": 0.16694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 1.9105936515933865e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.9032745731846598e-05,
+      "loss": 1.0823,
       "step": 5883
     },
     {
-      "epoch": 0.16159951662958996,
+      "epoch": 0.16696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.9105568840453837e-05,
-      "loss": 1.0368,
+      "learning_rate": 1.9032351351317283e-05,
+      "loss": 0.9273,
       "step": 5884
     },
     {
-      "epoch": 0.16162698085743318,
+      "epoch": 0.1669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.9105201092927027e-05,
-      "loss": 1.081,
+      "learning_rate": 1.903195689449129e-05,
+      "loss": 1.0632,
       "step": 5885
     },
     {
-      "epoch": 0.16165444508527643,
+      "epoch": 0.1670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9104833273356334e-05,
-      "loss": 1.1735,
+      "learning_rate": 1.9031562361371955e-05,
+      "loss": 1.0034,
       "step": 5886
     },
     {
-      "epoch": 0.16168190931311965,
+      "epoch": 0.16705448354143018,
       "grad_norm": 0.0,
-      "learning_rate": 1.9104465381744675e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.9031167751962604e-05,
+      "loss": 0.9451,
       "step": 5887
     },
     {
-      "epoch": 0.1617093735409629,
+      "epoch": 0.1670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.910409741809496e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.9030773066266573e-05,
+      "loss": 0.9405,
       "step": 5888
     },
     {
-      "epoch": 0.16173683776880612,
+      "epoch": 0.16711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.9103729382410098e-05,
-      "loss": 1.0589,
+      "learning_rate": 1.9030378304287198e-05,
+      "loss": 0.9706,
       "step": 5889
     },
     {
-      "epoch": 0.16176430199664937,
+      "epoch": 0.16713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9103361274693002e-05,
-      "loss": 1.079,
+      "learning_rate": 1.902998346602781e-05,
+      "loss": 1.0038,
       "step": 5890
     },
     {
-      "epoch": 0.1617917662244926,
+      "epoch": 0.16716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9102993094946583e-05,
-      "loss": 0.9503,
+      "learning_rate": 1.9029588551491743e-05,
+      "loss": 0.9496,
       "step": 5891
     },
     {
-      "epoch": 0.16181923045233584,
+      "epoch": 0.16719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.910262484317376e-05,
-      "loss": 1.0945,
+      "learning_rate": 1.902919356068234e-05,
+      "loss": 0.9821,
       "step": 5892
     },
     {
-      "epoch": 0.16184669468017907,
+      "epoch": 0.16722474460839953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9102256519377443e-05,
-      "loss": 1.106,
+      "learning_rate": 1.902879849360293e-05,
+      "loss": 0.9883,
       "step": 5893
     },
     {
-      "epoch": 0.16187415890802231,
+      "epoch": 0.16725312145289445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9101888123560546e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.9028403350256855e-05,
+      "loss": 1.1369,
       "step": 5894
     },
     {
-      "epoch": 0.16190162313586554,
+      "epoch": 0.16728149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.9101519655725988e-05,
-      "loss": 1.0971,
+      "learning_rate": 1.902800813064745e-05,
+      "loss": 1.082,
       "step": 5895
     },
     {
-      "epoch": 0.16192908736370876,
+      "epoch": 0.16730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.9101151115876674e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.9027612834778058e-05,
+      "loss": 0.9657,
       "step": 5896
     },
     {
-      "epoch": 0.161956551591552,
+      "epoch": 0.1673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.9100782504015533e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.902721746265201e-05,
+      "loss": 1.0138,
       "step": 5897
     },
     {
-      "epoch": 0.16198401581939523,
+      "epoch": 0.167366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9100413820145475e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.9026822014272653e-05,
+      "loss": 0.9352,
       "step": 5898
     },
     {
-      "epoch": 0.16201148004723848,
+      "epoch": 0.16739500567536889,
       "grad_norm": 0.0,
-      "learning_rate": 1.910004506426941e-05,
-      "loss": 0.957,
+      "learning_rate": 1.9026426489643325e-05,
+      "loss": 0.9442,
       "step": 5899
     },
     {
-      "epoch": 0.1620389442750817,
+      "epoch": 0.1674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.909967623639027e-05,
-      "loss": 1.0708,
+      "learning_rate": 1.9026030888767364e-05,
+      "loss": 1.0444,
       "step": 5900
     },
     {
-      "epoch": 0.16206640850292495,
+      "epoch": 0.1674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9099307336510968e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.902563521164811e-05,
+      "loss": 1.0089,
       "step": 5901
     },
     {
-      "epoch": 0.16209387273076817,
+      "epoch": 0.16748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.909893836463442e-05,
-      "loss": 1.1168,
+      "learning_rate": 1.9025239458288916e-05,
+      "loss": 1.0294,
       "step": 5902
     },
     {
-      "epoch": 0.16212133695861142,
+      "epoch": 0.16750851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 1.9098569320763546e-05,
-      "loss": 0.8955,
+      "learning_rate": 1.902484362869312e-05,
+      "loss": 0.9901,
       "step": 5903
     },
     {
-      "epoch": 0.16214880118645464,
+      "epoch": 0.16753688989784335,
       "grad_norm": 0.0,
-      "learning_rate": 1.909820020490127e-05,
-      "loss": 1.0936,
+      "learning_rate": 1.9024447722864062e-05,
+      "loss": 1.0133,
       "step": 5904
     },
     {
-      "epoch": 0.1621762654142979,
+      "epoch": 0.16756526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.9097831017050506e-05,
-      "loss": 1.048,
+      "learning_rate": 1.902405174080509e-05,
+      "loss": 1.1036,
       "step": 5905
     },
     {
-      "epoch": 0.1622037296421411,
+      "epoch": 0.16759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.9097461757214178e-05,
-      "loss": 1.039,
+      "learning_rate": 1.9023655682519548e-05,
+      "loss": 1.0472,
       "step": 5906
     },
     {
-      "epoch": 0.16223119386998436,
+      "epoch": 0.16762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9097092425395215e-05,
-      "loss": 1.072,
+      "learning_rate": 1.9023259548010777e-05,
+      "loss": 1.0041,
       "step": 5907
     },
     {
-      "epoch": 0.16225865809782758,
+      "epoch": 0.16765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9096723021596527e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.9022863337282125e-05,
+      "loss": 0.9723,
       "step": 5908
     },
     {
-      "epoch": 0.1622861223256708,
+      "epoch": 0.16767877412031781,
       "grad_norm": 0.0,
-      "learning_rate": 1.9096353545821045e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.9022467050336945e-05,
+      "loss": 1.0122,
       "step": 5909
     },
     {
-      "epoch": 0.16231358655351405,
+      "epoch": 0.1677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.9095983998071686e-05,
-      "loss": 1.027,
+      "learning_rate": 1.902207068717858e-05,
+      "loss": 1.0241,
       "step": 5910
     },
     {
-      "epoch": 0.16234105078135727,
+      "epoch": 0.16773552780930762,
       "grad_norm": 0.0,
-      "learning_rate": 1.9095614378351385e-05,
-      "loss": 1.1152,
+      "learning_rate": 1.902167424781038e-05,
+      "loss": 0.9487,
       "step": 5911
     },
     {
-      "epoch": 0.16236851500920052,
+      "epoch": 0.1677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.9095244686663058e-05,
-      "loss": 1.1649,
+      "learning_rate": 1.9021277732235687e-05,
+      "loss": 1.0908,
       "step": 5912
     },
     {
-      "epoch": 0.16239597923704374,
+      "epoch": 0.1677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.909487492300963e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.9020881140457857e-05,
+      "loss": 0.9404,
       "step": 5913
     },
     {
-      "epoch": 0.162423443464887,
+      "epoch": 0.16782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.9094505087394026e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.902048447248024e-05,
+      "loss": 1.0182,
       "step": 5914
     },
     {
-      "epoch": 0.1624509076927302,
+      "epoch": 0.16784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.909413517981918e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.9020087728306188e-05,
+      "loss": 0.9866,
       "step": 5915
     },
     {
-      "epoch": 0.16247837192057346,
+      "epoch": 0.16787741203178205,
       "grad_norm": 0.0,
-      "learning_rate": 1.909376520028801e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.9019690907939047e-05,
+      "loss": 1.0641,
       "step": 5916
     },
     {
-      "epoch": 0.16250583614841668,
+      "epoch": 0.16790578887627697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9093395148803447e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.901929401138217e-05,
+      "loss": 1.0016,
       "step": 5917
     },
     {
-      "epoch": 0.16253330037625993,
+      "epoch": 0.16793416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9093025025368423e-05,
-      "loss": 1.1389,
+      "learning_rate": 1.901889703863891e-05,
+      "loss": 0.9825,
       "step": 5918
     },
     {
-      "epoch": 0.16256076460410315,
+      "epoch": 0.16796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.909265482998586e-05,
-      "loss": 1.0074,
+      "learning_rate": 1.9018499989712627e-05,
+      "loss": 0.9775,
       "step": 5919
     },
     {
-      "epoch": 0.1625882288319464,
+      "epoch": 0.16799091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.909228456265869e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.9018102864606664e-05,
+      "loss": 0.9767,
       "step": 5920
     },
     {
-      "epoch": 0.16261569305978962,
+      "epoch": 0.16801929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9091914223389843e-05,
-      "loss": 1.067,
+      "learning_rate": 1.9017705663324382e-05,
+      "loss": 0.9908,
       "step": 5921
     },
     {
-      "epoch": 0.16264315728763284,
+      "epoch": 0.16804767309875143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9091543812182247e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.9017308385869136e-05,
+      "loss": 1.0011,
       "step": 5922
     },
     {
-      "epoch": 0.1626706215154761,
+      "epoch": 0.16807604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 1.909117332903884e-05,
-      "loss": 1.06,
+      "learning_rate": 1.901691103224428e-05,
+      "loss": 1.0515,
       "step": 5923
     },
     {
-      "epoch": 0.16269808574331932,
+      "epoch": 0.1681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.909080277396254e-05,
-      "loss": 1.074,
+      "learning_rate": 1.9016513602453172e-05,
+      "loss": 1.1144,
       "step": 5924
     },
     {
-      "epoch": 0.16272554997116256,
+      "epoch": 0.1681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9090432146956293e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.9016116096499163e-05,
+      "loss": 0.9018,
       "step": 5925
     },
     {
-      "epoch": 0.16275301419900579,
+      "epoch": 0.16816118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.9090061448023026e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.901571851438562e-05,
+      "loss": 0.9913,
       "step": 5926
     },
     {
-      "epoch": 0.16278047842684903,
+      "epoch": 0.16818955732122587,
       "grad_norm": 0.0,
-      "learning_rate": 1.9089690677165667e-05,
-      "loss": 1.0694,
+      "learning_rate": 1.90153208561159e-05,
+      "loss": 0.9883,
       "step": 5927
     },
     {
-      "epoch": 0.16280794265469226,
+      "epoch": 0.16821793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.9089319834387158e-05,
-      "loss": 1.0587,
+      "learning_rate": 1.9014923121693355e-05,
+      "loss": 1.104,
       "step": 5928
     },
     {
-      "epoch": 0.1628354068825355,
+      "epoch": 0.16824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.9088948919690428e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.901452531112135e-05,
+      "loss": 1.0536,
       "step": 5929
     },
     {
-      "epoch": 0.16286287111037873,
+      "epoch": 0.16827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9088577933078412e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.9014127424403248e-05,
+      "loss": 1.0265,
       "step": 5930
     },
     {
-      "epoch": 0.16289033533822198,
+      "epoch": 0.16830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.908820687455405e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.90137294615424e-05,
+      "loss": 0.9207,
       "step": 5931
     },
     {
-      "epoch": 0.1629177995660652,
+      "epoch": 0.16833144154370033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9087835744120277e-05,
-      "loss": 0.92,
+      "learning_rate": 1.9013331422542184e-05,
+      "loss": 0.9147,
       "step": 5932
     },
     {
-      "epoch": 0.16294526379390845,
+      "epoch": 0.16835981838819522,
       "grad_norm": 0.0,
-      "learning_rate": 1.908746454178002e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.9012933307405945e-05,
+      "loss": 1.0406,
       "step": 5933
     },
     {
-      "epoch": 0.16297272802175167,
+      "epoch": 0.16838819523269014,
       "grad_norm": 0.0,
-      "learning_rate": 1.9087093267536227e-05,
-      "loss": 1.1685,
+      "learning_rate": 1.9012535116137054e-05,
+      "loss": 1.0515,
       "step": 5934
     },
     {
-      "epoch": 0.16300019224959492,
+      "epoch": 0.16841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.9086721921391834e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.9012136848738874e-05,
+      "loss": 1.0546,
       "step": 5935
     },
     {
-      "epoch": 0.16302765647743814,
+      "epoch": 0.1684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.9086350503349777e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.901173850521477e-05,
+      "loss": 1.0135,
       "step": 5936
     },
     {
-      "epoch": 0.16305512070528136,
+      "epoch": 0.1684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.9085979013412993e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.9011340085568103e-05,
+      "loss": 1.0218,
       "step": 5937
     },
     {
-      "epoch": 0.1630825849331246,
+      "epoch": 0.16850170261066968,
       "grad_norm": 0.0,
-      "learning_rate": 1.9085607451584423e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.9010941589802245e-05,
+      "loss": 1.0179,
       "step": 5938
     },
     {
-      "epoch": 0.16311004916096783,
+      "epoch": 0.16853007945516457,
       "grad_norm": 0.0,
-      "learning_rate": 1.9085235817867008e-05,
-      "loss": 0.998,
+      "learning_rate": 1.901054301792056e-05,
+      "loss": 1.048,
       "step": 5939
     },
     {
-      "epoch": 0.16313751338881108,
+      "epoch": 0.1685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.908486411226369e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.9010144369926407e-05,
+      "loss": 1.0368,
       "step": 5940
     },
     {
-      "epoch": 0.1631649776166543,
+      "epoch": 0.16858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.9084492334777404e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.9009745645823162e-05,
+      "loss": 0.8713,
       "step": 5941
     },
     {
-      "epoch": 0.16319244184449755,
+      "epoch": 0.16861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.90841204854111e-05,
-      "loss": 1.0791,
+      "learning_rate": 1.900934684561419e-05,
+      "loss": 0.9048,
       "step": 5942
     },
     {
-      "epoch": 0.16321990607234077,
+      "epoch": 0.16864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.9083748564167714e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.900894796930286e-05,
+      "loss": 0.9547,
       "step": 5943
     },
     {
-      "epoch": 0.16324737030018402,
+      "epoch": 0.16867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.908337657105019e-05,
-      "loss": 1.1089,
+      "learning_rate": 1.9008549016892544e-05,
+      "loss": 1.0149,
       "step": 5944
     },
     {
-      "epoch": 0.16327483452802724,
+      "epoch": 0.16870034052213395,
       "grad_norm": 0.0,
-      "learning_rate": 1.9083004506061472e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.9008149988386608e-05,
+      "loss": 0.9929,
       "step": 5945
     },
     {
-      "epoch": 0.1633022987558705,
+      "epoch": 0.16872871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.9082632369204504e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.9007750883788422e-05,
+      "loss": 1.1247,
       "step": 5946
     },
     {
-      "epoch": 0.1633297629837137,
+      "epoch": 0.16875709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9082260160482233e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.900735170310136e-05,
+      "loss": 0.9759,
       "step": 5947
     },
     {
-      "epoch": 0.16335722721155696,
+      "epoch": 0.1687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9081887879897597e-05,
-      "loss": 0.994,
+      "learning_rate": 1.9006952446328797e-05,
+      "loss": 1.0185,
       "step": 5948
     },
     {
-      "epoch": 0.16338469143940018,
+      "epoch": 0.1688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.908151552745355e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.90065531134741e-05,
+      "loss": 0.9539,
       "step": 5949
     },
     {
-      "epoch": 0.1634121556672434,
+      "epoch": 0.1688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.908114310315303e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.9006153704540643e-05,
+      "loss": 1.0099,
       "step": 5950
     },
     {
-      "epoch": 0.16343961989508665,
+      "epoch": 0.1688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.9080770606998996e-05,
-      "loss": 1.077,
+      "learning_rate": 1.90057542195318e-05,
+      "loss": 1.0758,
       "step": 5951
     },
     {
-      "epoch": 0.16346708412292987,
+      "epoch": 0.1688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.908039803899438e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.9005354658450948e-05,
+      "loss": 0.9981,
       "step": 5952
     },
     {
-      "epoch": 0.16349454835077312,
+      "epoch": 0.16892735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.908002539914214e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.9004955021301458e-05,
+      "loss": 0.9448,
       "step": 5953
     },
     {
-      "epoch": 0.16352201257861634,
+      "epoch": 0.16895573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 1.9079652687445223e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.900455530808671e-05,
+      "loss": 1.0793,
       "step": 5954
     },
     {
-      "epoch": 0.1635494768064596,
+      "epoch": 0.16898410896708285,
       "grad_norm": 0.0,
-      "learning_rate": 1.9079279903906573e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.900415551881008e-05,
+      "loss": 1.0434,
       "step": 5955
     },
     {
-      "epoch": 0.16357694103430281,
+      "epoch": 0.16901248581157774,
       "grad_norm": 0.0,
-      "learning_rate": 1.9078907048529144e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.900375565347494e-05,
+      "loss": 1.0352,
       "step": 5956
     },
     {
-      "epoch": 0.16360440526214606,
+      "epoch": 0.16904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.9078534121315887e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.9003355712084673e-05,
+      "loss": 0.9385,
       "step": 5957
     },
     {
-      "epoch": 0.16363186948998928,
+      "epoch": 0.16906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.9078161122269755e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.900295569464266e-05,
+      "loss": 0.9962,
       "step": 5958
     },
     {
-      "epoch": 0.16365933371783253,
+      "epoch": 0.16909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.907778805139369e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.900255560115227e-05,
+      "loss": 0.9182,
       "step": 5959
     },
     {
-      "epoch": 0.16368679794567575,
+      "epoch": 0.16912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9077414908690654e-05,
-      "loss": 1.0207,
+      "learning_rate": 1.900215543161689e-05,
+      "loss": 1.0722,
       "step": 5960
     },
     {
-      "epoch": 0.163714262173519,
+      "epoch": 0.1691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.907704169416359e-05,
-      "loss": 1.1345,
+      "learning_rate": 1.9001755186039896e-05,
+      "loss": 0.9645,
       "step": 5961
     },
     {
-      "epoch": 0.16374172640136223,
+      "epoch": 0.16918274687854712,
       "grad_norm": 0.0,
-      "learning_rate": 1.9076668407815463e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.9001354864424673e-05,
+      "loss": 1.0315,
       "step": 5962
     },
     {
-      "epoch": 0.16376919062920545,
+      "epoch": 0.169211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.9076295049649214e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.9000954466774603e-05,
+      "loss": 1.0486,
       "step": 5963
     },
     {
-      "epoch": 0.1637966548570487,
+      "epoch": 0.1692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.9075921619667802e-05,
-      "loss": 1.0702,
+      "learning_rate": 1.9000553993093063e-05,
+      "loss": 0.9088,
       "step": 5964
     },
     {
-      "epoch": 0.16382411908489192,
+      "epoch": 0.16926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.9075548117874184e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.900015344338344e-05,
+      "loss": 1.007,
       "step": 5965
     },
     {
-      "epoch": 0.16385158331273517,
+      "epoch": 0.16929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.9075174544271316e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.8999752817649116e-05,
+      "loss": 1.1231,
       "step": 5966
     },
     {
-      "epoch": 0.1638790475405784,
+      "epoch": 0.16932463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.907480089886215e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.8999352115893476e-05,
+      "loss": 1.0846,
       "step": 5967
     },
     {
-      "epoch": 0.16390651176842164,
+      "epoch": 0.16935300794551647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9074427181649646e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.8998951338119904e-05,
+      "loss": 1.0361,
       "step": 5968
     },
     {
-      "epoch": 0.16393397599626486,
+      "epoch": 0.16938138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9074053392636755e-05,
-      "loss": 1.045,
+      "learning_rate": 1.8998550484331787e-05,
+      "loss": 0.9924,
       "step": 5969
     },
     {
-      "epoch": 0.1639614402241081,
+      "epoch": 0.16940976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.907367953182644e-05,
-      "loss": 1.1144,
+      "learning_rate": 1.8998149554532505e-05,
+      "loss": 1.0581,
       "step": 5970
     },
     {
-      "epoch": 0.16398890445195133,
+      "epoch": 0.16943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.907330559922166e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.899774854872545e-05,
+      "loss": 0.992,
       "step": 5971
     },
     {
-      "epoch": 0.16401636867979458,
+      "epoch": 0.16946651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 1.907293159482537e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.8997347466914014e-05,
+      "loss": 1.1344,
       "step": 5972
     },
     {
-      "epoch": 0.1640438329076378,
+      "epoch": 0.1694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.907255751864053e-05,
-      "loss": 1.0523,
+      "learning_rate": 1.8996946309101574e-05,
+      "loss": 1.0264,
       "step": 5973
     },
     {
-      "epoch": 0.16407129713548105,
+      "epoch": 0.16952326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 1.9072183370670102e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.8996545075291525e-05,
+      "loss": 0.9631,
       "step": 5974
     },
     {
-      "epoch": 0.16409876136332427,
+      "epoch": 0.1695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.9071809150917043e-05,
-      "loss": 1.022,
+      "learning_rate": 1.899614376548726e-05,
+      "loss": 0.9991,
       "step": 5975
     },
     {
-      "epoch": 0.1641262255911675,
+      "epoch": 0.1695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9071434859384315e-05,
-      "loss": 1.0552,
+      "learning_rate": 1.8995742379692158e-05,
+      "loss": 1.0496,
       "step": 5976
     },
     {
-      "epoch": 0.16415368981901074,
+      "epoch": 0.16960839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 1.9071060496074886e-05,
-      "loss": 1.0826,
+      "learning_rate": 1.899534091790962e-05,
+      "loss": 1.0117,
       "step": 5977
     },
     {
-      "epoch": 0.16418115404685396,
+      "epoch": 0.16963677639046537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9070686060991707e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.899493938014303e-05,
+      "loss": 1.0491,
       "step": 5978
     },
     {
-      "epoch": 0.1642086182746972,
+      "epoch": 0.16966515323496026,
       "grad_norm": 0.0,
-      "learning_rate": 1.907031155413775e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.8994537766395784e-05,
+      "loss": 0.9967,
       "step": 5979
     },
     {
-      "epoch": 0.16423608250254043,
+      "epoch": 0.16969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.906993697551597e-05,
-      "loss": 0.9786,
+      "learning_rate": 1.899413607667127e-05,
+      "loss": 1.0491,
       "step": 5980
     },
     {
-      "epoch": 0.16426354673038368,
+      "epoch": 0.16972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.9069562325129342e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.899373431097289e-05,
+      "loss": 0.8124,
       "step": 5981
     },
     {
-      "epoch": 0.1642910109582269,
+      "epoch": 0.16975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.9069187602980815e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.899333246930403e-05,
+      "loss": 1.0354,
       "step": 5982
     },
     {
-      "epoch": 0.16431847518607015,
+      "epoch": 0.16977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.9068812809073368e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.8992930551668084e-05,
+      "loss": 0.95,
       "step": 5983
     },
     {
-      "epoch": 0.16434593941391337,
+      "epoch": 0.16980703745743472,
       "grad_norm": 0.0,
-      "learning_rate": 1.9068437943409962e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.899252855806845e-05,
+      "loss": 0.9635,
       "step": 5984
     },
     {
-      "epoch": 0.16437340364175662,
+      "epoch": 0.16983541430192964,
       "grad_norm": 0.0,
-      "learning_rate": 1.906806300599356e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.8992126488508527e-05,
+      "loss": 0.9964,
       "step": 5985
     },
     {
-      "epoch": 0.16440086786959984,
+      "epoch": 0.16986379114642453,
       "grad_norm": 0.0,
-      "learning_rate": 1.906768799682713e-05,
-      "loss": 1.1035,
+      "learning_rate": 1.8991724342991703e-05,
+      "loss": 1.0557,
       "step": 5986
     },
     {
-      "epoch": 0.1644283320974431,
+      "epoch": 0.1698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9067312915913642e-05,
-      "loss": 1.1009,
+      "learning_rate": 1.899132212152138e-05,
+      "loss": 0.9405,
       "step": 5987
     },
     {
-      "epoch": 0.1644557963252863,
+      "epoch": 0.1699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9066937763256063e-05,
-      "loss": 1.0996,
+      "learning_rate": 1.8990919824100957e-05,
+      "loss": 0.9391,
       "step": 5988
     },
     {
-      "epoch": 0.16448326055312956,
+      "epoch": 0.1699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.906656253885736e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.8990517450733828e-05,
+      "loss": 1.0823,
       "step": 5989
     },
     {
-      "epoch": 0.16451072478097278,
+      "epoch": 0.16997729852440407,
       "grad_norm": 0.0,
-      "learning_rate": 1.90661872427205e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.8990115001423396e-05,
+      "loss": 1.0067,
       "step": 5990
     },
     {
-      "epoch": 0.164538189008816,
+      "epoch": 0.170005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.9065811874848455e-05,
-      "loss": 1.0166,
+      "learning_rate": 1.8989712476173054e-05,
+      "loss": 1.085,
       "step": 5991
     },
     {
-      "epoch": 0.16456565323665925,
+      "epoch": 0.17003405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9065436435244194e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.898930987498621e-05,
+      "loss": 1.0416,
       "step": 5992
     },
     {
-      "epoch": 0.16459311746450248,
+      "epoch": 0.17006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.906506092391069e-05,
-      "loss": 0.9166,
+      "learning_rate": 1.8988907197866262e-05,
+      "loss": 0.9643,
       "step": 5993
     },
     {
-      "epoch": 0.16462058169234572,
+      "epoch": 0.17009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.9064685340850913e-05,
-      "loss": 1.1465,
+      "learning_rate": 1.898850444481661e-05,
+      "loss": 0.9039,
       "step": 5994
     },
     {
-      "epoch": 0.16464804592018895,
+      "epoch": 0.17011918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.906430968606783e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.8988101615840656e-05,
+      "loss": 0.9786,
       "step": 5995
     },
     {
-      "epoch": 0.1646755101480322,
+      "epoch": 0.17014755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.9063933959564424e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.8987698710941802e-05,
+      "loss": 1.0805,
       "step": 5996
     },
     {
-      "epoch": 0.16470297437587542,
+      "epoch": 0.17017593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.906355816134366e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.8987295730123457e-05,
+      "loss": 1.0943,
       "step": 5997
     },
     {
-      "epoch": 0.16473043860371867,
+      "epoch": 0.17020431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 1.906318229140851e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.898689267338902e-05,
+      "loss": 1.1125,
       "step": 5998
     },
     {
-      "epoch": 0.1647579028315619,
+      "epoch": 0.17023269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.906280634976195e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.8986489540741895e-05,
+      "loss": 1.0146,
       "step": 5999
     },
     {
-      "epoch": 0.16478536705940514,
+      "epoch": 0.170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.906243033640696e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.8986086332185488e-05,
+      "loss": 1.0696,
       "step": 6000
     },
     {
-      "epoch": 0.16481283128724836,
+      "epoch": 0.1702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9062054251346506e-05,
-      "loss": 0.9686,
+      "learning_rate": 1.898568304772321e-05,
+      "loss": 0.9616,
       "step": 6001
     },
     {
-      "epoch": 0.1648402955150916,
+      "epoch": 0.1703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.9061678094583572e-05,
-      "loss": 1.0479,
+      "learning_rate": 1.898527968735846e-05,
+      "loss": 1.0242,
       "step": 6002
     },
     {
-      "epoch": 0.16486775974293483,
+      "epoch": 0.1703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9061301866121132e-05,
-      "loss": 0.995,
+      "learning_rate": 1.898487625109465e-05,
+      "loss": 1.0523,
       "step": 6003
     },
     {
-      "epoch": 0.16489522397077805,
+      "epoch": 0.17037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.9060925565962158e-05,
-      "loss": 1.0577,
+      "learning_rate": 1.8984472738935183e-05,
+      "loss": 1.0043,
       "step": 6004
     },
     {
-      "epoch": 0.1649226881986213,
+      "epoch": 0.17040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9060549194109633e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.8984069150883475e-05,
+      "loss": 1.0152,
       "step": 6005
     },
     {
-      "epoch": 0.16495015242646452,
+      "epoch": 0.17043132803632235,
       "grad_norm": 0.0,
-      "learning_rate": 1.906017275056653e-05,
-      "loss": 1.1005,
+      "learning_rate": 1.898366548694293e-05,
+      "loss": 1.0394,
       "step": 6006
     },
     {
-      "epoch": 0.16497761665430777,
+      "epoch": 0.17045970488081724,
       "grad_norm": 0.0,
-      "learning_rate": 1.9059796235335835e-05,
-      "loss": 1.0481,
+      "learning_rate": 1.898326174711696e-05,
+      "loss": 1.0821,
       "step": 6007
     },
     {
-      "epoch": 0.165005080882151,
+      "epoch": 0.17048808172531216,
       "grad_norm": 0.0,
-      "learning_rate": 1.905941964842052e-05,
-      "loss": 0.8976,
+      "learning_rate": 1.8982857931408972e-05,
+      "loss": 1.1725,
       "step": 6008
     },
     {
-      "epoch": 0.16503254510999424,
+      "epoch": 0.17051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.905904298982357e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.8982454039822377e-05,
+      "loss": 1.0322,
       "step": 6009
     },
     {
-      "epoch": 0.16506000933783746,
+      "epoch": 0.17054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.905866625954796e-05,
-      "loss": 1.0482,
+      "learning_rate": 1.8982050072360594e-05,
+      "loss": 0.9412,
       "step": 6010
     },
     {
-      "epoch": 0.1650874735656807,
+      "epoch": 0.17057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.905828945759667e-05,
-      "loss": 1.144,
+      "learning_rate": 1.8981646029027027e-05,
+      "loss": 0.9984,
       "step": 6011
     },
     {
-      "epoch": 0.16511493779352393,
+      "epoch": 0.1706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.905791258397269e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.898124190982509e-05,
+      "loss": 1.0714,
       "step": 6012
     },
     {
-      "epoch": 0.16514240202136718,
+      "epoch": 0.1706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.9057535638678998e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.8980837714758206e-05,
+      "loss": 1.0334,
       "step": 6013
     },
     {
-      "epoch": 0.1651698662492104,
+      "epoch": 0.1706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9057158621718574e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.8980433443829776e-05,
+      "loss": 1.0396,
       "step": 6014
     },
     {
-      "epoch": 0.16519733047705365,
+      "epoch": 0.1706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9056781533094404e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.8980029097043227e-05,
+      "loss": 0.9555,
       "step": 6015
     },
     {
-      "epoch": 0.16522479470489687,
+      "epoch": 0.17071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.905640437280947e-05,
-      "loss": 1.1124,
+      "learning_rate": 1.897962467440196e-05,
+      "loss": 1.0235,
       "step": 6016
     },
     {
-      "epoch": 0.1652522589327401,
+      "epoch": 0.17074347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9056027140866752e-05,
-      "loss": 1.1495,
+      "learning_rate": 1.8979220175909408e-05,
+      "loss": 1.0656,
       "step": 6017
     },
     {
-      "epoch": 0.16527972316058334,
+      "epoch": 0.17077185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.9055649837269244e-05,
-      "loss": 1.0911,
+      "learning_rate": 1.8978815601568975e-05,
+      "loss": 0.997,
       "step": 6018
     },
     {
-      "epoch": 0.16530718738842656,
+      "epoch": 0.17080022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.9055272462019925e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.8978410951384084e-05,
+      "loss": 1.0907,
       "step": 6019
     },
     {
-      "epoch": 0.1653346516162698,
+      "epoch": 0.17082860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 1.9054895015121784e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.8978006225358154e-05,
+      "loss": 1.0828,
       "step": 6020
     },
     {
-      "epoch": 0.16536211584411303,
+      "epoch": 0.17085698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9054517496577804e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.89776014234946e-05,
+      "loss": 1.022,
       "step": 6021
     },
     {
-      "epoch": 0.16538958007195628,
+      "epoch": 0.17088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.9054139906390978e-05,
-      "loss": 1.0994,
+      "learning_rate": 1.8977196545796844e-05,
+      "loss": 1.0431,
       "step": 6022
     },
     {
-      "epoch": 0.1654170442997995,
+      "epoch": 0.17091373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 1.9053762244564283e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.8976791592268303e-05,
+      "loss": 1.0537,
       "step": 6023
     },
     {
-      "epoch": 0.16544450852764275,
+      "epoch": 0.1709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.905338451110072e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.89763865629124e-05,
+      "loss": 1.0791,
       "step": 6024
     },
     {
-      "epoch": 0.16547197275548597,
+      "epoch": 0.17097048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 1.905300670600327e-05,
-      "loss": 1.049,
+      "learning_rate": 1.8975981457732557e-05,
+      "loss": 0.9958,
       "step": 6025
     },
     {
-      "epoch": 0.16549943698332922,
+      "epoch": 0.1709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.9052628829274923e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.8975576276732196e-05,
+      "loss": 0.9428,
       "step": 6026
     },
     {
-      "epoch": 0.16552690121117244,
+      "epoch": 0.1710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.905225088091867e-05,
-      "loss": 1.0709,
+      "learning_rate": 1.897517101991474e-05,
+      "loss": 1.1288,
       "step": 6027
     },
     {
-      "epoch": 0.1655543654390157,
+      "epoch": 0.17105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9051872860937504e-05,
-      "loss": 0.985,
+      "learning_rate": 1.897476568728361e-05,
+      "loss": 1.0599,
       "step": 6028
     },
     {
-      "epoch": 0.16558182966685892,
+      "epoch": 0.17108399545970487,
       "grad_norm": 0.0,
-      "learning_rate": 1.905149476933441e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.8974360278842226e-05,
+      "loss": 1.1563,
       "step": 6029
     },
     {
-      "epoch": 0.16560929389470216,
+      "epoch": 0.17111237230419976,
       "grad_norm": 0.0,
-      "learning_rate": 1.9051116606112388e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.8973954794594024e-05,
+      "loss": 1.0239,
       "step": 6030
     },
     {
-      "epoch": 0.16563675812254539,
+      "epoch": 0.17114074914869468,
       "grad_norm": 0.0,
-      "learning_rate": 1.9050738371274423e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.8973549234542418e-05,
+      "loss": 0.9961,
       "step": 6031
     },
     {
-      "epoch": 0.1656642223503886,
+      "epoch": 0.17116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.9050360064823506e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.8973143598690843e-05,
+      "loss": 1.0029,
       "step": 6032
     },
     {
-      "epoch": 0.16569168657823186,
+      "epoch": 0.17119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9049981686762642e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.8972737887042715e-05,
+      "loss": 1.0745,
       "step": 6033
     },
     {
-      "epoch": 0.16571915080607508,
+      "epoch": 0.17122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.904960323709481e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.897233209960147e-05,
+      "loss": 1.0445,
       "step": 6034
     },
     {
-      "epoch": 0.16574661503391833,
+      "epoch": 0.17125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.9049224715823017e-05,
-      "loss": 1.0694,
+      "learning_rate": 1.8971926236370532e-05,
+      "loss": 1.0122,
       "step": 6035
     },
     {
-      "epoch": 0.16577407926176155,
+      "epoch": 0.1712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.904884612295025e-05,
-      "loss": 1.0527,
+      "learning_rate": 1.8971520297353326e-05,
+      "loss": 1.097,
       "step": 6036
     },
     {
-      "epoch": 0.1658015434896048,
+      "epoch": 0.17131101021566403,
       "grad_norm": 0.0,
-      "learning_rate": 1.9048467458479505e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.897111428255329e-05,
+      "loss": 1.0404,
       "step": 6037
     },
     {
-      "epoch": 0.16582900771744802,
+      "epoch": 0.17133938706015892,
       "grad_norm": 0.0,
-      "learning_rate": 1.9048088722413786e-05,
-      "loss": 1.1996,
+      "learning_rate": 1.8970708191973847e-05,
+      "loss": 1.1627,
       "step": 6038
     },
     {
-      "epoch": 0.16585647194529127,
+      "epoch": 0.1713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.904770991475608e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.897030202561843e-05,
+      "loss": 1.0227,
       "step": 6039
     },
     {
-      "epoch": 0.1658839361731345,
+      "epoch": 0.1713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.9047331035509394e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.8969895783490466e-05,
+      "loss": 0.9823,
       "step": 6040
     },
     {
-      "epoch": 0.16591140040097774,
+      "epoch": 0.17142451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 1.9046952084676716e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.8969489465593393e-05,
+      "loss": 0.9386,
       "step": 6041
     },
     {
-      "epoch": 0.16593886462882096,
+      "epoch": 0.1714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.9046573062261046e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.8969083071930635e-05,
+      "loss": 1.0158,
       "step": 6042
     },
     {
-      "epoch": 0.1659663288566642,
+      "epoch": 0.17148127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 1.904619396826539e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.8968676602505626e-05,
+      "loss": 0.988,
       "step": 6043
     },
     {
-      "epoch": 0.16599379308450743,
+      "epoch": 0.17150964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.9045814802692742e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.896827005732181e-05,
+      "loss": 0.9766,
       "step": 6044
     },
     {
-      "epoch": 0.16602125731235065,
+      "epoch": 0.17153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.9045435565546102e-05,
-      "loss": 1.0582,
+      "learning_rate": 1.896786343638261e-05,
+      "loss": 1.0317,
       "step": 6045
     },
     {
-      "epoch": 0.1660487215401939,
+      "epoch": 0.17156640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.9045056256828475e-05,
-      "loss": 0.8449,
+      "learning_rate": 1.8967456739691464e-05,
+      "loss": 0.9296,
       "step": 6046
     },
     {
-      "epoch": 0.16607618576803712,
+      "epoch": 0.17159477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 1.9044676876542855e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.896704996725181e-05,
+      "loss": 0.9877,
       "step": 6047
     },
     {
-      "epoch": 0.16610364999588037,
+      "epoch": 0.17162315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 1.904429742469225e-05,
-      "loss": 1.054,
+      "learning_rate": 1.896664311906708e-05,
+      "loss": 0.9673,
       "step": 6048
     },
     {
-      "epoch": 0.1661311142237236,
+      "epoch": 0.17165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.904391790127966e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.8966236195140714e-05,
+      "loss": 1.1117,
       "step": 6049
     },
     {
-      "epoch": 0.16615857845156684,
+      "epoch": 0.17167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.904353830630809e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.8965829195476146e-05,
+      "loss": 0.9757,
       "step": 6050
     },
     {
-      "epoch": 0.16618604267941006,
+      "epoch": 0.1717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.904315863978054e-05,
-      "loss": 1.017,
+      "learning_rate": 1.8965422120076814e-05,
+      "loss": 0.9824,
       "step": 6051
     },
     {
-      "epoch": 0.1662135069072533,
+      "epoch": 0.1717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.9042778901700016e-05,
-      "loss": 1.0927,
+      "learning_rate": 1.8965014968946166e-05,
+      "loss": 0.9655,
       "step": 6052
     },
     {
-      "epoch": 0.16624097113509653,
+      "epoch": 0.17176503972758228,
       "grad_norm": 0.0,
-      "learning_rate": 1.904239909206952e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.8964607742087626e-05,
+      "loss": 0.8744,
       "step": 6053
     },
     {
-      "epoch": 0.16626843536293978,
+      "epoch": 0.1717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.9042019210892065e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.8964200439504644e-05,
+      "loss": 1.0587,
       "step": 6054
     },
     {
-      "epoch": 0.166295899590783,
+      "epoch": 0.17182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.904163925817065e-05,
-      "loss": 1.0998,
+      "learning_rate": 1.896379306120066e-05,
+      "loss": 1.054,
       "step": 6055
     },
     {
-      "epoch": 0.16632336381862625,
+      "epoch": 0.17185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.9041259233908282e-05,
-      "loss": 0.954,
+      "learning_rate": 1.896338560717911e-05,
+      "loss": 1.0566,
       "step": 6056
     },
     {
-      "epoch": 0.16635082804646947,
+      "epoch": 0.17187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.9040879138107965e-05,
-      "loss": 1.0372,
+      "learning_rate": 1.896297807744344e-05,
+      "loss": 1.0709,
       "step": 6057
     },
     {
-      "epoch": 0.1663782922743127,
+      "epoch": 0.17190692395005674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9040498970772717e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.896257047199709e-05,
+      "loss": 0.9067,
       "step": 6058
     },
     {
-      "epoch": 0.16640575650215594,
+      "epoch": 0.17193530079455163,
       "grad_norm": 0.0,
-      "learning_rate": 1.9040118731905534e-05,
-      "loss": 0.9404,
+      "learning_rate": 1.8962162790843504e-05,
+      "loss": 0.9309,
       "step": 6059
     },
     {
-      "epoch": 0.16643322072999917,
+      "epoch": 0.17196367763904655,
       "grad_norm": 0.0,
-      "learning_rate": 1.903973842150943e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.896175503398613e-05,
+      "loss": 1.0152,
       "step": 6060
     },
     {
-      "epoch": 0.16646068495784241,
+      "epoch": 0.17199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.9039358039587417e-05,
-      "loss": 1.0291,
+      "learning_rate": 1.8961347201428405e-05,
+      "loss": 0.9874,
       "step": 6061
     },
     {
-      "epoch": 0.16648814918568564,
+      "epoch": 0.17202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.90389775861425e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.8960939293173777e-05,
+      "loss": 0.9601,
       "step": 6062
     },
     {
-      "epoch": 0.16651561341352888,
+      "epoch": 0.1720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.903859706117769e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.8960531309225695e-05,
+      "loss": 1.0641,
       "step": 6063
     },
     {
-      "epoch": 0.1665430776413721,
+      "epoch": 0.1720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.9038216464696e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.89601232495876e-05,
+      "loss": 0.9454,
       "step": 6064
     },
     {
-      "epoch": 0.16657054186921535,
+      "epoch": 0.172105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.9037835796700442e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.895971511426294e-05,
+      "loss": 0.9394,
       "step": 6065
     },
     {
-      "epoch": 0.16659800609705858,
+      "epoch": 0.1721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.903745505719402e-05,
-      "loss": 1.1124,
+      "learning_rate": 1.8959306903255162e-05,
+      "loss": 0.9216,
       "step": 6066
     },
     {
-      "epoch": 0.16662547032490183,
+      "epoch": 0.17216231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 1.9037074246179762e-05,
-      "loss": 1.0617,
+      "learning_rate": 1.8958898616567724e-05,
+      "loss": 1.0641,
       "step": 6067
     },
     {
-      "epoch": 0.16665293455274505,
+      "epoch": 0.17219069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.9036693363660668e-05,
-      "loss": 1.0884,
+      "learning_rate": 1.895849025420406e-05,
+      "loss": 1.1009,
       "step": 6068
     },
     {
-      "epoch": 0.1666803987805883,
+      "epoch": 0.17221906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.9036312409639755e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.8958081816167628e-05,
+      "loss": 1.0573,
       "step": 6069
     },
     {
-      "epoch": 0.16670786300843152,
+      "epoch": 0.17224744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9035931384120036e-05,
-      "loss": 1.1341,
+      "learning_rate": 1.8957673302461875e-05,
+      "loss": 1.0599,
       "step": 6070
     },
     {
-      "epoch": 0.16673532723627474,
+      "epoch": 0.17227582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.9035550287104532e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.8957264713090256e-05,
+      "loss": 1.1403,
       "step": 6071
     },
     {
-      "epoch": 0.166762791464118,
+      "epoch": 0.17230419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 1.903516911859625e-05,
-      "loss": 1.0625,
+      "learning_rate": 1.8956856048056215e-05,
+      "loss": 0.9539,
       "step": 6072
     },
     {
-      "epoch": 0.1667902556919612,
+      "epoch": 0.17233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.903478787859821e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.895644730736321e-05,
+      "loss": 0.9782,
       "step": 6073
     },
     {
-      "epoch": 0.16681771991980446,
+      "epoch": 0.17236095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 1.903440656711343e-05,
-      "loss": 1.0329,
+      "learning_rate": 1.8956038491014693e-05,
+      "loss": 1.0813,
       "step": 6074
     },
     {
-      "epoch": 0.16684518414764768,
+      "epoch": 0.1723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.9034025184144925e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.8955629599014116e-05,
+      "loss": 1.0124,
       "step": 6075
     },
     {
-      "epoch": 0.16687264837549093,
+      "epoch": 0.1724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.9033643729695717e-05,
-      "loss": 1.088,
+      "learning_rate": 1.895522063136493e-05,
+      "loss": 1.0657,
       "step": 6076
     },
     {
-      "epoch": 0.16690011260333415,
+      "epoch": 0.17244608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 1.9033262203768818e-05,
-      "loss": 1.1022,
+      "learning_rate": 1.8954811588070598e-05,
+      "loss": 1.1488,
       "step": 6077
     },
     {
-      "epoch": 0.1669275768311774,
+      "epoch": 0.1724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.903288060636725e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.8954402469134566e-05,
+      "loss": 1.1279,
       "step": 6078
     },
     {
-      "epoch": 0.16695504105902062,
+      "epoch": 0.1725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.903249893749403e-05,
-      "loss": 1.0959,
+      "learning_rate": 1.8953993274560298e-05,
+      "loss": 1.0081,
       "step": 6079
     },
     {
-      "epoch": 0.16698250528686387,
+      "epoch": 0.17253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.903211719715218e-05,
-      "loss": 1.0472,
+      "learning_rate": 1.8953584004351243e-05,
+      "loss": 0.9474,
       "step": 6080
     },
     {
-      "epoch": 0.1670099695147071,
+      "epoch": 0.17255959137343926,
       "grad_norm": 0.0,
-      "learning_rate": 1.903173538534472e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.8953174658510863e-05,
+      "loss": 1.0272,
       "step": 6081
     },
     {
-      "epoch": 0.16703743374255034,
+      "epoch": 0.17258796821793418,
       "grad_norm": 0.0,
-      "learning_rate": 1.903135350207467e-05,
-      "loss": 0.973,
+      "learning_rate": 1.895276523704261e-05,
+      "loss": 1.0277,
       "step": 6082
     },
     {
-      "epoch": 0.16706489797039356,
+      "epoch": 0.17261634506242907,
       "grad_norm": 0.0,
-      "learning_rate": 1.9030971547345056e-05,
-      "loss": 1.0818,
+      "learning_rate": 1.8952355739949953e-05,
+      "loss": 0.9905,
       "step": 6083
     },
     {
-      "epoch": 0.1670923621982368,
+      "epoch": 0.17264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.903058952115889e-05,
-      "loss": 1.0878,
+      "learning_rate": 1.8951946167236342e-05,
+      "loss": 1.051,
       "step": 6084
     },
     {
-      "epoch": 0.16711982642608003,
+      "epoch": 0.17267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.903020742351921e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.895153651890524e-05,
+      "loss": 0.9937,
       "step": 6085
     },
     {
-      "epoch": 0.16714729065392325,
+      "epoch": 0.17270147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.9029825254429026e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.8951126794960103e-05,
+      "loss": 1.0082,
       "step": 6086
     },
     {
-      "epoch": 0.1671747548817665,
+      "epoch": 0.17272985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 1.9029443013891366e-05,
-      "loss": 0.9068,
+      "learning_rate": 1.8950716995404397e-05,
+      "loss": 0.837,
       "step": 6087
     },
     {
-      "epoch": 0.16720221910960972,
+      "epoch": 0.17275822928490353,
       "grad_norm": 0.0,
-      "learning_rate": 1.902906070190926e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.8950307120241585e-05,
+      "loss": 0.9932,
       "step": 6088
     },
     {
-      "epoch": 0.16722968333745297,
+      "epoch": 0.17278660612939842,
       "grad_norm": 0.0,
-      "learning_rate": 1.9028678318485726e-05,
-      "loss": 0.8692,
+      "learning_rate": 1.8949897169475126e-05,
+      "loss": 1.0025,
       "step": 6089
     },
     {
-      "epoch": 0.1672571475652962,
+      "epoch": 0.1728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.902829586362379e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.894948714310848e-05,
+      "loss": 0.9431,
       "step": 6090
     },
     {
-      "epoch": 0.16728461179313944,
+      "epoch": 0.1728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.902791333732648e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.8949077041145115e-05,
+      "loss": 1.0247,
       "step": 6091
     },
     {
-      "epoch": 0.16731207602098266,
+      "epoch": 0.17287173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 1.9027530739596826e-05,
-      "loss": 1.113,
+      "learning_rate": 1.8948666863588495e-05,
+      "loss": 1.0205,
       "step": 6092
     },
     {
-      "epoch": 0.1673395402488259,
+      "epoch": 0.17290011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 1.9027148070437848e-05,
-      "loss": 1.0984,
+      "learning_rate": 1.8948256610442085e-05,
+      "loss": 0.9798,
       "step": 6093
     },
     {
-      "epoch": 0.16736700447666913,
+      "epoch": 0.17292849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 1.902676532985258e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.8947846281709346e-05,
+      "loss": 0.9106,
       "step": 6094
     },
     {
-      "epoch": 0.16739446870451238,
+      "epoch": 0.17295686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 1.9026382517844047e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.894743587739375e-05,
+      "loss": 1.036,
       "step": 6095
     },
     {
-      "epoch": 0.1674219329323556,
+      "epoch": 0.17298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.902599963441528e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.8947025397498757e-05,
+      "loss": 0.9513,
       "step": 6096
     },
     {
-      "epoch": 0.16744939716019885,
+      "epoch": 0.17301362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.9025616679569306e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.8946614842027844e-05,
+      "loss": 0.8803,
       "step": 6097
     },
     {
-      "epoch": 0.16747686138804208,
+      "epoch": 0.17304199772985243,
       "grad_norm": 0.0,
-      "learning_rate": 1.902523365330916e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.8946204210984468e-05,
+      "loss": 1.0052,
       "step": 6098
     },
     {
-      "epoch": 0.1675043256158853,
+      "epoch": 0.17307037457434732,
       "grad_norm": 0.0,
-      "learning_rate": 1.9024850555637863e-05,
-      "loss": 1.0298,
+      "learning_rate": 1.8945793504372106e-05,
+      "loss": 0.8879,
       "step": 6099
     },
     {
-      "epoch": 0.16753178984372855,
+      "epoch": 0.17309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.9024467386558458e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.8945382722194224e-05,
+      "loss": 1.0776,
       "step": 6100
     },
     {
-      "epoch": 0.16755925407157177,
+      "epoch": 0.17312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.902408414607397e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.894497186445429e-05,
+      "loss": 0.9666,
       "step": 6101
     },
     {
-      "epoch": 0.16758671829941502,
+      "epoch": 0.173155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.902370083418743e-05,
-      "loss": 1.0254,
+      "learning_rate": 1.894456093115578e-05,
+      "loss": 0.9585,
       "step": 6102
     },
     {
-      "epoch": 0.16761418252725824,
+      "epoch": 0.1731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.9023317450901877e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.8944149922302156e-05,
+      "loss": 1.0223,
       "step": 6103
     },
     {
-      "epoch": 0.1676416467551015,
+      "epoch": 0.17321225879682178,
       "grad_norm": 0.0,
-      "learning_rate": 1.902293399622034e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.89437388378969e-05,
+      "loss": 1.0061,
       "step": 6104
     },
     {
-      "epoch": 0.1676691109829447,
+      "epoch": 0.1732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.9022550470145854e-05,
-      "loss": 1.1334,
+      "learning_rate": 1.8943327677943483e-05,
+      "loss": 1.0687,
       "step": 6105
     },
     {
-      "epoch": 0.16769657521078796,
+      "epoch": 0.17326901248581159,
       "grad_norm": 0.0,
-      "learning_rate": 1.9022166872681453e-05,
-      "loss": 1.0378,
+      "learning_rate": 1.8942916442445368e-05,
+      "loss": 0.9596,
       "step": 6106
     },
     {
-      "epoch": 0.16772403943863118,
+      "epoch": 0.17329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9021783203830173e-05,
-      "loss": 1.0053,
+      "learning_rate": 1.894250513140604e-05,
+      "loss": 1.0085,
       "step": 6107
     },
     {
-      "epoch": 0.16775150366647443,
+      "epoch": 0.17332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.902139946359505e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.894209374482897e-05,
+      "loss": 0.8933,
       "step": 6108
     },
     {
-      "epoch": 0.16777896789431765,
+      "epoch": 0.17335414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.902101565197912e-05,
-      "loss": 1.1313,
+      "learning_rate": 1.8941682282717625e-05,
+      "loss": 1.0448,
       "step": 6109
     },
     {
-      "epoch": 0.1678064321221609,
+      "epoch": 0.17338251986379113,
       "grad_norm": 0.0,
-      "learning_rate": 1.902063176898542e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.89412707450755e-05,
+      "loss": 0.9727,
       "step": 6110
     },
     {
-      "epoch": 0.16783389635000412,
+      "epoch": 0.17341089670828605,
       "grad_norm": 0.0,
-      "learning_rate": 1.902024781461699e-05,
-      "loss": 1.0062,
+      "learning_rate": 1.8940859131906047e-05,
+      "loss": 0.9091,
       "step": 6111
     },
     {
-      "epoch": 0.16786136057784734,
+      "epoch": 0.17343927355278094,
       "grad_norm": 0.0,
-      "learning_rate": 1.9019863788876857e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.894044744321276e-05,
+      "loss": 1.0644,
       "step": 6112
     },
     {
-      "epoch": 0.1678888248056906,
+      "epoch": 0.17346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.9019479691768073e-05,
-      "loss": 1.0763,
+      "learning_rate": 1.8940035678999114e-05,
+      "loss": 0.9865,
       "step": 6113
     },
     {
-      "epoch": 0.1679162890335338,
+      "epoch": 0.1734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.9019095523293674e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.8939623839268583e-05,
+      "loss": 0.9559,
       "step": 6114
     },
     {
-      "epoch": 0.16794375326137706,
+      "epoch": 0.1735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.9018711283456694e-05,
-      "loss": 1.145,
+      "learning_rate": 1.8939211924024644e-05,
+      "loss": 0.9358,
       "step": 6115
     },
     {
-      "epoch": 0.16797121748922028,
+      "epoch": 0.1735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.9018326972260177e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.8938799933270784e-05,
+      "loss": 0.9356,
       "step": 6116
     },
     {
-      "epoch": 0.16799868171706353,
+      "epoch": 0.1735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.9017942589707168e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.8938387867010476e-05,
+      "loss": 1.0628,
       "step": 6117
     },
     {
-      "epoch": 0.16802614594490675,
+      "epoch": 0.1736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.9017558135800697e-05,
-      "loss": 0.975,
+      "learning_rate": 1.893797572524721e-05,
+      "loss": 1.0325,
       "step": 6118
     },
     {
-      "epoch": 0.16805361017275,
+      "epoch": 0.17363791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.901717361054382e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.8937563507984454e-05,
+      "loss": 1.0102,
       "step": 6119
     },
     {
-      "epoch": 0.16808107440059322,
+      "epoch": 0.17366628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 1.901678901393957e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.8937151215225703e-05,
+      "loss": 1.1099,
       "step": 6120
     },
     {
-      "epoch": 0.16810853862843647,
+      "epoch": 0.17369466515323495,
       "grad_norm": 0.0,
-      "learning_rate": 1.901640434599099e-05,
-      "loss": 0.9747,
+      "learning_rate": 1.8936738846974433e-05,
+      "loss": 1.0567,
       "step": 6121
     },
     {
-      "epoch": 0.1681360028562797,
+      "epoch": 0.17372304199772987,
       "grad_norm": 0.0,
-      "learning_rate": 1.901601960670113e-05,
-      "loss": 1.1282,
+      "learning_rate": 1.8936326403234125e-05,
+      "loss": 1.0084,
       "step": 6122
     },
     {
-      "epoch": 0.16816346708412294,
+      "epoch": 0.17375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.9015634796073027e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.8935913884008265e-05,
+      "loss": 1.0388,
       "step": 6123
     },
     {
-      "epoch": 0.16819093131196616,
+      "epoch": 0.17377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.9015249914109728e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.8935501289300344e-05,
+      "loss": 1.0626,
       "step": 6124
     },
     {
-      "epoch": 0.16821839553980938,
+      "epoch": 0.17380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.9014864960814285e-05,
-      "loss": 1.0974,
+      "learning_rate": 1.893508861911384e-05,
+      "loss": 0.9721,
       "step": 6125
     },
     {
-      "epoch": 0.16824585976765263,
+      "epoch": 0.17383654937570941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9014479936189733e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.893467587345224e-05,
+      "loss": 0.8867,
       "step": 6126
     },
     {
-      "epoch": 0.16827332399549585,
+      "epoch": 0.1738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.9014094840239125e-05,
-      "loss": 1.1042,
+      "learning_rate": 1.893426305231903e-05,
+      "loss": 0.8752,
       "step": 6127
     },
     {
-      "epoch": 0.1683007882233391,
+      "epoch": 0.17389330306469922,
       "grad_norm": 0.0,
-      "learning_rate": 1.901370967296551e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.8933850155717702e-05,
+      "loss": 0.968,
       "step": 6128
     },
     {
-      "epoch": 0.16832825245118233,
+      "epoch": 0.1739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.9013324434371928e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.8933437183651737e-05,
+      "loss": 1.1057,
       "step": 6129
     },
     {
-      "epoch": 0.16835571667902557,
+      "epoch": 0.173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.9012939124461434e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.893302413612463e-05,
+      "loss": 1.0276,
       "step": 6130
     },
     {
-      "epoch": 0.1683831809068688,
+      "epoch": 0.17397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.9012553743237076e-05,
-      "loss": 0.965,
+      "learning_rate": 1.8932611013139862e-05,
+      "loss": 1.0238,
       "step": 6131
     },
     {
-      "epoch": 0.16841064513471204,
+      "epoch": 0.17400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.9012168290701904e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.893219781470093e-05,
+      "loss": 0.9627,
       "step": 6132
     },
     {
-      "epoch": 0.16843810936255527,
+      "epoch": 0.17403518728717365,
       "grad_norm": 0.0,
-      "learning_rate": 1.901178276685896e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.893178454081132e-05,
+      "loss": 1.0583,
       "step": 6133
     },
     {
-      "epoch": 0.16846557359039852,
+      "epoch": 0.17406356413166857,
       "grad_norm": 0.0,
-      "learning_rate": 1.90113971717113e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.8931371191474526e-05,
+      "loss": 1.0414,
       "step": 6134
     },
     {
-      "epoch": 0.16849303781824174,
+      "epoch": 0.17409194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 1.9011011505261975e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.893095776669404e-05,
+      "loss": 1.066,
       "step": 6135
     },
     {
-      "epoch": 0.16852050204608499,
+      "epoch": 0.17412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.901062576751404e-05,
-      "loss": 1.1176,
+      "learning_rate": 1.8930544266473348e-05,
+      "loss": 1.0276,
       "step": 6136
     },
     {
-      "epoch": 0.1685479662739282,
+      "epoch": 0.17414869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 1.9010239958470543e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.8930130690815947e-05,
+      "loss": 1.0654,
       "step": 6137
     },
     {
-      "epoch": 0.16857543050177146,
+      "epoch": 0.17417707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 1.9009854078134534e-05,
-      "loss": 1.1051,
+      "learning_rate": 1.8929717039725333e-05,
+      "loss": 1.0956,
       "step": 6138
     },
     {
-      "epoch": 0.16860289472961468,
+      "epoch": 0.174205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.900946812650907e-05,
-      "loss": 1.013,
+      "learning_rate": 1.8929303313204998e-05,
+      "loss": 1.0548,
       "step": 6139
     },
     {
-      "epoch": 0.1686303589574579,
+      "epoch": 0.17423382519863792,
       "grad_norm": 0.0,
-      "learning_rate": 1.9009082103597206e-05,
-      "loss": 1.0658,
+      "learning_rate": 1.8928889511258432e-05,
+      "loss": 0.9446,
       "step": 6140
     },
     {
-      "epoch": 0.16865782318530115,
+      "epoch": 0.1742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.9008696009401993e-05,
-      "loss": 0.9798,
+      "learning_rate": 1.892847563388914e-05,
+      "loss": 1.069,
       "step": 6141
     },
     {
-      "epoch": 0.16868528741314437,
+      "epoch": 0.1742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.9008309843926488e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.8928061681100615e-05,
+      "loss": 1.0217,
       "step": 6142
     },
     {
-      "epoch": 0.16871275164098762,
+      "epoch": 0.17431895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 1.900792360717375e-05,
-      "loss": 1.0841,
+      "learning_rate": 1.8927647652896347e-05,
+      "loss": 0.9617,
       "step": 6143
     },
     {
-      "epoch": 0.16874021586883084,
+      "epoch": 0.17434733257661747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9007537299146824e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.8927233549279837e-05,
+      "loss": 0.9677,
       "step": 6144
     },
     {
-      "epoch": 0.1687676800966741,
+      "epoch": 0.17437570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 1.900715091984878e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.8926819370254586e-05,
+      "loss": 1.0262,
       "step": 6145
     },
     {
-      "epoch": 0.1687951443245173,
+      "epoch": 0.17440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.9006764469282667e-05,
-      "loss": 1.1023,
+      "learning_rate": 1.892640511582409e-05,
+      "loss": 1.0459,
       "step": 6146
     },
     {
-      "epoch": 0.16882260855236056,
+      "epoch": 0.17443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.9006377947451544e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.892599078599185e-05,
+      "loss": 1.0418,
       "step": 6147
     },
     {
-      "epoch": 0.16885007278020378,
+      "epoch": 0.17446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.900599135435847e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.8925576380761366e-05,
+      "loss": 1.0609,
       "step": 6148
     },
     {
-      "epoch": 0.16887753700804703,
+      "epoch": 0.17448921679909193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9005604690006504e-05,
-      "loss": 1.0815,
+      "learning_rate": 1.8925161900136133e-05,
+      "loss": 1.1014,
       "step": 6149
     },
     {
-      "epoch": 0.16890500123589025,
+      "epoch": 0.17451759364358682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9005217954398705e-05,
-      "loss": 0.988,
+      "learning_rate": 1.8924747344119658e-05,
+      "loss": 0.9979,
       "step": 6150
     },
     {
-      "epoch": 0.1689324654637335,
+      "epoch": 0.17454597048808174,
       "grad_norm": 0.0,
-      "learning_rate": 1.9004831147538133e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.892433271271544e-05,
+      "loss": 1.0802,
       "step": 6151
     },
     {
-      "epoch": 0.16895992969157672,
+      "epoch": 0.17457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.9004444269427846e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.8923918005926984e-05,
+      "loss": 0.9784,
       "step": 6152
     },
     {
-      "epoch": 0.16898739391941994,
+      "epoch": 0.1746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.9004057320070914e-05,
-      "loss": 1.0819,
+      "learning_rate": 1.892350322375779e-05,
+      "loss": 1.0888,
       "step": 6153
     },
     {
-      "epoch": 0.1690148581472632,
+      "epoch": 0.1746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.900367029947039e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.8923088366211367e-05,
+      "loss": 0.9413,
       "step": 6154
     },
     {
-      "epoch": 0.1690423223751064,
+      "epoch": 0.17465947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.9003283207629336e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.892267343329121e-05,
+      "loss": 0.8551,
       "step": 6155
     },
     {
-      "epoch": 0.16906978660294966,
+      "epoch": 0.17468785471055617,
       "grad_norm": 0.0,
-      "learning_rate": 1.900289604455082e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.8922258425000833e-05,
+      "loss": 1.0883,
       "step": 6156
     },
     {
-      "epoch": 0.16909725083079288,
+      "epoch": 0.1747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.9002508810237905e-05,
-      "loss": 0.95,
+      "learning_rate": 1.8921843341343737e-05,
+      "loss": 1.0022,
       "step": 6157
     },
     {
-      "epoch": 0.16912471505863613,
+      "epoch": 0.17474460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 1.900212150469365e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.892142818232343e-05,
+      "loss": 1.0099,
       "step": 6158
     },
     {
-      "epoch": 0.16915217928647935,
+      "epoch": 0.17477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.9001734127921126e-05,
-      "loss": 0.9956,
+      "learning_rate": 1.8921012947943418e-05,
+      "loss": 0.8884,
       "step": 6159
     },
     {
-      "epoch": 0.1691796435143226,
+      "epoch": 0.17480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.9001346679923392e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.8920597638207208e-05,
+      "loss": 1.0276,
       "step": 6160
     },
     {
-      "epoch": 0.16920710774216582,
+      "epoch": 0.17482973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.9000959160703517e-05,
-      "loss": 1.1201,
+      "learning_rate": 1.892018225311831e-05,
+      "loss": 1.042,
       "step": 6161
     },
     {
-      "epoch": 0.16923457197000907,
+      "epoch": 0.17485811577752555,
       "grad_norm": 0.0,
-      "learning_rate": 1.9000571570264565e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.8919766792680227e-05,
+      "loss": 1.004,
       "step": 6162
     },
     {
-      "epoch": 0.1692620361978523,
+      "epoch": 0.17488649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 1.9000183908609605e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.8919351256896477e-05,
+      "loss": 1.0297,
       "step": 6163
     },
     {
-      "epoch": 0.16928950042569554,
+      "epoch": 0.17491486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 1.8999796175741708e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.8918935645770565e-05,
+      "loss": 1.146,
       "step": 6164
     },
     {
-      "epoch": 0.16931696465353877,
+      "epoch": 0.17494324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.8999408371663934e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.8918519959306003e-05,
+      "loss": 1.0668,
       "step": 6165
     },
     {
-      "epoch": 0.169344428881382,
+      "epoch": 0.1749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.8999020496379353e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.89181041975063e-05,
+      "loss": 0.963,
       "step": 6166
     },
     {
-      "epoch": 0.16937189310922524,
+      "epoch": 0.175,
       "grad_norm": 0.0,
-      "learning_rate": 1.8998632549891037e-05,
-      "loss": 1.0268,
+      "learning_rate": 1.891768836037497e-05,
+      "loss": 1.0323,
       "step": 6167
     },
     {
-      "epoch": 0.16939935733706846,
+      "epoch": 0.1750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.8998244532202054e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.8917272447915528e-05,
+      "loss": 0.9962,
       "step": 6168
     },
     {
-      "epoch": 0.1694268215649117,
+      "epoch": 0.1750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.8997856443315475e-05,
-      "loss": 1.1081,
+      "learning_rate": 1.8916856460131487e-05,
+      "loss": 0.9892,
       "step": 6169
     },
     {
-      "epoch": 0.16945428579275493,
+      "epoch": 0.17508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.899746828323437e-05,
-      "loss": 0.996,
+      "learning_rate": 1.8916440397026355e-05,
+      "loss": 1.0098,
       "step": 6170
     },
     {
-      "epoch": 0.16948175002059818,
+      "epoch": 0.17511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.8997080051961812e-05,
-      "loss": 1.1118,
+      "learning_rate": 1.891602425860365e-05,
+      "loss": 0.9894,
       "step": 6171
     },
     {
-      "epoch": 0.1695092142484414,
+      "epoch": 0.17514188422247445,
       "grad_norm": 0.0,
-      "learning_rate": 1.8996691749500874e-05,
-      "loss": 1.0913,
+      "learning_rate": 1.8915608044866885e-05,
+      "loss": 0.9725,
       "step": 6172
     },
     {
-      "epoch": 0.16953667847628465,
+      "epoch": 0.17517026106696934,
       "grad_norm": 0.0,
-      "learning_rate": 1.8996303375854622e-05,
-      "loss": 1.01,
+      "learning_rate": 1.8915191755819584e-05,
+      "loss": 0.9588,
       "step": 6173
     },
     {
-      "epoch": 0.16956414270412787,
+      "epoch": 0.17519863791146426,
       "grad_norm": 0.0,
-      "learning_rate": 1.8995914931026137e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.8914775391465254e-05,
+      "loss": 0.9913,
       "step": 6174
     },
     {
-      "epoch": 0.16959160693197112,
+      "epoch": 0.17522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.8995526415018483e-05,
-      "loss": 0.979,
+      "learning_rate": 1.8914358951807418e-05,
+      "loss": 0.9375,
       "step": 6175
     },
     {
-      "epoch": 0.16961907115981434,
+      "epoch": 0.17525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.899513782783474e-05,
-      "loss": 0.894,
+      "learning_rate": 1.891394243684959e-05,
+      "loss": 1.004,
       "step": 6176
     },
     {
-      "epoch": 0.1696465353876576,
+      "epoch": 0.17528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.8994749169477988e-05,
-      "loss": 1.1252,
+      "learning_rate": 1.8913525846595285e-05,
+      "loss": 1.1022,
       "step": 6177
     },
     {
-      "epoch": 0.1696739996155008,
+      "epoch": 0.1753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.899436043995129e-05,
-      "loss": 1.054,
+      "learning_rate": 1.8913109181048032e-05,
+      "loss": 1.015,
       "step": 6178
     },
     {
-      "epoch": 0.16970146384334406,
+      "epoch": 0.17534052213393872,
       "grad_norm": 0.0,
-      "learning_rate": 1.8993971639257733e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.8912692440211345e-05,
+      "loss": 1.0165,
       "step": 6179
     },
     {
-      "epoch": 0.16972892807118728,
+      "epoch": 0.1753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.8993582767400386e-05,
-      "loss": 1.005,
+      "learning_rate": 1.891227562408874e-05,
+      "loss": 1.0199,
       "step": 6180
     },
     {
-      "epoch": 0.1697563922990305,
+      "epoch": 0.1753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.8993193824382325e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.8911858732683746e-05,
+      "loss": 1.0703,
       "step": 6181
     },
     {
-      "epoch": 0.16978385652687375,
+      "epoch": 0.17542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.8992804810206634e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.891144176599988e-05,
+      "loss": 1.0002,
       "step": 6182
     },
     {
-      "epoch": 0.16981132075471697,
+      "epoch": 0.17545402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.899241572487639e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.8911024724040665e-05,
+      "loss": 0.9532,
       "step": 6183
     },
     {
-      "epoch": 0.16983878498256022,
+      "epoch": 0.17548240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.8992026568394668e-05,
-      "loss": 1.1165,
+      "learning_rate": 1.891060760680962e-05,
+      "loss": 1.1691,
       "step": 6184
     },
     {
-      "epoch": 0.16986624921040344,
+      "epoch": 0.17551078320090807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8991637340764547e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.8910190414310276e-05,
+      "loss": 1.0064,
       "step": 6185
     },
     {
-      "epoch": 0.1698937134382467,
+      "epoch": 0.17553916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.8991248041989107e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.8909773146546152e-05,
+      "loss": 0.9478,
       "step": 6186
     },
     {
-      "epoch": 0.1699211776660899,
+      "epoch": 0.17556753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.8990858672071434e-05,
-      "loss": 0.9326,
+      "learning_rate": 1.8909355803520776e-05,
+      "loss": 0.9826,
       "step": 6187
     },
     {
-      "epoch": 0.16994864189393316,
+      "epoch": 0.17559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.89904692310146e-05,
-      "loss": 1.002,
+      "learning_rate": 1.8908938385237667e-05,
+      "loss": 1.0161,
       "step": 6188
     },
     {
-      "epoch": 0.16997610612177638,
+      "epoch": 0.17562429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 1.8990079718821696e-05,
-      "loss": 1.0569,
+      "learning_rate": 1.8908520891700357e-05,
+      "loss": 0.9844,
       "step": 6189
     },
     {
-      "epoch": 0.17000357034961963,
+      "epoch": 0.1756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.8989690135495796e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.890810332291237e-05,
+      "loss": 1.0104,
       "step": 6190
     },
     {
-      "epoch": 0.17003103457746285,
+      "epoch": 0.17568104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.8989300481039987e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.8907685678877233e-05,
+      "loss": 1.0316,
       "step": 6191
     },
     {
-      "epoch": 0.1700584988053061,
+      "epoch": 0.1757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8988910755457347e-05,
-      "loss": 1.0374,
+      "learning_rate": 1.8907267959598473e-05,
+      "loss": 1.0681,
       "step": 6192
     },
     {
-      "epoch": 0.17008596303314932,
+      "epoch": 0.1757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.8988520958750966e-05,
-      "loss": 1.0542,
+      "learning_rate": 1.8906850165079627e-05,
+      "loss": 1.0576,
       "step": 6193
     },
     {
-      "epoch": 0.17011342726099254,
+      "epoch": 0.17576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.8988131090923922e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.890643229532421e-05,
+      "loss": 1.073,
       "step": 6194
     },
     {
-      "epoch": 0.1701408914888358,
+      "epoch": 0.17579455164585697,
       "grad_norm": 0.0,
-      "learning_rate": 1.898774115197931e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.890601435033576e-05,
+      "loss": 1.0051,
       "step": 6195
     },
     {
-      "epoch": 0.17016835571667902,
+      "epoch": 0.17582292849035186,
       "grad_norm": 0.0,
-      "learning_rate": 1.8987351141920204e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.890559633011781e-05,
+      "loss": 0.9583,
       "step": 6196
     },
     {
-      "epoch": 0.17019581994452226,
+      "epoch": 0.17585130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.898696106074969e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.8905178234673886e-05,
+      "loss": 0.8911,
       "step": 6197
     },
     {
-      "epoch": 0.17022328417236549,
+      "epoch": 0.17587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.8986570908470866e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.890476006400752e-05,
+      "loss": 0.9493,
       "step": 6198
     },
     {
-      "epoch": 0.17025074840020873,
+      "epoch": 0.17590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.898618068508681e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.890434181812225e-05,
+      "loss": 1.0546,
       "step": 6199
     },
     {
-      "epoch": 0.17027821262805196,
+      "epoch": 0.17593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.898579039060061e-05,
-      "loss": 1.061,
+      "learning_rate": 1.89039234970216e-05,
+      "loss": 1.0072,
       "step": 6200
     },
     {
-      "epoch": 0.1703056768558952,
+      "epoch": 0.17596481271282632,
       "grad_norm": 0.0,
-      "learning_rate": 1.8985400025015356e-05,
-      "loss": 1.0911,
+      "learning_rate": 1.8903505100709107e-05,
+      "loss": 1.0251,
       "step": 6201
     },
     {
-      "epoch": 0.17033314108373843,
+      "epoch": 0.17599318955732124,
       "grad_norm": 0.0,
-      "learning_rate": 1.8985009588334137e-05,
-      "loss": 1.0746,
+      "learning_rate": 1.8903086629188308e-05,
+      "loss": 1.042,
       "step": 6202
     },
     {
-      "epoch": 0.17036060531158168,
+      "epoch": 0.17602156640181613,
       "grad_norm": 0.0,
-      "learning_rate": 1.898461908056004e-05,
-      "loss": 1.0851,
+      "learning_rate": 1.8902668082462737e-05,
+      "loss": 1.0182,
       "step": 6203
     },
     {
-      "epoch": 0.1703880695394249,
+      "epoch": 0.176049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8984228501696158e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.8902249460535927e-05,
+      "loss": 0.9849,
       "step": 6204
     },
     {
-      "epoch": 0.17041553376726815,
+      "epoch": 0.1760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.8983837851745582e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.8901830763411417e-05,
+      "loss": 0.9788,
       "step": 6205
     },
     {
-      "epoch": 0.17044299799511137,
+      "epoch": 0.1761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.8983447130711397e-05,
-      "loss": 1.1254,
+      "learning_rate": 1.890141199109274e-05,
+      "loss": 1.0586,
       "step": 6206
     },
     {
-      "epoch": 0.1704704622229546,
+      "epoch": 0.17613507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 1.8983056338596702e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.890099314358344e-05,
+      "loss": 1.0198,
       "step": 6207
     },
     {
-      "epoch": 0.17049792645079784,
+      "epoch": 0.1761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.8982665475404585e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.890057422088705e-05,
+      "loss": 1.1155,
       "step": 6208
     },
     {
-      "epoch": 0.17052539067864106,
+      "epoch": 0.17619182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 1.8982274541138137e-05,
-      "loss": 1.0337,
+      "learning_rate": 1.8900155223007105e-05,
+      "loss": 1.0142,
       "step": 6209
     },
     {
-      "epoch": 0.1705528549064843,
+      "epoch": 0.17622020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.8981883535800458e-05,
-      "loss": 1.1036,
+      "learning_rate": 1.8899736149947152e-05,
+      "loss": 1.068,
       "step": 6210
     },
     {
-      "epoch": 0.17058031913432753,
+      "epoch": 0.17624858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.898149245939463e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.8899317001710726e-05,
+      "loss": 1.0121,
       "step": 6211
     },
     {
-      "epoch": 0.17060778336217078,
+      "epoch": 0.17627695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 1.898110131192376e-05,
-      "loss": 1.0581,
+      "learning_rate": 1.889889777830137e-05,
+      "loss": 1.055,
       "step": 6212
     },
     {
-      "epoch": 0.170635247590014,
+      "epoch": 0.17630533484676503,
       "grad_norm": 0.0,
-      "learning_rate": 1.8980710093390936e-05,
-      "loss": 1.085,
+      "learning_rate": 1.8898478479722624e-05,
+      "loss": 1.0463,
       "step": 6213
     },
     {
-      "epoch": 0.17066271181785725,
+      "epoch": 0.17633371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.8980318803799255e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.889805910597803e-05,
+      "loss": 0.9542,
       "step": 6214
     },
     {
-      "epoch": 0.17069017604570047,
+      "epoch": 0.17636208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 1.897992744315181e-05,
-      "loss": 1.002,
+      "learning_rate": 1.8897639657071135e-05,
+      "loss": 1.0124,
       "step": 6215
     },
     {
-      "epoch": 0.17071764027354372,
+      "epoch": 0.17639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.8979536011451704e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.8897220133005475e-05,
+      "loss": 1.0329,
       "step": 6216
     },
     {
-      "epoch": 0.17074510450138694,
+      "epoch": 0.1764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.897914450870203e-05,
-      "loss": 1.0449,
+      "learning_rate": 1.8896800533784593e-05,
+      "loss": 1.0152,
       "step": 6217
     },
     {
-      "epoch": 0.1707725687292302,
+      "epoch": 0.1764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.8978752934905887e-05,
-      "loss": 1.0486,
+      "learning_rate": 1.889638085941204e-05,
+      "loss": 1.0588,
       "step": 6218
     },
     {
-      "epoch": 0.1708000329570734,
+      "epoch": 0.1764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.8978361290066373e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.8895961109891358e-05,
+      "loss": 1.0178,
       "step": 6219
     },
     {
-      "epoch": 0.17082749718491663,
+      "epoch": 0.1765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.8977969574186585e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.8895541285226095e-05,
+      "loss": 0.9151,
       "step": 6220
     },
     {
-      "epoch": 0.17085496141275988,
+      "epoch": 0.17653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.8977577787269626e-05,
-      "loss": 1.1331,
+      "learning_rate": 1.889512138541979e-05,
+      "loss": 1.0054,
       "step": 6221
     },
     {
-      "epoch": 0.1708824256406031,
+      "epoch": 0.17656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.897718592931859e-05,
-      "loss": 1.0912,
+      "learning_rate": 1.8894701410475998e-05,
+      "loss": 0.9709,
       "step": 6222
     },
     {
-      "epoch": 0.17090988986844635,
+      "epoch": 0.17658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.8976794000336585e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.8894281360398263e-05,
+      "loss": 1.0278,
       "step": 6223
     },
     {
-      "epoch": 0.17093735409628957,
+      "epoch": 0.17661748013620884,
       "grad_norm": 0.0,
-      "learning_rate": 1.897640200032671e-05,
-      "loss": 1.0329,
+      "learning_rate": 1.889386123519013e-05,
+      "loss": 1.0596,
       "step": 6224
     },
     {
-      "epoch": 0.17096481832413282,
+      "epoch": 0.17664585698070376,
       "grad_norm": 0.0,
-      "learning_rate": 1.8976009929292065e-05,
-      "loss": 1.006,
+      "learning_rate": 1.889344103485516e-05,
+      "loss": 1.0366,
       "step": 6225
     },
     {
-      "epoch": 0.17099228255197604,
+      "epoch": 0.17667423382519865,
       "grad_norm": 0.0,
-      "learning_rate": 1.8975617787235747e-05,
-      "loss": 1.108,
+      "learning_rate": 1.8893020759396884e-05,
+      "loss": 1.0742,
       "step": 6226
     },
     {
-      "epoch": 0.1710197467798193,
+      "epoch": 0.17670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.8975225574160866e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.8892600408818866e-05,
+      "loss": 1.0915,
       "step": 6227
     },
     {
-      "epoch": 0.17104721100766251,
+      "epoch": 0.17673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.8974833290070526e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.889217998312465e-05,
+      "loss": 0.906,
       "step": 6228
     },
     {
-      "epoch": 0.17107467523550576,
+      "epoch": 0.1767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.897444093496783e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.8891759482317798e-05,
+      "loss": 0.9787,
       "step": 6229
     },
     {
-      "epoch": 0.17110213946334898,
+      "epoch": 0.1767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.897404850885588e-05,
-      "loss": 1.088,
+      "learning_rate": 1.8891338906401846e-05,
+      "loss": 0.9276,
       "step": 6230
     },
     {
-      "epoch": 0.17112960369119223,
+      "epoch": 0.1768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.8973656011737777e-05,
-      "loss": 0.9571,
+      "learning_rate": 1.889091825538036e-05,
+      "loss": 0.9793,
       "step": 6231
     },
     {
-      "epoch": 0.17115706791903545,
+      "epoch": 0.176844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.897326344361664e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.8890497529256882e-05,
+      "loss": 1.0713,
       "step": 6232
     },
     {
-      "epoch": 0.1711845321468787,
+      "epoch": 0.17687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.897287080449556e-05,
-      "loss": 1.1055,
+      "learning_rate": 1.8890076728034974e-05,
+      "loss": 1.0063,
       "step": 6233
     },
     {
-      "epoch": 0.17121199637472193,
+      "epoch": 0.17690124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.8972478094377653e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.8889655851718188e-05,
+      "loss": 0.9727,
       "step": 6234
     },
     {
-      "epoch": 0.17123946060256515,
+      "epoch": 0.17692962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 1.8972085313266025e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.888923490031008e-05,
+      "loss": 1.067,
       "step": 6235
     },
     {
-      "epoch": 0.1712669248304084,
+      "epoch": 0.17695800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.897169246116378e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.8888813873814208e-05,
+      "loss": 0.9386,
       "step": 6236
     },
     {
-      "epoch": 0.17129438905825162,
+      "epoch": 0.17698637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8971299538074027e-05,
-      "loss": 1.019,
+      "learning_rate": 1.8888392772234122e-05,
+      "loss": 1.1674,
       "step": 6237
     },
     {
-      "epoch": 0.17132185328609487,
+      "epoch": 0.17701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.897090654399988e-05,
-      "loss": 1.0559,
+      "learning_rate": 1.8887971595573385e-05,
+      "loss": 0.9949,
       "step": 6238
     },
     {
-      "epoch": 0.1713493175139381,
+      "epoch": 0.17704313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.897051347894445e-05,
-      "loss": 0.9912,
+      "learning_rate": 1.888755034383555e-05,
+      "loss": 1.1348,
       "step": 6239
     },
     {
-      "epoch": 0.17137678174178134,
+      "epoch": 0.17707150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 1.8970120342910834e-05,
-      "loss": 1.0769,
+      "learning_rate": 1.8887129017024183e-05,
+      "loss": 0.9881,
       "step": 6240
     },
     {
-      "epoch": 0.17140424596962456,
+      "epoch": 0.177099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.8969727135902153e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.888670761514283e-05,
+      "loss": 1.0163,
       "step": 6241
     },
     {
-      "epoch": 0.1714317101974678,
+      "epoch": 0.17712826333711693,
       "grad_norm": 0.0,
-      "learning_rate": 1.896933385792152e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.8886286138195063e-05,
+      "loss": 0.9533,
       "step": 6242
     },
     {
-      "epoch": 0.17145917442531103,
+      "epoch": 0.1771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.896894050897204e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.8885864586184435e-05,
+      "loss": 1.0103,
       "step": 6243
     },
     {
-      "epoch": 0.17148663865315428,
+      "epoch": 0.1771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.896854708905683e-05,
-      "loss": 0.9439,
+      "learning_rate": 1.888544295911451e-05,
+      "loss": 1.0114,
       "step": 6244
     },
     {
-      "epoch": 0.1715141028809975,
+      "epoch": 0.1772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.8968153598179e-05,
-      "loss": 1.0908,
+      "learning_rate": 1.888502125698885e-05,
+      "loss": 1.0644,
       "step": 6245
     },
     {
-      "epoch": 0.17154156710884075,
+      "epoch": 0.17724177071509647,
       "grad_norm": 0.0,
-      "learning_rate": 1.896776003634167e-05,
-      "loss": 0.8583,
+      "learning_rate": 1.8884599479811018e-05,
+      "loss": 1.0243,
       "step": 6246
     },
     {
-      "epoch": 0.17156903133668397,
+      "epoch": 0.17727014755959136,
       "grad_norm": 0.0,
-      "learning_rate": 1.896736640354794e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.888417762758457e-05,
+      "loss": 1.0432,
       "step": 6247
     },
     {
-      "epoch": 0.1715964955645272,
+      "epoch": 0.17729852440408628,
       "grad_norm": 0.0,
-      "learning_rate": 1.896697269980094e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.888375570031308e-05,
+      "loss": 1.1152,
       "step": 6248
     },
     {
-      "epoch": 0.17162395979237044,
+      "epoch": 0.17732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.8966578925103776e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.8883333698000102e-05,
+      "loss": 0.9403,
       "step": 6249
     },
     {
-      "epoch": 0.17165142402021366,
+      "epoch": 0.17735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8966185079459568e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.8882911620649204e-05,
+      "loss": 1.0138,
       "step": 6250
     },
     {
-      "epoch": 0.1716788882480569,
+      "epoch": 0.17738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.896579116287143e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.8882489468263954e-05,
+      "loss": 1.0191,
       "step": 6251
     },
     {
-      "epoch": 0.17170635247590013,
+      "epoch": 0.17741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.896539717534248e-05,
-      "loss": 1.1353,
+      "learning_rate": 1.8882067240847916e-05,
+      "loss": 0.981,
       "step": 6252
     },
     {
-      "epoch": 0.17173381670374338,
+      "epoch": 0.17744040862656071,
       "grad_norm": 0.0,
-      "learning_rate": 1.8965003116875832e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.888164493840466e-05,
+      "loss": 1.007,
       "step": 6253
     },
     {
-      "epoch": 0.1717612809315866,
+      "epoch": 0.17746878547105563,
       "grad_norm": 0.0,
-      "learning_rate": 1.8964608987474608e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.8881222560937745e-05,
+      "loss": 1.0659,
       "step": 6254
     },
     {
-      "epoch": 0.17178874515942985,
+      "epoch": 0.17749716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 1.8964214787141926e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.8880800108450748e-05,
+      "loss": 1.0757,
       "step": 6255
     },
     {
-      "epoch": 0.17181620938727307,
+      "epoch": 0.1775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.89638205158809e-05,
-      "loss": 0.9802,
+      "learning_rate": 1.888037758094723e-05,
+      "loss": 0.9007,
       "step": 6256
     },
     {
-      "epoch": 0.17184367361511632,
+      "epoch": 0.1775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.896342617369466e-05,
-      "loss": 0.9204,
+      "learning_rate": 1.8879954978430762e-05,
+      "loss": 0.8769,
       "step": 6257
     },
     {
-      "epoch": 0.17187113784295954,
+      "epoch": 0.17758229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 1.8963031760586316e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.8879532300904923e-05,
+      "loss": 0.9857,
       "step": 6258
     },
     {
-      "epoch": 0.1718986020708028,
+      "epoch": 0.1776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.8962637276558996e-05,
-      "loss": 0.9726,
+      "learning_rate": 1.8879109548373265e-05,
+      "loss": 1.0693,
       "step": 6259
     },
     {
-      "epoch": 0.171926066298646,
+      "epoch": 0.17763904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 1.8962242721615816e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.8878686720839378e-05,
+      "loss": 1.0665,
       "step": 6260
     },
     {
-      "epoch": 0.17195353052648923,
+      "epoch": 0.17766742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 1.8961848095759897e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.8878263818306818e-05,
+      "loss": 1.0103,
       "step": 6261
     },
     {
-      "epoch": 0.17198099475433248,
+      "epoch": 0.17769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.8961453398994368e-05,
-      "loss": 1.017,
+      "learning_rate": 1.887784084077917e-05,
+      "loss": 1.0178,
       "step": 6262
     },
     {
-      "epoch": 0.1720084589821757,
+      "epoch": 0.17772417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.8961058631322345e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.887741778826e-05,
+      "loss": 1.1224,
       "step": 6263
     },
     {
-      "epoch": 0.17203592321001895,
+      "epoch": 0.17775255391600453,
       "grad_norm": 0.0,
-      "learning_rate": 1.896066379274696e-05,
-      "loss": 1.0466,
+      "learning_rate": 1.8876994660752885e-05,
+      "loss": 1.0537,
       "step": 6264
     },
     {
-      "epoch": 0.17206338743786218,
+      "epoch": 0.17778093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.896026888327133e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.887657145826139e-05,
+      "loss": 1.0886,
       "step": 6265
     },
     {
-      "epoch": 0.17209085166570542,
+      "epoch": 0.17780930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 1.8959873902898582e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.88761481807891e-05,
+      "loss": 1.0518,
       "step": 6266
     },
     {
-      "epoch": 0.17211831589354865,
+      "epoch": 0.17783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.8959478851631838e-05,
-      "loss": 1.1627,
+      "learning_rate": 1.8875724828339593e-05,
+      "loss": 1.0211,
       "step": 6267
     },
     {
-      "epoch": 0.1721457801213919,
+      "epoch": 0.1778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.895908372947423e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.8875301400916433e-05,
+      "loss": 1.0887,
       "step": 6268
     },
     {
-      "epoch": 0.17217324434923512,
+      "epoch": 0.177894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.8958688536428882e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.8874877898523204e-05,
+      "loss": 0.9325,
       "step": 6269
     },
     {
-      "epoch": 0.17220070857707837,
+      "epoch": 0.17792281498297388,
       "grad_norm": 0.0,
-      "learning_rate": 1.8958293272498917e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.8874454321163483e-05,
+      "loss": 0.8967,
       "step": 6270
     },
     {
-      "epoch": 0.1722281728049216,
+      "epoch": 0.1779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.8957897937687467e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.887403066884085e-05,
+      "loss": 0.9767,
       "step": 6271
     },
     {
-      "epoch": 0.17225563703276484,
+      "epoch": 0.17797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.895750253199766e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.8873606941558878e-05,
+      "loss": 1.05,
       "step": 6272
     },
     {
-      "epoch": 0.17228310126060806,
+      "epoch": 0.17800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.895710705543262e-05,
-      "loss": 1.059,
+      "learning_rate": 1.8873183139321152e-05,
+      "loss": 0.9876,
       "step": 6273
     },
     {
-      "epoch": 0.1723105654884513,
+      "epoch": 0.17803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.895671150799548e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.8872759262131247e-05,
+      "loss": 0.9059,
       "step": 6274
     },
     {
-      "epoch": 0.17233802971629453,
+      "epoch": 0.17806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.8956315889689372e-05,
-      "loss": 1.0166,
+      "learning_rate": 1.8872335309992745e-05,
+      "loss": 1.0448,
       "step": 6275
     },
     {
-      "epoch": 0.17236549394413775,
+      "epoch": 0.17809307604994323,
       "grad_norm": 0.0,
-      "learning_rate": 1.895592020051742e-05,
-      "loss": 0.9635,
+      "learning_rate": 1.887191128290923e-05,
+      "loss": 1.0404,
       "step": 6276
     },
     {
-      "epoch": 0.172392958171981,
+      "epoch": 0.17812145289443815,
       "grad_norm": 0.0,
-      "learning_rate": 1.895552444048276e-05,
-      "loss": 1.1114,
+      "learning_rate": 1.8871487180884277e-05,
+      "loss": 1.0329,
       "step": 6277
     },
     {
-      "epoch": 0.17242042239982422,
+      "epoch": 0.17814982973893304,
       "grad_norm": 0.0,
-      "learning_rate": 1.895512860958852e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.887106300392148e-05,
+      "loss": 1.0083,
       "step": 6278
     },
     {
-      "epoch": 0.17244788662766747,
+      "epoch": 0.17817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.8954732707837835e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.8870638752024408e-05,
+      "loss": 0.9276,
       "step": 6279
     },
     {
-      "epoch": 0.1724753508555107,
+      "epoch": 0.1782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.8954336735233835e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.8870214425196658e-05,
+      "loss": 1.0446,
       "step": 6280
     },
     {
-      "epoch": 0.17250281508335394,
+      "epoch": 0.1782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.8953940691779653e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.8869790023441802e-05,
+      "loss": 1.0102,
       "step": 6281
     },
     {
-      "epoch": 0.17253027931119716,
+      "epoch": 0.1782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.8953544577478428e-05,
-      "loss": 1.0518,
+      "learning_rate": 1.8869365546763437e-05,
+      "loss": 1.0869,
       "step": 6282
     },
     {
-      "epoch": 0.1725577435390404,
+      "epoch": 0.1782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.8953148392333285e-05,
-      "loss": 0.9507,
+      "learning_rate": 1.8868940995165136e-05,
+      "loss": 1.0197,
       "step": 6283
     },
     {
-      "epoch": 0.17258520776688363,
+      "epoch": 0.1783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.895275213634737e-05,
-      "loss": 1.1077,
+      "learning_rate": 1.88685163686505e-05,
+      "loss": 0.9335,
       "step": 6284
     },
     {
-      "epoch": 0.17261267199472688,
+      "epoch": 0.17834846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 1.8952355809523806e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.8868091667223098e-05,
+      "loss": 0.9186,
       "step": 6285
     },
     {
-      "epoch": 0.1726401362225701,
+      "epoch": 0.17837684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 1.895195941186574e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.8867666890886532e-05,
+      "loss": 1.0013,
       "step": 6286
     },
     {
-      "epoch": 0.17266760045041335,
+      "epoch": 0.17840522133938705,
       "grad_norm": 0.0,
-      "learning_rate": 1.89515629433763e-05,
-      "loss": 1.014,
+      "learning_rate": 1.8867242039644387e-05,
+      "loss": 0.9525,
       "step": 6287
     },
     {
-      "epoch": 0.17269506467825657,
+      "epoch": 0.17843359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.8951166404058625e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.8866817113500247e-05,
+      "loss": 0.9891,
       "step": 6288
     },
     {
-      "epoch": 0.1727225289060998,
+      "epoch": 0.17846197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.8950769793915857e-05,
-      "loss": 1.1298,
+      "learning_rate": 1.8866392112457706e-05,
+      "loss": 0.9418,
       "step": 6289
     },
     {
-      "epoch": 0.17274999313394304,
+      "epoch": 0.17849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.8950373112951132e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.886596703652035e-05,
+      "loss": 1.0047,
       "step": 6290
     },
     {
-      "epoch": 0.17277745736178626,
+      "epoch": 0.17851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.8949976361167585e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.8865541885691775e-05,
+      "loss": 1.0409,
       "step": 6291
     },
     {
-      "epoch": 0.1728049215896295,
+      "epoch": 0.1785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.894957953856836e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.8865116659975567e-05,
+      "loss": 1.071,
       "step": 6292
     },
     {
-      "epoch": 0.17283238581747273,
+      "epoch": 0.1785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.8949182645156595e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.886469135937532e-05,
+      "loss": 1.0566,
       "step": 6293
     },
     {
-      "epoch": 0.17285985004531598,
+      "epoch": 0.17860385925085132,
       "grad_norm": 0.0,
-      "learning_rate": 1.894878568093543e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.8864265983894626e-05,
+      "loss": 1.1882,
       "step": 6294
     },
     {
-      "epoch": 0.1728873142731592,
+      "epoch": 0.1786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.894838864590801e-05,
-      "loss": 1.0855,
+      "learning_rate": 1.886384053353708e-05,
+      "loss": 0.9763,
       "step": 6295
     },
     {
-      "epoch": 0.17291477850100245,
+      "epoch": 0.1786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.8947991540077464e-05,
-      "loss": 0.9205,
+      "learning_rate": 1.8863415008306277e-05,
+      "loss": 0.9191,
       "step": 6296
     },
     {
-      "epoch": 0.17294224272884567,
+      "epoch": 0.17868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.8947594363446952e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.8862989408205803e-05,
+      "loss": 0.9691,
       "step": 6297
     },
     {
-      "epoch": 0.17296970695668892,
+      "epoch": 0.17871736662883086,
       "grad_norm": 0.0,
-      "learning_rate": 1.8947197116019603e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.8862563733239265e-05,
+      "loss": 1.0325,
       "step": 6298
     },
     {
-      "epoch": 0.17299717118453214,
+      "epoch": 0.17874574347332578,
       "grad_norm": 0.0,
-      "learning_rate": 1.8946799797798564e-05,
-      "loss": 1.2562,
+      "learning_rate": 1.886213798341025e-05,
+      "loss": 1.0562,
       "step": 6299
     },
     {
-      "epoch": 0.1730246354123754,
+      "epoch": 0.17877412031782067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8946402408786978e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.8861712158722355e-05,
+      "loss": 1.1302,
       "step": 6300
     },
     {
-      "epoch": 0.17305209964021862,
+      "epoch": 0.17880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.8946004948987995e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.8861286259179183e-05,
+      "loss": 0.9015,
       "step": 6301
     },
     {
-      "epoch": 0.17307956386806184,
+      "epoch": 0.17883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.894560741840475e-05,
-      "loss": 1.037,
+      "learning_rate": 1.8860860284784323e-05,
+      "loss": 1.0406,
       "step": 6302
     },
     {
-      "epoch": 0.17310702809590509,
+      "epoch": 0.17885925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.89452098170404e-05,
-      "loss": 0.94,
+      "learning_rate": 1.8860434235541383e-05,
+      "loss": 0.9924,
       "step": 6303
     },
     {
-      "epoch": 0.1731344923237483,
+      "epoch": 0.17888762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 1.8944812144898082e-05,
-      "loss": 1.1579,
+      "learning_rate": 1.8860008111453955e-05,
+      "loss": 1.0983,
       "step": 6304
     },
     {
-      "epoch": 0.17316195655159156,
+      "epoch": 0.17891600454029513,
       "grad_norm": 0.0,
-      "learning_rate": 1.8944414401980945e-05,
-      "loss": 0.971,
+      "learning_rate": 1.8859581912525636e-05,
+      "loss": 0.976,
       "step": 6305
     },
     {
-      "epoch": 0.17318942077943478,
+      "epoch": 0.17894438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 1.8944016588292138e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.8859155638760035e-05,
+      "loss": 1.0364,
       "step": 6306
     },
     {
-      "epoch": 0.17321688500727803,
+      "epoch": 0.1789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.8943618703834806e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.885872929016075e-05,
+      "loss": 1.006,
       "step": 6307
     },
     {
-      "epoch": 0.17324434923512125,
+      "epoch": 0.1790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.89432207486121e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.8858302866731377e-05,
+      "loss": 0.9473,
       "step": 6308
     },
     {
-      "epoch": 0.1732718134629645,
+      "epoch": 0.17902951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 1.8942822722627166e-05,
-      "loss": 1.0719,
+      "learning_rate": 1.8857876368475523e-05,
+      "loss": 0.9913,
       "step": 6309
     },
     {
-      "epoch": 0.17329927769080772,
+      "epoch": 0.17905788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 1.8942424625883157e-05,
-      "loss": 1.1171,
+      "learning_rate": 1.8857449795396786e-05,
+      "loss": 0.9434,
       "step": 6310
     },
     {
-      "epoch": 0.17332674191865097,
+      "epoch": 0.17908626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 1.8942026458383216e-05,
-      "loss": 1.054,
+      "learning_rate": 1.8857023147498776e-05,
+      "loss": 1.0181,
       "step": 6311
     },
     {
-      "epoch": 0.1733542061464942,
+      "epoch": 0.17911464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 1.89416282201305e-05,
-      "loss": 0.9645,
+      "learning_rate": 1.8856596424785093e-05,
+      "loss": 0.9352,
       "step": 6312
     },
     {
-      "epoch": 0.17338167037433744,
+      "epoch": 0.17914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.8941229911128158e-05,
-      "loss": 0.9052,
+      "learning_rate": 1.885616962725934e-05,
+      "loss": 1.0848,
       "step": 6313
     },
     {
-      "epoch": 0.17340913460218066,
+      "epoch": 0.17917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.8940831531379338e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.8855742754925122e-05,
+      "loss": 0.9811,
       "step": 6314
     },
     {
-      "epoch": 0.17343659883002388,
+      "epoch": 0.17919977298524403,
       "grad_norm": 0.0,
-      "learning_rate": 1.89404330808872e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.8855315807786048e-05,
+      "loss": 1.0414,
       "step": 6315
     },
     {
-      "epoch": 0.17346406305786713,
+      "epoch": 0.17922814982973892,
       "grad_norm": 0.0,
-      "learning_rate": 1.8940034559654893e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.885488878584573e-05,
+      "loss": 1.1918,
       "step": 6316
     },
     {
-      "epoch": 0.17349152728571035,
+      "epoch": 0.17925652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.8939635967685564e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.885446168910776e-05,
+      "loss": 1.0164,
       "step": 6317
     },
     {
-      "epoch": 0.1735189915135536,
+      "epoch": 0.17928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.8939237304982376e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.8854034517575758e-05,
+      "loss": 1.0471,
       "step": 6318
     },
     {
-      "epoch": 0.17354645574139682,
+      "epoch": 0.1793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.8938838571548478e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.885360727125333e-05,
+      "loss": 0.9198,
       "step": 6319
     },
     {
-      "epoch": 0.17357391996924007,
+      "epoch": 0.1793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.893843976738703e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.8853179950144077e-05,
+      "loss": 0.9755,
       "step": 6320
     },
     {
-      "epoch": 0.1736013841970833,
+      "epoch": 0.17937003405221338,
       "grad_norm": 0.0,
-      "learning_rate": 1.893804089250118e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.885275255425162e-05,
+      "loss": 0.9809,
       "step": 6321
     },
     {
-      "epoch": 0.17362884842492654,
+      "epoch": 0.1793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.893764194689409e-05,
-      "loss": 1.005,
+      "learning_rate": 1.8852325083579564e-05,
+      "loss": 0.9081,
       "step": 6322
     },
     {
-      "epoch": 0.17365631265276976,
+      "epoch": 0.1794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.893724293056892e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.885189753813152e-05,
+      "loss": 0.9806,
       "step": 6323
     },
     {
-      "epoch": 0.173683776880613,
+      "epoch": 0.17945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8936843843528815e-05,
-      "loss": 0.9947,
+      "learning_rate": 1.8851469917911098e-05,
+      "loss": 1.0782,
       "step": 6324
     },
     {
-      "epoch": 0.17371124110845623,
+      "epoch": 0.17948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.8936444685776943e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.8851042222921912e-05,
+      "loss": 0.9452,
       "step": 6325
     },
     {
-      "epoch": 0.17373870533629948,
+      "epoch": 0.17951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.8936045457316455e-05,
-      "loss": 0.9592,
+      "learning_rate": 1.8850614453167577e-05,
+      "loss": 0.9079,
       "step": 6326
     },
     {
-      "epoch": 0.1737661695641427,
+      "epoch": 0.17954029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.893564615815052e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.8850186608651702e-05,
+      "loss": 0.9182,
       "step": 6327
     },
     {
-      "epoch": 0.17379363379198595,
+      "epoch": 0.17956867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8935246788282285e-05,
-      "loss": 1.0509,
+      "learning_rate": 1.8849758689377902e-05,
+      "loss": 0.9227,
       "step": 6328
     },
     {
-      "epoch": 0.17382109801982917,
+      "epoch": 0.17959704880817254,
       "grad_norm": 0.0,
-      "learning_rate": 1.893484734771492e-05,
-      "loss": 0.8775,
+      "learning_rate": 1.8849330695349795e-05,
+      "loss": 1.0497,
       "step": 6329
     },
     {
-      "epoch": 0.1738485622476724,
+      "epoch": 0.17962542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 1.8934447836451583e-05,
-      "loss": 1.107,
+      "learning_rate": 1.884890262657099e-05,
+      "loss": 0.9795,
       "step": 6330
     },
     {
-      "epoch": 0.17387602647551564,
+      "epoch": 0.1796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.893404825449543e-05,
-      "loss": 1.0512,
+      "learning_rate": 1.884847448304511e-05,
+      "loss": 1.0322,
       "step": 6331
     },
     {
-      "epoch": 0.17390349070335887,
+      "epoch": 0.1796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.8933648601849627e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.8848046264775766e-05,
+      "loss": 0.9718,
       "step": 6332
     },
     {
-      "epoch": 0.17393095493120211,
+      "epoch": 0.1797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.893324887851734e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.8847617971766577e-05,
+      "loss": 1.1306,
       "step": 6333
     },
     {
-      "epoch": 0.17395841915904534,
+      "epoch": 0.179738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.893284908450172e-05,
-      "loss": 1.0259,
+      "learning_rate": 1.8847189604021165e-05,
+      "loss": 0.9569,
       "step": 6334
     },
     {
-      "epoch": 0.17398588338688858,
+      "epoch": 0.1797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.8932449219805944e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.8846761161543143e-05,
+      "loss": 1.0153,
       "step": 6335
     },
     {
-      "epoch": 0.1740133476147318,
+      "epoch": 0.17979568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.893204928443317e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.884633264433613e-05,
+      "loss": 1.112,
       "step": 6336
     },
     {
-      "epoch": 0.17404081184257505,
+      "epoch": 0.17982406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 1.8931649278386557e-05,
-      "loss": 1.011,
+      "learning_rate": 1.884590405240375e-05,
+      "loss": 0.8921,
       "step": 6337
     },
     {
-      "epoch": 0.17406827607041828,
+      "epoch": 0.17985244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.8931249201669278e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.8845475385749622e-05,
+      "loss": 1.0049,
       "step": 6338
     },
     {
-      "epoch": 0.17409574029826153,
+      "epoch": 0.17988081725312147,
       "grad_norm": 0.0,
-      "learning_rate": 1.8930849054284498e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.8845046644377363e-05,
+      "loss": 0.9366,
       "step": 6339
     },
     {
-      "epoch": 0.17412320452610475,
+      "epoch": 0.17990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.8930448836235377e-05,
-      "loss": 1.0512,
+      "learning_rate": 1.8844617828290597e-05,
+      "loss": 1.0223,
       "step": 6340
     },
     {
-      "epoch": 0.174150668753948,
+      "epoch": 0.17993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.8930048547525087e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.8844188937492948e-05,
+      "loss": 0.9737,
       "step": 6341
     },
     {
-      "epoch": 0.17417813298179122,
+      "epoch": 0.17996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.8929648188156796e-05,
-      "loss": 1.0758,
+      "learning_rate": 1.8843759971988036e-05,
+      "loss": 1.1011,
       "step": 6342
     },
     {
-      "epoch": 0.17420559720963444,
+      "epoch": 0.17999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.8929247758133665e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.8843330931779486e-05,
+      "loss": 0.9869,
       "step": 6343
     },
     {
-      "epoch": 0.1742330614374777,
+      "epoch": 0.1800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.8928847257458872e-05,
-      "loss": 1.0836,
+      "learning_rate": 1.8842901816870922e-05,
+      "loss": 0.9775,
       "step": 6344
     },
     {
-      "epoch": 0.1742605256653209,
+      "epoch": 0.18005107832009082,
       "grad_norm": 0.0,
-      "learning_rate": 1.892844668613558e-05,
-      "loss": 1.028,
+      "learning_rate": 1.884247262726597e-05,
+      "loss": 1.0895,
       "step": 6345
     },
     {
-      "epoch": 0.17428798989316416,
+      "epoch": 0.1800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8928046044166956e-05,
-      "loss": 1.033,
+      "learning_rate": 1.8842043362968256e-05,
+      "loss": 0.9847,
       "step": 6346
     },
     {
-      "epoch": 0.17431545412100738,
+      "epoch": 0.1801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.892764533155617e-05,
-      "loss": 0.958,
+      "learning_rate": 1.88416140239814e-05,
+      "loss": 1.0861,
       "step": 6347
     },
     {
-      "epoch": 0.17434291834885063,
+      "epoch": 0.18013620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.8927244548306403e-05,
-      "loss": 0.8536,
+      "learning_rate": 1.884118461030904e-05,
+      "loss": 0.9209,
       "step": 6348
     },
     {
-      "epoch": 0.17437038257669385,
+      "epoch": 0.18016458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.8926843694420816e-05,
-      "loss": 1.08,
+      "learning_rate": 1.884075512195479e-05,
+      "loss": 1.1302,
       "step": 6349
     },
     {
-      "epoch": 0.1743978468045371,
+      "epoch": 0.18019296254256525,
       "grad_norm": 0.0,
-      "learning_rate": 1.8926442769902585e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.8840325558922285e-05,
+      "loss": 1.0317,
       "step": 6350
     },
     {
-      "epoch": 0.17442531103238032,
+      "epoch": 0.18022133938706017,
       "grad_norm": 0.0,
-      "learning_rate": 1.892604177475488e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.8839895921215153e-05,
+      "loss": 1.0132,
       "step": 6351
     },
     {
-      "epoch": 0.17445277526022357,
+      "epoch": 0.18024971623155506,
       "grad_norm": 0.0,
-      "learning_rate": 1.8925640708980878e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.8839466208837022e-05,
+      "loss": 0.9623,
       "step": 6352
     },
     {
-      "epoch": 0.1744802394880668,
+      "epoch": 0.18027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.8925239572583742e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.8839036421791524e-05,
+      "loss": 0.8889,
       "step": 6353
     },
     {
-      "epoch": 0.17450770371591004,
+      "epoch": 0.18030646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.892483836556666e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.8838606560082287e-05,
+      "loss": 1.0383,
       "step": 6354
     },
     {
-      "epoch": 0.17453516794375326,
+      "epoch": 0.18033484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 1.8924437087932797e-05,
-      "loss": 1.086,
+      "learning_rate": 1.883817662371294e-05,
+      "loss": 1.11,
       "step": 6355
     },
     {
-      "epoch": 0.17456263217159648,
+      "epoch": 0.1803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.892403573968533e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.8837746612687123e-05,
+      "loss": 1.0094,
       "step": 6356
     },
     {
-      "epoch": 0.17459009639943973,
+      "epoch": 0.18039160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 1.8923634320827436e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.883731652700846e-05,
+      "loss": 0.8709,
       "step": 6357
     },
     {
-      "epoch": 0.17461756062728295,
+      "epoch": 0.1804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.8923232831362292e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.8836886366680587e-05,
+      "loss": 0.8553,
       "step": 6358
     },
     {
-      "epoch": 0.1746450248551262,
+      "epoch": 0.1804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.892283127129307e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.8836456131707143e-05,
+      "loss": 0.9492,
       "step": 6359
     },
     {
-      "epoch": 0.17467248908296942,
+      "epoch": 0.18047673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 1.8922429640622954e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.883602582209175e-05,
+      "loss": 0.9048,
       "step": 6360
     },
     {
-      "epoch": 0.17469995331081267,
+      "epoch": 0.18050510783200907,
       "grad_norm": 0.0,
-      "learning_rate": 1.892202793935512e-05,
-      "loss": 1.1099,
+      "learning_rate": 1.8835595437838052e-05,
+      "loss": 1.0671,
       "step": 6361
     },
     {
-      "epoch": 0.1747274175386559,
+      "epoch": 0.18053348467650399,
       "grad_norm": 0.0,
-      "learning_rate": 1.8921626167492742e-05,
-      "loss": 0.9404,
+      "learning_rate": 1.8835164978949683e-05,
+      "loss": 0.9635,
       "step": 6362
     },
     {
-      "epoch": 0.17475488176649914,
+      "epoch": 0.18056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.8921224325039e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.8834734445430277e-05,
+      "loss": 1.0303,
       "step": 6363
     },
     {
-      "epoch": 0.17478234599434236,
+      "epoch": 0.18059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.8920822411997077e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.8834303837283474e-05,
+      "loss": 1.116,
       "step": 6364
     },
     {
-      "epoch": 0.1748098102221856,
+      "epoch": 0.18061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.8920420428370153e-05,
-      "loss": 1.048,
+      "learning_rate": 1.883387315451291e-05,
+      "loss": 1.094,
       "step": 6365
     },
     {
-      "epoch": 0.17483727445002883,
+      "epoch": 0.18064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.8920018374161406e-05,
-      "loss": 1.0589,
+      "learning_rate": 1.883344239712222e-05,
+      "loss": 0.9243,
       "step": 6366
     },
     {
-      "epoch": 0.17486473867787208,
+      "epoch": 0.18067536889897842,
       "grad_norm": 0.0,
-      "learning_rate": 1.891961624937402e-05,
-      "loss": 1.0083,
+      "learning_rate": 1.8833011565115043e-05,
+      "loss": 0.986,
       "step": 6367
     },
     {
-      "epoch": 0.1748922029057153,
+      "epoch": 0.18070374574347334,
       "grad_norm": 0.0,
-      "learning_rate": 1.8919214054011173e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.8832580658495026e-05,
+      "loss": 1.0043,
       "step": 6368
     },
     {
-      "epoch": 0.17491966713355853,
+      "epoch": 0.18073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.8918811788076052e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.8832149677265796e-05,
+      "loss": 0.9064,
       "step": 6369
     },
     {
-      "epoch": 0.17494713136140178,
+      "epoch": 0.1807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.8918409451571834e-05,
-      "loss": 0.947,
+      "learning_rate": 1.8831718621431004e-05,
+      "loss": 1.0758,
       "step": 6370
     },
     {
-      "epoch": 0.174974595589245,
+      "epoch": 0.180788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.891800704450171e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.883128749099429e-05,
+      "loss": 1.0882,
       "step": 6371
     },
     {
-      "epoch": 0.17500205981708825,
+      "epoch": 0.1808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.8917604566868856e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.883085628595929e-05,
+      "loss": 0.9979,
       "step": 6372
     },
     {
-      "epoch": 0.17502952404493147,
+      "epoch": 0.18084562996594777,
       "grad_norm": 0.0,
-      "learning_rate": 1.8917202018676458e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.8830425006329657e-05,
+      "loss": 0.9074,
       "step": 6373
     },
     {
-      "epoch": 0.17505698827277472,
+      "epoch": 0.1808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.891679939992771e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.882999365210902e-05,
+      "loss": 0.9399,
       "step": 6374
     },
     {
-      "epoch": 0.17508445250061794,
+      "epoch": 0.18090238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 1.8916396710625785e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.882956222330103e-05,
+      "loss": 0.9673,
       "step": 6375
     },
     {
-      "epoch": 0.1751119167284612,
+      "epoch": 0.18093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8915993950773877e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.8829130719909332e-05,
+      "loss": 1.132,
       "step": 6376
     },
     {
-      "epoch": 0.1751393809563044,
+      "epoch": 0.18095913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.8915591120375174e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.882869914193757e-05,
+      "loss": 1.1569,
       "step": 6377
     },
     {
-      "epoch": 0.17516684518414766,
+      "epoch": 0.18098751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.891518821943286e-05,
-      "loss": 1.1138,
+      "learning_rate": 1.8828267489389393e-05,
+      "loss": 1.0083,
       "step": 6378
     },
     {
-      "epoch": 0.17519430941199088,
+      "epoch": 0.18101589103291715,
       "grad_norm": 0.0,
-      "learning_rate": 1.891478524795012e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.8827835762268438e-05,
+      "loss": 1.0799,
       "step": 6379
     },
     {
-      "epoch": 0.17522177363983413,
+      "epoch": 0.18104426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.891438220593015e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.882740396057836e-05,
+      "loss": 0.9386,
       "step": 6380
     },
     {
-      "epoch": 0.17524923786767735,
+      "epoch": 0.18107264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 1.891397909337613e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.8826972084322802e-05,
+      "loss": 1.0343,
       "step": 6381
     },
     {
-      "epoch": 0.1752767020955206,
+      "epoch": 0.18110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.891357591029126e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.8826540133505416e-05,
+      "loss": 1.0392,
       "step": 6382
     },
     {
-      "epoch": 0.17530416632336382,
+      "epoch": 0.1811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.891317265667872e-05,
-      "loss": 1.016,
+      "learning_rate": 1.882610810812985e-05,
+      "loss": 1.1032,
       "step": 6383
     },
     {
-      "epoch": 0.17533163055120704,
+      "epoch": 0.1811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8912769332541708e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.8825676008199753e-05,
+      "loss": 1.0299,
       "step": 6384
     },
     {
-      "epoch": 0.1753590947790503,
+      "epoch": 0.1811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.891236593788341e-05,
-      "loss": 1.0291,
+      "learning_rate": 1.882524383371877e-05,
+      "loss": 0.9916,
       "step": 6385
     },
     {
-      "epoch": 0.1753865590068935,
+      "epoch": 0.1812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.891196247270702e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.8824811584690557e-05,
+      "loss": 1.0369,
       "step": 6386
     },
     {
-      "epoch": 0.17541402323473676,
+      "epoch": 0.18124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.8911558937015736e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.8824379261118765e-05,
+      "loss": 1.0493,
       "step": 6387
     },
     {
-      "epoch": 0.17544148746257998,
+      "epoch": 0.18127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.8911155330812742e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.8823946863007044e-05,
+      "loss": 0.9352,
       "step": 6388
     },
     {
-      "epoch": 0.17546895169042323,
+      "epoch": 0.18129965947786605,
       "grad_norm": 0.0,
-      "learning_rate": 1.891075165410124e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.8823514390359046e-05,
+      "loss": 0.9841,
       "step": 6389
     },
     {
-      "epoch": 0.17549641591826645,
+      "epoch": 0.18132803632236094,
       "grad_norm": 0.0,
-      "learning_rate": 1.8910347906884413e-05,
-      "loss": 1.0374,
+      "learning_rate": 1.8823081843178426e-05,
+      "loss": 1.0529,
       "step": 6390
     },
     {
-      "epoch": 0.1755238801461097,
+      "epoch": 0.18135641316685586,
       "grad_norm": 0.0,
-      "learning_rate": 1.890994408916546e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.882264922146884e-05,
+      "loss": 0.9948,
       "step": 6391
     },
     {
-      "epoch": 0.17555134437395292,
+      "epoch": 0.18138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.890954020094758e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.8822216525233937e-05,
+      "loss": 0.9566,
       "step": 6392
     },
     {
-      "epoch": 0.17557880860179617,
+      "epoch": 0.18141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.8909136242233974e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.8821783754477373e-05,
+      "loss": 0.8743,
       "step": 6393
     },
     {
-      "epoch": 0.1756062728296394,
+      "epoch": 0.18144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.8908732213027826e-05,
-      "loss": 1.084,
+      "learning_rate": 1.8821350909202806e-05,
+      "loss": 1.0751,
       "step": 6394
     },
     {
-      "epoch": 0.17563373705748264,
+      "epoch": 0.1814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.8908328113332334e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.882091798941389e-05,
+      "loss": 1.0136,
       "step": 6395
     },
     {
-      "epoch": 0.17566120128532586,
+      "epoch": 0.1814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8907923943150704e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.8820484995114286e-05,
+      "loss": 0.9014,
       "step": 6396
     },
     {
-      "epoch": 0.17568866551316908,
+      "epoch": 0.1815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8907519702486128e-05,
-      "loss": 1.035,
+      "learning_rate": 1.8820051926307647e-05,
+      "loss": 1.0902,
       "step": 6397
     },
     {
-      "epoch": 0.17571612974101233,
+      "epoch": 0.1815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.8907115391341807e-05,
-      "loss": 1.1301,
+      "learning_rate": 1.8819618782997634e-05,
+      "loss": 0.9735,
       "step": 6398
     },
     {
-      "epoch": 0.17574359396885555,
+      "epoch": 0.18158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.8906711009720936e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.8819185565187904e-05,
+      "loss": 0.973,
       "step": 6399
     },
     {
-      "epoch": 0.1757710581966988,
+      "epoch": 0.18161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.8906306557626718e-05,
-      "loss": 1.0896,
+      "learning_rate": 1.881875227288212e-05,
+      "loss": 0.9258,
       "step": 6400
     },
     {
-      "epoch": 0.17579852242454203,
+      "epoch": 0.18164018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 1.890590203506235e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.8818318906083936e-05,
+      "loss": 1.0609,
       "step": 6401
     },
     {
-      "epoch": 0.17582598665238527,
+      "epoch": 0.18166855845629967,
       "grad_norm": 0.0,
-      "learning_rate": 1.890549744203104e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.8817885464797014e-05,
+      "loss": 0.8623,
       "step": 6402
     },
     {
-      "epoch": 0.1758534508802285,
+      "epoch": 0.18169693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 1.8905092778535982e-05,
-      "loss": 0.9856,
+      "learning_rate": 1.8817451949025018e-05,
+      "loss": 0.9808,
       "step": 6403
     },
     {
-      "epoch": 0.17588091510807174,
+      "epoch": 0.18172531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.8904688044580383e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.8817018358771612e-05,
+      "loss": 0.9532,
       "step": 6404
     },
     {
-      "epoch": 0.17590837933591497,
+      "epoch": 0.18175368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 1.8904283240167442e-05,
-      "loss": 1.101,
+      "learning_rate": 1.8816584694040455e-05,
+      "loss": 0.9568,
       "step": 6405
     },
     {
-      "epoch": 0.17593584356375822,
+      "epoch": 0.18178206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 1.890387836530036e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.8816150954835208e-05,
+      "loss": 0.9836,
       "step": 6406
     },
     {
-      "epoch": 0.17596330779160144,
+      "epoch": 0.1818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.8903473419982346e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.8815717141159538e-05,
+      "loss": 1.0502,
       "step": 6407
     },
     {
-      "epoch": 0.17599077201944469,
+      "epoch": 0.18183881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 1.8903068404216602e-05,
-      "loss": 1.1134,
+      "learning_rate": 1.881528325301711e-05,
+      "loss": 0.964,
       "step": 6408
     },
     {
-      "epoch": 0.1760182362472879,
+      "epoch": 0.1818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.890266331800633e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.8814849290411587e-05,
+      "loss": 1.0002,
       "step": 6409
     },
     {
-      "epoch": 0.17604570047513113,
+      "epoch": 0.1818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.8902258161354737e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.8814415253346638e-05,
+      "loss": 1.0955,
       "step": 6410
     },
     {
-      "epoch": 0.17607316470297438,
+      "epoch": 0.18192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.890185293426503e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.8813981141825927e-05,
+      "loss": 0.9325,
       "step": 6411
     },
     {
-      "epoch": 0.1761006289308176,
+      "epoch": 0.18195232690124857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8901447636740416e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.881354695585312e-05,
+      "loss": 0.869,
       "step": 6412
     },
     {
-      "epoch": 0.17612809315866085,
+      "epoch": 0.18198070374574346,
       "grad_norm": 0.0,
-      "learning_rate": 1.89010422687841e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.8813112695431885e-05,
+      "loss": 0.8888,
       "step": 6413
     },
     {
-      "epoch": 0.17615555738650407,
+      "epoch": 0.18200908059023838,
       "grad_norm": 0.0,
-      "learning_rate": 1.8900636830399285e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.8812678360565893e-05,
+      "loss": 1.0801,
       "step": 6414
     },
     {
-      "epoch": 0.17618302161434732,
+      "epoch": 0.18203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.8900231321589188e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.881224395125881e-05,
+      "loss": 0.9736,
       "step": 6415
     },
     {
-      "epoch": 0.17621048584219054,
+      "epoch": 0.18206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.8899825742357012e-05,
-      "loss": 0.966,
+      "learning_rate": 1.8811809467514302e-05,
+      "loss": 1.0884,
       "step": 6416
     },
     {
-      "epoch": 0.1762379500700338,
+      "epoch": 0.18209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.8899420092705966e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.881137490933605e-05,
+      "loss": 0.943,
       "step": 6417
     },
     {
-      "epoch": 0.176265414297877,
+      "epoch": 0.18212258796821792,
       "grad_norm": 0.0,
-      "learning_rate": 1.8899014372639265e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.8810940276727713e-05,
+      "loss": 0.9417,
       "step": 6418
     },
     {
-      "epoch": 0.17629287852572026,
+      "epoch": 0.18215096481271284,
       "grad_norm": 0.0,
-      "learning_rate": 1.889860858216011e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.881050556969297e-05,
+      "loss": 1.0446,
       "step": 6419
     },
     {
-      "epoch": 0.17632034275356348,
+      "epoch": 0.18217934165720773,
       "grad_norm": 0.0,
-      "learning_rate": 1.889820272127172e-05,
-      "loss": 1.031,
+      "learning_rate": 1.881007078823549e-05,
+      "loss": 1.0779,
       "step": 6420
     },
     {
-      "epoch": 0.17634780698140673,
+      "epoch": 0.18220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.8897796789977302e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.8809635932358946e-05,
+      "loss": 1.067,
       "step": 6421
     },
     {
-      "epoch": 0.17637527120924995,
+      "epoch": 0.1822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.889739078828007e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.880920100206701e-05,
+      "loss": 0.9064,
       "step": 6422
     },
     {
-      "epoch": 0.1764027354370932,
+      "epoch": 0.1822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.8896984716183235e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.880876599736336e-05,
+      "loss": 0.9287,
       "step": 6423
     },
     {
-      "epoch": 0.17643019966493642,
+      "epoch": 0.18229284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 1.889657857369001e-05,
-      "loss": 0.9259,
+      "learning_rate": 1.8808330918251667e-05,
+      "loss": 1.0022,
       "step": 6424
     },
     {
-      "epoch": 0.17645766389277964,
+      "epoch": 0.1823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.889617236080361e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.8807895764735605e-05,
+      "loss": 0.9531,
       "step": 6425
     },
     {
-      "epoch": 0.1764851281206229,
+      "epoch": 0.18234960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 1.889576607752725e-05,
-      "loss": 0.9009,
+      "learning_rate": 1.880746053681885e-05,
+      "loss": 1.1246,
       "step": 6426
     },
     {
-      "epoch": 0.1765125923484661,
+      "epoch": 0.18237797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.8895359723864138e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.8807025234505083e-05,
+      "loss": 0.9581,
       "step": 6427
     },
     {
-      "epoch": 0.17654005657630936,
+      "epoch": 0.18240635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.88949532998175e-05,
-      "loss": 1.1731,
+      "learning_rate": 1.8806589857797977e-05,
+      "loss": 1.0727,
       "step": 6428
     },
     {
-      "epoch": 0.17656752080415258,
+      "epoch": 0.18243473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.889454680539054e-05,
-      "loss": 0.912,
+      "learning_rate": 1.880615440670121e-05,
+      "loss": 1.1356,
       "step": 6429
     },
     {
-      "epoch": 0.17659498503199583,
+      "epoch": 0.18246311010215663,
       "grad_norm": 0.0,
-      "learning_rate": 1.8894140240586483e-05,
-      "loss": 1.1449,
+      "learning_rate": 1.8805718881218463e-05,
+      "loss": 1.0799,
       "step": 6430
     },
     {
-      "epoch": 0.17662244925983905,
+      "epoch": 0.18249148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 1.8893733605408548e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.8805283281353413e-05,
+      "loss": 1.0104,
       "step": 6431
     },
     {
-      "epoch": 0.1766499134876823,
+      "epoch": 0.18251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.889332689985994e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.8804847607109735e-05,
+      "loss": 1.0731,
       "step": 6432
     },
     {
-      "epoch": 0.17667737771552552,
+      "epoch": 0.18254824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.889292012394389e-05,
-      "loss": 1.0571,
+      "learning_rate": 1.8804411858491117e-05,
+      "loss": 0.9969,
       "step": 6433
     },
     {
-      "epoch": 0.17670484194336877,
+      "epoch": 0.1825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.8892513277663605e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.8803976035501234e-05,
+      "loss": 1.0557,
       "step": 6434
     },
     {
-      "epoch": 0.176732306171212,
+      "epoch": 0.1826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8892106361022316e-05,
-      "loss": 1.016,
+      "learning_rate": 1.8803540138143772e-05,
+      "loss": 1.064,
       "step": 6435
     },
     {
-      "epoch": 0.17675977039905524,
+      "epoch": 0.18263337116912598,
       "grad_norm": 0.0,
-      "learning_rate": 1.8891699374023235e-05,
-      "loss": 0.94,
+      "learning_rate": 1.880310416642241e-05,
+      "loss": 0.956,
       "step": 6436
     },
     {
-      "epoch": 0.17678723462689847,
+      "epoch": 0.1826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.8891292316669584e-05,
-      "loss": 0.9654,
+      "learning_rate": 1.880266812034083e-05,
+      "loss": 1.0348,
       "step": 6437
     },
     {
-      "epoch": 0.1768146988547417,
+      "epoch": 0.18269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.8890885188964585e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.880223199990272e-05,
+      "loss": 1.0868,
       "step": 6438
     },
     {
-      "epoch": 0.17684216308258494,
+      "epoch": 0.18271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8890477990911456e-05,
-      "loss": 1.0794,
+      "learning_rate": 1.880179580511176e-05,
+      "loss": 1.0597,
       "step": 6439
     },
     {
-      "epoch": 0.17686962731042816,
+      "epoch": 0.18274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.8890070722513424e-05,
-      "loss": 1.097,
+      "learning_rate": 1.8801359535971628e-05,
+      "loss": 1.0139,
       "step": 6440
     },
     {
-      "epoch": 0.1768970915382714,
+      "epoch": 0.18277525539160044,
       "grad_norm": 0.0,
-      "learning_rate": 1.888966338377371e-05,
-      "loss": 1.1772,
+      "learning_rate": 1.8800923192486023e-05,
+      "loss": 0.9448,
       "step": 6441
     },
     {
-      "epoch": 0.17692455576611463,
+      "epoch": 0.18280363223609536,
       "grad_norm": 0.0,
-      "learning_rate": 1.8889255974695533e-05,
-      "loss": 1.0243,
+      "learning_rate": 1.8800486774658623e-05,
+      "loss": 0.9473,
       "step": 6442
     },
     {
-      "epoch": 0.17695201999395788,
+      "epoch": 0.18283200908059025,
       "grad_norm": 0.0,
-      "learning_rate": 1.888884849528212e-05,
-      "loss": 1.0858,
+      "learning_rate": 1.8800050282493118e-05,
+      "loss": 1.036,
       "step": 6443
     },
     {
-      "epoch": 0.1769794842218011,
+      "epoch": 0.18286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.8888440945536695e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.8799613715993187e-05,
+      "loss": 1.0336,
       "step": 6444
     },
     {
-      "epoch": 0.17700694844964435,
+      "epoch": 0.18288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.8888033325462482e-05,
-      "loss": 1.0961,
+      "learning_rate": 1.8799177075162527e-05,
+      "loss": 1.0523,
       "step": 6445
     },
     {
-      "epoch": 0.17703441267748757,
+      "epoch": 0.1829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.8887625635062705e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.8798740360004825e-05,
+      "loss": 1.0652,
       "step": 6446
     },
     {
-      "epoch": 0.17706187690533082,
+      "epoch": 0.1829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.8887217874340592e-05,
-      "loss": 1.0143,
+      "learning_rate": 1.8798303570523762e-05,
+      "loss": 0.9563,
       "step": 6447
     },
     {
-      "epoch": 0.17708934113317404,
+      "epoch": 0.1829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.888681004329937e-05,
-      "loss": 1.1205,
+      "learning_rate": 1.879786670672304e-05,
+      "loss": 0.9966,
       "step": 6448
     },
     {
-      "epoch": 0.1771168053610173,
+      "epoch": 0.1830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.8886402141942262e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.8797429768606336e-05,
+      "loss": 1.0247,
       "step": 6449
     },
     {
-      "epoch": 0.1771442695888605,
+      "epoch": 0.18303064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 1.88859941702725e-05,
-      "loss": 1.1059,
+      "learning_rate": 1.879699275617735e-05,
+      "loss": 1.1024,
       "step": 6450
     },
     {
-      "epoch": 0.17717173381670373,
+      "epoch": 0.18305902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.888558612829331e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.879655566943977e-05,
+      "loss": 0.8653,
       "step": 6451
     },
     {
-      "epoch": 0.17719919804454698,
+      "epoch": 0.18308740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 1.888517801600792e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.879611850839729e-05,
+      "loss": 1.0274,
       "step": 6452
     },
     {
-      "epoch": 0.1772266622723902,
+      "epoch": 0.18311577752553915,
       "grad_norm": 0.0,
-      "learning_rate": 1.888476983341956e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.87956812730536e-05,
+      "loss": 1.0322,
       "step": 6453
     },
     {
-      "epoch": 0.17725412650023345,
+      "epoch": 0.18314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.8884361580531457e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.8795243963412394e-05,
+      "loss": 1.0427,
       "step": 6454
     },
     {
-      "epoch": 0.17728159072807667,
+      "epoch": 0.18317253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 1.8883953257346847e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.8794806579477372e-05,
+      "loss": 0.9465,
       "step": 6455
     },
     {
-      "epoch": 0.17730905495591992,
+      "epoch": 0.18320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.888354486386896e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.8794369121252217e-05,
+      "loss": 0.9563,
       "step": 6456
     },
     {
-      "epoch": 0.17733651918376314,
+      "epoch": 0.18322928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 1.888313640010102e-05,
-      "loss": 1.1272,
+      "learning_rate": 1.8793931588740636e-05,
+      "loss": 0.9644,
       "step": 6457
     },
     {
-      "epoch": 0.1773639834116064,
+      "epoch": 0.1832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.8882727866046264e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.879349398194632e-05,
+      "loss": 0.948,
       "step": 6458
     },
     {
-      "epoch": 0.1773914476394496,
+      "epoch": 0.18328603859250853,
       "grad_norm": 0.0,
-      "learning_rate": 1.8882319261707928e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.879305630087296e-05,
+      "loss": 0.9925,
       "step": 6459
     },
     {
-      "epoch": 0.17741891186729286,
+      "epoch": 0.18331441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.8881910587089236e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.879261854552426e-05,
+      "loss": 1.0117,
       "step": 6460
     },
     {
-      "epoch": 0.17744637609513608,
+      "epoch": 0.1833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.888150184219343e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.8792180715903917e-05,
+      "loss": 1.0381,
       "step": 6461
     },
     {
-      "epoch": 0.17747384032297933,
+      "epoch": 0.1833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.888109302702374e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.879174281201563e-05,
+      "loss": 0.975,
       "step": 6462
     },
     {
-      "epoch": 0.17750130455082255,
+      "epoch": 0.18339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.8880684141583404e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.8791304833863094e-05,
+      "loss": 1.1044,
       "step": 6463
     },
     {
-      "epoch": 0.17752876877866577,
+      "epoch": 0.18342792281498296,
       "grad_norm": 0.0,
-      "learning_rate": 1.8880275185875655e-05,
-      "loss": 1.062,
+      "learning_rate": 1.879086678145001e-05,
+      "loss": 1.0711,
       "step": 6464
     },
     {
-      "epoch": 0.17755623300650902,
+      "epoch": 0.18345629965947788,
       "grad_norm": 0.0,
-      "learning_rate": 1.8879866159903726e-05,
-      "loss": 1.1586,
+      "learning_rate": 1.879042865478008e-05,
+      "loss": 0.9766,
       "step": 6465
     },
     {
-      "epoch": 0.17758369723435224,
+      "epoch": 0.18348467650397277,
       "grad_norm": 0.0,
-      "learning_rate": 1.8879457063670857e-05,
-      "loss": 0.996,
+      "learning_rate": 1.8789990453857005e-05,
+      "loss": 0.9596,
       "step": 6466
     },
     {
-      "epoch": 0.1776111614621955,
+      "epoch": 0.18351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8879047897180288e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.8789552178684483e-05,
+      "loss": 0.994,
       "step": 6467
     },
     {
-      "epoch": 0.17763862569003872,
+      "epoch": 0.18354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.8878638660435248e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.8789113829266223e-05,
+      "loss": 1.0937,
       "step": 6468
     },
     {
-      "epoch": 0.17766608991788196,
+      "epoch": 0.18356980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.887822935343898e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.8788675405605915e-05,
+      "loss": 0.9268,
       "step": 6469
     },
     {
-      "epoch": 0.17769355414572519,
+      "epoch": 0.18359818388195231,
       "grad_norm": 0.0,
-      "learning_rate": 1.8877819976194724e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.878823690770728e-05,
+      "loss": 1.0343,
       "step": 6470
     },
     {
-      "epoch": 0.17772101837356843,
+      "epoch": 0.18362656072644723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8877410528705718e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.8787798335574007e-05,
+      "loss": 1.1396,
       "step": 6471
     },
     {
-      "epoch": 0.17774848260141166,
+      "epoch": 0.18365493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 1.8877001010975197e-05,
-      "loss": 1.1798,
+      "learning_rate": 1.8787359689209808e-05,
+      "loss": 1.0162,
       "step": 6472
     },
     {
-      "epoch": 0.1777759468292549,
+      "epoch": 0.183683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.887659142300641e-05,
-      "loss": 1.1611,
+      "learning_rate": 1.8786920968618387e-05,
+      "loss": 0.8803,
       "step": 6473
     },
     {
-      "epoch": 0.17780341105709813,
+      "epoch": 0.1837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.887618176480259e-05,
-      "loss": 1.0814,
+      "learning_rate": 1.878648217380345e-05,
+      "loss": 1.0203,
       "step": 6474
     },
     {
-      "epoch": 0.17783087528494138,
+      "epoch": 0.18374006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 1.8875772036366983e-05,
-      "loss": 0.8464,
+      "learning_rate": 1.8786043304768702e-05,
+      "loss": 0.9964,
       "step": 6475
     },
     {
-      "epoch": 0.1778583395127846,
+      "epoch": 0.18376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.887536223770283e-05,
-      "loss": 0.9956,
+      "learning_rate": 1.8785604361517852e-05,
+      "loss": 1.0846,
       "step": 6476
     },
     {
-      "epoch": 0.17788580374062785,
+      "epoch": 0.18379682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 1.887495236881337e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.8785165344054607e-05,
+      "loss": 0.983,
       "step": 6477
     },
     {
-      "epoch": 0.17791326796847107,
+      "epoch": 0.18382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.8874542429701856e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.878472625238268e-05,
+      "loss": 1.0284,
       "step": 6478
     },
     {
-      "epoch": 0.1779407321963143,
+      "epoch": 0.18385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.887413242037152e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.878428708650577e-05,
+      "loss": 1.1753,
       "step": 6479
     },
     {
-      "epoch": 0.17796819642415754,
+      "epoch": 0.18388195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.887372234082561e-05,
-      "loss": 1.0481,
+      "learning_rate": 1.8783847846427593e-05,
+      "loss": 1.0176,
       "step": 6480
     },
     {
-      "epoch": 0.17799566065200076,
+      "epoch": 0.18391032917139613,
       "grad_norm": 0.0,
-      "learning_rate": 1.8873312191067374e-05,
-      "loss": 1.0589,
+      "learning_rate": 1.878340853215186e-05,
+      "loss": 1.0819,
       "step": 6481
     },
     {
-      "epoch": 0.178023124879844,
+      "epoch": 0.18393870601589105,
       "grad_norm": 0.0,
-      "learning_rate": 1.887290197110005e-05,
-      "loss": 1.0116,
+      "learning_rate": 1.8782969143682278e-05,
+      "loss": 1.116,
       "step": 6482
     },
     {
-      "epoch": 0.17805058910768723,
+      "epoch": 0.18396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.887249168092689e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.878252968102256e-05,
+      "loss": 0.933,
       "step": 6483
     },
     {
-      "epoch": 0.17807805333553048,
+      "epoch": 0.18399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.887208132055114e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.878209014417642e-05,
+      "loss": 0.9852,
       "step": 6484
     },
     {
-      "epoch": 0.1781055175633737,
+      "epoch": 0.1840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.887167088997605e-05,
-      "loss": 1.1333,
+      "learning_rate": 1.8781650533147572e-05,
+      "loss": 1.01,
       "step": 6485
     },
     {
-      "epoch": 0.17813298179121695,
+      "epoch": 0.1840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.887126038920486e-05,
-      "loss": 1.1243,
+      "learning_rate": 1.8781210847939726e-05,
+      "loss": 1.0763,
       "step": 6486
     },
     {
-      "epoch": 0.17816044601906017,
+      "epoch": 0.18408059023836548,
       "grad_norm": 0.0,
-      "learning_rate": 1.887084981824082e-05,
-      "loss": 1.0757,
+      "learning_rate": 1.8780771088556595e-05,
+      "loss": 1.0972,
       "step": 6487
     },
     {
-      "epoch": 0.17818791024690342,
+      "epoch": 0.1841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.8870439177087184e-05,
-      "loss": 1.063,
+      "learning_rate": 1.8780331255001896e-05,
+      "loss": 0.9945,
       "step": 6488
     },
     {
-      "epoch": 0.17821537447474664,
+      "epoch": 0.18413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.8870028465747192e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.8779891347279346e-05,
+      "loss": 1.0423,
       "step": 6489
     },
     {
-      "epoch": 0.1782428387025899,
+      "epoch": 0.18416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.88696176842241e-05,
-      "loss": 0.9968,
+      "learning_rate": 1.877945136539266e-05,
+      "loss": 1.0781,
       "step": 6490
     },
     {
-      "epoch": 0.1782703029304331,
+      "epoch": 0.18419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.886920683252116e-05,
-      "loss": 1.1021,
+      "learning_rate": 1.8779011309345554e-05,
+      "loss": 1.1011,
       "step": 6491
     },
     {
-      "epoch": 0.17829776715827633,
+      "epoch": 0.18422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.8868795910641616e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.8778571179141743e-05,
+      "loss": 1.0535,
       "step": 6492
     },
     {
-      "epoch": 0.17832523138611958,
+      "epoch": 0.18425085130533483,
       "grad_norm": 0.0,
-      "learning_rate": 1.886838491858873e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.8778130974784944e-05,
+      "loss": 1.0689,
       "step": 6493
     },
     {
-      "epoch": 0.1783526956139628,
+      "epoch": 0.18427922814982975,
       "grad_norm": 0.0,
-      "learning_rate": 1.8867973856365738e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.877769069627888e-05,
+      "loss": 0.9885,
       "step": 6494
     },
     {
-      "epoch": 0.17838015984180605,
+      "epoch": 0.18430760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.886756272397591e-05,
-      "loss": 1.017,
+      "learning_rate": 1.8777250343627274e-05,
+      "loss": 1.0001,
       "step": 6495
     },
     {
-      "epoch": 0.17840762406964927,
+      "epoch": 0.18433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.886715152142249e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.8776809916833833e-05,
+      "loss": 0.9648,
       "step": 6496
     },
     {
-      "epoch": 0.17843508829749252,
+      "epoch": 0.1843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.8866740248708727e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.877636941590229e-05,
+      "loss": 1.0813,
       "step": 6497
     },
     {
-      "epoch": 0.17846255252533574,
+      "epoch": 0.1843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.8866328905837884e-05,
-      "loss": 1.0766,
+      "learning_rate": 1.8775928840836353e-05,
+      "loss": 0.9619,
       "step": 6498
     },
     {
-      "epoch": 0.178490016753179,
+      "epoch": 0.1844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.8865917492813213e-05,
-      "loss": 0.9112,
+      "learning_rate": 1.8775488191639755e-05,
+      "loss": 1.0073,
       "step": 6499
     },
     {
-      "epoch": 0.17851748098102221,
+      "epoch": 0.1844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.8865506009637965e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.8775047468316213e-05,
+      "loss": 1.06,
       "step": 6500
     },
     {
-      "epoch": 0.17854494520886546,
+      "epoch": 0.184477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.88650944563154e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.8774606670869445e-05,
+      "loss": 0.9951,
       "step": 6501
     },
     {
-      "epoch": 0.17857240943670868,
+      "epoch": 0.18450624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.8864682832848777e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.8774165799303188e-05,
+      "loss": 1.0329,
       "step": 6502
     },
     {
-      "epoch": 0.17859987366455193,
+      "epoch": 0.18453461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 1.8864271139241346e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.8773724853621154e-05,
+      "loss": 0.9731,
       "step": 6503
     },
     {
-      "epoch": 0.17862733789239515,
+      "epoch": 0.18456299659477865,
       "grad_norm": 0.0,
-      "learning_rate": 1.886385937549637e-05,
-      "loss": 1.1007,
+      "learning_rate": 1.8773283833827076e-05,
+      "loss": 0.8956,
       "step": 6504
     },
     {
-      "epoch": 0.17865480212023838,
+      "epoch": 0.18459137343927357,
       "grad_norm": 0.0,
-      "learning_rate": 1.8863447541617103e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.877284273992467e-05,
+      "loss": 0.987,
       "step": 6505
     },
     {
-      "epoch": 0.17868226634808163,
+      "epoch": 0.18461975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.8863035637606807e-05,
-      "loss": 0.8489,
+      "learning_rate": 1.877240157191767e-05,
+      "loss": 0.8973,
       "step": 6506
     },
     {
-      "epoch": 0.17870973057592485,
+      "epoch": 0.18464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.886262366346874e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.8771960329809794e-05,
+      "loss": 1.0651,
       "step": 6507
     },
     {
-      "epoch": 0.1787371948037681,
+      "epoch": 0.18467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.8862211619206164e-05,
-      "loss": 1.0556,
+      "learning_rate": 1.8771519013604782e-05,
+      "loss": 1.0259,
       "step": 6508
     },
     {
-      "epoch": 0.17876465903161132,
+      "epoch": 0.18470488081725311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8861799504822333e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.8771077623306353e-05,
+      "loss": 0.977,
       "step": 6509
     },
     {
-      "epoch": 0.17879212325945457,
+      "epoch": 0.184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.8861387320320515e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.8770636158918236e-05,
+      "loss": 0.9646,
       "step": 6510
     },
     {
-      "epoch": 0.1788195874872978,
+      "epoch": 0.18476163450624292,
       "grad_norm": 0.0,
-      "learning_rate": 1.8860975065703966e-05,
-      "loss": 1.1118,
+      "learning_rate": 1.877019462044416e-05,
+      "loss": 0.9446,
       "step": 6511
     },
     {
-      "epoch": 0.17884705171514104,
+      "epoch": 0.1847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.8860562740975948e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.8769753007887858e-05,
+      "loss": 0.942,
       "step": 6512
     },
     {
-      "epoch": 0.17887451594298426,
+      "epoch": 0.1848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.886015034613973e-05,
-      "loss": 1.0806,
+      "learning_rate": 1.8769311321253054e-05,
+      "loss": 1.1045,
       "step": 6513
     },
     {
-      "epoch": 0.1789019801708275,
+      "epoch": 0.18484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.8859737881198566e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.8768869560543488e-05,
+      "loss": 0.9527,
       "step": 6514
     },
     {
-      "epoch": 0.17892944439867073,
+      "epoch": 0.18487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.8859325346155727e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.8768427725762883e-05,
+      "loss": 1.0727,
       "step": 6515
     },
     {
-      "epoch": 0.17895690862651398,
+      "epoch": 0.18490351872871735,
       "grad_norm": 0.0,
-      "learning_rate": 1.885891274101447e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.8767985816914976e-05,
+      "loss": 1.0738,
       "step": 6516
     },
     {
-      "epoch": 0.1789843728543572,
+      "epoch": 0.18493189557321227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8858500065778067e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.8767543834003497e-05,
+      "loss": 1.0756,
       "step": 6517
     },
     {
-      "epoch": 0.17901183708220042,
+      "epoch": 0.18496027241770716,
       "grad_norm": 0.0,
-      "learning_rate": 1.8858087320449782e-05,
-      "loss": 1.1185,
+      "learning_rate": 1.8767101777032188e-05,
+      "loss": 1.0589,
       "step": 6518
     },
     {
-      "epoch": 0.17903930131004367,
+      "epoch": 0.18498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.885767450503287e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.876665964600477e-05,
+      "loss": 1.0507,
       "step": 6519
     },
     {
-      "epoch": 0.1790667655378869,
+      "epoch": 0.18501702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 1.8857261619530618e-05,
-      "loss": 1.0476,
+      "learning_rate": 1.8766217440924986e-05,
+      "loss": 1.057,
       "step": 6520
     },
     {
-      "epoch": 0.17909422976573014,
+      "epoch": 0.18504540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 1.8856848663946274e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.8765775161796565e-05,
+      "loss": 0.9144,
       "step": 6521
     },
     {
-      "epoch": 0.17912169399357336,
+      "epoch": 0.18507377979568673,
       "grad_norm": 0.0,
-      "learning_rate": 1.8856435638283114e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.8765332808623255e-05,
+      "loss": 1.0329,
       "step": 6522
     },
     {
-      "epoch": 0.1791491582214166,
+      "epoch": 0.18510215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 1.8856022542544404e-05,
-      "loss": 1.093,
+      "learning_rate": 1.8764890381408775e-05,
+      "loss": 1.1051,
       "step": 6523
     },
     {
-      "epoch": 0.17917662244925983,
+      "epoch": 0.1851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8855609376733415e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.876444788015688e-05,
+      "loss": 1.0051,
       "step": 6524
     },
     {
-      "epoch": 0.17920408667710308,
+      "epoch": 0.1851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.8855196140853415e-05,
-      "loss": 0.9856,
+      "learning_rate": 1.8764005304871297e-05,
+      "loss": 0.9879,
       "step": 6525
     },
     {
-      "epoch": 0.1792315509049463,
+      "epoch": 0.18518728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.885478283490767e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.8763562655555766e-05,
+      "loss": 1.0326,
       "step": 6526
     },
     {
-      "epoch": 0.17925901513278955,
+      "epoch": 0.18521566401816117,
       "grad_norm": 0.0,
-      "learning_rate": 1.8854369458899454e-05,
-      "loss": 1.0653,
+      "learning_rate": 1.8763119932214028e-05,
+      "loss": 1.0128,
       "step": 6527
     },
     {
-      "epoch": 0.17928647936063277,
+      "epoch": 0.18524404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 1.8853956012832037e-05,
-      "loss": 1.0848,
+      "learning_rate": 1.8762677134849823e-05,
+      "loss": 0.9412,
       "step": 6528
     },
     {
-      "epoch": 0.17931394358847602,
+      "epoch": 0.18527241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 1.885354249670869e-05,
-      "loss": 1.1043,
+      "learning_rate": 1.8762234263466892e-05,
+      "loss": 1.0144,
       "step": 6529
     },
     {
-      "epoch": 0.17934140781631924,
+      "epoch": 0.18530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.8853128910532684e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.8761791318068972e-05,
+      "loss": 0.9397,
       "step": 6530
     },
     {
-      "epoch": 0.1793688720441625,
+      "epoch": 0.18532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.8852715254307298e-05,
-      "loss": 1.0788,
+      "learning_rate": 1.8761348298659806e-05,
+      "loss": 1.1646,
       "step": 6531
     },
     {
-      "epoch": 0.1793963362720057,
+      "epoch": 0.18535754824063563,
       "grad_norm": 0.0,
-      "learning_rate": 1.885230152803579e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.876090520524314e-05,
+      "loss": 0.8765,
       "step": 6532
     },
     {
-      "epoch": 0.17942380049984893,
+      "epoch": 0.18538592508513052,
       "grad_norm": 0.0,
-      "learning_rate": 1.8851887731721456e-05,
-      "loss": 0.9755,
+      "learning_rate": 1.876046203782271e-05,
+      "loss": 0.9662,
       "step": 6533
     },
     {
-      "epoch": 0.17945126472769218,
+      "epoch": 0.18541430192962544,
       "grad_norm": 0.0,
-      "learning_rate": 1.8851473865367548e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.8760018796402267e-05,
+      "loss": 1.0827,
       "step": 6534
     },
     {
-      "epoch": 0.1794787289555354,
+      "epoch": 0.18544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.885105992897735e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.875957548098555e-05,
+      "loss": 1.0559,
       "step": 6535
     },
     {
-      "epoch": 0.17950619318337865,
+      "epoch": 0.1854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.8850645922554142e-05,
-      "loss": 1.1754,
+      "learning_rate": 1.8759132091576302e-05,
+      "loss": 0.9302,
       "step": 6536
     },
     {
-      "epoch": 0.17953365741122188,
+      "epoch": 0.1854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.8850231846101193e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.8758688628178276e-05,
+      "loss": 0.9247,
       "step": 6537
     },
     {
-      "epoch": 0.17956112163906512,
+      "epoch": 0.18552780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.884981769962178e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.875824509079521e-05,
+      "loss": 1.0093,
       "step": 6538
     },
     {
-      "epoch": 0.17958858586690835,
+      "epoch": 0.1855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.8849403483119185e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.8757801479430858e-05,
+      "loss": 0.9377,
       "step": 6539
     },
     {
-      "epoch": 0.1796160500947516,
+      "epoch": 0.1855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.8848989196596678e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.875735779408896e-05,
+      "loss": 0.9981,
       "step": 6540
     },
     {
-      "epoch": 0.17964351432259482,
+      "epoch": 0.18561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.8848574840057543e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.875691403477327e-05,
+      "loss": 1.0344,
       "step": 6541
     },
     {
-      "epoch": 0.17967097855043807,
+      "epoch": 0.18564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.8848160413505057e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.875647020148753e-05,
+      "loss": 1.06,
       "step": 6542
     },
     {
-      "epoch": 0.1796984427782813,
+      "epoch": 0.18566969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.8847745916942495e-05,
-      "loss": 1.0582,
+      "learning_rate": 1.8756026294235494e-05,
+      "loss": 0.9782,
       "step": 6543
     },
     {
-      "epoch": 0.17972590700612454,
+      "epoch": 0.18569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.8847331350373142e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.8755582313020912e-05,
+      "loss": 1.0138,
       "step": 6544
     },
     {
-      "epoch": 0.17975337123396776,
+      "epoch": 0.18572644721906925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8846916713800277e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.875513825784753e-05,
+      "loss": 0.988,
       "step": 6545
     },
     {
-      "epoch": 0.17978083546181098,
+      "epoch": 0.18575482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.884650200722718e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.8754694128719103e-05,
+      "loss": 0.9764,
       "step": 6546
     },
     {
-      "epoch": 0.17980829968965423,
+      "epoch": 0.18578320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.8846087230657128e-05,
-      "loss": 1.1155,
+      "learning_rate": 1.875424992563938e-05,
+      "loss": 0.9028,
       "step": 6547
     },
     {
-      "epoch": 0.17983576391749745,
+      "epoch": 0.1858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.884567238409341e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.8753805648612115e-05,
+      "loss": 1.045,
       "step": 6548
     },
     {
-      "epoch": 0.1798632281453407,
+      "epoch": 0.1858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.884525746753931e-05,
-      "loss": 1.1362,
+      "learning_rate": 1.8753361297641058e-05,
+      "loss": 0.9886,
       "step": 6549
     },
     {
-      "epoch": 0.17989069237318392,
+      "epoch": 0.1858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8844842480998102e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.8752916872729967e-05,
+      "loss": 0.989,
       "step": 6550
     },
     {
-      "epoch": 0.17991815660102717,
+      "epoch": 0.1858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.8844427424473075e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.8752472373882597e-05,
+      "loss": 1.0767,
       "step": 6551
     },
     {
-      "epoch": 0.1799456208288704,
+      "epoch": 0.1859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.8844012297967513e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.8752027801102694e-05,
+      "loss": 0.9443,
       "step": 6552
     },
     {
-      "epoch": 0.17997308505671364,
+      "epoch": 0.18595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.8843597101484697e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.8751583154394023e-05,
+      "loss": 1.0769,
       "step": 6553
     },
     {
-      "epoch": 0.18000054928455686,
+      "epoch": 0.18598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.8843181835027916e-05,
-      "loss": 1.12,
+      "learning_rate": 1.875113843376033e-05,
+      "loss": 1.0063,
       "step": 6554
     },
     {
-      "epoch": 0.1800280135124001,
+      "epoch": 0.18601021566401815,
       "grad_norm": 0.0,
-      "learning_rate": 1.8842766498600458e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.875069363920538e-05,
+      "loss": 0.9595,
       "step": 6555
     },
     {
-      "epoch": 0.18005547774024333,
+      "epoch": 0.18603859250851304,
       "grad_norm": 0.0,
-      "learning_rate": 1.88423510922056e-05,
-      "loss": 1.1068,
+      "learning_rate": 1.875024877073293e-05,
+      "loss": 1.0256,
       "step": 6556
     },
     {
-      "epoch": 0.18008294196808658,
+      "epoch": 0.18606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.884193561584664e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.8749803828346732e-05,
+      "loss": 1.0612,
       "step": 6557
     },
     {
-      "epoch": 0.1801104061959298,
+      "epoch": 0.18609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.884152006952686e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.874935881205055e-05,
+      "loss": 1.0345,
       "step": 6558
     },
     {
-      "epoch": 0.18013787042377302,
+      "epoch": 0.18612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.884110445324955e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.874891372184814e-05,
+      "loss": 0.9107,
       "step": 6559
     },
     {
-      "epoch": 0.18016533465161627,
+      "epoch": 0.18615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.8840688767017995e-05,
-      "loss": 1.0736,
+      "learning_rate": 1.8748468557743265e-05,
+      "loss": 0.8879,
       "step": 6560
     },
     {
-      "epoch": 0.1801927988794595,
+      "epoch": 0.1861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.8840273010835482e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.8748023319739678e-05,
+      "loss": 0.9162,
       "step": 6561
     },
     {
-      "epoch": 0.18022026310730274,
+      "epoch": 0.18620885357548242,
       "grad_norm": 0.0,
-      "learning_rate": 1.8839857184705307e-05,
-      "loss": 0.986,
+      "learning_rate": 1.8747578007841146e-05,
+      "loss": 0.9601,
       "step": 6562
     },
     {
-      "epoch": 0.18024772733514596,
+      "epoch": 0.1862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.883944128863076e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.874713262205143e-05,
+      "loss": 1.0666,
       "step": 6563
     },
     {
-      "epoch": 0.1802751915629892,
+      "epoch": 0.1862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.8839025322615126e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.8746687162374294e-05,
+      "loss": 0.9504,
       "step": 6564
     },
     {
-      "epoch": 0.18030265579083243,
+      "epoch": 0.18629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.8838609286661703e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.8746241628813498e-05,
+      "loss": 0.9437,
       "step": 6565
     },
     {
-      "epoch": 0.18033012001867568,
+      "epoch": 0.18632236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 1.8838193180773778e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.87457960213728e-05,
+      "loss": 1.1493,
       "step": 6566
     },
     {
-      "epoch": 0.1803575842465189,
+      "epoch": 0.18635073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.8837777004954646e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.8745350340055974e-05,
+      "loss": 1.032,
       "step": 6567
     },
     {
-      "epoch": 0.18038504847436215,
+      "epoch": 0.18637911464245177,
       "grad_norm": 0.0,
-      "learning_rate": 1.88373607592076e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.8744904584866782e-05,
+      "loss": 0.9407,
       "step": 6568
     },
     {
-      "epoch": 0.18041251270220537,
+      "epoch": 0.18640749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 1.883694444353593e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.8744458755808988e-05,
+      "loss": 1.0222,
       "step": 6569
     },
     {
-      "epoch": 0.18043997693004862,
+      "epoch": 0.18643586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.8836528057942934e-05,
-      "loss": 0.967,
+      "learning_rate": 1.8744012852886357e-05,
+      "loss": 1.001,
       "step": 6570
     },
     {
-      "epoch": 0.18046744115789184,
+      "epoch": 0.18646424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 1.88361116024319e-05,
-      "loss": 1.1533,
+      "learning_rate": 1.8743566876102655e-05,
+      "loss": 1.0004,
       "step": 6571
     },
     {
-      "epoch": 0.1804949053857351,
+      "epoch": 0.18649262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 1.8835695077006137e-05,
-      "loss": 1.0546,
+      "learning_rate": 1.874312082546165e-05,
+      "loss": 1.1311,
       "step": 6572
     },
     {
-      "epoch": 0.18052236961357832,
+      "epoch": 0.1865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.8835278481668923e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.874267470096711e-05,
+      "loss": 1.0044,
       "step": 6573
     },
     {
-      "epoch": 0.18054983384142154,
+      "epoch": 0.18654937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 1.883486181642357e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.8742228502622807e-05,
+      "loss": 0.9982,
       "step": 6574
     },
     {
-      "epoch": 0.18057729806926479,
+      "epoch": 0.186577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8834445081273364e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.8741782230432503e-05,
+      "loss": 0.9474,
       "step": 6575
     },
     {
-      "epoch": 0.180604762297108,
+      "epoch": 0.1866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.883402827622161e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.8741335884399974e-05,
+      "loss": 1.1313,
       "step": 6576
     },
     {
-      "epoch": 0.18063222652495126,
+      "epoch": 0.18663450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 1.88336114012716e-05,
-      "loss": 1.036,
+      "learning_rate": 1.8740889464528988e-05,
+      "loss": 0.977,
       "step": 6577
     },
     {
-      "epoch": 0.18065969075279448,
+      "epoch": 0.18666288308740067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8833194456426636e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.8740442970823315e-05,
+      "loss": 1.0845,
       "step": 6578
     },
     {
-      "epoch": 0.18068715498063773,
+      "epoch": 0.1866912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.8832777441690014e-05,
-      "loss": 1.0348,
+      "learning_rate": 1.8739996403286727e-05,
+      "loss": 0.9176,
       "step": 6579
     },
     {
-      "epoch": 0.18071461920848095,
+      "epoch": 0.18671963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.8832360357065035e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.8739549761923e-05,
+      "loss": 0.9128,
       "step": 6580
     },
     {
-      "epoch": 0.1807420834363242,
+      "epoch": 0.18674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.8831943202555e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.8739103046735896e-05,
+      "loss": 0.9819,
       "step": 6581
     },
     {
-      "epoch": 0.18076954766416742,
+      "epoch": 0.18677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.883152597816321e-05,
-      "loss": 1.0594,
+      "learning_rate": 1.8738656257729202e-05,
+      "loss": 1.0152,
       "step": 6582
     },
     {
-      "epoch": 0.18079701189201067,
+      "epoch": 0.18680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.8831108683892964e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.8738209394906683e-05,
+      "loss": 1.0577,
       "step": 6583
     },
     {
-      "epoch": 0.1808244761198539,
+      "epoch": 0.18683314415437002,
       "grad_norm": 0.0,
-      "learning_rate": 1.883069131974757e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.8737762458272114e-05,
+      "loss": 0.9373,
       "step": 6584
     },
     {
-      "epoch": 0.18085194034769714,
+      "epoch": 0.18686152099886494,
       "grad_norm": 0.0,
-      "learning_rate": 1.883027388573032e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.873731544782928e-05,
+      "loss": 1.1229,
       "step": 6585
     },
     {
-      "epoch": 0.18087940457554036,
+      "epoch": 0.18688989784335983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8829856381844527e-05,
-      "loss": 0.9953,
+      "learning_rate": 1.8736868363581943e-05,
+      "loss": 1.0989,
       "step": 6586
     },
     {
-      "epoch": 0.18090686880338358,
+      "epoch": 0.1869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8829438808093484e-05,
-      "loss": 1.0613,
+      "learning_rate": 1.873642120553389e-05,
+      "loss": 0.9413,
       "step": 6587
     },
     {
-      "epoch": 0.18093433303122683,
+      "epoch": 0.1869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.8829021164480506e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.8735973973688894e-05,
+      "loss": 0.9649,
       "step": 6588
     },
     {
-      "epoch": 0.18096179725907005,
+      "epoch": 0.1869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.8828603451008893e-05,
-      "loss": 1.1323,
+      "learning_rate": 1.8735526668050733e-05,
+      "loss": 1.0052,
       "step": 6589
     },
     {
-      "epoch": 0.1809892614869133,
+      "epoch": 0.18700340522133937,
       "grad_norm": 0.0,
-      "learning_rate": 1.8828185667681946e-05,
-      "loss": 0.814,
+      "learning_rate": 1.8735079288623182e-05,
+      "loss": 0.9961,
       "step": 6590
     },
     {
-      "epoch": 0.18101672571475652,
+      "epoch": 0.1870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.8827767814502978e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.8734631835410025e-05,
+      "loss": 0.9242,
       "step": 6591
     },
     {
-      "epoch": 0.18104418994259977,
+      "epoch": 0.18706015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 1.8827349891475288e-05,
-      "loss": 1.044,
+      "learning_rate": 1.8734184308415044e-05,
+      "loss": 1.0549,
       "step": 6592
     },
     {
-      "epoch": 0.181071654170443,
+      "epoch": 0.18708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.882693189860219e-05,
-      "loss": 1.0896,
+      "learning_rate": 1.8733736707642012e-05,
+      "loss": 1.0238,
       "step": 6593
     },
     {
-      "epoch": 0.18109911839828624,
+      "epoch": 0.18711691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.8826513835886985e-05,
-      "loss": 1.1302,
+      "learning_rate": 1.8733289033094715e-05,
+      "loss": 0.9416,
       "step": 6594
     },
     {
-      "epoch": 0.18112658262612946,
+      "epoch": 0.18714528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.882609570333298e-05,
-      "loss": 1.0091,
+      "learning_rate": 1.8732841284776932e-05,
+      "loss": 1.0197,
       "step": 6595
     },
     {
-      "epoch": 0.1811540468539727,
+      "epoch": 0.18717366628830873,
       "grad_norm": 0.0,
-      "learning_rate": 1.8825677500943493e-05,
-      "loss": 1.0762,
+      "learning_rate": 1.8732393462692447e-05,
+      "loss": 1.0037,
       "step": 6596
     },
     {
-      "epoch": 0.18118151108181593,
+      "epoch": 0.18720204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.8825259228721825e-05,
-      "loss": 0.947,
+      "learning_rate": 1.8731945566845042e-05,
+      "loss": 1.0752,
       "step": 6597
     },
     {
-      "epoch": 0.18120897530965918,
+      "epoch": 0.18723041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.8824840886671286e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.87314975972385e-05,
+      "loss": 1.0817,
       "step": 6598
     },
     {
-      "epoch": 0.1812364395375024,
+      "epoch": 0.18725879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.882442247479519e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.8731049553876604e-05,
+      "loss": 1.0675,
       "step": 6599
     },
     {
-      "epoch": 0.18126390376534562,
+      "epoch": 0.1872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.8824003993096843e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.873060143676314e-05,
+      "loss": 1.0848,
       "step": 6600
     },
     {
-      "epoch": 0.18129136799318887,
+      "epoch": 0.1873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.8823585441579556e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.8730153245901895e-05,
+      "loss": 0.9826,
       "step": 6601
     },
     {
-      "epoch": 0.1813188322210321,
+      "epoch": 0.1873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.882316682024665e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.8729704981296654e-05,
+      "loss": 1.0535,
       "step": 6602
     },
     {
-      "epoch": 0.18134629644887534,
+      "epoch": 0.187372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.8822748129101424e-05,
-      "loss": 1.0525,
+      "learning_rate": 1.87292566429512e-05,
+      "loss": 1.0521,
       "step": 6603
     },
     {
-      "epoch": 0.18137376067671857,
+      "epoch": 0.18740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.8822329368147197e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.872880823086932e-05,
+      "loss": 1.0164,
       "step": 6604
     },
     {
-      "epoch": 0.18140122490456181,
+      "epoch": 0.18742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.8821910537387288e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.8728359745054808e-05,
+      "loss": 1.0288,
       "step": 6605
     },
     {
-      "epoch": 0.18142868913240504,
+      "epoch": 0.18745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.8821491636825002e-05,
-      "loss": 1.044,
+      "learning_rate": 1.8727911185511448e-05,
+      "loss": 0.9577,
       "step": 6606
     },
     {
-      "epoch": 0.18145615336024828,
+      "epoch": 0.18748581157775254,
       "grad_norm": 0.0,
-      "learning_rate": 1.882107266646366e-05,
-      "loss": 1.0699,
+      "learning_rate": 1.8727462552243034e-05,
+      "loss": 1.0123,
       "step": 6607
     },
     {
-      "epoch": 0.1814836175880915,
+      "epoch": 0.18751418842224746,
       "grad_norm": 0.0,
-      "learning_rate": 1.8820653626306572e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.8727013845253346e-05,
+      "loss": 0.9388,
       "step": 6608
     },
     {
-      "epoch": 0.18151108181593475,
+      "epoch": 0.18754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.882023451635705e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.8726565064546183e-05,
+      "loss": 0.8242,
       "step": 6609
     },
     {
-      "epoch": 0.18153854604377798,
+      "epoch": 0.18757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8819815336618425e-05,
-      "loss": 1.006,
+      "learning_rate": 1.872611621012533e-05,
+      "loss": 1.0027,
       "step": 6610
     },
     {
-      "epoch": 0.18156601027162123,
+      "epoch": 0.18759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.8819396087094e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.8725667281994584e-05,
+      "loss": 0.9956,
       "step": 6611
     },
     {
-      "epoch": 0.18159347449946445,
+      "epoch": 0.187627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.88189767677871e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.8725218280157734e-05,
+      "loss": 1.0591,
       "step": 6612
     },
     {
-      "epoch": 0.18162093872730767,
+      "epoch": 0.1876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.8818557378701037e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.872476920461857e-05,
+      "loss": 1.114,
       "step": 6613
     },
     {
-      "epoch": 0.18164840295515092,
+      "epoch": 0.1876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.8818137919839132e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.8724320055380892e-05,
+      "loss": 0.9719,
       "step": 6614
     },
     {
-      "epoch": 0.18167586718299414,
+      "epoch": 0.1877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.8817718391204705e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.8723870832448487e-05,
+      "loss": 0.9687,
       "step": 6615
     },
     {
-      "epoch": 0.1817033314108374,
+      "epoch": 0.18774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.8817298792801075e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.8723421535825153e-05,
+      "loss": 1.0443,
       "step": 6616
     },
     {
-      "epoch": 0.1817307956386806,
+      "epoch": 0.18776958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.881687912463156e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.872297216551469e-05,
+      "loss": 0.9225,
       "step": 6617
     },
     {
-      "epoch": 0.18175825986652386,
+      "epoch": 0.18779795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 1.881645938669948e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.8722522721520885e-05,
+      "loss": 1.0028,
       "step": 6618
     },
     {
-      "epoch": 0.18178572409436708,
+      "epoch": 0.18782633371169127,
       "grad_norm": 0.0,
-      "learning_rate": 1.8816039579008162e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.872207320384754e-05,
+      "loss": 1.0437,
       "step": 6619
     },
     {
-      "epoch": 0.18181318832221033,
+      "epoch": 0.18785471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.8815619701560922e-05,
-      "loss": 1.0894,
+      "learning_rate": 1.872162361249845e-05,
+      "loss": 0.9514,
       "step": 6620
     },
     {
-      "epoch": 0.18184065255005355,
+      "epoch": 0.18788308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.8815199754361085e-05,
-      "loss": 0.898,
+      "learning_rate": 1.872117394747741e-05,
+      "loss": 0.9931,
       "step": 6621
     },
     {
-      "epoch": 0.1818681167778968,
+      "epoch": 0.18791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.8814779737411973e-05,
-      "loss": 1.1121,
+      "learning_rate": 1.8720724208788225e-05,
+      "loss": 1.0286,
       "step": 6622
     },
     {
-      "epoch": 0.18189558100574002,
+      "epoch": 0.18793984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 1.881435965071691e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.872027439643469e-05,
+      "loss": 1.0099,
       "step": 6623
     },
     {
-      "epoch": 0.18192304523358327,
+      "epoch": 0.1879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.8813939494279215e-05,
-      "loss": 1.033,
+      "learning_rate": 1.8719824510420606e-05,
+      "loss": 1.104,
       "step": 6624
     },
     {
-      "epoch": 0.1819505094614265,
+      "epoch": 0.18799659477866063,
       "grad_norm": 0.0,
-      "learning_rate": 1.8813519268102222e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.8719374550749768e-05,
+      "loss": 0.9702,
       "step": 6625
     },
     {
-      "epoch": 0.18197797368926974,
+      "epoch": 0.1880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.8813098972189247e-05,
-      "loss": 1.0546,
+      "learning_rate": 1.8718924517425987e-05,
+      "loss": 1.0352,
       "step": 6626
     },
     {
-      "epoch": 0.18200543791711296,
+      "epoch": 0.1880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.881267860654362e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.8718474410453056e-05,
+      "loss": 0.994,
       "step": 6627
     },
     {
-      "epoch": 0.18203290214495618,
+      "epoch": 0.1880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.8812258171168663e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.871802422983478e-05,
+      "loss": 1.0038,
       "step": 6628
     },
     {
-      "epoch": 0.18206036637279943,
+      "epoch": 0.18811010215664017,
       "grad_norm": 0.0,
-      "learning_rate": 1.8811837666067712e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.8717573975574962e-05,
+      "loss": 0.994,
       "step": 6629
     },
     {
-      "epoch": 0.18208783060064265,
+      "epoch": 0.18813847900113506,
       "grad_norm": 0.0,
-      "learning_rate": 1.881141709124408e-05,
-      "loss": 1.0957,
+      "learning_rate": 1.8717123647677404e-05,
+      "loss": 0.9963,
       "step": 6630
     },
     {
-      "epoch": 0.1821152948284859,
+      "epoch": 0.18816685584562998,
       "grad_norm": 0.0,
-      "learning_rate": 1.881099644670111e-05,
-      "loss": 1.0804,
+      "learning_rate": 1.871667324614591e-05,
+      "loss": 1.0499,
       "step": 6631
     },
     {
-      "epoch": 0.18214275905632912,
+      "epoch": 0.18819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.881057573244212e-05,
-      "loss": 0.9124,
+      "learning_rate": 1.871622277098429e-05,
+      "loss": 0.9779,
       "step": 6632
     },
     {
-      "epoch": 0.18217022328417237,
+      "epoch": 0.18822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.8810154948470444e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.8715772222196337e-05,
+      "loss": 0.9854,
       "step": 6633
     },
     {
-      "epoch": 0.1821976875120156,
+      "epoch": 0.18825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.8809734094789407e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.871532159978587e-05,
+      "loss": 1.1144,
       "step": 6634
     },
     {
-      "epoch": 0.18222515173985884,
+      "epoch": 0.18828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.880931317140234e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.8714870903756684e-05,
+      "loss": 0.8969,
       "step": 6635
     },
     {
-      "epoch": 0.18225261596770206,
+      "epoch": 0.1883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.880889217831258e-05,
-      "loss": 1.1302,
+      "learning_rate": 1.87144201341126e-05,
+      "loss": 1.0641,
       "step": 6636
     },
     {
-      "epoch": 0.1822800801955453,
+      "epoch": 0.18833711691259933,
       "grad_norm": 0.0,
-      "learning_rate": 1.880847111552345e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.8713969290857412e-05,
+      "loss": 1.0877,
       "step": 6637
     },
     {
-      "epoch": 0.18230754442338853,
+      "epoch": 0.18836549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 1.8808049983038284e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.8713518373994932e-05,
+      "loss": 1.0525,
       "step": 6638
     },
     {
-      "epoch": 0.18233500865123178,
+      "epoch": 0.1883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.8807628780860415e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.8713067383528975e-05,
+      "loss": 0.9858,
       "step": 6639
     },
     {
-      "epoch": 0.182362472879075,
+      "epoch": 0.188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.880720750899317e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.8712616319463343e-05,
+      "loss": 1.0497,
       "step": 6640
     },
     {
-      "epoch": 0.18238993710691823,
+      "epoch": 0.18845062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 1.8806786167439897e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.871216518180185e-05,
+      "loss": 1.0317,
       "step": 6641
     },
     {
-      "epoch": 0.18241740133476148,
+      "epoch": 0.1884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8806364756203915e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.871171397054831e-05,
+      "loss": 0.9879,
       "step": 6642
     },
     {
-      "epoch": 0.1824448655626047,
+      "epoch": 0.18850737797956868,
       "grad_norm": 0.0,
-      "learning_rate": 1.8805943275288564e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.8711262685706525e-05,
+      "loss": 1.0005,
       "step": 6643
     },
     {
-      "epoch": 0.18247232979044795,
+      "epoch": 0.18853575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 1.880552172469718e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.8710811327280314e-05,
+      "loss": 0.9255,
       "step": 6644
     },
     {
-      "epoch": 0.18249979401829117,
+      "epoch": 0.18856413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.8805100104433098e-05,
-      "loss": 1.0643,
+      "learning_rate": 1.8710359895273487e-05,
+      "loss": 1.0396,
       "step": 6645
     },
     {
-      "epoch": 0.18252725824613442,
+      "epoch": 0.18859250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.880467841449965e-05,
-      "loss": 0.9671,
+      "learning_rate": 1.870990838968986e-05,
+      "loss": 1.0442,
       "step": 6646
     },
     {
-      "epoch": 0.18255472247397764,
+      "epoch": 0.18862088535754823,
       "grad_norm": 0.0,
-      "learning_rate": 1.880425665490018e-05,
-      "loss": 1.106,
+      "learning_rate": 1.8709456810533248e-05,
+      "loss": 0.8489,
       "step": 6647
     },
     {
-      "epoch": 0.1825821867018209,
+      "epoch": 0.18864926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 1.8803834825638017e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.870900515780746e-05,
+      "loss": 1.0615,
       "step": 6648
     },
     {
-      "epoch": 0.1826096509296641,
+      "epoch": 0.18867763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.8803412926716508e-05,
-      "loss": 1.0225,
+      "learning_rate": 1.870855343151631e-05,
+      "loss": 1.0364,
       "step": 6649
     },
     {
-      "epoch": 0.18263711515750736,
+      "epoch": 0.18870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.880299095813898e-05,
-      "loss": 1.0675,
+      "learning_rate": 1.8708101631663623e-05,
+      "loss": 0.9388,
       "step": 6650
     },
     {
-      "epoch": 0.18266457938535058,
+      "epoch": 0.1887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.880256891990878e-05,
-      "loss": 1.0816,
+      "learning_rate": 1.870764975825321e-05,
+      "loss": 1.0686,
       "step": 6651
     },
     {
-      "epoch": 0.18269204361319383,
+      "epoch": 0.1887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8802146812029246e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.8707197811288887e-05,
+      "loss": 1.065,
       "step": 6652
     },
     {
-      "epoch": 0.18271950784103705,
+      "epoch": 0.18879114642451758,
       "grad_norm": 0.0,
-      "learning_rate": 1.8801724634503716e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.870674579077447e-05,
+      "loss": 0.9559,
       "step": 6653
     },
     {
-      "epoch": 0.18274697206888027,
+      "epoch": 0.1888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.880130238733553e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.8706293696713783e-05,
+      "loss": 1.0455,
       "step": 6654
     },
     {
-      "epoch": 0.18277443629672352,
+      "epoch": 0.18884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8800880070528032e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.8705841529110642e-05,
+      "loss": 0.9985,
       "step": 6655
     },
     {
-      "epoch": 0.18280190052456674,
+      "epoch": 0.18887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8800457684084557e-05,
-      "loss": 0.9,
+      "learning_rate": 1.8705389287968866e-05,
+      "loss": 0.9385,
       "step": 6656
     },
     {
-      "epoch": 0.18282936475241,
+      "epoch": 0.18890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.8800035228008458e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.8704936973292274e-05,
+      "loss": 0.9847,
       "step": 6657
     },
     {
-      "epoch": 0.1828568289802532,
+      "epoch": 0.18893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.879961270230307e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.8704484585084688e-05,
+      "loss": 1.0077,
       "step": 6658
     },
     {
-      "epoch": 0.18288429320809646,
+      "epoch": 0.18896140749148696,
       "grad_norm": 0.0,
-      "learning_rate": 1.8799190106971734e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.8704032123349932e-05,
+      "loss": 0.991,
       "step": 6659
     },
     {
-      "epoch": 0.18291175743593968,
+      "epoch": 0.18898978433598185,
       "grad_norm": 0.0,
-      "learning_rate": 1.87987674420178e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.8703579588091822e-05,
+      "loss": 1.1086,
       "step": 6660
     },
     {
-      "epoch": 0.18293922166378293,
+      "epoch": 0.18901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.879834470744461e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.870312697931419e-05,
+      "loss": 1.2551,
       "step": 6661
     },
     {
-      "epoch": 0.18296668589162615,
+      "epoch": 0.18904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.8797921903255507e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.8702674297020843e-05,
+      "loss": 1.0474,
       "step": 6662
     },
     {
-      "epoch": 0.1829941501194694,
+      "epoch": 0.1890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.8797499029453837e-05,
-      "loss": 0.9864,
+      "learning_rate": 1.8702221541215624e-05,
+      "loss": 0.9999,
       "step": 6663
     },
     {
-      "epoch": 0.18302161434731262,
+      "epoch": 0.1891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.8797076086042953e-05,
-      "loss": 1.0654,
+      "learning_rate": 1.8701768711902342e-05,
+      "loss": 1.097,
       "step": 6664
     },
     {
-      "epoch": 0.18304907857515587,
+      "epoch": 0.1891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.879665307302619e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.8701315809084835e-05,
+      "loss": 0.9676,
       "step": 6665
     },
     {
-      "epoch": 0.1830765428029991,
+      "epoch": 0.1891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.87962299904069e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.870086283276692e-05,
+      "loss": 0.9745,
       "step": 6666
     },
     {
-      "epoch": 0.18310400703084231,
+      "epoch": 0.1891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.8795806838188433e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.870040978295242e-05,
+      "loss": 1.0901,
       "step": 6667
     },
     {
-      "epoch": 0.18313147125868556,
+      "epoch": 0.18921679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.8795383616374136e-05,
-      "loss": 1.0629,
+      "learning_rate": 1.8699956659645173e-05,
+      "loss": 1.0785,
       "step": 6668
     },
     {
-      "epoch": 0.18315893548652878,
+      "epoch": 0.18924517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.8794960324967354e-05,
-      "loss": 0.9537,
+      "learning_rate": 1.8699503462849002e-05,
+      "loss": 0.776,
       "step": 6669
     },
     {
-      "epoch": 0.18318639971437203,
+      "epoch": 0.18927355278093075,
       "grad_norm": 0.0,
-      "learning_rate": 1.8794536963971442e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.869905019256773e-05,
+      "loss": 0.9168,
       "step": 6670
     },
     {
-      "epoch": 0.18321386394221525,
+      "epoch": 0.18930192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.8794113533389744e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.8698596848805194e-05,
+      "loss": 0.9562,
       "step": 6671
     },
     {
-      "epoch": 0.1832413281700585,
+      "epoch": 0.18933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.8793690033225615e-05,
-      "loss": 1.0108,
+      "learning_rate": 1.8698143431565215e-05,
+      "loss": 1.0419,
       "step": 6672
     },
     {
-      "epoch": 0.18326879239790173,
+      "epoch": 0.18935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.8793266463482407e-05,
-      "loss": 1.0324,
+      "learning_rate": 1.8697689940851633e-05,
+      "loss": 0.9139,
       "step": 6673
     },
     {
-      "epoch": 0.18329625662574497,
+      "epoch": 0.18938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.8792842824163463e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.869723637666827e-05,
+      "loss": 1.0325,
       "step": 6674
     },
     {
-      "epoch": 0.1833237208535882,
+      "epoch": 0.1894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.8792419115272143e-05,
-      "loss": 1.1329,
+      "learning_rate": 1.8696782739018958e-05,
+      "loss": 1.057,
       "step": 6675
     },
     {
-      "epoch": 0.18335118508143144,
+      "epoch": 0.1894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.8791995336811798e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.8696329027907535e-05,
+      "loss": 1.0143,
       "step": 6676
     },
     {
-      "epoch": 0.18337864930927467,
+      "epoch": 0.18947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.879157148878578e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.8695875243337827e-05,
+      "loss": 1.001,
       "step": 6677
     },
     {
-      "epoch": 0.18340611353711792,
+      "epoch": 0.1895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.8791147571197443e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.8695421385313673e-05,
+      "loss": 0.9438,
       "step": 6678
     },
     {
-      "epoch": 0.18343357776496114,
+      "epoch": 0.1895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.8790723584050138e-05,
-      "loss": 1.0254,
+      "learning_rate": 1.86949674538389e-05,
+      "loss": 1.0173,
       "step": 6679
     },
     {
-      "epoch": 0.18346104199280439,
+      "epoch": 0.18955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.8790299527347225e-05,
-      "loss": 0.9349,
+      "learning_rate": 1.869451344891735e-05,
+      "loss": 1.1224,
       "step": 6680
     },
     {
-      "epoch": 0.1834885062206476,
+      "epoch": 0.18958569807037456,
       "grad_norm": 0.0,
-      "learning_rate": 1.878987540109206e-05,
-      "loss": 1.0273,
+      "learning_rate": 1.869405937055285e-05,
+      "loss": 1.0324,
       "step": 6681
     },
     {
-      "epoch": 0.18351597044849083,
+      "epoch": 0.18961407491486948,
       "grad_norm": 0.0,
-      "learning_rate": 1.878945120528799e-05,
-      "loss": 1.0783,
+      "learning_rate": 1.869360521874924e-05,
+      "loss": 1.0217,
       "step": 6682
     },
     {
-      "epoch": 0.18354343467633408,
+      "epoch": 0.18964245175936437,
       "grad_norm": 0.0,
-      "learning_rate": 1.878902693993838e-05,
-      "loss": 1.1301,
+      "learning_rate": 1.869315099351036e-05,
+      "loss": 0.9627,
       "step": 6683
     },
     {
-      "epoch": 0.1835708989041773,
+      "epoch": 0.18967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8788602605046585e-05,
-      "loss": 0.9545,
+      "learning_rate": 1.869269669484004e-05,
+      "loss": 0.9624,
       "step": 6684
     },
     {
-      "epoch": 0.18359836313202055,
+      "epoch": 0.18969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.8788178200615963e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.869224232274212e-05,
+      "loss": 0.9735,
       "step": 6685
     },
     {
-      "epoch": 0.18362582735986377,
+      "epoch": 0.18972758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.8787753726649868e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.869178787722044e-05,
+      "loss": 0.9991,
       "step": 6686
     },
     {
-      "epoch": 0.18365329158770702,
+      "epoch": 0.18975595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 1.8787329183151662e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.8691333358278835e-05,
+      "loss": 1.0114,
       "step": 6687
     },
     {
-      "epoch": 0.18368075581555024,
+      "epoch": 0.18978433598183883,
       "grad_norm": 0.0,
-      "learning_rate": 1.8786904570124706e-05,
-      "loss": 1.0737,
+      "learning_rate": 1.869087876592115e-05,
+      "loss": 0.9047,
       "step": 6688
     },
     {
-      "epoch": 0.1837082200433935,
+      "epoch": 0.18981271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 1.8786479887572354e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.869042410015122e-05,
+      "loss": 0.9004,
       "step": 6689
     },
     {
-      "epoch": 0.1837356842712367,
+      "epoch": 0.1898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.8786055135497973e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.8689969360972883e-05,
+      "loss": 0.9784,
       "step": 6690
     },
     {
-      "epoch": 0.18376314849907996,
+      "epoch": 0.1898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.878563031390492e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.8689514548389987e-05,
+      "loss": 0.9589,
       "step": 6691
     },
     {
-      "epoch": 0.18379061272692318,
+      "epoch": 0.18989784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 1.8785205422796555e-05,
-      "loss": 1.0476,
+      "learning_rate": 1.8689059662406373e-05,
+      "loss": 0.8813,
       "step": 6692
     },
     {
-      "epoch": 0.18381807695476643,
+      "epoch": 0.18992622020431327,
       "grad_norm": 0.0,
-      "learning_rate": 1.878478046217624e-05,
-      "loss": 1.1106,
+      "learning_rate": 1.8688604703025878e-05,
+      "loss": 1.0809,
       "step": 6693
     },
     {
-      "epoch": 0.18384554118260965,
+      "epoch": 0.18995459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 1.8784355432047342e-05,
-      "loss": 1.1318,
+      "learning_rate": 1.8688149670252352e-05,
+      "loss": 0.9464,
       "step": 6694
     },
     {
-      "epoch": 0.18387300541045287,
+      "epoch": 0.18998297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 1.8783930332413222e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.8687694564089632e-05,
+      "loss": 0.982,
       "step": 6695
     },
     {
-      "epoch": 0.18390046963829612,
+      "epoch": 0.19001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.8783505163277236e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.8687239384541563e-05,
+      "loss": 1.07,
       "step": 6696
     },
     {
-      "epoch": 0.18392793386613934,
+      "epoch": 0.19003972758229284,
       "grad_norm": 0.0,
-      "learning_rate": 1.8783079924642764e-05,
-      "loss": 1.0273,
+      "learning_rate": 1.8686784131611998e-05,
+      "loss": 1.0825,
       "step": 6697
     },
     {
-      "epoch": 0.1839553980939826,
+      "epoch": 0.19006810442678773,
       "grad_norm": 0.0,
-      "learning_rate": 1.8782654616513154e-05,
-      "loss": 0.9766,
+      "learning_rate": 1.8686328805304775e-05,
+      "loss": 1.1033,
       "step": 6698
     },
     {
-      "epoch": 0.1839828623218258,
+      "epoch": 0.19009648127128265,
       "grad_norm": 0.0,
-      "learning_rate": 1.8782229238891782e-05,
-      "loss": 1.1065,
+      "learning_rate": 1.8685873405623742e-05,
+      "loss": 0.9903,
       "step": 6699
     },
     {
-      "epoch": 0.18401032654966906,
+      "epoch": 0.19012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.878180379178201e-05,
-      "loss": 1.0999,
+      "learning_rate": 1.8685417932572747e-05,
+      "loss": 1.0054,
       "step": 6700
     },
     {
-      "epoch": 0.18403779077751228,
+      "epoch": 0.19015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.8781378275187204e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.8684962386155633e-05,
+      "loss": 1.0027,
       "step": 6701
     },
     {
-      "epoch": 0.18406525500535553,
+      "epoch": 0.1901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.8780952689110733e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.8684506766376254e-05,
+      "loss": 1.0344,
       "step": 6702
     },
     {
-      "epoch": 0.18409271923319875,
+      "epoch": 0.1902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.8780527033555964e-05,
-      "loss": 1.124,
+      "learning_rate": 1.8684051073238455e-05,
+      "loss": 1.0887,
       "step": 6703
     },
     {
-      "epoch": 0.184120183461042,
+      "epoch": 0.19023836549375708,
       "grad_norm": 0.0,
-      "learning_rate": 1.8780101308526258e-05,
-      "loss": 1.057,
+      "learning_rate": 1.868359530674609e-05,
+      "loss": 1.0029,
       "step": 6704
     },
     {
-      "epoch": 0.18414764768888522,
+      "epoch": 0.190266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8779675514024994e-05,
-      "loss": 1.0071,
+      "learning_rate": 1.8683139466903e-05,
+      "loss": 0.998,
       "step": 6705
     },
     {
-      "epoch": 0.18417511191672847,
+      "epoch": 0.19029511918274689,
       "grad_norm": 0.0,
-      "learning_rate": 1.877924965005554e-05,
-      "loss": 1.1813,
+      "learning_rate": 1.868268355371304e-05,
+      "loss": 0.9958,
       "step": 6706
     },
     {
-      "epoch": 0.1842025761445717,
+      "epoch": 0.19032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.8778823716621254e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.8682227567180064e-05,
+      "loss": 0.9843,
       "step": 6707
     },
     {
-      "epoch": 0.18423004037241492,
+      "epoch": 0.19035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.877839771372552e-05,
-      "loss": 1.0481,
+      "learning_rate": 1.8681771507307922e-05,
+      "loss": 0.9983,
       "step": 6708
     },
     {
-      "epoch": 0.18425750460025817,
+      "epoch": 0.19038024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.87779716413717e-05,
-      "loss": 1.1167,
+      "learning_rate": 1.8681315374100468e-05,
+      "loss": 1.0114,
       "step": 6709
     },
     {
-      "epoch": 0.1842849688281014,
+      "epoch": 0.19040862656072643,
       "grad_norm": 0.0,
-      "learning_rate": 1.877754549956317e-05,
-      "loss": 1.1218,
+      "learning_rate": 1.868085916756155e-05,
+      "loss": 1.0055,
       "step": 6710
     },
     {
-      "epoch": 0.18431243305594464,
+      "epoch": 0.19043700340522135,
       "grad_norm": 0.0,
-      "learning_rate": 1.87771192883033e-05,
-      "loss": 1.0116,
+      "learning_rate": 1.8680402887695025e-05,
+      "loss": 0.9542,
       "step": 6711
     },
     {
-      "epoch": 0.18433989728378786,
+      "epoch": 0.19046538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.877669300759546e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.8679946534504742e-05,
+      "loss": 1.0431,
       "step": 6712
     },
     {
-      "epoch": 0.1843673615116311,
+      "epoch": 0.19049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.8776266657443022e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.8679490107994565e-05,
+      "loss": 1.0338,
       "step": 6713
     },
     {
-      "epoch": 0.18439482573947433,
+      "epoch": 0.190522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.877584023784937e-05,
-      "loss": 1.1299,
+      "learning_rate": 1.8679033608168344e-05,
+      "loss": 1.0225,
       "step": 6714
     },
     {
-      "epoch": 0.18442228996731758,
+      "epoch": 0.1905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.8775413748817866e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.867857703502994e-05,
+      "loss": 1.0086,
       "step": 6715
     },
     {
-      "epoch": 0.1844497541951608,
+      "epoch": 0.1905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.877498719035189e-05,
-      "loss": 1.0687,
+      "learning_rate": 1.86781203885832e-05,
+      "loss": 1.0062,
       "step": 6716
     },
     {
-      "epoch": 0.18447721842300405,
+      "epoch": 0.1906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.877456056245482e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.867766366883199e-05,
+      "loss": 0.9231,
       "step": 6717
     },
     {
-      "epoch": 0.18450468265084727,
+      "epoch": 0.1906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.8774133865130024e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.8677206875780162e-05,
+      "loss": 1.0038,
       "step": 6718
     },
     {
-      "epoch": 0.18453214687869052,
+      "epoch": 0.19066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.877370709838088e-05,
-      "loss": 0.9546,
+      "learning_rate": 1.867675000943158e-05,
+      "loss": 0.8687,
       "step": 6719
     },
     {
-      "epoch": 0.18455961110653374,
+      "epoch": 0.19069239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.877328026221077e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.8676293069790097e-05,
+      "loss": 0.9626,
       "step": 6720
     },
     {
-      "epoch": 0.184587075334377,
+      "epoch": 0.19072077185017025,
       "grad_norm": 0.0,
-      "learning_rate": 1.8772853356623066e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.867583605685958e-05,
+      "loss": 0.9021,
       "step": 6721
     },
     {
-      "epoch": 0.1846145395622202,
+      "epoch": 0.19074914869466517,
       "grad_norm": 0.0,
-      "learning_rate": 1.877242638162115e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.8675378970643888e-05,
+      "loss": 1.0056,
       "step": 6722
     },
     {
-      "epoch": 0.18464200379006343,
+      "epoch": 0.19077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.87719993372084e-05,
-      "loss": 0.981,
+      "learning_rate": 1.867492181114688e-05,
+      "loss": 0.8785,
       "step": 6723
     },
     {
-      "epoch": 0.18466946801790668,
+      "epoch": 0.19080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.877157222338819e-05,
-      "loss": 1.0859,
+      "learning_rate": 1.867446457837241e-05,
+      "loss": 1.0178,
       "step": 6724
     },
     {
-      "epoch": 0.1846969322457499,
+      "epoch": 0.19083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8771145040163904e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.8674007272324352e-05,
+      "loss": 1.0621,
       "step": 6725
     },
     {
-      "epoch": 0.18472439647359315,
+      "epoch": 0.19086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.877071778753892e-05,
-      "loss": 1.1173,
+      "learning_rate": 1.8673549893006566e-05,
+      "loss": 1.0023,
       "step": 6726
     },
     {
-      "epoch": 0.18475186070143637,
+      "epoch": 0.1908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.877029046551662e-05,
-      "loss": 1.0303,
+      "learning_rate": 1.8673092440422916e-05,
+      "loss": 1.0432,
       "step": 6727
     },
     {
-      "epoch": 0.18477932492927962,
+      "epoch": 0.19091940976163452,
       "grad_norm": 0.0,
-      "learning_rate": 1.8769863074100387e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.867263491457726e-05,
+      "loss": 0.9655,
       "step": 6728
     },
     {
-      "epoch": 0.18480678915712284,
+      "epoch": 0.1909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.8769435613293598e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.8672177315473465e-05,
+      "loss": 0.9913,
       "step": 6729
     },
     {
-      "epoch": 0.1848342533849661,
+      "epoch": 0.1909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.876900808309964e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.8671719643115405e-05,
+      "loss": 1.0467,
       "step": 6730
     },
     {
-      "epoch": 0.1848617176128093,
+      "epoch": 0.19100454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.876858048352189e-05,
-      "loss": 0.8989,
+      "learning_rate": 1.8671261897506934e-05,
+      "loss": 1.1085,
       "step": 6731
     },
     {
-      "epoch": 0.18488918184065256,
+      "epoch": 0.19103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.8768152814563735e-05,
-      "loss": 1.041,
+      "learning_rate": 1.8670804078651927e-05,
+      "loss": 1.0432,
       "step": 6732
     },
     {
-      "epoch": 0.18491664606849578,
+      "epoch": 0.19106129398410895,
       "grad_norm": 0.0,
-      "learning_rate": 1.876772507622856e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.8670346186554245e-05,
+      "loss": 1.1676,
       "step": 6733
     },
     {
-      "epoch": 0.18494411029633903,
+      "epoch": 0.19108967082860387,
       "grad_norm": 0.0,
-      "learning_rate": 1.8767297268519745e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.866988822121776e-05,
+      "loss": 0.9896,
       "step": 6734
     },
     {
-      "epoch": 0.18497157452418225,
+      "epoch": 0.19111804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.8766869391440678e-05,
-      "loss": 1.054,
+      "learning_rate": 1.8669430182646345e-05,
+      "loss": 1.0304,
       "step": 6735
     },
     {
-      "epoch": 0.18499903875202547,
+      "epoch": 0.19114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.8766441444994748e-05,
-      "loss": 1.1508,
+      "learning_rate": 1.8668972070843857e-05,
+      "loss": 1.0446,
       "step": 6736
     },
     {
-      "epoch": 0.18502650297986872,
+      "epoch": 0.19117480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.8766013429185335e-05,
-      "loss": 1.099,
+      "learning_rate": 1.8668513885814176e-05,
+      "loss": 1.0247,
       "step": 6737
     },
     {
-      "epoch": 0.18505396720771194,
+      "epoch": 0.19120317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 1.876558534401583e-05,
-      "loss": 0.8519,
+      "learning_rate": 1.866805562756117e-05,
+      "loss": 1.0112,
       "step": 6738
     },
     {
-      "epoch": 0.1850814314355552,
+      "epoch": 0.19123155505107833,
       "grad_norm": 0.0,
-      "learning_rate": 1.8765157189489615e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.86675972960887e-05,
+      "loss": 0.9807,
       "step": 6739
     },
     {
-      "epoch": 0.18510889566339842,
+      "epoch": 0.19125993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.8764728965610082e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.8667138891400655e-05,
+      "loss": 1.1387,
       "step": 6740
     },
     {
-      "epoch": 0.18513635989124166,
+      "epoch": 0.1912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.876430067238062e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.8666680413500892e-05,
+      "loss": 1.0684,
       "step": 6741
     },
     {
-      "epoch": 0.18516382411908489,
+      "epoch": 0.191316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.876387230980462e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.8666221862393295e-05,
+      "loss": 0.8984,
       "step": 6742
     },
     {
-      "epoch": 0.18519128834692813,
+      "epoch": 0.19134506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 1.876344387788546e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.8665763238081726e-05,
+      "loss": 1.028,
       "step": 6743
     },
     {
-      "epoch": 0.18521875257477136,
+      "epoch": 0.19137343927355277,
       "grad_norm": 0.0,
-      "learning_rate": 1.876301537662654e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.8665304540570074e-05,
+      "loss": 0.9394,
       "step": 6744
     },
     {
-      "epoch": 0.1852462168026146,
+      "epoch": 0.19140181611804769,
       "grad_norm": 0.0,
-      "learning_rate": 1.8762586806031247e-05,
-      "loss": 1.0449,
+      "learning_rate": 1.8664845769862196e-05,
+      "loss": 0.9223,
       "step": 6745
     },
     {
-      "epoch": 0.18527368103045783,
+      "epoch": 0.19143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.8762158166102974e-05,
-      "loss": 1.049,
+      "learning_rate": 1.866438692596198e-05,
+      "loss": 0.9813,
       "step": 6746
     },
     {
-      "epoch": 0.18530114525830108,
+      "epoch": 0.19145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.8761729456845112e-05,
-      "loss": 1.007,
+      "learning_rate": 1.8663928008873303e-05,
+      "loss": 0.9902,
       "step": 6747
     },
     {
-      "epoch": 0.1853286094861443,
+      "epoch": 0.19148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.8761300678261053e-05,
-      "loss": 1.066,
+      "learning_rate": 1.8663469018600027e-05,
+      "loss": 0.9337,
       "step": 6748
     },
     {
-      "epoch": 0.18535607371398752,
+      "epoch": 0.19151532349602723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8760871830354187e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.8663009955146046e-05,
+      "loss": 0.9606,
       "step": 6749
     },
     {
-      "epoch": 0.18538353794183077,
+      "epoch": 0.19154370034052212,
       "grad_norm": 0.0,
-      "learning_rate": 1.8760442913127913e-05,
-      "loss": 1.0362,
+      "learning_rate": 1.8662550818515228e-05,
+      "loss": 1.0516,
       "step": 6750
     },
     {
-      "epoch": 0.185411002169674,
+      "epoch": 0.19157207718501704,
       "grad_norm": 0.0,
-      "learning_rate": 1.876001392658562e-05,
-      "loss": 0.9391,
+      "learning_rate": 1.8662091608711453e-05,
+      "loss": 0.9283,
       "step": 6751
     },
     {
-      "epoch": 0.18543846639751724,
+      "epoch": 0.19160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.8759584870730702e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.8661632325738607e-05,
+      "loss": 1.0308,
       "step": 6752
     },
     {
-      "epoch": 0.18546593062536046,
+      "epoch": 0.1916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.8759155745566554e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.8661172969600556e-05,
+      "loss": 1.0178,
       "step": 6753
     },
     {
-      "epoch": 0.1854933948532037,
+      "epoch": 0.1916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.8758726551096574e-05,
-      "loss": 0.9698,
+      "learning_rate": 1.866071354030119e-05,
+      "loss": 0.902,
       "step": 6754
     },
     {
-      "epoch": 0.18552085908104693,
+      "epoch": 0.19168558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.8758297287324158e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.866025403784439e-05,
+      "loss": 1.0668,
       "step": 6755
     },
     {
-      "epoch": 0.18554832330889018,
+      "epoch": 0.19171396140749147,
       "grad_norm": 0.0,
-      "learning_rate": 1.87578679542527e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.8659794462234032e-05,
+      "loss": 1.0519,
       "step": 6756
     },
     {
-      "epoch": 0.1855757875367334,
+      "epoch": 0.1917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.87574385518856e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.8659334813474e-05,
+      "loss": 0.9778,
       "step": 6757
     },
     {
-      "epoch": 0.18560325176457665,
+      "epoch": 0.19177071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 1.8757009080226253e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.865887509156818e-05,
+      "loss": 0.9418,
       "step": 6758
     },
     {
-      "epoch": 0.18563071599241987,
+      "epoch": 0.19179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.8756579539278056e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.865841529652045e-05,
+      "loss": 0.9653,
       "step": 6759
     },
     {
-      "epoch": 0.18565818022026312,
+      "epoch": 0.19182746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.875614992904441e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.86579554283347e-05,
+      "loss": 0.9507,
       "step": 6760
     },
     {
-      "epoch": 0.18568564444810634,
+      "epoch": 0.19185584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.875572024952872e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.865749548701481e-05,
+      "loss": 1.0344,
       "step": 6761
     },
     {
-      "epoch": 0.18571310867594956,
+      "epoch": 0.19188422247446085,
       "grad_norm": 0.0,
-      "learning_rate": 1.8755290500734375e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.8657035472564665e-05,
+      "loss": 1.0712,
       "step": 6762
     },
     {
-      "epoch": 0.1857405729037928,
+      "epoch": 0.19191259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 1.875486068266478e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.8656575384988152e-05,
+      "loss": 1.1493,
       "step": 6763
     },
     {
-      "epoch": 0.18576803713163603,
+      "epoch": 0.19194097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 1.8754430795323334e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.865611522428916e-05,
+      "loss": 0.9563,
       "step": 6764
     },
     {
-      "epoch": 0.18579550135947928,
+      "epoch": 0.19196935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 1.875400083871344e-05,
-      "loss": 1.0052,
+      "learning_rate": 1.8655654990471572e-05,
+      "loss": 0.9048,
       "step": 6765
     },
     {
-      "epoch": 0.1858229655873225,
+      "epoch": 0.1919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.8753570812838506e-05,
-      "loss": 1.1562,
+      "learning_rate": 1.865519468353928e-05,
+      "loss": 1.1686,
       "step": 6766
     },
     {
-      "epoch": 0.18585042981516575,
+      "epoch": 0.1920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.8753140717701924e-05,
-      "loss": 1.1071,
+      "learning_rate": 1.8654734303496166e-05,
+      "loss": 1.1316,
       "step": 6767
     },
     {
-      "epoch": 0.18587789404300897,
+      "epoch": 0.1920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.8752710553307107e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.865427385034612e-05,
+      "loss": 0.8995,
       "step": 6768
     },
     {
-      "epoch": 0.18590535827085222,
+      "epoch": 0.1920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.8752280319657446e-05,
-      "loss": 1.0398,
+      "learning_rate": 1.865381332409304e-05,
+      "loss": 1.0448,
       "step": 6769
     },
     {
-      "epoch": 0.18593282249869544,
+      "epoch": 0.19211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.875185001675636e-05,
-      "loss": 1.1066,
+      "learning_rate": 1.865335272474081e-05,
+      "loss": 1.0381,
       "step": 6770
     },
     {
-      "epoch": 0.1859602867265387,
+      "epoch": 0.19213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.8751419644607245e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.8652892052293314e-05,
+      "loss": 0.9945,
       "step": 6771
     },
     {
-      "epoch": 0.18598775095438191,
+      "epoch": 0.19216799091940975,
       "grad_norm": 0.0,
-      "learning_rate": 1.8750989203213504e-05,
-      "loss": 1.2135,
+      "learning_rate": 1.865243130675445e-05,
+      "loss": 0.9272,
       "step": 6772
     },
     {
-      "epoch": 0.18601521518222516,
+      "epoch": 0.19219636776390464,
       "grad_norm": 0.0,
-      "learning_rate": 1.875055869257855e-05,
-      "loss": 1.1378,
+      "learning_rate": 1.8651970488128115e-05,
+      "loss": 1.0496,
       "step": 6773
     },
     {
-      "epoch": 0.18604267941006838,
+      "epoch": 0.19222474460839956,
       "grad_norm": 0.0,
-      "learning_rate": 1.875012811270579e-05,
-      "loss": 1.0512,
+      "learning_rate": 1.86515095964182e-05,
+      "loss": 0.9249,
       "step": 6774
     },
     {
-      "epoch": 0.18607014363791163,
+      "epoch": 0.19225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.8749697463598618e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.8651048631628585e-05,
+      "loss": 1.0366,
       "step": 6775
     },
     {
-      "epoch": 0.18609760786575485,
+      "epoch": 0.19228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.8749266745260455e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.865058759376318e-05,
+      "loss": 1.1197,
       "step": 6776
     },
     {
-      "epoch": 0.18612507209359808,
+      "epoch": 0.19230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.8748835957694703e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.8650126482825867e-05,
+      "loss": 1.0396,
       "step": 6777
     },
     {
-      "epoch": 0.18615253632144133,
+      "epoch": 0.1923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.8748405100904774e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.8649665298820552e-05,
+      "loss": 0.9961,
       "step": 6778
     },
     {
-      "epoch": 0.18618000054928455,
+      "epoch": 0.19236662883087402,
       "grad_norm": 0.0,
-      "learning_rate": 1.874797417489407e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.8649204041751123e-05,
+      "loss": 1.0006,
       "step": 6779
     },
     {
-      "epoch": 0.1862074647771278,
+      "epoch": 0.1923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.8747543179666012e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.8648742711621478e-05,
+      "loss": 1.0533,
       "step": 6780
     },
     {
-      "epoch": 0.18623492900497102,
+      "epoch": 0.1924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.8747112115224e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.8648281308435515e-05,
+      "loss": 0.9802,
       "step": 6781
     },
     {
-      "epoch": 0.18626239323281427,
+      "epoch": 0.19245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.8746680981571452e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.864781983219713e-05,
+      "loss": 1.0457,
       "step": 6782
     },
     {
-      "epoch": 0.1862898574606575,
+      "epoch": 0.19248013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.8746249778711772e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.8647358282910226e-05,
+      "loss": 1.0112,
       "step": 6783
     },
     {
-      "epoch": 0.18631732168850074,
+      "epoch": 0.19250851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 1.874581850664838e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.8646896660578697e-05,
+      "loss": 1.166,
       "step": 6784
     },
     {
-      "epoch": 0.18634478591634396,
+      "epoch": 0.19253688989784337,
       "grad_norm": 0.0,
-      "learning_rate": 1.874538716538468e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.864643496520644e-05,
+      "loss": 0.9377,
       "step": 6785
     },
     {
-      "epoch": 0.1863722501441872,
+      "epoch": 0.19256526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.8744955754924095e-05,
-      "loss": 1.0672,
+      "learning_rate": 1.8645973196797362e-05,
+      "loss": 1.0085,
       "step": 6786
     },
     {
-      "epoch": 0.18639971437203043,
+      "epoch": 0.19259364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.8744524275270026e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.8645511355355356e-05,
+      "loss": 1.0344,
       "step": 6787
     },
     {
-      "epoch": 0.18642717859987368,
+      "epoch": 0.19262202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.8744092726425895e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.8645049440884328e-05,
+      "loss": 0.9811,
       "step": 6788
     },
     {
-      "epoch": 0.1864546428277169,
+      "epoch": 0.19265039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 1.8743661108395116e-05,
-      "loss": 1.1691,
+      "learning_rate": 1.864458745338818e-05,
+      "loss": 0.9897,
       "step": 6789
     },
     {
-      "epoch": 0.18648210705556012,
+      "epoch": 0.1926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.8743229421181105e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.864412539287081e-05,
+      "loss": 1.0432,
       "step": 6790
     },
     {
-      "epoch": 0.18650957128340337,
+      "epoch": 0.19270715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 1.8742797664787278e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.8643663259336126e-05,
+      "loss": 0.9573,
       "step": 6791
     },
     {
-      "epoch": 0.1865370355112466,
+      "epoch": 0.1927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.8742365839217042e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.8643201052788033e-05,
+      "loss": 1.0044,
       "step": 6792
     },
     {
-      "epoch": 0.18656449973908984,
+      "epoch": 0.1927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.8741933944473826e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.864273877323043e-05,
+      "loss": 0.9218,
       "step": 6793
     },
     {
-      "epoch": 0.18659196396693306,
+      "epoch": 0.19279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.8741501980561043e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.864227642066722e-05,
+      "loss": 1.0276,
       "step": 6794
     },
     {
-      "epoch": 0.1866194281947763,
+      "epoch": 0.19282065834279227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8741069947482108e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.8641813995102317e-05,
+      "loss": 1.0012,
       "step": 6795
     },
     {
-      "epoch": 0.18664689242261953,
+      "epoch": 0.19284903518728716,
       "grad_norm": 0.0,
-      "learning_rate": 1.8740637845240442e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.8641351496539625e-05,
+      "loss": 0.9935,
       "step": 6796
     },
     {
-      "epoch": 0.18667435665046278,
+      "epoch": 0.19287741203178208,
       "grad_norm": 0.0,
-      "learning_rate": 1.8740205673839464e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.8640888924983045e-05,
+      "loss": 0.935,
       "step": 6797
     },
     {
-      "epoch": 0.186701820878306,
+      "epoch": 0.19290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.8739773433282593e-05,
-      "loss": 1.062,
+      "learning_rate": 1.8640426280436488e-05,
+      "loss": 1.0186,
       "step": 6798
     },
     {
-      "epoch": 0.18672928510614925,
+      "epoch": 0.19293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.8739341123573248e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.863996356290386e-05,
+      "loss": 0.9248,
       "step": 6799
     },
     {
-      "epoch": 0.18675674933399247,
+      "epoch": 0.19296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.8738908744714853e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.8639500772389073e-05,
+      "loss": 1.0796,
       "step": 6800
     },
     {
-      "epoch": 0.18678421356183572,
+      "epoch": 0.19299091940976162,
       "grad_norm": 0.0,
-      "learning_rate": 1.8738476296710825e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.863903790889604e-05,
+      "loss": 1.1368,
       "step": 6801
     },
     {
-      "epoch": 0.18681167778967894,
+      "epoch": 0.19301929625425654,
       "grad_norm": 0.0,
-      "learning_rate": 1.873804377956459e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.8638574972428657e-05,
+      "loss": 1.0037,
       "step": 6802
     },
     {
-      "epoch": 0.18683914201752216,
+      "epoch": 0.19304767309875143,
       "grad_norm": 0.0,
-      "learning_rate": 1.8737611193279564e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.8638111962990847e-05,
+      "loss": 0.9857,
       "step": 6803
     },
     {
-      "epoch": 0.1868666062453654,
+      "epoch": 0.19307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.8737178537859176e-05,
-      "loss": 0.9436,
+      "learning_rate": 1.8637648880586515e-05,
+      "loss": 1.006,
       "step": 6804
     },
     {
-      "epoch": 0.18689407047320863,
+      "epoch": 0.1931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.8736745813306846e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.8637185725219578e-05,
+      "loss": 1.0735,
       "step": 6805
     },
     {
-      "epoch": 0.18692153470105188,
+      "epoch": 0.1931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.8736313019626002e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.8636722496893943e-05,
+      "loss": 1.0378,
       "step": 6806
     },
     {
-      "epoch": 0.1869489989288951,
+      "epoch": 0.19316118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.8735880156820063e-05,
-      "loss": 1.0918,
+      "learning_rate": 1.8636259195613526e-05,
+      "loss": 0.9966,
       "step": 6807
     },
     {
-      "epoch": 0.18697646315673835,
+      "epoch": 0.1931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.8735447224892454e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.8635795821382237e-05,
+      "loss": 1.0597,
       "step": 6808
     },
     {
-      "epoch": 0.18700392738458158,
+      "epoch": 0.19321793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.8735014223846604e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.8635332374203993e-05,
+      "loss": 0.995,
       "step": 6809
     },
     {
-      "epoch": 0.18703139161242482,
+      "epoch": 0.19324631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.873458115368594e-05,
-      "loss": 0.9942,
+      "learning_rate": 1.8634868854082707e-05,
+      "loss": 0.9232,
       "step": 6810
     },
     {
-      "epoch": 0.18705885584026805,
+      "epoch": 0.19327468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.8734148014413884e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.86344052610223e-05,
+      "loss": 1.043,
       "step": 6811
     },
     {
-      "epoch": 0.1870863200681113,
+      "epoch": 0.19330306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.8733714806033867e-05,
-      "loss": 1.1082,
+      "learning_rate": 1.863394159502668e-05,
+      "loss": 1.1074,
       "step": 6812
     },
     {
-      "epoch": 0.18711378429595452,
+      "epoch": 0.19333144154370033,
       "grad_norm": 0.0,
-      "learning_rate": 1.8733281528549315e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.863347785609977e-05,
+      "loss": 1.0879,
       "step": 6813
     },
     {
-      "epoch": 0.18714124852379777,
+      "epoch": 0.19335981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.8732848181963656e-05,
-      "loss": 1.073,
+      "learning_rate": 1.8633014044245482e-05,
+      "loss": 0.9993,
       "step": 6814
     },
     {
-      "epoch": 0.187168712751641,
+      "epoch": 0.19338819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.873241476628032e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.8632550159467734e-05,
+      "loss": 0.907,
       "step": 6815
     },
     {
-      "epoch": 0.18719617697948424,
+      "epoch": 0.19341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.8731981281502734e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.863208620177045e-05,
+      "loss": 1.0929,
       "step": 6816
     },
     {
-      "epoch": 0.18722364120732746,
+      "epoch": 0.1934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.8731547727634333e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.863162217115755e-05,
+      "loss": 0.943,
       "step": 6817
     },
     {
-      "epoch": 0.18725110543517068,
+      "epoch": 0.1934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.8731114104678543e-05,
-      "loss": 1.0849,
+      "learning_rate": 1.8631158067632942e-05,
+      "loss": 0.9904,
       "step": 6818
     },
     {
-      "epoch": 0.18727856966301393,
+      "epoch": 0.1935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.8730680412638794e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.8630693891200557e-05,
+      "loss": 0.9371,
       "step": 6819
     },
     {
-      "epoch": 0.18730603389085715,
+      "epoch": 0.1935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.873024665151852e-05,
-      "loss": 1.0324,
+      "learning_rate": 1.863022964186431e-05,
+      "loss": 1.0044,
       "step": 6820
     },
     {
-      "epoch": 0.1873334981187004,
+      "epoch": 0.19355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.8729812821321155e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.862976531962813e-05,
+      "loss": 1.0111,
       "step": 6821
     },
     {
-      "epoch": 0.18736096234654362,
+      "epoch": 0.19358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.872937892205013e-05,
-      "loss": 1.146,
+      "learning_rate": 1.862930092449593e-05,
+      "loss": 0.9605,
       "step": 6822
     },
     {
-      "epoch": 0.18738842657438687,
+      "epoch": 0.19361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.8728944953708873e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.862883645647164e-05,
+      "loss": 1.0231,
       "step": 6823
     },
     {
-      "epoch": 0.1874158908022301,
+      "epoch": 0.19364358683314414,
       "grad_norm": 0.0,
-      "learning_rate": 1.8728510916300828e-05,
-      "loss": 1.017,
+      "learning_rate": 1.8628371915559183e-05,
+      "loss": 1.0116,
       "step": 6824
     },
     {
-      "epoch": 0.18744335503007334,
+      "epoch": 0.19367196367763906,
       "grad_norm": 0.0,
-      "learning_rate": 1.872807680982942e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.8627907301762475e-05,
+      "loss": 0.9222,
       "step": 6825
     },
     {
-      "epoch": 0.18747081925791656,
+      "epoch": 0.19370034052213395,
       "grad_norm": 0.0,
-      "learning_rate": 1.872764263429809e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.862744261508545e-05,
+      "loss": 1.0031,
       "step": 6826
     },
     {
-      "epoch": 0.1874982834857598,
+      "epoch": 0.19372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.872720838971027e-05,
-      "loss": 1.0614,
+      "learning_rate": 1.862697785553203e-05,
+      "loss": 1.0138,
       "step": 6827
     },
     {
-      "epoch": 0.18752574771360303,
+      "epoch": 0.19375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.872677407606939e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.8626513023106135e-05,
+      "loss": 0.9584,
       "step": 6828
     },
     {
-      "epoch": 0.18755321194144628,
+      "epoch": 0.1937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.8726339693378898e-05,
-      "loss": 0.956,
+      "learning_rate": 1.86260481178117e-05,
+      "loss": 1.035,
       "step": 6829
     },
     {
-      "epoch": 0.1875806761692895,
+      "epoch": 0.1938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.8725905241642227e-05,
-      "loss": 0.9318,
+      "learning_rate": 1.862558313965265e-05,
+      "loss": 0.9189,
       "step": 6830
     },
     {
-      "epoch": 0.18760814039713272,
+      "epoch": 0.1938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.8725470720862816e-05,
-      "loss": 1.0931,
+      "learning_rate": 1.8625118088632913e-05,
+      "loss": 1.065,
       "step": 6831
     },
     {
-      "epoch": 0.18763560462497597,
+      "epoch": 0.1938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.8725036131044098e-05,
-      "loss": 1.1438,
+      "learning_rate": 1.8624652964756412e-05,
+      "loss": 1.1252,
       "step": 6832
     },
     {
-      "epoch": 0.1876630688528192,
+      "epoch": 0.19389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.872460147218951e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.862418776802708e-05,
+      "loss": 1.0343,
       "step": 6833
     },
     {
-      "epoch": 0.18769053308066244,
+      "epoch": 0.19392735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.87241667443025e-05,
-      "loss": 1.0935,
+      "learning_rate": 1.8623722498448846e-05,
+      "loss": 1.108,
       "step": 6834
     },
     {
-      "epoch": 0.18771799730850566,
+      "epoch": 0.19395573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 1.8723731947386507e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.862325715602564e-05,
+      "loss": 1.0638,
       "step": 6835
     },
     {
-      "epoch": 0.1877454615363489,
+      "epoch": 0.19398410896708285,
       "grad_norm": 0.0,
-      "learning_rate": 1.872329708144496e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.8622791740761398e-05,
+      "loss": 1.0112,
       "step": 6836
     },
     {
-      "epoch": 0.18777292576419213,
+      "epoch": 0.19401248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.8722862146481313e-05,
-      "loss": 0.922,
+      "learning_rate": 1.8622326252660042e-05,
+      "loss": 0.9652,
       "step": 6837
     },
     {
-      "epoch": 0.18780038999203538,
+      "epoch": 0.19404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.8722427142498998e-05,
-      "loss": 0.989,
+      "learning_rate": 1.862186069172551e-05,
+      "loss": 1.0245,
       "step": 6838
     },
     {
-      "epoch": 0.1878278542198786,
+      "epoch": 0.19406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.8721992069501462e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.862139505796173e-05,
+      "loss": 0.9472,
       "step": 6839
     },
     {
-      "epoch": 0.18785531844772185,
+      "epoch": 0.19409761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 1.872155692749215e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.8620929351372644e-05,
+      "loss": 0.9942,
       "step": 6840
     },
     {
-      "epoch": 0.18788278267556507,
+      "epoch": 0.1941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.8721121716474497e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.8620463571962173e-05,
+      "loss": 0.8878,
       "step": 6841
     },
     {
-      "epoch": 0.18791024690340832,
+      "epoch": 0.19415437003405223,
       "grad_norm": 0.0,
-      "learning_rate": 1.872068643645195e-05,
-      "loss": 1.045,
+      "learning_rate": 1.8619997719734266e-05,
+      "loss": 0.9386,
       "step": 6842
     },
     {
-      "epoch": 0.18793771113125154,
+      "epoch": 0.1941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.8720251087427956e-05,
-      "loss": 0.954,
+      "learning_rate": 1.861953179469285e-05,
+      "loss": 0.9857,
       "step": 6843
     },
     {
-      "epoch": 0.18796517535909477,
+      "epoch": 0.194211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.8719815669405955e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.8619065796841858e-05,
+      "loss": 0.9743,
       "step": 6844
     },
     {
-      "epoch": 0.18799263958693802,
+      "epoch": 0.1942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.87193801823894e-05,
-      "loss": 1.0696,
+      "learning_rate": 1.861859972618523e-05,
+      "loss": 1.0219,
       "step": 6845
     },
     {
-      "epoch": 0.18802010381478124,
+      "epoch": 0.19426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.8718944626381727e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.8618133582726905e-05,
+      "loss": 1.0745,
       "step": 6846
     },
     {
-      "epoch": 0.18804756804262449,
+      "epoch": 0.19429625425652666,
       "grad_norm": 0.0,
-      "learning_rate": 1.8718509001386388e-05,
-      "loss": 1.1104,
+      "learning_rate": 1.8617667366470817e-05,
+      "loss": 0.906,
       "step": 6847
     },
     {
-      "epoch": 0.1880750322704677,
+      "epoch": 0.19432463110102158,
       "grad_norm": 0.0,
-      "learning_rate": 1.8718073307406826e-05,
-      "loss": 1.0939,
+      "learning_rate": 1.8617201077420907e-05,
+      "loss": 1.0061,
       "step": 6848
     },
     {
-      "epoch": 0.18810249649831096,
+      "epoch": 0.19435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.8717637544446496e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.861673471558111e-05,
+      "loss": 0.9526,
       "step": 6849
     },
     {
-      "epoch": 0.18812996072615418,
+      "epoch": 0.19438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.8717201712508835e-05,
-      "loss": 1.0766,
+      "learning_rate": 1.8616268280955366e-05,
+      "loss": 1.0056,
       "step": 6850
     },
     {
-      "epoch": 0.18815742495399743,
+      "epoch": 0.19440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.87167658115973e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.861580177354762e-05,
+      "loss": 0.9745,
       "step": 6851
     },
     {
-      "epoch": 0.18818488918184065,
+      "epoch": 0.19443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.871632984171534e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.8615335193361806e-05,
+      "loss": 0.971,
       "step": 6852
     },
     {
-      "epoch": 0.1882123534096839,
+      "epoch": 0.19446651532349601,
       "grad_norm": 0.0,
-      "learning_rate": 1.87158938028664e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.861486854040187e-05,
+      "loss": 0.9754,
       "step": 6853
     },
     {
-      "epoch": 0.18823981763752712,
+      "epoch": 0.19449489216799093,
       "grad_norm": 0.0,
-      "learning_rate": 1.871545769505393e-05,
-      "loss": 1.1093,
+      "learning_rate": 1.861440181467175e-05,
+      "loss": 0.8829,
       "step": 6854
     },
     {
-      "epoch": 0.18826728186537037,
+      "epoch": 0.19452326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 1.8715021518281385e-05,
-      "loss": 1.1859,
+      "learning_rate": 1.8613935016175396e-05,
+      "loss": 1.0357,
       "step": 6855
     },
     {
-      "epoch": 0.1882947460932136,
+      "epoch": 0.1945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8714585272552215e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.8613468144916742e-05,
+      "loss": 0.9201,
       "step": 6856
     },
     {
-      "epoch": 0.1883222103210568,
+      "epoch": 0.1945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.8714148957869867e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.8613001200899737e-05,
+      "loss": 0.9266,
       "step": 6857
     },
     {
-      "epoch": 0.18834967454890006,
+      "epoch": 0.19460839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 1.87137125742378e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.861253418412832e-05,
+      "loss": 0.9542,
       "step": 6858
     },
     {
-      "epoch": 0.18837713877674328,
+      "epoch": 0.1946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.871327612165946e-05,
-      "loss": 0.9454,
+      "learning_rate": 1.8612067094606442e-05,
+      "loss": 0.9961,
       "step": 6859
     },
     {
-      "epoch": 0.18840460300458653,
+      "epoch": 0.19466515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 1.8712839600138308e-05,
-      "loss": 1.018,
+      "learning_rate": 1.8611599932338046e-05,
+      "loss": 1.0648,
       "step": 6860
     },
     {
-      "epoch": 0.18843206723242975,
+      "epoch": 0.19469353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.8712403009677794e-05,
-      "loss": 1.1039,
+      "learning_rate": 1.861113269732708e-05,
+      "loss": 0.9937,
       "step": 6861
     },
     {
-      "epoch": 0.188459531460273,
+      "epoch": 0.19472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.871196635028137e-05,
-      "loss": 1.0555,
+      "learning_rate": 1.861066538957749e-05,
+      "loss": 0.9797,
       "step": 6862
     },
     {
-      "epoch": 0.18848699568811622,
+      "epoch": 0.19475028376844494,
       "grad_norm": 0.0,
-      "learning_rate": 1.8711529621952496e-05,
-      "loss": 1.03,
+      "learning_rate": 1.861019800909322e-05,
+      "loss": 0.9126,
       "step": 6863
     },
     {
-      "epoch": 0.18851445991595947,
+      "epoch": 0.19477866061293983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8711092824694624e-05,
-      "loss": 0.9931,
+      "learning_rate": 1.8609730555878217e-05,
+      "loss": 0.9823,
       "step": 6864
     },
     {
-      "epoch": 0.1885419241438027,
+      "epoch": 0.19480703745743475,
       "grad_norm": 0.0,
-      "learning_rate": 1.871065595851121e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.860926302993644e-05,
+      "loss": 1.1185,
       "step": 6865
     },
     {
-      "epoch": 0.18856938837164594,
+      "epoch": 0.19483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.8710219023405714e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.8608795431271823e-05,
+      "loss": 1.1388,
       "step": 6866
     },
     {
-      "epoch": 0.18859685259948916,
+      "epoch": 0.19486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.8709782019381587e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.860832775988833e-05,
+      "loss": 0.9626,
       "step": 6867
     },
     {
-      "epoch": 0.1886243168273324,
+      "epoch": 0.1948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.8709344946442297e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.8607860015789903e-05,
+      "loss": 1.0041,
       "step": 6868
     },
     {
-      "epoch": 0.18865178105517563,
+      "epoch": 0.1949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.8708907804591293e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.8607392198980496e-05,
+      "loss": 0.9861,
       "step": 6869
     },
     {
-      "epoch": 0.18867924528301888,
+      "epoch": 0.19494892167990918,
       "grad_norm": 0.0,
-      "learning_rate": 1.8708470593832035e-05,
-      "loss": 0.8712,
+      "learning_rate": 1.8606924309464062e-05,
+      "loss": 1.0771,
       "step": 6870
     },
     {
-      "epoch": 0.1887067095108621,
+      "epoch": 0.1949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.8708033314167988e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.8606456347244547e-05,
+      "loss": 0.856,
       "step": 6871
     },
     {
-      "epoch": 0.18873417373870532,
+      "epoch": 0.19500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.8707595965602605e-05,
-      "loss": 0.9491,
+      "learning_rate": 1.8605988312325915e-05,
+      "loss": 0.8942,
       "step": 6872
     },
     {
-      "epoch": 0.18876163796654857,
+      "epoch": 0.19503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.870715854813935e-05,
-      "loss": 1.0904,
+      "learning_rate": 1.8605520204712107e-05,
+      "loss": 0.9688,
       "step": 6873
     },
     {
-      "epoch": 0.1887891021943918,
+      "epoch": 0.19506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.8706721061781684e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.8605052024407083e-05,
+      "loss": 1.0965,
       "step": 6874
     },
     {
-      "epoch": 0.18881656642223504,
+      "epoch": 0.19509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.8706283506533073e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.86045837714148e-05,
+      "loss": 0.9963,
       "step": 6875
     },
     {
-      "epoch": 0.18884403065007827,
+      "epoch": 0.19511918274687853,
       "grad_norm": 0.0,
-      "learning_rate": 1.870584588239697e-05,
-      "loss": 1.1034,
+      "learning_rate": 1.860411544573921e-05,
+      "loss": 1.0806,
       "step": 6876
     },
     {
-      "epoch": 0.18887149487792151,
+      "epoch": 0.19514755959137345,
       "grad_norm": 0.0,
-      "learning_rate": 1.8705408189376842e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.8603647047384274e-05,
+      "loss": 1.0589,
       "step": 6877
     },
     {
-      "epoch": 0.18889895910576474,
+      "epoch": 0.19517593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.8704970427476152e-05,
-      "loss": 1.047,
+      "learning_rate": 1.860317857635394e-05,
+      "loss": 0.9798,
       "step": 6878
     },
     {
-      "epoch": 0.18892642333360798,
+      "epoch": 0.19520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.8704532596698365e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.8602710032652173e-05,
+      "loss": 0.9928,
       "step": 6879
     },
     {
-      "epoch": 0.1889538875614512,
+      "epoch": 0.1952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.8704094697046942e-05,
-      "loss": 1.1516,
+      "learning_rate": 1.8602241416282926e-05,
+      "loss": 0.9398,
       "step": 6880
     },
     {
-      "epoch": 0.18898135178929445,
+      "epoch": 0.195261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.8703656728525355e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.8601772727250158e-05,
+      "loss": 1.1038,
       "step": 6881
     },
     {
-      "epoch": 0.18900881601713768,
+      "epoch": 0.1952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.8703218691137062e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.860130396555783e-05,
+      "loss": 0.9179,
       "step": 6882
     },
     {
-      "epoch": 0.18903628024498093,
+      "epoch": 0.1953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.8702780584885532e-05,
-      "loss": 0.8592,
+      "learning_rate": 1.8600835131209902e-05,
+      "loss": 0.9968,
       "step": 6883
     },
     {
-      "epoch": 0.18906374447282415,
+      "epoch": 0.1953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.870234240977423e-05,
-      "loss": 1.09,
+      "learning_rate": 1.8600366224210332e-05,
+      "loss": 1.0048,
       "step": 6884
     },
     {
-      "epoch": 0.18909120870066737,
+      "epoch": 0.19537457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 1.8701904165806624e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.859989724456308e-05,
+      "loss": 1.0666,
       "step": 6885
     },
     {
-      "epoch": 0.18911867292851062,
+      "epoch": 0.19540295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.8701465852986184e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.8599428192272112e-05,
+      "loss": 0.9428,
       "step": 6886
     },
     {
-      "epoch": 0.18914613715635384,
+      "epoch": 0.19543132803632235,
       "grad_norm": 0.0,
-      "learning_rate": 1.8701027471316372e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.8598959067341386e-05,
+      "loss": 1.0036,
       "step": 6887
     },
     {
-      "epoch": 0.1891736013841971,
+      "epoch": 0.19545970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.8700589020800658e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.859848986977487e-05,
+      "loss": 1.0768,
       "step": 6888
     },
     {
-      "epoch": 0.1892010656120403,
+      "epoch": 0.19548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.8700150501442516e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.859802059957652e-05,
+      "loss": 1.008,
       "step": 6889
     },
     {
-      "epoch": 0.18922852983988356,
+      "epoch": 0.19551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.8699711913245414e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.85975512567503e-05,
+      "loss": 1.0828,
       "step": 6890
     },
     {
-      "epoch": 0.18925599406772678,
+      "epoch": 0.19554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.869927325621282e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.8597081841300184e-05,
+      "loss": 0.9761,
       "step": 6891
     },
     {
-      "epoch": 0.18928345829557003,
+      "epoch": 0.1955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.869883453034821e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.8596612353230127e-05,
+      "loss": 0.9493,
       "step": 6892
     },
     {
-      "epoch": 0.18931092252341325,
+      "epoch": 0.1956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.8698395735655047e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.8596142792544102e-05,
+      "loss": 0.9288,
       "step": 6893
     },
     {
-      "epoch": 0.1893383867512565,
+      "epoch": 0.19562996594778662,
       "grad_norm": 0.0,
-      "learning_rate": 1.869795687213681e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.8595673159246072e-05,
+      "loss": 1.0421,
       "step": 6894
     },
     {
-      "epoch": 0.18936585097909972,
+      "epoch": 0.1956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.869751793979697e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.8595203453340005e-05,
+      "loss": 0.8879,
       "step": 6895
     },
     {
-      "epoch": 0.18939331520694297,
+      "epoch": 0.1956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.8697078938638995e-05,
-      "loss": 1.1479,
+      "learning_rate": 1.859473367482987e-05,
+      "loss": 0.8936,
       "step": 6896
     },
     {
-      "epoch": 0.1894207794347862,
+      "epoch": 0.19571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.8696639868666365e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.8594263823719627e-05,
+      "loss": 1.0072,
       "step": 6897
     },
     {
-      "epoch": 0.1894482436626294,
+      "epoch": 0.19574347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.8696200729882554e-05,
-      "loss": 0.9517,
+      "learning_rate": 1.8593793900013254e-05,
+      "loss": 1.0633,
       "step": 6898
     },
     {
-      "epoch": 0.18947570789047266,
+      "epoch": 0.19577185017026108,
       "grad_norm": 0.0,
-      "learning_rate": 1.869576152229103e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.8593323903714718e-05,
+      "loss": 0.981,
       "step": 6899
     },
     {
-      "epoch": 0.18950317211831588,
+      "epoch": 0.19580022701475597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8695322245895275e-05,
-      "loss": 1.033,
+      "learning_rate": 1.859285383482799e-05,
+      "loss": 0.8991,
       "step": 6900
     },
     {
-      "epoch": 0.18953063634615913,
+      "epoch": 0.19582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.869488290069876e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.859238369335704e-05,
+      "loss": 1.0645,
       "step": 6901
     },
     {
-      "epoch": 0.18955810057400235,
+      "epoch": 0.19585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.8694443486704965e-05,
-      "loss": 0.886,
+      "learning_rate": 1.8591913479305834e-05,
+      "loss": 1.0467,
       "step": 6902
     },
     {
-      "epoch": 0.1895855648018456,
+      "epoch": 0.19588535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 1.8694004003917367e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.8591443192678353e-05,
+      "loss": 1.0975,
       "step": 6903
     },
     {
-      "epoch": 0.18961302902968882,
+      "epoch": 0.19591373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 1.8693564452339438e-05,
-      "loss": 1.0686,
+      "learning_rate": 1.8590972833478562e-05,
+      "loss": 1.0377,
       "step": 6904
     },
     {
-      "epoch": 0.18964049325753207,
+      "epoch": 0.19594211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.869312483197466e-05,
-      "loss": 1.0423,
+      "learning_rate": 1.859050240171044e-05,
+      "loss": 0.8591,
       "step": 6905
     },
     {
-      "epoch": 0.1896679574853753,
+      "epoch": 0.19597048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 1.8692685142826513e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.8590031897377954e-05,
+      "loss": 0.9405,
       "step": 6906
     },
     {
-      "epoch": 0.18969542171321854,
+      "epoch": 0.1959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.869224538489847e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.858956132048509e-05,
+      "loss": 0.9268,
       "step": 6907
     },
     {
-      "epoch": 0.18972288594106176,
+      "epoch": 0.1960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.8691805558194018e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.858909067103581e-05,
+      "loss": 0.9957,
       "step": 6908
     },
     {
-      "epoch": 0.189750350168905,
+      "epoch": 0.19605561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 1.8691365662716636e-05,
-      "loss": 1.1158,
+      "learning_rate": 1.8588619949034094e-05,
+      "loss": 1.0208,
       "step": 6909
     },
     {
-      "epoch": 0.18977781439674823,
+      "epoch": 0.19608399545970487,
       "grad_norm": 0.0,
-      "learning_rate": 1.8690925698469798e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.8588149154483922e-05,
+      "loss": 1.0469,
       "step": 6910
     },
     {
-      "epoch": 0.18980527862459146,
+      "epoch": 0.19611237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 1.869048566545699e-05,
-      "loss": 0.8789,
+      "learning_rate": 1.8587678287389265e-05,
+      "loss": 1.0103,
       "step": 6911
     },
     {
-      "epoch": 0.1898327428524347,
+      "epoch": 0.19614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.8690045563681692e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.8587207347754106e-05,
+      "loss": 0.939,
       "step": 6912
     },
     {
-      "epoch": 0.18986020708027793,
+      "epoch": 0.19616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.868960539314739e-05,
-      "loss": 1.1193,
+      "learning_rate": 1.858673633558242e-05,
+      "loss": 1.0533,
       "step": 6913
     },
     {
-      "epoch": 0.18988767130812118,
+      "epoch": 0.19619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.8689165153857563e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.8586265250878185e-05,
+      "loss": 1.0433,
       "step": 6914
     },
     {
-      "epoch": 0.1899151355359644,
+      "epoch": 0.19622587968217933,
       "grad_norm": 0.0,
-      "learning_rate": 1.8688724845815695e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.8585794093645385e-05,
+      "loss": 1.059,
       "step": 6915
     },
     {
-      "epoch": 0.18994259976380765,
+      "epoch": 0.19625425652667422,
       "grad_norm": 0.0,
-      "learning_rate": 1.868828446902527e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.8585322863887995e-05,
+      "loss": 1.0774,
       "step": 6916
     },
     {
-      "epoch": 0.18997006399165087,
+      "epoch": 0.19628263337116914,
       "grad_norm": 0.0,
-      "learning_rate": 1.8687844023489773e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.8584851561609995e-05,
+      "loss": 1.1074,
       "step": 6917
     },
     {
-      "epoch": 0.18999752821949412,
+      "epoch": 0.19631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.868740350921269e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.8584380186815366e-05,
+      "loss": 0.8846,
       "step": 6918
     },
     {
-      "epoch": 0.19002499244733734,
+      "epoch": 0.1963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.8686962926197502e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.8583908739508097e-05,
+      "loss": 1.189,
       "step": 6919
     },
     {
-      "epoch": 0.1900524566751806,
+      "epoch": 0.1963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.8686522274447702e-05,
-      "loss": 1.093,
+      "learning_rate": 1.858343721969216e-05,
+      "loss": 1.0758,
       "step": 6920
     },
     {
-      "epoch": 0.1900799209030238,
+      "epoch": 0.19639614074914868,
       "grad_norm": 0.0,
-      "learning_rate": 1.868608155396677e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.8582965627371546e-05,
+      "loss": 1.0699,
       "step": 6921
     },
     {
-      "epoch": 0.19010738513086706,
+      "epoch": 0.1964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.8685640764758196e-05,
-      "loss": 0.8796,
+      "learning_rate": 1.8582493962550235e-05,
+      "loss": 0.9032,
       "step": 6922
     },
     {
-      "epoch": 0.19013484935871028,
+      "epoch": 0.1964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.8685199906825473e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.8582022225232213e-05,
+      "loss": 1.0258,
       "step": 6923
     },
     {
-      "epoch": 0.19016231358655353,
+      "epoch": 0.19648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.868475898017208e-05,
-      "loss": 1.057,
+      "learning_rate": 1.858155041542146e-05,
+      "loss": 1.0788,
       "step": 6924
     },
     {
-      "epoch": 0.19018977781439675,
+      "epoch": 0.19650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.868431798480151e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.8581078533121965e-05,
+      "loss": 0.9514,
       "step": 6925
     },
     {
-      "epoch": 0.19021724204223997,
+      "epoch": 0.19653802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.868387692071725e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.8580606578337718e-05,
+      "loss": 1.1099,
       "step": 6926
     },
     {
-      "epoch": 0.19024470627008322,
+      "epoch": 0.19656640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.8683435787922798e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.8580134551072698e-05,
+      "loss": 0.8672,
       "step": 6927
     },
     {
-      "epoch": 0.19027217049792644,
+      "epoch": 0.19659477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.8682994586421635e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.8579662451330898e-05,
+      "loss": 0.8967,
       "step": 6928
     },
     {
-      "epoch": 0.1902996347257697,
+      "epoch": 0.19662315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 1.8682553316217255e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.8579190279116305e-05,
+      "loss": 1.0659,
       "step": 6929
     },
     {
-      "epoch": 0.1903270989536129,
+      "epoch": 0.19665153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.8682111977313146e-05,
-      "loss": 1.0495,
+      "learning_rate": 1.85787180344329e-05,
+      "loss": 1.1232,
       "step": 6930
     },
     {
-      "epoch": 0.19035456318145616,
+      "epoch": 0.1966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.868167056971281e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.8578245717284682e-05,
+      "loss": 0.9107,
       "step": 6931
     },
     {
-      "epoch": 0.19038202740929938,
+      "epoch": 0.1967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.868122909341973e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.8577773327675637e-05,
+      "loss": 0.9855,
       "step": 6932
     },
     {
-      "epoch": 0.19040949163714263,
+      "epoch": 0.1967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.8680787548437408e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.8577300865609755e-05,
+      "loss": 1.009,
       "step": 6933
     },
     {
-      "epoch": 0.19043695586498585,
+      "epoch": 0.1967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.8680345934769322e-05,
-      "loss": 1.1365,
+      "learning_rate": 1.8576828331091028e-05,
+      "loss": 0.9841,
       "step": 6934
     },
     {
-      "epoch": 0.1904644200928291,
+      "epoch": 0.1967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.8679904252418984e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.8576355724123446e-05,
+      "loss": 1.1076,
       "step": 6935
     },
     {
-      "epoch": 0.19049188432067232,
+      "epoch": 0.19682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.8679462501389883e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.8575883044711e-05,
+      "loss": 0.9853,
       "step": 6936
     },
     {
-      "epoch": 0.19051934854851557,
+      "epoch": 0.19685017026106696,
       "grad_norm": 0.0,
-      "learning_rate": 1.867902068168551e-05,
-      "loss": 1.0846,
+      "learning_rate": 1.8575410292857687e-05,
+      "loss": 1.1228,
       "step": 6937
     },
     {
-      "epoch": 0.1905468127763588,
+      "epoch": 0.19687854710556185,
       "grad_norm": 0.0,
-      "learning_rate": 1.867857879330936e-05,
-      "loss": 0.9035,
+      "learning_rate": 1.8574937468567495e-05,
+      "loss": 1.079,
       "step": 6938
     },
     {
-      "epoch": 0.19057427700420201,
+      "epoch": 0.19690692395005677,
       "grad_norm": 0.0,
-      "learning_rate": 1.8678136836264938e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.857446457184442e-05,
+      "loss": 0.9943,
       "step": 6939
     },
     {
-      "epoch": 0.19060174123204526,
+      "epoch": 0.19693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.8677694810555734e-05,
-      "loss": 1.007,
+      "learning_rate": 1.8573991602692457e-05,
+      "loss": 1.0206,
       "step": 6940
     },
     {
-      "epoch": 0.19062920545988848,
+      "epoch": 0.19696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.867725271618525e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.8573518561115604e-05,
+      "loss": 0.906,
       "step": 6941
     },
     {
-      "epoch": 0.19065666968773173,
+      "epoch": 0.19699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.8676810553156975e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.8573045447117854e-05,
+      "loss": 0.988,
       "step": 6942
     },
     {
-      "epoch": 0.19068413391557495,
+      "epoch": 0.19702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.8676368321474417e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.8572572260703205e-05,
+      "loss": 0.9206,
       "step": 6943
     },
     {
-      "epoch": 0.1907115981434182,
+      "epoch": 0.1970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.867592602114107e-05,
-      "loss": 1.01,
+      "learning_rate": 1.857209900187565e-05,
+      "loss": 0.9372,
       "step": 6944
     },
     {
-      "epoch": 0.19073906237126143,
+      "epoch": 0.19707718501702612,
       "grad_norm": 0.0,
-      "learning_rate": 1.867548365216044e-05,
-      "loss": 1.062,
+      "learning_rate": 1.8571625670639192e-05,
+      "loss": 0.9819,
       "step": 6945
     },
     {
-      "epoch": 0.19076652659910467,
+      "epoch": 0.197105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.867504121453602e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.8571152266997824e-05,
+      "loss": 1.0094,
       "step": 6946
     },
     {
-      "epoch": 0.1907939908269479,
+      "epoch": 0.1971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.8674598708271312e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.857067879095555e-05,
+      "loss": 0.9542,
       "step": 6947
     },
     {
-      "epoch": 0.19082145505479114,
+      "epoch": 0.19716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.867415613336982e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.8570205242516366e-05,
+      "loss": 1.0209,
       "step": 6948
     },
     {
-      "epoch": 0.19084891928263437,
+      "epoch": 0.19719069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.867371348983505e-05,
-      "loss": 0.9039,
+      "learning_rate": 1.856973162168427e-05,
+      "loss": 0.919,
       "step": 6949
     },
     {
-      "epoch": 0.19087638351047762,
+      "epoch": 0.19721906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.867327077767049e-05,
-      "loss": 1.0635,
+      "learning_rate": 1.856925792846327e-05,
+      "loss": 0.9335,
       "step": 6950
     },
     {
-      "epoch": 0.19090384773832084,
+      "epoch": 0.19724744608399547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8672827996879658e-05,
-      "loss": 1.0536,
+      "learning_rate": 1.856878416285736e-05,
+      "loss": 0.9459,
       "step": 6951
     },
     {
-      "epoch": 0.19093131196616406,
+      "epoch": 0.19727582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.867238514746605e-05,
-      "loss": 1.1114,
+      "learning_rate": 1.856831032487055e-05,
+      "loss": 1.0399,
       "step": 6952
     },
     {
-      "epoch": 0.1909587761940073,
+      "epoch": 0.19730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.867194222943317e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.8567836414506835e-05,
+      "loss": 1.0422,
       "step": 6953
     },
     {
-      "epoch": 0.19098624042185053,
+      "epoch": 0.19733257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 1.8671499242784525e-05,
-      "loss": 1.107,
+      "learning_rate": 1.856736243177022e-05,
+      "loss": 0.9853,
       "step": 6954
     },
     {
-      "epoch": 0.19101370464969378,
+      "epoch": 0.19736095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 1.867105618752362e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.8566888376664707e-05,
+      "loss": 0.9836,
       "step": 6955
     },
     {
-      "epoch": 0.191041168877537,
+      "epoch": 0.1973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.8670613063653956e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.8566414249194305e-05,
+      "loss": 0.9632,
       "step": 6956
     },
     {
-      "epoch": 0.19106863310538025,
+      "epoch": 0.19741770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 1.8670169871179046e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.8565940049363017e-05,
+      "loss": 1.1072,
       "step": 6957
     },
     {
-      "epoch": 0.19109609733322347,
+      "epoch": 0.1974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8669726610102392e-05,
-      "loss": 1.0275,
+      "learning_rate": 1.8565465777174848e-05,
+      "loss": 0.9996,
       "step": 6958
     },
     {
-      "epoch": 0.19112356156106672,
+      "epoch": 0.1974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.8669283280427504e-05,
-      "loss": 0.982,
+      "learning_rate": 1.8564991432633805e-05,
+      "loss": 1.0824,
       "step": 6959
     },
     {
-      "epoch": 0.19115102578890994,
+      "epoch": 0.19750283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 1.8668839882157886e-05,
-      "loss": 1.0088,
+      "learning_rate": 1.8564517015743894e-05,
+      "loss": 0.9532,
       "step": 6960
     },
     {
-      "epoch": 0.1911784900167532,
+      "epoch": 0.19753121452894437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8668396415297054e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.8564042526509123e-05,
+      "loss": 1.0707,
       "step": 6961
     },
     {
-      "epoch": 0.1912059542445964,
+      "epoch": 0.19755959137343929,
       "grad_norm": 0.0,
-      "learning_rate": 1.8667952879848505e-05,
-      "loss": 1.0793,
+      "learning_rate": 1.85635679649335e-05,
+      "loss": 1.0714,
       "step": 6962
     },
     {
-      "epoch": 0.19123341847243966,
+      "epoch": 0.19758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.866750927581576e-05,
-      "loss": 0.935,
+      "learning_rate": 1.8563093331021034e-05,
+      "loss": 1.0455,
       "step": 6963
     },
     {
-      "epoch": 0.19126088270028288,
+      "epoch": 0.19761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.8667065603202323e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.856261862477573e-05,
+      "loss": 0.9205,
       "step": 6964
     },
     {
-      "epoch": 0.19128834692812613,
+      "epoch": 0.19764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.8666621862011704e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.856214384620161e-05,
+      "loss": 1.081,
       "step": 6965
     },
     {
-      "epoch": 0.19131581115596935,
+      "epoch": 0.19767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.866617805224742e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.8561668995302668e-05,
+      "loss": 0.893,
       "step": 6966
     },
     {
-      "epoch": 0.19134327538381257,
+      "epoch": 0.19770147559591372,
       "grad_norm": 0.0,
-      "learning_rate": 1.8665734173912974e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.8561194072082925e-05,
+      "loss": 1.1189,
       "step": 6967
     },
     {
-      "epoch": 0.19137073961165582,
+      "epoch": 0.19772985244040864,
       "grad_norm": 0.0,
-      "learning_rate": 1.8665290227011882e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.856071907654639e-05,
+      "loss": 1.0211,
       "step": 6968
     },
     {
-      "epoch": 0.19139820383949904,
+      "epoch": 0.19775822928490353,
       "grad_norm": 0.0,
-      "learning_rate": 1.866484621154766e-05,
-      "loss": 1.0265,
+      "learning_rate": 1.8560244008697076e-05,
+      "loss": 0.9381,
       "step": 6969
     },
     {
-      "epoch": 0.1914256680673423,
+      "epoch": 0.1977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.8664402127523818e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.8559768868539e-05,
+      "loss": 1.0726,
       "step": 6970
     },
     {
-      "epoch": 0.1914531322951855,
+      "epoch": 0.1978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.866395797494387e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.8559293656076167e-05,
+      "loss": 1.0939,
       "step": 6971
     },
     {
-      "epoch": 0.19148059652302876,
+      "epoch": 0.1978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.8663513753811327e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.8558818371312598e-05,
+      "loss": 0.9178,
       "step": 6972
     },
     {
-      "epoch": 0.19150806075087198,
+      "epoch": 0.19787173666288307,
       "grad_norm": 0.0,
-      "learning_rate": 1.866306946412971e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.8558343014252304e-05,
+      "loss": 0.9688,
       "step": 6973
     },
     {
-      "epoch": 0.19153552497871523,
+      "epoch": 0.197900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.8662625105902532e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.8557867584899307e-05,
+      "loss": 0.9713,
       "step": 6974
     },
     {
-      "epoch": 0.19156298920655845,
+      "epoch": 0.19792849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 1.866218067913331e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.855739208325761e-05,
+      "loss": 1.0505,
       "step": 6975
     },
     {
-      "epoch": 0.1915904534344017,
+      "epoch": 0.19795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.8661736183825556e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.855691650933124e-05,
+      "loss": 1.0182,
       "step": 6976
     },
     {
-      "epoch": 0.19161791766224492,
+      "epoch": 0.19798524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.8661291619982793e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.8556440863124215e-05,
+      "loss": 0.9184,
       "step": 6977
     },
     {
-      "epoch": 0.19164538189008817,
+      "epoch": 0.19801362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.8660846987608533e-05,
-      "loss": 0.958,
+      "learning_rate": 1.855596514464055e-05,
+      "loss": 1.0832,
       "step": 6978
     },
     {
-      "epoch": 0.1916728461179314,
+      "epoch": 0.19804199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 1.8660402286706298e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.855548935388426e-05,
+      "loss": 0.9012,
       "step": 6979
     },
     {
-      "epoch": 0.19170031034577462,
+      "epoch": 0.19807037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 1.8659957517279602e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.8555013490859367e-05,
+      "loss": 1.1046,
       "step": 6980
     },
     {
-      "epoch": 0.19172777457361787,
+      "epoch": 0.19809875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.8659512679331973e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.855453755556989e-05,
+      "loss": 1.006,
       "step": 6981
     },
     {
-      "epoch": 0.1917552388014611,
+      "epoch": 0.19812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.8659067772866924e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.8554061548019847e-05,
+      "loss": 0.9231,
       "step": 6982
     },
     {
-      "epoch": 0.19178270302930434,
+      "epoch": 0.198155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.8658622797887975e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.8553585468213264e-05,
+      "loss": 1.0136,
       "step": 6983
     },
     {
-      "epoch": 0.19181016725714756,
+      "epoch": 0.1981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.8658177754398647e-05,
-      "loss": 1.0086,
+      "learning_rate": 1.855310931615416e-05,
+      "loss": 0.9612,
       "step": 6984
     },
     {
-      "epoch": 0.1918376314849908,
+      "epoch": 0.1982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.8657732642402466e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.855263309184656e-05,
+      "loss": 1.0546,
       "step": 6985
     },
     {
-      "epoch": 0.19186509571283403,
+      "epoch": 0.1982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.865728746190295e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.8552156795294482e-05,
+      "loss": 1.0896,
       "step": 6986
     },
     {
-      "epoch": 0.19189255994067728,
+      "epoch": 0.19826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.865684221290362e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.855168042650195e-05,
+      "loss": 0.9965,
       "step": 6987
     },
     {
-      "epoch": 0.1919200241685205,
+      "epoch": 0.19829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8656396895408004e-05,
-      "loss": 0.977,
+      "learning_rate": 1.855120398547299e-05,
+      "loss": 1.0424,
       "step": 6988
     },
     {
-      "epoch": 0.19194748839636375,
+      "epoch": 0.19832576617480135,
       "grad_norm": 0.0,
-      "learning_rate": 1.865595150941962e-05,
-      "loss": 0.9968,
+      "learning_rate": 1.8550727472211624e-05,
+      "loss": 0.981,
       "step": 6989
     },
     {
-      "epoch": 0.19197495262420697,
+      "epoch": 0.19835414301929624,
       "grad_norm": 0.0,
-      "learning_rate": 1.8655506054941997e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.855025088672188e-05,
+      "loss": 0.9515,
       "step": 6990
     },
     {
-      "epoch": 0.19200241685205022,
+      "epoch": 0.19838251986379116,
       "grad_norm": 0.0,
-      "learning_rate": 1.8655060531978657e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.8549774229007784e-05,
+      "loss": 0.9498,
       "step": 6991
     },
     {
-      "epoch": 0.19202988107989344,
+      "epoch": 0.19841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.8654614940533124e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.8549297499073358e-05,
+      "loss": 1.1767,
       "step": 6992
     },
     {
-      "epoch": 0.19205734530773666,
+      "epoch": 0.19843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.8654169280608924e-05,
-      "loss": 0.9768,
+      "learning_rate": 1.854882069692263e-05,
+      "loss": 0.9299,
       "step": 6993
     },
     {
-      "epoch": 0.1920848095355799,
+      "epoch": 0.19846765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.8653723552209588e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.8548343822559632e-05,
+      "loss": 0.9475,
       "step": 6994
     },
     {
-      "epoch": 0.19211227376342313,
+      "epoch": 0.1984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.8653277755338636e-05,
-      "loss": 1.1482,
+      "learning_rate": 1.8547866875988392e-05,
+      "loss": 0.9358,
       "step": 6995
     },
     {
-      "epoch": 0.19213973799126638,
+      "epoch": 0.1985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.86528318899996e-05,
-      "loss": 1.0872,
+      "learning_rate": 1.8547389857212933e-05,
+      "loss": 0.9912,
       "step": 6996
     },
     {
-      "epoch": 0.1921672022191096,
+      "epoch": 0.1985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8652385956196006e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.854691276623729e-05,
+      "loss": 0.9726,
       "step": 6997
     },
     {
-      "epoch": 0.19219466644695285,
+      "epoch": 0.1985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.8651939953931382e-05,
-      "loss": 1.062,
+      "learning_rate": 1.8546435603065488e-05,
+      "loss": 1.0623,
       "step": 6998
     },
     {
-      "epoch": 0.19222213067479607,
+      "epoch": 0.19860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.865149388320926e-05,
-      "loss": 1.0775,
+      "learning_rate": 1.8545958367701563e-05,
+      "loss": 1.0974,
       "step": 6999
     },
     {
-      "epoch": 0.19224959490263932,
+      "epoch": 0.19863791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.8651047744033165e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.8545481060149543e-05,
+      "loss": 1.0553,
       "step": 7000
     },
     {
-      "epoch": 0.19227705913048254,
+      "epoch": 0.19866628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 1.8650601536406626e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.854500368041346e-05,
+      "loss": 0.9204,
       "step": 7001
     },
     {
-      "epoch": 0.1923045233583258,
+      "epoch": 0.19869466515323497,
       "grad_norm": 0.0,
-      "learning_rate": 1.865015526033318e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.8544526228497342e-05,
+      "loss": 0.9772,
       "step": 7002
     },
     {
-      "epoch": 0.192331987586169,
+      "epoch": 0.19872304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 1.8649708915816357e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.8544048704405232e-05,
+      "loss": 0.9143,
       "step": 7003
     },
     {
-      "epoch": 0.19235945181401226,
+      "epoch": 0.19875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.8649262502859683e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.8543571108141158e-05,
+      "loss": 0.9504,
       "step": 7004
     },
     {
-      "epoch": 0.19238691604185548,
+      "epoch": 0.19877979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.8648816021466695e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.854309343970915e-05,
+      "loss": 0.9332,
       "step": 7005
     },
     {
-      "epoch": 0.1924143802696987,
+      "epoch": 0.19880817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.8648369471640926e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.8542615699113255e-05,
+      "loss": 1.0026,
       "step": 7006
     },
     {
-      "epoch": 0.19244184449754195,
+      "epoch": 0.1988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.8647922853385906e-05,
-      "loss": 1.0766,
+      "learning_rate": 1.85421378863575e-05,
+      "loss": 1.0354,
       "step": 7007
     },
     {
-      "epoch": 0.19246930872538517,
+      "epoch": 0.19886492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 1.8647476166705173e-05,
-      "loss": 1.0836,
+      "learning_rate": 1.854166000144592e-05,
+      "loss": 0.9905,
       "step": 7008
     },
     {
-      "epoch": 0.19249677295322842,
+      "epoch": 0.1988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.8647029411602256e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.854118204438255e-05,
+      "loss": 1.0522,
       "step": 7009
     },
     {
-      "epoch": 0.19252423718107164,
+      "epoch": 0.1989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.8646582588080694e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.854070401517144e-05,
+      "loss": 0.9101,
       "step": 7010
     },
     {
-      "epoch": 0.1925517014089149,
+      "epoch": 0.198950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.864613569614402e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.854022591381661e-05,
+      "loss": 1.0935,
       "step": 7011
     },
     {
-      "epoch": 0.19257916563675812,
+      "epoch": 0.19897843359818387,
       "grad_norm": 0.0,
-      "learning_rate": 1.864568873579577e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.8539747740322114e-05,
+      "loss": 1.0582,
       "step": 7012
     },
     {
-      "epoch": 0.19260662986460136,
+      "epoch": 0.19900681044267876,
       "grad_norm": 0.0,
-      "learning_rate": 1.8645241707039483e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.8539269494691984e-05,
+      "loss": 0.9469,
       "step": 7013
     },
     {
-      "epoch": 0.19263409409244459,
+      "epoch": 0.19903518728717368,
       "grad_norm": 0.0,
-      "learning_rate": 1.8644794609878696e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.853879117693026e-05,
+      "loss": 0.9819,
       "step": 7014
     },
     {
-      "epoch": 0.19266155832028783,
+      "epoch": 0.19906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.8644347444316945e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.8538312787040983e-05,
+      "loss": 0.8992,
       "step": 7015
     },
     {
-      "epoch": 0.19268902254813106,
+      "epoch": 0.19909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.864390021035777e-05,
-      "loss": 1.1196,
+      "learning_rate": 1.8537834325028196e-05,
+      "loss": 1.1005,
       "step": 7016
     },
     {
-      "epoch": 0.1927164867759743,
+      "epoch": 0.19912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.8643452908004704e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.8537355790895934e-05,
+      "loss": 0.9671,
       "step": 7017
     },
     {
-      "epoch": 0.19274395100381753,
+      "epoch": 0.19914869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 1.864300553726129e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.853687718464825e-05,
+      "loss": 0.8914,
       "step": 7018
     },
     {
-      "epoch": 0.19277141523166078,
+      "epoch": 0.19917707150964814,
       "grad_norm": 0.0,
-      "learning_rate": 1.8642558098131073e-05,
-      "loss": 1.052,
+      "learning_rate": 1.8536398506289176e-05,
+      "loss": 0.9772,
       "step": 7019
     },
     {
-      "epoch": 0.192798879459504,
+      "epoch": 0.19920544835414303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8642110590617588e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.853591975582276e-05,
+      "loss": 1.0508,
       "step": 7020
     },
     {
-      "epoch": 0.19282634368734722,
+      "epoch": 0.19923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.8641663014724374e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.853544093325305e-05,
+      "loss": 1.0912,
       "step": 7021
     },
     {
-      "epoch": 0.19285380791519047,
+      "epoch": 0.1992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.8641215370454974e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.8534962038584083e-05,
+      "loss": 0.9293,
       "step": 7022
     },
     {
-      "epoch": 0.1928812721430337,
+      "epoch": 0.1992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.8640767657812935e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.853448307181991e-05,
+      "loss": 0.9554,
       "step": 7023
     },
     {
-      "epoch": 0.19290873637087694,
+      "epoch": 0.19931895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 1.8640319876801792e-05,
-      "loss": 1.182,
+      "learning_rate": 1.8534004032964574e-05,
+      "loss": 0.9429,
       "step": 7024
     },
     {
-      "epoch": 0.19293620059872016,
+      "epoch": 0.1993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.863987202742509e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.8533524922022123e-05,
+      "loss": 1.0567,
       "step": 7025
     },
     {
-      "epoch": 0.1929636648265634,
+      "epoch": 0.19937570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 1.8639424109686378e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.8533045738996602e-05,
+      "loss": 0.9076,
       "step": 7026
     },
     {
-      "epoch": 0.19299112905440663,
+      "epoch": 0.19940408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.8638976123589195e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.853256648389206e-05,
+      "loss": 1.0133,
       "step": 7027
     },
     {
-      "epoch": 0.19301859328224988,
+      "epoch": 0.19943246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.863852806913708e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.8532087156712547e-05,
+      "loss": 1.0799,
       "step": 7028
     },
     {
-      "epoch": 0.1930460575100931,
+      "epoch": 0.19946083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.863807994633359e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.853160775746211e-05,
+      "loss": 1.0013,
       "step": 7029
     },
     {
-      "epoch": 0.19307352173793635,
+      "epoch": 0.19948921679909193,
       "grad_norm": 0.0,
-      "learning_rate": 1.8637631755182267e-05,
-      "loss": 0.8828,
+      "learning_rate": 1.85311282861448e-05,
+      "loss": 1.0809,
       "step": 7030
     },
     {
-      "epoch": 0.19310098596577957,
+      "epoch": 0.19951759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 1.8637183495686654e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.853064874276466e-05,
+      "loss": 1.0508,
       "step": 7031
     },
     {
-      "epoch": 0.19312845019362282,
+      "epoch": 0.19954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.8636735167850298e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.853016912732575e-05,
+      "loss": 0.9376,
       "step": 7032
     },
     {
-      "epoch": 0.19315591442146604,
+      "epoch": 0.19957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.8636286771676747e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.852968943983212e-05,
+      "loss": 0.9757,
       "step": 7033
     },
     {
-      "epoch": 0.19318337864930926,
+      "epoch": 0.1996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.8635838307169552e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.852920968028782e-05,
+      "loss": 1.0503,
       "step": 7034
     },
     {
-      "epoch": 0.1932108428771525,
+      "epoch": 0.1996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.863538977433226e-05,
-      "loss": 1.0286,
+      "learning_rate": 1.85287298486969e-05,
+      "loss": 1.0378,
       "step": 7035
     },
     {
-      "epoch": 0.19323830710499573,
+      "epoch": 0.19965947786606128,
       "grad_norm": 0.0,
-      "learning_rate": 1.8634941173168418e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.852824994506342e-05,
+      "loss": 1.0396,
       "step": 7036
     },
     {
-      "epoch": 0.19326577133283898,
+      "epoch": 0.1996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.8634492503681577e-05,
-      "loss": 0.925,
+      "learning_rate": 1.8527769969391425e-05,
+      "loss": 0.957,
       "step": 7037
     },
     {
-      "epoch": 0.1932932355606822,
+      "epoch": 0.19971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.8634043765875284e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.852728992168497e-05,
+      "loss": 0.9576,
       "step": 7038
     },
     {
-      "epoch": 0.19332069978852545,
+      "epoch": 0.19974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8633594959753094e-05,
-      "loss": 1.0781,
+      "learning_rate": 1.8526809801948123e-05,
+      "loss": 1.0345,
       "step": 7039
     },
     {
-      "epoch": 0.19334816401636867,
+      "epoch": 0.19977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.8633146085318558e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.8526329610184922e-05,
+      "loss": 0.8144,
       "step": 7040
     },
     {
-      "epoch": 0.19337562824421192,
+      "epoch": 0.19980136208853574,
       "grad_norm": 0.0,
-      "learning_rate": 1.8632697142575226e-05,
-      "loss": 1.0964,
+      "learning_rate": 1.8525849346399435e-05,
+      "loss": 0.9803,
       "step": 7041
     },
     {
-      "epoch": 0.19340309247205514,
+      "epoch": 0.19982973893303066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8632248131526648e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.8525369010595717e-05,
+      "loss": 0.9026,
       "step": 7042
     },
     {
-      "epoch": 0.1934305566998984,
+      "epoch": 0.19985811577752555,
       "grad_norm": 0.0,
-      "learning_rate": 1.8631799052176376e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.852488860277782e-05,
+      "loss": 0.9897,
       "step": 7043
     },
     {
-      "epoch": 0.19345802092774161,
+      "epoch": 0.19988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.863134990452797e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.852440812294981e-05,
+      "loss": 1.0151,
       "step": 7044
     },
     {
-      "epoch": 0.19348548515558486,
+      "epoch": 0.19991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.8630900688584984e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.852392757111574e-05,
+      "loss": 1.0304,
       "step": 7045
     },
     {
-      "epoch": 0.19351294938342808,
+      "epoch": 0.1999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.863045140435096e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.8523446947279668e-05,
+      "loss": 1.0368,
       "step": 7046
     },
     {
-      "epoch": 0.1935404136112713,
+      "epoch": 0.1999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.8630002051829467e-05,
-      "loss": 1.0997,
+      "learning_rate": 1.852296625144566e-05,
+      "loss": 1.0816,
       "step": 7047
     },
     {
-      "epoch": 0.19356787783911455,
+      "epoch": 0.2,
       "grad_norm": 0.0,
-      "learning_rate": 1.8629552631024052e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.852248548361777e-05,
+      "loss": 0.9651,
       "step": 7048
     },
     {
-      "epoch": 0.19359534206695778,
+      "epoch": 0.2000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.862910314193828e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.8522004643800064e-05,
+      "loss": 1.0133,
       "step": 7049
     },
     {
-      "epoch": 0.19362280629480103,
+      "epoch": 0.20005675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 1.8628653584575694e-05,
-      "loss": 1.0888,
+      "learning_rate": 1.85215237319966e-05,
+      "loss": 1.0083,
       "step": 7050
     },
     {
-      "epoch": 0.19365027052264425,
+      "epoch": 0.20008513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 1.8628203958939864e-05,
-      "loss": 1.0188,
+      "learning_rate": 1.8521042748211446e-05,
+      "loss": 0.9897,
       "step": 7051
     },
     {
-      "epoch": 0.1936777347504875,
+      "epoch": 0.20011350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 1.8627754265034338e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.8520561692448655e-05,
+      "loss": 1.0681,
       "step": 7052
     },
     {
-      "epoch": 0.19370519897833072,
+      "epoch": 0.20014188422247445,
       "grad_norm": 0.0,
-      "learning_rate": 1.862730450286268e-05,
-      "loss": 1.0637,
+      "learning_rate": 1.8520080564712303e-05,
+      "loss": 0.9805,
       "step": 7053
     },
     {
-      "epoch": 0.19373266320617397,
+      "epoch": 0.20017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.8626854672428446e-05,
-      "loss": 1.1153,
+      "learning_rate": 1.851959936500644e-05,
+      "loss": 0.97,
       "step": 7054
     },
     {
-      "epoch": 0.1937601274340172,
+      "epoch": 0.20019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.8626404773735194e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.8519118093335146e-05,
+      "loss": 1.0631,
       "step": 7055
     },
     {
-      "epoch": 0.19378759166186044,
+      "epoch": 0.20022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.8625954806786487e-05,
-      "loss": 1.0801,
+      "learning_rate": 1.8518636749702473e-05,
+      "loss": 0.973,
       "step": 7056
     },
     {
-      "epoch": 0.19381505588970366,
+      "epoch": 0.20025539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.8625504771585887e-05,
-      "loss": 0.9092,
+      "learning_rate": 1.8518155334112494e-05,
+      "loss": 1.1588,
       "step": 7057
     },
     {
-      "epoch": 0.1938425201175469,
+      "epoch": 0.2002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.862505466813695e-05,
-      "loss": 1.0864,
+      "learning_rate": 1.851767384656927e-05,
+      "loss": 0.9799,
       "step": 7058
     },
     {
-      "epoch": 0.19386998434539013,
+      "epoch": 0.20031214528944383,
       "grad_norm": 0.0,
-      "learning_rate": 1.862460449644324e-05,
-      "loss": 0.9444,
+      "learning_rate": 1.851719228707688e-05,
+      "loss": 1.0783,
       "step": 7059
     },
     {
-      "epoch": 0.19389744857323335,
+      "epoch": 0.20034052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.862415425650832e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.8516710655639377e-05,
+      "loss": 1.0345,
       "step": 7060
     },
     {
-      "epoch": 0.1939249128010766,
+      "epoch": 0.2003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.8623703948335747e-05,
-      "loss": 1.0948,
+      "learning_rate": 1.851622895226084e-05,
+      "loss": 1.025,
       "step": 7061
     },
     {
-      "epoch": 0.19395237702891982,
+      "epoch": 0.2003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.862325357192909e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.8515747176945333e-05,
+      "loss": 1.0486,
       "step": 7062
     },
     {
-      "epoch": 0.19397984125676307,
+      "epoch": 0.20042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.8622803127291914e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.851526532969693e-05,
+      "loss": 0.983,
       "step": 7063
     },
     {
-      "epoch": 0.1940073054846063,
+      "epoch": 0.20045402951191826,
       "grad_norm": 0.0,
-      "learning_rate": 1.8622352614427775e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.8514783410519693e-05,
+      "loss": 0.9422,
       "step": 7064
     },
     {
-      "epoch": 0.19403476971244954,
+      "epoch": 0.20048240635641318,
       "grad_norm": 0.0,
-      "learning_rate": 1.862190203334024e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.8514301419417697e-05,
+      "loss": 1.0078,
       "step": 7065
     },
     {
-      "epoch": 0.19406223394029276,
+      "epoch": 0.20051078320090807,
       "grad_norm": 0.0,
-      "learning_rate": 1.862145138403288e-05,
-      "loss": 1.082,
+      "learning_rate": 1.851381935639502e-05,
+      "loss": 0.9572,
       "step": 7066
     },
     {
-      "epoch": 0.194089698168136,
+      "epoch": 0.20053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.8621000666509257e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.8513337221455723e-05,
+      "loss": 0.9309,
       "step": 7067
     },
     {
-      "epoch": 0.19411716239597923,
+      "epoch": 0.20056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.8620549880772936e-05,
-      "loss": 0.906,
+      "learning_rate": 1.851285501460389e-05,
+      "loss": 0.9598,
       "step": 7068
     },
     {
-      "epoch": 0.19414462662382248,
+      "epoch": 0.20059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.8620099026827487e-05,
-      "loss": 1.0939,
+      "learning_rate": 1.851237273584358e-05,
+      "loss": 0.8198,
       "step": 7069
     },
     {
-      "epoch": 0.1941720908516657,
+      "epoch": 0.20062429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 1.8619648104676472e-05,
-      "loss": 1.2291,
+      "learning_rate": 1.851189038517888e-05,
+      "loss": 1.1971,
       "step": 7070
     },
     {
-      "epoch": 0.19419955507950895,
+      "epoch": 0.20065266742338253,
       "grad_norm": 0.0,
-      "learning_rate": 1.8619197114323466e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.8511407962613855e-05,
+      "loss": 0.9045,
       "step": 7071
     },
     {
-      "epoch": 0.19422701930735217,
+      "epoch": 0.20068104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.861874605577203e-05,
-      "loss": 1.0946,
+      "learning_rate": 1.851092546815259e-05,
+      "loss": 0.9296,
       "step": 7072
     },
     {
-      "epoch": 0.19425448353519542,
+      "epoch": 0.2007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.861829492902574e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.8510442901799153e-05,
+      "loss": 0.8705,
       "step": 7073
     },
     {
-      "epoch": 0.19428194776303864,
+      "epoch": 0.2007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.8617843734088162e-05,
-      "loss": 1.1378,
+      "learning_rate": 1.850996026355762e-05,
+      "loss": 0.998,
       "step": 7074
     },
     {
-      "epoch": 0.19430941199088186,
+      "epoch": 0.20076617480136208,
       "grad_norm": 0.0,
-      "learning_rate": 1.8617392470962865e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.8509477553432073e-05,
+      "loss": 1.0594,
       "step": 7075
     },
     {
-      "epoch": 0.1943368762187251,
+      "epoch": 0.200794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8616941139653422e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.8508994771426585e-05,
+      "loss": 0.9392,
       "step": 7076
     },
     {
-      "epoch": 0.19436434044656833,
+      "epoch": 0.20082292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 1.86164897401634e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.8508511917545236e-05,
+      "loss": 1.0864,
       "step": 7077
     },
     {
-      "epoch": 0.19439180467441158,
+      "epoch": 0.20085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.8616038272496374e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.85080289917921e-05,
+      "loss": 0.9911,
       "step": 7078
     },
     {
-      "epoch": 0.1944192689022548,
+      "epoch": 0.20087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.8615586736655918e-05,
-      "loss": 1.094,
+      "learning_rate": 1.850754599417127e-05,
+      "loss": 1.0247,
       "step": 7079
     },
     {
-      "epoch": 0.19444673313009805,
+      "epoch": 0.20090805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.86151351326456e-05,
-      "loss": 1.0383,
+      "learning_rate": 1.850706292468681e-05,
+      "loss": 0.89,
       "step": 7080
     },
     {
-      "epoch": 0.19447419735794128,
+      "epoch": 0.20093643586833143,
       "grad_norm": 0.0,
-      "learning_rate": 1.8614683460468997e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.8506579783342808e-05,
+      "loss": 1.0307,
       "step": 7081
     },
     {
-      "epoch": 0.19450166158578452,
+      "epoch": 0.20096481271282635,
       "grad_norm": 0.0,
-      "learning_rate": 1.8614231720129682e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.8506096570143343e-05,
+      "loss": 1.0515,
       "step": 7082
     },
     {
-      "epoch": 0.19452912581362775,
+      "epoch": 0.20099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.8613779911631228e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.85056132850925e-05,
+      "loss": 0.8666,
       "step": 7083
     },
     {
-      "epoch": 0.194556590041471,
+      "epoch": 0.20102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.861332803497721e-05,
-      "loss": 1.0855,
+      "learning_rate": 1.850512992819436e-05,
+      "loss": 0.9247,
       "step": 7084
     },
     {
-      "epoch": 0.19458405426931422,
+      "epoch": 0.201049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8612876090171204e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.8504646499453003e-05,
+      "loss": 1.0104,
       "step": 7085
     },
     {
-      "epoch": 0.19461151849715747,
+      "epoch": 0.2010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.8612424077216788e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.8504162998872518e-05,
+      "loss": 1.0024,
       "step": 7086
     },
     {
-      "epoch": 0.1946389827250007,
+      "epoch": 0.20110669693530078,
       "grad_norm": 0.0,
-      "learning_rate": 1.8611971996117534e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.850367942645698e-05,
+      "loss": 0.9445,
       "step": 7087
     },
     {
-      "epoch": 0.1946664469528439,
+      "epoch": 0.2011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.8611519846877024e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.8503195782210484e-05,
+      "loss": 0.9563,
       "step": 7088
     },
     {
-      "epoch": 0.19469391118068716,
+      "epoch": 0.20116345062429059,
       "grad_norm": 0.0,
-      "learning_rate": 1.8611067629498832e-05,
-      "loss": 1.0819,
+      "learning_rate": 1.850271206613711e-05,
+      "loss": 1.0835,
       "step": 7089
     },
     {
-      "epoch": 0.19472137540853038,
+      "epoch": 0.20119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8610615343986536e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.8502228278240945e-05,
+      "loss": 1.0559,
       "step": 7090
     },
     {
-      "epoch": 0.19474883963637363,
+      "epoch": 0.20122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.8610162990343718e-05,
-      "loss": 1.052,
+      "learning_rate": 1.8501744418526074e-05,
+      "loss": 0.9026,
       "step": 7091
     },
     {
-      "epoch": 0.19477630386421685,
+      "epoch": 0.20124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.8609710568573956e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.850126048699659e-05,
+      "loss": 0.9872,
       "step": 7092
     },
     {
-      "epoch": 0.1948037680920601,
+      "epoch": 0.20127695800227013,
       "grad_norm": 0.0,
-      "learning_rate": 1.8609258078680826e-05,
-      "loss": 1.0983,
+      "learning_rate": 1.8500776483656574e-05,
+      "loss": 1.0699,
       "step": 7093
     },
     {
-      "epoch": 0.19483123231990332,
+      "epoch": 0.20130533484676505,
       "grad_norm": 0.0,
-      "learning_rate": 1.860880552066791e-05,
-      "loss": 1.1075,
+      "learning_rate": 1.8500292408510114e-05,
+      "loss": 0.9359,
       "step": 7094
     },
     {
-      "epoch": 0.19485869654774657,
+      "epoch": 0.20133371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.8608352894538795e-05,
-      "loss": 1.1226,
+      "learning_rate": 1.8499808261561308e-05,
+      "loss": 0.9484,
       "step": 7095
     },
     {
-      "epoch": 0.1948861607755898,
+      "epoch": 0.20136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.8607900200297053e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.8499324042814236e-05,
+      "loss": 1.0403,
       "step": 7096
     },
     {
-      "epoch": 0.19491362500343304,
+      "epoch": 0.2013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.8607447437946272e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.849883975227299e-05,
+      "loss": 0.9987,
       "step": 7097
     },
     {
-      "epoch": 0.19494108923127626,
+      "epoch": 0.2014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.8606994607490034e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.8498355389941666e-05,
+      "loss": 0.915,
       "step": 7098
     },
     {
-      "epoch": 0.1949685534591195,
+      "epoch": 0.2014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.860654170893192e-05,
-      "loss": 1.1241,
+      "learning_rate": 1.8497870955824347e-05,
+      "loss": 1.1957,
       "step": 7099
     },
     {
-      "epoch": 0.19499601768696273,
+      "epoch": 0.2014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.8606088742275516e-05,
-      "loss": 0.9062,
+      "learning_rate": 1.8497386449925137e-05,
+      "loss": 0.9969,
       "step": 7100
     },
     {
-      "epoch": 0.19502348191480595,
+      "epoch": 0.2015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.86056357075244e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.8496901872248117e-05,
+      "loss": 0.962,
       "step": 7101
     },
     {
-      "epoch": 0.1950509461426492,
+      "epoch": 0.20153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.860518260468216e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.8496417222797385e-05,
+      "loss": 1.0173,
       "step": 7102
     },
     {
-      "epoch": 0.19507841037049242,
+      "epoch": 0.20156072644721906,
       "grad_norm": 0.0,
-      "learning_rate": 1.8604729433752387e-05,
-      "loss": 1.0091,
+      "learning_rate": 1.8495932501577036e-05,
+      "loss": 1.022,
       "step": 7103
     },
     {
-      "epoch": 0.19510587459833567,
+      "epoch": 0.20158910329171395,
       "grad_norm": 0.0,
-      "learning_rate": 1.8604276194738657e-05,
-      "loss": 1.104,
+      "learning_rate": 1.8495447708591163e-05,
+      "loss": 1.0125,
       "step": 7104
     },
     {
-      "epoch": 0.1951333388261789,
+      "epoch": 0.20161748013620887,
       "grad_norm": 0.0,
-      "learning_rate": 1.8603822887644565e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.8494962843843862e-05,
+      "loss": 1.0168,
       "step": 7105
     },
     {
-      "epoch": 0.19516080305402214,
+      "epoch": 0.20164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.860336951247369e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.8494477907339224e-05,
+      "loss": 0.92,
       "step": 7106
     },
     {
-      "epoch": 0.19518826728186536,
+      "epoch": 0.20167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.8602916069229624e-05,
-      "loss": 1.0018,
+      "learning_rate": 1.8493992899081354e-05,
+      "loss": 0.965,
       "step": 7107
     },
     {
-      "epoch": 0.1952157315097086,
+      "epoch": 0.20170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.8602462557915953e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.8493507819074342e-05,
+      "loss": 0.8959,
       "step": 7108
     },
     {
-      "epoch": 0.19524319573755183,
+      "epoch": 0.20173098751418841,
       "grad_norm": 0.0,
-      "learning_rate": 1.8602008978536267e-05,
-      "loss": 0.918,
+      "learning_rate": 1.849302266732229e-05,
+      "loss": 1.0577,
       "step": 7109
     },
     {
-      "epoch": 0.19527065996539508,
+      "epoch": 0.2017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.8601555331094156e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.8492537443829293e-05,
+      "loss": 1.0115,
       "step": 7110
     },
     {
-      "epoch": 0.1952981241932383,
+      "epoch": 0.20178774120317822,
       "grad_norm": 0.0,
-      "learning_rate": 1.8601101615593205e-05,
-      "loss": 1.1054,
+      "learning_rate": 1.849205214859945e-05,
+      "loss": 1.1033,
       "step": 7111
     },
     {
-      "epoch": 0.19532558842108155,
+      "epoch": 0.2018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.8600647832037003e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.8491566781636862e-05,
+      "loss": 0.9323,
       "step": 7112
     },
     {
-      "epoch": 0.19535305264892477,
+      "epoch": 0.201844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.860019398042915e-05,
-      "loss": 1.0473,
+      "learning_rate": 1.8491081342945624e-05,
+      "loss": 0.892,
       "step": 7113
     },
     {
-      "epoch": 0.19538051687676802,
+      "epoch": 0.20187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.8599740060773225e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.8490595832529844e-05,
+      "loss": 0.9025,
       "step": 7114
     },
     {
-      "epoch": 0.19540798110461124,
+      "epoch": 0.20190124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.859928607307283e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.8490110250393617e-05,
+      "loss": 0.9708,
       "step": 7115
     },
     {
-      "epoch": 0.19543544533245447,
+      "epoch": 0.20192962542565268,
       "grad_norm": 0.0,
-      "learning_rate": 1.859883201733155e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.8489624596541045e-05,
+      "loss": 0.9973,
       "step": 7116
     },
     {
-      "epoch": 0.19546290956029772,
+      "epoch": 0.20195800227014757,
       "grad_norm": 0.0,
-      "learning_rate": 1.859837789355298e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.8489138870976234e-05,
+      "loss": 1.1145,
       "step": 7117
     },
     {
-      "epoch": 0.19549037378814094,
+      "epoch": 0.20198637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8597923701740718e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.8488653073703287e-05,
+      "loss": 1.0608,
       "step": 7118
     },
     {
-      "epoch": 0.19551783801598419,
+      "epoch": 0.20201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.8597469441898347e-05,
-      "loss": 1.1685,
+      "learning_rate": 1.8488167204726308e-05,
+      "loss": 0.9257,
       "step": 7119
     },
     {
-      "epoch": 0.1955453022438274,
+      "epoch": 0.20204313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 1.8597015114029472e-05,
-      "loss": 1.2692,
+      "learning_rate": 1.8487681264049396e-05,
+      "loss": 0.9438,
       "step": 7120
     },
     {
-      "epoch": 0.19557276647167066,
+      "epoch": 0.20207150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 1.859656071813768e-05,
-      "loss": 1.0455,
+      "learning_rate": 1.848719525167666e-05,
+      "loss": 1.0494,
       "step": 7121
     },
     {
-      "epoch": 0.19560023069951388,
+      "epoch": 0.20209988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 1.8596106254226568e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.8486709167612203e-05,
+      "loss": 1.0665,
       "step": 7122
     },
     {
-      "epoch": 0.19562769492735713,
+      "epoch": 0.20212826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 1.8595651722299738e-05,
-      "loss": 1.027,
+      "learning_rate": 1.848622301186013e-05,
+      "loss": 1.1099,
       "step": 7123
     },
     {
-      "epoch": 0.19565515915520035,
+      "epoch": 0.2021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.859519712236078e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.8485736784424554e-05,
+      "loss": 1.0774,
       "step": 7124
     },
     {
-      "epoch": 0.1956826233830436,
+      "epoch": 0.2021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8594742454413293e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.8485250485309578e-05,
+      "loss": 1.0442,
       "step": 7125
     },
     {
-      "epoch": 0.19571008761088682,
+      "epoch": 0.20221339387060158,
       "grad_norm": 0.0,
-      "learning_rate": 1.8594287718460875e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.8484764114519306e-05,
+      "loss": 0.9344,
       "step": 7126
     },
     {
-      "epoch": 0.19573755183873007,
+      "epoch": 0.20224177071509647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8593832914507123e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.848427767205785e-05,
+      "loss": 1.0901,
       "step": 7127
     },
     {
-      "epoch": 0.1957650160665733,
+      "epoch": 0.20227014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 1.8593378042555637e-05,
-      "loss": 1.0706,
+      "learning_rate": 1.8483791157929323e-05,
+      "loss": 0.9547,
       "step": 7128
     },
     {
-      "epoch": 0.1957924802944165,
+      "epoch": 0.20229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.8592923102610014e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.8483304572137827e-05,
+      "loss": 1.0941,
       "step": 7129
     },
     {
-      "epoch": 0.19581994452225976,
+      "epoch": 0.20232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.8592468094673856e-05,
-      "loss": 0.936,
+      "learning_rate": 1.848281791468748e-05,
+      "loss": 1.0918,
       "step": 7130
     },
     {
-      "epoch": 0.19584740875010298,
+      "epoch": 0.20235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8592013018750758e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.8482331185582383e-05,
+      "loss": 0.9697,
       "step": 7131
     },
     {
-      "epoch": 0.19587487297794623,
+      "epoch": 0.20238365493757093,
       "grad_norm": 0.0,
-      "learning_rate": 1.8591557874844328e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.8481844384826658e-05,
+      "loss": 0.9608,
       "step": 7132
     },
     {
-      "epoch": 0.19590233720578945,
+      "epoch": 0.20241203178206582,
       "grad_norm": 0.0,
-      "learning_rate": 1.8591102662958164e-05,
-      "loss": 0.8645,
+      "learning_rate": 1.848135751242441e-05,
+      "loss": 1.0974,
       "step": 7133
     },
     {
-      "epoch": 0.1959298014336327,
+      "epoch": 0.20244040862656074,
       "grad_norm": 0.0,
-      "learning_rate": 1.8590647383095868e-05,
-      "loss": 0.977,
+      "learning_rate": 1.8480870568379757e-05,
+      "loss": 1.0432,
       "step": 7134
     },
     {
-      "epoch": 0.19595726566147592,
+      "epoch": 0.20246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.859019203526104e-05,
-      "loss": 0.951,
+      "learning_rate": 1.8480383552696805e-05,
+      "loss": 1.0049,
       "step": 7135
     },
     {
-      "epoch": 0.19598472988931917,
+      "epoch": 0.2024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.858973661945729e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.8479896465379676e-05,
+      "loss": 0.9692,
       "step": 7136
     },
     {
-      "epoch": 0.1960121941171624,
+      "epoch": 0.2025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.8589281135688213e-05,
-      "loss": 1.0658,
+      "learning_rate": 1.8479409306432474e-05,
+      "loss": 0.9767,
       "step": 7137
     },
     {
-      "epoch": 0.19603965834500564,
+      "epoch": 0.20255391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 1.8588825583957418e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.8478922075859326e-05,
+      "loss": 1.1163,
       "step": 7138
     },
     {
-      "epoch": 0.19606712257284886,
+      "epoch": 0.2025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.8588369964268506e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.847843477366434e-05,
+      "loss": 0.9914,
       "step": 7139
     },
     {
-      "epoch": 0.1960945868006921,
+      "epoch": 0.2026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.8587914276625084e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.8477947399851633e-05,
+      "loss": 0.8989,
       "step": 7140
     },
     {
-      "epoch": 0.19612205102853533,
+      "epoch": 0.20263904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 1.858745852103076e-05,
-      "loss": 1.048,
+      "learning_rate": 1.847745995442533e-05,
+      "loss": 1.041,
       "step": 7141
     },
     {
-      "epoch": 0.19614951525637855,
+      "epoch": 0.20266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.8587002697489137e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.8476972437389535e-05,
+      "loss": 1.0576,
       "step": 7142
     },
     {
-      "epoch": 0.1961769794842218,
+      "epoch": 0.20269580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.858654680600382e-05,
-      "loss": 1.0593,
+      "learning_rate": 1.8476484848748373e-05,
+      "loss": 1.004,
       "step": 7143
     },
     {
-      "epoch": 0.19620444371206502,
+      "epoch": 0.20272417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.8586090846578423e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.8475997188505962e-05,
+      "loss": 0.9675,
       "step": 7144
     },
     {
-      "epoch": 0.19623190793990827,
+      "epoch": 0.20275255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 1.8585634819216545e-05,
-      "loss": 1.0018,
+      "learning_rate": 1.8475509456666423e-05,
+      "loss": 1.0638,
       "step": 7145
     },
     {
-      "epoch": 0.1962593721677515,
+      "epoch": 0.20278093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.85851787239218e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.847502165323388e-05,
+      "loss": 0.9679,
       "step": 7146
     },
     {
-      "epoch": 0.19628683639559474,
+      "epoch": 0.20280930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 1.8584722560697798e-05,
-      "loss": 0.964,
+      "learning_rate": 1.847453377821244e-05,
+      "loss": 0.9399,
       "step": 7147
     },
     {
-      "epoch": 0.19631430062343797,
+      "epoch": 0.20283768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 1.858426632954814e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.8474045831606234e-05,
+      "loss": 1.0891,
       "step": 7148
     },
     {
-      "epoch": 0.19634176485128121,
+      "epoch": 0.2028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.8583810030476446e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.8473557813419388e-05,
+      "loss": 1.0695,
       "step": 7149
     },
     {
-      "epoch": 0.19636922907912444,
+      "epoch": 0.202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.858335366348632e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.847306972365601e-05,
+      "loss": 0.9977,
       "step": 7150
     },
     {
-      "epoch": 0.19639669330696768,
+      "epoch": 0.2029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.8582897228581375e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.8472581562320235e-05,
+      "loss": 1.0023,
       "step": 7151
     },
     {
-      "epoch": 0.1964241575348109,
+      "epoch": 0.2029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.8582440725765222e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.847209332941618e-05,
+      "loss": 0.9994,
       "step": 7152
     },
     {
-      "epoch": 0.19645162176265415,
+      "epoch": 0.20297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.8581984155041475e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.8471605024947978e-05,
+      "loss": 0.9914,
       "step": 7153
     },
     {
-      "epoch": 0.19647908599049738,
+      "epoch": 0.20300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8581527516413747e-05,
-      "loss": 1.1195,
+      "learning_rate": 1.8471116648919744e-05,
+      "loss": 1.0322,
       "step": 7154
     },
     {
-      "epoch": 0.1965065502183406,
+      "epoch": 0.20303632236095345,
       "grad_norm": 0.0,
-      "learning_rate": 1.8581070809885645e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.8470628201335604e-05,
+      "loss": 0.901,
       "step": 7155
     },
     {
-      "epoch": 0.19653401444618385,
+      "epoch": 0.20306469920544837,
       "grad_norm": 0.0,
-      "learning_rate": 1.8580614035460788e-05,
-      "loss": 0.97,
+      "learning_rate": 1.8470139682199693e-05,
+      "loss": 1.0494,
       "step": 7156
     },
     {
-      "epoch": 0.19656147867402707,
+      "epoch": 0.20309307604994326,
       "grad_norm": 0.0,
-      "learning_rate": 1.8580157193142788e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.8469651091516126e-05,
+      "loss": 1.0934,
       "step": 7157
     },
     {
-      "epoch": 0.19658894290187032,
+      "epoch": 0.20312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.8579700282935263e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.846916242928904e-05,
+      "loss": 0.9818,
       "step": 7158
     },
     {
-      "epoch": 0.19661640712971354,
+      "epoch": 0.20314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8579243304841826e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.8468673695522552e-05,
+      "loss": 1.1008,
       "step": 7159
     },
     {
-      "epoch": 0.1966438713575568,
+      "epoch": 0.20317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.8578786258866094e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.8468184890220803e-05,
+      "loss": 0.992,
       "step": 7160
     },
     {
-      "epoch": 0.1966713355854,
+      "epoch": 0.2032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.857832914501168e-05,
-      "loss": 0.912,
+      "learning_rate": 1.846769601338791e-05,
+      "loss": 1.0183,
       "step": 7161
     },
     {
-      "epoch": 0.19669879981324326,
+      "epoch": 0.20323496027241772,
       "grad_norm": 0.0,
-      "learning_rate": 1.85778719632822e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.846720706502801e-05,
+      "loss": 1.0741,
       "step": 7162
     },
     {
-      "epoch": 0.19672626404108648,
+      "epoch": 0.2032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.8577414713681275e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.8466718045145233e-05,
+      "loss": 1.0897,
       "step": 7163
     },
     {
-      "epoch": 0.19675372826892973,
+      "epoch": 0.2032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.8576957396212523e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.846622895374371e-05,
+      "loss": 1.0526,
       "step": 7164
     },
     {
-      "epoch": 0.19678119249677295,
+      "epoch": 0.20332009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.8576500010879566e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.8465739790827566e-05,
+      "loss": 0.9948,
       "step": 7165
     },
     {
-      "epoch": 0.1968086567246162,
+      "epoch": 0.20334846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 1.8576042557686013e-05,
-      "loss": 1.018,
+      "learning_rate": 1.8465250556400938e-05,
+      "loss": 1.0976,
       "step": 7166
     },
     {
-      "epoch": 0.19683612095245942,
+      "epoch": 0.20337684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 1.8575585036635494e-05,
-      "loss": 1.1467,
+      "learning_rate": 1.846476125046796e-05,
+      "loss": 0.8618,
       "step": 7167
     },
     {
-      "epoch": 0.19686358518030267,
+      "epoch": 0.20340522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 1.8575127447731623e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.8464271873032762e-05,
+      "loss": 1.086,
       "step": 7168
     },
     {
-      "epoch": 0.1968910494081459,
+      "epoch": 0.20343359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.857466979097802e-05,
-      "loss": 0.9635,
+      "learning_rate": 1.8463782424099478e-05,
+      "loss": 1.0213,
       "step": 7169
     },
     {
-      "epoch": 0.1969185136359891,
+      "epoch": 0.20346197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.857421206637831e-05,
-      "loss": 0.986,
+      "learning_rate": 1.8463292903672246e-05,
+      "loss": 1.054,
       "step": 7170
     },
     {
-      "epoch": 0.19694597786383236,
+      "epoch": 0.20349035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 1.8573754273936112e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.8462803311755196e-05,
+      "loss": 0.9692,
       "step": 7171
     },
     {
-      "epoch": 0.19697344209167558,
+      "epoch": 0.20351872871736662,
       "grad_norm": 0.0,
-      "learning_rate": 1.8573296413655048e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.846231364835247e-05,
+      "loss": 1.0312,
       "step": 7172
     },
     {
-      "epoch": 0.19700090631951883,
+      "epoch": 0.2035471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.8572838485538743e-05,
-      "loss": 1.1317,
+      "learning_rate": 1.8461823913468194e-05,
+      "loss": 0.9082,
       "step": 7173
     },
     {
-      "epoch": 0.19702837054736205,
+      "epoch": 0.20357548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.8572380489590817e-05,
-      "loss": 1.1864,
+      "learning_rate": 1.8461334107106515e-05,
+      "loss": 0.9216,
       "step": 7174
     },
     {
-      "epoch": 0.1970558347752053,
+      "epoch": 0.2036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.85719224258149e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.8460844229271566e-05,
+      "loss": 1.0453,
       "step": 7175
     },
     {
-      "epoch": 0.19708329900304852,
+      "epoch": 0.2036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.857146429421461e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.8460354279967484e-05,
+      "loss": 0.9645,
       "step": 7176
     },
     {
-      "epoch": 0.19711076323089177,
+      "epoch": 0.20366061293984108,
       "grad_norm": 0.0,
-      "learning_rate": 1.8571006094793573e-05,
-      "loss": 1.1302,
+      "learning_rate": 1.845986425919841e-05,
+      "loss": 0.9614,
       "step": 7177
     },
     {
-      "epoch": 0.197138227458735,
+      "epoch": 0.20368898978433597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8570547827555418e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.8459374166968483e-05,
+      "loss": 1.0727,
       "step": 7178
     },
     {
-      "epoch": 0.19716569168657824,
+      "epoch": 0.2037173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.8570089492503767e-05,
-      "loss": 0.9176,
+      "learning_rate": 1.8458884003281846e-05,
+      "loss": 0.9723,
       "step": 7179
     },
     {
-      "epoch": 0.19719315591442146,
+      "epoch": 0.20374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.856963108964225e-05,
-      "loss": 0.97,
+      "learning_rate": 1.845839376814263e-05,
+      "loss": 1.0656,
       "step": 7180
     },
     {
-      "epoch": 0.1972206201422647,
+      "epoch": 0.20377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8569172618974488e-05,
-      "loss": 1.0822,
+      "learning_rate": 1.845790346155498e-05,
+      "loss": 0.9467,
       "step": 7181
     },
     {
-      "epoch": 0.19724808437010793,
+      "epoch": 0.20380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.8568714080504118e-05,
-      "loss": 1.038,
+      "learning_rate": 1.8457413083523045e-05,
+      "loss": 1.0159,
       "step": 7182
     },
     {
-      "epoch": 0.19727554859795116,
+      "epoch": 0.20383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.8568255474234763e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.8456922634050957e-05,
+      "loss": 0.8266,
       "step": 7183
     },
     {
-      "epoch": 0.1973030128257944,
+      "epoch": 0.20385925085130532,
       "grad_norm": 0.0,
-      "learning_rate": 1.8567796800170048e-05,
-      "loss": 0.97,
+      "learning_rate": 1.8456432113142866e-05,
+      "loss": 0.951,
       "step": 7184
     },
     {
-      "epoch": 0.19733047705363763,
+      "epoch": 0.20388762769580024,
       "grad_norm": 0.0,
-      "learning_rate": 1.8567338058313605e-05,
-      "loss": 1.047,
+      "learning_rate": 1.845594152080291e-05,
+      "loss": 0.9202,
       "step": 7185
     },
     {
-      "epoch": 0.19735794128148088,
+      "epoch": 0.20391600454029513,
       "grad_norm": 0.0,
-      "learning_rate": 1.856687924866907e-05,
-      "loss": 1.149,
+      "learning_rate": 1.845545085703524e-05,
+      "loss": 0.9465,
       "step": 7186
     },
     {
-      "epoch": 0.1973854055093241,
+      "epoch": 0.20394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.8566420371240064e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.845496012184399e-05,
+      "loss": 1.0374,
       "step": 7187
     },
     {
-      "epoch": 0.19741286973716735,
+      "epoch": 0.2039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.856596142603022e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.845446931523332e-05,
+      "loss": 1.0287,
       "step": 7188
     },
     {
-      "epoch": 0.19744033396501057,
+      "epoch": 0.2040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.8565502413043175e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.845397843720736e-05,
+      "loss": 1.0651,
       "step": 7189
     },
     {
-      "epoch": 0.19746779819285382,
+      "epoch": 0.20402951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 1.856504333228256e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.845348748777027e-05,
+      "loss": 1.0349,
       "step": 7190
     },
     {
-      "epoch": 0.19749526242069704,
+      "epoch": 0.2040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.8564584183752e-05,
-      "loss": 1.0537,
+      "learning_rate": 1.845299646692619e-05,
+      "loss": 1.1381,
       "step": 7191
     },
     {
-      "epoch": 0.1975227266485403,
+      "epoch": 0.20408626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 1.856412496745513e-05,
-      "loss": 1.139,
+      "learning_rate": 1.8452505374679265e-05,
+      "loss": 1.024,
       "step": 7192
     },
     {
-      "epoch": 0.1975501908763835,
+      "epoch": 0.20411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.8563665683395588e-05,
-      "loss": 0.9394,
+      "learning_rate": 1.845201421103365e-05,
+      "loss": 0.9756,
       "step": 7193
     },
     {
-      "epoch": 0.19757765510422676,
+      "epoch": 0.20414301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.8563206331577005e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.845152297599349e-05,
+      "loss": 0.8821,
       "step": 7194
     },
     {
-      "epoch": 0.19760511933206998,
+      "epoch": 0.20417139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 1.856274691200302e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.8451031669562938e-05,
+      "loss": 0.9946,
       "step": 7195
     },
     {
-      "epoch": 0.1976325835599132,
+      "epoch": 0.20419977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 1.8562287424677262e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.8450540291746138e-05,
+      "loss": 0.9404,
       "step": 7196
     },
     {
-      "epoch": 0.19766004778775645,
+      "epoch": 0.20422814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.856182786960337e-05,
-      "loss": 0.9782,
+      "learning_rate": 1.8450048842547246e-05,
+      "loss": 0.9444,
       "step": 7197
     },
     {
-      "epoch": 0.19768751201559967,
+      "epoch": 0.20425652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.8561368246784978e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.8449557321970416e-05,
+      "loss": 0.9589,
       "step": 7198
     },
     {
-      "epoch": 0.19771497624344292,
+      "epoch": 0.20428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.8560908556225726e-05,
-      "loss": 0.8938,
+      "learning_rate": 1.8449065730019788e-05,
+      "loss": 1.0627,
       "step": 7199
     },
     {
-      "epoch": 0.19774244047128614,
+      "epoch": 0.2043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.856044879792925e-05,
-      "loss": 1.1297,
+      "learning_rate": 1.844857406669953e-05,
+      "loss": 1.0215,
       "step": 7200
     },
     {
-      "epoch": 0.1977699046991294,
+      "epoch": 0.2043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.8559988971899182e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.844808233201378e-05,
+      "loss": 1.044,
       "step": 7201
     },
     {
-      "epoch": 0.1977973689269726,
+      "epoch": 0.2043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.8559529078139166e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.84475905259667e-05,
+      "loss": 1.0236,
       "step": 7202
     },
     {
-      "epoch": 0.19782483315481586,
+      "epoch": 0.2043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.8559069116652844e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.8447098648562444e-05,
+      "loss": 1.0124,
       "step": 7203
     },
     {
-      "epoch": 0.19785229738265908,
+      "epoch": 0.20442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.855860908744385e-05,
-      "loss": 0.951,
+      "learning_rate": 1.8446606699805164e-05,
+      "loss": 1.0471,
       "step": 7204
     },
     {
-      "epoch": 0.19787976161050233,
+      "epoch": 0.20445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8558148990515828e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.844611467969902e-05,
+      "loss": 0.9612,
       "step": 7205
     },
     {
-      "epoch": 0.19790722583834555,
+      "epoch": 0.20448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.855768882587241e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.8445622588248168e-05,
+      "loss": 1.0225,
       "step": 7206
     },
     {
-      "epoch": 0.1979346900661888,
+      "epoch": 0.20451191827468784,
       "grad_norm": 0.0,
-      "learning_rate": 1.8557228593517252e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.8445130425456758e-05,
+      "loss": 0.9797,
       "step": 7207
     },
     {
-      "epoch": 0.19796215429403202,
+      "epoch": 0.20454029511918276,
       "grad_norm": 0.0,
-      "learning_rate": 1.8556768293453983e-05,
-      "loss": 0.945,
+      "learning_rate": 1.844463819132895e-05,
+      "loss": 1.024,
       "step": 7208
     },
     {
-      "epoch": 0.19798961852187527,
+      "epoch": 0.20456867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8556307925686244e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.8444145885868908e-05,
+      "loss": 1.0389,
       "step": 7209
     },
     {
-      "epoch": 0.1980170827497185,
+      "epoch": 0.20459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.8555847490217687e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.8443653509080787e-05,
+      "loss": 1.052,
       "step": 7210
     },
     {
-      "epoch": 0.19804454697756171,
+      "epoch": 0.20462542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.855538698705195e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.844316106096874e-05,
+      "loss": 0.9367,
       "step": 7211
     },
     {
-      "epoch": 0.19807201120540496,
+      "epoch": 0.2046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.8554926416192676e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.844266854153694e-05,
+      "loss": 1.0263,
       "step": 7212
     },
     {
-      "epoch": 0.19809947543324818,
+      "epoch": 0.2046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.855446577764351e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.8442175950789533e-05,
+      "loss": 0.9564,
       "step": 7213
     },
     {
-      "epoch": 0.19812693966109143,
+      "epoch": 0.2047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.85540050714081e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.8441683288730686e-05,
+      "loss": 0.901,
       "step": 7214
     },
     {
-      "epoch": 0.19815440388893465,
+      "epoch": 0.204738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8553544297490085e-05,
-      "loss": 1.1012,
+      "learning_rate": 1.8441190555364567e-05,
+      "loss": 1.0353,
       "step": 7215
     },
     {
-      "epoch": 0.1981818681167779,
+      "epoch": 0.20476730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.8553083455893116e-05,
-      "loss": 1.0519,
+      "learning_rate": 1.8440697750695326e-05,
+      "loss": 1.0619,
       "step": 7216
     },
     {
-      "epoch": 0.19820933234462113,
+      "epoch": 0.20479568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 1.8552622546620836e-05,
-      "loss": 1.1481,
+      "learning_rate": 1.844020487472713e-05,
+      "loss": 1.0202,
       "step": 7217
     },
     {
-      "epoch": 0.19823679657246437,
+      "epoch": 0.20482406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 1.8552161569676893e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.843971192746415e-05,
+      "loss": 0.9587,
       "step": 7218
     },
     {
-      "epoch": 0.1982642608003076,
+      "epoch": 0.20485244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 1.8551700525064937e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.8439218908910538e-05,
+      "loss": 1.0399,
       "step": 7219
     },
     {
-      "epoch": 0.19829172502815084,
+      "epoch": 0.20488081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.8551239412788615e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.8438725819070467e-05,
+      "loss": 0.9738,
       "step": 7220
     },
     {
-      "epoch": 0.19831918925599407,
+      "epoch": 0.20490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.855077823285157e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.8438232657948102e-05,
+      "loss": 1.0553,
       "step": 7221
     },
     {
-      "epoch": 0.19834665348383732,
+      "epoch": 0.20493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.855031698525746e-05,
-      "loss": 1.2288,
+      "learning_rate": 1.8437739425547602e-05,
+      "loss": 0.9926,
       "step": 7222
     },
     {
-      "epoch": 0.19837411771168054,
+      "epoch": 0.20496594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 1.8549855670009926e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.8437246121873143e-05,
+      "loss": 1.0064,
       "step": 7223
     },
     {
-      "epoch": 0.19840158193952376,
+      "epoch": 0.204994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.8549394287112625e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.8436752746928884e-05,
+      "loss": 0.9013,
       "step": 7224
     },
     {
-      "epoch": 0.198429046167367,
+      "epoch": 0.20502270147559593,
       "grad_norm": 0.0,
-      "learning_rate": 1.8548932836569206e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.8436259300718996e-05,
+      "loss": 1.04,
       "step": 7225
     },
     {
-      "epoch": 0.19845651039521023,
+      "epoch": 0.2050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.8548471318383314e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.8435765783247644e-05,
+      "loss": 1.0026,
       "step": 7226
     },
     {
-      "epoch": 0.19848397462305348,
+      "epoch": 0.2050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.854800973255861e-05,
-      "loss": 1.0612,
+      "learning_rate": 1.8435272194519002e-05,
+      "loss": 1.0098,
       "step": 7227
     },
     {
-      "epoch": 0.1985114388508967,
+      "epoch": 0.2051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.8547548079098738e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.8434778534537233e-05,
+      "loss": 1.0338,
       "step": 7228
     },
     {
-      "epoch": 0.19853890307873995,
+      "epoch": 0.20513620885357547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8547086358007358e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.8434284803306515e-05,
+      "loss": 1.0285,
       "step": 7229
     },
     {
-      "epoch": 0.19856636730658317,
+      "epoch": 0.20516458569807036,
       "grad_norm": 0.0,
-      "learning_rate": 1.8546624569288115e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.8433791000831012e-05,
+      "loss": 0.9766,
       "step": 7230
     },
     {
-      "epoch": 0.19859383153442642,
+      "epoch": 0.20519296254256528,
       "grad_norm": 0.0,
-      "learning_rate": 1.8546162712944672e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.84332971271149e-05,
+      "loss": 0.9423,
       "step": 7231
     },
     {
-      "epoch": 0.19862129576226964,
+      "epoch": 0.20522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.8545700788980678e-05,
-      "loss": 1.1136,
+      "learning_rate": 1.8432803182162346e-05,
+      "loss": 1.0652,
       "step": 7232
     },
     {
-      "epoch": 0.1986487599901129,
+      "epoch": 0.20524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.8545238797399792e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.8432309165977523e-05,
+      "loss": 1.0242,
       "step": 7233
     },
     {
-      "epoch": 0.1986762242179561,
+      "epoch": 0.20527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.8544776738205663e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.8431815078564606e-05,
+      "loss": 1.0087,
       "step": 7234
     },
     {
-      "epoch": 0.19870368844579936,
+      "epoch": 0.20530646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.854431461140195e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.8431320919927768e-05,
+      "loss": 0.9839,
       "step": 7235
     },
     {
-      "epoch": 0.19873115267364258,
+      "epoch": 0.20533484676503974,
       "grad_norm": 0.0,
-      "learning_rate": 1.8543852416992313e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.843082669007118e-05,
+      "loss": 0.9346,
       "step": 7236
     },
     {
-      "epoch": 0.1987586169014858,
+      "epoch": 0.20536322360953463,
       "grad_norm": 0.0,
-      "learning_rate": 1.8543390154980404e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.8430332388999027e-05,
+      "loss": 0.9331,
       "step": 7237
     },
     {
-      "epoch": 0.19878608112932905,
+      "epoch": 0.20539160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 1.8542927825369882e-05,
-      "loss": 1.1245,
+      "learning_rate": 1.8429838016715472e-05,
+      "loss": 1.0072,
       "step": 7238
     },
     {
-      "epoch": 0.19881354535717227,
+      "epoch": 0.2054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.8542465428164406e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.84293435732247e-05,
+      "loss": 1.089,
       "step": 7239
     },
     {
-      "epoch": 0.19884100958501552,
+      "epoch": 0.2054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.8542002963367636e-05,
-      "loss": 0.929,
+      "learning_rate": 1.842884905853088e-05,
+      "loss": 1.0562,
       "step": 7240
     },
     {
-      "epoch": 0.19886847381285874,
+      "epoch": 0.20547673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 1.8541540430983233e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.84283544726382e-05,
+      "loss": 0.9523,
       "step": 7241
     },
     {
-      "epoch": 0.198895938040702,
+      "epoch": 0.2055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.8541077831014848e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.8427859815550824e-05,
+      "loss": 1.0175,
       "step": 7242
     },
     {
-      "epoch": 0.1989234022685452,
+      "epoch": 0.20553348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 1.854061516346615e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.842736508727294e-05,
+      "loss": 1.054,
       "step": 7243
     },
     {
-      "epoch": 0.19895086649638846,
+      "epoch": 0.20556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.854015242834079e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.8426870287808723e-05,
+      "loss": 1.0735,
       "step": 7244
     },
     {
-      "epoch": 0.19897833072423168,
+      "epoch": 0.20559023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.853968962564244e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.8426375417162353e-05,
+      "loss": 1.0208,
       "step": 7245
     },
     {
-      "epoch": 0.19900579495207493,
+      "epoch": 0.20561861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 1.8539226755374758e-05,
-      "loss": 1.0215,
+      "learning_rate": 1.8425880475338015e-05,
+      "loss": 0.9774,
       "step": 7246
     },
     {
-      "epoch": 0.19903325917991815,
+      "epoch": 0.20564699205448353,
       "grad_norm": 0.0,
-      "learning_rate": 1.853876381754141e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.8425385462339884e-05,
+      "loss": 0.99,
       "step": 7247
     },
     {
-      "epoch": 0.1990607234077614,
+      "epoch": 0.20567536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.8538300812146047e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.842489037817214e-05,
+      "loss": 1.0305,
       "step": 7248
     },
     {
-      "epoch": 0.19908818763560462,
+      "epoch": 0.20570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.853783773919235e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.8424395222838972e-05,
+      "loss": 0.9762,
       "step": 7249
     },
     {
-      "epoch": 0.19911565186344785,
+      "epoch": 0.20573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.8537374598683967e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.842389999634456e-05,
+      "loss": 0.8121,
       "step": 7250
     },
     {
-      "epoch": 0.1991431160912911,
+      "epoch": 0.2057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.8536911390624574e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.8423404698693082e-05,
+      "loss": 0.9826,
       "step": 7251
     },
     {
-      "epoch": 0.19917058031913432,
+      "epoch": 0.205788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.8536448115017824e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.842290932988873e-05,
+      "loss": 0.9297,
       "step": 7252
     },
     {
-      "epoch": 0.19919804454697757,
+      "epoch": 0.20581725312145288,
       "grad_norm": 0.0,
-      "learning_rate": 1.8535984771867396e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.8422413889935678e-05,
+      "loss": 0.8785,
       "step": 7253
     },
     {
-      "epoch": 0.1992255087748208,
+      "epoch": 0.2058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.8535521361176946e-05,
-      "loss": 1.0495,
+      "learning_rate": 1.8421918378838125e-05,
+      "loss": 1.0914,
       "step": 7254
     },
     {
-      "epoch": 0.19925297300266404,
+      "epoch": 0.20587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.8535057882950147e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.8421422796600243e-05,
+      "loss": 1.0403,
       "step": 7255
     },
     {
-      "epoch": 0.19928043723050726,
+      "epoch": 0.20590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.8534594337190664e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.8420927143226226e-05,
+      "loss": 1.015,
       "step": 7256
     },
     {
-      "epoch": 0.1993079014583505,
+      "epoch": 0.20593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.8534130723902165e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.842043141872026e-05,
+      "loss": 1.0742,
       "step": 7257
     },
     {
-      "epoch": 0.19933536568619373,
+      "epoch": 0.20595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.8533667043088313e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.8419935623086532e-05,
+      "loss": 0.9916,
       "step": 7258
     },
     {
-      "epoch": 0.19936282991403698,
+      "epoch": 0.20598751418842226,
       "grad_norm": 0.0,
-      "learning_rate": 1.8533203294752787e-05,
-      "loss": 1.0299,
+      "learning_rate": 1.8419439756329226e-05,
+      "loss": 1.043,
       "step": 7259
     },
     {
-      "epoch": 0.1993902941418802,
+      "epoch": 0.20601589103291715,
       "grad_norm": 0.0,
-      "learning_rate": 1.853273947889925e-05,
-      "loss": 0.8716,
+      "learning_rate": 1.8418943818452536e-05,
+      "loss": 0.9937,
       "step": 7260
     },
     {
-      "epoch": 0.19941775836972345,
+      "epoch": 0.20604426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.8532275595531373e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.8418447809460648e-05,
+      "loss": 1.0833,
       "step": 7261
     },
     {
-      "epoch": 0.19944522259756667,
+      "epoch": 0.20607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.8531811644652827e-05,
-      "loss": 1.0183,
+      "learning_rate": 1.841795172935775e-05,
+      "loss": 0.925,
       "step": 7262
     },
     {
-      "epoch": 0.19947268682540992,
+      "epoch": 0.2061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.853134762626728e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.8417455578148042e-05,
+      "loss": 0.8857,
       "step": 7263
     },
     {
-      "epoch": 0.19950015105325314,
+      "epoch": 0.2061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.8530883540378407e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.8416959355835703e-05,
+      "loss": 1.0507,
       "step": 7264
     },
     {
-      "epoch": 0.19952761528109636,
+      "epoch": 0.2061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.853041938698988e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.8416463062424933e-05,
+      "loss": 0.9826,
       "step": 7265
     },
     {
-      "epoch": 0.1995550795089396,
+      "epoch": 0.2061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.8529955166105367e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.8415966697919924e-05,
+      "loss": 0.9995,
       "step": 7266
     },
     {
-      "epoch": 0.19958254373678283,
+      "epoch": 0.2062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.852949087772855e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.841547026232486e-05,
+      "loss": 1.024,
       "step": 7267
     },
     {
-      "epoch": 0.19961000796462608,
+      "epoch": 0.20624290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 1.8529026521863093e-05,
-      "loss": 1.1062,
+      "learning_rate": 1.8414973755643942e-05,
+      "loss": 1.121,
       "step": 7268
     },
     {
-      "epoch": 0.1996374721924693,
+      "epoch": 0.20627128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 1.8528562098512676e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.8414477177881366e-05,
+      "loss": 1.0197,
       "step": 7269
     },
     {
-      "epoch": 0.19966493642031255,
+      "epoch": 0.20629965947786605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8528097607680974e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.8413980529041318e-05,
+      "loss": 0.9521,
       "step": 7270
     },
     {
-      "epoch": 0.19969240064815577,
+      "epoch": 0.20632803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.8527633049371655e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.8413483809128002e-05,
+      "loss": 0.9537,
       "step": 7271
     },
     {
-      "epoch": 0.19971986487599902,
+      "epoch": 0.20635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.85271684235884e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.8412987018145607e-05,
+      "loss": 1.0333,
       "step": 7272
     },
     {
-      "epoch": 0.19974732910384224,
+      "epoch": 0.20638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.8526703730334892e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.8412490156098336e-05,
+      "loss": 0.8684,
       "step": 7273
     },
     {
-      "epoch": 0.1997747933316855,
+      "epoch": 0.20641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.85262389696148e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.8411993222990378e-05,
+      "loss": 1.0895,
       "step": 7274
     },
     {
-      "epoch": 0.1998022575595287,
+      "epoch": 0.2064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.8525774141431795e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.8411496218825938e-05,
+      "loss": 0.9418,
       "step": 7275
     },
     {
-      "epoch": 0.19982972178737196,
+      "epoch": 0.20646992054483543,
       "grad_norm": 0.0,
-      "learning_rate": 1.852530924578957e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.841099914360921e-05,
+      "loss": 0.9363,
       "step": 7276
     },
     {
-      "epoch": 0.19985718601521518,
+      "epoch": 0.20649829738933032,
       "grad_norm": 0.0,
-      "learning_rate": 1.8524844282691794e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.8410501997344394e-05,
+      "loss": 1.0569,
       "step": 7277
     },
     {
-      "epoch": 0.1998846502430584,
+      "epoch": 0.2065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8524379252142143e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.841000478003569e-05,
+      "loss": 1.033,
       "step": 7278
     },
     {
-      "epoch": 0.19991211447090165,
+      "epoch": 0.2065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.8523914154144308e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.84095074916873e-05,
+      "loss": 0.9981,
       "step": 7279
     },
     {
-      "epoch": 0.19993957869874487,
+      "epoch": 0.20658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.852344898870196e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.840901013230342e-05,
+      "loss": 1.0193,
       "step": 7280
     },
     {
-      "epoch": 0.19996704292658812,
+      "epoch": 0.20661180476730986,
       "grad_norm": 0.0,
-      "learning_rate": 1.852298375581878e-05,
-      "loss": 0.9542,
+      "learning_rate": 1.8408512701888256e-05,
+      "loss": 1.0761,
       "step": 7281
     },
     {
-      "epoch": 0.19999450715443134,
+      "epoch": 0.20664018161180478,
       "grad_norm": 0.0,
-      "learning_rate": 1.8522518455498452e-05,
-      "loss": 1.0472,
+      "learning_rate": 1.8408015200446e-05,
+      "loss": 0.8506,
       "step": 7282
     },
     {
-      "epoch": 0.2000219713822746,
+      "epoch": 0.20666855845629967,
       "grad_norm": 0.0,
-      "learning_rate": 1.852205308774466e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.840751762798087e-05,
+      "loss": 0.9056,
       "step": 7283
     },
     {
-      "epoch": 0.20004943561011782,
+      "epoch": 0.20669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.852158765256108e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.8407019984497057e-05,
+      "loss": 1.1485,
       "step": 7284
     },
     {
-      "epoch": 0.20007689983796106,
+      "epoch": 0.20672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.8521122149951396e-05,
-      "loss": 0.9521,
+      "learning_rate": 1.8406522269998766e-05,
+      "loss": 1.0279,
       "step": 7285
     },
     {
-      "epoch": 0.20010436406580429,
+      "epoch": 0.20675368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 1.8520656579919295e-05,
-      "loss": 0.9011,
+      "learning_rate": 1.8406024484490207e-05,
+      "loss": 1.0334,
       "step": 7286
     },
     {
-      "epoch": 0.20013182829364753,
+      "epoch": 0.20678206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 1.852019094246846e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.840552662797558e-05,
+      "loss": 0.9753,
       "step": 7287
     },
     {
-      "epoch": 0.20015929252149076,
+      "epoch": 0.20681044267877413,
       "grad_norm": 0.0,
-      "learning_rate": 1.851972523760257e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.840502870045909e-05,
+      "loss": 0.972,
       "step": 7288
     },
     {
-      "epoch": 0.200186756749334,
+      "epoch": 0.20683881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 1.8519259465325317e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.840453070194495e-05,
+      "loss": 1.0021,
       "step": 7289
     },
     {
-      "epoch": 0.20021422097717723,
+      "epoch": 0.2068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.8518793625640382e-05,
-      "loss": 0.99,
+      "learning_rate": 1.8404032632437356e-05,
+      "loss": 0.9167,
       "step": 7290
     },
     {
-      "epoch": 0.20024168520502045,
+      "epoch": 0.2068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.851832771855145e-05,
-      "loss": 1.1155,
+      "learning_rate": 1.840353449194052e-05,
+      "loss": 1.0366,
       "step": 7291
     },
     {
-      "epoch": 0.2002691494328637,
+      "epoch": 0.20692395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 1.8517861744062212e-05,
-      "loss": 1.007,
+      "learning_rate": 1.840303628045866e-05,
+      "loss": 0.9151,
       "step": 7292
     },
     {
-      "epoch": 0.20029661366070692,
+      "epoch": 0.20695232690124857,
       "grad_norm": 0.0,
-      "learning_rate": 1.851739570217635e-05,
-      "loss": 1.057,
+      "learning_rate": 1.8402537997995963e-05,
+      "loss": 0.9579,
       "step": 7293
     },
     {
-      "epoch": 0.20032407788855017,
+      "epoch": 0.20698070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 1.8516929592897555e-05,
-      "loss": 1.0784,
+      "learning_rate": 1.8402039644556654e-05,
+      "loss": 0.9017,
       "step": 7294
     },
     {
-      "epoch": 0.2003515421163934,
+      "epoch": 0.20700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.8516463416229517e-05,
-      "loss": 0.9701,
+      "learning_rate": 1.840154122014494e-05,
+      "loss": 1.0564,
       "step": 7295
     },
     {
-      "epoch": 0.20037900634423664,
+      "epoch": 0.20703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.851599717217592e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.840104272476503e-05,
+      "loss": 0.9798,
       "step": 7296
     },
     {
-      "epoch": 0.20040647057207986,
+      "epoch": 0.20706583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 1.851553086074045e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.8400544158421133e-05,
+      "loss": 0.9855,
       "step": 7297
     },
     {
-      "epoch": 0.2004339347999231,
+      "epoch": 0.20709421112372303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8515064481926805e-05,
-      "loss": 1.1212,
+      "learning_rate": 1.840004552111746e-05,
+      "loss": 1.0185,
       "step": 7298
     },
     {
-      "epoch": 0.20046139902776633,
+      "epoch": 0.20712258796821795,
       "grad_norm": 0.0,
-      "learning_rate": 1.851459803573867e-05,
-      "loss": 1.0299,
+      "learning_rate": 1.8399546812858225e-05,
+      "loss": 1.0134,
       "step": 7299
     },
     {
-      "epoch": 0.20048886325560958,
+      "epoch": 0.20715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.851413152217974e-05,
-      "loss": 1.0488,
+      "learning_rate": 1.8399048033647644e-05,
+      "loss": 1.0356,
       "step": 7300
     },
     {
-      "epoch": 0.2005163274834528,
+      "epoch": 0.20717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.85136649412537e-05,
-      "loss": 0.969,
+      "learning_rate": 1.8398549183489926e-05,
+      "loss": 0.9844,
       "step": 7301
     },
     {
-      "epoch": 0.20054379171129605,
+      "epoch": 0.2072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.8513198292964246e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.8398050262389285e-05,
+      "loss": 1.0352,
       "step": 7302
     },
     {
-      "epoch": 0.20057125593913927,
+      "epoch": 0.2072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.851273157731507e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.8397551270349936e-05,
+      "loss": 0.9954,
       "step": 7303
     },
     {
-      "epoch": 0.2005987201669825,
+      "epoch": 0.20726447219069238,
       "grad_norm": 0.0,
-      "learning_rate": 1.8512264794309863e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.839705220737609e-05,
+      "loss": 0.9693,
       "step": 7304
     },
     {
-      "epoch": 0.20062618439482574,
+      "epoch": 0.2072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.851179794395232e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.839655307347197e-05,
+      "loss": 1.0083,
       "step": 7305
     },
     {
-      "epoch": 0.20065364862266896,
+      "epoch": 0.20732122587968219,
       "grad_norm": 0.0,
-      "learning_rate": 1.8511331026246137e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.839605386864179e-05,
+      "loss": 1.0342,
       "step": 7306
     },
     {
-      "epoch": 0.2006811128505122,
+      "epoch": 0.20734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.8510864041195004e-05,
-      "loss": 1.1043,
+      "learning_rate": 1.8395554592889766e-05,
+      "loss": 1.0948,
       "step": 7307
     },
     {
-      "epoch": 0.20070857707835543,
+      "epoch": 0.20737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.851039698880262e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.8395055246220114e-05,
+      "loss": 0.9178,
       "step": 7308
     },
     {
-      "epoch": 0.20073604130619868,
+      "epoch": 0.20740635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.8509929869072677e-05,
-      "loss": 1.123,
+      "learning_rate": 1.839455582863705e-05,
+      "loss": 0.9418,
       "step": 7309
     },
     {
-      "epoch": 0.2007635055340419,
+      "epoch": 0.20743473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.850946268200887e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.8394056340144796e-05,
+      "loss": 0.9626,
       "step": 7310
     },
     {
-      "epoch": 0.20079096976188515,
+      "epoch": 0.20746311010215665,
       "grad_norm": 0.0,
-      "learning_rate": 1.85089954276149e-05,
-      "loss": 1.1092,
+      "learning_rate": 1.8393556780747573e-05,
+      "loss": 1.0145,
       "step": 7311
     },
     {
-      "epoch": 0.20081843398972837,
+      "epoch": 0.20749148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 1.850852810589446e-05,
-      "loss": 1.1107,
+      "learning_rate": 1.8393057150449597e-05,
+      "loss": 0.9738,
       "step": 7312
     },
     {
-      "epoch": 0.20084589821757162,
+      "epoch": 0.20751986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.8508060716851255e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.8392557449255094e-05,
+      "loss": 0.8407,
       "step": 7313
     },
     {
-      "epoch": 0.20087336244541484,
+      "epoch": 0.2075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.8507593260488974e-05,
-      "loss": 1.0378,
+      "learning_rate": 1.8392057677168274e-05,
+      "loss": 0.8499,
       "step": 7314
     },
     {
-      "epoch": 0.2009008266732581,
+      "epoch": 0.2075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.850712573681132e-05,
-      "loss": 1.046,
+      "learning_rate": 1.839155783419337e-05,
+      "loss": 1.0743,
       "step": 7315
     },
     {
-      "epoch": 0.20092829090110131,
+      "epoch": 0.20760499432463111,
       "grad_norm": 0.0,
-      "learning_rate": 1.8506658145821992e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.8391057920334602e-05,
+      "loss": 1.0468,
       "step": 7316
     },
     {
-      "epoch": 0.20095575512894456,
+      "epoch": 0.207633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.850619048752469e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.8390557935596187e-05,
+      "loss": 0.9048,
       "step": 7317
     },
     {
-      "epoch": 0.20098321935678778,
+      "epoch": 0.2076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.8505722761923116e-05,
-      "loss": 1.0206,
+      "learning_rate": 1.839005787998235e-05,
+      "loss": 1.0207,
       "step": 7318
     },
     {
-      "epoch": 0.201010683584631,
+      "epoch": 0.20769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.8505254969020963e-05,
-      "loss": 1.011,
+      "learning_rate": 1.838955775349732e-05,
+      "loss": 1.0441,
       "step": 7319
     },
     {
-      "epoch": 0.20103814781247425,
+      "epoch": 0.20771850170261066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8504787108821942e-05,
-      "loss": 1.1085,
+      "learning_rate": 1.8389057556145318e-05,
+      "loss": 1.0535,
       "step": 7320
     },
     {
-      "epoch": 0.20106561204031748,
+      "epoch": 0.20774687854710555,
       "grad_norm": 0.0,
-      "learning_rate": 1.850431918132975e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.838855728793057e-05,
+      "loss": 0.9949,
       "step": 7321
     },
     {
-      "epoch": 0.20109307626816073,
+      "epoch": 0.20777525539160047,
       "grad_norm": 0.0,
-      "learning_rate": 1.850385118654809e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.8388056948857303e-05,
+      "loss": 0.9929,
       "step": 7322
     },
     {
-      "epoch": 0.20112054049600395,
+      "epoch": 0.20780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.8503383124480667e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.838755653892974e-05,
+      "loss": 1.0891,
       "step": 7323
     },
     {
-      "epoch": 0.2011480047238472,
+      "epoch": 0.20783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.8502914995131183e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.8387056058152107e-05,
+      "loss": 0.9918,
       "step": 7324
     },
     {
-      "epoch": 0.20117546895169042,
+      "epoch": 0.20786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.850244679850334e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.838655550652864e-05,
+      "loss": 1.0203,
       "step": 7325
     },
     {
-      "epoch": 0.20120293317953367,
+      "epoch": 0.20788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.8501978534600847e-05,
-      "loss": 1.033,
+      "learning_rate": 1.8386054884063558e-05,
+      "loss": 0.9066,
       "step": 7326
     },
     {
-      "epoch": 0.2012303974073769,
+      "epoch": 0.2079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.8501510203427403e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.8385554190761096e-05,
+      "loss": 1.0225,
       "step": 7327
     },
     {
-      "epoch": 0.20125786163522014,
+      "epoch": 0.20794551645856982,
       "grad_norm": 0.0,
-      "learning_rate": 1.850104180498672e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.838505342662548e-05,
+      "loss": 1.0604,
       "step": 7328
     },
     {
-      "epoch": 0.20128532586306336,
+      "epoch": 0.2079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8500573339282496e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.838455259166094e-05,
+      "loss": 0.9851,
       "step": 7329
     },
     {
-      "epoch": 0.2013127900909066,
+      "epoch": 0.2080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.850010480631845e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.8384051685871703e-05,
+      "loss": 1.0554,
       "step": 7330
     },
     {
-      "epoch": 0.20134025431874983,
+      "epoch": 0.20803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.8499636206098278e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.838355070926201e-05,
+      "loss": 1.059,
       "step": 7331
     },
     {
-      "epoch": 0.20136771854659305,
+      "epoch": 0.20805902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.8499167538625694e-05,
-      "loss": 1.0915,
+      "learning_rate": 1.838304966183609e-05,
+      "loss": 0.9639,
       "step": 7332
     },
     {
-      "epoch": 0.2013951827744363,
+      "epoch": 0.20808740068104425,
       "grad_norm": 0.0,
-      "learning_rate": 1.8498698803904402e-05,
-      "loss": 1.0748,
+      "learning_rate": 1.8382548543598166e-05,
+      "loss": 1.1278,
       "step": 7333
     },
     {
-      "epoch": 0.20142264700227952,
+      "epoch": 0.20811577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 1.8498230001938117e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.838204735455248e-05,
+      "loss": 0.9385,
       "step": 7334
     },
     {
-      "epoch": 0.20145011123012277,
+      "epoch": 0.20814415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.8497761132730538e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.8381546094703263e-05,
+      "loss": 0.934,
       "step": 7335
     },
     {
-      "epoch": 0.201477575457966,
+      "epoch": 0.20817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.8497292196285386e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.8381044764054753e-05,
+      "loss": 1.0237,
       "step": 7336
     },
     {
-      "epoch": 0.20150503968580924,
+      "epoch": 0.20820090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 1.8496823192606363e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.8380543362611175e-05,
+      "loss": 0.9232,
       "step": 7337
     },
     {
-      "epoch": 0.20153250391365246,
+      "epoch": 0.20822928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 1.8496354121697186e-05,
-      "loss": 1.065,
+      "learning_rate": 1.8380041890376778e-05,
+      "loss": 1.0328,
       "step": 7338
     },
     {
-      "epoch": 0.2015599681414957,
+      "epoch": 0.20825766174801363,
       "grad_norm": 0.0,
-      "learning_rate": 1.8495884983561566e-05,
-      "loss": 1.1416,
+      "learning_rate": 1.8379540347355786e-05,
+      "loss": 1.0173,
       "step": 7339
     },
     {
-      "epoch": 0.20158743236933893,
+      "epoch": 0.20828603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.849541577820321e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.8379038733552435e-05,
+      "loss": 0.9576,
       "step": 7340
     },
     {
-      "epoch": 0.20161489659718218,
+      "epoch": 0.2083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.8494946505625836e-05,
-      "loss": 1.0992,
+      "learning_rate": 1.8378537048970974e-05,
+      "loss": 0.9352,
       "step": 7341
     },
     {
-      "epoch": 0.2016423608250254,
+      "epoch": 0.2083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.849447716583315e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.837803529361563e-05,
+      "loss": 0.9167,
       "step": 7342
     },
     {
-      "epoch": 0.20166982505286865,
+      "epoch": 0.20837116912599318,
       "grad_norm": 0.0,
-      "learning_rate": 1.8494007758828874e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.8377533467490648e-05,
+      "loss": 0.9959,
       "step": 7343
     },
     {
-      "epoch": 0.20169728928071187,
+      "epoch": 0.20839954597048807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8493538284616717e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.8377031570600267e-05,
+      "loss": 0.9463,
       "step": 7344
     },
     {
-      "epoch": 0.2017247535085551,
+      "epoch": 0.20842792281498299,
       "grad_norm": 0.0,
-      "learning_rate": 1.8493068743200396e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.837652960294872e-05,
+      "loss": 1.0743,
       "step": 7345
     },
     {
-      "epoch": 0.20175221773639834,
+      "epoch": 0.20845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.8492599134583625e-05,
-      "loss": 0.996,
+      "learning_rate": 1.837602756454025e-05,
+      "loss": 0.9938,
       "step": 7346
     },
     {
-      "epoch": 0.20177968196424156,
+      "epoch": 0.20848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.8492129458770122e-05,
-      "loss": 1.0593,
+      "learning_rate": 1.83755254553791e-05,
+      "loss": 0.95,
       "step": 7347
     },
     {
-      "epoch": 0.2018071461920848,
+      "epoch": 0.20851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.84916597157636e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.837502327546951e-05,
+      "loss": 0.9612,
       "step": 7348
     },
     {
-      "epoch": 0.20183461041992803,
+      "epoch": 0.20854143019296253,
       "grad_norm": 0.0,
-      "learning_rate": 1.8491189905567772e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.837452102481572e-05,
+      "loss": 1.0922,
       "step": 7349
     },
     {
-      "epoch": 0.20186207464777128,
+      "epoch": 0.20856980703745742,
       "grad_norm": 0.0,
-      "learning_rate": 1.8490720028186367e-05,
-      "loss": 0.9364,
+      "learning_rate": 1.8374018703421977e-05,
+      "loss": 0.9883,
       "step": 7350
     },
     {
-      "epoch": 0.2018895388756145,
+      "epoch": 0.20859818388195234,
       "grad_norm": 0.0,
-      "learning_rate": 1.8490250083623092e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.837351631129252e-05,
+      "loss": 0.981,
       "step": 7351
     },
     {
-      "epoch": 0.20191700310345775,
+      "epoch": 0.20862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.8489780071881668e-05,
-      "loss": 1.0173,
+      "learning_rate": 1.83730138484316e-05,
+      "loss": 1.0873,
       "step": 7352
     },
     {
-      "epoch": 0.20194446733130098,
+      "epoch": 0.2086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.848930999296582e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.8372511314843445e-05,
+      "loss": 0.8064,
       "step": 7353
     },
     {
-      "epoch": 0.20197193155914422,
+      "epoch": 0.208683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.848883984687926e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.837200871053232e-05,
+      "loss": 0.9952,
       "step": 7354
     },
     {
-      "epoch": 0.20199939578698745,
+      "epoch": 0.2087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.8488369633625713e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.837150603550246e-05,
+      "loss": 0.8692,
       "step": 7355
     },
     {
-      "epoch": 0.2020268600148307,
+      "epoch": 0.2087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.8487899353208897e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.837100328975811e-05,
+      "loss": 0.9576,
       "step": 7356
     },
     {
-      "epoch": 0.20205432424267392,
+      "epoch": 0.2087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.848742900563253e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.8370500473303515e-05,
+      "loss": 0.9895,
       "step": 7357
     },
     {
-      "epoch": 0.20208178847051717,
+      "epoch": 0.20879682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 1.848695859090034e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.836999758614293e-05,
+      "loss": 0.9766,
       "step": 7358
     },
     {
-      "epoch": 0.2021092526983604,
+      "epoch": 0.20882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.848648810901605e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.8369494628280598e-05,
+      "loss": 1.0006,
       "step": 7359
     },
     {
-      "epoch": 0.2021367169262036,
+      "epoch": 0.20885357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.8486017559983373e-05,
-      "loss": 0.9683,
+      "learning_rate": 1.836899159972077e-05,
+      "loss": 1.0686,
       "step": 7360
     },
     {
-      "epoch": 0.20216418115404686,
+      "epoch": 0.20888195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.8485546943806038e-05,
-      "loss": 0.9864,
+      "learning_rate": 1.8368488500467695e-05,
+      "loss": 0.9744,
       "step": 7361
     },
     {
-      "epoch": 0.20219164538189008,
+      "epoch": 0.20891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.8485076260487773e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.836798533052562e-05,
+      "loss": 1.0612,
       "step": 7362
     },
     {
-      "epoch": 0.20221910960973333,
+      "epoch": 0.20893870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 1.8484605510032293e-05,
-      "loss": 0.9597,
+      "learning_rate": 1.8367482089898792e-05,
+      "loss": 1.1021,
       "step": 7363
     },
     {
-      "epoch": 0.20224657383757655,
+      "epoch": 0.20896708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.848413469244333e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.8366978778591474e-05,
+      "loss": 0.9948,
       "step": 7364
     },
     {
-      "epoch": 0.2022740380654198,
+      "epoch": 0.20899545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 1.8483663807724607e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.8366475396607907e-05,
+      "loss": 1.0377,
       "step": 7365
     },
     {
-      "epoch": 0.20230150229326302,
+      "epoch": 0.2090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.8483192855879848e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.8365971943952345e-05,
+      "loss": 1.0159,
       "step": 7366
     },
     {
-      "epoch": 0.20232896652110627,
+      "epoch": 0.2090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8482721836912783e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.836546842062904e-05,
+      "loss": 0.9618,
       "step": 7367
     },
     {
-      "epoch": 0.2023564307489495,
+      "epoch": 0.2090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.8482250750827136e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.8364964826642255e-05,
+      "loss": 1.0199,
       "step": 7368
     },
     {
-      "epoch": 0.20238389497679274,
+      "epoch": 0.2091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.8481779597626636e-05,
-      "loss": 1.1157,
+      "learning_rate": 1.836446116199623e-05,
+      "loss": 1.0186,
       "step": 7369
     },
     {
-      "epoch": 0.20241135920463596,
+      "epoch": 0.20913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.8481308377315008e-05,
-      "loss": 0.9391,
+      "learning_rate": 1.8363957426695227e-05,
+      "loss": 1.0656,
       "step": 7370
     },
     {
-      "epoch": 0.2024388234324792,
+      "epoch": 0.20916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.8480837089895985e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.83634536207435e-05,
+      "loss": 1.0148,
       "step": 7371
     },
     {
-      "epoch": 0.20246628766032243,
+      "epoch": 0.20919409761634505,
       "grad_norm": 0.0,
-      "learning_rate": 1.8480365735373294e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.8362949744145303e-05,
+      "loss": 0.9766,
       "step": 7372
     },
     {
-      "epoch": 0.20249375188816565,
+      "epoch": 0.20922247446083994,
       "grad_norm": 0.0,
-      "learning_rate": 1.8479894313750663e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.8362445796904896e-05,
+      "loss": 1.0193,
       "step": 7373
     },
     {
-      "epoch": 0.2025212161160089,
+      "epoch": 0.20925085130533486,
       "grad_norm": 0.0,
-      "learning_rate": 1.8479422825031824e-05,
-      "loss": 0.9607,
+      "learning_rate": 1.8361941779026532e-05,
+      "loss": 0.9228,
       "step": 7374
     },
     {
-      "epoch": 0.20254868034385212,
+      "epoch": 0.20927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.8478951269220508e-05,
-      "loss": 1.0502,
+      "learning_rate": 1.836143769051447e-05,
+      "loss": 1.0446,
       "step": 7375
     },
     {
-      "epoch": 0.20257614457169537,
+      "epoch": 0.20930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.8478479646320446e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.836093353137297e-05,
+      "loss": 0.9104,
       "step": 7376
     },
     {
-      "epoch": 0.2026036087995386,
+      "epoch": 0.20933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.8478007956335364e-05,
-      "loss": 1.0252,
+      "learning_rate": 1.8360429301606285e-05,
+      "loss": 1.044,
       "step": 7377
     },
     {
-      "epoch": 0.20263107302738184,
+      "epoch": 0.2093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.8477536199269003e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.8359925001218676e-05,
+      "loss": 1.0263,
       "step": 7378
     },
     {
-      "epoch": 0.20265853725522506,
+      "epoch": 0.20939273552780932,
       "grad_norm": 0.0,
-      "learning_rate": 1.847706437512509e-05,
-      "loss": 1.0687,
+      "learning_rate": 1.8359420630214407e-05,
+      "loss": 1.0719,
       "step": 7379
     },
     {
-      "epoch": 0.2026860014830683,
+      "epoch": 0.2094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.8476592483907358e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.835891618859774e-05,
+      "loss": 0.9792,
       "step": 7380
     },
     {
-      "epoch": 0.20271346571091153,
+      "epoch": 0.2094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.847612052561955e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.835841167637293e-05,
+      "loss": 0.9051,
       "step": 7381
     },
     {
-      "epoch": 0.20274092993875478,
+      "epoch": 0.20947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.8475648500265385e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.835790709354424e-05,
+      "loss": 1.0656,
       "step": 7382
     },
     {
-      "epoch": 0.202768394166598,
+      "epoch": 0.20950624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 1.847517640784861e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.835740244011593e-05,
+      "loss": 0.8987,
       "step": 7383
     },
     {
-      "epoch": 0.20279585839444125,
+      "epoch": 0.20953461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 1.8474704248372956e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.835689771609227e-05,
+      "loss": 1.1108,
       "step": 7384
     },
     {
-      "epoch": 0.20282332262228447,
+      "epoch": 0.20956299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.847423202184216e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.835639292147752e-05,
+      "loss": 0.884,
       "step": 7385
     },
     {
-      "epoch": 0.2028507868501277,
+      "epoch": 0.20959137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 1.8473759728259955e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.8355888056275943e-05,
+      "loss": 1.071,
       "step": 7386
     },
     {
-      "epoch": 0.20287825107797094,
+      "epoch": 0.20961975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.8473287367630083e-05,
-      "loss": 1.001,
+      "learning_rate": 1.8355383120491803e-05,
+      "loss": 0.9833,
       "step": 7387
     },
     {
-      "epoch": 0.20290571530581417,
+      "epoch": 0.20964812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.8472814939956275e-05,
-      "loss": 0.941,
+      "learning_rate": 1.8354878114129368e-05,
+      "loss": 0.9076,
       "step": 7388
     },
     {
-      "epoch": 0.20293317953365742,
+      "epoch": 0.20967650397275822,
       "grad_norm": 0.0,
-      "learning_rate": 1.847234244524228e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.8354373037192896e-05,
+      "loss": 0.9953,
       "step": 7389
     },
     {
-      "epoch": 0.20296064376150064,
+      "epoch": 0.2097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.8471869883491824e-05,
-      "loss": 0.9557,
+      "learning_rate": 1.8353867889686664e-05,
+      "loss": 1.0235,
       "step": 7390
     },
     {
-      "epoch": 0.20298810798934389,
+      "epoch": 0.20973325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 1.8471397254708654e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.8353362671614933e-05,
+      "loss": 1.0324,
       "step": 7391
     },
     {
-      "epoch": 0.2030155722171871,
+      "epoch": 0.2097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.8470924558896505e-05,
-      "loss": 1.066,
+      "learning_rate": 1.8352857382981977e-05,
+      "loss": 0.9965,
       "step": 7392
     },
     {
-      "epoch": 0.20304303644503036,
+      "epoch": 0.2097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.8470451796059126e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.8352352023792054e-05,
+      "loss": 1.0066,
       "step": 7393
     },
     {
-      "epoch": 0.20307050067287358,
+      "epoch": 0.20981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.8469978966200245e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.835184659404944e-05,
+      "loss": 0.9989,
       "step": 7394
     },
     {
-      "epoch": 0.20309796490071683,
+      "epoch": 0.20984676503972757,
       "grad_norm": 0.0,
-      "learning_rate": 1.8469506069323612e-05,
-      "loss": 1.0482,
+      "learning_rate": 1.83513410937584e-05,
+      "loss": 0.9445,
       "step": 7395
     },
     {
-      "epoch": 0.20312542912856005,
+      "epoch": 0.2098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.8469033105432965e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.8350835522923206e-05,
+      "loss": 1.0095,
       "step": 7396
     },
     {
-      "epoch": 0.2031528933564033,
+      "epoch": 0.20990351872871738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8468560074532047e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.8350329881548133e-05,
+      "loss": 1.1174,
       "step": 7397
     },
     {
-      "epoch": 0.20318035758424652,
+      "epoch": 0.20993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.8468086976624603e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.8349824169637444e-05,
+      "loss": 0.9447,
       "step": 7398
     },
     {
-      "epoch": 0.20320782181208974,
+      "epoch": 0.20996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.8467613811714375e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.8349318387195416e-05,
+      "loss": 0.8416,
       "step": 7399
     },
     {
-      "epoch": 0.203235286039933,
+      "epoch": 0.20998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.8467140579805105e-05,
-      "loss": 1.0705,
+      "learning_rate": 1.834881253422632e-05,
+      "loss": 0.9836,
       "step": 7400
     },
     {
-      "epoch": 0.2032627502677762,
+      "epoch": 0.21001702610669692,
       "grad_norm": 0.0,
-      "learning_rate": 1.846666728090054e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.8348306610734433e-05,
+      "loss": 1.0144,
       "step": 7401
     },
     {
-      "epoch": 0.20329021449561946,
+      "epoch": 0.21004540295119184,
       "grad_norm": 0.0,
-      "learning_rate": 1.846619391500442e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.8347800616724022e-05,
+      "loss": 0.9676,
       "step": 7402
     },
     {
-      "epoch": 0.20331767872346268,
+      "epoch": 0.21007377979568673,
       "grad_norm": 0.0,
-      "learning_rate": 1.84657204821205e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.8347294552199366e-05,
+      "loss": 1.1057,
       "step": 7403
     },
     {
-      "epoch": 0.20334514295130593,
+      "epoch": 0.21010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.8465246982252518e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.8346788417164735e-05,
+      "loss": 1.0139,
       "step": 7404
     },
     {
-      "epoch": 0.20337260717914915,
+      "epoch": 0.2101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.846477341540422e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.8346282211624404e-05,
+      "loss": 0.952,
       "step": 7405
     },
     {
-      "epoch": 0.2034000714069924,
+      "epoch": 0.2101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.846429978157936e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.8345775935582657e-05,
+      "loss": 0.9596,
       "step": 7406
     },
     {
-      "epoch": 0.20342753563483562,
+      "epoch": 0.21018728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.846382608078168e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.8345269589043765e-05,
+      "loss": 1.0257,
       "step": 7407
     },
     {
-      "epoch": 0.20345499986267887,
+      "epoch": 0.2102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.846335231301493e-05,
-      "loss": 1.0774,
+      "learning_rate": 1.8344763172012005e-05,
+      "loss": 0.9267,
       "step": 7408
     },
     {
-      "epoch": 0.2034824640905221,
+      "epoch": 0.21024404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 1.8462878478282858e-05,
-      "loss": 1.0924,
+      "learning_rate": 1.8344256684491654e-05,
+      "loss": 0.9589,
       "step": 7409
     },
     {
-      "epoch": 0.20350992831836534,
+      "epoch": 0.21027241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 1.846240457658921e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.834375012648699e-05,
+      "loss": 0.8203,
       "step": 7410
     },
     {
-      "epoch": 0.20353739254620856,
+      "epoch": 0.21030079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.846193060793774e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.83432434980023e-05,
+      "loss": 0.9687,
       "step": 7411
     },
     {
-      "epoch": 0.2035648567740518,
+      "epoch": 0.21032917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 1.8461456572332203e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.834273679904185e-05,
+      "loss": 0.8729,
       "step": 7412
     },
     {
-      "epoch": 0.20359232100189503,
+      "epoch": 0.21035754824063563,
       "grad_norm": 0.0,
-      "learning_rate": 1.8460982469776337e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.834223002960993e-05,
+      "loss": 1.0542,
       "step": 7413
     },
     {
-      "epoch": 0.20361978522973825,
+      "epoch": 0.21038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.8460508300273904e-05,
-      "loss": 1.0572,
+      "learning_rate": 1.8341723189710824e-05,
+      "loss": 0.9509,
       "step": 7414
     },
     {
-      "epoch": 0.2036472494575815,
+      "epoch": 0.21041430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.846003406382865e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.83412162793488e-05,
+      "loss": 0.9088,
       "step": 7415
     },
     {
-      "epoch": 0.20367471368542472,
+      "epoch": 0.21044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.845955976044433e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.834070929852815e-05,
+      "loss": 0.9227,
       "step": 7416
     },
     {
-      "epoch": 0.20370217791326797,
+      "epoch": 0.2104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.8459085390124697e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.8340202247253152e-05,
+      "loss": 0.9191,
       "step": 7417
     },
     {
-      "epoch": 0.2037296421411112,
+      "epoch": 0.2104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.8458610952873502e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.833969512552809e-05,
+      "loss": 1.0043,
       "step": 7418
     },
     {
-      "epoch": 0.20375710636895444,
+      "epoch": 0.210527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.84581364486945e-05,
-      "loss": 1.0722,
+      "learning_rate": 1.833918793335725e-05,
+      "loss": 1.0207,
       "step": 7419
     },
     {
-      "epoch": 0.20378457059679767,
+      "epoch": 0.2105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.845766187759145e-05,
-      "loss": 1.054,
+      "learning_rate": 1.8338680670744913e-05,
+      "loss": 1.0222,
       "step": 7420
     },
     {
-      "epoch": 0.20381203482464091,
+      "epoch": 0.21058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.84571872395681e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.833817333769537e-05,
+      "loss": 1.036,
       "step": 7421
     },
     {
-      "epoch": 0.20383949905248414,
+      "epoch": 0.21061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.845671253462821e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.83376659342129e-05,
+      "loss": 0.9336,
       "step": 7422
     },
     {
-      "epoch": 0.20386696328032738,
+      "epoch": 0.21064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.8456237762775537e-05,
-      "loss": 1.1077,
+      "learning_rate": 1.8337158460301786e-05,
+      "loss": 0.9284,
       "step": 7423
     },
     {
-      "epoch": 0.2038944275081706,
+      "epoch": 0.21066969353007944,
       "grad_norm": 0.0,
-      "learning_rate": 1.8455762924013832e-05,
-      "loss": 0.9089,
+      "learning_rate": 1.8336650915966325e-05,
+      "loss": 0.9552,
       "step": 7424
     },
     {
-      "epoch": 0.20392189173601385,
+      "epoch": 0.21069807037457436,
       "grad_norm": 0.0,
-      "learning_rate": 1.8455288018346852e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.8336143301210794e-05,
+      "loss": 1.0821,
       "step": 7425
     },
     {
-      "epoch": 0.20394935596385708,
+      "epoch": 0.21072644721906925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8454813045778364e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.833563561603949e-05,
+      "loss": 0.9196,
       "step": 7426
     },
     {
-      "epoch": 0.2039768201917003,
+      "epoch": 0.21075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.8454338006312114e-05,
-      "loss": 0.9236,
+      "learning_rate": 1.83351278604567e-05,
+      "loss": 0.9997,
       "step": 7427
     },
     {
-      "epoch": 0.20400428441954355,
+      "epoch": 0.21078320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.8453862899951872e-05,
-      "loss": 1.02,
+      "learning_rate": 1.8334620034466706e-05,
+      "loss": 0.9113,
       "step": 7428
     },
     {
-      "epoch": 0.20403174864738677,
+      "epoch": 0.2108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.845338772670139e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.8334112138073805e-05,
+      "loss": 0.9274,
       "step": 7429
     },
     {
-      "epoch": 0.20405921287523002,
+      "epoch": 0.2108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.8452912486564433e-05,
-      "loss": 1.032,
+      "learning_rate": 1.833360417128228e-05,
+      "loss": 1.0439,
       "step": 7430
     },
     {
-      "epoch": 0.20408667710307324,
+      "epoch": 0.2108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8452437179544755e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.8333096134096427e-05,
+      "loss": 1.1146,
       "step": 7431
     },
     {
-      "epoch": 0.2041141413309165,
+      "epoch": 0.2108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.845196180564612e-05,
-      "loss": 0.9549,
+      "learning_rate": 1.8332588026520536e-05,
+      "loss": 1.0612,
       "step": 7432
     },
     {
-      "epoch": 0.2041416055587597,
+      "epoch": 0.21092508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.845148636487229e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.83320798485589e-05,
+      "loss": 0.8653,
       "step": 7433
     },
     {
-      "epoch": 0.20416906978660296,
+      "epoch": 0.21095346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 1.8451010857227023e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.8331571600215813e-05,
+      "loss": 0.9632,
       "step": 7434
     },
     {
-      "epoch": 0.20419653401444618,
+      "epoch": 0.21098183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 1.8450535282714087e-05,
-      "loss": 1.0489,
+      "learning_rate": 1.8331063281495564e-05,
+      "loss": 1.008,
       "step": 7435
     },
     {
-      "epoch": 0.20422399824228943,
+      "epoch": 0.21101021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 1.845005964133724e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.833055489240245e-05,
+      "loss": 1.0521,
       "step": 7436
     },
     {
-      "epoch": 0.20425146247013265,
+      "epoch": 0.21103859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.8449583933100254e-05,
-      "loss": 1.1472,
+      "learning_rate": 1.8330046432940764e-05,
+      "loss": 1.0359,
       "step": 7437
     },
     {
-      "epoch": 0.2042789266979759,
+      "epoch": 0.21106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.8449108158006882e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.83295379031148e-05,
+      "loss": 1.0756,
       "step": 7438
     },
     {
-      "epoch": 0.20430639092581912,
+      "epoch": 0.21109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.8448632316060894e-05,
-      "loss": 1.077,
+      "learning_rate": 1.8329029302928858e-05,
+      "loss": 0.9068,
       "step": 7439
     },
     {
-      "epoch": 0.20433385515366234,
+      "epoch": 0.21112372304199772,
       "grad_norm": 0.0,
-      "learning_rate": 1.8448156407266052e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.8328520632387226e-05,
+      "loss": 1.1057,
       "step": 7440
     },
     {
-      "epoch": 0.2043613193815056,
+      "epoch": 0.2111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.8447680431626128e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.832801189149421e-05,
+      "loss": 1.0424,
       "step": 7441
     },
     {
-      "epoch": 0.2043887836093488,
+      "epoch": 0.21118047673098753,
       "grad_norm": 0.0,
-      "learning_rate": 1.8447204389144886e-05,
-      "loss": 1.088,
+      "learning_rate": 1.8327503080254107e-05,
+      "loss": 1.0059,
       "step": 7442
     },
     {
-      "epoch": 0.20441624783719206,
+      "epoch": 0.2112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.8446728279826086e-05,
-      "loss": 1.1049,
+      "learning_rate": 1.83269941986712e-05,
+      "loss": 1.0249,
       "step": 7443
     },
     {
-      "epoch": 0.20444371206503528,
+      "epoch": 0.2112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.84462521036735e-05,
-      "loss": 1.044,
+      "learning_rate": 1.832648524674981e-05,
+      "loss": 1.0664,
       "step": 7444
     },
     {
-      "epoch": 0.20447117629287853,
+      "epoch": 0.2112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.84457758606909e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.832597622449422e-05,
+      "loss": 1.1174,
       "step": 7445
     },
     {
-      "epoch": 0.20449864052072175,
+      "epoch": 0.21129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.8445299550882045e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.8325467131908735e-05,
+      "loss": 0.9395,
       "step": 7446
     },
     {
-      "epoch": 0.204526104748565,
+      "epoch": 0.21132236095346196,
       "grad_norm": 0.0,
-      "learning_rate": 1.844482317425071e-05,
-      "loss": 1.061,
+      "learning_rate": 1.8324957968997652e-05,
+      "loss": 1.0506,
       "step": 7447
     },
     {
-      "epoch": 0.20455356897640822,
+      "epoch": 0.21135073779795688,
       "grad_norm": 0.0,
-      "learning_rate": 1.8444346730800665e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.8324448735765278e-05,
+      "loss": 1.0117,
       "step": 7448
     },
     {
-      "epoch": 0.20458103320425147,
+      "epoch": 0.21137911464245177,
       "grad_norm": 0.0,
-      "learning_rate": 1.8443870220535676e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.8323939432215908e-05,
+      "loss": 0.9359,
       "step": 7449
     },
     {
-      "epoch": 0.2046084974320947,
+      "epoch": 0.21140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.8443393643459516e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.832343005835385e-05,
+      "loss": 1.0237,
       "step": 7450
     },
     {
-      "epoch": 0.20463596165993794,
+      "epoch": 0.21143586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.8442916999575954e-05,
-      "loss": 0.9912,
+      "learning_rate": 1.8322920614183405e-05,
+      "loss": 0.9547,
       "step": 7451
     },
     {
-      "epoch": 0.20466342588778116,
+      "epoch": 0.21146424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 1.844244028888876e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.8322411099708876e-05,
+      "loss": 1.0231,
       "step": 7452
     },
     {
-      "epoch": 0.20469089011562439,
+      "epoch": 0.21149262202043131,
       "grad_norm": 0.0,
-      "learning_rate": 1.844196351140171e-05,
-      "loss": 1.0857,
+      "learning_rate": 1.8321901514934562e-05,
+      "loss": 1.0715,
       "step": 7453
     },
     {
-      "epoch": 0.20471835434346763,
+      "epoch": 0.21152099886492623,
       "grad_norm": 0.0,
-      "learning_rate": 1.8441486667118577e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.8321391859864777e-05,
+      "loss": 1.0689,
       "step": 7454
     },
     {
-      "epoch": 0.20474581857131086,
+      "epoch": 0.21154937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 1.8441009756043126e-05,
-      "loss": 1.0905,
+      "learning_rate": 1.8320882134503818e-05,
+      "loss": 0.9749,
       "step": 7455
     },
     {
-      "epoch": 0.2047732827991541,
+      "epoch": 0.211577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.844053277817914e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.8320372338855993e-05,
+      "loss": 1.0152,
       "step": 7456
     },
     {
-      "epoch": 0.20480074702699733,
+      "epoch": 0.2116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.844005573353039e-05,
-      "loss": 1.0647,
+      "learning_rate": 1.831986247292561e-05,
+      "loss": 0.932,
       "step": 7457
     },
     {
-      "epoch": 0.20482821125484058,
+      "epoch": 0.21163450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 1.843957862210065e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.8319352536716974e-05,
+      "loss": 0.9852,
       "step": 7458
     },
     {
-      "epoch": 0.2048556754826838,
+      "epoch": 0.2116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.843910144389369e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.831884253023439e-05,
+      "loss": 1.0253,
       "step": 7459
     },
     {
-      "epoch": 0.20488313971052705,
+      "epoch": 0.21169125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 1.8438624198913296e-05,
-      "loss": 0.978,
+      "learning_rate": 1.8318332453482176e-05,
+      "loss": 1.0856,
       "step": 7460
     },
     {
-      "epoch": 0.20491060393837027,
+      "epoch": 0.21171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.8438146887163232e-05,
-      "loss": 0.934,
+      "learning_rate": 1.831782230646463e-05,
+      "loss": 1.0414,
       "step": 7461
     },
     {
-      "epoch": 0.20493806816621352,
+      "epoch": 0.21174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.8437669508647286e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.8317312089186065e-05,
+      "loss": 0.9156,
       "step": 7462
     },
     {
-      "epoch": 0.20496553239405674,
+      "epoch": 0.21177639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 1.843719206336923e-05,
-      "loss": 1.074,
+      "learning_rate": 1.8316801801650796e-05,
+      "loss": 1.0968,
       "step": 7463
     },
     {
-      "epoch": 0.2049929966219,
+      "epoch": 0.21180476730987513,
       "grad_norm": 0.0,
-      "learning_rate": 1.843671455133284e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.8316291443863125e-05,
+      "loss": 0.9616,
       "step": 7464
     },
     {
-      "epoch": 0.2050204608497432,
+      "epoch": 0.21183314415437005,
       "grad_norm": 0.0,
-      "learning_rate": 1.84362369725419e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.831578101582736e-05,
+      "loss": 0.9858,
       "step": 7465
     },
     {
-      "epoch": 0.20504792507758646,
+      "epoch": 0.21186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.843575932700018e-05,
-      "loss": 1.1298,
+      "learning_rate": 1.831527051754783e-05,
+      "loss": 0.9716,
       "step": 7466
     },
     {
-      "epoch": 0.20507538930542968,
+      "epoch": 0.21188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.8435281614711472e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.8314759949028827e-05,
+      "loss": 0.9835,
       "step": 7467
     },
     {
-      "epoch": 0.2051028535332729,
+      "epoch": 0.2119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8434803835679545e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.8314249310274676e-05,
+      "loss": 1.1211,
       "step": 7468
     },
     {
-      "epoch": 0.20513031776111615,
+      "epoch": 0.2119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.843432598990818e-05,
-      "loss": 0.8705,
+      "learning_rate": 1.8313738601289686e-05,
+      "loss": 1.0212,
       "step": 7469
     },
     {
-      "epoch": 0.20515778198895937,
+      "epoch": 0.21197502837684448,
       "grad_norm": 0.0,
-      "learning_rate": 1.8433848077401164e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.8313227822078174e-05,
+      "loss": 1.0113,
       "step": 7470
     },
     {
-      "epoch": 0.20518524621680262,
+      "epoch": 0.2120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.8433370098162272e-05,
-      "loss": 1.0917,
+      "learning_rate": 1.8312716972644456e-05,
+      "loss": 0.9471,
       "step": 7471
     },
     {
-      "epoch": 0.20521271044464584,
+      "epoch": 0.21203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.843289205219529e-05,
-      "loss": 1.0166,
+      "learning_rate": 1.8312206052992838e-05,
+      "loss": 1.0494,
       "step": 7472
     },
     {
-      "epoch": 0.2052401746724891,
+      "epoch": 0.21206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.8432413939504e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.8311695063127647e-05,
+      "loss": 0.9381,
       "step": 7473
     },
     {
-      "epoch": 0.2052676389003323,
+      "epoch": 0.21208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.8431935760092186e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.831118400305319e-05,
+      "loss": 0.9118,
       "step": 7474
     },
     {
-      "epoch": 0.20529510312817556,
+      "epoch": 0.21211691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.8431457513963628e-05,
-      "loss": 1.04,
+      "learning_rate": 1.8310672872773788e-05,
+      "loss": 1.0026,
       "step": 7475
     },
     {
-      "epoch": 0.20532256735601878,
+      "epoch": 0.21214528944381386,
       "grad_norm": 0.0,
-      "learning_rate": 1.843097920112211e-05,
-      "loss": 0.953,
+      "learning_rate": 1.8310161672293763e-05,
+      "loss": 1.0252,
       "step": 7476
     },
     {
-      "epoch": 0.20535003158386203,
+      "epoch": 0.21217366628830875,
       "grad_norm": 0.0,
-      "learning_rate": 1.8430500821571424e-05,
-      "loss": 1.0529,
+      "learning_rate": 1.8309650401617425e-05,
+      "loss": 0.8637,
       "step": 7477
     },
     {
-      "epoch": 0.20537749581170525,
+      "epoch": 0.21220204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.8430022375315347e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.83091390607491e-05,
+      "loss": 1.1166,
       "step": 7478
     },
     {
-      "epoch": 0.2054049600395485,
+      "epoch": 0.21223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.842954386235767e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.8308627649693102e-05,
+      "loss": 1.111,
       "step": 7479
     },
     {
-      "epoch": 0.20543242426739172,
+      "epoch": 0.2122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.8429065282702173e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.830811616845375e-05,
+      "loss": 1.0094,
       "step": 7480
     },
     {
-      "epoch": 0.20545988849523494,
+      "epoch": 0.2122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.8428586636352645e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.8307604617035372e-05,
+      "loss": 0.9517,
       "step": 7481
     },
     {
-      "epoch": 0.2054873527230782,
+      "epoch": 0.2123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.842810792331288e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.830709299544228e-05,
+      "loss": 0.9597,
       "step": 7482
     },
     {
-      "epoch": 0.20551481695092141,
+      "epoch": 0.2123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.8427629143586656e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.8306581303678804e-05,
+      "loss": 0.9599,
       "step": 7483
     },
     {
-      "epoch": 0.20554228117876466,
+      "epoch": 0.212372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.8427150297177768e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.830606954174926e-05,
+      "loss": 0.9596,
       "step": 7484
     },
     {
-      "epoch": 0.20556974540660788,
+      "epoch": 0.21240068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 1.842667138409e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.830555770965797e-05,
+      "loss": 1.185,
       "step": 7485
     },
     {
-      "epoch": 0.20559720963445113,
+      "epoch": 0.21242905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 1.8426192404327147e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.8305045807409266e-05,
+      "loss": 0.9902,
       "step": 7486
     },
     {
-      "epoch": 0.20562467386229435,
+      "epoch": 0.21245743473325765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8425713357892993e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.8304533835007466e-05,
+      "loss": 1.082,
       "step": 7487
     },
     {
-      "epoch": 0.2056521380901376,
+      "epoch": 0.21248581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.8425234244791333e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.8304021792456894e-05,
+      "loss": 1.1565,
       "step": 7488
     },
     {
-      "epoch": 0.20567960231798083,
+      "epoch": 0.21251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.8424755065025957e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.8303509679761877e-05,
+      "loss": 0.9453,
       "step": 7489
     },
     {
-      "epoch": 0.20570706654582407,
+      "epoch": 0.21254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.8424275818600654e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.830299749692674e-05,
+      "loss": 1.0139,
       "step": 7490
     },
     {
-      "epoch": 0.2057345307736673,
+      "epoch": 0.21257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8423796505519214e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.830248524395581e-05,
+      "loss": 0.9254,
       "step": 7491
     },
     {
-      "epoch": 0.20576199500151054,
+      "epoch": 0.21259931895573211,
       "grad_norm": 0.0,
-      "learning_rate": 1.842331712578544e-05,
-      "loss": 0.8852,
+      "learning_rate": 1.8301972920853416e-05,
+      "loss": 1.0298,
       "step": 7492
     },
     {
-      "epoch": 0.20578945922935377,
+      "epoch": 0.212627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.842283767940311e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.830146052762388e-05,
+      "loss": 0.9833,
       "step": 7493
     },
     {
-      "epoch": 0.205816923457197,
+      "epoch": 0.21265607264472192,
       "grad_norm": 0.0,
-      "learning_rate": 1.8422358166376034e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.8300948064271536e-05,
+      "loss": 1.093,
       "step": 7494
     },
     {
-      "epoch": 0.20584438768504024,
+      "epoch": 0.2126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.8421878586707993e-05,
-      "loss": 1.042,
+      "learning_rate": 1.8300435530800712e-05,
+      "loss": 0.925,
       "step": 7495
     },
     {
-      "epoch": 0.20587185191288346,
+      "epoch": 0.2127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.842139894040279e-05,
-      "loss": 1.009,
+      "learning_rate": 1.8299922927215733e-05,
+      "loss": 0.9834,
       "step": 7496
     },
     {
-      "epoch": 0.2058993161407267,
+      "epoch": 0.21274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.842091922746421e-05,
-      "loss": 1.0358,
+      "learning_rate": 1.8299410253520932e-05,
+      "loss": 1.0134,
       "step": 7497
     },
     {
-      "epoch": 0.20592678036856993,
+      "epoch": 0.21276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.8420439447896058e-05,
-      "loss": 0.9087,
+      "learning_rate": 1.829889750972064e-05,
+      "loss": 0.8639,
       "step": 7498
     },
     {
-      "epoch": 0.20595424459641318,
+      "epoch": 0.21279795686719638,
       "grad_norm": 0.0,
-      "learning_rate": 1.841995960170213e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.829838469581919e-05,
+      "loss": 0.9184,
       "step": 7499
     },
     {
-      "epoch": 0.2059817088242564,
+      "epoch": 0.21282633371169127,
       "grad_norm": 0.0,
-      "learning_rate": 1.8419479688886215e-05,
-      "loss": 1.1478,
+      "learning_rate": 1.8297871811820907e-05,
+      "loss": 0.9333,
       "step": 7500
     },
     {
-      "epoch": 0.20600917305209965,
+      "epoch": 0.21285471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.841899970945212e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.8297358857730134e-05,
+      "loss": 0.9309,
       "step": 7501
     },
     {
-      "epoch": 0.20603663727994287,
+      "epoch": 0.21288308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.8418519663403636e-05,
-      "loss": 0.8854,
+      "learning_rate": 1.8296845833551192e-05,
+      "loss": 1.1852,
       "step": 7502
     },
     {
-      "epoch": 0.20606410150778612,
+      "epoch": 0.21291146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 1.8418039550744565e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.8296332739288422e-05,
+      "loss": 1.0369,
       "step": 7503
     },
     {
-      "epoch": 0.20609156573562934,
+      "epoch": 0.21293984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 1.8417559371478702e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.8295819574946158e-05,
+      "loss": 1.0132,
       "step": 7504
     },
     {
-      "epoch": 0.2061190299634726,
+      "epoch": 0.21296821793416573,
       "grad_norm": 0.0,
-      "learning_rate": 1.8417079125609853e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.8295306340528733e-05,
+      "loss": 0.9572,
       "step": 7505
     },
     {
-      "epoch": 0.2061464941913158,
+      "epoch": 0.21299659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 1.8416598813141808e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.829479303604048e-05,
+      "loss": 1.0179,
       "step": 7506
     },
     {
-      "epoch": 0.20617395841915906,
+      "epoch": 0.2130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.8416118434078375e-05,
-      "loss": 0.9627,
+      "learning_rate": 1.829427966148574e-05,
+      "loss": 1.0238,
       "step": 7507
     },
     {
-      "epoch": 0.20620142264700228,
+      "epoch": 0.2130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.841563798842335e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.8293766216868842e-05,
+      "loss": 0.9968,
       "step": 7508
     },
     {
-      "epoch": 0.2062288868748455,
+      "epoch": 0.21308172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 1.8415157476180542e-05,
-      "loss": 1.0735,
+      "learning_rate": 1.8293252702194134e-05,
+      "loss": 0.8939,
       "step": 7509
     },
     {
-      "epoch": 0.20625635110268875,
+      "epoch": 0.21311010215664017,
       "grad_norm": 0.0,
-      "learning_rate": 1.8414676897353744e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.8292739117465945e-05,
+      "loss": 1.1105,
       "step": 7510
     },
     {
-      "epoch": 0.20628381533053197,
+      "epoch": 0.21313847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 1.841419625194677e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.8292225462688616e-05,
+      "loss": 1.0125,
       "step": 7511
     },
     {
-      "epoch": 0.20631127955837522,
+      "epoch": 0.21316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.841371553996341e-05,
-      "loss": 0.9395,
+      "learning_rate": 1.8291711737866484e-05,
+      "loss": 1.0736,
       "step": 7512
     },
     {
-      "epoch": 0.20633874378621844,
+      "epoch": 0.21319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.8413234761407476e-05,
-      "loss": 1.0,
+      "learning_rate": 1.8291197943003895e-05,
+      "loss": 1.0361,
       "step": 7513
     },
     {
-      "epoch": 0.2063662080140617,
+      "epoch": 0.21322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.8412753916282765e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.829068407810518e-05,
+      "loss": 0.9844,
       "step": 7514
     },
     {
-      "epoch": 0.2063936722419049,
+      "epoch": 0.21325198637911463,
       "grad_norm": 0.0,
-      "learning_rate": 1.841227300459309e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.8290170143174685e-05,
+      "loss": 0.9504,
       "step": 7515
     },
     {
-      "epoch": 0.20642113646974816,
+      "epoch": 0.21328036322360955,
       "grad_norm": 0.0,
-      "learning_rate": 1.841179202634225e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.8289656138216748e-05,
+      "loss": 1.0349,
       "step": 7516
     },
     {
-      "epoch": 0.20644860069759138,
+      "epoch": 0.21330874006810444,
       "grad_norm": 0.0,
-      "learning_rate": 1.841131098153406e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.8289142063235718e-05,
+      "loss": 0.9719,
       "step": 7517
     },
     {
-      "epoch": 0.20647606492543463,
+      "epoch": 0.21333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.8410829870172315e-05,
-      "loss": 0.953,
+      "learning_rate": 1.828862791823593e-05,
+      "loss": 0.9694,
       "step": 7518
     },
     {
-      "epoch": 0.20650352915327785,
+      "epoch": 0.2133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.8410348692260824e-05,
-      "loss": 0.9223,
+      "learning_rate": 1.8288113703221726e-05,
+      "loss": 1.0527,
       "step": 7519
     },
     {
-      "epoch": 0.2065309933811211,
+      "epoch": 0.2133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.8409867447803396e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.8287599418197457e-05,
+      "loss": 0.9153,
       "step": 7520
     },
     {
-      "epoch": 0.20655845760896432,
+      "epoch": 0.21342224744608398,
       "grad_norm": 0.0,
-      "learning_rate": 1.840938613680384e-05,
-      "loss": 0.9212,
+      "learning_rate": 1.8287085063167464e-05,
+      "loss": 1.0159,
       "step": 7521
     },
     {
-      "epoch": 0.20658592183680755,
+      "epoch": 0.2134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.8408904759265965e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.8286570638136086e-05,
+      "loss": 0.8586,
       "step": 7522
     },
     {
-      "epoch": 0.2066133860646508,
+      "epoch": 0.2134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8408423315193578e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.8286056143107677e-05,
+      "loss": 0.9221,
       "step": 7523
     },
     {
-      "epoch": 0.20664085029249402,
+      "epoch": 0.21350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.8407941804590487e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.8285541578086578e-05,
+      "loss": 0.8843,
       "step": 7524
     },
     {
-      "epoch": 0.20666831452033727,
+      "epoch": 0.21353575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.8407460227460507e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.828502694307714e-05,
+      "loss": 1.071,
       "step": 7525
     },
     {
-      "epoch": 0.2066957787481805,
+      "epoch": 0.21356413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.840697858380744e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.8284512238083706e-05,
+      "loss": 0.974,
       "step": 7526
     },
     {
-      "epoch": 0.20672324297602374,
+      "epoch": 0.21359250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.8406496873635104e-05,
-      "loss": 0.988,
+      "learning_rate": 1.828399746311062e-05,
+      "loss": 0.9772,
       "step": 7527
     },
     {
-      "epoch": 0.20675070720386696,
+      "epoch": 0.21362088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.840601509694731e-05,
-      "loss": 1.1283,
+      "learning_rate": 1.828348261816224e-05,
+      "loss": 0.9535,
       "step": 7528
     },
     {
-      "epoch": 0.2067781714317102,
+      "epoch": 0.21364926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 1.8405533253747868e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.8282967703242905e-05,
+      "loss": 0.8447,
       "step": 7529
     },
     {
-      "epoch": 0.20680563565955343,
+      "epoch": 0.21367763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.840505134404059e-05,
-      "loss": 1.1017,
+      "learning_rate": 1.8282452718356976e-05,
+      "loss": 1.0707,
       "step": 7530
     },
     {
-      "epoch": 0.20683309988739668,
+      "epoch": 0.2137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.8404569367829287e-05,
-      "loss": 1.0036,
+      "learning_rate": 1.8281937663508792e-05,
+      "loss": 1.0317,
       "step": 7531
     },
     {
-      "epoch": 0.2068605641152399,
+      "epoch": 0.2137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.840408732511778e-05,
-      "loss": 1.0569,
+      "learning_rate": 1.828142253870271e-05,
+      "loss": 1.076,
       "step": 7532
     },
     {
-      "epoch": 0.20688802834308315,
+      "epoch": 0.2137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8403605215909875e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.8280907343943077e-05,
+      "loss": 0.9007,
       "step": 7533
     },
     {
-      "epoch": 0.20691549257092637,
+      "epoch": 0.2137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.8403123040209396e-05,
-      "loss": 1.1053,
+      "learning_rate": 1.828039207923425e-05,
+      "loss": 0.9906,
       "step": 7534
     },
     {
-      "epoch": 0.2069429567987696,
+      "epoch": 0.2138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.840264079802015e-05,
-      "loss": 1.011,
+      "learning_rate": 1.8279876744580577e-05,
+      "loss": 1.0619,
       "step": 7535
     },
     {
-      "epoch": 0.20697042102661284,
+      "epoch": 0.21384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8402158489345953e-05,
-      "loss": 0.968,
+      "learning_rate": 1.8279361339986414e-05,
+      "loss": 1.1074,
       "step": 7536
     },
     {
-      "epoch": 0.20699788525445606,
+      "epoch": 0.21387627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.8401676114190624e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.8278845865456115e-05,
+      "loss": 0.9841,
       "step": 7537
     },
     {
-      "epoch": 0.2070253494822993,
+      "epoch": 0.21390465380249715,
       "grad_norm": 0.0,
-      "learning_rate": 1.8401193672557977e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.8278330320994035e-05,
+      "loss": 0.9232,
       "step": 7538
     },
     {
-      "epoch": 0.20705281371014253,
+      "epoch": 0.21393303064699207,
       "grad_norm": 0.0,
-      "learning_rate": 1.8400711164451835e-05,
-      "loss": 1.1074,
+      "learning_rate": 1.827781470660452e-05,
+      "loss": 1.0953,
       "step": 7539
     },
     {
-      "epoch": 0.20708027793798578,
+      "epoch": 0.21396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.840022858987601e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.8277299022291935e-05,
+      "loss": 1.0913,
       "step": 7540
     },
     {
-      "epoch": 0.207107742165829,
+      "epoch": 0.21398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.8399745948834323e-05,
-      "loss": 0.9798,
+      "learning_rate": 1.8276783268060633e-05,
+      "loss": 1.0456,
       "step": 7541
     },
     {
-      "epoch": 0.20713520639367225,
+      "epoch": 0.21401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.8399263241330594e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.827626744391497e-05,
+      "loss": 0.9163,
       "step": 7542
     },
     {
-      "epoch": 0.20716267062151547,
+      "epoch": 0.21404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.839878046736864e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.8275751549859306e-05,
+      "loss": 0.906,
       "step": 7543
     },
     {
-      "epoch": 0.20719013484935872,
+      "epoch": 0.2140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.8398297626952278e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.8275235585897995e-05,
+      "loss": 0.9531,
       "step": 7544
     },
     {
-      "epoch": 0.20721759907720194,
+      "epoch": 0.21410329171396142,
       "grad_norm": 0.0,
-      "learning_rate": 1.8397814720085335e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.82747195520354e-05,
+      "loss": 0.9955,
       "step": 7545
     },
     {
-      "epoch": 0.2072450633050452,
+      "epoch": 0.2141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.8397331746771628e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.827420344827587e-05,
+      "loss": 0.9641,
       "step": 7546
     },
     {
-      "epoch": 0.2072725275328884,
+      "epoch": 0.2141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.8396848707014984e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.827368727462378e-05,
+      "loss": 1.0737,
       "step": 7547
     },
     {
-      "epoch": 0.20729999176073163,
+      "epoch": 0.21418842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.8396365600819216e-05,
-      "loss": 0.988,
+      "learning_rate": 1.8273171031083477e-05,
+      "loss": 0.9352,
       "step": 7548
     },
     {
-      "epoch": 0.20732745598857488,
+      "epoch": 0.21421679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.839588242818815e-05,
-      "loss": 1.0669,
+      "learning_rate": 1.8272654717659327e-05,
+      "loss": 1.1259,
       "step": 7549
     },
     {
-      "epoch": 0.2073549202164181,
+      "epoch": 0.21424517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.8395399189125612e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.827213833435569e-05,
+      "loss": 0.9319,
       "step": 7550
     },
     {
-      "epoch": 0.20738238444426135,
+      "epoch": 0.21427355278093077,
       "grad_norm": 0.0,
-      "learning_rate": 1.8394915883635423e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.827162188117693e-05,
+      "loss": 0.9369,
       "step": 7551
     },
     {
-      "epoch": 0.20740984867210457,
+      "epoch": 0.21430192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.8394432511721408e-05,
-      "loss": 1.134,
+      "learning_rate": 1.827110535812741e-05,
+      "loss": 0.9954,
       "step": 7552
     },
     {
-      "epoch": 0.20743731289994782,
+      "epoch": 0.21433030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.8393949073387393e-05,
-      "loss": 1.1309,
+      "learning_rate": 1.8270588765211487e-05,
+      "loss": 1.037,
       "step": 7553
     },
     {
-      "epoch": 0.20746477712779104,
+      "epoch": 0.21435868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.8393465568637196e-05,
-      "loss": 1.113,
+      "learning_rate": 1.827007210243353e-05,
+      "loss": 0.8869,
       "step": 7554
     },
     {
-      "epoch": 0.2074922413556343,
+      "epoch": 0.21438706015891032,
       "grad_norm": 0.0,
-      "learning_rate": 1.839298199747465e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.82695553697979e-05,
+      "loss": 1.0845,
       "step": 7555
     },
     {
-      "epoch": 0.20751970558347752,
+      "epoch": 0.21441543700340523,
       "grad_norm": 0.0,
-      "learning_rate": 1.8392498359903586e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.8269038567308968e-05,
+      "loss": 1.011,
       "step": 7556
     },
     {
-      "epoch": 0.20754716981132076,
+      "epoch": 0.21444381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 1.8392014655927818e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.8268521694971096e-05,
+      "loss": 1.0132,
       "step": 7557
     },
     {
-      "epoch": 0.20757463403916399,
+      "epoch": 0.214472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.839153088555118e-05,
-      "loss": 1.0368,
+      "learning_rate": 1.8268004752788646e-05,
+      "loss": 1.052,
       "step": 7558
     },
     {
-      "epoch": 0.20760209826700723,
+      "epoch": 0.2145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.83910470487775e-05,
-      "loss": 1.073,
+      "learning_rate": 1.826748774076599e-05,
+      "loss": 0.9739,
       "step": 7559
     },
     {
-      "epoch": 0.20762956249485046,
+      "epoch": 0.21452894438138478,
       "grad_norm": 0.0,
-      "learning_rate": 1.8390563145610603e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.8266970658907493e-05,
+      "loss": 1.0864,
       "step": 7560
     },
     {
-      "epoch": 0.2076570267226937,
+      "epoch": 0.21455732122587967,
       "grad_norm": 0.0,
-      "learning_rate": 1.8390079176054322e-05,
-      "loss": 1.0248,
+      "learning_rate": 1.826645350721753e-05,
+      "loss": 1.0158,
       "step": 7561
     },
     {
-      "epoch": 0.20768449095053693,
+      "epoch": 0.2145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.8389595140112484e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.826593628570045e-05,
+      "loss": 1.0988,
       "step": 7562
     },
     {
-      "epoch": 0.20771195517838015,
+      "epoch": 0.21461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.838911103778892e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.8265418994360643e-05,
+      "loss": 1.1136,
       "step": 7563
     },
     {
-      "epoch": 0.2077394194062234,
+      "epoch": 0.21464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.838862686908746e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.826490163320247e-05,
+      "loss": 0.9208,
       "step": 7564
     },
     {
-      "epoch": 0.20776688363406662,
+      "epoch": 0.21467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8388142634011933e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.82643842022303e-05,
+      "loss": 1.033,
       "step": 7565
     },
     {
-      "epoch": 0.20779434786190987,
+      "epoch": 0.21469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.8387658332566177e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.8263866701448502e-05,
+      "loss": 0.9415,
       "step": 7566
     },
     {
-      "epoch": 0.2078218120897531,
+      "epoch": 0.21472758229284902,
       "grad_norm": 0.0,
-      "learning_rate": 1.8387173964754015e-05,
-      "loss": 0.949,
+      "learning_rate": 1.8263349130861453e-05,
+      "loss": 0.9735,
       "step": 7567
     },
     {
-      "epoch": 0.20784927631759634,
+      "epoch": 0.21475595913734394,
       "grad_norm": 0.0,
-      "learning_rate": 1.8386689530579285e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.8262831490473522e-05,
+      "loss": 0.9525,
       "step": 7568
     },
     {
-      "epoch": 0.20787674054543956,
+      "epoch": 0.21478433598183883,
       "grad_norm": 0.0,
-      "learning_rate": 1.8386205030045818e-05,
-      "loss": 0.986,
+      "learning_rate": 1.8262313780289084e-05,
+      "loss": 1.0366,
       "step": 7569
     },
     {
-      "epoch": 0.2079042047732828,
+      "epoch": 0.2148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.8385720463157447e-05,
-      "loss": 1.0797,
+      "learning_rate": 1.826179600031251e-05,
+      "loss": 0.982,
       "step": 7570
     },
     {
-      "epoch": 0.20793166900112603,
+      "epoch": 0.2148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.838523582991801e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.826127815054817e-05,
+      "loss": 0.9799,
       "step": 7571
     },
     {
-      "epoch": 0.20795913322896928,
+      "epoch": 0.2148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.8384751130331337e-05,
-      "loss": 0.9588,
+      "learning_rate": 1.8260760231000444e-05,
+      "loss": 1.0931,
       "step": 7572
     },
     {
-      "epoch": 0.2079865974568125,
+      "epoch": 0.21489784335981837,
       "grad_norm": 0.0,
-      "learning_rate": 1.8384266364401265e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.8260242241673706e-05,
+      "loss": 1.0443,
       "step": 7573
     },
     {
-      "epoch": 0.20801406168465575,
+      "epoch": 0.2149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.8383781532131634e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.825972418257233e-05,
+      "loss": 0.9308,
       "step": 7574
     },
     {
-      "epoch": 0.20804152591249897,
+      "epoch": 0.21495459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 1.838329663352627e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.8259206053700692e-05,
+      "loss": 1.1551,
       "step": 7575
     },
     {
-      "epoch": 0.2080689901403422,
+      "epoch": 0.21498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.8382811668589017e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.825868785506317e-05,
+      "loss": 0.9303,
       "step": 7576
     },
     {
-      "epoch": 0.20809645436818544,
+      "epoch": 0.21501135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.8382326637323717e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.825816958666414e-05,
+      "loss": 0.905,
       "step": 7577
     },
     {
-      "epoch": 0.20812391859602866,
+      "epoch": 0.21503972758229284,
       "grad_norm": 0.0,
-      "learning_rate": 1.8381841539734195e-05,
-      "loss": 1.0482,
+      "learning_rate": 1.825765124850798e-05,
+      "loss": 0.9711,
       "step": 7578
     },
     {
-      "epoch": 0.2081513828238719,
+      "epoch": 0.21506810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 1.8381356375824297e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.825713284059907e-05,
+      "loss": 0.9771,
       "step": 7579
     },
     {
-      "epoch": 0.20817884705171513,
+      "epoch": 0.21509648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.838087114559786e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.8256614362941785e-05,
+      "loss": 0.906,
       "step": 7580
     },
     {
-      "epoch": 0.20820631127955838,
+      "epoch": 0.21512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.8380385849058728e-05,
-      "loss": 1.0748,
+      "learning_rate": 1.8256095815540512e-05,
+      "loss": 1.0811,
       "step": 7581
     },
     {
-      "epoch": 0.2082337755074016,
+      "epoch": 0.21515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.8379900486210735e-05,
-      "loss": 1.1172,
+      "learning_rate": 1.8255577198399625e-05,
+      "loss": 1.038,
       "step": 7582
     },
     {
-      "epoch": 0.20826123973524485,
+      "epoch": 0.2151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.8379415057057725e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.82550585115235e-05,
+      "loss": 1.1177,
       "step": 7583
     },
     {
-      "epoch": 0.20828870396308807,
+      "epoch": 0.2152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.8378929561603532e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.825453975491653e-05,
+      "loss": 1.0051,
       "step": 7584
     },
     {
-      "epoch": 0.20831616819093132,
+      "epoch": 0.2152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.8378443999852007e-05,
-      "loss": 0.954,
+      "learning_rate": 1.825402092858309e-05,
+      "loss": 1.0513,
       "step": 7585
     },
     {
-      "epoch": 0.20834363241877454,
+      "epoch": 0.215266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8377958371806987e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.825350203252757e-05,
+      "loss": 1.0151,
       "step": 7586
     },
     {
-      "epoch": 0.2083710966466178,
+      "epoch": 0.21529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.837747267747231e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.825298306675434e-05,
+      "loss": 1.1118,
       "step": 7587
     },
     {
-      "epoch": 0.20839856087446101,
+      "epoch": 0.21532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.8376986916851832e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.8252464031267794e-05,
+      "loss": 1.1037,
       "step": 7588
     },
     {
-      "epoch": 0.20842602510230424,
+      "epoch": 0.21535187287173665,
       "grad_norm": 0.0,
-      "learning_rate": 1.8376501089949382e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.8251944926072313e-05,
+      "loss": 0.8811,
       "step": 7589
     },
     {
-      "epoch": 0.20845348933014748,
+      "epoch": 0.21538024971623154,
       "grad_norm": 0.0,
-      "learning_rate": 1.8376015196768816e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.825142575117228e-05,
+      "loss": 1.0029,
       "step": 7590
     },
     {
-      "epoch": 0.2084809535579907,
+      "epoch": 0.21540862656072646,
       "grad_norm": 0.0,
-      "learning_rate": 1.837552923731397e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.8250906506572087e-05,
+      "loss": 0.9814,
       "step": 7591
     },
     {
-      "epoch": 0.20850841778583395,
+      "epoch": 0.21543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.8375043211588695e-05,
-      "loss": 1.1184,
+      "learning_rate": 1.8250387192276115e-05,
+      "loss": 1.0792,
       "step": 7592
     },
     {
-      "epoch": 0.20853588201367718,
+      "epoch": 0.21546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.837455711959683e-05,
-      "loss": 0.8673,
+      "learning_rate": 1.824986780828875e-05,
+      "loss": 0.9471,
       "step": 7593
     },
     {
-      "epoch": 0.20856334624152043,
+      "epoch": 0.21549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.8374070961342228e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.824934835461438e-05,
+      "loss": 1.0402,
       "step": 7594
     },
     {
-      "epoch": 0.20859081046936365,
+      "epoch": 0.215522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8373584736828732e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.8248828831257396e-05,
+      "loss": 0.9873,
       "step": 7595
     },
     {
-      "epoch": 0.2086182746972069,
+      "epoch": 0.21555051078320092,
       "grad_norm": 0.0,
-      "learning_rate": 1.8373098446060188e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.8248309238222183e-05,
+      "loss": 0.9774,
       "step": 7596
     },
     {
-      "epoch": 0.20864573892505012,
+      "epoch": 0.2155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.8372612089040446e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.824778957551313e-05,
+      "loss": 1.0164,
       "step": 7597
     },
     {
-      "epoch": 0.20867320315289337,
+      "epoch": 0.2156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.837212566577336e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.824726984313463e-05,
+      "loss": 1.039,
       "step": 7598
     },
     {
-      "epoch": 0.2087006673807366,
+      "epoch": 0.21563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.8371639176262764e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.824675004109107e-05,
+      "loss": 1.064,
       "step": 7599
     },
     {
-      "epoch": 0.20872813160857984,
+      "epoch": 0.21566401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.8371152620512522e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.824623016938684e-05,
+      "loss": 0.9111,
       "step": 7600
     },
     {
-      "epoch": 0.20875559583642306,
+      "epoch": 0.21569239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.8370665998526475e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.8245710228026336e-05,
+      "loss": 0.9532,
       "step": 7601
     },
     {
-      "epoch": 0.20878306006426628,
+      "epoch": 0.21572077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.8370179310308475e-05,
-      "loss": 1.1234,
+      "learning_rate": 1.8245190217013943e-05,
+      "loss": 0.8456,
       "step": 7602
     },
     {
-      "epoch": 0.20881052429210953,
+      "epoch": 0.21574914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.836969255586238e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.8244670136354062e-05,
+      "loss": 0.9973,
       "step": 7603
     },
     {
-      "epoch": 0.20883798851995275,
+      "epoch": 0.21577752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.8369205735192032e-05,
-      "loss": 1.0708,
+      "learning_rate": 1.8244149986051075e-05,
+      "loss": 0.9921,
       "step": 7604
     },
     {
-      "epoch": 0.208865452747796,
+      "epoch": 0.21580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.8368718848301283e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.8243629766109393e-05,
+      "loss": 0.9565,
       "step": 7605
     },
     {
-      "epoch": 0.20889291697563922,
+      "epoch": 0.21583427922814982,
       "grad_norm": 0.0,
-      "learning_rate": 1.8368231895193993e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.824310947653339e-05,
+      "loss": 0.8923,
       "step": 7606
     },
     {
-      "epoch": 0.20892038120348247,
+      "epoch": 0.2158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.8367744875874006e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.8242589117327477e-05,
+      "loss": 1.1177,
       "step": 7607
     },
     {
-      "epoch": 0.2089478454313257,
+      "epoch": 0.21589103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 1.8367257790345184e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.824206868849604e-05,
+      "loss": 1.0024,
       "step": 7608
     },
     {
-      "epoch": 0.20897530965916894,
+      "epoch": 0.2159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.8366770638611376e-05,
-      "loss": 1.0632,
+      "learning_rate": 1.8241548190043476e-05,
+      "loss": 1.0139,
       "step": 7609
     },
     {
-      "epoch": 0.20900277388701216,
+      "epoch": 0.2159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.836628342067644e-05,
-      "loss": 1.0953,
+      "learning_rate": 1.824102762197419e-05,
+      "loss": 0.961,
       "step": 7610
     },
     {
-      "epoch": 0.2090302381148554,
+      "epoch": 0.2159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.8365796136544227e-05,
-      "loss": 0.8605,
+      "learning_rate": 1.824050698429257e-05,
+      "loss": 1.0372,
       "step": 7611
     },
     {
-      "epoch": 0.20905770234269863,
+      "epoch": 0.21600454029511917,
       "grad_norm": 0.0,
-      "learning_rate": 1.8365308786218594e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.8239986277003016e-05,
+      "loss": 0.9476,
       "step": 7612
     },
     {
-      "epoch": 0.20908516657054188,
+      "epoch": 0.21603291713961406,
       "grad_norm": 0.0,
-      "learning_rate": 1.8364821369703398e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.823946550010993e-05,
+      "loss": 1.0321,
       "step": 7613
     },
     {
-      "epoch": 0.2091126307983851,
+      "epoch": 0.21606129398410898,
       "grad_norm": 0.0,
-      "learning_rate": 1.8364333887002494e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.8238944653617706e-05,
+      "loss": 0.9907,
       "step": 7614
     },
     {
-      "epoch": 0.20914009502622835,
+      "epoch": 0.21608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.8363846338119745e-05,
-      "loss": 1.1322,
+      "learning_rate": 1.8238423737530748e-05,
+      "loss": 0.9511,
       "step": 7615
     },
     {
-      "epoch": 0.20916755925407157,
+      "epoch": 0.21611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.8363358723058996e-05,
-      "loss": 1.0714,
+      "learning_rate": 1.8237902751853453e-05,
+      "loss": 0.936,
       "step": 7616
     },
     {
-      "epoch": 0.2091950234819148,
+      "epoch": 0.21614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.8362871041824116e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.8237381696590227e-05,
+      "loss": 0.9702,
       "step": 7617
     },
     {
-      "epoch": 0.20922248770975804,
+      "epoch": 0.21617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.8362383294418965e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.8236860571745463e-05,
+      "loss": 1.0991,
       "step": 7618
     },
     {
-      "epoch": 0.20924995193760126,
+      "epoch": 0.21620317820658344,
       "grad_norm": 0.0,
-      "learning_rate": 1.8361895480847395e-05,
-      "loss": 1.0569,
+      "learning_rate": 1.823633937732357e-05,
+      "loss": 0.9526,
       "step": 7619
     },
     {
-      "epoch": 0.2092774161654445,
+      "epoch": 0.21623155505107833,
       "grad_norm": 0.0,
-      "learning_rate": 1.8361407601113268e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.8235818113328944e-05,
+      "loss": 1.0938,
       "step": 7620
     },
     {
-      "epoch": 0.20930488039328773,
+      "epoch": 0.21625993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.8360919655220447e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.8235296779765996e-05,
+      "loss": 0.9856,
       "step": 7621
     },
     {
-      "epoch": 0.20933234462113098,
+      "epoch": 0.2162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.836043164317279e-05,
-      "loss": 1.0886,
+      "learning_rate": 1.8234775376639128e-05,
+      "loss": 1.0559,
       "step": 7622
     },
     {
-      "epoch": 0.2093598088489742,
+      "epoch": 0.216316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.8359943564974162e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.8234253903952735e-05,
+      "loss": 0.9848,
       "step": 7623
     },
     {
-      "epoch": 0.20938727307681745,
+      "epoch": 0.21634506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 1.8359455420628416e-05,
-      "loss": 1.032,
+      "learning_rate": 1.8233732361711233e-05,
+      "loss": 0.9884,
       "step": 7624
     },
     {
-      "epoch": 0.20941473730466068,
+      "epoch": 0.2163734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.8358967210139426e-05,
-      "loss": 0.9969,
+      "learning_rate": 1.8233210749919025e-05,
+      "loss": 0.9264,
       "step": 7625
     },
     {
-      "epoch": 0.20944220153250392,
+      "epoch": 0.21640181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 1.8358478933511047e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.8232689068580516e-05,
+      "loss": 0.9565,
       "step": 7626
     },
     {
-      "epoch": 0.20946966576034715,
+      "epoch": 0.21643019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.835799059074715e-05,
-      "loss": 1.0816,
+      "learning_rate": 1.8232167317700108e-05,
+      "loss": 0.998,
       "step": 7627
     },
     {
-      "epoch": 0.2094971299881904,
+      "epoch": 0.21645856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.8357502181851584e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.8231645497282218e-05,
+      "loss": 1.0472,
       "step": 7628
     },
     {
-      "epoch": 0.20952459421603362,
+      "epoch": 0.21648694665153234,
       "grad_norm": 0.0,
-      "learning_rate": 1.835701370682823e-05,
-      "loss": 0.9699,
+      "learning_rate": 1.8231123607331245e-05,
+      "loss": 0.9695,
       "step": 7629
     },
     {
-      "epoch": 0.20955205844387684,
+      "epoch": 0.21651532349602723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8356525165680947e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.8230601647851602e-05,
+      "loss": 1.0414,
       "step": 7630
     },
     {
-      "epoch": 0.2095795226717201,
+      "epoch": 0.21654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.8356036558413596e-05,
-      "loss": 1.1015,
+      "learning_rate": 1.82300796188477e-05,
+      "loss": 1.0346,
       "step": 7631
     },
     {
-      "epoch": 0.2096069868995633,
+      "epoch": 0.21657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.8355547885030048e-05,
-      "loss": 0.957,
+      "learning_rate": 1.8229557520323942e-05,
+      "loss": 1.037,
       "step": 7632
     },
     {
-      "epoch": 0.20963445112740656,
+      "epoch": 0.21660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.835505914553417e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.822903535228474e-05,
+      "loss": 0.9832,
       "step": 7633
     },
     {
-      "epoch": 0.20966191535524978,
+      "epoch": 0.2166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.8354570339929826e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.8228513114734508e-05,
+      "loss": 1.0255,
       "step": 7634
     },
     {
-      "epoch": 0.20968937958309303,
+      "epoch": 0.2166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.835408146822089e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.8227990807677657e-05,
+      "loss": 1.0191,
       "step": 7635
     },
     {
-      "epoch": 0.20971684381093625,
+      "epoch": 0.2166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.835359253041122e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.8227468431118595e-05,
+      "loss": 0.9415,
       "step": 7636
     },
     {
-      "epoch": 0.2097443080387795,
+      "epoch": 0.2167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.8353103526504693e-05,
-      "loss": 0.9364,
+      "learning_rate": 1.8226945985061743e-05,
+      "loss": 1.0407,
       "step": 7637
     },
     {
-      "epoch": 0.20977177226662272,
+      "epoch": 0.21674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.8352614456505174e-05,
-      "loss": 0.8102,
+      "learning_rate": 1.8226423469511503e-05,
+      "loss": 1.028,
       "step": 7638
     },
     {
-      "epoch": 0.20979923649446597,
+      "epoch": 0.21677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.8352125320416533e-05,
-      "loss": 0.985,
+      "learning_rate": 1.8225900884472296e-05,
+      "loss": 1.0008,
       "step": 7639
     },
     {
-      "epoch": 0.2098267007223092,
+      "epoch": 0.21679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.8351636118242642e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.8225378229948533e-05,
+      "loss": 0.9998,
       "step": 7640
     },
     {
-      "epoch": 0.20985416495015244,
+      "epoch": 0.21682746878547104,
       "grad_norm": 0.0,
-      "learning_rate": 1.835114684998737e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.8224855505944634e-05,
+      "loss": 0.9921,
       "step": 7641
     },
     {
-      "epoch": 0.20988162917799566,
+      "epoch": 0.21685584562996596,
       "grad_norm": 0.0,
-      "learning_rate": 1.835065751565459e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.8224332712465008e-05,
+      "loss": 1.0163,
       "step": 7642
     },
     {
-      "epoch": 0.20990909340583888,
+      "epoch": 0.21688422247446085,
       "grad_norm": 0.0,
-      "learning_rate": 1.8350168115248175e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.8223809849514074e-05,
+      "loss": 1.0116,
       "step": 7643
     },
     {
-      "epoch": 0.20993655763368213,
+      "epoch": 0.21691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.8349678648771992e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.8223286917096247e-05,
+      "loss": 0.9701,
       "step": 7644
     },
     {
-      "epoch": 0.20996402186152535,
+      "epoch": 0.21694097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.834918911622992e-05,
-      "loss": 1.0713,
+      "learning_rate": 1.822276391521595e-05,
+      "loss": 1.0367,
       "step": 7645
     },
     {
-      "epoch": 0.2099914860893686,
+      "epoch": 0.2169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.834869951762583e-05,
-      "loss": 0.9678,
+      "learning_rate": 1.8222240843877593e-05,
+      "loss": 1.0648,
       "step": 7646
     },
     {
-      "epoch": 0.21001895031721182,
+      "epoch": 0.2169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.834820985296359e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.82217177030856e-05,
+      "loss": 0.9348,
       "step": 7647
     },
     {
-      "epoch": 0.21004641454505507,
+      "epoch": 0.2170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.8347720122247087e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.8221194492844383e-05,
+      "loss": 1.1225,
       "step": 7648
     },
     {
-      "epoch": 0.2100738787728983,
+      "epoch": 0.2170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.8347230325480184e-05,
-      "loss": 0.9831,
+      "learning_rate": 1.8220671213158373e-05,
+      "loss": 0.944,
       "step": 7649
     },
     {
-      "epoch": 0.21010134300074154,
+      "epoch": 0.21708286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 1.834674046266676e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.8220147864031984e-05,
+      "loss": 0.9353,
       "step": 7650
     },
     {
-      "epoch": 0.21012880722858476,
+      "epoch": 0.21711123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 1.8346250533810698e-05,
-      "loss": 0.9567,
+      "learning_rate": 1.8219624445469633e-05,
+      "loss": 1.0425,
       "step": 7651
     },
     {
-      "epoch": 0.210156271456428,
+      "epoch": 0.21713961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 1.8345760538915863e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.8219100957475747e-05,
+      "loss": 1.032,
       "step": 7652
     },
     {
-      "epoch": 0.21018373568427123,
+      "epoch": 0.21716799091940975,
       "grad_norm": 0.0,
-      "learning_rate": 1.834527047798614e-05,
-      "loss": 1.1455,
+      "learning_rate": 1.8218577400054744e-05,
+      "loss": 0.9197,
       "step": 7653
     },
     {
-      "epoch": 0.21021119991211448,
+      "epoch": 0.21719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.8344780351025405e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.8218053773211047e-05,
+      "loss": 0.8071,
       "step": 7654
     },
     {
-      "epoch": 0.2102386641399577,
+      "epoch": 0.21722474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.8344290158037536e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.8217530076949083e-05,
+      "loss": 0.9397,
       "step": 7655
     },
     {
-      "epoch": 0.21026612836780095,
+      "epoch": 0.21725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.834379989902641e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.821700631127327e-05,
+      "loss": 1.0876,
       "step": 7656
     },
     {
-      "epoch": 0.21029359259564417,
+      "epoch": 0.21728149829738932,
       "grad_norm": 0.0,
-      "learning_rate": 1.8343309573995908e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.821648247618804e-05,
+      "loss": 0.9105,
       "step": 7657
     },
     {
-      "epoch": 0.2103210568234874,
+      "epoch": 0.2173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.8342819182949906e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.8215958571697808e-05,
+      "loss": 0.9761,
       "step": 7658
     },
     {
-      "epoch": 0.21034852105133064,
+      "epoch": 0.21733825198637913,
       "grad_norm": 0.0,
-      "learning_rate": 1.834232872589229e-05,
-      "loss": 0.9758,
+      "learning_rate": 1.821543459780701e-05,
+      "loss": 1.0484,
       "step": 7659
     },
     {
-      "epoch": 0.21037598527917387,
+      "epoch": 0.21736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.8341838202826937e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.821491055452006e-05,
+      "loss": 0.9368,
       "step": 7660
     },
     {
-      "epoch": 0.21040344950701712,
+      "epoch": 0.2173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.834134761375773e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.8214386441841396e-05,
+      "loss": 0.9129,
       "step": 7661
     },
     {
-      "epoch": 0.21043091373486034,
+      "epoch": 0.2174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.8340856958688548e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.821386225977544e-05,
+      "loss": 0.9608,
       "step": 7662
     },
     {
-      "epoch": 0.21045837796270359,
+      "epoch": 0.21745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.8340366237623274e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.8213338008326623e-05,
+      "loss": 0.9273,
       "step": 7663
     },
     {
-      "epoch": 0.2104858421905468,
+      "epoch": 0.21748013620885356,
       "grad_norm": 0.0,
-      "learning_rate": 1.8339875450565792e-05,
-      "loss": 0.9087,
+      "learning_rate": 1.8212813687499365e-05,
+      "loss": 1.095,
       "step": 7664
     },
     {
-      "epoch": 0.21051330641839006,
+      "epoch": 0.21750851305334848,
       "grad_norm": 0.0,
-      "learning_rate": 1.833938459751998e-05,
-      "loss": 1.016,
+      "learning_rate": 1.8212289297298104e-05,
+      "loss": 1.06,
       "step": 7665
     },
     {
-      "epoch": 0.21054077064623328,
+      "epoch": 0.21753688989784337,
       "grad_norm": 0.0,
-      "learning_rate": 1.8338893678489738e-05,
-      "loss": 1.0682,
+      "learning_rate": 1.8211764837727263e-05,
+      "loss": 1.0283,
       "step": 7666
     },
     {
-      "epoch": 0.21056823487407653,
+      "epoch": 0.21756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.833840269347893e-05,
-      "loss": 0.9091,
+      "learning_rate": 1.821124030879128e-05,
+      "loss": 0.9201,
       "step": 7667
     },
     {
-      "epoch": 0.21059569910191975,
+      "epoch": 0.21759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.8337911642491454e-05,
-      "loss": 1.1298,
+      "learning_rate": 1.8210715710494576e-05,
+      "loss": 0.9227,
       "step": 7668
     },
     {
-      "epoch": 0.210623163329763,
+      "epoch": 0.21762202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.833742052553119e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.821019104284159e-05,
+      "loss": 1.0414,
       "step": 7669
     },
     {
-      "epoch": 0.21065062755760622,
+      "epoch": 0.21765039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 1.8336929342602025e-05,
-      "loss": 1.057,
+      "learning_rate": 1.820966630583675e-05,
+      "loss": 1.0477,
       "step": 7670
     },
     {
-      "epoch": 0.21067809178544944,
+      "epoch": 0.21767877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.8336438093707848e-05,
-      "loss": 1.0929,
+      "learning_rate": 1.820914149948449e-05,
+      "loss": 1.0003,
       "step": 7671
     },
     {
-      "epoch": 0.2107055560132927,
+      "epoch": 0.21770715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 1.833594677885254e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.8208616623789246e-05,
+      "loss": 1.0158,
       "step": 7672
     },
     {
-      "epoch": 0.2107330202411359,
+      "epoch": 0.2177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.8335455398039994e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.8208091678755444e-05,
+      "loss": 0.9887,
       "step": 7673
     },
     {
-      "epoch": 0.21076048446897916,
+      "epoch": 0.2177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.8334963951274096e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.8207566664387526e-05,
+      "loss": 0.9406,
       "step": 7674
     },
     {
-      "epoch": 0.21078794869682238,
+      "epoch": 0.21779228149829738,
       "grad_norm": 0.0,
-      "learning_rate": 1.833447243855873e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.820704158068992e-05,
+      "loss": 0.9858,
       "step": 7675
     },
     {
-      "epoch": 0.21081541292466563,
+      "epoch": 0.2178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.8333980859897795e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.820651642766707e-05,
+      "loss": 0.9922,
       "step": 7676
     },
     {
-      "epoch": 0.21084287715250885,
+      "epoch": 0.21784903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 1.8333489215295172e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.8205991205323402e-05,
+      "loss": 1.0831,
       "step": 7677
     },
     {
-      "epoch": 0.2108703413803521,
+      "epoch": 0.21787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.833299750475476e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.820546591366336e-05,
+      "loss": 0.8902,
       "step": 7678
     },
     {
-      "epoch": 0.21089780560819532,
+      "epoch": 0.21790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.8332505728280437e-05,
-      "loss": 0.9881,
+      "learning_rate": 1.8204940552691375e-05,
+      "loss": 1.0555,
       "step": 7679
     },
     {
-      "epoch": 0.21092526983603857,
+      "epoch": 0.21793416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.8332013885876105e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.820441512241189e-05,
+      "loss": 1.1048,
       "step": 7680
     },
     {
-      "epoch": 0.2109527340638818,
+      "epoch": 0.21796254256526673,
       "grad_norm": 0.0,
-      "learning_rate": 1.833152197754565e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.8203889622829344e-05,
+      "loss": 1.0366,
       "step": 7681
     },
     {
-      "epoch": 0.21098019829172504,
+      "epoch": 0.21799091940976165,
       "grad_norm": 0.0,
-      "learning_rate": 1.8331030003292966e-05,
-      "loss": 1.0344,
+      "learning_rate": 1.8203364053948173e-05,
+      "loss": 1.0515,
       "step": 7682
     },
     {
-      "epoch": 0.21100766251956826,
+      "epoch": 0.21801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.8330537963121945e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.8202838415772812e-05,
+      "loss": 0.8494,
       "step": 7683
     },
     {
-      "epoch": 0.21103512674741148,
+      "epoch": 0.21804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.8330045857036483e-05,
-      "loss": 1.0556,
+      "learning_rate": 1.820231270830771e-05,
+      "loss": 1.0419,
       "step": 7684
     },
     {
-      "epoch": 0.21106259097525473,
+      "epoch": 0.2180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.832955368504047e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.8201786931557305e-05,
+      "loss": 0.9569,
       "step": 7685
     },
     {
-      "epoch": 0.21109005520309795,
+      "epoch": 0.2181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.83290614471378e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.8201261085526036e-05,
+      "loss": 0.8791,
       "step": 7686
     },
     {
-      "epoch": 0.2111175194309412,
+      "epoch": 0.21813280363223608,
       "grad_norm": 0.0,
-      "learning_rate": 1.8328569143332372e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.8200735170218348e-05,
+      "loss": 0.9358,
       "step": 7687
     },
     {
-      "epoch": 0.21114498365878442,
+      "epoch": 0.218161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.8328076773628082e-05,
-      "loss": 1.1449,
+      "learning_rate": 1.820020918563868e-05,
+      "loss": 0.9111,
       "step": 7688
     },
     {
-      "epoch": 0.21117244788662767,
+      "epoch": 0.21818955732122589,
       "grad_norm": 0.0,
-      "learning_rate": 1.8327584338028813e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.8199683131791474e-05,
+      "loss": 0.9991,
       "step": 7689
     },
     {
-      "epoch": 0.2111999121144709,
+      "epoch": 0.21821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.8327091836538478e-05,
-      "loss": 1.1003,
+      "learning_rate": 1.819915700868118e-05,
+      "loss": 1.004,
       "step": 7690
     },
     {
-      "epoch": 0.21122737634231414,
+      "epoch": 0.21824631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.8326599269160964e-05,
-      "loss": 1.1422,
+      "learning_rate": 1.8198630816312236e-05,
+      "loss": 0.9184,
       "step": 7691
     },
     {
-      "epoch": 0.21125484057015737,
+      "epoch": 0.21827468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.8326106635900173e-05,
-      "loss": 1.1313,
+      "learning_rate": 1.8198104554689086e-05,
+      "loss": 0.9849,
       "step": 7692
     },
     {
-      "epoch": 0.21128230479800061,
+      "epoch": 0.21830306469920543,
       "grad_norm": 0.0,
-      "learning_rate": 1.832561393676e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.819757822381618e-05,
+      "loss": 1.0385,
       "step": 7693
     },
     {
-      "epoch": 0.21130976902584384,
+      "epoch": 0.21833144154370035,
       "grad_norm": 0.0,
-      "learning_rate": 1.8325121171744345e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.8197051823697964e-05,
+      "loss": 1.0491,
       "step": 7694
     },
     {
-      "epoch": 0.21133723325368708,
+      "epoch": 0.21835981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.832462834085711e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.8196525354338882e-05,
+      "loss": 1.0165,
       "step": 7695
     },
     {
-      "epoch": 0.2113646974815303,
+      "epoch": 0.21838819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.8324135444102185e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.8195998815743382e-05,
+      "loss": 0.9298,
       "step": 7696
     },
     {
-      "epoch": 0.21139216170937353,
+      "epoch": 0.218416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.832364248148348e-05,
-      "loss": 0.968,
+      "learning_rate": 1.819547220791591e-05,
+      "loss": 0.9825,
       "step": 7697
     },
     {
-      "epoch": 0.21141962593721678,
+      "epoch": 0.2184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.8323149453004887e-05,
-      "loss": 0.9118,
+      "learning_rate": 1.819494553086092e-05,
+      "loss": 1.0347,
       "step": 7698
     },
     {
-      "epoch": 0.21144709016506,
+      "epoch": 0.21847332576617481,
       "grad_norm": 0.0,
-      "learning_rate": 1.8322656358670314e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.8194418784582854e-05,
+      "loss": 0.9427,
       "step": 7699
     },
     {
-      "epoch": 0.21147455439290325,
+      "epoch": 0.2185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.832216319848366e-05,
-      "loss": 0.9859,
+      "learning_rate": 1.8193891969086164e-05,
+      "loss": 1.0542,
       "step": 7700
     },
     {
-      "epoch": 0.21150201862074647,
+      "epoch": 0.2185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.832166997244883e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.81933650843753e-05,
+      "loss": 1.0818,
       "step": 7701
     },
     {
-      "epoch": 0.21152948284858972,
+      "epoch": 0.21855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.832117668056972e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.8192838130454712e-05,
+      "loss": 1.0407,
       "step": 7702
     },
     {
-      "epoch": 0.21155694707643294,
+      "epoch": 0.21858683314415436,
       "grad_norm": 0.0,
-      "learning_rate": 1.8320683322850236e-05,
-      "loss": 1.0873,
+      "learning_rate": 1.8192311107328855e-05,
+      "loss": 0.9837,
       "step": 7703
     },
     {
-      "epoch": 0.2115844113042762,
+      "epoch": 0.21861520998864925,
       "grad_norm": 0.0,
-      "learning_rate": 1.832018989929428e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.8191784015002177e-05,
+      "loss": 0.9959,
       "step": 7704
     },
     {
-      "epoch": 0.2116118755321194,
+      "epoch": 0.21864358683314417,
       "grad_norm": 0.0,
-      "learning_rate": 1.8319696409905768e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.8191256853479128e-05,
+      "loss": 1.0904,
       "step": 7705
     },
     {
-      "epoch": 0.21163933975996266,
+      "epoch": 0.21867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.8319202854688588e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.8190729622764167e-05,
+      "loss": 0.9642,
       "step": 7706
     },
     {
-      "epoch": 0.21166680398780588,
+      "epoch": 0.21870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.8318709233646655e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.8190202322861742e-05,
+      "loss": 1.108,
       "step": 7707
     },
     {
-      "epoch": 0.21169426821564913,
+      "epoch": 0.21872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.8318215546783872e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.8189674953776312e-05,
+      "loss": 1.0019,
       "step": 7708
     },
     {
-      "epoch": 0.21172173244349235,
+      "epoch": 0.21875709421112371,
       "grad_norm": 0.0,
-      "learning_rate": 1.8317721794104142e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.818914751551233e-05,
+      "loss": 0.8325,
       "step": 7709
     },
     {
-      "epoch": 0.2117491966713356,
+      "epoch": 0.2187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.8317227975611378e-05,
-      "loss": 1.0663,
+      "learning_rate": 1.8188620008074252e-05,
+      "loss": 1.0645,
       "step": 7710
     },
     {
-      "epoch": 0.21177666089917882,
+      "epoch": 0.21881384790011352,
       "grad_norm": 0.0,
-      "learning_rate": 1.8316734091309484e-05,
-      "loss": 0.9472,
+      "learning_rate": 1.818809243146653e-05,
+      "loss": 0.9355,
       "step": 7711
     },
     {
-      "epoch": 0.21180412512702204,
+      "epoch": 0.2188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.831624014120237e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.8187564785693627e-05,
+      "loss": 0.9619,
       "step": 7712
     },
     {
-      "epoch": 0.2118315893548653,
+      "epoch": 0.2188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.8315746125293944e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.818703707075999e-05,
+      "loss": 1.0009,
       "step": 7713
     },
     {
-      "epoch": 0.2118590535827085,
+      "epoch": 0.21889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.831525204358811e-05,
-      "loss": 0.997,
+      "learning_rate": 1.8186509286670086e-05,
+      "loss": 0.9741,
       "step": 7714
     },
     {
-      "epoch": 0.21188651781055176,
+      "epoch": 0.21892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.831475789608878e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.818598143342837e-05,
+      "loss": 0.8382,
       "step": 7715
     },
     {
-      "epoch": 0.21191398203839498,
+      "epoch": 0.21895573212258798,
       "grad_norm": 0.0,
-      "learning_rate": 1.8314263682799866e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.8185453511039303e-05,
+      "loss": 0.896,
       "step": 7716
     },
     {
-      "epoch": 0.21194144626623823,
+      "epoch": 0.21898410896708287,
       "grad_norm": 0.0,
-      "learning_rate": 1.8313769403725277e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.8184925519507337e-05,
+      "loss": 0.9991,
       "step": 7717
     },
     {
-      "epoch": 0.21196891049408145,
+      "epoch": 0.21901248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.8313275058868923e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.818439745883694e-05,
+      "loss": 1.0156,
       "step": 7718
     },
     {
-      "epoch": 0.2119963747219247,
+      "epoch": 0.21904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.831278064823472e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.8183869329032573e-05,
+      "loss": 0.9918,
       "step": 7719
     },
     {
-      "epoch": 0.21202383894976792,
+      "epoch": 0.21906923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 1.8312286171826575e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.8183341130098692e-05,
+      "loss": 1.0723,
       "step": 7720
     },
     {
-      "epoch": 0.21205130317761117,
+      "epoch": 0.21909761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 1.83117916296484e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.818281286203976e-05,
+      "loss": 1.0405,
       "step": 7721
     },
     {
-      "epoch": 0.2120787674054544,
+      "epoch": 0.21912599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.8311297021704112e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.818228452486024e-05,
+      "loss": 0.9031,
       "step": 7722
     },
     {
-      "epoch": 0.21210623163329764,
+      "epoch": 0.21915437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 1.831080234799762e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.818175611856459e-05,
+      "loss": 0.981,
       "step": 7723
     },
     {
-      "epoch": 0.21213369586114086,
+      "epoch": 0.2191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.8310307608532846e-05,
-      "loss": 1.0231,
+      "learning_rate": 1.8181227643157286e-05,
+      "loss": 1.0148,
       "step": 7724
     },
     {
-      "epoch": 0.21216116008898409,
+      "epoch": 0.219211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.8309812803313692e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.8180699098642778e-05,
+      "loss": 1.078,
       "step": 7725
     },
     {
-      "epoch": 0.21218862431682733,
+      "epoch": 0.21923950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 1.8309317932344087e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.818017048502554e-05,
+      "loss": 0.9474,
       "step": 7726
     },
     {
-      "epoch": 0.21221608854467056,
+      "epoch": 0.21926787741203177,
       "grad_norm": 0.0,
-      "learning_rate": 1.8308822995627934e-05,
-      "loss": 1.006,
+      "learning_rate": 1.8179641802310035e-05,
+      "loss": 1.0276,
       "step": 7727
     },
     {
-      "epoch": 0.2122435527725138,
+      "epoch": 0.21929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 1.830832799316916e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.8179113050500727e-05,
+      "loss": 1.0282,
       "step": 7728
     },
     {
-      "epoch": 0.21227101700035703,
+      "epoch": 0.21932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.8307832924971673e-05,
-      "loss": 1.0675,
+      "learning_rate": 1.8178584229602082e-05,
+      "loss": 0.8849,
       "step": 7729
     },
     {
-      "epoch": 0.21229848122820028,
+      "epoch": 0.21935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.8307337791039392e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.817805533961857e-05,
+      "loss": 0.9011,
       "step": 7730
     },
     {
-      "epoch": 0.2123259454560435,
+      "epoch": 0.21938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.8306842591376243e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.817752638055466e-05,
+      "loss": 1.0949,
       "step": 7731
     },
     {
-      "epoch": 0.21235340968388675,
+      "epoch": 0.21940976163450623,
       "grad_norm": 0.0,
-      "learning_rate": 1.8306347325986132e-05,
-      "loss": 0.8902,
+      "learning_rate": 1.817699735241481e-05,
+      "loss": 0.9925,
       "step": 7732
     },
     {
-      "epoch": 0.21238087391172997,
+      "epoch": 0.21943813847900112,
       "grad_norm": 0.0,
-      "learning_rate": 1.8305851994872987e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.8176468255203502e-05,
+      "loss": 1.0344,
       "step": 7733
     },
     {
-      "epoch": 0.21240833813957322,
+      "epoch": 0.21946651532349604,
       "grad_norm": 0.0,
-      "learning_rate": 1.830535659804072e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.8175939088925195e-05,
+      "loss": 1.0693,
       "step": 7734
     },
     {
-      "epoch": 0.21243580236741644,
+      "epoch": 0.21949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.830486113549326e-05,
-      "loss": 1.0883,
+      "learning_rate": 1.8175409853584364e-05,
+      "loss": 0.9925,
       "step": 7735
     },
     {
-      "epoch": 0.2124632665952597,
+      "epoch": 0.2195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.8304365607234513e-05,
-      "loss": 1.0581,
+      "learning_rate": 1.817488054918548e-05,
+      "loss": 0.9563,
       "step": 7736
     },
     {
-      "epoch": 0.2124907308231029,
+      "epoch": 0.2195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.8303870013268416e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.817435117573301e-05,
+      "loss": 0.9352,
       "step": 7737
     },
     {
-      "epoch": 0.21251819505094613,
+      "epoch": 0.21958002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 1.8303374353598878e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.817382173323143e-05,
+      "loss": 1.0923,
       "step": 7738
     },
     {
-      "epoch": 0.21254565927878938,
+      "epoch": 0.2196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.8302878628229826e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.817329222168521e-05,
+      "loss": 1.0311,
       "step": 7739
     },
     {
-      "epoch": 0.2125731235066326,
+      "epoch": 0.2196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.8302382837165183e-05,
-      "loss": 1.1046,
+      "learning_rate": 1.8172762641098827e-05,
+      "loss": 1.0757,
       "step": 7740
     },
     {
-      "epoch": 0.21260058773447585,
+      "epoch": 0.21966515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 1.8301886980408876e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.8172232991476747e-05,
+      "loss": 1.1563,
       "step": 7741
     },
     {
-      "epoch": 0.21262805196231907,
+      "epoch": 0.21969353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.830139105796482e-05,
-      "loss": 1.1488,
+      "learning_rate": 1.8171703272823443e-05,
+      "loss": 0.8861,
       "step": 7742
     },
     {
-      "epoch": 0.21265551619016232,
+      "epoch": 0.21972190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.830089506983694e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.8171173485143402e-05,
+      "loss": 1.0359,
       "step": 7743
     },
     {
-      "epoch": 0.21268298041800554,
+      "epoch": 0.21975028376844494,
       "grad_norm": 0.0,
-      "learning_rate": 1.8300399016029166e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.817064362844109e-05,
+      "loss": 0.9296,
       "step": 7744
     },
     {
-      "epoch": 0.2127104446458488,
+      "epoch": 0.21977866061293985,
       "grad_norm": 0.0,
-      "learning_rate": 1.8299902896545414e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.817011370272098e-05,
+      "loss": 1.0758,
       "step": 7745
     },
     {
-      "epoch": 0.212737908873692,
+      "epoch": 0.21980703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 1.829940671138962e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.8169583707987552e-05,
+      "loss": 1.0537,
       "step": 7746
     },
     {
-      "epoch": 0.21276537310153526,
+      "epoch": 0.21983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.8298910460565708e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.8169053644245287e-05,
+      "loss": 0.8928,
       "step": 7747
     },
     {
-      "epoch": 0.21279283732937848,
+      "epoch": 0.21986379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 1.8298414144077593e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.8168523511498657e-05,
+      "loss": 0.9878,
       "step": 7748
     },
     {
-      "epoch": 0.21282030155722173,
+      "epoch": 0.2198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.829791776192922e-05,
-      "loss": 0.9165,
+      "learning_rate": 1.816799330975214e-05,
+      "loss": 1.0079,
       "step": 7749
     },
     {
-      "epoch": 0.21284776578506495,
+      "epoch": 0.2199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.8297421314124502e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.816746303901022e-05,
+      "loss": 1.0059,
       "step": 7750
     },
     {
-      "epoch": 0.2128752300129082,
+      "epoch": 0.2199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.8296924800667373e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.816693269927737e-05,
+      "loss": 0.9798,
       "step": 7751
     },
     {
-      "epoch": 0.21290269424075142,
+      "epoch": 0.2199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.8296428221561765e-05,
-      "loss": 1.105,
+      "learning_rate": 1.8166402290558073e-05,
+      "loss": 0.8838,
       "step": 7752
     },
     {
-      "epoch": 0.21293015846859464,
+      "epoch": 0.22000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.82959315768116e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.8165871812856807e-05,
+      "loss": 0.9402,
       "step": 7753
     },
     {
-      "epoch": 0.2129576226964379,
+      "epoch": 0.22003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.829543486642081e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.8165341266178056e-05,
+      "loss": 0.9815,
       "step": 7754
     },
     {
-      "epoch": 0.21298508692428111,
+      "epoch": 0.22006242905788875,
       "grad_norm": 0.0,
-      "learning_rate": 1.8294938090393327e-05,
-      "loss": 1.0912,
+      "learning_rate": 1.81648106505263e-05,
+      "loss": 1.0131,
       "step": 7755
     },
     {
-      "epoch": 0.21301255115212436,
+      "epoch": 0.22009080590238367,
       "grad_norm": 0.0,
-      "learning_rate": 1.829444124873308e-05,
-      "loss": 0.9068,
+      "learning_rate": 1.8164279965906023e-05,
+      "loss": 0.9814,
       "step": 7756
     },
     {
-      "epoch": 0.21304001537996758,
+      "epoch": 0.22011918274687856,
       "grad_norm": 0.0,
-      "learning_rate": 1.8293944341444e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.8163749212321704e-05,
+      "loss": 0.9104,
       "step": 7757
     },
     {
-      "epoch": 0.21306747960781083,
+      "epoch": 0.22014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.829344736853002e-05,
-      "loss": 1.0582,
+      "learning_rate": 1.8163218389777832e-05,
+      "loss": 1.0395,
       "step": 7758
     },
     {
-      "epoch": 0.21309494383565405,
+      "epoch": 0.22017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.8292950329995074e-05,
-      "loss": 1.1563,
+      "learning_rate": 1.8162687498278884e-05,
+      "loss": 0.8998,
       "step": 7759
     },
     {
-      "epoch": 0.2131224080634973,
+      "epoch": 0.22020431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.8292453225843093e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.8162156537829347e-05,
+      "loss": 1.082,
       "step": 7760
     },
     {
-      "epoch": 0.21314987229134053,
+      "epoch": 0.2202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.829195605607801e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.816162550843371e-05,
+      "loss": 0.9783,
       "step": 7761
     },
     {
-      "epoch": 0.21317733651918377,
+      "epoch": 0.22026106696935302,
       "grad_norm": 0.0,
-      "learning_rate": 1.8291458820703756e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.8161094410096455e-05,
+      "loss": 1.0436,
       "step": 7762
     },
     {
-      "epoch": 0.213204800747027,
+      "epoch": 0.2202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.8290961519724266e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.8160563242822068e-05,
+      "loss": 1.0891,
       "step": 7763
     },
     {
-      "epoch": 0.21323226497487024,
+      "epoch": 0.2203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.829046415314348e-05,
-      "loss": 1.0401,
+      "learning_rate": 1.8160032006615032e-05,
+      "loss": 0.9182,
       "step": 7764
     },
     {
-      "epoch": 0.21325972920271347,
+      "epoch": 0.22034619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.8289966720965334e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.8159500701479844e-05,
+      "loss": 1.0521,
       "step": 7765
     },
     {
-      "epoch": 0.2132871934305567,
+      "epoch": 0.22037457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 1.8289469223193754e-05,
-      "loss": 1.0722,
+      "learning_rate": 1.8158969327420984e-05,
+      "loss": 0.918,
       "step": 7766
     },
     {
-      "epoch": 0.21331465765839994,
+      "epoch": 0.22040295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.8288971659832684e-05,
-      "loss": 0.9318,
+      "learning_rate": 1.8158437884442947e-05,
+      "loss": 1.0988,
       "step": 7767
     },
     {
-      "epoch": 0.21334212188624316,
+      "epoch": 0.22043132803632237,
       "grad_norm": 0.0,
-      "learning_rate": 1.8288474030886062e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.8157906372550217e-05,
+      "loss": 0.9557,
       "step": 7768
     },
     {
-      "epoch": 0.2133695861140864,
+      "epoch": 0.22045970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.828797633635782e-05,
-      "loss": 0.9674,
+      "learning_rate": 1.8157374791747285e-05,
+      "loss": 0.9102,
       "step": 7769
     },
     {
-      "epoch": 0.21339705034192963,
+      "epoch": 0.22048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.82874785762519e-05,
-      "loss": 0.967,
+      "learning_rate": 1.815684314203864e-05,
+      "loss": 1.052,
       "step": 7770
     },
     {
-      "epoch": 0.21342451456977288,
+      "epoch": 0.22051645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.8286980750572237e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.8156311423428773e-05,
+      "loss": 0.9971,
       "step": 7771
     },
     {
-      "epoch": 0.2134519787976161,
+      "epoch": 0.22054483541430192,
       "grad_norm": 0.0,
-      "learning_rate": 1.8286482859322775e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.8155779635922178e-05,
+      "loss": 0.9808,
       "step": 7772
     },
     {
-      "epoch": 0.21347944302545935,
+      "epoch": 0.2205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.828598490250745e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.8155247779523348e-05,
+      "loss": 1.0685,
       "step": 7773
     },
     {
-      "epoch": 0.21350690725330257,
+      "epoch": 0.22060158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 1.8285486880130202e-05,
-      "loss": 1.002,
+      "learning_rate": 1.8154715854236775e-05,
+      "loss": 0.9676,
       "step": 7774
     },
     {
-      "epoch": 0.21353437148114582,
+      "epoch": 0.2206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.8284988792194976e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.8154183860066944e-05,
+      "loss": 1.015,
       "step": 7775
     },
     {
-      "epoch": 0.21356183570898904,
+      "epoch": 0.2206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.8284490638705702e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.815365179701836e-05,
+      "loss": 1.1113,
       "step": 7776
     },
     {
-      "epoch": 0.2135892999368323,
+      "epoch": 0.22068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.8283992419666337e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.8153119665095514e-05,
+      "loss": 0.9219,
       "step": 7777
     },
     {
-      "epoch": 0.2136167641646755,
+      "epoch": 0.22071509648127127,
       "grad_norm": 0.0,
-      "learning_rate": 1.8283494135080812e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.8152587464302896e-05,
+      "loss": 1.1039,
       "step": 7778
     },
     {
-      "epoch": 0.21364422839251873,
+      "epoch": 0.2207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.828299578495307e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.8152055194645006e-05,
+      "loss": 0.9821,
       "step": 7779
     },
     {
-      "epoch": 0.21367169262036198,
+      "epoch": 0.22077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.8282497369287058e-05,
-      "loss": 1.0265,
+      "learning_rate": 1.815152285612634e-05,
+      "loss": 0.9359,
       "step": 7780
     },
     {
-      "epoch": 0.2136991568482052,
+      "epoch": 0.22080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.828199888808672e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.8150990448751393e-05,
+      "loss": 1.1432,
       "step": 7781
     },
     {
-      "epoch": 0.21372662107604845,
+      "epoch": 0.22082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.8281500341355997e-05,
-      "loss": 0.949,
+      "learning_rate": 1.8150457972524667e-05,
+      "loss": 1.0428,
       "step": 7782
     },
     {
-      "epoch": 0.21375408530389167,
+      "epoch": 0.22085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.828100172909884e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.8149925427450653e-05,
+      "loss": 1.1014,
       "step": 7783
     },
     {
-      "epoch": 0.21378154953173492,
+      "epoch": 0.22088535754824062,
       "grad_norm": 0.0,
-      "learning_rate": 1.8280503051319182e-05,
-      "loss": 1.0837,
+      "learning_rate": 1.8149392813533852e-05,
+      "loss": 1.0145,
       "step": 7784
     },
     {
-      "epoch": 0.21380901375957814,
+      "epoch": 0.22091373439273554,
       "grad_norm": 0.0,
-      "learning_rate": 1.828000430802098e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.8148860130778765e-05,
+      "loss": 0.9385,
       "step": 7785
     },
     {
-      "epoch": 0.2138364779874214,
+      "epoch": 0.22094211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.827950549920818e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.814832737918989e-05,
+      "loss": 0.9982,
       "step": 7786
     },
     {
-      "epoch": 0.2138639422152646,
+      "epoch": 0.2209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.827900662488472e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.8147794558771728e-05,
+      "loss": 1.0975,
       "step": 7787
     },
     {
-      "epoch": 0.21389140644310786,
+      "epoch": 0.2209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.8278507685054556e-05,
-      "loss": 0.9689,
+      "learning_rate": 1.814726166952878e-05,
+      "loss": 0.9729,
       "step": 7788
     },
     {
-      "epoch": 0.21391887067095108,
+      "epoch": 0.2210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.827800867972163e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.814672871146554e-05,
+      "loss": 1.0341,
       "step": 7789
     },
     {
-      "epoch": 0.21394633489879433,
+      "epoch": 0.22105561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 1.8277509608889895e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.814619568458652e-05,
+      "loss": 1.037,
       "step": 7790
     },
     {
-      "epoch": 0.21397379912663755,
+      "epoch": 0.2210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.8277010472563295e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.814566258889622e-05,
+      "loss": 1.01,
       "step": 7791
     },
     {
-      "epoch": 0.21400126335448078,
+      "epoch": 0.22111237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 1.8276511270745784e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.8145129424399142e-05,
+      "loss": 0.9422,
       "step": 7792
     },
     {
-      "epoch": 0.21402872758232402,
+      "epoch": 0.22114074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.8276012003441307e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.814459619109979e-05,
+      "loss": 1.0901,
       "step": 7793
     },
     {
-      "epoch": 0.21405619181016725,
+      "epoch": 0.22116912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.827551267065382e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.8144062889002668e-05,
+      "loss": 0.999,
       "step": 7794
     },
     {
-      "epoch": 0.2140836560380105,
+      "epoch": 0.22119750283768444,
       "grad_norm": 0.0,
-      "learning_rate": 1.8275013272387272e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.814352951811228e-05,
+      "loss": 1.0437,
       "step": 7795
     },
     {
-      "epoch": 0.21411112026585372,
+      "epoch": 0.22122587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 1.8274513808645612e-05,
-      "loss": 0.8558,
+      "learning_rate": 1.814299607843313e-05,
+      "loss": 1.0963,
       "step": 7796
     },
     {
-      "epoch": 0.21413858449369697,
+      "epoch": 0.22125425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.827401427943279e-05,
-      "loss": 1.157,
+      "learning_rate": 1.814246256996973e-05,
+      "loss": 1.0034,
       "step": 7797
     },
     {
-      "epoch": 0.2141660487215402,
+      "epoch": 0.22128263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.827351468475277e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.814192899272658e-05,
+      "loss": 0.9559,
       "step": 7798
     },
     {
-      "epoch": 0.21419351294938344,
+      "epoch": 0.22131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.827301502460949e-05,
-      "loss": 1.0951,
+      "learning_rate": 1.8141395346708192e-05,
+      "loss": 0.9955,
       "step": 7799
     },
     {
-      "epoch": 0.21422097717722666,
+      "epoch": 0.2213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.8272515299006916e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.814086163191907e-05,
+      "loss": 1.0579,
       "step": 7800
     },
     {
-      "epoch": 0.2142484414050699,
+      "epoch": 0.2213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.8272015507948994e-05,
-      "loss": 1.1224,
+      "learning_rate": 1.814032784836372e-05,
+      "loss": 0.9186,
       "step": 7801
     },
     {
-      "epoch": 0.21427590563291313,
+      "epoch": 0.2213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.827151565143968e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.8139793996046662e-05,
+      "loss": 0.9715,
       "step": 7802
     },
     {
-      "epoch": 0.21430336986075638,
+      "epoch": 0.2214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.8271015729482927e-05,
-      "loss": 1.0316,
+      "learning_rate": 1.8139260074972392e-05,
+      "loss": 0.9743,
       "step": 7803
     },
     {
-      "epoch": 0.2143308340885996,
+      "epoch": 0.22145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.82705157420827e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.813872608514543e-05,
+      "loss": 0.9395,
       "step": 7804
     },
     {
-      "epoch": 0.21435829831644285,
+      "epoch": 0.22148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.8270015689242944e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.8138192026570282e-05,
+      "loss": 0.9482,
       "step": 7805
     },
     {
-      "epoch": 0.21438576254428607,
+      "epoch": 0.22150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.8269515570967623e-05,
-      "loss": 1.1041,
+      "learning_rate": 1.813765789925146e-05,
+      "loss": 0.8566,
       "step": 7806
     },
     {
-      "epoch": 0.2144132267721293,
+      "epoch": 0.22153802497162314,
       "grad_norm": 0.0,
-      "learning_rate": 1.826901538726069e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.8137123703193474e-05,
+      "loss": 0.9187,
       "step": 7807
     },
     {
-      "epoch": 0.21444069099997254,
+      "epoch": 0.22156640181611806,
       "grad_norm": 0.0,
-      "learning_rate": 1.8268515138126107e-05,
-      "loss": 0.8588,
+      "learning_rate": 1.8136589438400843e-05,
+      "loss": 1.1276,
       "step": 7808
     },
     {
-      "epoch": 0.21446815522781576,
+      "epoch": 0.22159477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.8268014823567824e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.8136055104878072e-05,
+      "loss": 0.9683,
       "step": 7809
     },
     {
-      "epoch": 0.214495619455659,
+      "epoch": 0.22162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.8267514443589812e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.8135520702629677e-05,
+      "loss": 1.0086,
       "step": 7810
     },
     {
-      "epoch": 0.21452308368350223,
+      "epoch": 0.22165153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.8267013998196018e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.8134986231660174e-05,
+      "loss": 1.0132,
       "step": 7811
     },
     {
-      "epoch": 0.21455054791134548,
+      "epoch": 0.2216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.826651348739041e-05,
-      "loss": 1.0409,
+      "learning_rate": 1.8134451691974076e-05,
+      "loss": 0.8855,
       "step": 7812
     },
     {
-      "epoch": 0.2145780121391887,
+      "epoch": 0.2217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8266012911176944e-05,
-      "loss": 1.1047,
+      "learning_rate": 1.81339170835759e-05,
+      "loss": 1.0689,
       "step": 7813
     },
     {
-      "epoch": 0.21460547636703195,
+      "epoch": 0.2217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.826551226955958e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.813338240647016e-05,
+      "loss": 1.0682,
       "step": 7814
     },
     {
-      "epoch": 0.21463294059487517,
+      "epoch": 0.2217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.8265011562542284e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.8132847660661374e-05,
+      "loss": 0.999,
       "step": 7815
     },
     {
-      "epoch": 0.21466040482271842,
+      "epoch": 0.22179341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.826451079012901e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.8132312846154058e-05,
+      "loss": 1.065,
       "step": 7816
     },
     {
-      "epoch": 0.21468786905056164,
+      "epoch": 0.22182179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 1.826400995232373e-05,
-      "loss": 1.0127,
+      "learning_rate": 1.8131777962952732e-05,
+      "loss": 1.0956,
       "step": 7817
     },
     {
-      "epoch": 0.2147153332784049,
+      "epoch": 0.22185017026106696,
       "grad_norm": 0.0,
-      "learning_rate": 1.82635090491304e-05,
-      "loss": 1.0332,
+      "learning_rate": 1.813124301106191e-05,
+      "loss": 0.993,
       "step": 7818
     },
     {
-      "epoch": 0.2147427975062481,
+      "epoch": 0.22187854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 1.8263008080552985e-05,
-      "loss": 1.0036,
+      "learning_rate": 1.813070799048611e-05,
+      "loss": 0.9368,
       "step": 7819
     },
     {
-      "epoch": 0.21477026173409133,
+      "epoch": 0.22190692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.8262507046595454e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.8130172901229857e-05,
+      "loss": 1.0063,
       "step": 7820
     },
     {
-      "epoch": 0.21479772596193458,
+      "epoch": 0.22193530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.8262005947261762e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.8129637743297667e-05,
+      "loss": 0.983,
       "step": 7821
     },
     {
-      "epoch": 0.2148251901897778,
+      "epoch": 0.22196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.826150478255588e-05,
-      "loss": 0.9968,
+      "learning_rate": 1.8129102516694063e-05,
+      "loss": 1.114,
       "step": 7822
     },
     {
-      "epoch": 0.21485265441762105,
+      "epoch": 0.22199205448354142,
       "grad_norm": 0.0,
-      "learning_rate": 1.826100355248177e-05,
-      "loss": 1.052,
+      "learning_rate": 1.8128567221423565e-05,
+      "loss": 1.0292,
       "step": 7823
     },
     {
-      "epoch": 0.21488011864546427,
+      "epoch": 0.2220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.8260502257043404e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.812803185749069e-05,
+      "loss": 1.02,
       "step": 7824
     },
     {
-      "epoch": 0.21490758287330752,
+      "epoch": 0.22204880817253123,
       "grad_norm": 0.0,
-      "learning_rate": 1.8260000896244743e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.8127496424899968e-05,
+      "loss": 0.9906,
       "step": 7825
     },
     {
-      "epoch": 0.21493504710115074,
+      "epoch": 0.2220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.8259499470089754e-05,
-      "loss": 1.0579,
+      "learning_rate": 1.8126960923655915e-05,
+      "loss": 0.9189,
       "step": 7826
     },
     {
-      "epoch": 0.214962511328994,
+      "epoch": 0.222105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.8258997978582403e-05,
-      "loss": 1.0336,
+      "learning_rate": 1.812642535376306e-05,
+      "loss": 1.0056,
       "step": 7827
     },
     {
-      "epoch": 0.21498997555683722,
+      "epoch": 0.2221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.8258496421726666e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.812588971522593e-05,
+      "loss": 0.9342,
       "step": 7828
     },
     {
-      "epoch": 0.21501743978468046,
+      "epoch": 0.22216231555051077,
       "grad_norm": 0.0,
-      "learning_rate": 1.8257994799526505e-05,
-      "loss": 1.1213,
+      "learning_rate": 1.8125354008049037e-05,
+      "loss": 0.9833,
       "step": 7829
     },
     {
-      "epoch": 0.21504490401252369,
+      "epoch": 0.22219069239500566,
       "grad_norm": 0.0,
-      "learning_rate": 1.8257493111985887e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.8124818232236917e-05,
+      "loss": 0.9535,
       "step": 7830
     },
     {
-      "epoch": 0.21507236824036693,
+      "epoch": 0.22221906923950058,
       "grad_norm": 0.0,
-      "learning_rate": 1.825699135910879e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.812428238779409e-05,
+      "loss": 0.9302,
       "step": 7831
     },
     {
-      "epoch": 0.21509983246821016,
+      "epoch": 0.22224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.8256489540899177e-05,
-      "loss": 1.0565,
+      "learning_rate": 1.8123746474725087e-05,
+      "loss": 1.0027,
       "step": 7832
     },
     {
-      "epoch": 0.21512729669605338,
+      "epoch": 0.22227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.8255987657361017e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.812321049303443e-05,
+      "loss": 0.9315,
       "step": 7833
     },
     {
-      "epoch": 0.21515476092389663,
+      "epoch": 0.22230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.8255485708498287e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.812267444272665e-05,
+      "loss": 0.9847,
       "step": 7834
     },
     {
-      "epoch": 0.21518222515173985,
+      "epoch": 0.22233257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 1.8254983694314957e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.812213832380627e-05,
+      "loss": 1.0935,
       "step": 7835
     },
     {
-      "epoch": 0.2152096893795831,
+      "epoch": 0.22236095346197504,
       "grad_norm": 0.0,
-      "learning_rate": 1.8254481614815e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.8121602136277828e-05,
+      "loss": 1.0518,
       "step": 7836
     },
     {
-      "epoch": 0.21523715360742632,
+      "epoch": 0.22238933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 1.8253979470002385e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.8121065880145846e-05,
+      "loss": 1.0937,
       "step": 7837
     },
     {
-      "epoch": 0.21526461783526957,
+      "epoch": 0.22241770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 1.8253477259881087e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.8120529555414855e-05,
+      "loss": 0.9798,
       "step": 7838
     },
     {
-      "epoch": 0.2152920820631128,
+      "epoch": 0.2224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8252974984455076e-05,
-      "loss": 1.0188,
+      "learning_rate": 1.8119993162089387e-05,
+      "loss": 0.9093,
       "step": 7839
     },
     {
-      "epoch": 0.21531954629095604,
+      "epoch": 0.2224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.8252472643728335e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.811945670017397e-05,
+      "loss": 0.9592,
       "step": 7840
     },
     {
-      "epoch": 0.21534701051879926,
+      "epoch": 0.22250283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 1.8251970237704833e-05,
-      "loss": 0.9627,
+      "learning_rate": 1.811892016967314e-05,
+      "loss": 0.9955,
       "step": 7841
     },
     {
-      "epoch": 0.2153744747466425,
+      "epoch": 0.2225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.8251467766388546e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.811838357059142e-05,
+      "loss": 1.0943,
       "step": 7842
     },
     {
-      "epoch": 0.21540193897448573,
+      "epoch": 0.22255959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 1.825096522978345e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.8117846902933355e-05,
+      "loss": 1.0003,
       "step": 7843
     },
     {
-      "epoch": 0.21542940320232898,
+      "epoch": 0.22258796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.8250462627893516e-05,
-      "loss": 1.12,
+      "learning_rate": 1.811731016670347e-05,
+      "loss": 1.0271,
       "step": 7844
     },
     {
-      "epoch": 0.2154568674301722,
+      "epoch": 0.22261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.824995996072273e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.81167733619063e-05,
+      "loss": 0.9084,
       "step": 7845
     },
     {
-      "epoch": 0.21548433165801542,
+      "epoch": 0.22264472190692394,
       "grad_norm": 0.0,
-      "learning_rate": 1.824945722827507e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.8116236488546388e-05,
+      "loss": 0.8946,
       "step": 7846
     },
     {
-      "epoch": 0.21551179588585867,
+      "epoch": 0.22267309875141883,
       "grad_norm": 0.0,
-      "learning_rate": 1.82489544305545e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.8115699546628254e-05,
+      "loss": 0.8899,
       "step": 7847
     },
     {
-      "epoch": 0.2155392601137019,
+      "epoch": 0.22270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.8248451567565013e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.8115162536156445e-05,
+      "loss": 1.0276,
       "step": 7848
     },
     {
-      "epoch": 0.21556672434154514,
+      "epoch": 0.22272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.824794863931058e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.811462545713549e-05,
+      "loss": 0.9598,
       "step": 7849
     },
     {
-      "epoch": 0.21559418856938836,
+      "epoch": 0.22275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.8247445645795182e-05,
-      "loss": 1.0263,
+      "learning_rate": 1.811408830956993e-05,
+      "loss": 0.9578,
       "step": 7850
     },
     {
-      "epoch": 0.2156216527972316,
+      "epoch": 0.2227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.82469425870228e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.81135510934643e-05,
+      "loss": 1.0252,
       "step": 7851
     },
     {
-      "epoch": 0.21564911702507483,
+      "epoch": 0.2228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.8246439462997413e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.811301380882314e-05,
+      "loss": 1.0544,
       "step": 7852
     },
     {
-      "epoch": 0.21567658125291808,
+      "epoch": 0.22284335981838818,
       "grad_norm": 0.0,
-      "learning_rate": 1.8245936273723003e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.8112476455650987e-05,
+      "loss": 0.9787,
       "step": 7853
     },
     {
-      "epoch": 0.2157040454807613,
+      "epoch": 0.2228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.8245433019203555e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.8111939033952382e-05,
+      "loss": 1.0655,
       "step": 7854
     },
     {
-      "epoch": 0.21573150970860455,
+      "epoch": 0.22290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.8244929699443038e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.8111401543731862e-05,
+      "loss": 1.0372,
       "step": 7855
     },
     {
-      "epoch": 0.21575897393644777,
+      "epoch": 0.22292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.8244426314445453e-05,
-      "loss": 1.0235,
+      "learning_rate": 1.8110863984993972e-05,
+      "loss": 0.8988,
       "step": 7856
     },
     {
-      "epoch": 0.21578643816429102,
+      "epoch": 0.22295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.8243922864214767e-05,
-      "loss": 1.0563,
+      "learning_rate": 1.8110326357743247e-05,
+      "loss": 1.0685,
       "step": 7857
     },
     {
-      "epoch": 0.21581390239213424,
+      "epoch": 0.22298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.824341934875497e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.810978866198423e-05,
+      "loss": 1.0578,
       "step": 7858
     },
     {
-      "epoch": 0.2158413666199775,
+      "epoch": 0.22301362088535756,
       "grad_norm": 0.0,
-      "learning_rate": 1.824291576807005e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.8109250897721465e-05,
+      "loss": 0.9211,
       "step": 7859
     },
     {
-      "epoch": 0.21586883084782071,
+      "epoch": 0.22304199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 1.8242412122163987e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.8108713064959493e-05,
+      "loss": 0.9717,
       "step": 7860
     },
     {
-      "epoch": 0.21589629507566394,
+      "epoch": 0.22307037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 1.8241908411040763e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.8108175163702855e-05,
+      "loss": 1.0399,
       "step": 7861
     },
     {
-      "epoch": 0.21592375930350718,
+      "epoch": 0.22309875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.8241404634704368e-05,
-      "loss": 1.0664,
+      "learning_rate": 1.8107637193956102e-05,
+      "loss": 0.9305,
       "step": 7862
     },
     {
-      "epoch": 0.2159512235313504,
+      "epoch": 0.2231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.824090079315879e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.8107099155723767e-05,
+      "loss": 0.9897,
       "step": 7863
     },
     {
-      "epoch": 0.21597868775919365,
+      "epoch": 0.223155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.824039688640801e-05,
-      "loss": 1.092,
+      "learning_rate": 1.8106561049010404e-05,
+      "loss": 1.1283,
       "step": 7864
     },
     {
-      "epoch": 0.21600615198703688,
+      "epoch": 0.2231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.8239892914456017e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.8106022873820553e-05,
+      "loss": 0.9719,
       "step": 7865
     },
     {
-      "epoch": 0.21603361621488013,
+      "epoch": 0.2232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.8239388877306803e-05,
-      "loss": 0.8445,
+      "learning_rate": 1.8105484630158768e-05,
+      "loss": 0.9671,
       "step": 7866
     },
     {
-      "epoch": 0.21606108044272335,
+      "epoch": 0.2232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.823888477496435e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.8104946318029585e-05,
+      "loss": 0.9945,
       "step": 7867
     },
     {
-      "epoch": 0.2160885446705666,
+      "epoch": 0.22326901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 1.823838060743265e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.8104407937437558e-05,
+      "loss": 0.9671,
       "step": 7868
     },
     {
-      "epoch": 0.21611600889840982,
+      "epoch": 0.22329738933030646,
       "grad_norm": 0.0,
-      "learning_rate": 1.8237876374715688e-05,
-      "loss": 1.0482,
+      "learning_rate": 1.810386948838723e-05,
+      "loss": 1.0787,
       "step": 7869
     },
     {
-      "epoch": 0.21614347312625307,
+      "epoch": 0.22332576617480135,
       "grad_norm": 0.0,
-      "learning_rate": 1.823737207681746e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.8103330970883156e-05,
+      "loss": 0.9566,
       "step": 7870
     },
     {
-      "epoch": 0.2161709373540963,
+      "epoch": 0.22335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.8236867713741957e-05,
-      "loss": 0.9372,
+      "learning_rate": 1.810279238492988e-05,
+      "loss": 1.074,
       "step": 7871
     },
     {
-      "epoch": 0.21619840158193954,
+      "epoch": 0.22338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.823636328549316e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.810225373053195e-05,
+      "loss": 0.9885,
       "step": 7872
     },
     {
-      "epoch": 0.21622586580978276,
+      "epoch": 0.22341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.823585879207507e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.8101715007693923e-05,
+      "loss": 0.9389,
       "step": 7873
     },
     {
-      "epoch": 0.21625333003762598,
+      "epoch": 0.22343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.8235354233491675e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.8101176216420343e-05,
+      "loss": 0.9829,
       "step": 7874
     },
     {
-      "epoch": 0.21628079426546923,
+      "epoch": 0.2234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.823484960974696e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.810063735671576e-05,
+      "loss": 0.9799,
       "step": 7875
     },
     {
-      "epoch": 0.21630825849331245,
+      "epoch": 0.22349602724177073,
       "grad_norm": 0.0,
-      "learning_rate": 1.8234344920844936e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.8100098428584733e-05,
+      "loss": 1.0074,
       "step": 7876
     },
     {
-      "epoch": 0.2163357227211557,
+      "epoch": 0.22352440408626562,
       "grad_norm": 0.0,
-      "learning_rate": 1.8233840166789577e-05,
-      "loss": 1.0719,
+      "learning_rate": 1.809955943203181e-05,
+      "loss": 0.9315,
       "step": 7877
     },
     {
-      "epoch": 0.21636318694899892,
+      "epoch": 0.2235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8233335347584892e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.8099020367061547e-05,
+      "loss": 1.0142,
       "step": 7878
     },
     {
-      "epoch": 0.21639065117684217,
+      "epoch": 0.2235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.8232830463234863e-05,
-      "loss": 1.0747,
+      "learning_rate": 1.809848123367849e-05,
+      "loss": 0.9672,
       "step": 7879
     },
     {
-      "epoch": 0.2164181154046854,
+      "epoch": 0.22360953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.823232551374349e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.80979420318872e-05,
+      "loss": 0.968,
       "step": 7880
     },
     {
-      "epoch": 0.21644557963252864,
+      "epoch": 0.22363791146424516,
       "grad_norm": 0.0,
-      "learning_rate": 1.8231820499114773e-05,
-      "loss": 1.0091,
+      "learning_rate": 1.809740276169223e-05,
+      "loss": 0.9436,
       "step": 7881
     },
     {
-      "epoch": 0.21647304386037186,
+      "epoch": 0.22366628830874008,
       "grad_norm": 0.0,
-      "learning_rate": 1.82313154193527e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.8096863423098136e-05,
+      "loss": 0.8457,
       "step": 7882
     },
     {
-      "epoch": 0.2165005080882151,
+      "epoch": 0.22369466515323497,
       "grad_norm": 0.0,
-      "learning_rate": 1.8230810274461273e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.809632401610947e-05,
+      "loss": 1.0229,
       "step": 7883
     },
     {
-      "epoch": 0.21652797231605833,
+      "epoch": 0.22372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.8230305064444483e-05,
-      "loss": 1.0732,
+      "learning_rate": 1.809578454073079e-05,
+      "loss": 0.9693,
       "step": 7884
     },
     {
-      "epoch": 0.21655543654390158,
+      "epoch": 0.22375141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.8229799789306337e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.8095244996966655e-05,
+      "loss": 0.8691,
       "step": 7885
     },
     {
-      "epoch": 0.2165829007717448,
+      "epoch": 0.22377979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.8229294449050826e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.809470538482163e-05,
+      "loss": 1.1315,
       "step": 7886
     },
     {
-      "epoch": 0.21661036499958802,
+      "epoch": 0.22380817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.8228789043681944e-05,
-      "loss": 1.0926,
+      "learning_rate": 1.8094165704300253e-05,
+      "loss": 0.9922,
       "step": 7887
     },
     {
-      "epoch": 0.21663782922743127,
+      "epoch": 0.22383654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 1.8228283573203702e-05,
-      "loss": 1.0786,
+      "learning_rate": 1.80936259554071e-05,
+      "loss": 1.0311,
       "step": 7888
     },
     {
-      "epoch": 0.2166652934552745,
+      "epoch": 0.22386492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 1.822777803762009e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.8093086138146723e-05,
+      "loss": 1.0676,
       "step": 7889
     },
     {
-      "epoch": 0.21669275768311774,
+      "epoch": 0.2238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.822727243693511e-05,
-      "loss": 1.0202,
+      "learning_rate": 1.8092546252523685e-05,
+      "loss": 1.0621,
       "step": 7890
     },
     {
-      "epoch": 0.21672022191096096,
+      "epoch": 0.2239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.8226766771152764e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.8092006298542544e-05,
+      "loss": 1.0565,
       "step": 7891
     },
     {
-      "epoch": 0.2167476861388042,
+      "epoch": 0.22395005675368898,
       "grad_norm": 0.0,
-      "learning_rate": 1.8226261040277056e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.8091466276207864e-05,
+      "loss": 0.9675,
       "step": 7892
     },
     {
-      "epoch": 0.21677515036664743,
+      "epoch": 0.22397843359818387,
       "grad_norm": 0.0,
-      "learning_rate": 1.822575524431198e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.80909261855242e-05,
+      "loss": 1.0822,
       "step": 7893
     },
     {
-      "epoch": 0.21680261459449068,
+      "epoch": 0.22400681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 1.822524938326154e-05,
-      "loss": 1.0221,
+      "learning_rate": 1.8090386026496123e-05,
+      "loss": 1.0111,
       "step": 7894
     },
     {
-      "epoch": 0.2168300788223339,
+      "epoch": 0.22403518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.822474345712975e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.808984579912819e-05,
+      "loss": 0.9957,
       "step": 7895
     },
     {
-      "epoch": 0.21685754305017715,
+      "epoch": 0.22406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.82242374659206e-05,
-      "loss": 1.0626,
+      "learning_rate": 1.8089305503424966e-05,
+      "loss": 1.0381,
       "step": 7896
     },
     {
-      "epoch": 0.21688500727802038,
+      "epoch": 0.22409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.8223731409638094e-05,
-      "loss": 0.8917,
+      "learning_rate": 1.8088765139391014e-05,
+      "loss": 0.9923,
       "step": 7897
     },
     {
-      "epoch": 0.21691247150586362,
+      "epoch": 0.22412031782065833,
       "grad_norm": 0.0,
-      "learning_rate": 1.822322528828624e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.8088224707030903e-05,
+      "loss": 0.953,
       "step": 7898
     },
     {
-      "epoch": 0.21693993573370685,
+      "epoch": 0.22414869466515325,
       "grad_norm": 0.0,
-      "learning_rate": 1.8222719101869044e-05,
-      "loss": 1.049,
+      "learning_rate": 1.808768420634919e-05,
+      "loss": 1.0862,
       "step": 7899
     },
     {
-      "epoch": 0.2169673999615501,
+      "epoch": 0.22417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.8222212850390507e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.8087143637350445e-05,
+      "loss": 1.1074,
       "step": 7900
     },
     {
-      "epoch": 0.21699486418939332,
+      "epoch": 0.22420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.822170653385464e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.8086603000039237e-05,
+      "loss": 1.0195,
       "step": 7901
     },
     {
-      "epoch": 0.21702232841723654,
+      "epoch": 0.2242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.8221200152265446e-05,
-      "loss": 0.9936,
+      "learning_rate": 1.8086062294420127e-05,
+      "loss": 0.9576,
       "step": 7902
     },
     {
-      "epoch": 0.2170497926450798,
+      "epoch": 0.2242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.822069370562693e-05,
-      "loss": 1.047,
+      "learning_rate": 1.8085521520497686e-05,
+      "loss": 1.0851,
       "step": 7903
     },
     {
-      "epoch": 0.217077256872923,
+      "epoch": 0.22429057888762768,
       "grad_norm": 0.0,
-      "learning_rate": 1.82201871939431e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.8084980678276482e-05,
+      "loss": 0.9754,
       "step": 7904
     },
     {
-      "epoch": 0.21710472110076626,
+      "epoch": 0.2243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.8219680617217967e-05,
-      "loss": 0.9683,
+      "learning_rate": 1.8084439767761084e-05,
+      "loss": 1.1398,
       "step": 7905
     },
     {
-      "epoch": 0.21713218532860948,
+      "epoch": 0.2243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.8219173975455535e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.8083898788956058e-05,
+      "loss": 1.0647,
       "step": 7906
     },
     {
-      "epoch": 0.21715964955645273,
+      "epoch": 0.22437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.821866726865982e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.8083357741865976e-05,
+      "loss": 1.0289,
       "step": 7907
     },
     {
-      "epoch": 0.21718711378429595,
+      "epoch": 0.22440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.821816049683482e-05,
-      "loss": 0.904,
+      "learning_rate": 1.8082816626495407e-05,
+      "loss": 0.9761,
       "step": 7908
     },
     {
-      "epoch": 0.2172145780121392,
+      "epoch": 0.22443246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.821765365998455e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.8082275442848923e-05,
+      "loss": 1.0649,
       "step": 7909
     },
     {
-      "epoch": 0.21724204223998242,
+      "epoch": 0.22446083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.8217146758113026e-05,
-      "loss": 1.0899,
+      "learning_rate": 1.8081734190931096e-05,
+      "loss": 1.0669,
       "step": 7910
     },
     {
-      "epoch": 0.21726950646782567,
+      "epoch": 0.22448921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 1.821663979122425e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.80811928707465e-05,
+      "loss": 0.9711,
       "step": 7911
     },
     {
-      "epoch": 0.2172969706956689,
+      "epoch": 0.22451759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 1.821613275932224e-05,
-      "loss": 1.1296,
+      "learning_rate": 1.80806514822997e-05,
+      "loss": 0.9438,
       "step": 7912
     },
     {
-      "epoch": 0.21732443492351214,
+      "epoch": 0.22454597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.8215625662411004e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.8080110025595273e-05,
+      "loss": 0.9833,
       "step": 7913
     },
     {
-      "epoch": 0.21735189915135536,
+      "epoch": 0.2245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.8215118500494554e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.8079568500637795e-05,
+      "loss": 1.0,
       "step": 7914
     },
     {
-      "epoch": 0.21737936337919858,
+      "epoch": 0.2246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.8214611273576904e-05,
-      "loss": 0.955,
+      "learning_rate": 1.807902690743184e-05,
+      "loss": 1.0821,
       "step": 7915
     },
     {
-      "epoch": 0.21740682760704183,
+      "epoch": 0.22463110102156642,
       "grad_norm": 0.0,
-      "learning_rate": 1.821410398166207e-05,
-      "loss": 1.0791,
+      "learning_rate": 1.8078485245981982e-05,
+      "loss": 0.8733,
       "step": 7916
     },
     {
-      "epoch": 0.21743429183488505,
+      "epoch": 0.2246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.8213596624754063e-05,
-      "loss": 0.8924,
+      "learning_rate": 1.8077943516292795e-05,
+      "loss": 0.8976,
       "step": 7917
     },
     {
-      "epoch": 0.2174617560627283,
+      "epoch": 0.2246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.8213089202856896e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.8077401718368853e-05,
+      "loss": 0.9564,
       "step": 7918
     },
     {
-      "epoch": 0.21748922029057152,
+      "epoch": 0.22471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.8212581715974586e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.807685985221474e-05,
+      "loss": 0.9425,
       "step": 7919
     },
     {
-      "epoch": 0.21751668451841477,
+      "epoch": 0.22474460839954596,
       "grad_norm": 0.0,
-      "learning_rate": 1.821207416411115e-05,
-      "loss": 1.1525,
+      "learning_rate": 1.8076317917835027e-05,
+      "loss": 0.9051,
       "step": 7920
     },
     {
-      "epoch": 0.217544148746258,
+      "epoch": 0.22477298524404085,
       "grad_norm": 0.0,
-      "learning_rate": 1.82115665472706e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.8075775915234294e-05,
+      "loss": 0.8612,
       "step": 7921
     },
     {
-      "epoch": 0.21757161297410124,
+      "epoch": 0.22480136208853577,
       "grad_norm": 0.0,
-      "learning_rate": 1.8211058865456955e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.8075233844417116e-05,
+      "loss": 0.7729,
       "step": 7922
     },
     {
-      "epoch": 0.21759907720194446,
+      "epoch": 0.22482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.8210551118674233e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.8074691705388076e-05,
+      "loss": 0.9667,
       "step": 7923
     },
     {
-      "epoch": 0.2176265414297877,
+      "epoch": 0.22485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.8210043306926452e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.8074149498151753e-05,
+      "loss": 1.041,
       "step": 7924
     },
     {
-      "epoch": 0.21765400565763093,
+      "epoch": 0.22488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.820953543021762e-05,
-      "loss": 1.0828,
+      "learning_rate": 1.807360722271273e-05,
+      "loss": 0.9606,
       "step": 7925
     },
     {
-      "epoch": 0.21768146988547418,
+      "epoch": 0.22491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.8209027488551774e-05,
-      "loss": 1.01,
+      "learning_rate": 1.8073064879075577e-05,
+      "loss": 0.9373,
       "step": 7926
     },
     {
-      "epoch": 0.2177089341133174,
+      "epoch": 0.2249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.820851948193292e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.8072522467244885e-05,
+      "loss": 0.9496,
       "step": 7927
     },
     {
-      "epoch": 0.21773639834116063,
+      "epoch": 0.22497162315550512,
       "grad_norm": 0.0,
-      "learning_rate": 1.8208011410365073e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.8071979987225233e-05,
+      "loss": 0.98,
       "step": 7928
     },
     {
-      "epoch": 0.21776386256900387,
+      "epoch": 0.225,
       "grad_norm": 0.0,
-      "learning_rate": 1.8207503273852267e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.8071437439021204e-05,
+      "loss": 0.9417,
       "step": 7929
     },
     {
-      "epoch": 0.2177913267968471,
+      "epoch": 0.2250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.8206995072398515e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.8070894822637375e-05,
+      "loss": 0.9362,
       "step": 7930
     },
     {
-      "epoch": 0.21781879102469034,
+      "epoch": 0.22505675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.8206486806007836e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.8070352138078342e-05,
+      "loss": 0.9944,
       "step": 7931
     },
     {
-      "epoch": 0.21784625525253357,
+      "epoch": 0.22508513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 1.8205978474684255e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.8069809385348677e-05,
+      "loss": 0.9855,
       "step": 7932
     },
     {
-      "epoch": 0.21787371948037682,
+      "epoch": 0.22511350737797958,
       "grad_norm": 0.0,
-      "learning_rate": 1.8205470078431797e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.806926656445297e-05,
+      "loss": 0.9311,
       "step": 7933
     },
     {
-      "epoch": 0.21790118370822004,
+      "epoch": 0.22514188422247447,
       "grad_norm": 0.0,
-      "learning_rate": 1.820496161725448e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.806872367539581e-05,
+      "loss": 0.8207,
       "step": 7934
     },
     {
-      "epoch": 0.21792864793606329,
+      "epoch": 0.22517026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.8204453091156327e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.8068180718181773e-05,
+      "loss": 1.0557,
       "step": 7935
     },
     {
-      "epoch": 0.2179561121639065,
+      "epoch": 0.22519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.8203944500141365e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.8067637692815455e-05,
+      "loss": 0.9159,
       "step": 7936
     },
     {
-      "epoch": 0.21798357639174976,
+      "epoch": 0.22522701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.8203435844213613e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.8067094599301433e-05,
+      "loss": 1.0025,
       "step": 7937
     },
     {
-      "epoch": 0.21801104061959298,
+      "epoch": 0.22525539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.82029271233771e-05,
-      "loss": 0.94,
+      "learning_rate": 1.8066551437644306e-05,
+      "loss": 0.9013,
       "step": 7938
     },
     {
-      "epoch": 0.21803850484743623,
+      "epoch": 0.22528376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.8202418337635854e-05,
-      "loss": 0.8438,
+      "learning_rate": 1.8066008207848653e-05,
+      "loss": 1.1318,
       "step": 7939
     },
     {
-      "epoch": 0.21806596907527945,
+      "epoch": 0.22531214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.8201909486993892e-05,
-      "loss": 1.0488,
+      "learning_rate": 1.806546490991907e-05,
+      "loss": 1.071,
       "step": 7940
     },
     {
-      "epoch": 0.21809343330312267,
+      "epoch": 0.2253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.820140057145525e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.8064921543860138e-05,
+      "loss": 1.0137,
       "step": 7941
     },
     {
-      "epoch": 0.21812089753096592,
+      "epoch": 0.2253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.8200891591023943e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.806437810967645e-05,
+      "loss": 1.0071,
       "step": 7942
     },
     {
-      "epoch": 0.21814836175880914,
+      "epoch": 0.22539727582292848,
       "grad_norm": 0.0,
-      "learning_rate": 1.820038254570401e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.8063834607372603e-05,
+      "loss": 0.9547,
       "step": 7943
     },
     {
-      "epoch": 0.2181758259866524,
+      "epoch": 0.22542565266742337,
       "grad_norm": 0.0,
-      "learning_rate": 1.819987343549947e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.806329103695318e-05,
+      "loss": 0.9978,
       "step": 7944
     },
     {
-      "epoch": 0.2182032902144956,
+      "epoch": 0.22545402951191829,
       "grad_norm": 0.0,
-      "learning_rate": 1.819936426041436e-05,
-      "loss": 0.8884,
+      "learning_rate": 1.8062747398422775e-05,
+      "loss": 1.0697,
       "step": 7945
     },
     {
-      "epoch": 0.21823075444233886,
+      "epoch": 0.22548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.81988550204527e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.8062203691785977e-05,
+      "loss": 0.9829,
       "step": 7946
     },
     {
-      "epoch": 0.21825821867018208,
+      "epoch": 0.22551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.8198345715618524e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.8061659917047384e-05,
+      "loss": 1.0438,
       "step": 7947
     },
     {
-      "epoch": 0.21828568289802533,
+      "epoch": 0.22553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.819783634591586e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.806111607421159e-05,
+      "loss": 0.9528,
       "step": 7948
     },
     {
-      "epoch": 0.21831314712586855,
+      "epoch": 0.22556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.8197326911348742e-05,
-      "loss": 0.8643,
+      "learning_rate": 1.8060572163283186e-05,
+      "loss": 0.9296,
       "step": 7949
     },
     {
-      "epoch": 0.2183406113537118,
+      "epoch": 0.22559591373439272,
       "grad_norm": 0.0,
-      "learning_rate": 1.8196817411921197e-05,
-      "loss": 1.1148,
+      "learning_rate": 1.8060028184266766e-05,
+      "loss": 1.1524,
       "step": 7950
     },
     {
-      "epoch": 0.21836807558155502,
+      "epoch": 0.22562429057888764,
       "grad_norm": 0.0,
-      "learning_rate": 1.8196307847637258e-05,
-      "loss": 1.0594,
+      "learning_rate": 1.8059484137166925e-05,
+      "loss": 0.9826,
       "step": 7951
     },
     {
-      "epoch": 0.21839553980939827,
+      "epoch": 0.22565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8195798218500955e-05,
-      "loss": 1.1447,
+      "learning_rate": 1.8058940021988257e-05,
+      "loss": 1.0109,
       "step": 7952
     },
     {
-      "epoch": 0.2184230040372415,
+      "epoch": 0.2256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.8195288524516322e-05,
-      "loss": 0.9542,
+      "learning_rate": 1.8058395838735364e-05,
+      "loss": 0.9218,
       "step": 7953
     },
     {
-      "epoch": 0.21845046826508474,
+      "epoch": 0.2257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.8194778765687393e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.8057851587412837e-05,
+      "loss": 0.9121,
       "step": 7954
     },
     {
-      "epoch": 0.21847793249292796,
+      "epoch": 0.2257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.8194268942018197e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.8057307268025273e-05,
+      "loss": 1.0127,
       "step": 7955
     },
     {
-      "epoch": 0.21850539672077118,
+      "epoch": 0.2257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.8193759053512777e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.8056762880577277e-05,
+      "loss": 1.0422,
       "step": 7956
     },
     {
-      "epoch": 0.21853286094861443,
+      "epoch": 0.225794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8193249100175156e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.805621842507344e-05,
+      "loss": 1.0754,
       "step": 7957
     },
     {
-      "epoch": 0.21856032517645765,
+      "epoch": 0.22582292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 1.8192739082009377e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.8055673901518365e-05,
+      "loss": 1.0226,
       "step": 7958
     },
     {
-      "epoch": 0.2185877894043009,
+      "epoch": 0.22585130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.819222899901947e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.8055129309916652e-05,
+      "loss": 1.01,
       "step": 7959
     },
     {
-      "epoch": 0.21861525363214412,
+      "epoch": 0.22587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.8191718851209473e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.8054584650272897e-05,
+      "loss": 1.0438,
       "step": 7960
     },
     {
-      "epoch": 0.21864271785998737,
+      "epoch": 0.22590805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.8191208638583426e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.8054039922591705e-05,
+      "loss": 0.9987,
       "step": 7961
     },
     {
-      "epoch": 0.2186701820878306,
+      "epoch": 0.22593643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 1.8190698361145365e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.8053495126877677e-05,
+      "loss": 0.9929,
       "step": 7962
     },
     {
-      "epoch": 0.21869764631567384,
+      "epoch": 0.22596481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.8190188018899322e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.8052950263135408e-05,
+      "loss": 0.9777,
       "step": 7963
     },
     {
-      "epoch": 0.21872511054351707,
+      "epoch": 0.22599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.818967761184934e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.805240533136951e-05,
+      "loss": 0.9422,
       "step": 7964
     },
     {
-      "epoch": 0.21875257477136031,
+      "epoch": 0.22602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.8189167139999457e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.8051860331584582e-05,
+      "loss": 0.9334,
       "step": 7965
     },
     {
-      "epoch": 0.21878003899920354,
+      "epoch": 0.226049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8188656603353712e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.805131526378523e-05,
+      "loss": 0.8978,
       "step": 7966
     },
     {
-      "epoch": 0.21880750322704678,
+      "epoch": 0.2260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.8188146001916146e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.805077012797605e-05,
+      "loss": 1.1444,
       "step": 7967
     },
     {
-      "epoch": 0.21883496745489,
+      "epoch": 0.2261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.8187635335690793e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.8050224924161658e-05,
+      "loss": 1.0506,
       "step": 7968
     },
     {
-      "epoch": 0.21886243168273323,
+      "epoch": 0.2261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.81871246046817e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.8049679652346653e-05,
+      "loss": 0.9066,
       "step": 7969
     },
     {
-      "epoch": 0.21888989591057648,
+      "epoch": 0.22616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.8186613808892902e-05,
-      "loss": 1.0164,
+      "learning_rate": 1.804913431253564e-05,
+      "loss": 0.9376,
       "step": 7970
     },
     {
-      "epoch": 0.2189173601384197,
+      "epoch": 0.22619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.818610294832845e-05,
-      "loss": 0.985,
+      "learning_rate": 1.804858890473323e-05,
+      "loss": 0.9059,
       "step": 7971
     },
     {
-      "epoch": 0.21894482436626295,
+      "epoch": 0.22622020431328035,
       "grad_norm": 0.0,
-      "learning_rate": 1.8185592022992376e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.804804342894403e-05,
+      "loss": 1.0349,
       "step": 7972
     },
     {
-      "epoch": 0.21897228859410617,
+      "epoch": 0.22624858115777527,
       "grad_norm": 0.0,
-      "learning_rate": 1.818508103288873e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.8047497885172643e-05,
+      "loss": 1.0327,
       "step": 7973
     },
     {
-      "epoch": 0.21899975282194942,
+      "epoch": 0.22627695800227016,
       "grad_norm": 0.0,
-      "learning_rate": 1.8184569978021548e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.804695227342368e-05,
+      "loss": 0.8487,
       "step": 7974
     },
     {
-      "epoch": 0.21902721704979264,
+      "epoch": 0.22630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.818405885839488e-05,
-      "loss": 0.919,
+      "learning_rate": 1.8046406593701748e-05,
+      "loss": 0.9662,
       "step": 7975
     },
     {
-      "epoch": 0.2190546812776359,
+      "epoch": 0.22633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.818354767401277e-05,
-      "loss": 1.0894,
+      "learning_rate": 1.8045860846011456e-05,
+      "loss": 1.0429,
       "step": 7976
     },
     {
-      "epoch": 0.2190821455054791,
+      "epoch": 0.22636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.818303642487926e-05,
-      "loss": 1.1045,
+      "learning_rate": 1.804531503035742e-05,
+      "loss": 0.9508,
       "step": 7977
     },
     {
-      "epoch": 0.21910960973332236,
+      "epoch": 0.2263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.81825251109984e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.8044769146744242e-05,
+      "loss": 0.9725,
       "step": 7978
     },
     {
-      "epoch": 0.21913707396116558,
+      "epoch": 0.22641884222474462,
       "grad_norm": 0.0,
-      "learning_rate": 1.8182013732374223e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.804422319517654e-05,
+      "loss": 0.9556,
       "step": 7979
     },
     {
-      "epoch": 0.21916453818900883,
+      "epoch": 0.2264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.818150228901079e-05,
-      "loss": 1.0171,
+      "learning_rate": 1.804367717565892e-05,
+      "loss": 0.9001,
       "step": 7980
     },
     {
-      "epoch": 0.21919200241685205,
+      "epoch": 0.2264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.818099078091214e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.8043131088196003e-05,
+      "loss": 1.0255,
       "step": 7981
     },
     {
-      "epoch": 0.21921946664469527,
+      "epoch": 0.22650397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 1.8180479208082323e-05,
-      "loss": 1.0561,
+      "learning_rate": 1.8042584932792392e-05,
+      "loss": 0.989,
       "step": 7982
     },
     {
-      "epoch": 0.21924693087253852,
+      "epoch": 0.22653234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 1.8179967570525386e-05,
-      "loss": 1.0546,
+      "learning_rate": 1.804203870945271e-05,
+      "loss": 0.9955,
       "step": 7983
     },
     {
-      "epoch": 0.21927439510038174,
+      "epoch": 0.22656072644721906,
       "grad_norm": 0.0,
-      "learning_rate": 1.8179455868245374e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.804149241818156e-05,
+      "loss": 1.1599,
       "step": 7984
     },
     {
-      "epoch": 0.219301859328225,
+      "epoch": 0.22658910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 1.817894410124634e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.804094605898356e-05,
+      "loss": 0.9493,
       "step": 7985
     },
     {
-      "epoch": 0.2193293235560682,
+      "epoch": 0.22661748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.8178432269532334e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.8040399631863334e-05,
+      "loss": 0.9731,
       "step": 7986
     },
     {
-      "epoch": 0.21935678778391146,
+      "epoch": 0.22664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.81779203731074e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.8039853136825485e-05,
+      "loss": 1.0403,
       "step": 7987
     },
     {
-      "epoch": 0.21938425201175468,
+      "epoch": 0.22667423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.8177408411975594e-05,
-      "loss": 0.9165,
+      "learning_rate": 1.803930657387464e-05,
+      "loss": 1.0149,
       "step": 7988
     },
     {
-      "epoch": 0.21941171623959793,
+      "epoch": 0.22670261066969352,
       "grad_norm": 0.0,
-      "learning_rate": 1.8176896386140966e-05,
-      "loss": 0.9931,
+      "learning_rate": 1.8038759943015406e-05,
+      "loss": 1.067,
       "step": 7989
     },
     {
-      "epoch": 0.21943918046744115,
+      "epoch": 0.2267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.8176384295607567e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.8038213244252408e-05,
+      "loss": 1.0581,
       "step": 7990
     },
     {
-      "epoch": 0.2194666446952844,
+      "epoch": 0.22675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 1.8175872140379448e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.803766647759026e-05,
+      "loss": 0.9147,
       "step": 7991
     },
     {
-      "epoch": 0.21949410892312762,
+      "epoch": 0.2267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.817535992046066e-05,
-      "loss": 1.1091,
+      "learning_rate": 1.8037119643033586e-05,
+      "loss": 1.0111,
       "step": 7992
     },
     {
-      "epoch": 0.21952157315097087,
+      "epoch": 0.2268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.8174847635855256e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.8036572740586997e-05,
+      "loss": 0.9139,
       "step": 7993
     },
     {
-      "epoch": 0.2195490373788141,
+      "epoch": 0.22684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.8174335286567294e-05,
-      "loss": 1.019,
+      "learning_rate": 1.803602577025512e-05,
+      "loss": 1.0207,
       "step": 7994
     },
     {
-      "epoch": 0.21957650160665731,
+      "epoch": 0.22687287173666287,
       "grad_norm": 0.0,
-      "learning_rate": 1.8173822872600822e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.803547873204257e-05,
+      "loss": 0.9706,
       "step": 7995
     },
     {
-      "epoch": 0.21960396583450056,
+      "epoch": 0.2269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.81733103939599e-05,
-      "loss": 1.0498,
+      "learning_rate": 1.8034931625953967e-05,
+      "loss": 1.0046,
       "step": 7996
     },
     {
-      "epoch": 0.21963143006234379,
+      "epoch": 0.22692962542565268,
       "grad_norm": 0.0,
-      "learning_rate": 1.8172797850648584e-05,
-      "loss": 1.032,
+      "learning_rate": 1.803438445199394e-05,
+      "loss": 1.0099,
       "step": 7997
     },
     {
-      "epoch": 0.21965889429018703,
+      "epoch": 0.22695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.817228524267092e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.8033837210167107e-05,
+      "loss": 1.0412,
       "step": 7998
     },
     {
-      "epoch": 0.21968635851803026,
+      "epoch": 0.22698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.817177257003097e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.8033289900478086e-05,
+      "loss": 0.9779,
       "step": 7999
     },
     {
-      "epoch": 0.2197138227458735,
+      "epoch": 0.22701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.817125983273279e-05,
-      "loss": 1.1141,
+      "learning_rate": 1.8032742522931507e-05,
+      "loss": 0.9762,
       "step": 8000
     },
     {
-      "epoch": 0.21974128697371673,
+      "epoch": 0.22704313280363222,
       "grad_norm": 0.0,
-      "learning_rate": 1.8170747030780436e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.8032195077531988e-05,
+      "loss": 0.9183,
       "step": 8001
     },
     {
-      "epoch": 0.21976875120155998,
+      "epoch": 0.22707150964812714,
       "grad_norm": 0.0,
-      "learning_rate": 1.8170234164177968e-05,
-      "loss": 0.9414,
+      "learning_rate": 1.8031647564284158e-05,
+      "loss": 0.977,
       "step": 8002
     },
     {
-      "epoch": 0.2197962154294032,
+      "epoch": 0.22709988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 1.8169721232929443e-05,
-      "loss": 1.0864,
+      "learning_rate": 1.8031099983192638e-05,
+      "loss": 0.9611,
       "step": 8003
     },
     {
-      "epoch": 0.21982367965724645,
+      "epoch": 0.22712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.816920823703892e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.803055233426206e-05,
+      "loss": 1.0699,
       "step": 8004
     },
     {
-      "epoch": 0.21985114388508967,
+      "epoch": 0.2271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.8168695176510452e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.8030004617497043e-05,
+      "loss": 0.917,
       "step": 8005
     },
     {
-      "epoch": 0.21987860811293292,
+      "epoch": 0.2271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8168182051348107e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.8029456832902215e-05,
+      "loss": 0.936,
       "step": 8006
     },
     {
-      "epoch": 0.21990607234077614,
+      "epoch": 0.22721339387060158,
       "grad_norm": 0.0,
-      "learning_rate": 1.8167668861555942e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.8028908980482203e-05,
+      "loss": 1.0458,
       "step": 8007
     },
     {
-      "epoch": 0.2199335365686194,
+      "epoch": 0.2272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.8167155607138015e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.802836106024164e-05,
+      "loss": 1.0955,
       "step": 8008
     },
     {
-      "epoch": 0.2199610007964626,
+      "epoch": 0.22727014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 1.816664228809839e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.8027813072185147e-05,
+      "loss": 1.0774,
       "step": 8009
     },
     {
-      "epoch": 0.21998846502430583,
+      "epoch": 0.22729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.816612890444113e-05,
-      "loss": 0.8974,
+      "learning_rate": 1.8027265016317358e-05,
+      "loss": 0.8791,
       "step": 8010
     },
     {
-      "epoch": 0.22001592925214908,
+      "epoch": 0.22732690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.816561545617029e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.80267168926429e-05,
+      "loss": 0.9434,
       "step": 8011
     },
     {
-      "epoch": 0.2200433934799923,
+      "epoch": 0.22735527809307604,
       "grad_norm": 0.0,
-      "learning_rate": 1.8165101943289944e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.8026168701166402e-05,
+      "loss": 1.0052,
       "step": 8012
     },
     {
-      "epoch": 0.22007085770783555,
+      "epoch": 0.22738365493757096,
       "grad_norm": 0.0,
-      "learning_rate": 1.8164588365804144e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.8025620441892498e-05,
+      "loss": 1.049,
       "step": 8013
     },
     {
-      "epoch": 0.22009832193567877,
+      "epoch": 0.22741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.816407472371696e-05,
-      "loss": 1.0708,
+      "learning_rate": 1.8025072114825817e-05,
+      "loss": 1.0266,
       "step": 8014
     },
     {
-      "epoch": 0.22012578616352202,
+      "epoch": 0.22744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.816356101703245e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.802452371997099e-05,
+      "loss": 1.0226,
       "step": 8015
     },
     {
-      "epoch": 0.22015325039136524,
+      "epoch": 0.22746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.8163047245754687e-05,
-      "loss": 1.0743,
+      "learning_rate": 1.8023975257332652e-05,
+      "loss": 1.0072,
       "step": 8016
     },
     {
-      "epoch": 0.2201807146192085,
+      "epoch": 0.2274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.8162533409887735e-05,
-      "loss": 1.1412,
+      "learning_rate": 1.8023426726915434e-05,
+      "loss": 1.0342,
       "step": 8017
     },
     {
-      "epoch": 0.2202081788470517,
+      "epoch": 0.2275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.8162019509435652e-05,
-      "loss": 0.9273,
+      "learning_rate": 1.802287812872397e-05,
+      "loss": 0.9958,
       "step": 8018
     },
     {
-      "epoch": 0.22023564307489496,
+      "epoch": 0.2275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.816150554440251e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.802232946276289e-05,
+      "loss": 0.9159,
       "step": 8019
     },
     {
-      "epoch": 0.22026310730273818,
+      "epoch": 0.2275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.8160991514792378e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.802178072903684e-05,
+      "loss": 1.002,
       "step": 8020
     },
     {
-      "epoch": 0.22029057153058143,
+      "epoch": 0.22761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.816047742060931e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.802123192755044e-05,
+      "loss": 0.8851,
       "step": 8021
     },
     {
-      "epoch": 0.22031803575842465,
+      "epoch": 0.22763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.8159963261857397e-05,
-      "loss": 1.0701,
+      "learning_rate": 1.8020683058308334e-05,
+      "loss": 1.0423,
       "step": 8022
     },
     {
-      "epoch": 0.22034549998626787,
+      "epoch": 0.22766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.8159449038540685e-05,
-      "loss": 1.0243,
+      "learning_rate": 1.8020134121315162e-05,
+      "loss": 1.0914,
       "step": 8023
     },
     {
-      "epoch": 0.22037296421411112,
+      "epoch": 0.22769580022701474,
       "grad_norm": 0.0,
-      "learning_rate": 1.8158934750663254e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.8019585116575554e-05,
+      "loss": 0.9833,
       "step": 8024
     },
     {
-      "epoch": 0.22040042844195434,
+      "epoch": 0.22772417707150966,
       "grad_norm": 0.0,
-      "learning_rate": 1.815842039822917e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.801903604409415e-05,
+      "loss": 1.0729,
       "step": 8025
     },
     {
-      "epoch": 0.2204278926697976,
+      "epoch": 0.22775255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 1.8157905981242504e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.801848690387559e-05,
+      "loss": 0.9164,
       "step": 8026
     },
     {
-      "epoch": 0.22045535689764081,
+      "epoch": 0.22778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.8157391499707326e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.801793769592451e-05,
+      "loss": 0.9959,
       "step": 8027
     },
     {
-      "epoch": 0.22048282112548406,
+      "epoch": 0.22780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.8156876953627707e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.801738842024555e-05,
+      "loss": 0.9894,
       "step": 8028
     },
     {
-      "epoch": 0.22051028535332728,
+      "epoch": 0.2278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.8156362343007717e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.8016839076843347e-05,
+      "loss": 0.9519,
       "step": 8029
     },
     {
-      "epoch": 0.22053774958117053,
+      "epoch": 0.2278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.815584766785143e-05,
-      "loss": 0.9549,
+      "learning_rate": 1.8016289665722545e-05,
+      "loss": 0.9305,
       "step": 8030
     },
     {
-      "epoch": 0.22056521380901375,
+      "epoch": 0.227894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.8155332928162918e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.8015740186887785e-05,
+      "loss": 0.9181,
       "step": 8031
     },
     {
-      "epoch": 0.220592678036857,
+      "epoch": 0.2279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.815481812394625e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.8015190640343707e-05,
+      "loss": 0.9884,
       "step": 8032
     },
     {
-      "epoch": 0.22062014226470023,
+      "epoch": 0.22795119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.8154303255205503e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.801464102609495e-05,
+      "loss": 1.067,
       "step": 8033
     },
     {
-      "epoch": 0.22064760649254347,
+      "epoch": 0.22797956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 1.815378832194475e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.8014091344146166e-05,
+      "loss": 1.0988,
       "step": 8034
     },
     {
-      "epoch": 0.2206750707203867,
+      "epoch": 0.22800794551645856,
       "grad_norm": 0.0,
-      "learning_rate": 1.8153273324168065e-05,
-      "loss": 0.936,
+      "learning_rate": 1.801354159450199e-05,
+      "loss": 1.1248,
       "step": 8035
     },
     {
-      "epoch": 0.22070253494822992,
+      "epoch": 0.22803632236095348,
       "grad_norm": 0.0,
-      "learning_rate": 1.8152758261879523e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.8012991777167065e-05,
+      "loss": 1.0085,
       "step": 8036
     },
     {
-      "epoch": 0.22072999917607317,
+      "epoch": 0.22806469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.81522431350832e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.8012441892146043e-05,
+      "loss": 1.005,
       "step": 8037
     },
     {
-      "epoch": 0.2207574634039164,
+      "epoch": 0.22809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.8151727943783173e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.8011891939443562e-05,
+      "loss": 0.9686,
       "step": 8038
     },
     {
-      "epoch": 0.22078492763175964,
+      "epoch": 0.22812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.8151212687983514e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.801134191906427e-05,
+      "loss": 1.0614,
       "step": 8039
     },
     {
-      "epoch": 0.22081239185960286,
+      "epoch": 0.22814982973893302,
       "grad_norm": 0.0,
-      "learning_rate": 1.8150697367688306e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.801079183101281e-05,
+      "loss": 1.0506,
       "step": 8040
     },
     {
-      "epoch": 0.2208398560874461,
+      "epoch": 0.2281782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.815018198290162e-05,
-      "loss": 0.8778,
+      "learning_rate": 1.8010241675293834e-05,
+      "loss": 0.9485,
       "step": 8041
     },
     {
-      "epoch": 0.22086732031528933,
+      "epoch": 0.22820658342792283,
       "grad_norm": 0.0,
-      "learning_rate": 1.814966653362754e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.8009691451911986e-05,
+      "loss": 1.1129,
       "step": 8042
     },
     {
-      "epoch": 0.22089478454313258,
+      "epoch": 0.22823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.8149151019870138e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.800914116087191e-05,
+      "loss": 0.9255,
       "step": 8043
     },
     {
-      "epoch": 0.2209222487709758,
+      "epoch": 0.2282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.81486354416335e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.800859080217827e-05,
+      "loss": 0.987,
       "step": 8044
     },
     {
-      "epoch": 0.22094971299881905,
+      "epoch": 0.2282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.8148119798921698e-05,
-      "loss": 1.1006,
+      "learning_rate": 1.8008040375835692e-05,
+      "loss": 0.9733,
       "step": 8045
     },
     {
-      "epoch": 0.22097717722666227,
+      "epoch": 0.22832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.814760409173882e-05,
-      "loss": 1.006,
+      "learning_rate": 1.800748988184884e-05,
+      "loss": 0.999,
       "step": 8046
     },
     {
-      "epoch": 0.22100464145450552,
+      "epoch": 0.22834846765039726,
       "grad_norm": 0.0,
-      "learning_rate": 1.8147088320088938e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.8006939320222363e-05,
+      "loss": 0.9547,
       "step": 8047
     },
     {
-      "epoch": 0.22103210568234874,
+      "epoch": 0.22837684449489218,
       "grad_norm": 0.0,
-      "learning_rate": 1.814657248397614e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.800638869096091e-05,
+      "loss": 1.0816,
       "step": 8048
     },
     {
-      "epoch": 0.221059569910192,
+      "epoch": 0.22840522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 1.8146056583404505e-05,
-      "loss": 1.0659,
+      "learning_rate": 1.800583799406913e-05,
+      "loss": 1.0304,
       "step": 8049
     },
     {
-      "epoch": 0.2210870341380352,
+      "epoch": 0.22843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.8145540618378118e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.8005287229551675e-05,
+      "loss": 0.9132,
       "step": 8050
     },
     {
-      "epoch": 0.22111449836587843,
+      "epoch": 0.22846197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.814502458890105e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.8004736397413204e-05,
+      "loss": 0.9043,
       "step": 8051
     },
     {
-      "epoch": 0.22114196259372168,
+      "epoch": 0.22849035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 1.8144508494977398e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.800418549765836e-05,
+      "loss": 0.9254,
       "step": 8052
     },
     {
-      "epoch": 0.2211694268215649,
+      "epoch": 0.22851872871736664,
       "grad_norm": 0.0,
-      "learning_rate": 1.814399233661124e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.80036345302918e-05,
+      "loss": 1.0521,
       "step": 8053
     },
     {
-      "epoch": 0.22119689104940815,
+      "epoch": 0.22854710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 1.814347611380666e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.8003083495318187e-05,
+      "loss": 1.0177,
       "step": 8054
     },
     {
-      "epoch": 0.22122435527725137,
+      "epoch": 0.22857548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.814295982656774e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.8002532392742164e-05,
+      "loss": 0.9624,
       "step": 8055
     },
     {
-      "epoch": 0.22125181950509462,
+      "epoch": 0.2286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.814244347489857e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.8001981222568386e-05,
+      "loss": 0.9329,
       "step": 8056
     },
     {
-      "epoch": 0.22127928373293784,
+      "epoch": 0.2286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.8141927058803233e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.8001429984801517e-05,
+      "loss": 0.9516,
       "step": 8057
     },
     {
-      "epoch": 0.2213067479607811,
+      "epoch": 0.22866061293984108,
       "grad_norm": 0.0,
-      "learning_rate": 1.8141410578285815e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.8000878679446207e-05,
+      "loss": 1.0449,
       "step": 8058
     },
     {
-      "epoch": 0.2213342121886243,
+      "epoch": 0.228688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.8140894033350402e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.800032730650712e-05,
+      "loss": 1.0089,
       "step": 8059
     },
     {
-      "epoch": 0.22136167641646756,
+      "epoch": 0.22871736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 1.8140377424001082e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.7999775865988903e-05,
+      "loss": 0.9129,
       "step": 8060
     },
     {
-      "epoch": 0.22138914064431078,
+      "epoch": 0.22874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.8139860750241943e-05,
-      "loss": 0.8948,
+      "learning_rate": 1.7999224357896222e-05,
+      "loss": 0.9648,
       "step": 8061
     },
     {
-      "epoch": 0.22141660487215403,
+      "epoch": 0.22877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.813934401207707e-05,
-      "loss": 0.8775,
+      "learning_rate": 1.7998672782233732e-05,
+      "loss": 1.1185,
       "step": 8062
     },
     {
-      "epoch": 0.22144406909999725,
+      "epoch": 0.22880249716231554,
       "grad_norm": 0.0,
-      "learning_rate": 1.8138827209510557e-05,
-      "loss": 0.988,
+      "learning_rate": 1.7998121139006096e-05,
+      "loss": 0.9142,
       "step": 8063
     },
     {
-      "epoch": 0.22147153332784048,
+      "epoch": 0.22883087400681043,
       "grad_norm": 0.0,
-      "learning_rate": 1.813831034254649e-05,
-      "loss": 0.999,
+      "learning_rate": 1.799756942821797e-05,
+      "loss": 1.0509,
       "step": 8064
     },
     {
-      "epoch": 0.22149899755568372,
+      "epoch": 0.22885925085130535,
       "grad_norm": 0.0,
-      "learning_rate": 1.813779341118896e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.7997017649874015e-05,
+      "loss": 1.0072,
       "step": 8065
     },
     {
-      "epoch": 0.22152646178352695,
+      "epoch": 0.22888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.8137276415442055e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.7996465803978893e-05,
+      "loss": 0.9755,
       "step": 8066
     },
     {
-      "epoch": 0.2215539260113702,
+      "epoch": 0.22891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.8136759355309867e-05,
-      "loss": 0.9155,
+      "learning_rate": 1.7995913890537268e-05,
+      "loss": 1.0437,
       "step": 8067
     },
     {
-      "epoch": 0.22158139023921342,
+      "epoch": 0.22894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.813624223079649e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.799536190955379e-05,
+      "loss": 0.9085,
       "step": 8068
     },
     {
-      "epoch": 0.22160885446705667,
+      "epoch": 0.2289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.8135725041906008e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.799480986103314e-05,
+      "loss": 1.0414,
       "step": 8069
     },
     {
-      "epoch": 0.2216363186948999,
+      "epoch": 0.22900113507377978,
       "grad_norm": 0.0,
-      "learning_rate": 1.8135207788642522e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.7994257744979965e-05,
+      "loss": 1.0619,
       "step": 8070
     },
     {
-      "epoch": 0.22166378292274314,
+      "epoch": 0.2290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.8134690471010116e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.799370556139894e-05,
+      "loss": 1.0325,
       "step": 8071
     },
     {
-      "epoch": 0.22169124715058636,
+      "epoch": 0.22905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.8134173089012893e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.7993153310294722e-05,
+      "loss": 0.9071,
       "step": 8072
     },
     {
-      "epoch": 0.2217187113784296,
+      "epoch": 0.22908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.813365564265494e-05,
-      "loss": 0.971,
+      "learning_rate": 1.799260099167198e-05,
+      "loss": 1.0049,
       "step": 8073
     },
     {
-      "epoch": 0.22174617560627283,
+      "epoch": 0.22911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.8133138131940352e-05,
-      "loss": 1.1163,
+      "learning_rate": 1.7992048605535378e-05,
+      "loss": 1.0597,
       "step": 8074
     },
     {
-      "epoch": 0.22177363983411608,
+      "epoch": 0.22914301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.8132620556873223e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.7991496151889583e-05,
+      "loss": 1.092,
       "step": 8075
     },
     {
-      "epoch": 0.2218011040619593,
+      "epoch": 0.22917139614074916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8132102917457656e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.799094363073926e-05,
+      "loss": 1.0349,
       "step": 8076
     },
     {
-      "epoch": 0.22182856828980252,
+      "epoch": 0.22919977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 1.8131585213697732e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.7990391042089076e-05,
+      "loss": 0.9137,
       "step": 8077
     },
     {
-      "epoch": 0.22185603251764577,
+      "epoch": 0.22922814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.8131067445597562e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.7989838385943697e-05,
+      "loss": 0.9916,
       "step": 8078
     },
     {
-      "epoch": 0.221883496745489,
+      "epoch": 0.22925652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.8130549613161238e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.79892856623078e-05,
+      "loss": 0.8367,
       "step": 8079
     },
     {
-      "epoch": 0.22191096097333224,
+      "epoch": 0.2292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.8130031716392853e-05,
-      "loss": 0.922,
+      "learning_rate": 1.7988732871186045e-05,
+      "loss": 0.9803,
       "step": 8080
     },
     {
-      "epoch": 0.22193842520117546,
+      "epoch": 0.2293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.8129513755296506e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.7988180012583105e-05,
+      "loss": 0.9939,
       "step": 8081
     },
     {
-      "epoch": 0.2219658894290187,
+      "epoch": 0.2293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.81289957298763e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.798762708650365e-05,
+      "loss": 0.9602,
       "step": 8082
     },
     {
-      "epoch": 0.22199335365686193,
+      "epoch": 0.2293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.8128477640136332e-05,
-      "loss": 1.084,
+      "learning_rate": 1.7987074092952347e-05,
+      "loss": 0.9822,
       "step": 8083
     },
     {
-      "epoch": 0.22202081788470518,
+      "epoch": 0.2293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.8127959486080697e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.7986521031933876e-05,
+      "loss": 0.9798,
       "step": 8084
     },
     {
-      "epoch": 0.2220482821125484,
+      "epoch": 0.22942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.81274412677135e-05,
-      "loss": 0.946,
+      "learning_rate": 1.7985967903452897e-05,
+      "loss": 1.0492,
       "step": 8085
     },
     {
-      "epoch": 0.22207574634039165,
+      "epoch": 0.22945516458569806,
       "grad_norm": 0.0,
-      "learning_rate": 1.8126922985038837e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.798541470751409e-05,
+      "loss": 0.9487,
       "step": 8086
     },
     {
-      "epoch": 0.22210321056823487,
+      "epoch": 0.22948354143019295,
       "grad_norm": 0.0,
-      "learning_rate": 1.8126404638060813e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.7984861444122127e-05,
+      "loss": 0.9566,
       "step": 8087
     },
     {
-      "epoch": 0.22213067479607812,
+      "epoch": 0.22951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.8125886226783528e-05,
-      "loss": 1.1065,
+      "learning_rate": 1.7984308113281677e-05,
+      "loss": 1.0779,
       "step": 8088
     },
     {
-      "epoch": 0.22215813902392134,
+      "epoch": 0.22954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.8125367751211082e-05,
-      "loss": 0.8257,
+      "learning_rate": 1.798375471499742e-05,
+      "loss": 1.0995,
       "step": 8089
     },
     {
-      "epoch": 0.22218560325176456,
+      "epoch": 0.22956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.812484921134758e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.7983201249274028e-05,
+      "loss": 1.0651,
       "step": 8090
     },
     {
-      "epoch": 0.2222130674796078,
+      "epoch": 0.22959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.8124330607197124e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.7982647716116173e-05,
+      "loss": 0.9857,
       "step": 8091
     },
     {
-      "epoch": 0.22224053170745103,
+      "epoch": 0.22962542565266741,
       "grad_norm": 0.0,
-      "learning_rate": 1.8123811938763814e-05,
-      "loss": 1.0535,
+      "learning_rate": 1.798209411552854e-05,
+      "loss": 1.0072,
       "step": 8092
     },
     {
-      "epoch": 0.22226799593529428,
+      "epoch": 0.22965380249716233,
       "grad_norm": 0.0,
-      "learning_rate": 1.8123293206051763e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.7981540447515797e-05,
+      "loss": 1.1223,
       "step": 8093
     },
     {
-      "epoch": 0.2222954601631375,
+      "epoch": 0.22968217934165722,
       "grad_norm": 0.0,
-      "learning_rate": 1.8122774409065066e-05,
-      "loss": 0.997,
+      "learning_rate": 1.798098671208262e-05,
+      "loss": 0.8677,
       "step": 8094
     },
     {
-      "epoch": 0.22232292439098075,
+      "epoch": 0.2297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.8122255547807833e-05,
-      "loss": 1.0071,
+      "learning_rate": 1.7980432909233693e-05,
+      "loss": 0.8889,
       "step": 8095
     },
     {
-      "epoch": 0.22235038861882397,
+      "epoch": 0.229738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8121736622284167e-05,
-      "loss": 1.0872,
+      "learning_rate": 1.7979879038973686e-05,
+      "loss": 1.0608,
       "step": 8096
     },
     {
-      "epoch": 0.22237785284666722,
+      "epoch": 0.22976730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.8121217632498172e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.797932510130729e-05,
+      "loss": 0.9563,
       "step": 8097
     },
     {
-      "epoch": 0.22240531707451044,
+      "epoch": 0.22979568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 1.812069857845396e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.797877109623917e-05,
+      "loss": 0.9713,
       "step": 8098
     },
     {
-      "epoch": 0.2224327813023537,
+      "epoch": 0.22982406356413168,
       "grad_norm": 0.0,
-      "learning_rate": 1.8120179460155638e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.7978217023774012e-05,
+      "loss": 1.0295,
       "step": 8099
     },
     {
-      "epoch": 0.22246024553019692,
+      "epoch": 0.22985244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 1.8119660277607306e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.7977662883916497e-05,
+      "loss": 0.9346,
       "step": 8100
     },
     {
-      "epoch": 0.22248770975804016,
+      "epoch": 0.22988081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.811914103081308e-05,
-      "loss": 1.0978,
+      "learning_rate": 1.7977108676671307e-05,
+      "loss": 1.0381,
       "step": 8101
     },
     {
-      "epoch": 0.22251517398588339,
+      "epoch": 0.22990919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.8118621719777062e-05,
-      "loss": 1.031,
+      "learning_rate": 1.7976554402043115e-05,
+      "loss": 0.98,
       "step": 8102
     },
     {
-      "epoch": 0.22254263821372663,
+      "epoch": 0.22993757094211123,
       "grad_norm": 0.0,
-      "learning_rate": 1.8118102344503364e-05,
-      "loss": 1.0164,
+      "learning_rate": 1.7976000060036617e-05,
+      "loss": 0.8614,
       "step": 8103
     },
     {
-      "epoch": 0.22257010244156986,
+      "epoch": 0.22996594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 1.81175829049961e-05,
-      "loss": 1.029,
+      "learning_rate": 1.7975445650656484e-05,
+      "loss": 0.9798,
       "step": 8104
     },
     {
-      "epoch": 0.22259756666941308,
+      "epoch": 0.22999432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.811706340125937e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.7974891173907406e-05,
+      "loss": 0.9175,
       "step": 8105
     },
     {
-      "epoch": 0.22262503089725633,
+      "epoch": 0.23002270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 1.8116543833297294e-05,
-      "loss": 0.9523,
+      "learning_rate": 1.7974336629794062e-05,
+      "loss": 0.9991,
       "step": 8106
     },
     {
-      "epoch": 0.22265249512509955,
+      "epoch": 0.2300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.8116024201113977e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.7973782018321134e-05,
+      "loss": 1.0163,
       "step": 8107
     },
     {
-      "epoch": 0.2226799593529428,
+      "epoch": 0.2300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.8115504504713534e-05,
-      "loss": 1.0966,
+      "learning_rate": 1.797322733949332e-05,
+      "loss": 0.9666,
       "step": 8108
     },
     {
-      "epoch": 0.22270742358078602,
+      "epoch": 0.23010783200908058,
       "grad_norm": 0.0,
-      "learning_rate": 1.811498474410008e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.7972672593315292e-05,
+      "loss": 0.9756,
       "step": 8109
     },
     {
-      "epoch": 0.22273488780862927,
+      "epoch": 0.23013620885357547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8114464919277714e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.7972117779791735e-05,
+      "loss": 1.0862,
       "step": 8110
     },
     {
-      "epoch": 0.2227623520364725,
+      "epoch": 0.23016458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 1.8113945030250565e-05,
-      "loss": 0.9831,
+      "learning_rate": 1.7971562898927348e-05,
+      "loss": 0.8679,
       "step": 8111
     },
     {
-      "epoch": 0.22278981626431574,
+      "epoch": 0.23019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.8113425077022736e-05,
-      "loss": 0.959,
+      "learning_rate": 1.7971007950726806e-05,
+      "loss": 0.9542,
       "step": 8112
     },
     {
-      "epoch": 0.22281728049215896,
+      "epoch": 0.23022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.8112905059598347e-05,
-      "loss": 0.9652,
+      "learning_rate": 1.79704529351948e-05,
+      "loss": 1.0025,
       "step": 8113
     },
     {
-      "epoch": 0.2228447447200022,
+      "epoch": 0.23024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.811238497798151e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.7969897852336028e-05,
+      "loss": 0.9942,
       "step": 8114
     },
     {
-      "epoch": 0.22287220894784543,
+      "epoch": 0.23027809307604993,
       "grad_norm": 0.0,
-      "learning_rate": 1.8111864832176342e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.7969342702155164e-05,
+      "loss": 1.0552,
       "step": 8115
     },
     {
-      "epoch": 0.22289967317568868,
+      "epoch": 0.23030646992054485,
       "grad_norm": 0.0,
-      "learning_rate": 1.8111344622186955e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.7968787484656905e-05,
+      "loss": 1.0689,
       "step": 8116
     },
     {
-      "epoch": 0.2229271374035319,
+      "epoch": 0.23033484676503974,
       "grad_norm": 0.0,
-      "learning_rate": 1.8110824348017468e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.796823219984594e-05,
+      "loss": 0.9917,
       "step": 8117
     },
     {
-      "epoch": 0.22295460163137512,
+      "epoch": 0.23036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.8110304009671997e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.796767684772696e-05,
+      "loss": 0.9622,
       "step": 8118
     },
     {
-      "epoch": 0.22298206585921837,
+      "epoch": 0.2303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.810978360715466e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.7967121428304656e-05,
+      "loss": 1.0369,
       "step": 8119
     },
     {
-      "epoch": 0.2230095300870616,
+      "epoch": 0.2304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.8109263140469574e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.7966565941583718e-05,
+      "loss": 0.9795,
       "step": 8120
     },
     {
-      "epoch": 0.22303699431490484,
+      "epoch": 0.23044835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 1.8108742609620854e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.796601038756884e-05,
+      "loss": 1.0292,
       "step": 8121
     },
     {
-      "epoch": 0.22306445854274806,
+      "epoch": 0.2304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.8108222014612623e-05,
-      "loss": 1.1967,
+      "learning_rate": 1.7965454766264714e-05,
+      "loss": 1.0277,
       "step": 8122
     },
     {
-      "epoch": 0.2230919227705913,
+      "epoch": 0.2305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.8107701355449e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.7964899077676034e-05,
+      "loss": 1.0336,
       "step": 8123
     },
     {
-      "epoch": 0.22311938699843453,
+      "epoch": 0.23053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.8107180632134102e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.7964343321807493e-05,
+      "loss": 0.8228,
       "step": 8124
     },
     {
-      "epoch": 0.22314685122627778,
+      "epoch": 0.23056186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.810665984467205e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.7963787498663783e-05,
+      "loss": 0.9844,
       "step": 8125
     },
     {
-      "epoch": 0.223174315454121,
+      "epoch": 0.23059023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.8106138993066967e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.796323160824961e-05,
+      "loss": 1.0378,
       "step": 8126
     },
     {
-      "epoch": 0.22320177968196425,
+      "epoch": 0.23061861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 1.8105618077322972e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.796267565056965e-05,
+      "loss": 1.0042,
       "step": 8127
     },
     {
-      "epoch": 0.22322924390980747,
+      "epoch": 0.23064699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 1.8105097097444184e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.796211962562862e-05,
+      "loss": 0.9745,
       "step": 8128
     },
     {
-      "epoch": 0.22325670813765072,
+      "epoch": 0.23067536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.810457605343473e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.7961563533431206e-05,
+      "loss": 1.0039,
       "step": 8129
     },
     {
-      "epoch": 0.22328417236549394,
+      "epoch": 0.23070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.810405494529873e-05,
-      "loss": 1.0354,
+      "learning_rate": 1.7961007373982108e-05,
+      "loss": 0.9578,
       "step": 8130
     },
     {
-      "epoch": 0.22331163659333716,
+      "epoch": 0.2307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.810353377304031e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.7960451147286017e-05,
+      "loss": 0.8898,
       "step": 8131
     },
     {
-      "epoch": 0.22333910082118041,
+      "epoch": 0.2307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.810301253666359e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.795989485334764e-05,
+      "loss": 1.0394,
       "step": 8132
     },
     {
-      "epoch": 0.22336656504902364,
+      "epoch": 0.23078887627695802,
       "grad_norm": 0.0,
-      "learning_rate": 1.8102491236172698e-05,
-      "loss": 1.047,
+      "learning_rate": 1.7959338492171675e-05,
+      "loss": 0.8997,
       "step": 8133
     },
     {
-      "epoch": 0.22339402927686688,
+      "epoch": 0.2308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.8101969871571755e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.795878206376282e-05,
+      "loss": 0.924,
       "step": 8134
     },
     {
-      "epoch": 0.2234214935047101,
+      "epoch": 0.2308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.810144844286489e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.795822556812578e-05,
+      "loss": 1.0362,
       "step": 8135
     },
     {
-      "epoch": 0.22344895773255335,
+      "epoch": 0.23087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.8100926950056222e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.7957669005265243e-05,
+      "loss": 1.0117,
       "step": 8136
     },
     {
-      "epoch": 0.22347642196039658,
+      "epoch": 0.23090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.8100405393149886e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.795711237518592e-05,
+      "loss": 0.928,
       "step": 8137
     },
     {
-      "epoch": 0.22350388618823983,
+      "epoch": 0.23093076049943245,
       "grad_norm": 0.0,
-      "learning_rate": 1.8099883772150003e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.795655567789251e-05,
+      "loss": 0.9921,
       "step": 8138
     },
     {
-      "epoch": 0.22353135041608305,
+      "epoch": 0.23095913734392737,
       "grad_norm": 0.0,
-      "learning_rate": 1.8099362087060704e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.795599891338972e-05,
+      "loss": 1.022,
       "step": 8139
     },
     {
-      "epoch": 0.2235588146439263,
+      "epoch": 0.23098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.8098840337886115e-05,
-      "loss": 1.0316,
+      "learning_rate": 1.7955442081682246e-05,
+      "loss": 0.9989,
       "step": 8140
     },
     {
-      "epoch": 0.22358627887176952,
+      "epoch": 0.23101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.8098318524630363e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.7954885182774798e-05,
+      "loss": 0.9827,
       "step": 8141
     },
     {
-      "epoch": 0.22361374309961277,
+      "epoch": 0.23104426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.8097796647297578e-05,
-      "loss": 0.9674,
+      "learning_rate": 1.7954328216672077e-05,
+      "loss": 0.9646,
       "step": 8142
     },
     {
-      "epoch": 0.223641207327456,
+      "epoch": 0.23107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.8097274705891887e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.7953771183378786e-05,
+      "loss": 0.9447,
       "step": 8143
     },
     {
-      "epoch": 0.22366867155529924,
+      "epoch": 0.2311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.8096752700417424e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.7953214082899634e-05,
+      "loss": 0.8394,
       "step": 8144
     },
     {
-      "epoch": 0.22369613578314246,
+      "epoch": 0.23112939841089672,
       "grad_norm": 0.0,
-      "learning_rate": 1.8096230630878315e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.7952656915239325e-05,
+      "loss": 0.8814,
       "step": 8145
     },
     {
-      "epoch": 0.22372360001098568,
+      "epoch": 0.2311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8095708497278693e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.7952099680402564e-05,
+      "loss": 1.0264,
       "step": 8146
     },
     {
-      "epoch": 0.22375106423882893,
+      "epoch": 0.2311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.8095186299622693e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.795154237839406e-05,
+      "loss": 0.9651,
       "step": 8147
     },
     {
-      "epoch": 0.22377852846667215,
+      "epoch": 0.23121452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.809466403791444e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.795098500921852e-05,
+      "loss": 0.8941,
       "step": 8148
     },
     {
-      "epoch": 0.2238059926945154,
+      "epoch": 0.23124290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 1.8094141712158074e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.7950427572880655e-05,
+      "loss": 1.0581,
       "step": 8149
     },
     {
-      "epoch": 0.22383345692235862,
+      "epoch": 0.23127128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 1.8093619322357722e-05,
-      "loss": 0.923,
+      "learning_rate": 1.794987006938517e-05,
+      "loss": 1.0046,
       "step": 8150
     },
     {
-      "epoch": 0.22386092115020187,
+      "epoch": 0.23129965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.8093096868517515e-05,
-      "loss": 0.956,
+      "learning_rate": 1.794931249873677e-05,
+      "loss": 1.2108,
       "step": 8151
     },
     {
-      "epoch": 0.2238883853780451,
+      "epoch": 0.23132803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.8092574350641593e-05,
-      "loss": 1.0056,
+      "learning_rate": 1.7948754860940175e-05,
+      "loss": 1.0579,
       "step": 8152
     },
     {
-      "epoch": 0.22391584960588834,
+      "epoch": 0.23135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.809205176873409e-05,
-      "loss": 1.0844,
+      "learning_rate": 1.7948197156000088e-05,
+      "loss": 0.9166,
       "step": 8153
     },
     {
-      "epoch": 0.22394331383373156,
+      "epoch": 0.23138479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 1.8091529122799137e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.7947639383921225e-05,
+      "loss": 1.0578,
       "step": 8154
     },
     {
-      "epoch": 0.2239707780615748,
+      "epoch": 0.23141316685584562,
       "grad_norm": 0.0,
-      "learning_rate": 1.8091006412840867e-05,
-      "loss": 1.0755,
+      "learning_rate": 1.7947081544708292e-05,
+      "loss": 0.9455,
       "step": 8155
     },
     {
-      "epoch": 0.22399824228941803,
+      "epoch": 0.23144154370034054,
       "grad_norm": 0.0,
-      "learning_rate": 1.809048363886343e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.7946523638366006e-05,
+      "loss": 0.9062,
       "step": 8156
     },
     {
-      "epoch": 0.22402570651726128,
+      "epoch": 0.23146992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 1.808996080087094e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.7945965664899076e-05,
+      "loss": 1.04,
       "step": 8157
     },
     {
-      "epoch": 0.2240531707451045,
+      "epoch": 0.2314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.8089437898867553e-05,
-      "loss": 1.005,
+      "learning_rate": 1.794540762431222e-05,
+      "loss": 1.0296,
       "step": 8158
     },
     {
-      "epoch": 0.22408063497294772,
+      "epoch": 0.2315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.8088914932857402e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.7944849516610147e-05,
+      "loss": 0.8615,
       "step": 8159
     },
     {
-      "epoch": 0.22410809920079097,
+      "epoch": 0.23155505107832008,
       "grad_norm": 0.0,
-      "learning_rate": 1.808839190284462e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.794429134179757e-05,
+      "loss": 1.1806,
       "step": 8160
     },
     {
-      "epoch": 0.2241355634286342,
+      "epoch": 0.23158342792281497,
       "grad_norm": 0.0,
-      "learning_rate": 1.808786880883335e-05,
-      "loss": 0.942,
+      "learning_rate": 1.794373309987921e-05,
+      "loss": 1.1083,
       "step": 8161
     },
     {
-      "epoch": 0.22416302765647744,
+      "epoch": 0.2316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.8087345650827725e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.794317479085978e-05,
+      "loss": 0.9639,
       "step": 8162
     },
     {
-      "epoch": 0.22419049188432066,
+      "epoch": 0.23164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.8086822428831895e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.794261641474399e-05,
+      "loss": 0.8312,
       "step": 8163
     },
     {
-      "epoch": 0.2242179561121639,
+      "epoch": 0.23166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.808629914284999e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.794205797153657e-05,
+      "loss": 0.9669,
       "step": 8164
     },
     {
-      "epoch": 0.22424542034000713,
+      "epoch": 0.23169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.808577579288615e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.7941499461242225e-05,
+      "loss": 0.9909,
       "step": 8165
     },
     {
-      "epoch": 0.22427288456785038,
+      "epoch": 0.23172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.8085252378944522e-05,
-      "loss": 1.0771,
+      "learning_rate": 1.794094088386568e-05,
+      "loss": 0.9225,
       "step": 8166
     },
     {
-      "epoch": 0.2243003487956936,
+      "epoch": 0.23175368898978432,
       "grad_norm": 0.0,
-      "learning_rate": 1.8084728901029248e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.7940382239411646e-05,
+      "loss": 1.0309,
       "step": 8167
     },
     {
-      "epoch": 0.22432781302353685,
+      "epoch": 0.23178206583427924,
       "grad_norm": 0.0,
-      "learning_rate": 1.8084205359144466e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.7939823527884845e-05,
+      "loss": 1.0427,
       "step": 8168
     },
     {
-      "epoch": 0.22435527725138008,
+      "epoch": 0.23181044267877413,
       "grad_norm": 0.0,
-      "learning_rate": 1.8083681753294317e-05,
-      "loss": 1.0407,
+      "learning_rate": 1.7939264749290005e-05,
+      "loss": 0.9369,
       "step": 8169
     },
     {
-      "epoch": 0.22438274147922332,
+      "epoch": 0.231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.808315808348295e-05,
-      "loss": 1.0191,
+      "learning_rate": 1.7938705903631833e-05,
+      "loss": 0.9911,
       "step": 8170
     },
     {
-      "epoch": 0.22441020570706655,
+      "epoch": 0.2318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.80826343497145e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.7938146990915055e-05,
+      "loss": 0.9606,
       "step": 8171
     },
     {
-      "epoch": 0.22443766993490977,
+      "epoch": 0.2318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.8082110551993125e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.79375880111444e-05,
+      "loss": 1.0794,
       "step": 8172
     },
     {
-      "epoch": 0.22446513416275302,
+      "epoch": 0.2319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.8081586690322953e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.7937028964324575e-05,
+      "loss": 1.0498,
       "step": 8173
     },
     {
-      "epoch": 0.22449259839059624,
+      "epoch": 0.2319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.8081062764708143e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.793646985046031e-05,
+      "loss": 0.9568,
       "step": 8174
     },
     {
-      "epoch": 0.2245200626184395,
+      "epoch": 0.23198070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 1.8080538775152828e-05,
-      "loss": 0.8902,
+      "learning_rate": 1.793591066955633e-05,
+      "loss": 1.0378,
       "step": 8175
     },
     {
-      "epoch": 0.2245475268462827,
+      "epoch": 0.23200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.8080014721661163e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.793535142161735e-05,
+      "loss": 0.9846,
       "step": 8176
     },
     {
-      "epoch": 0.22457499107412596,
+      "epoch": 0.23203745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.807949060423729e-05,
-      "loss": 0.9766,
+      "learning_rate": 1.793479210664811e-05,
+      "loss": 0.8977,
       "step": 8177
     },
     {
-      "epoch": 0.22460245530196918,
+      "epoch": 0.23206583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 1.807896642288536e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.7934232724653318e-05,
+      "loss": 0.9235,
       "step": 8178
     },
     {
-      "epoch": 0.22462991952981243,
+      "epoch": 0.23209421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 1.8078442177609515e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.7933673275637704e-05,
+      "loss": 0.9688,
       "step": 8179
     },
     {
-      "epoch": 0.22465738375765565,
+      "epoch": 0.23212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.807791786841391e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.7933113759606e-05,
+      "loss": 0.8817,
       "step": 8180
     },
     {
-      "epoch": 0.2246848479854989,
+      "epoch": 0.23215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.8077393495302687e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.7932554176562923e-05,
+      "loss": 0.9848,
       "step": 8181
     },
     {
-      "epoch": 0.22471231221334212,
+      "epoch": 0.23217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.807686905828e-05,
-      "loss": 1.1123,
+      "learning_rate": 1.7931994526513207e-05,
+      "loss": 0.9849,
       "step": 8182
     },
     {
-      "epoch": 0.22473977644118537,
+      "epoch": 0.2322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.8076344557349995e-05,
-      "loss": 0.8495,
+      "learning_rate": 1.7931434809461577e-05,
+      "loss": 1.0518,
       "step": 8183
     },
     {
-      "epoch": 0.2247672406690286,
+      "epoch": 0.2322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.8075819992516822e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.793087502541276e-05,
+      "loss": 0.8728,
       "step": 8184
     },
     {
-      "epoch": 0.2247947048968718,
+      "epoch": 0.2322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.8075295363784634e-05,
-      "loss": 1.106,
+      "learning_rate": 1.7930315174371484e-05,
+      "loss": 0.9105,
       "step": 8185
     },
     {
-      "epoch": 0.22482216912471506,
+      "epoch": 0.2322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.8074770671157584e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.792975525634248e-05,
+      "loss": 0.995,
       "step": 8186
     },
     {
-      "epoch": 0.22484963335255828,
+      "epoch": 0.23232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.8074245914639817e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.7929195271330475e-05,
+      "loss": 0.904,
       "step": 8187
     },
     {
-      "epoch": 0.22487709758040153,
+      "epoch": 0.23234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.807372109423549e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.7928635219340204e-05,
+      "loss": 0.9222,
       "step": 8188
     },
     {
-      "epoch": 0.22490456180824475,
+      "epoch": 0.23237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.8073196209948755e-05,
-      "loss": 0.9557,
+      "learning_rate": 1.7928075100376396e-05,
+      "loss": 1.0976,
       "step": 8189
     },
     {
-      "epoch": 0.224932026036088,
+      "epoch": 0.23240635641316684,
       "grad_norm": 0.0,
-      "learning_rate": 1.8072671261783763e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.792751491444378e-05,
+      "loss": 0.9435,
       "step": 8190
     },
     {
-      "epoch": 0.22495949026393122,
+      "epoch": 0.23243473325766176,
       "grad_norm": 0.0,
-      "learning_rate": 1.8072146249744668e-05,
-      "loss": 0.9048,
+      "learning_rate": 1.7926954661547084e-05,
+      "loss": 0.9745,
       "step": 8191
     },
     {
-      "epoch": 0.22498695449177447,
+      "epoch": 0.23246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.8071621173835625e-05,
-      "loss": 1.1149,
+      "learning_rate": 1.792639434169105e-05,
+      "loss": 1.0427,
       "step": 8192
     },
     {
-      "epoch": 0.2250144187196177,
+      "epoch": 0.23249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.807109603406079e-05,
-      "loss": 1.0141,
+      "learning_rate": 1.7925833954880408e-05,
+      "loss": 1.0587,
       "step": 8193
     },
     {
-      "epoch": 0.22504188294746094,
+      "epoch": 0.23251986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.8070570830424316e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.792527350111989e-05,
+      "loss": 0.955,
       "step": 8194
     },
     {
-      "epoch": 0.22506934717530416,
+      "epoch": 0.2325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.807004556293036e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.7924712980414227e-05,
+      "loss": 1.0511,
       "step": 8195
     },
     {
-      "epoch": 0.2250968114031474,
+      "epoch": 0.23257661748013622,
       "grad_norm": 0.0,
-      "learning_rate": 1.806952023158308e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.792415239276816e-05,
+      "loss": 1.0132,
       "step": 8196
     },
     {
-      "epoch": 0.22512427563099063,
+      "epoch": 0.2326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.8068994836386627e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.7923591738186423e-05,
+      "loss": 0.9878,
       "step": 8197
     },
     {
-      "epoch": 0.22515173985883388,
+      "epoch": 0.232633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.8068469377345162e-05,
-      "loss": 0.9968,
+      "learning_rate": 1.7923031016673748e-05,
+      "loss": 0.9768,
       "step": 8198
     },
     {
-      "epoch": 0.2251792040866771,
+      "epoch": 0.23266174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 1.806794385446284e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.7922470228234875e-05,
+      "loss": 0.9834,
       "step": 8199
     },
     {
-      "epoch": 0.22520666831452033,
+      "epoch": 0.23269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.8067418267743823e-05,
-      "loss": 0.8938,
+      "learning_rate": 1.792190937287454e-05,
+      "loss": 0.9578,
       "step": 8200
     },
     {
-      "epoch": 0.22523413254236357,
+      "epoch": 0.23271850170261066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8066892617192264e-05,
-      "loss": 1.0074,
+      "learning_rate": 1.7921348450597478e-05,
+      "loss": 1.0502,
       "step": 8201
     },
     {
-      "epoch": 0.2252615967702068,
+      "epoch": 0.23274687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.806636690281233e-05,
-      "loss": 1.0082,
+      "learning_rate": 1.7920787461408432e-05,
+      "loss": 0.9534,
       "step": 8202
     },
     {
-      "epoch": 0.22528906099805004,
+      "epoch": 0.23277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.8065841124608177e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.7920226405312136e-05,
+      "loss": 0.925,
       "step": 8203
     },
     {
-      "epoch": 0.22531652522589327,
+      "epoch": 0.23280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.8065315282583963e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.7919665282313335e-05,
+      "loss": 1.0602,
       "step": 8204
     },
     {
-      "epoch": 0.22534398945373652,
+      "epoch": 0.23283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.806478937674385e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.791910409241676e-05,
+      "loss": 1.0189,
       "step": 8205
     },
     {
-      "epoch": 0.22537145368157974,
+      "epoch": 0.23286038592508512,
       "grad_norm": 0.0,
-      "learning_rate": 1.8064263407092e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.791854283562716e-05,
+      "loss": 0.9495,
       "step": 8206
     },
     {
-      "epoch": 0.22539891790942299,
+      "epoch": 0.23288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.806373737363257e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.7917981511949273e-05,
+      "loss": 0.99,
       "step": 8207
     },
     {
-      "epoch": 0.2254263821372662,
+      "epoch": 0.23291713961407493,
       "grad_norm": 0.0,
-      "learning_rate": 1.806321127636973e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.791742012138784e-05,
+      "loss": 1.0888,
       "step": 8208
     },
     {
-      "epoch": 0.22545384636510946,
+      "epoch": 0.2329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.8062685115307643e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.7916858663947602e-05,
+      "loss": 0.9669,
       "step": 8209
     },
     {
-      "epoch": 0.22548131059295268,
+      "epoch": 0.2329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.8062158890450467e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.7916297139633305e-05,
+      "loss": 1.0545,
       "step": 8210
     },
     {
-      "epoch": 0.22550877482079593,
+      "epoch": 0.2330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.806163260180236e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.7915735548449687e-05,
+      "loss": 1.0537,
       "step": 8211
     },
     {
-      "epoch": 0.22553623904863915,
+      "epoch": 0.23303064699205447,
       "grad_norm": 0.0,
-      "learning_rate": 1.8061106249367498e-05,
-      "loss": 0.9611,
+      "learning_rate": 1.7915173890401502e-05,
+      "loss": 1.0143,
       "step": 8212
     },
     {
-      "epoch": 0.22556370327648237,
+      "epoch": 0.2330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.806057983315004e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.7914612165493483e-05,
+      "loss": 0.9899,
       "step": 8213
     },
     {
-      "epoch": 0.22559116750432562,
+      "epoch": 0.23308740068104428,
       "grad_norm": 0.0,
-      "learning_rate": 1.8060053353154153e-05,
-      "loss": 1.1001,
+      "learning_rate": 1.791405037373038e-05,
+      "loss": 1.0369,
       "step": 8214
     },
     {
-      "epoch": 0.22561863173216884,
+      "epoch": 0.23311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8059526809384e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.7913488515116935e-05,
+      "loss": 1.0925,
       "step": 8215
     },
     {
-      "epoch": 0.2256460959600121,
+      "epoch": 0.23314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.805900020184375e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.7912926589657897e-05,
+      "loss": 0.9898,
       "step": 8216
     },
     {
-      "epoch": 0.2256735601878553,
+      "epoch": 0.23317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.8058473530537568e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.7912364597358015e-05,
+      "loss": 0.955,
       "step": 8217
     },
     {
-      "epoch": 0.22570102441569856,
+      "epoch": 0.23320090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 1.8057946795469625e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.791180253822203e-05,
+      "loss": 1.0034,
       "step": 8218
     },
     {
-      "epoch": 0.22572848864354178,
+      "epoch": 0.23322928490351874,
       "grad_norm": 0.0,
-      "learning_rate": 1.805741999664408e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.7911240412254698e-05,
+      "loss": 0.987,
       "step": 8219
     },
     {
-      "epoch": 0.22575595287138503,
+      "epoch": 0.23325766174801363,
       "grad_norm": 0.0,
-      "learning_rate": 1.805689313406511e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.791067821946076e-05,
+      "loss": 0.8772,
       "step": 8220
     },
     {
-      "epoch": 0.22578341709922825,
+      "epoch": 0.23328603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.805636620773688e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.7910115959844967e-05,
+      "loss": 0.8957,
       "step": 8221
     },
     {
-      "epoch": 0.2258108813270715,
+      "epoch": 0.2333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.805583921766356e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.7909553633412068e-05,
+      "loss": 0.991,
       "step": 8222
     },
     {
-      "epoch": 0.22583834555491472,
+      "epoch": 0.2333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.805531216384932e-05,
-      "loss": 0.9529,
+      "learning_rate": 1.7908991240166817e-05,
+      "loss": 0.8822,
       "step": 8223
     },
     {
-      "epoch": 0.22586580978275797,
+      "epoch": 0.23337116912599318,
       "grad_norm": 0.0,
-      "learning_rate": 1.805478504629833e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.790842878011396e-05,
+      "loss": 0.9078,
       "step": 8224
     },
     {
-      "epoch": 0.2258932740106012,
+      "epoch": 0.2333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.805425786501476e-05,
-      "loss": 0.9786,
+      "learning_rate": 1.790786625325825e-05,
+      "loss": 1.0603,
       "step": 8225
     },
     {
-      "epoch": 0.2259207382384444,
+      "epoch": 0.23342792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 1.8053730620002783e-05,
-      "loss": 0.9953,
+      "learning_rate": 1.790730365960444e-05,
+      "loss": 0.888,
       "step": 8226
     },
     {
-      "epoch": 0.22594820246628766,
+      "epoch": 0.23345629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.8053203311266566e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.7906740999157275e-05,
+      "loss": 1.101,
       "step": 8227
     },
     {
-      "epoch": 0.22597566669413088,
+      "epoch": 0.23348467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.8052675938810288e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.7906178271921518e-05,
+      "loss": 0.9059,
       "step": 8228
     },
     {
-      "epoch": 0.22600313092197413,
+      "epoch": 0.23351305334846764,
       "grad_norm": 0.0,
-      "learning_rate": 1.8052148502638117e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.7905615477901916e-05,
+      "loss": 0.9586,
       "step": 8229
     },
     {
-      "epoch": 0.22603059514981735,
+      "epoch": 0.23354143019296253,
       "grad_norm": 0.0,
-      "learning_rate": 1.805162100275423e-05,
-      "loss": 0.9395,
+      "learning_rate": 1.7905052617103226e-05,
+      "loss": 0.9746,
       "step": 8230
     },
     {
-      "epoch": 0.2260580593776606,
+      "epoch": 0.23356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.80510934391628e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.79044896895302e-05,
+      "loss": 0.9798,
       "step": 8231
     },
     {
-      "epoch": 0.22608552360550382,
+      "epoch": 0.23359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.8050565811867996e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.7903926695187595e-05,
+      "loss": 0.9923,
       "step": 8232
     },
     {
-      "epoch": 0.22611298783334707,
+      "epoch": 0.23362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.8050038120874e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.7903363634080168e-05,
+      "loss": 1.021,
       "step": 8233
     },
     {
-      "epoch": 0.2261404520611903,
+      "epoch": 0.2336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.8049510366184982e-05,
-      "loss": 1.0944,
+      "learning_rate": 1.7902800506212667e-05,
+      "loss": 0.8841,
       "step": 8234
     },
     {
-      "epoch": 0.22616791628903354,
+      "epoch": 0.233683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.804898254780512e-05,
-      "loss": 1.0644,
+      "learning_rate": 1.790223731158986e-05,
+      "loss": 0.9562,
       "step": 8235
     },
     {
-      "epoch": 0.22619538051687677,
+      "epoch": 0.2337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.8048454665738592e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.7901674050216498e-05,
+      "loss": 0.9772,
       "step": 8236
     },
     {
-      "epoch": 0.22622284474472001,
+      "epoch": 0.2337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.804792671998957e-05,
-      "loss": 0.993,
+      "learning_rate": 1.790111072209734e-05,
+      "loss": 1.0275,
       "step": 8237
     },
     {
-      "epoch": 0.22625030897256324,
+      "epoch": 0.23376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.8047398710562235e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.7900547327237143e-05,
+      "loss": 1.0379,
       "step": 8238
     },
     {
-      "epoch": 0.22627777320040646,
+      "epoch": 0.23379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.8046870637460767e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.789998386564067e-05,
+      "loss": 0.9332,
       "step": 8239
     },
     {
-      "epoch": 0.2263052374282497,
+      "epoch": 0.23382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.804634250068934e-05,
-      "loss": 1.0354,
+      "learning_rate": 1.7899420337312675e-05,
+      "loss": 1.0139,
       "step": 8240
     },
     {
-      "epoch": 0.22633270165609293,
+      "epoch": 0.23385357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.804581430025213e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.7898856742257925e-05,
+      "loss": 0.9749,
       "step": 8241
     },
     {
-      "epoch": 0.22636016588393618,
+      "epoch": 0.23388195232690126,
       "grad_norm": 0.0,
-      "learning_rate": 1.8045286036153328e-05,
-      "loss": 1.024,
+      "learning_rate": 1.7898293080481176e-05,
+      "loss": 0.9821,
       "step": 8242
     },
     {
-      "epoch": 0.2263876301117794,
+      "epoch": 0.23391032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.8044757708397103e-05,
-      "loss": 1.1326,
+      "learning_rate": 1.7897729351987186e-05,
+      "loss": 1.1137,
       "step": 8243
     },
     {
-      "epoch": 0.22641509433962265,
+      "epoch": 0.23393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.804422931698764e-05,
-      "loss": 0.8382,
+      "learning_rate": 1.7897165556780726e-05,
+      "loss": 1.0421,
       "step": 8244
     },
     {
-      "epoch": 0.22644255856746587,
+      "epoch": 0.23396708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 1.8043700861929117e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.7896601694866555e-05,
+      "loss": 1.0096,
       "step": 8245
     },
     {
-      "epoch": 0.22647002279530912,
+      "epoch": 0.2339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.8043172343225716e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.789603776624943e-05,
+      "loss": 0.89,
       "step": 8246
     },
     {
-      "epoch": 0.22649748702315234,
+      "epoch": 0.2340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.804264376088162e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.789547377093412e-05,
+      "loss": 1.0445,
       "step": 8247
     },
     {
-      "epoch": 0.2265249512509956,
+      "epoch": 0.2340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8042115114901014e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.789490970892539e-05,
+      "loss": 0.8019,
       "step": 8248
     },
     {
-      "epoch": 0.2265524154788388,
+      "epoch": 0.2340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.8041586405288076e-05,
-      "loss": 0.9818,
+      "learning_rate": 1.7894345580228e-05,
+      "loss": 1.1509,
       "step": 8249
     },
     {
-      "epoch": 0.22657987970668206,
+      "epoch": 0.2341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.804105763204699e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.7893781384846722e-05,
+      "loss": 0.9966,
       "step": 8250
     },
     {
-      "epoch": 0.22660734393452528,
+      "epoch": 0.23413734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.8040528795181945e-05,
-      "loss": 0.9581,
+      "learning_rate": 1.7893217122786316e-05,
+      "loss": 0.9659,
       "step": 8251
     },
     {
-      "epoch": 0.22663480816236853,
+      "epoch": 0.23416572077185016,
       "grad_norm": 0.0,
-      "learning_rate": 1.8039999894697118e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.789265279405155e-05,
+      "loss": 1.0173,
       "step": 8252
     },
     {
-      "epoch": 0.22666227239021175,
+      "epoch": 0.23419409761634508,
       "grad_norm": 0.0,
-      "learning_rate": 1.80394709305967e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.789208839864719e-05,
+      "loss": 0.975,
       "step": 8253
     },
     {
-      "epoch": 0.22668973661805497,
+      "epoch": 0.23422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.8038941902884873e-05,
-      "loss": 1.0348,
+      "learning_rate": 1.7891523936578004e-05,
+      "loss": 0.9995,
       "step": 8254
     },
     {
-      "epoch": 0.22671720084589822,
+      "epoch": 0.23425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.803841281156582e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.7890959407848766e-05,
+      "loss": 0.9776,
       "step": 8255
     },
     {
-      "epoch": 0.22674466507374144,
+      "epoch": 0.23427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.8037883656643735e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.7890394812464233e-05,
+      "loss": 1.0209,
       "step": 8256
     },
     {
-      "epoch": 0.2267721293015847,
+      "epoch": 0.23430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.80373544381228e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.788983015042918e-05,
+      "loss": 0.9016,
       "step": 8257
     },
     {
-      "epoch": 0.2267995935294279,
+      "epoch": 0.2343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.8036825156007207e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.788926542174838e-05,
+      "loss": 0.8873,
       "step": 8258
     },
     {
-      "epoch": 0.22682705775727116,
+      "epoch": 0.23436435868331443,
       "grad_norm": 0.0,
-      "learning_rate": 1.8036295810301135e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.78887006264266e-05,
+      "loss": 1.0692,
       "step": 8259
     },
     {
-      "epoch": 0.22685452198511438,
+      "epoch": 0.23439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.803576640100878e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.7888135764468612e-05,
+      "loss": 1.0322,
       "step": 8260
     },
     {
-      "epoch": 0.22688198621295763,
+      "epoch": 0.2344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.8035236928134327e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.7887570835879183e-05,
+      "loss": 1.0194,
       "step": 8261
     },
     {
-      "epoch": 0.22690945044080085,
+      "epoch": 0.2344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.8034707391681968e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.788700584066309e-05,
+      "loss": 1.0526,
       "step": 8262
     },
     {
-      "epoch": 0.2269369146686441,
+      "epoch": 0.23447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.8034177791655893e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.788644077882511e-05,
+      "loss": 0.9475,
       "step": 8263
     },
     {
-      "epoch": 0.22696437889648732,
+      "epoch": 0.23450624290578886,
       "grad_norm": 0.0,
-      "learning_rate": 1.803364812806029e-05,
-      "loss": 1.0455,
+      "learning_rate": 1.7885875650370002e-05,
+      "loss": 0.9207,
       "step": 8264
     },
     {
-      "epoch": 0.22699184312433057,
+      "epoch": 0.23453461975028378,
       "grad_norm": 0.0,
-      "learning_rate": 1.803311840089935e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.7885310455302554e-05,
+      "loss": 1.0329,
       "step": 8265
     },
     {
-      "epoch": 0.2270193073521738,
+      "epoch": 0.23456299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.8032588610177268e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.7884745193627532e-05,
+      "loss": 1.0857,
       "step": 8266
     },
     {
-      "epoch": 0.22704677158001701,
+      "epoch": 0.23459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.8032058755898228e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.7884179865349713e-05,
+      "loss": 0.9591,
       "step": 8267
     },
     {
-      "epoch": 0.22707423580786026,
+      "epoch": 0.23461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.803152883806643e-05,
-      "loss": 1.0758,
+      "learning_rate": 1.788361447047387e-05,
+      "loss": 0.9046,
       "step": 8268
     },
     {
-      "epoch": 0.22710170003570349,
+      "epoch": 0.23464812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.803099885668607e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.7883049009004782e-05,
+      "loss": 1.0711,
       "step": 8269
     },
     {
-      "epoch": 0.22712916426354673,
+      "epoch": 0.23467650397275822,
       "grad_norm": 0.0,
-      "learning_rate": 1.8030468811761332e-05,
-      "loss": 1.1194,
+      "learning_rate": 1.7882483480947226e-05,
+      "loss": 0.8507,
       "step": 8270
     },
     {
-      "epoch": 0.22715662849138996,
+      "epoch": 0.23470488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 1.8029938703296412e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.7881917886305978e-05,
+      "loss": 1.0934,
       "step": 8271
     },
     {
-      "epoch": 0.2271840927192332,
+      "epoch": 0.23473325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 1.802940853129551e-05,
-      "loss": 0.924,
+      "learning_rate": 1.7881352225085812e-05,
+      "loss": 0.9877,
       "step": 8272
     },
     {
-      "epoch": 0.22721155694707643,
+      "epoch": 0.2347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.8028878295762817e-05,
-      "loss": 0.9115,
+      "learning_rate": 1.7880786497291514e-05,
+      "loss": 1.075,
       "step": 8273
     },
     {
-      "epoch": 0.22723902117491968,
+      "epoch": 0.2347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.8028347996702527e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.7880220702927855e-05,
+      "loss": 1.0016,
       "step": 8274
     },
     {
-      "epoch": 0.2272664854027629,
+      "epoch": 0.23481838819523268,
       "grad_norm": 0.0,
-      "learning_rate": 1.802781763411884e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.787965484199962e-05,
+      "loss": 0.9722,
       "step": 8275
     },
     {
-      "epoch": 0.22729394963060615,
+      "epoch": 0.2348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.8027287208015947e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.7879088914511583e-05,
+      "loss": 0.918,
       "step": 8276
     },
     {
-      "epoch": 0.22732141385844937,
+      "epoch": 0.23487514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 1.802675671839805e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.787852292046853e-05,
+      "loss": 0.9811,
       "step": 8277
     },
     {
-      "epoch": 0.22734887808629262,
+      "epoch": 0.23490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.8026226165269347e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.7877956859875238e-05,
+      "loss": 0.9258,
       "step": 8278
     },
     {
-      "epoch": 0.22737634231413584,
+      "epoch": 0.23493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.802569554863403e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.7877390732736492e-05,
+      "loss": 1.0096,
       "step": 8279
     },
     {
-      "epoch": 0.22740380654197906,
+      "epoch": 0.23496027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.8025164868496304e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.787682453905707e-05,
+      "loss": 0.9001,
       "step": 8280
     },
     {
-      "epoch": 0.2274312707698223,
+      "epoch": 0.23498864926220203,
       "grad_norm": 0.0,
-      "learning_rate": 1.802463412486036e-05,
-      "loss": 1.0807,
+      "learning_rate": 1.7876258278841764e-05,
+      "loss": 0.8944,
       "step": 8281
     },
     {
-      "epoch": 0.22745873499766553,
+      "epoch": 0.23501702610669695,
       "grad_norm": 0.0,
-      "learning_rate": 1.8024103317730408e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.7875691952095344e-05,
+      "loss": 0.8878,
       "step": 8282
     },
     {
-      "epoch": 0.22748619922550878,
+      "epoch": 0.23504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.802357244711064e-05,
-      "loss": 1.069,
+      "learning_rate": 1.78751255588226e-05,
+      "loss": 1.0652,
       "step": 8283
     },
     {
-      "epoch": 0.227513663453352,
+      "epoch": 0.23507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.8023041513005256e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.787455909902832e-05,
+      "loss": 0.8293,
       "step": 8284
     },
     {
-      "epoch": 0.22754112768119525,
+      "epoch": 0.2351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.8022510515418464e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.7873992572717282e-05,
+      "loss": 0.9232,
       "step": 8285
     },
     {
-      "epoch": 0.22756859190903847,
+      "epoch": 0.2351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.802197945435446e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.7873425979894276e-05,
+      "loss": 0.9506,
       "step": 8286
     },
     {
-      "epoch": 0.22759605613688172,
+      "epoch": 0.23515891032917138,
       "grad_norm": 0.0,
-      "learning_rate": 1.8021448329817443e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.7872859320564085e-05,
+      "loss": 0.93,
       "step": 8287
     },
     {
-      "epoch": 0.22762352036472494,
+      "epoch": 0.2351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.8020917141811627e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.78722925947315e-05,
+      "loss": 0.9409,
       "step": 8288
     },
     {
-      "epoch": 0.2276509845925682,
+      "epoch": 0.23521566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 1.8020385890341203e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.7871725802401303e-05,
+      "loss": 0.9148,
       "step": 8289
     },
     {
-      "epoch": 0.2276784488204114,
+      "epoch": 0.23524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.8019854575410377e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.787115894357829e-05,
+      "loss": 0.9531,
       "step": 8290
     },
     {
-      "epoch": 0.22770591304825466,
+      "epoch": 0.23527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.801932319702336e-05,
-      "loss": 1.0077,
+      "learning_rate": 1.7870592018267237e-05,
+      "loss": 0.8171,
       "step": 8291
     },
     {
-      "epoch": 0.22773337727609788,
+      "epoch": 0.23530079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.801879175518435e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.7870025026472942e-05,
+      "loss": 0.9694,
       "step": 8292
     },
     {
-      "epoch": 0.22776084150394113,
+      "epoch": 0.23532917139614076,
       "grad_norm": 0.0,
-      "learning_rate": 1.801826024989755e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.7869457968200192e-05,
+      "loss": 0.9008,
       "step": 8293
     },
     {
-      "epoch": 0.22778830573178435,
+      "epoch": 0.23535754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 1.801772868116717e-05,
-      "loss": 1.0184,
+      "learning_rate": 1.7868890843453776e-05,
+      "loss": 1.0605,
       "step": 8294
     },
     {
-      "epoch": 0.22781576995962757,
+      "epoch": 0.23538592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.801719704899742e-05,
-      "loss": 1.0282,
+      "learning_rate": 1.7868323652238486e-05,
+      "loss": 1.0262,
       "step": 8295
     },
     {
-      "epoch": 0.22784323418747082,
+      "epoch": 0.23541430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.80166653533925e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.786775639455911e-05,
+      "loss": 0.9299,
       "step": 8296
     },
     {
-      "epoch": 0.22787069841531404,
+      "epoch": 0.2354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.8016133594356614e-05,
-      "loss": 1.059,
+      "learning_rate": 1.7867189070420446e-05,
+      "loss": 1.021,
       "step": 8297
     },
     {
-      "epoch": 0.2278981626431573,
+      "epoch": 0.2354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.8015601771893976e-05,
-      "loss": 1.0948,
+      "learning_rate": 1.786662167982728e-05,
+      "loss": 0.9982,
       "step": 8298
     },
     {
-      "epoch": 0.22792562687100051,
+      "epoch": 0.23549943246311011,
       "grad_norm": 0.0,
-      "learning_rate": 1.8015069886008794e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.786605422278441e-05,
+      "loss": 0.9451,
       "step": 8299
     },
     {
-      "epoch": 0.22795309109884376,
+      "epoch": 0.235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.801453793670527e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.7865486699296624e-05,
+      "loss": 0.9269,
       "step": 8300
     },
     {
-      "epoch": 0.22798055532668698,
+      "epoch": 0.2355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.801400592398762e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.7864919109368716e-05,
+      "loss": 0.9669,
       "step": 8301
     },
     {
-      "epoch": 0.22800801955453023,
+      "epoch": 0.23558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.801347384786005e-05,
-      "loss": 1.1072,
+      "learning_rate": 1.7864351453005486e-05,
+      "loss": 0.9821,
       "step": 8302
     },
     {
-      "epoch": 0.22803548378237345,
+      "epoch": 0.23561293984108966,
       "grad_norm": 0.0,
-      "learning_rate": 1.801294170832677e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.7863783730211722e-05,
+      "loss": 0.99,
       "step": 8303
     },
     {
-      "epoch": 0.2280629480102167,
+      "epoch": 0.23564131668558455,
       "grad_norm": 0.0,
-      "learning_rate": 1.8012409505391995e-05,
-      "loss": 0.9855,
+      "learning_rate": 1.7863215940992228e-05,
+      "loss": 0.9837,
       "step": 8304
     },
     {
-      "epoch": 0.22809041223805993,
+      "epoch": 0.23566969353007947,
       "grad_norm": 0.0,
-      "learning_rate": 1.8011877239059928e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.7862648085351795e-05,
+      "loss": 0.8922,
       "step": 8305
     },
     {
-      "epoch": 0.22811787646590317,
+      "epoch": 0.23569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.8011344909334787e-05,
-      "loss": 1.0766,
+      "learning_rate": 1.7862080163295218e-05,
+      "loss": 0.9986,
       "step": 8306
     },
     {
-      "epoch": 0.2281453406937464,
+      "epoch": 0.23572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.801081251622078e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.7861512174827297e-05,
+      "loss": 0.9045,
       "step": 8307
     },
     {
-      "epoch": 0.22817280492158962,
+      "epoch": 0.23575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.8010280059722123e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.786094411995283e-05,
+      "loss": 0.9594,
       "step": 8308
     },
     {
-      "epoch": 0.22820026914943287,
+      "epoch": 0.23578320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.800974753984303e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.7860375998676613e-05,
+      "loss": 0.8879,
       "step": 8309
     },
     {
-      "epoch": 0.2282277333772761,
+      "epoch": 0.2358115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.8009214956587706e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.7859807811003446e-05,
+      "loss": 0.9639,
       "step": 8310
     },
     {
-      "epoch": 0.22825519760511934,
+      "epoch": 0.23583995459704882,
       "grad_norm": 0.0,
-      "learning_rate": 1.8008682309960374e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.7859239556938132e-05,
+      "loss": 1.0112,
       "step": 8311
     },
     {
-      "epoch": 0.22828266183296256,
+      "epoch": 0.2358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8008149599965246e-05,
-      "loss": 1.0442,
+      "learning_rate": 1.7858671236485466e-05,
+      "loss": 1.071,
       "step": 8312
     },
     {
-      "epoch": 0.2283101260608058,
+      "epoch": 0.2358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.8007616826606535e-05,
-      "loss": 1.1183,
+      "learning_rate": 1.7858102849650252e-05,
+      "loss": 1.0746,
       "step": 8313
     },
     {
-      "epoch": 0.22833759028864903,
+      "epoch": 0.23592508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 1.8007083989888458e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.7857534396437287e-05,
+      "loss": 0.9335,
       "step": 8314
     },
     {
-      "epoch": 0.22836505451649228,
+      "epoch": 0.23595346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 1.800655108981523e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.7856965876851383e-05,
+      "loss": 0.9516,
       "step": 8315
     },
     {
-      "epoch": 0.2283925187443355,
+      "epoch": 0.23598183881952328,
       "grad_norm": 0.0,
-      "learning_rate": 1.8006018126391067e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.785639729089733e-05,
+      "loss": 0.9535,
       "step": 8316
     },
     {
-      "epoch": 0.22841998297217875,
+      "epoch": 0.23601021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 1.800548509962019e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.7855828638579934e-05,
+      "loss": 0.9885,
       "step": 8317
     },
     {
-      "epoch": 0.22844744720002197,
+      "epoch": 0.23603859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.8004952009506813e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.7855259919904004e-05,
+      "loss": 0.9565,
       "step": 8318
     },
     {
-      "epoch": 0.22847491142786522,
+      "epoch": 0.23606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.8004418856055155e-05,
-      "loss": 0.8459,
+      "learning_rate": 1.785469113487434e-05,
+      "loss": 0.9274,
       "step": 8319
     },
     {
-      "epoch": 0.22850237565570844,
+      "epoch": 0.23609534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.8003885639269438e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.7854122283495747e-05,
+      "loss": 0.9391,
       "step": 8320
     },
     {
-      "epoch": 0.22852983988355166,
+      "epoch": 0.23612372304199772,
       "grad_norm": 0.0,
-      "learning_rate": 1.8003352359153872e-05,
-      "loss": 0.9655,
+      "learning_rate": 1.7853553365773027e-05,
+      "loss": 0.9514,
       "step": 8321
     },
     {
-      "epoch": 0.2285573041113949,
+      "epoch": 0.23615209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.8002819015712683e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.785298438171099e-05,
+      "loss": 0.934,
       "step": 8322
     },
     {
-      "epoch": 0.22858476833923813,
+      "epoch": 0.23618047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 1.800228560895009e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.785241533131444e-05,
+      "loss": 0.9064,
       "step": 8323
     },
     {
-      "epoch": 0.22861223256708138,
+      "epoch": 0.2362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.8001752138870316e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.7851846214588187e-05,
+      "loss": 0.8847,
       "step": 8324
     },
     {
-      "epoch": 0.2286396967949246,
+      "epoch": 0.2362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.8001218605477578e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.7851277031537036e-05,
+      "loss": 1.0292,
       "step": 8325
     },
     {
-      "epoch": 0.22866716102276785,
+      "epoch": 0.23626560726447218,
       "grad_norm": 0.0,
-      "learning_rate": 1.80006850087761e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.7850707782165796e-05,
+      "loss": 1.0054,
       "step": 8326
     },
     {
-      "epoch": 0.22869462525061107,
+      "epoch": 0.23629398410896707,
       "grad_norm": 0.0,
-      "learning_rate": 1.80001513487701e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.785013846647927e-05,
+      "loss": 0.9852,
       "step": 8327
     },
     {
-      "epoch": 0.22872208947845432,
+      "epoch": 0.23632236095346199,
       "grad_norm": 0.0,
-      "learning_rate": 1.7999617625463805e-05,
-      "loss": 0.9044,
+      "learning_rate": 1.7849569084482275e-05,
+      "loss": 1.0027,
       "step": 8328
     },
     {
-      "epoch": 0.22874955370629754,
+      "epoch": 0.23635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.7999083838861436e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.7848999636179615e-05,
+      "loss": 0.9527,
       "step": 8329
     },
     {
-      "epoch": 0.2287770179341408,
+      "epoch": 0.23637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.799854998896722e-05,
-      "loss": 1.1327,
+      "learning_rate": 1.7848430121576105e-05,
+      "loss": 0.9433,
       "step": 8330
     },
     {
-      "epoch": 0.228804482161984,
+      "epoch": 0.23640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.7998016075785373e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.784786054067655e-05,
+      "loss": 0.9986,
       "step": 8331
     },
     {
-      "epoch": 0.22883194638982726,
+      "epoch": 0.23643586833144153,
       "grad_norm": 0.0,
-      "learning_rate": 1.7997482099320128e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.7847290893485766e-05,
+      "loss": 0.9616,
       "step": 8332
     },
     {
-      "epoch": 0.22885941061767048,
+      "epoch": 0.23646424517593645,
       "grad_norm": 0.0,
-      "learning_rate": 1.799694805957571e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.7846721180008563e-05,
+      "loss": 0.9374,
       "step": 8333
     },
     {
-      "epoch": 0.2288868748455137,
+      "epoch": 0.23649262202043134,
       "grad_norm": 0.0,
-      "learning_rate": 1.7996413956556335e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.7846151400249754e-05,
+      "loss": 1.0554,
       "step": 8334
     },
     {
-      "epoch": 0.22891433907335695,
+      "epoch": 0.23652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.799587979026624e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.784558155421415e-05,
+      "loss": 1.0493,
       "step": 8335
     },
     {
-      "epoch": 0.22894180330120018,
+      "epoch": 0.2365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.7995345560709645e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.7845011641906565e-05,
+      "loss": 0.957,
       "step": 8336
     },
     {
-      "epoch": 0.22896926752904342,
+      "epoch": 0.236577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.7994811267890778e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.7844441663331817e-05,
+      "loss": 1.009,
       "step": 8337
     },
     {
-      "epoch": 0.22899673175688665,
+      "epoch": 0.23660612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 1.799427691181387e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.7843871618494714e-05,
+      "loss": 1.0489,
       "step": 8338
     },
     {
-      "epoch": 0.2290241959847299,
+      "epoch": 0.2366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.7993742492483145e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.784330150740008e-05,
+      "loss": 0.8914,
       "step": 8339
     },
     {
-      "epoch": 0.22905166021257312,
+      "epoch": 0.2366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.799320800990283e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.784273133005272e-05,
+      "loss": 1.1019,
       "step": 8340
     },
     {
-      "epoch": 0.22907912444041637,
+      "epoch": 0.23669125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 1.799267346407716e-05,
-      "loss": 1.053,
+      "learning_rate": 1.784216108645746e-05,
+      "loss": 1.0304,
       "step": 8341
     },
     {
-      "epoch": 0.2291065886682596,
+      "epoch": 0.23671963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.7992138855010362e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.784159077661911e-05,
+      "loss": 0.9526,
       "step": 8342
     },
     {
-      "epoch": 0.22913405289610284,
+      "epoch": 0.23674801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 1.7991604182706663e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.7841020400542493e-05,
+      "loss": 1.0461,
       "step": 8343
     },
     {
-      "epoch": 0.22916151712394606,
+      "epoch": 0.23677639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 1.79910694471703e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.784044995823242e-05,
+      "loss": 0.9564,
       "step": 8344
     },
     {
-      "epoch": 0.2291889813517893,
+      "epoch": 0.23680476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 1.7990534648405496e-05,
-      "loss": 1.0922,
+      "learning_rate": 1.7839879449693716e-05,
+      "loss": 0.9371,
       "step": 8345
     },
     {
-      "epoch": 0.22921644557963253,
+      "epoch": 0.23683314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 1.7989999786416487e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.78393088749312e-05,
+      "loss": 1.0595,
       "step": 8346
     },
     {
-      "epoch": 0.22924390980747578,
+      "epoch": 0.23686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.7989464861207508e-05,
-      "loss": 1.0777,
+      "learning_rate": 1.783873823394969e-05,
+      "loss": 0.9156,
       "step": 8347
     },
     {
-      "epoch": 0.229271374035319,
+      "epoch": 0.23688989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 1.7988929872782782e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.7838167526754002e-05,
+      "loss": 1.0165,
       "step": 8348
     },
     {
-      "epoch": 0.22929883826316222,
+      "epoch": 0.2369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.7988394821146556e-05,
-      "loss": 1.0589,
+      "learning_rate": 1.783759675334896e-05,
+      "loss": 1.0051,
       "step": 8349
     },
     {
-      "epoch": 0.22932630249100547,
+      "epoch": 0.2369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.798785970630305e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.783702591373939e-05,
+      "loss": 1.0204,
       "step": 8350
     },
     {
-      "epoch": 0.2293537667188487,
+      "epoch": 0.2369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.7987324528256508e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.7836455007930107e-05,
+      "loss": 0.99,
       "step": 8351
     },
     {
-      "epoch": 0.22938123094669194,
+      "epoch": 0.2370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.7986789287011156e-05,
-      "loss": 1.027,
+      "learning_rate": 1.7835884035925936e-05,
+      "loss": 1.0292,
       "step": 8352
     },
     {
-      "epoch": 0.22940869517453516,
+      "epoch": 0.23703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.7986253982571235e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.7835312997731704e-05,
+      "loss": 1.0076,
       "step": 8353
     },
     {
-      "epoch": 0.2294361594023784,
+      "epoch": 0.23706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.798571861494098e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.783474189335223e-05,
+      "loss": 1.1458,
       "step": 8354
     },
     {
-      "epoch": 0.22946362363022163,
+      "epoch": 0.23708853575482405,
       "grad_norm": 0.0,
-      "learning_rate": 1.7985183184124626e-05,
-      "loss": 0.9069,
+      "learning_rate": 1.7834170722792334e-05,
+      "loss": 1.0803,
       "step": 8355
     },
     {
-      "epoch": 0.22949108785806488,
+      "epoch": 0.23711691259931897,
       "grad_norm": 0.0,
-      "learning_rate": 1.798464769012641e-05,
-      "loss": 1.046,
+      "learning_rate": 1.783359948605685e-05,
+      "loss": 0.8764,
       "step": 8356
     },
     {
-      "epoch": 0.2295185520859081,
+      "epoch": 0.23714528944381386,
       "grad_norm": 0.0,
-      "learning_rate": 1.7984112132950565e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.7833028183150597e-05,
+      "loss": 0.9552,
       "step": 8357
     },
     {
-      "epoch": 0.22954601631375135,
+      "epoch": 0.23717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7983576512601337e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.7832456814078404e-05,
+      "loss": 0.9186,
       "step": 8358
     },
     {
-      "epoch": 0.22957348054159457,
+      "epoch": 0.23720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.7983040829082954e-05,
-      "loss": 1.017,
+      "learning_rate": 1.7831885378845095e-05,
+      "loss": 0.9794,
       "step": 8359
     },
     {
-      "epoch": 0.22960094476943782,
+      "epoch": 0.23723041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.7982505082399663e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.78313138774555e-05,
+      "loss": 0.8585,
       "step": 8360
     },
     {
-      "epoch": 0.22962840899728104,
+      "epoch": 0.2372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.79819692725557e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.7830742309914444e-05,
+      "loss": 0.9966,
       "step": 8361
     },
     {
-      "epoch": 0.22965587322512426,
+      "epoch": 0.23728717366628832,
       "grad_norm": 0.0,
-      "learning_rate": 1.79814333995553e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.7830170676226757e-05,
+      "loss": 1.0054,
       "step": 8362
     },
     {
-      "epoch": 0.2296833374529675,
+      "epoch": 0.2373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.798089746340271e-05,
-      "loss": 0.9969,
+      "learning_rate": 1.782959897639726e-05,
+      "loss": 0.9966,
       "step": 8363
     },
     {
-      "epoch": 0.22971080168081073,
+      "epoch": 0.2373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.798036146410217e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.7829027210430796e-05,
+      "loss": 0.9228,
       "step": 8364
     },
     {
-      "epoch": 0.22973826590865398,
+      "epoch": 0.23737230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 1.7979825401657916e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.7828455378332186e-05,
+      "loss": 1.0394,
       "step": 8365
     },
     {
-      "epoch": 0.2297657301364972,
+      "epoch": 0.23740068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 1.7979289276074193e-05,
-      "loss": 1.0129,
+      "learning_rate": 1.782788348010626e-05,
+      "loss": 1.0972,
       "step": 8366
     },
     {
-      "epoch": 0.22979319436434045,
+      "epoch": 0.23742905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 1.7978753087355243e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.782731151575785e-05,
+      "loss": 1.0331,
       "step": 8367
     },
     {
-      "epoch": 0.22982065859218367,
+      "epoch": 0.23745743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 1.797821683550531e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.782673948529179e-05,
+      "loss": 1.0696,
       "step": 8368
     },
     {
-      "epoch": 0.22984812282002692,
+      "epoch": 0.23748581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.7977680520528632e-05,
-      "loss": 0.9758,
+      "learning_rate": 1.7826167388712905e-05,
+      "loss": 1.0023,
       "step": 8369
     },
     {
-      "epoch": 0.22987558704787014,
+      "epoch": 0.23751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.7977144142429455e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.7825595226026036e-05,
+      "loss": 0.9653,
       "step": 8370
     },
     {
-      "epoch": 0.2299030512757134,
+      "epoch": 0.23754256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.7976607701212023e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.782502299723601e-05,
+      "loss": 0.9427,
       "step": 8371
     },
     {
-      "epoch": 0.22993051550355662,
+      "epoch": 0.23757094211123722,
       "grad_norm": 0.0,
-      "learning_rate": 1.7976071196880584e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.7824450702347663e-05,
+      "loss": 1.1085,
       "step": 8372
     },
     {
-      "epoch": 0.22995797973139986,
+      "epoch": 0.23759931895573214,
       "grad_norm": 0.0,
-      "learning_rate": 1.7975534629439376e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.7823878341365835e-05,
+      "loss": 1.0368,
       "step": 8373
     },
     {
-      "epoch": 0.22998544395924309,
+      "epoch": 0.23762769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 1.797499799889265e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.782330591429535e-05,
+      "loss": 0.99,
       "step": 8374
     },
     {
-      "epoch": 0.2300129081870863,
+      "epoch": 0.2376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.7974461305244652e-05,
-      "loss": 0.9777,
+      "learning_rate": 1.7822733421141046e-05,
+      "loss": 0.9345,
       "step": 8375
     },
     {
-      "epoch": 0.23004037241492956,
+      "epoch": 0.2376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.7973924548499627e-05,
-      "loss": 0.9899,
+      "learning_rate": 1.7822160861907765e-05,
+      "loss": 1.0132,
       "step": 8376
     },
     {
-      "epoch": 0.23006783664277278,
+      "epoch": 0.23771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.797338772866182e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.7821588236600337e-05,
+      "loss": 0.9051,
       "step": 8377
     },
     {
-      "epoch": 0.23009530087061603,
+      "epoch": 0.23774120317820657,
       "grad_norm": 0.0,
-      "learning_rate": 1.7972850845735484e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.7821015545223604e-05,
+      "loss": 1.0371,
       "step": 8378
     },
     {
-      "epoch": 0.23012276509845925,
+      "epoch": 0.2377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.7972313899724864e-05,
-      "loss": 0.9317,
+      "learning_rate": 1.78204427877824e-05,
+      "loss": 0.9925,
       "step": 8379
     },
     {
-      "epoch": 0.2301502293263025,
+      "epoch": 0.23779795686719638,
       "grad_norm": 0.0,
-      "learning_rate": 1.7971776890634206e-05,
-      "loss": 1.005,
+      "learning_rate": 1.7819869964281566e-05,
+      "loss": 0.8797,
       "step": 8380
     },
     {
-      "epoch": 0.23017769355414572,
+      "epoch": 0.23782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.797123981846776e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.7819297074725937e-05,
+      "loss": 0.9899,
       "step": 8381
     },
     {
-      "epoch": 0.23020515778198897,
+      "epoch": 0.23785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.7970702683229778e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.7818724119120352e-05,
+      "loss": 1.0592,
       "step": 8382
     },
     {
-      "epoch": 0.2302326220098322,
+      "epoch": 0.23788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.7970165484924513e-05,
-      "loss": 1.1575,
+      "learning_rate": 1.7818151097469658e-05,
+      "loss": 0.8995,
       "step": 8383
     },
     {
-      "epoch": 0.23026008623767544,
+      "epoch": 0.23791146424517592,
       "grad_norm": 0.0,
-      "learning_rate": 1.7969628223556208e-05,
-      "loss": 1.0184,
+      "learning_rate": 1.7817578009778687e-05,
+      "loss": 0.8972,
       "step": 8384
     },
     {
-      "epoch": 0.23028755046551866,
+      "epoch": 0.23793984108967084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7969090899129116e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.7817004856052286e-05,
+      "loss": 1.0263,
       "step": 8385
     },
     {
-      "epoch": 0.2303150146933619,
+      "epoch": 0.23796821793416573,
       "grad_norm": 0.0,
-      "learning_rate": 1.796855351164749e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.781643163629529e-05,
+      "loss": 0.9336,
       "step": 8386
     },
     {
-      "epoch": 0.23034247892120513,
+      "epoch": 0.23799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.7968016061115584e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.7815858350512542e-05,
+      "loss": 1.0015,
       "step": 8387
     },
     {
-      "epoch": 0.23036994314904835,
+      "epoch": 0.2380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.7967478547537645e-05,
-      "loss": 1.033,
+      "learning_rate": 1.7815284998708892e-05,
+      "loss": 0.9092,
       "step": 8388
     },
     {
-      "epoch": 0.2303974073768916,
+      "epoch": 0.2380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.7966940970917935e-05,
-      "loss": 1.045,
+      "learning_rate": 1.781471158088918e-05,
+      "loss": 0.8797,
       "step": 8389
     },
     {
-      "epoch": 0.23042487160473482,
+      "epoch": 0.23808172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 1.7966403331260697e-05,
-      "loss": 0.881,
+      "learning_rate": 1.7814138097058246e-05,
+      "loss": 0.9889,
       "step": 8390
     },
     {
-      "epoch": 0.23045233583257807,
+      "epoch": 0.2381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.7965865628570192e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.7813564547220933e-05,
+      "loss": 1.0023,
       "step": 8391
     },
     {
-      "epoch": 0.2304798000604213,
+      "epoch": 0.23813847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 1.7965327862850674e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.781299093138209e-05,
+      "loss": 1.0002,
       "step": 8392
     },
     {
-      "epoch": 0.23050726428826454,
+      "epoch": 0.23816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.7964790034106392e-05,
-      "loss": 0.9364,
+      "learning_rate": 1.7812417249546564e-05,
+      "loss": 1.0243,
       "step": 8393
     },
     {
-      "epoch": 0.23053472851610776,
+      "epoch": 0.23819523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.796425214234161e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.7811843501719196e-05,
+      "loss": 1.0035,
       "step": 8394
     },
     {
-      "epoch": 0.230562192743951,
+      "epoch": 0.23822360953461974,
       "grad_norm": 0.0,
-      "learning_rate": 1.7963714187560576e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.7811269687904837e-05,
+      "loss": 1.0035,
       "step": 8395
     },
     {
-      "epoch": 0.23058965697179423,
+      "epoch": 0.23825198637911466,
       "grad_norm": 0.0,
-      "learning_rate": 1.7963176169767553e-05,
-      "loss": 1.013,
+      "learning_rate": 1.7810695808108333e-05,
+      "loss": 1.06,
       "step": 8396
     },
     {
-      "epoch": 0.23061712119963748,
+      "epoch": 0.23828036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.7962638088966798e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.7810121862334528e-05,
+      "loss": 0.9734,
       "step": 8397
     },
     {
-      "epoch": 0.2306445854274807,
+      "epoch": 0.23830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.7962099945162564e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.780954785058827e-05,
+      "loss": 1.016,
       "step": 8398
     },
     {
-      "epoch": 0.23067204965532395,
+      "epoch": 0.23833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.796156173835911e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.7808973772874414e-05,
+      "loss": 1.0068,
       "step": 8399
     },
     {
-      "epoch": 0.23069951388316717,
+      "epoch": 0.2383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.7961023468560696e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.7808399629197802e-05,
+      "loss": 1.0271,
       "step": 8400
     },
     {
-      "epoch": 0.23072697811101042,
+      "epoch": 0.2383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.7960485135771583e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.780782541956329e-05,
+      "loss": 0.9351,
       "step": 8401
     },
     {
-      "epoch": 0.23075444233885364,
+      "epoch": 0.238422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7959946739996023e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.7807251143975727e-05,
+      "loss": 0.9377,
       "step": 8402
     },
     {
-      "epoch": 0.23078190656669686,
+      "epoch": 0.2384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.795940828123828e-05,
-      "loss": 1.0358,
+      "learning_rate": 1.780667680243996e-05,
+      "loss": 1.0398,
       "step": 8403
     },
     {
-      "epoch": 0.23080937079454011,
+      "epoch": 0.23847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.7958869759502623e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.7806102394960846e-05,
+      "loss": 1.1302,
       "step": 8404
     },
     {
-      "epoch": 0.23083683502238334,
+      "epoch": 0.23850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.79583311747933e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.7805527921543232e-05,
+      "loss": 1.0442,
       "step": 8405
     },
     {
-      "epoch": 0.23086429925022658,
+      "epoch": 0.23853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.795779252711458e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.7804953382191974e-05,
+      "loss": 0.9541,
       "step": 8406
     },
     {
-      "epoch": 0.2308917634780698,
+      "epoch": 0.23856413166855844,
       "grad_norm": 0.0,
-      "learning_rate": 1.7957253816470724e-05,
-      "loss": 0.9416,
+      "learning_rate": 1.7804378776911923e-05,
+      "loss": 0.9564,
       "step": 8407
     },
     {
-      "epoch": 0.23091922770591305,
+      "epoch": 0.23859250851305336,
       "grad_norm": 0.0,
-      "learning_rate": 1.795671504286599e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.7803804105707935e-05,
+      "loss": 0.8895,
       "step": 8408
     },
     {
-      "epoch": 0.23094669193375628,
+      "epoch": 0.23862088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.7956176206304646e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.780322936858486e-05,
+      "loss": 1.0505,
       "step": 8409
     },
     {
-      "epoch": 0.23097415616159953,
+      "epoch": 0.23864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.7955637306790955e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.7802654565547557e-05,
+      "loss": 0.9769,
       "step": 8410
     },
     {
-      "epoch": 0.23100162038944275,
+      "epoch": 0.23867763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.795509834432918e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.780207969660088e-05,
+      "loss": 0.9046,
       "step": 8411
     },
     {
-      "epoch": 0.231029084617286,
+      "epoch": 0.2387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.7954559318923585e-05,
-      "loss": 1.0426,
+      "learning_rate": 1.7801504761749686e-05,
+      "loss": 0.8781,
       "step": 8412
     },
     {
-      "epoch": 0.23105654884512922,
+      "epoch": 0.23873439273552782,
       "grad_norm": 0.0,
-      "learning_rate": 1.7954020230578434e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.7800929760998832e-05,
+      "loss": 1.0064,
       "step": 8413
     },
     {
-      "epoch": 0.23108401307297247,
+      "epoch": 0.2387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.7953481079297996e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.780035469435317e-05,
+      "loss": 1.0484,
       "step": 8414
     },
     {
-      "epoch": 0.2311114773008157,
+      "epoch": 0.2387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.7952941865086534e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.779977956181756e-05,
+      "loss": 0.9202,
       "step": 8415
     },
     {
-      "epoch": 0.2311389415286589,
+      "epoch": 0.23881952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 1.7952402587948313e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.779920436339687e-05,
+      "loss": 0.8359,
       "step": 8416
     },
     {
-      "epoch": 0.23116640575650216,
+      "epoch": 0.23884790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 1.7951863247887606e-05,
-      "loss": 1.001,
+      "learning_rate": 1.779862909909594e-05,
+      "loss": 0.9492,
       "step": 8417
     },
     {
-      "epoch": 0.23119386998434538,
+      "epoch": 0.23887627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.7951323844908675e-05,
-      "loss": 0.9629,
+      "learning_rate": 1.7798053768919646e-05,
+      "loss": 1.0414,
       "step": 8418
     },
     {
-      "epoch": 0.23122133421218863,
+      "epoch": 0.23890465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.795078437901579e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.7797478372872838e-05,
+      "loss": 1.054,
       "step": 8419
     },
     {
-      "epoch": 0.23124879844003185,
+      "epoch": 0.23893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.7950244850213217e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.779690291096038e-05,
+      "loss": 1.0424,
       "step": 8420
     },
     {
-      "epoch": 0.2312762626678751,
+      "epoch": 0.23896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.794970525850523e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.7796327383187136e-05,
+      "loss": 0.9677,
       "step": 8421
     },
     {
-      "epoch": 0.23130372689571832,
+      "epoch": 0.23898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.7949165603896096e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.779575178955796e-05,
+      "loss": 1.0249,
       "step": 8422
     },
     {
-      "epoch": 0.23133119112356157,
+      "epoch": 0.23901816118047672,
       "grad_norm": 0.0,
-      "learning_rate": 1.794862588639008e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.779517613007772e-05,
+      "loss": 1.0133,
       "step": 8423
     },
     {
-      "epoch": 0.2313586553514048,
+      "epoch": 0.2390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.7948086105991462e-05,
-      "loss": 1.1434,
+      "learning_rate": 1.7794600404751276e-05,
+      "loss": 1.0045,
       "step": 8424
     },
     {
-      "epoch": 0.23138611957924804,
+      "epoch": 0.23907491486946653,
       "grad_norm": 0.0,
-      "learning_rate": 1.7947546262704504e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.7794024613583493e-05,
+      "loss": 1.0774,
       "step": 8425
     },
     {
-      "epoch": 0.23141358380709126,
+      "epoch": 0.2391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7947006356533483e-05,
-      "loss": 1.2233,
+      "learning_rate": 1.7793448756579233e-05,
+      "loss": 1.0076,
       "step": 8426
     },
     {
-      "epoch": 0.2314410480349345,
+      "epoch": 0.2391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.794646638748267e-05,
-      "loss": 1.0675,
+      "learning_rate": 1.7792872833743363e-05,
+      "loss": 0.8775,
       "step": 8427
     },
     {
-      "epoch": 0.23146851226277773,
+      "epoch": 0.2391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.794592635555634e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.7792296845080744e-05,
+      "loss": 0.945,
       "step": 8428
     },
     {
-      "epoch": 0.23149597649062095,
+      "epoch": 0.23918842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.794538626075876e-05,
-      "loss": 1.0386,
+      "learning_rate": 1.7791720790596242e-05,
+      "loss": 1.0206,
       "step": 8429
     },
     {
-      "epoch": 0.2315234407184642,
+      "epoch": 0.23921679909194096,
       "grad_norm": 0.0,
-      "learning_rate": 1.7944846103094203e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.7791144670294725e-05,
+      "loss": 0.9323,
       "step": 8430
     },
     {
-      "epoch": 0.23155090494630742,
+      "epoch": 0.23924517593643588,
       "grad_norm": 0.0,
-      "learning_rate": 1.794430588256695e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.7790568484181063e-05,
+      "loss": 1.0195,
       "step": 8431
     },
     {
-      "epoch": 0.23157836917415067,
+      "epoch": 0.23927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7943765599181268e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.7789992232260113e-05,
+      "loss": 1.0212,
       "step": 8432
     },
     {
-      "epoch": 0.2316058334019939,
+      "epoch": 0.23930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.7943225252941436e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.778941591453675e-05,
+      "loss": 0.9639,
       "step": 8433
     },
     {
-      "epoch": 0.23163329762983714,
+      "epoch": 0.23933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.7942684843851732e-05,
-      "loss": 1.068,
+      "learning_rate": 1.7788839531015846e-05,
+      "loss": 0.9628,
       "step": 8434
     },
     {
-      "epoch": 0.23166076185768036,
+      "epoch": 0.23935868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.7942144371916425e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.7788263081702256e-05,
+      "loss": 0.8031,
       "step": 8435
     },
     {
-      "epoch": 0.2316882260855236,
+      "epoch": 0.23938706015891034,
       "grad_norm": 0.0,
-      "learning_rate": 1.7941603837139798e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.778768656660086e-05,
+      "loss": 0.8933,
       "step": 8436
     },
     {
-      "epoch": 0.23171569031336683,
+      "epoch": 0.23941543700340523,
       "grad_norm": 0.0,
-      "learning_rate": 1.7941063239526124e-05,
-      "loss": 1.038,
+      "learning_rate": 1.778710998571653e-05,
+      "loss": 1.0819,
       "step": 8437
     },
     {
-      "epoch": 0.23174315454121008,
+      "epoch": 0.23944381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 1.794052257907968e-05,
-      "loss": 1.0801,
+      "learning_rate": 1.7786533339054126e-05,
+      "loss": 1.0255,
       "step": 8438
     },
     {
-      "epoch": 0.2317706187690533,
+      "epoch": 0.239472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.7939981855804752e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.7785956626618527e-05,
+      "loss": 1.0064,
       "step": 8439
     },
     {
-      "epoch": 0.23179808299689655,
+      "epoch": 0.2395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.793944106970561e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.7785379848414607e-05,
+      "loss": 0.9346,
       "step": 8440
     },
     {
-      "epoch": 0.23182554722473978,
+      "epoch": 0.23952894438138478,
       "grad_norm": 0.0,
-      "learning_rate": 1.793890022078653e-05,
-      "loss": 0.9996,
+      "learning_rate": 1.7784803004447226e-05,
+      "loss": 1.1198,
       "step": 8441
     },
     {
-      "epoch": 0.23185301145258302,
+      "epoch": 0.2395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.7938359309051797e-05,
-      "loss": 0.9501,
+      "learning_rate": 1.7784226094721268e-05,
+      "loss": 0.933,
       "step": 8442
     },
     {
-      "epoch": 0.23188047568042625,
+      "epoch": 0.23958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.7937818334505693e-05,
-      "loss": 0.947,
+      "learning_rate": 1.7783649119241603e-05,
+      "loss": 0.936,
       "step": 8443
     },
     {
-      "epoch": 0.23190793990826947,
+      "epoch": 0.23961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.7937277297152493e-05,
-      "loss": 1.075,
+      "learning_rate": 1.7783072078013103e-05,
+      "loss": 0.9595,
       "step": 8444
     },
     {
-      "epoch": 0.23193540413611272,
+      "epoch": 0.23964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.793673619699648e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.778249497104064e-05,
+      "loss": 0.9498,
       "step": 8445
     },
     {
-      "epoch": 0.23196286836395594,
+      "epoch": 0.23967082860385924,
       "grad_norm": 0.0,
-      "learning_rate": 1.7936195034041938e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.7781917798329095e-05,
+      "loss": 1.0234,
       "step": 8446
     },
     {
-      "epoch": 0.2319903325917992,
+      "epoch": 0.23969920544835413,
       "grad_norm": 0.0,
-      "learning_rate": 1.7935653808293146e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.778134055988334e-05,
+      "loss": 1.0554,
       "step": 8447
     },
     {
-      "epoch": 0.2320177968196424,
+      "epoch": 0.23972758229284905,
       "grad_norm": 0.0,
-      "learning_rate": 1.793511251975439e-05,
-      "loss": 1.032,
+      "learning_rate": 1.7780763255708252e-05,
+      "loss": 0.9862,
       "step": 8448
     },
     {
-      "epoch": 0.23204526104748566,
+      "epoch": 0.23975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.7934571168429944e-05,
-      "loss": 1.1029,
+      "learning_rate": 1.7780185885808708e-05,
+      "loss": 0.9772,
       "step": 8449
     },
     {
-      "epoch": 0.23207272527532888,
+      "epoch": 0.23978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.79340297543241e-05,
-      "loss": 0.9378,
+      "learning_rate": 1.7779608450189582e-05,
+      "loss": 1.016,
       "step": 8450
     },
     {
-      "epoch": 0.23210018950317213,
+      "epoch": 0.2398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.7933488277441143e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.7779030948855756e-05,
+      "loss": 0.8566,
       "step": 8451
     },
     {
-      "epoch": 0.23212765373101535,
+      "epoch": 0.2398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.793294673778535e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.7778453381812102e-05,
+      "loss": 0.8473,
       "step": 8452
     },
     {
-      "epoch": 0.2321551179588586,
+      "epoch": 0.2398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.793240513536101e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.777787574906351e-05,
+      "loss": 0.9327,
       "step": 8453
     },
     {
-      "epoch": 0.23218258218670182,
+      "epoch": 0.2398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.7931863470172403e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.7777298050614845e-05,
+      "loss": 0.9819,
       "step": 8454
     },
     {
-      "epoch": 0.23221004641454507,
+      "epoch": 0.23992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.7931321742223828e-05,
-      "loss": 1.0532,
+      "learning_rate": 1.7776720286470998e-05,
+      "loss": 0.9824,
       "step": 8455
     },
     {
-      "epoch": 0.2322375106423883,
+      "epoch": 0.23995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.7930779951519557e-05,
-      "loss": 0.8542,
+      "learning_rate": 1.7776142456636843e-05,
+      "loss": 1.1274,
       "step": 8456
     },
     {
-      "epoch": 0.2322649748702315,
+      "epoch": 0.23998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.7930238098063885e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.7775564561117265e-05,
+      "loss": 0.9524,
       "step": 8457
     },
     {
-      "epoch": 0.23229243909807476,
+      "epoch": 0.24001135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.79296961818611e-05,
-      "loss": 0.8125,
+      "learning_rate": 1.777498659991714e-05,
+      "loss": 1.0365,
       "step": 8458
     },
     {
-      "epoch": 0.23231990332591798,
+      "epoch": 0.24003972758229286,
       "grad_norm": 0.0,
-      "learning_rate": 1.7929154202915486e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.7774408573041356e-05,
+      "loss": 0.9536,
       "step": 8459
     },
     {
-      "epoch": 0.23234736755376123,
+      "epoch": 0.24006810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 1.7928612161231328e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.7773830480494792e-05,
+      "loss": 1.0342,
       "step": 8460
     },
     {
-      "epoch": 0.23237483178160445,
+      "epoch": 0.24009648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.7928070056812922e-05,
-      "loss": 0.9544,
+      "learning_rate": 1.7773252322282335e-05,
+      "loss": 1.0269,
       "step": 8461
     },
     {
-      "epoch": 0.2324022960094477,
+      "epoch": 0.24012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.7927527889664557e-05,
-      "loss": 1.084,
+      "learning_rate": 1.7772674098408863e-05,
+      "loss": 0.9932,
       "step": 8462
     },
     {
-      "epoch": 0.23242976023729092,
+      "epoch": 0.2401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.792698565979052e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.7772095808879265e-05,
+      "loss": 0.8202,
       "step": 8463
     },
     {
-      "epoch": 0.23245722446513417,
+      "epoch": 0.2401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.7926443367195097e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.7771517453698423e-05,
+      "loss": 0.976,
       "step": 8464
     },
     {
-      "epoch": 0.2324846886929774,
+      "epoch": 0.2402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.7925901011882588e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.7770939032871227e-05,
+      "loss": 0.9784,
       "step": 8465
     },
     {
-      "epoch": 0.23251215292082064,
+      "epoch": 0.2402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.792535859385728e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.777036054640256e-05,
+      "loss": 1.0186,
       "step": 8466
     },
     {
-      "epoch": 0.23253961714866386,
+      "epoch": 0.240266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.7924816113123468e-05,
-      "loss": 0.922,
+      "learning_rate": 1.7769781994297305e-05,
+      "loss": 1.0792,
       "step": 8467
     },
     {
-      "epoch": 0.2325670813765071,
+      "epoch": 0.24029511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 1.7924273569685432e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.7769203376560353e-05,
+      "loss": 0.9953,
       "step": 8468
     },
     {
-      "epoch": 0.23259454560435033,
+      "epoch": 0.24032349602724176,
       "grad_norm": 0.0,
-      "learning_rate": 1.7923730963547484e-05,
-      "loss": 1.0169,
+      "learning_rate": 1.776862469319659e-05,
+      "loss": 1.1203,
       "step": 8469
     },
     {
-      "epoch": 0.23262200983219355,
+      "epoch": 0.24035187287173665,
       "grad_norm": 0.0,
-      "learning_rate": 1.79231882947139e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.7768045944210906e-05,
+      "loss": 1.1005,
       "step": 8470
     },
     {
-      "epoch": 0.2326494740600368,
+      "epoch": 0.24038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.7922645563188984e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.7767467129608186e-05,
+      "loss": 0.928,
       "step": 8471
     },
     {
-      "epoch": 0.23267693828788003,
+      "epoch": 0.24040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.7922102768977026e-05,
-      "loss": 0.997,
+      "learning_rate": 1.7766888249393325e-05,
+      "loss": 1.0504,
       "step": 8472
     },
     {
-      "epoch": 0.23270440251572327,
+      "epoch": 0.24043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.7921559912082322e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.7766309303571212e-05,
+      "loss": 0.9954,
       "step": 8473
     },
     {
-      "epoch": 0.2327318667435665,
+      "epoch": 0.24046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.792101699250917e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.776573029214673e-05,
+      "loss": 0.9778,
       "step": 8474
     },
     {
-      "epoch": 0.23275933097140974,
+      "epoch": 0.2404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.7920474010261863e-05,
-      "loss": 0.8871,
+      "learning_rate": 1.776515121512478e-05,
+      "loss": 1.1823,
       "step": 8475
     },
     {
-      "epoch": 0.23278679519925297,
+      "epoch": 0.24052213393870603,
       "grad_norm": 0.0,
-      "learning_rate": 1.7919930965344696e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.7764572072510248e-05,
+      "loss": 0.9761,
       "step": 8476
     },
     {
-      "epoch": 0.23281425942709622,
+      "epoch": 0.24055051078320092,
       "grad_norm": 0.0,
-      "learning_rate": 1.7919387857761966e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.7763992864308024e-05,
+      "loss": 0.9767,
       "step": 8477
     },
     {
-      "epoch": 0.23284172365493944,
+      "epoch": 0.2405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.7918844687517977e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.7763413590523004e-05,
+      "loss": 1.056,
       "step": 8478
     },
     {
-      "epoch": 0.23286918788278269,
+      "epoch": 0.2406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.7918301454617017e-05,
-      "loss": 1.1089,
+      "learning_rate": 1.776283425116008e-05,
+      "loss": 1.0311,
       "step": 8479
     },
     {
-      "epoch": 0.2328966521106259,
+      "epoch": 0.24063564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.7917758159063387e-05,
-      "loss": 0.936,
+      "learning_rate": 1.7762254846224144e-05,
+      "loss": 0.9532,
       "step": 8480
     },
     {
-      "epoch": 0.23292411633846916,
+      "epoch": 0.24066401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.7917214800861388e-05,
-      "loss": 1.0243,
+      "learning_rate": 1.7761675375720096e-05,
+      "loss": 0.9496,
       "step": 8481
     },
     {
-      "epoch": 0.23295158056631238,
+      "epoch": 0.24069239500567538,
       "grad_norm": 0.0,
-      "learning_rate": 1.791667138001532e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.776109583965283e-05,
+      "loss": 0.9189,
       "step": 8482
     },
     {
-      "epoch": 0.2329790447941556,
+      "epoch": 0.24072077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.7916127896529484e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.7760516238027233e-05,
+      "loss": 1.136,
       "step": 8483
     },
     {
-      "epoch": 0.23300650902199885,
+      "epoch": 0.24074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.7915584350408174e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.775993657084821e-05,
+      "loss": 0.9478,
       "step": 8484
     },
     {
-      "epoch": 0.23303397324984207,
+      "epoch": 0.24077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.7915040741655696e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.775935683812065e-05,
+      "loss": 0.888,
       "step": 8485
     },
     {
-      "epoch": 0.23306143747768532,
+      "epoch": 0.24080590238365493,
       "grad_norm": 0.0,
-      "learning_rate": 1.791449707027635e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.775877703984946e-05,
+      "loss": 0.9818,
       "step": 8486
     },
     {
-      "epoch": 0.23308890170552854,
+      "epoch": 0.24083427922814982,
       "grad_norm": 0.0,
-      "learning_rate": 1.7913953336274438e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.7758197176039526e-05,
+      "loss": 0.9786,
       "step": 8487
     },
     {
-      "epoch": 0.2331163659333718,
+      "epoch": 0.24086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.7913409539654257e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.7757617246695755e-05,
+      "loss": 0.8965,
       "step": 8488
     },
     {
-      "epoch": 0.233143830161215,
+      "epoch": 0.24089103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 1.7912865680420117e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.775703725182304e-05,
+      "loss": 0.8493,
       "step": 8489
     },
     {
-      "epoch": 0.23317129438905826,
+      "epoch": 0.2409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.791232175857632e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.7756457191426282e-05,
+      "loss": 0.8683,
       "step": 8490
     },
     {
-      "epoch": 0.23319875861690148,
+      "epoch": 0.2409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.791177777412717e-05,
-      "loss": 1.1743,
+      "learning_rate": 1.7755877065510386e-05,
+      "loss": 1.0242,
       "step": 8491
     },
     {
-      "epoch": 0.23322622284474473,
+      "epoch": 0.24097616345062428,
       "grad_norm": 0.0,
-      "learning_rate": 1.7911233727076967e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.7755296874080242e-05,
+      "loss": 0.9791,
       "step": 8492
     },
     {
-      "epoch": 0.23325368707258795,
+      "epoch": 0.2410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.7910689617430018e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.775471661714076e-05,
+      "loss": 0.9512,
       "step": 8493
     },
     {
-      "epoch": 0.2332811513004312,
+      "epoch": 0.24103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.791014544519063e-05,
-      "loss": 0.9755,
+      "learning_rate": 1.775413629469684e-05,
+      "loss": 1.0071,
       "step": 8494
     },
     {
-      "epoch": 0.23330861552827442,
+      "epoch": 0.24106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.7909601210363104e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.775355590675338e-05,
+      "loss": 0.9753,
       "step": 8495
     },
     {
-      "epoch": 0.23333607975611767,
+      "epoch": 0.24108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.7909056912951752e-05,
-      "loss": 1.0344,
+      "learning_rate": 1.775297545331528e-05,
+      "loss": 0.9821,
       "step": 8496
     },
     {
-      "epoch": 0.2333635439839609,
+      "epoch": 0.24111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.7908512552960878e-05,
-      "loss": 0.979,
+      "learning_rate": 1.7752394934387455e-05,
+      "loss": 0.963,
       "step": 8497
     },
     {
-      "epoch": 0.2333910082118041,
+      "epoch": 0.24114642451759363,
       "grad_norm": 0.0,
-      "learning_rate": 1.7907968130394787e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.77518143499748e-05,
+      "loss": 0.9975,
       "step": 8498
     },
     {
-      "epoch": 0.23341847243964736,
+      "epoch": 0.24117480136208855,
       "grad_norm": 0.0,
-      "learning_rate": 1.790742364525779e-05,
-      "loss": 1.0285,
+      "learning_rate": 1.775123370008222e-05,
+      "loss": 1.0464,
       "step": 8499
     },
     {
-      "epoch": 0.23344593666749058,
+      "epoch": 0.24120317820658344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7906879097554196e-05,
-      "loss": 0.9523,
+      "learning_rate": 1.7750652984714617e-05,
+      "loss": 0.9218,
       "step": 8500
     },
     {
-      "epoch": 0.23347340089533383,
+      "epoch": 0.24123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.790633448728831e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.7750072203876905e-05,
+      "loss": 0.9599,
       "step": 8501
     },
     {
-      "epoch": 0.23350086512317705,
+      "epoch": 0.2412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.7905789814464443e-05,
-      "loss": 0.9454,
+      "learning_rate": 1.774949135757398e-05,
+      "loss": 0.8682,
       "step": 8502
     },
     {
-      "epoch": 0.2335283293510203,
+      "epoch": 0.2412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.7905245079086906e-05,
-      "loss": 1.0277,
+      "learning_rate": 1.7748910445810756e-05,
+      "loss": 1.0491,
       "step": 8503
     },
     {
-      "epoch": 0.23355579357886352,
+      "epoch": 0.24131668558456298,
       "grad_norm": 0.0,
-      "learning_rate": 1.7904700281160005e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.774832946859214e-05,
+      "loss": 0.9524,
       "step": 8504
     },
     {
-      "epoch": 0.23358325780670677,
+      "epoch": 0.2413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.7904155420688055e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.774774842592303e-05,
+      "loss": 1.0581,
       "step": 8505
     },
     {
-      "epoch": 0.23361072203455,
+      "epoch": 0.2413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.7903610497675367e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.7747167317808346e-05,
+      "loss": 0.9272,
       "step": 8506
     },
     {
-      "epoch": 0.23363818626239324,
+      "epoch": 0.24140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.7903065512126246e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.774658614425299e-05,
+      "loss": 1.0197,
       "step": 8507
     },
     {
-      "epoch": 0.23366565049023647,
+      "epoch": 0.24143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.7902520464045016e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.774600490526187e-05,
+      "loss": 0.9078,
       "step": 8508
     },
     {
-      "epoch": 0.23369311471807971,
+      "epoch": 0.24145856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.790197535343598e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.77454236008399e-05,
+      "loss": 0.9566,
       "step": 8509
     },
     {
-      "epoch": 0.23372057894592294,
+      "epoch": 0.24148694665153234,
       "grad_norm": 0.0,
-      "learning_rate": 1.7901430180303455e-05,
-      "loss": 1.0921,
+      "learning_rate": 1.774484223099199e-05,
+      "loss": 1.0909,
       "step": 8510
     },
     {
-      "epoch": 0.23374804317376616,
+      "epoch": 0.24151532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 1.790088494465175e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.774426079572305e-05,
+      "loss": 0.9745,
       "step": 8511
     },
     {
-      "epoch": 0.2337755074016094,
+      "epoch": 0.24154370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.7900339646485187e-05,
-      "loss": 0.984,
+      "learning_rate": 1.774367929503799e-05,
+      "loss": 1.02,
       "step": 8512
     },
     {
-      "epoch": 0.23380297162945263,
+      "epoch": 0.24157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.7899794285808078e-05,
-      "loss": 0.8849,
+      "learning_rate": 1.7743097728941723e-05,
+      "loss": 0.9829,
       "step": 8513
     },
     {
-      "epoch": 0.23383043585729588,
+      "epoch": 0.2416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.789924886262473e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.774251609743916e-05,
+      "loss": 0.9391,
       "step": 8514
     },
     {
-      "epoch": 0.2338579000851391,
+      "epoch": 0.2416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.7898703376939467e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.7741934400535216e-05,
+      "loss": 0.928,
       "step": 8515
     },
     {
-      "epoch": 0.23388536431298235,
+      "epoch": 0.24165720771850172,
       "grad_norm": 0.0,
-      "learning_rate": 1.7898157828756605e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.7741352638234807e-05,
+      "loss": 1.0232,
       "step": 8516
     },
     {
-      "epoch": 0.23391282854082557,
+      "epoch": 0.2416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.789761221808046e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.774077081054284e-05,
+      "loss": 1.025,
       "step": 8517
     },
     {
-      "epoch": 0.23394029276866882,
+      "epoch": 0.2417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.7897066544915345e-05,
-      "loss": 1.2063,
+      "learning_rate": 1.774018891746424e-05,
+      "loss": 0.918,
       "step": 8518
     },
     {
-      "epoch": 0.23396775699651204,
+      "epoch": 0.24174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.7896520809265583e-05,
-      "loss": 1.007,
+      "learning_rate": 1.773960695900391e-05,
+      "loss": 1.0599,
       "step": 8519
     },
     {
-      "epoch": 0.2339952212243553,
+      "epoch": 0.24177071509648126,
       "grad_norm": 0.0,
-      "learning_rate": 1.789597501113549e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.7739024935166773e-05,
+      "loss": 0.9611,
       "step": 8520
     },
     {
-      "epoch": 0.2340226854521985,
+      "epoch": 0.24179909194097615,
       "grad_norm": 0.0,
-      "learning_rate": 1.789542915052938e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.7738442845957745e-05,
+      "loss": 1.0627,
       "step": 8521
     },
     {
-      "epoch": 0.23405014968004176,
+      "epoch": 0.24182746878547107,
       "grad_norm": 0.0,
-      "learning_rate": 1.789488322745158e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.773786069138174e-05,
+      "loss": 0.9617,
       "step": 8522
     },
     {
-      "epoch": 0.23407761390788498,
+      "epoch": 0.24185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.7894337241906403e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.7737278471443685e-05,
+      "loss": 1.0335,
       "step": 8523
     },
     {
-      "epoch": 0.2341050781357282,
+      "epoch": 0.24188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7893791193898174e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.773669618614849e-05,
+      "loss": 1.0392,
       "step": 8524
     },
     {
-      "epoch": 0.23413254236357145,
+      "epoch": 0.24191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.7893245083431214e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.7736113835501068e-05,
+      "loss": 0.9794,
       "step": 8525
     },
     {
-      "epoch": 0.23416000659141467,
+      "epoch": 0.24194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.7892698910509834e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.773553141950635e-05,
+      "loss": 0.9582,
       "step": 8526
     },
     {
-      "epoch": 0.23418747081925792,
+      "epoch": 0.2419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.7892152675138368e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.7734948938169244e-05,
+      "loss": 0.953,
       "step": 8527
     },
     {
-      "epoch": 0.23421493504710114,
+      "epoch": 0.24199772985244042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7891606377321133e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.7734366391494686e-05,
+      "loss": 0.9497,
       "step": 8528
     },
     {
-      "epoch": 0.2342423992749444,
+      "epoch": 0.2420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.7891060017062448e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.773378377948758e-05,
+      "loss": 0.9,
       "step": 8529
     },
     {
-      "epoch": 0.2342698635027876,
+      "epoch": 0.2420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.7890513594366644e-05,
-      "loss": 0.8789,
+      "learning_rate": 1.773320110215286e-05,
+      "loss": 0.9517,
       "step": 8530
     },
     {
-      "epoch": 0.23429732773063086,
+      "epoch": 0.24208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.7889967109238037e-05,
-      "loss": 1.1033,
+      "learning_rate": 1.7732618359495436e-05,
+      "loss": 0.9677,
       "step": 8531
     },
     {
-      "epoch": 0.23432479195847408,
+      "epoch": 0.24211123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 1.7889420561680954e-05,
-      "loss": 0.8341,
+      "learning_rate": 1.773203555152024e-05,
+      "loss": 0.9241,
       "step": 8532
     },
     {
-      "epoch": 0.23435225618631733,
+      "epoch": 0.24213961407491488,
       "grad_norm": 0.0,
-      "learning_rate": 1.7888873951699718e-05,
-      "loss": 0.9851,
+      "learning_rate": 1.7731452678232196e-05,
+      "loss": 0.9701,
       "step": 8533
     },
     {
-      "epoch": 0.23437972041416055,
+      "epoch": 0.24216799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 1.7888327279298655e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.7730869739636222e-05,
+      "loss": 1.0366,
       "step": 8534
     },
     {
-      "epoch": 0.2344071846420038,
+      "epoch": 0.24219636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.7887780544482096e-05,
-      "loss": 1.1058,
+      "learning_rate": 1.773028673573724e-05,
+      "loss": 0.9966,
       "step": 8535
     },
     {
-      "epoch": 0.23443464886984702,
+      "epoch": 0.24222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.7887233747254357e-05,
-      "loss": 0.929,
+      "learning_rate": 1.7729703666540183e-05,
+      "loss": 1.127,
       "step": 8536
     },
     {
-      "epoch": 0.23446211309769027,
+      "epoch": 0.24225312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 1.7886686887619772e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.7729120532049967e-05,
+      "loss": 0.9765,
       "step": 8537
     },
     {
-      "epoch": 0.2344895773255335,
+      "epoch": 0.24228149829738932,
       "grad_norm": 0.0,
-      "learning_rate": 1.7886139965582664e-05,
-      "loss": 0.9521,
+      "learning_rate": 1.7728537332271526e-05,
+      "loss": 0.9599,
       "step": 8538
     },
     {
-      "epoch": 0.23451704155337671,
+      "epoch": 0.24230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.7885592981147365e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.7727954067209782e-05,
+      "loss": 1.0105,
       "step": 8539
     },
     {
-      "epoch": 0.23454450578121996,
+      "epoch": 0.24233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.78850459343182e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.7727370736869662e-05,
+      "loss": 0.9273,
       "step": 8540
     },
     {
-      "epoch": 0.23457197000906319,
+      "epoch": 0.242366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7884498825099492e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.77267873412561e-05,
+      "loss": 0.9559,
       "step": 8541
     },
     {
-      "epoch": 0.23459943423690643,
+      "epoch": 0.2423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.7883951653495578e-05,
-      "loss": 1.0113,
+      "learning_rate": 1.772620388037401e-05,
+      "loss": 0.9157,
       "step": 8542
     },
     {
-      "epoch": 0.23462689846474966,
+      "epoch": 0.24242338251986378,
       "grad_norm": 0.0,
-      "learning_rate": 1.7883404419510784e-05,
-      "loss": 1.0372,
+      "learning_rate": 1.772562035422833e-05,
+      "loss": 0.923,
       "step": 8543
     },
     {
-      "epoch": 0.2346543626925929,
+      "epoch": 0.24245175936435867,
       "grad_norm": 0.0,
-      "learning_rate": 1.7882857123149443e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.7725036762823987e-05,
+      "loss": 0.9973,
       "step": 8544
     },
     {
-      "epoch": 0.23468182692043613,
+      "epoch": 0.24248013620885359,
       "grad_norm": 0.0,
-      "learning_rate": 1.7882309764415876e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.7724453106165915e-05,
+      "loss": 0.8442,
       "step": 8545
     },
     {
-      "epoch": 0.23470929114827938,
+      "epoch": 0.24250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.788176234331443e-05,
-      "loss": 1.001,
+      "learning_rate": 1.7723869384259038e-05,
+      "loss": 0.8405,
       "step": 8546
     },
     {
-      "epoch": 0.2347367553761226,
+      "epoch": 0.24253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7881214859849422e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.772328559710829e-05,
+      "loss": 0.9565,
       "step": 8547
     },
     {
-      "epoch": 0.23476421960396585,
+      "epoch": 0.24256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.788066731402519e-05,
-      "loss": 1.0275,
+      "learning_rate": 1.77227017447186e-05,
+      "loss": 1.0905,
       "step": 8548
     },
     {
-      "epoch": 0.23479168383180907,
+      "epoch": 0.24259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.7880119705846064e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.7722117827094904e-05,
+      "loss": 1.0572,
       "step": 8549
     },
     {
-      "epoch": 0.23481914805965232,
+      "epoch": 0.24262202043132802,
       "grad_norm": 0.0,
-      "learning_rate": 1.787957203531638e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.7721533844242126e-05,
+      "loss": 1.1836,
       "step": 8550
     },
     {
-      "epoch": 0.23484661228749554,
+      "epoch": 0.24265039727582294,
       "grad_norm": 0.0,
-      "learning_rate": 1.7879024302440468e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.772094979616521e-05,
+      "loss": 1.0289,
       "step": 8551
     },
     {
-      "epoch": 0.23487407651533876,
+      "epoch": 0.24267877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.7878476507222668e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.772036568286908e-05,
+      "loss": 0.9548,
       "step": 8552
     },
     {
-      "epoch": 0.234901540743182,
+      "epoch": 0.2427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.7877928649667305e-05,
-      "loss": 1.0426,
+      "learning_rate": 1.7719781504358678e-05,
+      "loss": 1.0242,
       "step": 8553
     },
     {
-      "epoch": 0.23492900497102523,
+      "epoch": 0.2427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7877380729778723e-05,
-      "loss": 1.0132,
+      "learning_rate": 1.7719197260638933e-05,
+      "loss": 1.0425,
       "step": 8554
     },
     {
-      "epoch": 0.23495646919886848,
+      "epoch": 0.2427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.787683274756125e-05,
-      "loss": 1.0754,
+      "learning_rate": 1.771861295171478e-05,
+      "loss": 0.9153,
       "step": 8555
     },
     {
-      "epoch": 0.2349839334267117,
+      "epoch": 0.2427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.7876284703019226e-05,
-      "loss": 0.7054,
+      "learning_rate": 1.7718028577591156e-05,
+      "loss": 0.9728,
       "step": 8556
     },
     {
-      "epoch": 0.23501139765455495,
+      "epoch": 0.2428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.7875736596156988e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.7717444138273e-05,
+      "loss": 0.9543,
       "step": 8557
     },
     {
-      "epoch": 0.23503886188239817,
+      "epoch": 0.24284903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 1.787518842697887e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.7716859633765248e-05,
+      "loss": 0.9874,
       "step": 8558
     },
     {
-      "epoch": 0.23506632611024142,
+      "epoch": 0.24287741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.7874640195489213e-05,
-      "loss": 0.922,
+      "learning_rate": 1.7716275064072828e-05,
+      "loss": 0.9847,
       "step": 8559
     },
     {
-      "epoch": 0.23509379033808464,
+      "epoch": 0.24290578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.787409190169235e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.7715690429200695e-05,
+      "loss": 0.9785,
       "step": 8560
     },
     {
-      "epoch": 0.2351212545659279,
+      "epoch": 0.24293416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.7873543545592622e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.7715105729153774e-05,
+      "loss": 1.1307,
       "step": 8561
     },
     {
-      "epoch": 0.2351487187937711,
+      "epoch": 0.24296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.7872995127194367e-05,
-      "loss": 1.001,
+      "learning_rate": 1.771452096393701e-05,
+      "loss": 0.9576,
       "step": 8562
     },
     {
-      "epoch": 0.23517618302161436,
+      "epoch": 0.24299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.7872446646501927e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.7713936133555336e-05,
+      "loss": 1.0134,
       "step": 8563
     },
     {
-      "epoch": 0.23520364724945758,
+      "epoch": 0.24301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.7871898103519636e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.77133512380137e-05,
+      "loss": 0.9946,
       "step": 8564
     },
     {
-      "epoch": 0.2352311114773008,
+      "epoch": 0.24304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.787134949825184e-05,
-      "loss": 0.976,
+      "learning_rate": 1.771276627731704e-05,
+      "loss": 0.9744,
       "step": 8565
     },
     {
-      "epoch": 0.23525857570514405,
+      "epoch": 0.2430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.7870800830702875e-05,
-      "loss": 1.0702,
+      "learning_rate": 1.7712181251470297e-05,
+      "loss": 1.038,
       "step": 8566
     },
     {
-      "epoch": 0.23528603993298727,
+      "epoch": 0.2431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.787025210087709e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.771159616047841e-05,
+      "loss": 0.9919,
       "step": 8567
     },
     {
-      "epoch": 0.23531350416083052,
+      "epoch": 0.2431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.7869703308778818e-05,
-      "loss": 1.007,
+      "learning_rate": 1.7711011004346323e-05,
+      "loss": 1.0149,
       "step": 8568
     },
     {
-      "epoch": 0.23534096838867374,
+      "epoch": 0.243161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.786915445441241e-05,
-      "loss": 1.0852,
+      "learning_rate": 1.7710425783078982e-05,
+      "loss": 0.9149,
       "step": 8569
     },
     {
-      "epoch": 0.235368432616517,
+      "epoch": 0.24318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.7868605537782198e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.7709840496681327e-05,
+      "loss": 1.0512,
       "step": 8570
     },
     {
-      "epoch": 0.23539589684436021,
+      "epoch": 0.24321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.7868056558892534e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.7709255145158303e-05,
+      "loss": 0.8593,
       "step": 8571
     },
     {
-      "epoch": 0.23542336107220346,
+      "epoch": 0.24324631101021565,
       "grad_norm": 0.0,
-      "learning_rate": 1.7867507517747758e-05,
-      "loss": 0.9909,
+      "learning_rate": 1.770866972851485e-05,
+      "loss": 0.8949,
       "step": 8572
     },
     {
-      "epoch": 0.23545082530004668,
+      "epoch": 0.24327468785471057,
       "grad_norm": 0.0,
-      "learning_rate": 1.7866958414352216e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.7708084246755924e-05,
+      "loss": 1.13,
       "step": 8573
     },
     {
-      "epoch": 0.23547828952788993,
+      "epoch": 0.24330306469920546,
       "grad_norm": 0.0,
-      "learning_rate": 1.786640924871025e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.7707498699886462e-05,
+      "loss": 0.9983,
       "step": 8574
     },
     {
-      "epoch": 0.23550575375573315,
+      "epoch": 0.24333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.7865860020826208e-05,
-      "loss": 0.9521,
+      "learning_rate": 1.770691308791141e-05,
+      "loss": 0.9921,
       "step": 8575
     },
     {
-      "epoch": 0.2355332179835764,
+      "epoch": 0.24335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.786531073070443e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.7706327410835717e-05,
+      "loss": 0.9735,
       "step": 8576
     },
     {
-      "epoch": 0.23556068221141963,
+      "epoch": 0.24338819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.7864761378349274e-05,
-      "loss": 0.999,
+      "learning_rate": 1.770574166866433e-05,
+      "loss": 0.9386,
       "step": 8577
     },
     {
-      "epoch": 0.23558814643926285,
+      "epoch": 0.243416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.7864211963765076e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.7705155861402195e-05,
+      "loss": 0.8967,
       "step": 8578
     },
     {
-      "epoch": 0.2356156106671061,
+      "epoch": 0.24344494892167992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7863662486956184e-05,
-      "loss": 0.984,
+      "learning_rate": 1.770456998905427e-05,
+      "loss": 0.9099,
       "step": 8579
     },
     {
-      "epoch": 0.23564307489494932,
+      "epoch": 0.2434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.7863112947926953e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.7703984051625487e-05,
+      "loss": 1.0296,
       "step": 8580
     },
     {
-      "epoch": 0.23567053912279257,
+      "epoch": 0.2435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.7862563346681725e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.7703398049120804e-05,
+      "loss": 1.0226,
       "step": 8581
     },
     {
-      "epoch": 0.2356980033506358,
+      "epoch": 0.24353007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 1.786201368322485e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.7702811981545177e-05,
+      "loss": 0.9473,
       "step": 8582
     },
     {
-      "epoch": 0.23572546757847904,
+      "epoch": 0.24355845629965947,
       "grad_norm": 0.0,
-      "learning_rate": 1.7861463957560678e-05,
-      "loss": 1.014,
+      "learning_rate": 1.7702225848903548e-05,
+      "loss": 1.0121,
       "step": 8583
     },
     {
-      "epoch": 0.23575293180632226,
+      "epoch": 0.24358683314415436,
       "grad_norm": 0.0,
-      "learning_rate": 1.7860914169693556e-05,
-      "loss": 1.1064,
+      "learning_rate": 1.770163965120087e-05,
+      "loss": 1.0073,
       "step": 8584
     },
     {
-      "epoch": 0.2357803960341655,
+      "epoch": 0.24361520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 1.7860364319627836e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.7701053388442103e-05,
+      "loss": 1.1016,
       "step": 8585
     },
     {
-      "epoch": 0.23580786026200873,
+      "epoch": 0.24364358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.785981440736787e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.7700467060632184e-05,
+      "loss": 0.9973,
       "step": 8586
     },
     {
-      "epoch": 0.23583532448985198,
+      "epoch": 0.24367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.7859264432918006e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.7699880667776078e-05,
+      "loss": 0.9662,
       "step": 8587
     },
     {
-      "epoch": 0.2358627887176952,
+      "epoch": 0.24370034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.78587143962826e-05,
-      "loss": 1.0749,
+      "learning_rate": 1.769929420987873e-05,
+      "loss": 0.9897,
       "step": 8588
     },
     {
-      "epoch": 0.23589025294553845,
+      "epoch": 0.24372871736662882,
       "grad_norm": 0.0,
-      "learning_rate": 1.7858164297466e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.76987076869451e-05,
+      "loss": 1.0181,
       "step": 8589
     },
     {
-      "epoch": 0.23591771717338167,
+      "epoch": 0.2437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.785761413647256e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.769812109898014e-05,
+      "loss": 0.924,
       "step": 8590
     },
     {
-      "epoch": 0.23594518140122492,
+      "epoch": 0.24378547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.7857063913306636e-05,
-      "loss": 0.8347,
+      "learning_rate": 1.7697534445988804e-05,
+      "loss": 0.9176,
       "step": 8591
     },
     {
-      "epoch": 0.23597264562906814,
+      "epoch": 0.2438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.7856513627972575e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.769694772797605e-05,
+      "loss": 0.9429,
       "step": 8592
     },
     {
-      "epoch": 0.23600010985691136,
+      "epoch": 0.2438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.785596328047474e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.769636094494683e-05,
+      "loss": 1.0103,
       "step": 8593
     },
     {
-      "epoch": 0.2360275740847546,
+      "epoch": 0.2438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.7855412870817475e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.7695774096906105e-05,
+      "loss": 0.9462,
       "step": 8594
     },
     {
-      "epoch": 0.23605503831259783,
+      "epoch": 0.24389897843359817,
       "grad_norm": 0.0,
-      "learning_rate": 1.7854862399005143e-05,
-      "loss": 0.8916,
+      "learning_rate": 1.7695187183858832e-05,
+      "loss": 1.0049,
       "step": 8595
     },
     {
-      "epoch": 0.23608250254044108,
+      "epoch": 0.2439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.7854311865042094e-05,
-      "loss": 1.0596,
+      "learning_rate": 1.7694600205809963e-05,
+      "loss": 0.9749,
       "step": 8596
     },
     {
-      "epoch": 0.2361099667682843,
+      "epoch": 0.24395573212258798,
       "grad_norm": 0.0,
-      "learning_rate": 1.785376126893269e-05,
-      "loss": 1.034,
+      "learning_rate": 1.769401316276446e-05,
+      "loss": 0.9745,
       "step": 8597
     },
     {
-      "epoch": 0.23613743099612755,
+      "epoch": 0.24398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.7853210610681283e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.7693426054727287e-05,
+      "loss": 1.0124,
       "step": 8598
     },
     {
-      "epoch": 0.23616489522397077,
+      "epoch": 0.24401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.7852659890292235e-05,
-      "loss": 1.0865,
+      "learning_rate": 1.7692838881703397e-05,
+      "loss": 0.912,
       "step": 8599
     },
     {
-      "epoch": 0.23619235945181402,
+      "epoch": 0.24404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.78521091077699e-05,
-      "loss": 1.011,
+      "learning_rate": 1.7692251643697748e-05,
+      "loss": 0.9543,
       "step": 8600
     },
     {
-      "epoch": 0.23621982367965724,
+      "epoch": 0.24406923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 1.785155826311863e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.7691664340715307e-05,
+      "loss": 1.0016,
       "step": 8601
     },
     {
-      "epoch": 0.2362472879075005,
+      "epoch": 0.24409761634506244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7851007356342795e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.7691076972761028e-05,
+      "loss": 0.9088,
       "step": 8602
     },
     {
-      "epoch": 0.2362747521353437,
+      "epoch": 0.24412599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.7850456387446745e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.7690489539839877e-05,
+      "loss": 0.9838,
       "step": 8603
     },
     {
-      "epoch": 0.23630221636318696,
+      "epoch": 0.24415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.7849905356434846e-05,
-      "loss": 0.9678,
+      "learning_rate": 1.768990204195682e-05,
+      "loss": 1.0116,
       "step": 8604
     },
     {
-      "epoch": 0.23632968059103018,
+      "epoch": 0.2441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.7849354263311454e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.7689314479116806e-05,
+      "loss": 0.8807,
       "step": 8605
     },
     {
-      "epoch": 0.2363571448188734,
+      "epoch": 0.244211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.784880310808093e-05,
-      "loss": 0.8123,
+      "learning_rate": 1.7688726851324812e-05,
+      "loss": 0.8358,
       "step": 8606
     },
     {
-      "epoch": 0.23638460904671665,
+      "epoch": 0.24423950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 1.7848251890747636e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.76881391585858e-05,
+      "loss": 0.9498,
       "step": 8607
     },
     {
-      "epoch": 0.23641207327455988,
+      "epoch": 0.2442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.784770061131593e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.7687551400904725e-05,
+      "loss": 1.0954,
       "step": 8608
     },
     {
-      "epoch": 0.23643953750240312,
+      "epoch": 0.24429625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 1.7847149269790183e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.768696357828656e-05,
+      "loss": 0.8835,
       "step": 8609
     },
     {
-      "epoch": 0.23646700173024635,
+      "epoch": 0.24432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.7846597866174748e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.7686375690736265e-05,
+      "loss": 1.0208,
       "step": 8610
     },
     {
-      "epoch": 0.2364944659580896,
+      "epoch": 0.24435300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.784604640047399e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.7685787738258816e-05,
+      "loss": 0.9261,
       "step": 8611
     },
     {
-      "epoch": 0.23652193018593282,
+      "epoch": 0.24438138479001134,
       "grad_norm": 0.0,
-      "learning_rate": 1.7845494872692276e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.768519972085917e-05,
+      "loss": 0.9945,
       "step": 8612
     },
     {
-      "epoch": 0.23654939441377607,
+      "epoch": 0.24440976163450626,
       "grad_norm": 0.0,
-      "learning_rate": 1.7844943282833968e-05,
-      "loss": 1.0406,
+      "learning_rate": 1.7684611638542293e-05,
+      "loss": 1.1082,
       "step": 8613
     },
     {
-      "epoch": 0.2365768586416193,
+      "epoch": 0.24443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.7844391630903428e-05,
-      "loss": 0.8812,
+      "learning_rate": 1.7684023491313155e-05,
+      "loss": 1.0361,
       "step": 8614
     },
     {
-      "epoch": 0.23660432286946254,
+      "epoch": 0.24446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.7843839916905024e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.768343527917673e-05,
+      "loss": 0.9603,
       "step": 8615
     },
     {
-      "epoch": 0.23663178709730576,
+      "epoch": 0.24449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.784328814084312e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.768284700213798e-05,
+      "loss": 0.9516,
       "step": 8616
     },
     {
-      "epoch": 0.236659251325149,
+      "epoch": 0.2445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.784273630272208e-05,
-      "loss": 1.0661,
+      "learning_rate": 1.7682258660201874e-05,
+      "loss": 1.1172,
       "step": 8617
     },
     {
-      "epoch": 0.23668671555299223,
+      "epoch": 0.2445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.7842184402546275e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.7681670253373387e-05,
+      "loss": 0.9117,
       "step": 8618
     },
     {
-      "epoch": 0.23671417978083545,
+      "epoch": 0.2445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.7841632440320065e-05,
-      "loss": 1.1659,
+      "learning_rate": 1.7681081781657485e-05,
+      "loss": 1.0685,
       "step": 8619
     },
     {
-      "epoch": 0.2367416440086787,
+      "epoch": 0.2446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.7841080416047828e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.768049324505914e-05,
+      "loss": 0.9523,
       "step": 8620
     },
     {
-      "epoch": 0.23676910823652192,
+      "epoch": 0.24463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.784052832973392e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.7679904643583322e-05,
+      "loss": 0.9841,
       "step": 8621
     },
     {
-      "epoch": 0.23679657246436517,
+      "epoch": 0.24466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.7839976181382717e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.7679315977235006e-05,
+      "loss": 1.0131,
       "step": 8622
     },
     {
-      "epoch": 0.2368240366922084,
+      "epoch": 0.24469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.7839423970998588e-05,
-      "loss": 1.022,
+      "learning_rate": 1.7678727246019162e-05,
+      "loss": 1.1353,
       "step": 8623
     },
     {
-      "epoch": 0.23685150092005164,
+      "epoch": 0.24472190692395004,
       "grad_norm": 0.0,
-      "learning_rate": 1.7838871698585897e-05,
-      "loss": 0.8335,
+      "learning_rate": 1.7678138449940764e-05,
+      "loss": 1.0485,
       "step": 8624
     },
     {
-      "epoch": 0.23687896514789486,
+      "epoch": 0.24475028376844496,
       "grad_norm": 0.0,
-      "learning_rate": 1.7838319364149015e-05,
-      "loss": 1.0894,
+      "learning_rate": 1.7677549589004786e-05,
+      "loss": 0.9159,
       "step": 8625
     },
     {
-      "epoch": 0.2369064293757381,
+      "epoch": 0.24477866061293985,
       "grad_norm": 0.0,
-      "learning_rate": 1.7837766967692317e-05,
-      "loss": 1.0352,
+      "learning_rate": 1.7676960663216197e-05,
+      "loss": 0.8851,
       "step": 8626
     },
     {
-      "epoch": 0.23693389360358133,
+      "epoch": 0.24480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.7837214509220166e-05,
-      "loss": 1.026,
+      "learning_rate": 1.7676371672579983e-05,
+      "loss": 0.9812,
       "step": 8627
     },
     {
-      "epoch": 0.23696135783142458,
+      "epoch": 0.24483541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.7836661988736943e-05,
-      "loss": 1.0619,
+      "learning_rate": 1.767578261710111e-05,
+      "loss": 0.9312,
       "step": 8628
     },
     {
-      "epoch": 0.2369888220592678,
+      "epoch": 0.2448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.783610940624701e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.7675193496784553e-05,
+      "loss": 1.0163,
       "step": 8629
     },
     {
-      "epoch": 0.23701628628711105,
+      "epoch": 0.2448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.783555676175475e-05,
-      "loss": 1.089,
+      "learning_rate": 1.7674604311635296e-05,
+      "loss": 0.969,
       "step": 8630
     },
     {
-      "epoch": 0.23704375051495427,
+      "epoch": 0.2449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.7835004055264528e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.7674015061658308e-05,
+      "loss": 1.0709,
       "step": 8631
     },
     {
-      "epoch": 0.2370712147427975,
+      "epoch": 0.2449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.7834451286780717e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.767342574685857e-05,
+      "loss": 0.9381,
       "step": 8632
     },
     {
-      "epoch": 0.23709867897064074,
+      "epoch": 0.24497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.783389845630769e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.767283636724106e-05,
+      "loss": 1.0358,
       "step": 8633
     },
     {
-      "epoch": 0.23712614319848396,
+      "epoch": 0.24500567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.783334556384983e-05,
-      "loss": 0.8358,
+      "learning_rate": 1.767224692281076e-05,
+      "loss": 1.0526,
       "step": 8634
     },
     {
-      "epoch": 0.2371536074263272,
+      "epoch": 0.24503405221339386,
       "grad_norm": 0.0,
-      "learning_rate": 1.7832792609411503e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.767165741357264e-05,
+      "loss": 1.0248,
       "step": 8635
     },
     {
-      "epoch": 0.23718107165417043,
+      "epoch": 0.24506242905788878,
       "grad_norm": 0.0,
-      "learning_rate": 1.7832239592997085e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.767106783953169e-05,
+      "loss": 0.8599,
       "step": 8636
     },
     {
-      "epoch": 0.23720853588201368,
+      "epoch": 0.24509080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.7831686514610956e-05,
-      "loss": 0.8501,
+      "learning_rate": 1.767047820069288e-05,
+      "loss": 0.955,
       "step": 8637
     },
     {
-      "epoch": 0.2372360001098569,
+      "epoch": 0.24511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.783113337425749e-05,
-      "loss": 1.0252,
+      "learning_rate": 1.76698884970612e-05,
+      "loss": 0.9895,
       "step": 8638
     },
     {
-      "epoch": 0.23726346433770015,
+      "epoch": 0.24514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7830580171941058e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.7669298728641622e-05,
+      "loss": 1.0337,
       "step": 8639
     },
     {
-      "epoch": 0.23729092856554337,
+      "epoch": 0.24517593643586832,
       "grad_norm": 0.0,
-      "learning_rate": 1.7830026907666047e-05,
-      "loss": 1.1738,
+      "learning_rate": 1.766870889543914e-05,
+      "loss": 0.9717,
       "step": 8640
     },
     {
-      "epoch": 0.23731839279338662,
+      "epoch": 0.2452043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.7829473581436828e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.7668118997458725e-05,
+      "loss": 0.9265,
       "step": 8641
     },
     {
-      "epoch": 0.23734585702122984,
+      "epoch": 0.24523269012485813,
       "grad_norm": 0.0,
-      "learning_rate": 1.7828920193257783e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.7667529034705365e-05,
+      "loss": 0.9806,
       "step": 8642
     },
     {
-      "epoch": 0.2373733212490731,
+      "epoch": 0.24526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.782836674313329e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.766693900718404e-05,
+      "loss": 0.9938,
       "step": 8643
     },
     {
-      "epoch": 0.23740078547691632,
+      "epoch": 0.2452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7827813231067726e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.766634891489974e-05,
+      "loss": 0.9636,
       "step": 8644
     },
     {
-      "epoch": 0.23742824970475956,
+      "epoch": 0.2453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.782725965706547e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.7665758757857445e-05,
+      "loss": 0.9777,
       "step": 8645
     },
     {
-      "epoch": 0.23745571393260279,
+      "epoch": 0.24534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.78267060211309e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.7665168536062142e-05,
+      "loss": 1.0368,
       "step": 8646
     },
     {
-      "epoch": 0.237483178160446,
+      "epoch": 0.24537457434733256,
       "grad_norm": 0.0,
-      "learning_rate": 1.7826152323268407e-05,
-      "loss": 1.0554,
+      "learning_rate": 1.7664578249518817e-05,
+      "loss": 0.869,
       "step": 8647
     },
     {
-      "epoch": 0.23751064238828926,
+      "epoch": 0.24540295119182748,
       "grad_norm": 0.0,
-      "learning_rate": 1.782559856348236e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.766398789823245e-05,
+      "loss": 1.0695,
       "step": 8648
     },
     {
-      "epoch": 0.23753810661613248,
+      "epoch": 0.24543132803632237,
       "grad_norm": 0.0,
-      "learning_rate": 1.782504474177715e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.7663397482208037e-05,
+      "loss": 0.9226,
       "step": 8649
     },
     {
-      "epoch": 0.23756557084397573,
+      "epoch": 0.24545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.7824490858157154e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.766280700145056e-05,
+      "loss": 0.9357,
       "step": 8650
     },
     {
-      "epoch": 0.23759303507181895,
+      "epoch": 0.24548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.7823936912626754e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.7662216455965004e-05,
+      "loss": 0.9831,
       "step": 8651
     },
     {
-      "epoch": 0.2376204992996622,
+      "epoch": 0.24551645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.7823382905190335e-05,
-      "loss": 0.9122,
+      "learning_rate": 1.7661625845756366e-05,
+      "loss": 1.1235,
       "step": 8652
     },
     {
-      "epoch": 0.23764796352750542,
+      "epoch": 0.24554483541430194,
       "grad_norm": 0.0,
-      "learning_rate": 1.7822828835852283e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.7661035170829627e-05,
+      "loss": 1.0669,
       "step": 8653
     },
     {
-      "epoch": 0.23767542775534867,
+      "epoch": 0.24557321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 1.7822274704616974e-05,
-      "loss": 1.077,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 0.9077,
       "step": 8654
     },
     {
-      "epoch": 0.2377028919831919,
+      "epoch": 0.24560158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 1.78217205114888e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.7659853626841818e-05,
+      "loss": 0.8493,
       "step": 8655
     },
     {
-      "epoch": 0.23773035621103514,
+      "epoch": 0.2456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.7821166256472142e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.7659262757790726e-05,
+      "loss": 1.0079,
       "step": 8656
     },
     {
-      "epoch": 0.23775782043887836,
+      "epoch": 0.2456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.782061193957139e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.7658671824041496e-05,
+      "loss": 0.9729,
       "step": 8657
     },
     {
-      "epoch": 0.2377852846667216,
+      "epoch": 0.24568671963677638,
       "grad_norm": 0.0,
-      "learning_rate": 1.782005756079092e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.7658080825599122e-05,
+      "loss": 1.0054,
       "step": 8658
     },
     {
-      "epoch": 0.23781274889456483,
+      "epoch": 0.2457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.7819503120135136e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.7657489762468595e-05,
+      "loss": 0.8821,
       "step": 8659
     },
     {
-      "epoch": 0.23784021312240805,
+      "epoch": 0.24574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.7818948617608406e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.7656898634654907e-05,
+      "loss": 0.997,
       "step": 8660
     },
     {
-      "epoch": 0.2378676773502513,
+      "epoch": 0.24577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.781839405321513e-05,
-      "loss": 1.0336,
+      "learning_rate": 1.765630744216305e-05,
+      "loss": 1.0369,
       "step": 8661
     },
     {
-      "epoch": 0.23789514157809452,
+      "epoch": 0.24580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.781783942695969e-05,
-      "loss": 1.057,
+      "learning_rate": 1.7655716184998025e-05,
+      "loss": 1.0188,
       "step": 8662
     },
     {
-      "epoch": 0.23792260580593777,
+      "epoch": 0.24582860385925084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7817284738846473e-05,
-      "loss": 1.088,
+      "learning_rate": 1.7655124863164817e-05,
+      "loss": 0.9552,
       "step": 8663
     },
     {
-      "epoch": 0.237950070033781,
+      "epoch": 0.24585698070374573,
       "grad_norm": 0.0,
-      "learning_rate": 1.7816729988879874e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.7654533476668427e-05,
+      "loss": 0.9536,
       "step": 8664
     },
     {
-      "epoch": 0.23797753426162424,
+      "epoch": 0.24588535754824065,
       "grad_norm": 0.0,
-      "learning_rate": 1.781617517706428e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.7653942025513847e-05,
+      "loss": 0.9857,
       "step": 8665
     },
     {
-      "epoch": 0.23800499848946746,
+      "epoch": 0.24591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.7815620303404082e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.7653350509706077e-05,
+      "loss": 0.9076,
       "step": 8666
     },
     {
-      "epoch": 0.2380324627173107,
+      "epoch": 0.24594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7815065367903664e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.7652758929250112e-05,
+      "loss": 0.9467,
       "step": 8667
     },
     {
-      "epoch": 0.23805992694515393,
+      "epoch": 0.2459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.7814510370567425e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.7652167284150944e-05,
+      "loss": 1.0393,
       "step": 8668
     },
     {
-      "epoch": 0.23808739117299718,
+      "epoch": 0.2459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.781395531139975e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.765157557441358e-05,
+      "loss": 1.0024,
       "step": 8669
     },
     {
-      "epoch": 0.2381148554008404,
+      "epoch": 0.24602724177071508,
       "grad_norm": 0.0,
-      "learning_rate": 1.7813400190405036e-05,
-      "loss": 1.0567,
+      "learning_rate": 1.765098380004301e-05,
+      "loss": 1.0684,
       "step": 8670
     },
     {
-      "epoch": 0.23814231962868365,
+      "epoch": 0.24605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.7812845007587677e-05,
-      "loss": 1.1051,
+      "learning_rate": 1.7650391961044234e-05,
+      "loss": 1.0448,
       "step": 8671
     },
     {
-      "epoch": 0.23816978385652687,
+      "epoch": 0.24608399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 1.7812289762952058e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.7649800057422256e-05,
+      "loss": 1.0056,
       "step": 8672
     },
     {
-      "epoch": 0.2381972480843701,
+      "epoch": 0.24611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.7811734456502573e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.7649208089182072e-05,
+      "loss": 0.9301,
       "step": 8673
     },
     {
-      "epoch": 0.23822471231221334,
+      "epoch": 0.24614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.781117908824362e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.7648616056328687e-05,
+      "loss": 0.8479,
       "step": 8674
     },
     {
-      "epoch": 0.23825217654005656,
+      "epoch": 0.24616912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.7810623658179598e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.7648023958867094e-05,
+      "loss": 0.8598,
       "step": 8675
     },
     {
-      "epoch": 0.23827964076789981,
+      "epoch": 0.24619750283768446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7810068166314892e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.7647431796802302e-05,
+      "loss": 0.9489,
       "step": 8676
     },
     {
-      "epoch": 0.23830710499574304,
+      "epoch": 0.24622587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 1.7809512612653902e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.764683957013931e-05,
+      "loss": 0.9719,
       "step": 8677
     },
     {
-      "epoch": 0.23833456922358628,
+      "epoch": 0.24625425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.780895699720102e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.7646247278883113e-05,
+      "loss": 0.9026,
       "step": 8678
     },
     {
-      "epoch": 0.2383620334514295,
+      "epoch": 0.24628263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 1.7808401319960648e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.7645654923038732e-05,
+      "loss": 0.913,
       "step": 8679
     },
     {
-      "epoch": 0.23838949767927275,
+      "epoch": 0.246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.7807845580937183e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.7645062502611154e-05,
+      "loss": 1.0721,
       "step": 8680
     },
     {
-      "epoch": 0.23841696190711598,
+      "epoch": 0.2463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.7807289780135015e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.764447001760539e-05,
+      "loss": 0.9429,
       "step": 8681
     },
     {
-      "epoch": 0.23844442613495923,
+      "epoch": 0.2463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.7806733917558547e-05,
-      "loss": 1.1161,
+      "learning_rate": 1.7643877468026445e-05,
+      "loss": 1.0844,
       "step": 8682
     },
     {
-      "epoch": 0.23847189036280245,
+      "epoch": 0.2463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.780617799321218e-05,
-      "loss": 0.975,
+      "learning_rate": 1.7643284853879322e-05,
+      "loss": 1.0862,
       "step": 8683
     },
     {
-      "epoch": 0.2384993545906457,
+      "epoch": 0.2464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.7805622007100302e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.764269217516903e-05,
+      "loss": 1.0695,
       "step": 8684
     },
     {
-      "epoch": 0.23852681881848892,
+      "epoch": 0.24645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.780506595922732e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.764209943190057e-05,
+      "loss": 1.0243,
       "step": 8685
     },
     {
-      "epoch": 0.23855428304633217,
+      "epoch": 0.24648127128263336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7804509849597638e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.7641506624078958e-05,
+      "loss": 1.0039,
       "step": 8686
     },
     {
-      "epoch": 0.2385817472741754,
+      "epoch": 0.24650964812712825,
       "grad_norm": 0.0,
-      "learning_rate": 1.7803953678215648e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.7640913751709196e-05,
+      "loss": 0.9589,
       "step": 8687
     },
     {
-      "epoch": 0.2386092115020186,
+      "epoch": 0.24653802497162317,
       "grad_norm": 0.0,
-      "learning_rate": 1.7803397445085756e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.764032081479629e-05,
+      "loss": 0.8382,
       "step": 8688
     },
     {
-      "epoch": 0.23863667572986186,
+      "epoch": 0.24656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.780284115021236e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.7639727813345248e-05,
+      "loss": 0.9308,
       "step": 8689
     },
     {
-      "epoch": 0.23866413995770508,
+      "epoch": 0.24659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.780228479359986e-05,
-      "loss": 0.9674,
+      "learning_rate": 1.7639134747361084e-05,
+      "loss": 0.9923,
       "step": 8690
     },
     {
-      "epoch": 0.23869160418554833,
+      "epoch": 0.24662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.7801728375252658e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.7638541616848806e-05,
+      "loss": 1.0212,
       "step": 8691
     },
     {
-      "epoch": 0.23871906841339155,
+      "epoch": 0.24665153234960271,
       "grad_norm": 0.0,
-      "learning_rate": 1.780117189517516e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.7637948421813423e-05,
+      "loss": 1.1012,
       "step": 8692
     },
     {
-      "epoch": 0.2387465326412348,
+      "epoch": 0.24667990919409763,
       "grad_norm": 0.0,
-      "learning_rate": 1.7800615353371772e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.7637355162259947e-05,
+      "loss": 0.9129,
       "step": 8693
     },
     {
-      "epoch": 0.23877399686907802,
+      "epoch": 0.24670828603859252,
       "grad_norm": 0.0,
-      "learning_rate": 1.7800058749846892e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.7636761838193386e-05,
+      "loss": 0.9568,
       "step": 8694
     },
     {
-      "epoch": 0.23880146109692127,
+      "epoch": 0.2467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7799502084604924e-05,
-      "loss": 0.9522,
+      "learning_rate": 1.7636168449618754e-05,
+      "loss": 1.0342,
       "step": 8695
     },
     {
-      "epoch": 0.2388289253247645,
+      "epoch": 0.2467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.7798945357650277e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.7635574996541066e-05,
+      "loss": 0.9588,
       "step": 8696
     },
     {
-      "epoch": 0.23885638955260774,
+      "epoch": 0.24679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.779838856898735e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.7634981478965335e-05,
+      "loss": 1.0756,
       "step": 8697
     },
     {
-      "epoch": 0.23888385378045096,
+      "epoch": 0.24682179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 1.7797831718620556e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.763438789689657e-05,
+      "loss": 0.8963,
       "step": 8698
     },
     {
-      "epoch": 0.2389113180082942,
+      "epoch": 0.24685017026106698,
       "grad_norm": 0.0,
-      "learning_rate": 1.7797274806554295e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.763379425033979e-05,
+      "loss": 1.0553,
       "step": 8699
     },
     {
-      "epoch": 0.23893878223613743,
+      "epoch": 0.24687854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 1.7796717832792977e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.7633200539300005e-05,
+      "loss": 0.8902,
       "step": 8700
     },
     {
-      "epoch": 0.23896624646398065,
+      "epoch": 0.24690692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.7796160797341005e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.7632606763782227e-05,
+      "loss": 0.9317,
       "step": 8701
     },
     {
-      "epoch": 0.2389937106918239,
+      "epoch": 0.24693530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.779560370020279e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.7632012923791482e-05,
+      "loss": 1.1197,
       "step": 8702
     },
     {
-      "epoch": 0.23902117491966712,
+      "epoch": 0.24696367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 1.7795046541382735e-05,
-      "loss": 1.0343,
+      "learning_rate": 1.7631419019332782e-05,
+      "loss": 1.0338,
       "step": 8703
     },
     {
-      "epoch": 0.23904863914751037,
+      "epoch": 0.24699205448354142,
       "grad_norm": 0.0,
-      "learning_rate": 1.7794489320885256e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.7630825050411138e-05,
+      "loss": 1.0724,
       "step": 8704
     },
     {
-      "epoch": 0.2390761033753536,
+      "epoch": 0.24702043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 1.779393203871476e-05,
-      "loss": 1.0782,
+      "learning_rate": 1.763023101703157e-05,
+      "loss": 1.0453,
       "step": 8705
     },
     {
-      "epoch": 0.23910356760319684,
+      "epoch": 0.24704880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 1.779337469487565e-05,
-      "loss": 1.1307,
+      "learning_rate": 1.7629636919199104e-05,
+      "loss": 1.0272,
       "step": 8706
     },
     {
-      "epoch": 0.23913103183104006,
+      "epoch": 0.2470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.7792817289372343e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.762904275691875e-05,
+      "loss": 1.0452,
       "step": 8707
     },
     {
-      "epoch": 0.2391584960588833,
+      "epoch": 0.247105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.7792259822209244e-05,
-      "loss": 1.0144,
+      "learning_rate": 1.7628448530195527e-05,
+      "loss": 0.9823,
       "step": 8708
     },
     {
-      "epoch": 0.23918596028672653,
+      "epoch": 0.24713393870601588,
       "grad_norm": 0.0,
-      "learning_rate": 1.7791702293390773e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.7627854239034458e-05,
+      "loss": 1.0887,
       "step": 8709
     },
     {
-      "epoch": 0.23921342451456978,
+      "epoch": 0.24716231555051077,
       "grad_norm": 0.0,
-      "learning_rate": 1.7791144702921333e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.7627259883440562e-05,
+      "loss": 1.1693,
       "step": 8710
     },
     {
-      "epoch": 0.239240888742413,
+      "epoch": 0.24719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.7790587050805338e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.7626665463418857e-05,
+      "loss": 0.8755,
       "step": 8711
     },
     {
-      "epoch": 0.23926835297025625,
+      "epoch": 0.24721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.7790029337047197e-05,
-      "loss": 0.8473,
+      "learning_rate": 1.7626070978974367e-05,
+      "loss": 0.9677,
       "step": 8712
     },
     {
-      "epoch": 0.23929581719809948,
+      "epoch": 0.24724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.778947156165133e-05,
-      "loss": 1.0273,
+      "learning_rate": 1.7625476430112113e-05,
+      "loss": 1.0793,
       "step": 8713
     },
     {
-      "epoch": 0.2393232814259427,
+      "epoch": 0.24727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.7788913724622142e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.7624881816837116e-05,
+      "loss": 1.0626,
       "step": 8714
     },
     {
-      "epoch": 0.23935074565378595,
+      "epoch": 0.24730419977298523,
       "grad_norm": 0.0,
-      "learning_rate": 1.7788355825964057e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.76242871391544e-05,
+      "loss": 0.9073,
       "step": 8715
     },
     {
-      "epoch": 0.23937820988162917,
+      "epoch": 0.24733257661748015,
       "grad_norm": 0.0,
-      "learning_rate": 1.7787797865681483e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.7623692397068985e-05,
+      "loss": 0.9958,
       "step": 8716
     },
     {
-      "epoch": 0.23940567410947242,
+      "epoch": 0.24736095346197504,
       "grad_norm": 0.0,
-      "learning_rate": 1.778723984377883e-05,
-      "loss": 0.865,
+      "learning_rate": 1.7623097590585902e-05,
+      "loss": 0.9452,
       "step": 8717
     },
     {
-      "epoch": 0.23943313833731564,
+      "epoch": 0.24738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7786681760260526e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.7622502719710168e-05,
+      "loss": 1.0641,
       "step": 8718
     },
     {
-      "epoch": 0.2394606025651589,
+      "epoch": 0.2474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.7786123615130978e-05,
-      "loss": 1.0188,
+      "learning_rate": 1.762190778444681e-05,
+      "loss": 0.9488,
       "step": 8719
     },
     {
-      "epoch": 0.2394880667930021,
+      "epoch": 0.2474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.77855654083946e-05,
-      "loss": 1.0984,
+      "learning_rate": 1.762131278480086e-05,
+      "loss": 1.0462,
       "step": 8720
     },
     {
-      "epoch": 0.23951553102084536,
+      "epoch": 0.24747446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 1.7785007140055816e-05,
-      "loss": 1.0735,
+      "learning_rate": 1.7620717720777335e-05,
+      "loss": 1.0445,
       "step": 8721
     },
     {
-      "epoch": 0.23954299524868858,
+      "epoch": 0.2475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.7784448810119038e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.762012259238126e-05,
+      "loss": 0.9208,
       "step": 8722
     },
     {
-      "epoch": 0.23957045947653183,
+      "epoch": 0.2475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.7783890418588687e-05,
-      "loss": 0.925,
+      "learning_rate": 1.7619527399617676e-05,
+      "loss": 0.9928,
       "step": 8723
     },
     {
-      "epoch": 0.23959792370437505,
+      "epoch": 0.24755959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 1.7783331965469178e-05,
-      "loss": 1.0668,
+      "learning_rate": 1.7618932142491595e-05,
+      "loss": 1.0656,
       "step": 8724
     },
     {
-      "epoch": 0.2396253879322183,
+      "epoch": 0.24758796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.7782773450764935e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.7618336821008052e-05,
+      "loss": 1.0432,
       "step": 8725
     },
     {
-      "epoch": 0.23965285216006152,
+      "epoch": 0.24761634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 1.778221487448037e-05,
-      "loss": 0.9947,
+      "learning_rate": 1.761774143517208e-05,
+      "loss": 0.9618,
       "step": 8726
     },
     {
-      "epoch": 0.23968031638790474,
+      "epoch": 0.24764472190692394,
       "grad_norm": 0.0,
-      "learning_rate": 1.778165623661991e-05,
-      "loss": 1.009,
+      "learning_rate": 1.76171459849887e-05,
+      "loss": 1.1056,
       "step": 8727
     },
     {
-      "epoch": 0.239707780615748,
+      "epoch": 0.24767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.778109753718797e-05,
-      "loss": 0.9956,
+      "learning_rate": 1.7616550470462946e-05,
+      "loss": 0.809,
       "step": 8728
     },
     {
-      "epoch": 0.2397352448435912,
+      "epoch": 0.24770147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.7780538776188973e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.7615954891599846e-05,
+      "loss": 1.0281,
       "step": 8729
     },
     {
-      "epoch": 0.23976270907143446,
+      "epoch": 0.24772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.777997995362734e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.7615359248404435e-05,
+      "loss": 1.0513,
       "step": 8730
     },
     {
-      "epoch": 0.23979017329927768,
+      "epoch": 0.24775822928490351,
       "grad_norm": 0.0,
-      "learning_rate": 1.777942106950749e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.761476354088174e-05,
+      "loss": 1.0391,
       "step": 8731
     },
     {
-      "epoch": 0.23981763752712093,
+      "epoch": 0.2477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.7778862123833845e-05,
-      "loss": 0.911,
+      "learning_rate": 1.76141677690368e-05,
+      "loss": 0.9241,
       "step": 8732
     },
     {
-      "epoch": 0.23984510175496415,
+      "epoch": 0.24781498297389332,
       "grad_norm": 0.0,
-      "learning_rate": 1.777830311661083e-05,
-      "loss": 1.042,
+      "learning_rate": 1.7613571932874638e-05,
+      "loss": 0.9768,
       "step": 8733
     },
     {
-      "epoch": 0.2398725659828074,
+      "epoch": 0.2478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.7777744047842872e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.7612976032400294e-05,
+      "loss": 0.8853,
       "step": 8734
     },
     {
-      "epoch": 0.23990003021065062,
+      "epoch": 0.2478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.7777184917534386e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.7612380067618796e-05,
+      "loss": 0.9733,
       "step": 8735
     },
     {
-      "epoch": 0.23992749443849387,
+      "epoch": 0.24790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.7776625725689803e-05,
-      "loss": 0.9187,
+      "learning_rate": 1.7611784038535184e-05,
+      "loss": 0.9558,
       "step": 8736
     },
     {
-      "epoch": 0.2399549586663371,
+      "epoch": 0.24792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.7776066472313545e-05,
-      "loss": 0.8504,
+      "learning_rate": 1.761118794515449e-05,
+      "loss": 1.0505,
       "step": 8737
     },
     {
-      "epoch": 0.23998242289418034,
+      "epoch": 0.24795686719636775,
       "grad_norm": 0.0,
-      "learning_rate": 1.7775507157410033e-05,
-      "loss": 0.9912,
+      "learning_rate": 1.761059178748175e-05,
+      "loss": 1.0649,
       "step": 8738
     },
     {
-      "epoch": 0.24000988712202356,
+      "epoch": 0.24798524404086267,
       "grad_norm": 0.0,
-      "learning_rate": 1.77749477809837e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.7609995565521997e-05,
+      "loss": 0.8604,
       "step": 8739
     },
     {
-      "epoch": 0.2400373513498668,
+      "epoch": 0.24801362088535756,
       "grad_norm": 0.0,
-      "learning_rate": 1.777438834303897e-05,
-      "loss": 0.94,
+      "learning_rate": 1.760939927928027e-05,
+      "loss": 0.9388,
       "step": 8740
     },
     {
-      "epoch": 0.24006481557771003,
+      "epoch": 0.24804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7773828843580264e-05,
-      "loss": 0.8921,
+      "learning_rate": 1.7608802928761607e-05,
+      "loss": 0.9746,
       "step": 8741
     },
     {
-      "epoch": 0.24009227980555325,
+      "epoch": 0.24807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.7773269282612014e-05,
-      "loss": 0.9956,
+      "learning_rate": 1.760820651397104e-05,
+      "loss": 0.9803,
       "step": 8742
     },
     {
-      "epoch": 0.2401197440333965,
+      "epoch": 0.24809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.7772709660138646e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.7607610034913612e-05,
+      "loss": 0.9309,
       "step": 8743
     },
     {
-      "epoch": 0.24014720826123973,
+      "epoch": 0.2481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.777214997616459e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.7607013491594363e-05,
+      "loss": 0.8969,
       "step": 8744
     },
     {
-      "epoch": 0.24017467248908297,
+      "epoch": 0.24815550510783202,
       "grad_norm": 0.0,
-      "learning_rate": 1.7771590230694272e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.7606416884018324e-05,
+      "loss": 1.037,
       "step": 8745
     },
     {
-      "epoch": 0.2402021367169262,
+      "epoch": 0.2481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.777103042373212e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.7605820212190547e-05,
+      "loss": 1.0164,
       "step": 8746
     },
     {
-      "epoch": 0.24022960094476944,
+      "epoch": 0.2482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.777047055528257e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.7605223476116056e-05,
+      "loss": 0.9858,
       "step": 8747
     },
     {
-      "epoch": 0.24025706517261267,
+      "epoch": 0.24824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 1.776991062535004e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.7604626675799906e-05,
+      "loss": 1.0216,
       "step": 8748
     },
     {
-      "epoch": 0.24028452940045592,
+      "epoch": 0.24826901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 1.7769350633938973e-05,
-      "loss": 0.907,
+      "learning_rate": 1.7604029811247132e-05,
+      "loss": 1.1393,
       "step": 8749
     },
     {
-      "epoch": 0.24031199362829914,
+      "epoch": 0.24829738933030646,
       "grad_norm": 0.0,
-      "learning_rate": 1.776879058105379e-05,
-      "loss": 1.0803,
+      "learning_rate": 1.7603432882462773e-05,
+      "loss": 1.1622,
       "step": 8750
     },
     {
-      "epoch": 0.24033945785614239,
+      "epoch": 0.24832576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.7768230466698934e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.7602835889451875e-05,
+      "loss": 0.9677,
       "step": 8751
     },
     {
-      "epoch": 0.2403669220839856,
+      "epoch": 0.24835414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.776767029087882e-05,
-      "loss": 0.957,
+      "learning_rate": 1.7602238832219486e-05,
+      "loss": 0.9548,
       "step": 8752
     },
     {
-      "epoch": 0.24039438631182886,
+      "epoch": 0.24838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.7767110053597896e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.760164171077064e-05,
+      "loss": 0.9413,
       "step": 8753
     },
     {
-      "epoch": 0.24042185053967208,
+      "epoch": 0.24841089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.7766549754860585e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.7601044525110385e-05,
+      "loss": 1.0564,
       "step": 8754
     },
     {
-      "epoch": 0.2404493147675153,
+      "epoch": 0.24843927355278092,
       "grad_norm": 0.0,
-      "learning_rate": 1.7765989394671325e-05,
-      "loss": 1.0845,
+      "learning_rate": 1.7600447275243766e-05,
+      "loss": 1.0289,
       "step": 8755
     },
     {
-      "epoch": 0.24047677899535855,
+      "epoch": 0.24846765039727584,
       "grad_norm": 0.0,
-      "learning_rate": 1.776542897303455e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.7599849961175825e-05,
+      "loss": 1.011,
       "step": 8756
     },
     {
-      "epoch": 0.24050424322320177,
+      "epoch": 0.24849602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 1.7764868489954695e-05,
-      "loss": 1.0174,
+      "learning_rate": 1.7599252582911613e-05,
+      "loss": 0.9802,
       "step": 8757
     },
     {
-      "epoch": 0.24053170745104502,
+      "epoch": 0.2485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.7764307945436188e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.759865514045617e-05,
+      "loss": 0.9528,
       "step": 8758
     },
     {
-      "epoch": 0.24055917167888824,
+      "epoch": 0.2485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.776374733948347e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.7598057633814547e-05,
+      "loss": 1.0711,
       "step": 8759
     },
     {
-      "epoch": 0.2405866359067315,
+      "epoch": 0.24858115777525538,
       "grad_norm": 0.0,
-      "learning_rate": 1.7763186672100977e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.759746006299179e-05,
+      "loss": 1.0264,
       "step": 8760
     },
     {
-      "epoch": 0.2406141001345747,
+      "epoch": 0.24860953461975027,
       "grad_norm": 0.0,
-      "learning_rate": 1.7762625943293143e-05,
-      "loss": 1.0654,
+      "learning_rate": 1.7596862427992945e-05,
+      "loss": 0.8923,
       "step": 8761
     },
     {
-      "epoch": 0.24064156436241796,
+      "epoch": 0.2486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.7762065153064405e-05,
-      "loss": 1.0334,
+      "learning_rate": 1.7596264728823063e-05,
+      "loss": 0.9815,
       "step": 8762
     },
     {
-      "epoch": 0.24066902859026118,
+      "epoch": 0.24866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.7761504301419202e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.759566696548719e-05,
+      "loss": 0.9076,
       "step": 8763
     },
     {
-      "epoch": 0.24069649281810443,
+      "epoch": 0.24869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.776094338836197e-05,
-      "loss": 1.0748,
+      "learning_rate": 1.7595069137990377e-05,
+      "loss": 0.9312,
       "step": 8764
     },
     {
-      "epoch": 0.24072395704594765,
+      "epoch": 0.24872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.776038241389715e-05,
-      "loss": 0.8842,
+      "learning_rate": 1.7594471246337676e-05,
+      "loss": 1.0101,
       "step": 8765
     },
     {
-      "epoch": 0.2407514212737909,
+      "epoch": 0.24875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.775982137802917e-05,
-      "loss": 1.0551,
+      "learning_rate": 1.7593873290534128e-05,
+      "loss": 1.0176,
       "step": 8766
     },
     {
-      "epoch": 0.24077888550163412,
+      "epoch": 0.24877979568671962,
       "grad_norm": 0.0,
-      "learning_rate": 1.7759260280762486e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.75932752705848e-05,
+      "loss": 0.9065,
       "step": 8767
     },
     {
-      "epoch": 0.24080634972947734,
+      "epoch": 0.24880817253121454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7758699122101524e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.759267718649473e-05,
+      "loss": 0.9375,
       "step": 8768
     },
     {
-      "epoch": 0.2408338139573206,
+      "epoch": 0.24883654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 1.7758137902050735e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.7592079038268975e-05,
+      "loss": 0.9883,
       "step": 8769
     },
     {
-      "epoch": 0.2408612781851638,
+      "epoch": 0.2488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.775757662061455e-05,
-      "loss": 0.9604,
+      "learning_rate": 1.7591480825912585e-05,
+      "loss": 0.9867,
       "step": 8770
     },
     {
-      "epoch": 0.24088874241300706,
+      "epoch": 0.2488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7757015277797415e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.7590882549430617e-05,
+      "loss": 0.8681,
       "step": 8771
     },
     {
-      "epoch": 0.24091620664085028,
+      "epoch": 0.2489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.7756453873603768e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.7590284208828122e-05,
+      "loss": 0.96,
       "step": 8772
     },
     {
-      "epoch": 0.24094367086869353,
+      "epoch": 0.248950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.7755892408038056e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.7589685804110157e-05,
+      "loss": 0.8923,
       "step": 8773
     },
     {
-      "epoch": 0.24097113509653675,
+      "epoch": 0.2489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.7755330881104717e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.7589087335281775e-05,
+      "loss": 1.0305,
       "step": 8774
     },
     {
-      "epoch": 0.24099859932438,
+      "epoch": 0.24900681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 1.7754769292808194e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.758848880234803e-05,
+      "loss": 1.0048,
       "step": 8775
     },
     {
-      "epoch": 0.24102606355222322,
+      "epoch": 0.24903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.7754207643152933e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.7587890205313977e-05,
+      "loss": 0.9798,
       "step": 8776
     },
     {
-      "epoch": 0.24105352778006647,
+      "epoch": 0.24906356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 1.7753645932143386e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.7587291544184678e-05,
+      "loss": 0.8959,
       "step": 8777
     },
     {
-      "epoch": 0.2410809920079097,
+      "epoch": 0.24909194097616344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7753084159783977e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.7586692818965183e-05,
+      "loss": 1.0164,
       "step": 8778
     },
     {
-      "epoch": 0.24110845623575294,
+      "epoch": 0.24912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.775252232607917e-05,
-      "loss": 1.0631,
+      "learning_rate": 1.758609402966055e-05,
+      "loss": 0.9214,
       "step": 8779
     },
     {
-      "epoch": 0.24113592046359617,
+      "epoch": 0.24914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.7751960431033402e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.758549517627585e-05,
+      "loss": 0.966,
       "step": 8780
     },
     {
-      "epoch": 0.2411633846914394,
+      "epoch": 0.24917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.775139847465112e-05,
-      "loss": 0.9396,
+      "learning_rate": 1.7584896258816122e-05,
+      "loss": 0.9548,
       "step": 8781
     },
     {
-      "epoch": 0.24119084891928264,
+      "epoch": 0.24920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.775083645693677e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.7584297277286436e-05,
+      "loss": 1.0595,
       "step": 8782
     },
     {
-      "epoch": 0.24121831314712586,
+      "epoch": 0.2492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.7750274377894802e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.7583698231691856e-05,
+      "loss": 1.0653,
       "step": 8783
     },
     {
-      "epoch": 0.2412457773749691,
+      "epoch": 0.2492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.7749712237529655e-05,
-      "loss": 1.116,
+      "learning_rate": 1.758309912203743e-05,
+      "loss": 0.9593,
       "step": 8784
     },
     {
-      "epoch": 0.24127324160281233,
+      "epoch": 0.2492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7749150035845786e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.7582499948328225e-05,
+      "loss": 1.0041,
       "step": 8785
     },
     {
-      "epoch": 0.24130070583065558,
+      "epoch": 0.2493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.7748587772847643e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.75819007105693e-05,
+      "loss": 0.9015,
       "step": 8786
     },
     {
-      "epoch": 0.2413281700584988,
+      "epoch": 0.24934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.774802544853967e-05,
-      "loss": 0.969,
+      "learning_rate": 1.7581301408765727e-05,
+      "loss": 0.9682,
       "step": 8787
     },
     {
-      "epoch": 0.24135563428634205,
+      "epoch": 0.24937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.774746306292632e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.7580702042922556e-05,
+      "loss": 0.9829,
       "step": 8788
     },
     {
-      "epoch": 0.24138309851418527,
+      "epoch": 0.24940408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.7746900616012037e-05,
-      "loss": 1.015,
+      "learning_rate": 1.7580102613044853e-05,
+      "loss": 0.9566,
       "step": 8789
     },
     {
-      "epoch": 0.24141056274202852,
+      "epoch": 0.24943246311010214,
       "grad_norm": 0.0,
-      "learning_rate": 1.7746338107801277e-05,
-      "loss": 1.0506,
+      "learning_rate": 1.7579503119137683e-05,
+      "loss": 1.055,
       "step": 8790
     },
     {
-      "epoch": 0.24143802696987174,
+      "epoch": 0.24946083995459706,
       "grad_norm": 0.0,
-      "learning_rate": 1.774577553829849e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.7578903561206108e-05,
+      "loss": 0.9541,
       "step": 8791
     },
     {
-      "epoch": 0.241465491197715,
+      "epoch": 0.24948921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 1.774521290750813e-05,
-      "loss": 1.0044,
+      "learning_rate": 1.7578303939255197e-05,
+      "loss": 0.9613,
       "step": 8792
     },
     {
-      "epoch": 0.2414929554255582,
+      "epoch": 0.24951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.774465021543464e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.7577704253290007e-05,
+      "loss": 0.9897,
       "step": 8793
     },
     {
-      "epoch": 0.24152041965340146,
+      "epoch": 0.24954597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 1.774408746208248e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.7577104503315613e-05,
+      "loss": 0.9136,
       "step": 8794
     },
     {
-      "epoch": 0.24154788388124468,
+      "epoch": 0.2495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.7743524647456097e-05,
-      "loss": 1.1086,
+      "learning_rate": 1.7576504689337074e-05,
+      "loss": 0.9767,
       "step": 8795
     },
     {
-      "epoch": 0.2415753481090879,
+      "epoch": 0.24960272417707152,
       "grad_norm": 0.0,
-      "learning_rate": 1.774296177155995e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.7575904811359463e-05,
+      "loss": 0.9122,
       "step": 8796
     },
     {
-      "epoch": 0.24160281233693115,
+      "epoch": 0.2496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.774239883439849e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.7575304869387843e-05,
+      "loss": 1.0441,
       "step": 8797
     },
     {
-      "epoch": 0.24163027656477437,
+      "epoch": 0.2496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.7741835835976175e-05,
-      "loss": 1.0578,
+      "learning_rate": 1.7574704863427277e-05,
+      "loss": 1.0068,
       "step": 8798
     },
     {
-      "epoch": 0.24165774079261762,
+      "epoch": 0.24968785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 1.7741272776297452e-05,
-      "loss": 0.9046,
+      "learning_rate": 1.7574104793482846e-05,
+      "loss": 0.9509,
       "step": 8799
     },
     {
-      "epoch": 0.24168520502046084,
+      "epoch": 0.24971623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.7740709655366778e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.7573504659559604e-05,
+      "loss": 0.9482,
       "step": 8800
     },
     {
-      "epoch": 0.2417126692483041,
+      "epoch": 0.24974460839954596,
       "grad_norm": 0.0,
-      "learning_rate": 1.7740146473188616e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.7572904461662633e-05,
+      "loss": 0.9083,
       "step": 8801
     },
     {
-      "epoch": 0.2417401334761473,
+      "epoch": 0.24977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.7739583229767413e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.7572304199796995e-05,
+      "loss": 0.9618,
       "step": 8802
     },
     {
-      "epoch": 0.24176759770399056,
+      "epoch": 0.24980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.773901992510763e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.7571703873967766e-05,
+      "loss": 0.9263,
       "step": 8803
     },
     {
-      "epoch": 0.24179506193183378,
+      "epoch": 0.24982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.7738456559213726e-05,
-      "loss": 1.123,
+      "learning_rate": 1.7571103484180008e-05,
+      "loss": 0.9186,
       "step": 8804
     },
     {
-      "epoch": 0.24182252615967703,
+      "epoch": 0.24985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.7737893132090152e-05,
-      "loss": 1.0082,
+      "learning_rate": 1.7570503030438802e-05,
+      "loss": 0.9784,
       "step": 8805
     },
     {
-      "epoch": 0.24184999038752025,
+      "epoch": 0.24988649262202042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7737329643741374e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.756990251274922e-05,
+      "loss": 0.9687,
       "step": 8806
     },
     {
-      "epoch": 0.2418774546153635,
+      "epoch": 0.2499148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.7736766094171845e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.7569301931116326e-05,
+      "loss": 0.9389,
       "step": 8807
     },
     {
-      "epoch": 0.24190491884320672,
+      "epoch": 0.24994324631101023,
       "grad_norm": 0.0,
-      "learning_rate": 1.7736202483386026e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.75687012855452e-05,
+      "loss": 0.963,
       "step": 8808
     },
     {
-      "epoch": 0.24193238307104994,
+      "epoch": 0.2499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.7735638811388375e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.7568100576040916e-05,
+      "loss": 1.0299,
       "step": 8809
     },
     {
-      "epoch": 0.2419598472988932,
+      "epoch": 0.25,
       "grad_norm": 0.0,
-      "learning_rate": 1.7735075078183354e-05,
-      "loss": 1.07,
+      "learning_rate": 1.7567499802608544e-05,
+      "loss": 0.8824,
       "step": 8810
     },
     {
-      "epoch": 0.24198731152673641,
+      "epoch": 0.2500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.7734511283775425e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.756689896525316e-05,
+      "loss": 1.0324,
       "step": 8811
     },
     {
-      "epoch": 0.24201477575457966,
+      "epoch": 0.2500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.7733947428169044e-05,
-      "loss": 0.95,
+      "learning_rate": 1.7566298063979844e-05,
+      "loss": 1.0819,
       "step": 8812
     },
     {
-      "epoch": 0.24204223998242289,
+      "epoch": 0.2500851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.7733383511368677e-05,
-      "loss": 0.916,
+      "learning_rate": 1.7565697098793663e-05,
+      "loss": 0.9228,
       "step": 8813
     },
     {
-      "epoch": 0.24206970421026613,
+      "epoch": 0.25011350737797955,
       "grad_norm": 0.0,
-      "learning_rate": 1.773281953337878e-05,
-      "loss": 1.1158,
+      "learning_rate": 1.7565096069699703e-05,
+      "loss": 1.026,
       "step": 8814
     },
     {
-      "epoch": 0.24209716843810936,
+      "epoch": 0.25014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7732255494203823e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.7564494976703034e-05,
+      "loss": 0.9896,
       "step": 8815
     },
     {
-      "epoch": 0.2421246326659526,
+      "epoch": 0.2501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.7731691393848264e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.7563893819808736e-05,
+      "loss": 1.0139,
       "step": 8816
     },
     {
-      "epoch": 0.24215209689379583,
+      "epoch": 0.25019863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.773112723231657e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.756329259902189e-05,
+      "loss": 0.8813,
       "step": 8817
     },
     {
-      "epoch": 0.24217956112163908,
+      "epoch": 0.25022701475595915,
       "grad_norm": 0.0,
-      "learning_rate": 1.77305630096132e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.7562691314347566e-05,
+      "loss": 0.9969,
       "step": 8818
     },
     {
-      "epoch": 0.2422070253494823,
+      "epoch": 0.250255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7729998725742623e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.7562089965790855e-05,
+      "loss": 0.8759,
       "step": 8819
     },
     {
-      "epoch": 0.24223448957732555,
+      "epoch": 0.25028376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.7729434380709302e-05,
-      "loss": 0.8296,
+      "learning_rate": 1.7561488553356824e-05,
+      "loss": 1.018,
       "step": 8820
     },
     {
-      "epoch": 0.24226195380516877,
+      "epoch": 0.2503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.77288699745177e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.756088707705056e-05,
+      "loss": 0.8551,
       "step": 8821
     },
     {
-      "epoch": 0.242289418033012,
+      "epoch": 0.2503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.7728305507172285e-05,
-      "loss": 1.0206,
+      "learning_rate": 1.7560285536877147e-05,
+      "loss": 0.9184,
       "step": 8822
     },
     {
-      "epoch": 0.24231688226085524,
+      "epoch": 0.2503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7727740978677523e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.755968393284166e-05,
+      "loss": 0.8664,
       "step": 8823
     },
     {
-      "epoch": 0.24234434648869846,
+      "epoch": 0.2503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.7727176389037882e-05,
-      "loss": 1.0417,
+      "learning_rate": 1.7559082264949182e-05,
+      "loss": 0.9549,
       "step": 8824
     },
     {
-      "epoch": 0.2423718107165417,
+      "epoch": 0.2504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.7726611738257827e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.75584805332048e-05,
+      "loss": 1.0728,
       "step": 8825
     },
     {
-      "epoch": 0.24239927494438493,
+      "epoch": 0.25045402951191825,
       "grad_norm": 0.0,
-      "learning_rate": 1.7726047026341828e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.755787873761359e-05,
+      "loss": 0.9756,
       "step": 8826
     },
     {
-      "epoch": 0.24242673917222818,
+      "epoch": 0.25048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.7725482253294353e-05,
-      "loss": 0.9396,
+      "learning_rate": 1.755727687818064e-05,
+      "loss": 1.0006,
       "step": 8827
     },
     {
-      "epoch": 0.2424542034000714,
+      "epoch": 0.2505107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 1.7724917419119868e-05,
-      "loss": 0.8579,
+      "learning_rate": 1.755667495491103e-05,
+      "loss": 0.9063,
       "step": 8828
     },
     {
-      "epoch": 0.24248166762791465,
+      "epoch": 0.25053916004540294,
       "grad_norm": 0.0,
-      "learning_rate": 1.7724352523822845e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.755607296780985e-05,
+      "loss": 0.9843,
       "step": 8829
     },
     {
-      "epoch": 0.24250913185575787,
+      "epoch": 0.25056753688989786,
       "grad_norm": 0.0,
-      "learning_rate": 1.7723787567407753e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.755547091688218e-05,
+      "loss": 0.9801,
       "step": 8830
     },
     {
-      "epoch": 0.24253659608360112,
+      "epoch": 0.2505959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.772322254987906e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.7554868802133106e-05,
+      "loss": 1.0599,
       "step": 8831
     },
     {
-      "epoch": 0.24256406031144434,
+      "epoch": 0.25062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.7722657471241242e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.7554266623567718e-05,
+      "loss": 1.0281,
       "step": 8832
     },
     {
-      "epoch": 0.2425915245392876,
+      "epoch": 0.25065266742338255,
       "grad_norm": 0.0,
-      "learning_rate": 1.7722092331498764e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.75536643811911e-05,
+      "loss": 0.947,
       "step": 8833
     },
     {
-      "epoch": 0.2426189887671308,
+      "epoch": 0.2506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.77215271306561e-05,
-      "loss": 0.9812,
+      "learning_rate": 1.755306207500834e-05,
+      "loss": 1.0408,
       "step": 8834
     },
     {
-      "epoch": 0.24264645299497406,
+      "epoch": 0.2507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.7720961868717724e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.7552459705024526e-05,
+      "loss": 1.0132,
       "step": 8835
     },
     {
-      "epoch": 0.24267391722281728,
+      "epoch": 0.2507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.7720396545688104e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.7551857271244744e-05,
+      "loss": 0.8492,
       "step": 8836
     },
     {
-      "epoch": 0.2427013814506605,
+      "epoch": 0.2507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.771983116157172e-05,
-      "loss": 1.1085,
+      "learning_rate": 1.7551254773674083e-05,
+      "loss": 1.0596,
       "step": 8837
     },
     {
-      "epoch": 0.24272884567850375,
+      "epoch": 0.25079455164585696,
       "grad_norm": 0.0,
-      "learning_rate": 1.771926571637304e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.7550652212317634e-05,
+      "loss": 0.9611,
       "step": 8838
     },
     {
-      "epoch": 0.24275630990634697,
+      "epoch": 0.25082292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.7718700210096538e-05,
-      "loss": 1.038,
+      "learning_rate": 1.7550049587180485e-05,
+      "loss": 0.7995,
       "step": 8839
     },
     {
-      "epoch": 0.24278377413419022,
+      "epoch": 0.2508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.7718134642746688e-05,
-      "loss": 1.011,
+      "learning_rate": 1.754944689826773e-05,
+      "loss": 0.7791,
       "step": 8840
     },
     {
-      "epoch": 0.24281123836203344,
+      "epoch": 0.25087968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.7717569014327972e-05,
-      "loss": 0.9326,
+      "learning_rate": 1.754884414558446e-05,
+      "loss": 0.9638,
       "step": 8841
     },
     {
-      "epoch": 0.2428387025898767,
+      "epoch": 0.25090805902383656,
       "grad_norm": 0.0,
-      "learning_rate": 1.771700332484486e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.754824132913576e-05,
+      "loss": 1.0816,
       "step": 8842
     },
     {
-      "epoch": 0.24286616681771991,
+      "epoch": 0.2509364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.7716437574301823e-05,
-      "loss": 0.998,
+      "learning_rate": 1.754763844892673e-05,
+      "loss": 0.9412,
       "step": 8843
     },
     {
-      "epoch": 0.24289363104556316,
+      "epoch": 0.25096481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.7715871762703348e-05,
-      "loss": 0.903,
+      "learning_rate": 1.7547035504962455e-05,
+      "loss": 0.9378,
       "step": 8844
     },
     {
-      "epoch": 0.24292109527340638,
+      "epoch": 0.25099318955732125,
       "grad_norm": 0.0,
-      "learning_rate": 1.7715305890053906e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.754643249724803e-05,
+      "loss": 1.0235,
       "step": 8845
     },
     {
-      "epoch": 0.24294855950124963,
+      "epoch": 0.2510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.7714739956357974e-05,
-      "loss": 1.055,
+      "learning_rate": 1.7545829425788555e-05,
+      "loss": 0.9556,
       "step": 8846
     },
     {
-      "epoch": 0.24297602372909285,
+      "epoch": 0.251049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.7714173961620032e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.754522629058912e-05,
+      "loss": 0.9419,
       "step": 8847
     },
     {
-      "epoch": 0.2430034879569361,
+      "epoch": 0.2510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.7713607905844557e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.7544623091654817e-05,
+      "loss": 0.9249,
       "step": 8848
     },
     {
-      "epoch": 0.24303095218477933,
+      "epoch": 0.2511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.771304178903603e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.7544019828990744e-05,
+      "loss": 1.006,
       "step": 8849
     },
     {
-      "epoch": 0.24305841641262255,
+      "epoch": 0.2511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.7712475611198922e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.7543416502601994e-05,
+      "loss": 1.0126,
       "step": 8850
     },
     {
-      "epoch": 0.2430858806404658,
+      "epoch": 0.2511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.7711909372337727e-05,
-      "loss": 1.0919,
+      "learning_rate": 1.754281311249367e-05,
+      "loss": 0.977,
       "step": 8851
     },
     {
-      "epoch": 0.24311334486830902,
+      "epoch": 0.2511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.7711343072456915e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.7542209658670862e-05,
+      "loss": 0.9639,
       "step": 8852
     },
     {
-      "epoch": 0.24314080909615227,
+      "epoch": 0.25122020431328035,
       "grad_norm": 0.0,
-      "learning_rate": 1.7710776711560972e-05,
-      "loss": 1.021,
+      "learning_rate": 1.7541606141138672e-05,
+      "loss": 1.0082,
       "step": 8853
     },
     {
-      "epoch": 0.2431682733239955,
+      "epoch": 0.25124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.771021028965437e-05,
-      "loss": 0.9882,
+      "learning_rate": 1.754100255990219e-05,
+      "loss": 1.0636,
       "step": 8854
     },
     {
-      "epoch": 0.24319573755183874,
+      "epoch": 0.2512769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.7709643806741604e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.7540398914966528e-05,
+      "loss": 1.0366,
       "step": 8855
     },
     {
-      "epoch": 0.24322320177968196,
+      "epoch": 0.25130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.7709077262827147e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.7539795206336773e-05,
+      "loss": 1.0555,
       "step": 8856
     },
     {
-      "epoch": 0.2432506660075252,
+      "epoch": 0.25133371169125995,
       "grad_norm": 0.0,
-      "learning_rate": 1.7708510657915487e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.7539191434018034e-05,
+      "loss": 0.9541,
       "step": 8857
     },
     {
-      "epoch": 0.24327813023536843,
+      "epoch": 0.2513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.77079439920111e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.75385875980154e-05,
+      "loss": 0.8282,
       "step": 8858
     },
     {
-      "epoch": 0.24330559446321168,
+      "epoch": 0.25139046538024973,
       "grad_norm": 0.0,
-      "learning_rate": 1.7707377265118476e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.753798369833398e-05,
+      "loss": 1.0555,
       "step": 8859
     },
     {
-      "epoch": 0.2433330586910549,
+      "epoch": 0.2514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7706810477242102e-05,
-      "loss": 1.004,
+      "learning_rate": 1.7537379734978876e-05,
+      "loss": 1.0468,
       "step": 8860
     },
     {
-      "epoch": 0.24336052291889815,
+      "epoch": 0.2514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.7706243628386453e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.7536775707955182e-05,
+      "loss": 1.0359,
       "step": 8861
     },
     {
-      "epoch": 0.24338798714674137,
+      "epoch": 0.2514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7705676718556023e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.7536171617268004e-05,
+      "loss": 0.8684,
       "step": 8862
     },
     {
-      "epoch": 0.2434154513745846,
+      "epoch": 0.2515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.770510974775529e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.753556746292245e-05,
+      "loss": 0.9166,
       "step": 8863
     },
     {
-      "epoch": 0.24344291560242784,
+      "epoch": 0.2515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.7704542715988745e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.7534963244923618e-05,
+      "loss": 1.0201,
       "step": 8864
     },
     {
-      "epoch": 0.24347037983027106,
+      "epoch": 0.25156072644721905,
       "grad_norm": 0.0,
-      "learning_rate": 1.7703975623260876e-05,
-      "loss": 1.002,
+      "learning_rate": 1.7534358963276606e-05,
+      "loss": 1.0473,
       "step": 8865
     },
     {
-      "epoch": 0.2434978440581143,
+      "epoch": 0.25158910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 1.7703408469576166e-05,
-      "loss": 0.8996,
+      "learning_rate": 1.7533754617986533e-05,
+      "loss": 0.9533,
       "step": 8866
     },
     {
-      "epoch": 0.24352530828595753,
+      "epoch": 0.2516174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 1.7702841254939107e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.7533150209058492e-05,
+      "loss": 1.0204,
       "step": 8867
     },
     {
-      "epoch": 0.24355277251380078,
+      "epoch": 0.25164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.770227397935418e-05,
-      "loss": 1.01,
+      "learning_rate": 1.7532545736497596e-05,
+      "loss": 1.0272,
       "step": 8868
     },
     {
-      "epoch": 0.243580236741644,
+      "epoch": 0.25167423382519866,
       "grad_norm": 0.0,
-      "learning_rate": 1.770170664282588e-05,
-      "loss": 0.959,
+      "learning_rate": 1.7531941200308946e-05,
+      "loss": 0.9093,
       "step": 8869
     },
     {
-      "epoch": 0.24360770096948725,
+      "epoch": 0.2517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.770113924535869e-05,
-      "loss": 1.0823,
+      "learning_rate": 1.753133660049765e-05,
+      "loss": 0.988,
       "step": 8870
     },
     {
-      "epoch": 0.24363516519733047,
+      "epoch": 0.25173098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.7700571786957107e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.7530731937068815e-05,
+      "loss": 0.9501,
       "step": 8871
     },
     {
-      "epoch": 0.24366262942517372,
+      "epoch": 0.2517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.770000426762562e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.753012721002755e-05,
+      "loss": 0.9921,
       "step": 8872
     },
     {
-      "epoch": 0.24369009365301694,
+      "epoch": 0.2517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.769943668736871e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.752952241937896e-05,
+      "loss": 0.8772,
       "step": 8873
     },
     {
-      "epoch": 0.2437175578808602,
+      "epoch": 0.2518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.7698869046190878e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.7528917565128155e-05,
+      "loss": 0.8755,
       "step": 8874
     },
     {
-      "epoch": 0.2437450221087034,
+      "epoch": 0.251844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.769830134409661e-05,
-      "loss": 0.9842,
+      "learning_rate": 1.752831264728025e-05,
+      "loss": 0.9656,
       "step": 8875
     },
     {
-      "epoch": 0.24377248633654663,
+      "epoch": 0.2518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.76977335810904e-05,
-      "loss": 1.0037,
+      "learning_rate": 1.7527707665840346e-05,
+      "loss": 0.947,
       "step": 8876
     },
     {
-      "epoch": 0.24379995056438988,
+      "epoch": 0.25190124858115776,
       "grad_norm": 0.0,
-      "learning_rate": 1.7697165757176743e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.7527102620813553e-05,
+      "loss": 1.0178,
       "step": 8877
     },
     {
-      "epoch": 0.2438274147922331,
+      "epoch": 0.25192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.7696597872360128e-05,
-      "loss": 1.0252,
+      "learning_rate": 1.752649751220499e-05,
+      "loss": 0.8695,
       "step": 8878
     },
     {
-      "epoch": 0.24385487902007635,
+      "epoch": 0.2519580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.769602992664505e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.7525892340019766e-05,
+      "loss": 0.9932,
       "step": 8879
     },
     {
-      "epoch": 0.24388234324791958,
+      "epoch": 0.25198637911464244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7695461920036e-05,
-      "loss": 0.9747,
+      "learning_rate": 1.7525287104262987e-05,
+      "loss": 0.8247,
       "step": 8880
     },
     {
-      "epoch": 0.24390980747576282,
+      "epoch": 0.25201475595913736,
       "grad_norm": 0.0,
-      "learning_rate": 1.7694893852537472e-05,
-      "loss": 1.0779,
+      "learning_rate": 1.752468180493977e-05,
+      "loss": 1.0127,
       "step": 8881
     },
     {
-      "epoch": 0.24393727170360605,
+      "epoch": 0.2520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.769432572415397e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.752407644205523e-05,
+      "loss": 1.0518,
       "step": 8882
     },
     {
-      "epoch": 0.2439647359314493,
+      "epoch": 0.25207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.7693757534889975e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.752347101561448e-05,
+      "loss": 0.9966,
       "step": 8883
     },
     {
-      "epoch": 0.24399220015929252,
+      "epoch": 0.252099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.7693189284749992e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.7522865525622628e-05,
+      "loss": 1.0996,
       "step": 8884
     },
     {
-      "epoch": 0.24401966438713577,
+      "epoch": 0.2521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.7692620973738517e-05,
-      "loss": 1.057,
+      "learning_rate": 1.7522259972084793e-05,
+      "loss": 0.8813,
       "step": 8885
     },
     {
-      "epoch": 0.244047128614979,
+      "epoch": 0.2521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.7692052601860045e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.7521654355006094e-05,
+      "loss": 0.9245,
       "step": 8886
     },
     {
-      "epoch": 0.24407459284282224,
+      "epoch": 0.2521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.7691484169119074e-05,
-      "loss": 1.068,
+      "learning_rate": 1.752104867439164e-05,
+      "loss": 0.9632,
       "step": 8887
     },
     {
-      "epoch": 0.24410205707066546,
+      "epoch": 0.2522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.7690915675520097e-05,
-      "loss": 0.8824,
+      "learning_rate": 1.7520442930246546e-05,
+      "loss": 0.9857,
       "step": 8888
     },
     {
-      "epoch": 0.2441295212985087,
+      "epoch": 0.25224177071509646,
       "grad_norm": 0.0,
-      "learning_rate": 1.7690347121067623e-05,
-      "loss": 1.0272,
+      "learning_rate": 1.7519837122575936e-05,
+      "loss": 1.0023,
       "step": 8889
     },
     {
-      "epoch": 0.24415698552635193,
+      "epoch": 0.2522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.7689778505766135e-05,
-      "loss": 1.0143,
+      "learning_rate": 1.7519231251384923e-05,
+      "loss": 1.054,
       "step": 8890
     },
     {
-      "epoch": 0.24418444975419515,
+      "epoch": 0.2522985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.7689209829620147e-05,
-      "loss": 1.1416,
+      "learning_rate": 1.751862531667863e-05,
+      "loss": 0.9589,
       "step": 8891
     },
     {
-      "epoch": 0.2442119139820384,
+      "epoch": 0.25232690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.768864109263415e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.751801931846216e-05,
+      "loss": 1.0415,
       "step": 8892
     },
     {
-      "epoch": 0.24423937820988162,
+      "epoch": 0.25235527809307606,
       "grad_norm": 0.0,
-      "learning_rate": 1.7688072294812647e-05,
-      "loss": 1.1445,
+      "learning_rate": 1.751741325674065e-05,
+      "loss": 1.097,
       "step": 8893
     },
     {
-      "epoch": 0.24426684243772487,
+      "epoch": 0.2523836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 1.7687503436160136e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.7516807131519216e-05,
+      "loss": 0.989,
       "step": 8894
     },
     {
-      "epoch": 0.2442943066655681,
+      "epoch": 0.25241203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.7686934516681124e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.751620094280297e-05,
+      "loss": 0.9121,
       "step": 8895
     },
     {
-      "epoch": 0.24432177089341134,
+      "epoch": 0.25244040862656075,
       "grad_norm": 0.0,
-      "learning_rate": 1.7686365536380102e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.7515594690597037e-05,
+      "loss": 0.9318,
       "step": 8896
     },
     {
-      "epoch": 0.24434923512125456,
+      "epoch": 0.2524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.7685796495261586e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.7514988374906535e-05,
+      "loss": 0.9277,
       "step": 8897
     },
     {
-      "epoch": 0.2443766993490978,
+      "epoch": 0.2524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.7685227393330064e-05,
-      "loss": 1.0656,
+      "learning_rate": 1.7514381995736592e-05,
+      "loss": 0.9368,
       "step": 8898
     },
     {
-      "epoch": 0.24440416357694103,
+      "epoch": 0.2525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.768465823059005e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.7513775553092328e-05,
+      "loss": 0.9211,
       "step": 8899
     },
     {
-      "epoch": 0.24443162780478428,
+      "epoch": 0.2525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.7684089007046042e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.7513169046978862e-05,
+      "loss": 0.978,
       "step": 8900
     },
     {
-      "epoch": 0.2444590920326275,
+      "epoch": 0.25258229284903516,
       "grad_norm": 0.0,
-      "learning_rate": 1.7683519722702547e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.7512562477401317e-05,
+      "loss": 0.9907,
       "step": 8901
     },
     {
-      "epoch": 0.24448655626047075,
+      "epoch": 0.2526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.768295037756407e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.7511955844364822e-05,
+      "loss": 1.0281,
       "step": 8902
     },
     {
-      "epoch": 0.24451402048831397,
+      "epoch": 0.252639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.7682380971635106e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.75113491478745e-05,
+      "loss": 0.9802,
       "step": 8903
     },
     {
-      "epoch": 0.2445414847161572,
+      "epoch": 0.25266742338251985,
       "grad_norm": 0.0,
-      "learning_rate": 1.7681811504920173e-05,
-      "loss": 0.8846,
+      "learning_rate": 1.7510742387935472e-05,
+      "loss": 0.9885,
       "step": 8904
     },
     {
-      "epoch": 0.24456894894400044,
+      "epoch": 0.25269580022701477,
       "grad_norm": 0.0,
-      "learning_rate": 1.768124197742377e-05,
-      "loss": 1.0702,
+      "learning_rate": 1.7510135564552863e-05,
+      "loss": 0.9972,
       "step": 8905
     },
     {
-      "epoch": 0.24459641317184366,
+      "epoch": 0.2527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.7680672389150407e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.7509528677731806e-05,
+      "loss": 1.0062,
       "step": 8906
     },
     {
-      "epoch": 0.2446238773996869,
+      "epoch": 0.25275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7680102740104588e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.750892172747742e-05,
+      "loss": 0.9767,
       "step": 8907
     },
     {
-      "epoch": 0.24465134162753013,
+      "epoch": 0.25278093076049946,
       "grad_norm": 0.0,
-      "learning_rate": 1.767953303029082e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.7508314713794835e-05,
+      "loss": 1.0803,
       "step": 8908
     },
     {
-      "epoch": 0.24467880585537338,
+      "epoch": 0.2528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.767896325971361e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.750770763668918e-05,
+      "loss": 0.9093,
       "step": 8909
     },
     {
-      "epoch": 0.2447062700832166,
+      "epoch": 0.25283768444948923,
       "grad_norm": 0.0,
-      "learning_rate": 1.767839342837747e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.7507100496165575e-05,
+      "loss": 0.9785,
       "step": 8910
     },
     {
-      "epoch": 0.24473373431105985,
+      "epoch": 0.2528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.7677823536286908e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.750649329222916e-05,
+      "loss": 1.0118,
       "step": 8911
     },
     {
-      "epoch": 0.24476119853890307,
+      "epoch": 0.252894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7677253583446434e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.7505886024885055e-05,
+      "loss": 0.8594,
       "step": 8912
     },
     {
-      "epoch": 0.24478866276674632,
+      "epoch": 0.2529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.7676683569860555e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.7505278694138397e-05,
+      "loss": 1.0207,
       "step": 8913
     },
     {
-      "epoch": 0.24481612699458954,
+      "epoch": 0.2529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.767611349553378e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.750467129999431e-05,
+      "loss": 1.0447,
       "step": 8914
     },
     {
-      "epoch": 0.2448435912224328,
+      "epoch": 0.2529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.767554336047062e-05,
-      "loss": 1.1203,
+      "learning_rate": 1.750406384245793e-05,
+      "loss": 0.9709,
       "step": 8915
     },
     {
-      "epoch": 0.24487105545027602,
+      "epoch": 0.25300794551645855,
       "grad_norm": 0.0,
-      "learning_rate": 1.767497316467559e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.7503456321534383e-05,
+      "loss": 0.9977,
       "step": 8916
     },
     {
-      "epoch": 0.24489851967811924,
+      "epoch": 0.25303632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.7674402908153198e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.7502848737228803e-05,
+      "loss": 1.029,
       "step": 8917
     },
     {
-      "epoch": 0.24492598390596249,
+      "epoch": 0.25306469920544833,
       "grad_norm": 0.0,
-      "learning_rate": 1.767383259090796e-05,
-      "loss": 1.011,
+      "learning_rate": 1.7502241089546323e-05,
+      "loss": 1.0437,
       "step": 8918
     },
     {
-      "epoch": 0.2449534481338057,
+      "epoch": 0.25309307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.7673262212944383e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.7501633378492077e-05,
+      "loss": 1.0286,
       "step": 8919
     },
     {
-      "epoch": 0.24498091236164896,
+      "epoch": 0.25312145289443816,
       "grad_norm": 0.0,
-      "learning_rate": 1.7672691774266986e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.7501025604071195e-05,
+      "loss": 0.8714,
       "step": 8920
     },
     {
-      "epoch": 0.24500837658949218,
+      "epoch": 0.253149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.7672121274880278e-05,
-      "loss": 1.1003,
+      "learning_rate": 1.7500417766288812e-05,
+      "loss": 0.9651,
       "step": 8921
     },
     {
-      "epoch": 0.24503584081733543,
+      "epoch": 0.25317820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.7671550714788775e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.7499809865150063e-05,
+      "loss": 0.9512,
       "step": 8922
     },
     {
-      "epoch": 0.24506330504517865,
+      "epoch": 0.2532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.7670980093996987e-05,
-      "loss": 1.0459,
+      "learning_rate": 1.7499201900660083e-05,
+      "loss": 0.9521,
       "step": 8923
     },
     {
-      "epoch": 0.2450907692730219,
+      "epoch": 0.2532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.7670409412509437e-05,
-      "loss": 0.9029,
+      "learning_rate": 1.749859387282401e-05,
+      "loss": 1.0243,
       "step": 8924
     },
     {
-      "epoch": 0.24511823350086512,
+      "epoch": 0.2532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.7669838670330638e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.7497985781646974e-05,
+      "loss": 1.0199,
       "step": 8925
     },
     {
-      "epoch": 0.24514569772870837,
+      "epoch": 0.2532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.76692678674651e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.7497377627134117e-05,
+      "loss": 0.9519,
       "step": 8926
     },
     {
-      "epoch": 0.2451731619565516,
+      "epoch": 0.2533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.766869700391735e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.7496769409290572e-05,
+      "loss": 0.9446,
       "step": 8927
     },
     {
-      "epoch": 0.24520062618439484,
+      "epoch": 0.25334846765039726,
       "grad_norm": 0.0,
-      "learning_rate": 1.766812607969189e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.7496161128121482e-05,
+      "loss": 1.0608,
       "step": 8928
     },
     {
-      "epoch": 0.24522809041223806,
+      "epoch": 0.2533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.7667555094793253e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.749555278363198e-05,
+      "loss": 0.9379,
       "step": 8929
     },
     {
-      "epoch": 0.24525555464008128,
+      "epoch": 0.2534052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.766698404922595e-05,
-      "loss": 0.949,
+      "learning_rate": 1.749494437582721e-05,
+      "loss": 0.8561,
       "step": 8930
     },
     {
-      "epoch": 0.24528301886792453,
+      "epoch": 0.25343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.7666412942994496e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.74943359047123e-05,
+      "loss": 0.9733,
       "step": 8931
     },
     {
-      "epoch": 0.24531048309576775,
+      "epoch": 0.25346197502837686,
       "grad_norm": 0.0,
-      "learning_rate": 1.7665841776103416e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.7493727370292405e-05,
+      "loss": 0.9107,
       "step": 8932
     },
     {
-      "epoch": 0.245337947323611,
+      "epoch": 0.2534903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.7665270548557225e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.7493118772572655e-05,
+      "loss": 0.8714,
       "step": 8933
     },
     {
-      "epoch": 0.24536541155145422,
+      "epoch": 0.25351872871736664,
       "grad_norm": 0.0,
-      "learning_rate": 1.7664699260360443e-05,
-      "loss": 1.0518,
+      "learning_rate": 1.7492510111558196e-05,
+      "loss": 0.9196,
       "step": 8934
     },
     {
-      "epoch": 0.24539287577929747,
+      "epoch": 0.2535471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.7664127911517596e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.7491901387254163e-05,
+      "loss": 0.9586,
       "step": 8935
     },
     {
-      "epoch": 0.2454203400071407,
+      "epoch": 0.2535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.7663556502033193e-05,
-      "loss": 1.0302,
+      "learning_rate": 1.7491292599665704e-05,
+      "loss": 1.0678,
       "step": 8936
     },
     {
-      "epoch": 0.24544780423498394,
+      "epoch": 0.2536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.766298503191177e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.7490683748797964e-05,
+      "loss": 0.8449,
       "step": 8937
     },
     {
-      "epoch": 0.24547526846282716,
+      "epoch": 0.2536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.7662413501157836e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.7490074834656077e-05,
+      "loss": 0.9202,
       "step": 8938
     },
     {
-      "epoch": 0.2455027326906704,
+      "epoch": 0.2536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.766184190977592e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.7489465857245193e-05,
+      "loss": 0.8793,
       "step": 8939
     },
     {
-      "epoch": 0.24553019691851363,
+      "epoch": 0.25368898978433596,
       "grad_norm": 0.0,
-      "learning_rate": 1.7661270257770538e-05,
-      "loss": 1.0077,
+      "learning_rate": 1.7488856816570455e-05,
+      "loss": 0.9125,
       "step": 8940
     },
     {
-      "epoch": 0.24555766114635688,
+      "epoch": 0.2537173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.7660698545146227e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.7488247712637006e-05,
+      "loss": 1.0054,
       "step": 8941
     },
     {
-      "epoch": 0.2455851253742001,
+      "epoch": 0.2537457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.7660126771907496e-05,
-      "loss": 0.9557,
+      "learning_rate": 1.7487638545449993e-05,
+      "loss": 0.8951,
       "step": 8942
     },
     {
-      "epoch": 0.24561258960204335,
+      "epoch": 0.25377412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.7659554938058872e-05,
-      "loss": 0.983,
+      "learning_rate": 1.7487029315014558e-05,
+      "loss": 0.9027,
       "step": 8943
     },
     {
-      "epoch": 0.24564005382988657,
+      "epoch": 0.25380249716231557,
       "grad_norm": 0.0,
-      "learning_rate": 1.765898304360489e-05,
-      "loss": 1.0977,
+      "learning_rate": 1.748642002133585e-05,
+      "loss": 0.9996,
       "step": 8944
     },
     {
-      "epoch": 0.2456675180577298,
+      "epoch": 0.2538308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.7658411088550065e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.7485810664419015e-05,
+      "loss": 1.0469,
       "step": 8945
     },
     {
-      "epoch": 0.24569498228557304,
+      "epoch": 0.25385925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.765783907289892e-05,
-      "loss": 0.9454,
+      "learning_rate": 1.7485201244269204e-05,
+      "loss": 1.0377,
       "step": 8946
     },
     {
-      "epoch": 0.24572244651341626,
+      "epoch": 0.2538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.765726699665599e-05,
-      "loss": 0.981,
+      "learning_rate": 1.748459176089156e-05,
+      "loss": 1.009,
       "step": 8947
     },
     {
-      "epoch": 0.24574991074125951,
+      "epoch": 0.2539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.7656694859825794e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.7483982214291233e-05,
+      "loss": 0.961,
       "step": 8948
     },
     {
-      "epoch": 0.24577737496910274,
+      "epoch": 0.25394438138479003,
       "grad_norm": 0.0,
-      "learning_rate": 1.765612266241287e-05,
-      "loss": 1.141,
+      "learning_rate": 1.7483372604473373e-05,
+      "loss": 0.9069,
       "step": 8949
     },
     {
-      "epoch": 0.24580483919694598,
+      "epoch": 0.2539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.765555040442173e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.7482762931443125e-05,
+      "loss": 0.9093,
       "step": 8950
     },
     {
-      "epoch": 0.2458323034247892,
+      "epoch": 0.2540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.7654978085856912e-05,
-      "loss": 0.8987,
+      "learning_rate": 1.7482153195205646e-05,
+      "loss": 0.919,
       "step": 8951
     },
     {
-      "epoch": 0.24585976765263245,
+      "epoch": 0.25402951191827466,
       "grad_norm": 0.0,
-      "learning_rate": 1.7654405706722944e-05,
-      "loss": 1.1206,
+      "learning_rate": 1.7481543395766078e-05,
+      "loss": 0.9918,
       "step": 8952
     },
     {
-      "epoch": 0.24588723188047568,
+      "epoch": 0.2540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.7653833267024352e-05,
-      "loss": 1.0597,
+      "learning_rate": 1.7480933533129582e-05,
+      "loss": 0.9447,
       "step": 8953
     },
     {
-      "epoch": 0.24591469610831893,
+      "epoch": 0.2540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.7653260766765665e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.74803236073013e-05,
+      "loss": 0.8662,
       "step": 8954
     },
     {
-      "epoch": 0.24594216033616215,
+      "epoch": 0.25411464245175935,
       "grad_norm": 0.0,
-      "learning_rate": 1.7652688205951414e-05,
-      "loss": 0.912,
+      "learning_rate": 1.7479713618286388e-05,
+      "loss": 0.9032,
       "step": 8955
     },
     {
-      "epoch": 0.2459696245640054,
+      "epoch": 0.25414301929625427,
       "grad_norm": 0.0,
-      "learning_rate": 1.7652115584586132e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.7479103566090003e-05,
+      "loss": 0.9465,
       "step": 8956
     },
     {
-      "epoch": 0.24599708879184862,
+      "epoch": 0.25417139614074913,
       "grad_norm": 0.0,
-      "learning_rate": 1.7651542902674345e-05,
-      "loss": 1.1075,
+      "learning_rate": 1.747849345071729e-05,
+      "loss": 0.9562,
       "step": 8957
     },
     {
-      "epoch": 0.24602455301969184,
+      "epoch": 0.25419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.7650970160220587e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.7477883272173408e-05,
+      "loss": 0.9279,
       "step": 8958
     },
     {
-      "epoch": 0.2460520172475351,
+      "epoch": 0.25422814982973896,
       "grad_norm": 0.0,
-      "learning_rate": 1.765039735722939e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.7477273030463512e-05,
+      "loss": 0.8417,
       "step": 8959
     },
     {
-      "epoch": 0.2460794814753783,
+      "epoch": 0.2542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.7649824493705287e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.747666272559275e-05,
+      "loss": 0.8952,
       "step": 8960
     },
     {
-      "epoch": 0.24610694570322156,
+      "epoch": 0.25428490351872873,
       "grad_norm": 0.0,
-      "learning_rate": 1.7649251569652808e-05,
-      "loss": 0.9931,
+      "learning_rate": 1.7476052357566287e-05,
+      "loss": 0.9753,
       "step": 8961
     },
     {
-      "epoch": 0.24613440993106478,
+      "epoch": 0.2543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.7648678585076488e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.747544192638927e-05,
+      "loss": 0.9742,
       "step": 8962
     },
     {
-      "epoch": 0.24616187415890803,
+      "epoch": 0.2543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.764810553998086e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.747483143206686e-05,
+      "loss": 1.0278,
       "step": 8963
     },
     {
-      "epoch": 0.24618933838675125,
+      "epoch": 0.25437003405221337,
       "grad_norm": 0.0,
-      "learning_rate": 1.7647532434370456e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.7474220874604213e-05,
+      "loss": 1.0451,
       "step": 8964
     },
     {
-      "epoch": 0.2462168026145945,
+      "epoch": 0.2543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.764695926824982e-05,
-      "loss": 0.9103,
+      "learning_rate": 1.7473610254006484e-05,
+      "loss": 0.9789,
       "step": 8965
     },
     {
-      "epoch": 0.24624426684243772,
+      "epoch": 0.2544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.7646386041623473e-05,
-      "loss": 0.9098,
+      "learning_rate": 1.7472999570278837e-05,
+      "loss": 0.8198,
       "step": 8966
     },
     {
-      "epoch": 0.24627173107028097,
+      "epoch": 0.25445516458569806,
       "grad_norm": 0.0,
-      "learning_rate": 1.764581275449596e-05,
-      "loss": 1.0939,
+      "learning_rate": 1.7472388823426426e-05,
+      "loss": 1.056,
       "step": 8967
     },
     {
-      "epoch": 0.2462991952981242,
+      "epoch": 0.25448354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.7645239406871814e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.747177801345441e-05,
+      "loss": 0.9493,
       "step": 8968
     },
     {
-      "epoch": 0.24632665952596744,
+      "epoch": 0.25451191827468783,
       "grad_norm": 0.0,
-      "learning_rate": 1.7644665998755574e-05,
-      "loss": 0.8545,
+      "learning_rate": 1.7471167140367948e-05,
+      "loss": 0.8911,
       "step": 8969
     },
     {
-      "epoch": 0.24635412375381066,
+      "epoch": 0.25454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.7644092530151776e-05,
-      "loss": 0.963,
+      "learning_rate": 1.7470556204172204e-05,
+      "loss": 0.9714,
       "step": 8970
     },
     {
-      "epoch": 0.24638158798165388,
+      "epoch": 0.25456867196367766,
       "grad_norm": 0.0,
-      "learning_rate": 1.7643519001064954e-05,
-      "loss": 1.1061,
+      "learning_rate": 1.7469945204872333e-05,
+      "loss": 0.9906,
       "step": 8971
     },
     {
-      "epoch": 0.24640905220949713,
+      "epoch": 0.2545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.764294541149965e-05,
-      "loss": 0.9281,
+      "learning_rate": 1.7469334142473502e-05,
+      "loss": 0.981,
       "step": 8972
     },
     {
-      "epoch": 0.24643651643734035,
+      "epoch": 0.25462542565266744,
       "grad_norm": 0.0,
-      "learning_rate": 1.7642371761460403e-05,
-      "loss": 1.0346,
+      "learning_rate": 1.7468723016980866e-05,
+      "loss": 1.0119,
       "step": 8973
     },
     {
-      "epoch": 0.2464639806651836,
+      "epoch": 0.2546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.7641798050951745e-05,
-      "loss": 0.8797,
+      "learning_rate": 1.7468111828399594e-05,
+      "loss": 0.9625,
       "step": 8974
     },
     {
-      "epoch": 0.24649144489302682,
+      "epoch": 0.2546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.7641224279978225e-05,
-      "loss": 1.09,
+      "learning_rate": 1.7467500576734842e-05,
+      "loss": 1.0379,
       "step": 8975
     },
     {
-      "epoch": 0.24651890912087007,
+      "epoch": 0.2547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.764065044854438e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.746688926199178e-05,
+      "loss": 0.9909,
       "step": 8976
     },
     {
-      "epoch": 0.2465463733487133,
+      "epoch": 0.254738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.7640076556654747e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.7466277884175572e-05,
+      "loss": 1.0437,
       "step": 8977
     },
     {
-      "epoch": 0.24657383757655654,
+      "epoch": 0.2547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.763950260431387e-05,
-      "loss": 0.8662,
+      "learning_rate": 1.7465666443291374e-05,
+      "loss": 1.1022,
       "step": 8978
     },
     {
-      "epoch": 0.24660130180439976,
+      "epoch": 0.25479568671963676,
       "grad_norm": 0.0,
-      "learning_rate": 1.7638928591526287e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.7465054939344357e-05,
+      "loss": 0.9736,
       "step": 8979
     },
     {
-      "epoch": 0.246628766032243,
+      "epoch": 0.2548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.7638354518296544e-05,
-      "loss": 1.0248,
+      "learning_rate": 1.7464443372339688e-05,
+      "loss": 0.9988,
       "step": 8980
     },
     {
-      "epoch": 0.24665623026008623,
+      "epoch": 0.25485244040862653,
       "grad_norm": 0.0,
-      "learning_rate": 1.763778038462918e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.7463831742282526e-05,
+      "loss": 0.8109,
       "step": 8981
     },
     {
-      "epoch": 0.24668369448792948,
+      "epoch": 0.25488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.7637206190528736e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.7463220049178046e-05,
+      "loss": 0.9764,
       "step": 8982
     },
     {
-      "epoch": 0.2467111587157727,
+      "epoch": 0.25490919409761637,
       "grad_norm": 0.0,
-      "learning_rate": 1.7636631935999763e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.7462608293031407e-05,
+      "loss": 1.0772,
       "step": 8983
     },
     {
-      "epoch": 0.24673862294361595,
+      "epoch": 0.2549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.76360576210468e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.7461996473847783e-05,
+      "loss": 0.8696,
       "step": 8984
     },
     {
-      "epoch": 0.24676608717145918,
+      "epoch": 0.25496594778660614,
       "grad_norm": 0.0,
-      "learning_rate": 1.763548324567439e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.7461384591632335e-05,
+      "loss": 1.0001,
       "step": 8985
     },
     {
-      "epoch": 0.2467935513993024,
+      "epoch": 0.254994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.7634908809887078e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.7460772646390243e-05,
+      "loss": 1.0035,
       "step": 8986
     },
     {
-      "epoch": 0.24682101562714565,
+      "epoch": 0.2550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.763433431368941e-05,
-      "loss": 1.054,
+      "learning_rate": 1.7460160638126663e-05,
+      "loss": 1.0175,
       "step": 8987
     },
     {
-      "epoch": 0.24684847985498887,
+      "epoch": 0.25505107832009083,
       "grad_norm": 0.0,
-      "learning_rate": 1.763375975708593e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.7459548566846773e-05,
+      "loss": 0.9693,
       "step": 8988
     },
     {
-      "epoch": 0.24687594408283212,
+      "epoch": 0.2550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.763318514008119e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.745893643255574e-05,
+      "loss": 0.9529,
       "step": 8989
     },
     {
-      "epoch": 0.24690340831067534,
+      "epoch": 0.2551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.7632610462679728e-05,
-      "loss": 0.8767,
+      "learning_rate": 1.7458324235258737e-05,
+      "loss": 1.0083,
       "step": 8990
     },
     {
-      "epoch": 0.2469308725385186,
+      "epoch": 0.25513620885357546,
       "grad_norm": 0.0,
-      "learning_rate": 1.76320357248861e-05,
-      "loss": 1.0787,
+      "learning_rate": 1.7457711974960934e-05,
+      "loss": 0.9703,
       "step": 8991
     },
     {
-      "epoch": 0.2469583367663618,
+      "epoch": 0.2551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.7631460926704845e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.74570996516675e-05,
+      "loss": 1.1212,
       "step": 8992
     },
     {
-      "epoch": 0.24698580099420506,
+      "epoch": 0.2551929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.7630886068140517e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.7456487265383614e-05,
+      "loss": 1.0193,
       "step": 8993
     },
     {
-      "epoch": 0.24701326522204828,
+      "epoch": 0.25522133938706015,
       "grad_norm": 0.0,
-      "learning_rate": 1.763031114919766e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.745587481611444e-05,
+      "loss": 0.9706,
       "step": 8994
     },
     {
-      "epoch": 0.24704072944989153,
+      "epoch": 0.25524971623155507,
       "grad_norm": 0.0,
-      "learning_rate": 1.762973616988083e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.745526230386516e-05,
+      "loss": 0.8932,
       "step": 8995
     },
     {
-      "epoch": 0.24706819367773475,
+      "epoch": 0.25527809307604993,
       "grad_norm": 0.0,
-      "learning_rate": 1.762916113019457e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.7454649728640944e-05,
+      "loss": 1.0075,
       "step": 8996
     },
     {
-      "epoch": 0.247095657905578,
+      "epoch": 0.25530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.762858603014343e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.7454037090446968e-05,
+      "loss": 0.9648,
       "step": 8997
     },
     {
-      "epoch": 0.24712312213342122,
+      "epoch": 0.2553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.7628010869731965e-05,
-      "loss": 1.0671,
+      "learning_rate": 1.7453424389288404e-05,
+      "loss": 1.0944,
       "step": 8998
     },
     {
-      "epoch": 0.24715058636126444,
+      "epoch": 0.2553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.762743564896472e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.745281162517043e-05,
+      "loss": 1.0303,
       "step": 8999
     },
     {
-      "epoch": 0.2471780505891077,
+      "epoch": 0.25539160045402953,
       "grad_norm": 0.0,
-      "learning_rate": 1.7626860367846252e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.7452198798098217e-05,
+      "loss": 0.9361,
       "step": 9000
     },
     {
-      "epoch": 0.2472055148169509,
+      "epoch": 0.2554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7626285026381112e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.7451585908076948e-05,
+      "loss": 0.9443,
       "step": 9001
     },
     {
-      "epoch": 0.24723297904479416,
+      "epoch": 0.2554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.762570962457385e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.74509729551118e-05,
+      "loss": 1.1018,
       "step": 9002
     },
     {
-      "epoch": 0.24726044327263738,
+      "epoch": 0.25547673098751417,
       "grad_norm": 0.0,
-      "learning_rate": 1.7625134162429015e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.745035993920795e-05,
+      "loss": 1.0009,
       "step": 9003
     },
     {
-      "epoch": 0.24728790750048063,
+      "epoch": 0.2555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.762455863995117e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.744974686037057e-05,
+      "loss": 0.8754,
       "step": 9004
     },
     {
-      "epoch": 0.24731537172832385,
+      "epoch": 0.255533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.762398305714486e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.7449133718604845e-05,
+      "loss": 0.9922,
       "step": 9005
     },
     {
-      "epoch": 0.2473428359561671,
+      "epoch": 0.25556186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.7623407414014645e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.7448520513915955e-05,
+      "loss": 1.178,
       "step": 9006
     },
     {
-      "epoch": 0.24737030018401032,
+      "epoch": 0.25559023836549377,
       "grad_norm": 0.0,
-      "learning_rate": 1.762283171056508e-05,
-      "loss": 0.972,
+      "learning_rate": 1.7447907246309072e-05,
+      "loss": 1.0175,
       "step": 9007
     },
     {
-      "epoch": 0.24739776441185357,
+      "epoch": 0.25561861520998863,
       "grad_norm": 0.0,
-      "learning_rate": 1.7622255946800712e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.744729391578939e-05,
+      "loss": 0.9973,
       "step": 9008
     },
     {
-      "epoch": 0.2474252286396968,
+      "epoch": 0.25564699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 1.7621680122726102e-05,
-      "loss": 0.945,
+      "learning_rate": 1.7446680522362073e-05,
+      "loss": 0.9768,
       "step": 9009
     },
     {
-      "epoch": 0.24745269286754004,
+      "epoch": 0.25567536889897846,
       "grad_norm": 0.0,
-      "learning_rate": 1.7621104238345812e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.7446067066032312e-05,
+      "loss": 0.8979,
       "step": 9010
     },
     {
-      "epoch": 0.24748015709538326,
+      "epoch": 0.2557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.762052829366439e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.744545354680529e-05,
+      "loss": 1.0576,
       "step": 9011
     },
     {
-      "epoch": 0.24750762132322648,
+      "epoch": 0.25573212258796824,
       "grad_norm": 0.0,
-      "learning_rate": 1.7619952288686395e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.7444839964686186e-05,
+      "loss": 0.9692,
       "step": 9012
     },
     {
-      "epoch": 0.24753508555106973,
+      "epoch": 0.2557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.761937622341639e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.7444226319680188e-05,
+      "loss": 0.8579,
       "step": 9013
     },
     {
-      "epoch": 0.24756254977891295,
+      "epoch": 0.255788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.7618800097858924e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.7443612611792473e-05,
+      "loss": 1.0057,
       "step": 9014
     },
     {
-      "epoch": 0.2475900140067562,
+      "epoch": 0.25581725312145287,
       "grad_norm": 0.0,
-      "learning_rate": 1.7618223912018565e-05,
-      "loss": 0.8673,
+      "learning_rate": 1.744299884102823e-05,
+      "loss": 1.0295,
       "step": 9015
     },
     {
-      "epoch": 0.24761747823459943,
+      "epoch": 0.2558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.7617647665899865e-05,
-      "loss": 0.8315,
+      "learning_rate": 1.7442385007392636e-05,
+      "loss": 1.0068,
       "step": 9016
     },
     {
-      "epoch": 0.24764494246244267,
+      "epoch": 0.2558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.7617071359507385e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.7441771110890884e-05,
+      "loss": 0.8499,
       "step": 9017
     },
     {
-      "epoch": 0.2476724066902859,
+      "epoch": 0.25590238365493756,
       "grad_norm": 0.0,
-      "learning_rate": 1.7616494992845688e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.7441157151528157e-05,
+      "loss": 0.9609,
       "step": 9018
     },
     {
-      "epoch": 0.24769987091812914,
+      "epoch": 0.2559307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.7615918565919332e-05,
-      "loss": 1.0597,
+      "learning_rate": 1.7440543129309643e-05,
+      "loss": 1.0969,
       "step": 9019
     },
     {
-      "epoch": 0.24772733514597237,
+      "epoch": 0.25595913734392733,
       "grad_norm": 0.0,
-      "learning_rate": 1.7615342078732877e-05,
-      "loss": 0.9083,
+      "learning_rate": 1.7439929044240522e-05,
+      "loss": 1.0021,
       "step": 9020
     },
     {
-      "epoch": 0.24775479937381562,
+      "epoch": 0.25598751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.7614765531290886e-05,
-      "loss": 0.9423,
+      "learning_rate": 1.7439314896325988e-05,
+      "loss": 0.9981,
       "step": 9021
     },
     {
-      "epoch": 0.24778226360165884,
+      "epoch": 0.25601589103291716,
       "grad_norm": 0.0,
-      "learning_rate": 1.7614188923597918e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.743870068557123e-05,
+      "loss": 1.0104,
       "step": 9022
     },
     {
-      "epoch": 0.24780972782950209,
+      "epoch": 0.256044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.761361225565854e-05,
-      "loss": 1.0707,
+      "learning_rate": 1.743808641198143e-05,
+      "loss": 0.9424,
       "step": 9023
     },
     {
-      "epoch": 0.2478371920573453,
+      "epoch": 0.25607264472190694,
       "grad_norm": 0.0,
-      "learning_rate": 1.7613035527477314e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.743747207556178e-05,
+      "loss": 0.8839,
       "step": 9024
     },
     {
-      "epoch": 0.24786465628518853,
+      "epoch": 0.2561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.76124587390588e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.743685767631747e-05,
+      "loss": 1.0042,
       "step": 9025
     },
     {
-      "epoch": 0.24789212051303178,
+      "epoch": 0.2561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.7611881890407564e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.7436243214253686e-05,
+      "loss": 0.9697,
       "step": 9026
     },
     {
-      "epoch": 0.247919584740875,
+      "epoch": 0.2561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.761130498152817e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.7435628689375622e-05,
+      "loss": 1.0327,
       "step": 9027
     },
     {
-      "epoch": 0.24794704896871825,
+      "epoch": 0.2561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.7610728012425184e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.7435014101688474e-05,
+      "loss": 0.8754,
       "step": 9028
     },
     {
-      "epoch": 0.24797451319656147,
+      "epoch": 0.2562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.761015098310317e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.743439945119742e-05,
+      "loss": 1.0306,
       "step": 9029
     },
     {
-      "epoch": 0.24800197742440472,
+      "epoch": 0.25624290578887626,
       "grad_norm": 0.0,
-      "learning_rate": 1.7609573893566692e-05,
-      "loss": 0.9373,
+      "learning_rate": 1.7433784737907662e-05,
+      "loss": 0.9953,
       "step": 9030
     },
     {
-      "epoch": 0.24802944165224794,
+      "epoch": 0.2562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.760899674382032e-05,
-      "loss": 1.0484,
+      "learning_rate": 1.743316996182439e-05,
+      "loss": 0.91,
       "step": 9031
     },
     {
-      "epoch": 0.2480569058800912,
+      "epoch": 0.25629965947786604,
       "grad_norm": 0.0,
-      "learning_rate": 1.760841953386861e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.74325551229528e-05,
+      "loss": 0.9887,
       "step": 9032
     },
     {
-      "epoch": 0.2480843701079344,
+      "epoch": 0.25632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.7607842263716145e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.7431940221298082e-05,
+      "loss": 0.9624,
       "step": 9033
     },
     {
-      "epoch": 0.24811183433577766,
+      "epoch": 0.25635641316685587,
       "grad_norm": 0.0,
-      "learning_rate": 1.7607264933367487e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.7431325256865427e-05,
+      "loss": 0.8724,
       "step": 9034
     },
     {
-      "epoch": 0.24813929856362088,
+      "epoch": 0.2563847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.76066875428272e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.7430710229660037e-05,
+      "loss": 0.9746,
       "step": 9035
     },
     {
-      "epoch": 0.24816676279146413,
+      "epoch": 0.25641316685584564,
       "grad_norm": 0.0,
-      "learning_rate": 1.7606110092099852e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.74300951396871e-05,
+      "loss": 0.9318,
       "step": 9036
     },
     {
-      "epoch": 0.24819422701930735,
+      "epoch": 0.2564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.7605532581190017e-05,
-      "loss": 0.9786,
+      "learning_rate": 1.7429479986951822e-05,
+      "loss": 0.9634,
       "step": 9037
     },
     {
-      "epoch": 0.2482216912471506,
+      "epoch": 0.2564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.760495501010226e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.7428864771459387e-05,
+      "loss": 0.9923,
       "step": 9038
     },
     {
-      "epoch": 0.24824915547499382,
+      "epoch": 0.25649829738933033,
       "grad_norm": 0.0,
-      "learning_rate": 1.7604377378841155e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.7428249493215e-05,
+      "loss": 1.0334,
       "step": 9039
     },
     {
-      "epoch": 0.24827661970283704,
+      "epoch": 0.2565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.760379968741127e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.7427634152223857e-05,
+      "loss": 1.0047,
       "step": 9040
     },
     {
-      "epoch": 0.2483040839306803,
+      "epoch": 0.2565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.7603221935817174e-05,
-      "loss": 1.1073,
+      "learning_rate": 1.742701874849115e-05,
+      "loss": 0.9982,
       "step": 9041
     },
     {
-      "epoch": 0.2483315481585235,
+      "epoch": 0.25658342792281497,
       "grad_norm": 0.0,
-      "learning_rate": 1.7602644124063442e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.7426403282022084e-05,
+      "loss": 0.9948,
       "step": 9042
     },
     {
-      "epoch": 0.24835901238636676,
+      "epoch": 0.2566118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.7602066252154647e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.7425787752821853e-05,
+      "loss": 1.016,
       "step": 9043
     },
     {
-      "epoch": 0.24838647661420998,
+      "epoch": 0.25664018161180474,
       "grad_norm": 0.0,
-      "learning_rate": 1.7601488320095358e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.7425172160895664e-05,
+      "loss": 0.9994,
       "step": 9044
     },
     {
-      "epoch": 0.24841394084205323,
+      "epoch": 0.25666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.7600910327890147e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.742455650624871e-05,
+      "loss": 0.9549,
       "step": 9045
     },
     {
-      "epoch": 0.24844140506989645,
+      "epoch": 0.25669693530079457,
       "grad_norm": 0.0,
-      "learning_rate": 1.760033227554359e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.7423940788886192e-05,
+      "loss": 0.9517,
       "step": 9046
     },
     {
-      "epoch": 0.2484688692977397,
+      "epoch": 0.25672531214528943,
       "grad_norm": 0.0,
-      "learning_rate": 1.7599754163060264e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.7423325008813315e-05,
+      "loss": 1.04,
       "step": 9047
     },
     {
-      "epoch": 0.24849633352558292,
+      "epoch": 0.25675368898978435,
       "grad_norm": 0.0,
-      "learning_rate": 1.7599175990444733e-05,
-      "loss": 1.0828,
+      "learning_rate": 1.7422709166035273e-05,
+      "loss": 0.9477,
       "step": 9048
     },
     {
-      "epoch": 0.24852379775342617,
+      "epoch": 0.2567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.7598597757701583e-05,
-      "loss": 1.098,
+      "learning_rate": 1.7422093260557277e-05,
+      "loss": 0.9937,
       "step": 9049
     },
     {
-      "epoch": 0.2485512619812694,
+      "epoch": 0.2568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.7598019464835383e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.7421477292384524e-05,
+      "loss": 0.9758,
       "step": 9050
     },
     {
-      "epoch": 0.24857872620911264,
+      "epoch": 0.25683881952326904,
       "grad_norm": 0.0,
-      "learning_rate": 1.7597441111850708e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.7420861261522223e-05,
+      "loss": 0.8509,
       "step": 9051
     },
     {
-      "epoch": 0.24860619043695587,
+      "epoch": 0.2568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.759686269875214e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.742024516797557e-05,
+      "loss": 0.9223,
       "step": 9052
     },
     {
-      "epoch": 0.2486336546647991,
+      "epoch": 0.2568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.7596284225544246e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.741962901174977e-05,
+      "loss": 1.0774,
       "step": 9053
     },
     {
-      "epoch": 0.24866111889264234,
+      "epoch": 0.25692395005675367,
       "grad_norm": 0.0,
-      "learning_rate": 1.759570569223161e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.741901279285003e-05,
+      "loss": 0.962,
       "step": 9054
     },
     {
-      "epoch": 0.24868858312048556,
+      "epoch": 0.2569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.7595127098818808e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.741839651128156e-05,
+      "loss": 0.7991,
       "step": 9055
     },
     {
-      "epoch": 0.2487160473483288,
+      "epoch": 0.2569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.759454844531042e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.741778016704956e-05,
+      "loss": 0.9187,
       "step": 9056
     },
     {
-      "epoch": 0.24874351157617203,
+      "epoch": 0.25700908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.7593969731711024e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.7417163760159238e-05,
+      "loss": 1.0131,
       "step": 9057
     },
     {
-      "epoch": 0.24877097580401528,
+      "epoch": 0.2570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.7593390958025196e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.7416547290615798e-05,
+      "loss": 0.9218,
       "step": 9058
     },
     {
-      "epoch": 0.2487984400318585,
+      "epoch": 0.25706583427922813,
       "grad_norm": 0.0,
-      "learning_rate": 1.759281212425752e-05,
-      "loss": 1.0765,
+      "learning_rate": 1.741593075842445e-05,
+      "loss": 1.0318,
       "step": 9059
     },
     {
-      "epoch": 0.24882590425970175,
+      "epoch": 0.25709421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 1.759223323041257e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.7415314163590405e-05,
+      "loss": 1.0955,
       "step": 9060
     },
     {
-      "epoch": 0.24885336848754497,
+      "epoch": 0.2571225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.7591654276494933e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.7414697506118862e-05,
+      "loss": 0.9567,
       "step": 9061
     },
     {
-      "epoch": 0.24888083271538822,
+      "epoch": 0.2571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.759107526250918e-05,
-      "loss": 1.1543,
+      "learning_rate": 1.741408078601504e-05,
+      "loss": 0.9685,
       "step": 9062
     },
     {
-      "epoch": 0.24890829694323144,
+      "epoch": 0.25717934165720774,
       "grad_norm": 0.0,
-      "learning_rate": 1.7590496188459904e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.7413464003284143e-05,
+      "loss": 1.0286,
       "step": 9063
     },
     {
-      "epoch": 0.2489357611710747,
+      "epoch": 0.2572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.758991705435168e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.7412847157931382e-05,
+      "loss": 0.9884,
       "step": 9064
     },
     {
-      "epoch": 0.2489632253989179,
+      "epoch": 0.2572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.7589337860189096e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.741223024996197e-05,
+      "loss": 0.9122,
       "step": 9065
     },
     {
-      "epoch": 0.24899068962676113,
+      "epoch": 0.2572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.758875860597673e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.7411613279381116e-05,
+      "loss": 1.0282,
       "step": 9066
     },
     {
-      "epoch": 0.24901815385460438,
+      "epoch": 0.2572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.758817929171916e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.7410996246194028e-05,
+      "loss": 0.9897,
       "step": 9067
     },
     {
-      "epoch": 0.2490456180824476,
+      "epoch": 0.2573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.7587599917420983e-05,
-      "loss": 1.081,
+      "learning_rate": 1.7410379150405924e-05,
+      "loss": 0.9986,
       "step": 9068
     },
     {
-      "epoch": 0.24907308231029085,
+      "epoch": 0.25734960272417706,
       "grad_norm": 0.0,
-      "learning_rate": 1.7587020483086777e-05,
-      "loss": 0.959,
+      "learning_rate": 1.7409761992022013e-05,
+      "loss": 0.9887,
       "step": 9069
     },
     {
-      "epoch": 0.24910054653813407,
+      "epoch": 0.257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.7586440988721123e-05,
-      "loss": 0.9786,
+      "learning_rate": 1.7409144771047508e-05,
+      "loss": 0.9843,
       "step": 9070
     },
     {
-      "epoch": 0.24912801076597732,
+      "epoch": 0.25740635641316684,
       "grad_norm": 0.0,
-      "learning_rate": 1.758586143432861e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.7408527487487626e-05,
+      "loss": 0.954,
       "step": 9071
     },
     {
-      "epoch": 0.24915547499382054,
+      "epoch": 0.25743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.7585281819913823e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.740791014134758e-05,
+      "loss": 0.9478,
       "step": 9072
     },
     {
-      "epoch": 0.2491829392216638,
+      "epoch": 0.25746311010215667,
       "grad_norm": 0.0,
-      "learning_rate": 1.7584702145481344e-05,
-      "loss": 1.021,
+      "learning_rate": 1.7407292732632582e-05,
+      "loss": 0.866,
       "step": 9073
     },
     {
-      "epoch": 0.249210403449507,
+      "epoch": 0.2574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.7584122411035765e-05,
-      "loss": 1.0169,
+      "learning_rate": 1.740667526134785e-05,
+      "loss": 0.9198,
       "step": 9074
     },
     {
-      "epoch": 0.24923786767735026,
+      "epoch": 0.25751986379114644,
       "grad_norm": 0.0,
-      "learning_rate": 1.7583542616581676e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.7406057727498602e-05,
+      "loss": 1.0606,
       "step": 9075
     },
     {
-      "epoch": 0.24926533190519348,
+      "epoch": 0.2575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.7582962762123658e-05,
-      "loss": 0.9295,
+      "learning_rate": 1.740544013109005e-05,
+      "loss": 0.9838,
       "step": 9076
     },
     {
-      "epoch": 0.24929279613303673,
+      "epoch": 0.2575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.7582382847666298e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.7404822472127406e-05,
+      "loss": 0.9631,
       "step": 9077
     },
     {
-      "epoch": 0.24932026036087995,
+      "epoch": 0.2576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.7581802873214194e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.74042047506159e-05,
+      "loss": 0.97,
       "step": 9078
     },
     {
-      "epoch": 0.2493477245887232,
+      "epoch": 0.257633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.7581222838771927e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.7403586966560743e-05,
+      "loss": 0.925,
       "step": 9079
     },
     {
-      "epoch": 0.24937518881656642,
+      "epoch": 0.2576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.7580642744344084e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.7402969119967154e-05,
+      "loss": 1.0544,
       "step": 9080
     },
     {
-      "epoch": 0.24940265304440964,
+      "epoch": 0.25769012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.7580062589935263e-05,
-      "loss": 0.992,
+      "learning_rate": 1.7402351210840352e-05,
+      "loss": 1.0105,
       "step": 9081
     },
     {
-      "epoch": 0.2494301172722529,
+      "epoch": 0.2577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.757948237555005e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.7401733239185557e-05,
+      "loss": 0.9454,
       "step": 9082
     },
     {
-      "epoch": 0.24945758150009611,
+      "epoch": 0.25774687854710554,
       "grad_norm": 0.0,
-      "learning_rate": 1.7578902101193035e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.7401115205007987e-05,
+      "loss": 0.9199,
       "step": 9083
     },
     {
-      "epoch": 0.24948504572793936,
+      "epoch": 0.25777525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.7578321766868812e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.7400497108312867e-05,
+      "loss": 1.0692,
       "step": 9084
     },
     {
-      "epoch": 0.24951250995578259,
+      "epoch": 0.25780363223609537,
       "grad_norm": 0.0,
-      "learning_rate": 1.7577741372581972e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.7399878949105414e-05,
+      "loss": 0.9834,
       "step": 9085
     },
     {
-      "epoch": 0.24953997418362583,
+      "epoch": 0.25783200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 1.7577160918337105e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.7399260727390846e-05,
+      "loss": 0.9417,
       "step": 9086
     },
     {
-      "epoch": 0.24956743841146906,
+      "epoch": 0.25786038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.757658040413881e-05,
-      "loss": 0.8408,
+      "learning_rate": 1.7398642443174395e-05,
+      "loss": 1.0114,
       "step": 9087
     },
     {
-      "epoch": 0.2495949026393123,
+      "epoch": 0.25788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.757599982999167e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.739802409646128e-05,
+      "loss": 1.1028,
       "step": 9088
     },
     {
-      "epoch": 0.24962236686715553,
+      "epoch": 0.2579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.7575419195900286e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.739740568725672e-05,
+      "loss": 1.0221,
       "step": 9089
     },
     {
-      "epoch": 0.24964983109499878,
+      "epoch": 0.25794551645856983,
       "grad_norm": 0.0,
-      "learning_rate": 1.7574838501869255e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.739678721556594e-05,
+      "loss": 0.9357,
       "step": 9090
     },
     {
-      "epoch": 0.249677295322842,
+      "epoch": 0.2579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.7574257747903165e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.739616868139417e-05,
+      "loss": 0.9579,
       "step": 9091
     },
     {
-      "epoch": 0.24970475955068525,
+      "epoch": 0.2580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.7573676934006614e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.7395550084746632e-05,
+      "loss": 0.9443,
       "step": 9092
     },
     {
-      "epoch": 0.24973222377852847,
+      "epoch": 0.25803064699205447,
       "grad_norm": 0.0,
-      "learning_rate": 1.7573096060184197e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.7394931425628543e-05,
+      "loss": 1.0238,
       "step": 9093
     },
     {
-      "epoch": 0.2497596880063717,
+      "epoch": 0.2580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.757251512644051e-05,
-      "loss": 0.9057,
+      "learning_rate": 1.7394312704045143e-05,
+      "loss": 0.9444,
       "step": 9094
     },
     {
-      "epoch": 0.24978715223421494,
+      "epoch": 0.25808740068104424,
       "grad_norm": 0.0,
-      "learning_rate": 1.757193413278015e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.7393693920001647e-05,
+      "loss": 0.9968,
       "step": 9095
     },
     {
-      "epoch": 0.24981461646205816,
+      "epoch": 0.25811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.7571353079207715e-05,
-      "loss": 0.8328,
+      "learning_rate": 1.7393075073503286e-05,
+      "loss": 1.1219,
       "step": 9096
     },
     {
-      "epoch": 0.2498420806899014,
+      "epoch": 0.2581441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.75707719657278e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.739245616455529e-05,
+      "loss": 1.0191,
       "step": 9097
     },
     {
-      "epoch": 0.24986954491774463,
+      "epoch": 0.25817253121452893,
       "grad_norm": 0.0,
-      "learning_rate": 1.7570190792345006e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.7391837193162885e-05,
+      "loss": 0.9777,
       "step": 9098
     },
     {
-      "epoch": 0.24989700914558788,
+      "epoch": 0.25820090805902385,
       "grad_norm": 0.0,
-      "learning_rate": 1.756960955906393e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.7391218159331295e-05,
+      "loss": 1.0008,
       "step": 9099
     },
     {
-      "epoch": 0.2499244733734311,
+      "epoch": 0.2582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.7569028265889172e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.7390599063065753e-05,
+      "loss": 0.932,
       "step": 9100
     },
     {
-      "epoch": 0.24995193760127435,
+      "epoch": 0.2582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.756844691282533e-05,
-      "loss": 1.0746,
+      "learning_rate": 1.738997990437149e-05,
+      "loss": 0.9869,
       "step": 9101
     },
     {
-      "epoch": 0.24997940182911757,
+      "epoch": 0.25828603859250854,
       "grad_norm": 0.0,
-      "learning_rate": 1.7567865499877e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.7389360683253737e-05,
+      "loss": 0.9772,
       "step": 9102
     },
     {
-      "epoch": 0.2500068660569608,
+      "epoch": 0.2583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.756728402704879e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.738874139971772e-05,
+      "loss": 0.8549,
       "step": 9103
     },
     {
-      "epoch": 0.25003433028480404,
+      "epoch": 0.2583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.75667024943453e-05,
-      "loss": 1.0795,
+      "learning_rate": 1.7388122053768673e-05,
+      "loss": 1.0065,
       "step": 9104
     },
     {
-      "epoch": 0.25006179451264726,
+      "epoch": 0.25837116912599317,
       "grad_norm": 0.0,
-      "learning_rate": 1.7566120901771126e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.7387502645411826e-05,
+      "loss": 1.016,
       "step": 9105
     },
     {
-      "epoch": 0.25008925874049054,
+      "epoch": 0.2583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.7565539249330873e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.7386883174652415e-05,
+      "loss": 0.9743,
       "step": 9106
     },
     {
-      "epoch": 0.25011672296833376,
+      "epoch": 0.25842792281498295,
       "grad_norm": 0.0,
-      "learning_rate": 1.7564957537029142e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.7386263641495668e-05,
+      "loss": 1.0257,
       "step": 9107
     },
     {
-      "epoch": 0.250144187196177,
+      "epoch": 0.25845629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.7564375764870537e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.7385644045946816e-05,
+      "loss": 0.9642,
       "step": 9108
     },
     {
-      "epoch": 0.2501716514240202,
+      "epoch": 0.2584846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.7563793932859662e-05,
-      "loss": 0.9689,
+      "learning_rate": 1.73850243880111e-05,
+      "loss": 0.973,
       "step": 9109
     },
     {
-      "epoch": 0.2501991156518634,
+      "epoch": 0.25851305334846764,
       "grad_norm": 0.0,
-      "learning_rate": 1.756321204100112e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.738440466769375e-05,
+      "loss": 0.9658,
       "step": 9110
     },
     {
-      "epoch": 0.2502265798797067,
+      "epoch": 0.25854143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 1.756263008929951e-05,
-      "loss": 0.8959,
+      "learning_rate": 1.7383784885000004e-05,
+      "loss": 0.9774,
       "step": 9111
     },
     {
-      "epoch": 0.2502540441075499,
+      "epoch": 0.2585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.7562048077759448e-05,
-      "loss": 1.1292,
+      "learning_rate": 1.7383165039935094e-05,
+      "loss": 0.9515,
       "step": 9112
     },
     {
-      "epoch": 0.25028150833539314,
+      "epoch": 0.2585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.7561466006385525e-05,
-      "loss": 1.0729,
+      "learning_rate": 1.7382545132504255e-05,
+      "loss": 1.015,
       "step": 9113
     },
     {
-      "epoch": 0.25030897256323636,
+      "epoch": 0.25862656072644724,
       "grad_norm": 0.0,
-      "learning_rate": 1.756088387518236e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.7381925162712727e-05,
+      "loss": 1.0648,
       "step": 9114
     },
     {
-      "epoch": 0.25033643679107964,
+      "epoch": 0.2586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.756030168415455e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.7381305130565747e-05,
+      "loss": 0.9422,
       "step": 9115
     },
     {
-      "epoch": 0.25036390101892286,
+      "epoch": 0.258683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.7559719433306707e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.738068503606855e-05,
+      "loss": 0.9792,
       "step": 9116
     },
     {
-      "epoch": 0.2503913652467661,
+      "epoch": 0.2587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.7559137122643432e-05,
-      "loss": 0.9899,
+      "learning_rate": 1.7380064879226374e-05,
+      "loss": 1.0379,
       "step": 9117
     },
     {
-      "epoch": 0.2504188294746093,
+      "epoch": 0.2587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.755855475216934e-05,
-      "loss": 0.8289,
+      "learning_rate": 1.7379444660044456e-05,
+      "loss": 0.962,
       "step": 9118
     },
     {
-      "epoch": 0.2504462937024526,
+      "epoch": 0.2587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.755797232188903e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.737882437852804e-05,
+      "loss": 0.8831,
       "step": 9119
     },
     {
-      "epoch": 0.2504737579302958,
+      "epoch": 0.25879682179341656,
       "grad_norm": 0.0,
-      "learning_rate": 1.7557389831807116e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.7378204034682364e-05,
+      "loss": 0.9544,
       "step": 9120
     },
     {
-      "epoch": 0.250501222158139,
+      "epoch": 0.2588251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.755680728192821e-05,
-      "loss": 0.9674,
+      "learning_rate": 1.7377583628512665e-05,
+      "loss": 0.9732,
       "step": 9121
     },
     {
-      "epoch": 0.25052868638598225,
+      "epoch": 0.25885357548240634,
       "grad_norm": 0.0,
-      "learning_rate": 1.7556224672256916e-05,
-      "loss": 1.0225,
+      "learning_rate": 1.7376963160024184e-05,
+      "loss": 0.948,
       "step": 9122
     },
     {
-      "epoch": 0.25055615061382547,
+      "epoch": 0.25888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.7555642002797845e-05,
-      "loss": 0.9645,
+      "learning_rate": 1.7376342629222165e-05,
+      "loss": 0.9091,
       "step": 9123
     },
     {
-      "epoch": 0.25058361484166874,
+      "epoch": 0.2589103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.7555059273555608e-05,
-      "loss": 1.0179,
+      "learning_rate": 1.7375722036111848e-05,
+      "loss": 1.1245,
       "step": 9124
     },
     {
-      "epoch": 0.25061107906951197,
+      "epoch": 0.25893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.7554476484534813e-05,
-      "loss": 1.0963,
+      "learning_rate": 1.7375101380698475e-05,
+      "loss": 1.0428,
       "step": 9125
     },
     {
-      "epoch": 0.2506385432973552,
+      "epoch": 0.25896708286038594,
       "grad_norm": 0.0,
-      "learning_rate": 1.755389363574008e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.737448066298729e-05,
+      "loss": 0.8579,
       "step": 9126
     },
     {
-      "epoch": 0.2506660075251984,
+      "epoch": 0.2589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.755331072717601e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.7373859882983537e-05,
+      "loss": 0.935,
       "step": 9127
     },
     {
-      "epoch": 0.2506934717530417,
+      "epoch": 0.2590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.7552727758847218e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.7373239040692457e-05,
+      "loss": 0.8934,
       "step": 9128
     },
     {
-      "epoch": 0.2507209359808849,
+      "epoch": 0.2590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.755214473075832e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.7372618136119292e-05,
+      "loss": 0.9477,
       "step": 9129
     },
     {
-      "epoch": 0.25074840020872813,
+      "epoch": 0.2590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.755156164291393e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.7371997169269293e-05,
+      "loss": 0.9005,
       "step": 9130
     },
     {
-      "epoch": 0.25077586443657135,
+      "epoch": 0.2591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.7550978495318655e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.7371376140147705e-05,
+      "loss": 1.0615,
       "step": 9131
     },
     {
-      "epoch": 0.2508033286644146,
+      "epoch": 0.25913734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.7550395287977114e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.7370755048759767e-05,
+      "loss": 1.071,
       "step": 9132
     },
     {
-      "epoch": 0.25083079289225785,
+      "epoch": 0.2591657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.7549812020893923e-05,
-      "loss": 1.085,
+      "learning_rate": 1.737013389511073e-05,
+      "loss": 0.8247,
       "step": 9133
     },
     {
-      "epoch": 0.25085825712010107,
+      "epoch": 0.25919409761634504,
       "grad_norm": 0.0,
-      "learning_rate": 1.7549228694073696e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.7369512679205844e-05,
+      "loss": 1.0092,
       "step": 9134
     },
     {
-      "epoch": 0.2508857213479443,
+      "epoch": 0.25922247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.7548645307521044e-05,
-      "loss": 0.93,
+      "learning_rate": 1.736889140105035e-05,
+      "loss": 0.9442,
       "step": 9135
     },
     {
-      "epoch": 0.2509131855757875,
+      "epoch": 0.2592508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.754806186124059e-05,
-      "loss": 1.0728,
+      "learning_rate": 1.7368270060649503e-05,
+      "loss": 0.9615,
       "step": 9136
     },
     {
-      "epoch": 0.2509406498036308,
+      "epoch": 0.25927922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 1.754747835523694e-05,
-      "loss": 0.9489,
+      "learning_rate": 1.7367648658008544e-05,
+      "loss": 0.9078,
       "step": 9137
     },
     {
-      "epoch": 0.250968114031474,
+      "epoch": 0.25930760499432465,
       "grad_norm": 0.0,
-      "learning_rate": 1.7546894789514724e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.7367027193132723e-05,
+      "loss": 0.9521,
       "step": 9138
     },
     {
-      "epoch": 0.25099557825931723,
+      "epoch": 0.2593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.7546311164078547e-05,
-      "loss": 1.0905,
+      "learning_rate": 1.7366405666027296e-05,
+      "loss": 0.9362,
       "step": 9139
     },
     {
-      "epoch": 0.25102304248716045,
+      "epoch": 0.2593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.754572747893304e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.736578407669751e-05,
+      "loss": 0.9418,
       "step": 9140
     },
     {
-      "epoch": 0.25105050671500373,
+      "epoch": 0.2593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.7545143734082808e-05,
-      "loss": 1.0402,
+      "learning_rate": 1.7365162425148615e-05,
+      "loss": 0.9932,
       "step": 9141
     },
     {
-      "epoch": 0.25107797094284695,
+      "epoch": 0.2594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7544559929532478e-05,
-      "loss": 1.056,
+      "learning_rate": 1.736454071138586e-05,
+      "loss": 0.9677,
       "step": 9142
     },
     {
-      "epoch": 0.25110543517069017,
+      "epoch": 0.2594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.7543976065286668e-05,
-      "loss": 0.9244,
+      "learning_rate": 1.73639189354145e-05,
+      "loss": 0.8522,
       "step": 9143
     },
     {
-      "epoch": 0.2511328993985334,
+      "epoch": 0.25947786606129397,
       "grad_norm": 0.0,
-      "learning_rate": 1.754339214135e-05,
-      "loss": 0.9513,
+      "learning_rate": 1.7363297097239784e-05,
+      "loss": 0.9832,
       "step": 9144
     },
     {
-      "epoch": 0.25116036362637667,
+      "epoch": 0.2595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.754280815772709e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.7362675196866968e-05,
+      "loss": 0.9812,
       "step": 9145
     },
     {
-      "epoch": 0.2511878278542199,
+      "epoch": 0.25953461975028375,
       "grad_norm": 0.0,
-      "learning_rate": 1.7542224114422555e-05,
-      "loss": 1.004,
+      "learning_rate": 1.73620532343013e-05,
+      "loss": 0.9958,
       "step": 9146
     },
     {
-      "epoch": 0.2512152920820631,
+      "epoch": 0.25956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.7541640011441024e-05,
-      "loss": 1.0733,
+      "learning_rate": 1.736143120954804e-05,
+      "loss": 0.9956,
       "step": 9147
     },
     {
-      "epoch": 0.25124275630990633,
+      "epoch": 0.2595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.754105584878712e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.736080912261244e-05,
+      "loss": 0.8895,
       "step": 9148
     },
     {
-      "epoch": 0.25127022053774956,
+      "epoch": 0.25961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.7540471626465456e-05,
-      "loss": 1.045,
+      "learning_rate": 1.7360186973499752e-05,
+      "loss": 0.9951,
       "step": 9149
     },
     {
-      "epoch": 0.25129768476559283,
+      "epoch": 0.25964812712826335,
       "grad_norm": 0.0,
-      "learning_rate": 1.7539887344480662e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.735956476221524e-05,
+      "loss": 1.0687,
       "step": 9150
     },
     {
-      "epoch": 0.25132514899343605,
+      "epoch": 0.2596765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.753930300283736e-05,
-      "loss": 1.1654,
+      "learning_rate": 1.735894248876415e-05,
+      "loss": 0.9838,
       "step": 9151
     },
     {
-      "epoch": 0.2513526132212793,
+      "epoch": 0.2597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.7538718601540168e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.735832015315174e-05,
+      "loss": 1.0184,
       "step": 9152
     },
     {
-      "epoch": 0.2513800774491225,
+      "epoch": 0.25973325766174804,
       "grad_norm": 0.0,
-      "learning_rate": 1.7538134140593717e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.735769775538327e-05,
+      "loss": 1.0076,
       "step": 9153
     },
     {
-      "epoch": 0.2514075416769658,
+      "epoch": 0.2597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.753754962000263e-05,
-      "loss": 1.0681,
+      "learning_rate": 1.7357075295464e-05,
+      "loss": 0.9194,
       "step": 9154
     },
     {
-      "epoch": 0.251435005904809,
+      "epoch": 0.2597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.7536965039771527e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.7356452773399178e-05,
+      "loss": 0.9975,
       "step": 9155
     },
     {
-      "epoch": 0.2514624701326522,
+      "epoch": 0.2598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.753638039990504e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.735583018919407e-05,
+      "loss": 0.9272,
       "step": 9156
     },
     {
-      "epoch": 0.25148993436049544,
+      "epoch": 0.2598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.753579570040779e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.7355207542853938e-05,
+      "loss": 0.9406,
       "step": 9157
     },
     {
-      "epoch": 0.2515173985883387,
+      "epoch": 0.25987514188422245,
       "grad_norm": 0.0,
-      "learning_rate": 1.7535210941284406e-05,
-      "loss": 0.986,
+      "learning_rate": 1.7354584834384036e-05,
+      "loss": 0.9551,
       "step": 9158
     },
     {
-      "epoch": 0.25154486281618194,
+      "epoch": 0.25990351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.7534626122539515e-05,
-      "loss": 0.972,
+      "learning_rate": 1.735396206378962e-05,
+      "loss": 1.021,
       "step": 9159
     },
     {
-      "epoch": 0.25157232704402516,
+      "epoch": 0.2599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.7534041244177743e-05,
-      "loss": 1.0717,
+      "learning_rate": 1.7353339231075964e-05,
+      "loss": 0.9184,
       "step": 9160
     },
     {
-      "epoch": 0.2515997912718684,
+      "epoch": 0.25996027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.7533456306203715e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.7352716336248313e-05,
+      "loss": 1.0486,
       "step": 9161
     },
     {
-      "epoch": 0.2516272554997116,
+      "epoch": 0.25998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.753287130862207e-05,
-      "loss": 0.9591,
+      "learning_rate": 1.7352093379311942e-05,
+      "loss": 1.0144,
       "step": 9162
     },
     {
-      "epoch": 0.2516547197275549,
+      "epoch": 0.2600170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.7532286251437424e-05,
-      "loss": 0.986,
+      "learning_rate": 1.7351470360272107e-05,
+      "loss": 1.0297,
       "step": 9163
     },
     {
-      "epoch": 0.2516821839553981,
+      "epoch": 0.26004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.753170113465441e-05,
-      "loss": 1.019,
+      "learning_rate": 1.7350847279134072e-05,
+      "loss": 0.9957,
       "step": 9164
     },
     {
-      "epoch": 0.2517096481832413,
+      "epoch": 0.26007377979568674,
       "grad_norm": 0.0,
-      "learning_rate": 1.7531115958277664e-05,
-      "loss": 1.0763,
+      "learning_rate": 1.7350224135903097e-05,
+      "loss": 1.0318,
       "step": 9165
     },
     {
-      "epoch": 0.25173711241108454,
+      "epoch": 0.2601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.7530530722311808e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.734960093058445e-05,
+      "loss": 0.8897,
       "step": 9166
     },
     {
-      "epoch": 0.2517645766389278,
+      "epoch": 0.2601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.7529945426761473e-05,
-      "loss": 0.9113,
+      "learning_rate": 1.7348977663183392e-05,
+      "loss": 0.9836,
       "step": 9167
     },
     {
-      "epoch": 0.25179204086677104,
+      "epoch": 0.2601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.75293600716313e-05,
-      "loss": 1.1279,
+      "learning_rate": 1.734835433370519e-05,
+      "loss": 0.9211,
       "step": 9168
     },
     {
-      "epoch": 0.25181950509461426,
+      "epoch": 0.2601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.752877465692591e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.734773094215511e-05,
+      "loss": 1.0753,
       "step": 9169
     },
     {
-      "epoch": 0.2518469693224575,
+      "epoch": 0.2602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.752818918264994e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.7347107488538413e-05,
+      "loss": 0.986,
       "step": 9170
     },
     {
-      "epoch": 0.25187443355030076,
+      "epoch": 0.26024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.7527603648808016e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.7346483972860373e-05,
+      "loss": 0.9306,
       "step": 9171
     },
     {
-      "epoch": 0.251901897778144,
+      "epoch": 0.260272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.752701805540478e-05,
-      "loss": 1.0148,
+      "learning_rate": 1.734586039512625e-05,
+      "loss": 0.9219,
       "step": 9172
     },
     {
-      "epoch": 0.2519293620059872,
+      "epoch": 0.26030079455164584,
       "grad_norm": 0.0,
-      "learning_rate": 1.7526432402444862e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.734523675534132e-05,
+      "loss": 1.0444,
       "step": 9173
     },
     {
-      "epoch": 0.2519568262338304,
+      "epoch": 0.26032917139614076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7525846689932896e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.7344613053510838e-05,
+      "loss": 0.9518,
       "step": 9174
     },
     {
-      "epoch": 0.25198429046167364,
+      "epoch": 0.2603575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.752526091787351e-05,
-      "loss": 0.8778,
+      "learning_rate": 1.734398928964008e-05,
+      "loss": 0.9672,
       "step": 9175
     },
     {
-      "epoch": 0.2520117546895169,
+      "epoch": 0.26038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.752467508627135e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.7343365463734314e-05,
+      "loss": 0.9106,
       "step": 9176
     },
     {
-      "epoch": 0.25203921891736014,
+      "epoch": 0.26041430192962545,
       "grad_norm": 0.0,
-      "learning_rate": 1.7524089195131045e-05,
-      "loss": 0.988,
+      "learning_rate": 1.7342741575798813e-05,
+      "loss": 1.0428,
       "step": 9177
     },
     {
-      "epoch": 0.25206668314520336,
+      "epoch": 0.2604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.7523503244457233e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.7342117625838842e-05,
+      "loss": 1.1374,
       "step": 9178
     },
     {
-      "epoch": 0.2520941473730466,
+      "epoch": 0.2604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.7522917234254545e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.7341493613859672e-05,
+      "loss": 1.0506,
       "step": 9179
     },
     {
-      "epoch": 0.25212161160088986,
+      "epoch": 0.2604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.7522331164527627e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.734086953986658e-05,
+      "loss": 0.8936,
       "step": 9180
     },
     {
-      "epoch": 0.2521490758287331,
+      "epoch": 0.260527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.7521745035281106e-05,
-      "loss": 1.1018,
+      "learning_rate": 1.7340245403864825e-05,
+      "loss": 0.925,
       "step": 9181
     },
     {
-      "epoch": 0.2521765400565763,
+      "epoch": 0.2605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.7521158846519628e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.7339621205859692e-05,
+      "loss": 0.9938,
       "step": 9182
     },
     {
-      "epoch": 0.2522040042844195,
+      "epoch": 0.26058456299659477,
       "grad_norm": 0.0,
-      "learning_rate": 1.7520572598247827e-05,
-      "loss": 1.03,
+      "learning_rate": 1.733899694585645e-05,
+      "loss": 0.9399,
       "step": 9183
     },
     {
-      "epoch": 0.2522314685122628,
+      "epoch": 0.2606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.7519986290470342e-05,
-      "loss": 1.1274,
+      "learning_rate": 1.7338372623860372e-05,
+      "loss": 1.0,
       "step": 9184
     },
     {
-      "epoch": 0.252258932740106,
+      "epoch": 0.26064131668558455,
       "grad_norm": 0.0,
-      "learning_rate": 1.751939992319181e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.733774823987673e-05,
+      "loss": 0.9934,
       "step": 9185
     },
     {
-      "epoch": 0.25228639696794924,
+      "epoch": 0.26066969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.7518813496416877e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.733712379391079e-05,
+      "loss": 1.0128,
       "step": 9186
     },
     {
-      "epoch": 0.25231386119579247,
+      "epoch": 0.2606980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.751822701015018e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.7336499285967847e-05,
+      "loss": 0.9643,
       "step": 9187
     },
     {
-      "epoch": 0.2523413254236357,
+      "epoch": 0.26072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.7517640464396356e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.733587471605316e-05,
+      "loss": 1.0702,
       "step": 9188
     },
     {
-      "epoch": 0.25236878965147896,
+      "epoch": 0.26075482406356415,
       "grad_norm": 0.0,
-      "learning_rate": 1.7517053859160047e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.733525008417201e-05,
+      "loss": 0.9159,
       "step": 9189
     },
     {
-      "epoch": 0.2523962538793222,
+      "epoch": 0.260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.7516467194445897e-05,
-      "loss": 1.0954,
+      "learning_rate": 1.7334625390329672e-05,
+      "loss": 0.9749,
       "step": 9190
     },
     {
-      "epoch": 0.2524237181071654,
+      "epoch": 0.2608115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.7515880470258548e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.7334000634531426e-05,
+      "loss": 0.9306,
       "step": 9191
     },
     {
-      "epoch": 0.25245118233500863,
+      "epoch": 0.2608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.7515293686602643e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.7333375816782542e-05,
+      "loss": 1.0225,
       "step": 9192
     },
     {
-      "epoch": 0.2524786465628519,
+      "epoch": 0.2608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.7514706843482822e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.733275093708831e-05,
+      "loss": 0.9354,
       "step": 9193
     },
     {
-      "epoch": 0.2525061107906951,
+      "epoch": 0.2608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.7514119940903728e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.7332125995453992e-05,
+      "loss": 0.9403,
       "step": 9194
     },
     {
-      "epoch": 0.25253357501853835,
+      "epoch": 0.2609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.7513532978870005e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.7331500991884883e-05,
+      "loss": 1.0435,
       "step": 9195
     },
     {
-      "epoch": 0.25256103924638157,
+      "epoch": 0.2609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.7512945957386303e-05,
-      "loss": 0.8694,
+      "learning_rate": 1.7330875926386252e-05,
+      "loss": 0.9783,
       "step": 9196
     },
     {
-      "epoch": 0.25258850347422485,
+      "epoch": 0.26098183881952325,
       "grad_norm": 0.0,
-      "learning_rate": 1.751235887645726e-05,
-      "loss": 1.01,
+      "learning_rate": 1.7330250798963386e-05,
+      "loss": 0.9935,
       "step": 9197
     },
     {
-      "epoch": 0.25261596770206807,
+      "epoch": 0.26101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.7511771736087523e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.732962560962156e-05,
+      "loss": 0.8939,
       "step": 9198
     },
     {
-      "epoch": 0.2526434319299113,
+      "epoch": 0.2610385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 1.7511184536281742e-05,
-      "loss": 0.9616,
+      "learning_rate": 1.7329000358366057e-05,
+      "loss": 0.9756,
       "step": 9199
     },
     {
-      "epoch": 0.2526708961577545,
+      "epoch": 0.26106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.7510597277044555e-05,
-      "loss": 0.9209,
+      "learning_rate": 1.732837504520216e-05,
+      "loss": 1.0076,
       "step": 9200
     },
     {
-      "epoch": 0.25269836038559773,
+      "epoch": 0.26109534619750285,
       "grad_norm": 0.0,
-      "learning_rate": 1.7510009958380613e-05,
-      "loss": 0.859,
+      "learning_rate": 1.732774967013515e-05,
+      "loss": 0.8791,
       "step": 9201
     },
     {
-      "epoch": 0.252725824613441,
+      "epoch": 0.2611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.7509422580294563e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.7327124233170306e-05,
+      "loss": 0.9173,
       "step": 9202
     },
     {
-      "epoch": 0.25275328884128423,
+      "epoch": 0.26115209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.7508835142791056e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.732649873431291e-05,
+      "loss": 0.9626,
       "step": 9203
     },
     {
-      "epoch": 0.25278075306912745,
+      "epoch": 0.2611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.7508247645874736e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.7325873173568258e-05,
+      "loss": 0.8394,
       "step": 9204
     },
     {
-      "epoch": 0.25280821729697067,
+      "epoch": 0.2612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.750766008955025e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.7325247550941627e-05,
+      "loss": 1.0064,
       "step": 9205
     },
     {
-      "epoch": 0.25283568152481395,
+      "epoch": 0.2612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.7507072473822252e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.7324621866438297e-05,
+      "loss": 0.9768,
       "step": 9206
     },
     {
-      "epoch": 0.25286314575265717,
+      "epoch": 0.2612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.7506484798695388e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.7323996120063554e-05,
+      "loss": 0.9986,
       "step": 9207
     },
     {
-      "epoch": 0.2528906099805004,
+      "epoch": 0.2612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.7505897064174306e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.7323370311822693e-05,
+      "loss": 0.9078,
       "step": 9208
     },
     {
-      "epoch": 0.2529180742083436,
+      "epoch": 0.26132236095346195,
       "grad_norm": 0.0,
-      "learning_rate": 1.750530927026366e-05,
-      "loss": 0.9395,
+      "learning_rate": 1.732274444172099e-05,
+      "loss": 1.158,
       "step": 9209
     },
     {
-      "epoch": 0.2529455384361869,
+      "epoch": 0.26135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.75047214169681e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.7322118509763734e-05,
+      "loss": 1.0223,
       "step": 9210
     },
     {
-      "epoch": 0.2529730026640301,
+      "epoch": 0.2613791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.7504133504292277e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.7321492515956216e-05,
+      "loss": 0.9566,
       "step": 9211
     },
     {
-      "epoch": 0.25300046689187333,
+      "epoch": 0.26140749148694664,
       "grad_norm": 0.0,
-      "learning_rate": 1.7503545532240843e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.732086646030372e-05,
+      "loss": 0.9299,
       "step": 9212
     },
     {
-      "epoch": 0.25302793111971655,
+      "epoch": 0.26143586833144156,
       "grad_norm": 0.0,
-      "learning_rate": 1.7502957500818453e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.7320240342811537e-05,
+      "loss": 1.0069,
       "step": 9213
     },
     {
-      "epoch": 0.25305539534755983,
+      "epoch": 0.2614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.7502369410029753e-05,
-      "loss": 1.0745,
+      "learning_rate": 1.7319614163484956e-05,
+      "loss": 0.9531,
       "step": 9214
     },
     {
-      "epoch": 0.25308285957540305,
+      "epoch": 0.26149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.7501781259879402e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.7318987922329263e-05,
+      "loss": 0.9904,
       "step": 9215
     },
     {
-      "epoch": 0.2531103238032463,
+      "epoch": 0.26152099886492625,
       "grad_norm": 0.0,
-      "learning_rate": 1.750119305037205e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.731836161934975e-05,
+      "loss": 1.0216,
       "step": 9216
     },
     {
-      "epoch": 0.2531377880310895,
+      "epoch": 0.2615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.7500604781512354e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.731773525455171e-05,
+      "loss": 0.9755,
       "step": 9217
     },
     {
-      "epoch": 0.2531652522589327,
+      "epoch": 0.261577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.7500016453304968e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.731710882794043e-05,
+      "loss": 0.9045,
       "step": 9218
     },
     {
-      "epoch": 0.253192716486776,
+      "epoch": 0.2616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.7499428065754545e-05,
-      "loss": 0.9571,
+      "learning_rate": 1.7316482339521204e-05,
+      "loss": 1.0804,
       "step": 9219
     },
     {
-      "epoch": 0.2532201807146192,
+      "epoch": 0.2616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.7498839618865743e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.7315855789299322e-05,
+      "loss": 0.9224,
       "step": 9220
     },
     {
-      "epoch": 0.25324764494246244,
+      "epoch": 0.26166288308740066,
       "grad_norm": 0.0,
-      "learning_rate": 1.749825111264322e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.7315229177280075e-05,
+      "loss": 0.9694,
       "step": 9221
     },
     {
-      "epoch": 0.25327510917030566,
+      "epoch": 0.26169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.7497662547091624e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.7314602503468758e-05,
+      "loss": 0.8924,
       "step": 9222
     },
     {
-      "epoch": 0.25330257339814893,
+      "epoch": 0.2617196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.749707392221562e-05,
-      "loss": 1.0493,
+      "learning_rate": 1.7313975767870666e-05,
+      "loss": 0.9369,
       "step": 9223
     },
     {
-      "epoch": 0.25333003762599215,
+      "epoch": 0.26174801362088534,
       "grad_norm": 0.0,
-      "learning_rate": 1.7496485238019863e-05,
-      "loss": 1.111,
+      "learning_rate": 1.7313348970491093e-05,
+      "loss": 0.9501,
       "step": 9224
     },
     {
-      "epoch": 0.2533575018538354,
+      "epoch": 0.26177639046538026,
       "grad_norm": 0.0,
-      "learning_rate": 1.7495896494509014e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.7312722111335333e-05,
+      "loss": 0.998,
       "step": 9225
     },
     {
-      "epoch": 0.2533849660816786,
+      "epoch": 0.2618047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.7495307691687726e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.731209519040868e-05,
+      "loss": 0.9966,
       "step": 9226
     },
     {
-      "epoch": 0.2534124303095219,
+      "epoch": 0.26183314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.749471882956066e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.731146820771643e-05,
+      "loss": 1.0186,
       "step": 9227
     },
     {
-      "epoch": 0.2534398945373651,
+      "epoch": 0.26186152099886495,
       "grad_norm": 0.0,
-      "learning_rate": 1.7494129908132475e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.731084116326388e-05,
+      "loss": 1.0185,
       "step": 9228
     },
     {
-      "epoch": 0.2534673587652083,
+      "epoch": 0.2618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.7493540927407833e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.7310214057056326e-05,
+      "loss": 0.9804,
       "step": 9229
     },
     {
-      "epoch": 0.25349482299305154,
+      "epoch": 0.2619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.7492951887391393e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.7309586889099063e-05,
+      "loss": 1.0028,
       "step": 9230
     },
     {
-      "epoch": 0.25352228722089476,
+      "epoch": 0.2619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.749236278808781e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.7308959659397393e-05,
+      "loss": 0.8694,
       "step": 9231
     },
     {
-      "epoch": 0.25354975144873804,
+      "epoch": 0.2619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.7491773629501754e-05,
-      "loss": 1.0515,
+      "learning_rate": 1.730833236795661e-05,
+      "loss": 1.002,
       "step": 9232
     },
     {
-      "epoch": 0.25357721567658126,
+      "epoch": 0.2620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.7491184411637886e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.730770501478202e-05,
+      "loss": 0.8429,
       "step": 9233
     },
     {
-      "epoch": 0.2536046799044245,
+      "epoch": 0.2620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.7490595134500858e-05,
-      "loss": 1.0943,
+      "learning_rate": 1.730707759987891e-05,
+      "loss": 1.0479,
       "step": 9234
     },
     {
-      "epoch": 0.2536321441322677,
+      "epoch": 0.2620601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.7490005798095345e-05,
-      "loss": 0.9208,
+      "learning_rate": 1.7306450123252592e-05,
+      "loss": 0.983,
       "step": 9235
     },
     {
-      "epoch": 0.253659608360111,
+      "epoch": 0.26208853575482405,
       "grad_norm": 0.0,
-      "learning_rate": 1.7489416402426003e-05,
-      "loss": 1.0231,
+      "learning_rate": 1.730582258490836e-05,
+      "loss": 1.0597,
       "step": 9236
     },
     {
-      "epoch": 0.2536870725879542,
+      "epoch": 0.26211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.7488826947497498e-05,
-      "loss": 0.9881,
+      "learning_rate": 1.7305194984851515e-05,
+      "loss": 1.02,
       "step": 9237
     },
     {
-      "epoch": 0.2537145368157974,
+      "epoch": 0.2621452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.748823743331449e-05,
-      "loss": 1.018,
+      "learning_rate": 1.7304567323087362e-05,
+      "loss": 1.0718,
       "step": 9238
     },
     {
-      "epoch": 0.25374200104364064,
+      "epoch": 0.26217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7487647859881655e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.73039395996212e-05,
+      "loss": 0.9708,
       "step": 9239
     },
     {
-      "epoch": 0.2537694652714839,
+      "epoch": 0.26220204313280365,
       "grad_norm": 0.0,
-      "learning_rate": 1.748705822720364e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.730331181445833e-05,
+      "loss": 1.0368,
       "step": 9240
     },
     {
-      "epoch": 0.25379692949932714,
+      "epoch": 0.2622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.748646853528513e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.7302683967604062e-05,
+      "loss": 0.9983,
       "step": 9241
     },
     {
-      "epoch": 0.25382439372717036,
+      "epoch": 0.2622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.748587878413077e-05,
-      "loss": 1.0581,
+      "learning_rate": 1.730205605906369e-05,
+      "loss": 0.8524,
       "step": 9242
     },
     {
-      "epoch": 0.2538518579550136,
+      "epoch": 0.2622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.7485288973745245e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.730142808884252e-05,
+      "loss": 0.9476,
       "step": 9243
     },
     {
-      "epoch": 0.2538793221828568,
+      "epoch": 0.2623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.748469910413321e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.7300800056945865e-05,
+      "loss": 1.0072,
       "step": 9244
     },
     {
-      "epoch": 0.2539067864107001,
+      "epoch": 0.2623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.7484109175299342e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.730017196337902e-05,
+      "loss": 1.0895,
       "step": 9245
     },
     {
-      "epoch": 0.2539342506385433,
+      "epoch": 0.262372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.7483519187248297e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.729954380814729e-05,
+      "loss": 1.0362,
       "step": 9246
     },
     {
-      "epoch": 0.2539617148663865,
+      "epoch": 0.2624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.7482929139984752e-05,
-      "loss": 1.092,
+      "learning_rate": 1.7298915591255993e-05,
+      "loss": 0.9064,
       "step": 9247
     },
     {
-      "epoch": 0.25398917909422974,
+      "epoch": 0.26242905788876275,
       "grad_norm": 0.0,
-      "learning_rate": 1.7482339033513374e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.7298287312710423e-05,
+      "loss": 0.9333,
       "step": 9248
     },
     {
-      "epoch": 0.254016643322073,
+      "epoch": 0.26245743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 1.748174886783883e-05,
-      "loss": 1.0786,
+      "learning_rate": 1.7297658972515895e-05,
+      "loss": 1.1161,
       "step": 9249
     },
     {
-      "epoch": 0.25404410754991624,
+      "epoch": 0.2624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.748115864296579e-05,
-      "loss": 0.904,
+      "learning_rate": 1.729703057067771e-05,
+      "loss": 0.9807,
       "step": 9250
     },
     {
-      "epoch": 0.25407157177775946,
+      "epoch": 0.26251418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.7480568358898924e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.7296402107201182e-05,
+      "loss": 0.9396,
       "step": 9251
     },
     {
-      "epoch": 0.2540990360056027,
+      "epoch": 0.26254256526674236,
       "grad_norm": 0.0,
-      "learning_rate": 1.7479978015642908e-05,
-      "loss": 0.9992,
+      "learning_rate": 1.729577358209162e-05,
+      "loss": 1.0162,
       "step": 9252
     },
     {
-      "epoch": 0.25412650023344596,
+      "epoch": 0.2625709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.7479387613202402e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.729514499535433e-05,
+      "loss": 0.8221,
       "step": 9253
     },
     {
-      "epoch": 0.2541539644612892,
+      "epoch": 0.26259931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 1.747879715158209e-05,
-      "loss": 0.9478,
+      "learning_rate": 1.7294516346994615e-05,
+      "loss": 0.9553,
       "step": 9254
     },
     {
-      "epoch": 0.2541814286891324,
+      "epoch": 0.262627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.7478206630786636e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.7293887637017798e-05,
+      "loss": 0.8331,
       "step": 9255
     },
     {
-      "epoch": 0.2542088929169756,
+      "epoch": 0.2626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.7477616050820714e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.7293258865429186e-05,
+      "loss": 0.9341,
       "step": 9256
     },
     {
-      "epoch": 0.25423635714481885,
+      "epoch": 0.2626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.7477025411688997e-05,
-      "loss": 1.023,
+      "learning_rate": 1.7292630032234086e-05,
+      "loss": 0.8464,
       "step": 9257
     },
     {
-      "epoch": 0.2542638213726621,
+      "epoch": 0.2627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.747643471339616e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.7292001137437814e-05,
+      "loss": 0.9588,
       "step": 9258
     },
     {
-      "epoch": 0.25429128560050535,
+      "epoch": 0.2627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.747584395594687e-05,
-      "loss": 0.9938,
+      "learning_rate": 1.7291372181045678e-05,
+      "loss": 0.8554,
       "step": 9259
     },
     {
-      "epoch": 0.25431874982834857,
+      "epoch": 0.26276958002270145,
       "grad_norm": 0.0,
-      "learning_rate": 1.7475253139345814e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.7290743163062995e-05,
+      "loss": 0.9484,
       "step": 9260
     },
     {
-      "epoch": 0.2543462140561918,
+      "epoch": 0.26279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.7474662263597656e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.7290114083495074e-05,
+      "loss": 0.8868,
       "step": 9261
     },
     {
-      "epoch": 0.25437367828403507,
+      "epoch": 0.2628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.7474071328707073e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.7289484942347238e-05,
+      "loss": 1.0292,
       "step": 9262
     },
     {
-      "epoch": 0.2544011425118783,
+      "epoch": 0.26285471055618614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7473480334678747e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.728885573962479e-05,
+      "loss": 1.0001,
       "step": 9263
     },
     {
-      "epoch": 0.2544286067397215,
+      "epoch": 0.26288308740068106,
       "grad_norm": 0.0,
-      "learning_rate": 1.747288928151735e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.728822647533305e-05,
+      "loss": 0.9054,
       "step": 9264
     },
     {
-      "epoch": 0.25445607096756473,
+      "epoch": 0.2629114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 1.7472298169227554e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.7287597149477336e-05,
+      "loss": 1.0363,
       "step": 9265
     },
     {
-      "epoch": 0.254483535195408,
+      "epoch": 0.26293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.747170699781404e-05,
-      "loss": 0.9129,
+      "learning_rate": 1.7286967762062958e-05,
+      "loss": 1.0357,
       "step": 9266
     },
     {
-      "epoch": 0.2545109994232512,
+      "epoch": 0.2629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.747111576728149e-05,
-      "loss": 0.998,
+      "learning_rate": 1.7286338313095237e-05,
+      "loss": 0.9485,
       "step": 9267
     },
     {
-      "epoch": 0.25453846365109445,
+      "epoch": 0.2629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.7470524477634575e-05,
-      "loss": 1.1207,
+      "learning_rate": 1.728570880257949e-05,
+      "loss": 0.961,
       "step": 9268
     },
     {
-      "epoch": 0.25456592787893767,
+      "epoch": 0.2630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.7469933128877978e-05,
-      "loss": 1.056,
+      "learning_rate": 1.7285079230521034e-05,
+      "loss": 0.905,
       "step": 9269
     },
     {
-      "epoch": 0.2545933921067809,
+      "epoch": 0.2630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.7469341721016374e-05,
-      "loss": 1.0962,
+      "learning_rate": 1.7284449596925183e-05,
+      "loss": 0.9652,
       "step": 9270
     },
     {
-      "epoch": 0.25462085633462417,
+      "epoch": 0.2630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.746875025405445e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.728381990179726e-05,
+      "loss": 0.8705,
       "step": 9271
     },
     {
-      "epoch": 0.2546483205624674,
+      "epoch": 0.26311010215664016,
       "grad_norm": 0.0,
-      "learning_rate": 1.7468158727996875e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.7283190145142582e-05,
+      "loss": 1.0064,
       "step": 9272
     },
     {
-      "epoch": 0.2546757847903106,
+      "epoch": 0.2631384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.746756714284834e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.7282560326966475e-05,
+      "loss": 0.9745,
       "step": 9273
     },
     {
-      "epoch": 0.25470324901815383,
+      "epoch": 0.26316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.746697549861352e-05,
-      "loss": 1.0652,
+      "learning_rate": 1.7281930447274247e-05,
+      "loss": 0.9886,
       "step": 9274
     },
     {
-      "epoch": 0.2547307132459971,
+      "epoch": 0.26319523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.7466383795297096e-05,
-      "loss": 1.1061,
+      "learning_rate": 1.728130050607123e-05,
+      "loss": 0.8678,
       "step": 9275
     },
     {
-      "epoch": 0.25475817747384033,
+      "epoch": 0.26322360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 1.746579203290375e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.7280670503362732e-05,
+      "loss": 0.995,
       "step": 9276
     },
     {
-      "epoch": 0.25478564170168355,
+      "epoch": 0.2632519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.746520021143817e-05,
-      "loss": 1.1213,
+      "learning_rate": 1.728004043915409e-05,
+      "loss": 0.9945,
       "step": 9277
     },
     {
-      "epoch": 0.2548131059295268,
+      "epoch": 0.26328036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.746460833090503e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.727941031345062e-05,
+      "loss": 1.0094,
       "step": 9278
     },
     {
-      "epoch": 0.25484057015737005,
+      "epoch": 0.26330874006810445,
       "grad_norm": 0.0,
-      "learning_rate": 1.746401639130902e-05,
-      "loss": 0.968,
+      "learning_rate": 1.727878012625764e-05,
+      "loss": 1.0638,
       "step": 9279
     },
     {
-      "epoch": 0.25486803438521327,
+      "epoch": 0.2633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.746342439265482e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.727814987758048e-05,
+      "loss": 0.9515,
       "step": 9280
     },
     {
-      "epoch": 0.2548954986130565,
+      "epoch": 0.2633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.7462832334947113e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.7277519567424458e-05,
+      "loss": 0.9443,
       "step": 9281
     },
     {
-      "epoch": 0.2549229628408997,
+      "epoch": 0.2633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.7462240218190588e-05,
-      "loss": 1.0542,
+      "learning_rate": 1.7276889195794905e-05,
+      "loss": 1.0428,
       "step": 9282
     },
     {
-      "epoch": 0.25495042706874294,
+      "epoch": 0.263422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.746164804238993e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.727625876269714e-05,
+      "loss": 0.9746,
       "step": 9283
     },
     {
-      "epoch": 0.2549778912965862,
+      "epoch": 0.26345062429057886,
       "grad_norm": 0.0,
-      "learning_rate": 1.746105580754982e-05,
-      "loss": 0.9128,
+      "learning_rate": 1.7275628268136487e-05,
+      "loss": 1.0823,
       "step": 9284
     },
     {
-      "epoch": 0.25500535552442943,
+      "epoch": 0.2634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.7460463513674947e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.7274997712118276e-05,
+      "loss": 0.9008,
       "step": 9285
     },
     {
-      "epoch": 0.25503281975227265,
+      "epoch": 0.2635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.7459871160769993e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.7274367094647836e-05,
+      "loss": 1.0526,
       "step": 9286
     },
     {
-      "epoch": 0.2550602839801159,
+      "epoch": 0.26353575482406355,
       "grad_norm": 0.0,
-      "learning_rate": 1.7459278748839654e-05,
-      "loss": 1.069,
+      "learning_rate": 1.7273736415730488e-05,
+      "loss": 1.0591,
       "step": 9287
     },
     {
-      "epoch": 0.25508774820795915,
+      "epoch": 0.26356413166855847,
       "grad_norm": 0.0,
-      "learning_rate": 1.7458686277888607e-05,
-      "loss": 1.0929,
+      "learning_rate": 1.7273105675371562e-05,
+      "loss": 0.9861,
       "step": 9288
     },
     {
-      "epoch": 0.2551152124358024,
+      "epoch": 0.2635925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.7458093747921553e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.7272474873576385e-05,
+      "loss": 0.9512,
       "step": 9289
     },
     {
-      "epoch": 0.2551426766636456,
+      "epoch": 0.26362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.745750115894316e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.727184401035029e-05,
+      "loss": 1.0558,
       "step": 9290
     },
     {
-      "epoch": 0.2551701408914888,
+      "epoch": 0.26364926220204316,
       "grad_norm": 0.0,
-      "learning_rate": 1.7456908510958142e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.7271213085698593e-05,
+      "loss": 0.9839,
       "step": 9291
     },
     {
-      "epoch": 0.2551976051193321,
+      "epoch": 0.263677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.7456315803971166e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.7270582099626637e-05,
+      "loss": 0.9518,
       "step": 9292
     },
     {
-      "epoch": 0.2552250693471753,
+      "epoch": 0.26370601589103293,
       "grad_norm": 0.0,
-      "learning_rate": 1.7455723037986937e-05,
-      "loss": 0.8291,
+      "learning_rate": 1.7269951052139745e-05,
+      "loss": 0.9352,
       "step": 9293
     },
     {
-      "epoch": 0.25525253357501854,
+      "epoch": 0.2637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.7455130213010135e-05,
-      "loss": 0.9686,
+      "learning_rate": 1.7269319943243252e-05,
+      "loss": 0.9463,
       "step": 9294
     },
     {
-      "epoch": 0.25527999780286176,
+      "epoch": 0.2637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.7454537329045457e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.7268688772942487e-05,
+      "loss": 1.0039,
       "step": 9295
     },
     {
-      "epoch": 0.255307462030705,
+      "epoch": 0.2637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.7453944386097593e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.7268057541242778e-05,
+      "loss": 0.9784,
       "step": 9296
     },
     {
-      "epoch": 0.25533492625854826,
+      "epoch": 0.2638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.745335138417123e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.7267426248149467e-05,
+      "loss": 0.9904,
       "step": 9297
     },
     {
-      "epoch": 0.2553623904863915,
+      "epoch": 0.2638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.7452758323271068e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.7266794893667877e-05,
+      "loss": 0.9969,
       "step": 9298
     },
     {
-      "epoch": 0.2553898547142347,
+      "epoch": 0.26387627695800225,
       "grad_norm": 0.0,
-      "learning_rate": 1.7452165203401792e-05,
-      "loss": 1.0173,
+      "learning_rate": 1.726616347780334e-05,
+      "loss": 0.9223,
       "step": 9299
     },
     {
-      "epoch": 0.2554173189420779,
+      "epoch": 0.26390465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.74515720245681e-05,
-      "loss": 1.0538,
+      "learning_rate": 1.7265532000561193e-05,
+      "loss": 1.0209,
       "step": 9300
     },
     {
-      "epoch": 0.2554447831699212,
+      "epoch": 0.26393303064699203,
       "grad_norm": 0.0,
-      "learning_rate": 1.7450978786774684e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.7264900461946777e-05,
+      "loss": 1.0285,
       "step": 9301
     },
     {
-      "epoch": 0.2554722473977644,
+      "epoch": 0.26396140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.7450385490026236e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.7264268861965416e-05,
+      "loss": 0.916,
       "step": 9302
     },
     {
-      "epoch": 0.25549971162560764,
+      "epoch": 0.26398978433598186,
       "grad_norm": 0.0,
-      "learning_rate": 1.7449792134327453e-05,
-      "loss": 1.0001,
+      "learning_rate": 1.7263637200622447e-05,
+      "loss": 1.0321,
       "step": 9303
     },
     {
-      "epoch": 0.25552717585345086,
+      "epoch": 0.2640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.7449198719683028e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.726300547792321e-05,
+      "loss": 0.9254,
       "step": 9304
     },
     {
-      "epoch": 0.25555464008129414,
+      "epoch": 0.26404653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 1.744860524609766e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.726237369387304e-05,
+      "loss": 1.0241,
       "step": 9305
     },
     {
-      "epoch": 0.25558210430913736,
+      "epoch": 0.2640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.744801171357604e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.7261741848477277e-05,
+      "loss": 1.0356,
       "step": 9306
     },
     {
-      "epoch": 0.2556095685369806,
+      "epoch": 0.2641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7447418122122863e-05,
-      "loss": 1.0265,
+      "learning_rate": 1.726110994174125e-05,
+      "loss": 0.8735,
       "step": 9307
     },
     {
-      "epoch": 0.2556370327648238,
+      "epoch": 0.2641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.7446824471742833e-05,
-      "loss": 1.0455,
+      "learning_rate": 1.72604779736703e-05,
+      "loss": 0.8722,
       "step": 9308
     },
     {
-      "epoch": 0.2556644969926671,
+      "epoch": 0.2641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.7446230762440642e-05,
-      "loss": 0.955,
+      "learning_rate": 1.7259845944269765e-05,
+      "loss": 1.0392,
       "step": 9309
     },
     {
-      "epoch": 0.2556919612205103,
+      "epoch": 0.2641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.7445636994220988e-05,
-      "loss": 0.8837,
+      "learning_rate": 1.7259213853544992e-05,
+      "loss": 0.989,
       "step": 9310
     },
     {
-      "epoch": 0.2557194254483535,
+      "epoch": 0.26421679909194096,
       "grad_norm": 0.0,
-      "learning_rate": 1.7445043167088573e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.7258581701501305e-05,
+      "loss": 0.9407,
       "step": 9311
     },
     {
-      "epoch": 0.25574688967619674,
+      "epoch": 0.26424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.744444928104809e-05,
-      "loss": 1.0025,
+      "learning_rate": 1.7257949488144058e-05,
+      "loss": 0.8298,
       "step": 9312
     },
     {
-      "epoch": 0.25577435390403996,
+      "epoch": 0.2642735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.7443855336104243e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.7257317213478583e-05,
+      "loss": 0.9978,
       "step": 9313
     },
     {
-      "epoch": 0.25580181813188324,
+      "epoch": 0.26430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.7443261332261724e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.7256684877510223e-05,
+      "loss": 1.0486,
       "step": 9314
     },
     {
-      "epoch": 0.25582928235972646,
+      "epoch": 0.26433030646992056,
       "grad_norm": 0.0,
-      "learning_rate": 1.744266726952524e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.725605248024432e-05,
+      "loss": 0.9142,
       "step": 9315
     },
     {
-      "epoch": 0.2558567465875697,
+      "epoch": 0.2643586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.744207314789949e-05,
-      "loss": 1.195,
+      "learning_rate": 1.7255420021686214e-05,
+      "loss": 0.9787,
       "step": 9316
     },
     {
-      "epoch": 0.2558842108154129,
+      "epoch": 0.26438706015891034,
       "grad_norm": 0.0,
-      "learning_rate": 1.744147896738918e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.7254787501841252e-05,
+      "loss": 0.9949,
       "step": 9317
     },
     {
-      "epoch": 0.2559116750432562,
+      "epoch": 0.2644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.7440884727998998e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.725415492071477e-05,
+      "loss": 0.9019,
       "step": 9318
     },
     {
-      "epoch": 0.2559391392710994,
+      "epoch": 0.2644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.7440290429733657e-05,
-      "loss": 1.0071,
+      "learning_rate": 1.725352227831212e-05,
+      "loss": 0.8635,
       "step": 9319
     },
     {
-      "epoch": 0.2559666034989426,
+      "epoch": 0.264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.7439696072597856e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.725288957463864e-05,
+      "loss": 1.01,
       "step": 9320
     },
     {
-      "epoch": 0.25599406772678585,
+      "epoch": 0.2645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.7439101656596298e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.7252256809699672e-05,
+      "loss": 0.954,
       "step": 9321
     },
     {
-      "epoch": 0.2560215319546291,
+      "epoch": 0.2645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.7438507181733685e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.7251623983500566e-05,
+      "loss": 1.0459,
       "step": 9322
     },
     {
-      "epoch": 0.25604899618247234,
+      "epoch": 0.26455732122587966,
       "grad_norm": 0.0,
-      "learning_rate": 1.7437912648014724e-05,
-      "loss": 1.1216,
+      "learning_rate": 1.725099109604667e-05,
+      "loss": 1.0792,
       "step": 9323
     },
     {
-      "epoch": 0.25607646041031557,
+      "epoch": 0.2645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.7437318055444116e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.725035814734332e-05,
+      "loss": 0.9249,
       "step": 9324
     },
     {
-      "epoch": 0.2561039246381588,
+      "epoch": 0.2646140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.7436723404026566e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.7249725137395876e-05,
+      "loss": 0.8573,
       "step": 9325
     },
     {
-      "epoch": 0.256131388866002,
+      "epoch": 0.26464245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 1.743612869376678e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.7249092066209674e-05,
+      "loss": 0.9313,
       "step": 9326
     },
     {
-      "epoch": 0.2561588530938453,
+      "epoch": 0.26467082860385927,
       "grad_norm": 0.0,
-      "learning_rate": 1.7435533924669463e-05,
-      "loss": 1.0551,
+      "learning_rate": 1.7248458933790068e-05,
+      "loss": 1.0219,
       "step": 9327
     },
     {
-      "epoch": 0.2561863173216885,
+      "epoch": 0.2646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.743493909673932e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.72478257401424e-05,
+      "loss": 0.9731,
       "step": 9328
     },
     {
-      "epoch": 0.2562137815495317,
+      "epoch": 0.26472758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.743434420998106e-05,
-      "loss": 1.0544,
+      "learning_rate": 1.724719248527202e-05,
+      "loss": 0.993,
       "step": 9329
     },
     {
-      "epoch": 0.25624124577737495,
+      "epoch": 0.26475595913734395,
       "grad_norm": 0.0,
-      "learning_rate": 1.743374926439939e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.7246559169184284e-05,
+      "loss": 0.9265,
       "step": 9330
     },
     {
-      "epoch": 0.2562687100052182,
+      "epoch": 0.2647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.7433154259999018e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.7245925791884538e-05,
+      "loss": 0.9473,
       "step": 9331
     },
     {
-      "epoch": 0.25629617423306145,
+      "epoch": 0.26481271282633373,
       "grad_norm": 0.0,
-      "learning_rate": 1.7432559196784647e-05,
-      "loss": 1.0746,
+      "learning_rate": 1.7245292353378128e-05,
+      "loss": 0.8876,
       "step": 9332
     },
     {
-      "epoch": 0.25632363846090467,
+      "epoch": 0.2648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.7431964074760988e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.724465885367041e-05,
+      "loss": 1.0649,
       "step": 9333
     },
     {
-      "epoch": 0.2563511026887479,
+      "epoch": 0.2648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.743136889393275e-05,
-      "loss": 1.0709,
+      "learning_rate": 1.724402529276673e-05,
+      "loss": 0.9272,
       "step": 9334
     },
     {
-      "epoch": 0.25637856691659117,
+      "epoch": 0.26489784335981836,
       "grad_norm": 0.0,
-      "learning_rate": 1.7430773654304646e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.7243391670672445e-05,
+      "loss": 1.131,
       "step": 9335
     },
     {
-      "epoch": 0.2564060311444344,
+      "epoch": 0.2649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.743017835588138e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.7242757987392903e-05,
+      "loss": 1.0486,
       "step": 9336
     },
     {
-      "epoch": 0.2564334953722776,
+      "epoch": 0.2649545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.7429582998667665e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.7242124242933462e-05,
+      "loss": 0.9348,
       "step": 9337
     },
     {
-      "epoch": 0.25646095960012083,
+      "epoch": 0.26498297389330305,
       "grad_norm": 0.0,
-      "learning_rate": 1.742898758266821e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.724149043729947e-05,
+      "loss": 0.849,
       "step": 9338
     },
     {
-      "epoch": 0.25648842382796405,
+      "epoch": 0.26501135073779797,
       "grad_norm": 0.0,
-      "learning_rate": 1.7428392107887727e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.7240856570496283e-05,
+      "loss": 1.0269,
       "step": 9339
     },
     {
-      "epoch": 0.25651588805580733,
+      "epoch": 0.26503972758229283,
       "grad_norm": 0.0,
-      "learning_rate": 1.7427796574330932e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.724022264252925e-05,
+      "loss": 1.0272,
       "step": 9340
     },
     {
-      "epoch": 0.25654335228365055,
+      "epoch": 0.26506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.7427200982002533e-05,
-      "loss": 0.9052,
+      "learning_rate": 1.7239588653403737e-05,
+      "loss": 1.0032,
       "step": 9341
     },
     {
-      "epoch": 0.25657081651149377,
+      "epoch": 0.26509648127128266,
       "grad_norm": 0.0,
-      "learning_rate": 1.742660533090724e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.7238954603125093e-05,
+      "loss": 0.9149,
       "step": 9342
     },
     {
-      "epoch": 0.256598280739337,
+      "epoch": 0.2651248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.742600962104977e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.7238320491698674e-05,
+      "loss": 1.0194,
       "step": 9343
     },
     {
-      "epoch": 0.25662574496718027,
+      "epoch": 0.26515323496027243,
       "grad_norm": 0.0,
-      "learning_rate": 1.7425413852434835e-05,
-      "loss": 1.0598,
+      "learning_rate": 1.7237686319129835e-05,
+      "loss": 0.9307,
       "step": 9344
     },
     {
-      "epoch": 0.2566532091950235,
+      "epoch": 0.2651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.7424818025067153e-05,
-      "loss": 0.973,
+      "learning_rate": 1.7237052085423937e-05,
+      "loss": 0.9677,
       "step": 9345
     },
     {
-      "epoch": 0.2566806734228667,
+      "epoch": 0.2652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.742422213895143e-05,
-      "loss": 0.924,
+      "learning_rate": 1.723641779058633e-05,
+      "loss": 0.9522,
       "step": 9346
     },
     {
-      "epoch": 0.25670813765070993,
+      "epoch": 0.26523836549375707,
       "grad_norm": 0.0,
-      "learning_rate": 1.7423626194092387e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.723578343462238e-05,
+      "loss": 0.8301,
       "step": 9347
     },
     {
-      "epoch": 0.2567356018785532,
+      "epoch": 0.265266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.742303019049474e-05,
-      "loss": 1.0037,
+      "learning_rate": 1.723514901753744e-05,
+      "loss": 0.9649,
       "step": 9348
     },
     {
-      "epoch": 0.25676306610639643,
+      "epoch": 0.2652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.7422434128163197e-05,
-      "loss": 1.0882,
+      "learning_rate": 1.723451453933687e-05,
+      "loss": 0.9919,
       "step": 9349
     },
     {
-      "epoch": 0.25679053033423965,
+      "epoch": 0.26532349602724176,
       "grad_norm": 0.0,
-      "learning_rate": 1.7421838007102483e-05,
-      "loss": 0.9282,
+      "learning_rate": 1.7233880000026034e-05,
+      "loss": 1.0782,
       "step": 9350
     },
     {
-      "epoch": 0.2568179945620829,
+      "epoch": 0.26535187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.7421241827317314e-05,
-      "loss": 1.1208,
+      "learning_rate": 1.723324539961029e-05,
+      "loss": 1.0051,
       "step": 9351
     },
     {
-      "epoch": 0.2568454587899261,
+      "epoch": 0.26538024971623153,
       "grad_norm": 0.0,
-      "learning_rate": 1.7420645588812402e-05,
-      "loss": 1.0341,
+      "learning_rate": 1.7232610738094992e-05,
+      "loss": 0.9897,
       "step": 9352
     },
     {
-      "epoch": 0.2568729230177694,
+      "epoch": 0.26540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.742004929159247e-05,
-      "loss": 1.0991,
+      "learning_rate": 1.7231976015485505e-05,
+      "loss": 1.0836,
       "step": 9353
     },
     {
-      "epoch": 0.2569003872456126,
+      "epoch": 0.26543700340522136,
       "grad_norm": 0.0,
-      "learning_rate": 1.7419452935662233e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.72313412317872e-05,
+      "loss": 0.9338,
       "step": 9354
     },
     {
-      "epoch": 0.2569278514734558,
+      "epoch": 0.2654653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.741885652102641e-05,
-      "loss": 0.8755,
+      "learning_rate": 1.723070638700542e-05,
+      "loss": 0.9218,
       "step": 9355
     },
     {
-      "epoch": 0.25695531570129904,
+      "epoch": 0.26549375709421114,
       "grad_norm": 0.0,
-      "learning_rate": 1.741826004768972e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.7230071481145546e-05,
+      "loss": 1.0128,
       "step": 9356
     },
     {
-      "epoch": 0.2569827799291423,
+      "epoch": 0.265522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.7417663515656883e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.7229436514212926e-05,
+      "loss": 1.0416,
       "step": 9357
     },
     {
-      "epoch": 0.25701024415698553,
+      "epoch": 0.2655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.7417066924932623e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.7228801486212937e-05,
+      "loss": 0.9667,
       "step": 9358
     },
     {
-      "epoch": 0.25703770838482876,
+      "epoch": 0.2655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.741647027552165e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.7228166397150932e-05,
+      "loss": 0.946,
       "step": 9359
     },
     {
-      "epoch": 0.257065172612672,
+      "epoch": 0.2656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.7415873567428696e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.7227531247032286e-05,
+      "loss": 0.9508,
       "step": 9360
     },
     {
-      "epoch": 0.25709263684051525,
+      "epoch": 0.2656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.7415276800658473e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.7226896035862357e-05,
+      "loss": 1.0574,
       "step": 9361
     },
     {
-      "epoch": 0.2571201010683585,
+      "epoch": 0.26566401816118046,
       "grad_norm": 0.0,
-      "learning_rate": 1.741467997521571e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.7226260763646513e-05,
+      "loss": 0.8796,
       "step": 9362
     },
     {
-      "epoch": 0.2571475652962017,
+      "epoch": 0.2656923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.741408309110513e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.7225625430390117e-05,
+      "loss": 0.9946,
       "step": 9363
     },
     {
-      "epoch": 0.2571750295240449,
+      "epoch": 0.26572077185017023,
       "grad_norm": 0.0,
-      "learning_rate": 1.7413486148331448e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.7224990036098536e-05,
+      "loss": 0.829,
       "step": 9364
     },
     {
-      "epoch": 0.25720249375188814,
+      "epoch": 0.26574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.7412889146899393e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.7224354580777144e-05,
+      "loss": 0.9834,
       "step": 9365
     },
     {
-      "epoch": 0.2572299579797314,
+      "epoch": 0.26577752553916006,
       "grad_norm": 0.0,
-      "learning_rate": 1.7412292086813688e-05,
-      "loss": 1.0691,
+      "learning_rate": 1.7223719064431304e-05,
+      "loss": 0.9991,
       "step": 9366
     },
     {
-      "epoch": 0.25725742220757464,
+      "epoch": 0.2658059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.741169496807906e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.7223083487066382e-05,
+      "loss": 1.0188,
       "step": 9367
     },
     {
-      "epoch": 0.25728488643541786,
+      "epoch": 0.26583427922814984,
       "grad_norm": 0.0,
-      "learning_rate": 1.7411097790700227e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.722244784868775e-05,
+      "loss": 0.9554,
       "step": 9368
     },
     {
-      "epoch": 0.2573123506632611,
+      "epoch": 0.2658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.7410500554681916e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.7221812149300773e-05,
+      "loss": 0.9552,
       "step": 9369
     },
     {
-      "epoch": 0.25733981489110436,
+      "epoch": 0.2658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.7409903260028858e-05,
-      "loss": 1.009,
+      "learning_rate": 1.722117638891083e-05,
+      "loss": 0.9896,
       "step": 9370
     },
     {
-      "epoch": 0.2573672791189476,
+      "epoch": 0.26591940976163453,
       "grad_norm": 0.0,
-      "learning_rate": 1.740930590674577e-05,
-      "loss": 1.0531,
+      "learning_rate": 1.7220540567523284e-05,
+      "loss": 0.9965,
       "step": 9371
     },
     {
-      "epoch": 0.2573947433467908,
+      "epoch": 0.2659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.7408708494837386e-05,
-      "loss": 0.9812,
+      "learning_rate": 1.7219904685143506e-05,
+      "loss": 1.1033,
       "step": 9372
     },
     {
-      "epoch": 0.257422207574634,
+      "epoch": 0.2659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.7408111024308433e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.7219268741776868e-05,
+      "loss": 0.9799,
       "step": 9373
     },
     {
-      "epoch": 0.2574496718024773,
+      "epoch": 0.26600454029511916,
       "grad_norm": 0.0,
-      "learning_rate": 1.7407513495163634e-05,
-      "loss": 1.0847,
+      "learning_rate": 1.7218632737428743e-05,
+      "loss": 1.0666,
       "step": 9374
     },
     {
-      "epoch": 0.2574771360303205,
+      "epoch": 0.2660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.740691590740772e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.72179966721045e-05,
+      "loss": 0.9825,
       "step": 9375
     },
     {
-      "epoch": 0.25750460025816374,
+      "epoch": 0.266061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.740631826104542e-05,
-      "loss": 1.0348,
+      "learning_rate": 1.7217360545809518e-05,
+      "loss": 0.922,
       "step": 9376
     },
     {
-      "epoch": 0.25753206448600696,
+      "epoch": 0.26608967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 1.7405720556081456e-05,
-      "loss": 0.9101,
+      "learning_rate": 1.7216724358549165e-05,
+      "loss": 0.9987,
       "step": 9377
     },
     {
-      "epoch": 0.2575595287138502,
+      "epoch": 0.26611804767309877,
       "grad_norm": 0.0,
-      "learning_rate": 1.7405122792520564e-05,
-      "loss": 0.8997,
+      "learning_rate": 1.7216088110328823e-05,
+      "loss": 1.0163,
       "step": 9378
     },
     {
-      "epoch": 0.25758699294169346,
+      "epoch": 0.2661464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.740452497036747e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.7215451801153852e-05,
+      "loss": 0.8994,
       "step": 9379
     },
     {
-      "epoch": 0.2576144571695367,
+      "epoch": 0.26617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.740392708962691e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.721481543102964e-05,
+      "loss": 1.0334,
       "step": 9380
     },
     {
-      "epoch": 0.2576419213973799,
+      "epoch": 0.2662031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.740332915030361e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.7214178999961556e-05,
+      "loss": 0.9859,
       "step": 9381
     },
     {
-      "epoch": 0.2576693856252231,
+      "epoch": 0.2662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.74027311524023e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.7213542507954978e-05,
+      "loss": 0.964,
       "step": 9382
     },
     {
-      "epoch": 0.2576968498530664,
+      "epoch": 0.26625993189557323,
       "grad_norm": 0.0,
-      "learning_rate": 1.7402133095927715e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.7212905955015285e-05,
+      "loss": 0.9849,
       "step": 9383
     },
     {
-      "epoch": 0.2577243140809096,
+      "epoch": 0.2662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.7401534980884586e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.7212269341147845e-05,
+      "loss": 0.9471,
       "step": 9384
     },
     {
-      "epoch": 0.25775177830875284,
+      "epoch": 0.266316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.7400936807277645e-05,
-      "loss": 0.9454,
+      "learning_rate": 1.7211632666358045e-05,
+      "loss": 0.7941,
       "step": 9385
     },
     {
-      "epoch": 0.25777924253659606,
+      "epoch": 0.26634506242905787,
       "grad_norm": 0.0,
-      "learning_rate": 1.7400338575111624e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.7210995930651262e-05,
+      "loss": 1.0847,
       "step": 9386
     },
     {
-      "epoch": 0.25780670676443934,
+      "epoch": 0.2663734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.739974028439126e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.721035913403287e-05,
+      "loss": 0.9108,
       "step": 9387
     },
     {
-      "epoch": 0.25783417099228256,
+      "epoch": 0.2664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.7399141935121283e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.7209722276508253e-05,
+      "loss": 0.9084,
       "step": 9388
     },
     {
-      "epoch": 0.2578616352201258,
+      "epoch": 0.26643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.739854352730643e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.7209085358082785e-05,
+      "loss": 0.9973,
       "step": 9389
     },
     {
-      "epoch": 0.257889099447969,
+      "epoch": 0.26645856980703747,
       "grad_norm": 0.0,
-      "learning_rate": 1.7397945060951432e-05,
-      "loss": 0.9665,
+      "learning_rate": 1.720844837876185e-05,
+      "loss": 1.0007,
       "step": 9390
     },
     {
-      "epoch": 0.2579165636758122,
+      "epoch": 0.26648694665153233,
       "grad_norm": 0.0,
-      "learning_rate": 1.739734653606103e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.7207811338550824e-05,
+      "loss": 0.8613,
       "step": 9391
     },
     {
-      "epoch": 0.2579440279036555,
+      "epoch": 0.26651532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 1.7396747952639956e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.7207174237455095e-05,
+      "loss": 1.075,
       "step": 9392
     },
     {
-      "epoch": 0.2579714921314987,
+      "epoch": 0.26654370034052216,
       "grad_norm": 0.0,
-      "learning_rate": 1.7396149310692944e-05,
-      "loss": 1.021,
+      "learning_rate": 1.720653707548004e-05,
+      "loss": 1.0143,
       "step": 9393
     },
     {
-      "epoch": 0.25799895635934195,
+      "epoch": 0.266572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.7395550610224737e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.7205899852631044e-05,
+      "loss": 0.9501,
       "step": 9394
     },
     {
-      "epoch": 0.25802642058718517,
+      "epoch": 0.26660045402951194,
       "grad_norm": 0.0,
-      "learning_rate": 1.7394951851240066e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.720526256891349e-05,
+      "loss": 0.849,
       "step": 9395
     },
     {
-      "epoch": 0.25805388481502844,
+      "epoch": 0.2666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.7394353033743674e-05,
-      "loss": 0.875,
+      "learning_rate": 1.7204625224332752e-05,
+      "loss": 0.9082,
       "step": 9396
     },
     {
-      "epoch": 0.25808134904287167,
+      "epoch": 0.2666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.7393754157740293e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.7203987818894226e-05,
+      "loss": 1.1013,
       "step": 9397
     },
     {
-      "epoch": 0.2581088132707149,
+      "epoch": 0.26668558456299657,
       "grad_norm": 0.0,
-      "learning_rate": 1.739315522323467e-05,
-      "loss": 0.8088,
+      "learning_rate": 1.720335035260329e-05,
+      "loss": 1.0353,
       "step": 9398
     },
     {
-      "epoch": 0.2581362774985581,
+      "epoch": 0.2667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.7392556230231537e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.720271282546533e-05,
+      "loss": 0.9284,
       "step": 9399
     },
     {
-      "epoch": 0.2581637417264014,
+      "epoch": 0.2667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.7391957178735635e-05,
-      "loss": 1.0181,
+      "learning_rate": 1.7202075237485732e-05,
+      "loss": 0.9028,
       "step": 9400
     },
     {
-      "epoch": 0.2581912059542446,
+      "epoch": 0.26677071509648126,
       "grad_norm": 0.0,
-      "learning_rate": 1.7391358068751705e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.720143758866988e-05,
+      "loss": 1.0701,
       "step": 9401
     },
     {
-      "epoch": 0.25821867018208783,
+      "epoch": 0.2667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.7390758900284487e-05,
-      "loss": 0.9909,
+      "learning_rate": 1.720079987902316e-05,
+      "loss": 0.9598,
       "step": 9402
     },
     {
-      "epoch": 0.25824613440993105,
+      "epoch": 0.26682746878547103,
       "grad_norm": 0.0,
-      "learning_rate": 1.739015967333872e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.720016210855096e-05,
+      "loss": 1.0081,
       "step": 9403
     },
     {
-      "epoch": 0.2582735986377743,
+      "epoch": 0.26685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.738956038791915e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.7199524277258667e-05,
+      "loss": 0.952,
       "step": 9404
     },
     {
-      "epoch": 0.25830106286561755,
+      "epoch": 0.26688422247446086,
       "grad_norm": 0.0,
-      "learning_rate": 1.7388961044030513e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.719888638515167e-05,
+      "loss": 0.9627,
       "step": 9405
     },
     {
-      "epoch": 0.25832852709346077,
+      "epoch": 0.2669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.7388361641677556e-05,
-      "loss": 0.8643,
+      "learning_rate": 1.7198248432235352e-05,
+      "loss": 0.8742,
       "step": 9406
     },
     {
-      "epoch": 0.258355991321304,
+      "epoch": 0.26694097616345064,
       "grad_norm": 0.0,
-      "learning_rate": 1.7387762180865017e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.719761041851511e-05,
+      "loss": 1.0494,
       "step": 9407
     },
     {
-      "epoch": 0.2583834555491472,
+      "epoch": 0.2669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.7387162661597643e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.7196972343996328e-05,
+      "loss": 0.9641,
       "step": 9408
     },
     {
-      "epoch": 0.2584109197769905,
+      "epoch": 0.2669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.7386563083880178e-05,
-      "loss": 1.0529,
+      "learning_rate": 1.7196334208684398e-05,
+      "loss": 0.9985,
       "step": 9409
     },
     {
-      "epoch": 0.2584383840048337,
+      "epoch": 0.26702610669693533,
       "grad_norm": 0.0,
-      "learning_rate": 1.7385963447717365e-05,
-      "loss": 0.8959,
+      "learning_rate": 1.719569601258471e-05,
+      "loss": 1.0222,
       "step": 9410
     },
     {
-      "epoch": 0.25846584823267693,
+      "epoch": 0.2670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.7385363753113944e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.7195057755702653e-05,
+      "loss": 0.9941,
       "step": 9411
     },
     {
-      "epoch": 0.25849331246052015,
+      "epoch": 0.2670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.738476400007467e-05,
-      "loss": 1.052,
+      "learning_rate": 1.719441943804362e-05,
+      "loss": 0.8682,
       "step": 9412
     },
     {
-      "epoch": 0.25852077668836343,
+      "epoch": 0.26711123723041996,
       "grad_norm": 0.0,
-      "learning_rate": 1.7384164188604276e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.7193781059613002e-05,
+      "loss": 1.1422,
       "step": 9413
     },
     {
-      "epoch": 0.25854824091620665,
+      "epoch": 0.2671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.738356431870752e-05,
-      "loss": 1.061,
+      "learning_rate": 1.719314262041619e-05,
+      "loss": 1.0756,
       "step": 9414
     },
     {
-      "epoch": 0.25857570514404987,
+      "epoch": 0.26716799091940974,
       "grad_norm": 0.0,
-      "learning_rate": 1.738296439038914e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.719250412045858e-05,
+      "loss": 1.0452,
       "step": 9415
     },
     {
-      "epoch": 0.2586031693718931,
+      "epoch": 0.26719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.738236440365389e-05,
-      "loss": 0.8619,
+      "learning_rate": 1.719186555974557e-05,
+      "loss": 0.9981,
       "step": 9416
     },
     {
-      "epoch": 0.25863063359973637,
+      "epoch": 0.26722474460839957,
       "grad_norm": 0.0,
-      "learning_rate": 1.7381764358506507e-05,
-      "loss": 0.9264,
+      "learning_rate": 1.7191226938282543e-05,
+      "loss": 0.9507,
       "step": 9417
     },
     {
-      "epoch": 0.2586580978275796,
+      "epoch": 0.2672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.738116425495175e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.71905882560749e-05,
+      "loss": 0.9296,
       "step": 9418
     },
     {
-      "epoch": 0.2586855620554228,
+      "epoch": 0.26728149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.7380564092994357e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.7189949513128033e-05,
+      "loss": 0.9428,
       "step": 9419
     },
     {
-      "epoch": 0.25871302628326603,
+      "epoch": 0.2673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.7379963872639087e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.718931070944734e-05,
+      "loss": 1.0111,
       "step": 9420
     },
     {
-      "epoch": 0.25874049051110926,
+      "epoch": 0.2673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.7379363593890683e-05,
-      "loss": 0.9544,
+      "learning_rate": 1.718867184503822e-05,
+      "loss": 0.9684,
       "step": 9421
     },
     {
-      "epoch": 0.25876795473895253,
+      "epoch": 0.26736662883087403,
       "grad_norm": 0.0,
-      "learning_rate": 1.7378763256753893e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.718803291990606e-05,
+      "loss": 0.9899,
       "step": 9422
     },
     {
-      "epoch": 0.25879541896679575,
+      "epoch": 0.2673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.7378162861233473e-05,
-      "loss": 1.0753,
+      "learning_rate": 1.7187393934056267e-05,
+      "loss": 1.0239,
       "step": 9423
     },
     {
-      "epoch": 0.258822883194639,
+      "epoch": 0.2674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.737756240733417e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.718675488749423e-05,
+      "loss": 1.0011,
       "step": 9424
     },
     {
-      "epoch": 0.2588503474224822,
+      "epoch": 0.26745175936435867,
       "grad_norm": 0.0,
-      "learning_rate": 1.7376961895060735e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.7186115780225357e-05,
+      "loss": 0.9495,
       "step": 9425
     },
     {
-      "epoch": 0.2588778116503255,
+      "epoch": 0.2674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.737636132441792e-05,
-      "loss": 1.1006,
+      "learning_rate": 1.718547661225504e-05,
+      "loss": 0.9043,
       "step": 9426
     },
     {
-      "epoch": 0.2589052758781687,
+      "epoch": 0.26750851305334844,
       "grad_norm": 0.0,
-      "learning_rate": 1.7375760695410476e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.7184837383588675e-05,
+      "loss": 0.8727,
       "step": 9427
     },
     {
-      "epoch": 0.2589327401060119,
+      "epoch": 0.26753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7375160008043157e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.718419809423167e-05,
+      "loss": 1.1013,
       "step": 9428
     },
     {
-      "epoch": 0.25896020433385514,
+      "epoch": 0.26756526674233827,
       "grad_norm": 0.0,
-      "learning_rate": 1.7374559262320716e-05,
-      "loss": 1.0554,
+      "learning_rate": 1.7183558744189416e-05,
+      "loss": 0.9702,
       "step": 9429
     },
     {
-      "epoch": 0.2589876685616984,
+      "epoch": 0.26759364358683313,
       "grad_norm": 0.0,
-      "learning_rate": 1.7373958458247906e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.7182919333467323e-05,
+      "loss": 0.9784,
       "step": 9430
     },
     {
-      "epoch": 0.25901513278954164,
+      "epoch": 0.26762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.7373357595829475e-05,
-      "loss": 0.936,
+      "learning_rate": 1.7182279862070786e-05,
+      "loss": 1.111,
       "step": 9431
     },
     {
-      "epoch": 0.25904259701738486,
+      "epoch": 0.2676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.737275667507019e-05,
-      "loss": 0.957,
+      "learning_rate": 1.718164033000521e-05,
+      "loss": 1.0119,
       "step": 9432
     },
     {
-      "epoch": 0.2590700612452281,
+      "epoch": 0.2676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.7372155695974795e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.718100073727599e-05,
+      "loss": 1.0025,
       "step": 9433
     },
     {
-      "epoch": 0.2590975254730713,
+      "epoch": 0.26770715096481273,
       "grad_norm": 0.0,
-      "learning_rate": 1.7371554658548048e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.7180361083888536e-05,
+      "loss": 1.0734,
       "step": 9434
     },
     {
-      "epoch": 0.2591249897009146,
+      "epoch": 0.2677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7370953562794703e-05,
-      "loss": 0.955,
+      "learning_rate": 1.717972136984825e-05,
+      "loss": 0.9869,
       "step": 9435
     },
     {
-      "epoch": 0.2591524539287578,
+      "epoch": 0.2677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.737035240871952e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.7179081595160535e-05,
+      "loss": 1.0248,
       "step": 9436
     },
     {
-      "epoch": 0.259179918156601,
+      "epoch": 0.26779228149829737,
       "grad_norm": 0.0,
-      "learning_rate": 1.7369751196327255e-05,
-      "loss": 0.9285,
+      "learning_rate": 1.7178441759830797e-05,
+      "loss": 0.9088,
       "step": 9437
     },
     {
-      "epoch": 0.25920738238444424,
+      "epoch": 0.2678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.7369149925622663e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.7177801863864437e-05,
+      "loss": 0.999,
       "step": 9438
     },
     {
-      "epoch": 0.2592348466122875,
+      "epoch": 0.2678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.7368548596610502e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.7177161907266858e-05,
+      "loss": 0.9736,
       "step": 9439
     },
     {
-      "epoch": 0.25926231084013074,
+      "epoch": 0.26787741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.736794720929553e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.7176521890043476e-05,
+      "loss": 0.8928,
       "step": 9440
     },
     {
-      "epoch": 0.25928977506797396,
+      "epoch": 0.267905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7367345763682505e-05,
-      "loss": 1.11,
+      "learning_rate": 1.7175881812199687e-05,
+      "loss": 0.9422,
       "step": 9441
     },
     {
-      "epoch": 0.2593172392958172,
+      "epoch": 0.26793416572077183,
       "grad_norm": 0.0,
-      "learning_rate": 1.736674425977619e-05,
-      "loss": 0.93,
+      "learning_rate": 1.7175241673740906e-05,
+      "loss": 0.8455,
       "step": 9442
     },
     {
-      "epoch": 0.25934470352366046,
+      "epoch": 0.26796254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.7366142697581337e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.7174601474672533e-05,
+      "loss": 0.8621,
       "step": 9443
     },
     {
-      "epoch": 0.2593721677515037,
+      "epoch": 0.2679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.736554107710271e-05,
-      "loss": 0.8776,
+      "learning_rate": 1.717396121499998e-05,
+      "loss": 0.998,
       "step": 9444
     },
     {
-      "epoch": 0.2593996319793469,
+      "epoch": 0.2680192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.736493939834507e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.7173320894728653e-05,
+      "loss": 1.0327,
       "step": 9445
     },
     {
-      "epoch": 0.2594270962071901,
+      "epoch": 0.26804767309875144,
       "grad_norm": 0.0,
-      "learning_rate": 1.7364337661313178e-05,
-      "loss": 0.9291,
+      "learning_rate": 1.717268051386396e-05,
+      "loss": 1.0715,
       "step": 9446
     },
     {
-      "epoch": 0.25945456043503334,
+      "epoch": 0.2680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.736373586601179e-05,
-      "loss": 0.8714,
+      "learning_rate": 1.7172040072411316e-05,
+      "loss": 1.0359,
       "step": 9447
     },
     {
-      "epoch": 0.2594820246628766,
+      "epoch": 0.2681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.7363134012445675e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.7171399570376124e-05,
+      "loss": 0.9969,
       "step": 9448
     },
     {
-      "epoch": 0.25950948889071984,
+      "epoch": 0.26813280363223607,
       "grad_norm": 0.0,
-      "learning_rate": 1.7362532100619592e-05,
-      "loss": 0.894,
+      "learning_rate": 1.71707590077638e-05,
+      "loss": 0.9968,
       "step": 9449
     },
     {
-      "epoch": 0.25953695311856306,
+      "epoch": 0.268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.7361930130538297e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.717011838457975e-05,
+      "loss": 1.0083,
       "step": 9450
     },
     {
-      "epoch": 0.2595644173464063,
+      "epoch": 0.2681895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.7361328102206566e-05,
-      "loss": 1.0447,
+      "learning_rate": 1.716947770082939e-05,
+      "loss": 1.1098,
       "step": 9451
     },
     {
-      "epoch": 0.25959188157424956,
+      "epoch": 0.26821793416572076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7360726015629153e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.7168836956518128e-05,
+      "loss": 1.0183,
       "step": 9452
     },
     {
-      "epoch": 0.2596193458020928,
+      "epoch": 0.2682463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.7360123870810822e-05,
-      "loss": 1.1187,
+      "learning_rate": 1.7168196151651378e-05,
+      "loss": 0.9739,
       "step": 9453
     },
     {
-      "epoch": 0.259646810029936,
+      "epoch": 0.26827468785471054,
       "grad_norm": 0.0,
-      "learning_rate": 1.735952166775634e-05,
-      "loss": 1.0359,
+      "learning_rate": 1.7167555286234556e-05,
+      "loss": 1.0311,
       "step": 9454
     },
     {
-      "epoch": 0.2596742742577792,
+      "epoch": 0.26830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.7358919406470477e-05,
-      "loss": 0.959,
+      "learning_rate": 1.716691436027307e-05,
+      "loss": 0.9085,
       "step": 9455
     },
     {
-      "epoch": 0.2597017384856225,
+      "epoch": 0.26833144154370037,
       "grad_norm": 0.0,
-      "learning_rate": 1.7358317086957987e-05,
-      "loss": 0.9326,
+      "learning_rate": 1.7166273373772334e-05,
+      "loss": 0.9468,
       "step": 9456
     },
     {
-      "epoch": 0.2597292027134657,
+      "epoch": 0.2683598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.7357714709223644e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.7165632326737767e-05,
+      "loss": 1.0157,
       "step": 9457
     },
     {
-      "epoch": 0.25975666694130894,
+      "epoch": 0.26838819523269014,
       "grad_norm": 0.0,
-      "learning_rate": 1.7357112273272215e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.7164991219174784e-05,
+      "loss": 0.9386,
       "step": 9458
     },
     {
-      "epoch": 0.25978413116915217,
+      "epoch": 0.268416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.735650977910846e-05,
-      "loss": 0.8408,
+      "learning_rate": 1.7164350051088793e-05,
+      "loss": 1.0291,
       "step": 9459
     },
     {
-      "epoch": 0.2598115953969954,
+      "epoch": 0.2684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.7355907226737154e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.7163708822485216e-05,
+      "loss": 1.0673,
       "step": 9460
     },
     {
-      "epoch": 0.25983905962483866,
+      "epoch": 0.2684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.735530461616306e-05,
-      "loss": 0.85,
+      "learning_rate": 1.7163067533369472e-05,
+      "loss": 1.1302,
       "step": 9461
     },
     {
-      "epoch": 0.2598665238526819,
+      "epoch": 0.2685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.735470194739094e-05,
-      "loss": 1.0502,
+      "learning_rate": 1.716242618374697e-05,
+      "loss": 1.0967,
       "step": 9462
     },
     {
-      "epoch": 0.2598939880805251,
+      "epoch": 0.2685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.7354099220425577e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.7161784773623134e-05,
+      "loss": 0.9901,
       "step": 9463
     },
     {
-      "epoch": 0.25992145230836833,
+      "epoch": 0.26855845629965946,
       "grad_norm": 0.0,
-      "learning_rate": 1.7353496435271727e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.7161143303003382e-05,
+      "loss": 0.8763,
       "step": 9464
     },
     {
-      "epoch": 0.2599489165362116,
+      "epoch": 0.2685868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.7352893591934168e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.716050177189313e-05,
+      "loss": 1.0714,
       "step": 9465
     },
     {
-      "epoch": 0.2599763807640548,
+      "epoch": 0.26861520998864924,
       "grad_norm": 0.0,
-      "learning_rate": 1.7352290690417666e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.7159860180297795e-05,
+      "loss": 1.0128,
       "step": 9466
     },
     {
-      "epoch": 0.26000384499189805,
+      "epoch": 0.26864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.7351687730726992e-05,
-      "loss": 1.0077,
+      "learning_rate": 1.7159218528222803e-05,
+      "loss": 0.9579,
       "step": 9467
     },
     {
-      "epoch": 0.26003130921974127,
+      "epoch": 0.26867196367763907,
       "grad_norm": 0.0,
-      "learning_rate": 1.735108471286692e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.7158576815673568e-05,
+      "loss": 0.9193,
       "step": 9468
     },
     {
-      "epoch": 0.26005877344758455,
+      "epoch": 0.26870034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.7350481636842213e-05,
-      "loss": 0.7898,
+      "learning_rate": 1.7157935042655515e-05,
+      "loss": 0.9946,
       "step": 9469
     },
     {
-      "epoch": 0.26008623767542777,
+      "epoch": 0.26872871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.734987850265765e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.7157293209174058e-05,
+      "loss": 1.0389,
       "step": 9470
     },
     {
-      "epoch": 0.260113701903271,
+      "epoch": 0.2687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.7349275310318e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.7156651315234628e-05,
+      "loss": 1.0089,
       "step": 9471
     },
     {
-      "epoch": 0.2601411661311142,
+      "epoch": 0.2687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.7348672059828037e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.715600936084264e-05,
+      "loss": 1.0941,
       "step": 9472
     },
     {
-      "epoch": 0.26016863035895743,
+      "epoch": 0.26881384790011353,
       "grad_norm": 0.0,
-      "learning_rate": 1.7348068751192538e-05,
-      "loss": 1.0643,
+      "learning_rate": 1.715536734600352e-05,
+      "loss": 1.0469,
       "step": 9473
     },
     {
-      "epoch": 0.2601960945868007,
+      "epoch": 0.2688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7347465384416268e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.7154725270722693e-05,
+      "loss": 0.8741,
       "step": 9474
     },
     {
-      "epoch": 0.26022355881464393,
+      "epoch": 0.2688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.734686195950401e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.7154083135005575e-05,
+      "loss": 1.0177,
       "step": 9475
     },
     {
-      "epoch": 0.26025102304248715,
+      "epoch": 0.26889897843359817,
       "grad_norm": 0.0,
-      "learning_rate": 1.7346258476460528e-05,
-      "loss": 0.9299,
+      "learning_rate": 1.71534409388576e-05,
+      "loss": 0.9804,
       "step": 9476
     },
     {
-      "epoch": 0.26027848727033037,
+      "epoch": 0.2689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.734565493529061e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.7152798682284183e-05,
+      "loss": 0.8639,
       "step": 9477
     },
     {
-      "epoch": 0.26030595149817365,
+      "epoch": 0.26895573212258794,
       "grad_norm": 0.0,
-      "learning_rate": 1.734505133599902e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.715215636529076e-05,
+      "loss": 1.0414,
       "step": 9478
     },
     {
-      "epoch": 0.26033341572601687,
+      "epoch": 0.26898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.734444767859054e-05,
-      "loss": 1.019,
+      "learning_rate": 1.7151513987882745e-05,
+      "loss": 0.8134,
       "step": 9479
     },
     {
-      "epoch": 0.2603608799538601,
+      "epoch": 0.2690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.7343843963069948e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.7150871550065574e-05,
+      "loss": 1.0611,
       "step": 9480
     },
     {
-      "epoch": 0.2603883441817033,
+      "epoch": 0.26904086265607263,
       "grad_norm": 0.0,
-      "learning_rate": 1.7343240189442015e-05,
-      "loss": 0.987,
+      "learning_rate": 1.7150229051844667e-05,
+      "loss": 1.0837,
       "step": 9481
     },
     {
-      "epoch": 0.2604158084095466,
+      "epoch": 0.26906923950056755,
       "grad_norm": 0.0,
-      "learning_rate": 1.7342636357711518e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.7149586493225453e-05,
+      "loss": 1.0239,
       "step": 9482
     },
     {
-      "epoch": 0.2604432726373898,
+      "epoch": 0.2690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.7342032467883245e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.7148943874213362e-05,
+      "loss": 1.0003,
       "step": 9483
     },
     {
-      "epoch": 0.26047073686523303,
+      "epoch": 0.2691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.7341428519961962e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.7148301194813823e-05,
+      "loss": 0.9084,
       "step": 9484
     },
     {
-      "epoch": 0.26049820109307625,
+      "epoch": 0.26915437003405224,
       "grad_norm": 0.0,
-      "learning_rate": 1.7340824513952454e-05,
-      "loss": 1.0754,
+      "learning_rate": 1.7147658455032263e-05,
+      "loss": 0.9983,
       "step": 9485
     },
     {
-      "epoch": 0.2605256653209195,
+      "epoch": 0.2691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.7340220449859498e-05,
-      "loss": 0.9701,
+      "learning_rate": 1.7147015654874106e-05,
+      "loss": 1.0125,
       "step": 9486
     },
     {
-      "epoch": 0.26055312954876275,
+      "epoch": 0.269211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7339616327687875e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.7146372794344787e-05,
+      "loss": 1.0356,
       "step": 9487
     },
     {
-      "epoch": 0.260580593776606,
+      "epoch": 0.26923950056753687,
       "grad_norm": 0.0,
-      "learning_rate": 1.7339012147442365e-05,
-      "loss": 1.051,
+      "learning_rate": 1.714572987344974e-05,
+      "loss": 1.071,
       "step": 9488
     },
     {
-      "epoch": 0.2606080580044492,
+      "epoch": 0.2692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.733840790912775e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.714508689219439e-05,
+      "loss": 0.9365,
       "step": 9489
     },
     {
-      "epoch": 0.2606355222322924,
+      "epoch": 0.2692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.7337803612748808e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.7144443850584166e-05,
+      "loss": 1.0554,
       "step": 9490
     },
     {
-      "epoch": 0.2606629864601357,
+      "epoch": 0.26932463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.733719925831032e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.7143800748624507e-05,
+      "loss": 0.8802,
       "step": 9491
     },
     {
-      "epoch": 0.2606904506879789,
+      "epoch": 0.2693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.7336594845817068e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.714315758632084e-05,
+      "loss": 0.872,
       "step": 9492
     },
     {
-      "epoch": 0.26071791491582214,
+      "epoch": 0.26938138479001134,
       "grad_norm": 0.0,
-      "learning_rate": 1.7335990375273838e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.7142514363678602e-05,
+      "loss": 0.9712,
       "step": 9493
     },
     {
-      "epoch": 0.26074537914366536,
+      "epoch": 0.26940976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.7335385846685412e-05,
-      "loss": 1.0394,
+      "learning_rate": 1.7141871080703222e-05,
+      "loss": 0.8458,
       "step": 9494
     },
     {
-      "epoch": 0.26077284337150863,
+      "epoch": 0.2694381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.7334781260056567e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.7141227737400136e-05,
+      "loss": 0.9512,
       "step": 9495
     },
     {
-      "epoch": 0.26080030759935185,
+      "epoch": 0.269466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.7334176615392094e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.7140584333774782e-05,
+      "loss": 0.9561,
       "step": 9496
     },
     {
-      "epoch": 0.2608277718271951,
+      "epoch": 0.26949489216799094,
       "grad_norm": 0.0,
-      "learning_rate": 1.7333571912696775e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.713994086983259e-05,
+      "loss": 0.9,
       "step": 9497
     },
     {
-      "epoch": 0.2608552360550383,
+      "epoch": 0.2695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.733296715197539e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.7139297345578992e-05,
+      "loss": 0.9124,
       "step": 9498
     },
     {
-      "epoch": 0.2608827002828816,
+      "epoch": 0.2695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.7332362333232736e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.7138653761019432e-05,
+      "loss": 0.9386,
       "step": 9499
     },
     {
-      "epoch": 0.2609101645107248,
+      "epoch": 0.2695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.733175745647359e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.7138010116159342e-05,
+      "loss": 1.0498,
       "step": 9500
     },
     {
-      "epoch": 0.260937628738568,
+      "epoch": 0.2696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.7331152521702733e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.713736641100416e-05,
+      "loss": 0.9769,
       "step": 9501
     },
     {
-      "epoch": 0.26096509296641124,
+      "epoch": 0.2696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.733054752892496e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.7136722645559322e-05,
+      "loss": 0.9625,
       "step": 9502
     },
     {
-      "epoch": 0.26099255719425446,
+      "epoch": 0.26966515323496026,
       "grad_norm": 0.0,
-      "learning_rate": 1.7329942478145057e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.7136078819830267e-05,
+      "loss": 1.0466,
       "step": 9503
     },
     {
-      "epoch": 0.26102002142209774,
+      "epoch": 0.2696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.732933736936781e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.713543493382244e-05,
+      "loss": 0.9867,
       "step": 9504
     },
     {
-      "epoch": 0.26104748564994096,
+      "epoch": 0.26972190692395004,
       "grad_norm": 0.0,
-      "learning_rate": 1.7328732202598006e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.7134790987541263e-05,
+      "loss": 0.9984,
       "step": 9505
     },
     {
-      "epoch": 0.2610749498777842,
+      "epoch": 0.26975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.732812697784043e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.713414698099219e-05,
+      "loss": 0.9325,
       "step": 9506
     },
     {
-      "epoch": 0.2611024141056274,
+      "epoch": 0.2697786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.732752169509988e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.7133502914180657e-05,
+      "loss": 1.0257,
       "step": 9507
     },
     {
-      "epoch": 0.2611298783334707,
+      "epoch": 0.26980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.732691635438114e-05,
-      "loss": 1.1237,
+      "learning_rate": 1.7132858787112107e-05,
+      "loss": 0.9166,
       "step": 9508
     },
     {
-      "epoch": 0.2611573425613139,
+      "epoch": 0.26983541430192964,
       "grad_norm": 0.0,
-      "learning_rate": 1.7326310955688997e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.7132214599791975e-05,
+      "loss": 1.0424,
       "step": 9509
     },
     {
-      "epoch": 0.2611848067891571,
+      "epoch": 0.2698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.7325705499028244e-05,
-      "loss": 1.0252,
+      "learning_rate": 1.7131570352225703e-05,
+      "loss": 1.0311,
       "step": 9510
     },
     {
-      "epoch": 0.26121227101700034,
+      "epoch": 0.2698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.732509998440367e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.713092604441874e-05,
+      "loss": 0.9852,
       "step": 9511
     },
     {
-      "epoch": 0.2612397352448436,
+      "epoch": 0.2699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.732449441182007e-05,
-      "loss": 1.0638,
+      "learning_rate": 1.7130281676376522e-05,
+      "loss": 0.9557,
       "step": 9512
     },
     {
-      "epoch": 0.26126719947268684,
+      "epoch": 0.2699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.732388878128223e-05,
-      "loss": 0.8746,
+      "learning_rate": 1.71296372481045e-05,
+      "loss": 1.0364,
       "step": 9513
     },
     {
-      "epoch": 0.26129466370053006,
+      "epoch": 0.2699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.732328309279495e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.71289927596081e-05,
+      "loss": 1.0027,
       "step": 9514
     },
     {
-      "epoch": 0.2613221279283733,
+      "epoch": 0.27000567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.7322677346363014e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.712834821089279e-05,
+      "loss": 0.8164,
       "step": 9515
     },
     {
-      "epoch": 0.2613495921562165,
+      "epoch": 0.2700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.7322071541991218e-05,
-      "loss": 0.9406,
+      "learning_rate": 1.7127703601963995e-05,
+      "loss": 0.9355,
       "step": 9516
     },
     {
-      "epoch": 0.2613770563840598,
+      "epoch": 0.27006242905788874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7321465679684358e-05,
-      "loss": 0.892,
+      "learning_rate": 1.712705893282717e-05,
+      "loss": 0.9637,
       "step": 9517
     },
     {
-      "epoch": 0.261404520611903,
+      "epoch": 0.27009080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.7320859759447223e-05,
-      "loss": 1.0351,
+      "learning_rate": 1.7126414203487757e-05,
+      "loss": 0.9038,
       "step": 9518
     },
     {
-      "epoch": 0.2614319848397462,
+      "epoch": 0.27011918274687857,
       "grad_norm": 0.0,
-      "learning_rate": 1.7320253781284612e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.71257694139512e-05,
+      "loss": 1.003,
       "step": 9519
     },
     {
-      "epoch": 0.26145944906758944,
+      "epoch": 0.27014755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.7319647745201316e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.712512456422295e-05,
+      "loss": 0.9545,
       "step": 9520
     },
     {
-      "epoch": 0.2614869132954327,
+      "epoch": 0.27017593643586835,
       "grad_norm": 0.0,
-      "learning_rate": 1.7319041651202132e-05,
-      "loss": 1.0394,
+      "learning_rate": 1.7124479654308457e-05,
+      "loss": 1.0233,
       "step": 9521
     },
     {
-      "epoch": 0.26151437752327594,
+      "epoch": 0.2702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.7318435499291854e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.7123834684213157e-05,
+      "loss": 0.9813,
       "step": 9522
     },
     {
-      "epoch": 0.26154184175111916,
+      "epoch": 0.2702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.731782928947528e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.7123189653942512e-05,
+      "loss": 1.0329,
       "step": 9523
     },
     {
-      "epoch": 0.2615693059789624,
+      "epoch": 0.270261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.7317223021757207e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.712254456350196e-05,
+      "loss": 0.953,
       "step": 9524
     },
     {
-      "epoch": 0.26159677020680566,
+      "epoch": 0.2702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7316616696142434e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.7121899412896954e-05,
+      "loss": 0.8936,
       "step": 9525
     },
     {
-      "epoch": 0.2616242344346489,
+      "epoch": 0.2703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.731601031263575e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.7121254202132943e-05,
+      "loss": 0.9886,
       "step": 9526
     },
     {
-      "epoch": 0.2616516986624921,
+      "epoch": 0.27034619750283767,
       "grad_norm": 0.0,
-      "learning_rate": 1.731540387124196e-05,
-      "loss": 0.9299,
+      "learning_rate": 1.712060893121538e-05,
+      "loss": 1.0007,
       "step": 9527
     },
     {
-      "epoch": 0.2616791628903353,
+      "epoch": 0.2703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.7314797371965863e-05,
-      "loss": 1.0851,
+      "learning_rate": 1.7119963600149715e-05,
+      "loss": 0.9645,
       "step": 9528
     },
     {
-      "epoch": 0.26170662711817855,
+      "epoch": 0.27040295119182745,
       "grad_norm": 0.0,
-      "learning_rate": 1.731419081481225e-05,
-      "loss": 1.0072,
+      "learning_rate": 1.711931820894139e-05,
+      "loss": 1.0198,
       "step": 9529
     },
     {
-      "epoch": 0.2617340913460218,
+      "epoch": 0.27043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.7313584199785934e-05,
-      "loss": 1.093,
+      "learning_rate": 1.7118672757595873e-05,
+      "loss": 1.0485,
       "step": 9530
     },
     {
-      "epoch": 0.26176155557386505,
+      "epoch": 0.2704597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.7312977526891704e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.71180272461186e-05,
+      "loss": 0.8716,
       "step": 9531
     },
     {
-      "epoch": 0.26178901980170827,
+      "epoch": 0.27048808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.731237079613436e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.7117381674515034e-05,
+      "loss": 1.0685,
       "step": 9532
     },
     {
-      "epoch": 0.2618164840295515,
+      "epoch": 0.27051645856980705,
       "grad_norm": 0.0,
-      "learning_rate": 1.731176400751871e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.7116736042790624e-05,
+      "loss": 0.8965,
       "step": 9533
     },
     {
-      "epoch": 0.26184394825739477,
+      "epoch": 0.2705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.731115716104955e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.7116090350950825e-05,
+      "loss": 0.9326,
       "step": 9534
     },
     {
-      "epoch": 0.261871412485238,
+      "epoch": 0.2705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.7310550256731678e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.7115444599001093e-05,
+      "loss": 1.0013,
       "step": 9535
     },
     {
-      "epoch": 0.2618988767130812,
+      "epoch": 0.27060158910329174,
       "grad_norm": 0.0,
-      "learning_rate": 1.7309943294569903e-05,
-      "loss": 0.8925,
+      "learning_rate": 1.7114798786946875e-05,
+      "loss": 0.8425,
       "step": 9536
     },
     {
-      "epoch": 0.26192634094092443,
+      "epoch": 0.2706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.7309336274569025e-05,
-      "loss": 0.996,
+      "learning_rate": 1.7114152914793636e-05,
+      "loss": 0.9581,
       "step": 9537
     },
     {
-      "epoch": 0.2619538051687677,
+      "epoch": 0.2706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.730872919673385e-05,
-      "loss": 0.9591,
+      "learning_rate": 1.711350698254683e-05,
+      "loss": 0.8829,
       "step": 9538
     },
     {
-      "epoch": 0.2619812693966109,
+      "epoch": 0.2706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.7308122061069176e-05,
-      "loss": 1.1246,
+      "learning_rate": 1.7112860990211908e-05,
+      "loss": 1.0294,
       "step": 9539
     },
     {
-      "epoch": 0.26200873362445415,
+      "epoch": 0.2707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.730751486757981e-05,
-      "loss": 1.025,
+      "learning_rate": 1.7112214937794327e-05,
+      "loss": 0.9381,
       "step": 9540
     },
     {
-      "epoch": 0.26203619785229737,
+      "epoch": 0.27074347332576615,
       "grad_norm": 0.0,
-      "learning_rate": 1.7306907616270555e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.711156882529955e-05,
+      "loss": 1.0034,
       "step": 9541
     },
     {
-      "epoch": 0.2620636620801406,
+      "epoch": 0.27077185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.730630030714622e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.7110922652733028e-05,
+      "loss": 0.9896,
       "step": 9542
     },
     {
-      "epoch": 0.26209112630798387,
+      "epoch": 0.270800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.73056929402116e-05,
-      "loss": 1.0437,
+      "learning_rate": 1.7110276420100226e-05,
+      "loss": 0.9833,
       "step": 9543
     },
     {
-      "epoch": 0.2621185905358271,
+      "epoch": 0.27082860385925084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7305085515471513e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.71096301274066e-05,
+      "loss": 0.8908,
       "step": 9544
     },
     {
-      "epoch": 0.2621460547636703,
+      "epoch": 0.27085698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.7304478032930757e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.710898377465761e-05,
+      "loss": 0.9146,
       "step": 9545
     },
     {
-      "epoch": 0.26217351899151353,
+      "epoch": 0.2708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.730387049259414e-05,
-      "loss": 0.8869,
+      "learning_rate": 1.710833736185871e-05,
+      "loss": 0.9607,
       "step": 9546
     },
     {
-      "epoch": 0.2622009832193568,
+      "epoch": 0.27091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.7303262894466474e-05,
-      "loss": 1.0448,
+      "learning_rate": 1.7107690889015364e-05,
+      "loss": 1.0295,
       "step": 9547
     },
     {
-      "epoch": 0.26222844744720003,
+      "epoch": 0.27094211123723044,
       "grad_norm": 0.0,
-      "learning_rate": 1.730265523855256e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.710704435613304e-05,
+      "loss": 1.1541,
       "step": 9548
     },
     {
-      "epoch": 0.26225591167504325,
+      "epoch": 0.2709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.730204752485721e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.7106397763217185e-05,
+      "loss": 0.8927,
       "step": 9549
     },
     {
-      "epoch": 0.2622833759028865,
+      "epoch": 0.2709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.7301439753385235e-05,
-      "loss": 1.0286,
+      "learning_rate": 1.7105751110273275e-05,
+      "loss": 0.9812,
       "step": 9550
     },
     {
-      "epoch": 0.26231084013072975,
+      "epoch": 0.2710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.7300831924141433e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.710510439730676e-05,
+      "loss": 1.0438,
       "step": 9551
     },
     {
-      "epoch": 0.26233830435857297,
+      "epoch": 0.27105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.7300224037130628e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.7104457624323118e-05,
+      "loss": 0.9377,
       "step": 9552
     },
     {
-      "epoch": 0.2623657685864162,
+      "epoch": 0.2710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.729961609235762e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.7103810791327796e-05,
+      "loss": 1.0259,
       "step": 9553
     },
     {
-      "epoch": 0.2623932328142594,
+      "epoch": 0.27111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.729900808982722e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.7103163898326265e-05,
+      "loss": 0.9118,
       "step": 9554
     },
     {
-      "epoch": 0.26242069704210264,
+      "epoch": 0.2711407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.7298400029544242e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.7102516945323994e-05,
+      "loss": 0.9479,
       "step": 9555
     },
     {
-      "epoch": 0.2624481612699459,
+      "epoch": 0.27116912599318954,
       "grad_norm": 0.0,
-      "learning_rate": 1.7297791911513497e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.7101869932326436e-05,
+      "loss": 1.0839,
       "step": 9556
     },
     {
-      "epoch": 0.26247562549778913,
+      "epoch": 0.27119750283768446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7297183735739795e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.7101222859339067e-05,
+      "loss": 0.9315,
       "step": 9557
     },
     {
-      "epoch": 0.26250308972563235,
+      "epoch": 0.2712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.7296575502227947e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.710057572636735e-05,
+      "loss": 0.9084,
       "step": 9558
     },
     {
-      "epoch": 0.2625305539534756,
+      "epoch": 0.27125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.7295967210982766e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.709992853341675e-05,
+      "loss": 0.9327,
       "step": 9559
     },
     {
-      "epoch": 0.26255801818131885,
+      "epoch": 0.27128263337116915,
       "grad_norm": 0.0,
-      "learning_rate": 1.729535886200907e-05,
-      "loss": 1.0648,
+      "learning_rate": 1.7099281280492733e-05,
+      "loss": 0.9783,
       "step": 9560
     },
     {
-      "epoch": 0.2625854824091621,
+      "epoch": 0.271311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.7294750455311667e-05,
-      "loss": 0.962,
+      "learning_rate": 1.7098633967600772e-05,
+      "loss": 0.991,
       "step": 9561
     },
     {
-      "epoch": 0.2626129466370053,
+      "epoch": 0.2713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.729414199089537e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.7097986594746328e-05,
+      "loss": 1.1127,
       "step": 9562
     },
     {
-      "epoch": 0.2626404108648485,
+      "epoch": 0.2713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.7293533468765004e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.709733916193487e-05,
+      "loss": 1.0654,
       "step": 9563
     },
     {
-      "epoch": 0.2626678750926918,
+      "epoch": 0.2713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.7292924888925368e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.7096691669171872e-05,
+      "loss": 1.0663,
       "step": 9564
     },
     {
-      "epoch": 0.262695339320535,
+      "epoch": 0.2714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.729231625138129e-05,
-      "loss": 1.063,
+      "learning_rate": 1.70960441164628e-05,
+      "loss": 0.9347,
       "step": 9565
     },
     {
-      "epoch": 0.26272280354837824,
+      "epoch": 0.27145289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 1.729170755613758e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.7095396503813125e-05,
+      "loss": 1.0025,
       "step": 9566
     },
     {
-      "epoch": 0.26275026777622146,
+      "epoch": 0.2714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.7291098803199052e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.7094748831228318e-05,
+      "loss": 0.9908,
       "step": 9567
     },
     {
-      "epoch": 0.2627777320040647,
+      "epoch": 0.27150964812712824,
       "grad_norm": 0.0,
-      "learning_rate": 1.729048999257053e-05,
-      "loss": 0.983,
+      "learning_rate": 1.7094101098713845e-05,
+      "loss": 1.0004,
       "step": 9568
     },
     {
-      "epoch": 0.26280519623190796,
+      "epoch": 0.27153802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.7289881124256823e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.7093453306275184e-05,
+      "loss": 0.9997,
       "step": 9569
     },
     {
-      "epoch": 0.2628326604597512,
+      "epoch": 0.2715664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 1.7289272198262756e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.70928054539178e-05,
+      "loss": 1.0668,
       "step": 9570
     },
     {
-      "epoch": 0.2628601246875944,
+      "epoch": 0.27159477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 1.7288663214593146e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.7092157541647173e-05,
+      "loss": 0.9927,
       "step": 9571
     },
     {
-      "epoch": 0.2628875889154376,
+      "epoch": 0.27162315550510785,
       "grad_norm": 0.0,
-      "learning_rate": 1.7288054173252806e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.709150956946877e-05,
+      "loss": 0.8965,
       "step": 9572
     },
     {
-      "epoch": 0.2629150531432809,
+      "epoch": 0.2716515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.7287445074246557e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.709086153738807e-05,
+      "loss": 0.9601,
       "step": 9573
     },
     {
-      "epoch": 0.2629425173711241,
+      "epoch": 0.2716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.7286835917579222e-05,
-      "loss": 1.0113,
+      "learning_rate": 1.7090213445410544e-05,
+      "loss": 0.9347,
       "step": 9574
     },
     {
-      "epoch": 0.26296998159896734,
+      "epoch": 0.2717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.7286226703255618e-05,
-      "loss": 1.0835,
+      "learning_rate": 1.7089565293541664e-05,
+      "loss": 0.994,
       "step": 9575
     },
     {
-      "epoch": 0.26299744582681056,
+      "epoch": 0.2717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.7285617431280566e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.7088917081786908e-05,
+      "loss": 0.9246,
       "step": 9576
     },
     {
-      "epoch": 0.26302491005465384,
+      "epoch": 0.2717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.7285008101658885e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.708826881015175e-05,
+      "loss": 0.966,
       "step": 9577
     },
     {
-      "epoch": 0.26305237428249706,
+      "epoch": 0.2717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.7284398714395402e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.708762047864167e-05,
+      "loss": 0.9585,
       "step": 9578
     },
     {
-      "epoch": 0.2630798385103403,
+      "epoch": 0.2718217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.7283789269494932e-05,
-      "loss": 1.0833,
+      "learning_rate": 1.7086972087262138e-05,
+      "loss": 0.9635,
       "step": 9579
     },
     {
-      "epoch": 0.2631073027381835,
+      "epoch": 0.27185017026106695,
       "grad_norm": 0.0,
-      "learning_rate": 1.72831797669623e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.7086323636018635e-05,
+      "loss": 0.8598,
       "step": 9580
     },
     {
-      "epoch": 0.2631347669660267,
+      "epoch": 0.27187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.7282570206802326e-05,
-      "loss": 0.9317,
+      "learning_rate": 1.7085675124916638e-05,
+      "loss": 1.0648,
       "step": 9581
     },
     {
-      "epoch": 0.26316223119387,
+      "epoch": 0.2719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.728196058901984e-05,
-      "loss": 1.0903,
+      "learning_rate": 1.7085026553961622e-05,
+      "loss": 1.067,
       "step": 9582
     },
     {
-      "epoch": 0.2631896954217132,
+      "epoch": 0.27193530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.7281350913619658e-05,
-      "loss": 0.9266,
+      "learning_rate": 1.7084377923159074e-05,
+      "loss": 0.8547,
       "step": 9583
     },
     {
-      "epoch": 0.26321715964955644,
+      "epoch": 0.27196367763904655,
       "grad_norm": 0.0,
-      "learning_rate": 1.728074118060661e-05,
-      "loss": 1.0601,
+      "learning_rate": 1.708372923251446e-05,
+      "loss": 0.873,
       "step": 9584
     },
     {
-      "epoch": 0.26324462387739966,
+      "epoch": 0.2719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.728013138998551e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.708308048203327e-05,
+      "loss": 0.9227,
       "step": 9585
     },
     {
-      "epoch": 0.26327208810524294,
+      "epoch": 0.2720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.7279521541761203e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.7082431671720982e-05,
+      "loss": 0.9596,
       "step": 9586
     },
     {
-      "epoch": 0.26329955233308616,
+      "epoch": 0.2720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.727891163593849e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.7081782801583074e-05,
+      "loss": 0.9713,
       "step": 9587
     },
     {
-      "epoch": 0.2633270165609294,
+      "epoch": 0.2720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.7278301672522215e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.708113387162503e-05,
+      "loss": 1.0332,
       "step": 9588
     },
     {
-      "epoch": 0.2633544807887726,
+      "epoch": 0.272105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.7277691651517195e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.708048488185233e-05,
+      "loss": 0.8727,
       "step": 9589
     },
     {
-      "epoch": 0.2633819450166159,
+      "epoch": 0.2721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.7277081572928263e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.7079835832270455e-05,
+      "loss": 0.9529,
       "step": 9590
     },
     {
-      "epoch": 0.2634094092444591,
+      "epoch": 0.2721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.7276471436760237e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.7079186722884882e-05,
+      "loss": 0.9716,
       "step": 9591
     },
     {
-      "epoch": 0.2634368734723023,
+      "epoch": 0.27219069239500565,
       "grad_norm": 0.0,
-      "learning_rate": 1.7275861243017956e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.707853755370111e-05,
+      "loss": 1.0261,
       "step": 9592
     },
     {
-      "epoch": 0.26346433770014555,
+      "epoch": 0.27221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.727525099170624e-05,
-      "loss": 1.0019,
+      "learning_rate": 1.707788832472461e-05,
+      "loss": 1.0426,
       "step": 9593
     },
     {
-      "epoch": 0.26349180192798877,
+      "epoch": 0.2722474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.727464068282992e-05,
-      "loss": 1.0026,
+      "learning_rate": 1.7077239035960868e-05,
+      "loss": 1.0351,
       "step": 9594
     },
     {
-      "epoch": 0.26351926615583204,
+      "epoch": 0.27227582292849034,
       "grad_norm": 0.0,
-      "learning_rate": 1.7274030316393826e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.7076589687415374e-05,
+      "loss": 0.9715,
       "step": 9595
     },
     {
-      "epoch": 0.26354673038367527,
+      "epoch": 0.27230419977298526,
       "grad_norm": 0.0,
-      "learning_rate": 1.7273419892402785e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.7075940279093602e-05,
+      "loss": 1.0213,
       "step": 9596
     },
     {
-      "epoch": 0.2635741946115185,
+      "epoch": 0.2723325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.7272809410861628e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.707529081100105e-05,
+      "loss": 0.9058,
       "step": 9597
     },
     {
-      "epoch": 0.2636016588393617,
+      "epoch": 0.27236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.7272198871775184e-05,
-      "loss": 0.9481,
+      "learning_rate": 1.7074641283143195e-05,
+      "loss": 0.9946,
       "step": 9598
     },
     {
-      "epoch": 0.263629123067205,
+      "epoch": 0.27238933030646995,
       "grad_norm": 0.0,
-      "learning_rate": 1.727158827514829e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.7073991695525528e-05,
+      "loss": 0.9721,
       "step": 9599
     },
     {
-      "epoch": 0.2636565872950482,
+      "epoch": 0.2724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.727097762098577e-05,
-      "loss": 0.9812,
+      "learning_rate": 1.7073342048153535e-05,
+      "loss": 0.9218,
       "step": 9600
     },
     {
-      "epoch": 0.2636840515228914,
+      "epoch": 0.2724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.727036690929246e-05,
-      "loss": 0.954,
+      "learning_rate": 1.7072692341032705e-05,
+      "loss": 1.0539,
       "step": 9601
     },
     {
-      "epoch": 0.26371151575073465,
+      "epoch": 0.2724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.7269756140073193e-05,
-      "loss": 1.0116,
+      "learning_rate": 1.7072042574168524e-05,
+      "loss": 0.971,
       "step": 9602
     },
     {
-      "epoch": 0.2637389799785779,
+      "epoch": 0.2725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.7269145313332795e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.707139274756648e-05,
+      "loss": 0.9074,
       "step": 9603
     },
     {
-      "epoch": 0.26376644420642115,
+      "epoch": 0.27253121452894435,
       "grad_norm": 0.0,
-      "learning_rate": 1.7268534429076105e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.7070742861232066e-05,
+      "loss": 1.0184,
       "step": 9604
     },
     {
-      "epoch": 0.26379390843426437,
+      "epoch": 0.27255959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.7267923487307955e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.7070092915170767e-05,
+      "loss": 1.0672,
       "step": 9605
     },
     {
-      "epoch": 0.2638213726621076,
+      "epoch": 0.2725879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.726731248803318e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.706944290938808e-05,
+      "loss": 0.9492,
       "step": 9606
     },
     {
-      "epoch": 0.26384883688995087,
+      "epoch": 0.27261634506242904,
       "grad_norm": 0.0,
-      "learning_rate": 1.726670143125661e-05,
-      "loss": 1.0427,
+      "learning_rate": 1.7068792843889486e-05,
+      "loss": 0.9314,
       "step": 9607
     },
     {
-      "epoch": 0.2638763011177941,
+      "epoch": 0.27264472190692396,
       "grad_norm": 0.0,
-      "learning_rate": 1.726609031698309e-05,
-      "loss": 0.8092,
+      "learning_rate": 1.706814271868048e-05,
+      "loss": 0.8251,
       "step": 9608
     },
     {
-      "epoch": 0.2639037653456373,
+      "epoch": 0.2726730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.7265479145217445e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.706749253376656e-05,
+      "loss": 0.9783,
       "step": 9609
     },
     {
-      "epoch": 0.26393122957348053,
+      "epoch": 0.27270147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.7264867915964513e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.7066842289153208e-05,
+      "loss": 0.9225,
       "step": 9610
     },
     {
-      "epoch": 0.26395869380132375,
+      "epoch": 0.27272985244040865,
       "grad_norm": 0.0,
-      "learning_rate": 1.7264256629229134e-05,
-      "loss": 1.0215,
+      "learning_rate": 1.7066191984845924e-05,
+      "loss": 1.0684,
       "step": 9611
     },
     {
-      "epoch": 0.26398615802916703,
+      "epoch": 0.2727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.7263645285016143e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.70655416208502e-05,
+      "loss": 0.995,
       "step": 9612
     },
     {
-      "epoch": 0.26401362225701025,
+      "epoch": 0.2727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.7263033883330377e-05,
-      "loss": 0.9689,
+      "learning_rate": 1.706489119717153e-05,
+      "loss": 1.0076,
       "step": 9613
     },
     {
-      "epoch": 0.26404108648485347,
+      "epoch": 0.2728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.726242242417667e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.70642407138154e-05,
+      "loss": 1.0728,
       "step": 9614
     },
     {
-      "epoch": 0.2640685507126967,
+      "epoch": 0.2728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.7261810907559867e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.7063590170787314e-05,
+      "loss": 1.0639,
       "step": 9615
     },
     {
-      "epoch": 0.26409601494053997,
+      "epoch": 0.2728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.72611993334848e-05,
-      "loss": 1.0754,
+      "learning_rate": 1.7062939568092765e-05,
+      "loss": 1.0017,
       "step": 9616
     },
     {
-      "epoch": 0.2641234791683832,
+      "epoch": 0.272900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.7260587701956315e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.7062288905737245e-05,
+      "loss": 1.0122,
       "step": 9617
     },
     {
-      "epoch": 0.2641509433962264,
+      "epoch": 0.2729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.7259976012979247e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.7061638183726254e-05,
+      "loss": 0.9161,
       "step": 9618
     },
     {
-      "epoch": 0.26417840762406963,
+      "epoch": 0.27295686719636775,
       "grad_norm": 0.0,
-      "learning_rate": 1.7259364266558437e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.7060987402065292e-05,
+      "loss": 1.0642,
       "step": 9619
     },
     {
-      "epoch": 0.2642058718519129,
+      "epoch": 0.27298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.725875246269872e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.7060336560759848e-05,
+      "loss": 1.0485,
       "step": 9620
     },
     {
-      "epoch": 0.26423333607975613,
+      "epoch": 0.2730136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.7258140601404945e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.7059685659815424e-05,
+      "loss": 0.8932,
       "step": 9621
     },
     {
-      "epoch": 0.26426080030759935,
+      "epoch": 0.27304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.725752868268195e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.7059034699237523e-05,
+      "loss": 1.0124,
       "step": 9622
     },
     {
-      "epoch": 0.2642882645354426,
+      "epoch": 0.27307037457434735,
       "grad_norm": 0.0,
-      "learning_rate": 1.725691670653458e-05,
-      "loss": 0.9652,
+      "learning_rate": 1.7058383679031632e-05,
+      "loss": 0.9811,
       "step": 9623
     },
     {
-      "epoch": 0.2643157287632858,
+      "epoch": 0.2730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.725630467296767e-05,
-      "loss": 0.9427,
+      "learning_rate": 1.7057732599203262e-05,
+      "loss": 0.9754,
       "step": 9624
     },
     {
-      "epoch": 0.2643431929911291,
+      "epoch": 0.2731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.7255692581986068e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.7057081459757904e-05,
+      "loss": 0.9168,
       "step": 9625
     },
     {
-      "epoch": 0.2643706572189723,
+      "epoch": 0.273155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.7255080433594616e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.705643026070106e-05,
+      "loss": 1.0309,
       "step": 9626
     },
     {
-      "epoch": 0.2643981214468155,
+      "epoch": 0.2731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.7254468227798152e-05,
-      "loss": 1.016,
+      "learning_rate": 1.7055779002038235e-05,
+      "loss": 0.939,
       "step": 9627
     },
     {
-      "epoch": 0.26442558567465874,
+      "epoch": 0.2732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.725385596460153e-05,
-      "loss": 1.0591,
+      "learning_rate": 1.705512768377493e-05,
+      "loss": 1.0179,
       "step": 9628
     },
     {
-      "epoch": 0.264453049902502,
+      "epoch": 0.2732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.7253243644009592e-05,
-      "loss": 1.0813,
+      "learning_rate": 1.7054476305916644e-05,
+      "loss": 0.9008,
       "step": 9629
     },
     {
-      "epoch": 0.26448051413034523,
+      "epoch": 0.2732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.7252631266027177e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.7053824868468876e-05,
+      "loss": 0.9011,
       "step": 9630
     },
     {
-      "epoch": 0.26450797835818846,
+      "epoch": 0.27329738933030645,
       "grad_norm": 0.0,
-      "learning_rate": 1.7252018830659136e-05,
-      "loss": 0.9404,
+      "learning_rate": 1.705317337143713e-05,
+      "loss": 0.9196,
       "step": 9631
     },
     {
-      "epoch": 0.2645354425860317,
+      "epoch": 0.27332576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.725140633791031e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.7052521814826915e-05,
+      "loss": 0.9648,
       "step": 9632
     },
     {
-      "epoch": 0.26456290681387495,
+      "epoch": 0.2733541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.7250793787785552e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.7051870198643734e-05,
+      "loss": 0.9893,
       "step": 9633
     },
     {
-      "epoch": 0.2645903710417182,
+      "epoch": 0.27338251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.7250181180289703e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.7051218522893086e-05,
+      "loss": 1.0893,
       "step": 9634
     },
     {
-      "epoch": 0.2646178352695614,
+      "epoch": 0.27341089670828606,
       "grad_norm": 0.0,
-      "learning_rate": 1.7249568515427613e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.7050566787580477e-05,
+      "loss": 1.0053,
       "step": 9635
     },
     {
-      "epoch": 0.2646452994974046,
+      "epoch": 0.2734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.7248955793204126e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.7049914992711412e-05,
+      "loss": 0.9132,
       "step": 9636
     },
     {
-      "epoch": 0.26467276372524784,
+      "epoch": 0.27346765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 1.7248343013624094e-05,
-      "loss": 1.056,
+      "learning_rate": 1.70492631382914e-05,
+      "loss": 1.1185,
       "step": 9637
     },
     {
-      "epoch": 0.2647002279530911,
+      "epoch": 0.2734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.7247730176692367e-05,
-      "loss": 0.9495,
+      "learning_rate": 1.7048611224325947e-05,
+      "loss": 0.9412,
       "step": 9638
     },
     {
-      "epoch": 0.26472769218093434,
+      "epoch": 0.2735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.724711728241379e-05,
-      "loss": 1.145,
+      "learning_rate": 1.704795925082056e-05,
+      "loss": 1.0093,
       "step": 9639
     },
     {
-      "epoch": 0.26475515640877756,
+      "epoch": 0.2735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.7246504330793213e-05,
-      "loss": 0.9491,
+      "learning_rate": 1.7047307217780737e-05,
+      "loss": 0.9076,
       "step": 9640
     },
     {
-      "epoch": 0.2647826206366208,
+      "epoch": 0.2735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.7245891321835487e-05,
-      "loss": 0.9017,
+      "learning_rate": 1.7046655125211996e-05,
+      "loss": 0.9507,
       "step": 9641
     },
     {
-      "epoch": 0.26481008486446406,
+      "epoch": 0.2736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.724527825554546e-05,
-      "loss": 1.0804,
+      "learning_rate": 1.7046002973119847e-05,
+      "loss": 0.919,
       "step": 9642
     },
     {
-      "epoch": 0.2648375490923073,
+      "epoch": 0.27363791146424515,
       "grad_norm": 0.0,
-      "learning_rate": 1.7244665131927986e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.704535076150979e-05,
+      "loss": 0.9823,
       "step": 9643
     },
     {
-      "epoch": 0.2648650133201505,
+      "epoch": 0.27366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.7244051950987918e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.704469849038734e-05,
+      "loss": 0.9163,
       "step": 9644
     },
     {
-      "epoch": 0.2648924775479937,
+      "epoch": 0.273694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.72434387127301e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.7044046159758004e-05,
+      "loss": 0.9311,
       "step": 9645
     },
     {
-      "epoch": 0.264919941775837,
+      "epoch": 0.27372304199772984,
       "grad_norm": 0.0,
-      "learning_rate": 1.7242825417159395e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.7043393769627293e-05,
+      "loss": 0.995,
       "step": 9646
     },
     {
-      "epoch": 0.2649474060036802,
+      "epoch": 0.27375141884222476,
       "grad_norm": 0.0,
-      "learning_rate": 1.7242212064280644e-05,
-      "loss": 0.9239,
+      "learning_rate": 1.704274132000072e-05,
+      "loss": 0.9429,
       "step": 9647
     },
     {
-      "epoch": 0.26497487023152344,
+      "epoch": 0.2737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.724159865409871e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.704208881088379e-05,
+      "loss": 1.0209,
       "step": 9648
     },
     {
-      "epoch": 0.26500233445936666,
+      "epoch": 0.27380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.724098518661844e-05,
-      "loss": 0.8699,
+      "learning_rate": 1.704143624228203e-05,
+      "loss": 0.9317,
       "step": 9649
     },
     {
-      "epoch": 0.2650297986872099,
+      "epoch": 0.27383654937570945,
       "grad_norm": 0.0,
-      "learning_rate": 1.724037166184469e-05,
-      "loss": 1.073,
+      "learning_rate": 1.7040783614200932e-05,
+      "loss": 0.9752,
       "step": 9650
     },
     {
-      "epoch": 0.26505726291505316,
+      "epoch": 0.2738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.7239758079782317e-05,
-      "loss": 0.962,
+      "learning_rate": 1.7040130926646023e-05,
+      "loss": 0.995,
       "step": 9651
     },
     {
-      "epoch": 0.2650847271428964,
+      "epoch": 0.2738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7239144440436173e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.7039478179622804e-05,
+      "loss": 0.934,
       "step": 9652
     },
     {
-      "epoch": 0.2651121913707396,
+      "epoch": 0.2739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.7238530743811113e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.7038825373136806e-05,
+      "loss": 0.8797,
       "step": 9653
     },
     {
-      "epoch": 0.2651396555985828,
+      "epoch": 0.273950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.7237916989911993e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.7038172507193526e-05,
+      "loss": 0.8624,
       "step": 9654
     },
     {
-      "epoch": 0.2651671198264261,
+      "epoch": 0.27397843359818386,
       "grad_norm": 0.0,
-      "learning_rate": 1.723730317874367e-05,
-      "loss": 1.063,
+      "learning_rate": 1.703751958179849e-05,
+      "loss": 0.9074,
       "step": 9655
     },
     {
-      "epoch": 0.2651945840542693,
+      "epoch": 0.27400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.7236689310311002e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.703686659695721e-05,
+      "loss": 0.9505,
       "step": 9656
     },
     {
-      "epoch": 0.26522204828211254,
+      "epoch": 0.2740351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.7236075384618848e-05,
-      "loss": 0.8517,
+      "learning_rate": 1.7036213552675202e-05,
+      "loss": 0.9815,
       "step": 9657
     },
     {
-      "epoch": 0.26524951250995576,
+      "epoch": 0.27406356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 1.7235461401672058e-05,
-      "loss": 0.9581,
+      "learning_rate": 1.703556044895798e-05,
+      "loss": 0.9696,
       "step": 9658
     },
     {
-      "epoch": 0.26527697673779904,
+      "epoch": 0.27409194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 1.7234847361475496e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.7034907285811062e-05,
+      "loss": 0.9706,
       "step": 9659
     },
     {
-      "epoch": 0.26530444096564226,
+      "epoch": 0.2741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.7234233264034022e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.703425406323997e-05,
+      "loss": 1.033,
       "step": 9660
     },
     {
-      "epoch": 0.2653319051934855,
+      "epoch": 0.27414869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.7233619109352488e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.7033600781250213e-05,
+      "loss": 0.9025,
       "step": 9661
     },
     {
-      "epoch": 0.2653593694213287,
+      "epoch": 0.27417707150964815,
       "grad_norm": 0.0,
-      "learning_rate": 1.7233004897435758e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.7032947439847316e-05,
+      "loss": 1.0637,
       "step": 9662
     },
     {
-      "epoch": 0.2653868336491719,
+      "epoch": 0.274205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.7232390628288695e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.7032294039036797e-05,
+      "loss": 1.0247,
       "step": 9663
     },
     {
-      "epoch": 0.2654142978770152,
+      "epoch": 0.2742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.7231776301916154e-05,
-      "loss": 1.0723,
+      "learning_rate": 1.703164057882417e-05,
+      "loss": 0.9498,
       "step": 9664
     },
     {
-      "epoch": 0.2654417621048584,
+      "epoch": 0.2742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.7231161918322993e-05,
-      "loss": 0.943,
+      "learning_rate": 1.7030987059214963e-05,
+      "loss": 0.9084,
       "step": 9665
     },
     {
-      "epoch": 0.26546922633270165,
+      "epoch": 0.2742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7230547477514084e-05,
-      "loss": 1.0537,
+      "learning_rate": 1.7030333480214693e-05,
+      "loss": 0.9756,
       "step": 9666
     },
     {
-      "epoch": 0.26549669056054487,
+      "epoch": 0.27431895573212256,
       "grad_norm": 0.0,
-      "learning_rate": 1.722993297949428e-05,
-      "loss": 1.0605,
+      "learning_rate": 1.7029679841828878e-05,
+      "loss": 0.9137,
       "step": 9667
     },
     {
-      "epoch": 0.26552415478838814,
+      "epoch": 0.2743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.7229318424268443e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.7029026144063042e-05,
+      "loss": 0.9093,
       "step": 9668
     },
     {
-      "epoch": 0.26555161901623137,
+      "epoch": 0.2743757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.7228703811841444e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.7028372386922703e-05,
+      "loss": 1.0598,
       "step": 9669
     },
     {
-      "epoch": 0.2655790832440746,
+      "epoch": 0.27440408626560725,
       "grad_norm": 0.0,
-      "learning_rate": 1.7228089142218135e-05,
-      "loss": 1.01,
+      "learning_rate": 1.7027718570413392e-05,
+      "loss": 0.9501,
       "step": 9670
     },
     {
-      "epoch": 0.2656065474719178,
+      "epoch": 0.27443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.7227474415403388e-05,
-      "loss": 1.0756,
+      "learning_rate": 1.7027064694540623e-05,
+      "loss": 1.0735,
       "step": 9671
     },
     {
-      "epoch": 0.2656340116997611,
+      "epoch": 0.274460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.7226859631402063e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.702641075930992e-05,
+      "loss": 0.9226,
       "step": 9672
     },
     {
-      "epoch": 0.2656614759276043,
+      "epoch": 0.27448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.7226244790219024e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.7025756764726815e-05,
+      "loss": 1.0083,
       "step": 9673
     },
     {
-      "epoch": 0.26568894015544753,
+      "epoch": 0.27451759364358685,
       "grad_norm": 0.0,
-      "learning_rate": 1.7225629891859143e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.7025102710796825e-05,
+      "loss": 0.9807,
       "step": 9674
     },
     {
-      "epoch": 0.26571640438329075,
+      "epoch": 0.2745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.7225014936327273e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.7024448597525478e-05,
+      "loss": 1.0627,
       "step": 9675
     },
     {
-      "epoch": 0.26574386861113397,
+      "epoch": 0.27457434733257663,
       "grad_norm": 0.0,
-      "learning_rate": 1.7224399923628285e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.7023794424918298e-05,
+      "loss": 0.94,
       "step": 9676
     },
     {
-      "epoch": 0.26577133283897725,
+      "epoch": 0.2746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.722378485376705e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.7023140192980806e-05,
+      "loss": 0.9567,
       "step": 9677
     },
     {
-      "epoch": 0.26579879706682047,
+      "epoch": 0.2746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.722316972674843e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.7022485901718533e-05,
+      "loss": 0.9915,
       "step": 9678
     },
     {
-      "epoch": 0.2658262612946637,
+      "epoch": 0.2746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.7222554542577296e-05,
-      "loss": 0.986,
+      "learning_rate": 1.702183155113701e-05,
+      "loss": 0.9929,
       "step": 9679
     },
     {
-      "epoch": 0.2658537255225069,
+      "epoch": 0.2746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.7221939301258508e-05,
-      "loss": 0.9264,
+      "learning_rate": 1.7021177141241757e-05,
+      "loss": 1.0731,
       "step": 9680
     },
     {
-      "epoch": 0.2658811897503502,
+      "epoch": 0.2747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.722132400279694e-05,
-      "loss": 0.9038,
+      "learning_rate": 1.702052267203831e-05,
+      "loss": 1.0367,
       "step": 9681
     },
     {
-      "epoch": 0.2659086539781934,
+      "epoch": 0.27474460839954595,
       "grad_norm": 0.0,
-      "learning_rate": 1.7220708647197463e-05,
-      "loss": 1.1146,
+      "learning_rate": 1.701986814353219e-05,
+      "loss": 0.9193,
       "step": 9682
     },
     {
-      "epoch": 0.26593611820603663,
+      "epoch": 0.27477298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.7220093234464936e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.7019213555728925e-05,
+      "loss": 0.9995,
       "step": 9683
     },
     {
-      "epoch": 0.26596358243387985,
+      "epoch": 0.27480136208853573,
       "grad_norm": 0.0,
-      "learning_rate": 1.7219477764604238e-05,
-      "loss": 0.943,
+      "learning_rate": 1.701855890863405e-05,
+      "loss": 1.0267,
       "step": 9684
     },
     {
-      "epoch": 0.26599104666172313,
+      "epoch": 0.27482973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.7218862237620235e-05,
-      "loss": 1.0167,
+      "learning_rate": 1.7017904202253093e-05,
+      "loss": 0.9977,
       "step": 9685
     },
     {
-      "epoch": 0.26601851088956635,
+      "epoch": 0.27485811577752556,
       "grad_norm": 0.0,
-      "learning_rate": 1.7218246653517797e-05,
-      "loss": 1.046,
+      "learning_rate": 1.7017249436591584e-05,
+      "loss": 0.9789,
       "step": 9686
     },
     {
-      "epoch": 0.26604597511740957,
+      "epoch": 0.2748864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.7217631012301796e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.7016594611655054e-05,
+      "loss": 0.9973,
       "step": 9687
     },
     {
-      "epoch": 0.2660734393452528,
+      "epoch": 0.27491486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 1.7217015313977105e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.7015939727449033e-05,
+      "loss": 0.8649,
       "step": 9688
     },
     {
-      "epoch": 0.266100903573096,
+      "epoch": 0.2749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.7216399558548586e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.701528478397905e-05,
+      "loss": 1.0213,
       "step": 9689
     },
     {
-      "epoch": 0.2661283678009393,
+      "epoch": 0.2749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.7215783746021126e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.7014629781250648e-05,
+      "loss": 0.9687,
       "step": 9690
     },
     {
-      "epoch": 0.2661558320287825,
+      "epoch": 0.275,
       "grad_norm": 0.0,
-      "learning_rate": 1.7215167876399587e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.701397471926935e-05,
+      "loss": 1.0318,
       "step": 9691
     },
     {
-      "epoch": 0.26618329625662573,
+      "epoch": 0.2750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.7214551949688843e-05,
-      "loss": 0.9,
+      "learning_rate": 1.701331959804069e-05,
+      "loss": 0.9601,
       "step": 9692
     },
     {
-      "epoch": 0.26621076048446896,
+      "epoch": 0.2750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.721393596589377e-05,
-      "loss": 0.8607,
+      "learning_rate": 1.7012664417570206e-05,
+      "loss": 0.847,
       "step": 9693
     },
     {
-      "epoch": 0.26623822471231223,
+      "epoch": 0.27508513053348466,
       "grad_norm": 0.0,
-      "learning_rate": 1.7213319925019246e-05,
-      "loss": 1.0046,
+      "learning_rate": 1.701200917786343e-05,
+      "loss": 1.0306,
       "step": 9694
     },
     {
-      "epoch": 0.26626568894015545,
+      "epoch": 0.27511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.7212703827070133e-05,
-      "loss": 0.941,
+      "learning_rate": 1.70113538789259e-05,
+      "loss": 0.9854,
       "step": 9695
     },
     {
-      "epoch": 0.2662931531679987,
+      "epoch": 0.2751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.7212087672051318e-05,
-      "loss": 1.004,
+      "learning_rate": 1.7010698520763145e-05,
+      "loss": 0.9968,
       "step": 9696
     },
     {
-      "epoch": 0.2663206173958419,
+      "epoch": 0.27517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.7211471459967673e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.7010043103380706e-05,
+      "loss": 0.9409,
       "step": 9697
     },
     {
-      "epoch": 0.2663480816236852,
+      "epoch": 0.27519863791146426,
       "grad_norm": 0.0,
-      "learning_rate": 1.721085519082407e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.7009387626784118e-05,
+      "loss": 0.9733,
       "step": 9698
     },
     {
-      "epoch": 0.2663755458515284,
+      "epoch": 0.2752270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.721023886462539e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.7008732090978917e-05,
+      "loss": 1.0019,
       "step": 9699
     },
     {
-      "epoch": 0.2664030100793716,
+      "epoch": 0.27525539160045404,
       "grad_norm": 0.0,
-      "learning_rate": 1.7209622481376504e-05,
-      "loss": 0.978,
+      "learning_rate": 1.700807649597064e-05,
+      "loss": 0.9865,
       "step": 9700
     },
     {
-      "epoch": 0.26643047430721484,
+      "epoch": 0.2752837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.720900604108229e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.7007420841764826e-05,
+      "loss": 0.9975,
       "step": 9701
     },
     {
-      "epoch": 0.2664579385350581,
+      "epoch": 0.2753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.7208389543747633e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.7006765128367012e-05,
+      "loss": 0.9818,
       "step": 9702
     },
     {
-      "epoch": 0.26648540276290134,
+      "epoch": 0.2753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.7207772989377405e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.7006109355782743e-05,
+      "loss": 1.0243,
       "step": 9703
     },
     {
-      "epoch": 0.26651286699074456,
+      "epoch": 0.2753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7207156377976482e-05,
-      "loss": 0.9912,
+      "learning_rate": 1.700545352401755e-05,
+      "loss": 0.9428,
       "step": 9704
     },
     {
-      "epoch": 0.2665403312185878,
+      "epoch": 0.2753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.7206539709549748e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.7004797633076977e-05,
+      "loss": 0.9011,
       "step": 9705
     },
     {
-      "epoch": 0.266567795446431,
+      "epoch": 0.27542565266742336,
       "grad_norm": 0.0,
-      "learning_rate": 1.720592298410208e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.7004141682966563e-05,
+      "loss": 0.9641,
       "step": 9706
     },
     {
-      "epoch": 0.2665952596742743,
+      "epoch": 0.2754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.7205306201638357e-05,
-      "loss": 0.9881,
+      "learning_rate": 1.7003485673691845e-05,
+      "loss": 0.8766,
       "step": 9707
     },
     {
-      "epoch": 0.2666227239021175,
+      "epoch": 0.2754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.7204689362163463e-05,
-      "loss": 0.927,
+      "learning_rate": 1.7002829605258372e-05,
+      "loss": 1.0045,
       "step": 9708
     },
     {
-      "epoch": 0.2666501881299607,
+      "epoch": 0.27551078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 1.7204072465682272e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.7002173477671685e-05,
+      "loss": 0.8654,
       "step": 9709
     },
     {
-      "epoch": 0.26667765235780394,
+      "epoch": 0.27553916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.720345551219967e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.7001517290937325e-05,
+      "loss": 1.0092,
       "step": 9710
     },
     {
-      "epoch": 0.2667051165856472,
+      "epoch": 0.2755675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.720283850172054e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.700086104506083e-05,
+      "loss": 0.9746,
       "step": 9711
     },
     {
-      "epoch": 0.26673258081349044,
+      "epoch": 0.27559591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.720222143424976e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.700020474004775e-05,
+      "loss": 1.0514,
       "step": 9712
     },
     {
-      "epoch": 0.26676004504133366,
+      "epoch": 0.27562429057888765,
       "grad_norm": 0.0,
-      "learning_rate": 1.7201604309792216e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.699954837590362e-05,
+      "loss": 0.8964,
       "step": 9713
     },
     {
-      "epoch": 0.2667875092691769,
+      "epoch": 0.2756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.7200987128352788e-05,
-      "loss": 0.9118,
+      "learning_rate": 1.6998891952633994e-05,
+      "loss": 0.8891,
       "step": 9714
     },
     {
-      "epoch": 0.26681497349702016,
+      "epoch": 0.27568104426787743,
       "grad_norm": 0.0,
-      "learning_rate": 1.720036988993636e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.6998235470244413e-05,
+      "loss": 0.947,
       "step": 9715
     },
     {
-      "epoch": 0.2668424377248634,
+      "epoch": 0.2757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.7199752594547818e-05,
-      "loss": 0.8821,
+      "learning_rate": 1.6997578928740423e-05,
+      "loss": 1.0577,
       "step": 9716
     },
     {
-      "epoch": 0.2668699019527066,
+      "epoch": 0.2757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.719913524219204e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.699692232812757e-05,
+      "loss": 0.9694,
       "step": 9717
     },
     {
-      "epoch": 0.2668973661805498,
+      "epoch": 0.27576617480136206,
       "grad_norm": 0.0,
-      "learning_rate": 1.719851783287392e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.6996265668411398e-05,
+      "loss": 0.9648,
       "step": 9718
     },
     {
-      "epoch": 0.26692483040839304,
+      "epoch": 0.275794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.7197900366598336e-05,
-      "loss": 1.0375,
+      "learning_rate": 1.699560894959746e-05,
+      "loss": 0.9577,
       "step": 9719
     },
     {
-      "epoch": 0.2669522946362363,
+      "epoch": 0.2758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.7197282843370176e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.6994952171691293e-05,
+      "loss": 1.0157,
       "step": 9720
     },
     {
-      "epoch": 0.26697975886407954,
+      "epoch": 0.27585130533484675,
       "grad_norm": 0.0,
-      "learning_rate": 1.7196665263194328e-05,
-      "loss": 1.0077,
+      "learning_rate": 1.6994295334698453e-05,
+      "loss": 1.0726,
       "step": 9721
     },
     {
-      "epoch": 0.26700722309192276,
+      "epoch": 0.27587968217934167,
       "grad_norm": 0.0,
-      "learning_rate": 1.7196047626075673e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.6993638438624485e-05,
+      "loss": 1.0064,
       "step": 9722
     },
     {
-      "epoch": 0.267034687319766,
+      "epoch": 0.2759080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.7195429932019108e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.6992981483474934e-05,
+      "loss": 0.9893,
       "step": 9723
     },
     {
-      "epoch": 0.26706215154760926,
+      "epoch": 0.27593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.719481218102951e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.699232446925536e-05,
+      "loss": 0.9848,
       "step": 9724
     },
     {
-      "epoch": 0.2670896157754525,
+      "epoch": 0.27596481271282636,
       "grad_norm": 0.0,
-      "learning_rate": 1.719419437311177e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.6991667395971306e-05,
+      "loss": 1.009,
       "step": 9725
     },
     {
-      "epoch": 0.2671170800032957,
+      "epoch": 0.2759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.719357650827078e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.6991010263628323e-05,
+      "loss": 1.0769,
       "step": 9726
     },
     {
-      "epoch": 0.2671445442311389,
+      "epoch": 0.27602156640181613,
       "grad_norm": 0.0,
-      "learning_rate": 1.7192958586511425e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.699035307223196e-05,
+      "loss": 0.9579,
       "step": 9727
     },
     {
-      "epoch": 0.2671720084589822,
+      "epoch": 0.276049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.7192340607838596e-05,
-      "loss": 0.9678,
+      "learning_rate": 1.698969582178777e-05,
+      "loss": 1.0547,
       "step": 9728
     },
     {
-      "epoch": 0.2671994726868254,
+      "epoch": 0.2760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.719172257225718e-05,
-      "loss": 0.8847,
+      "learning_rate": 1.6989038512301312e-05,
+      "loss": 0.9896,
       "step": 9729
     },
     {
-      "epoch": 0.26722693691466864,
+      "epoch": 0.2761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.719110447977207e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.6988381143778127e-05,
+      "loss": 0.893,
       "step": 9730
     },
     {
-      "epoch": 0.26725440114251187,
+      "epoch": 0.2761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.7190486330388155e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.698772371622377e-05,
+      "loss": 0.9482,
       "step": 9731
     },
     {
-      "epoch": 0.2672818653703551,
+      "epoch": 0.2761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.7189868124110332e-05,
-      "loss": 0.9441,
+      "learning_rate": 1.6987066229643797e-05,
+      "loss": 1.0653,
       "step": 9732
     },
     {
-      "epoch": 0.26730932959819836,
+      "epoch": 0.27619182746878546,
       "grad_norm": 0.0,
-      "learning_rate": 1.7189249860943483e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.6986408684043766e-05,
+      "loss": 0.9683,
       "step": 9733
     },
     {
-      "epoch": 0.2673367938260416,
+      "epoch": 0.27622020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.7188631540892504e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.6985751079429226e-05,
+      "loss": 0.9627,
       "step": 9734
     },
     {
-      "epoch": 0.2673642580538848,
+      "epoch": 0.27624858115777523,
       "grad_norm": 0.0,
-      "learning_rate": 1.718801316396229e-05,
-      "loss": 0.9869,
+      "learning_rate": 1.6985093415805728e-05,
+      "loss": 0.9644,
       "step": 9735
     },
     {
-      "epoch": 0.26739172228172803,
+      "epoch": 0.27627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.718739473015773e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.6984435693178837e-05,
+      "loss": 0.9625,
       "step": 9736
     },
     {
-      "epoch": 0.2674191865095713,
+      "epoch": 0.27630533484676506,
       "grad_norm": 0.0,
-      "learning_rate": 1.7186776239483716e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.6983777911554102e-05,
+      "loss": 0.9298,
       "step": 9737
     },
     {
-      "epoch": 0.2674466507374145,
+      "epoch": 0.2763337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.718615769194515e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.698312007093708e-05,
+      "loss": 0.9477,
       "step": 9738
     },
     {
-      "epoch": 0.26747411496525775,
+      "epoch": 0.27636208853575484,
       "grad_norm": 0.0,
-      "learning_rate": 1.7185539087546917e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.6982462171333327e-05,
+      "loss": 0.9716,
       "step": 9739
     },
     {
-      "epoch": 0.26750157919310097,
+      "epoch": 0.2763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.7184920426293915e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.6981804212748406e-05,
+      "loss": 0.8628,
       "step": 9740
     },
     {
-      "epoch": 0.26752904342094425,
+      "epoch": 0.2764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.718430170819104e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.698114619518787e-05,
+      "loss": 0.9435,
       "step": 9741
     },
     {
-      "epoch": 0.26755650764878747,
+      "epoch": 0.2764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.7183682933243192e-05,
-      "loss": 1.0138,
+      "learning_rate": 1.6980488118657276e-05,
+      "loss": 0.9397,
       "step": 9742
     },
     {
-      "epoch": 0.2675839718766307,
+      "epoch": 0.2764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7183064101455255e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.697982998316219e-05,
+      "loss": 0.9825,
       "step": 9743
     },
     {
-      "epoch": 0.2676114361044739,
+      "epoch": 0.2765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.7182445212832135e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.6979171788708158e-05,
+      "loss": 1.004,
       "step": 9744
     },
     {
-      "epoch": 0.26763890033231713,
+      "epoch": 0.27653234960272416,
       "grad_norm": 0.0,
-      "learning_rate": 1.7181826267378726e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.6978513535300756e-05,
+      "loss": 0.9902,
       "step": 9745
     },
     {
-      "epoch": 0.2676663645601604,
+      "epoch": 0.2765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.7181207265099927e-05,
-      "loss": 1.0464,
+      "learning_rate": 1.6977855222945532e-05,
+      "loss": 0.9646,
       "step": 9746
     },
     {
-      "epoch": 0.26769382878800363,
+      "epoch": 0.276589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.7180588206000632e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.6977196851648053e-05,
+      "loss": 0.9703,
       "step": 9747
     },
     {
-      "epoch": 0.26772129301584685,
+      "epoch": 0.27661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.7179969090085742e-05,
-      "loss": 0.9092,
+      "learning_rate": 1.6976538421413877e-05,
+      "loss": 0.8339,
       "step": 9748
     },
     {
-      "epoch": 0.26774875724369007,
+      "epoch": 0.27664585698070376,
       "grad_norm": 0.0,
-      "learning_rate": 1.717934991736016e-05,
-      "loss": 1.012,
+      "learning_rate": 1.6975879932248572e-05,
+      "loss": 0.989,
       "step": 9749
     },
     {
-      "epoch": 0.26777622147153335,
+      "epoch": 0.2766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.7178730687828772e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.6975221384157692e-05,
+      "loss": 0.9668,
       "step": 9750
     },
     {
-      "epoch": 0.26780368569937657,
+      "epoch": 0.27670261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.717811140149649e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.6974562777146804e-05,
+      "loss": 0.985,
       "step": 9751
     },
     {
-      "epoch": 0.2678311499272198,
+      "epoch": 0.2767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.717749205836821e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.697390411122147e-05,
+      "loss": 0.9734,
       "step": 9752
     },
     {
-      "epoch": 0.267858614155063,
+      "epoch": 0.2767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.7176872658448836e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.6973245386387254e-05,
+      "loss": 1.0173,
       "step": 9753
     },
     {
-      "epoch": 0.2678860783829063,
+      "epoch": 0.27678774120317823,
       "grad_norm": 0.0,
-      "learning_rate": 1.7176253201743263e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.6972586602649724e-05,
+      "loss": 1.0329,
       "step": 9754
     },
     {
-      "epoch": 0.2679135426107495,
+      "epoch": 0.2768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.7175633688256393e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.6971927760014437e-05,
+      "loss": 0.9137,
       "step": 9755
     },
     {
-      "epoch": 0.26794100683859273,
+      "epoch": 0.276844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.7175014117993132e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.6971268858486968e-05,
+      "loss": 0.9621,
       "step": 9756
     },
     {
-      "epoch": 0.26796847106643595,
+      "epoch": 0.27687287173666286,
       "grad_norm": 0.0,
-      "learning_rate": 1.717439449095838e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.6970609898072874e-05,
+      "loss": 0.8975,
       "step": 9757
     },
     {
-      "epoch": 0.2679959352942792,
+      "epoch": 0.2769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.717377480715704e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.6969950878777725e-05,
+      "loss": 0.9341,
       "step": 9758
     },
     {
-      "epoch": 0.26802339952212245,
+      "epoch": 0.2769296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.7173155066594012e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.6969291800607087e-05,
+      "loss": 1.0447,
       "step": 9759
     },
     {
-      "epoch": 0.2680508637499657,
+      "epoch": 0.27695800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.7172535269274204e-05,
-      "loss": 1.0885,
+      "learning_rate": 1.696863266356653e-05,
+      "loss": 0.9779,
       "step": 9760
     },
     {
-      "epoch": 0.2680783279778089,
+      "epoch": 0.27698637911464247,
       "grad_norm": 0.0,
-      "learning_rate": 1.7171915415202522e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.696797346766162e-05,
+      "loss": 0.9496,
       "step": 9761
     },
     {
-      "epoch": 0.2681057922056521,
+      "epoch": 0.2770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.717129550438386e-05,
-      "loss": 0.9244,
+      "learning_rate": 1.696731421289792e-05,
+      "loss": 0.9344,
       "step": 9762
     },
     {
-      "epoch": 0.2681332564334954,
+      "epoch": 0.27704313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.7170675536823136e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.6966654899281008e-05,
+      "loss": 0.9439,
       "step": 9763
     },
     {
-      "epoch": 0.2681607206613386,
+      "epoch": 0.2770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.7170055512525244e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.6965995526816447e-05,
+      "loss": 1.0322,
       "step": 9764
     },
     {
-      "epoch": 0.26818818488918184,
+      "epoch": 0.277099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.71694354314951e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.696533609550981e-05,
+      "loss": 0.9849,
       "step": 9765
     },
     {
-      "epoch": 0.26821564911702506,
+      "epoch": 0.27712826333711693,
       "grad_norm": 0.0,
-      "learning_rate": 1.7168815293737604e-05,
-      "loss": 1.0592,
+      "learning_rate": 1.6964676605366666e-05,
+      "loss": 0.9915,
       "step": 9766
     },
     {
-      "epoch": 0.26824311334486833,
+      "epoch": 0.2771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.7168195099257668e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.6964017056392586e-05,
+      "loss": 0.9724,
       "step": 9767
     },
     {
-      "epoch": 0.26827057757271155,
+      "epoch": 0.2771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.7167574848060192e-05,
-      "loss": 0.9938,
+      "learning_rate": 1.6963357448593138e-05,
+      "loss": 0.9838,
       "step": 9768
     },
     {
-      "epoch": 0.2682980418005548,
+      "epoch": 0.27721339387060157,
       "grad_norm": 0.0,
-      "learning_rate": 1.7166954540150086e-05,
-      "loss": 0.8975,
+      "learning_rate": 1.69626977819739e-05,
+      "loss": 1.0791,
       "step": 9769
     },
     {
-      "epoch": 0.268325506028398,
+      "epoch": 0.2772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.7166334175532262e-05,
-      "loss": 1.1041,
+      "learning_rate": 1.696203805654044e-05,
+      "loss": 0.8985,
       "step": 9770
     },
     {
-      "epoch": 0.2683529702562412,
+      "epoch": 0.2772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.7165713754211627e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.696137827229833e-05,
+      "loss": 0.823,
       "step": 9771
     },
     {
-      "epoch": 0.2683804344840845,
+      "epoch": 0.27729852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.7165093276193087e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.6960718429253148e-05,
+      "loss": 1.0191,
       "step": 9772
     },
     {
-      "epoch": 0.2684078987119277,
+      "epoch": 0.27732690124858117,
       "grad_norm": 0.0,
-      "learning_rate": 1.7164472741481554e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.696005852741046e-05,
+      "loss": 0.8746,
       "step": 9773
     },
     {
-      "epoch": 0.26843536293977094,
+      "epoch": 0.27735527809307603,
       "grad_norm": 0.0,
-      "learning_rate": 1.7163852150081938e-05,
-      "loss": 0.8674,
+      "learning_rate": 1.6959398566775847e-05,
+      "loss": 0.9864,
       "step": 9774
     },
     {
-      "epoch": 0.26846282716761416,
+      "epoch": 0.27738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.7163231501999148e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.6958738547354884e-05,
+      "loss": 0.8915,
       "step": 9775
     },
     {
-      "epoch": 0.26849029139545744,
+      "epoch": 0.27741203178206586,
       "grad_norm": 0.0,
-      "learning_rate": 1.7162610797238094e-05,
-      "loss": 0.996,
+      "learning_rate": 1.695807846915314e-05,
+      "loss": 1.0839,
       "step": 9776
     },
     {
-      "epoch": 0.26851775562330066,
+      "epoch": 0.2774404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.716199003580369e-05,
-      "loss": 0.9803,
+      "learning_rate": 1.6957418332176195e-05,
+      "loss": 1.0331,
       "step": 9777
     },
     {
-      "epoch": 0.2685452198511439,
+      "epoch": 0.27746878547105563,
       "grad_norm": 0.0,
-      "learning_rate": 1.716136921770085e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.6956758136429627e-05,
+      "loss": 0.9398,
       "step": 9778
     },
     {
-      "epoch": 0.2685726840789871,
+      "epoch": 0.2774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.716074834293448e-05,
-      "loss": 0.9604,
+      "learning_rate": 1.6956097881919006e-05,
+      "loss": 0.8641,
       "step": 9779
     },
     {
-      "epoch": 0.2686001483068304,
+      "epoch": 0.2775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7160127411509493e-05,
-      "loss": 0.9481,
+      "learning_rate": 1.6955437568649916e-05,
+      "loss": 0.9992,
       "step": 9780
     },
     {
-      "epoch": 0.2686276125346736,
+      "epoch": 0.27755391600454027,
       "grad_norm": 0.0,
-      "learning_rate": 1.715950642343081e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.6954777196627934e-05,
+      "loss": 0.9771,
       "step": 9781
     },
     {
-      "epoch": 0.2686550767625168,
+      "epoch": 0.2775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.7158885378703333e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.6954116765858634e-05,
+      "loss": 0.9809,
       "step": 9782
     },
     {
-      "epoch": 0.26868254099036004,
+      "epoch": 0.2776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.7158264277331983e-05,
-      "loss": 1.0352,
+      "learning_rate": 1.69534562763476e-05,
+      "loss": 0.9878,
       "step": 9783
     },
     {
-      "epoch": 0.26871000521820326,
+      "epoch": 0.27763904653802496,
       "grad_norm": 0.0,
-      "learning_rate": 1.7157643119321678e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.69527957281004e-05,
+      "loss": 1.0061,
       "step": 9784
     },
     {
-      "epoch": 0.26873746944604654,
+      "epoch": 0.2776674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.7157021904677323e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.6952135121122634e-05,
+      "loss": 0.9738,
       "step": 9785
     },
     {
-      "epoch": 0.26876493367388976,
+      "epoch": 0.27769580022701473,
       "grad_norm": 0.0,
-      "learning_rate": 1.715640063340384e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.6951474455419862e-05,
+      "loss": 0.9358,
       "step": 9786
     },
     {
-      "epoch": 0.268792397901733,
+      "epoch": 0.27772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.715577930550614e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.6950813730997673e-05,
+      "loss": 0.9002,
       "step": 9787
     },
     {
-      "epoch": 0.2688198621295762,
+      "epoch": 0.27775255391600456,
       "grad_norm": 0.0,
-      "learning_rate": 1.7155157920989146e-05,
-      "loss": 0.8339,
+      "learning_rate": 1.695015294786165e-05,
+      "loss": 1.0134,
       "step": 9788
     },
     {
-      "epoch": 0.2688473263574195,
+      "epoch": 0.2777809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.7154536479857773e-05,
-      "loss": 1.0315,
+      "learning_rate": 1.6949492106017374e-05,
+      "loss": 1.0112,
       "step": 9789
     },
     {
-      "epoch": 0.2688747905852627,
+      "epoch": 0.27780930760499434,
       "grad_norm": 0.0,
-      "learning_rate": 1.7153914982116932e-05,
-      "loss": 1.0735,
+      "learning_rate": 1.6948831205470424e-05,
+      "loss": 1.0105,
       "step": 9790
     },
     {
-      "epoch": 0.2689022548131059,
+      "epoch": 0.2778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.7153293427771545e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.6948170246226385e-05,
+      "loss": 0.8409,
       "step": 9791
     },
     {
-      "epoch": 0.26892971904094914,
+      "epoch": 0.2778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.715267181682653e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.694750922829084e-05,
+      "loss": 0.8751,
       "step": 9792
     },
     {
-      "epoch": 0.2689571832687924,
+      "epoch": 0.277894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7152050149286804e-05,
-      "loss": 1.0861,
+      "learning_rate": 1.6946848151669367e-05,
+      "loss": 0.9571,
       "step": 9793
     },
     {
-      "epoch": 0.26898464749663564,
+      "epoch": 0.2779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.7151428425157288e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.694618701636756e-05,
+      "loss": 0.9894,
       "step": 9794
     },
     {
-      "epoch": 0.26901211172447886,
+      "epoch": 0.2779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.71508066444429e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.6945525822391e-05,
+      "loss": 0.9671,
       "step": 9795
     },
     {
-      "epoch": 0.2690395759523221,
+      "epoch": 0.27797956867196366,
       "grad_norm": 0.0,
-      "learning_rate": 1.7150184807148558e-05,
-      "loss": 0.964,
+      "learning_rate": 1.6944864569745273e-05,
+      "loss": 0.963,
       "step": 9796
     },
     {
-      "epoch": 0.26906704018016536,
+      "epoch": 0.2780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.7149562913279186e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.6944203258435957e-05,
+      "loss": 0.968,
       "step": 9797
     },
     {
-      "epoch": 0.2690945044080086,
+      "epoch": 0.27803632236095344,
       "grad_norm": 0.0,
-      "learning_rate": 1.7148940962839706e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.6943541888468647e-05,
+      "loss": 0.9695,
       "step": 9798
     },
     {
-      "epoch": 0.2691219686358518,
+      "epoch": 0.27806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.714831895583503e-05,
-      "loss": 0.9661,
+      "learning_rate": 1.6942880459848927e-05,
+      "loss": 0.8482,
       "step": 9799
     },
     {
-      "epoch": 0.269149432863695,
+      "epoch": 0.27809307604994327,
       "grad_norm": 0.0,
-      "learning_rate": 1.7147696892270088e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.6942218972582384e-05,
+      "loss": 0.9257,
       "step": 9800
     },
     {
-      "epoch": 0.26917689709153825,
+      "epoch": 0.2781214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.71470747721498e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.69415574266746e-05,
+      "loss": 0.9778,
       "step": 9801
     },
     {
-      "epoch": 0.2692043613193815,
+      "epoch": 0.27814982973893304,
       "grad_norm": 0.0,
-      "learning_rate": 1.7146452595479085e-05,
-      "loss": 0.8893,
+      "learning_rate": 1.6940895822131173e-05,
+      "loss": 0.9258,
       "step": 9802
     },
     {
-      "epoch": 0.26923182554722475,
+      "epoch": 0.2781782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.7145830362262872e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.6940234158957686e-05,
+      "loss": 0.8731,
       "step": 9803
     },
     {
-      "epoch": 0.26925928977506797,
+      "epoch": 0.2782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.714520807250608e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.693957243715973e-05,
+      "loss": 1.1211,
       "step": 9804
     },
     {
-      "epoch": 0.2692867540029112,
+      "epoch": 0.27823496027241773,
       "grad_norm": 0.0,
-      "learning_rate": 1.7144585726213636e-05,
-      "loss": 0.8748,
+      "learning_rate": 1.6938910656742895e-05,
+      "loss": 0.9271,
       "step": 9805
     },
     {
-      "epoch": 0.26931421823075447,
+      "epoch": 0.2782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.7143963323390456e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.6938248817712767e-05,
+      "loss": 1.0969,
       "step": 9806
     },
     {
-      "epoch": 0.2693416824585977,
+      "epoch": 0.2782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.7143340864041476e-05,
-      "loss": 1.1069,
+      "learning_rate": 1.693758692007494e-05,
+      "loss": 1.0143,
       "step": 9807
     },
     {
-      "epoch": 0.2693691466864409,
+      "epoch": 0.27832009080590236,
       "grad_norm": 0.0,
-      "learning_rate": 1.7142718348171613e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.6936924963835006e-05,
+      "loss": 0.8639,
       "step": 9808
     },
     {
-      "epoch": 0.26939661091428413,
+      "epoch": 0.2783484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.71420957757858e-05,
-      "loss": 1.016,
+      "learning_rate": 1.6936262948998552e-05,
+      "loss": 1.0151,
       "step": 9809
     },
     {
-      "epoch": 0.2694240751421274,
+      "epoch": 0.2783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.7141473146888955e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.6935600875571175e-05,
+      "loss": 0.9825,
       "step": 9810
     },
     {
-      "epoch": 0.2694515393699706,
+      "epoch": 0.27840522133938705,
       "grad_norm": 0.0,
-      "learning_rate": 1.714085046148601e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.6934938743558467e-05,
+      "loss": 0.9553,
       "step": 9811
     },
     {
-      "epoch": 0.26947900359781385,
+      "epoch": 0.27843359818388197,
       "grad_norm": 0.0,
-      "learning_rate": 1.714022771958189e-05,
-      "loss": 1.0568,
+      "learning_rate": 1.6934276552966018e-05,
+      "loss": 0.8992,
       "step": 9812
     },
     {
-      "epoch": 0.26950646782565707,
+      "epoch": 0.27846197502837683,
       "grad_norm": 0.0,
-      "learning_rate": 1.7139604921181522e-05,
-      "loss": 0.8945,
+      "learning_rate": 1.6933614303799424e-05,
+      "loss": 0.9848,
       "step": 9813
     },
     {
-      "epoch": 0.2695339320535003,
+      "epoch": 0.27849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.7138982066289832e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.6932951996064276e-05,
+      "loss": 0.9436,
       "step": 9814
     },
     {
-      "epoch": 0.26956139628134357,
+      "epoch": 0.2785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.7138359154911755e-05,
-      "loss": 0.9256,
+      "learning_rate": 1.693228962976617e-05,
+      "loss": 0.9488,
       "step": 9815
     },
     {
-      "epoch": 0.2695888605091868,
+      "epoch": 0.2785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.7137736187052212e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.6931627204910707e-05,
+      "loss": 0.9583,
       "step": 9816
     },
     {
-      "epoch": 0.26961632473703,
+      "epoch": 0.27857548240635643,
       "grad_norm": 0.0,
-      "learning_rate": 1.7137113162716135e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.6930964721503477e-05,
+      "loss": 0.9084,
       "step": 9817
     },
     {
-      "epoch": 0.26964378896487323,
+      "epoch": 0.2786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.7136490081908456e-05,
-      "loss": 1.1207,
+      "learning_rate": 1.693030217955007e-05,
+      "loss": 0.9865,
       "step": 9818
     },
     {
-      "epoch": 0.2696712531927165,
+      "epoch": 0.2786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.7135866944634104e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.6929639579056095e-05,
+      "loss": 0.9458,
       "step": 9819
     },
     {
-      "epoch": 0.26969871742055973,
+      "epoch": 0.27866061293984107,
       "grad_norm": 0.0,
-      "learning_rate": 1.7135243750898007e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.692897692002714e-05,
+      "loss": 0.8996,
       "step": 9820
     },
     {
-      "epoch": 0.26972618164840295,
+      "epoch": 0.278688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.7134620500705094e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.6928314202468806e-05,
+      "loss": 1.0341,
       "step": 9821
     },
     {
-      "epoch": 0.2697536458762462,
+      "epoch": 0.2787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.7133997194060304e-05,
-      "loss": 1.1062,
+      "learning_rate": 1.6927651426386692e-05,
+      "loss": 0.8465,
       "step": 9822
     },
     {
-      "epoch": 0.26978111010408945,
+      "epoch": 0.27874574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.7133373830968568e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.6926988591786392e-05,
+      "loss": 0.9747,
       "step": 9823
     },
     {
-      "epoch": 0.26980857433193267,
+      "epoch": 0.2787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.713275041143481e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.6926325698673513e-05,
+      "loss": 1.0259,
       "step": 9824
     },
     {
-      "epoch": 0.2698360385597759,
+      "epoch": 0.27880249716231553,
       "grad_norm": 0.0,
-      "learning_rate": 1.713212693546397e-05,
-      "loss": 1.0886,
+      "learning_rate": 1.6925662747053646e-05,
+      "loss": 0.9654,
       "step": 9825
     },
     {
-      "epoch": 0.2698635027876191,
+      "epoch": 0.27883087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.7131503403060975e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.692499973693239e-05,
+      "loss": 0.9984,
       "step": 9826
     },
     {
-      "epoch": 0.26989096701546234,
+      "epoch": 0.27885925085130536,
       "grad_norm": 0.0,
-      "learning_rate": 1.7130879814230764e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.6924336668315357e-05,
+      "loss": 1.0177,
       "step": 9827
     },
     {
-      "epoch": 0.2699184312433056,
+      "epoch": 0.2788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.7130256168978273e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.6923673541208142e-05,
+      "loss": 0.9485,
       "step": 9828
     },
     {
-      "epoch": 0.26994589547114883,
+      "epoch": 0.27891600454029514,
       "grad_norm": 0.0,
-      "learning_rate": 1.7129632467308433e-05,
-      "loss": 1.0269,
+      "learning_rate": 1.6923010355616342e-05,
+      "loss": 1.0103,
       "step": 9829
     },
     {
-      "epoch": 0.26997335969899205,
+      "epoch": 0.27894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.7129008709226176e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.692234711154556e-05,
+      "loss": 0.919,
       "step": 9830
     },
     {
-      "epoch": 0.2700008239268353,
+      "epoch": 0.2789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.712838489473644e-05,
-      "loss": 1.1125,
+      "learning_rate": 1.6921683809001403e-05,
+      "loss": 0.988,
       "step": 9831
     },
     {
-      "epoch": 0.27002828815467855,
+      "epoch": 0.27900113507377977,
       "grad_norm": 0.0,
-      "learning_rate": 1.7127761023844166e-05,
-      "loss": 1.1042,
+      "learning_rate": 1.6921020447989472e-05,
+      "loss": 0.9691,
       "step": 9832
     },
     {
-      "epoch": 0.2700557523825218,
+      "epoch": 0.2790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.7127137096554284e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.6920357028515368e-05,
+      "loss": 0.9513,
       "step": 9833
     },
     {
-      "epoch": 0.270083216610365,
+      "epoch": 0.2790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.712651311287173e-05,
-      "loss": 0.912,
+      "learning_rate": 1.69196935505847e-05,
+      "loss": 0.9617,
       "step": 9834
     },
     {
-      "epoch": 0.2701106808382082,
+      "epoch": 0.27908626560726446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7125889072801446e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.691903001420307e-05,
+      "loss": 1.1294,
       "step": 9835
     },
     {
-      "epoch": 0.2701381450660515,
+      "epoch": 0.2791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.7125264976348365e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.691836641937608e-05,
+      "loss": 0.9564,
       "step": 9836
     },
     {
-      "epoch": 0.2701656092938947,
+      "epoch": 0.27914301929625424,
       "grad_norm": 0.0,
-      "learning_rate": 1.712464082351743e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.691770276610934e-05,
+      "loss": 0.947,
       "step": 9837
     },
     {
-      "epoch": 0.27019307352173794,
+      "epoch": 0.27917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.7124016614313577e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.691703905440845e-05,
+      "loss": 1.0251,
       "step": 9838
     },
     {
-      "epoch": 0.27022053774958116,
+      "epoch": 0.27919977298524407,
       "grad_norm": 0.0,
-      "learning_rate": 1.7123392348741744e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.6916375284279024e-05,
+      "loss": 1.0725,
       "step": 9839
     },
     {
-      "epoch": 0.2702480019774244,
+      "epoch": 0.2792281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.7122768026806872e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.6915711455726665e-05,
+      "loss": 0.9552,
       "step": 9840
     },
     {
-      "epoch": 0.27027546620526766,
+      "epoch": 0.27925652667423384,
       "grad_norm": 0.0,
-      "learning_rate": 1.71221436485139e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.6915047568756978e-05,
+      "loss": 0.8916,
       "step": 9841
     },
     {
-      "epoch": 0.2703029304331109,
+      "epoch": 0.2792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.7121519213867766e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.6914383623375577e-05,
+      "loss": 0.9339,
       "step": 9842
     },
     {
-      "epoch": 0.2703303946609541,
+      "epoch": 0.2793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.712089472287342e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.6913719619588062e-05,
+      "loss": 0.9411,
       "step": 9843
     },
     {
-      "epoch": 0.2703578588887973,
+      "epoch": 0.2793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.712027017553579e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.6913055557400052e-05,
+      "loss": 0.9532,
       "step": 9844
     },
     {
-      "epoch": 0.2703853231166406,
+      "epoch": 0.2793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.7119645571859828e-05,
-      "loss": 0.88,
+      "learning_rate": 1.6912391436817147e-05,
+      "loss": 0.8994,
       "step": 9845
     },
     {
-      "epoch": 0.2704127873444838,
+      "epoch": 0.2793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.711902091185047e-05,
-      "loss": 1.0838,
+      "learning_rate": 1.691172725784496e-05,
+      "loss": 0.9722,
       "step": 9846
     },
     {
-      "epoch": 0.27044025157232704,
+      "epoch": 0.27942678774120316,
       "grad_norm": 0.0,
-      "learning_rate": 1.711839619551266e-05,
-      "loss": 1.0721,
+      "learning_rate": 1.6911063020489104e-05,
+      "loss": 0.9063,
       "step": 9847
     },
     {
-      "epoch": 0.27046771580017026,
+      "epoch": 0.2794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.7117771422851347e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.6910398724755187e-05,
+      "loss": 1.0082,
       "step": 9848
     },
     {
-      "epoch": 0.27049518002801354,
+      "epoch": 0.27948354143019294,
       "grad_norm": 0.0,
-      "learning_rate": 1.7117146593871464e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.6909734370648822e-05,
+      "loss": 1.0424,
       "step": 9849
     },
     {
-      "epoch": 0.27052264425585676,
+      "epoch": 0.27951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.7116521708577964e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.6909069958175622e-05,
+      "loss": 1.0959,
       "step": 9850
     },
     {
-      "epoch": 0.2705501084837,
+      "epoch": 0.27954029511918277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7115896766975786e-05,
-      "loss": 1.0037,
+      "learning_rate": 1.6908405487341195e-05,
+      "loss": 0.989,
       "step": 9851
     },
     {
-      "epoch": 0.2705775727115432,
+      "epoch": 0.27956867196367763,
       "grad_norm": 0.0,
-      "learning_rate": 1.7115271769069876e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.6907740958151158e-05,
+      "loss": 0.9211,
       "step": 9852
     },
     {
-      "epoch": 0.2706050369393864,
+      "epoch": 0.27959704880817254,
       "grad_norm": 0.0,
-      "learning_rate": 1.711464671486518e-05,
-      "loss": 0.8611,
+      "learning_rate": 1.6907076370611118e-05,
+      "loss": 1.0191,
       "step": 9853
     },
     {
-      "epoch": 0.2706325011672297,
+      "epoch": 0.2796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.711402160436664e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.69064117247267e-05,
+      "loss": 1.0896,
       "step": 9854
     },
     {
-      "epoch": 0.2706599653950729,
+      "epoch": 0.2796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.711339643757921e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.6905747020503508e-05,
+      "loss": 1.0707,
       "step": 9855
     },
     {
-      "epoch": 0.27068742962291614,
+      "epoch": 0.27968217934165723,
       "grad_norm": 0.0,
-      "learning_rate": 1.711277121450783e-05,
-      "loss": 0.8702,
+      "learning_rate": 1.690508225794716e-05,
+      "loss": 0.9525,
       "step": 9856
     },
     {
-      "epoch": 0.27071489385075936,
+      "epoch": 0.2797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.711214593515745e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.6904417437063275e-05,
+      "loss": 0.9408,
       "step": 9857
     },
     {
-      "epoch": 0.27074235807860264,
+      "epoch": 0.279738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.7111520599533015e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.6903752557857462e-05,
+      "loss": 1.0753,
       "step": 9858
     },
     {
-      "epoch": 0.27076982230644586,
+      "epoch": 0.27976730987514187,
       "grad_norm": 0.0,
-      "learning_rate": 1.7110895207639477e-05,
-      "loss": 0.9042,
+      "learning_rate": 1.6903087620335345e-05,
+      "loss": 1.0043,
       "step": 9859
     },
     {
-      "epoch": 0.2707972865342891,
+      "epoch": 0.2797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.711026975948178e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.690242262450253e-05,
+      "loss": 0.9836,
       "step": 9860
     },
     {
-      "epoch": 0.2708247507621323,
+      "epoch": 0.27982406356413164,
       "grad_norm": 0.0,
-      "learning_rate": 1.7109644255064873e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.6901757570364648e-05,
+      "loss": 0.863,
       "step": 9861
     },
     {
-      "epoch": 0.2708522149899756,
+      "epoch": 0.27985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.7109018694393708e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.6901092457927304e-05,
+      "loss": 1.0193,
       "step": 9862
     },
     {
-      "epoch": 0.2708796792178188,
+      "epoch": 0.27988081725312147,
       "grad_norm": 0.0,
-      "learning_rate": 1.7108393077473233e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.6900427287196125e-05,
+      "loss": 0.9277,
       "step": 9863
     },
     {
-      "epoch": 0.270907143445662,
+      "epoch": 0.27990919409761633,
       "grad_norm": 0.0,
-      "learning_rate": 1.7107767404308394e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.6899762058176726e-05,
+      "loss": 0.8938,
       "step": 9864
     },
     {
-      "epoch": 0.27093460767350525,
+      "epoch": 0.27993757094211125,
       "grad_norm": 0.0,
-      "learning_rate": 1.7107141674904153e-05,
-      "loss": 1.0418,
+      "learning_rate": 1.6899096770874725e-05,
+      "loss": 1.0368,
       "step": 9865
     },
     {
-      "epoch": 0.27096207190134847,
+      "epoch": 0.2799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.7106515889265455e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.6898431425295746e-05,
+      "loss": 0.8856,
       "step": 9866
     },
     {
-      "epoch": 0.27098953612919174,
+      "epoch": 0.279994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.7105890047397244e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.6897766021445407e-05,
+      "loss": 1.0454,
       "step": 9867
     },
     {
-      "epoch": 0.27101700035703497,
+      "epoch": 0.28002270147559594,
       "grad_norm": 0.0,
-      "learning_rate": 1.710526414930448e-05,
-      "loss": 1.1725,
+      "learning_rate": 1.6897100559329324e-05,
+      "loss": 0.8597,
       "step": 9868
     },
     {
-      "epoch": 0.2710444645848782,
+      "epoch": 0.2800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.7104638194992117e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.6896435038953126e-05,
+      "loss": 1.04,
       "step": 9869
     },
     {
-      "epoch": 0.2710719288127214,
+      "epoch": 0.2800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.71040121844651e-05,
-      "loss": 0.96,
+      "learning_rate": 1.689576946032243e-05,
+      "loss": 0.906,
       "step": 9870
     },
     {
-      "epoch": 0.2710993930405647,
+      "epoch": 0.28010783200908057,
       "grad_norm": 0.0,
-      "learning_rate": 1.710338611772839e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.689510382344286e-05,
+      "loss": 1.1369,
       "step": 9871
     },
     {
-      "epoch": 0.2711268572684079,
+      "epoch": 0.2801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.7102759994786938e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.689443812832004e-05,
+      "loss": 0.9812,
       "step": 9872
     },
     {
-      "epoch": 0.2711543214962511,
+      "epoch": 0.2801645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.7102133815645695e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.6893772374959586e-05,
+      "loss": 0.9937,
       "step": 9873
     },
     {
-      "epoch": 0.27118178572409435,
+      "epoch": 0.28019296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.710150758030962e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.689310656336713e-05,
+      "loss": 0.991,
       "step": 9874
     },
     {
-      "epoch": 0.2712092499519376,
+      "epoch": 0.2802213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.7100881288783665e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.6892440693548294e-05,
+      "loss": 0.9985,
       "step": 9875
     },
     {
-      "epoch": 0.27123671417978085,
+      "epoch": 0.28024971623155503,
       "grad_norm": 0.0,
-      "learning_rate": 1.710025494107279e-05,
-      "loss": 0.9374,
+      "learning_rate": 1.6891774765508698e-05,
+      "loss": 0.9813,
       "step": 9876
     },
     {
-      "epoch": 0.27126417840762407,
+      "epoch": 0.28027809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.7099628537181944e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.6891108779253974e-05,
+      "loss": 0.9256,
       "step": 9877
     },
     {
-      "epoch": 0.2712916426354673,
+      "epoch": 0.2803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.7099002077116087e-05,
-      "loss": 0.922,
+      "learning_rate": 1.6890442734789743e-05,
+      "loss": 1.0555,
       "step": 9878
     },
     {
-      "epoch": 0.2713191068633105,
+      "epoch": 0.2803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.7098375560880176e-05,
-      "loss": 1.0337,
+      "learning_rate": 1.6889776632121634e-05,
+      "loss": 1.0535,
       "step": 9879
     },
     {
-      "epoch": 0.2713465710911538,
+      "epoch": 0.28036322360953464,
       "grad_norm": 0.0,
-      "learning_rate": 1.7097748988479167e-05,
-      "loss": 0.9567,
+      "learning_rate": 1.6889110471255272e-05,
+      "loss": 1.131,
       "step": 9880
     },
     {
-      "epoch": 0.271374035318997,
+      "epoch": 0.2803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.709712235991802e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.6888444252196284e-05,
+      "loss": 0.9763,
       "step": 9881
     },
     {
-      "epoch": 0.27140149954684023,
+      "epoch": 0.2804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7096495675201693e-05,
-      "loss": 0.9654,
+      "learning_rate": 1.6887777974950296e-05,
+      "loss": 0.9099,
       "step": 9882
     },
     {
-      "epoch": 0.27142896377468345,
+      "epoch": 0.2804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.709586893433514e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.688711163952294e-05,
+      "loss": 0.9499,
       "step": 9883
     },
     {
-      "epoch": 0.27145642800252673,
+      "epoch": 0.2804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.7095242137323326e-05,
-      "loss": 0.9661,
+      "learning_rate": 1.688644524591984e-05,
+      "loss": 0.9242,
       "step": 9884
     },
     {
-      "epoch": 0.27148389223036995,
+      "epoch": 0.2805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.7094615284171207e-05,
-      "loss": 1.004,
+      "learning_rate": 1.688577879414663e-05,
+      "loss": 1.0291,
       "step": 9885
     },
     {
-      "epoch": 0.27151135645821317,
+      "epoch": 0.28053348467650396,
       "grad_norm": 0.0,
-      "learning_rate": 1.7093988374883745e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.6885112284208934e-05,
+      "loss": 1.0049,
       "step": 9886
     },
     {
-      "epoch": 0.2715388206860564,
+      "epoch": 0.2805618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.7093361409465896e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.6884445716112388e-05,
+      "loss": 1.0122,
       "step": 9887
     },
     {
-      "epoch": 0.27156628491389967,
+      "epoch": 0.28059023836549374,
       "grad_norm": 0.0,
-      "learning_rate": 1.7092734387922627e-05,
-      "loss": 0.8846,
+      "learning_rate": 1.6883779089862618e-05,
+      "loss": 0.9954,
       "step": 9888
     },
     {
-      "epoch": 0.2715937491417429,
+      "epoch": 0.28061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.7092107310258893e-05,
-      "loss": 1.1412,
+      "learning_rate": 1.6883112405465257e-05,
+      "loss": 1.0063,
       "step": 9889
     },
     {
-      "epoch": 0.2716212133695861,
+      "epoch": 0.28064699205448357,
       "grad_norm": 0.0,
-      "learning_rate": 1.7091480176479663e-05,
-      "loss": 0.9281,
+      "learning_rate": 1.6882445662925934e-05,
+      "loss": 0.9678,
       "step": 9890
     },
     {
-      "epoch": 0.27164867759742933,
+      "epoch": 0.28067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.709085298658989e-05,
-      "loss": 0.8715,
+      "learning_rate": 1.6881778862250285e-05,
+      "loss": 0.9465,
       "step": 9891
     },
     {
-      "epoch": 0.2716761418252726,
+      "epoch": 0.28070374574347334,
       "grad_norm": 0.0,
-      "learning_rate": 1.7090225740594543e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.6881112003443944e-05,
+      "loss": 0.9783,
       "step": 9892
     },
     {
-      "epoch": 0.27170360605311583,
+      "epoch": 0.2807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.7089598438498583e-05,
-      "loss": 0.8914,
+      "learning_rate": 1.6880445086512535e-05,
+      "loss": 1.0703,
       "step": 9893
     },
     {
-      "epoch": 0.27173107028095905,
+      "epoch": 0.2807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.7088971080306976e-05,
-      "loss": 1.0854,
+      "learning_rate": 1.68797781114617e-05,
+      "loss": 1.0677,
       "step": 9894
     },
     {
-      "epoch": 0.2717585345088023,
+      "epoch": 0.280788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.7088343666024684e-05,
-      "loss": 0.936,
+      "learning_rate": 1.687911107829707e-05,
+      "loss": 0.9608,
       "step": 9895
     },
     {
-      "epoch": 0.2717859987366455,
+      "epoch": 0.2808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.708771619565667e-05,
-      "loss": 0.9269,
+      "learning_rate": 1.6878443987024276e-05,
+      "loss": 1.027,
       "step": 9896
     },
     {
-      "epoch": 0.2718134629644888,
+      "epoch": 0.2808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.7087088669207902e-05,
-      "loss": 1.0658,
+      "learning_rate": 1.687777683764896e-05,
+      "loss": 0.9608,
       "step": 9897
     },
     {
-      "epoch": 0.271840927192332,
+      "epoch": 0.28087400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.708646108668334e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.6877109630176757e-05,
+      "loss": 1.0435,
       "step": 9898
     },
     {
-      "epoch": 0.2718683914201752,
+      "epoch": 0.2809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.7085833448087953e-05,
-      "loss": 1.0385,
+      "learning_rate": 1.6876442364613292e-05,
+      "loss": 1.0796,
       "step": 9899
     },
     {
-      "epoch": 0.27189585564801844,
+      "epoch": 0.28093076049943244,
       "grad_norm": 0.0,
-      "learning_rate": 1.7085205753426707e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.6875775040964214e-05,
+      "loss": 0.9976,
       "step": 9900
     },
     {
-      "epoch": 0.2719233198758617,
+      "epoch": 0.28095913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.7084578002704568e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.6875107659235155e-05,
+      "loss": 0.9783,
       "step": 9901
     },
     {
-      "epoch": 0.27195078410370493,
+      "epoch": 0.28098751418842227,
       "grad_norm": 0.0,
-      "learning_rate": 1.7083950195926508e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.687444021943175e-05,
+      "loss": 0.9782,
       "step": 9902
     },
     {
-      "epoch": 0.27197824833154816,
+      "epoch": 0.28101589103291713,
       "grad_norm": 0.0,
-      "learning_rate": 1.7083322333097485e-05,
-      "loss": 0.9162,
+      "learning_rate": 1.6873772721559644e-05,
+      "loss": 0.8469,
       "step": 9903
     },
     {
-      "epoch": 0.2720057125593914,
+      "epoch": 0.28104426787741205,
       "grad_norm": 0.0,
-      "learning_rate": 1.7082694414222474e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.6873105165624465e-05,
+      "loss": 0.9931,
       "step": 9904
     },
     {
-      "epoch": 0.27203317678723465,
+      "epoch": 0.2810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.7082066439306442e-05,
-      "loss": 1.1289,
+      "learning_rate": 1.6872437551631863e-05,
+      "loss": 0.8819,
       "step": 9905
     },
     {
-      "epoch": 0.2720606410150779,
+      "epoch": 0.2811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.7081438408354356e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.687176987958747e-05,
+      "loss": 0.8802,
       "step": 9906
     },
     {
-      "epoch": 0.2720881052429211,
+      "epoch": 0.28112939841089674,
       "grad_norm": 0.0,
-      "learning_rate": 1.7080810321371183e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.6871102149496925e-05,
+      "loss": 0.8851,
       "step": 9907
     },
     {
-      "epoch": 0.2721155694707643,
+      "epoch": 0.2811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.7080182178361903e-05,
-      "loss": 1.0173,
+      "learning_rate": 1.6870434361365875e-05,
+      "loss": 0.9837,
       "step": 9908
     },
     {
-      "epoch": 0.27214303369860754,
+      "epoch": 0.2811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.7079553979331474e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.6869766515199956e-05,
+      "loss": 0.924,
       "step": 9909
     },
     {
-      "epoch": 0.2721704979264508,
+      "epoch": 0.28121452894438137,
       "grad_norm": 0.0,
-      "learning_rate": 1.7078925724284875e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.686909861100481e-05,
+      "loss": 0.8929,
       "step": 9910
     },
     {
-      "epoch": 0.27219796215429404,
+      "epoch": 0.2812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.7078297413227067e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.686843064878608e-05,
+      "loss": 0.8789,
       "step": 9911
     },
     {
-      "epoch": 0.27222542638213726,
+      "epoch": 0.28127128263337114,
       "grad_norm": 0.0,
-      "learning_rate": 1.7077669046163034e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.686776262854941e-05,
+      "loss": 0.9624,
       "step": 9912
     },
     {
-      "epoch": 0.2722528906099805,
+      "epoch": 0.28129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.707704062309774e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.6867094550300436e-05,
+      "loss": 0.9424,
       "step": 9913
     },
     {
-      "epoch": 0.27228035483782376,
+      "epoch": 0.281328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.707641214403616e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.686642641404481e-05,
+      "loss": 0.966,
       "step": 9914
     },
     {
-      "epoch": 0.272307819065667,
+      "epoch": 0.28135641316685583,
       "grad_norm": 0.0,
-      "learning_rate": 1.7075783608983265e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.6865758219788166e-05,
+      "loss": 0.8988,
       "step": 9915
     },
     {
-      "epoch": 0.2723352832935102,
+      "epoch": 0.28138479001135075,
       "grad_norm": 0.0,
-      "learning_rate": 1.707515501794403e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.6865089967536156e-05,
+      "loss": 0.9128,
       "step": 9916
     },
     {
-      "epoch": 0.2723627475213534,
+      "epoch": 0.2814131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.7074526370923424e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.6864421657294424e-05,
+      "loss": 1.076,
       "step": 9917
     },
     {
-      "epoch": 0.2723902117491967,
+      "epoch": 0.2814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.7073897667926432e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.6863753289068612e-05,
+      "loss": 0.9047,
       "step": 9918
     },
     {
-      "epoch": 0.2724176759770399,
+      "epoch": 0.28146992054483544,
       "grad_norm": 0.0,
-      "learning_rate": 1.707326890895802e-05,
-      "loss": 0.8721,
+      "learning_rate": 1.686308486286437e-05,
+      "loss": 0.9813,
       "step": 9919
     },
     {
-      "epoch": 0.27244514020488314,
+      "epoch": 0.2814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.7072640094023157e-05,
-      "loss": 1.0496,
+      "learning_rate": 1.686241637868734e-05,
+      "loss": 0.9482,
       "step": 9920
     },
     {
-      "epoch": 0.27247260443272636,
+      "epoch": 0.2815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.707201122312683e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.686174783654317e-05,
+      "loss": 1.01,
       "step": 9921
     },
     {
-      "epoch": 0.2725000686605696,
+      "epoch": 0.2815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.7071382296274017e-05,
-      "loss": 1.0448,
+      "learning_rate": 1.6861079236437503e-05,
+      "loss": 0.9663,
       "step": 9922
     },
     {
-      "epoch": 0.27252753288841286,
+      "epoch": 0.281583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.7070753313469683e-05,
-      "loss": 0.875,
+      "learning_rate": 1.6860410578375997e-05,
+      "loss": 0.9584,
       "step": 9923
     },
     {
-      "epoch": 0.2725549971162561,
+      "epoch": 0.28161180476730985,
       "grad_norm": 0.0,
-      "learning_rate": 1.707012427471881e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.6859741862364294e-05,
+      "loss": 0.9766,
       "step": 9924
     },
     {
-      "epoch": 0.2725824613440993,
+      "epoch": 0.28164018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 1.7069495180026373e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.6859073088408043e-05,
+      "loss": 0.9721,
       "step": 9925
     },
     {
-      "epoch": 0.2726099255719425,
+      "epoch": 0.2816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.7068866029397355e-05,
-      "loss": 0.9676,
+      "learning_rate": 1.685840425651289e-05,
+      "loss": 0.9644,
       "step": 9926
     },
     {
-      "epoch": 0.2726373897997858,
+      "epoch": 0.28169693530079454,
       "grad_norm": 0.0,
-      "learning_rate": 1.7068236822836728e-05,
-      "loss": 0.9439,
+      "learning_rate": 1.6857735366684492e-05,
+      "loss": 0.8982,
       "step": 9927
     },
     {
-      "epoch": 0.272664854027629,
+      "epoch": 0.28172531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.7067607560349474e-05,
-      "loss": 0.8618,
+      "learning_rate": 1.6857066418928495e-05,
+      "loss": 0.9355,
       "step": 9928
     },
     {
-      "epoch": 0.27269231825547224,
+      "epoch": 0.2817536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.7066978241940573e-05,
-      "loss": 0.8693,
+      "learning_rate": 1.6856397413250548e-05,
+      "loss": 1.0898,
       "step": 9929
     },
     {
-      "epoch": 0.27271978248331546,
+      "epoch": 0.2817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.7066348867615002e-05,
-      "loss": 1.0477,
+      "learning_rate": 1.68557283496563e-05,
+      "loss": 1.0086,
       "step": 9930
     },
     {
-      "epoch": 0.27274724671115874,
+      "epoch": 0.28181044267877414,
       "grad_norm": 0.0,
-      "learning_rate": 1.706571943737774e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.6855059228151413e-05,
+      "loss": 0.9656,
       "step": 9931
     },
     {
-      "epoch": 0.27277471093900196,
+      "epoch": 0.281838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.7065089951233772e-05,
-      "loss": 0.964,
+      "learning_rate": 1.685439004874153e-05,
+      "loss": 0.9312,
       "step": 9932
     },
     {
-      "epoch": 0.2728021751668452,
+      "epoch": 0.2818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.7064460409188074e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.6853720811432307e-05,
+      "loss": 0.8835,
       "step": 9933
     },
     {
-      "epoch": 0.2728296393946884,
+      "epoch": 0.2818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.7063830811245625e-05,
-      "loss": 0.8785,
+      "learning_rate": 1.68530515162294e-05,
+      "loss": 0.8972,
       "step": 9934
     },
     {
-      "epoch": 0.2728571036225316,
+      "epoch": 0.2819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.7063201157411413e-05,
-      "loss": 0.9306,
+      "learning_rate": 1.6852382163138453e-05,
+      "loss": 0.9809,
       "step": 9935
     },
     {
-      "epoch": 0.2728845678503749,
+      "epoch": 0.2819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.706257144769042e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.6851712752165124e-05,
+      "loss": 0.9589,
       "step": 9936
     },
     {
-      "epoch": 0.2729120320782181,
+      "epoch": 0.28198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.706194168208762e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.6851043283315075e-05,
+      "loss": 0.8887,
       "step": 9937
     },
     {
-      "epoch": 0.27293949630606135,
+      "epoch": 0.2820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.7061311860608008e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.6850373756593952e-05,
+      "loss": 0.952,
       "step": 9938
     },
     {
-      "epoch": 0.27296696053390457,
+      "epoch": 0.28203745743473324,
       "grad_norm": 0.0,
-      "learning_rate": 1.706068198325656e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.6849704172007414e-05,
+      "loss": 0.9364,
       "step": 9939
     },
     {
-      "epoch": 0.27299442476174784,
+      "epoch": 0.28206583427922816,
       "grad_norm": 0.0,
-      "learning_rate": 1.706005205003826e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.6849034529561116e-05,
+      "loss": 0.8024,
       "step": 9940
     },
     {
-      "epoch": 0.27302188898959107,
+      "epoch": 0.282094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.7059422060958093e-05,
-      "loss": 0.963,
+      "learning_rate": 1.684836482926072e-05,
+      "loss": 0.9613,
       "step": 9941
     },
     {
-      "epoch": 0.2730493532174343,
+      "epoch": 0.28212258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.7058792016021044e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.6847695071111876e-05,
+      "loss": 1.0187,
       "step": 9942
     },
     {
-      "epoch": 0.2730768174452775,
+      "epoch": 0.28215096481271285,
       "grad_norm": 0.0,
-      "learning_rate": 1.70581619152321e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.6847025255120244e-05,
+      "loss": 0.9966,
       "step": 9943
     },
     {
-      "epoch": 0.2731042816731208,
+      "epoch": 0.2821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.7057531758596245e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.684635538129148e-05,
+      "loss": 0.9998,
       "step": 9944
     },
     {
-      "epoch": 0.273131745900964,
+      "epoch": 0.2822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.7056901546118464e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.6845685449631247e-05,
+      "loss": 1.0488,
       "step": 9945
     },
     {
-      "epoch": 0.27315921012880723,
+      "epoch": 0.2822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.7056271277803742e-05,
-      "loss": 0.8561,
+      "learning_rate": 1.68450154601452e-05,
+      "loss": 0.9566,
       "step": 9946
     },
     {
-      "epoch": 0.27318667435665045,
+      "epoch": 0.2822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.7055640953657068e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.6844345412838997e-05,
+      "loss": 0.995,
       "step": 9947
     },
     {
-      "epoch": 0.27321413858449367,
+      "epoch": 0.2822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.7055010573683434e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.6843675307718306e-05,
+      "loss": 0.935,
       "step": 9948
     },
     {
-      "epoch": 0.27324160281233695,
+      "epoch": 0.28232122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 1.7054380137887822e-05,
-      "loss": 0.946,
+      "learning_rate": 1.6843005144788778e-05,
+      "loss": 0.9924,
       "step": 9949
     },
     {
-      "epoch": 0.27326906704018017,
+      "epoch": 0.2823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.705374964627522e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.6842334924056082e-05,
+      "loss": 0.9381,
       "step": 9950
     },
     {
-      "epoch": 0.2732965312680234,
+      "epoch": 0.28237797956867194,
       "grad_norm": 0.0,
-      "learning_rate": 1.705311909885062e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.684166464552587e-05,
+      "loss": 0.8763,
       "step": 9951
     },
     {
-      "epoch": 0.2733239954958666,
+      "epoch": 0.28240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.705248849561901e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.6840994309203815e-05,
+      "loss": 0.8757,
       "step": 9952
     },
     {
-      "epoch": 0.2733514597237099,
+      "epoch": 0.2824347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.7051857836585376e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.6840323915095566e-05,
+      "loss": 0.9778,
       "step": 9953
     },
     {
-      "epoch": 0.2733789239515531,
+      "epoch": 0.28246311010215663,
       "grad_norm": 0.0,
-      "learning_rate": 1.7051227121754716e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.68396534632068e-05,
+      "loss": 0.9757,
       "step": 9954
     },
     {
-      "epoch": 0.27340638817939633,
+      "epoch": 0.28249148694665155,
       "grad_norm": 0.0,
-      "learning_rate": 1.705059635113201e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.683898295354317e-05,
+      "loss": 0.9632,
       "step": 9955
     },
     {
-      "epoch": 0.27343385240723955,
+      "epoch": 0.2825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.704996552472226e-05,
-      "loss": 0.9349,
+      "learning_rate": 1.6838312386110347e-05,
+      "loss": 0.881,
       "step": 9956
     },
     {
-      "epoch": 0.27346131663508283,
+      "epoch": 0.2825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.704933464253045e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.683764176091399e-05,
+      "loss": 0.958,
       "step": 9957
     },
     {
-      "epoch": 0.27348878086292605,
+      "epoch": 0.2825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.704870370456157e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.6836971077959766e-05,
+      "loss": 0.9832,
       "step": 9958
     },
     {
-      "epoch": 0.27351624509076927,
+      "epoch": 0.2826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.704807271082062e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.6836300337253337e-05,
+      "loss": 1.0065,
       "step": 9959
     },
     {
-      "epoch": 0.2735437093186125,
+      "epoch": 0.282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.704744166131259e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.6835629538800376e-05,
+      "loss": 0.9143,
       "step": 9960
     },
     {
-      "epoch": 0.2735711735464557,
+      "epoch": 0.2826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.7046810556042467e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.6834958682606545e-05,
+      "loss": 1.0318,
       "step": 9961
     },
     {
-      "epoch": 0.273598637774299,
+      "epoch": 0.2826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.7046179395015254e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.6834287768677507e-05,
+      "loss": 0.9108,
       "step": 9962
     },
     {
-      "epoch": 0.2736261020021422,
+      "epoch": 0.28271850170261065,
       "grad_norm": 0.0,
-      "learning_rate": 1.7045548178235938e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.6833616797018936e-05,
+      "loss": 0.9397,
       "step": 9963
     },
     {
-      "epoch": 0.27365356622998543,
+      "epoch": 0.28274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.7044916905709512e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.6832945767636496e-05,
+      "loss": 0.8994,
       "step": 9964
     },
     {
-      "epoch": 0.27368103045782866,
+      "epoch": 0.2827752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.704428557744098e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.6832274680535855e-05,
+      "loss": 1.054,
       "step": 9965
     },
     {
-      "epoch": 0.27370849468567193,
+      "epoch": 0.28280363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.704365419343533e-05,
-      "loss": 0.978,
+      "learning_rate": 1.6831603535722682e-05,
+      "loss": 0.9495,
       "step": 9966
     },
     {
-      "epoch": 0.27373595891351515,
+      "epoch": 0.28283200908059025,
       "grad_norm": 0.0,
-      "learning_rate": 1.704302275369756e-05,
-      "loss": 1.063,
+      "learning_rate": 1.683093233320265e-05,
+      "loss": 0.9698,
       "step": 9967
     },
     {
-      "epoch": 0.2737634231413584,
+      "epoch": 0.2828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.704239125823266e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.6830261072981423e-05,
+      "loss": 0.9318,
       "step": 9968
     },
     {
-      "epoch": 0.2737908873692016,
+      "epoch": 0.28288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.7041759707045643e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.6829589755064673e-05,
+      "loss": 0.9869,
       "step": 9969
     },
     {
-      "epoch": 0.2738183515970449,
+      "epoch": 0.28291713961407494,
       "grad_norm": 0.0,
-      "learning_rate": 1.7041128100141485e-05,
-      "loss": 1.0381,
+      "learning_rate": 1.6828918379458072e-05,
+      "loss": 0.9861,
       "step": 9970
     },
     {
-      "epoch": 0.2738458158248881,
+      "epoch": 0.2829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.70404964375252e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.6828246946167293e-05,
+      "loss": 0.8488,
       "step": 9971
     },
     {
-      "epoch": 0.2738732800527313,
+      "epoch": 0.2829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.703986471920178e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.6827575455198002e-05,
+      "loss": 0.8104,
       "step": 9972
     },
     {
-      "epoch": 0.27390074428057454,
+      "epoch": 0.2830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.703923294517622e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.6826903906555875e-05,
+      "loss": 0.9307,
       "step": 9973
     },
     {
-      "epoch": 0.27392820850841776,
+      "epoch": 0.2830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.7038601115453527e-05,
-      "loss": 0.883,
+      "learning_rate": 1.6826232300246585e-05,
+      "loss": 0.8825,
       "step": 9974
     },
     {
-      "epoch": 0.27395567273626104,
+      "epoch": 0.28305902383654935,
       "grad_norm": 0.0,
-      "learning_rate": 1.7037969230038692e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.6825560636275806e-05,
+      "loss": 0.903,
       "step": 9975
     },
     {
-      "epoch": 0.27398313696410426,
+      "epoch": 0.28308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.703733728893672e-05,
-      "loss": 1.1035,
+      "learning_rate": 1.6824888914649205e-05,
+      "loss": 0.9882,
       "step": 9976
     },
     {
-      "epoch": 0.2740106011919475,
+      "epoch": 0.2831157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.7036705292152607e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.682421713537246e-05,
+      "loss": 0.9249,
       "step": 9977
     },
     {
-      "epoch": 0.2740380654197907,
+      "epoch": 0.28314415437003404,
       "grad_norm": 0.0,
-      "learning_rate": 1.703607323969136e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.6823545298451248e-05,
+      "loss": 0.9248,
       "step": 9978
     },
     {
-      "epoch": 0.274065529647634,
+      "epoch": 0.28317253121452896,
       "grad_norm": 0.0,
-      "learning_rate": 1.7035441131557974e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.6822873403891246e-05,
+      "loss": 0.8915,
       "step": 9979
     },
     {
-      "epoch": 0.2740929938754772,
+      "epoch": 0.2832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.7034808967757446e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.6822201451698123e-05,
+      "loss": 0.9709,
       "step": 9980
     },
     {
-      "epoch": 0.2741204581033204,
+      "epoch": 0.28322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.7034176748294796e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.6821529441877557e-05,
+      "loss": 0.9589,
       "step": 9981
     },
     {
-      "epoch": 0.27414792233116364,
+      "epoch": 0.28325766174801364,
       "grad_norm": 0.0,
-      "learning_rate": 1.7033544473175006e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.6820857374435223e-05,
+      "loss": 0.9381,
       "step": 9982
     },
     {
-      "epoch": 0.2741753865590069,
+      "epoch": 0.2832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.703291214240309e-05,
-      "loss": 1.0486,
+      "learning_rate": 1.6820185249376804e-05,
+      "loss": 0.9118,
       "step": 9983
     },
     {
-      "epoch": 0.27420285078685014,
+      "epoch": 0.2833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.703227975598405e-05,
-      "loss": 0.9172,
+      "learning_rate": 1.6819513066707967e-05,
+      "loss": 0.9688,
       "step": 9984
     },
     {
-      "epoch": 0.27423031501469336,
+      "epoch": 0.2833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.703164731392289e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.6818840826434405e-05,
+      "loss": 1.0194,
       "step": 9985
     },
     {
-      "epoch": 0.2742577792425366,
+      "epoch": 0.2833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.7031014816224612e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.6818168528561783e-05,
+      "loss": 1.0255,
       "step": 9986
     },
     {
-      "epoch": 0.2742852434703798,
+      "epoch": 0.2833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.703038226289422e-05,
-      "loss": 0.8781,
+      "learning_rate": 1.6817496173095786e-05,
+      "loss": 0.895,
       "step": 9987
     },
     {
-      "epoch": 0.2743127076982231,
+      "epoch": 0.28342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.7029749653936718e-05,
-      "loss": 0.8461,
+      "learning_rate": 1.6816823760042092e-05,
+      "loss": 0.9354,
       "step": 9988
     },
     {
-      "epoch": 0.2743401719260663,
+      "epoch": 0.2834562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.7029116989357115e-05,
-      "loss": 1.0524,
+      "learning_rate": 1.681615128940638e-05,
+      "loss": 0.8889,
       "step": 9989
     },
     {
-      "epoch": 0.2743676361539095,
+      "epoch": 0.28348467650397274,
       "grad_norm": 0.0,
-      "learning_rate": 1.7028484269160417e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.681547876119434e-05,
+      "loss": 1.0103,
       "step": 9990
     },
     {
-      "epoch": 0.27439510038175274,
+      "epoch": 0.28351305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.702785149335163e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.6814806175411633e-05,
+      "loss": 1.0701,
       "step": 9991
     },
     {
-      "epoch": 0.274422564609596,
+      "epoch": 0.2835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.7027218661935755e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.6814133532063955e-05,
+      "loss": 1.0123,
       "step": 9992
     },
     {
-      "epoch": 0.27445002883743924,
+      "epoch": 0.28356980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.702658577491781e-05,
-      "loss": 0.9122,
+      "learning_rate": 1.681346083115699e-05,
+      "loss": 1.0108,
       "step": 9993
     },
     {
-      "epoch": 0.27447749306528246,
+      "epoch": 0.28359818388195235,
       "grad_norm": 0.0,
-      "learning_rate": 1.702595283230279e-05,
-      "loss": 0.8708,
+      "learning_rate": 1.681278807269641e-05,
+      "loss": 0.893,
       "step": 9994
     },
     {
-      "epoch": 0.2745049572931257,
+      "epoch": 0.2836265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.7025319834095715e-05,
-      "loss": 0.986,
+      "learning_rate": 1.6812115256687906e-05,
+      "loss": 0.9708,
       "step": 9995
     },
     {
-      "epoch": 0.27453242152096896,
+      "epoch": 0.2836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.7024686780301586e-05,
-      "loss": 1.0011,
+      "learning_rate": 1.6811442383137154e-05,
+      "loss": 1.0083,
       "step": 9996
     },
     {
-      "epoch": 0.2745598857488122,
+      "epoch": 0.283683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.7024053670925412e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.6810769452049846e-05,
+      "loss": 0.9977,
       "step": 9997
     },
     {
-      "epoch": 0.2745873499766554,
+      "epoch": 0.2837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.702342050597221e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.681009646343166e-05,
+      "loss": 0.9319,
       "step": 9998
     },
     {
-      "epoch": 0.2746148142044986,
+      "epoch": 0.2837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.7022787285446976e-05,
-      "loss": 1.0076,
+      "learning_rate": 1.6809423417288283e-05,
+      "loss": 0.8995,
       "step": 9999
     },
     {
-      "epoch": 0.2746422784323419,
+      "epoch": 0.28376844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.7022154009354735e-05,
-      "loss": 0.976,
+      "learning_rate": 1.68087503136254e-05,
+      "loss": 0.8365,
       "step": 10000
     },
     {
-      "epoch": 0.2746697426601851,
+      "epoch": 0.2837968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.702152067770049e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.68080771524487e-05,
+      "loss": 0.9613,
       "step": 10001
     },
     {
-      "epoch": 0.27469720688802834,
+      "epoch": 0.28382519863791145,
       "grad_norm": 0.0,
-      "learning_rate": 1.7020887290489257e-05,
-      "loss": 1.0676,
+      "learning_rate": 1.6807403933763865e-05,
+      "loss": 0.9199,
       "step": 10002
     },
     {
-      "epoch": 0.27472467111587157,
+      "epoch": 0.28385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.702025384772604e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.6806730657576585e-05,
+      "loss": 1.0168,
       "step": 10003
     },
     {
-      "epoch": 0.2747521353437148,
+      "epoch": 0.2838819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.7019620349415856e-05,
-      "loss": 1.033,
+      "learning_rate": 1.6806057323892542e-05,
+      "loss": 0.8998,
       "step": 10004
     },
     {
-      "epoch": 0.27477959957155806,
+      "epoch": 0.28391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7018986795563718e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.6805383932717428e-05,
+      "loss": 1.0095,
       "step": 10005
     },
     {
-      "epoch": 0.2748070637994013,
+      "epoch": 0.28393870601589105,
       "grad_norm": 0.0,
-      "learning_rate": 1.701835318617463e-05,
-      "loss": 1.0669,
+      "learning_rate": 1.6804710484056926e-05,
+      "loss": 0.9357,
       "step": 10006
     },
     {
-      "epoch": 0.2748345280272445,
+      "epoch": 0.2839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.7017719521253623e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.6804036977916735e-05,
+      "loss": 0.921,
       "step": 10007
     },
     {
-      "epoch": 0.27486199225508773,
+      "epoch": 0.2839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.7017085800805696e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.6803363414302535e-05,
+      "loss": 1.0444,
       "step": 10008
     },
     {
-      "epoch": 0.274889456482931,
+      "epoch": 0.2840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.7016452024835867e-05,
-      "loss": 0.9283,
+      "learning_rate": 1.6802689793220017e-05,
+      "loss": 1.0453,
       "step": 10009
     },
     {
-      "epoch": 0.2749169207107742,
+      "epoch": 0.2840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.7015818193349155e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.6802016114674874e-05,
+      "loss": 0.943,
       "step": 10010
     },
     {
-      "epoch": 0.27494438493861745,
+      "epoch": 0.2840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.701518430635057e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.6801342378672797e-05,
+      "loss": 1.1157,
       "step": 10011
     },
     {
-      "epoch": 0.27497184916646067,
+      "epoch": 0.2841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.701455036384513e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.6800668585219474e-05,
+      "loss": 0.8655,
       "step": 10012
     },
     {
-      "epoch": 0.27499931339430395,
+      "epoch": 0.2841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.701391636583785e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.6799994734320596e-05,
+      "loss": 0.8215,
       "step": 10013
     },
     {
-      "epoch": 0.27502677762214717,
+      "epoch": 0.28416572077185015,
       "grad_norm": 0.0,
-      "learning_rate": 1.7013282312333746e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.6799320825981856e-05,
+      "loss": 1.0405,
       "step": 10014
     },
     {
-      "epoch": 0.2750542418499904,
+      "epoch": 0.28419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.7012648203337833e-05,
-      "loss": 1.078,
+      "learning_rate": 1.679864686020895e-05,
+      "loss": 1.0312,
       "step": 10015
     },
     {
-      "epoch": 0.2750817060778336,
+      "epoch": 0.28422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.7012014038855137e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.6797972837007568e-05,
+      "loss": 0.9077,
       "step": 10016
     },
     {
-      "epoch": 0.27510917030567683,
+      "epoch": 0.28425085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.7011379818890668e-05,
-      "loss": 0.94,
+      "learning_rate": 1.6797298756383404e-05,
+      "loss": 1.0053,
       "step": 10017
     },
     {
-      "epoch": 0.2751366345335201,
+      "epoch": 0.28427922814982975,
       "grad_norm": 0.0,
-      "learning_rate": 1.7010745543449443e-05,
-      "loss": 1.024,
+      "learning_rate": 1.679662461834215e-05,
+      "loss": 0.9331,
       "step": 10018
     },
     {
-      "epoch": 0.27516409876136333,
+      "epoch": 0.2843076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.7010111212536486e-05,
-      "loss": 1.0411,
+      "learning_rate": 1.6795950422889503e-05,
+      "loss": 0.8827,
       "step": 10019
     },
     {
-      "epoch": 0.27519156298920655,
+      "epoch": 0.28433598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 1.7009476826156815e-05,
-      "loss": 0.972,
+      "learning_rate": 1.6795276170031157e-05,
+      "loss": 0.9426,
       "step": 10020
     },
     {
-      "epoch": 0.27521902721704977,
+      "epoch": 0.2843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.7008842384315444e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.679460185977281e-05,
+      "loss": 1.0273,
       "step": 10021
     },
     {
-      "epoch": 0.27524649144489305,
+      "epoch": 0.2843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.7008207887017402e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.679392749212015e-05,
+      "loss": 0.8826,
       "step": 10022
     },
     {
-      "epoch": 0.27527395567273627,
+      "epoch": 0.2844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.70075733342677e-05,
-      "loss": 0.7652,
+      "learning_rate": 1.6793253067078886e-05,
+      "loss": 1.0435,
       "step": 10023
     },
     {
-      "epoch": 0.2753014199005795,
+      "epoch": 0.2844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.7006938726071366e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.6792578584654706e-05,
+      "loss": 0.9376,
       "step": 10024
     },
     {
-      "epoch": 0.2753288841284227,
+      "epoch": 0.284477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.7006304062433417e-05,
-      "loss": 0.9549,
+      "learning_rate": 1.679190404485331e-05,
+      "loss": 0.9648,
       "step": 10025
     },
     {
-      "epoch": 0.275356348356266,
+      "epoch": 0.28450624290578885,
       "grad_norm": 0.0,
-      "learning_rate": 1.7005669343358877e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.679122944768039e-05,
+      "loss": 0.9939,
       "step": 10026
     },
     {
-      "epoch": 0.2753838125841092,
+      "epoch": 0.28453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.700503456885277e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.6790554793141654e-05,
+      "loss": 0.9768,
       "step": 10027
     },
     {
-      "epoch": 0.27541127681195243,
+      "epoch": 0.2845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.700439973892011e-05,
-      "loss": 0.9855,
+      "learning_rate": 1.6789880081242794e-05,
+      "loss": 0.93,
       "step": 10028
     },
     {
-      "epoch": 0.27543874103979565,
+      "epoch": 0.28459137343927354,
       "grad_norm": 0.0,
-      "learning_rate": 1.700376485356593e-05,
-      "loss": 1.0056,
+      "learning_rate": 1.6789205311989518e-05,
+      "loss": 0.9234,
       "step": 10029
     },
     {
-      "epoch": 0.2754662052676389,
+      "epoch": 0.28461975028376846,
       "grad_norm": 0.0,
-      "learning_rate": 1.7003129912795246e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.6788530485387513e-05,
+      "loss": 0.9963,
       "step": 10030
     },
     {
-      "epoch": 0.27549366949548215,
+      "epoch": 0.2846481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.700249491661309e-05,
-      "loss": 0.986,
+      "learning_rate": 1.678785560144249e-05,
+      "loss": 0.9396,
       "step": 10031
     },
     {
-      "epoch": 0.2755211337233254,
+      "epoch": 0.28467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.700185986502448e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.678718066016014e-05,
+      "loss": 1.0046,
       "step": 10032
     },
     {
-      "epoch": 0.2755485979511686,
+      "epoch": 0.28470488081725315,
       "grad_norm": 0.0,
-      "learning_rate": 1.700122475803444e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.6786505661546172e-05,
+      "loss": 0.9563,
       "step": 10033
     },
     {
-      "epoch": 0.2755760621790118,
+      "epoch": 0.284733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.7000589595648006e-05,
-      "loss": 0.931,
+      "learning_rate": 1.6785830605606287e-05,
+      "loss": 1.0193,
       "step": 10034
     },
     {
-      "epoch": 0.2756035264068551,
+      "epoch": 0.2847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.6999954377870187e-05,
-      "loss": 0.9969,
+      "learning_rate": 1.6785155492346188e-05,
+      "loss": 0.8937,
       "step": 10035
     },
     {
-      "epoch": 0.2756309906346983,
+      "epoch": 0.2847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.699931910470602e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.6784480321771574e-05,
+      "loss": 0.9768,
       "step": 10036
     },
     {
-      "epoch": 0.27565845486254154,
+      "epoch": 0.2848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.699868377616053e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.6783805093888144e-05,
+      "loss": 1.0139,
       "step": 10037
     },
     {
-      "epoch": 0.27568591909038476,
+      "epoch": 0.28484676503972756,
       "grad_norm": 0.0,
-      "learning_rate": 1.699804839223874e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.6783129808701618e-05,
+      "loss": 0.9794,
       "step": 10038
     },
     {
-      "epoch": 0.27571338331822803,
+      "epoch": 0.28487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.699741295294568e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.6782454466217683e-05,
+      "loss": 0.9445,
       "step": 10039
     },
     {
-      "epoch": 0.27574084754607125,
+      "epoch": 0.2849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.6996777458286383e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.678177906644205e-05,
+      "loss": 1.0317,
       "step": 10040
     },
     {
-      "epoch": 0.2757683117739145,
+      "epoch": 0.28493189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 1.699614190826587e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.6781103609380425e-05,
+      "loss": 1.0329,
       "step": 10041
     },
     {
-      "epoch": 0.2757957760017577,
+      "epoch": 0.28496027241770716,
       "grad_norm": 0.0,
-      "learning_rate": 1.699550630288917e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.6780428095038512e-05,
+      "loss": 0.9013,
       "step": 10042
     },
     {
-      "epoch": 0.2758232402296009,
+      "epoch": 0.284988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.6994870642161315e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.6779752523422025e-05,
+      "loss": 0.9462,
       "step": 10043
     },
     {
-      "epoch": 0.2758507044574442,
+      "epoch": 0.28501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.699423492608734e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.6779076894536657e-05,
+      "loss": 0.9559,
       "step": 10044
     },
     {
-      "epoch": 0.2758781686852874,
+      "epoch": 0.28504540295119185,
       "grad_norm": 0.0,
-      "learning_rate": 1.6993599154672265e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.677840120838812e-05,
+      "loss": 0.8673,
       "step": 10045
     },
     {
-      "epoch": 0.27590563291313064,
+      "epoch": 0.2850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.699296332792112e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.6777725464982127e-05,
+      "loss": 0.9176,
       "step": 10046
     },
     {
-      "epoch": 0.27593309714097386,
+      "epoch": 0.2851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.6992327445838947e-05,
-      "loss": 0.8752,
+      "learning_rate": 1.677704966432438e-05,
+      "loss": 0.9131,
       "step": 10047
     },
     {
-      "epoch": 0.27596056136881714,
+      "epoch": 0.2851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.6991691508430765e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.677637380642059e-05,
+      "loss": 1.0452,
       "step": 10048
     },
     {
-      "epoch": 0.27598802559666036,
+      "epoch": 0.2851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.6991055515701615e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.677569789127647e-05,
+      "loss": 0.9289,
       "step": 10049
     },
     {
-      "epoch": 0.2760154898245036,
+      "epoch": 0.2851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.6990419467656524e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.677502191889772e-05,
+      "loss": 0.9775,
       "step": 10050
     },
     {
-      "epoch": 0.2760429540523468,
+      "epoch": 0.2852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.6989783364300527e-05,
-      "loss": 1.0802,
+      "learning_rate": 1.6774345889290062e-05,
+      "loss": 0.8712,
       "step": 10051
     },
     {
-      "epoch": 0.2760704182801901,
+      "epoch": 0.2852440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.698914720563866e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.6773669802459194e-05,
+      "loss": 0.8479,
       "step": 10052
     },
     {
-      "epoch": 0.2760978825080333,
+      "epoch": 0.28527241770715095,
       "grad_norm": 0.0,
-      "learning_rate": 1.6988510991675945e-05,
-      "loss": 1.0664,
+      "learning_rate": 1.6772993658410834e-05,
+      "loss": 1.0708,
       "step": 10053
     },
     {
-      "epoch": 0.2761253467358765,
+      "epoch": 0.28530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.698787472241743e-05,
-      "loss": 0.906,
+      "learning_rate": 1.6772317457150694e-05,
+      "loss": 1.0053,
       "step": 10054
     },
     {
-      "epoch": 0.27615281096371974,
+      "epoch": 0.2853291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.6987238397868143e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.677164119868448e-05,
+      "loss": 0.8672,
       "step": 10055
     },
     {
-      "epoch": 0.27618027519156296,
+      "epoch": 0.28535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.6986602018033118e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.6770964883017913e-05,
+      "loss": 0.9792,
       "step": 10056
     },
     {
-      "epoch": 0.27620773941940624,
+      "epoch": 0.28538592508513055,
       "grad_norm": 0.0,
-      "learning_rate": 1.698596558291739e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.6770288510156697e-05,
+      "loss": 0.8649,
       "step": 10057
     },
     {
-      "epoch": 0.27623520364724946,
+      "epoch": 0.2854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.6985329092525997e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.6769612080106554e-05,
+      "loss": 1.0617,
       "step": 10058
     },
     {
-      "epoch": 0.2762626678750927,
+      "epoch": 0.28544267877412033,
       "grad_norm": 0.0,
-      "learning_rate": 1.6984692546863978e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.676893559287319e-05,
+      "loss": 0.9629,
       "step": 10059
     },
     {
-      "epoch": 0.2762901321029359,
+      "epoch": 0.2854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.6984055945936362e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.6768259048462324e-05,
+      "loss": 0.9921,
       "step": 10060
     },
     {
-      "epoch": 0.2763175963307792,
+      "epoch": 0.2854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.6983419289748192e-05,
-      "loss": 1.0772,
+      "learning_rate": 1.6767582446879668e-05,
+      "loss": 0.9587,
       "step": 10061
     },
     {
-      "epoch": 0.2763450605586224,
+      "epoch": 0.285527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.69827825783045e-05,
-      "loss": 0.968,
+      "learning_rate": 1.676690578813094e-05,
+      "loss": 0.9879,
       "step": 10062
     },
     {
-      "epoch": 0.2763725247864656,
+      "epoch": 0.2855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.6982145811610332e-05,
-      "loss": 1.0623,
+      "learning_rate": 1.6766229072221856e-05,
+      "loss": 1.0021,
       "step": 10063
     },
     {
-      "epoch": 0.27639998901430884,
+      "epoch": 0.2855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.698150898967072e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.6765552299158128e-05,
+      "loss": 0.9613,
       "step": 10064
     },
     {
-      "epoch": 0.2764274532421521,
+      "epoch": 0.28561293984108965,
       "grad_norm": 0.0,
-      "learning_rate": 1.69808721124907e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.6764875468945477e-05,
+      "loss": 0.9664,
       "step": 10065
     },
     {
-      "epoch": 0.27645491746999534,
+      "epoch": 0.28564131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.698023518007532e-05,
-      "loss": 0.9414,
+      "learning_rate": 1.6764198581589622e-05,
+      "loss": 0.9728,
       "step": 10066
     },
     {
-      "epoch": 0.27648238169783856,
+      "epoch": 0.2856696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.6979598192429616e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.6763521637096272e-05,
+      "loss": 0.9936,
       "step": 10067
     },
     {
-      "epoch": 0.2765098459256818,
+      "epoch": 0.28569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.6978961149558627e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.6762844635471153e-05,
+      "loss": 0.9737,
       "step": 10068
     },
     {
-      "epoch": 0.276537310153525,
+      "epoch": 0.28572644721906926,
       "grad_norm": 0.0,
-      "learning_rate": 1.6978324051467394e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.6762167576719983e-05,
+      "loss": 1.0716,
       "step": 10069
     },
     {
-      "epoch": 0.2765647743813683,
+      "epoch": 0.2857548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.6977686898160957e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.6761490460848476e-05,
+      "loss": 0.9276,
       "step": 10070
     },
     {
-      "epoch": 0.2765922386092115,
+      "epoch": 0.28578320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.6977049689644357e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.676081328786236e-05,
+      "loss": 1.0096,
       "step": 10071
     },
     {
-      "epoch": 0.2766197028370547,
+      "epoch": 0.2858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.6976412425922642e-05,
-      "loss": 1.04,
+      "learning_rate": 1.676013605776735e-05,
+      "loss": 0.8519,
       "step": 10072
     },
     {
-      "epoch": 0.27664716706489795,
+      "epoch": 0.2858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.6975775107000843e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.675945877056916e-05,
+      "loss": 0.962,
       "step": 10073
     },
     {
-      "epoch": 0.2766746312927412,
+      "epoch": 0.2858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.6975137732884017e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.6758781426273523e-05,
+      "loss": 0.8548,
       "step": 10074
     },
     {
-      "epoch": 0.27670209552058445,
+      "epoch": 0.2858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.6974500303577196e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.6758104024886156e-05,
+      "loss": 1.1035,
       "step": 10075
     },
     {
-      "epoch": 0.27672955974842767,
+      "epoch": 0.2859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.6973862819085423e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.675742656641278e-05,
+      "loss": 1.0142,
       "step": 10076
     },
     {
-      "epoch": 0.2767570239762709,
+      "epoch": 0.28595346197502836,
       "grad_norm": 0.0,
-      "learning_rate": 1.697322527941375e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.6756749050859117e-05,
+      "loss": 0.9699,
       "step": 10077
     },
     {
-      "epoch": 0.27678448820411417,
+      "epoch": 0.28598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.6972587684567215e-05,
-      "loss": 0.927,
+      "learning_rate": 1.675607147823089e-05,
+      "loss": 1.1038,
       "step": 10078
     },
     {
-      "epoch": 0.2768119524319574,
+      "epoch": 0.2860102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.697195003455087e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.6755393848533825e-05,
+      "loss": 0.8659,
       "step": 10079
     },
     {
-      "epoch": 0.2768394166598006,
+      "epoch": 0.28603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.6971312329369748e-05,
-      "loss": 0.9683,
+      "learning_rate": 1.6754716161773645e-05,
+      "loss": 0.8988,
       "step": 10080
     },
     {
-      "epoch": 0.27686688088764383,
+      "epoch": 0.28606696935300796,
       "grad_norm": 0.0,
-      "learning_rate": 1.697067456902891e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.6754038417956075e-05,
+      "loss": 0.8589,
       "step": 10081
     },
     {
-      "epoch": 0.27689434511548705,
+      "epoch": 0.2860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.6970036753533392e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.6753360617086835e-05,
+      "loss": 0.9694,
       "step": 10082
     },
     {
-      "epoch": 0.2769218093433303,
+      "epoch": 0.28612372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.6969398882888243e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.6752682759171657e-05,
+      "loss": 0.8881,
       "step": 10083
     },
     {
-      "epoch": 0.27694927357117355,
+      "epoch": 0.2861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.696876095709851e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.6752004844216264e-05,
+      "loss": 0.9368,
       "step": 10084
     },
     {
-      "epoch": 0.27697673779901677,
+      "epoch": 0.2861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.696812297616924e-05,
-      "loss": 0.975,
+      "learning_rate": 1.675132687222638e-05,
+      "loss": 1.0561,
       "step": 10085
     },
     {
-      "epoch": 0.27700420202686,
+      "epoch": 0.2862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.6967484940105482e-05,
-      "loss": 0.8054,
+      "learning_rate": 1.6750648843207736e-05,
+      "loss": 0.974,
       "step": 10086
     },
     {
-      "epoch": 0.27703166625470327,
+      "epoch": 0.2862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.6966846848912287e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.674997075716606e-05,
+      "loss": 1.0264,
       "step": 10087
     },
     {
-      "epoch": 0.2770591304825465,
+      "epoch": 0.2862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.69662087025947e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.6749292614107076e-05,
+      "loss": 0.8469,
       "step": 10088
     },
     {
-      "epoch": 0.2770865947103897,
+      "epoch": 0.28629398410896706,
       "grad_norm": 0.0,
-      "learning_rate": 1.696557050115777e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.6748614414036514e-05,
+      "loss": 0.9269,
       "step": 10089
     },
     {
-      "epoch": 0.27711405893823293,
+      "epoch": 0.286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.6964932244606546e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.6747936156960102e-05,
+      "loss": 0.9087,
       "step": 10090
     },
     {
-      "epoch": 0.2771415231660762,
+      "epoch": 0.2863507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.6964293932946083e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.6747257842883567e-05,
+      "loss": 0.9964,
       "step": 10091
     },
     {
-      "epoch": 0.27716898739391943,
+      "epoch": 0.28637911464245175,
       "grad_norm": 0.0,
-      "learning_rate": 1.696365556618143e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.6746579471812648e-05,
+      "loss": 1.0261,
       "step": 10092
     },
     {
-      "epoch": 0.27719645162176265,
+      "epoch": 0.28640749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 1.6963017144317633e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.6745901043753065e-05,
+      "loss": 0.9794,
       "step": 10093
     },
     {
-      "epoch": 0.2772239158496059,
+      "epoch": 0.2864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.6962378667359748e-05,
-      "loss": 1.0491,
+      "learning_rate": 1.6745222558710556e-05,
+      "loss": 0.9862,
       "step": 10094
     },
     {
-      "epoch": 0.27725138007744915,
+      "epoch": 0.28646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.696174013531283e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.6744544016690844e-05,
+      "loss": 0.8842,
       "step": 10095
     },
     {
-      "epoch": 0.27727884430529237,
+      "epoch": 0.28649262202043135,
       "grad_norm": 0.0,
-      "learning_rate": 1.696110154818192e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.674386541769967e-05,
+      "loss": 0.8536,
       "step": 10096
     },
     {
-      "epoch": 0.2773063085331356,
+      "epoch": 0.2865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.6960462905972082e-05,
-      "loss": 0.922,
+      "learning_rate": 1.6743186761742758e-05,
+      "loss": 1.0094,
       "step": 10097
     },
     {
-      "epoch": 0.2773337727609788,
+      "epoch": 0.28654937570942113,
       "grad_norm": 0.0,
-      "learning_rate": 1.6959824208688368e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.674250804882585e-05,
+      "loss": 1.0322,
       "step": 10098
     },
     {
-      "epoch": 0.27736123698882204,
+      "epoch": 0.286577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.6959185456335826e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.6741829278954664e-05,
+      "loss": 1.0242,
       "step": 10099
     },
     {
-      "epoch": 0.2773887012166653,
+      "epoch": 0.2866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.6958546648919513e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.674115045213495e-05,
+      "loss": 0.9568,
       "step": 10100
     },
     {
-      "epoch": 0.27741616544450853,
+      "epoch": 0.28663450624290576,
       "grad_norm": 0.0,
-      "learning_rate": 1.6957907786444482e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.674047156837243e-05,
+      "loss": 1.0736,
       "step": 10101
     },
     {
-      "epoch": 0.27744362967235175,
+      "epoch": 0.2866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.6957268868915796e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.673979262767285e-05,
+      "loss": 0.8956,
       "step": 10102
     },
     {
-      "epoch": 0.277471093900195,
+      "epoch": 0.2866912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.6956629896338497e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.6739113630041937e-05,
+      "loss": 1.0563,
       "step": 10103
     },
     {
-      "epoch": 0.27749855812803825,
+      "epoch": 0.28671963677639045,
       "grad_norm": 0.0,
-      "learning_rate": 1.6955990868717652e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.6738434575485425e-05,
+      "loss": 0.9891,
       "step": 10104
     },
     {
-      "epoch": 0.2775260223558815,
+      "epoch": 0.28674801362088537,
       "grad_norm": 0.0,
-      "learning_rate": 1.6955351786058308e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.6737755464009054e-05,
+      "loss": 0.8733,
       "step": 10105
     },
     {
-      "epoch": 0.2775534865837247,
+      "epoch": 0.2867763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.695471264836553e-05,
-      "loss": 1.006,
+      "learning_rate": 1.6737076295618565e-05,
+      "loss": 0.9749,
       "step": 10106
     },
     {
-      "epoch": 0.2775809508115679,
+      "epoch": 0.28680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.695407345564437e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.6736397070319682e-05,
+      "loss": 0.9367,
       "step": 10107
     },
     {
-      "epoch": 0.2776084150394112,
+      "epoch": 0.28683314415437006,
       "grad_norm": 0.0,
-      "learning_rate": 1.6953434207899886e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.6735717788118156e-05,
+      "loss": 1.0301,
       "step": 10108
     },
     {
-      "epoch": 0.2776358792672544,
+      "epoch": 0.2868615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.695279490513714e-05,
-      "loss": 1.0469,
+      "learning_rate": 1.673503844901972e-05,
+      "loss": 1.0486,
       "step": 10109
     },
     {
-      "epoch": 0.27766334349509764,
+      "epoch": 0.28688989784335983,
       "grad_norm": 0.0,
-      "learning_rate": 1.6952155547361185e-05,
-      "loss": 0.898,
+      "learning_rate": 1.6734359053030105e-05,
+      "loss": 0.8591,
       "step": 10110
     },
     {
-      "epoch": 0.27769080772294086,
+      "epoch": 0.2869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.6951516134577085e-05,
-      "loss": 1.019,
+      "learning_rate": 1.673367960015506e-05,
+      "loss": 0.9708,
       "step": 10111
     },
     {
-      "epoch": 0.2777182719507841,
+      "epoch": 0.2869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.695087666678989e-05,
-      "loss": 0.8549,
+      "learning_rate": 1.673300009040032e-05,
+      "loss": 0.9122,
       "step": 10112
     },
     {
-      "epoch": 0.27774573617862736,
+      "epoch": 0.2869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.6950237144004675e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.6732320523771626e-05,
+      "loss": 0.9739,
       "step": 10113
     },
     {
-      "epoch": 0.2777732004064706,
+      "epoch": 0.2870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.6949597566226485e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.6731640900274718e-05,
+      "loss": 0.9295,
       "step": 10114
     },
     {
-      "epoch": 0.2778006646343138,
+      "epoch": 0.2870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.694895793346039e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.6730961219915336e-05,
+      "loss": 0.9082,
       "step": 10115
     },
     {
-      "epoch": 0.277828128862157,
+      "epoch": 0.28706015891032916,
       "grad_norm": 0.0,
-      "learning_rate": 1.694831824571145e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.673028148269922e-05,
+      "loss": 0.9344,
       "step": 10116
     },
     {
-      "epoch": 0.2778555930900003,
+      "epoch": 0.28708853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.694767850298472e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.6729601688632117e-05,
+      "loss": 0.9014,
       "step": 10117
     },
     {
-      "epoch": 0.2778830573178435,
+      "epoch": 0.28711691259931893,
       "grad_norm": 0.0,
-      "learning_rate": 1.694703870528527e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.672892183771977e-05,
+      "loss": 0.9194,
       "step": 10118
     },
     {
-      "epoch": 0.27791052154568674,
+      "epoch": 0.28714528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.6946398852618156e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.672824192996791e-05,
+      "loss": 1.1315,
       "step": 10119
     },
     {
-      "epoch": 0.27793798577352996,
+      "epoch": 0.28717366628830876,
       "grad_norm": 0.0,
-      "learning_rate": 1.6945758944988443e-05,
-      "loss": 1.049,
+      "learning_rate": 1.672756196538229e-05,
+      "loss": 0.9282,
       "step": 10120
     },
     {
-      "epoch": 0.27796545000137324,
+      "epoch": 0.2872020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.6945118982401197e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.6726881943968653e-05,
+      "loss": 0.957,
       "step": 10121
     },
     {
-      "epoch": 0.27799291422921646,
+      "epoch": 0.28723041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.6944478964861478e-05,
-      "loss": 0.944,
+      "learning_rate": 1.672620186573274e-05,
+      "loss": 0.9746,
       "step": 10122
     },
     {
-      "epoch": 0.2780203784570597,
+      "epoch": 0.2872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.6943838892374352e-05,
-      "loss": 0.8781,
+      "learning_rate": 1.67255217306803e-05,
+      "loss": 0.9067,
       "step": 10123
     },
     {
-      "epoch": 0.2780478426849029,
+      "epoch": 0.2872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.6943198764944882e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.6724841538817073e-05,
+      "loss": 0.9413,
       "step": 10124
     },
     {
-      "epoch": 0.2780753069127461,
+      "epoch": 0.2873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.694255858257814e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.672416129014881e-05,
+      "loss": 0.9435,
       "step": 10125
     },
     {
-      "epoch": 0.2781027711405894,
+      "epoch": 0.2873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.6941918345279175e-05,
-      "loss": 0.8643,
+      "learning_rate": 1.672348098468125e-05,
+      "loss": 0.9693,
       "step": 10126
     },
     {
-      "epoch": 0.2781302353684326,
+      "epoch": 0.287372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.6941278053053068e-05,
-      "loss": 0.8875,
+      "learning_rate": 1.672280062242015e-05,
+      "loss": 1.0221,
       "step": 10127
     },
     {
-      "epoch": 0.27815769959627584,
+      "epoch": 0.28740068104426786,
       "grad_norm": 0.0,
-      "learning_rate": 1.6940637705904884e-05,
-      "loss": 1.053,
+      "learning_rate": 1.672212020337125e-05,
+      "loss": 0.9113,
       "step": 10128
     },
     {
-      "epoch": 0.27818516382411906,
+      "epoch": 0.2874290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.693999730383968e-05,
-      "loss": 0.907,
+      "learning_rate": 1.6721439727540294e-05,
+      "loss": 0.9841,
       "step": 10129
     },
     {
-      "epoch": 0.27821262805196234,
+      "epoch": 0.2874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.693935684686253e-05,
-      "loss": 0.8124,
+      "learning_rate": 1.6720759194933037e-05,
+      "loss": 0.9207,
       "step": 10130
     },
     {
-      "epoch": 0.27824009227980556,
+      "epoch": 0.28748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.6938716334978503e-05,
-      "loss": 0.9218,
+      "learning_rate": 1.6720078605555227e-05,
+      "loss": 0.9318,
       "step": 10131
     },
     {
-      "epoch": 0.2782675565076488,
+      "epoch": 0.28751418842224746,
       "grad_norm": 0.0,
-      "learning_rate": 1.6938075768192663e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.6719397959412606e-05,
+      "loss": 0.8866,
       "step": 10132
     },
     {
-      "epoch": 0.278295020735492,
+      "epoch": 0.2875425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.6937435146510083e-05,
-      "loss": 0.8507,
+      "learning_rate": 1.671871725651093e-05,
+      "loss": 0.9418,
       "step": 10133
     },
     {
-      "epoch": 0.2783224849633353,
+      "epoch": 0.28757094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.6936794469935824e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.671803649685595e-05,
+      "loss": 0.9501,
       "step": 10134
     },
     {
-      "epoch": 0.2783499491911785,
+      "epoch": 0.2875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.6936153738474962e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.6717355680453412e-05,
+      "loss": 0.9222,
       "step": 10135
     },
     {
-      "epoch": 0.2783774134190217,
+      "epoch": 0.287627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.6935512952132565e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.671667480730907e-05,
+      "loss": 0.9849,
       "step": 10136
     },
     {
-      "epoch": 0.27840487764686495,
+      "epoch": 0.2876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.6934872110913706e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.6715993877428672e-05,
+      "loss": 0.9528,
       "step": 10137
     },
     {
-      "epoch": 0.27843234187470817,
+      "epoch": 0.2876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.693423121482345e-05,
-      "loss": 1.0703,
+      "learning_rate": 1.6715312890817973e-05,
+      "loss": 0.9492,
       "step": 10138
     },
     {
-      "epoch": 0.27845980610255144,
+      "epoch": 0.2877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.693359026386687e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.6714631847482724e-05,
+      "loss": 1.0588,
       "step": 10139
     },
     {
-      "epoch": 0.27848727033039467,
+      "epoch": 0.28774120317820656,
       "grad_norm": 0.0,
-      "learning_rate": 1.6932949258049037e-05,
-      "loss": 1.0571,
+      "learning_rate": 1.6713950747428675e-05,
+      "loss": 1.0107,
       "step": 10140
     },
     {
-      "epoch": 0.2785147345582379,
+      "epoch": 0.2877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.6932308197375023e-05,
-      "loss": 0.908,
+      "learning_rate": 1.6713269590661587e-05,
+      "loss": 1.041,
       "step": 10141
     },
     {
-      "epoch": 0.2785421987860811,
+      "epoch": 0.2877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.6931667081849904e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.6712588377187207e-05,
+      "loss": 0.9328,
       "step": 10142
     },
     {
-      "epoch": 0.2785696630139244,
+      "epoch": 0.28782633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.693102591147875e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.671190710701129e-05,
+      "loss": 0.9569,
       "step": 10143
     },
     {
-      "epoch": 0.2785971272417676,
+      "epoch": 0.28785471055618617,
       "grad_norm": 0.0,
-      "learning_rate": 1.6930384686266635e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.6711225780139596e-05,
+      "loss": 0.9159,
       "step": 10144
     },
     {
-      "epoch": 0.2786245914696108,
+      "epoch": 0.287883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.6929743406218627e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.671054439657787e-05,
+      "loss": 0.9653,
       "step": 10145
     },
     {
-      "epoch": 0.27865205569745405,
+      "epoch": 0.28791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.692910207133981e-05,
-      "loss": 1.0725,
+      "learning_rate": 1.6709862956331877e-05,
+      "loss": 0.9777,
       "step": 10146
     },
     {
-      "epoch": 0.2786795199252973,
+      "epoch": 0.28793984108967086,
       "grad_norm": 0.0,
-      "learning_rate": 1.692846068163525e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.6709181459407368e-05,
+      "loss": 1.0216,
       "step": 10147
     },
     {
-      "epoch": 0.27870698415314055,
+      "epoch": 0.2879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.6927819237110027e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.6708499905810107e-05,
+      "loss": 0.9482,
       "step": 10148
     },
     {
-      "epoch": 0.27873444838098377,
+      "epoch": 0.28799659477866063,
       "grad_norm": 0.0,
-      "learning_rate": 1.6927177737769214e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.670781829554584e-05,
+      "loss": 0.996,
       "step": 10149
     },
     {
-      "epoch": 0.278761912608827,
+      "epoch": 0.2880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.692653618361789e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.6707136628620335e-05,
+      "loss": 1.0591,
       "step": 10150
     },
     {
-      "epoch": 0.2787893768366702,
+      "epoch": 0.2880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.6925894574661124e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.6706454905039343e-05,
+      "loss": 0.9516,
       "step": 10151
     },
     {
-      "epoch": 0.2788168410645135,
+      "epoch": 0.28808172531214526,
       "grad_norm": 0.0,
-      "learning_rate": 1.6925252910904003e-05,
-      "loss": 1.1014,
+      "learning_rate": 1.6705773124808622e-05,
+      "loss": 1.0047,
       "step": 10152
     },
     {
-      "epoch": 0.2788443052923567,
+      "epoch": 0.2881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.692461119235159e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.670509128793394e-05,
+      "loss": 1.0153,
       "step": 10153
     },
     {
-      "epoch": 0.27887176952019993,
+      "epoch": 0.2881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.692396941900898e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.6704409394421044e-05,
+      "loss": 1.0819,
       "step": 10154
     },
     {
-      "epoch": 0.27889923374804315,
+      "epoch": 0.28816685584562995,
       "grad_norm": 0.0,
-      "learning_rate": 1.6923327590881234e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.6703727444275704e-05,
+      "loss": 0.9752,
       "step": 10155
     },
     {
-      "epoch": 0.27892669797588643,
+      "epoch": 0.28819523269012487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6922685707973446e-05,
-      "loss": 1.0206,
+      "learning_rate": 1.6703045437503677e-05,
+      "loss": 1.031,
       "step": 10156
     },
     {
-      "epoch": 0.27895416220372965,
+      "epoch": 0.28822360953461973,
       "grad_norm": 0.0,
-      "learning_rate": 1.692204377029068e-05,
-      "loss": 1.0334,
+      "learning_rate": 1.6702363374110726e-05,
+      "loss": 0.9553,
       "step": 10157
     },
     {
-      "epoch": 0.27898162643157287,
+      "epoch": 0.28825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.6921401777838026e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.6701681254102608e-05,
+      "loss": 0.9656,
       "step": 10158
     },
     {
-      "epoch": 0.2790090906594161,
+      "epoch": 0.28828036322360956,
       "grad_norm": 0.0,
-      "learning_rate": 1.692075973062056e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.6700999077485085e-05,
+      "loss": 0.9686,
       "step": 10159
     },
     {
-      "epoch": 0.27903655488725937,
+      "epoch": 0.2883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.692011762864336e-05,
-      "loss": 1.0628,
+      "learning_rate": 1.6700316844263922e-05,
+      "loss": 0.9866,
       "step": 10160
     },
     {
-      "epoch": 0.2790640191151026,
+      "epoch": 0.28833711691259933,
       "grad_norm": 0.0,
-      "learning_rate": 1.6919475471911513e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.6699634554444885e-05,
+      "loss": 0.9723,
       "step": 10161
     },
     {
-      "epoch": 0.2790914833429458,
+      "epoch": 0.2883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.6918833260430093e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.669895220803373e-05,
+      "loss": 1.0749,
       "step": 10162
     },
     {
-      "epoch": 0.27911894757078903,
+      "epoch": 0.2883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.6918190994204186e-05,
-      "loss": 1.014,
+      "learning_rate": 1.6698269805036228e-05,
+      "loss": 0.9195,
       "step": 10163
     },
     {
-      "epoch": 0.27914641179863225,
+      "epoch": 0.28842224744608397,
       "grad_norm": 0.0,
-      "learning_rate": 1.691754867323887e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.669758734545814e-05,
+      "loss": 0.9361,
       "step": 10164
     },
     {
-      "epoch": 0.27917387602647553,
+      "epoch": 0.2884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.691690629753923e-05,
-      "loss": 0.9887,
+      "learning_rate": 1.6696904829305226e-05,
+      "loss": 1.0061,
       "step": 10165
     },
     {
-      "epoch": 0.27920134025431875,
+      "epoch": 0.2884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.6916263867110347e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.669622225658326e-05,
+      "loss": 0.9543,
       "step": 10166
     },
     {
-      "epoch": 0.279228804482162,
+      "epoch": 0.28850737797956866,
       "grad_norm": 0.0,
-      "learning_rate": 1.6915621381957302e-05,
-      "loss": 0.9992,
+      "learning_rate": 1.6695539627298e-05,
+      "loss": 0.9605,
       "step": 10167
     },
     {
-      "epoch": 0.2792562687100052,
+      "epoch": 0.2885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.6914978842085187e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.669485694145522e-05,
+      "loss": 0.899,
       "step": 10168
     },
     {
-      "epoch": 0.2792837329378485,
+      "epoch": 0.28856413166855843,
       "grad_norm": 0.0,
-      "learning_rate": 1.691433624749908e-05,
-      "loss": 1.012,
+      "learning_rate": 1.669417419906068e-05,
+      "loss": 0.9472,
       "step": 10169
     },
     {
-      "epoch": 0.2793111971656917,
+      "epoch": 0.28859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.6913693598204067e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.669349140012015e-05,
+      "loss": 1.001,
       "step": 10170
     },
     {
-      "epoch": 0.2793386613935349,
+      "epoch": 0.28862088535754826,
       "grad_norm": 0.0,
-      "learning_rate": 1.691305089420523e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.66928085446394e-05,
+      "loss": 1.039,
       "step": 10171
     },
     {
-      "epoch": 0.27936612562137814,
+      "epoch": 0.2886492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.6912408135507656e-05,
-      "loss": 1.0357,
+      "learning_rate": 1.669212563262419e-05,
+      "loss": 0.9663,
       "step": 10172
     },
     {
-      "epoch": 0.2793935898492214,
+      "epoch": 0.28867763904653804,
       "grad_norm": 0.0,
-      "learning_rate": 1.6911765322116432e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.6691442664080298e-05,
+      "loss": 1.0723,
       "step": 10173
     },
     {
-      "epoch": 0.27942105407706463,
+      "epoch": 0.2887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.6911122454036644e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.6690759639013487e-05,
+      "loss": 1.0672,
       "step": 10174
     },
     {
-      "epoch": 0.27944851830490786,
+      "epoch": 0.2887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.691047953127338e-05,
-      "loss": 0.942,
+      "learning_rate": 1.669007655742953e-05,
+      "loss": 0.8937,
       "step": 10175
     },
     {
-      "epoch": 0.2794759825327511,
+      "epoch": 0.2887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.690983655383172e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.6689393419334195e-05,
+      "loss": 1.0168,
       "step": 10176
     },
     {
-      "epoch": 0.2795034467605943,
+      "epoch": 0.2887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.690919352171676e-05,
-      "loss": 0.935,
+      "learning_rate": 1.6688710224733253e-05,
+      "loss": 0.8842,
       "step": 10177
     },
     {
-      "epoch": 0.2795309109884376,
+      "epoch": 0.2888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.6908550434933583e-05,
-      "loss": 1.0308,
+      "learning_rate": 1.6688026973632475e-05,
+      "loss": 1.0835,
       "step": 10178
     },
     {
-      "epoch": 0.2795583752162808,
+      "epoch": 0.28884790011350736,
       "grad_norm": 0.0,
-      "learning_rate": 1.690790729348728e-05,
-      "loss": 0.9285,
+      "learning_rate": 1.6687343666037632e-05,
+      "loss": 0.9792,
       "step": 10179
     },
     {
-      "epoch": 0.279585839444124,
+      "epoch": 0.2888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.690726409738294e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.6686660301954497e-05,
+      "loss": 0.8977,
       "step": 10180
     },
     {
-      "epoch": 0.27961330367196724,
+      "epoch": 0.28890465380249714,
       "grad_norm": 0.0,
-      "learning_rate": 1.690662084662565e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.668597688138884e-05,
+      "loss": 0.9464,
       "step": 10181
     },
     {
-      "epoch": 0.2796407678998105,
+      "epoch": 0.28893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.6905977541220498e-05,
-      "loss": 1.0855,
+      "learning_rate": 1.668529340434644e-05,
+      "loss": 0.9909,
       "step": 10182
     },
     {
-      "epoch": 0.27966823212765374,
+      "epoch": 0.28896140749148697,
       "grad_norm": 0.0,
-      "learning_rate": 1.690533418117258e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.668460987083306e-05,
+      "loss": 0.9716,
       "step": 10183
     },
     {
-      "epoch": 0.27969569635549696,
+      "epoch": 0.2889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.690469076648698e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.6683926280854484e-05,
+      "loss": 0.9364,
       "step": 10184
     },
     {
-      "epoch": 0.2797231605833402,
+      "epoch": 0.28901816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 1.6904047297168795e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.6683242634416484e-05,
+      "loss": 0.8679,
       "step": 10185
     },
     {
-      "epoch": 0.27975062481118346,
+      "epoch": 0.2890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.6903403773223113e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.668255893152483e-05,
+      "loss": 1.0388,
       "step": 10186
     },
     {
-      "epoch": 0.2797780890390267,
+      "epoch": 0.2890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.6902760194655023e-05,
-      "loss": 1.1025,
+      "learning_rate": 1.66818751721853e-05,
+      "loss": 0.9517,
       "step": 10187
     },
     {
-      "epoch": 0.2798055532668699,
+      "epoch": 0.28910329171396143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6902116561469625e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.6681191356403672e-05,
+      "loss": 0.9863,
       "step": 10188
     },
     {
-      "epoch": 0.2798330174947131,
+      "epoch": 0.2891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.6901472873672002e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.668050748418572e-05,
+      "loss": 0.9805,
       "step": 10189
     },
     {
-      "epoch": 0.2798604817225564,
+      "epoch": 0.2891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.6900829131267256e-05,
-      "loss": 0.9374,
+      "learning_rate": 1.667982355553722e-05,
+      "loss": 1.0341,
       "step": 10190
     },
     {
-      "epoch": 0.2798879459503996,
+      "epoch": 0.28918842224744606,
       "grad_norm": 0.0,
-      "learning_rate": 1.690018533426048e-05,
-      "loss": 1.0209,
+      "learning_rate": 1.6679139570463948e-05,
+      "loss": 1.0002,
       "step": 10191
     },
     {
-      "epoch": 0.27991541017824284,
+      "epoch": 0.289216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.6899541482656756e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.667845552897169e-05,
+      "loss": 0.8661,
       "step": 10192
     },
     {
-      "epoch": 0.27994287440608606,
+      "epoch": 0.2892451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.689889757646119e-05,
-      "loss": 0.9408,
+      "learning_rate": 1.6677771431066213e-05,
+      "loss": 0.8698,
       "step": 10193
     },
     {
-      "epoch": 0.2799703386339293,
+      "epoch": 0.28927355278093075,
       "grad_norm": 0.0,
-      "learning_rate": 1.6898253615678876e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.6677087276753302e-05,
+      "loss": 0.9868,
       "step": 10194
     },
     {
-      "epoch": 0.27999780286177256,
+      "epoch": 0.28930192962542567,
       "grad_norm": 0.0,
-      "learning_rate": 1.689760960031491e-05,
-      "loss": 0.9069,
+      "learning_rate": 1.6676403066038734e-05,
+      "loss": 1.0123,
       "step": 10195
     },
     {
-      "epoch": 0.2800252670896158,
+      "epoch": 0.28933030646992053,
       "grad_norm": 0.0,
-      "learning_rate": 1.689696553037438e-05,
-      "loss": 1.045,
+      "learning_rate": 1.667571879892829e-05,
+      "loss": 0.9043,
       "step": 10196
     },
     {
-      "epoch": 0.280052731317459,
+      "epoch": 0.28935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.689632140586239e-05,
-      "loss": 0.9882,
+      "learning_rate": 1.6675034475427747e-05,
+      "loss": 0.8604,
       "step": 10197
     },
     {
-      "epoch": 0.2800801955453022,
+      "epoch": 0.2893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.6895677226784034e-05,
-      "loss": 0.8927,
+      "learning_rate": 1.667435009554289e-05,
+      "loss": 0.9126,
       "step": 10198
     },
     {
-      "epoch": 0.2801076597731455,
+      "epoch": 0.2894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.6895032993144407e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.6673665659279496e-05,
+      "loss": 1.0352,
       "step": 10199
     },
     {
-      "epoch": 0.2801351240009887,
+      "epoch": 0.28944381384790013,
       "grad_norm": 0.0,
-      "learning_rate": 1.6894388704948608e-05,
-      "loss": 1.0127,
+      "learning_rate": 1.667298116664335e-05,
+      "loss": 0.8523,
       "step": 10200
     },
     {
-      "epoch": 0.28016258822883194,
+      "epoch": 0.289472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.6893744362201735e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.6672296617640232e-05,
+      "loss": 0.8985,
       "step": 10201
     },
     {
-      "epoch": 0.28019005245667516,
+      "epoch": 0.2895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.6893099964908885e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.6671612012275924e-05,
+      "loss": 1.0641,
       "step": 10202
     },
     {
-      "epoch": 0.28021751668451844,
+      "epoch": 0.28952894438138477,
       "grad_norm": 0.0,
-      "learning_rate": 1.689245551307516e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.667092735055621e-05,
+      "loss": 1.0157,
       "step": 10203
     },
     {
-      "epoch": 0.28024498091236166,
+      "epoch": 0.2895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.6891811006705655e-05,
-      "loss": 0.8874,
+      "learning_rate": 1.667024263248687e-05,
+      "loss": 1.0452,
       "step": 10204
     },
     {
-      "epoch": 0.2802724451402049,
+      "epoch": 0.2895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.6891166445805477e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.6669557858073696e-05,
+      "loss": 0.9522,
       "step": 10205
     },
     {
-      "epoch": 0.2802999093680481,
+      "epoch": 0.28961407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.6890521830379715e-05,
-      "loss": 0.8989,
+      "learning_rate": 1.6668873027322462e-05,
+      "loss": 0.9034,
       "step": 10206
     },
     {
-      "epoch": 0.2803273735958913,
+      "epoch": 0.28964245175936437,
       "grad_norm": 0.0,
-      "learning_rate": 1.6889877160433478e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.666818814023896e-05,
+      "loss": 0.9162,
       "step": 10207
     },
     {
-      "epoch": 0.2803548378237346,
+      "epoch": 0.28967082860385923,
       "grad_norm": 0.0,
-      "learning_rate": 1.688923243597186e-05,
-      "loss": 0.9909,
+      "learning_rate": 1.666750319682897e-05,
+      "loss": 1.0865,
       "step": 10208
     },
     {
-      "epoch": 0.2803823020515778,
+      "epoch": 0.28969920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.688858765699997e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.6666818197098286e-05,
+      "loss": 0.9926,
       "step": 10209
     },
     {
-      "epoch": 0.28040976627942105,
+      "epoch": 0.28972758229284906,
       "grad_norm": 0.0,
-      "learning_rate": 1.6887942823522907e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.6666133141052687e-05,
+      "loss": 0.9739,
       "step": 10210
     },
     {
-      "epoch": 0.28043723050726427,
+      "epoch": 0.2897559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.688729793554577e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.666544802869796e-05,
+      "loss": 0.9695,
       "step": 10211
     },
     {
-      "epoch": 0.28046469473510754,
+      "epoch": 0.28978433598183884,
       "grad_norm": 0.0,
-      "learning_rate": 1.6886652993073666e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.6664762860039893e-05,
+      "loss": 0.9166,
       "step": 10212
     },
     {
-      "epoch": 0.28049215896295077,
+      "epoch": 0.2898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.6886007996111694e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.666407763508428e-05,
+      "loss": 1.052,
       "step": 10213
     },
     {
-      "epoch": 0.280519623190794,
+      "epoch": 0.2898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.6885362944664962e-05,
-      "loss": 0.8824,
+      "learning_rate": 1.66633923538369e-05,
+      "loss": 1.0115,
       "step": 10214
     },
     {
-      "epoch": 0.2805470874186372,
+      "epoch": 0.28986946651532347,
       "grad_norm": 0.0,
-      "learning_rate": 1.6884717838738568e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.6662707016303542e-05,
+      "loss": 0.9583,
       "step": 10215
     },
     {
-      "epoch": 0.2805745516464805,
+      "epoch": 0.2898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.6884072678337618e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.6662021622490003e-05,
+      "loss": 0.9238,
       "step": 10216
     },
     {
-      "epoch": 0.2806020158743237,
+      "epoch": 0.2899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.6883427463467222e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.6661336172402068e-05,
+      "loss": 1.0939,
       "step": 10217
     },
     {
-      "epoch": 0.28062948010216693,
+      "epoch": 0.28995459704880816,
       "grad_norm": 0.0,
-      "learning_rate": 1.6882782194132482e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.6660650666045525e-05,
+      "loss": 1.0071,
       "step": 10218
     },
     {
-      "epoch": 0.28065694433001015,
+      "epoch": 0.2899829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.68821368703385e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.6659965103426166e-05,
+      "loss": 0.9791,
       "step": 10219
     },
     {
-      "epoch": 0.28068440855785337,
+      "epoch": 0.29001135073779793,
       "grad_norm": 0.0,
-      "learning_rate": 1.688149149209039e-05,
-      "loss": 0.8741,
+      "learning_rate": 1.6659279484549782e-05,
+      "loss": 0.8704,
       "step": 10220
     },
     {
-      "epoch": 0.28071187278569665,
+      "epoch": 0.29003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.688084605939325e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.6658593809422168e-05,
+      "loss": 0.9224,
       "step": 10221
     },
     {
-      "epoch": 0.28073933701353987,
+      "epoch": 0.29006810442678777,
       "grad_norm": 0.0,
-      "learning_rate": 1.688020057225219e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.665790807804911e-05,
+      "loss": 0.855,
       "step": 10222
     },
     {
-      "epoch": 0.2807668012413831,
+      "epoch": 0.2900964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.687955503067232e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.6657222290436404e-05,
+      "loss": 0.9383,
       "step": 10223
     },
     {
-      "epoch": 0.2807942654692263,
+      "epoch": 0.29012485811577754,
       "grad_norm": 0.0,
-      "learning_rate": 1.6878909434658745e-05,
-      "loss": 1.0418,
+      "learning_rate": 1.665653644658984e-05,
+      "loss": 0.8061,
       "step": 10224
     },
     {
-      "epoch": 0.2808217296970696,
+      "epoch": 0.2901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.6878263784216577e-05,
-      "loss": 0.9537,
+      "learning_rate": 1.6655850546515216e-05,
+      "loss": 1.0114,
       "step": 10225
     },
     {
-      "epoch": 0.2808491939249128,
+      "epoch": 0.2901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.6877618079350916e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.6655164590218325e-05,
+      "loss": 0.9953,
       "step": 10226
     },
     {
-      "epoch": 0.28087665815275603,
+      "epoch": 0.29020998864926223,
       "grad_norm": 0.0,
-      "learning_rate": 1.6876972320066877e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.6654478577704955e-05,
+      "loss": 0.9528,
       "step": 10227
     },
     {
-      "epoch": 0.28090412238059925,
+      "epoch": 0.2902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.687632650636957e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.665379250898091e-05,
+      "loss": 0.9641,
       "step": 10228
     },
     {
-      "epoch": 0.28093158660844253,
+      "epoch": 0.290266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.6875680638264107e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.6653106384051977e-05,
+      "loss": 0.9447,
       "step": 10229
     },
     {
-      "epoch": 0.28095905083628575,
+      "epoch": 0.29029511918274686,
       "grad_norm": 0.0,
-      "learning_rate": 1.6875034715755593e-05,
-      "loss": 0.948,
+      "learning_rate": 1.6652420202923956e-05,
+      "loss": 1.065,
       "step": 10230
     },
     {
-      "epoch": 0.28098651506412897,
+      "epoch": 0.2903234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.687438873884914e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.665173396560264e-05,
+      "loss": 1.0095,
       "step": 10231
     },
     {
-      "epoch": 0.2810139792919722,
+      "epoch": 0.29035187287173664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6873742707549862e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.6651047672093836e-05,
+      "loss": 0.9404,
       "step": 10232
     },
     {
-      "epoch": 0.2810414435198154,
+      "epoch": 0.29038024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.687309662186287e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.6650361322403324e-05,
+      "loss": 0.894,
       "step": 10233
     },
     {
-      "epoch": 0.2810689077476587,
+      "epoch": 0.29040862656072647,
       "grad_norm": 0.0,
-      "learning_rate": 1.6872450481793277e-05,
-      "loss": 1.0542,
+      "learning_rate": 1.6649674916536916e-05,
+      "loss": 0.885,
       "step": 10234
     },
     {
-      "epoch": 0.2810963719755019,
+      "epoch": 0.29043700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.687180428734619e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.6648988454500407e-05,
+      "loss": 0.9014,
       "step": 10235
     },
     {
-      "epoch": 0.28112383620334513,
+      "epoch": 0.29046538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.6871158038526725e-05,
-      "loss": 1.0397,
+      "learning_rate": 1.6648301936299592e-05,
+      "loss": 1.1212,
       "step": 10236
     },
     {
-      "epoch": 0.28115130043118836,
+      "epoch": 0.2904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.6870511735340002e-05,
-      "loss": 0.9812,
+      "learning_rate": 1.664761536194027e-05,
+      "loss": 0.9393,
       "step": 10237
     },
     {
-      "epoch": 0.28117876465903163,
+      "epoch": 0.290522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.686986537779112e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.664692873142824e-05,
+      "loss": 0.9719,
       "step": 10238
     },
     {
-      "epoch": 0.28120622888687485,
+      "epoch": 0.29055051078320093,
       "grad_norm": 0.0,
-      "learning_rate": 1.686921896588521e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.6646242044769308e-05,
+      "loss": 0.9331,
       "step": 10239
     },
     {
-      "epoch": 0.2812336931147181,
+      "epoch": 0.2905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.686857249962737e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.664555530196927e-05,
+      "loss": 0.9982,
       "step": 10240
     },
     {
-      "epoch": 0.2812611573425613,
+      "epoch": 0.2906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.6867925979022734e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.6644868503033927e-05,
+      "loss": 0.9877,
       "step": 10241
     },
     {
-      "epoch": 0.2812886215704046,
+      "epoch": 0.29063564131668557,
       "grad_norm": 0.0,
-      "learning_rate": 1.6867279404076396e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.6644181647969082e-05,
+      "loss": 0.93,
       "step": 10242
     },
     {
-      "epoch": 0.2813160857982478,
+      "epoch": 0.2906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.686663277479349e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.6643494736780534e-05,
+      "loss": 1.0529,
       "step": 10243
     },
     {
-      "epoch": 0.281343550026091,
+      "epoch": 0.29069239500567534,
       "grad_norm": 0.0,
-      "learning_rate": 1.686598609117913e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.664280776947409e-05,
+      "loss": 1.0376,
       "step": 10244
     },
     {
-      "epoch": 0.28137101425393424,
+      "epoch": 0.29072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.686533935323842e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.6642120746055547e-05,
+      "loss": 0.946,
       "step": 10245
     },
     {
-      "epoch": 0.28139847848177746,
+      "epoch": 0.29074914869466517,
       "grad_norm": 0.0,
-      "learning_rate": 1.6864692560976486e-05,
-      "loss": 1.0712,
+      "learning_rate": 1.6641433666530714e-05,
+      "loss": 1.0069,
       "step": 10246
     },
     {
-      "epoch": 0.28142594270962074,
+      "epoch": 0.29077752553916003,
       "grad_norm": 0.0,
-      "learning_rate": 1.6864045714398453e-05,
-      "loss": 0.9091,
+      "learning_rate": 1.6640746530905395e-05,
+      "loss": 0.9666,
       "step": 10247
     },
     {
-      "epoch": 0.28145340693746396,
+      "epoch": 0.29080590238365495,
       "grad_norm": 0.0,
-      "learning_rate": 1.6863398813509423e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.6640059339185387e-05,
+      "loss": 0.9433,
       "step": 10248
     },
     {
-      "epoch": 0.2814808711653072,
+      "epoch": 0.2908342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.6862751858314528e-05,
-      "loss": 1.0834,
+      "learning_rate": 1.66393720913765e-05,
+      "loss": 1.0053,
       "step": 10249
     },
     {
-      "epoch": 0.2815083353931504,
+      "epoch": 0.2908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.6862104848818882e-05,
-      "loss": 1.009,
+      "learning_rate": 1.663868478748454e-05,
+      "loss": 0.9163,
       "step": 10250
     },
     {
-      "epoch": 0.2815357996209937,
+      "epoch": 0.29089103291713964,
       "grad_norm": 0.0,
-      "learning_rate": 1.68614577850276e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.663799742751531e-05,
+      "loss": 1.0036,
       "step": 10251
     },
     {
-      "epoch": 0.2815632638488369,
+      "epoch": 0.2909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.686081066694581e-05,
-      "loss": 0.981,
+      "learning_rate": 1.6637310011474618e-05,
+      "loss": 1.0304,
       "step": 10252
     },
     {
-      "epoch": 0.2815907280766801,
+      "epoch": 0.2909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.686016349457863e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.663662253936827e-05,
+      "loss": 0.9609,
       "step": 10253
     },
     {
-      "epoch": 0.28161819230452334,
+      "epoch": 0.29097616345062427,
       "grad_norm": 0.0,
-      "learning_rate": 1.6859516267931172e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.663593501120207e-05,
+      "loss": 0.829,
       "step": 10254
     },
     {
-      "epoch": 0.2816456565323666,
+      "epoch": 0.2910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.6858868987008572e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.6635247426981833e-05,
+      "loss": 1.0197,
       "step": 10255
     },
     {
-      "epoch": 0.28167312076020984,
+      "epoch": 0.2910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.685822165181594e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.663455978671336e-05,
+      "loss": 0.9952,
       "step": 10256
     },
     {
-      "epoch": 0.28170058498805306,
+      "epoch": 0.29106129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.68575742623584e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.6633872090402465e-05,
+      "loss": 0.9796,
       "step": 10257
     },
     {
-      "epoch": 0.2817280492158963,
+      "epoch": 0.2910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.685692681864108e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.6633184338054947e-05,
+      "loss": 0.9834,
       "step": 10258
     },
     {
-      "epoch": 0.2817555134437395,
+      "epoch": 0.29111804767309873,
       "grad_norm": 0.0,
-      "learning_rate": 1.68562793206691e-05,
-      "loss": 1.0001,
+      "learning_rate": 1.663249652967663e-05,
+      "loss": 0.8614,
       "step": 10259
     },
     {
-      "epoch": 0.2817829776715828,
+      "epoch": 0.29114642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.6855631768447575e-05,
-      "loss": 1.014,
+      "learning_rate": 1.6631808665273315e-05,
+      "loss": 0.9581,
       "step": 10260
     },
     {
-      "epoch": 0.281810441899426,
+      "epoch": 0.2911748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.685498416198164e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.663112074485081e-05,
+      "loss": 0.9569,
       "step": 10261
     },
     {
-      "epoch": 0.2818379061272692,
+      "epoch": 0.2912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.6854336501276416e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.6630432768414937e-05,
+      "loss": 1.0609,
       "step": 10262
     },
     {
-      "epoch": 0.28186537035511244,
+      "epoch": 0.29123155505107834,
       "grad_norm": 0.0,
-      "learning_rate": 1.6853688786337024e-05,
-      "loss": 1.0467,
+      "learning_rate": 1.66297447359715e-05,
+      "loss": 0.9196,
       "step": 10263
     },
     {
-      "epoch": 0.2818928345829557,
+      "epoch": 0.2912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.6853041017168594e-05,
-      "loss": 0.9441,
+      "learning_rate": 1.6629056647526306e-05,
+      "loss": 0.967,
       "step": 10264
     },
     {
-      "epoch": 0.28192029881079894,
+      "epoch": 0.2912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.6852393193776248e-05,
-      "loss": 1.015,
+      "learning_rate": 1.6628368503085174e-05,
+      "loss": 0.9626,
       "step": 10265
     },
     {
-      "epoch": 0.28194776303864216,
+      "epoch": 0.291316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.685174531616511e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.6627680302653914e-05,
+      "loss": 0.9044,
       "step": 10266
     },
     {
-      "epoch": 0.2819752272664854,
+      "epoch": 0.2913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.6851097384340313e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.6626992046238343e-05,
+      "loss": 0.9201,
       "step": 10267
     },
     {
-      "epoch": 0.28200269149432866,
+      "epoch": 0.2913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.685044939830698e-05,
-      "loss": 0.8445,
+      "learning_rate": 1.6626303733844272e-05,
+      "loss": 0.9098,
       "step": 10268
     },
     {
-      "epoch": 0.2820301557221719,
+      "epoch": 0.29140181611804766,
       "grad_norm": 0.0,
-      "learning_rate": 1.684980135807023e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.6625615365477515e-05,
+      "loss": 0.8735,
       "step": 10269
     },
     {
-      "epoch": 0.2820576199500151,
+      "epoch": 0.2914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.6849153263635206e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.662492694114389e-05,
+      "loss": 0.8757,
       "step": 10270
     },
     {
-      "epoch": 0.2820850841778583,
+      "epoch": 0.29145856980703744,
       "grad_norm": 0.0,
-      "learning_rate": 1.6848505115007027e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.6624238460849203e-05,
+      "loss": 0.9169,
       "step": 10271
     },
     {
-      "epoch": 0.28211254840570155,
+      "epoch": 0.29148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.6847856912190823e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.6623549924599283e-05,
+      "loss": 0.9701,
       "step": 10272
     },
     {
-      "epoch": 0.2821400126335448,
+      "epoch": 0.29151532349602727,
       "grad_norm": 0.0,
-      "learning_rate": 1.684720865519172e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.6622861332399936e-05,
+      "loss": 0.9759,
       "step": 10273
     },
     {
-      "epoch": 0.28216747686138804,
+      "epoch": 0.2915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.6846560344014852e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.6622172684256984e-05,
+      "loss": 0.9519,
       "step": 10274
     },
     {
-      "epoch": 0.28219494108923127,
+      "epoch": 0.29157207718501704,
       "grad_norm": 0.0,
-      "learning_rate": 1.6845911978665344e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.6621483980176237e-05,
+      "loss": 0.9264,
       "step": 10275
     },
     {
-      "epoch": 0.2822224053170745,
+      "epoch": 0.2916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.6845263559148334e-05,
-      "loss": 1.0324,
+      "learning_rate": 1.662079522016352e-05,
+      "loss": 0.9414,
       "step": 10276
     },
     {
-      "epoch": 0.28224986954491776,
+      "epoch": 0.2916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.684461508546894e-05,
-      "loss": 1.025,
+      "learning_rate": 1.6620106404224644e-05,
+      "loss": 0.9426,
       "step": 10277
     },
     {
-      "epoch": 0.282277333772761,
+      "epoch": 0.2916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.6843966557632306e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.6619417532365437e-05,
+      "loss": 1.0314,
       "step": 10278
     },
     {
-      "epoch": 0.2823047980006042,
+      "epoch": 0.2916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.6843317975643555e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.6618728604591708e-05,
+      "loss": 0.8917,
       "step": 10279
     },
     {
-      "epoch": 0.28233226222844743,
+      "epoch": 0.2917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.684266933950782e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.6618039620909285e-05,
+      "loss": 0.9344,
       "step": 10280
     },
     {
-      "epoch": 0.2823597264562907,
+      "epoch": 0.29174233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.6842020649230236e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.6617350581323985e-05,
+      "loss": 1.1064,
       "step": 10281
     },
     {
-      "epoch": 0.2823871906841339,
+      "epoch": 0.2917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.6841371904815934e-05,
-      "loss": 1.0843,
+      "learning_rate": 1.6616661485841622e-05,
+      "loss": 0.9436,
       "step": 10282
     },
     {
-      "epoch": 0.28241465491197715,
+      "epoch": 0.29179909194097614,
       "grad_norm": 0.0,
-      "learning_rate": 1.6840723106270047e-05,
-      "loss": 0.8893,
+      "learning_rate": 1.6615972334468023e-05,
+      "loss": 1.0105,
       "step": 10283
     },
     {
-      "epoch": 0.28244211913982037,
+      "epoch": 0.29182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.6840074253597708e-05,
-      "loss": 0.8968,
+      "learning_rate": 1.661528312720901e-05,
+      "loss": 0.9797,
       "step": 10284
     },
     {
-      "epoch": 0.28246958336766365,
+      "epoch": 0.29185584562996597,
       "grad_norm": 0.0,
-      "learning_rate": 1.6839425346804056e-05,
-      "loss": 1.1024,
+      "learning_rate": 1.6614593864070404e-05,
+      "loss": 0.8805,
       "step": 10285
     },
     {
-      "epoch": 0.28249704759550687,
+      "epoch": 0.29188422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 1.6838776385894218e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.6613904545058025e-05,
+      "loss": 0.9598,
       "step": 10286
     },
     {
-      "epoch": 0.2825245118233501,
+      "epoch": 0.29191259931895575,
       "grad_norm": 0.0,
-      "learning_rate": 1.6838127370873333e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.6613215170177698e-05,
+      "loss": 0.9339,
       "step": 10287
     },
     {
-      "epoch": 0.2825519760511933,
+      "epoch": 0.2919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.6837478301746535e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.6612525739435245e-05,
+      "loss": 0.9551,
       "step": 10288
     },
     {
-      "epoch": 0.28257944027903653,
+      "epoch": 0.2919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.683682917851896e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.661183625283649e-05,
+      "loss": 0.9582,
       "step": 10289
     },
     {
-      "epoch": 0.2826069045068798,
+      "epoch": 0.29199772985244044,
       "grad_norm": 0.0,
-      "learning_rate": 1.6836180001195745e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.6611146710387254e-05,
+      "loss": 0.9484,
       "step": 10290
     },
     {
-      "epoch": 0.28263436873472303,
+      "epoch": 0.2920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.6835530769782025e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.6610457112093364e-05,
+      "loss": 0.8951,
       "step": 10291
     },
     {
-      "epoch": 0.28266183296256625,
+      "epoch": 0.2920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.6834881484282938e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.660976745796065e-05,
+      "loss": 0.9579,
       "step": 10292
     },
     {
-      "epoch": 0.28268929719040947,
+      "epoch": 0.29208286038592507,
       "grad_norm": 0.0,
-      "learning_rate": 1.6834232144703622e-05,
-      "loss": 0.9136,
+      "learning_rate": 1.660907774799493e-05,
+      "loss": 1.0284,
       "step": 10293
     },
     {
-      "epoch": 0.28271676141825275,
+      "epoch": 0.29211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.6833582751049215e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.6608387982202034e-05,
+      "loss": 0.8521,
       "step": 10294
     },
     {
-      "epoch": 0.28274422564609597,
+      "epoch": 0.29213961407491484,
       "grad_norm": 0.0,
-      "learning_rate": 1.683293330332485e-05,
-      "loss": 0.931,
+      "learning_rate": 1.6607698160587792e-05,
+      "loss": 0.9425,
       "step": 10295
     },
     {
-      "epoch": 0.2827716898739392,
+      "epoch": 0.29216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.6832283801535672e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.6607008283158023e-05,
+      "loss": 0.9779,
       "step": 10296
     },
     {
-      "epoch": 0.2827991541017824,
+      "epoch": 0.2921963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.6831634245686815e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.6606318349918557e-05,
+      "loss": 1.0224,
       "step": 10297
     },
     {
-      "epoch": 0.2828266183296257,
+      "epoch": 0.29222474460839953,
       "grad_norm": 0.0,
-      "learning_rate": 1.6830984635783426e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.660562836087522e-05,
+      "loss": 1.0372,
       "step": 10298
     },
     {
-      "epoch": 0.2828540825574689,
+      "epoch": 0.29225312145289445,
       "grad_norm": 0.0,
-      "learning_rate": 1.683033497183064e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.660493831603385e-05,
+      "loss": 0.9023,
       "step": 10299
     },
     {
-      "epoch": 0.28288154678531213,
+      "epoch": 0.2922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.6829685253833597e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.6604248215400266e-05,
+      "loss": 1.0209,
       "step": 10300
     },
     {
-      "epoch": 0.28290901101315535,
+      "epoch": 0.2923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.6829035481797436e-05,
-      "loss": 1.0361,
+      "learning_rate": 1.6603558058980296e-05,
+      "loss": 1.0225,
       "step": 10301
     },
     {
-      "epoch": 0.2829364752409986,
+      "epoch": 0.29233825198637914,
       "grad_norm": 0.0,
-      "learning_rate": 1.6828385655727306e-05,
-      "loss": 0.9303,
+      "learning_rate": 1.6602867846779782e-05,
+      "loss": 0.85,
       "step": 10302
     },
     {
-      "epoch": 0.28296393946884185,
+      "epoch": 0.292366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.682773577562834e-05,
-      "loss": 1.0872,
+      "learning_rate": 1.6602177578804544e-05,
+      "loss": 0.9618,
       "step": 10303
     },
     {
-      "epoch": 0.2829914036966851,
+      "epoch": 0.2923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.6827085841505687e-05,
-      "loss": 1.1099,
+      "learning_rate": 1.6601487255060416e-05,
+      "loss": 0.9409,
       "step": 10304
     },
     {
-      "epoch": 0.2830188679245283,
+      "epoch": 0.2924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.6826435853364486e-05,
-      "loss": 0.8921,
+      "learning_rate": 1.6600796875553225e-05,
+      "loss": 0.9452,
       "step": 10305
     },
     {
-      "epoch": 0.2830463321523715,
+      "epoch": 0.2924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.6825785811209877e-05,
-      "loss": 1.0157,
+      "learning_rate": 1.6600106440288807e-05,
+      "loss": 0.9583,
       "step": 10306
     },
     {
-      "epoch": 0.2830737963802148,
+      "epoch": 0.2924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.6825135715047008e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.6599415949272995e-05,
+      "loss": 0.968,
       "step": 10307
     },
     {
-      "epoch": 0.283101260608058,
+      "epoch": 0.29250851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 1.682448556488102e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.6598725402511618e-05,
+      "loss": 0.8703,
       "step": 10308
     },
     {
-      "epoch": 0.28312872483590124,
+      "epoch": 0.2925368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.682383536071706e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.6598034800010512e-05,
+      "loss": 1.0996,
       "step": 10309
     },
     {
-      "epoch": 0.28315618906374446,
+      "epoch": 0.29256526674233824,
       "grad_norm": 0.0,
-      "learning_rate": 1.6823185102560275e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.659734414177551e-05,
+      "loss": 0.8692,
       "step": 10310
     },
     {
-      "epoch": 0.28318365329158773,
+      "epoch": 0.29259364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.68225347904158e-05,
-      "loss": 1.0695,
+      "learning_rate": 1.6596653427812443e-05,
+      "loss": 0.9474,
       "step": 10311
     },
     {
-      "epoch": 0.28321111751943095,
+      "epoch": 0.292622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.682188442428879e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.6595962658127152e-05,
+      "loss": 0.9875,
       "step": 10312
     },
     {
-      "epoch": 0.2832385817472742,
+      "epoch": 0.2926503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.6821234004184387e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.6595271832725467e-05,
+      "loss": 0.9297,
       "step": 10313
     },
     {
-      "epoch": 0.2832660459751174,
+      "epoch": 0.29267877412031784,
       "grad_norm": 0.0,
-      "learning_rate": 1.682058353010774e-05,
-      "loss": 1.0842,
+      "learning_rate": 1.659458095161322e-05,
+      "loss": 0.9109,
       "step": 10314
     },
     {
-      "epoch": 0.2832935102029606,
+      "epoch": 0.2927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.6819933002063994e-05,
-      "loss": 1.0149,
+      "learning_rate": 1.6593890014796256e-05,
+      "loss": 0.9767,
       "step": 10315
     },
     {
-      "epoch": 0.2833209744308039,
+      "epoch": 0.2927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.6819282420058293e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.6593199022280405e-05,
+      "loss": 0.8983,
       "step": 10316
     },
     {
-      "epoch": 0.2833484386586471,
+      "epoch": 0.2927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.681863178409579e-05,
-      "loss": 0.96,
+      "learning_rate": 1.6592507974071504e-05,
+      "loss": 0.9545,
       "step": 10317
     },
     {
-      "epoch": 0.28337590288649034,
+      "epoch": 0.2927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.6817981094181633e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.659181687017539e-05,
+      "loss": 1.0243,
       "step": 10318
     },
     {
-      "epoch": 0.28340336711433356,
+      "epoch": 0.2928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.681733035032097e-05,
-      "loss": 0.8715,
+      "learning_rate": 1.6591125710597908e-05,
+      "loss": 0.8605,
       "step": 10319
     },
     {
-      "epoch": 0.28343083134217684,
+      "epoch": 0.29284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.681667955251894e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.659043449534489e-05,
+      "loss": 0.9138,
       "step": 10320
     },
     {
-      "epoch": 0.28345829557002006,
+      "epoch": 0.2928774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.6816028700780708e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.658974322442217e-05,
+      "loss": 0.9234,
       "step": 10321
     },
     {
-      "epoch": 0.2834857597978633,
+      "epoch": 0.29290578887627694,
       "grad_norm": 0.0,
-      "learning_rate": 1.6815377795111415e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.65890518978356e-05,
+      "loss": 0.9239,
       "step": 10322
     },
     {
-      "epoch": 0.2835132240257065,
+      "epoch": 0.29293416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 1.6814726835516214e-05,
-      "loss": 0.9028,
+      "learning_rate": 1.6588360515591007e-05,
+      "loss": 0.8954,
       "step": 10323
     },
     {
-      "epoch": 0.2835406882535498,
+      "epoch": 0.2929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.6814075822000254e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.6587669077694242e-05,
+      "loss": 1.01,
       "step": 10324
     },
     {
-      "epoch": 0.283568152481393,
+      "epoch": 0.29299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.6813424754568686e-05,
-      "loss": 0.8208,
+      "learning_rate": 1.6586977584151135e-05,
+      "loss": 1.0904,
       "step": 10325
     },
     {
-      "epoch": 0.2835956167092362,
+      "epoch": 0.29301929625425654,
       "grad_norm": 0.0,
-      "learning_rate": 1.681277363322666e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.6586286034967535e-05,
+      "loss": 1.0764,
       "step": 10326
     },
     {
-      "epoch": 0.28362308093707944,
+      "epoch": 0.2930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.6812122457979332e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.658559443014928e-05,
+      "loss": 0.9535,
       "step": 10327
     },
     {
-      "epoch": 0.28365054516492266,
+      "epoch": 0.2930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.6811471228831853e-05,
-      "loss": 0.8633,
+      "learning_rate": 1.6584902769702214e-05,
+      "loss": 1.0176,
       "step": 10328
     },
     {
-      "epoch": 0.28367800939276594,
+      "epoch": 0.2931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.6810819945789373e-05,
-      "loss": 0.981,
+      "learning_rate": 1.6584211053632174e-05,
+      "loss": 1.0875,
       "step": 10329
     },
     {
-      "epoch": 0.28370547362060916,
+      "epoch": 0.2931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.681016860885705e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.658351928194501e-05,
+      "loss": 0.9424,
       "step": 10330
     },
     {
-      "epoch": 0.2837329378484524,
+      "epoch": 0.293161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.6809517218040032e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.6582827454646565e-05,
+      "loss": 0.8866,
       "step": 10331
     },
     {
-      "epoch": 0.2837604020762956,
+      "epoch": 0.29318955732122587,
       "grad_norm": 0.0,
-      "learning_rate": 1.680886577334348e-05,
-      "loss": 0.885,
+      "learning_rate": 1.658213557174268e-05,
+      "loss": 1.017,
       "step": 10332
     },
     {
-      "epoch": 0.2837878663041389,
+      "epoch": 0.2932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.680821427477254e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.6581443633239197e-05,
+      "loss": 0.9946,
       "step": 10333
     },
     {
-      "epoch": 0.2838153305319821,
+      "epoch": 0.29324631101021564,
       "grad_norm": 0.0,
-      "learning_rate": 1.6807562722332375e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.6580751639141966e-05,
+      "loss": 0.959,
       "step": 10334
     },
     {
-      "epoch": 0.2838427947598253,
+      "epoch": 0.29327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.680691111602814e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.658005958945683e-05,
+      "loss": 0.984,
       "step": 10335
     },
     {
-      "epoch": 0.28387025898766854,
+      "epoch": 0.2933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.680625945586498e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.6579367484189632e-05,
+      "loss": 0.875,
       "step": 10336
     },
     {
-      "epoch": 0.2838977232155118,
+      "epoch": 0.29333144154370033,
       "grad_norm": 0.0,
-      "learning_rate": 1.680560774184806e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.6578675323346224e-05,
+      "loss": 0.8768,
       "step": 10337
     },
     {
-      "epoch": 0.28392518744335504,
+      "epoch": 0.29335981838819525,
       "grad_norm": 0.0,
-      "learning_rate": 1.6804955973982538e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.6577983106932448e-05,
+      "loss": 1.012,
       "step": 10338
     },
     {
-      "epoch": 0.28395265167119826,
+      "epoch": 0.2933881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.6804304152273565e-05,
-      "loss": 0.9905,
+      "learning_rate": 1.6577290834954153e-05,
+      "loss": 0.9374,
       "step": 10339
     },
     {
-      "epoch": 0.2839801158990415,
+      "epoch": 0.293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.6803652276726306e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.657659850741719e-05,
+      "loss": 0.9537,
       "step": 10340
     },
     {
-      "epoch": 0.2840075801268847,
+      "epoch": 0.2934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.6803000347345913e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.6575906124327403e-05,
+      "loss": 0.917,
       "step": 10341
     },
     {
-      "epoch": 0.284035044354728,
+      "epoch": 0.2934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.6802348364137544e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.657521368569064e-05,
+      "loss": 0.9595,
       "step": 10342
     },
     {
-      "epoch": 0.2840625085825712,
+      "epoch": 0.2935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.680169632710636e-05,
-      "loss": 0.936,
+      "learning_rate": 1.657452119151275e-05,
+      "loss": 1.1124,
       "step": 10343
     },
     {
-      "epoch": 0.2840899728104144,
+      "epoch": 0.29353007945516457,
       "grad_norm": 0.0,
-      "learning_rate": 1.6801044236257524e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.6573828641799585e-05,
+      "loss": 1.0206,
       "step": 10344
     },
     {
-      "epoch": 0.28411743703825765,
+      "epoch": 0.2935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.680039209159619e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.6573136036556994e-05,
+      "loss": 1.0439,
       "step": 10345
     },
     {
-      "epoch": 0.2841449012661009,
+      "epoch": 0.29358683314415435,
       "grad_norm": 0.0,
-      "learning_rate": 1.679973989312752e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.6572443375790824e-05,
+      "loss": 1.0724,
       "step": 10346
     },
     {
-      "epoch": 0.28417236549394415,
+      "epoch": 0.29361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.6799087640856672e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.6571750659506937e-05,
+      "loss": 1.0187,
       "step": 10347
     },
     {
-      "epoch": 0.28419982972178737,
+      "epoch": 0.2936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.6798435334788808e-05,
-      "loss": 0.936,
+      "learning_rate": 1.657105788771117e-05,
+      "loss": 0.8777,
       "step": 10348
     },
     {
-      "epoch": 0.2842272939496306,
+      "epoch": 0.29367196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.6797782974929092e-05,
-      "loss": 0.9122,
+      "learning_rate": 1.6570365060409388e-05,
+      "loss": 0.9233,
       "step": 10349
     },
     {
-      "epoch": 0.28425475817747387,
+      "epoch": 0.29370034052213395,
       "grad_norm": 0.0,
-      "learning_rate": 1.679713056128268e-05,
-      "loss": 0.91,
+      "learning_rate": 1.6569672177607432e-05,
+      "loss": 0.9647,
       "step": 10350
     },
     {
-      "epoch": 0.2842822224053171,
+      "epoch": 0.2937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.6796478093854744e-05,
-      "loss": 0.9372,
+      "learning_rate": 1.656897923931116e-05,
+      "loss": 0.9073,
       "step": 10351
     },
     {
-      "epoch": 0.2843096866331603,
+      "epoch": 0.2937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.679582557265044e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.6568286245526425e-05,
+      "loss": 1.0138,
       "step": 10352
     },
     {
-      "epoch": 0.28433715086100353,
+      "epoch": 0.29378547105561864,
       "grad_norm": 0.0,
-      "learning_rate": 1.679517299767493e-05,
-      "loss": 0.9036,
+      "learning_rate": 1.6567593196259084e-05,
+      "loss": 1.0235,
       "step": 10353
     },
     {
-      "epoch": 0.28436461508884675,
+      "epoch": 0.2938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.6794520368933377e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.6566900091514987e-05,
+      "loss": 0.8978,
       "step": 10354
     },
     {
-      "epoch": 0.28439207931669,
+      "epoch": 0.2938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.679386768643095e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.6566206931299987e-05,
+      "loss": 0.9675,
       "step": 10355
     },
     {
-      "epoch": 0.28441954354453325,
+      "epoch": 0.2938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.6793214950172807e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.6565513715619944e-05,
+      "loss": 0.9174,
       "step": 10356
     },
     {
-      "epoch": 0.28444700777237647,
+      "epoch": 0.2938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.6792562160164117e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.656482044448071e-05,
+      "loss": 0.9528,
       "step": 10357
     },
     {
-      "epoch": 0.2844744720002197,
+      "epoch": 0.29392735527809305,
       "grad_norm": 0.0,
-      "learning_rate": 1.6791909316410047e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.6564127117888148e-05,
+      "loss": 0.9989,
       "step": 10358
     },
     {
-      "epoch": 0.28450193622806297,
+      "epoch": 0.29395573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 1.6791256418915756e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.65634337358481e-05,
+      "loss": 0.927,
       "step": 10359
     },
     {
-      "epoch": 0.2845294004559062,
+      "epoch": 0.2939841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.6790603467686416e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.6562740298366437e-05,
+      "loss": 0.909,
       "step": 10360
     },
     {
-      "epoch": 0.2845568646837494,
+      "epoch": 0.29401248581157774,
       "grad_norm": 0.0,
-      "learning_rate": 1.678995046272719e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.6562046805449014e-05,
+      "loss": 0.992,
       "step": 10361
     },
     {
-      "epoch": 0.28458432891159263,
+      "epoch": 0.29404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.6789297404043244e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.6561353257101684e-05,
+      "loss": 0.8493,
       "step": 10362
     },
     {
-      "epoch": 0.2846117931394359,
+      "epoch": 0.2940692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.6788644291639747e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.656065965333031e-05,
+      "loss": 0.9653,
       "step": 10363
     },
     {
-      "epoch": 0.28463925736727913,
+      "epoch": 0.29409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.678799112552187e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.655996599414075e-05,
+      "loss": 0.8084,
       "step": 10364
     },
     {
-      "epoch": 0.28466672159512235,
+      "epoch": 0.29412599318955734,
       "grad_norm": 0.0,
-      "learning_rate": 1.6787337905694775e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.655927227953886e-05,
+      "loss": 0.9351,
       "step": 10365
     },
     {
-      "epoch": 0.2846941858229656,
+      "epoch": 0.2941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.6786684632163636e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.65585785095305e-05,
+      "loss": 1.0904,
       "step": 10366
     },
     {
-      "epoch": 0.2847216500508088,
+      "epoch": 0.2941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.6786031304933616e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.655788468412153e-05,
+      "loss": 0.896,
       "step": 10367
     },
     {
-      "epoch": 0.28474911427865207,
+      "epoch": 0.294211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.678537792400989e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.6557190803317818e-05,
+      "loss": 1.036,
       "step": 10368
     },
     {
-      "epoch": 0.2847765785064953,
+      "epoch": 0.2942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.6784724489397623e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.655649686712522e-05,
+      "loss": 0.9383,
       "step": 10369
     },
     {
-      "epoch": 0.2848040427343385,
+      "epoch": 0.2942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.678407100110199e-05,
-      "loss": 1.071,
+      "learning_rate": 1.65558028755496e-05,
+      "loss": 0.9862,
       "step": 10370
     },
     {
-      "epoch": 0.28483150696218174,
+      "epoch": 0.29429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.678341745912816e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.655510882859682e-05,
+      "loss": 0.953,
       "step": 10371
     },
     {
-      "epoch": 0.284858971190025,
+      "epoch": 0.2943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.67827638634813e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.6554414726272738e-05,
+      "loss": 0.8811,
       "step": 10372
     },
     {
-      "epoch": 0.28488643541786823,
+      "epoch": 0.29435300794551644,
       "grad_norm": 0.0,
-      "learning_rate": 1.678211021416659e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.655372056858322e-05,
+      "loss": 0.8859,
       "step": 10373
     },
     {
-      "epoch": 0.28491389964571145,
+      "epoch": 0.29438138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 1.6781456511189193e-05,
-      "loss": 1.0166,
+      "learning_rate": 1.655302635553413e-05,
+      "loss": 0.8231,
       "step": 10374
     },
     {
-      "epoch": 0.2849413638735547,
+      "epoch": 0.2944097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.6780802754554286e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.655233208713133e-05,
+      "loss": 1.0352,
       "step": 10375
     },
     {
-      "epoch": 0.28496882810139795,
+      "epoch": 0.29443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.678014894426704e-05,
-      "loss": 0.966,
+      "learning_rate": 1.655163776338069e-05,
+      "loss": 0.9446,
       "step": 10376
     },
     {
-      "epoch": 0.2849962923292412,
+      "epoch": 0.29446651532349605,
       "grad_norm": 0.0,
-      "learning_rate": 1.677949508033263e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.6550943384288066e-05,
+      "loss": 0.9072,
       "step": 10377
     },
     {
-      "epoch": 0.2850237565570844,
+      "epoch": 0.2944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.677884116275623e-05,
-      "loss": 1.1042,
+      "learning_rate": 1.6550248949859333e-05,
+      "loss": 0.9999,
       "step": 10378
     },
     {
-      "epoch": 0.2850512207849276,
+      "epoch": 0.2945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.6778187191543015e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.6549554460100354e-05,
+      "loss": 1.0473,
       "step": 10379
     },
     {
-      "epoch": 0.28507868501277084,
+      "epoch": 0.2945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.677753316669815e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.6548859915016992e-05,
+      "loss": 0.8761,
       "step": 10380
     },
     {
-      "epoch": 0.2851061492406141,
+      "epoch": 0.2945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.6776879088226824e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.6548165314615114e-05,
+      "loss": 0.9584,
       "step": 10381
     },
     {
-      "epoch": 0.28513361346845734,
+      "epoch": 0.2946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.6776224956134202e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.6547470658900592e-05,
+      "loss": 1.012,
       "step": 10382
     },
     {
-      "epoch": 0.28516107769630056,
+      "epoch": 0.29463677639046537,
       "grad_norm": 0.0,
-      "learning_rate": 1.6775570770425465e-05,
-      "loss": 0.9983,
+      "learning_rate": 1.6546775947879288e-05,
+      "loss": 1.0417,
       "step": 10383
     },
     {
-      "epoch": 0.2851885419241438,
+      "epoch": 0.2946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.6774916531105783e-05,
-      "loss": 0.9777,
+      "learning_rate": 1.6546081181557076e-05,
+      "loss": 0.9629,
       "step": 10384
     },
     {
-      "epoch": 0.28521600615198706,
+      "epoch": 0.29469353007945515,
       "grad_norm": 0.0,
-      "learning_rate": 1.6774262238180344e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.654538635993982e-05,
+      "loss": 0.9156,
       "step": 10385
     },
     {
-      "epoch": 0.2852434703798303,
+      "epoch": 0.29472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.6773607891654315e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.6544691483033392e-05,
+      "loss": 0.8688,
       "step": 10386
     },
     {
-      "epoch": 0.2852709346076735,
+      "epoch": 0.294750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.677295349153288e-05,
-      "loss": 1.0968,
+      "learning_rate": 1.654399655084366e-05,
+      "loss": 0.9031,
       "step": 10387
     },
     {
-      "epoch": 0.2852983988355167,
+      "epoch": 0.29477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.6772299037821206e-05,
-      "loss": 1.0647,
+      "learning_rate": 1.6543301563376498e-05,
+      "loss": 0.9926,
       "step": 10388
     },
     {
-      "epoch": 0.28532586306336,
+      "epoch": 0.29480703745743475,
       "grad_norm": 0.0,
-      "learning_rate": 1.6771644530524485e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.6542606520637772e-05,
+      "loss": 0.8979,
       "step": 10389
     },
     {
-      "epoch": 0.2853533272912032,
+      "epoch": 0.2948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.6770989969647887e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.6541911422633355e-05,
+      "loss": 1.0027,
       "step": 10390
     },
     {
-      "epoch": 0.28538079151904644,
+      "epoch": 0.2948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.6770335355196592e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.654121626936912e-05,
+      "loss": 0.9398,
       "step": 10391
     },
     {
-      "epoch": 0.28540825574688966,
+      "epoch": 0.2948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.6769680687175782e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.654052106085093e-05,
+      "loss": 0.9873,
       "step": 10392
     },
     {
-      "epoch": 0.28543571997473294,
+      "epoch": 0.2949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6769025965590638e-05,
-      "loss": 0.8591,
+      "learning_rate": 1.6539825797084674e-05,
+      "loss": 0.9636,
       "step": 10393
     },
     {
-      "epoch": 0.28546318420257616,
+      "epoch": 0.2949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.6768371190446337e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.6539130478076208e-05,
+      "loss": 1.0197,
       "step": 10394
     },
     {
-      "epoch": 0.2854906484304194,
+      "epoch": 0.2949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.676771636174806e-05,
-      "loss": 0.897,
+      "learning_rate": 1.6538435103831416e-05,
+      "loss": 0.9421,
       "step": 10395
     },
     {
-      "epoch": 0.2855181126582626,
+      "epoch": 0.295005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.676706147950099e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.653773967435617e-05,
+      "loss": 1.0065,
       "step": 10396
     },
     {
-      "epoch": 0.2855455768861058,
+      "epoch": 0.29503405221339385,
       "grad_norm": 0.0,
-      "learning_rate": 1.6766406543710312e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.6537044189656343e-05,
+      "loss": 0.8915,
       "step": 10397
     },
     {
-      "epoch": 0.2855730411139491,
+      "epoch": 0.29506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.67657515543812e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.6536348649737806e-05,
+      "loss": 0.9531,
       "step": 10398
     },
     {
-      "epoch": 0.2856005053417923,
+      "epoch": 0.2950908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.6765096511518842e-05,
-      "loss": 0.8389,
+      "learning_rate": 1.653565305460644e-05,
+      "loss": 0.9549,
       "step": 10399
     },
     {
-      "epoch": 0.28562796956963554,
+      "epoch": 0.29511918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.6764441415128422e-05,
-      "loss": 0.9469,
+      "learning_rate": 1.653495740426812e-05,
+      "loss": 1.0071,
       "step": 10400
     },
     {
-      "epoch": 0.28565543379747876,
+      "epoch": 0.29514755959137345,
       "grad_norm": 0.0,
-      "learning_rate": 1.676378626521512e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.653426169872872e-05,
+      "loss": 0.9521,
       "step": 10401
     },
     {
-      "epoch": 0.28568289802532204,
+      "epoch": 0.2951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.6763131061784118e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.6533565937994116e-05,
+      "loss": 0.9363,
       "step": 10402
     },
     {
-      "epoch": 0.28571036225316526,
+      "epoch": 0.29520431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 1.6762475804840606e-05,
-      "loss": 1.019,
+      "learning_rate": 1.6532870122070187e-05,
+      "loss": 1.0706,
       "step": 10403
     },
     {
-      "epoch": 0.2857378264810085,
+      "epoch": 0.2952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.6761820494389767e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.6532174250962814e-05,
+      "loss": 1.0808,
       "step": 10404
     },
     {
-      "epoch": 0.2857652907088517,
+      "epoch": 0.295261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.6761165130436784e-05,
-      "loss": 1.0482,
+      "learning_rate": 1.653147832467787e-05,
+      "loss": 1.0013,
       "step": 10405
     },
     {
-      "epoch": 0.285792754936695,
+      "epoch": 0.2952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.6760509712986844e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.6530782343221235e-05,
+      "loss": 0.9346,
       "step": 10406
     },
     {
-      "epoch": 0.2858202191645382,
+      "epoch": 0.2953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.675985424204513e-05,
-      "loss": 0.9887,
+      "learning_rate": 1.6530086306598786e-05,
+      "loss": 0.9448,
       "step": 10407
     },
     {
-      "epoch": 0.2858476833923814,
+      "epoch": 0.2953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.6759198717616835e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.652939021481641e-05,
+      "loss": 0.8943,
       "step": 10408
     },
     {
-      "epoch": 0.28587514762022465,
+      "epoch": 0.29537457434733255,
       "grad_norm": 0.0,
-      "learning_rate": 1.6758543139707137e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.6528694067879977e-05,
+      "loss": 1.0765,
       "step": 10409
     },
     {
-      "epoch": 0.28590261184806787,
+      "epoch": 0.29540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.6757887508321228e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.652799786579537e-05,
+      "loss": 0.9034,
       "step": 10410
     },
     {
-      "epoch": 0.28593007607591114,
+      "epoch": 0.2954313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.6757231823464297e-05,
-      "loss": 1.0254,
+      "learning_rate": 1.6527301608568476e-05,
+      "loss": 1.0261,
       "step": 10411
     },
     {
-      "epoch": 0.28595754030375437,
+      "epoch": 0.29545970488081724,
       "grad_norm": 0.0,
-      "learning_rate": 1.6756576085141532e-05,
-      "loss": 0.8468,
+      "learning_rate": 1.652660529620517e-05,
+      "loss": 0.9861,
       "step": 10412
     },
     {
-      "epoch": 0.2859850045315976,
+      "epoch": 0.29548808172531216,
       "grad_norm": 0.0,
-      "learning_rate": 1.6755920293358116e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.652590892871133e-05,
+      "loss": 1.0605,
       "step": 10413
     },
     {
-      "epoch": 0.2860124687594408,
+      "epoch": 0.295516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.6755264448119244e-05,
-      "loss": 0.947,
+      "learning_rate": 1.652521250609285e-05,
+      "loss": 1.0175,
       "step": 10414
     },
     {
-      "epoch": 0.2860399329872841,
+      "epoch": 0.29554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.6754608549430102e-05,
-      "loss": 0.8251,
+      "learning_rate": 1.6524516028355608e-05,
+      "loss": 0.9542,
       "step": 10415
     },
     {
-      "epoch": 0.2860673972151273,
+      "epoch": 0.29557321225879685,
       "grad_norm": 0.0,
-      "learning_rate": 1.6753952597295882e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.6523819495505483e-05,
+      "loss": 0.952,
       "step": 10416
     },
     {
-      "epoch": 0.2860948614429705,
+      "epoch": 0.2956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.675329659172177e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.652312290754836e-05,
+      "loss": 0.9928,
       "step": 10417
     },
     {
-      "epoch": 0.28612232567081375,
+      "epoch": 0.2956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.675264053271296e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.652242626449013e-05,
+      "loss": 0.9196,
       "step": 10418
     },
     {
-      "epoch": 0.286149789898657,
+      "epoch": 0.2956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.6751984420274646e-05,
-      "loss": 0.88,
+      "learning_rate": 1.6521729566336668e-05,
+      "loss": 0.897,
       "step": 10419
     },
     {
-      "epoch": 0.28617725412650025,
+      "epoch": 0.2956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.6751328254412014e-05,
-      "loss": 0.8665,
+      "learning_rate": 1.6521032813093866e-05,
+      "loss": 0.8728,
       "step": 10420
     },
     {
-      "epoch": 0.28620471835434347,
+      "epoch": 0.29571509648127126,
       "grad_norm": 0.0,
-      "learning_rate": 1.6750672035130257e-05,
-      "loss": 1.0534,
+      "learning_rate": 1.6520336004767603e-05,
+      "loss": 1.0795,
       "step": 10421
     },
     {
-      "epoch": 0.2862321825821867,
+      "epoch": 0.29574347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.6750015762434568e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.6519639141363768e-05,
+      "loss": 0.8537,
       "step": 10422
     },
     {
-      "epoch": 0.2862596468100299,
+      "epoch": 0.2957718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.674935943633014e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.6518942222888253e-05,
+      "loss": 0.9287,
       "step": 10423
     },
     {
-      "epoch": 0.2862871110378732,
+      "epoch": 0.29580022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.6748703056822164e-05,
-      "loss": 0.9755,
+      "learning_rate": 1.6518245249346937e-05,
+      "loss": 0.956,
       "step": 10424
     },
     {
-      "epoch": 0.2863145752657164,
+      "epoch": 0.29582860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 1.6748046623915838e-05,
-      "loss": 0.979,
+      "learning_rate": 1.651754822074571e-05,
+      "loss": 0.8387,
       "step": 10425
     },
     {
-      "epoch": 0.28634203949355963,
+      "epoch": 0.2958569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.6747390137616353e-05,
-      "loss": 0.9059,
+      "learning_rate": 1.651685113709046e-05,
+      "loss": 0.8933,
       "step": 10426
     },
     {
-      "epoch": 0.28636950372140285,
+      "epoch": 0.29588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.6746733597928903e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.6516153998387077e-05,
+      "loss": 0.9779,
       "step": 10427
     },
     {
-      "epoch": 0.28639696794924613,
+      "epoch": 0.29591373439273555,
       "grad_norm": 0.0,
-      "learning_rate": 1.6746077004858682e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.6515456804641446e-05,
+      "loss": 1.0585,
       "step": 10428
     },
     {
-      "epoch": 0.28642443217708935,
+      "epoch": 0.2959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.6745420358410887e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.651475955585946e-05,
+      "loss": 0.9295,
       "step": 10429
     },
     {
-      "epoch": 0.28645189640493257,
+      "epoch": 0.2959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.674476365859071e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.651406225204701e-05,
+      "loss": 0.9521,
       "step": 10430
     },
     {
-      "epoch": 0.2864793606327758,
+      "epoch": 0.2959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.6744106905403354e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.651336489320998e-05,
+      "loss": 0.9576,
       "step": 10431
     },
     {
-      "epoch": 0.28650682486061907,
+      "epoch": 0.2960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.6743450098854013e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.6512667479354263e-05,
+      "loss": 0.9755,
       "step": 10432
     },
     {
-      "epoch": 0.2865342890884623,
+      "epoch": 0.29605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.674279323894788e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.6511970010485755e-05,
+      "loss": 0.8557,
       "step": 10433
     },
     {
-      "epoch": 0.2865617533163055,
+      "epoch": 0.2960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.6742136325690154e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.651127248661034e-05,
+      "loss": 1.0968,
       "step": 10434
     },
     {
-      "epoch": 0.28658921754414873,
+      "epoch": 0.2961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.6741479359086036e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.6510574907733916e-05,
+      "loss": 0.9675,
       "step": 10435
     },
     {
-      "epoch": 0.28661668177199195,
+      "epoch": 0.29614074914869465,
       "grad_norm": 0.0,
-      "learning_rate": 1.6740822339140724e-05,
-      "loss": 0.851,
+      "learning_rate": 1.650987727386237e-05,
+      "loss": 0.9892,
       "step": 10436
     },
     {
-      "epoch": 0.28664414599983523,
+      "epoch": 0.29616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.674016526585941e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.65091795850016e-05,
+      "loss": 0.8173,
       "step": 10437
     },
     {
-      "epoch": 0.28667161022767845,
+      "epoch": 0.2961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.67395081392473e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.6508481841157497e-05,
+      "loss": 0.9106,
       "step": 10438
     },
     {
-      "epoch": 0.2866990744555217,
+      "epoch": 0.29622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.673885095930959e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.6507784042335958e-05,
+      "loss": 0.9489,
       "step": 10439
     },
     {
-      "epoch": 0.2867265386833649,
+      "epoch": 0.29625425652667425,
       "grad_norm": 0.0,
-      "learning_rate": 1.673819372605148e-05,
-      "loss": 1.0509,
+      "learning_rate": 1.6507086188542875e-05,
+      "loss": 1.0163,
       "step": 10440
     },
     {
-      "epoch": 0.2867540029112082,
+      "epoch": 0.2962826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.6737536439478175e-05,
-      "loss": 0.8582,
+      "learning_rate": 1.650638827978414e-05,
+      "loss": 0.9752,
       "step": 10441
     },
     {
-      "epoch": 0.2867814671390514,
+      "epoch": 0.29631101021566403,
       "grad_norm": 0.0,
-      "learning_rate": 1.673687909959487e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.6505690316065644e-05,
+      "loss": 0.917,
       "step": 10442
     },
     {
-      "epoch": 0.2868089313668946,
+      "epoch": 0.2963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.6736221706406767e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.65049922973933e-05,
+      "loss": 0.9523,
       "step": 10443
     },
     {
-      "epoch": 0.28683639559473784,
+      "epoch": 0.2963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.6735564259919068e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.6504294223772987e-05,
+      "loss": 0.9373,
       "step": 10444
     },
     {
-      "epoch": 0.2868638598225811,
+      "epoch": 0.2963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6734906760136977e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.650359609521061e-05,
+      "loss": 0.9915,
       "step": 10445
     },
     {
-      "epoch": 0.28689132405042433,
+      "epoch": 0.2964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.6734249207065695e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.6502897911712067e-05,
+      "loss": 0.8809,
       "step": 10446
     },
     {
-      "epoch": 0.28691878827826756,
+      "epoch": 0.2964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.6733591600710424e-05,
-      "loss": 1.034,
+      "learning_rate": 1.650219967328325e-05,
+      "loss": 1.0578,
       "step": 10447
     },
     {
-      "epoch": 0.2869462525061108,
+      "epoch": 0.29648127128263335,
       "grad_norm": 0.0,
-      "learning_rate": 1.6732933941076366e-05,
-      "loss": 0.8783,
+      "learning_rate": 1.6501501379930066e-05,
+      "loss": 1.0373,
       "step": 10448
     },
     {
-      "epoch": 0.286973716733954,
+      "epoch": 0.29650964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.6732276228168728e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.65008030316584e-05,
+      "loss": 0.9699,
       "step": 10449
     },
     {
-      "epoch": 0.2870011809617973,
+      "epoch": 0.2965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.6731618461992715e-05,
-      "loss": 0.9511,
+      "learning_rate": 1.6500104628474163e-05,
+      "loss": 1.0515,
       "step": 10450
     },
     {
-      "epoch": 0.2870286451896405,
+      "epoch": 0.29656640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.6730960642553526e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.649940617038325e-05,
+      "loss": 0.8617,
       "step": 10451
     },
     {
-      "epoch": 0.2870561094174837,
+      "epoch": 0.29659477866061296,
       "grad_norm": 0.0,
-      "learning_rate": 1.673030276985637e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.6498707657391558e-05,
+      "loss": 0.9484,
       "step": 10452
     },
     {
-      "epoch": 0.28708357364532694,
+      "epoch": 0.2966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.672964484390645e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.649800908950499e-05,
+      "loss": 0.9896,
       "step": 10453
     },
     {
-      "epoch": 0.2871110378731702,
+      "epoch": 0.29665153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.6728986864708977e-05,
-      "loss": 1.0539,
+      "learning_rate": 1.6497310466729448e-05,
+      "loss": 1.0672,
       "step": 10454
     },
     {
-      "epoch": 0.28713850210101344,
+      "epoch": 0.2966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.672832883226915e-05,
-      "loss": 1.0163,
+      "learning_rate": 1.6496611789070836e-05,
+      "loss": 1.0078,
       "step": 10455
     },
     {
-      "epoch": 0.28716596632885666,
+      "epoch": 0.2967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.672767074659218e-05,
-      "loss": 0.922,
+      "learning_rate": 1.649591305653505e-05,
+      "loss": 0.9779,
       "step": 10456
     },
     {
-      "epoch": 0.2871934305566999,
+      "epoch": 0.2967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.6727012607683277e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.649521426912799e-05,
+      "loss": 1.0903,
       "step": 10457
     },
     {
-      "epoch": 0.28722089478454316,
+      "epoch": 0.2967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.6726354415547636e-05,
-      "loss": 1.047,
+      "learning_rate": 1.649451542685557e-05,
+      "loss": 0.9252,
       "step": 10458
     },
     {
-      "epoch": 0.2872483590123864,
+      "epoch": 0.2967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.672569617019048e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.6493816529723683e-05,
+      "loss": 0.9895,
       "step": 10459
     },
     {
-      "epoch": 0.2872758232402296,
+      "epoch": 0.29682179341657206,
       "grad_norm": 0.0,
-      "learning_rate": 1.6725037871617006e-05,
-      "loss": 1.0282,
+      "learning_rate": 1.6493117577738235e-05,
+      "loss": 1.0348,
       "step": 10460
     },
     {
-      "epoch": 0.2873032874680728,
+      "epoch": 0.29685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.672437951983243e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.6492418570905134e-05,
+      "loss": 1.0025,
       "step": 10461
     },
     {
-      "epoch": 0.28733075169591604,
+      "epoch": 0.2968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.6723721114841958e-05,
-      "loss": 1.0505,
+      "learning_rate": 1.649171950923028e-05,
+      "loss": 0.9625,
       "step": 10462
     },
     {
-      "epoch": 0.2873582159237593,
+      "epoch": 0.29690692395005674,
       "grad_norm": 0.0,
-      "learning_rate": 1.6723062656650802e-05,
-      "loss": 0.865,
+      "learning_rate": 1.6491020392719575e-05,
+      "loss": 0.9555,
       "step": 10463
     },
     {
-      "epoch": 0.28738568015160254,
+      "epoch": 0.29693530079455166,
       "grad_norm": 0.0,
-      "learning_rate": 1.6722404145264167e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.6490321221378933e-05,
+      "loss": 0.9288,
       "step": 10464
     },
     {
-      "epoch": 0.28741314437944576,
+      "epoch": 0.2969636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.6721745580687268e-05,
-      "loss": 0.951,
+      "learning_rate": 1.648962199521426e-05,
+      "loss": 0.9069,
       "step": 10465
     },
     {
-      "epoch": 0.287440608607289,
+      "epoch": 0.29699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6721086962925316e-05,
-      "loss": 0.9255,
+      "learning_rate": 1.6488922714231453e-05,
+      "loss": 1.0653,
       "step": 10466
     },
     {
-      "epoch": 0.28746807283513226,
+      "epoch": 0.29702043132803635,
       "grad_norm": 0.0,
-      "learning_rate": 1.6720428291983516e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.6488223378436423e-05,
+      "loss": 0.979,
       "step": 10467
     },
     {
-      "epoch": 0.2874955370629755,
+      "epoch": 0.2970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.6719769567867087e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.6487523987835082e-05,
+      "loss": 0.8968,
       "step": 10468
     },
     {
-      "epoch": 0.2875230012908187,
+      "epoch": 0.2970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.671911079058124e-05,
-      "loss": 0.9701,
+      "learning_rate": 1.648682454243333e-05,
+      "loss": 1.0175,
       "step": 10469
     },
     {
-      "epoch": 0.2875504655186619,
+      "epoch": 0.297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.671845196013118e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.6486125042237084e-05,
+      "loss": 1.1928,
       "step": 10470
     },
     {
-      "epoch": 0.2875779297465052,
+      "epoch": 0.2971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.671779307652213e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.6485425487252245e-05,
+      "loss": 1.0895,
       "step": 10471
     },
     {
-      "epoch": 0.2876053939743484,
+      "epoch": 0.29716231555051076,
       "grad_norm": 0.0,
-      "learning_rate": 1.67171341397593e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.648472587748473e-05,
+      "loss": 0.8467,
       "step": 10472
     },
     {
-      "epoch": 0.28763285820219164,
+      "epoch": 0.2971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.67164751498479e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.6484026212940442e-05,
+      "loss": 0.8208,
       "step": 10473
     },
     {
-      "epoch": 0.28766032243003486,
+      "epoch": 0.2972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.6715816106793145e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.6483326493625292e-05,
+      "loss": 0.9478,
       "step": 10474
     },
     {
-      "epoch": 0.2876877866578781,
+      "epoch": 0.29724744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 1.6715157010600256e-05,
-      "loss": 1.0998,
+      "learning_rate": 1.648262671954519e-05,
+      "loss": 1.0533,
       "step": 10475
     },
     {
-      "epoch": 0.28771525088572136,
+      "epoch": 0.29727582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.671449786127444e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.648192689070605e-05,
+      "loss": 0.9829,
       "step": 10476
     },
     {
-      "epoch": 0.2877427151135646,
+      "epoch": 0.2973041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.671383865882092e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.6481227007113784e-05,
+      "loss": 0.8857,
       "step": 10477
     },
     {
-      "epoch": 0.2877701793414078,
+      "epoch": 0.29733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.6713179403244905e-05,
-      "loss": 0.903,
+      "learning_rate": 1.6480527068774298e-05,
+      "loss": 0.9582,
       "step": 10478
     },
     {
-      "epoch": 0.287797643569251,
+      "epoch": 0.29736095346197505,
       "grad_norm": 0.0,
-      "learning_rate": 1.6712520094551616e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.6479827075693512e-05,
+      "loss": 1.0173,
       "step": 10479
     },
     {
-      "epoch": 0.2878251077970943,
+      "epoch": 0.2973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.6711860732746267e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.6479127027877335e-05,
+      "loss": 0.9938,
       "step": 10480
     },
     {
-      "epoch": 0.2878525720249375,
+      "epoch": 0.2974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.6711201317834073e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.647842692533168e-05,
+      "loss": 0.9803,
       "step": 10481
     },
     {
-      "epoch": 0.28788003625278075,
+      "epoch": 0.2974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.6710541849820256e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.647772676806246e-05,
+      "loss": 1.0032,
       "step": 10482
     },
     {
-      "epoch": 0.28790750048062397,
+      "epoch": 0.2974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6709882328710036e-05,
-      "loss": 1.0228,
+      "learning_rate": 1.647702655607559e-05,
+      "loss": 1.0016,
       "step": 10483
     },
     {
-      "epoch": 0.28793496470846724,
+      "epoch": 0.29750283768444946,
       "grad_norm": 0.0,
-      "learning_rate": 1.6709222754508624e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.647632628937699e-05,
+      "loss": 0.8601,
       "step": 10484
     },
     {
-      "epoch": 0.28796242893631047,
+      "epoch": 0.2975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.6708563127221245e-05,
-      "loss": 0.9029,
+      "learning_rate": 1.6475625967972573e-05,
+      "loss": 0.9631,
       "step": 10485
     },
     {
-      "epoch": 0.2879898931641537,
+      "epoch": 0.2975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.6707903446853114e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.647492559186825e-05,
+      "loss": 0.9857,
       "step": 10486
     },
     {
-      "epoch": 0.2880173573919969,
+      "epoch": 0.29758796821793415,
       "grad_norm": 0.0,
-      "learning_rate": 1.670724371340945e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.6474225161069936e-05,
+      "loss": 0.9233,
       "step": 10487
     },
     {
-      "epoch": 0.2880448216198402,
+      "epoch": 0.29761634506242907,
       "grad_norm": 0.0,
-      "learning_rate": 1.670658392689548e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.647352467558355e-05,
+      "loss": 0.8458,
       "step": 10488
     },
     {
-      "epoch": 0.2880722858476834,
+      "epoch": 0.2976447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 1.670592408731642e-05,
-      "loss": 0.938,
+      "learning_rate": 1.6472824135415013e-05,
+      "loss": 0.98,
       "step": 10489
     },
     {
-      "epoch": 0.28809975007552663,
+      "epoch": 0.29767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.6705264194677488e-05,
-      "loss": 0.9726,
+      "learning_rate": 1.647212354057024e-05,
+      "loss": 0.9778,
       "step": 10490
     },
     {
-      "epoch": 0.28812721430336985,
+      "epoch": 0.29770147559591376,
       "grad_norm": 0.0,
-      "learning_rate": 1.6704604248983912e-05,
-      "loss": 1.0442,
+      "learning_rate": 1.647142289105515e-05,
+      "loss": 0.9829,
       "step": 10491
     },
     {
-      "epoch": 0.28815467853121307,
+      "epoch": 0.2977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.6703944250240907e-05,
-      "loss": 0.9201,
+      "learning_rate": 1.6470722186875656e-05,
+      "loss": 0.945,
       "step": 10492
     },
     {
-      "epoch": 0.28818214275905635,
+      "epoch": 0.29775822928490353,
       "grad_norm": 0.0,
-      "learning_rate": 1.67032841984537e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.6470021428037682e-05,
+      "loss": 1.0518,
       "step": 10493
     },
     {
-      "epoch": 0.28820960698689957,
+      "epoch": 0.2977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.670262409362751e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.646932061454715e-05,
+      "loss": 0.9154,
       "step": 10494
     },
     {
-      "epoch": 0.2882370712147428,
+      "epoch": 0.2978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.6701963935767566e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.6468619746409972e-05,
+      "loss": 0.9088,
       "step": 10495
     },
     {
-      "epoch": 0.288264535442586,
+      "epoch": 0.2978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.670130372487908e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.646791882363207e-05,
+      "loss": 0.9492,
       "step": 10496
     },
     {
-      "epoch": 0.2882919996704293,
+      "epoch": 0.2978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.670064346096729e-05,
-      "loss": 0.9423,
+      "learning_rate": 1.6467217846219372e-05,
+      "loss": 1.0526,
       "step": 10497
     },
     {
-      "epoch": 0.2883194638982725,
+      "epoch": 0.297900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.669998314403741e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.6466516814177792e-05,
+      "loss": 0.8661,
       "step": 10498
     },
     {
-      "epoch": 0.28834692812611573,
+      "epoch": 0.29792849035187285,
       "grad_norm": 0.0,
-      "learning_rate": 1.6699322774094667e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.6465815727513253e-05,
+      "loss": 1.0073,
       "step": 10499
     },
     {
-      "epoch": 0.28837439235395895,
+      "epoch": 0.29795686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 1.6698662351144294e-05,
-      "loss": 0.9101,
+      "learning_rate": 1.646511458623168e-05,
+      "loss": 0.8922,
       "step": 10500
     },
     {
-      "epoch": 0.28840185658180223,
+      "epoch": 0.29798524404086263,
       "grad_norm": 0.0,
-      "learning_rate": 1.66980018751915e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.6464413390338993e-05,
+      "loss": 0.982,
       "step": 10501
     },
     {
-      "epoch": 0.28842932080964545,
+      "epoch": 0.29801362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.6697341346241528e-05,
-      "loss": 0.9887,
+      "learning_rate": 1.6463712139841114e-05,
+      "loss": 0.9609,
       "step": 10502
     },
     {
-      "epoch": 0.28845678503748867,
+      "epoch": 0.29804199772985246,
       "grad_norm": 0.0,
-      "learning_rate": 1.6696680764299595e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.6463010834743965e-05,
+      "loss": 0.8873,
       "step": 10503
     },
     {
-      "epoch": 0.2884842492653319,
+      "epoch": 0.2980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.6696020129370925e-05,
-      "loss": 1.0283,
+      "learning_rate": 1.646230947505348e-05,
+      "loss": 1.0826,
       "step": 10504
     },
     {
-      "epoch": 0.2885117134931751,
+      "epoch": 0.29809875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.6695359441460755e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.646160806077557e-05,
+      "loss": 0.9301,
       "step": 10505
     },
     {
-      "epoch": 0.2885391777210184,
+      "epoch": 0.2981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.669469870057431e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.646090659191617e-05,
+      "loss": 0.9974,
       "step": 10506
     },
     {
-      "epoch": 0.2885666419488616,
+      "epoch": 0.298155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.669403790671681e-05,
-      "loss": 1.0594,
+      "learning_rate": 1.64602050684812e-05,
+      "loss": 0.9464,
       "step": 10507
     },
     {
-      "epoch": 0.28859410617670483,
+      "epoch": 0.2981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.6693377059893492e-05,
-      "loss": 1.091,
+      "learning_rate": 1.645950349047659e-05,
+      "loss": 0.8804,
       "step": 10508
     },
     {
-      "epoch": 0.28862157040454806,
+      "epoch": 0.2982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.669271616010958e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.645880185790826e-05,
+      "loss": 0.906,
       "step": 10509
     },
     {
-      "epoch": 0.28864903463239133,
+      "epoch": 0.2982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.6692055207370307e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.6458100170782145e-05,
+      "loss": 0.9689,
       "step": 10510
     },
     {
-      "epoch": 0.28867649886023455,
+      "epoch": 0.29826901248581156,
       "grad_norm": 0.0,
-      "learning_rate": 1.66913942016809e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.645739842910416e-05,
+      "loss": 0.9296,
       "step": 10511
     },
     {
-      "epoch": 0.2887039630880778,
+      "epoch": 0.2982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.6690733143046588e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.645669663288025e-05,
+      "loss": 1.0331,
       "step": 10512
     },
     {
-      "epoch": 0.288731427315921,
+      "epoch": 0.2983257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.669007203147261e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.6455994782116325e-05,
+      "loss": 1.0778,
       "step": 10513
     },
     {
-      "epoch": 0.2887588915437643,
+      "epoch": 0.29835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.6689410866964185e-05,
-      "loss": 0.894,
+      "learning_rate": 1.6455292876818325e-05,
+      "loss": 1.0017,
       "step": 10514
     },
     {
-      "epoch": 0.2887863557716075,
+      "epoch": 0.29838251986379116,
       "grad_norm": 0.0,
-      "learning_rate": 1.6688749649526554e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.6454590916992176e-05,
+      "loss": 0.9957,
       "step": 10515
     },
     {
-      "epoch": 0.2888138199994507,
+      "epoch": 0.298410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.668808837916494e-05,
-      "loss": 1.0047,
+      "learning_rate": 1.6453888902643807e-05,
+      "loss": 1.0057,
       "step": 10516
     },
     {
-      "epoch": 0.28884128422729394,
+      "epoch": 0.29843927355278094,
       "grad_norm": 0.0,
-      "learning_rate": 1.668742705588458e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.645318683377915e-05,
+      "loss": 1.0489,
       "step": 10517
     },
     {
-      "epoch": 0.28886874845513716,
+      "epoch": 0.2984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.668676567969071e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.6452484710404132e-05,
+      "loss": 0.9353,
       "step": 10518
     },
     {
-      "epoch": 0.28889621268298044,
+      "epoch": 0.2984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.668610425058856e-05,
-      "loss": 0.9766,
+      "learning_rate": 1.6451782532524686e-05,
+      "loss": 0.9721,
       "step": 10519
     },
     {
-      "epoch": 0.28892367691082366,
+      "epoch": 0.2985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.668544276858336e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.6451080300146746e-05,
+      "loss": 0.9506,
       "step": 10520
     },
     {
-      "epoch": 0.2889511411386669,
+      "epoch": 0.2985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.6684781233680352e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.6450378013276233e-05,
+      "loss": 0.7543,
       "step": 10521
     },
     {
-      "epoch": 0.2889786053665101,
+      "epoch": 0.2985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.668411964588476e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.644967567191909e-05,
+      "loss": 0.9481,
       "step": 10522
     },
     {
-      "epoch": 0.2890060695943534,
+      "epoch": 0.29860953461975026,
       "grad_norm": 0.0,
-      "learning_rate": 1.6683458005201828e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.6448973276081252e-05,
+      "loss": 0.8884,
       "step": 10523
     },
     {
-      "epoch": 0.2890335338221966,
+      "epoch": 0.2986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.668279631163679e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.6448270825768644e-05,
+      "loss": 0.9091,
       "step": 10524
     },
     {
-      "epoch": 0.2890609980500398,
+      "epoch": 0.2986662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.668213456519487e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.64475683209872e-05,
+      "loss": 1.0452,
       "step": 10525
     },
     {
-      "epoch": 0.28908846227788304,
+      "epoch": 0.29869466515323495,
       "grad_norm": 0.0,
-      "learning_rate": 1.668147276588132e-05,
-      "loss": 0.846,
+      "learning_rate": 1.644686576174286e-05,
+      "loss": 0.9901,
       "step": 10526
     },
     {
-      "epoch": 0.2891159265057263,
+      "epoch": 0.29872304199772987,
       "grad_norm": 0.0,
-      "learning_rate": 1.6680810913701368e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.644616314804155e-05,
+      "loss": 0.9002,
       "step": 10527
     },
     {
-      "epoch": 0.28914339073356954,
+      "epoch": 0.2987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.668014900866025e-05,
-      "loss": 1.0952,
+      "learning_rate": 1.6445460479889214e-05,
+      "loss": 0.9273,
       "step": 10528
     },
     {
-      "epoch": 0.28917085496141276,
+      "epoch": 0.29877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.667948705076321e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.6444757757291783e-05,
+      "loss": 0.8894,
       "step": 10529
     },
     {
-      "epoch": 0.289198319189256,
+      "epoch": 0.29880817253121456,
       "grad_norm": 0.0,
-      "learning_rate": 1.6678825040015475e-05,
-      "loss": 0.941,
+      "learning_rate": 1.6444054980255192e-05,
+      "loss": 0.9747,
       "step": 10530
     },
     {
-      "epoch": 0.2892257834170992,
+      "epoch": 0.2988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.6678162976422294e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.644335214878538e-05,
+      "loss": 0.8592,
       "step": 10531
     },
     {
-      "epoch": 0.2892532476449425,
+      "epoch": 0.29886492622020433,
       "grad_norm": 0.0,
-      "learning_rate": 1.6677500859988896e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.644264926288828e-05,
+      "loss": 0.8994,
       "step": 10532
     },
     {
-      "epoch": 0.2892807118727857,
+      "epoch": 0.2988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.6676838690720528e-05,
-      "loss": 1.0183,
+      "learning_rate": 1.644194632256984e-05,
+      "loss": 0.9432,
       "step": 10533
     },
     {
-      "epoch": 0.2893081761006289,
+      "epoch": 0.2989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.6676176468622424e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.6441243327835983e-05,
+      "loss": 1.0289,
       "step": 10534
     },
     {
-      "epoch": 0.28933564032847214,
+      "epoch": 0.29895005675368896,
       "grad_norm": 0.0,
-      "learning_rate": 1.6675514193699828e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.6440540278692656e-05,
+      "loss": 0.8501,
       "step": 10535
     },
     {
-      "epoch": 0.2893631045563154,
+      "epoch": 0.2989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.667485186595798e-05,
-      "loss": 0.9244,
+      "learning_rate": 1.6439837175145793e-05,
+      "loss": 0.9233,
       "step": 10536
     },
     {
-      "epoch": 0.28939056878415864,
+      "epoch": 0.2990068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.6674189485402114e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.643913401720134e-05,
+      "loss": 0.9986,
       "step": 10537
     },
     {
-      "epoch": 0.28941803301200186,
+      "epoch": 0.29903518728717365,
       "grad_norm": 0.0,
-      "learning_rate": 1.667352705203748e-05,
-      "loss": 0.9378,
+      "learning_rate": 1.6438430804865232e-05,
+      "loss": 0.9184,
       "step": 10538
     },
     {
-      "epoch": 0.2894454972398451,
+      "epoch": 0.29906356413166857,
       "grad_norm": 0.0,
-      "learning_rate": 1.6672864565869308e-05,
-      "loss": 1.0072,
+      "learning_rate": 1.643772753814341e-05,
+      "loss": 0.7796,
       "step": 10539
     },
     {
-      "epoch": 0.28947296146768836,
+      "epoch": 0.29909194097616343,
       "grad_norm": 0.0,
-      "learning_rate": 1.6672202026902852e-05,
-      "loss": 1.0346,
+      "learning_rate": 1.6437024217041813e-05,
+      "loss": 0.8943,
       "step": 10540
     },
     {
-      "epoch": 0.2895004256955316,
+      "epoch": 0.29912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.6671539435143344e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.643632084156638e-05,
+      "loss": 0.9514,
       "step": 10541
     },
     {
-      "epoch": 0.2895278899233748,
+      "epoch": 0.29914869466515326,
       "grad_norm": 0.0,
-      "learning_rate": 1.6670876790596034e-05,
-      "loss": 1.0028,
+      "learning_rate": 1.643561741172306e-05,
+      "loss": 0.9556,
       "step": 10542
     },
     {
-      "epoch": 0.289555354151218,
+      "epoch": 0.2991770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.6670214093266164e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.6434913927517787e-05,
+      "loss": 1.0236,
       "step": 10543
     },
     {
-      "epoch": 0.28958281837906125,
+      "epoch": 0.29920544835414303,
       "grad_norm": 0.0,
-      "learning_rate": 1.666955134315898e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.643421038895651e-05,
+      "loss": 0.9366,
       "step": 10544
     },
     {
-      "epoch": 0.2896102826069045,
+      "epoch": 0.2992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.6668888540279714e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.6433506796045166e-05,
+      "loss": 0.9635,
       "step": 10545
     },
     {
-      "epoch": 0.28963774683474774,
+      "epoch": 0.2992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.6668225684633627e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.6432803148789704e-05,
+      "loss": 0.9949,
       "step": 10546
     },
     {
-      "epoch": 0.28966521106259097,
+      "epoch": 0.2992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.666756277622595e-05,
-      "loss": 0.995,
+      "learning_rate": 1.643209944719606e-05,
+      "loss": 0.9774,
       "step": 10547
     },
     {
-      "epoch": 0.2896926752904342,
+      "epoch": 0.2993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.666689981506193e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.6431395691270188e-05,
+      "loss": 0.9965,
       "step": 10548
     },
     {
-      "epoch": 0.28972013951827746,
+      "epoch": 0.2993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.6666236801146826e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.6430691881018024e-05,
+      "loss": 0.9183,
       "step": 10549
     },
     {
-      "epoch": 0.2897476037461207,
+      "epoch": 0.29937570942111236,
       "grad_norm": 0.0,
-      "learning_rate": 1.6665573734485868e-05,
-      "loss": 1.0808,
+      "learning_rate": 1.6429988016445518e-05,
+      "loss": 0.9218,
       "step": 10550
     },
     {
-      "epoch": 0.2897750679739639,
+      "epoch": 0.29940408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.666491061508431e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.6429284097558614e-05,
+      "loss": 0.8699,
       "step": 10551
     },
     {
-      "epoch": 0.28980253220180713,
+      "epoch": 0.29943246311010213,
       "grad_norm": 0.0,
-      "learning_rate": 1.6664247442947393e-05,
-      "loss": 1.0711,
+      "learning_rate": 1.6428580124363257e-05,
+      "loss": 0.9561,
       "step": 10552
     },
     {
-      "epoch": 0.2898299964296504,
+      "epoch": 0.29946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.6663584218080376e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.6427876096865394e-05,
+      "loss": 0.9237,
       "step": 10553
     },
     {
-      "epoch": 0.2898574606574936,
+      "epoch": 0.29948921679909196,
       "grad_norm": 0.0,
-      "learning_rate": 1.6662920940488495e-05,
-      "loss": 0.9516,
+      "learning_rate": 1.6427172015070977e-05,
+      "loss": 0.9894,
       "step": 10554
     },
     {
-      "epoch": 0.28988492488533685,
+      "epoch": 0.2995175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.6662257610177003e-05,
-      "loss": 1.0129,
+      "learning_rate": 1.6426467878985946e-05,
+      "loss": 0.8495,
       "step": 10555
     },
     {
-      "epoch": 0.28991238911318007,
+      "epoch": 0.29954597048808174,
       "grad_norm": 0.0,
-      "learning_rate": 1.666159422715115e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.642576368861625e-05,
+      "loss": 0.943,
       "step": 10556
     },
     {
-      "epoch": 0.2899398533410233,
+      "epoch": 0.2995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.6660930791416185e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.6425059443967844e-05,
+      "loss": 0.9335,
       "step": 10557
     },
     {
-      "epoch": 0.28996731756886657,
+      "epoch": 0.2996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.666026730297735e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.642435514504667e-05,
+      "loss": 0.9211,
       "step": 10558
     },
     {
-      "epoch": 0.2899947817967098,
+      "epoch": 0.2996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6659603761839905e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.6423650791858678e-05,
+      "loss": 0.979,
       "step": 10559
     },
     {
-      "epoch": 0.290022246024553,
+      "epoch": 0.2996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.665894016800909e-05,
-      "loss": 0.9365,
+      "learning_rate": 1.6422946384409817e-05,
+      "loss": 0.933,
       "step": 10560
     },
     {
-      "epoch": 0.29004971025239623,
+      "epoch": 0.2996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.6658276521490163e-05,
-      "loss": 0.904,
+      "learning_rate": 1.6422241922706043e-05,
+      "loss": 0.9383,
       "step": 10561
     },
     {
-      "epoch": 0.2900771744802395,
+      "epoch": 0.29971623155505106,
       "grad_norm": 0.0,
-      "learning_rate": 1.6657612822288375e-05,
-      "loss": 1.101,
+      "learning_rate": 1.6421537406753302e-05,
+      "loss": 0.9873,
       "step": 10562
     },
     {
-      "epoch": 0.29010463870808273,
+      "epoch": 0.299744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6656949070408974e-05,
-      "loss": 0.975,
+      "learning_rate": 1.6420832836557546e-05,
+      "loss": 0.9335,
       "step": 10563
     },
     {
-      "epoch": 0.29013210293592595,
+      "epoch": 0.29977298524404083,
       "grad_norm": 0.0,
-      "learning_rate": 1.6656285265857212e-05,
-      "loss": 0.966,
+      "learning_rate": 1.6420128212124726e-05,
+      "loss": 1.1077,
       "step": 10564
     },
     {
-      "epoch": 0.29015956716376917,
+      "epoch": 0.29980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.6655621408638348e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.6419423533460795e-05,
+      "loss": 1.0472,
       "step": 10565
     },
     {
-      "epoch": 0.29018703139161245,
+      "epoch": 0.29982973893303067,
       "grad_norm": 0.0,
-      "learning_rate": 1.6654957498757626e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.6418718800571705e-05,
+      "loss": 0.9636,
       "step": 10566
     },
     {
-      "epoch": 0.29021449561945567,
+      "epoch": 0.2998581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.66542935362203e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.641801401346341e-05,
+      "loss": 1.0147,
       "step": 10567
     },
     {
-      "epoch": 0.2902419598472989,
+      "epoch": 0.29988649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 1.6653629521031628e-05,
-      "loss": 1.0449,
+      "learning_rate": 1.6417309172141856e-05,
+      "loss": 0.9598,
       "step": 10568
     },
     {
-      "epoch": 0.2902694240751421,
+      "epoch": 0.2999148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.665296545319686e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.6416604276613007e-05,
+      "loss": 0.9754,
       "step": 10569
     },
     {
-      "epoch": 0.29029688830298533,
+      "epoch": 0.2999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.6652301332721256e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.6415899326882815e-05,
+      "loss": 0.9392,
       "step": 10570
     },
     {
-      "epoch": 0.2903243525308286,
+      "epoch": 0.29997162315550513,
       "grad_norm": 0.0,
-      "learning_rate": 1.6651637159610063e-05,
-      "loss": 0.886,
+      "learning_rate": 1.6415194322957232e-05,
+      "loss": 0.8689,
       "step": 10571
     },
     {
-      "epoch": 0.29035181675867183,
+      "epoch": 0.3,
       "grad_norm": 0.0,
-      "learning_rate": 1.6650972933868543e-05,
-      "loss": 0.921,
+      "learning_rate": 1.6414489264842212e-05,
+      "loss": 0.9235,
       "step": 10572
     },
     {
-      "epoch": 0.29037928098651505,
+      "epoch": 0.3000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.6650308655501946e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.6413784152543713e-05,
+      "loss": 1.0757,
       "step": 10573
     },
     {
-      "epoch": 0.2904067452143583,
+      "epoch": 0.30005675368898976,
       "grad_norm": 0.0,
-      "learning_rate": 1.664964432451553e-05,
-      "loss": 0.9423,
+      "learning_rate": 1.6413078986067693e-05,
+      "loss": 0.8762,
       "step": 10574
     },
     {
-      "epoch": 0.29043420944220155,
+      "epoch": 0.3000851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.6648979940914557e-05,
-      "loss": 1.0771,
+      "learning_rate": 1.6412373765420105e-05,
+      "loss": 0.8889,
       "step": 10575
     },
     {
-      "epoch": 0.2904616736700448,
+      "epoch": 0.3001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.664831550470428e-05,
-      "loss": 1.0082,
+      "learning_rate": 1.6411668490606907e-05,
+      "loss": 1.011,
       "step": 10576
     },
     {
-      "epoch": 0.290489137897888,
+      "epoch": 0.30014188422247445,
       "grad_norm": 0.0,
-      "learning_rate": 1.664765101588995e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.641096316163406e-05,
+      "loss": 0.9703,
       "step": 10577
     },
     {
-      "epoch": 0.2905166021257312,
+      "epoch": 0.30017026106696937,
       "grad_norm": 0.0,
-      "learning_rate": 1.6646986474476835e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.6410257778507518e-05,
+      "loss": 0.9016,
       "step": 10578
     },
     {
-      "epoch": 0.2905440663535745,
+      "epoch": 0.30019863791146423,
       "grad_norm": 0.0,
-      "learning_rate": 1.6646321880470184e-05,
-      "loss": 1.02,
+      "learning_rate": 1.640955234123324e-05,
+      "loss": 0.8976,
       "step": 10579
     },
     {
-      "epoch": 0.2905715305814177,
+      "epoch": 0.30022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.6645657233875264e-05,
-      "loss": 1.0405,
+      "learning_rate": 1.6408846849817182e-05,
+      "loss": 0.9744,
       "step": 10580
     },
     {
-      "epoch": 0.29059899480926094,
+      "epoch": 0.300255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.6644992534697326e-05,
-      "loss": 0.9678,
+      "learning_rate": 1.6408141304265312e-05,
+      "loss": 0.8988,
       "step": 10581
     },
     {
-      "epoch": 0.29062645903710416,
+      "epoch": 0.3002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.6644327782941633e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.6407435704583582e-05,
+      "loss": 0.9768,
       "step": 10582
     },
     {
-      "epoch": 0.29065392326494743,
+      "epoch": 0.30031214528944383,
       "grad_norm": 0.0,
-      "learning_rate": 1.6643662978613448e-05,
-      "loss": 0.9796,
+      "learning_rate": 1.6406730050777954e-05,
+      "loss": 1.0469,
       "step": 10583
     },
     {
-      "epoch": 0.29068138749279065,
+      "epoch": 0.3003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.6642998121718026e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.640602434285439e-05,
+      "loss": 0.9585,
       "step": 10584
     },
     {
-      "epoch": 0.2907088517206339,
+      "epoch": 0.3003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.6642333212260627e-05,
-      "loss": 0.8537,
+      "learning_rate": 1.640531858081885e-05,
+      "loss": 0.8976,
       "step": 10585
     },
     {
-      "epoch": 0.2907363159484771,
+      "epoch": 0.30039727582292847,
       "grad_norm": 0.0,
-      "learning_rate": 1.664166825024652e-05,
-      "loss": 0.856,
+      "learning_rate": 1.6404612764677294e-05,
+      "loss": 0.9512,
       "step": 10586
     },
     {
-      "epoch": 0.2907637801763203,
+      "epoch": 0.3004256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.6641003235680958e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.640390689443569e-05,
+      "loss": 0.9843,
       "step": 10587
     },
     {
-      "epoch": 0.2907912444041636,
+      "epoch": 0.3004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.6640338168569207e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.6403200970099996e-05,
+      "loss": 0.8703,
       "step": 10588
     },
     {
-      "epoch": 0.2908187086320068,
+      "epoch": 0.30048240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.6639673048916528e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.6402494991676174e-05,
+      "loss": 1.1187,
       "step": 10589
     },
     {
-      "epoch": 0.29084617285985004,
+      "epoch": 0.30051078320090807,
       "grad_norm": 0.0,
-      "learning_rate": 1.663900787672818e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.640178895917019e-05,
+      "loss": 0.9835,
       "step": 10590
     },
     {
-      "epoch": 0.29087363708769326,
+      "epoch": 0.30053916004540293,
       "grad_norm": 0.0,
-      "learning_rate": 1.6638342652009437e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.6401082872588004e-05,
+      "loss": 0.9063,
       "step": 10591
     },
     {
-      "epoch": 0.29090110131553654,
+      "epoch": 0.30056753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.6637677374765552e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.6400376731935587e-05,
+      "loss": 0.9394,
       "step": 10592
     },
     {
-      "epoch": 0.29092856554337976,
+      "epoch": 0.30059591373439276,
       "grad_norm": 0.0,
-      "learning_rate": 1.663701204500179e-05,
-      "loss": 1.1329,
+      "learning_rate": 1.6399670537218896e-05,
+      "loss": 0.9068,
       "step": 10593
     },
     {
-      "epoch": 0.290956029771223,
+      "epoch": 0.3006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.663634666272342e-05,
-      "loss": 0.9018,
+      "learning_rate": 1.6398964288443902e-05,
+      "loss": 0.9621,
       "step": 10594
     },
     {
-      "epoch": 0.2909834939990662,
+      "epoch": 0.30065266742338254,
       "grad_norm": 0.0,
-      "learning_rate": 1.6635681227935704e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.639825798561657e-05,
+      "loss": 0.9208,
       "step": 10595
     },
     {
-      "epoch": 0.2910109582269095,
+      "epoch": 0.3006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.663501574064391e-05,
-      "loss": 0.963,
+      "learning_rate": 1.6397551628742866e-05,
+      "loss": 0.9844,
       "step": 10596
     },
     {
-      "epoch": 0.2910384224547527,
+      "epoch": 0.3007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.66343502008533e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.6396845217828758e-05,
+      "loss": 0.8936,
       "step": 10597
     },
     {
-      "epoch": 0.2910658866825959,
+      "epoch": 0.30073779795686717,
       "grad_norm": 0.0,
-      "learning_rate": 1.663368460856914e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.6396138752880205e-05,
+      "loss": 1.0128,
       "step": 10598
     },
     {
-      "epoch": 0.29109335091043914,
+      "epoch": 0.3007661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.6633018963796697e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.639543223390318e-05,
+      "loss": 0.9504,
       "step": 10599
     },
     {
-      "epoch": 0.29112081513828236,
+      "epoch": 0.300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.6632353266541243e-05,
-      "loss": 0.92,
+      "learning_rate": 1.639472566090366e-05,
+      "loss": 0.9443,
       "step": 10600
     },
     {
-      "epoch": 0.29114827936612564,
+      "epoch": 0.30082292849035186,
       "grad_norm": 0.0,
-      "learning_rate": 1.663168751680804e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.6394019033887597e-05,
+      "loss": 0.8777,
       "step": 10601
     },
     {
-      "epoch": 0.29117574359396886,
+      "epoch": 0.3008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.663102171460235e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.639331235286097e-05,
+      "loss": 0.9083,
       "step": 10602
     },
     {
-      "epoch": 0.2912032078218121,
+      "epoch": 0.30087968217934163,
       "grad_norm": 0.0,
-      "learning_rate": 1.6630355859929454e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.6392605617829745e-05,
+      "loss": 1.0005,
       "step": 10603
     },
     {
-      "epoch": 0.2912306720496553,
+      "epoch": 0.30090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.662968995279461e-05,
-      "loss": 1.0568,
+      "learning_rate": 1.6391898828799895e-05,
+      "loss": 0.9467,
       "step": 10604
     },
     {
-      "epoch": 0.2912581362774986,
+      "epoch": 0.30093643586833146,
       "grad_norm": 0.0,
-      "learning_rate": 1.6629023993203094e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.639119198577739e-05,
+      "loss": 0.9923,
       "step": 10605
     },
     {
-      "epoch": 0.2912856005053418,
+      "epoch": 0.3009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.662835798116017e-05,
-      "loss": 0.7948,
+      "learning_rate": 1.6390485088768193e-05,
+      "loss": 0.8663,
       "step": 10606
     },
     {
-      "epoch": 0.291313064733185,
+      "epoch": 0.30099318955732124,
       "grad_norm": 0.0,
-      "learning_rate": 1.6627691916671113e-05,
-      "loss": 0.9359,
+      "learning_rate": 1.6389778137778284e-05,
+      "loss": 0.9248,
       "step": 10607
     },
     {
-      "epoch": 0.29134052896102824,
+      "epoch": 0.3010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.662702579974119e-05,
-      "loss": 1.0524,
+      "learning_rate": 1.6389071132813634e-05,
+      "loss": 0.9839,
       "step": 10608
     },
     {
-      "epoch": 0.2913679931888715,
+      "epoch": 0.301049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.662635963037567e-05,
-      "loss": 1.026,
+      "learning_rate": 1.638836407388021e-05,
+      "loss": 0.9924,
       "step": 10609
     },
     {
-      "epoch": 0.29139545741671474,
+      "epoch": 0.30107832009080593,
       "grad_norm": 0.0,
-      "learning_rate": 1.6625693408579827e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.638765696098399e-05,
+      "loss": 0.9823,
       "step": 10610
     },
     {
-      "epoch": 0.29142292164455796,
+      "epoch": 0.3011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.6625027134358932e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.6386949794130943e-05,
+      "loss": 0.9968,
       "step": 10611
     },
     {
-      "epoch": 0.2914503858724012,
+      "epoch": 0.3011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.6624360807718255e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.6386242573327048e-05,
+      "loss": 0.9883,
       "step": 10612
     },
     {
-      "epoch": 0.2914778501002444,
+      "epoch": 0.30116345062429056,
       "grad_norm": 0.0,
-      "learning_rate": 1.6623694428663068e-05,
-      "loss": 1.016,
+      "learning_rate": 1.6385535298578267e-05,
+      "loss": 0.9621,
       "step": 10613
     },
     {
-      "epoch": 0.2915053143280877,
+      "epoch": 0.3011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.6623027997198645e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.6384827969890587e-05,
+      "loss": 0.882,
       "step": 10614
     },
     {
-      "epoch": 0.2915327785559309,
+      "epoch": 0.30122020431328034,
       "grad_norm": 0.0,
-      "learning_rate": 1.662236151333026e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.6384120587269978e-05,
+      "loss": 0.9125,
       "step": 10615
     },
     {
-      "epoch": 0.2915602427837741,
+      "epoch": 0.30124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.6621694977063185e-05,
-      "loss": 1.0569,
+      "learning_rate": 1.6383413150722417e-05,
+      "loss": 1.0491,
       "step": 10616
     },
     {
-      "epoch": 0.29158770701161735,
+      "epoch": 0.30127695800227017,
       "grad_norm": 0.0,
-      "learning_rate": 1.6621028388402695e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.6382705660253873e-05,
+      "loss": 1.0549,
       "step": 10617
     },
     {
-      "epoch": 0.2916151712394606,
+      "epoch": 0.301305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.6620361747354064e-05,
-      "loss": 0.8998,
+      "learning_rate": 1.638199811587033e-05,
+      "loss": 1.0292,
       "step": 10618
     },
     {
-      "epoch": 0.29164263546730385,
+      "epoch": 0.30133371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.6619695053922564e-05,
-      "loss": 0.8669,
+      "learning_rate": 1.6381290517577766e-05,
+      "loss": 0.9708,
       "step": 10619
     },
     {
-      "epoch": 0.29167009969514707,
+      "epoch": 0.3013620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.6619028308113474e-05,
-      "loss": 0.8352,
+      "learning_rate": 1.638058286538215e-05,
+      "loss": 0.8629,
       "step": 10620
     },
     {
-      "epoch": 0.2916975639229903,
+      "epoch": 0.3013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.661836150993207e-05,
-      "loss": 0.8817,
+      "learning_rate": 1.6379875159289466e-05,
+      "loss": 0.9368,
       "step": 10621
     },
     {
-      "epoch": 0.29172502815083357,
+      "epoch": 0.30141884222474463,
       "grad_norm": 0.0,
-      "learning_rate": 1.661769465938362e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.6379167399305687e-05,
+      "loss": 0.9712,
       "step": 10622
     },
     {
-      "epoch": 0.2917524923786768,
+      "epoch": 0.3014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.6617027756473413e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.6378459585436793e-05,
+      "loss": 0.8391,
       "step": 10623
     },
     {
-      "epoch": 0.29177995660652,
+      "epoch": 0.3014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.6616360801206716e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.637775171768877e-05,
+      "loss": 0.9541,
       "step": 10624
     },
     {
-      "epoch": 0.29180742083436323,
+      "epoch": 0.30150397275822927,
       "grad_norm": 0.0,
-      "learning_rate": 1.661569379358881e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.6377043796067585e-05,
+      "loss": 0.9861,
       "step": 10625
     },
     {
-      "epoch": 0.29183488506220645,
+      "epoch": 0.3015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.6615026733624974e-05,
-      "loss": 1.0286,
+      "learning_rate": 1.637633582057923e-05,
+      "loss": 0.9559,
       "step": 10626
     },
     {
-      "epoch": 0.2918623492900497,
+      "epoch": 0.3015607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.6614359621320482e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.6375627791229675e-05,
+      "loss": 0.9572,
       "step": 10627
     },
     {
-      "epoch": 0.29188981351789295,
+      "epoch": 0.30158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.6613692456680613e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.6374919708024907e-05,
+      "loss": 1.0222,
       "step": 10628
     },
     {
-      "epoch": 0.29191727774573617,
+      "epoch": 0.30161748013620887,
       "grad_norm": 0.0,
-      "learning_rate": 1.661302523971065e-05,
-      "loss": 1.024,
+      "learning_rate": 1.6374211570970907e-05,
+      "loss": 0.8411,
       "step": 10629
     },
     {
-      "epoch": 0.2919447419735794,
+      "epoch": 0.30164585698070373,
       "grad_norm": 0.0,
-      "learning_rate": 1.6612357970415872e-05,
-      "loss": 0.9008,
+      "learning_rate": 1.6373503380073655e-05,
+      "loss": 0.9449,
       "step": 10630
     },
     {
-      "epoch": 0.29197220620142267,
+      "epoch": 0.30167423382519865,
       "grad_norm": 0.0,
-      "learning_rate": 1.661169064880155e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.637279513533913e-05,
+      "loss": 1.0096,
       "step": 10631
     },
     {
-      "epoch": 0.2919996704292659,
+      "epoch": 0.3017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.6611023274872976e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.6372086836773325e-05,
+      "loss": 1.0592,
       "step": 10632
     },
     {
-      "epoch": 0.2920271346571091,
+      "epoch": 0.3017309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.661035584863542e-05,
-      "loss": 1.0776,
+      "learning_rate": 1.6371378484382212e-05,
+      "loss": 0.8973,
       "step": 10633
     },
     {
-      "epoch": 0.29205459888495233,
+      "epoch": 0.30175936435868334,
       "grad_norm": 0.0,
-      "learning_rate": 1.6609688370094172e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.637067007817178e-05,
+      "loss": 0.8808,
       "step": 10634
     },
     {
-      "epoch": 0.2920820631127956,
+      "epoch": 0.3017877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.660902083925451e-05,
-      "loss": 0.9018,
+      "learning_rate": 1.636996161814801e-05,
+      "loss": 0.9849,
       "step": 10635
     },
     {
-      "epoch": 0.29210952734063883,
+      "epoch": 0.3018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.6608353256121713e-05,
-      "loss": 0.8815,
+      "learning_rate": 1.6369253104316885e-05,
+      "loss": 1.077,
       "step": 10636
     },
     {
-      "epoch": 0.29213699156848205,
+      "epoch": 0.30184449489216797,
       "grad_norm": 0.0,
-      "learning_rate": 1.660768562070106e-05,
-      "loss": 0.9015,
+      "learning_rate": 1.63685445366844e-05,
+      "loss": 1.0875,
       "step": 10637
     },
     {
-      "epoch": 0.2921644557963253,
+      "epoch": 0.3018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.6607017932997845e-05,
-      "loss": 1.0351,
+      "learning_rate": 1.6367835915256523e-05,
+      "loss": 0.9701,
       "step": 10638
     },
     {
-      "epoch": 0.2921919200241685,
+      "epoch": 0.3019012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.6606350193017346e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.636712724003926e-05,
+      "loss": 0.8933,
       "step": 10639
     },
     {
-      "epoch": 0.29221938425201177,
+      "epoch": 0.30192962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 1.6605682400764843e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.636641851103858e-05,
+      "loss": 0.9684,
       "step": 10640
     },
     {
-      "epoch": 0.292246848479855,
+      "epoch": 0.3019580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.660501455624562e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.6365709728260483e-05,
+      "loss": 0.9907,
       "step": 10641
     },
     {
-      "epoch": 0.2922743127076982,
+      "epoch": 0.30198637911464243,
       "grad_norm": 0.0,
-      "learning_rate": 1.6604346659464967e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.6365000891710944e-05,
+      "loss": 0.9632,
       "step": 10642
     },
     {
-      "epoch": 0.29230177693554144,
+      "epoch": 0.30201475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.660367871042816e-05,
-      "loss": 0.9017,
+      "learning_rate": 1.6364292001395958e-05,
+      "loss": 0.9077,
       "step": 10643
     },
     {
-      "epoch": 0.2923292411633847,
+      "epoch": 0.30204313280363226,
       "grad_norm": 0.0,
-      "learning_rate": 1.6603010709140494e-05,
-      "loss": 0.91,
+      "learning_rate": 1.6363583057321512e-05,
+      "loss": 1.0091,
       "step": 10644
     },
     {
-      "epoch": 0.29235670539122793,
+      "epoch": 0.3020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.660234265560725e-05,
-      "loss": 1.013,
+      "learning_rate": 1.6362874059493593e-05,
+      "loss": 0.9235,
       "step": 10645
     },
     {
-      "epoch": 0.29238416961907115,
+      "epoch": 0.30209988649262204,
       "grad_norm": 0.0,
-      "learning_rate": 1.6601674549833707e-05,
-      "loss": 1.0052,
+      "learning_rate": 1.636216500791819e-05,
+      "loss": 0.9545,
       "step": 10646
     },
     {
-      "epoch": 0.2924116338469144,
+      "epoch": 0.3021282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.6601006391825165e-05,
-      "loss": 1.076,
+      "learning_rate": 1.6361455902601296e-05,
+      "loss": 0.9405,
       "step": 10647
     },
     {
-      "epoch": 0.29243909807475765,
+      "epoch": 0.3021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.6600338181586898e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.6360746743548896e-05,
+      "loss": 0.8851,
       "step": 10648
     },
     {
-      "epoch": 0.2924665623026009,
+      "epoch": 0.3021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.65996699191242e-05,
-      "loss": 1.0575,
+      "learning_rate": 1.6360037530766983e-05,
+      "loss": 0.9026,
       "step": 10649
     },
     {
-      "epoch": 0.2924940265304441,
+      "epoch": 0.3022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.6599001604442363e-05,
-      "loss": 0.9022,
+      "learning_rate": 1.6359328264261548e-05,
+      "loss": 0.9704,
       "step": 10650
     },
     {
-      "epoch": 0.2925214907582873,
+      "epoch": 0.3022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.6598333237546663e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.6358618944038578e-05,
+      "loss": 0.9035,
       "step": 10651
     },
     {
-      "epoch": 0.29254895498613054,
+      "epoch": 0.30227014755959136,
       "grad_norm": 0.0,
-      "learning_rate": 1.6597664818442394e-05,
-      "loss": 0.8412,
+      "learning_rate": 1.635790957010407e-05,
+      "loss": 1.044,
       "step": 10652
     },
     {
-      "epoch": 0.2925764192139738,
+      "epoch": 0.3022985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.6596996347134847e-05,
-      "loss": 0.8523,
+      "learning_rate": 1.6357200142464013e-05,
+      "loss": 0.9728,
       "step": 10653
     },
     {
-      "epoch": 0.29260388344181704,
+      "epoch": 0.30232690124858114,
       "grad_norm": 0.0,
-      "learning_rate": 1.6596327823629308e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.6356490661124402e-05,
+      "loss": 1.0575,
       "step": 10654
     },
     {
-      "epoch": 0.29263134766966026,
+      "epoch": 0.30235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.659565924793107e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.6355781126091227e-05,
+      "loss": 0.9234,
       "step": 10655
     },
     {
-      "epoch": 0.2926588118975035,
+      "epoch": 0.30238365493757097,
       "grad_norm": 0.0,
-      "learning_rate": 1.6594990620045423e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.6355071537370484e-05,
+      "loss": 0.9292,
       "step": 10656
     },
     {
-      "epoch": 0.29268627612534676,
+      "epoch": 0.3024120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.659432193997765e-05,
-      "loss": 1.1029,
+      "learning_rate": 1.635436189496817e-05,
+      "loss": 0.9547,
       "step": 10657
     },
     {
-      "epoch": 0.29271374035319,
+      "epoch": 0.30244040862656074,
       "grad_norm": 0.0,
-      "learning_rate": 1.6593653207733053e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.635365219889027e-05,
+      "loss": 0.9179,
       "step": 10658
     },
     {
-      "epoch": 0.2927412045810332,
+      "epoch": 0.3024687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.6592984423316914e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.6352942449142787e-05,
+      "loss": 0.8601,
       "step": 10659
     },
     {
-      "epoch": 0.2927686688088764,
+      "epoch": 0.3024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.659231558673453e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.635223264573171e-05,
+      "loss": 0.9018,
       "step": 10660
     },
     {
-      "epoch": 0.2927961330367197,
+      "epoch": 0.3025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.659164669799119e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.6351522788663044e-05,
+      "loss": 0.9764,
       "step": 10661
     },
     {
-      "epoch": 0.2928235972645629,
+      "epoch": 0.3025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.6590977757092188e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.6350812877942778e-05,
+      "loss": 1.0361,
       "step": 10662
     },
     {
-      "epoch": 0.29285106149240614,
+      "epoch": 0.3025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.659030876404282e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.6350102913576906e-05,
+      "loss": 0.8914,
       "step": 10663
     },
     {
-      "epoch": 0.29287852572024936,
+      "epoch": 0.30261066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.6589639718848373e-05,
-      "loss": 1.0129,
+      "learning_rate": 1.6349392895571433e-05,
+      "loss": 0.8987,
       "step": 10664
     },
     {
-      "epoch": 0.2929059899480926,
+      "epoch": 0.302639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.6588970621514146e-05,
-      "loss": 1.0704,
+      "learning_rate": 1.6348682823932352e-05,
+      "loss": 0.9612,
       "step": 10665
     },
     {
-      "epoch": 0.29293345417593586,
+      "epoch": 0.30266742338251984,
       "grad_norm": 0.0,
-      "learning_rate": 1.658830147204543e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.634797269866566e-05,
+      "loss": 1.0013,
       "step": 10666
     },
     {
-      "epoch": 0.2929609184037791,
+      "epoch": 0.30269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.658763227044752e-05,
-      "loss": 1.0362,
+      "learning_rate": 1.6347262519777358e-05,
+      "loss": 1.0813,
       "step": 10667
     },
     {
-      "epoch": 0.2929883826316223,
+      "epoch": 0.30272417707150967,
       "grad_norm": 0.0,
-      "learning_rate": 1.6586963016725714e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.6346552287273446e-05,
+      "loss": 1.0125,
       "step": 10668
     },
     {
-      "epoch": 0.2930158468594655,
+      "epoch": 0.30275255391600453,
       "grad_norm": 0.0,
-      "learning_rate": 1.65862937108853e-05,
-      "loss": 0.8255,
+      "learning_rate": 1.634584200115992e-05,
+      "loss": 0.9713,
       "step": 10669
     },
     {
-      "epoch": 0.2930433110873088,
+      "epoch": 0.30278093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.6585624352931583e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.6345131661442782e-05,
+      "loss": 0.8082,
       "step": 10670
     },
     {
-      "epoch": 0.293070775315152,
+      "epoch": 0.3028093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.658495494286986e-05,
-      "loss": 1.032,
+      "learning_rate": 1.6344421268128032e-05,
+      "loss": 0.939,
       "step": 10671
     },
     {
-      "epoch": 0.29309823954299524,
+      "epoch": 0.3028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.6584285480705413e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.634371082122167e-05,
+      "loss": 0.9722,
       "step": 10672
     },
     {
-      "epoch": 0.29312570377083846,
+      "epoch": 0.30286606129398413,
       "grad_norm": 0.0,
-      "learning_rate": 1.6583615966443555e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.63430003207297e-05,
+      "loss": 1.0053,
       "step": 10673
     },
     {
-      "epoch": 0.29315316799868174,
+      "epoch": 0.302894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.6582946400089574e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.6342289766658114e-05,
+      "loss": 1.0135,
       "step": 10674
     },
     {
-      "epoch": 0.29318063222652496,
+      "epoch": 0.3029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.6582276781648773e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.6341579159012928e-05,
+      "loss": 1.0538,
       "step": 10675
     },
     {
-      "epoch": 0.2932080964543682,
+      "epoch": 0.30295119182746877,
       "grad_norm": 0.0,
-      "learning_rate": 1.6581607111126447e-05,
-      "loss": 1.0231,
+      "learning_rate": 1.6340868497800135e-05,
+      "loss": 1.0508,
       "step": 10676
     },
     {
-      "epoch": 0.2932355606822114,
+      "epoch": 0.3029795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.6580937388527898e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.634015778302574e-05,
+      "loss": 1.0195,
       "step": 10677
     },
     {
-      "epoch": 0.2932630249100547,
+      "epoch": 0.30300794551645854,
       "grad_norm": 0.0,
-      "learning_rate": 1.6580267613858422e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.633944701469575e-05,
+      "loss": 0.9279,
       "step": 10678
     },
     {
-      "epoch": 0.2932904891378979,
+      "epoch": 0.30303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.6579597787123318e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.6338736192816163e-05,
+      "loss": 0.8671,
       "step": 10679
     },
     {
-      "epoch": 0.2933179533657411,
+      "epoch": 0.3030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.6578927908327892e-05,
-      "loss": 0.9611,
+      "learning_rate": 1.6338025317392986e-05,
+      "loss": 0.9258,
       "step": 10680
     },
     {
-      "epoch": 0.29334541759358435,
+      "epoch": 0.30309307604994323,
       "grad_norm": 0.0,
-      "learning_rate": 1.6578257977477433e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.633731438843223e-05,
+      "loss": 0.9894,
       "step": 10681
     },
     {
-      "epoch": 0.29337288182142757,
+      "epoch": 0.30312145289443815,
       "grad_norm": 0.0,
-      "learning_rate": 1.6577587994577254e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.633660340593989e-05,
+      "loss": 1.0419,
       "step": 10682
     },
     {
-      "epoch": 0.29340034604927084,
+      "epoch": 0.303149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.657691795963265e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.6335892369921974e-05,
+      "loss": 0.863,
       "step": 10683
     },
     {
-      "epoch": 0.29342781027711407,
+      "epoch": 0.3031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.6576247872648922e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.6335181280384494e-05,
+      "loss": 0.9607,
       "step": 10684
     },
     {
-      "epoch": 0.2934552745049573,
+      "epoch": 0.30320658342792284,
       "grad_norm": 0.0,
-      "learning_rate": 1.6575577733631375e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.6334470137333452e-05,
+      "loss": 0.9632,
       "step": 10685
     },
     {
-      "epoch": 0.2934827387328005,
+      "epoch": 0.3032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.657490754258531e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.6333758940774854e-05,
+      "loss": 0.7791,
       "step": 10686
     },
     {
-      "epoch": 0.2935102029606438,
+      "epoch": 0.3032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.657423729951603e-05,
-      "loss": 1.0859,
+      "learning_rate": 1.6333047690714714e-05,
+      "loss": 0.8973,
       "step": 10687
     },
     {
-      "epoch": 0.293537667188487,
+      "epoch": 0.30329171396140747,
       "grad_norm": 0.0,
-      "learning_rate": 1.6573567004428837e-05,
-      "loss": 0.8809,
+      "learning_rate": 1.6332336387159032e-05,
+      "loss": 0.8868,
       "step": 10688
     },
     {
-      "epoch": 0.2935651314163302,
+      "epoch": 0.3033200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.6572896657329037e-05,
-      "loss": 0.9338,
+      "learning_rate": 1.633162503011382e-05,
+      "loss": 0.9203,
       "step": 10689
     },
     {
-      "epoch": 0.29359259564417345,
+      "epoch": 0.3033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.657222625822193e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.633091361958509e-05,
+      "loss": 1.0095,
       "step": 10690
     },
     {
-      "epoch": 0.2936200598720167,
+      "epoch": 0.30337684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 1.6571555807112822e-05,
-      "loss": 0.9537,
+      "learning_rate": 1.6330202155578845e-05,
+      "loss": 0.9753,
       "step": 10691
     },
     {
-      "epoch": 0.29364752409985995,
+      "epoch": 0.3034052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.6570885304007023e-05,
-      "loss": 1.0869,
+      "learning_rate": 1.6329490638101096e-05,
+      "loss": 0.8387,
       "step": 10692
     },
     {
-      "epoch": 0.29367498832770317,
+      "epoch": 0.30343359818388194,
       "grad_norm": 0.0,
-      "learning_rate": 1.6570214748909833e-05,
-      "loss": 0.8878,
+      "learning_rate": 1.6328779067157862e-05,
+      "loss": 0.8518,
       "step": 10693
     },
     {
-      "epoch": 0.2937024525555464,
+      "epoch": 0.30346197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.6569544141826558e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.632806744275514e-05,
+      "loss": 0.9065,
       "step": 10694
     },
     {
-      "epoch": 0.2937299167833896,
+      "epoch": 0.3034903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.6568873482762504e-05,
-      "loss": 1.002,
+      "learning_rate": 1.632735576489895e-05,
+      "loss": 0.9462,
       "step": 10695
     },
     {
-      "epoch": 0.2937573810112329,
+      "epoch": 0.3035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.6568202771722977e-05,
-      "loss": 0.9359,
+      "learning_rate": 1.6326644033595303e-05,
+      "loss": 0.9423,
       "step": 10696
     },
     {
-      "epoch": 0.2937848452390761,
+      "epoch": 0.30354710556186154,
       "grad_norm": 0.0,
-      "learning_rate": 1.656753200871329e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.6325932248850206e-05,
+      "loss": 0.9388,
       "step": 10697
     },
     {
-      "epoch": 0.29381230946691933,
+      "epoch": 0.3035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.6566861193738745e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.6325220410669676e-05,
+      "loss": 0.8938,
       "step": 10698
     },
     {
-      "epoch": 0.29383977369476255,
+      "epoch": 0.3036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.6566190326804647e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.6324508519059728e-05,
+      "loss": 1.0059,
       "step": 10699
     },
     {
-      "epoch": 0.29386723792260583,
+      "epoch": 0.3036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.656551940791631e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.6323796574026368e-05,
+      "loss": 0.9777,
       "step": 10700
     },
     {
-      "epoch": 0.29389470215044905,
+      "epoch": 0.3036606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.656484843707904e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.6323084575575617e-05,
+      "loss": 0.901,
       "step": 10701
     },
     {
-      "epoch": 0.29392216637829227,
+      "epoch": 0.303688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.6564177414298146e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.6322372523713487e-05,
+      "loss": 0.7912,
       "step": 10702
     },
     {
-      "epoch": 0.2939496306061355,
+      "epoch": 0.30371736662883086,
       "grad_norm": 0.0,
-      "learning_rate": 1.6563506339578935e-05,
-      "loss": 1.026,
+      "learning_rate": 1.632166041844599e-05,
+      "loss": 0.8844,
       "step": 10703
     },
     {
-      "epoch": 0.29397709483397877,
+      "epoch": 0.3037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.6562835212926725e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.6320948259779143e-05,
+      "loss": 0.9914,
       "step": 10704
     },
     {
-      "epoch": 0.294004559061822,
+      "epoch": 0.30377412031782064,
       "grad_norm": 0.0,
-      "learning_rate": 1.6562164034346817e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.632023604771896e-05,
+      "loss": 1.007,
       "step": 10705
     },
     {
-      "epoch": 0.2940320232896652,
+      "epoch": 0.30380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.6561492803844525e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.6319523782271462e-05,
+      "loss": 1.008,
       "step": 10706
     },
     {
-      "epoch": 0.29405948751750843,
+      "epoch": 0.30383087400681047,
       "grad_norm": 0.0,
-      "learning_rate": 1.656082152142516e-05,
-      "loss": 1.0468,
+      "learning_rate": 1.6318811463442666e-05,
+      "loss": 0.9541,
       "step": 10707
     },
     {
-      "epoch": 0.29408695174535165,
+      "epoch": 0.30385925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.6560150187094034e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.631809909123858e-05,
+      "loss": 1.0979,
       "step": 10708
     },
     {
-      "epoch": 0.29411441597319493,
+      "epoch": 0.30388762769580024,
       "grad_norm": 0.0,
-      "learning_rate": 1.6559478800856464e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.631738666566523e-05,
+      "loss": 0.9221,
       "step": 10709
     },
     {
-      "epoch": 0.29414188020103815,
+      "epoch": 0.3039160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.655880736271775e-05,
-      "loss": 1.1725,
+      "learning_rate": 1.6316674186728628e-05,
+      "loss": 0.8671,
       "step": 10710
     },
     {
-      "epoch": 0.2941693444288814,
+      "epoch": 0.30394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.6558135872683215e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.6315961654434797e-05,
+      "loss": 0.9666,
       "step": 10711
     },
     {
-      "epoch": 0.2941968086567246,
+      "epoch": 0.3039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.6557464330758167e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.6315249068789755e-05,
+      "loss": 0.9554,
       "step": 10712
     },
     {
-      "epoch": 0.2942242728845679,
+      "epoch": 0.3040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.655679273694792e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.6314536429799516e-05,
+      "loss": 0.9442,
       "step": 10713
     },
     {
-      "epoch": 0.2942517371124111,
+      "epoch": 0.3040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.655612109125779e-05,
-      "loss": 0.8907,
+      "learning_rate": 1.6313823737470105e-05,
+      "loss": 1.0348,
       "step": 10714
     },
     {
-      "epoch": 0.2942792013402543,
+      "epoch": 0.30405788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 1.6555449393693092e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.6313110991807546e-05,
+      "loss": 0.8562,
       "step": 10715
     },
     {
-      "epoch": 0.29430666556809754,
+      "epoch": 0.3040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.6554777644259138e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.631239819281785e-05,
+      "loss": 0.8571,
       "step": 10716
     },
     {
-      "epoch": 0.2943341297959408,
+      "epoch": 0.30411464245175934,
       "grad_norm": 0.0,
-      "learning_rate": 1.6554105842961245e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.6311685340507046e-05,
+      "loss": 0.9235,
       "step": 10717
     },
     {
-      "epoch": 0.29436159402378403,
+      "epoch": 0.30414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.6553433989804725e-05,
-      "loss": 0.8951,
+      "learning_rate": 1.6310972434881152e-05,
+      "loss": 1.0948,
       "step": 10718
     },
     {
-      "epoch": 0.29438905825162726,
+      "epoch": 0.3041713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.65527620847949e-05,
-      "loss": 0.9041,
+      "learning_rate": 1.6310259475946187e-05,
+      "loss": 0.9343,
       "step": 10719
     },
     {
-      "epoch": 0.2944165224794705,
+      "epoch": 0.30419977298524403,
       "grad_norm": 0.0,
-      "learning_rate": 1.655209012793708e-05,
-      "loss": 0.915,
+      "learning_rate": 1.630954646370818e-05,
+      "loss": 1.015,
       "step": 10720
     },
     {
-      "epoch": 0.2944439867073137,
+      "epoch": 0.30422814982973895,
       "grad_norm": 0.0,
-      "learning_rate": 1.6551418119236585e-05,
-      "loss": 0.8205,
+      "learning_rate": 1.630883339817315e-05,
+      "loss": 0.8925,
       "step": 10721
     },
     {
-      "epoch": 0.294471450935157,
+      "epoch": 0.3042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.6550746058698736e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.6308120279347118e-05,
+      "loss": 1.0146,
       "step": 10722
     },
     {
-      "epoch": 0.2944989151630002,
+      "epoch": 0.3042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.6550073946328843e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.630740710723611e-05,
+      "loss": 1.0412,
       "step": 10723
     },
     {
-      "epoch": 0.2945263793908434,
+      "epoch": 0.30431328036322364,
       "grad_norm": 0.0,
-      "learning_rate": 1.654940178213223e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.6306693881846152e-05,
+      "loss": 0.9601,
       "step": 10724
     },
     {
-      "epoch": 0.29455384361868664,
+      "epoch": 0.3043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.654872956611421e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.630598060318327e-05,
+      "loss": 0.9308,
       "step": 10725
     },
     {
-      "epoch": 0.2945813078465299,
+      "epoch": 0.3043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.6548057298280105e-05,
-      "loss": 0.9808,
+      "learning_rate": 1.6305267271253483e-05,
+      "loss": 0.9736,
       "step": 10726
     },
     {
-      "epoch": 0.29460877207437314,
+      "epoch": 0.30439841089670827,
       "grad_norm": 0.0,
-      "learning_rate": 1.654738497863524e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.630455388606282e-05,
+      "loss": 0.9914,
       "step": 10727
     },
     {
-      "epoch": 0.29463623630221636,
+      "epoch": 0.3044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.6546712607184924e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.6303840447617306e-05,
+      "loss": 0.9999,
       "step": 10728
     },
     {
-      "epoch": 0.2946637005300596,
+      "epoch": 0.30445516458569805,
       "grad_norm": 0.0,
-      "learning_rate": 1.6546040183934482e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.630312695592297e-05,
+      "loss": 0.842,
       "step": 10729
     },
     {
-      "epoch": 0.29469116475790286,
+      "epoch": 0.30448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.6545367708889235e-05,
-      "loss": 1.0901,
+      "learning_rate": 1.6302413410985838e-05,
+      "loss": 0.9444,
       "step": 10730
     },
     {
-      "epoch": 0.2947186289857461,
+      "epoch": 0.3045119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.6544695182054503e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.6301699812811936e-05,
+      "loss": 0.879,
       "step": 10731
     },
     {
-      "epoch": 0.2947460932135893,
+      "epoch": 0.30454029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.654402260343561e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.6300986161407292e-05,
+      "loss": 0.9154,
       "step": 10732
     },
     {
-      "epoch": 0.2947735574414325,
+      "epoch": 0.30456867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.6543349973037872e-05,
-      "loss": 0.954,
+      "learning_rate": 1.6300272456777933e-05,
+      "loss": 1.0007,
       "step": 10733
     },
     {
-      "epoch": 0.29480102166927574,
+      "epoch": 0.3045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.6542677290866612e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.6299558698929894e-05,
+      "loss": 0.9115,
       "step": 10734
     },
     {
-      "epoch": 0.294828485897119,
+      "epoch": 0.3046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.654200455692716e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.6298844887869196e-05,
+      "loss": 0.8141,
       "step": 10735
     },
     {
-      "epoch": 0.29485595012496224,
+      "epoch": 0.30465380249716234,
       "grad_norm": 0.0,
-      "learning_rate": 1.6541331771224832e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.6298131023601872e-05,
+      "loss": 0.9774,
       "step": 10736
     },
     {
-      "epoch": 0.29488341435280546,
+      "epoch": 0.3046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.6540658933764954e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.6297417106133954e-05,
+      "loss": 0.9836,
       "step": 10737
     },
     {
-      "epoch": 0.2949108785806487,
+      "epoch": 0.3047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.6539986044552848e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.629670313547147e-05,
+      "loss": 0.9286,
       "step": 10738
     },
     {
-      "epoch": 0.29493834280849196,
+      "epoch": 0.304738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.6539313103593842e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.629598911162045e-05,
+      "loss": 0.9032,
       "step": 10739
     },
     {
-      "epoch": 0.2949658070363352,
+      "epoch": 0.3047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.6538640110893253e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.629527503458693e-05,
+      "loss": 0.9317,
       "step": 10740
     },
     {
-      "epoch": 0.2949932712641784,
+      "epoch": 0.30479568671963675,
       "grad_norm": 0.0,
-      "learning_rate": 1.6537967066456412e-05,
-      "loss": 0.973,
+      "learning_rate": 1.6294560904376938e-05,
+      "loss": 1.0262,
       "step": 10741
     },
     {
-      "epoch": 0.2950207354920216,
+      "epoch": 0.30482406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 1.6537293970288644e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.6293846720996504e-05,
+      "loss": 0.9872,
       "step": 10742
     },
     {
-      "epoch": 0.2950481997198649,
+      "epoch": 0.3048524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.6536620822395274e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.6293132484451667e-05,
+      "loss": 0.9216,
       "step": 10743
     },
     {
-      "epoch": 0.2950756639477081,
+      "epoch": 0.30488081725312144,
       "grad_norm": 0.0,
-      "learning_rate": 1.653594762278163e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.6292418194748452e-05,
+      "loss": 0.8067,
       "step": 10744
     },
     {
-      "epoch": 0.29510312817555134,
+      "epoch": 0.30490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.653527437145303e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.6291703851892904e-05,
+      "loss": 0.9594,
       "step": 10745
     },
     {
-      "epoch": 0.29513059240339456,
+      "epoch": 0.3049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.6534601068414812e-05,
-      "loss": 0.973,
+      "learning_rate": 1.629098945589105e-05,
+      "loss": 0.9642,
       "step": 10746
     },
     {
-      "epoch": 0.2951580566312378,
+      "epoch": 0.30496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.65339277136723e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.6290275006748922e-05,
+      "loss": 1.0432,
       "step": 10747
     },
     {
-      "epoch": 0.29518552085908106,
+      "epoch": 0.30499432463110104,
       "grad_norm": 0.0,
-      "learning_rate": 1.653325430723082e-05,
-      "loss": 0.9583,
+      "learning_rate": 1.628956050447256e-05,
+      "loss": 0.9166,
       "step": 10748
     },
     {
-      "epoch": 0.2952129850869243,
+      "epoch": 0.3050227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.65325808490957e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.6288845949068e-05,
+      "loss": 0.8435,
       "step": 10749
     },
     {
-      "epoch": 0.2952404493147675,
+      "epoch": 0.3050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.653190733927227e-05,
-      "loss": 0.989,
+      "learning_rate": 1.628813134054127e-05,
+      "loss": 1.0121,
       "step": 10750
     },
     {
-      "epoch": 0.2952679135426107,
+      "epoch": 0.3050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.653123377776586e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.6287416678898415e-05,
+      "loss": 0.9499,
       "step": 10751
     },
     {
-      "epoch": 0.295295377770454,
+      "epoch": 0.3051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.65305601645818e-05,
-      "loss": 0.9039,
+      "learning_rate": 1.6286701964145468e-05,
+      "loss": 0.9555,
       "step": 10752
     },
     {
-      "epoch": 0.2953228419982972,
+      "epoch": 0.3051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.6529886499725414e-05,
-      "loss": 0.9255,
+      "learning_rate": 1.6285987196288465e-05,
+      "loss": 0.8549,
       "step": 10753
     },
     {
-      "epoch": 0.29535030622614045,
+      "epoch": 0.30516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.652921278320204e-05,
-      "loss": 1.0517,
+      "learning_rate": 1.6285272375333453e-05,
+      "loss": 0.873,
       "step": 10754
     },
     {
-      "epoch": 0.29537777045398367,
+      "epoch": 0.3051929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.6528539015017006e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.6284557501286457e-05,
+      "loss": 0.856,
       "step": 10755
     },
     {
-      "epoch": 0.29540523468182694,
+      "epoch": 0.30522133938706014,
       "grad_norm": 0.0,
-      "learning_rate": 1.6527865195175638e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.628384257415352e-05,
+      "loss": 1.0299,
       "step": 10756
     },
     {
-      "epoch": 0.29543269890967017,
+      "epoch": 0.30524971623155506,
       "grad_norm": 0.0,
-      "learning_rate": 1.6527191323683275e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.6283127593940686e-05,
+      "loss": 0.9889,
       "step": 10757
     },
     {
-      "epoch": 0.2954601631375134,
+      "epoch": 0.3052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.6526517400545247e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.628241256065399e-05,
+      "loss": 0.9406,
       "step": 10758
     },
     {
-      "epoch": 0.2954876273653566,
+      "epoch": 0.30530646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.6525843425766882e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.628169747429947e-05,
+      "loss": 0.9857,
       "step": 10759
     },
     {
-      "epoch": 0.29551509159319983,
+      "epoch": 0.30533484676503975,
       "grad_norm": 0.0,
-      "learning_rate": 1.6525169399353518e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.628098233488317e-05,
+      "loss": 1.0875,
       "step": 10760
     },
     {
-      "epoch": 0.2955425558210431,
+      "epoch": 0.3053632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.6524495321310487e-05,
-      "loss": 0.983,
+      "learning_rate": 1.628026714241113e-05,
+      "loss": 1.0994,
       "step": 10761
     },
     {
-      "epoch": 0.29557002004888633,
+      "epoch": 0.3053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.652382119164312e-05,
-      "loss": 0.939,
+      "learning_rate": 1.6279551896889393e-05,
+      "loss": 0.9375,
       "step": 10762
     },
     {
-      "epoch": 0.29559748427672955,
+      "epoch": 0.3054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.6523147010356753e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.6278836598324e-05,
+      "loss": 1.0269,
       "step": 10763
     },
     {
-      "epoch": 0.29562494850457277,
+      "epoch": 0.3054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.652247277745672e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.627812124672099e-05,
+      "loss": 1.0278,
       "step": 10764
     },
     {
-      "epoch": 0.29565241273241605,
+      "epoch": 0.3054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.6521798492948353e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.6277405842086405e-05,
+      "loss": 0.9344,
       "step": 10765
     },
     {
-      "epoch": 0.29567987696025927,
+      "epoch": 0.30550510783200907,
       "grad_norm": 0.0,
-      "learning_rate": 1.6521124156836996e-05,
-      "loss": 1.0413,
+      "learning_rate": 1.627669038442629e-05,
+      "loss": 0.858,
       "step": 10766
     },
     {
-      "epoch": 0.2957073411881025,
+      "epoch": 0.305533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.6520449769127976e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.6275974873746693e-05,
+      "loss": 1.01,
       "step": 10767
     },
     {
-      "epoch": 0.2957348054159457,
+      "epoch": 0.30556186152099885,
       "grad_norm": 0.0,
-      "learning_rate": 1.651977532982663e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.6275259310053653e-05,
+      "loss": 0.8703,
       "step": 10768
     },
     {
-      "epoch": 0.295762269643789,
+      "epoch": 0.30559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.6519100838938296e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.6274543693353218e-05,
+      "loss": 0.8699,
       "step": 10769
     },
     {
-      "epoch": 0.2957897338716322,
+      "epoch": 0.3056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.651842629646831e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.6273828023651426e-05,
+      "loss": 0.9174,
       "step": 10770
     },
     {
-      "epoch": 0.29581719809947543,
+      "epoch": 0.30564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.6517751702422016e-05,
-      "loss": 0.994,
+      "learning_rate": 1.627311230095433e-05,
+      "loss": 0.9144,
       "step": 10771
     },
     {
-      "epoch": 0.29584466232731865,
+      "epoch": 0.30567536889897845,
       "grad_norm": 0.0,
-      "learning_rate": 1.651707705680474e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.627239652526797e-05,
+      "loss": 0.9479,
       "step": 10772
     },
     {
-      "epoch": 0.2958721265551619,
+      "epoch": 0.3057037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.651640235962183e-05,
-      "loss": 1.0679,
+      "learning_rate": 1.62716806965984e-05,
+      "loss": 0.9667,
       "step": 10773
     },
     {
-      "epoch": 0.29589959078300515,
+      "epoch": 0.3057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.6515727610878616e-05,
-      "loss": 1.1462,
+      "learning_rate": 1.6270964814951654e-05,
+      "loss": 0.9136,
       "step": 10774
     },
     {
-      "epoch": 0.29592705501084837,
+      "epoch": 0.3057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.6515052810580443e-05,
-      "loss": 0.9882,
+      "learning_rate": 1.627024888033379e-05,
+      "loss": 0.9069,
       "step": 10775
     },
     {
-      "epoch": 0.2959545192386916,
+      "epoch": 0.305788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.651437795873265e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.626953289275085e-05,
+      "loss": 1.0186,
       "step": 10776
     },
     {
-      "epoch": 0.2959819834665348,
+      "epoch": 0.3058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.6513703055340575e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.6268816852208887e-05,
+      "loss": 0.8292,
       "step": 10777
     },
     {
-      "epoch": 0.2960094476943781,
+      "epoch": 0.3058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.6513028100409555e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.626810075871394e-05,
+      "loss": 0.9482,
       "step": 10778
     },
     {
-      "epoch": 0.2960369119222213,
+      "epoch": 0.3058740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.6512353093944935e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.6267384612272073e-05,
+      "loss": 1.0394,
       "step": 10779
     },
     {
-      "epoch": 0.29606437615006453,
+      "epoch": 0.30590238365493755,
       "grad_norm": 0.0,
-      "learning_rate": 1.6511678035952056e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.626666841288932e-05,
+      "loss": 1.0409,
       "step": 10780
     },
     {
-      "epoch": 0.29609184037790776,
+      "epoch": 0.30593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.651100292643626e-05,
-      "loss": 0.9095,
+      "learning_rate": 1.6265952160571736e-05,
+      "loss": 0.9036,
       "step": 10781
     },
     {
-      "epoch": 0.29611930460575103,
+      "epoch": 0.3059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.6510327765402882e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.6265235855325377e-05,
+      "loss": 0.9096,
       "step": 10782
     },
     {
-      "epoch": 0.29614676883359425,
+      "epoch": 0.30598751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.6509652552857274e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.6264519497156287e-05,
+      "loss": 0.9297,
       "step": 10783
     },
     {
-      "epoch": 0.2961742330614375,
+      "epoch": 0.30601589103291715,
       "grad_norm": 0.0,
-      "learning_rate": 1.650897728880477e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.626380308607052e-05,
+      "loss": 1.0445,
       "step": 10784
     },
     {
-      "epoch": 0.2962016972892807,
+      "epoch": 0.306044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.6508301973250714e-05,
-      "loss": 1.0416,
+      "learning_rate": 1.6263086622074127e-05,
+      "loss": 0.9416,
       "step": 10785
     },
     {
-      "epoch": 0.296229161517124,
+      "epoch": 0.30607264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 1.650762660620046e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.626237010517316e-05,
+      "loss": 1.0028,
       "step": 10786
     },
     {
-      "epoch": 0.2962566257449672,
+      "epoch": 0.30610102156640184,
       "grad_norm": 0.0,
-      "learning_rate": 1.6506951187659335e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.6261653535373668e-05,
+      "loss": 1.0365,
       "step": 10787
     },
     {
-      "epoch": 0.2962840899728104,
+      "epoch": 0.3061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.6506275717632693e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.626093691268171e-05,
+      "loss": 0.9462,
       "step": 10788
     },
     {
-      "epoch": 0.29631155420065364,
+      "epoch": 0.3061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.650560019612588e-05,
-      "loss": 0.9448,
+      "learning_rate": 1.6260220237103338e-05,
+      "loss": 0.9707,
       "step": 10789
     },
     {
-      "epoch": 0.29633901842849686,
+      "epoch": 0.3061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.6504924623144237e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.62595035086446e-05,
+      "loss": 0.889,
       "step": 10790
     },
     {
-      "epoch": 0.29636648265634014,
+      "epoch": 0.3062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.6504248998693114e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.6258786727311558e-05,
+      "loss": 0.9832,
       "step": 10791
     },
     {
-      "epoch": 0.29639394688418336,
+      "epoch": 0.30624290578887625,
       "grad_norm": 0.0,
-      "learning_rate": 1.650357332277785e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.625806989311026e-05,
+      "loss": 0.9692,
       "step": 10792
     },
     {
-      "epoch": 0.2964214111120266,
+      "epoch": 0.30627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.6502897595403795e-05,
-      "loss": 1.1059,
+      "learning_rate": 1.6257353006046768e-05,
+      "loss": 0.8666,
       "step": 10793
     },
     {
-      "epoch": 0.2964488753398698,
+      "epoch": 0.3062996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.6502221816576292e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.625663606612713e-05,
+      "loss": 1.0072,
       "step": 10794
     },
     {
-      "epoch": 0.2964763395677131,
+      "epoch": 0.30632803632236094,
       "grad_norm": 0.0,
-      "learning_rate": 1.65015459863007e-05,
-      "loss": 0.8848,
+      "learning_rate": 1.625591907335741e-05,
+      "loss": 0.8774,
       "step": 10795
     },
     {
-      "epoch": 0.2965038037955563,
+      "epoch": 0.30635641316685586,
       "grad_norm": 0.0,
-      "learning_rate": 1.650087010458235e-05,
-      "loss": 1.0372,
+      "learning_rate": 1.6255202027743658e-05,
+      "loss": 1.0963,
       "step": 10796
     },
     {
-      "epoch": 0.2965312680233995,
+      "epoch": 0.3063847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.65001941714266e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.6254484929291933e-05,
+      "loss": 0.9024,
       "step": 10797
     },
     {
-      "epoch": 0.29655873225124274,
+      "epoch": 0.30641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.6499518186838795e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.6253767778008293e-05,
+      "loss": 0.952,
       "step": 10798
     },
     {
-      "epoch": 0.296586196479086,
+      "epoch": 0.30644154370034055,
       "grad_norm": 0.0,
-      "learning_rate": 1.6498842150824283e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.6253050573898796e-05,
+      "loss": 1.0424,
       "step": 10799
     },
     {
-      "epoch": 0.29661366070692924,
+      "epoch": 0.3064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.6498166063388413e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.6252333316969497e-05,
+      "loss": 0.9215,
       "step": 10800
     },
     {
-      "epoch": 0.29664112493477246,
+      "epoch": 0.3064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.6497489924536542e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.6251616007226464e-05,
+      "loss": 1.0745,
       "step": 10801
     },
     {
-      "epoch": 0.2966685891626157,
+      "epoch": 0.3065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.6496813734274014e-05,
-      "loss": 0.8265,
+      "learning_rate": 1.6250898644675745e-05,
+      "loss": 0.9877,
       "step": 10802
     },
     {
-      "epoch": 0.2966960533904589,
+      "epoch": 0.3065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.6496137492606174e-05,
-      "loss": 0.9913,
+      "learning_rate": 1.6250181229323403e-05,
+      "loss": 0.9437,
       "step": 10803
     },
     {
-      "epoch": 0.2967235176183022,
+      "epoch": 0.306583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.6495461199538382e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.6249463761175502e-05,
+      "loss": 0.9753,
       "step": 10804
     },
     {
-      "epoch": 0.2967509818461454,
+      "epoch": 0.30661180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.649478485507598e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.6248746240238095e-05,
+      "loss": 0.8991,
       "step": 10805
     },
     {
-      "epoch": 0.2967784460739886,
+      "epoch": 0.3066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.6494108459224327e-05,
-      "loss": 0.918,
+      "learning_rate": 1.624802866651725e-05,
+      "loss": 1.1112,
       "step": 10806
     },
     {
-      "epoch": 0.29680591030183184,
+      "epoch": 0.30666855845629964,
       "grad_norm": 0.0,
-      "learning_rate": 1.6493432011988777e-05,
-      "loss": 0.8,
+      "learning_rate": 1.624731104001903e-05,
+      "loss": 0.8798,
       "step": 10807
     },
     {
-      "epoch": 0.2968333745296751,
+      "epoch": 0.30669693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 1.649275551337467e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.6246593360749488e-05,
+      "loss": 1.0854,
       "step": 10808
     },
     {
-      "epoch": 0.29686083875751834,
+      "epoch": 0.3067253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.649207896338737e-05,
-      "loss": 1.013,
+      "learning_rate": 1.6245875628714692e-05,
+      "loss": 1.0138,
       "step": 10809
     },
     {
-      "epoch": 0.29688830298536156,
+      "epoch": 0.30675368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 1.6491402362032227e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.6245157843920708e-05,
+      "loss": 0.9993,
       "step": 10810
     },
     {
-      "epoch": 0.2969157672132048,
+      "epoch": 0.30678206583427925,
       "grad_norm": 0.0,
-      "learning_rate": 1.6490725709314595e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.624444000637359e-05,
+      "loss": 1.0559,
       "step": 10811
     },
     {
-      "epoch": 0.29694323144104806,
+      "epoch": 0.3068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.6490049005239824e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.6243722116079406e-05,
+      "loss": 0.9777,
       "step": 10812
     },
     {
-      "epoch": 0.2969706956688913,
+      "epoch": 0.306838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.648937224981327e-05,
-      "loss": 0.8435,
+      "learning_rate": 1.6243004173044224e-05,
+      "loss": 0.9765,
       "step": 10813
     },
     {
-      "epoch": 0.2969981598967345,
+      "epoch": 0.3068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.648869544304029e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.6242286177274103e-05,
+      "loss": 0.8608,
       "step": 10814
     },
     {
-      "epoch": 0.2970256241245777,
+      "epoch": 0.3068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.648801858492624e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.6241568128775113e-05,
+      "loss": 0.9856,
       "step": 10815
     },
     {
-      "epoch": 0.29705308835242095,
+      "epoch": 0.3069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.6487341675476476e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.6240850027553313e-05,
+      "loss": 1.0641,
       "step": 10816
     },
     {
-      "epoch": 0.2970805525802642,
+      "epoch": 0.3069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.648666471469635e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.6240131873614774e-05,
+      "loss": 0.8108,
       "step": 10817
     },
     {
-      "epoch": 0.29710801680810744,
+      "epoch": 0.3069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.6485987702591216e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.623941366696556e-05,
+      "loss": 0.9292,
       "step": 10818
     },
     {
-      "epoch": 0.29713548103595067,
+      "epoch": 0.30700908059023835,
       "grad_norm": 0.0,
-      "learning_rate": 1.6485310639166437e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.6238695407611737e-05,
+      "loss": 1.0324,
       "step": 10819
     },
     {
-      "epoch": 0.2971629452637939,
+      "epoch": 0.30703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.648463352442737e-05,
-      "loss": 1.0535,
+      "learning_rate": 1.6237977095559374e-05,
+      "loss": 1.019,
       "step": 10820
     },
     {
-      "epoch": 0.29719040949163716,
+      "epoch": 0.3070658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.648395635837937e-05,
-      "loss": 1.0262,
+      "learning_rate": 1.623725873081454e-05,
+      "loss": 0.9693,
       "step": 10821
     },
     {
-      "epoch": 0.2972178737194804,
+      "epoch": 0.30709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.6483279141027795e-05,
-      "loss": 1.0181,
+      "learning_rate": 1.62365403133833e-05,
+      "loss": 1.0234,
       "step": 10822
     },
     {
-      "epoch": 0.2972453379473236,
+      "epoch": 0.30712258796821795,
       "grad_norm": 0.0,
-      "learning_rate": 1.6482601872378006e-05,
-      "loss": 1.014,
+      "learning_rate": 1.6235821843271723e-05,
+      "loss": 0.8978,
       "step": 10823
     },
     {
-      "epoch": 0.29727280217516683,
+      "epoch": 0.3071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.6481924552435357e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.6235103320485878e-05,
+      "loss": 0.9305,
       "step": 10824
     },
     {
-      "epoch": 0.2973002664030101,
+      "epoch": 0.3071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.6481247181205212e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.6234384745031838e-05,
+      "loss": 0.9513,
       "step": 10825
     },
     {
-      "epoch": 0.2973277306308533,
+      "epoch": 0.3072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.6480569758692925e-05,
-      "loss": 0.8477,
+      "learning_rate": 1.6233666116915667e-05,
+      "loss": 0.8904,
       "step": 10826
     },
     {
-      "epoch": 0.29735519485869655,
+      "epoch": 0.3072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.6479892284903865e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.623294743614344e-05,
+      "loss": 0.9208,
       "step": 10827
     },
     {
-      "epoch": 0.29738265908653977,
+      "epoch": 0.3072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.6479214759843385e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.6232228702721223e-05,
+      "loss": 0.9305,
       "step": 10828
     },
     {
-      "epoch": 0.297410123314383,
+      "epoch": 0.3072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.6478537183516847e-05,
-      "loss": 0.9148,
+      "learning_rate": 1.623150991665509e-05,
+      "loss": 1.0305,
       "step": 10829
     },
     {
-      "epoch": 0.29743758754222627,
+      "epoch": 0.3073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.6477859555929615e-05,
-      "loss": 1.01,
+      "learning_rate": 1.6230791077951117e-05,
+      "loss": 0.9831,
       "step": 10830
     },
     {
-      "epoch": 0.2974650517700695,
+      "epoch": 0.30734960272417705,
       "grad_norm": 0.0,
-      "learning_rate": 1.647718187708705e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.6230072186615367e-05,
+      "loss": 1.0435,
       "step": 10831
     },
     {
-      "epoch": 0.2974925159979127,
+      "epoch": 0.30737797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.6476504146994506e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.6229353242653922e-05,
+      "loss": 0.9376,
       "step": 10832
     },
     {
-      "epoch": 0.29751998022575593,
+      "epoch": 0.3074063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.647582636565736e-05,
-      "loss": 1.0132,
+      "learning_rate": 1.6228634246072848e-05,
+      "loss": 1.0735,
       "step": 10833
     },
     {
-      "epoch": 0.2975474444535992,
+      "epoch": 0.30743473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.6475148533080963e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.622791519687822e-05,
+      "loss": 1.0211,
       "step": 10834
     },
     {
-      "epoch": 0.29757490868144243,
+      "epoch": 0.30746311010215666,
       "grad_norm": 0.0,
-      "learning_rate": 1.6474470649270684e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.6227196095076113e-05,
+      "loss": 0.9936,
       "step": 10835
     },
     {
-      "epoch": 0.29760237290928565,
+      "epoch": 0.3074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.6473792714231886e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.62264769406726e-05,
+      "loss": 0.9876,
       "step": 10836
     },
     {
-      "epoch": 0.29762983713712887,
+      "epoch": 0.30751986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.6473114727969932e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.6225757733673754e-05,
+      "loss": 0.9192,
       "step": 10837
     },
     {
-      "epoch": 0.29765730136497215,
+      "epoch": 0.3075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.6472436690490186e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.6225038474085657e-05,
+      "loss": 0.8431,
       "step": 10838
     },
     {
-      "epoch": 0.29768476559281537,
+      "epoch": 0.3075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.6471758601798016e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.6224319161914378e-05,
+      "loss": 0.9332,
       "step": 10839
     },
     {
-      "epoch": 0.2977122298206586,
+      "epoch": 0.3076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.6471080461898783e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.6223599797165996e-05,
+      "loss": 0.9275,
       "step": 10840
     },
     {
-      "epoch": 0.2977396940485018,
+      "epoch": 0.307633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.6470402270797852e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.6222880379846584e-05,
+      "loss": 0.9825,
       "step": 10841
     },
     {
-      "epoch": 0.29776715827634503,
+      "epoch": 0.3076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.6469724028500597e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.6222160909962226e-05,
+      "loss": 1.0269,
       "step": 10842
     },
     {
-      "epoch": 0.2977946225041883,
+      "epoch": 0.30769012485811575,
       "grad_norm": 0.0,
-      "learning_rate": 1.646904573501238e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.622144138751899e-05,
+      "loss": 0.9896,
       "step": 10843
     },
     {
-      "epoch": 0.29782208673203153,
+      "epoch": 0.30771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.6468367390338562e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.622072181252296e-05,
+      "loss": 0.9271,
       "step": 10844
     },
     {
-      "epoch": 0.29784955095987475,
+      "epoch": 0.3077468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.646768899448452e-05,
-      "loss": 0.9183,
+      "learning_rate": 1.6220002184980218e-05,
+      "loss": 1.0423,
       "step": 10845
     },
     {
-      "epoch": 0.297877015187718,
+      "epoch": 0.30777525539160044,
       "grad_norm": 0.0,
-      "learning_rate": 1.6467010547455613e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.6219282504896834e-05,
+      "loss": 0.919,
       "step": 10846
     },
     {
-      "epoch": 0.29790447941556125,
+      "epoch": 0.30780363223609536,
       "grad_norm": 0.0,
-      "learning_rate": 1.6466332049257215e-05,
-      "loss": 0.8747,
+      "learning_rate": 1.6218562772278895e-05,
+      "loss": 0.8627,
       "step": 10847
     },
     {
-      "epoch": 0.2979319436434045,
+      "epoch": 0.3078320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.6465653499894697e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.6217842987132472e-05,
+      "loss": 0.9222,
       "step": 10848
     },
     {
-      "epoch": 0.2979594078712477,
+      "epoch": 0.30786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.646497489937342e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.6217123149463653e-05,
+      "loss": 0.9644,
       "step": 10849
     },
     {
-      "epoch": 0.2979868720990909,
+      "epoch": 0.30788876276958005,
       "grad_norm": 0.0,
-      "learning_rate": 1.646429624769876e-05,
-      "loss": 0.9436,
+      "learning_rate": 1.6216403259278514e-05,
+      "loss": 1.0583,
       "step": 10850
     },
     {
-      "epoch": 0.2980143363269342,
+      "epoch": 0.3079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.6463617544876084e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.6215683316583133e-05,
+      "loss": 0.9552,
       "step": 10851
     },
     {
-      "epoch": 0.2980418005547774,
+      "epoch": 0.3079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.6462938790910755e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.62149633213836e-05,
+      "loss": 0.9914,
       "step": 10852
     },
     {
-      "epoch": 0.29806926478262064,
+      "epoch": 0.3079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.6462259985808157e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.6214243273685994e-05,
+      "loss": 1.1143,
       "step": 10853
     },
     {
-      "epoch": 0.29809672901046386,
+      "epoch": 0.3080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.6461581129573655e-05,
-      "loss": 1.0459,
+      "learning_rate": 1.6213523173496393e-05,
+      "loss": 0.8773,
       "step": 10854
     },
     {
-      "epoch": 0.2981241932383071,
+      "epoch": 0.30803064699205446,
       "grad_norm": 0.0,
-      "learning_rate": 1.646090222221262e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.6212803020820883e-05,
+      "loss": 0.9367,
       "step": 10855
     },
     {
-      "epoch": 0.29815165746615035,
+      "epoch": 0.3080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.6460223263730425e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.621208281566555e-05,
+      "loss": 0.9533,
       "step": 10856
     },
     {
-      "epoch": 0.2981791216939936,
+      "epoch": 0.3080874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.6459544254132437e-05,
-      "loss": 0.929,
+      "learning_rate": 1.6211362558036468e-05,
+      "loss": 1.0315,
       "step": 10857
     },
     {
-      "epoch": 0.2982065859218368,
+      "epoch": 0.30811577752553915,
       "grad_norm": 0.0,
-      "learning_rate": 1.6458865193424033e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.621064224793973e-05,
+      "loss": 0.9283,
       "step": 10858
     },
     {
-      "epoch": 0.29823405014968,
+      "epoch": 0.30814415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.6458186081610587e-05,
-      "loss": 0.9049,
+      "learning_rate": 1.6209921885381418e-05,
+      "loss": 1.0201,
       "step": 10859
     },
     {
-      "epoch": 0.2982615143775233,
+      "epoch": 0.3081725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 1.6457506918697473e-05,
-      "loss": 1.0879,
+      "learning_rate": 1.6209201470367618e-05,
+      "loss": 0.9283,
       "step": 10860
     },
     {
-      "epoch": 0.2982889786053665,
+      "epoch": 0.30820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.6456827704690065e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.6208481002904412e-05,
+      "loss": 0.9649,
       "step": 10861
     },
     {
-      "epoch": 0.29831644283320974,
+      "epoch": 0.30822928490351875,
       "grad_norm": 0.0,
-      "learning_rate": 1.645614843959373e-05,
-      "loss": 1.007,
+      "learning_rate": 1.620776048299789e-05,
+      "loss": 0.98,
       "step": 10862
     },
     {
-      "epoch": 0.29834390706105296,
+      "epoch": 0.3082576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.6455469123413846e-05,
-      "loss": 0.8342,
+      "learning_rate": 1.6207039910654133e-05,
+      "loss": 1.0453,
       "step": 10863
     },
     {
-      "epoch": 0.29837137128889624,
+      "epoch": 0.3082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.6454789756155792e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.6206319285879233e-05,
+      "loss": 0.8112,
       "step": 10864
     },
     {
-      "epoch": 0.29839883551673946,
+      "epoch": 0.3083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.6454110337824945e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.6205598608679272e-05,
+      "loss": 0.9188,
       "step": 10865
     },
     {
-      "epoch": 0.2984262997445827,
+      "epoch": 0.3083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.6453430868426675e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.6204877879060346e-05,
+      "loss": 0.9817,
       "step": 10866
     },
     {
-      "epoch": 0.2984537639724259,
+      "epoch": 0.3083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.6452751347966358e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.6204157097028534e-05,
+      "loss": 0.8538,
       "step": 10867
     },
     {
-      "epoch": 0.2984812282002691,
+      "epoch": 0.3083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.6452071776449372e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.6203436262589927e-05,
+      "loss": 1.0651,
       "step": 10868
     },
     {
-      "epoch": 0.2985086924281124,
+      "epoch": 0.308427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.64513921538811e-05,
-      "loss": 0.9129,
+      "learning_rate": 1.620271537575062e-05,
+      "loss": 0.9377,
       "step": 10869
     },
     {
-      "epoch": 0.2985361566559556,
+      "epoch": 0.30845629965947785,
       "grad_norm": 0.0,
-      "learning_rate": 1.645071248026691e-05,
-      "loss": 0.9755,
+      "learning_rate": 1.6201994436516693e-05,
+      "loss": 0.9546,
       "step": 10870
     },
     {
-      "epoch": 0.29856362088379884,
+      "epoch": 0.30848467650397277,
       "grad_norm": 0.0,
-      "learning_rate": 1.6450032755612184e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.6201273444894243e-05,
+      "loss": 0.914,
       "step": 10871
     },
     {
-      "epoch": 0.29859108511164206,
+      "epoch": 0.3085130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.64493529799223e-05,
-      "loss": 0.9115,
+      "learning_rate": 1.6200552400889355e-05,
+      "loss": 0.9735,
       "step": 10872
     },
     {
-      "epoch": 0.29861854933948534,
+      "epoch": 0.30854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.644867315320264e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.6199831304508126e-05,
+      "loss": 1.0009,
       "step": 10873
     },
     {
-      "epoch": 0.29864601356732856,
+      "epoch": 0.30856980703745746,
       "grad_norm": 0.0,
-      "learning_rate": 1.6447993275458577e-05,
-      "loss": 0.9364,
+      "learning_rate": 1.6199110155756636e-05,
+      "loss": 0.9111,
       "step": 10874
     },
     {
-      "epoch": 0.2986734777951718,
+      "epoch": 0.3085981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.6447313346695498e-05,
-      "loss": 0.9965,
+      "learning_rate": 1.619838895464099e-05,
+      "loss": 1.0148,
       "step": 10875
     },
     {
-      "epoch": 0.298700942023015,
+      "epoch": 0.30862656072644723,
       "grad_norm": 0.0,
-      "learning_rate": 1.6446633366918772e-05,
-      "loss": 1.024,
+      "learning_rate": 1.6197667701167272e-05,
+      "loss": 0.9979,
       "step": 10876
     },
     {
-      "epoch": 0.2987284062508583,
+      "epoch": 0.3086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.644595333613379e-05,
-      "loss": 0.9016,
+      "learning_rate": 1.6196946395341576e-05,
+      "loss": 0.8987,
       "step": 10877
     },
     {
-      "epoch": 0.2987558704787015,
+      "epoch": 0.308683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.6445273254345928e-05,
-      "loss": 0.8581,
+      "learning_rate": 1.6196225037169995e-05,
+      "loss": 0.8376,
       "step": 10878
     },
     {
-      "epoch": 0.2987833347065447,
+      "epoch": 0.3087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.6444593121560567e-05,
-      "loss": 1.0177,
+      "learning_rate": 1.6195503626658623e-05,
+      "loss": 0.9161,
       "step": 10879
     },
     {
-      "epoch": 0.29881079893438794,
+      "epoch": 0.3087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.6443912937783088e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.6194782163813557e-05,
+      "loss": 0.9623,
       "step": 10880
     },
     {
-      "epoch": 0.2988382631622312,
+      "epoch": 0.3087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.6443232703018874e-05,
-      "loss": 1.0395,
+      "learning_rate": 1.6194060648640882e-05,
+      "loss": 0.9459,
       "step": 10881
     },
     {
-      "epoch": 0.29886572739007444,
+      "epoch": 0.30879682179341655,
       "grad_norm": 0.0,
-      "learning_rate": 1.6442552417273305e-05,
-      "loss": 0.9089,
+      "learning_rate": 1.6193339081146698e-05,
+      "loss": 0.9654,
       "step": 10882
     },
     {
-      "epoch": 0.29889319161791766,
+      "epoch": 0.30882519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.644187208055177e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.6192617461337102e-05,
+      "loss": 0.9209,
       "step": 10883
     },
     {
-      "epoch": 0.2989206558457609,
+      "epoch": 0.3088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.6441191692859645e-05,
-      "loss": 0.853,
+      "learning_rate": 1.619189578921819e-05,
+      "loss": 1.042,
       "step": 10884
     },
     {
-      "epoch": 0.2989481200736041,
+      "epoch": 0.30888195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.6440511254202315e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.6191174064796056e-05,
+      "loss": 0.9866,
       "step": 10885
     },
     {
-      "epoch": 0.2989755843014474,
+      "epoch": 0.30891032917139616,
       "grad_norm": 0.0,
-      "learning_rate": 1.6439830764585166e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.6190452288076795e-05,
+      "loss": 0.9733,
       "step": 10886
     },
     {
-      "epoch": 0.2990030485292906,
+      "epoch": 0.308938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.6439150224013584e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.6189730459066505e-05,
+      "loss": 0.9308,
       "step": 10887
     },
     {
-      "epoch": 0.2990305127571338,
+      "epoch": 0.30896708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.643846963249295e-05,
-      "loss": 1.0792,
+      "learning_rate": 1.6189008577771286e-05,
+      "loss": 0.914,
       "step": 10888
     },
     {
-      "epoch": 0.29905797698497705,
+      "epoch": 0.3089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.643778899002865e-05,
-      "loss": 0.8841,
+      "learning_rate": 1.6188286644197227e-05,
+      "loss": 0.9786,
       "step": 10889
     },
     {
-      "epoch": 0.2990854412128203,
+      "epoch": 0.3090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.6437108296626067e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.6187564658350438e-05,
+      "loss": 0.9831,
       "step": 10890
     },
     {
-      "epoch": 0.29911290544066355,
+      "epoch": 0.3090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.6436427552290592e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.618684262023701e-05,
+      "loss": 0.8934,
       "step": 10891
     },
     {
-      "epoch": 0.29914036966850677,
+      "epoch": 0.3090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.643574675702761e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.6186120529863045e-05,
+      "loss": 0.9721,
       "step": 10892
     },
     {
-      "epoch": 0.29916783389635,
+      "epoch": 0.3091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.6435065910842506e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.6185398387234644e-05,
+      "loss": 0.9868,
       "step": 10893
     },
     {
-      "epoch": 0.29919529812419327,
+      "epoch": 0.30913734392735526,
       "grad_norm": 0.0,
-      "learning_rate": 1.6434385013740665e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.61846761923579e-05,
+      "loss": 0.9452,
       "step": 10894
     },
     {
-      "epoch": 0.2992227623520365,
+      "epoch": 0.30916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.6433704065727482e-05,
-      "loss": 0.9011,
+      "learning_rate": 1.6183953945238924e-05,
+      "loss": 0.8914,
       "step": 10895
     },
     {
-      "epoch": 0.2992502265798797,
+      "epoch": 0.3091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.6433023066808338e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.6183231645883804e-05,
+      "loss": 0.8788,
       "step": 10896
     },
     {
-      "epoch": 0.29927769080772293,
+      "epoch": 0.30922247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.643234201698862e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.6182509294298654e-05,
+      "loss": 0.9312,
       "step": 10897
     },
     {
-      "epoch": 0.29930515503556615,
+      "epoch": 0.30925085130533486,
       "grad_norm": 0.0,
-      "learning_rate": 1.6431660916273728e-05,
-      "loss": 0.8811,
+      "learning_rate": 1.618178689048957e-05,
+      "loss": 1.0237,
       "step": 10898
     },
     {
-      "epoch": 0.2993326192634094,
+      "epoch": 0.3092792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.643097976466904e-05,
-      "loss": 1.0421,
+      "learning_rate": 1.618106443446265e-05,
+      "loss": 0.9251,
       "step": 10899
     },
     {
-      "epoch": 0.29936008349125265,
+      "epoch": 0.30930760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.6430298562179947e-05,
-      "loss": 0.9735,
+      "learning_rate": 1.6180341926224006e-05,
+      "loss": 0.9162,
       "step": 10900
     },
     {
-      "epoch": 0.29938754771909587,
+      "epoch": 0.3093359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.6429617308811844e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.6179619365779732e-05,
+      "loss": 0.9769,
       "step": 10901
     },
     {
-      "epoch": 0.2994150119469391,
+      "epoch": 0.3093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.6428936004570112e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.6178896753135938e-05,
+      "loss": 1.0652,
       "step": 10902
     },
     {
-      "epoch": 0.29944247617478237,
+      "epoch": 0.3093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.6428254649460155e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.6178174088298725e-05,
+      "loss": 0.9721,
       "step": 10903
     },
     {
-      "epoch": 0.2994699404026256,
+      "epoch": 0.3094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.6427573243487357e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.6177451371274197e-05,
+      "loss": 0.9834,
       "step": 10904
     },
     {
-      "epoch": 0.2994974046304688,
+      "epoch": 0.3094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.6426891786657105e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.6176728602068458e-05,
+      "loss": 0.9244,
       "step": 10905
     },
     {
-      "epoch": 0.29952486885831203,
+      "epoch": 0.30947786606129396,
       "grad_norm": 0.0,
-      "learning_rate": 1.6426210278974797e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.6176005780687617e-05,
+      "loss": 1.0288,
       "step": 10906
     },
     {
-      "epoch": 0.2995523330861553,
+      "epoch": 0.3095062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.6425528720445825e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.6175282907137773e-05,
+      "loss": 1.004,
       "step": 10907
     },
     {
-      "epoch": 0.29957979731399853,
+      "epoch": 0.3095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.6424847111075582e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.6174559981425045e-05,
+      "loss": 0.8921,
       "step": 10908
     },
     {
-      "epoch": 0.29960726154184175,
+      "epoch": 0.30956299659477865,
       "grad_norm": 0.0,
-      "learning_rate": 1.6424165450869456e-05,
-      "loss": 0.904,
+      "learning_rate": 1.6173837003555525e-05,
+      "loss": 0.9913,
       "step": 10909
     },
     {
-      "epoch": 0.299634725769685,
+      "epoch": 0.30959137343927357,
       "grad_norm": 0.0,
-      "learning_rate": 1.642348373983285e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.6173113973535327e-05,
+      "loss": 1.0405,
       "step": 10910
     },
     {
-      "epoch": 0.2996621899975282,
+      "epoch": 0.3096197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.6422801977971147e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.617239089137056e-05,
+      "loss": 0.9445,
       "step": 10911
     },
     {
-      "epoch": 0.29968965422537147,
+      "epoch": 0.30964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.642212016528975e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.6171667757067328e-05,
+      "loss": 0.9157,
       "step": 10912
     },
     {
-      "epoch": 0.2997171184532147,
+      "epoch": 0.30967650397275825,
       "grad_norm": 0.0,
-      "learning_rate": 1.6421438301794054e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.6170944570631743e-05,
+      "loss": 0.9606,
       "step": 10913
     },
     {
-      "epoch": 0.2997445826810579,
+      "epoch": 0.3097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.6420756387489442e-05,
-      "loss": 1.0522,
+      "learning_rate": 1.617022133206991e-05,
+      "loss": 0.8976,
       "step": 10914
     },
     {
-      "epoch": 0.29977204690890114,
+      "epoch": 0.30973325766174803,
       "grad_norm": 0.0,
-      "learning_rate": 1.6420074422381325e-05,
-      "loss": 1.0424,
+      "learning_rate": 1.616949804138794e-05,
+      "loss": 0.9787,
       "step": 10915
     },
     {
-      "epoch": 0.2997995111367444,
+      "epoch": 0.3097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.641939240647509e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.6168774698591943e-05,
+      "loss": 1.0429,
       "step": 10916
     },
     {
-      "epoch": 0.29982697536458763,
+      "epoch": 0.3097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.6418710339776137e-05,
-      "loss": 1.0825,
+      "learning_rate": 1.6168051303688028e-05,
+      "loss": 0.8945,
       "step": 10917
     },
     {
-      "epoch": 0.29985443959243085,
+      "epoch": 0.30981838819523266,
       "grad_norm": 0.0,
-      "learning_rate": 1.6418028222289856e-05,
-      "loss": 0.8814,
+      "learning_rate": 1.616732785668231e-05,
+      "loss": 1.0163,
       "step": 10918
     },
     {
-      "epoch": 0.2998819038202741,
+      "epoch": 0.3098467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.641734605402166e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.6166604357580894e-05,
+      "loss": 0.8299,
       "step": 10919
     },
     {
-      "epoch": 0.29990936804811735,
+      "epoch": 0.3098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.6416663834976928e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.616588080638989e-05,
+      "loss": 1.0983,
       "step": 10920
     },
     {
-      "epoch": 0.2999368322759606,
+      "epoch": 0.30990351872871735,
       "grad_norm": 0.0,
-      "learning_rate": 1.6415981565161065e-05,
-      "loss": 0.941,
+      "learning_rate": 1.616515720311542e-05,
+      "loss": 1.0114,
       "step": 10921
     },
     {
-      "epoch": 0.2999642965038038,
+      "epoch": 0.30993189557321227,
       "grad_norm": 0.0,
-      "learning_rate": 1.6415299244579475e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.6164433547763584e-05,
+      "loss": 0.8673,
       "step": 10922
     },
     {
-      "epoch": 0.299991760731647,
+      "epoch": 0.30996027241770713,
       "grad_norm": 0.0,
-      "learning_rate": 1.6414616873237552e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.6163709840340503e-05,
+      "loss": 0.9635,
       "step": 10923
     },
     {
-      "epoch": 0.30001922495949024,
+      "epoch": 0.30998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.6413934451140693e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.616298608085229e-05,
+      "loss": 0.9265,
       "step": 10924
     },
     {
-      "epoch": 0.3000466891873335,
+      "epoch": 0.31001702610669696,
       "grad_norm": 0.0,
-      "learning_rate": 1.6413251978294306e-05,
+      "learning_rate": 1.6162262269305054e-05,
       "loss": 0.9002,
       "step": 10925
     },
     {
-      "epoch": 0.30007415341517674,
+      "epoch": 0.3100454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.641256945470378e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.6161538405704914e-05,
+      "loss": 0.9355,
       "step": 10926
     },
     {
-      "epoch": 0.30010161764301996,
+      "epoch": 0.31007377979568673,
       "grad_norm": 0.0,
-      "learning_rate": 1.641188688037452e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.616081449005798e-05,
+      "loss": 0.8767,
       "step": 10927
     },
     {
-      "epoch": 0.3001290818708632,
+      "epoch": 0.3101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.641120425531193e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.616009052237037e-05,
+      "loss": 0.9931,
       "step": 10928
     },
     {
-      "epoch": 0.30015654609870646,
+      "epoch": 0.3101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.6410521579521405e-05,
-      "loss": 0.9444,
+      "learning_rate": 1.61593665026482e-05,
+      "loss": 0.9416,
       "step": 10929
     },
     {
-      "epoch": 0.3001840103265497,
+      "epoch": 0.3101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.6409838853008352e-05,
-      "loss": 0.8924,
+      "learning_rate": 1.615864243089758e-05,
+      "loss": 1.0902,
       "step": 10930
     },
     {
-      "epoch": 0.3002114745543929,
+      "epoch": 0.3101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.640915607577817e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.6157918307124635e-05,
+      "loss": 0.9362,
       "step": 10931
     },
     {
-      "epoch": 0.3002389387822361,
+      "epoch": 0.3102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.6408473247836263e-05,
-      "loss": 0.8178,
+      "learning_rate": 1.6157194131335473e-05,
+      "loss": 0.943,
       "step": 10932
     },
     {
-      "epoch": 0.3002664030100794,
+      "epoch": 0.31024404086265606,
       "grad_norm": 0.0,
-      "learning_rate": 1.6407790369188036e-05,
-      "loss": 1.0277,
+      "learning_rate": 1.6156469903536217e-05,
+      "loss": 0.9126,
       "step": 10933
     },
     {
-      "epoch": 0.3002938672379226,
+      "epoch": 0.31027241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 1.6407107439838887e-05,
-      "loss": 0.9172,
+      "learning_rate": 1.6155745623732987e-05,
+      "loss": 0.7945,
       "step": 10934
     },
     {
-      "epoch": 0.30032133146576584,
+      "epoch": 0.31030079455164583,
       "grad_norm": 0.0,
-      "learning_rate": 1.6406424459794222e-05,
-      "loss": 0.9295,
+      "learning_rate": 1.6155021291931895e-05,
+      "loss": 0.98,
       "step": 10935
     },
     {
-      "epoch": 0.30034879569360906,
+      "epoch": 0.31032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.6405741429059444e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.6154296908139057e-05,
+      "loss": 0.9103,
       "step": 10936
     },
     {
-      "epoch": 0.3003762599214523,
+      "epoch": 0.31035754824063566,
       "grad_norm": 0.0,
-      "learning_rate": 1.6405058347639965e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.6153572472360602e-05,
+      "loss": 0.9235,
       "step": 10937
     },
     {
-      "epoch": 0.30040372414929556,
+      "epoch": 0.3103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.6404375215541174e-05,
-      "loss": 0.8051,
+      "learning_rate": 1.615284798460264e-05,
+      "loss": 1.0215,
       "step": 10938
     },
     {
-      "epoch": 0.3004311883771388,
+      "epoch": 0.31041430192962544,
       "grad_norm": 0.0,
-      "learning_rate": 1.640369203276849e-05,
-      "loss": 0.8919,
+      "learning_rate": 1.6152123444871296e-05,
+      "loss": 0.9962,
       "step": 10939
     },
     {
-      "epoch": 0.300458652604982,
+      "epoch": 0.3104426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.6403008799327317e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.615139885317269e-05,
+      "loss": 0.9537,
       "step": 10940
     },
     {
-      "epoch": 0.3004861168328252,
+      "epoch": 0.3104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.6402325515223053e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.615067420951294e-05,
+      "loss": 1.0112,
       "step": 10941
     },
     {
-      "epoch": 0.3005135810606685,
+      "epoch": 0.3104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.6401642180461112e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.6149949513898165e-05,
+      "loss": 0.9578,
       "step": 10942
     },
     {
-      "epoch": 0.3005410452885117,
+      "epoch": 0.310527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.6400958795046896e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.6149224766334492e-05,
+      "loss": 1.0146,
       "step": 10943
     },
     {
-      "epoch": 0.30056850951635494,
+      "epoch": 0.3105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.6400275358985815e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.6148499966828045e-05,
+      "loss": 0.9717,
       "step": 10944
     },
     {
-      "epoch": 0.30059597374419816,
+      "epoch": 0.31058456299659476,
       "grad_norm": 0.0,
-      "learning_rate": 1.6399591872283277e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.6147775115384942e-05,
+      "loss": 1.0149,
       "step": 10945
     },
     {
-      "epoch": 0.30062343797204144,
+      "epoch": 0.3106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.639890833494469e-05,
-      "loss": 0.9541,
+      "learning_rate": 1.6147050212011303e-05,
+      "loss": 0.9683,
       "step": 10946
     },
     {
-      "epoch": 0.30065090219988466,
+      "epoch": 0.3106413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.6398224746975463e-05,
-      "loss": 0.8507,
+      "learning_rate": 1.6146325256713253e-05,
+      "loss": 0.911,
       "step": 10947
     },
     {
-      "epoch": 0.3006783664277279,
+      "epoch": 0.31066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.6397541108381002e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.6145600249496922e-05,
+      "loss": 0.9695,
       "step": 10948
     },
     {
-      "epoch": 0.3007058306555711,
+      "epoch": 0.31069807037457436,
       "grad_norm": 0.0,
-      "learning_rate": 1.6396857419166717e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.6144875190368426e-05,
+      "loss": 0.9639,
       "step": 10949
     },
     {
-      "epoch": 0.3007332948834143,
+      "epoch": 0.3107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.6396173679338015e-05,
-      "loss": 0.9087,
+      "learning_rate": 1.6144150079333892e-05,
+      "loss": 1.0632,
       "step": 10950
     },
     {
-      "epoch": 0.3007607591112576,
+      "epoch": 0.31075482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.6395489888900314e-05,
-      "loss": 1.0737,
+      "learning_rate": 1.6143424916399448e-05,
+      "loss": 0.8776,
       "step": 10951
     },
     {
-      "epoch": 0.3007882233391008,
+      "epoch": 0.310783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.639480604785902e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.6142699701571218e-05,
+      "loss": 0.9749,
       "step": 10952
     },
     {
-      "epoch": 0.30081568756694405,
+      "epoch": 0.3108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.639412215621954e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.614197443485533e-05,
+      "loss": 0.8743,
       "step": 10953
     },
     {
-      "epoch": 0.30084315179478727,
+      "epoch": 0.31083995459704883,
       "grad_norm": 0.0,
-      "learning_rate": 1.639343821398729e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.6141249116257906e-05,
+      "loss": 0.8668,
       "step": 10954
     },
     {
-      "epoch": 0.30087061602263054,
+      "epoch": 0.3108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.639275422116768e-05,
-      "loss": 0.9053,
+      "learning_rate": 1.614052374578507e-05,
+      "loss": 0.8708,
       "step": 10955
     },
     {
-      "epoch": 0.30089808025047377,
+      "epoch": 0.3108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.6392070177766123e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.613979832344296e-05,
+      "loss": 0.933,
       "step": 10956
     },
     {
-      "epoch": 0.300925544478317,
+      "epoch": 0.31092508513053346,
       "grad_norm": 0.0,
-      "learning_rate": 1.639138608378803e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.6139072849237692e-05,
+      "loss": 0.8665,
       "step": 10957
     },
     {
-      "epoch": 0.3009530087061602,
+      "epoch": 0.3109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.6390701939238814e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.6138347323175402e-05,
+      "loss": 0.7799,
       "step": 10958
     },
     {
-      "epoch": 0.3009804729340035,
+      "epoch": 0.3109818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.639001774412389e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.613762174526222e-05,
+      "loss": 0.9221,
       "step": 10959
     },
     {
-      "epoch": 0.3010079371618467,
+      "epoch": 0.31101021566401815,
       "grad_norm": 0.0,
-      "learning_rate": 1.6389333498448673e-05,
-      "loss": 0.8618,
+      "learning_rate": 1.613689611550427e-05,
+      "loss": 0.95,
       "step": 10960
     },
     {
-      "epoch": 0.3010354013896899,
+      "epoch": 0.31103859250851307,
       "grad_norm": 0.0,
-      "learning_rate": 1.638864920221857e-05,
-      "loss": 0.9796,
+      "learning_rate": 1.613617043390768e-05,
+      "loss": 0.916,
       "step": 10961
     },
     {
-      "epoch": 0.30106286561753315,
+      "epoch": 0.3110669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 1.6387964855438998e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.6135444700478583e-05,
+      "loss": 1.0871,
       "step": 10962
     },
     {
-      "epoch": 0.30109032984537637,
+      "epoch": 0.31109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.6387280458115378e-05,
-      "loss": 1.014,
+      "learning_rate": 1.613471891522311e-05,
+      "loss": 1.0554,
       "step": 10963
     },
     {
-      "epoch": 0.30111779407321965,
+      "epoch": 0.31112372304199776,
       "grad_norm": 0.0,
-      "learning_rate": 1.638659601025312e-05,
-      "loss": 0.8571,
+      "learning_rate": 1.613399307814739e-05,
+      "loss": 0.9566,
       "step": 10964
     },
     {
-      "epoch": 0.30114525830106287,
+      "epoch": 0.3111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.6385911511857638e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.6133267189257552e-05,
+      "loss": 0.9173,
       "step": 10965
     },
     {
-      "epoch": 0.3011727225289061,
+      "epoch": 0.31118047673098753,
       "grad_norm": 0.0,
-      "learning_rate": 1.6385226962934354e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.6132541248559732e-05,
+      "loss": 0.9852,
       "step": 10966
     },
     {
-      "epoch": 0.3012001867567493,
+      "epoch": 0.3112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.638454236348868e-05,
-      "loss": 0.9782,
+      "learning_rate": 1.6131815256060064e-05,
+      "loss": 0.9011,
       "step": 10967
     },
     {
-      "epoch": 0.3012276509845926,
+      "epoch": 0.3112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.6383857713526033e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.6131089211764673e-05,
+      "loss": 0.9869,
       "step": 10968
     },
     {
-      "epoch": 0.3012551152124358,
+      "epoch": 0.31126560726447217,
       "grad_norm": 0.0,
-      "learning_rate": 1.638317301305183e-05,
-      "loss": 0.995,
+      "learning_rate": 1.6130363115679696e-05,
+      "loss": 0.8043,
       "step": 10969
     },
     {
-      "epoch": 0.30128257944027903,
+      "epoch": 0.3112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.638248826207149e-05,
-      "loss": 0.9571,
+      "learning_rate": 1.6129636967811267e-05,
+      "loss": 0.9601,
       "step": 10970
     },
     {
-      "epoch": 0.30131004366812225,
+      "epoch": 0.311322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.638180346059043e-05,
-      "loss": 0.8674,
+      "learning_rate": 1.612891076816552e-05,
+      "loss": 1.0462,
       "step": 10971
     },
     {
-      "epoch": 0.30133750789596553,
+      "epoch": 0.31135073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.638111860861407e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.612818451674859e-05,
+      "loss": 1.0649,
       "step": 10972
     },
     {
-      "epoch": 0.30136497212380875,
+      "epoch": 0.31137911464245177,
       "grad_norm": 0.0,
-      "learning_rate": 1.6380433706147828e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.6127458213566604e-05,
+      "loss": 0.9725,
       "step": 10973
     },
     {
-      "epoch": 0.30139243635165197,
+      "epoch": 0.31140749148694663,
       "grad_norm": 0.0,
-      "learning_rate": 1.6379748753197123e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.6126731858625705e-05,
+      "loss": 0.8616,
       "step": 10974
     },
     {
-      "epoch": 0.3014199005794952,
+      "epoch": 0.31143586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.6379063749767374e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.6126005451932028e-05,
+      "loss": 0.9911,
       "step": 10975
     },
     {
-      "epoch": 0.30144736480733847,
+      "epoch": 0.31146424517593646,
       "grad_norm": 0.0,
-      "learning_rate": 1.6378378695864002e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.612527899349171e-05,
+      "loss": 0.9502,
       "step": 10976
     },
     {
-      "epoch": 0.3014748290351817,
+      "epoch": 0.3114926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.6377693591492426e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.612455248331088e-05,
+      "loss": 0.983,
       "step": 10977
     },
     {
-      "epoch": 0.3015022932630249,
+      "epoch": 0.31152099886492624,
       "grad_norm": 0.0,
-      "learning_rate": 1.637700843665807e-05,
-      "loss": 0.9115,
+      "learning_rate": 1.6123825921395682e-05,
+      "loss": 0.8939,
       "step": 10978
     },
     {
-      "epoch": 0.30152975749086813,
+      "epoch": 0.3115493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.6376323231366352e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.6123099307752252e-05,
+      "loss": 0.9427,
       "step": 10979
     },
     {
-      "epoch": 0.30155722171871135,
+      "epoch": 0.311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.6375637975622693e-05,
-      "loss": 1.0148,
+      "learning_rate": 1.6122372642386728e-05,
+      "loss": 0.9527,
       "step": 10980
     },
     {
-      "epoch": 0.30158468594655463,
+      "epoch": 0.31160612939841087,
       "grad_norm": 0.0,
-      "learning_rate": 1.637495266943252e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.6121645925305244e-05,
+      "loss": 0.9721,
       "step": 10981
     },
     {
-      "epoch": 0.30161215017439785,
+      "epoch": 0.3116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.6374267312801248e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.6120919156513944e-05,
+      "loss": 1.0254,
       "step": 10982
     },
     {
-      "epoch": 0.3016396144022411,
+      "epoch": 0.3116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.6373581905734302e-05,
-      "loss": 0.983,
+      "learning_rate": 1.6120192336018963e-05,
+      "loss": 0.8701,
       "step": 10983
     },
     {
-      "epoch": 0.3016670786300843,
+      "epoch": 0.31169125993189556,
       "grad_norm": 0.0,
-      "learning_rate": 1.6372896448237116e-05,
-      "loss": 1.0324,
+      "learning_rate": 1.611946546382644e-05,
+      "loss": 1.0251,
       "step": 10984
     },
     {
-      "epoch": 0.3016945428579276,
+      "epoch": 0.3117196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.63722109403151e-05,
-      "loss": 0.8038,
+      "learning_rate": 1.6118738539942524e-05,
+      "loss": 1.028,
       "step": 10985
     },
     {
-      "epoch": 0.3017220070857708,
+      "epoch": 0.31174801362088533,
       "grad_norm": 0.0,
-      "learning_rate": 1.637152538197368e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.6118011564373347e-05,
+      "loss": 1.0859,
       "step": 10986
     },
     {
-      "epoch": 0.301749471313614,
+      "epoch": 0.31177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.6370839773218284e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.6117284537125047e-05,
+      "loss": 0.8818,
       "step": 10987
     },
     {
-      "epoch": 0.30177693554145724,
+      "epoch": 0.31180476730987516,
       "grad_norm": 0.0,
-      "learning_rate": 1.6370154114054337e-05,
-      "loss": 0.9562,
+      "learning_rate": 1.6116557458203774e-05,
+      "loss": 0.9316,
       "step": 10988
     },
     {
-      "epoch": 0.3018043997693005,
+      "epoch": 0.31183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.6369468404487264e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.611583032761566e-05,
+      "loss": 1.014,
       "step": 10989
     },
     {
-      "epoch": 0.30183186399714373,
+      "epoch": 0.31186152099886494,
       "grad_norm": 0.0,
-      "learning_rate": 1.636878264452249e-05,
-      "loss": 0.8424,
+      "learning_rate": 1.611510314536685e-05,
+      "loss": 1.0561,
       "step": 10990
     },
     {
-      "epoch": 0.30185932822498696,
+      "epoch": 0.3118898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.636809683416544e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.61143759114635e-05,
+      "loss": 1.086,
       "step": 10991
     },
     {
-      "epoch": 0.3018867924528302,
+      "epoch": 0.3119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.6367410973421538e-05,
-      "loss": 1.0562,
+      "learning_rate": 1.6113648625911733e-05,
+      "loss": 0.9797,
       "step": 10992
     },
     {
-      "epoch": 0.3019142566806734,
+      "epoch": 0.31194665153234963,
       "grad_norm": 0.0,
-      "learning_rate": 1.636672506229622e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.6112921288717706e-05,
+      "loss": 0.8768,
       "step": 10993
     },
     {
-      "epoch": 0.3019417209085167,
+      "epoch": 0.3119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.63660391007949e-05,
-      "loss": 0.973,
+      "learning_rate": 1.6112193899887555e-05,
+      "loss": 0.9704,
       "step": 10994
     },
     {
-      "epoch": 0.3019691851363599,
+      "epoch": 0.3120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.636535308892302e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.6111466459427428e-05,
+      "loss": 0.8743,
       "step": 10995
     },
     {
-      "epoch": 0.3019966493642031,
+      "epoch": 0.31203178206583426,
       "grad_norm": 0.0,
-      "learning_rate": 1.6364667026685996e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.611073896734347e-05,
+      "loss": 1.0666,
       "step": 10996
     },
     {
-      "epoch": 0.30202411359204634,
+      "epoch": 0.3120601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.6363980914089262e-05,
-      "loss": 0.9446,
+      "learning_rate": 1.611001142364182e-05,
+      "loss": 1.0364,
       "step": 10997
     },
     {
-      "epoch": 0.3020515778198896,
+      "epoch": 0.31208853575482404,
       "grad_norm": 0.0,
-      "learning_rate": 1.6363294751138247e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.6109283828328636e-05,
+      "loss": 0.9242,
       "step": 10998
     },
     {
-      "epoch": 0.30207904204773284,
+      "epoch": 0.31211691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.636260853783838e-05,
-      "loss": 0.955,
+      "learning_rate": 1.6108556181410056e-05,
+      "loss": 0.9723,
       "step": 10999
     },
     {
-      "epoch": 0.30210650627557606,
+      "epoch": 0.31214528944381387,
       "grad_norm": 0.0,
-      "learning_rate": 1.6361922274195085e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.6107828482892225e-05,
+      "loss": 0.9255,
       "step": 11000
     },
     {
-      "epoch": 0.3021339705034193,
+      "epoch": 0.3121736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.63612359602138e-05,
-      "loss": 0.8796,
+      "learning_rate": 1.6107100732781295e-05,
+      "loss": 0.9785,
       "step": 11001
     },
     {
-      "epoch": 0.30216143473126256,
+      "epoch": 0.31220204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.636054959589995e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.610637293108341e-05,
+      "loss": 0.9166,
       "step": 11002
     },
     {
-      "epoch": 0.3021888989591058,
+      "epoch": 0.3122304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.6359863181258967e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.6105645077804713e-05,
+      "loss": 0.8083,
       "step": 11003
     },
     {
-      "epoch": 0.302216363186949,
+      "epoch": 0.3122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.6359176716296287e-05,
-      "loss": 0.9503,
+      "learning_rate": 1.6104917172951363e-05,
+      "loss": 1.0451,
       "step": 11004
     },
     {
-      "epoch": 0.3022438274147922,
+      "epoch": 0.31228717366628833,
       "grad_norm": 0.0,
-      "learning_rate": 1.6358490201017336e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.61041892165295e-05,
+      "loss": 1.0045,
       "step": 11005
     },
     {
-      "epoch": 0.30227129164263544,
+      "epoch": 0.3123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.635780363542755e-05,
-      "loss": 1.1022,
+      "learning_rate": 1.6103461208545277e-05,
+      "loss": 0.921,
       "step": 11006
     },
     {
-      "epoch": 0.3022987558704787,
+      "epoch": 0.3123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.6357117019532352e-05,
-      "loss": 0.9018,
+      "learning_rate": 1.6102733149004847e-05,
+      "loss": 0.9498,
       "step": 11007
     },
     {
-      "epoch": 0.30232622009832194,
+      "epoch": 0.31237230419977297,
       "grad_norm": 0.0,
-      "learning_rate": 1.6356430353337185e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.610200503791435e-05,
+      "loss": 1.0286,
       "step": 11008
     },
     {
-      "epoch": 0.30235368432616516,
+      "epoch": 0.3124006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.635574363684748e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.6101276875279946e-05,
+      "loss": 1.0224,
       "step": 11009
     },
     {
-      "epoch": 0.3023811485540084,
+      "epoch": 0.3124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.6355056870068667e-05,
-      "loss": 1.047,
+      "learning_rate": 1.610054866110778e-05,
+      "loss": 0.9473,
       "step": 11010
     },
     {
-      "epoch": 0.30240861278185166,
+      "epoch": 0.31245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.6354370053006185e-05,
-      "loss": 1.004,
+      "learning_rate": 1.6099820395404005e-05,
+      "loss": 0.9296,
       "step": 11011
     },
     {
-      "epoch": 0.3024360770096949,
+      "epoch": 0.31248581157775257,
       "grad_norm": 0.0,
-      "learning_rate": 1.6353683185665464e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.6099092078174773e-05,
+      "loss": 0.9406,
       "step": 11012
     },
     {
-      "epoch": 0.3024635412375381,
+      "epoch": 0.31251418842224743,
       "grad_norm": 0.0,
-      "learning_rate": 1.635299626805194e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.6098363709426238e-05,
+      "loss": 0.8775,
       "step": 11013
     },
     {
-      "epoch": 0.3024910054653813,
+      "epoch": 0.31254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.6352309300171052e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.6097635289164548e-05,
+      "loss": 0.885,
       "step": 11014
     },
     {
-      "epoch": 0.3025184696932246,
+      "epoch": 0.3125709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.6351622282028227e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.6096906817395862e-05,
+      "loss": 0.7901,
       "step": 11015
     },
     {
-      "epoch": 0.3025459339210678,
+      "epoch": 0.3125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.635093521362891e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.6096178294126326e-05,
+      "loss": 1.032,
       "step": 11016
     },
     {
-      "epoch": 0.30257339814891104,
+      "epoch": 0.31262769580022703,
       "grad_norm": 0.0,
-      "learning_rate": 1.6350248094978532e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.60954497193621e-05,
+      "loss": 0.8896,
       "step": 11017
     },
     {
-      "epoch": 0.30260086237675426,
+      "epoch": 0.3126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.634956092608253e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.6094721093109335e-05,
+      "loss": 1.007,
       "step": 11018
     },
     {
-      "epoch": 0.3026283266045975,
+      "epoch": 0.3126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.6348873706946347e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.609399241537419e-05,
+      "loss": 0.9253,
       "step": 11019
     },
     {
-      "epoch": 0.30265579083244076,
+      "epoch": 0.31271282633371167,
       "grad_norm": 0.0,
-      "learning_rate": 1.6348186437575413e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.6093263686162813e-05,
+      "loss": 0.9374,
       "step": 11020
     },
     {
-      "epoch": 0.302683255060284,
+      "epoch": 0.3127412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.634749911797517e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.6092534905481367e-05,
+      "loss": 0.8542,
       "step": 11021
     },
     {
-      "epoch": 0.3027107192881272,
+      "epoch": 0.3127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.6346811748151054e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.6091806073336e-05,
+      "loss": 1.0214,
       "step": 11022
     },
     {
-      "epoch": 0.3027381835159704,
+      "epoch": 0.31279795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 1.63461243281085e-05,
-      "loss": 0.955,
+      "learning_rate": 1.609107718973288e-05,
+      "loss": 0.9655,
       "step": 11023
     },
     {
-      "epoch": 0.3027656477438137,
+      "epoch": 0.3128263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.634543685785296e-05,
-      "loss": 0.992,
+      "learning_rate": 1.6090348254678154e-05,
+      "loss": 1.0406,
       "step": 11024
     },
     {
-      "epoch": 0.3027931119716569,
+      "epoch": 0.31285471055618613,
       "grad_norm": 0.0,
-      "learning_rate": 1.6344749337389862e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.6089619268177985e-05,
+      "loss": 1.0,
       "step": 11025
     },
     {
-      "epoch": 0.30282057619950015,
+      "epoch": 0.31288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.634406176672465e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.6088890230238524e-05,
+      "loss": 0.9974,
       "step": 11026
     },
     {
-      "epoch": 0.30284804042734337,
+      "epoch": 0.31291146424517596,
       "grad_norm": 0.0,
-      "learning_rate": 1.6343374145862762e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.6088161140865936e-05,
+      "loss": 1.0058,
       "step": 11027
     },
     {
-      "epoch": 0.30287550465518664,
+      "epoch": 0.3129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.6342686474809645e-05,
-      "loss": 0.9935,
+      "learning_rate": 1.6087432000066376e-05,
+      "loss": 0.9915,
       "step": 11028
     },
     {
-      "epoch": 0.30290296888302987,
+      "epoch": 0.31296821793416574,
       "grad_norm": 0.0,
-      "learning_rate": 1.6341998753570734e-05,
-      "loss": 0.8915,
+      "learning_rate": 1.6086702807846007e-05,
+      "loss": 1.0405,
       "step": 11029
     },
     {
-      "epoch": 0.3029304331108731,
+      "epoch": 0.3129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.634131098215147e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.608597356421098e-05,
+      "loss": 0.9461,
       "step": 11030
     },
     {
-      "epoch": 0.3029578973387163,
+      "epoch": 0.3130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.6340623160557298e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.6085244269167467e-05,
+      "loss": 1.0045,
       "step": 11031
     },
     {
-      "epoch": 0.30298536156655953,
+      "epoch": 0.31305334846765037,
       "grad_norm": 0.0,
-      "learning_rate": 1.633993528879366e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.608451492272162e-05,
+      "loss": 0.9756,
       "step": 11032
     },
     {
-      "epoch": 0.3030128257944028,
+      "epoch": 0.3130817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.6339247366865998e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.60837855248796e-05,
+      "loss": 1.0856,
       "step": 11033
     },
     {
-      "epoch": 0.30304029002224603,
+      "epoch": 0.3131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.6338559394779757e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.6083056075647572e-05,
+      "loss": 0.9481,
       "step": 11034
     },
     {
-      "epoch": 0.30306775425008925,
+      "epoch": 0.31313847900113506,
       "grad_norm": 0.0,
-      "learning_rate": 1.6337871372540376e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.6082326575031696e-05,
+      "loss": 0.9498,
       "step": 11035
     },
     {
-      "epoch": 0.30309521847793247,
+      "epoch": 0.31316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.63371833001533e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.608159702303813e-05,
+      "loss": 0.9146,
       "step": 11036
     },
     {
-      "epoch": 0.30312268270577575,
+      "epoch": 0.31319523269012484,
       "grad_norm": 0.0,
-      "learning_rate": 1.633649517762398e-05,
-      "loss": 0.9876,
+      "learning_rate": 1.6080867419673045e-05,
+      "loss": 0.9818,
       "step": 11037
     },
     {
-      "epoch": 0.30315014693361897,
+      "epoch": 0.31322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.6335807004957853e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.6080137764942597e-05,
+      "loss": 0.8999,
       "step": 11038
     },
     {
-      "epoch": 0.3031776111614622,
+      "epoch": 0.31325198637911467,
       "grad_norm": 0.0,
-      "learning_rate": 1.6335118782160366e-05,
-      "loss": 0.9645,
+      "learning_rate": 1.6079408058852956e-05,
+      "loss": 0.8853,
       "step": 11039
     },
     {
-      "epoch": 0.3032050753893054,
+      "epoch": 0.3132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.6334430509236966e-05,
-      "loss": 1.0768,
+      "learning_rate": 1.6078678301410276e-05,
+      "loss": 0.9154,
       "step": 11040
     },
     {
-      "epoch": 0.3032325396171487,
+      "epoch": 0.31330874006810444,
       "grad_norm": 0.0,
-      "learning_rate": 1.63337421861931e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.6077948492620734e-05,
+      "loss": 0.9655,
       "step": 11041
     },
     {
-      "epoch": 0.3032600038449919,
+      "epoch": 0.3133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.6333053813034206e-05,
-      "loss": 0.8524,
+      "learning_rate": 1.6077218632490484e-05,
+      "loss": 1.0774,
       "step": 11042
     },
     {
-      "epoch": 0.30328746807283513,
+      "epoch": 0.3133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.6332365389765745e-05,
-      "loss": 0.9295,
+      "learning_rate": 1.6076488721025695e-05,
+      "loss": 0.8804,
       "step": 11043
     },
     {
-      "epoch": 0.30331493230067835,
+      "epoch": 0.31339387060158913,
       "grad_norm": 0.0,
-      "learning_rate": 1.633167691639315e-05,
-      "loss": 1.046,
+      "learning_rate": 1.6075758758232535e-05,
+      "loss": 0.9253,
       "step": 11044
     },
     {
-      "epoch": 0.3033423965285216,
+      "epoch": 0.313422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.6330988392921878e-05,
-      "loss": 1.0331,
+      "learning_rate": 1.607502874411716e-05,
+      "loss": 1.0716,
       "step": 11045
     },
     {
-      "epoch": 0.30336986075636485,
+      "epoch": 0.3134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.6330299819357375e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.6074298678685753e-05,
+      "loss": 1.0073,
       "step": 11046
     },
     {
-      "epoch": 0.30339732498420807,
+      "epoch": 0.31347900113507376,
       "grad_norm": 0.0,
-      "learning_rate": 1.6329611195705086e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.6073568561944467e-05,
+      "loss": 0.9696,
       "step": 11047
     },
     {
-      "epoch": 0.3034247892120513,
+      "epoch": 0.3135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.632892252197046e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.6072838393899476e-05,
+      "loss": 0.9328,
       "step": 11048
     },
     {
-      "epoch": 0.3034522534398945,
+      "epoch": 0.31353575482406354,
       "grad_norm": 0.0,
-      "learning_rate": 1.632823379815895e-05,
-      "loss": 0.8815,
+      "learning_rate": 1.607210817455695e-05,
+      "loss": 0.8406,
       "step": 11049
     },
     {
-      "epoch": 0.3034797176677378,
+      "epoch": 0.31356413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.6327545024276e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.6071377903923048e-05,
+      "loss": 0.9985,
       "step": 11050
     },
     {
-      "epoch": 0.303507181895581,
+      "epoch": 0.31359250851305337,
       "grad_norm": 0.0,
-      "learning_rate": 1.632685620032706e-05,
-      "loss": 0.9301,
+      "learning_rate": 1.6070647582003945e-05,
+      "loss": 0.9758,
       "step": 11051
     },
     {
-      "epoch": 0.30353464612342423,
+      "epoch": 0.31362088535754823,
       "grad_norm": 0.0,
-      "learning_rate": 1.632616732631759e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.606991720880581e-05,
+      "loss": 0.9196,
       "step": 11052
     },
     {
-      "epoch": 0.30356211035126746,
+      "epoch": 0.31364926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 1.632547840225303e-05,
-      "loss": 0.9701,
+      "learning_rate": 1.606918678433481e-05,
+      "loss": 0.9909,
       "step": 11053
     },
     {
-      "epoch": 0.30358957457911073,
+      "epoch": 0.313677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.6324789428138834e-05,
-      "loss": 0.8998,
+      "learning_rate": 1.6068456308597115e-05,
+      "loss": 0.8399,
       "step": 11054
     },
     {
-      "epoch": 0.30361703880695395,
+      "epoch": 0.3137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.6324100403980457e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.60677257815989e-05,
+      "loss": 0.9225,
       "step": 11055
     },
     {
-      "epoch": 0.3036445030347972,
+      "epoch": 0.31373439273552783,
       "grad_norm": 0.0,
-      "learning_rate": 1.6323411329783344e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.6066995203346332e-05,
+      "loss": 0.9672,
       "step": 11056
     },
     {
-      "epoch": 0.3036719672626404,
+      "epoch": 0.3137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.6322722205552954e-05,
-      "loss": 1.0972,
+      "learning_rate": 1.606626457384558e-05,
+      "loss": 0.9852,
       "step": 11057
     },
     {
-      "epoch": 0.3036994314904836,
+      "epoch": 0.3137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.6322033031294738e-05,
-      "loss": 1.0071,
+      "learning_rate": 1.6065533893102823e-05,
+      "loss": 0.9684,
       "step": 11058
     },
     {
-      "epoch": 0.3037268957183269,
+      "epoch": 0.31381952326901247,
       "grad_norm": 0.0,
-      "learning_rate": 1.6321343807014144e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.6064803161124224e-05,
+      "loss": 0.8776,
       "step": 11059
     },
     {
-      "epoch": 0.3037543599461701,
+      "epoch": 0.3138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.6320654532716627e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.6064072377915965e-05,
+      "loss": 1.0068,
       "step": 11060
     },
     {
-      "epoch": 0.30378182417401334,
+      "epoch": 0.31387627695800224,
       "grad_norm": 0.0,
-      "learning_rate": 1.631996520840765e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.606334154348421e-05,
+      "loss": 0.908,
       "step": 11061
     },
     {
-      "epoch": 0.30380928840185656,
+      "epoch": 0.31390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.6319275834092655e-05,
-      "loss": 1.0889,
+      "learning_rate": 1.6062610657835137e-05,
+      "loss": 0.8927,
       "step": 11062
     },
     {
-      "epoch": 0.30383675262969984,
+      "epoch": 0.3139330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 1.6318586409777103e-05,
-      "loss": 1.0892,
+      "learning_rate": 1.606187972097492e-05,
+      "loss": 0.81,
       "step": 11063
     },
     {
-      "epoch": 0.30386421685754306,
+      "epoch": 0.31396140749148693,
       "grad_norm": 0.0,
-      "learning_rate": 1.631789693546645e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.6061148732909734e-05,
+      "loss": 0.9355,
       "step": 11064
     },
     {
-      "epoch": 0.3038916810853863,
+      "epoch": 0.31398978433598185,
       "grad_norm": 0.0,
-      "learning_rate": 1.6317207411166146e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.6060417693645753e-05,
+      "loss": 0.9021,
       "step": 11065
     },
     {
-      "epoch": 0.3039191453132295,
+      "epoch": 0.3140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.6316517836881652e-05,
-      "loss": 1.0841,
+      "learning_rate": 1.6059686603189146e-05,
+      "loss": 0.9673,
       "step": 11066
     },
     {
-      "epoch": 0.3039466095410728,
+      "epoch": 0.3140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.631582821261842e-05,
-      "loss": 0.978,
+      "learning_rate": 1.6058955461546098e-05,
+      "loss": 0.9302,
       "step": 11067
     },
     {
-      "epoch": 0.303974073768916,
+      "epoch": 0.31407491486946654,
       "grad_norm": 0.0,
-      "learning_rate": 1.631513853838191e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.605822426872278e-05,
+      "loss": 0.9526,
       "step": 11068
     },
     {
-      "epoch": 0.3040015379967592,
+      "epoch": 0.3141032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.6314448814177578e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.6057493024725367e-05,
+      "loss": 0.9101,
       "step": 11069
     },
     {
-      "epoch": 0.30402900222460244,
+      "epoch": 0.3141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.6313759040010883e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.6056761729560043e-05,
+      "loss": 0.8477,
       "step": 11070
     },
     {
-      "epoch": 0.3040564664524457,
+      "epoch": 0.31416004540295117,
       "grad_norm": 0.0,
-      "learning_rate": 1.6313069215887275e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.6056030383232978e-05,
+      "loss": 1.0448,
       "step": 11071
     },
     {
-      "epoch": 0.30408393068028894,
+      "epoch": 0.3141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.6312379341812223e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.605529898575035e-05,
+      "loss": 0.9504,
       "step": 11072
     },
     {
-      "epoch": 0.30411139490813216,
+      "epoch": 0.314216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.631168941779118e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.605456753711834e-05,
+      "loss": 0.835,
       "step": 11073
     },
     {
-      "epoch": 0.3041388591359754,
+      "epoch": 0.31424517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.6310999443829602e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.6053836037343125e-05,
+      "loss": 1.0737,
       "step": 11074
     },
     {
-      "epoch": 0.3041663233638186,
+      "epoch": 0.3142735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.6310309419932955e-05,
-      "loss": 0.9947,
+      "learning_rate": 1.6053104486430887e-05,
+      "loss": 0.943,
       "step": 11075
     },
     {
-      "epoch": 0.3041937875916619,
+      "epoch": 0.31430192962542564,
       "grad_norm": 0.0,
-      "learning_rate": 1.6309619346106694e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.6052372884387804e-05,
+      "loss": 0.9262,
       "step": 11076
     },
     {
-      "epoch": 0.3042212518195051,
+      "epoch": 0.31433030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.630892922235628e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.6051641231220055e-05,
+      "loss": 0.9447,
       "step": 11077
     },
     {
-      "epoch": 0.3042487160473483,
+      "epoch": 0.3143586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.6308239048687177e-05,
-      "loss": 0.9812,
+      "learning_rate": 1.605090952693382e-05,
+      "loss": 1.0289,
       "step": 11078
     },
     {
-      "epoch": 0.30427618027519154,
+      "epoch": 0.3143870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.630754882510484e-05,
-      "loss": 1.011,
+      "learning_rate": 1.605017777153528e-05,
+      "loss": 0.908,
       "step": 11079
     },
     {
-      "epoch": 0.3043036445030348,
+      "epoch": 0.31441543700340524,
       "grad_norm": 0.0,
-      "learning_rate": 1.6306858551614736e-05,
-      "loss": 1.064,
+      "learning_rate": 1.6049445965030616e-05,
+      "loss": 0.9991,
       "step": 11080
     },
     {
-      "epoch": 0.30433110873087804,
+      "epoch": 0.3144438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.6306168228222323e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.604871410742601e-05,
+      "loss": 1.0103,
       "step": 11081
     },
     {
-      "epoch": 0.30435857295872126,
+      "epoch": 0.314472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.6305477854933066e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.604798219872764e-05,
+      "loss": 0.9304,
       "step": 11082
     },
     {
-      "epoch": 0.3043860371865645,
+      "epoch": 0.3145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.630478743175242e-05,
-      "loss": 0.8736,
+      "learning_rate": 1.60472502389417e-05,
+      "loss": 1.0641,
       "step": 11083
     },
     {
-      "epoch": 0.30441350141440776,
+      "epoch": 0.3145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.6304096958685857e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.604651822807436e-05,
+      "loss": 0.9549,
       "step": 11084
     },
     {
-      "epoch": 0.304440965642251,
+      "epoch": 0.3145573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.630340643573884e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.604578616613181e-05,
+      "loss": 0.9227,
       "step": 11085
     },
     {
-      "epoch": 0.3044684298700942,
+      "epoch": 0.31458569807037456,
       "grad_norm": 0.0,
-      "learning_rate": 1.6302715862916824e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.6045054053120235e-05,
+      "loss": 0.9078,
       "step": 11086
     },
     {
-      "epoch": 0.3044958940979374,
+      "epoch": 0.3146140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.6302025240225282e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.6044321889045814e-05,
+      "loss": 0.925,
       "step": 11087
     },
     {
-      "epoch": 0.30452335832578065,
+      "epoch": 0.31464245175936434,
       "grad_norm": 0.0,
-      "learning_rate": 1.6301334567669676e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.6043589673914737e-05,
+      "loss": 0.938,
       "step": 11088
     },
     {
-      "epoch": 0.3045508225536239,
+      "epoch": 0.31467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.6300643845255468e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.6042857407733184e-05,
+      "loss": 0.8815,
       "step": 11089
     },
     {
-      "epoch": 0.30457828678146714,
+      "epoch": 0.31469920544835417,
       "grad_norm": 0.0,
-      "learning_rate": 1.6299953072988126e-05,
-      "loss": 1.0677,
+      "learning_rate": 1.6042125090507343e-05,
+      "loss": 0.903,
       "step": 11090
     },
     {
-      "epoch": 0.30460575100931037,
+      "epoch": 0.31472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.6299262250873115e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.60413927222434e-05,
+      "loss": 0.9866,
       "step": 11091
     },
     {
-      "epoch": 0.3046332152371536,
+      "epoch": 0.31475595913734394,
       "grad_norm": 0.0,
-      "learning_rate": 1.62985713789159e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.604066030294754e-05,
+      "loss": 0.9017,
       "step": 11092
     },
     {
-      "epoch": 0.30466067946499686,
+      "epoch": 0.3147843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.6297880457121947e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.6039927832625953e-05,
+      "loss": 1.0692,
       "step": 11093
     },
     {
-      "epoch": 0.3046881436928401,
+      "epoch": 0.3148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.6297189485496728e-05,
-      "loss": 0.8913,
+      "learning_rate": 1.6039195311284825e-05,
+      "loss": 0.8818,
       "step": 11094
     },
     {
-      "epoch": 0.3047156079206833,
+      "epoch": 0.3148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.6296498464045706e-05,
-      "loss": 0.9114,
+      "learning_rate": 1.6038462738930342e-05,
+      "loss": 0.8592,
       "step": 11095
     },
     {
-      "epoch": 0.30474307214852653,
+      "epoch": 0.3148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.6295807392774348e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.6037730115568687e-05,
+      "loss": 0.9781,
       "step": 11096
     },
     {
-      "epoch": 0.3047705363763698,
+      "epoch": 0.3148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.629511627168812e-05,
-      "loss": 0.8719,
+      "learning_rate": 1.6036997441206063e-05,
+      "loss": 1.0569,
       "step": 11097
     },
     {
-      "epoch": 0.304798000604213,
+      "epoch": 0.31492622020431327,
       "grad_norm": 0.0,
-      "learning_rate": 1.62944251007925e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.6036264715848644e-05,
+      "loss": 1.014,
       "step": 11098
     },
     {
-      "epoch": 0.30482546483205625,
+      "epoch": 0.3149545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.6293733880092945e-05,
-      "loss": 0.9567,
+      "learning_rate": 1.603553193950263e-05,
+      "loss": 0.9374,
       "step": 11099
     },
     {
-      "epoch": 0.30485292905989947,
+      "epoch": 0.31498297389330304,
       "grad_norm": 0.0,
-      "learning_rate": 1.629304260959493e-05,
-      "loss": 0.9116,
+      "learning_rate": 1.6034799112174205e-05,
+      "loss": 0.9561,
       "step": 11100
     },
     {
-      "epoch": 0.3048803932877427,
+      "epoch": 0.31501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.6292351289303926e-05,
-      "loss": 1.024,
+      "learning_rate": 1.6034066233869557e-05,
+      "loss": 0.9005,
       "step": 11101
     },
     {
-      "epoch": 0.30490785751558597,
+      "epoch": 0.31503972758229287,
       "grad_norm": 0.0,
-      "learning_rate": 1.6291659919225402e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.6033333304594886e-05,
+      "loss": 0.9993,
       "step": 11102
     },
     {
-      "epoch": 0.3049353217434292,
+      "epoch": 0.31506810442678773,
       "grad_norm": 0.0,
-      "learning_rate": 1.6290968499364828e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.6032600324356373e-05,
+      "loss": 0.8974,
       "step": 11103
     },
     {
-      "epoch": 0.3049627859712724,
+      "epoch": 0.31509648127128265,
       "grad_norm": 0.0,
-      "learning_rate": 1.629027702972767e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.6031867293160215e-05,
+      "loss": 1.0151,
       "step": 11104
     },
     {
-      "epoch": 0.30499025019911563,
+      "epoch": 0.3151248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.628958551031941e-05,
-      "loss": 0.9542,
+      "learning_rate": 1.6031134211012603e-05,
+      "loss": 0.944,
       "step": 11105
     },
     {
-      "epoch": 0.3050177144269589,
+      "epoch": 0.3151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.628889394114551e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.6030401077919727e-05,
+      "loss": 1.004,
       "step": 11106
     },
     {
-      "epoch": 0.30504517865480213,
+      "epoch": 0.31518161180476734,
       "grad_norm": 0.0,
-      "learning_rate": 1.6288202322211445e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.6029667893887787e-05,
+      "loss": 0.8523,
       "step": 11107
     },
     {
-      "epoch": 0.30507264288264535,
+      "epoch": 0.3152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.6287510653522684e-05,
-      "loss": 0.9136,
+      "learning_rate": 1.6028934658922967e-05,
+      "loss": 0.8255,
       "step": 11108
     },
     {
-      "epoch": 0.30510010711048857,
+      "epoch": 0.3152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.6286818935084706e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.6028201373031466e-05,
+      "loss": 0.924,
       "step": 11109
     },
     {
-      "epoch": 0.30512757133833185,
+      "epoch": 0.31526674233825197,
       "grad_norm": 0.0,
-      "learning_rate": 1.628612716690298e-05,
-      "loss": 0.8855,
+      "learning_rate": 1.602746803621948e-05,
+      "loss": 1.0176,
       "step": 11110
     },
     {
-      "epoch": 0.30515503556617507,
+      "epoch": 0.3152951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.6285435348982982e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.6026734648493195e-05,
+      "loss": 0.924,
       "step": 11111
     },
     {
-      "epoch": 0.3051824997940183,
+      "epoch": 0.31532349602724175,
       "grad_norm": 0.0,
-      "learning_rate": 1.6284743481330187e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.6026001209858818e-05,
+      "loss": 0.9449,
       "step": 11112
     },
     {
-      "epoch": 0.3052099640218615,
+      "epoch": 0.31535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.6284051563950065e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.602526772032253e-05,
+      "loss": 0.982,
       "step": 11113
     },
     {
-      "epoch": 0.30523742824970473,
+      "epoch": 0.3153802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.6283359596848094e-05,
-      "loss": 0.9541,
+      "learning_rate": 1.6024534179890542e-05,
+      "loss": 1.1372,
       "step": 11114
     },
     {
-      "epoch": 0.305264892477548,
+      "epoch": 0.31540862656072643,
       "grad_norm": 0.0,
-      "learning_rate": 1.6282667580029747e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.602380058856904e-05,
+      "loss": 0.8728,
       "step": 11115
     },
     {
-      "epoch": 0.30529235670539123,
+      "epoch": 0.31543700340522135,
       "grad_norm": 0.0,
-      "learning_rate": 1.62819755135005e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.6023066946364223e-05,
+      "loss": 0.9366,
       "step": 11116
     },
     {
-      "epoch": 0.30531982093323445,
+      "epoch": 0.3154653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.628128339726583e-05,
-      "loss": 0.9481,
+      "learning_rate": 1.602233325328229e-05,
+      "loss": 0.9782,
       "step": 11117
     },
     {
-      "epoch": 0.3053472851610777,
+      "epoch": 0.3154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.628059123133121e-05,
-      "loss": 0.974,
+      "learning_rate": 1.602159950932944e-05,
+      "loss": 0.969,
       "step": 11118
     },
     {
-      "epoch": 0.30537474938892095,
+      "epoch": 0.31552213393870604,
       "grad_norm": 0.0,
-      "learning_rate": 1.6279899015702124e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.6020865714511863e-05,
+      "loss": 1.0166,
       "step": 11119
     },
     {
-      "epoch": 0.3054022136167642,
+      "epoch": 0.3155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.627920675038404e-05,
-      "loss": 0.9063,
+      "learning_rate": 1.6020131868835762e-05,
+      "loss": 0.8088,
       "step": 11120
     },
     {
-      "epoch": 0.3054296778446074,
+      "epoch": 0.3155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.627851443538244e-05,
-      "loss": 0.991,
+      "learning_rate": 1.601939797230734e-05,
+      "loss": 0.8344,
       "step": 11121
     },
     {
-      "epoch": 0.3054571420724506,
+      "epoch": 0.3156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.6277822070702804e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.6018664024932792e-05,
+      "loss": 1.022,
       "step": 11122
     },
     {
-      "epoch": 0.3054846063002939,
+      "epoch": 0.3156356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.627712965635061e-05,
-      "loss": 0.931,
+      "learning_rate": 1.601793002671832e-05,
+      "loss": 0.9823,
       "step": 11123
     },
     {
-      "epoch": 0.3055120705281371,
+      "epoch": 0.3156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.6276437192331333e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.601719597767012e-05,
+      "loss": 0.9303,
       "step": 11124
     },
     {
-      "epoch": 0.30553953475598034,
+      "epoch": 0.31569239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.6275744678650452e-05,
-      "loss": 0.9266,
+      "learning_rate": 1.6016461877794397e-05,
+      "loss": 0.8902,
       "step": 11125
     },
     {
-      "epoch": 0.30556699898382356,
+      "epoch": 0.3157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.6275052115313447e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.601572772709735e-05,
+      "loss": 0.9354,
       "step": 11126
     },
     {
-      "epoch": 0.3055944632116668,
+      "epoch": 0.31574914869466514,
       "grad_norm": 0.0,
-      "learning_rate": 1.6274359502325804e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.6014993525585176e-05,
+      "loss": 1.074,
       "step": 11127
     },
     {
-      "epoch": 0.30562192743951005,
+      "epoch": 0.31577752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.6273666839692993e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.6014259273264085e-05,
+      "loss": 0.9233,
       "step": 11128
     },
     {
-      "epoch": 0.3056493916673533,
+      "epoch": 0.3158059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.6272974127420505e-05,
-      "loss": 0.9291,
+      "learning_rate": 1.6013524970140276e-05,
+      "loss": 0.9483,
       "step": 11129
     },
     {
-      "epoch": 0.3056768558951965,
+      "epoch": 0.31583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.627228136551381e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.601279061621995e-05,
+      "loss": 0.8837,
       "step": 11130
     },
     {
-      "epoch": 0.3057043201230397,
+      "epoch": 0.31586265607264474,
       "grad_norm": 0.0,
-      "learning_rate": 1.6271588553978398e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.6012056211509312e-05,
+      "loss": 0.8457,
       "step": 11131
     },
     {
-      "epoch": 0.305731784350883,
+      "epoch": 0.3158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.6270895692819747e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.6011321756014566e-05,
+      "loss": 1.0807,
       "step": 11132
     },
     {
-      "epoch": 0.3057592485787262,
+      "epoch": 0.3159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.6270202782043342e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.6010587249741915e-05,
+      "loss": 0.8654,
       "step": 11133
     },
     {
-      "epoch": 0.30578671280656944,
+      "epoch": 0.3159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.6269509821654663e-05,
-      "loss": 1.0848,
+      "learning_rate": 1.600985269269756e-05,
+      "loss": 0.9647,
       "step": 11134
     },
     {
-      "epoch": 0.30581417703441266,
+      "epoch": 0.3159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.6268816811659194e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.6009118084887712e-05,
+      "loss": 0.8101,
       "step": 11135
     },
     {
-      "epoch": 0.30584164126225594,
+      "epoch": 0.3160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.6268123752062414e-05,
-      "loss": 0.9483,
+      "learning_rate": 1.600838342631857e-05,
+      "loss": 0.9031,
       "step": 11136
     },
     {
-      "epoch": 0.30586910549009916,
+      "epoch": 0.31603291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.6267430642869816e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.6007648716996348e-05,
+      "loss": 0.9034,
       "step": 11137
     },
     {
-      "epoch": 0.3058965697179424,
+      "epoch": 0.316061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.6266737484086877e-05,
-      "loss": 0.9317,
+      "learning_rate": 1.6006913956927242e-05,
+      "loss": 0.9301,
       "step": 11138
     },
     {
-      "epoch": 0.3059240339457856,
+      "epoch": 0.31608967082860384,
       "grad_norm": 0.0,
-      "learning_rate": 1.6266044275719084e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.600617914611747e-05,
+      "loss": 1.014,
       "step": 11139
     },
     {
-      "epoch": 0.3059514981736288,
+      "epoch": 0.31611804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.626535101777192e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.600544428457323e-05,
+      "loss": 0.8989,
       "step": 11140
     },
     {
-      "epoch": 0.3059789624014721,
+      "epoch": 0.3161464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.6264657710250875e-05,
-      "loss": 0.9166,
+      "learning_rate": 1.6004709372300732e-05,
+      "loss": 0.8125,
       "step": 11141
     },
     {
-      "epoch": 0.3060064266293153,
+      "epoch": 0.31617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.626396435316143e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.600397440930618e-05,
+      "loss": 0.9567,
       "step": 11142
     },
     {
-      "epoch": 0.30603389085715854,
+      "epoch": 0.31620317820658345,
       "grad_norm": 0.0,
-      "learning_rate": 1.626327094650907e-05,
-      "loss": 1.0599,
+      "learning_rate": 1.600323939559579e-05,
+      "loss": 0.9727,
       "step": 11143
     },
     {
-      "epoch": 0.30606135508500176,
+      "epoch": 0.3162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.626257749029929e-05,
-      "loss": 0.8276,
+      "learning_rate": 1.600250433117577e-05,
+      "loss": 0.9701,
       "step": 11144
     },
     {
-      "epoch": 0.30608881931284504,
+      "epoch": 0.3162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.6261883984537567e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.6001769216052323e-05,
+      "loss": 0.9313,
       "step": 11145
     },
     {
-      "epoch": 0.30611628354068826,
+      "epoch": 0.3162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.626119042922939e-05,
-      "loss": 0.929,
+      "learning_rate": 1.600103405023166e-05,
+      "loss": 0.9257,
       "step": 11146
     },
     {
-      "epoch": 0.3061437477685315,
+      "epoch": 0.316316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.6260496824380254e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.6000298833719996e-05,
+      "loss": 0.8689,
       "step": 11147
     },
     {
-      "epoch": 0.3061712119963747,
+      "epoch": 0.3163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.6259803169995638e-05,
-      "loss": 0.8775,
+      "learning_rate": 1.5999563566523537e-05,
+      "loss": 0.926,
       "step": 11148
     },
     {
-      "epoch": 0.306198676224218,
+      "epoch": 0.31637343927355277,
       "grad_norm": 0.0,
-      "learning_rate": 1.6259109466081034e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.5998828248648494e-05,
+      "loss": 0.8802,
       "step": 11149
     },
     {
-      "epoch": 0.3062261404520612,
+      "epoch": 0.3164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.6258415712641935e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.5998092880101076e-05,
+      "loss": 0.9551,
       "step": 11150
     },
     {
-      "epoch": 0.3062536046799044,
+      "epoch": 0.31643019296254254,
       "grad_norm": 0.0,
-      "learning_rate": 1.6257721909683827e-05,
-      "loss": 0.992,
+      "learning_rate": 1.5997357460887502e-05,
+      "loss": 1.0917,
       "step": 11151
     },
     {
-      "epoch": 0.30628106890774764,
+      "epoch": 0.31645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.6257028057212197e-05,
-      "loss": 1.0602,
+      "learning_rate": 1.5996621991013982e-05,
+      "loss": 0.9923,
       "step": 11152
     },
     {
-      "epoch": 0.30630853313559087,
+      "epoch": 0.3164869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.6256334155232535e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.5995886470486723e-05,
+      "loss": 0.8795,
       "step": 11153
     },
     {
-      "epoch": 0.30633599736343414,
+      "epoch": 0.31651532349602723,
       "grad_norm": 0.0,
-      "learning_rate": 1.6255640203750337e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.5995150899311938e-05,
+      "loss": 0.964,
       "step": 11154
     },
     {
-      "epoch": 0.30636346159127736,
+      "epoch": 0.31654370034052215,
       "grad_norm": 0.0,
-      "learning_rate": 1.625494620277109e-05,
-      "loss": 1.033,
+      "learning_rate": 1.5994415277495852e-05,
+      "loss": 0.8153,
       "step": 11155
     },
     {
-      "epoch": 0.3063909258191206,
+      "epoch": 0.316572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.6254252152300285e-05,
-      "loss": 1.0211,
+      "learning_rate": 1.5993679605044662e-05,
+      "loss": 0.8934,
       "step": 11156
     },
     {
-      "epoch": 0.3064183900469638,
+      "epoch": 0.3166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.6253558052343416e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.5992943881964594e-05,
+      "loss": 0.8784,
       "step": 11157
     },
     {
-      "epoch": 0.3064458542748071,
+      "epoch": 0.3166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.625286390290597e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.5992208108261864e-05,
+      "loss": 0.8962,
       "step": 11158
     },
     {
-      "epoch": 0.3064733185026503,
+      "epoch": 0.3166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.6252169703993445e-05,
-      "loss": 0.925,
+      "learning_rate": 1.599147228394268e-05,
+      "loss": 0.9931,
       "step": 11159
     },
     {
-      "epoch": 0.3065007827304935,
+      "epoch": 0.3166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.625147545561133e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.599073640901326e-05,
+      "loss": 0.9657,
       "step": 11160
     },
     {
-      "epoch": 0.30652824695833675,
+      "epoch": 0.3167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.6250781157765123e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.599000048347982e-05,
+      "loss": 1.0274,
       "step": 11161
     },
     {
-      "epoch": 0.30655571118618,
+      "epoch": 0.3167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.625008681046031e-05,
-      "loss": 1.02,
+      "learning_rate": 1.5989264507348576e-05,
+      "loss": 1.003,
       "step": 11162
     },
     {
-      "epoch": 0.30658317541402325,
+      "epoch": 0.31677071509648125,
       "grad_norm": 0.0,
-      "learning_rate": 1.6249392413702393e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.598852848062575e-05,
+      "loss": 1.0076,
       "step": 11163
     },
     {
-      "epoch": 0.30661063964186647,
+      "epoch": 0.31679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.6248697967496863e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.598779240331755e-05,
+      "loss": 1.0659,
       "step": 11164
     },
     {
-      "epoch": 0.3066381038697097,
+      "epoch": 0.3168274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 1.624800347184921e-05,
-      "loss": 0.8312,
+      "learning_rate": 1.5987056275430196e-05,
+      "loss": 0.9701,
       "step": 11165
     },
     {
-      "epoch": 0.3066655680975529,
+      "epoch": 0.31685584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.6247308926764932e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.5986320096969915e-05,
+      "loss": 0.97,
       "step": 11166
     },
     {
-      "epoch": 0.3066930323253962,
+      "epoch": 0.31688422247446085,
       "grad_norm": 0.0,
-      "learning_rate": 1.624661433224953e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.5985583867942916e-05,
+      "loss": 0.9399,
       "step": 11167
     },
     {
-      "epoch": 0.3067204965532394,
+      "epoch": 0.3169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.6245919688308495e-05,
-      "loss": 0.8721,
+      "learning_rate": 1.5984847588355423e-05,
+      "loss": 1.0328,
       "step": 11168
     },
     {
-      "epoch": 0.30674796078108263,
+      "epoch": 0.3169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.6245224994947323e-05,
-      "loss": 0.871,
+      "learning_rate": 1.598411125821365e-05,
+      "loss": 1.0053,
       "step": 11169
     },
     {
-      "epoch": 0.30677542500892585,
+      "epoch": 0.31696935300794554,
       "grad_norm": 0.0,
-      "learning_rate": 1.6244530252171512e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.598337487752382e-05,
+      "loss": 1.0141,
       "step": 11170
     },
     {
-      "epoch": 0.3068028892367691,
+      "epoch": 0.3169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.6243835459986562e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.598263844629216e-05,
+      "loss": 1.0558,
       "step": 11171
     },
     {
-      "epoch": 0.30683035346461235,
+      "epoch": 0.3170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.624314061839796e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.5981901964524877e-05,
+      "loss": 0.8681,
       "step": 11172
     },
     {
-      "epoch": 0.30685781769245557,
+      "epoch": 0.3170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.624244572741122e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.5981165432228197e-05,
+      "loss": 0.8975,
       "step": 11173
     },
     {
-      "epoch": 0.3068852819202988,
+      "epoch": 0.3170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.624175078703182e-05,
-      "loss": 1.057,
+      "learning_rate": 1.598042884940835e-05,
+      "loss": 1.0048,
       "step": 11174
     },
     {
-      "epoch": 0.30691274614814207,
+      "epoch": 0.31711123723041995,
       "grad_norm": 0.0,
-      "learning_rate": 1.624105579726528e-05,
-      "loss": 1.0618,
+      "learning_rate": 1.597969221607155e-05,
+      "loss": 0.8887,
       "step": 11175
     },
     {
-      "epoch": 0.3069402103759853,
+      "epoch": 0.31713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6240360758117085e-05,
-      "loss": 1.063,
+      "learning_rate": 1.5978955532224022e-05,
+      "loss": 0.9715,
       "step": 11176
     },
     {
-      "epoch": 0.3069676746038285,
+      "epoch": 0.3171679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.623966566959274e-05,
-      "loss": 0.7973,
+      "learning_rate": 1.5978218797871987e-05,
+      "loss": 0.9063,
       "step": 11177
     },
     {
-      "epoch": 0.30699513883167173,
+      "epoch": 0.31719636776390464,
       "grad_norm": 0.0,
-      "learning_rate": 1.623897053169774e-05,
-      "loss": 0.9596,
+      "learning_rate": 1.597748201302167e-05,
+      "loss": 0.9451,
       "step": 11178
     },
     {
-      "epoch": 0.307022603059515,
+      "epoch": 0.31722474460839956,
       "grad_norm": 0.0,
-      "learning_rate": 1.623827534443759e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.597674517767929e-05,
+      "loss": 0.9375,
       "step": 11179
     },
     {
-      "epoch": 0.30705006728735823,
+      "epoch": 0.3172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.623758010781779e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.5976008291851076e-05,
+      "loss": 0.8923,
       "step": 11180
     },
     {
-      "epoch": 0.30707753151520145,
+      "epoch": 0.31728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.623688482184384e-05,
-      "loss": 0.946,
+      "learning_rate": 1.5975271355543252e-05,
+      "loss": 1.0171,
       "step": 11181
     },
     {
-      "epoch": 0.3071049957430447,
+      "epoch": 0.31730987514188425,
       "grad_norm": 0.0,
-      "learning_rate": 1.623618948652124e-05,
-      "loss": 1.0693,
+      "learning_rate": 1.5974534368762045e-05,
+      "loss": 0.9459,
       "step": 11182
     },
     {
-      "epoch": 0.3071324599708879,
+      "epoch": 0.3173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.6235494101855493e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.5973797331513674e-05,
+      "loss": 1.0429,
       "step": 11183
     },
     {
-      "epoch": 0.30715992419873117,
+      "epoch": 0.317366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.62347986678521e-05,
-      "loss": 0.9192,
+      "learning_rate": 1.597306024380437e-05,
+      "loss": 0.8542,
       "step": 11184
     },
     {
-      "epoch": 0.3071873884265744,
+      "epoch": 0.3173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.623410318451657e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.5972323105640356e-05,
+      "loss": 0.8467,
       "step": 11185
     },
     {
-      "epoch": 0.3072148526544176,
+      "epoch": 0.3174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.62334076518544e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.5971585917027864e-05,
+      "loss": 0.9766,
       "step": 11186
     },
     {
-      "epoch": 0.30724231688226084,
+      "epoch": 0.3174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.6232712069871088e-05,
-      "loss": 1.1676,
+      "learning_rate": 1.5970848677973117e-05,
+      "loss": 0.9102,
       "step": 11187
     },
     {
-      "epoch": 0.3072697811101041,
+      "epoch": 0.31748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.623201643857215e-05,
-      "loss": 0.937,
+      "learning_rate": 1.5970111388482335e-05,
+      "loss": 0.9228,
       "step": 11188
     },
     {
-      "epoch": 0.30729724533794733,
+      "epoch": 0.3175085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.6231320757963082e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.5969374048561763e-05,
+      "loss": 0.9862,
       "step": 11189
     },
     {
-      "epoch": 0.30732470956579055,
+      "epoch": 0.31753688989784334,
       "grad_norm": 0.0,
-      "learning_rate": 1.6230625028049386e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.5968636658217614e-05,
+      "loss": 1.0428,
       "step": 11190
     },
     {
-      "epoch": 0.3073521737936338,
+      "epoch": 0.31756526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.6229929248836577e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.5967899217456126e-05,
+      "loss": 0.9112,
       "step": 11191
     },
     {
-      "epoch": 0.30737963802147705,
+      "epoch": 0.3175936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 1.622923342033015e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.5967161726283527e-05,
+      "loss": 0.9984,
       "step": 11192
     },
     {
-      "epoch": 0.3074071022493203,
+      "epoch": 0.31762202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.622853754253562e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.5966424184706043e-05,
+      "loss": 1.0514,
       "step": 11193
     },
     {
-      "epoch": 0.3074345664771635,
+      "epoch": 0.31765039727582295,
       "grad_norm": 0.0,
-      "learning_rate": 1.6227841615458486e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.5965686592729903e-05,
+      "loss": 0.9409,
       "step": 11194
     },
     {
-      "epoch": 0.3074620307050067,
+      "epoch": 0.3176787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.6227145639104257e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.5964948950361343e-05,
+      "loss": 0.9622,
       "step": 11195
     },
     {
-      "epoch": 0.30748949493284994,
+      "epoch": 0.3177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.622644961347844e-05,
-      "loss": 0.9874,
+      "learning_rate": 1.5964211257606587e-05,
+      "loss": 0.9713,
       "step": 11196
     },
     {
-      "epoch": 0.3075169591606932,
+      "epoch": 0.3177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.622575353858654e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.5963473514471874e-05,
+      "loss": 0.9467,
       "step": 11197
     },
     {
-      "epoch": 0.30754442338853644,
+      "epoch": 0.3177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.6225057414434065e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.5962735720963434e-05,
+      "loss": 0.963,
       "step": 11198
     },
     {
-      "epoch": 0.30757188761637966,
+      "epoch": 0.3177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.622436124102653e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.5961997877087495e-05,
+      "loss": 1.0161,
       "step": 11199
     },
     {
-      "epoch": 0.3075993518442229,
+      "epoch": 0.3178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.6223665018369434e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.5961259982850293e-05,
+      "loss": 0.9849,
       "step": 11200
     },
     {
-      "epoch": 0.30762681607206616,
+      "epoch": 0.3178490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.622296874646829e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.5960522038258057e-05,
+      "loss": 0.9231,
       "step": 11201
     },
     {
-      "epoch": 0.3076542802999094,
+      "epoch": 0.31787741203178205,
       "grad_norm": 0.0,
-      "learning_rate": 1.6222272425328608e-05,
-      "loss": 1.0973,
+      "learning_rate": 1.5959784043317027e-05,
+      "loss": 0.9852,
       "step": 11202
     },
     {
-      "epoch": 0.3076817445277526,
+      "epoch": 0.31790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.6221576054955896e-05,
-      "loss": 0.9529,
+      "learning_rate": 1.595904599803343e-05,
+      "loss": 0.9043,
       "step": 11203
     },
     {
-      "epoch": 0.3077092087555958,
+      "epoch": 0.3179341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 1.622087963535566e-05,
-      "loss": 0.9556,
+      "learning_rate": 1.5958307902413505e-05,
+      "loss": 1.0425,
       "step": 11204
     },
     {
-      "epoch": 0.3077366729834391,
+      "epoch": 0.31796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.622018316653342e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.5957569756463485e-05,
+      "loss": 1.0155,
       "step": 11205
     },
     {
-      "epoch": 0.3077641372112823,
+      "epoch": 0.31799091940976165,
       "grad_norm": 0.0,
-      "learning_rate": 1.6219486648494677e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.5956831560189608e-05,
+      "loss": 0.9029,
       "step": 11206
     },
     {
-      "epoch": 0.30779160143912554,
+      "epoch": 0.3180192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.621879008124495e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.59560933135981e-05,
+      "loss": 1.1194,
       "step": 11207
     },
     {
-      "epoch": 0.30781906566696876,
+      "epoch": 0.3180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.6218093464789747e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.5955355016695212e-05,
+      "loss": 0.9537,
       "step": 11208
     },
     {
-      "epoch": 0.307846529894812,
+      "epoch": 0.3180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.621739679913458e-05,
-      "loss": 0.9593,
+      "learning_rate": 1.595461666948717e-05,
+      "loss": 1.0564,
       "step": 11209
     },
     {
-      "epoch": 0.30787399412265526,
+      "epoch": 0.3181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.621670008428496e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.5953878271980212e-05,
+      "loss": 0.8149,
       "step": 11210
     },
     {
-      "epoch": 0.3079014583504985,
+      "epoch": 0.3181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.62160033202464e-05,
-      "loss": 1.0406,
+      "learning_rate": 1.5953139824180576e-05,
+      "loss": 0.9756,
       "step": 11211
     },
     {
-      "epoch": 0.3079289225783417,
+      "epoch": 0.318161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.6215306507024412e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.59524013260945e-05,
+      "loss": 0.9451,
       "step": 11212
     },
     {
-      "epoch": 0.3079563868061849,
+      "epoch": 0.3181895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.621460964462451e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.5951662777728225e-05,
+      "loss": 1.0269,
       "step": 11213
     },
     {
-      "epoch": 0.3079838510340282,
+      "epoch": 0.31821793416572075,
       "grad_norm": 0.0,
-      "learning_rate": 1.6213912733052216e-05,
-      "loss": 1.0414,
+      "learning_rate": 1.5950924179087985e-05,
+      "loss": 0.9466,
       "step": 11214
     },
     {
-      "epoch": 0.3080113152618714,
+      "epoch": 0.31824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.6213215772313034e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.5950185530180023e-05,
+      "loss": 0.994,
       "step": 11215
     },
     {
-      "epoch": 0.30803877948971464,
+      "epoch": 0.3182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.6212518762412482e-05,
-      "loss": 0.9446,
+      "learning_rate": 1.5949446831010576e-05,
+      "loss": 0.921,
       "step": 11216
     },
     {
-      "epoch": 0.30806624371755786,
+      "epoch": 0.31830306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.6211821703356075e-05,
-      "loss": 1.0047,
+      "learning_rate": 1.5948708081585885e-05,
+      "loss": 0.9004,
       "step": 11217
     },
     {
-      "epoch": 0.30809370794540114,
+      "epoch": 0.31833144154370036,
       "grad_norm": 0.0,
-      "learning_rate": 1.621112459514933e-05,
-      "loss": 0.953,
+      "learning_rate": 1.5947969281912188e-05,
+      "loss": 0.8983,
       "step": 11218
     },
     {
-      "epoch": 0.30812117217324436,
+      "epoch": 0.3183598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.621042743779776e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.5947230431995725e-05,
+      "loss": 0.9459,
       "step": 11219
     },
     {
-      "epoch": 0.3081486364010876,
+      "epoch": 0.31838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.6209730231306883e-05,
-      "loss": 0.8648,
+      "learning_rate": 1.5946491531842744e-05,
+      "loss": 0.9922,
       "step": 11220
     },
     {
-      "epoch": 0.3081761006289308,
+      "epoch": 0.318416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.6209032975682216e-05,
-      "loss": 1.0354,
+      "learning_rate": 1.5945752581459476e-05,
+      "loss": 0.9477,
       "step": 11221
     },
     {
-      "epoch": 0.308203564856774,
+      "epoch": 0.3184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.6208335670929273e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.5945013580852172e-05,
+      "loss": 0.9196,
       "step": 11222
     },
     {
-      "epoch": 0.3082310290846173,
+      "epoch": 0.3184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.6207638317053574e-05,
-      "loss": 0.8642,
+      "learning_rate": 1.594427453002707e-05,
+      "loss": 0.9487,
       "step": 11223
     },
     {
-      "epoch": 0.3082584933124605,
+      "epoch": 0.3185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.6206940914060635e-05,
-      "loss": 0.9645,
+      "learning_rate": 1.5943535428990417e-05,
+      "loss": 0.8608,
       "step": 11224
     },
     {
-      "epoch": 0.30828595754030375,
+      "epoch": 0.3185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.6206243461955976e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.5942796277748448e-05,
+      "loss": 0.9597,
       "step": 11225
     },
     {
-      "epoch": 0.30831342176814697,
+      "epoch": 0.31855845629965945,
       "grad_norm": 0.0,
-      "learning_rate": 1.6205545960745114e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.5942057076307413e-05,
+      "loss": 0.8337,
       "step": 11226
     },
     {
-      "epoch": 0.30834088599599024,
+      "epoch": 0.31858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.6204848410433568e-05,
-      "loss": 0.943,
+      "learning_rate": 1.5941317824673557e-05,
+      "loss": 0.9199,
       "step": 11227
     },
     {
-      "epoch": 0.30836835022383347,
+      "epoch": 0.3186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.620415081102686e-05,
-      "loss": 1.0166,
+      "learning_rate": 1.594057852285312e-05,
+      "loss": 1.0803,
       "step": 11228
     },
     {
-      "epoch": 0.3083958144516767,
+      "epoch": 0.31864358683314414,
       "grad_norm": 0.0,
-      "learning_rate": 1.6203453162530504e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.5939839170852352e-05,
+      "loss": 1.0274,
       "step": 11229
     },
     {
-      "epoch": 0.3084232786795199,
+      "epoch": 0.31867196367763906,
       "grad_norm": 0.0,
-      "learning_rate": 1.6202755464950026e-05,
-      "loss": 0.9345,
+      "learning_rate": 1.5939099768677495e-05,
+      "loss": 0.9519,
       "step": 11230
     },
     {
-      "epoch": 0.3084507429073632,
+      "epoch": 0.3187003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 1.6202057718290944e-05,
-      "loss": 0.8818,
+      "learning_rate": 1.593836031633479e-05,
+      "loss": 0.9004,
       "step": 11231
     },
     {
-      "epoch": 0.3084782071352064,
+      "epoch": 0.31872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.620135992255878e-05,
-      "loss": 0.9802,
+      "learning_rate": 1.5937620813830495e-05,
+      "loss": 0.9054,
       "step": 11232
     },
     {
-      "epoch": 0.3085056713630496,
+      "epoch": 0.31875709421112375,
       "grad_norm": 0.0,
-      "learning_rate": 1.6200662077759056e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.593688126117085e-05,
+      "loss": 1.0251,
       "step": 11233
     },
     {
-      "epoch": 0.30853313559089285,
+      "epoch": 0.3187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.619996418389729e-05,
-      "loss": 1.1013,
+      "learning_rate": 1.5936141658362098e-05,
+      "loss": 1.1187,
       "step": 11234
     },
     {
-      "epoch": 0.30856059981873607,
+      "epoch": 0.3188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.6199266240979003e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.5935402005410493e-05,
+      "loss": 1.1443,
       "step": 11235
     },
     {
-      "epoch": 0.30858806404657935,
+      "epoch": 0.3188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.619856824900972e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.593466230232228e-05,
+      "loss": 0.9493,
       "step": 11236
     },
     {
-      "epoch": 0.30861552827442257,
+      "epoch": 0.3188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.6197870207994965e-05,
-      "loss": 0.978,
+      "learning_rate": 1.5933922549103708e-05,
+      "loss": 0.9777,
       "step": 11237
     },
     {
-      "epoch": 0.3086429925022658,
+      "epoch": 0.31889897843359816,
       "grad_norm": 0.0,
-      "learning_rate": 1.619717211794026e-05,
-      "loss": 1.0501,
+      "learning_rate": 1.5933182745761023e-05,
+      "loss": 0.8681,
       "step": 11238
     },
     {
-      "epoch": 0.308670456730109,
+      "epoch": 0.31892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.619647397885113e-05,
-      "loss": 1.069,
+      "learning_rate": 1.593244289230048e-05,
+      "loss": 0.8682,
       "step": 11239
     },
     {
-      "epoch": 0.3086979209579523,
+      "epoch": 0.318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.6195775790733096e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.5931702988728323e-05,
+      "loss": 0.8554,
       "step": 11240
     },
     {
-      "epoch": 0.3087253851857955,
+      "epoch": 0.31898410896708285,
       "grad_norm": 0.0,
-      "learning_rate": 1.619507755359169e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.5930963035050805e-05,
+      "loss": 0.9485,
       "step": 11241
     },
     {
-      "epoch": 0.30875284941363873,
+      "epoch": 0.31901248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.6194379267432423e-05,
-      "loss": 0.8412,
+      "learning_rate": 1.5930223031274176e-05,
+      "loss": 0.879,
       "step": 11242
     },
     {
-      "epoch": 0.30878031364148195,
+      "epoch": 0.3190408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.619368093226083e-05,
-      "loss": 0.9971,
+      "learning_rate": 1.5929482977404685e-05,
+      "loss": 1.0547,
       "step": 11243
     },
     {
-      "epoch": 0.30880777786932523,
+      "epoch": 0.31906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.6192982548082433e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.5928742873448585e-05,
+      "loss": 0.9805,
       "step": 11244
     },
     {
-      "epoch": 0.30883524209716845,
+      "epoch": 0.31909761634506245,
       "grad_norm": 0.0,
-      "learning_rate": 1.6192284114902764e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.592800271941213e-05,
+      "loss": 0.9254,
       "step": 11245
     },
     {
-      "epoch": 0.30886270632501167,
+      "epoch": 0.3191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.619158563272734e-05,
-      "loss": 1.0143,
+      "learning_rate": 1.5927262515301565e-05,
+      "loss": 1.0441,
       "step": 11246
     },
     {
-      "epoch": 0.3088901705528549,
+      "epoch": 0.3191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.6190887101561692e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.5926522261123155e-05,
+      "loss": 0.9436,
       "step": 11247
     },
     {
-      "epoch": 0.3089176347806981,
+      "epoch": 0.3191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.6190188521411348e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.5925781956883136e-05,
+      "loss": 0.9808,
       "step": 11248
     },
     {
-      "epoch": 0.3089450990085414,
+      "epoch": 0.319211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.6189489892281835e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.5925041602587773e-05,
+      "loss": 0.9837,
       "step": 11249
     },
     {
-      "epoch": 0.3089725632363846,
+      "epoch": 0.3192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.618879121417868e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.592430119824332e-05,
+      "loss": 0.9709,
       "step": 11250
     },
     {
-      "epoch": 0.30900002746422783,
+      "epoch": 0.3192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.618809248710741e-05,
-      "loss": 0.9098,
+      "learning_rate": 1.592356074385603e-05,
+      "loss": 0.946,
       "step": 11251
     },
     {
-      "epoch": 0.30902749169207105,
+      "epoch": 0.3192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.6187393711073562e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.592282023943215e-05,
+      "loss": 0.9938,
       "step": 11252
     },
     {
-      "epoch": 0.30905495591991433,
+      "epoch": 0.31932463110102155,
       "grad_norm": 0.0,
-      "learning_rate": 1.6186694886082652e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.5922079684977944e-05,
+      "loss": 0.8816,
       "step": 11253
     },
     {
-      "epoch": 0.30908242014775755,
+      "epoch": 0.31935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.6185996012140215e-05,
-      "loss": 0.8684,
+      "learning_rate": 1.5921339080499666e-05,
+      "loss": 0.8569,
       "step": 11254
     },
     {
-      "epoch": 0.3091098843756008,
+      "epoch": 0.3193813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.6185297089251785e-05,
-      "loss": 1.0619,
+      "learning_rate": 1.592059842600357e-05,
+      "loss": 0.9272,
       "step": 11255
     },
     {
-      "epoch": 0.309137348603444,
+      "epoch": 0.31940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.618459811742289e-05,
-      "loss": 1.0052,
+      "learning_rate": 1.5919857721495914e-05,
+      "loss": 1.0296,
       "step": 11256
     },
     {
-      "epoch": 0.3091648128312873,
+      "epoch": 0.31943813847900115,
       "grad_norm": 0.0,
-      "learning_rate": 1.6183899096659053e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.5919116966982952e-05,
+      "loss": 0.9807,
       "step": 11257
     },
     {
-      "epoch": 0.3091922770591305,
+      "epoch": 0.319466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.6183200026965817e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.5918376162470942e-05,
+      "loss": 1.0633,
       "step": 11258
     },
     {
-      "epoch": 0.3092197412869737,
+      "epoch": 0.31949489216799093,
       "grad_norm": 0.0,
-      "learning_rate": 1.61825009083487e-05,
-      "loss": 0.964,
+      "learning_rate": 1.5917635307966144e-05,
+      "loss": 0.966,
       "step": 11259
     },
     {
-      "epoch": 0.30924720551481694,
+      "epoch": 0.3195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.6181801740813245e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.5916894403474814e-05,
+      "loss": 0.9666,
       "step": 11260
     },
     {
-      "epoch": 0.30927466974266016,
+      "epoch": 0.3195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.618110252436498e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.591615344900321e-05,
+      "loss": 1.0521,
       "step": 11261
     },
     {
-      "epoch": 0.30930213397050343,
+      "epoch": 0.3195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.6180403259009438e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.5915412444557595e-05,
+      "loss": 0.8845,
       "step": 11262
     },
     {
-      "epoch": 0.30932959819834666,
+      "epoch": 0.3196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.617970394475215e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.5914671390144222e-05,
+      "loss": 0.9802,
       "step": 11263
     },
     {
-      "epoch": 0.3093570624261899,
+      "epoch": 0.3196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.6179004581598652e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.5913930285769356e-05,
+      "loss": 0.9714,
       "step": 11264
     },
     {
-      "epoch": 0.3093845266540331,
+      "epoch": 0.31966515323496025,
       "grad_norm": 0.0,
-      "learning_rate": 1.6178305169554472e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.5913189131439254e-05,
+      "loss": 0.9247,
       "step": 11265
     },
     {
-      "epoch": 0.3094119908818764,
+      "epoch": 0.31969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.6177605708625154e-05,
-      "loss": 0.9818,
+      "learning_rate": 1.5912447927160176e-05,
+      "loss": 1.021,
       "step": 11266
     },
     {
-      "epoch": 0.3094394551097196,
+      "epoch": 0.3197219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.6176906198816225e-05,
-      "loss": 1.014,
+      "learning_rate": 1.5911706672938388e-05,
+      "loss": 0.8979,
       "step": 11267
     },
     {
-      "epoch": 0.3094669193375628,
+      "epoch": 0.31975028376844494,
       "grad_norm": 0.0,
-      "learning_rate": 1.617620664013322e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.5910965368780146e-05,
+      "loss": 0.909,
       "step": 11268
     },
     {
-      "epoch": 0.30949438356540604,
+      "epoch": 0.31977866061293986,
       "grad_norm": 0.0,
-      "learning_rate": 1.6175507032581675e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.5910224014691717e-05,
+      "loss": 0.9894,
       "step": 11269
     },
     {
-      "epoch": 0.3095218477932493,
+      "epoch": 0.3198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.617480737616713e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.5909482610679355e-05,
+      "loss": 0.9915,
       "step": 11270
     },
     {
-      "epoch": 0.30954931202109254,
+      "epoch": 0.31983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.6174107670895116e-05,
-      "loss": 1.0365,
+      "learning_rate": 1.590874115674933e-05,
+      "loss": 0.913,
       "step": 11271
     },
     {
-      "epoch": 0.30957677624893576,
+      "epoch": 0.3198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.6173407916771168e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.59079996529079e-05,
+      "loss": 0.9659,
       "step": 11272
     },
     {
-      "epoch": 0.309604240476779,
+      "epoch": 0.3198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.617270811380083e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.5907258099161335e-05,
+      "loss": 0.9607,
       "step": 11273
     },
     {
-      "epoch": 0.30963170470462226,
+      "epoch": 0.3199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6172008261989627e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.5906516495515895e-05,
+      "loss": 0.9723,
       "step": 11274
     },
     {
-      "epoch": 0.3096591689324655,
+      "epoch": 0.3199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.6171308361343107e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.5905774841977843e-05,
+      "loss": 0.9962,
       "step": 11275
     },
     {
-      "epoch": 0.3096866331603087,
+      "epoch": 0.3199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.617060841186681e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.590503313855345e-05,
+      "loss": 0.9471,
       "step": 11276
     },
     {
-      "epoch": 0.3097140973881519,
+      "epoch": 0.32000567536889896,
       "grad_norm": 0.0,
-      "learning_rate": 1.6169908413566263e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.5904291385248973e-05,
+      "loss": 0.929,
       "step": 11277
     },
     {
-      "epoch": 0.30974156161599514,
+      "epoch": 0.32003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.6169208366447008e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.5903549582070682e-05,
+      "loss": 0.9803,
       "step": 11278
     },
     {
-      "epoch": 0.3097690258438384,
+      "epoch": 0.3200624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.6168508270514593e-05,
-      "loss": 1.0939,
+      "learning_rate": 1.590280772902484e-05,
+      "loss": 0.9163,
       "step": 11279
     },
     {
-      "epoch": 0.30979649007168164,
+      "epoch": 0.32009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.6167808125774547e-05,
-      "loss": 0.8304,
+      "learning_rate": 1.5902065826117717e-05,
+      "loss": 0.9757,
       "step": 11280
     },
     {
-      "epoch": 0.30982395429952486,
+      "epoch": 0.32011918274687856,
       "grad_norm": 0.0,
-      "learning_rate": 1.616710793223242e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.590132387335558e-05,
+      "loss": 0.8527,
       "step": 11281
     },
     {
-      "epoch": 0.3098514185273681,
+      "epoch": 0.3201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.616640768989374e-05,
-      "loss": 0.9036,
+      "learning_rate": 1.5900581870744692e-05,
+      "loss": 0.9661,
       "step": 11282
     },
     {
-      "epoch": 0.30987888275521136,
+      "epoch": 0.32017593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.6165707398764056e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.5899839818291323e-05,
+      "loss": 0.907,
       "step": 11283
     },
     {
-      "epoch": 0.3099063469830546,
+      "epoch": 0.32020431328036325,
       "grad_norm": 0.0,
-      "learning_rate": 1.6165007058848906e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.589909771600175e-05,
+      "loss": 1.0776,
       "step": 11284
     },
     {
-      "epoch": 0.3099338112108978,
+      "epoch": 0.3202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.616430667015383e-05,
-      "loss": 0.8968,
+      "learning_rate": 1.5898355563882227e-05,
+      "loss": 0.9417,
       "step": 11285
     },
     {
-      "epoch": 0.309961275438741,
+      "epoch": 0.320261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.6163606232684378e-05,
-      "loss": 0.8725,
+      "learning_rate": 1.589761336193903e-05,
+      "loss": 0.9811,
       "step": 11286
     },
     {
-      "epoch": 0.3099887396665843,
+      "epoch": 0.3202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.6162905746446078e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.5896871110178428e-05,
+      "loss": 0.8925,
       "step": 11287
     },
     {
-      "epoch": 0.3100162038944275,
+      "epoch": 0.3203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.616220521144449e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.589612880860669e-05,
+      "loss": 0.9472,
       "step": 11288
     },
     {
-      "epoch": 0.31004366812227074,
+      "epoch": 0.32034619750283766,
       "grad_norm": 0.0,
-      "learning_rate": 1.6161504627685138e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.5895386457230088e-05,
+      "loss": 1.0178,
       "step": 11289
     },
     {
-      "epoch": 0.31007113235011396,
+      "epoch": 0.3203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.6160803995173576e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.589464405605489e-05,
+      "loss": 0.9614,
       "step": 11290
     },
     {
-      "epoch": 0.3100985965779572,
+      "epoch": 0.3204029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.6160103313915353e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.589390160508737e-05,
+      "loss": 0.9409,
       "step": 11291
     },
     {
-      "epoch": 0.31012606080580046,
+      "epoch": 0.32043132803632235,
       "grad_norm": 0.0,
-      "learning_rate": 1.6159402583915998e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.58931591043338e-05,
+      "loss": 0.9321,
       "step": 11292
     },
     {
-      "epoch": 0.3101535250336437,
+      "epoch": 0.32045970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.615870180518107e-05,
-      "loss": 0.9964,
+      "learning_rate": 1.589241655380045e-05,
+      "loss": 0.8492,
       "step": 11293
     },
     {
-      "epoch": 0.3101809892614869,
+      "epoch": 0.3204880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.6158000977716108e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.589167395349359e-05,
+      "loss": 0.9595,
       "step": 11294
     },
     {
-      "epoch": 0.3102084534893301,
+      "epoch": 0.32051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.6157300101526655e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.5890931303419496e-05,
+      "loss": 0.9518,
       "step": 11295
     },
     {
-      "epoch": 0.3102359177171734,
+      "epoch": 0.32054483541430195,
       "grad_norm": 0.0,
-      "learning_rate": 1.6156599176618256e-05,
-      "loss": 1.0601,
+      "learning_rate": 1.589018860358444e-05,
+      "loss": 0.9327,
       "step": 11296
     },
     {
-      "epoch": 0.3102633819450166,
+      "epoch": 0.3205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.615589820299646e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.5889445853994695e-05,
+      "loss": 0.9812,
       "step": 11297
     },
     {
-      "epoch": 0.31029084617285985,
+      "epoch": 0.32060158910329173,
       "grad_norm": 0.0,
-      "learning_rate": 1.6155197180666818e-05,
-      "loss": 1.092,
+      "learning_rate": 1.5888703054656536e-05,
+      "loss": 0.9509,
       "step": 11298
     },
     {
-      "epoch": 0.31031831040070307,
+      "epoch": 0.3206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.6154496109634864e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.5887960205576243e-05,
+      "loss": 0.9995,
       "step": 11299
     },
     {
-      "epoch": 0.31034577462854634,
+      "epoch": 0.3206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.615379498990616e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.5887217306760082e-05,
+      "loss": 0.8574,
       "step": 11300
     },
     {
-      "epoch": 0.31037323885638957,
+      "epoch": 0.32068671963677636,
       "grad_norm": 0.0,
-      "learning_rate": 1.6153093821486244e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.5886474358214333e-05,
+      "loss": 0.9227,
       "step": 11301
     },
     {
-      "epoch": 0.3104007030842328,
+      "epoch": 0.3207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.6152392604380664e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.5885731359945266e-05,
+      "loss": 0.8575,
       "step": 11302
     },
     {
-      "epoch": 0.310428167312076,
+      "epoch": 0.3207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.615169133859497e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.5884988311959167e-05,
+      "loss": 0.9563,
       "step": 11303
     },
     {
-      "epoch": 0.31045563153991923,
+      "epoch": 0.32077185017026105,
       "grad_norm": 0.0,
-      "learning_rate": 1.6150990024134714e-05,
-      "loss": 0.9581,
+      "learning_rate": 1.5884245214262303e-05,
+      "loss": 0.9628,
       "step": 11304
     },
     {
-      "epoch": 0.3104830957677625,
+      "epoch": 0.32080022701475597,
       "grad_norm": 0.0,
-      "learning_rate": 1.615028866100544e-05,
-      "loss": 1.0422,
+      "learning_rate": 1.5883502066860956e-05,
+      "loss": 0.9051,
       "step": 11305
     },
     {
-      "epoch": 0.31051055999560573,
+      "epoch": 0.3208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.6149587249212705e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.5882758869761404e-05,
+      "loss": 1.0173,
       "step": 11306
     },
     {
-      "epoch": 0.31053802422344895,
+      "epoch": 0.32085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.6148885788762047e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.5882015622969923e-05,
+      "loss": 0.9649,
       "step": 11307
     },
     {
-      "epoch": 0.31056548845129217,
+      "epoch": 0.32088535754824066,
       "grad_norm": 0.0,
-      "learning_rate": 1.6148184279659026e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.588127232649279e-05,
+      "loss": 1.032,
       "step": 11308
     },
     {
-      "epoch": 0.31059295267913545,
+      "epoch": 0.3209137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.6147482721909188e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.588052898033629e-05,
+      "loss": 0.9544,
       "step": 11309
     },
     {
-      "epoch": 0.31062041690697867,
+      "epoch": 0.32094211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.6146781115518086e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.5879785584506692e-05,
+      "loss": 0.8775,
       "step": 11310
     },
     {
-      "epoch": 0.3106478811348219,
+      "epoch": 0.3209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.614607946049127e-05,
-      "loss": 0.8451,
+      "learning_rate": 1.5879042139010283e-05,
+      "loss": 0.9226,
       "step": 11311
     },
     {
-      "epoch": 0.3106753453626651,
+      "epoch": 0.3209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.6145377756834296e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.587829864385334e-05,
+      "loss": 0.9475,
       "step": 11312
     },
     {
-      "epoch": 0.3107028095905084,
+      "epoch": 0.3210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.614467600455271e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.5877555099042145e-05,
+      "loss": 0.8845,
       "step": 11313
     },
     {
-      "epoch": 0.3107302738183516,
+      "epoch": 0.32105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.6143974203652066e-05,
-      "loss": 0.9099,
+      "learning_rate": 1.587681150458298e-05,
+      "loss": 1.0576,
       "step": 11314
     },
     {
-      "epoch": 0.31075773804619483,
+      "epoch": 0.3210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.614327235413792e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.587606786048212e-05,
+      "loss": 0.8374,
       "step": 11315
     },
     {
-      "epoch": 0.31078520227403805,
+      "epoch": 0.32111237230419976,
       "grad_norm": 0.0,
-      "learning_rate": 1.614257045601582e-05,
-      "loss": 0.9306,
+      "learning_rate": 1.5875324166745855e-05,
+      "loss": 0.7706,
       "step": 11316
     },
     {
-      "epoch": 0.3108126665018813,
+      "epoch": 0.32114074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.614186850929133e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.5874580423380458e-05,
+      "loss": 0.8789,
       "step": 11317
     },
     {
-      "epoch": 0.31084013072972455,
+      "epoch": 0.32116912599318953,
       "grad_norm": 0.0,
-      "learning_rate": 1.614116651396999e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.5873836630392218e-05,
+      "loss": 1.0382,
       "step": 11318
     },
     {
-      "epoch": 0.31086759495756777,
+      "epoch": 0.32119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.6140464470057367e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.5873092787787418e-05,
+      "loss": 0.9264,
       "step": 11319
     },
     {
-      "epoch": 0.310895059185411,
+      "epoch": 0.32122587968217936,
       "grad_norm": 0.0,
-      "learning_rate": 1.613976237755901e-05,
-      "loss": 0.965,
+      "learning_rate": 1.5872348895572335e-05,
+      "loss": 0.9338,
       "step": 11320
     },
     {
-      "epoch": 0.3109225234132542,
+      "epoch": 0.3212542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.6139060236480472e-05,
-      "loss": 1.074,
+      "learning_rate": 1.587160495375326e-05,
+      "loss": 1.0344,
       "step": 11321
     },
     {
-      "epoch": 0.3109499876410975,
+      "epoch": 0.32128263337116914,
       "grad_norm": 0.0,
-      "learning_rate": 1.613835804682731e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.587086096233647e-05,
+      "loss": 0.9526,
       "step": 11322
     },
     {
-      "epoch": 0.3109774518689407,
+      "epoch": 0.321311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.613765580860509e-05,
-      "loss": 1.0231,
+      "learning_rate": 1.5870116921328263e-05,
+      "loss": 1.0087,
       "step": 11323
     },
     {
-      "epoch": 0.31100491609678393,
+      "epoch": 0.3213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.6136953521819353e-05,
-      "loss": 1.1154,
+      "learning_rate": 1.5869372830734906e-05,
+      "loss": 0.9839,
       "step": 11324
     },
     {
-      "epoch": 0.31103238032462716,
+      "epoch": 0.3213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.613625118647566e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.5868628690562695e-05,
+      "loss": 0.8506,
       "step": 11325
     },
     {
-      "epoch": 0.31105984455247043,
+      "epoch": 0.3213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6135548802579574e-05,
-      "loss": 0.8434,
+      "learning_rate": 1.5867884500817914e-05,
+      "loss": 0.985,
       "step": 11326
     },
     {
-      "epoch": 0.31108730878031365,
+      "epoch": 0.3214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.6134846370136653e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.5867140261506845e-05,
+      "loss": 0.8806,
       "step": 11327
     },
     {
-      "epoch": 0.3111147730081569,
+      "epoch": 0.32145289443813846,
       "grad_norm": 0.0,
-      "learning_rate": 1.6134143889152446e-05,
-      "loss": 1.0041,
+      "learning_rate": 1.5866395972635788e-05,
+      "loss": 1.0475,
       "step": 11328
     },
     {
-      "epoch": 0.3111422372360001,
+      "epoch": 0.3214812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.6133441359632516e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.586565163421101e-05,
+      "loss": 1.0355,
       "step": 11329
     },
     {
-      "epoch": 0.3111697014638433,
+      "epoch": 0.3215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.6132738781582423e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.5864907246238814e-05,
+      "loss": 0.8801,
       "step": 11330
     },
     {
-      "epoch": 0.3111971656916866,
+      "epoch": 0.32153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.6132036155007726e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.586416280872548e-05,
+      "loss": 0.9749,
       "step": 11331
     },
     {
-      "epoch": 0.3112246299195298,
+      "epoch": 0.32156640181611806,
       "grad_norm": 0.0,
-      "learning_rate": 1.6131333479913983e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.5863418321677303e-05,
+      "loss": 0.9711,
       "step": 11332
     },
     {
-      "epoch": 0.31125209414737304,
+      "epoch": 0.3215947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.6130630756306757e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.5862673785100567e-05,
+      "loss": 0.9888,
       "step": 11333
     },
     {
-      "epoch": 0.31127955837521626,
+      "epoch": 0.32162315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 1.6129927984191603e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.5861929199001562e-05,
+      "loss": 1.0872,
       "step": 11334
     },
     {
-      "epoch": 0.31130702260305954,
+      "epoch": 0.3216515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.6129225163574086e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.5861184563386576e-05,
+      "loss": 0.9345,
       "step": 11335
     },
     {
-      "epoch": 0.31133448683090276,
+      "epoch": 0.3216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.6128522294459762e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.5860439878261905e-05,
+      "loss": 0.9358,
       "step": 11336
     },
     {
-      "epoch": 0.311361951058746,
+      "epoch": 0.32170828603859253,
       "grad_norm": 0.0,
-      "learning_rate": 1.6127819376854198e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.585969514363383e-05,
+      "loss": 0.9968,
       "step": 11337
     },
     {
-      "epoch": 0.3113894152865892,
+      "epoch": 0.3217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.6127116410762957e-05,
-      "loss": 1.012,
+      "learning_rate": 1.585895035950865e-05,
+      "loss": 1.0179,
       "step": 11338
     },
     {
-      "epoch": 0.3114168795144325,
+      "epoch": 0.3217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.612641339619159e-05,
-      "loss": 1.036,
+      "learning_rate": 1.5858205525892647e-05,
+      "loss": 0.9794,
       "step": 11339
     },
     {
-      "epoch": 0.3114443437422757,
+      "epoch": 0.32179341657207716,
       "grad_norm": 0.0,
-      "learning_rate": 1.6125710333145672e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.5857460642792124e-05,
+      "loss": 0.914,
       "step": 11340
     },
     {
-      "epoch": 0.3114718079701189,
+      "epoch": 0.3218217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.612500722163076e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.5856715710213367e-05,
+      "loss": 1.0312,
       "step": 11341
     },
     {
-      "epoch": 0.31149927219796214,
+      "epoch": 0.321850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.612430406165242e-05,
-      "loss": 0.994,
+      "learning_rate": 1.5855970728162668e-05,
+      "loss": 0.9347,
       "step": 11342
     },
     {
-      "epoch": 0.31152673642580536,
+      "epoch": 0.32187854710556185,
       "grad_norm": 0.0,
-      "learning_rate": 1.612360085321621e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.5855225696646322e-05,
+      "loss": 0.9457,
       "step": 11343
     },
     {
-      "epoch": 0.31155420065364864,
+      "epoch": 0.32190692395005677,
       "grad_norm": 0.0,
-      "learning_rate": 1.61228975963277e-05,
-      "loss": 1.0486,
+      "learning_rate": 1.5854480615670618e-05,
+      "loss": 0.8392,
       "step": 11344
     },
     {
-      "epoch": 0.31158166488149186,
+      "epoch": 0.3219353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.6122194290992453e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.5853735485241858e-05,
+      "loss": 0.8729,
       "step": 11345
     },
     {
-      "epoch": 0.3116091291093351,
+      "epoch": 0.32196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.612149093721603e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.5852990305366326e-05,
+      "loss": 0.9809,
       "step": 11346
     },
     {
-      "epoch": 0.3116365933371783,
+      "epoch": 0.32199205448354146,
       "grad_norm": 0.0,
-      "learning_rate": 1.6120787535004005e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.5852245076050325e-05,
+      "loss": 0.9567,
       "step": 11347
     },
     {
-      "epoch": 0.3116640575650216,
+      "epoch": 0.3220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.6120084084361935e-05,
-      "loss": 0.9197,
+      "learning_rate": 1.585149979730015e-05,
+      "loss": 0.92,
       "step": 11348
     },
     {
-      "epoch": 0.3116915217928648,
+      "epoch": 0.32204880817253123,
       "grad_norm": 0.0,
-      "learning_rate": 1.611938058529539e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.5850754469122084e-05,
+      "loss": 0.8961,
       "step": 11349
     },
     {
-      "epoch": 0.311718986020708,
+      "epoch": 0.3220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.6118677037809934e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.585000909152244e-05,
+      "loss": 0.914,
       "step": 11350
     },
     {
-      "epoch": 0.31174645024855124,
+      "epoch": 0.322105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.6117973441911136e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.5849263664507504e-05,
+      "loss": 0.8499,
       "step": 11351
     },
     {
-      "epoch": 0.3117739144763945,
+      "epoch": 0.32213393870601587,
       "grad_norm": 0.0,
-      "learning_rate": 1.611726979760456e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.5848518188083576e-05,
+      "loss": 0.861,
       "step": 11352
     },
     {
-      "epoch": 0.31180137870423774,
+      "epoch": 0.3221623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.6116566104895782e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.584777266225695e-05,
+      "loss": 0.9117,
       "step": 11353
     },
     {
-      "epoch": 0.31182884293208096,
+      "epoch": 0.3221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.6115862363790356e-05,
-      "loss": 1.0406,
+      "learning_rate": 1.5847027087033926e-05,
+      "loss": 0.9922,
       "step": 11354
     },
     {
-      "epoch": 0.3118563071599242,
+      "epoch": 0.32221906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.611515857429386e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.5846281462420804e-05,
+      "loss": 1.0124,
       "step": 11355
     },
     {
-      "epoch": 0.3118837713877674,
+      "epoch": 0.32224744608399547,
       "grad_norm": 0.0,
-      "learning_rate": 1.6114454736411868e-05,
-      "loss": 0.9529,
+      "learning_rate": 1.5845535788423876e-05,
+      "loss": 0.8573,
       "step": 11356
     },
     {
-      "epoch": 0.3119112356156107,
+      "epoch": 0.32227582292849033,
       "grad_norm": 0.0,
-      "learning_rate": 1.611375085014993e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.5844790065049444e-05,
+      "loss": 1.0802,
       "step": 11357
     },
     {
-      "epoch": 0.3119386998434539,
+      "epoch": 0.32230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.6113046915513635e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.584404429230381e-05,
+      "loss": 1.0594,
       "step": 11358
     },
     {
-      "epoch": 0.3119661640712971,
+      "epoch": 0.32233257661748016,
       "grad_norm": 0.0,
-      "learning_rate": 1.6112342932508543e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.584329847019327e-05,
+      "loss": 0.9426,
       "step": 11359
     },
     {
-      "epoch": 0.31199362829914035,
+      "epoch": 0.322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.6111638901140227e-05,
-      "loss": 1.0375,
+      "learning_rate": 1.5842552598724122e-05,
+      "loss": 1.02,
       "step": 11360
     },
     {
-      "epoch": 0.3120210925269836,
+      "epoch": 0.32238933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 1.611093482141425e-05,
-      "loss": 1.0132,
+      "learning_rate": 1.5841806677902675e-05,
+      "loss": 0.8604,
       "step": 11361
     },
     {
-      "epoch": 0.31204855675482684,
+      "epoch": 0.3224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.61102306933362e-05,
-      "loss": 0.8691,
+      "learning_rate": 1.5841060707735222e-05,
+      "loss": 0.9111,
       "step": 11362
     },
     {
-      "epoch": 0.31207602098267007,
+      "epoch": 0.3224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.610952651691163e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.584031468822807e-05,
+      "loss": 0.9337,
       "step": 11363
     },
     {
-      "epoch": 0.3121034852105133,
+      "epoch": 0.3224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.610882229214612e-05,
-      "loss": 0.9207,
+      "learning_rate": 1.5839568619387512e-05,
+      "loss": 1.0135,
       "step": 11364
     },
     {
-      "epoch": 0.31213094943835656,
+      "epoch": 0.3225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.6108118019045246e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.583882250121986e-05,
+      "loss": 0.9004,
       "step": 11365
     },
     {
-      "epoch": 0.3121584136661998,
+      "epoch": 0.3225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.6107413697614573e-05,
-      "loss": 1.0182,
+      "learning_rate": 1.5838076333731406e-05,
+      "loss": 0.9231,
       "step": 11366
     },
     {
-      "epoch": 0.312185877894043,
+      "epoch": 0.32255959137343926,
       "grad_norm": 0.0,
-      "learning_rate": 1.6106709327859678e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.5837330116928464e-05,
+      "loss": 1.0934,
       "step": 11367
     },
     {
-      "epoch": 0.31221334212188623,
+      "epoch": 0.3225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.6106004909786134e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.583658385081733e-05,
+      "loss": 0.8775,
       "step": 11368
     },
     {
-      "epoch": 0.3122408063497295,
+      "epoch": 0.32261634506242903,
       "grad_norm": 0.0,
-      "learning_rate": 1.610530044339951e-05,
-      "loss": 1.0108,
+      "learning_rate": 1.583583753540431e-05,
+      "loss": 0.9351,
       "step": 11369
     },
     {
-      "epoch": 0.3122682705775727,
+      "epoch": 0.32264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.610459592870539e-05,
-      "loss": 1.008,
+      "learning_rate": 1.5835091170695708e-05,
+      "loss": 1.014,
       "step": 11370
     },
     {
-      "epoch": 0.31229573480541595,
+      "epoch": 0.32267309875141886,
       "grad_norm": 0.0,
-      "learning_rate": 1.610389136570934e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.583434475669783e-05,
+      "loss": 1.0194,
       "step": 11371
     },
     {
-      "epoch": 0.31232319903325917,
+      "epoch": 0.3227014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.610318675441694e-05,
-      "loss": 0.9113,
+      "learning_rate": 1.5833598293416978e-05,
+      "loss": 1.0771,
       "step": 11372
     },
     {
-      "epoch": 0.3123506632611024,
+      "epoch": 0.32272985244040864,
       "grad_norm": 0.0,
-      "learning_rate": 1.6102482094833756e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.583285178085946e-05,
+      "loss": 0.88,
       "step": 11373
     },
     {
-      "epoch": 0.31237812748894567,
+      "epoch": 0.3227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.610177738696537e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.583210521903158e-05,
+      "loss": 0.9224,
       "step": 11374
     },
     {
-      "epoch": 0.3124055917167889,
+      "epoch": 0.3227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.6101072630817363e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.5831358607939643e-05,
+      "loss": 1.0235,
       "step": 11375
     },
     {
-      "epoch": 0.3124330559446321,
+      "epoch": 0.3228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.6100367826395304e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.583061194758996e-05,
+      "loss": 0.8339,
       "step": 11376
     },
     {
-      "epoch": 0.31246052017247533,
+      "epoch": 0.3228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.6099662973704767e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.5829865237988838e-05,
+      "loss": 0.9426,
       "step": 11377
     },
     {
-      "epoch": 0.3124879844003186,
+      "epoch": 0.3228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.609895807275134e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.582911847914258e-05,
+      "loss": 0.9888,
       "step": 11378
     },
     {
-      "epoch": 0.31251544862816183,
+      "epoch": 0.32290011350737796,
       "grad_norm": 0.0,
-      "learning_rate": 1.609825312354059e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.5828371671057495e-05,
+      "loss": 0.9834,
       "step": 11379
     },
     {
-      "epoch": 0.31254291285600505,
+      "epoch": 0.3229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.60975481260781e-05,
-      "loss": 0.8924,
+      "learning_rate": 1.5827624813739893e-05,
+      "loss": 0.9538,
       "step": 11380
     },
     {
-      "epoch": 0.31257037708384827,
+      "epoch": 0.32295686719636774,
       "grad_norm": 0.0,
-      "learning_rate": 1.609684308036945e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.5826877907196084e-05,
+      "loss": 0.9501,
       "step": 11381
     },
     {
-      "epoch": 0.31259784131169155,
+      "epoch": 0.32298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.6096137986420216e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.5826130951432373e-05,
+      "loss": 1.0262,
       "step": 11382
     },
     {
-      "epoch": 0.31262530553953477,
+      "epoch": 0.32301362088535757,
       "grad_norm": 0.0,
-      "learning_rate": 1.6095432844235973e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.5825383946455075e-05,
+      "loss": 0.9441,
       "step": 11383
     },
     {
-      "epoch": 0.312652769767378,
+      "epoch": 0.3230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.6094727653822307e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.5824636892270496e-05,
+      "loss": 0.8712,
       "step": 11384
     },
     {
-      "epoch": 0.3126802339952212,
+      "epoch": 0.32307037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 1.6094022415184792e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.5823889788884944e-05,
+      "loss": 0.8765,
       "step": 11385
     },
     {
-      "epoch": 0.31270769822306443,
+      "epoch": 0.3230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.6093317128329016e-05,
-      "loss": 0.8154,
+      "learning_rate": 1.5823142636304736e-05,
+      "loss": 1.0189,
       "step": 11386
     },
     {
-      "epoch": 0.3127351624509077,
+      "epoch": 0.3231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.609261179326055e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.582239543453618e-05,
+      "loss": 1.0279,
       "step": 11387
     },
     {
-      "epoch": 0.31276262667875093,
+      "epoch": 0.32315550510783203,
       "grad_norm": 0.0,
-      "learning_rate": 1.6091906409984984e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.5821648183585588e-05,
+      "loss": 0.939,
       "step": 11388
     },
     {
-      "epoch": 0.31279009090659415,
+      "epoch": 0.3231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.609120097850789e-05,
-      "loss": 0.965,
+      "learning_rate": 1.582090088345927e-05,
+      "loss": 0.9232,
       "step": 11389
     },
     {
-      "epoch": 0.3128175551344374,
+      "epoch": 0.3232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.6090495498834858e-05,
-      "loss": 0.8633,
+      "learning_rate": 1.5820153534163543e-05,
+      "loss": 1.0107,
       "step": 11390
     },
     {
-      "epoch": 0.31284501936228065,
+      "epoch": 0.32324063564131666,
       "grad_norm": 0.0,
-      "learning_rate": 1.6089789970971465e-05,
-      "loss": 1.0848,
+      "learning_rate": 1.5819406135704717e-05,
+      "loss": 0.9984,
       "step": 11391
     },
     {
-      "epoch": 0.3128724835901239,
+      "epoch": 0.3232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.6089084394923292e-05,
-      "loss": 0.957,
+      "learning_rate": 1.581865868808911e-05,
+      "loss": 1.0094,
       "step": 11392
     },
     {
-      "epoch": 0.3128999478179671,
+      "epoch": 0.3232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.6088378770695926e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.5817911191323025e-05,
+      "loss": 0.9912,
       "step": 11393
     },
     {
-      "epoch": 0.3129274120458103,
+      "epoch": 0.32332576617480135,
       "grad_norm": 0.0,
-      "learning_rate": 1.6087673098294953e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.5817163645412784e-05,
+      "loss": 0.8811,
       "step": 11394
     },
     {
-      "epoch": 0.3129548762736536,
+      "epoch": 0.32335414301929627,
       "grad_norm": 0.0,
-      "learning_rate": 1.6086967377725947e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.58164160503647e-05,
+      "loss": 1.0307,
       "step": 11395
     },
     {
-      "epoch": 0.3129823405014968,
+      "epoch": 0.32338251986379113,
       "grad_norm": 0.0,
-      "learning_rate": 1.60862616089945e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.5815668406185093e-05,
+      "loss": 0.9369,
       "step": 11396
     },
     {
-      "epoch": 0.31300980472934004,
+      "epoch": 0.32341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.6085555792106192e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.5814920712880267e-05,
+      "loss": 0.922,
       "step": 11397
     },
     {
-      "epoch": 0.31303726895718326,
+      "epoch": 0.3234392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.608484992706661e-05,
-      "loss": 0.9165,
+      "learning_rate": 1.581417297045655e-05,
+      "loss": 0.9765,
       "step": 11398
     },
     {
-      "epoch": 0.3130647331850265,
+      "epoch": 0.3234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.6084144013881336e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.581342517892025e-05,
+      "loss": 0.9127,
       "step": 11399
     },
     {
-      "epoch": 0.31309219741286975,
+      "epoch": 0.32349602724177073,
       "grad_norm": 0.0,
-      "learning_rate": 1.608343805255596e-05,
-      "loss": 1.1015,
+      "learning_rate": 1.5812677338277684e-05,
+      "loss": 0.9576,
       "step": 11400
     },
     {
-      "epoch": 0.313119661640713,
+      "epoch": 0.3235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.608273204309607e-05,
-      "loss": 1.0342,
+      "learning_rate": 1.5811929448535173e-05,
+      "loss": 0.845,
       "step": 11401
     },
     {
-      "epoch": 0.3131471258685562,
+      "epoch": 0.3235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.608202598550724e-05,
-      "loss": 1.0164,
+      "learning_rate": 1.5811181509699033e-05,
+      "loss": 1.0322,
       "step": 11402
     },
     {
-      "epoch": 0.3131745900963994,
+      "epoch": 0.32358115777525537,
       "grad_norm": 0.0,
-      "learning_rate": 1.608131987979507e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.5810433521775583e-05,
+      "loss": 0.9544,
       "step": 11403
     },
     {
-      "epoch": 0.3132020543242427,
+      "epoch": 0.3236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.608061372596514e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.5809685484771136e-05,
+      "loss": 0.8891,
       "step": 11404
     },
     {
-      "epoch": 0.3132295185520859,
+      "epoch": 0.3236379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.6079907524023036e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.5808937398692016e-05,
+      "loss": 1.0131,
       "step": 11405
     },
     {
-      "epoch": 0.31325698277992914,
+      "epoch": 0.32366628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 1.6079201273974355e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.5808189263544542e-05,
+      "loss": 0.9753,
       "step": 11406
     },
     {
-      "epoch": 0.31328444700777236,
+      "epoch": 0.323694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.607849497582467e-05,
-      "loss": 0.9011,
+      "learning_rate": 1.5807441079335033e-05,
+      "loss": 0.9309,
       "step": 11407
     },
     {
-      "epoch": 0.31331191123561564,
+      "epoch": 0.32372304199772983,
       "grad_norm": 0.0,
-      "learning_rate": 1.6077788629579584e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.5806692846069807e-05,
+      "loss": 0.8524,
       "step": 11408
     },
     {
-      "epoch": 0.31333937546345886,
+      "epoch": 0.32375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.607708223524468e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.5805944563755183e-05,
+      "loss": 0.9542,
       "step": 11409
     },
     {
-      "epoch": 0.3133668396913021,
+      "epoch": 0.32377979568671966,
       "grad_norm": 0.0,
-      "learning_rate": 1.6076375792825545e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.580519623239749e-05,
+      "loss": 0.9872,
       "step": 11410
     },
     {
-      "epoch": 0.3133943039191453,
+      "epoch": 0.3238081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.6075669302327775e-05,
-      "loss": 1.0507,
+      "learning_rate": 1.580444785200304e-05,
+      "loss": 0.9274,
       "step": 11411
     },
     {
-      "epoch": 0.3134217681469885,
+      "epoch": 0.32383654937570944,
       "grad_norm": 0.0,
-      "learning_rate": 1.607496276375695e-05,
-      "loss": 0.9009,
+      "learning_rate": 1.5803699422578157e-05,
+      "loss": 0.9155,
       "step": 11412
     },
     {
-      "epoch": 0.3134492323748318,
+      "epoch": 0.3238649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.607425617711867e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.580295094412916e-05,
+      "loss": 0.9211,
       "step": 11413
     },
     {
-      "epoch": 0.313476696602675,
+      "epoch": 0.3238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.6073549542418522e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.5802202416662386e-05,
+      "loss": 0.937,
       "step": 11414
     },
     {
-      "epoch": 0.31350416083051824,
+      "epoch": 0.32392167990919407,
       "grad_norm": 0.0,
-      "learning_rate": 1.6072842859662098e-05,
-      "loss": 0.9395,
+      "learning_rate": 1.5801453840184143e-05,
+      "loss": 1.0465,
       "step": 11415
     },
     {
-      "epoch": 0.31353162505836146,
+      "epoch": 0.323950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.607213612885499e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.5800705214700758e-05,
+      "loss": 0.9333,
       "step": 11416
     },
     {
-      "epoch": 0.31355908928620474,
+      "epoch": 0.3239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.607142935000278e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.5799956540218553e-05,
+      "loss": 0.9093,
       "step": 11417
     },
     {
-      "epoch": 0.31358655351404796,
+      "epoch": 0.32400681044267876,
       "grad_norm": 0.0,
-      "learning_rate": 1.607072252311108e-05,
-      "loss": 1.0938,
+      "learning_rate": 1.5799207816743857e-05,
+      "loss": 0.8482,
       "step": 11418
     },
     {
-      "epoch": 0.3136140177418912,
+      "epoch": 0.3240351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.6070015648185466e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.5798459044282993e-05,
+      "loss": 0.9884,
       "step": 11419
     },
     {
-      "epoch": 0.3136414819697344,
+      "epoch": 0.32406356413166854,
       "grad_norm": 0.0,
-      "learning_rate": 1.606930872523154e-05,
-      "loss": 1.0739,
+      "learning_rate": 1.579771022284228e-05,
+      "loss": 1.0344,
       "step": 11420
     },
     {
-      "epoch": 0.3136689461975777,
+      "epoch": 0.32409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.606860175425489e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.5796961352428052e-05,
+      "loss": 0.9695,
       "step": 11421
     },
     {
-      "epoch": 0.3136964104254209,
+      "epoch": 0.32412031782065837,
       "grad_norm": 0.0,
-      "learning_rate": 1.6067894735261115e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.579621243304663e-05,
+      "loss": 1.0598,
       "step": 11422
     },
     {
-      "epoch": 0.3137238746532641,
+      "epoch": 0.3241486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.6067187668255804e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.5795463464704342e-05,
+      "loss": 0.9722,
       "step": 11423
     },
     {
-      "epoch": 0.31375133888110734,
+      "epoch": 0.32417707150964814,
       "grad_norm": 0.0,
-      "learning_rate": 1.6066480553244555e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.5794714447407514e-05,
+      "loss": 0.8957,
       "step": 11424
     },
     {
-      "epoch": 0.31377880310895057,
+      "epoch": 0.324205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.606577339023296e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.5793965381162467e-05,
+      "loss": 0.9292,
       "step": 11425
     },
     {
-      "epoch": 0.31380626733679384,
+      "epoch": 0.3242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.6065066179226615e-05,
-      "loss": 0.8853,
+      "learning_rate": 1.579321626597554e-05,
+      "loss": 0.9901,
       "step": 11426
     },
     {
-      "epoch": 0.31383373156463706,
+      "epoch": 0.32426220204313283,
       "grad_norm": 0.0,
-      "learning_rate": 1.6064358920231122e-05,
-      "loss": 1.0319,
+      "learning_rate": 1.5792467101853054e-05,
+      "loss": 1.0222,
       "step": 11427
     },
     {
-      "epoch": 0.3138611957924803,
+      "epoch": 0.3242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.606365161325207e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.5791717888801334e-05,
+      "loss": 0.9578,
       "step": 11428
     },
     {
-      "epoch": 0.3138886600203235,
+      "epoch": 0.3243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.6062944258295056e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.5790968626826716e-05,
+      "loss": 0.9806,
       "step": 11429
     },
     {
-      "epoch": 0.3139161242481668,
+      "epoch": 0.32434733257661746,
       "grad_norm": 0.0,
-      "learning_rate": 1.606223685536568e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.5790219315935526e-05,
+      "loss": 0.9241,
       "step": 11430
     },
     {
-      "epoch": 0.31394358847601,
+      "epoch": 0.3243757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.606152940446954e-05,
-      "loss": 0.963,
+      "learning_rate": 1.5789469956134095e-05,
+      "loss": 0.9325,
       "step": 11431
     },
     {
-      "epoch": 0.3139710527038532,
+      "epoch": 0.32440408626560724,
       "grad_norm": 0.0,
-      "learning_rate": 1.6060821905612226e-05,
-      "loss": 1.0419,
+      "learning_rate": 1.578872054742875e-05,
+      "loss": 0.9322,
       "step": 11432
     },
     {
-      "epoch": 0.31399851693169645,
+      "epoch": 0.32443246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.6060114358799345e-05,
-      "loss": 1.0402,
+      "learning_rate": 1.5787971089825817e-05,
+      "loss": 0.9929,
       "step": 11433
     },
     {
-      "epoch": 0.3140259811595397,
+      "epoch": 0.32446083995459707,
       "grad_norm": 0.0,
-      "learning_rate": 1.6059406764036492e-05,
-      "loss": 1.0462,
+      "learning_rate": 1.5787221583331636e-05,
+      "loss": 0.9808,
       "step": 11434
     },
     {
-      "epoch": 0.31405344538738295,
+      "epoch": 0.32448921679909193,
       "grad_norm": 0.0,
-      "learning_rate": 1.6058699121329263e-05,
-      "loss": 1.009,
+      "learning_rate": 1.5786472027952533e-05,
+      "loss": 0.8708,
       "step": 11435
     },
     {
-      "epoch": 0.31408090961522617,
+      "epoch": 0.32451759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 1.6057991430683263e-05,
-      "loss": 0.9891,
+      "learning_rate": 1.5785722423694842e-05,
+      "loss": 0.8421,
       "step": 11436
     },
     {
-      "epoch": 0.3141083738430694,
+      "epoch": 0.3245459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.6057283692104085e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.5784972770564895e-05,
+      "loss": 0.9572,
       "step": 11437
     },
     {
-      "epoch": 0.3141358380709126,
+      "epoch": 0.3245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.6056575905597337e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.578422306856902e-05,
+      "loss": 0.9592,
       "step": 11438
     },
     {
-      "epoch": 0.3141633022987559,
+      "epoch": 0.32460272417707153,
       "grad_norm": 0.0,
-      "learning_rate": 1.605586807116861e-05,
-      "loss": 1.0699,
+      "learning_rate": 1.5783473317713553e-05,
+      "loss": 0.9415,
       "step": 11439
     },
     {
-      "epoch": 0.3141907665265991,
+      "epoch": 0.3246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6055160188823507e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.578272351800483e-05,
+      "loss": 0.9656,
       "step": 11440
     },
     {
-      "epoch": 0.31421823075444233,
+      "epoch": 0.3246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.6054452258567634e-05,
-      "loss": 0.9084,
+      "learning_rate": 1.5781973669449175e-05,
+      "loss": 0.9197,
       "step": 11441
     },
     {
-      "epoch": 0.31424569498228555,
+      "epoch": 0.32468785471055617,
       "grad_norm": 0.0,
-      "learning_rate": 1.605374428040659e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.5781223772052934e-05,
+      "loss": 0.7952,
       "step": 11442
     },
     {
-      "epoch": 0.3142731592101288,
+      "epoch": 0.3247162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.605303625434597e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.5780473825822434e-05,
+      "loss": 0.9616,
       "step": 11443
     },
     {
-      "epoch": 0.31430062343797205,
+      "epoch": 0.324744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6052328180391393e-05,
-      "loss": 0.8782,
+      "learning_rate": 1.5779723830764013e-05,
+      "loss": 0.9204,
       "step": 11444
     },
     {
-      "epoch": 0.31432808766581527,
+      "epoch": 0.32477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.6051620058548444e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.5778973786884005e-05,
+      "loss": 0.9844,
       "step": 11445
     },
     {
-      "epoch": 0.3143555518936585,
+      "epoch": 0.32480136208853577,
       "grad_norm": 0.0,
-      "learning_rate": 1.6050911888822733e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.5778223694188746e-05,
+      "loss": 0.9984,
       "step": 11446
     },
     {
-      "epoch": 0.31438301612150177,
+      "epoch": 0.32482973893303063,
       "grad_norm": 0.0,
-      "learning_rate": 1.6050203671219867e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.577747355268457e-05,
+      "loss": 0.9263,
       "step": 11447
     },
     {
-      "epoch": 0.314410480349345,
+      "epoch": 0.32485811577752555,
       "grad_norm": 0.0,
-      "learning_rate": 1.6049495405745444e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.5776723362377816e-05,
+      "loss": 0.8805,
       "step": 11448
     },
     {
-      "epoch": 0.3144379445771882,
+      "epoch": 0.3248864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.6048787092405073e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.577597312327482e-05,
+      "loss": 0.8879,
       "step": 11449
     },
     {
-      "epoch": 0.31446540880503143,
+      "epoch": 0.3249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.604807873120435e-05,
-      "loss": 0.8034,
+      "learning_rate": 1.577522283538192e-05,
+      "loss": 0.9815,
       "step": 11450
     },
     {
-      "epoch": 0.31449287303287465,
+      "epoch": 0.32494324631101024,
       "grad_norm": 0.0,
-      "learning_rate": 1.6047370322148886e-05,
-      "loss": 1.1122,
+      "learning_rate": 1.577447249870545e-05,
+      "loss": 0.9393,
       "step": 11451
     },
     {
-      "epoch": 0.31452033726071793,
+      "epoch": 0.3249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.6046661865244288e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.5773722113251755e-05,
+      "loss": 0.9838,
       "step": 11452
     },
     {
-      "epoch": 0.31454780148856115,
+      "epoch": 0.325,
       "grad_norm": 0.0,
-      "learning_rate": 1.604595336049616e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.5772971679027168e-05,
+      "loss": 0.8106,
       "step": 11453
     },
     {
-      "epoch": 0.3145752657164044,
+      "epoch": 0.32502837684449487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6045244807910102e-05,
-      "loss": 1.116,
+      "learning_rate": 1.5772221196038032e-05,
+      "loss": 0.892,
       "step": 11454
     },
     {
-      "epoch": 0.3146027299442476,
+      "epoch": 0.3250567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.604453620749173e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.5771470664290682e-05,
+      "loss": 0.9358,
       "step": 11455
     },
     {
-      "epoch": 0.31463019417209087,
+      "epoch": 0.3250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.6043827559246647e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.5770720083791463e-05,
+      "loss": 0.951,
       "step": 11456
     },
     {
-      "epoch": 0.3146576583999341,
+      "epoch": 0.32511350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 1.6043118863180453e-05,
-      "loss": 1.0068,
+      "learning_rate": 1.5769969454546704e-05,
+      "loss": 0.9392,
       "step": 11457
     },
     {
-      "epoch": 0.3146851226277773,
+      "epoch": 0.3251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.6042410119298768e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.576921877656276e-05,
+      "loss": 0.8944,
       "step": 11458
     },
     {
-      "epoch": 0.31471258685562054,
+      "epoch": 0.32517026106696933,
       "grad_norm": 0.0,
-      "learning_rate": 1.6041701327607193e-05,
-      "loss": 1.004,
+      "learning_rate": 1.5768468049845963e-05,
+      "loss": 1.0092,
       "step": 11459
     },
     {
-      "epoch": 0.3147400510834638,
+      "epoch": 0.32519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.6040992488111334e-05,
-      "loss": 0.8318,
+      "learning_rate": 1.5767717274402657e-05,
+      "loss": 0.8975,
       "step": 11460
     },
     {
-      "epoch": 0.31476751531130703,
+      "epoch": 0.32522701475595917,
       "grad_norm": 0.0,
-      "learning_rate": 1.60402836008168e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.576696645023919e-05,
+      "loss": 0.941,
       "step": 11461
     },
     {
-      "epoch": 0.31479497953915025,
+      "epoch": 0.325255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.6039574665729208e-05,
-      "loss": 0.8517,
+      "learning_rate": 1.576621557736189e-05,
+      "loss": 0.8573,
       "step": 11462
     },
     {
-      "epoch": 0.3148224437669935,
+      "epoch": 0.32528376844494894,
       "grad_norm": 0.0,
-      "learning_rate": 1.6038865682854158e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.5765464655777114e-05,
+      "loss": 0.9674,
       "step": 11463
     },
     {
-      "epoch": 0.3148499079948367,
+      "epoch": 0.3253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.6038156652197265e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.5764713685491195e-05,
+      "loss": 1.0109,
       "step": 11464
     },
     {
-      "epoch": 0.31487737222268,
+      "epoch": 0.3253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.603744757376414e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.5763962666510483e-05,
+      "loss": 1.037,
       "step": 11465
     },
     {
-      "epoch": 0.3149048364505232,
+      "epoch": 0.3253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.6036738447560387e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.5763211598841318e-05,
+      "loss": 0.9707,
       "step": 11466
     },
     {
-      "epoch": 0.3149323006783664,
+      "epoch": 0.3253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.603602927359162e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.5762460482490042e-05,
+      "loss": 1.0406,
       "step": 11467
     },
     {
-      "epoch": 0.31495976490620964,
+      "epoch": 0.3254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.6035320051863458e-05,
-      "loss": 1.1068,
+      "learning_rate": 1.576170931746301e-05,
+      "loss": 1.0479,
       "step": 11468
     },
     {
-      "epoch": 0.3149872291340529,
+      "epoch": 0.32545402951191826,
       "grad_norm": 0.0,
-      "learning_rate": 1.60346107823815e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.5760958103766556e-05,
+      "loss": 0.954,
       "step": 11469
     },
     {
-      "epoch": 0.31501469336189614,
+      "epoch": 0.3254824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.6033901465151364e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.576020684140703e-05,
+      "loss": 0.9005,
       "step": 11470
     },
     {
-      "epoch": 0.31504215758973936,
+      "epoch": 0.32551078320090804,
       "grad_norm": 0.0,
-      "learning_rate": 1.6033192100178664e-05,
-      "loss": 0.8907,
+      "learning_rate": 1.5759455530390782e-05,
+      "loss": 0.9528,
       "step": 11471
     },
     {
-      "epoch": 0.3150696218175826,
+      "epoch": 0.32553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.603248268746901e-05,
-      "loss": 1.0141,
+      "learning_rate": 1.575870417072415e-05,
+      "loss": 0.9455,
       "step": 11472
     },
     {
-      "epoch": 0.31509708604542586,
+      "epoch": 0.32556753688989787,
       "grad_norm": 0.0,
-      "learning_rate": 1.603177322702802e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.5757952762413484e-05,
+      "loss": 0.8655,
       "step": 11473
     },
     {
-      "epoch": 0.3151245502732691,
+      "epoch": 0.32559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.60310637188613e-05,
-      "loss": 1.016,
+      "learning_rate": 1.5757201305465134e-05,
+      "loss": 0.9849,
       "step": 11474
     },
     {
-      "epoch": 0.3151520145011123,
+      "epoch": 0.32562429057888764,
       "grad_norm": 0.0,
-      "learning_rate": 1.6030354162974468e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.5756449799885448e-05,
+      "loss": 0.9401,
       "step": 11475
     },
     {
-      "epoch": 0.3151794787289555,
+      "epoch": 0.3256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.602964455937314e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.5755698245680768e-05,
+      "loss": 0.936,
       "step": 11476
     },
     {
-      "epoch": 0.3152069429567988,
+      "epoch": 0.3256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.6028934908062926e-05,
-      "loss": 0.928,
+      "learning_rate": 1.575494664285745e-05,
+      "loss": 0.9796,
       "step": 11477
     },
     {
-      "epoch": 0.315234407184642,
+      "epoch": 0.3257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.6028225209049444e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.5754194991421835e-05,
+      "loss": 0.8907,
       "step": 11478
     },
     {
-      "epoch": 0.31526187141248524,
+      "epoch": 0.3257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.602751546233831e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.575344329138028e-05,
+      "loss": 0.9365,
       "step": 11479
     },
     {
-      "epoch": 0.31528933564032846,
+      "epoch": 0.3257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.602680566793514e-05,
-      "loss": 0.8539,
+      "learning_rate": 1.575269154273913e-05,
+      "loss": 0.845,
       "step": 11480
     },
     {
-      "epoch": 0.3153167998681717,
+      "epoch": 0.32579455164585697,
       "grad_norm": 0.0,
-      "learning_rate": 1.6026095825845548e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.5751939745504733e-05,
+      "loss": 1.0008,
       "step": 11481
     },
     {
-      "epoch": 0.31534426409601496,
+      "epoch": 0.3258229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.6025385936075148e-05,
-      "loss": 0.7907,
+      "learning_rate": 1.5751187899683445e-05,
+      "loss": 1.0467,
       "step": 11482
     },
     {
-      "epoch": 0.3153717283238582,
+      "epoch": 0.32585130533484674,
       "grad_norm": 0.0,
-      "learning_rate": 1.6024675998629565e-05,
-      "loss": 0.866,
+      "learning_rate": 1.5750436005281612e-05,
+      "loss": 0.9822,
       "step": 11483
     },
     {
-      "epoch": 0.3153991925517014,
+      "epoch": 0.32587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.602396601351441e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.574968406230559e-05,
+      "loss": 0.9522,
       "step": 11484
     },
     {
-      "epoch": 0.3154266567795446,
+      "epoch": 0.32590805902383657,
       "grad_norm": 0.0,
-      "learning_rate": 1.6023255980735305e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.5748932070761727e-05,
+      "loss": 0.9091,
       "step": 11485
     },
     {
-      "epoch": 0.3154541210073879,
+      "epoch": 0.32593643586833143,
       "grad_norm": 0.0,
-      "learning_rate": 1.602254590029786e-05,
-      "loss": 0.8812,
+      "learning_rate": 1.5748180030656377e-05,
+      "loss": 0.941,
       "step": 11486
     },
     {
-      "epoch": 0.3154815852352311,
+      "epoch": 0.32596481271282635,
       "grad_norm": 0.0,
-      "learning_rate": 1.6021835772207702e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.5747427941995893e-05,
+      "loss": 0.9206,
       "step": 11487
     },
     {
-      "epoch": 0.31550904946307434,
+      "epoch": 0.3259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.6021125596470447e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.5746675804786625e-05,
+      "loss": 0.9796,
       "step": 11488
     },
     {
-      "epoch": 0.31553651369091756,
+      "epoch": 0.3260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.6020415373091712e-05,
-      "loss": 0.8874,
+      "learning_rate": 1.5745923619034928e-05,
+      "loss": 0.9875,
       "step": 11489
     },
     {
-      "epoch": 0.31556397791876084,
+      "epoch": 0.32604994324631104,
       "grad_norm": 0.0,
-      "learning_rate": 1.6019705102077122e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.574517138474716e-05,
+      "loss": 0.8385,
       "step": 11490
     },
     {
-      "epoch": 0.31559144214660406,
+      "epoch": 0.3260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.6018994783432293e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.5744419101929665e-05,
+      "loss": 0.9192,
       "step": 11491
     },
     {
-      "epoch": 0.3156189063744473,
+      "epoch": 0.3261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.6018284417162844e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.5743666770588807e-05,
+      "loss": 1.1035,
       "step": 11492
     },
     {
-      "epoch": 0.3156463706022905,
+      "epoch": 0.32613507377979567,
       "grad_norm": 0.0,
-      "learning_rate": 1.60175740032744e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.5742914390730938e-05,
+      "loss": 1.0124,
       "step": 11493
     },
     {
-      "epoch": 0.3156738348301337,
+      "epoch": 0.3261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.6016863541772578e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.574216196236241e-05,
+      "loss": 0.9682,
       "step": 11494
     },
     {
-      "epoch": 0.315701299057977,
+      "epoch": 0.32619182746878544,
       "grad_norm": 0.0,
-      "learning_rate": 1.6016153032663e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.5741409485489584e-05,
+      "loss": 1.0385,
       "step": 11495
     },
     {
-      "epoch": 0.3157287632858202,
+      "epoch": 0.32622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.6015442475951287e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.574065696011881e-05,
+      "loss": 0.992,
       "step": 11496
     },
     {
-      "epoch": 0.31575622751366345,
+      "epoch": 0.3262485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.6014731871643065e-05,
-      "loss": 1.0439,
+      "learning_rate": 1.5739904386256456e-05,
+      "loss": 1.0016,
       "step": 11497
     },
     {
-      "epoch": 0.31578369174150667,
+      "epoch": 0.32627695800227013,
       "grad_norm": 0.0,
-      "learning_rate": 1.6014021219743955e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.5739151763908867e-05,
+      "loss": 0.9821,
       "step": 11498
     },
     {
-      "epoch": 0.31581115596934994,
+      "epoch": 0.32630533484676505,
       "grad_norm": 0.0,
-      "learning_rate": 1.6013310520259578e-05,
-      "loss": 0.9913,
+      "learning_rate": 1.5738399093082404e-05,
+      "loss": 1.0428,
       "step": 11499
     },
     {
-      "epoch": 0.31583862019719317,
+      "epoch": 0.3263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.6012599773195564e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.573764637378343e-05,
+      "loss": 0.8846,
       "step": 11500
     },
     {
-      "epoch": 0.3158660844250364,
+      "epoch": 0.3263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.6011888978557525e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.5736893606018297e-05,
+      "loss": 0.9968,
       "step": 11501
     },
     {
-      "epoch": 0.3158935486528796,
+      "epoch": 0.32639046538024974,
       "grad_norm": 0.0,
-      "learning_rate": 1.6011178136351095e-05,
-      "loss": 1.0073,
+      "learning_rate": 1.5736140789793367e-05,
+      "loss": 0.9337,
       "step": 11502
     },
     {
-      "epoch": 0.3159210128807229,
+      "epoch": 0.3264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.6010467246581894e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.5735387925114992e-05,
+      "loss": 0.7919,
       "step": 11503
     },
     {
-      "epoch": 0.3159484771085661,
+      "epoch": 0.3264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.6009756309255548e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.5734635011989544e-05,
+      "loss": 0.8554,
       "step": 11504
     },
     {
-      "epoch": 0.3159759413364093,
+      "epoch": 0.3264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.6009045324377683e-05,
-      "loss": 1.0981,
+      "learning_rate": 1.5733882050423376e-05,
+      "loss": 0.9105,
       "step": 11505
     },
     {
-      "epoch": 0.31600340556425255,
+      "epoch": 0.3265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.600833429195392e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.5733129040422844e-05,
+      "loss": 0.9231,
       "step": 11506
     },
     {
-      "epoch": 0.31603086979209577,
+      "epoch": 0.3265323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.6007623211989892e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.573237598199432e-05,
+      "loss": 0.8616,
       "step": 11507
     },
     {
-      "epoch": 0.31605833401993905,
+      "epoch": 0.32656072644721906,
       "grad_norm": 0.0,
-      "learning_rate": 1.6006912084491222e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.5731622875144155e-05,
+      "loss": 0.961,
       "step": 11508
     },
     {
-      "epoch": 0.31608579824778227,
+      "epoch": 0.326589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.6006200909463537e-05,
-      "loss": 0.8986,
+      "learning_rate": 1.5730869719878713e-05,
+      "loss": 1.004,
       "step": 11509
     },
     {
-      "epoch": 0.3161132624756255,
+      "epoch": 0.32661748013620884,
       "grad_norm": 0.0,
-      "learning_rate": 1.6005489686912462e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.5730116516204356e-05,
+      "loss": 0.903,
       "step": 11510
     },
     {
-      "epoch": 0.3161407267034687,
+      "epoch": 0.32664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.600477841684363e-05,
-      "loss": 0.8021,
+      "learning_rate": 1.572936326412745e-05,
+      "loss": 1.0392,
       "step": 11511
     },
     {
-      "epoch": 0.316168190931312,
+      "epoch": 0.3266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.600406709926266e-05,
-      "loss": 1.035,
+      "learning_rate": 1.572860996365435e-05,
+      "loss": 0.9652,
       "step": 11512
     },
     {
-      "epoch": 0.3161956551591552,
+      "epoch": 0.3267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.600335573417519e-05,
-      "loss": 0.8405,
+      "learning_rate": 1.572785661479143e-05,
+      "loss": 0.926,
       "step": 11513
     },
     {
-      "epoch": 0.31622311938699843,
+      "epoch": 0.32673098751418844,
       "grad_norm": 0.0,
-      "learning_rate": 1.600264432158684e-05,
-      "loss": 1.0003,
+      "learning_rate": 1.5727103217545043e-05,
+      "loss": 0.9806,
       "step": 11514
     },
     {
-      "epoch": 0.31625058361484165,
+      "epoch": 0.3267593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.6001932861503246e-05,
-      "loss": 0.9759,
+      "learning_rate": 1.572634977192156e-05,
+      "loss": 0.8633,
       "step": 11515
     },
     {
-      "epoch": 0.31627804784268493,
+      "epoch": 0.3267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.6001221353930033e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.5725596277927342e-05,
+      "loss": 0.9585,
       "step": 11516
     },
     {
-      "epoch": 0.31630551207052815,
+      "epoch": 0.3268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.600050979887283e-05,
-      "loss": 0.8869,
+      "learning_rate": 1.572484273556876e-05,
+      "loss": 0.9529,
       "step": 11517
     },
     {
-      "epoch": 0.31633297629837137,
+      "epoch": 0.326844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.5999798196337274e-05,
-      "loss": 1.0436,
+      "learning_rate": 1.572408914485217e-05,
+      "loss": 0.975,
       "step": 11518
     },
     {
-      "epoch": 0.3163604405262146,
+      "epoch": 0.3268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.599908654632899e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.572333550578394e-05,
+      "loss": 0.9051,
       "step": 11519
     },
     {
-      "epoch": 0.3163879047540578,
+      "epoch": 0.32690124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.5998374848853605e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.572258181837044e-05,
+      "loss": 0.9786,
       "step": 11520
     },
     {
-      "epoch": 0.3164153689819011,
+      "epoch": 0.3269296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.5997663103916756e-05,
-      "loss": 0.9057,
+      "learning_rate": 1.572182808261803e-05,
+      "loss": 1.0157,
       "step": 11521
     },
     {
-      "epoch": 0.3164428332097443,
+      "epoch": 0.32695800227014754,
       "grad_norm": 0.0,
-      "learning_rate": 1.5996951311524073e-05,
-      "loss": 0.992,
+      "learning_rate": 1.5721074298533086e-05,
+      "loss": 0.9024,
       "step": 11522
     },
     {
-      "epoch": 0.31647029743758753,
+      "epoch": 0.32698637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.599623947168119e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.572032046612197e-05,
+      "loss": 0.9291,
       "step": 11523
     },
     {
-      "epoch": 0.31649776166543075,
+      "epoch": 0.32701475595913737,
       "grad_norm": 0.0,
-      "learning_rate": 1.5995527584393736e-05,
-      "loss": 1.0077,
+      "learning_rate": 1.5719566585391048e-05,
+      "loss": 0.745,
       "step": 11524
     },
     {
-      "epoch": 0.31652522589327403,
+      "epoch": 0.32704313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 1.5994815649667348e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.571881265634669e-05,
+      "loss": 0.9517,
       "step": 11525
     },
     {
-      "epoch": 0.31655269012111725,
+      "epoch": 0.32707150964812715,
       "grad_norm": 0.0,
-      "learning_rate": 1.5994103667507653e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.5718058678995264e-05,
+      "loss": 0.9819,
       "step": 11526
     },
     {
-      "epoch": 0.3165801543489605,
+      "epoch": 0.327099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.599339163792029e-05,
-      "loss": 0.8783,
+      "learning_rate": 1.571730465334314e-05,
+      "loss": 0.9728,
       "step": 11527
     },
     {
-      "epoch": 0.3166076185768037,
+      "epoch": 0.3271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.5992679560910892e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.5716550579396687e-05,
+      "loss": 0.9045,
       "step": 11528
     },
     {
-      "epoch": 0.316635082804647,
+      "epoch": 0.3271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.599196743648509e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.5715796457162274e-05,
+      "loss": 0.9994,
       "step": 11529
     },
     {
-      "epoch": 0.3166625470324902,
+      "epoch": 0.3271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.5991255264648523e-05,
-      "loss": 0.991,
+      "learning_rate": 1.5715042286646272e-05,
+      "loss": 0.942,
       "step": 11530
     },
     {
-      "epoch": 0.3166900112603334,
+      "epoch": 0.3272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.599054304540682e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.571428806785505e-05,
+      "loss": 0.9418,
       "step": 11531
     },
     {
-      "epoch": 0.31671747548817664,
+      "epoch": 0.32724177071509647,
       "grad_norm": 0.0,
-      "learning_rate": 1.5989830778765618e-05,
-      "loss": 1.0383,
+      "learning_rate": 1.571353380079498e-05,
+      "loss": 0.9395,
       "step": 11532
     },
     {
-      "epoch": 0.31674493971601986,
+      "epoch": 0.3272701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.598911846473056e-05,
-      "loss": 0.9002,
+      "learning_rate": 1.571277948547243e-05,
+      "loss": 0.8811,
       "step": 11533
     },
     {
-      "epoch": 0.31677240394386313,
+      "epoch": 0.32729852440408624,
       "grad_norm": 0.0,
-      "learning_rate": 1.5988406103307273e-05,
-      "loss": 0.9503,
+      "learning_rate": 1.571202512189378e-05,
+      "loss": 1.0225,
       "step": 11534
     },
     {
-      "epoch": 0.31679986817170636,
+      "epoch": 0.32732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.59876936945014e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.5711270710065397e-05,
+      "loss": 0.9222,
       "step": 11535
     },
     {
-      "epoch": 0.3168273323995496,
+      "epoch": 0.3273552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 1.5986981238318574e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.571051624999365e-05,
+      "loss": 0.9703,
       "step": 11536
     },
     {
-      "epoch": 0.3168547966273928,
+      "epoch": 0.32738365493757093,
       "grad_norm": 0.0,
-      "learning_rate": 1.598626873476443e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.570976174168492e-05,
+      "loss": 1.022,
       "step": 11537
     },
     {
-      "epoch": 0.3168822608552361,
+      "epoch": 0.32741203178206585,
       "grad_norm": 0.0,
-      "learning_rate": 1.5985556183844616e-05,
-      "loss": 0.951,
+      "learning_rate": 1.570900718514557e-05,
+      "loss": 0.9508,
       "step": 11538
     },
     {
-      "epoch": 0.3169097250830793,
+      "epoch": 0.3274404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.5984843585564757e-05,
-      "loss": 1.0044,
+      "learning_rate": 1.5708252580381983e-05,
+      "loss": 0.9822,
       "step": 11539
     },
     {
-      "epoch": 0.3169371893109225,
+      "epoch": 0.3274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.5984130939930503e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.570749792740053e-05,
+      "loss": 0.945,
       "step": 11540
     },
     {
-      "epoch": 0.31696465353876574,
+      "epoch": 0.32749716231555054,
       "grad_norm": 0.0,
-      "learning_rate": 1.598341824694748e-05,
-      "loss": 0.968,
+      "learning_rate": 1.5706743226207586e-05,
+      "loss": 0.923,
       "step": 11541
     },
     {
-      "epoch": 0.316992117766609,
+      "epoch": 0.3275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.5982705506621338e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.5705988476809522e-05,
+      "loss": 0.9183,
       "step": 11542
     },
     {
-      "epoch": 0.31701958199445224,
+      "epoch": 0.3275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.5981992718957712e-05,
-      "loss": 1.024,
+      "learning_rate": 1.570523367921272e-05,
+      "loss": 0.9128,
       "step": 11543
     },
     {
-      "epoch": 0.31704704622229546,
+      "epoch": 0.3275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.598127988396224e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.5704478833423552e-05,
+      "loss": 1.0389,
       "step": 11544
     },
     {
-      "epoch": 0.3170745104501387,
+      "epoch": 0.3276106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.598056700164057e-05,
-      "loss": 0.897,
+      "learning_rate": 1.5703723939448395e-05,
+      "loss": 0.9803,
       "step": 11545
     },
     {
-      "epoch": 0.3171019746779819,
+      "epoch": 0.32763904653802495,
       "grad_norm": 0.0,
-      "learning_rate": 1.5979854071998336e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.5702968997293625e-05,
+      "loss": 1.0346,
       "step": 11546
     },
     {
-      "epoch": 0.3171294389058252,
+      "epoch": 0.32766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.597914109504118e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.570221400696562e-05,
+      "loss": 0.8712,
       "step": 11547
     },
     {
-      "epoch": 0.3171569031336684,
+      "epoch": 0.3276958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.597842807077474e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.570145896847076e-05,
+      "loss": 0.9493,
       "step": 11548
     },
     {
-      "epoch": 0.3171843673615116,
+      "epoch": 0.32772417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.5977714999204662e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.5700703881815415e-05,
+      "loss": 0.8597,
       "step": 11549
     },
     {
-      "epoch": 0.31721183158935484,
+      "epoch": 0.32775255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 1.5977001880336588e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.569994874700597e-05,
+      "loss": 1.0652,
       "step": 11550
     },
     {
-      "epoch": 0.3172392958171981,
+      "epoch": 0.3277809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.597628871417616e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.56991935640488e-05,
+      "loss": 0.9891,
       "step": 11551
     },
     {
-      "epoch": 0.31726676004504134,
+      "epoch": 0.3278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.597557550072902e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.5698438332950287e-05,
+      "loss": 0.978,
       "step": 11552
     },
     {
-      "epoch": 0.31729422427288456,
+      "epoch": 0.32783768444948924,
       "grad_norm": 0.0,
-      "learning_rate": 1.5974862240000817e-05,
-      "loss": 0.9255,
+      "learning_rate": 1.569768305371681e-05,
+      "loss": 0.9812,
       "step": 11553
     },
     {
-      "epoch": 0.3173216885007278,
+      "epoch": 0.3278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.5974148931997184e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.5696927726354746e-05,
+      "loss": 0.9289,
       "step": 11554
     },
     {
-      "epoch": 0.31734915272857106,
+      "epoch": 0.327894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.5973435576723772e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.5696172350870484e-05,
+      "loss": 0.8001,
       "step": 11555
     },
     {
-      "epoch": 0.3173766169564143,
+      "epoch": 0.3279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.5972722174186228e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.569541692727039e-05,
+      "loss": 0.9044,
       "step": 11556
     },
     {
-      "epoch": 0.3174040811842575,
+      "epoch": 0.3279511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.5972008724390184e-05,
-      "loss": 0.988,
+      "learning_rate": 1.5694661455560852e-05,
+      "loss": 0.9847,
       "step": 11557
     },
     {
-      "epoch": 0.3174315454121007,
+      "epoch": 0.32797956867196365,
       "grad_norm": 0.0,
-      "learning_rate": 1.5971295227341295e-05,
-      "loss": 0.8646,
+      "learning_rate": 1.569390593574826e-05,
+      "loss": 0.9704,
       "step": 11558
     },
     {
-      "epoch": 0.31745900963994395,
+      "epoch": 0.32800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.597058168304521e-05,
-      "loss": 0.9239,
+      "learning_rate": 1.5693150367838986e-05,
+      "loss": 0.9562,
       "step": 11559
     },
     {
-      "epoch": 0.3174864738677872,
+      "epoch": 0.3280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.596986809150757e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.5692394751839413e-05,
+      "loss": 1.0901,
       "step": 11560
     },
     {
-      "epoch": 0.31751393809563044,
+      "epoch": 0.32806469920544834,
       "grad_norm": 0.0,
-      "learning_rate": 1.5969154452734018e-05,
-      "loss": 0.8276,
+      "learning_rate": 1.5691639087755928e-05,
+      "loss": 0.8848,
       "step": 11561
     },
     {
-      "epoch": 0.31754140232347366,
+      "epoch": 0.32809307604994326,
       "grad_norm": 0.0,
-      "learning_rate": 1.5968440766730206e-05,
-      "loss": 1.0119,
+      "learning_rate": 1.569088337559491e-05,
+      "loss": 0.8735,
       "step": 11562
     },
     {
-      "epoch": 0.3175688665513169,
+      "epoch": 0.3281214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.596772703350178e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.5690127615362746e-05,
+      "loss": 0.9195,
       "step": 11563
     },
     {
-      "epoch": 0.31759633077916016,
+      "epoch": 0.32814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.5967013253054383e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.5689371807065816e-05,
+      "loss": 0.9684,
       "step": 11564
     },
     {
-      "epoch": 0.3176237950070034,
+      "epoch": 0.32817820658342794,
       "grad_norm": 0.0,
-      "learning_rate": 1.5966299425393665e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.5688615950710507e-05,
+      "loss": 1.0361,
       "step": 11565
     },
     {
-      "epoch": 0.3176512592348466,
+      "epoch": 0.3282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.5965585550525277e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.5687860046303205e-05,
+      "loss": 0.9762,
       "step": 11566
     },
     {
-      "epoch": 0.3176787234626898,
+      "epoch": 0.3282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.5964871628454863e-05,
-      "loss": 0.9017,
+      "learning_rate": 1.5687104093850294e-05,
+      "loss": 0.8493,
       "step": 11567
     },
     {
-      "epoch": 0.3177061876905331,
+      "epoch": 0.3282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5964157659188078e-05,
-      "loss": 0.991,
+      "learning_rate": 1.5686348093358156e-05,
+      "loss": 0.9075,
       "step": 11568
     },
     {
-      "epoch": 0.3177336519183763,
+      "epoch": 0.3282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.5963443642730563e-05,
-      "loss": 0.8449,
+      "learning_rate": 1.5685592044833185e-05,
+      "loss": 0.9546,
       "step": 11569
     },
     {
-      "epoch": 0.31776111614621955,
+      "epoch": 0.3283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.5962729579087974e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.568483594828176e-05,
+      "loss": 0.9124,
       "step": 11570
     },
     {
-      "epoch": 0.31778858037406277,
+      "epoch": 0.32834846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 1.5962015468265958e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.568407980371027e-05,
+      "loss": 1.0675,
       "step": 11571
     },
     {
-      "epoch": 0.31781604460190604,
+      "epoch": 0.3283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.596130131027017e-05,
-      "loss": 0.9511,
+      "learning_rate": 1.5683323611125095e-05,
+      "loss": 0.9722,
       "step": 11572
     },
     {
-      "epoch": 0.31784350882974927,
+      "epoch": 0.32840522133938704,
       "grad_norm": 0.0,
-      "learning_rate": 1.596058710510625e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.568256737053264e-05,
+      "loss": 1.0499,
       "step": 11573
     },
     {
-      "epoch": 0.3178709730575925,
+      "epoch": 0.32843359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.5959872852779862e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.568181108193928e-05,
+      "loss": 0.9059,
       "step": 11574
     },
     {
-      "epoch": 0.3178984372854357,
+      "epoch": 0.3284619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.595915855329665e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.5681054745351408e-05,
+      "loss": 0.8611,
       "step": 11575
     },
     {
-      "epoch": 0.31792590151327893,
+      "epoch": 0.32849035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 1.5958444206662266e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.568029836077541e-05,
+      "loss": 0.958,
       "step": 11576
     },
     {
-      "epoch": 0.3179533657411222,
+      "epoch": 0.32851872871736665,
       "grad_norm": 0.0,
-      "learning_rate": 1.595772981288236e-05,
-      "loss": 1.04,
+      "learning_rate": 1.5679541928217674e-05,
+      "loss": 0.92,
       "step": 11577
     },
     {
-      "epoch": 0.31798082996896543,
+      "epoch": 0.3285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.5957015371962593e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.5678785447684592e-05,
+      "loss": 1.0367,
       "step": 11578
     },
     {
-      "epoch": 0.31800829419680865,
+      "epoch": 0.3285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.5956300883908613e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.5678028919182558e-05,
+      "loss": 0.8104,
       "step": 11579
     },
     {
-      "epoch": 0.31803575842465187,
+      "epoch": 0.3286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.5955586348726072e-05,
-      "loss": 0.9215,
+      "learning_rate": 1.5677272342717957e-05,
+      "loss": 0.8755,
       "step": 11580
     },
     {
-      "epoch": 0.31806322265249515,
+      "epoch": 0.3286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.5954871766420624e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.567651571829718e-05,
+      "loss": 1.0068,
       "step": 11581
     },
     {
-      "epoch": 0.31809068688033837,
+      "epoch": 0.3286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.5954157136997923e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.567575904592662e-05,
+      "loss": 0.9486,
       "step": 11582
     },
     {
-      "epoch": 0.3181181511081816,
+      "epoch": 0.32868898978433597,
       "grad_norm": 0.0,
-      "learning_rate": 1.5953442460463628e-05,
-      "loss": 0.9882,
+      "learning_rate": 1.5675002325612665e-05,
+      "loss": 0.8382,
       "step": 11583
     },
     {
-      "epoch": 0.3181456153360248,
+      "epoch": 0.3287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.5952727736823384e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.5674245557361716e-05,
+      "loss": 0.9942,
       "step": 11584
     },
     {
-      "epoch": 0.3181730795638681,
+      "epoch": 0.32874574347332575,
       "grad_norm": 0.0,
-      "learning_rate": 1.5952012966082857e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.5673488741180157e-05,
+      "loss": 0.9785,
       "step": 11585
     },
     {
-      "epoch": 0.3182005437917113,
+      "epoch": 0.32877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.59512981482477e-05,
-      "loss": 0.8149,
+      "learning_rate": 1.567273187707438e-05,
+      "loss": 0.9173,
       "step": 11586
     },
     {
-      "epoch": 0.31822800801955453,
+      "epoch": 0.3288024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.595058328332356e-05,
-      "loss": 1.0127,
+      "learning_rate": 1.5671974965050787e-05,
+      "loss": 0.9202,
       "step": 11587
     },
     {
-      "epoch": 0.31825547224739775,
+      "epoch": 0.32883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.5949868371316103e-05,
-      "loss": 0.8491,
+      "learning_rate": 1.5671218005115767e-05,
+      "loss": 1.0105,
       "step": 11588
     },
     {
-      "epoch": 0.318282936475241,
+      "epoch": 0.32885925085130535,
       "grad_norm": 0.0,
-      "learning_rate": 1.5949153412230984e-05,
-      "loss": 0.9394,
+      "learning_rate": 1.567046099727571e-05,
+      "loss": 1.0457,
       "step": 11589
     },
     {
-      "epoch": 0.31831040070308425,
+      "epoch": 0.3288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.5948438406073858e-05,
-      "loss": 0.9239,
+      "learning_rate": 1.5669703941537018e-05,
+      "loss": 0.9044,
       "step": 11590
     },
     {
-      "epoch": 0.31833786493092747,
+      "epoch": 0.3289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.5947723352850384e-05,
-      "loss": 0.9406,
+      "learning_rate": 1.5668946837906077e-05,
+      "loss": 1.0013,
       "step": 11591
     },
     {
-      "epoch": 0.3183653291587707,
+      "epoch": 0.32894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.5947008252566218e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.5668189686389293e-05,
+      "loss": 0.957,
       "step": 11592
     },
     {
-      "epoch": 0.3183927933866139,
+      "epoch": 0.3289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.594629310522702e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.5667432486993054e-05,
+      "loss": 0.8899,
       "step": 11593
     },
     {
-      "epoch": 0.3184202576144572,
+      "epoch": 0.3290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.5945577910838444e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.5666675239723757e-05,
+      "loss": 0.9631,
       "step": 11594
     },
     {
-      "epoch": 0.3184477218423004,
+      "epoch": 0.3290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.5944862669406157e-05,
-      "loss": 0.7421,
+      "learning_rate": 1.5665917944587797e-05,
+      "loss": 0.9833,
       "step": 11595
     },
     {
-      "epoch": 0.31847518607014363,
+      "epoch": 0.3290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.594414738093581e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.5665160601591576e-05,
+      "loss": 1.0098,
       "step": 11596
     },
     {
-      "epoch": 0.31850265029798686,
+      "epoch": 0.32908626560726445,
       "grad_norm": 0.0,
-      "learning_rate": 1.5943432045433067e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.566440321074149e-05,
+      "loss": 0.9854,
       "step": 11597
     },
     {
-      "epoch": 0.31853011452583013,
+      "epoch": 0.32911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.594271666290359e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.5663645772043935e-05,
+      "loss": 0.8593,
       "step": 11598
     },
     {
-      "epoch": 0.31855757875367335,
+      "epoch": 0.3291430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.5942001233353034e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.5662888285505305e-05,
+      "loss": 1.0611,
       "step": 11599
     },
     {
-      "epoch": 0.3185850429815166,
+      "epoch": 0.32917139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 1.5941285756787067e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.5662130751132008e-05,
+      "loss": 0.9971,
       "step": 11600
     },
     {
-      "epoch": 0.3186125072093598,
+      "epoch": 0.32919977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 1.594057023321134e-05,
-      "loss": 1.1127,
+      "learning_rate": 1.5661373168930437e-05,
+      "loss": 1.0654,
       "step": 11601
     },
     {
-      "epoch": 0.318639971437203,
+      "epoch": 0.3292281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.593985466263152e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.566061553890699e-05,
+      "loss": 0.812,
       "step": 11602
     },
     {
-      "epoch": 0.3186674356650463,
+      "epoch": 0.32925652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.5939139045053274e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.5659857861068074e-05,
+      "loss": 0.909,
       "step": 11603
     },
     {
-      "epoch": 0.3186948998928895,
+      "epoch": 0.32928490351872874,
       "grad_norm": 0.0,
-      "learning_rate": 1.593842338048226e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.5659100135420077e-05,
+      "loss": 0.9579,
       "step": 11604
     },
     {
-      "epoch": 0.31872236412073274,
+      "epoch": 0.3293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.5937707668924135e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.5658342361969413e-05,
+      "loss": 1.0282,
       "step": 11605
     },
     {
-      "epoch": 0.31874982834857596,
+      "epoch": 0.3293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5936991910384566e-05,
-      "loss": 1.0086,
+      "learning_rate": 1.5657584540722476e-05,
+      "loss": 0.8632,
       "step": 11606
     },
     {
-      "epoch": 0.31877729257641924,
+      "epoch": 0.3293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.593627610486922e-05,
-      "loss": 0.9629,
+      "learning_rate": 1.5656826671685665e-05,
+      "loss": 0.9324,
       "step": 11607
     },
     {
-      "epoch": 0.31880475680426246,
+      "epoch": 0.3293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.593556025238376e-05,
-      "loss": 0.9136,
+      "learning_rate": 1.5656068754865388e-05,
+      "loss": 0.9984,
       "step": 11608
     },
     {
-      "epoch": 0.3188322210321057,
+      "epoch": 0.32942678774120315,
       "grad_norm": 0.0,
-      "learning_rate": 1.5934844352933847e-05,
-      "loss": 1.0585,
+      "learning_rate": 1.5655310790268045e-05,
+      "loss": 1.0369,
       "step": 11609
     },
     {
-      "epoch": 0.3188596852599489,
+      "epoch": 0.32945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.5934128406525146e-05,
-      "loss": 0.8413,
+      "learning_rate": 1.5654552777900032e-05,
+      "loss": 0.8265,
       "step": 11610
     },
     {
-      "epoch": 0.3188871494877922,
+      "epoch": 0.329483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.593341241316332e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.565379471776776e-05,
+      "loss": 1.0623,
       "step": 11611
     },
     {
-      "epoch": 0.3189146137156354,
+      "epoch": 0.32951191827468784,
       "grad_norm": 0.0,
-      "learning_rate": 1.593269637285404e-05,
-      "loss": 1.0056,
+      "learning_rate": 1.5653036609877633e-05,
+      "loss": 0.8454,
       "step": 11612
     },
     {
-      "epoch": 0.3189420779434786,
+      "epoch": 0.32954029511918276,
       "grad_norm": 0.0,
-      "learning_rate": 1.5931980285602967e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.5652278454236047e-05,
+      "loss": 1.0547,
       "step": 11613
     },
     {
-      "epoch": 0.31896954217132184,
+      "epoch": 0.3295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.5931264151415765e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.5651520250849413e-05,
+      "loss": 1.0972,
       "step": 11614
     },
     {
-      "epoch": 0.31899700639916506,
+      "epoch": 0.32959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.593054797029811e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.5650761999724134e-05,
+      "loss": 0.872,
       "step": 11615
     },
     {
-      "epoch": 0.31902447062700834,
+      "epoch": 0.32962542565266745,
       "grad_norm": 0.0,
-      "learning_rate": 1.5929831742255655e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.5650003700866613e-05,
+      "loss": 1.0136,
       "step": 11616
     },
     {
-      "epoch": 0.31905193485485156,
+      "epoch": 0.3296538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.5929115467294078e-05,
-      "loss": 1.0337,
+      "learning_rate": 1.564924535428326e-05,
+      "loss": 0.9646,
       "step": 11617
     },
     {
-      "epoch": 0.3190793990826948,
+      "epoch": 0.3296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.5928399145419045e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.5648486959980474e-05,
+      "loss": 1.0415,
       "step": 11618
     },
     {
-      "epoch": 0.319106863310538,
+      "epoch": 0.3297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.5927682776636216e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.5647728517964665e-05,
+      "loss": 1.0104,
       "step": 11619
     },
     {
-      "epoch": 0.3191343275383813,
+      "epoch": 0.329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.592696636095127e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.564697002824224e-05,
+      "loss": 1.0117,
       "step": 11620
     },
     {
-      "epoch": 0.3191617917662245,
+      "epoch": 0.3297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.5926249898369867e-05,
-      "loss": 0.9281,
+      "learning_rate": 1.5646211490819604e-05,
+      "loss": 1.0577,
       "step": 11621
     },
     {
-      "epoch": 0.3191892559940677,
+      "epoch": 0.32979568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 1.592553338889768e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.564545290570317e-05,
+      "loss": 0.9843,
       "step": 11622
     },
     {
-      "epoch": 0.31921672022191094,
+      "epoch": 0.3298240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.5924816832540377e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.5644694272899338e-05,
+      "loss": 0.915,
       "step": 11623
     },
     {
-      "epoch": 0.3192441844497542,
+      "epoch": 0.32985244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.5924100229303632e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.5643935592414518e-05,
+      "loss": 1.0241,
       "step": 11624
     },
     {
-      "epoch": 0.31927164867759744,
+      "epoch": 0.32988081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.5923383579193107e-05,
-      "loss": 1.0799,
+      "learning_rate": 1.5643176864255123e-05,
+      "loss": 0.7844,
       "step": 11625
     },
     {
-      "epoch": 0.31929911290544066,
+      "epoch": 0.3299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.5922666882214477e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.564241808842756e-05,
+      "loss": 0.9129,
       "step": 11626
     },
     {
-      "epoch": 0.3193265771332839,
+      "epoch": 0.32993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.5921950138373414e-05,
-      "loss": 0.974,
+      "learning_rate": 1.5641659264938236e-05,
+      "loss": 0.9547,
       "step": 11627
     },
     {
-      "epoch": 0.3193540413611271,
+      "epoch": 0.32996594778660615,
       "grad_norm": 0.0,
-      "learning_rate": 1.5921233347675586e-05,
-      "loss": 0.935,
+      "learning_rate": 1.5640900393793564e-05,
+      "loss": 0.8848,
       "step": 11628
     },
     {
-      "epoch": 0.3193815055889704,
+      "epoch": 0.329994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.5920516510126668e-05,
-      "loss": 0.9782,
+      "learning_rate": 1.564014147499995e-05,
+      "loss": 0.9225,
       "step": 11629
     },
     {
-      "epoch": 0.3194089698168136,
+      "epoch": 0.3300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.591979962573233e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.563938250856381e-05,
+      "loss": 0.8575,
       "step": 11630
     },
     {
-      "epoch": 0.3194364340446568,
+      "epoch": 0.3300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.591908269449824e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.5638623494491552e-05,
+      "loss": 0.9419,
       "step": 11631
     },
     {
-      "epoch": 0.31946389827250005,
+      "epoch": 0.3300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.5918365716430076e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.563786443278959e-05,
+      "loss": 0.9366,
       "step": 11632
     },
     {
-      "epoch": 0.3194913625003433,
+      "epoch": 0.3301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.5917648691533513e-05,
-      "loss": 1.0488,
+      "learning_rate": 1.563710532346433e-05,
+      "loss": 0.9633,
       "step": 11633
     },
     {
-      "epoch": 0.31951882672818654,
+      "epoch": 0.3301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.591693161981422e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.5636346166522192e-05,
+      "loss": 0.9361,
       "step": 11634
     },
     {
-      "epoch": 0.31954629095602977,
+      "epoch": 0.3301645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.5916214501277873e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.563558696196958e-05,
+      "loss": 0.9008,
       "step": 11635
     },
     {
-      "epoch": 0.319573755183873,
+      "epoch": 0.33019296254256525,
       "grad_norm": 0.0,
-      "learning_rate": 1.5915497335930145e-05,
-      "loss": 1.024,
+      "learning_rate": 1.5634827709812914e-05,
+      "loss": 1.0271,
       "step": 11636
     },
     {
-      "epoch": 0.31960121941171626,
+      "epoch": 0.33022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.591478012377671e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.5634068410058607e-05,
+      "loss": 0.9229,
       "step": 11637
     },
     {
-      "epoch": 0.3196286836395595,
+      "epoch": 0.330249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.5914062864823245e-05,
-      "loss": 0.92,
+      "learning_rate": 1.5633309062713065e-05,
+      "loss": 0.957,
       "step": 11638
     },
     {
-      "epoch": 0.3196561478674027,
+      "epoch": 0.33027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.591334555907542e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.5632549667782714e-05,
+      "loss": 0.9113,
       "step": 11639
     },
     {
-      "epoch": 0.31968361209524593,
+      "epoch": 0.33030646992054485,
       "grad_norm": 0.0,
-      "learning_rate": 1.5912628206538915e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.563179022527396e-05,
+      "loss": 0.8516,
       "step": 11640
     },
     {
-      "epoch": 0.31971107632308915,
+      "epoch": 0.3303348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.5911910807219407e-05,
-      "loss": 0.973,
+      "learning_rate": 1.563103073519322e-05,
+      "loss": 0.9327,
       "step": 11641
     },
     {
-      "epoch": 0.3197385405509324,
+      "epoch": 0.33036322360953463,
       "grad_norm": 0.0,
-      "learning_rate": 1.5911193361122572e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.563027119754691e-05,
+      "loss": 0.9659,
       "step": 11642
     },
     {
-      "epoch": 0.31976600477877565,
+      "epoch": 0.3303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.591047586825408e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.562951161234145e-05,
+      "loss": 0.9917,
       "step": 11643
     },
     {
-      "epoch": 0.31979346900661887,
+      "epoch": 0.3304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.590975832861962e-05,
-      "loss": 0.8354,
+      "learning_rate": 1.562875197958325e-05,
+      "loss": 0.9935,
       "step": 11644
     },
     {
-      "epoch": 0.3198209332344621,
+      "epoch": 0.3304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.590904074222486e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.562799229927873e-05,
+      "loss": 0.9518,
       "step": 11645
     },
     {
-      "epoch": 0.31984839746230537,
+      "epoch": 0.3304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.590832310907548e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.5627232571434304e-05,
+      "loss": 1.0048,
       "step": 11646
     },
     {
-      "epoch": 0.3198758616901486,
+      "epoch": 0.3305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.5907605429177158e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.5626472796056394e-05,
+      "loss": 0.9467,
       "step": 11647
     },
     {
-      "epoch": 0.3199033259179918,
+      "epoch": 0.33053348467650395,
       "grad_norm": 0.0,
-      "learning_rate": 1.5906887702535577e-05,
-      "loss": 0.962,
+      "learning_rate": 1.562571297315141e-05,
+      "loss": 0.8772,
       "step": 11648
     },
     {
-      "epoch": 0.31993079014583503,
+      "epoch": 0.33056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.590616992915641e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.5624953102725783e-05,
+      "loss": 0.9012,
       "step": 11649
     },
     {
-      "epoch": 0.3199582543736783,
+      "epoch": 0.3305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.590545210904534e-05,
-      "loss": 0.956,
+      "learning_rate": 1.562419318478592e-05,
+      "loss": 0.9972,
       "step": 11650
     },
     {
-      "epoch": 0.31998571860152153,
+      "epoch": 0.33061861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 1.590473424220804e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.5623433219338244e-05,
+      "loss": 0.9224,
       "step": 11651
     },
     {
-      "epoch": 0.32001318282936475,
+      "epoch": 0.33064699205448356,
       "grad_norm": 0.0,
-      "learning_rate": 1.5904016328650206e-05,
-      "loss": 1.0685,
+      "learning_rate": 1.5622673206389178e-05,
+      "loss": 0.9378,
       "step": 11652
     },
     {
-      "epoch": 0.32004064705720797,
+      "epoch": 0.3306753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.5903298368377497e-05,
-      "loss": 0.9503,
+      "learning_rate": 1.5621913145945135e-05,
+      "loss": 0.9355,
       "step": 11653
     },
     {
-      "epoch": 0.3200681112850512,
+      "epoch": 0.33070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.590258036139561e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.562115303801254e-05,
+      "loss": 0.8332,
       "step": 11654
     },
     {
-      "epoch": 0.32009557551289447,
+      "epoch": 0.3307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.590186230771022e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.5620392882597816e-05,
+      "loss": 0.9008,
       "step": 11655
     },
     {
-      "epoch": 0.3201230397407377,
+      "epoch": 0.3307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.590114420732701e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.561963267970737e-05,
+      "loss": 0.9323,
       "step": 11656
     },
     {
-      "epoch": 0.3201505039685809,
+      "epoch": 0.330788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.590042606025166e-05,
-      "loss": 0.925,
+      "learning_rate": 1.5618872429347646e-05,
+      "loss": 0.9976,
       "step": 11657
     },
     {
-      "epoch": 0.32017796819642413,
+      "epoch": 0.3308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.5899707866489852e-05,
-      "loss": 0.8039,
+      "learning_rate": 1.5618112131525048e-05,
+      "loss": 0.9323,
       "step": 11658
     },
     {
-      "epoch": 0.3202054324242674,
+      "epoch": 0.3308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.5898989626047274e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.5617351786246007e-05,
+      "loss": 0.9527,
       "step": 11659
     },
     {
-      "epoch": 0.32023289665211063,
+      "epoch": 0.33087400681044266,
       "grad_norm": 0.0,
-      "learning_rate": 1.5898271338929606e-05,
-      "loss": 0.965,
+      "learning_rate": 1.5616591393516946e-05,
+      "loss": 1.0078,
       "step": 11660
     },
     {
-      "epoch": 0.32026036087995385,
+      "epoch": 0.33090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.5897553005142527e-05,
-      "loss": 0.978,
+      "learning_rate": 1.561583095334428e-05,
+      "loss": 0.8725,
       "step": 11661
     },
     {
-      "epoch": 0.3202878251077971,
+      "epoch": 0.3309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.589683462469172e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.5615070465734436e-05,
+      "loss": 0.946,
       "step": 11662
     },
     {
-      "epoch": 0.32031528933564035,
+      "epoch": 0.33095913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.5896116197582882e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.5614309930693844e-05,
+      "loss": 0.8888,
       "step": 11663
     },
     {
-      "epoch": 0.3203427535634836,
+      "epoch": 0.33098751418842226,
       "grad_norm": 0.0,
-      "learning_rate": 1.589539772382168e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.561354934822892e-05,
+      "loss": 0.9193,
       "step": 11664
     },
     {
-      "epoch": 0.3203702177913268,
+      "epoch": 0.3310158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.5894679203413815e-05,
-      "loss": 1.0434,
+      "learning_rate": 1.5612788718346092e-05,
+      "loss": 1.0215,
       "step": 11665
     },
     {
-      "epoch": 0.32039768201917,
+      "epoch": 0.33104426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.589396063636496e-05,
-      "loss": 0.9389,
+      "learning_rate": 1.561202804105179e-05,
+      "loss": 0.9757,
       "step": 11666
     },
     {
-      "epoch": 0.3204251462470133,
+      "epoch": 0.33107264472190695,
       "grad_norm": 0.0,
-      "learning_rate": 1.5893242022680805e-05,
-      "loss": 0.8364,
+      "learning_rate": 1.561126731635243e-05,
+      "loss": 0.9619,
       "step": 11667
     },
     {
-      "epoch": 0.3204526104748565,
+      "epoch": 0.3311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.589252336236704e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.5610506544254446e-05,
+      "loss": 0.9274,
       "step": 11668
     },
     {
-      "epoch": 0.32048007470269974,
+      "epoch": 0.3311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.5891804655429345e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.5609745724764264e-05,
+      "loss": 0.9883,
       "step": 11669
     },
     {
-      "epoch": 0.32050753893054296,
+      "epoch": 0.3311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.589108590187341e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.56089848578883e-05,
+      "loss": 0.8898,
       "step": 11670
     },
     {
-      "epoch": 0.3205350031583862,
+      "epoch": 0.3311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.5890367101704926e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.5608223943632993e-05,
+      "loss": 0.9084,
       "step": 11671
     },
     {
-      "epoch": 0.32056246738622945,
+      "epoch": 0.33121452894438136,
       "grad_norm": 0.0,
-      "learning_rate": 1.588964825492957e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.5607462982004763e-05,
+      "loss": 0.8961,
       "step": 11672
     },
     {
-      "epoch": 0.3205899316140727,
+      "epoch": 0.3312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.5888929361553038e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.5606701973010047e-05,
+      "loss": 1.0252,
       "step": 11673
     },
     {
-      "epoch": 0.3206173958419159,
+      "epoch": 0.3312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.5888210421581016e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.5605940916655266e-05,
+      "loss": 0.9186,
       "step": 11674
     },
     {
-      "epoch": 0.3206448600697591,
+      "epoch": 0.33129965947786605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5887491435019186e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.5605179812946844e-05,
+      "loss": 1.0703,
       "step": 11675
     },
     {
-      "epoch": 0.3206723242976024,
+      "epoch": 0.33132803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.588677240187325e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.560441866189122e-05,
+      "loss": 0.9765,
       "step": 11676
     },
     {
-      "epoch": 0.3206997885254456,
+      "epoch": 0.3313564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.5886053322148888e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.5603657463494823e-05,
+      "loss": 1.0729,
       "step": 11677
     },
     {
-      "epoch": 0.32072725275328884,
+      "epoch": 0.33138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.5885334195851793e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.5602896217764075e-05,
+      "loss": 0.9307,
       "step": 11678
     },
     {
-      "epoch": 0.32075471698113206,
+      "epoch": 0.33141316685584565,
       "grad_norm": 0.0,
-      "learning_rate": 1.588461502298765e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.560213492470541e-05,
+      "loss": 0.9486,
       "step": 11679
     },
     {
-      "epoch": 0.32078218120897534,
+      "epoch": 0.3314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.5883895803562158e-05,
-      "loss": 0.9245,
+      "learning_rate": 1.560137358432526e-05,
+      "loss": 0.9542,
       "step": 11680
     },
     {
-      "epoch": 0.32080964543681856,
+      "epoch": 0.33146992054483543,
       "grad_norm": 0.0,
-      "learning_rate": 1.5883176537581005e-05,
-      "loss": 0.931,
+      "learning_rate": 1.560061219663006e-05,
+      "loss": 1.0114,
       "step": 11681
     },
     {
-      "epoch": 0.3208371096646618,
+      "epoch": 0.3314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.5882457225049876e-05,
-      "loss": 0.9501,
+      "learning_rate": 1.5599850761626235e-05,
+      "loss": 0.9255,
       "step": 11682
     },
     {
-      "epoch": 0.320864573892505,
+      "epoch": 0.3315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.588173786597447e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.5599089279320215e-05,
+      "loss": 1.0724,
       "step": 11683
     },
     {
-      "epoch": 0.3208920381203482,
+      "epoch": 0.3315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.5881018460360474e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.5598327749718443e-05,
+      "loss": 0.9933,
       "step": 11684
     },
     {
-      "epoch": 0.3209195023481915,
+      "epoch": 0.331583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.588029900821358e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.559756617282734e-05,
+      "loss": 0.884,
       "step": 11685
     },
     {
-      "epoch": 0.3209469665760347,
+      "epoch": 0.3316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.5879579509539485e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.5596804548653347e-05,
+      "loss": 0.875,
       "step": 11686
     },
     {
-      "epoch": 0.32097443080387794,
+      "epoch": 0.33164018161180475,
       "grad_norm": 0.0,
-      "learning_rate": 1.587885996434388e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.559604287720289e-05,
+      "loss": 1.0083,
       "step": 11687
     },
     {
-      "epoch": 0.32100189503172116,
+      "epoch": 0.33166855845629967,
       "grad_norm": 0.0,
-      "learning_rate": 1.5878140372632456e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.559528115848241e-05,
+      "loss": 0.943,
       "step": 11688
     },
     {
-      "epoch": 0.32102935925956444,
+      "epoch": 0.3316969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.5877420734410907e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.5594519392498337e-05,
+      "loss": 0.9143,
       "step": 11689
     },
     {
-      "epoch": 0.32105682348740766,
+      "epoch": 0.33172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.5876701049684933e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.559375757925711e-05,
+      "loss": 0.8424,
       "step": 11690
     },
     {
-      "epoch": 0.3210842877152509,
+      "epoch": 0.33175368898978436,
       "grad_norm": 0.0,
-      "learning_rate": 1.587598131846022e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.559299571876516e-05,
+      "loss": 0.9317,
       "step": 11691
     },
     {
-      "epoch": 0.3211117519430941,
+      "epoch": 0.3317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.587526154074247e-05,
-      "loss": 1.0348,
+      "learning_rate": 1.5592233811028924e-05,
+      "loss": 0.9493,
       "step": 11692
     },
     {
-      "epoch": 0.3211392161709374,
+      "epoch": 0.33181044267877413,
       "grad_norm": 0.0,
-      "learning_rate": 1.587454171653737e-05,
-      "loss": 0.8272,
+      "learning_rate": 1.5591471856054836e-05,
+      "loss": 0.9796,
       "step": 11693
     },
     {
-      "epoch": 0.3211666803987806,
+      "epoch": 0.331838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.5873821845850623e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.5590709853849334e-05,
+      "loss": 0.941,
       "step": 11694
     },
     {
-      "epoch": 0.3211941446266238,
+      "epoch": 0.3318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.587310192868792e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.558994780441886e-05,
+      "loss": 0.9668,
       "step": 11695
     },
     {
-      "epoch": 0.32122160885446704,
+      "epoch": 0.3318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.5872381965054966e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.558918570776984e-05,
+      "loss": 0.9836,
       "step": 11696
     },
     {
-      "epoch": 0.32124907308231027,
+      "epoch": 0.3319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.5871661954957446e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.5588423563908716e-05,
+      "loss": 0.9044,
       "step": 11697
     },
     {
-      "epoch": 0.32127653731015354,
+      "epoch": 0.3319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.5870941898401063e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.5587661372841928e-05,
+      "loss": 0.8127,
       "step": 11698
     },
     {
-      "epoch": 0.32130400153799676,
+      "epoch": 0.33198070374574346,
       "grad_norm": 0.0,
-      "learning_rate": 1.5870221795391514e-05,
-      "loss": 0.8859,
+      "learning_rate": 1.5586899134575916e-05,
+      "loss": 0.9943,
       "step": 11699
     },
     {
-      "epoch": 0.32133146576584,
+      "epoch": 0.33200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.5869501645934495e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.5586136849117114e-05,
+      "loss": 0.9092,
       "step": 11700
     },
     {
-      "epoch": 0.3213589299936832,
+      "epoch": 0.3320374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.5868781450035707e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.5585374516471964e-05,
+      "loss": 0.9337,
       "step": 11701
     },
     {
-      "epoch": 0.3213863942215265,
+      "epoch": 0.33206583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 1.5868061207700845e-05,
-      "loss": 0.9828,
+      "learning_rate": 1.55846121366469e-05,
+      "loss": 0.827,
       "step": 11702
     },
     {
-      "epoch": 0.3214138584493697,
+      "epoch": 0.33209421112372306,
       "grad_norm": 0.0,
-      "learning_rate": 1.586734091893561e-05,
-      "loss": 0.939,
+      "learning_rate": 1.558384970964837e-05,
+      "loss": 0.943,
       "step": 11703
     },
     {
-      "epoch": 0.3214413226772129,
+      "epoch": 0.3321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.58666205837457e-05,
-      "loss": 0.9059,
+      "learning_rate": 1.558308723548281e-05,
+      "loss": 0.9853,
       "step": 11704
     },
     {
-      "epoch": 0.32146878690505615,
+      "epoch": 0.33215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.586590020213682e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.5582324714156663e-05,
+      "loss": 0.9569,
       "step": 11705
     },
     {
-      "epoch": 0.3214962511328994,
+      "epoch": 0.3321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.5865179774114664e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.5581562145676365e-05,
+      "loss": 0.9697,
       "step": 11706
     },
     {
-      "epoch": 0.32152371536074265,
+      "epoch": 0.3322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.586445929968493e-05,
-      "loss": 0.968,
+      "learning_rate": 1.5580799530048362e-05,
+      "loss": 0.9958,
       "step": 11707
     },
     {
-      "epoch": 0.32155117958858587,
+      "epoch": 0.3322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.5863738778853327e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.5580036867279096e-05,
+      "loss": 0.975,
       "step": 11708
     },
     {
-      "epoch": 0.3215786438164291,
+      "epoch": 0.3322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.5863018211625547e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.557927415737501e-05,
+      "loss": 0.8923,
       "step": 11709
     },
     {
-      "epoch": 0.3216061080442723,
+      "epoch": 0.3322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.5862297598007298e-05,
-      "loss": 1.0526,
+      "learning_rate": 1.5578511400342537e-05,
+      "loss": 0.9038,
       "step": 11710
     },
     {
-      "epoch": 0.3216335722721156,
+      "epoch": 0.33232122587968216,
       "grad_norm": 0.0,
-      "learning_rate": 1.586157693800428e-05,
-      "loss": 0.9011,
+      "learning_rate": 1.557774859618813e-05,
+      "loss": 0.8675,
       "step": 11711
     },
     {
-      "epoch": 0.3216610364999588,
+      "epoch": 0.3323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.586085623162219e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.5576985744918232e-05,
+      "loss": 0.9778,
       "step": 11712
     },
     {
-      "epoch": 0.32168850072780203,
+      "epoch": 0.332377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.5860135478866743e-05,
-      "loss": 0.9046,
+      "learning_rate": 1.5576222846539285e-05,
+      "loss": 1.0151,
       "step": 11713
     },
     {
-      "epoch": 0.32171596495564525,
+      "epoch": 0.33240635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.5859414679743626e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.557545990105773e-05,
+      "loss": 1.0109,
       "step": 11714
     },
     {
-      "epoch": 0.3217434291834885,
+      "epoch": 0.33243473325766176,
       "grad_norm": 0.0,
-      "learning_rate": 1.5858693834258557e-05,
-      "loss": 1.0789,
+      "learning_rate": 1.5574696908480018e-05,
+      "loss": 0.9178,
       "step": 11715
     },
     {
-      "epoch": 0.32177089341133175,
+      "epoch": 0.3324631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.585797294241723e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.557393386881259e-05,
+      "loss": 0.9382,
       "step": 11716
     },
     {
-      "epoch": 0.32179835763917497,
+      "epoch": 0.33249148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 1.5857252004225353e-05,
-      "loss": 0.958,
+      "learning_rate": 1.5573170782061888e-05,
+      "loss": 1.1355,
       "step": 11717
     },
     {
-      "epoch": 0.3218258218670182,
+      "epoch": 0.3325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.5856531019688628e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.5572407648234365e-05,
+      "loss": 0.9235,
       "step": 11718
     },
     {
-      "epoch": 0.32185328609486147,
+      "epoch": 0.3325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.5855809988812764e-05,
-      "loss": 1.0049,
+      "learning_rate": 1.5571644467336465e-05,
+      "loss": 0.8749,
       "step": 11719
     },
     {
-      "epoch": 0.3218807503227047,
+      "epoch": 0.3325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.585508891160346e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.5570881239374633e-05,
+      "loss": 0.8278,
       "step": 11720
     },
     {
-      "epoch": 0.3219082145505479,
+      "epoch": 0.3326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.585436778806642e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.5570117964355317e-05,
+      "loss": 0.9069,
       "step": 11721
     },
     {
-      "epoch": 0.32193567877839113,
+      "epoch": 0.332633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5853646618207362e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.5569354642284963e-05,
+      "loss": 0.9421,
       "step": 11722
     },
     {
-      "epoch": 0.32196314300623435,
+      "epoch": 0.33266174801362086,
       "grad_norm": 0.0,
-      "learning_rate": 1.585292540203198e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.556859127317002e-05,
+      "loss": 1.039,
       "step": 11723
     },
     {
-      "epoch": 0.32199060723407763,
+      "epoch": 0.3326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.585220413954599e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.5567827857016936e-05,
+      "loss": 0.9115,
       "step": 11724
     },
     {
-      "epoch": 0.32201807146192085,
+      "epoch": 0.3327185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.585148283075509e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.556706439383216e-05,
+      "loss": 0.9756,
       "step": 11725
     },
     {
-      "epoch": 0.3220455356897641,
+      "epoch": 0.33274687854710555,
       "grad_norm": 0.0,
-      "learning_rate": 1.585076147566499e-05,
-      "loss": 0.927,
+      "learning_rate": 1.5566300883622142e-05,
+      "loss": 0.9137,
       "step": 11726
     },
     {
-      "epoch": 0.3220729999176073,
+      "epoch": 0.33277525539160047,
       "grad_norm": 0.0,
-      "learning_rate": 1.58500400742814e-05,
-      "loss": 0.941,
+      "learning_rate": 1.556553732639333e-05,
+      "loss": 0.9072,
       "step": 11727
     },
     {
-      "epoch": 0.32210046414545057,
+      "epoch": 0.3328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.584931862661003e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.5564773722152173e-05,
+      "loss": 0.9695,
       "step": 11728
     },
     {
-      "epoch": 0.3221279283732938,
+      "epoch": 0.33283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.5848597132656586e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.556401007090512e-05,
+      "loss": 0.9132,
       "step": 11729
     },
     {
-      "epoch": 0.322155392601137,
+      "epoch": 0.33286038592508516,
       "grad_norm": 0.0,
-      "learning_rate": 1.5847875592426772e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.556324637265863e-05,
+      "loss": 1.1664,
       "step": 11730
     },
     {
-      "epoch": 0.32218285682898024,
+      "epoch": 0.33288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.58471540059263e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.5562482627419144e-05,
+      "loss": 0.8641,
       "step": 11731
     },
     {
-      "epoch": 0.3222103210568235,
+      "epoch": 0.33291713961407493,
       "grad_norm": 0.0,
-      "learning_rate": 1.5846432373160882e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.5561718835193117e-05,
+      "loss": 1.0375,
       "step": 11732
     },
     {
-      "epoch": 0.32223778528466673,
+      "epoch": 0.3329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.584571069413623e-05,
-      "loss": 0.904,
+      "learning_rate": 1.5560954995987e-05,
+      "loss": 1.0115,
       "step": 11733
     },
     {
-      "epoch": 0.32226524951250995,
+      "epoch": 0.3329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.5844988968858042e-05,
-      "loss": 1.1237,
+      "learning_rate": 1.5560191109807246e-05,
+      "loss": 0.8712,
       "step": 11734
     },
     {
-      "epoch": 0.3222927137403532,
+      "epoch": 0.33300227014755956,
       "grad_norm": 0.0,
-      "learning_rate": 1.5844267197332042e-05,
-      "loss": 0.915,
+      "learning_rate": 1.5559427176660315e-05,
+      "loss": 0.9632,
       "step": 11735
     },
     {
-      "epoch": 0.3223201779681964,
+      "epoch": 0.3330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.584354537956394e-05,
-      "loss": 0.8423,
+      "learning_rate": 1.5558663196552646e-05,
+      "loss": 0.9519,
       "step": 11736
     },
     {
-      "epoch": 0.3223476421960397,
+      "epoch": 0.3330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.584282351555944e-05,
-      "loss": 0.9544,
+      "learning_rate": 1.55578991694907e-05,
+      "loss": 0.8748,
       "step": 11737
     },
     {
-      "epoch": 0.3223751064238829,
+      "epoch": 0.33308740068104425,
       "grad_norm": 0.0,
-      "learning_rate": 1.5842101605324257e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.555713509548093e-05,
+      "loss": 0.9514,
       "step": 11738
     },
     {
-      "epoch": 0.3224025706517261,
+      "epoch": 0.33311577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 1.5841379648864104e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.555637097452979e-05,
+      "loss": 0.9116,
       "step": 11739
     },
     {
-      "epoch": 0.32243003487956934,
+      "epoch": 0.33314415437003403,
       "grad_norm": 0.0,
-      "learning_rate": 1.5840657646184693e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.5555606806643733e-05,
+      "loss": 1.0097,
       "step": 11740
     },
     {
-      "epoch": 0.3224574991074126,
+      "epoch": 0.33317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.5839935597291733e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.555484259182922e-05,
+      "loss": 0.974,
       "step": 11741
     },
     {
-      "epoch": 0.32248496333525584,
+      "epoch": 0.33320090805902386,
       "grad_norm": 0.0,
-      "learning_rate": 1.5839213502190945e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.5554078330092697e-05,
+      "loss": 1.0047,
       "step": 11742
     },
     {
-      "epoch": 0.32251242756309906,
+      "epoch": 0.3332292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.5838491360888033e-05,
-      "loss": 0.9484,
+      "learning_rate": 1.5553314021440627e-05,
+      "loss": 0.9509,
       "step": 11743
     },
     {
-      "epoch": 0.3225398917909423,
+      "epoch": 0.33325766174801363,
       "grad_norm": 0.0,
-      "learning_rate": 1.5837769173388716e-05,
-      "loss": 0.8706,
+      "learning_rate": 1.5552549665879462e-05,
+      "loss": 0.9689,
       "step": 11744
     },
     {
-      "epoch": 0.32256735601878556,
+      "epoch": 0.3332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.583704693969871e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.5551785263415667e-05,
+      "loss": 0.8757,
       "step": 11745
     },
     {
-      "epoch": 0.3225948202466288,
+      "epoch": 0.3333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.583632465982373e-05,
-      "loss": 0.947,
+      "learning_rate": 1.5551020814055687e-05,
+      "loss": 0.9882,
       "step": 11746
     },
     {
-      "epoch": 0.322622284474472,
+      "epoch": 0.3333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.5835602333769484e-05,
-      "loss": 1.0742,
+      "learning_rate": 1.555025631780598e-05,
+      "loss": 0.8241,
       "step": 11747
     },
     {
-      "epoch": 0.3226497487023152,
+      "epoch": 0.3333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.5834879961541694e-05,
-      "loss": 0.9197,
+      "learning_rate": 1.5549491774673016e-05,
+      "loss": 0.965,
       "step": 11748
     },
     {
-      "epoch": 0.32267721293015844,
+      "epoch": 0.3333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.5834157543146075e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.5548727184663245e-05,
+      "loss": 0.9505,
       "step": 11749
     },
     {
-      "epoch": 0.3227046771580017,
+      "epoch": 0.33342792281498296,
       "grad_norm": 0.0,
-      "learning_rate": 1.5833435078588336e-05,
-      "loss": 0.9207,
+      "learning_rate": 1.5547962547783126e-05,
+      "loss": 1.0348,
       "step": 11750
     },
     {
-      "epoch": 0.32273214138584494,
+      "epoch": 0.3334562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.5832712567874206e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.5547197864039113e-05,
+      "loss": 0.8598,
       "step": 11751
     },
     {
-      "epoch": 0.32275960561368816,
+      "epoch": 0.33348467650397273,
       "grad_norm": 0.0,
-      "learning_rate": 1.583199001100939e-05,
-      "loss": 0.803,
+      "learning_rate": 1.5546433133437675e-05,
+      "loss": 0.9768,
       "step": 11752
     },
     {
-      "epoch": 0.3227870698415314,
+      "epoch": 0.33351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.583126740799961e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.5545668355985263e-05,
+      "loss": 0.9384,
       "step": 11753
     },
     {
-      "epoch": 0.32281453406937466,
+      "epoch": 0.33354143019296256,
       "grad_norm": 0.0,
-      "learning_rate": 1.5830544758850587e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.5544903531688343e-05,
+      "loss": 1.0133,
       "step": 11754
     },
     {
-      "epoch": 0.3228419982972179,
+      "epoch": 0.3335698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.5829822063568033e-05,
-      "loss": 1.0009,
+      "learning_rate": 1.5544138660553375e-05,
+      "loss": 0.9092,
       "step": 11755
     },
     {
-      "epoch": 0.3228694625250611,
+      "epoch": 0.33359818388195234,
       "grad_norm": 0.0,
-      "learning_rate": 1.582909932215767e-05,
-      "loss": 0.9283,
+      "learning_rate": 1.5543373742586816e-05,
+      "loss": 1.0234,
       "step": 11756
     },
     {
-      "epoch": 0.3228969267529043,
+      "epoch": 0.3336265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.5828376534625214e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.554260877779513e-05,
+      "loss": 0.9785,
       "step": 11757
     },
     {
-      "epoch": 0.3229243909807476,
+      "epoch": 0.3336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.5827653700976385e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.5541843766184783e-05,
+      "loss": 1.0247,
       "step": 11758
     },
     {
-      "epoch": 0.3229518552085908,
+      "epoch": 0.333683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.5826930821216902e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.554107870776223e-05,
+      "loss": 0.9893,
       "step": 11759
     },
     {
-      "epoch": 0.32297931943643404,
+      "epoch": 0.3337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.5826207895352487e-05,
-      "loss": 1.0533,
+      "learning_rate": 1.5540313602533932e-05,
+      "loss": 0.8763,
       "step": 11760
     },
     {
-      "epoch": 0.32300678366427726,
+      "epoch": 0.3337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.5825484923388857e-05,
-      "loss": 0.9351,
+      "learning_rate": 1.5539548450506362e-05,
+      "loss": 0.9671,
       "step": 11761
     },
     {
-      "epoch": 0.32303424789212054,
+      "epoch": 0.33376844494892166,
       "grad_norm": 0.0,
-      "learning_rate": 1.5824761905331733e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.5538783251685975e-05,
+      "loss": 0.8397,
       "step": 11762
     },
     {
-      "epoch": 0.32306171211996376,
+      "epoch": 0.3337968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.5824038841186838e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.5538018006079235e-05,
+      "loss": 0.8497,
       "step": 11763
     },
     {
-      "epoch": 0.323089176347807,
+      "epoch": 0.3338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.582331573095989e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.553725271369261e-05,
+      "loss": 0.9214,
       "step": 11764
     },
     {
-      "epoch": 0.3231166405756502,
+      "epoch": 0.33385357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.5822592574656616e-05,
-      "loss": 0.863,
+      "learning_rate": 1.5536487374532562e-05,
+      "loss": 0.8946,
       "step": 11765
     },
     {
-      "epoch": 0.3231441048034934,
+      "epoch": 0.33388195232690127,
       "grad_norm": 0.0,
-      "learning_rate": 1.582186937228273e-05,
-      "loss": 1.098,
+      "learning_rate": 1.5535721988605558e-05,
+      "loss": 0.991,
       "step": 11766
     },
     {
-      "epoch": 0.3231715690313367,
+      "epoch": 0.3339103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.582114612384396e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.5534956555918056e-05,
+      "loss": 0.9214,
       "step": 11767
     },
     {
-      "epoch": 0.3231990332591799,
+      "epoch": 0.33393870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 1.5820422829346024e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.553419107647653e-05,
+      "loss": 0.8542,
       "step": 11768
     },
     {
-      "epoch": 0.32322649748702315,
+      "epoch": 0.3339670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.5819699488794654e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.5533425550287447e-05,
+      "loss": 0.8455,
       "step": 11769
     },
     {
-      "epoch": 0.32325396171486637,
+      "epoch": 0.3339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.5818976102195565e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.5532659977357266e-05,
+      "loss": 1.0161,
       "step": 11770
     },
     {
-      "epoch": 0.32328142594270964,
+      "epoch": 0.33402383654937573,
       "grad_norm": 0.0,
-      "learning_rate": 1.581825266955448e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.553189435769246e-05,
+      "loss": 0.878,
       "step": 11771
     },
     {
-      "epoch": 0.32330889017055287,
+      "epoch": 0.3340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.581752919087713e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.553112869129949e-05,
+      "loss": 1.0506,
       "step": 11772
     },
     {
-      "epoch": 0.3233363543983961,
+      "epoch": 0.3340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.5816805666169232e-05,
-      "loss": 1.0625,
+      "learning_rate": 1.5530362978184827e-05,
+      "loss": 0.9628,
       "step": 11773
     },
     {
-      "epoch": 0.3233638186262393,
+      "epoch": 0.33410896708286036,
       "grad_norm": 0.0,
-      "learning_rate": 1.5816082095436517e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.5529597218354943e-05,
+      "loss": 1.0132,
       "step": 11774
     },
     {
-      "epoch": 0.3233912828540826,
+      "epoch": 0.3341373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.5815358478684705e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.55288314118163e-05,
+      "loss": 0.9111,
       "step": 11775
     },
     {
-      "epoch": 0.3234187470819258,
+      "epoch": 0.3341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.581463481591953e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.552806555857537e-05,
+      "loss": 0.9834,
       "step": 11776
     },
     {
-      "epoch": 0.323446211309769,
+      "epoch": 0.33419409761634505,
       "grad_norm": 0.0,
-      "learning_rate": 1.5813911107146704e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.5527299658638627e-05,
+      "loss": 0.9263,
       "step": 11777
     },
     {
-      "epoch": 0.32347367553761225,
+      "epoch": 0.33422247446083997,
       "grad_norm": 0.0,
-      "learning_rate": 1.5813187352371965e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.552653371201253e-05,
+      "loss": 1.0678,
       "step": 11778
     },
     {
-      "epoch": 0.32350113976545547,
+      "epoch": 0.33425085130533483,
       "grad_norm": 0.0,
-      "learning_rate": 1.5812463551601035e-05,
-      "loss": 0.965,
+      "learning_rate": 1.5525767718703552e-05,
+      "loss": 1.0248,
       "step": 11779
     },
     {
-      "epoch": 0.32352860399329875,
+      "epoch": 0.33427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.5811739704839642e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.5525001678718167e-05,
+      "loss": 0.9447,
       "step": 11780
     },
     {
-      "epoch": 0.32355606822114197,
+      "epoch": 0.33430760499432466,
       "grad_norm": 0.0,
-      "learning_rate": 1.5811015812093515e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.5524235592062845e-05,
+      "loss": 0.9326,
       "step": 11781
     },
     {
-      "epoch": 0.3235835324489852,
+      "epoch": 0.3343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.5810291873368376e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.552346945874406e-05,
+      "loss": 0.9082,
       "step": 11782
     },
     {
-      "epoch": 0.3236109966768284,
+      "epoch": 0.33436435868331443,
       "grad_norm": 0.0,
-      "learning_rate": 1.580956788866996e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.5522703278768278e-05,
+      "loss": 0.8873,
       "step": 11783
     },
     {
-      "epoch": 0.3236384609046717,
+      "epoch": 0.3343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.5808843858003994e-05,
-      "loss": 0.933,
+      "learning_rate": 1.5521937052141972e-05,
+      "loss": 1.0598,
       "step": 11784
     },
     {
-      "epoch": 0.3236659251325149,
+      "epoch": 0.3344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.5808119781376202e-05,
-      "loss": 0.9544,
+      "learning_rate": 1.552117077887162e-05,
+      "loss": 0.9309,
       "step": 11785
     },
     {
-      "epoch": 0.32369338936035813,
+      "epoch": 0.33444948921679907,
       "grad_norm": 0.0,
-      "learning_rate": 1.580739565879232e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.5520404458963683e-05,
+      "loss": 0.7731,
       "step": 11786
     },
     {
-      "epoch": 0.32372085358820135,
+      "epoch": 0.334477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.5806671490258072e-05,
-      "loss": 0.868,
+      "learning_rate": 1.5519638092424647e-05,
+      "loss": 1.0327,
       "step": 11787
     },
     {
-      "epoch": 0.32374831781604463,
+      "epoch": 0.3345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.580594727577919e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.5518871679260978e-05,
+      "loss": 0.8913,
       "step": 11788
     },
     {
-      "epoch": 0.32377578204388785,
+      "epoch": 0.33453461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 1.5805223015361402e-05,
-      "loss": 1.0,
+      "learning_rate": 1.5518105219479152e-05,
+      "loss": 0.977,
       "step": 11789
     },
     {
-      "epoch": 0.32380324627173107,
+      "epoch": 0.33456299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.5804498709010445e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.5517338713085642e-05,
+      "loss": 0.9596,
       "step": 11790
     },
     {
-      "epoch": 0.3238307104995743,
+      "epoch": 0.33459137343927353,
       "grad_norm": 0.0,
-      "learning_rate": 1.5803774356732044e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.5516572160086923e-05,
+      "loss": 0.8931,
       "step": 11791
     },
     {
-      "epoch": 0.3238581747274175,
+      "epoch": 0.33461975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.580304995853193e-05,
-      "loss": 1.045,
+      "learning_rate": 1.5515805560489475e-05,
+      "loss": 0.9304,
       "step": 11792
     },
     {
-      "epoch": 0.3238856389552608,
+      "epoch": 0.33464812712826336,
       "grad_norm": 0.0,
-      "learning_rate": 1.5802325514415836e-05,
-      "loss": 1.001,
+      "learning_rate": 1.551503891429977e-05,
+      "loss": 0.8849,
       "step": 11793
     },
     {
-      "epoch": 0.323913103183104,
+      "epoch": 0.3346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.5801601024389498e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.5514272221524278e-05,
+      "loss": 0.9069,
       "step": 11794
     },
     {
-      "epoch": 0.32394056741094723,
+      "epoch": 0.33470488081725314,
       "grad_norm": 0.0,
-      "learning_rate": 1.5800876488458646e-05,
-      "loss": 1.0202,
+      "learning_rate": 1.551350548216948e-05,
+      "loss": 0.8975,
       "step": 11795
     },
     {
-      "epoch": 0.32396803163879045,
+      "epoch": 0.334733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.580015190662901e-05,
-      "loss": 0.932,
+      "learning_rate": 1.5512738696241864e-05,
+      "loss": 0.8988,
       "step": 11796
     },
     {
-      "epoch": 0.32399549586663373,
+      "epoch": 0.3347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.5799427278906327e-05,
-      "loss": 0.8655,
+      "learning_rate": 1.5511971863747886e-05,
+      "loss": 0.9328,
       "step": 11797
     },
     {
-      "epoch": 0.32402296009447695,
+      "epoch": 0.33479001135073777,
       "grad_norm": 0.0,
-      "learning_rate": 1.5798702605296328e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.551120498469404e-05,
+      "loss": 0.9431,
       "step": 11798
     },
     {
-      "epoch": 0.3240504243223202,
+      "epoch": 0.3348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.5797977885804746e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.5510438059086793e-05,
+      "loss": 0.9168,
       "step": 11799
     },
     {
-      "epoch": 0.3240778885501634,
+      "epoch": 0.3348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.579725312043732e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.550967108693263e-05,
+      "loss": 1.0009,
       "step": 11800
     },
     {
-      "epoch": 0.3241053527780067,
+      "epoch": 0.33487514188422246,
       "grad_norm": 0.0,
-      "learning_rate": 1.5796528309199782e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.5508904068238025e-05,
+      "loss": 0.9903,
       "step": 11801
     },
     {
-      "epoch": 0.3241328170058499,
+      "epoch": 0.3349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.5795803452097863e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.5508137003009463e-05,
+      "loss": 0.9223,
       "step": 11802
     },
     {
-      "epoch": 0.3241602812336931,
+      "epoch": 0.33493189557321223,
       "grad_norm": 0.0,
-      "learning_rate": 1.5795078549137305e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.550736989125342e-05,
+      "loss": 0.9604,
       "step": 11803
     },
     {
-      "epoch": 0.32418774546153634,
+      "epoch": 0.33496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.579435360032384e-05,
-      "loss": 0.935,
+      "learning_rate": 1.5506602732976373e-05,
+      "loss": 0.8964,
       "step": 11804
     },
     {
-      "epoch": 0.32421520968937956,
+      "epoch": 0.33498864926220207,
       "grad_norm": 0.0,
-      "learning_rate": 1.5793628605663205e-05,
-      "loss": 1.0392,
+      "learning_rate": 1.5505835528184808e-05,
+      "loss": 0.888,
       "step": 11805
     },
     {
-      "epoch": 0.32424267391722283,
+      "epoch": 0.3350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.5792903565161135e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.5505068276885205e-05,
+      "loss": 0.9167,
       "step": 11806
     },
     {
-      "epoch": 0.32427013814506606,
+      "epoch": 0.33504540295119184,
       "grad_norm": 0.0,
-      "learning_rate": 1.5792178478823372e-05,
-      "loss": 0.945,
+      "learning_rate": 1.550430097908404e-05,
+      "loss": 0.9546,
       "step": 11807
     },
     {
-      "epoch": 0.3242976023729093,
+      "epoch": 0.3350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.5791453346655644e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.5503533634787794e-05,
+      "loss": 0.9162,
       "step": 11808
     },
     {
-      "epoch": 0.3243250666007525,
+      "epoch": 0.3351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.5790728168663695e-05,
-      "loss": 1.017,
+      "learning_rate": 1.5502766244002955e-05,
+      "loss": 0.8571,
       "step": 11809
     },
     {
-      "epoch": 0.3243525308285958,
+      "epoch": 0.33513053348467653,
       "grad_norm": 0.0,
-      "learning_rate": 1.5790002944853264e-05,
-      "loss": 0.948,
+      "learning_rate": 1.5501998806736e-05,
+      "loss": 1.0476,
       "step": 11810
     },
     {
-      "epoch": 0.324379995056439,
+      "epoch": 0.3351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5789277675230087e-05,
-      "loss": 0.999,
+      "learning_rate": 1.550123132299342e-05,
+      "loss": 0.9725,
       "step": 11811
     },
     {
-      "epoch": 0.3244074592842822,
+      "epoch": 0.3351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.57885523597999e-05,
-      "loss": 1.0329,
+      "learning_rate": 1.5500463792781687e-05,
+      "loss": 0.8823,
       "step": 11812
     },
     {
-      "epoch": 0.32443492351212544,
+      "epoch": 0.33521566401816116,
       "grad_norm": 0.0,
-      "learning_rate": 1.5787826998568445e-05,
-      "loss": 0.8154,
+      "learning_rate": 1.5499696216107293e-05,
+      "loss": 0.8929,
       "step": 11813
     },
     {
-      "epoch": 0.3244623877399687,
+      "epoch": 0.3352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.578710159154146e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.5498928592976712e-05,
+      "loss": 0.9221,
       "step": 11814
     },
     {
-      "epoch": 0.32448985196781194,
+      "epoch": 0.33527241770715094,
       "grad_norm": 0.0,
-      "learning_rate": 1.5786376138724687e-05,
-      "loss": 0.9041,
+      "learning_rate": 1.5498160923396438e-05,
+      "loss": 0.9136,
       "step": 11815
     },
     {
-      "epoch": 0.32451731619565516,
+      "epoch": 0.33530079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.5785650640123862e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.549739320737295e-05,
+      "loss": 1.0259,
       "step": 11816
     },
     {
-      "epoch": 0.3245447804234984,
+      "epoch": 0.33532917139614077,
       "grad_norm": 0.0,
-      "learning_rate": 1.5784925095744732e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.549662544491273e-05,
+      "loss": 0.8714,
       "step": 11817
     },
     {
-      "epoch": 0.3245722446513416,
+      "epoch": 0.33535754824063563,
       "grad_norm": 0.0,
-      "learning_rate": 1.5784199505593032e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.5495857636022275e-05,
+      "loss": 0.958,
       "step": 11818
     },
     {
-      "epoch": 0.3245997088791849,
+      "epoch": 0.33538592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.5783473869674506e-05,
-      "loss": 0.8736,
+      "learning_rate": 1.5495089780708062e-05,
+      "loss": 0.8987,
       "step": 11819
     },
     {
-      "epoch": 0.3246271731070281,
+      "epoch": 0.3354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.5782748187994892e-05,
-      "loss": 1.03,
+      "learning_rate": 1.5494321878976578e-05,
+      "loss": 0.9052,
       "step": 11820
     },
     {
-      "epoch": 0.3246546373348713,
+      "epoch": 0.3354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.5782022460559937e-05,
-      "loss": 1.0265,
+      "learning_rate": 1.549355393083431e-05,
+      "loss": 0.9947,
       "step": 11821
     },
     {
-      "epoch": 0.32468210156271454,
+      "epoch": 0.33547105561861523,
       "grad_norm": 0.0,
-      "learning_rate": 1.578129668737538e-05,
-      "loss": 0.994,
+      "learning_rate": 1.5492785936287744e-05,
+      "loss": 0.9201,
       "step": 11822
     },
     {
-      "epoch": 0.3247095657905578,
+      "epoch": 0.3354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.5780570868446964e-05,
-      "loss": 0.8897,
+      "learning_rate": 1.5492017895343366e-05,
+      "loss": 1.0039,
       "step": 11823
     },
     {
-      "epoch": 0.32473703001840104,
+      "epoch": 0.335527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.577984500378043e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.5491249808007672e-05,
+      "loss": 1.0356,
       "step": 11824
     },
     {
-      "epoch": 0.32476449424624426,
+      "epoch": 0.33555618615209987,
       "grad_norm": 0.0,
-      "learning_rate": 1.5779119093381525e-05,
-      "loss": 1.0374,
+      "learning_rate": 1.5490481674287142e-05,
+      "loss": 0.9113,
       "step": 11825
     },
     {
-      "epoch": 0.3247919584740875,
+      "epoch": 0.3355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.577839313725599e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.548971349418826e-05,
+      "loss": 1.0214,
       "step": 11826
     },
     {
-      "epoch": 0.32481942270193076,
+      "epoch": 0.3356129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.577766713540957e-05,
-      "loss": 0.923,
+      "learning_rate": 1.548894526771753e-05,
+      "loss": 1.0264,
       "step": 11827
     },
     {
-      "epoch": 0.324846886929774,
+      "epoch": 0.33564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.5776941087848013e-05,
-      "loss": 1.1179,
+      "learning_rate": 1.548817699488143e-05,
+      "loss": 0.9193,
       "step": 11828
     },
     {
-      "epoch": 0.3248743511576172,
+      "epoch": 0.33566969353007947,
       "grad_norm": 0.0,
-      "learning_rate": 1.5776214994577055e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.548740867568645e-05,
+      "loss": 0.9417,
       "step": 11829
     },
     {
-      "epoch": 0.3249018153854604,
+      "epoch": 0.33569807037457433,
       "grad_norm": 0.0,
-      "learning_rate": 1.577548885560245e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.5486640310139085e-05,
+      "loss": 0.9635,
       "step": 11830
     },
     {
-      "epoch": 0.32492927961330365,
+      "epoch": 0.33572644721906925,
       "grad_norm": 0.0,
-      "learning_rate": 1.5774762670929942e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.5485871898245824e-05,
+      "loss": 0.9469,
       "step": 11831
     },
     {
-      "epoch": 0.3249567438411469,
+      "epoch": 0.3357548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.5774036440565272e-05,
-      "loss": 0.9031,
+      "learning_rate": 1.548510344001315e-05,
+      "loss": 0.8982,
       "step": 11832
     },
     {
-      "epoch": 0.32498420806899014,
+      "epoch": 0.335783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.577331016451419e-05,
-      "loss": 1.0314,
+      "learning_rate": 1.5484334935447564e-05,
+      "loss": 0.8895,
       "step": 11833
     },
     {
-      "epoch": 0.32501167229683336,
+      "epoch": 0.33581157775255394,
       "grad_norm": 0.0,
-      "learning_rate": 1.577258384278244e-05,
-      "loss": 0.8897,
+      "learning_rate": 1.5483566384555556e-05,
+      "loss": 0.9773,
       "step": 11834
     },
     {
-      "epoch": 0.3250391365246766,
+      "epoch": 0.3358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.577185747537577e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.5482797787343615e-05,
+      "loss": 0.9113,
       "step": 11835
     },
     {
-      "epoch": 0.32506660075251986,
+      "epoch": 0.3358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.577113106229993e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.5482029143818237e-05,
+      "loss": 0.8174,
       "step": 11836
     },
     {
-      "epoch": 0.3250940649803631,
+      "epoch": 0.33589670828603857,
       "grad_norm": 0.0,
-      "learning_rate": 1.5770404603560663e-05,
-      "loss": 0.8715,
+      "learning_rate": 1.5481260453985906e-05,
+      "loss": 0.9516,
       "step": 11837
     },
     {
-      "epoch": 0.3251215292082063,
+      "epoch": 0.3359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.5769678099163725e-05,
-      "loss": 1.0456,
+      "learning_rate": 1.548049171785313e-05,
+      "loss": 0.9799,
       "step": 11838
     },
     {
-      "epoch": 0.3251489934360495,
+      "epoch": 0.3359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.5768951549114855e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.547972293542639e-05,
+      "loss": 0.9999,
       "step": 11839
     },
     {
-      "epoch": 0.3251764576638928,
+      "epoch": 0.33598183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 1.5768224953419806e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.547895410671218e-05,
+      "loss": 0.9123,
       "step": 11840
     },
     {
-      "epoch": 0.325203921891736,
+      "epoch": 0.3360102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.5767498312084325e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.5478185231717002e-05,
+      "loss": 1.0031,
       "step": 11841
     },
     {
-      "epoch": 0.32523138611957925,
+      "epoch": 0.33603859250851303,
       "grad_norm": 0.0,
-      "learning_rate": 1.5766771625114164e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.5477416310447346e-05,
+      "loss": 0.9034,
       "step": 11842
     },
     {
-      "epoch": 0.32525885034742247,
+      "epoch": 0.33606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.5766044892515074e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.5476647342909714e-05,
+      "loss": 0.9662,
       "step": 11843
     },
     {
-      "epoch": 0.3252863145752657,
+      "epoch": 0.33609534619750286,
       "grad_norm": 0.0,
-      "learning_rate": 1.5765318114292804e-05,
-      "loss": 0.973,
+      "learning_rate": 1.547587832911059e-05,
+      "loss": 0.9675,
       "step": 11844
     },
     {
-      "epoch": 0.32531377880310897,
+      "epoch": 0.3361237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.57645912904531e-05,
-      "loss": 1.0391,
+      "learning_rate": 1.5475109269056473e-05,
+      "loss": 0.984,
       "step": 11845
     },
     {
-      "epoch": 0.3253412430309522,
+      "epoch": 0.33615209988649264,
       "grad_norm": 0.0,
-      "learning_rate": 1.5763864421001722e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.547434016275387e-05,
+      "loss": 1.0226,
       "step": 11846
     },
     {
-      "epoch": 0.3253687072587954,
+      "epoch": 0.3361804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.5763137505944414e-05,
-      "loss": 0.8317,
+      "learning_rate": 1.5473571010209262e-05,
+      "loss": 0.89,
       "step": 11847
     },
     {
-      "epoch": 0.32539617148663863,
+      "epoch": 0.3362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.5762410545286927e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.5472801811429155e-05,
+      "loss": 1.0015,
       "step": 11848
     },
     {
-      "epoch": 0.3254236357144819,
+      "epoch": 0.3362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.576168353903502e-05,
-      "loss": 0.8928,
+      "learning_rate": 1.5472032566420047e-05,
+      "loss": 0.8943,
       "step": 11849
     },
     {
-      "epoch": 0.32545109994232513,
+      "epoch": 0.3362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.5760956487194435e-05,
-      "loss": 0.923,
+      "learning_rate": 1.547126327518844e-05,
+      "loss": 0.8719,
       "step": 11850
     },
     {
-      "epoch": 0.32547856417016835,
+      "epoch": 0.3362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.576022938977093e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.5470493937740816e-05,
+      "loss": 0.9885,
       "step": 11851
     },
     {
-      "epoch": 0.32550602839801157,
+      "epoch": 0.33632236095346196,
       "grad_norm": 0.0,
-      "learning_rate": 1.5759502246770265e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.5469724554083686e-05,
+      "loss": 1.0066,
       "step": 11852
     },
     {
-      "epoch": 0.32553349262585485,
+      "epoch": 0.3363507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.5758775058198184e-05,
-      "loss": 1.0339,
+      "learning_rate": 1.546895512422355e-05,
+      "loss": 1.0147,
       "step": 11853
     },
     {
-      "epoch": 0.32556095685369807,
+      "epoch": 0.33637911464245174,
       "grad_norm": 0.0,
-      "learning_rate": 1.5758047824060447e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.5468185648166902e-05,
+      "loss": 0.9429,
       "step": 11854
     },
     {
-      "epoch": 0.3255884210815413,
+      "epoch": 0.33640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.5757320544362804e-05,
-      "loss": 0.8452,
+      "learning_rate": 1.5467416125920244e-05,
+      "loss": 0.9221,
       "step": 11855
     },
     {
-      "epoch": 0.3256158853093845,
+      "epoch": 0.33643586833144157,
       "grad_norm": 0.0,
-      "learning_rate": 1.575659321911101e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.5466646557490077e-05,
+      "loss": 0.9955,
       "step": 11856
     },
     {
-      "epoch": 0.32564334953722773,
+      "epoch": 0.3364642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.575586584831082e-05,
-      "loss": 0.9378,
+      "learning_rate": 1.5465876942882902e-05,
+      "loss": 0.9793,
       "step": 11857
     },
     {
-      "epoch": 0.325670813765071,
+      "epoch": 0.33649262202043134,
       "grad_norm": 0.0,
-      "learning_rate": 1.575513843196799e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.546510728210522e-05,
+      "loss": 0.944,
       "step": 11858
     },
     {
-      "epoch": 0.32569827799291423,
+      "epoch": 0.3365209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.5754410970088274e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.5464337575163525e-05,
+      "loss": 0.8857,
       "step": 11859
     },
     {
-      "epoch": 0.32572574222075745,
+      "epoch": 0.3365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.575368346267743e-05,
-      "loss": 0.9941,
+      "learning_rate": 1.546356782206433e-05,
+      "loss": 0.9902,
       "step": 11860
     },
     {
-      "epoch": 0.3257532064486007,
+      "epoch": 0.33657775255391603,
       "grad_norm": 0.0,
-      "learning_rate": 1.5752955909741215e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.5462798022814133e-05,
+      "loss": 0.9751,
       "step": 11861
     },
     {
-      "epoch": 0.32578067067644395,
+      "epoch": 0.3366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.5752228311285382e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.546202817741943e-05,
+      "loss": 1.0139,
       "step": 11862
     },
     {
-      "epoch": 0.32580813490428717,
+      "epoch": 0.3366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.5751500667315697e-05,
-      "loss": 0.8959,
+      "learning_rate": 1.546125828588673e-05,
+      "loss": 0.9763,
       "step": 11863
     },
     {
-      "epoch": 0.3258355991321304,
+      "epoch": 0.33666288308740067,
       "grad_norm": 0.0,
-      "learning_rate": 1.5750772977837903e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.546048834822254e-05,
+      "loss": 1.0481,
       "step": 11864
     },
     {
-      "epoch": 0.3258630633599736,
+      "epoch": 0.3366912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.5750045242857766e-05,
-      "loss": 0.954,
+      "learning_rate": 1.5459718364433357e-05,
+      "loss": 0.9462,
       "step": 11865
     },
     {
-      "epoch": 0.3258905275878169,
+      "epoch": 0.33671963677639044,
       "grad_norm": 0.0,
-      "learning_rate": 1.574931746238105e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.545894833452569e-05,
+      "loss": 0.9354,
       "step": 11866
     },
     {
-      "epoch": 0.3259179918156601,
+      "epoch": 0.33674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.5748589636413502e-05,
-      "loss": 0.9655,
+      "learning_rate": 1.545817825850604e-05,
+      "loss": 0.851,
       "step": 11867
     },
     {
-      "epoch": 0.32594545604350333,
+      "epoch": 0.33677639046538027,
       "grad_norm": 0.0,
-      "learning_rate": 1.5747861764960885e-05,
-      "loss": 1.0492,
+      "learning_rate": 1.545740813638091e-05,
+      "loss": 1.0447,
       "step": 11868
     },
     {
-      "epoch": 0.32597292027134656,
+      "epoch": 0.33680476730987513,
       "grad_norm": 0.0,
-      "learning_rate": 1.5747133848028964e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.545663796815681e-05,
+      "loss": 1.01,
       "step": 11869
     },
     {
-      "epoch": 0.32600038449918983,
+      "epoch": 0.33683314415437005,
       "grad_norm": 0.0,
-      "learning_rate": 1.5746405885623493e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.5455867753840243e-05,
+      "loss": 0.9179,
       "step": 11870
     },
     {
-      "epoch": 0.32602784872703305,
+      "epoch": 0.3368615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.574567787775023e-05,
-      "loss": 1.0755,
+      "learning_rate": 1.5455097493437718e-05,
+      "loss": 0.9136,
       "step": 11871
     },
     {
-      "epoch": 0.3260553129548763,
+      "epoch": 0.3368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.5744949824414938e-05,
-      "loss": 0.9522,
+      "learning_rate": 1.545432718695574e-05,
+      "loss": 0.9434,
       "step": 11872
     },
     {
-      "epoch": 0.3260827771827195,
+      "epoch": 0.33691827468785474,
       "grad_norm": 0.0,
-      "learning_rate": 1.574422172562338e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.545355683440081e-05,
+      "loss": 0.9366,
       "step": 11873
     },
     {
-      "epoch": 0.3261102414105627,
+      "epoch": 0.3369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.574349358138131e-05,
-      "loss": 0.989,
+      "learning_rate": 1.5452786435779444e-05,
+      "loss": 0.8569,
       "step": 11874
     },
     {
-      "epoch": 0.326137705638406,
+      "epoch": 0.3369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.5742765391694498e-05,
-      "loss": 0.9913,
+      "learning_rate": 1.545201599109815e-05,
+      "loss": 0.9066,
       "step": 11875
     },
     {
-      "epoch": 0.3261651698662492,
+      "epoch": 0.33700340522133937,
       "grad_norm": 0.0,
-      "learning_rate": 1.57420371565687e-05,
-      "loss": 0.9121,
+      "learning_rate": 1.5451245500363424e-05,
+      "loss": 0.9722,
       "step": 11876
     },
     {
-      "epoch": 0.32619263409409244,
+      "epoch": 0.3370317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.5741308876009683e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.5450474963581787e-05,
+      "loss": 0.9255,
       "step": 11877
     },
     {
-      "epoch": 0.32622009832193566,
+      "epoch": 0.33706015891032914,
       "grad_norm": 0.0,
-      "learning_rate": 1.5740580550023204e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.5449704380759738e-05,
+      "loss": 0.9821,
       "step": 11878
     },
     {
-      "epoch": 0.32624756254977894,
+      "epoch": 0.33708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.5739852178615027e-05,
-      "loss": 1.0249,
+      "learning_rate": 1.5448933751903796e-05,
+      "loss": 1.0112,
       "step": 11879
     },
     {
-      "epoch": 0.32627502677762216,
+      "epoch": 0.337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.573912376179092e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.5448163077020463e-05,
+      "loss": 0.8719,
       "step": 11880
     },
     {
-      "epoch": 0.3263024910054654,
+      "epoch": 0.33714528944381383,
       "grad_norm": 0.0,
-      "learning_rate": 1.573839529955664e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.5447392356116255e-05,
+      "loss": 0.887,
       "step": 11881
     },
     {
-      "epoch": 0.3263299552333086,
+      "epoch": 0.33717366628830875,
       "grad_norm": 0.0,
-      "learning_rate": 1.573766679191795e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.5446621589197673e-05,
+      "loss": 0.9603,
       "step": 11882
     },
     {
-      "epoch": 0.3263574194611519,
+      "epoch": 0.3372020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.573693823888062e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.544585077627124e-05,
+      "loss": 0.9767,
       "step": 11883
     },
     {
-      "epoch": 0.3263848836889951,
+      "epoch": 0.3372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.573620964045042e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.5445079917343456e-05,
+      "loss": 0.9689,
       "step": 11884
     },
     {
-      "epoch": 0.3264123479168383,
+      "epoch": 0.33725879682179344,
       "grad_norm": 0.0,
-      "learning_rate": 1.5735480996633095e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.544430901242084e-05,
+      "loss": 0.9739,
       "step": 11885
     },
     {
-      "epoch": 0.32643981214468154,
+      "epoch": 0.3372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.573475230743443e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.54435380615099e-05,
+      "loss": 1.0322,
       "step": 11886
     },
     {
-      "epoch": 0.32646727637252476,
+      "epoch": 0.3373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.573402357286018e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.544276706461715e-05,
+      "loss": 0.9531,
       "step": 11887
     },
     {
-      "epoch": 0.32649474060036804,
+      "epoch": 0.3373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.5733294792916115e-05,
-      "loss": 0.9032,
+      "learning_rate": 1.54419960217491e-05,
+      "loss": 0.9935,
       "step": 11888
     },
     {
-      "epoch": 0.32652220482821126,
+      "epoch": 0.337372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.5732565967608003e-05,
-      "loss": 0.9038,
+      "learning_rate": 1.5441224932912265e-05,
+      "loss": 0.9661,
       "step": 11889
     },
     {
-      "epoch": 0.3265496690560545,
+      "epoch": 0.3374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.5731837096941608e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.544045379811316e-05,
+      "loss": 0.8973,
       "step": 11890
     },
     {
-      "epoch": 0.3265771332838977,
+      "epoch": 0.33742905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 1.5731108180922695e-05,
-      "loss": 1.0426,
+      "learning_rate": 1.5439682617358295e-05,
+      "loss": 0.871,
       "step": 11891
     },
     {
-      "epoch": 0.326604597511741,
+      "epoch": 0.3374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.5730379219557035e-05,
-      "loss": 1.0471,
+      "learning_rate": 1.5438911390654193e-05,
+      "loss": 1.0229,
       "step": 11892
     },
     {
-      "epoch": 0.3266320617395842,
+      "epoch": 0.33748581157775254,
       "grad_norm": 0.0,
-      "learning_rate": 1.5729650212850397e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.5438140118007353e-05,
+      "loss": 0.9491,
       "step": 11893
     },
     {
-      "epoch": 0.3266595259674274,
+      "epoch": 0.33751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.5728921160808546e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.5437368799424307e-05,
+      "loss": 1.0707,
       "step": 11894
     },
     {
-      "epoch": 0.32668699019527064,
+      "epoch": 0.3375425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.5728192063437253e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.5436597434911556e-05,
+      "loss": 0.8924,
       "step": 11895
     },
     {
-      "epoch": 0.3267144544231139,
+      "epoch": 0.3375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.5727462920742284e-05,
-      "loss": 1.0129,
+      "learning_rate": 1.5435826024475623e-05,
+      "loss": 0.8298,
       "step": 11896
     },
     {
-      "epoch": 0.32674191865095714,
+      "epoch": 0.33759931895573214,
       "grad_norm": 0.0,
-      "learning_rate": 1.572673373272941e-05,
-      "loss": 1.0138,
+      "learning_rate": 1.5435054568123027e-05,
+      "loss": 1.0583,
       "step": 11897
     },
     {
-      "epoch": 0.32676938287880036,
+      "epoch": 0.337627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.5726004499404397e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.5434283065860277e-05,
+      "loss": 0.9035,
       "step": 11898
     },
     {
-      "epoch": 0.3267968471066436,
+      "epoch": 0.3376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.5725275220773024e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.5433511517693897e-05,
+      "loss": 0.9971,
       "step": 11899
     },
     {
-      "epoch": 0.3268243113344868,
+      "epoch": 0.3376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.572454589684105e-05,
-      "loss": 1.0199,
+      "learning_rate": 1.54327399236304e-05,
+      "loss": 0.9098,
       "step": 11900
     },
     {
-      "epoch": 0.3268517755623301,
+      "epoch": 0.3377128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.5723816527614255e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.54319682836763e-05,
+      "loss": 0.9179,
       "step": 11901
     },
     {
-      "epoch": 0.3268792397901733,
+      "epoch": 0.3377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.572308711309841e-05,
-      "loss": 1.027,
+      "learning_rate": 1.5431196597838126e-05,
+      "loss": 0.9598,
       "step": 11902
     },
     {
-      "epoch": 0.3269067040180165,
+      "epoch": 0.33776958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.5722357653299276e-05,
-      "loss": 1.0328,
+      "learning_rate": 1.5430424866122386e-05,
+      "loss": 0.8868,
       "step": 11903
     },
     {
-      "epoch": 0.32693416824585975,
+      "epoch": 0.3377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.5721628148222636e-05,
-      "loss": 1.0156,
+      "learning_rate": 1.5429653088535604e-05,
+      "loss": 0.9536,
       "step": 11904
     },
     {
-      "epoch": 0.326961632473703,
+      "epoch": 0.33782633371169124,
       "grad_norm": 0.0,
-      "learning_rate": 1.5720898597874254e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.5428881265084303e-05,
+      "loss": 1.0327,
       "step": 11905
     },
     {
-      "epoch": 0.32698909670154624,
+      "epoch": 0.33785471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.572016900225991e-05,
-      "loss": 0.8841,
+      "learning_rate": 1.5428109395774993e-05,
+      "loss": 1.0341,
       "step": 11906
     },
     {
-      "epoch": 0.32701656092938947,
+      "epoch": 0.33788308740068107,
       "grad_norm": 0.0,
-      "learning_rate": 1.571943936138537e-05,
-      "loss": 0.8783,
+      "learning_rate": 1.54273374806142e-05,
+      "loss": 1.0667,
       "step": 11907
     },
     {
-      "epoch": 0.3270440251572327,
+      "epoch": 0.33791146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 1.5718709675256414e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.542656551960844e-05,
+      "loss": 0.8965,
       "step": 11908
     },
     {
-      "epoch": 0.32707148938507596,
+      "epoch": 0.33793984108967084,
       "grad_norm": 0.0,
-      "learning_rate": 1.571797994387881e-05,
-      "loss": 0.9652,
+      "learning_rate": 1.542579351276424e-05,
+      "loss": 0.9319,
       "step": 11909
     },
     {
-      "epoch": 0.3270989536129192,
+      "epoch": 0.3379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.571725016725833e-05,
-      "loss": 1.035,
+      "learning_rate": 1.542502146008812e-05,
+      "loss": 0.9349,
       "step": 11910
     },
     {
-      "epoch": 0.3271264178407624,
+      "epoch": 0.3379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.5716520345400756e-05,
-      "loss": 0.8818,
+      "learning_rate": 1.5424249361586596e-05,
+      "loss": 0.9209,
       "step": 11911
     },
     {
-      "epoch": 0.32715388206860563,
+      "epoch": 0.3380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.571579047831186e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.54234772172662e-05,
+      "loss": 0.9437,
       "step": 11912
     },
     {
-      "epoch": 0.32718134629644885,
+      "epoch": 0.3380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.5715060565997412e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.542270502713344e-05,
+      "loss": 1.016,
       "step": 11913
     },
     {
-      "epoch": 0.3272088105242921,
+      "epoch": 0.3380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.5714330608463195e-05,
-      "loss": 0.929,
+      "learning_rate": 1.5421932791194855e-05,
+      "loss": 0.9245,
       "step": 11914
     },
     {
-      "epoch": 0.32723627475213535,
+      "epoch": 0.33811010215664017,
       "grad_norm": 0.0,
-      "learning_rate": 1.5713600605714976e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.5421160509456956e-05,
+      "loss": 1.0346,
       "step": 11915
     },
     {
-      "epoch": 0.32726373897997857,
+      "epoch": 0.3381384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.571287055775854e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.5420388181926268e-05,
+      "loss": 0.9869,
       "step": 11916
     },
     {
-      "epoch": 0.3272912032078218,
+      "epoch": 0.33816685584562994,
       "grad_norm": 0.0,
-      "learning_rate": 1.5712140464599656e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.5419615808609322e-05,
+      "loss": 0.9555,
       "step": 11917
     },
     {
-      "epoch": 0.32731866743566507,
+      "epoch": 0.33819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.5711410326244105e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.5418843389512637e-05,
+      "loss": 0.9577,
       "step": 11918
     },
     {
-      "epoch": 0.3273461316635083,
+      "epoch": 0.3382236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.5710680142697666e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.5418070924642736e-05,
+      "loss": 0.9608,
       "step": 11919
     },
     {
-      "epoch": 0.3273735958913515,
+      "epoch": 0.33825198637911463,
       "grad_norm": 0.0,
-      "learning_rate": 1.5709949913966108e-05,
-      "loss": 0.887,
+      "learning_rate": 1.5417298414006147e-05,
+      "loss": 0.9333,
       "step": 11920
     },
     {
-      "epoch": 0.32740106011919473,
+      "epoch": 0.33828036322360955,
       "grad_norm": 0.0,
-      "learning_rate": 1.570921964005522e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.5416525857609396e-05,
+      "loss": 0.8756,
       "step": 11921
     },
     {
-      "epoch": 0.327428524347038,
+      "epoch": 0.3383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.5708489320970772e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.5415753255459005e-05,
+      "loss": 0.9477,
       "step": 11922
     },
     {
-      "epoch": 0.32745598857488123,
+      "epoch": 0.3383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.5707758956718544e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.5414980607561506e-05,
+      "loss": 0.9323,
       "step": 11923
     },
     {
-      "epoch": 0.32748345280272445,
+      "epoch": 0.33836549375709424,
       "grad_norm": 0.0,
-      "learning_rate": 1.570702854730432e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.5414207913923422e-05,
+      "loss": 0.9353,
       "step": 11924
     },
     {
-      "epoch": 0.32751091703056767,
+      "epoch": 0.3383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.5706298092733872e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.5413435174551277e-05,
+      "loss": 0.8973,
       "step": 11925
     },
     {
-      "epoch": 0.3275383812584109,
+      "epoch": 0.338422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.570556759301298e-05,
-      "loss": 1.0928,
+      "learning_rate": 1.5412662389451605e-05,
+      "loss": 0.8525,
       "step": 11926
     },
     {
-      "epoch": 0.32756584548625417,
+      "epoch": 0.33845062429057887,
       "grad_norm": 0.0,
-      "learning_rate": 1.5704837048147433e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.5411889558630928e-05,
+      "loss": 1.0208,
       "step": 11927
     },
     {
-      "epoch": 0.3275933097140974,
+      "epoch": 0.3384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.5704106458143005e-05,
-      "loss": 0.8904,
+      "learning_rate": 1.541111668209578e-05,
+      "loss": 0.9828,
       "step": 11928
     },
     {
-      "epoch": 0.3276207739419406,
+      "epoch": 0.33850737797956865,
       "grad_norm": 0.0,
-      "learning_rate": 1.5703375823005476e-05,
-      "loss": 0.8927,
+      "learning_rate": 1.5410343759852682e-05,
+      "loss": 0.8694,
       "step": 11929
     },
     {
-      "epoch": 0.32764823816978383,
+      "epoch": 0.33853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.5702645142740628e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.5409570791908172e-05,
+      "loss": 0.9118,
       "step": 11930
     },
     {
-      "epoch": 0.3276757023976271,
+      "epoch": 0.3385641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.5701914417354242e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.5408797778268772e-05,
+      "loss": 0.9685,
       "step": 11931
     },
     {
-      "epoch": 0.32770316662547033,
+      "epoch": 0.33859250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.57011836468521e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.540802471894101e-05,
+      "loss": 1.0514,
       "step": 11932
     },
     {
-      "epoch": 0.32773063085331355,
+      "epoch": 0.33862088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.5700452831239986e-05,
-      "loss": 0.9478,
+      "learning_rate": 1.5407251613931425e-05,
+      "loss": 0.8604,
       "step": 11933
     },
     {
-      "epoch": 0.3277580950811568,
+      "epoch": 0.3386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.5699721970523682e-05,
-      "loss": 0.9145,
+      "learning_rate": 1.5406478463246543e-05,
+      "loss": 1.0204,
       "step": 11934
     },
     {
-      "epoch": 0.32778555930900005,
+      "epoch": 0.338677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.5698991064708965e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.540570526689289e-05,
+      "loss": 0.8762,
       "step": 11935
     },
     {
-      "epoch": 0.3278130235368433,
+      "epoch": 0.33870601589103294,
       "grad_norm": 0.0,
-      "learning_rate": 1.569826011380163e-05,
-      "loss": 0.9347,
+      "learning_rate": 1.5404932024877005e-05,
+      "loss": 0.8464,
       "step": 11936
     },
     {
-      "epoch": 0.3278404877646865,
+      "epoch": 0.3387343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.569752911780745e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.5404158737205418e-05,
+      "loss": 0.9158,
       "step": 11937
     },
     {
-      "epoch": 0.3278679519925297,
+      "epoch": 0.3387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.5696798076732213e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.5403385403884653e-05,
+      "loss": 1.0051,
       "step": 11938
     },
     {
-      "epoch": 0.32789541622037294,
+      "epoch": 0.3387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.56960669905817e-05,
-      "loss": 0.9996,
+      "learning_rate": 1.5402612024921254e-05,
+      "loss": 0.9783,
       "step": 11939
     },
     {
-      "epoch": 0.3279228804482162,
+      "epoch": 0.3388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.5695335859361706e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.5401838600321746e-05,
+      "loss": 0.9734,
       "step": 11940
     },
     {
-      "epoch": 0.32795034467605944,
+      "epoch": 0.3388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.5694604683078002e-05,
-      "loss": 1.0612,
+      "learning_rate": 1.5401065130092665e-05,
+      "loss": 1.0426,
       "step": 11941
     },
     {
-      "epoch": 0.32797780890390266,
+      "epoch": 0.33887627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.569387346173638e-05,
-      "loss": 0.958,
+      "learning_rate": 1.5400291614240542e-05,
+      "loss": 1.0607,
       "step": 11942
     },
     {
-      "epoch": 0.3280052731317459,
+      "epoch": 0.3389046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.569314219534263e-05,
-      "loss": 0.7393,
+      "learning_rate": 1.5399518052771918e-05,
+      "loss": 0.9435,
       "step": 11943
     },
     {
-      "epoch": 0.32803273735958915,
+      "epoch": 0.33893303064699204,
       "grad_norm": 0.0,
-      "learning_rate": 1.569241088390253e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.539874444569332e-05,
+      "loss": 0.8705,
       "step": 11944
     },
     {
-      "epoch": 0.3280602015874324,
+      "epoch": 0.33896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.5691679527421866e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.539797079301128e-05,
+      "loss": 0.8116,
       "step": 11945
     },
     {
-      "epoch": 0.3280876658152756,
+      "epoch": 0.3389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.569094812590644e-05,
-      "loss": 0.9051,
+      "learning_rate": 1.539719709473235e-05,
+      "loss": 0.9039,
       "step": 11946
     },
     {
-      "epoch": 0.3281151300431188,
+      "epoch": 0.33901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.5690216679362018e-05,
-      "loss": 0.972,
+      "learning_rate": 1.5396423350863043e-05,
+      "loss": 0.8349,
       "step": 11947
     },
     {
-      "epoch": 0.3281425942709621,
+      "epoch": 0.33904653802497164,
       "grad_norm": 0.0,
-      "learning_rate": 1.5689485187794403e-05,
-      "loss": 1.071,
+      "learning_rate": 1.5395649561409903e-05,
+      "loss": 0.9558,
       "step": 11948
     },
     {
-      "epoch": 0.3281700584988053,
+      "epoch": 0.3390749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.5688753651209375e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.5394875726379476e-05,
+      "loss": 0.9474,
       "step": 11949
     },
     {
-      "epoch": 0.32819752272664854,
+      "epoch": 0.3391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.5688022069612725e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.539410184577829e-05,
+      "loss": 0.9398,
       "step": 11950
     },
     {
-      "epoch": 0.32822498695449176,
+      "epoch": 0.3391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.568729044301024e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.539332791961288e-05,
+      "loss": 0.9745,
       "step": 11951
     },
     {
-      "epoch": 0.328252451182335,
+      "epoch": 0.3391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.5686558771407713e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.539255394788979e-05,
+      "loss": 0.9569,
       "step": 11952
     },
     {
-      "epoch": 0.32827991541017826,
+      "epoch": 0.3391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.5685827054810928e-05,
-      "loss": 0.9202,
+      "learning_rate": 1.539177993061556e-05,
+      "loss": 0.9897,
       "step": 11953
     },
     {
-      "epoch": 0.3283073796380215,
+      "epoch": 0.33921679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.5685095293225676e-05,
-      "loss": 0.8664,
+      "learning_rate": 1.5391005867796712e-05,
+      "loss": 0.9386,
       "step": 11954
     },
     {
-      "epoch": 0.3283348438658647,
+      "epoch": 0.3392451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.568436348665775e-05,
-      "loss": 0.8975,
+      "learning_rate": 1.5390231759439798e-05,
+      "loss": 0.9475,
       "step": 11955
     },
     {
-      "epoch": 0.3283623080937079,
+      "epoch": 0.33927355278093074,
       "grad_norm": 0.0,
-      "learning_rate": 1.5683631635112935e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.5389457605551353e-05,
+      "loss": 0.7526,
       "step": 11956
     },
     {
-      "epoch": 0.3283897723215512,
+      "epoch": 0.33930192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.568289973859703e-05,
-      "loss": 0.9954,
+      "learning_rate": 1.5388683406137923e-05,
+      "loss": 0.9416,
       "step": 11957
     },
     {
-      "epoch": 0.3284172365493944,
+      "epoch": 0.3393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.568216779711582e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.5387909161206037e-05,
+      "loss": 0.9241,
       "step": 11958
     },
     {
-      "epoch": 0.32844470077723764,
+      "epoch": 0.33935868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.5681435810675092e-05,
-      "loss": 0.9454,
+      "learning_rate": 1.5387134870762243e-05,
+      "loss": 0.8185,
       "step": 11959
     },
     {
-      "epoch": 0.32847216500508086,
+      "epoch": 0.33938706015891035,
       "grad_norm": 0.0,
-      "learning_rate": 1.5680703779280647e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.538636053481308e-05,
+      "loss": 1.0011,
       "step": 11960
     },
     {
-      "epoch": 0.32849962923292414,
+      "epoch": 0.3394154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.5679971702938273e-05,
-      "loss": 0.9758,
+      "learning_rate": 1.5385586153365082e-05,
+      "loss": 0.8423,
       "step": 11961
     },
     {
-      "epoch": 0.32852709346076736,
+      "epoch": 0.3394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.567923958165376e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.53848117264248e-05,
+      "loss": 0.9387,
       "step": 11962
     },
     {
-      "epoch": 0.3285545576886106,
+      "epoch": 0.339472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.5678507415432908e-05,
-      "loss": 0.9562,
+      "learning_rate": 1.5384037253998765e-05,
+      "loss": 1.0075,
       "step": 11963
     },
     {
-      "epoch": 0.3285820219164538,
+      "epoch": 0.3395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.5677775204281503e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.538326273609353e-05,
+      "loss": 0.9249,
       "step": 11964
     },
     {
-      "epoch": 0.3286094861442971,
+      "epoch": 0.3395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.567704294820534e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.5382488172715633e-05,
+      "loss": 0.9886,
       "step": 11965
     },
     {
-      "epoch": 0.3286369503721403,
+      "epoch": 0.33955732122587967,
       "grad_norm": 0.0,
-      "learning_rate": 1.5676310647210216e-05,
-      "loss": 1.032,
+      "learning_rate": 1.5381713563871616e-05,
+      "loss": 0.9076,
       "step": 11966
     },
     {
-      "epoch": 0.3286644145999835,
+      "epoch": 0.3395856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.5675578301301924e-05,
-      "loss": 0.9273,
+      "learning_rate": 1.5380938909568025e-05,
+      "loss": 0.9471,
       "step": 11967
     },
     {
-      "epoch": 0.32869187882782674,
+      "epoch": 0.33961407491486945,
       "grad_norm": 0.0,
-      "learning_rate": 1.567484591048626e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.5380164209811396e-05,
+      "loss": 1.0051,
       "step": 11968
     },
     {
-      "epoch": 0.32871934305566997,
+      "epoch": 0.33964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.567411347476901e-05,
-      "loss": 1.015,
+      "learning_rate": 1.5379389464608282e-05,
+      "loss": 1.005,
       "step": 11969
     },
     {
-      "epoch": 0.32874680728351324,
+      "epoch": 0.3396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.5673380994155984e-05,
-      "loss": 0.9021,
+      "learning_rate": 1.537861467396522e-05,
+      "loss": 0.8452,
       "step": 11970
     },
     {
-      "epoch": 0.32877427151135646,
+      "epoch": 0.33969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.5672648468652963e-05,
-      "loss": 0.9567,
+      "learning_rate": 1.5377839837888763e-05,
+      "loss": 0.9686,
       "step": 11971
     },
     {
-      "epoch": 0.3288017357391997,
+      "epoch": 0.33972758229284905,
       "grad_norm": 0.0,
-      "learning_rate": 1.5671915898265755e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.5377064956385447e-05,
+      "loss": 0.9901,
       "step": 11972
     },
     {
-      "epoch": 0.3288291999670429,
+      "epoch": 0.3397559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.5671183283000152e-05,
-      "loss": 0.9597,
+      "learning_rate": 1.537629002946182e-05,
+      "loss": 0.9125,
       "step": 11973
     },
     {
-      "epoch": 0.3288566641948862,
+      "epoch": 0.3397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.5670450622861942e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.5375515057124437e-05,
+      "loss": 0.898,
       "step": 11974
     },
     {
-      "epoch": 0.3288841284227294,
+      "epoch": 0.3398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.566971791785694e-05,
-      "loss": 0.855,
+      "learning_rate": 1.537474003937983e-05,
+      "loss": 0.9587,
       "step": 11975
     },
     {
-      "epoch": 0.3289115926505726,
+      "epoch": 0.3398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.5668985167990927e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.5373964976234556e-05,
+      "loss": 0.955,
       "step": 11976
     },
     {
-      "epoch": 0.32893905687841585,
+      "epoch": 0.3398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.566825237326971e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.5373189867695154e-05,
+      "loss": 0.9994,
       "step": 11977
     },
     {
-      "epoch": 0.3289665211062591,
+      "epoch": 0.3398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.5667519533699086e-05,
-      "loss": 0.8222,
+      "learning_rate": 1.5372414713768177e-05,
+      "loss": 0.8852,
       "step": 11978
     },
     {
-      "epoch": 0.32899398533410235,
+      "epoch": 0.3399262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.5666786649284848e-05,
-      "loss": 0.8558,
+      "learning_rate": 1.5371639514460172e-05,
+      "loss": 0.9267,
       "step": 11979
     },
     {
-      "epoch": 0.32902144956194557,
+      "epoch": 0.33995459704880815,
       "grad_norm": 0.0,
-      "learning_rate": 1.5666053720032802e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.5370864269777687e-05,
+      "loss": 1.0325,
       "step": 11980
     },
     {
-      "epoch": 0.3290489137897888,
+      "epoch": 0.33998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.5665320745948745e-05,
-      "loss": 0.8958,
+      "learning_rate": 1.537008897972727e-05,
+      "loss": 0.9104,
       "step": 11981
     },
     {
-      "epoch": 0.329076378017632,
+      "epoch": 0.340011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.5664587727038472e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.5369313644315474e-05,
+      "loss": 0.8774,
       "step": 11982
     },
     {
-      "epoch": 0.3291038422454753,
+      "epoch": 0.34003972758229284,
       "grad_norm": 0.0,
-      "learning_rate": 1.566385466330779e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.536853826354884e-05,
+      "loss": 1.0072,
       "step": 11983
     },
     {
-      "epoch": 0.3291313064733185,
+      "epoch": 0.34006810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 1.5663121554762494e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.536776283743392e-05,
+      "loss": 0.8294,
       "step": 11984
     },
     {
-      "epoch": 0.32915877070116173,
+      "epoch": 0.3400964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.5662388401408387e-05,
-      "loss": 0.901,
+      "learning_rate": 1.536698736597727e-05,
+      "loss": 0.956,
       "step": 11985
     },
     {
-      "epoch": 0.32918623492900495,
+      "epoch": 0.34012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.5661655203251268e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.5366211849185436e-05,
+      "loss": 0.9179,
       "step": 11986
     },
     {
-      "epoch": 0.3292136991568482,
+      "epoch": 0.34015323496027244,
       "grad_norm": 0.0,
-      "learning_rate": 1.5660921960296948e-05,
-      "loss": 1.007,
+      "learning_rate": 1.5365436287064967e-05,
+      "loss": 0.9654,
       "step": 11987
     },
     {
-      "epoch": 0.32924116338469145,
+      "epoch": 0.3401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.566018867255121e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.5364660679622417e-05,
+      "loss": 0.9854,
       "step": 11988
     },
     {
-      "epoch": 0.32926862761253467,
+      "epoch": 0.3402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.565945534001987e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.5363885026864337e-05,
+      "loss": 0.9992,
       "step": 11989
     },
     {
-      "epoch": 0.3292960918403779,
+      "epoch": 0.3402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.565872196270873e-05,
-      "loss": 0.8763,
+      "learning_rate": 1.5363109328797282e-05,
+      "loss": 1.0187,
       "step": 11990
     },
     {
-      "epoch": 0.32932355606822117,
+      "epoch": 0.340266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.565798854062359e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.5362333585427798e-05,
+      "loss": 0.9359,
       "step": 11991
     },
     {
-      "epoch": 0.3293510202960644,
+      "epoch": 0.34029511918274685,
       "grad_norm": 0.0,
-      "learning_rate": 1.565725507377025e-05,
-      "loss": 0.9583,
+      "learning_rate": 1.5361557796762444e-05,
+      "loss": 0.971,
       "step": 11992
     },
     {
-      "epoch": 0.3293784845239076,
+      "epoch": 0.34032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.565652156215452e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.536078196280777e-05,
+      "loss": 0.9081,
       "step": 11993
     },
     {
-      "epoch": 0.32940594875175083,
+      "epoch": 0.3403518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.5655788005782193e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.5360006083570326e-05,
+      "loss": 0.9414,
       "step": 11994
     },
     {
-      "epoch": 0.32943341297959405,
+      "epoch": 0.34038024971623154,
       "grad_norm": 0.0,
-      "learning_rate": 1.5655054404659087e-05,
-      "loss": 0.84,
+      "learning_rate": 1.535923015905667e-05,
+      "loss": 0.9408,
       "step": 11995
     },
     {
-      "epoch": 0.32946087720743733,
+      "epoch": 0.34040862656072646,
       "grad_norm": 0.0,
-      "learning_rate": 1.5654320758790998e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.535845418927336e-05,
+      "loss": 0.9582,
       "step": 11996
     },
     {
-      "epoch": 0.32948834143528055,
+      "epoch": 0.3404370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.565358706818373e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.5357678174226946e-05,
+      "loss": 0.9205,
       "step": 11997
     },
     {
-      "epoch": 0.3295158056631238,
+      "epoch": 0.34046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.5652853332843096e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.535690211392398e-05,
+      "loss": 0.9001,
       "step": 11998
     },
     {
-      "epoch": 0.329543269890967,
+      "epoch": 0.34049375709421115,
       "grad_norm": 0.0,
-      "learning_rate": 1.5652119552774895e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.5356126008371025e-05,
+      "loss": 1.0219,
       "step": 11999
     },
     {
-      "epoch": 0.32957073411881027,
+      "epoch": 0.340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.5651385727984936e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.535534985757463e-05,
+      "loss": 1.0245,
       "step": 12000
     },
     {
-      "epoch": 0.3295981983466535,
+      "epoch": 0.3405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.5650651858479024e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.5354573661541354e-05,
+      "loss": 0.8966,
       "step": 12001
     },
     {
-      "epoch": 0.3296256625744967,
+      "epoch": 0.3405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.564991794426296e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.5353797420277755e-05,
+      "loss": 0.925,
       "step": 12002
     },
     {
-      "epoch": 0.32965312680233994,
+      "epoch": 0.3406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.564918398534256e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.535302113379039e-05,
+      "loss": 1.0078,
       "step": 12003
     },
     {
-      "epoch": 0.3296805910301832,
+      "epoch": 0.3406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.564844998172363e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.535224480208581e-05,
+      "loss": 0.9706,
       "step": 12004
     },
     {
-      "epoch": 0.32970805525802643,
+      "epoch": 0.34066401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.5647715933411975e-05,
-      "loss": 0.8992,
+      "learning_rate": 1.5351468425170583e-05,
+      "loss": 0.976,
       "step": 12005
     },
     {
-      "epoch": 0.32973551948586965,
+      "epoch": 0.3406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.5646981840413404e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.535069200305126e-05,
+      "loss": 0.8744,
       "step": 12006
     },
     {
-      "epoch": 0.3297629837137129,
+      "epoch": 0.34072077185017025,
       "grad_norm": 0.0,
-      "learning_rate": 1.5646247702733727e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.5349915535734398e-05,
+      "loss": 0.791,
       "step": 12007
     },
     {
-      "epoch": 0.3297904479415561,
+      "epoch": 0.34074914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.5645513520378747e-05,
-      "loss": 1.0412,
+      "learning_rate": 1.5349139023226563e-05,
+      "loss": 0.9231,
       "step": 12008
     },
     {
-      "epoch": 0.3298179121693994,
+      "epoch": 0.34077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.5644779293354278e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.5348362465534305e-05,
+      "loss": 0.8931,
       "step": 12009
     },
     {
-      "epoch": 0.3298453763972426,
+      "epoch": 0.34080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.564404502166613e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.5347585862664192e-05,
+      "loss": 0.8911,
       "step": 12010
     },
     {
-      "epoch": 0.3298728406250858,
+      "epoch": 0.34083427922814985,
       "grad_norm": 0.0,
-      "learning_rate": 1.564331070532011e-05,
-      "loss": 0.9358,
+      "learning_rate": 1.5346809214622785e-05,
+      "loss": 0.8923,
       "step": 12011
     },
     {
-      "epoch": 0.32990030485292904,
+      "epoch": 0.3408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.5642576344322033e-05,
-      "loss": 0.9192,
+      "learning_rate": 1.5346032521416632e-05,
+      "loss": 0.8578,
       "step": 12012
     },
     {
-      "epoch": 0.3299277690807723,
+      "epoch": 0.3408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.5641841938677703e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.5345255783052307e-05,
+      "loss": 0.9222,
       "step": 12013
     },
     {
-      "epoch": 0.32995523330861554,
+      "epoch": 0.3409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.5641107488392937e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.5344478999536366e-05,
+      "loss": 0.9585,
       "step": 12014
     },
     {
-      "epoch": 0.32998269753645876,
+      "epoch": 0.3409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.564037299347354e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.534370217087537e-05,
+      "loss": 0.9979,
       "step": 12015
     },
     {
-      "epoch": 0.330010161764302,
+      "epoch": 0.3409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.5639638453925328e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.5342925297075885e-05,
+      "loss": 0.9107,
       "step": 12016
     },
     {
-      "epoch": 0.33003762599214526,
+      "epoch": 0.3410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.5638903869754113e-05,
-      "loss": 0.975,
+      "learning_rate": 1.5342148378144464e-05,
+      "loss": 0.9797,
       "step": 12017
     },
     {
-      "epoch": 0.3300650902199885,
+      "epoch": 0.3410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.5638169240965706e-05,
-      "loss": 1.0446,
+      "learning_rate": 1.534137141408768e-05,
+      "loss": 0.9932,
       "step": 12018
     },
     {
-      "epoch": 0.3300925544478317,
+      "epoch": 0.34106129398410895,
       "grad_norm": 0.0,
-      "learning_rate": 1.563743456756592e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.5340594404912087e-05,
+      "loss": 0.9357,
       "step": 12019
     },
     {
-      "epoch": 0.3301200186756749,
+      "epoch": 0.34108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.5636699849560568e-05,
-      "loss": 0.9654,
+      "learning_rate": 1.533981735062426e-05,
+      "loss": 1.1046,
       "step": 12020
     },
     {
-      "epoch": 0.33014748290351814,
+      "epoch": 0.3411180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.5635965086955462e-05,
-      "loss": 0.897,
+      "learning_rate": 1.533904025123075e-05,
+      "loss": 0.9664,
       "step": 12021
     },
     {
-      "epoch": 0.3301749471313614,
+      "epoch": 0.34114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.5635230279756417e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.5338263106738126e-05,
+      "loss": 0.7995,
       "step": 12022
     },
     {
-      "epoch": 0.33020241135920464,
+      "epoch": 0.34117480136208855,
       "grad_norm": 0.0,
-      "learning_rate": 1.563449542796925e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.5337485917152956e-05,
+      "loss": 0.9787,
       "step": 12023
     },
     {
-      "epoch": 0.33022987558704786,
+      "epoch": 0.3412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.5633760531599768e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.53367086824818e-05,
+      "loss": 1.1063,
       "step": 12024
     },
     {
-      "epoch": 0.3302573398148911,
+      "epoch": 0.34123155505107833,
       "grad_norm": 0.0,
-      "learning_rate": 1.5633025590653793e-05,
-      "loss": 0.8548,
+      "learning_rate": 1.5335931402731232e-05,
+      "loss": 0.9011,
       "step": 12025
     },
     {
-      "epoch": 0.33028480404273436,
+      "epoch": 0.3412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.5632290605137138e-05,
-      "loss": 0.9596,
+      "learning_rate": 1.5335154077907806e-05,
+      "loss": 0.9328,
       "step": 12026
     },
     {
-      "epoch": 0.3303122682705776,
+      "epoch": 0.3412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.563155557505562e-05,
-      "loss": 1.004,
+      "learning_rate": 1.5334376708018098e-05,
+      "loss": 0.8712,
       "step": 12027
     },
     {
-      "epoch": 0.3303397324984208,
+      "epoch": 0.341316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.563082050041505e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.5333599293068665e-05,
+      "loss": 1.0127,
       "step": 12028
     },
     {
-      "epoch": 0.330367196726264,
+      "epoch": 0.3413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.5630085381221246e-05,
-      "loss": 0.973,
+      "learning_rate": 1.5332821833066083e-05,
+      "loss": 1.0676,
       "step": 12029
     },
     {
-      "epoch": 0.3303946609541073,
+      "epoch": 0.3413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.5629350217480025e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.5332044328016916e-05,
+      "loss": 0.976,
       "step": 12030
     },
     {
-      "epoch": 0.3304221251819505,
+      "epoch": 0.34140181611804765,
       "grad_norm": 0.0,
-      "learning_rate": 1.5628615009197208e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.533126677792773e-05,
+      "loss": 0.9658,
       "step": 12031
     },
     {
-      "epoch": 0.33044958940979374,
+      "epoch": 0.34143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.562787975637861e-05,
-      "loss": 0.9057,
+      "learning_rate": 1.5330489182805087e-05,
+      "loss": 0.9143,
       "step": 12032
     },
     {
-      "epoch": 0.33047705363763696,
+      "epoch": 0.3414585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.5627144459030043e-05,
-      "loss": 1.0251,
+      "learning_rate": 1.532971154265557e-05,
+      "loss": 0.9508,
       "step": 12033
     },
     {
-      "epoch": 0.3305045178654802,
+      "epoch": 0.34148694665153234,
       "grad_norm": 0.0,
-      "learning_rate": 1.562640911715733e-05,
-      "loss": 1.0207,
+      "learning_rate": 1.5328933857485735e-05,
+      "loss": 0.9295,
       "step": 12034
     },
     {
-      "epoch": 0.33053198209332346,
+      "epoch": 0.34151532349602726,
       "grad_norm": 0.0,
-      "learning_rate": 1.5625673730766292e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.5328156127302164e-05,
+      "loss": 0.9628,
       "step": 12035
     },
     {
-      "epoch": 0.3305594463211667,
+      "epoch": 0.3415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.562493829986274e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.532737835211141e-05,
+      "loss": 0.927,
       "step": 12036
     },
     {
-      "epoch": 0.3305869105490099,
+      "epoch": 0.34157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.56242028244525e-05,
-      "loss": 1.0011,
+      "learning_rate": 1.532660053192006e-05,
+      "loss": 0.9144,
       "step": 12037
     },
     {
-      "epoch": 0.3306143747768531,
+      "epoch": 0.3416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.562346730454139e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.532582266673467e-05,
+      "loss": 0.9295,
       "step": 12038
     },
     {
-      "epoch": 0.3306418390046964,
+      "epoch": 0.3416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.562273174013522e-05,
-      "loss": 1.0186,
+      "learning_rate": 1.5325044756561816e-05,
+      "loss": 0.9703,
       "step": 12039
     },
     {
-      "epoch": 0.3306693032325396,
+      "epoch": 0.3416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.5621996131239827e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.532426680140807e-05,
+      "loss": 1.0531,
       "step": 12040
     },
     {
-      "epoch": 0.33069676746038285,
+      "epoch": 0.3416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.562126047786102e-05,
-      "loss": 1.0547,
+      "learning_rate": 1.5323488801280006e-05,
+      "loss": 0.8558,
       "step": 12041
     },
     {
-      "epoch": 0.33072423168822607,
+      "epoch": 0.3417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.5620524780004624e-05,
-      "loss": 1.0877,
+      "learning_rate": 1.532271075618419e-05,
+      "loss": 0.8641,
       "step": 12042
     },
     {
-      "epoch": 0.33075169591606934,
+      "epoch": 0.34174233825198636,
       "grad_norm": 0.0,
-      "learning_rate": 1.5619789037676458e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.53219326661272e-05,
+      "loss": 0.9608,
       "step": 12043
     },
     {
-      "epoch": 0.33077916014391257,
+      "epoch": 0.34177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.5619053250882345e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.5321154531115603e-05,
+      "loss": 0.9629,
       "step": 12044
     },
     {
-      "epoch": 0.3308066243717558,
+      "epoch": 0.3417990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.56183174196281e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.5320376351155975e-05,
+      "loss": 0.911,
       "step": 12045
     },
     {
-      "epoch": 0.330834088599599,
+      "epoch": 0.34182746878547104,
       "grad_norm": 0.0,
-      "learning_rate": 1.561758154391956e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.531959812625489e-05,
+      "loss": 0.914,
       "step": 12046
     },
     {
-      "epoch": 0.33086155282744223,
+      "epoch": 0.34185584562996596,
       "grad_norm": 0.0,
-      "learning_rate": 1.5616845623762536e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.5318819856418918e-05,
+      "loss": 0.929,
       "step": 12047
     },
     {
-      "epoch": 0.3308890170552855,
+      "epoch": 0.3418842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.561610965916285e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.531804154165464e-05,
+      "loss": 0.8919,
       "step": 12048
     },
     {
-      "epoch": 0.3309164812831287,
+      "epoch": 0.34191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.5615373650126334e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.531726318196862e-05,
+      "loss": 0.9711,
       "step": 12049
     },
     {
-      "epoch": 0.33094394551097195,
+      "epoch": 0.34194097616345065,
       "grad_norm": 0.0,
-      "learning_rate": 1.5614637596658803e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.5316484777367443e-05,
+      "loss": 0.8405,
       "step": 12050
     },
     {
-      "epoch": 0.33097140973881517,
+      "epoch": 0.3419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.5613901498766084e-05,
-      "loss": 0.9497,
+      "learning_rate": 1.5315706327857678e-05,
+      "loss": 0.9098,
       "step": 12051
     },
     {
-      "epoch": 0.33099887396665845,
+      "epoch": 0.3419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.5613165356454005e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.5314927833445904e-05,
+      "loss": 0.9437,
       "step": 12052
     },
     {
-      "epoch": 0.33102633819450167,
+      "epoch": 0.3420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.5612429169728387e-05,
-      "loss": 0.8856,
+      "learning_rate": 1.5314149294138693e-05,
+      "loss": 0.8776,
       "step": 12053
     },
     {
-      "epoch": 0.3310538024223449,
+      "epoch": 0.3420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.5611692938595054e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.531337070994263e-05,
+      "loss": 0.9392,
       "step": 12054
     },
     {
-      "epoch": 0.3310812666501881,
+      "epoch": 0.34208286038592506,
       "grad_norm": 0.0,
-      "learning_rate": 1.5610956663059834e-05,
-      "loss": 0.913,
+      "learning_rate": 1.531259208086428e-05,
+      "loss": 0.9629,
       "step": 12055
     },
     {
-      "epoch": 0.3311087308780314,
+      "epoch": 0.34211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.5610220343128547e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.5311813406910227e-05,
+      "loss": 0.9365,
       "step": 12056
     },
     {
-      "epoch": 0.3311361951058746,
+      "epoch": 0.3421396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.5609483978807025e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.5311034688087048e-05,
+      "loss": 1.0,
       "step": 12057
     },
     {
-      "epoch": 0.33116365933371783,
+      "epoch": 0.34216799091940975,
       "grad_norm": 0.0,
-      "learning_rate": 1.5608747570101096e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.531025592440132e-05,
+      "loss": 0.9291,
       "step": 12058
     },
     {
-      "epoch": 0.33119112356156105,
+      "epoch": 0.34219636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.5608011117016582e-05,
-      "loss": 0.922,
+      "learning_rate": 1.530947711585962e-05,
+      "loss": 1.0382,
       "step": 12059
     },
     {
-      "epoch": 0.33121858778940433,
+      "epoch": 0.3422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.5607274619559313e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.5308698262468533e-05,
+      "loss": 0.8628,
       "step": 12060
     },
     {
-      "epoch": 0.33124605201724755,
+      "epoch": 0.34225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.5606538077735115e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.530791936423463e-05,
+      "loss": 0.9105,
       "step": 12061
     },
     {
-      "epoch": 0.33127351624509077,
+      "epoch": 0.34228149829738935,
       "grad_norm": 0.0,
-      "learning_rate": 1.5605801491549813e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.5307140421164494e-05,
+      "loss": 0.9311,
       "step": 12062
     },
     {
-      "epoch": 0.331300980472934,
+      "epoch": 0.3423098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.5605064861009237e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.53063614332647e-05,
+      "loss": 0.9046,
       "step": 12063
     },
     {
-      "epoch": 0.3313284447007772,
+      "epoch": 0.3423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.5604328186119223e-05,
-      "loss": 0.8914,
+      "learning_rate": 1.530558240054184e-05,
+      "loss": 1.0538,
       "step": 12064
     },
     {
-      "epoch": 0.3313559089286205,
+      "epoch": 0.342366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.560359146688559e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.5304803323002483e-05,
+      "loss": 0.9279,
       "step": 12065
     },
     {
-      "epoch": 0.3313833731564637,
+      "epoch": 0.3423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.560285470331417e-05,
-      "loss": 0.8455,
+      "learning_rate": 1.5304024200653215e-05,
+      "loss": 0.945,
       "step": 12066
     },
     {
-      "epoch": 0.33141083738430693,
+      "epoch": 0.3424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.5602117895410794e-05,
-      "loss": 0.975,
+      "learning_rate": 1.5303245033500616e-05,
+      "loss": 0.9629,
       "step": 12067
     },
     {
-      "epoch": 0.33143830161215015,
+      "epoch": 0.3424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.5601381043181293e-05,
-      "loss": 0.879,
+      "learning_rate": 1.5302465821551268e-05,
+      "loss": 1.0608,
       "step": 12068
     },
     {
-      "epoch": 0.33146576583999343,
+      "epoch": 0.3424801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.560064414663149e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.5301686564811752e-05,
+      "loss": 0.9699,
       "step": 12069
     },
     {
-      "epoch": 0.33149323006783665,
+      "epoch": 0.34250851305334845,
       "grad_norm": 0.0,
-      "learning_rate": 1.559990720576723e-05,
-      "loss": 1.016,
+      "learning_rate": 1.530090726328865e-05,
+      "loss": 1.1001,
       "step": 12070
     },
     {
-      "epoch": 0.3315206942956799,
+      "epoch": 0.34253688989784337,
       "grad_norm": 0.0,
-      "learning_rate": 1.5599170220594328e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.5300127916988548e-05,
+      "loss": 0.905,
       "step": 12071
     },
     {
-      "epoch": 0.3315481585235231,
+      "epoch": 0.3425652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.5598433191118622e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.5299348525918026e-05,
+      "loss": 0.9852,
       "step": 12072
     },
     {
-      "epoch": 0.3315756227513664,
+      "epoch": 0.34259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.559769611734595e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.529856909008367e-05,
+      "loss": 0.8536,
       "step": 12073
     },
     {
-      "epoch": 0.3316030869792096,
+      "epoch": 0.34262202043132806,
       "grad_norm": 0.0,
-      "learning_rate": 1.559695899928213e-05,
-      "loss": 0.8542,
+      "learning_rate": 1.5297789609492062e-05,
+      "loss": 0.9484,
       "step": 12074
     },
     {
-      "epoch": 0.3316305512070528,
+      "epoch": 0.3426503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.5596221836933007e-05,
-      "loss": 0.8821,
+      "learning_rate": 1.529701008414979e-05,
+      "loss": 1.0428,
       "step": 12075
     },
     {
-      "epoch": 0.33165801543489604,
+      "epoch": 0.34267877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.559548463030441e-05,
-      "loss": 0.9217,
+      "learning_rate": 1.529623051406343e-05,
+      "loss": 0.8926,
       "step": 12076
     },
     {
-      "epoch": 0.33168547966273926,
+      "epoch": 0.3427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.5594747379402168e-05,
-      "loss": 1.127,
+      "learning_rate": 1.529545089923958e-05,
+      "loss": 0.9389,
       "step": 12077
     },
     {
-      "epoch": 0.33171294389058253,
+      "epoch": 0.3427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.5594010084232118e-05,
-      "loss": 0.94,
+      "learning_rate": 1.529467123968481e-05,
+      "loss": 0.9021,
       "step": 12078
     },
     {
-      "epoch": 0.33174040811842576,
+      "epoch": 0.3427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.5593272744800098e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.5293891535405716e-05,
+      "loss": 0.9971,
       "step": 12079
     },
     {
-      "epoch": 0.331767872346269,
+      "epoch": 0.3427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.559253536111193e-05,
-      "loss": 0.9033,
+      "learning_rate": 1.5293111786408886e-05,
+      "loss": 0.9471,
       "step": 12080
     },
     {
-      "epoch": 0.3317953365741122,
+      "epoch": 0.3428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.559179793317346e-05,
-      "loss": 1.0277,
+      "learning_rate": 1.52923319927009e-05,
+      "loss": 0.8985,
       "step": 12081
     },
     {
-      "epoch": 0.3318228008019555,
+      "epoch": 0.34284903518728715,
       "grad_norm": 0.0,
-      "learning_rate": 1.5591060460990517e-05,
-      "loss": 0.928,
+      "learning_rate": 1.529155215428835e-05,
+      "loss": 1.0167,
       "step": 12082
     },
     {
-      "epoch": 0.3318502650297987,
+      "epoch": 0.34287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.559032294456894e-05,
-      "loss": 0.9259,
+      "learning_rate": 1.5290772271177818e-05,
+      "loss": 0.9661,
       "step": 12083
     },
     {
-      "epoch": 0.3318777292576419,
+      "epoch": 0.342905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.558958538391456e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.5289992343375897e-05,
+      "loss": 0.9781,
       "step": 12084
     },
     {
-      "epoch": 0.33190519348548514,
+      "epoch": 0.34293416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.5588847779033218e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.5289212370889174e-05,
+      "loss": 0.8527,
       "step": 12085
     },
     {
-      "epoch": 0.3319326577133284,
+      "epoch": 0.34296254256526676,
       "grad_norm": 0.0,
-      "learning_rate": 1.5588110129930746e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.5288432353724233e-05,
+      "loss": 0.9002,
       "step": 12086
     },
     {
-      "epoch": 0.33196012194117164,
+      "epoch": 0.3429909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.558737243661298e-05,
-      "loss": 0.8909,
+      "learning_rate": 1.5287652291887667e-05,
+      "loss": 0.9285,
       "step": 12087
     },
     {
-      "epoch": 0.33198758616901486,
+      "epoch": 0.34301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.5586634699085762e-05,
-      "loss": 1.0309,
+      "learning_rate": 1.5286872185386067e-05,
+      "loss": 0.9393,
       "step": 12088
     },
     {
-      "epoch": 0.3320150503968581,
+      "epoch": 0.3430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.5585896917354925e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.5286092034226017e-05,
+      "loss": 0.9242,
       "step": 12089
     },
     {
-      "epoch": 0.3320425146247013,
+      "epoch": 0.3430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.558515909142631e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.5285311838414114e-05,
+      "loss": 0.9161,
       "step": 12090
     },
     {
-      "epoch": 0.3320699788525446,
+      "epoch": 0.3431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.5584421221305752e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.528453159795694e-05,
+      "loss": 0.8792,
       "step": 12091
     },
     {
-      "epoch": 0.3320974430803878,
+      "epoch": 0.3431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.558368330699909e-05,
-      "loss": 0.9766,
+      "learning_rate": 1.5283751312861093e-05,
+      "loss": 1.068,
       "step": 12092
     },
     {
-      "epoch": 0.332124907308231,
+      "epoch": 0.343161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.5582945348512163e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.5282970983133155e-05,
+      "loss": 0.985,
       "step": 12093
     },
     {
-      "epoch": 0.33215237153607424,
+      "epoch": 0.34318955732122586,
       "grad_norm": 0.0,
-      "learning_rate": 1.5582207345850807e-05,
-      "loss": 0.8147,
+      "learning_rate": 1.528219060877973e-05,
+      "loss": 0.9738,
       "step": 12094
     },
     {
-      "epoch": 0.3321798357639175,
+      "epoch": 0.3432179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.5581469299020864e-05,
-      "loss": 0.8448,
+      "learning_rate": 1.5281410189807405e-05,
+      "loss": 0.8889,
       "step": 12095
     },
     {
-      "epoch": 0.33220729999176074,
+      "epoch": 0.3432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.5580731208028178e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.5280629726222766e-05,
+      "loss": 0.9497,
       "step": 12096
     },
     {
-      "epoch": 0.33223476421960396,
+      "epoch": 0.34327468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.557999307287858e-05,
-      "loss": 0.8548,
+      "learning_rate": 1.527984921803241e-05,
+      "loss": 1.0056,
       "step": 12097
     },
     {
-      "epoch": 0.3322622284474472,
+      "epoch": 0.34330306469920546,
       "grad_norm": 0.0,
-      "learning_rate": 1.557925489357792e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.5279068665242936e-05,
+      "loss": 0.9317,
       "step": 12098
     },
     {
-      "epoch": 0.33228969267529046,
+      "epoch": 0.3433314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.557851667013203e-05,
-      "loss": 0.9501,
+      "learning_rate": 1.5278288067860925e-05,
+      "loss": 0.9431,
       "step": 12099
     },
     {
-      "epoch": 0.3323171569031337,
+      "epoch": 0.34335981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.557777840254676e-05,
-      "loss": 1.1099,
+      "learning_rate": 1.527750742589298e-05,
+      "loss": 0.9613,
       "step": 12100
     },
     {
-      "epoch": 0.3323446211309769,
+      "epoch": 0.34338819523269015,
       "grad_norm": 0.0,
-      "learning_rate": 1.557704009082794e-05,
-      "loss": 1.1087,
+      "learning_rate": 1.527672673934569e-05,
+      "loss": 1.0122,
       "step": 12101
     },
     {
-      "epoch": 0.3323720853588201,
+      "epoch": 0.343416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5576301734981423e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.527594600822566e-05,
+      "loss": 0.9746,
       "step": 12102
     },
     {
-      "epoch": 0.33239954958666335,
+      "epoch": 0.3434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.5575563335013046e-05,
-      "loss": 1.0151,
+      "learning_rate": 1.527516523253947e-05,
+      "loss": 0.9981,
       "step": 12103
     },
     {
-      "epoch": 0.3324270138145066,
+      "epoch": 0.3434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.5574824890928652e-05,
-      "loss": 1.0555,
+      "learning_rate": 1.5274384412293724e-05,
+      "loss": 0.8325,
       "step": 12104
     },
     {
-      "epoch": 0.33245447804234984,
+      "epoch": 0.3435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.5574086402734085e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.5273603547495016e-05,
+      "loss": 1.021,
       "step": 12105
     },
     {
-      "epoch": 0.33248194227019306,
+      "epoch": 0.34353007945516456,
       "grad_norm": 0.0,
-      "learning_rate": 1.5573347870435184e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.527282263814994e-05,
+      "loss": 1.1122,
       "step": 12106
     },
     {
-      "epoch": 0.3325094064980363,
+      "epoch": 0.3435584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.55726092940378e-05,
-      "loss": 0.913,
+      "learning_rate": 1.5272041684265095e-05,
+      "loss": 0.8884,
       "step": 12107
     },
     {
-      "epoch": 0.33253687072587956,
+      "epoch": 0.3435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.557187067354777e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.5271260685847078e-05,
+      "loss": 0.8708,
       "step": 12108
     },
     {
-      "epoch": 0.3325643349537228,
+      "epoch": 0.34361520998864925,
       "grad_norm": 0.0,
-      "learning_rate": 1.557113200897094e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.5270479642902484e-05,
+      "loss": 0.8896,
       "step": 12109
     },
     {
-      "epoch": 0.332591799181566,
+      "epoch": 0.34364358683314417,
       "grad_norm": 0.0,
-      "learning_rate": 1.5570393300313156e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.5269698555437913e-05,
+      "loss": 0.9862,
       "step": 12110
     },
     {
-      "epoch": 0.3326192634094092,
+      "epoch": 0.343671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.5569654547580266e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.5268917423459958e-05,
+      "loss": 0.937,
       "step": 12111
     },
     {
-      "epoch": 0.3326467276372525,
+      "epoch": 0.34370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.5568915750778112e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.5268136246975226e-05,
+      "loss": 0.9776,
       "step": 12112
     },
     {
-      "epoch": 0.3326741918650957,
+      "epoch": 0.34372871736662886,
       "grad_norm": 0.0,
-      "learning_rate": 1.5568176909912535e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.5267355025990304e-05,
+      "loss": 0.8703,
       "step": 12113
     },
     {
-      "epoch": 0.33270165609293895,
+      "epoch": 0.3437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.5567438024989386e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.52665737605118e-05,
+      "loss": 0.8737,
       "step": 12114
     },
     {
-      "epoch": 0.33272912032078217,
+      "epoch": 0.34378547105561863,
       "grad_norm": 0.0,
-      "learning_rate": 1.556669909601451e-05,
-      "loss": 0.9358,
+      "learning_rate": 1.526579245054632e-05,
+      "loss": 0.8988,
       "step": 12115
     },
     {
-      "epoch": 0.3327565845486254,
+      "epoch": 0.3438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.5565960122993758e-05,
-      "loss": 0.8439,
+      "learning_rate": 1.526501109610044e-05,
+      "loss": 0.8909,
       "step": 12116
     },
     {
-      "epoch": 0.33278404877646867,
+      "epoch": 0.3438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.556522110593297e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.5264229697180782e-05,
+      "loss": 1.0817,
       "step": 12117
     },
     {
-      "epoch": 0.3328115130043119,
+      "epoch": 0.34387060158910326,
       "grad_norm": 0.0,
-      "learning_rate": 1.5564482044837998e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.526344825379394e-05,
+      "loss": 0.8762,
       "step": 12118
     },
     {
-      "epoch": 0.3328389772321551,
+      "epoch": 0.3438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.556374293971469e-05,
-      "loss": 1.0368,
+      "learning_rate": 1.526266676594651e-05,
+      "loss": 1.0697,
       "step": 12119
     },
     {
-      "epoch": 0.33286644145999833,
+      "epoch": 0.3439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.5563003790568892e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.5261885233645097e-05,
+      "loss": 0.97,
       "step": 12120
     },
     {
-      "epoch": 0.3328939056878416,
+      "epoch": 0.34395573212258795,
       "grad_norm": 0.0,
-      "learning_rate": 1.5562264597406452e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.5261103656896304e-05,
+      "loss": 0.932,
       "step": 12121
     },
     {
-      "epoch": 0.33292136991568483,
+      "epoch": 0.34398410896708287,
       "grad_norm": 0.0,
-      "learning_rate": 1.556152536023322e-05,
-      "loss": 1.0235,
+      "learning_rate": 1.5260322035706735e-05,
+      "loss": 0.9359,
       "step": 12122
     },
     {
-      "epoch": 0.33294883414352805,
+      "epoch": 0.34401248581157773,
       "grad_norm": 0.0,
-      "learning_rate": 1.5560786079055045e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.5259540370082984e-05,
+      "loss": 0.8331,
       "step": 12123
     },
     {
-      "epoch": 0.33297629837137127,
+      "epoch": 0.34404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.556004675387778e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.5258758660031663e-05,
+      "loss": 0.9098,
       "step": 12124
     },
     {
-      "epoch": 0.33300376259921455,
+      "epoch": 0.34406923950056756,
       "grad_norm": 0.0,
-      "learning_rate": 1.5559307384707265e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.5257976905559368e-05,
+      "loss": 0.9742,
       "step": 12125
     },
     {
-      "epoch": 0.33303122682705777,
+      "epoch": 0.3440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.5558567971549362e-05,
-      "loss": 0.8161,
+      "learning_rate": 1.5257195106672709e-05,
+      "loss": 0.97,
       "step": 12126
     },
     {
-      "epoch": 0.333058691054901,
+      "epoch": 0.34412599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.5557828514409914e-05,
-      "loss": 1.0293,
+      "learning_rate": 1.5256413263378287e-05,
+      "loss": 0.963,
       "step": 12127
     },
     {
-      "epoch": 0.3330861552827442,
+      "epoch": 0.3441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.5557089013294772e-05,
-      "loss": 1.0614,
+      "learning_rate": 1.5255631375682702e-05,
+      "loss": 0.82,
       "step": 12128
     },
     {
-      "epoch": 0.33311361951058743,
+      "epoch": 0.3441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.5556349468209794e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.5254849443592568e-05,
+      "loss": 0.9248,
       "step": 12129
     },
     {
-      "epoch": 0.3331410837384307,
+      "epoch": 0.344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.555560987916082e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.5254067467114479e-05,
+      "loss": 0.9433,
       "step": 12130
     },
     {
-      "epoch": 0.33316854796627393,
+      "epoch": 0.3442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.5554870246153714e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.5253285446255048e-05,
+      "loss": 0.8888,
       "step": 12131
     },
     {
-      "epoch": 0.33319601219411715,
+      "epoch": 0.3442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.5554130569194322e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.525250338102088e-05,
+      "loss": 1.0079,
       "step": 12132
     },
     {
-      "epoch": 0.3332234764219604,
+      "epoch": 0.34429625425652666,
       "grad_norm": 0.0,
-      "learning_rate": 1.5553390848288496e-05,
-      "loss": 0.9489,
+      "learning_rate": 1.525172127141858e-05,
+      "loss": 0.8655,
       "step": 12133
     },
     {
-      "epoch": 0.33325094064980365,
+      "epoch": 0.34432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.5552651083442093e-05,
-      "loss": 0.8762,
+      "learning_rate": 1.525093911745475e-05,
+      "loss": 1.0115,
       "step": 12134
     },
     {
-      "epoch": 0.33327840487764687,
+      "epoch": 0.34435300794551643,
       "grad_norm": 0.0,
-      "learning_rate": 1.555191127466096e-05,
-      "loss": 0.978,
+      "learning_rate": 1.5250156919136005e-05,
+      "loss": 1.0022,
       "step": 12135
     },
     {
-      "epoch": 0.3333058691054901,
+      "epoch": 0.34438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.5551171421950955e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.524937467646895e-05,
+      "loss": 0.9941,
       "step": 12136
     },
     {
-      "epoch": 0.3333333333333333,
+      "epoch": 0.34440976163450626,
       "grad_norm": 0.0,
-      "learning_rate": 1.5550431525317932e-05,
-      "loss": 1.0744,
+      "learning_rate": 1.5248592389460189e-05,
+      "loss": 0.8901,
       "step": 12137
     },
     {
-      "epoch": 0.3333607975611766,
+      "epoch": 0.3444381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.554969158476775e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.524781005811633e-05,
+      "loss": 0.9062,
       "step": 12138
     },
     {
-      "epoch": 0.3333882617890198,
+      "epoch": 0.34446651532349604,
       "grad_norm": 0.0,
-      "learning_rate": 1.5548951600306252e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.5247027682443987e-05,
+      "loss": 0.9479,
       "step": 12139
     },
     {
-      "epoch": 0.33341572601686303,
+      "epoch": 0.3444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.55482115719393e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.5246245262449763e-05,
+      "loss": 0.9511,
       "step": 12140
     },
     {
-      "epoch": 0.33344319024470626,
+      "epoch": 0.3445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.554747149967275e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.5245462798140272e-05,
+      "loss": 0.9855,
       "step": 12141
     },
     {
-      "epoch": 0.3334706544725495,
+      "epoch": 0.3445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.5546731383512456e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.5244680289522118e-05,
+      "loss": 1.0461,
       "step": 12142
     },
     {
-      "epoch": 0.33349811870039275,
+      "epoch": 0.3445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.5545991223464275e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.5243897736601914e-05,
+      "loss": 0.9141,
       "step": 12143
     },
     {
-      "epoch": 0.333525582928236,
+      "epoch": 0.3446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.5545251019534065e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.5243115139386274e-05,
+      "loss": 0.9974,
       "step": 12144
     },
     {
-      "epoch": 0.3335530471560792,
+      "epoch": 0.34463677639046536,
       "grad_norm": 0.0,
-      "learning_rate": 1.5544510771727677e-05,
-      "loss": 1.0878,
+      "learning_rate": 1.5242332497881801e-05,
+      "loss": 0.9705,
       "step": 12145
     },
     {
-      "epoch": 0.3335805113839224,
+      "epoch": 0.3446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.5543770480050976e-05,
-      "loss": 1.008,
+      "learning_rate": 1.5241549812095114e-05,
+      "loss": 0.8673,
       "step": 12146
     },
     {
-      "epoch": 0.3336079756117657,
+      "epoch": 0.3446935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.5543030144509813e-05,
-      "loss": 0.8798,
+      "learning_rate": 1.524076708203281e-05,
+      "loss": 0.9666,
       "step": 12147
     },
     {
-      "epoch": 0.3336354398396089,
+      "epoch": 0.34472190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.554228976511005e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.523998430770152e-05,
+      "loss": 0.9247,
       "step": 12148
     },
     {
-      "epoch": 0.33366290406745214,
+      "epoch": 0.34475028376844497,
       "grad_norm": 0.0,
-      "learning_rate": 1.554154934185754e-05,
-      "loss": 1.0831,
+      "learning_rate": 1.5239201489107846e-05,
+      "loss": 0.9435,
       "step": 12149
     },
     {
-      "epoch": 0.33369036829529536,
+      "epoch": 0.3447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.5540808874758152e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.5238418626258398e-05,
+      "loss": 1.0093,
       "step": 12150
     },
     {
-      "epoch": 0.33371783252313864,
+      "epoch": 0.34480703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 1.554006836381773e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.5237635719159795e-05,
+      "loss": 0.9227,
       "step": 12151
     },
     {
-      "epoch": 0.33374529675098186,
+      "epoch": 0.3448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.553932780904215e-05,
-      "loss": 0.9996,
+      "learning_rate": 1.5236852767818651e-05,
+      "loss": 0.9675,
       "step": 12152
     },
     {
-      "epoch": 0.3337727609788251,
+      "epoch": 0.3448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.5538587210437253e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.523606977224157e-05,
+      "loss": 0.9304,
       "step": 12153
     },
     {
-      "epoch": 0.3338002252066683,
+      "epoch": 0.34489216799091943,
       "grad_norm": 0.0,
-      "learning_rate": 1.5537846568008915e-05,
-      "loss": 0.8773,
+      "learning_rate": 1.5235286732435174e-05,
+      "loss": 0.9055,
       "step": 12154
     },
     {
-      "epoch": 0.3338276894345116,
+      "epoch": 0.3449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.553710588176299e-05,
-      "loss": 0.952,
+      "learning_rate": 1.5234503648406075e-05,
+      "loss": 1.0191,
       "step": 12155
     },
     {
-      "epoch": 0.3338551536623548,
+      "epoch": 0.3449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.5536365151705336e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.523372052016089e-05,
+      "loss": 0.911,
       "step": 12156
     },
     {
-      "epoch": 0.333882617890198,
+      "epoch": 0.34497729852440406,
       "grad_norm": 0.0,
-      "learning_rate": 1.5535624377841814e-05,
-      "loss": 1.0075,
+      "learning_rate": 1.5232937347706234e-05,
+      "loss": 0.9194,
       "step": 12157
     },
     {
-      "epoch": 0.33391008211804124,
+      "epoch": 0.345005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.553488356017829e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.5232154131048718e-05,
+      "loss": 0.9057,
       "step": 12158
     },
     {
-      "epoch": 0.33393754634588446,
+      "epoch": 0.3450340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.553414269872062e-05,
-      "loss": 0.9607,
+      "learning_rate": 1.5231370870194964e-05,
+      "loss": 1.0247,
       "step": 12159
     },
     {
-      "epoch": 0.33396501057372774,
+      "epoch": 0.34506242905788875,
       "grad_norm": 0.0,
-      "learning_rate": 1.5533401793474677e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.5230587565151582e-05,
+      "loss": 0.9729,
       "step": 12160
     },
     {
-      "epoch": 0.33399247480157096,
+      "epoch": 0.34509080590238367,
       "grad_norm": 0.0,
-      "learning_rate": 1.5532660844446314e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.5229804215925192e-05,
+      "loss": 0.9457,
       "step": 12161
     },
     {
-      "epoch": 0.3340199390294142,
+      "epoch": 0.34511918274687853,
       "grad_norm": 0.0,
-      "learning_rate": 1.5531919851641387e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.5229020822522413e-05,
+      "loss": 0.9837,
       "step": 12162
     },
     {
-      "epoch": 0.3340474032572574,
+      "epoch": 0.34514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.5531178815065776e-05,
-      "loss": 1.029,
+      "learning_rate": 1.5228237384949858e-05,
+      "loss": 0.9478,
       "step": 12163
     },
     {
-      "epoch": 0.3340748674851007,
+      "epoch": 0.34517593643586836,
       "grad_norm": 0.0,
-      "learning_rate": 1.553043773472533e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.5227453903214147e-05,
+      "loss": 0.9029,
       "step": 12164
     },
     {
-      "epoch": 0.3341023317129439,
+      "epoch": 0.3452043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.5529696610625923e-05,
-      "loss": 0.962,
+      "learning_rate": 1.5226670377321898e-05,
+      "loss": 0.9267,
       "step": 12165
     },
     {
-      "epoch": 0.3341297959407871,
+      "epoch": 0.34523269012485813,
       "grad_norm": 0.0,
-      "learning_rate": 1.552895544277341e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.522588680727973e-05,
+      "loss": 0.9565,
       "step": 12166
     },
     {
-      "epoch": 0.33415726016863034,
+      "epoch": 0.345261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.5528214231173665e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.5225103193094261e-05,
+      "loss": 0.9581,
       "step": 12167
     },
     {
-      "epoch": 0.3341847243964736,
+      "epoch": 0.3452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.5527472975832543e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.5224319534772111e-05,
+      "loss": 0.9721,
       "step": 12168
     },
     {
-      "epoch": 0.33421218862431684,
+      "epoch": 0.34531782065834277,
       "grad_norm": 0.0,
-      "learning_rate": 1.5526731676755913e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.5223535832319898e-05,
+      "loss": 0.9273,
       "step": 12169
     },
     {
-      "epoch": 0.33423965285216006,
+      "epoch": 0.3453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.5525990333949644e-05,
-      "loss": 0.8233,
+      "learning_rate": 1.5222752085744244e-05,
+      "loss": 0.9799,
       "step": 12170
     },
     {
-      "epoch": 0.3342671170800033,
+      "epoch": 0.3453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.55252489474196e-05,
-      "loss": 0.9436,
+      "learning_rate": 1.5221968295051766e-05,
+      "loss": 0.9827,
       "step": 12171
     },
     {
-      "epoch": 0.3342945813078465,
+      "epoch": 0.34540295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.5524507517171642e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.522118446024909e-05,
+      "loss": 0.7911,
       "step": 12172
     },
     {
-      "epoch": 0.3343220455356898,
+      "epoch": 0.34543132803632237,
       "grad_norm": 0.0,
-      "learning_rate": 1.5523766043211646e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.5220400581342833e-05,
+      "loss": 0.834,
       "step": 12173
     },
     {
-      "epoch": 0.334349509763533,
+      "epoch": 0.34545970488081723,
       "grad_norm": 0.0,
-      "learning_rate": 1.552302452554547e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.5219616658339615e-05,
+      "loss": 0.9542,
       "step": 12174
     },
     {
-      "epoch": 0.3343769739913762,
+      "epoch": 0.34548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.5522282964178982e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.5218832691246066e-05,
+      "loss": 0.9548,
       "step": 12175
     },
     {
-      "epoch": 0.33440443821921945,
+      "epoch": 0.34551645856980706,
       "grad_norm": 0.0,
-      "learning_rate": 1.552154135911806e-05,
-      "loss": 1.0014,
+      "learning_rate": 1.52180486800688e-05,
+      "loss": 1.0161,
       "step": 12176
     },
     {
-      "epoch": 0.3344319024470627,
+      "epoch": 0.3455448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.5520799710368555e-05,
-      "loss": 1.0197,
+      "learning_rate": 1.521726462481444e-05,
+      "loss": 1.0543,
       "step": 12177
     },
     {
-      "epoch": 0.33445936667490594,
+      "epoch": 0.34557321225879684,
       "grad_norm": 0.0,
-      "learning_rate": 1.552005801793635e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.5216480525489612e-05,
+      "loss": 0.9133,
       "step": 12178
     },
     {
-      "epoch": 0.33448683090274917,
+      "epoch": 0.3456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.5519316281827302e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.5215696382100941e-05,
+      "loss": 0.998,
       "step": 12179
     },
     {
-      "epoch": 0.3345142951305924,
+      "epoch": 0.3456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.5518574502047293e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.5214912194655048e-05,
+      "loss": 0.9678,
       "step": 12180
     },
     {
-      "epoch": 0.33454175935843566,
+      "epoch": 0.3456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.5517832678602177e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.5214127963158553e-05,
+      "loss": 0.9132,
       "step": 12181
     },
     {
-      "epoch": 0.3345692235862789,
+      "epoch": 0.3456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.551709081149784e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.521334368761809e-05,
+      "loss": 0.9034,
       "step": 12182
     },
     {
-      "epoch": 0.3345966878141221,
+      "epoch": 0.3457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.5516348900740136e-05,
-      "loss": 1.0876,
+      "learning_rate": 1.5212559368040278e-05,
+      "loss": 0.9177,
       "step": 12183
     },
     {
-      "epoch": 0.33462415204196533,
+      "epoch": 0.34574347332576616,
       "grad_norm": 0.0,
-      "learning_rate": 1.5515606946334944e-05,
-      "loss": 0.7913,
+      "learning_rate": 1.521177500443174e-05,
+      "loss": 0.8323,
       "step": 12184
     },
     {
-      "epoch": 0.33465161626980855,
+      "epoch": 0.3457718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.5514864948288138e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.5210990596799103e-05,
+      "loss": 0.959,
       "step": 12185
     },
     {
-      "epoch": 0.3346790804976518,
+      "epoch": 0.34580022701475593,
       "grad_norm": 0.0,
-      "learning_rate": 1.551412290660558e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.5210206145148996e-05,
+      "loss": 0.8902,
       "step": 12186
     },
     {
-      "epoch": 0.33470654472549505,
+      "epoch": 0.34582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.5513380821293146e-05,
-      "loss": 0.9,
+      "learning_rate": 1.5209421649488045e-05,
+      "loss": 0.9676,
       "step": 12187
     },
     {
-      "epoch": 0.33473400895333827,
+      "epoch": 0.34585698070374576,
       "grad_norm": 0.0,
-      "learning_rate": 1.5512638692356706e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.5208637109822873e-05,
+      "loss": 0.8889,
       "step": 12188
     },
     {
-      "epoch": 0.3347614731811815,
+      "epoch": 0.3458853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.5511896519802136e-05,
-      "loss": 0.8705,
+      "learning_rate": 1.5207852526160112e-05,
+      "loss": 0.968,
       "step": 12189
     },
     {
-      "epoch": 0.33478893740902477,
+      "epoch": 0.34591373439273554,
       "grad_norm": 0.0,
-      "learning_rate": 1.5511154303635307e-05,
-      "loss": 0.9955,
+      "learning_rate": 1.5207067898506384e-05,
+      "loss": 1.045,
       "step": 12190
     },
     {
-      "epoch": 0.334816401636868,
+      "epoch": 0.3459421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.5510412043862086e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.520628322686832e-05,
+      "loss": 0.9187,
       "step": 12191
     },
     {
-      "epoch": 0.3348438658647112,
+      "epoch": 0.3459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.5509669740488352e-05,
-      "loss": 0.8888,
+      "learning_rate": 1.5205498511252548e-05,
+      "loss": 1.0996,
       "step": 12192
     },
     {
-      "epoch": 0.33487133009255443,
+      "epoch": 0.34599886492622023,
       "grad_norm": 0.0,
-      "learning_rate": 1.5508927393519978e-05,
-      "loss": 0.8081,
+      "learning_rate": 1.5204713751665697e-05,
+      "loss": 1.0123,
       "step": 12193
     },
     {
-      "epoch": 0.3348987943203977,
+      "epoch": 0.3460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.5508185002962837e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.520392894811439e-05,
+      "loss": 0.9809,
       "step": 12194
     },
     {
-      "epoch": 0.33492625854824093,
+      "epoch": 0.34605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.5507442568822802e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.5203144100605267e-05,
+      "loss": 1.1122,
       "step": 12195
     },
     {
-      "epoch": 0.33495372277608415,
+      "epoch": 0.34608399545970486,
       "grad_norm": 0.0,
-      "learning_rate": 1.5506700091105748e-05,
-      "loss": 0.9817,
+      "learning_rate": 1.520235920914495e-05,
+      "loss": 0.9914,
       "step": 12196
     },
     {
-      "epoch": 0.33498118700392737,
+      "epoch": 0.3461123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.5505957569817545e-05,
-      "loss": 0.8816,
+      "learning_rate": 1.520157427374007e-05,
+      "loss": 0.9084,
       "step": 12197
     },
     {
-      "epoch": 0.3350086512317706,
+      "epoch": 0.34614074914869464,
       "grad_norm": 0.0,
-      "learning_rate": 1.550521500496408e-05,
-      "loss": 0.935,
+      "learning_rate": 1.520078929439726e-05,
+      "loss": 0.8293,
       "step": 12198
     },
     {
-      "epoch": 0.33503611545961387,
+      "epoch": 0.34616912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.5504472396551217e-05,
-      "loss": 1.0989,
+      "learning_rate": 1.520000427112315e-05,
+      "loss": 0.8868,
       "step": 12199
     },
     {
-      "epoch": 0.3350635796874571,
+      "epoch": 0.34619750283768447,
       "grad_norm": 0.0,
-      "learning_rate": 1.550372974458484e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.5199219203924367e-05,
+      "loss": 0.8848,
       "step": 12200
     },
     {
-      "epoch": 0.3350910439153003,
+      "epoch": 0.3462258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.550298704907082e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.5198434092807546e-05,
+      "loss": 0.9541,
       "step": 12201
     },
     {
-      "epoch": 0.33511850814314353,
+      "epoch": 0.34625425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.550224431001503e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.5197648937779319e-05,
+      "loss": 0.8717,
       "step": 12202
     },
     {
-      "epoch": 0.3351459723709868,
+      "epoch": 0.3462826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.5501501527423357e-05,
-      "loss": 0.8838,
+      "learning_rate": 1.5196863738846319e-05,
+      "loss": 0.9533,
       "step": 12203
     },
     {
-      "epoch": 0.33517343659883003,
+      "epoch": 0.346311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.550075870130167e-05,
-      "loss": 1.0665,
+      "learning_rate": 1.5196078496015174e-05,
+      "loss": 0.9615,
       "step": 12204
     },
     {
-      "epoch": 0.33520090082667325,
+      "epoch": 0.34633938706015893,
       "grad_norm": 0.0,
-      "learning_rate": 1.5500015831655852e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.5195293209292525e-05,
+      "loss": 0.92,
       "step": 12205
     },
     {
-      "epoch": 0.3352283650545165,
+      "epoch": 0.3463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.5499272918491777e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.5194507878684999e-05,
+      "loss": 1.0433,
       "step": 12206
     },
     {
-      "epoch": 0.33525582928235975,
+      "epoch": 0.3463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.5498529961815324e-05,
-      "loss": 1.0815,
+      "learning_rate": 1.5193722504199232e-05,
+      "loss": 1.0147,
       "step": 12207
     },
     {
-      "epoch": 0.335283293510203,
+      "epoch": 0.34642451759364357,
       "grad_norm": 0.0,
-      "learning_rate": 1.5497786961632372e-05,
-      "loss": 0.854,
+      "learning_rate": 1.5192937085841854e-05,
+      "loss": 1.0074,
       "step": 12208
     },
     {
-      "epoch": 0.3353107577380462,
+      "epoch": 0.3464528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.54970439179488e-05,
-      "loss": 0.986,
+      "learning_rate": 1.5192151623619506e-05,
+      "loss": 0.9589,
       "step": 12209
     },
     {
-      "epoch": 0.3353382219658894,
+      "epoch": 0.3464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.549630083077049e-05,
-      "loss": 0.8457,
+      "learning_rate": 1.5191366117538822e-05,
+      "loss": 0.9019,
       "step": 12210
     },
     {
-      "epoch": 0.33536568619373264,
+      "epoch": 0.34650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.5495557700103316e-05,
-      "loss": 1.0657,
+      "learning_rate": 1.5190580567606433e-05,
+      "loss": 1.0348,
       "step": 12211
     },
     {
-      "epoch": 0.3353931504215759,
+      "epoch": 0.34653802497162317,
       "grad_norm": 0.0,
-      "learning_rate": 1.549481452595316e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.5189794973828975e-05,
+      "loss": 0.9561,
       "step": 12212
     },
     {
-      "epoch": 0.33542061464941914,
+      "epoch": 0.34656640181611803,
       "grad_norm": 0.0,
-      "learning_rate": 1.5494071308325904e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.5189009336213087e-05,
+      "loss": 0.9453,
       "step": 12213
     },
     {
-      "epoch": 0.33544807887726236,
+      "epoch": 0.34659477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.5493328047227426e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.5188223654765404e-05,
+      "loss": 0.921,
       "step": 12214
     },
     {
-      "epoch": 0.3354755431051056,
+      "epoch": 0.3466231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.549258474266361e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.5187437929492561e-05,
+      "loss": 0.841,
       "step": 12215
     },
     {
-      "epoch": 0.33550300733294885,
+      "epoch": 0.3466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.5491841394640336e-05,
-      "loss": 0.8878,
+      "learning_rate": 1.5186652160401196e-05,
+      "loss": 0.9457,
       "step": 12216
     },
     {
-      "epoch": 0.3355304715607921,
+      "epoch": 0.34667990919409764,
       "grad_norm": 0.0,
-      "learning_rate": 1.5491098003163488e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.518586634749795e-05,
+      "loss": 0.8877,
       "step": 12217
     },
     {
-      "epoch": 0.3355579357886353,
+      "epoch": 0.3467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.549035456823894e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.5185080490789457e-05,
+      "loss": 0.9928,
       "step": 12218
     },
     {
-      "epoch": 0.3355854000164785,
+      "epoch": 0.3467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.5489611089872583e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.5184294590282356e-05,
+      "loss": 0.963,
       "step": 12219
     },
     {
-      "epoch": 0.3356128642443218,
+      "epoch": 0.34676503972758227,
       "grad_norm": 0.0,
-      "learning_rate": 1.54888675680703e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.5183508645983285e-05,
+      "loss": 1.175,
       "step": 12220
     },
     {
-      "epoch": 0.335640328472165,
+      "epoch": 0.3467934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.5488124002837964e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.5182722657898882e-05,
+      "loss": 0.9609,
       "step": 12221
     },
     {
-      "epoch": 0.33566779270000824,
+      "epoch": 0.3468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.548738039418147e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.5181936626035791e-05,
+      "loss": 1.0167,
       "step": 12222
     },
     {
-      "epoch": 0.33569525692785146,
+      "epoch": 0.34685017026106696,
       "grad_norm": 0.0,
-      "learning_rate": 1.5486636742106697e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.5181150550400647e-05,
+      "loss": 0.9245,
       "step": 12223
     },
     {
-      "epoch": 0.3357227211556947,
+      "epoch": 0.3468785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.5485893046619526e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.5180364431000092e-05,
+      "loss": 0.9887,
       "step": 12224
     },
     {
-      "epoch": 0.33575018538353796,
+      "epoch": 0.34690692395005673,
       "grad_norm": 0.0,
-      "learning_rate": 1.5485149307725842e-05,
-      "loss": 0.8991,
+      "learning_rate": 1.5179578267840762e-05,
+      "loss": 0.8426,
       "step": 12225
     },
     {
-      "epoch": 0.3357776496113812,
+      "epoch": 0.34693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.5484405525431536e-05,
-      "loss": 0.999,
+      "learning_rate": 1.5178792060929307e-05,
+      "loss": 0.8728,
       "step": 12226
     },
     {
-      "epoch": 0.3358051138392244,
+      "epoch": 0.34696367763904656,
       "grad_norm": 0.0,
-      "learning_rate": 1.548366169974249e-05,
-      "loss": 0.8891,
+      "learning_rate": 1.517800581027236e-05,
+      "loss": 0.9204,
       "step": 12227
     },
     {
-      "epoch": 0.3358325780670676,
+      "epoch": 0.3469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.5482917830664585e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.5177219515876565e-05,
+      "loss": 1.002,
       "step": 12228
     },
     {
-      "epoch": 0.3358600422949109,
+      "epoch": 0.34702043132803634,
       "grad_norm": 0.0,
-      "learning_rate": 1.548217391820371e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.5176433177748564e-05,
+      "loss": 0.9046,
       "step": 12229
     },
     {
-      "epoch": 0.3358875065227541,
+      "epoch": 0.3470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.5481429962365753e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.5175646795895e-05,
+      "loss": 0.8596,
       "step": 12230
     },
     {
-      "epoch": 0.33591497075059734,
+      "epoch": 0.3470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.54806859631566e-05,
-      "loss": 0.8682,
+      "learning_rate": 1.5174860370322515e-05,
+      "loss": 0.8937,
       "step": 12231
     },
     {
-      "epoch": 0.33594243497844056,
+      "epoch": 0.347105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.5479941920582135e-05,
-      "loss": 0.8915,
+      "learning_rate": 1.5174073901037749e-05,
+      "loss": 0.9454,
       "step": 12232
     },
     {
-      "epoch": 0.33596989920628384,
+      "epoch": 0.3471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.5479197834648246e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.517328738804735e-05,
+      "loss": 1.0313,
       "step": 12233
     },
     {
-      "epoch": 0.33599736343412706,
+      "epoch": 0.3471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.5478453705360824e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.5172500831357963e-05,
+      "loss": 0.7902,
       "step": 12234
     },
     {
-      "epoch": 0.3360248276619703,
+      "epoch": 0.34719069239500566,
       "grad_norm": 0.0,
-      "learning_rate": 1.547770953272575e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.5171714230976224e-05,
+      "loss": 0.8441,
       "step": 12235
     },
     {
-      "epoch": 0.3360522918898135,
+      "epoch": 0.3472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.547696531674892e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.5170927586908787e-05,
+      "loss": 0.8222,
       "step": 12236
     },
     {
-      "epoch": 0.3360797561176567,
+      "epoch": 0.34724744608399544,
       "grad_norm": 0.0,
-      "learning_rate": 1.5476221057436217e-05,
-      "loss": 1.1117,
+      "learning_rate": 1.5170140899162289e-05,
+      "loss": 0.9647,
       "step": 12237
     },
     {
-      "epoch": 0.3361072203455,
+      "epoch": 0.34727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.547547675479353e-05,
-      "loss": 0.9591,
+      "learning_rate": 1.516935416774338e-05,
+      "loss": 0.9277,
       "step": 12238
     },
     {
-      "epoch": 0.3361346845733432,
+      "epoch": 0.34730419977298527,
       "grad_norm": 0.0,
-      "learning_rate": 1.5474732408826754e-05,
-      "loss": 0.9828,
+      "learning_rate": 1.5168567392658701e-05,
+      "loss": 0.8911,
       "step": 12239
     },
     {
-      "epoch": 0.33616214880118644,
+      "epoch": 0.3473325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.5473988019541774e-05,
-      "loss": 1.14,
+      "learning_rate": 1.5167780573914903e-05,
+      "loss": 0.8413,
       "step": 12240
     },
     {
-      "epoch": 0.33618961302902967,
+      "epoch": 0.34736095346197504,
       "grad_norm": 0.0,
-      "learning_rate": 1.547324358694448e-05,
-      "loss": 1.0577,
+      "learning_rate": 1.5166993711518631e-05,
+      "loss": 0.9268,
       "step": 12241
     },
     {
-      "epoch": 0.33621707725687294,
+      "epoch": 0.3473893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.5472499111040763e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.516620680547653e-05,
+      "loss": 1.1222,
       "step": 12242
     },
     {
-      "epoch": 0.33624454148471616,
+      "epoch": 0.3474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.547175459183651e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.516541985579525e-05,
+      "loss": 0.9011,
       "step": 12243
     },
     {
-      "epoch": 0.3362720057125594,
+      "epoch": 0.34744608399545973,
       "grad_norm": 0.0,
-      "learning_rate": 1.547101002933762e-05,
-      "loss": 1.067,
+      "learning_rate": 1.5164632862481434e-05,
+      "loss": 0.9002,
       "step": 12244
     },
     {
-      "epoch": 0.3362994699404026,
+      "epoch": 0.3474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.5470265423549976e-05,
-      "loss": 0.939,
+      "learning_rate": 1.516384582554173e-05,
+      "loss": 0.9231,
       "step": 12245
     },
     {
-      "epoch": 0.3363269341682459,
+      "epoch": 0.3475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.5469520774479472e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.5163058744982793e-05,
+      "loss": 1.01,
       "step": 12246
     },
     {
-      "epoch": 0.3363543983960891,
+      "epoch": 0.34753121452894437,
       "grad_norm": 0.0,
-      "learning_rate": 1.5468776082132002e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.5162271620811262e-05,
+      "loss": 1.0219,
       "step": 12247
     },
     {
-      "epoch": 0.3363818626239323,
+      "epoch": 0.3475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.546803134651346e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.516148445303379e-05,
+      "loss": 0.8734,
       "step": 12248
     },
     {
-      "epoch": 0.33640932685177555,
+      "epoch": 0.34758796821793414,
       "grad_norm": 0.0,
-      "learning_rate": 1.5467286567629733e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.5160697241657032e-05,
+      "loss": 1.0398,
       "step": 12249
     },
     {
-      "epoch": 0.33643679107961877,
+      "epoch": 0.34761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.5466541745486715e-05,
-      "loss": 0.945,
+      "learning_rate": 1.5159909986687631e-05,
+      "loss": 0.959,
       "step": 12250
     },
     {
-      "epoch": 0.33646425530746205,
+      "epoch": 0.34764472190692397,
       "grad_norm": 0.0,
-      "learning_rate": 1.5465796880090305e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.5159122688132235e-05,
+      "loss": 0.9307,
       "step": 12251
     },
     {
-      "epoch": 0.33649171953530527,
+      "epoch": 0.34767309875141883,
       "grad_norm": 0.0,
-      "learning_rate": 1.546505197144639e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.5158335345997502e-05,
+      "loss": 0.9519,
       "step": 12252
     },
     {
-      "epoch": 0.3365191837631485,
+      "epoch": 0.34770147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.546430701956087e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.5157547960290077e-05,
+      "loss": 0.9052,
       "step": 12253
     },
     {
-      "epoch": 0.3365466479909917,
+      "epoch": 0.3477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.5463562024439633e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.515676053101661e-05,
+      "loss": 0.8808,
       "step": 12254
     },
     {
-      "epoch": 0.336574112218835,
+      "epoch": 0.3477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.546281698608858e-05,
-      "loss": 0.8693,
+      "learning_rate": 1.5155973058183757e-05,
+      "loss": 0.9252,
       "step": 12255
     },
     {
-      "epoch": 0.3366015764466782,
+      "epoch": 0.34778660612939843,
       "grad_norm": 0.0,
-      "learning_rate": 1.54620719045136e-05,
-      "loss": 0.977,
+      "learning_rate": 1.515518554179817e-05,
+      "loss": 1.0422,
       "step": 12256
     },
     {
-      "epoch": 0.33662904067452143,
+      "epoch": 0.3478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.5461326779720592e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.5154397981866495e-05,
+      "loss": 0.8994,
       "step": 12257
     },
     {
-      "epoch": 0.33665650490236465,
+      "epoch": 0.3478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.546058161171545e-05,
-      "loss": 0.955,
+      "learning_rate": 1.5153610378395392e-05,
+      "loss": 1.0262,
       "step": 12258
     },
     {
-      "epoch": 0.3366839691302079,
+      "epoch": 0.34787173666288307,
       "grad_norm": 0.0,
-      "learning_rate": 1.5459836400504075e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.5152822731391508e-05,
+      "loss": 0.9487,
       "step": 12259
     },
     {
-      "epoch": 0.33671143335805115,
+      "epoch": 0.347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.5459091146092358e-05,
-      "loss": 0.885,
+      "learning_rate": 1.5152035040861499e-05,
+      "loss": 0.9442,
       "step": 12260
     },
     {
-      "epoch": 0.33673889758589437,
+      "epoch": 0.3479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.5458345848486195e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.515124730681202e-05,
+      "loss": 0.8938,
       "step": 12261
     },
     {
-      "epoch": 0.3367663618137376,
+      "epoch": 0.34795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.5457600507691487e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.515045952924972e-05,
+      "loss": 0.9116,
       "step": 12262
     },
     {
-      "epoch": 0.33679382604158087,
+      "epoch": 0.3479852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.545685512371413e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.514967170818126e-05,
+      "loss": 0.9391,
       "step": 12263
     },
     {
-      "epoch": 0.3368212902694241,
+      "epoch": 0.34801362088535753,
       "grad_norm": 0.0,
-      "learning_rate": 1.545610969656002e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.514888384361329e-05,
+      "loss": 0.9768,
       "step": 12264
     },
     {
-      "epoch": 0.3368487544972673,
+      "epoch": 0.34804199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 1.545536422623506e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.5148095935552469e-05,
+      "loss": 1.0163,
       "step": 12265
     },
     {
-      "epoch": 0.33687621872511053,
+      "epoch": 0.3480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.545461871274514e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.514730798400545e-05,
+      "loss": 0.8799,
       "step": 12266
     },
     {
-      "epoch": 0.33690368295295375,
+      "epoch": 0.3480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.5453873156096165e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.5146519988978888e-05,
+      "loss": 0.9602,
       "step": 12267
     },
     {
-      "epoch": 0.33693114718079703,
+      "epoch": 0.34812712826333714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5453127556294035e-05,
-      "loss": 1.0521,
+      "learning_rate": 1.514573195047944e-05,
+      "loss": 0.8941,
       "step": 12268
     },
     {
-      "epoch": 0.33695861140864025,
+      "epoch": 0.348155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.545238191334465e-05,
-      "loss": 0.8944,
+      "learning_rate": 1.5144943868513764e-05,
+      "loss": 0.992,
       "step": 12269
     },
     {
-      "epoch": 0.3369860756364835,
+      "epoch": 0.3481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.54516362272539e-05,
-      "loss": 1.0474,
+      "learning_rate": 1.5144155743088513e-05,
+      "loss": 0.9152,
       "step": 12270
     },
     {
-      "epoch": 0.3370135398643267,
+      "epoch": 0.34821225879682177,
       "grad_norm": 0.0,
-      "learning_rate": 1.5450890498027696e-05,
-      "loss": 0.9342,
+      "learning_rate": 1.514336757421035e-05,
+      "loss": 0.9254,
       "step": 12271
     },
     {
-      "epoch": 0.33704100409216997,
+      "epoch": 0.3482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.5450144725671937e-05,
-      "loss": 0.9095,
+      "learning_rate": 1.5142579361885927e-05,
+      "loss": 0.929,
       "step": 12272
     },
     {
-      "epoch": 0.3370684683200132,
+      "epoch": 0.3482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.5449398910192517e-05,
-      "loss": 0.8172,
+      "learning_rate": 1.5141791106121909e-05,
+      "loss": 0.8573,
       "step": 12273
     },
     {
-      "epoch": 0.3370959325478564,
+      "epoch": 0.34829738933030646,
       "grad_norm": 0.0,
-      "learning_rate": 1.5448653051595346e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.5141002806924948e-05,
+      "loss": 0.8822,
       "step": 12274
     },
     {
-      "epoch": 0.33712339677569964,
+      "epoch": 0.3483257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.544790714988632e-05,
-      "loss": 0.9077,
+      "learning_rate": 1.5140214464301704e-05,
+      "loss": 0.8511,
       "step": 12275
     },
     {
-      "epoch": 0.3371508610035429,
+      "epoch": 0.34835414301929624,
       "grad_norm": 0.0,
-      "learning_rate": 1.5447161205071342e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.5139426078258842e-05,
+      "loss": 0.83,
       "step": 12276
     },
     {
-      "epoch": 0.33717832523138613,
+      "epoch": 0.34838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.5446415217156316e-05,
-      "loss": 0.8966,
+      "learning_rate": 1.5138637648803011e-05,
+      "loss": 0.9413,
       "step": 12277
     },
     {
-      "epoch": 0.33720578945922935,
+      "epoch": 0.348410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.5445669186147144e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.5137849175940881e-05,
+      "loss": 0.9961,
       "step": 12278
     },
     {
-      "epoch": 0.3372332536870726,
+      "epoch": 0.3484392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.5444923112049724e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.5137060659679104e-05,
+      "loss": 0.9959,
       "step": 12279
     },
     {
-      "epoch": 0.3372607179149158,
+      "epoch": 0.34846765039727584,
       "grad_norm": 0.0,
-      "learning_rate": 1.5444176994869966e-05,
-      "loss": 0.914,
+      "learning_rate": 1.513627210002435e-05,
+      "loss": 0.9293,
       "step": 12280
     },
     {
-      "epoch": 0.3372881821427591,
+      "epoch": 0.3484960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.544343083461377e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.5135483496983273e-05,
+      "loss": 1.0329,
       "step": 12281
     },
     {
-      "epoch": 0.3373156463706023,
+      "epoch": 0.3485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.5442684631287042e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.5134694850562533e-05,
+      "loss": 0.9716,
       "step": 12282
     },
     {
-      "epoch": 0.3373431105984455,
+      "epoch": 0.3485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5441938384895684e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.5133906160768797e-05,
+      "loss": 0.8366,
       "step": 12283
     },
     {
-      "epoch": 0.33737057482628874,
+      "epoch": 0.3485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.54411920954456e-05,
-      "loss": 0.8499,
+      "learning_rate": 1.5133117427608726e-05,
+      "loss": 0.8767,
       "step": 12284
     },
     {
-      "epoch": 0.337398039054132,
+      "epoch": 0.3486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.5440445762942696e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.5132328651088979e-05,
+      "loss": 1.057,
       "step": 12285
     },
     {
-      "epoch": 0.33742550328197524,
+      "epoch": 0.34863791146424516,
       "grad_norm": 0.0,
-      "learning_rate": 1.5439699387392878e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.513153983121622e-05,
+      "loss": 0.7888,
       "step": 12286
     },
     {
-      "epoch": 0.33745296750981846,
+      "epoch": 0.3486662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.5438952968802056e-05,
-      "loss": 0.7666,
+      "learning_rate": 1.5130750967997117e-05,
+      "loss": 1.0201,
       "step": 12287
     },
     {
-      "epoch": 0.3374804317376617,
+      "epoch": 0.34869466515323494,
       "grad_norm": 0.0,
-      "learning_rate": 1.5438206507176126e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.512996206143833e-05,
+      "loss": 0.9846,
       "step": 12288
     },
     {
-      "epoch": 0.33750789596550496,
+      "epoch": 0.34872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.5437460002521004e-05,
-      "loss": 1.073,
+      "learning_rate": 1.5129173111546518e-05,
+      "loss": 0.9224,
       "step": 12289
     },
     {
-      "epoch": 0.3375353601933482,
+      "epoch": 0.34875141884222477,
       "grad_norm": 0.0,
-      "learning_rate": 1.5436713454842588e-05,
-      "loss": 0.9758,
+      "learning_rate": 1.5128384118328355e-05,
+      "loss": 0.9565,
       "step": 12290
     },
     {
-      "epoch": 0.3375628244211914,
+      "epoch": 0.34877979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.543596686414679e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.5127595081790501e-05,
+      "loss": 0.8846,
       "step": 12291
     },
     {
-      "epoch": 0.3375902886490346,
+      "epoch": 0.34880817253121454,
       "grad_norm": 0.0,
-      "learning_rate": 1.5435220230439517e-05,
-      "loss": 0.9103,
+      "learning_rate": 1.5126806001939618e-05,
+      "loss": 0.9818,
       "step": 12292
     },
     {
-      "epoch": 0.33761775287687784,
+      "epoch": 0.3488365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.5434473553726673e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.5126016878782372e-05,
+      "loss": 0.8112,
       "step": 12293
     },
     {
-      "epoch": 0.3376452171047211,
+      "epoch": 0.3488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.5433726834014175e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.5125227712325436e-05,
+      "loss": 1.0157,
       "step": 12294
     },
     {
-      "epoch": 0.33767268133256434,
+      "epoch": 0.3488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.5432980071307924e-05,
-      "loss": 1.0715,
+      "learning_rate": 1.5124438502575472e-05,
+      "loss": 0.9772,
       "step": 12295
     },
     {
-      "epoch": 0.33770014556040756,
+      "epoch": 0.3489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.5432233265613826e-05,
-      "loss": 0.9145,
+      "learning_rate": 1.5123649249539141e-05,
+      "loss": 0.911,
       "step": 12296
     },
     {
-      "epoch": 0.3377276097882508,
+      "epoch": 0.348950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.54314864169378e-05,
-      "loss": 0.965,
+      "learning_rate": 1.5122859953223116e-05,
+      "loss": 0.9741,
       "step": 12297
     },
     {
-      "epoch": 0.33775507401609406,
+      "epoch": 0.34897843359818387,
       "grad_norm": 0.0,
-      "learning_rate": 1.5430739525285745e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.5122070613634066e-05,
+      "loss": 0.9056,
       "step": 12298
     },
     {
-      "epoch": 0.3377825382439373,
+      "epoch": 0.3490068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.5429992590663577e-05,
-      "loss": 0.8838,
+      "learning_rate": 1.512128123077865e-05,
+      "loss": 0.9345,
       "step": 12299
     },
     {
-      "epoch": 0.3378100024717805,
+      "epoch": 0.34903518728717364,
       "grad_norm": 0.0,
-      "learning_rate": 1.5429245613077207e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.5120491804663543e-05,
+      "loss": 0.9962,
       "step": 12300
     },
     {
-      "epoch": 0.3378374666996237,
+      "epoch": 0.34906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.5428498592532538e-05,
-      "loss": 0.8735,
+      "learning_rate": 1.5119702335295415e-05,
+      "loss": 0.933,
       "step": 12301
     },
     {
-      "epoch": 0.337864930927467,
+      "epoch": 0.3490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.542775152903549e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.5118912822680924e-05,
+      "loss": 0.8535,
       "step": 12302
     },
     {
-      "epoch": 0.3378923951553102,
+      "epoch": 0.34912031782065833,
       "grad_norm": 0.0,
-      "learning_rate": 1.5427004422591965e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.5118123266826752e-05,
+      "loss": 0.9591,
       "step": 12303
     },
     {
-      "epoch": 0.33791985938315344,
+      "epoch": 0.34914869466515325,
       "grad_norm": 0.0,
-      "learning_rate": 1.542625727320788e-05,
-      "loss": 1.132,
+      "learning_rate": 1.5117333667739564e-05,
+      "loss": 0.9385,
       "step": 12304
     },
     {
-      "epoch": 0.33794732361099666,
+      "epoch": 0.3491770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.5425510080889147e-05,
-      "loss": 0.9588,
+      "learning_rate": 1.5116544025426025e-05,
+      "loss": 0.9086,
       "step": 12305
     },
     {
-      "epoch": 0.3379747878388399,
+      "epoch": 0.349205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.5424762845641673e-05,
-      "loss": 0.9342,
+      "learning_rate": 1.511575433989281e-05,
+      "loss": 0.9655,
       "step": 12306
     },
     {
-      "epoch": 0.33800225206668316,
+      "epoch": 0.34923382519863794,
       "grad_norm": 0.0,
-      "learning_rate": 1.5424015567471378e-05,
-      "loss": 0.8956,
+      "learning_rate": 1.5114964611146587e-05,
+      "loss": 0.9892,
       "step": 12307
     },
     {
-      "epoch": 0.3380297162945264,
+      "epoch": 0.3492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.542326824638417e-05,
-      "loss": 0.9529,
+      "learning_rate": 1.5114174839194029e-05,
+      "loss": 0.911,
       "step": 12308
     },
     {
-      "epoch": 0.3380571805223696,
+      "epoch": 0.3492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.542252088238596e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.5113385024041804e-05,
+      "loss": 1.0422,
       "step": 12309
     },
     {
-      "epoch": 0.3380846447502128,
+      "epoch": 0.34931895573212257,
       "grad_norm": 0.0,
-      "learning_rate": 1.5421773475482665e-05,
-      "loss": 0.8727,
+      "learning_rate": 1.5112595165696592e-05,
+      "loss": 0.9539,
       "step": 12310
     },
     {
-      "epoch": 0.3381121089780561,
+      "epoch": 0.3493473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.54210260256802e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.5111805264165051e-05,
+      "loss": 0.9408,
       "step": 12311
     },
     {
-      "epoch": 0.3381395732058993,
+      "epoch": 0.34937570942111235,
       "grad_norm": 0.0,
-      "learning_rate": 1.5420278532984476e-05,
-      "loss": 0.9942,
+      "learning_rate": 1.5111015319453866e-05,
+      "loss": 0.8398,
       "step": 12312
     },
     {
-      "epoch": 0.33816703743374255,
+      "epoch": 0.34940408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.541953099740141e-05,
-      "loss": 0.9259,
+      "learning_rate": 1.5110225331569704e-05,
+      "loss": 0.9006,
       "step": 12313
     },
     {
-      "epoch": 0.33819450166158577,
+      "epoch": 0.3494324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 1.5418783418936914e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.510943530051924e-05,
+      "loss": 0.9892,
       "step": 12314
     },
     {
-      "epoch": 0.33822196588942904,
+      "epoch": 0.34946083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.541803579759691e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.5108645226309142e-05,
+      "loss": 0.8918,
       "step": 12315
     },
     {
-      "epoch": 0.33824943011727227,
+      "epoch": 0.34948921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 1.54172881333873e-05,
-      "loss": 0.9114,
+      "learning_rate": 1.5107855108946091e-05,
+      "loss": 0.9296,
       "step": 12316
     },
     {
-      "epoch": 0.3382768943451155,
+      "epoch": 0.3495175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.541654042631401e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.5107064948436758e-05,
+      "loss": 0.9438,
       "step": 12317
     },
     {
-      "epoch": 0.3383043585729587,
+      "epoch": 0.3495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.5415792676382958e-05,
-      "loss": 1.039,
+      "learning_rate": 1.5106274744787819e-05,
+      "loss": 0.9574,
       "step": 12318
     },
     {
-      "epoch": 0.33833182280080193,
+      "epoch": 0.34957434733257664,
       "grad_norm": 0.0,
-      "learning_rate": 1.5415044883600052e-05,
-      "loss": 0.933,
+      "learning_rate": 1.5105484498005946e-05,
+      "loss": 0.8928,
       "step": 12319
     },
     {
-      "epoch": 0.3383592870286452,
+      "epoch": 0.3496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.5414297047971214e-05,
-      "loss": 0.9052,
+      "learning_rate": 1.5104694208097816e-05,
+      "loss": 0.9855,
       "step": 12320
     },
     {
-      "epoch": 0.3383867512564884,
+      "epoch": 0.3496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.5413549169502363e-05,
-      "loss": 0.8987,
+      "learning_rate": 1.5103903875070106e-05,
+      "loss": 0.9009,
       "step": 12321
     },
     {
-      "epoch": 0.33841421548433165,
+      "epoch": 0.3496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.5412801248199412e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.5103113498929489e-05,
+      "loss": 0.8338,
       "step": 12322
     },
     {
-      "epoch": 0.33844167971217487,
+      "epoch": 0.3496878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.5412053284068277e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.5102323079682641e-05,
+      "loss": 0.8382,
       "step": 12323
     },
     {
-      "epoch": 0.33846914394001815,
+      "epoch": 0.3497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.541130527711489e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.5101532617336245e-05,
+      "loss": 0.8992,
       "step": 12324
     },
     {
-      "epoch": 0.33849660816786137,
+      "epoch": 0.34974460839954596,
       "grad_norm": 0.0,
-      "learning_rate": 1.541055722734515e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.5100742111896973e-05,
+      "loss": 0.9331,
       "step": 12325
     },
     {
-      "epoch": 0.3385240723957046,
+      "epoch": 0.3497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.540980913476499e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.50999515633715e-05,
+      "loss": 0.972,
       "step": 12326
     },
     {
-      "epoch": 0.3385515366235478,
+      "epoch": 0.34980136208853574,
       "grad_norm": 0.0,
-      "learning_rate": 1.5409060999380322e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.5099160971766509e-05,
+      "loss": 0.9006,
       "step": 12327
     },
     {
-      "epoch": 0.3385790008513911,
+      "epoch": 0.34982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.5408312821197073e-05,
-      "loss": 0.9806,
+      "learning_rate": 1.509837033708868e-05,
+      "loss": 0.9731,
       "step": 12328
     },
     {
-      "epoch": 0.3386064650792343,
+      "epoch": 0.3498581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.5407564600221153e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.5097579659344682e-05,
+      "loss": 0.8493,
       "step": 12329
     },
     {
-      "epoch": 0.33863392930707753,
+      "epoch": 0.34988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.5406816336458485e-05,
-      "loss": 0.8798,
+      "learning_rate": 1.5096788938541198e-05,
+      "loss": 1.0816,
       "step": 12330
     },
     {
-      "epoch": 0.33866139353492075,
+      "epoch": 0.34991486946651534,
       "grad_norm": 0.0,
-      "learning_rate": 1.5406068029915e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.5095998174684915e-05,
+      "loss": 0.9272,
       "step": 12331
     },
     {
-      "epoch": 0.338688857762764,
+      "epoch": 0.3499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.5405319680596603e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.50952073677825e-05,
+      "loss": 0.9139,
       "step": 12332
     },
     {
-      "epoch": 0.33871632199060725,
+      "epoch": 0.3499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.5404571288509226e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.5094416517840642e-05,
+      "loss": 0.9917,
       "step": 12333
     },
     {
-      "epoch": 0.33874378621845047,
+      "epoch": 0.35,
       "grad_norm": 0.0,
-      "learning_rate": 1.5403822853658788e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.509362562486602e-05,
+      "loss": 1.0053,
       "step": 12334
     },
     {
-      "epoch": 0.3387712504462937,
+      "epoch": 0.3500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.540307437605121e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.5092834688865311e-05,
+      "loss": 0.9512,
       "step": 12335
     },
     {
-      "epoch": 0.3387987146741369,
+      "epoch": 0.3500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.5402325855692414e-05,
-      "loss": 1.041,
+      "learning_rate": 1.50920437098452e-05,
+      "loss": 1.0323,
       "step": 12336
     },
     {
-      "epoch": 0.3388261789019802,
+      "epoch": 0.35008513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 1.5401577292588323e-05,
-      "loss": 0.99,
+      "learning_rate": 1.5091252687812367e-05,
+      "loss": 1.0418,
       "step": 12337
     },
     {
-      "epoch": 0.3388536431298234,
+      "epoch": 0.3501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.540082868674486e-05,
-      "loss": 0.9008,
+      "learning_rate": 1.5090461622773496e-05,
+      "loss": 0.8704,
       "step": 12338
     },
     {
-      "epoch": 0.33888110735766663,
+      "epoch": 0.35014188422247444,
       "grad_norm": 0.0,
-      "learning_rate": 1.540008003816795e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.5089670514735261e-05,
+      "loss": 0.8813,
       "step": 12339
     },
     {
-      "epoch": 0.33890857158550985,
+      "epoch": 0.35017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.5399331346863514e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.5088879363704351e-05,
+      "loss": 0.9422,
       "step": 12340
     },
     {
-      "epoch": 0.33893603581335313,
+      "epoch": 0.35019863791146427,
       "grad_norm": 0.0,
-      "learning_rate": 1.5398582612837478e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.5088088169687453e-05,
+      "loss": 0.8946,
       "step": 12341
     },
     {
-      "epoch": 0.33896350004119635,
+      "epoch": 0.35022701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.539783383609576e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.5087296932691244e-05,
+      "loss": 0.9437,
       "step": 12342
     },
     {
-      "epoch": 0.3389909642690396,
+      "epoch": 0.35025539160045405,
       "grad_norm": 0.0,
-      "learning_rate": 1.5397085016644297e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.5086505652722407e-05,
+      "loss": 0.9729,
       "step": 12343
     },
     {
-      "epoch": 0.3390184284968828,
+      "epoch": 0.3502837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.5396336154489003e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.5085714329787631e-05,
+      "loss": 0.9624,
       "step": 12344
     },
     {
-      "epoch": 0.339045892724726,
+      "epoch": 0.3503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.539558724963581e-05,
-      "loss": 1.0074,
+      "learning_rate": 1.5084922963893597e-05,
+      "loss": 0.9275,
       "step": 12345
     },
     {
-      "epoch": 0.3390733569525693,
+      "epoch": 0.3503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.5394838302090636e-05,
-      "loss": 0.9201,
+      "learning_rate": 1.5084131555046985e-05,
+      "loss": 0.7721,
       "step": 12346
     },
     {
-      "epoch": 0.3391008211804125,
+      "epoch": 0.3503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.5394089311859414e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.5083340103254491e-05,
+      "loss": 0.9282,
       "step": 12347
     },
     {
-      "epoch": 0.33912828540825574,
+      "epoch": 0.3503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.5393340278948068e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.5082548608522794e-05,
+      "loss": 0.9379,
       "step": 12348
     },
     {
-      "epoch": 0.33915574963609896,
+      "epoch": 0.35042565266742337,
       "grad_norm": 0.0,
-      "learning_rate": 1.5392591203362522e-05,
-      "loss": 0.8704,
+      "learning_rate": 1.5081757070858581e-05,
+      "loss": 0.9751,
       "step": 12349
     },
     {
-      "epoch": 0.33918321386394223,
+      "epoch": 0.3504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.539184208510871e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.5080965490268532e-05,
+      "loss": 1.0254,
       "step": 12350
     },
     {
-      "epoch": 0.33921067809178546,
+      "epoch": 0.35048240635641315,
       "grad_norm": 0.0,
-      "learning_rate": 1.539109292419255e-05,
-      "loss": 0.94,
+      "learning_rate": 1.5080173866759345e-05,
+      "loss": 0.8377,
       "step": 12351
     },
     {
-      "epoch": 0.3392381423196287,
+      "epoch": 0.35051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.5390343720619976e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.5079382200337697e-05,
+      "loss": 0.948,
       "step": 12352
     },
     {
-      "epoch": 0.3392656065474719,
+      "epoch": 0.350539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.538959447439692e-05,
-      "loss": 0.9062,
+      "learning_rate": 1.507859049101028e-05,
+      "loss": 1.0005,
       "step": 12353
     },
     {
-      "epoch": 0.3392930707753152,
+      "epoch": 0.35056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.5388845185529296e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.5077798738783784e-05,
+      "loss": 1.0872,
       "step": 12354
     },
     {
-      "epoch": 0.3393205350031584,
+      "epoch": 0.35059591373439275,
       "grad_norm": 0.0,
-      "learning_rate": 1.5388095854023045e-05,
-      "loss": 0.7879,
+      "learning_rate": 1.5077006943664891e-05,
+      "loss": 1.0114,
       "step": 12355
     },
     {
-      "epoch": 0.3393479992310016,
+      "epoch": 0.3506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.5387346479884094e-05,
-      "loss": 0.8711,
+      "learning_rate": 1.5076215105660292e-05,
+      "loss": 0.9229,
       "step": 12356
     },
     {
-      "epoch": 0.33937546345884484,
+      "epoch": 0.3506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.5386597063118372e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.5075423224776677e-05,
+      "loss": 0.9676,
       "step": 12357
     },
     {
-      "epoch": 0.3394029276866881,
+      "epoch": 0.35068104426787744,
       "grad_norm": 0.0,
-      "learning_rate": 1.5385847603731803e-05,
-      "loss": 0.8927,
+      "learning_rate": 1.5074631301020737e-05,
+      "loss": 0.7983,
       "step": 12358
     },
     {
-      "epoch": 0.33943039191453134,
+      "epoch": 0.3507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.5385098101730325e-05,
-      "loss": 0.9342,
+      "learning_rate": 1.5073839334399153e-05,
+      "loss": 1.0152,
       "step": 12359
     },
     {
-      "epoch": 0.33945785614237456,
+      "epoch": 0.3507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.5384348557119864e-05,
-      "loss": 0.8777,
+      "learning_rate": 1.5073047324918621e-05,
+      "loss": 0.8668,
       "step": 12360
     },
     {
-      "epoch": 0.3394853203702178,
+      "epoch": 0.3507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.538359896990635e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.5072255272585833e-05,
+      "loss": 0.9483,
       "step": 12361
     },
     {
-      "epoch": 0.339512784598061,
+      "epoch": 0.350794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.5382849340095714e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.5071463177407473e-05,
+      "loss": 0.9092,
       "step": 12362
     },
     {
-      "epoch": 0.3395402488259043,
+      "epoch": 0.35082292849035185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5382099667693894e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.5070671039390237e-05,
+      "loss": 0.8652,
       "step": 12363
     },
     {
-      "epoch": 0.3395677130537475,
+      "epoch": 0.35085130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.5381349952706815e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.5069878858540814e-05,
+      "loss": 0.8986,
       "step": 12364
     },
     {
-      "epoch": 0.3395951772815907,
+      "epoch": 0.3508796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.538060019514041e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.50690866348659e-05,
+      "loss": 0.8674,
       "step": 12365
     },
     {
-      "epoch": 0.33962264150943394,
+      "epoch": 0.35090805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.537985039500061e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.5068294368372179e-05,
+      "loss": 1.0164,
       "step": 12366
     },
     {
-      "epoch": 0.3396501057372772,
+      "epoch": 0.35093643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 1.5379100552293352e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.506750205906635e-05,
+      "loss": 0.9027,
       "step": 12367
     },
     {
-      "epoch": 0.33967756996512044,
+      "epoch": 0.3509648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.537835066702457e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.5066709706955105e-05,
+      "loss": 1.0913,
       "step": 12368
     },
     {
-      "epoch": 0.33970503419296366,
+      "epoch": 0.35099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.5377600739200188e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.5065917312045131e-05,
+      "loss": 1.0618,
       "step": 12369
     },
     {
-      "epoch": 0.3397324984208069,
+      "epoch": 0.35102156640181614,
       "grad_norm": 0.0,
-      "learning_rate": 1.537685076882615e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.5065124874343127e-05,
+      "loss": 0.8523,
       "step": 12370
     },
     {
-      "epoch": 0.33975996264865016,
+      "epoch": 0.351049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.537610075590838e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.5064332393855787e-05,
+      "loss": 0.9692,
       "step": 12371
     },
     {
-      "epoch": 0.3397874268764934,
+      "epoch": 0.3510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.5375350700452827e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.5063539870589803e-05,
+      "loss": 0.9383,
       "step": 12372
     },
     {
-      "epoch": 0.3398148911043366,
+      "epoch": 0.3511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.537460060246541e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.506274730455187e-05,
+      "loss": 0.9474,
       "step": 12373
     },
     {
-      "epoch": 0.3398423553321798,
+      "epoch": 0.3511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.5373850461952073e-05,
-      "loss": 1.0345,
+      "learning_rate": 1.5061954695748684e-05,
+      "loss": 0.834,
       "step": 12374
     },
     {
-      "epoch": 0.33986981956002305,
+      "epoch": 0.35116345062429055,
       "grad_norm": 0.0,
-      "learning_rate": 1.5373100278918748e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.506116204418694e-05,
+      "loss": 0.9061,
       "step": 12375
     },
     {
-      "epoch": 0.3398972837878663,
+      "epoch": 0.35119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.537235005337138e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.506036934987333e-05,
+      "loss": 0.9041,
       "step": 12376
     },
     {
-      "epoch": 0.33992474801570954,
+      "epoch": 0.3512202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.537159978531589e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.5059576612814551e-05,
+      "loss": 0.9755,
       "step": 12377
     },
     {
-      "epoch": 0.33995221224355276,
+      "epoch": 0.35124858115777524,
       "grad_norm": 0.0,
-      "learning_rate": 1.5370849474758222e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.5058783833017304e-05,
+      "loss": 0.9733,
       "step": 12378
     },
     {
-      "epoch": 0.339979676471396,
+      "epoch": 0.35127695800227016,
       "grad_norm": 0.0,
-      "learning_rate": 1.5370099121704317e-05,
-      "loss": 0.998,
+      "learning_rate": 1.5057991010488281e-05,
+      "loss": 0.969,
       "step": 12379
     },
     {
-      "epoch": 0.34000714069923926,
+      "epoch": 0.351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.5369348726160102e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.5057198145234182e-05,
+      "loss": 0.9351,
       "step": 12380
     },
     {
-      "epoch": 0.3400346049270825,
+      "epoch": 0.35133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.5368598288131523e-05,
-      "loss": 0.956,
+      "learning_rate": 1.5056405237261702e-05,
+      "loss": 1.0224,
       "step": 12381
     },
     {
-      "epoch": 0.3400620691549257,
+      "epoch": 0.35136208853575485,
       "grad_norm": 0.0,
-      "learning_rate": 1.5367847807624513e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.5055612286577541e-05,
+      "loss": 0.9664,
       "step": 12382
     },
     {
-      "epoch": 0.3400895333827689,
+      "epoch": 0.3513904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.5367097284645014e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.5054819293188394e-05,
+      "loss": 0.9372,
       "step": 12383
     },
     {
-      "epoch": 0.3401169976106122,
+      "epoch": 0.3514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.5366346719198963e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.5054026257100964e-05,
+      "loss": 0.9138,
       "step": 12384
     },
     {
-      "epoch": 0.3401444618384554,
+      "epoch": 0.3514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.5365596111292296e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.5053233178321942e-05,
+      "loss": 0.925,
       "step": 12385
     },
     {
-      "epoch": 0.34017192606629865,
+      "epoch": 0.3514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.5364845460930954e-05,
-      "loss": 0.9263,
+      "learning_rate": 1.5052440056858038e-05,
+      "loss": 0.8338,
       "step": 12386
     },
     {
-      "epoch": 0.34019939029414187,
+      "epoch": 0.3515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.5364094768120875e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.505164689271594e-05,
+      "loss": 1.0523,
       "step": 12387
     },
     {
-      "epoch": 0.3402268545219851,
+      "epoch": 0.35153234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 1.5363344032868005e-05,
-      "loss": 0.8917,
+      "learning_rate": 1.5050853685902358e-05,
+      "loss": 0.8882,
       "step": 12388
     },
     {
-      "epoch": 0.34025431874982837,
+      "epoch": 0.3515607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.536259325517828e-05,
-      "loss": 1.0301,
+      "learning_rate": 1.5050060436423986e-05,
+      "loss": 0.8747,
       "step": 12389
     },
     {
-      "epoch": 0.3402817829776716,
+      "epoch": 0.35158910329171394,
       "grad_norm": 0.0,
-      "learning_rate": 1.5361842435057634e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.5049267144287527e-05,
+      "loss": 1.0095,
       "step": 12390
     },
     {
-      "epoch": 0.3403092472055148,
+      "epoch": 0.35161748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.536109157251202e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.5048473809499682e-05,
+      "loss": 0.8686,
       "step": 12391
     },
     {
-      "epoch": 0.34033671143335803,
+      "epoch": 0.3516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.536034066754737e-05,
-      "loss": 0.9825,
+      "learning_rate": 1.5047680432067151e-05,
+      "loss": 0.9426,
       "step": 12392
     },
     {
-      "epoch": 0.3403641756612013,
+      "epoch": 0.35167423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.5359589720169625e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.5046887011996635e-05,
+      "loss": 1.0825,
       "step": 12393
     },
     {
-      "epoch": 0.34039163988904453,
+      "epoch": 0.35170261066969355,
       "grad_norm": 0.0,
-      "learning_rate": 1.5358838730384733e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.504609354929484e-05,
+      "loss": 0.9286,
       "step": 12394
     },
     {
-      "epoch": 0.34041910411688775,
+      "epoch": 0.3517309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.535808769819863e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.5045300043968465e-05,
+      "loss": 0.9352,
       "step": 12395
     },
     {
-      "epoch": 0.34044656834473097,
+      "epoch": 0.3517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.5357336623617265e-05,
-      "loss": 1.0207,
+      "learning_rate": 1.5044506496024216e-05,
+      "loss": 0.9829,
       "step": 12396
     },
     {
-      "epoch": 0.34047403257257425,
+      "epoch": 0.3517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.5356585506646578e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.504371290546879e-05,
+      "loss": 0.9436,
       "step": 12397
     },
     {
-      "epoch": 0.34050149680041747,
+      "epoch": 0.3518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.535583434729251e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.5042919272308895e-05,
+      "loss": 0.7895,
       "step": 12398
     },
     {
-      "epoch": 0.3405289610282607,
+      "epoch": 0.351844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.5355083145561008e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.5042125596551235e-05,
+      "loss": 0.8368,
       "step": 12399
     },
     {
-      "epoch": 0.3405564252561039,
+      "epoch": 0.3518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.5354331901458014e-05,
-      "loss": 0.8927,
+      "learning_rate": 1.5041331878202514e-05,
+      "loss": 0.8851,
       "step": 12400
     },
     {
-      "epoch": 0.34058388948394713,
+      "epoch": 0.3519012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.5353580614989468e-05,
-      "loss": 0.8136,
+      "learning_rate": 1.5040538117269435e-05,
+      "loss": 1.0091,
       "step": 12401
     },
     {
-      "epoch": 0.3406113537117904,
+      "epoch": 0.35192962542565265,
       "grad_norm": 0.0,
-      "learning_rate": 1.535282928616132e-05,
-      "loss": 1.0574,
+      "learning_rate": 1.5039744313758706e-05,
+      "loss": 0.9401,
       "step": 12402
     },
     {
-      "epoch": 0.34063881793963363,
+      "epoch": 0.35195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.5352077914979513e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.5038950467677029e-05,
+      "loss": 0.9273,
       "step": 12403
     },
     {
-      "epoch": 0.34066628216747685,
+      "epoch": 0.3519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.535132650144999e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.5038156579031109e-05,
+      "loss": 1.0218,
       "step": 12404
     },
     {
-      "epoch": 0.3406937463953201,
+      "epoch": 0.35201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.5350575045578704e-05,
-      "loss": 0.971,
+      "learning_rate": 1.5037362647827656e-05,
+      "loss": 0.9839,
       "step": 12405
     },
     {
-      "epoch": 0.34072121062316335,
+      "epoch": 0.35204313280363225,
       "grad_norm": 0.0,
-      "learning_rate": 1.5349823547371592e-05,
-      "loss": 0.946,
+      "learning_rate": 1.5036568674073375e-05,
+      "loss": 0.891,
       "step": 12406
     },
     {
-      "epoch": 0.34074867485100657,
+      "epoch": 0.3520715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.5349072006834606e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.5035774657774972e-05,
+      "loss": 1.0002,
       "step": 12407
     },
     {
-      "epoch": 0.3407761390788498,
+      "epoch": 0.352099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.534832042397369e-05,
-      "loss": 1.0351,
+      "learning_rate": 1.5034980598939151e-05,
+      "loss": 0.8955,
       "step": 12408
     },
     {
-      "epoch": 0.340803603306693,
+      "epoch": 0.3521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.534756879879479e-05,
-      "loss": 1.0082,
+      "learning_rate": 1.5034186497572624e-05,
+      "loss": 0.9452,
       "step": 12409
     },
     {
-      "epoch": 0.3408310675345363,
+      "epoch": 0.3521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.5346817131303856e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.5033392353682097e-05,
+      "loss": 0.7757,
       "step": 12410
     },
     {
-      "epoch": 0.3408585317623795,
+      "epoch": 0.3521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.5346065421506833e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.5032598167274279e-05,
+      "loss": 0.9484,
       "step": 12411
     },
     {
-      "epoch": 0.34088599599022273,
+      "epoch": 0.3522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.534531366940967e-05,
-      "loss": 0.921,
+      "learning_rate": 1.503180393835588e-05,
+      "loss": 1.0011,
       "step": 12412
     },
     {
-      "epoch": 0.34091346021806596,
+      "epoch": 0.3522417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.534456187501831e-05,
-      "loss": 1.0458,
+      "learning_rate": 1.5031009666933606e-05,
+      "loss": 0.8755,
       "step": 12413
     },
     {
-      "epoch": 0.3409409244459092,
+      "epoch": 0.35227014755959135,
       "grad_norm": 0.0,
-      "learning_rate": 1.5343810038338708e-05,
-      "loss": 1.0645,
+      "learning_rate": 1.5030215353014166e-05,
+      "loss": 1.0069,
       "step": 12414
     },
     {
-      "epoch": 0.34096838867375245,
+      "epoch": 0.35229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.5343058159376815e-05,
-      "loss": 0.989,
+      "learning_rate": 1.5029420996604272e-05,
+      "loss": 0.9095,
       "step": 12415
     },
     {
-      "epoch": 0.3409958529015957,
+      "epoch": 0.3523269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.5342306238138577e-05,
-      "loss": 0.8832,
+      "learning_rate": 1.5028626597710632e-05,
+      "loss": 0.9103,
       "step": 12416
     },
     {
-      "epoch": 0.3410233171294389,
+      "epoch": 0.35235527809307604,
       "grad_norm": 0.0,
-      "learning_rate": 1.5341554274629938e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.5027832156339957e-05,
+      "loss": 0.9602,
       "step": 12417
     },
     {
-      "epoch": 0.3410507813572821,
+      "epoch": 0.35238365493757096,
       "grad_norm": 0.0,
-      "learning_rate": 1.5340802268856853e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.5027037672498957e-05,
+      "loss": 0.8628,
       "step": 12418
     },
     {
-      "epoch": 0.3410782455851254,
+      "epoch": 0.3524120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.534005022082527e-05,
-      "loss": 0.997,
+      "learning_rate": 1.5026243146194346e-05,
+      "loss": 1.0218,
       "step": 12419
     },
     {
-      "epoch": 0.3411057098129686,
+      "epoch": 0.35244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.533929813054115e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.5025448577432831e-05,
+      "loss": 1.0043,
       "step": 12420
     },
     {
-      "epoch": 0.34113317404081184,
+      "epoch": 0.35246878547105565,
       "grad_norm": 0.0,
-      "learning_rate": 1.5338545998010426e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.5024653966221125e-05,
+      "loss": 0.8933,
       "step": 12421
     },
     {
-      "epoch": 0.34116063826865506,
+      "epoch": 0.3524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.533779382323906e-05,
-      "loss": 0.9931,
+      "learning_rate": 1.5023859312565945e-05,
+      "loss": 0.9028,
       "step": 12422
     },
     {
-      "epoch": 0.34118810249649834,
+      "epoch": 0.3525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.5337041606233005e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.5023064616473996e-05,
+      "loss": 0.9153,
       "step": 12423
     },
     {
-      "epoch": 0.34121556672434156,
+      "epoch": 0.3525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.5336289346998205e-05,
-      "loss": 1.0138,
+      "learning_rate": 1.5022269877951996e-05,
+      "loss": 0.9976,
       "step": 12424
     },
     {
-      "epoch": 0.3412430309521848,
+      "epoch": 0.3525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.533553704554062e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.5021475097006657e-05,
+      "loss": 0.9416,
       "step": 12425
     },
     {
-      "epoch": 0.341270495180028,
+      "epoch": 0.35261066969353005,
       "grad_norm": 0.0,
-      "learning_rate": 1.5334784701866195e-05,
-      "loss": 0.8572,
+      "learning_rate": 1.5020680273644692e-05,
+      "loss": 0.8826,
       "step": 12426
     },
     {
-      "epoch": 0.3412979594078712,
+      "epoch": 0.35263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.5334032315980888e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.5019885407872814e-05,
+      "loss": 1.0034,
       "step": 12427
     },
     {
-      "epoch": 0.3413254236357145,
+      "epoch": 0.3526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.5333279887890652e-05,
-      "loss": 0.899,
+      "learning_rate": 1.5019090499697739e-05,
+      "loss": 0.8596,
       "step": 12428
     },
     {
-      "epoch": 0.3413528878635577,
+      "epoch": 0.35269580022701474,
       "grad_norm": 0.0,
-      "learning_rate": 1.5332527417601443e-05,
-      "loss": 0.9107,
+      "learning_rate": 1.501829554912618e-05,
+      "loss": 0.8575,
       "step": 12429
     },
     {
-      "epoch": 0.34138035209140094,
+      "epoch": 0.35272417707150966,
       "grad_norm": 0.0,
-      "learning_rate": 1.5331774905119207e-05,
-      "loss": 1.0226,
+      "learning_rate": 1.5017500556164855e-05,
+      "loss": 0.9368,
       "step": 12430
     },
     {
-      "epoch": 0.34140781631924416,
+      "epoch": 0.3527525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.5331022350449903e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.5016705520820476e-05,
+      "loss": 0.8815,
       "step": 12431
     },
     {
-      "epoch": 0.34143528054708744,
+      "epoch": 0.35278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.533026975359949e-05,
-      "loss": 1.0866,
+      "learning_rate": 1.5015910443099759e-05,
+      "loss": 0.8275,
       "step": 12432
     },
     {
-      "epoch": 0.34146274477493066,
+      "epoch": 0.35280930760499435,
       "grad_norm": 0.0,
-      "learning_rate": 1.5329517114573914e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.5015115323009423e-05,
+      "loss": 1.0001,
       "step": 12433
     },
     {
-      "epoch": 0.3414902090027739,
+      "epoch": 0.3528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.532876443337913e-05,
-      "loss": 1.0313,
+      "learning_rate": 1.5014320160556182e-05,
+      "loss": 0.8422,
       "step": 12434
     },
     {
-      "epoch": 0.3415176732306171,
+      "epoch": 0.3528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.5328011710021105e-05,
-      "loss": 0.8329,
+      "learning_rate": 1.501352495574675e-05,
+      "loss": 0.9802,
       "step": 12435
     },
     {
-      "epoch": 0.3415451374584604,
+      "epoch": 0.352894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.532725894450579e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.5012729708587852e-05,
+      "loss": 0.9683,
       "step": 12436
     },
     {
-      "epoch": 0.3415726016863036,
+      "epoch": 0.3529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.5326506136839128e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.5011934419086199e-05,
+      "loss": 0.8659,
       "step": 12437
     },
     {
-      "epoch": 0.3416000659141468,
+      "epoch": 0.3529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.532575328702709e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.5011139087248508e-05,
+      "loss": 0.9251,
       "step": 12438
     },
     {
-      "epoch": 0.34162753014199004,
+      "epoch": 0.3529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.532500039507563e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.5010343713081504e-05,
+      "loss": 1.0423,
       "step": 12439
     },
     {
-      "epoch": 0.34165499436983326,
+      "epoch": 0.3530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.5324247460990705e-05,
-      "loss": 0.936,
+      "learning_rate": 1.50095482965919e-05,
+      "loss": 0.9583,
       "step": 12440
     },
     {
-      "epoch": 0.34168245859767654,
+      "epoch": 0.35303632236095345,
       "grad_norm": 0.0,
-      "learning_rate": 1.532349448477827e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.5008752837786413e-05,
+      "loss": 0.9462,
       "step": 12441
     },
     {
-      "epoch": 0.34170992282551976,
+      "epoch": 0.35306469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.532274146644429e-05,
-      "loss": 0.997,
+      "learning_rate": 1.500795733667177e-05,
+      "loss": 1.0008,
       "step": 12442
     },
     {
-      "epoch": 0.341737387053363,
+      "epoch": 0.3530930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.5321988405994713e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.5007161793254686e-05,
+      "loss": 0.9193,
       "step": 12443
     },
     {
-      "epoch": 0.3417648512812062,
+      "epoch": 0.35312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.53212353034355e-05,
-      "loss": 0.925,
+      "learning_rate": 1.5006366207541877e-05,
+      "loss": 1.0104,
       "step": 12444
     },
     {
-      "epoch": 0.3417923155090495,
+      "epoch": 0.35314982973893305,
       "grad_norm": 0.0,
-      "learning_rate": 1.5320482158772617e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.5005570579540073e-05,
+      "loss": 0.8317,
       "step": 12445
     },
     {
-      "epoch": 0.3418197797368927,
+      "epoch": 0.3531782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.5319728972012017e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.5004774909255985e-05,
+      "loss": 0.8672,
       "step": 12446
     },
     {
-      "epoch": 0.3418472439647359,
+      "epoch": 0.3532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.531897574315966e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.5003979196696343e-05,
+      "loss": 0.991,
       "step": 12447
     },
     {
-      "epoch": 0.34187470819257915,
+      "epoch": 0.3532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.5318222472221504e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.5003183441867858e-05,
+      "loss": 0.9101,
       "step": 12448
     },
     {
-      "epoch": 0.3419021724204224,
+      "epoch": 0.3532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5317469159203514e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.5002387644777263e-05,
+      "loss": 1.0438,
       "step": 12449
     },
     {
-      "epoch": 0.34192963664826564,
+      "epoch": 0.3532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.5316715804111646e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.5001591805431272e-05,
+      "loss": 0.9485,
       "step": 12450
     },
     {
-      "epoch": 0.34195710087610887,
+      "epoch": 0.3533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.5315962406951867e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.5000795923836611e-05,
+      "loss": 0.8956,
       "step": 12451
     },
     {
-      "epoch": 0.3419845651039521,
+      "epoch": 0.3533484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.5315208967730132e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.0441,
       "step": 12452
     },
     {
-      "epoch": 0.34201202933179536,
+      "epoch": 0.35337684449489215,
       "grad_norm": 0.0,
-      "learning_rate": 1.5314455486452405e-05,
-      "loss": 1.0152,
+      "learning_rate": 1.499920403392817e-05,
+      "loss": 0.9614,
       "step": 12453
     },
     {
-      "epoch": 0.3420394935596386,
+      "epoch": 0.35340522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 1.531370196312465e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.4998408025627831e-05,
+      "loss": 1.015,
       "step": 12454
     },
     {
-      "epoch": 0.3420669577874818,
+      "epoch": 0.3534335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 1.531294839775282e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.499761197510572e-05,
+      "loss": 0.9275,
       "step": 12455
     },
     {
-      "epoch": 0.34209442201532503,
+      "epoch": 0.35346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.531219479034289e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.4996815882368554e-05,
+      "loss": 0.759,
       "step": 12456
     },
     {
-      "epoch": 0.34212188624316825,
+      "epoch": 0.35349035187287176,
       "grad_norm": 0.0,
-      "learning_rate": 1.5311441140900816e-05,
-      "loss": 1.0737,
+      "learning_rate": 1.499601974742306e-05,
+      "loss": 0.866,
       "step": 12457
     },
     {
-      "epoch": 0.3421493504710115,
+      "epoch": 0.3535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.5310687449432558e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.4995223570275963e-05,
+      "loss": 1.0375,
       "step": 12458
     },
     {
-      "epoch": 0.34217681469885475,
+      "epoch": 0.35354710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 1.5309933715944093e-05,
-      "loss": 0.8907,
+      "learning_rate": 1.4994427350933987e-05,
+      "loss": 0.8018,
       "step": 12459
     },
     {
-      "epoch": 0.34220427892669797,
+      "epoch": 0.3535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.530917994044137e-05,
-      "loss": 1.0076,
+      "learning_rate": 1.499363108940386e-05,
+      "loss": 0.8875,
       "step": 12460
     },
     {
-      "epoch": 0.3422317431545412,
+      "epoch": 0.3536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.5308426122930358e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.4992834785692303e-05,
+      "loss": 0.9244,
       "step": 12461
     },
     {
-      "epoch": 0.34225920738238447,
+      "epoch": 0.3536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.5307672263417024e-05,
-      "loss": 0.8835,
+      "learning_rate": 1.499203843980605e-05,
+      "loss": 1.0135,
       "step": 12462
     },
     {
-      "epoch": 0.3422866716102277,
+      "epoch": 0.3536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.530691836190733e-05,
-      "loss": 0.8814,
+      "learning_rate": 1.4991242051751824e-05,
+      "loss": 0.9139,
       "step": 12463
     },
     {
-      "epoch": 0.3423141358380709,
+      "epoch": 0.353688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.530616441840724e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.4990445621536349e-05,
+      "loss": 1.0035,
       "step": 12464
     },
     {
-      "epoch": 0.34234160006591413,
+      "epoch": 0.35371736662883085,
       "grad_norm": 0.0,
-      "learning_rate": 1.5305410432922725e-05,
-      "loss": 0.922,
+      "learning_rate": 1.4989649149166358e-05,
+      "loss": 1.001,
       "step": 12465
     },
     {
-      "epoch": 0.3423690642937574,
+      "epoch": 0.35374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.5304656405459746e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.4988852634648577e-05,
+      "loss": 0.9185,
       "step": 12466
     },
     {
-      "epoch": 0.34239652852160063,
+      "epoch": 0.3537741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.530390233602427e-05,
-      "loss": 0.8812,
+      "learning_rate": 1.498805607798973e-05,
+      "loss": 0.9672,
       "step": 12467
     },
     {
-      "epoch": 0.34242399274944385,
+      "epoch": 0.35380249716231554,
       "grad_norm": 0.0,
-      "learning_rate": 1.530314822462227e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.4987259479196551e-05,
+      "loss": 0.9369,
       "step": 12468
     },
     {
-      "epoch": 0.34245145697728707,
+      "epoch": 0.35383087400681046,
       "grad_norm": 0.0,
-      "learning_rate": 1.53023940712597e-05,
-      "loss": 0.9899,
+      "learning_rate": 1.4986462838275769e-05,
+      "loss": 0.9721,
       "step": 12469
     },
     {
-      "epoch": 0.3424789212051303,
+      "epoch": 0.3538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.5301639875942537e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.4985666155234109e-05,
+      "loss": 0.9366,
       "step": 12470
     },
     {
-      "epoch": 0.34250638543297357,
+      "epoch": 0.35388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.5300885638676747e-05,
-      "loss": 0.9259,
+      "learning_rate": 1.4984869430078304e-05,
+      "loss": 0.9642,
       "step": 12471
     },
     {
-      "epoch": 0.3425338496608168,
+      "epoch": 0.3539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.530013135946829e-05,
-      "loss": 0.8303,
+      "learning_rate": 1.4984072662815082e-05,
+      "loss": 0.9339,
       "step": 12472
     },
     {
-      "epoch": 0.34256131388866,
+      "epoch": 0.35394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.529937703832315e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.4983275853451176e-05,
+      "loss": 0.8722,
       "step": 12473
     },
     {
-      "epoch": 0.34258877811650323,
+      "epoch": 0.3539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.5298622675247283e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.4982479001993311e-05,
+      "loss": 0.8822,
       "step": 12474
     },
     {
-      "epoch": 0.3426162423443465,
+      "epoch": 0.3540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.5297868270246655e-05,
-      "loss": 0.8763,
+      "learning_rate": 1.4981682108448225e-05,
+      "loss": 1.0213,
       "step": 12475
     },
     {
-      "epoch": 0.34264370657218973,
+      "epoch": 0.3540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.5297113823327248e-05,
-      "loss": 0.9709,
+      "learning_rate": 1.4980885172822648e-05,
+      "loss": 0.9149,
       "step": 12476
     },
     {
-      "epoch": 0.34267117080003295,
+      "epoch": 0.35405788876276956,
       "grad_norm": 0.0,
-      "learning_rate": 1.529635933449502e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.4980088195123306e-05,
+      "loss": 1.017,
       "step": 12477
     },
     {
-      "epoch": 0.3426986350278762,
+      "epoch": 0.35408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.5295604803755948e-05,
-      "loss": 1.0541,
+      "learning_rate": 1.4979291175356934e-05,
+      "loss": 0.8961,
       "step": 12478
     },
     {
-      "epoch": 0.34272609925571945,
+      "epoch": 0.3541146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.5294850231116e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.4978494113530268e-05,
+      "loss": 1.0319,
       "step": 12479
     },
     {
-      "epoch": 0.3427535634835627,
+      "epoch": 0.35414301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.5294095616581147e-05,
-      "loss": 1.0086,
+      "learning_rate": 1.497769700965004e-05,
+      "loss": 0.9322,
       "step": 12480
     },
     {
-      "epoch": 0.3427810277114059,
+      "epoch": 0.35417139614074916,
       "grad_norm": 0.0,
-      "learning_rate": 1.5293340960157356e-05,
-      "loss": 1.0205,
+      "learning_rate": 1.497689986372298e-05,
+      "loss": 0.9074,
       "step": 12481
     },
     {
-      "epoch": 0.3428084919392491,
+      "epoch": 0.354199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.5292586261850598e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.4976102675755824e-05,
+      "loss": 0.8972,
       "step": 12482
     },
     {
-      "epoch": 0.34283595616709234,
+      "epoch": 0.35422814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.529183152166685e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.4975305445755306e-05,
+      "loss": 0.8662,
       "step": 12483
     },
     {
-      "epoch": 0.3428634203949356,
+      "epoch": 0.35425652667423385,
       "grad_norm": 0.0,
-      "learning_rate": 1.5291076739612084e-05,
-      "loss": 1.0225,
+      "learning_rate": 1.4974508173728156e-05,
+      "loss": 0.9613,
       "step": 12484
     },
     {
-      "epoch": 0.34289088462277884,
+      "epoch": 0.3542849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.5290321915692265e-05,
-      "loss": 1.0371,
+      "learning_rate": 1.4973710859681112e-05,
+      "loss": 0.9371,
       "step": 12485
     },
     {
-      "epoch": 0.34291834885062206,
+      "epoch": 0.3543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.5289567049913375e-05,
-      "loss": 1.011,
+      "learning_rate": 1.4972913503620912e-05,
+      "loss": 0.9525,
       "step": 12486
     },
     {
-      "epoch": 0.3429458130784653,
+      "epoch": 0.3543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5288812142281377e-05,
-      "loss": 0.898,
+      "learning_rate": 1.4972116105554287e-05,
+      "loss": 0.9745,
       "step": 12487
     },
     {
-      "epoch": 0.34297327730630855,
+      "epoch": 0.3543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.5288057192802252e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.4971318665487974e-05,
+      "loss": 0.9583,
       "step": 12488
     },
     {
-      "epoch": 0.3430007415341518,
+      "epoch": 0.35439841089670826,
       "grad_norm": 0.0,
-      "learning_rate": 1.5287302201481966e-05,
-      "loss": 0.9562,
+      "learning_rate": 1.4970521183428706e-05,
+      "loss": 0.9039,
       "step": 12489
     },
     {
-      "epoch": 0.343028205761995,
+      "epoch": 0.3544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.52865471683265e-05,
-      "loss": 0.8475,
+      "learning_rate": 1.4969723659383224e-05,
+      "loss": 0.8039,
       "step": 12490
     },
     {
-      "epoch": 0.3430556699898382,
+      "epoch": 0.3544551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.528579209334182e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.4968926093358263e-05,
+      "loss": 0.9363,
       "step": 12491
     },
     {
-      "epoch": 0.3430831342176815,
+      "epoch": 0.35448354143019295,
       "grad_norm": 0.0,
-      "learning_rate": 1.5285036976533912e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.4968128485360558e-05,
+      "loss": 0.9429,
       "step": 12492
     },
     {
-      "epoch": 0.3431105984455247,
+      "epoch": 0.35451191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.5284281817908743e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.4967330835396851e-05,
+      "loss": 0.8681,
       "step": 12493
     },
     {
-      "epoch": 0.34313806267336794,
+      "epoch": 0.3545402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.5283526617472285e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.4966533143473876e-05,
+      "loss": 0.9812,
       "step": 12494
     },
     {
-      "epoch": 0.34316552690121116,
+      "epoch": 0.35456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.5282771375230523e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.4965735409598373e-05,
+      "loss": 0.93,
       "step": 12495
     },
     {
-      "epoch": 0.3431929911290544,
+      "epoch": 0.35459704880817255,
       "grad_norm": 0.0,
-      "learning_rate": 1.5282016091189423e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.4964937633777079e-05,
+      "loss": 1.0078,
       "step": 12496
     },
     {
-      "epoch": 0.34322045535689766,
+      "epoch": 0.3546254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.528126076535497e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.4964139816016737e-05,
+      "loss": 0.8735,
       "step": 12497
     },
     {
-      "epoch": 0.3432479195847409,
+      "epoch": 0.35465380249716233,
       "grad_norm": 0.0,
-      "learning_rate": 1.528050539773313e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.4963341956324077e-05,
+      "loss": 0.9096,
       "step": 12498
     },
     {
-      "epoch": 0.3432753838125841,
+      "epoch": 0.3546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.5279749988329894e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.4962544054705848e-05,
+      "loss": 0.7929,
       "step": 12499
     },
     {
-      "epoch": 0.3433028480404273,
+      "epoch": 0.3547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.5278994537151226e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.4961746111168785e-05,
+      "loss": 0.8577,
       "step": 12500
     },
     {
-      "epoch": 0.3433303122682706,
+      "epoch": 0.354738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.5278239044203108e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.496094812571963e-05,
+      "loss": 0.9965,
       "step": 12501
     },
     {
-      "epoch": 0.3433577764961138,
+      "epoch": 0.3547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.5277483509491518e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.496015009836512e-05,
+      "loss": 0.9391,
       "step": 12502
     },
     {
-      "epoch": 0.34338524072395704,
+      "epoch": 0.3547956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.5276727933022437e-05,
-      "loss": 0.8927,
+      "learning_rate": 1.4959352029112004e-05,
+      "loss": 0.878,
       "step": 12503
     },
     {
-      "epoch": 0.34341270495180026,
+      "epoch": 0.35482406356413165,
       "grad_norm": 0.0,
-      "learning_rate": 1.5275972314801842e-05,
-      "loss": 1.0995,
+      "learning_rate": 1.4958553917967018e-05,
+      "loss": 0.9203,
       "step": 12504
     },
     {
-      "epoch": 0.34344016917964354,
+      "epoch": 0.35485244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 1.5275216654835703e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.4957755764936903e-05,
+      "loss": 0.8997,
       "step": 12505
     },
     {
-      "epoch": 0.34346763340748676,
+      "epoch": 0.35488081725312143,
       "grad_norm": 0.0,
-      "learning_rate": 1.527446095313001e-05,
-      "loss": 1.018,
+      "learning_rate": 1.4956957570028401e-05,
+      "loss": 1.137,
       "step": 12506
     },
     {
-      "epoch": 0.34349509763533,
+      "epoch": 0.35490919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.527370520969074e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.4956159333248258e-05,
+      "loss": 0.8054,
       "step": 12507
     },
     {
-      "epoch": 0.3435225618631732,
+      "epoch": 0.35493757094211126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5272949424523872e-05,
-      "loss": 1.0227,
+      "learning_rate": 1.495536105460321e-05,
+      "loss": 0.8929,
       "step": 12508
     },
     {
-      "epoch": 0.3435500260910164,
+      "epoch": 0.3549659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.5272193597635382e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.4954562734100008e-05,
+      "loss": 0.8938,
       "step": 12509
     },
     {
-      "epoch": 0.3435774903188597,
+      "epoch": 0.35499432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.5271437729031258e-05,
-      "loss": 1.0593,
+      "learning_rate": 1.4953764371745392e-05,
+      "loss": 1.0594,
       "step": 12510
     },
     {
-      "epoch": 0.3436049545467029,
+      "epoch": 0.3550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.5270681818717473e-05,
-      "loss": 0.91,
+      "learning_rate": 1.4952965967546106e-05,
+      "loss": 0.8786,
       "step": 12511
     },
     {
-      "epoch": 0.34363241877454614,
+      "epoch": 0.3550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.5269925866700015e-05,
-      "loss": 0.806,
+      "learning_rate": 1.495216752150889e-05,
+      "loss": 1.0092,
       "step": 12512
     },
     {
-      "epoch": 0.34365988300238937,
+      "epoch": 0.3550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.5269169872984862e-05,
-      "loss": 1.0986,
+      "learning_rate": 1.4951369033640497e-05,
+      "loss": 0.9255,
       "step": 12513
     },
     {
-      "epoch": 0.34368734723023264,
+      "epoch": 0.3551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.5268413837577996e-05,
-      "loss": 0.986,
+      "learning_rate": 1.4950570503947668e-05,
+      "loss": 0.893,
       "step": 12514
     },
     {
-      "epoch": 0.34371481145807586,
+      "epoch": 0.3551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.52676577604854e-05,
-      "loss": 0.9652,
+      "learning_rate": 1.494977193243714e-05,
+      "loss": 0.9891,
       "step": 12515
     },
     {
-      "epoch": 0.3437422756859191,
+      "epoch": 0.35516458569807036,
       "grad_norm": 0.0,
-      "learning_rate": 1.5266901641713052e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.4948973319115671e-05,
+      "loss": 1.0713,
       "step": 12516
     },
     {
-      "epoch": 0.3437697399137623,
+      "epoch": 0.35519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.526614548126694e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.4948174663990002e-05,
+      "loss": 0.9272,
       "step": 12517
     },
     {
-      "epoch": 0.3437972041416056,
+      "epoch": 0.3552213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.526538927915304e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.494737596706688e-05,
+      "loss": 0.9585,
       "step": 12518
     },
     {
-      "epoch": 0.3438246683694488,
+      "epoch": 0.35524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.5264633035377347e-05,
-      "loss": 1.0874,
+      "learning_rate": 1.4946577228353048e-05,
+      "loss": 0.8957,
       "step": 12519
     },
     {
-      "epoch": 0.343852132597292,
+      "epoch": 0.35527809307604996,
       "grad_norm": 0.0,
-      "learning_rate": 1.5263876749945837e-05,
-      "loss": 0.8925,
+      "learning_rate": 1.4945778447855259e-05,
+      "loss": 0.9668,
       "step": 12520
     },
     {
-      "epoch": 0.34387959682513525,
+      "epoch": 0.3553064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.5263120422864493e-05,
-      "loss": 0.9281,
+      "learning_rate": 1.4944979625580253e-05,
+      "loss": 0.9919,
       "step": 12521
     },
     {
-      "epoch": 0.34390706105297847,
+      "epoch": 0.35533484676503974,
       "grad_norm": 0.0,
-      "learning_rate": 1.5262364054139306e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.4944180761534785e-05,
+      "loss": 0.8218,
       "step": 12522
     },
     {
-      "epoch": 0.34393452528082175,
+      "epoch": 0.3553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.5261607643776247e-05,
-      "loss": 1.1175,
+      "learning_rate": 1.4943381855725599e-05,
+      "loss": 0.9418,
       "step": 12523
     },
     {
-      "epoch": 0.34396198950866497,
+      "epoch": 0.3553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.526085119178132e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.4942582908159446e-05,
+      "loss": 0.9661,
       "step": 12524
     },
     {
-      "epoch": 0.3439894537365082,
+      "epoch": 0.3554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.5260094698160493e-05,
-      "loss": 0.8317,
+      "learning_rate": 1.4941783918843069e-05,
+      "loss": 0.9653,
       "step": 12525
     },
     {
-      "epoch": 0.3440169179643514,
+      "epoch": 0.3554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.5259338162919762e-05,
-      "loss": 0.963,
+      "learning_rate": 1.4940984887783226e-05,
+      "loss": 0.9486,
       "step": 12526
     },
     {
-      "epoch": 0.3440443821921947,
+      "epoch": 0.3554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.5258581586065113e-05,
-      "loss": 1.0033,
+      "learning_rate": 1.4940185814986659e-05,
+      "loss": 1.0162,
       "step": 12527
     },
     {
-      "epoch": 0.3440718464200379,
+      "epoch": 0.35550510783200906,
       "grad_norm": 0.0,
-      "learning_rate": 1.5257824967602528e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.493938670046012e-05,
+      "loss": 0.883,
       "step": 12528
     },
     {
-      "epoch": 0.34409931064788113,
+      "epoch": 0.355533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.5257068307537992e-05,
-      "loss": 1.0123,
+      "learning_rate": 1.4938587544210358e-05,
+      "loss": 0.8602,
       "step": 12529
     },
     {
-      "epoch": 0.34412677487572435,
+      "epoch": 0.3555618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.52563116058775e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.4937788346244126e-05,
+      "loss": 0.9087,
       "step": 12530
     },
     {
-      "epoch": 0.3441542391035676,
+      "epoch": 0.35559023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.5255554862627028e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.4936989106568176e-05,
+      "loss": 0.9063,
       "step": 12531
     },
     {
-      "epoch": 0.34418170333141085,
+      "epoch": 0.35561861520998866,
       "grad_norm": 0.0,
-      "learning_rate": 1.5254798077792576e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.4936189825189256e-05,
+      "loss": 0.8483,
       "step": 12532
     },
     {
-      "epoch": 0.34420916755925407,
+      "epoch": 0.3556469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.5254041251380124e-05,
-      "loss": 1.0305,
+      "learning_rate": 1.4935390502114118e-05,
+      "loss": 1.0033,
       "step": 12533
     },
     {
-      "epoch": 0.3442366317870973,
+      "epoch": 0.35567536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.5253284383395663e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.4934591137349514e-05,
+      "loss": 0.9423,
       "step": 12534
     },
     {
-      "epoch": 0.3442640960149405,
+      "epoch": 0.3557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.525252747384518e-05,
-      "loss": 0.8347,
+      "learning_rate": 1.4933791730902199e-05,
+      "loss": 0.9531,
       "step": 12535
     },
     {
-      "epoch": 0.3442915602427838,
+      "epoch": 0.3557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.5251770522734669e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.4932992282778923e-05,
+      "loss": 0.8717,
       "step": 12536
     },
     {
-      "epoch": 0.344319024470627,
+      "epoch": 0.35576049943246313,
       "grad_norm": 0.0,
-      "learning_rate": 1.525101353007011e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.493219279298644e-05,
+      "loss": 1.0458,
       "step": 12537
     },
     {
-      "epoch": 0.34434648869847023,
+      "epoch": 0.355788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.52502564958575e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.4931393261531499e-05,
+      "loss": 1.0298,
       "step": 12538
     },
     {
-      "epoch": 0.34437395292631345,
+      "epoch": 0.3558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.5249499420102827e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.493059368842086e-05,
+      "loss": 0.9688,
       "step": 12539
     },
     {
-      "epoch": 0.34440141715415673,
+      "epoch": 0.35584562996594776,
       "grad_norm": 0.0,
-      "learning_rate": 1.524874230281208e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.4929794073661274e-05,
+      "loss": 0.9772,
       "step": 12540
     },
     {
-      "epoch": 0.34442888138199995,
+      "epoch": 0.3558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.5247985143991253e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.4928994417259496e-05,
+      "loss": 0.8847,
       "step": 12541
     },
     {
-      "epoch": 0.3444563456098432,
+      "epoch": 0.3559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.5247227943646332e-05,
-      "loss": 0.8425,
+      "learning_rate": 1.492819471922228e-05,
+      "loss": 0.9833,
       "step": 12542
     },
     {
-      "epoch": 0.3444838098376864,
+      "epoch": 0.35593076049943245,
       "grad_norm": 0.0,
-      "learning_rate": 1.524647070178331e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.4927394979556384e-05,
+      "loss": 0.8823,
       "step": 12543
     },
     {
-      "epoch": 0.34451127406552967,
+      "epoch": 0.35595913734392737,
       "grad_norm": 0.0,
-      "learning_rate": 1.5245713418408183e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.4926595198268561e-05,
+      "loss": 0.8962,
       "step": 12544
     },
     {
-      "epoch": 0.3445387382933729,
+      "epoch": 0.3559875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.5244956093526942e-05,
-      "loss": 0.8633,
+      "learning_rate": 1.4925795375365564e-05,
+      "loss": 0.7679,
       "step": 12545
     },
     {
-      "epoch": 0.3445662025212161,
+      "epoch": 0.35601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.524419872714557e-05,
-      "loss": 0.9004,
+      "learning_rate": 1.4924995510854153e-05,
+      "loss": 1.0043,
       "step": 12546
     },
     {
-      "epoch": 0.34459366674905934,
+      "epoch": 0.35604426787741206,
       "grad_norm": 0.0,
-      "learning_rate": 1.5243441319270066e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.4924195604741085e-05,
+      "loss": 1.022,
       "step": 12547
     },
     {
-      "epoch": 0.3446211309769026,
+      "epoch": 0.3560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.5242683869906429e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.4923395657033113e-05,
+      "loss": 1.0923,
       "step": 12548
     },
     {
-      "epoch": 0.34464859520474583,
+      "epoch": 0.35610102156640183,
       "grad_norm": 0.0,
-      "learning_rate": 1.5241926379060642e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.4922595667736999e-05,
+      "loss": 0.8751,
       "step": 12549
     },
     {
-      "epoch": 0.34467605943258905,
+      "epoch": 0.3561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.5241168846738705e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.4921795636859497e-05,
+      "loss": 1.0258,
       "step": 12550
     },
     {
-      "epoch": 0.3447035236604323,
+      "epoch": 0.3561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.5240411272946612e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.4920995564407368e-05,
+      "loss": 0.9367,
       "step": 12551
     },
     {
-      "epoch": 0.3447309878882755,
+      "epoch": 0.35618615209988647,
       "grad_norm": 0.0,
-      "learning_rate": 1.5239653657690352e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.4920195450387365e-05,
+      "loss": 0.952,
       "step": 12552
     },
     {
-      "epoch": 0.3447584521161188,
+      "epoch": 0.3562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.5238896000975922e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.491939529480625e-05,
+      "loss": 0.9581,
       "step": 12553
     },
     {
-      "epoch": 0.344785916343962,
+      "epoch": 0.3562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.5238138302809318e-05,
-      "loss": 0.8659,
+      "learning_rate": 1.4918595097670782e-05,
+      "loss": 0.8901,
       "step": 12554
     },
     {
-      "epoch": 0.3448133805718052,
+      "epoch": 0.35627128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 1.5237380563196533e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.4917794858987721e-05,
+      "loss": 0.9492,
       "step": 12555
     },
     {
-      "epoch": 0.34484084479964844,
+      "epoch": 0.35629965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.5236622782143568e-05,
-      "loss": 0.9024,
+      "learning_rate": 1.4916994578763826e-05,
+      "loss": 0.8675,
       "step": 12556
     },
     {
-      "epoch": 0.3448683090274917,
+      "epoch": 0.35632803632236093,
       "grad_norm": 0.0,
-      "learning_rate": 1.5235864959656414e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.4916194257005857e-05,
+      "loss": 0.861,
       "step": 12557
     },
     {
-      "epoch": 0.34489577325533494,
+      "epoch": 0.35635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.523510709574107e-05,
-      "loss": 0.8145,
+      "learning_rate": 1.4915393893720575e-05,
+      "loss": 1.0075,
       "step": 12558
     },
     {
-      "epoch": 0.34492323748317816,
+      "epoch": 0.35638479001135076,
       "grad_norm": 0.0,
-      "learning_rate": 1.5234349190403528e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.4914593488914738e-05,
+      "loss": 0.9234,
       "step": 12559
     },
     {
-      "epoch": 0.3449507017110214,
+      "epoch": 0.3564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.5233591243649788e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.4913793042595109e-05,
+      "loss": 0.8635,
       "step": 12560
     },
     {
-      "epoch": 0.34497816593886466,
+      "epoch": 0.35644154370034054,
       "grad_norm": 0.0,
-      "learning_rate": 1.5232833255485846e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.4912992554768448e-05,
+      "loss": 0.9814,
       "step": 12561
     },
     {
-      "epoch": 0.3450056301667079,
+      "epoch": 0.3564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.5232075225917701e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.4912192025441518e-05,
+      "loss": 0.9714,
       "step": 12562
     },
     {
-      "epoch": 0.3450330943945511,
+      "epoch": 0.3564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.5231317154951351e-05,
-      "loss": 0.922,
+      "learning_rate": 1.4911391454621085e-05,
+      "loss": 0.9393,
       "step": 12563
     },
     {
-      "epoch": 0.3450605586223943,
+      "epoch": 0.3565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.5230559042592793e-05,
-      "loss": 1.0162,
+      "learning_rate": 1.4910590842313908e-05,
+      "loss": 0.9222,
       "step": 12564
     },
     {
-      "epoch": 0.34508802285023754,
+      "epoch": 0.3565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.5229800888848022e-05,
-      "loss": 0.968,
+      "learning_rate": 1.4909790188526747e-05,
+      "loss": 0.9785,
       "step": 12565
     },
     {
-      "epoch": 0.3451154870780808,
+      "epoch": 0.356583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.5229042693723046e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.4908989493266366e-05,
+      "loss": 0.9423,
       "step": 12566
     },
     {
-      "epoch": 0.34514295130592404,
+      "epoch": 0.35661180476730986,
       "grad_norm": 0.0,
-      "learning_rate": 1.5228284457223856e-05,
-      "loss": 1.0241,
+      "learning_rate": 1.4908188756539533e-05,
+      "loss": 0.935,
       "step": 12567
     },
     {
-      "epoch": 0.34517041553376726,
+      "epoch": 0.3566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.5227526179356456e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.4907387978353007e-05,
+      "loss": 0.9653,
       "step": 12568
     },
     {
-      "epoch": 0.3451978797616105,
+      "epoch": 0.35666855845629963,
       "grad_norm": 0.0,
-      "learning_rate": 1.5226767860126843e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.4906587158713552e-05,
+      "loss": 0.9702,
       "step": 12569
     },
     {
-      "epoch": 0.34522534398945376,
+      "epoch": 0.35669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.5226009499541017e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.4905786297627937e-05,
+      "loss": 0.9947,
       "step": 12570
     },
     {
-      "epoch": 0.345252808217297,
+      "epoch": 0.35672531214528946,
       "grad_norm": 0.0,
-      "learning_rate": 1.5225251097604981e-05,
-      "loss": 0.9448,
+      "learning_rate": 1.4904985395102927e-05,
+      "loss": 0.9487,
       "step": 12571
     },
     {
-      "epoch": 0.3452802724451402,
+      "epoch": 0.3567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.5224492654324733e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.490418445114528e-05,
+      "loss": 0.9324,
       "step": 12572
     },
     {
-      "epoch": 0.3453077366729834,
+      "epoch": 0.35678206583427924,
       "grad_norm": 0.0,
-      "learning_rate": 1.5223734169706274e-05,
-      "loss": 1.0376,
+      "learning_rate": 1.4903383465761771e-05,
+      "loss": 0.9723,
       "step": 12573
     },
     {
-      "epoch": 0.3453352009008267,
+      "epoch": 0.3568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.5222975643755609e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.4902582438959161e-05,
+      "loss": 1.0097,
       "step": 12574
     },
     {
-      "epoch": 0.3453626651286699,
+      "epoch": 0.356838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.5222217076478738e-05,
-      "loss": 1.0116,
+      "learning_rate": 1.4901781370744215e-05,
+      "loss": 1.0099,
       "step": 12575
     },
     {
-      "epoch": 0.34539012935651314,
+      "epoch": 0.35686719636776393,
       "grad_norm": 0.0,
-      "learning_rate": 1.522145846788166e-05,
+      "learning_rate": 1.49009802611237e-05,
       "loss": 0.9172,
       "step": 12576
     },
     {
-      "epoch": 0.34541759358435636,
+      "epoch": 0.3568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.5220699817970383e-05,
-      "loss": 1.048,
+      "learning_rate": 1.4900179110104387e-05,
+      "loss": 1.0687,
       "step": 12577
     },
     {
-      "epoch": 0.3454450578121996,
+      "epoch": 0.3569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.5219941126750905e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.4899377917693041e-05,
+      "loss": 1.0618,
       "step": 12578
     },
     {
-      "epoch": 0.34547252204004286,
+      "epoch": 0.35695232690124856,
       "grad_norm": 0.0,
-      "learning_rate": 1.5219182394229228e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.4898576683896427e-05,
+      "loss": 0.9304,
       "step": 12579
     },
     {
-      "epoch": 0.3454999862678861,
+      "epoch": 0.3569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.5218423620411367e-05,
-      "loss": 1.0626,
+      "learning_rate": 1.4897775408721318e-05,
+      "loss": 0.8425,
       "step": 12580
     },
     {
-      "epoch": 0.3455274504957293,
+      "epoch": 0.3570090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.521766480530331e-05,
-      "loss": 0.9205,
+      "learning_rate": 1.4896974092174481e-05,
+      "loss": 1.0195,
       "step": 12581
     },
     {
-      "epoch": 0.3455549147235725,
+      "epoch": 0.35703745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.5216905948911067e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.4896172734262679e-05,
+      "loss": 0.842,
       "step": 12582
     },
     {
-      "epoch": 0.3455823789514158,
+      "epoch": 0.35706583427922817,
       "grad_norm": 0.0,
-      "learning_rate": 1.5216147051240647e-05,
-      "loss": 1.0098,
+      "learning_rate": 1.489537133499269e-05,
+      "loss": 0.8395,
       "step": 12583
     },
     {
-      "epoch": 0.345609843179259,
+      "epoch": 0.357094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.5215388112298048e-05,
-      "loss": 0.9039,
+      "learning_rate": 1.4894569894371276e-05,
+      "loss": 0.9179,
       "step": 12584
     },
     {
-      "epoch": 0.34563730740710225,
+      "epoch": 0.35712258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.5214629132089278e-05,
-      "loss": 1.088,
+      "learning_rate": 1.4893768412405214e-05,
+      "loss": 0.8899,
       "step": 12585
     },
     {
-      "epoch": 0.34566477163494547,
+      "epoch": 0.3571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.5213870110620346e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.4892966889101265e-05,
+      "loss": 0.9915,
       "step": 12586
     },
     {
-      "epoch": 0.34569223586278874,
+      "epoch": 0.3571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.521311104789725e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.489216532446621e-05,
+      "loss": 0.9284,
       "step": 12587
     },
     {
-      "epoch": 0.34571970009063197,
+      "epoch": 0.35720771850170263,
       "grad_norm": 0.0,
-      "learning_rate": 1.5212351943926003e-05,
-      "loss": 0.9583,
+      "learning_rate": 1.4891363718506812e-05,
+      "loss": 0.8703,
       "step": 12588
     },
     {
-      "epoch": 0.3457471643184752,
+      "epoch": 0.3572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.521159279871261e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.4890562071229844e-05,
+      "loss": 0.979,
       "step": 12589
     },
     {
-      "epoch": 0.3457746285463184,
+      "epoch": 0.3572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.5210833612263072e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.488976038264208e-05,
+      "loss": 1.1383,
       "step": 12590
     },
     {
-      "epoch": 0.34580209277416163,
+      "epoch": 0.35729284903518727,
       "grad_norm": 0.0,
-      "learning_rate": 1.52100743845834e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.4888958652750292e-05,
+      "loss": 0.8967,
       "step": 12591
     },
     {
-      "epoch": 0.3458295570020049,
+      "epoch": 0.3573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.5209315115679604e-05,
-      "loss": 1.0557,
+      "learning_rate": 1.4888156881561246e-05,
+      "loss": 1.0821,
       "step": 12592
     },
     {
-      "epoch": 0.3458570212298481,
+      "epoch": 0.3573496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.5208555805557685e-05,
-      "loss": 0.7975,
+      "learning_rate": 1.4887355069081721e-05,
+      "loss": 0.9548,
       "step": 12593
     },
     {
-      "epoch": 0.34588448545769135,
+      "epoch": 0.35737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.5207796454223655e-05,
-      "loss": 0.929,
+      "learning_rate": 1.488655321531849e-05,
+      "loss": 0.8625,
       "step": 12594
     },
     {
-      "epoch": 0.34591194968553457,
+      "epoch": 0.35740635641316687,
       "grad_norm": 0.0,
-      "learning_rate": 1.5207037061683524e-05,
-      "loss": 0.971,
+      "learning_rate": 1.4885751320278323e-05,
+      "loss": 0.9667,
       "step": 12595
     },
     {
-      "epoch": 0.34593941391337785,
+      "epoch": 0.35743473325766173,
       "grad_norm": 0.0,
-      "learning_rate": 1.5206277627943302e-05,
-      "loss": 0.963,
+      "learning_rate": 1.4884949383967993e-05,
+      "loss": 0.8236,
       "step": 12596
     },
     {
-      "epoch": 0.34596687814122107,
+      "epoch": 0.35746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.5205518153008992e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.4884147406394277e-05,
+      "loss": 0.8411,
       "step": 12597
     },
     {
-      "epoch": 0.3459943423690643,
+      "epoch": 0.35749148694665156,
       "grad_norm": 0.0,
-      "learning_rate": 1.5204758636886605e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.4883345387563952e-05,
+      "loss": 0.8875,
       "step": 12598
     },
     {
-      "epoch": 0.3460218065969075,
+      "epoch": 0.3575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.5203999079582152e-05,
-      "loss": 0.8436,
+      "learning_rate": 1.4882543327483784e-05,
+      "loss": 1.0917,
       "step": 12599
     },
     {
-      "epoch": 0.3460492708247508,
+      "epoch": 0.35754824063564133,
       "grad_norm": 0.0,
-      "learning_rate": 1.5203239481101641e-05,
-      "loss": 0.8602,
+      "learning_rate": 1.4881741226160556e-05,
+      "loss": 0.8853,
       "step": 12600
     },
     {
-      "epoch": 0.346076735052594,
+      "epoch": 0.3575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.5202479841451088e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.4880939083601037e-05,
+      "loss": 0.9456,
       "step": 12601
     },
     {
-      "epoch": 0.34610419928043723,
+      "epoch": 0.3576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.5201720160636497e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.4880136899812013e-05,
+      "loss": 0.8391,
       "step": 12602
     },
     {
-      "epoch": 0.34613166350828045,
+      "epoch": 0.35763337116912597,
       "grad_norm": 0.0,
-      "learning_rate": 1.5200960438663879e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.4879334674800246e-05,
+      "loss": 1.0014,
       "step": 12603
     },
     {
-      "epoch": 0.3461591277361237,
+      "epoch": 0.3576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.5200200675539249e-05,
-      "loss": 0.991,
+      "learning_rate": 1.4878532408572523e-05,
+      "loss": 0.8521,
       "step": 12604
     },
     {
-      "epoch": 0.34618659196396695,
+      "epoch": 0.3576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.5199440871268618e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.4877730101135619e-05,
+      "loss": 0.8228,
       "step": 12605
     },
     {
-      "epoch": 0.34621405619181017,
+      "epoch": 0.35771850170261066,
       "grad_norm": 0.0,
-      "learning_rate": 1.5198681025857993e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.4876927752496303e-05,
+      "loss": 0.9796,
       "step": 12606
     },
     {
-      "epoch": 0.3462415204196534,
+      "epoch": 0.3577468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.5197921139313394e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.4876125362661364e-05,
+      "loss": 1.0476,
       "step": 12607
     },
     {
-      "epoch": 0.3462689846474966,
+      "epoch": 0.35777525539160043,
       "grad_norm": 0.0,
-      "learning_rate": 1.519716121164083e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.4875322931637574e-05,
+      "loss": 1.0331,
       "step": 12608
     },
     {
-      "epoch": 0.3462964488753399,
+      "epoch": 0.35780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.519640124284631e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.4874520459431713e-05,
+      "loss": 0.8751,
       "step": 12609
     },
     {
-      "epoch": 0.3463239131031831,
+      "epoch": 0.35783200908059026,
       "grad_norm": 0.0,
-      "learning_rate": 1.5195641232935852e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.487371794605056e-05,
+      "loss": 0.8454,
       "step": 12610
     },
     {
-      "epoch": 0.34635137733102633,
+      "epoch": 0.3578603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.519488118191547e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.487291539150089e-05,
+      "loss": 0.8894,
       "step": 12611
     },
     {
-      "epoch": 0.34637884155886955,
+      "epoch": 0.35788876276958004,
       "grad_norm": 0.0,
-      "learning_rate": 1.5194121089791174e-05,
-      "loss": 0.9107,
+      "learning_rate": 1.4872112795789486e-05,
+      "loss": 0.965,
       "step": 12612
     },
     {
-      "epoch": 0.34640630578671283,
+      "epoch": 0.3579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.519336095656898e-05,
-      "loss": 0.907,
+      "learning_rate": 1.4871310158923122e-05,
+      "loss": 0.9837,
       "step": 12613
     },
     {
-      "epoch": 0.34643377001455605,
+      "epoch": 0.3579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.5192600782254899e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.4870507480908587e-05,
+      "loss": 1.0385,
       "step": 12614
     },
     {
-      "epoch": 0.3464612342423993,
+      "epoch": 0.35797389330306467,
       "grad_norm": 0.0,
-      "learning_rate": 1.5191840566854952e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.4869704761752654e-05,
+      "loss": 0.8383,
       "step": 12615
     },
     {
-      "epoch": 0.3464886984702425,
+      "epoch": 0.3580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.5191080310375155e-05,
-      "loss": 1.0596,
+      "learning_rate": 1.4868902001462107e-05,
+      "loss": 0.7949,
       "step": 12616
     },
     {
-      "epoch": 0.3465161626980857,
+      "epoch": 0.3580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.5190320012821516e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.4868099200043723e-05,
+      "loss": 0.914,
       "step": 12617
     },
     {
-      "epoch": 0.346543626925929,
+      "epoch": 0.35805902383654936,
       "grad_norm": 0.0,
-      "learning_rate": 1.5189559674200053e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.4867296357504292e-05,
+      "loss": 0.9231,
       "step": 12618
     },
     {
-      "epoch": 0.3465710911537722,
+      "epoch": 0.3580874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.5188799294516784e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.4866493473850586e-05,
+      "loss": 1.0289,
       "step": 12619
     },
     {
-      "epoch": 0.34659855538161544,
+      "epoch": 0.35811577752553914,
       "grad_norm": 0.0,
-      "learning_rate": 1.5188038873777728e-05,
-      "loss": 1.0812,
+      "learning_rate": 1.4865690549089391e-05,
+      "loss": 0.91,
       "step": 12620
     },
     {
-      "epoch": 0.34662601960945866,
+      "epoch": 0.35814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.5187278411988895e-05,
-      "loss": 0.8349,
+      "learning_rate": 1.4864887583227492e-05,
+      "loss": 0.9669,
       "step": 12621
     },
     {
-      "epoch": 0.34665348383730193,
+      "epoch": 0.35817253121452897,
       "grad_norm": 0.0,
-      "learning_rate": 1.5186517909156311e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.4864084576271666e-05,
+      "loss": 0.8911,
       "step": 12622
     },
     {
-      "epoch": 0.34668094806514516,
+      "epoch": 0.3582009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.5185757365285984e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.4863281528228698e-05,
+      "loss": 0.886,
       "step": 12623
     },
     {
-      "epoch": 0.3467084122929884,
+      "epoch": 0.35822928490351874,
       "grad_norm": 0.0,
-      "learning_rate": 1.5184996780383938e-05,
-      "loss": 1.0133,
+      "learning_rate": 1.4862478439105372e-05,
+      "loss": 1.0027,
       "step": 12624
     },
     {
-      "epoch": 0.3467358765208316,
+      "epoch": 0.3582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.5184236154456188e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.4861675308908476e-05,
+      "loss": 0.9277,
       "step": 12625
     },
     {
-      "epoch": 0.3467633407486749,
+      "epoch": 0.3582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.5183475487508754e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.4860872137644786e-05,
+      "loss": 0.9298,
       "step": 12626
     },
     {
-      "epoch": 0.3467908049765181,
+      "epoch": 0.35831441543700343,
       "grad_norm": 0.0,
-      "learning_rate": 1.5182714779547652e-05,
-      "loss": 1.0082,
+      "learning_rate": 1.4860068925321093e-05,
+      "loss": 0.9534,
       "step": 12627
     },
     {
-      "epoch": 0.3468182692043613,
+      "epoch": 0.3583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.5181954030578906e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.485926567194418e-05,
+      "loss": 0.9191,
       "step": 12628
     },
     {
-      "epoch": 0.34684573343220454,
+      "epoch": 0.3583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.518119324060853e-05,
-      "loss": 0.9898,
+      "learning_rate": 1.485846237752083e-05,
+      "loss": 0.9306,
       "step": 12629
     },
     {
-      "epoch": 0.34687319766004776,
+      "epoch": 0.35839954597048806,
       "grad_norm": 0.0,
-      "learning_rate": 1.5180432409642547e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.4857659042057828e-05,
+      "loss": 0.8779,
       "step": 12630
     },
     {
-      "epoch": 0.34690066188789104,
+      "epoch": 0.358427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.5179671537686976e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.4856855665561964e-05,
+      "loss": 0.9864,
       "step": 12631
     },
     {
-      "epoch": 0.34692812611573426,
+      "epoch": 0.35845629965947784,
       "grad_norm": 0.0,
-      "learning_rate": 1.5178910624747838e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.4856052248040023e-05,
+      "loss": 0.9386,
       "step": 12632
     },
     {
-      "epoch": 0.3469555903435775,
+      "epoch": 0.35848467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.5178149670831153e-05,
-      "loss": 1.0584,
+      "learning_rate": 1.485524878949879e-05,
+      "loss": 0.87,
       "step": 12633
     },
     {
-      "epoch": 0.3469830545714207,
+      "epoch": 0.35851305334846767,
       "grad_norm": 0.0,
-      "learning_rate": 1.5177388675942943e-05,
-      "loss": 1.0828,
+      "learning_rate": 1.4854445289945049e-05,
+      "loss": 0.9501,
       "step": 12634
     },
     {
-      "epoch": 0.347010518799264,
+      "epoch": 0.35854143019296253,
       "grad_norm": 0.0,
-      "learning_rate": 1.5176627640089226e-05,
-      "loss": 1.014,
+      "learning_rate": 1.4853641749385595e-05,
+      "loss": 0.8722,
       "step": 12635
     },
     {
-      "epoch": 0.3470379830271072,
+      "epoch": 0.35856980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.5175866563276027e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.4852838167827209e-05,
+      "loss": 0.9913,
       "step": 12636
     },
     {
-      "epoch": 0.3470654472549504,
+      "epoch": 0.3585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.5175105445509366e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.4852034545276681e-05,
+      "loss": 0.9714,
       "step": 12637
     },
     {
-      "epoch": 0.34709291148279364,
+      "epoch": 0.3586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.5174344286795266e-05,
-      "loss": 0.9516,
+      "learning_rate": 1.4851230881740796e-05,
+      "loss": 0.9552,
       "step": 12638
     },
     {
-      "epoch": 0.3471203757106369,
+      "epoch": 0.35865493757094213,
       "grad_norm": 0.0,
-      "learning_rate": 1.5173583087139751e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.485042717722635e-05,
+      "loss": 0.9641,
       "step": 12639
     },
     {
-      "epoch": 0.34714783993848014,
+      "epoch": 0.358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.5172821846548841e-05,
-      "loss": 0.8977,
+      "learning_rate": 1.4849623431740126e-05,
+      "loss": 1.0843,
       "step": 12640
     },
     {
-      "epoch": 0.34717530416632336,
+      "epoch": 0.3587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.5172060565028563e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.4848819645288915e-05,
+      "loss": 0.9934,
       "step": 12641
     },
     {
-      "epoch": 0.3472027683941666,
+      "epoch": 0.35874006810442677,
       "grad_norm": 0.0,
-      "learning_rate": 1.5171299242584938e-05,
-      "loss": 1.032,
+      "learning_rate": 1.4848015817879507e-05,
+      "loss": 1.0449,
       "step": 12642
     },
     {
-      "epoch": 0.3472302326220098,
+      "epoch": 0.3587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.5170537879223988e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.484721194951869e-05,
+      "loss": 0.8673,
       "step": 12643
     },
     {
-      "epoch": 0.3472576968498531,
+      "epoch": 0.3587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.5169776474951742e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.4846408040213256e-05,
+      "loss": 0.8844,
       "step": 12644
     },
     {
-      "epoch": 0.3472851610776963,
+      "epoch": 0.35882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.516901502977422e-05,
-      "loss": 1.086,
+      "learning_rate": 1.4845604089969996e-05,
+      "loss": 0.7623,
       "step": 12645
     },
     {
-      "epoch": 0.3473126253055395,
+      "epoch": 0.3588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.5168253543697448e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.48448000987957e-05,
+      "loss": 0.9403,
       "step": 12646
     },
     {
-      "epoch": 0.34734008953338275,
+      "epoch": 0.35888195232690123,
       "grad_norm": 0.0,
-      "learning_rate": 1.5167492016727452e-05,
-      "loss": 0.9299,
+      "learning_rate": 1.484399606669716e-05,
+      "loss": 0.9967,
       "step": 12647
     },
     {
-      "epoch": 0.347367553761226,
+      "epoch": 0.35891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.5166730448870262e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.4843191993681167e-05,
+      "loss": 0.9617,
       "step": 12648
     },
     {
-      "epoch": 0.34739501798906924,
+      "epoch": 0.358938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.5165968840131896e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.4842387879754516e-05,
+      "loss": 0.9159,
       "step": 12649
     },
     {
-      "epoch": 0.34742248221691246,
+      "epoch": 0.3589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.5165207190518383e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.4841583724923994e-05,
+      "loss": 0.9645,
       "step": 12650
     },
     {
-      "epoch": 0.3474499464447557,
+      "epoch": 0.35899545970488084,
       "grad_norm": 0.0,
-      "learning_rate": 1.5164445500035751e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.4840779529196398e-05,
+      "loss": 0.8674,
       "step": 12651
     },
     {
-      "epoch": 0.34747741067259896,
+      "epoch": 0.3590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.5163683768690027e-05,
-      "loss": 0.9542,
+      "learning_rate": 1.483997529257852e-05,
+      "loss": 0.9907,
       "step": 12652
     },
     {
-      "epoch": 0.3475048749004422,
+      "epoch": 0.3590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.5162921996487234e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.4839171015077152e-05,
+      "loss": 0.8963,
       "step": 12653
     },
     {
-      "epoch": 0.3475323391282854,
+      "epoch": 0.35908059023836547,
       "grad_norm": 0.0,
-      "learning_rate": 1.5162160183433403e-05,
-      "loss": 1.0528,
+      "learning_rate": 1.4838366696699087e-05,
+      "loss": 0.8049,
       "step": 12654
     },
     {
-      "epoch": 0.3475598033561286,
+      "epoch": 0.3591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.516139832953456e-05,
-      "loss": 0.903,
+      "learning_rate": 1.4837562337451124e-05,
+      "loss": 0.8615,
       "step": 12655
     },
     {
-      "epoch": 0.3475872675839719,
+      "epoch": 0.3591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.5160636434796735e-05,
-      "loss": 0.9699,
+      "learning_rate": 1.4836757937340053e-05,
+      "loss": 1.005,
       "step": 12656
     },
     {
-      "epoch": 0.3476147318118151,
+      "epoch": 0.35916572077185016,
       "grad_norm": 0.0,
-      "learning_rate": 1.5159874499225956e-05,
-      "loss": 0.9441,
+      "learning_rate": 1.483595349637267e-05,
+      "loss": 0.9838,
       "step": 12657
     },
     {
-      "epoch": 0.34764219603965835,
+      "epoch": 0.3591940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.515911252282825e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.4835149014555774e-05,
+      "loss": 0.9541,
       "step": 12658
     },
     {
-      "epoch": 0.34766966026750157,
+      "epoch": 0.35922247446083994,
       "grad_norm": 0.0,
-      "learning_rate": 1.5158350505609648e-05,
-      "loss": 0.9588,
+      "learning_rate": 1.4834344491896156e-05,
+      "loss": 0.937,
       "step": 12659
     },
     {
-      "epoch": 0.3476971244953448,
+      "epoch": 0.35925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.5157588447576181e-05,
-      "loss": 0.8735,
+      "learning_rate": 1.4833539928400609e-05,
+      "loss": 0.8156,
       "step": 12660
     },
     {
-      "epoch": 0.34772458872318807,
+      "epoch": 0.35927922814982977,
       "grad_norm": 0.0,
-      "learning_rate": 1.5156826348733872e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.4832735324075932e-05,
+      "loss": 0.8465,
       "step": 12661
     },
     {
-      "epoch": 0.3477520529510313,
+      "epoch": 0.3593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.5156064209088759e-05,
-      "loss": 0.936,
+      "learning_rate": 1.4831930678928928e-05,
+      "loss": 0.9215,
       "step": 12662
     },
     {
-      "epoch": 0.3477795171788745,
+      "epoch": 0.35933598183881954,
       "grad_norm": 0.0,
-      "learning_rate": 1.5155302028646867e-05,
-      "loss": 1.0149,
+      "learning_rate": 1.4831125992966386e-05,
+      "loss": 0.9553,
       "step": 12663
     },
     {
-      "epoch": 0.34780698140671773,
+      "epoch": 0.3593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.515453980741423e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.4830321266195105e-05,
+      "loss": 0.9613,
       "step": 12664
     },
     {
-      "epoch": 0.347834445634561,
+      "epoch": 0.3593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.5153777545396877e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.4829516498621885e-05,
+      "loss": 0.9723,
       "step": 12665
     },
     {
-      "epoch": 0.34786190986240423,
+      "epoch": 0.3594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.5153015242600839e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.4828711690253521e-05,
+      "loss": 0.9027,
       "step": 12666
     },
     {
-      "epoch": 0.34788937409024745,
+      "epoch": 0.3594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.5152252899032145e-05,
-      "loss": 0.944,
+      "learning_rate": 1.482790684109681e-05,
+      "loss": 0.9325,
       "step": 12667
     },
     {
-      "epoch": 0.34791683831809067,
+      "epoch": 0.359477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.5151490514696834e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.4827101951158555e-05,
+      "loss": 0.9546,
       "step": 12668
     },
     {
-      "epoch": 0.34794430254593395,
+      "epoch": 0.35950624290578886,
       "grad_norm": 0.0,
-      "learning_rate": 1.5150728089600936e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.4826297020445553e-05,
+      "loss": 0.9765,
       "step": 12669
     },
     {
-      "epoch": 0.34797176677377717,
+      "epoch": 0.3595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.514996562375048e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.4825492048964603e-05,
+      "loss": 0.8887,
       "step": 12670
     },
     {
-      "epoch": 0.3479992310016204,
+      "epoch": 0.35956299659477864,
       "grad_norm": 0.0,
-      "learning_rate": 1.51492031171515e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.4824687036722505e-05,
+      "loss": 0.8816,
       "step": 12671
     },
     {
-      "epoch": 0.3480266952294636,
+      "epoch": 0.35959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.5148440569810033e-05,
-      "loss": 0.9481,
+      "learning_rate": 1.4823881983726059e-05,
+      "loss": 1.0031,
       "step": 12672
     },
     {
-      "epoch": 0.34805415945730683,
+      "epoch": 0.35961975028376847,
       "grad_norm": 0.0,
-      "learning_rate": 1.5147677981732107e-05,
-      "loss": 0.9578,
+      "learning_rate": 1.4823076889982064e-05,
+      "loss": 0.9384,
       "step": 12673
     },
     {
-      "epoch": 0.3480816236851501,
+      "epoch": 0.35964812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.5146915352923758e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.4822271755497322e-05,
+      "loss": 1.0017,
       "step": 12674
     },
     {
-      "epoch": 0.34810908791299333,
+      "epoch": 0.35967650397275824,
       "grad_norm": 0.0,
-      "learning_rate": 1.5146152683391026e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.4821466580278634e-05,
+      "loss": 0.9819,
       "step": 12675
     },
     {
-      "epoch": 0.34813655214083655,
+      "epoch": 0.3597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.5145389973139936e-05,
-      "loss": 0.8773,
+      "learning_rate": 1.48206613643328e-05,
+      "loss": 0.9796,
       "step": 12676
     },
     {
-      "epoch": 0.3481640163686798,
+      "epoch": 0.359733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.5144627222176528e-05,
-      "loss": 0.8877,
+      "learning_rate": 1.4819856107666622e-05,
+      "loss": 0.8973,
       "step": 12677
     },
     {
-      "epoch": 0.34819148059652305,
+      "epoch": 0.35976163450624293,
       "grad_norm": 0.0,
-      "learning_rate": 1.514386443050684e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.4819050810286903e-05,
+      "loss": 0.9555,
       "step": 12678
     },
     {
-      "epoch": 0.34821894482436627,
+      "epoch": 0.3597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.51431015981369e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.4818245472200449e-05,
+      "loss": 0.9017,
       "step": 12679
     },
     {
-      "epoch": 0.3482464090522095,
+      "epoch": 0.3598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.5142338725072752e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.4817440093414055e-05,
+      "loss": 0.9258,
       "step": 12680
     },
     {
-      "epoch": 0.3482738732800527,
+      "epoch": 0.35984676503972757,
       "grad_norm": 0.0,
-      "learning_rate": 1.5141575811320426e-05,
-      "loss": 1.025,
+      "learning_rate": 1.4816634673934532e-05,
+      "loss": 0.9889,
       "step": 12681
     },
     {
-      "epoch": 0.348301337507896,
+      "epoch": 0.3598751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.514081285688596e-05,
-      "loss": 1.0734,
+      "learning_rate": 1.4815829213768675e-05,
+      "loss": 0.8776,
       "step": 12682
     },
     {
-      "epoch": 0.3483288017357392,
+      "epoch": 0.35990351872871734,
       "grad_norm": 0.0,
-      "learning_rate": 1.514004986177539e-05,
-      "loss": 0.9178,
+      "learning_rate": 1.4815023712923296e-05,
+      "loss": 0.926,
       "step": 12683
     },
     {
-      "epoch": 0.34835626596358243,
+      "epoch": 0.35993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.5139286825994757e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.481421817140519e-05,
+      "loss": 0.913,
       "step": 12684
     },
     {
-      "epoch": 0.34838373019142566,
+      "epoch": 0.35996027241770717,
       "grad_norm": 0.0,
-      "learning_rate": 1.5138523749550094e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.481341258922117e-05,
+      "loss": 0.9346,
       "step": 12685
     },
     {
-      "epoch": 0.3484111944192689,
+      "epoch": 0.35998864926220203,
       "grad_norm": 0.0,
-      "learning_rate": 1.5137760632447443e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.4812606966378039e-05,
+      "loss": 0.9121,
       "step": 12686
     },
     {
-      "epoch": 0.34843865864711215,
+      "epoch": 0.36001702610669695,
       "grad_norm": 0.0,
-      "learning_rate": 1.5136997474692838e-05,
-      "loss": 0.9114,
+      "learning_rate": 1.4811801302882596e-05,
+      "loss": 0.9297,
       "step": 12687
     },
     {
-      "epoch": 0.3484661228749554,
+      "epoch": 0.3600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.513623427629232e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.4810995598741655e-05,
+      "loss": 1.0428,
       "step": 12688
     },
     {
-      "epoch": 0.3484935871027986,
+      "epoch": 0.3600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.5135471037251925e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.4810189853962018e-05,
+      "loss": 0.8886,
       "step": 12689
     },
     {
-      "epoch": 0.3485210513306418,
+      "epoch": 0.36010215664018164,
       "grad_norm": 0.0,
-      "learning_rate": 1.5134707757577696e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.480938406855049e-05,
+      "loss": 1.0159,
       "step": 12690
     },
     {
-      "epoch": 0.3485485155584851,
+      "epoch": 0.3601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.5133944437275668e-05,
-      "loss": 0.986,
+      "learning_rate": 1.4808578242513878e-05,
+      "loss": 1.0598,
       "step": 12691
     },
     {
-      "epoch": 0.3485759797863283,
+      "epoch": 0.3601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5133181076351884e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.4807772375858989e-05,
+      "loss": 1.0892,
       "step": 12692
     },
     {
-      "epoch": 0.34860344401417154,
+      "epoch": 0.36018728717366627,
       "grad_norm": 0.0,
-      "learning_rate": 1.5132417674812385e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.4806966468592632e-05,
+      "loss": 0.9434,
       "step": 12693
     },
     {
-      "epoch": 0.34863090824201476,
+      "epoch": 0.3602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.5131654232663205e-05,
-      "loss": 0.9968,
+      "learning_rate": 1.480616052072161e-05,
+      "loss": 0.8591,
       "step": 12694
     },
     {
-      "epoch": 0.34865837246985804,
+      "epoch": 0.36024404086265605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5130890749910391e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.480535453225274e-05,
+      "loss": 0.8547,
       "step": 12695
     },
     {
-      "epoch": 0.34868583669770126,
+      "epoch": 0.36027241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.5130127226559985e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.4804548503192821e-05,
+      "loss": 0.9592,
       "step": 12696
     },
     {
-      "epoch": 0.3487133009255445,
+      "epoch": 0.3603007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.512936366261802e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.4803742433548664e-05,
+      "loss": 0.9983,
       "step": 12697
     },
     {
-      "epoch": 0.3487407651533877,
+      "epoch": 0.36032917139614073,
       "grad_norm": 0.0,
-      "learning_rate": 1.5128600058090546e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.4802936323327079e-05,
+      "loss": 0.7476,
       "step": 12698
     },
     {
-      "epoch": 0.3487682293812309,
+      "epoch": 0.36035754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 1.5127836412983598e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.4802130172534876e-05,
+      "loss": 0.8205,
       "step": 12699
     },
     {
-      "epoch": 0.3487956936090742,
+      "epoch": 0.3603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.5127072727303227e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.4801323981178863e-05,
+      "loss": 0.8869,
       "step": 12700
     },
     {
-      "epoch": 0.3488231578369174,
+      "epoch": 0.3604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.5126309001055468e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.4800517749265848e-05,
+      "loss": 0.9827,
       "step": 12701
     },
     {
-      "epoch": 0.34885062206476064,
+      "epoch": 0.36044267877412034,
       "grad_norm": 0.0,
-      "learning_rate": 1.5125545234246366e-05,
-      "loss": 0.9909,
+      "learning_rate": 1.4799711476802649e-05,
+      "loss": 0.9837,
       "step": 12702
     },
     {
-      "epoch": 0.34887808629260386,
+      "epoch": 0.3604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.5124781426881962e-05,
-      "loss": 0.9441,
+      "learning_rate": 1.4798905163796069e-05,
+      "loss": 0.9696,
       "step": 12703
     },
     {
-      "epoch": 0.34890555052044714,
+      "epoch": 0.3604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.5124017578968312e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.479809881025292e-05,
+      "loss": 0.984,
       "step": 12704
     },
     {
-      "epoch": 0.34893301474829036,
+      "epoch": 0.360527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.5123253690511441e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.4797292416180016e-05,
+      "loss": 0.8744,
       "step": 12705
     },
     {
-      "epoch": 0.3489604789761336,
+      "epoch": 0.3605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.5122489761517405e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.4796485981584169e-05,
+      "loss": 0.9288,
       "step": 12706
     },
     {
-      "epoch": 0.3489879432039768,
+      "epoch": 0.3605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.5121725791992243e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.4795679506472184e-05,
+      "loss": 0.8521,
       "step": 12707
     },
     {
-      "epoch": 0.3490154074318201,
+      "epoch": 0.36061293984108966,
       "grad_norm": 0.0,
-      "learning_rate": 1.5120961781942007e-05,
-      "loss": 1.0384,
+      "learning_rate": 1.4794872990850881e-05,
+      "loss": 0.9924,
       "step": 12708
     },
     {
-      "epoch": 0.3490428716596633,
+      "epoch": 0.3606413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.5120197731372734e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.4794066434727073e-05,
+      "loss": 1.0455,
       "step": 12709
     },
     {
-      "epoch": 0.3490703358875065,
+      "epoch": 0.36066969353007944,
       "grad_norm": 0.0,
-      "learning_rate": 1.5119433640290477e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.479325983810757e-05,
+      "loss": 0.9704,
       "step": 12710
     },
     {
-      "epoch": 0.34909780011534974,
+      "epoch": 0.36069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.5118669508701276e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.4792453200999185e-05,
+      "loss": 0.9236,
       "step": 12711
     },
     {
-      "epoch": 0.34912526434319296,
+      "epoch": 0.3607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.5117905336611178e-05,
-      "loss": 0.964,
+      "learning_rate": 1.479164652340873e-05,
+      "loss": 1.0108,
       "step": 12712
     },
     {
-      "epoch": 0.34915272857103624,
+      "epoch": 0.36075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.5117141124026234e-05,
-      "loss": 0.9344,
+      "learning_rate": 1.4790839805343025e-05,
+      "loss": 0.9361,
       "step": 12713
     },
     {
-      "epoch": 0.34918019279887946,
+      "epoch": 0.36078320090805904,
       "grad_norm": 0.0,
-      "learning_rate": 1.5116376870952483e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.4790033046808878e-05,
+      "loss": 0.9872,
       "step": 12714
     },
     {
-      "epoch": 0.3492076570267227,
+      "epoch": 0.3608115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.5115612577395978e-05,
-      "loss": 0.8608,
+      "learning_rate": 1.4789226247813104e-05,
+      "loss": 0.9348,
       "step": 12715
     },
     {
-      "epoch": 0.3492351212545659,
+      "epoch": 0.3608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.5114848243362765e-05,
-      "loss": 0.9394,
+      "learning_rate": 1.4788419408362527e-05,
+      "loss": 0.8743,
       "step": 12716
     },
     {
-      "epoch": 0.3492625854824092,
+      "epoch": 0.3608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.511408386885889e-05,
-      "loss": 0.9859,
+      "learning_rate": 1.4787612528463952e-05,
+      "loss": 0.891,
       "step": 12717
     },
     {
-      "epoch": 0.3492900497102524,
+      "epoch": 0.3608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.5113319453890403e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.47868056081242e-05,
+      "loss": 0.7779,
       "step": 12718
     },
     {
-      "epoch": 0.3493175139380956,
+      "epoch": 0.3609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.5112554998463354e-05,
-      "loss": 0.806,
+      "learning_rate": 1.4785998647350087e-05,
+      "loss": 0.8946,
       "step": 12719
     },
     {
-      "epoch": 0.34934497816593885,
+      "epoch": 0.36095346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 1.5111790502583788e-05,
-      "loss": 0.8623,
+      "learning_rate": 1.4785191646148427e-05,
+      "loss": 0.9118,
       "step": 12720
     },
     {
-      "epoch": 0.3493724423937821,
+      "epoch": 0.3609818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.5111025966257752e-05,
-      "loss": 0.8535,
+      "learning_rate": 1.4784384604526037e-05,
+      "loss": 0.8943,
       "step": 12721
     },
     {
-      "epoch": 0.34939990662162534,
+      "epoch": 0.36101021566401814,
       "grad_norm": 0.0,
-      "learning_rate": 1.5110261389491306e-05,
-      "loss": 0.9041,
+      "learning_rate": 1.4783577522489733e-05,
+      "loss": 0.758,
       "step": 12722
     },
     {
-      "epoch": 0.34942737084946857,
+      "epoch": 0.36103859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.5109496772290486e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.478277040004634e-05,
+      "loss": 0.9819,
       "step": 12723
     },
     {
-      "epoch": 0.3494548350773118,
+      "epoch": 0.36106696935300797,
       "grad_norm": 0.0,
-      "learning_rate": 1.5108732114661352e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.4781963237202665e-05,
+      "loss": 0.986,
       "step": 12724
     },
     {
-      "epoch": 0.349482299305155,
+      "epoch": 0.36109534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.510796741660995e-05,
-      "loss": 0.953,
+      "learning_rate": 1.4781156033965536e-05,
+      "loss": 0.9031,
       "step": 12725
     },
     {
-      "epoch": 0.3495097635329983,
+      "epoch": 0.36112372304199775,
       "grad_norm": 0.0,
-      "learning_rate": 1.5107202678142328e-05,
-      "loss": 0.8888,
+      "learning_rate": 1.4780348790341767e-05,
+      "loss": 0.8964,
       "step": 12726
     },
     {
-      "epoch": 0.3495372277608415,
+      "epoch": 0.3611520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.510643789926454e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.4779541506338178e-05,
+      "loss": 0.8995,
       "step": 12727
     },
     {
-      "epoch": 0.34956469198868473,
+      "epoch": 0.3611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.5105673079982644e-05,
-      "loss": 0.9901,
+      "learning_rate": 1.4778734181961582e-05,
+      "loss": 0.9503,
       "step": 12728
     },
     {
-      "epoch": 0.34959215621652795,
+      "epoch": 0.3612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.5104908220302679e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.4777926817218808e-05,
+      "loss": 0.8943,
       "step": 12729
     },
     {
-      "epoch": 0.3496196204443712,
+      "epoch": 0.3612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.5104143320230703e-05,
-      "loss": 0.9021,
+      "learning_rate": 1.4777119412116667e-05,
+      "loss": 0.8529,
       "step": 12730
     },
     {
-      "epoch": 0.34964708467221445,
+      "epoch": 0.3612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.510337837977277e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.4776311966661987e-05,
+      "loss": 0.9639,
       "step": 12731
     },
     {
-      "epoch": 0.34967454890005767,
+      "epoch": 0.36129398410896707,
       "grad_norm": 0.0,
-      "learning_rate": 1.5102613398934931e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.4775504480861584e-05,
+      "loss": 1.0502,
       "step": 12732
     },
     {
-      "epoch": 0.3497020131279009,
+      "epoch": 0.361322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.5101848377723234e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.477469695472228e-05,
+      "loss": 0.9608,
       "step": 12733
     },
     {
-      "epoch": 0.34972947735574417,
+      "epoch": 0.36135073779795684,
       "grad_norm": 0.0,
-      "learning_rate": 1.5101083316143742e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.4773889388250896e-05,
+      "loss": 0.9814,
       "step": 12734
     },
     {
-      "epoch": 0.3497569415835874,
+      "epoch": 0.36137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.5100318214202499e-05,
-      "loss": 0.9969,
+      "learning_rate": 1.4773081781454254e-05,
+      "loss": 0.8156,
       "step": 12735
     },
     {
-      "epoch": 0.3497844058114306,
+      "epoch": 0.3614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.5099553071905564e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.4772274134339178e-05,
+      "loss": 1.0166,
       "step": 12736
     },
     {
-      "epoch": 0.34981187003927383,
+      "epoch": 0.36143586833144153,
       "grad_norm": 0.0,
-      "learning_rate": 1.509878788925899e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.4771466446912485e-05,
+      "loss": 0.9135,
       "step": 12737
     },
     {
-      "epoch": 0.34983933426711705,
+      "epoch": 0.36146424517593645,
       "grad_norm": 0.0,
-      "learning_rate": 1.509802266626883e-05,
-      "loss": 0.8472,
+      "learning_rate": 1.4770658719180999e-05,
+      "loss": 0.9308,
       "step": 12738
     },
     {
-      "epoch": 0.34986679849496033,
+      "epoch": 0.3614926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.5097257402941136e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.4769850951151547e-05,
+      "loss": 0.9205,
       "step": 12739
     },
     {
-      "epoch": 0.34989426272280355,
+      "epoch": 0.3615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.5096492099281977e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.476904314283095e-05,
+      "loss": 0.9994,
       "step": 12740
     },
     {
-      "epoch": 0.34992172695064677,
+      "epoch": 0.36154937570942114,
       "grad_norm": 0.0,
-      "learning_rate": 1.5095726755297391e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.476823529422603e-05,
+      "loss": 0.952,
       "step": 12741
     },
     {
-      "epoch": 0.34994919117849,
+      "epoch": 0.361577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.509496137099344e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.4767427405343613e-05,
+      "loss": 0.8712,
       "step": 12742
     },
     {
-      "epoch": 0.34997665540633327,
+      "epoch": 0.3616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.5094195946376187e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.4766619476190522e-05,
+      "loss": 0.8983,
       "step": 12743
     },
     {
-      "epoch": 0.3500041196341765,
+      "epoch": 0.3616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.5093430481451678e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.4765811506773582e-05,
+      "loss": 0.8852,
       "step": 12744
     },
     {
-      "epoch": 0.3500315838620197,
+      "epoch": 0.3616628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.5092664976225972e-05,
-      "loss": 0.9604,
+      "learning_rate": 1.4765003497099615e-05,
+      "loss": 0.8871,
       "step": 12745
     },
     {
-      "epoch": 0.35005904808986293,
+      "epoch": 0.36169125993189555,
       "grad_norm": 0.0,
-      "learning_rate": 1.5091899430705135e-05,
-      "loss": 1.0701,
+      "learning_rate": 1.4764195447175454e-05,
+      "loss": 0.926,
       "step": 12746
     },
     {
-      "epoch": 0.3500865123177062,
+      "epoch": 0.36171963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.509113384489521e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.4763387357007917e-05,
+      "loss": 0.8706,
       "step": 12747
     },
     {
-      "epoch": 0.35011397654554943,
+      "epoch": 0.3617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.5090368218802267e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.4762579226603833e-05,
+      "loss": 0.9591,
       "step": 12748
     },
     {
-      "epoch": 0.35014144077339265,
+      "epoch": 0.36177639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 1.5089602552432356e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.4761771055970032e-05,
+      "loss": 0.9238,
       "step": 12749
     },
     {
-      "epoch": 0.3501689050012359,
+      "epoch": 0.36180476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 1.508883684579154e-05,
-      "loss": 1.0132,
+      "learning_rate": 1.4760962845113334e-05,
+      "loss": 0.864,
       "step": 12750
     },
     {
-      "epoch": 0.35019636922907915,
+      "epoch": 0.36183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.5088071098885873e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.4760154594040566e-05,
+      "loss": 0.9429,
       "step": 12751
     },
     {
-      "epoch": 0.3502238334569224,
+      "epoch": 0.3618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.5087305311721424e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.4759346302758561e-05,
+      "loss": 0.9517,
       "step": 12752
     },
     {
-      "epoch": 0.3502512976847656,
+      "epoch": 0.36188989784335984,
       "grad_norm": 0.0,
-      "learning_rate": 1.5086539484304234e-05,
-      "loss": 0.8966,
+      "learning_rate": 1.4758537971274143e-05,
+      "loss": 0.9623,
       "step": 12753
     },
     {
-      "epoch": 0.3502787619126088,
+      "epoch": 0.3619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.5085773616640379e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.4757729599594144e-05,
+      "loss": 0.822,
       "step": 12754
     },
     {
-      "epoch": 0.35030622614045204,
+      "epoch": 0.3619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.5085007708735912e-05,
-      "loss": 1.0056,
+      "learning_rate": 1.4756921187725383e-05,
+      "loss": 0.9108,
       "step": 12755
     },
     {
-      "epoch": 0.3503336903682953,
+      "epoch": 0.3619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.5084241760596897e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.4756112735674698e-05,
+      "loss": 0.9152,
       "step": 12756
     },
     {
-      "epoch": 0.35036115459613854,
+      "epoch": 0.3620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.508347577222939e-05,
-      "loss": 1.03,
+      "learning_rate": 1.4755304243448918e-05,
+      "loss": 0.9174,
       "step": 12757
     },
     {
-      "epoch": 0.35038861882398176,
+      "epoch": 0.3620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.5082709743639452e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.4754495711054866e-05,
+      "loss": 0.9748,
       "step": 12758
     },
     {
-      "epoch": 0.350416083051825,
+      "epoch": 0.36206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.5081943674833146e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.4753687138499377e-05,
+      "loss": 0.939,
       "step": 12759
     },
     {
-      "epoch": 0.35044354727966825,
+      "epoch": 0.3620885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 1.5081177565816534e-05,
-      "loss": 1.0473,
+      "learning_rate": 1.4752878525789279e-05,
+      "loss": 0.8723,
       "step": 12760
     },
     {
-      "epoch": 0.3504710115075115,
+      "epoch": 0.36211691259931894,
       "grad_norm": 0.0,
-      "learning_rate": 1.5080411416595678e-05,
-      "loss": 0.8316,
+      "learning_rate": 1.4752069872931399e-05,
+      "loss": 0.9601,
       "step": 12761
     },
     {
-      "epoch": 0.3504984757353547,
+      "epoch": 0.36214528944381386,
       "grad_norm": 0.0,
-      "learning_rate": 1.5079645227176634e-05,
-      "loss": 0.9505,
+      "learning_rate": 1.4751261179932572e-05,
+      "loss": 0.8148,
       "step": 12762
     },
     {
-      "epoch": 0.3505259399631979,
+      "epoch": 0.3621736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.5078878997565473e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.475045244679963e-05,
+      "loss": 0.9077,
       "step": 12763
     },
     {
-      "epoch": 0.3505534041910412,
+      "epoch": 0.36220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.5078112727768255e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.4749643673539405e-05,
+      "loss": 0.8768,
       "step": 12764
     },
     {
-      "epoch": 0.3505808684188844,
+      "epoch": 0.36223041997729855,
       "grad_norm": 0.0,
-      "learning_rate": 1.5077346417791039e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.474883486015872e-05,
+      "loss": 0.8337,
       "step": 12765
     },
     {
-      "epoch": 0.35060833264672764,
+      "epoch": 0.3622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.5076580067639897e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.474802600666442e-05,
+      "loss": 0.9733,
       "step": 12766
     },
     {
-      "epoch": 0.35063579687457086,
+      "epoch": 0.3622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.5075813677320882e-05,
-      "loss": 0.9104,
+      "learning_rate": 1.4747217113063328e-05,
+      "loss": 0.9668,
       "step": 12767
     },
     {
-      "epoch": 0.3506632611024141,
+      "epoch": 0.3623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.5075047246840062e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.4746408179362278e-05,
+      "loss": 1.0083,
       "step": 12768
     },
     {
-      "epoch": 0.35069072533025736,
+      "epoch": 0.3623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.5074280776203508e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.4745599205568107e-05,
+      "loss": 0.9699,
       "step": 12769
     },
     {
-      "epoch": 0.3507181895581006,
+      "epoch": 0.362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.5073514265417273e-05,
-      "loss": 1.0158,
+      "learning_rate": 1.4744790191687646e-05,
+      "loss": 0.9258,
       "step": 12770
     },
     {
-      "epoch": 0.3507456537859438,
+      "epoch": 0.36240068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 1.507274771448743e-05,
-      "loss": 0.9773,
+      "learning_rate": 1.4743981137727728e-05,
+      "loss": 1.0708,
       "step": 12771
     },
     {
-      "epoch": 0.350773118013787,
+      "epoch": 0.3624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.5071981123420045e-05,
-      "loss": 0.882,
+      "learning_rate": 1.4743172043695189e-05,
+      "loss": 0.9869,
       "step": 12772
     },
     {
-      "epoch": 0.3508005822416303,
+      "epoch": 0.36245743473325764,
       "grad_norm": 0.0,
-      "learning_rate": 1.5071214492221175e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.4742362909596861e-05,
+      "loss": 1.0729,
       "step": 12773
     },
     {
-      "epoch": 0.3508280464694735,
+      "epoch": 0.36248581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.5070447820896894e-05,
-      "loss": 0.8836,
+      "learning_rate": 1.4741553735439581e-05,
+      "loss": 0.9781,
       "step": 12774
     },
     {
-      "epoch": 0.35085551069731674,
+      "epoch": 0.3625141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.5069681109453268e-05,
-      "loss": 0.8999,
+      "learning_rate": 1.4740744521230184e-05,
+      "loss": 0.9285,
       "step": 12775
     },
     {
-      "epoch": 0.35088297492515996,
+      "epoch": 0.36254256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.5068914357896361e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.4739935266975503e-05,
+      "loss": 0.967,
       "step": 12776
     },
     {
-      "epoch": 0.35091043915300324,
+      "epoch": 0.36257094211123725,
       "grad_norm": 0.0,
-      "learning_rate": 1.5068147566232236e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.473912597268238e-05,
+      "loss": 0.9287,
       "step": 12777
     },
     {
-      "epoch": 0.35093790338084646,
+      "epoch": 0.3625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.5067380734466967e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.4738316638357644e-05,
+      "loss": 0.9269,
       "step": 12778
     },
     {
-      "epoch": 0.3509653676086897,
+      "epoch": 0.362627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.5066613862606618e-05,
-      "loss": 0.9329,
+      "learning_rate": 1.4737507264008133e-05,
+      "loss": 1.0466,
       "step": 12779
     },
     {
-      "epoch": 0.3509928318365329,
+      "epoch": 0.3626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.5065846950657259e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.4736697849640688e-05,
+      "loss": 0.812,
       "step": 12780
     },
     {
-      "epoch": 0.3510202960643761,
+      "epoch": 0.3626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.5065079998624957e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.4735888395262144e-05,
+      "loss": 0.9784,
       "step": 12781
     },
     {
-      "epoch": 0.3510477602922194,
+      "epoch": 0.3627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.5064313006515777e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.4735078900879334e-05,
+      "loss": 0.9813,
       "step": 12782
     },
     {
-      "epoch": 0.3510752245200626,
+      "epoch": 0.3627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.5063545974335791e-05,
-      "loss": 0.9887,
+      "learning_rate": 1.4734269366499104e-05,
+      "loss": 0.9287,
       "step": 12783
     },
     {
-      "epoch": 0.35110268874790584,
+      "epoch": 0.3627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.5062778902091074e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.4733459792128283e-05,
+      "loss": 0.9772,
       "step": 12784
     },
     {
-      "epoch": 0.35113015297574907,
+      "epoch": 0.36279795686719635,
       "grad_norm": 0.0,
-      "learning_rate": 1.5062011789787683e-05,
-      "loss": 0.922,
+      "learning_rate": 1.4732650177773717e-05,
+      "loss": 0.8983,
       "step": 12785
     },
     {
-      "epoch": 0.35115761720359234,
+      "epoch": 0.36282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5061244637431691e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.4731840523442243e-05,
+      "loss": 0.8666,
       "step": 12786
     },
     {
-      "epoch": 0.35118508143143556,
+      "epoch": 0.3628547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.5060477445029177e-05,
-      "loss": 0.951,
+      "learning_rate": 1.4731030829140698e-05,
+      "loss": 0.9851,
       "step": 12787
     },
     {
-      "epoch": 0.3512125456592788,
+      "epoch": 0.36288308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.5059710212586205e-05,
-      "loss": 0.991,
+      "learning_rate": 1.4730221094875922e-05,
+      "loss": 0.908,
       "step": 12788
     },
     {
-      "epoch": 0.351240009887122,
+      "epoch": 0.36291146424517595,
       "grad_norm": 0.0,
-      "learning_rate": 1.5058942940108843e-05,
-      "loss": 0.9136,
+      "learning_rate": 1.4729411320654758e-05,
+      "loss": 0.9038,
       "step": 12789
     },
     {
-      "epoch": 0.3512674741149653,
+      "epoch": 0.3629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.5058175627603166e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.4728601506484043e-05,
+      "loss": 0.8604,
       "step": 12790
     },
     {
-      "epoch": 0.3512949383428085,
+      "epoch": 0.3629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.5057408275075243e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.472779165237062e-05,
+      "loss": 0.92,
       "step": 12791
     },
     {
-      "epoch": 0.3513224025706517,
+      "epoch": 0.3629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.5056640882531151e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.4726981758321324e-05,
+      "loss": 1.0583,
       "step": 12792
     },
     {
-      "epoch": 0.35134986679849495,
+      "epoch": 0.3630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.5055873449976952e-05,
-      "loss": 0.953,
+      "learning_rate": 1.4726171824343005e-05,
+      "loss": 0.819,
       "step": 12793
     },
     {
-      "epoch": 0.35137733102633817,
+      "epoch": 0.3630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.5055105977418727e-05,
-      "loss": 0.8833,
+      "learning_rate": 1.4725361850442502e-05,
+      "loss": 0.8305,
       "step": 12794
     },
     {
-      "epoch": 0.35140479525418145,
+      "epoch": 0.3630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.5054338464862544e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.4724551836626648e-05,
+      "loss": 0.9459,
       "step": 12795
     },
     {
-      "epoch": 0.35143225948202467,
+      "epoch": 0.3631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.5053570912314478e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.4723741782902297e-05,
+      "loss": 1.0269,
       "step": 12796
     },
     {
-      "epoch": 0.3514597237098679,
+      "epoch": 0.36313847900113505,
       "grad_norm": 0.0,
-      "learning_rate": 1.5052803319780602e-05,
-      "loss": 0.9024,
+      "learning_rate": 1.4722931689276287e-05,
+      "loss": 0.946,
       "step": 12797
     },
     {
-      "epoch": 0.3514871879377111,
+      "epoch": 0.36316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.5052035687266985e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.4722121555755457e-05,
+      "loss": 0.9395,
       "step": 12798
     },
     {
-      "epoch": 0.3515146521655544,
+      "epoch": 0.3631952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.505126801477971e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.4721311382346657e-05,
+      "loss": 0.872,
       "step": 12799
     },
     {
-      "epoch": 0.3515421163933976,
+      "epoch": 0.36322360953461974,
       "grad_norm": 0.0,
-      "learning_rate": 1.5050500302324844e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.4720501169056726e-05,
+      "loss": 0.9402,
       "step": 12800
     },
     {
-      "epoch": 0.35156958062124083,
+      "epoch": 0.36325198637911466,
       "grad_norm": 0.0,
-      "learning_rate": 1.5049732549908462e-05,
-      "loss": 0.9987,
+      "learning_rate": 1.4719690915892513e-05,
+      "loss": 0.9438,
       "step": 12801
     },
     {
-      "epoch": 0.35159704484908405,
+      "epoch": 0.3632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.5048964757536641e-05,
-      "loss": 0.9578,
+      "learning_rate": 1.4718880622860856e-05,
+      "loss": 0.9664,
       "step": 12802
     },
     {
-      "epoch": 0.3516245090769273,
+      "epoch": 0.36330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.5048196925215455e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.4718070289968602e-05,
+      "loss": 0.97,
       "step": 12803
     },
     {
-      "epoch": 0.35165197330477055,
+      "epoch": 0.36333711691259934,
       "grad_norm": 0.0,
-      "learning_rate": 1.504742905295098e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.4717259917222597e-05,
+      "loss": 0.8683,
       "step": 12804
     },
     {
-      "epoch": 0.35167943753261377,
+      "epoch": 0.3633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.504666114074929e-05,
-      "loss": 1.0176,
+      "learning_rate": 1.4716449504629685e-05,
+      "loss": 0.8124,
       "step": 12805
     },
     {
-      "epoch": 0.351706901760457,
+      "epoch": 0.3633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.5045893188616463e-05,
-      "loss": 0.913,
+      "learning_rate": 1.4715639052196712e-05,
+      "loss": 0.8096,
       "step": 12806
     },
     {
-      "epoch": 0.3517343659883002,
+      "epoch": 0.363422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.5045125196558572e-05,
-      "loss": 1.0621,
+      "learning_rate": 1.4714828559930523e-05,
+      "loss": 1.0227,
       "step": 12807
     },
     {
-      "epoch": 0.3517618302161435,
+      "epoch": 0.3634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.50443571645817e-05,
-      "loss": 0.993,
+      "learning_rate": 1.4714018027837966e-05,
+      "loss": 0.9012,
       "step": 12808
     },
     {
-      "epoch": 0.3517892944439867,
+      "epoch": 0.36347900113507375,
       "grad_norm": 0.0,
-      "learning_rate": 1.504358909269192e-05,
-      "loss": 0.9954,
+      "learning_rate": 1.4713207455925886e-05,
+      "loss": 1.0353,
       "step": 12809
     },
     {
-      "epoch": 0.35181675867182993,
+      "epoch": 0.36350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.5042820980895302e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.4712396844201134e-05,
+      "loss": 0.9818,
       "step": 12810
     },
     {
-      "epoch": 0.35184422289967315,
+      "epoch": 0.3635357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.504205282919794e-05,
-      "loss": 0.9208,
+      "learning_rate": 1.4711586192670551e-05,
+      "loss": 0.8859,
       "step": 12811
     },
     {
-      "epoch": 0.35187168712751643,
+      "epoch": 0.36356413166855844,
       "grad_norm": 0.0,
-      "learning_rate": 1.50412846376059e-05,
-      "loss": 0.983,
+      "learning_rate": 1.4710775501340988e-05,
+      "loss": 1.0063,
       "step": 12812
     },
     {
-      "epoch": 0.35189915135535965,
+      "epoch": 0.36359250851305336,
       "grad_norm": 0.0,
-      "learning_rate": 1.5040516406125262e-05,
-      "loss": 0.8635,
+      "learning_rate": 1.4709964770219293e-05,
+      "loss": 0.9562,
       "step": 12813
     },
     {
-      "epoch": 0.3519266155832029,
+      "epoch": 0.3636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.5039748134762107e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.4709153999312313e-05,
+      "loss": 0.9735,
       "step": 12814
     },
     {
-      "epoch": 0.3519540798110461,
+      "epoch": 0.36364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.503897982352251e-05,
-      "loss": 1.0261,
+      "learning_rate": 1.4708343188626899e-05,
+      "loss": 0.9913,
       "step": 12815
     },
     {
-      "epoch": 0.35198154403888937,
+      "epoch": 0.36367763904653805,
       "grad_norm": 0.0,
-      "learning_rate": 1.5038211472412556e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.4707532338169897e-05,
+      "loss": 0.8967,
       "step": 12816
     },
     {
-      "epoch": 0.3520090082667326,
+      "epoch": 0.3637060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.503744308143832e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.4706721447948158e-05,
+      "loss": 0.8577,
       "step": 12817
     },
     {
-      "epoch": 0.3520364724945758,
+      "epoch": 0.3637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.5036674650605883e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.4705910517968533e-05,
+      "loss": 0.9577,
       "step": 12818
     },
     {
-      "epoch": 0.35206393672241904,
+      "epoch": 0.3637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.5035906179921327e-05,
-      "loss": 0.8769,
+      "learning_rate": 1.4705099548237869e-05,
+      "loss": 1.084,
       "step": 12819
     },
     {
-      "epoch": 0.35209140095026226,
+      "epoch": 0.3637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.5035137669390727e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.4704288538763019e-05,
+      "loss": 0.9008,
       "step": 12820
     },
     {
-      "epoch": 0.35211886517810553,
+      "epoch": 0.3638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.5034369119020169e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.470347748955083e-05,
+      "loss": 0.8323,
       "step": 12821
     },
     {
-      "epoch": 0.35214632940594875,
+      "epoch": 0.36384790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 1.5033600528815733e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.4702666400608157e-05,
+      "loss": 0.9642,
       "step": 12822
     },
     {
-      "epoch": 0.352173793633792,
+      "epoch": 0.3638762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.5032831898783497e-05,
-      "loss": 0.8478,
+      "learning_rate": 1.4701855271941848e-05,
+      "loss": 0.9524,
       "step": 12823
     },
     {
-      "epoch": 0.3522012578616352,
+      "epoch": 0.36390465380249715,
       "grad_norm": 0.0,
-      "learning_rate": 1.503206322892955e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.4701044103558757e-05,
+      "loss": 0.9164,
       "step": 12824
     },
     {
-      "epoch": 0.3522287220894785,
+      "epoch": 0.36393303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.5031294519259967e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.4700232895465733e-05,
+      "loss": 0.8227,
       "step": 12825
     },
     {
-      "epoch": 0.3522561863173217,
+      "epoch": 0.3639614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 1.5030525769780833e-05,
-      "loss": 1.0958,
+      "learning_rate": 1.4699421647669634e-05,
+      "loss": 0.9102,
       "step": 12826
     },
     {
-      "epoch": 0.3522836505451649,
+      "epoch": 0.36398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.5029756980498229e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.4698610360177306e-05,
+      "loss": 0.8471,
       "step": 12827
     },
     {
-      "epoch": 0.35231111477300814,
+      "epoch": 0.36401816118047675,
       "grad_norm": 0.0,
-      "learning_rate": 1.5028988151418244e-05,
-      "loss": 0.9202,
+      "learning_rate": 1.4697799032995608e-05,
+      "loss": 0.848,
       "step": 12828
     },
     {
-      "epoch": 0.3523385790008514,
+      "epoch": 0.3640465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.5028219282546952e-05,
-      "loss": 0.858,
+      "learning_rate": 1.4696987666131388e-05,
+      "loss": 0.9479,
       "step": 12829
     },
     {
-      "epoch": 0.35236604322869464,
+      "epoch": 0.3640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.5027450373890439e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.4696176259591501e-05,
+      "loss": 0.8852,
       "step": 12830
     },
     {
-      "epoch": 0.35239350745653786,
+      "epoch": 0.3641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.50266814254548e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.4695364813382806e-05,
+      "loss": 0.9702,
       "step": 12831
     },
     {
-      "epoch": 0.3524209716843811,
+      "epoch": 0.3641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.5025912437246104e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.4694553327512151e-05,
+      "loss": 0.936,
       "step": 12832
     },
     {
-      "epoch": 0.3524484359122243,
+      "epoch": 0.3641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.5025143409270442e-05,
-      "loss": 0.947,
+      "learning_rate": 1.4693741801986392e-05,
+      "loss": 0.9384,
       "step": 12833
     },
     {
-      "epoch": 0.3524759001400676,
+      "epoch": 0.3641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.5024374341533903e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.4692930236812389e-05,
+      "loss": 1.0175,
       "step": 12834
     },
     {
-      "epoch": 0.3525033643679108,
+      "epoch": 0.364216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.5023605234042564e-05,
-      "loss": 0.8474,
+      "learning_rate": 1.469211863199699e-05,
+      "loss": 1.0099,
       "step": 12835
     },
     {
-      "epoch": 0.352530828595754,
+      "epoch": 0.36424517593643585,
       "grad_norm": 0.0,
-      "learning_rate": 1.5022836086802515e-05,
-      "loss": 0.8557,
+      "learning_rate": 1.4691306987547054e-05,
+      "loss": 0.9156,
       "step": 12836
     },
     {
-      "epoch": 0.35255829282359724,
+      "epoch": 0.36427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.5022066899819842e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.4690495303469436e-05,
+      "loss": 0.8938,
       "step": 12837
     },
     {
-      "epoch": 0.3525857570514405,
+      "epoch": 0.3643019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.502129767310063e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.4689683579770994e-05,
+      "loss": 0.8361,
       "step": 12838
     },
     {
-      "epoch": 0.35261322127928374,
+      "epoch": 0.36433030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.5020528406650965e-05,
-      "loss": 0.8569,
+      "learning_rate": 1.4688871816458585e-05,
+      "loss": 0.958,
       "step": 12839
     },
     {
-      "epoch": 0.35264068550712696,
+      "epoch": 0.36435868331441545,
       "grad_norm": 0.0,
-      "learning_rate": 1.5019759100476936e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.4688060013539068e-05,
+      "loss": 1.035,
       "step": 12840
     },
     {
-      "epoch": 0.3526681497349702,
+      "epoch": 0.3643870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.5018989754584627e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.4687248171019293e-05,
+      "loss": 0.9662,
       "step": 12841
     },
     {
-      "epoch": 0.35269561396281346,
+      "epoch": 0.36441543700340523,
       "grad_norm": 0.0,
-      "learning_rate": 1.5018220368980128e-05,
-      "loss": 0.8865,
+      "learning_rate": 1.4686436288906124e-05,
+      "loss": 0.9041,
       "step": 12842
     },
     {
-      "epoch": 0.3527230781906567,
+      "epoch": 0.3644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.5017450943669526e-05,
-      "loss": 0.9176,
+      "learning_rate": 1.4685624367206414e-05,
+      "loss": 0.9714,
       "step": 12843
     },
     {
-      "epoch": 0.3527505424184999,
+      "epoch": 0.364472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.5016681478658907e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.468481240592703e-05,
+      "loss": 0.9995,
       "step": 12844
     },
     {
-      "epoch": 0.3527780066463431,
+      "epoch": 0.3645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.5015911973954363e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.4684000405074818e-05,
+      "loss": 0.8637,
       "step": 12845
     },
     {
-      "epoch": 0.3528054708741864,
+      "epoch": 0.3645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.501514242956198e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.4683188364656648e-05,
+      "loss": 1.0257,
       "step": 12846
     },
     {
-      "epoch": 0.3528329351020296,
+      "epoch": 0.3645573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.5014372845487844e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.4682376284679376e-05,
+      "loss": 0.8682,
       "step": 12847
     },
     {
-      "epoch": 0.35286039932987284,
+      "epoch": 0.36458569807037455,
       "grad_norm": 0.0,
-      "learning_rate": 1.5013603221738051e-05,
-      "loss": 0.9096,
+      "learning_rate": 1.4681564165149863e-05,
+      "loss": 0.9157,
       "step": 12848
     },
     {
-      "epoch": 0.35288786355771606,
+      "epoch": 0.36461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.5012833558318689e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.4680752006074961e-05,
+      "loss": 1.0112,
       "step": 12849
     },
     {
-      "epoch": 0.3529153277855593,
+      "epoch": 0.3646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.5012063855235842e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.4679939807461541e-05,
+      "loss": 0.8735,
       "step": 12850
     },
     {
-      "epoch": 0.35294279201340256,
+      "epoch": 0.36467082860385924,
       "grad_norm": 0.0,
-      "learning_rate": 1.5011294112495604e-05,
-      "loss": 0.9879,
+      "learning_rate": 1.4679127569316455e-05,
+      "loss": 1.0056,
       "step": 12851
     },
     {
-      "epoch": 0.3529702562412458,
+      "epoch": 0.36469920544835416,
       "grad_norm": 0.0,
-      "learning_rate": 1.5010524330104073e-05,
-      "loss": 0.8293,
+      "learning_rate": 1.4678315291646572e-05,
+      "loss": 0.9802,
       "step": 12852
     },
     {
-      "epoch": 0.352997720469089,
+      "epoch": 0.364727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.5009754508067325e-05,
-      "loss": 0.8696,
+      "learning_rate": 1.4677502974458746e-05,
+      "loss": 1.063,
       "step": 12853
     },
     {
-      "epoch": 0.3530251846969322,
+      "epoch": 0.36475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.500898464639146e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.4676690617759845e-05,
+      "loss": 1.0199,
       "step": 12854
     },
     {
-      "epoch": 0.3530526489247755,
+      "epoch": 0.3647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.500821474508257e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.4675878221556727e-05,
+      "loss": 0.8524,
       "step": 12855
     },
     {
-      "epoch": 0.3530801131526187,
+      "epoch": 0.3648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.5007444804146744e-05,
-      "loss": 0.8835,
+      "learning_rate": 1.4675065785856256e-05,
+      "loss": 1.109,
       "step": 12856
     },
     {
-      "epoch": 0.35310757738046195,
+      "epoch": 0.3648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.5006674823590074e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.4674253310665294e-05,
+      "loss": 0.9111,
       "step": 12857
     },
     {
-      "epoch": 0.35313504160830517,
+      "epoch": 0.3648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.5005904803418656e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.4673440795990705e-05,
+      "loss": 0.904,
       "step": 12858
     },
     {
-      "epoch": 0.35316250583614844,
+      "epoch": 0.3648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.5005134743638579e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.4672628241839349e-05,
+      "loss": 0.927,
       "step": 12859
     },
     {
-      "epoch": 0.35318997006399167,
+      "epoch": 0.36492622020431326,
       "grad_norm": 0.0,
-      "learning_rate": 1.5004364644255938e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.4671815648218092e-05,
+      "loss": 0.988,
       "step": 12860
     },
     {
-      "epoch": 0.3532174342918349,
+      "epoch": 0.36495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.5003594505276824e-05,
-      "loss": 0.92,
+      "learning_rate": 1.4671003015133803e-05,
+      "loss": 0.916,
       "step": 12861
     },
     {
-      "epoch": 0.3532448985196781,
+      "epoch": 0.3649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.500282432670733e-05,
-      "loss": 0.8875,
+      "learning_rate": 1.4670190342593338e-05,
+      "loss": 0.8822,
       "step": 12862
     },
     {
-      "epoch": 0.35327236274752133,
+      "epoch": 0.36501135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.5002054108553555e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.4669377630603565e-05,
+      "loss": 0.9355,
       "step": 12863
     },
     {
-      "epoch": 0.3532998269753646,
+      "epoch": 0.36503972758229286,
       "grad_norm": 0.0,
-      "learning_rate": 1.500128385082159e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.466856487917135e-05,
+      "loss": 0.9453,
       "step": 12864
     },
     {
-      "epoch": 0.3533272912032078,
+      "epoch": 0.3650681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.500051355351753e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.466775208830356e-05,
+      "loss": 1.0282,
       "step": 12865
     },
     {
-      "epoch": 0.35335475543105105,
+      "epoch": 0.36509648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.499974321664747e-05,
-      "loss": 0.8932,
+      "learning_rate": 1.4666939258007054e-05,
+      "loss": 0.9493,
       "step": 12866
     },
     {
-      "epoch": 0.35338221965889427,
+      "epoch": 0.36512485811577755,
       "grad_norm": 0.0,
-      "learning_rate": 1.4998972840217503e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.4666126388288703e-05,
+      "loss": 0.9281,
       "step": 12867
     },
     {
-      "epoch": 0.35340968388673755,
+      "epoch": 0.3651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.4998202424233728e-05,
-      "loss": 0.8654,
+      "learning_rate": 1.4665313479155375e-05,
+      "loss": 0.9637,
       "step": 12868
     },
     {
-      "epoch": 0.35343714811458077,
+      "epoch": 0.3651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.4997431968702238e-05,
-      "loss": 0.9656,
+      "learning_rate": 1.4664500530613932e-05,
+      "loss": 0.8991,
       "step": 12869
     },
     {
-      "epoch": 0.353464612342424,
+      "epoch": 0.3652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.4996661473629133e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.4663687542671245e-05,
+      "loss": 0.9779,
       "step": 12870
     },
     {
-      "epoch": 0.3534920765702672,
+      "epoch": 0.3652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.4995890939020503e-05,
-      "loss": 1.0607,
+      "learning_rate": 1.4662874515334178e-05,
+      "loss": 0.9846,
       "step": 12871
     },
     {
-      "epoch": 0.3535195407981105,
+      "epoch": 0.36526674233825196,
       "grad_norm": 0.0,
-      "learning_rate": 1.4995120364882452e-05,
-      "loss": 0.8753,
+      "learning_rate": 1.4662061448609604e-05,
+      "loss": 0.9615,
       "step": 12872
     },
     {
-      "epoch": 0.3535470050259537,
+      "epoch": 0.3652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.4994349751221073e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.4661248342504383e-05,
+      "loss": 0.9031,
       "step": 12873
     },
     {
-      "epoch": 0.35357446925379693,
+      "epoch": 0.3653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.4993579098042464e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.4660435197025391e-05,
+      "loss": 0.8938,
       "step": 12874
     },
     {
-      "epoch": 0.35360193348164015,
+      "epoch": 0.36535187287173665,
       "grad_norm": 0.0,
-      "learning_rate": 1.4992808405352721e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.4659622012179493e-05,
+      "loss": 0.9798,
       "step": 12875
     },
     {
-      "epoch": 0.3536293977094834,
+      "epoch": 0.36538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.4992037673157948e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.4658808787973556e-05,
+      "loss": 0.8718,
       "step": 12876
     },
     {
-      "epoch": 0.35365686193732665,
+      "epoch": 0.3654086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.4991266901464237e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.4657995524414453e-05,
+      "loss": 0.9139,
       "step": 12877
     },
     {
-      "epoch": 0.35368432616516987,
+      "epoch": 0.36543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.4990496090277689e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.4657182221509051e-05,
+      "loss": 1.0076,
       "step": 12878
     },
     {
-      "epoch": 0.3537117903930131,
+      "epoch": 0.36546538024971625,
       "grad_norm": 0.0,
-      "learning_rate": 1.49897252396044e-05,
-      "loss": 0.774,
+      "learning_rate": 1.4656368879264225e-05,
+      "loss": 0.9622,
       "step": 12879
     },
     {
-      "epoch": 0.3537392546208563,
+      "epoch": 0.3654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.4988954349450475e-05,
-      "loss": 0.9969,
+      "learning_rate": 1.4655555497686837e-05,
+      "loss": 0.9373,
       "step": 12880
     },
     {
-      "epoch": 0.3537667188486996,
+      "epoch": 0.36552213393870603,
       "grad_norm": 0.0,
-      "learning_rate": 1.498818341982201e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.4654742076783767e-05,
+      "loss": 0.8267,
       "step": 12881
     },
     {
-      "epoch": 0.3537941830765428,
+      "epoch": 0.3655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.4987412450725108e-05,
-      "loss": 0.936,
+      "learning_rate": 1.465392861656188e-05,
+      "loss": 0.8947,
       "step": 12882
     },
     {
-      "epoch": 0.35382164730438603,
+      "epoch": 0.3655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.4986641442165861e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.4653115117028045e-05,
+      "loss": 1.0233,
       "step": 12883
     },
     {
-      "epoch": 0.35384911153222925,
+      "epoch": 0.3656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.4985870394150381e-05,
-      "loss": 0.9256,
+      "learning_rate": 1.4652301578189141e-05,
+      "loss": 0.9846,
       "step": 12884
     },
     {
-      "epoch": 0.35387657576007253,
+      "epoch": 0.3656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.498509930668476e-05,
-      "loss": 1.0298,
+      "learning_rate": 1.4651488000052036e-05,
+      "loss": 0.8945,
       "step": 12885
     },
     {
-      "epoch": 0.35390403998791575,
+      "epoch": 0.3656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.4984328179775103e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.4650674382623606e-05,
+      "loss": 0.9693,
       "step": 12886
     },
     {
-      "epoch": 0.353931504215759,
+      "epoch": 0.36569239500567535,
       "grad_norm": 0.0,
-      "learning_rate": 1.4983557013427513e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.4649860725910716e-05,
+      "loss": 0.9135,
       "step": 12887
     },
     {
-      "epoch": 0.3539589684436022,
+      "epoch": 0.36572077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.4982785807648086e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.4649047029920245e-05,
+      "loss": 0.9866,
       "step": 12888
     },
     {
-      "epoch": 0.3539864326714454,
+      "epoch": 0.3657491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.4982014562442928e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.4648233294659066e-05,
+      "loss": 0.9877,
       "step": 12889
     },
     {
-      "epoch": 0.3540138968992887,
+      "epoch": 0.36577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.4981243277818142e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.4647419520134047e-05,
+      "loss": 0.9714,
       "step": 12890
     },
     {
-      "epoch": 0.3540413611271319,
+      "epoch": 0.36580590238365496,
       "grad_norm": 0.0,
-      "learning_rate": 1.498047195377983e-05,
-      "loss": 0.976,
+      "learning_rate": 1.464660570635207e-05,
+      "loss": 1.0271,
       "step": 12891
     },
     {
-      "epoch": 0.35406882535497514,
+      "epoch": 0.3658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.4979700590334095e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.4645791853320005e-05,
+      "loss": 0.8931,
       "step": 12892
     },
     {
-      "epoch": 0.35409628958281836,
+      "epoch": 0.36586265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.4978929187487037e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.4644977961044725e-05,
+      "loss": 0.9916,
       "step": 12893
     },
     {
-      "epoch": 0.35412375381066163,
+      "epoch": 0.3658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.4978157745244763e-05,
-      "loss": 1.0025,
+      "learning_rate": 1.4644164029533113e-05,
+      "loss": 0.8785,
       "step": 12894
     },
     {
-      "epoch": 0.35415121803850486,
+      "epoch": 0.3659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.4977386263613379e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.4643350058792036e-05,
+      "loss": 1.0405,
       "step": 12895
     },
     {
-      "epoch": 0.3541786822663481,
+      "epoch": 0.3659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.4976614742598985e-05,
-      "loss": 0.9842,
+      "learning_rate": 1.464253604882837e-05,
+      "loss": 0.8698,
       "step": 12896
     },
     {
-      "epoch": 0.3542061464941913,
+      "epoch": 0.3659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.497584318220769e-05,
-      "loss": 0.8193,
+      "learning_rate": 1.4641721999648994e-05,
+      "loss": 0.9104,
       "step": 12897
     },
     {
-      "epoch": 0.3542336107220346,
+      "epoch": 0.3660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.4975071582445595e-05,
-      "loss": 0.8986,
+      "learning_rate": 1.4640907911260787e-05,
+      "loss": 0.9662,
       "step": 12898
     },
     {
-      "epoch": 0.3542610749498778,
+      "epoch": 0.36603291713961406,
       "grad_norm": 0.0,
-      "learning_rate": 1.4974299943318807e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.4640093783670619e-05,
+      "loss": 0.9013,
       "step": 12899
     },
     {
-      "epoch": 0.354288539177721,
+      "epoch": 0.36606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.497352826483343e-05,
-      "loss": 0.943,
+      "learning_rate": 1.463927961688537e-05,
+      "loss": 1.0025,
       "step": 12900
     },
     {
-      "epoch": 0.35431600340556424,
+      "epoch": 0.3660896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.4972756546995571e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.463846541091192e-05,
+      "loss": 0.976,
       "step": 12901
     },
     {
-      "epoch": 0.35434346763340746,
+      "epoch": 0.36611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.4971984789811336e-05,
-      "loss": 1.0411,
+      "learning_rate": 1.4637651165757143e-05,
+      "loss": 0.9989,
       "step": 12902
     },
     {
-      "epoch": 0.35437093186125074,
+      "epoch": 0.36614642451759366,
       "grad_norm": 0.0,
-      "learning_rate": 1.4971212993286827e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.4636836881427918e-05,
+      "loss": 0.9379,
       "step": 12903
     },
     {
-      "epoch": 0.35439839608909396,
+      "epoch": 0.3661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.4970441157428162e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.4636022557931124e-05,
+      "loss": 0.9778,
       "step": 12904
     },
     {
-      "epoch": 0.3544258603169372,
+      "epoch": 0.36620317820658344,
       "grad_norm": 0.0,
-      "learning_rate": 1.496966928224144e-05,
-      "loss": 1.027,
+      "learning_rate": 1.4635208195273638e-05,
+      "loss": 1.0027,
       "step": 12905
     },
     {
-      "epoch": 0.3544533245447804,
+      "epoch": 0.3662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.4968897367732766e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.4634393793462341e-05,
+      "loss": 0.8889,
       "step": 12906
     },
     {
-      "epoch": 0.3544807887726237,
+      "epoch": 0.3662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.4968125413908252e-05,
-      "loss": 0.8672,
+      "learning_rate": 1.4633579352504109e-05,
+      "loss": 0.9816,
       "step": 12907
     },
     {
-      "epoch": 0.3545082530004669,
+      "epoch": 0.3662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.496735342077401e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.4632764872405827e-05,
+      "loss": 0.937,
       "step": 12908
     },
     {
-      "epoch": 0.3545357172283101,
+      "epoch": 0.366316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.4966581388336139e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.4631950353174368e-05,
+      "loss": 0.8452,
       "step": 12909
     },
     {
-      "epoch": 0.35456318145615334,
+      "epoch": 0.3663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.4965809316600752e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.463113579481662e-05,
+      "loss": 1.0046,
       "step": 12910
     },
     {
-      "epoch": 0.3545906456839966,
+      "epoch": 0.36637343927355276,
       "grad_norm": 0.0,
-      "learning_rate": 1.4965037205573962e-05,
-      "loss": 0.9338,
+      "learning_rate": 1.463032119733946e-05,
+      "loss": 0.9767,
       "step": 12911
     },
     {
-      "epoch": 0.35461810991183984,
+      "epoch": 0.3664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.4964265055261872e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.4629506560749766e-05,
+      "loss": 0.9122,
       "step": 12912
     },
     {
-      "epoch": 0.35464557413968306,
+      "epoch": 0.3664301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.4963492865670592e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.4628691885054423e-05,
+      "loss": 0.9273,
       "step": 12913
     },
     {
-      "epoch": 0.3546730383675263,
+      "epoch": 0.36645856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.4962720636806235e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.462787717026031e-05,
+      "loss": 0.8827,
       "step": 12914
     },
     {
-      "epoch": 0.3547005025953695,
+      "epoch": 0.36648694665153236,
       "grad_norm": 0.0,
-      "learning_rate": 1.496194836867491e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.4627062416374314e-05,
+      "loss": 0.925,
       "step": 12915
     },
     {
-      "epoch": 0.3547279668232128,
+      "epoch": 0.3665153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.496117606128273e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.462624762340331e-05,
+      "loss": 0.9576,
       "step": 12916
     },
     {
-      "epoch": 0.354755431051056,
+      "epoch": 0.36654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.4960403714635802e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.4625432791354187e-05,
+      "loss": 0.9695,
       "step": 12917
     },
     {
-      "epoch": 0.3547828952788992,
+      "epoch": 0.36657207718501705,
       "grad_norm": 0.0,
-      "learning_rate": 1.4959631328740236e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.4624617920233826e-05,
+      "loss": 0.8121,
       "step": 12918
     },
     {
-      "epoch": 0.35481035950674245,
+      "epoch": 0.3666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.4958858903602148e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.462380301004911e-05,
+      "loss": 0.9024,
       "step": 12919
     },
     {
-      "epoch": 0.3548378237345857,
+      "epoch": 0.36662883087400683,
       "grad_norm": 0.0,
-      "learning_rate": 1.4958086439227647e-05,
-      "loss": 0.9512,
+      "learning_rate": 1.4622988060806917e-05,
+      "loss": 0.9734,
       "step": 12920
     },
     {
-      "epoch": 0.35486528796242894,
+      "epoch": 0.3666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.4957313935622843e-05,
-      "loss": 0.928,
+      "learning_rate": 1.4622173072514141e-05,
+      "loss": 0.9552,
       "step": 12921
     },
     {
-      "epoch": 0.35489275219027216,
+      "epoch": 0.3666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.4956541392793855e-05,
-      "loss": 1.0304,
+      "learning_rate": 1.4621358045177658e-05,
+      "loss": 0.9864,
       "step": 12922
     },
     {
-      "epoch": 0.3549202164181154,
+      "epoch": 0.36671396140749146,
       "grad_norm": 0.0,
-      "learning_rate": 1.4955768810746788e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.4620542978804357e-05,
+      "loss": 0.9946,
       "step": 12923
     },
     {
-      "epoch": 0.35494768064595866,
+      "epoch": 0.3667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.4954996189487763e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.4619727873401122e-05,
+      "loss": 0.9641,
       "step": 12924
     },
     {
-      "epoch": 0.3549751448738019,
+      "epoch": 0.3667707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.4954223529022885e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.461891272897484e-05,
+      "loss": 0.9528,
       "step": 12925
     },
     {
-      "epoch": 0.3550026091016451,
+      "epoch": 0.36679909194097615,
       "grad_norm": 0.0,
-      "learning_rate": 1.4953450829358272e-05,
-      "loss": 1.0215,
+      "learning_rate": 1.4618097545532393e-05,
+      "loss": 0.9555,
       "step": 12926
     },
     {
-      "epoch": 0.3550300733294883,
+      "epoch": 0.36682746878547107,
       "grad_norm": 0.0,
-      "learning_rate": 1.4952678090500037e-05,
-      "loss": 1.0183,
+      "learning_rate": 1.4617282323080666e-05,
+      "loss": 0.958,
       "step": 12927
     },
     {
-      "epoch": 0.35505753755733155,
+      "epoch": 0.3668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.4951905312454295e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.461646706162655e-05,
+      "loss": 0.876,
       "step": 12928
     },
     {
-      "epoch": 0.3550850017851748,
+      "epoch": 0.36688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.4951132495227161e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.4615651761176928e-05,
+      "loss": 0.9509,
       "step": 12929
     },
     {
-      "epoch": 0.35511246601301805,
+      "epoch": 0.36691259931895576,
       "grad_norm": 0.0,
-      "learning_rate": 1.4950359638824747e-05,
-      "loss": 0.9896,
+      "learning_rate": 1.4614836421738692e-05,
+      "loss": 0.9032,
       "step": 12930
     },
     {
-      "epoch": 0.35513993024086127,
+      "epoch": 0.3669409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.4949586743253171e-05,
-      "loss": 1.0344,
+      "learning_rate": 1.461402104331872e-05,
+      "loss": 0.9235,
       "step": 12931
     },
     {
-      "epoch": 0.3551673944687045,
+      "epoch": 0.36696935300794553,
       "grad_norm": 0.0,
-      "learning_rate": 1.4948813808518548e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.4613205625923907e-05,
+      "loss": 0.9059,
       "step": 12932
     },
     {
-      "epoch": 0.35519485869654777,
+      "epoch": 0.3669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.4948040834626992e-05,
-      "loss": 0.832,
+      "learning_rate": 1.461239016956114e-05,
+      "loss": 0.7857,
       "step": 12933
     },
     {
-      "epoch": 0.355222322924391,
+      "epoch": 0.3670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.494726782158462e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.4611574674237302e-05,
+      "loss": 0.9603,
       "step": 12934
     },
     {
-      "epoch": 0.3552497871522342,
+      "epoch": 0.36705448354143017,
       "grad_norm": 0.0,
-      "learning_rate": 1.4946494769397548e-05,
-      "loss": 1.0232,
+      "learning_rate": 1.461075913995929e-05,
+      "loss": 0.9212,
       "step": 12935
     },
     {
-      "epoch": 0.35527725138007743,
+      "epoch": 0.3670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.4945721678071896e-05,
-      "loss": 0.996,
+      "learning_rate": 1.4609943566733988e-05,
+      "loss": 0.9847,
       "step": 12936
     },
     {
-      "epoch": 0.3553047156079207,
+      "epoch": 0.36711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.4944948547613778e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.460912795456828e-05,
+      "loss": 0.8544,
       "step": 12937
     },
     {
-      "epoch": 0.35533217983576393,
+      "epoch": 0.36713961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 1.4944175378029308e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.4608312303469067e-05,
+      "loss": 1.0172,
       "step": 12938
     },
     {
-      "epoch": 0.35535964406360715,
+      "epoch": 0.36716799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 1.4943402169324607e-05,
-      "loss": 0.9813,
+      "learning_rate": 1.460749661344323e-05,
+      "loss": 0.8598,
       "step": 12939
     },
     {
-      "epoch": 0.35538710829145037,
+      "epoch": 0.36719636776390463,
       "grad_norm": 0.0,
-      "learning_rate": 1.4942628921505798e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.4606680884497664e-05,
+      "loss": 0.9561,
       "step": 12940
     },
     {
-      "epoch": 0.35541457251929365,
+      "epoch": 0.36722474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.4941855634578992e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.4605865116639255e-05,
+      "loss": 1.0012,
       "step": 12941
     },
     {
-      "epoch": 0.35544203674713687,
+      "epoch": 0.36725312145289446,
       "grad_norm": 0.0,
-      "learning_rate": 1.494108230855031e-05,
-      "loss": 0.95,
+      "learning_rate": 1.4605049309874899e-05,
+      "loss": 0.863,
       "step": 12942
     },
     {
-      "epoch": 0.3554695009749801,
+      "epoch": 0.3672814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.4940308943425871e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.4604233464211484e-05,
+      "loss": 0.8333,
       "step": 12943
     },
     {
-      "epoch": 0.3554969652028233,
+      "epoch": 0.36730987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.4939535539211796e-05,
-      "loss": 0.963,
+      "learning_rate": 1.4603417579655901e-05,
+      "loss": 0.848,
       "step": 12944
     },
     {
-      "epoch": 0.35552442943066653,
+      "epoch": 0.3673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.4938762095914197e-05,
-      "loss": 0.8552,
+      "learning_rate": 1.4602601656215041e-05,
+      "loss": 0.9441,
       "step": 12945
     },
     {
-      "epoch": 0.3555518936585098,
+      "epoch": 0.367366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.4937988613539203e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.4601785693895801e-05,
+      "loss": 0.891,
       "step": 12946
     },
     {
-      "epoch": 0.35557935788635303,
+      "epoch": 0.3673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.493721509209293e-05,
-      "loss": 0.957,
+      "learning_rate": 1.4600969692705067e-05,
+      "loss": 0.9686,
       "step": 12947
     },
     {
-      "epoch": 0.35560682211419625,
+      "epoch": 0.3674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.4936441531581498e-05,
-      "loss": 1.035,
+      "learning_rate": 1.4600153652649737e-05,
+      "loss": 0.8912,
       "step": 12948
     },
     {
-      "epoch": 0.3556342863420395,
+      "epoch": 0.3674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.4935667932011031e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.4599337573736703e-05,
+      "loss": 0.8712,
       "step": 12949
     },
     {
-      "epoch": 0.35566175056988275,
+      "epoch": 0.36748013620885356,
       "grad_norm": 0.0,
-      "learning_rate": 1.4934894293387642e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.4598521455972857e-05,
+      "loss": 0.9627,
       "step": 12950
     },
     {
-      "epoch": 0.35568921479772597,
+      "epoch": 0.3675085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.493412061571746e-05,
-      "loss": 1.0311,
+      "learning_rate": 1.4597705299365094e-05,
+      "loss": 0.8785,
       "step": 12951
     },
     {
-      "epoch": 0.3557166790255692,
+      "epoch": 0.36753688989784333,
       "grad_norm": 0.0,
-      "learning_rate": 1.4933346899006607e-05,
-      "loss": 0.8786,
+      "learning_rate": 1.4596889103920306e-05,
+      "loss": 0.9261,
       "step": 12952
     },
     {
-      "epoch": 0.3557441432534124,
+      "epoch": 0.36756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.4932573143261203e-05,
-      "loss": 0.96,
+      "learning_rate": 1.459607286964539e-05,
+      "loss": 0.8728,
       "step": 12953
     },
     {
-      "epoch": 0.3557716074812557,
+      "epoch": 0.36759364358683316,
       "grad_norm": 0.0,
-      "learning_rate": 1.4931799348487363e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.459525659654724e-05,
+      "loss": 0.9337,
       "step": 12954
     },
     {
-      "epoch": 0.3557990717090989,
+      "epoch": 0.367622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.4931025514691222e-05,
-      "loss": 0.8665,
+      "learning_rate": 1.4594440284632754e-05,
+      "loss": 0.9398,
       "step": 12955
     },
     {
-      "epoch": 0.35582653593694213,
+      "epoch": 0.36765039727582294,
       "grad_norm": 0.0,
-      "learning_rate": 1.4930251641878897e-05,
-      "loss": 0.9977,
+      "learning_rate": 1.4593623933908822e-05,
+      "loss": 0.9554,
       "step": 12956
     },
     {
-      "epoch": 0.35585400016478536,
+      "epoch": 0.3676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.4929477730056507e-05,
-      "loss": 0.9523,
+      "learning_rate": 1.4592807544382339e-05,
+      "loss": 0.8775,
       "step": 12957
     },
     {
-      "epoch": 0.3558814643926286,
+      "epoch": 0.3677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.4928703779230186e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.459199111606021e-05,
+      "loss": 0.9851,
       "step": 12958
     },
     {
-      "epoch": 0.35590892862047185,
+      "epoch": 0.3677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.4927929789406046e-05,
-      "loss": 0.9223,
+      "learning_rate": 1.4591174648949323e-05,
+      "loss": 0.9461,
       "step": 12959
     },
     {
-      "epoch": 0.3559363928483151,
+      "epoch": 0.3677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.492715576059022e-05,
-      "loss": 0.969,
+      "learning_rate": 1.4590358143056575e-05,
+      "loss": 1.0164,
       "step": 12960
     },
     {
-      "epoch": 0.3559638570761583,
+      "epoch": 0.3677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.4926381692788827e-05,
-      "loss": 0.9747,
+      "learning_rate": 1.4589541598388865e-05,
+      "loss": 0.9271,
       "step": 12961
     },
     {
-      "epoch": 0.3559913213040015,
+      "epoch": 0.36782065834279226,
       "grad_norm": 0.0,
-      "learning_rate": 1.4925607586007994e-05,
-      "loss": 0.9166,
+      "learning_rate": 1.4588725014953096e-05,
+      "loss": 0.8732,
       "step": 12962
     },
     {
-      "epoch": 0.3560187855318448,
+      "epoch": 0.3678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.4924833440253843e-05,
-      "loss": 0.8377,
+      "learning_rate": 1.4587908392756159e-05,
+      "loss": 0.863,
       "step": 12963
     },
     {
-      "epoch": 0.356046249759688,
+      "epoch": 0.3678774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.4924059255532506e-05,
-      "loss": 0.9489,
+      "learning_rate": 1.458709173180495e-05,
+      "loss": 0.9364,
       "step": 12964
     },
     {
-      "epoch": 0.35607371398753124,
+      "epoch": 0.36790578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.4923285031850105e-05,
-      "loss": 1.018,
+      "learning_rate": 1.4586275032106373e-05,
+      "loss": 0.9398,
       "step": 12965
     },
     {
-      "epoch": 0.35610117821537446,
+      "epoch": 0.36793416572077187,
       "grad_norm": 0.0,
-      "learning_rate": 1.4922510769212763e-05,
-      "loss": 1.0149,
+      "learning_rate": 1.4585458293667327e-05,
+      "loss": 1.0085,
       "step": 12966
     },
     {
-      "epoch": 0.35612864244321774,
+      "epoch": 0.3679625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.4921736467626612e-05,
-      "loss": 0.976,
+      "learning_rate": 1.4584641516494706e-05,
+      "loss": 0.8472,
       "step": 12967
     },
     {
-      "epoch": 0.35615610667106096,
+      "epoch": 0.36799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.4920962127097776e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.4583824700595411e-05,
+      "loss": 0.9815,
       "step": 12968
     },
     {
-      "epoch": 0.3561835708989042,
+      "epoch": 0.3680192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.4920187747632378e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.4583007845976345e-05,
+      "loss": 0.826,
       "step": 12969
     },
     {
-      "epoch": 0.3562110351267474,
+      "epoch": 0.3680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.4919413329236552e-05,
-      "loss": 1.0149,
+      "learning_rate": 1.4582190952644407e-05,
+      "loss": 0.9597,
       "step": 12970
     },
     {
-      "epoch": 0.3562384993545906,
+      "epoch": 0.36807604994324633,
       "grad_norm": 0.0,
-      "learning_rate": 1.4918638871916418e-05,
-      "loss": 0.9814,
+      "learning_rate": 1.4581374020606492e-05,
+      "loss": 0.9495,
       "step": 12971
     },
     {
-      "epoch": 0.3562659635824339,
+      "epoch": 0.3681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.491786437567811e-05,
-      "loss": 0.8374,
+      "learning_rate": 1.4580557049869507e-05,
+      "loss": 1.0104,
       "step": 12972
     },
     {
-      "epoch": 0.3562934278102771,
+      "epoch": 0.3681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.4917089840527756e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.4579740040440351e-05,
+      "loss": 0.9424,
       "step": 12973
     },
     {
-      "epoch": 0.35632089203812034,
+      "epoch": 0.36816118047673096,
       "grad_norm": 0.0,
-      "learning_rate": 1.4916315266471482e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.4578922992325924e-05,
+      "loss": 1.0191,
       "step": 12974
     },
     {
-      "epoch": 0.35634835626596356,
+      "epoch": 0.3681895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.4915540653515415e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.4578105905533126e-05,
+      "loss": 0.8929,
       "step": 12975
     },
     {
-      "epoch": 0.35637582049380684,
+      "epoch": 0.3682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.4914766001665685e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.4577288780068868e-05,
+      "loss": 0.8642,
       "step": 12976
     },
     {
-      "epoch": 0.35640328472165006,
+      "epoch": 0.36824631101021565,
       "grad_norm": 0.0,
-      "learning_rate": 1.4913991310928428e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.4576471615940042e-05,
+      "loss": 0.9149,
       "step": 12977
     },
     {
-      "epoch": 0.3564307489494933,
+      "epoch": 0.36827468785471057,
       "grad_norm": 0.0,
-      "learning_rate": 1.4913216581309764e-05,
-      "loss": 0.9103,
+      "learning_rate": 1.4575654413153553e-05,
+      "loss": 0.9384,
       "step": 12978
     },
     {
-      "epoch": 0.3564582131773365,
+      "epoch": 0.36830306469920543,
       "grad_norm": 0.0,
-      "learning_rate": 1.4912441812815828e-05,
-      "loss": 0.9212,
+      "learning_rate": 1.4574837171716309e-05,
+      "loss": 0.9525,
       "step": 12979
     },
     {
-      "epoch": 0.3564856774051798,
+      "epoch": 0.36833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.491166700545275e-05,
-      "loss": 0.9683,
+      "learning_rate": 1.4574019891635209e-05,
+      "loss": 0.9303,
       "step": 12980
     },
     {
-      "epoch": 0.356513141633023,
+      "epoch": 0.36835981838819526,
       "grad_norm": 0.0,
-      "learning_rate": 1.4910892159226659e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.4573202572917156e-05,
+      "loss": 0.9749,
       "step": 12981
     },
     {
-      "epoch": 0.3565406058608662,
+      "epoch": 0.3683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.491011727414369e-05,
-      "loss": 0.9193,
+      "learning_rate": 1.4572385215569055e-05,
+      "loss": 0.9491,
       "step": 12982
     },
     {
-      "epoch": 0.35656807008870944,
+      "epoch": 0.36841657207718503,
       "grad_norm": 0.0,
-      "learning_rate": 1.4909342350209968e-05,
-      "loss": 0.9188,
+      "learning_rate": 1.4571567819597812e-05,
+      "loss": 0.9173,
       "step": 12983
     },
     {
-      "epoch": 0.35659553431655266,
+      "epoch": 0.3684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.490856738743163e-05,
-      "loss": 0.8236,
+      "learning_rate": 1.457075038501033e-05,
+      "loss": 0.8445,
       "step": 12984
     },
     {
-      "epoch": 0.35662299854439594,
+      "epoch": 0.3684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.4907792385814803e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.4569932911813514e-05,
+      "loss": 0.9032,
       "step": 12985
     },
     {
-      "epoch": 0.35665046277223916,
+      "epoch": 0.36850170261066967,
       "grad_norm": 0.0,
-      "learning_rate": 1.4907017345365622e-05,
-      "loss": 0.9655,
+      "learning_rate": 1.4569115400014269e-05,
+      "loss": 1.0009,
       "step": 12986
     },
     {
-      "epoch": 0.3566779270000824,
+      "epoch": 0.3685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.4906242266090218e-05,
-      "loss": 0.954,
+      "learning_rate": 1.4568297849619502e-05,
+      "loss": 0.8731,
       "step": 12987
     },
     {
-      "epoch": 0.3567053912279256,
+      "epoch": 0.3685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.490546714799473e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.4567480260636115e-05,
+      "loss": 0.9833,
       "step": 12988
     },
     {
-      "epoch": 0.3567328554557689,
+      "epoch": 0.36858683314415436,
       "grad_norm": 0.0,
-      "learning_rate": 1.4904691991085281e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.4566662633071019e-05,
+      "loss": 1.0111,
       "step": 12989
     },
     {
-      "epoch": 0.3567603196836121,
+      "epoch": 0.3686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.490391679536801e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.4565844966931119e-05,
+      "loss": 0.8573,
       "step": 12990
     },
     {
-      "epoch": 0.3567877839114553,
+      "epoch": 0.36864358683314413,
       "grad_norm": 0.0,
-      "learning_rate": 1.4903141560849051e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.4565027262223317e-05,
+      "loss": 0.9227,
       "step": 12991
     },
     {
-      "epoch": 0.35681524813929855,
+      "epoch": 0.36867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.4902366287534535e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.4564209518954528e-05,
+      "loss": 0.8974,
       "step": 12992
     },
     {
-      "epoch": 0.3568427123671418,
+      "epoch": 0.36870034052213396,
       "grad_norm": 0.0,
-      "learning_rate": 1.4901590975430598e-05,
-      "loss": 0.971,
+      "learning_rate": 1.4563391737131656e-05,
+      "loss": 0.8892,
       "step": 12993
     },
     {
-      "epoch": 0.35687017659498504,
+      "epoch": 0.3687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.4900815624543375e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.4562573916761609e-05,
+      "loss": 1.0071,
       "step": 12994
     },
     {
-      "epoch": 0.35689764082282827,
+      "epoch": 0.36875709421112374,
       "grad_norm": 0.0,
-      "learning_rate": 1.4900040234878998e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.4561756057851295e-05,
+      "loss": 0.9447,
       "step": 12995
     },
     {
-      "epoch": 0.3569251050506715,
+      "epoch": 0.3687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.489926480644361e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.4560938160407622e-05,
+      "loss": 0.985,
       "step": 12996
     },
     {
-      "epoch": 0.3569525692785147,
+      "epoch": 0.3688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.4898489339243337e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.45601202244375e-05,
+      "loss": 0.8234,
       "step": 12997
     },
     {
-      "epoch": 0.356980033506358,
+      "epoch": 0.3688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.489771383328432e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.4559302249947834e-05,
+      "loss": 0.8974,
       "step": 12998
     },
     {
-      "epoch": 0.3570074977342012,
+      "epoch": 0.3688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.4896938288572694e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.455848423694554e-05,
+      "loss": 1.0356,
       "step": 12999
     },
     {
-      "epoch": 0.35703496196204443,
+      "epoch": 0.3688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.4896162705114597e-05,
-      "loss": 0.912,
+      "learning_rate": 1.4557666185437526e-05,
+      "loss": 0.9615,
       "step": 13000
     },
     {
-      "epoch": 0.35706242618988765,
+      "epoch": 0.36892735527809306,
       "grad_norm": 0.0,
-      "learning_rate": 1.489538708291616e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.4556848095430698e-05,
+      "loss": 0.8345,
       "step": 13001
     },
     {
-      "epoch": 0.3570898904177309,
+      "epoch": 0.368955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.4894611421983527e-05,
-      "loss": 1.1011,
+      "learning_rate": 1.4556029966931972e-05,
+      "loss": 0.9146,
       "step": 13002
     },
     {
-      "epoch": 0.35711735464557415,
+      "epoch": 0.36898410896708284,
       "grad_norm": 0.0,
-      "learning_rate": 1.4893835722322832e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.4555211799948255e-05,
+      "loss": 1.0316,
       "step": 13003
     },
     {
-      "epoch": 0.35714481887341737,
+      "epoch": 0.36901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.4893059983940212e-05,
-      "loss": 0.956,
+      "learning_rate": 1.4554393594486457e-05,
+      "loss": 1.0104,
       "step": 13004
     },
     {
-      "epoch": 0.3571722831012606,
+      "epoch": 0.36904086265607267,
       "grad_norm": 0.0,
-      "learning_rate": 1.4892284206841808e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.4553575350553495e-05,
+      "loss": 0.7872,
       "step": 13005
     },
     {
-      "epoch": 0.35719974732910387,
+      "epoch": 0.3690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.4891508391033753e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.4552757068156275e-05,
+      "loss": 0.9327,
       "step": 13006
     },
     {
-      "epoch": 0.3572272115569471,
+      "epoch": 0.36909761634506244,
       "grad_norm": 0.0,
-      "learning_rate": 1.4890732536522192e-05,
-      "loss": 1.0237,
+      "learning_rate": 1.4551938747301712e-05,
+      "loss": 0.9256,
       "step": 13007
     },
     {
-      "epoch": 0.3572546757847903,
+      "epoch": 0.3691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.488995664331326e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.4551120387996718e-05,
+      "loss": 1.0081,
       "step": 13008
     },
     {
-      "epoch": 0.35728214001263353,
+      "epoch": 0.3691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.4889180711413096e-05,
-      "loss": 0.9768,
+      "learning_rate": 1.4550301990248205e-05,
+      "loss": 0.9814,
       "step": 13009
     },
     {
-      "epoch": 0.35730960424047675,
+      "epoch": 0.36918274687854713,
       "grad_norm": 0.0,
-      "learning_rate": 1.488840474082784e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.454948355406309e-05,
+      "loss": 1.0334,
       "step": 13010
     },
     {
-      "epoch": 0.35733706846832003,
+      "epoch": 0.369211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.4887628731563635e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.4548665079448276e-05,
+      "loss": 1.0637,
       "step": 13011
     },
     {
-      "epoch": 0.35736453269616325,
+      "epoch": 0.3692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.4886852683626616e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.454784656641069e-05,
+      "loss": 0.8897,
       "step": 13012
     },
     {
-      "epoch": 0.35739199692400647,
+      "epoch": 0.36926787741203176,
       "grad_norm": 0.0,
-      "learning_rate": 1.4886076597022925e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.4547028014957238e-05,
+      "loss": 0.9597,
       "step": 13013
     },
     {
-      "epoch": 0.3574194611518497,
+      "epoch": 0.3692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.4885300471758703e-05,
-      "loss": 0.8605,
+      "learning_rate": 1.4546209425094832e-05,
+      "loss": 1.0327,
       "step": 13014
     },
     {
-      "epoch": 0.35744692537969297,
+      "epoch": 0.36932463110102154,
       "grad_norm": 0.0,
-      "learning_rate": 1.4884524307840091e-05,
-      "loss": 0.9846,
+      "learning_rate": 1.4545390796830394e-05,
+      "loss": 0.8462,
       "step": 13015
     },
     {
-      "epoch": 0.3574743896075362,
+      "epoch": 0.36935300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4883748105273235e-05,
-      "loss": 1.0132,
+      "learning_rate": 1.4544572130170837e-05,
+      "loss": 1.0142,
       "step": 13016
     },
     {
-      "epoch": 0.3575018538353794,
+      "epoch": 0.36938138479001137,
       "grad_norm": 0.0,
-      "learning_rate": 1.4882971864064268e-05,
-      "loss": 0.9629,
+      "learning_rate": 1.4543753425123073e-05,
+      "loss": 0.8584,
       "step": 13017
     },
     {
-      "epoch": 0.35752931806322263,
+      "epoch": 0.36940976163450623,
       "grad_norm": 0.0,
-      "learning_rate": 1.4882195584219336e-05,
-      "loss": 1.0499,
+      "learning_rate": 1.4542934681694019e-05,
+      "loss": 0.8669,
       "step": 13018
     },
     {
-      "epoch": 0.3575567822910659,
+      "epoch": 0.36943813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.4881419265744583e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.4542115899890594e-05,
+      "loss": 0.8874,
       "step": 13019
     },
     {
-      "epoch": 0.35758424651890913,
+      "epoch": 0.369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.488064290864615e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.454129707971971e-05,
+      "loss": 0.8217,
       "step": 13020
     },
     {
-      "epoch": 0.35761171074675235,
+      "epoch": 0.3694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.487986651293018e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.4540478221188284e-05,
+      "loss": 1.0011,
       "step": 13021
     },
     {
-      "epoch": 0.3576391749745956,
+      "epoch": 0.36952326901248583,
       "grad_norm": 0.0,
-      "learning_rate": 1.4879090078602813e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.4539659324303235e-05,
+      "loss": 0.828,
       "step": 13022
     },
     {
-      "epoch": 0.3576666392024388,
+      "epoch": 0.3695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.4878313605670197e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.4538840389071482e-05,
+      "loss": 0.9319,
       "step": 13023
     },
     {
-      "epoch": 0.3576941034302821,
+      "epoch": 0.3695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.4877537094138477e-05,
-      "loss": 0.959,
+      "learning_rate": 1.4538021415499938e-05,
+      "loss": 0.9898,
       "step": 13024
     },
     {
-      "epoch": 0.3577215676581253,
+      "epoch": 0.36960839954597047,
       "grad_norm": 0.0,
-      "learning_rate": 1.487676054401379e-05,
-      "loss": 0.8762,
+      "learning_rate": 1.4537202403595523e-05,
+      "loss": 0.9712,
       "step": 13025
     },
     {
-      "epoch": 0.3577490318859685,
+      "epoch": 0.3696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.4875983955302287e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.4536383353365158e-05,
+      "loss": 0.8665,
       "step": 13026
     },
     {
-      "epoch": 0.35777649611381174,
+      "epoch": 0.3696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.4875207328010109e-05,
-      "loss": 1.0485,
+      "learning_rate": 1.4535564264815757e-05,
+      "loss": 0.9919,
       "step": 13027
     },
     {
-      "epoch": 0.357803960341655,
+      "epoch": 0.36969353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.4874430662143402e-05,
-      "loss": 0.8835,
+      "learning_rate": 1.453474513795424e-05,
+      "loss": 1.0788,
       "step": 13028
     },
     {
-      "epoch": 0.35783142456949824,
+      "epoch": 0.36972190692395007,
       "grad_norm": 0.0,
-      "learning_rate": 1.4873653957708315e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.4533925972787526e-05,
+      "loss": 0.923,
       "step": 13029
     },
     {
-      "epoch": 0.35785888879734146,
+      "epoch": 0.36975028376844493,
       "grad_norm": 0.0,
-      "learning_rate": 1.4872877214710983e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.4533106769322538e-05,
+      "loss": 0.8911,
       "step": 13030
     },
     {
-      "epoch": 0.3578863530251847,
+      "epoch": 0.36977866061293985,
       "grad_norm": 0.0,
-      "learning_rate": 1.4872100433157561e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.4532287527566193e-05,
+      "loss": 0.9019,
       "step": 13031
     },
     {
-      "epoch": 0.35791381725302795,
+      "epoch": 0.3698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.4871323613054196e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.4531468247525411e-05,
+      "loss": 0.9722,
       "step": 13032
     },
     {
-      "epoch": 0.3579412814808712,
+      "epoch": 0.3698354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.4870546754407032e-05,
-      "loss": 1.0369,
+      "learning_rate": 1.4530648929207114e-05,
+      "loss": 0.8823,
       "step": 13033
     },
     {
-      "epoch": 0.3579687457087144,
+      "epoch": 0.36986379114642454,
       "grad_norm": 0.0,
-      "learning_rate": 1.4869769857222211e-05,
-      "loss": 1.0044,
+      "learning_rate": 1.4529829572618222e-05,
+      "loss": 0.9095,
       "step": 13034
     },
     {
-      "epoch": 0.3579962099365576,
+      "epoch": 0.3698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.4868992921505889e-05,
-      "loss": 0.9591,
+      "learning_rate": 1.4529010177765654e-05,
+      "loss": 0.9082,
       "step": 13035
     },
     {
-      "epoch": 0.35802367416440084,
+      "epoch": 0.3699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.4868215947264208e-05,
-      "loss": 0.94,
+      "learning_rate": 1.4528190744656334e-05,
+      "loss": 0.9335,
       "step": 13036
     },
     {
-      "epoch": 0.3580511383922441,
+      "epoch": 0.36994892167990917,
       "grad_norm": 0.0,
-      "learning_rate": 1.4867438934503316e-05,
-      "loss": 0.909,
+      "learning_rate": 1.4527371273297184e-05,
+      "loss": 0.9746,
       "step": 13037
     },
     {
-      "epoch": 0.35807860262008734,
+      "epoch": 0.3699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.4866661883229361e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.4526551763695124e-05,
+      "loss": 1.0002,
       "step": 13038
     },
     {
-      "epoch": 0.35810606684793056,
+      "epoch": 0.370005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.4865884793448493e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.4525732215857078e-05,
+      "loss": 0.862,
       "step": 13039
     },
     {
-      "epoch": 0.3581335310757738,
+      "epoch": 0.37003405221339386,
       "grad_norm": 0.0,
-      "learning_rate": 1.486510766516686e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.452491262978997e-05,
+      "loss": 1.0093,
       "step": 13040
     },
     {
-      "epoch": 0.35816099530361706,
+      "epoch": 0.3700624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.486433049839061e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.4524093005500721e-05,
+      "loss": 0.9817,
       "step": 13041
     },
     {
-      "epoch": 0.3581884595314603,
+      "epoch": 0.37009080590238363,
       "grad_norm": 0.0,
-      "learning_rate": 1.4863553293125891e-05,
-      "loss": 0.9016,
+      "learning_rate": 1.4523273342996254e-05,
+      "loss": 0.9019,
       "step": 13042
     },
     {
-      "epoch": 0.3582159237593035,
+      "epoch": 0.37011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.4862776049378855e-05,
-      "loss": 0.9083,
+      "learning_rate": 1.4522453642283496e-05,
+      "loss": 0.9972,
       "step": 13043
     },
     {
-      "epoch": 0.3582433879871467,
+      "epoch": 0.37014755959137347,
       "grad_norm": 0.0,
-      "learning_rate": 1.4861998767155654e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.4521633903369365e-05,
+      "loss": 1.046,
       "step": 13044
     },
     {
-      "epoch": 0.35827085221499,
+      "epoch": 0.3701759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.4861221446462435e-05,
-      "loss": 0.9116,
+      "learning_rate": 1.4520814126260791e-05,
+      "loss": 0.9724,
       "step": 13045
     },
     {
-      "epoch": 0.3582983164428332,
+      "epoch": 0.37020431328036324,
       "grad_norm": 0.0,
-      "learning_rate": 1.4860444087305346e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.4519994310964697e-05,
+      "loss": 0.9466,
       "step": 13046
     },
     {
-      "epoch": 0.35832578067067644,
+      "epoch": 0.3702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.4859666689690544e-05,
-      "loss": 0.8876,
+      "learning_rate": 1.451917445748801e-05,
+      "loss": 1.0786,
       "step": 13047
     },
     {
-      "epoch": 0.35835324489851966,
+      "epoch": 0.370261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.4858889253624176e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.4518354565837652e-05,
+      "loss": 0.903,
       "step": 13048
     },
     {
-      "epoch": 0.35838070912636294,
+      "epoch": 0.3702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.485811177911239e-05,
-      "loss": 1.0649,
+      "learning_rate": 1.451753463602055e-05,
+      "loss": 1.0561,
       "step": 13049
     },
     {
-      "epoch": 0.35840817335420616,
+      "epoch": 0.3703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.4857334266161346e-05,
-      "loss": 0.9616,
+      "learning_rate": 1.451671466804363e-05,
+      "loss": 0.9779,
       "step": 13050
     },
     {
-      "epoch": 0.3584356375820494,
+      "epoch": 0.3703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.4856556714777188e-05,
-      "loss": 0.8471,
+      "learning_rate": 1.4515894661913817e-05,
+      "loss": 0.9562,
       "step": 13051
     },
     {
-      "epoch": 0.3584631018098926,
+      "epoch": 0.37037457434733256,
       "grad_norm": 0.0,
-      "learning_rate": 1.4855779124966076e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.4515074617638036e-05,
+      "loss": 0.9142,
       "step": 13052
     },
     {
-      "epoch": 0.3584905660377358,
+      "epoch": 0.3704029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.4855001496734158e-05,
-      "loss": 0.8929,
+      "learning_rate": 1.451425453522322e-05,
+      "loss": 0.8959,
       "step": 13053
     },
     {
-      "epoch": 0.3585180302655791,
+      "epoch": 0.37043132803632234,
       "grad_norm": 0.0,
-      "learning_rate": 1.4854223830087586e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.4513434414676293e-05,
+      "loss": 0.8534,
       "step": 13054
     },
     {
-      "epoch": 0.3585454944934223,
+      "epoch": 0.37045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.4853446125032514e-05,
-      "loss": 0.8083,
+      "learning_rate": 1.4512614256004182e-05,
+      "loss": 0.9822,
       "step": 13055
     },
     {
-      "epoch": 0.35857295872126554,
+      "epoch": 0.37048808172531217,
       "grad_norm": 0.0,
-      "learning_rate": 1.4852668381575099e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.4511794059213818e-05,
+      "loss": 0.8193,
       "step": 13056
     },
     {
-      "epoch": 0.35860042294910877,
+      "epoch": 0.37051645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.485189059972149e-05,
-      "loss": 0.9851,
+      "learning_rate": 1.4510973824312125e-05,
+      "loss": 0.9896,
       "step": 13057
     },
     {
-      "epoch": 0.35862788717695204,
+      "epoch": 0.37054483541430194,
       "grad_norm": 0.0,
-      "learning_rate": 1.4851112779477845e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.4510153551306032e-05,
+      "loss": 0.8315,
       "step": 13058
     },
     {
-      "epoch": 0.35865535140479526,
+      "epoch": 0.3705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.4850334920850316e-05,
-      "loss": 1.0097,
+      "learning_rate": 1.450933324020247e-05,
+      "loss": 0.826,
       "step": 13059
     },
     {
-      "epoch": 0.3586828156326385,
+      "epoch": 0.3706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.4849557023845057e-05,
-      "loss": 0.9484,
+      "learning_rate": 1.4508512891008368e-05,
+      "loss": 0.907,
       "step": 13060
     },
     {
-      "epoch": 0.3587102798604817,
+      "epoch": 0.37062996594778663,
       "grad_norm": 0.0,
-      "learning_rate": 1.4848779088468226e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.4507692503730654e-05,
+      "loss": 0.9649,
       "step": 13061
     },
     {
-      "epoch": 0.358737744088325,
+      "epoch": 0.3706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.4848001114725975e-05,
-      "loss": 1.0078,
+      "learning_rate": 1.4506872078376257e-05,
+      "loss": 0.9517,
       "step": 13062
     },
     {
-      "epoch": 0.3587652083161682,
+      "epoch": 0.3706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.4847223102624459e-05,
-      "loss": 0.8717,
+      "learning_rate": 1.4506051614952114e-05,
+      "loss": 1.0138,
       "step": 13063
     },
     {
-      "epoch": 0.3587926725440114,
+      "epoch": 0.37071509648127127,
       "grad_norm": 0.0,
-      "learning_rate": 1.484644505216984e-05,
-      "loss": 0.8854,
+      "learning_rate": 1.4505231113465148e-05,
+      "loss": 0.883,
       "step": 13064
     },
     {
-      "epoch": 0.35882013677185465,
+      "epoch": 0.3707434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.484566696336827e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.4504410573922292e-05,
+      "loss": 0.966,
       "step": 13065
     },
     {
-      "epoch": 0.35884760099969787,
+      "epoch": 0.37077185017026104,
       "grad_norm": 0.0,
-      "learning_rate": 1.4844888836225904e-05,
-      "loss": 1.0529,
+      "learning_rate": 1.4503589996330478e-05,
+      "loss": 1.0065,
       "step": 13066
     },
     {
-      "epoch": 0.35887506522754115,
+      "epoch": 0.37080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.4844110670748904e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.4502769380696635e-05,
+      "loss": 1.0011,
       "step": 13067
     },
     {
-      "epoch": 0.35890252945538437,
+      "epoch": 0.37082860385925087,
       "grad_norm": 0.0,
-      "learning_rate": 1.484333246694342e-05,
-      "loss": 0.9245,
+      "learning_rate": 1.45019487270277e-05,
+      "loss": 0.9388,
       "step": 13068
     },
     {
-      "epoch": 0.3589299936832276,
+      "epoch": 0.37085698070374573,
       "grad_norm": 0.0,
-      "learning_rate": 1.4842554224815616e-05,
-      "loss": 0.977,
+      "learning_rate": 1.4501128035330595e-05,
+      "loss": 1.0318,
       "step": 13069
     },
     {
-      "epoch": 0.3589574579110708,
+      "epoch": 0.37088535754824065,
       "grad_norm": 0.0,
-      "learning_rate": 1.4841775944371648e-05,
-      "loss": 0.9592,
+      "learning_rate": 1.4500307305612267e-05,
+      "loss": 1.0507,
       "step": 13070
     },
     {
-      "epoch": 0.3589849221389141,
+      "epoch": 0.3709137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.4840997625617668e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.4499486537879639e-05,
+      "loss": 0.9053,
       "step": 13071
     },
     {
-      "epoch": 0.3590123863667573,
+      "epoch": 0.3709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.4840219268559845e-05,
-      "loss": 0.8089,
+      "learning_rate": 1.4498665732139643e-05,
+      "loss": 0.8468,
       "step": 13072
     },
     {
-      "epoch": 0.35903985059460053,
+      "epoch": 0.37097048808172534,
       "grad_norm": 0.0,
-      "learning_rate": 1.4839440873204332e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.4497844888399216e-05,
+      "loss": 1.0218,
       "step": 13073
     },
     {
-      "epoch": 0.35906731482244375,
+      "epoch": 0.3709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.4838662439557286e-05,
-      "loss": 0.8829,
+      "learning_rate": 1.4497024006665295e-05,
+      "loss": 0.981,
       "step": 13074
     },
     {
-      "epoch": 0.359094779050287,
+      "epoch": 0.3710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.4837883967624866e-05,
-      "loss": 1.0155,
+      "learning_rate": 1.4496203086944805e-05,
+      "loss": 0.9108,
       "step": 13075
     },
     {
-      "epoch": 0.35912224327813025,
+      "epoch": 0.37105561861520997,
       "grad_norm": 0.0,
-      "learning_rate": 1.4837105457413239e-05,
-      "loss": 1.0006,
+      "learning_rate": 1.4495382129244686e-05,
+      "loss": 0.9088,
       "step": 13076
     },
     {
-      "epoch": 0.35914970750597347,
+      "epoch": 0.3710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.483632690892856e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.4494561133571873e-05,
+      "loss": 1.0029,
       "step": 13077
     },
     {
-      "epoch": 0.3591771717338167,
+      "epoch": 0.3711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.4835548322176984e-05,
-      "loss": 0.8701,
+      "learning_rate": 1.4493740099933302e-05,
+      "loss": 0.9778,
       "step": 13078
     },
     {
-      "epoch": 0.3592046359616599,
+      "epoch": 0.37114074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.4834769697164681e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.4492919028335905e-05,
+      "loss": 0.9711,
       "step": 13079
     },
     {
-      "epoch": 0.3592321001895032,
+      "epoch": 0.3711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.4833991033897807e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.4492097918786621e-05,
+      "loss": 0.9512,
       "step": 13080
     },
     {
-      "epoch": 0.3592595644173464,
+      "epoch": 0.37119750283768443,
       "grad_norm": 0.0,
-      "learning_rate": 1.483321233238252e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.4491276771292382e-05,
+      "loss": 0.9422,
       "step": 13081
     },
     {
-      "epoch": 0.35928702864518963,
+      "epoch": 0.37122587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 1.4832433592624988e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.4490455585860124e-05,
+      "loss": 1.0333,
       "step": 13082
     },
     {
-      "epoch": 0.35931449287303285,
+      "epoch": 0.3712542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.4831654814631369e-05,
-      "loss": 1.0461,
+      "learning_rate": 1.4489634362496787e-05,
+      "loss": 0.8991,
       "step": 13083
     },
     {
-      "epoch": 0.35934195710087613,
+      "epoch": 0.3712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.4830875998407827e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.4488813101209309e-05,
+      "loss": 0.9798,
       "step": 13084
     },
     {
-      "epoch": 0.35936942132871935,
+      "epoch": 0.37131101021566404,
       "grad_norm": 0.0,
-      "learning_rate": 1.4830097143960523e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.4487991802004625e-05,
+      "loss": 0.8631,
       "step": 13085
     },
     {
-      "epoch": 0.3593968855565626,
+      "epoch": 0.3713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.4829318251295615e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.4487170464889668e-05,
+      "loss": 0.9131,
       "step": 13086
     },
     {
-      "epoch": 0.3594243497844058,
+      "epoch": 0.3713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.4828539320419271e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.4486349089871386e-05,
+      "loss": 0.8568,
       "step": 13087
     },
     {
-      "epoch": 0.35945181401224907,
+      "epoch": 0.3713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.4827760351337657e-05,
-      "loss": 0.9008,
+      "learning_rate": 1.448552767695671e-05,
+      "loss": 0.9304,
       "step": 13088
     },
     {
-      "epoch": 0.3594792782400923,
+      "epoch": 0.3714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.4826981344056931e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.4484706226152576e-05,
+      "loss": 0.9089,
       "step": 13089
     },
     {
-      "epoch": 0.3595067424679355,
+      "epoch": 0.3714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.4826202298583259e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.448388473746593e-05,
+      "loss": 0.9667,
       "step": 13090
     },
     {
-      "epoch": 0.35953420669577874,
+      "epoch": 0.37148127128263336,
       "grad_norm": 0.0,
-      "learning_rate": 1.4825423214922804e-05,
-      "loss": 0.8512,
+      "learning_rate": 1.448306321090371e-05,
+      "loss": 0.9475,
       "step": 13091
     },
     {
-      "epoch": 0.35956167092362196,
+      "epoch": 0.3715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.4824644093081735e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.448224164647285e-05,
+      "loss": 0.9407,
       "step": 13092
     },
     {
-      "epoch": 0.35958913515146523,
+      "epoch": 0.37153802497162314,
       "grad_norm": 0.0,
-      "learning_rate": 1.482386493306621e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.4481420044180295e-05,
+      "loss": 0.8365,
       "step": 13093
     },
     {
-      "epoch": 0.35961659937930845,
+      "epoch": 0.37156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.4823085734882398e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.4480598404032984e-05,
+      "loss": 0.9959,
       "step": 13094
     },
     {
-      "epoch": 0.3596440636071517,
+      "epoch": 0.3715947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.482230649853646e-05,
-      "loss": 0.9248,
+      "learning_rate": 1.4479776726037858e-05,
+      "loss": 0.9772,
       "step": 13095
     },
     {
-      "epoch": 0.3596715278349949,
+      "epoch": 0.3716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.4821527224034569e-05,
-      "loss": 0.8781,
+      "learning_rate": 1.4478955010201855e-05,
+      "loss": 1.0157,
       "step": 13096
     },
     {
-      "epoch": 0.3596989920628382,
+      "epoch": 0.37165153234960274,
       "grad_norm": 0.0,
-      "learning_rate": 1.4820747911382886e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.4478133256531917e-05,
+      "loss": 0.8354,
       "step": 13097
     },
     {
-      "epoch": 0.3597264562906814,
+      "epoch": 0.3716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.4819968560587578e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.4477311465034984e-05,
+      "loss": 0.9895,
       "step": 13098
     },
     {
-      "epoch": 0.3597539205185246,
+      "epoch": 0.3717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.4819189171654812e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.4476489635718001e-05,
+      "loss": 0.8467,
       "step": 13099
     },
     {
-      "epoch": 0.35978138474636784,
+      "epoch": 0.3717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.4818409744590753e-05,
-      "loss": 0.8701,
+      "learning_rate": 1.447566776858791e-05,
+      "loss": 0.9958,
       "step": 13100
     },
     {
-      "epoch": 0.3598088489742111,
+      "epoch": 0.3717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.4817630279401572e-05,
-      "loss": 0.8688,
+      "learning_rate": 1.4474845863651655e-05,
+      "loss": 1.0245,
       "step": 13101
     },
     {
-      "epoch": 0.35983631320205434,
+      "epoch": 0.3717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.4816850776093432e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.4474023920916174e-05,
+      "loss": 0.9484,
       "step": 13102
     },
     {
-      "epoch": 0.35986377742989756,
+      "epoch": 0.37182179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 1.4816071234672505e-05,
-      "loss": 0.8832,
+      "learning_rate": 1.447320194038841e-05,
+      "loss": 0.8465,
       "step": 13103
     },
     {
-      "epoch": 0.3598912416577408,
+      "epoch": 0.371850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.4815291655144957e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.4472379922075308e-05,
+      "loss": 1.0231,
       "step": 13104
     },
     {
-      "epoch": 0.359918705885584,
+      "epoch": 0.37187854710556184,
       "grad_norm": 0.0,
-      "learning_rate": 1.481451203751695e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.4471557865983816e-05,
+      "loss": 1.0349,
       "step": 13105
     },
     {
-      "epoch": 0.3599461701134273,
+      "epoch": 0.37190692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.4813732381794666e-05,
-      "loss": 0.9263,
+      "learning_rate": 1.4470735772120868e-05,
+      "loss": 0.967,
       "step": 13106
     },
     {
-      "epoch": 0.3599736343412705,
+      "epoch": 0.37193530079455167,
       "grad_norm": 0.0,
-      "learning_rate": 1.4812952687984264e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.4469913640493417e-05,
+      "loss": 0.8462,
       "step": 13107
     },
     {
-      "epoch": 0.3600010985691137,
+      "epoch": 0.37196367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 1.4812172956091914e-05,
-      "loss": 0.8662,
+      "learning_rate": 1.4469091471108403e-05,
+      "loss": 0.9681,
       "step": 13108
     },
     {
-      "epoch": 0.36002856279695694,
+      "epoch": 0.37199205448354145,
       "grad_norm": 0.0,
-      "learning_rate": 1.4811393186123792e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.4468269263972773e-05,
+      "loss": 0.9159,
       "step": 13109
     },
     {
-      "epoch": 0.3600560270248002,
+      "epoch": 0.3720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.481061337808606e-05,
-      "loss": 1.0014,
+      "learning_rate": 1.4467447019093473e-05,
+      "loss": 0.9747,
       "step": 13110
     },
     {
-      "epoch": 0.36008349125264344,
+      "epoch": 0.3720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.4809833531984892e-05,
-      "loss": 1.0354,
+      "learning_rate": 1.4466624736477447e-05,
+      "loss": 1.0234,
       "step": 13111
     },
     {
-      "epoch": 0.36011095548048666,
+      "epoch": 0.3720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.480905364782646e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.446580241613164e-05,
+      "loss": 0.8799,
       "step": 13112
     },
     {
-      "epoch": 0.3601384197083299,
+      "epoch": 0.372105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.480827372561693e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.4464980058062999e-05,
+      "loss": 1.0431,
       "step": 13113
     },
     {
-      "epoch": 0.36016588393617316,
+      "epoch": 0.3721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.4807493765362475e-05,
-      "loss": 1.0462,
+      "learning_rate": 1.446415766227847e-05,
+      "loss": 0.962,
       "step": 13114
     },
     {
-      "epoch": 0.3601933481640164,
+      "epoch": 0.37216231555051077,
       "grad_norm": 0.0,
-      "learning_rate": 1.4806713767069272e-05,
-      "loss": 1.0399,
+      "learning_rate": 1.4463335228785005e-05,
+      "loss": 0.8046,
       "step": 13115
     },
     {
-      "epoch": 0.3602208123918596,
+      "epoch": 0.3721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.4805933730743484e-05,
-      "loss": 0.8733,
+      "learning_rate": 1.4462512757589541e-05,
+      "loss": 0.9419,
       "step": 13116
     },
     {
-      "epoch": 0.3602482766197028,
+      "epoch": 0.37221906923950054,
       "grad_norm": 0.0,
-      "learning_rate": 1.480515365639129e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.4461690248699034e-05,
+      "loss": 0.8796,
       "step": 13117
     },
     {
-      "epoch": 0.36027574084754604,
+      "epoch": 0.37224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4804373544018853e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.4460867702120432e-05,
+      "loss": 0.8823,
       "step": 13118
     },
     {
-      "epoch": 0.3603032050753893,
+      "epoch": 0.3722758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.4803593393632354e-05,
-      "loss": 1.0068,
+      "learning_rate": 1.4460045117860674e-05,
+      "loss": 0.8778,
       "step": 13119
     },
     {
-      "epoch": 0.36033066930323254,
+      "epoch": 0.37230419977298523,
       "grad_norm": 0.0,
-      "learning_rate": 1.4802813205237964e-05,
-      "loss": 0.9389,
+      "learning_rate": 1.4459222495926719e-05,
+      "loss": 0.8175,
       "step": 13120
     },
     {
-      "epoch": 0.36035813353107576,
+      "epoch": 0.37233257661748015,
       "grad_norm": 0.0,
-      "learning_rate": 1.4802032978841857e-05,
-      "loss": 0.8569,
+      "learning_rate": 1.4458399836325508e-05,
+      "loss": 0.9158,
       "step": 13121
     },
     {
-      "epoch": 0.360385597758919,
+      "epoch": 0.372360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.4801252714450203e-05,
-      "loss": 0.8794,
+      "learning_rate": 1.4457577139063995e-05,
+      "loss": 0.9,
       "step": 13122
     },
     {
-      "epoch": 0.36041306198676226,
+      "epoch": 0.3723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.4800472412069177e-05,
-      "loss": 0.8732,
+      "learning_rate": 1.4456754404149124e-05,
+      "loss": 0.8988,
       "step": 13123
     },
     {
-      "epoch": 0.3604405262146055,
+      "epoch": 0.37241770715096484,
       "grad_norm": 0.0,
-      "learning_rate": 1.4799692071704954e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.4455931631587854e-05,
+      "loss": 0.9172,
       "step": 13124
     },
     {
-      "epoch": 0.3604679904424487,
+      "epoch": 0.3724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.4798911693363712e-05,
-      "loss": 0.8535,
+      "learning_rate": 1.4455108821387124e-05,
+      "loss": 0.8142,
       "step": 13125
     },
     {
-      "epoch": 0.3604954546702919,
+      "epoch": 0.3724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4798131277051617e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.4454285973553891e-05,
+      "loss": 0.9551,
       "step": 13126
     },
     {
-      "epoch": 0.3605229188981352,
+      "epoch": 0.3725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.4797350822774852e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.4453463088095108e-05,
+      "loss": 0.9703,
       "step": 13127
     },
     {
-      "epoch": 0.3605503831259784,
+      "epoch": 0.3725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.4796570330539588e-05,
-      "loss": 0.8848,
+      "learning_rate": 1.445264016501772e-05,
+      "loss": 0.8816,
       "step": 13128
     },
     {
-      "epoch": 0.36057784735382165,
+      "epoch": 0.37255959137343925,
       "grad_norm": 0.0,
-      "learning_rate": 1.4795789800352e-05,
-      "loss": 0.9041,
+      "learning_rate": 1.4451817204328678e-05,
+      "loss": 0.8533,
       "step": 13129
     },
     {
-      "epoch": 0.36060531158166487,
+      "epoch": 0.37258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.4795009232218265e-05,
-      "loss": 0.8288,
+      "learning_rate": 1.4450994206034936e-05,
+      "loss": 0.9131,
       "step": 13130
     },
     {
-      "epoch": 0.3606327758095081,
+      "epoch": 0.3726163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.479422862614456e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.4450171170143447e-05,
+      "loss": 1.0147,
       "step": 13131
     },
     {
-      "epoch": 0.36066024003735137,
+      "epoch": 0.37264472190692394,
       "grad_norm": 0.0,
-      "learning_rate": 1.4793447982137064e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.4449348096661164e-05,
+      "loss": 0.9507,
       "step": 13132
     },
     {
-      "epoch": 0.3606877042651946,
+      "epoch": 0.37267309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.479266730020195e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.4448524985595033e-05,
+      "loss": 0.8923,
       "step": 13133
     },
     {
-      "epoch": 0.3607151684930378,
+      "epoch": 0.3727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.4791886580345391e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.4447701836952017e-05,
+      "loss": 0.997,
       "step": 13134
     },
     {
-      "epoch": 0.36074263272088103,
+      "epoch": 0.3727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.4791105822573575e-05,
-      "loss": 0.9414,
+      "learning_rate": 1.444687865073906e-05,
+      "loss": 0.8665,
       "step": 13135
     },
     {
-      "epoch": 0.3607700969487243,
+      "epoch": 0.37275822928490354,
       "grad_norm": 0.0,
-      "learning_rate": 1.479032502689267e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.4446055426963121e-05,
+      "loss": 0.9351,
       "step": 13136
     },
     {
-      "epoch": 0.3607975611765675,
+      "epoch": 0.3727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.478954419330886e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.444523216563115e-05,
+      "loss": 0.8895,
       "step": 13137
     },
     {
-      "epoch": 0.36082502540441075,
+      "epoch": 0.3728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.478876332182832e-05,
-      "loss": 0.9588,
+      "learning_rate": 1.4444408866750104e-05,
+      "loss": 1.0993,
       "step": 13138
     },
     {
-      "epoch": 0.36085248963225397,
+      "epoch": 0.3728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.478798241245723e-05,
-      "loss": 0.8827,
+      "learning_rate": 1.4443585530326938e-05,
+      "loss": 0.9674,
       "step": 13139
     },
     {
-      "epoch": 0.36087995386009725,
+      "epoch": 0.3728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.4787201465201771e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.4442762156368604e-05,
+      "loss": 1.0312,
       "step": 13140
     },
     {
-      "epoch": 0.36090741808794047,
+      "epoch": 0.372900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.4786420480068118e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.4441938744882058e-05,
+      "loss": 1.0238,
       "step": 13141
     },
     {
-      "epoch": 0.3609348823157837,
+      "epoch": 0.37292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.4785639457062449e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.4441115295874256e-05,
+      "loss": 0.9998,
       "step": 13142
     },
     {
-      "epoch": 0.3609623465436269,
+      "epoch": 0.3729568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.4784858396190948e-05,
-      "loss": 1.048,
+      "learning_rate": 1.4440291809352153e-05,
+      "loss": 0.9138,
       "step": 13143
     },
     {
-      "epoch": 0.3609898107714702,
+      "epoch": 0.37298524404086264,
       "grad_norm": 0.0,
-      "learning_rate": 1.4784077297459795e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.4439468285322705e-05,
+      "loss": 0.9292,
       "step": 13144
     },
     {
-      "epoch": 0.3610172749993134,
+      "epoch": 0.37301362088535756,
       "grad_norm": 0.0,
-      "learning_rate": 1.478329616087517e-05,
-      "loss": 0.9053,
+      "learning_rate": 1.443864472379287e-05,
+      "loss": 0.9471,
       "step": 13145
     },
     {
-      "epoch": 0.36104473922715663,
+      "epoch": 0.3730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.4782514986443253e-05,
-      "loss": 0.938,
+      "learning_rate": 1.4437821124769605e-05,
+      "loss": 1.0322,
       "step": 13146
     },
     {
-      "epoch": 0.36107220345499985,
+      "epoch": 0.37307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.4781733774170224e-05,
-      "loss": 0.9992,
+      "learning_rate": 1.4436997488259863e-05,
+      "loss": 0.9094,
       "step": 13147
     },
     {
-      "epoch": 0.3610996676828431,
+      "epoch": 0.37309875141884224,
       "grad_norm": 0.0,
-      "learning_rate": 1.4780952524062264e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.4436173814270604e-05,
+      "loss": 0.9663,
       "step": 13148
     },
     {
-      "epoch": 0.36112713191068635,
+      "epoch": 0.3731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.4780171236125555e-05,
-      "loss": 0.8121,
+      "learning_rate": 1.4435350102808787e-05,
+      "loss": 0.9785,
       "step": 13149
     },
     {
-      "epoch": 0.36115459613852957,
+      "epoch": 0.373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.4779389910366285e-05,
-      "loss": 0.9949,
+      "learning_rate": 1.4434526353881366e-05,
+      "loss": 0.9958,
       "step": 13150
     },
     {
-      "epoch": 0.3611820603663728,
+      "epoch": 0.3731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.4778608546790625e-05,
-      "loss": 0.8555,
+      "learning_rate": 1.4433702567495304e-05,
+      "loss": 1.0253,
       "step": 13151
     },
     {
-      "epoch": 0.361209524594216,
+      "epoch": 0.3732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.4777827145404766e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.4432878743657555e-05,
+      "loss": 0.8965,
       "step": 13152
     },
     {
-      "epoch": 0.3612369888220593,
+      "epoch": 0.3732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.4777045706214886e-05,
-      "loss": 1.0432,
+      "learning_rate": 1.443205488237508e-05,
+      "loss": 0.956,
       "step": 13153
     },
     {
-      "epoch": 0.3612644530499025,
+      "epoch": 0.37326901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 1.4776264229227171e-05,
-      "loss": 0.9172,
+      "learning_rate": 1.4431230983654839e-05,
+      "loss": 0.9866,
       "step": 13154
     },
     {
-      "epoch": 0.36129191727774573,
+      "epoch": 0.3732973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.4775482714447802e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.4430407047503792e-05,
+      "loss": 1.0255,
       "step": 13155
     },
     {
-      "epoch": 0.36131938150558895,
+      "epoch": 0.37332576617480134,
       "grad_norm": 0.0,
-      "learning_rate": 1.4774701161882965e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.4429583073928897e-05,
+      "loss": 0.8565,
       "step": 13156
     },
     {
-      "epoch": 0.36134684573343223,
+      "epoch": 0.37335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.4773919571538843e-05,
-      "loss": 1.0206,
+      "learning_rate": 1.4428759062937112e-05,
+      "loss": 0.9221,
       "step": 13157
     },
     {
-      "epoch": 0.36137430996127545,
+      "epoch": 0.3733825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.4773137943421621e-05,
-      "loss": 1.0445,
+      "learning_rate": 1.44279350145354e-05,
+      "loss": 0.8652,
       "step": 13158
     },
     {
-      "epoch": 0.3614017741891187,
+      "epoch": 0.37341089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.4772356277537481e-05,
-      "loss": 0.8601,
+      "learning_rate": 1.4427110928730727e-05,
+      "loss": 1.0637,
       "step": 13159
     },
     {
-      "epoch": 0.3614292384169619,
+      "epoch": 0.37343927355278095,
       "grad_norm": 0.0,
-      "learning_rate": 1.4771574573892611e-05,
-      "loss": 0.843,
+      "learning_rate": 1.4426286805530043e-05,
+      "loss": 0.8726,
       "step": 13160
     },
     {
-      "epoch": 0.3614567026448051,
+      "epoch": 0.3734676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.4770792832493192e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.4425462644940317e-05,
+      "loss": 0.9546,
       "step": 13161
     },
     {
-      "epoch": 0.3614841668726484,
+      "epoch": 0.3734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.4770011053345414e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.4424638446968513e-05,
+      "loss": 0.7975,
       "step": 13162
     },
     {
-      "epoch": 0.3615116311004916,
+      "epoch": 0.3735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.476922923645546e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.4423814211621586e-05,
+      "loss": 0.8386,
       "step": 13163
     },
     {
-      "epoch": 0.36153909532833484,
+      "epoch": 0.3735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.4768447381829518e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.4422989938906502e-05,
+      "loss": 1.0553,
       "step": 13164
     },
     {
-      "epoch": 0.36156655955617806,
+      "epoch": 0.3735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.4767665489473774e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.4422165628830223e-05,
+      "loss": 0.9945,
       "step": 13165
     },
     {
-      "epoch": 0.36159402378402133,
+      "epoch": 0.37360953461975027,
       "grad_norm": 0.0,
-      "learning_rate": 1.4766883559394412e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.4421341281399713e-05,
+      "loss": 0.9158,
       "step": 13166
     },
     {
-      "epoch": 0.36162148801186456,
+      "epoch": 0.3736379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.476610159159762e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.442051689662193e-05,
+      "loss": 0.9669,
       "step": 13167
     },
     {
-      "epoch": 0.3616489522397078,
+      "epoch": 0.37366628830874005,
       "grad_norm": 0.0,
-      "learning_rate": 1.4765319586089589e-05,
-      "loss": 1.0185,
+      "learning_rate": 1.4419692474503847e-05,
+      "loss": 1.0284,
       "step": 13168
     },
     {
-      "epoch": 0.361676416467551,
+      "epoch": 0.37369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.4764537542876501e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.4418868015052424e-05,
+      "loss": 0.9725,
       "step": 13169
     },
     {
-      "epoch": 0.3617038806953943,
+      "epoch": 0.3737230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.4763755461964547e-05,
-      "loss": 0.7294,
+      "learning_rate": 1.4418043518274621e-05,
+      "loss": 0.967,
       "step": 13170
     },
     {
-      "epoch": 0.3617313449232375,
+      "epoch": 0.37375141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.4762973343359915e-05,
-      "loss": 1.004,
+      "learning_rate": 1.4417218984177406e-05,
+      "loss": 0.9284,
       "step": 13171
     },
     {
-      "epoch": 0.3617588091510807,
+      "epoch": 0.37377979568671965,
       "grad_norm": 0.0,
-      "learning_rate": 1.4762191187068793e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.4416394412767747e-05,
+      "loss": 0.87,
       "step": 13172
     },
     {
-      "epoch": 0.36178627337892394,
+      "epoch": 0.3738081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.4761408993097368e-05,
-      "loss": 1.0457,
+      "learning_rate": 1.44155698040526e-05,
+      "loss": 0.8725,
       "step": 13173
     },
     {
-      "epoch": 0.36181373760676716,
+      "epoch": 0.3738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.4760626761451833e-05,
-      "loss": 0.8338,
+      "learning_rate": 1.441474515803894e-05,
+      "loss": 1.0347,
       "step": 13174
     },
     {
-      "epoch": 0.36184120183461044,
+      "epoch": 0.3738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.4759844492138373e-05,
-      "loss": 0.864,
+      "learning_rate": 1.441392047473373e-05,
+      "loss": 0.9733,
       "step": 13175
     },
     {
-      "epoch": 0.36186866606245366,
+      "epoch": 0.3738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.4759062185163183e-05,
-      "loss": 0.8999,
+      "learning_rate": 1.4413095754143934e-05,
+      "loss": 0.9392,
       "step": 13176
     },
     {
-      "epoch": 0.3618961302902969,
+      "epoch": 0.3739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.4758279840532445e-05,
-      "loss": 0.959,
+      "learning_rate": 1.4412270996276518e-05,
+      "loss": 0.9074,
       "step": 13177
     },
     {
-      "epoch": 0.3619235945181401,
+      "epoch": 0.373950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.4757497458252355e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.4411446201138453e-05,
+      "loss": 0.9321,
       "step": 13178
     },
     {
-      "epoch": 0.3619510587459834,
+      "epoch": 0.3739784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.4756715038329103e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.4410621368736703e-05,
+      "loss": 0.7933,
       "step": 13179
     },
     {
-      "epoch": 0.3619785229738266,
+      "epoch": 0.37400681044267875,
       "grad_norm": 0.0,
-      "learning_rate": 1.4755932580768879e-05,
-      "loss": 0.8805,
+      "learning_rate": 1.4409796499078238e-05,
+      "loss": 0.8649,
       "step": 13180
     },
     {
-      "epoch": 0.3620059872016698,
+      "epoch": 0.37403518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.4755150085577874e-05,
-      "loss": 0.8941,
+      "learning_rate": 1.440897159217002e-05,
+      "loss": 0.8488,
       "step": 13181
     },
     {
-      "epoch": 0.36203345142951304,
+      "epoch": 0.3740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.475436755276228e-05,
-      "loss": 1.0487,
+      "learning_rate": 1.4408146648019025e-05,
+      "loss": 0.9146,
       "step": 13182
     },
     {
-      "epoch": 0.3620609156573563,
+      "epoch": 0.37409194097616344,
       "grad_norm": 0.0,
-      "learning_rate": 1.4753584982328288e-05,
-      "loss": 0.9436,
+      "learning_rate": 1.4407321666632213e-05,
+      "loss": 0.8929,
       "step": 13183
     },
     {
-      "epoch": 0.36208837988519954,
+      "epoch": 0.37412031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.475280237428209e-05,
-      "loss": 1.0327,
+      "learning_rate": 1.4406496648016557e-05,
+      "loss": 0.953,
       "step": 13184
     },
     {
-      "epoch": 0.36211584411304276,
+      "epoch": 0.3741486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.475201972862988e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.4405671592179028e-05,
+      "loss": 0.9077,
       "step": 13185
     },
     {
-      "epoch": 0.362143308340886,
+      "epoch": 0.37417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.4751237045377847e-05,
-      "loss": 0.8877,
+      "learning_rate": 1.4404846499126594e-05,
+      "loss": 0.918,
       "step": 13186
     },
     {
-      "epoch": 0.3621707725687292,
+      "epoch": 0.37420544835414304,
       "grad_norm": 0.0,
-      "learning_rate": 1.4750454324532184e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.4404021368866222e-05,
+      "loss": 1.0576,
       "step": 13187
     },
     {
-      "epoch": 0.3621982367965725,
+      "epoch": 0.3742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.4749671566099091e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.4403196201404884e-05,
+      "loss": 0.9917,
       "step": 13188
     },
     {
-      "epoch": 0.3622257010244157,
+      "epoch": 0.3742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.4748888770084757e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.4402370996749549e-05,
+      "loss": 0.9596,
       "step": 13189
     },
     {
-      "epoch": 0.3622531652522589,
+      "epoch": 0.3742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.4748105936495372e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.4401545754907187e-05,
+      "loss": 0.865,
       "step": 13190
     },
     {
-      "epoch": 0.36228062948010215,
+      "epoch": 0.3743189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.4747323065337136e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.440072047588477e-05,
+      "loss": 1.0005,
       "step": 13191
     },
     {
-      "epoch": 0.3623080937079454,
+      "epoch": 0.37434733257661745,
       "grad_norm": 0.0,
-      "learning_rate": 1.4746540156616237e-05,
-      "loss": 0.8602,
+      "learning_rate": 1.4399895159689272e-05,
+      "loss": 1.0602,
       "step": 13192
     },
     {
-      "epoch": 0.36233555793578864,
+      "epoch": 0.37437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.4745757210338878e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.4399069806327662e-05,
+      "loss": 1.0127,
       "step": 13193
     },
     {
-      "epoch": 0.36236302216363186,
+      "epoch": 0.3744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.4744974226511248e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.4398244415806908e-05,
+      "loss": 0.9648,
       "step": 13194
     },
     {
-      "epoch": 0.3623904863914751,
+      "epoch": 0.37443246311010214,
       "grad_norm": 0.0,
-      "learning_rate": 1.4744191205139544e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.4397418988133989e-05,
+      "loss": 0.8314,
       "step": 13195
     },
     {
-      "epoch": 0.36241795061931836,
+      "epoch": 0.37446083995459706,
       "grad_norm": 0.0,
-      "learning_rate": 1.4743408146229963e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.4396593523315874e-05,
+      "loss": 0.8631,
       "step": 13196
     },
     {
-      "epoch": 0.3624454148471616,
+      "epoch": 0.3744892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 1.4742625049788698e-05,
-      "loss": 1.0418,
+      "learning_rate": 1.4395768021359533e-05,
+      "loss": 0.9077,
       "step": 13197
     },
     {
-      "epoch": 0.3624728790750048,
+      "epoch": 0.37451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.4741841915821944e-05,
-      "loss": 0.929,
+      "learning_rate": 1.4394942482271942e-05,
+      "loss": 0.8973,
       "step": 13198
     },
     {
-      "epoch": 0.362500343302848,
+      "epoch": 0.37454597048808175,
       "grad_norm": 0.0,
-      "learning_rate": 1.4741058744335903e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.4394116906060078e-05,
+      "loss": 0.8319,
       "step": 13199
     },
     {
-      "epoch": 0.36252780753069125,
+      "epoch": 0.3745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.4740275535336768e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.439329129273091e-05,
+      "loss": 0.8381,
       "step": 13200
     },
     {
-      "epoch": 0.3625552717585345,
+      "epoch": 0.3746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.4739492288830737e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.4392465642291412e-05,
+      "loss": 0.8561,
       "step": 13201
     },
     {
-      "epoch": 0.36258273598637775,
+      "epoch": 0.3746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.4738709004824005e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.439163995474856e-05,
+      "loss": 1.0219,
       "step": 13202
     },
     {
-      "epoch": 0.36261020021422097,
+      "epoch": 0.3746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.4737925683322776e-05,
-      "loss": 0.8841,
+      "learning_rate": 1.4390814230109327e-05,
+      "loss": 1.0033,
       "step": 13203
     },
     {
-      "epoch": 0.3626376644420642,
+      "epoch": 0.3746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.473714232433324e-05,
-      "loss": 1.0576,
+      "learning_rate": 1.4389988468380688e-05,
+      "loss": 0.9,
       "step": 13204
     },
     {
-      "epoch": 0.36266512866990747,
+      "epoch": 0.37471623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.47363589278616e-05,
-      "loss": 0.8837,
+      "learning_rate": 1.4389162669569621e-05,
+      "loss": 0.9251,
       "step": 13205
     },
     {
-      "epoch": 0.3626925928977507,
+      "epoch": 0.374744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4735575493914056e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.4388336833683099e-05,
+      "loss": 0.9346,
       "step": 13206
     },
     {
-      "epoch": 0.3627200571255939,
+      "epoch": 0.37477298524404085,
       "grad_norm": 0.0,
-      "learning_rate": 1.47347920224968e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.4387510960728098e-05,
+      "loss": 0.8459,
       "step": 13207
     },
     {
-      "epoch": 0.36274752135343713,
+      "epoch": 0.37480136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.4734008513616037e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.4386685050711593e-05,
+      "loss": 1.0206,
       "step": 13208
     },
     {
-      "epoch": 0.3627749855812804,
+      "epoch": 0.3748297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.4733224967277969e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.4385859103640565e-05,
+      "loss": 0.9932,
       "step": 13209
     },
     {
-      "epoch": 0.36280244980912363,
+      "epoch": 0.37485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.4732441383488785e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.4385033119521987e-05,
+      "loss": 0.9156,
       "step": 13210
     },
     {
-      "epoch": 0.36282991403696685,
+      "epoch": 0.37488649262202045,
       "grad_norm": 0.0,
-      "learning_rate": 1.4731657762254694e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.4384207098362837e-05,
+      "loss": 0.8867,
       "step": 13211
     },
     {
-      "epoch": 0.36285737826481007,
+      "epoch": 0.3749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.4730874103581894e-05,
-      "loss": 0.8608,
+      "learning_rate": 1.4383381040170095e-05,
+      "loss": 0.9455,
       "step": 13212
     },
     {
-      "epoch": 0.3628848424926533,
+      "epoch": 0.3749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.4730090407476588e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.4382554944950733e-05,
+      "loss": 0.8869,
       "step": 13213
     },
     {
-      "epoch": 0.36291230672049657,
+      "epoch": 0.3749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.4729306673944969e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.4381728812711733e-05,
+      "loss": 0.8669,
       "step": 13214
     },
     {
-      "epoch": 0.3629397709483398,
+      "epoch": 0.375,
       "grad_norm": 0.0,
-      "learning_rate": 1.472852290299325e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.4380902643460073e-05,
+      "loss": 1.1051,
       "step": 13215
     },
     {
-      "epoch": 0.362967235176183,
+      "epoch": 0.3750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.4727739094627622e-05,
-      "loss": 0.8652,
+      "learning_rate": 1.4380076437202733e-05,
+      "loss": 0.919,
       "step": 13216
     },
     {
-      "epoch": 0.36299469940402623,
+      "epoch": 0.3750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.4726955248854294e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.4379250193946685e-05,
+      "loss": 0.8999,
       "step": 13217
     },
     {
-      "epoch": 0.3630221636318695,
+      "epoch": 0.3750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.4726171365679462e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.437842391369892e-05,
+      "loss": 1.0208,
       "step": 13218
     },
     {
-      "epoch": 0.36304962785971273,
+      "epoch": 0.37511350737797955,
       "grad_norm": 0.0,
-      "learning_rate": 1.4725387445109331e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.437759759646641e-05,
+      "loss": 0.9829,
       "step": 13219
     },
     {
-      "epoch": 0.36307709208755595,
+      "epoch": 0.37514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.4724603487150106e-05,
-      "loss": 0.9044,
+      "learning_rate": 1.4376771242256134e-05,
+      "loss": 0.9566,
       "step": 13220
     },
     {
-      "epoch": 0.3631045563153992,
+      "epoch": 0.3751702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.472381949180799e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.4375944851075074e-05,
+      "loss": 0.9809,
       "step": 13221
     },
     {
-      "epoch": 0.36313202054324245,
+      "epoch": 0.37519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.4723035459089181e-05,
-      "loss": 0.9076,
+      "learning_rate": 1.4375118422930215e-05,
+      "loss": 0.9337,
       "step": 13222
     },
     {
-      "epoch": 0.36315948477108567,
+      "epoch": 0.37522701475595915,
       "grad_norm": 0.0,
-      "learning_rate": 1.4722251388999883e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.4374291957828534e-05,
+      "loss": 0.8539,
       "step": 13223
     },
     {
-      "epoch": 0.3631869489989289,
+      "epoch": 0.375255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.4721467281546309e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.4373465455777009e-05,
+      "loss": 0.89,
       "step": 13224
     },
     {
-      "epoch": 0.3632144132267721,
+      "epoch": 0.37528376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.4720683136734656e-05,
-      "loss": 0.951,
+      "learning_rate": 1.4372638916782627e-05,
+      "loss": 0.9598,
       "step": 13225
     },
     {
-      "epoch": 0.36324187745461534,
+      "epoch": 0.3753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.4719898954571128e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.4371812340852366e-05,
+      "loss": 0.9363,
       "step": 13226
     },
     {
-      "epoch": 0.3632693416824586,
+      "epoch": 0.3753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.4719114735061932e-05,
-      "loss": 0.974,
+      "learning_rate": 1.437098572799321e-05,
+      "loss": 1.0474,
       "step": 13227
     },
     {
-      "epoch": 0.36329680591030183,
+      "epoch": 0.3753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.4718330478213272e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.4370159078212142e-05,
+      "loss": 1.0129,
       "step": 13228
     },
     {
-      "epoch": 0.36332427013814506,
+      "epoch": 0.3753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.4717546184031352e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.4369332391516142e-05,
+      "loss": 1.0369,
       "step": 13229
     },
     {
-      "epoch": 0.3633517343659883,
+      "epoch": 0.3754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.4716761852522381e-05,
-      "loss": 0.9373,
+      "learning_rate": 1.4368505667912196e-05,
+      "loss": 0.8808,
       "step": 13230
     },
     {
-      "epoch": 0.36337919859383155,
+      "epoch": 0.37545402951191825,
       "grad_norm": 0.0,
-      "learning_rate": 1.471597748369256e-05,
-      "loss": 1.0141,
+      "learning_rate": 1.4367678907407285e-05,
+      "loss": 1.1089,
       "step": 13231
     },
     {
-      "epoch": 0.3634066628216748,
+      "epoch": 0.37548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.4715193077548103e-05,
-      "loss": 0.8502,
+      "learning_rate": 1.4366852110008397e-05,
+      "loss": 0.8794,
       "step": 13232
     },
     {
-      "epoch": 0.363434127049518,
+      "epoch": 0.3755107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 1.471440863409521e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.4366025275722512e-05,
+      "loss": 0.9621,
       "step": 13233
     },
     {
-      "epoch": 0.3634615912773612,
+      "epoch": 0.37553916004540294,
       "grad_norm": 0.0,
-      "learning_rate": 1.4713624153340087e-05,
-      "loss": 0.9574,
+      "learning_rate": 1.4365198404556612e-05,
+      "loss": 0.9234,
       "step": 13234
     },
     {
-      "epoch": 0.3634890555052045,
+      "epoch": 0.37556753688989786,
       "grad_norm": 0.0,
-      "learning_rate": 1.4712839635288947e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.4364371496517688e-05,
+      "loss": 0.9142,
       "step": 13235
     },
     {
-      "epoch": 0.3635165197330477,
+      "epoch": 0.3755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.4712055079947994e-05,
-      "loss": 0.961,
+      "learning_rate": 1.436354455161272e-05,
+      "loss": 0.9843,
       "step": 13236
     },
     {
-      "epoch": 0.36354398396089094,
+      "epoch": 0.37562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.4711270487323434e-05,
-      "loss": 0.8796,
+      "learning_rate": 1.4362717569848697e-05,
+      "loss": 1.0175,
       "step": 13237
     },
     {
-      "epoch": 0.36357144818873416,
+      "epoch": 0.37565266742338255,
       "grad_norm": 0.0,
-      "learning_rate": 1.471048585742148e-05,
-      "loss": 1.028,
+      "learning_rate": 1.43618905512326e-05,
+      "loss": 1.0336,
       "step": 13238
     },
     {
-      "epoch": 0.36359891241657744,
+      "epoch": 0.3756810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.4709701190248333e-05,
-      "loss": 1.0184,
+      "learning_rate": 1.436106349577142e-05,
+      "loss": 1.0142,
       "step": 13239
     },
     {
-      "epoch": 0.36362637664442066,
+      "epoch": 0.3757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.470891648581021e-05,
-      "loss": 0.8635,
+      "learning_rate": 1.436023640347214e-05,
+      "loss": 0.8299,
       "step": 13240
     },
     {
-      "epoch": 0.3636538408722639,
+      "epoch": 0.3757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.4708131744113315e-05,
-      "loss": 1.0062,
+      "learning_rate": 1.4359409274341747e-05,
+      "loss": 0.9292,
       "step": 13241
     },
     {
-      "epoch": 0.3636813051001071,
+      "epoch": 0.3757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.4707346965163855e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.4358582108387226e-05,
+      "loss": 0.9865,
       "step": 13242
     },
     {
-      "epoch": 0.3637087693279503,
+      "epoch": 0.37579455164585696,
       "grad_norm": 0.0,
-      "learning_rate": 1.4706562148968044e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.4357754905615571e-05,
+      "loss": 0.8818,
       "step": 13243
     },
     {
-      "epoch": 0.3637362335557936,
+      "epoch": 0.37582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.470577729553209e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.435692766603376e-05,
+      "loss": 0.9808,
       "step": 13244
     },
     {
-      "epoch": 0.3637636977836368,
+      "epoch": 0.3758513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.4704992404862203e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.4356100389648784e-05,
+      "loss": 0.9953,
       "step": 13245
     },
     {
-      "epoch": 0.36379116201148004,
+      "epoch": 0.37587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.4704207476964595e-05,
-      "loss": 0.8877,
+      "learning_rate": 1.4355273076467635e-05,
+      "loss": 0.9026,
       "step": 13246
     },
     {
-      "epoch": 0.36381862623932326,
+      "epoch": 0.37590805902383656,
       "grad_norm": 0.0,
-      "learning_rate": 1.4703422511845475e-05,
-      "loss": 1.0604,
+      "learning_rate": 1.4354445726497298e-05,
+      "loss": 0.9268,
       "step": 13247
     },
     {
-      "epoch": 0.36384609046716654,
+      "epoch": 0.3759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.4702637509511051e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.4353618339744761e-05,
+      "loss": 0.9482,
       "step": 13248
     },
     {
-      "epoch": 0.36387355469500976,
+      "epoch": 0.37596481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.4701852469967541e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.4352790916217016e-05,
+      "loss": 0.9689,
       "step": 13249
     },
     {
-      "epoch": 0.363901018922853,
+      "epoch": 0.37599318955732125,
       "grad_norm": 0.0,
-      "learning_rate": 1.4701067393221153e-05,
-      "loss": 0.8871,
+      "learning_rate": 1.4351963455921052e-05,
+      "loss": 0.9159,
       "step": 13250
     },
     {
-      "epoch": 0.3639284831506962,
+      "epoch": 0.3760215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.4700282279278097e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.4351135958863854e-05,
+      "loss": 0.8519,
       "step": 13251
     },
     {
-      "epoch": 0.3639559473785395,
+      "epoch": 0.376049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.4699497128144588e-05,
-      "loss": 0.993,
+      "learning_rate": 1.4350308425052413e-05,
+      "loss": 0.9026,
       "step": 13252
     },
     {
-      "epoch": 0.3639834116063827,
+      "epoch": 0.3760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.4698711939826837e-05,
-      "loss": 0.8511,
+      "learning_rate": 1.4349480854493724e-05,
+      "loss": 1.0298,
       "step": 13253
     },
     {
-      "epoch": 0.3640108758342259,
+      "epoch": 0.3761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.4697926714331057e-05,
-      "loss": 0.8826,
+      "learning_rate": 1.4348653247194777e-05,
+      "loss": 0.9972,
       "step": 13254
     },
     {
-      "epoch": 0.36403834006206914,
+      "epoch": 0.3761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.4697141451663459e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.4347825603162555e-05,
+      "loss": 1.0251,
       "step": 13255
     },
     {
-      "epoch": 0.36406580428991236,
+      "epoch": 0.3761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.469635615183026e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.434699792240406e-05,
+      "loss": 1.0754,
       "step": 13256
     },
     {
-      "epoch": 0.36409326851775564,
+      "epoch": 0.3761918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.4695570814837674e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.4346170204926275e-05,
+      "loss": 0.8495,
       "step": 13257
     },
     {
-      "epoch": 0.36412073274559886,
+      "epoch": 0.37622020431328035,
       "grad_norm": 0.0,
-      "learning_rate": 1.4694785440691907e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.4345342450736195e-05,
+      "loss": 0.9765,
       "step": 13258
     },
     {
-      "epoch": 0.3641481969734421,
+      "epoch": 0.37624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.4694000029399184e-05,
-      "loss": 0.8082,
+      "learning_rate": 1.4344514659840814e-05,
+      "loss": 0.9417,
       "step": 13259
     },
     {
-      "epoch": 0.3641756612012853,
+      "epoch": 0.3762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.4693214580965713e-05,
-      "loss": 1.077,
+      "learning_rate": 1.4343686832247121e-05,
+      "loss": 0.9666,
       "step": 13260
     },
     {
-      "epoch": 0.3642031254291286,
+      "epoch": 0.37630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.4692429095397712e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.4342858967962108e-05,
+      "loss": 0.9087,
       "step": 13261
     },
     {
-      "epoch": 0.3642305896569718,
+      "epoch": 0.37633371169125995,
       "grad_norm": 0.0,
-      "learning_rate": 1.4691643572701393e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.4342031066992775e-05,
+      "loss": 0.8956,
       "step": 13262
     },
     {
-      "epoch": 0.364258053884815,
+      "epoch": 0.3763620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.4690858012882972e-05,
-      "loss": 0.8801,
+      "learning_rate": 1.4341203129346108e-05,
+      "loss": 0.9552,
       "step": 13263
     },
     {
-      "epoch": 0.36428551811265825,
+      "epoch": 0.37639046538024973,
       "grad_norm": 0.0,
-      "learning_rate": 1.4690072415948666e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.4340375155029102e-05,
+      "loss": 0.8934,
       "step": 13264
     },
     {
-      "epoch": 0.3643129823405015,
+      "epoch": 0.3764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.4689286781904693e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.4339547144048753e-05,
+      "loss": 1.0401,
       "step": 13265
     },
     {
-      "epoch": 0.36434044656834474,
+      "epoch": 0.3764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.468850111075726e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.4338719096412056e-05,
+      "loss": 0.8837,
       "step": 13266
     },
     {
-      "epoch": 0.36436791079618797,
+      "epoch": 0.3764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.4687715402512595e-05,
-      "loss": 0.9385,
+      "learning_rate": 1.4337891012125997e-05,
+      "loss": 0.946,
       "step": 13267
     },
     {
-      "epoch": 0.3643953750240312,
+      "epoch": 0.3765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.4686929657176906e-05,
-      "loss": 1.0224,
+      "learning_rate": 1.4337062891197582e-05,
+      "loss": 0.9679,
       "step": 13268
     },
     {
-      "epoch": 0.3644228392518744,
+      "epoch": 0.3765323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.468614387475642e-05,
-      "loss": 0.9084,
+      "learning_rate": 1.4336234733633802e-05,
+      "loss": 0.9078,
       "step": 13269
     },
     {
-      "epoch": 0.3644503034797177,
+      "epoch": 0.37656072644721905,
       "grad_norm": 0.0,
-      "learning_rate": 1.4685358055257343e-05,
-      "loss": 0.8907,
+      "learning_rate": 1.4335406539441655e-05,
+      "loss": 0.9485,
       "step": 13270
     },
     {
-      "epoch": 0.3644777677075609,
+      "epoch": 0.37658910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 1.4684572198685903e-05,
-      "loss": 1.0548,
+      "learning_rate": 1.4334578308628131e-05,
+      "loss": 0.8313,
       "step": 13271
     },
     {
-      "epoch": 0.36450523193540413,
+      "epoch": 0.3766174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 1.468378630504831e-05,
-      "loss": 1.0566,
+      "learning_rate": 1.433375004120023e-05,
+      "loss": 1.0583,
       "step": 13272
     },
     {
-      "epoch": 0.36453269616324735,
+      "epoch": 0.37664585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.4683000374350785e-05,
-      "loss": 0.9856,
+      "learning_rate": 1.4332921737164946e-05,
+      "loss": 0.9456,
       "step": 13273
     },
     {
-      "epoch": 0.3645601603910906,
+      "epoch": 0.37667423382519866,
       "grad_norm": 0.0,
-      "learning_rate": 1.4682214406599548e-05,
-      "loss": 0.9299,
+      "learning_rate": 1.433209339652928e-05,
+      "loss": 0.9285,
       "step": 13274
     },
     {
-      "epoch": 0.36458762461893385,
+      "epoch": 0.3767026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.4681428401800817e-05,
-      "loss": 0.871,
+      "learning_rate": 1.4331265019300222e-05,
+      "loss": 1.0026,
       "step": 13275
     },
     {
-      "epoch": 0.36461508884677707,
+      "epoch": 0.37673098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.4680642359960811e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.4330436605484776e-05,
+      "loss": 0.8692,
       "step": 13276
     },
     {
-      "epoch": 0.3646425530746203,
+      "epoch": 0.3767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.4679856281085748e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.432960815508994e-05,
+      "loss": 1.0141,
       "step": 13277
     },
     {
-      "epoch": 0.36467001730246357,
+      "epoch": 0.3767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.467907016518185e-05,
-      "loss": 0.9373,
+      "learning_rate": 1.4328779668122706e-05,
+      "loss": 0.806,
       "step": 13278
     },
     {
-      "epoch": 0.3646974815303068,
+      "epoch": 0.3768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.4678284012255338e-05,
-      "loss": 0.8037,
+      "learning_rate": 1.4327951144590078e-05,
+      "loss": 0.9606,
       "step": 13279
     },
     {
-      "epoch": 0.36472494575815,
+      "epoch": 0.376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.467749782231243e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.4327122584499052e-05,
+      "loss": 0.9455,
       "step": 13280
     },
     {
-      "epoch": 0.36475240998599323,
+      "epoch": 0.3768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.4676711595359348e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.4326293987856623e-05,
+      "loss": 0.8682,
       "step": 13281
     },
     {
-      "epoch": 0.36477987421383645,
+      "epoch": 0.37690124858115776,
       "grad_norm": 0.0,
-      "learning_rate": 1.467592533140231e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.4325465354669796e-05,
+      "loss": 1.0683,
       "step": 13282
     },
     {
-      "epoch": 0.36480733844167973,
+      "epoch": 0.37692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.467513903044754e-05,
-      "loss": 0.9048,
+      "learning_rate": 1.432463668494557e-05,
+      "loss": 1.1434,
       "step": 13283
     },
     {
-      "epoch": 0.36483480266952295,
+      "epoch": 0.3769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.467435269250126e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.4323807978690945e-05,
+      "loss": 0.9585,
       "step": 13284
     },
     {
-      "epoch": 0.36486226689736617,
+      "epoch": 0.37698637911464244,
       "grad_norm": 0.0,
-      "learning_rate": 1.4673566317569692e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.4322979235912917e-05,
+      "loss": 0.9274,
       "step": 13285
     },
     {
-      "epoch": 0.3648897311252094,
+      "epoch": 0.37701475595913736,
       "grad_norm": 0.0,
-      "learning_rate": 1.4672779905659056e-05,
-      "loss": 0.8217,
+      "learning_rate": 1.4322150456618489e-05,
+      "loss": 0.909,
       "step": 13286
     },
     {
-      "epoch": 0.36491719535305267,
+      "epoch": 0.3770431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.4671993456775573e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.4321321640814665e-05,
+      "loss": 0.8645,
       "step": 13287
     },
     {
-      "epoch": 0.3649446595808959,
+      "epoch": 0.37707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.467120697092547e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.432049278850844e-05,
+      "loss": 0.9685,
       "step": 13288
     },
     {
-      "epoch": 0.3649721238087391,
+      "epoch": 0.377099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.467042044811497e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.4319663899706818e-05,
+      "loss": 1.0515,
       "step": 13289
     },
     {
-      "epoch": 0.36499958803658233,
+      "epoch": 0.3771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.4669633888350292e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.43188349744168e-05,
+      "loss": 0.9209,
       "step": 13290
     },
     {
-      "epoch": 0.3650270522644256,
+      "epoch": 0.3771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.466884729163766e-05,
-      "loss": 0.8666,
+      "learning_rate": 1.4318006012645391e-05,
+      "loss": 0.8953,
       "step": 13291
     },
     {
-      "epoch": 0.36505451649226883,
+      "epoch": 0.3771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.4668060657983303e-05,
-      "loss": 1.0615,
+      "learning_rate": 1.431717701439959e-05,
+      "loss": 1.0173,
       "step": 13292
     },
     {
-      "epoch": 0.36508198072011205,
+      "epoch": 0.3772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.466727398739344e-05,
-      "loss": 0.966,
+      "learning_rate": 1.4316347979686402e-05,
+      "loss": 0.929,
       "step": 13293
     },
     {
-      "epoch": 0.3651094449479553,
+      "epoch": 0.37724177071509646,
       "grad_norm": 0.0,
-      "learning_rate": 1.4666487279874297e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.431551890851283e-05,
+      "loss": 0.9422,
       "step": 13294
     },
     {
-      "epoch": 0.3651369091757985,
+      "epoch": 0.3772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.4665700535432102e-05,
-      "loss": 0.8554,
+      "learning_rate": 1.4314689800885871e-05,
+      "loss": 0.9084,
       "step": 13295
     },
     {
-      "epoch": 0.3651643734036418,
+      "epoch": 0.3772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.4664913754073075e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.4313860656812537e-05,
+      "loss": 0.8494,
       "step": 13296
     },
     {
-      "epoch": 0.365191837631485,
+      "epoch": 0.37732690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.4664126935803442e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.4313031476299828e-05,
+      "loss": 0.7873,
       "step": 13297
     },
     {
-      "epoch": 0.3652193018593282,
+      "epoch": 0.37735527809307606,
       "grad_norm": 0.0,
-      "learning_rate": 1.466334008062943e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.4312202259354746e-05,
+      "loss": 0.9917,
       "step": 13298
     },
     {
-      "epoch": 0.36524676608717144,
+      "epoch": 0.3773836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 1.4662553188557266e-05,
-      "loss": 1.0504,
+      "learning_rate": 1.4311373005984297e-05,
+      "loss": 0.9353,
       "step": 13299
     },
     {
-      "epoch": 0.3652742303150147,
+      "epoch": 0.37741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.4661766259593177e-05,
-      "loss": 1.082,
+      "learning_rate": 1.4310543716195489e-05,
+      "loss": 0.9027,
       "step": 13300
     },
     {
-      "epoch": 0.36530169454285794,
+      "epoch": 0.37744040862656075,
       "grad_norm": 0.0,
-      "learning_rate": 1.4660979293743388e-05,
-      "loss": 0.7763,
+      "learning_rate": 1.4309714389995324e-05,
+      "loss": 0.9318,
       "step": 13301
     },
     {
-      "epoch": 0.36532915877070116,
+      "epoch": 0.3774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.4660192291014119e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.4308885027390806e-05,
+      "loss": 0.8527,
       "step": 13302
     },
     {
-      "epoch": 0.3653566229985444,
+      "epoch": 0.3774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.465940525141161e-05,
-      "loss": 0.9829,
+      "learning_rate": 1.4308055628388946e-05,
+      "loss": 0.9124,
       "step": 13303
     },
     {
-      "epoch": 0.36538408722638765,
+      "epoch": 0.3775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.4658618174942078e-05,
-      "loss": 0.9522,
+      "learning_rate": 1.4307226192996745e-05,
+      "loss": 0.958,
       "step": 13304
     },
     {
-      "epoch": 0.3654115514542309,
+      "epoch": 0.3775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.4657831061611756e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.4306396721221209e-05,
+      "loss": 0.8787,
       "step": 13305
     },
     {
-      "epoch": 0.3654390156820741,
+      "epoch": 0.37758229284903516,
       "grad_norm": 0.0,
-      "learning_rate": 1.4657043911426872e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.4305567213069346e-05,
+      "loss": 1.0361,
       "step": 13306
     },
     {
-      "epoch": 0.3654664799099173,
+      "epoch": 0.3776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.4656256724393649e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.4304737668548165e-05,
+      "loss": 0.9249,
       "step": 13307
     },
     {
-      "epoch": 0.36549394413776054,
+      "epoch": 0.377639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.4655469500518322e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.4303908087664673e-05,
+      "loss": 0.9263,
       "step": 13308
     },
     {
-      "epoch": 0.3655214083656038,
+      "epoch": 0.37766742338251985,
       "grad_norm": 0.0,
-      "learning_rate": 1.4654682239807116e-05,
-      "loss": 0.8484,
+      "learning_rate": 1.4303078470425873e-05,
+      "loss": 0.8784,
       "step": 13309
     },
     {
-      "epoch": 0.36554887259344704,
+      "epoch": 0.37769580022701477,
       "grad_norm": 0.0,
-      "learning_rate": 1.465389494226626e-05,
-      "loss": 0.9486,
+      "learning_rate": 1.4302248816838777e-05,
+      "loss": 1.0169,
       "step": 13310
     },
     {
-      "epoch": 0.36557633682129026,
+      "epoch": 0.3777241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.4653107607901985e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.4301419126910391e-05,
+      "loss": 0.8672,
       "step": 13311
     },
     {
-      "epoch": 0.3656038010491335,
+      "epoch": 0.37775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.4652320236720518e-05,
-      "loss": 0.8554,
+      "learning_rate": 1.4300589400647726e-05,
+      "loss": 0.9935,
       "step": 13312
     },
     {
-      "epoch": 0.36563126527697676,
+      "epoch": 0.37778093076049946,
       "grad_norm": 0.0,
-      "learning_rate": 1.4651532828728097e-05,
-      "loss": 0.9483,
+      "learning_rate": 1.4299759638057786e-05,
+      "loss": 0.9082,
       "step": 13313
     },
     {
-      "epoch": 0.36565872950482,
+      "epoch": 0.3778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.4650745383930943e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.4298929839147586e-05,
+      "loss": 0.8893,
       "step": 13314
     },
     {
-      "epoch": 0.3656861937326632,
+      "epoch": 0.37783768444948923,
       "grad_norm": 0.0,
-      "learning_rate": 1.464995790233529e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.4298100003924129e-05,
+      "loss": 0.9745,
       "step": 13315
     },
     {
-      "epoch": 0.3657136579605064,
+      "epoch": 0.3778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.4649170383947367e-05,
-      "loss": 0.8293,
+      "learning_rate": 1.4297270132394432e-05,
+      "loss": 1.0305,
       "step": 13316
     },
     {
-      "epoch": 0.3657411221883497,
+      "epoch": 0.377894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.4648382828773408e-05,
-      "loss": 0.9618,
+      "learning_rate": 1.42964402245655e-05,
+      "loss": 0.8779,
       "step": 13317
     },
     {
-      "epoch": 0.3657685864161929,
+      "epoch": 0.3779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.4647595236819648e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.4295610280444345e-05,
+      "loss": 0.8843,
       "step": 13318
     },
     {
-      "epoch": 0.36579605064403614,
+      "epoch": 0.3779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.4646807608092308e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.4294780300037973e-05,
+      "loss": 0.9149,
       "step": 13319
     },
     {
-      "epoch": 0.36582351487187936,
+      "epoch": 0.3779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.4646019942597629e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.4293950283353404e-05,
+      "loss": 1.0524,
       "step": 13320
     },
     {
-      "epoch": 0.3658509790997226,
+      "epoch": 0.37800794551645855,
       "grad_norm": 0.0,
-      "learning_rate": 1.4645232240341842e-05,
-      "loss": 1.0718,
+      "learning_rate": 1.4293120230397639e-05,
+      "loss": 0.8845,
       "step": 13321
     },
     {
-      "epoch": 0.36587844332756586,
+      "epoch": 0.37803632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.4644444501331175e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.4292290141177697e-05,
+      "loss": 0.9035,
       "step": 13322
     },
     {
-      "epoch": 0.3659059075554091,
+      "epoch": 0.37806469920544833,
       "grad_norm": 0.0,
-      "learning_rate": 1.4643656725571865e-05,
-      "loss": 1.043,
+      "learning_rate": 1.4291460015700587e-05,
+      "loss": 0.8978,
       "step": 13323
     },
     {
-      "epoch": 0.3659333717832523,
+      "epoch": 0.37809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.4642868913070146e-05,
-      "loss": 1.0443,
+      "learning_rate": 1.4290629853973321e-05,
+      "loss": 0.879,
       "step": 13324
     },
     {
-      "epoch": 0.3659608360110955,
+      "epoch": 0.37812145289443816,
       "grad_norm": 0.0,
-      "learning_rate": 1.4642081063832247e-05,
-      "loss": 0.9039,
+      "learning_rate": 1.4289799656002912e-05,
+      "loss": 1.0386,
       "step": 13325
     },
     {
-      "epoch": 0.3659883002389388,
+      "epoch": 0.378149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.4641293177864406e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.4288969421796372e-05,
+      "loss": 0.9594,
       "step": 13326
     },
     {
-      "epoch": 0.366015764466782,
+      "epoch": 0.37817820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.4640505255172855e-05,
-      "loss": 0.9284,
+      "learning_rate": 1.4288139151360716e-05,
+      "loss": 0.9781,
       "step": 13327
     },
     {
-      "epoch": 0.36604322869462524,
+      "epoch": 0.3782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.4639717295763828e-05,
-      "loss": 0.962,
+      "learning_rate": 1.4287308844702954e-05,
+      "loss": 0.9058,
       "step": 13328
     },
     {
-      "epoch": 0.36607069292246847,
+      "epoch": 0.3782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.4638929299643562e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.4286478501830102e-05,
+      "loss": 0.9858,
       "step": 13329
     },
     {
-      "epoch": 0.36609815715031174,
+      "epoch": 0.3782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.463814126681829e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.4285648122749174e-05,
+      "loss": 1.0005,
       "step": 13330
     },
     {
-      "epoch": 0.36612562137815496,
+      "epoch": 0.3782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.4637353197294248e-05,
-      "loss": 0.9777,
+      "learning_rate": 1.4284817707467182e-05,
+      "loss": 0.8797,
       "step": 13331
     },
     {
-      "epoch": 0.3661530856059982,
+      "epoch": 0.3783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.463656509107767e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.4283987255991143e-05,
+      "loss": 0.9776,
       "step": 13332
     },
     {
-      "epoch": 0.3661805498338414,
+      "epoch": 0.37834846765039726,
       "grad_norm": 0.0,
-      "learning_rate": 1.4635776948174796e-05,
-      "loss": 0.913,
+      "learning_rate": 1.4283156768328073e-05,
+      "loss": 1.0056,
       "step": 13333
     },
     {
-      "epoch": 0.36620801406168463,
+      "epoch": 0.3783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.4634988768591855e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.4282326244484983e-05,
+      "loss": 0.8365,
       "step": 13334
     },
     {
-      "epoch": 0.3662354782895279,
+      "epoch": 0.3784052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.4634200552335089e-05,
-      "loss": 0.9121,
+      "learning_rate": 1.4281495684468892e-05,
+      "loss": 0.9914,
       "step": 13335
     },
     {
-      "epoch": 0.3662629425173711,
+      "epoch": 0.37843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.4633412299410735e-05,
-      "loss": 1.0243,
+      "learning_rate": 1.4280665088286811e-05,
+      "loss": 0.9272,
       "step": 13336
     },
     {
-      "epoch": 0.36629040674521435,
+      "epoch": 0.37846197502837686,
       "grad_norm": 0.0,
-      "learning_rate": 1.4632624009825027e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.4279834455945765e-05,
+      "loss": 0.8438,
       "step": 13337
     },
     {
-      "epoch": 0.36631787097305757,
+      "epoch": 0.3784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.4631835683584201e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.4279003787452763e-05,
+      "loss": 0.914,
       "step": 13338
     },
     {
-      "epoch": 0.36634533520090085,
+      "epoch": 0.37851872871736664,
       "grad_norm": 0.0,
-      "learning_rate": 1.46310473206945e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.427817308281482e-05,
+      "loss": 0.8439,
       "step": 13339
     },
     {
-      "epoch": 0.36637279942874407,
+      "epoch": 0.3785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.4630258921162157e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.4277342342038963e-05,
+      "loss": 0.7931,
       "step": 13340
     },
     {
-      "epoch": 0.3664002636565873,
+      "epoch": 0.3785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.462947048499341e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.4276511565132198e-05,
+      "loss": 0.9984,
       "step": 13341
     },
     {
-      "epoch": 0.3664277278844305,
+      "epoch": 0.3786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.4628682012194503e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.427568075210155e-05,
+      "loss": 0.9398,
       "step": 13342
     },
     {
-      "epoch": 0.3664551921122738,
+      "epoch": 0.3786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.4627893502771668e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.4274849902954034e-05,
+      "loss": 0.9229,
       "step": 13343
     },
     {
-      "epoch": 0.366482656340117,
+      "epoch": 0.3786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.4627104956731151e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.4274019017696668e-05,
+      "loss": 0.9857,
       "step": 13344
     },
     {
-      "epoch": 0.36651012056796023,
+      "epoch": 0.37868898978433596,
       "grad_norm": 0.0,
-      "learning_rate": 1.4626316374079185e-05,
-      "loss": 0.8734,
+      "learning_rate": 1.4273188096336472e-05,
+      "loss": 0.9686,
       "step": 13345
     },
     {
-      "epoch": 0.36653758479580345,
+      "epoch": 0.3787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.4625527754822008e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.4272357138880462e-05,
+      "loss": 1.0005,
       "step": 13346
     },
     {
-      "epoch": 0.3665650490236467,
+      "epoch": 0.3787457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.4624739098965868e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.4271526145335663e-05,
+      "loss": 0.8724,
       "step": 13347
     },
     {
-      "epoch": 0.36659251325148995,
+      "epoch": 0.37877412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 1.4623950406517001e-05,
-      "loss": 0.9721,
+      "learning_rate": 1.4270695115709088e-05,
+      "loss": 0.9824,
       "step": 13348
     },
     {
-      "epoch": 0.36661997747933317,
+      "epoch": 0.37880249716231557,
       "grad_norm": 0.0,
-      "learning_rate": 1.4623161677481645e-05,
-      "loss": 0.942,
+      "learning_rate": 1.426986405000776e-05,
+      "loss": 0.921,
       "step": 13349
     },
     {
-      "epoch": 0.3666474417071764,
+      "epoch": 0.3788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.4622372911866043e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.42690329482387e-05,
+      "loss": 0.889,
       "step": 13350
     },
     {
-      "epoch": 0.3666749059350196,
+      "epoch": 0.37885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.4621584109676439e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.4268201810408924e-05,
+      "loss": 0.9415,
       "step": 13351
     },
     {
-      "epoch": 0.3667023701628629,
+      "epoch": 0.3788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.4620795270919069e-05,
-      "loss": 0.9628,
+      "learning_rate": 1.4267370636525457e-05,
+      "loss": 0.9581,
       "step": 13352
     },
     {
-      "epoch": 0.3667298343907061,
+      "epoch": 0.3789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.4620006395600177e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.4266539426595316e-05,
+      "loss": 0.9439,
       "step": 13353
     },
     {
-      "epoch": 0.36675729861854933,
+      "epoch": 0.37894438138479003,
       "grad_norm": 0.0,
-      "learning_rate": 1.4619217483726005e-05,
-      "loss": 0.9005,
+      "learning_rate": 1.4265708180625529e-05,
+      "loss": 0.944,
       "step": 13354
     },
     {
-      "epoch": 0.36678476284639255,
+      "epoch": 0.3789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.4618428535302793e-05,
-      "loss": 0.8707,
+      "learning_rate": 1.426487689862311e-05,
+      "loss": 1.0625,
       "step": 13355
     },
     {
-      "epoch": 0.36681222707423583,
+      "epoch": 0.3790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.4617639550336788e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.4264045580595082e-05,
+      "loss": 0.8943,
       "step": 13356
     },
     {
-      "epoch": 0.36683969130207905,
+      "epoch": 0.37902951191827466,
       "grad_norm": 0.0,
-      "learning_rate": 1.4616850528834232e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.4263214226548475e-05,
+      "loss": 0.9289,
       "step": 13357
     },
     {
-      "epoch": 0.3668671555299223,
+      "epoch": 0.3790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.4616061470801362e-05,
-      "loss": 0.8853,
+      "learning_rate": 1.4262382836490302e-05,
+      "loss": 0.8783,
       "step": 13358
     },
     {
-      "epoch": 0.3668946197577655,
+      "epoch": 0.3790862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4615272376244428e-05,
-      "loss": 0.856,
+      "learning_rate": 1.4261551410427592e-05,
+      "loss": 0.8367,
       "step": 13359
     },
     {
-      "epoch": 0.36692208398560877,
+      "epoch": 0.37911464245175935,
       "grad_norm": 0.0,
-      "learning_rate": 1.461448324516967e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.4260719948367364e-05,
+      "loss": 0.9661,
       "step": 13360
     },
     {
-      "epoch": 0.366949548213452,
+      "epoch": 0.37914301929625427,
       "grad_norm": 0.0,
-      "learning_rate": 1.4613694077583333e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.4259888450316641e-05,
+      "loss": 0.8583,
       "step": 13361
     },
     {
-      "epoch": 0.3669770124412952,
+      "epoch": 0.37917139614074913,
       "grad_norm": 0.0,
-      "learning_rate": 1.4612904873491664e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.4259056916282455e-05,
+      "loss": 0.8958,
       "step": 13362
     },
     {
-      "epoch": 0.36700447666913844,
+      "epoch": 0.37919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.46121156329009e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.4258225346271817e-05,
+      "loss": 1.1178,
       "step": 13363
     },
     {
-      "epoch": 0.36703194089698166,
+      "epoch": 0.37922814982973896,
       "grad_norm": 0.0,
-      "learning_rate": 1.4611326355817292e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.4257393740291762e-05,
+      "loss": 0.9767,
       "step": 13364
     },
     {
-      "epoch": 0.36705940512482493,
+      "epoch": 0.3792565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.4610537042247088e-05,
-      "loss": 0.891,
+      "learning_rate": 1.4256562098349312e-05,
+      "loss": 0.8781,
       "step": 13365
     },
     {
-      "epoch": 0.36708686935266815,
+      "epoch": 0.37928490351872873,
       "grad_norm": 0.0,
-      "learning_rate": 1.4609747692196526e-05,
-      "loss": 0.8724,
+      "learning_rate": 1.4255730420451485e-05,
+      "loss": 1.03,
       "step": 13366
     },
     {
-      "epoch": 0.3671143335805114,
+      "epoch": 0.3793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.4608958305671851e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.4254898706605316e-05,
+      "loss": 1.0193,
       "step": 13367
     },
     {
-      "epoch": 0.3671417978083546,
+      "epoch": 0.3793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.4608168882679317e-05,
-      "loss": 0.8465,
+      "learning_rate": 1.4254066956817824e-05,
+      "loss": 0.8828,
       "step": 13368
     },
     {
-      "epoch": 0.3671692620361979,
+      "epoch": 0.37937003405221337,
       "grad_norm": 0.0,
-      "learning_rate": 1.4607379423225165e-05,
-      "loss": 1.0149,
+      "learning_rate": 1.4253235171096037e-05,
+      "loss": 0.8827,
       "step": 13369
     },
     {
-      "epoch": 0.3671967262640411,
+      "epoch": 0.3793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.4606589927315641e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.4252403349446986e-05,
+      "loss": 0.8807,
       "step": 13370
     },
     {
-      "epoch": 0.3672241904918843,
+      "epoch": 0.3794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.4605800394956994e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.4251571491877689e-05,
+      "loss": 0.8506,
       "step": 13371
     },
     {
-      "epoch": 0.36725165471972754,
+      "epoch": 0.37945516458569806,
       "grad_norm": 0.0,
-      "learning_rate": 1.460501082615547e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.425073959839518e-05,
+      "loss": 0.9089,
       "step": 13372
     },
     {
-      "epoch": 0.3672791189475708,
+      "epoch": 0.37948354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.4604221220917318e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.4249907669006478e-05,
+      "loss": 0.867,
       "step": 13373
     },
     {
-      "epoch": 0.36730658317541404,
+      "epoch": 0.37951191827468783,
       "grad_norm": 0.0,
-      "learning_rate": 1.4603431579248783e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.4249075703718615e-05,
+      "loss": 0.9512,
       "step": 13374
     },
     {
-      "epoch": 0.36733404740325726,
+      "epoch": 0.37954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.4602641901156113e-05,
-      "loss": 1.0036,
+      "learning_rate": 1.424824370253862e-05,
+      "loss": 0.9548,
       "step": 13375
     },
     {
-      "epoch": 0.3673615116311005,
+      "epoch": 0.37956867196367766,
       "grad_norm": 0.0,
-      "learning_rate": 1.460185218664556e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.424741166547352e-05,
+      "loss": 0.8052,
       "step": 13376
     },
     {
-      "epoch": 0.3673889758589437,
+      "epoch": 0.3795970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.4601062435723368e-05,
-      "loss": 0.8921,
+      "learning_rate": 1.4246579592530342e-05,
+      "loss": 0.8968,
       "step": 13377
     },
     {
-      "epoch": 0.367416440086787,
+      "epoch": 0.37962542565266744,
       "grad_norm": 0.0,
-      "learning_rate": 1.4600272648395788e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.4245747483716117e-05,
+      "loss": 0.8806,
       "step": 13378
     },
     {
-      "epoch": 0.3674439043146302,
+      "epoch": 0.3796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.4599482824669067e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.424491533903787e-05,
+      "loss": 0.9982,
       "step": 13379
     },
     {
-      "epoch": 0.3674713685424734,
+      "epoch": 0.3796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.459869296454946e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.4244083158502633e-05,
+      "loss": 0.9301,
       "step": 13380
     },
     {
-      "epoch": 0.36749883277031664,
+      "epoch": 0.3797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.4597903068043212e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.4243250942117437e-05,
+      "loss": 0.8617,
       "step": 13381
     },
     {
-      "epoch": 0.3675262969981599,
+      "epoch": 0.379738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.459711313515657e-05,
-      "loss": 1.0426,
+      "learning_rate": 1.4242418689889306e-05,
+      "loss": 0.94,
       "step": 13382
     },
     {
-      "epoch": 0.36755376122600314,
+      "epoch": 0.3797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.4596323165895791e-05,
-      "loss": 1.0216,
+      "learning_rate": 1.4241586401825274e-05,
+      "loss": 0.9682,
       "step": 13383
     },
     {
-      "epoch": 0.36758122545384636,
+      "epoch": 0.37979568671963676,
       "grad_norm": 0.0,
-      "learning_rate": 1.4595533160267124e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.4240754077932373e-05,
+      "loss": 0.9393,
       "step": 13384
     },
     {
-      "epoch": 0.3676086896816896,
+      "epoch": 0.3798240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.4594743118276818e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.4239921718217632e-05,
+      "loss": 0.934,
       "step": 13385
     },
     {
-      "epoch": 0.36763615390953286,
+      "epoch": 0.37985244040862653,
       "grad_norm": 0.0,
-      "learning_rate": 1.4593953039931123e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.4239089322688078e-05,
+      "loss": 0.8239,
       "step": 13386
     },
     {
-      "epoch": 0.3676636181373761,
+      "epoch": 0.37988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.4593162925236293e-05,
-      "loss": 1.0809,
+      "learning_rate": 1.4238256891350748e-05,
+      "loss": 0.8707,
       "step": 13387
     },
     {
-      "epoch": 0.3676910823652193,
+      "epoch": 0.37990919409761637,
       "grad_norm": 0.0,
-      "learning_rate": 1.4592372774198578e-05,
-      "loss": 1.0386,
+      "learning_rate": 1.4237424424212673e-05,
+      "loss": 0.9882,
       "step": 13388
     },
     {
-      "epoch": 0.3677185465930625,
+      "epoch": 0.3799375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.4591582586824231e-05,
-      "loss": 0.8477,
+      "learning_rate": 1.4236591921280883e-05,
+      "loss": 0.8975,
       "step": 13389
     },
     {
-      "epoch": 0.36774601082090574,
+      "epoch": 0.37996594778660614,
       "grad_norm": 0.0,
-      "learning_rate": 1.4590792363119503e-05,
-      "loss": 0.8839,
+      "learning_rate": 1.4235759382562407e-05,
+      "loss": 0.944,
       "step": 13390
     },
     {
-      "epoch": 0.367773475048749,
+      "epoch": 0.379994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.4590002103090648e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.4234926808064284e-05,
+      "loss": 0.9919,
       "step": 13391
     },
     {
-      "epoch": 0.36780093927659224,
+      "epoch": 0.3800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.4589211806743921e-05,
-      "loss": 1.0415,
+      "learning_rate": 1.4234094197793543e-05,
+      "loss": 0.9795,
       "step": 13392
     },
     {
-      "epoch": 0.36782840350443546,
+      "epoch": 0.38005107832009083,
       "grad_norm": 0.0,
-      "learning_rate": 1.458842147408557e-05,
-      "loss": 0.907,
+      "learning_rate": 1.4233261551757221e-05,
+      "loss": 0.8756,
       "step": 13393
     },
     {
-      "epoch": 0.3678558677322787,
+      "epoch": 0.3800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.4587631105121851e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.4232428869962345e-05,
+      "loss": 0.9322,
       "step": 13394
     },
     {
-      "epoch": 0.36788333196012196,
+      "epoch": 0.3801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.458684069985902e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.4231596152415954e-05,
+      "loss": 0.9572,
       "step": 13395
     },
     {
-      "epoch": 0.3679107961879652,
+      "epoch": 0.38013620885357546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4586050258303323e-05,
-      "loss": 0.8968,
+      "learning_rate": 1.4230763399125079e-05,
+      "loss": 0.9144,
       "step": 13396
     },
     {
-      "epoch": 0.3679382604158084,
+      "epoch": 0.3801645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.4585259780461025e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.4229930610096754e-05,
+      "loss": 0.9801,
       "step": 13397
     },
     {
-      "epoch": 0.3679657246436516,
+      "epoch": 0.3801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.4584469266338374e-05,
-      "loss": 0.829,
+      "learning_rate": 1.4229097785338018e-05,
+      "loss": 0.9358,
       "step": 13398
     },
     {
-      "epoch": 0.3679931888714949,
+      "epoch": 0.38022133938706015,
       "grad_norm": 0.0,
-      "learning_rate": 1.4583678715941628e-05,
-      "loss": 1.0419,
+      "learning_rate": 1.42282649248559e-05,
+      "loss": 0.847,
       "step": 13399
     },
     {
-      "epoch": 0.3680206530993381,
+      "epoch": 0.38024971623155507,
       "grad_norm": 0.0,
-      "learning_rate": 1.4582888129277038e-05,
-      "loss": 0.9165,
+      "learning_rate": 1.422743202865744e-05,
+      "loss": 1.119,
       "step": 13400
     },
     {
-      "epoch": 0.36804811732718135,
+      "epoch": 0.38027809307604993,
       "grad_norm": 0.0,
-      "learning_rate": 1.4582097506350864e-05,
-      "loss": 0.951,
+      "learning_rate": 1.4226599096749673e-05,
+      "loss": 0.9343,
       "step": 13401
     },
     {
-      "epoch": 0.36807558155502457,
+      "epoch": 0.38030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.4581306847169357e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.4225766129139631e-05,
+      "loss": 1.0069,
       "step": 13402
     },
     {
-      "epoch": 0.3681030457828678,
+      "epoch": 0.3803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.4580516151738776e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.4224933125834353e-05,
+      "loss": 0.8766,
       "step": 13403
     },
     {
-      "epoch": 0.36813051001071107,
+      "epoch": 0.3803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.4579725420065379e-05,
-      "loss": 0.8642,
+      "learning_rate": 1.4224100086840875e-05,
+      "loss": 0.8568,
       "step": 13404
     },
     {
-      "epoch": 0.3681579742385543,
+      "epoch": 0.38039160045402953,
       "grad_norm": 0.0,
-      "learning_rate": 1.4578934652155418e-05,
-      "loss": 1.0525,
+      "learning_rate": 1.4223267012166234e-05,
+      "loss": 0.925,
       "step": 13405
     },
     {
-      "epoch": 0.3681854384663975,
+      "epoch": 0.3804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.4578143848015154e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.4222433901817468e-05,
+      "loss": 0.9336,
       "step": 13406
     },
     {
-      "epoch": 0.36821290269424073,
+      "epoch": 0.3804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.4577353007650842e-05,
-      "loss": 0.952,
+      "learning_rate": 1.422160075580161e-05,
+      "loss": 0.9221,
       "step": 13407
     },
     {
-      "epoch": 0.368240366922084,
+      "epoch": 0.38047673098751417,
       "grad_norm": 0.0,
-      "learning_rate": 1.457656213106874e-05,
-      "loss": 0.9042,
+      "learning_rate": 1.4220767574125702e-05,
+      "loss": 0.9584,
       "step": 13408
     },
     {
-      "epoch": 0.3682678311499272,
+      "epoch": 0.3805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.4575771218275106e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.4219934356796783e-05,
+      "loss": 0.7552,
       "step": 13409
     },
     {
-      "epoch": 0.36829529537777045,
+      "epoch": 0.380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.4574980269276198e-05,
-      "loss": 1.0194,
+      "learning_rate": 1.4219101103821884e-05,
+      "loss": 0.8963,
       "step": 13410
     },
     {
-      "epoch": 0.36832275960561367,
+      "epoch": 0.38056186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.457418928407827e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.4218267815208053e-05,
+      "loss": 0.9085,
       "step": 13411
     },
     {
-      "epoch": 0.36835022383345695,
+      "epoch": 0.38059023836549377,
       "grad_norm": 0.0,
-      "learning_rate": 1.457339826268759e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.4217434490962321e-05,
+      "loss": 1.0292,
       "step": 13412
     },
     {
-      "epoch": 0.36837768806130017,
+      "epoch": 0.38061861520998863,
       "grad_norm": 0.0,
-      "learning_rate": 1.457260720511041e-05,
-      "loss": 1.0956,
+      "learning_rate": 1.421660113109173e-05,
+      "loss": 1.002,
       "step": 13413
     },
     {
-      "epoch": 0.3684051522891434,
+      "epoch": 0.38064699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 1.4571816111352988e-05,
-      "loss": 0.8942,
+      "learning_rate": 1.421576773560332e-05,
+      "loss": 0.8838,
       "step": 13414
     },
     {
-      "epoch": 0.3684326165169866,
+      "epoch": 0.38067536889897846,
       "grad_norm": 0.0,
-      "learning_rate": 1.4571024981421586e-05,
-      "loss": 1.0018,
+      "learning_rate": 1.4214934304504133e-05,
+      "loss": 0.9165,
       "step": 13415
     },
     {
-      "epoch": 0.36846008074482983,
+      "epoch": 0.3807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.4570233815322465e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.4214100837801206e-05,
+      "loss": 0.8367,
       "step": 13416
     },
     {
-      "epoch": 0.3684875449726731,
+      "epoch": 0.38073212258796824,
       "grad_norm": 0.0,
-      "learning_rate": 1.4569442613061886e-05,
-      "loss": 0.917,
+      "learning_rate": 1.4213267335501578e-05,
+      "loss": 0.9372,
       "step": 13417
     },
     {
-      "epoch": 0.36851500920051633,
+      "epoch": 0.3807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.4568651374646104e-05,
-      "loss": 0.909,
+      "learning_rate": 1.4212433797612293e-05,
+      "loss": 0.9322,
       "step": 13418
     },
     {
-      "epoch": 0.36854247342835955,
+      "epoch": 0.380788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.4567860100081383e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.421160022414039e-05,
+      "loss": 0.8928,
       "step": 13419
     },
     {
-      "epoch": 0.3685699376562028,
+      "epoch": 0.38081725312145287,
       "grad_norm": 0.0,
-      "learning_rate": 1.4567068789373984e-05,
-      "loss": 0.8527,
+      "learning_rate": 1.4210766615092905e-05,
+      "loss": 0.9648,
       "step": 13420
     },
     {
-      "epoch": 0.36859740188404605,
+      "epoch": 0.3808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.456627744253017e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.4209932970476891e-05,
+      "loss": 0.8883,
       "step": 13421
     },
     {
-      "epoch": 0.36862486611188927,
+      "epoch": 0.3808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.4565486059556197e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.4209099290299384e-05,
+      "loss": 0.8723,
       "step": 13422
     },
     {
-      "epoch": 0.3686523303397325,
+      "epoch": 0.38090238365493756,
       "grad_norm": 0.0,
-      "learning_rate": 1.4564694640458328e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.4208265574567426e-05,
+      "loss": 1.0291,
       "step": 13423
     },
     {
-      "epoch": 0.3686797945675757,
+      "epoch": 0.3809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.4563903185242832e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.4207431823288058e-05,
+      "loss": 1.0076,
       "step": 13424
     },
     {
-      "epoch": 0.368707258795419,
+      "epoch": 0.38095913734392733,
       "grad_norm": 0.0,
-      "learning_rate": 1.4563111693915963e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.4206598036468326e-05,
+      "loss": 0.9305,
       "step": 13425
     },
     {
-      "epoch": 0.3687347230232622,
+      "epoch": 0.38098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.4562320166483984e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.4205764214115272e-05,
+      "loss": 1.0878,
       "step": 13426
     },
     {
-      "epoch": 0.36876218725110543,
+      "epoch": 0.38101589103291716,
       "grad_norm": 0.0,
-      "learning_rate": 1.4561528602953167e-05,
-      "loss": 1.0052,
+      "learning_rate": 1.4204930356235936e-05,
+      "loss": 0.9339,
       "step": 13427
     },
     {
-      "epoch": 0.36878965147894865,
+      "epoch": 0.381044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.4560737003329763e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.4204096462837362e-05,
+      "loss": 0.8067,
       "step": 13428
     },
     {
-      "epoch": 0.3688171157067919,
+      "epoch": 0.38107264472190694,
       "grad_norm": 0.0,
-      "learning_rate": 1.4559945367620045e-05,
-      "loss": 0.906,
+      "learning_rate": 1.4203262533926601e-05,
+      "loss": 0.9899,
       "step": 13429
     },
     {
-      "epoch": 0.36884457993463515,
+      "epoch": 0.3811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.4559153695830272e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.420242856951069e-05,
+      "loss": 0.956,
       "step": 13430
     },
     {
-      "epoch": 0.3688720441624784,
+      "epoch": 0.3811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.4558361987966707e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.4201594569596675e-05,
+      "loss": 1.0309,
       "step": 13431
     },
     {
-      "epoch": 0.3688995083903216,
+      "epoch": 0.3811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.4557570244035617e-05,
-      "loss": 0.8699,
+      "learning_rate": 1.4200760534191606e-05,
+      "loss": 0.9361,
       "step": 13432
     },
     {
-      "epoch": 0.3689269726181648,
+      "epoch": 0.3811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.4556778464043269e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.419992646330252e-05,
+      "loss": 0.9115,
       "step": 13433
     },
     {
-      "epoch": 0.3689544368460081,
+      "epoch": 0.3812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.4555986647995921e-05,
-      "loss": 0.8975,
+      "learning_rate": 1.4199092356936468e-05,
+      "loss": 0.9774,
       "step": 13434
     },
     {
-      "epoch": 0.3689819010738513,
+      "epoch": 0.38124290578887626,
       "grad_norm": 0.0,
-      "learning_rate": 1.4555194795899841e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.4198258215100496e-05,
+      "loss": 0.8784,
       "step": 13435
     },
     {
-      "epoch": 0.36900936530169454,
+      "epoch": 0.3812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.4554402907761298e-05,
-      "loss": 0.923,
+      "learning_rate": 1.4197424037801643e-05,
+      "loss": 0.958,
       "step": 13436
     },
     {
-      "epoch": 0.36903682952953776,
+      "epoch": 0.38129965947786604,
       "grad_norm": 0.0,
-      "learning_rate": 1.4553610983586556e-05,
-      "loss": 0.9038,
+      "learning_rate": 1.4196589825046962e-05,
+      "loss": 0.8943,
       "step": 13437
     },
     {
-      "epoch": 0.36906429375738103,
+      "epoch": 0.38132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.4552819023381876e-05,
-      "loss": 0.93,
+      "learning_rate": 1.4195755576843502e-05,
+      "loss": 0.9747,
       "step": 13438
     },
     {
-      "epoch": 0.36909175798522426,
+      "epoch": 0.38135641316685587,
       "grad_norm": 0.0,
-      "learning_rate": 1.455202702715353e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.4194921293198304e-05,
+      "loss": 0.8332,
       "step": 13439
     },
     {
-      "epoch": 0.3691192222130675,
+      "epoch": 0.3813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.4551234994907783e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.4194086974118415e-05,
+      "loss": 0.9014,
       "step": 13440
     },
     {
-      "epoch": 0.3691466864409107,
+      "epoch": 0.38141316685584564,
       "grad_norm": 0.0,
-      "learning_rate": 1.4550442926650902e-05,
-      "loss": 0.9907,
+      "learning_rate": 1.4193252619610886e-05,
+      "loss": 0.8985,
       "step": 13441
     },
     {
-      "epoch": 0.369174150668754,
+      "epoch": 0.3814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.4549650822389156e-05,
-      "loss": 0.8892,
+      "learning_rate": 1.4192418229682768e-05,
+      "loss": 0.8602,
       "step": 13442
     },
     {
-      "epoch": 0.3692016148965972,
+      "epoch": 0.3814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.4548858682128808e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.4191583804341098e-05,
+      "loss": 0.8675,
       "step": 13443
     },
     {
-      "epoch": 0.3692290791244404,
+      "epoch": 0.38149829738933033,
       "grad_norm": 0.0,
-      "learning_rate": 1.4548066505876126e-05,
-      "loss": 0.8539,
+      "learning_rate": 1.4190749343592934e-05,
+      "loss": 0.8897,
       "step": 13444
     },
     {
-      "epoch": 0.36925654335228364,
+      "epoch": 0.3815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.4547274293637386e-05,
-      "loss": 0.938,
+      "learning_rate": 1.4189914847445323e-05,
+      "loss": 0.9032,
       "step": 13445
     },
     {
-      "epoch": 0.36928400758012686,
+      "epoch": 0.3815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.4546482045418846e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.4189080315905313e-05,
+      "loss": 0.9719,
       "step": 13446
     },
     {
-      "epoch": 0.36931147180797014,
+      "epoch": 0.38158342792281497,
       "grad_norm": 0.0,
-      "learning_rate": 1.4545689761226779e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.418824574897995e-05,
+      "loss": 0.9576,
       "step": 13447
     },
     {
-      "epoch": 0.36933893603581336,
+      "epoch": 0.3816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.4544897441067454e-05,
-      "loss": 1.0147,
+      "learning_rate": 1.4187411146676292e-05,
+      "loss": 0.9469,
       "step": 13448
     },
     {
-      "epoch": 0.3693664002636566,
+      "epoch": 0.38164018161180474,
       "grad_norm": 0.0,
-      "learning_rate": 1.4544105084947139e-05,
-      "loss": 0.8229,
+      "learning_rate": 1.418657650900138e-05,
+      "loss": 0.9831,
       "step": 13449
     },
     {
-      "epoch": 0.3693938644914998,
+      "epoch": 0.38166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.4543312692872107e-05,
-      "loss": 0.967,
+      "learning_rate": 1.4185741835962268e-05,
+      "loss": 0.9666,
       "step": 13450
     },
     {
-      "epoch": 0.3694213287193431,
+      "epoch": 0.38169693530079457,
       "grad_norm": 0.0,
-      "learning_rate": 1.4542520264848624e-05,
-      "loss": 1.0072,
+      "learning_rate": 1.4184907127566006e-05,
+      "loss": 0.8305,
       "step": 13451
     },
     {
-      "epoch": 0.3694487929471863,
+      "epoch": 0.38172531214528943,
       "grad_norm": 0.0,
-      "learning_rate": 1.4541727800882959e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.4184072383819646e-05,
+      "loss": 0.9081,
       "step": 13452
     },
     {
-      "epoch": 0.3694762571750295,
+      "epoch": 0.38175368898978435,
       "grad_norm": 0.0,
-      "learning_rate": 1.4540935300981386e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.4183237604730243e-05,
+      "loss": 0.9283,
       "step": 13453
     },
     {
-      "epoch": 0.36950372140287274,
+      "epoch": 0.3817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.4540142765150175e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.4182402790304839e-05,
+      "loss": 0.8854,
       "step": 13454
     },
     {
-      "epoch": 0.369531185630716,
+      "epoch": 0.3818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.4539350193395594e-05,
-      "loss": 0.8529,
+      "learning_rate": 1.4181567940550492e-05,
+      "loss": 0.8588,
       "step": 13455
     },
     {
-      "epoch": 0.36955864985855924,
+      "epoch": 0.38183881952326904,
       "grad_norm": 0.0,
-      "learning_rate": 1.4538557585723917e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.4180733055474254e-05,
+      "loss": 0.9771,
       "step": 13456
     },
     {
-      "epoch": 0.36958611408640246,
+      "epoch": 0.3818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.4537764942141416e-05,
-      "loss": 0.8673,
+      "learning_rate": 1.4179898135083172e-05,
+      "loss": 0.9908,
       "step": 13457
     },
     {
-      "epoch": 0.3696135783142457,
+      "epoch": 0.3818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.4536972262654358e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.4179063179384307e-05,
+      "loss": 0.8852,
       "step": 13458
     },
     {
-      "epoch": 0.3696410425420889,
+      "epoch": 0.38192395005675367,
       "grad_norm": 0.0,
-      "learning_rate": 1.4536179547269018e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.4178228188384704e-05,
+      "loss": 0.9616,
       "step": 13459
     },
     {
-      "epoch": 0.3696685067699322,
+      "epoch": 0.3819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.4535386795991668e-05,
-      "loss": 0.8691,
+      "learning_rate": 1.4177393162091421e-05,
+      "loss": 0.9999,
       "step": 13460
     },
     {
-      "epoch": 0.3696959709977754,
+      "epoch": 0.3819807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.4534594008828584e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.417655810051151e-05,
+      "loss": 0.8669,
       "step": 13461
     },
     {
-      "epoch": 0.3697234352256186,
+      "epoch": 0.38200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.4533801185786032e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.4175723003652027e-05,
+      "loss": 0.9454,
       "step": 13462
     },
     {
-      "epoch": 0.36975089945346185,
+      "epoch": 0.3820374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.4533008326870292e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.4174887871520022e-05,
+      "loss": 0.9726,
       "step": 13463
     },
     {
-      "epoch": 0.3697783636813051,
+      "epoch": 0.38206583427922813,
       "grad_norm": 0.0,
-      "learning_rate": 1.4532215432087631e-05,
-      "loss": 0.911,
+      "learning_rate": 1.4174052704122552e-05,
+      "loss": 0.9509,
       "step": 13464
     },
     {
-      "epoch": 0.36980582790914834,
+      "epoch": 0.38209421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 1.453142250144433e-05,
-      "loss": 1.0588,
+      "learning_rate": 1.417321750146667e-05,
+      "loss": 0.9925,
       "step": 13465
     },
     {
-      "epoch": 0.36983329213699156,
+      "epoch": 0.3821225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.4530629534946657e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.4172382263559432e-05,
+      "loss": 0.9464,
       "step": 13466
     },
     {
-      "epoch": 0.3698607563648348,
+      "epoch": 0.3821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.4529836532600884e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.4171546990407896e-05,
+      "loss": 0.8789,
       "step": 13467
     },
     {
-      "epoch": 0.36988822059267806,
+      "epoch": 0.38217934165720774,
       "grad_norm": 0.0,
-      "learning_rate": 1.4529043494413296e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.4170711682019111e-05,
+      "loss": 0.8909,
       "step": 13468
     },
     {
-      "epoch": 0.3699156848205213,
+      "epoch": 0.3822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.4528250420390159e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.4169876338400141e-05,
+      "loss": 0.9947,
       "step": 13469
     },
     {
-      "epoch": 0.3699431490483645,
+      "epoch": 0.3822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.4527457310537752e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.416904095955804e-05,
+      "loss": 0.8649,
       "step": 13470
     },
     {
-      "epoch": 0.3699706132762077,
+      "epoch": 0.3822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.4526664164862345e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.4168205545499856e-05,
+      "loss": 0.7772,
       "step": 13471
     },
     {
-      "epoch": 0.36999807750405095,
+      "epoch": 0.3822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.452587098337022e-05,
-      "loss": 0.835,
+      "learning_rate": 1.4167370096232657e-05,
+      "loss": 0.9055,
       "step": 13472
     },
     {
-      "epoch": 0.3700255417318942,
+      "epoch": 0.3823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.4525077766067653e-05,
-      "loss": 1.037,
+      "learning_rate": 1.4166534611763495e-05,
+      "loss": 0.9679,
       "step": 13473
     },
     {
-      "epoch": 0.37005300595973745,
+      "epoch": 0.38234960272417706,
       "grad_norm": 0.0,
-      "learning_rate": 1.4524284512960914e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.4165699092099425e-05,
+      "loss": 0.9016,
       "step": 13474
     },
     {
-      "epoch": 0.37008047018758067,
+      "epoch": 0.382377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.4523491224056286e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.4164863537247509e-05,
+      "loss": 0.8421,
       "step": 13475
     },
     {
-      "epoch": 0.3701079344154239,
+      "epoch": 0.38240635641316684,
       "grad_norm": 0.0,
-      "learning_rate": 1.4522697899360042e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.4164027947214804e-05,
+      "loss": 0.8721,
       "step": 13476
     },
     {
-      "epoch": 0.37013539864326717,
+      "epoch": 0.38243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.452190453887846e-05,
-      "loss": 0.9192,
+      "learning_rate": 1.4163192322008367e-05,
+      "loss": 0.8475,
       "step": 13477
     },
     {
-      "epoch": 0.3701628628711104,
+      "epoch": 0.38246311010215667,
       "grad_norm": 0.0,
-      "learning_rate": 1.4521111142617818e-05,
-      "loss": 0.7604,
+      "learning_rate": 1.4162356661635262e-05,
+      "loss": 0.9843,
       "step": 13478
     },
     {
-      "epoch": 0.3701903270989536,
+      "epoch": 0.3824914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.4520317710584393e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.4161520966102538e-05,
+      "loss": 0.8874,
       "step": 13479
     },
     {
-      "epoch": 0.37021779132679683,
+      "epoch": 0.38251986379114644,
       "grad_norm": 0.0,
-      "learning_rate": 1.4519524242784464e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.4160685235417262e-05,
+      "loss": 0.9256,
       "step": 13480
     },
     {
-      "epoch": 0.3702452555546401,
+      "epoch": 0.3825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.4518730739224309e-05,
-      "loss": 0.9245,
+      "learning_rate": 1.4159849469586489e-05,
+      "loss": 0.8896,
       "step": 13481
     },
     {
-      "epoch": 0.37027271978248333,
+      "epoch": 0.3825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.4517937199910204e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.4159013668617278e-05,
+      "loss": 0.9666,
       "step": 13482
     },
     {
-      "epoch": 0.37030018401032655,
+      "epoch": 0.3826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.451714362484843e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.4158177832516695e-05,
+      "loss": 0.9799,
       "step": 13483
     },
     {
-      "epoch": 0.37032764823816977,
+      "epoch": 0.382633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.4516350014045269e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.4157341961291798e-05,
+      "loss": 0.9574,
       "step": 13484
     },
     {
-      "epoch": 0.370355112466013,
+      "epoch": 0.3826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.4515556367506993e-05,
-      "loss": 0.8539,
+      "learning_rate": 1.4156506054949643e-05,
+      "loss": 0.9881,
       "step": 13485
     },
     {
-      "epoch": 0.37038257669385627,
+      "epoch": 0.38269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.451476268523989e-05,
-      "loss": 1.0234,
+      "learning_rate": 1.4155670113497297e-05,
+      "loss": 0.9844,
       "step": 13486
     },
     {
-      "epoch": 0.3704100409216995,
+      "epoch": 0.3827185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.4513968967250232e-05,
-      "loss": 0.9017,
+      "learning_rate": 1.4154834136941817e-05,
+      "loss": 0.8238,
       "step": 13487
     },
     {
-      "epoch": 0.3704375051495427,
+      "epoch": 0.38274687854710554,
       "grad_norm": 0.0,
-      "learning_rate": 1.4513175213544304e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.4153998125290269e-05,
+      "loss": 0.9243,
       "step": 13488
     },
     {
-      "epoch": 0.37046496937738593,
+      "epoch": 0.38277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.451238142412839e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.4153162078549708e-05,
+      "loss": 0.9908,
       "step": 13489
     },
     {
-      "epoch": 0.3704924336052292,
+      "epoch": 0.38280363223609537,
       "grad_norm": 0.0,
-      "learning_rate": 1.451158759900876e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.4152325996727205e-05,
+      "loss": 1.0161,
       "step": 13490
     },
     {
-      "epoch": 0.37051989783307243,
+      "epoch": 0.38283200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 1.4510793738191701e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.4151489879829814e-05,
+      "loss": 0.886,
       "step": 13491
     },
     {
-      "epoch": 0.37054736206091565,
+      "epoch": 0.38286038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.45099998416835e-05,
-      "loss": 0.9167,
+      "learning_rate": 1.4150653727864605e-05,
+      "loss": 0.9633,
       "step": 13492
     },
     {
-      "epoch": 0.3705748262887589,
+      "epoch": 0.38288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.4509205909490431e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.4149817540838636e-05,
+      "loss": 0.9967,
       "step": 13493
     },
     {
-      "epoch": 0.37060229051660215,
+      "epoch": 0.3829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.4508411941618777e-05,
-      "loss": 0.985,
+      "learning_rate": 1.414898131875897e-05,
+      "loss": 0.8744,
       "step": 13494
     },
     {
-      "epoch": 0.37062975474444537,
+      "epoch": 0.38294551645856983,
       "grad_norm": 0.0,
-      "learning_rate": 1.4507617938074823e-05,
-      "loss": 0.8615,
+      "learning_rate": 1.4148145061632673e-05,
+      "loss": 0.9343,
       "step": 13495
     },
     {
-      "epoch": 0.3706572189722886,
+      "epoch": 0.3829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.4506823898864848e-05,
-      "loss": 0.975,
+      "learning_rate": 1.414730876946681e-05,
+      "loss": 0.972,
       "step": 13496
     },
     {
-      "epoch": 0.3706846832001318,
+      "epoch": 0.3830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.4506029823995137e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.4146472442268439e-05,
+      "loss": 0.9818,
       "step": 13497
     },
     {
-      "epoch": 0.37071214742797504,
+      "epoch": 0.38303064699205447,
       "grad_norm": 0.0,
-      "learning_rate": 1.4505235713471974e-05,
-      "loss": 0.8319,
+      "learning_rate": 1.4145636080044631e-05,
+      "loss": 0.9293,
       "step": 13498
     },
     {
-      "epoch": 0.3707396116558183,
+      "epoch": 0.3830590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.450444156730164e-05,
-      "loss": 0.8626,
+      "learning_rate": 1.414479968280245e-05,
+      "loss": 0.9876,
       "step": 13499
     },
     {
-      "epoch": 0.37076707588366153,
+      "epoch": 0.38308740068104424,
       "grad_norm": 0.0,
-      "learning_rate": 1.4503647385490417e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.414396325054896e-05,
+      "loss": 0.9841,
       "step": 13500
     },
     {
-      "epoch": 0.37079454011150476,
+      "epoch": 0.38311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.4502853168044597e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.4143126783291224e-05,
+      "loss": 0.9554,
       "step": 13501
     },
     {
-      "epoch": 0.370822004339348,
+      "epoch": 0.3831441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.4502058914970454e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.4142290281036312e-05,
+      "loss": 0.9617,
       "step": 13502
     },
     {
-      "epoch": 0.37084946856719125,
+      "epoch": 0.38317253121452893,
       "grad_norm": 0.0,
-      "learning_rate": 1.4501264626274277e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.4141453743791288e-05,
+      "loss": 0.9755,
       "step": 13503
     },
     {
-      "epoch": 0.3708769327950345,
+      "epoch": 0.38320090805902385,
       "grad_norm": 0.0,
-      "learning_rate": 1.450047030196235e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.4140617171563216e-05,
+      "loss": 0.8588,
       "step": 13504
     },
     {
-      "epoch": 0.3709043970228777,
+      "epoch": 0.3832292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.4499675942040963e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.4139780564359162e-05,
+      "loss": 1.0101,
       "step": 13505
     },
     {
-      "epoch": 0.3709318612507209,
+      "epoch": 0.3832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.4498881546516392e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.4138943922186202e-05,
+      "loss": 0.9976,
       "step": 13506
     },
     {
-      "epoch": 0.3709593254785642,
+      "epoch": 0.38328603859250854,
       "grad_norm": 0.0,
-      "learning_rate": 1.4498087115394933e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.4138107245051394e-05,
+      "loss": 0.9117,
       "step": 13507
     },
     {
-      "epoch": 0.3709867897064074,
+      "epoch": 0.3833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.4497292648682862e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.4137270532961806e-05,
+      "loss": 0.9598,
       "step": 13508
     },
     {
-      "epoch": 0.37101425393425064,
+      "epoch": 0.3833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.449649814638647e-05,
-      "loss": 0.8949,
+      "learning_rate": 1.413643378592451e-05,
+      "loss": 0.8824,
       "step": 13509
     },
     {
-      "epoch": 0.37104171816209386,
+      "epoch": 0.38337116912599317,
       "grad_norm": 0.0,
-      "learning_rate": 1.4495703608512044e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.4135597003946573e-05,
+      "loss": 0.9759,
       "step": 13510
     },
     {
-      "epoch": 0.3710691823899371,
+      "epoch": 0.3833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.449490903506587e-05,
-      "loss": 0.8201,
+      "learning_rate": 1.4134760187035059e-05,
+      "loss": 0.8735,
       "step": 13511
     },
     {
-      "epoch": 0.37109664661778036,
+      "epoch": 0.38342792281498295,
       "grad_norm": 0.0,
-      "learning_rate": 1.449411442605423e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.413392333519704e-05,
+      "loss": 0.9319,
       "step": 13512
     },
     {
-      "epoch": 0.3711241108456236,
+      "epoch": 0.38345629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.4493319781483423e-05,
-      "loss": 0.8668,
+      "learning_rate": 1.4133086448439587e-05,
+      "loss": 1.0627,
       "step": 13513
     },
     {
-      "epoch": 0.3711515750734668,
+      "epoch": 0.3834846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.4492525101359724e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.4132249526769765e-05,
+      "loss": 0.9473,
       "step": 13514
     },
     {
-      "epoch": 0.37117903930131,
+      "epoch": 0.38351305334846764,
       "grad_norm": 0.0,
-      "learning_rate": 1.4491730385689428e-05,
-      "loss": 0.9591,
+      "learning_rate": 1.4131412570194648e-05,
+      "loss": 1.0031,
       "step": 13515
     },
     {
-      "epoch": 0.3712065035291533,
+      "epoch": 0.38354143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 1.4490935634478823e-05,
-      "loss": 0.9116,
+      "learning_rate": 1.4130575578721302e-05,
+      "loss": 0.9714,
       "step": 13516
     },
     {
-      "epoch": 0.3712339677569965,
+      "epoch": 0.3835698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.4490140847734194e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.41297385523568e-05,
+      "loss": 0.8675,
       "step": 13517
     },
     {
-      "epoch": 0.37126143198483974,
+      "epoch": 0.3835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.4489346025461829e-05,
-      "loss": 0.9129,
+      "learning_rate": 1.4128901491108208e-05,
+      "loss": 0.9827,
       "step": 13518
     },
     {
-      "epoch": 0.37128889621268296,
+      "epoch": 0.38362656072644724,
       "grad_norm": 0.0,
-      "learning_rate": 1.4488551167668022e-05,
-      "loss": 0.9902,
+      "learning_rate": 1.4128064394982601e-05,
+      "loss": 0.8766,
       "step": 13519
     },
     {
-      "epoch": 0.37131636044052624,
+      "epoch": 0.3836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.4487756274359059e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.4127227263987047e-05,
+      "loss": 1.0094,
       "step": 13520
     },
     {
-      "epoch": 0.37134382466836946,
+      "epoch": 0.383683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.448696134554123e-05,
-      "loss": 0.8862,
+      "learning_rate": 1.412639009812862e-05,
+      "loss": 0.8596,
       "step": 13521
     },
     {
-      "epoch": 0.3713712888962127,
+      "epoch": 0.3837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.4486166381220826e-05,
-      "loss": 1.0545,
+      "learning_rate": 1.412555289741439e-05,
+      "loss": 0.908,
       "step": 13522
     },
     {
-      "epoch": 0.3713987531240559,
+      "epoch": 0.3837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.4485371381404133e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.412471566185143e-05,
+      "loss": 0.898,
       "step": 13523
     },
     {
-      "epoch": 0.3714262173518991,
+      "epoch": 0.3837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.4484576346097443e-05,
-      "loss": 0.8753,
+      "learning_rate": 1.412387839144681e-05,
+      "loss": 0.9109,
       "step": 13524
     },
     {
-      "epoch": 0.3714536815797424,
+      "epoch": 0.38379682179341656,
       "grad_norm": 0.0,
-      "learning_rate": 1.4483781275307055e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.4123041086207606e-05,
+      "loss": 0.9963,
       "step": 13525
     },
     {
-      "epoch": 0.3714811458075856,
+      "epoch": 0.3838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.4482986169039247e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.4122203746140886e-05,
+      "loss": 0.9318,
       "step": 13526
     },
     {
-      "epoch": 0.37150861003542884,
+      "epoch": 0.38385357548240634,
       "grad_norm": 0.0,
-      "learning_rate": 1.4482191027300314e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.4121366371253728e-05,
+      "loss": 0.8675,
       "step": 13527
     },
     {
-      "epoch": 0.37153607426327206,
+      "epoch": 0.38388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.4481395850096553e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.4120528961553199e-05,
+      "loss": 0.9452,
       "step": 13528
     },
     {
-      "epoch": 0.37156353849111534,
+      "epoch": 0.3839103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.4480600637434253e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.4119691517046379e-05,
+      "loss": 0.8725,
       "step": 13529
     },
     {
-      "epoch": 0.37159100271895856,
+      "epoch": 0.38393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.44798053893197e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.4118854037740341e-05,
+      "loss": 0.9239,
       "step": 13530
     },
     {
-      "epoch": 0.3716184669468018,
+      "epoch": 0.38396708286038594,
       "grad_norm": 0.0,
-      "learning_rate": 1.4479010105759198e-05,
-      "loss": 1.0106,
+      "learning_rate": 1.4118016523642158e-05,
+      "loss": 0.9247,
       "step": 13531
     },
     {
-      "epoch": 0.371645931174645,
+      "epoch": 0.3839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.4478214786759028e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.4117178974758903e-05,
+      "loss": 0.8668,
       "step": 13532
     },
     {
-      "epoch": 0.3716733954024883,
+      "epoch": 0.3840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.4477419432325493e-05,
-      "loss": 0.8684,
+      "learning_rate": 1.4116341391097652e-05,
+      "loss": 1.0156,
       "step": 13533
     },
     {
-      "epoch": 0.3717008596303315,
+      "epoch": 0.3840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.4476624042464877e-05,
-      "loss": 0.8974,
+      "learning_rate": 1.4115503772665483e-05,
+      "loss": 0.8564,
       "step": 13534
     },
     {
-      "epoch": 0.3717283238581747,
+      "epoch": 0.3840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.4475828617183479e-05,
-      "loss": 0.7864,
+      "learning_rate": 1.4114666119469463e-05,
+      "loss": 0.9891,
       "step": 13535
     },
     {
-      "epoch": 0.37175578808601795,
+      "epoch": 0.3841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.447503315648759e-05,
-      "loss": 0.9958,
+      "learning_rate": 1.4113828431516676e-05,
+      "loss": 1.0218,
       "step": 13536
     },
     {
-      "epoch": 0.3717832523138612,
+      "epoch": 0.38413734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.4474237660383507e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.4112990708814195e-05,
+      "loss": 0.8958,
       "step": 13537
     },
     {
-      "epoch": 0.37181071654170444,
+      "epoch": 0.3841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.4473442128877519e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.4112152951369097e-05,
+      "loss": 0.9057,
       "step": 13538
     },
     {
-      "epoch": 0.37183818076954767,
+      "epoch": 0.38419409761634504,
       "grad_norm": 0.0,
-      "learning_rate": 1.447264656197593e-05,
-      "loss": 0.941,
+      "learning_rate": 1.4111315159188461e-05,
+      "loss": 0.8172,
       "step": 13539
     },
     {
-      "epoch": 0.3718656449973909,
+      "epoch": 0.38422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.4471850959685023e-05,
-      "loss": 0.97,
+      "learning_rate": 1.411047733227936e-05,
+      "loss": 0.9689,
       "step": 13540
     },
     {
-      "epoch": 0.3718931092252341,
+      "epoch": 0.3842508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.4471055322011101e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.4109639470648868e-05,
+      "loss": 1.0025,
       "step": 13541
     },
     {
-      "epoch": 0.3719205734530774,
+      "epoch": 0.38427922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 1.4470259648960458e-05,
-      "loss": 0.908,
+      "learning_rate": 1.4108801574304072e-05,
+      "loss": 0.9518,
       "step": 13542
     },
     {
-      "epoch": 0.3719480376809206,
+      "epoch": 0.38430760499432465,
       "grad_norm": 0.0,
-      "learning_rate": 1.446946394053939e-05,
-      "loss": 0.8859,
+      "learning_rate": 1.4107963643252038e-05,
+      "loss": 0.9892,
       "step": 13543
     },
     {
-      "epoch": 0.37197550190876383,
+      "epoch": 0.3843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.4468668196754188e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.4107125677499856e-05,
+      "loss": 0.9177,
       "step": 13544
     },
     {
-      "epoch": 0.37200296613660705,
+      "epoch": 0.3843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.4467872417611157e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.4106287677054597e-05,
+      "loss": 0.8376,
       "step": 13545
     },
     {
-      "epoch": 0.3720304303644503,
+      "epoch": 0.3843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.4467076603116588e-05,
-      "loss": 0.9255,
+      "learning_rate": 1.4105449641923341e-05,
+      "loss": 1.0765,
       "step": 13546
     },
     {
-      "epoch": 0.37205789459229355,
+      "epoch": 0.3844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.4466280753276775e-05,
-      "loss": 0.8688,
+      "learning_rate": 1.410461157211317e-05,
+      "loss": 0.9201,
       "step": 13547
     },
     {
-      "epoch": 0.37208535882013677,
+      "epoch": 0.3844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.4465484868098022e-05,
-      "loss": 0.8446,
+      "learning_rate": 1.4103773467631157e-05,
+      "loss": 0.9908,
       "step": 13548
     },
     {
-      "epoch": 0.37211282304798,
+      "epoch": 0.38447786606129397,
       "grad_norm": 0.0,
-      "learning_rate": 1.4464688947586623e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.4102935328484385e-05,
+      "loss": 0.8802,
       "step": 13549
     },
     {
-      "epoch": 0.37214028727582327,
+      "epoch": 0.3845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.446389299174887e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.4102097154679938e-05,
+      "loss": 0.9147,
       "step": 13550
     },
     {
-      "epoch": 0.3721677515036665,
+      "epoch": 0.38453461975028375,
       "grad_norm": 0.0,
-      "learning_rate": 1.4463097000591074e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.4101258946224888e-05,
+      "loss": 0.9078,
       "step": 13551
     },
     {
-      "epoch": 0.3721952157315097,
+      "epoch": 0.38456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.4462300974119521e-05,
-      "loss": 0.943,
+      "learning_rate": 1.4100420703126318e-05,
+      "loss": 0.9374,
       "step": 13552
     },
     {
-      "epoch": 0.37222267995935293,
+      "epoch": 0.3845913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.4461504912340515e-05,
-      "loss": 0.8929,
+      "learning_rate": 1.4099582425391314e-05,
+      "loss": 0.8732,
       "step": 13553
     },
     {
-      "epoch": 0.37225014418719615,
+      "epoch": 0.38461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.4460708815260356e-05,
-      "loss": 0.8692,
+      "learning_rate": 1.4098744113026951e-05,
+      "loss": 0.9956,
       "step": 13554
     },
     {
-      "epoch": 0.37227760841503943,
+      "epoch": 0.38464812712826335,
       "grad_norm": 0.0,
-      "learning_rate": 1.445991268288534e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.4097905766040312e-05,
+      "loss": 0.8551,
       "step": 13555
     },
     {
-      "epoch": 0.37230507264288265,
+      "epoch": 0.3846765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.4459116515221765e-05,
-      "loss": 0.9831,
+      "learning_rate": 1.409706738443848e-05,
+      "loss": 0.9171,
       "step": 13556
     },
     {
-      "epoch": 0.37233253687072587,
+      "epoch": 0.3847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.4458320312275938e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.4096228968228536e-05,
+      "loss": 1.0029,
       "step": 13557
     },
     {
-      "epoch": 0.3723600010985691,
+      "epoch": 0.38473325766174804,
       "grad_norm": 0.0,
-      "learning_rate": 1.4457524074054148e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.409539051741756e-05,
+      "loss": 0.8823,
       "step": 13558
     },
     {
-      "epoch": 0.37238746532641237,
+      "epoch": 0.3847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.4456727800562703e-05,
-      "loss": 1.0489,
+      "learning_rate": 1.4094552032012635e-05,
+      "loss": 1.0027,
       "step": 13559
     },
     {
-      "epoch": 0.3724149295542556,
+      "epoch": 0.3847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.4455931491807902e-05,
-      "loss": 0.9248,
+      "learning_rate": 1.4093713512020848e-05,
+      "loss": 0.8735,
       "step": 13560
     },
     {
-      "epoch": 0.3724423937820988,
+      "epoch": 0.3848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.4455135147796044e-05,
-      "loss": 1.0721,
+      "learning_rate": 1.4092874957449278e-05,
+      "loss": 0.9194,
       "step": 13561
     },
     {
-      "epoch": 0.37246985800994203,
+      "epoch": 0.3848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.445433876853343e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.409203636830501e-05,
+      "loss": 0.9693,
       "step": 13562
     },
     {
-      "epoch": 0.3724973222377853,
+      "epoch": 0.38487514188422245,
       "grad_norm": 0.0,
-      "learning_rate": 1.4453542354026366e-05,
-      "loss": 0.8929,
+      "learning_rate": 1.4091197744595127e-05,
+      "loss": 0.8871,
       "step": 13563
     },
     {
-      "epoch": 0.37252478646562853,
+      "epoch": 0.38490351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.4452745904281144e-05,
-      "loss": 0.8827,
+      "learning_rate": 1.4090359086326712e-05,
+      "loss": 0.967,
       "step": 13564
     },
     {
-      "epoch": 0.37255225069347175,
+      "epoch": 0.3849318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.445194941930408e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.4089520393506848e-05,
+      "loss": 0.9157,
       "step": 13565
     },
     {
-      "epoch": 0.372579714921315,
+      "epoch": 0.38496027241770714,
       "grad_norm": 0.0,
-      "learning_rate": 1.4451152899101462e-05,
-      "loss": 0.9967,
+      "learning_rate": 1.4088681666142621e-05,
+      "loss": 0.8933,
       "step": 13566
     },
     {
-      "epoch": 0.3726071791491582,
+      "epoch": 0.38498864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.4450356343679596e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.408784290424112e-05,
+      "loss": 0.8422,
       "step": 13567
     },
     {
-      "epoch": 0.3726346433770015,
+      "epoch": 0.3850170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.4449559753044788e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.4087004107809424e-05,
+      "loss": 0.9214,
       "step": 13568
     },
     {
-      "epoch": 0.3726621076048447,
+      "epoch": 0.38504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.4448763127203344e-05,
-      "loss": 0.8862,
+      "learning_rate": 1.408616527685462e-05,
+      "loss": 1.0362,
       "step": 13569
     },
     {
-      "epoch": 0.3726895718326879,
+      "epoch": 0.38507377979568674,
       "grad_norm": 0.0,
-      "learning_rate": 1.444796646616156e-05,
-      "loss": 0.8676,
+      "learning_rate": 1.4085326411383796e-05,
+      "loss": 0.9253,
       "step": 13570
     },
     {
-      "epoch": 0.37271703606053114,
+      "epoch": 0.3851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.444716976992574e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.4084487511404036e-05,
+      "loss": 1.0077,
       "step": 13571
     },
     {
-      "epoch": 0.3727445002883744,
+      "epoch": 0.3851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.4446373038502193e-05,
-      "loss": 0.9391,
+      "learning_rate": 1.4083648576922424e-05,
+      "loss": 0.9126,
       "step": 13572
     },
     {
-      "epoch": 0.37277196451621764,
+      "epoch": 0.3851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.444557627189722e-05,
-      "loss": 0.8798,
+      "learning_rate": 1.4082809607946053e-05,
+      "loss": 0.9982,
       "step": 13573
     },
     {
-      "epoch": 0.37279942874406086,
+      "epoch": 0.3851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.4444779470117125e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.4081970604482003e-05,
+      "loss": 0.9355,
       "step": 13574
     },
     {
-      "epoch": 0.3728268929719041,
+      "epoch": 0.3852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.4443982633168214e-05,
-      "loss": 1.0055,
+      "learning_rate": 1.4081131566537365e-05,
+      "loss": 1.0021,
       "step": 13575
     },
     {
-      "epoch": 0.37285435719974735,
+      "epoch": 0.38524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.4443185761056789e-05,
-      "loss": 0.8742,
+      "learning_rate": 1.408029249411922e-05,
+      "loss": 0.8268,
       "step": 13576
     },
     {
-      "epoch": 0.3728818214275906,
+      "epoch": 0.385272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.4442388853789158e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.4079453387234667e-05,
+      "loss": 0.9336,
       "step": 13577
     },
     {
-      "epoch": 0.3729092856554338,
+      "epoch": 0.38530079455164584,
       "grad_norm": 0.0,
-      "learning_rate": 1.4441591911371626e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.4078614245890787e-05,
+      "loss": 1.0225,
       "step": 13578
     },
     {
-      "epoch": 0.372936749883277,
+      "epoch": 0.38532917139614076,
       "grad_norm": 0.0,
-      "learning_rate": 1.4440794933810498e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.4077775070094667e-05,
+      "loss": 0.9746,
       "step": 13579
     },
     {
-      "epoch": 0.37296421411112024,
+      "epoch": 0.3853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.443999792111208e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.40769358598534e-05,
+      "loss": 0.829,
       "step": 13580
     },
     {
-      "epoch": 0.3729916783389635,
+      "epoch": 0.38538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.4439200873282682e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.4076096615174071e-05,
+      "loss": 0.7601,
       "step": 13581
     },
     {
-      "epoch": 0.37301914256680674,
+      "epoch": 0.38541430192962545,
       "grad_norm": 0.0,
-      "learning_rate": 1.4438403790328601e-05,
-      "loss": 0.8794,
+      "learning_rate": 1.4075257336063767e-05,
+      "loss": 0.7696,
       "step": 13582
     },
     {
-      "epoch": 0.37304660679464996,
+      "epoch": 0.3854426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.4437606672256152e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.4074418022529589e-05,
+      "loss": 0.9553,
       "step": 13583
     },
     {
-      "epoch": 0.3730740710224932,
+      "epoch": 0.3854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.4436809519071641e-05,
-      "loss": 0.8565,
+      "learning_rate": 1.4073578674578615e-05,
+      "loss": 0.8412,
       "step": 13584
     },
     {
-      "epoch": 0.37310153525033646,
+      "epoch": 0.3854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.4436012330781375e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.4072739292217939e-05,
+      "loss": 0.9452,
       "step": 13585
     },
     {
-      "epoch": 0.3731289994781797,
+      "epoch": 0.385527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.443521510739166e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.407189987545465e-05,
+      "loss": 1.0103,
       "step": 13586
     },
     {
-      "epoch": 0.3731564637060229,
+      "epoch": 0.3855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.4434417848908807e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.4071060424295841e-05,
+      "loss": 0.8955,
       "step": 13587
     },
     {
-      "epoch": 0.3731839279338661,
+      "epoch": 0.38558456299659477,
       "grad_norm": 0.0,
-      "learning_rate": 1.4433620555339118e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.4070220938748602e-05,
+      "loss": 0.9821,
       "step": 13588
     },
     {
-      "epoch": 0.3732113921617094,
+      "epoch": 0.3856129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.443282322668891e-05,
-      "loss": 0.8801,
+      "learning_rate": 1.406938141882002e-05,
+      "loss": 0.9695,
       "step": 13589
     },
     {
-      "epoch": 0.3732388563895526,
+      "epoch": 0.38564131668558455,
       "grad_norm": 0.0,
-      "learning_rate": 1.4432025862964487e-05,
-      "loss": 1.0107,
+      "learning_rate": 1.4068541864517193e-05,
+      "loss": 0.9828,
       "step": 13590
     },
     {
-      "epoch": 0.37326632061739584,
+      "epoch": 0.38566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.4431228464172158e-05,
-      "loss": 0.9066,
+      "learning_rate": 1.4067702275847212e-05,
+      "loss": 0.9124,
       "step": 13591
     },
     {
-      "epoch": 0.37329378484523906,
+      "epoch": 0.3856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.4430431030318231e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.4066862652817164e-05,
+      "loss": 0.9063,
       "step": 13592
     },
     {
-      "epoch": 0.3733212490730823,
+      "epoch": 0.38572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.442963356140902e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.4066022995434145e-05,
+      "loss": 0.8731,
       "step": 13593
     },
     {
-      "epoch": 0.37334871330092556,
+      "epoch": 0.38575482406356415,
       "grad_norm": 0.0,
-      "learning_rate": 1.4428836057450833e-05,
-      "loss": 1.0181,
+      "learning_rate": 1.406518330370525e-05,
+      "loss": 0.9989,
       "step": 13594
     },
     {
-      "epoch": 0.3733761775287688,
+      "epoch": 0.385783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.4428038518449973e-05,
-      "loss": 0.8868,
+      "learning_rate": 1.4064343577637565e-05,
+      "loss": 0.9968,
       "step": 13595
     },
     {
-      "epoch": 0.373403641756612,
+      "epoch": 0.3858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.4427240944412763e-05,
-      "loss": 1.0329,
+      "learning_rate": 1.4063503817238189e-05,
+      "loss": 0.9883,
       "step": 13596
     },
     {
-      "epoch": 0.3734311059844552,
+      "epoch": 0.3858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.4426443335345508e-05,
-      "loss": 0.9264,
+      "learning_rate": 1.4062664022514212e-05,
+      "loss": 0.8848,
       "step": 13597
     },
     {
-      "epoch": 0.3734585702122985,
+      "epoch": 0.3858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.4425645691254518e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.406182419347273e-05,
+      "loss": 0.9814,
       "step": 13598
     },
     {
-      "epoch": 0.3734860344401417,
+      "epoch": 0.3858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.4424848012146104e-05,
-      "loss": 0.8284,
+      "learning_rate": 1.4060984330120836e-05,
+      "loss": 0.9165,
       "step": 13599
     },
     {
-      "epoch": 0.37351349866798494,
+      "epoch": 0.3859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.4424050298026577e-05,
-      "loss": 1.0126,
+      "learning_rate": 1.4060144432465627e-05,
+      "loss": 0.9974,
       "step": 13600
     },
     {
-      "epoch": 0.37354096289582817,
+      "epoch": 0.3859534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.4423252548902256e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.4059304500514195e-05,
+      "loss": 0.8704,
       "step": 13601
     },
     {
-      "epoch": 0.37356842712367144,
+      "epoch": 0.38598183881952325,
       "grad_norm": 0.0,
-      "learning_rate": 1.4422454764779442e-05,
-      "loss": 1.0387,
+      "learning_rate": 1.4058464534273632e-05,
+      "loss": 0.9155,
       "step": 13602
     },
     {
-      "epoch": 0.37359589135151466,
+      "epoch": 0.38601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.4421656945664453e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.4057624533751042e-05,
+      "loss": 0.9316,
       "step": 13603
     },
     {
-      "epoch": 0.3736233555793579,
+      "epoch": 0.3860385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 1.4420859091563603e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.4056784498953511e-05,
+      "loss": 1.035,
       "step": 13604
     },
     {
-      "epoch": 0.3736508198072011,
+      "epoch": 0.38606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.4420061202483205e-05,
-      "loss": 0.8553,
+      "learning_rate": 1.4055944429888139e-05,
+      "loss": 0.8561,
       "step": 13605
     },
     {
-      "epoch": 0.37367828403504433,
+      "epoch": 0.38609534619750285,
       "grad_norm": 0.0,
-      "learning_rate": 1.4419263278429567e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.4055104326562024e-05,
+      "loss": 0.8342,
       "step": 13606
     },
     {
-      "epoch": 0.3737057482628876,
+      "epoch": 0.3861237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.441846531940901e-05,
-      "loss": 1.0627,
+      "learning_rate": 1.4054264188982261e-05,
+      "loss": 0.9075,
       "step": 13607
     },
     {
-      "epoch": 0.3737332124907308,
+      "epoch": 0.38615209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.441766732542784e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.4053424017155945e-05,
+      "loss": 0.924,
       "step": 13608
     },
     {
-      "epoch": 0.37376067671857405,
+      "epoch": 0.3861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.4416869296492378e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.4052583811090173e-05,
+      "loss": 0.8918,
       "step": 13609
     },
     {
-      "epoch": 0.37378814094641727,
+      "epoch": 0.3862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.4416071232608935e-05,
-      "loss": 0.8421,
+      "learning_rate": 1.4051743570792046e-05,
+      "loss": 0.8485,
       "step": 13610
     },
     {
-      "epoch": 0.37381560517426055,
+      "epoch": 0.3862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.4415273133783823e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.4050903296268658e-05,
+      "loss": 0.9542,
       "step": 13611
     },
     {
-      "epoch": 0.37384306940210377,
+      "epoch": 0.3862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.4414475000023358e-05,
-      "loss": 0.8162,
+      "learning_rate": 1.4050062987527104e-05,
+      "loss": 0.9597,
       "step": 13612
     },
     {
-      "epoch": 0.373870533629947,
+      "epoch": 0.3862939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.4413676831333864e-05,
-      "loss": 1.0159,
+      "learning_rate": 1.404922264457449e-05,
+      "loss": 0.9935,
       "step": 13613
     },
     {
-      "epoch": 0.3738979978577902,
+      "epoch": 0.38632236095346195,
       "grad_norm": 0.0,
-      "learning_rate": 1.4412878627721645e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.404838226741791e-05,
+      "loss": 0.9188,
       "step": 13614
     },
     {
-      "epoch": 0.3739254620856335,
+      "epoch": 0.38635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.4412080389193019e-05,
-      "loss": 0.9208,
+      "learning_rate": 1.4047541856064464e-05,
+      "loss": 0.9537,
       "step": 13615
     },
     {
-      "epoch": 0.3739529263134767,
+      "epoch": 0.3863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.4411282115754306e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.4046701410521247e-05,
+      "loss": 0.8551,
       "step": 13616
     },
     {
-      "epoch": 0.37398039054131993,
+      "epoch": 0.38640749148694664,
       "grad_norm": 0.0,
-      "learning_rate": 1.4410483807411821e-05,
-      "loss": 0.8703,
+      "learning_rate": 1.4045860930795362e-05,
+      "loss": 1.0273,
       "step": 13617
     },
     {
-      "epoch": 0.37400785476916315,
+      "epoch": 0.38643586833144156,
       "grad_norm": 0.0,
-      "learning_rate": 1.440968546417188e-05,
-      "loss": 0.8948,
+      "learning_rate": 1.4045020416893911e-05,
+      "loss": 0.9428,
       "step": 13618
     },
     {
-      "epoch": 0.37403531899700637,
+      "epoch": 0.3864642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.4408887086040797e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.4044179868823986e-05,
+      "loss": 0.9687,
       "step": 13619
     },
     {
-      "epoch": 0.37406278322484965,
+      "epoch": 0.38649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.4408088673024891e-05,
-      "loss": 0.8709,
+      "learning_rate": 1.4043339286592691e-05,
+      "loss": 1.0078,
       "step": 13620
     },
     {
-      "epoch": 0.37409024745269287,
+      "epoch": 0.38652099886492625,
       "grad_norm": 0.0,
-      "learning_rate": 1.4407290225130484e-05,
-      "loss": 0.8905,
+      "learning_rate": 1.4042498670207131e-05,
+      "loss": 0.962,
       "step": 13621
     },
     {
-      "epoch": 0.3741177116805361,
+      "epoch": 0.3865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.4406491742363889e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.4041658019674405e-05,
+      "loss": 0.9742,
       "step": 13622
     },
     {
-      "epoch": 0.3741451759083793,
+      "epoch": 0.386577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.4405693224731421e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.4040817335001608e-05,
+      "loss": 0.9011,
       "step": 13623
     },
     {
-      "epoch": 0.3741726401362226,
+      "epoch": 0.3866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.44048946722394e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.4039976616195848e-05,
+      "loss": 0.9728,
       "step": 13624
     },
     {
-      "epoch": 0.3742001043640658,
+      "epoch": 0.3866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.4404096084894152e-05,
-      "loss": 1.0257,
+      "learning_rate": 1.4039135863264221e-05,
+      "loss": 0.9106,
       "step": 13625
     },
     {
-      "epoch": 0.37422756859190903,
+      "epoch": 0.38666288308740066,
       "grad_norm": 0.0,
-      "learning_rate": 1.4403297462701988e-05,
-      "loss": 1.0095,
+      "learning_rate": 1.4038295076213833e-05,
+      "loss": 0.9235,
       "step": 13626
     },
     {
-      "epoch": 0.37425503281975225,
+      "epoch": 0.38669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.4402498805669228e-05,
-      "loss": 0.8608,
+      "learning_rate": 1.4037454255051786e-05,
+      "loss": 0.9141,
       "step": 13627
     },
     {
-      "epoch": 0.37428249704759553,
+      "epoch": 0.3867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.440170011380219e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.403661339978518e-05,
+      "loss": 0.86,
       "step": 13628
     },
     {
-      "epoch": 0.37430996127543875,
+      "epoch": 0.38674801362088534,
       "grad_norm": 0.0,
-      "learning_rate": 1.44009013871072e-05,
-      "loss": 0.8683,
+      "learning_rate": 1.403577251042112e-05,
+      "loss": 0.8888,
       "step": 13629
     },
     {
-      "epoch": 0.374337425503282,
+      "epoch": 0.38677639046538026,
       "grad_norm": 0.0,
-      "learning_rate": 1.440010262559057e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.4034931586966707e-05,
+      "loss": 0.9783,
       "step": 13630
     },
     {
-      "epoch": 0.3743648897311252,
+      "epoch": 0.3868047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.4399303829258624e-05,
-      "loss": 1.0873,
+      "learning_rate": 1.4034090629429047e-05,
+      "loss": 1.0879,
       "step": 13631
     },
     {
-      "epoch": 0.37439235395896847,
+      "epoch": 0.38683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.439850499811768e-05,
-      "loss": 0.9004,
+      "learning_rate": 1.4033249637815242e-05,
+      "loss": 0.8742,
       "step": 13632
     },
     {
-      "epoch": 0.3744198181868117,
+      "epoch": 0.38686152099886495,
       "grad_norm": 0.0,
-      "learning_rate": 1.4397706132174064e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.4032408612132395e-05,
+      "loss": 0.8358,
       "step": 13633
     },
     {
-      "epoch": 0.3744472824146549,
+      "epoch": 0.3868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.4396907231434096e-05,
-      "loss": 0.9798,
+      "learning_rate": 1.4031567552387613e-05,
+      "loss": 0.8784,
       "step": 13634
     },
     {
-      "epoch": 0.37447474664249814,
+      "epoch": 0.3869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.439610829590409e-05,
-      "loss": 0.9875,
+      "learning_rate": 1.4030726458587997e-05,
+      "loss": 0.918,
       "step": 13635
     },
     {
-      "epoch": 0.37450221087034136,
+      "epoch": 0.3869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.4395309325590372e-05,
-      "loss": 0.8741,
+      "learning_rate": 1.4029885330740652e-05,
+      "loss": 0.9112,
       "step": 13636
     },
     {
-      "epoch": 0.37452967509818463,
+      "epoch": 0.3869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.439451032049927e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.4029044168852689e-05,
+      "loss": 0.8746,
       "step": 13637
     },
     {
-      "epoch": 0.37455713932602785,
+      "epoch": 0.3870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.4393711280637095e-05,
-      "loss": 1.0764,
+      "learning_rate": 1.4028202972931206e-05,
+      "loss": 0.9448,
       "step": 13638
     },
     {
-      "epoch": 0.3745846035538711,
+      "epoch": 0.3870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.4392912206010175e-05,
-      "loss": 0.929,
+      "learning_rate": 1.402736174298331e-05,
+      "loss": 0.8313,
       "step": 13639
     },
     {
-      "epoch": 0.3746120677817143,
+      "epoch": 0.3870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.439211309662483e-05,
-      "loss": 0.8836,
+      "learning_rate": 1.402652047901611e-05,
+      "loss": 0.8634,
       "step": 13640
     },
     {
-      "epoch": 0.3746395320095576,
+      "epoch": 0.38708853575482405,
       "grad_norm": 0.0,
-      "learning_rate": 1.4391313952487388e-05,
-      "loss": 1.0279,
+      "learning_rate": 1.402567918103671e-05,
+      "loss": 0.8962,
       "step": 13641
     },
     {
-      "epoch": 0.3746669962374008,
+      "epoch": 0.38711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.439051477360417e-05,
-      "loss": 0.8849,
+      "learning_rate": 1.4024837849052219e-05,
+      "loss": 0.8576,
       "step": 13642
     },
     {
-      "epoch": 0.374694460465244,
+      "epoch": 0.3871452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.4389715559981496e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.4023996483069738e-05,
+      "loss": 1.0539,
       "step": 13643
     },
     {
-      "epoch": 0.37472192469308724,
+      "epoch": 0.38717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.4388916311625694e-05,
-      "loss": 0.8832,
+      "learning_rate": 1.402315508309638e-05,
+      "loss": 0.9715,
       "step": 13644
     },
     {
-      "epoch": 0.3747493889209305,
+      "epoch": 0.38720204313280365,
       "grad_norm": 0.0,
-      "learning_rate": 1.4388117028543084e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.4022313649139252e-05,
+      "loss": 0.9398,
       "step": 13645
     },
     {
-      "epoch": 0.37477685314877374,
+      "epoch": 0.3872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.4387317710739995e-05,
-      "loss": 1.0297,
+      "learning_rate": 1.4021472181205455e-05,
+      "loss": 0.8498,
       "step": 13646
     },
     {
-      "epoch": 0.37480431737661696,
+      "epoch": 0.3872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.4386518358222744e-05,
-      "loss": 0.8283,
+      "learning_rate": 1.4020630679302105e-05,
+      "loss": 0.9058,
       "step": 13647
     },
     {
-      "epoch": 0.3748317816044602,
+      "epoch": 0.3872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.4385718970997662e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.401978914343631e-05,
+      "loss": 0.9099,
       "step": 13648
     },
     {
-      "epoch": 0.3748592458323034,
+      "epoch": 0.3873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.4384919549071077e-05,
-      "loss": 0.9976,
+      "learning_rate": 1.401894757361517e-05,
+      "loss": 0.9833,
       "step": 13649
     },
     {
-      "epoch": 0.3748867100601467,
+      "epoch": 0.3873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.4384120092449307e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.4018105969845798e-05,
+      "loss": 0.8356,
       "step": 13650
     },
     {
-      "epoch": 0.3749141742879899,
+      "epoch": 0.387372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.4383320601138678e-05,
-      "loss": 0.8414,
+      "learning_rate": 1.401726433213531e-05,
+      "loss": 0.9886,
       "step": 13651
     },
     {
-      "epoch": 0.3749416385158331,
+      "epoch": 0.3874006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.4382521075145523e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.4016422660490807e-05,
+      "loss": 0.8752,
       "step": 13652
     },
     {
-      "epoch": 0.37496910274367634,
+      "epoch": 0.38742905788876275,
       "grad_norm": 0.0,
-      "learning_rate": 1.4381721514476161e-05,
-      "loss": 1.0459,
+      "learning_rate": 1.4015580954919399e-05,
+      "loss": 0.9297,
       "step": 13653
     },
     {
-      "epoch": 0.3749965669715196,
+      "epoch": 0.38745743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 1.438092191913692e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.4014739215428202e-05,
+      "loss": 1.0198,
       "step": 13654
     },
     {
-      "epoch": 0.37502403119936284,
+      "epoch": 0.3874858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.438012228913413e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.4013897442024323e-05,
+      "loss": 0.8653,
       "step": 13655
     },
     {
-      "epoch": 0.37505149542720606,
+      "epoch": 0.38751418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.4379322624474115e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.4013055634714867e-05,
+      "loss": 1.0004,
       "step": 13656
     },
     {
-      "epoch": 0.3750789596550493,
+      "epoch": 0.38754256526674236,
       "grad_norm": 0.0,
-      "learning_rate": 1.4378522925163205e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.4012213793506953e-05,
+      "loss": 0.9881,
       "step": 13657
     },
     {
-      "epoch": 0.37510642388289256,
+      "epoch": 0.3875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.4377723191207724e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.4011371918407686e-05,
+      "loss": 0.8707,
       "step": 13658
     },
     {
-      "epoch": 0.3751338881107358,
+      "epoch": 0.38759931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 1.4376923422614e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.4010530009424183e-05,
+      "loss": 1.0424,
       "step": 13659
     },
     {
-      "epoch": 0.375161352338579,
+      "epoch": 0.387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.4376123619388364e-05,
-      "loss": 0.8142,
+      "learning_rate": 1.4009688066563552e-05,
+      "loss": 0.8535,
       "step": 13660
     },
     {
-      "epoch": 0.3751888165664222,
+      "epoch": 0.3876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.4375323781537143e-05,
-      "loss": 1.0228,
+      "learning_rate": 1.4008846089832906e-05,
+      "loss": 0.9977,
       "step": 13661
     },
     {
-      "epoch": 0.37521628079426544,
+      "epoch": 0.3876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.4374523909066664e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.4008004079239358e-05,
+      "loss": 0.8802,
       "step": 13662
     },
     {
-      "epoch": 0.3752437450221087,
+      "epoch": 0.3877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.437372400198326e-05,
-      "loss": 0.8181,
+      "learning_rate": 1.4007162034790015e-05,
+      "loss": 0.9321,
       "step": 13663
     },
     {
-      "epoch": 0.37527120924995194,
+      "epoch": 0.3877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.4372924060293253e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.4006319956491998e-05,
+      "loss": 0.8367,
       "step": 13664
     },
     {
-      "epoch": 0.37529867347779516,
+      "epoch": 0.38776958002270145,
       "grad_norm": 0.0,
-      "learning_rate": 1.4372124084002981e-05,
-      "loss": 0.9155,
+      "learning_rate": 1.4005477844352415e-05,
+      "loss": 1.059,
       "step": 13665
     },
     {
-      "epoch": 0.3753261377056384,
+      "epoch": 0.38779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.4371324073118769e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.400463569837838e-05,
+      "loss": 0.8497,
       "step": 13666
     },
     {
-      "epoch": 0.37535360193348166,
+      "epoch": 0.3878263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.4370524027646947e-05,
-      "loss": 1.1219,
+      "learning_rate": 1.4003793518577007e-05,
+      "loss": 0.9682,
       "step": 13667
     },
     {
-      "epoch": 0.3753810661613249,
+      "epoch": 0.38785471055618614,
       "grad_norm": 0.0,
-      "learning_rate": 1.4369723947593842e-05,
-      "loss": 1.004,
+      "learning_rate": 1.400295130495541e-05,
+      "loss": 0.7921,
       "step": 13668
     },
     {
-      "epoch": 0.3754085303891681,
+      "epoch": 0.38788308740068106,
       "grad_norm": 0.0,
-      "learning_rate": 1.4368923832965795e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.4002109057520707e-05,
+      "loss": 0.9578,
       "step": 13669
     },
     {
-      "epoch": 0.3754359946170113,
+      "epoch": 0.3879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 1.4368123683769126e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.4001266776280005e-05,
+      "loss": 0.916,
       "step": 13670
     },
     {
-      "epoch": 0.3754634588448546,
+      "epoch": 0.38793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.4367323500010171e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.4000424461240426e-05,
+      "loss": 0.9869,
       "step": 13671
     },
     {
-      "epoch": 0.3754909230726978,
+      "epoch": 0.3879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.4366523281695264e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.3999582112409078e-05,
+      "loss": 0.9831,
       "step": 13672
     },
     {
-      "epoch": 0.37551838730054105,
+      "epoch": 0.3879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.436572302883073e-05,
-      "loss": 0.8311,
+      "learning_rate": 1.3998739729793083e-05,
+      "loss": 0.9439,
       "step": 13673
     },
     {
-      "epoch": 0.37554585152838427,
+      "epoch": 0.3880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.4364922741422905e-05,
-      "loss": 1.0215,
+      "learning_rate": 1.3997897313399552e-05,
+      "loss": 0.9266,
       "step": 13674
     },
     {
-      "epoch": 0.3755733157562275,
+      "epoch": 0.3880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.4364122419478119e-05,
-      "loss": 0.8746,
+      "learning_rate": 1.3997054863235601e-05,
+      "loss": 1.0042,
       "step": 13675
     },
     {
-      "epoch": 0.37560077998407077,
+      "epoch": 0.3880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.4363322063002707e-05,
-      "loss": 0.9859,
+      "learning_rate": 1.3996212379308351e-05,
+      "loss": 0.9635,
       "step": 13676
     },
     {
-      "epoch": 0.375628244211914,
+      "epoch": 0.38811010215664016,
       "grad_norm": 0.0,
-      "learning_rate": 1.4362521672003002e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.3995369861624914e-05,
+      "loss": 0.8843,
       "step": 13677
     },
     {
-      "epoch": 0.3756557084397572,
+      "epoch": 0.3881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.4361721246485337e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.399452731019241e-05,
+      "loss": 0.9915,
       "step": 13678
     },
     {
-      "epoch": 0.37568317266760043,
+      "epoch": 0.38816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.4360920786456039e-05,
-      "loss": 0.8838,
+      "learning_rate": 1.3993684725017954e-05,
+      "loss": 0.9308,
       "step": 13679
     },
     {
-      "epoch": 0.3757106368954437,
+      "epoch": 0.38819523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.4360120291921448e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.399284210610866e-05,
+      "loss": 0.985,
       "step": 13680
     },
     {
-      "epoch": 0.3757381011232869,
+      "epoch": 0.38822360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 1.43593197628879e-05,
-      "loss": 1.0138,
+      "learning_rate": 1.3991999453471651e-05,
+      "loss": 0.9348,
       "step": 13681
     },
     {
-      "epoch": 0.37576556535113015,
+      "epoch": 0.3882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.4358519199361722e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.3991156767114044e-05,
+      "loss": 0.8997,
       "step": 13682
     },
     {
-      "epoch": 0.37579302957897337,
+      "epoch": 0.38828036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.4357718601349253e-05,
-      "loss": 0.8545,
+      "learning_rate": 1.3990314047042958e-05,
+      "loss": 1.0493,
       "step": 13683
     },
     {
-      "epoch": 0.37582049380681665,
+      "epoch": 0.38830874006810445,
       "grad_norm": 0.0,
-      "learning_rate": 1.4356917968856827e-05,
-      "loss": 1.0344,
+      "learning_rate": 1.3989471293265506e-05,
+      "loss": 0.906,
       "step": 13684
     },
     {
-      "epoch": 0.37584795803465987,
+      "epoch": 0.3883371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.4356117301890777e-05,
-      "loss": 0.987,
+      "learning_rate": 1.3988628505788814e-05,
+      "loss": 0.8797,
       "step": 13685
     },
     {
-      "epoch": 0.3758754222625031,
+      "epoch": 0.3883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.4355316600457438e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.398778568462e-05,
+      "loss": 0.9417,
       "step": 13686
     },
     {
-      "epoch": 0.3759028864903463,
+      "epoch": 0.3883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.4354515864563152e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.3986942829766175e-05,
+      "loss": 0.9067,
       "step": 13687
     },
     {
-      "epoch": 0.37593035071818953,
+      "epoch": 0.388422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.4353715094214244e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.3986099941234468e-05,
+      "loss": 0.9104,
       "step": 13688
     },
     {
-      "epoch": 0.3759578149460328,
+      "epoch": 0.38845062429057886,
       "grad_norm": 0.0,
-      "learning_rate": 1.435291428941706e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.3985257019031992e-05,
+      "loss": 0.9542,
       "step": 13689
     },
     {
-      "epoch": 0.37598527917387603,
+      "epoch": 0.3884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.435211345017793e-05,
-      "loss": 0.9468,
+      "learning_rate": 1.3984414063165874e-05,
+      "loss": 1.0767,
       "step": 13690
     },
     {
-      "epoch": 0.37601274340171925,
+      "epoch": 0.3885073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.4351312576503191e-05,
-      "loss": 0.9283,
+      "learning_rate": 1.3983571073643232e-05,
+      "loss": 0.7938,
       "step": 13691
     },
     {
-      "epoch": 0.3760402076295625,
+      "epoch": 0.38853575482406355,
       "grad_norm": 0.0,
-      "learning_rate": 1.4350511668399183e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.3982728050471184e-05,
+      "loss": 0.9866,
       "step": 13692
     },
     {
-      "epoch": 0.37606767185740575,
+      "epoch": 0.38856413166855847,
       "grad_norm": 0.0,
-      "learning_rate": 1.4349710725872243e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.3981884993656853e-05,
+      "loss": 0.9522,
       "step": 13693
     },
     {
-      "epoch": 0.37609513608524897,
+      "epoch": 0.3885925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.4348909748928706e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.3981041903207364e-05,
+      "loss": 0.9173,
       "step": 13694
     },
     {
-      "epoch": 0.3761226003130922,
+      "epoch": 0.38862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.4348108737574909e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.3980198779129834e-05,
+      "loss": 0.9335,
       "step": 13695
     },
     {
-      "epoch": 0.3761500645409354,
+      "epoch": 0.38864926220204316,
       "grad_norm": 0.0,
-      "learning_rate": 1.4347307691817192e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.3979355621431382e-05,
+      "loss": 0.9019,
       "step": 13696
     },
     {
-      "epoch": 0.3761775287687787,
+      "epoch": 0.388677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.4346506611661894e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.3978512430119136e-05,
+      "loss": 0.934,
       "step": 13697
     },
     {
-      "epoch": 0.3762049929966219,
+      "epoch": 0.38870601589103293,
       "grad_norm": 0.0,
-      "learning_rate": 1.434570549711535e-05,
-      "loss": 1.0146,
+      "learning_rate": 1.397766920520022e-05,
+      "loss": 0.9029,
       "step": 13698
     },
     {
-      "epoch": 0.37623245722446513,
+      "epoch": 0.3887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.43449043481839e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.3976825946681748e-05,
+      "loss": 1.0693,
       "step": 13699
     },
     {
-      "epoch": 0.37625992145230835,
+      "epoch": 0.3887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.4344103164873886e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.3975982654570852e-05,
+      "loss": 0.8519,
       "step": 13700
     },
     {
-      "epoch": 0.3762873856801516,
+      "epoch": 0.3887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.4343301947191645e-05,
-      "loss": 1.0143,
+      "learning_rate": 1.3975139328874651e-05,
+      "loss": 0.9105,
       "step": 13701
     },
     {
-      "epoch": 0.37631484990799485,
+      "epoch": 0.3888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.4342500695143518e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.397429596960027e-05,
+      "loss": 0.9525,
       "step": 13702
     },
     {
-      "epoch": 0.3763423141358381,
+      "epoch": 0.3888479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.434169940873584e-05,
-      "loss": 0.936,
+      "learning_rate": 1.3973452576754833e-05,
+      "loss": 0.8674,
       "step": 13703
     },
     {
-      "epoch": 0.3763697783636813,
+      "epoch": 0.38887627695800225,
       "grad_norm": 0.0,
-      "learning_rate": 1.4340898087974954e-05,
-      "loss": 0.8735,
+      "learning_rate": 1.3972609150345462e-05,
+      "loss": 0.9175,
       "step": 13704
     },
     {
-      "epoch": 0.3763972425915245,
+      "epoch": 0.38890465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.4340096732867203e-05,
-      "loss": 0.8837,
+      "learning_rate": 1.3971765690379284e-05,
+      "loss": 0.901,
       "step": 13705
     },
     {
-      "epoch": 0.3764247068193678,
+      "epoch": 0.38893303064699203,
       "grad_norm": 0.0,
-      "learning_rate": 1.4339295343418927e-05,
-      "loss": 1.095,
+      "learning_rate": 1.397092219686342e-05,
+      "loss": 1.0338,
       "step": 13706
     },
     {
-      "epoch": 0.376452171047211,
+      "epoch": 0.38896140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.433849391963646e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.3970078669805001e-05,
+      "loss": 0.9414,
       "step": 13707
     },
     {
-      "epoch": 0.37647963527505424,
+      "epoch": 0.38898978433598186,
       "grad_norm": 0.0,
-      "learning_rate": 1.4337692461526154e-05,
-      "loss": 0.9162,
+      "learning_rate": 1.3969235109211149e-05,
+      "loss": 0.9027,
       "step": 13708
     },
     {
-      "epoch": 0.37650709950289746,
+      "epoch": 0.3890181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.4336890969094343e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.3968391515088987e-05,
+      "loss": 0.9025,
       "step": 13709
     },
     {
-      "epoch": 0.37653456373074073,
+      "epoch": 0.38904653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 1.433608944234737e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.3967547887445645e-05,
+      "loss": 1.0408,
       "step": 13710
     },
     {
-      "epoch": 0.37656202795858396,
+      "epoch": 0.3890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.433528788129158e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.3966704226288247e-05,
+      "loss": 0.9308,
       "step": 13711
     },
     {
-      "epoch": 0.3765894921864272,
+      "epoch": 0.3891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.4334486285933308e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.396586053162392e-05,
+      "loss": 0.8612,
       "step": 13712
     },
     {
-      "epoch": 0.3766169564142704,
+      "epoch": 0.3891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.4333684656278906e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.3965016803459792e-05,
+      "loss": 0.9675,
       "step": 13713
     },
     {
-      "epoch": 0.3766444206421136,
+      "epoch": 0.3891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.433288299233471e-05,
-      "loss": 0.8267,
+      "learning_rate": 1.3964173041802986e-05,
+      "loss": 0.952,
       "step": 13714
     },
     {
-      "epoch": 0.3766718848699569,
+      "epoch": 0.3891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.4332081294107067e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.3963329246660636e-05,
+      "loss": 0.9402,
       "step": 13715
     },
     {
-      "epoch": 0.3766993490978001,
+      "epoch": 0.38921679909194096,
       "grad_norm": 0.0,
-      "learning_rate": 1.4331279561602318e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.3962485418039865e-05,
+      "loss": 0.9987,
       "step": 13716
     },
     {
-      "epoch": 0.37672681332564334,
+      "epoch": 0.38924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.433047779482681e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.3961641555947797e-05,
+      "loss": 0.9807,
       "step": 13717
     },
     {
-      "epoch": 0.37675427755348656,
+      "epoch": 0.3892735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.432967599378688e-05,
-      "loss": 0.997,
+      "learning_rate": 1.396079766039157e-05,
+      "loss": 0.8969,
       "step": 13718
     },
     {
-      "epoch": 0.37678174178132984,
+      "epoch": 0.38930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.4328874158488877e-05,
-      "loss": 1.0509,
+      "learning_rate": 1.3959953731378305e-05,
+      "loss": 0.9177,
       "step": 13719
     },
     {
-      "epoch": 0.37680920600917306,
+      "epoch": 0.38933030646992056,
       "grad_norm": 0.0,
-      "learning_rate": 1.4328072288939145e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.3959109768915133e-05,
+      "loss": 0.9459,
       "step": 13720
     },
     {
-      "epoch": 0.3768366702370163,
+      "epoch": 0.3893586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.4327270385144032e-05,
-      "loss": 0.9071,
+      "learning_rate": 1.3958265773009184e-05,
+      "loss": 0.9168,
       "step": 13721
     },
     {
-      "epoch": 0.3768641344648595,
+      "epoch": 0.38938706015891034,
       "grad_norm": 0.0,
-      "learning_rate": 1.4326468447109877e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.3957421743667582e-05,
+      "loss": 1.0418,
       "step": 13722
     },
     {
-      "epoch": 0.3768915986927028,
+      "epoch": 0.3894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.4325666474843027e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.3956577680897465e-05,
+      "loss": 0.9589,
       "step": 13723
     },
     {
-      "epoch": 0.376919062920546,
+      "epoch": 0.3894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.4324864468349828e-05,
-      "loss": 0.834,
+      "learning_rate": 1.3955733584705957e-05,
+      "loss": 1.0278,
       "step": 13724
     },
     {
-      "epoch": 0.3769465271483892,
+      "epoch": 0.389472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.432406242763663e-05,
-      "loss": 0.8747,
+      "learning_rate": 1.3954889455100192e-05,
+      "loss": 1.0245,
       "step": 13725
     },
     {
-      "epoch": 0.37697399137623244,
+      "epoch": 0.3895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.4323260352709769e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.3954045292087294e-05,
+      "loss": 1.0445,
       "step": 13726
     },
     {
-      "epoch": 0.37700145560407566,
+      "epoch": 0.3895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.43224582435756e-05,
-      "loss": 0.9052,
+      "learning_rate": 1.39532010956744e-05,
+      "loss": 0.8757,
       "step": 13727
     },
     {
-      "epoch": 0.37702891983191894,
+      "epoch": 0.38955732122587966,
       "grad_norm": 0.0,
-      "learning_rate": 1.432165610024047e-05,
-      "loss": 0.8546,
+      "learning_rate": 1.3952356865868634e-05,
+      "loss": 0.7472,
       "step": 13728
     },
     {
-      "epoch": 0.37705638405976216,
+      "epoch": 0.3895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.432085392271072e-05,
-      "loss": 1.0389,
+      "learning_rate": 1.3951512602677137e-05,
+      "loss": 0.8171,
       "step": 13729
     },
     {
-      "epoch": 0.3770838482876054,
+      "epoch": 0.3896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.4320051710992698e-05,
-      "loss": 0.7949,
+      "learning_rate": 1.3950668306107035e-05,
+      "loss": 0.9558,
       "step": 13730
     },
     {
-      "epoch": 0.3771113125154486,
+      "epoch": 0.38964245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 1.4319249465092758e-05,
-      "loss": 0.8576,
+      "learning_rate": 1.3949823976165459e-05,
+      "loss": 0.99,
       "step": 13731
     },
     {
-      "epoch": 0.3771387767432919,
+      "epoch": 0.38967082860385927,
       "grad_norm": 0.0,
-      "learning_rate": 1.431844718501724e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.3948979612859543e-05,
+      "loss": 1.0927,
       "step": 13732
     },
     {
-      "epoch": 0.3771662409711351,
+      "epoch": 0.3896992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.4317644870772496e-05,
-      "loss": 0.9365,
+      "learning_rate": 1.3948135216196419e-05,
+      "loss": 0.9189,
       "step": 13733
     },
     {
-      "epoch": 0.3771937051989783,
+      "epoch": 0.38972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.4316842522364873e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.3947290786183219e-05,
+      "loss": 1.0226,
       "step": 13734
     },
     {
-      "epoch": 0.37722116942682155,
+      "epoch": 0.38975595913734395,
       "grad_norm": 0.0,
-      "learning_rate": 1.431604013980072e-05,
-      "loss": 1.0498,
+      "learning_rate": 1.3946446322827074e-05,
+      "loss": 0.8744,
       "step": 13735
     },
     {
-      "epoch": 0.3772486336546648,
+      "epoch": 0.3897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.4315237723086383e-05,
-      "loss": 1.0694,
+      "learning_rate": 1.3945601826135122e-05,
+      "loss": 0.9576,
       "step": 13736
     },
     {
-      "epoch": 0.37727609788250804,
+      "epoch": 0.38981271282633373,
       "grad_norm": 0.0,
-      "learning_rate": 1.431443527222822e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.3944757296114495e-05,
+      "loss": 0.9943,
       "step": 13737
     },
     {
-      "epoch": 0.37730356211035126,
+      "epoch": 0.3898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.4313632787232567e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.3943912732772324e-05,
+      "loss": 1.0063,
       "step": 13738
     },
     {
-      "epoch": 0.3773310263381945,
+      "epoch": 0.3898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.4312830268105784e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.3943068136115745e-05,
+      "loss": 1.0759,
       "step": 13739
     },
     {
-      "epoch": 0.37735849056603776,
+      "epoch": 0.38989784335981836,
       "grad_norm": 0.0,
-      "learning_rate": 1.4312027714854218e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.3942223506151895e-05,
+      "loss": 1.0068,
       "step": 13740
     },
     {
-      "epoch": 0.377385954793881,
+      "epoch": 0.3899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.4311225127484216e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.39413788428879e-05,
+      "loss": 0.8505,
       "step": 13741
     },
     {
-      "epoch": 0.3774134190217242,
+      "epoch": 0.3899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.431042250600213e-05,
-      "loss": 0.8992,
+      "learning_rate": 1.3940534146330906e-05,
+      "loss": 0.8435,
       "step": 13742
     },
     {
-      "epoch": 0.3774408832495674,
+      "epoch": 0.38998297389330305,
       "grad_norm": 0.0,
-      "learning_rate": 1.4309619850414316e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.393968941648804e-05,
+      "loss": 0.9496,
       "step": 13743
     },
     {
-      "epoch": 0.37746834747741065,
+      "epoch": 0.39001135073779797,
       "grad_norm": 0.0,
-      "learning_rate": 1.4308817160727115e-05,
-      "loss": 0.8632,
+      "learning_rate": 1.393884465336644e-05,
+      "loss": 0.973,
       "step": 13744
     },
     {
-      "epoch": 0.3774958117052539,
+      "epoch": 0.39003972758229283,
       "grad_norm": 0.0,
-      "learning_rate": 1.4308014436946888e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.3937999856973243e-05,
+      "loss": 0.9851,
       "step": 13745
     },
     {
-      "epoch": 0.37752327593309715,
+      "epoch": 0.39006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.4307211679079979e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.3937155027315586e-05,
+      "loss": 0.9157,
       "step": 13746
     },
     {
-      "epoch": 0.37755074016094037,
+      "epoch": 0.39009648127128266,
       "grad_norm": 0.0,
-      "learning_rate": 1.4306408887132745e-05,
-      "loss": 0.968,
+      "learning_rate": 1.39363101644006e-05,
+      "loss": 0.9643,
       "step": 13747
     },
     {
-      "epoch": 0.3775782043887836,
+      "epoch": 0.3901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.4305606061111532e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.3935465268235429e-05,
+      "loss": 0.8305,
       "step": 13748
     },
     {
-      "epoch": 0.37760566861662687,
+      "epoch": 0.39015323496027243,
       "grad_norm": 0.0,
-      "learning_rate": 1.4304803201022703e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.3934620338827204e-05,
+      "loss": 0.8918,
       "step": 13749
     },
     {
-      "epoch": 0.3776331328444701,
+      "epoch": 0.3901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.4304000306872597e-05,
-      "loss": 0.9978,
+      "learning_rate": 1.3933775376183065e-05,
+      "loss": 0.8395,
       "step": 13750
     },
     {
-      "epoch": 0.3776605970723133,
+      "epoch": 0.3902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.4303197378667576e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.3932930380310145e-05,
+      "loss": 0.9103,
       "step": 13751
     },
     {
-      "epoch": 0.37768806130015653,
+      "epoch": 0.39023836549375707,
       "grad_norm": 0.0,
-      "learning_rate": 1.430239441641399e-05,
-      "loss": 1.0347,
+      "learning_rate": 1.393208535121559e-05,
+      "loss": 0.9324,
       "step": 13752
     },
     {
-      "epoch": 0.3777155255279998,
+      "epoch": 0.390266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.4301591420118195e-05,
-      "loss": 0.97,
+      "learning_rate": 1.3931240288906534e-05,
+      "loss": 0.9179,
       "step": 13753
     },
     {
-      "epoch": 0.37774298975584303,
+      "epoch": 0.3902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.4300788389786537e-05,
-      "loss": 0.9953,
+      "learning_rate": 1.393039519339011e-05,
+      "loss": 0.8386,
       "step": 13754
     },
     {
-      "epoch": 0.37777045398368625,
+      "epoch": 0.39032349602724176,
       "grad_norm": 0.0,
-      "learning_rate": 1.429998532542538e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.3929550064673462e-05,
+      "loss": 0.9775,
       "step": 13755
     },
     {
-      "epoch": 0.37779791821152947,
+      "epoch": 0.39035187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.429918222704107e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.3928704902763733e-05,
+      "loss": 0.9537,
       "step": 13756
     },
     {
-      "epoch": 0.3778253824393727,
+      "epoch": 0.39038024971623153,
       "grad_norm": 0.0,
-      "learning_rate": 1.429837909463997e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.392785970766805e-05,
+      "loss": 0.9454,
       "step": 13757
     },
     {
-      "epoch": 0.37785284666721597,
+      "epoch": 0.39040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4297575928228426e-05,
-      "loss": 1.0751,
+      "learning_rate": 1.392701447939356e-05,
+      "loss": 0.8869,
       "step": 13758
     },
     {
-      "epoch": 0.3778803108950592,
+      "epoch": 0.39043700340522136,
       "grad_norm": 0.0,
-      "learning_rate": 1.4296772727812796e-05,
-      "loss": 0.8522,
+      "learning_rate": 1.3926169217947407e-05,
+      "loss": 0.8387,
       "step": 13759
     },
     {
-      "epoch": 0.3779077751229024,
+      "epoch": 0.3904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.4295969493399435e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.3925323923336724e-05,
+      "loss": 0.9605,
       "step": 13760
     },
     {
-      "epoch": 0.37793523935074563,
+      "epoch": 0.39049375709421114,
       "grad_norm": 0.0,
-      "learning_rate": 1.4295166224994702e-05,
-      "loss": 0.8779,
+      "learning_rate": 1.3924478595568654e-05,
+      "loss": 0.964,
       "step": 13761
     },
     {
-      "epoch": 0.3779627035785889,
+      "epoch": 0.390522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.429436292260495e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.3923633234650336e-05,
+      "loss": 0.899,
       "step": 13762
     },
     {
-      "epoch": 0.37799016780643213,
+      "epoch": 0.3905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.4293559586236532e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.3922787840588913e-05,
+      "loss": 0.9016,
       "step": 13763
     },
     {
-      "epoch": 0.37801763203427535,
+      "epoch": 0.3905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.4292756215895811e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.3921942413391522e-05,
+      "loss": 0.9705,
       "step": 13764
     },
     {
-      "epoch": 0.3780450962621186,
+      "epoch": 0.3906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.429195281158914e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.392109695306531e-05,
+      "loss": 0.9413,
       "step": 13765
     },
     {
-      "epoch": 0.37807256048996185,
+      "epoch": 0.3906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.4291149373322872e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.3920251459617413e-05,
+      "loss": 0.9825,
       "step": 13766
     },
     {
-      "epoch": 0.37810002471780507,
+      "epoch": 0.39066401816118046,
       "grad_norm": 0.0,
-      "learning_rate": 1.4290345901103372e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.3919405933054977e-05,
+      "loss": 0.8273,
       "step": 13767
     },
     {
-      "epoch": 0.3781274889456483,
+      "epoch": 0.3906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.428954239493699e-05,
-      "loss": 1.0238,
+      "learning_rate": 1.3918560373385144e-05,
+      "loss": 0.8929,
       "step": 13768
     },
     {
-      "epoch": 0.3781549531734915,
+      "epoch": 0.39072077185017023,
       "grad_norm": 0.0,
-      "learning_rate": 1.4288738854830089e-05,
-      "loss": 1.038,
+      "learning_rate": 1.3917714780615055e-05,
+      "loss": 1.014,
       "step": 13769
     },
     {
-      "epoch": 0.37818241740133474,
+      "epoch": 0.39074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.4287935280789025e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.3916869154751853e-05,
+      "loss": 0.8862,
       "step": 13770
     },
     {
-      "epoch": 0.378209881629178,
+      "epoch": 0.39077752553916006,
       "grad_norm": 0.0,
-      "learning_rate": 1.4287131672820156e-05,
-      "loss": 0.9523,
+      "learning_rate": 1.391602349580268e-05,
+      "loss": 0.84,
       "step": 13771
     },
     {
-      "epoch": 0.37823734585702123,
+      "epoch": 0.3908059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.4286328030929837e-05,
-      "loss": 0.8809,
+      "learning_rate": 1.3915177803774682e-05,
+      "loss": 0.9004,
       "step": 13772
     },
     {
-      "epoch": 0.37826481008486446,
+      "epoch": 0.39083427922814984,
       "grad_norm": 0.0,
-      "learning_rate": 1.4285524355124436e-05,
-      "loss": 1.0651,
+      "learning_rate": 1.3914332078675e-05,
+      "loss": 0.8896,
       "step": 13773
     },
     {
-      "epoch": 0.3782922743127077,
+      "epoch": 0.3908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.4284720645410304e-05,
-      "loss": 0.961,
+      "learning_rate": 1.3913486320510779e-05,
+      "loss": 0.8788,
       "step": 13774
     },
     {
-      "epoch": 0.37831973854055095,
+      "epoch": 0.3908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.42839169017938e-05,
-      "loss": 0.9016,
+      "learning_rate": 1.3912640529289163e-05,
+      "loss": 0.8906,
       "step": 13775
     },
     {
-      "epoch": 0.3783472027683942,
+      "epoch": 0.39091940976163453,
       "grad_norm": 0.0,
-      "learning_rate": 1.428311312428129e-05,
-      "loss": 0.8303,
+      "learning_rate": 1.3911794705017299e-05,
+      "loss": 0.9658,
       "step": 13776
     },
     {
-      "epoch": 0.3783746669962374,
+      "epoch": 0.3909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.4282309312879127e-05,
-      "loss": 1.0355,
+      "learning_rate": 1.3910948847702326e-05,
+      "loss": 0.9262,
       "step": 13777
     },
     {
-      "epoch": 0.3784021312240806,
+      "epoch": 0.3909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.4281505467593672e-05,
-      "loss": 0.893,
+      "learning_rate": 1.3910102957351392e-05,
+      "loss": 0.952,
       "step": 13778
     },
     {
-      "epoch": 0.3784295954519239,
+      "epoch": 0.39100454029511916,
       "grad_norm": 0.0,
-      "learning_rate": 1.4280701588431295e-05,
-      "loss": 0.8148,
+      "learning_rate": 1.3909257033971645e-05,
+      "loss": 0.886,
       "step": 13779
     },
     {
-      "epoch": 0.3784570596797671,
+      "epoch": 0.3910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.427989767539834e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.390841107757023e-05,
+      "loss": 0.9382,
       "step": 13780
     },
     {
-      "epoch": 0.37848452390761034,
+      "epoch": 0.391061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.427909372850118e-05,
-      "loss": 1.0248,
+      "learning_rate": 1.3907565088154286e-05,
+      "loss": 0.9,
       "step": 13781
     },
     {
-      "epoch": 0.37851198813545356,
+      "epoch": 0.39108967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 1.4278289747746176e-05,
-      "loss": 0.9009,
+      "learning_rate": 1.3906719065730965e-05,
+      "loss": 0.9467,
       "step": 13782
     },
     {
-      "epoch": 0.3785394523632968,
+      "epoch": 0.39111804767309877,
       "grad_norm": 0.0,
-      "learning_rate": 1.4277485733139682e-05,
-      "loss": 1.0137,
+      "learning_rate": 1.3905873010307413e-05,
+      "loss": 0.8146,
       "step": 13783
     },
     {
-      "epoch": 0.37856691659114006,
+      "epoch": 0.3911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.4276681684688068e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.390502692189078e-05,
+      "loss": 0.9371,
       "step": 13784
     },
     {
-      "epoch": 0.3785943808189833,
+      "epoch": 0.39117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.427587760239769e-05,
-      "loss": 0.9032,
+      "learning_rate": 1.3904180800488202e-05,
+      "loss": 0.8558,
       "step": 13785
     },
     {
-      "epoch": 0.3786218450468265,
+      "epoch": 0.3912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.4275073486274911e-05,
-      "loss": 0.9359,
+      "learning_rate": 1.3903334646106838e-05,
+      "loss": 0.9452,
       "step": 13786
     },
     {
-      "epoch": 0.3786493092746697,
+      "epoch": 0.3912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.4274269336326094e-05,
-      "loss": 0.9511,
+      "learning_rate": 1.3902488458753834e-05,
+      "loss": 0.9286,
       "step": 13787
     },
     {
-      "epoch": 0.378676773502513,
+      "epoch": 0.39125993189557323,
       "grad_norm": 0.0,
-      "learning_rate": 1.4273465152557604e-05,
-      "loss": 0.9593,
+      "learning_rate": 1.3901642238436326e-05,
+      "loss": 0.9285,
       "step": 13788
     },
     {
-      "epoch": 0.3787042377303562,
+      "epoch": 0.3912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.4272660934975801e-05,
-      "loss": 0.9851,
+      "learning_rate": 1.3900795985161475e-05,
+      "loss": 0.9262,
       "step": 13789
     },
     {
-      "epoch": 0.37873170195819944,
+      "epoch": 0.391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.4271856683587053e-05,
-      "loss": 1.0348,
+      "learning_rate": 1.3899949698936426e-05,
+      "loss": 0.9041,
       "step": 13790
     },
     {
-      "epoch": 0.37875916618604266,
+      "epoch": 0.39134506242905787,
       "grad_norm": 0.0,
-      "learning_rate": 1.4271052398397715e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.3899103379768327e-05,
+      "loss": 0.9006,
       "step": 13791
     },
     {
-      "epoch": 0.37878663041388594,
+      "epoch": 0.3913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.4270248079414159e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.3898257027664324e-05,
+      "loss": 0.8718,
       "step": 13792
     },
     {
-      "epoch": 0.37881409464172916,
+      "epoch": 0.3914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.4269443726642744e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.389741064263157e-05,
+      "loss": 0.9016,
       "step": 13793
     },
     {
-      "epoch": 0.3788415588695724,
+      "epoch": 0.39143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.426863934008984e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.3896564224677214e-05,
+      "loss": 0.9479,
       "step": 13794
     },
     {
-      "epoch": 0.3788690230974156,
+      "epoch": 0.39145856980703747,
       "grad_norm": 0.0,
-      "learning_rate": 1.4267834919761805e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.3895717773808402e-05,
+      "loss": 1.0021,
       "step": 13795
     },
     {
-      "epoch": 0.3788964873252588,
+      "epoch": 0.39148694665153233,
       "grad_norm": 0.0,
-      "learning_rate": 1.4267030465665009e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.3894871290032286e-05,
+      "loss": 0.8148,
       "step": 13796
     },
     {
-      "epoch": 0.3789239515531021,
+      "epoch": 0.39151532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 1.4266225977805816e-05,
-      "loss": 0.9845,
+      "learning_rate": 1.3894024773356021e-05,
+      "loss": 0.9361,
       "step": 13797
     },
     {
-      "epoch": 0.3789514157809453,
+      "epoch": 0.39154370034052216,
       "grad_norm": 0.0,
-      "learning_rate": 1.4265421456190586e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.3893178223786751e-05,
+      "loss": 1.0211,
       "step": 13798
     },
     {
-      "epoch": 0.37897888000878854,
+      "epoch": 0.391572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.4264616900825692e-05,
-      "loss": 0.9828,
+      "learning_rate": 1.3892331641331632e-05,
+      "loss": 0.9157,
       "step": 13799
     },
     {
-      "epoch": 0.37900634423663176,
+      "epoch": 0.39160045402951194,
       "grad_norm": 0.0,
-      "learning_rate": 1.4263812311717498e-05,
-      "loss": 1.015,
+      "learning_rate": 1.3891485025997813e-05,
+      "loss": 0.8831,
       "step": 13800
     },
     {
-      "epoch": 0.37903380846447504,
+      "epoch": 0.3916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.4263007688872368e-05,
-      "loss": 0.9238,
+      "learning_rate": 1.3890638377792442e-05,
+      "loss": 0.907,
       "step": 13801
     },
     {
-      "epoch": 0.37906127269231826,
+      "epoch": 0.3916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.426220303229667e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.3889791696722676e-05,
+      "loss": 0.9561,
       "step": 13802
     },
     {
-      "epoch": 0.3790887369201615,
+      "epoch": 0.39168558456299657,
       "grad_norm": 0.0,
-      "learning_rate": 1.426139834199677e-05,
-      "loss": 1.0266,
+      "learning_rate": 1.3888944982795665e-05,
+      "loss": 0.9609,
       "step": 13803
     },
     {
-      "epoch": 0.3791162011480047,
+      "epoch": 0.3917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.4260593617979036e-05,
-      "loss": 1.0382,
+      "learning_rate": 1.388809823601856e-05,
+      "loss": 1.0447,
       "step": 13804
     },
     {
-      "epoch": 0.379143665375848,
+      "epoch": 0.3917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.4259788860249837e-05,
-      "loss": 1.02,
+      "learning_rate": 1.3887251456398512e-05,
+      "loss": 1.0009,
       "step": 13805
     },
     {
-      "epoch": 0.3791711296036912,
+      "epoch": 0.39177071509648126,
       "grad_norm": 0.0,
-      "learning_rate": 1.4258984068815536e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.3886404643942679e-05,
+      "loss": 0.9053,
       "step": 13806
     },
     {
-      "epoch": 0.3791985938315344,
+      "epoch": 0.3917990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.4258179243682501e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.3885557798658212e-05,
+      "loss": 0.8792,
       "step": 13807
     },
     {
-      "epoch": 0.37922605805937765,
+      "epoch": 0.39182746878547103,
       "grad_norm": 0.0,
-      "learning_rate": 1.4257374384857108e-05,
-      "loss": 0.9239,
+      "learning_rate": 1.388471092055226e-05,
+      "loss": 0.971,
       "step": 13808
     },
     {
-      "epoch": 0.37925352228722087,
+      "epoch": 0.39185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.4256569492345717e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.3883864009631984e-05,
+      "loss": 0.9134,
       "step": 13809
     },
     {
-      "epoch": 0.37928098651506414,
+      "epoch": 0.39188422247446086,
       "grad_norm": 0.0,
-      "learning_rate": 1.4255764566154699e-05,
-      "loss": 0.8871,
+      "learning_rate": 1.3883017065904534e-05,
+      "loss": 0.9606,
       "step": 13810
     },
     {
-      "epoch": 0.37930845074290737,
+      "epoch": 0.3919125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.4254959606290423e-05,
-      "loss": 0.9358,
+      "learning_rate": 1.388217008937706e-05,
+      "loss": 1.1116,
       "step": 13811
     },
     {
-      "epoch": 0.3793359149707506,
+      "epoch": 0.39194097616345064,
       "grad_norm": 0.0,
-      "learning_rate": 1.4254154612759259e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.3881323080056723e-05,
+      "loss": 0.8625,
       "step": 13812
     },
     {
-      "epoch": 0.3793633791985938,
+      "epoch": 0.3919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.4253349585567575e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.3880476037950676e-05,
+      "loss": 0.9318,
       "step": 13813
     },
     {
-      "epoch": 0.3793908434264371,
+      "epoch": 0.3919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.4252544524721744e-05,
-      "loss": 0.9282,
+      "learning_rate": 1.3879628963066076e-05,
+      "loss": 1.0018,
       "step": 13814
     },
     {
-      "epoch": 0.3794183076542803,
+      "epoch": 0.39202610669693533,
       "grad_norm": 0.0,
-      "learning_rate": 1.425173943022813e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.3878781855410072e-05,
+      "loss": 0.8717,
       "step": 13815
     },
     {
-      "epoch": 0.37944577188212353,
+      "epoch": 0.3920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.4250934302093107e-05,
-      "loss": 0.961,
+      "learning_rate": 1.3877934714989824e-05,
+      "loss": 0.8972,
       "step": 13816
     },
     {
-      "epoch": 0.37947323610996675,
+      "epoch": 0.3920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.425012914032305e-05,
-      "loss": 0.8835,
+      "learning_rate": 1.3877087541812491e-05,
+      "loss": 0.9429,
       "step": 13817
     },
     {
-      "epoch": 0.37950070033781,
+      "epoch": 0.39211123723041996,
       "grad_norm": 0.0,
-      "learning_rate": 1.424932394492432e-05,
-      "loss": 0.9989,
+      "learning_rate": 1.387624033588522e-05,
+      "loss": 0.9256,
       "step": 13818
     },
     {
-      "epoch": 0.37952816456565325,
+      "epoch": 0.3921396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.4248518715903292e-05,
-      "loss": 0.8503,
+      "learning_rate": 1.3875393097215173e-05,
+      "loss": 0.9634,
       "step": 13819
     },
     {
-      "epoch": 0.37955562879349647,
+      "epoch": 0.39216799091940974,
       "grad_norm": 0.0,
-      "learning_rate": 1.4247713453266342e-05,
-      "loss": 0.8767,
+      "learning_rate": 1.3874545825809509e-05,
+      "loss": 0.8824,
       "step": 13820
     },
     {
-      "epoch": 0.3795830930213397,
+      "epoch": 0.39219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.4246908157019836e-05,
-      "loss": 0.948,
+      "learning_rate": 1.3873698521675383e-05,
+      "loss": 0.8734,
       "step": 13821
     },
     {
-      "epoch": 0.3796105572491829,
+      "epoch": 0.39222474460839957,
       "grad_norm": 0.0,
-      "learning_rate": 1.4246102827170148e-05,
-      "loss": 0.901,
+      "learning_rate": 1.3872851184819948e-05,
+      "loss": 0.9262,
       "step": 13822
     },
     {
-      "epoch": 0.3796380214770262,
+      "epoch": 0.3922531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.4245297463723651e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.3872003815250367e-05,
+      "loss": 0.8997,
       "step": 13823
     },
     {
-      "epoch": 0.3796654857048694,
+      "epoch": 0.39228149829738934,
       "grad_norm": 0.0,
-      "learning_rate": 1.4244492066686713e-05,
-      "loss": 0.8845,
+      "learning_rate": 1.3871156412973797e-05,
+      "loss": 0.971,
       "step": 13824
     },
     {
-      "epoch": 0.37969294993271263,
+      "epoch": 0.3923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.424368663606571e-05,
-      "loss": 0.942,
+      "learning_rate": 1.387030897799739e-05,
+      "loss": 0.9329,
       "step": 13825
     },
     {
-      "epoch": 0.37972041416055585,
+      "epoch": 0.3923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.4242881171867019e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.3869461510328314e-05,
+      "loss": 0.87,
       "step": 13826
     },
     {
-      "epoch": 0.37974787838839913,
+      "epoch": 0.39236662883087403,
       "grad_norm": 0.0,
-      "learning_rate": 1.4242075674097002e-05,
-      "loss": 1.073,
+      "learning_rate": 1.386861400997372e-05,
+      "loss": 0.9349,
       "step": 13827
     },
     {
-      "epoch": 0.37977534261624235,
+      "epoch": 0.3923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.4241270142762043e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.3867766476940771e-05,
+      "loss": 0.823,
       "step": 13828
     },
     {
-      "epoch": 0.37980280684408557,
+      "epoch": 0.3924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.4240464577868513e-05,
-      "loss": 0.9942,
+      "learning_rate": 1.3866918911236624e-05,
+      "loss": 1.0377,
       "step": 13829
     },
     {
-      "epoch": 0.3798302710719288,
+      "epoch": 0.39245175936435867,
       "grad_norm": 0.0,
-      "learning_rate": 1.4239658979422781e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.386607131286844e-05,
+      "loss": 0.9067,
       "step": 13830
     },
     {
-      "epoch": 0.37985773529977207,
+      "epoch": 0.3924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.4238853347431226e-05,
-      "loss": 0.8495,
+      "learning_rate": 1.3865223681843376e-05,
+      "loss": 0.9372,
       "step": 13831
     },
     {
-      "epoch": 0.3798851995276153,
+      "epoch": 0.39250851305334844,
       "grad_norm": 0.0,
-      "learning_rate": 1.4238047681900222e-05,
-      "loss": 1.0511,
+      "learning_rate": 1.3864376018168594e-05,
+      "loss": 0.8533,
       "step": 13832
     },
     {
-      "epoch": 0.3799126637554585,
+      "epoch": 0.39253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.4237241982836143e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.3863528321851256e-05,
+      "loss": 0.9862,
       "step": 13833
     },
     {
-      "epoch": 0.37994012798330173,
+      "epoch": 0.39256526674233827,
       "grad_norm": 0.0,
-      "learning_rate": 1.4236436250245362e-05,
-      "loss": 0.9165,
+      "learning_rate": 1.386268059289852e-05,
+      "loss": 0.9292,
       "step": 13834
     },
     {
-      "epoch": 0.379967592211145,
+      "epoch": 0.39259364358683313,
       "grad_norm": 0.0,
-      "learning_rate": 1.423563048413426e-05,
-      "loss": 1.0674,
+      "learning_rate": 1.3861832831317546e-05,
+      "loss": 0.8841,
       "step": 13835
     },
     {
-      "epoch": 0.37999505643898823,
+      "epoch": 0.39262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.4234824684509206e-05,
-      "loss": 0.9942,
+      "learning_rate": 1.3860985037115497e-05,
+      "loss": 0.9513,
       "step": 13836
     },
     {
-      "epoch": 0.38002252066683145,
+      "epoch": 0.3926503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.423401885137658e-05,
-      "loss": 0.8732,
+      "learning_rate": 1.3860137210299533e-05,
+      "loss": 1.0537,
       "step": 13837
     },
     {
-      "epoch": 0.3800499848946747,
+      "epoch": 0.3926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.4233212984742756e-05,
-      "loss": 0.9581,
+      "learning_rate": 1.3859289350876822e-05,
+      "loss": 0.9168,
       "step": 13838
     },
     {
-      "epoch": 0.3800774491225179,
+      "epoch": 0.39270715096481273,
       "grad_norm": 0.0,
-      "learning_rate": 1.4232407084614113e-05,
-      "loss": 1.0465,
+      "learning_rate": 1.3858441458854512e-05,
+      "loss": 1.0008,
       "step": 13839
     },
     {
-      "epoch": 0.3801049133503612,
+      "epoch": 0.3927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.4231601150997023e-05,
-      "loss": 0.9592,
+      "learning_rate": 1.3857593534239779e-05,
+      "loss": 1.0131,
       "step": 13840
     },
     {
-      "epoch": 0.3801323775782044,
+      "epoch": 0.3927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.423079518389787e-05,
-      "loss": 0.9101,
+      "learning_rate": 1.3856745577039781e-05,
+      "loss": 0.9344,
       "step": 13841
     },
     {
-      "epoch": 0.3801598418060476,
+      "epoch": 0.39279228149829737,
       "grad_norm": 0.0,
-      "learning_rate": 1.4229989183323021e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.3855897587261673e-05,
+      "loss": 0.9395,
       "step": 13842
     },
     {
-      "epoch": 0.38018730603389084,
+      "epoch": 0.3928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.4229183149278862e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.385504956491263e-05,
+      "loss": 0.9003,
       "step": 13843
     },
     {
-      "epoch": 0.3802147702617341,
+      "epoch": 0.3928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.4228377081771767e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.385420150999981e-05,
+      "loss": 1.0042,
       "step": 13844
     },
     {
-      "epoch": 0.38024223448957734,
+      "epoch": 0.39287741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.4227570980808117e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.3853353422530375e-05,
+      "loss": 0.8443,
       "step": 13845
     },
     {
-      "epoch": 0.38026969871742056,
+      "epoch": 0.392905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.4226764846394285e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.3852505302511492e-05,
+      "loss": 0.869,
       "step": 13846
     },
     {
-      "epoch": 0.3802971629452638,
+      "epoch": 0.39293416572077183,
       "grad_norm": 0.0,
-      "learning_rate": 1.4225958678536653e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.3851657149950325e-05,
+      "loss": 0.8067,
       "step": 13847
     },
     {
-      "epoch": 0.38032462717310705,
+      "epoch": 0.39296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.4225152477241598e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.3850808964854032e-05,
+      "loss": 1.0316,
       "step": 13848
     },
     {
-      "epoch": 0.3803520914009503,
+      "epoch": 0.3929909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.42243462425155e-05,
-      "loss": 1.0463,
+      "learning_rate": 1.3849960747229784e-05,
+      "loss": 0.9077,
       "step": 13849
     },
     {
-      "epoch": 0.3803795556287935,
+      "epoch": 0.3930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.4223539974364739e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.3849112497084747e-05,
+      "loss": 0.953,
       "step": 13850
     },
     {
-      "epoch": 0.3804070198566367,
+      "epoch": 0.39304767309875144,
       "grad_norm": 0.0,
-      "learning_rate": 1.4222733672795694e-05,
-      "loss": 0.8998,
+      "learning_rate": 1.3848264214426081e-05,
+      "loss": 0.8128,
       "step": 13851
     },
     {
-      "epoch": 0.38043448408447994,
+      "epoch": 0.3930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.4221927337814745e-05,
-      "loss": 0.8714,
+      "learning_rate": 1.3847415899260955e-05,
+      "loss": 1.0235,
       "step": 13852
     },
     {
-      "epoch": 0.3804619483123232,
+      "epoch": 0.3931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.4221120969428271e-05,
-      "loss": 0.941,
+      "learning_rate": 1.3846567551596534e-05,
+      "loss": 0.8986,
       "step": 13853
     },
     {
-      "epoch": 0.38048941254016644,
+      "epoch": 0.39313280363223607,
       "grad_norm": 0.0,
-      "learning_rate": 1.4220314567642652e-05,
-      "loss": 0.9357,
+      "learning_rate": 1.3845719171439986e-05,
+      "loss": 0.8523,
       "step": 13854
     },
     {
-      "epoch": 0.38051687676800966,
+      "epoch": 0.393161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.4219508132464271e-05,
-      "loss": 0.9782,
+      "learning_rate": 1.3844870758798474e-05,
+      "loss": 0.9521,
       "step": 13855
     },
     {
-      "epoch": 0.3805443409958529,
+      "epoch": 0.3931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.4218701663899504e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.3844022313679167e-05,
+      "loss": 0.8939,
       "step": 13856
     },
     {
-      "epoch": 0.38057180522369616,
+      "epoch": 0.39321793416572076,
       "grad_norm": 0.0,
-      "learning_rate": 1.4217895161954739e-05,
-      "loss": 0.9285,
+      "learning_rate": 1.3843173836089228e-05,
+      "loss": 0.9526,
       "step": 13857
     },
     {
-      "epoch": 0.3805992694515394,
+      "epoch": 0.3932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.421708862663635e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.3842325326035829e-05,
+      "loss": 0.8502,
       "step": 13858
     },
     {
-      "epoch": 0.3806267336793826,
+      "epoch": 0.39327468785471054,
       "grad_norm": 0.0,
-      "learning_rate": 1.4216282057950724e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.384147678352613e-05,
+      "loss": 0.9346,
       "step": 13859
     },
     {
-      "epoch": 0.3806541979072258,
+      "epoch": 0.39330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.421547545590424e-05,
-      "loss": 0.902,
+      "learning_rate": 1.384062820856731e-05,
+      "loss": 1.1395,
       "step": 13860
     },
     {
-      "epoch": 0.3806816621350691,
+      "epoch": 0.39333144154370037,
       "grad_norm": 0.0,
-      "learning_rate": 1.4214668820503283e-05,
-      "loss": 0.941,
+      "learning_rate": 1.3839779601166532e-05,
+      "loss": 0.8685,
       "step": 13861
     },
     {
-      "epoch": 0.3807091263629123,
+      "epoch": 0.3933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.4213862151754234e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.383893096133096e-05,
+      "loss": 1.0161,
       "step": 13862
     },
     {
-      "epoch": 0.38073659059075554,
+      "epoch": 0.39338819523269014,
       "grad_norm": 0.0,
-      "learning_rate": 1.4213055449663472e-05,
-      "loss": 0.9956,
+      "learning_rate": 1.3838082289067765e-05,
+      "loss": 0.7704,
       "step": 13863
     },
     {
-      "epoch": 0.38076405481859876,
+      "epoch": 0.393416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.4212248714237389e-05,
-      "loss": 0.801,
+      "learning_rate": 1.3837233584384118e-05,
+      "loss": 0.8249,
       "step": 13864
     },
     {
-      "epoch": 0.380791519046442,
+      "epoch": 0.3934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.4211441945482358e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.3836384847287186e-05,
+      "loss": 0.9061,
       "step": 13865
     },
     {
-      "epoch": 0.38081898327428526,
+      "epoch": 0.3934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.4210635143404767e-05,
-      "loss": 0.995,
+      "learning_rate": 1.3835536077784135e-05,
+      "loss": 0.8668,
       "step": 13866
     },
     {
-      "epoch": 0.3808464475021285,
+      "epoch": 0.3935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.4209828308011003e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.3834687275882144e-05,
+      "loss": 0.994,
       "step": 13867
     },
     {
-      "epoch": 0.3808739117299717,
+      "epoch": 0.3935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.4209021439307443e-05,
-      "loss": 0.955,
+      "learning_rate": 1.3833838441588374e-05,
+      "loss": 0.8522,
       "step": 13868
     },
     {
-      "epoch": 0.3809013759578149,
+      "epoch": 0.39355845629965946,
       "grad_norm": 0.0,
-      "learning_rate": 1.4208214537300478e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.3832989574909999e-05,
+      "loss": 1.0071,
       "step": 13869
     },
     {
-      "epoch": 0.3809288401856582,
+      "epoch": 0.3935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.4207407601996488e-05,
-      "loss": 1.0207,
+      "learning_rate": 1.3832140675854189e-05,
+      "loss": 0.9086,
       "step": 13870
     },
     {
-      "epoch": 0.3809563044135014,
+      "epoch": 0.39361520998864924,
       "grad_norm": 0.0,
-      "learning_rate": 1.4206600633401858e-05,
-      "loss": 0.8659,
+      "learning_rate": 1.3831291744428117e-05,
+      "loss": 0.9227,
       "step": 13871
     },
     {
-      "epoch": 0.38098376864134464,
+      "epoch": 0.39364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.4205793631522976e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.3830442780638946e-05,
+      "loss": 1.0189,
       "step": 13872
     },
     {
-      "epoch": 0.38101123286918787,
+      "epoch": 0.39367196367763907,
       "grad_norm": 0.0,
-      "learning_rate": 1.4204986596366225e-05,
-      "loss": 0.9385,
+      "learning_rate": 1.3829593784493855e-05,
+      "loss": 0.9517,
       "step": 13873
     },
     {
-      "epoch": 0.38103869709703114,
+      "epoch": 0.39370034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.4204179527937994e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.3828744756000015e-05,
+      "loss": 0.8601,
       "step": 13874
     },
     {
-      "epoch": 0.38106616132487436,
+      "epoch": 0.39372871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.4203372426244662e-05,
-      "loss": 0.997,
+      "learning_rate": 1.3827895695164595e-05,
+      "loss": 0.7895,
       "step": 13875
     },
     {
-      "epoch": 0.3810936255527176,
+      "epoch": 0.3937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.420256529129262e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.3827046601994765e-05,
+      "loss": 1.0492,
       "step": 13876
     },
     {
-      "epoch": 0.3811210897805608,
+      "epoch": 0.3937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.4201758123088257e-05,
-      "loss": 0.8051,
+      "learning_rate": 1.3826197476497705e-05,
+      "loss": 0.8737,
       "step": 13877
     },
     {
-      "epoch": 0.38114855400840403,
+      "epoch": 0.39381384790011353,
       "grad_norm": 0.0,
-      "learning_rate": 1.420095092163795e-05,
-      "loss": 0.9403,
+      "learning_rate": 1.3825348318680582e-05,
+      "loss": 0.924,
       "step": 13878
     },
     {
-      "epoch": 0.3811760182362473,
+      "epoch": 0.3938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.4200143686948095e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.3824499128550569e-05,
+      "loss": 0.8423,
       "step": 13879
     },
     {
-      "epoch": 0.3812034824640905,
+      "epoch": 0.3938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.4199336419025076e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.3823649906114837e-05,
+      "loss": 0.9807,
       "step": 13880
     },
     {
-      "epoch": 0.38123094669193375,
+      "epoch": 0.39389897843359817,
       "grad_norm": 0.0,
-      "learning_rate": 1.4198529117875284e-05,
-      "loss": 0.9068,
+      "learning_rate": 1.3822800651380565e-05,
+      "loss": 0.7928,
       "step": 13881
     },
     {
-      "epoch": 0.38125841091977697,
+      "epoch": 0.3939273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.41977217835051e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.3821951364354924e-05,
+      "loss": 0.8655,
       "step": 13882
     },
     {
-      "epoch": 0.38128587514762025,
+      "epoch": 0.39395573212258794,
       "grad_norm": 0.0,
-      "learning_rate": 1.4196914415920912e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.3821102045045088e-05,
+      "loss": 0.8925,
       "step": 13883
     },
     {
-      "epoch": 0.38131333937546347,
+      "epoch": 0.39398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.4196107015129115e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.3820252693458231e-05,
+      "loss": 0.9717,
       "step": 13884
     },
     {
-      "epoch": 0.3813408036033067,
+      "epoch": 0.3940124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.4195299581136097e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.3819403309601528e-05,
+      "loss": 0.982,
       "step": 13885
     },
     {
-      "epoch": 0.3813682678311499,
+      "epoch": 0.39404086265607263,
       "grad_norm": 0.0,
-      "learning_rate": 1.419449211394824e-05,
-      "loss": 1.0608,
+      "learning_rate": 1.3818553893482154e-05,
+      "loss": 0.8487,
       "step": 13886
     },
     {
-      "epoch": 0.3813957320589932,
+      "epoch": 0.39406923950056755,
       "grad_norm": 0.0,
-      "learning_rate": 1.4193684613571935e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.3817704445107283e-05,
+      "loss": 0.981,
       "step": 13887
     },
     {
-      "epoch": 0.3814231962868364,
+      "epoch": 0.3940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.4192877080013576e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.381685496448409e-05,
+      "loss": 0.9807,
       "step": 13888
     },
     {
-      "epoch": 0.38145066051467963,
+      "epoch": 0.3941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.419206951327955e-05,
-      "loss": 0.8925,
+      "learning_rate": 1.3816005451619754e-05,
+      "loss": 0.8261,
       "step": 13889
     },
     {
-      "epoch": 0.38147812474252285,
+      "epoch": 0.39415437003405224,
       "grad_norm": 0.0,
-      "learning_rate": 1.4191261913376241e-05,
-      "loss": 0.8905,
+      "learning_rate": 1.3815155906521445e-05,
+      "loss": 0.849,
       "step": 13890
     },
     {
-      "epoch": 0.38150558897036607,
+      "epoch": 0.3941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.4190454280310051e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.3814306329196345e-05,
+      "loss": 0.882,
       "step": 13891
     },
     {
-      "epoch": 0.38153305319820935,
+      "epoch": 0.394211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.4189646614087358e-05,
-      "loss": 0.786,
+      "learning_rate": 1.381345671965163e-05,
+      "loss": 0.9118,
       "step": 13892
     },
     {
-      "epoch": 0.38156051742605257,
+      "epoch": 0.39423950056753687,
       "grad_norm": 0.0,
-      "learning_rate": 1.418883891471456e-05,
-      "loss": 1.0441,
+      "learning_rate": 1.3812607077894471e-05,
+      "loss": 0.9675,
       "step": 13893
     },
     {
-      "epoch": 0.3815879816538958,
+      "epoch": 0.3942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.4188031182198047e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.3811757403932048e-05,
+      "loss": 0.8817,
       "step": 13894
     },
     {
-      "epoch": 0.381615445881739,
+      "epoch": 0.3942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.4187223416544207e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.3810907697771543e-05,
+      "loss": 1.0108,
       "step": 13895
     },
     {
-      "epoch": 0.3816429101095823,
+      "epoch": 0.39432463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.4186415617759432e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.3810057959420124e-05,
+      "loss": 0.8345,
       "step": 13896
     },
     {
-      "epoch": 0.3816703743374255,
+      "epoch": 0.3943530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.4185607785850122e-05,
-      "loss": 0.8603,
+      "learning_rate": 1.3809208188884978e-05,
+      "loss": 0.9331,
       "step": 13897
     },
     {
-      "epoch": 0.38169783856526873,
+      "epoch": 0.39438138479001134,
       "grad_norm": 0.0,
-      "learning_rate": 1.4184799920822657e-05,
-      "loss": 0.9096,
+      "learning_rate": 1.380835838617328e-05,
+      "loss": 1.015,
       "step": 13898
     },
     {
-      "epoch": 0.38172530279311195,
+      "epoch": 0.39440976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.418399202268343e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.3807508551292207e-05,
+      "loss": 0.9108,
       "step": 13899
     },
     {
-      "epoch": 0.38175276702095523,
+      "epoch": 0.3944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.4183184091438844e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.3806658684248935e-05,
+      "loss": 0.8598,
       "step": 13900
     },
     {
-      "epoch": 0.38178023124879845,
+      "epoch": 0.394466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.4182376127095284e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.380580878505065e-05,
+      "loss": 0.8629,
       "step": 13901
     },
     {
-      "epoch": 0.3818076954766417,
+      "epoch": 0.39449489216799094,
       "grad_norm": 0.0,
-      "learning_rate": 1.4181568129659144e-05,
-      "loss": 0.9176,
+      "learning_rate": 1.3804958853704524e-05,
+      "loss": 0.9199,
       "step": 13902
     },
     {
-      "epoch": 0.3818351597044849,
+      "epoch": 0.3945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.4180760099136814e-05,
-      "loss": 0.9676,
+      "learning_rate": 1.380410889021774e-05,
+      "loss": 0.9389,
       "step": 13903
     },
     {
-      "epoch": 0.3818626239323281,
+      "epoch": 0.3945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.4179952035534692e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.3803258894597478e-05,
+      "loss": 0.8932,
       "step": 13904
     },
     {
-      "epoch": 0.3818900881601714,
+      "epoch": 0.3945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.417914393885917e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.3802408866850917e-05,
+      "loss": 1.0344,
       "step": 13905
     },
     {
-      "epoch": 0.3819175523880146,
+      "epoch": 0.3946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.4178335809116646e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.3801558806985237e-05,
+      "loss": 0.955,
       "step": 13906
     },
     {
-      "epoch": 0.38194501661585784,
+      "epoch": 0.3946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.4177527646313505e-05,
-      "loss": 0.8236,
+      "learning_rate": 1.3800708715007618e-05,
+      "loss": 0.9318,
       "step": 13907
     },
     {
-      "epoch": 0.38197248084370106,
+      "epoch": 0.39466515323496026,
       "grad_norm": 0.0,
-      "learning_rate": 1.4176719450456151e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.3799858590925244e-05,
+      "loss": 1.0442,
       "step": 13908
     },
     {
-      "epoch": 0.38199994507154433,
+      "epoch": 0.3946935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.4175911221550975e-05,
-      "loss": 0.8383,
+      "learning_rate": 1.3799008434745291e-05,
+      "loss": 0.9638,
       "step": 13909
     },
     {
-      "epoch": 0.38202740929938755,
+      "epoch": 0.39472190692395004,
       "grad_norm": 0.0,
-      "learning_rate": 1.4175102959604368e-05,
-      "loss": 1.1057,
+      "learning_rate": 1.3798158246474946e-05,
+      "loss": 0.9997,
       "step": 13910
     },
     {
-      "epoch": 0.3820548735272308,
+      "epoch": 0.39475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.4174294664622733e-05,
-      "loss": 0.8749,
+      "learning_rate": 1.3797308026121386e-05,
+      "loss": 0.9687,
       "step": 13911
     },
     {
-      "epoch": 0.382082337755074,
+      "epoch": 0.3947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.417348633661246e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.3796457773691793e-05,
+      "loss": 0.9525,
       "step": 13912
     },
     {
-      "epoch": 0.3821098019829173,
+      "epoch": 0.39480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.4172677975579944e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.3795607489193351e-05,
+      "loss": 1.018,
       "step": 13913
     },
     {
-      "epoch": 0.3821372662107605,
+      "epoch": 0.39483541430192964,
       "grad_norm": 0.0,
-      "learning_rate": 1.4171869581531587e-05,
-      "loss": 1.0178,
+      "learning_rate": 1.3794757172633244e-05,
+      "loss": 0.974,
       "step": 13914
     },
     {
-      "epoch": 0.3821647304386037,
+      "epoch": 0.3948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.4171061154473782e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.3793906824018652e-05,
+      "loss": 0.9452,
       "step": 13915
     },
     {
-      "epoch": 0.38219219466644694,
+      "epoch": 0.3948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.4170252694412922e-05,
-      "loss": 0.934,
+      "learning_rate": 1.3793056443356758e-05,
+      "loss": 0.8829,
       "step": 13916
     },
     {
-      "epoch": 0.38221965889429016,
+      "epoch": 0.3949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.416944420135541e-05,
-      "loss": 1.0438,
+      "learning_rate": 1.3792206030654747e-05,
+      "loss": 0.9895,
       "step": 13917
     },
     {
-      "epoch": 0.38224712312213344,
+      "epoch": 0.3949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.416863567530764e-05,
-      "loss": 0.9482,
+      "learning_rate": 1.3791355585919801e-05,
+      "loss": 1.0075,
       "step": 13918
     },
     {
-      "epoch": 0.38227458734997666,
+      "epoch": 0.3949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.416782711627601e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.3790505109159101e-05,
+      "loss": 0.821,
       "step": 13919
     },
     {
-      "epoch": 0.3823020515778199,
+      "epoch": 0.39500567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.4167018524266919e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.3789654600379833e-05,
+      "loss": 0.8643,
       "step": 13920
     },
     {
-      "epoch": 0.3823295158056631,
+      "epoch": 0.3950340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.416620989928676e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.3788804059589186e-05,
+      "loss": 0.9462,
       "step": 13921
     },
     {
-      "epoch": 0.3823569800335064,
+      "epoch": 0.39506242905788874,
       "grad_norm": 0.0,
-      "learning_rate": 1.4165401241341936e-05,
-      "loss": 1.0502,
+      "learning_rate": 1.3787953486794342e-05,
+      "loss": 0.9818,
       "step": 13922
     },
     {
-      "epoch": 0.3823844442613496,
+      "epoch": 0.39509080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.4164592550438845e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.3787102882002482e-05,
+      "loss": 0.907,
       "step": 13923
     },
     {
-      "epoch": 0.3824119084891928,
+      "epoch": 0.39511918274687857,
       "grad_norm": 0.0,
-      "learning_rate": 1.4163783826583881e-05,
-      "loss": 0.8157,
+      "learning_rate": 1.3786252245220793e-05,
+      "loss": 1.0088,
       "step": 13924
     },
     {
-      "epoch": 0.38243937271703604,
+      "epoch": 0.39514755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.416297506978345e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.3785401576456463e-05,
+      "loss": 0.9605,
       "step": 13925
     },
     {
-      "epoch": 0.3824668369448793,
+      "epoch": 0.39517593643586835,
       "grad_norm": 0.0,
-      "learning_rate": 1.4162166280043948e-05,
-      "loss": 0.978,
+      "learning_rate": 1.3784550875716673e-05,
+      "loss": 0.8277,
       "step": 13926
     },
     {
-      "epoch": 0.38249430117272254,
+      "epoch": 0.3952043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.4161357457371771e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.3783700143008614e-05,
+      "loss": 0.918,
       "step": 13927
     },
     {
-      "epoch": 0.38252176540056576,
+      "epoch": 0.3952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.4160548601773321e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.3782849378339469e-05,
+      "loss": 0.9145,
       "step": 13928
     },
     {
-      "epoch": 0.382549229628409,
+      "epoch": 0.395261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.4159739713255002e-05,
-      "loss": 0.984,
+      "learning_rate": 1.3781998581716427e-05,
+      "loss": 0.8172,
       "step": 13929
     },
     {
-      "epoch": 0.38257669385625226,
+      "epoch": 0.3952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.4158930791823214e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.3781147753146671e-05,
+      "loss": 0.8864,
       "step": 13930
     },
     {
-      "epoch": 0.3826041580840955,
+      "epoch": 0.3953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.4158121837484346e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.378029689263739e-05,
+      "loss": 1.0336,
       "step": 13931
     },
     {
-      "epoch": 0.3826316223119387,
+      "epoch": 0.39534619750283767,
       "grad_norm": 0.0,
-      "learning_rate": 1.4157312850244815e-05,
-      "loss": 0.9031,
+      "learning_rate": 1.3779446000195773e-05,
+      "loss": 1.0138,
       "step": 13932
     },
     {
-      "epoch": 0.3826590865397819,
+      "epoch": 0.3953745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.4156503830111012e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.3778595075829002e-05,
+      "loss": 0.9579,
       "step": 13933
     },
     {
-      "epoch": 0.38268655076762514,
+      "epoch": 0.39540295119182745,
       "grad_norm": 0.0,
-      "learning_rate": 1.415569477708934e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.3777744119544271e-05,
+      "loss": 0.9721,
       "step": 13934
     },
     {
-      "epoch": 0.3827140149954684,
+      "epoch": 0.39543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.4154885691186201e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.3776893131348765e-05,
+      "loss": 0.9614,
       "step": 13935
     },
     {
-      "epoch": 0.38274147922331164,
+      "epoch": 0.3954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.4154076572407996e-05,
-      "loss": 1.02,
+      "learning_rate": 1.3776042111249676e-05,
+      "loss": 0.9737,
       "step": 13936
     },
     {
-      "epoch": 0.38276894345115486,
+      "epoch": 0.39548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.415326742076113e-05,
-      "loss": 0.888,
+      "learning_rate": 1.3775191059254185e-05,
+      "loss": 0.964,
       "step": 13937
     },
     {
-      "epoch": 0.3827964076789981,
+      "epoch": 0.39551645856980705,
       "grad_norm": 0.0,
-      "learning_rate": 1.4152458236252003e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.3774339975369489e-05,
+      "loss": 0.8179,
       "step": 13938
     },
     {
-      "epoch": 0.38282387190684136,
+      "epoch": 0.3955448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.4151649018887018e-05,
-      "loss": 0.8826,
+      "learning_rate": 1.3773488859602772e-05,
+      "loss": 0.9527,
       "step": 13939
     },
     {
-      "epoch": 0.3828513361346846,
+      "epoch": 0.3955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.4150839768672576e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.3772637711961223e-05,
+      "loss": 0.9319,
       "step": 13940
     },
     {
-      "epoch": 0.3828788003625278,
+      "epoch": 0.39560158910329174,
       "grad_norm": 0.0,
-      "learning_rate": 1.4150030485615084e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.3771786532452038e-05,
+      "loss": 0.915,
       "step": 13941
     },
     {
-      "epoch": 0.382906264590371,
+      "epoch": 0.3956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.4149221169720941e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.3770935321082399e-05,
+      "loss": 0.8839,
       "step": 13942
     },
     {
-      "epoch": 0.3829337288182143,
+      "epoch": 0.3956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.4148411820996555e-05,
-      "loss": 0.855,
+      "learning_rate": 1.3770084077859502e-05,
+      "loss": 0.8754,
       "step": 13943
     },
     {
-      "epoch": 0.3829611930460575,
+      "epoch": 0.3956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.4147602439448327e-05,
-      "loss": 0.9016,
+      "learning_rate": 1.3769232802790534e-05,
+      "loss": 0.9483,
       "step": 13944
     },
     {
-      "epoch": 0.38298865727390075,
+      "epoch": 0.3957150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.414679302508266e-05,
-      "loss": 0.8845,
+      "learning_rate": 1.3768381495882688e-05,
+      "loss": 0.9848,
       "step": 13945
     },
     {
-      "epoch": 0.38301612150174397,
+      "epoch": 0.39574347332576615,
       "grad_norm": 0.0,
-      "learning_rate": 1.4145983577905962e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.3767530157143154e-05,
+      "loss": 0.9648,
       "step": 13946
     },
     {
-      "epoch": 0.3830435857295872,
+      "epoch": 0.39577185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.4145174097924637e-05,
-      "loss": 0.9317,
+      "learning_rate": 1.3766678786579121e-05,
+      "loss": 0.9125,
       "step": 13947
     },
     {
-      "epoch": 0.38307104995743047,
+      "epoch": 0.395800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.4144364585145088e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.3765827384197787e-05,
+      "loss": 0.932,
       "step": 13948
     },
     {
-      "epoch": 0.3830985141852737,
+      "epoch": 0.39582860385925084,
       "grad_norm": 0.0,
-      "learning_rate": 1.4143555039573721e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.3764975950006339e-05,
+      "loss": 0.9032,
       "step": 13949
     },
     {
-      "epoch": 0.3831259784131169,
+      "epoch": 0.39585698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.4142745461216942e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.3764124484011965e-05,
+      "loss": 1.0319,
       "step": 13950
     },
     {
-      "epoch": 0.38315344264096013,
+      "epoch": 0.3958853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.4141935850081155e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.3763272986221864e-05,
+      "loss": 0.9464,
       "step": 13951
     },
     {
-      "epoch": 0.3831809068688034,
+      "epoch": 0.39591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.4141126206172766e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.376242145664323e-05,
+      "loss": 0.906,
       "step": 13952
     },
     {
-      "epoch": 0.3832083710966466,
+      "epoch": 0.39594211123723044,
       "grad_norm": 0.0,
-      "learning_rate": 1.4140316529498186e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.3761569895283253e-05,
+      "loss": 0.9283,
       "step": 13953
     },
     {
-      "epoch": 0.38323583532448985,
+      "epoch": 0.3959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.4139506820063815e-05,
-      "loss": 1.0642,
+      "learning_rate": 1.3760718302149125e-05,
+      "loss": 0.9045,
       "step": 13954
     },
     {
-      "epoch": 0.38326329955233307,
+      "epoch": 0.3959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.4138697077876063e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.3759866677248041e-05,
+      "loss": 0.8772,
       "step": 13955
     },
     {
-      "epoch": 0.38329076378017635,
+      "epoch": 0.3960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.4137887302941337e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.3759015020587194e-05,
+      "loss": 0.807,
       "step": 13956
     },
     {
-      "epoch": 0.38331822800801957,
+      "epoch": 0.39605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.4137077495266045e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.3758163332173772e-05,
+      "loss": 0.8831,
       "step": 13957
     },
     {
-      "epoch": 0.3833456922358628,
+      "epoch": 0.3960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.413626765485659e-05,
-      "loss": 0.8549,
+      "learning_rate": 1.3757311612014984e-05,
+      "loss": 0.7693,
       "step": 13958
     },
     {
-      "epoch": 0.383373156463706,
+      "epoch": 0.39611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.4135457781719386e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.3756459860118013e-05,
+      "loss": 1.0241,
       "step": 13959
     },
     {
-      "epoch": 0.38340062069154923,
+      "epoch": 0.3961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.4134647875860836e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.3755608076490054e-05,
+      "loss": 0.887,
       "step": 13960
     },
     {
-      "epoch": 0.3834280849193925,
+      "epoch": 0.39616912599318954,
       "grad_norm": 0.0,
-      "learning_rate": 1.4133837937287353e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.3754756261138308e-05,
+      "loss": 0.9341,
       "step": 13961
     },
     {
-      "epoch": 0.38345554914723573,
+      "epoch": 0.39619750283768446,
       "grad_norm": 0.0,
-      "learning_rate": 1.4133027966005344e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.3753904414069967e-05,
+      "loss": 0.789,
       "step": 13962
     },
     {
-      "epoch": 0.38348301337507895,
+      "epoch": 0.3962258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.4132217962021214e-05,
-      "loss": 1.0018,
+      "learning_rate": 1.3753052535292226e-05,
+      "loss": 0.9334,
       "step": 13963
     },
     {
-      "epoch": 0.3835104776029222,
+      "epoch": 0.39625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.4131407925341374e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.3752200624812282e-05,
+      "loss": 0.9669,
       "step": 13964
     },
     {
-      "epoch": 0.38353794183076545,
+      "epoch": 0.39628263337116915,
       "grad_norm": 0.0,
-      "learning_rate": 1.4130597855972235e-05,
-      "loss": 0.8956,
+      "learning_rate": 1.3751348682637328e-05,
+      "loss": 0.7889,
       "step": 13965
     },
     {
-      "epoch": 0.38356540605860867,
+      "epoch": 0.396311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.4129787753920205e-05,
-      "loss": 0.843,
+      "learning_rate": 1.3750496708774565e-05,
+      "loss": 0.9434,
       "step": 13966
     },
     {
-      "epoch": 0.3835928702864519,
+      "epoch": 0.3963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.4128977619191698e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.3749644703231188e-05,
+      "loss": 0.9832,
       "step": 13967
     },
     {
-      "epoch": 0.3836203345142951,
+      "epoch": 0.3963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.4128167451793116e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.3748792666014394e-05,
+      "loss": 0.8974,
       "step": 13968
     },
     {
-      "epoch": 0.3836477987421384,
+      "epoch": 0.3963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.4127357251730876e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.3747940597131382e-05,
+      "loss": 0.8868,
       "step": 13969
     },
     {
-      "epoch": 0.3836752629699816,
+      "epoch": 0.3964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.412654701901139e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.3747088496589343e-05,
+      "loss": 0.9296,
       "step": 13970
     },
     {
-      "epoch": 0.38370272719782483,
+      "epoch": 0.39645289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 1.4125736753641059e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.374623636439548e-05,
+      "loss": 0.9672,
       "step": 13971
     },
     {
-      "epoch": 0.38373019142566805,
+      "epoch": 0.3964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.4124926455626304e-05,
-      "loss": 0.8538,
+      "learning_rate": 1.374538420055699e-05,
+      "loss": 0.889,
       "step": 13972
     },
     {
-      "epoch": 0.3837576556535113,
+      "epoch": 0.39650964812712824,
       "grad_norm": 0.0,
-      "learning_rate": 1.4124116124973533e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.3744532005081072e-05,
+      "loss": 0.8626,
       "step": 13973
     },
     {
-      "epoch": 0.38378511988135455,
+      "epoch": 0.39653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.412330576168916e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.3743679777974923e-05,
+      "loss": 0.8569,
       "step": 13974
     },
     {
-      "epoch": 0.3838125841091978,
+      "epoch": 0.3965664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 1.412249536577959e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.3742827519245742e-05,
+      "loss": 0.9112,
       "step": 13975
     },
     {
-      "epoch": 0.383840048337041,
+      "epoch": 0.39659477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 1.4121684937251243e-05,
-      "loss": 0.8614,
+      "learning_rate": 1.3741975228900732e-05,
+      "loss": 0.9022,
       "step": 13976
     },
     {
-      "epoch": 0.3838675125648842,
+      "epoch": 0.39662315550510785,
       "grad_norm": 0.0,
-      "learning_rate": 1.412087447611053e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.3741122906947086e-05,
+      "loss": 0.9102,
       "step": 13977
     },
     {
-      "epoch": 0.3838949767927275,
+      "epoch": 0.3966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.4120063982363856e-05,
-      "loss": 1.0404,
+      "learning_rate": 1.374027055339201e-05,
+      "loss": 0.9073,
       "step": 13978
     },
     {
-      "epoch": 0.3839224410205707,
+      "epoch": 0.3966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.4119253456017643e-05,
-      "loss": 1.0292,
+      "learning_rate": 1.37394181682427e-05,
+      "loss": 0.8293,
       "step": 13979
     },
     {
-      "epoch": 0.38394990524841394,
+      "epoch": 0.3967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.4118442897078302e-05,
-      "loss": 0.9077,
+      "learning_rate": 1.3738565751506352e-05,
+      "loss": 0.9695,
       "step": 13980
     },
     {
-      "epoch": 0.38397736947625716,
+      "epoch": 0.3967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.4117632305552244e-05,
-      "loss": 0.9985,
+      "learning_rate": 1.3737713303190173e-05,
+      "loss": 0.9296,
       "step": 13981
     },
     {
-      "epoch": 0.38400483370410043,
+      "epoch": 0.3967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.4116821681445884e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.3736860823301364e-05,
+      "loss": 0.8237,
       "step": 13982
     },
     {
-      "epoch": 0.38403229793194366,
+      "epoch": 0.3967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.4116011024765638e-05,
-      "loss": 1.038,
+      "learning_rate": 1.3736008311847122e-05,
+      "loss": 0.8983,
       "step": 13983
     },
     {
-      "epoch": 0.3840597621597869,
+      "epoch": 0.3968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.4115200335517916e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.373515576883465e-05,
+      "loss": 0.8361,
       "step": 13984
     },
     {
-      "epoch": 0.3840872263876301,
+      "epoch": 0.39685017026106695,
       "grad_norm": 0.0,
-      "learning_rate": 1.4114389613709136e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.373430319427115e-05,
+      "loss": 0.9157,
       "step": 13985
     },
     {
-      "epoch": 0.3841146906154733,
+      "epoch": 0.39687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.4113578859345711e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.3733450588163822e-05,
+      "loss": 0.8539,
       "step": 13986
     },
     {
-      "epoch": 0.3841421548433166,
+      "epoch": 0.3969069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.4112768072434057e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.3732597950519868e-05,
+      "loss": 0.9666,
       "step": 13987
     },
     {
-      "epoch": 0.3841696190711598,
+      "epoch": 0.39693530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.411195725298059e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.3731745281346492e-05,
+      "loss": 0.9203,
       "step": 13988
     },
     {
-      "epoch": 0.38419708329900304,
+      "epoch": 0.39696367763904655,
       "grad_norm": 0.0,
-      "learning_rate": 1.4111146400991725e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.3730892580650898e-05,
+      "loss": 0.9694,
       "step": 13989
     },
     {
-      "epoch": 0.38422454752684626,
+      "epoch": 0.3969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.4110335516473874e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.3730039848440286e-05,
+      "loss": 0.9733,
       "step": 13990
     },
     {
-      "epoch": 0.38425201175468954,
+      "epoch": 0.3970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.4109524599433457e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.3729187084721858e-05,
+      "loss": 0.9702,
       "step": 13991
     },
     {
-      "epoch": 0.38427947598253276,
+      "epoch": 0.3970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.410871364987689e-05,
-      "loss": 0.8629,
+      "learning_rate": 1.372833428950282e-05,
+      "loss": 0.9664,
       "step": 13992
     },
     {
-      "epoch": 0.384306940210376,
+      "epoch": 0.3970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.410790266781059e-05,
-      "loss": 0.962,
+      "learning_rate": 1.3727481462790376e-05,
+      "loss": 0.9759,
       "step": 13993
     },
     {
-      "epoch": 0.3843344044382192,
+      "epoch": 0.397105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.4107091653240973e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.3726628604591725e-05,
+      "loss": 0.9636,
       "step": 13994
     },
     {
-      "epoch": 0.3843618686660625,
+      "epoch": 0.3971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.4106280606174454e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.3725775714914078e-05,
+      "loss": 0.8407,
       "step": 13995
     },
     {
-      "epoch": 0.3843893328939057,
+      "epoch": 0.3971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.4105469526617452e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.3724922793764634e-05,
+      "loss": 0.8644,
       "step": 13996
     },
     {
-      "epoch": 0.3844167971217489,
+      "epoch": 0.39719069239500565,
       "grad_norm": 0.0,
-      "learning_rate": 1.4104658414576389e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.3724069841150603e-05,
+      "loss": 0.907,
       "step": 13997
     },
     {
-      "epoch": 0.38444426134959214,
+      "epoch": 0.39721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.4103847270057675e-05,
-      "loss": 0.9166,
+      "learning_rate": 1.3723216857079183e-05,
+      "loss": 0.9664,
       "step": 13998
     },
     {
-      "epoch": 0.38447172557743536,
+      "epoch": 0.3972474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.4103036093067732e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.3722363841557584e-05,
+      "loss": 0.8872,
       "step": 13999
     },
     {
-      "epoch": 0.38449918980527864,
+      "epoch": 0.39727582292849034,
       "grad_norm": 0.0,
-      "learning_rate": 1.410222488361298e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.3721510794593012e-05,
+      "loss": 0.8683,
       "step": 14000
     },
     {
-      "epoch": 0.38452665403312186,
+      "epoch": 0.39730419977298526,
       "grad_norm": 0.0,
-      "learning_rate": 1.4101413641699836e-05,
-      "loss": 0.909,
+      "learning_rate": 1.3720657716192669e-05,
+      "loss": 0.8874,
       "step": 14001
     },
     {
-      "epoch": 0.3845541182609651,
+      "epoch": 0.3973325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.4100602367334715e-05,
-      "loss": 0.8781,
+      "learning_rate": 1.3719804606363765e-05,
+      "loss": 0.8706,
       "step": 14002
     },
     {
-      "epoch": 0.3845815824888083,
+      "epoch": 0.39736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.4099791060524042e-05,
-      "loss": 0.8573,
+      "learning_rate": 1.3718951465113499e-05,
+      "loss": 0.9254,
       "step": 14003
     },
     {
-      "epoch": 0.3846090467166516,
+      "epoch": 0.39738933030646995,
       "grad_norm": 0.0,
-      "learning_rate": 1.4098979721274232e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.3718098292449087e-05,
+      "loss": 0.8588,
       "step": 14004
     },
     {
-      "epoch": 0.3846365109444948,
+      "epoch": 0.3974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.409816834959171e-05,
-      "loss": 0.932,
+      "learning_rate": 1.3717245088377732e-05,
+      "loss": 0.8931,
       "step": 14005
     },
     {
-      "epoch": 0.384663975172338,
+      "epoch": 0.3974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.4097356945482892e-05,
-      "loss": 0.8866,
+      "learning_rate": 1.3716391852906638e-05,
+      "loss": 0.8305,
       "step": 14006
     },
     {
-      "epoch": 0.38469143940018125,
+      "epoch": 0.3974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.4096545508954194e-05,
-      "loss": 0.9621,
+      "learning_rate": 1.371553858604302e-05,
+      "loss": 0.8738,
       "step": 14007
     },
     {
-      "epoch": 0.3847189036280245,
+      "epoch": 0.3975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.4095734040012043e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.3714685287794075e-05,
+      "loss": 0.8989,
       "step": 14008
     },
     {
-      "epoch": 0.38474636785586774,
+      "epoch": 0.39753121452894435,
       "grad_norm": 0.0,
-      "learning_rate": 1.409492253866286e-05,
-      "loss": 0.9254,
+      "learning_rate": 1.3713831958167018e-05,
+      "loss": 0.9684,
       "step": 14009
     },
     {
-      "epoch": 0.38477383208371096,
+      "epoch": 0.39755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.4094111004913065e-05,
-      "loss": 0.8817,
+      "learning_rate": 1.371297859716906e-05,
+      "loss": 1.0417,
       "step": 14010
     },
     {
-      "epoch": 0.3848012963115542,
+      "epoch": 0.3975879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.4093299438769076e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.3712125204807397e-05,
+      "loss": 0.9893,
       "step": 14011
     },
     {
-      "epoch": 0.3848287605393974,
+      "epoch": 0.39761634506242904,
       "grad_norm": 0.0,
-      "learning_rate": 1.4092487840237314e-05,
-      "loss": 1.0035,
+      "learning_rate": 1.371127178108925e-05,
+      "loss": 0.9237,
       "step": 14012
     },
     {
-      "epoch": 0.3848562247672407,
+      "epoch": 0.39764472190692396,
       "grad_norm": 0.0,
-      "learning_rate": 1.4091676209324208e-05,
-      "loss": 0.8691,
+      "learning_rate": 1.3710418326021824e-05,
+      "loss": 0.877,
       "step": 14013
     },
     {
-      "epoch": 0.3848836889950839,
+      "epoch": 0.3976730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.409086454603617e-05,
-      "loss": 1.079,
+      "learning_rate": 1.3709564839612325e-05,
+      "loss": 0.9062,
       "step": 14014
     },
     {
-      "epoch": 0.3849111532229271,
+      "epoch": 0.39770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.4090052850379628e-05,
-      "loss": 0.9423,
+      "learning_rate": 1.370871132186797e-05,
+      "loss": 0.8866,
       "step": 14015
     },
     {
-      "epoch": 0.38493861745077035,
+      "epoch": 0.39772985244040865,
       "grad_norm": 0.0,
-      "learning_rate": 1.4089241122361003e-05,
-      "loss": 0.884,
+      "learning_rate": 1.3707857772795961e-05,
+      "loss": 0.8363,
       "step": 14016
     },
     {
-      "epoch": 0.3849660816786136,
+      "epoch": 0.3977582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.4088429361986721e-05,
-      "loss": 0.8647,
+      "learning_rate": 1.3707004192403507e-05,
+      "loss": 0.9248,
       "step": 14017
     },
     {
-      "epoch": 0.38499354590645685,
+      "epoch": 0.3977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.40876175692632e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.3706150580697826e-05,
+      "loss": 0.9214,
       "step": 14018
     },
     {
-      "epoch": 0.38502101013430007,
+      "epoch": 0.3978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.4086805744196865e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.3705296937686123e-05,
+      "loss": 0.9372,
       "step": 14019
     },
     {
-      "epoch": 0.3850484743621433,
+      "epoch": 0.3978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.4085993886794142e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.3704443263375611e-05,
+      "loss": 0.9359,
       "step": 14020
     },
     {
-      "epoch": 0.38507593858998657,
+      "epoch": 0.3978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.4085181997061452e-05,
-      "loss": 1.006,
+      "learning_rate": 1.3703589557773498e-05,
+      "loss": 0.8507,
       "step": 14021
     },
     {
-      "epoch": 0.3851034028178298,
+      "epoch": 0.397900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.408437007500522e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.3702735820887001e-05,
+      "loss": 0.8394,
       "step": 14022
     },
     {
-      "epoch": 0.385130867045673,
+      "epoch": 0.3979284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.4083558120631868e-05,
-      "loss": 0.9306,
+      "learning_rate": 1.3701882052723328e-05,
+      "loss": 0.9782,
       "step": 14023
     },
     {
-      "epoch": 0.38515833127351623,
+      "epoch": 0.39795686719636775,
       "grad_norm": 0.0,
-      "learning_rate": 1.4082746133947824e-05,
-      "loss": 0.9188,
+      "learning_rate": 1.3701028253289688e-05,
+      "loss": 0.8505,
       "step": 14024
     },
     {
-      "epoch": 0.3851857955013595,
+      "epoch": 0.39798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.4081934114959511e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.3700174422593298e-05,
+      "loss": 0.9926,
       "step": 14025
     },
     {
-      "epoch": 0.38521325972920273,
+      "epoch": 0.3980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.4081122063673354e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.3699320560641364e-05,
+      "loss": 0.9208,
       "step": 14026
     },
     {
-      "epoch": 0.38524072395704595,
+      "epoch": 0.39804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.4080309980095776e-05,
-      "loss": 0.9747,
+      "learning_rate": 1.369846666744111e-05,
+      "loss": 0.9676,
       "step": 14027
     },
     {
-      "epoch": 0.38526818818488917,
+      "epoch": 0.39807037457434735,
       "grad_norm": 0.0,
-      "learning_rate": 1.4079497864233206e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.3697612742999738e-05,
+      "loss": 0.9296,
       "step": 14028
     },
     {
-      "epoch": 0.3852956524127324,
+      "epoch": 0.3980987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.407868571609207e-05,
-      "loss": 0.8425,
+      "learning_rate": 1.3696758787324464e-05,
+      "loss": 0.9475,
       "step": 14029
     },
     {
-      "epoch": 0.38532311664057567,
+      "epoch": 0.3981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.4077873535678795e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.3695904800422506e-05,
+      "loss": 0.9563,
       "step": 14030
     },
     {
-      "epoch": 0.3853505808684189,
+      "epoch": 0.398155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.4077061322999798e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.3695050782301068e-05,
+      "loss": 0.946,
       "step": 14031
     },
     {
-      "epoch": 0.3853780450962621,
+      "epoch": 0.3981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.407624907806152e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.3694196732967378e-05,
+      "loss": 0.8388,
       "step": 14032
     },
     {
-      "epoch": 0.38540550932410533,
+      "epoch": 0.3982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.4075436800870374e-05,
-      "loss": 0.9866,
+      "learning_rate": 1.3693342652428637e-05,
+      "loss": 0.9332,
       "step": 14033
     },
     {
-      "epoch": 0.3854329735519486,
+      "epoch": 0.3982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.4074624491432798e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.3692488540692064e-05,
+      "loss": 0.8176,
       "step": 14034
     },
     {
-      "epoch": 0.38546043777979183,
+      "epoch": 0.3982690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.4073812149755213e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.3691634397764875e-05,
+      "loss": 0.9229,
       "step": 14035
     },
     {
-      "epoch": 0.38548790200763505,
+      "epoch": 0.39829738933030645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4072999775844047e-05,
-      "loss": 0.9887,
+      "learning_rate": 1.3690780223654286e-05,
+      "loss": 0.8465,
       "step": 14036
     },
     {
-      "epoch": 0.3855153662354783,
+      "epoch": 0.39832576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.407218736970573e-05,
-      "loss": 0.8937,
+      "learning_rate": 1.368992601836751e-05,
+      "loss": 0.9351,
       "step": 14037
     },
     {
-      "epoch": 0.38554283046332155,
+      "epoch": 0.3983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.4071374931346691e-05,
-      "loss": 0.9564,
+      "learning_rate": 1.3689071781911763e-05,
+      "loss": 0.9258,
       "step": 14038
     },
     {
-      "epoch": 0.38557029469116477,
+      "epoch": 0.39838251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.4070562460773353e-05,
-      "loss": 0.9788,
+      "learning_rate": 1.368821751429426e-05,
+      "loss": 0.8687,
       "step": 14039
     },
     {
-      "epoch": 0.385597758919008,
+      "epoch": 0.39841089670828606,
       "grad_norm": 0.0,
-      "learning_rate": 1.406974995799215e-05,
-      "loss": 0.8959,
+      "learning_rate": 1.3687363215522218e-05,
+      "loss": 0.8645,
       "step": 14040
     },
     {
-      "epoch": 0.3856252231468512,
+      "epoch": 0.3984392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.4068937423009507e-05,
-      "loss": 0.8828,
+      "learning_rate": 1.3686508885602852e-05,
+      "loss": 1.0078,
       "step": 14041
     },
     {
-      "epoch": 0.38565268737469444,
+      "epoch": 0.39846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 1.406812485583186e-05,
-      "loss": 0.8846,
+      "learning_rate": 1.368565452454338e-05,
+      "loss": 0.8593,
       "step": 14042
     },
     {
-      "epoch": 0.3856801516025377,
+      "epoch": 0.3984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.4067312256465626e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.3684800132351022e-05,
+      "loss": 0.8278,
       "step": 14043
     },
     {
-      "epoch": 0.38570761583038093,
+      "epoch": 0.3985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.406649962491725e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.3683945709032989e-05,
+      "loss": 0.9878,
       "step": 14044
     },
     {
-      "epoch": 0.38573508005822416,
+      "epoch": 0.3985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.4065686961193148e-05,
-      "loss": 0.8251,
+      "learning_rate": 1.36830912545965e-05,
+      "loss": 0.8791,
       "step": 14045
     },
     {
-      "epoch": 0.3857625442860674,
+      "epoch": 0.3985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.4064874265299757e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.3682236769048776e-05,
+      "loss": 0.9576,
       "step": 14046
     },
     {
-      "epoch": 0.38579000851391065,
+      "epoch": 0.3986095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.4064061537243508e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.3681382252397033e-05,
+      "loss": 0.9208,
       "step": 14047
     },
     {
-      "epoch": 0.3858174727417539,
+      "epoch": 0.39863791146424515,
       "grad_norm": 0.0,
-      "learning_rate": 1.4063248777030827e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.3680527704648485e-05,
+      "loss": 0.9581,
       "step": 14048
     },
     {
-      "epoch": 0.3858449369695971,
+      "epoch": 0.39866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.4062435984668148e-05,
-      "loss": 0.9419,
+      "learning_rate": 1.3679673125810356e-05,
+      "loss": 0.9698,
       "step": 14049
     },
     {
-      "epoch": 0.3858724011974403,
+      "epoch": 0.398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.4061623160161906e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.3678818515889863e-05,
+      "loss": 0.8327,
       "step": 14050
     },
     {
-      "epoch": 0.3858998654252836,
+      "epoch": 0.39872304199772984,
       "grad_norm": 0.0,
-      "learning_rate": 1.4060810303518522e-05,
-      "loss": 0.8815,
+      "learning_rate": 1.3677963874894225e-05,
+      "loss": 1.0735,
       "step": 14051
     },
     {
-      "epoch": 0.3859273296531268,
+      "epoch": 0.39875141884222476,
       "grad_norm": 0.0,
-      "learning_rate": 1.4059997414744437e-05,
-      "loss": 0.9349,
+      "learning_rate": 1.3677109202830658e-05,
+      "loss": 0.8836,
       "step": 14052
     },
     {
-      "epoch": 0.38595479388097004,
+      "epoch": 0.3987797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.405918449384608e-05,
-      "loss": 1.0408,
+      "learning_rate": 1.3676254499706387e-05,
+      "loss": 0.8907,
       "step": 14053
     },
     {
-      "epoch": 0.38598225810881326,
+      "epoch": 0.39880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.4058371540829884e-05,
-      "loss": 0.9406,
+      "learning_rate": 1.367539976552863e-05,
+      "loss": 0.8814,
       "step": 14054
     },
     {
-      "epoch": 0.3860097223366565,
+      "epoch": 0.39883654937570945,
       "grad_norm": 0.0,
-      "learning_rate": 1.4057558555702278e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.3674545000304602e-05,
+      "loss": 0.9973,
       "step": 14055
     },
     {
-      "epoch": 0.38603718656449976,
+      "epoch": 0.3988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.40567455384697e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.3673690204041531e-05,
+      "loss": 0.9298,
       "step": 14056
     },
     {
-      "epoch": 0.386064650792343,
+      "epoch": 0.3988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.4055932489138577e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.3672835376746631e-05,
+      "loss": 0.8771,
       "step": 14057
     },
     {
-      "epoch": 0.3860921150201862,
+      "epoch": 0.3989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.4055119407715346e-05,
-      "loss": 0.8399,
+      "learning_rate": 1.3671980518427126e-05,
+      "loss": 0.8794,
       "step": 14058
     },
     {
-      "epoch": 0.3861195792480294,
+      "epoch": 0.398950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.4054306294206442e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.3671125629090239e-05,
+      "loss": 0.974,
       "step": 14059
     },
     {
-      "epoch": 0.3861470434758727,
+      "epoch": 0.39897843359818386,
       "grad_norm": 0.0,
-      "learning_rate": 1.405349314861829e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.3670270708743187e-05,
+      "loss": 0.8234,
       "step": 14060
     },
     {
-      "epoch": 0.3861745077037159,
+      "epoch": 0.39900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.4052679970957335e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.3669415757393195e-05,
+      "loss": 0.7819,
       "step": 14061
     },
     {
-      "epoch": 0.38620197193155914,
+      "epoch": 0.3990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.4051866761230004e-05,
-      "loss": 0.939,
+      "learning_rate": 1.3668560775047478e-05,
+      "loss": 0.9624,
       "step": 14062
     },
     {
-      "epoch": 0.38622943615940236,
+      "epoch": 0.39906356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 1.4051053519442734e-05,
-      "loss": 0.8399,
+      "learning_rate": 1.366770576171327e-05,
+      "loss": 0.9672,
       "step": 14063
     },
     {
-      "epoch": 0.38625690038724564,
+      "epoch": 0.39909194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 1.4050240245601957e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.3666850717397783e-05,
+      "loss": 0.8745,
       "step": 14064
     },
     {
-      "epoch": 0.38628436461508886,
+      "epoch": 0.3991203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.4049426939714117e-05,
-      "loss": 1.0975,
+      "learning_rate": 1.366599564210824e-05,
+      "loss": 0.9058,
       "step": 14065
     },
     {
-      "epoch": 0.3863118288429321,
+      "epoch": 0.39914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.4048613601785638e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.366514053585187e-05,
+      "loss": 0.9412,
       "step": 14066
     },
     {
-      "epoch": 0.3863392930707753,
+      "epoch": 0.39917707150964815,
       "grad_norm": 0.0,
-      "learning_rate": 1.404780023182296e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.3664285398635895e-05,
+      "loss": 0.8697,
       "step": 14067
     },
     {
-      "epoch": 0.3863667572986185,
+      "epoch": 0.399205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.4046986829832518e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.3663430230467535e-05,
+      "loss": 0.9342,
       "step": 14068
     },
     {
-      "epoch": 0.3863942215264618,
+      "epoch": 0.3992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.404617339582075e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.3662575031354016e-05,
+      "loss": 0.924,
       "step": 14069
     },
     {
-      "epoch": 0.386421685754305,
+      "epoch": 0.3992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.4045359929794091e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.3661719801302562e-05,
+      "loss": 0.8571,
       "step": 14070
     },
     {
-      "epoch": 0.38644914998214824,
+      "epoch": 0.3992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.4044546431758977e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.3660864540320392e-05,
+      "loss": 0.9515,
       "step": 14071
     },
     {
-      "epoch": 0.38647661420999146,
+      "epoch": 0.39931895573212256,
       "grad_norm": 0.0,
-      "learning_rate": 1.4043732901721843e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.3660009248414735e-05,
+      "loss": 1.0246,
       "step": 14072
     },
     {
-      "epoch": 0.38650407843783474,
+      "epoch": 0.3993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.4042919339689129e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.3659153925592821e-05,
+      "loss": 0.8429,
       "step": 14073
     },
     {
-      "epoch": 0.38653154266567796,
+      "epoch": 0.3993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.4042105745667274e-05,
-      "loss": 0.9927,
+      "learning_rate": 1.3658298571861866e-05,
+      "loss": 1.0394,
       "step": 14074
     },
     {
-      "epoch": 0.3865590068935212,
+      "epoch": 0.39940408626560725,
       "grad_norm": 0.0,
-      "learning_rate": 1.4041292119662708e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.3657443187229098e-05,
+      "loss": 0.8828,
       "step": 14075
     },
     {
-      "epoch": 0.3865864711213644,
+      "epoch": 0.39943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.4040478461681874e-05,
-      "loss": 1.0503,
+      "learning_rate": 1.3656587771701744e-05,
+      "loss": 0.9893,
       "step": 14076
     },
     {
-      "epoch": 0.3866139353492077,
+      "epoch": 0.399460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.4039664771731212e-05,
-      "loss": 0.933,
+      "learning_rate": 1.3655732325287028e-05,
+      "loss": 0.8787,
       "step": 14077
     },
     {
-      "epoch": 0.3866413995770509,
+      "epoch": 0.39948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.4038851049817157e-05,
-      "loss": 1.0475,
+      "learning_rate": 1.3654876847992176e-05,
+      "loss": 0.8988,
       "step": 14078
     },
     {
-      "epoch": 0.3866688638048941,
+      "epoch": 0.39951759364358685,
       "grad_norm": 0.0,
-      "learning_rate": 1.4038037295946146e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.3654021339824417e-05,
+      "loss": 0.9186,
       "step": 14079
     },
     {
-      "epoch": 0.38669632803273735,
+      "epoch": 0.3995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.403722351012462e-05,
-      "loss": 0.902,
+      "learning_rate": 1.3653165800790973e-05,
+      "loss": 0.9092,
       "step": 14080
     },
     {
-      "epoch": 0.38672379226058057,
+      "epoch": 0.39957434733257663,
       "grad_norm": 0.0,
-      "learning_rate": 1.4036409692359022e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.3652310230899074e-05,
+      "loss": 0.9479,
       "step": 14081
     },
     {
-      "epoch": 0.38675125648842384,
+      "epoch": 0.3996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.4035595842655782e-05,
-      "loss": 0.9516,
+      "learning_rate": 1.3651454630155945e-05,
+      "loss": 0.9221,
       "step": 14082
     },
     {
-      "epoch": 0.38677872071626707,
+      "epoch": 0.3996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.4034781961021344e-05,
-      "loss": 0.894,
+      "learning_rate": 1.3650598998568818e-05,
+      "loss": 0.844,
       "step": 14083
     },
     {
-      "epoch": 0.3868061849441103,
+      "epoch": 0.3996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.403396804746215e-05,
-      "loss": 0.9342,
+      "learning_rate": 1.3649743336144917e-05,
+      "loss": 1.1044,
       "step": 14084
     },
     {
-      "epoch": 0.3868336491719535,
+      "epoch": 0.3996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.4033154101984638e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.3648887642891466e-05,
+      "loss": 0.9187,
       "step": 14085
     },
     {
-      "epoch": 0.3868611133997968,
+      "epoch": 0.3997162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.4032340124595249e-05,
-      "loss": 0.8642,
+      "learning_rate": 1.36480319188157e-05,
+      "loss": 0.9168,
       "step": 14086
     },
     {
-      "epoch": 0.38688857762764,
+      "epoch": 0.39974460839954595,
       "grad_norm": 0.0,
-      "learning_rate": 1.4031526115300422e-05,
-      "loss": 0.9973,
+      "learning_rate": 1.3647176163924846e-05,
+      "loss": 0.9726,
       "step": 14087
     },
     {
-      "epoch": 0.38691604185548323,
+      "epoch": 0.39977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.40307120741066e-05,
-      "loss": 0.9099,
+      "learning_rate": 1.3646320378226128e-05,
+      "loss": 0.8259,
       "step": 14088
     },
     {
-      "epoch": 0.38694350608332645,
+      "epoch": 0.39980136208853573,
       "grad_norm": 0.0,
-      "learning_rate": 1.4029898001020223e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.3645464561726779e-05,
+      "loss": 1.0223,
       "step": 14089
     },
     {
-      "epoch": 0.3869709703111697,
+      "epoch": 0.39982973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.4029083896047731e-05,
-      "loss": 0.8722,
+      "learning_rate": 1.3644608714434027e-05,
+      "loss": 0.8282,
       "step": 14090
     },
     {
-      "epoch": 0.38699843453901295,
+      "epoch": 0.39985811577752556,
       "grad_norm": 0.0,
-      "learning_rate": 1.4028269759195567e-05,
-      "loss": 0.995,
+      "learning_rate": 1.3643752836355103e-05,
+      "loss": 0.9061,
       "step": 14091
     },
     {
-      "epoch": 0.38702589876685617,
+      "epoch": 0.3998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.4027455590470172e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.3642896927497228e-05,
+      "loss": 0.8778,
       "step": 14092
     },
     {
-      "epoch": 0.3870533629946994,
+      "epoch": 0.39991486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 1.402664138987799e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.3642040987867647e-05,
+      "loss": 0.9539,
       "step": 14093
     },
     {
-      "epoch": 0.3870808272225426,
+      "epoch": 0.3999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.402582715742546e-05,
-      "loss": 0.9808,
+      "learning_rate": 1.364118501747358e-05,
+      "loss": 0.9563,
       "step": 14094
     },
     {
-      "epoch": 0.3871082914503859,
+      "epoch": 0.3999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.4025012893119024e-05,
-      "loss": 0.8576,
+      "learning_rate": 1.3640329016322259e-05,
+      "loss": 0.8923,
       "step": 14095
     },
     {
-      "epoch": 0.3871357556782291,
+      "epoch": 0.4,
       "grad_norm": 0.0,
-      "learning_rate": 1.4024198596965126e-05,
-      "loss": 0.9112,
+      "learning_rate": 1.3639472984420913e-05,
+      "loss": 0.9051,
       "step": 14096
     },
     {
-      "epoch": 0.38716321990607233,
+      "epoch": 0.4000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.4023384268970215e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.3638616921776775e-05,
+      "loss": 0.9249,
       "step": 14097
     },
     {
-      "epoch": 0.38719068413391555,
+      "epoch": 0.4000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.4022569909140727e-05,
-      "loss": 0.9273,
+      "learning_rate": 1.363776082839708e-05,
+      "loss": 0.8635,
       "step": 14098
     },
     {
-      "epoch": 0.38721814836175883,
+      "epoch": 0.40008513053348466,
       "grad_norm": 0.0,
-      "learning_rate": 1.4021755517483107e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.3636904704289053e-05,
+      "loss": 1.0508,
       "step": 14099
     },
     {
-      "epoch": 0.38724561258960205,
+      "epoch": 0.40011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.40209410940038e-05,
-      "loss": 0.8904,
+      "learning_rate": 1.363604854945993e-05,
+      "loss": 0.951,
       "step": 14100
     },
     {
-      "epoch": 0.38727307681744527,
+      "epoch": 0.4001418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.4020126638709249e-05,
-      "loss": 0.9439,
+      "learning_rate": 1.3635192363916941e-05,
+      "loss": 0.9568,
       "step": 14101
     },
     {
-      "epoch": 0.3873005410452885,
+      "epoch": 0.40017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.4019312151605897e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.3634336147667317e-05,
+      "loss": 0.8249,
       "step": 14102
     },
     {
-      "epoch": 0.38732800527313177,
+      "epoch": 0.40019863791146426,
       "grad_norm": 0.0,
-      "learning_rate": 1.4018497632700194e-05,
-      "loss": 0.9529,
+      "learning_rate": 1.3633479900718292e-05,
+      "loss": 0.9201,
       "step": 14103
     },
     {
-      "epoch": 0.387355469500975,
+      "epoch": 0.4002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.401768308199858e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.3632623623077104e-05,
+      "loss": 0.9312,
       "step": 14104
     },
     {
-      "epoch": 0.3873829337288182,
+      "epoch": 0.40025539160045404,
       "grad_norm": 0.0,
-      "learning_rate": 1.4016868499507501e-05,
-      "loss": 0.8958,
+      "learning_rate": 1.3631767314750977e-05,
+      "loss": 0.9136,
       "step": 14105
     },
     {
-      "epoch": 0.38741039795666143,
+      "epoch": 0.4002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.4016053885233402e-05,
-      "loss": 0.919,
+      "learning_rate": 1.3630910975747146e-05,
+      "loss": 0.9041,
       "step": 14106
     },
     {
-      "epoch": 0.38743786218450466,
+      "epoch": 0.4003121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.4015239239182727e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.363005460607285e-05,
+      "loss": 0.8525,
       "step": 14107
     },
     {
-      "epoch": 0.38746532641234793,
+      "epoch": 0.4003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.4014424561361926e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.362919820573532e-05,
+      "loss": 0.9945,
       "step": 14108
     },
     {
-      "epoch": 0.38749279064019115,
+      "epoch": 0.4003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.4013609851777443e-05,
-      "loss": 1.0148,
+      "learning_rate": 1.3628341774741789e-05,
+      "loss": 0.8693,
       "step": 14109
     },
     {
-      "epoch": 0.3875202548680344,
+      "epoch": 0.4003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.4012795110435726e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.3627485313099493e-05,
+      "loss": 0.9584,
       "step": 14110
     },
     {
-      "epoch": 0.3875477190958776,
+      "epoch": 0.40042565266742336,
       "grad_norm": 0.0,
-      "learning_rate": 1.4011980337343217e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.3626628820815662e-05,
+      "loss": 0.8205,
       "step": 14111
     },
     {
-      "epoch": 0.3875751833237209,
+      "epoch": 0.4004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.4011165532506366e-05,
-      "loss": 0.919,
+      "learning_rate": 1.3625772297897535e-05,
+      "loss": 0.9595,
       "step": 14112
     },
     {
-      "epoch": 0.3876026475515641,
+      "epoch": 0.4004824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.4010350695931618e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.3624915744352349e-05,
+      "loss": 0.8733,
       "step": 14113
     },
     {
-      "epoch": 0.3876301117794073,
+      "epoch": 0.40051078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 1.4009535827625424e-05,
-      "loss": 1.043,
+      "learning_rate": 1.3624059160187337e-05,
+      "loss": 0.9373,
       "step": 14114
     },
     {
-      "epoch": 0.38765757600725054,
+      "epoch": 0.40053916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.400872092759423e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.3623202545409733e-05,
+      "loss": 0.9684,
       "step": 14115
     },
     {
-      "epoch": 0.3876850402350938,
+      "epoch": 0.4005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.4007905995844482e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.3622345900026774e-05,
+      "loss": 0.9533,
       "step": 14116
     },
     {
-      "epoch": 0.38771250446293704,
+      "epoch": 0.40059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.4007091032382629e-05,
-      "loss": 0.8707,
+      "learning_rate": 1.3621489224045696e-05,
+      "loss": 0.9745,
       "step": 14117
     },
     {
-      "epoch": 0.38773996869078026,
+      "epoch": 0.40062429057888765,
       "grad_norm": 0.0,
-      "learning_rate": 1.400627603721512e-05,
-      "loss": 0.8285,
+      "learning_rate": 1.3620632517473736e-05,
+      "loss": 0.9072,
       "step": 14118
     },
     {
-      "epoch": 0.3877674329186235,
+      "epoch": 0.4006526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.4005461010348403e-05,
-      "loss": 1.006,
+      "learning_rate": 1.361977578031813e-05,
+      "loss": 0.9825,
       "step": 14119
     },
     {
-      "epoch": 0.3877948971464667,
+      "epoch": 0.40068104426787743,
       "grad_norm": 0.0,
-      "learning_rate": 1.4004645951788925e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.3618919012586116e-05,
+      "loss": 1.0121,
       "step": 14120
     },
     {
-      "epoch": 0.38782236137431,
+      "epoch": 0.4007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.400383086154314e-05,
-      "loss": 0.7762,
+      "learning_rate": 1.3618062214284931e-05,
+      "loss": 0.9414,
       "step": 14121
     },
     {
-      "epoch": 0.3878498256021532,
+      "epoch": 0.4007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.4003015739617493e-05,
-      "loss": 0.9436,
+      "learning_rate": 1.361720538542181e-05,
+      "loss": 0.9224,
       "step": 14122
     },
     {
-      "epoch": 0.3878772898299964,
+      "epoch": 0.40076617480136206,
       "grad_norm": 0.0,
-      "learning_rate": 1.4002200586018435e-05,
-      "loss": 1.0111,
+      "learning_rate": 1.3616348526003996e-05,
+      "loss": 0.9776,
       "step": 14123
     },
     {
-      "epoch": 0.38790475405783964,
+      "epoch": 0.400794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.4001385400752416e-05,
-      "loss": 0.8556,
+      "learning_rate": 1.361549163603872e-05,
+      "loss": 0.7985,
       "step": 14124
     },
     {
-      "epoch": 0.3879322182856829,
+      "epoch": 0.4008229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.4000570183825886e-05,
-      "loss": 0.9448,
+      "learning_rate": 1.3614634715533223e-05,
+      "loss": 1.0133,
       "step": 14125
     },
     {
-      "epoch": 0.38795968251352614,
+      "epoch": 0.40085130533484675,
       "grad_norm": 0.0,
-      "learning_rate": 1.3999754935245293e-05,
-      "loss": 0.8557,
+      "learning_rate": 1.3613777764494747e-05,
+      "loss": 0.9629,
       "step": 14126
     },
     {
-      "epoch": 0.38798714674136936,
+      "epoch": 0.40087968217934167,
       "grad_norm": 0.0,
-      "learning_rate": 1.3998939655017093e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.3612920782930525e-05,
+      "loss": 0.9365,
       "step": 14127
     },
     {
-      "epoch": 0.3880146109692126,
+      "epoch": 0.4009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.3998124343147727e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.36120637708478e-05,
+      "loss": 1.0044,
       "step": 14128
     },
     {
-      "epoch": 0.38804207519705586,
+      "epoch": 0.40093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.399730899964366e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.3611206728253813e-05,
+      "loss": 0.9489,
       "step": 14129
     },
     {
-      "epoch": 0.3880695394248991,
+      "epoch": 0.40096481271282636,
       "grad_norm": 0.0,
-      "learning_rate": 1.3996493624511334e-05,
-      "loss": 0.865,
+      "learning_rate": 1.3610349655155797e-05,
+      "loss": 0.8655,
       "step": 14130
     },
     {
-      "epoch": 0.3880970036527423,
+      "epoch": 0.4009931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.3995678217757199e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.3609492551560996e-05,
+      "loss": 1.0665,
       "step": 14131
     },
     {
-      "epoch": 0.3881244678805855,
+      "epoch": 0.40102156640181613,
       "grad_norm": 0.0,
-      "learning_rate": 1.399486277938771e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.3608635417476647e-05,
+      "loss": 0.952,
       "step": 14132
     },
     {
-      "epoch": 0.3881519321084288,
+      "epoch": 0.401049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.3994047309409322e-05,
-      "loss": 1.0714,
+      "learning_rate": 1.3607778252909996e-05,
+      "loss": 0.9134,
       "step": 14133
     },
     {
-      "epoch": 0.388179396336272,
+      "epoch": 0.4010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.3993231807828485e-05,
-      "loss": 0.9776,
+      "learning_rate": 1.3606921057868277e-05,
+      "loss": 0.9415,
       "step": 14134
     },
     {
-      "epoch": 0.38820686056411524,
+      "epoch": 0.4011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.3992416274651648e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.3606063832358736e-05,
+      "loss": 0.862,
       "step": 14135
     },
     {
-      "epoch": 0.38823432479195846,
+      "epoch": 0.4011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.3991600709885269e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.360520657638861e-05,
+      "loss": 0.8842,
       "step": 14136
     },
     {
-      "epoch": 0.3882617890198017,
+      "epoch": 0.4011634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.39907851135358e-05,
-      "loss": 1.0338,
+      "learning_rate": 1.3604349289965141e-05,
+      "loss": 0.8164,
       "step": 14137
     },
     {
-      "epoch": 0.38828925324764496,
+      "epoch": 0.40119182746878546,
       "grad_norm": 0.0,
-      "learning_rate": 1.3989969485609691e-05,
-      "loss": 0.9347,
+      "learning_rate": 1.3603491973095575e-05,
+      "loss": 1.0043,
       "step": 14138
     },
     {
-      "epoch": 0.3883167174754882,
+      "epoch": 0.40122020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.3989153826113401e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.3602634625787147e-05,
+      "loss": 0.8722,
       "step": 14139
     },
     {
-      "epoch": 0.3883441817033314,
+      "epoch": 0.40124858115777523,
       "grad_norm": 0.0,
-      "learning_rate": 1.3988338135053376e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.3601777248047105e-05,
+      "loss": 0.8312,
       "step": 14140
     },
     {
-      "epoch": 0.3883716459311746,
+      "epoch": 0.40127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.3987522412436078e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.3600919839882687e-05,
+      "loss": 0.9265,
       "step": 14141
     },
     {
-      "epoch": 0.3883991101590179,
+      "epoch": 0.40130533484676506,
       "grad_norm": 0.0,
-      "learning_rate": 1.3986706658267961e-05,
-      "loss": 0.9122,
+      "learning_rate": 1.3600062401301139e-05,
+      "loss": 0.8938,
       "step": 14142
     },
     {
-      "epoch": 0.3884265743868611,
+      "epoch": 0.4013337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.3985890872555472e-05,
-      "loss": 0.8898,
+      "learning_rate": 1.35992049323097e-05,
+      "loss": 0.9271,
       "step": 14143
     },
     {
-      "epoch": 0.38845403861470434,
+      "epoch": 0.40136208853575484,
       "grad_norm": 0.0,
-      "learning_rate": 1.398507505530507e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.3598347432915617e-05,
+      "loss": 0.8718,
       "step": 14144
     },
     {
-      "epoch": 0.38848150284254757,
+      "epoch": 0.4013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.3984259206523211e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.359748990312613e-05,
+      "loss": 0.897,
       "step": 14145
     },
     {
-      "epoch": 0.38850896707039084,
+      "epoch": 0.4014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.3983443326216352e-05,
-      "loss": 0.9181,
+      "learning_rate": 1.3596632342948484e-05,
+      "loss": 1.043,
       "step": 14146
     },
     {
-      "epoch": 0.38853643129823406,
+      "epoch": 0.4014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.3982627414390945e-05,
-      "loss": 0.9151,
+      "learning_rate": 1.3595774752389926e-05,
+      "loss": 0.8918,
       "step": 14147
     },
     {
-      "epoch": 0.3885638955260773,
+      "epoch": 0.4014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.3981811471053447e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.3594917131457695e-05,
+      "loss": 0.9951,
       "step": 14148
     },
     {
-      "epoch": 0.3885913597539205,
+      "epoch": 0.4015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.3980995496210317e-05,
-      "loss": 0.9796,
+      "learning_rate": 1.3594059480159034e-05,
+      "loss": 0.9488,
       "step": 14149
     },
     {
-      "epoch": 0.38861882398176373,
+      "epoch": 0.40153234960272416,
       "grad_norm": 0.0,
-      "learning_rate": 1.3980179489868006e-05,
-      "loss": 0.8884,
+      "learning_rate": 1.3593201798501192e-05,
+      "loss": 0.8777,
       "step": 14150
     },
     {
-      "epoch": 0.388646288209607,
+      "epoch": 0.4015607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.3979363452032972e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.3592344086491416e-05,
+      "loss": 1.0378,
       "step": 14151
     },
     {
-      "epoch": 0.3886737524374502,
+      "epoch": 0.401589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.3978547382711673e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.3591486344136949e-05,
+      "loss": 0.8334,
       "step": 14152
     },
     {
-      "epoch": 0.38870121666529345,
+      "epoch": 0.40161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.397773128191057e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.3590628571445033e-05,
+      "loss": 0.8509,
       "step": 14153
     },
     {
-      "epoch": 0.38872868089313667,
+      "epoch": 0.40164585698070376,
       "grad_norm": 0.0,
-      "learning_rate": 1.3976915149636115e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.3589770768422916e-05,
+      "loss": 0.9514,
       "step": 14154
     },
     {
-      "epoch": 0.38875614512097995,
+      "epoch": 0.4016742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.3976098985894762e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.3588912935077845e-05,
+      "loss": 0.9347,
       "step": 14155
     },
     {
-      "epoch": 0.38878360934882317,
+      "epoch": 0.40170261066969354,
       "grad_norm": 0.0,
-      "learning_rate": 1.3975282790692978e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.3588055071417063e-05,
+      "loss": 0.9684,
       "step": 14156
     },
     {
-      "epoch": 0.3888110735766664,
+      "epoch": 0.4017309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.3974466564037215e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.3587197177447817e-05,
+      "loss": 0.9735,
       "step": 14157
     },
     {
-      "epoch": 0.3888385378045096,
+      "epoch": 0.4017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.3973650305933934e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.3586339253177361e-05,
+      "loss": 0.8907,
       "step": 14158
     },
     {
-      "epoch": 0.3888660020323529,
+      "epoch": 0.40178774120317823,
       "grad_norm": 0.0,
-      "learning_rate": 1.3972834016389594e-05,
-      "loss": 1.0042,
+      "learning_rate": 1.3585481298612933e-05,
+      "loss": 1.0084,
       "step": 14159
     },
     {
-      "epoch": 0.3888934662601961,
+      "epoch": 0.4018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.3972017695410647e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.3584623313761783e-05,
+      "loss": 0.9351,
       "step": 14160
     },
     {
-      "epoch": 0.38892093048803933,
+      "epoch": 0.401844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.3971201343003561e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.358376529863116e-05,
+      "loss": 1.0032,
       "step": 14161
     },
     {
-      "epoch": 0.38894839471588255,
+      "epoch": 0.40187287173666286,
       "grad_norm": 0.0,
-      "learning_rate": 1.3970384959174794e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.358290725322831e-05,
+      "loss": 0.8275,
       "step": 14162
     },
     {
-      "epoch": 0.38897585894372577,
+      "epoch": 0.4019012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.3969568543930798e-05,
-      "loss": 0.879,
+      "learning_rate": 1.358204917756048e-05,
+      "loss": 0.9126,
       "step": 14163
     },
     {
-      "epoch": 0.38900332317156905,
+      "epoch": 0.4019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.396875209727804e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.3581191071634918e-05,
+      "loss": 0.9272,
       "step": 14164
     },
     {
-      "epoch": 0.38903078739941227,
+      "epoch": 0.40195800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 1.3967935619222982e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.3580332935458878e-05,
+      "loss": 0.9374,
       "step": 14165
     },
     {
-      "epoch": 0.3890582516272555,
+      "epoch": 0.40198637911464247,
       "grad_norm": 0.0,
-      "learning_rate": 1.3967119109772076e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.3579474769039602e-05,
+      "loss": 0.9744,
       "step": 14166
     },
     {
-      "epoch": 0.3890857158550987,
+      "epoch": 0.4020147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.3966302568931786e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.3578616572384345e-05,
+      "loss": 0.8313,
       "step": 14167
     },
     {
-      "epoch": 0.389113180082942,
+      "epoch": 0.40204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.3965485996708576e-05,
-      "loss": 0.937,
+      "learning_rate": 1.357775834550035e-05,
+      "loss": 0.993,
       "step": 14168
     },
     {
-      "epoch": 0.3891406443107852,
+      "epoch": 0.4020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.3964669393108904e-05,
-      "loss": 0.9794,
+      "learning_rate": 1.3576900088394871e-05,
+      "loss": 0.9125,
       "step": 14169
     },
     {
-      "epoch": 0.38916810853862843,
+      "epoch": 0.402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.3963852758139234e-05,
-      "loss": 0.8973,
+      "learning_rate": 1.3576041801075154e-05,
+      "loss": 0.9073,
       "step": 14170
     },
     {
-      "epoch": 0.38919557276647165,
+      "epoch": 0.40212826333711693,
       "grad_norm": 0.0,
-      "learning_rate": 1.3963036091806025e-05,
-      "loss": 0.9798,
+      "learning_rate": 1.3575183483548453e-05,
+      "loss": 0.9034,
       "step": 14171
     },
     {
-      "epoch": 0.38922303699431493,
+      "epoch": 0.4021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.3962219394115737e-05,
-      "loss": 0.9537,
+      "learning_rate": 1.3574325135822016e-05,
+      "loss": 0.9525,
       "step": 14172
     },
     {
-      "epoch": 0.38925050122215815,
+      "epoch": 0.4021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.3961402665074838e-05,
-      "loss": 0.913,
+      "learning_rate": 1.3573466757903094e-05,
+      "loss": 0.9802,
       "step": 14173
     },
     {
-      "epoch": 0.3892779654500014,
+      "epoch": 0.40221339387060157,
       "grad_norm": 0.0,
-      "learning_rate": 1.3960585904689782e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.3572608349798937e-05,
+      "loss": 0.8829,
       "step": 14174
     },
     {
-      "epoch": 0.3893054296778446,
+      "epoch": 0.4022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.395976911296704e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.3571749911516798e-05,
+      "loss": 0.7846,
       "step": 14175
     },
     {
-      "epoch": 0.3893328939056878,
+      "epoch": 0.4022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.3958952289913066e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.3570891443063925e-05,
+      "loss": 1.0746,
       "step": 14176
     },
     {
-      "epoch": 0.3893603581335311,
+      "epoch": 0.40229852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.3958135435534333e-05,
-      "loss": 0.9236,
+      "learning_rate": 1.3570032944447574e-05,
+      "loss": 0.9413,
       "step": 14177
     },
     {
-      "epoch": 0.3893878223613743,
+      "epoch": 0.40232690124858117,
       "grad_norm": 0.0,
-      "learning_rate": 1.3957318549837298e-05,
-      "loss": 0.8611,
+      "learning_rate": 1.3569174415674993e-05,
+      "loss": 0.8506,
       "step": 14178
     },
     {
-      "epoch": 0.38941528658921754,
+      "epoch": 0.40235527809307603,
       "grad_norm": 0.0,
-      "learning_rate": 1.3956501632828424e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.3568315856753436e-05,
+      "loss": 0.8839,
       "step": 14179
     },
     {
-      "epoch": 0.38944275081706076,
+      "epoch": 0.40238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.3955684684514176e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.3567457267690152e-05,
+      "loss": 0.8998,
       "step": 14180
     },
     {
-      "epoch": 0.38947021504490403,
+      "epoch": 0.40241203178206586,
       "grad_norm": 0.0,
-      "learning_rate": 1.3954867704901021e-05,
-      "loss": 0.9351,
+      "learning_rate": 1.3566598648492398e-05,
+      "loss": 1.0348,
       "step": 14181
     },
     {
-      "epoch": 0.38949767927274725,
+      "epoch": 0.4024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.3954050693995421e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.3565739999167425e-05,
+      "loss": 0.9623,
       "step": 14182
     },
     {
-      "epoch": 0.3895251435005905,
+      "epoch": 0.40246878547105563,
       "grad_norm": 0.0,
-      "learning_rate": 1.395323365180384e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.3564881319722485e-05,
+      "loss": 0.9379,
       "step": 14183
     },
     {
-      "epoch": 0.3895526077284337,
+      "epoch": 0.4024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.3952416578332738e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.3564022610164833e-05,
+      "loss": 0.8325,
       "step": 14184
     },
     {
-      "epoch": 0.389580071956277,
+      "epoch": 0.4025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.395159947358859e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.3563163870501723e-05,
+      "loss": 0.8933,
       "step": 14185
     },
     {
-      "epoch": 0.3896075361841202,
+      "epoch": 0.40255391600454027,
       "grad_norm": 0.0,
-      "learning_rate": 1.3950782337577857e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.3562305100740404e-05,
+      "loss": 0.9225,
       "step": 14186
     },
     {
-      "epoch": 0.3896350004119634,
+      "epoch": 0.4025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.3949965170307002e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.3561446300888137e-05,
+      "loss": 0.9133,
       "step": 14187
     },
     {
-      "epoch": 0.38966246463980664,
+      "epoch": 0.4026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.3949147971782491e-05,
-      "loss": 0.9363,
+      "learning_rate": 1.3560587470952172e-05,
+      "loss": 0.9233,
       "step": 14188
     },
     {
-      "epoch": 0.38968992886764986,
+      "epoch": 0.40263904653802496,
       "grad_norm": 0.0,
-      "learning_rate": 1.3948330742010794e-05,
-      "loss": 0.9983,
+      "learning_rate": 1.3559728610939767e-05,
+      "loss": 0.9408,
       "step": 14189
     },
     {
-      "epoch": 0.38971739309549314,
+      "epoch": 0.4026674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.3947513480998373e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.3558869720858167e-05,
+      "loss": 0.8918,
       "step": 14190
     },
     {
-      "epoch": 0.38974485732333636,
+      "epoch": 0.40269580022701473,
       "grad_norm": 0.0,
-      "learning_rate": 1.3946696188751698e-05,
-      "loss": 0.8541,
+      "learning_rate": 1.3558010800714643e-05,
+      "loss": 0.8717,
       "step": 14191
     },
     {
-      "epoch": 0.3897723215511796,
+      "epoch": 0.40272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.3945878865277234e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.3557151850516439e-05,
+      "loss": 1.0041,
       "step": 14192
     },
     {
-      "epoch": 0.3897997857790228,
+      "epoch": 0.40275255391600456,
       "grad_norm": 0.0,
-      "learning_rate": 1.3945061510581448e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.355629287027081e-05,
+      "loss": 0.866,
       "step": 14193
     },
     {
-      "epoch": 0.3898272500068661,
+      "epoch": 0.4027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.3944244124670805e-05,
-      "loss": 0.9635,
+      "learning_rate": 1.3555433859985019e-05,
+      "loss": 1.0156,
       "step": 14194
     },
     {
-      "epoch": 0.3898547142347093,
+      "epoch": 0.40280930760499434,
       "grad_norm": 0.0,
-      "learning_rate": 1.3943426707551777e-05,
-      "loss": 0.9562,
+      "learning_rate": 1.3554574819666315e-05,
+      "loss": 0.8217,
       "step": 14195
     },
     {
-      "epoch": 0.3898821784625525,
+      "epoch": 0.4028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.3942609259230826e-05,
-      "loss": 0.939,
+      "learning_rate": 1.3553715749321962e-05,
+      "loss": 1.026,
       "step": 14196
     },
     {
-      "epoch": 0.38990964269039574,
+      "epoch": 0.4028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.3941791779714426e-05,
-      "loss": 0.8706,
+      "learning_rate": 1.355285664895921e-05,
+      "loss": 0.9281,
       "step": 14197
     },
     {
-      "epoch": 0.389937106918239,
+      "epoch": 0.402894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3940974269009043e-05,
-      "loss": 0.9758,
+      "learning_rate": 1.3551997518585317e-05,
+      "loss": 0.8234,
       "step": 14198
     },
     {
-      "epoch": 0.38996457114608224,
+      "epoch": 0.4029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.3940156727121141e-05,
-      "loss": 0.8829,
+      "learning_rate": 1.3551138358207545e-05,
+      "loss": 0.8779,
       "step": 14199
     },
     {
-      "epoch": 0.38999203537392546,
+      "epoch": 0.4029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.3939339154057194e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.3550279167833145e-05,
+      "loss": 0.9014,
       "step": 14200
     },
     {
-      "epoch": 0.3900194996017687,
+      "epoch": 0.40297956867196366,
       "grad_norm": 0.0,
-      "learning_rate": 1.3938521549823673e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.3549419947469377e-05,
+      "loss": 0.9302,
       "step": 14201
     },
     {
-      "epoch": 0.3900469638296119,
+      "epoch": 0.4030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.3937703914427043e-05,
-      "loss": 0.8448,
+      "learning_rate": 1.3548560697123501e-05,
+      "loss": 0.9576,
       "step": 14202
     },
     {
-      "epoch": 0.3900744280574552,
+      "epoch": 0.40303632236095344,
       "grad_norm": 0.0,
-      "learning_rate": 1.3936886247873776e-05,
-      "loss": 0.8678,
+      "learning_rate": 1.3547701416802772e-05,
+      "loss": 0.9485,
       "step": 14203
     },
     {
-      "epoch": 0.3901018922852984,
+      "epoch": 0.40306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.3936068550170335e-05,
-      "loss": 0.9671,
+      "learning_rate": 1.3546842106514448e-05,
+      "loss": 0.916,
       "step": 14204
     },
     {
-      "epoch": 0.3901293565131416,
+      "epoch": 0.40309307604994327,
       "grad_norm": 0.0,
-      "learning_rate": 1.3935250821323199e-05,
-      "loss": 0.888,
+      "learning_rate": 1.3545982766265792e-05,
+      "loss": 0.8469,
       "step": 14205
     },
     {
-      "epoch": 0.39015682074098484,
+      "epoch": 0.4031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.3934433061338833e-05,
-      "loss": 0.8644,
+      "learning_rate": 1.354512339606406e-05,
+      "loss": 0.8489,
       "step": 14206
     },
     {
-      "epoch": 0.3901842849688281,
+      "epoch": 0.40314982973893304,
       "grad_norm": 0.0,
-      "learning_rate": 1.3933615270223706e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.3544263995916508e-05,
+      "loss": 0.963,
       "step": 14207
     },
     {
-      "epoch": 0.39021174919667134,
+      "epoch": 0.4031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.3932797447984293e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.3543404565830403e-05,
+      "loss": 0.7941,
       "step": 14208
     },
     {
-      "epoch": 0.39023921342451456,
+      "epoch": 0.4032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.3931979594627065e-05,
-      "loss": 1.0267,
+      "learning_rate": 1.3542545105813002e-05,
+      "loss": 0.8193,
       "step": 14209
     },
     {
-      "epoch": 0.3902666776523578,
+      "epoch": 0.40323496027241773,
       "grad_norm": 0.0,
-      "learning_rate": 1.3931161710158489e-05,
-      "loss": 0.9545,
+      "learning_rate": 1.3541685615871556e-05,
+      "loss": 0.8937,
       "step": 14210
     },
     {
-      "epoch": 0.39029414188020106,
+      "epoch": 0.4032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.393034379458504e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.3540826096013335e-05,
+      "loss": 0.95,
       "step": 14211
     },
     {
-      "epoch": 0.3903216061080443,
+      "epoch": 0.4032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.3929525847913188e-05,
-      "loss": 0.8563,
+      "learning_rate": 1.35399665462456e-05,
+      "loss": 0.8317,
       "step": 14212
     },
     {
-      "epoch": 0.3903490703358875,
+      "epoch": 0.40332009080590236,
       "grad_norm": 0.0,
-      "learning_rate": 1.3928707870149405e-05,
-      "loss": 1.0289,
+      "learning_rate": 1.3539106966575605e-05,
+      "loss": 0.8074,
       "step": 14213
     },
     {
-      "epoch": 0.3903765345637307,
+      "epoch": 0.4033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.3927889861300163e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.3538247357010617e-05,
+      "loss": 0.8472,
       "step": 14214
     },
     {
-      "epoch": 0.39040399879157395,
+      "epoch": 0.4033768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.3927071821371938e-05,
-      "loss": 0.8727,
+      "learning_rate": 1.3537387717557893e-05,
+      "loss": 0.8694,
       "step": 14215
     },
     {
-      "epoch": 0.3904314630194172,
+      "epoch": 0.40340522133938705,
       "grad_norm": 0.0,
-      "learning_rate": 1.3926253750371195e-05,
-      "loss": 0.952,
+      "learning_rate": 1.3536528048224698e-05,
+      "loss": 0.9474,
       "step": 14216
     },
     {
-      "epoch": 0.39045892724726045,
+      "epoch": 0.40343359818388197,
       "grad_norm": 0.0,
-      "learning_rate": 1.3925435648304415e-05,
-      "loss": 0.8821,
+      "learning_rate": 1.3535668349018287e-05,
+      "loss": 0.9844,
       "step": 14217
     },
     {
-      "epoch": 0.39048639147510367,
+      "epoch": 0.40346197502837683,
       "grad_norm": 0.0,
-      "learning_rate": 1.3924617515178066e-05,
-      "loss": 0.8953,
+      "learning_rate": 1.353480861994593e-05,
+      "loss": 0.9279,
       "step": 14218
     },
     {
-      "epoch": 0.3905138557029469,
+      "epoch": 0.40349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.3923799350998623e-05,
-      "loss": 0.8882,
+      "learning_rate": 1.3533948861014885e-05,
+      "loss": 1.0409,
       "step": 14219
     },
     {
-      "epoch": 0.39054131993079017,
+      "epoch": 0.4035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.3922981155772559e-05,
-      "loss": 0.9512,
+      "learning_rate": 1.3533089072232413e-05,
+      "loss": 0.9995,
       "step": 14220
     },
     {
-      "epoch": 0.3905687841586334,
+      "epoch": 0.4035471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.3922162929506351e-05,
-      "loss": 0.8397,
+      "learning_rate": 1.3532229253605783e-05,
+      "loss": 0.8368,
       "step": 14221
     },
     {
-      "epoch": 0.3905962483864766,
+      "epoch": 0.40357548240635643,
       "grad_norm": 0.0,
-      "learning_rate": 1.3921344672206469e-05,
-      "loss": 0.82,
+      "learning_rate": 1.3531369405142251e-05,
+      "loss": 0.9133,
       "step": 14222
     },
     {
-      "epoch": 0.39062371261431983,
+      "epoch": 0.4036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.3920526383879387e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.3530509526849086e-05,
+      "loss": 0.9565,
       "step": 14223
     },
     {
-      "epoch": 0.3906511768421631,
+      "epoch": 0.4036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.3919708064531585e-05,
-      "loss": 0.8122,
+      "learning_rate": 1.3529649618733545e-05,
+      "loss": 0.8373,
       "step": 14224
     },
     {
-      "epoch": 0.3906786410700063,
+      "epoch": 0.40366061293984107,
       "grad_norm": 0.0,
-      "learning_rate": 1.3918889714169532e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.3528789680802896e-05,
+      "loss": 0.9486,
       "step": 14225
     },
     {
-      "epoch": 0.39070610529784955,
+      "epoch": 0.403688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.3918071332799708e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.3527929713064401e-05,
+      "loss": 0.909,
       "step": 14226
     },
     {
-      "epoch": 0.39073356952569277,
+      "epoch": 0.4037173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.3917252920428586e-05,
-      "loss": 0.891,
+      "learning_rate": 1.3527069715525322e-05,
+      "loss": 0.9018,
       "step": 14227
     },
     {
-      "epoch": 0.39076103375353605,
+      "epoch": 0.40374574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.3916434477062636e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.3526209688192934e-05,
+      "loss": 0.925,
       "step": 14228
     },
     {
-      "epoch": 0.39078849798137927,
+      "epoch": 0.4037741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.3915616002708345e-05,
-      "loss": 0.9252,
+      "learning_rate": 1.352534963107449e-05,
+      "loss": 0.8229,
       "step": 14229
     },
     {
-      "epoch": 0.3908159622092225,
+      "epoch": 0.40380249716231553,
       "grad_norm": 0.0,
-      "learning_rate": 1.3914797497372183e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.352448954417726e-05,
+      "loss": 0.9642,
       "step": 14230
     },
     {
-      "epoch": 0.3908434264370657,
+      "epoch": 0.40383087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.3913978961060623e-05,
-      "loss": 0.8686,
+      "learning_rate": 1.352362942750851e-05,
+      "loss": 1.0441,
       "step": 14231
     },
     {
-      "epoch": 0.39087089066490893,
+      "epoch": 0.40385925085130536,
       "grad_norm": 0.0,
-      "learning_rate": 1.3913160393780147e-05,
-      "loss": 0.8706,
+      "learning_rate": 1.3522769281075502e-05,
+      "loss": 0.9313,
       "step": 14232
     },
     {
-      "epoch": 0.3908983548927522,
+      "epoch": 0.4038876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.3912341795537235e-05,
-      "loss": 0.9854,
+      "learning_rate": 1.3521909104885503e-05,
+      "loss": 0.9449,
       "step": 14233
     },
     {
-      "epoch": 0.39092581912059543,
+      "epoch": 0.40391600454029514,
       "grad_norm": 0.0,
-      "learning_rate": 1.3911523166338355e-05,
-      "loss": 1.0407,
+      "learning_rate": 1.3521048898945778e-05,
+      "loss": 0.9097,
       "step": 14234
     },
     {
-      "epoch": 0.39095328334843865,
+      "epoch": 0.40394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3910704506189987e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.3520188663263598e-05,
+      "loss": 0.8728,
       "step": 14235
     },
     {
-      "epoch": 0.3909807475762819,
+      "epoch": 0.4039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.3909885815098612e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.3519328397846225e-05,
+      "loss": 0.9063,
       "step": 14236
     },
     {
-      "epoch": 0.39100821180412515,
+      "epoch": 0.40400113507377977,
       "grad_norm": 0.0,
-      "learning_rate": 1.3909067093070705e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.3518468102700926e-05,
+      "loss": 1.0174,
       "step": 14237
     },
     {
-      "epoch": 0.39103567603196837,
+      "epoch": 0.4040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.3908248340112747e-05,
-      "loss": 0.905,
+      "learning_rate": 1.3517607777834972e-05,
+      "loss": 0.8932,
       "step": 14238
     },
     {
-      "epoch": 0.3910631402598116,
+      "epoch": 0.4040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.3907429556231211e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.3516747423255626e-05,
+      "loss": 1.0027,
       "step": 14239
     },
     {
-      "epoch": 0.3910906044876548,
+      "epoch": 0.40408626560726446,
       "grad_norm": 0.0,
-      "learning_rate": 1.390661074143258e-05,
-      "loss": 0.8622,
+      "learning_rate": 1.351588703897015e-05,
+      "loss": 0.9352,
       "step": 14240
     },
     {
-      "epoch": 0.3911180687154981,
+      "epoch": 0.4041146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.3905791895723332e-05,
-      "loss": 0.893,
+      "learning_rate": 1.3515026624985823e-05,
+      "loss": 0.939,
       "step": 14241
     },
     {
-      "epoch": 0.3911455329433413,
+      "epoch": 0.40414301929625424,
       "grad_norm": 0.0,
-      "learning_rate": 1.3904973019109949e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.3514166181309908e-05,
+      "loss": 0.9959,
       "step": 14242
     },
     {
-      "epoch": 0.39117299717118453,
+      "epoch": 0.40417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.3904154111598902e-05,
-      "loss": 0.9085,
+      "learning_rate": 1.3513305707949675e-05,
+      "loss": 0.7712,
       "step": 14243
     },
     {
-      "epoch": 0.39120046139902775,
+      "epoch": 0.40419977298524407,
       "grad_norm": 0.0,
-      "learning_rate": 1.3903335173196674e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.3512445204912387e-05,
+      "loss": 1.0507,
       "step": 14244
     },
     {
-      "epoch": 0.391227925626871,
+      "epoch": 0.4042281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.390251620390975e-05,
-      "loss": 0.96,
+      "learning_rate": 1.351158467220532e-05,
+      "loss": 0.8724,
       "step": 14245
     },
     {
-      "epoch": 0.39125538985471425,
+      "epoch": 0.40425652667423384,
       "grad_norm": 0.0,
-      "learning_rate": 1.3901697203744607e-05,
-      "loss": 1.0282,
+      "learning_rate": 1.351072410983574e-05,
+      "loss": 0.9167,
       "step": 14246
     },
     {
-      "epoch": 0.3912828540825575,
+      "epoch": 0.4042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.3900878172707722e-05,
-      "loss": 1.0225,
+      "learning_rate": 1.3509863517810913e-05,
+      "loss": 0.8942,
       "step": 14247
     },
     {
-      "epoch": 0.3913103183104007,
+      "epoch": 0.4043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.3900059110805578e-05,
-      "loss": 1.0501,
+      "learning_rate": 1.3509002896138113e-05,
+      "loss": 0.89,
       "step": 14248
     },
     {
-      "epoch": 0.3913377825382439,
+      "epoch": 0.4043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.3899240018044655e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.3508142244824604e-05,
+      "loss": 0.9271,
       "step": 14249
     },
     {
-      "epoch": 0.3913652467660872,
+      "epoch": 0.4043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.3898420894431434e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.3507281563877663e-05,
+      "loss": 0.9748,
       "step": 14250
     },
     {
-      "epoch": 0.3913927109939304,
+      "epoch": 0.4043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.3897601739972402e-05,
-      "loss": 0.9654,
+      "learning_rate": 1.3506420853304557e-05,
+      "loss": 0.9236,
       "step": 14251
     },
     {
-      "epoch": 0.39142017522177364,
+      "epoch": 0.40442678774120316,
       "grad_norm": 0.0,
-      "learning_rate": 1.389678255467403e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.3505560113112555e-05,
+      "loss": 0.9731,
       "step": 14252
     },
     {
-      "epoch": 0.39144763944961686,
+      "epoch": 0.4044551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.3895963338542807e-05,
-      "loss": 0.9,
+      "learning_rate": 1.3504699343308932e-05,
+      "loss": 0.9041,
       "step": 14253
     },
     {
-      "epoch": 0.39147510367746013,
+      "epoch": 0.40448354143019294,
       "grad_norm": 0.0,
-      "learning_rate": 1.3895144091585211e-05,
+      "learning_rate": 1.3503838543900952e-05,
       "loss": 0.9734,
       "step": 14254
     },
     {
-      "epoch": 0.39150256790530336,
+      "epoch": 0.40451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.3894324813807727e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.3502977714895894e-05,
+      "loss": 0.9558,
       "step": 14255
     },
     {
-      "epoch": 0.3915300321331466,
+      "epoch": 0.40454029511918277,
       "grad_norm": 0.0,
-      "learning_rate": 1.3893505505216835e-05,
-      "loss": 0.948,
+      "learning_rate": 1.3502116856301022e-05,
+      "loss": 0.8781,
       "step": 14256
     },
     {
-      "epoch": 0.3915574963609898,
+      "epoch": 0.40456867196367763,
       "grad_norm": 0.0,
-      "learning_rate": 1.3892686165819022e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.3501255968123613e-05,
+      "loss": 0.9635,
       "step": 14257
     },
     {
-      "epoch": 0.391584960588833,
+      "epoch": 0.40459704880817254,
       "grad_norm": 0.0,
-      "learning_rate": 1.389186679562077e-05,
-      "loss": 0.9204,
+      "learning_rate": 1.3500395050370939e-05,
+      "loss": 0.8921,
       "step": 14258
     },
     {
-      "epoch": 0.3916124248166763,
+      "epoch": 0.4046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.3891047394628555e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.3499534103050269e-05,
+      "loss": 0.9327,
       "step": 14259
     },
     {
-      "epoch": 0.3916398890445195,
+      "epoch": 0.4046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.389022796284887e-05,
-      "loss": 1.0564,
+      "learning_rate": 1.3498673126168877e-05,
+      "loss": 0.9776,
       "step": 14260
     },
     {
-      "epoch": 0.39166735327236274,
+      "epoch": 0.40468217934165723,
       "grad_norm": 0.0,
-      "learning_rate": 1.3889408500288192e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.3497812119734037e-05,
+      "loss": 0.9163,
       "step": 14261
     },
     {
-      "epoch": 0.39169481750020596,
+      "epoch": 0.4047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.388858900695301e-05,
-      "loss": 1.0024,
+      "learning_rate": 1.3496951083753021e-05,
+      "loss": 0.8616,
       "step": 14262
     },
     {
-      "epoch": 0.39172228172804924,
+      "epoch": 0.404738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.3887769482849805e-05,
-      "loss": 0.8853,
+      "learning_rate": 1.3496090018233101e-05,
+      "loss": 0.8687,
       "step": 14263
     },
     {
-      "epoch": 0.39174974595589246,
+      "epoch": 0.40476730987514187,
       "grad_norm": 0.0,
-      "learning_rate": 1.3886949927985058e-05,
-      "loss": 0.9483,
+      "learning_rate": 1.3495228923181552e-05,
+      "loss": 0.9041,
       "step": 14264
     },
     {
-      "epoch": 0.3917772101837357,
+      "epoch": 0.4047956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.388613034236526e-05,
-      "loss": 1.025,
+      "learning_rate": 1.3494367798605645e-05,
+      "loss": 0.9723,
       "step": 14265
     },
     {
-      "epoch": 0.3918046744115789,
+      "epoch": 0.40482406356413164,
       "grad_norm": 0.0,
-      "learning_rate": 1.3885310725996894e-05,
-      "loss": 0.8849,
+      "learning_rate": 1.3493506644512659e-05,
+      "loss": 0.9616,
       "step": 14266
     },
     {
-      "epoch": 0.3918321386394222,
+      "epoch": 0.40485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.3884491078886444e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.3492645460909866e-05,
+      "loss": 0.9046,
       "step": 14267
     },
     {
-      "epoch": 0.3918596028672654,
+      "epoch": 0.40488081725312147,
       "grad_norm": 0.0,
-      "learning_rate": 1.3883671401040398e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.3491784247804535e-05,
+      "loss": 0.9223,
       "step": 14268
     },
     {
-      "epoch": 0.3918870670951086,
+      "epoch": 0.40490919409761633,
       "grad_norm": 0.0,
-      "learning_rate": 1.3882851692465233e-05,
-      "loss": 0.981,
+      "learning_rate": 1.3490923005203951e-05,
+      "loss": 0.8704,
       "step": 14269
     },
     {
-      "epoch": 0.39191453132295184,
+      "epoch": 0.40493757094211125,
       "grad_norm": 0.0,
-      "learning_rate": 1.3882031953167448e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.3490061733115383e-05,
+      "loss": 0.8922,
       "step": 14270
     },
     {
-      "epoch": 0.39194199555079506,
+      "epoch": 0.4049659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.388121218315352e-05,
-      "loss": 0.8767,
+      "learning_rate": 1.3489200431546103e-05,
+      "loss": 0.9627,
       "step": 14271
     },
     {
-      "epoch": 0.39196945977863834,
+      "epoch": 0.404994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.3880392382429938e-05,
-      "loss": 1.058,
+      "learning_rate": 1.348833910050339e-05,
+      "loss": 0.9836,
       "step": 14272
     },
     {
-      "epoch": 0.39199692400648156,
+      "epoch": 0.40502270147559594,
       "grad_norm": 0.0,
-      "learning_rate": 1.387957255100319e-05,
-      "loss": 0.9541,
+      "learning_rate": 1.3487477739994522e-05,
+      "loss": 1.0171,
       "step": 14273
     },
     {
-      "epoch": 0.3920243882343248,
+      "epoch": 0.4050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.387875268887976e-05,
-      "loss": 0.8524,
+      "learning_rate": 1.3486616350026771e-05,
+      "loss": 1.0769,
       "step": 14274
     },
     {
-      "epoch": 0.392051852462168,
+      "epoch": 0.4050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.3877932796066133e-05,
-      "loss": 0.8275,
+      "learning_rate": 1.3485754930607418e-05,
+      "loss": 0.8737,
       "step": 14275
     },
     {
-      "epoch": 0.3920793166900113,
+      "epoch": 0.40510783200908057,
       "grad_norm": 0.0,
-      "learning_rate": 1.3877112872568801e-05,
-      "loss": 1.0688,
+      "learning_rate": 1.3484893481743735e-05,
+      "loss": 0.983,
       "step": 14276
     },
     {
-      "epoch": 0.3921067809178545,
+      "epoch": 0.4051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.3876292918394251e-05,
-      "loss": 0.9994,
+      "learning_rate": 1.3484032003443002e-05,
+      "loss": 0.918,
       "step": 14277
     },
     {
-      "epoch": 0.3921342451456977,
+      "epoch": 0.4051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.3875472933548972e-05,
-      "loss": 0.9472,
+      "learning_rate": 1.348317049571249e-05,
+      "loss": 0.91,
       "step": 14278
     },
     {
-      "epoch": 0.39216170937354095,
+      "epoch": 0.40519296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.3874652918039448e-05,
-      "loss": 0.9655,
+      "learning_rate": 1.3482308958559483e-05,
+      "loss": 0.8387,
       "step": 14279
     },
     {
-      "epoch": 0.3921891736013842,
+      "epoch": 0.4052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.387383287187217e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.3481447391991254e-05,
+      "loss": 0.8626,
       "step": 14280
     },
     {
-      "epoch": 0.39221663782922744,
+      "epoch": 0.40524971623155503,
       "grad_norm": 0.0,
-      "learning_rate": 1.3873012795053624e-05,
-      "loss": 0.9803,
+      "learning_rate": 1.3480585796015082e-05,
+      "loss": 1.0355,
       "step": 14281
     },
     {
-      "epoch": 0.39224410205707066,
+      "epoch": 0.40527809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.3872192687590302e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.3479724170638248e-05,
+      "loss": 0.98,
       "step": 14282
     },
     {
-      "epoch": 0.3922715662849139,
+      "epoch": 0.4053064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.3871372549488693e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.3478862515868025e-05,
+      "loss": 0.9513,
       "step": 14283
     },
     {
-      "epoch": 0.3922990305127571,
+      "epoch": 0.4053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.387055238075528e-05,
-      "loss": 0.8507,
+      "learning_rate": 1.3478000831711695e-05,
+      "loss": 0.902,
       "step": 14284
     },
     {
-      "epoch": 0.3923264947406004,
+      "epoch": 0.40536322360953464,
       "grad_norm": 0.0,
-      "learning_rate": 1.386973218139656e-05,
-      "loss": 0.8582,
+      "learning_rate": 1.3477139118176536e-05,
+      "loss": 0.922,
       "step": 14285
     },
     {
-      "epoch": 0.3923539589684436,
+      "epoch": 0.4053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.3868911951419021e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.3476277375269828e-05,
+      "loss": 0.7798,
       "step": 14286
     },
     {
-      "epoch": 0.3923814231962868,
+      "epoch": 0.4054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.386809169082915e-05,
-      "loss": 0.8635,
+      "learning_rate": 1.3475415602998844e-05,
+      "loss": 1.0759,
       "step": 14287
     },
     {
-      "epoch": 0.39240888742413005,
+      "epoch": 0.4054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.386727139963344e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.3474553801370871e-05,
+      "loss": 0.9047,
       "step": 14288
     },
     {
-      "epoch": 0.3924363516519733,
+      "epoch": 0.4054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.3866451077838382e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.3473691970393187e-05,
+      "loss": 0.9977,
       "step": 14289
     },
     {
-      "epoch": 0.39246381587981655,
+      "epoch": 0.4055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.3865630725450465e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.3472830110073071e-05,
+      "loss": 0.9434,
       "step": 14290
     },
     {
-      "epoch": 0.39249128010765977,
+      "epoch": 0.40553348467650396,
       "grad_norm": 0.0,
-      "learning_rate": 1.3864810342476178e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.34719682204178e-05,
+      "loss": 0.9628,
       "step": 14291
     },
     {
-      "epoch": 0.392518744335503,
+      "epoch": 0.4055618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.3863989928922017e-05,
-      "loss": 0.9635,
+      "learning_rate": 1.347110630143466e-05,
+      "loss": 0.9391,
       "step": 14292
     },
     {
-      "epoch": 0.39254620856334627,
+      "epoch": 0.40559023836549374,
       "grad_norm": 0.0,
-      "learning_rate": 1.3863169484794469e-05,
-      "loss": 1.0135,
+      "learning_rate": 1.347024435313093e-05,
+      "loss": 0.9308,
       "step": 14293
     },
     {
-      "epoch": 0.3925736727911895,
+      "epoch": 0.40561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.386234901010003e-05,
-      "loss": 0.8876,
+      "learning_rate": 1.3469382375513886e-05,
+      "loss": 0.965,
       "step": 14294
     },
     {
-      "epoch": 0.3926011370190327,
+      "epoch": 0.40564699205448357,
       "grad_norm": 0.0,
-      "learning_rate": 1.3861528504845188e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.3468520368590813e-05,
+      "loss": 0.9616,
       "step": 14295
     },
     {
-      "epoch": 0.39262860124687593,
+      "epoch": 0.40567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.3860707969036436e-05,
-      "loss": 0.997,
+      "learning_rate": 1.3467658332368996e-05,
+      "loss": 0.9775,
       "step": 14296
     },
     {
-      "epoch": 0.39265606547471915,
+      "epoch": 0.40570374574347334,
       "grad_norm": 0.0,
-      "learning_rate": 1.3859887402680268e-05,
-      "loss": 0.8466,
+      "learning_rate": 1.3466796266855713e-05,
+      "loss": 0.8726,
       "step": 14297
     },
     {
-      "epoch": 0.39268352970256243,
+      "epoch": 0.4057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.3859066805783177e-05,
-      "loss": 0.8638,
+      "learning_rate": 1.3465934172058244e-05,
+      "loss": 0.9091,
       "step": 14298
     },
     {
-      "epoch": 0.39271099393040565,
+      "epoch": 0.4057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.385824617835165e-05,
-      "loss": 0.9427,
+      "learning_rate": 1.3465072047983875e-05,
+      "loss": 0.9659,
       "step": 14299
     },
     {
-      "epoch": 0.39273845815824887,
+      "epoch": 0.405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.385742552039219e-05,
-      "loss": 0.8336,
+      "learning_rate": 1.3464209894639885e-05,
+      "loss": 1.0234,
       "step": 14300
     },
     {
-      "epoch": 0.3927659223860921,
+      "epoch": 0.4058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.3856604831911283e-05,
-      "loss": 0.9069,
+      "learning_rate": 1.3463347712033558e-05,
+      "loss": 0.9158,
       "step": 14301
     },
     {
-      "epoch": 0.39279338661393537,
+      "epoch": 0.4058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.3855784112915424e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.3462485500172177e-05,
+      "loss": 0.9729,
       "step": 14302
     },
     {
-      "epoch": 0.3928208508417786,
+      "epoch": 0.40587400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.3854963363411106e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.3461623259063031e-05,
+      "loss": 0.9003,
       "step": 14303
     },
     {
-      "epoch": 0.3928483150696218,
+      "epoch": 0.4059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.3854142583404828e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.3460760988713393e-05,
+      "loss": 0.8542,
       "step": 14304
     },
     {
-      "epoch": 0.39287577929746503,
+      "epoch": 0.40593076049943244,
       "grad_norm": 0.0,
-      "learning_rate": 1.385332177290308e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.3459898689130548e-05,
+      "loss": 0.8381,
       "step": 14305
     },
     {
-      "epoch": 0.3929032435253083,
+      "epoch": 0.40595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.3852500931912355e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.345903636032179e-05,
+      "loss": 0.9701,
       "step": 14306
     },
     {
-      "epoch": 0.39293070775315153,
+      "epoch": 0.40598751418842227,
       "grad_norm": 0.0,
-      "learning_rate": 1.3851680060439155e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.3458174002294394e-05,
+      "loss": 0.9718,
       "step": 14307
     },
     {
-      "epoch": 0.39295817198099475,
+      "epoch": 0.40601589103291713,
       "grad_norm": 0.0,
-      "learning_rate": 1.3850859158489964e-05,
-      "loss": 0.8463,
+      "learning_rate": 1.3457311615055643e-05,
+      "loss": 0.9367,
       "step": 14308
     },
     {
-      "epoch": 0.392985636208838,
+      "epoch": 0.40604426787741205,
       "grad_norm": 0.0,
-      "learning_rate": 1.385003822607129e-05,
-      "loss": 0.8978,
+      "learning_rate": 1.345644919861283e-05,
+      "loss": 0.9099,
       "step": 14309
     },
     {
-      "epoch": 0.3930131004366812,
+      "epoch": 0.4060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.3849217263189621e-05,
-      "loss": 0.9292,
+      "learning_rate": 1.3455586752973234e-05,
+      "loss": 1.0078,
       "step": 14310
     },
     {
-      "epoch": 0.39304056466452447,
+      "epoch": 0.4061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.384839626985145e-05,
-      "loss": 0.8941,
+      "learning_rate": 1.3454724278144142e-05,
+      "loss": 0.9303,
       "step": 14311
     },
     {
-      "epoch": 0.3930680288923677,
+      "epoch": 0.40612939841089674,
       "grad_norm": 0.0,
-      "learning_rate": 1.3847575246063279e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.3453861774132836e-05,
+      "loss": 0.9547,
       "step": 14312
     },
     {
-      "epoch": 0.3930954931202109,
+      "epoch": 0.4061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.3846754191831604e-05,
-      "loss": 0.9468,
+      "learning_rate": 1.3452999240946606e-05,
+      "loss": 0.9812,
       "step": 14313
     },
     {
-      "epoch": 0.39312295734805414,
+      "epoch": 0.4061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.3845933107162918e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.3452136678592737e-05,
+      "loss": 0.9545,
       "step": 14314
     },
     {
-      "epoch": 0.3931504215758974,
+      "epoch": 0.40621452894438137,
       "grad_norm": 0.0,
-      "learning_rate": 1.384511199206372e-05,
-      "loss": 1.0346,
+      "learning_rate": 1.3451274087078514e-05,
+      "loss": 0.9913,
       "step": 14315
     },
     {
-      "epoch": 0.39317788580374063,
+      "epoch": 0.4062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.3844290846540508e-05,
-      "loss": 0.9066,
+      "learning_rate": 1.3450411466411224e-05,
+      "loss": 0.8968,
       "step": 14316
     },
     {
-      "epoch": 0.39320535003158386,
+      "epoch": 0.40627128263337114,
       "grad_norm": 0.0,
-      "learning_rate": 1.3843469670599775e-05,
-      "loss": 0.884,
+      "learning_rate": 1.3449548816598152e-05,
+      "loss": 0.9656,
       "step": 14317
     },
     {
-      "epoch": 0.3932328142594271,
+      "epoch": 0.40629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.3842648464248021e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.3448686137646587e-05,
+      "loss": 0.9181,
       "step": 14318
     },
     {
-      "epoch": 0.39326027848727035,
+      "epoch": 0.406328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.3841827227491746e-05,
-      "loss": 0.834,
+      "learning_rate": 1.3447823429563813e-05,
+      "loss": 1.0258,
       "step": 14319
     },
     {
-      "epoch": 0.3932877427151136,
+      "epoch": 0.40635641316685583,
       "grad_norm": 0.0,
-      "learning_rate": 1.3841005960337443e-05,
-      "loss": 0.8203,
+      "learning_rate": 1.3446960692357122e-05,
+      "loss": 0.9492,
       "step": 14320
     },
     {
-      "epoch": 0.3933152069429568,
+      "epoch": 0.40638479001135075,
       "grad_norm": 0.0,
-      "learning_rate": 1.3840184662791616e-05,
-      "loss": 1.1285,
+      "learning_rate": 1.34460979260338e-05,
+      "loss": 0.8802,
       "step": 14321
     },
     {
-      "epoch": 0.3933426711708,
+      "epoch": 0.4064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.3839363334860761e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.3445235130601129e-05,
+      "loss": 0.944,
       "step": 14322
     },
     {
-      "epoch": 0.3933701353986433,
+      "epoch": 0.4064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.3838541976551371e-05,
-      "loss": 1.0318,
+      "learning_rate": 1.3444372306066407e-05,
+      "loss": 0.852,
       "step": 14323
     },
     {
-      "epoch": 0.3933975996264865,
+      "epoch": 0.40646992054483544,
       "grad_norm": 0.0,
-      "learning_rate": 1.3837720587869955e-05,
-      "loss": 0.9803,
+      "learning_rate": 1.3443509452436917e-05,
+      "loss": 0.9875,
       "step": 14324
     },
     {
-      "epoch": 0.39342506385432974,
+      "epoch": 0.4064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.3836899168823006e-05,
-      "loss": 0.8929,
+      "learning_rate": 1.3442646569719944e-05,
+      "loss": 0.9626,
       "step": 14325
     },
     {
-      "epoch": 0.39345252808217296,
+      "epoch": 0.4065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.3836077719417022e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.3441783657922782e-05,
+      "loss": 0.8694,
       "step": 14326
     },
     {
-      "epoch": 0.3934799923100162,
+      "epoch": 0.4065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.383525623965851e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.3440920717052721e-05,
+      "loss": 0.848,
       "step": 14327
     },
     {
-      "epoch": 0.39350745653785946,
+      "epoch": 0.406583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.383443472955396e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.3440057747117049e-05,
+      "loss": 0.8667,
       "step": 14328
     },
     {
-      "epoch": 0.3935349207657027,
+      "epoch": 0.40661180476730985,
       "grad_norm": 0.0,
-      "learning_rate": 1.3833613189109878e-05,
-      "loss": 0.8111,
+      "learning_rate": 1.3439194748123048e-05,
+      "loss": 0.8591,
       "step": 14329
     },
     {
-      "epoch": 0.3935623849935459,
+      "epoch": 0.40664018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 1.3832791618332769e-05,
-      "loss": 0.9945,
+      "learning_rate": 1.3438331720078019e-05,
+      "loss": 0.9164,
       "step": 14330
     },
     {
-      "epoch": 0.3935898492213891,
+      "epoch": 0.4066685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.3831970017229121e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.3437468662989251e-05,
+      "loss": 1.0431,
       "step": 14331
     },
     {
-      "epoch": 0.3936173134492324,
+      "epoch": 0.40669693530079454,
       "grad_norm": 0.0,
-      "learning_rate": 1.3831148385805444e-05,
-      "loss": 1.0606,
+      "learning_rate": 1.3436605576864023e-05,
+      "loss": 0.8838,
       "step": 14332
     },
     {
-      "epoch": 0.3936447776770756,
+      "epoch": 0.40672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.3830326724068237e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.3435742461709635e-05,
+      "loss": 0.8778,
       "step": 14333
     },
     {
-      "epoch": 0.39367224190491884,
+      "epoch": 0.4067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.3829505032024004e-05,
-      "loss": 1.0029,
+      "learning_rate": 1.343487931753338e-05,
+      "loss": 0.9594,
       "step": 14334
     },
     {
-      "epoch": 0.39369970613276206,
+      "epoch": 0.4067820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.382868330967924e-05,
-      "loss": 0.849,
+      "learning_rate": 1.3434016144342543e-05,
+      "loss": 0.8501,
       "step": 14335
     },
     {
-      "epoch": 0.39372717036060534,
+      "epoch": 0.40681044267877414,
       "grad_norm": 0.0,
-      "learning_rate": 1.3827861557040452e-05,
-      "loss": 0.9847,
+      "learning_rate": 1.343315294214442e-05,
+      "loss": 0.9147,
       "step": 14336
     },
     {
-      "epoch": 0.39375463458844856,
+      "epoch": 0.406838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.3827039774114139e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.3432289710946296e-05,
+      "loss": 0.9583,
       "step": 14337
     },
     {
-      "epoch": 0.3937820988162918,
+      "epoch": 0.4068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.382621796090681e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.3431426450755465e-05,
+      "loss": 0.8976,
       "step": 14338
     },
     {
-      "epoch": 0.393809563044135,
+      "epoch": 0.4068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.3825396117424959e-05,
-      "loss": 0.8725,
+      "learning_rate": 1.3430563161579224e-05,
+      "loss": 0.8316,
       "step": 14339
     },
     {
-      "epoch": 0.3938370272719782,
+      "epoch": 0.4069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.3824574243675089e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.3429699843424861e-05,
+      "loss": 0.9248,
       "step": 14340
     },
     {
-      "epoch": 0.3938644914998215,
+      "epoch": 0.4069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.382375233966371e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.3428836496299668e-05,
+      "loss": 0.9213,
       "step": 14341
     },
     {
-      "epoch": 0.3938919557276647,
+      "epoch": 0.40698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.382293040539732e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.342797312021094e-05,
+      "loss": 0.9184,
       "step": 14342
     },
     {
-      "epoch": 0.39391941995550794,
+      "epoch": 0.4070090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.3822108440882424e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.3427109715165969e-05,
+      "loss": 0.895,
       "step": 14343
     },
     {
-      "epoch": 0.39394688418335116,
+      "epoch": 0.40703745743473324,
       "grad_norm": 0.0,
-      "learning_rate": 1.3821286446125526e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.3426246281172048e-05,
+      "loss": 0.9753,
       "step": 14344
     },
     {
-      "epoch": 0.39397434841119444,
+      "epoch": 0.40706583427922816,
       "grad_norm": 0.0,
-      "learning_rate": 1.3820464421133129e-05,
-      "loss": 0.8488,
+      "learning_rate": 1.3425382818236469e-05,
+      "loss": 0.9518,
       "step": 14345
     },
     {
-      "epoch": 0.39400181263903766,
+      "epoch": 0.407094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.3819642365911739e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.3424519326366531e-05,
+      "loss": 0.9157,
       "step": 14346
     },
     {
-      "epoch": 0.3940292768668809,
+      "epoch": 0.40712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.3818820280467856e-05,
-      "loss": 0.8592,
+      "learning_rate": 1.3423655805569524e-05,
+      "loss": 0.939,
       "step": 14347
     },
     {
-      "epoch": 0.3940567410947241,
+      "epoch": 0.40715096481271285,
       "grad_norm": 0.0,
-      "learning_rate": 1.3817998164807988e-05,
-      "loss": 0.968,
+      "learning_rate": 1.3422792255852739e-05,
+      "loss": 0.8992,
       "step": 14348
     },
     {
-      "epoch": 0.3940842053225674,
+      "epoch": 0.4071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.381717601893864e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.3421928677223474e-05,
+      "loss": 0.9944,
       "step": 14349
     },
     {
-      "epoch": 0.3941116695504106,
+      "epoch": 0.4072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.381635384286632e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.3421065069689028e-05,
+      "loss": 0.8723,
       "step": 14350
     },
     {
-      "epoch": 0.3941391337782538,
+      "epoch": 0.4072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.3815531636597528e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.342020143325669e-05,
+      "loss": 0.8992,
       "step": 14351
     },
     {
-      "epoch": 0.39416659800609705,
+      "epoch": 0.4072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.3814709400138769e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.3419337767933753e-05,
+      "loss": 0.9035,
       "step": 14352
     },
     {
-      "epoch": 0.39419406223394027,
+      "epoch": 0.4072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.3813887133496556e-05,
-      "loss": 0.8997,
+      "learning_rate": 1.341847407372752e-05,
+      "loss": 0.9686,
       "step": 14353
     },
     {
-      "epoch": 0.39422152646178354,
+      "epoch": 0.40732122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 1.3813064836677388e-05,
-      "loss": 0.7921,
+      "learning_rate": 1.3417610350645283e-05,
+      "loss": 0.837,
       "step": 14354
     },
     {
-      "epoch": 0.39424899068962677,
+      "epoch": 0.4073496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.3812242509687774e-05,
-      "loss": 0.9558,
+      "learning_rate": 1.3416746598694336e-05,
+      "loss": 0.9409,
       "step": 14355
     },
     {
-      "epoch": 0.39427645491747,
+      "epoch": 0.40737797956867194,
       "grad_norm": 0.0,
-      "learning_rate": 1.3811420152534218e-05,
-      "loss": 1.0431,
+      "learning_rate": 1.3415882817881975e-05,
+      "loss": 0.9227,
       "step": 14356
     },
     {
-      "epoch": 0.3943039191453132,
+      "epoch": 0.40740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.3810597765223235e-05,
-      "loss": 0.85,
+      "learning_rate": 1.3415019008215501e-05,
+      "loss": 0.874,
       "step": 14357
     },
     {
-      "epoch": 0.3943313833731565,
+      "epoch": 0.4074347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.3809775347761324e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.341415516970221e-05,
+      "loss": 0.9647,
       "step": 14358
     },
     {
-      "epoch": 0.3943588476009997,
+      "epoch": 0.40746311010215663,
       "grad_norm": 0.0,
-      "learning_rate": 1.3808952900154993e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.3413291302349394e-05,
+      "loss": 0.848,
       "step": 14359
     },
     {
-      "epoch": 0.39438631182884293,
+      "epoch": 0.40749148694665155,
       "grad_norm": 0.0,
-      "learning_rate": 1.3808130422410755e-05,
-      "loss": 0.8716,
+      "learning_rate": 1.3412427406164353e-05,
+      "loss": 0.8574,
       "step": 14360
     },
     {
-      "epoch": 0.39441377605668615,
+      "epoch": 0.4075198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.3807307914535108e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.3411563481154387e-05,
+      "loss": 0.7858,
       "step": 14361
     },
     {
-      "epoch": 0.3944412402845294,
+      "epoch": 0.4075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.3806485376534573e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.3410699527326788e-05,
+      "loss": 0.9324,
       "step": 14362
     },
     {
-      "epoch": 0.39446870451237265,
+      "epoch": 0.4075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.3805662808415648e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.3409835544688856e-05,
+      "loss": 0.9104,
       "step": 14363
     },
     {
-      "epoch": 0.39449616874021587,
+      "epoch": 0.4076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.3804840210184842e-05,
-      "loss": 1.0079,
+      "learning_rate": 1.340897153324789e-05,
+      "loss": 0.9141,
       "step": 14364
     },
     {
-      "epoch": 0.3945236329680591,
+      "epoch": 0.407633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.380401758184867e-05,
-      "loss": 0.782,
+      "learning_rate": 1.340810749301119e-05,
+      "loss": 0.9539,
       "step": 14365
     },
     {
-      "epoch": 0.3945510971959023,
+      "epoch": 0.4076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.3803194923413639e-05,
-      "loss": 0.9802,
+      "learning_rate": 1.3407243423986052e-05,
+      "loss": 0.9205,
       "step": 14366
     },
     {
-      "epoch": 0.3945785614237456,
+      "epoch": 0.4076901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.3802372234886251e-05,
-      "loss": 0.8481,
+      "learning_rate": 1.3406379326179778e-05,
+      "loss": 0.9323,
       "step": 14367
     },
     {
-      "epoch": 0.3946060256515888,
+      "epoch": 0.40771850170261065,
       "grad_norm": 0.0,
-      "learning_rate": 1.3801549516273023e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.3405515199599662e-05,
+      "loss": 0.9081,
       "step": 14368
     },
     {
-      "epoch": 0.39463348987943203,
+      "epoch": 0.40774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.3800726767580463e-05,
-      "loss": 1.0332,
+      "learning_rate": 1.3404651044253006e-05,
+      "loss": 0.9907,
       "step": 14369
     },
     {
-      "epoch": 0.39466095410727525,
+      "epoch": 0.4077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.3799903988815081e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.340378686014711e-05,
+      "loss": 0.8883,
       "step": 14370
     },
     {
-      "epoch": 0.39468841833511853,
+      "epoch": 0.40780363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.3799081179983388e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.3402922647289277e-05,
+      "loss": 0.9226,
       "step": 14371
     },
     {
-      "epoch": 0.39471588256296175,
+      "epoch": 0.40783200908059025,
       "grad_norm": 0.0,
-      "learning_rate": 1.3798258341091892e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.3402058405686798e-05,
+      "loss": 0.9118,
       "step": 14372
     },
     {
-      "epoch": 0.39474334679080497,
+      "epoch": 0.4078603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.3797435472147102e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.3401194135346978e-05,
+      "loss": 0.8558,
       "step": 14373
     },
     {
-      "epoch": 0.3947708110186482,
+      "epoch": 0.40788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.3796612573155534e-05,
-      "loss": 1.0335,
+      "learning_rate": 1.3400329836277122e-05,
+      "loss": 1.0351,
       "step": 14374
     },
     {
-      "epoch": 0.39479827524649147,
+      "epoch": 0.40791713961407494,
       "grad_norm": 0.0,
-      "learning_rate": 1.3795789644123698e-05,
-      "loss": 0.9583,
+      "learning_rate": 1.3399465508484526e-05,
+      "loss": 0.9672,
       "step": 14375
     },
     {
-      "epoch": 0.3948257394743347,
+      "epoch": 0.4079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.3794966685058102e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.3398601151976491e-05,
+      "loss": 0.7903,
       "step": 14376
     },
     {
-      "epoch": 0.3948532037021779,
+      "epoch": 0.4079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.3794143695965261e-05,
-      "loss": 0.9906,
+      "learning_rate": 1.3397736766760322e-05,
+      "loss": 0.9157,
       "step": 14377
     },
     {
-      "epoch": 0.39488066793002113,
+      "epoch": 0.4080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.3793320676851685e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.3396872352843318e-05,
+      "loss": 0.9552,
       "step": 14378
     },
     {
-      "epoch": 0.39490813215786436,
+      "epoch": 0.4080306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.3792497627723887e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.3396007910232773e-05,
+      "loss": 0.7886,
       "step": 14379
     },
     {
-      "epoch": 0.39493559638570763,
+      "epoch": 0.40805902383654935,
       "grad_norm": 0.0,
-      "learning_rate": 1.3791674548588374e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.3395143438936003e-05,
+      "loss": 0.9583,
       "step": 14380
     },
     {
-      "epoch": 0.39496306061355085,
+      "epoch": 0.40808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.3790851439451669e-05,
-      "loss": 0.9204,
+      "learning_rate": 1.3394278938960303e-05,
+      "loss": 0.943,
       "step": 14381
     },
     {
-      "epoch": 0.3949905248413941,
+      "epoch": 0.4081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.3790028300320277e-05,
-      "loss": 0.9471,
+      "learning_rate": 1.339341441031298e-05,
+      "loss": 1.016,
       "step": 14382
     },
     {
-      "epoch": 0.3950179890692373,
+      "epoch": 0.40814415437003404,
       "grad_norm": 0.0,
-      "learning_rate": 1.3789205131200713e-05,
-      "loss": 0.8821,
+      "learning_rate": 1.3392549853001326e-05,
+      "loss": 0.7882,
       "step": 14383
     },
     {
-      "epoch": 0.3950454532970806,
+      "epoch": 0.40817253121452896,
       "grad_norm": 0.0,
-      "learning_rate": 1.378838193209949e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.3391685267032654e-05,
+      "loss": 0.9324,
       "step": 14384
     },
     {
-      "epoch": 0.3950729175249238,
+      "epoch": 0.4082009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.3787558703023121e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.3390820652414265e-05,
+      "loss": 1.044,
       "step": 14385
     },
     {
-      "epoch": 0.395100381752767,
+      "epoch": 0.40822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.3786735443978124e-05,
-      "loss": 0.9681,
+      "learning_rate": 1.338995600915346e-05,
+      "loss": 0.9042,
       "step": 14386
     },
     {
-      "epoch": 0.39512784598061024,
+      "epoch": 0.40825766174801364,
       "grad_norm": 0.0,
-      "learning_rate": 1.3785912154971004e-05,
-      "loss": 0.889,
+      "learning_rate": 1.3389091337257543e-05,
+      "loss": 0.9753,
       "step": 14387
     },
     {
-      "epoch": 0.3951553102084535,
+      "epoch": 0.4082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.3785088836008283e-05,
-      "loss": 0.9063,
+      "learning_rate": 1.338822663673382e-05,
+      "loss": 0.8623,
       "step": 14388
     },
     {
-      "epoch": 0.39518277443629674,
+      "epoch": 0.4083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.3784265487096473e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.3387361907589596e-05,
+      "loss": 0.8027,
       "step": 14389
     },
     {
-      "epoch": 0.39521023866413996,
+      "epoch": 0.4083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.3783442108242089e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.3386497149832175e-05,
+      "loss": 0.7882,
       "step": 14390
     },
     {
-      "epoch": 0.3952377028919832,
+      "epoch": 0.4083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.3782618699451644e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.3385632363468858e-05,
+      "loss": 0.8401,
       "step": 14391
     },
     {
-      "epoch": 0.3952651671198264,
+      "epoch": 0.4083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.3781795260731656e-05,
-      "loss": 1.0229,
+      "learning_rate": 1.338476754850695e-05,
+      "loss": 0.9522,
       "step": 14392
     },
     {
-      "epoch": 0.3952926313476697,
+      "epoch": 0.40842792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.3780971792088637e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.3383902704953761e-05,
+      "loss": 0.8026,
       "step": 14393
     },
     {
-      "epoch": 0.3953200955755129,
+      "epoch": 0.4084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.3780148293529107e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.3383037832816594e-05,
+      "loss": 0.934,
       "step": 14394
     },
     {
-      "epoch": 0.3953475598033561,
+      "epoch": 0.40848467650397274,
       "grad_norm": 0.0,
-      "learning_rate": 1.3779324765059578e-05,
-      "loss": 0.9803,
+      "learning_rate": 1.3382172932102755e-05,
+      "loss": 0.9206,
       "step": 14395
     },
     {
-      "epoch": 0.39537502403119934,
+      "epoch": 0.40851305334846766,
       "grad_norm": 0.0,
-      "learning_rate": 1.3778501206686567e-05,
-      "loss": 0.8765,
+      "learning_rate": 1.3381308002819546e-05,
+      "loss": 0.8118,
       "step": 14396
     },
     {
-      "epoch": 0.3954024882590426,
+      "epoch": 0.4085414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.3777677618416592e-05,
-      "loss": 0.8907,
+      "learning_rate": 1.3380443044974279e-05,
+      "loss": 0.8244,
       "step": 14397
     },
     {
-      "epoch": 0.39542995248688584,
+      "epoch": 0.40856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.3776854000256171e-05,
-      "loss": 0.8216,
+      "learning_rate": 1.3379578058574259e-05,
+      "loss": 0.9391,
       "step": 14398
     },
     {
-      "epoch": 0.39545741671472906,
+      "epoch": 0.40859818388195235,
       "grad_norm": 0.0,
-      "learning_rate": 1.3776030352211814e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.3378713043626788e-05,
+      "loss": 0.933,
       "step": 14399
     },
     {
-      "epoch": 0.3954848809425723,
+      "epoch": 0.4086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.3775206674290042e-05,
-      "loss": 0.947,
+      "learning_rate": 1.3377848000139178e-05,
+      "loss": 0.9721,
       "step": 14400
     },
     {
-      "epoch": 0.39551234517041556,
+      "epoch": 0.4086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.3774382966497375e-05,
-      "loss": 1.0174,
+      "learning_rate": 1.3376982928118734e-05,
+      "loss": 0.9763,
       "step": 14401
     },
     {
-      "epoch": 0.3955398093982588,
+      "epoch": 0.408683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.3773559228840327e-05,
-      "loss": 0.9911,
+      "learning_rate": 1.337611782757276e-05,
+      "loss": 0.9555,
       "step": 14402
     },
     {
-      "epoch": 0.395567273626102,
+      "epoch": 0.4087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.3772735461325416e-05,
-      "loss": 0.9748,
+      "learning_rate": 1.337525269850857e-05,
+      "loss": 0.9113,
       "step": 14403
     },
     {
-      "epoch": 0.3955947378539452,
+      "epoch": 0.4087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.377191166395916e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.337438754093347e-05,
+      "loss": 1.0229,
       "step": 14404
     },
     {
-      "epoch": 0.39562220208178844,
+      "epoch": 0.40876844494892167,
       "grad_norm": 0.0,
-      "learning_rate": 1.3771087836748079e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.3373522354854765e-05,
+      "loss": 1.0052,
       "step": 14405
     },
     {
-      "epoch": 0.3956496663096317,
+      "epoch": 0.4087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.377026397969869e-05,
-      "loss": 0.8029,
+      "learning_rate": 1.3372657140279763e-05,
+      "loss": 0.8274,
       "step": 14406
     },
     {
-      "epoch": 0.39567713053747494,
+      "epoch": 0.40882519863791145,
       "grad_norm": 0.0,
-      "learning_rate": 1.3769440092817515e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.3371791897215778e-05,
+      "loss": 0.9505,
       "step": 14407
     },
     {
-      "epoch": 0.39570459476531816,
+      "epoch": 0.40885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.3768616176111063e-05,
-      "loss": 0.8475,
+      "learning_rate": 1.3370926625670115e-05,
+      "loss": 1.0245,
       "step": 14408
     },
     {
-      "epoch": 0.3957320589931614,
+      "epoch": 0.4088819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.3767792229585863e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.3370061325650081e-05,
+      "loss": 0.859,
       "step": 14409
     },
     {
-      "epoch": 0.39575952322100466,
+      "epoch": 0.40891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.3766968253248436e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.3369195997162987e-05,
+      "loss": 0.8903,
       "step": 14410
     },
     {
-      "epoch": 0.3957869874488479,
+      "epoch": 0.40893870601589105,
       "grad_norm": 0.0,
-      "learning_rate": 1.3766144247105292e-05,
-      "loss": 0.9477,
+      "learning_rate": 1.3368330640216147e-05,
+      "loss": 0.9851,
       "step": 14411
     },
     {
-      "epoch": 0.3958144516766911,
+      "epoch": 0.4089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.3765320211162955e-05,
-      "loss": 0.9378,
+      "learning_rate": 1.3367465254816865e-05,
+      "loss": 0.8866,
       "step": 14412
     },
     {
-      "epoch": 0.3958419159045343,
+      "epoch": 0.4089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.376449614542795e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.336659984097245e-05,
+      "loss": 0.9112,
       "step": 14413
     },
     {
-      "epoch": 0.3958693801323776,
+      "epoch": 0.4090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.3763672049906792e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.3365734398690217e-05,
+      "loss": 0.8802,
       "step": 14414
     },
     {
-      "epoch": 0.3958968443602208,
+      "epoch": 0.4090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.3762847924606e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.3364868927977475e-05,
+      "loss": 0.9024,
       "step": 14415
     },
     {
-      "epoch": 0.39592430858806404,
+      "epoch": 0.4090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.37620237695321e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.3364003428841532e-05,
+      "loss": 0.9349,
       "step": 14416
     },
     {
-      "epoch": 0.39595177281590727,
+      "epoch": 0.4091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.376119958469161e-05,
-      "loss": 0.985,
+      "learning_rate": 1.3363137901289702e-05,
+      "loss": 0.9365,
       "step": 14417
     },
     {
-      "epoch": 0.39597923704375054,
+      "epoch": 0.4091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.3760375370091052e-05,
-      "loss": 0.7847,
+      "learning_rate": 1.3362272345329294e-05,
+      "loss": 0.938,
       "step": 14418
     },
     {
-      "epoch": 0.39600670127159376,
+      "epoch": 0.40916572077185015,
       "grad_norm": 0.0,
-      "learning_rate": 1.3759551125736949e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.336140676096762e-05,
+      "loss": 0.892,
       "step": 14419
     },
     {
-      "epoch": 0.396034165499437,
+      "epoch": 0.40919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.375872685163582e-05,
-      "loss": 0.9441,
+      "learning_rate": 1.3360541148211995e-05,
+      "loss": 0.9367,
       "step": 14420
     },
     {
-      "epoch": 0.3960616297272802,
+      "epoch": 0.40922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.3757902547794186e-05,
-      "loss": 0.8874,
+      "learning_rate": 1.3359675507069725e-05,
+      "loss": 0.8832,
       "step": 14421
     },
     {
-      "epoch": 0.39608909395512343,
+      "epoch": 0.40925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.3757078214218575e-05,
-      "loss": 0.987,
+      "learning_rate": 1.3358809837548128e-05,
+      "loss": 0.9391,
       "step": 14422
     },
     {
-      "epoch": 0.3961165581829667,
+      "epoch": 0.40927922814982975,
       "grad_norm": 0.0,
-      "learning_rate": 1.3756253850915505e-05,
-      "loss": 0.8367,
+      "learning_rate": 1.3357944139654508e-05,
+      "loss": 0.9272,
       "step": 14423
     },
     {
-      "epoch": 0.3961440224108099,
+      "epoch": 0.4093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.3755429457891498e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.3357078413396188e-05,
+      "loss": 0.9028,
       "step": 14424
     },
     {
-      "epoch": 0.39617148663865315,
+      "epoch": 0.40933598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 1.375460503515308e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.3356212658780471e-05,
+      "loss": 0.8757,
       "step": 14425
     },
     {
-      "epoch": 0.39619895086649637,
+      "epoch": 0.4093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.3753780582706772e-05,
-      "loss": 0.8942,
+      "learning_rate": 1.335534687581468e-05,
+      "loss": 1.042,
       "step": 14426
     },
     {
-      "epoch": 0.39622641509433965,
+      "epoch": 0.4093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.3752956100559096e-05,
-      "loss": 0.8177,
+      "learning_rate": 1.335448106450612e-05,
+      "loss": 0.9204,
       "step": 14427
     },
     {
-      "epoch": 0.39625387932218287,
+      "epoch": 0.4094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.3752131588716582e-05,
-      "loss": 0.8017,
+      "learning_rate": 1.3353615224862109e-05,
+      "loss": 0.9908,
       "step": 14428
     },
     {
-      "epoch": 0.3962813435500261,
+      "epoch": 0.4094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.3751307047185745e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.3352749356889957e-05,
+      "loss": 0.855,
       "step": 14429
     },
     {
-      "epoch": 0.3963088077778693,
+      "epoch": 0.409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.3750482475973115e-05,
-      "loss": 1.0144,
+      "learning_rate": 1.3351883460596984e-05,
+      "loss": 1.0352,
       "step": 14430
     },
     {
-      "epoch": 0.3963362720057126,
+      "epoch": 0.40950624290578885,
       "grad_norm": 0.0,
-      "learning_rate": 1.3749657875085215e-05,
-      "loss": 0.886,
+      "learning_rate": 1.3351017535990499e-05,
+      "loss": 0.9203,
       "step": 14431
     },
     {
-      "epoch": 0.3963637362335558,
+      "epoch": 0.40953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.374883324452857e-05,
-      "loss": 0.833,
+      "learning_rate": 1.335015158307782e-05,
+      "loss": 0.8952,
       "step": 14432
     },
     {
-      "epoch": 0.39639120046139903,
+      "epoch": 0.4095629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.3748008584309705e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.3349285601866254e-05,
+      "loss": 1.002,
       "step": 14433
     },
     {
-      "epoch": 0.39641866468924225,
+      "epoch": 0.40959137343927354,
       "grad_norm": 0.0,
-      "learning_rate": 1.3747183894435143e-05,
-      "loss": 0.8436,
+      "learning_rate": 1.3348419592363126e-05,
+      "loss": 0.887,
       "step": 14434
     },
     {
-      "epoch": 0.39644612891708547,
+      "epoch": 0.40961975028376846,
       "grad_norm": 0.0,
-      "learning_rate": 1.374635917491141e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.334755355457575e-05,
+      "loss": 0.8501,
       "step": 14435
     },
     {
-      "epoch": 0.39647359314492875,
+      "epoch": 0.4096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.3745534425745031e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.3346687488511436e-05,
+      "loss": 0.8925,
       "step": 14436
     },
     {
-      "epoch": 0.39650105737277197,
+      "epoch": 0.40967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.3744709646942535e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.3345821394177501e-05,
+      "loss": 0.92,
       "step": 14437
     },
     {
-      "epoch": 0.3965285216006152,
+      "epoch": 0.40970488081725315,
       "grad_norm": 0.0,
-      "learning_rate": 1.3743884838510446e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.3344955271581264e-05,
+      "loss": 0.8517,
       "step": 14438
     },
     {
-      "epoch": 0.3965559858284584,
+      "epoch": 0.409733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.3743060000455288e-05,
-      "loss": 0.9137,
+      "learning_rate": 1.334408912073004e-05,
+      "loss": 0.8753,
       "step": 14439
     },
     {
-      "epoch": 0.3965834500563017,
+      "epoch": 0.4097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.374223513278359e-05,
-      "loss": 0.9256,
+      "learning_rate": 1.3343222941631142e-05,
+      "loss": 0.9818,
       "step": 14440
     },
     {
-      "epoch": 0.3966109142841449,
+      "epoch": 0.4097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.3741410235501877e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.334235673429189e-05,
+      "loss": 0.8057,
       "step": 14441
     },
     {
-      "epoch": 0.39663837851198813,
+      "epoch": 0.4098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.3740585308616679e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.3341490498719605e-05,
+      "loss": 0.9117,
       "step": 14442
     },
     {
-      "epoch": 0.39666584273983135,
+      "epoch": 0.40984676503972756,
       "grad_norm": 0.0,
-      "learning_rate": 1.3739760352134521e-05,
-      "loss": 0.9073,
+      "learning_rate": 1.3340624234921592e-05,
+      "loss": 1.017,
       "step": 14443
     },
     {
-      "epoch": 0.39669330696767463,
+      "epoch": 0.40987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.373893536606193e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.3339757942905182e-05,
+      "loss": 0.9257,
       "step": 14444
     },
     {
-      "epoch": 0.39672077119551785,
+      "epoch": 0.4099035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.3738110350405433e-05,
-      "loss": 0.8769,
+      "learning_rate": 1.3338891622677688e-05,
+      "loss": 0.8585,
       "step": 14445
     },
     {
-      "epoch": 0.3967482354233611,
+      "epoch": 0.40993189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 1.3737285305171562e-05,
-      "loss": 0.8385,
+      "learning_rate": 1.333802527424642e-05,
+      "loss": 0.9175,
       "step": 14446
     },
     {
-      "epoch": 0.3967756996512043,
+      "epoch": 0.40996027241770716,
       "grad_norm": 0.0,
-      "learning_rate": 1.3736460230366838e-05,
-      "loss": 0.928,
+      "learning_rate": 1.3337158897618707e-05,
+      "loss": 0.8872,
       "step": 14447
     },
     {
-      "epoch": 0.3968031638790475,
+      "epoch": 0.409988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.3735635125997794e-05,
-      "loss": 0.8535,
+      "learning_rate": 1.333629249280186e-05,
+      "loss": 0.9811,
       "step": 14448
     },
     {
-      "epoch": 0.3968306281068908,
+      "epoch": 0.41001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.3734809992070959e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.3335426059803203e-05,
+      "loss": 0.8626,
       "step": 14449
     },
     {
-      "epoch": 0.396858092334734,
+      "epoch": 0.41004540295119185,
       "grad_norm": 0.0,
-      "learning_rate": 1.3733984828592861e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.333455959863005e-05,
+      "loss": 0.9377,
       "step": 14450
     },
     {
-      "epoch": 0.39688555656257724,
+      "epoch": 0.4100737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.3733159635570029e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.3333693109289722e-05,
+      "loss": 0.8159,
       "step": 14451
     },
     {
-      "epoch": 0.39691302079042046,
+      "epoch": 0.4101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.373233441300899e-05,
-      "loss": 1.0233,
+      "learning_rate": 1.3332826591789543e-05,
+      "loss": 0.945,
       "step": 14452
     },
     {
-      "epoch": 0.39694048501826373,
+      "epoch": 0.4101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.3731509160916274e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.3331960046136822e-05,
+      "loss": 1.0721,
       "step": 14453
     },
     {
-      "epoch": 0.39696794924610695,
+      "epoch": 0.4101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.3730683879298417e-05,
-      "loss": 0.948,
+      "learning_rate": 1.3331093472338888e-05,
+      "loss": 0.8862,
       "step": 14454
     },
     {
-      "epoch": 0.3969954134739502,
+      "epoch": 0.4101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.372985856816194e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.3330226870403058e-05,
+      "loss": 0.9097,
       "step": 14455
     },
     {
-      "epoch": 0.3970228777017934,
+      "epoch": 0.4102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.3729033227513377e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.3329360240336651e-05,
+      "loss": 0.934,
       "step": 14456
     },
     {
-      "epoch": 0.3970503419296367,
+      "epoch": 0.4102440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.372820785735926e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.3328493582146983e-05,
+      "loss": 1.012,
       "step": 14457
     },
     {
-      "epoch": 0.3970778061574799,
+      "epoch": 0.41027241770715095,
       "grad_norm": 0.0,
-      "learning_rate": 1.3727382457706123e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.3327626895841388e-05,
+      "loss": 0.8435,
       "step": 14458
     },
     {
-      "epoch": 0.3971052703853231,
+      "epoch": 0.41030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.3726557028560485e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.332676018142718e-05,
+      "loss": 0.8884,
       "step": 14459
     },
     {
-      "epoch": 0.39713273461316634,
+      "epoch": 0.4103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.3725731569928886e-05,
-      "loss": 0.8809,
+      "learning_rate": 1.3325893438911672e-05,
+      "loss": 0.8722,
       "step": 14460
     },
     {
-      "epoch": 0.39716019884100956,
+      "epoch": 0.41035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.3724906081817858e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.3325026668302197e-05,
+      "loss": 0.8713,
       "step": 14461
     },
     {
-      "epoch": 0.39718766306885284,
+      "epoch": 0.41038592508513055,
       "grad_norm": 0.0,
-      "learning_rate": 1.3724080564233928e-05,
-      "loss": 1.0242,
+      "learning_rate": 1.3324159869606072e-05,
+      "loss": 0.918,
       "step": 14462
     },
     {
-      "epoch": 0.39721512729669606,
+      "epoch": 0.4104143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.3723255017183632e-05,
-      "loss": 0.9124,
+      "learning_rate": 1.3323293042830617e-05,
+      "loss": 0.9203,
       "step": 14463
     },
     {
-      "epoch": 0.3972425915245393,
+      "epoch": 0.41044267877412033,
       "grad_norm": 0.0,
-      "learning_rate": 1.3722429440673497e-05,
-      "loss": 0.9218,
+      "learning_rate": 1.3322426187983158e-05,
+      "loss": 0.9441,
       "step": 14464
     },
     {
-      "epoch": 0.3972700557523825,
+      "epoch": 0.4104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.3721603834710057e-05,
-      "loss": 1.0039,
+      "learning_rate": 1.3321559305071014e-05,
+      "loss": 0.9295,
       "step": 14465
     },
     {
-      "epoch": 0.3972975199802258,
+      "epoch": 0.4104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.3720778199299853e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.3320692394101514e-05,
+      "loss": 0.8319,
       "step": 14466
     },
     {
-      "epoch": 0.397324984208069,
+      "epoch": 0.410527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.3719952534449405e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.3319825455081969e-05,
+      "loss": 1.0112,
       "step": 14467
     },
     {
-      "epoch": 0.3973524484359122,
+      "epoch": 0.4105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.3719126840165253e-05,
-      "loss": 0.9882,
+      "learning_rate": 1.3318958488019716e-05,
+      "loss": 0.9007,
       "step": 14468
     },
     {
-      "epoch": 0.39737991266375544,
+      "epoch": 0.4105845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.371830111645393e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.3318091492922068e-05,
+      "loss": 1.0247,
       "step": 14469
     },
     {
-      "epoch": 0.3974073768915987,
+      "epoch": 0.41061293984108965,
       "grad_norm": 0.0,
-      "learning_rate": 1.3717475363321966e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.331722446979635e-05,
+      "loss": 0.9653,
       "step": 14470
     },
     {
-      "epoch": 0.39743484111944194,
+      "epoch": 0.41064131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.3716649580775898e-05,
-      "loss": 0.9995,
+      "learning_rate": 1.331635741864989e-05,
+      "loss": 0.9288,
       "step": 14471
     },
     {
-      "epoch": 0.39746230534728516,
+      "epoch": 0.4106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.3715823768822261e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.3315490339490012e-05,
+      "loss": 0.9947,
       "step": 14472
     },
     {
-      "epoch": 0.3974897695751284,
+      "epoch": 0.41069807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.3714997927467582e-05,
-      "loss": 0.937,
+      "learning_rate": 1.3314623232324035e-05,
+      "loss": 0.9586,
       "step": 14473
     },
     {
-      "epoch": 0.3975172338029716,
+      "epoch": 0.41072644721906926,
       "grad_norm": 0.0,
-      "learning_rate": 1.3714172056718405e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.3313756097159288e-05,
+      "loss": 0.8536,
       "step": 14474
     },
     {
-      "epoch": 0.3975446980308149,
+      "epoch": 0.4107548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.3713346156581258e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.3312888934003095e-05,
+      "loss": 0.8097,
       "step": 14475
     },
     {
-      "epoch": 0.3975721622586581,
+      "epoch": 0.41078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 1.3712520227062678e-05,
-      "loss": 0.91,
+      "learning_rate": 1.3312021742862777e-05,
+      "loss": 0.9317,
       "step": 14476
     },
     {
-      "epoch": 0.3975996264865013,
+      "epoch": 0.4108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.3711694268169198e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.3311154523745663e-05,
+      "loss": 0.9924,
       "step": 14477
     },
     {
-      "epoch": 0.39762709071434454,
+      "epoch": 0.4108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.3710868279907359e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.3310287276659082e-05,
+      "loss": 0.8944,
       "step": 14478
     },
     {
-      "epoch": 0.3976545549421878,
+      "epoch": 0.4108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.3710042262283689e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.3309420001610352e-05,
+      "loss": 0.8625,
       "step": 14479
     },
     {
-      "epoch": 0.39768201917003104,
+      "epoch": 0.4108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.3709216215304728e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.3308552698606805e-05,
+      "loss": 0.8987,
       "step": 14480
     },
     {
-      "epoch": 0.39770948339787426,
+      "epoch": 0.4109250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.3708390138977014e-05,
-      "loss": 0.9171,
+      "learning_rate": 1.330768536765576e-05,
+      "loss": 0.928,
       "step": 14481
     },
     {
-      "epoch": 0.3977369476257175,
+      "epoch": 0.41095346197502836,
       "grad_norm": 0.0,
-      "learning_rate": 1.370756403330708e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.3306818008764553e-05,
+      "loss": 0.9952,
       "step": 14482
     },
     {
-      "epoch": 0.39776441185356076,
+      "epoch": 0.41098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.3706737898301461e-05,
-      "loss": 0.8905,
+      "learning_rate": 1.3305950621940504e-05,
+      "loss": 0.9083,
       "step": 14483
     },
     {
-      "epoch": 0.397791876081404,
+      "epoch": 0.4110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.3705911733966697e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.3305083207190941e-05,
+      "loss": 0.9471,
       "step": 14484
     },
     {
-      "epoch": 0.3978193403092472,
+      "epoch": 0.41103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.3705085540309325e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.3304215764523192e-05,
+      "loss": 0.8351,
       "step": 14485
     },
     {
-      "epoch": 0.3978468045370904,
+      "epoch": 0.41106696935300796,
       "grad_norm": 0.0,
-      "learning_rate": 1.3704259317335879e-05,
-      "loss": 0.8649,
+      "learning_rate": 1.3303348293944585e-05,
+      "loss": 0.8772,
       "step": 14486
     },
     {
-      "epoch": 0.39787426876493365,
+      "epoch": 0.4110953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.3703433065052903e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.3302480795462441e-05,
+      "loss": 0.8427,
       "step": 14487
     },
     {
-      "epoch": 0.3979017329927769,
+      "epoch": 0.41112372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.3702606783466922e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.3301613269084099e-05,
+      "loss": 0.8835,
       "step": 14488
     },
     {
-      "epoch": 0.39792919722062015,
+      "epoch": 0.4111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.3701780472584488e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.330074571481688e-05,
+      "loss": 0.8504,
       "step": 14489
     },
     {
-      "epoch": 0.39795666144846337,
+      "epoch": 0.4111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.3700954132412135e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.3299878132668114e-05,
+      "loss": 0.9599,
       "step": 14490
     },
     {
-      "epoch": 0.3979841256763066,
+      "epoch": 0.4112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.3700127762956394e-05,
-      "loss": 0.8724,
+      "learning_rate": 1.3299010522645129e-05,
+      "loss": 0.943,
       "step": 14491
     },
     {
-      "epoch": 0.39801158990414987,
+      "epoch": 0.4112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.3699301364223809e-05,
-      "loss": 0.8339,
+      "learning_rate": 1.3298142884755252e-05,
+      "loss": 0.9208,
       "step": 14492
     },
     {
-      "epoch": 0.3980390541319931,
+      "epoch": 0.4112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.369847493622092e-05,
-      "loss": 0.9578,
+      "learning_rate": 1.3297275219005817e-05,
+      "loss": 0.954,
       "step": 14493
     },
     {
-      "epoch": 0.3980665183598363,
+      "epoch": 0.41129398410896706,
       "grad_norm": 0.0,
-      "learning_rate": 1.3697648478954267e-05,
-      "loss": 0.965,
+      "learning_rate": 1.3296407525404145e-05,
+      "loss": 0.8791,
       "step": 14494
     },
     {
-      "epoch": 0.39809398258767953,
+      "epoch": 0.411322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.3696821992430383e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.3295539803957573e-05,
+      "loss": 0.8681,
       "step": 14495
     },
     {
-      "epoch": 0.3981214468155228,
+      "epoch": 0.4113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.3695995476655814e-05,
-      "loss": 0.8845,
+      "learning_rate": 1.329467205467343e-05,
+      "loss": 0.8946,
       "step": 14496
     },
     {
-      "epoch": 0.398148911043366,
+      "epoch": 0.41137911464245175,
       "grad_norm": 0.0,
-      "learning_rate": 1.3695168931637095e-05,
-      "loss": 0.8833,
+      "learning_rate": 1.3293804277559042e-05,
+      "loss": 0.8485,
       "step": 14497
     },
     {
-      "epoch": 0.39817637527120925,
+      "epoch": 0.41140749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 1.369434235738077e-05,
-      "loss": 0.861,
+      "learning_rate": 1.329293647262174e-05,
+      "loss": 0.844,
       "step": 14498
     },
     {
-      "epoch": 0.39820383949905247,
+      "epoch": 0.4114358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.3693515753893378e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.329206863986886e-05,
+      "loss": 0.9152,
       "step": 14499
     },
     {
-      "epoch": 0.3982313037268957,
+      "epoch": 0.41146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.3692689121181455e-05,
-      "loss": 0.9688,
+      "learning_rate": 1.3291200779307724e-05,
+      "loss": 0.8273,
       "step": 14500
     },
     {
-      "epoch": 0.39825876795473897,
+      "epoch": 0.41149262202043135,
       "grad_norm": 0.0,
-      "learning_rate": 1.3691862459251546e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.329033289094567e-05,
+      "loss": 0.9668,
       "step": 14501
     },
     {
-      "epoch": 0.3982862321825822,
+      "epoch": 0.4115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.3691035768110197e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.3289464974790022e-05,
+      "loss": 0.9908,
       "step": 14502
     },
     {
-      "epoch": 0.3983136964104254,
+      "epoch": 0.41154937570942113,
       "grad_norm": 0.0,
-      "learning_rate": 1.3690209047763938e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.328859703084812e-05,
+      "loss": 0.9591,
       "step": 14503
     },
     {
-      "epoch": 0.39834116063826863,
+      "epoch": 0.411577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.3689382298219315e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.3287729059127288e-05,
+      "loss": 0.9568,
       "step": 14504
     },
     {
-      "epoch": 0.3983686248661119,
+      "epoch": 0.4116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.368855551948287e-05,
-      "loss": 0.899,
+      "learning_rate": 1.3286861059634863e-05,
+      "loss": 1.031,
       "step": 14505
     },
     {
-      "epoch": 0.39839608909395513,
+      "epoch": 0.41163450624290576,
       "grad_norm": 0.0,
-      "learning_rate": 1.3687728711561148e-05,
-      "loss": 0.9795,
+      "learning_rate": 1.3285993032378175e-05,
+      "loss": 1.0084,
       "step": 14506
     },
     {
-      "epoch": 0.39842355332179835,
+      "epoch": 0.4116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.3686901874460688e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.3285124977364555e-05,
+      "loss": 0.8872,
       "step": 14507
     },
     {
-      "epoch": 0.3984510175496416,
+      "epoch": 0.4116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.368607500818803e-05,
-      "loss": 1.0366,
+      "learning_rate": 1.3284256894601337e-05,
+      "loss": 0.9468,
       "step": 14508
     },
     {
-      "epoch": 0.39847848177748485,
+      "epoch": 0.41171963677639045,
       "grad_norm": 0.0,
-      "learning_rate": 1.3685248112749716e-05,
-      "loss": 1.0119,
+      "learning_rate": 1.3283388784095854e-05,
+      "loss": 0.9204,
       "step": 14509
     },
     {
-      "epoch": 0.39850594600532807,
+      "epoch": 0.41174801362088537,
       "grad_norm": 0.0,
-      "learning_rate": 1.3684421188152299e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.3282520645855436e-05,
+      "loss": 0.934,
       "step": 14510
     },
     {
-      "epoch": 0.3985334102331713,
+      "epoch": 0.4117763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.3683594234402308e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.328165247988742e-05,
+      "loss": 0.8409,
       "step": 14511
     },
     {
-      "epoch": 0.3985608744610145,
+      "epoch": 0.41180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.3682767251506295e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.3280784286199137e-05,
+      "loss": 0.9083,
       "step": 14512
     },
     {
-      "epoch": 0.39858833868885774,
+      "epoch": 0.41183314415437006,
       "grad_norm": 0.0,
-      "learning_rate": 1.3681940239470801e-05,
-      "loss": 1.0217,
+      "learning_rate": 1.3279916064797924e-05,
+      "loss": 0.9088,
       "step": 14513
     },
     {
-      "epoch": 0.398615802916701,
+      "epoch": 0.4118615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.3681113198302369e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.327904781569111e-05,
+      "loss": 0.8962,
       "step": 14514
     },
     {
-      "epoch": 0.39864326714454423,
+      "epoch": 0.41188989784335983,
       "grad_norm": 0.0,
-      "learning_rate": 1.3680286128007543e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.3278179538886034e-05,
+      "loss": 0.9481,
       "step": 14515
     },
     {
-      "epoch": 0.39867073137238745,
+      "epoch": 0.4119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.3679459028592867e-05,
-      "loss": 0.9143,
+      "learning_rate": 1.327731123439003e-05,
+      "loss": 0.9442,
       "step": 14516
     },
     {
-      "epoch": 0.3986981956002307,
+      "epoch": 0.4119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.3678631900064887e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.3276442902210425e-05,
+      "loss": 0.996,
       "step": 14517
     },
     {
-      "epoch": 0.39872565982807395,
+      "epoch": 0.4119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.3677804742430147e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.3275574542354562e-05,
+      "loss": 0.8873,
       "step": 14518
     },
     {
-      "epoch": 0.3987531240559172,
+      "epoch": 0.4120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.3676977555695193e-05,
-      "loss": 0.8369,
+      "learning_rate": 1.3274706154829772e-05,
+      "loss": 0.866,
       "step": 14519
     },
     {
-      "epoch": 0.3987805882837604,
+      "epoch": 0.4120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.3676150339866562e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.3273837739643396e-05,
+      "loss": 0.8191,
       "step": 14520
     },
     {
-      "epoch": 0.3988080525116036,
+      "epoch": 0.41206015891032916,
       "grad_norm": 0.0,
-      "learning_rate": 1.367532309495081e-05,
-      "loss": 0.9217,
+      "learning_rate": 1.3272969296802762e-05,
+      "loss": 0.9943,
       "step": 14521
     },
     {
-      "epoch": 0.3988355167394469,
+      "epoch": 0.41208853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.367449582095448e-05,
-      "loss": 0.9579,
+      "learning_rate": 1.327210082631521e-05,
+      "loss": 0.993,
       "step": 14522
     },
     {
-      "epoch": 0.3988629809672901,
+      "epoch": 0.41211691259931893,
       "grad_norm": 0.0,
-      "learning_rate": 1.3673668517884111e-05,
-      "loss": 0.8461,
+      "learning_rate": 1.3271232328188077e-05,
+      "loss": 0.9843,
       "step": 14523
     },
     {
-      "epoch": 0.39889044519513334,
+      "epoch": 0.41214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.3672841185746256e-05,
-      "loss": 0.903,
+      "learning_rate": 1.3270363802428697e-05,
+      "loss": 0.8658,
       "step": 14524
     },
     {
-      "epoch": 0.39891790942297656,
+      "epoch": 0.41217366628830876,
       "grad_norm": 0.0,
-      "learning_rate": 1.367201382454746e-05,
-      "loss": 1.0071,
+      "learning_rate": 1.3269495249044404e-05,
+      "loss": 1.0643,
       "step": 14525
     },
     {
-      "epoch": 0.39894537365081983,
+      "epoch": 0.4122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.3671186434294265e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.3268626668042539e-05,
+      "loss": 0.9421,
       "step": 14526
     },
     {
-      "epoch": 0.39897283787866306,
+      "epoch": 0.41223041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.3670359014993222e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.326775805943044e-05,
+      "loss": 0.8632,
       "step": 14527
     },
     {
-      "epoch": 0.3990003021065063,
+      "epoch": 0.4122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.3669531566650878e-05,
-      "loss": 0.873,
+      "learning_rate": 1.3266889423215438e-05,
+      "loss": 0.9599,
       "step": 14528
     },
     {
-      "epoch": 0.3990277663343495,
+      "epoch": 0.4122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.3668704089273777e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.3266020759404878e-05,
+      "loss": 0.8907,
       "step": 14529
     },
     {
-      "epoch": 0.3990552305621927,
+      "epoch": 0.4123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.3667876582868471e-05,
-      "loss": 0.8893,
+      "learning_rate": 1.3265152068006093e-05,
+      "loss": 1.0132,
       "step": 14530
     },
     {
-      "epoch": 0.399082694790036,
+      "epoch": 0.4123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.3667049047441503e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.3264283349026419e-05,
+      "loss": 0.8474,
       "step": 14531
     },
     {
-      "epoch": 0.3991101590178792,
+      "epoch": 0.412372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.3666221482999422e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.3263414602473201e-05,
+      "loss": 0.9416,
       "step": 14532
     },
     {
-      "epoch": 0.39913762324572244,
+      "epoch": 0.41240068104426786,
       "grad_norm": 0.0,
-      "learning_rate": 1.3665393889548776e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.326254582835377e-05,
+      "loss": 0.8662,
       "step": 14533
     },
     {
-      "epoch": 0.39916508747356566,
+      "epoch": 0.4124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.3664566267096117e-05,
-      "loss": 0.9014,
+      "learning_rate": 1.3261677026675467e-05,
+      "loss": 0.9548,
       "step": 14534
     },
     {
-      "epoch": 0.39919255170140894,
+      "epoch": 0.4124574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.3663738615647986e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.3260808197445632e-05,
+      "loss": 1.0189,
       "step": 14535
     },
     {
-      "epoch": 0.39922001592925216,
+      "epoch": 0.41248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.3662910935210935e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.3259939340671606e-05,
+      "loss": 0.8415,
       "step": 14536
     },
     {
-      "epoch": 0.3992474801570954,
+      "epoch": 0.41251418842224746,
       "grad_norm": 0.0,
-      "learning_rate": 1.3662083225791519e-05,
-      "loss": 0.8291,
+      "learning_rate": 1.3259070456360726e-05,
+      "loss": 0.8702,
       "step": 14537
     },
     {
-      "epoch": 0.3992749443849386,
+      "epoch": 0.4125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.366125548739628e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.3258201544520328e-05,
+      "loss": 0.9791,
       "step": 14538
     },
     {
-      "epoch": 0.3993024086127819,
+      "epoch": 0.41257094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.3660427720031767e-05,
-      "loss": 0.9859,
+      "learning_rate": 1.3257332605157758e-05,
+      "loss": 0.9278,
       "step": 14539
     },
     {
-      "epoch": 0.3993298728406251,
+      "epoch": 0.4125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.3659599923704534e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.325646363828035e-05,
+      "loss": 0.91,
       "step": 14540
     },
     {
-      "epoch": 0.3993573370684683,
+      "epoch": 0.412627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.3658772098421128e-05,
-      "loss": 0.9181,
+      "learning_rate": 1.3255594643895448e-05,
+      "loss": 0.8247,
       "step": 14541
     },
     {
-      "epoch": 0.39938480129631154,
+      "epoch": 0.4126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.3657944244188102e-05,
-      "loss": 0.8566,
+      "learning_rate": 1.3254725622010393e-05,
+      "loss": 0.8215,
       "step": 14542
     },
     {
-      "epoch": 0.39941226552415476,
+      "epoch": 0.4126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.3657116361012002e-05,
-      "loss": 0.946,
+      "learning_rate": 1.3253856572632526e-05,
+      "loss": 0.8915,
       "step": 14543
     },
     {
-      "epoch": 0.39943972975199804,
+      "epoch": 0.4127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.365628844889938e-05,
-      "loss": 0.8582,
+      "learning_rate": 1.3252987495769185e-05,
+      "loss": 0.9695,
       "step": 14544
     },
     {
-      "epoch": 0.39946719397984126,
+      "epoch": 0.41274120317820656,
       "grad_norm": 0.0,
-      "learning_rate": 1.3655460507856788e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.325211839142771e-05,
+      "loss": 0.9251,
       "step": 14545
     },
     {
-      "epoch": 0.3994946582076845,
+      "epoch": 0.4127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.3654632537890779e-05,
-      "loss": 0.8479,
+      "learning_rate": 1.3251249259615448e-05,
+      "loss": 0.7983,
       "step": 14546
     },
     {
-      "epoch": 0.3995221224355277,
+      "epoch": 0.4127979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.36538045390079e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.3250380100339738e-05,
+      "loss": 0.9892,
       "step": 14547
     },
     {
-      "epoch": 0.399549586663371,
+      "epoch": 0.41282633371169125,
       "grad_norm": 0.0,
-      "learning_rate": 1.3652976511214701e-05,
-      "loss": 0.937,
+      "learning_rate": 1.3249510913607916e-05,
+      "loss": 0.9733,
       "step": 14548
     },
     {
-      "epoch": 0.3995770508912142,
+      "epoch": 0.41285471055618617,
       "grad_norm": 0.0,
-      "learning_rate": 1.3652148454517741e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.3248641699427332e-05,
+      "loss": 0.9302,
       "step": 14549
     },
     {
-      "epoch": 0.3996045151190574,
+      "epoch": 0.412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.3651320368923568e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.3247772457805328e-05,
+      "loss": 0.9943,
       "step": 14550
     },
     {
-      "epoch": 0.39963197934690065,
+      "epoch": 0.41291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.3650492254438732e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.324690318874924e-05,
+      "loss": 0.8909,
       "step": 14551
     },
     {
-      "epoch": 0.3996594435747439,
+      "epoch": 0.41293984108967086,
       "grad_norm": 0.0,
-      "learning_rate": 1.364966411106979e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.324603389226642e-05,
+      "loss": 0.8736,
       "step": 14552
     },
     {
-      "epoch": 0.39968690780258714,
+      "epoch": 0.4129682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.3648835938823287e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.3245164568364201e-05,
+      "loss": 0.8486,
       "step": 14553
     },
     {
-      "epoch": 0.39971437203043036,
+      "epoch": 0.41299659477866063,
       "grad_norm": 0.0,
-      "learning_rate": 1.3648007737705784e-05,
-      "loss": 0.8859,
+      "learning_rate": 1.3244295217049932e-05,
+      "loss": 0.86,
       "step": 14554
     },
     {
-      "epoch": 0.3997418362582736,
+      "epoch": 0.4130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.3647179507723828e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.3243425838330957e-05,
+      "loss": 0.999,
       "step": 14555
     },
     {
-      "epoch": 0.3997693004861168,
+      "epoch": 0.4130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.3646351248883978e-05,
-      "loss": 0.9028,
+      "learning_rate": 1.3242556432214615e-05,
+      "loss": 0.9606,
       "step": 14556
     },
     {
-      "epoch": 0.3997967647139601,
+      "epoch": 0.41308172531214526,
       "grad_norm": 0.0,
-      "learning_rate": 1.364552296119278e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.3241686998708256e-05,
+      "loss": 0.8739,
       "step": 14557
     },
     {
-      "epoch": 0.3998242289418033,
+      "epoch": 0.4131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.36446946446568e-05,
-      "loss": 0.9465,
+      "learning_rate": 1.324081753781922e-05,
+      "loss": 0.8925,
       "step": 14558
     },
     {
-      "epoch": 0.3998516931696465,
+      "epoch": 0.4131384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.3643866299282578e-05,
-      "loss": 1.0053,
+      "learning_rate": 1.3239948049554854e-05,
+      "loss": 0.887,
       "step": 14559
     },
     {
-      "epoch": 0.39987915739748975,
+      "epoch": 0.41316685584562995,
       "grad_norm": 0.0,
-      "learning_rate": 1.3643037925076675e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.32390785339225e-05,
+      "loss": 0.8687,
       "step": 14560
     },
     {
-      "epoch": 0.399906621625333,
+      "epoch": 0.41319523269012487,
       "grad_norm": 0.0,
-      "learning_rate": 1.3642209522045646e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.3238208990929505e-05,
+      "loss": 0.9123,
       "step": 14561
     },
     {
-      "epoch": 0.39993408585317625,
+      "epoch": 0.41322360953461973,
       "grad_norm": 0.0,
-      "learning_rate": 1.3641381090196047e-05,
-      "loss": 0.9071,
+      "learning_rate": 1.3237339420583213e-05,
+      "loss": 0.9421,
       "step": 14562
     },
     {
-      "epoch": 0.39996155008101947,
+      "epoch": 0.41325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.3640552629534426e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.3236469822890968e-05,
+      "loss": 0.9296,
       "step": 14563
     },
     {
-      "epoch": 0.3999890143088627,
+      "epoch": 0.41328036322360956,
       "grad_norm": 0.0,
-      "learning_rate": 1.3639724140067343e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.3235600197860116e-05,
+      "loss": 0.9183,
       "step": 14564
     },
     {
-      "epoch": 0.40001647853670597,
+      "epoch": 0.4133087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.3638895621801351e-05,
-      "loss": 0.8876,
+      "learning_rate": 1.3234730545498004e-05,
+      "loss": 0.8517,
       "step": 14565
     },
     {
-      "epoch": 0.4000439427645492,
+      "epoch": 0.41333711691259933,
       "grad_norm": 0.0,
-      "learning_rate": 1.3638067074743013e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.323386086581198e-05,
+      "loss": 1.0568,
       "step": 14566
     },
     {
-      "epoch": 0.4000714069923924,
+      "epoch": 0.4133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.3637238498898875e-05,
-      "loss": 0.9683,
+      "learning_rate": 1.323299115880939e-05,
+      "loss": 0.9694,
       "step": 14567
     },
     {
-      "epoch": 0.40009887122023563,
+      "epoch": 0.4133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.3636409894275495e-05,
-      "loss": 0.8966,
+      "learning_rate": 1.3232121424497572e-05,
+      "loss": 0.9164,
       "step": 14568
     },
     {
-      "epoch": 0.40012633544807885,
+      "epoch": 0.41342224744608397,
       "grad_norm": 0.0,
-      "learning_rate": 1.3635581260879436e-05,
-      "loss": 0.8762,
+      "learning_rate": 1.3231251662883884e-05,
+      "loss": 0.829,
       "step": 14569
     },
     {
-      "epoch": 0.40015379967592213,
+      "epoch": 0.4134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.3634752598717249e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.3230381873975667e-05,
+      "loss": 1.0267,
       "step": 14570
     },
     {
-      "epoch": 0.40018126390376535,
+      "epoch": 0.4134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.363392390779549e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.3229512057780267e-05,
+      "loss": 0.9698,
       "step": 14571
     },
     {
-      "epoch": 0.40020872813160857,
+      "epoch": 0.41350737797956866,
       "grad_norm": 0.0,
-      "learning_rate": 1.3633095188120718e-05,
-      "loss": 0.9523,
+      "learning_rate": 1.3228642214305033e-05,
+      "loss": 0.9295,
       "step": 14572
     },
     {
-      "epoch": 0.4002361923594518,
+      "epoch": 0.4135357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.3632266439699489e-05,
-      "loss": 0.9105,
+      "learning_rate": 1.3227772343557316e-05,
+      "loss": 0.9429,
       "step": 14573
     },
     {
-      "epoch": 0.40026365658729507,
+      "epoch": 0.41356413166855843,
       "grad_norm": 0.0,
-      "learning_rate": 1.363143766253836e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.322690244554446e-05,
+      "loss": 0.9011,
       "step": 14574
     },
     {
-      "epoch": 0.4002911208151383,
+      "epoch": 0.41359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.3630608856643893e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.3226032520273812e-05,
+      "loss": 0.9344,
       "step": 14575
     },
     {
-      "epoch": 0.4003185850429815,
+      "epoch": 0.41362088535754826,
       "grad_norm": 0.0,
-      "learning_rate": 1.3629780022022638e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.3225162567752725e-05,
+      "loss": 0.9391,
       "step": 14576
     },
     {
-      "epoch": 0.40034604927082473,
+      "epoch": 0.4136492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.3628951158681159e-05,
-      "loss": 0.8841,
+      "learning_rate": 1.3224292587988545e-05,
+      "loss": 0.869,
       "step": 14577
     },
     {
-      "epoch": 0.400373513498668,
+      "epoch": 0.41367763904653804,
       "grad_norm": 0.0,
-      "learning_rate": 1.3628122266626017e-05,
-      "loss": 1.0032,
+      "learning_rate": 1.322342258098862e-05,
+      "loss": 0.8201,
       "step": 14578
     },
     {
-      "epoch": 0.40040097772651123,
+      "epoch": 0.4137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.362729334586376e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.3222552546760297e-05,
+      "loss": 1.0504,
       "step": 14579
     },
     {
-      "epoch": 0.40042844195435445,
+      "epoch": 0.4137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.3626464396400954e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.3221682485310929e-05,
+      "loss": 1.0353,
       "step": 14580
     },
     {
-      "epoch": 0.4004559061821977,
+      "epoch": 0.4137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.3625635418244159e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.3220812396647867e-05,
+      "loss": 0.8658,
       "step": 14581
     },
     {
-      "epoch": 0.4004833704100409,
+      "epoch": 0.4137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.3624806411399933e-05,
-      "loss": 1.025,
+      "learning_rate": 1.3219942280778456e-05,
+      "loss": 0.8345,
       "step": 14582
     },
     {
-      "epoch": 0.40051083463788417,
+      "epoch": 0.4138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.3623977375874834e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.3219072137710048e-05,
+      "loss": 1.031,
       "step": 14583
     },
     {
-      "epoch": 0.4005382988657274,
+      "epoch": 0.41384790011350736,
       "grad_norm": 0.0,
-      "learning_rate": 1.362314831167542e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.3218201967449993e-05,
+      "loss": 0.9769,
       "step": 14584
     },
     {
-      "epoch": 0.4005657630935706,
+      "epoch": 0.4138762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.3622319218808253e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.3217331770005639e-05,
+      "loss": 0.9029,
       "step": 14585
     },
     {
-      "epoch": 0.40059322732141384,
+      "epoch": 0.41390465380249714,
       "grad_norm": 0.0,
-      "learning_rate": 1.3621490097279895e-05,
-      "loss": 0.905,
+      "learning_rate": 1.321646154538434e-05,
+      "loss": 0.9644,
       "step": 14586
     },
     {
-      "epoch": 0.4006206915492571,
+      "epoch": 0.41393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.3620660947096904e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.3215591293593449e-05,
+      "loss": 0.9341,
       "step": 14587
     },
     {
-      "epoch": 0.40064815577710033,
+      "epoch": 0.41396140749148697,
       "grad_norm": 0.0,
-      "learning_rate": 1.3619831768265842e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.321472101464031e-05,
+      "loss": 1.0121,
       "step": 14588
     },
     {
-      "epoch": 0.40067562000494356,
+      "epoch": 0.4139897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.3619002560793267e-05,
-      "loss": 0.9853,
+      "learning_rate": 1.3213850708532278e-05,
+      "loss": 0.9025,
       "step": 14589
     },
     {
-      "epoch": 0.4007030842327868,
+      "epoch": 0.41401816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 1.3618173324685744e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.3212980375276708e-05,
+      "loss": 0.9892,
       "step": 14590
     },
     {
-      "epoch": 0.40073054846063005,
+      "epoch": 0.4140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.361734405994983e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.3212110014880945e-05,
+      "loss": 0.9677,
       "step": 14591
     },
     {
-      "epoch": 0.4007580126884733,
+      "epoch": 0.4140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.3616514766592087e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.3211239627352345e-05,
+      "loss": 0.9195,
       "step": 14592
     },
     {
-      "epoch": 0.4007854769163165,
+      "epoch": 0.41410329171396143,
       "grad_norm": 0.0,
-      "learning_rate": 1.3615685444619083e-05,
-      "loss": 0.8121,
+      "learning_rate": 1.3210369212698262e-05,
+      "loss": 0.8575,
       "step": 14593
     },
     {
-      "epoch": 0.4008129411441597,
+      "epoch": 0.4141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.3614856094037375e-05,
-      "loss": 1.0622,
+      "learning_rate": 1.3209498770926043e-05,
+      "loss": 0.9605,
       "step": 14594
     },
     {
-      "epoch": 0.40084040537200294,
+      "epoch": 0.4141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.3614026714853522e-05,
-      "loss": 1.0061,
+      "learning_rate": 1.3208628302043042e-05,
+      "loss": 1.0013,
       "step": 14595
     },
     {
-      "epoch": 0.4008678695998462,
+      "epoch": 0.41418842224744606,
       "grad_norm": 0.0,
-      "learning_rate": 1.3613197307074091e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.3207757806056619e-05,
+      "loss": 0.9097,
       "step": 14596
     },
     {
-      "epoch": 0.40089533382768944,
+      "epoch": 0.414216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.3612367870705643e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.320688728297412e-05,
+      "loss": 0.834,
       "step": 14597
     },
     {
-      "epoch": 0.40092279805553266,
+      "epoch": 0.4142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.361153840575474e-05,
-      "loss": 1.0256,
+      "learning_rate": 1.3206016732802898e-05,
+      "loss": 1.0315,
       "step": 14598
     },
     {
-      "epoch": 0.4009502622833759,
+      "epoch": 0.41427355278093075,
       "grad_norm": 0.0,
-      "learning_rate": 1.361070891222795e-05,
-      "loss": 0.8719,
+      "learning_rate": 1.3205146155550309e-05,
+      "loss": 0.8726,
       "step": 14599
     },
     {
-      "epoch": 0.40097772651121916,
+      "epoch": 0.41430192962542567,
       "grad_norm": 0.0,
-      "learning_rate": 1.3609879390131828e-05,
-      "loss": 0.9883,
+      "learning_rate": 1.3204275551223709e-05,
+      "loss": 1.0088,
       "step": 14600
     },
     {
-      "epoch": 0.4010051907390624,
+      "epoch": 0.41433030646992053,
       "grad_norm": 0.0,
-      "learning_rate": 1.3609049839472943e-05,
-      "loss": 0.947,
+      "learning_rate": 1.3203404919830447e-05,
+      "loss": 0.9359,
       "step": 14601
     },
     {
-      "epoch": 0.4010326549669056,
+      "epoch": 0.41435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.360822026025786e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.3202534261377875e-05,
+      "loss": 0.8907,
       "step": 14602
     },
     {
-      "epoch": 0.4010601191947488,
+      "epoch": 0.4143870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.3607390652493139e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.3201663575873359e-05,
+      "loss": 0.874,
       "step": 14603
     },
     {
-      "epoch": 0.4010875834225921,
+      "epoch": 0.4144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.3606561016185344e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.3200792863324246e-05,
+      "loss": 0.7891,
       "step": 14604
     },
     {
-      "epoch": 0.4011150476504353,
+      "epoch": 0.41444381384790013,
       "grad_norm": 0.0,
-      "learning_rate": 1.3605731351341044e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.3199922123737887e-05,
+      "loss": 0.9456,
       "step": 14605
     },
     {
-      "epoch": 0.40114251187827854,
+      "epoch": 0.414472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.3604901657966803e-05,
-      "loss": 0.8692,
+      "learning_rate": 1.3199051357121647e-05,
+      "loss": 0.9782,
       "step": 14606
     },
     {
-      "epoch": 0.40116997610612176,
+      "epoch": 0.4145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.360407193606918e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.3198180563482877e-05,
+      "loss": 1.0405,
       "step": 14607
     },
     {
-      "epoch": 0.401197440333965,
+      "epoch": 0.41452894438138477,
       "grad_norm": 0.0,
-      "learning_rate": 1.3603242185654743e-05,
-      "loss": 0.8382,
+      "learning_rate": 1.3197309742828927e-05,
+      "loss": 0.9145,
       "step": 14608
     },
     {
-      "epoch": 0.40122490456180826,
+      "epoch": 0.4145573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.360241240673006e-05,
-      "loss": 0.8848,
+      "learning_rate": 1.319643889516716e-05,
+      "loss": 0.8445,
       "step": 14609
     },
     {
-      "epoch": 0.4012523687896515,
+      "epoch": 0.4145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.3601582599301693e-05,
-      "loss": 0.8503,
+      "learning_rate": 1.3195568020504932e-05,
+      "loss": 1.049,
       "step": 14610
     },
     {
-      "epoch": 0.4012798330174947,
+      "epoch": 0.41461407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.3600752763376213e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.3194697118849598e-05,
+      "loss": 0.9107,
       "step": 14611
     },
     {
-      "epoch": 0.4013072972453379,
+      "epoch": 0.41464245175936437,
       "grad_norm": 0.0,
-      "learning_rate": 1.3599922898960177e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.3193826190208507e-05,
+      "loss": 0.8927,
       "step": 14612
     },
     {
-      "epoch": 0.4013347614731812,
+      "epoch": 0.41467082860385923,
       "grad_norm": 0.0,
-      "learning_rate": 1.3599093006060157e-05,
-      "loss": 0.9988,
+      "learning_rate": 1.3192955234589028e-05,
+      "loss": 0.983,
       "step": 14613
     },
     {
-      "epoch": 0.4013622257010244,
+      "epoch": 0.41469920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.3598263084682726e-05,
-      "loss": 0.8737,
+      "learning_rate": 1.3192084251998515e-05,
+      "loss": 0.8817,
       "step": 14614
     },
     {
-      "epoch": 0.40138968992886764,
+      "epoch": 0.41472758229284906,
       "grad_norm": 0.0,
-      "learning_rate": 1.3597433134834436e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.3191213242444318e-05,
+      "loss": 0.9825,
       "step": 14615
     },
     {
-      "epoch": 0.40141715415671086,
+      "epoch": 0.4147559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.3596603156521866e-05,
-      "loss": 0.881,
+      "learning_rate": 1.31903422059338e-05,
+      "loss": 0.8434,
       "step": 14616
     },
     {
-      "epoch": 0.40144461838455414,
+      "epoch": 0.41478433598183884,
       "grad_norm": 0.0,
-      "learning_rate": 1.3595773149751577e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.318947114247432e-05,
+      "loss": 0.9081,
       "step": 14617
     },
     {
-      "epoch": 0.40147208261239736,
+      "epoch": 0.4148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.3594943114530139e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.3188600052073235e-05,
+      "loss": 1.0483,
       "step": 14618
     },
     {
-      "epoch": 0.4014995468402406,
+      "epoch": 0.4148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.3594113050864115e-05,
-      "loss": 1.0738,
+      "learning_rate": 1.3187728934737897e-05,
+      "loss": 0.8585,
       "step": 14619
     },
     {
-      "epoch": 0.4015270110680838,
+      "epoch": 0.41486946651532347,
       "grad_norm": 0.0,
-      "learning_rate": 1.3593282958760081e-05,
-      "loss": 0.8013,
+      "learning_rate": 1.3186857790475673e-05,
+      "loss": 0.9467,
       "step": 14620
     },
     {
-      "epoch": 0.4015544752959271,
+      "epoch": 0.4148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.3592452838224597e-05,
-      "loss": 0.9494,
+      "learning_rate": 1.318598661929392e-05,
+      "loss": 0.8167,
       "step": 14621
     },
     {
-      "epoch": 0.4015819395237703,
+      "epoch": 0.4149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.3591622689264237e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.318511542119999e-05,
+      "loss": 0.9128,
       "step": 14622
     },
     {
-      "epoch": 0.4016094037516135,
+      "epoch": 0.41495459704880816,
       "grad_norm": 0.0,
-      "learning_rate": 1.3590792511885566e-05,
-      "loss": 1.0724,
+      "learning_rate": 1.318424419620125e-05,
+      "loss": 0.9253,
       "step": 14623
     },
     {
-      "epoch": 0.40163686797945675,
+      "epoch": 0.4149829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.3589962306095152e-05,
-      "loss": 0.951,
+      "learning_rate": 1.3183372944305055e-05,
+      "loss": 0.8444,
       "step": 14624
     },
     {
-      "epoch": 0.40166433220729997,
+      "epoch": 0.41501135073779793,
       "grad_norm": 0.0,
-      "learning_rate": 1.358913207189957e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.3182501665518766e-05,
+      "loss": 0.8403,
       "step": 14625
     },
     {
-      "epoch": 0.40169179643514324,
+      "epoch": 0.41503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.358830180930538e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.318163035984974e-05,
+      "loss": 0.968,
       "step": 14626
     },
     {
-      "epoch": 0.40171926066298647,
+      "epoch": 0.41506810442678777,
       "grad_norm": 0.0,
-      "learning_rate": 1.3587471518319158e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.3180759027305342e-05,
+      "loss": 0.9355,
       "step": 14627
     },
     {
-      "epoch": 0.4017467248908297,
+      "epoch": 0.4150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.3586641198947472e-05,
-      "loss": 0.914,
+      "learning_rate": 1.317988766789293e-05,
+      "loss": 0.9164,
       "step": 14628
     },
     {
-      "epoch": 0.4017741891186729,
+      "epoch": 0.41512485811577754,
       "grad_norm": 0.0,
-      "learning_rate": 1.358581085119689e-05,
-      "loss": 0.8054,
+      "learning_rate": 1.317901628161986e-05,
+      "loss": 0.8456,
       "step": 14629
     },
     {
-      "epoch": 0.4018016533465162,
+      "epoch": 0.4151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.3584980475073985e-05,
-      "loss": 0.858,
+      "learning_rate": 1.3178144868493502e-05,
+      "loss": 0.9249,
       "step": 14630
     },
     {
-      "epoch": 0.4018291175743594,
+      "epoch": 0.4151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.3584150070585327e-05,
-      "loss": 0.8956,
+      "learning_rate": 1.3177273428521211e-05,
+      "loss": 0.8504,
       "step": 14631
     },
     {
-      "epoch": 0.40185658180220263,
+      "epoch": 0.41520998864926223,
       "grad_norm": 0.0,
-      "learning_rate": 1.3583319637737482e-05,
-      "loss": 0.989,
+      "learning_rate": 1.3176401961710343e-05,
+      "loss": 0.9502,
       "step": 14632
     },
     {
-      "epoch": 0.40188404603004585,
+      "epoch": 0.4152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.3582489176537025e-05,
-      "loss": 0.9408,
+      "learning_rate": 1.3175530468068267e-05,
+      "loss": 0.9315,
       "step": 14633
     },
     {
-      "epoch": 0.4019115102578891,
+      "epoch": 0.415266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.3581658686990528e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.3174658947602346e-05,
+      "loss": 0.8885,
       "step": 14634
     },
     {
-      "epoch": 0.40193897448573235,
+      "epoch": 0.41529511918274686,
       "grad_norm": 0.0,
-      "learning_rate": 1.3580828169104558e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.3173787400319938e-05,
+      "loss": 1.0204,
       "step": 14635
     },
     {
-      "epoch": 0.40196643871357557,
+      "epoch": 0.4153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.357999762288569e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.3172915826228398e-05,
+      "loss": 0.9695,
       "step": 14636
     },
     {
-      "epoch": 0.4019939029414188,
+      "epoch": 0.41535187287173664,
       "grad_norm": 0.0,
-      "learning_rate": 1.3579167048340494e-05,
-      "loss": 1.0094,
+      "learning_rate": 1.3172044225335103e-05,
+      "loss": 0.8114,
       "step": 14637
     },
     {
-      "epoch": 0.402021367169262,
+      "epoch": 0.41538024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.3578336445475542e-05,
-      "loss": 0.944,
+      "learning_rate": 1.3171172597647407e-05,
+      "loss": 0.8145,
       "step": 14638
     },
     {
-      "epoch": 0.4020488313971053,
+      "epoch": 0.41540862656072647,
       "grad_norm": 0.0,
-      "learning_rate": 1.3577505814297403e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.3170300943172669e-05,
+      "loss": 0.9421,
       "step": 14639
     },
     {
-      "epoch": 0.4020762956249485,
+      "epoch": 0.41543700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.3576675154812656e-05,
-      "loss": 0.8688,
+      "learning_rate": 1.3169429261918258e-05,
+      "loss": 1.0016,
       "step": 14640
     },
     {
-      "epoch": 0.40210375985279173,
+      "epoch": 0.41546538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.3575844467027865e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.316855755389154e-05,
+      "loss": 0.8801,
       "step": 14641
     },
     {
-      "epoch": 0.40213122408063495,
+      "epoch": 0.4154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.3575013750949613e-05,
-      "loss": 0.997,
+      "learning_rate": 1.3167685819099869e-05,
+      "loss": 0.9076,
       "step": 14642
     },
     {
-      "epoch": 0.40215868830847823,
+      "epoch": 0.415522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.3574183006584465e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.3166814057550614e-05,
+      "loss": 0.834,
       "step": 14643
     },
     {
-      "epoch": 0.40218615253632145,
+      "epoch": 0.41555051078320093,
       "grad_norm": 0.0,
-      "learning_rate": 1.3573352233938994e-05,
-      "loss": 0.9627,
+      "learning_rate": 1.3165942269251139e-05,
+      "loss": 0.9518,
       "step": 14644
     },
     {
-      "epoch": 0.40221361676416467,
+      "epoch": 0.4155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.3572521433019778e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.316507045420881e-05,
+      "loss": 1.0314,
       "step": 14645
     },
     {
-      "epoch": 0.4022410809920079,
+      "epoch": 0.4156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.3571690603833391e-05,
-      "loss": 1.0089,
+      "learning_rate": 1.3164198612430985e-05,
+      "loss": 0.8684,
       "step": 14646
     },
     {
-      "epoch": 0.40226854521985117,
+      "epoch": 0.41563564131668557,
       "grad_norm": 0.0,
-      "learning_rate": 1.3570859746386402e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.3163326743925034e-05,
+      "loss": 1.0095,
       "step": 14647
     },
     {
-      "epoch": 0.4022960094476944,
+      "epoch": 0.4156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.3570028860685387e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.3162454848698318e-05,
+      "loss": 0.9553,
       "step": 14648
     },
     {
-      "epoch": 0.4023234736755376,
+      "epoch": 0.41569239500567534,
       "grad_norm": 0.0,
-      "learning_rate": 1.356919794673692e-05,
-      "loss": 0.896,
+      "learning_rate": 1.3161582926758202e-05,
+      "loss": 0.9376,
       "step": 14649
     },
     {
-      "epoch": 0.40235093790338083,
+      "epoch": 0.41572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.356836700454758e-05,
-      "loss": 0.9014,
+      "learning_rate": 1.316071097811206e-05,
+      "loss": 0.8712,
       "step": 14650
     },
     {
-      "epoch": 0.40237840213122406,
+      "epoch": 0.41574914869466517,
       "grad_norm": 0.0,
-      "learning_rate": 1.3567536034123934e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.3159839002767245e-05,
+      "loss": 1.0194,
       "step": 14651
     },
     {
-      "epoch": 0.40240586635906733,
+      "epoch": 0.41577752553916003,
       "grad_norm": 0.0,
-      "learning_rate": 1.3566705035472563e-05,
-      "loss": 0.9759,
+      "learning_rate": 1.3158967000731129e-05,
+      "loss": 0.9805,
       "step": 14652
     },
     {
-      "epoch": 0.40243333058691055,
+      "epoch": 0.41580590238365495,
       "grad_norm": 0.0,
-      "learning_rate": 1.356587400860004e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.3158094972011077e-05,
+      "loss": 0.9432,
       "step": 14653
     },
     {
-      "epoch": 0.4024607948147538,
+      "epoch": 0.4158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.3565042953512942e-05,
-      "loss": 0.8542,
+      "learning_rate": 1.3157222916614454e-05,
+      "loss": 0.9367,
       "step": 14654
     },
     {
-      "epoch": 0.402488259042597,
+      "epoch": 0.4158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.356421187021784e-05,
-      "loss": 0.8729,
+      "learning_rate": 1.3156350834548625e-05,
+      "loss": 0.9107,
       "step": 14655
     },
     {
-      "epoch": 0.4025157232704403,
+      "epoch": 0.41589103291713964,
       "grad_norm": 0.0,
-      "learning_rate": 1.3563380758721314e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.315547872582096e-05,
+      "loss": 0.9403,
       "step": 14656
     },
     {
-      "epoch": 0.4025431874982835,
+      "epoch": 0.4159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.3562549619029938e-05,
-      "loss": 0.8955,
+      "learning_rate": 1.3154606590438828e-05,
+      "loss": 1.0266,
       "step": 14657
     },
     {
-      "epoch": 0.4025706517261267,
+      "epoch": 0.4159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.3561718451150294e-05,
-      "loss": 0.9979,
+      "learning_rate": 1.315373442840959e-05,
+      "loss": 0.9629,
       "step": 14658
     },
     {
-      "epoch": 0.40259811595396994,
+      "epoch": 0.41597616345062427,
       "grad_norm": 0.0,
-      "learning_rate": 1.356088725508895e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.3152862239740616e-05,
+      "loss": 0.8471,
       "step": 14659
     },
     {
-      "epoch": 0.4026255801818132,
+      "epoch": 0.4160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.3560056030852486e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.3151990024439272e-05,
+      "loss": 0.9273,
       "step": 14660
     },
     {
-      "epoch": 0.40265304440965644,
+      "epoch": 0.4160329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.3559224778447485e-05,
-      "loss": 0.9863,
+      "learning_rate": 1.3151117782512928e-05,
+      "loss": 0.8701,
       "step": 14661
     },
     {
-      "epoch": 0.40268050863749966,
+      "epoch": 0.41606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.3558393497880515e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.3150245513968947e-05,
+      "loss": 0.8601,
       "step": 14662
     },
     {
-      "epoch": 0.4027079728653429,
+      "epoch": 0.4160896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.3557562189158156e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.3149373218814703e-05,
+      "loss": 0.9757,
       "step": 14663
     },
     {
-      "epoch": 0.4027354370931861,
+      "epoch": 0.41611804767309873,
       "grad_norm": 0.0,
-      "learning_rate": 1.355673085228699e-05,
-      "loss": 0.9124,
+      "learning_rate": 1.314850089705756e-05,
+      "loss": 0.8305,
       "step": 14664
     },
     {
-      "epoch": 0.4027629013210294,
+      "epoch": 0.41614642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.3555899487273588e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.3147628548704893e-05,
+      "loss": 0.9306,
       "step": 14665
     },
     {
-      "epoch": 0.4027903655488726,
+      "epoch": 0.4161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.3555068094124535e-05,
-      "loss": 0.9981,
+      "learning_rate": 1.3146756173764061e-05,
+      "loss": 0.8874,
       "step": 14666
     },
     {
-      "epoch": 0.4028178297767158,
+      "epoch": 0.4162031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.3554236672846405e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.3145883772242443e-05,
+      "loss": 0.9318,
       "step": 14667
     },
     {
-      "epoch": 0.40284529400455904,
+      "epoch": 0.41623155505107834,
       "grad_norm": 0.0,
-      "learning_rate": 1.3553405223445777e-05,
-      "loss": 0.972,
+      "learning_rate": 1.31450113441474e-05,
+      "loss": 0.9116,
       "step": 14668
     },
     {
-      "epoch": 0.4028727582324023,
+      "epoch": 0.4162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.355257374592923e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.3144138889486304e-05,
+      "loss": 0.9014,
       "step": 14669
     },
     {
-      "epoch": 0.40290022246024554,
+      "epoch": 0.4162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.3551742240303346e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.3143266408266528e-05,
+      "loss": 0.8896,
       "step": 14670
     },
     {
-      "epoch": 0.40292768668808876,
+      "epoch": 0.416316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.35509107065747e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.3142393900495438e-05,
+      "loss": 0.9376,
       "step": 14671
     },
     {
-      "epoch": 0.402955150915932,
+      "epoch": 0.4163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.3550079144749872e-05,
-      "loss": 1.0442,
+      "learning_rate": 1.3141521366180407e-05,
+      "loss": 0.9088,
       "step": 14672
     },
     {
-      "epoch": 0.40298261514377526,
+      "epoch": 0.4163734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.3549247554835445e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.3140648805328801e-05,
+      "loss": 0.8607,
       "step": 14673
     },
     {
-      "epoch": 0.4030100793716185,
+      "epoch": 0.41640181611804766,
       "grad_norm": 0.0,
-      "learning_rate": 1.3548415936837998e-05,
-      "loss": 0.8588,
+      "learning_rate": 1.3139776217947996e-05,
+      "loss": 1.0578,
       "step": 14674
     },
     {
-      "epoch": 0.4030375435994617,
+      "epoch": 0.4164301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.3547584290764103e-05,
-      "loss": 0.9831,
+      "learning_rate": 1.3138903604045358e-05,
+      "loss": 0.9068,
       "step": 14675
     },
     {
-      "epoch": 0.4030650078273049,
+      "epoch": 0.41645856980703744,
       "grad_norm": 0.0,
-      "learning_rate": 1.3546752616620354e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.3138030963628263e-05,
+      "loss": 0.9275,
       "step": 14676
     },
     {
-      "epoch": 0.40309247205514814,
+      "epoch": 0.41648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.354592091441332e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.3137158296704078e-05,
+      "loss": 0.99,
       "step": 14677
     },
     {
-      "epoch": 0.4031199362829914,
+      "epoch": 0.41651532349602727,
       "grad_norm": 0.0,
-      "learning_rate": 1.3545089184149588e-05,
-      "loss": 0.991,
+      "learning_rate": 1.3136285603280174e-05,
+      "loss": 1.054,
       "step": 14678
     },
     {
-      "epoch": 0.40314740051083464,
+      "epoch": 0.4165437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.3544257425835737e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.3135412883363928e-05,
+      "loss": 0.8929,
       "step": 14679
     },
     {
-      "epoch": 0.40317486473867786,
+      "epoch": 0.41657207718501704,
       "grad_norm": 0.0,
-      "learning_rate": 1.3543425639478347e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.3134540136962704e-05,
+      "loss": 0.8638,
       "step": 14680
     },
     {
-      "epoch": 0.4032023289665211,
+      "epoch": 0.4166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.3542593825084e-05,
-      "loss": 0.8414,
+      "learning_rate": 1.3133667364083882e-05,
+      "loss": 0.9583,
       "step": 14681
     },
     {
-      "epoch": 0.40322979319436436,
+      "epoch": 0.4166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.3541761982659283e-05,
-      "loss": 0.8565,
+      "learning_rate": 1.3132794564734829e-05,
+      "loss": 0.8878,
       "step": 14682
     },
     {
-      "epoch": 0.4032572574222076,
+      "epoch": 0.4166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.3540930112210769e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.3131921738922921e-05,
+      "loss": 0.9544,
       "step": 14683
     },
     {
-      "epoch": 0.4032847216500508,
+      "epoch": 0.4166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.3540098213745045e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.3131048886655529e-05,
+      "loss": 0.8914,
       "step": 14684
     },
     {
-      "epoch": 0.403312185877894,
+      "epoch": 0.4167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.3539266287268693e-05,
-      "loss": 0.8353,
+      "learning_rate": 1.3130176007940027e-05,
+      "loss": 0.9212,
       "step": 14685
     },
     {
-      "epoch": 0.4033396501057373,
+      "epoch": 0.41674233825198637,
       "grad_norm": 0.0,
-      "learning_rate": 1.3538434332788295e-05,
-      "loss": 0.907,
+      "learning_rate": 1.3129303102783785e-05,
+      "loss": 0.9707,
       "step": 14686
     },
     {
-      "epoch": 0.4033671143335805,
+      "epoch": 0.4167707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.3537602350310434e-05,
-      "loss": 0.9766,
+      "learning_rate": 1.312843017119418e-05,
+      "loss": 0.8619,
       "step": 14687
     },
     {
-      "epoch": 0.40339457856142374,
+      "epoch": 0.41679909194097614,
       "grad_norm": 0.0,
-      "learning_rate": 1.3536770339841694e-05,
-      "loss": 1.0555,
+      "learning_rate": 1.3127557213178586e-05,
+      "loss": 0.9343,
       "step": 14688
     },
     {
-      "epoch": 0.40342204278926697,
+      "epoch": 0.41682746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.3535938301388654e-05,
-      "loss": 0.9829,
+      "learning_rate": 1.3126684228744375e-05,
+      "loss": 0.8765,
       "step": 14689
     },
     {
-      "epoch": 0.4034495070171102,
+      "epoch": 0.41685584562996597,
       "grad_norm": 0.0,
-      "learning_rate": 1.3535106234957905e-05,
-      "loss": 0.9168,
+      "learning_rate": 1.312581121789892e-05,
+      "loss": 1.0118,
       "step": 14690
     },
     {
-      "epoch": 0.40347697124495346,
+      "epoch": 0.41688422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 1.3534274140556023e-05,
-      "loss": 0.9181,
+      "learning_rate": 1.3124938180649601e-05,
+      "loss": 0.9834,
       "step": 14691
     },
     {
-      "epoch": 0.4035044354727967,
+      "epoch": 0.41691259931895575,
       "grad_norm": 0.0,
-      "learning_rate": 1.3533442018189595e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.3124065117003786e-05,
+      "loss": 0.9202,
       "step": 14692
     },
     {
-      "epoch": 0.4035318997006399,
+      "epoch": 0.4169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.3532609867865207e-05,
-      "loss": 0.9686,
+      "learning_rate": 1.3123192026968851e-05,
+      "loss": 1.0343,
       "step": 14693
     },
     {
-      "epoch": 0.40355936392848313,
+      "epoch": 0.4169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.353177768958944e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.3122318910552174e-05,
+      "loss": 0.9337,
       "step": 14694
     },
     {
-      "epoch": 0.4035868281563264,
+      "epoch": 0.41699772985244044,
       "grad_norm": 0.0,
-      "learning_rate": 1.353094548336888e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.3121445767761132e-05,
+      "loss": 0.9228,
       "step": 14695
     },
     {
-      "epoch": 0.4036142923841696,
+      "epoch": 0.4170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.353011324921011e-05,
-      "loss": 0.8743,
+      "learning_rate": 1.3120572598603094e-05,
+      "loss": 0.89,
       "step": 14696
     },
     {
-      "epoch": 0.40364175661201285,
+      "epoch": 0.4170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.352928098711972e-05,
-      "loss": 0.9507,
+      "learning_rate": 1.3119699403085439e-05,
+      "loss": 0.8914,
       "step": 14697
     },
     {
-      "epoch": 0.40366922083985607,
+      "epoch": 0.41708286038592507,
       "grad_norm": 0.0,
-      "learning_rate": 1.3528448697104289e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.3118826181215545e-05,
+      "loss": 0.9528,
       "step": 14698
     },
     {
-      "epoch": 0.40369668506769935,
+      "epoch": 0.41711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.3527616379170407e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.3117952933000783e-05,
+      "loss": 0.8743,
       "step": 14699
     },
     {
-      "epoch": 0.40372414929554257,
+      "epoch": 0.41713961407491484,
       "grad_norm": 0.0,
-      "learning_rate": 1.3526784033324653e-05,
-      "loss": 0.8286,
+      "learning_rate": 1.3117079658448535e-05,
+      "loss": 0.8774,
       "step": 14700
     },
     {
-      "epoch": 0.4037516135233858,
+      "epoch": 0.41716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.3525951659573622e-05,
-      "loss": 1.0121,
+      "learning_rate": 1.3116206357566173e-05,
+      "loss": 0.7841,
       "step": 14701
     },
     {
-      "epoch": 0.403779077751229,
+      "epoch": 0.4171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.3525119257923898e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.3115333030361078e-05,
+      "loss": 0.8919,
       "step": 14702
     },
     {
-      "epoch": 0.40380654197907223,
+      "epoch": 0.41722474460839953,
       "grad_norm": 0.0,
-      "learning_rate": 1.3524286828382061e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.311445967684062e-05,
+      "loss": 0.9589,
       "step": 14703
     },
     {
-      "epoch": 0.4038340062069155,
+      "epoch": 0.41725312145289445,
       "grad_norm": 0.0,
-      "learning_rate": 1.3523454370954699e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.3113586297012187e-05,
+      "loss": 0.8765,
       "step": 14704
     },
     {
-      "epoch": 0.40386147043475873,
+      "epoch": 0.4172814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.3522621885648406e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.3112712890883149e-05,
+      "loss": 0.9438,
       "step": 14705
     },
     {
-      "epoch": 0.40388893466260195,
+      "epoch": 0.4173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.3521789372469762e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.3111839458460884e-05,
+      "loss": 0.8348,
       "step": 14706
     },
     {
-      "epoch": 0.40391639889044517,
+      "epoch": 0.41733825198637914,
       "grad_norm": 0.0,
-      "learning_rate": 1.3520956831425357e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.3110965999752772e-05,
+      "loss": 0.8812,
       "step": 14707
     },
     {
-      "epoch": 0.40394386311828845,
+      "epoch": 0.417366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.3520124262521779e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.3110092514766191e-05,
+      "loss": 0.8884,
       "step": 14708
     },
     {
-      "epoch": 0.40397132734613167,
+      "epoch": 0.4173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.3519291665765611e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.3109219003508517e-05,
+      "loss": 0.8738,
       "step": 14709
     },
     {
-      "epoch": 0.4039987915739749,
+      "epoch": 0.4174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.3518459041163445e-05,
-      "loss": 1.0494,
+      "learning_rate": 1.3108345465987128e-05,
+      "loss": 0.9628,
       "step": 14710
     },
     {
-      "epoch": 0.4040262558018181,
+      "epoch": 0.4174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.3517626388721872e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.310747190220941e-05,
+      "loss": 0.9659,
       "step": 14711
     },
     {
-      "epoch": 0.4040537200296614,
+      "epoch": 0.4174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.3516793708447471e-05,
-      "loss": 0.8153,
+      "learning_rate": 1.3106598312182736e-05,
+      "loss": 0.9011,
       "step": 14712
     },
     {
-      "epoch": 0.4040811842575046,
+      "epoch": 0.41750851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 1.3515961000346835e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.3105724695914486e-05,
+      "loss": 0.8744,
       "step": 14713
     },
     {
-      "epoch": 0.40410864848534783,
+      "epoch": 0.4175368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.3515128264426563e-05,
-      "loss": 0.9578,
+      "learning_rate": 1.3104851053412039e-05,
+      "loss": 0.9595,
       "step": 14714
     },
     {
-      "epoch": 0.40413611271319105,
+      "epoch": 0.41756526674233824,
       "grad_norm": 0.0,
-      "learning_rate": 1.3514295500693227e-05,
-      "loss": 0.8931,
+      "learning_rate": 1.3103977384682778e-05,
+      "loss": 0.9396,
       "step": 14715
     },
     {
-      "epoch": 0.40416357694103433,
+      "epoch": 0.41759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.3513462709153422e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.3103103689734079e-05,
+      "loss": 0.8223,
       "step": 14716
     },
     {
-      "epoch": 0.40419104116887755,
+      "epoch": 0.417622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.3512629889813745e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.3102229968573321e-05,
+      "loss": 0.9323,
       "step": 14717
     },
     {
-      "epoch": 0.4042185053967208,
+      "epoch": 0.4176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.3511797042680774e-05,
-      "loss": 0.899,
+      "learning_rate": 1.310135622120789e-05,
+      "loss": 0.9877,
       "step": 14718
     },
     {
-      "epoch": 0.404245969624564,
+      "epoch": 0.41767877412031784,
       "grad_norm": 0.0,
-      "learning_rate": 1.3510964167761107e-05,
-      "loss": 0.8536,
+      "learning_rate": 1.3100482447645164e-05,
+      "loss": 0.799,
       "step": 14719
     },
     {
-      "epoch": 0.4042734338524072,
+      "epoch": 0.4177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.3510131265061332e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.309960864789252e-05,
+      "loss": 0.8825,
       "step": 14720
     },
     {
-      "epoch": 0.4043008980802505,
+      "epoch": 0.4177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.3509298334588037e-05,
-      "loss": 0.987,
+      "learning_rate": 1.3098734821957347e-05,
+      "loss": 0.9486,
       "step": 14721
     },
     {
-      "epoch": 0.4043283623080937,
+      "epoch": 0.4177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.3508465376347816e-05,
-      "loss": 0.9138,
+      "learning_rate": 1.309786096984702e-05,
+      "loss": 0.9041,
       "step": 14722
     },
     {
-      "epoch": 0.40435582653593694,
+      "epoch": 0.4177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.350763239034726e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.309698709156892e-05,
+      "loss": 1.0315,
       "step": 14723
     },
     {
-      "epoch": 0.40438329076378016,
+      "epoch": 0.4178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.3506799376592954e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.3096113187130432e-05,
+      "loss": 0.8781,
       "step": 14724
     },
     {
-      "epoch": 0.40441075499162343,
+      "epoch": 0.41784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.3505966335091493e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.3095239256538939e-05,
+      "loss": 0.9629,
       "step": 14725
     },
     {
-      "epoch": 0.40443821921946665,
+      "epoch": 0.4178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.3505133265849472e-05,
-      "loss": 0.9469,
+      "learning_rate": 1.3094365299801818e-05,
+      "loss": 1.0045,
       "step": 14726
     },
     {
-      "epoch": 0.4044656834473099,
+      "epoch": 0.41790578887627694,
       "grad_norm": 0.0,
-      "learning_rate": 1.3504300168873474e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.3093491316926454e-05,
+      "loss": 0.9731,
       "step": 14727
     },
     {
-      "epoch": 0.4044931476751531,
+      "epoch": 0.41793416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 1.3503467044170098e-05,
-      "loss": 1.0013,
+      "learning_rate": 1.3092617307920233e-05,
+      "loss": 0.865,
       "step": 14728
     },
     {
-      "epoch": 0.4045206119029964,
+      "epoch": 0.4179625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.3502633891745935e-05,
-      "loss": 0.9107,
+      "learning_rate": 1.3091743272790533e-05,
+      "loss": 0.8376,
       "step": 14729
     },
     {
-      "epoch": 0.4045480761308396,
+      "epoch": 0.41799091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.3501800711607575e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.309086921154474e-05,
+      "loss": 0.7502,
       "step": 14730
     },
     {
-      "epoch": 0.4045755403586828,
+      "epoch": 0.41801929625425654,
       "grad_norm": 0.0,
-      "learning_rate": 1.350096750376161e-05,
-      "loss": 0.9595,
+      "learning_rate": 1.3089995124190234e-05,
+      "loss": 0.8343,
       "step": 14731
     },
     {
-      "epoch": 0.40460300458652604,
+      "epoch": 0.4180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.3500134268214636e-05,
-      "loss": 0.9128,
+      "learning_rate": 1.3089121010734398e-05,
+      "loss": 0.9171,
       "step": 14732
     },
     {
-      "epoch": 0.40463046881436926,
+      "epoch": 0.4180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.349930100497324e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.3088246871184622e-05,
+      "loss": 1.0019,
       "step": 14733
     },
     {
-      "epoch": 0.40465793304221254,
+      "epoch": 0.4181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.3498467714044024e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.3087372705548282e-05,
+      "loss": 0.7906,
       "step": 14734
     },
     {
-      "epoch": 0.40468539727005576,
+      "epoch": 0.4181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.3497634395433575e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.3086498513832768e-05,
+      "loss": 0.9413,
       "step": 14735
     },
     {
-      "epoch": 0.404712861497899,
+      "epoch": 0.418161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.3496801049148484e-05,
-      "loss": 0.8318,
+      "learning_rate": 1.3085624296045465e-05,
+      "loss": 0.9284,
       "step": 14736
     },
     {
-      "epoch": 0.4047403257257422,
+      "epoch": 0.41818955732122587,
       "grad_norm": 0.0,
-      "learning_rate": 1.349596767519535e-05,
-      "loss": 0.914,
+      "learning_rate": 1.3084750052193749e-05,
+      "loss": 0.8481,
       "step": 14737
     },
     {
-      "epoch": 0.4047677899535855,
+      "epoch": 0.4182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.349513427358077e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.3083875782285015e-05,
+      "loss": 0.9667,
       "step": 14738
     },
     {
-      "epoch": 0.4047952541814287,
+      "epoch": 0.41824631101021564,
       "grad_norm": 0.0,
-      "learning_rate": 1.3494300844311328e-05,
-      "loss": 0.8064,
+      "learning_rate": 1.3083001486326646e-05,
+      "loss": 0.8974,
       "step": 14739
     },
     {
-      "epoch": 0.4048227184092719,
+      "epoch": 0.41827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.3493467387393626e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.308212716432602e-05,
+      "loss": 0.8669,
       "step": 14740
     },
     {
-      "epoch": 0.40485018263711514,
+      "epoch": 0.4183030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.3492633902834259e-05,
-      "loss": 0.8201,
+      "learning_rate": 1.3081252816290527e-05,
+      "loss": 0.9401,
       "step": 14741
     },
     {
-      "epoch": 0.4048776468649584,
+      "epoch": 0.41833144154370033,
       "grad_norm": 0.0,
-      "learning_rate": 1.349180039063982e-05,
-      "loss": 1.013,
+      "learning_rate": 1.3080378442227554e-05,
+      "loss": 1.0125,
       "step": 14742
     },
     {
-      "epoch": 0.40490511109280164,
+      "epoch": 0.41835981838819525,
       "grad_norm": 0.0,
-      "learning_rate": 1.34909668508169e-05,
-      "loss": 0.8427,
+      "learning_rate": 1.3079504042144487e-05,
+      "loss": 0.8951,
       "step": 14743
     },
     {
-      "epoch": 0.40493257532064486,
+      "epoch": 0.4183881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.34901332833721e-05,
-      "loss": 0.9433,
+      "learning_rate": 1.307862961604871e-05,
+      "loss": 0.8957,
       "step": 14744
     },
     {
-      "epoch": 0.4049600395484881,
+      "epoch": 0.418416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.3489299688312013e-05,
-      "loss": 0.8449,
+      "learning_rate": 1.3077755163947612e-05,
+      "loss": 0.918,
       "step": 14745
     },
     {
-      "epoch": 0.4049875037763313,
+      "epoch": 0.4184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.3488466065643236e-05,
-      "loss": 1.0219,
+      "learning_rate": 1.3076880685848576e-05,
+      "loss": 0.9178,
       "step": 14746
     },
     {
-      "epoch": 0.4050149680041746,
+      "epoch": 0.4184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.3487632415372364e-05,
-      "loss": 0.8079,
+      "learning_rate": 1.3076006181758989e-05,
+      "loss": 0.9826,
       "step": 14747
     },
     {
-      "epoch": 0.4050424322320178,
+      "epoch": 0.4185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.3486798737505992e-05,
-      "loss": 0.798,
+      "learning_rate": 1.307513165168624e-05,
+      "loss": 0.9489,
       "step": 14748
     },
     {
-      "epoch": 0.405069896459861,
+      "epoch": 0.41853007945516457,
       "grad_norm": 0.0,
-      "learning_rate": 1.3485965032050719e-05,
-      "loss": 0.8872,
+      "learning_rate": 1.307425709563772e-05,
+      "loss": 0.9165,
       "step": 14749
     },
     {
-      "epoch": 0.40509736068770424,
+      "epoch": 0.4185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.3485131299013141e-05,
-      "loss": 0.8389,
+      "learning_rate": 1.307338251362081e-05,
+      "loss": 0.8333,
       "step": 14750
     },
     {
-      "epoch": 0.4051248249155475,
+      "epoch": 0.41858683314415435,
       "grad_norm": 0.0,
-      "learning_rate": 1.3484297538399851e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.3072507905642897e-05,
+      "loss": 0.9007,
       "step": 14751
     },
     {
-      "epoch": 0.40515228914339074,
+      "epoch": 0.41861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.3483463750217453e-05,
-      "loss": 0.9404,
+      "learning_rate": 1.3071633271711375e-05,
+      "loss": 0.9119,
       "step": 14752
     },
     {
-      "epoch": 0.40517975337123396,
+      "epoch": 0.4186435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.3482629934472537e-05,
-      "loss": 0.803,
+      "learning_rate": 1.3070758611833628e-05,
+      "loss": 0.9046,
       "step": 14753
     },
     {
-      "epoch": 0.4052072175990772,
+      "epoch": 0.41867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.3481796091171709e-05,
-      "loss": 0.9686,
+      "learning_rate": 1.3069883926017042e-05,
+      "loss": 0.9789,
       "step": 14754
     },
     {
-      "epoch": 0.40523468182692046,
+      "epoch": 0.41870034052213395,
       "grad_norm": 0.0,
-      "learning_rate": 1.3480962220321558e-05,
-      "loss": 0.8142,
+      "learning_rate": 1.3069009214269011e-05,
+      "loss": 0.9496,
       "step": 14755
     },
     {
-      "epoch": 0.4052621460547637,
+      "epoch": 0.4187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.3480128321928685e-05,
-      "loss": 0.8621,
+      "learning_rate": 1.306813447659692e-05,
+      "loss": 0.8159,
       "step": 14756
     },
     {
-      "epoch": 0.4052896102826069,
+      "epoch": 0.4187570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.3479294395999689e-05,
-      "loss": 0.9014,
+      "learning_rate": 1.306725971300816e-05,
+      "loss": 0.8786,
       "step": 14757
     },
     {
-      "epoch": 0.4053170745104501,
+      "epoch": 0.41878547105561864,
       "grad_norm": 0.0,
-      "learning_rate": 1.3478460442541168e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.3066384923510121e-05,
+      "loss": 0.8892,
       "step": 14758
     },
     {
-      "epoch": 0.40534453873829335,
+      "epoch": 0.4188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.3477626461559722e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.306551010811019e-05,
+      "loss": 0.9765,
       "step": 14759
     },
     {
-      "epoch": 0.4053720029661366,
+      "epoch": 0.4188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.3476792453061946e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.3064635266815757e-05,
+      "loss": 0.8513,
       "step": 14760
     },
     {
-      "epoch": 0.40539946719397985,
+      "epoch": 0.4188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.3475958417054445e-05,
-      "loss": 0.9594,
+      "learning_rate": 1.306376039963421e-05,
+      "loss": 0.7565,
       "step": 14761
     },
     {
-      "epoch": 0.40542693142182307,
+      "epoch": 0.4188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.3475124353543813e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.3062885506572944e-05,
+      "loss": 1.0297,
       "step": 14762
     },
     {
-      "epoch": 0.4054543956496663,
+      "epoch": 0.41892735527809305,
       "grad_norm": 0.0,
-      "learning_rate": 1.3474290262536649e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.3062010587639345e-05,
+      "loss": 0.8804,
       "step": 14763
     },
     {
-      "epoch": 0.40548185987750957,
+      "epoch": 0.41895573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 1.3473456144039556e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.3061135642840804e-05,
+      "loss": 0.8663,
       "step": 14764
     },
     {
-      "epoch": 0.4055093241053528,
+      "epoch": 0.4189841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.3472621998059133e-05,
-      "loss": 0.9217,
+      "learning_rate": 1.3060260672184713e-05,
+      "loss": 0.9591,
       "step": 14765
     },
     {
-      "epoch": 0.405536788333196,
+      "epoch": 0.41901248581157774,
       "grad_norm": 0.0,
-      "learning_rate": 1.3471787824601978e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.3059385675678465e-05,
+      "loss": 0.9256,
       "step": 14766
     },
     {
-      "epoch": 0.40556425256103923,
+      "epoch": 0.41904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.3470953623674696e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.3058510653329446e-05,
+      "loss": 0.9279,
       "step": 14767
     },
     {
-      "epoch": 0.4055917167888825,
+      "epoch": 0.4190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.347011939528388e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.305763560514505e-05,
+      "loss": 1.0288,
       "step": 14768
     },
     {
-      "epoch": 0.4056191810167257,
+      "epoch": 0.41909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.3469285139436137e-05,
-      "loss": 0.9011,
+      "learning_rate": 1.3056760531132671e-05,
+      "loss": 0.9179,
       "step": 14769
     },
     {
-      "epoch": 0.40564664524456895,
+      "epoch": 0.41912599318955734,
       "grad_norm": 0.0,
-      "learning_rate": 1.3468450856138068e-05,
-      "loss": 1.0051,
+      "learning_rate": 1.3055885431299695e-05,
+      "loss": 0.883,
       "step": 14770
     },
     {
-      "epoch": 0.40567410947241217,
+      "epoch": 0.4191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.3467616545396271e-05,
-      "loss": 0.9444,
+      "learning_rate": 1.3055010305653516e-05,
+      "loss": 0.9242,
       "step": 14771
     },
     {
-      "epoch": 0.4057015737002554,
+      "epoch": 0.4191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.3466782207217348e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.3054135154201535e-05,
+      "loss": 0.9029,
       "step": 14772
     },
     {
-      "epoch": 0.40572903792809867,
+      "epoch": 0.419211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.3465947841607902e-05,
-      "loss": 0.8987,
+      "learning_rate": 1.3053259976951134e-05,
+      "loss": 0.8816,
       "step": 14773
     },
     {
-      "epoch": 0.4057565021559419,
+      "epoch": 0.4192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.3465113448574534e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.3052384773909706e-05,
+      "loss": 0.9315,
       "step": 14774
     },
     {
-      "epoch": 0.4057839663837851,
+      "epoch": 0.4192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.3464279028123845e-05,
-      "loss": 0.9919,
+      "learning_rate": 1.3051509545084648e-05,
+      "loss": 0.8264,
       "step": 14775
     },
     {
-      "epoch": 0.40581143061162833,
+      "epoch": 0.41929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.3463444580262442e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.3050634290483352e-05,
+      "loss": 0.9366,
       "step": 14776
     },
     {
-      "epoch": 0.4058388948394716,
+      "epoch": 0.4193246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.3462610104996921e-05,
-      "loss": 1.0641,
+      "learning_rate": 1.3049759010113212e-05,
+      "loss": 0.889,
       "step": 14777
     },
     {
-      "epoch": 0.40586635906731483,
+      "epoch": 0.41935300794551644,
       "grad_norm": 0.0,
-      "learning_rate": 1.3461775602333886e-05,
-      "loss": 1.027,
+      "learning_rate": 1.3048883703981617e-05,
+      "loss": 0.843,
       "step": 14778
     },
     {
-      "epoch": 0.40589382329515805,
+      "epoch": 0.41938138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 1.3460941072279944e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.3048008372095968e-05,
+      "loss": 0.9382,
       "step": 14779
     },
     {
-      "epoch": 0.4059212875230013,
+      "epoch": 0.4194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.3460106514841696e-05,
-      "loss": 1.0189,
+      "learning_rate": 1.3047133014463656e-05,
+      "loss": 0.9498,
       "step": 14780
     },
     {
-      "epoch": 0.40594875175084455,
+      "epoch": 0.41943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.3459271930025741e-05,
-      "loss": 0.894,
+      "learning_rate": 1.3046257631092072e-05,
+      "loss": 0.932,
       "step": 14781
     },
     {
-      "epoch": 0.40597621597868777,
+      "epoch": 0.41946651532349605,
       "grad_norm": 0.0,
-      "learning_rate": 1.3458437317838693e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.3045382221988612e-05,
+      "loss": 0.9539,
       "step": 14782
     },
     {
-      "epoch": 0.406003680206531,
+      "epoch": 0.4194948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.3457602678287145e-05,
-      "loss": 0.9101,
+      "learning_rate": 1.3044506787160676e-05,
+      "loss": 0.9434,
       "step": 14783
     },
     {
-      "epoch": 0.4060311444343742,
+      "epoch": 0.4195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.3456768011377706e-05,
-      "loss": 1.0184,
+      "learning_rate": 1.3043631326615648e-05,
+      "loss": 0.876,
       "step": 14784
     },
     {
-      "epoch": 0.40605860866221744,
+      "epoch": 0.4195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.3455933317116978e-05,
-      "loss": 0.8928,
+      "learning_rate": 1.3042755840360935e-05,
+      "loss": 1.0499,
       "step": 14785
     },
     {
-      "epoch": 0.4060860728900607,
+      "epoch": 0.4195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.345509859551157e-05,
-      "loss": 0.869,
+      "learning_rate": 1.3041880328403921e-05,
+      "loss": 0.9976,
       "step": 14786
     },
     {
-      "epoch": 0.40611353711790393,
+      "epoch": 0.4196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.345426384656808e-05,
-      "loss": 0.9793,
+      "learning_rate": 1.3041004790752013e-05,
+      "loss": 0.8766,
       "step": 14787
     },
     {
-      "epoch": 0.40614100134574715,
+      "epoch": 0.41963677639046537,
       "grad_norm": 0.0,
-      "learning_rate": 1.3453429070293118e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.3040129227412597e-05,
+      "loss": 1.0195,
       "step": 14788
     },
     {
-      "epoch": 0.4061684655735904,
+      "epoch": 0.4196651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.3452594266693288e-05,
-      "loss": 0.8089,
+      "learning_rate": 1.3039253638393075e-05,
+      "loss": 0.9391,
       "step": 14789
     },
     {
-      "epoch": 0.40619592980143365,
+      "epoch": 0.41969353007945515,
       "grad_norm": 0.0,
-      "learning_rate": 1.3451759435775195e-05,
-      "loss": 0.7872,
+      "learning_rate": 1.303837802370084e-05,
+      "loss": 0.8951,
       "step": 14790
     },
     {
-      "epoch": 0.4062233940292769,
+      "epoch": 0.41972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.3450924577545443e-05,
-      "loss": 0.8942,
+      "learning_rate": 1.3037502383343287e-05,
+      "loss": 1.0076,
       "step": 14791
     },
     {
-      "epoch": 0.4062508582571201,
+      "epoch": 0.419750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.345008969201064e-05,
-      "loss": 0.8418,
+      "learning_rate": 1.3036626717327818e-05,
+      "loss": 0.9271,
       "step": 14792
     },
     {
-      "epoch": 0.4062783224849633,
+      "epoch": 0.41977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.344925477917739e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.3035751025661824e-05,
+      "loss": 0.9439,
       "step": 14793
     },
     {
-      "epoch": 0.4063057867128066,
+      "epoch": 0.41980703745743475,
       "grad_norm": 0.0,
-      "learning_rate": 1.3448419839052303e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.3034875308352706e-05,
+      "loss": 0.9236,
       "step": 14794
     },
     {
-      "epoch": 0.4063332509406498,
+      "epoch": 0.4198354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.344758487164198e-05,
-      "loss": 0.8857,
+      "learning_rate": 1.303399956540786e-05,
+      "loss": 0.7115,
       "step": 14795
     },
     {
-      "epoch": 0.40636071516849304,
+      "epoch": 0.4198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.3446749876953028e-05,
-      "loss": 0.8898,
+      "learning_rate": 1.3033123796834683e-05,
+      "loss": 0.9112,
       "step": 14796
     },
     {
-      "epoch": 0.40638817939633626,
+      "epoch": 0.4198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.344591485499206e-05,
-      "loss": 0.8986,
+      "learning_rate": 1.3032248002640574e-05,
+      "loss": 0.8923,
       "step": 14797
     },
     {
-      "epoch": 0.4064156436241795,
+      "epoch": 0.4199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.3445079805765677e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.303137218283293e-05,
+      "loss": 0.9008,
       "step": 14798
     },
     {
-      "epoch": 0.40644310785202276,
+      "epoch": 0.4199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.3444244729280485e-05,
-      "loss": 1.0332,
+      "learning_rate": 1.3030496337419148e-05,
+      "loss": 0.8383,
       "step": 14799
     },
     {
-      "epoch": 0.406470572079866,
+      "epoch": 0.4199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.34434096255431e-05,
-      "loss": 0.9939,
+      "learning_rate": 1.3029620466406627e-05,
+      "loss": 0.9781,
       "step": 14800
     },
     {
-      "epoch": 0.4064980363077092,
+      "epoch": 0.420005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.3442574494560118e-05,
-      "loss": 0.8486,
+      "learning_rate": 1.3028744569802765e-05,
+      "loss": 0.9091,
       "step": 14801
     },
     {
-      "epoch": 0.4065255005355524,
+      "epoch": 0.42003405221339385,
       "grad_norm": 0.0,
-      "learning_rate": 1.344173933633816e-05,
-      "loss": 0.8978,
+      "learning_rate": 1.3027868647614961e-05,
+      "loss": 0.9442,
       "step": 14802
     },
     {
-      "epoch": 0.4065529647633957,
+      "epoch": 0.42006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.3440904150883825e-05,
-      "loss": 0.9402,
+      "learning_rate": 1.3026992699850616e-05,
+      "loss": 0.8774,
       "step": 14803
     },
     {
-      "epoch": 0.4065804289912389,
+      "epoch": 0.4200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.3440068938203719e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.302611672651713e-05,
+      "loss": 0.8354,
       "step": 14804
     },
     {
-      "epoch": 0.40660789321908214,
+      "epoch": 0.42011918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.3439233698304462e-05,
-      "loss": 0.8537,
+      "learning_rate": 1.3025240727621894e-05,
+      "loss": 0.9088,
       "step": 14805
     },
     {
-      "epoch": 0.40663535744692536,
+      "epoch": 0.42014755959137345,
       "grad_norm": 0.0,
-      "learning_rate": 1.3438398431192652e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.302436470317232e-05,
+      "loss": 0.9131,
       "step": 14806
     },
     {
-      "epoch": 0.40666282167476864,
+      "epoch": 0.4201759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.3437563136874899e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.3023488653175797e-05,
+      "loss": 1.0014,
       "step": 14807
     },
     {
-      "epoch": 0.40669028590261186,
+      "epoch": 0.42020431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 1.343672781535782e-05,
-      "loss": 0.8872,
+      "learning_rate": 1.302261257763973e-05,
+      "loss": 0.9655,
       "step": 14808
     },
     {
-      "epoch": 0.4067177501304551,
+      "epoch": 0.4202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.3435892466648016e-05,
-      "loss": 0.9064,
+      "learning_rate": 1.3021736476571518e-05,
+      "loss": 0.9158,
       "step": 14809
     },
     {
-      "epoch": 0.4067452143582983,
+      "epoch": 0.420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.3435057090752101e-05,
-      "loss": 0.8955,
+      "learning_rate": 1.3020860349978564e-05,
+      "loss": 0.8512,
       "step": 14810
     },
     {
-      "epoch": 0.4067726785861416,
+      "epoch": 0.4202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3434221687676683e-05,
-      "loss": 0.916,
+      "learning_rate": 1.3019984197868266e-05,
+      "loss": 0.8858,
       "step": 14811
     },
     {
-      "epoch": 0.4068001428139848,
+      "epoch": 0.4203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.3433386257428374e-05,
-      "loss": 1.0967,
+      "learning_rate": 1.3019108020248025e-05,
+      "loss": 0.8804,
       "step": 14812
     },
     {
-      "epoch": 0.406827607041828,
+      "epoch": 0.4203461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.343255080001378e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.3018231817125247e-05,
+      "loss": 0.8834,
       "step": 14813
     },
     {
-      "epoch": 0.40685507126967124,
+      "epoch": 0.42037457434733255,
       "grad_norm": 0.0,
-      "learning_rate": 1.3431715315439518e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.3017355588507325e-05,
+      "loss": 0.9124,
       "step": 14814
     },
     {
-      "epoch": 0.40688253549751446,
+      "epoch": 0.42040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.3430879803712194e-05,
-      "loss": 0.8996,
+      "learning_rate": 1.3016479334401666e-05,
+      "loss": 0.8043,
       "step": 14815
     },
     {
-      "epoch": 0.40690999972535774,
+      "epoch": 0.4204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.3430044264838418e-05,
-      "loss": 1.0466,
+      "learning_rate": 1.3015603054815669e-05,
+      "loss": 0.9685,
       "step": 14816
     },
     {
-      "epoch": 0.40693746395320096,
+      "epoch": 0.42045970488081724,
       "grad_norm": 0.0,
-      "learning_rate": 1.3429208698824808e-05,
-      "loss": 0.9245,
+      "learning_rate": 1.3014726749756738e-05,
+      "loss": 0.811,
       "step": 14817
     },
     {
-      "epoch": 0.4069649281810442,
+      "epoch": 0.42048808172531216,
       "grad_norm": 0.0,
-      "learning_rate": 1.3428373105677965e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.3013850419232275e-05,
+      "loss": 0.8155,
       "step": 14818
     },
     {
-      "epoch": 0.4069923924088874,
+      "epoch": 0.420516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.3427537485404507e-05,
-      "loss": 1.0291,
+      "learning_rate": 1.3012974063249684e-05,
+      "loss": 0.8643,
       "step": 14819
     },
     {
-      "epoch": 0.4070198566367307,
+      "epoch": 0.42054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.3426701838011048e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.3012097681816364e-05,
+      "loss": 0.9922,
       "step": 14820
     },
     {
-      "epoch": 0.4070473208645739,
+      "epoch": 0.42057321225879685,
       "grad_norm": 0.0,
-      "learning_rate": 1.3425866163504193e-05,
-      "loss": 0.8086,
+      "learning_rate": 1.3011221274939718e-05,
+      "loss": 1.0224,
       "step": 14821
     },
     {
-      "epoch": 0.4070747850924171,
+      "epoch": 0.4206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.342503046189056e-05,
-      "loss": 1.0887,
+      "learning_rate": 1.3010344842627155e-05,
+      "loss": 0.9345,
       "step": 14822
     },
     {
-      "epoch": 0.40710224932026035,
+      "epoch": 0.4206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.3424194733176758e-05,
-      "loss": 0.9992,
+      "learning_rate": 1.3009468384886068e-05,
+      "loss": 0.9432,
       "step": 14823
     },
     {
-      "epoch": 0.4071297135481036,
+      "epoch": 0.4206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.3423358977369403e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.3008591901723869e-05,
+      "loss": 0.8461,
       "step": 14824
     },
     {
-      "epoch": 0.40715717777594684,
+      "epoch": 0.4206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.34225231944751e-05,
-      "loss": 0.8782,
+      "learning_rate": 1.3007715393147957e-05,
+      "loss": 0.9383,
       "step": 14825
     },
     {
-      "epoch": 0.40718464200379006,
+      "epoch": 0.42071509648127126,
       "grad_norm": 0.0,
-      "learning_rate": 1.3421687384500474e-05,
-      "loss": 0.8347,
+      "learning_rate": 1.3006838859165742e-05,
+      "loss": 0.8117,
       "step": 14826
     },
     {
-      "epoch": 0.4072121062316333,
+      "epoch": 0.42074347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.3420851547452132e-05,
-      "loss": 1.021,
+      "learning_rate": 1.3005962299784623e-05,
+      "loss": 0.8281,
       "step": 14827
     },
     {
-      "epoch": 0.4072395704594765,
+      "epoch": 0.4207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.3420015683336684e-05,
-      "loss": 0.9031,
+      "learning_rate": 1.3005085715012003e-05,
+      "loss": 0.9401,
       "step": 14828
     },
     {
-      "epoch": 0.4072670346873198,
+      "epoch": 0.42080022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 1.341917979216075e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.300420910485529e-05,
+      "loss": 0.9327,
       "step": 14829
     },
     {
-      "epoch": 0.407294498915163,
+      "epoch": 0.42082860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 1.3418343873930942e-05,
-      "loss": 0.952,
+      "learning_rate": 1.3003332469321888e-05,
+      "loss": 0.9197,
       "step": 14830
     },
     {
-      "epoch": 0.4073219631430062,
+      "epoch": 0.4208569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.3417507928653868e-05,
-      "loss": 0.9105,
+      "learning_rate": 1.30024558084192e-05,
+      "loss": 0.9161,
       "step": 14831
     },
     {
-      "epoch": 0.40734942737084945,
+      "epoch": 0.42088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.341667195633615e-05,
-      "loss": 0.9679,
+      "learning_rate": 1.3001579122154633e-05,
+      "loss": 0.9068,
       "step": 14832
     },
     {
-      "epoch": 0.4073768915986927,
+      "epoch": 0.42091373439273555,
       "grad_norm": 0.0,
-      "learning_rate": 1.34158359569844e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.3000702410535595e-05,
+      "loss": 0.7946,
       "step": 14833
     },
     {
-      "epoch": 0.40740435582653595,
+      "epoch": 0.4209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.3414999930605233e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.2999825673569489e-05,
+      "loss": 0.9238,
       "step": 14834
     },
     {
-      "epoch": 0.40743182005437917,
+      "epoch": 0.4209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.3414163877205264e-05,
-      "loss": 0.9404,
+      "learning_rate": 1.2998948911263716e-05,
+      "loss": 0.8648,
       "step": 14835
     },
     {
-      "epoch": 0.4074592842822224,
+      "epoch": 0.4209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.3413327796791106e-05,
-      "loss": 0.8759,
+      "learning_rate": 1.299807212362569e-05,
+      "loss": 1.0103,
       "step": 14836
     },
     {
-      "epoch": 0.40748674851006567,
+      "epoch": 0.4210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.3412491689369378e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.2997195310662816e-05,
+      "loss": 0.9696,
       "step": 14837
     },
     {
-      "epoch": 0.4075142127379089,
+      "epoch": 0.42105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.3411655554946694e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.2996318472382494e-05,
+      "loss": 0.8331,
       "step": 14838
     },
     {
-      "epoch": 0.4075416769657521,
+      "epoch": 0.4210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.341081939352967e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.2995441608792137e-05,
+      "loss": 0.93,
       "step": 14839
     },
     {
-      "epoch": 0.40756914119359533,
+      "epoch": 0.4211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.3409983205124919e-05,
-      "loss": 0.8708,
+      "learning_rate": 1.299456471989915e-05,
+      "loss": 0.7998,
       "step": 14840
     },
     {
-      "epoch": 0.40759660542143855,
+      "epoch": 0.42114074914869465,
       "grad_norm": 0.0,
-      "learning_rate": 1.3409146989739064e-05,
-      "loss": 0.9516,
+      "learning_rate": 1.2993687805710941e-05,
+      "loss": 0.9044,
       "step": 14841
     },
     {
-      "epoch": 0.40762406964928183,
+      "epoch": 0.42116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.3408310747378717e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.2992810866234916e-05,
+      "loss": 0.9294,
       "step": 14842
     },
     {
-      "epoch": 0.40765153387712505,
+      "epoch": 0.4211975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.340747447805049e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.2991933901478484e-05,
+      "loss": 0.9015,
       "step": 14843
     },
     {
-      "epoch": 0.40767899810496827,
+      "epoch": 0.42122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.340663818176101e-05,
-      "loss": 0.8426,
+      "learning_rate": 1.2991056911449052e-05,
+      "loss": 0.8974,
       "step": 14844
     },
     {
-      "epoch": 0.4077064623328115,
+      "epoch": 0.42125425652667425,
       "grad_norm": 0.0,
-      "learning_rate": 1.3405801858516887e-05,
-      "loss": 0.8798,
+      "learning_rate": 1.2990179896154025e-05,
+      "loss": 0.8742,
       "step": 14845
     },
     {
-      "epoch": 0.40773392656065477,
+      "epoch": 0.4212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.3404965508324743e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.2989302855600816e-05,
+      "loss": 0.9155,
       "step": 14846
     },
     {
-      "epoch": 0.407761390788498,
+      "epoch": 0.42131101021566403,
       "grad_norm": 0.0,
-      "learning_rate": 1.340412913119119e-05,
-      "loss": 0.841,
+      "learning_rate": 1.298842578979683e-05,
+      "loss": 0.8563,
       "step": 14847
     },
     {
-      "epoch": 0.4077888550163412,
+      "epoch": 0.4213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.3403292727122849e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.298754869874948e-05,
+      "loss": 0.9525,
       "step": 14848
     },
     {
-      "epoch": 0.40781631924418443,
+      "epoch": 0.4213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.3402456296126339e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.2986671582466168e-05,
+      "loss": 0.8851,
       "step": 14849
     },
     {
-      "epoch": 0.4078437834720277,
+      "epoch": 0.4213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.3401619838208276e-05,
-      "loss": 0.9843,
+      "learning_rate": 1.298579444095431e-05,
+      "loss": 0.8573,
       "step": 14850
     },
     {
-      "epoch": 0.40787124769987093,
+      "epoch": 0.4214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.340078335337528e-05,
-      "loss": 0.8968,
+      "learning_rate": 1.298491727422131e-05,
+      "loss": 0.8211,
       "step": 14851
     },
     {
-      "epoch": 0.40789871192771415,
+      "epoch": 0.4214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.3399946841633965e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.2984040082274581e-05,
+      "loss": 0.9351,
       "step": 14852
     },
     {
-      "epoch": 0.4079261761555574,
+      "epoch": 0.42148127128263335,
       "grad_norm": 0.0,
-      "learning_rate": 1.3399110302990959e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.2983162865121533e-05,
+      "loss": 0.9507,
       "step": 14853
     },
     {
-      "epoch": 0.4079536403834006,
+      "epoch": 0.42150964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.3398273737452874e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.2982285622769573e-05,
+      "loss": 0.8811,
       "step": 14854
     },
     {
-      "epoch": 0.40798110461124387,
+      "epoch": 0.4215380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.3397437145026328e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.2981408355226109e-05,
+      "loss": 0.9215,
       "step": 14855
     },
     {
-      "epoch": 0.4080085688390871,
+      "epoch": 0.42156640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.3396600525717946e-05,
-      "loss": 0.9739,
+      "learning_rate": 1.2980531062498557e-05,
+      "loss": 0.837,
       "step": 14856
     },
     {
-      "epoch": 0.4080360330669303,
+      "epoch": 0.42159477866061296,
       "grad_norm": 0.0,
-      "learning_rate": 1.3395763879534345e-05,
-      "loss": 0.9282,
+      "learning_rate": 1.2979653744594324e-05,
+      "loss": 0.9046,
       "step": 14857
     },
     {
-      "epoch": 0.40806349729477354,
+      "epoch": 0.4216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.3394927206482145e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.2978776401520825e-05,
+      "loss": 0.9322,
       "step": 14858
     },
     {
-      "epoch": 0.4080909615226168,
+      "epoch": 0.42165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 1.3394090506567965e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.2977899033285465e-05,
+      "loss": 0.9408,
       "step": 14859
     },
     {
-      "epoch": 0.40811842575046003,
+      "epoch": 0.4216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.3393253779798425e-05,
-      "loss": 1.0502,
+      "learning_rate": 1.297702163989566e-05,
+      "loss": 0.8022,
       "step": 14860
     },
     {
-      "epoch": 0.40814588997830326,
+      "epoch": 0.4217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.3392417026180147e-05,
-      "loss": 0.7176,
+      "learning_rate": 1.2976144221358818e-05,
+      "loss": 0.844,
       "step": 14861
     },
     {
-      "epoch": 0.4081733542061465,
+      "epoch": 0.4217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.3391580245719754e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.297526677768235e-05,
+      "loss": 0.9859,
       "step": 14862
     },
     {
-      "epoch": 0.40820081843398975,
+      "epoch": 0.4217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.3390743438423864e-05,
-      "loss": 0.899,
+      "learning_rate": 1.2974389308873672e-05,
+      "loss": 0.9038,
       "step": 14863
     },
     {
-      "epoch": 0.408228282661833,
+      "epoch": 0.4217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.3389906604299095e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.2973511814940194e-05,
+      "loss": 0.8363,
       "step": 14864
     },
     {
-      "epoch": 0.4082557468896762,
+      "epoch": 0.42182179341657206,
       "grad_norm": 0.0,
-      "learning_rate": 1.3389069743352075e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.2972634295889327e-05,
+      "loss": 0.912,
       "step": 14865
     },
     {
-      "epoch": 0.4082832111175194,
+      "epoch": 0.42185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.3388232855589421e-05,
-      "loss": 0.8291,
+      "learning_rate": 1.2971756751728486e-05,
+      "loss": 0.9167,
       "step": 14866
     },
     {
-      "epoch": 0.40831067534536264,
+      "epoch": 0.4218785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.3387395941017758e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.2970879182465082e-05,
+      "loss": 0.9475,
       "step": 14867
     },
     {
-      "epoch": 0.4083381395732059,
+      "epoch": 0.42190692395005674,
       "grad_norm": 0.0,
-      "learning_rate": 1.3386558999643702e-05,
-      "loss": 0.9155,
+      "learning_rate": 1.297000158810653e-05,
+      "loss": 0.9921,
       "step": 14868
     },
     {
-      "epoch": 0.40836560380104914,
+      "epoch": 0.42193530079455166,
       "grad_norm": 0.0,
-      "learning_rate": 1.338572203147388e-05,
-      "loss": 0.9482,
+      "learning_rate": 1.2969123968660235e-05,
+      "loss": 0.8803,
       "step": 14869
     },
     {
-      "epoch": 0.40839306802889236,
+      "epoch": 0.4219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.3384885036514916e-05,
-      "loss": 0.964,
+      "learning_rate": 1.296824632413362e-05,
+      "loss": 0.9471,
       "step": 14870
     },
     {
-      "epoch": 0.4084205322567356,
+      "epoch": 0.42199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.3384048014773433e-05,
-      "loss": 0.8801,
+      "learning_rate": 1.2967368654534096e-05,
+      "loss": 0.9179,
       "step": 14871
     },
     {
-      "epoch": 0.40844799648457886,
+      "epoch": 0.42202043132803635,
       "grad_norm": 0.0,
-      "learning_rate": 1.3383210966256046e-05,
-      "loss": 0.9064,
+      "learning_rate": 1.2966490959869072e-05,
+      "loss": 1.0009,
       "step": 14872
     },
     {
-      "epoch": 0.4084754607124221,
+      "epoch": 0.4220488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.3382373890969387e-05,
-      "loss": 1.058,
+      "learning_rate": 1.2965613240145969e-05,
+      "loss": 0.8489,
       "step": 14873
     },
     {
-      "epoch": 0.4085029249402653,
+      "epoch": 0.4220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.3381536788920076e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.2964735495372197e-05,
+      "loss": 0.8841,
       "step": 14874
     },
     {
-      "epoch": 0.4085303891681085,
+      "epoch": 0.422105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.338069966011473e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.2963857725555169e-05,
+      "loss": 0.9173,
       "step": 14875
     },
     {
-      "epoch": 0.4085578533959518,
+      "epoch": 0.4221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.3379862504559986e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.2962979930702305e-05,
+      "loss": 0.996,
       "step": 14876
     },
     {
-      "epoch": 0.408585317623795,
+      "epoch": 0.42216231555051076,
       "grad_norm": 0.0,
-      "learning_rate": 1.3379025322262455e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.2962102110821013e-05,
+      "loss": 0.9236,
       "step": 14877
     },
     {
-      "epoch": 0.40861278185163824,
+      "epoch": 0.4221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.3378188113228771e-05,
-      "loss": 0.998,
+      "learning_rate": 1.2961224265918712e-05,
+      "loss": 1.0611,
       "step": 14878
     },
     {
-      "epoch": 0.40864024607948146,
+      "epoch": 0.4222190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.3377350877465553e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.2960346396002817e-05,
+      "loss": 0.9443,
       "step": 14879
     },
     {
-      "epoch": 0.4086677103073247,
+      "epoch": 0.42224744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 1.3376513614979425e-05,
-      "loss": 0.8821,
+      "learning_rate": 1.2959468501080744e-05,
+      "loss": 0.902,
       "step": 14880
     },
     {
-      "epoch": 0.40869517453516796,
+      "epoch": 0.42227582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.3375676325777014e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.2958590581159907e-05,
+      "loss": 0.8217,
       "step": 14881
     },
     {
-      "epoch": 0.4087226387630112,
+      "epoch": 0.4223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.3374839009864943e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.2957712636247722e-05,
+      "loss": 1.0046,
       "step": 14882
     },
     {
-      "epoch": 0.4087501029908544,
+      "epoch": 0.42233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.3374001667249842e-05,
-      "loss": 0.858,
+      "learning_rate": 1.2956834666351603e-05,
+      "loss": 0.8854,
       "step": 14883
     },
     {
-      "epoch": 0.4087775672186976,
+      "epoch": 0.42236095346197505,
       "grad_norm": 0.0,
-      "learning_rate": 1.3373164297938331e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.2955956671478973e-05,
+      "loss": 0.8788,
       "step": 14884
     },
     {
-      "epoch": 0.4088050314465409,
+      "epoch": 0.4223893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.3372326901937039e-05,
-      "loss": 0.999,
+      "learning_rate": 1.2955078651637241e-05,
+      "loss": 0.9162,
       "step": 14885
     },
     {
-      "epoch": 0.4088324956743841,
+      "epoch": 0.4224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.3371489479252589e-05,
-      "loss": 0.8233,
+      "learning_rate": 1.2954200606833826e-05,
+      "loss": 0.8855,
       "step": 14886
     },
     {
-      "epoch": 0.40885995990222734,
+      "epoch": 0.4224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.3370652029891607e-05,
-      "loss": 0.866,
+      "learning_rate": 1.295332253707615e-05,
+      "loss": 0.9508,
       "step": 14887
     },
     {
-      "epoch": 0.40888742413007056,
+      "epoch": 0.4224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.3369814553860721e-05,
-      "loss": 0.8646,
+      "learning_rate": 1.2952444442371624e-05,
+      "loss": 0.9457,
       "step": 14888
     },
     {
-      "epoch": 0.40891488835791384,
+      "epoch": 0.42250283768444946,
       "grad_norm": 0.0,
-      "learning_rate": 1.3368977051166556e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.2951566322727664e-05,
+      "loss": 0.9635,
       "step": 14889
     },
     {
-      "epoch": 0.40894235258575706,
+      "epoch": 0.4225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.3368139521815742e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.2950688178151696e-05,
+      "loss": 0.949,
       "step": 14890
     },
     {
-      "epoch": 0.4089698168136003,
+      "epoch": 0.4225595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.3367301965814903e-05,
-      "loss": 0.8663,
+      "learning_rate": 1.2949810008651129e-05,
+      "loss": 0.8917,
       "step": 14891
     },
     {
-      "epoch": 0.4089972810414435,
+      "epoch": 0.42258796821793415,
       "grad_norm": 0.0,
-      "learning_rate": 1.3366464383170663e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.2948931814233382e-05,
+      "loss": 0.8585,
       "step": 14892
     },
     {
-      "epoch": 0.4090247452692867,
+      "epoch": 0.42261634506242907,
       "grad_norm": 0.0,
-      "learning_rate": 1.3365626773889656e-05,
-      "loss": 0.8895,
+      "learning_rate": 1.2948053594905878e-05,
+      "loss": 0.9658,
       "step": 14893
     },
     {
-      "epoch": 0.40905220949713,
+      "epoch": 0.4226447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 1.3364789137978507e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.2947175350676033e-05,
+      "loss": 0.8617,
       "step": 14894
     },
     {
-      "epoch": 0.4090796737249732,
+      "epoch": 0.42267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.336395147544384e-05,
-      "loss": 1.0124,
+      "learning_rate": 1.2946297081551267e-05,
+      "loss": 0.8767,
       "step": 14895
     },
     {
-      "epoch": 0.40910713795281645,
+      "epoch": 0.42270147559591376,
       "grad_norm": 0.0,
-      "learning_rate": 1.3363113786292287e-05,
-      "loss": 1.0582,
+      "learning_rate": 1.2945418787538992e-05,
+      "loss": 0.9109,
       "step": 14896
     },
     {
-      "epoch": 0.40913460218065967,
+      "epoch": 0.4227298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.3362276070530477e-05,
-      "loss": 1.0287,
+      "learning_rate": 1.2944540468646639e-05,
+      "loss": 0.8973,
       "step": 14897
     },
     {
-      "epoch": 0.40916206640850294,
+      "epoch": 0.42275822928490353,
       "grad_norm": 0.0,
-      "learning_rate": 1.3361438328165036e-05,
-      "loss": 0.8999,
+      "learning_rate": 1.2943662124881615e-05,
+      "loss": 0.8906,
       "step": 14898
     },
     {
-      "epoch": 0.40918953063634617,
+      "epoch": 0.4227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.336060055920259e-05,
-      "loss": 0.8785,
+      "learning_rate": 1.2942783756251345e-05,
+      "loss": 0.8997,
       "step": 14899
     },
     {
-      "epoch": 0.4092169948641894,
+      "epoch": 0.4228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.3359762763649773e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.2941905362763252e-05,
+      "loss": 0.9885,
       "step": 14900
     },
     {
-      "epoch": 0.4092444590920326,
+      "epoch": 0.4228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.3358924941513209e-05,
-      "loss": 0.8851,
+      "learning_rate": 1.2941026944424748e-05,
+      "loss": 0.9321,
       "step": 14901
     },
     {
-      "epoch": 0.4092719233198759,
+      "epoch": 0.4228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.3358087092799532e-05,
-      "loss": 0.904,
+      "learning_rate": 1.2940148501243259e-05,
+      "loss": 0.953,
       "step": 14902
     },
     {
-      "epoch": 0.4092993875477191,
+      "epoch": 0.422900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.3357249217515371e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.2939270033226204e-05,
+      "loss": 0.9471,
       "step": 14903
     },
     {
-      "epoch": 0.40932685177556233,
+      "epoch": 0.42292849035187285,
       "grad_norm": 0.0,
-      "learning_rate": 1.335641131566735e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.2938391540381001e-05,
+      "loss": 0.9164,
       "step": 14904
     },
     {
-      "epoch": 0.40935431600340555,
+      "epoch": 0.42295686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 1.3355573387262104e-05,
-      "loss": 1.0909,
+      "learning_rate": 1.2937513022715076e-05,
+      "loss": 0.8838,
       "step": 14905
     },
     {
-      "epoch": 0.40938178023124877,
+      "epoch": 0.42298524404086263,
       "grad_norm": 0.0,
-      "learning_rate": 1.335473543230626e-05,
-      "loss": 0.8811,
+      "learning_rate": 1.2936634480235843e-05,
+      "loss": 0.9711,
       "step": 14906
     },
     {
-      "epoch": 0.40940924445909205,
+      "epoch": 0.42301362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.3353897450806455e-05,
-      "loss": 0.8143,
+      "learning_rate": 1.2935755912950728e-05,
+      "loss": 0.8664,
       "step": 14907
     },
     {
-      "epoch": 0.40943670868693527,
+      "epoch": 0.42304199772985246,
       "grad_norm": 0.0,
-      "learning_rate": 1.3353059442769308e-05,
-      "loss": 0.8688,
+      "learning_rate": 1.2934877320867152e-05,
+      "loss": 0.9849,
       "step": 14908
     },
     {
-      "epoch": 0.4094641729147785,
+      "epoch": 0.4230703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.3352221408201457e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.2933998703992531e-05,
+      "loss": 0.8864,
       "step": 14909
     },
     {
-      "epoch": 0.4094916371426217,
+      "epoch": 0.42309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.3351383347109534e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.2933120062334294e-05,
+      "loss": 0.7795,
       "step": 14910
     },
     {
-      "epoch": 0.409519101370465,
+      "epoch": 0.4231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.3350545259500166e-05,
-      "loss": 0.9187,
+      "learning_rate": 1.2932241395899862e-05,
+      "loss": 0.7975,
       "step": 14911
     },
     {
-      "epoch": 0.4095465655983082,
+      "epoch": 0.423155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.334970714537999e-05,
-      "loss": 1.0014,
+      "learning_rate": 1.2931362704696652e-05,
+      "loss": 0.9157,
       "step": 14912
     },
     {
-      "epoch": 0.40957402982615143,
+      "epoch": 0.4231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.3348869004755626e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.293048398873209e-05,
+      "loss": 0.9119,
       "step": 14913
     },
     {
-      "epoch": 0.40960149405399465,
+      "epoch": 0.4232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.334803083763372e-05,
-      "loss": 0.9317,
+      "learning_rate": 1.2929605248013601e-05,
+      "loss": 0.9165,
       "step": 14914
     },
     {
-      "epoch": 0.40962895828183793,
+      "epoch": 0.4232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.3347192644020894e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.2928726482548602e-05,
+      "loss": 0.8972,
       "step": 14915
     },
     {
-      "epoch": 0.40965642250968115,
+      "epoch": 0.42326901248581156,
       "grad_norm": 0.0,
-      "learning_rate": 1.3346354423923784e-05,
-      "loss": 0.886,
+      "learning_rate": 1.2927847692344514e-05,
+      "loss": 0.9889,
       "step": 14916
     },
     {
-      "epoch": 0.40968388673752437,
+      "epoch": 0.4232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.3345516177349021e-05,
-      "loss": 0.8917,
+      "learning_rate": 1.2926968877408773e-05,
+      "loss": 0.8712,
       "step": 14917
     },
     {
-      "epoch": 0.4097113509653676,
+      "epoch": 0.4233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.3344677904303238e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.2926090037748793e-05,
+      "loss": 0.8884,
       "step": 14918
     },
     {
-      "epoch": 0.40973881519321087,
+      "epoch": 0.42335414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.334383960479307e-05,
-      "loss": 0.8477,
+      "learning_rate": 1.2925211173371994e-05,
+      "loss": 1.0076,
       "step": 14919
     },
     {
-      "epoch": 0.4097662794210541,
+      "epoch": 0.42338251986379116,
       "grad_norm": 0.0,
-      "learning_rate": 1.3343001278825144e-05,
-      "loss": 0.9115,
+      "learning_rate": 1.2924332284285809e-05,
+      "loss": 0.9945,
       "step": 14920
     },
     {
-      "epoch": 0.4097937436488973,
+      "epoch": 0.423410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.33421629264061e-05,
-      "loss": 0.9493,
+      "learning_rate": 1.2923453370497658e-05,
+      "loss": 0.9263,
       "step": 14921
     },
     {
-      "epoch": 0.40982120787674053,
+      "epoch": 0.42343927355278094,
       "grad_norm": 0.0,
-      "learning_rate": 1.3341324547542566e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.2922574432014965e-05,
+      "loss": 0.856,
       "step": 14922
     },
     {
-      "epoch": 0.40984867210458376,
+      "epoch": 0.4234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.334048614224118e-05,
-      "loss": 0.8006,
+      "learning_rate": 1.2921695468845152e-05,
+      "loss": 0.9477,
       "step": 14923
     },
     {
-      "epoch": 0.40987613633242703,
+      "epoch": 0.4234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.3339647710508574e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.2920816480995645e-05,
+      "loss": 0.8925,
       "step": 14924
     },
     {
-      "epoch": 0.40990360056027025,
+      "epoch": 0.4235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.3338809252351379e-05,
-      "loss": 0.811,
+      "learning_rate": 1.2919937468473873e-05,
+      "loss": 0.8782,
       "step": 14925
     },
     {
-      "epoch": 0.4099310647881135,
+      "epoch": 0.4235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.3337970767776233e-05,
-      "loss": 0.9699,
+      "learning_rate": 1.2919058431287257e-05,
+      "loss": 0.8806,
       "step": 14926
     },
     {
-      "epoch": 0.4099585290159567,
+      "epoch": 0.4235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.3337132256789772e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.2918179369443224e-05,
+      "loss": 0.8265,
       "step": 14927
     },
     {
-      "epoch": 0.4099859932438,
+      "epoch": 0.42360953461975026,
       "grad_norm": 0.0,
-      "learning_rate": 1.3336293719398623e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.2917300282949199e-05,
+      "loss": 0.9127,
       "step": 14928
     },
     {
-      "epoch": 0.4100134574716432,
+      "epoch": 0.4236379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.3335455155609429e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.2916421171812605e-05,
+      "loss": 0.8773,
       "step": 14929
     },
     {
-      "epoch": 0.4100409216994864,
+      "epoch": 0.4236662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.3334616565428819e-05,
-      "loss": 0.903,
+      "learning_rate": 1.2915542036040871e-05,
+      "loss": 0.9404,
       "step": 14930
     },
     {
-      "epoch": 0.41006838592732964,
+      "epoch": 0.42369466515323495,
       "grad_norm": 0.0,
-      "learning_rate": 1.3333777948863436e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.2914662875641425e-05,
+      "loss": 0.8435,
       "step": 14931
     },
     {
-      "epoch": 0.4100958501551729,
+      "epoch": 0.42372304199772987,
       "grad_norm": 0.0,
-      "learning_rate": 1.3332939305919907e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.2913783690621688e-05,
+      "loss": 0.9475,
       "step": 14932
     },
     {
-      "epoch": 0.41012331438301614,
+      "epoch": 0.4237514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.333210063660487e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.2912904480989088e-05,
+      "loss": 1.0021,
       "step": 14933
     },
     {
-      "epoch": 0.41015077861085936,
+      "epoch": 0.42377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.3331261940924964e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.2912025246751054e-05,
+      "loss": 0.9116,
       "step": 14934
     },
     {
-      "epoch": 0.4101782428387026,
+      "epoch": 0.42380817253121456,
       "grad_norm": 0.0,
-      "learning_rate": 1.3330423218886824e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.2911145987915015e-05,
+      "loss": 0.9799,
       "step": 14935
     },
     {
-      "epoch": 0.4102057070665458,
+      "epoch": 0.4238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.3329584470497083e-05,
-      "loss": 1.0072,
+      "learning_rate": 1.291026670448839e-05,
+      "loss": 0.8702,
       "step": 14936
     },
     {
-      "epoch": 0.4102331712943891,
+      "epoch": 0.42386492622020433,
       "grad_norm": 0.0,
-      "learning_rate": 1.332874569576238e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.2909387396478614e-05,
+      "loss": 0.9053,
       "step": 14937
     },
     {
-      "epoch": 0.4102606355222323,
+      "epoch": 0.4238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.3327906894689353e-05,
-      "loss": 0.923,
+      "learning_rate": 1.2908508063893112e-05,
+      "loss": 1.0275,
       "step": 14938
     },
     {
-      "epoch": 0.4102880997500755,
+      "epoch": 0.4239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.3327068067284639e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.2907628706739312e-05,
+      "loss": 0.8908,
       "step": 14939
     },
     {
-      "epoch": 0.41031556397791874,
+      "epoch": 0.42395005675368896,
       "grad_norm": 0.0,
-      "learning_rate": 1.332622921355487e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.2906749325024638e-05,
+      "loss": 0.8469,
       "step": 14940
     },
     {
-      "epoch": 0.410343028205762,
+      "epoch": 0.4239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.332539033350669e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.2905869918756526e-05,
+      "loss": 0.7975,
       "step": 14941
     },
     {
-      "epoch": 0.41037049243360524,
+      "epoch": 0.4240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.3324551427146735e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.29049904879424e-05,
+      "loss": 0.8658,
       "step": 14942
     },
     {
-      "epoch": 0.41039795666144846,
+      "epoch": 0.42403518728717365,
       "grad_norm": 0.0,
-      "learning_rate": 1.3323712494481638e-05,
-      "loss": 1.009,
+      "learning_rate": 1.2904111032589688e-05,
+      "loss": 0.9394,
       "step": 14943
     },
     {
-      "epoch": 0.4104254208892917,
+      "epoch": 0.42406356413166857,
       "grad_norm": 0.0,
-      "learning_rate": 1.332287353551804e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.2903231552705819e-05,
+      "loss": 0.9044,
       "step": 14944
     },
     {
-      "epoch": 0.41045288511713496,
+      "epoch": 0.42409194097616343,
       "grad_norm": 0.0,
-      "learning_rate": 1.332203455026258e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.2902352048298224e-05,
+      "loss": 0.8726,
       "step": 14945
     },
     {
-      "epoch": 0.4104803493449782,
+      "epoch": 0.42412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.3321195538721896e-05,
-      "loss": 0.8953,
+      "learning_rate": 1.2901472519374327e-05,
+      "loss": 0.9056,
       "step": 14946
     },
     {
-      "epoch": 0.4105078135728214,
+      "epoch": 0.42414869466515326,
       "grad_norm": 0.0,
-      "learning_rate": 1.3320356500902629e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.2900592965941563e-05,
+      "loss": 0.8937,
       "step": 14947
     },
     {
-      "epoch": 0.4105352778006646,
+      "epoch": 0.4241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.3319517436811411e-05,
-      "loss": 0.8977,
+      "learning_rate": 1.2899713388007362e-05,
+      "loss": 0.9642,
       "step": 14948
     },
     {
-      "epoch": 0.41056274202850784,
+      "epoch": 0.42420544835414303,
       "grad_norm": 0.0,
-      "learning_rate": 1.3318678346454885e-05,
-      "loss": 0.866,
+      "learning_rate": 1.289883378557915e-05,
+      "loss": 0.9146,
       "step": 14949
     },
     {
-      "epoch": 0.4105902062563511,
+      "epoch": 0.4242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.3317839229839693e-05,
-      "loss": 0.8855,
+      "learning_rate": 1.2897954158664358e-05,
+      "loss": 0.8973,
       "step": 14950
     },
     {
-      "epoch": 0.41061767048419434,
+      "epoch": 0.4242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.3317000086972474e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.289707450727042e-05,
+      "loss": 0.8717,
       "step": 14951
     },
     {
-      "epoch": 0.41064513471203756,
+      "epoch": 0.4242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.3316160917859859e-05,
-      "loss": 0.8756,
+      "learning_rate": 1.289619483140476e-05,
+      "loss": 0.8213,
       "step": 14952
     },
     {
-      "epoch": 0.4106725989398808,
+      "epoch": 0.4243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.3315321722508501e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.2895315131074812e-05,
+      "loss": 0.8425,
       "step": 14953
     },
     {
-      "epoch": 0.41070006316772406,
+      "epoch": 0.4243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.331448250092503e-05,
-      "loss": 1.0333,
+      "learning_rate": 1.289443540628801e-05,
+      "loss": 0.8955,
       "step": 14954
     },
     {
-      "epoch": 0.4107275273955673,
+      "epoch": 0.42437570942111236,
       "grad_norm": 0.0,
-      "learning_rate": 1.331364325311609e-05,
-      "loss": 0.933,
+      "learning_rate": 1.289355565705178e-05,
+      "loss": 0.9148,
       "step": 14955
     },
     {
-      "epoch": 0.4107549916234105,
+      "epoch": 0.42440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.331280397908832e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.2892675883373555e-05,
+      "loss": 0.9412,
       "step": 14956
     },
     {
-      "epoch": 0.4107824558512537,
+      "epoch": 0.42443246311010213,
       "grad_norm": 0.0,
-      "learning_rate": 1.3311964678848361e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.2891796085260766e-05,
+      "loss": 0.9294,
       "step": 14957
     },
     {
-      "epoch": 0.410809920079097,
+      "epoch": 0.42446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.3311125352402856e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.2890916262720848e-05,
+      "loss": 0.8941,
       "step": 14958
     },
     {
-      "epoch": 0.4108373843069402,
+      "epoch": 0.42448921679909196,
       "grad_norm": 0.0,
-      "learning_rate": 1.3310285999758445e-05,
-      "loss": 0.9726,
+      "learning_rate": 1.2890036415761232e-05,
+      "loss": 0.9721,
       "step": 14959
     },
     {
-      "epoch": 0.41086484853478344,
+      "epoch": 0.4245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.3309446620921768e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.2889156544389342e-05,
+      "loss": 0.8649,
       "step": 14960
     },
     {
-      "epoch": 0.41089231276262667,
+      "epoch": 0.42454597048808174,
       "grad_norm": 0.0,
-      "learning_rate": 1.3308607215899468e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.288827664861262e-05,
+      "loss": 1.0153,
       "step": 14961
     },
     {
-      "epoch": 0.4109197769904699,
+      "epoch": 0.4245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.3307767784698186e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.2887396728438498e-05,
+      "loss": 0.8404,
       "step": 14962
     },
     {
-      "epoch": 0.41094724121831316,
+      "epoch": 0.4246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.3306928327324564e-05,
-      "loss": 0.8417,
+      "learning_rate": 1.2886516783874403e-05,
+      "loss": 0.9088,
       "step": 14963
     },
     {
-      "epoch": 0.4109747054461564,
+      "epoch": 0.4246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.3306088843785243e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.2885636814927769e-05,
+      "loss": 0.881,
       "step": 14964
     },
     {
-      "epoch": 0.4110021696739996,
+      "epoch": 0.4246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.3305249334086869e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.2884756821606037e-05,
+      "loss": 0.9255,
       "step": 14965
     },
     {
-      "epoch": 0.41102963390184283,
+      "epoch": 0.4246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.330440979823608e-05,
-      "loss": 0.8528,
+      "learning_rate": 1.288387680391663e-05,
+      "loss": 1.0188,
       "step": 14966
     },
     {
-      "epoch": 0.4110570981296861,
+      "epoch": 0.42471623155505106,
       "grad_norm": 0.0,
-      "learning_rate": 1.330357023623952e-05,
-      "loss": 0.8326,
+      "learning_rate": 1.2882996761866985e-05,
+      "loss": 0.9317,
       "step": 14967
     },
     {
-      "epoch": 0.4110845623575293,
+      "epoch": 0.424744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.3302730648103833e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.2882116695464542e-05,
+      "loss": 0.9773,
       "step": 14968
     },
     {
-      "epoch": 0.41111202658537255,
+      "epoch": 0.42477298524404083,
       "grad_norm": 0.0,
-      "learning_rate": 1.3301891033835664e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.2881236604716728e-05,
+      "loss": 0.928,
       "step": 14969
     },
     {
-      "epoch": 0.41113949081321577,
+      "epoch": 0.42480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.3301051393441654e-05,
-      "loss": 1.0616,
+      "learning_rate": 1.2880356489630974e-05,
+      "loss": 0.9995,
       "step": 14970
     },
     {
-      "epoch": 0.41116695504105905,
+      "epoch": 0.42482973893303067,
       "grad_norm": 0.0,
-      "learning_rate": 1.3300211726928444e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.2879476350214724e-05,
+      "loss": 0.9616,
       "step": 14971
     },
     {
-      "epoch": 0.41119441926890227,
+      "epoch": 0.4248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.329937203430268e-05,
-      "loss": 1.0719,
+      "learning_rate": 1.2878596186475408e-05,
+      "loss": 0.8861,
       "step": 14972
     },
     {
-      "epoch": 0.4112218834967455,
+      "epoch": 0.42488649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 1.3298532315571007e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.2877715998420457e-05,
+      "loss": 0.9871,
       "step": 14973
     },
     {
-      "epoch": 0.4112493477245887,
+      "epoch": 0.4249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.3297692570740071e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.2876835786057313e-05,
+      "loss": 0.9198,
       "step": 14974
     },
     {
-      "epoch": 0.41127681195243193,
+      "epoch": 0.4249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.3296852799816514e-05,
-      "loss": 0.9562,
+      "learning_rate": 1.2875955549393408e-05,
+      "loss": 0.9006,
       "step": 14975
     },
     {
-      "epoch": 0.4113042761802752,
+      "epoch": 0.42497162315550513,
       "grad_norm": 0.0,
-      "learning_rate": 1.3296013002806977e-05,
-      "loss": 0.8223,
+      "learning_rate": 1.2875075288436177e-05,
+      "loss": 0.9341,
       "step": 14976
     },
     {
-      "epoch": 0.41133174040811843,
+      "epoch": 0.425,
       "grad_norm": 0.0,
-      "learning_rate": 1.3295173179718113e-05,
-      "loss": 0.9339,
+      "learning_rate": 1.2874195003193052e-05,
+      "loss": 0.9135,
       "step": 14977
     },
     {
-      "epoch": 0.41135920463596165,
+      "epoch": 0.4250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.3294333330556559e-05,
-      "loss": 1.0271,
+      "learning_rate": 1.2873314693671475e-05,
+      "loss": 0.9138,
       "step": 14978
     },
     {
-      "epoch": 0.41138666886380487,
+      "epoch": 0.42505675368898976,
       "grad_norm": 0.0,
-      "learning_rate": 1.3293493455328963e-05,
-      "loss": 0.8423,
+      "learning_rate": 1.287243435987888e-05,
+      "loss": 0.8701,
       "step": 14979
     },
     {
-      "epoch": 0.41141413309164815,
+      "epoch": 0.4250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.3292653554041974e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.2871554001822701e-05,
+      "loss": 0.9718,
       "step": 14980
     },
     {
-      "epoch": 0.41144159731949137,
+      "epoch": 0.4251135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.3291813626702233e-05,
-      "loss": 0.9975,
+      "learning_rate": 1.2870673619510382e-05,
+      "loss": 0.9157,
       "step": 14981
     },
     {
-      "epoch": 0.4114690615473346,
+      "epoch": 0.42514188422247445,
       "grad_norm": 0.0,
-      "learning_rate": 1.3290973673316388e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.2869793212949349e-05,
+      "loss": 0.9651,
       "step": 14982
     },
     {
-      "epoch": 0.4114965257751778,
+      "epoch": 0.42517026106696937,
       "grad_norm": 0.0,
-      "learning_rate": 1.3290133693891085e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.2868912782147043e-05,
+      "loss": 0.9903,
       "step": 14983
     },
     {
-      "epoch": 0.4115239900030211,
+      "epoch": 0.42519863791146423,
       "grad_norm": 0.0,
-      "learning_rate": 1.3289293688432968e-05,
-      "loss": 1.024,
+      "learning_rate": 1.2868032327110904e-05,
+      "loss": 0.9156,
       "step": 14984
     },
     {
-      "epoch": 0.4115514542308643,
+      "epoch": 0.42522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.3288453656948684e-05,
-      "loss": 0.9651,
+      "learning_rate": 1.2867151847848364e-05,
+      "loss": 0.9073,
       "step": 14985
     },
     {
-      "epoch": 0.41157891845870753,
+      "epoch": 0.425255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.3287613599444882e-05,
-      "loss": 0.937,
+      "learning_rate": 1.286627134436687e-05,
+      "loss": 0.9838,
       "step": 14986
     },
     {
-      "epoch": 0.41160638268655075,
+      "epoch": 0.4252837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.3286773515928211e-05,
-      "loss": 0.8674,
+      "learning_rate": 1.2865390816673846e-05,
+      "loss": 1.0334,
       "step": 14987
     },
     {
-      "epoch": 0.411633846914394,
+      "epoch": 0.42531214528944383,
       "grad_norm": 0.0,
-      "learning_rate": 1.3285933406405312e-05,
-      "loss": 0.983,
+      "learning_rate": 1.2864510264776739e-05,
+      "loss": 0.9475,
       "step": 14988
     },
     {
-      "epoch": 0.41166131114223725,
+      "epoch": 0.4253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.3285093270882834e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.2863629688682988e-05,
+      "loss": 0.8888,
       "step": 14989
     },
     {
-      "epoch": 0.4116887753700805,
+      "epoch": 0.4253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.328425310936743e-05,
-      "loss": 0.8614,
+      "learning_rate": 1.2862749088400026e-05,
+      "loss": 0.8562,
       "step": 14990
     },
     {
-      "epoch": 0.4117162395979237,
+      "epoch": 0.42539727582292847,
       "grad_norm": 0.0,
-      "learning_rate": 1.328341292186574e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.2861868463935294e-05,
+      "loss": 0.9876,
       "step": 14991
     },
     {
-      "epoch": 0.4117437038257669,
+      "epoch": 0.4254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.3282572708384413e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.2860987815296233e-05,
+      "loss": 0.9647,
       "step": 14992
     },
     {
-      "epoch": 0.4117711680536102,
+      "epoch": 0.4254540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.32817324689301e-05,
-      "loss": 0.9775,
+      "learning_rate": 1.2860107142490274e-05,
+      "loss": 0.9578,
       "step": 14993
     },
     {
-      "epoch": 0.4117986322814534,
+      "epoch": 0.42548240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.328089220350945e-05,
-      "loss": 0.8925,
+      "learning_rate": 1.2859226445524865e-05,
+      "loss": 0.9148,
       "step": 14994
     },
     {
-      "epoch": 0.41182609650929664,
+      "epoch": 0.42551078320090807,
       "grad_norm": 0.0,
-      "learning_rate": 1.3280051912129111e-05,
-      "loss": 1.0312,
+      "learning_rate": 1.285834572440744e-05,
+      "loss": 0.8826,
       "step": 14995
     },
     {
-      "epoch": 0.41185356073713986,
+      "epoch": 0.42553916004540293,
       "grad_norm": 0.0,
-      "learning_rate": 1.327921159479573e-05,
-      "loss": 0.8744,
+      "learning_rate": 1.2857464979145442e-05,
+      "loss": 0.9122,
       "step": 14996
     },
     {
-      "epoch": 0.41188102496498313,
+      "epoch": 0.42556753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.3278371251515956e-05,
-      "loss": 0.9893,
+      "learning_rate": 1.2856584209746306e-05,
+      "loss": 0.8412,
       "step": 14997
     },
     {
-      "epoch": 0.41190848919282635,
+      "epoch": 0.42559591373439276,
       "grad_norm": 0.0,
-      "learning_rate": 1.3277530882296436e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.2855703416217478e-05,
+      "loss": 0.8884,
       "step": 14998
     },
     {
-      "epoch": 0.4119359534206696,
+      "epoch": 0.4256242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.3276690487143826e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.2854822598566394e-05,
+      "loss": 0.891,
       "step": 14999
     },
     {
-      "epoch": 0.4119634176485128,
+      "epoch": 0.42565266742338254,
       "grad_norm": 0.0,
-      "learning_rate": 1.3275850066064773e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.2853941756800494e-05,
+      "loss": 0.9268,
       "step": 15000
     },
     {
-      "epoch": 0.411990881876356,
+      "epoch": 0.4256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.3275009619065919e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.2853060890927217e-05,
+      "loss": 0.9172,
       "step": 15001
     },
     {
-      "epoch": 0.4120183461041993,
+      "epoch": 0.4257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.3274169146153928e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.285218000095401e-05,
+      "loss": 0.879,
       "step": 15002
     },
     {
-      "epoch": 0.4120458103320425,
+      "epoch": 0.42573779795686717,
       "grad_norm": 0.0,
-      "learning_rate": 1.3273328647335437e-05,
-      "loss": 0.9338,
+      "learning_rate": 1.2851299086888313e-05,
+      "loss": 0.9872,
       "step": 15003
     },
     {
-      "epoch": 0.41207327455988574,
+      "epoch": 0.4257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.3272488122617104e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.285041814873756e-05,
+      "loss": 0.9607,
       "step": 15004
     },
     {
-      "epoch": 0.41210073878772896,
+      "epoch": 0.425794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.3271647572005575e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.28495371865092e-05,
+      "loss": 0.9434,
       "step": 15005
     },
     {
-      "epoch": 0.41212820301557224,
+      "epoch": 0.42582292849035186,
       "grad_norm": 0.0,
-      "learning_rate": 1.3270806995507506e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.284865620021067e-05,
+      "loss": 0.9925,
       "step": 15006
     },
     {
-      "epoch": 0.41215566724341546,
+      "epoch": 0.4258513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.3269966393129543e-05,
-      "loss": 1.0889,
+      "learning_rate": 1.2847775189849412e-05,
+      "loss": 0.8271,
       "step": 15007
     },
     {
-      "epoch": 0.4121831314712587,
+      "epoch": 0.42587968217934163,
       "grad_norm": 0.0,
-      "learning_rate": 1.3269125764878339e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.2846894155432868e-05,
+      "loss": 0.9417,
       "step": 15008
     },
     {
-      "epoch": 0.4122105956991019,
+      "epoch": 0.42590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.3268285110760548e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.2846013096968483e-05,
+      "loss": 0.7647,
       "step": 15009
     },
     {
-      "epoch": 0.4122380599269452,
+      "epoch": 0.42593643586833146,
       "grad_norm": 0.0,
-      "learning_rate": 1.3267444430782818e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.28451320144637e-05,
+      "loss": 0.9508,
       "step": 15010
     },
     {
-      "epoch": 0.4122655241547884,
+      "epoch": 0.4259648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.32666037249518e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.2844250907925953e-05,
+      "loss": 0.9322,
       "step": 15011
     },
     {
-      "epoch": 0.4122929883826316,
+      "epoch": 0.42599318955732124,
       "grad_norm": 0.0,
-      "learning_rate": 1.3265762993274152e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.2843369777362695e-05,
+      "loss": 1.0329,
       "step": 15012
     },
     {
-      "epoch": 0.41232045261047484,
+      "epoch": 0.4260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.3264922235756519e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.2842488622781364e-05,
+      "loss": 0.9601,
       "step": 15013
     },
     {
-      "epoch": 0.4123479168383181,
+      "epoch": 0.426049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.3264081452405558e-05,
-      "loss": 0.9827,
+      "learning_rate": 1.2841607444189402e-05,
+      "loss": 0.9922,
       "step": 15014
     },
     {
-      "epoch": 0.41237538106616134,
+      "epoch": 0.42607832009080593,
       "grad_norm": 0.0,
-      "learning_rate": 1.3263240643227921e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.2840726241594258e-05,
+      "loss": 1.0216,
       "step": 15015
     },
     {
-      "epoch": 0.41240284529400456,
+      "epoch": 0.4261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.3262399808230257e-05,
-      "loss": 0.8289,
+      "learning_rate": 1.2839845015003368e-05,
+      "loss": 0.7919,
       "step": 15016
     },
     {
-      "epoch": 0.4124303095218478,
+      "epoch": 0.4261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.3261558947419222e-05,
-      "loss": 0.8837,
+      "learning_rate": 1.283896376442418e-05,
+      "loss": 0.929,
       "step": 15017
     },
     {
-      "epoch": 0.412457773749691,
+      "epoch": 0.42616345062429056,
       "grad_norm": 0.0,
-      "learning_rate": 1.3260718060801473e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.2838082489864137e-05,
+      "loss": 0.962,
       "step": 15018
     },
     {
-      "epoch": 0.4124852379775343,
+      "epoch": 0.4261918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.3259877148383657e-05,
-      "loss": 0.9452,
+      "learning_rate": 1.2837201191330686e-05,
+      "loss": 0.8789,
       "step": 15019
     },
     {
-      "epoch": 0.4125127022053775,
+      "epoch": 0.42622020431328034,
       "grad_norm": 0.0,
-      "learning_rate": 1.3259036210172428e-05,
-      "loss": 0.8693,
+      "learning_rate": 1.2836319868831269e-05,
+      "loss": 0.9013,
       "step": 15020
     },
     {
-      "epoch": 0.4125401664332207,
+      "epoch": 0.42624858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.3258195246174448e-05,
-      "loss": 1.036,
+      "learning_rate": 1.2835438522373326e-05,
+      "loss": 0.9543,
       "step": 15021
     },
     {
-      "epoch": 0.41256763066106394,
+      "epoch": 0.42627695800227017,
       "grad_norm": 0.0,
-      "learning_rate": 1.3257354256396362e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.283455715196431e-05,
+      "loss": 0.8925,
       "step": 15022
     },
     {
-      "epoch": 0.4125950948889072,
+      "epoch": 0.426305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.3256513240844826e-05,
-      "loss": 0.9556,
+      "learning_rate": 1.2833675757611666e-05,
+      "loss": 0.9712,
       "step": 15023
     },
     {
-      "epoch": 0.41262255911675044,
+      "epoch": 0.42633371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 1.3255672199526498e-05,
-      "loss": 0.9507,
+      "learning_rate": 1.2832794339322827e-05,
+      "loss": 0.9122,
       "step": 15024
     },
     {
-      "epoch": 0.41265002334459366,
+      "epoch": 0.4263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.3254831132448027e-05,
-      "loss": 0.8429,
+      "learning_rate": 1.2831912897105252e-05,
+      "loss": 1.0062,
       "step": 15025
     },
     {
-      "epoch": 0.4126774875724369,
+      "epoch": 0.4263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.3253990039616076e-05,
-      "loss": 0.9029,
+      "learning_rate": 1.283103143096638e-05,
+      "loss": 0.8428,
       "step": 15026
     },
     {
-      "epoch": 0.41270495180028016,
+      "epoch": 0.42641884222474463,
       "grad_norm": 0.0,
-      "learning_rate": 1.3253148921037292e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.283014994091366e-05,
+      "loss": 0.8755,
       "step": 15027
     },
     {
-      "epoch": 0.4127324160281234,
+      "epoch": 0.4264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.3252307776718336e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.2829268426954535e-05,
+      "loss": 0.9138,
       "step": 15028
     },
     {
-      "epoch": 0.4127598802559666,
+      "epoch": 0.4264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.3251466606665856e-05,
-      "loss": 0.957,
+      "learning_rate": 1.2828386889096453e-05,
+      "loss": 0.8324,
       "step": 15029
     },
     {
-      "epoch": 0.4127873444838098,
+      "epoch": 0.42650397275822927,
       "grad_norm": 0.0,
-      "learning_rate": 1.3250625410886518e-05,
-      "loss": 0.9116,
+      "learning_rate": 1.282750532734686e-05,
+      "loss": 0.9758,
       "step": 15030
     },
     {
-      "epoch": 0.41281480871165305,
+      "epoch": 0.4265323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.3249784189386971e-05,
-      "loss": 0.9209,
+      "learning_rate": 1.28266237417132e-05,
+      "loss": 0.9473,
       "step": 15031
     },
     {
-      "epoch": 0.4128422729394963,
+      "epoch": 0.4265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.3248942942173873e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.2825742132202925e-05,
+      "loss": 0.9557,
       "step": 15032
     },
     {
-      "epoch": 0.41286973716733955,
+      "epoch": 0.42658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.3248101669253881e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.282486049882348e-05,
+      "loss": 0.9701,
       "step": 15033
     },
     {
-      "epoch": 0.41289720139518277,
+      "epoch": 0.42661748013620887,
       "grad_norm": 0.0,
-      "learning_rate": 1.324726037063365e-05,
-      "loss": 0.927,
+      "learning_rate": 1.2823978841582308e-05,
+      "loss": 0.8615,
       "step": 15034
     },
     {
-      "epoch": 0.412924665623026,
+      "epoch": 0.42664585698070373,
       "grad_norm": 0.0,
-      "learning_rate": 1.3246419046319833e-05,
-      "loss": 1.0356,
+      "learning_rate": 1.2823097160486861e-05,
+      "loss": 0.8932,
       "step": 15035
     },
     {
-      "epoch": 0.41295212985086927,
+      "epoch": 0.42667423382519865,
       "grad_norm": 0.0,
-      "learning_rate": 1.3245577696319094e-05,
-      "loss": 1.0829,
+      "learning_rate": 1.2822215455544587e-05,
+      "loss": 0.7378,
       "step": 15036
     },
     {
-      "epoch": 0.4129795940787125,
+      "epoch": 0.4267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.3244736320638087e-05,
-      "loss": 1.0586,
+      "learning_rate": 1.282133372676293e-05,
+      "loss": 0.9389,
       "step": 15037
     },
     {
-      "epoch": 0.4130070583065557,
+      "epoch": 0.4267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.324389491928347e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.2820451974149341e-05,
+      "loss": 0.8498,
       "step": 15038
     },
     {
-      "epoch": 0.41303452253439893,
+      "epoch": 0.42675936435868334,
       "grad_norm": 0.0,
-      "learning_rate": 1.3243053492261901e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.2819570197711269e-05,
+      "loss": 0.9473,
       "step": 15039
     },
     {
-      "epoch": 0.4130619867622422,
+      "epoch": 0.4267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.3242212039580033e-05,
-      "loss": 0.8788,
+      "learning_rate": 1.2818688397456159e-05,
+      "loss": 0.7496,
       "step": 15040
     },
     {
-      "epoch": 0.4130894509900854,
+      "epoch": 0.4268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.3241370561244527e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.281780657339146e-05,
+      "loss": 0.9398,
       "step": 15041
     },
     {
-      "epoch": 0.41311691521792865,
+      "epoch": 0.42684449489216797,
       "grad_norm": 0.0,
-      "learning_rate": 1.3240529057262045e-05,
-      "loss": 0.954,
+      "learning_rate": 1.2816924725524624e-05,
+      "loss": 0.9817,
       "step": 15042
     },
     {
-      "epoch": 0.41314437944577187,
+      "epoch": 0.4268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.3239687527639243e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.28160428538631e-05,
+      "loss": 0.9376,
       "step": 15043
     },
     {
-      "epoch": 0.4131718436736151,
+      "epoch": 0.4269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.3238845972382772e-05,
-      "loss": 0.8759,
+      "learning_rate": 1.2815160958414332e-05,
+      "loss": 0.9199,
       "step": 15044
     },
     {
-      "epoch": 0.41319930790145837,
+      "epoch": 0.42692962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 1.3238004391499304e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.2814279039185775e-05,
+      "loss": 0.9589,
       "step": 15045
     },
     {
-      "epoch": 0.4132267721293016,
+      "epoch": 0.4269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.3237162784995487e-05,
-      "loss": 0.8333,
+      "learning_rate": 1.2813397096184876e-05,
+      "loss": 0.8359,
       "step": 15046
     },
     {
-      "epoch": 0.4132542363571448,
+      "epoch": 0.42698637911464243,
       "grad_norm": 0.0,
-      "learning_rate": 1.3236321152877986e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.2812515129419085e-05,
+      "loss": 0.908,
       "step": 15047
     },
     {
-      "epoch": 0.41328170058498803,
+      "epoch": 0.42701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.3235479495153457e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.2811633138895851e-05,
+      "loss": 0.9573,
       "step": 15048
     },
     {
-      "epoch": 0.4133091648128313,
+      "epoch": 0.42704313280363226,
       "grad_norm": 0.0,
-      "learning_rate": 1.3234637811828562e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.2810751124622626e-05,
+      "loss": 0.9765,
       "step": 15049
     },
     {
-      "epoch": 0.41333662904067453,
+      "epoch": 0.4270715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.3233796102909957e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.2809869086606863e-05,
+      "loss": 0.9867,
       "step": 15050
     },
     {
-      "epoch": 0.41336409326851775,
+      "epoch": 0.42709988649262204,
       "grad_norm": 0.0,
-      "learning_rate": 1.323295436840431e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.2808987024856006e-05,
+      "loss": 0.7929,
       "step": 15051
     },
     {
-      "epoch": 0.413391557496361,
+      "epoch": 0.4271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.323211260831827e-05,
-      "loss": 0.9762,
+      "learning_rate": 1.2808104939377512e-05,
+      "loss": 0.8848,
       "step": 15052
     },
     {
-      "epoch": 0.41341902172420425,
+      "epoch": 0.4271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.3231270822658506e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.280722283017883e-05,
+      "loss": 0.8683,
       "step": 15053
     },
     {
-      "epoch": 0.41344648595204747,
+      "epoch": 0.4271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.3230429011431674e-05,
-      "loss": 1.0633,
+      "learning_rate": 1.2806340697267406e-05,
+      "loss": 0.9277,
       "step": 15054
     },
     {
-      "epoch": 0.4134739501798907,
+      "epoch": 0.4272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.3229587174644439e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.2805458540650696e-05,
+      "loss": 0.9983,
       "step": 15055
     },
     {
-      "epoch": 0.4135014144077339,
+      "epoch": 0.4272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.3228745312303457e-05,
-      "loss": 0.7478,
+      "learning_rate": 1.2804576360336156e-05,
+      "loss": 0.8937,
       "step": 15056
     },
     {
-      "epoch": 0.41352887863557714,
+      "epoch": 0.42727014755959136,
       "grad_norm": 0.0,
-      "learning_rate": 1.3227903424415391e-05,
-      "loss": 0.961,
+      "learning_rate": 1.2803694156331233e-05,
+      "loss": 0.8428,
       "step": 15057
     },
     {
-      "epoch": 0.4135563428634204,
+      "epoch": 0.4272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.3227061510986902e-05,
-      "loss": 0.7562,
+      "learning_rate": 1.2802811928643375e-05,
+      "loss": 0.8476,
       "step": 15058
     },
     {
-      "epoch": 0.41358380709126363,
+      "epoch": 0.42732690124858114,
       "grad_norm": 0.0,
-      "learning_rate": 1.3226219572024657e-05,
-      "loss": 0.9427,
+      "learning_rate": 1.2801929677280043e-05,
+      "loss": 0.9206,
       "step": 15059
     },
     {
-      "epoch": 0.41361127131910685,
+      "epoch": 0.42735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.3225377607535307e-05,
-      "loss": 0.9497,
+      "learning_rate": 1.2801047402248686e-05,
+      "loss": 1.0377,
       "step": 15060
     },
     {
-      "epoch": 0.4136387355469501,
+      "epoch": 0.42738365493757097,
       "grad_norm": 0.0,
-      "learning_rate": 1.3224535617525522e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.280016510355675e-05,
+      "loss": 0.8101,
       "step": 15061
     },
     {
-      "epoch": 0.41366619977479335,
+      "epoch": 0.4274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.3223693602001966e-05,
-      "loss": 0.8589,
+      "learning_rate": 1.2799282781211698e-05,
+      "loss": 0.857,
       "step": 15062
     },
     {
-      "epoch": 0.4136936640026366,
+      "epoch": 0.42744040862656074,
       "grad_norm": 0.0,
-      "learning_rate": 1.3222851560971294e-05,
-      "loss": 0.8709,
+      "learning_rate": 1.2798400435220977e-05,
+      "loss": 0.8711,
       "step": 15063
     },
     {
-      "epoch": 0.4137211282304798,
+      "epoch": 0.4274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.3222009494440171e-05,
-      "loss": 0.8356,
+      "learning_rate": 1.2797518065592043e-05,
+      "loss": 1.0103,
       "step": 15064
     },
     {
-      "epoch": 0.413748592458323,
+      "epoch": 0.4274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.3221167402415263e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.2796635672332345e-05,
+      "loss": 0.9278,
       "step": 15065
     },
     {
-      "epoch": 0.4137760566861663,
+      "epoch": 0.4275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.3220325284903228e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.2795753255449343e-05,
+      "loss": 0.8594,
       "step": 15066
     },
     {
-      "epoch": 0.4138035209140095,
+      "epoch": 0.4275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.3219483141910736e-05,
-      "loss": 0.915,
+      "learning_rate": 1.2794870814950486e-05,
+      "loss": 0.8272,
       "step": 15067
     },
     {
-      "epoch": 0.41383098514185274,
+      "epoch": 0.4275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.3218640973444446e-05,
-      "loss": 0.8639,
+      "learning_rate": 1.2793988350843232e-05,
+      "loss": 0.8577,
       "step": 15068
     },
     {
-      "epoch": 0.41385844936969596,
+      "epoch": 0.42761066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.3217798779511018e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.279310586313503e-05,
+      "loss": 0.9853,
       "step": 15069
     },
     {
-      "epoch": 0.4138859135975392,
+      "epoch": 0.427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.3216956560117122e-05,
-      "loss": 0.8845,
+      "learning_rate": 1.2792223351833338e-05,
+      "loss": 0.9066,
       "step": 15070
     },
     {
-      "epoch": 0.41391337782538246,
+      "epoch": 0.42766742338251984,
       "grad_norm": 0.0,
-      "learning_rate": 1.321611431526942e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.279134081694561e-05,
+      "loss": 1.0106,
       "step": 15071
     },
     {
-      "epoch": 0.4139408420532257,
+      "epoch": 0.42769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.3215272044974573e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.27904582584793e-05,
+      "loss": 0.8934,
       "step": 15072
     },
     {
-      "epoch": 0.4139683062810689,
+      "epoch": 0.42772417707150967,
       "grad_norm": 0.0,
-      "learning_rate": 1.321442974923925e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.2789575676441866e-05,
+      "loss": 0.9336,
       "step": 15073
     },
     {
-      "epoch": 0.4139957705089121,
+      "epoch": 0.42775255391600453,
       "grad_norm": 0.0,
-      "learning_rate": 1.3213587428070114e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.2788693070840758e-05,
+      "loss": 0.9509,
       "step": 15074
     },
     {
-      "epoch": 0.4140232347367554,
+      "epoch": 0.42778093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 1.321274508147383e-05,
-      "loss": 0.8754,
+      "learning_rate": 1.2787810441683435e-05,
+      "loss": 0.8844,
       "step": 15075
     },
     {
-      "epoch": 0.4140506989645986,
+      "epoch": 0.4278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.321190270945706e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.2786927788977353e-05,
+      "loss": 0.9508,
       "step": 15076
     },
     {
-      "epoch": 0.41407816319244184,
+      "epoch": 0.4278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.3211060312026475e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.2786045112729964e-05,
+      "loss": 0.7767,
       "step": 15077
     },
     {
-      "epoch": 0.41410562742028506,
+      "epoch": 0.42786606129398413,
       "grad_norm": 0.0,
-      "learning_rate": 1.3210217889188734e-05,
-      "loss": 1.0282,
+      "learning_rate": 1.2785162412948728e-05,
+      "loss": 0.8362,
       "step": 15078
     },
     {
-      "epoch": 0.41413309164812834,
+      "epoch": 0.427894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3209375440950506e-05,
-      "loss": 1.0549,
+      "learning_rate": 1.2784279689641102e-05,
+      "loss": 0.8243,
       "step": 15079
     },
     {
-      "epoch": 0.41416055587597156,
+      "epoch": 0.4279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.320853296731846e-05,
-      "loss": 0.9639,
+      "learning_rate": 1.278339694281454e-05,
+      "loss": 0.8438,
       "step": 15080
     },
     {
-      "epoch": 0.4141880201038148,
+      "epoch": 0.42795119182746877,
       "grad_norm": 0.0,
-      "learning_rate": 1.3207690468299253e-05,
-      "loss": 0.8495,
+      "learning_rate": 1.2782514172476494e-05,
+      "loss": 0.9789,
       "step": 15081
     },
     {
-      "epoch": 0.414215484331658,
+      "epoch": 0.4279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.3206847943899561e-05,
-      "loss": 0.8891,
+      "learning_rate": 1.2781631378634432e-05,
+      "loss": 0.9532,
       "step": 15082
     },
     {
-      "epoch": 0.4142429485595012,
+      "epoch": 0.42800794551645854,
       "grad_norm": 0.0,
-      "learning_rate": 1.3206005394126042e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.2780748561295802e-05,
+      "loss": 1.033,
       "step": 15083
     },
     {
-      "epoch": 0.4142704127873445,
+      "epoch": 0.42803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.3205162818985368e-05,
-      "loss": 1.0245,
+      "learning_rate": 1.2779865720468065e-05,
+      "loss": 0.776,
       "step": 15084
     },
     {
-      "epoch": 0.4142978770151877,
+      "epoch": 0.4280646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.3204320218484203e-05,
-      "loss": 0.949,
+      "learning_rate": 1.2778982856158673e-05,
+      "loss": 0.8957,
       "step": 15085
     },
     {
-      "epoch": 0.41432534124303094,
+      "epoch": 0.42809307604994323,
       "grad_norm": 0.0,
-      "learning_rate": 1.320347759262922e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.2778099968375092e-05,
+      "loss": 0.8625,
       "step": 15086
     },
     {
-      "epoch": 0.41435280547087416,
+      "epoch": 0.42812145289443815,
       "grad_norm": 0.0,
-      "learning_rate": 1.3202634941427077e-05,
-      "loss": 1.0925,
+      "learning_rate": 1.2777217057124776e-05,
+      "loss": 0.7971,
       "step": 15087
     },
     {
-      "epoch": 0.41438026969871744,
+      "epoch": 0.428149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.3201792264884446e-05,
-      "loss": 0.9013,
+      "learning_rate": 1.277633412241518e-05,
+      "loss": 0.9617,
       "step": 15088
     },
     {
-      "epoch": 0.41440773392656066,
+      "epoch": 0.4281782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.3200949563007994e-05,
-      "loss": 0.8591,
+      "learning_rate": 1.2775451164253767e-05,
+      "loss": 0.8109,
       "step": 15089
     },
     {
-      "epoch": 0.4144351981544039,
+      "epoch": 0.42820658342792284,
       "grad_norm": 0.0,
-      "learning_rate": 1.3200106835804391e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.2774568182647995e-05,
+      "loss": 0.9376,
       "step": 15090
     },
     {
-      "epoch": 0.4144626623822471,
+      "epoch": 0.4282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.3199264083280304e-05,
-      "loss": 0.917,
+      "learning_rate": 1.2773685177605314e-05,
+      "loss": 0.999,
       "step": 15091
     },
     {
-      "epoch": 0.4144901266100904,
+      "epoch": 0.4282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.3198421305442398e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.2772802149133196e-05,
+      "loss": 0.9912,
       "step": 15092
     },
     {
-      "epoch": 0.4145175908379336,
+      "epoch": 0.42829171396140747,
       "grad_norm": 0.0,
-      "learning_rate": 1.3197578502297343e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.2771919097239092e-05,
+      "loss": 0.9567,
       "step": 15093
     },
     {
-      "epoch": 0.4145450550657768,
+      "epoch": 0.4283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.319673567385181e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.2771036021930463e-05,
+      "loss": 0.8869,
       "step": 15094
     },
     {
-      "epoch": 0.41457251929362005,
+      "epoch": 0.4283484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.3195892820112466e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.2770152923214767e-05,
+      "loss": 0.9889,
       "step": 15095
     },
     {
-      "epoch": 0.41459998352146327,
+      "epoch": 0.42837684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 1.3195049941085979e-05,
-      "loss": 0.9021,
+      "learning_rate": 1.2769269801099467e-05,
+      "loss": 0.913,
       "step": 15096
     },
     {
-      "epoch": 0.41462744774930654,
+      "epoch": 0.4284052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.3194207036779018e-05,
-      "loss": 0.9121,
+      "learning_rate": 1.2768386655592018e-05,
+      "loss": 0.9227,
       "step": 15097
     },
     {
-      "epoch": 0.41465491197714976,
+      "epoch": 0.42843359818388194,
       "grad_norm": 0.0,
-      "learning_rate": 1.3193364107198257e-05,
-      "loss": 0.8789,
+      "learning_rate": 1.2767503486699884e-05,
+      "loss": 0.9038,
       "step": 15098
     },
     {
-      "epoch": 0.414682376204993,
+      "epoch": 0.42846197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.319252115235036e-05,
-      "loss": 1.0448,
+      "learning_rate": 1.2766620294430523e-05,
+      "loss": 0.9632,
       "step": 15099
     },
     {
-      "epoch": 0.4147098404328362,
+      "epoch": 0.4284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.3191678172242e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.2765737078791394e-05,
+      "loss": 0.8636,
       "step": 15100
     },
     {
-      "epoch": 0.4147373046606795,
+      "epoch": 0.4285187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.3190835166879844e-05,
-      "loss": 0.8871,
+      "learning_rate": 1.2764853839789964e-05,
+      "loss": 1.0034,
       "step": 15101
     },
     {
-      "epoch": 0.4147647688885227,
+      "epoch": 0.42854710556186154,
       "grad_norm": 0.0,
-      "learning_rate": 1.3189992136270568e-05,
-      "loss": 0.7857,
+      "learning_rate": 1.2763970577433685e-05,
+      "loss": 0.8584,
       "step": 15102
     },
     {
-      "epoch": 0.4147922331163659,
+      "epoch": 0.4285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.3189149080420833e-05,
-      "loss": 0.8634,
+      "learning_rate": 1.2763087291730023e-05,
+      "loss": 0.9133,
       "step": 15103
     },
     {
-      "epoch": 0.41481969734420915,
+      "epoch": 0.4286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.318830599933732e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.276220398268644e-05,
+      "loss": 0.9133,
       "step": 15104
     },
     {
-      "epoch": 0.4148471615720524,
+      "epoch": 0.4286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.3187462893026692e-05,
-      "loss": 0.9661,
+      "learning_rate": 1.2761320650310396e-05,
+      "loss": 0.8242,
       "step": 15105
     },
     {
-      "epoch": 0.41487462579989565,
+      "epoch": 0.4286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.3186619761495627e-05,
-      "loss": 0.8945,
+      "learning_rate": 1.2760437294609354e-05,
+      "loss": 0.8878,
       "step": 15106
     },
     {
-      "epoch": 0.41490209002773887,
+      "epoch": 0.428688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.3185776604750788e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.275955391559077e-05,
+      "loss": 0.8464,
       "step": 15107
     },
     {
-      "epoch": 0.4149295542555821,
+      "epoch": 0.42871736662883086,
       "grad_norm": 0.0,
-      "learning_rate": 1.318493342279885e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.2758670513262113e-05,
+      "loss": 0.896,
       "step": 15108
     },
     {
-      "epoch": 0.41495701848342537,
+      "epoch": 0.4287457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.3184090215646488e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.2757787087630837e-05,
+      "loss": 0.9259,
       "step": 15109
     },
     {
-      "epoch": 0.4149844827112686,
+      "epoch": 0.42877412031782064,
       "grad_norm": 0.0,
-      "learning_rate": 1.3183246983300369e-05,
-      "loss": 0.8728,
+      "learning_rate": 1.2756903638704414e-05,
+      "loss": 0.8377,
       "step": 15110
     },
     {
-      "epoch": 0.4150119469391118,
+      "epoch": 0.42880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.3182403725767168e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.2756020166490303e-05,
+      "loss": 0.9115,
       "step": 15111
     },
     {
-      "epoch": 0.41503941116695503,
+      "epoch": 0.42883087400681047,
       "grad_norm": 0.0,
-      "learning_rate": 1.3181560443053556e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.2755136670995965e-05,
+      "loss": 0.9152,
       "step": 15112
     },
     {
-      "epoch": 0.41506687539479825,
+      "epoch": 0.42885925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.3180717135166203e-05,
-      "loss": 0.9785,
+      "learning_rate": 1.2754253152228864e-05,
+      "loss": 0.8796,
       "step": 15113
     },
     {
-      "epoch": 0.41509433962264153,
+      "epoch": 0.42888762769580024,
       "grad_norm": 0.0,
-      "learning_rate": 1.3179873802111787e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.2753369610196463e-05,
+      "loss": 0.9837,
       "step": 15114
     },
     {
-      "epoch": 0.41512180385048475,
+      "epoch": 0.4289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.3179030443896976e-05,
-      "loss": 1.0139,
+      "learning_rate": 1.2752486044906221e-05,
+      "loss": 0.9797,
       "step": 15115
     },
     {
-      "epoch": 0.41514926807832797,
+      "epoch": 0.42894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.3178187060528444e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.2751602456365608e-05,
+      "loss": 0.9336,
       "step": 15116
     },
     {
-      "epoch": 0.4151767323061712,
+      "epoch": 0.4289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.3177343652012863e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.2750718844582087e-05,
+      "loss": 0.9894,
       "step": 15117
     },
     {
-      "epoch": 0.41520419653401447,
+      "epoch": 0.4290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.3176500218356913e-05,
-      "loss": 0.8446,
+      "learning_rate": 1.2749835209563119e-05,
+      "loss": 0.9704,
       "step": 15118
     },
     {
-      "epoch": 0.4152316607618577,
+      "epoch": 0.4290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.3175656759567264e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.2748951551316168e-05,
+      "loss": 0.9329,
       "step": 15119
     },
     {
-      "epoch": 0.4152591249897009,
+      "epoch": 0.42905788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 1.3174813275650583e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.2748067869848703e-05,
+      "loss": 0.9477,
       "step": 15120
     },
     {
-      "epoch": 0.41528658921754413,
+      "epoch": 0.4290862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.3173969766613552e-05,
-      "loss": 0.8888,
+      "learning_rate": 1.2747184165168186e-05,
+      "loss": 1.0391,
       "step": 15121
     },
     {
-      "epoch": 0.4153140534453874,
+      "epoch": 0.42911464245175934,
       "grad_norm": 0.0,
-      "learning_rate": 1.3173126232462843e-05,
-      "loss": 1.0058,
+      "learning_rate": 1.2746300437282076e-05,
+      "loss": 0.8678,
       "step": 15122
     },
     {
-      "epoch": 0.41534151767323063,
+      "epoch": 0.42914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.317228267320513e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.2745416686197842e-05,
+      "loss": 1.0296,
       "step": 15123
     },
     {
-      "epoch": 0.41536898190107385,
+      "epoch": 0.4291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.3171439088847089e-05,
-      "loss": 0.9772,
+      "learning_rate": 1.2744532911922956e-05,
+      "loss": 0.8527,
       "step": 15124
     },
     {
-      "epoch": 0.4153964461289171,
+      "epoch": 0.42919977298524403,
       "grad_norm": 0.0,
-      "learning_rate": 1.317059547939539e-05,
-      "loss": 1.0425,
+      "learning_rate": 1.2743649114464873e-05,
+      "loss": 0.8421,
       "step": 15125
     },
     {
-      "epoch": 0.4154239103567603,
+      "epoch": 0.42922814982973895,
       "grad_norm": 0.0,
-      "learning_rate": 1.3169751844856714e-05,
-      "loss": 0.93,
+      "learning_rate": 1.2742765293831063e-05,
+      "loss": 0.9287,
       "step": 15126
     },
     {
-      "epoch": 0.41545137458460357,
+      "epoch": 0.4292565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.3168908185237734e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.2741881450028992e-05,
+      "loss": 0.9638,
       "step": 15127
     },
     {
-      "epoch": 0.4154788388124468,
+      "epoch": 0.4292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.3168064500545121e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.2740997583066126e-05,
+      "loss": 0.8516,
       "step": 15128
     },
     {
-      "epoch": 0.41550630304029,
+      "epoch": 0.42931328036322364,
       "grad_norm": 0.0,
-      "learning_rate": 1.3167220790785558e-05,
-      "loss": 0.9671,
+      "learning_rate": 1.2740113692949927e-05,
+      "loss": 0.9264,
       "step": 15129
     },
     {
-      "epoch": 0.41553376726813324,
+      "epoch": 0.4293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.3166377055965719e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.273922977968787e-05,
+      "loss": 0.7955,
       "step": 15130
     },
     {
-      "epoch": 0.4155612314959765,
+      "epoch": 0.4293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.3165533296092274e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.273834584328741e-05,
+      "loss": 0.975,
       "step": 15131
     },
     {
-      "epoch": 0.41558869572381973,
+      "epoch": 0.42939841089670827,
       "grad_norm": 0.0,
-      "learning_rate": 1.3164689511171905e-05,
-      "loss": 0.8276,
+      "learning_rate": 1.2737461883756021e-05,
+      "loss": 0.9325,
       "step": 15132
     },
     {
-      "epoch": 0.41561615995166296,
+      "epoch": 0.4294267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.3163845701211286e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.273657790110117e-05,
+      "loss": 0.9379,
       "step": 15133
     },
     {
-      "epoch": 0.4156436241795062,
+      "epoch": 0.42945516458569805,
       "grad_norm": 0.0,
-      "learning_rate": 1.3163001866217095e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.2735693895330324e-05,
+      "loss": 0.8753,
       "step": 15134
     },
     {
-      "epoch": 0.41567108840734945,
+      "epoch": 0.42948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.3162158006196008e-05,
-      "loss": 0.8818,
+      "learning_rate": 1.2734809866450944e-05,
+      "loss": 0.8719,
       "step": 15135
     },
     {
-      "epoch": 0.4156985526351927,
+      "epoch": 0.4295119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.3161314121154702e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.2733925814470507e-05,
+      "loss": 0.8782,
       "step": 15136
     },
     {
-      "epoch": 0.4157260168630359,
+      "epoch": 0.42954029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.3160470211099853e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.2733041739396473e-05,
+      "loss": 0.8794,
       "step": 15137
     },
     {
-      "epoch": 0.4157534810908791,
+      "epoch": 0.42956867196367765,
       "grad_norm": 0.0,
-      "learning_rate": 1.3159626276038142e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.2732157641236315e-05,
+      "loss": 0.9298,
       "step": 15138
     },
     {
-      "epoch": 0.41578094531872234,
+      "epoch": 0.4295970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.3158782315976243e-05,
-      "loss": 0.8459,
+      "learning_rate": 1.2731273519997493e-05,
+      "loss": 0.9064,
       "step": 15139
     },
     {
-      "epoch": 0.4158084095465656,
+      "epoch": 0.4296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.3157938330920835e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.2730389375687487e-05,
+      "loss": 0.9296,
       "step": 15140
     },
     {
-      "epoch": 0.41583587377440884,
+      "epoch": 0.42965380249716234,
       "grad_norm": 0.0,
-      "learning_rate": 1.3157094320878595e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.2729505208313755e-05,
+      "loss": 0.8345,
       "step": 15141
     },
     {
-      "epoch": 0.41586333800225206,
+      "epoch": 0.4296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.3156250285856201e-05,
-      "loss": 1.0008,
+      "learning_rate": 1.272862101788377e-05,
+      "loss": 0.8709,
       "step": 15142
     },
     {
-      "epoch": 0.4158908022300953,
+      "epoch": 0.4297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.3155406225860335e-05,
-      "loss": 1.0193,
+      "learning_rate": 1.2727736804405005e-05,
+      "loss": 0.8841,
       "step": 15143
     },
     {
-      "epoch": 0.41591826645793856,
+      "epoch": 0.429738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.3154562140897667e-05,
-      "loss": 0.9389,
+      "learning_rate": 1.2726852567884922e-05,
+      "loss": 0.8384,
       "step": 15144
     },
     {
-      "epoch": 0.4159457306857818,
+      "epoch": 0.4297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.3153718030974885e-05,
-      "loss": 0.9381,
+      "learning_rate": 1.272596830833099e-05,
+      "loss": 0.9384,
       "step": 15145
     },
     {
-      "epoch": 0.415973194913625,
+      "epoch": 0.42979568671963675,
       "grad_norm": 0.0,
-      "learning_rate": 1.3152873896098664e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.2725084025750683e-05,
+      "loss": 0.869,
       "step": 15146
     },
     {
-      "epoch": 0.4160006591414682,
+      "epoch": 0.42982406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 1.3152029736275685e-05,
-      "loss": 1.0433,
+      "learning_rate": 1.2724199720151469e-05,
+      "loss": 0.9957,
       "step": 15147
     },
     {
-      "epoch": 0.4160281233693115,
+      "epoch": 0.4298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.3151185551512625e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.272331539154082e-05,
+      "loss": 1.0251,
       "step": 15148
     },
     {
-      "epoch": 0.4160555875971547,
+      "epoch": 0.42988081725312144,
       "grad_norm": 0.0,
-      "learning_rate": 1.3150341341816162e-05,
-      "loss": 0.9983,
+      "learning_rate": 1.27224310399262e-05,
+      "loss": 0.9344,
       "step": 15149
     },
     {
-      "epoch": 0.41608305182499794,
+      "epoch": 0.42990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.3149497107192982e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.2721546665315086e-05,
+      "loss": 0.8828,
       "step": 15150
     },
     {
-      "epoch": 0.41611051605284116,
+      "epoch": 0.4299375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.3148652847649759e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.2720662267714942e-05,
+      "loss": 0.9891,
       "step": 15151
     },
     {
-      "epoch": 0.4161379802806844,
+      "epoch": 0.42996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.3147808563193172e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.2719777847133241e-05,
+      "loss": 0.9352,
       "step": 15152
     },
     {
-      "epoch": 0.41616544450852766,
+      "epoch": 0.42999432463110104,
       "grad_norm": 0.0,
-      "learning_rate": 1.3146964253829905e-05,
-      "loss": 0.8852,
+      "learning_rate": 1.2718893403577457e-05,
+      "loss": 0.8085,
       "step": 15153
     },
     {
-      "epoch": 0.4161929087363709,
+      "epoch": 0.4300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.314611991956664e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.2718008937055056e-05,
+      "loss": 0.9807,
       "step": 15154
     },
     {
-      "epoch": 0.4162203729642141,
+      "epoch": 0.4300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.3145275560410055e-05,
-      "loss": 0.7978,
+      "learning_rate": 1.2717124447573515e-05,
+      "loss": 0.9476,
       "step": 15155
     },
     {
-      "epoch": 0.4162478371920573,
+      "epoch": 0.4300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.3144431176366833e-05,
-      "loss": 0.843,
+      "learning_rate": 1.2716239935140295e-05,
+      "loss": 0.8804,
       "step": 15156
     },
     {
-      "epoch": 0.4162753014199006,
+      "epoch": 0.4301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.3143586767443652e-05,
-      "loss": 0.8568,
+      "learning_rate": 1.2715355399762878e-05,
+      "loss": 0.8659,
       "step": 15157
     },
     {
-      "epoch": 0.4163027656477438,
+      "epoch": 0.4301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.3142742333647193e-05,
-      "loss": 0.7642,
+      "learning_rate": 1.2714470841448733e-05,
+      "loss": 0.9486,
       "step": 15158
     },
     {
-      "epoch": 0.41633022987558704,
+      "epoch": 0.43016458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.3141897874984142e-05,
-      "loss": 0.8657,
+      "learning_rate": 1.2713586260205331e-05,
+      "loss": 0.8022,
       "step": 15159
     },
     {
-      "epoch": 0.41635769410343026,
+      "epoch": 0.4301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.3141053391461178e-05,
-      "loss": 1.0258,
+      "learning_rate": 1.2712701656040142e-05,
+      "loss": 0.8016,
       "step": 15160
     },
     {
-      "epoch": 0.41638515833127354,
+      "epoch": 0.43022133938706014,
       "grad_norm": 0.0,
-      "learning_rate": 1.3140208883084982e-05,
-      "loss": 0.949,
+      "learning_rate": 1.2711817028960642e-05,
+      "loss": 0.8881,
       "step": 15161
     },
     {
-      "epoch": 0.41641262255911676,
+      "epoch": 0.43024971623155506,
       "grad_norm": 0.0,
-      "learning_rate": 1.3139364349862236e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.2710932378974299e-05,
+      "loss": 0.9162,
       "step": 15162
     },
     {
-      "epoch": 0.41644008678696,
+      "epoch": 0.4302780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.3138519791799624e-05,
-      "loss": 0.8953,
+      "learning_rate": 1.2710047706088588e-05,
+      "loss": 0.9064,
       "step": 15163
     },
     {
-      "epoch": 0.4164675510148032,
+      "epoch": 0.43030646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.3137675208903828e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.2709163010310985e-05,
+      "loss": 0.8299,
       "step": 15164
     },
     {
-      "epoch": 0.4164950152426464,
+      "epoch": 0.43033484676503975,
       "grad_norm": 0.0,
-      "learning_rate": 1.3136830601181527e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.2708278291648961e-05,
+      "loss": 0.8816,
       "step": 15165
     },
     {
-      "epoch": 0.4165224794704897,
+      "epoch": 0.4303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.313598596863941e-05,
-      "loss": 0.8542,
+      "learning_rate": 1.2707393550109985e-05,
+      "loss": 1.0359,
       "step": 15166
     },
     {
-      "epoch": 0.4165499436983329,
+      "epoch": 0.4303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.3135141311284159e-05,
-      "loss": 0.918,
+      "learning_rate": 1.2706508785701537e-05,
+      "loss": 0.9791,
       "step": 15167
     },
     {
-      "epoch": 0.41657740792617615,
+      "epoch": 0.4304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.3134296629122456e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.2705623998431086e-05,
+      "loss": 0.8894,
       "step": 15168
     },
     {
-      "epoch": 0.41660487215401937,
+      "epoch": 0.4304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.313345192216098e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.2704739188306107e-05,
+      "loss": 0.8797,
       "step": 15169
     },
     {
-      "epoch": 0.41663233638186264,
+      "epoch": 0.4304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.3132607190406421e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.2703854355334073e-05,
+      "loss": 1.0231,
       "step": 15170
     },
     {
-      "epoch": 0.41665980060970587,
+      "epoch": 0.43050510783200907,
       "grad_norm": 0.0,
-      "learning_rate": 1.313176243386546e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.2702969499522463e-05,
+      "loss": 0.8755,
       "step": 15171
     },
     {
-      "epoch": 0.4166872648375491,
+      "epoch": 0.430533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.3130917652544782e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.270208462087875e-05,
+      "loss": 0.7999,
       "step": 15172
     },
     {
-      "epoch": 0.4167147290653923,
+      "epoch": 0.43056186152099885,
       "grad_norm": 0.0,
-      "learning_rate": 1.313007284645107e-05,
-      "loss": 0.8739,
+      "learning_rate": 1.2701199719410402e-05,
+      "loss": 0.9766,
       "step": 15173
     },
     {
-      "epoch": 0.4167421932932356,
+      "epoch": 0.43059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.312922801559101e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.2700314795124902e-05,
+      "loss": 0.9756,
       "step": 15174
     },
     {
-      "epoch": 0.4167696575210788,
+      "epoch": 0.4306186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.3128383159971282e-05,
-      "loss": 1.0038,
+      "learning_rate": 1.269942984802972e-05,
+      "loss": 0.8892,
       "step": 15175
     },
     {
-      "epoch": 0.41679712174892203,
+      "epoch": 0.43064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.312753827959858e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.269854487813233e-05,
+      "loss": 0.9624,
       "step": 15176
     },
     {
-      "epoch": 0.41682458597676525,
+      "epoch": 0.43067536889897845,
       "grad_norm": 0.0,
-      "learning_rate": 1.3126693374479582e-05,
-      "loss": 0.907,
+      "learning_rate": 1.2697659885440213e-05,
+      "loss": 0.9781,
       "step": 15177
     },
     {
-      "epoch": 0.41685205020460847,
+      "epoch": 0.4307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.3125848444620972e-05,
-      "loss": 0.8833,
+      "learning_rate": 1.2696774869960842e-05,
+      "loss": 0.9974,
       "step": 15178
     },
     {
-      "epoch": 0.41687951443245175,
+      "epoch": 0.4307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.312500349002944e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.2695889831701691e-05,
+      "loss": 0.9667,
       "step": 15179
     },
     {
-      "epoch": 0.41690697866029497,
+      "epoch": 0.4307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.3124158510711674e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.2695004770670239e-05,
+      "loss": 0.9668,
       "step": 15180
     },
     {
-      "epoch": 0.4169344428881382,
+      "epoch": 0.430788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.3123313506674349e-05,
-      "loss": 0.9273,
+      "learning_rate": 1.269411968687396e-05,
+      "loss": 0.8507,
       "step": 15181
     },
     {
-      "epoch": 0.4169619071159814,
+      "epoch": 0.4308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.3122468477924162e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.2693234580320333e-05,
+      "loss": 0.9255,
       "step": 15182
     },
     {
-      "epoch": 0.4169893713438247,
+      "epoch": 0.4308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.3121623424467793e-05,
-      "loss": 0.9873,
+      "learning_rate": 1.2692349451016828e-05,
+      "loss": 0.8994,
       "step": 15183
     },
     {
-      "epoch": 0.4170168355716679,
+      "epoch": 0.4308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.312077834631193e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.2691464298970928e-05,
+      "loss": 0.9792,
       "step": 15184
     },
     {
-      "epoch": 0.41704429979951113,
+      "epoch": 0.43090238365493755,
       "grad_norm": 0.0,
-      "learning_rate": 1.3119933243463261e-05,
-      "loss": 0.8386,
+      "learning_rate": 1.2690579124190108e-05,
+      "loss": 0.9399,
       "step": 15185
     },
     {
-      "epoch": 0.41707176402735435,
+      "epoch": 0.43093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.3119088115928471e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.2689693926681845e-05,
+      "loss": 0.8142,
       "step": 15186
     },
     {
-      "epoch": 0.41709922825519763,
+      "epoch": 0.4309591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.3118242963714249e-05,
-      "loss": 0.9112,
+      "learning_rate": 1.2688808706453616e-05,
+      "loss": 0.8658,
       "step": 15187
     },
     {
-      "epoch": 0.41712669248304085,
+      "epoch": 0.43098751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.311739778682728e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.26879234635129e-05,
+      "loss": 0.8564,
       "step": 15188
     },
     {
-      "epoch": 0.41715415671088407,
+      "epoch": 0.43101589103291715,
       "grad_norm": 0.0,
-      "learning_rate": 1.3116552585274252e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.2687038197867174e-05,
+      "loss": 0.9475,
       "step": 15189
     },
     {
-      "epoch": 0.4171816209387273,
+      "epoch": 0.431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.3115707359061852e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.2686152909523917e-05,
+      "loss": 0.8423,
       "step": 15190
     },
     {
-      "epoch": 0.4172090851665705,
+      "epoch": 0.43107264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 1.3114862108196768e-05,
-      "loss": 0.8705,
+      "learning_rate": 1.2685267598490606e-05,
+      "loss": 1.0438,
       "step": 15191
     },
     {
-      "epoch": 0.4172365493944138,
+      "epoch": 0.43110102156640184,
       "grad_norm": 0.0,
-      "learning_rate": 1.311401683268569e-05,
-      "loss": 1.0276,
+      "learning_rate": 1.2684382264774713e-05,
+      "loss": 0.9422,
       "step": 15192
     },
     {
-      "epoch": 0.417264013622257,
+      "epoch": 0.4311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.3113171532535303e-05,
-      "loss": 0.9223,
+      "learning_rate": 1.2683496908383725e-05,
+      "loss": 0.8582,
       "step": 15193
     },
     {
-      "epoch": 0.41729147785010023,
+      "epoch": 0.4311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.3112326207752299e-05,
-      "loss": 0.8457,
+      "learning_rate": 1.268261152932512e-05,
+      "loss": 0.9049,
       "step": 15194
     },
     {
-      "epoch": 0.41731894207794346,
+      "epoch": 0.4311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.3111480858343365e-05,
-      "loss": 0.942,
+      "learning_rate": 1.2681726127606374e-05,
+      "loss": 0.8903,
       "step": 15195
     },
     {
-      "epoch": 0.41734640630578673,
+      "epoch": 0.4312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.3110635484315186e-05,
-      "loss": 0.9782,
+      "learning_rate": 1.2680840703234968e-05,
+      "loss": 1.0267,
       "step": 15196
     },
     {
-      "epoch": 0.41737387053362995,
+      "epoch": 0.43124290578887625,
       "grad_norm": 0.0,
-      "learning_rate": 1.3109790085674454e-05,
-      "loss": 0.9672,
+      "learning_rate": 1.2679955256218379e-05,
+      "loss": 0.7664,
       "step": 15197
     },
     {
-      "epoch": 0.4174013347614732,
+      "epoch": 0.43127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.310894466242786e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.2679069786564089e-05,
+      "loss": 0.9212,
       "step": 15198
     },
     {
-      "epoch": 0.4174287989893164,
+      "epoch": 0.4312996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.3108099214582094e-05,
-      "loss": 0.8832,
+      "learning_rate": 1.2678184294279573e-05,
+      "loss": 0.881,
       "step": 15199
     },
     {
-      "epoch": 0.4174562632171597,
+      "epoch": 0.43132803632236094,
       "grad_norm": 0.0,
-      "learning_rate": 1.3107253742143838e-05,
-      "loss": 0.9207,
+      "learning_rate": 1.2677298779372314e-05,
+      "loss": 0.9928,
       "step": 15200
     },
     {
-      "epoch": 0.4174837274450029,
+      "epoch": 0.43135641316685586,
       "grad_norm": 0.0,
-      "learning_rate": 1.310640824511979e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.2676413241849792e-05,
+      "loss": 0.9744,
       "step": 15201
     },
     {
-      "epoch": 0.4175111916728461,
+      "epoch": 0.4313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.3105562723516636e-05,
-      "loss": 1.0043,
+      "learning_rate": 1.267552768171949e-05,
+      "loss": 0.8987,
       "step": 15202
     },
     {
-      "epoch": 0.41753865590068934,
+      "epoch": 0.43141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.3104717177341067e-05,
-      "loss": 0.986,
+      "learning_rate": 1.2674642098988884e-05,
+      "loss": 0.8741,
       "step": 15203
     },
     {
-      "epoch": 0.41756612012853256,
+      "epoch": 0.43144154370034055,
       "grad_norm": 0.0,
-      "learning_rate": 1.3103871606599774e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.2673756493665454e-05,
+      "loss": 0.8733,
       "step": 15204
     },
     {
-      "epoch": 0.41759358435637584,
+      "epoch": 0.4314699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.3103026011299442e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.2672870865756685e-05,
+      "loss": 0.8263,
       "step": 15205
     },
     {
-      "epoch": 0.41762104858421906,
+      "epoch": 0.4314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.3102180391446771e-05,
-      "loss": 0.9627,
+      "learning_rate": 1.2671985215270055e-05,
+      "loss": 0.9401,
       "step": 15206
     },
     {
-      "epoch": 0.4176485128120623,
+      "epoch": 0.4315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.3101334747048447e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.2671099542213046e-05,
+      "loss": 0.8361,
       "step": 15207
     },
     {
-      "epoch": 0.4176759770399055,
+      "epoch": 0.4315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.3100489078111158e-05,
-      "loss": 0.7866,
+      "learning_rate": 1.2670213846593137e-05,
+      "loss": 0.9289,
       "step": 15208
     },
     {
-      "epoch": 0.4177034412677488,
+      "epoch": 0.431583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.3099643384641598e-05,
-      "loss": 0.989,
+      "learning_rate": 1.2669328128417815e-05,
+      "loss": 0.8983,
       "step": 15209
     },
     {
-      "epoch": 0.417730905495592,
+      "epoch": 0.43161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.3098797666646462e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.2668442387694555e-05,
+      "loss": 0.8498,
       "step": 15210
     },
     {
-      "epoch": 0.4177583697234352,
+      "epoch": 0.4316401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.3097951924132439e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.2667556624430844e-05,
+      "loss": 0.8774,
       "step": 15211
     },
     {
-      "epoch": 0.41778583395127844,
+      "epoch": 0.43166855845629964,
       "grad_norm": 0.0,
-      "learning_rate": 1.3097106157106216e-05,
-      "loss": 0.8619,
+      "learning_rate": 1.2666670838634163e-05,
+      "loss": 0.9172,
       "step": 15212
     },
     {
-      "epoch": 0.4178132981791217,
+      "epoch": 0.43169693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 1.3096260365574493e-05,
-      "loss": 1.0025,
+      "learning_rate": 1.2665785030311993e-05,
+      "loss": 0.8419,
       "step": 15213
     },
     {
-      "epoch": 0.41784076240696494,
+      "epoch": 0.4317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.3095414549543956e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.2664899199471816e-05,
+      "loss": 1.0169,
       "step": 15214
     },
     {
-      "epoch": 0.41786822663480816,
+      "epoch": 0.43175368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 1.3094568709021304e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.2664013346121115e-05,
+      "loss": 0.9366,
       "step": 15215
     },
     {
-      "epoch": 0.4178956908626514,
+      "epoch": 0.43178206583427925,
       "grad_norm": 0.0,
-      "learning_rate": 1.3093722844013224e-05,
-      "loss": 1.0203,
+      "learning_rate": 1.2663127470267376e-05,
+      "loss": 0.9688,
       "step": 15216
     },
     {
-      "epoch": 0.41792315509049466,
+      "epoch": 0.4318104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.3092876954526408e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.2662241571918077e-05,
+      "loss": 0.8565,
       "step": 15217
     },
     {
-      "epoch": 0.4179506193183379,
+      "epoch": 0.431838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.3092031040567555e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.2661355651080707e-05,
+      "loss": 0.859,
       "step": 15218
     },
     {
-      "epoch": 0.4179780835461811,
+      "epoch": 0.4318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.3091185102143354e-05,
-      "loss": 0.8457,
+      "learning_rate": 1.2660469707762744e-05,
+      "loss": 0.8287,
       "step": 15219
     },
     {
-      "epoch": 0.4180055477740243,
+      "epoch": 0.4318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.3090339139260494e-05,
-      "loss": 0.7811,
+      "learning_rate": 1.2659583741971675e-05,
+      "loss": 0.8639,
       "step": 15220
     },
     {
-      "epoch": 0.41803301200186754,
+      "epoch": 0.4319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.3089493151925679e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.2658697753714982e-05,
+      "loss": 0.8411,
       "step": 15221
     },
     {
-      "epoch": 0.4180604762297108,
+      "epoch": 0.4319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.3088647140145596e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.265781174300015e-05,
+      "loss": 0.8886,
       "step": 15222
     },
     {
-      "epoch": 0.41808794045755404,
+      "epoch": 0.4319807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.3087801103926943e-05,
-      "loss": 0.955,
+      "learning_rate": 1.265692570983466e-05,
+      "loss": 0.9564,
       "step": 15223
     },
     {
-      "epoch": 0.41811540468539726,
+      "epoch": 0.43200908059023835,
       "grad_norm": 0.0,
-      "learning_rate": 1.3086955043276408e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.2656039654226e-05,
+      "loss": 0.9222,
       "step": 15224
     },
     {
-      "epoch": 0.4181428689132405,
+      "epoch": 0.43203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.3086108958200689e-05,
-      "loss": 0.918,
+      "learning_rate": 1.2655153576181655e-05,
+      "loss": 0.9562,
       "step": 15225
     },
     {
-      "epoch": 0.41817033314108376,
+      "epoch": 0.4320658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.3085262848706482e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.265426747570911e-05,
+      "loss": 0.9392,
       "step": 15226
     },
     {
-      "epoch": 0.418197797368927,
+      "epoch": 0.43209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.308441671480048e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.2653381352815846e-05,
+      "loss": 0.9077,
       "step": 15227
     },
     {
-      "epoch": 0.4182252615967702,
+      "epoch": 0.43212258796821795,
       "grad_norm": 0.0,
-      "learning_rate": 1.3083570556489378e-05,
-      "loss": 0.9818,
+      "learning_rate": 1.265249520750935e-05,
+      "loss": 0.8784,
       "step": 15228
     },
     {
-      "epoch": 0.4182527258246134,
+      "epoch": 0.4321509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.308272437377987e-05,
-      "loss": 0.894,
+      "learning_rate": 1.265160903979711e-05,
+      "loss": 0.9927,
       "step": 15229
     },
     {
-      "epoch": 0.4182801900524567,
+      "epoch": 0.4321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.3081878166678654e-05,
-      "loss": 0.8781,
+      "learning_rate": 1.2650722849686607e-05,
+      "loss": 0.9007,
       "step": 15230
     },
     {
-      "epoch": 0.4183076542802999,
+      "epoch": 0.4322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.3081031935192425e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.264983663718533e-05,
+      "loss": 0.8393,
       "step": 15231
     },
     {
-      "epoch": 0.41833511850814314,
+      "epoch": 0.4322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.3080185679327877e-05,
-      "loss": 0.9573,
+      "learning_rate": 1.2648950402300763e-05,
+      "loss": 1.0748,
       "step": 15232
     },
     {
-      "epoch": 0.41836258273598637,
+      "epoch": 0.4322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.3079339399091706e-05,
-      "loss": 0.9202,
+      "learning_rate": 1.2648064145040392e-05,
+      "loss": 0.9157,
       "step": 15233
     },
     {
-      "epoch": 0.4183900469638296,
+      "epoch": 0.4322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.3078493094490607e-05,
-      "loss": 1.007,
+      "learning_rate": 1.2647177865411706e-05,
+      "loss": 0.9625,
       "step": 15234
     },
     {
-      "epoch": 0.41841751119167286,
+      "epoch": 0.4323212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.3077646765531282e-05,
-      "loss": 1.0452,
+      "learning_rate": 1.2646291563422191e-05,
+      "loss": 0.9468,
       "step": 15235
     },
     {
-      "epoch": 0.4184449754195161,
+      "epoch": 0.43234960272417705,
       "grad_norm": 0.0,
-      "learning_rate": 1.3076800412220423e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.264540523907933e-05,
+      "loss": 0.988,
       "step": 15236
     },
     {
-      "epoch": 0.4184724396473593,
+      "epoch": 0.43237797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.3075954034564727e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.2644518892390611e-05,
+      "loss": 1.0238,
       "step": 15237
     },
     {
-      "epoch": 0.41849990387520253,
+      "epoch": 0.4324063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.3075107632570888e-05,
-      "loss": 0.8707,
+      "learning_rate": 1.2643632523363523e-05,
+      "loss": 0.8789,
       "step": 15238
     },
     {
-      "epoch": 0.4185273681030458,
+      "epoch": 0.43243473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.3074261206245614e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.2642746132005552e-05,
+      "loss": 0.9899,
       "step": 15239
     },
     {
-      "epoch": 0.418554832330889,
+      "epoch": 0.43246311010215666,
       "grad_norm": 0.0,
-      "learning_rate": 1.3073414755595586e-05,
-      "loss": 0.9916,
+      "learning_rate": 1.2641859718324185e-05,
+      "loss": 0.9152,
       "step": 15240
     },
     {
-      "epoch": 0.41858229655873225,
+      "epoch": 0.4324914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.3072568280627511e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.2640973282326914e-05,
+      "loss": 0.8838,
       "step": 15241
     },
     {
-      "epoch": 0.41860976078657547,
+      "epoch": 0.43251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.307172178134809e-05,
-      "loss": 0.8488,
+      "learning_rate": 1.264008682402122e-05,
+      "loss": 0.9015,
       "step": 15242
     },
     {
-      "epoch": 0.41863722501441875,
+      "epoch": 0.4325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.3070875257764015e-05,
-      "loss": 0.9344,
+      "learning_rate": 1.2639200343414593e-05,
+      "loss": 0.9586,
       "step": 15243
     },
     {
-      "epoch": 0.41866468924226197,
+      "epoch": 0.4325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.3070028709881984e-05,
-      "loss": 0.9037,
+      "learning_rate": 1.2638313840514525e-05,
+      "loss": 0.8608,
       "step": 15244
     },
     {
-      "epoch": 0.4186921534701052,
+      "epoch": 0.4326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.3069182137708698e-05,
-      "loss": 0.9352,
+      "learning_rate": 1.2637427315328502e-05,
+      "loss": 0.898,
       "step": 15245
     },
     {
-      "epoch": 0.4187196176979484,
+      "epoch": 0.432633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.3068335541250852e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.2636540767864008e-05,
+      "loss": 0.9783,
       "step": 15246
     },
     {
-      "epoch": 0.41874708192579163,
+      "epoch": 0.4326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.3067488920515148e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.2635654198128536e-05,
+      "loss": 0.9141,
       "step": 15247
     },
     {
-      "epoch": 0.4187745461536349,
+      "epoch": 0.43269012485811575,
       "grad_norm": 0.0,
-      "learning_rate": 1.3066642275508286e-05,
-      "loss": 0.9014,
+      "learning_rate": 1.2634767606129577e-05,
+      "loss": 1.0524,
       "step": 15248
     },
     {
-      "epoch": 0.41880201038147813,
+      "epoch": 0.43271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.3065795606236956e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.2633880991874617e-05,
+      "loss": 0.8062,
       "step": 15249
     },
     {
-      "epoch": 0.41882947460932135,
+      "epoch": 0.4327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.306494891270787e-05,
-      "loss": 1.0362,
+      "learning_rate": 1.2632994355371144e-05,
+      "loss": 0.7749,
       "step": 15250
     },
     {
-      "epoch": 0.41885693883716457,
+      "epoch": 0.43277525539160044,
       "grad_norm": 0.0,
-      "learning_rate": 1.3064102194927719e-05,
-      "loss": 0.942,
+      "learning_rate": 1.2632107696626652e-05,
+      "loss": 0.9547,
       "step": 15251
     },
     {
-      "epoch": 0.41888440306500785,
+      "epoch": 0.43280363223609536,
       "grad_norm": 0.0,
-      "learning_rate": 1.3063255452903204e-05,
-      "loss": 0.9087,
+      "learning_rate": 1.2631221015648626e-05,
+      "loss": 0.99,
       "step": 15252
     },
     {
-      "epoch": 0.41891186729285107,
+      "epoch": 0.4328320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.3062408686641022e-05,
-      "loss": 0.8332,
+      "learning_rate": 1.2630334312444557e-05,
+      "loss": 0.912,
       "step": 15253
     },
     {
-      "epoch": 0.4189393315206943,
+      "epoch": 0.43286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.306156189614788e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.2629447587021937e-05,
+      "loss": 0.867,
       "step": 15254
     },
     {
-      "epoch": 0.4189667957485375,
+      "epoch": 0.43288876276958005,
       "grad_norm": 0.0,
-      "learning_rate": 1.3060715081430476e-05,
-      "loss": 0.9121,
+      "learning_rate": 1.2628560839388255e-05,
+      "loss": 1.0072,
       "step": 15255
     },
     {
-      "epoch": 0.4189942599763808,
+      "epoch": 0.4329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.3059868242495503e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.2627674069551003e-05,
+      "loss": 0.9211,
       "step": 15256
     },
     {
-      "epoch": 0.419021724204224,
+      "epoch": 0.4329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.3059021379349673e-05,
-      "loss": 1.0483,
+      "learning_rate": 1.2626787277517667e-05,
+      "loss": 0.8074,
       "step": 15257
     },
     {
-      "epoch": 0.41904918843206723,
+      "epoch": 0.4329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.3058174491999676e-05,
-      "loss": 1.0017,
+      "learning_rate": 1.2625900463295743e-05,
+      "loss": 0.7344,
       "step": 15258
     },
     {
-      "epoch": 0.41907665265991045,
+      "epoch": 0.4330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.305732758045222e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.2625013626892719e-05,
+      "loss": 0.8412,
       "step": 15259
     },
     {
-      "epoch": 0.4191041168877537,
+      "epoch": 0.43303064699205446,
       "grad_norm": 0.0,
-      "learning_rate": 1.3056480644714006e-05,
-      "loss": 0.9059,
+      "learning_rate": 1.2624126768316088e-05,
+      "loss": 0.9353,
       "step": 15260
     },
     {
-      "epoch": 0.41913158111559695,
+      "epoch": 0.4330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.305563368479173e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.2623239887573338e-05,
+      "loss": 0.9255,
       "step": 15261
     },
     {
-      "epoch": 0.4191590453434402,
+      "epoch": 0.4330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.3054786700692099e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.2622352984671964e-05,
+      "loss": 0.9831,
       "step": 15262
     },
     {
-      "epoch": 0.4191865095712834,
+      "epoch": 0.43311577752553915,
       "grad_norm": 0.0,
-      "learning_rate": 1.3053939692421813e-05,
-      "loss": 0.8365,
+      "learning_rate": 1.2621466059619459e-05,
+      "loss": 0.9215,
       "step": 15263
     },
     {
-      "epoch": 0.4192139737991266,
+      "epoch": 0.43314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.305309265998757e-05,
-      "loss": 1.0085,
+      "learning_rate": 1.2620579112423308e-05,
+      "loss": 0.8963,
       "step": 15264
     },
     {
-      "epoch": 0.4192414380269699,
+      "epoch": 0.4331725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 1.3052245603396075e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.2619692143091009e-05,
+      "loss": 1.0117,
       "step": 15265
     },
     {
-      "epoch": 0.4192689022548131,
+      "epoch": 0.43320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.3051398522654033e-05,
-      "loss": 1.0379,
+      "learning_rate": 1.2618805151630054e-05,
+      "loss": 0.9148,
       "step": 15266
     },
     {
-      "epoch": 0.41929636648265634,
+      "epoch": 0.43322928490351875,
       "grad_norm": 0.0,
-      "learning_rate": 1.3050551417768143e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.2617918138047931e-05,
+      "loss": 1.0457,
       "step": 15267
     },
     {
-      "epoch": 0.41932383071049956,
+      "epoch": 0.4332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.3049704288745106e-05,
-      "loss": 1.0334,
+      "learning_rate": 1.261703110235214e-05,
+      "loss": 1.0,
       "step": 15268
     },
     {
-      "epoch": 0.41935129493834283,
+      "epoch": 0.4332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.3048857135591632e-05,
-      "loss": 1.0393,
+      "learning_rate": 1.2616144044550167e-05,
+      "loss": 0.8303,
       "step": 15269
     },
     {
-      "epoch": 0.41937875916618605,
+      "epoch": 0.4333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.3048009958314415e-05,
-      "loss": 0.9044,
+      "learning_rate": 1.2615256964649509e-05,
+      "loss": 0.8666,
       "step": 15270
     },
     {
-      "epoch": 0.4194062233940293,
+      "epoch": 0.4333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.3047162756920167e-05,
-      "loss": 0.8569,
+      "learning_rate": 1.2614369862657657e-05,
+      "loss": 0.9876,
       "step": 15271
     },
     {
-      "epoch": 0.4194336876218725,
+      "epoch": 0.4333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.3046315531415586e-05,
-      "loss": 0.9517,
+      "learning_rate": 1.2613482738582104e-05,
+      "loss": 0.8499,
       "step": 15272
     },
     {
-      "epoch": 0.4194611518497157,
+      "epoch": 0.4333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.3045468281807372e-05,
-      "loss": 0.8839,
+      "learning_rate": 1.2612595592430347e-05,
+      "loss": 0.9307,
       "step": 15273
     },
     {
-      "epoch": 0.419488616077559,
+      "epoch": 0.433427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.3044621008102239e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.2611708424209877e-05,
+      "loss": 0.9146,
       "step": 15274
     },
     {
-      "epoch": 0.4195160803054022,
+      "epoch": 0.43345629965947785,
       "grad_norm": 0.0,
-      "learning_rate": 1.3043773710306881e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.2610821233928188e-05,
+      "loss": 0.9798,
       "step": 15275
     },
     {
-      "epoch": 0.41954354453324544,
+      "epoch": 0.43348467650397277,
       "grad_norm": 0.0,
-      "learning_rate": 1.3042926388428008e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.2609934021592776e-05,
+      "loss": 0.9216,
       "step": 15276
     },
     {
-      "epoch": 0.41957100876108866,
+      "epoch": 0.4335130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.3042079042472322e-05,
-      "loss": 0.8685,
+      "learning_rate": 1.2609046787211132e-05,
+      "loss": 0.8959,
       "step": 15277
     },
     {
-      "epoch": 0.41959847298893194,
+      "epoch": 0.43354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.3041231672446529e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.260815953079075e-05,
+      "loss": 0.9581,
       "step": 15278
     },
     {
-      "epoch": 0.41962593721677516,
+      "epoch": 0.43356980703745746,
       "grad_norm": 0.0,
-      "learning_rate": 1.3040384278357335e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.2607272252339133e-05,
+      "loss": 0.9691,
       "step": 15279
     },
     {
-      "epoch": 0.4196534014446184,
+      "epoch": 0.4335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.303953686021144e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.260638495186377e-05,
+      "loss": 1.1309,
       "step": 15280
     },
     {
-      "epoch": 0.4196808656724616,
+      "epoch": 0.43362656072644723,
       "grad_norm": 0.0,
-      "learning_rate": 1.3038689418015551e-05,
-      "loss": 0.8717,
+      "learning_rate": 1.2605497629372153e-05,
+      "loss": 0.9077,
       "step": 15281
     },
     {
-      "epoch": 0.4197083299003049,
+      "epoch": 0.4336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.3037841951776375e-05,
-      "loss": 0.981,
+      "learning_rate": 1.2604610284871782e-05,
+      "loss": 0.8498,
       "step": 15282
     },
     {
-      "epoch": 0.4197357941281481,
+      "epoch": 0.433683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.3036994461500618e-05,
-      "loss": 0.8563,
+      "learning_rate": 1.260372291837015e-05,
+      "loss": 0.9052,
       "step": 15283
     },
     {
-      "epoch": 0.4197632583559913,
+      "epoch": 0.4337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.3036146947194986e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.260283552987475e-05,
+      "loss": 0.8258,
       "step": 15284
     },
     {
-      "epoch": 0.41979072258383454,
+      "epoch": 0.4337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.3035299408866178e-05,
-      "loss": 0.9461,
+      "learning_rate": 1.2601948119393085e-05,
+      "loss": 0.8976,
       "step": 15285
     },
     {
-      "epoch": 0.41981818681167776,
+      "epoch": 0.4337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.303445184652091e-05,
-      "loss": 0.8461,
+      "learning_rate": 1.2601060686932649e-05,
+      "loss": 1.0012,
       "step": 15286
     },
     {
-      "epoch": 0.41984565103952104,
+      "epoch": 0.43379682179341655,
       "grad_norm": 0.0,
-      "learning_rate": 1.3033604260165884e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.2600173232500935e-05,
+      "loss": 0.9774,
       "step": 15287
     },
     {
-      "epoch": 0.41987311526736426,
+      "epoch": 0.43382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.30327566498078e-05,
-      "loss": 0.9291,
+      "learning_rate": 1.259928575610544e-05,
+      "loss": 0.9551,
       "step": 15288
     },
     {
-      "epoch": 0.4199005794952075,
+      "epoch": 0.4338535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.3031909015453377e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.2598398257753661e-05,
+      "loss": 0.9183,
       "step": 15289
     },
     {
-      "epoch": 0.4199280437230507,
+      "epoch": 0.43388195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.3031061357109312e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.2597510737453098e-05,
+      "loss": 0.8777,
       "step": 15290
     },
     {
-      "epoch": 0.419955507950894,
+      "epoch": 0.43391032917139616,
       "grad_norm": 0.0,
-      "learning_rate": 1.3030213674782318e-05,
-      "loss": 0.9597,
+      "learning_rate": 1.2596623195211242e-05,
+      "loss": 0.8613,
       "step": 15291
     },
     {
-      "epoch": 0.4199829721787372,
+      "epoch": 0.433938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.3029365968479097e-05,
-      "loss": 0.9326,
+      "learning_rate": 1.2595735631035593e-05,
+      "loss": 0.9422,
       "step": 15292
     },
     {
-      "epoch": 0.4200104364065804,
+      "epoch": 0.43396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.302851823820636e-05,
-      "loss": 0.9093,
+      "learning_rate": 1.259484804493365e-05,
+      "loss": 0.926,
       "step": 15293
     },
     {
-      "epoch": 0.42003790063442364,
+      "epoch": 0.4339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.3027670483970813e-05,
-      "loss": 0.8645,
+      "learning_rate": 1.2593960436912907e-05,
+      "loss": 1.0214,
       "step": 15294
     },
     {
-      "epoch": 0.4200653648622669,
+      "epoch": 0.4340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.3026822705779167e-05,
-      "loss": 0.9832,
+      "learning_rate": 1.2593072806980865e-05,
+      "loss": 0.8572,
       "step": 15295
     },
     {
-      "epoch": 0.42009282909011014,
+      "epoch": 0.4340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.3025974903638125e-05,
-      "loss": 0.9093,
+      "learning_rate": 1.2592185155145023e-05,
+      "loss": 0.9919,
       "step": 15296
     },
     {
-      "epoch": 0.42012029331795336,
+      "epoch": 0.4340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.3025127077554397e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.2591297481412872e-05,
+      "loss": 0.9872,
       "step": 15297
     },
     {
-      "epoch": 0.4201477575457966,
+      "epoch": 0.4341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.3024279227534692e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.2590409785791917e-05,
+      "loss": 0.9121,
       "step": 15298
     },
     {
-      "epoch": 0.4201752217736398,
+      "epoch": 0.43413734392735526,
       "grad_norm": 0.0,
-      "learning_rate": 1.3023431353585721e-05,
-      "loss": 0.936,
+      "learning_rate": 1.2589522068289655e-05,
+      "loss": 0.7882,
       "step": 15299
     },
     {
-      "epoch": 0.4202026860014831,
+      "epoch": 0.43416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.3022583455714186e-05,
-      "loss": 0.9512,
+      "learning_rate": 1.2588634328913581e-05,
+      "loss": 0.8831,
       "step": 15300
     },
     {
-      "epoch": 0.4202301502293263,
+      "epoch": 0.4341940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.3021735533926803e-05,
-      "loss": 0.974,
+      "learning_rate": 1.2587746567671197e-05,
+      "loss": 0.9539,
       "step": 15301
     },
     {
-      "epoch": 0.4202576144571695,
+      "epoch": 0.43422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.3020887588230276e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.2586858784570002e-05,
+      "loss": 0.9753,
       "step": 15302
     },
     {
-      "epoch": 0.42028507868501275,
+      "epoch": 0.43425085130533486,
       "grad_norm": 0.0,
-      "learning_rate": 1.3020039618631315e-05,
-      "loss": 0.965,
+      "learning_rate": 1.2585970979617494e-05,
+      "loss": 0.9725,
       "step": 15303
     },
     {
-      "epoch": 0.420312542912856,
+      "epoch": 0.4342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.3019191625136633e-05,
-      "loss": 0.9507,
+      "learning_rate": 1.2585083152821174e-05,
+      "loss": 1.0308,
       "step": 15304
     },
     {
-      "epoch": 0.42034000714069925,
+      "epoch": 0.43430760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.3018343607752938e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.2584195304188542e-05,
+      "loss": 0.9289,
       "step": 15305
     },
     {
-      "epoch": 0.42036747136854247,
+      "epoch": 0.4343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.3017495566486934e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.2583307433727094e-05,
+      "loss": 1.0144,
       "step": 15306
     },
     {
-      "epoch": 0.4203949355963857,
+      "epoch": 0.4343643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.3016647501345342e-05,
-      "loss": 1.039,
+      "learning_rate": 1.2582419541444331e-05,
+      "loss": 0.9529,
       "step": 15307
     },
     {
-      "epoch": 0.42042239982422897,
+      "epoch": 0.4343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.3015799412334867e-05,
-      "loss": 0.8532,
+      "learning_rate": 1.2581531627347754e-05,
+      "loss": 0.9682,
       "step": 15308
     },
     {
-      "epoch": 0.4204498640520722,
+      "epoch": 0.4344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.3014951299462213e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.2580643691444866e-05,
+      "loss": 0.9446,
       "step": 15309
     },
     {
-      "epoch": 0.4204773282799154,
+      "epoch": 0.4344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.3014103162734102e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.2579755733743164e-05,
+      "loss": 0.8885,
       "step": 15310
     },
     {
-      "epoch": 0.42050479250775863,
+      "epoch": 0.43447786606129396,
       "grad_norm": 0.0,
-      "learning_rate": 1.3013255002157237e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.2578867754250146e-05,
+      "loss": 0.9421,
       "step": 15311
     },
     {
-      "epoch": 0.4205322567356019,
+      "epoch": 0.4345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.301240681773833e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.257797975297332e-05,
+      "loss": 0.9532,
       "step": 15312
     },
     {
-      "epoch": 0.4205597209634451,
+      "epoch": 0.4345346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.3011558609484096e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.2577091729920184e-05,
+      "loss": 1.0043,
       "step": 15313
     },
     {
-      "epoch": 0.42058718519128835,
+      "epoch": 0.43456299659477865,
       "grad_norm": 0.0,
-      "learning_rate": 1.3010710377401242e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.2576203685098233e-05,
+      "loss": 0.9546,
       "step": 15314
     },
     {
-      "epoch": 0.42061464941913157,
+      "epoch": 0.43459137343927357,
       "grad_norm": 0.0,
-      "learning_rate": 1.3009862121496481e-05,
-      "loss": 0.8942,
+      "learning_rate": 1.2575315618514977e-05,
+      "loss": 0.8824,
       "step": 15315
     },
     {
-      "epoch": 0.4206421136469748,
+      "epoch": 0.4346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.3009013841776526e-05,
-      "loss": 0.841,
+      "learning_rate": 1.2574427530177915e-05,
+      "loss": 0.7345,
       "step": 15316
     },
     {
-      "epoch": 0.42066957787481807,
+      "epoch": 0.43464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.3008165538248085e-05,
-      "loss": 0.9005,
+      "learning_rate": 1.2573539420094548e-05,
+      "loss": 0.964,
       "step": 15317
     },
     {
-      "epoch": 0.4206970421026613,
+      "epoch": 0.43467650397275825,
       "grad_norm": 0.0,
-      "learning_rate": 1.3007317210917877e-05,
-      "loss": 0.8733,
+      "learning_rate": 1.2572651288272375e-05,
+      "loss": 0.8543,
       "step": 15318
     },
     {
-      "epoch": 0.4207245063305045,
+      "epoch": 0.4347048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.3006468859792612e-05,
-      "loss": 0.9521,
+      "learning_rate": 1.2571763134718904e-05,
+      "loss": 0.9098,
       "step": 15319
     },
     {
-      "epoch": 0.42075197055834773,
+      "epoch": 0.43473325766174803,
       "grad_norm": 0.0,
-      "learning_rate": 1.3005620484878994e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.2570874959441635e-05,
+      "loss": 0.8913,
       "step": 15320
     },
     {
-      "epoch": 0.420779434786191,
+      "epoch": 0.4347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.3004772086183745e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.2569986762448063e-05,
+      "loss": 0.889,
       "step": 15321
     },
     {
-      "epoch": 0.42080689901403423,
+      "epoch": 0.4347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.3003923663713578e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.2569098543745703e-05,
+      "loss": 0.8799,
       "step": 15322
     },
     {
-      "epoch": 0.42083436324187745,
+      "epoch": 0.43481838819523266,
       "grad_norm": 0.0,
-      "learning_rate": 1.3003075217475204e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.256821030334205e-05,
+      "loss": 0.8643,
       "step": 15323
     },
     {
-      "epoch": 0.4208618274697207,
+      "epoch": 0.4348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.300222674747533e-05,
-      "loss": 1.005,
+      "learning_rate": 1.2567322041244612e-05,
+      "loss": 0.8035,
       "step": 15324
     },
     {
-      "epoch": 0.42088929169756395,
+      "epoch": 0.4348751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.300137825372068e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.2566433757460886e-05,
+      "loss": 0.853,
       "step": 15325
     },
     {
-      "epoch": 0.42091675592540717,
+      "epoch": 0.43490351872871735,
       "grad_norm": 0.0,
-      "learning_rate": 1.3000529736217959e-05,
-      "loss": 0.8545,
+      "learning_rate": 1.2565545451998382e-05,
+      "loss": 0.9659,
       "step": 15326
     },
     {
-      "epoch": 0.4209442201532504,
+      "epoch": 0.43493189557321227,
       "grad_norm": 0.0,
-      "learning_rate": 1.2999681194973888e-05,
-      "loss": 1.0609,
+      "learning_rate": 1.2564657124864599e-05,
+      "loss": 0.9651,
       "step": 15327
     },
     {
-      "epoch": 0.4209716843810936,
+      "epoch": 0.43496027241770713,
       "grad_norm": 0.0,
-      "learning_rate": 1.2998832629995177e-05,
-      "loss": 0.9214,
+      "learning_rate": 1.2563768776067039e-05,
+      "loss": 1.0034,
       "step": 15328
     },
     {
-      "epoch": 0.42099914860893684,
+      "epoch": 0.43498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.2997984041288536e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.256288040561321e-05,
+      "loss": 0.9237,
       "step": 15329
     },
     {
-      "epoch": 0.4210266128367801,
+      "epoch": 0.43501702610669696,
       "grad_norm": 0.0,
-      "learning_rate": 1.299713542886069e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.2561992013510618e-05,
+      "loss": 0.8547,
       "step": 15330
     },
     {
-      "epoch": 0.42105407706462333,
+      "epoch": 0.4350454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.2996286792718345e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.2561103599766761e-05,
+      "loss": 1.0139,
       "step": 15331
     },
     {
-      "epoch": 0.42108154129246655,
+      "epoch": 0.43507377979568673,
       "grad_norm": 0.0,
-      "learning_rate": 1.2995438132868217e-05,
-      "loss": 0.92,
+      "learning_rate": 1.2560215164389149e-05,
+      "loss": 0.9864,
       "step": 15332
     },
     {
-      "epoch": 0.4211090055203098,
+      "epoch": 0.4351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.2994589449317023e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.2559326707385285e-05,
+      "loss": 0.956,
       "step": 15333
     },
     {
-      "epoch": 0.42113646974815305,
+      "epoch": 0.4351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.2993740742071477e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.2558438228762673e-05,
+      "loss": 0.9103,
       "step": 15334
     },
     {
-      "epoch": 0.4211639339759963,
+      "epoch": 0.4351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.2992892011138296e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.2557549728528815e-05,
+      "loss": 0.9881,
       "step": 15335
     },
     {
-      "epoch": 0.4211913982038395,
+      "epoch": 0.4351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.2992043256524192e-05,
-      "loss": 0.8885,
+      "learning_rate": 1.2556661206691221e-05,
+      "loss": 0.8237,
       "step": 15336
     },
     {
-      "epoch": 0.4212188624316827,
+      "epoch": 0.4352156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.2991194478235885e-05,
-      "loss": 0.8577,
+      "learning_rate": 1.2555772663257397e-05,
+      "loss": 0.9039,
       "step": 15337
     },
     {
-      "epoch": 0.421246326659526,
+      "epoch": 0.43524404086265606,
       "grad_norm": 0.0,
-      "learning_rate": 1.2990345676280085e-05,
-      "loss": 0.9446,
+      "learning_rate": 1.2554884098234843e-05,
+      "loss": 0.8851,
       "step": 15338
     },
     {
-      "epoch": 0.4212737908873692,
+      "epoch": 0.43527241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 1.2989496850663517e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.255399551163107e-05,
+      "loss": 0.818,
       "step": 15339
     },
     {
-      "epoch": 0.42130125511521244,
+      "epoch": 0.43530079455164583,
       "grad_norm": 0.0,
-      "learning_rate": 1.2988648001392889e-05,
-      "loss": 0.9566,
+      "learning_rate": 1.2553106903453583e-05,
+      "loss": 0.9899,
       "step": 15340
     },
     {
-      "epoch": 0.42132871934305566,
+      "epoch": 0.43532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.298779912847492e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.2552218273709889e-05,
+      "loss": 0.8233,
       "step": 15341
     },
     {
-      "epoch": 0.4213561835708989,
+      "epoch": 0.43535754824063566,
       "grad_norm": 0.0,
-      "learning_rate": 1.298695023191633e-05,
-      "loss": 0.983,
+      "learning_rate": 1.2551329622407486e-05,
+      "loss": 1.0166,
       "step": 15342
     },
     {
-      "epoch": 0.42138364779874216,
+      "epoch": 0.4353859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.2986101311723829e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.2550440949553892e-05,
+      "loss": 0.8959,
       "step": 15343
     },
     {
-      "epoch": 0.4214111120265854,
+      "epoch": 0.43541430192962544,
       "grad_norm": 0.0,
-      "learning_rate": 1.2985252367904137e-05,
-      "loss": 0.9225,
+      "learning_rate": 1.254955225515661e-05,
+      "loss": 0.8012,
       "step": 15344
     },
     {
-      "epoch": 0.4214385762544286,
+      "epoch": 0.4354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.2984403400463973e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.254866353922314e-05,
+      "loss": 0.8971,
       "step": 15345
     },
     {
-      "epoch": 0.4214660404822718,
+      "epoch": 0.4354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.2983554409410054e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.2547774801760997e-05,
+      "loss": 0.9373,
       "step": 15346
     },
     {
-      "epoch": 0.4214935047101151,
+      "epoch": 0.4354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.2982705394749097e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.2546886042777686e-05,
+      "loss": 0.8799,
       "step": 15347
     },
     {
-      "epoch": 0.4215209689379583,
+      "epoch": 0.435527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.298185635648782e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.2545997262280713e-05,
+      "loss": 0.8432,
       "step": 15348
     },
     {
-      "epoch": 0.42154843316580154,
+      "epoch": 0.4355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.298100729463294e-05,
-      "loss": 0.943,
+      "learning_rate": 1.2545108460277591e-05,
+      "loss": 0.8383,
       "step": 15349
     },
     {
-      "epoch": 0.42157589739364476,
+      "epoch": 0.43558456299659476,
       "grad_norm": 0.0,
-      "learning_rate": 1.2980158209191173e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.254421963677582e-05,
+      "loss": 0.9567,
       "step": 15350
     },
     {
-      "epoch": 0.42160336162148804,
+      "epoch": 0.4356129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.2979309100169246e-05,
-      "loss": 0.9162,
+      "learning_rate": 1.254333079178291e-05,
+      "loss": 0.8982,
       "step": 15351
     },
     {
-      "epoch": 0.42163082584933126,
+      "epoch": 0.4356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.2978459967573866e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.2542441925306375e-05,
+      "loss": 0.9732,
       "step": 15352
     },
     {
-      "epoch": 0.4216582900771745,
+      "epoch": 0.43566969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.2977610811411758e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.2541553037353716e-05,
+      "loss": 0.8935,
       "step": 15353
     },
     {
-      "epoch": 0.4216857543050177,
+      "epoch": 0.43569807037457436,
       "grad_norm": 0.0,
-      "learning_rate": 1.2976761631689641e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.2540664127932444e-05,
+      "loss": 0.9894,
       "step": 15354
     },
     {
-      "epoch": 0.4217132185328609,
+      "epoch": 0.4357264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.2975912428414232e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.2539775197050065e-05,
+      "loss": 0.9398,
       "step": 15355
     },
     {
-      "epoch": 0.4217406827607042,
+      "epoch": 0.43575482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.297506320159225e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.2538886244714096e-05,
+      "loss": 0.8484,
       "step": 15356
     },
     {
-      "epoch": 0.4217681469885474,
+      "epoch": 0.435783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.2974213951230418e-05,
-      "loss": 0.9768,
+      "learning_rate": 1.2537997270932041e-05,
+      "loss": 0.8796,
       "step": 15357
     },
     {
-      "epoch": 0.42179561121639064,
+      "epoch": 0.4358115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.2973364677335451e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.2537108275711405e-05,
+      "loss": 1.0286,
       "step": 15358
     },
     {
-      "epoch": 0.42182307544423386,
+      "epoch": 0.43583995459704883,
       "grad_norm": 0.0,
-      "learning_rate": 1.2972515379914073e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.2536219259059703e-05,
+      "loss": 0.923,
       "step": 15359
     },
     {
-      "epoch": 0.42185053967207714,
+      "epoch": 0.4358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.2971666058973002e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.2535330220984445e-05,
+      "loss": 0.9248,
       "step": 15360
     },
     {
-      "epoch": 0.42187800389992036,
+      "epoch": 0.4358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.2970816714518954e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.2534441161493134e-05,
+      "loss": 0.7752,
       "step": 15361
     },
     {
-      "epoch": 0.4219054681277636,
+      "epoch": 0.43592508513053346,
       "grad_norm": 0.0,
-      "learning_rate": 1.2969967346558656e-05,
-      "loss": 0.9842,
+      "learning_rate": 1.2533552080593286e-05,
+      "loss": 1.002,
       "step": 15362
     },
     {
-      "epoch": 0.4219329323556068,
+      "epoch": 0.4359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.296911795509883e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.253266297829241e-05,
+      "loss": 0.8741,
       "step": 15363
     },
     {
-      "epoch": 0.4219603965834501,
+      "epoch": 0.4359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.2968268540146185e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.2531773854598019e-05,
+      "loss": 0.9221,
       "step": 15364
     },
     {
-      "epoch": 0.4219878608112933,
+      "epoch": 0.43601021566401815,
       "grad_norm": 0.0,
-      "learning_rate": 1.2967419101707453e-05,
-      "loss": 1.0207,
+      "learning_rate": 1.2530884709517616e-05,
+      "loss": 0.975,
       "step": 15365
     },
     {
-      "epoch": 0.4220153250391365,
+      "epoch": 0.43603859250851307,
       "grad_norm": 0.0,
-      "learning_rate": 1.2966569639789353e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.2529995543058721e-05,
+      "loss": 0.8948,
       "step": 15366
     },
     {
-      "epoch": 0.42204278926697975,
+      "epoch": 0.4360669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 1.2965720154398604e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.2529106355228836e-05,
+      "loss": 0.9197,
       "step": 15367
     },
     {
-      "epoch": 0.42207025349482297,
+      "epoch": 0.43609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.2964870645541927e-05,
-      "loss": 0.887,
+      "learning_rate": 1.2528217146035478e-05,
+      "loss": 0.9172,
       "step": 15368
     },
     {
-      "epoch": 0.42209771772266624,
+      "epoch": 0.43612372304199776,
       "grad_norm": 0.0,
-      "learning_rate": 1.2964021113226046e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.2527327915486154e-05,
+      "loss": 0.7613,
       "step": 15369
     },
     {
-      "epoch": 0.42212518195050946,
+      "epoch": 0.4361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.296317155745768e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.252643866358838e-05,
+      "loss": 0.9375,
       "step": 15370
     },
     {
-      "epoch": 0.4221526461783527,
+      "epoch": 0.43618047673098753,
       "grad_norm": 0.0,
-      "learning_rate": 1.2962321978243555e-05,
-      "loss": 0.8701,
+      "learning_rate": 1.2525549390349665e-05,
+      "loss": 0.8953,
       "step": 15371
     },
     {
-      "epoch": 0.4221801104061959,
+      "epoch": 0.4362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.2961472375590392e-05,
-      "loss": 0.9358,
+      "learning_rate": 1.2524660095777516e-05,
+      "loss": 0.8624,
       "step": 15372
     },
     {
-      "epoch": 0.4222075746340392,
+      "epoch": 0.4362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.2960622749504909e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.2523770779879457e-05,
+      "loss": 0.8771,
       "step": 15373
     },
     {
-      "epoch": 0.4222350388618824,
+      "epoch": 0.43626560726447217,
       "grad_norm": 0.0,
-      "learning_rate": 1.2959773099993834e-05,
-      "loss": 0.92,
+      "learning_rate": 1.2522881442662989e-05,
+      "loss": 0.9965,
       "step": 15374
     },
     {
-      "epoch": 0.4222625030897256,
+      "epoch": 0.4362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.295892342706389e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.2521992084135626e-05,
+      "loss": 0.9476,
       "step": 15375
     },
     {
-      "epoch": 0.42228996731756885,
+      "epoch": 0.436322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.2958073730721795e-05,
-      "loss": 0.8224,
+      "learning_rate": 1.2521102704304884e-05,
+      "loss": 0.8734,
       "step": 15376
     },
     {
-      "epoch": 0.4223174315454121,
+      "epoch": 0.43635073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.2957224010974275e-05,
-      "loss": 0.9877,
+      "learning_rate": 1.2520213303178274e-05,
+      "loss": 0.8948,
       "step": 15377
     },
     {
-      "epoch": 0.42234489577325535,
+      "epoch": 0.43637911464245177,
       "grad_norm": 0.0,
-      "learning_rate": 1.2956374267828053e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.2519323880763312e-05,
+      "loss": 0.8629,
       "step": 15378
     },
     {
-      "epoch": 0.42237236000109857,
+      "epoch": 0.43640749148694663,
       "grad_norm": 0.0,
-      "learning_rate": 1.2955524501289854e-05,
-      "loss": 0.939,
+      "learning_rate": 1.2518434437067507e-05,
+      "loss": 0.7977,
       "step": 15379
     },
     {
-      "epoch": 0.4223998242289418,
+      "epoch": 0.43643586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.2954674711366398e-05,
-      "loss": 1.0573,
+      "learning_rate": 1.251754497209837e-05,
+      "loss": 0.9614,
       "step": 15380
     },
     {
-      "epoch": 0.422427288456785,
+      "epoch": 0.43646424517593646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2953824898064412e-05,
-      "loss": 0.8255,
+      "learning_rate": 1.251665548586342e-05,
+      "loss": 0.8568,
       "step": 15381
     },
     {
-      "epoch": 0.4224547526846283,
+      "epoch": 0.4364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.2952975061390618e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.2515765978370167e-05,
+      "loss": 0.8828,
       "step": 15382
     },
     {
-      "epoch": 0.4224822169124715,
+      "epoch": 0.43652099886492624,
       "grad_norm": 0.0,
-      "learning_rate": 1.2952125201351743e-05,
-      "loss": 0.921,
+      "learning_rate": 1.2514876449626126e-05,
+      "loss": 0.9843,
       "step": 15383
     },
     {
-      "epoch": 0.42250968114031473,
+      "epoch": 0.4365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.2951275317954512e-05,
-      "loss": 0.8984,
+      "learning_rate": 1.2513986899638809e-05,
+      "loss": 1.0537,
       "step": 15384
     },
     {
-      "epoch": 0.42253714536815795,
+      "epoch": 0.436577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.2950425411205643e-05,
-      "loss": 0.9406,
+      "learning_rate": 1.2513097328415733e-05,
+      "loss": 0.8456,
       "step": 15385
     },
     {
-      "epoch": 0.42256460959600123,
+      "epoch": 0.43660612939841087,
       "grad_norm": 0.0,
-      "learning_rate": 1.2949575481111867e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.251220773596441e-05,
+      "loss": 0.9893,
       "step": 15386
     },
     {
-      "epoch": 0.42259207382384445,
+      "epoch": 0.4366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.294872552767991e-05,
-      "loss": 1.0212,
+      "learning_rate": 1.2511318122292358e-05,
+      "loss": 0.8875,
       "step": 15387
     },
     {
-      "epoch": 0.42261953805168767,
+      "epoch": 0.4366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.294787555091649e-05,
-      "loss": 1.0118,
+      "learning_rate": 1.2510428487407088e-05,
+      "loss": 0.9125,
       "step": 15388
     },
     {
-      "epoch": 0.4226470022795309,
+      "epoch": 0.43669125993189556,
       "grad_norm": 0.0,
-      "learning_rate": 1.2947025550828338e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.2509538831316112e-05,
+      "loss": 0.9021,
       "step": 15389
     },
     {
-      "epoch": 0.42267446650737417,
+      "epoch": 0.4367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.294617552742218e-05,
-      "loss": 0.9917,
+      "learning_rate": 1.2508649154026953e-05,
+      "loss": 0.9938,
       "step": 15390
     },
     {
-      "epoch": 0.4227019307352174,
+      "epoch": 0.43674801362088533,
       "grad_norm": 0.0,
-      "learning_rate": 1.294532548070474e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.2507759455547121e-05,
+      "loss": 0.867,
       "step": 15391
     },
     {
-      "epoch": 0.4227293949630606,
+      "epoch": 0.43677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.2944475410682741e-05,
-      "loss": 0.8928,
+      "learning_rate": 1.250686973588413e-05,
+      "loss": 0.8469,
       "step": 15392
     },
     {
-      "epoch": 0.42275685919090383,
+      "epoch": 0.43680476730987516,
       "grad_norm": 0.0,
-      "learning_rate": 1.2943625317362916e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.2505979995045498e-05,
+      "loss": 0.9564,
       "step": 15393
     },
     {
-      "epoch": 0.42278432341874705,
+      "epoch": 0.43683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.2942775200751985e-05,
-      "loss": 0.8619,
+      "learning_rate": 1.2505090233038743e-05,
+      "loss": 0.9137,
       "step": 15394
     },
     {
-      "epoch": 0.42281178764659033,
+      "epoch": 0.43686152099886494,
       "grad_norm": 0.0,
-      "learning_rate": 1.2941925060856677e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.2504200449871378e-05,
+      "loss": 0.82,
       "step": 15395
     },
     {
-      "epoch": 0.42283925187443355,
+      "epoch": 0.4368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.294107489768372e-05,
-      "loss": 0.8764,
+      "learning_rate": 1.2503310645550918e-05,
+      "loss": 0.9621,
       "step": 15396
     },
     {
-      "epoch": 0.4228667161022768,
+      "epoch": 0.4369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.2940224711239835e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.2502420820084879e-05,
+      "loss": 0.8511,
       "step": 15397
     },
     {
-      "epoch": 0.42289418033012,
+      "epoch": 0.43694665153234963,
       "grad_norm": 0.0,
-      "learning_rate": 1.2939374501531754e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.2501530973480783e-05,
+      "loss": 0.9661,
       "step": 15398
     },
     {
-      "epoch": 0.42292164455796327,
+      "epoch": 0.4369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.2938524268566208e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.2500641105746137e-05,
+      "loss": 0.7917,
       "step": 15399
     },
     {
-      "epoch": 0.4229491087858065,
+      "epoch": 0.4370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.2937674012349915e-05,
-      "loss": 0.8324,
+      "learning_rate": 1.2499751216888464e-05,
+      "loss": 0.8523,
       "step": 15400
     },
     {
-      "epoch": 0.4229765730136497,
+      "epoch": 0.43703178206583426,
       "grad_norm": 0.0,
-      "learning_rate": 1.2936823732889607e-05,
-      "loss": 0.9444,
+      "learning_rate": 1.2498861306915283e-05,
+      "loss": 0.9639,
       "step": 15401
     },
     {
-      "epoch": 0.42300403724149294,
+      "epoch": 0.4370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.2935973430192017e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.2497971375834106e-05,
+      "loss": 1.0098,
       "step": 15402
     },
     {
-      "epoch": 0.4230315014693362,
+      "epoch": 0.43708853575482404,
       "grad_norm": 0.0,
-      "learning_rate": 1.2935123104263864e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.2497081423652455e-05,
+      "loss": 0.8893,
       "step": 15403
     },
     {
-      "epoch": 0.42305896569717943,
+      "epoch": 0.43711691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.293427275511188e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.2496191450377844e-05,
+      "loss": 0.9988,
       "step": 15404
     },
     {
-      "epoch": 0.42308642992502266,
+      "epoch": 0.43714528944381387,
       "grad_norm": 0.0,
-      "learning_rate": 1.2933422382742795e-05,
-      "loss": 0.7981,
+      "learning_rate": 1.249530145601779e-05,
+      "loss": 0.8294,
       "step": 15405
     },
     {
-      "epoch": 0.4231138941528659,
+      "epoch": 0.4371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.2932571987163333e-05,
-      "loss": 0.8582,
+      "learning_rate": 1.2494411440579814e-05,
+      "loss": 0.9612,
       "step": 15406
     },
     {
-      "epoch": 0.42314135838070915,
+      "epoch": 0.43720204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 1.2931721568380229e-05,
-      "loss": 1.0881,
+      "learning_rate": 1.2493521404071432e-05,
+      "loss": 0.9424,
       "step": 15407
     },
     {
-      "epoch": 0.4231688226085524,
+      "epoch": 0.4372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.2930871126400208e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.2492631346500163e-05,
+      "loss": 1.0049,
       "step": 15408
     },
     {
-      "epoch": 0.4231962868363956,
+      "epoch": 0.4372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.2930020661229997e-05,
-      "loss": 0.9417,
+      "learning_rate": 1.2491741267873522e-05,
+      "loss": 0.8901,
       "step": 15409
     },
     {
-      "epoch": 0.4232237510642388,
+      "epoch": 0.43728717366628833,
       "grad_norm": 0.0,
-      "learning_rate": 1.2929170172876329e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.2490851168199036e-05,
+      "loss": 0.9009,
       "step": 15410
     },
     {
-      "epoch": 0.42325121529208204,
+      "epoch": 0.4373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.2928319661345932e-05,
-      "loss": 0.9703,
+      "learning_rate": 1.2489961047484215e-05,
+      "loss": 0.8317,
       "step": 15411
     },
     {
-      "epoch": 0.4232786795199253,
+      "epoch": 0.4373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.2927469126645533e-05,
-      "loss": 0.8507,
+      "learning_rate": 1.248907090573658e-05,
+      "loss": 0.9449,
       "step": 15412
     },
     {
-      "epoch": 0.42330614374776854,
+      "epoch": 0.43737230419977297,
       "grad_norm": 0.0,
-      "learning_rate": 1.2926618568781869e-05,
-      "loss": 0.8997,
+      "learning_rate": 1.2488180742963654e-05,
+      "loss": 0.9928,
       "step": 15413
     },
     {
-      "epoch": 0.42333360797561176,
+      "epoch": 0.4374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.2925767987761661e-05,
-      "loss": 0.9918,
+      "learning_rate": 1.2487290559172953e-05,
+      "loss": 0.9015,
       "step": 15414
     },
     {
-      "epoch": 0.423361072203455,
+      "epoch": 0.4374290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.2924917383591647e-05,
-      "loss": 0.8692,
+      "learning_rate": 1.2486400354371995e-05,
+      "loss": 0.8026,
       "step": 15415
     },
     {
-      "epoch": 0.42338853643129826,
+      "epoch": 0.43745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.292406675627855e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.2485510128568302e-05,
+      "loss": 0.9518,
       "step": 15416
     },
     {
-      "epoch": 0.4234160006591415,
+      "epoch": 0.43748581157775257,
       "grad_norm": 0.0,
-      "learning_rate": 1.2923216105829103e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.2484619881769393e-05,
+      "loss": 0.8883,
       "step": 15417
     },
     {
-      "epoch": 0.4234434648869847,
+      "epoch": 0.43751418842224743,
       "grad_norm": 0.0,
-      "learning_rate": 1.2922365432250042e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.2483729613982789e-05,
+      "loss": 0.8374,
       "step": 15418
     },
     {
-      "epoch": 0.4234709291148279,
+      "epoch": 0.43754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.2921514735548095e-05,
-      "loss": 1.0306,
+      "learning_rate": 1.2482839325216008e-05,
+      "loss": 0.9046,
       "step": 15419
     },
     {
-      "epoch": 0.4234983933426712,
+      "epoch": 0.4375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.2920664015729986e-05,
-      "loss": 0.9807,
+      "learning_rate": 1.2481949015476573e-05,
+      "loss": 1.0464,
       "step": 15420
     },
     {
-      "epoch": 0.4235258575705144,
+      "epoch": 0.4375993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.2919813272802455e-05,
-      "loss": 0.9341,
+      "learning_rate": 1.2481058684772006e-05,
+      "loss": 1.0369,
       "step": 15421
     },
     {
-      "epoch": 0.42355332179835764,
+      "epoch": 0.43762769580022703,
       "grad_norm": 0.0,
-      "learning_rate": 1.291896250677223e-05,
-      "loss": 0.931,
+      "learning_rate": 1.2480168333109818e-05,
+      "loss": 1.0157,
       "step": 15422
     },
     {
-      "epoch": 0.42358078602620086,
+      "epoch": 0.4376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.2918111717646045e-05,
-      "loss": 1.0122,
+      "learning_rate": 1.247927796049754e-05,
+      "loss": 0.9761,
       "step": 15423
     },
     {
-      "epoch": 0.4236082502540441,
+      "epoch": 0.4376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.2917260905430626e-05,
-      "loss": 0.8891,
+      "learning_rate": 1.2478387566942689e-05,
+      "loss": 0.9554,
       "step": 15424
     },
     {
-      "epoch": 0.42363571448188736,
+      "epoch": 0.43771282633371167,
       "grad_norm": 0.0,
-      "learning_rate": 1.2916410070132711e-05,
-      "loss": 0.9296,
+      "learning_rate": 1.2477497152452788e-05,
+      "loss": 0.8943,
       "step": 15425
     },
     {
-      "epoch": 0.4236631787097306,
+      "epoch": 0.4377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.291555921175903e-05,
-      "loss": 0.8835,
+      "learning_rate": 1.2476606717035355e-05,
+      "loss": 0.973,
       "step": 15426
     },
     {
-      "epoch": 0.4236906429375738,
+      "epoch": 0.4377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.2914708330316314e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.2475716260697917e-05,
+      "loss": 0.87,
       "step": 15427
     },
     {
-      "epoch": 0.423718107165417,
+      "epoch": 0.43779795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 1.29138574258113e-05,
-      "loss": 1.0527,
+      "learning_rate": 1.2474825783447993e-05,
+      "loss": 0.9879,
       "step": 15428
     },
     {
-      "epoch": 0.4237455713932603,
+      "epoch": 0.4378263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.2913006498250714e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.2473935285293099e-05,
+      "loss": 0.8622,
       "step": 15429
     },
     {
-      "epoch": 0.4237730356211035,
+      "epoch": 0.43785471055618613,
       "grad_norm": 0.0,
-      "learning_rate": 1.291215554764129e-05,
-      "loss": 0.8184,
+      "learning_rate": 1.2473044766240766e-05,
+      "loss": 0.8893,
       "step": 15430
     },
     {
-      "epoch": 0.42380049984894674,
+      "epoch": 0.43788308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.291130457398977e-05,
-      "loss": 0.8704,
+      "learning_rate": 1.2472154226298513e-05,
+      "loss": 0.8691,
       "step": 15431
     },
     {
-      "epoch": 0.42382796407678996,
+      "epoch": 0.43791146424517596,
       "grad_norm": 0.0,
-      "learning_rate": 1.2910453577302874e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.247126366547386e-05,
+      "loss": 0.9655,
       "step": 15432
     },
     {
-      "epoch": 0.42385542830463324,
+      "epoch": 0.4379398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.2909602557587344e-05,
-      "loss": 0.9079,
+      "learning_rate": 1.2470373083774335e-05,
+      "loss": 0.8057,
       "step": 15433
     },
     {
-      "epoch": 0.42388289253247646,
+      "epoch": 0.43796821793416574,
       "grad_norm": 0.0,
-      "learning_rate": 1.290875151484991e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.2469482481207456e-05,
+      "loss": 0.897,
       "step": 15434
     },
     {
-      "epoch": 0.4239103567603197,
+      "epoch": 0.4379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.290790044909731e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.2468591857780747e-05,
+      "loss": 0.9365,
       "step": 15435
     },
     {
-      "epoch": 0.4239378209881629,
+      "epoch": 0.4380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.2907049360336275e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.246770121350173e-05,
+      "loss": 0.8759,
       "step": 15436
     },
     {
-      "epoch": 0.4239652852160061,
+      "epoch": 0.43805334846765037,
       "grad_norm": 0.0,
-      "learning_rate": 1.2906198248573541e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.2466810548377932e-05,
+      "loss": 0.868,
       "step": 15437
     },
     {
-      "epoch": 0.4239927494438494,
+      "epoch": 0.4380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.2905347113815836e-05,
-      "loss": 0.7922,
+      "learning_rate": 1.2465919862416874e-05,
+      "loss": 1.0137,
       "step": 15438
     },
     {
-      "epoch": 0.4240202136716926,
+      "epoch": 0.4381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.2904495956069902e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.2465029155626081e-05,
+      "loss": 0.8893,
       "step": 15439
     },
     {
-      "epoch": 0.42404767789953585,
+      "epoch": 0.43813847900113506,
       "grad_norm": 0.0,
-      "learning_rate": 1.2903644775342472e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.2464138428013072e-05,
+      "loss": 0.8649,
       "step": 15440
     },
     {
-      "epoch": 0.42407514212737907,
+      "epoch": 0.43816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2902793571640278e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.246324767958538e-05,
+      "loss": 0.8905,
       "step": 15441
     },
     {
-      "epoch": 0.42410260635522234,
+      "epoch": 0.43819523269012484,
       "grad_norm": 0.0,
-      "learning_rate": 1.2901942344970056e-05,
-      "loss": 0.8486,
+      "learning_rate": 1.2462356910350524e-05,
+      "loss": 0.9274,
       "step": 15442
     },
     {
-      "epoch": 0.42413007058306557,
+      "epoch": 0.43822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.2901091095338545e-05,
-      "loss": 0.8739,
+      "learning_rate": 1.2461466120316024e-05,
+      "loss": 0.9726,
       "step": 15443
     },
     {
-      "epoch": 0.4241575348109088,
+      "epoch": 0.43825198637911467,
       "grad_norm": 0.0,
-      "learning_rate": 1.2900239822752474e-05,
-      "loss": 0.805,
+      "learning_rate": 1.2460575309489414e-05,
+      "loss": 0.9015,
       "step": 15444
     },
     {
-      "epoch": 0.424184999038752,
+      "epoch": 0.4382803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.2899388527218583e-05,
-      "loss": 0.8275,
+      "learning_rate": 1.245968447787821e-05,
+      "loss": 1.0099,
       "step": 15445
     },
     {
-      "epoch": 0.4242124632665953,
+      "epoch": 0.43830874006810444,
       "grad_norm": 0.0,
-      "learning_rate": 1.2898537208743607e-05,
-      "loss": 0.8729,
+      "learning_rate": 1.245879362548994e-05,
+      "loss": 0.9427,
       "step": 15446
     },
     {
-      "epoch": 0.4242399274944385,
+      "epoch": 0.4383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.2897685867334285e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.2457902752332131e-05,
+      "loss": 0.8752,
       "step": 15447
     },
     {
-      "epoch": 0.42426739172228173,
+      "epoch": 0.4383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.2896834502997346e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.245701185841231e-05,
+      "loss": 0.9006,
       "step": 15448
     },
     {
-      "epoch": 0.42429485595012495,
+      "epoch": 0.43839387060158913,
       "grad_norm": 0.0,
-      "learning_rate": 1.2895983115739532e-05,
-      "loss": 0.9325,
+      "learning_rate": 1.2456120943737996e-05,
+      "loss": 0.8766,
       "step": 15449
     },
     {
-      "epoch": 0.42432232017796817,
+      "epoch": 0.438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2895131705567577e-05,
-      "loss": 0.882,
+      "learning_rate": 1.245523000831672e-05,
+      "loss": 0.744,
       "step": 15450
     },
     {
-      "epoch": 0.42434978440581145,
+      "epoch": 0.4384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.289428027248822e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.2454339052156007e-05,
+      "loss": 0.8572,
       "step": 15451
     },
     {
-      "epoch": 0.42437724863365467,
+      "epoch": 0.43847900113507376,
       "grad_norm": 0.0,
-      "learning_rate": 1.2893428816508195e-05,
-      "loss": 1.0514,
+      "learning_rate": 1.245344807526338e-05,
+      "loss": 0.9109,
       "step": 15452
     },
     {
-      "epoch": 0.4244047128614979,
+      "epoch": 0.4385073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.289257733763424e-05,
-      "loss": 0.9554,
+      "learning_rate": 1.245255707764636e-05,
+      "loss": 1.0157,
       "step": 15453
     },
     {
-      "epoch": 0.4244321770893411,
+      "epoch": 0.43853575482406354,
       "grad_norm": 0.0,
-      "learning_rate": 1.2891725835873092e-05,
-      "loss": 0.951,
+      "learning_rate": 1.2451666059312488e-05,
+      "loss": 0.8327,
       "step": 15454
     },
     {
-      "epoch": 0.4244596413171844,
+      "epoch": 0.43856413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.2890874311231492e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.2450775020269283e-05,
+      "loss": 0.8789,
       "step": 15455
     },
     {
-      "epoch": 0.4244871055450276,
+      "epoch": 0.43859250851305337,
       "grad_norm": 0.0,
-      "learning_rate": 1.2890022763716173e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.2449883960524267e-05,
+      "loss": 0.9889,
       "step": 15456
     },
     {
-      "epoch": 0.42451456977287083,
+      "epoch": 0.43862088535754823,
       "grad_norm": 0.0,
-      "learning_rate": 1.2889171193333873e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.2448992880084973e-05,
+      "loss": 1.005,
       "step": 15457
     },
     {
-      "epoch": 0.42454203400071405,
+      "epoch": 0.43864926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 1.2888319600091332e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.244810177895893e-05,
+      "loss": 0.9157,
       "step": 15458
     },
     {
-      "epoch": 0.42456949822855733,
+      "epoch": 0.438677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.288746798399529e-05,
-      "loss": 0.9151,
+      "learning_rate": 1.2447210657153655e-05,
+      "loss": 0.9216,
       "step": 15459
     },
     {
-      "epoch": 0.42459696245640055,
+      "epoch": 0.4387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.2886616345052481e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.2446319514676685e-05,
+      "loss": 0.984,
       "step": 15460
     },
     {
-      "epoch": 0.42462442668424377,
+      "epoch": 0.43873439273552783,
       "grad_norm": 0.0,
-      "learning_rate": 1.2885764683269646e-05,
-      "loss": 0.9504,
+      "learning_rate": 1.2445428351535542e-05,
+      "loss": 1.0345,
       "step": 15461
     },
     {
-      "epoch": 0.424651890912087,
+      "epoch": 0.4387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.288491299865352e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.2444537167737759e-05,
+      "loss": 0.9522,
       "step": 15462
     },
     {
-      "epoch": 0.4246793551399302,
+      "epoch": 0.4387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.2884061291210848e-05,
-      "loss": 0.8283,
+      "learning_rate": 1.2443645963290856e-05,
+      "loss": 1.0924,
       "step": 15463
     },
     {
-      "epoch": 0.4247068193677735,
+      "epoch": 0.43881952326901247,
       "grad_norm": 0.0,
-      "learning_rate": 1.2883209560948366e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.244275473820237e-05,
+      "loss": 1.014,
       "step": 15464
     },
     {
-      "epoch": 0.4247342835956167,
+      "epoch": 0.4388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.2882357807872811e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.2441863492479825e-05,
+      "loss": 0.8185,
       "step": 15465
     },
     {
-      "epoch": 0.42476174782345993,
+      "epoch": 0.43887627695800224,
       "grad_norm": 0.0,
-      "learning_rate": 1.2881506031990925e-05,
-      "loss": 0.8859,
+      "learning_rate": 1.2440972226130746e-05,
+      "loss": 0.9408,
       "step": 15466
     },
     {
-      "epoch": 0.42478921205130316,
+      "epoch": 0.43890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.2880654233309451e-05,
-      "loss": 0.9362,
+      "learning_rate": 1.2440080939162666e-05,
+      "loss": 0.9034,
       "step": 15467
     },
     {
-      "epoch": 0.42481667627914643,
+      "epoch": 0.4389330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 1.287980241183512e-05,
-      "loss": 0.9549,
+      "learning_rate": 1.2439189631583113e-05,
+      "loss": 0.8758,
       "step": 15468
     },
     {
-      "epoch": 0.42484414050698965,
+      "epoch": 0.43896140749148693,
       "grad_norm": 0.0,
-      "learning_rate": 1.287895056757468e-05,
-      "loss": 0.9698,
+      "learning_rate": 1.2438298303399616e-05,
+      "loss": 0.9144,
       "step": 15469
     },
     {
-      "epoch": 0.4248716047348329,
+      "epoch": 0.43898978433598185,
       "grad_norm": 0.0,
-      "learning_rate": 1.2878098700534864e-05,
-      "loss": 1.0673,
+      "learning_rate": 1.2437406954619701e-05,
+      "loss": 0.9377,
       "step": 15470
     },
     {
-      "epoch": 0.4248990689626761,
+      "epoch": 0.4390181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.2877246810722422e-05,
-      "loss": 0.9308,
+      "learning_rate": 1.2436515585250904e-05,
+      "loss": 0.8206,
       "step": 15471
     },
     {
-      "epoch": 0.4249265331905194,
+      "epoch": 0.4390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.2876394898144084e-05,
-      "loss": 0.8633,
+      "learning_rate": 1.2435624195300748e-05,
+      "loss": 0.8721,
       "step": 15472
     },
     {
-      "epoch": 0.4249539974183626,
+      "epoch": 0.43907491486946654,
       "grad_norm": 0.0,
-      "learning_rate": 1.2875542962806595e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.2434732784776765e-05,
+      "loss": 0.7533,
       "step": 15473
     },
     {
-      "epoch": 0.4249814616462058,
+      "epoch": 0.4391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.2874691004716697e-05,
-      "loss": 0.9808,
+      "learning_rate": 1.2433841353686486e-05,
+      "loss": 0.8624,
       "step": 15474
     },
     {
-      "epoch": 0.42500892587404904,
+      "epoch": 0.4391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2873839023881134e-05,
-      "loss": 1.0092,
+      "learning_rate": 1.2432949902037439e-05,
+      "loss": 0.8413,
       "step": 15475
     },
     {
-      "epoch": 0.42503639010189226,
+      "epoch": 0.43916004540295117,
       "grad_norm": 0.0,
-      "learning_rate": 1.287298702030664e-05,
-      "loss": 0.8525,
+      "learning_rate": 1.2432058429837153e-05,
+      "loss": 1.0386,
       "step": 15476
     },
     {
-      "epoch": 0.42506385432973554,
+      "epoch": 0.4391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.2872134993999958e-05,
-      "loss": 1.0364,
+      "learning_rate": 1.243116693709316e-05,
+      "loss": 0.9736,
       "step": 15477
     },
     {
-      "epoch": 0.42509131855757876,
+      "epoch": 0.439216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.2871282944967836e-05,
-      "loss": 0.961,
+      "learning_rate": 1.2430275423812995e-05,
+      "loss": 0.9065,
       "step": 15478
     },
     {
-      "epoch": 0.425118782785422,
+      "epoch": 0.43924517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 1.2870430873217007e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.242938389000418e-05,
+      "loss": 0.8297,
       "step": 15479
     },
     {
-      "epoch": 0.4251462470132652,
+      "epoch": 0.4392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.2869578778754219e-05,
-      "loss": 0.7975,
+      "learning_rate": 1.2428492335674252e-05,
+      "loss": 0.8749,
       "step": 15480
     },
     {
-      "epoch": 0.4251737112411085,
+      "epoch": 0.43930192962542564,
       "grad_norm": 0.0,
-      "learning_rate": 1.2868726661586212e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.242760076083074e-05,
+      "loss": 0.985,
       "step": 15481
     },
     {
-      "epoch": 0.4252011754689517,
+      "epoch": 0.43933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.2867874521719728e-05,
-      "loss": 0.8062,
+      "learning_rate": 1.2426709165481177e-05,
+      "loss": 0.8946,
       "step": 15482
     },
     {
-      "epoch": 0.4252286396967949,
+      "epoch": 0.4393586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.2867022359161512e-05,
-      "loss": 0.853,
+      "learning_rate": 1.2425817549633088e-05,
+      "loss": 0.8887,
       "step": 15483
     },
     {
-      "epoch": 0.42525610392463814,
+      "epoch": 0.4393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.2866170173918302e-05,
+      "learning_rate": 1.242492591329401e-05,
       "loss": 0.9318,
       "step": 15484
     },
     {
-      "epoch": 0.4252835681524814,
+      "epoch": 0.43941543700340524,
       "grad_norm": 0.0,
-      "learning_rate": 1.2865317965996843e-05,
-      "loss": 0.9598,
+      "learning_rate": 1.2424034256471478e-05,
+      "loss": 0.9009,
       "step": 15485
     },
     {
-      "epoch": 0.42531103238032464,
+      "epoch": 0.4394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.2864465735403879e-05,
-      "loss": 0.8828,
+      "learning_rate": 1.2423142579173018e-05,
+      "loss": 0.9467,
       "step": 15486
     },
     {
-      "epoch": 0.42533849660816786,
+      "epoch": 0.439472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.2863613482146155e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.2422250881406159e-05,
+      "loss": 0.9215,
       "step": 15487
     },
     {
-      "epoch": 0.4253659608360111,
+      "epoch": 0.4395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.286276120623041e-05,
-      "loss": 0.8225,
+      "learning_rate": 1.2421359163178443e-05,
+      "loss": 1.0098,
       "step": 15488
     },
     {
-      "epoch": 0.4253934250638543,
+      "epoch": 0.4395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.2861908907663386e-05,
-      "loss": 0.883,
+      "learning_rate": 1.2420467424497398e-05,
+      "loss": 0.8036,
       "step": 15489
     },
     {
-      "epoch": 0.4254208892916976,
+      "epoch": 0.4395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.2861056586451834e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.2419575665370554e-05,
+      "loss": 1.0215,
       "step": 15490
     },
     {
-      "epoch": 0.4254483535195408,
+      "epoch": 0.43958569807037456,
       "grad_norm": 0.0,
-      "learning_rate": 1.2860204242602495e-05,
-      "loss": 0.895,
+      "learning_rate": 1.2418683885805442e-05,
+      "loss": 0.9447,
       "step": 15491
     },
     {
-      "epoch": 0.425475817747384,
+      "epoch": 0.4396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.2859351876122109e-05,
-      "loss": 0.9837,
+      "learning_rate": 1.2417792085809604e-05,
+      "loss": 0.8406,
       "step": 15492
     },
     {
-      "epoch": 0.42550328197522724,
+      "epoch": 0.43964245175936434,
       "grad_norm": 0.0,
-      "learning_rate": 1.2858499487017427e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.2416900265390565e-05,
+      "loss": 0.9681,
       "step": 15493
     },
     {
-      "epoch": 0.4255307462030705,
+      "epoch": 0.43967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.2857647075295185e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.241600842455586e-05,
+      "loss": 0.8462,
       "step": 15494
     },
     {
-      "epoch": 0.42555821043091374,
+      "epoch": 0.43969920544835417,
       "grad_norm": 0.0,
-      "learning_rate": 1.2856794640962136e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.2415116563313025e-05,
+      "loss": 0.9457,
       "step": 15495
     },
     {
-      "epoch": 0.42558567465875696,
+      "epoch": 0.43972758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 1.2855942184025021e-05,
-      "loss": 0.892,
+      "learning_rate": 1.2414224681669593e-05,
+      "loss": 0.8834,
       "step": 15496
     },
     {
-      "epoch": 0.4256131388866002,
+      "epoch": 0.43975595913734394,
       "grad_norm": 0.0,
-      "learning_rate": 1.2855089704490582e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.2413332779633093e-05,
+      "loss": 0.889,
       "step": 15497
     },
     {
-      "epoch": 0.42564060311444346,
+      "epoch": 0.4397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.2854237202365569e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.2412440857211066e-05,
+      "loss": 0.9093,
       "step": 15498
     },
     {
-      "epoch": 0.4256680673422867,
+      "epoch": 0.4398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.2853384677656728e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.241154891441104e-05,
+      "loss": 1.0337,
       "step": 15499
     },
     {
-      "epoch": 0.4256955315701299,
+      "epoch": 0.4398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.28525321303708e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.2410656951240552e-05,
+      "loss": 0.9206,
       "step": 15500
     },
     {
-      "epoch": 0.4257229957979731,
+      "epoch": 0.4398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.2851679560514532e-05,
-      "loss": 0.9236,
+      "learning_rate": 1.2409764967707139e-05,
+      "loss": 0.8982,
       "step": 15501
     },
     {
-      "epoch": 0.4257504600258164,
+      "epoch": 0.4398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.2850826968094672e-05,
-      "loss": 0.9685,
+      "learning_rate": 1.2408872963818332e-05,
+      "loss": 0.9913,
       "step": 15502
     },
     {
-      "epoch": 0.4257779242536596,
+      "epoch": 0.43992622020431327,
       "grad_norm": 0.0,
-      "learning_rate": 1.2849974353117967e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.2407980939581665e-05,
+      "loss": 0.87,
       "step": 15503
     },
     {
-      "epoch": 0.42580538848150284,
+      "epoch": 0.4399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.2849121715591157e-05,
-      "loss": 0.9569,
+      "learning_rate": 1.2407088895004676e-05,
+      "loss": 0.9087,
       "step": 15504
     },
     {
-      "epoch": 0.42583285270934607,
+      "epoch": 0.43998297389330304,
       "grad_norm": 0.0,
-      "learning_rate": 1.2848269055520994e-05,
-      "loss": 0.9048,
+      "learning_rate": 1.2406196830094901e-05,
+      "loss": 0.9454,
       "step": 15505
     },
     {
-      "epoch": 0.4258603169371893,
+      "epoch": 0.44001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.2847416372914221e-05,
-      "loss": 0.8941,
+      "learning_rate": 1.2405304744859871e-05,
+      "loss": 0.9495,
       "step": 15506
     },
     {
-      "epoch": 0.42588778116503256,
+      "epoch": 0.44003972758229287,
       "grad_norm": 0.0,
-      "learning_rate": 1.284656366777759e-05,
-      "loss": 1.0406,
+      "learning_rate": 1.240441263930712e-05,
+      "loss": 0.8858,
       "step": 15507
     },
     {
-      "epoch": 0.4259152453928758,
+      "epoch": 0.44006810442678773,
       "grad_norm": 0.0,
-      "learning_rate": 1.2845710940117843e-05,
-      "loss": 0.886,
+      "learning_rate": 1.2403520513444192e-05,
+      "loss": 0.895,
       "step": 15508
     },
     {
-      "epoch": 0.425942709620719,
+      "epoch": 0.44009648127128265,
       "grad_norm": 0.0,
-      "learning_rate": 1.2844858189941729e-05,
-      "loss": 0.9577,
+      "learning_rate": 1.2402628367278618e-05,
+      "loss": 0.8401,
       "step": 15509
     },
     {
-      "epoch": 0.42597017384856223,
+      "epoch": 0.4401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.2844005417255994e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.240173620081793e-05,
+      "loss": 0.8578,
       "step": 15510
     },
     {
-      "epoch": 0.4259976380764055,
+      "epoch": 0.4401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.284315262206739e-05,
-      "loss": 0.91,
+      "learning_rate": 1.2400844014069675e-05,
+      "loss": 0.8592,
       "step": 15511
     },
     {
-      "epoch": 0.4260251023042487,
+      "epoch": 0.44018161180476734,
       "grad_norm": 0.0,
-      "learning_rate": 1.2842299804382656e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.239995180704138e-05,
+      "loss": 0.9375,
       "step": 15512
     },
     {
-      "epoch": 0.42605256653209195,
+      "epoch": 0.4402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.2841446964208546e-05,
-      "loss": 0.9069,
+      "learning_rate": 1.2399059579740582e-05,
+      "loss": 0.7436,
       "step": 15513
     },
     {
-      "epoch": 0.42608003075993517,
+      "epoch": 0.4402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.2840594101551809e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.2398167332174821e-05,
+      "loss": 0.9428,
       "step": 15514
     },
     {
-      "epoch": 0.42610749498777845,
+      "epoch": 0.44026674233825197,
       "grad_norm": 0.0,
-      "learning_rate": 1.2839741216419192e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.2397275064351635e-05,
+      "loss": 0.8955,
       "step": 15515
     },
     {
-      "epoch": 0.42613495921562167,
+      "epoch": 0.4402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.2838888308817439e-05,
-      "loss": 0.8314,
+      "learning_rate": 1.2396382776278557e-05,
+      "loss": 0.9562,
       "step": 15516
     },
     {
-      "epoch": 0.4261624234434649,
+      "epoch": 0.44032349602724175,
       "grad_norm": 0.0,
-      "learning_rate": 1.2838035378753304e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.2395490467963124e-05,
+      "loss": 0.9843,
       "step": 15517
     },
     {
-      "epoch": 0.4261898876713081,
+      "epoch": 0.44035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.2837182426233534e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.239459813941288e-05,
+      "loss": 0.8891,
       "step": 15518
     },
     {
-      "epoch": 0.42621735189915133,
+      "epoch": 0.4403802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.2836329451264877e-05,
-      "loss": 0.8428,
+      "learning_rate": 1.2393705790635356e-05,
+      "loss": 0.8698,
       "step": 15519
     },
     {
-      "epoch": 0.4262448161269946,
+      "epoch": 0.44040862656072643,
       "grad_norm": 0.0,
-      "learning_rate": 1.2835476453854084e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.2392813421638088e-05,
+      "loss": 0.8691,
       "step": 15520
     },
     {
-      "epoch": 0.42627228035483783,
+      "epoch": 0.44043700340522135,
       "grad_norm": 0.0,
-      "learning_rate": 1.2834623434007902e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.2391921032428622e-05,
+      "loss": 0.9096,
       "step": 15521
     },
     {
-      "epoch": 0.42629974458268105,
+      "epoch": 0.4404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.283377039173308e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.2391028623014487e-05,
+      "loss": 0.9628,
       "step": 15522
     },
     {
-      "epoch": 0.42632720881052427,
+      "epoch": 0.4404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.2832917327036372e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.239013619340323e-05,
+      "loss": 0.9795,
       "step": 15523
     },
     {
-      "epoch": 0.42635467303836755,
+      "epoch": 0.44052213393870604,
       "grad_norm": 0.0,
-      "learning_rate": 1.2832064239924522e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.2389243743602383e-05,
+      "loss": 0.9271,
       "step": 15524
     },
     {
-      "epoch": 0.42638213726621077,
+      "epoch": 0.4405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.2831211130404282e-05,
-      "loss": 0.9,
+      "learning_rate": 1.2388351273619488e-05,
+      "loss": 1.0026,
       "step": 15525
     },
     {
-      "epoch": 0.426409601494054,
+      "epoch": 0.4405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.2830357998482405e-05,
-      "loss": 0.8275,
+      "learning_rate": 1.2387458783462082e-05,
+      "loss": 0.8364,
       "step": 15526
     },
     {
-      "epoch": 0.4264370657218972,
+      "epoch": 0.4406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.2829504844165641e-05,
-      "loss": 1.0213,
+      "learning_rate": 1.2386566273137703e-05,
+      "loss": 0.8928,
       "step": 15527
     },
     {
-      "epoch": 0.4264645299497405,
+      "epoch": 0.4406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.2828651667460737e-05,
-      "loss": 0.7364,
+      "learning_rate": 1.2385673742653894e-05,
+      "loss": 1.0088,
       "step": 15528
     },
     {
-      "epoch": 0.4264919941775837,
+      "epoch": 0.4406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.2827798468374445e-05,
-      "loss": 0.8298,
+      "learning_rate": 1.2384781192018192e-05,
+      "loss": 0.8534,
       "step": 15529
     },
     {
-      "epoch": 0.42651945840542693,
+      "epoch": 0.44069239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.2826945246913511e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.2383888621238132e-05,
+      "loss": 0.9303,
       "step": 15530
     },
     {
-      "epoch": 0.42654692263327015,
+      "epoch": 0.4407207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.28260920030847e-05,
-      "loss": 0.902,
+      "learning_rate": 1.2382996030321258e-05,
+      "loss": 0.8665,
       "step": 15531
     },
     {
-      "epoch": 0.4265743868611134,
+      "epoch": 0.44074914869466514,
       "grad_norm": 0.0,
-      "learning_rate": 1.282523873689475e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.2382103419275113e-05,
+      "loss": 0.9482,
       "step": 15532
     },
     {
-      "epoch": 0.42660185108895665,
+      "epoch": 0.44077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.2824385448350413e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.2381210788107232e-05,
+      "loss": 0.8828,
       "step": 15533
     },
     {
-      "epoch": 0.4266293153167999,
+      "epoch": 0.4408059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.2823532137458449e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.2380318136825155e-05,
+      "loss": 0.9261,
       "step": 15534
     },
     {
-      "epoch": 0.4266567795446431,
+      "epoch": 0.44083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.2822678804225608e-05,
-      "loss": 0.9415,
+      "learning_rate": 1.2379425465436426e-05,
+      "loss": 0.8964,
       "step": 15535
     },
     {
-      "epoch": 0.4266842437724863,
+      "epoch": 0.44086265607264474,
       "grad_norm": 0.0,
-      "learning_rate": 1.2821825448658634e-05,
-      "loss": 0.8481,
+      "learning_rate": 1.2378532773948582e-05,
+      "loss": 0.7365,
       "step": 15536
     },
     {
-      "epoch": 0.4267117080003296,
+      "epoch": 0.4408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.2820972070764283e-05,
-      "loss": 0.8862,
+      "learning_rate": 1.2377640062369163e-05,
+      "loss": 0.8463,
       "step": 15537
     },
     {
-      "epoch": 0.4267391722281728,
+      "epoch": 0.4409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.2820118670549312e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.2376747330705711e-05,
+      "loss": 0.9538,
       "step": 15538
     },
     {
-      "epoch": 0.42676663645601604,
+      "epoch": 0.4409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.2819265248020466e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.2375854578965771e-05,
+      "loss": 0.9277,
       "step": 15539
     },
     {
-      "epoch": 0.42679410068385926,
+      "epoch": 0.4409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.2818411803184505e-05,
-      "loss": 0.937,
+      "learning_rate": 1.237496180715688e-05,
+      "loss": 0.9209,
       "step": 15540
     },
     {
-      "epoch": 0.42682156491170253,
+      "epoch": 0.4410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.2817558336048175e-05,
-      "loss": 0.8511,
+      "learning_rate": 1.2374069015286579e-05,
+      "loss": 0.9492,
       "step": 15541
     },
     {
-      "epoch": 0.42684902913954575,
+      "epoch": 0.44103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.2816704846618233e-05,
-      "loss": 0.9106,
+      "learning_rate": 1.2373176203362412e-05,
+      "loss": 0.8852,
       "step": 15542
     },
     {
-      "epoch": 0.426876493367389,
+      "epoch": 0.441061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.2815851334901435e-05,
-      "loss": 0.8755,
+      "learning_rate": 1.2372283371391918e-05,
+      "loss": 0.9388,
       "step": 15543
     },
     {
-      "epoch": 0.4269039575952322,
+      "epoch": 0.44108967082860384,
       "grad_norm": 0.0,
-      "learning_rate": 1.2814997800904524e-05,
-      "loss": 0.8875,
+      "learning_rate": 1.2371390519382635e-05,
+      "loss": 0.8954,
       "step": 15544
     },
     {
-      "epoch": 0.4269314218230754,
+      "epoch": 0.44111804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 1.281414424463426e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.2370497647342114e-05,
+      "loss": 0.9859,
       "step": 15545
     },
     {
-      "epoch": 0.4269588860509187,
+      "epoch": 0.4411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.2813290666097402e-05,
-      "loss": 0.9288,
+      "learning_rate": 1.2369604755277893e-05,
+      "loss": 0.9163,
       "step": 15546
     },
     {
-      "epoch": 0.4269863502787619,
+      "epoch": 0.44117480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 1.2812437065300698e-05,
-      "loss": 0.9938,
+      "learning_rate": 1.2368711843197513e-05,
+      "loss": 0.8846,
       "step": 15547
     },
     {
-      "epoch": 0.42701381450660514,
+      "epoch": 0.44120317820658345,
       "grad_norm": 0.0,
-      "learning_rate": 1.2811583442250899e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.2367818911108518e-05,
+      "loss": 0.9218,
       "step": 15548
     },
     {
-      "epoch": 0.42704127873444836,
+      "epoch": 0.4412315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.2810729796954764e-05,
-      "loss": 0.8976,
+      "learning_rate": 1.2366925959018452e-05,
+      "loss": 0.823,
       "step": 15549
     },
     {
-      "epoch": 0.42706874296229164,
+      "epoch": 0.4412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.2809876129419045e-05,
-      "loss": 0.8926,
+      "learning_rate": 1.2366032986934856e-05,
+      "loss": 1.0009,
       "step": 15550
     },
     {
-      "epoch": 0.42709620719013486,
+      "epoch": 0.4412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.28090224396505e-05,
-      "loss": 0.9176,
+      "learning_rate": 1.236513999486527e-05,
+      "loss": 0.721,
       "step": 15551
     },
     {
-      "epoch": 0.4271236714179781,
+      "epoch": 0.441316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2808168727655882e-05,
-      "loss": 0.9513,
+      "learning_rate": 1.236424698281724e-05,
+      "loss": 0.9349,
       "step": 15552
     },
     {
-      "epoch": 0.4271511356458213,
+      "epoch": 0.4413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.2807314993441944e-05,
-      "loss": 0.7719,
+      "learning_rate": 1.2363353950798311e-05,
+      "loss": 0.8958,
       "step": 15553
     },
     {
-      "epoch": 0.4271785998736646,
+      "epoch": 0.44137343927355277,
       "grad_norm": 0.0,
-      "learning_rate": 1.280646123701544e-05,
-      "loss": 0.8917,
+      "learning_rate": 1.2362460898816025e-05,
+      "loss": 1.0089,
       "step": 15554
     },
     {
-      "epoch": 0.4272060641015078,
+      "epoch": 0.4414018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.2805607458383132e-05,
-      "loss": 0.8659,
+      "learning_rate": 1.2361567826877925e-05,
+      "loss": 0.9124,
       "step": 15555
     },
     {
-      "epoch": 0.427233528329351,
+      "epoch": 0.44143019296254254,
       "grad_norm": 0.0,
-      "learning_rate": 1.2804753657551768e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.2360674734991558e-05,
+      "loss": 0.8175,
       "step": 15556
     },
     {
-      "epoch": 0.42726099255719424,
+      "epoch": 0.44145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.2803899834528103e-05,
-      "loss": 0.8802,
+      "learning_rate": 1.2359781623164465e-05,
+      "loss": 0.9625,
       "step": 15557
     },
     {
-      "epoch": 0.42728845678503746,
+      "epoch": 0.4414869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.2803045989318903e-05,
-      "loss": 0.8833,
+      "learning_rate": 1.2358888491404186e-05,
+      "loss": 0.9505,
       "step": 15558
     },
     {
-      "epoch": 0.42731592101288074,
+      "epoch": 0.44151532349602723,
       "grad_norm": 0.0,
-      "learning_rate": 1.2802192121930917e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.2357995339718273e-05,
+      "loss": 0.9238,
       "step": 15559
     },
     {
-      "epoch": 0.42734338524072396,
+      "epoch": 0.44154370034052215,
       "grad_norm": 0.0,
-      "learning_rate": 1.2801338232370897e-05,
-      "loss": 0.9769,
+      "learning_rate": 1.2357102168114267e-05,
+      "loss": 1.019,
       "step": 15560
     },
     {
-      "epoch": 0.4273708494685672,
+      "epoch": 0.441572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.2800484320645605e-05,
-      "loss": 0.8415,
+      "learning_rate": 1.235620897659971e-05,
+      "loss": 0.8827,
       "step": 15561
     },
     {
-      "epoch": 0.4273983136964104,
+      "epoch": 0.4416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.2799630386761797e-05,
-      "loss": 1.008,
+      "learning_rate": 1.2355315765182154e-05,
+      "loss": 0.9444,
       "step": 15562
     },
     {
-      "epoch": 0.4274257779242537,
+      "epoch": 0.4416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.2798776430726231e-05,
-      "loss": 0.7906,
+      "learning_rate": 1.235442253386914e-05,
+      "loss": 0.8328,
       "step": 15563
     },
     {
-      "epoch": 0.4274532421520969,
+      "epoch": 0.4416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.2797922452545662e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.235352928266821e-05,
+      "loss": 0.953,
       "step": 15564
     },
     {
-      "epoch": 0.4274807063799401,
+      "epoch": 0.4416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.2797068452226842e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.2352636011586915e-05,
+      "loss": 0.9109,
       "step": 15565
     },
     {
-      "epoch": 0.42750817060778334,
+      "epoch": 0.4417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.2796214429776535e-05,
-      "loss": 0.8366,
+      "learning_rate": 1.2351742720632798e-05,
+      "loss": 0.8955,
       "step": 15566
     },
     {
-      "epoch": 0.4275356348356266,
+      "epoch": 0.4417423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.2795360385201499e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.2350849409813403e-05,
+      "loss": 0.8834,
       "step": 15567
     },
     {
-      "epoch": 0.42756309906346984,
+      "epoch": 0.44177071509648125,
       "grad_norm": 0.0,
-      "learning_rate": 1.2794506318508487e-05,
-      "loss": 0.9212,
+      "learning_rate": 1.2349956079136276e-05,
+      "loss": 0.9488,
       "step": 15568
     },
     {
-      "epoch": 0.42759056329131306,
+      "epoch": 0.44179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.2793652229704256e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.2349062728608967e-05,
+      "loss": 0.8614,
       "step": 15569
     },
     {
-      "epoch": 0.4276180275191563,
+      "epoch": 0.4418274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 1.2792798118795568e-05,
-      "loss": 0.9064,
+      "learning_rate": 1.2348169358239019e-05,
+      "loss": 0.9249,
       "step": 15570
     },
     {
-      "epoch": 0.4276454917469995,
+      "epoch": 0.44185584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.2791943985789182e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.2347275968033978e-05,
+      "loss": 0.8873,
       "step": 15571
     },
     {
-      "epoch": 0.4276729559748428,
+      "epoch": 0.44188422247446085,
       "grad_norm": 0.0,
-      "learning_rate": 1.2791089830691849e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.2346382558001392e-05,
+      "loss": 0.9171,
       "step": 15572
     },
     {
-      "epoch": 0.427700420202686,
+      "epoch": 0.4419125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.2790235653510336e-05,
-      "loss": 0.8373,
+      "learning_rate": 1.234548912814881e-05,
+      "loss": 0.9082,
       "step": 15573
     },
     {
-      "epoch": 0.4277278844305292,
+      "epoch": 0.4419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.2789381454251397e-05,
-      "loss": 0.9063,
+      "learning_rate": 1.234459567848377e-05,
+      "loss": 0.9305,
       "step": 15574
     },
     {
-      "epoch": 0.42775534865837245,
+      "epoch": 0.44196935300794554,
       "grad_norm": 0.0,
-      "learning_rate": 1.278852723292179e-05,
-      "loss": 0.9401,
+      "learning_rate": 1.234370220901383e-05,
+      "loss": 0.8777,
       "step": 15575
     },
     {
-      "epoch": 0.4277828128862157,
+      "epoch": 0.4419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.2787672989528276e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.2342808719746528e-05,
+      "loss": 0.8447,
       "step": 15576
     },
     {
-      "epoch": 0.42781027711405895,
+      "epoch": 0.4420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.2786818724077614e-05,
-      "loss": 0.8509,
+      "learning_rate": 1.2341915210689419e-05,
+      "loss": 0.8388,
       "step": 15577
     },
     {
-      "epoch": 0.42783774134190217,
+      "epoch": 0.4420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.2785964436576562e-05,
-      "loss": 0.8154,
+      "learning_rate": 1.2341021681850045e-05,
+      "loss": 0.8862,
       "step": 15578
     },
     {
-      "epoch": 0.4278652055697454,
+      "epoch": 0.4420828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.2785110127031884e-05,
-      "loss": 0.902,
+      "learning_rate": 1.2340128133235956e-05,
+      "loss": 0.9109,
       "step": 15579
     },
     {
-      "epoch": 0.42789266979758867,
+      "epoch": 0.44211123723041995,
       "grad_norm": 0.0,
-      "learning_rate": 1.278425579545033e-05,
-      "loss": 0.7917,
+      "learning_rate": 1.23392345648547e-05,
+      "loss": 0.8468,
       "step": 15580
     },
     {
-      "epoch": 0.4279201340254319,
+      "epoch": 0.44213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.2783401441838668e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.2338340976713823e-05,
+      "loss": 0.9785,
       "step": 15581
     },
     {
-      "epoch": 0.4279475982532751,
+      "epoch": 0.4421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.2782547066203657e-05,
-      "loss": 1.0198,
+      "learning_rate": 1.2337447368820876e-05,
+      "loss": 0.847,
       "step": 15582
     },
     {
-      "epoch": 0.42797506248111833,
+      "epoch": 0.44219636776390464,
       "grad_norm": 0.0,
-      "learning_rate": 1.2781692668552055e-05,
-      "loss": 0.8856,
+      "learning_rate": 1.2336553741183408e-05,
+      "loss": 0.9478,
       "step": 15583
     },
     {
-      "epoch": 0.42800252670896155,
+      "epoch": 0.44222474460839956,
       "grad_norm": 0.0,
-      "learning_rate": 1.2780838248890622e-05,
-      "loss": 0.8777,
+      "learning_rate": 1.2335660093808962e-05,
+      "loss": 0.9481,
       "step": 15584
     },
     {
-      "epoch": 0.4280299909368048,
+      "epoch": 0.4422531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.277998380722612e-05,
-      "loss": 0.9313,
+      "learning_rate": 1.233476642670509e-05,
+      "loss": 0.9306,
       "step": 15585
     },
     {
-      "epoch": 0.42805745516464805,
+      "epoch": 0.44228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.2779129343565309e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.2333872739879343e-05,
+      "loss": 0.892,
       "step": 15586
     },
     {
-      "epoch": 0.42808491939249127,
+      "epoch": 0.44230987514188425,
       "grad_norm": 0.0,
-      "learning_rate": 1.2778274857914949e-05,
-      "loss": 0.814,
+      "learning_rate": 1.2332979033339267e-05,
+      "loss": 1.0011,
       "step": 15587
     },
     {
-      "epoch": 0.4281123836203345,
+      "epoch": 0.4423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.2777420350281806e-05,
-      "loss": 0.9015,
+      "learning_rate": 1.2332085307092412e-05,
+      "loss": 0.9309,
       "step": 15588
     },
     {
-      "epoch": 0.42813984784817777,
+      "epoch": 0.442366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.2776565820672635e-05,
-      "loss": 0.9684,
+      "learning_rate": 1.2331191561146329e-05,
+      "loss": 0.8353,
       "step": 15589
     },
     {
-      "epoch": 0.428167312076021,
+      "epoch": 0.4423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.2775711269094199e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.2330297795508566e-05,
+      "loss": 0.8192,
       "step": 15590
     },
     {
-      "epoch": 0.4281947763038642,
+      "epoch": 0.4424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.2774856695553265e-05,
-      "loss": 0.9696,
+      "learning_rate": 1.2329404010186673e-05,
+      "loss": 1.0688,
       "step": 15591
     },
     {
-      "epoch": 0.42822224053170743,
+      "epoch": 0.4424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.2774002100056582e-05,
-      "loss": 1.0036,
+      "learning_rate": 1.2328510205188195e-05,
+      "loss": 0.9189,
       "step": 15592
     },
     {
-      "epoch": 0.4282497047595507,
+      "epoch": 0.44248013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.2773147482610923e-05,
-      "loss": 0.9561,
+      "learning_rate": 1.232761638052069e-05,
+      "loss": 0.9551,
       "step": 15593
     },
     {
-      "epoch": 0.42827716898739393,
+      "epoch": 0.4425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.2772292843223048e-05,
-      "loss": 1.009,
+      "learning_rate": 1.2326722536191707e-05,
+      "loss": 0.8742,
       "step": 15594
     },
     {
-      "epoch": 0.42830463321523715,
+      "epoch": 0.44253688989784334,
       "grad_norm": 0.0,
-      "learning_rate": 1.2771438181899717e-05,
-      "loss": 1.0023,
+      "learning_rate": 1.232582867220879e-05,
+      "loss": 0.9067,
       "step": 15595
     },
     {
-      "epoch": 0.4283320974430804,
+      "epoch": 0.44256526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.2770583498647698e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.23249347885795e-05,
+      "loss": 0.934,
       "step": 15596
     },
     {
-      "epoch": 0.4283595616709236,
+      "epoch": 0.4425936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 1.2769728793473743e-05,
-      "loss": 0.8363,
+      "learning_rate": 1.2324040885311376e-05,
+      "loss": 0.9295,
       "step": 15597
     },
     {
-      "epoch": 0.42838702589876687,
+      "epoch": 0.44262202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.2768874066384623e-05,
-      "loss": 0.8149,
+      "learning_rate": 1.2323146962411974e-05,
+      "loss": 0.8794,
       "step": 15598
     },
     {
-      "epoch": 0.4284144901266101,
+      "epoch": 0.44265039727582295,
       "grad_norm": 0.0,
-      "learning_rate": 1.2768019317387102e-05,
-      "loss": 0.9997,
+      "learning_rate": 1.2322253019888846e-05,
+      "loss": 0.8144,
       "step": 15599
     },
     {
-      "epoch": 0.4284419543544533,
+      "epoch": 0.4426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.2767164546487936e-05,
-      "loss": 0.8782,
+      "learning_rate": 1.2321359057749542e-05,
+      "loss": 0.8466,
       "step": 15600
     },
     {
-      "epoch": 0.42846941858229654,
+      "epoch": 0.4427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.276630975369389e-05,
-      "loss": 0.9996,
+      "learning_rate": 1.2320465076001618e-05,
+      "loss": 0.9247,
       "step": 15601
     },
     {
-      "epoch": 0.4284968828101398,
+      "epoch": 0.4427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.2765454939011734e-05,
-      "loss": 0.8893,
+      "learning_rate": 1.2319571074652614e-05,
+      "loss": 0.8722,
       "step": 15602
     },
     {
-      "epoch": 0.42852434703798303,
+      "epoch": 0.4427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.2764600102448228e-05,
-      "loss": 0.8936,
+      "learning_rate": 1.2318677053710095e-05,
+      "loss": 1.0721,
       "step": 15603
     },
     {
-      "epoch": 0.42855181126582625,
+      "epoch": 0.4427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.2763745244010134e-05,
-      "loss": 0.8794,
+      "learning_rate": 1.2317783013181606e-05,
+      "loss": 0.9601,
       "step": 15604
     },
     {
-      "epoch": 0.4285792754936695,
+      "epoch": 0.4428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.2762890363704217e-05,
-      "loss": 0.9611,
+      "learning_rate": 1.2316888953074695e-05,
+      "loss": 0.8698,
       "step": 15605
     },
     {
-      "epoch": 0.42860673972151275,
+      "epoch": 0.4428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.2762035461537238e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.2315994873396923e-05,
+      "loss": 0.8674,
       "step": 15606
     },
     {
-      "epoch": 0.428634203949356,
+      "epoch": 0.44287741203178205,
       "grad_norm": 0.0,
-      "learning_rate": 1.276118053751597e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.2315100774155837e-05,
+      "loss": 0.9676,
       "step": 15607
     },
     {
-      "epoch": 0.4286616681771992,
+      "epoch": 0.44290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.2760325591647168e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.231420665535899e-05,
+      "loss": 0.9512,
       "step": 15608
     },
     {
-      "epoch": 0.4286891324050424,
+      "epoch": 0.4429341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 1.27594706239376e-05,
-      "loss": 0.9501,
+      "learning_rate": 1.2313312517013936e-05,
+      "loss": 0.9423,
       "step": 15609
     },
     {
-      "epoch": 0.4287165966328857,
+      "epoch": 0.44296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.2758615634394031e-05,
-      "loss": 0.9446,
+      "learning_rate": 1.231241835912823e-05,
+      "loss": 0.8766,
       "step": 15610
     },
     {
-      "epoch": 0.4287440608607289,
+      "epoch": 0.44299091940976165,
       "grad_norm": 0.0,
-      "learning_rate": 1.2757760623023235e-05,
-      "loss": 0.8157,
+      "learning_rate": 1.2311524181709417e-05,
+      "loss": 0.9698,
       "step": 15611
     },
     {
-      "epoch": 0.42877152508857214,
+      "epoch": 0.4430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.2756905589831959e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.2310629984765056e-05,
+      "loss": 0.9102,
       "step": 15612
     },
     {
-      "epoch": 0.42879898931641536,
+      "epoch": 0.4430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.275605053482698e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.2309735768302704e-05,
+      "loss": 0.9297,
       "step": 15613
     },
     {
-      "epoch": 0.4288264535442586,
+      "epoch": 0.4430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.2755195458015064e-05,
-      "loss": 0.846,
+      "learning_rate": 1.2308841532329908e-05,
+      "loss": 0.8924,
       "step": 15614
     },
     {
-      "epoch": 0.42885391777210186,
+      "epoch": 0.4431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.2754340359402975e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.230794727685422e-05,
+      "loss": 0.8589,
       "step": 15615
     },
     {
-      "epoch": 0.4288813819999451,
+      "epoch": 0.4431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.2753485238997476e-05,
-      "loss": 0.8664,
+      "learning_rate": 1.2307053001883202e-05,
+      "loss": 1.0473,
       "step": 15616
     },
     {
-      "epoch": 0.4289088462277883,
+      "epoch": 0.443161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.2752630096805336e-05,
-      "loss": 0.8097,
+      "learning_rate": 1.2306158707424402e-05,
+      "loss": 0.9079,
       "step": 15617
     },
     {
-      "epoch": 0.4289363104556315,
+      "epoch": 0.4431895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.2751774932833318e-05,
-      "loss": 0.956,
+      "learning_rate": 1.2305264393485377e-05,
+      "loss": 0.9196,
       "step": 15618
     },
     {
-      "epoch": 0.4289637746834748,
+      "epoch": 0.44321793416572075,
       "grad_norm": 0.0,
-      "learning_rate": 1.2750919747088195e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.230437006007368e-05,
+      "loss": 0.8556,
       "step": 15619
     },
     {
-      "epoch": 0.428991238911318,
+      "epoch": 0.44324631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.2750064539576728e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.2303475707196866e-05,
+      "loss": 0.9369,
       "step": 15620
     },
     {
-      "epoch": 0.42901870313916124,
+      "epoch": 0.4432746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.2749209310305683e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.2302581334862486e-05,
+      "loss": 1.0637,
       "step": 15621
     },
     {
-      "epoch": 0.42904616736700446,
+      "epoch": 0.44330306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.2748354059281829e-05,
-      "loss": 1.0453,
+      "learning_rate": 1.2301686943078099e-05,
+      "loss": 0.7954,
       "step": 15622
     },
     {
-      "epoch": 0.42907363159484774,
+      "epoch": 0.44333144154370036,
       "grad_norm": 0.0,
-      "learning_rate": 1.2747498786511937e-05,
-      "loss": 1.0142,
+      "learning_rate": 1.230079253185126e-05,
+      "loss": 0.9182,
       "step": 15623
     },
     {
-      "epoch": 0.42910109582269096,
+      "epoch": 0.4433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.2746643492002769e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.2299898101189524e-05,
+      "loss": 0.849,
       "step": 15624
     },
     {
-      "epoch": 0.4291285600505342,
+      "epoch": 0.44338819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.274578817576109e-05,
-      "loss": 0.8412,
+      "learning_rate": 1.2299003651100442e-05,
+      "loss": 0.8501,
       "step": 15625
     },
     {
-      "epoch": 0.4291560242783774,
+      "epoch": 0.443416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.2744932837793673e-05,
-      "loss": 0.8396,
+      "learning_rate": 1.2298109181591578e-05,
+      "loss": 0.8031,
       "step": 15626
     },
     {
-      "epoch": 0.4291834885062206,
+      "epoch": 0.4434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.2744077478107286e-05,
-      "loss": 0.9655,
+      "learning_rate": 1.2297214692670481e-05,
+      "loss": 0.9244,
       "step": 15627
     },
     {
-      "epoch": 0.4292109527340639,
+      "epoch": 0.4434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.2743222096708695e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.2296320184344704e-05,
+      "loss": 0.8854,
       "step": 15628
     },
     {
-      "epoch": 0.4292384169619071,
+      "epoch": 0.4435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.2742366693604666e-05,
-      "loss": 0.9382,
+      "learning_rate": 1.2295425656621812e-05,
+      "loss": 0.9232,
       "step": 15629
     },
     {
-      "epoch": 0.42926588118975034,
+      "epoch": 0.4435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2741511268801969e-05,
-      "loss": 1.0281,
+      "learning_rate": 1.2294531109509351e-05,
+      "loss": 0.8321,
       "step": 15630
     },
     {
-      "epoch": 0.42929334541759356,
+      "epoch": 0.44355845629965945,
       "grad_norm": 0.0,
-      "learning_rate": 1.2740655822307378e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.2293636543014888e-05,
+      "loss": 0.8937,
       "step": 15631
     },
     {
-      "epoch": 0.42932080964543684,
+      "epoch": 0.44358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.2739800354127651e-05,
-      "loss": 0.8085,
+      "learning_rate": 1.229274195714597e-05,
+      "loss": 0.9508,
       "step": 15632
     },
     {
-      "epoch": 0.42934827387328006,
+      "epoch": 0.4436152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.2738944864269564e-05,
-      "loss": 0.9331,
+      "learning_rate": 1.229184735191016e-05,
+      "loss": 0.9843,
       "step": 15633
     },
     {
-      "epoch": 0.4293757381011233,
+      "epoch": 0.44364358683314414,
       "grad_norm": 0.0,
-      "learning_rate": 1.2738089352739884e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.2290952727315013e-05,
+      "loss": 0.8568,
       "step": 15634
     },
     {
-      "epoch": 0.4294032023289665,
+      "epoch": 0.44367196367763906,
       "grad_norm": 0.0,
-      "learning_rate": 1.2737233819545382e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.2290058083368084e-05,
+      "loss": 0.8783,
       "step": 15635
     },
     {
-      "epoch": 0.4294306665568098,
+      "epoch": 0.4437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 1.2736378264692826e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.228916342007693e-05,
+      "loss": 0.8949,
       "step": 15636
     },
     {
-      "epoch": 0.429458130784653,
+      "epoch": 0.44372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.2735522688188981e-05,
-      "loss": 0.8782,
+      "learning_rate": 1.2288268737449111e-05,
+      "loss": 1.0045,
       "step": 15637
     },
     {
-      "epoch": 0.4294855950124962,
+      "epoch": 0.44375709421112375,
       "grad_norm": 0.0,
-      "learning_rate": 1.2734667090040624e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.2287374035492184e-05,
+      "loss": 0.9272,
       "step": 15638
     },
     {
-      "epoch": 0.42951305924033945,
+      "epoch": 0.4437854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.2733811470254525e-05,
-      "loss": 0.8153,
+      "learning_rate": 1.2286479314213703e-05,
+      "loss": 0.9014,
       "step": 15639
     },
     {
-      "epoch": 0.42954052346818267,
+      "epoch": 0.4438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.2732955828837445e-05,
-      "loss": 1.0307,
+      "learning_rate": 1.2285584573621231e-05,
+      "loss": 0.9289,
       "step": 15640
     },
     {
-      "epoch": 0.42956798769602594,
+      "epoch": 0.4438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2732100165796162e-05,
-      "loss": 0.83,
+      "learning_rate": 1.2284689813722326e-05,
+      "loss": 0.9151,
       "step": 15641
     },
     {
-      "epoch": 0.42959545192386916,
+      "epoch": 0.4438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.2731244481137443e-05,
-      "loss": 0.8508,
+      "learning_rate": 1.228379503452454e-05,
+      "loss": 0.8524,
       "step": 15642
     },
     {
-      "epoch": 0.4296229161517124,
+      "epoch": 0.44389897843359816,
       "grad_norm": 0.0,
-      "learning_rate": 1.273038877486806e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.2282900236035435e-05,
+      "loss": 0.9347,
       "step": 15643
     },
     {
-      "epoch": 0.4296503803795556,
+      "epoch": 0.44392735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.2729533046994786e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.228200541826257e-05,
+      "loss": 0.8495,
       "step": 15644
     },
     {
-      "epoch": 0.4296778446073989,
+      "epoch": 0.443955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.2728677297524388e-05,
-      "loss": 0.8777,
+      "learning_rate": 1.22811105812135e-05,
+      "loss": 0.9032,
       "step": 15645
     },
     {
-      "epoch": 0.4297053088352421,
+      "epoch": 0.44398410896708285,
       "grad_norm": 0.0,
-      "learning_rate": 1.2727821526463634e-05,
-      "loss": 0.9151,
+      "learning_rate": 1.2280215724895786e-05,
+      "loss": 0.7262,
       "step": 15646
     },
     {
-      "epoch": 0.4297327730630853,
+      "epoch": 0.44401248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.2726965733819307e-05,
-      "loss": 0.854,
+      "learning_rate": 1.2279320849316991e-05,
+      "loss": 1.0054,
       "step": 15647
     },
     {
-      "epoch": 0.42976023729092855,
+      "epoch": 0.4440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.2726109919598165e-05,
-      "loss": 0.8465,
+      "learning_rate": 1.227842595448467e-05,
+      "loss": 0.9454,
       "step": 15648
     },
     {
-      "epoch": 0.4297877015187718,
+      "epoch": 0.44406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.2725254083806989e-05,
-      "loss": 0.9202,
+      "learning_rate": 1.2277531040406379e-05,
+      "loss": 0.7999,
       "step": 15649
     },
     {
-      "epoch": 0.42981516574661505,
+      "epoch": 0.44409761634506245,
       "grad_norm": 0.0,
-      "learning_rate": 1.2724398226452543e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.2276636107089683e-05,
+      "loss": 0.8718,
       "step": 15650
     },
     {
-      "epoch": 0.42984262997445827,
+      "epoch": 0.4441259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.2723542347541607e-05,
-      "loss": 0.962,
+      "learning_rate": 1.2275741154542143e-05,
+      "loss": 0.9178,
       "step": 15651
     },
     {
-      "epoch": 0.4298700942023015,
+      "epoch": 0.4441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.2722686447080948e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.227484618277131e-05,
+      "loss": 0.8962,
       "step": 15652
     },
     {
-      "epoch": 0.4298975584301447,
+      "epoch": 0.4441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.2721830525077336e-05,
-      "loss": 0.866,
+      "learning_rate": 1.2273951191784748e-05,
+      "loss": 0.8963,
       "step": 15653
     },
     {
-      "epoch": 0.429925022657988,
+      "epoch": 0.444211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.2720974581537546e-05,
-      "loss": 0.89,
+      "learning_rate": 1.2273056181590022e-05,
+      "loss": 0.9153,
       "step": 15654
     },
     {
-      "epoch": 0.4299524868858312,
+      "epoch": 0.4442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.2720118616468356e-05,
-      "loss": 0.8627,
+      "learning_rate": 1.2272161152194688e-05,
+      "loss": 0.9135,
       "step": 15655
     },
     {
-      "epoch": 0.42997995111367443,
+      "epoch": 0.4442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.2719262629876532e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.2271266103606305e-05,
+      "loss": 1.0407,
       "step": 15656
     },
     {
-      "epoch": 0.43000741534151765,
+      "epoch": 0.4442962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.2718406621768846e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.2270371035832436e-05,
+      "loss": 0.8724,
       "step": 15657
     },
     {
-      "epoch": 0.43003487956936093,
+      "epoch": 0.44432463110102155,
       "grad_norm": 0.0,
-      "learning_rate": 1.271755059215208e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.2269475948880643e-05,
+      "loss": 0.9755,
       "step": 15658
     },
     {
-      "epoch": 0.43006234379720415,
+      "epoch": 0.44435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2716694541032996e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.226858084275848e-05,
+      "loss": 0.8762,
       "step": 15659
     },
     {
-      "epoch": 0.43008980802504737,
+      "epoch": 0.4443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.2715838468418373e-05,
-      "loss": 0.8013,
+      "learning_rate": 1.2267685717473513e-05,
+      "loss": 0.9146,
       "step": 15660
     },
     {
-      "epoch": 0.4301172722528906,
+      "epoch": 0.44440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.2714982374314986e-05,
-      "loss": 0.9699,
+      "learning_rate": 1.2266790573033307e-05,
+      "loss": 0.8976,
       "step": 15661
     },
     {
-      "epoch": 0.43014473648073387,
+      "epoch": 0.44443813847900115,
       "grad_norm": 0.0,
-      "learning_rate": 1.2714126258729604e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.2265895409445413e-05,
+      "loss": 0.9991,
       "step": 15662
     },
     {
-      "epoch": 0.4301722007085771,
+      "epoch": 0.444466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.2713270121669006e-05,
-      "loss": 0.8508,
+      "learning_rate": 1.2265000226717405e-05,
+      "loss": 0.8445,
       "step": 15663
     },
     {
-      "epoch": 0.4301996649364203,
+      "epoch": 0.44449489216799093,
       "grad_norm": 0.0,
-      "learning_rate": 1.2712413963139962e-05,
-      "loss": 0.8482,
+      "learning_rate": 1.2264105024856836e-05,
+      "loss": 0.8844,
       "step": 15664
     },
     {
-      "epoch": 0.43022712916426353,
+      "epoch": 0.4445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.271155778314925e-05,
-      "loss": 1.0069,
+      "learning_rate": 1.2263209803871271e-05,
+      "loss": 0.8916,
       "step": 15665
     },
     {
-      "epoch": 0.43025459339210675,
+      "epoch": 0.4445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.271070158170364e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.226231456376827e-05,
+      "loss": 0.8545,
       "step": 15666
     },
     {
-      "epoch": 0.43028205761995003,
+      "epoch": 0.4445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.2709845358809912e-05,
-      "loss": 0.8308,
+      "learning_rate": 1.2261419304555395e-05,
+      "loss": 0.892,
       "step": 15667
     },
     {
-      "epoch": 0.43030952184779325,
+      "epoch": 0.4446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.2708989114474837e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.226052402624021e-05,
+      "loss": 0.8096,
       "step": 15668
     },
     {
-      "epoch": 0.4303369860756365,
+      "epoch": 0.4446367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.2708132848705189e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.2259628728830277e-05,
+      "loss": 0.9156,
       "step": 15669
     },
     {
-      "epoch": 0.4303644503034797,
+      "epoch": 0.44466515323496025,
       "grad_norm": 0.0,
-      "learning_rate": 1.2707276561507747e-05,
-      "loss": 0.9928,
+      "learning_rate": 1.2258733412333159e-05,
+      "loss": 0.86,
       "step": 15670
     },
     {
-      "epoch": 0.43039191453132297,
+      "epoch": 0.44469353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 1.2706420252889281e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.2257838076756418e-05,
+      "loss": 0.9475,
       "step": 15671
     },
     {
-      "epoch": 0.4304193787591662,
+      "epoch": 0.4447219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.2705563922856572e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.2256942722107618e-05,
+      "loss": 0.9418,
       "step": 15672
     },
     {
-      "epoch": 0.4304468429870094,
+      "epoch": 0.44475028376844494,
       "grad_norm": 0.0,
-      "learning_rate": 1.2704707571416393e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.2256047348394322e-05,
+      "loss": 0.9155,
       "step": 15673
     },
     {
-      "epoch": 0.43047430721485264,
+      "epoch": 0.44477866061293986,
       "grad_norm": 0.0,
-      "learning_rate": 1.2703851198575519e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.2255151955624092e-05,
+      "loss": 0.8687,
       "step": 15674
     },
     {
-      "epoch": 0.4305017714426959,
+      "epoch": 0.4448070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.2702994804340729e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.225425654380449e-05,
+      "loss": 0.9402,
       "step": 15675
     },
     {
-      "epoch": 0.43052923567053913,
+      "epoch": 0.44483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.2702138388718797e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.2253361112943084e-05,
+      "loss": 0.883,
       "step": 15676
     },
     {
-      "epoch": 0.43055669989838236,
+      "epoch": 0.4448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.2701281951716497e-05,
-      "loss": 0.9038,
+      "learning_rate": 1.2252465663047434e-05,
+      "loss": 0.7651,
       "step": 15677
     },
     {
-      "epoch": 0.4305841641262256,
+      "epoch": 0.4448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.2700425493340608e-05,
-      "loss": 0.8642,
+      "learning_rate": 1.2251570194125106e-05,
+      "loss": 0.8624,
       "step": 15678
     },
     {
-      "epoch": 0.4306116283540688,
+      "epoch": 0.4449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.269956901359791e-05,
-      "loss": 0.8251,
+      "learning_rate": 1.2250674706183664e-05,
+      "loss": 0.9186,
       "step": 15679
     },
     {
-      "epoch": 0.4306390925819121,
+      "epoch": 0.4449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.2698712512495173e-05,
-      "loss": 0.8976,
+      "learning_rate": 1.2249779199230672e-05,
+      "loss": 0.9581,
       "step": 15680
     },
     {
-      "epoch": 0.4306665568097553,
+      "epoch": 0.4449772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.2697855990039176e-05,
-      "loss": 0.8807,
+      "learning_rate": 1.2248883673273693e-05,
+      "loss": 0.783,
       "step": 15681
     },
     {
-      "epoch": 0.4306940210375985,
+      "epoch": 0.44500567536889896,
       "grad_norm": 0.0,
-      "learning_rate": 1.26969994462367e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.2247988128320292e-05,
+      "loss": 0.7298,
       "step": 15682
     },
     {
-      "epoch": 0.43072148526544174,
+      "epoch": 0.44503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.2696142881094519e-05,
-      "loss": 0.8838,
+      "learning_rate": 1.2247092564378032e-05,
+      "loss": 0.9979,
       "step": 15683
     },
     {
-      "epoch": 0.430748949493285,
+      "epoch": 0.4450624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.2695286294619411e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.2246196981454483e-05,
+      "loss": 0.8999,
       "step": 15684
     },
     {
-      "epoch": 0.43077641372112824,
+      "epoch": 0.44509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.2694429686818155e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.2245301379557209e-05,
+      "loss": 0.9511,
       "step": 15685
     },
     {
-      "epoch": 0.43080387794897146,
+      "epoch": 0.44511918274687856,
       "grad_norm": 0.0,
-      "learning_rate": 1.2693573057697526e-05,
-      "loss": 0.8547,
+      "learning_rate": 1.224440575869377e-05,
+      "loss": 0.9029,
       "step": 15686
     },
     {
-      "epoch": 0.4308313421768147,
+      "epoch": 0.4451475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.2692716407264305e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.2243510118871736e-05,
+      "loss": 0.9384,
       "step": 15687
     },
     {
-      "epoch": 0.43085880640465796,
+      "epoch": 0.44517593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 1.2691859735525267e-05,
-      "loss": 0.9563,
+      "learning_rate": 1.2242614460098672e-05,
+      "loss": 0.868,
       "step": 15688
     },
     {
-      "epoch": 0.4308862706325012,
+      "epoch": 0.44520431328036325,
       "grad_norm": 0.0,
-      "learning_rate": 1.269100304248719e-05,
-      "loss": 0.8228,
+      "learning_rate": 1.2241718782382139e-05,
+      "loss": 0.8198,
       "step": 15689
     },
     {
-      "epoch": 0.4309137348603444,
+      "epoch": 0.4452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.2690146328156859e-05,
-      "loss": 0.86,
+      "learning_rate": 1.2240823085729712e-05,
+      "loss": 0.865,
       "step": 15690
     },
     {
-      "epoch": 0.4309411990881876,
+      "epoch": 0.445261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.268928959254105e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.2239927370148948e-05,
+      "loss": 0.8866,
       "step": 15691
     },
     {
-      "epoch": 0.43096866331603084,
+      "epoch": 0.4452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.2688432835646534e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.2239031635647417e-05,
+      "loss": 0.9274,
       "step": 15692
     },
     {
-      "epoch": 0.4309961275438741,
+      "epoch": 0.4453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.2687576057480098e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.2238135882232687e-05,
+      "loss": 0.903,
       "step": 15693
     },
     {
-      "epoch": 0.43102359177171734,
+      "epoch": 0.44534619750283766,
       "grad_norm": 0.0,
-      "learning_rate": 1.2686719258048518e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.2237240109912323e-05,
+      "loss": 0.9503,
       "step": 15694
     },
     {
-      "epoch": 0.43105105599956056,
+      "epoch": 0.4453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.2685862437358579e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.2236344318693889e-05,
+      "loss": 0.9361,
       "step": 15695
     },
     {
-      "epoch": 0.4310785202274038,
+      "epoch": 0.4454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.2685005595417053e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.2235448508584956e-05,
+      "loss": 1.0135,
       "step": 15696
     },
     {
-      "epoch": 0.43110598445524706,
+      "epoch": 0.44543132803632235,
       "grad_norm": 0.0,
-      "learning_rate": 1.2684148732230724e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.2234552679593089e-05,
+      "loss": 1.019,
       "step": 15697
     },
     {
-      "epoch": 0.4311334486830903,
+      "epoch": 0.44545970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.2683291847806368e-05,
-      "loss": 0.8931,
+      "learning_rate": 1.2233656831725853e-05,
+      "loss": 1.0238,
       "step": 15698
     },
     {
-      "epoch": 0.4311609129109335,
+      "epoch": 0.4454880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.268243494215077e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.2232760964990818e-05,
+      "loss": 0.9933,
       "step": 15699
     },
     {
-      "epoch": 0.4311883771387767,
+      "epoch": 0.44551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.2681578015270705e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.2231865079395549e-05,
+      "loss": 0.9844,
       "step": 15700
     },
     {
-      "epoch": 0.43121584136662,
+      "epoch": 0.44554483541430195,
       "grad_norm": 0.0,
-      "learning_rate": 1.2680721067172956e-05,
-      "loss": 0.8438,
+      "learning_rate": 1.2230969174947619e-05,
+      "loss": 0.8647,
       "step": 15701
     },
     {
-      "epoch": 0.4312433055944632,
+      "epoch": 0.4455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.2679864097864305e-05,
-      "loss": 0.9522,
+      "learning_rate": 1.2230073251654591e-05,
+      "loss": 0.9589,
       "step": 15702
     },
     {
-      "epoch": 0.43127076982230644,
+      "epoch": 0.44560158910329173,
       "grad_norm": 0.0,
-      "learning_rate": 1.2679007107351535e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.2229177309524032e-05,
+      "loss": 0.9389,
       "step": 15703
     },
     {
-      "epoch": 0.43129823405014966,
+      "epoch": 0.4456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.2678150095641417e-05,
-      "loss": 0.897,
+      "learning_rate": 1.2228281348563513e-05,
+      "loss": 0.8446,
       "step": 15704
     },
     {
-      "epoch": 0.43132569827799294,
+      "epoch": 0.4456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.2677293062740739e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.2227385368780605e-05,
+      "loss": 0.8165,
       "step": 15705
     },
     {
-      "epoch": 0.43135316250583616,
+      "epoch": 0.44568671963677636,
       "grad_norm": 0.0,
-      "learning_rate": 1.2676436008656283e-05,
-      "loss": 0.8945,
+      "learning_rate": 1.2226489370182865e-05,
+      "loss": 0.8632,
       "step": 15706
     },
     {
-      "epoch": 0.4313806267336794,
+      "epoch": 0.4457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.267557893339483e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.2225593352777873e-05,
+      "loss": 0.8979,
       "step": 15707
     },
     {
-      "epoch": 0.4314080909615226,
+      "epoch": 0.4457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.2674721836963157e-05,
-      "loss": 0.8242,
+      "learning_rate": 1.2224697316573197e-05,
+      "loss": 0.8784,
       "step": 15708
     },
     {
-      "epoch": 0.4314355551893658,
+      "epoch": 0.44577185017026105,
       "grad_norm": 0.0,
-      "learning_rate": 1.267386471936805e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.22238012615764e-05,
+      "loss": 0.9025,
       "step": 15709
     },
     {
-      "epoch": 0.4314630194172091,
+      "epoch": 0.44580022701475597,
       "grad_norm": 0.0,
-      "learning_rate": 1.2673007580616287e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.2222905187795053e-05,
+      "loss": 0.9718,
       "step": 15710
     },
     {
-      "epoch": 0.4314904836450523,
+      "epoch": 0.4458286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.2672150420714657e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.222200909523673e-05,
+      "loss": 0.962,
       "step": 15711
     },
     {
-      "epoch": 0.43151794787289555,
+      "epoch": 0.44585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.2671293239669937e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.2221112983908994e-05,
+      "loss": 0.8921,
       "step": 15712
     },
     {
-      "epoch": 0.43154541210073877,
+      "epoch": 0.44588535754824066,
       "grad_norm": 0.0,
-      "learning_rate": 1.2670436037488907e-05,
-      "loss": 0.9803,
+      "learning_rate": 1.2220216853819413e-05,
+      "loss": 0.9449,
       "step": 15713
     },
     {
-      "epoch": 0.43157287632858204,
+      "epoch": 0.4459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.2669578814178357e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.2219320704975564e-05,
+      "loss": 0.9909,
       "step": 15714
     },
     {
-      "epoch": 0.43160034055642527,
+      "epoch": 0.44594211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.2668721569745062e-05,
-      "loss": 0.769,
+      "learning_rate": 1.2218424537385012e-05,
+      "loss": 0.9008,
       "step": 15715
     },
     {
-      "epoch": 0.4316278047842685,
+      "epoch": 0.4459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.2667864304195809e-05,
-      "loss": 0.883,
+      "learning_rate": 1.2217528351055329e-05,
+      "loss": 0.8514,
       "step": 15716
     },
     {
-      "epoch": 0.4316552690121117,
+      "epoch": 0.4459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.266700701753738e-05,
-      "loss": 0.9083,
+      "learning_rate": 1.2216632145994086e-05,
+      "loss": 0.7814,
       "step": 15717
     },
     {
-      "epoch": 0.431682733239955,
+      "epoch": 0.4460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.2666149709776558e-05,
-      "loss": 0.9033,
+      "learning_rate": 1.2215735922208849e-05,
+      "loss": 1.0045,
       "step": 15718
     },
     {
-      "epoch": 0.4317101974677982,
+      "epoch": 0.44605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.2665292380920129e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.2214839679707193e-05,
+      "loss": 0.9492,
       "step": 15719
     },
     {
-      "epoch": 0.43173766169564143,
+      "epoch": 0.4460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.2664435030974873e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.2213943418496685e-05,
+      "loss": 0.9537,
       "step": 15720
     },
     {
-      "epoch": 0.43176512592348465,
+      "epoch": 0.44611237230419976,
       "grad_norm": 0.0,
-      "learning_rate": 1.2663577659947573e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.2213047138584899e-05,
+      "loss": 1.0387,
       "step": 15721
     },
     {
-      "epoch": 0.43179259015132787,
+      "epoch": 0.44614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.2662720267845014e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.2212150839979402e-05,
+      "loss": 0.9313,
       "step": 15722
     },
     {
-      "epoch": 0.43182005437917115,
+      "epoch": 0.44616912599318953,
       "grad_norm": 0.0,
-      "learning_rate": 1.2661862854673985e-05,
-      "loss": 0.8687,
+      "learning_rate": 1.2211254522687767e-05,
+      "loss": 0.8848,
       "step": 15723
     },
     {
-      "epoch": 0.43184751860701437,
+      "epoch": 0.44619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.2661005420441266e-05,
-      "loss": 0.8107,
+      "learning_rate": 1.2210358186717569e-05,
+      "loss": 0.8852,
       "step": 15724
     },
     {
-      "epoch": 0.4318749828348576,
+      "epoch": 0.44622587968217936,
       "grad_norm": 0.0,
-      "learning_rate": 1.2660147965153638e-05,
-      "loss": 0.926,
+      "learning_rate": 1.2209461832076375e-05,
+      "loss": 0.8804,
       "step": 15725
     },
     {
-      "epoch": 0.4319024470627008,
+      "epoch": 0.4462542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.2659290488817891e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.2208565458771755e-05,
+      "loss": 0.8668,
       "step": 15726
     },
     {
-      "epoch": 0.4319299112905441,
+      "epoch": 0.44628263337116914,
       "grad_norm": 0.0,
-      "learning_rate": 1.2658432991440808e-05,
-      "loss": 0.908,
+      "learning_rate": 1.2207669066811288e-05,
+      "loss": 0.9496,
       "step": 15727
     },
     {
-      "epoch": 0.4319573755183873,
+      "epoch": 0.446311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.2657575473029172e-05,
-      "loss": 0.8945,
+      "learning_rate": 1.220677265620254e-05,
+      "loss": 0.8396,
       "step": 15728
     },
     {
-      "epoch": 0.43198483974623053,
+      "epoch": 0.4463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.2656717933589769e-05,
-      "loss": 0.7991,
+      "learning_rate": 1.220587622695308e-05,
+      "loss": 0.8502,
       "step": 15729
     },
     {
-      "epoch": 0.43201230397407375,
+      "epoch": 0.4463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.2655860373129385e-05,
-      "loss": 0.8747,
+      "learning_rate": 1.2204979779070485e-05,
+      "loss": 0.9158,
       "step": 15730
     },
     {
-      "epoch": 0.43203976820191703,
+      "epoch": 0.4463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.2655002791654804e-05,
-      "loss": 1.0175,
+      "learning_rate": 1.2204083312562331e-05,
+      "loss": 0.8931,
       "step": 15731
     },
     {
-      "epoch": 0.43206723242976025,
+      "epoch": 0.4464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.2654145189172816e-05,
-      "loss": 0.8757,
+      "learning_rate": 1.2203186827436185e-05,
+      "loss": 0.9678,
       "step": 15732
     },
     {
-      "epoch": 0.43209469665760347,
+      "epoch": 0.44645289443813846,
       "grad_norm": 0.0,
-      "learning_rate": 1.2653287565690197e-05,
-      "loss": 0.891,
+      "learning_rate": 1.2202290323699617e-05,
+      "loss": 0.8837,
       "step": 15733
     },
     {
-      "epoch": 0.4321221608854467,
+      "epoch": 0.4464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.2652429921213742e-05,
-      "loss": 0.975,
+      "learning_rate": 1.2201393801360208e-05,
+      "loss": 0.867,
       "step": 15734
     },
     {
-      "epoch": 0.4321496251132899,
+      "epoch": 0.4465096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.2651572255750237e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.2200497260425526e-05,
+      "loss": 0.8834,
       "step": 15735
     },
     {
-      "epoch": 0.4321770893411332,
+      "epoch": 0.44653802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.2650714569306464e-05,
-      "loss": 0.9103,
+      "learning_rate": 1.2199600700903143e-05,
+      "loss": 1.0549,
       "step": 15736
     },
     {
-      "epoch": 0.4322045535689764,
+      "epoch": 0.44656640181611806,
       "grad_norm": 0.0,
-      "learning_rate": 1.2649856861889207e-05,
-      "loss": 0.7989,
+      "learning_rate": 1.2198704122800633e-05,
+      "loss": 0.8408,
       "step": 15737
     },
     {
-      "epoch": 0.43223201779681963,
+      "epoch": 0.4465947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.2648999133505262e-05,
-      "loss": 0.9439,
+      "learning_rate": 1.2197807526125573e-05,
+      "loss": 0.9863,
       "step": 15738
     },
     {
-      "epoch": 0.43225948202466286,
+      "epoch": 0.44662315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 1.2648141384161406e-05,
-      "loss": 0.8917,
+      "learning_rate": 1.2196910910885534e-05,
+      "loss": 0.7618,
       "step": 15739
     },
     {
-      "epoch": 0.43228694625250613,
+      "epoch": 0.4466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.264728361386443e-05,
-      "loss": 0.9229,
+      "learning_rate": 1.2196014277088089e-05,
+      "loss": 0.9901,
       "step": 15740
     },
     {
-      "epoch": 0.43231441048034935,
+      "epoch": 0.4466799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.264642582262112e-05,
-      "loss": 0.9765,
+      "learning_rate": 1.2195117624740812e-05,
+      "loss": 0.9734,
       "step": 15741
     },
     {
-      "epoch": 0.4323418747081926,
+      "epoch": 0.44670828603859253,
       "grad_norm": 0.0,
-      "learning_rate": 1.2645568010438262e-05,
-      "loss": 0.9645,
+      "learning_rate": 1.2194220953851278e-05,
+      "loss": 0.8744,
       "step": 15742
     },
     {
-      "epoch": 0.4323693389360358,
+      "epoch": 0.4467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.264471017732265e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.2193324264427062e-05,
+      "loss": 0.891,
       "step": 15743
     },
     {
-      "epoch": 0.4323968031638791,
+      "epoch": 0.4467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.2643852323281064e-05,
-      "loss": 0.9396,
+      "learning_rate": 1.2192427556475736e-05,
+      "loss": 0.8556,
       "step": 15744
     },
     {
-      "epoch": 0.4324242673917223,
+      "epoch": 0.44679341657207716,
       "grad_norm": 0.0,
-      "learning_rate": 1.2642994448320295e-05,
-      "loss": 0.9586,
+      "learning_rate": 1.2191530830004877e-05,
+      "loss": 0.8486,
       "step": 15745
     },
     {
-      "epoch": 0.4324517316195655,
+      "epoch": 0.4468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.2642136552447127e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.2190634085022056e-05,
+      "loss": 0.9557,
       "step": 15746
     },
     {
-      "epoch": 0.43247919584740874,
+      "epoch": 0.446850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.2641278635668352e-05,
-      "loss": 0.9275,
+      "learning_rate": 1.2189737321534853e-05,
+      "loss": 0.7736,
       "step": 15747
     },
     {
-      "epoch": 0.43250666007525196,
+      "epoch": 0.44687854710556185,
       "grad_norm": 0.0,
-      "learning_rate": 1.2640420697990761e-05,
-      "loss": 0.8759,
+      "learning_rate": 1.2188840539550841e-05,
+      "loss": 0.9964,
       "step": 15748
     },
     {
-      "epoch": 0.43253412430309524,
+      "epoch": 0.44690692395005677,
       "grad_norm": 0.0,
-      "learning_rate": 1.2639562739421136e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.2187943739077595e-05,
+      "loss": 0.8366,
       "step": 15749
     },
     {
-      "epoch": 0.43256158853093846,
+      "epoch": 0.4469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.2638704759966269e-05,
-      "loss": 0.9972,
+      "learning_rate": 1.2187046920122686e-05,
+      "loss": 1.0168,
       "step": 15750
     },
     {
-      "epoch": 0.4325890527587817,
+      "epoch": 0.44696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.2637846759632947e-05,
-      "loss": 1.003,
+      "learning_rate": 1.2186150082693697e-05,
+      "loss": 1.1106,
       "step": 15751
     },
     {
-      "epoch": 0.4326165169866249,
+      "epoch": 0.44699205448354146,
       "grad_norm": 0.0,
-      "learning_rate": 1.263698873842796e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.2185253226798198e-05,
+      "loss": 0.8866,
       "step": 15752
     },
     {
-      "epoch": 0.4326439812144682,
+      "epoch": 0.4470204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.2636130696358098e-05,
-      "loss": 0.8251,
+      "learning_rate": 1.2184356352443766e-05,
+      "loss": 1.0148,
       "step": 15753
     },
     {
-      "epoch": 0.4326714454423114,
+      "epoch": 0.44704880817253123,
       "grad_norm": 0.0,
-      "learning_rate": 1.2635272633430145e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.2183459459637981e-05,
+      "loss": 0.9038,
       "step": 15754
     },
     {
-      "epoch": 0.4326989096701546,
+      "epoch": 0.4470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.2634414549650898e-05,
-      "loss": 0.8607,
+      "learning_rate": 1.2182562548388414e-05,
+      "loss": 0.9202,
       "step": 15755
     },
     {
-      "epoch": 0.43272637389799784,
+      "epoch": 0.447105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.2633556445027141e-05,
-      "loss": 0.9983,
+      "learning_rate": 1.2181665618702643e-05,
+      "loss": 0.8892,
       "step": 15756
     },
     {
-      "epoch": 0.4327538381258411,
+      "epoch": 0.44713393870601587,
       "grad_norm": 0.0,
-      "learning_rate": 1.2632698319565666e-05,
-      "loss": 0.848,
+      "learning_rate": 1.2180768670588244e-05,
+      "loss": 0.8419,
       "step": 15757
     },
     {
-      "epoch": 0.43278130235368434,
+      "epoch": 0.4471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.263184017327326e-05,
-      "loss": 0.9984,
+      "learning_rate": 1.2179871704052795e-05,
+      "loss": 0.8772,
       "step": 15758
     },
     {
-      "epoch": 0.43280876658152756,
+      "epoch": 0.4471906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.263098200615672e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.2178974719103873e-05,
+      "loss": 0.9859,
       "step": 15759
     },
     {
-      "epoch": 0.4328362308093708,
+      "epoch": 0.44721906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 1.2630123818222827e-05,
-      "loss": 0.9855,
+      "learning_rate": 1.217807771574905e-05,
+      "loss": 0.8382,
       "step": 15760
     },
     {
-      "epoch": 0.432863695037214,
+      "epoch": 0.44724744608399547,
       "grad_norm": 0.0,
-      "learning_rate": 1.2629265609478373e-05,
-      "loss": 0.7518,
+      "learning_rate": 1.2177180693995908e-05,
+      "loss": 0.9,
       "step": 15761
     },
     {
-      "epoch": 0.4328911592650573,
+      "epoch": 0.44727582292849033,
       "grad_norm": 0.0,
-      "learning_rate": 1.2628407379930157e-05,
-      "loss": 0.9634,
+      "learning_rate": 1.2176283653852024e-05,
+      "loss": 0.8324,
       "step": 15762
     },
     {
-      "epoch": 0.4329186234929005,
+      "epoch": 0.44730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.2627549129584961e-05,
-      "loss": 0.8496,
+      "learning_rate": 1.2175386595324974e-05,
+      "loss": 0.9533,
       "step": 15763
     },
     {
-      "epoch": 0.4329460877207437,
+      "epoch": 0.44733257661748016,
       "grad_norm": 0.0,
-      "learning_rate": 1.2626690858449577e-05,
-      "loss": 0.8924,
+      "learning_rate": 1.2174489518422333e-05,
+      "loss": 0.7776,
       "step": 15764
     },
     {
-      "epoch": 0.43297355194858694,
+      "epoch": 0.447360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.26258325665308e-05,
-      "loss": 1.0136,
+      "learning_rate": 1.2173592423151683e-05,
+      "loss": 0.9439,
       "step": 15765
     },
     {
-      "epoch": 0.4330010161764302,
+      "epoch": 0.44738933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 1.2624974253835416e-05,
-      "loss": 0.8785,
+      "learning_rate": 1.2172695309520602e-05,
+      "loss": 0.9342,
       "step": 15766
     },
     {
-      "epoch": 0.43302848040427344,
+      "epoch": 0.4474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.2624115920370221e-05,
-      "loss": 0.97,
+      "learning_rate": 1.2171798177536665e-05,
+      "loss": 0.9559,
       "step": 15767
     },
     {
-      "epoch": 0.43305594463211666,
+      "epoch": 0.4474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.2623257566142005e-05,
-      "loss": 0.8583,
+      "learning_rate": 1.2170901027207447e-05,
+      "loss": 0.8911,
       "step": 15768
     },
     {
-      "epoch": 0.4330834088599599,
+      "epoch": 0.4474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.2622399191157557e-05,
-      "loss": 0.837,
+      "learning_rate": 1.2170003858540536e-05,
+      "loss": 0.8017,
       "step": 15769
     },
     {
-      "epoch": 0.43311087308780316,
+      "epoch": 0.4475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.2621540795423672e-05,
-      "loss": 0.929,
+      "learning_rate": 1.21691066715435e-05,
+      "loss": 0.9022,
       "step": 15770
     },
     {
-      "epoch": 0.4331383373156464,
+      "epoch": 0.4475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.2620682378947142e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.2168209466223928e-05,
+      "loss": 0.9729,
       "step": 15771
     },
     {
-      "epoch": 0.4331658015434896,
+      "epoch": 0.44755959137343926,
       "grad_norm": 0.0,
-      "learning_rate": 1.2619823941734755e-05,
-      "loss": 0.851,
+      "learning_rate": 1.2167312242589391e-05,
+      "loss": 0.9734,
       "step": 15772
     },
     {
-      "epoch": 0.4331932657713328,
+      "epoch": 0.4475879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.2618965483793307e-05,
-      "loss": 1.0128,
+      "learning_rate": 1.216641500064747e-05,
+      "loss": 0.7922,
       "step": 15773
     },
     {
-      "epoch": 0.43322072999917605,
+      "epoch": 0.44761634506242903,
       "grad_norm": 0.0,
-      "learning_rate": 1.261810700512959e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.2165517740405743e-05,
+      "loss": 0.7977,
       "step": 15774
     },
     {
-      "epoch": 0.4332481942270193,
+      "epoch": 0.44764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.2617248505750398e-05,
-      "loss": 0.955,
+      "learning_rate": 1.216462046187179e-05,
+      "loss": 0.8614,
       "step": 15775
     },
     {
-      "epoch": 0.43327565845486254,
+      "epoch": 0.44767309875141886,
       "grad_norm": 0.0,
-      "learning_rate": 1.2616389985662518e-05,
-      "loss": 0.8211,
+      "learning_rate": 1.2163723165053192e-05,
+      "loss": 0.8313,
       "step": 15776
     },
     {
-      "epoch": 0.43330312268270577,
+      "epoch": 0.4477014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.2615531444872749e-05,
-      "loss": 0.8667,
+      "learning_rate": 1.2162825849957525e-05,
+      "loss": 0.9365,
       "step": 15777
     },
     {
-      "epoch": 0.433330586910549,
+      "epoch": 0.44772985244040864,
       "grad_norm": 0.0,
-      "learning_rate": 1.2614672883387882e-05,
-      "loss": 0.9091,
+      "learning_rate": 1.2161928516592373e-05,
+      "loss": 0.9496,
       "step": 15778
     },
     {
-      "epoch": 0.43335805113839226,
+      "epoch": 0.4477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.261381430121471e-05,
-      "loss": 0.8396,
+      "learning_rate": 1.2161031164965315e-05,
+      "loss": 0.9861,
       "step": 15779
     },
     {
-      "epoch": 0.4333855153662355,
+      "epoch": 0.4477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.2612955698360028e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.2160133795083926e-05,
+      "loss": 0.8648,
       "step": 15780
     },
     {
-      "epoch": 0.4334129795940787,
+      "epoch": 0.4478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.2612097074830627e-05,
-      "loss": 0.8949,
+      "learning_rate": 1.2159236406955793e-05,
+      "loss": 0.9358,
       "step": 15781
     },
     {
-      "epoch": 0.43344044382192193,
+      "epoch": 0.4478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.2611238430633303e-05,
-      "loss": 0.8884,
+      "learning_rate": 1.2158339000588493e-05,
+      "loss": 1.0009,
       "step": 15782
     },
     {
-      "epoch": 0.4334679080497652,
+      "epoch": 0.4478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.261037976577485e-05,
-      "loss": 0.9126,
+      "learning_rate": 1.2157441575989604e-05,
+      "loss": 0.918,
       "step": 15783
     },
     {
-      "epoch": 0.4334953722776084,
+      "epoch": 0.44790011350737796,
       "grad_norm": 0.0,
-      "learning_rate": 1.2609521080262056e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.2156544133166708e-05,
+      "loss": 0.9863,
       "step": 15784
     },
     {
-      "epoch": 0.43352283650545165,
+      "epoch": 0.4479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.2608662374101726e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.215564667212739e-05,
+      "loss": 0.9085,
       "step": 15785
     },
     {
-      "epoch": 0.43355030073329487,
+      "epoch": 0.44795686719636774,
       "grad_norm": 0.0,
-      "learning_rate": 1.2607803647300647e-05,
-      "loss": 0.9478,
+      "learning_rate": 1.2154749192879227e-05,
+      "loss": 0.8034,
       "step": 15786
     },
     {
-      "epoch": 0.4335777649611381,
+      "epoch": 0.44798524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.2606944899865616e-05,
-      "loss": 0.9638,
+      "learning_rate": 1.21538516954298e-05,
+      "loss": 0.8466,
       "step": 15787
     },
     {
-      "epoch": 0.43360522918898137,
+      "epoch": 0.44801362088535757,
       "grad_norm": 0.0,
-      "learning_rate": 1.2606086131803427e-05,
-      "loss": 1.0291,
+      "learning_rate": 1.2152954179786691e-05,
+      "loss": 0.9278,
       "step": 15788
     },
     {
-      "epoch": 0.4336326934168246,
+      "epoch": 0.4480419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.2605227343120873e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.2152056645957481e-05,
+      "loss": 1.0146,
       "step": 15789
     },
     {
-      "epoch": 0.4336601576446678,
+      "epoch": 0.44807037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 1.2604368533824751e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.2151159093949751e-05,
+      "loss": 0.8949,
       "step": 15790
     },
     {
-      "epoch": 0.43368762187251103,
+      "epoch": 0.4480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.2603509703921859e-05,
-      "loss": 0.8594,
+      "learning_rate": 1.2150261523771083e-05,
+      "loss": 0.9381,
       "step": 15791
     },
     {
-      "epoch": 0.4337150861003543,
+      "epoch": 0.4481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.260265085341899e-05,
-      "loss": 0.9092,
+      "learning_rate": 1.2149363935429061e-05,
+      "loss": 0.8864,
       "step": 15792
     },
     {
-      "epoch": 0.43374255032819753,
+      "epoch": 0.44815550510783203,
       "grad_norm": 0.0,
-      "learning_rate": 1.2601791982322937e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.2148466328931268e-05,
+      "loss": 0.9485,
       "step": 15793
     },
     {
-      "epoch": 0.43377001455604075,
+      "epoch": 0.4481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.26009330906405e-05,
-      "loss": 0.9188,
+      "learning_rate": 1.2147568704285278e-05,
+      "loss": 0.8966,
       "step": 15794
     },
     {
-      "epoch": 0.43379747878388397,
+      "epoch": 0.4482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.2600074178378471e-05,
-      "loss": 0.9057,
+      "learning_rate": 1.2146671061498682e-05,
+      "loss": 0.8966,
       "step": 15795
     },
     {
-      "epoch": 0.43382494301172725,
+      "epoch": 0.44824063564131666,
       "grad_norm": 0.0,
-      "learning_rate": 1.2599215245543647e-05,
-      "loss": 0.9049,
+      "learning_rate": 1.214577340057906e-05,
+      "loss": 1.043,
       "step": 15796
     },
     {
-      "epoch": 0.43385240723957047,
+      "epoch": 0.4482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.2598356292142828e-05,
-      "loss": 0.8737,
+      "learning_rate": 1.2144875721533989e-05,
+      "loss": 0.9531,
       "step": 15797
     },
     {
-      "epoch": 0.4338798714674137,
+      "epoch": 0.4482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.2597497318182805e-05,
-      "loss": 0.942,
+      "learning_rate": 1.2143978024371056e-05,
+      "loss": 0.9146,
       "step": 15798
     },
     {
-      "epoch": 0.4339073356952569,
+      "epoch": 0.44832576617480135,
       "grad_norm": 0.0,
-      "learning_rate": 1.2596638323670381e-05,
-      "loss": 0.8444,
+      "learning_rate": 1.214308030909785e-05,
+      "loss": 0.9922,
       "step": 15799
     },
     {
-      "epoch": 0.4339347999231002,
+      "epoch": 0.44835414301929627,
       "grad_norm": 0.0,
-      "learning_rate": 1.2595779308612346e-05,
-      "loss": 0.8237,
+      "learning_rate": 1.2142182575721946e-05,
+      "loss": 0.8016,
       "step": 15800
     },
     {
-      "epoch": 0.4339622641509434,
+      "epoch": 0.44838251986379113,
       "grad_norm": 0.0,
-      "learning_rate": 1.2594920273015499e-05,
-      "loss": 0.8312,
+      "learning_rate": 1.2141284824250927e-05,
+      "loss": 0.7686,
       "step": 15801
     },
     {
-      "epoch": 0.43398972837878663,
+      "epoch": 0.44841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.2594061216886637e-05,
-      "loss": 0.9913,
+      "learning_rate": 1.2140387054692381e-05,
+      "loss": 0.9145,
       "step": 15802
     },
     {
-      "epoch": 0.43401719260662985,
+      "epoch": 0.4484392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.2593202140232561e-05,
-      "loss": 0.9778,
+      "learning_rate": 1.2139489267053892e-05,
+      "loss": 1.0256,
       "step": 15803
     },
     {
-      "epoch": 0.4340446568344731,
+      "epoch": 0.4484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.2592343043060064e-05,
-      "loss": 0.7689,
+      "learning_rate": 1.2138591461343037e-05,
+      "loss": 0.8936,
       "step": 15804
     },
     {
-      "epoch": 0.43407212106231635,
+      "epoch": 0.44849602724177073,
       "grad_norm": 0.0,
-      "learning_rate": 1.2591483925375943e-05,
-      "loss": 0.8622,
+      "learning_rate": 1.2137693637567405e-05,
+      "loss": 0.9188,
       "step": 15805
     },
     {
-      "epoch": 0.4340995852901596,
+      "epoch": 0.4485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.2590624787186998e-05,
-      "loss": 0.8729,
+      "learning_rate": 1.2136795795734577e-05,
+      "loss": 0.9437,
       "step": 15806
     },
     {
-      "epoch": 0.4341270495180028,
+      "epoch": 0.4485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.2589765628500028e-05,
-      "loss": 0.9576,
+      "learning_rate": 1.2135897935852143e-05,
+      "loss": 0.8252,
       "step": 15807
     },
     {
-      "epoch": 0.434154513745846,
+      "epoch": 0.44858115777525537,
       "grad_norm": 0.0,
-      "learning_rate": 1.2588906449321827e-05,
-      "loss": 0.9588,
+      "learning_rate": 1.213500005792768e-05,
+      "loss": 0.9346,
       "step": 15808
     },
     {
-      "epoch": 0.4341819779736893,
+      "epoch": 0.4486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.2588047249659196e-05,
-      "loss": 0.9517,
+      "learning_rate": 1.2134102161968775e-05,
+      "loss": 0.9817,
       "step": 15809
     },
     {
-      "epoch": 0.4342094422015325,
+      "epoch": 0.4486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.2587188029518931e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.2133204247983016e-05,
+      "loss": 0.9791,
       "step": 15810
     },
     {
-      "epoch": 0.43423690642937574,
+      "epoch": 0.44866628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 1.2586328788907835e-05,
-      "loss": 0.7939,
+      "learning_rate": 1.2132306315977983e-05,
+      "loss": 1.0544,
       "step": 15811
     },
     {
-      "epoch": 0.43426437065721896,
+      "epoch": 0.448694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.2585469527832703e-05,
-      "loss": 0.9033,
+      "learning_rate": 1.2131408365961263e-05,
+      "loss": 0.9027,
       "step": 15812
     },
     {
-      "epoch": 0.43429183488506223,
+      "epoch": 0.44872304199772983,
       "grad_norm": 0.0,
-      "learning_rate": 1.2584610246300332e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.213051039794044e-05,
+      "loss": 0.985,
       "step": 15813
     },
     {
-      "epoch": 0.43431929911290545,
+      "epoch": 0.44875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.2583750944317525e-05,
-      "loss": 0.957,
+      "learning_rate": 1.21296124119231e-05,
+      "loss": 0.8929,
       "step": 15814
     },
     {
-      "epoch": 0.4343467633407487,
+      "epoch": 0.44877979568671966,
       "grad_norm": 0.0,
-      "learning_rate": 1.2582891621891083e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.2128714407916827e-05,
+      "loss": 1.0296,
       "step": 15815
     },
     {
-      "epoch": 0.4343742275685919,
+      "epoch": 0.4488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.2582032279027801e-05,
-      "loss": 0.998,
+      "learning_rate": 1.212781638592921e-05,
+      "loss": 0.8221,
       "step": 15816
     },
     {
-      "epoch": 0.4344016917964351,
+      "epoch": 0.44883654937570944,
       "grad_norm": 0.0,
-      "learning_rate": 1.2581172915734476e-05,
-      "loss": 0.8653,
+      "learning_rate": 1.2126918345967832e-05,
+      "loss": 1.0135,
       "step": 15817
     },
     {
-      "epoch": 0.4344291560242784,
+      "epoch": 0.4488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.2580313532017914e-05,
-      "loss": 0.8446,
+      "learning_rate": 1.212602028804028e-05,
+      "loss": 0.9286,
       "step": 15818
     },
     {
-      "epoch": 0.4344566202521216,
+      "epoch": 0.4488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.2579454127884909e-05,
-      "loss": 0.8521,
+      "learning_rate": 1.2125122212154137e-05,
+      "loss": 0.9131,
       "step": 15819
     },
     {
-      "epoch": 0.43448408447996484,
+      "epoch": 0.44892167990919407,
       "grad_norm": 0.0,
-      "learning_rate": 1.2578594703342267e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.2124224118316995e-05,
+      "loss": 0.8609,
       "step": 15820
     },
     {
-      "epoch": 0.43451154870780806,
+      "epoch": 0.448950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.2577735258396785e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.2123326006536429e-05,
+      "loss": 0.9699,
       "step": 15821
     },
     {
-      "epoch": 0.43453901293565134,
+      "epoch": 0.4489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.2576875793055256e-05,
-      "loss": 0.9033,
+      "learning_rate": 1.2122427876820038e-05,
+      "loss": 0.9197,
       "step": 15822
     },
     {
-      "epoch": 0.43456647716349456,
+      "epoch": 0.44900681044267876,
       "grad_norm": 0.0,
-      "learning_rate": 1.2576016307324497e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.2121529729175402e-05,
+      "loss": 0.9674,
       "step": 15823
     },
     {
-      "epoch": 0.4345939413913378,
+      "epoch": 0.4490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.2575156801211295e-05,
-      "loss": 0.9419,
+      "learning_rate": 1.2120631563610108e-05,
+      "loss": 0.84,
       "step": 15824
     },
     {
-      "epoch": 0.434621405619181,
+      "epoch": 0.44906356413166854,
       "grad_norm": 0.0,
-      "learning_rate": 1.2574297274722454e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.2119733380131746e-05,
+      "loss": 0.8262,
       "step": 15825
     },
     {
-      "epoch": 0.4346488698470243,
+      "epoch": 0.44909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.2573437727864775e-05,
-      "loss": 0.8838,
+      "learning_rate": 1.2118835178747902e-05,
+      "loss": 0.8926,
       "step": 15826
     },
     {
-      "epoch": 0.4346763340748675,
+      "epoch": 0.44912031782065837,
       "grad_norm": 0.0,
-      "learning_rate": 1.257257816064506e-05,
-      "loss": 0.907,
+      "learning_rate": 1.211793695946616e-05,
+      "loss": 0.903,
       "step": 15827
     },
     {
-      "epoch": 0.4347037983027107,
+      "epoch": 0.4491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.2571718573070113e-05,
-      "loss": 1.0154,
+      "learning_rate": 1.211703872229411e-05,
+      "loss": 0.8275,
       "step": 15828
     },
     {
-      "epoch": 0.43473126253055394,
+      "epoch": 0.44917707150964814,
       "grad_norm": 0.0,
-      "learning_rate": 1.2570858965146734e-05,
-      "loss": 0.8028,
+      "learning_rate": 1.2116140467239337e-05,
+      "loss": 0.8452,
       "step": 15829
     },
     {
-      "epoch": 0.43475872675839716,
+      "epoch": 0.449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.2569999336881717e-05,
-      "loss": 0.9601,
+      "learning_rate": 1.2115242194309431e-05,
+      "loss": 1.0328,
       "step": 15830
     },
     {
-      "epoch": 0.43478619098624044,
+      "epoch": 0.4492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.2569139688281874e-05,
-      "loss": 0.8707,
+      "learning_rate": 1.2114343903511979e-05,
+      "loss": 0.9888,
       "step": 15831
     },
     {
-      "epoch": 0.43481365521408366,
+      "epoch": 0.44926220204313283,
       "grad_norm": 0.0,
-      "learning_rate": 1.2568280019354001e-05,
-      "loss": 0.8871,
+      "learning_rate": 1.2113445594854572e-05,
+      "loss": 0.8544,
       "step": 15832
     },
     {
-      "epoch": 0.4348411194419269,
+      "epoch": 0.4492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.2567420330104902e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.2112547268344793e-05,
+      "loss": 0.8997,
       "step": 15833
     },
     {
-      "epoch": 0.4348685836697701,
+      "epoch": 0.4493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.2566560620541379e-05,
-      "loss": 0.8444,
+      "learning_rate": 1.211164892399023e-05,
+      "loss": 1.0301,
       "step": 15834
     },
     {
-      "epoch": 0.4348960478976134,
+      "epoch": 0.44934733257661746,
       "grad_norm": 0.0,
-      "learning_rate": 1.2565700890670234e-05,
-      "loss": 0.8853,
+      "learning_rate": 1.2110750561798476e-05,
+      "loss": 0.9178,
       "step": 15835
     },
     {
-      "epoch": 0.4349235121254566,
+      "epoch": 0.4493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.2564841140498269e-05,
-      "loss": 0.8333,
+      "learning_rate": 1.2109852181777117e-05,
+      "loss": 0.9711,
       "step": 15836
     },
     {
-      "epoch": 0.4349509763532998,
+      "epoch": 0.44940408626560724,
       "grad_norm": 0.0,
-      "learning_rate": 1.2563981370032287e-05,
-      "loss": 1.0096,
+      "learning_rate": 1.2108953783933741e-05,
+      "loss": 0.9591,
       "step": 15837
     },
     {
-      "epoch": 0.43497844058114304,
+      "epoch": 0.44943246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.2563121579279094e-05,
-      "loss": 0.8884,
+      "learning_rate": 1.2108055368275938e-05,
+      "loss": 0.961,
       "step": 15838
     },
     {
-      "epoch": 0.4350059048089863,
+      "epoch": 0.44946083995459707,
       "grad_norm": 0.0,
-      "learning_rate": 1.256226176824549e-05,
-      "loss": 1.02,
+      "learning_rate": 1.2107156934811296e-05,
+      "loss": 0.8568,
       "step": 15839
     },
     {
-      "epoch": 0.43503336903682954,
+      "epoch": 0.44948921679909193,
       "grad_norm": 0.0,
-      "learning_rate": 1.2561401936938275e-05,
-      "loss": 0.9838,
+      "learning_rate": 1.2106258483547406e-05,
+      "loss": 0.9225,
       "step": 15840
     },
     {
-      "epoch": 0.43506083326467276,
+      "epoch": 0.44951759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 1.2560542085364258e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.2105360014491854e-05,
+      "loss": 0.8829,
       "step": 15841
     },
     {
-      "epoch": 0.435088297492516,
+      "epoch": 0.4495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.255968221353024e-05,
-      "loss": 0.906,
+      "learning_rate": 1.2104461527652233e-05,
+      "loss": 0.9062,
       "step": 15842
     },
     {
-      "epoch": 0.4351157617203592,
+      "epoch": 0.4495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.2558822321443024e-05,
-      "loss": 0.8721,
+      "learning_rate": 1.2103563023036131e-05,
+      "loss": 0.9294,
       "step": 15843
     },
     {
-      "epoch": 0.4351432259482025,
+      "epoch": 0.44960272417707153,
       "grad_norm": 0.0,
-      "learning_rate": 1.2557962409109417e-05,
-      "loss": 0.8465,
+      "learning_rate": 1.2102664500651136e-05,
+      "loss": 0.8955,
       "step": 15844
     },
     {
-      "epoch": 0.4351706901760457,
+      "epoch": 0.4496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.2557102476536219e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.210176596050484e-05,
+      "loss": 0.8855,
       "step": 15845
     },
     {
-      "epoch": 0.4351981544038889,
+      "epoch": 0.4496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.2556242523730231e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.2100867402604832e-05,
+      "loss": 0.9076,
       "step": 15846
     },
     {
-      "epoch": 0.43522561863173215,
+      "epoch": 0.44968785471055617,
       "grad_norm": 0.0,
-      "learning_rate": 1.2555382550698267e-05,
-      "loss": 0.8373,
+      "learning_rate": 1.2099968826958705e-05,
+      "loss": 0.7759,
       "step": 15847
     },
     {
-      "epoch": 0.4352530828595754,
+      "epoch": 0.4497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.2554522557447128e-05,
-      "loss": 0.8378,
+      "learning_rate": 1.2099070233574043e-05,
+      "loss": 1.0219,
       "step": 15848
     },
     {
-      "epoch": 0.43528054708741865,
+      "epoch": 0.449744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.2553662543983613e-05,
-      "loss": 0.8372,
+      "learning_rate": 1.2098171622458446e-05,
+      "loss": 0.8243,
       "step": 15849
     },
     {
-      "epoch": 0.43530801131526187,
+      "epoch": 0.44977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.2552802510314535e-05,
-      "loss": 0.9502,
+      "learning_rate": 1.2097272993619495e-05,
+      "loss": 0.9088,
       "step": 15850
     },
     {
-      "epoch": 0.4353354755431051,
+      "epoch": 0.44980136208853577,
       "grad_norm": 0.0,
-      "learning_rate": 1.2551942456446691e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.2096374347064785e-05,
+      "loss": 0.8353,
       "step": 15851
     },
     {
-      "epoch": 0.43536293977094837,
+      "epoch": 0.44982973893303063,
       "grad_norm": 0.0,
-      "learning_rate": 1.255108238238689e-05,
-      "loss": 0.8852,
+      "learning_rate": 1.2095475682801907e-05,
+      "loss": 0.8457,
       "step": 15852
     },
     {
-      "epoch": 0.4353904039987916,
+      "epoch": 0.44985811577752555,
       "grad_norm": 0.0,
-      "learning_rate": 1.2550222288141937e-05,
-      "loss": 0.9251,
+      "learning_rate": 1.2094577000838452e-05,
+      "loss": 0.8608,
       "step": 15853
     },
     {
-      "epoch": 0.4354178682266348,
+      "epoch": 0.4498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.254936217371864e-05,
-      "loss": 0.8663,
+      "learning_rate": 1.2093678301182013e-05,
+      "loss": 0.9246,
       "step": 15854
     },
     {
-      "epoch": 0.43544533245447803,
+      "epoch": 0.4499148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.25485020391238e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.2092779583840173e-05,
+      "loss": 0.8489,
       "step": 15855
     },
     {
-      "epoch": 0.43547279668232125,
+      "epoch": 0.44994324631101024,
       "grad_norm": 0.0,
-      "learning_rate": 1.2547641884364226e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.2091880848820536e-05,
+      "loss": 0.8429,
       "step": 15856
     },
     {
-      "epoch": 0.4355002609101645,
+      "epoch": 0.4499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.254678170944672e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.2090982096130686e-05,
+      "loss": 0.8913,
       "step": 15857
     },
     {
-      "epoch": 0.43552772513800775,
+      "epoch": 0.45,
       "grad_norm": 0.0,
-      "learning_rate": 1.2545921514378091e-05,
-      "loss": 0.8117,
+      "learning_rate": 1.2090083325778214e-05,
+      "loss": 0.8661,
       "step": 15858
     },
     {
-      "epoch": 0.43555518936585097,
+      "epoch": 0.45002837684449487,
       "grad_norm": 0.0,
-      "learning_rate": 1.2545061299165147e-05,
-      "loss": 0.889,
+      "learning_rate": 1.2089184537770717e-05,
+      "loss": 0.8474,
       "step": 15859
     },
     {
-      "epoch": 0.4355826535936942,
+      "epoch": 0.4500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.2544201063814692e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.208828573211578e-05,
+      "loss": 0.8646,
       "step": 15860
     },
     {
-      "epoch": 0.43561011782153747,
+      "epoch": 0.4500851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.254334080833353e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.2087386908821005e-05,
+      "loss": 0.9217,
       "step": 15861
     },
     {
-      "epoch": 0.4356375820493807,
+      "epoch": 0.45011350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 1.2542480532728471e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.2086488067893976e-05,
+      "loss": 1.0156,
       "step": 15862
     },
     {
-      "epoch": 0.4356650462772239,
+      "epoch": 0.4501418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.2541620237006324e-05,
-      "loss": 0.9607,
+      "learning_rate": 1.2085589209342287e-05,
+      "loss": 0.8471,
       "step": 15863
     },
     {
-      "epoch": 0.43569251050506713,
+      "epoch": 0.45017026106696933,
       "grad_norm": 0.0,
-      "learning_rate": 1.254075992117389e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.2084690333173536e-05,
+      "loss": 0.8874,
       "step": 15864
     },
     {
-      "epoch": 0.4357199747329104,
+      "epoch": 0.45019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.2539899585237983e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.2083791439395309e-05,
+      "loss": 0.919,
       "step": 15865
     },
     {
-      "epoch": 0.43574743896075363,
+      "epoch": 0.45022701475595917,
       "grad_norm": 0.0,
-      "learning_rate": 1.2539039229205403e-05,
-      "loss": 0.8374,
+      "learning_rate": 1.2082892528015205e-05,
+      "loss": 0.8554,
       "step": 15866
     },
     {
-      "epoch": 0.43577490318859685,
+      "epoch": 0.450255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.2538178853082963e-05,
-      "loss": 0.8626,
+      "learning_rate": 1.208199359904081e-05,
+      "loss": 0.9309,
       "step": 15867
     },
     {
-      "epoch": 0.4358023674164401,
+      "epoch": 0.45028376844494894,
       "grad_norm": 0.0,
-      "learning_rate": 1.2537318456877468e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.2081094652479725e-05,
+      "loss": 0.8757,
       "step": 15868
     },
     {
-      "epoch": 0.4358298316442833,
+      "epoch": 0.4503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.2536458040595725e-05,
-      "loss": 0.8079,
+      "learning_rate": 1.2080195688339537e-05,
+      "loss": 0.975,
       "step": 15869
     },
     {
-      "epoch": 0.43585729587212657,
+      "epoch": 0.4503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.2535597604244542e-05,
-      "loss": 1.0115,
+      "learning_rate": 1.2079296706627845e-05,
+      "loss": 0.9376,
       "step": 15870
     },
     {
-      "epoch": 0.4358847600999698,
+      "epoch": 0.4503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.2534737147830734e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.207839770735224e-05,
+      "loss": 0.8573,
       "step": 15871
     },
     {
-      "epoch": 0.435912224327813,
+      "epoch": 0.4503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.2533876671361102e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.2077498690520316e-05,
+      "loss": 0.9714,
       "step": 15872
     },
     {
-      "epoch": 0.43593968855565624,
+      "epoch": 0.4504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.2533016174842454e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.2076599656139667e-05,
+      "loss": 0.9758,
       "step": 15873
     },
     {
-      "epoch": 0.4359671527834995,
+      "epoch": 0.45045402951191826,
       "grad_norm": 0.0,
-      "learning_rate": 1.2532155658281605e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.2075700604217889e-05,
+      "loss": 0.8718,
       "step": 15874
     },
     {
-      "epoch": 0.43599461701134273,
+      "epoch": 0.4504824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.2531295121685356e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.207480153476257e-05,
+      "loss": 0.7808,
       "step": 15875
     },
     {
-      "epoch": 0.43602208123918595,
+      "epoch": 0.45051078320090804,
       "grad_norm": 0.0,
-      "learning_rate": 1.2530434565060522e-05,
-      "loss": 1.0129,
+      "learning_rate": 1.2073902447781311e-05,
+      "loss": 0.9376,
       "step": 15876
     },
     {
-      "epoch": 0.4360495454670292,
+      "epoch": 0.45053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.2529573988413906e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.2073003343281709e-05,
+      "loss": 0.7451,
       "step": 15877
     },
     {
-      "epoch": 0.43607700969487245,
+      "epoch": 0.45056753688989787,
       "grad_norm": 0.0,
-      "learning_rate": 1.2528713391752323e-05,
-      "loss": 0.925,
+      "learning_rate": 1.207210422127135e-05,
+      "loss": 0.8898,
       "step": 15878
     },
     {
-      "epoch": 0.4361044739227157,
+      "epoch": 0.45059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.252785277508258e-05,
-      "loss": 0.9236,
+      "learning_rate": 1.2071205081757835e-05,
+      "loss": 0.9722,
       "step": 15879
     },
     {
-      "epoch": 0.4361319381505589,
+      "epoch": 0.45062429057888764,
       "grad_norm": 0.0,
-      "learning_rate": 1.2526992138411487e-05,
-      "loss": 0.8472,
+      "learning_rate": 1.2070305924748758e-05,
+      "loss": 0.8156,
       "step": 15880
     },
     {
-      "epoch": 0.4361594023784021,
+      "epoch": 0.4506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.2526131481745852e-05,
-      "loss": 0.8326,
+      "learning_rate": 1.2069406750251713e-05,
+      "loss": 0.9147,
       "step": 15881
     },
     {
-      "epoch": 0.43618686660624534,
+      "epoch": 0.4506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.2525270805092485e-05,
-      "loss": 0.981,
+      "learning_rate": 1.2068507558274297e-05,
+      "loss": 0.9131,
       "step": 15882
     },
     {
-      "epoch": 0.4362143308340886,
+      "epoch": 0.4507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.25244101084582e-05,
-      "loss": 0.9168,
+      "learning_rate": 1.2067608348824102e-05,
+      "loss": 0.9069,
       "step": 15883
     },
     {
-      "epoch": 0.43624179506193184,
+      "epoch": 0.4507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.2523549391849805e-05,
-      "loss": 0.8926,
+      "learning_rate": 1.2066709121908731e-05,
+      "loss": 0.9704,
       "step": 15884
     },
     {
-      "epoch": 0.43626925928977506,
+      "epoch": 0.4507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.2522688655274107e-05,
-      "loss": 0.8546,
+      "learning_rate": 1.206580987753577e-05,
+      "loss": 0.9384,
       "step": 15885
     },
     {
-      "epoch": 0.4362967235176183,
+      "epoch": 0.45079455164585697,
       "grad_norm": 0.0,
-      "learning_rate": 1.2521827898737922e-05,
-      "loss": 0.922,
+      "learning_rate": 1.2064910615712824e-05,
+      "loss": 0.9741,
       "step": 15886
     },
     {
-      "epoch": 0.43632418774546156,
+      "epoch": 0.4508229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.2520967122248056e-05,
-      "loss": 0.9607,
+      "learning_rate": 1.2064011336447484e-05,
+      "loss": 0.9296,
       "step": 15887
     },
     {
-      "epoch": 0.4363516519733048,
+      "epoch": 0.45085130533484674,
       "grad_norm": 0.0,
-      "learning_rate": 1.252010632581132e-05,
-      "loss": 1.0444,
+      "learning_rate": 1.2063112039747347e-05,
+      "loss": 1.0309,
       "step": 15888
     },
     {
-      "epoch": 0.436379116201148,
+      "epoch": 0.45087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.251924550943453e-05,
-      "loss": 0.8135,
+      "learning_rate": 1.2062212725620011e-05,
+      "loss": 0.8687,
       "step": 15889
     },
     {
-      "epoch": 0.4364065804289912,
+      "epoch": 0.45090805902383657,
       "grad_norm": 0.0,
-      "learning_rate": 1.2518384673124491e-05,
-      "loss": 0.905,
+      "learning_rate": 1.206131339407307e-05,
+      "loss": 0.9332,
       "step": 15890
     },
     {
-      "epoch": 0.4364340446568345,
+      "epoch": 0.45093643586833143,
       "grad_norm": 0.0,
-      "learning_rate": 1.2517523816888018e-05,
-      "loss": 0.9181,
+      "learning_rate": 1.2060414045114122e-05,
+      "loss": 0.8056,
       "step": 15891
     },
     {
-      "epoch": 0.4364615088846777,
+      "epoch": 0.45096481271282635,
       "grad_norm": 0.0,
-      "learning_rate": 1.2516662940731923e-05,
-      "loss": 0.8193,
+      "learning_rate": 1.2059514678750764e-05,
+      "loss": 0.9719,
       "step": 15892
     },
     {
-      "epoch": 0.43648897311252094,
+      "epoch": 0.4509931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.2515802044663012e-05,
-      "loss": 0.9722,
+      "learning_rate": 1.2058615294990594e-05,
+      "loss": 0.9939,
       "step": 15893
     },
     {
-      "epoch": 0.43651643734036416,
+      "epoch": 0.4510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.2514941128688105e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.2057715893841208e-05,
+      "loss": 0.9445,
       "step": 15894
     },
     {
-      "epoch": 0.43654390156820744,
+      "epoch": 0.45104994324631104,
       "grad_norm": 0.0,
-      "learning_rate": 1.2514080192814006e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.2056816475310202e-05,
+      "loss": 0.9797,
       "step": 15895
     },
     {
-      "epoch": 0.43657136579605066,
+      "epoch": 0.4510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.2513219237047531e-05,
-      "loss": 0.8951,
+      "learning_rate": 1.2055917039405176e-05,
+      "loss": 0.8671,
       "step": 15896
     },
     {
-      "epoch": 0.4365988300238939,
+      "epoch": 0.4511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.2512358261395497e-05,
-      "loss": 0.948,
+      "learning_rate": 1.2055017586133729e-05,
+      "loss": 0.9977,
       "step": 15897
     },
     {
-      "epoch": 0.4366262942517371,
+      "epoch": 0.45113507377979567,
       "grad_norm": 0.0,
-      "learning_rate": 1.2511497265864705e-05,
-      "loss": 0.9851,
+      "learning_rate": 1.205411811550345e-05,
+      "loss": 0.9643,
       "step": 15898
     },
     {
-      "epoch": 0.4366537584795803,
+      "epoch": 0.4511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.2510636250461976e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.2053218627521946e-05,
+      "loss": 0.9036,
       "step": 15899
     },
     {
-      "epoch": 0.4366812227074236,
+      "epoch": 0.45119182746878544,
       "grad_norm": 0.0,
-      "learning_rate": 1.2509775215194123e-05,
-      "loss": 0.923,
+      "learning_rate": 1.2052319122196812e-05,
+      "loss": 0.9579,
       "step": 15900
     },
     {
-      "epoch": 0.4367086869352668,
+      "epoch": 0.45122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.2508914160067951e-05,
-      "loss": 0.9003,
+      "learning_rate": 1.2051419599535648e-05,
+      "loss": 0.8593,
       "step": 15901
     },
     {
-      "epoch": 0.43673615116311004,
+      "epoch": 0.4512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.250805308509028e-05,
-      "loss": 0.9405,
+      "learning_rate": 1.2050520059546048e-05,
+      "loss": 1.0412,
       "step": 15902
     },
     {
-      "epoch": 0.43676361539095326,
+      "epoch": 0.45127695800227013,
       "grad_norm": 0.0,
-      "learning_rate": 1.2507191990267924e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.2049620502235613e-05,
+      "loss": 0.8816,
       "step": 15903
     },
     {
-      "epoch": 0.43679107961879654,
+      "epoch": 0.45130533484676505,
       "grad_norm": 0.0,
-      "learning_rate": 1.2506330875607693e-05,
-      "loss": 0.9585,
+      "learning_rate": 1.2048720927611944e-05,
+      "loss": 0.962,
       "step": 15904
     },
     {
-      "epoch": 0.43681854384663976,
+      "epoch": 0.4513337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.2505469741116398e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.2047821335682632e-05,
+      "loss": 0.8881,
       "step": 15905
     },
     {
-      "epoch": 0.436846008074483,
+      "epoch": 0.4513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.2504608586800862e-05,
-      "loss": 1.0005,
+      "learning_rate": 1.2046921726455284e-05,
+      "loss": 0.8184,
       "step": 15906
     },
     {
-      "epoch": 0.4368734723023262,
+      "epoch": 0.45139046538024974,
       "grad_norm": 0.0,
-      "learning_rate": 1.2503747412667888e-05,
-      "loss": 0.9246,
+      "learning_rate": 1.2046022099937497e-05,
+      "loss": 0.8462,
       "step": 15907
     },
     {
-      "epoch": 0.4369009365301695,
+      "epoch": 0.4514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.2502886218724295e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.204512245613687e-05,
+      "loss": 0.8817,
       "step": 15908
     },
     {
-      "epoch": 0.4369284007580127,
+      "epoch": 0.4514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.25020250049769e-05,
-      "loss": 0.9376,
+      "learning_rate": 1.2044222795060999e-05,
+      "loss": 0.8341,
       "step": 15909
     },
     {
-      "epoch": 0.4369558649858559,
+      "epoch": 0.4514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.250116377143251e-05,
-      "loss": 0.82,
+      "learning_rate": 1.204332311671749e-05,
+      "loss": 0.9329,
       "step": 15910
     },
     {
-      "epoch": 0.43698332921369915,
+      "epoch": 0.4515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.2500302518097945e-05,
-      "loss": 0.8662,
+      "learning_rate": 1.2042423421113936e-05,
+      "loss": 0.9875,
       "step": 15911
     },
     {
-      "epoch": 0.43701079344154237,
+      "epoch": 0.4515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.2499441244980019e-05,
-      "loss": 0.934,
+      "learning_rate": 1.2041523708257938e-05,
+      "loss": 0.9535,
       "step": 15912
     },
     {
-      "epoch": 0.43703825766938564,
+      "epoch": 0.45156072644721906,
       "grad_norm": 0.0,
-      "learning_rate": 1.2498579952085543e-05,
-      "loss": 0.8759,
+      "learning_rate": 1.2040623978157097e-05,
+      "loss": 0.8139,
       "step": 15913
     },
     {
-      "epoch": 0.43706572189722886,
+      "epoch": 0.451589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.2497718639421335e-05,
-      "loss": 1.0195,
+      "learning_rate": 1.2039724230819019e-05,
+      "loss": 0.8412,
       "step": 15914
     },
     {
-      "epoch": 0.4370931861250721,
+      "epoch": 0.45161748013620884,
       "grad_norm": 0.0,
-      "learning_rate": 1.2496857306994213e-05,
-      "loss": 0.8828,
+      "learning_rate": 1.2038824466251295e-05,
+      "loss": 0.897,
       "step": 15915
     },
     {
-      "epoch": 0.4371206503529153,
+      "epoch": 0.45164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.2495995954810986e-05,
-      "loss": 0.8824,
+      "learning_rate": 1.203792468446153e-05,
+      "loss": 0.9217,
       "step": 15916
     },
     {
-      "epoch": 0.4371481145807586,
+      "epoch": 0.4516742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.2495134582878472e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.2037024885457325e-05,
+      "loss": 0.956,
       "step": 15917
     },
     {
-      "epoch": 0.4371755788086018,
+      "epoch": 0.4517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.2494273191203488e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.2036125069246278e-05,
+      "loss": 0.8284,
       "step": 15918
     },
     {
-      "epoch": 0.437203043036445,
+      "epoch": 0.45173098751418844,
       "grad_norm": 0.0,
-      "learning_rate": 1.2493411779792848e-05,
-      "loss": 0.9071,
+      "learning_rate": 1.2035225235835988e-05,
+      "loss": 0.9545,
       "step": 15919
     },
     {
-      "epoch": 0.43723050726428825,
+      "epoch": 0.4517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.2492550348653367e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.2034325385234062e-05,
+      "loss": 0.9724,
       "step": 15920
     },
     {
-      "epoch": 0.4372579714921315,
+      "epoch": 0.4517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.2491688897791862e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.2033425517448097e-05,
+      "loss": 0.8305,
       "step": 15921
     },
     {
-      "epoch": 0.43728543571997475,
+      "epoch": 0.4518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.2490827427215151e-05,
-      "loss": 0.87,
+      "learning_rate": 1.2032525632485696e-05,
+      "loss": 0.9355,
       "step": 15922
     },
     {
-      "epoch": 0.43731289994781797,
+      "epoch": 0.451844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.2489965936930046e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.2031625730354459e-05,
+      "loss": 0.9618,
       "step": 15923
     },
     {
-      "epoch": 0.4373403641756612,
+      "epoch": 0.4518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.248910442694337e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.2030725811061989e-05,
+      "loss": 0.9107,
       "step": 15924
     },
     {
-      "epoch": 0.4373678284035044,
+      "epoch": 0.45190124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.248824289726193e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.2029825874615889e-05,
+      "loss": 0.7849,
       "step": 15925
     },
     {
-      "epoch": 0.4373952926313477,
+      "epoch": 0.4519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.2487381347892554e-05,
-      "loss": 0.8751,
+      "learning_rate": 1.2028925921023753e-05,
+      "loss": 0.861,
       "step": 15926
     },
     {
-      "epoch": 0.4374227568591909,
+      "epoch": 0.45195800227014754,
       "grad_norm": 0.0,
-      "learning_rate": 1.248651977884205e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.2028025950293191e-05,
+      "loss": 0.8532,
       "step": 15927
     },
     {
-      "epoch": 0.43745022108703413,
+      "epoch": 0.45198637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.248565819011724e-05,
-      "loss": 0.9899,
+      "learning_rate": 1.2027125962431805e-05,
+      "loss": 0.9748,
       "step": 15928
     },
     {
-      "epoch": 0.43747768531487735,
+      "epoch": 0.45201475595913737,
       "grad_norm": 0.0,
-      "learning_rate": 1.2484796581724935e-05,
-      "loss": 1.0157,
+      "learning_rate": 1.202622595744719e-05,
+      "loss": 0.8682,
       "step": 15929
     },
     {
-      "epoch": 0.43750514954272063,
+      "epoch": 0.45204313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 1.2483934953671963e-05,
-      "loss": 0.8754,
+      "learning_rate": 1.2025325935346957e-05,
+      "loss": 0.8712,
       "step": 15930
     },
     {
-      "epoch": 0.43753261377056385,
+      "epoch": 0.45207150964812715,
       "grad_norm": 0.0,
-      "learning_rate": 1.2483073305965135e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.2024425896138705e-05,
+      "loss": 0.9247,
       "step": 15931
     },
     {
-      "epoch": 0.43756007799840707,
+      "epoch": 0.452099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.2482211638611265e-05,
-      "loss": 0.8794,
+      "learning_rate": 1.2023525839830038e-05,
+      "loss": 0.9291,
       "step": 15932
     },
     {
-      "epoch": 0.4375875422262503,
+      "epoch": 0.4521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.2481349951617179e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.202262576642855e-05,
+      "loss": 0.7158,
       "step": 15933
     },
     {
-      "epoch": 0.43761500645409357,
+      "epoch": 0.4521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.2480488244989688e-05,
-      "loss": 0.8879,
+      "learning_rate": 1.2021725675941857e-05,
+      "loss": 0.9507,
       "step": 15934
     },
     {
-      "epoch": 0.4376424706819368,
+      "epoch": 0.4521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.2479626518735614e-05,
-      "loss": 0.947,
+      "learning_rate": 1.2020825568377556e-05,
+      "loss": 0.9578,
       "step": 15935
     },
     {
-      "epoch": 0.43766993490978,
+      "epoch": 0.4522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.2478764772861775e-05,
-      "loss": 0.89,
+      "learning_rate": 1.2019925443743248e-05,
+      "loss": 0.9444,
       "step": 15936
     },
     {
-      "epoch": 0.43769739913762323,
+      "epoch": 0.45224177071509647,
       "grad_norm": 0.0,
-      "learning_rate": 1.247790300737499e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.2019025302046541e-05,
+      "loss": 0.9034,
       "step": 15937
     },
     {
-      "epoch": 0.43772486336546645,
+      "epoch": 0.4522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.2477041222282074e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.2018125143295038e-05,
+      "loss": 1.0527,
       "step": 15938
     },
     {
-      "epoch": 0.43775232759330973,
+      "epoch": 0.45229852440408624,
       "grad_norm": 0.0,
-      "learning_rate": 1.247617941758985e-05,
-      "loss": 0.9201,
+      "learning_rate": 1.2017224967496336e-05,
+      "loss": 0.888,
       "step": 15939
     },
     {
-      "epoch": 0.43777979182115295,
+      "epoch": 0.45232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.2475317593305136e-05,
-      "loss": 0.9059,
+      "learning_rate": 1.201632477465805e-05,
+      "loss": 0.8708,
       "step": 15940
     },
     {
-      "epoch": 0.4378072560489962,
+      "epoch": 0.4523552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 1.2474455749434749e-05,
-      "loss": 0.8826,
+      "learning_rate": 1.2015424564787777e-05,
+      "loss": 0.9026,
       "step": 15941
     },
     {
-      "epoch": 0.4378347202768394,
+      "epoch": 0.45238365493757093,
       "grad_norm": 0.0,
-      "learning_rate": 1.2473593885985508e-05,
-      "loss": 0.8628,
+      "learning_rate": 1.2014524337893119e-05,
+      "loss": 0.8513,
       "step": 15942
     },
     {
-      "epoch": 0.43786218450468267,
+      "epoch": 0.45241203178206585,
       "grad_norm": 0.0,
-      "learning_rate": 1.2472732002964238e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.2013624093981685e-05,
+      "loss": 0.812,
       "step": 15943
     },
     {
-      "epoch": 0.4378896487325259,
+      "epoch": 0.4524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.2471870100377752e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.2012723833061077e-05,
+      "loss": 0.8379,
       "step": 15944
     },
     {
-      "epoch": 0.4379171129603691,
+      "epoch": 0.4524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.2471008178232873e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.2011823555138902e-05,
+      "loss": 1.0161,
       "step": 15945
     },
     {
-      "epoch": 0.43794457718821234,
+      "epoch": 0.45249716231555054,
       "grad_norm": 0.0,
-      "learning_rate": 1.2470146236536417e-05,
-      "loss": 0.8854,
+      "learning_rate": 1.201092326022276e-05,
+      "loss": 0.8635,
       "step": 15946
     },
     {
-      "epoch": 0.4379720414160556,
+      "epoch": 0.4525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.2469284275295212e-05,
-      "loss": 0.8092,
+      "learning_rate": 1.2010022948320263e-05,
+      "loss": 0.9093,
       "step": 15947
     },
     {
-      "epoch": 0.43799950564389883,
+      "epoch": 0.4525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.2468422294516072e-05,
-      "loss": 0.8219,
+      "learning_rate": 1.2009122619439011e-05,
+      "loss": 0.9259,
       "step": 15948
     },
     {
-      "epoch": 0.43802696987174206,
+      "epoch": 0.4525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.2467560294205816e-05,
-      "loss": 0.8533,
+      "learning_rate": 1.2008222273586607e-05,
+      "loss": 0.8402,
       "step": 15949
     },
     {
-      "epoch": 0.4380544340995853,
+      "epoch": 0.4526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.2466698274371271e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.2007321910770662e-05,
+      "loss": 0.8945,
       "step": 15950
     },
     {
-      "epoch": 0.4380818983274285,
+      "epoch": 0.45263904653802495,
       "grad_norm": 0.0,
-      "learning_rate": 1.2465836235019251e-05,
-      "loss": 0.8737,
+      "learning_rate": 1.2006421530998778e-05,
+      "loss": 1.0536,
       "step": 15951
     },
     {
-      "epoch": 0.4381093625552718,
+      "epoch": 0.45266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.246497417615658e-05,
-      "loss": 0.917,
+      "learning_rate": 1.2005521134278559e-05,
+      "loss": 0.9499,
       "step": 15952
     },
     {
-      "epoch": 0.438136826783115,
+      "epoch": 0.4526958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.246411209779008e-05,
-      "loss": 0.8246,
+      "learning_rate": 1.2004620720617613e-05,
+      "loss": 0.9075,
       "step": 15953
     },
     {
-      "epoch": 0.4381642910109582,
+      "epoch": 0.45272417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.2463249999926567e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.2003720290023549e-05,
+      "loss": 0.8393,
       "step": 15954
     },
     {
-      "epoch": 0.43819175523880144,
+      "epoch": 0.45275255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 1.246238788257287e-05,
-      "loss": 0.9387,
+      "learning_rate": 1.2002819842503967e-05,
+      "loss": 0.9651,
       "step": 15955
     },
     {
-      "epoch": 0.4382192194666447,
+      "epoch": 0.4527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.2461525745735805e-05,
-      "loss": 0.8192,
+      "learning_rate": 1.2001919378066476e-05,
+      "loss": 0.8716,
       "step": 15956
     },
     {
-      "epoch": 0.43824668369448794,
+      "epoch": 0.4528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.2460663589422192e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.2001018896718684e-05,
+      "loss": 0.8092,
       "step": 15957
     },
     {
-      "epoch": 0.43827414792233116,
+      "epoch": 0.45283768444948924,
       "grad_norm": 0.0,
-      "learning_rate": 1.2459801413638858e-05,
-      "loss": 0.8564,
+      "learning_rate": 1.2000118398468193e-05,
+      "loss": 0.8941,
       "step": 15958
     },
     {
-      "epoch": 0.4383016121501744,
+      "epoch": 0.4528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.2458939218392623e-05,
-      "loss": 0.7669,
+      "learning_rate": 1.1999217883322612e-05,
+      "loss": 0.9238,
       "step": 15959
     },
     {
-      "epoch": 0.43832907637801766,
+      "epoch": 0.452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.2458077003690308e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.1998317351289547e-05,
+      "loss": 0.8379,
       "step": 15960
     },
     {
-      "epoch": 0.4383565406058609,
+      "epoch": 0.4529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.2457214769538735e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.1997416802376606e-05,
+      "loss": 0.9184,
       "step": 15961
     },
     {
-      "epoch": 0.4383840048337041,
+      "epoch": 0.4529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.2456352515944728e-05,
-      "loss": 0.9707,
+      "learning_rate": 1.1996516236591398e-05,
+      "loss": 0.8425,
       "step": 15962
     },
     {
-      "epoch": 0.4384114690615473,
+      "epoch": 0.45297956867196365,
       "grad_norm": 0.0,
-      "learning_rate": 1.2455490242915104e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.1995615653941526e-05,
+      "loss": 0.827,
       "step": 15963
     },
     {
-      "epoch": 0.43843893328939054,
+      "epoch": 0.45300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.2454627950456695e-05,
-      "loss": 0.9884,
+      "learning_rate": 1.1994715054434601e-05,
+      "loss": 0.9575,
       "step": 15964
     },
     {
-      "epoch": 0.4384663975172338,
+      "epoch": 0.4530363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.2453765638576318e-05,
-      "loss": 0.8892,
+      "learning_rate": 1.1993814438078226e-05,
+      "loss": 0.8219,
       "step": 15965
     },
     {
-      "epoch": 0.43849386174507704,
+      "epoch": 0.45306469920544834,
       "grad_norm": 0.0,
-      "learning_rate": 1.2452903307280794e-05,
-      "loss": 0.9835,
+      "learning_rate": 1.199291380488001e-05,
+      "loss": 0.9581,
       "step": 15966
     },
     {
-      "epoch": 0.43852132597292026,
+      "epoch": 0.45309307604994326,
       "grad_norm": 0.0,
-      "learning_rate": 1.2452040956576953e-05,
-      "loss": 1.0114,
+      "learning_rate": 1.1992013154847559e-05,
+      "loss": 0.9182,
       "step": 15967
     },
     {
-      "epoch": 0.4385487902007635,
+      "epoch": 0.4531214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.2451178586471613e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.199111248798849e-05,
+      "loss": 0.9499,
       "step": 15968
     },
     {
-      "epoch": 0.43857625442860676,
+      "epoch": 0.45314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.2450316196971597e-05,
-      "loss": 0.8875,
+      "learning_rate": 1.19902118043104e-05,
+      "loss": 0.8375,
       "step": 15969
     },
     {
-      "epoch": 0.43860371865645,
+      "epoch": 0.45317820658342794,
       "grad_norm": 0.0,
-      "learning_rate": 1.2449453788083728e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.1989311103820902e-05,
+      "loss": 0.8916,
       "step": 15970
     },
     {
-      "epoch": 0.4386311828842932,
+      "epoch": 0.4532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.2448591359814837e-05,
-      "loss": 0.8328,
+      "learning_rate": 1.1988410386527605e-05,
+      "loss": 0.7809,
       "step": 15971
     },
     {
-      "epoch": 0.4386586471121364,
+      "epoch": 0.4532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.2447728912171739e-05,
-      "loss": 0.953,
+      "learning_rate": 1.1987509652438117e-05,
+      "loss": 0.9986,
       "step": 15972
     },
     {
-      "epoch": 0.4386861113399797,
+      "epoch": 0.4532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.2446866445161264e-05,
-      "loss": 0.944,
+      "learning_rate": 1.1986608901560041e-05,
+      "loss": 0.8785,
       "step": 15973
     },
     {
-      "epoch": 0.4387135755678229,
+      "epoch": 0.4532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.2446003958790231e-05,
-      "loss": 0.9694,
+      "learning_rate": 1.1985708133900996e-05,
+      "loss": 1.0249,
       "step": 15974
     },
     {
-      "epoch": 0.43874103979566614,
+      "epoch": 0.4533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.2445141453065468e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.198480734946858e-05,
+      "loss": 0.9552,
       "step": 15975
     },
     {
-      "epoch": 0.43876850402350936,
+      "epoch": 0.45334846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 1.24442789279938e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.198390654827041e-05,
+      "loss": 0.8918,
       "step": 15976
     },
     {
-      "epoch": 0.4387959682513526,
+      "epoch": 0.4533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.2443416383582052e-05,
-      "loss": 0.8797,
+      "learning_rate": 1.198300573031409e-05,
+      "loss": 0.9622,
       "step": 15977
     },
     {
-      "epoch": 0.43882343247919586,
+      "epoch": 0.45340522133938704,
       "grad_norm": 0.0,
-      "learning_rate": 1.2442553819837042e-05,
-      "loss": 0.874,
+      "learning_rate": 1.1982104895607237e-05,
+      "loss": 0.9718,
       "step": 15978
     },
     {
-      "epoch": 0.4388508967070391,
+      "epoch": 0.45343359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.2441691236765606e-05,
-      "loss": 0.9759,
+      "learning_rate": 1.1981204044157451e-05,
+      "loss": 0.8998,
       "step": 15979
     },
     {
-      "epoch": 0.4388783609348823,
+      "epoch": 0.4534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.244082863437456e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.1980303175972343e-05,
+      "loss": 0.865,
       "step": 15980
     },
     {
-      "epoch": 0.4389058251627255,
+      "epoch": 0.45349035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 1.243996601267073e-05,
-      "loss": 0.8017,
+      "learning_rate": 1.1979402291059528e-05,
+      "loss": 0.8443,
       "step": 15981
     },
     {
-      "epoch": 0.4389332893905688,
+      "epoch": 0.45351872871736665,
       "grad_norm": 0.0,
-      "learning_rate": 1.2439103371660946e-05,
-      "loss": 0.9505,
+      "learning_rate": 1.1978501389426611e-05,
+      "loss": 0.7745,
       "step": 15982
     },
     {
-      "epoch": 0.438960753618412,
+      "epoch": 0.4535471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.2438240711352032e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.1977600471081203e-05,
+      "loss": 0.9978,
       "step": 15983
     },
     {
-      "epoch": 0.43898821784625525,
+      "epoch": 0.4535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.2437378031750814e-05,
-      "loss": 0.9868,
+      "learning_rate": 1.1976699536030918e-05,
+      "loss": 0.9158,
       "step": 15984
     },
     {
-      "epoch": 0.43901568207409847,
+      "epoch": 0.4536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.2436515332864114e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.1975798584283362e-05,
+      "loss": 0.8642,
       "step": 15985
     },
     {
-      "epoch": 0.43904314630194174,
+      "epoch": 0.4536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.2435652614698763e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.1974897615846148e-05,
+      "loss": 1.0098,
       "step": 15986
     },
     {
-      "epoch": 0.43907061052978497,
+      "epoch": 0.4536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.2434789877261584e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.1973996630726882e-05,
+      "loss": 1.0734,
       "step": 15987
     },
     {
-      "epoch": 0.4390980747576282,
+      "epoch": 0.45368898978433597,
       "grad_norm": 0.0,
-      "learning_rate": 1.2433927120559404e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.1973095628933182e-05,
+      "loss": 0.9185,
       "step": 15988
     },
     {
-      "epoch": 0.4391255389854714,
+      "epoch": 0.4537173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.243306434459905e-05,
-      "loss": 0.8281,
+      "learning_rate": 1.1972194610472655e-05,
+      "loss": 0.9017,
       "step": 15989
     },
     {
-      "epoch": 0.43915300321331463,
+      "epoch": 0.45374574347332575,
       "grad_norm": 0.0,
-      "learning_rate": 1.2432201549387349e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.1971293575352902e-05,
+      "loss": 0.8118,
       "step": 15990
     },
     {
-      "epoch": 0.4391804674411579,
+      "epoch": 0.45377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.2431338734931126e-05,
-      "loss": 0.894,
+      "learning_rate": 1.197039252358155e-05,
+      "loss": 0.8166,
       "step": 15991
     },
     {
-      "epoch": 0.43920793166900113,
+      "epoch": 0.4538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.2430475901237212e-05,
-      "loss": 0.8683,
+      "learning_rate": 1.1969491455166207e-05,
+      "loss": 0.8795,
       "step": 15992
     },
     {
-      "epoch": 0.43923539589684435,
+      "epoch": 0.45383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.2429613048312428e-05,
-      "loss": 0.9811,
+      "learning_rate": 1.1968590370114477e-05,
+      "loss": 0.9559,
       "step": 15993
     },
     {
-      "epoch": 0.43926286012468757,
+      "epoch": 0.45385925085130535,
       "grad_norm": 0.0,
-      "learning_rate": 1.2428750176163604e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.1967689268433978e-05,
+      "loss": 0.9036,
       "step": 15994
     },
     {
-      "epoch": 0.43929032435253085,
+      "epoch": 0.4538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.242788728479757e-05,
-      "loss": 0.875,
+      "learning_rate": 1.1966788150132318e-05,
+      "loss": 0.8484,
       "step": 15995
     },
     {
-      "epoch": 0.43931778858037407,
+      "epoch": 0.4539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.2427024374221152e-05,
-      "loss": 0.886,
+      "learning_rate": 1.1965887015217107e-05,
+      "loss": 0.9054,
       "step": 15996
     },
     {
-      "epoch": 0.4393452528082173,
+      "epoch": 0.45394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.2426161444441175e-05,
-      "loss": 0.8853,
+      "learning_rate": 1.1964985863695966e-05,
+      "loss": 0.91,
       "step": 15997
     },
     {
-      "epoch": 0.4393727170360605,
+      "epoch": 0.4539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.2425298495464471e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.1964084695576496e-05,
+      "loss": 0.877,
       "step": 15998
     },
     {
-      "epoch": 0.4394001812639038,
+      "epoch": 0.4540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.2424435527297864e-05,
-      "loss": 0.9319,
+      "learning_rate": 1.1963183510866317e-05,
+      "loss": 0.9429,
       "step": 15999
     },
     {
-      "epoch": 0.439427645491747,
+      "epoch": 0.4540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.2423572539948183e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.1962282309573037e-05,
+      "loss": 0.8823,
       "step": 16000
     },
     {
-      "epoch": 0.43945510971959023,
+      "epoch": 0.4540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.2422709533422261e-05,
-      "loss": 0.989,
+      "learning_rate": 1.1961381091704271e-05,
+      "loss": 0.9221,
       "step": 16001
     },
     {
-      "epoch": 0.43948257394743345,
+      "epoch": 0.45408626560726445,
       "grad_norm": 0.0,
-      "learning_rate": 1.2421846507726918e-05,
-      "loss": 0.9495,
+      "learning_rate": 1.1960479857267632e-05,
+      "loss": 0.9001,
       "step": 16002
     },
     {
-      "epoch": 0.43951003817527673,
+      "epoch": 0.45411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.2420983462868991e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.195957860627073e-05,
+      "loss": 0.9371,
       "step": 16003
     },
     {
-      "epoch": 0.43953750240311995,
+      "epoch": 0.4541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.2420120398855304e-05,
-      "loss": 0.8677,
+      "learning_rate": 1.1958677338721181e-05,
+      "loss": 0.9423,
       "step": 16004
     },
     {
-      "epoch": 0.43956496663096317,
+      "epoch": 0.45417139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 1.2419257315692685e-05,
-      "loss": 0.8061,
+      "learning_rate": 1.1957776054626593e-05,
+      "loss": 0.7964,
       "step": 16005
     },
     {
-      "epoch": 0.4395924308588064,
+      "epoch": 0.45419977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 1.2418394213387967e-05,
-      "loss": 0.8248,
+      "learning_rate": 1.1956874753994585e-05,
+      "loss": 0.8996,
       "step": 16006
     },
     {
-      "epoch": 0.4396198950866496,
+      "epoch": 0.4542281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.2417531091947978e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.1955973436832768e-05,
+      "loss": 0.8883,
       "step": 16007
     },
     {
-      "epoch": 0.4396473593144929,
+      "epoch": 0.45425652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.2416667951379547e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.1955072103148755e-05,
+      "loss": 0.9449,
       "step": 16008
     },
     {
-      "epoch": 0.4396748235423361,
+      "epoch": 0.45428490351872874,
       "grad_norm": 0.0,
-      "learning_rate": 1.2415804791689498e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.1954170752950163e-05,
+      "loss": 0.8698,
       "step": 16009
     },
     {
-      "epoch": 0.43970228777017933,
+      "epoch": 0.4543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.2414941612884673e-05,
-      "loss": 0.8604,
+      "learning_rate": 1.1953269386244598e-05,
+      "loss": 0.975,
       "step": 16010
     },
     {
-      "epoch": 0.43972975199802256,
+      "epoch": 0.4543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.241407841497189e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.1952368003039681e-05,
+      "loss": 0.808,
       "step": 16011
     },
     {
-      "epoch": 0.43975721622586583,
+      "epoch": 0.4543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.2413215197957986e-05,
-      "loss": 0.8811,
+      "learning_rate": 1.1951466603343025e-05,
+      "loss": 0.8901,
       "step": 16012
     },
     {
-      "epoch": 0.43978468045370905,
+      "epoch": 0.4543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.2412351961849784e-05,
-      "loss": 1.0165,
+      "learning_rate": 1.195056518716224e-05,
+      "loss": 0.8163,
       "step": 16013
     },
     {
-      "epoch": 0.4398121446815523,
+      "epoch": 0.45442678774120315,
       "grad_norm": 0.0,
-      "learning_rate": 1.241148870665412e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.1949663754504944e-05,
+      "loss": 0.8837,
       "step": 16014
     },
     {
-      "epoch": 0.4398396089093955,
+      "epoch": 0.45445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.2410625432377827e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.1948762305378751e-05,
+      "loss": 0.8688,
       "step": 16015
     },
     {
-      "epoch": 0.4398670731372388,
+      "epoch": 0.454483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.240976213902773e-05,
-      "loss": 1.0875,
+      "learning_rate": 1.1947860839791276e-05,
+      "loss": 0.8203,
       "step": 16016
     },
     {
-      "epoch": 0.439894537365082,
+      "epoch": 0.45451191827468784,
       "grad_norm": 0.0,
-      "learning_rate": 1.240889882661066e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.1946959357750132e-05,
+      "loss": 0.8373,
       "step": 16017
     },
     {
-      "epoch": 0.4399220015929252,
+      "epoch": 0.45454029511918276,
       "grad_norm": 0.0,
-      "learning_rate": 1.2408035495133451e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.1946057859262936e-05,
+      "loss": 0.9549,
       "step": 16018
     },
     {
-      "epoch": 0.43994946582076844,
+      "epoch": 0.4545686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.2407172144602931e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.1945156344337303e-05,
+      "loss": 0.8099,
       "step": 16019
     },
     {
-      "epoch": 0.43997693004861166,
+      "epoch": 0.45459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.2406308775025933e-05,
-      "loss": 0.9255,
+      "learning_rate": 1.1944254812980841e-05,
+      "loss": 0.9709,
       "step": 16020
     },
     {
-      "epoch": 0.44000439427645494,
+      "epoch": 0.45462542565266745,
       "grad_norm": 0.0,
-      "learning_rate": 1.240544538640929e-05,
-      "loss": 0.8218,
+      "learning_rate": 1.1943353265201177e-05,
+      "loss": 0.9062,
       "step": 16021
     },
     {
-      "epoch": 0.44003185850429816,
+      "epoch": 0.4546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.240458197875983e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.194245170100592e-05,
+      "loss": 0.9382,
       "step": 16022
     },
     {
-      "epoch": 0.4400593227321414,
+      "epoch": 0.4546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.2403718552084385e-05,
-      "loss": 0.8429,
+      "learning_rate": 1.1941550120402683e-05,
+      "loss": 0.8557,
       "step": 16023
     },
     {
-      "epoch": 0.4400867869599846,
+      "epoch": 0.4547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.240285510638979e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.1940648523399085e-05,
+      "loss": 0.9271,
       "step": 16024
     },
     {
-      "epoch": 0.4401142511878279,
+      "epoch": 0.454738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.2401991641682868e-05,
-      "loss": 0.9625,
+      "learning_rate": 1.1939746910002745e-05,
+      "loss": 0.9989,
       "step": 16025
     },
     {
-      "epoch": 0.4401417154156711,
+      "epoch": 0.4547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.2401128157970463e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.1938845280221275e-05,
+      "loss": 0.8996,
       "step": 16026
     },
     {
-      "epoch": 0.4401691796435143,
+      "epoch": 0.45479568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 1.2400264655259401e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.193794363406229e-05,
+      "loss": 0.9485,
       "step": 16027
     },
     {
-      "epoch": 0.44019664387135754,
+      "epoch": 0.4548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.2399401133556518e-05,
-      "loss": 0.8827,
+      "learning_rate": 1.1937041971533407e-05,
+      "loss": 0.8887,
       "step": 16028
     },
     {
-      "epoch": 0.4402241080992008,
+      "epoch": 0.45485244040862655,
       "grad_norm": 0.0,
-      "learning_rate": 1.239853759286864e-05,
-      "loss": 0.9004,
+      "learning_rate": 1.1936140292642247e-05,
+      "loss": 0.9808,
       "step": 16029
     },
     {
-      "epoch": 0.44025157232704404,
+      "epoch": 0.45488081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.2397674033202604e-05,
-      "loss": 0.9024,
+      "learning_rate": 1.1935238597396421e-05,
+      "loss": 0.7949,
       "step": 16030
     },
     {
-      "epoch": 0.44027903655488726,
+      "epoch": 0.4549091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.2396810454565244e-05,
-      "loss": 0.8505,
+      "learning_rate": 1.1934336885803548e-05,
+      "loss": 0.924,
       "step": 16031
     },
     {
-      "epoch": 0.4403065007827305,
+      "epoch": 0.45493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.239594685696339e-05,
-      "loss": 0.9359,
+      "learning_rate": 1.1933435157871245e-05,
+      "loss": 0.8395,
       "step": 16032
     },
     {
-      "epoch": 0.4403339650105737,
+      "epoch": 0.45496594778660615,
       "grad_norm": 0.0,
-      "learning_rate": 1.2395083240403877e-05,
-      "loss": 0.913,
+      "learning_rate": 1.1932533413607128e-05,
+      "loss": 0.9522,
       "step": 16033
     },
     {
-      "epoch": 0.440361429238417,
+      "epoch": 0.454994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.2394219604893539e-05,
-      "loss": 1.11,
+      "learning_rate": 1.1931631653018812e-05,
+      "loss": 0.8533,
       "step": 16034
     },
     {
-      "epoch": 0.4403888934662602,
+      "epoch": 0.4550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.2393355950439205e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.1930729876113918e-05,
+      "loss": 0.8814,
       "step": 16035
     },
     {
-      "epoch": 0.4404163576941034,
+      "epoch": 0.4550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.2392492277047714e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.1929828082900062e-05,
+      "loss": 0.8327,
       "step": 16036
     },
     {
-      "epoch": 0.44044382192194664,
+      "epoch": 0.4550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.2391628584725897e-05,
-      "loss": 0.8915,
+      "learning_rate": 1.192892627338486e-05,
+      "loss": 0.8452,
       "step": 16037
     },
     {
-      "epoch": 0.4404712861497899,
+      "epoch": 0.4551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.2390764873480585e-05,
-      "loss": 0.9723,
+      "learning_rate": 1.1928024447575934e-05,
+      "loss": 0.85,
       "step": 16038
     },
     {
-      "epoch": 0.44049875037763314,
+      "epoch": 0.4551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.238990114331862e-05,
-      "loss": 0.8836,
+      "learning_rate": 1.1927122605480899e-05,
+      "loss": 1.0073,
       "step": 16039
     },
     {
-      "epoch": 0.44052621460547636,
+      "epoch": 0.4551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.238903739424683e-05,
-      "loss": 0.8703,
+      "learning_rate": 1.1926220747107372e-05,
+      "loss": 0.9376,
       "step": 16040
     },
     {
-      "epoch": 0.4405536788333196,
+      "epoch": 0.45519296254256525,
       "grad_norm": 0.0,
-      "learning_rate": 1.2388173626272048e-05,
-      "loss": 0.8349,
+      "learning_rate": 1.192531887246297e-05,
+      "loss": 0.8592,
       "step": 16041
     },
     {
-      "epoch": 0.44058114306116286,
+      "epoch": 0.45522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.2387309839401113e-05,
-      "loss": 0.899,
+      "learning_rate": 1.1924416981555314e-05,
+      "loss": 0.9031,
       "step": 16042
     },
     {
-      "epoch": 0.4406086072890061,
+      "epoch": 0.455249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.2386446033640857e-05,
-      "loss": 0.8854,
+      "learning_rate": 1.1923515074392022e-05,
+      "loss": 0.8862,
       "step": 16043
     },
     {
-      "epoch": 0.4406360715168493,
+      "epoch": 0.45527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.2385582208998117e-05,
-      "loss": 0.9677,
+      "learning_rate": 1.1922613150980708e-05,
+      "loss": 0.925,
       "step": 16044
     },
     {
-      "epoch": 0.4406635357446925,
+      "epoch": 0.45530646992054485,
       "grad_norm": 0.0,
-      "learning_rate": 1.2384718365479729e-05,
-      "loss": 0.9921,
+      "learning_rate": 1.1921711211329e-05,
+      "loss": 1.0089,
       "step": 16045
     },
     {
-      "epoch": 0.44069099997253575,
+      "epoch": 0.4553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.2383854503092521e-05,
-      "loss": 0.9823,
+      "learning_rate": 1.1920809255444507e-05,
+      "loss": 0.9448,
       "step": 16046
     },
     {
-      "epoch": 0.440718464200379,
+      "epoch": 0.45536322360953463,
       "grad_norm": 0.0,
-      "learning_rate": 1.2382990621843334e-05,
-      "loss": 0.9442,
+      "learning_rate": 1.1919907283334854e-05,
+      "loss": 1.0126,
       "step": 16047
     },
     {
-      "epoch": 0.44074592842822224,
+      "epoch": 0.4553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.2382126721739004e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.1919005295007655e-05,
+      "loss": 0.8771,
       "step": 16048
     },
     {
-      "epoch": 0.44077339265606547,
+      "epoch": 0.4554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.2381262802786362e-05,
-      "loss": 0.7965,
+      "learning_rate": 1.1918103290470535e-05,
+      "loss": 0.8765,
       "step": 16049
     },
     {
-      "epoch": 0.4408008568839087,
+      "epoch": 0.4554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.2380398864992245e-05,
-      "loss": 0.9644,
+      "learning_rate": 1.1917201269731109e-05,
+      "loss": 0.9325,
       "step": 16050
     },
     {
-      "epoch": 0.44082832111175196,
+      "epoch": 0.4554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.2379534908363493e-05,
-      "loss": 0.8704,
+      "learning_rate": 1.1916299232796997e-05,
+      "loss": 0.9351,
       "step": 16051
     },
     {
-      "epoch": 0.4408557853395952,
+      "epoch": 0.4555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.237867093290694e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.191539717967582e-05,
+      "loss": 0.8805,
       "step": 16052
     },
     {
-      "epoch": 0.4408832495674384,
+      "epoch": 0.45553348467650395,
       "grad_norm": 0.0,
-      "learning_rate": 1.2377806938629418e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.1914495110375199e-05,
+      "loss": 0.9759,
       "step": 16053
     },
     {
-      "epoch": 0.44091071379528163,
+      "epoch": 0.45556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.237694292553777e-05,
-      "loss": 0.8061,
+      "learning_rate": 1.1913593024902748e-05,
+      "loss": 0.9126,
       "step": 16054
     },
     {
-      "epoch": 0.4409381780231249,
+      "epoch": 0.4555902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.2376078893638826e-05,
-      "loss": 0.868,
+      "learning_rate": 1.1912690923266095e-05,
+      "loss": 0.8929,
       "step": 16055
     },
     {
-      "epoch": 0.4409656422509681,
+      "epoch": 0.45561861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 1.2375214842939426e-05,
-      "loss": 0.9104,
+      "learning_rate": 1.1911788805472853e-05,
+      "loss": 0.8643,
       "step": 16056
     },
     {
-      "epoch": 0.44099310647881135,
+      "epoch": 0.45564699205448356,
       "grad_norm": 0.0,
-      "learning_rate": 1.2374350773446409e-05,
-      "loss": 1.0168,
+      "learning_rate": 1.1910886671530644e-05,
+      "loss": 0.8749,
       "step": 16057
     },
     {
-      "epoch": 0.44102057070665457,
+      "epoch": 0.4556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.2373486685166604e-05,
-      "loss": 0.9457,
+      "learning_rate": 1.1909984521447091e-05,
+      "loss": 0.7935,
       "step": 16058
     },
     {
-      "epoch": 0.4410480349344978,
+      "epoch": 0.45570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.2372622578106857e-05,
-      "loss": 0.9275,
+      "learning_rate": 1.1909082355229812e-05,
+      "loss": 0.892,
       "step": 16059
     },
     {
-      "epoch": 0.44107549916234107,
+      "epoch": 0.4557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.2371758452273999e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.190818017288643e-05,
+      "loss": 0.8882,
       "step": 16060
     },
     {
-      "epoch": 0.4411029633901843,
+      "epoch": 0.4557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.2370894307674868e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.1907277974424562e-05,
+      "loss": 0.8821,
       "step": 16061
     },
     {
-      "epoch": 0.4411304276180275,
+      "epoch": 0.455788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.2370030144316305e-05,
-      "loss": 0.8722,
+      "learning_rate": 1.1906375759851834e-05,
+      "loss": 0.8327,
       "step": 16062
     },
     {
-      "epoch": 0.44115789184587073,
+      "epoch": 0.4558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.2369165962205145e-05,
-      "loss": 1.0568,
+      "learning_rate": 1.1905473529175864e-05,
+      "loss": 0.9045,
       "step": 16063
     },
     {
-      "epoch": 0.441185356073714,
+      "epoch": 0.4558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.2368301761348227e-05,
-      "loss": 0.8916,
+      "learning_rate": 1.1904571282404271e-05,
+      "loss": 0.9839,
       "step": 16064
     },
     {
-      "epoch": 0.44121282030155723,
+      "epoch": 0.45587400681044266,
       "grad_norm": 0.0,
-      "learning_rate": 1.2367437541752387e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.1903669019544682e-05,
+      "loss": 0.8322,
       "step": 16065
     },
     {
-      "epoch": 0.44124028452940045,
+      "epoch": 0.45590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.2366573303424465e-05,
-      "loss": 0.9002,
+      "learning_rate": 1.1902766740604714e-05,
+      "loss": 0.8847,
       "step": 16066
     },
     {
-      "epoch": 0.44126774875724367,
+      "epoch": 0.4559307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.2365709046371299e-05,
-      "loss": 0.8829,
+      "learning_rate": 1.1901864445591987e-05,
+      "loss": 0.9021,
       "step": 16067
     },
     {
-      "epoch": 0.44129521298508695,
+      "epoch": 0.45595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.2364844770599724e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.1900962134514127e-05,
+      "loss": 0.8859,
       "step": 16068
     },
     {
-      "epoch": 0.44132267721293017,
+      "epoch": 0.45598751418842226,
       "grad_norm": 0.0,
-      "learning_rate": 1.2363980476116582e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.1900059807378757e-05,
+      "loss": 0.8465,
       "step": 16069
     },
     {
-      "epoch": 0.4413501414407734,
+      "epoch": 0.4560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.236311616292871e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.1899157464193493e-05,
+      "loss": 0.9252,
       "step": 16070
     },
     {
-      "epoch": 0.4413776056686166,
+      "epoch": 0.45604426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.236225183104295e-05,
-      "loss": 1.0247,
+      "learning_rate": 1.189825510496596e-05,
+      "loss": 0.8835,
       "step": 16071
     },
     {
-      "epoch": 0.44140506989645983,
+      "epoch": 0.45607264472190695,
       "grad_norm": 0.0,
-      "learning_rate": 1.2361387480466139e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.1897352729703786e-05,
+      "loss": 0.7802,
       "step": 16072
     },
     {
-      "epoch": 0.4414325341243031,
+      "epoch": 0.4561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.2360523111205112e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.1896450338414584e-05,
+      "loss": 0.9049,
       "step": 16073
     },
     {
-      "epoch": 0.44145999835214633,
+      "epoch": 0.4561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.2359658723266714e-05,
-      "loss": 0.8966,
+      "learning_rate": 1.1895547931105982e-05,
+      "loss": 1.0201,
       "step": 16074
     },
     {
-      "epoch": 0.44148746257998955,
+      "epoch": 0.4561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.235879431665778e-05,
-      "loss": 0.8484,
+      "learning_rate": 1.18946455077856e-05,
+      "loss": 0.7984,
       "step": 16075
     },
     {
-      "epoch": 0.4415149268078328,
+      "epoch": 0.4561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.2357929891385154e-05,
-      "loss": 0.9872,
+      "learning_rate": 1.1893743068461064e-05,
+      "loss": 0.9276,
       "step": 16076
     },
     {
-      "epoch": 0.44154239103567605,
+      "epoch": 0.45621452894438136,
       "grad_norm": 0.0,
-      "learning_rate": 1.2357065447455674e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.1892840613139995e-05,
+      "loss": 0.8734,
       "step": 16077
     },
     {
-      "epoch": 0.4415698552635193,
+      "epoch": 0.4562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.2356200984876177e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.1891938141830015e-05,
+      "loss": 0.7728,
       "step": 16078
     },
     {
-      "epoch": 0.4415973194913625,
+      "epoch": 0.4562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.2355336503653507e-05,
-      "loss": 0.8674,
+      "learning_rate": 1.189103565453875e-05,
+      "loss": 0.8426,
       "step": 16079
     },
     {
-      "epoch": 0.4416247837192057,
+      "epoch": 0.45629965947786605,
       "grad_norm": 0.0,
-      "learning_rate": 1.2354472003794503e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.1890133151273822e-05,
+      "loss": 0.8976,
       "step": 16080
     },
     {
-      "epoch": 0.441652247947049,
+      "epoch": 0.45632803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.2353607485306001e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.1889230632042851e-05,
+      "loss": 0.83,
       "step": 16081
     },
     {
-      "epoch": 0.4416797121748922,
+      "epoch": 0.4563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.2352742948194845e-05,
-      "loss": 0.8844,
+      "learning_rate": 1.1888328096853465e-05,
+      "loss": 0.8744,
       "step": 16082
     },
     {
-      "epoch": 0.44170717640273544,
+      "epoch": 0.45638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.2351878392467877e-05,
-      "loss": 0.9649,
+      "learning_rate": 1.1887425545713291e-05,
+      "loss": 0.8984,
       "step": 16083
     },
     {
-      "epoch": 0.44173464063057866,
+      "epoch": 0.45641316685584565,
       "grad_norm": 0.0,
-      "learning_rate": 1.2351013818131939e-05,
-      "loss": 0.8684,
+      "learning_rate": 1.1886522978629947e-05,
+      "loss": 0.9606,
       "step": 16084
     },
     {
-      "epoch": 0.4417621048584219,
+      "epoch": 0.4564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.2350149225193865e-05,
-      "loss": 0.9064,
+      "learning_rate": 1.1885620395611054e-05,
+      "loss": 0.8798,
       "step": 16085
     },
     {
-      "epoch": 0.44178956908626515,
+      "epoch": 0.45646992054483543,
       "grad_norm": 0.0,
-      "learning_rate": 1.23492846136605e-05,
-      "loss": 0.9094,
+      "learning_rate": 1.1884717796664246e-05,
+      "loss": 0.9293,
       "step": 16086
     },
     {
-      "epoch": 0.4418170333141084,
+      "epoch": 0.4564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.2348419983538684e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.188381518179714e-05,
+      "loss": 0.8072,
       "step": 16087
     },
     {
-      "epoch": 0.4418444975419516,
+      "epoch": 0.4565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.2347555334835262e-05,
-      "loss": 0.8069,
+      "learning_rate": 1.1882912551017362e-05,
+      "loss": 0.9212,
       "step": 16088
     },
     {
-      "epoch": 0.4418719617697948,
+      "epoch": 0.4565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.2346690667557075e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.188200990433254e-05,
+      "loss": 0.9512,
       "step": 16089
     },
     {
-      "epoch": 0.4418994259976381,
+      "epoch": 0.456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.2345825981710957e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.1881107241750289e-05,
+      "loss": 0.8698,
       "step": 16090
     },
     {
-      "epoch": 0.4419268902254813,
+      "epoch": 0.4566118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.234496127730376e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.1880204563278244e-05,
+      "loss": 0.9896,
       "step": 16091
     },
     {
-      "epoch": 0.44195435445332454,
+      "epoch": 0.45664018161180475,
       "grad_norm": 0.0,
-      "learning_rate": 1.234409655434232e-05,
-      "loss": 0.9005,
+      "learning_rate": 1.1879301868924029e-05,
+      "loss": 0.8167,
       "step": 16092
     },
     {
-      "epoch": 0.44198181868116776,
+      "epoch": 0.45666855845629967,
       "grad_norm": 0.0,
-      "learning_rate": 1.2343231812833476e-05,
-      "loss": 0.9836,
+      "learning_rate": 1.1878399158695267e-05,
+      "loss": 1.0118,
       "step": 16093
     },
     {
-      "epoch": 0.44200928290901104,
+      "epoch": 0.4566969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.2342367052784076e-05,
-      "loss": 1.0088,
+      "learning_rate": 1.187749643259958e-05,
+      "loss": 0.8527,
       "step": 16094
     },
     {
-      "epoch": 0.44203674713685426,
+      "epoch": 0.45672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.2341502274200963e-05,
-      "loss": 1.0236,
+      "learning_rate": 1.1876593690644595e-05,
+      "loss": 0.8034,
       "step": 16095
     },
     {
-      "epoch": 0.4420642113646975,
+      "epoch": 0.45675368898978436,
       "grad_norm": 0.0,
-      "learning_rate": 1.2340637477090977e-05,
-      "loss": 0.8309,
+      "learning_rate": 1.1875690932837942e-05,
+      "loss": 0.9097,
       "step": 16096
     },
     {
-      "epoch": 0.4420916755925407,
+      "epoch": 0.4567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.2339772661460958e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.1874788159187243e-05,
+      "loss": 0.8402,
       "step": 16097
     },
     {
-      "epoch": 0.442119139820384,
+      "epoch": 0.45681044267877413,
       "grad_norm": 0.0,
-      "learning_rate": 1.2338907827317755e-05,
-      "loss": 0.9469,
+      "learning_rate": 1.187388536970012e-05,
+      "loss": 0.9584,
       "step": 16098
     },
     {
-      "epoch": 0.4421466040482272,
+      "epoch": 0.456838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.2338042974668203e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.1872982564384208e-05,
+      "loss": 0.7984,
       "step": 16099
     },
     {
-      "epoch": 0.4421740682760704,
+      "epoch": 0.4568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.2337178103519152e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.1872079743247127e-05,
+      "loss": 0.9021,
       "step": 16100
     },
     {
-      "epoch": 0.44220153250391364,
+      "epoch": 0.4568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.2336313213877442e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.1871176906296502e-05,
+      "loss": 1.0009,
       "step": 16101
     },
     {
-      "epoch": 0.44222899673175686,
+      "epoch": 0.4569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.2335448305749915e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.1870274053539966e-05,
+      "loss": 0.8922,
       "step": 16102
     },
     {
-      "epoch": 0.44225646095960014,
+      "epoch": 0.4569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.2334583379143418e-05,
-      "loss": 0.9801,
+      "learning_rate": 1.1869371184985139e-05,
+      "loss": 0.9117,
       "step": 16103
     },
     {
-      "epoch": 0.44228392518744336,
+      "epoch": 0.45698070374574346,
       "grad_norm": 0.0,
-      "learning_rate": 1.2333718434064795e-05,
-      "loss": 0.9183,
+      "learning_rate": 1.1868468300639646e-05,
+      "loss": 0.916,
       "step": 16104
     },
     {
-      "epoch": 0.4423113894152866,
+      "epoch": 0.45700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.2332853470520885e-05,
-      "loss": 0.8325,
+      "learning_rate": 1.1867565400511119e-05,
+      "loss": 0.9541,
       "step": 16105
     },
     {
-      "epoch": 0.4423388536431298,
+      "epoch": 0.4570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.2331988488518535e-05,
-      "loss": 0.8659,
+      "learning_rate": 1.1866662484607186e-05,
+      "loss": 0.9626,
       "step": 16106
     },
     {
-      "epoch": 0.4423663178709731,
+      "epoch": 0.45706583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 1.2331123488064586e-05,
-      "loss": 0.8641,
+      "learning_rate": 1.186575955293547e-05,
+      "loss": 0.9178,
       "step": 16107
     },
     {
-      "epoch": 0.4423937820988163,
+      "epoch": 0.45709421112372306,
       "grad_norm": 0.0,
-      "learning_rate": 1.233025846916589e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.1864856605503596e-05,
+      "loss": 0.8685,
       "step": 16108
     },
     {
-      "epoch": 0.4424212463266595,
+      "epoch": 0.4571225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.2329393431829284e-05,
-      "loss": 0.8494,
+      "learning_rate": 1.1863953642319199e-05,
+      "loss": 0.8206,
       "step": 16109
     },
     {
-      "epoch": 0.44244871055450274,
+      "epoch": 0.45715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.2328528376061618e-05,
-      "loss": 0.8457,
+      "learning_rate": 1.1863050663389899e-05,
+      "loss": 0.8952,
       "step": 16110
     },
     {
-      "epoch": 0.442476174782346,
+      "epoch": 0.4571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.2327663301869728e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.1862147668723328e-05,
+      "loss": 0.9252,
       "step": 16111
     },
     {
-      "epoch": 0.44250363901018924,
+      "epoch": 0.4572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.2326798209260469e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.1861244658327111e-05,
+      "loss": 0.9057,
       "step": 16112
     },
     {
-      "epoch": 0.44253110323803246,
+      "epoch": 0.4572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.2325933098240679e-05,
-      "loss": 0.8817,
+      "learning_rate": 1.1860341632208874e-05,
+      "loss": 1.0225,
       "step": 16113
     },
     {
-      "epoch": 0.4425585674658757,
+      "epoch": 0.4572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.2325067968817204e-05,
-      "loss": 0.8953,
+      "learning_rate": 1.1859438590376251e-05,
+      "loss": 0.8469,
       "step": 16114
     },
     {
-      "epoch": 0.4425860316937189,
+      "epoch": 0.4572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.2324202820996893e-05,
-      "loss": 0.8662,
+      "learning_rate": 1.1858535532836866e-05,
+      "loss": 0.856,
       "step": 16115
     },
     {
-      "epoch": 0.4426134959215622,
+      "epoch": 0.45732122587968216,
       "grad_norm": 0.0,
-      "learning_rate": 1.232333765478659e-05,
-      "loss": 0.848,
+      "learning_rate": 1.1857632459598346e-05,
+      "loss": 0.9516,
       "step": 16116
     },
     {
-      "epoch": 0.4426409601494054,
+      "epoch": 0.4573496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.2322472470193137e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.1856729370668325e-05,
+      "loss": 1.0113,
       "step": 16117
     },
     {
-      "epoch": 0.4426684243772486,
+      "epoch": 0.457377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.232160726722338e-05,
-      "loss": 0.9478,
+      "learning_rate": 1.1855826266054425e-05,
+      "loss": 0.965,
       "step": 16118
     },
     {
-      "epoch": 0.44269588860509185,
+      "epoch": 0.45740635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.2320742045884172e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.1854923145764279e-05,
+      "loss": 0.8672,
       "step": 16119
     },
     {
-      "epoch": 0.4427233528329351,
+      "epoch": 0.45743473325766176,
       "grad_norm": 0.0,
-      "learning_rate": 1.2319876806182352e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.1854020009805511e-05,
+      "loss": 0.9708,
       "step": 16120
     },
     {
-      "epoch": 0.44275081706077835,
+      "epoch": 0.4574631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.2319011548124766e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.1853116858185754e-05,
+      "loss": 0.9513,
       "step": 16121
     },
     {
-      "epoch": 0.44277828128862157,
+      "epoch": 0.45749148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 1.2318146271718265e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.1852213690912632e-05,
+      "loss": 0.8508,
       "step": 16122
     },
     {
-      "epoch": 0.4428057455164648,
+      "epoch": 0.4575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.2317280976969692e-05,
-      "loss": 0.9152,
+      "learning_rate": 1.1851310507993784e-05,
+      "loss": 0.8807,
       "step": 16123
     },
     {
-      "epoch": 0.44283320974430807,
+      "epoch": 0.4575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.2316415663885895e-05,
-      "loss": 1.0223,
+      "learning_rate": 1.1850407309436831e-05,
+      "loss": 0.9282,
       "step": 16124
     },
     {
-      "epoch": 0.4428606739721513,
+      "epoch": 0.4575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.231555033247372e-05,
-      "loss": 0.9675,
+      "learning_rate": 1.1849504095249402e-05,
+      "loss": 0.9585,
       "step": 16125
     },
     {
-      "epoch": 0.4428881381999945,
+      "epoch": 0.4576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.2314684982740013e-05,
-      "loss": 0.9192,
+      "learning_rate": 1.1848600865439131e-05,
+      "loss": 0.8961,
       "step": 16126
     },
     {
-      "epoch": 0.44291560242783773,
+      "epoch": 0.457633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.2313819614691623e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.1847697620013645e-05,
+      "loss": 0.9299,
       "step": 16127
     },
     {
-      "epoch": 0.44294306665568095,
+      "epoch": 0.45766174801362086,
       "grad_norm": 0.0,
-      "learning_rate": 1.2312954228335395e-05,
-      "loss": 0.9301,
+      "learning_rate": 1.1846794358980573e-05,
+      "loss": 0.8652,
       "step": 16128
     },
     {
-      "epoch": 0.4429705308835242,
+      "epoch": 0.4576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.2312088823678174e-05,
-      "loss": 0.8914,
+      "learning_rate": 1.1845891082347545e-05,
+      "loss": 1.0438,
       "step": 16129
     },
     {
-      "epoch": 0.44299799511136745,
+      "epoch": 0.4577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.2311223400726813e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.1844987790122195e-05,
+      "loss": 0.9524,
       "step": 16130
     },
     {
-      "epoch": 0.44302545933921067,
+      "epoch": 0.45774687854710555,
       "grad_norm": 0.0,
-      "learning_rate": 1.2310357959488159e-05,
-      "loss": 0.8356,
+      "learning_rate": 1.1844084482312148e-05,
+      "loss": 1.0616,
       "step": 16131
     },
     {
-      "epoch": 0.4430529235670539,
+      "epoch": 0.45777525539160047,
       "grad_norm": 0.0,
-      "learning_rate": 1.2309492499969059e-05,
-      "loss": 0.9943,
+      "learning_rate": 1.1843181158925036e-05,
+      "loss": 0.8762,
       "step": 16132
     },
     {
-      "epoch": 0.44308038779489717,
+      "epoch": 0.4578036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.2308627022176356e-05,
-      "loss": 0.8691,
+      "learning_rate": 1.1842277819968493e-05,
+      "loss": 0.8398,
       "step": 16133
     },
     {
-      "epoch": 0.4431078520227404,
+      "epoch": 0.45783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.2307761526116901e-05,
-      "loss": 0.8991,
+      "learning_rate": 1.1841374465450146e-05,
+      "loss": 0.822,
       "step": 16134
     },
     {
-      "epoch": 0.4431353162505836,
+      "epoch": 0.45786038592508516,
       "grad_norm": 0.0,
-      "learning_rate": 1.2306896011797544e-05,
-      "loss": 0.8568,
+      "learning_rate": 1.1840471095377623e-05,
+      "loss": 0.9466,
       "step": 16135
     },
     {
-      "epoch": 0.44316278047842683,
+      "epoch": 0.45788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.2306030479225137e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.183956770975856e-05,
+      "loss": 0.8609,
       "step": 16136
     },
     {
-      "epoch": 0.4431902447062701,
+      "epoch": 0.45791713961407493,
       "grad_norm": 0.0,
-      "learning_rate": 1.2305164928406518e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.1838664308600585e-05,
+      "loss": 0.8407,
       "step": 16137
     },
     {
-      "epoch": 0.44321770893411333,
+      "epoch": 0.4579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.2304299359348541e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.1837760891911332e-05,
+      "loss": 0.881,
       "step": 16138
     },
     {
-      "epoch": 0.44324517316195655,
+      "epoch": 0.4579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.2303433772058058e-05,
-      "loss": 0.8064,
+      "learning_rate": 1.1836857459698427e-05,
+      "loss": 0.9619,
       "step": 16139
     },
     {
-      "epoch": 0.4432726373897998,
+      "epoch": 0.45800227014755956,
       "grad_norm": 0.0,
-      "learning_rate": 1.2302568166541914e-05,
-      "loss": 0.9077,
+      "learning_rate": 1.1835954011969508e-05,
+      "loss": 0.8104,
       "step": 16140
     },
     {
-      "epoch": 0.443300101617643,
+      "epoch": 0.4580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.2301702542806958e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.1835050548732202e-05,
+      "loss": 0.9669,
       "step": 16141
     },
     {
-      "epoch": 0.44332756584548627,
+      "epoch": 0.4580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.230083690086004e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.183414706999414e-05,
+      "loss": 1.0205,
       "step": 16142
     },
     {
-      "epoch": 0.4433550300733295,
+      "epoch": 0.45808740068104425,
       "grad_norm": 0.0,
-      "learning_rate": 1.2299971240708008e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.1833243575762956e-05,
+      "loss": 0.8468,
       "step": 16143
     },
     {
-      "epoch": 0.4433824943011727,
+      "epoch": 0.45811577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 1.2299105562357716e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.1832340066046279e-05,
+      "loss": 0.9825,
       "step": 16144
     },
     {
-      "epoch": 0.44340995852901594,
+      "epoch": 0.45814415437003403,
       "grad_norm": 0.0,
-      "learning_rate": 1.2298239865816008e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.1831436540851743e-05,
+      "loss": 0.8827,
       "step": 16145
     },
     {
-      "epoch": 0.4434374227568592,
+      "epoch": 0.45817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.2297374151089733e-05,
-      "loss": 0.8165,
+      "learning_rate": 1.1830533000186984e-05,
+      "loss": 0.9673,
       "step": 16146
     },
     {
-      "epoch": 0.44346488698470243,
+      "epoch": 0.45820090805902386,
       "grad_norm": 0.0,
-      "learning_rate": 1.2296508418185748e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.182962944405963e-05,
+      "loss": 0.9075,
       "step": 16147
     },
     {
-      "epoch": 0.44349235121254565,
+      "epoch": 0.4582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.22956426671109e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.1828725872477313e-05,
+      "loss": 0.847,
       "step": 16148
     },
     {
-      "epoch": 0.4435198154403889,
+      "epoch": 0.45825766174801363,
       "grad_norm": 0.0,
-      "learning_rate": 1.2294776897872033e-05,
-      "loss": 0.928,
+      "learning_rate": 1.1827822285447666e-05,
+      "loss": 0.8253,
       "step": 16149
     },
     {
-      "epoch": 0.44354727966823215,
+      "epoch": 0.4582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.2293911110476005e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.1826918682978321e-05,
+      "loss": 0.8169,
       "step": 16150
     },
     {
-      "epoch": 0.4435747438960754,
+      "epoch": 0.4583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.2293045304929664e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.1826015065076912e-05,
+      "loss": 0.8194,
       "step": 16151
     },
     {
-      "epoch": 0.4436022081239186,
+      "epoch": 0.4583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.2292179481239861e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.182511143175107e-05,
+      "loss": 0.8063,
       "step": 16152
     },
     {
-      "epoch": 0.4436296723517618,
+      "epoch": 0.4583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.2291313639413443e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.1824207783008434e-05,
+      "loss": 0.8914,
       "step": 16153
     },
     {
-      "epoch": 0.44365713657960504,
+      "epoch": 0.4583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.2290447779457266e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.182330411885663e-05,
+      "loss": 1.0051,
       "step": 16154
     },
     {
-      "epoch": 0.4436846008074483,
+      "epoch": 0.45842792281498296,
       "grad_norm": 0.0,
-      "learning_rate": 1.2289581901378178e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.1822400439303295e-05,
+      "loss": 0.9047,
       "step": 16155
     },
     {
-      "epoch": 0.44371206503529154,
+      "epoch": 0.4584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.2288716005183032e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.1821496744356062e-05,
+      "loss": 0.8352,
       "step": 16156
     },
     {
-      "epoch": 0.44373952926313476,
+      "epoch": 0.45848467650397273,
       "grad_norm": 0.0,
-      "learning_rate": 1.2287850090878681e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.1820593034022565e-05,
+      "loss": 0.8419,
       "step": 16157
     },
     {
-      "epoch": 0.443766993490978,
+      "epoch": 0.45851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.2286984158471969e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.1819689308310433e-05,
+      "loss": 0.8853,
       "step": 16158
     },
     {
-      "epoch": 0.44379445771882126,
+      "epoch": 0.45854143019296256,
       "grad_norm": 0.0,
-      "learning_rate": 1.2286118207969755e-05,
-      "loss": 0.906,
+      "learning_rate": 1.1818785567227307e-05,
+      "loss": 0.8907,
       "step": 16159
     },
     {
-      "epoch": 0.4438219219466645,
+      "epoch": 0.4585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.2285252239378888e-05,
-      "loss": 1.0349,
+      "learning_rate": 1.1817881810780815e-05,
+      "loss": 0.8947,
       "step": 16160
     },
     {
-      "epoch": 0.4438493861745077,
+      "epoch": 0.45859818388195234,
       "grad_norm": 0.0,
-      "learning_rate": 1.2284386252706219e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.1816978038978596e-05,
+      "loss": 0.8416,
       "step": 16161
     },
     {
-      "epoch": 0.4438768504023509,
+      "epoch": 0.4586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.2283520247958602e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.1816074251828281e-05,
+      "loss": 0.9476,
       "step": 16162
     },
     {
-      "epoch": 0.4439043146301942,
+      "epoch": 0.4586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.228265422514289e-05,
-      "loss": 0.82,
+      "learning_rate": 1.1815170449337507e-05,
+      "loss": 0.8619,
       "step": 16163
     },
     {
-      "epoch": 0.4439317788580374,
+      "epoch": 0.458683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.228178818426593e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.1814266631513906e-05,
+      "loss": 0.9351,
       "step": 16164
     },
     {
-      "epoch": 0.44395924308588064,
+      "epoch": 0.4587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.228092212533458e-05,
-      "loss": 0.7881,
+      "learning_rate": 1.1813362798365111e-05,
+      "loss": 0.8603,
       "step": 16165
     },
     {
-      "epoch": 0.44398670731372386,
+      "epoch": 0.4587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.228005604835569e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.181245894989876e-05,
+      "loss": 0.9743,
       "step": 16166
     },
     {
-      "epoch": 0.4440141715415671,
+      "epoch": 0.45876844494892166,
       "grad_norm": 0.0,
-      "learning_rate": 1.2279189953336113e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.1811555086122486e-05,
+      "loss": 0.7537,
       "step": 16167
     },
     {
-      "epoch": 0.44404163576941036,
+      "epoch": 0.4587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.2278323840282705e-05,
-      "loss": 0.8201,
+      "learning_rate": 1.1810651207043925e-05,
+      "loss": 0.8016,
       "step": 16168
     },
     {
-      "epoch": 0.4440690999972536,
+      "epoch": 0.4588251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.2277457709202314e-05,
-      "loss": 0.8098,
+      "learning_rate": 1.1809747312670711e-05,
+      "loss": 0.8914,
       "step": 16169
     },
     {
-      "epoch": 0.4440965642250968,
+      "epoch": 0.45885357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.2276591560101794e-05,
-      "loss": 1.0435,
+      "learning_rate": 1.1808843403010484e-05,
+      "loss": 0.9777,
       "step": 16170
     },
     {
-      "epoch": 0.44412402845294,
+      "epoch": 0.45888195232690127,
       "grad_norm": 0.0,
-      "learning_rate": 1.2275725392988002e-05,
-      "loss": 1.0264,
+      "learning_rate": 1.1807939478070873e-05,
+      "loss": 0.9545,
       "step": 16171
     },
     {
-      "epoch": 0.4441514926807833,
+      "epoch": 0.4589103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.2274859207867789e-05,
-      "loss": 1.014,
+      "learning_rate": 1.1807035537859514e-05,
+      "loss": 0.9003,
       "step": 16172
     },
     {
-      "epoch": 0.4441789569086265,
+      "epoch": 0.45893870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 1.2273993004748006e-05,
-      "loss": 0.9534,
+      "learning_rate": 1.1806131582384049e-05,
+      "loss": 0.8545,
       "step": 16173
     },
     {
-      "epoch": 0.44420642113646974,
+      "epoch": 0.4589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.2273126783635512e-05,
-      "loss": 0.907,
+      "learning_rate": 1.1805227611652106e-05,
+      "loss": 0.936,
       "step": 16174
     },
     {
-      "epoch": 0.44423388536431296,
+      "epoch": 0.4589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.2272260544537158e-05,
-      "loss": 0.8437,
+      "learning_rate": 1.1804323625671326e-05,
+      "loss": 0.9986,
       "step": 16175
     },
     {
-      "epoch": 0.44426134959215624,
+      "epoch": 0.45902383654937573,
       "grad_norm": 0.0,
-      "learning_rate": 1.2271394287459798e-05,
-      "loss": 0.9733,
+      "learning_rate": 1.180341962444934e-05,
+      "loss": 0.8803,
       "step": 16176
     },
     {
-      "epoch": 0.44428881381999946,
+      "epoch": 0.4590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.2270528012410285e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.180251560799379e-05,
+      "loss": 0.9626,
       "step": 16177
     },
     {
-      "epoch": 0.4443162780478427,
+      "epoch": 0.4590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.2269661719395477e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.180161157631231e-05,
+      "loss": 0.8481,
       "step": 16178
     },
     {
-      "epoch": 0.4443437422756859,
+      "epoch": 0.45910896708286036,
       "grad_norm": 0.0,
-      "learning_rate": 1.2268795408422228e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.1800707529412534e-05,
+      "loss": 0.8823,
       "step": 16179
     },
     {
-      "epoch": 0.4443712065035291,
+      "epoch": 0.4591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.2267929079497389e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.1799803467302102e-05,
+      "loss": 0.9167,
       "step": 16180
     },
     {
-      "epoch": 0.4443986707313724,
+      "epoch": 0.4591657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.2267062732627816e-05,
-      "loss": 1.0125,
+      "learning_rate": 1.1798899389988652e-05,
+      "loss": 0.8483,
       "step": 16181
     },
     {
-      "epoch": 0.4444261349592156,
+      "epoch": 0.45919409761634505,
       "grad_norm": 0.0,
-      "learning_rate": 1.2266196367820365e-05,
-      "loss": 0.8647,
+      "learning_rate": 1.1797995297479812e-05,
+      "loss": 0.9059,
       "step": 16182
     },
     {
-      "epoch": 0.44445359918705885,
+      "epoch": 0.45922247446083997,
       "grad_norm": 0.0,
-      "learning_rate": 1.2265329985081891e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.1797091189783227e-05,
+      "loss": 0.8324,
       "step": 16183
     },
     {
-      "epoch": 0.44448106341490207,
+      "epoch": 0.45925085130533483,
       "grad_norm": 0.0,
-      "learning_rate": 1.2264463584419248e-05,
-      "loss": 0.8898,
+      "learning_rate": 1.1796187066906535e-05,
+      "loss": 1.0718,
       "step": 16184
     },
     {
-      "epoch": 0.44450852764274534,
+      "epoch": 0.45927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.2263597165839292e-05,
-      "loss": 0.8196,
+      "learning_rate": 1.1795282928857368e-05,
+      "loss": 0.979,
       "step": 16185
     },
     {
-      "epoch": 0.44453599187058856,
+      "epoch": 0.45930760499432466,
       "grad_norm": 0.0,
-      "learning_rate": 1.2262730729348877e-05,
-      "loss": 0.8507,
+      "learning_rate": 1.1794378775643365e-05,
+      "loss": 0.8617,
       "step": 16186
     },
     {
-      "epoch": 0.4445634560984318,
+      "epoch": 0.4593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.2261864274954862e-05,
-      "loss": 0.8537,
+      "learning_rate": 1.1793474607272164e-05,
+      "loss": 0.9531,
       "step": 16187
     },
     {
-      "epoch": 0.444590920326275,
+      "epoch": 0.45936435868331443,
       "grad_norm": 0.0,
-      "learning_rate": 1.2260997802664101e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.1792570423751404e-05,
+      "loss": 0.8531,
       "step": 16188
     },
     {
-      "epoch": 0.4446183845541183,
+      "epoch": 0.4593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.2260131312483447e-05,
-      "loss": 0.9664,
+      "learning_rate": 1.1791666225088719e-05,
+      "loss": 0.903,
       "step": 16189
     },
     {
-      "epoch": 0.4446458487819615,
+      "epoch": 0.4594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.2259264804419761e-05,
-      "loss": 0.9191,
+      "learning_rate": 1.1790762011291747e-05,
+      "loss": 0.8793,
       "step": 16190
     },
     {
-      "epoch": 0.4446733130098047,
+      "epoch": 0.45944948921679907,
       "grad_norm": 0.0,
-      "learning_rate": 1.2258398278479895e-05,
-      "loss": 0.8823,
+      "learning_rate": 1.178985778236813e-05,
+      "loss": 1.0061,
       "step": 16191
     },
     {
-      "epoch": 0.44470077723764795,
+      "epoch": 0.459477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.225753173467071e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.1788953538325508e-05,
+      "loss": 0.8379,
       "step": 16192
     },
     {
-      "epoch": 0.4447282414654912,
+      "epoch": 0.4595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.2256665172999058e-05,
-      "loss": 0.8881,
+      "learning_rate": 1.1788049279171507e-05,
+      "loss": 0.8014,
       "step": 16193
     },
     {
-      "epoch": 0.44475570569333445,
+      "epoch": 0.45953461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 1.2255798593471795e-05,
-      "loss": 0.9262,
+      "learning_rate": 1.178714500491378e-05,
+      "loss": 0.9272,
       "step": 16194
     },
     {
-      "epoch": 0.44478316992117767,
+      "epoch": 0.45956299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.2254931996095782e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.1786240715559956e-05,
+      "loss": 0.9672,
       "step": 16195
     },
     {
-      "epoch": 0.4448106341490209,
+      "epoch": 0.45959137343927353,
       "grad_norm": 0.0,
-      "learning_rate": 1.2254065380877873e-05,
-      "loss": 1.0551,
+      "learning_rate": 1.1785336411117675e-05,
+      "loss": 0.9357,
       "step": 16196
     },
     {
-      "epoch": 0.4448380983768641,
+      "epoch": 0.45961975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.2253198747824924e-05,
-      "loss": 0.9274,
+      "learning_rate": 1.1784432091594578e-05,
+      "loss": 0.958,
       "step": 16197
     },
     {
-      "epoch": 0.4448655626047074,
+      "epoch": 0.45964812712826336,
       "grad_norm": 0.0,
-      "learning_rate": 1.2252332096943795e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.1783527756998306e-05,
+      "loss": 0.9492,
       "step": 16198
     },
     {
-      "epoch": 0.4448930268325506,
+      "epoch": 0.4596765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.2251465428241342e-05,
-      "loss": 0.8482,
+      "learning_rate": 1.178262340733649e-05,
+      "loss": 0.8914,
       "step": 16199
     },
     {
-      "epoch": 0.44492049106039383,
+      "epoch": 0.45970488081725314,
       "grad_norm": 0.0,
-      "learning_rate": 1.2250598741724423e-05,
-      "loss": 0.8492,
+      "learning_rate": 1.1781719042616777e-05,
+      "loss": 0.9597,
       "step": 16200
     },
     {
-      "epoch": 0.44494795528823705,
+      "epoch": 0.459733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.2249732037399896e-05,
-      "loss": 0.9548,
+      "learning_rate": 1.1780814662846804e-05,
+      "loss": 0.8152,
       "step": 16201
     },
     {
-      "epoch": 0.44497541951608033,
+      "epoch": 0.4597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.2248865315274614e-05,
-      "loss": 0.9374,
+      "learning_rate": 1.177991026803421e-05,
+      "loss": 0.9537,
       "step": 16202
     },
     {
-      "epoch": 0.44500288374392355,
+      "epoch": 0.45979001135073777,
       "grad_norm": 0.0,
-      "learning_rate": 1.224799857535544e-05,
-      "loss": 0.8448,
+      "learning_rate": 1.177900585818663e-05,
+      "loss": 0.8352,
       "step": 16203
     },
     {
-      "epoch": 0.44503034797176677,
+      "epoch": 0.4598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.2247131817649232e-05,
-      "loss": 0.9933,
+      "learning_rate": 1.1778101433311711e-05,
+      "loss": 1.0023,
       "step": 16204
     },
     {
-      "epoch": 0.44505781219961,
+      "epoch": 0.4598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.2246265042162846e-05,
-      "loss": 0.8252,
+      "learning_rate": 1.1777196993417087e-05,
+      "loss": 0.8757,
       "step": 16205
     },
     {
-      "epoch": 0.44508527642745327,
+      "epoch": 0.45987514188422246,
       "grad_norm": 0.0,
-      "learning_rate": 1.2245398248903138e-05,
-      "loss": 1.0396,
+      "learning_rate": 1.17762925385104e-05,
+      "loss": 0.9211,
       "step": 16206
     },
     {
-      "epoch": 0.4451127406552965,
+      "epoch": 0.4599035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.2244531437876971e-05,
-      "loss": 0.8441,
+      "learning_rate": 1.1775388068599289e-05,
+      "loss": 0.8739,
       "step": 16207
     },
     {
-      "epoch": 0.4451402048831397,
+      "epoch": 0.45993189557321223,
       "grad_norm": 0.0,
-      "learning_rate": 1.2243664609091205e-05,
-      "loss": 1.009,
+      "learning_rate": 1.1774483583691399e-05,
+      "loss": 0.8269,
       "step": 16208
     },
     {
-      "epoch": 0.44516766911098293,
+      "epoch": 0.45996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.2242797762552692e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.1773579083794363e-05,
+      "loss": 0.9752,
       "step": 16209
     },
     {
-      "epoch": 0.44519513333882615,
+      "epoch": 0.45998864926220207,
       "grad_norm": 0.0,
-      "learning_rate": 1.2241930898268296e-05,
-      "loss": 0.9035,
+      "learning_rate": 1.1772674568915827e-05,
+      "loss": 0.9712,
       "step": 16210
     },
     {
-      "epoch": 0.44522259756666943,
+      "epoch": 0.4600170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.2241064016244872e-05,
-      "loss": 0.8422,
+      "learning_rate": 1.1771770039063431e-05,
+      "loss": 0.9559,
       "step": 16211
     },
     {
-      "epoch": 0.44525006179451265,
+      "epoch": 0.46004540295119184,
       "grad_norm": 0.0,
-      "learning_rate": 1.2240197116489284e-05,
-      "loss": 1.0296,
+      "learning_rate": 1.177086549424481e-05,
+      "loss": 0.8422,
       "step": 16212
     },
     {
-      "epoch": 0.4452775260223559,
+      "epoch": 0.4600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.2239330199008387e-05,
-      "loss": 0.856,
+      "learning_rate": 1.1769960934467608e-05,
+      "loss": 0.878,
       "step": 16213
     },
     {
-      "epoch": 0.4453049902501991,
+      "epoch": 0.4601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.2238463263809042e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.176905635973947e-05,
+      "loss": 0.9428,
       "step": 16214
     },
     {
-      "epoch": 0.44533245447804237,
+      "epoch": 0.46013053348467653,
       "grad_norm": 0.0,
-      "learning_rate": 1.223759631089811e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.1768151770068033e-05,
+      "loss": 0.9047,
       "step": 16215
     },
     {
-      "epoch": 0.4453599187058856,
+      "epoch": 0.4601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.223672934028245e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.1767247165460935e-05,
+      "loss": 0.9308,
       "step": 16216
     },
     {
-      "epoch": 0.4453873829337288,
+      "epoch": 0.4601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.2235862351968919e-05,
-      "loss": 0.8704,
+      "learning_rate": 1.1766342545925824e-05,
+      "loss": 0.7324,
       "step": 16217
     },
     {
-      "epoch": 0.44541484716157204,
+      "epoch": 0.46021566401816116,
       "grad_norm": 0.0,
-      "learning_rate": 1.2234995345964378e-05,
-      "loss": 0.865,
+      "learning_rate": 1.1765437911470339e-05,
+      "loss": 0.8854,
       "step": 16218
     },
     {
-      "epoch": 0.4454423113894153,
+      "epoch": 0.4602440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.223412832227569e-05,
-      "loss": 0.8747,
+      "learning_rate": 1.1764533262102119e-05,
+      "loss": 0.8415,
       "step": 16219
     },
     {
-      "epoch": 0.44546977561725853,
+      "epoch": 0.46027241770715094,
       "grad_norm": 0.0,
-      "learning_rate": 1.2233261280909714e-05,
-      "loss": 0.8199,
+      "learning_rate": 1.1763628597828804e-05,
+      "loss": 0.9643,
       "step": 16220
     },
     {
-      "epoch": 0.44549723984510176,
+      "epoch": 0.46030079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.2232394221873308e-05,
-      "loss": 0.9531,
+      "learning_rate": 1.1762723918658043e-05,
+      "loss": 0.8957,
       "step": 16221
     },
     {
-      "epoch": 0.445524704072945,
+      "epoch": 0.46032917139614077,
       "grad_norm": 0.0,
-      "learning_rate": 1.2231527145173335e-05,
-      "loss": 0.8495,
+      "learning_rate": 1.1761819224597473e-05,
+      "loss": 0.9963,
       "step": 16222
     },
     {
-      "epoch": 0.4455521683007882,
+      "epoch": 0.46035754824063563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2230660050816651e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.1760914515654735e-05,
+      "loss": 0.9041,
       "step": 16223
     },
     {
-      "epoch": 0.4455796325286315,
+      "epoch": 0.46038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.2229792938810125e-05,
-      "loss": 0.916,
+      "learning_rate": 1.1760009791837477e-05,
+      "loss": 0.8864,
       "step": 16224
     },
     {
-      "epoch": 0.4456070967564747,
+      "epoch": 0.4604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.2228925809160614e-05,
-      "loss": 0.983,
+      "learning_rate": 1.1759105053153334e-05,
+      "loss": 0.915,
       "step": 16225
     },
     {
-      "epoch": 0.4456345609843179,
+      "epoch": 0.4604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.2228058661874978e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.1758200299609952e-05,
+      "loss": 0.9153,
       "step": 16226
     },
     {
-      "epoch": 0.44566202521216114,
+      "epoch": 0.46047105561861523,
       "grad_norm": 0.0,
-      "learning_rate": 1.2227191496960077e-05,
-      "loss": 0.843,
+      "learning_rate": 1.1757295531214974e-05,
+      "loss": 0.8143,
       "step": 16227
     },
     {
-      "epoch": 0.4456894894400044,
+      "epoch": 0.4604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.2226324314422778e-05,
-      "loss": 0.945,
+      "learning_rate": 1.1756390747976038e-05,
+      "loss": 0.8993,
       "step": 16228
     },
     {
-      "epoch": 0.44571695366784764,
+      "epoch": 0.460527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.2225457114269934e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.1755485949900796e-05,
+      "loss": 0.8616,
       "step": 16229
     },
     {
-      "epoch": 0.44574441789569086,
+      "epoch": 0.46055618615209987,
       "grad_norm": 0.0,
-      "learning_rate": 1.2224589896508412e-05,
-      "loss": 0.8086,
+      "learning_rate": 1.1754581136996882e-05,
+      "loss": 0.8656,
       "step": 16230
     },
     {
-      "epoch": 0.4457718821235341,
+      "epoch": 0.4605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.2223722661145075e-05,
-      "loss": 0.9513,
+      "learning_rate": 1.1753676309271943e-05,
+      "loss": 0.8708,
       "step": 16231
     },
     {
-      "epoch": 0.44579934635137736,
+      "epoch": 0.4606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.2222855408186783e-05,
-      "loss": 0.8812,
+      "learning_rate": 1.1752771466733622e-05,
+      "loss": 0.8421,
       "step": 16232
     },
     {
-      "epoch": 0.4458268105792206,
+      "epoch": 0.46064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.2221988137640396e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.175186660938956e-05,
+      "loss": 0.9218,
       "step": 16233
     },
     {
-      "epoch": 0.4458542748070638,
+      "epoch": 0.46066969353007947,
       "grad_norm": 0.0,
-      "learning_rate": 1.2221120849512777e-05,
-      "loss": 0.9792,
+      "learning_rate": 1.1750961737247404e-05,
+      "loss": 0.9307,
       "step": 16234
     },
     {
-      "epoch": 0.445881739034907,
+      "epoch": 0.46069807037457433,
       "grad_norm": 0.0,
-      "learning_rate": 1.2220253543810792e-05,
-      "loss": 0.9914,
+      "learning_rate": 1.1750056850314794e-05,
+      "loss": 0.8329,
       "step": 16235
     },
     {
-      "epoch": 0.44590920326275024,
+      "epoch": 0.46072644721906925,
       "grad_norm": 0.0,
-      "learning_rate": 1.2219386220541301e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.1749151948599374e-05,
+      "loss": 0.8425,
       "step": 16236
     },
     {
-      "epoch": 0.4459366674905935,
+      "epoch": 0.4607548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.2218518879711168e-05,
-      "loss": 0.8944,
+      "learning_rate": 1.1748247032108789e-05,
+      "loss": 0.9759,
       "step": 16237
     },
     {
-      "epoch": 0.44596413171843674,
+      "epoch": 0.460783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.221765152132725e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.1747342100850685e-05,
+      "loss": 0.8948,
       "step": 16238
     },
     {
-      "epoch": 0.44599159594627996,
+      "epoch": 0.46081157775255394,
       "grad_norm": 0.0,
-      "learning_rate": 1.2216784145396417e-05,
-      "loss": 0.9527,
+      "learning_rate": 1.1746437154832704e-05,
+      "loss": 0.9234,
       "step": 16239
     },
     {
-      "epoch": 0.4460190601741232,
+      "epoch": 0.4608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.221591675192553e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.1745532194062487e-05,
+      "loss": 0.8752,
       "step": 16240
     },
     {
-      "epoch": 0.44604652440196646,
+      "epoch": 0.4608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.2215049340921447e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.1744627218547683e-05,
+      "loss": 0.8531,
       "step": 16241
     },
     {
-      "epoch": 0.4460739886298097,
+      "epoch": 0.46089670828603857,
       "grad_norm": 0.0,
-      "learning_rate": 1.2214181912391038e-05,
-      "loss": 0.93,
+      "learning_rate": 1.1743722228295934e-05,
+      "loss": 0.9374,
       "step": 16242
     },
     {
-      "epoch": 0.4461014528576529,
+      "epoch": 0.4609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.2213314466341166e-05,
-      "loss": 1.0321,
+      "learning_rate": 1.1742817223314887e-05,
+      "loss": 0.8695,
       "step": 16243
     },
     {
-      "epoch": 0.4461289170854961,
+      "epoch": 0.4609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.2212447002778692e-05,
-      "loss": 0.8432,
+      "learning_rate": 1.174191220361218e-05,
+      "loss": 0.9908,
       "step": 16244
     },
     {
-      "epoch": 0.4461563813133394,
+      "epoch": 0.46098183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 1.2211579521710477e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.1741007169195466e-05,
+      "loss": 0.851,
       "step": 16245
     },
     {
-      "epoch": 0.4461838455411826,
+      "epoch": 0.4610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.2210712023143392e-05,
-      "loss": 0.8531,
+      "learning_rate": 1.1740102120072385e-05,
+      "loss": 1.0023,
       "step": 16246
     },
     {
-      "epoch": 0.44621130976902584,
+      "epoch": 0.46103859250851303,
       "grad_norm": 0.0,
-      "learning_rate": 1.2209844507084293e-05,
-      "loss": 0.916,
+      "learning_rate": 1.1739197056250582e-05,
+      "loss": 0.8974,
       "step": 16247
     },
     {
-      "epoch": 0.44623877399686906,
+      "epoch": 0.46106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.2208976973540053e-05,
-      "loss": 1.0002,
+      "learning_rate": 1.1738291977737706e-05,
+      "loss": 0.9608,
       "step": 16248
     },
     {
-      "epoch": 0.4462662382247123,
+      "epoch": 0.46109534619750286,
       "grad_norm": 0.0,
-      "learning_rate": 1.220810942251753e-05,
-      "loss": 1.0044,
+      "learning_rate": 1.17373868845414e-05,
+      "loss": 0.8551,
       "step": 16249
     },
     {
-      "epoch": 0.44629370245255556,
+      "epoch": 0.4611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.2207241854023586e-05,
-      "loss": 0.8486,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 0.8953,
       "step": 16250
     },
     {
-      "epoch": 0.4463211666803988,
+      "epoch": 0.46115209988649264,
       "grad_norm": 0.0,
-      "learning_rate": 1.2206374268065095e-05,
-      "loss": 1.0294,
+      "learning_rate": 1.173557665412907e-05,
+      "loss": 0.9034,
       "step": 16251
     },
     {
-      "epoch": 0.446348630908242,
+      "epoch": 0.4611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.2205506664648914e-05,
-      "loss": 1.0087,
+      "learning_rate": 1.1734671516928346e-05,
+      "loss": 0.9608,
       "step": 16252
     },
     {
-      "epoch": 0.4463760951360852,
+      "epoch": 0.4612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.220463904378191e-05,
-      "loss": 1.095,
+      "learning_rate": 1.173376636507477e-05,
+      "loss": 0.9077,
       "step": 16253
     },
     {
-      "epoch": 0.4464035593639285,
+      "epoch": 0.4612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.2203771405470947e-05,
-      "loss": 1.059,
+      "learning_rate": 1.1732861198575993e-05,
+      "loss": 0.9592,
       "step": 16254
     },
     {
-      "epoch": 0.4464310235917717,
+      "epoch": 0.4612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.220290374972289e-05,
-      "loss": 0.8575,
+      "learning_rate": 1.1731956017439661e-05,
+      "loss": 1.0182,
       "step": 16255
     },
     {
-      "epoch": 0.44645848781961495,
+      "epoch": 0.4612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.2202036076544609e-05,
-      "loss": 0.835,
+      "learning_rate": 1.1731050821673418e-05,
+      "loss": 0.8925,
       "step": 16256
     },
     {
-      "epoch": 0.44648595204745817,
+      "epoch": 0.46132236095346196,
       "grad_norm": 0.0,
-      "learning_rate": 1.2201168385942963e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.1730145611284907e-05,
+      "loss": 0.9707,
       "step": 16257
     },
     {
-      "epoch": 0.44651341627530144,
+      "epoch": 0.4613507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.2200300677924819e-05,
-      "loss": 0.8937,
+      "learning_rate": 1.1729240386281782e-05,
+      "loss": 0.8756,
       "step": 16258
     },
     {
-      "epoch": 0.44654088050314467,
+      "epoch": 0.46137911464245174,
       "grad_norm": 0.0,
-      "learning_rate": 1.2199432952497043e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.1728335146671683e-05,
+      "loss": 0.8728,
       "step": 16259
     },
     {
-      "epoch": 0.4465683447309879,
+      "epoch": 0.46140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.2198565209666509e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.1727429892462262e-05,
+      "loss": 0.8174,
       "step": 16260
     },
     {
-      "epoch": 0.4465958089588311,
+      "epoch": 0.46143586833144157,
       "grad_norm": 0.0,
-      "learning_rate": 1.219769744944007e-05,
-      "loss": 0.9285,
+      "learning_rate": 1.1726524623661161e-05,
+      "loss": 1.0594,
       "step": 16261
     },
     {
-      "epoch": 0.44662327318667433,
+      "epoch": 0.4614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.2196829671824595e-05,
-      "loss": 1.0131,
+      "learning_rate": 1.172561934027603e-05,
+      "loss": 0.8145,
       "step": 16262
     },
     {
-      "epoch": 0.4466507374145176,
+      "epoch": 0.46149262202043134,
       "grad_norm": 0.0,
-      "learning_rate": 1.2195961876826957e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.1724714042314515e-05,
+      "loss": 0.9141,
       "step": 16263
     },
     {
-      "epoch": 0.44667820164236083,
+      "epoch": 0.4615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.2195094064454016e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.1723808729784265e-05,
+      "loss": 0.7852,
       "step": 16264
     },
     {
-      "epoch": 0.44670566587020405,
+      "epoch": 0.4615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.2194226234712639e-05,
-      "loss": 0.932,
+      "learning_rate": 1.1722903402692922e-05,
+      "loss": 0.8866,
       "step": 16265
     },
     {
-      "epoch": 0.44673313009804727,
+      "epoch": 0.46157775255391603,
       "grad_norm": 0.0,
-      "learning_rate": 1.2193358387609698e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.1721998061048136e-05,
+      "loss": 0.924,
       "step": 16266
     },
     {
-      "epoch": 0.44676059432589055,
+      "epoch": 0.4616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.219249052315205e-05,
-      "loss": 0.8415,
+      "learning_rate": 1.1721092704857556e-05,
+      "loss": 0.8367,
       "step": 16267
     },
     {
-      "epoch": 0.44678805855373377,
+      "epoch": 0.4616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.2191622641346572e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.172018733412883e-05,
+      "loss": 1.053,
       "step": 16268
     },
     {
-      "epoch": 0.446815522781577,
+      "epoch": 0.46166288308740067,
       "grad_norm": 0.0,
-      "learning_rate": 1.2190754742200127e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.1719281948869605e-05,
+      "loss": 1.0277,
       "step": 16269
     },
     {
-      "epoch": 0.4468429870094202,
+      "epoch": 0.4616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.2189886825719579e-05,
-      "loss": 0.8816,
+      "learning_rate": 1.1718376549087525e-05,
+      "loss": 0.9961,
       "step": 16270
     },
     {
-      "epoch": 0.4468704512372635,
+      "epoch": 0.46171963677639044,
       "grad_norm": 0.0,
-      "learning_rate": 1.21890188919118e-05,
-      "loss": 1.0076,
+      "learning_rate": 1.1717471134790243e-05,
+      "loss": 0.8891,
       "step": 16271
     },
     {
-      "epoch": 0.4468979154651067,
+      "epoch": 0.46174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.2188150940783658e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.1716565705985405e-05,
+      "loss": 0.946,
       "step": 16272
     },
     {
-      "epoch": 0.44692537969294993,
+      "epoch": 0.46177639046538027,
       "grad_norm": 0.0,
-      "learning_rate": 1.2187282972342014e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.1715660262680657e-05,
+      "loss": 0.966,
       "step": 16273
     },
     {
-      "epoch": 0.44695284392079315,
+      "epoch": 0.46180476730987513,
       "grad_norm": 0.0,
-      "learning_rate": 1.2186414986593742e-05,
-      "loss": 0.8717,
+      "learning_rate": 1.171475480488365e-05,
+      "loss": 0.9424,
       "step": 16274
     },
     {
-      "epoch": 0.4469803081486364,
+      "epoch": 0.46183314415437005,
       "grad_norm": 0.0,
-      "learning_rate": 1.2185546983545706e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.1713849332602036e-05,
+      "loss": 0.8962,
       "step": 16275
     },
     {
-      "epoch": 0.44700777237647965,
+      "epoch": 0.4618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.2184678963204778e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.1712943845843459e-05,
+      "loss": 0.7607,
       "step": 16276
     },
     {
-      "epoch": 0.44703523660432287,
+      "epoch": 0.4618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.2183810925577822e-05,
-      "loss": 0.8544,
+      "learning_rate": 1.1712038344615564e-05,
+      "loss": 0.9572,
       "step": 16277
     },
     {
-      "epoch": 0.4470627008321661,
+      "epoch": 0.46191827468785474,
       "grad_norm": 0.0,
-      "learning_rate": 1.218294287067171e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.1711132828926007e-05,
+      "loss": 0.902,
       "step": 16278
     },
     {
-      "epoch": 0.4470901650600093,
+      "epoch": 0.4619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.2182074798493306e-05,
-      "loss": 1.0117,
+      "learning_rate": 1.1710227298782435e-05,
+      "loss": 0.8696,
       "step": 16279
     },
     {
-      "epoch": 0.4471176292878526,
+      "epoch": 0.4619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.2181206709049482e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.1709321754192492e-05,
+      "loss": 0.8722,
       "step": 16280
     },
     {
-      "epoch": 0.4471450935156958,
+      "epoch": 0.46200340522133937,
       "grad_norm": 0.0,
-      "learning_rate": 1.2180338602347107e-05,
-      "loss": 0.9816,
+      "learning_rate": 1.1708416195163836e-05,
+      "loss": 0.9744,
       "step": 16281
     },
     {
-      "epoch": 0.44717255774353903,
+      "epoch": 0.4620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.2179470478393045e-05,
-      "loss": 0.858,
+      "learning_rate": 1.170751062170411e-05,
+      "loss": 0.8165,
       "step": 16282
     },
     {
-      "epoch": 0.44720002197138226,
+      "epoch": 0.46206015891032914,
       "grad_norm": 0.0,
-      "learning_rate": 1.217860233719417e-05,
-      "loss": 0.9031,
+      "learning_rate": 1.1706605033820966e-05,
+      "loss": 0.9365,
       "step": 16283
     },
     {
-      "epoch": 0.44722748619922553,
+      "epoch": 0.46208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.2177734178757352e-05,
-      "loss": 0.9456,
+      "learning_rate": 1.1705699431522049e-05,
+      "loss": 0.8312,
       "step": 16284
     },
     {
-      "epoch": 0.44725495042706875,
+      "epoch": 0.462116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.2176866003089453e-05,
-      "loss": 0.9082,
+      "learning_rate": 1.1704793814815017e-05,
+      "loss": 0.8506,
       "step": 16285
     },
     {
-      "epoch": 0.447282414654912,
+      "epoch": 0.46214528944381383,
       "grad_norm": 0.0,
-      "learning_rate": 1.2175997810197351e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.1703888183707513e-05,
+      "loss": 0.9363,
       "step": 16286
     },
     {
-      "epoch": 0.4473098788827552,
+      "epoch": 0.46217366628830875,
       "grad_norm": 0.0,
-      "learning_rate": 1.2175129600087909e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.1702982538207186e-05,
+      "loss": 0.8241,
       "step": 16287
     },
     {
-      "epoch": 0.4473373431105985,
+      "epoch": 0.4622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.2174261372768002e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.1702076878321692e-05,
+      "loss": 0.9152,
       "step": 16288
     },
     {
-      "epoch": 0.4473648073384417,
+      "epoch": 0.4622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.2173393128244493e-05,
-      "loss": 0.8766,
+      "learning_rate": 1.1701171204058677e-05,
+      "loss": 0.9358,
       "step": 16289
     },
     {
-      "epoch": 0.4473922715662849,
+      "epoch": 0.46225879682179344,
       "grad_norm": 0.0,
-      "learning_rate": 1.2172524866524258e-05,
-      "loss": 0.8212,
+      "learning_rate": 1.1700265515425792e-05,
+      "loss": 0.7594,
       "step": 16290
     },
     {
-      "epoch": 0.44741973579412814,
+      "epoch": 0.4622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.2171656587614162e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.1699359812430689e-05,
+      "loss": 0.9168,
       "step": 16291
     },
     {
-      "epoch": 0.44744720002197136,
+      "epoch": 0.4623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.217078829152108e-05,
-      "loss": 0.8941,
+      "learning_rate": 1.1698454095081018e-05,
+      "loss": 0.906,
       "step": 16292
     },
     {
-      "epoch": 0.44747466424981464,
+      "epoch": 0.4623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.2169919978251883e-05,
-      "loss": 0.839,
+      "learning_rate": 1.1697548363384428e-05,
+      "loss": 0.9438,
       "step": 16293
     },
     {
-      "epoch": 0.44750212847765786,
+      "epoch": 0.462372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.2169051647813432e-05,
-      "loss": 0.9085,
+      "learning_rate": 1.169664261734857e-05,
+      "loss": 0.947,
       "step": 16294
     },
     {
-      "epoch": 0.4475295927055011,
+      "epoch": 0.4624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.2168183300212608e-05,
-      "loss": 1.0731,
+      "learning_rate": 1.1695736856981097e-05,
+      "loss": 0.9322,
       "step": 16295
     },
     {
-      "epoch": 0.4475570569333443,
+      "epoch": 0.46242905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 1.216731493545628e-05,
-      "loss": 0.8878,
+      "learning_rate": 1.1694831082289659e-05,
+      "loss": 0.7805,
       "step": 16296
     },
     {
-      "epoch": 0.4475845211611876,
+      "epoch": 0.4624574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.2166446553551312e-05,
-      "loss": 0.961,
+      "learning_rate": 1.1693925293281905e-05,
+      "loss": 0.794,
       "step": 16297
     },
     {
-      "epoch": 0.4476119853890308,
+      "epoch": 0.46248581157775254,
       "grad_norm": 0.0,
-      "learning_rate": 1.2165578154504582e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.1693019489965485e-05,
+      "loss": 0.8834,
       "step": 16298
     },
     {
-      "epoch": 0.447639449616874,
+      "epoch": 0.46251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.216470973832296e-05,
-      "loss": 1.0363,
+      "learning_rate": 1.1692113672348058e-05,
+      "loss": 0.8588,
       "step": 16299
     },
     {
-      "epoch": 0.44766691384471724,
+      "epoch": 0.4625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.2163841305013316e-05,
-      "loss": 0.8256,
+      "learning_rate": 1.1691207840437271e-05,
+      "loss": 0.9302,
       "step": 16300
     },
     {
-      "epoch": 0.4476943780725605,
+      "epoch": 0.4625709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.2162972854582518e-05,
-      "loss": 0.9909,
+      "learning_rate": 1.1690301994240772e-05,
+      "loss": 1.0101,
       "step": 16301
     },
     {
-      "epoch": 0.44772184230040374,
+      "epoch": 0.46259931895573214,
       "grad_norm": 0.0,
-      "learning_rate": 1.2162104387037444e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.1689396133766217e-05,
+      "loss": 0.8171,
       "step": 16302
     },
     {
-      "epoch": 0.44774930652824696,
+      "epoch": 0.462627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.216123590238496e-05,
-      "loss": 0.8829,
+      "learning_rate": 1.1688490259021258e-05,
+      "loss": 0.9855,
       "step": 16303
     },
     {
-      "epoch": 0.4477767707560902,
+      "epoch": 0.4626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.2160367400631943e-05,
-      "loss": 0.912,
+      "learning_rate": 1.1687584370013545e-05,
+      "loss": 0.9221,
       "step": 16304
     },
     {
-      "epoch": 0.4478042349839334,
+      "epoch": 0.4626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.2159498881785261e-05,
-      "loss": 0.881,
+      "learning_rate": 1.1686678466750729e-05,
+      "loss": 0.9641,
       "step": 16305
     },
     {
-      "epoch": 0.4478316992117767,
+      "epoch": 0.4627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.2158630345851787e-05,
-      "loss": 0.8579,
+      "learning_rate": 1.1685772549240469e-05,
+      "loss": 0.9628,
       "step": 16306
     },
     {
-      "epoch": 0.4478591634396199,
+      "epoch": 0.4627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.2157761792838396e-05,
-      "loss": 1.0277,
+      "learning_rate": 1.168486661749041e-05,
+      "loss": 0.9694,
       "step": 16307
     },
     {
-      "epoch": 0.4478866276674631,
+      "epoch": 0.46276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.2156893222751958e-05,
-      "loss": 0.8652,
+      "learning_rate": 1.1683960671508206e-05,
+      "loss": 0.8253,
       "step": 16308
     },
     {
-      "epoch": 0.44791409189530634,
+      "epoch": 0.4627979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.2156024635599342e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.1683054711301514e-05,
+      "loss": 0.9684,
       "step": 16309
     },
     {
-      "epoch": 0.4479415561231496,
+      "epoch": 0.46282633371169124,
       "grad_norm": 0.0,
-      "learning_rate": 1.2155156031387422e-05,
-      "loss": 0.9347,
+      "learning_rate": 1.168214873687798e-05,
+      "loss": 0.8419,
       "step": 16310
     },
     {
-      "epoch": 0.44796902035099284,
+      "epoch": 0.46285471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.2154287410123078e-05,
-      "loss": 0.8894,
+      "learning_rate": 1.168124274824526e-05,
+      "loss": 0.9348,
       "step": 16311
     },
     {
-      "epoch": 0.44799648457883606,
+      "epoch": 0.46288308740068107,
       "grad_norm": 0.0,
-      "learning_rate": 1.2153418771813174e-05,
-      "loss": 0.8981,
+      "learning_rate": 1.1680336745411005e-05,
+      "loss": 0.906,
       "step": 16312
     },
     {
-      "epoch": 0.4480239488066793,
+      "epoch": 0.46291146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 1.215255011646459e-05,
-      "loss": 1.0174,
+      "learning_rate": 1.1679430728382875e-05,
+      "loss": 0.8327,
       "step": 16313
     },
     {
-      "epoch": 0.44805141303452256,
+      "epoch": 0.46293984108967084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2151681444084194e-05,
-      "loss": 0.9599,
+      "learning_rate": 1.1678524697168513e-05,
+      "loss": 1.0353,
       "step": 16314
     },
     {
-      "epoch": 0.4480788772623658,
+      "epoch": 0.4629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.2150812754678858e-05,
-      "loss": 0.8615,
+      "learning_rate": 1.1677618651775582e-05,
+      "loss": 0.9576,
       "step": 16315
     },
     {
-      "epoch": 0.448106341490209,
+      "epoch": 0.4629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.2149944048255464e-05,
-      "loss": 0.9982,
+      "learning_rate": 1.1676712592211729e-05,
+      "loss": 0.8309,
       "step": 16316
     },
     {
-      "epoch": 0.4481338057180522,
+      "epoch": 0.4630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.2149075324820876e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.1675806518484609e-05,
+      "loss": 0.8663,
       "step": 16317
     },
     {
-      "epoch": 0.44816126994589545,
+      "epoch": 0.4630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.2148206584381969e-05,
-      "loss": 0.951,
+      "learning_rate": 1.1674900430601876e-05,
+      "loss": 0.9765,
       "step": 16318
     },
     {
-      "epoch": 0.4481887341737387,
+      "epoch": 0.4630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.2147337826945625e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.1673994328571186e-05,
+      "loss": 1.0224,
       "step": 16319
     },
     {
-      "epoch": 0.44821619840158194,
+      "epoch": 0.46311010215664017,
       "grad_norm": 0.0,
-      "learning_rate": 1.214646905251871e-05,
-      "loss": 0.9819,
+      "learning_rate": 1.1673088212400188e-05,
+      "loss": 0.8795,
       "step": 16320
     },
     {
-      "epoch": 0.44824366262942517,
+      "epoch": 0.4631384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.2145600261108099e-05,
-      "loss": 0.9491,
+      "learning_rate": 1.1672182082096538e-05,
+      "loss": 0.9177,
       "step": 16321
     },
     {
-      "epoch": 0.4482711268572684,
+      "epoch": 0.46316685584562994,
       "grad_norm": 0.0,
-      "learning_rate": 1.2144731452720671e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.1671275937667895e-05,
+      "loss": 0.8498,
       "step": 16322
     },
     {
-      "epoch": 0.44829859108511166,
+      "epoch": 0.46319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.2143862627363294e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.1670369779121904e-05,
+      "loss": 0.9351,
       "step": 16323
     },
     {
-      "epoch": 0.4483260553129549,
+      "epoch": 0.4632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.2142993785042847e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.1669463606466227e-05,
+      "loss": 0.8373,
       "step": 16324
     },
     {
-      "epoch": 0.4483535195407981,
+      "epoch": 0.46325198637911463,
       "grad_norm": 0.0,
-      "learning_rate": 1.2142124925766204e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.1668557419708514e-05,
+      "loss": 0.8202,
       "step": 16325
     },
     {
-      "epoch": 0.44838098376864133,
+      "epoch": 0.46328036322360955,
       "grad_norm": 0.0,
-      "learning_rate": 1.2141256049540236e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.1667651218856427e-05,
+      "loss": 0.9314,
       "step": 16326
     },
     {
-      "epoch": 0.4484084479964846,
+      "epoch": 0.4633087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.214038715637182e-05,
-      "loss": 0.9903,
+      "learning_rate": 1.1666745003917608e-05,
+      "loss": 0.9232,
       "step": 16327
     },
     {
-      "epoch": 0.4484359122243278,
+      "epoch": 0.4633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.2139518246267837e-05,
-      "loss": 0.9889,
+      "learning_rate": 1.166583877489972e-05,
+      "loss": 1.0433,
       "step": 16328
     },
     {
-      "epoch": 0.44846337645217105,
+      "epoch": 0.46336549375709424,
       "grad_norm": 0.0,
-      "learning_rate": 1.2138649319235152e-05,
-      "loss": 0.9923,
+      "learning_rate": 1.1664932531810422e-05,
+      "loss": 0.9662,
       "step": 16329
     },
     {
-      "epoch": 0.44849084068001427,
+      "epoch": 0.4633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.2137780375280645e-05,
-      "loss": 0.8587,
+      "learning_rate": 1.166402627465736e-05,
+      "loss": 1.0019,
       "step": 16330
     },
     {
-      "epoch": 0.4485183049078575,
+      "epoch": 0.463422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2136911414411194e-05,
-      "loss": 0.9367,
+      "learning_rate": 1.1663120003448195e-05,
+      "loss": 0.9045,
       "step": 16331
     },
     {
-      "epoch": 0.44854576913570077,
+      "epoch": 0.46345062429057887,
       "grad_norm": 0.0,
-      "learning_rate": 1.2136042436633672e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.166221371819058e-05,
+      "loss": 0.9522,
       "step": 16332
     },
     {
-      "epoch": 0.448573233363544,
+      "epoch": 0.4634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.2135173441954953e-05,
-      "loss": 0.9952,
+      "learning_rate": 1.1661307418892172e-05,
+      "loss": 0.8829,
       "step": 16333
     },
     {
-      "epoch": 0.4486006975913872,
+      "epoch": 0.46350737797956865,
       "grad_norm": 0.0,
-      "learning_rate": 1.2134304430381916e-05,
-      "loss": 1.0253,
+      "learning_rate": 1.1660401105560624e-05,
+      "loss": 0.9286,
       "step": 16334
     },
     {
-      "epoch": 0.44862816181923043,
+      "epoch": 0.46353575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.2133435401921434e-05,
-      "loss": 0.9946,
+      "learning_rate": 1.1659494778203593e-05,
+      "loss": 0.8949,
       "step": 16335
     },
     {
-      "epoch": 0.4486556260470737,
+      "epoch": 0.4635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.2132566356580388e-05,
-      "loss": 0.9833,
+      "learning_rate": 1.1658588436828738e-05,
+      "loss": 0.9201,
       "step": 16336
     },
     {
-      "epoch": 0.44868309027491693,
+      "epoch": 0.46359250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.213169729436565e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.1657682081443708e-05,
+      "loss": 0.8687,
       "step": 16337
     },
     {
-      "epoch": 0.44871055450276015,
+      "epoch": 0.46362088535754825,
       "grad_norm": 0.0,
-      "learning_rate": 1.2130828215284094e-05,
-      "loss": 0.9293,
+      "learning_rate": 1.1656775712056163e-05,
+      "loss": 0.9193,
       "step": 16338
     },
     {
-      "epoch": 0.44873801873060337,
+      "epoch": 0.4636492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.21299591193426e-05,
-      "loss": 0.8118,
+      "learning_rate": 1.1655869328673764e-05,
+      "loss": 1.0629,
       "step": 16339
     },
     {
-      "epoch": 0.44876548295844665,
+      "epoch": 0.463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.212909000654805e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.1654962931304159e-05,
+      "loss": 0.938,
       "step": 16340
     },
     {
-      "epoch": 0.44879294718628987,
+      "epoch": 0.46370601589103294,
       "grad_norm": 0.0,
-      "learning_rate": 1.2128220876907311e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.1654056519955005e-05,
+      "loss": 0.8706,
       "step": 16341
     },
     {
-      "epoch": 0.4488204114141331,
+      "epoch": 0.4637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.2127351730427264e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.1653150094633967e-05,
+      "loss": 0.9938,
       "step": 16342
     },
     {
-      "epoch": 0.4488478756419763,
+      "epoch": 0.4637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.2126482567114788e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.1652243655348691e-05,
+      "loss": 0.8869,
       "step": 16343
     },
     {
-      "epoch": 0.44887533986981953,
+      "epoch": 0.4637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.2125613386976758e-05,
-      "loss": 0.8652,
+      "learning_rate": 1.1651337202106841e-05,
+      "loss": 0.9071,
       "step": 16344
     },
     {
-      "epoch": 0.4489028040976628,
+      "epoch": 0.4638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.2124744190020047e-05,
-      "loss": 0.879,
+      "learning_rate": 1.165043073491607e-05,
+      "loss": 0.8456,
       "step": 16345
     },
     {
-      "epoch": 0.44893026832550603,
+      "epoch": 0.4638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.2123874976251542e-05,
-      "loss": 0.845,
+      "learning_rate": 1.1649524253784037e-05,
+      "loss": 0.9033,
       "step": 16346
     },
     {
-      "epoch": 0.44895773255334925,
+      "epoch": 0.46387627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.212300574567811e-05,
-      "loss": 1.0684,
+      "learning_rate": 1.16486177587184e-05,
+      "loss": 0.8334,
       "step": 16347
     },
     {
-      "epoch": 0.4489851967811925,
+      "epoch": 0.4639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.212213649830664e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.1647711249726813e-05,
+      "loss": 0.9618,
       "step": 16348
     },
     {
-      "epoch": 0.44901266100903575,
+      "epoch": 0.46393303064699204,
       "grad_norm": 0.0,
-      "learning_rate": 1.2121267234144003e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.1646804726816937e-05,
+      "loss": 0.9159,
       "step": 16349
     },
     {
-      "epoch": 0.449040125236879,
+      "epoch": 0.46396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.2120397953197075e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.1645898189996425e-05,
+      "loss": 0.8672,
       "step": 16350
     },
     {
-      "epoch": 0.4490675894647222,
+      "epoch": 0.4639897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.211952865547274e-05,
-      "loss": 0.986,
+      "learning_rate": 1.1644991639272938e-05,
+      "loss": 0.9454,
       "step": 16351
     },
     {
-      "epoch": 0.4490950536925654,
+      "epoch": 0.46401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.2118659340977874e-05,
-      "loss": 0.974,
+      "learning_rate": 1.164408507465413e-05,
+      "loss": 0.9098,
       "step": 16352
     },
     {
-      "epoch": 0.4491225179204087,
+      "epoch": 0.46404653802497164,
       "grad_norm": 0.0,
-      "learning_rate": 1.2117790009719353e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.1643178496147667e-05,
+      "loss": 0.9888,
       "step": 16353
     },
     {
-      "epoch": 0.4491499821482519,
+      "epoch": 0.4640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.2116920661704055e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.1642271903761198e-05,
+      "loss": 0.8764,
       "step": 16354
     },
     {
-      "epoch": 0.44917744637609514,
+      "epoch": 0.4641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.2116051296938862e-05,
-      "loss": 1.0543,
+      "learning_rate": 1.1641365297502384e-05,
+      "loss": 0.9444,
       "step": 16355
     },
     {
-      "epoch": 0.44920491060393836,
+      "epoch": 0.4641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2115181915430655e-05,
-      "loss": 0.9379,
+      "learning_rate": 1.1640458677378885e-05,
+      "loss": 0.9368,
       "step": 16356
     },
     {
-      "epoch": 0.4492323748317816,
+      "epoch": 0.4641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.2114312517186305e-05,
-      "loss": 0.8287,
+      "learning_rate": 1.163955204339836e-05,
+      "loss": 0.9489,
       "step": 16357
     },
     {
-      "epoch": 0.44925983905962485,
+      "epoch": 0.4641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.2113443102212697e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.1638645395568458e-05,
+      "loss": 0.9591,
       "step": 16358
     },
     {
-      "epoch": 0.4492873032874681,
+      "epoch": 0.46421679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.2112573670516705e-05,
-      "loss": 0.969,
+      "learning_rate": 1.1637738733896848e-05,
+      "loss": 0.9218,
       "step": 16359
     },
     {
-      "epoch": 0.4493147675153113,
+      "epoch": 0.4642451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.2111704222105216e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.1636832058391188e-05,
+      "loss": 0.9025,
       "step": 16360
     },
     {
-      "epoch": 0.4493422317431545,
+      "epoch": 0.46427355278093074,
       "grad_norm": 0.0,
-      "learning_rate": 1.2110834756985106e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.1635925369059131e-05,
+      "loss": 0.8811,
       "step": 16361
     },
     {
-      "epoch": 0.4493696959709978,
+      "epoch": 0.46430192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.210996527516325e-05,
-      "loss": 0.8031,
+      "learning_rate": 1.1635018665908342e-05,
+      "loss": 0.8931,
       "step": 16362
     },
     {
-      "epoch": 0.449397160198841,
+      "epoch": 0.4643303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.2109095776646534e-05,
-      "loss": 0.8545,
+      "learning_rate": 1.1634111948946475e-05,
+      "loss": 0.8418,
       "step": 16363
     },
     {
-      "epoch": 0.44942462442668424,
+      "epoch": 0.46435868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.2108226261441835e-05,
-      "loss": 0.9714,
+      "learning_rate": 1.1633205218181192e-05,
+      "loss": 0.9187,
       "step": 16364
     },
     {
-      "epoch": 0.44945208865452746,
+      "epoch": 0.46438706015891035,
       "grad_norm": 0.0,
-      "learning_rate": 1.2107356729556032e-05,
-      "loss": 0.956,
+      "learning_rate": 1.163229847362015e-05,
+      "loss": 0.9176,
       "step": 16365
     },
     {
-      "epoch": 0.44947955288237074,
+      "epoch": 0.4644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.2106487180996007e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.1631391715271008e-05,
+      "loss": 0.9575,
       "step": 16366
     },
     {
-      "epoch": 0.44950701711021396,
+      "epoch": 0.4644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.2105617615768639e-05,
-      "loss": 0.7824,
+      "learning_rate": 1.1630484943141428e-05,
+      "loss": 0.9806,
       "step": 16367
     },
     {
-      "epoch": 0.4495344813380572,
+      "epoch": 0.464472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.210474803388081e-05,
-      "loss": 0.9058,
+      "learning_rate": 1.1629578157239067e-05,
+      "loss": 0.9764,
       "step": 16368
     },
     {
-      "epoch": 0.4495619455659004,
+      "epoch": 0.4645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.2103878435339398e-05,
-      "loss": 0.9728,
+      "learning_rate": 1.1628671357571587e-05,
+      "loss": 0.8661,
       "step": 16369
     },
     {
-      "epoch": 0.4495894097937436,
+      "epoch": 0.4645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.2103008820151285e-05,
-      "loss": 0.8915,
+      "learning_rate": 1.1627764544146651e-05,
+      "loss": 0.8982,
       "step": 16370
     },
     {
-      "epoch": 0.4496168740215869,
+      "epoch": 0.46455732122587967,
       "grad_norm": 0.0,
-      "learning_rate": 1.2102139188323351e-05,
-      "loss": 1.0767,
+      "learning_rate": 1.1626857716971908e-05,
+      "loss": 0.8892,
       "step": 16371
     },
     {
-      "epoch": 0.4496443382494301,
+      "epoch": 0.4645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.210126953986248e-05,
-      "loss": 1.0101,
+      "learning_rate": 1.162595087605503e-05,
+      "loss": 0.8473,
       "step": 16372
     },
     {
-      "epoch": 0.44967180247727334,
+      "epoch": 0.46461407491486945,
       "grad_norm": 0.0,
-      "learning_rate": 1.210039987477555e-05,
-      "loss": 0.7957,
+      "learning_rate": 1.162504402140367e-05,
+      "loss": 0.7432,
       "step": 16373
     },
     {
-      "epoch": 0.44969926670511656,
+      "epoch": 0.46464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.209953019306944e-05,
-      "loss": 0.9746,
+      "learning_rate": 1.162413715302549e-05,
+      "loss": 0.9501,
       "step": 16374
     },
     {
-      "epoch": 0.44972673093295984,
+      "epoch": 0.4646708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.2098660494751035e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.162323027092815e-05,
+      "loss": 0.9172,
       "step": 16375
     },
     {
-      "epoch": 0.44975419516080306,
+      "epoch": 0.46469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.2097790779827216e-05,
-      "loss": 0.8498,
+      "learning_rate": 1.1622323375119312e-05,
+      "loss": 0.9611,
       "step": 16376
     },
     {
-      "epoch": 0.4497816593886463,
+      "epoch": 0.46472758229284905,
       "grad_norm": 0.0,
-      "learning_rate": 1.2096921048304864e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.1621416465606634e-05,
+      "loss": 1.0099,
       "step": 16377
     },
     {
-      "epoch": 0.4498091236164895,
+      "epoch": 0.4647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.2096051300190859e-05,
-      "loss": 0.8497,
+      "learning_rate": 1.162050954239778e-05,
+      "loss": 0.9394,
       "step": 16378
     },
     {
-      "epoch": 0.4498365878443328,
+      "epoch": 0.4647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.2095181535492085e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.1619602605500408e-05,
+      "loss": 0.8558,
       "step": 16379
     },
     {
-      "epoch": 0.449864052072176,
+      "epoch": 0.4648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.2094311754215421e-05,
-      "loss": 0.934,
+      "learning_rate": 1.161869565492218e-05,
+      "loss": 0.9245,
       "step": 16380
     },
     {
-      "epoch": 0.4498915163000192,
+      "epoch": 0.4648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.2093441956367756e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.1617788690670756e-05,
+      "loss": 0.984,
       "step": 16381
     },
     {
-      "epoch": 0.44991898052786244,
+      "epoch": 0.4648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.2092572141955963e-05,
-      "loss": 0.9025,
+      "learning_rate": 1.16168817127538e-05,
+      "loss": 0.9095,
       "step": 16382
     },
     {
-      "epoch": 0.44994644475570567,
+      "epoch": 0.4648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.2091702310986929e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.1615974721178972e-05,
+      "loss": 0.8662,
       "step": 16383
     },
     {
-      "epoch": 0.44997390898354894,
+      "epoch": 0.4649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.2090832463467539e-05,
-      "loss": 0.9232,
+      "learning_rate": 1.1615067715953931e-05,
+      "loss": 0.8559,
       "step": 16384
     },
     {
-      "epoch": 0.45000137321139216,
+      "epoch": 0.46495459704880815,
       "grad_norm": 0.0,
-      "learning_rate": 1.208996259940467e-05,
-      "loss": 0.903,
+      "learning_rate": 1.161416069708634e-05,
+      "loss": 0.9886,
       "step": 16385
     },
     {
-      "epoch": 0.4500288374392354,
+      "epoch": 0.46498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.2089092718805206e-05,
-      "loss": 0.954,
+      "learning_rate": 1.1613253664583864e-05,
+      "loss": 0.9401,
       "step": 16386
     },
     {
-      "epoch": 0.4500563016670786,
+      "epoch": 0.465011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.2088222821676034e-05,
-      "loss": 0.9962,
+      "learning_rate": 1.161234661845416e-05,
+      "loss": 1.0137,
       "step": 16387
     },
     {
-      "epoch": 0.4500837658949219,
+      "epoch": 0.46503972758229284,
       "grad_norm": 0.0,
-      "learning_rate": 1.2087352908024035e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.161143955870489e-05,
+      "loss": 0.8605,
       "step": 16388
     },
     {
-      "epoch": 0.4501112301227651,
+      "epoch": 0.46506810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 1.208648297785609e-05,
-      "loss": 1.1248,
+      "learning_rate": 1.161053248534372e-05,
+      "loss": 0.8595,
       "step": 16389
     },
     {
-      "epoch": 0.4501386943506083,
+      "epoch": 0.4650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.2085613031179082e-05,
-      "loss": 0.8839,
+      "learning_rate": 1.1609625398378308e-05,
+      "loss": 0.7185,
       "step": 16390
     },
     {
-      "epoch": 0.45016615857845155,
+      "epoch": 0.46512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.2084743067999895e-05,
-      "loss": 0.8559,
+      "learning_rate": 1.1608718297816322e-05,
+      "loss": 0.9144,
       "step": 16391
     },
     {
-      "epoch": 0.4501936228062948,
+      "epoch": 0.46515323496027244,
       "grad_norm": 0.0,
-      "learning_rate": 1.2083873088325415e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.1607811183665414e-05,
+      "loss": 0.9603,
       "step": 16392
     },
     {
-      "epoch": 0.45022108703413805,
+      "epoch": 0.4651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.2083003092162527e-05,
-      "loss": 0.8856,
+      "learning_rate": 1.1606904055933256e-05,
+      "loss": 0.8436,
       "step": 16393
     },
     {
-      "epoch": 0.45024855126198127,
+      "epoch": 0.4652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.2082133079518106e-05,
-      "loss": 0.898,
+      "learning_rate": 1.1605996914627509e-05,
+      "loss": 0.9214,
       "step": 16394
     },
     {
-      "epoch": 0.4502760154898245,
+      "epoch": 0.4652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.2081263050399044e-05,
-      "loss": 0.8874,
+      "learning_rate": 1.160508975975583e-05,
+      "loss": 0.8819,
       "step": 16395
     },
     {
-      "epoch": 0.45030347971766777,
+      "epoch": 0.465266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.2080393004812226e-05,
-      "loss": 0.8436,
+      "learning_rate": 1.1604182591325888e-05,
+      "loss": 0.9065,
       "step": 16396
     },
     {
-      "epoch": 0.450330943945511,
+      "epoch": 0.46529511918274685,
       "grad_norm": 0.0,
-      "learning_rate": 1.2079522942764528e-05,
-      "loss": 0.8928,
+      "learning_rate": 1.1603275409345343e-05,
+      "loss": 0.8972,
       "step": 16397
     },
     {
-      "epoch": 0.4503584081733542,
+      "epoch": 0.46532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.2078652864262838e-05,
-      "loss": 0.7699,
+      "learning_rate": 1.160236821382186e-05,
+      "loss": 0.8396,
       "step": 16398
     },
     {
-      "epoch": 0.45038587240119743,
+      "epoch": 0.4653518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.2077782769314045e-05,
-      "loss": 0.905,
+      "learning_rate": 1.1601461004763097e-05,
+      "loss": 0.8979,
       "step": 16399
     },
     {
-      "epoch": 0.45041333662904065,
+      "epoch": 0.46538024971623154,
       "grad_norm": 0.0,
-      "learning_rate": 1.2076912657925028e-05,
-      "loss": 0.9292,
+      "learning_rate": 1.1600553782176724e-05,
+      "loss": 0.8728,
       "step": 16400
     },
     {
-      "epoch": 0.4504408008568839,
+      "epoch": 0.46540862656072646,
       "grad_norm": 0.0,
-      "learning_rate": 1.2076042530102674e-05,
-      "loss": 0.857,
+      "learning_rate": 1.15996465460704e-05,
+      "loss": 0.9577,
       "step": 16401
     },
     {
-      "epoch": 0.45046826508472715,
+      "epoch": 0.4654370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.2075172385853868e-05,
-      "loss": 0.7526,
+      "learning_rate": 1.1598739296451789e-05,
+      "loss": 0.8661,
       "step": 16402
     },
     {
-      "epoch": 0.45049572931257037,
+      "epoch": 0.46546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.2074302225185492e-05,
-      "loss": 0.8714,
+      "learning_rate": 1.1597832033328558e-05,
+      "loss": 0.8475,
       "step": 16403
     },
     {
-      "epoch": 0.4505231935404136,
+      "epoch": 0.46549375709421115,
       "grad_norm": 0.0,
-      "learning_rate": 1.2073432048104437e-05,
-      "loss": 0.914,
+      "learning_rate": 1.1596924756708366e-05,
+      "loss": 0.8693,
       "step": 16404
     },
     {
-      "epoch": 0.45055065776825687,
+      "epoch": 0.465522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.2072561854617582e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.1596017466598879e-05,
+      "loss": 0.8738,
       "step": 16405
     },
     {
-      "epoch": 0.4505781219961001,
+      "epoch": 0.4655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.2071691644731815e-05,
-      "loss": 0.9636,
+      "learning_rate": 1.159511016300776e-05,
+      "loss": 0.892,
       "step": 16406
     },
     {
-      "epoch": 0.4506055862239433,
+      "epoch": 0.4655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.2070821418454019e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.1594202845942674e-05,
+      "loss": 0.8705,
       "step": 16407
     },
     {
-      "epoch": 0.45063305045178653,
+      "epoch": 0.4656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.2069951175791087e-05,
-      "loss": 0.9716,
+      "learning_rate": 1.1593295515411287e-05,
+      "loss": 0.896,
       "step": 16408
     },
     {
-      "epoch": 0.4506605146796298,
+      "epoch": 0.4656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.2069080916749896e-05,
-      "loss": 0.871,
+      "learning_rate": 1.1592388171421259e-05,
+      "loss": 0.8556,
       "step": 16409
     },
     {
-      "epoch": 0.45068797890747303,
+      "epoch": 0.46566401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 1.2068210641337335e-05,
-      "loss": 0.8895,
+      "learning_rate": 1.1591480813980257e-05,
+      "loss": 0.9053,
       "step": 16410
     },
     {
-      "epoch": 0.45071544313531625,
+      "epoch": 0.4656923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.2067340349560292e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.1590573443095947e-05,
+      "loss": 0.8126,
       "step": 16411
     },
     {
-      "epoch": 0.4507429073631595,
+      "epoch": 0.46572077185017025,
       "grad_norm": 0.0,
-      "learning_rate": 1.206647004142565e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.1589666058775986e-05,
+      "loss": 0.8944,
       "step": 16412
     },
     {
-      "epoch": 0.4507703715910027,
+      "epoch": 0.46574914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.2065599716940297e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.158875866102805e-05,
+      "loss": 0.9091,
       "step": 16413
     },
     {
-      "epoch": 0.45079783581884597,
+      "epoch": 0.46577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.206472937611112e-05,
-      "loss": 0.9867,
+      "learning_rate": 1.1587851249859795e-05,
+      "loss": 0.8676,
       "step": 16414
     },
     {
-      "epoch": 0.4508253000466892,
+      "epoch": 0.46580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.2063859018945e-05,
-      "loss": 0.8773,
+      "learning_rate": 1.1586943825278892e-05,
+      "loss": 0.8319,
       "step": 16415
     },
     {
-      "epoch": 0.4508527642745324,
+      "epoch": 0.46583427922814985,
       "grad_norm": 0.0,
-      "learning_rate": 1.2062988645448835e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.1586036387293e-05,
+      "loss": 0.8285,
       "step": 16416
     },
     {
-      "epoch": 0.45088022850237564,
+      "epoch": 0.4658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.2062118255629502e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.158512893590979e-05,
+      "loss": 0.9255,
       "step": 16417
     },
     {
-      "epoch": 0.4509076927302189,
+      "epoch": 0.4658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.2061247849493887e-05,
-      "loss": 0.8851,
+      "learning_rate": 1.1584221471136924e-05,
+      "loss": 0.9077,
       "step": 16418
     },
     {
-      "epoch": 0.45093515695806213,
+      "epoch": 0.4659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.2060377427048882e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.1583313992982067e-05,
+      "loss": 0.9119,
       "step": 16419
     },
     {
-      "epoch": 0.45096262118590535,
+      "epoch": 0.4659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.2059506988301377e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.1582406501452886e-05,
+      "loss": 0.9488,
       "step": 16420
     },
     {
-      "epoch": 0.4509900854137486,
+      "epoch": 0.4659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.205863653325825e-05,
-      "loss": 0.8576,
+      "learning_rate": 1.1581498996557046e-05,
+      "loss": 0.8949,
       "step": 16421
     },
     {
-      "epoch": 0.45101754964159185,
+      "epoch": 0.4660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.2057766061926394e-05,
-      "loss": 0.8999,
+      "learning_rate": 1.1580591478302212e-05,
+      "loss": 0.9144,
       "step": 16422
     },
     {
-      "epoch": 0.4510450138694351,
+      "epoch": 0.4660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.20568955743127e-05,
-      "loss": 1.039,
+      "learning_rate": 1.1579683946696054e-05,
+      "loss": 0.9346,
       "step": 16423
     },
     {
-      "epoch": 0.4510724780972783,
+      "epoch": 0.46606129398410895,
       "grad_norm": 0.0,
-      "learning_rate": 1.2056025070424048e-05,
-      "loss": 0.8427,
+      "learning_rate": 1.1578776401746232e-05,
+      "loss": 0.8943,
       "step": 16424
     },
     {
-      "epoch": 0.4510999423251215,
+      "epoch": 0.46608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.205515455026733e-05,
-      "loss": 1.0288,
+      "learning_rate": 1.1577868843460415e-05,
+      "loss": 0.923,
       "step": 16425
     },
     {
-      "epoch": 0.45112740655296474,
+      "epoch": 0.4661180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.2054284013849431e-05,
-      "loss": 0.888,
+      "learning_rate": 1.1576961271846269e-05,
+      "loss": 0.9435,
       "step": 16426
     },
     {
-      "epoch": 0.451154870780808,
+      "epoch": 0.46614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.2053413461177243e-05,
-      "loss": 0.9365,
+      "learning_rate": 1.1576053686911459e-05,
+      "loss": 0.9007,
       "step": 16427
     },
     {
-      "epoch": 0.45118233500865124,
+      "epoch": 0.46617480136208855,
       "grad_norm": 0.0,
-      "learning_rate": 1.2052542892257653e-05,
-      "loss": 0.9993,
+      "learning_rate": 1.1575146088663656e-05,
+      "loss": 0.9222,
       "step": 16428
     },
     {
-      "epoch": 0.45120979923649446,
+      "epoch": 0.4662031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.2051672307097548e-05,
-      "loss": 0.9796,
+      "learning_rate": 1.157423847711052e-05,
+      "loss": 0.949,
       "step": 16429
     },
     {
-      "epoch": 0.4512372634643377,
+      "epoch": 0.46623155505107833,
       "grad_norm": 0.0,
-      "learning_rate": 1.2050801705703817e-05,
-      "loss": 0.9841,
+      "learning_rate": 1.1573330852259724e-05,
+      "loss": 0.8777,
       "step": 16430
     },
     {
-      "epoch": 0.45126472769218096,
+      "epoch": 0.4662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.2049931088083344e-05,
-      "loss": 0.9932,
+      "learning_rate": 1.157242321411893e-05,
+      "loss": 0.9108,
       "step": 16431
     },
     {
-      "epoch": 0.4512921919200242,
+      "epoch": 0.4662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.2049060454243031e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.1571515562695805e-05,
+      "loss": 0.9149,
       "step": 16432
     },
     {
-      "epoch": 0.4513196561478674,
+      "epoch": 0.466316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.2048189804189752e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.1570607897998018e-05,
+      "loss": 1.0117,
       "step": 16433
     },
     {
-      "epoch": 0.4513471203757106,
+      "epoch": 0.4663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.2047319137930404e-05,
-      "loss": 0.9738,
+      "learning_rate": 1.1569700220033237e-05,
+      "loss": 0.8988,
       "step": 16434
     },
     {
-      "epoch": 0.4513745846035539,
+      "epoch": 0.4663734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.204644845547187e-05,
-      "loss": 0.8194,
+      "learning_rate": 1.1568792528809127e-05,
+      "loss": 0.9345,
       "step": 16435
     },
     {
-      "epoch": 0.4514020488313971,
+      "epoch": 0.46640181611804765,
       "grad_norm": 0.0,
-      "learning_rate": 1.2045577756821047e-05,
-      "loss": 0.903,
+      "learning_rate": 1.1567884824333353e-05,
+      "loss": 0.7209,
       "step": 16436
     },
     {
-      "epoch": 0.45142951305924034,
+      "epoch": 0.46643019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.2044707041984822e-05,
-      "loss": 0.8744,
+      "learning_rate": 1.156697710661359e-05,
+      "loss": 0.909,
       "step": 16437
     },
     {
-      "epoch": 0.45145697728708356,
+      "epoch": 0.4664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.2043836310970078e-05,
-      "loss": 0.8241,
+      "learning_rate": 1.15660693756575e-05,
+      "loss": 0.9308,
       "step": 16438
     },
     {
-      "epoch": 0.4514844415149268,
+      "epoch": 0.46648694665153234,
       "grad_norm": 0.0,
-      "learning_rate": 1.2042965563783712e-05,
-      "loss": 0.9018,
+      "learning_rate": 1.156516163147275e-05,
+      "loss": 0.7232,
       "step": 16439
     },
     {
-      "epoch": 0.45151190574277006,
+      "epoch": 0.46651532349602726,
       "grad_norm": 0.0,
-      "learning_rate": 1.2042094800432613e-05,
-      "loss": 0.9138,
+      "learning_rate": 1.156425387406701e-05,
+      "loss": 0.8658,
       "step": 16440
     },
     {
-      "epoch": 0.4515393699706133,
+      "epoch": 0.4665437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.2041224020923665e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.1563346103447948e-05,
+      "loss": 0.8863,
       "step": 16441
     },
     {
-      "epoch": 0.4515668341984565,
+      "epoch": 0.46657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.2040353225263763e-05,
-      "loss": 0.8834,
+      "learning_rate": 1.1562438319623231e-05,
+      "loss": 0.89,
       "step": 16442
     },
     {
-      "epoch": 0.4515942984262997,
+      "epoch": 0.4666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.2039482413459798e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.1561530522600524e-05,
+      "loss": 0.9858,
       "step": 16443
     },
     {
-      "epoch": 0.451621762654143,
+      "epoch": 0.4666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.2038611585518659e-05,
-      "loss": 0.7799,
+      "learning_rate": 1.1560622712387502e-05,
+      "loss": 1.0635,
       "step": 16444
     },
     {
-      "epoch": 0.4516492268819862,
+      "epoch": 0.4666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.203774074144723e-05,
-      "loss": 0.9337,
+      "learning_rate": 1.1559714888991831e-05,
+      "loss": 0.8614,
       "step": 16445
     },
     {
-      "epoch": 0.45167669110982944,
+      "epoch": 0.4666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.2036869881252411e-05,
-      "loss": 0.98,
+      "learning_rate": 1.1558807052421174e-05,
+      "loss": 0.9047,
       "step": 16446
     },
     {
-      "epoch": 0.45170415533767266,
+      "epoch": 0.4667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.2035999004941087e-05,
-      "loss": 0.8323,
+      "learning_rate": 1.1557899202683208e-05,
+      "loss": 0.918,
       "step": 16447
     },
     {
-      "epoch": 0.45173161956551594,
+      "epoch": 0.46674233825198636,
       "grad_norm": 0.0,
-      "learning_rate": 1.203512811252015e-05,
-      "loss": 1.0031,
+      "learning_rate": 1.1556991339785595e-05,
+      "loss": 0.9861,
       "step": 16448
     },
     {
-      "epoch": 0.45175908379335916,
+      "epoch": 0.46677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.2034257203996493e-05,
-      "loss": 0.9584,
+      "learning_rate": 1.1556083463736005e-05,
+      "loss": 0.8835,
       "step": 16449
     },
     {
-      "epoch": 0.4517865480212024,
+      "epoch": 0.4667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.2033386279377005e-05,
-      "loss": 0.8495,
+      "learning_rate": 1.1555175574542108e-05,
+      "loss": 0.9417,
       "step": 16450
     },
     {
-      "epoch": 0.4518140122490456,
+      "epoch": 0.46682746878547104,
       "grad_norm": 0.0,
-      "learning_rate": 1.2032515338668573e-05,
-      "loss": 0.822,
+      "learning_rate": 1.1554267672211576e-05,
+      "loss": 0.9302,
       "step": 16451
     },
     {
-      "epoch": 0.4518414764768888,
+      "epoch": 0.46685584562996596,
       "grad_norm": 0.0,
-      "learning_rate": 1.2031644381878098e-05,
-      "loss": 0.8509,
+      "learning_rate": 1.1553359756752074e-05,
+      "loss": 0.8166,
       "step": 16452
     },
     {
-      "epoch": 0.4518689407047321,
+      "epoch": 0.4668842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.2030773409012462e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.1552451828171271e-05,
+      "loss": 0.9097,
       "step": 16453
     },
     {
-      "epoch": 0.4518964049325753,
+      "epoch": 0.46691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.202990242007856e-05,
-      "loss": 0.931,
+      "learning_rate": 1.155154388647684e-05,
+      "loss": 0.9465,
       "step": 16454
     },
     {
-      "epoch": 0.45192386916041855,
+      "epoch": 0.46694097616345065,
       "grad_norm": 0.0,
-      "learning_rate": 1.2029031415083286e-05,
-      "loss": 0.9267,
+      "learning_rate": 1.1550635931676446e-05,
+      "loss": 0.8157,
       "step": 16455
     },
     {
-      "epoch": 0.45195133338826177,
+      "epoch": 0.4669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.202816039403353e-05,
-      "loss": 0.8689,
+      "learning_rate": 1.1549727963777761e-05,
+      "loss": 0.8796,
       "step": 16456
     },
     {
-      "epoch": 0.45197879761610504,
+      "epoch": 0.4669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.2027289356936178e-05,
-      "loss": 0.8212,
+      "learning_rate": 1.1548819982788455e-05,
+      "loss": 0.889,
       "step": 16457
     },
     {
-      "epoch": 0.45200626184394826,
+      "epoch": 0.4670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.2026418303798133e-05,
-      "loss": 0.9065,
+      "learning_rate": 1.1547911988716194e-05,
+      "loss": 0.9272,
       "step": 16458
     },
     {
-      "epoch": 0.4520337260717915,
+      "epoch": 0.4670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.2025547234626279e-05,
-      "loss": 0.862,
+      "learning_rate": 1.1547003981568654e-05,
+      "loss": 0.9299,
       "step": 16459
     },
     {
-      "epoch": 0.4520611902996347,
+      "epoch": 0.46708286038592506,
       "grad_norm": 0.0,
-      "learning_rate": 1.202467614942751e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.15460959613535e-05,
+      "loss": 0.897,
       "step": 16460
     },
     {
-      "epoch": 0.452088654527478,
+      "epoch": 0.46711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.2023805048208721e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.1545187928078407e-05,
+      "loss": 0.9211,
       "step": 16461
     },
     {
-      "epoch": 0.4521161187553212,
+      "epoch": 0.4671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.20229339309768e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.1544279881751042e-05,
+      "loss": 0.8841,
       "step": 16462
     },
     {
-      "epoch": 0.4521435829831644,
+      "epoch": 0.46716799091940975,
       "grad_norm": 0.0,
-      "learning_rate": 1.2022062797738642e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.154337182237907e-05,
+      "loss": 0.8645,
       "step": 16463
     },
     {
-      "epoch": 0.45217104721100765,
+      "epoch": 0.46719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.2021191648501147e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.1542463749970172e-05,
+      "loss": 0.8961,
       "step": 16464
     },
     {
-      "epoch": 0.45219851143885087,
+      "epoch": 0.4672247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.2020320483271193e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.1541555664532011e-05,
+      "loss": 0.8619,
       "step": 16465
     },
     {
-      "epoch": 0.45222597566669415,
+      "epoch": 0.46725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.201944930205568e-05,
-      "loss": 0.8303,
+      "learning_rate": 1.1540647566072259e-05,
+      "loss": 0.8127,
       "step": 16466
     },
     {
-      "epoch": 0.45225343989453737,
+      "epoch": 0.46728149829738935,
       "grad_norm": 0.0,
-      "learning_rate": 1.2018578104861507e-05,
-      "loss": 0.8955,
+      "learning_rate": 1.153973945459859e-05,
+      "loss": 0.8131,
       "step": 16467
     },
     {
-      "epoch": 0.4522809041223806,
+      "epoch": 0.4673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.2017706891695561e-05,
-      "loss": 0.8748,
+      "learning_rate": 1.1538831330118674e-05,
+      "loss": 0.9215,
       "step": 16468
     },
     {
-      "epoch": 0.4523083683502238,
+      "epoch": 0.4673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.2016835662564734e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.1537923192640178e-05,
+      "loss": 0.8759,
       "step": 16469
     },
     {
-      "epoch": 0.4523358325780671,
+      "epoch": 0.467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.2015964417475926e-05,
-      "loss": 0.8309,
+      "learning_rate": 1.1537015042170776e-05,
+      "loss": 0.7792,
       "step": 16470
     },
     {
-      "epoch": 0.4523632968059103,
+      "epoch": 0.4673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.2015093156436022e-05,
-      "loss": 0.8203,
+      "learning_rate": 1.1536106878718138e-05,
+      "loss": 0.904,
       "step": 16471
     },
     {
-      "epoch": 0.45239076103375353,
+      "epoch": 0.4674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.2014221879451924e-05,
-      "loss": 0.9623,
+      "learning_rate": 1.153519870228994e-05,
+      "loss": 0.926,
       "step": 16472
     },
     {
-      "epoch": 0.45241822526159675,
+      "epoch": 0.4674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.2013350586530521e-05,
-      "loss": 0.984,
+      "learning_rate": 1.1534290512893844e-05,
+      "loss": 0.9424,
       "step": 16473
     },
     {
-      "epoch": 0.45244568948944003,
+      "epoch": 0.4674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.2012479277678706e-05,
-      "loss": 0.9611,
+      "learning_rate": 1.1533382310537531e-05,
+      "loss": 0.8559,
       "step": 16474
     },
     {
-      "epoch": 0.45247315371728325,
+      "epoch": 0.46750851305334845,
       "grad_norm": 0.0,
-      "learning_rate": 1.2011607952903376e-05,
-      "loss": 0.7738,
+      "learning_rate": 1.1532474095228667e-05,
+      "loss": 0.893,
       "step": 16475
     },
     {
-      "epoch": 0.45250061794512647,
+      "epoch": 0.46753688989784337,
       "grad_norm": 0.0,
-      "learning_rate": 1.2010736612211428e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.1531565866974926e-05,
+      "loss": 1.0237,
       "step": 16476
     },
     {
-      "epoch": 0.4525280821729697,
+      "epoch": 0.4675652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.2009865255609752e-05,
-      "loss": 0.9698,
+      "learning_rate": 1.153065762578398e-05,
+      "loss": 0.8652,
       "step": 16477
     },
     {
-      "epoch": 0.4525555464008129,
+      "epoch": 0.46759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.200899388310524e-05,
-      "loss": 0.8591,
+      "learning_rate": 1.1529749371663498e-05,
+      "loss": 0.8726,
       "step": 16478
     },
     {
-      "epoch": 0.4525830106286562,
+      "epoch": 0.46762202043132806,
       "grad_norm": 0.0,
-      "learning_rate": 1.2008122494704795e-05,
-      "loss": 0.8626,
+      "learning_rate": 1.1528841104621155e-05,
+      "loss": 0.962,
       "step": 16479
     },
     {
-      "epoch": 0.4526104748564994,
+      "epoch": 0.4676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.2007251090415304e-05,
-      "loss": 0.9626,
+      "learning_rate": 1.1527932824664624e-05,
+      "loss": 0.8832,
       "step": 16480
     },
     {
-      "epoch": 0.45263793908434263,
+      "epoch": 0.46767877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.2006379670243664e-05,
-      "loss": 0.8042,
+      "learning_rate": 1.1527024531801576e-05,
+      "loss": 0.8797,
       "step": 16481
     },
     {
-      "epoch": 0.45266540331218585,
+      "epoch": 0.4677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.2005508234196774e-05,
-      "loss": 0.9159,
+      "learning_rate": 1.1526116226039683e-05,
+      "loss": 0.9489,
       "step": 16482
     },
     {
-      "epoch": 0.45269286754002913,
+      "epoch": 0.4677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.2004636782281522e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.1525207907386615e-05,
+      "loss": 0.8302,
       "step": 16483
     },
     {
-      "epoch": 0.45272033176787235,
+      "epoch": 0.4677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.2003765314504812e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.1524299575850048e-05,
+      "loss": 0.877,
       "step": 16484
     },
     {
-      "epoch": 0.4527477959957156,
+      "epoch": 0.4677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.2002893830873531e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.1523391231437657e-05,
+      "loss": 0.9542,
       "step": 16485
     },
     {
-      "epoch": 0.4527752602235588,
+      "epoch": 0.4678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.2002022331394581e-05,
-      "loss": 0.9557,
+      "learning_rate": 1.1522482874157108e-05,
+      "loss": 0.9932,
       "step": 16486
     },
     {
-      "epoch": 0.45280272445140207,
+      "epoch": 0.46784903518728715,
       "grad_norm": 0.0,
-      "learning_rate": 1.2001150816074851e-05,
-      "loss": 1.0066,
+      "learning_rate": 1.1521574504016082e-05,
+      "loss": 0.8757,
       "step": 16487
     },
     {
-      "epoch": 0.4528301886792453,
+      "epoch": 0.46787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.2000279284921247e-05,
-      "loss": 0.91,
+      "learning_rate": 1.1520666121022246e-05,
+      "loss": 0.9525,
       "step": 16488
     },
     {
-      "epoch": 0.4528576529070885,
+      "epoch": 0.467905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.1999407737940652e-05,
-      "loss": 0.8134,
+      "learning_rate": 1.1519757725183272e-05,
+      "loss": 0.922,
       "step": 16489
     },
     {
-      "epoch": 0.45288511713493174,
+      "epoch": 0.46793416572077184,
       "grad_norm": 0.0,
-      "learning_rate": 1.199853617513997e-05,
-      "loss": 0.9648,
+      "learning_rate": 1.1518849316506837e-05,
+      "loss": 0.8588,
       "step": 16490
     },
     {
-      "epoch": 0.452912581362775,
+      "epoch": 0.46796254256526676,
       "grad_norm": 0.0,
-      "learning_rate": 1.1997664596526097e-05,
-      "loss": 1.0048,
+      "learning_rate": 1.1517940895000616e-05,
+      "loss": 0.7337,
       "step": 16491
     },
     {
-      "epoch": 0.45294004559061823,
+      "epoch": 0.4679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.1996793002105928e-05,
-      "loss": 0.9418,
+      "learning_rate": 1.151703246067228e-05,
+      "loss": 0.9023,
       "step": 16492
     },
     {
-      "epoch": 0.45296750981846146,
+      "epoch": 0.46801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.199592139188636e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.15161240135295e-05,
+      "loss": 0.9536,
       "step": 16493
     },
     {
-      "epoch": 0.4529949740463047,
+      "epoch": 0.4680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.1995049765874284e-05,
-      "loss": 0.9611,
+      "learning_rate": 1.1515215553579955e-05,
+      "loss": 0.9031,
       "step": 16494
     },
     {
-      "epoch": 0.4530224382741479,
+      "epoch": 0.4680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.1994178124076606e-05,
-      "loss": 0.919,
+      "learning_rate": 1.1514307080831316e-05,
+      "loss": 0.9965,
       "step": 16495
     },
     {
-      "epoch": 0.4530499025019912,
+      "epoch": 0.4681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.1993306466500218e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.1513398595291254e-05,
+      "loss": 0.8084,
       "step": 16496
     },
     {
-      "epoch": 0.4530773667298344,
+      "epoch": 0.4681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.1992434793152016e-05,
-      "loss": 0.8657,
+      "learning_rate": 1.1512490096967449e-05,
+      "loss": 0.947,
       "step": 16497
     },
     {
-      "epoch": 0.4531048309576776,
+      "epoch": 0.468161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.1991563104038896e-05,
-      "loss": 0.8828,
+      "learning_rate": 1.151158158586757e-05,
+      "loss": 0.9691,
       "step": 16498
     },
     {
-      "epoch": 0.45313229518552084,
+      "epoch": 0.46818955732122586,
       "grad_norm": 0.0,
-      "learning_rate": 1.1990691399167762e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.1510673061999296e-05,
+      "loss": 0.9252,
       "step": 16499
     },
     {
-      "epoch": 0.4531597594133641,
+      "epoch": 0.4682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.1989819678545504e-05,
-      "loss": 1.0597,
+      "learning_rate": 1.1509764525370294e-05,
+      "loss": 0.8102,
       "step": 16500
     },
     {
-      "epoch": 0.45318722364120734,
+      "epoch": 0.4682463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.1988947942179021e-05,
-      "loss": 0.8374,
+      "learning_rate": 1.150885597598825e-05,
+      "loss": 0.9027,
       "step": 16501
     },
     {
-      "epoch": 0.45321468786905056,
+      "epoch": 0.46827468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.1988076190075212e-05,
-      "loss": 1.0021,
+      "learning_rate": 1.1507947413860827e-05,
+      "loss": 0.9237,
       "step": 16502
     },
     {
-      "epoch": 0.4532421520968938,
+      "epoch": 0.46830306469920546,
       "grad_norm": 0.0,
-      "learning_rate": 1.1987204422240974e-05,
-      "loss": 0.9915,
+      "learning_rate": 1.1507038838995705e-05,
+      "loss": 0.9644,
       "step": 16503
     },
     {
-      "epoch": 0.45326961632473706,
+      "epoch": 0.4683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.1986332638683207e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.1506130251400558e-05,
+      "loss": 0.9375,
       "step": 16504
     },
     {
-      "epoch": 0.4532970805525803,
+      "epoch": 0.46835981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 1.1985460839408807e-05,
-      "loss": 1.0084,
+      "learning_rate": 1.1505221651083063e-05,
+      "loss": 0.9161,
       "step": 16505
     },
     {
-      "epoch": 0.4533245447804235,
+      "epoch": 0.46838819523269015,
       "grad_norm": 0.0,
-      "learning_rate": 1.1984589024424666e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.1504313038050893e-05,
+      "loss": 0.8284,
       "step": 16506
     },
     {
-      "epoch": 0.4533520090082667,
+      "epoch": 0.468416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.1983717193737692e-05,
-      "loss": 0.8322,
+      "learning_rate": 1.150340441231172e-05,
+      "loss": 0.87,
       "step": 16507
     },
     {
-      "epoch": 0.45337947323610994,
+      "epoch": 0.4684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.198284534735478e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.1502495773873226e-05,
+      "loss": 0.8517,
       "step": 16508
     },
     {
-      "epoch": 0.4534069374639532,
+      "epoch": 0.4684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.1981973485282827e-05,
-      "loss": 0.873,
+      "learning_rate": 1.1501587122743081e-05,
+      "loss": 1.0031,
       "step": 16509
     },
     {
-      "epoch": 0.45343440169179644,
+      "epoch": 0.4685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.198110160752873e-05,
-      "loss": 1.0516,
+      "learning_rate": 1.1500678458928962e-05,
+      "loss": 0.8494,
       "step": 16510
     },
     {
-      "epoch": 0.45346186591963966,
+      "epoch": 0.46853007945516456,
       "grad_norm": 0.0,
-      "learning_rate": 1.1980229714099392e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.1499769782438547e-05,
+      "loss": 0.9038,
       "step": 16511
     },
     {
-      "epoch": 0.4534893301474829,
+      "epoch": 0.4685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.197935780500171e-05,
-      "loss": 0.9894,
+      "learning_rate": 1.1498861093279503e-05,
+      "loss": 0.8815,
       "step": 16512
     },
     {
-      "epoch": 0.45351679437532616,
+      "epoch": 0.4685868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.1978485880242578e-05,
-      "loss": 0.8609,
+      "learning_rate": 1.1497952391459517e-05,
+      "loss": 0.8266,
       "step": 16513
     },
     {
-      "epoch": 0.4535442586031694,
+      "epoch": 0.46861520998864925,
       "grad_norm": 0.0,
-      "learning_rate": 1.1977613939828904e-05,
-      "loss": 0.9112,
+      "learning_rate": 1.1497043676986256e-05,
+      "loss": 0.9851,
       "step": 16514
     },
     {
-      "epoch": 0.4535717228310126,
+      "epoch": 0.46864358683314417,
       "grad_norm": 0.0,
-      "learning_rate": 1.1976741983767579e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.1496134949867405e-05,
+      "loss": 0.8853,
       "step": 16515
     },
     {
-      "epoch": 0.4535991870588558,
+      "epoch": 0.468671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.197587001206551e-05,
-      "loss": 0.8817,
+      "learning_rate": 1.1495226210110633e-05,
+      "loss": 0.9241,
       "step": 16516
     },
     {
-      "epoch": 0.4536266512866991,
+      "epoch": 0.46870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.1974998024729589e-05,
-      "loss": 0.946,
+      "learning_rate": 1.1494317457723615e-05,
+      "loss": 0.9029,
       "step": 16517
     },
     {
-      "epoch": 0.4536541155145423,
+      "epoch": 0.46872871736662886,
       "grad_norm": 0.0,
-      "learning_rate": 1.1974126021766716e-05,
-      "loss": 0.9582,
+      "learning_rate": 1.1493408692714035e-05,
+      "loss": 0.9599,
       "step": 16518
     },
     {
-      "epoch": 0.45368157974238554,
+      "epoch": 0.4687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.1973254003183795e-05,
-      "loss": 0.9104,
+      "learning_rate": 1.149249991508956e-05,
+      "loss": 0.8054,
       "step": 16519
     },
     {
-      "epoch": 0.45370904397022876,
+      "epoch": 0.46878547105561863,
       "grad_norm": 0.0,
-      "learning_rate": 1.197238196898773e-05,
-      "loss": 0.8902,
+      "learning_rate": 1.1491591124857873e-05,
+      "loss": 0.9128,
       "step": 16520
     },
     {
-      "epoch": 0.453736508198072,
+      "epoch": 0.4688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.197150991918541e-05,
-      "loss": 0.9018,
+      "learning_rate": 1.149068232202665e-05,
+      "loss": 0.8438,
       "step": 16521
     },
     {
-      "epoch": 0.45376397242591526,
+      "epoch": 0.4688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1970637853783737e-05,
-      "loss": 0.8069,
+      "learning_rate": 1.1489773506603566e-05,
+      "loss": 0.7239,
       "step": 16522
     },
     {
-      "epoch": 0.4537914366537585,
+      "epoch": 0.46887060158910326,
       "grad_norm": 0.0,
-      "learning_rate": 1.196976577278962e-05,
-      "loss": 0.9303,
+      "learning_rate": 1.1488864678596296e-05,
+      "loss": 0.927,
       "step": 16523
     },
     {
-      "epoch": 0.4538189008816017,
+      "epoch": 0.4688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.196889367620995e-05,
-      "loss": 0.9151,
+      "learning_rate": 1.1487955838012523e-05,
+      "loss": 0.8446,
       "step": 16524
     },
     {
-      "epoch": 0.4538463651094449,
+      "epoch": 0.4689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.1968021564051629e-05,
-      "loss": 1.0192,
+      "learning_rate": 1.1487046984859917e-05,
+      "loss": 0.8542,
       "step": 16525
     },
     {
-      "epoch": 0.4538738293372882,
+      "epoch": 0.46895573212258795,
       "grad_norm": 0.0,
-      "learning_rate": 1.1967149436321562e-05,
-      "loss": 0.9344,
+      "learning_rate": 1.1486138119146162e-05,
+      "loss": 0.9171,
       "step": 16526
     },
     {
-      "epoch": 0.4539012935651314,
+      "epoch": 0.46898410896708287,
       "grad_norm": 0.0,
-      "learning_rate": 1.1966277293026643e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.1485229240878924e-05,
+      "loss": 0.9346,
       "step": 16527
     },
     {
-      "epoch": 0.45392875779297465,
+      "epoch": 0.46901248581157773,
       "grad_norm": 0.0,
-      "learning_rate": 1.1965405134173779e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.1484320350065894e-05,
+      "loss": 0.8829,
       "step": 16528
     },
     {
-      "epoch": 0.45395622202081787,
+      "epoch": 0.46904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.1964532959769868e-05,
-      "loss": 0.7886,
+      "learning_rate": 1.1483411446714744e-05,
+      "loss": 0.8365,
       "step": 16529
     },
     {
-      "epoch": 0.45398368624866114,
+      "epoch": 0.46906923950056756,
       "grad_norm": 0.0,
-      "learning_rate": 1.196366076982181e-05,
-      "loss": 0.9828,
+      "learning_rate": 1.1482502530833147e-05,
+      "loss": 1.0234,
       "step": 16530
     },
     {
-      "epoch": 0.45401115047650437,
+      "epoch": 0.4690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.1962788564336506e-05,
-      "loss": 0.917,
+      "learning_rate": 1.1481593602428786e-05,
+      "loss": 0.9781,
       "step": 16531
     },
     {
-      "epoch": 0.4540386147043476,
+      "epoch": 0.46912599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.1961916343320865e-05,
-      "loss": 0.9035,
+      "learning_rate": 1.1480684661509338e-05,
+      "loss": 0.993,
       "step": 16532
     },
     {
-      "epoch": 0.4540660789321908,
+      "epoch": 0.4691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.196104410678178e-05,
-      "loss": 0.8529,
+      "learning_rate": 1.1479775708082479e-05,
+      "loss": 0.9545,
       "step": 16533
     },
     {
-      "epoch": 0.45409354316003403,
+      "epoch": 0.4691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.1960171854726148e-05,
-      "loss": 0.8736,
+      "learning_rate": 1.1478866742155889e-05,
+      "loss": 0.972,
       "step": 16534
     },
     {
-      "epoch": 0.4541210073878773,
+      "epoch": 0.469211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.1959299587160882e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.1477957763737243e-05,
+      "loss": 0.8992,
       "step": 16535
     },
     {
-      "epoch": 0.45414847161572053,
+      "epoch": 0.4692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.195842730409288e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.1477048772834225e-05,
+      "loss": 0.9169,
       "step": 16536
     },
     {
-      "epoch": 0.45417593584356375,
+      "epoch": 0.4692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.1957555005529037e-05,
-      "loss": 1.0103,
+      "learning_rate": 1.1476139769454506e-05,
+      "loss": 0.8003,
       "step": 16537
     },
     {
-      "epoch": 0.45420340007140697,
+      "epoch": 0.46929625425652666,
       "grad_norm": 0.0,
-      "learning_rate": 1.1956682691476267e-05,
-      "loss": 0.9365,
+      "learning_rate": 1.147523075360577e-05,
+      "loss": 0.9362,
       "step": 16538
     },
     {
-      "epoch": 0.45423086429925025,
+      "epoch": 0.46932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.1955810361941462e-05,
-      "loss": 0.86,
+      "learning_rate": 1.1474321725295694e-05,
+      "loss": 0.9053,
       "step": 16539
     },
     {
-      "epoch": 0.45425832852709347,
+      "epoch": 0.46935300794551643,
       "grad_norm": 0.0,
-      "learning_rate": 1.1954938016931528e-05,
-      "loss": 0.8403,
+      "learning_rate": 1.1473412684531955e-05,
+      "loss": 0.8727,
       "step": 16540
     },
     {
-      "epoch": 0.4542857927549367,
+      "epoch": 0.46938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.1954065656453369e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.1472503631322233e-05,
+      "loss": 1.0294,
       "step": 16541
     },
     {
-      "epoch": 0.4543132569827799,
+      "epoch": 0.46940976163450626,
       "grad_norm": 0.0,
-      "learning_rate": 1.1953193280513883e-05,
-      "loss": 0.9944,
+      "learning_rate": 1.1471594565674207e-05,
+      "loss": 0.9494,
       "step": 16542
     },
     {
-      "epoch": 0.4543407212106232,
+      "epoch": 0.4694381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.1952320889119975e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.1470685487595557e-05,
+      "loss": 0.9202,
       "step": 16543
     },
     {
-      "epoch": 0.4543681854384664,
+      "epoch": 0.46946651532349604,
       "grad_norm": 0.0,
-      "learning_rate": 1.1951448482278554e-05,
-      "loss": 0.921,
+      "learning_rate": 1.1469776397093954e-05,
+      "loss": 0.8973,
       "step": 16544
     },
     {
-      "epoch": 0.45439564966630963,
+      "epoch": 0.4694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.195057605999651e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.1468867294177091e-05,
+      "loss": 0.9338,
       "step": 16545
     },
     {
-      "epoch": 0.45442311389415285,
+      "epoch": 0.4695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.1949703622280752e-05,
-      "loss": 0.912,
+      "learning_rate": 1.1467958178852638e-05,
+      "loss": 0.8811,
       "step": 16546
     },
     {
-      "epoch": 0.4544505781219961,
+      "epoch": 0.4695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.1948831169138188e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.1467049051128275e-05,
+      "loss": 0.974,
       "step": 16547
     },
     {
-      "epoch": 0.45447804234983935,
+      "epoch": 0.4695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.1947958700575718e-05,
-      "loss": 0.8717,
+      "learning_rate": 1.1466139911011684e-05,
+      "loss": 0.8195,
       "step": 16548
     },
     {
-      "epoch": 0.45450550657768257,
+      "epoch": 0.4696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.194708621660024e-05,
-      "loss": 0.9318,
+      "learning_rate": 1.1465230758510544e-05,
+      "loss": 0.8756,
       "step": 16549
     },
     {
-      "epoch": 0.4545329708055258,
+      "epoch": 0.46963677639046536,
       "grad_norm": 0.0,
-      "learning_rate": 1.1946213717218661e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.1464321593632533e-05,
+      "loss": 0.8668,
       "step": 16550
     },
     {
-      "epoch": 0.454560435033369,
+      "epoch": 0.4696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.1945341202437887e-05,
-      "loss": 0.8288,
+      "learning_rate": 1.146341241638533e-05,
+      "loss": 0.8128,
       "step": 16551
     },
     {
-      "epoch": 0.4545878992612123,
+      "epoch": 0.4696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.194446867226482e-05,
-      "loss": 1.046,
+      "learning_rate": 1.1462503226776618e-05,
+      "loss": 0.8286,
       "step": 16552
     },
     {
-      "epoch": 0.4546153634890555,
+      "epoch": 0.46972190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.1943596126706365e-05,
-      "loss": 0.9098,
+      "learning_rate": 1.1461594024814075e-05,
+      "loss": 0.9275,
       "step": 16553
     },
     {
-      "epoch": 0.45464282771689873,
+      "epoch": 0.46975028376844497,
       "grad_norm": 0.0,
-      "learning_rate": 1.1942723565769423e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.1460684810505383e-05,
+      "loss": 0.797,
       "step": 16554
     },
     {
-      "epoch": 0.45467029194474196,
+      "epoch": 0.4697786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.1941850989460899e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.145977558385822e-05,
+      "loss": 0.9097,
       "step": 16555
     },
     {
-      "epoch": 0.45469775617258523,
+      "epoch": 0.46980703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 1.1940978397787702e-05,
-      "loss": 0.8929,
+      "learning_rate": 1.1458866344880266e-05,
+      "loss": 0.8122,
       "step": 16556
     },
     {
-      "epoch": 0.45472522040042845,
+      "epoch": 0.4698354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.1940105790756727e-05,
-      "loss": 0.9447,
+      "learning_rate": 1.1457957093579202e-05,
+      "loss": 0.981,
       "step": 16557
     },
     {
-      "epoch": 0.4547526846282717,
+      "epoch": 0.4698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.1939233168374884e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.1457047829962708e-05,
+      "loss": 0.961,
       "step": 16558
     },
     {
-      "epoch": 0.4547801488561149,
+      "epoch": 0.46989216799091943,
       "grad_norm": 0.0,
-      "learning_rate": 1.1938360530649078e-05,
-      "loss": 0.8465,
+      "learning_rate": 1.145613855403847e-05,
+      "loss": 0.8892,
       "step": 16559
     },
     {
-      "epoch": 0.4548076130839581,
+      "epoch": 0.4699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.1937487877586213e-05,
-      "loss": 0.9497,
+      "learning_rate": 1.1455229265814163e-05,
+      "loss": 0.7836,
       "step": 16560
     },
     {
-      "epoch": 0.4548350773118014,
+      "epoch": 0.4699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.1936615209193194e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.1454319965297464e-05,
+      "loss": 0.942,
       "step": 16561
     },
     {
-      "epoch": 0.4548625415396446,
+      "epoch": 0.46997729852440406,
       "grad_norm": 0.0,
-      "learning_rate": 1.1935742525476923e-05,
-      "loss": 0.8196,
+      "learning_rate": 1.1453410652496064e-05,
+      "loss": 0.8665,
       "step": 16562
     },
     {
-      "epoch": 0.45489000576748784,
+      "epoch": 0.470005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.1934869826444309e-05,
-      "loss": 0.8168,
+      "learning_rate": 1.145250132741764e-05,
+      "loss": 0.919,
       "step": 16563
     },
     {
-      "epoch": 0.45491746999533106,
+      "epoch": 0.4700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.1933997112102256e-05,
-      "loss": 0.8489,
+      "learning_rate": 1.1451591990069867e-05,
+      "loss": 0.9344,
       "step": 16564
     },
     {
-      "epoch": 0.45494493422317434,
+      "epoch": 0.47006242905788875,
       "grad_norm": 0.0,
-      "learning_rate": 1.1933124382457668e-05,
-      "loss": 0.9559,
+      "learning_rate": 1.1450682640460434e-05,
+      "loss": 0.8555,
       "step": 16565
     },
     {
-      "epoch": 0.45497239845101756,
+      "epoch": 0.47009080590238367,
       "grad_norm": 0.0,
-      "learning_rate": 1.1932251637517449e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.1449773278597018e-05,
+      "loss": 0.8358,
       "step": 16566
     },
     {
-      "epoch": 0.4549998626788608,
+      "epoch": 0.47011918274687853,
       "grad_norm": 0.0,
-      "learning_rate": 1.193137887728851e-05,
-      "loss": 1.01,
+      "learning_rate": 1.1448863904487304e-05,
+      "loss": 0.8931,
       "step": 16567
     },
     {
-      "epoch": 0.455027326906704,
+      "epoch": 0.47014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.1930506101777753e-05,
-      "loss": 0.8575,
+      "learning_rate": 1.144795451813897e-05,
+      "loss": 0.9275,
       "step": 16568
     },
     {
-      "epoch": 0.4550547911345473,
+      "epoch": 0.47017593643586836,
       "grad_norm": 0.0,
-      "learning_rate": 1.1929633310992082e-05,
-      "loss": 0.9657,
+      "learning_rate": 1.14470451195597e-05,
+      "loss": 0.9636,
       "step": 16569
     },
     {
-      "epoch": 0.4550822553623905,
+      "epoch": 0.4702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.1928760504938406e-05,
-      "loss": 0.9024,
+      "learning_rate": 1.1446135708757176e-05,
+      "loss": 0.9918,
       "step": 16570
     },
     {
-      "epoch": 0.4551097195902337,
+      "epoch": 0.47023269012485813,
       "grad_norm": 0.0,
-      "learning_rate": 1.1927887683623626e-05,
-      "loss": 0.8851,
+      "learning_rate": 1.1445226285739075e-05,
+      "loss": 0.9234,
       "step": 16571
     },
     {
-      "epoch": 0.45513718381807694,
+      "epoch": 0.470261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.1927014847054656e-05,
-      "loss": 0.9678,
+      "learning_rate": 1.1444316850513086e-05,
+      "loss": 0.817,
       "step": 16572
     },
     {
-      "epoch": 0.45516464804592016,
+      "epoch": 0.4702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1926141995238398e-05,
-      "loss": 0.9774,
+      "learning_rate": 1.1443407403086886e-05,
+      "loss": 0.9188,
       "step": 16573
     },
     {
-      "epoch": 0.45519211227376344,
+      "epoch": 0.47031782065834277,
       "grad_norm": 0.0,
-      "learning_rate": 1.1925269128181758e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.1442497943468158e-05,
+      "loss": 0.863,
       "step": 16574
     },
     {
-      "epoch": 0.45521957650160666,
+      "epoch": 0.4703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.1924396245891641e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.1441588471664587e-05,
+      "loss": 0.8771,
       "step": 16575
     },
     {
-      "epoch": 0.4552470407294499,
+      "epoch": 0.4703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.192352334837496e-05,
-      "loss": 0.9704,
+      "learning_rate": 1.1440678987683853e-05,
+      "loss": 0.7784,
       "step": 16576
     },
     {
-      "epoch": 0.4552745049572931,
+      "epoch": 0.47040295119182746,
       "grad_norm": 0.0,
-      "learning_rate": 1.1922650435638611e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.1439769491533636e-05,
+      "loss": 0.8241,
       "step": 16577
     },
     {
-      "epoch": 0.4553019691851364,
+      "epoch": 0.47043132803632237,
       "grad_norm": 0.0,
-      "learning_rate": 1.1921777507689509e-05,
-      "loss": 0.9767,
+      "learning_rate": 1.1438859983221625e-05,
+      "loss": 0.8536,
       "step": 16578
     },
     {
-      "epoch": 0.4553294334129796,
+      "epoch": 0.47045970488081723,
       "grad_norm": 0.0,
-      "learning_rate": 1.1920904564534561e-05,
-      "loss": 0.8393,
+      "learning_rate": 1.1437950462755498e-05,
+      "loss": 0.9648,
       "step": 16579
     },
     {
-      "epoch": 0.4553568976408228,
+      "epoch": 0.47048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.1920031606180671e-05,
-      "loss": 0.7746,
+      "learning_rate": 1.143704093014294e-05,
+      "loss": 0.8518,
       "step": 16580
     },
     {
-      "epoch": 0.45538436186866604,
+      "epoch": 0.47051645856980706,
       "grad_norm": 0.0,
-      "learning_rate": 1.1919158632634747e-05,
-      "loss": 1.002,
+      "learning_rate": 1.143613138539163e-05,
+      "loss": 0.8684,
       "step": 16581
     },
     {
-      "epoch": 0.4554118260965093,
+      "epoch": 0.4705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.1918285643903697e-05,
-      "loss": 0.9572,
+      "learning_rate": 1.1435221828509253e-05,
+      "loss": 0.9709,
       "step": 16582
     },
     {
-      "epoch": 0.45543929032435254,
+      "epoch": 0.47057321225879684,
       "grad_norm": 0.0,
-      "learning_rate": 1.1917412639994425e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.1434312259503498e-05,
+      "loss": 0.8825,
       "step": 16583
     },
     {
-      "epoch": 0.45546675455219576,
+      "epoch": 0.4706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.1916539620913848e-05,
-      "loss": 0.82,
+      "learning_rate": 1.1433402678382037e-05,
+      "loss": 0.7996,
       "step": 16584
     },
     {
-      "epoch": 0.455494218780039,
+      "epoch": 0.4706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.1915666586668862e-05,
-      "loss": 0.9643,
+      "learning_rate": 1.1432493085152563e-05,
+      "loss": 0.9798,
       "step": 16585
     },
     {
-      "epoch": 0.45552168300788226,
+      "epoch": 0.4706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.191479353726638e-05,
-      "loss": 0.8676,
+      "learning_rate": 1.1431583479822754e-05,
+      "loss": 0.933,
       "step": 16586
     },
     {
-      "epoch": 0.4555491472357255,
+      "epoch": 0.4706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.191392047271331e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.1430673862400295e-05,
+      "loss": 0.9002,
       "step": 16587
     },
     {
-      "epoch": 0.4555766114635687,
+      "epoch": 0.4707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.1913047393016563e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.1429764232892868e-05,
+      "loss": 0.8213,
       "step": 16588
     },
     {
-      "epoch": 0.4556040756914119,
+      "epoch": 0.47074347332576616,
       "grad_norm": 0.0,
-      "learning_rate": 1.1912174298183041e-05,
-      "loss": 0.9098,
+      "learning_rate": 1.142885459130816e-05,
+      "loss": 0.8171,
       "step": 16589
     },
     {
-      "epoch": 0.45563153991925515,
+      "epoch": 0.4707718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.1911301188219655e-05,
-      "loss": 0.986,
+      "learning_rate": 1.1427944937653855e-05,
+      "loss": 0.8646,
       "step": 16590
     },
     {
-      "epoch": 0.4556590041470984,
+      "epoch": 0.47080022701475593,
       "grad_norm": 0.0,
-      "learning_rate": 1.1910428063133315e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.142703527193763e-05,
+      "loss": 0.8422,
       "step": 16591
     },
     {
-      "epoch": 0.45568646837494164,
+      "epoch": 0.47082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.1909554922930927e-05,
-      "loss": 0.9237,
+      "learning_rate": 1.1426125594167179e-05,
+      "loss": 0.8091,
       "step": 16592
     },
     {
-      "epoch": 0.45571393260278487,
+      "epoch": 0.47085698070374576,
       "grad_norm": 0.0,
-      "learning_rate": 1.1908681767619402e-05,
-      "loss": 0.8705,
+      "learning_rate": 1.1425215904350179e-05,
+      "loss": 0.9407,
       "step": 16593
     },
     {
-      "epoch": 0.4557413968306281,
+      "epoch": 0.4708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.1907808597205648e-05,
-      "loss": 0.891,
+      "learning_rate": 1.1424306202494315e-05,
+      "loss": 0.8339,
       "step": 16594
     },
     {
-      "epoch": 0.45576886105847136,
+      "epoch": 0.47091373439273554,
       "grad_norm": 0.0,
-      "learning_rate": 1.190693541169657e-05,
-      "loss": 0.9053,
+      "learning_rate": 1.1423396488607275e-05,
+      "loss": 0.8624,
       "step": 16595
     },
     {
-      "epoch": 0.4557963252863146,
+      "epoch": 0.4709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.1906062211099083e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.1422486762696736e-05,
+      "loss": 0.9097,
       "step": 16596
     },
     {
-      "epoch": 0.4558237895141578,
+      "epoch": 0.4709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.1905188995420094e-05,
-      "loss": 0.9462,
+      "learning_rate": 1.1421577024770392e-05,
+      "loss": 0.9191,
       "step": 16597
     },
     {
-      "epoch": 0.45585125374200103,
+      "epoch": 0.47099886492622023,
       "grad_norm": 0.0,
-      "learning_rate": 1.190431576466651e-05,
-      "loss": 0.894,
+      "learning_rate": 1.1420667274835922e-05,
+      "loss": 0.7971,
       "step": 16598
     },
     {
-      "epoch": 0.4558787179698443,
+      "epoch": 0.4710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.1903442518845241e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.141975751290101e-05,
+      "loss": 0.9649,
       "step": 16599
     },
     {
-      "epoch": 0.4559061821976875,
+      "epoch": 0.47105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.1902569257963203e-05,
-      "loss": 0.7555,
+      "learning_rate": 1.1418847738973346e-05,
+      "loss": 0.8073,
       "step": 16600
     },
     {
-      "epoch": 0.45593364642553075,
+      "epoch": 0.47108399545970486,
       "grad_norm": 0.0,
-      "learning_rate": 1.1901695982027295e-05,
-      "loss": 0.8611,
+      "learning_rate": 1.1417937953060607e-05,
+      "loss": 0.9216,
       "step": 16601
     },
     {
-      "epoch": 0.45596111065337397,
+      "epoch": 0.4711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.1900822691044432e-05,
-      "loss": 0.9062,
+      "learning_rate": 1.1417028155170487e-05,
+      "loss": 0.8884,
       "step": 16602
     },
     {
-      "epoch": 0.4559885748812172,
+      "epoch": 0.47114074914869464,
       "grad_norm": 0.0,
-      "learning_rate": 1.1899949385021525e-05,
-      "loss": 0.8177,
+      "learning_rate": 1.1416118345310664e-05,
+      "loss": 0.7776,
       "step": 16603
     },
     {
-      "epoch": 0.45601603910906047,
+      "epoch": 0.47116912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.1899076063965482e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.1415208523488825e-05,
+      "loss": 0.9188,
       "step": 16604
     },
     {
-      "epoch": 0.4560435033369037,
+      "epoch": 0.47119750283768447,
       "grad_norm": 0.0,
-      "learning_rate": 1.1898202727883212e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.1414298689712655e-05,
+      "loss": 0.9169,
       "step": 16605
     },
     {
-      "epoch": 0.4560709675647469,
+      "epoch": 0.4712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.189732937678163e-05,
-      "loss": 0.8416,
+      "learning_rate": 1.1413388843989843e-05,
+      "loss": 0.9098,
       "step": 16606
     },
     {
-      "epoch": 0.45609843179259013,
+      "epoch": 0.47125425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.1896456010667641e-05,
-      "loss": 0.8546,
+      "learning_rate": 1.141247898632807e-05,
+      "loss": 0.9597,
       "step": 16607
     },
     {
-      "epoch": 0.4561258960204334,
+      "epoch": 0.4712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.1895582629548157e-05,
-      "loss": 0.8555,
+      "learning_rate": 1.1411569116735024e-05,
+      "loss": 0.989,
       "step": 16608
     },
     {
-      "epoch": 0.45615336024827663,
+      "epoch": 0.471311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.1894709233430091e-05,
-      "loss": 0.966,
+      "learning_rate": 1.1410659235218393e-05,
+      "loss": 0.8368,
       "step": 16609
     },
     {
-      "epoch": 0.45618082447611985,
+      "epoch": 0.47133938706015893,
       "grad_norm": 0.0,
-      "learning_rate": 1.189383582232035e-05,
-      "loss": 0.819,
+      "learning_rate": 1.1409749341785859e-05,
+      "loss": 0.8988,
       "step": 16610
     },
     {
-      "epoch": 0.45620828870396307,
+      "epoch": 0.4713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.1892962396225845e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.1408839436445105e-05,
+      "loss": 0.9153,
       "step": 16611
     },
     {
-      "epoch": 0.45623575293180635,
+      "epoch": 0.4713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.1892088955153494e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.1407929519203821e-05,
+      "loss": 0.9656,
       "step": 16612
     },
     {
-      "epoch": 0.45626321715964957,
+      "epoch": 0.47142451759364357,
       "grad_norm": 0.0,
-      "learning_rate": 1.1891215499110197e-05,
-      "loss": 0.8844,
+      "learning_rate": 1.1407019590069698e-05,
+      "loss": 0.9816,
       "step": 16613
     },
     {
-      "epoch": 0.4562906813874928,
+      "epoch": 0.4714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.189034202810287e-05,
-      "loss": 1.032,
+      "learning_rate": 1.1406109649050416e-05,
+      "loss": 0.9312,
       "step": 16614
     },
     {
-      "epoch": 0.456318145615336,
+      "epoch": 0.4714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.1889468542138425e-05,
-      "loss": 0.9689,
+      "learning_rate": 1.140519969615366e-05,
+      "loss": 0.8594,
       "step": 16615
     },
     {
-      "epoch": 0.45634560984317923,
+      "epoch": 0.47150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.1888595041223777e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.140428973138712e-05,
+      "loss": 0.993,
       "step": 16616
     },
     {
-      "epoch": 0.4563730740710225,
+      "epoch": 0.47153802497162317,
       "grad_norm": 0.0,
-      "learning_rate": 1.1887721525365832e-05,
-      "loss": 0.8945,
+      "learning_rate": 1.1403379754758483e-05,
+      "loss": 0.9195,
       "step": 16617
     },
     {
-      "epoch": 0.45640053829886573,
+      "epoch": 0.47156640181611803,
       "grad_norm": 0.0,
-      "learning_rate": 1.18868479945715e-05,
-      "loss": 0.8424,
+      "learning_rate": 1.1402469766275432e-05,
+      "loss": 0.8971,
       "step": 16618
     },
     {
-      "epoch": 0.45642800252670895,
+      "epoch": 0.47159477866061295,
       "grad_norm": 0.0,
-      "learning_rate": 1.1885974448847695e-05,
-      "loss": 0.9056,
+      "learning_rate": 1.1401559765945657e-05,
+      "loss": 0.9295,
       "step": 16619
     },
     {
-      "epoch": 0.4564554667545522,
+      "epoch": 0.4716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.1885100888201333e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.1400649753776845e-05,
+      "loss": 0.9772,
       "step": 16620
     },
     {
-      "epoch": 0.45648293098239545,
+      "epoch": 0.4716515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.1884227312639323e-05,
-      "loss": 0.9929,
+      "learning_rate": 1.1399739729776679e-05,
+      "loss": 0.855,
       "step": 16621
     },
     {
-      "epoch": 0.4565103952102387,
+      "epoch": 0.47167990919409764,
       "grad_norm": 0.0,
-      "learning_rate": 1.1883353722168573e-05,
-      "loss": 0.9046,
+      "learning_rate": 1.1398829693952851e-05,
+      "loss": 0.9884,
       "step": 16622
     },
     {
-      "epoch": 0.4565378594380819,
+      "epoch": 0.4717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.1882480116796003e-05,
-      "loss": 0.9718,
+      "learning_rate": 1.1397919646313044e-05,
+      "loss": 0.9331,
       "step": 16623
     },
     {
-      "epoch": 0.4565653236659251,
+      "epoch": 0.4717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.188160649652852e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.1397009586864949e-05,
+      "loss": 0.9081,
       "step": 16624
     },
     {
-      "epoch": 0.4565927878937684,
+      "epoch": 0.47176503972758227,
       "grad_norm": 0.0,
-      "learning_rate": 1.1880732861373034e-05,
-      "loss": 0.7908,
+      "learning_rate": 1.1396099515616248e-05,
+      "loss": 0.8614,
       "step": 16625
     },
     {
-      "epoch": 0.4566202521216116,
+      "epoch": 0.4717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.1879859211336466e-05,
-      "loss": 0.8814,
+      "learning_rate": 1.1395189432574634e-05,
+      "loss": 0.9969,
       "step": 16626
     },
     {
-      "epoch": 0.45664771634945484,
+      "epoch": 0.4718217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.1878985546425721e-05,
-      "loss": 0.9484,
+      "learning_rate": 1.1394279337747789e-05,
+      "loss": 0.94,
       "step": 16627
     },
     {
-      "epoch": 0.45667518057729806,
+      "epoch": 0.47185017026106696,
       "grad_norm": 0.0,
-      "learning_rate": 1.1878111866647715e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.1393369231143406e-05,
+      "loss": 0.8972,
       "step": 16628
     },
     {
-      "epoch": 0.4567026448051413,
+      "epoch": 0.4718785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.1877238172009362e-05,
-      "loss": 0.8956,
+      "learning_rate": 1.1392459112769172e-05,
+      "loss": 0.9557,
       "step": 16629
     },
     {
-      "epoch": 0.45673010903298455,
+      "epoch": 0.47190692395005673,
       "grad_norm": 0.0,
-      "learning_rate": 1.1876364462517569e-05,
-      "loss": 0.8408,
+      "learning_rate": 1.1391548982632772e-05,
+      "loss": 0.8923,
       "step": 16630
     },
     {
-      "epoch": 0.4567575732608278,
+      "epoch": 0.47193530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.1875490738179257e-05,
-      "loss": 0.8249,
+      "learning_rate": 1.1390638840741894e-05,
+      "loss": 0.8275,
       "step": 16631
     },
     {
-      "epoch": 0.456785037488671,
+      "epoch": 0.47196367763904656,
       "grad_norm": 0.0,
-      "learning_rate": 1.1874616999001336e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.1389728687104228e-05,
+      "loss": 0.8966,
       "step": 16632
     },
     {
-      "epoch": 0.4568125017165142,
+      "epoch": 0.4719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.1873743244990718e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.1388818521727463e-05,
+      "loss": 0.9781,
       "step": 16633
     },
     {
-      "epoch": 0.4568399659443575,
+      "epoch": 0.47202043132803634,
       "grad_norm": 0.0,
-      "learning_rate": 1.1872869476154318e-05,
-      "loss": 1.009,
+      "learning_rate": 1.1387908344619283e-05,
+      "loss": 0.8713,
       "step": 16634
     },
     {
-      "epoch": 0.4568674301722007,
+      "epoch": 0.4720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.1871995692499049e-05,
-      "loss": 0.967,
+      "learning_rate": 1.138699815578738e-05,
+      "loss": 0.8972,
       "step": 16635
     },
     {
-      "epoch": 0.45689489440004394,
+      "epoch": 0.4720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.1871121894031825e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.1386087955239442e-05,
+      "loss": 0.9251,
       "step": 16636
     },
     {
-      "epoch": 0.45692235862788716,
+      "epoch": 0.472105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.1870248080759562e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.1385177742983155e-05,
+      "loss": 0.842,
       "step": 16637
     },
     {
-      "epoch": 0.45694982285573044,
+      "epoch": 0.4721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.1869374252689167e-05,
-      "loss": 0.9046,
+      "learning_rate": 1.1384267519026211e-05,
+      "loss": 1.0086,
       "step": 16638
     },
     {
-      "epoch": 0.45697728708357366,
+      "epoch": 0.4721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.186850040982756e-05,
-      "loss": 0.8952,
+      "learning_rate": 1.1383357283376295e-05,
+      "loss": 0.9578,
       "step": 16639
     },
     {
-      "epoch": 0.4570047513114169,
+      "epoch": 0.47219069239500566,
       "grad_norm": 0.0,
-      "learning_rate": 1.1867626552181657e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.13824470360411e-05,
+      "loss": 0.9716,
       "step": 16640
     },
     {
-      "epoch": 0.4570322155392601,
+      "epoch": 0.4722190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.1866752679758369e-05,
-      "loss": 0.9192,
+      "learning_rate": 1.1381536777028313e-05,
+      "loss": 0.9172,
       "step": 16641
     },
     {
-      "epoch": 0.4570596797671033,
+      "epoch": 0.47224744608399544,
       "grad_norm": 0.0,
-      "learning_rate": 1.186587879256461e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.1380626506345618e-05,
+      "loss": 0.8832,
       "step": 16642
     },
     {
-      "epoch": 0.4570871439949466,
+      "epoch": 0.47227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.1865004890607292e-05,
-      "loss": 0.8615,
+      "learning_rate": 1.1379716224000715e-05,
+      "loss": 0.9199,
       "step": 16643
     },
     {
-      "epoch": 0.4571146082227898,
+      "epoch": 0.47230419977298527,
       "grad_norm": 0.0,
-      "learning_rate": 1.1864130973893335e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.1378805930001284e-05,
+      "loss": 0.9566,
       "step": 16644
     },
     {
-      "epoch": 0.45714207245063304,
+      "epoch": 0.4723325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.1863257042429655e-05,
-      "loss": 0.8675,
+      "learning_rate": 1.1377895624355017e-05,
+      "loss": 1.002,
       "step": 16645
     },
     {
-      "epoch": 0.45716953667847626,
+      "epoch": 0.47236095346197504,
       "grad_norm": 0.0,
-      "learning_rate": 1.1862383096223158e-05,
-      "loss": 0.9198,
+      "learning_rate": 1.1376985307069606e-05,
+      "loss": 0.9631,
       "step": 16646
     },
     {
-      "epoch": 0.45719700090631954,
+      "epoch": 0.4723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.1861509135280768e-05,
-      "loss": 0.8449,
+      "learning_rate": 1.1376074978152736e-05,
+      "loss": 0.8966,
       "step": 16647
     },
     {
-      "epoch": 0.45722446513416276,
+      "epoch": 0.4724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.1860635159609397e-05,
-      "loss": 0.9974,
+      "learning_rate": 1.13751646376121e-05,
+      "loss": 0.924,
       "step": 16648
     },
     {
-      "epoch": 0.457251929362006,
+      "epoch": 0.47244608399545973,
       "grad_norm": 0.0,
-      "learning_rate": 1.1859761169215957e-05,
-      "loss": 0.8746,
+      "learning_rate": 1.1374254285455387e-05,
+      "loss": 0.8457,
       "step": 16649
     },
     {
-      "epoch": 0.4572793935898492,
+      "epoch": 0.4724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.1858887164107369e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.1373343921690283e-05,
+      "loss": 0.8738,
       "step": 16650
     },
     {
-      "epoch": 0.4573068578176925,
+      "epoch": 0.4725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.1858013144290543e-05,
-      "loss": 0.924,
+      "learning_rate": 1.1372433546324484e-05,
+      "loss": 0.9562,
       "step": 16651
     },
     {
-      "epoch": 0.4573343220455357,
+      "epoch": 0.47253121452894437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1857139109772399e-05,
-      "loss": 0.9459,
+      "learning_rate": 1.1371523159365675e-05,
+      "loss": 0.9002,
       "step": 16652
     },
     {
-      "epoch": 0.4573617862733789,
+      "epoch": 0.4725595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.1856265060559853e-05,
-      "loss": 0.8686,
+      "learning_rate": 1.1370612760821551e-05,
+      "loss": 0.8779,
       "step": 16653
     },
     {
-      "epoch": 0.45738925050122214,
+      "epoch": 0.47258796821793414,
       "grad_norm": 0.0,
-      "learning_rate": 1.1855390996659814e-05,
-      "loss": 0.9028,
+      "learning_rate": 1.1369702350699798e-05,
+      "loss": 0.8409,
       "step": 16654
     },
     {
-      "epoch": 0.45741671472906537,
+      "epoch": 0.47261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.1854516918079205e-05,
-      "loss": 0.876,
+      "learning_rate": 1.1368791929008107e-05,
+      "loss": 0.8294,
       "step": 16655
     },
     {
-      "epoch": 0.45744417895690864,
+      "epoch": 0.47264472190692397,
       "grad_norm": 0.0,
-      "learning_rate": 1.1853642824824942e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.1367881495754172e-05,
+      "loss": 0.765,
       "step": 16656
     },
     {
-      "epoch": 0.45747164318475186,
+      "epoch": 0.47267309875141883,
       "grad_norm": 0.0,
-      "learning_rate": 1.1852768716903938e-05,
-      "loss": 0.888,
+      "learning_rate": 1.1366971050945676e-05,
+      "loss": 1.0111,
       "step": 16657
     },
     {
-      "epoch": 0.4574991074125951,
+      "epoch": 0.47270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.1851894594323106e-05,
-      "loss": 0.8141,
+      "learning_rate": 1.1366060594590319e-05,
+      "loss": 0.9521,
       "step": 16658
     },
     {
-      "epoch": 0.4575265716404383,
+      "epoch": 0.4727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.185102045708937e-05,
-      "loss": 0.8379,
+      "learning_rate": 1.136515012669578e-05,
+      "loss": 0.8929,
       "step": 16659
     },
     {
-      "epoch": 0.4575540358682816,
+      "epoch": 0.4727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.1850146305209645e-05,
-      "loss": 0.8381,
+      "learning_rate": 1.1364239647269763e-05,
+      "loss": 0.8667,
       "step": 16660
     },
     {
-      "epoch": 0.4575815000961248,
+      "epoch": 0.47278660612939843,
       "grad_norm": 0.0,
-      "learning_rate": 1.1849272138690844e-05,
-      "loss": 0.7999,
+      "learning_rate": 1.1363329156319952e-05,
+      "loss": 0.8808,
       "step": 16661
     },
     {
-      "epoch": 0.457608964323968,
+      "epoch": 0.4728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.1848397957539886e-05,
-      "loss": 0.9712,
+      "learning_rate": 1.1362418653854035e-05,
+      "loss": 0.94,
       "step": 16662
     },
     {
-      "epoch": 0.45763642855181125,
+      "epoch": 0.4728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.1847523761763685e-05,
-      "loss": 1.0015,
+      "learning_rate": 1.1361508139879708e-05,
+      "loss": 0.8618,
       "step": 16663
     },
     {
-      "epoch": 0.4576638927796545,
+      "epoch": 0.47287173666288307,
       "grad_norm": 0.0,
-      "learning_rate": 1.1846649551369163e-05,
-      "loss": 0.9353,
+      "learning_rate": 1.1360597614404663e-05,
+      "loss": 0.972,
       "step": 16664
     },
     {
-      "epoch": 0.45769135700749775,
+      "epoch": 0.472900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.1845775326363234e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.1359687077436582e-05,
+      "loss": 0.8773,
       "step": 16665
     },
     {
-      "epoch": 0.45771882123534097,
+      "epoch": 0.4729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.1844901086752813e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.1358776528983166e-05,
+      "loss": 1.0186,
       "step": 16666
     },
     {
-      "epoch": 0.4577462854631842,
+      "epoch": 0.47295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.184402683254482e-05,
-      "loss": 0.9128,
+      "learning_rate": 1.1357865969052107e-05,
+      "loss": 0.9619,
       "step": 16667
     },
     {
-      "epoch": 0.4577737496910274,
+      "epoch": 0.4729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.1843152563746179e-05,
-      "loss": 0.9487,
+      "learning_rate": 1.1356955397651093e-05,
+      "loss": 0.9661,
       "step": 16668
     },
     {
-      "epoch": 0.4578012139188707,
+      "epoch": 0.47301362088535753,
       "grad_norm": 0.0,
-      "learning_rate": 1.1842278280363794e-05,
-      "loss": 0.9215,
+      "learning_rate": 1.1356044814787812e-05,
+      "loss": 0.8736,
       "step": 16669
     },
     {
-      "epoch": 0.4578286781467139,
+      "epoch": 0.47304199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 1.184140398240459e-05,
-      "loss": 0.9253,
+      "learning_rate": 1.1355134220469961e-05,
+      "loss": 0.8675,
       "step": 16670
     },
     {
-      "epoch": 0.45785614237455713,
+      "epoch": 0.4730703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.1840529669875487e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.135422361470523e-05,
+      "loss": 0.7816,
       "step": 16671
     },
     {
-      "epoch": 0.45788360660240035,
+      "epoch": 0.4730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.18396553427834e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.1353312997501313e-05,
+      "loss": 1.0001,
       "step": 16672
     },
     {
-      "epoch": 0.4579110708302436,
+      "epoch": 0.47312712826333714,
       "grad_norm": 0.0,
-      "learning_rate": 1.1838781001135248e-05,
-      "loss": 0.8876,
+      "learning_rate": 1.1352402368865898e-05,
+      "loss": 0.9216,
       "step": 16673
     },
     {
-      "epoch": 0.45793853505808685,
+      "epoch": 0.473155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.1837906644937947e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.135149172880668e-05,
+      "loss": 0.8655,
       "step": 16674
     },
     {
-      "epoch": 0.45796599928593007,
+      "epoch": 0.4731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.1837032274198417e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.1350581077331353e-05,
+      "loss": 0.8992,
       "step": 16675
     },
     {
-      "epoch": 0.4579934635137733,
+      "epoch": 0.47321225879682177,
       "grad_norm": 0.0,
-      "learning_rate": 1.1836157888923576e-05,
-      "loss": 0.8904,
+      "learning_rate": 1.1349670414447604e-05,
+      "loss": 0.8818,
       "step": 16676
     },
     {
-      "epoch": 0.45802092774161657,
+      "epoch": 0.4732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.1835283489120344e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.1348759740163129e-05,
+      "loss": 0.9524,
       "step": 16677
     },
     {
-      "epoch": 0.4580483919694598,
+      "epoch": 0.4732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.1834409074795634e-05,
-      "loss": 0.8677,
+      "learning_rate": 1.1347849054485622e-05,
+      "loss": 0.9121,
       "step": 16678
     },
     {
-      "epoch": 0.458075856197303,
+      "epoch": 0.47329738933030646,
       "grad_norm": 0.0,
-      "learning_rate": 1.1833534645956373e-05,
-      "loss": 0.9524,
+      "learning_rate": 1.134693835742277e-05,
+      "loss": 0.8799,
       "step": 16679
     },
     {
-      "epoch": 0.45810332042514623,
+      "epoch": 0.4733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.1832660202609475e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.1346027648982272e-05,
+      "loss": 1.0119,
       "step": 16680
     },
     {
-      "epoch": 0.4581307846529895,
+      "epoch": 0.47335414301929624,
       "grad_norm": 0.0,
-      "learning_rate": 1.1831785744761857e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.1345116929171815e-05,
+      "loss": 0.9909,
       "step": 16681
     },
     {
-      "epoch": 0.45815824888083273,
+      "epoch": 0.47338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.1830911272420445e-05,
-      "loss": 0.912,
+      "learning_rate": 1.1344206197999095e-05,
+      "loss": 0.9624,
       "step": 16682
     },
     {
-      "epoch": 0.45818571310867595,
+      "epoch": 0.473410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.1830036785592149e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.1343295455471808e-05,
+      "loss": 0.8779,
       "step": 16683
     },
     {
-      "epoch": 0.4582131773365192,
+      "epoch": 0.4734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.1829162284283894e-05,
-      "loss": 1.0535,
+      "learning_rate": 1.1342384701597644e-05,
+      "loss": 0.8538,
       "step": 16684
     },
     {
-      "epoch": 0.4582406415643624,
+      "epoch": 0.47346765039727584,
       "grad_norm": 0.0,
-      "learning_rate": 1.1828287768502603e-05,
-      "loss": 0.9122,
+      "learning_rate": 1.1341473936384292e-05,
+      "loss": 0.9121,
       "step": 16685
     },
     {
-      "epoch": 0.45826810579220567,
+      "epoch": 0.4734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.1827413238255186e-05,
-      "loss": 0.9492,
+      "learning_rate": 1.1340563159839452e-05,
+      "loss": 0.8998,
       "step": 16686
     },
     {
-      "epoch": 0.4582955700200489,
+      "epoch": 0.4735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.1826538693548568e-05,
-      "loss": 0.8972,
+      "learning_rate": 1.1339652371970815e-05,
+      "loss": 0.8535,
       "step": 16687
     },
     {
-      "epoch": 0.4583230342478921,
+      "epoch": 0.4735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1825664134389669e-05,
-      "loss": 0.9472,
+      "learning_rate": 1.1338741572786074e-05,
+      "loss": 1.0869,
       "step": 16688
     },
     {
-      "epoch": 0.45835049847573534,
+      "epoch": 0.4735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.182478956078541e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.133783076229292e-05,
+      "loss": 0.9359,
       "step": 16689
     },
     {
-      "epoch": 0.4583779627035786,
+      "epoch": 0.4736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.1823914972742707e-05,
-      "loss": 0.8317,
+      "learning_rate": 1.1336919940499054e-05,
+      "loss": 0.9165,
       "step": 16690
     },
     {
-      "epoch": 0.45840542693142183,
+      "epoch": 0.47363791146424516,
       "grad_norm": 0.0,
-      "learning_rate": 1.1823040370268484e-05,
-      "loss": 0.7793,
+      "learning_rate": 1.1336009107412162e-05,
+      "loss": 0.9022,
       "step": 16691
     },
     {
-      "epoch": 0.45843289115926505,
+      "epoch": 0.4736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.182216575336966e-05,
-      "loss": 1.0403,
+      "learning_rate": 1.1335098263039943e-05,
+      "loss": 0.8732,
       "step": 16692
     },
     {
-      "epoch": 0.4584603553871083,
+      "epoch": 0.47369466515323494,
       "grad_norm": 0.0,
-      "learning_rate": 1.1821291122053152e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.1334187407390088e-05,
+      "loss": 0.952,
       "step": 16693
     },
     {
-      "epoch": 0.45848781961495155,
+      "epoch": 0.47372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.1820416476325884e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.1333276540470293e-05,
+      "loss": 0.8117,
       "step": 16694
     },
     {
-      "epoch": 0.4585152838427948,
+      "epoch": 0.47375141884222477,
       "grad_norm": 0.0,
-      "learning_rate": 1.1819541816194775e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.1332365662288249e-05,
+      "loss": 0.8842,
       "step": 16695
     },
     {
-      "epoch": 0.458542748070638,
+      "epoch": 0.47377979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.1818667141666749e-05,
-      "loss": 0.9256,
+      "learning_rate": 1.133145477285165e-05,
+      "loss": 1.0301,
       "step": 16696
     },
     {
-      "epoch": 0.4585702122984812,
+      "epoch": 0.47380817253121454,
       "grad_norm": 0.0,
-      "learning_rate": 1.1817792452748723e-05,
-      "loss": 0.8256,
+      "learning_rate": 1.1330543872168198e-05,
+      "loss": 0.8065,
       "step": 16697
     },
     {
-      "epoch": 0.45859767652632444,
+      "epoch": 0.4738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.1816917749447616e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.1329632960245582e-05,
+      "loss": 0.9238,
       "step": 16698
     },
     {
-      "epoch": 0.4586251407541677,
+      "epoch": 0.4738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.1816043031770355e-05,
-      "loss": 0.92,
+      "learning_rate": 1.1328722037091494e-05,
+      "loss": 0.9163,
       "step": 16699
     },
     {
-      "epoch": 0.45865260498201094,
+      "epoch": 0.4738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.1815168299723858e-05,
-      "loss": 1.0317,
+      "learning_rate": 1.1327811102713633e-05,
+      "loss": 0.8873,
       "step": 16700
     },
     {
-      "epoch": 0.45868006920985416,
+      "epoch": 0.4739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.1814293553315046e-05,
-      "loss": 0.991,
+      "learning_rate": 1.1326900157119692e-05,
+      "loss": 0.8771,
       "step": 16701
     },
     {
-      "epoch": 0.4587075334376974,
+      "epoch": 0.473950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.181341879255084e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.1325989200317362e-05,
+      "loss": 0.9332,
       "step": 16702
     },
     {
-      "epoch": 0.45873499766554066,
+      "epoch": 0.47397843359818387,
       "grad_norm": 0.0,
-      "learning_rate": 1.1812544017438163e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.1325078232314347e-05,
+      "loss": 0.9278,
       "step": 16703
     },
     {
-      "epoch": 0.4587624618933839,
+      "epoch": 0.4740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.1811669227983935e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.132416725311833e-05,
+      "loss": 0.8056,
       "step": 16704
     },
     {
-      "epoch": 0.4587899261212271,
+      "epoch": 0.47403518728717364,
       "grad_norm": 0.0,
-      "learning_rate": 1.1810794424195078e-05,
-      "loss": 0.89,
+      "learning_rate": 1.1323256262737018e-05,
+      "loss": 0.8779,
       "step": 16705
     },
     {
-      "epoch": 0.4588173903490703,
+      "epoch": 0.47406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.1809919606078515e-05,
-      "loss": 0.8861,
+      "learning_rate": 1.1322345261178098e-05,
+      "loss": 0.9639,
       "step": 16706
     },
     {
-      "epoch": 0.4588448545769136,
+      "epoch": 0.4740919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.1809044773641165e-05,
-      "loss": 0.9565,
+      "learning_rate": 1.132143424844927e-05,
+      "loss": 0.8868,
       "step": 16707
     },
     {
-      "epoch": 0.4588723188047568,
+      "epoch": 0.47412031782065833,
       "grad_norm": 0.0,
-      "learning_rate": 1.1808169926889955e-05,
-      "loss": 1.0093,
+      "learning_rate": 1.1320523224558228e-05,
+      "loss": 0.8636,
       "step": 16708
     },
     {
-      "epoch": 0.45889978303260004,
+      "epoch": 0.47414869466515325,
       "grad_norm": 0.0,
-      "learning_rate": 1.1807295065831803e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.1319612189512665e-05,
+      "loss": 0.951,
       "step": 16709
     },
     {
-      "epoch": 0.45892724726044326,
+      "epoch": 0.4741770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.1806420190473631e-05,
-      "loss": 0.9587,
+      "learning_rate": 1.131870114332028e-05,
+      "loss": 0.8692,
       "step": 16710
     },
     {
-      "epoch": 0.4589547114882865,
+      "epoch": 0.474205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.1805545300822361e-05,
-      "loss": 0.8409,
+      "learning_rate": 1.1317790085988767e-05,
+      "loss": 0.8936,
       "step": 16711
     },
     {
-      "epoch": 0.45898217571612976,
+      "epoch": 0.47423382519863794,
       "grad_norm": 0.0,
-      "learning_rate": 1.1804670396884921e-05,
-      "loss": 0.9661,
+      "learning_rate": 1.131687901752582e-05,
+      "loss": 0.9061,
       "step": 16712
     },
     {
-      "epoch": 0.459009639943973,
+      "epoch": 0.4742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.1803795478668229e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.1315967937939138e-05,
+      "loss": 0.9944,
       "step": 16713
     },
     {
-      "epoch": 0.4590371041718162,
+      "epoch": 0.4742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.1802920546179206e-05,
-      "loss": 0.8293,
+      "learning_rate": 1.1315056847236419e-05,
+      "loss": 0.9171,
       "step": 16714
     },
     {
-      "epoch": 0.4590645683996594,
+      "epoch": 0.47431895573212257,
       "grad_norm": 0.0,
-      "learning_rate": 1.180204559942478e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.1314145745425354e-05,
+      "loss": 0.9997,
       "step": 16715
     },
     {
-      "epoch": 0.4590920326275027,
+      "epoch": 0.4743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.1801170638411873e-05,
-      "loss": 0.8635,
+      "learning_rate": 1.1313234632513637e-05,
+      "loss": 0.8782,
       "step": 16716
     },
     {
-      "epoch": 0.4591194968553459,
+      "epoch": 0.47437570942111235,
       "grad_norm": 0.0,
-      "learning_rate": 1.1800295663147403e-05,
-      "loss": 0.8333,
+      "learning_rate": 1.131232350850897e-05,
+      "loss": 0.8308,
       "step": 16717
     },
     {
-      "epoch": 0.45914696108318914,
+      "epoch": 0.47440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.1799420673638297e-05,
-      "loss": 0.8437,
+      "learning_rate": 1.1311412373419051e-05,
+      "loss": 0.8502,
       "step": 16718
     },
     {
-      "epoch": 0.45917442531103236,
+      "epoch": 0.4744324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 1.1798545669891477e-05,
-      "loss": 0.8839,
+      "learning_rate": 1.131050122725157e-05,
+      "loss": 0.839,
       "step": 16719
     },
     {
-      "epoch": 0.45920188953887564,
+      "epoch": 0.47446083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.1797670651913868e-05,
-      "loss": 0.9691,
+      "learning_rate": 1.1309590070014222e-05,
+      "loss": 0.848,
       "step": 16720
     },
     {
-      "epoch": 0.45922935376671886,
+      "epoch": 0.47448921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 1.1796795619712391e-05,
-      "loss": 0.8843,
+      "learning_rate": 1.1308678901714713e-05,
+      "loss": 0.8661,
       "step": 16721
     },
     {
-      "epoch": 0.4592568179945621,
+      "epoch": 0.4745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.1795920573293974e-05,
-      "loss": 0.9743,
+      "learning_rate": 1.1307767722360734e-05,
+      "loss": 0.8627,
       "step": 16722
     },
     {
-      "epoch": 0.4592842822224053,
+      "epoch": 0.4745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.1795045512665532e-05,
-      "loss": 1.0088,
+      "learning_rate": 1.130685653195998e-05,
+      "loss": 0.9305,
       "step": 16723
     },
     {
-      "epoch": 0.4593117464502485,
+      "epoch": 0.47457434733257664,
       "grad_norm": 0.0,
-      "learning_rate": 1.1794170437834001e-05,
-      "loss": 0.8367,
+      "learning_rate": 1.1305945330520153e-05,
+      "loss": 0.8228,
       "step": 16724
     },
     {
-      "epoch": 0.4593392106780918,
+      "epoch": 0.4746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.1793295348806295e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.1305034118048945e-05,
+      "loss": 0.8323,
       "step": 16725
     },
     {
-      "epoch": 0.459366674905935,
+      "epoch": 0.4746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.1792420245589341e-05,
-      "loss": 0.8282,
+      "learning_rate": 1.1304122894554055e-05,
+      "loss": 0.8737,
       "step": 16726
     },
     {
-      "epoch": 0.45939413913377825,
+      "epoch": 0.4746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.1791545128190066e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.1303211660043181e-05,
+      "loss": 0.9363,
       "step": 16727
     },
     {
-      "epoch": 0.45942160336162147,
+      "epoch": 0.4746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.1790669996615392e-05,
-      "loss": 0.9673,
+      "learning_rate": 1.1302300414524021e-05,
+      "loss": 0.8794,
       "step": 16728
     },
     {
-      "epoch": 0.45944906758946474,
+      "epoch": 0.4747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.1789794850872243e-05,
-      "loss": 0.8831,
+      "learning_rate": 1.130138915800427e-05,
+      "loss": 0.7628,
       "step": 16729
     },
     {
-      "epoch": 0.45947653181730796,
+      "epoch": 0.47474460839954596,
       "grad_norm": 0.0,
-      "learning_rate": 1.1788919690967542e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.1300477890491623e-05,
+      "loss": 0.8408,
       "step": 16730
     },
     {
-      "epoch": 0.4595039960451512,
+      "epoch": 0.4747729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.1788044516908217e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.1299566611993784e-05,
+      "loss": 0.9909,
       "step": 16731
     },
     {
-      "epoch": 0.4595314602729944,
+      "epoch": 0.47480136208853574,
       "grad_norm": 0.0,
-      "learning_rate": 1.1787169328701191e-05,
-      "loss": 0.8346,
+      "learning_rate": 1.1298655322518448e-05,
+      "loss": 0.8422,
       "step": 16732
     },
     {
-      "epoch": 0.4595589245008377,
+      "epoch": 0.47482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.178629412635339e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.1297744022073312e-05,
+      "loss": 0.9706,
       "step": 16733
     },
     {
-      "epoch": 0.4595863887286809,
+      "epoch": 0.4748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.1785418909871735e-05,
-      "loss": 0.8599,
+      "learning_rate": 1.129683271066607e-05,
+      "loss": 0.7503,
       "step": 16734
     },
     {
-      "epoch": 0.4596138529565241,
+      "epoch": 0.47488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.1784543679263156e-05,
-      "loss": 1.0161,
+      "learning_rate": 1.1295921388304427e-05,
+      "loss": 0.8628,
       "step": 16735
     },
     {
-      "epoch": 0.45964131718436735,
+      "epoch": 0.47491486946651534,
       "grad_norm": 0.0,
-      "learning_rate": 1.1783668434534576e-05,
-      "loss": 0.8434,
+      "learning_rate": 1.1295010054996079e-05,
+      "loss": 0.9592,
       "step": 16736
     },
     {
-      "epoch": 0.45966878141221057,
+      "epoch": 0.4749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.178279317569292e-05,
-      "loss": 0.8197,
+      "learning_rate": 1.129409871074872e-05,
+      "loss": 0.8743,
       "step": 16737
     },
     {
-      "epoch": 0.45969624564005385,
+      "epoch": 0.4749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.1781917902745114e-05,
-      "loss": 0.967,
+      "learning_rate": 1.1293187355570054e-05,
+      "loss": 0.8531,
       "step": 16738
     },
     {
-      "epoch": 0.45972370986789707,
+      "epoch": 0.475,
       "grad_norm": 0.0,
-      "learning_rate": 1.1781042615698084e-05,
-      "loss": 0.8856,
+      "learning_rate": 1.1292275989467773e-05,
+      "loss": 0.8299,
       "step": 16739
     },
     {
-      "epoch": 0.4597511740957403,
+      "epoch": 0.4750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.1780167314558756e-05,
-      "loss": 0.9128,
+      "learning_rate": 1.1291364612449583e-05,
+      "loss": 0.8645,
       "step": 16740
     },
     {
-      "epoch": 0.4597786383235835,
+      "epoch": 0.4750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.1779291999334053e-05,
-      "loss": 0.8458,
+      "learning_rate": 1.1290453224523175e-05,
+      "loss": 0.8533,
       "step": 16741
     },
     {
-      "epoch": 0.4598061025514268,
+      "epoch": 0.47508513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 1.1778416670030901e-05,
-      "loss": 0.8112,
+      "learning_rate": 1.1289541825696248e-05,
+      "loss": 0.8963,
       "step": 16742
     },
     {
-      "epoch": 0.45983356677927,
+      "epoch": 0.4751135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.1777541326656227e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.1288630415976508e-05,
+      "loss": 0.9293,
       "step": 16743
     },
     {
-      "epoch": 0.45986103100711323,
+      "epoch": 0.47514188422247444,
       "grad_norm": 0.0,
-      "learning_rate": 1.177666596921696e-05,
-      "loss": 0.8417,
+      "learning_rate": 1.1287718995371647e-05,
+      "loss": 0.9459,
       "step": 16744
     },
     {
-      "epoch": 0.45988849523495645,
+      "epoch": 0.47517026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.1775790597720025e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.1286807563889366e-05,
+      "loss": 0.8978,
       "step": 16745
     },
     {
-      "epoch": 0.45991595946279973,
+      "epoch": 0.47519863791146427,
       "grad_norm": 0.0,
-      "learning_rate": 1.1774915212172342e-05,
-      "loss": 1.0187,
+      "learning_rate": 1.1285896121537363e-05,
+      "loss": 0.8511,
       "step": 16746
     },
     {
-      "epoch": 0.45994342369064295,
+      "epoch": 0.47522701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.1774039812580847e-05,
-      "loss": 0.9535,
+      "learning_rate": 1.1284984668323338e-05,
+      "loss": 0.7693,
       "step": 16747
     },
     {
-      "epoch": 0.45997088791848617,
+      "epoch": 0.47525539160045405,
       "grad_norm": 0.0,
-      "learning_rate": 1.1773164398952461e-05,
-      "loss": 0.8426,
+      "learning_rate": 1.128407320425499e-05,
+      "loss": 0.8654,
       "step": 16748
     },
     {
-      "epoch": 0.4599983521463294,
+      "epoch": 0.4752837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.1772288971294108e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.1283161729340017e-05,
+      "loss": 0.8587,
       "step": 16749
     },
     {
-      "epoch": 0.4600258163741726,
+      "epoch": 0.4753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.1771413529612722e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.1282250243586118e-05,
+      "loss": 0.8516,
       "step": 16750
     },
     {
-      "epoch": 0.4600532806020159,
+      "epoch": 0.4753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.1770538073915223e-05,
-      "loss": 0.8108,
+      "learning_rate": 1.1281338747000998e-05,
+      "loss": 1.0669,
       "step": 16751
     },
     {
-      "epoch": 0.4600807448298591,
+      "epoch": 0.4753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1769662604208543e-05,
-      "loss": 0.8791,
+      "learning_rate": 1.1280427239592351e-05,
+      "loss": 0.7972,
       "step": 16752
     },
     {
-      "epoch": 0.46010820905770233,
+      "epoch": 0.4753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.1768787120499608e-05,
-      "loss": 0.9686,
+      "learning_rate": 1.1279515721367875e-05,
+      "loss": 0.9016,
       "step": 16753
     },
     {
-      "epoch": 0.46013567328554555,
+      "epoch": 0.47542565266742337,
       "grad_norm": 0.0,
-      "learning_rate": 1.176791162279534e-05,
-      "loss": 0.9138,
+      "learning_rate": 1.1278604192335273e-05,
+      "loss": 0.8084,
       "step": 16754
     },
     {
-      "epoch": 0.46016313751338883,
+      "epoch": 0.4754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.1767036111102672e-05,
-      "loss": 0.8365,
+      "learning_rate": 1.1277692652502246e-05,
+      "loss": 0.8394,
       "step": 16755
     },
     {
-      "epoch": 0.46019060174123205,
+      "epoch": 0.47548240635641315,
       "grad_norm": 0.0,
-      "learning_rate": 1.1766160585428532e-05,
-      "loss": 0.9839,
+      "learning_rate": 1.1276781101876488e-05,
+      "loss": 0.8997,
       "step": 16756
     },
     {
-      "epoch": 0.4602180659690753,
+      "epoch": 0.47551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.1765285045779847e-05,
-      "loss": 0.9237,
+      "learning_rate": 1.1275869540465705e-05,
+      "loss": 0.8837,
       "step": 16757
     },
     {
-      "epoch": 0.4602455301969185,
+      "epoch": 0.475539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.1764409492163537e-05,
-      "loss": 0.9201,
+      "learning_rate": 1.1274957968277593e-05,
+      "loss": 0.7972,
       "step": 16758
     },
     {
-      "epoch": 0.46027299442476177,
+      "epoch": 0.47556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.1763533924586539e-05,
-      "loss": 0.9717,
+      "learning_rate": 1.1274046385319857e-05,
+      "loss": 0.791,
       "step": 16759
     },
     {
-      "epoch": 0.460300458652605,
+      "epoch": 0.47559591373439275,
       "grad_norm": 0.0,
-      "learning_rate": 1.1762658343055779e-05,
-      "loss": 0.9865,
+      "learning_rate": 1.127313479160019e-05,
+      "loss": 0.8325,
       "step": 16760
     },
     {
-      "epoch": 0.4603279228804482,
+      "epoch": 0.4756242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.1761782747578178e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.1272223187126298e-05,
+      "loss": 0.8442,
       "step": 16761
     },
     {
-      "epoch": 0.46035538710829144,
+      "epoch": 0.4756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.1760907138160676e-05,
-      "loss": 0.9834,
+      "learning_rate": 1.1271311571905884e-05,
+      "loss": 0.8545,
       "step": 16762
     },
     {
-      "epoch": 0.46038285133613466,
+      "epoch": 0.47568104426787744,
       "grad_norm": 0.0,
-      "learning_rate": 1.176003151481019e-05,
-      "loss": 0.8381,
+      "learning_rate": 1.1270399945946638e-05,
+      "loss": 0.8552,
       "step": 16763
     },
     {
-      "epoch": 0.46041031556397793,
+      "epoch": 0.4757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.1759155877533654e-05,
-      "loss": 0.9364,
+      "learning_rate": 1.1269488309256267e-05,
+      "loss": 0.9221,
       "step": 16764
     },
     {
-      "epoch": 0.46043777979182116,
+      "epoch": 0.4757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.1758280226337997e-05,
-      "loss": 0.823,
+      "learning_rate": 1.1268576661842472e-05,
+      "loss": 0.9942,
       "step": 16765
     },
     {
-      "epoch": 0.4604652440196644,
+      "epoch": 0.4757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.1757404561230144e-05,
-      "loss": 1.0248,
+      "learning_rate": 1.1267665003712953e-05,
+      "loss": 0.8158,
       "step": 16766
     },
     {
-      "epoch": 0.4604927082475076,
+      "epoch": 0.475794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.1756528882217024e-05,
-      "loss": 0.9365,
+      "learning_rate": 1.1266753334875411e-05,
+      "loss": 1.0778,
       "step": 16767
     },
     {
-      "epoch": 0.4605201724753509,
+      "epoch": 0.47582292849035185,
       "grad_norm": 0.0,
-      "learning_rate": 1.175565318930557e-05,
-      "loss": 0.9096,
+      "learning_rate": 1.1265841655337547e-05,
+      "loss": 0.8804,
       "step": 16768
     },
     {
-      "epoch": 0.4605476367031941,
+      "epoch": 0.47585130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.1754777482502706e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.1264929965107064e-05,
+      "loss": 0.8779,
       "step": 16769
     },
     {
-      "epoch": 0.4605751009310373,
+      "epoch": 0.4758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.1753901761815363e-05,
-      "loss": 1.0405,
+      "learning_rate": 1.1264018264191653e-05,
+      "loss": 0.8771,
       "step": 16770
     },
     {
-      "epoch": 0.46060256515888054,
+      "epoch": 0.47590805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.1753026027250471e-05,
-      "loss": 0.9376,
+      "learning_rate": 1.126310655259903e-05,
+      "loss": 1.01,
       "step": 16771
     },
     {
-      "epoch": 0.4606300293867238,
+      "epoch": 0.47593643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 1.175215027881496e-05,
-      "loss": 0.8793,
+      "learning_rate": 1.1262194830336888e-05,
+      "loss": 0.9235,
       "step": 16772
     },
     {
-      "epoch": 0.46065749361456704,
+      "epoch": 0.4759648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.1751274516515752e-05,
-      "loss": 1.0171,
+      "learning_rate": 1.1261283097412928e-05,
+      "loss": 0.8202,
       "step": 16773
     },
     {
-      "epoch": 0.46068495784241026,
+      "epoch": 0.47599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.1750398740359786e-05,
-      "loss": 0.8894,
+      "learning_rate": 1.1260371353834851e-05,
+      "loss": 0.9378,
       "step": 16774
     },
     {
-      "epoch": 0.4607124220702535,
+      "epoch": 0.47602156640181614,
       "grad_norm": 0.0,
-      "learning_rate": 1.1749522950353984e-05,
-      "loss": 1.006,
+      "learning_rate": 1.1259459599610365e-05,
+      "loss": 0.8671,
       "step": 16775
     },
     {
-      "epoch": 0.4607398862980967,
+      "epoch": 0.476049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.1748647146505282e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.1258547834747167e-05,
+      "loss": 0.9054,
       "step": 16776
     },
     {
-      "epoch": 0.46076735052594,
+      "epoch": 0.4760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.1747771328820604e-05,
-      "loss": 0.7768,
+      "learning_rate": 1.1257636059252954e-05,
+      "loss": 0.9002,
       "step": 16777
     },
     {
-      "epoch": 0.4607948147537832,
+      "epoch": 0.4761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.1746895497306883e-05,
-      "loss": 0.9711,
+      "learning_rate": 1.1256724273135438e-05,
+      "loss": 0.9762,
       "step": 16778
     },
     {
-      "epoch": 0.4608222789816264,
+      "epoch": 0.4761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.1746019651971045e-05,
-      "loss": 0.8608,
+      "learning_rate": 1.1255812476402315e-05,
+      "loss": 0.913,
       "step": 16779
     },
     {
-      "epoch": 0.46084974320946964,
+      "epoch": 0.47616345062429055,
       "grad_norm": 0.0,
-      "learning_rate": 1.1745143792820028e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.1254900669061282e-05,
+      "loss": 0.8652,
       "step": 16780
     },
     {
-      "epoch": 0.4608772074373129,
+      "epoch": 0.47619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.1744267919860757e-05,
-      "loss": 0.8302,
+      "learning_rate": 1.125398885112005e-05,
+      "loss": 0.9182,
       "step": 16781
     },
     {
-      "epoch": 0.46090467166515614,
+      "epoch": 0.4762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.1743392033100159e-05,
-      "loss": 0.8364,
+      "learning_rate": 1.1253077022586321e-05,
+      "loss": 0.9541,
       "step": 16782
     },
     {
-      "epoch": 0.46093213589299936,
+      "epoch": 0.47624858115777524,
       "grad_norm": 0.0,
-      "learning_rate": 1.174251613254517e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.1252165183467792e-05,
+      "loss": 0.9293,
       "step": 16783
     },
     {
-      "epoch": 0.4609596001208426,
+      "epoch": 0.47627695800227016,
       "grad_norm": 0.0,
-      "learning_rate": 1.1741640218202722e-05,
-      "loss": 0.8606,
+      "learning_rate": 1.1251253333772166e-05,
+      "loss": 0.9076,
       "step": 16784
     },
     {
-      "epoch": 0.46098706434868586,
+      "epoch": 0.476305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.1740764290079736e-05,
-      "loss": 0.9581,
+      "learning_rate": 1.1250341473507147e-05,
+      "loss": 0.9701,
       "step": 16785
     },
     {
-      "epoch": 0.4610145285765291,
+      "epoch": 0.47633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.173988834818315e-05,
-      "loss": 0.8842,
+      "learning_rate": 1.124942960268044e-05,
+      "loss": 0.9431,
       "step": 16786
     },
     {
-      "epoch": 0.4610419928043723,
+      "epoch": 0.47636208853575485,
       "grad_norm": 0.0,
-      "learning_rate": 1.1739012392519893e-05,
-      "loss": 0.9432,
+      "learning_rate": 1.1248517721299741e-05,
+      "loss": 0.958,
       "step": 16787
     },
     {
-      "epoch": 0.4610694570322155,
+      "epoch": 0.4763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.17381364230969e-05,
-      "loss": 0.8518,
+      "learning_rate": 1.1247605829372758e-05,
+      "loss": 0.8732,
       "step": 16788
     },
     {
-      "epoch": 0.4610969212600588,
+      "epoch": 0.4764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.1737260439921097e-05,
-      "loss": 0.9697,
+      "learning_rate": 1.1246693926907193e-05,
+      "loss": 0.8649,
       "step": 16789
     },
     {
-      "epoch": 0.461124385487902,
+      "epoch": 0.4764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.1736384442999413e-05,
-      "loss": 0.8257,
+      "learning_rate": 1.1245782013910748e-05,
+      "loss": 0.8875,
       "step": 16790
     },
     {
-      "epoch": 0.46115184971574524,
+      "epoch": 0.4764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.1735508432338784e-05,
-      "loss": 0.9239,
+      "learning_rate": 1.1244870090391127e-05,
+      "loss": 0.8915,
       "step": 16791
     },
     {
-      "epoch": 0.46117931394358846,
+      "epoch": 0.4765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.1734632407946143e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.1243958156356034e-05,
+      "loss": 1.0132,
       "step": 16792
     },
     {
-      "epoch": 0.4612067781714317,
+      "epoch": 0.47653234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 1.1733756369828418e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.1243046211813167e-05,
+      "loss": 0.8401,
       "step": 16793
     },
     {
-      "epoch": 0.46123424239927496,
+      "epoch": 0.4765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.173288031799254e-05,
-      "loss": 0.8798,
+      "learning_rate": 1.1242134256770237e-05,
+      "loss": 0.917,
       "step": 16794
     },
     {
-      "epoch": 0.4612617066271182,
+      "epoch": 0.47658910329171394,
       "grad_norm": 0.0,
-      "learning_rate": 1.1732004252445439e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.124122229123494e-05,
+      "loss": 0.9261,
       "step": 16795
     },
     {
-      "epoch": 0.4612891708549614,
+      "epoch": 0.47661748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.1731128173194053e-05,
-      "loss": 0.8375,
+      "learning_rate": 1.1240310315214982e-05,
+      "loss": 0.8679,
       "step": 16796
     },
     {
-      "epoch": 0.4613166350828046,
+      "epoch": 0.4766458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.173025208024531e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.1239398328718068e-05,
+      "loss": 0.9491,
       "step": 16797
     },
     {
-      "epoch": 0.4613440993106479,
+      "epoch": 0.47667423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.172937597360614e-05,
-      "loss": 0.9032,
+      "learning_rate": 1.1238486331751902e-05,
+      "loss": 0.8006,
       "step": 16798
     },
     {
-      "epoch": 0.4613715635384911,
+      "epoch": 0.47670261066969355,
       "grad_norm": 0.0,
-      "learning_rate": 1.1728499853283476e-05,
-      "loss": 0.9126,
+      "learning_rate": 1.1237574324324185e-05,
+      "loss": 0.8843,
       "step": 16799
     },
     {
-      "epoch": 0.46139902776633435,
+      "epoch": 0.4767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.1727623719284255e-05,
-      "loss": 0.9532,
+      "learning_rate": 1.123666230644262e-05,
+      "loss": 0.9138,
       "step": 16800
     },
     {
-      "epoch": 0.46142649199417757,
+      "epoch": 0.4767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.1726747571615406e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.1235750278114916e-05,
+      "loss": 0.9373,
       "step": 16801
     },
     {
-      "epoch": 0.46145395622202084,
+      "epoch": 0.4767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.172587141028386e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.1234838239348773e-05,
+      "loss": 0.9207,
       "step": 16802
     },
     {
-      "epoch": 0.46148142044986407,
+      "epoch": 0.4768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.1724995235296552e-05,
-      "loss": 0.8763,
+      "learning_rate": 1.1233926190151896e-05,
+      "loss": 0.9137,
       "step": 16803
     },
     {
-      "epoch": 0.4615088846777073,
+      "epoch": 0.476844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.1724119046660415e-05,
-      "loss": 0.8629,
+      "learning_rate": 1.1233014130531988e-05,
+      "loss": 0.8762,
       "step": 16804
     },
     {
-      "epoch": 0.4615363489055505,
+      "epoch": 0.4768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.1723242844382376e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.1232102060496754e-05,
+      "loss": 0.9139,
       "step": 16805
     },
     {
-      "epoch": 0.46156381313339373,
+      "epoch": 0.4769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.1722366628469375e-05,
-      "loss": 0.912,
+      "learning_rate": 1.12311899800539e-05,
+      "loss": 0.8311,
       "step": 16806
     },
     {
-      "epoch": 0.461591277361237,
+      "epoch": 0.47692962542565265,
       "grad_norm": 0.0,
-      "learning_rate": 1.1721490398928341e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.1230277889211128e-05,
+      "loss": 0.8257,
       "step": 16807
     },
     {
-      "epoch": 0.46161874158908023,
+      "epoch": 0.47695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.1720614155766207e-05,
-      "loss": 1.0105,
+      "learning_rate": 1.1229365787976145e-05,
+      "loss": 0.9075,
       "step": 16808
     },
     {
-      "epoch": 0.46164620581692345,
+      "epoch": 0.4769863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.1719737898989911e-05,
-      "loss": 0.904,
+      "learning_rate": 1.1228453676356653e-05,
+      "loss": 0.8843,
       "step": 16809
     },
     {
-      "epoch": 0.46167367004476667,
+      "epoch": 0.47701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.171886162860638e-05,
-      "loss": 1.0068,
+      "learning_rate": 1.1227541554360354e-05,
+      "loss": 0.9007,
       "step": 16810
     },
     {
-      "epoch": 0.46170113427260995,
+      "epoch": 0.47704313280363225,
       "grad_norm": 0.0,
-      "learning_rate": 1.1717985344622549e-05,
-      "loss": 0.9356,
+      "learning_rate": 1.122662942199496e-05,
+      "loss": 0.9096,
       "step": 16811
     },
     {
-      "epoch": 0.46172859850045317,
+      "epoch": 0.4770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.1717109047045355e-05,
-      "loss": 0.9805,
+      "learning_rate": 1.1225717279268174e-05,
+      "loss": 0.8689,
       "step": 16812
     },
     {
-      "epoch": 0.4617560627282964,
+      "epoch": 0.477099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.1716232735881727e-05,
-      "loss": 0.9174,
+      "learning_rate": 1.1224805126187697e-05,
+      "loss": 0.9122,
       "step": 16813
     },
     {
-      "epoch": 0.4617835269561396,
+      "epoch": 0.4771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.1715356411138601e-05,
-      "loss": 0.9098,
+      "learning_rate": 1.1223892962761233e-05,
+      "loss": 0.9139,
       "step": 16814
     },
     {
-      "epoch": 0.4618109911839829,
+      "epoch": 0.4771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.1714480072822912e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.1222980788996494e-05,
+      "loss": 0.9296,
       "step": 16815
     },
     {
-      "epoch": 0.4618384554118261,
+      "epoch": 0.4771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.1713603720941594e-05,
-      "loss": 0.93,
+      "learning_rate": 1.1222068604901181e-05,
+      "loss": 0.8801,
       "step": 16816
     },
     {
-      "epoch": 0.46186591963966933,
+      "epoch": 0.4772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.1712727355501574e-05,
-      "loss": 0.8566,
+      "learning_rate": 1.1221156410482998e-05,
+      "loss": 0.9968,
       "step": 16817
     },
     {
-      "epoch": 0.46189338386751255,
+      "epoch": 0.4772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.1711850976509797e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.1220244205749653e-05,
+      "loss": 0.813,
       "step": 16818
     },
     {
-      "epoch": 0.4619208480953558,
+      "epoch": 0.47727014755959135,
       "grad_norm": 0.0,
-      "learning_rate": 1.171097458397319e-05,
-      "loss": 0.9815,
+      "learning_rate": 1.1219331990708849e-05,
+      "loss": 0.9901,
       "step": 16819
     },
     {
-      "epoch": 0.46194831232319905,
+      "epoch": 0.47729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.1710098177898687e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.1218419765368296e-05,
+      "loss": 0.8441,
       "step": 16820
     },
     {
-      "epoch": 0.46197577655104227,
+      "epoch": 0.4773269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.170922175829323e-05,
-      "loss": 0.9044,
+      "learning_rate": 1.1217507529735692e-05,
+      "loss": 0.901,
       "step": 16821
     },
     {
-      "epoch": 0.4620032407788855,
+      "epoch": 0.47735527809307604,
       "grad_norm": 0.0,
-      "learning_rate": 1.1708345325163744e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.121659528381875e-05,
+      "loss": 0.9273,
       "step": 16822
     },
     {
-      "epoch": 0.4620307050067287,
+      "epoch": 0.47738365493757096,
       "grad_norm": 0.0,
-      "learning_rate": 1.1707468878517167e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.1215683027625173e-05,
+      "loss": 0.9398,
       "step": 16823
     },
     {
-      "epoch": 0.462058169234572,
+      "epoch": 0.4774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.1706592418360439e-05,
-      "loss": 0.7939,
+      "learning_rate": 1.1214770761162666e-05,
+      "loss": 0.8288,
       "step": 16824
     },
     {
-      "epoch": 0.4620856334624152,
+      "epoch": 0.47744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.1705715944700491e-05,
-      "loss": 0.9724,
+      "learning_rate": 1.1213858484438935e-05,
+      "loss": 0.96,
       "step": 16825
     },
     {
-      "epoch": 0.46211309769025843,
+      "epoch": 0.47746878547105565,
       "grad_norm": 0.0,
-      "learning_rate": 1.1704839457544256e-05,
-      "loss": 0.8953,
+      "learning_rate": 1.1212946197461688e-05,
+      "loss": 0.9634,
       "step": 16826
     },
     {
-      "epoch": 0.46214056191810166,
+      "epoch": 0.4774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.170396295689867e-05,
-      "loss": 1.0208,
+      "learning_rate": 1.121203390023863e-05,
+      "loss": 0.8717,
       "step": 16827
     },
     {
-      "epoch": 0.46216802614594493,
+      "epoch": 0.4775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.170308644277067e-05,
-      "loss": 0.8274,
+      "learning_rate": 1.1211121592777466e-05,
+      "loss": 1.0235,
       "step": 16828
     },
     {
-      "epoch": 0.46219549037378815,
+      "epoch": 0.4775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.1702209915167189e-05,
-      "loss": 0.9186,
+      "learning_rate": 1.1210209275085906e-05,
+      "loss": 0.8689,
       "step": 16829
     },
     {
-      "epoch": 0.4622229546016314,
+      "epoch": 0.4775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.1701333374095165e-05,
-      "loss": 0.9448,
+      "learning_rate": 1.1209296947171653e-05,
+      "loss": 0.9108,
       "step": 16830
     },
     {
-      "epoch": 0.4622504188294746,
+      "epoch": 0.47761066969353005,
       "grad_norm": 0.0,
-      "learning_rate": 1.170045681956153e-05,
-      "loss": 0.9236,
+      "learning_rate": 1.120838460904241e-05,
+      "loss": 0.9398,
       "step": 16831
     },
     {
-      "epoch": 0.4622778830573178,
+      "epoch": 0.47763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.1699580251573226e-05,
-      "loss": 0.8847,
+      "learning_rate": 1.1207472260705893e-05,
+      "loss": 0.9674,
       "step": 16832
     },
     {
-      "epoch": 0.4623053472851611,
+      "epoch": 0.4776674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.1698703670137184e-05,
-      "loss": 1.0075,
+      "learning_rate": 1.1206559902169805e-05,
+      "loss": 0.8752,
       "step": 16833
     },
     {
-      "epoch": 0.4623328115130043,
+      "epoch": 0.47769580022701474,
       "grad_norm": 0.0,
-      "learning_rate": 1.1697827075260337e-05,
-      "loss": 0.8136,
+      "learning_rate": 1.1205647533441843e-05,
+      "loss": 0.7673,
       "step": 16834
     },
     {
-      "epoch": 0.46236027574084754,
+      "epoch": 0.47772417707150966,
       "grad_norm": 0.0,
-      "learning_rate": 1.1696950466949626e-05,
-      "loss": 0.8851,
+      "learning_rate": 1.1204735154529728e-05,
+      "loss": 0.7841,
       "step": 16835
     },
     {
-      "epoch": 0.46238773996869076,
+      "epoch": 0.4777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.1696073845211987e-05,
-      "loss": 0.8789,
+      "learning_rate": 1.1203822765441162e-05,
+      "loss": 0.8473,
       "step": 16836
     },
     {
-      "epoch": 0.46241520419653404,
+      "epoch": 0.47778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.1695197210054355e-05,
-      "loss": 0.9589,
+      "learning_rate": 1.120291036618385e-05,
+      "loss": 0.9645,
       "step": 16837
     },
     {
-      "epoch": 0.46244266842437726,
+      "epoch": 0.47780930760499435,
       "grad_norm": 0.0,
-      "learning_rate": 1.1694320561483664e-05,
-      "loss": 0.887,
+      "learning_rate": 1.1201997956765499e-05,
+      "loss": 0.9067,
       "step": 16838
     },
     {
-      "epoch": 0.4624701326522205,
+      "epoch": 0.4778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.1693443899506853e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.1201085537193817e-05,
+      "loss": 0.9037,
       "step": 16839
     },
     {
-      "epoch": 0.4624975968800637,
+      "epoch": 0.4778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.169256722413086e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.1200173107476514e-05,
+      "loss": 0.8487,
       "step": 16840
     },
     {
-      "epoch": 0.462525061107907,
+      "epoch": 0.477894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1691690535362617e-05,
-      "loss": 1.0325,
+      "learning_rate": 1.119926066762129e-05,
+      "loss": 0.8269,
       "step": 16841
     },
     {
-      "epoch": 0.4625525253357502,
+      "epoch": 0.4779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.1690813833209065e-05,
-      "loss": 0.9661,
+      "learning_rate": 1.119834821763586e-05,
+      "loss": 0.9212,
       "step": 16842
     },
     {
-      "epoch": 0.4625799895635934,
+      "epoch": 0.4779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.1689937117677136e-05,
-      "loss": 0.9556,
+      "learning_rate": 1.119743575752793e-05,
+      "loss": 0.965,
       "step": 16843
     },
     {
-      "epoch": 0.46260745379143664,
+      "epoch": 0.4779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.1689060388773774e-05,
-      "loss": 0.8607,
+      "learning_rate": 1.1196523287305204e-05,
+      "loss": 0.8436,
       "step": 16844
     },
     {
-      "epoch": 0.46263491801927986,
+      "epoch": 0.4780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.1688183646505911e-05,
-      "loss": 0.8221,
+      "learning_rate": 1.1195610806975394e-05,
+      "loss": 0.8721,
       "step": 16845
     },
     {
-      "epoch": 0.46266238224712314,
+      "epoch": 0.47803632236095345,
       "grad_norm": 0.0,
-      "learning_rate": 1.168730689088048e-05,
-      "loss": 0.85,
+      "learning_rate": 1.1194698316546205e-05,
+      "loss": 0.8376,
       "step": 16846
     },
     {
-      "epoch": 0.46268984647496636,
+      "epoch": 0.47806469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.1686430121904429e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.1193785816025345e-05,
+      "loss": 0.7803,
       "step": 16847
     },
     {
-      "epoch": 0.4627173107028096,
+      "epoch": 0.4780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.1685553339584687e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.1192873305420525e-05,
+      "loss": 0.89,
       "step": 16848
     },
     {
-      "epoch": 0.4627447749306528,
+      "epoch": 0.47812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.1684676543928197e-05,
-      "loss": 0.7906,
+      "learning_rate": 1.1191960784739447e-05,
+      "loss": 0.8228,
       "step": 16849
     },
     {
-      "epoch": 0.4627722391584961,
+      "epoch": 0.47814982973893305,
       "grad_norm": 0.0,
-      "learning_rate": 1.168379973494189e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.1191048253989825e-05,
+      "loss": 0.8414,
       "step": 16850
     },
     {
-      "epoch": 0.4627997033863393,
+      "epoch": 0.4781782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.1682922912632708e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.1190135713179362e-05,
+      "loss": 0.8082,
       "step": 16851
     },
     {
-      "epoch": 0.4628271676141825,
+      "epoch": 0.4782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.1682046077007592e-05,
-      "loss": 0.8699,
+      "learning_rate": 1.1189223162315773e-05,
+      "loss": 0.9106,
       "step": 16852
     },
     {
-      "epoch": 0.46285463184202574,
+      "epoch": 0.4782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.168116922807347e-05,
-      "loss": 0.8842,
+      "learning_rate": 1.118831060140676e-05,
+      "loss": 0.9499,
       "step": 16853
     },
     {
-      "epoch": 0.462882096069869,
+      "epoch": 0.4782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.168029236583729e-05,
-      "loss": 0.8468,
+      "learning_rate": 1.1187398030460035e-05,
+      "loss": 0.8838,
       "step": 16854
     },
     {
-      "epoch": 0.46290956029771224,
+      "epoch": 0.4782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.1679415490305983e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.1186485449483306e-05,
+      "loss": 0.8732,
       "step": 16855
     },
     {
-      "epoch": 0.46293702452555546,
+      "epoch": 0.4783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.167853860148649e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.1185572858484282e-05,
+      "loss": 0.9211,
       "step": 16856
     },
     {
-      "epoch": 0.4629644887533987,
+      "epoch": 0.4783484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.1677661699385753e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.1184660257470666e-05,
+      "loss": 0.7862,
       "step": 16857
     },
     {
-      "epoch": 0.4629919529812419,
+      "epoch": 0.47837684449489215,
       "grad_norm": 0.0,
-      "learning_rate": 1.1676784784010704e-05,
-      "loss": 0.8608,
+      "learning_rate": 1.1183747646450173e-05,
+      "loss": 0.7307,
       "step": 16858
     },
     {
-      "epoch": 0.4630194172090852,
+      "epoch": 0.47840522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 1.1675907855368287e-05,
-      "loss": 0.8182,
+      "learning_rate": 1.1182835025430514e-05,
+      "loss": 1.0201,
       "step": 16859
     },
     {
-      "epoch": 0.4630468814369284,
+      "epoch": 0.4784335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 1.1675030913465433e-05,
-      "loss": 0.8665,
+      "learning_rate": 1.1181922394419393e-05,
+      "loss": 0.8629,
       "step": 16860
     },
     {
-      "epoch": 0.4630743456647716,
+      "epoch": 0.47846197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.1674153958309089e-05,
-      "loss": 0.971,
+      "learning_rate": 1.1181009753424519e-05,
+      "loss": 0.7769,
       "step": 16861
     },
     {
-      "epoch": 0.46310180989261485,
+      "epoch": 0.47849035187287176,
       "grad_norm": 0.0,
-      "learning_rate": 1.167327698990619e-05,
-      "loss": 0.8764,
+      "learning_rate": 1.1180097102453604e-05,
+      "loss": 0.9625,
       "step": 16862
     },
     {
-      "epoch": 0.4631292741204581,
+      "epoch": 0.4785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.1672400008263676e-05,
-      "loss": 0.8826,
+      "learning_rate": 1.1179184441514354e-05,
+      "loss": 0.9376,
       "step": 16863
     },
     {
-      "epoch": 0.46315673834830134,
+      "epoch": 0.47854710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 1.1671523013388484e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.1178271770614482e-05,
+      "loss": 0.9147,
       "step": 16864
     },
     {
-      "epoch": 0.46318420257614457,
+      "epoch": 0.4785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.1670646005287556e-05,
-      "loss": 0.8715,
+      "learning_rate": 1.1177359089761693e-05,
+      "loss": 0.9342,
       "step": 16865
     },
     {
-      "epoch": 0.4632116668039878,
+      "epoch": 0.4786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.1669768983967826e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.11764463989637e-05,
+      "loss": 0.9307,
       "step": 16866
     },
     {
-      "epoch": 0.46323913103183106,
+      "epoch": 0.4786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.1668891949436239e-05,
-      "loss": 0.8468,
+      "learning_rate": 1.1175533698228214e-05,
+      "loss": 0.7562,
       "step": 16867
     },
     {
-      "epoch": 0.4632665952596743,
+      "epoch": 0.4786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.1668014901699733e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.1174620987562936e-05,
+      "loss": 0.918,
       "step": 16868
     },
     {
-      "epoch": 0.4632940594875175,
+      "epoch": 0.478688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1667137840765247e-05,
-      "loss": 0.9051,
+      "learning_rate": 1.1173708266975588e-05,
+      "loss": 0.9679,
       "step": 16869
     },
     {
-      "epoch": 0.46332152371536073,
+      "epoch": 0.47871736662883085,
       "grad_norm": 0.0,
-      "learning_rate": 1.1666260766639721e-05,
-      "loss": 0.9431,
+      "learning_rate": 1.117279553647387e-05,
+      "loss": 0.7333,
       "step": 16870
     },
     {
-      "epoch": 0.46334898794320395,
+      "epoch": 0.47874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.1665383679330094e-05,
-      "loss": 0.8598,
+      "learning_rate": 1.1171882796065496e-05,
+      "loss": 0.9415,
       "step": 16871
     },
     {
-      "epoch": 0.4633764521710472,
+      "epoch": 0.4787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.1664506578843305e-05,
-      "loss": 0.9022,
+      "learning_rate": 1.1170970045758173e-05,
+      "loss": 0.892,
       "step": 16872
     },
     {
-      "epoch": 0.46340391639889045,
+      "epoch": 0.47880249716231554,
       "grad_norm": 0.0,
-      "learning_rate": 1.1663629465186297e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.1170057285559619e-05,
+      "loss": 0.8511,
       "step": 16873
     },
     {
-      "epoch": 0.46343138062673367,
+      "epoch": 0.47883087400681046,
       "grad_norm": 0.0,
-      "learning_rate": 1.1662752338366008e-05,
-      "loss": 0.9114,
+      "learning_rate": 1.1169144515477536e-05,
+      "loss": 0.9828,
       "step": 16874
     },
     {
-      "epoch": 0.4634588448545769,
+      "epoch": 0.4788592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.1661875198389376e-05,
-      "loss": 0.8485,
+      "learning_rate": 1.1168231735519634e-05,
+      "loss": 0.9202,
       "step": 16875
     },
     {
-      "epoch": 0.46348630908242017,
+      "epoch": 0.47888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.1660998045263347e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.116731894569363e-05,
+      "loss": 0.8531,
       "step": 16876
     },
     {
-      "epoch": 0.4635137733102634,
+      "epoch": 0.4789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.1660120878994858e-05,
-      "loss": 0.9885,
+      "learning_rate": 1.116640614600723e-05,
+      "loss": 0.856,
       "step": 16877
     },
     {
-      "epoch": 0.4635412375381066,
+      "epoch": 0.47894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1659243699590845e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.1165493336468142e-05,
+      "loss": 0.844,
       "step": 16878
     },
     {
-      "epoch": 0.46356870176594983,
+      "epoch": 0.4789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.1658366507058256e-05,
-      "loss": 0.9397,
+      "learning_rate": 1.116458051708408e-05,
+      "loss": 0.975,
       "step": 16879
     },
     {
-      "epoch": 0.4635961659937931,
+      "epoch": 0.4790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.165748930140403e-05,
-      "loss": 0.946,
+      "learning_rate": 1.1163667687862755e-05,
+      "loss": 0.8423,
       "step": 16880
     },
     {
-      "epoch": 0.46362363022163633,
+      "epoch": 0.4790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.1656612082635105e-05,
-      "loss": 0.8455,
+      "learning_rate": 1.116275484881188e-05,
+      "loss": 0.8832,
       "step": 16881
     },
     {
-      "epoch": 0.46365109444947955,
+      "epoch": 0.47905788876276956,
       "grad_norm": 0.0,
-      "learning_rate": 1.1655734850758424e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.1161841999939156e-05,
+      "loss": 0.8421,
       "step": 16882
     },
     {
-      "epoch": 0.46367855867732277,
+      "epoch": 0.47908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.165485760578093e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.1160929141252303e-05,
+      "loss": 0.9282,
       "step": 16883
     },
     {
-      "epoch": 0.46370602290516605,
+      "epoch": 0.4791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.165398034770956e-05,
-      "loss": 0.9781,
+      "learning_rate": 1.1160016272759031e-05,
+      "loss": 0.8816,
       "step": 16884
     },
     {
-      "epoch": 0.46373348713300927,
+      "epoch": 0.47914301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.1653103076551253e-05,
-      "loss": 0.7823,
+      "learning_rate": 1.1159103394467048e-05,
+      "loss": 0.9385,
       "step": 16885
     },
     {
-      "epoch": 0.4637609513608525,
+      "epoch": 0.47917139614074916,
       "grad_norm": 0.0,
-      "learning_rate": 1.1652225792312959e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.1158190506384069e-05,
+      "loss": 0.8848,
       "step": 16886
     },
     {
-      "epoch": 0.4637884155886957,
+      "epoch": 0.479199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.165134849500161e-05,
-      "loss": 0.9861,
+      "learning_rate": 1.11572776085178e-05,
+      "loss": 0.8617,
       "step": 16887
     },
     {
-      "epoch": 0.46381587981653893,
+      "epoch": 0.47922814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 1.1650471184624157e-05,
-      "loss": 1.0076,
+      "learning_rate": 1.1156364700875952e-05,
+      "loss": 0.8609,
       "step": 16888
     },
     {
-      "epoch": 0.4638433440443822,
+      "epoch": 0.47925652667423385,
       "grad_norm": 0.0,
-      "learning_rate": 1.1649593861187536e-05,
-      "loss": 0.8549,
+      "learning_rate": 1.1155451783466244e-05,
+      "loss": 0.9055,
       "step": 16889
     },
     {
-      "epoch": 0.46387080827222543,
+      "epoch": 0.4792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.1648716524698684e-05,
-      "loss": 0.8177,
+      "learning_rate": 1.1154538856296381e-05,
+      "loss": 0.9209,
       "step": 16890
     },
     {
-      "epoch": 0.46389827250006865,
+      "epoch": 0.4793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.1647839175164552e-05,
-      "loss": 0.9268,
+      "learning_rate": 1.115362591937408e-05,
+      "loss": 0.9554,
       "step": 16891
     },
     {
-      "epoch": 0.4639257367279119,
+      "epoch": 0.4793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.164696181259208e-05,
-      "loss": 0.856,
+      "learning_rate": 1.1152712972707045e-05,
+      "loss": 0.8759,
       "step": 16892
     },
     {
-      "epoch": 0.46395320095575515,
+      "epoch": 0.4793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.1646084436988205e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.1151800016302994e-05,
+      "loss": 0.8621,
       "step": 16893
     },
     {
-      "epoch": 0.4639806651835984,
+      "epoch": 0.47939841089670826,
       "grad_norm": 0.0,
-      "learning_rate": 1.1645207048359873e-05,
-      "loss": 0.8618,
+      "learning_rate": 1.1150887050169636e-05,
+      "loss": 0.9259,
       "step": 16894
     },
     {
-      "epoch": 0.4640081294114416,
+      "epoch": 0.4794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.1644329646714028e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.1149974074314678e-05,
+      "loss": 1.0403,
       "step": 16895
     },
     {
-      "epoch": 0.4640355936392848,
+      "epoch": 0.4794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.1643452232057608e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.1149061088745842e-05,
+      "loss": 0.9417,
       "step": 16896
     },
     {
-      "epoch": 0.4640630578671281,
+      "epoch": 0.47948354143019295,
       "grad_norm": 0.0,
-      "learning_rate": 1.1642574804397556e-05,
-      "loss": 0.936,
+      "learning_rate": 1.1148148093470838e-05,
+      "loss": 0.9515,
       "step": 16897
     },
     {
-      "epoch": 0.4640905220949713,
+      "epoch": 0.47951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.1641697363740819e-05,
-      "loss": 1.0388,
+      "learning_rate": 1.1147235088497372e-05,
+      "loss": 1.0031,
       "step": 16898
     },
     {
-      "epoch": 0.46411798632281454,
+      "epoch": 0.4795402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.1640819910094332e-05,
-      "loss": 0.7684,
+      "learning_rate": 1.114632207383316e-05,
+      "loss": 0.938,
       "step": 16899
     },
     {
-      "epoch": 0.46414545055065776,
+      "epoch": 0.47956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.1639942443465047e-05,
-      "loss": 0.9607,
+      "learning_rate": 1.1145409049485916e-05,
+      "loss": 0.8475,
       "step": 16900
     },
     {
-      "epoch": 0.464172914778501,
+      "epoch": 0.47959704880817255,
       "grad_norm": 0.0,
-      "learning_rate": 1.16390649638599e-05,
-      "loss": 0.8153,
+      "learning_rate": 1.1144496015463346e-05,
+      "loss": 0.8655,
       "step": 16901
     },
     {
-      "epoch": 0.46420037900634425,
+      "epoch": 0.4796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.1638187471285836e-05,
-      "loss": 0.8423,
+      "learning_rate": 1.1143582971773172e-05,
+      "loss": 0.7908,
       "step": 16902
     },
     {
-      "epoch": 0.4642278432341875,
+      "epoch": 0.47965380249716233,
       "grad_norm": 0.0,
-      "learning_rate": 1.1637309965749796e-05,
-      "loss": 0.9613,
+      "learning_rate": 1.1142669918423098e-05,
+      "loss": 0.8641,
       "step": 16903
     },
     {
-      "epoch": 0.4642553074620307,
+      "epoch": 0.4796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.1636432447258728e-05,
-      "loss": 0.8626,
+      "learning_rate": 1.114175685542084e-05,
+      "loss": 0.8943,
       "step": 16904
     },
     {
-      "epoch": 0.4642827716898739,
+      "epoch": 0.4797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.1635554915819573e-05,
-      "loss": 0.8989,
+      "learning_rate": 1.1140843782774108e-05,
+      "loss": 0.8235,
       "step": 16905
     },
     {
-      "epoch": 0.4643102359177172,
+      "epoch": 0.479738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.1634677371439271e-05,
-      "loss": 0.8451,
+      "learning_rate": 1.1139930700490622e-05,
+      "loss": 0.8591,
       "step": 16906
     },
     {
-      "epoch": 0.4643377001455604,
+      "epoch": 0.4797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.1633799814124773e-05,
-      "loss": 0.8671,
+      "learning_rate": 1.1139017608578088e-05,
+      "loss": 0.8871,
       "step": 16907
     },
     {
-      "epoch": 0.46436516437340364,
+      "epoch": 0.4797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.1632922243883015e-05,
-      "loss": 1.0214,
+      "learning_rate": 1.1138104507044222e-05,
+      "loss": 0.8785,
       "step": 16908
     },
     {
-      "epoch": 0.46439262860124686,
+      "epoch": 0.47982406356413165,
       "grad_norm": 0.0,
-      "learning_rate": 1.1632044660720944e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.1137191395896736e-05,
+      "loss": 0.8146,
       "step": 16909
     },
     {
-      "epoch": 0.46442009282909014,
+      "epoch": 0.47985244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 1.1631167064645505e-05,
-      "loss": 0.9729,
+      "learning_rate": 1.1136278275143344e-05,
+      "loss": 0.9185,
       "step": 16910
     },
     {
-      "epoch": 0.46444755705693336,
+      "epoch": 0.47988081725312143,
       "grad_norm": 0.0,
-      "learning_rate": 1.1630289455663637e-05,
-      "loss": 0.9453,
+      "learning_rate": 1.1135365144791754e-05,
+      "loss": 0.8655,
       "step": 16911
     },
     {
-      "epoch": 0.4644750212847766,
+      "epoch": 0.47990919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.1629411833782292e-05,
-      "loss": 0.8101,
+      "learning_rate": 1.1134452004849687e-05,
+      "loss": 0.9684,
       "step": 16912
     },
     {
-      "epoch": 0.4645024855126198,
+      "epoch": 0.47993757094211126,
       "grad_norm": 0.0,
-      "learning_rate": 1.1628534199008406e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.1133538855324853e-05,
+      "loss": 0.8155,
       "step": 16913
     },
     {
-      "epoch": 0.464529949740463,
+      "epoch": 0.4799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.1627656551348928e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.1132625696224966e-05,
+      "loss": 0.8752,
       "step": 16914
     },
     {
-      "epoch": 0.4645574139683063,
+      "epoch": 0.47999432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.16267788908108e-05,
-      "loss": 0.8041,
+      "learning_rate": 1.1131712527557737e-05,
+      "loss": 0.8943,
       "step": 16915
     },
     {
-      "epoch": 0.4645848781961495,
+      "epoch": 0.4800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.1625901217400966e-05,
-      "loss": 0.8522,
+      "learning_rate": 1.1130799349330881e-05,
+      "loss": 0.9028,
       "step": 16916
     },
     {
-      "epoch": 0.46461234242399274,
+      "epoch": 0.4800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.1625023531126374e-05,
-      "loss": 0.8394,
+      "learning_rate": 1.1129886161552116e-05,
+      "loss": 0.9135,
       "step": 16917
     },
     {
-      "epoch": 0.46463980665183596,
+      "epoch": 0.4800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.1624145831993968e-05,
-      "loss": 0.8577,
+      "learning_rate": 1.1128972964229146e-05,
+      "loss": 0.9746,
       "step": 16918
     },
     {
-      "epoch": 0.46466727087967924,
+      "epoch": 0.4801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.1623268120010684e-05,
-      "loss": 1.0677,
+      "learning_rate": 1.112805975736969e-05,
+      "loss": 0.9939,
       "step": 16919
     },
     {
-      "epoch": 0.46469473510752246,
+      "epoch": 0.4801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.162239039518348e-05,
-      "loss": 0.8044,
+      "learning_rate": 1.1127146540981468e-05,
+      "loss": 0.9041,
       "step": 16920
     },
     {
-      "epoch": 0.4647221993353657,
+      "epoch": 0.48016458569807036,
       "grad_norm": 0.0,
-      "learning_rate": 1.1621512657519297e-05,
-      "loss": 0.8681,
+      "learning_rate": 1.1126233315072188e-05,
+      "loss": 0.8648,
       "step": 16921
     },
     {
-      "epoch": 0.4647496635632089,
+      "epoch": 0.48019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.162063490702507e-05,
-      "loss": 1.0278,
+      "learning_rate": 1.1125320079649562e-05,
+      "loss": 0.9039,
       "step": 16922
     },
     {
-      "epoch": 0.4647771277910522,
+      "epoch": 0.4802213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.1619757143707755e-05,
-      "loss": 0.881,
+      "learning_rate": 1.112440683472131e-05,
+      "loss": 0.8194,
       "step": 16923
     },
     {
-      "epoch": 0.4648045920188954,
+      "epoch": 0.48024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.1618879367574297e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.1123493580295142e-05,
+      "loss": 0.881,
       "step": 16924
     },
     {
-      "epoch": 0.4648320562467386,
+      "epoch": 0.48027809307604996,
       "grad_norm": 0.0,
-      "learning_rate": 1.1618001578631638e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.1122580316378769e-05,
+      "loss": 0.8829,
       "step": 16925
     },
     {
-      "epoch": 0.46485952047458184,
+      "epoch": 0.4803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.1617123776886719e-05,
-      "loss": 0.8837,
+      "learning_rate": 1.112166704297991e-05,
+      "loss": 0.9741,
       "step": 16926
     },
     {
-      "epoch": 0.46488698470242507,
+      "epoch": 0.48033484676503974,
       "grad_norm": 0.0,
-      "learning_rate": 1.1616245962346494e-05,
-      "loss": 0.9963,
+      "learning_rate": 1.1120753760106284e-05,
+      "loss": 0.8947,
       "step": 16927
     },
     {
-      "epoch": 0.46491444893026834,
+      "epoch": 0.4803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.1615368135017902e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.11198404677656e-05,
+      "loss": 0.9743,
       "step": 16928
     },
     {
-      "epoch": 0.46494191315811156,
+      "epoch": 0.4803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.1614490294907895e-05,
-      "loss": 0.9702,
+      "learning_rate": 1.1118927165965569e-05,
+      "loss": 0.9189,
       "step": 16929
     },
     {
-      "epoch": 0.4649693773859548,
+      "epoch": 0.4804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.1613612442023415e-05,
-      "loss": 0.8231,
+      "learning_rate": 1.1118013854713913e-05,
+      "loss": 0.9046,
       "step": 16930
     },
     {
-      "epoch": 0.464996841613798,
+      "epoch": 0.4804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.1612734576371407e-05,
-      "loss": 0.9632,
+      "learning_rate": 1.1117100534018343e-05,
+      "loss": 0.8976,
       "step": 16931
     },
     {
-      "epoch": 0.4650243058416413,
+      "epoch": 0.4804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.1611856697958819e-05,
-      "loss": 0.7847,
+      "learning_rate": 1.1116187203886575e-05,
+      "loss": 0.9025,
       "step": 16932
     },
     {
-      "epoch": 0.4650517700694845,
+      "epoch": 0.48050510783200906,
       "grad_norm": 0.0,
-      "learning_rate": 1.1610978806792597e-05,
-      "loss": 0.9749,
+      "learning_rate": 1.1115273864326324e-05,
+      "loss": 0.9824,
       "step": 16933
     },
     {
-      "epoch": 0.4650792342973277,
+      "epoch": 0.480533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.1610100902879684e-05,
-      "loss": 0.8858,
+      "learning_rate": 1.1114360515345301e-05,
+      "loss": 0.9715,
       "step": 16934
     },
     {
-      "epoch": 0.46510669852517095,
+      "epoch": 0.4805618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.160922298622703e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.1113447156951229e-05,
+      "loss": 0.7452,
       "step": 16935
     },
     {
-      "epoch": 0.4651341627530142,
+      "epoch": 0.48059023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.1608345056841581e-05,
-      "loss": 0.9756,
+      "learning_rate": 1.1112533789151816e-05,
+      "loss": 0.8552,
       "step": 16936
     },
     {
-      "epoch": 0.46516162698085745,
+      "epoch": 0.48061861520998866,
       "grad_norm": 0.0,
-      "learning_rate": 1.1607467114730284e-05,
-      "loss": 0.8978,
+      "learning_rate": 1.1111620411954782e-05,
+      "loss": 0.8902,
       "step": 16937
     },
     {
-      "epoch": 0.46518909120870067,
+      "epoch": 0.4806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.1606589159900081e-05,
-      "loss": 0.891,
+      "learning_rate": 1.111070702536784e-05,
+      "loss": 0.9445,
       "step": 16938
     },
     {
-      "epoch": 0.4652165554365439,
+      "epoch": 0.48067536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.1605711192357926e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.1109793629398705e-05,
+      "loss": 0.879,
       "step": 16939
     },
     {
-      "epoch": 0.4652440196643871,
+      "epoch": 0.4807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.160483321211076e-05,
-      "loss": 0.9536,
+      "learning_rate": 1.1108880224055093e-05,
+      "loss": 0.9284,
       "step": 16940
     },
     {
-      "epoch": 0.4652714838922304,
+      "epoch": 0.4807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.1603955219165534e-05,
-      "loss": 0.9295,
+      "learning_rate": 1.110796680934472e-05,
+      "loss": 0.9289,
       "step": 16941
     },
     {
-      "epoch": 0.4652989481200736,
+      "epoch": 0.48076049943246313,
       "grad_norm": 0.0,
-      "learning_rate": 1.1603077213529192e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.1107053385275302e-05,
+      "loss": 0.8797,
       "step": 16942
     },
     {
-      "epoch": 0.46532641234791683,
+      "epoch": 0.480788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.1602199195208681e-05,
-      "loss": 1.0353,
+      "learning_rate": 1.1106139951854555e-05,
+      "loss": 0.8567,
       "step": 16943
     },
     {
-      "epoch": 0.46535387657576005,
+      "epoch": 0.4808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.1601321164210953e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.1105226509090194e-05,
+      "loss": 0.9131,
       "step": 16944
     },
     {
-      "epoch": 0.4653813408036033,
+      "epoch": 0.48084562996594776,
       "grad_norm": 0.0,
-      "learning_rate": 1.1600443120542949e-05,
-      "loss": 0.8096,
+      "learning_rate": 1.1104313056989936e-05,
+      "loss": 0.9461,
       "step": 16945
     },
     {
-      "epoch": 0.46540880503144655,
+      "epoch": 0.4808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.1599565064211617e-05,
-      "loss": 0.9991,
+      "learning_rate": 1.1103399595561493e-05,
+      "loss": 0.9669,
       "step": 16946
     },
     {
-      "epoch": 0.46543626925928977,
+      "epoch": 0.4809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.1598686995223907e-05,
-      "loss": 0.8941,
+      "learning_rate": 1.1102486124812586e-05,
+      "loss": 0.8824,
       "step": 16947
     },
     {
-      "epoch": 0.465463733487133,
+      "epoch": 0.48093076049943245,
       "grad_norm": 0.0,
-      "learning_rate": 1.1597808913586768e-05,
-      "loss": 0.888,
+      "learning_rate": 1.1101572644750931e-05,
+      "loss": 0.9519,
       "step": 16948
     },
     {
-      "epoch": 0.46549119771497627,
+      "epoch": 0.48095913734392737,
       "grad_norm": 0.0,
-      "learning_rate": 1.1596930819307146e-05,
-      "loss": 0.9764,
+      "learning_rate": 1.1100659155384235e-05,
+      "loss": 0.918,
       "step": 16949
     },
     {
-      "epoch": 0.4655186619428195,
+      "epoch": 0.4809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.1596052712391988e-05,
-      "loss": 0.8848,
+      "learning_rate": 1.1099745656720229e-05,
+      "loss": 0.914,
       "step": 16950
     },
     {
-      "epoch": 0.4655461261706627,
+      "epoch": 0.48101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.1595174592848244e-05,
-      "loss": 0.9307,
+      "learning_rate": 1.1098832148766621e-05,
+      "loss": 0.7916,
       "step": 16951
     },
     {
-      "epoch": 0.46557359039850593,
+      "epoch": 0.48104426787741206,
       "grad_norm": 0.0,
-      "learning_rate": 1.1594296460682859e-05,
-      "loss": 0.9369,
+      "learning_rate": 1.1097918631531123e-05,
+      "loss": 0.8546,
       "step": 16952
     },
     {
-      "epoch": 0.46560105462634915,
+      "epoch": 0.4810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.1593418315902785e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.1097005105021463e-05,
+      "loss": 0.8531,
       "step": 16953
     },
     {
-      "epoch": 0.46562851885419243,
+      "epoch": 0.48110102156640183,
       "grad_norm": 0.0,
-      "learning_rate": 1.1592540158514966e-05,
-      "loss": 0.8016,
+      "learning_rate": 1.109609156924535e-05,
+      "loss": 0.9038,
       "step": 16954
     },
     {
-      "epoch": 0.46565598308203565,
+      "epoch": 0.4811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.1591661988526353e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.1095178024210502e-05,
+      "loss": 0.9409,
       "step": 16955
     },
     {
-      "epoch": 0.4656834473098789,
+      "epoch": 0.4811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.1590783805943895e-05,
-      "loss": 0.8832,
+      "learning_rate": 1.1094264469924636e-05,
+      "loss": 0.9246,
       "step": 16956
     },
     {
-      "epoch": 0.4657109115377221,
+      "epoch": 0.48118615209988647,
       "grad_norm": 0.0,
-      "learning_rate": 1.158990561077454e-05,
-      "loss": 0.9526,
+      "learning_rate": 1.1093350906395469e-05,
+      "loss": 0.8721,
       "step": 16957
     },
     {
-      "epoch": 0.46573837576556537,
+      "epoch": 0.4812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.1589027403025233e-05,
-      "loss": 0.8986,
+      "learning_rate": 1.1092437333630716e-05,
+      "loss": 0.7865,
       "step": 16958
     },
     {
-      "epoch": 0.4657658399934086,
+      "epoch": 0.4812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.1588149182702924e-05,
-      "loss": 0.9205,
+      "learning_rate": 1.1091523751638098e-05,
+      "loss": 0.9081,
       "step": 16959
     },
     {
-      "epoch": 0.4657933042212518,
+      "epoch": 0.48127128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 1.1587270949814569e-05,
-      "loss": 0.827,
+      "learning_rate": 1.1090610160425327e-05,
+      "loss": 0.8276,
       "step": 16960
     },
     {
-      "epoch": 0.46582076844909504,
+      "epoch": 0.48129965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.158639270436711e-05,
-      "loss": 0.8818,
+      "learning_rate": 1.1089696560000128e-05,
+      "loss": 0.9392,
       "step": 16961
     },
     {
-      "epoch": 0.4658482326769383,
+      "epoch": 0.48132803632236093,
       "grad_norm": 0.0,
-      "learning_rate": 1.1585514446367494e-05,
-      "loss": 0.9934,
+      "learning_rate": 1.108878295037021e-05,
+      "loss": 0.8796,
       "step": 16962
     },
     {
-      "epoch": 0.46587569690478153,
+      "epoch": 0.48135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.1584636175822674e-05,
-      "loss": 0.8061,
+      "learning_rate": 1.1087869331543293e-05,
+      "loss": 0.9037,
       "step": 16963
     },
     {
-      "epoch": 0.46590316113262475,
+      "epoch": 0.48138479001135076,
       "grad_norm": 0.0,
-      "learning_rate": 1.1583757892739602e-05,
-      "loss": 0.6989,
+      "learning_rate": 1.1086955703527093e-05,
+      "loss": 0.9245,
       "step": 16964
     },
     {
-      "epoch": 0.465930625360468,
+      "epoch": 0.4814131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.1582879597125222e-05,
-      "loss": 1.0012,
+      "learning_rate": 1.1086042066329334e-05,
+      "loss": 0.8706,
       "step": 16965
     },
     {
-      "epoch": 0.4659580895883112,
+      "epoch": 0.48144154370034054,
       "grad_norm": 0.0,
-      "learning_rate": 1.1582001288986486e-05,
-      "loss": 0.9763,
+      "learning_rate": 1.1085128419957724e-05,
+      "loss": 0.8774,
       "step": 16966
     },
     {
-      "epoch": 0.4659855538161545,
+      "epoch": 0.4814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.1581122968330342e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.1084214764419989e-05,
+      "loss": 0.8906,
       "step": 16967
     },
     {
-      "epoch": 0.4660130180439977,
+      "epoch": 0.4814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.1580244635163742e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.1083301099723844e-05,
+      "loss": 0.9373,
       "step": 16968
     },
     {
-      "epoch": 0.4660404822718409,
+      "epoch": 0.4815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.1579366289493637e-05,
-      "loss": 0.9483,
+      "learning_rate": 1.1082387425877004e-05,
+      "loss": 0.8179,
       "step": 16969
     },
     {
-      "epoch": 0.46606794649968414,
+      "epoch": 0.4815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.157848793132697e-05,
-      "loss": 0.8858,
+      "learning_rate": 1.108147374288719e-05,
+      "loss": 0.8816,
       "step": 16970
     },
     {
-      "epoch": 0.4660954107275274,
+      "epoch": 0.481583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.1577609560670693e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.1080560050762116e-05,
+      "loss": 1.002,
       "step": 16971
     },
     {
-      "epoch": 0.46612287495537064,
+      "epoch": 0.48161180476730986,
       "grad_norm": 0.0,
-      "learning_rate": 1.1576731177531762e-05,
-      "loss": 0.8665,
+      "learning_rate": 1.1079646349509505e-05,
+      "loss": 0.834,
       "step": 16972
     },
     {
-      "epoch": 0.46615033918321386,
+      "epoch": 0.4816401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.1575852781917123e-05,
-      "loss": 0.9514,
+      "learning_rate": 1.1078732639137071e-05,
+      "loss": 0.9076,
       "step": 16973
     },
     {
-      "epoch": 0.4661778034110571,
+      "epoch": 0.48166855845629963,
       "grad_norm": 0.0,
-      "learning_rate": 1.1574974373833725e-05,
-      "loss": 0.9687,
+      "learning_rate": 1.1077818919652534e-05,
+      "loss": 0.895,
       "step": 16974
     },
     {
-      "epoch": 0.46620526763890036,
+      "epoch": 0.48169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.157409595328852e-05,
-      "loss": 0.902,
+      "learning_rate": 1.1076905191063614e-05,
+      "loss": 1.0153,
       "step": 16975
     },
     {
-      "epoch": 0.4662327318667436,
+      "epoch": 0.48172531214528946,
       "grad_norm": 0.0,
-      "learning_rate": 1.157321752028846e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.1075991453378026e-05,
+      "loss": 0.8811,
       "step": 16976
     },
     {
-      "epoch": 0.4662601960945868,
+      "epoch": 0.4817536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.1572339074840493e-05,
-      "loss": 0.9065,
+      "learning_rate": 1.1075077706603493e-05,
+      "loss": 0.8941,
       "step": 16977
     },
     {
-      "epoch": 0.46628766032243,
+      "epoch": 0.48178206583427924,
       "grad_norm": 0.0,
-      "learning_rate": 1.1571460616951567e-05,
-      "loss": 0.9602,
+      "learning_rate": 1.1074163950747727e-05,
+      "loss": 0.8945,
       "step": 16978
     },
     {
-      "epoch": 0.4663151245502733,
+      "epoch": 0.4818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.1570582146628637e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.1073250185818449e-05,
+      "loss": 0.7791,
       "step": 16979
     },
     {
-      "epoch": 0.4663425887781165,
+      "epoch": 0.481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.1569703663878656e-05,
-      "loss": 0.9751,
+      "learning_rate": 1.107233641182338e-05,
+      "loss": 0.9214,
       "step": 16980
     },
     {
-      "epoch": 0.46637005300595974,
+      "epoch": 0.48186719636776393,
       "grad_norm": 0.0,
-      "learning_rate": 1.1568825168708568e-05,
-      "loss": 0.9085,
+      "learning_rate": 1.1071422628770237e-05,
+      "loss": 0.8813,
       "step": 16981
     },
     {
-      "epoch": 0.46639751723380296,
+      "epoch": 0.4818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.1567946661125328e-05,
-      "loss": 0.8632,
+      "learning_rate": 1.1070508836666738e-05,
+      "loss": 0.8736,
       "step": 16982
     },
     {
-      "epoch": 0.4664249814616462,
+      "epoch": 0.4819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.1567068141135888e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.1069595035520604e-05,
+      "loss": 0.9717,
       "step": 16983
     },
     {
-      "epoch": 0.46645244568948946,
+      "epoch": 0.48195232690124856,
       "grad_norm": 0.0,
-      "learning_rate": 1.1566189608747196e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.1068681225339552e-05,
+      "loss": 0.8875,
       "step": 16984
     },
     {
-      "epoch": 0.4664799099173327,
+      "epoch": 0.4819807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.1565311063966206e-05,
-      "loss": 0.9335,
+      "learning_rate": 1.1067767406131303e-05,
+      "loss": 0.9261,
       "step": 16985
     },
     {
-      "epoch": 0.4665073741451759,
+      "epoch": 0.4820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.1564432506799872e-05,
-      "loss": 0.9085,
+      "learning_rate": 1.1066853577903572e-05,
+      "loss": 1.0107,
       "step": 16986
     },
     {
-      "epoch": 0.4665348383730191,
+      "epoch": 0.48203745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.1563553937255136e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.1065939740664083e-05,
+      "loss": 0.8535,
       "step": 16987
     },
     {
-      "epoch": 0.4665623026008624,
+      "epoch": 0.48206583427922817,
       "grad_norm": 0.0,
-      "learning_rate": 1.1562675355338961e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.1065025894420551e-05,
+      "loss": 0.8476,
       "step": 16988
     },
     {
-      "epoch": 0.4665897668287056,
+      "epoch": 0.482094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.156179676105829e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.10641120391807e-05,
+      "loss": 0.9138,
       "step": 16989
     },
     {
-      "epoch": 0.46661723105654884,
+      "epoch": 0.48212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.1560918154420079e-05,
-      "loss": 0.8597,
+      "learning_rate": 1.1063198174952245e-05,
+      "loss": 0.8738,
       "step": 16990
     },
     {
-      "epoch": 0.46664469528439206,
+      "epoch": 0.4821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.1560039535431277e-05,
-      "loss": 0.9167,
+      "learning_rate": 1.106228430174291e-05,
+      "loss": 0.8974,
       "step": 16991
     },
     {
-      "epoch": 0.46667215951223534,
+      "epoch": 0.4821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.1559160904098841e-05,
-      "loss": 0.8846,
+      "learning_rate": 1.1061370419560408e-05,
+      "loss": 0.8578,
       "step": 16992
     },
     {
-      "epoch": 0.46669962374007856,
+      "epoch": 0.48220771850170263,
       "grad_norm": 0.0,
-      "learning_rate": 1.1558282260429719e-05,
-      "loss": 0.9396,
+      "learning_rate": 1.1060456528412464e-05,
+      "loss": 0.8788,
       "step": 16993
     },
     {
-      "epoch": 0.4667270879679218,
+      "epoch": 0.4822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.1557403604430862e-05,
-      "loss": 1.0428,
+      "learning_rate": 1.1059542628306797e-05,
+      "loss": 0.9774,
       "step": 16994
     },
     {
-      "epoch": 0.466754552195765,
+      "epoch": 0.4822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.1556524936109228e-05,
-      "loss": 0.968,
+      "learning_rate": 1.1058628719251123e-05,
+      "loss": 0.8895,
       "step": 16995
     },
     {
-      "epoch": 0.4667820164236082,
+      "epoch": 0.48229284903518727,
       "grad_norm": 0.0,
-      "learning_rate": 1.1555646255471761e-05,
-      "loss": 0.9731,
+      "learning_rate": 1.1057714801253165e-05,
+      "loss": 0.9988,
       "step": 16996
     },
     {
-      "epoch": 0.4668094806514515,
+      "epoch": 0.4823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.1554767562525422e-05,
-      "loss": 1.0059,
+      "learning_rate": 1.1056800874320646e-05,
+      "loss": 0.8809,
       "step": 16997
     },
     {
-      "epoch": 0.4668369448792947,
+      "epoch": 0.4823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.155388885727716e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.1055886938461282e-05,
+      "loss": 1.0122,
       "step": 16998
     },
     {
-      "epoch": 0.46686440910713795,
+      "epoch": 0.48237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.1553010139733924e-05,
-      "loss": 0.8427,
+      "learning_rate": 1.1054972993682794e-05,
+      "loss": 0.7949,
       "step": 16999
     },
     {
-      "epoch": 0.46689187333498117,
+      "epoch": 0.48240635641316687,
       "grad_norm": 0.0,
-      "learning_rate": 1.1552131409902674e-05,
-      "loss": 0.8021,
+      "learning_rate": 1.1054059039992897e-05,
+      "loss": 1.0315,
       "step": 17000
     },
     {
-      "epoch": 0.46691933756282444,
+      "epoch": 0.48243473325766173,
       "grad_norm": 0.0,
-      "learning_rate": 1.155125266779036e-05,
-      "loss": 0.8764,
+      "learning_rate": 1.1053145077399318e-05,
+      "loss": 0.9264,
       "step": 17001
     },
     {
-      "epoch": 0.46694680179066766,
+      "epoch": 0.48246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.155037391340393e-05,
-      "loss": 0.9114,
+      "learning_rate": 1.1052231105909778e-05,
+      "loss": 0.8347,
       "step": 17002
     },
     {
-      "epoch": 0.4669742660185109,
+      "epoch": 0.48249148694665156,
       "grad_norm": 0.0,
-      "learning_rate": 1.154949514675034e-05,
-      "loss": 0.777,
+      "learning_rate": 1.1051317125531989e-05,
+      "loss": 0.8567,
       "step": 17003
     },
     {
-      "epoch": 0.4670017302463541,
+      "epoch": 0.4825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.154861636783655e-05,
-      "loss": 0.8769,
+      "learning_rate": 1.1050403136273681e-05,
+      "loss": 0.9104,
       "step": 17004
     },
     {
-      "epoch": 0.4670291944741974,
+      "epoch": 0.48254824063564133,
       "grad_norm": 0.0,
-      "learning_rate": 1.1547737576669505e-05,
-      "loss": 0.8605,
+      "learning_rate": 1.1049489138142571e-05,
+      "loss": 0.7928,
       "step": 17005
     },
     {
-      "epoch": 0.4670566587020406,
+      "epoch": 0.4825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.1546858773256157e-05,
-      "loss": 0.856,
+      "learning_rate": 1.1048575131146377e-05,
+      "loss": 0.9427,
       "step": 17006
     },
     {
-      "epoch": 0.4670841229298838,
+      "epoch": 0.4826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.1545979957603468e-05,
-      "loss": 0.879,
+      "learning_rate": 1.1047661115292825e-05,
+      "loss": 0.8016,
       "step": 17007
     },
     {
-      "epoch": 0.46711158715772705,
+      "epoch": 0.48263337116912597,
       "grad_norm": 0.0,
-      "learning_rate": 1.1545101129718383e-05,
-      "loss": 0.7822,
+      "learning_rate": 1.1046747090589629e-05,
+      "loss": 0.9362,
       "step": 17008
     },
     {
-      "epoch": 0.46713905138557027,
+      "epoch": 0.4826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.1544222289607863e-05,
-      "loss": 0.8911,
+      "learning_rate": 1.1045833057044515e-05,
+      "loss": 0.8351,
       "step": 17009
     },
     {
-      "epoch": 0.46716651561341355,
+      "epoch": 0.4826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.1543343437278856e-05,
-      "loss": 0.8976,
+      "learning_rate": 1.1044919014665198e-05,
+      "loss": 0.8695,
       "step": 17010
     },
     {
-      "epoch": 0.46719397984125677,
+      "epoch": 0.48271850170261066,
       "grad_norm": 0.0,
-      "learning_rate": 1.1542464572738317e-05,
-      "loss": 0.8687,
+      "learning_rate": 1.1044004963459409e-05,
+      "loss": 0.8419,
       "step": 17011
     },
     {
-      "epoch": 0.4672214440691,
+      "epoch": 0.4827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.1541585695993202e-05,
-      "loss": 0.984,
+      "learning_rate": 1.104309090343486e-05,
+      "loss": 0.7379,
       "step": 17012
     },
     {
-      "epoch": 0.4672489082969432,
+      "epoch": 0.48277525539160043,
       "grad_norm": 0.0,
-      "learning_rate": 1.1540706807050466e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.1042176834599276e-05,
+      "loss": 0.8501,
       "step": 17013
     },
     {
-      "epoch": 0.4672763725247865,
+      "epoch": 0.48280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.1539827905917058e-05,
-      "loss": 0.8728,
+      "learning_rate": 1.1041262756960378e-05,
+      "loss": 0.9544,
       "step": 17014
     },
     {
-      "epoch": 0.4673038367526297,
+      "epoch": 0.48283200908059026,
       "grad_norm": 0.0,
-      "learning_rate": 1.1538948992599934e-05,
-      "loss": 0.9428,
+      "learning_rate": 1.1040348670525889e-05,
+      "loss": 0.9568,
       "step": 17015
     },
     {
-      "epoch": 0.46733130098047293,
+      "epoch": 0.4828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.1538070067106052e-05,
-      "loss": 0.9084,
+      "learning_rate": 1.1039434575303522e-05,
+      "loss": 0.9238,
       "step": 17016
     },
     {
-      "epoch": 0.46735876520831615,
+      "epoch": 0.48288876276958004,
       "grad_norm": 0.0,
-      "learning_rate": 1.1537191129442364e-05,
-      "loss": 0.9799,
+      "learning_rate": 1.1038520471301008e-05,
+      "loss": 0.8782,
       "step": 17017
     },
     {
-      "epoch": 0.46738622943615943,
+      "epoch": 0.4829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.1536312179615823e-05,
-      "loss": 0.9324,
+      "learning_rate": 1.1037606358526065e-05,
+      "loss": 0.8341,
       "step": 17018
     },
     {
-      "epoch": 0.46741369366400265,
+      "epoch": 0.4829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.1535433217633384e-05,
-      "loss": 0.7937,
+      "learning_rate": 1.1036692236986416e-05,
+      "loss": 0.8747,
       "step": 17019
     },
     {
-      "epoch": 0.46744115789184587,
+      "epoch": 0.48297389330306467,
       "grad_norm": 0.0,
-      "learning_rate": 1.1534554243502005e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.1035778106689776e-05,
+      "loss": 0.8934,
       "step": 17020
     },
     {
-      "epoch": 0.4674686221196891,
+      "epoch": 0.4830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.1533675257228638e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.1034863967643877e-05,
+      "loss": 0.9539,
       "step": 17021
     },
     {
-      "epoch": 0.4674960863475323,
+      "epoch": 0.4830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.1532796258820238e-05,
-      "loss": 0.901,
+      "learning_rate": 1.1033949819856432e-05,
+      "loss": 0.9078,
       "step": 17022
     },
     {
-      "epoch": 0.4675235505753756,
+      "epoch": 0.48305902383654936,
       "grad_norm": 0.0,
-      "learning_rate": 1.153191724828376e-05,
-      "loss": 0.8719,
+      "learning_rate": 1.1033035663335167e-05,
+      "loss": 0.8992,
       "step": 17023
     },
     {
-      "epoch": 0.4675510148032188,
+      "epoch": 0.4830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.153103822562616e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.1032121498087807e-05,
+      "loss": 0.8742,
       "step": 17024
     },
     {
-      "epoch": 0.46757847903106203,
+      "epoch": 0.48311577752553914,
       "grad_norm": 0.0,
-      "learning_rate": 1.153015919085439e-05,
-      "loss": 0.9243,
+      "learning_rate": 1.1031207324122066e-05,
+      "loss": 0.8462,
       "step": 17025
     },
     {
-      "epoch": 0.46760594325890525,
+      "epoch": 0.48314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.152928014397541e-05,
-      "loss": 0.8411,
+      "learning_rate": 1.1030293141445672e-05,
+      "loss": 0.9258,
       "step": 17026
     },
     {
-      "epoch": 0.46763340748674853,
+      "epoch": 0.48317253121452897,
       "grad_norm": 0.0,
-      "learning_rate": 1.152840108499617e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.1029378950066343e-05,
+      "loss": 0.8917,
       "step": 17027
     },
     {
-      "epoch": 0.46766087171459175,
+      "epoch": 0.4832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.1527522013923634e-05,
-      "loss": 0.9286,
+      "learning_rate": 1.1028464749991807e-05,
+      "loss": 0.8795,
       "step": 17028
     },
     {
-      "epoch": 0.467688335942435,
+      "epoch": 0.48322928490351874,
       "grad_norm": 0.0,
-      "learning_rate": 1.1526642930764748e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.1027550541229783e-05,
+      "loss": 0.7897,
       "step": 17029
     },
     {
-      "epoch": 0.4677158001702782,
+      "epoch": 0.4832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.1525763835526473e-05,
-      "loss": 1.0662,
+      "learning_rate": 1.102663632378799e-05,
+      "loss": 0.8484,
       "step": 17030
     },
     {
-      "epoch": 0.46774326439812147,
+      "epoch": 0.4832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.1524884728215763e-05,
-      "loss": 0.8824,
+      "learning_rate": 1.1025722097674158e-05,
+      "loss": 0.9008,
       "step": 17031
     },
     {
-      "epoch": 0.4677707286259647,
+      "epoch": 0.48331441543700343,
       "grad_norm": 0.0,
-      "learning_rate": 1.1524005608839575e-05,
-      "loss": 0.951,
+      "learning_rate": 1.1024807862896005e-05,
+      "loss": 0.9301,
       "step": 17032
     },
     {
-      "epoch": 0.4677981928538079,
+      "epoch": 0.4833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.1523126477404865e-05,
-      "loss": 0.9167,
+      "learning_rate": 1.1023893619461253e-05,
+      "loss": 0.8922,
       "step": 17033
     },
     {
-      "epoch": 0.46782565708165114,
+      "epoch": 0.4833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.1522247333918589e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.1022979367377622e-05,
+      "loss": 0.8643,
       "step": 17034
     },
     {
-      "epoch": 0.46785312130949436,
+      "epoch": 0.48339954597048806,
       "grad_norm": 0.0,
-      "learning_rate": 1.15213681783877e-05,
-      "loss": 0.82,
+      "learning_rate": 1.1022065106652842e-05,
+      "loss": 0.8756,
       "step": 17035
     },
     {
-      "epoch": 0.46788058553733763,
+      "epoch": 0.483427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.1520489010819158e-05,
-      "loss": 0.8708,
+      "learning_rate": 1.1021150837294632e-05,
+      "loss": 0.9736,
       "step": 17036
     },
     {
-      "epoch": 0.46790804976518086,
+      "epoch": 0.48345629965947784,
       "grad_norm": 0.0,
-      "learning_rate": 1.1519609831219918e-05,
-      "loss": 0.7912,
+      "learning_rate": 1.1020236559310714e-05,
+      "loss": 1.0368,
       "step": 17037
     },
     {
-      "epoch": 0.4679355139930241,
+      "epoch": 0.48348467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.1518730639596936e-05,
-      "loss": 0.8221,
+      "learning_rate": 1.1019322272708812e-05,
+      "loss": 1.0088,
       "step": 17038
     },
     {
-      "epoch": 0.4679629782208673,
+      "epoch": 0.48351305334846767,
       "grad_norm": 0.0,
-      "learning_rate": 1.1517851435957168e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.1018407977496652e-05,
+      "loss": 0.9561,
       "step": 17039
     },
     {
-      "epoch": 0.4679904424487106,
+      "epoch": 0.48354143019296253,
       "grad_norm": 0.0,
-      "learning_rate": 1.1516972220307573e-05,
-      "loss": 0.9139,
+      "learning_rate": 1.1017493673681949e-05,
+      "loss": 0.9762,
       "step": 17040
     },
     {
-      "epoch": 0.4680179066765538,
+      "epoch": 0.48356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.1516092992655105e-05,
-      "loss": 0.857,
+      "learning_rate": 1.1016579361272432e-05,
+      "loss": 0.8651,
       "step": 17041
     },
     {
-      "epoch": 0.468045370904397,
+      "epoch": 0.4835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.1515213753006723e-05,
-      "loss": 0.9278,
+      "learning_rate": 1.1015665040275827e-05,
+      "loss": 0.8137,
       "step": 17042
     },
     {
-      "epoch": 0.46807283513224024,
+      "epoch": 0.4836265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.1514334501369381e-05,
-      "loss": 1.0064,
+      "learning_rate": 1.1014750710699853e-05,
+      "loss": 0.8788,
       "step": 17043
     },
     {
-      "epoch": 0.4681002993600835,
+      "epoch": 0.48365493757094213,
       "grad_norm": 0.0,
-      "learning_rate": 1.151345523775004e-05,
-      "loss": 0.8968,
+      "learning_rate": 1.101383637255223e-05,
+      "loss": 1.0157,
       "step": 17044
     },
     {
-      "epoch": 0.46812776358792674,
+      "epoch": 0.483683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1512575962155657e-05,
-      "loss": 0.7832,
+      "learning_rate": 1.1012922025840691e-05,
+      "loss": 1.0009,
       "step": 17045
     },
     {
-      "epoch": 0.46815522781576996,
+      "epoch": 0.4837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.151169667459318e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.101200767057295e-05,
+      "loss": 0.82,
       "step": 17046
     },
     {
-      "epoch": 0.4681826920436132,
+      "epoch": 0.48374006810442677,
       "grad_norm": 0.0,
-      "learning_rate": 1.1510817375069579e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.1011093306756737e-05,
+      "loss": 0.7448,
       "step": 17047
     },
     {
-      "epoch": 0.4682101562714564,
+      "epoch": 0.4837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.1509938063591804e-05,
-      "loss": 0.9443,
+      "learning_rate": 1.1010178934399773e-05,
+      "loss": 0.9193,
       "step": 17048
     },
     {
-      "epoch": 0.4682376204992997,
+      "epoch": 0.4837968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.1509058740166813e-05,
-      "loss": 0.8895,
+      "learning_rate": 1.100926455350978e-05,
+      "loss": 0.8176,
       "step": 17049
     },
     {
-      "epoch": 0.4682650847271429,
+      "epoch": 0.48382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.1508179404801567e-05,
-      "loss": 0.901,
+      "learning_rate": 1.1008350164094486e-05,
+      "loss": 0.8932,
       "step": 17050
     },
     {
-      "epoch": 0.4682925489549861,
+      "epoch": 0.4838535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.1507300057503021e-05,
-      "loss": 1.002,
+      "learning_rate": 1.1007435766161612e-05,
+      "loss": 0.7619,
       "step": 17051
     },
     {
-      "epoch": 0.46832001318282934,
+      "epoch": 0.48388195232690123,
       "grad_norm": 0.0,
-      "learning_rate": 1.1506420698278131e-05,
-      "loss": 0.9797,
+      "learning_rate": 1.1006521359718885e-05,
+      "loss": 0.9353,
       "step": 17052
     },
     {
-      "epoch": 0.4683474774106726,
+      "epoch": 0.48391032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 1.1505541327133857e-05,
-      "loss": 0.9368,
+      "learning_rate": 1.1005606944774025e-05,
+      "loss": 0.9141,
       "step": 17053
     },
     {
-      "epoch": 0.46837494163851584,
+      "epoch": 0.483938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.150466194407716e-05,
-      "loss": 0.9537,
+      "learning_rate": 1.1004692521334757e-05,
+      "loss": 0.9705,
       "step": 17054
     },
     {
-      "epoch": 0.46840240586635906,
+      "epoch": 0.4839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.1503782549114992e-05,
-      "loss": 0.8868,
+      "learning_rate": 1.1003778089408808e-05,
+      "loss": 0.8903,
       "step": 17055
     },
     {
-      "epoch": 0.4684298700942023,
+      "epoch": 0.48399545970488084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1502903142254316e-05,
-      "loss": 0.8308,
+      "learning_rate": 1.1002863649003898e-05,
+      "loss": 0.9424,
       "step": 17056
     },
     {
-      "epoch": 0.46845733432204556,
+      "epoch": 0.4840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.1502023723502086e-05,
-      "loss": 0.8673,
+      "learning_rate": 1.1001949200127756e-05,
+      "loss": 0.8033,
       "step": 17057
     },
     {
-      "epoch": 0.4684847985498888,
+      "epoch": 0.4840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.1501144292865265e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.1001034742788104e-05,
+      "loss": 0.8338,
       "step": 17058
     },
     {
-      "epoch": 0.468512262777732,
+      "epoch": 0.48408059023836547,
       "grad_norm": 0.0,
-      "learning_rate": 1.1500264850350806e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.1000120276992665e-05,
+      "loss": 0.9107,
       "step": 17059
     },
     {
-      "epoch": 0.4685397270055752,
+      "epoch": 0.4841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.1499385395965673e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.0999205802749165e-05,
+      "loss": 0.9382,
       "step": 17060
     },
     {
-      "epoch": 0.46856719123341845,
+      "epoch": 0.4841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.1498505929716821e-05,
-      "loss": 0.9631,
+      "learning_rate": 1.0998291320065331e-05,
+      "loss": 0.7609,
       "step": 17061
     },
     {
-      "epoch": 0.4685946554612617,
+      "epoch": 0.48416572077185016,
       "grad_norm": 0.0,
-      "learning_rate": 1.1497626451611207e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.0997376828948885e-05,
+      "loss": 0.8841,
       "step": 17062
     },
     {
-      "epoch": 0.46862211968910494,
+      "epoch": 0.4841940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.1496746961655796e-05,
-      "loss": 0.9323,
+      "learning_rate": 1.0996462329407549e-05,
+      "loss": 0.8438,
       "step": 17063
     },
     {
-      "epoch": 0.46864958391694816,
+      "epoch": 0.48422247446083994,
       "grad_norm": 0.0,
-      "learning_rate": 1.1495867459857542e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.0995547821449053e-05,
+      "loss": 0.9058,
       "step": 17064
     },
     {
-      "epoch": 0.4686770481447914,
+      "epoch": 0.48425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.1494987946223406e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.099463330508112e-05,
+      "loss": 0.9209,
       "step": 17065
     },
     {
-      "epoch": 0.46870451237263466,
+      "epoch": 0.48427922814982977,
       "grad_norm": 0.0,
-      "learning_rate": 1.1494108420760346e-05,
-      "loss": 0.9528,
+      "learning_rate": 1.0993718780311475e-05,
+      "loss": 0.8684,
       "step": 17066
     },
     {
-      "epoch": 0.4687319766004779,
+      "epoch": 0.4843076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.1493228883475319e-05,
-      "loss": 0.8661,
+      "learning_rate": 1.0992804247147841e-05,
+      "loss": 0.8123,
       "step": 17067
     },
     {
-      "epoch": 0.4687594408283211,
+      "epoch": 0.48433598183881954,
       "grad_norm": 0.0,
-      "learning_rate": 1.149234933437529e-05,
-      "loss": 0.9741,
+      "learning_rate": 1.0991889705597946e-05,
+      "loss": 0.9012,
       "step": 17068
     },
     {
-      "epoch": 0.4687869050561643,
+      "epoch": 0.4843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.1491469773467214e-05,
-      "loss": 0.8449,
+      "learning_rate": 1.0990975155669516e-05,
+      "loss": 0.9158,
       "step": 17069
     },
     {
-      "epoch": 0.4688143692840076,
+      "epoch": 0.4843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.149059020075805e-05,
-      "loss": 0.9727,
+      "learning_rate": 1.0990060597370271e-05,
+      "loss": 0.8356,
       "step": 17070
     },
     {
-      "epoch": 0.4688418335118508,
+      "epoch": 0.4844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.1489710616254759e-05,
-      "loss": 0.9111,
+      "learning_rate": 1.0989146030707942e-05,
+      "loss": 0.8399,
       "step": 17071
     },
     {
-      "epoch": 0.46886929773969405,
+      "epoch": 0.4844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.1488831019964301e-05,
-      "loss": 1.0274,
+      "learning_rate": 1.098823145569025e-05,
+      "loss": 0.813,
       "step": 17072
     },
     {
-      "epoch": 0.46889676196753727,
+      "epoch": 0.484477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.1487951411893637e-05,
-      "loss": 0.9051,
+      "learning_rate": 1.0987316872324925e-05,
+      "loss": 0.7815,
       "step": 17073
     },
     {
-      "epoch": 0.46892422619538054,
+      "epoch": 0.48450624290578886,
       "grad_norm": 0.0,
-      "learning_rate": 1.1487071792049722e-05,
-      "loss": 0.8344,
+      "learning_rate": 1.0986402280619689e-05,
+      "loss": 0.8252,
       "step": 17074
     },
     {
-      "epoch": 0.46895169042322377,
+      "epoch": 0.4845346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.1486192160439523e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.0985487680582267e-05,
+      "loss": 0.9134,
       "step": 17075
     },
     {
-      "epoch": 0.468979154651067,
+      "epoch": 0.48456299659477864,
       "grad_norm": 0.0,
-      "learning_rate": 1.1485312517069992e-05,
-      "loss": 0.8385,
+      "learning_rate": 1.0984573072220388e-05,
+      "loss": 0.8363,
       "step": 17076
     },
     {
-      "epoch": 0.4690066188789102,
+      "epoch": 0.48459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.1484432861948097e-05,
-      "loss": 0.8377,
+      "learning_rate": 1.0983658455541774e-05,
+      "loss": 0.9463,
       "step": 17077
     },
     {
-      "epoch": 0.46903408310675343,
+      "epoch": 0.48461975028376847,
       "grad_norm": 0.0,
-      "learning_rate": 1.148355319508079e-05,
-      "loss": 0.8058,
+      "learning_rate": 1.0982743830554157e-05,
+      "loss": 0.9157,
       "step": 17078
     },
     {
-      "epoch": 0.4690615473345967,
+      "epoch": 0.48464812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.1482673516475035e-05,
-      "loss": 0.9231,
+      "learning_rate": 1.0981829197265254e-05,
+      "loss": 0.8974,
       "step": 17079
     },
     {
-      "epoch": 0.46908901156243993,
+      "epoch": 0.48467650397275824,
       "grad_norm": 0.0,
-      "learning_rate": 1.1481793826137796e-05,
-      "loss": 0.9568,
+      "learning_rate": 1.0980914555682797e-05,
+      "loss": 0.9617,
       "step": 17080
     },
     {
-      "epoch": 0.46911647579028315,
+      "epoch": 0.4847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.1480914124076033e-05,
-      "loss": 0.962,
+      "learning_rate": 1.0979999905814512e-05,
+      "loss": 0.9702,
       "step": 17081
     },
     {
-      "epoch": 0.46914394001812637,
+      "epoch": 0.484733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.1480034410296696e-05,
-      "loss": 0.8262,
+      "learning_rate": 1.0979085247668123e-05,
+      "loss": 0.8618,
       "step": 17082
     },
     {
-      "epoch": 0.46917140424596965,
+      "epoch": 0.48476163450624293,
       "grad_norm": 0.0,
-      "learning_rate": 1.1479154684806757e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.0978170581251358e-05,
+      "loss": 0.8666,
       "step": 17083
     },
     {
-      "epoch": 0.46919886847381287,
+      "epoch": 0.4847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.1478274947613175e-05,
-      "loss": 0.9689,
+      "learning_rate": 1.097725590657194e-05,
+      "loss": 0.8547,
       "step": 17084
     },
     {
-      "epoch": 0.4692263327016561,
+      "epoch": 0.4848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.1477395198722909e-05,
-      "loss": 0.971,
+      "learning_rate": 1.0976341223637599e-05,
+      "loss": 0.8453,
       "step": 17085
     },
     {
-      "epoch": 0.4692537969294993,
+      "epoch": 0.48484676503972757,
       "grad_norm": 0.0,
-      "learning_rate": 1.1476515438142917e-05,
-      "loss": 0.9603,
+      "learning_rate": 1.0975426532456059e-05,
+      "loss": 0.8535,
       "step": 17086
     },
     {
-      "epoch": 0.4692812611573426,
+      "epoch": 0.4848751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.1475635665880165e-05,
-      "loss": 0.9389,
+      "learning_rate": 1.0974511833035048e-05,
+      "loss": 0.924,
       "step": 17087
     },
     {
-      "epoch": 0.4693087253851858,
+      "epoch": 0.48490351872871734,
       "grad_norm": 0.0,
-      "learning_rate": 1.147475588194161e-05,
-      "loss": 0.8711,
+      "learning_rate": 1.0973597125382289e-05,
+      "loss": 0.8397,
       "step": 17088
     },
     {
-      "epoch": 0.46933618961302903,
+      "epoch": 0.48493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.1473876086334217e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.0972682409505515e-05,
+      "loss": 0.7233,
       "step": 17089
     },
     {
-      "epoch": 0.46936365384087225,
+      "epoch": 0.48496027241770717,
       "grad_norm": 0.0,
-      "learning_rate": 1.1472996279064948e-05,
-      "loss": 1.0239,
+      "learning_rate": 1.097176768541245e-05,
+      "loss": 0.8657,
       "step": 17090
     },
     {
-      "epoch": 0.4693911180687155,
+      "epoch": 0.48498864926220203,
       "grad_norm": 0.0,
-      "learning_rate": 1.1472116460140757e-05,
-      "loss": 1.1042,
+      "learning_rate": 1.0970852953110812e-05,
+      "loss": 0.8319,
       "step": 17091
     },
     {
-      "epoch": 0.46941858229655875,
+      "epoch": 0.48501702610669695,
       "grad_norm": 0.0,
-      "learning_rate": 1.1471236629568613e-05,
-      "loss": 0.8511,
+      "learning_rate": 1.0969938212608345e-05,
+      "loss": 0.8402,
       "step": 17092
     },
     {
-      "epoch": 0.46944604652440197,
+      "epoch": 0.4850454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.1470356787355474e-05,
-      "loss": 0.9966,
+      "learning_rate": 1.0969023463912763e-05,
+      "loss": 0.8701,
       "step": 17093
     },
     {
-      "epoch": 0.4694735107522452,
+      "epoch": 0.4850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.1469476933508303e-05,
-      "loss": 0.8909,
+      "learning_rate": 1.0968108707031792e-05,
+      "loss": 0.9046,
       "step": 17094
     },
     {
-      "epoch": 0.4695009749800884,
+      "epoch": 0.48510215664018164,
       "grad_norm": 0.0,
-      "learning_rate": 1.1468597068034061e-05,
-      "loss": 0.8724,
+      "learning_rate": 1.0967193941973167e-05,
+      "loss": 1.0093,
       "step": 17095
     },
     {
-      "epoch": 0.4695284392079317,
+      "epoch": 0.4851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.1467717190939709e-05,
-      "loss": 0.9372,
+      "learning_rate": 1.096627916874461e-05,
+      "loss": 0.9496,
       "step": 17096
     },
     {
-      "epoch": 0.4695559034357749,
+      "epoch": 0.4851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.1466837302232212e-05,
-      "loss": 0.7937,
+      "learning_rate": 1.0965364387353852e-05,
+      "loss": 0.9779,
       "step": 17097
     },
     {
-      "epoch": 0.46958336766361813,
+      "epoch": 0.48518728717366627,
       "grad_norm": 0.0,
-      "learning_rate": 1.146595740191853e-05,
-      "loss": 0.8408,
+      "learning_rate": 1.0964449597808612e-05,
+      "loss": 0.8835,
       "step": 17098
     },
     {
-      "epoch": 0.46961083189146136,
+      "epoch": 0.4852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.1465077490005623e-05,
-      "loss": 0.8651,
+      "learning_rate": 1.0963534800116628e-05,
+      "loss": 0.8932,
       "step": 17099
     },
     {
-      "epoch": 0.46963829611930463,
+      "epoch": 0.48524404086265605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1464197566500456e-05,
-      "loss": 0.9249,
+      "learning_rate": 1.0962619994285623e-05,
+      "loss": 0.8987,
       "step": 17100
     },
     {
-      "epoch": 0.46966576034714785,
+      "epoch": 0.48527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.1463317631409996e-05,
-      "loss": 0.8709,
+      "learning_rate": 1.096170518032332e-05,
+      "loss": 0.9232,
       "step": 17101
     },
     {
-      "epoch": 0.4696932245749911,
+      "epoch": 0.4853007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.1462437684741192e-05,
-      "loss": 0.9107,
+      "learning_rate": 1.096079035823745e-05,
+      "loss": 0.9058,
       "step": 17102
     },
     {
-      "epoch": 0.4697206888028343,
+      "epoch": 0.48532917139614073,
       "grad_norm": 0.0,
-      "learning_rate": 1.1461557726501017e-05,
-      "loss": 0.8752,
+      "learning_rate": 1.0959875528035743e-05,
+      "loss": 0.8978,
       "step": 17103
     },
     {
-      "epoch": 0.4697481530306775,
+      "epoch": 0.48535754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 1.1460677756696435e-05,
-      "loss": 0.9123,
+      "learning_rate": 1.0958960689725924e-05,
+      "loss": 0.8232,
       "step": 17104
     },
     {
-      "epoch": 0.4697756172585208,
+      "epoch": 0.4853859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.1459797775334402e-05,
-      "loss": 0.8478,
+      "learning_rate": 1.0958045843315722e-05,
+      "loss": 0.9231,
       "step": 17105
     },
     {
-      "epoch": 0.469803081486364,
+      "epoch": 0.4854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.1458917782421883e-05,
-      "loss": 0.9824,
+      "learning_rate": 1.0957130988812863e-05,
+      "loss": 0.9313,
       "step": 17106
     },
     {
-      "epoch": 0.46983054571420724,
+      "epoch": 0.48544267877412034,
       "grad_norm": 0.0,
-      "learning_rate": 1.1458037777965842e-05,
-      "loss": 0.9784,
+      "learning_rate": 1.0956216126225074e-05,
+      "loss": 0.9985,
       "step": 17107
     },
     {
-      "epoch": 0.46985800994205046,
+      "epoch": 0.4854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.1457157761973241e-05,
-      "loss": 0.9844,
+      "learning_rate": 1.0955301255560085e-05,
+      "loss": 0.8129,
       "step": 17108
     },
     {
-      "epoch": 0.46988547416989374,
+      "epoch": 0.4854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.1456277734451041e-05,
-      "loss": 0.8732,
+      "learning_rate": 1.0954386376825624e-05,
+      "loss": 0.9647,
       "step": 17109
     },
     {
-      "epoch": 0.46991293839773696,
+      "epoch": 0.485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1455397695406214e-05,
-      "loss": 0.9022,
+      "learning_rate": 1.0953471490029417e-05,
+      "loss": 0.893,
       "step": 17110
     },
     {
-      "epoch": 0.4699404026255802,
+      "epoch": 0.4855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.145451764484571e-05,
-      "loss": 1.0014,
+      "learning_rate": 1.0952556595179193e-05,
+      "loss": 0.9093,
       "step": 17111
     },
     {
-      "epoch": 0.4699678668534234,
+      "epoch": 0.4855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.1453637582776502e-05,
-      "loss": 0.8549,
+      "learning_rate": 1.0951641692282682e-05,
+      "loss": 0.8551,
       "step": 17112
     },
     {
-      "epoch": 0.4699953310812667,
+      "epoch": 0.48561293984108966,
       "grad_norm": 0.0,
-      "learning_rate": 1.1452757509205553e-05,
-      "loss": 0.9425,
+      "learning_rate": 1.0950726781347612e-05,
+      "loss": 1.0056,
       "step": 17113
     },
     {
-      "epoch": 0.4700227953091099,
+      "epoch": 0.4856413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.1451877424139819e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.0949811862381708e-05,
+      "loss": 0.9038,
       "step": 17114
     },
     {
-      "epoch": 0.4700502595369531,
+      "epoch": 0.48566969353007944,
       "grad_norm": 0.0,
-      "learning_rate": 1.1450997327586268e-05,
-      "loss": 0.9871,
+      "learning_rate": 1.09488969353927e-05,
+      "loss": 0.8653,
       "step": 17115
     },
     {
-      "epoch": 0.47007772376479634,
+      "epoch": 0.48569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.1450117219551869e-05,
-      "loss": 0.8475,
+      "learning_rate": 1.0947982000388322e-05,
+      "loss": 0.8704,
       "step": 17116
     },
     {
-      "epoch": 0.47010518799263956,
+      "epoch": 0.4857264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.1449237100043578e-05,
-      "loss": 0.7905,
+      "learning_rate": 1.0947067057376291e-05,
+      "loss": 0.9951,
       "step": 17117
     },
     {
-      "epoch": 0.47013265222048284,
+      "epoch": 0.48575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.1448356969068362e-05,
-      "loss": 0.886,
+      "learning_rate": 1.0946152106364344e-05,
+      "loss": 0.9244,
       "step": 17118
     },
     {
-      "epoch": 0.47016011644832606,
+      "epoch": 0.48578320090805904,
       "grad_norm": 0.0,
-      "learning_rate": 1.1447476826633183e-05,
-      "loss": 0.8059,
+      "learning_rate": 1.0945237147360208e-05,
+      "loss": 0.8852,
       "step": 17119
     },
     {
-      "epoch": 0.4701875806761693,
+      "epoch": 0.4858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.1446596672745007e-05,
-      "loss": 0.8541,
+      "learning_rate": 1.0944322180371614e-05,
+      "loss": 0.9646,
       "step": 17120
     },
     {
-      "epoch": 0.4702150449040125,
+      "epoch": 0.4858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.14457165074108e-05,
-      "loss": 0.8741,
+      "learning_rate": 1.0943407205406284e-05,
+      "loss": 0.9275,
       "step": 17121
     },
     {
-      "epoch": 0.4702425091318558,
+      "epoch": 0.4858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1444836330637525e-05,
-      "loss": 0.8914,
+      "learning_rate": 1.0942492222471954e-05,
+      "loss": 0.7253,
       "step": 17122
     },
     {
-      "epoch": 0.470269973359699,
+      "epoch": 0.4858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.1443956142432141e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.094157723157635e-05,
+      "loss": 0.8759,
       "step": 17123
     },
     {
-      "epoch": 0.4702974375875422,
+      "epoch": 0.4859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.144307594280162e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.0940662232727198e-05,
+      "loss": 0.9012,
       "step": 17124
     },
     {
-      "epoch": 0.47032490181538544,
+      "epoch": 0.48595346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 1.1442195731752924e-05,
-      "loss": 0.9888,
+      "learning_rate": 1.0939747225932229e-05,
+      "loss": 0.8898,
       "step": 17125
     },
     {
-      "epoch": 0.4703523660432287,
+      "epoch": 0.4859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.1441315509293011e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.0938832211199177e-05,
+      "loss": 0.8176,
       "step": 17126
     },
     {
-      "epoch": 0.47037983027107194,
+      "epoch": 0.48601021566401814,
       "grad_norm": 0.0,
-      "learning_rate": 1.1440435275428857e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.0937917188535765e-05,
+      "loss": 0.9729,
       "step": 17127
     },
     {
-      "epoch": 0.47040729449891516,
+      "epoch": 0.48603859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.1439555030167417e-05,
-      "loss": 0.9434,
+      "learning_rate": 1.0937002157949723e-05,
+      "loss": 0.9669,
       "step": 17128
     },
     {
-      "epoch": 0.4704347587267584,
+      "epoch": 0.48606696935300797,
       "grad_norm": 0.0,
-      "learning_rate": 1.1438674773515666e-05,
-      "loss": 0.827,
+      "learning_rate": 1.0936087119448784e-05,
+      "loss": 0.8902,
       "step": 17129
     },
     {
-      "epoch": 0.4704622229546016,
+      "epoch": 0.48609534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.1437794505480558e-05,
-      "loss": 1.0624,
+      "learning_rate": 1.0935172073040676e-05,
+      "loss": 1.0114,
       "step": 17130
     },
     {
-      "epoch": 0.4704896871824449,
+      "epoch": 0.48612372304199775,
       "grad_norm": 0.0,
-      "learning_rate": 1.1436914226069062e-05,
-      "loss": 0.9754,
+      "learning_rate": 1.0934257018733125e-05,
+      "loss": 0.8715,
       "step": 17131
     },
     {
-      "epoch": 0.4705171514102881,
+      "epoch": 0.4861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.1436033935288146e-05,
-      "loss": 0.8453,
+      "learning_rate": 1.0933341956533865e-05,
+      "loss": 0.8595,
       "step": 17132
     },
     {
-      "epoch": 0.4705446156381313,
+      "epoch": 0.4861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.1435153633144773e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.093242688645062e-05,
+      "loss": 0.9316,
       "step": 17133
     },
     {
-      "epoch": 0.47057207986597455,
+      "epoch": 0.4862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.143427331964591e-05,
-      "loss": 0.9162,
+      "learning_rate": 1.0931511808491126e-05,
+      "loss": 0.9188,
       "step": 17134
     },
     {
-      "epoch": 0.4705995440938178,
+      "epoch": 0.4862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.1433392994798519e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.0930596722663109e-05,
+      "loss": 0.9418,
       "step": 17135
     },
     {
-      "epoch": 0.47062700832166104,
+      "epoch": 0.4862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.1432512658609569e-05,
-      "loss": 0.8253,
+      "learning_rate": 1.0929681628974301e-05,
+      "loss": 0.8727,
       "step": 17136
     },
     {
-      "epoch": 0.47065447254950427,
+      "epoch": 0.48629398410896707,
       "grad_norm": 0.0,
-      "learning_rate": 1.1431632311086025e-05,
-      "loss": 0.9826,
+      "learning_rate": 1.092876652743243e-05,
+      "loss": 0.8587,
       "step": 17137
     },
     {
-      "epoch": 0.4706819367773475,
+      "epoch": 0.486322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.1430751952234846e-05,
-      "loss": 0.9345,
+      "learning_rate": 1.0927851418045224e-05,
+      "loss": 0.9843,
       "step": 17138
     },
     {
-      "epoch": 0.47070940100519076,
+      "epoch": 0.48635073779795684,
       "grad_norm": 0.0,
-      "learning_rate": 1.1429871582063007e-05,
-      "loss": 0.9291,
+      "learning_rate": 1.092693630082042e-05,
+      "loss": 0.9775,
       "step": 17139
     },
     {
-      "epoch": 0.470736865233034,
+      "epoch": 0.48637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.142899120057747e-05,
-      "loss": 0.9706,
+      "learning_rate": 1.092602117576574e-05,
+      "loss": 0.8555,
       "step": 17140
     },
     {
-      "epoch": 0.4707643294608772,
+      "epoch": 0.4864074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 1.1428110807785202e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.0925106042888915e-05,
+      "loss": 0.9383,
       "step": 17141
     },
     {
-      "epoch": 0.47079179368872043,
+      "epoch": 0.48643586833144153,
       "grad_norm": 0.0,
-      "learning_rate": 1.1427230403693165e-05,
-      "loss": 0.8708,
+      "learning_rate": 1.092419090219768e-05,
+      "loss": 0.9725,
       "step": 17142
     },
     {
-      "epoch": 0.47081925791656365,
+      "epoch": 0.48646424517593645,
       "grad_norm": 0.0,
-      "learning_rate": 1.1426349988308334e-05,
-      "loss": 0.9318,
+      "learning_rate": 1.0923275753699765e-05,
+      "loss": 0.9478,
       "step": 17143
     },
     {
-      "epoch": 0.4708467221444069,
+      "epoch": 0.4864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.1425469561637663e-05,
-      "loss": 0.9177,
+      "learning_rate": 1.09223605974029e-05,
+      "loss": 0.898,
       "step": 17144
     },
     {
-      "epoch": 0.47087418637225015,
+      "epoch": 0.4865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.142458912368813e-05,
-      "loss": 0.8506,
+      "learning_rate": 1.0921445433314808e-05,
+      "loss": 0.8659,
       "step": 17145
     },
     {
-      "epoch": 0.47090165060009337,
+      "epoch": 0.48654937570942114,
       "grad_norm": 0.0,
-      "learning_rate": 1.1423708674466693e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.0920530261443228e-05,
+      "loss": 0.8704,
       "step": 17146
     },
     {
-      "epoch": 0.4709291148279366,
+      "epoch": 0.486577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.1422828213980321e-05,
-      "loss": 0.9141,
+      "learning_rate": 1.0919615081795887e-05,
+      "loss": 0.909,
       "step": 17147
     },
     {
-      "epoch": 0.47095657905577987,
+      "epoch": 0.4866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.1421947742235983e-05,
-      "loss": 0.9256,
+      "learning_rate": 1.0918699894380515e-05,
+      "loss": 0.8979,
       "step": 17148
     },
     {
-      "epoch": 0.4709840432836231,
+      "epoch": 0.4866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.1421067259240643e-05,
-      "loss": 0.9886,
+      "learning_rate": 1.0917784699204843e-05,
+      "loss": 0.914,
       "step": 17149
     },
     {
-      "epoch": 0.4710115075114663,
+      "epoch": 0.4866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.1420186765001269e-05,
-      "loss": 0.9276,
+      "learning_rate": 1.0916869496276604e-05,
+      "loss": 0.881,
       "step": 17150
     },
     {
-      "epoch": 0.47103897173930953,
+      "epoch": 0.48669125993189555,
       "grad_norm": 0.0,
-      "learning_rate": 1.1419306259524822e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.0915954285603529e-05,
+      "loss": 0.8903,
       "step": 17151
     },
     {
-      "epoch": 0.4710664359671528,
+      "epoch": 0.48671963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.141842574281828e-05,
-      "loss": 0.9064,
+      "learning_rate": 1.0915039067193345e-05,
+      "loss": 0.8595,
       "step": 17152
     },
     {
-      "epoch": 0.47109390019499603,
+      "epoch": 0.4867480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.1417545214888603e-05,
-      "loss": 0.8948,
+      "learning_rate": 1.0914123841053785e-05,
+      "loss": 0.948,
       "step": 17153
     },
     {
-      "epoch": 0.47112136442283925,
+      "epoch": 0.48677639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 1.1416664675742758e-05,
-      "loss": 0.8647,
+      "learning_rate": 1.0913208607192581e-05,
+      "loss": 0.8137,
       "step": 17154
     },
     {
-      "epoch": 0.47114882865068247,
+      "epoch": 0.48680476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 1.1415784125387717e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.0912293365617462e-05,
+      "loss": 0.954,
       "step": 17155
     },
     {
-      "epoch": 0.4711762928785257,
+      "epoch": 0.48683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.1414903563830437e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.0911378116336157e-05,
+      "loss": 0.9306,
       "step": 17156
     },
     {
-      "epoch": 0.47120375710636897,
+      "epoch": 0.4868615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.14140229910779e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.0910462859356404e-05,
+      "loss": 0.9365,
       "step": 17157
     },
     {
-      "epoch": 0.4712312213342122,
+      "epoch": 0.48688989784335984,
       "grad_norm": 0.0,
-      "learning_rate": 1.141314240713706e-05,
-      "loss": 0.9543,
+      "learning_rate": 1.0909547594685928e-05,
+      "loss": 0.8102,
       "step": 17158
     },
     {
-      "epoch": 0.4712586855620554,
+      "epoch": 0.4869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.1412261812014888e-05,
-      "loss": 0.854,
+      "learning_rate": 1.0908632322332464e-05,
+      "loss": 0.9225,
       "step": 17159
     },
     {
-      "epoch": 0.47128614978989863,
+      "epoch": 0.4869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.1411381205718357e-05,
-      "loss": 0.9096,
+      "learning_rate": 1.0907717042303743e-05,
+      "loss": 0.8781,
       "step": 17160
     },
     {
-      "epoch": 0.4713136140177419,
+      "epoch": 0.4869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.1410500588254431e-05,
-      "loss": 0.9327,
+      "learning_rate": 1.0906801754607495e-05,
+      "loss": 0.8237,
       "step": 17161
     },
     {
-      "epoch": 0.47134107824558513,
+      "epoch": 0.4870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.1409619959630076e-05,
-      "loss": 0.8795,
+      "learning_rate": 1.0905886459251449e-05,
+      "loss": 0.8786,
       "step": 17162
     },
     {
-      "epoch": 0.47136854247342835,
+      "epoch": 0.4870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.1408739319852265e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.090497115624334e-05,
+      "loss": 0.8323,
       "step": 17163
     },
     {
-      "epoch": 0.4713960067012716,
+      "epoch": 0.48706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.1407858668927958e-05,
-      "loss": 0.9951,
+      "learning_rate": 1.0904055845590899e-05,
+      "loss": 0.9219,
       "step": 17164
     },
     {
-      "epoch": 0.47142347092911485,
+      "epoch": 0.4870885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 1.140697800686413e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.0903140527301859e-05,
+      "loss": 0.9718,
       "step": 17165
     },
     {
-      "epoch": 0.4714509351569581,
+      "epoch": 0.48711691259931894,
       "grad_norm": 0.0,
-      "learning_rate": 1.1406097333667751e-05,
-      "loss": 0.8607,
+      "learning_rate": 1.0902225201383949e-05,
+      "loss": 0.8392,
       "step": 17166
     },
     {
-      "epoch": 0.4714783993848013,
+      "epoch": 0.48714528944381386,
       "grad_norm": 0.0,
-      "learning_rate": 1.140521664934578e-05,
-      "loss": 0.8657,
+      "learning_rate": 1.0901309867844904e-05,
+      "loss": 0.9228,
       "step": 17167
     },
     {
-      "epoch": 0.4715058636126445,
+      "epoch": 0.4871736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.1404335953905192e-05,
-      "loss": 0.9008,
+      "learning_rate": 1.0900394526692453e-05,
+      "loss": 0.97,
       "step": 17168
     },
     {
-      "epoch": 0.47153332784048774,
+      "epoch": 0.48720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.1403455247352957e-05,
-      "loss": 0.9147,
+      "learning_rate": 1.0899479177934331e-05,
+      "loss": 0.8917,
       "step": 17169
     },
     {
-      "epoch": 0.471560792068331,
+      "epoch": 0.48723041997729855,
       "grad_norm": 0.0,
-      "learning_rate": 1.1402574529696038e-05,
-      "loss": 0.8467,
+      "learning_rate": 1.0898563821578265e-05,
+      "loss": 0.8593,
       "step": 17170
     },
     {
-      "epoch": 0.47158825629617424,
+      "epoch": 0.4872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.1401693800941404e-05,
-      "loss": 0.9292,
+      "learning_rate": 1.0897648457631991e-05,
+      "loss": 0.9474,
       "step": 17171
     },
     {
-      "epoch": 0.47161572052401746,
+      "epoch": 0.4872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.1400813061096027e-05,
-      "loss": 0.8499,
+      "learning_rate": 1.0896733086103239e-05,
+      "loss": 0.9622,
       "step": 17172
     },
     {
-      "epoch": 0.4716431847518607,
+      "epoch": 0.4873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.1399932310166877e-05,
-      "loss": 0.9757,
+      "learning_rate": 1.0895817706999746e-05,
+      "loss": 0.998,
       "step": 17173
     },
     {
-      "epoch": 0.47167064897970395,
+      "epoch": 0.4873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.1399051548160917e-05,
-      "loss": 0.8325,
+      "learning_rate": 1.0894902320329236e-05,
+      "loss": 0.9229,
       "step": 17174
     },
     {
-      "epoch": 0.4716981132075472,
+      "epoch": 0.487372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.1398170775085122e-05,
-      "loss": 1.0323,
+      "learning_rate": 1.0893986926099449e-05,
+      "loss": 0.8943,
       "step": 17175
     },
     {
-      "epoch": 0.4717255774353904,
+      "epoch": 0.48740068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 1.1397289990946456e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.0893071524318112e-05,
+      "loss": 0.89,
       "step": 17176
     },
     {
-      "epoch": 0.4717530416632336,
+      "epoch": 0.4874290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.1396409195751892e-05,
-      "loss": 0.8351,
+      "learning_rate": 1.0892156114992963e-05,
+      "loss": 0.7759,
       "step": 17177
     },
     {
-      "epoch": 0.4717805058910769,
+      "epoch": 0.48745743473325764,
       "grad_norm": 0.0,
-      "learning_rate": 1.1395528389508397e-05,
-      "loss": 1.0558,
+      "learning_rate": 1.0891240698131729e-05,
+      "loss": 0.9278,
       "step": 17178
     },
     {
-      "epoch": 0.4718079701189201,
+      "epoch": 0.48748581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.139464757222294e-05,
-      "loss": 0.9301,
+      "learning_rate": 1.0890325273742141e-05,
+      "loss": 0.9802,
       "step": 17179
     },
     {
-      "epoch": 0.47183543434676334,
+      "epoch": 0.4875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.1393766743902491e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.0889409841831942e-05,
+      "loss": 0.9559,
       "step": 17180
     },
     {
-      "epoch": 0.47186289857460656,
+      "epoch": 0.48754256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.1392885904554021e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.0888494402408856e-05,
+      "loss": 0.9358,
       "step": 17181
     },
     {
-      "epoch": 0.47189036280244984,
+      "epoch": 0.48757094211123725,
       "grad_norm": 0.0,
-      "learning_rate": 1.1392005054184497e-05,
-      "loss": 0.8745,
+      "learning_rate": 1.0887578955480616e-05,
+      "loss": 0.8203,
       "step": 17182
     },
     {
-      "epoch": 0.47191782703029306,
+      "epoch": 0.4875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.1391124192800888e-05,
-      "loss": 0.8771,
+      "learning_rate": 1.088666350105496e-05,
+      "loss": 0.9448,
       "step": 17183
     },
     {
-      "epoch": 0.4719452912581363,
+      "epoch": 0.487627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.1390243320410171e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.0885748039139615e-05,
+      "loss": 1.0283,
       "step": 17184
     },
     {
-      "epoch": 0.4719727554859795,
+      "epoch": 0.4876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.1389362437019306e-05,
-      "loss": 1.0222,
+      "learning_rate": 1.0884832569742316e-05,
+      "loss": 0.8613,
       "step": 17185
     },
     {
-      "epoch": 0.4720002197138227,
+      "epoch": 0.4876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.1388481542635268e-05,
-      "loss": 0.991,
+      "learning_rate": 1.0883917092870797e-05,
+      "loss": 0.8306,
       "step": 17186
     },
     {
-      "epoch": 0.472027683941666,
+      "epoch": 0.4877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.1387600637265025e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.0883001608532791e-05,
+      "loss": 0.926,
       "step": 17187
     },
     {
-      "epoch": 0.4720551481695092,
+      "epoch": 0.4877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.138671972091555e-05,
-      "loss": 0.8012,
+      "learning_rate": 1.088208611673603e-05,
+      "loss": 0.8885,
       "step": 17188
     },
     {
-      "epoch": 0.47208261239735244,
+      "epoch": 0.4877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.138583879359381e-05,
-      "loss": 0.8502,
+      "learning_rate": 1.0881170617488248e-05,
+      "loss": 0.7813,
       "step": 17189
     },
     {
-      "epoch": 0.47211007662519566,
+      "epoch": 0.48779795686719635,
       "grad_norm": 0.0,
-      "learning_rate": 1.138495785530678e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.088025511079718e-05,
+      "loss": 0.9652,
       "step": 17190
     },
     {
-      "epoch": 0.47213754085303894,
+      "epoch": 0.48782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.1384076906061422e-05,
-      "loss": 1.0004,
+      "learning_rate": 1.0879339596670558e-05,
+      "loss": 0.8958,
       "step": 17191
     },
     {
-      "epoch": 0.47216500508088216,
+      "epoch": 0.4878547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 1.1383195945864715e-05,
-      "loss": 0.9475,
+      "learning_rate": 1.0878424075116111e-05,
+      "loss": 0.8642,
       "step": 17192
     },
     {
-      "epoch": 0.4721924693087254,
+      "epoch": 0.48788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.1382314974723628e-05,
-      "loss": 0.9476,
+      "learning_rate": 1.087750854614158e-05,
+      "loss": 0.8,
       "step": 17193
     },
     {
-      "epoch": 0.4722199335365686,
+      "epoch": 0.48791146424517595,
       "grad_norm": 0.0,
-      "learning_rate": 1.1381433992645125e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.0876593009754692e-05,
+      "loss": 0.9307,
       "step": 17194
     },
     {
-      "epoch": 0.4722473977644119,
+      "epoch": 0.4879398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.1380552999636181e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.0875677465963186e-05,
+      "loss": 0.9391,
       "step": 17195
     },
     {
-      "epoch": 0.4722748619922551,
+      "epoch": 0.4879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.137967199570377e-05,
-      "loss": 0.8412,
+      "learning_rate": 1.087476191477479e-05,
+      "loss": 0.9042,
       "step": 17196
     },
     {
-      "epoch": 0.4723023262200983,
+      "epoch": 0.4879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.1378790980854859e-05,
-      "loss": 0.9131,
+      "learning_rate": 1.0873846356197244e-05,
+      "loss": 0.8843,
       "step": 17197
     },
     {
-      "epoch": 0.47232979044794154,
+      "epoch": 0.4880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.1377909955096417e-05,
-      "loss": 0.8312,
+      "learning_rate": 1.0872930790238278e-05,
+      "loss": 0.8851,
       "step": 17198
     },
     {
-      "epoch": 0.47235725467578477,
+      "epoch": 0.4880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.137702891843542e-05,
-      "loss": 1.0102,
+      "learning_rate": 1.0872015216905625e-05,
+      "loss": 0.8638,
       "step": 17199
     },
     {
-      "epoch": 0.47238471890362804,
+      "epoch": 0.4880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.1376147870878835e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.0871099636207022e-05,
+      "loss": 0.9952,
       "step": 17200
     },
     {
-      "epoch": 0.47241218313147126,
+      "epoch": 0.4881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.1375266812433638e-05,
-      "loss": 1.0112,
+      "learning_rate": 1.08701840481502e-05,
+      "loss": 0.8006,
       "step": 17201
     },
     {
-      "epoch": 0.4724396473593145,
+      "epoch": 0.48813847900113505,
       "grad_norm": 0.0,
-      "learning_rate": 1.1374385743106797e-05,
-      "loss": 0.9063,
+      "learning_rate": 1.0869268452742893e-05,
+      "loss": 0.903,
       "step": 17202
     },
     {
-      "epoch": 0.4724671115871577,
+      "epoch": 0.48816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.137350466290528e-05,
-      "loss": 0.9089,
+      "learning_rate": 1.0868352849992836e-05,
+      "loss": 0.8574,
       "step": 17203
     },
     {
-      "epoch": 0.472494575815001,
+      "epoch": 0.4881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.1372623571836066e-05,
-      "loss": 0.9333,
+      "learning_rate": 1.0867437239907766e-05,
+      "loss": 0.8523,
       "step": 17204
     },
     {
-      "epoch": 0.4725220400428442,
+      "epoch": 0.48822360953461974,
       "grad_norm": 0.0,
-      "learning_rate": 1.1371742469906122e-05,
-      "loss": 1.0145,
+      "learning_rate": 1.0866521622495417e-05,
+      "loss": 0.8983,
       "step": 17205
     },
     {
-      "epoch": 0.4725495042706874,
+      "epoch": 0.48825198637911466,
       "grad_norm": 0.0,
-      "learning_rate": 1.1370861357122416e-05,
-      "loss": 0.7898,
+      "learning_rate": 1.0865605997763516e-05,
+      "loss": 0.7896,
       "step": 17206
     },
     {
-      "epoch": 0.47257696849853065,
+      "epoch": 0.4882803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.1369980233491926e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.0864690365719803e-05,
+      "loss": 0.9352,
       "step": 17207
     },
     {
-      "epoch": 0.4726044327263739,
+      "epoch": 0.48830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.1369099099021623e-05,
-      "loss": 0.9199,
+      "learning_rate": 1.0863774726372014e-05,
+      "loss": 0.8347,
       "step": 17208
     },
     {
-      "epoch": 0.47263189695421715,
+      "epoch": 0.48833711691259934,
       "grad_norm": 0.0,
-      "learning_rate": 1.1368217953718479e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.086285907972788e-05,
+      "loss": 0.8837,
       "step": 17209
     },
     {
-      "epoch": 0.47265936118206037,
+      "epoch": 0.4883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.1367336797589461e-05,
-      "loss": 0.8171,
+      "learning_rate": 1.0861943425795132e-05,
+      "loss": 0.8837,
       "step": 17210
     },
     {
-      "epoch": 0.4726868254099036,
+      "epoch": 0.4883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.136645563064155e-05,
-      "loss": 0.9895,
+      "learning_rate": 1.0861027764581515e-05,
+      "loss": 0.8597,
       "step": 17211
     },
     {
-      "epoch": 0.4727142896377468,
+      "epoch": 0.488422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1365574452881707e-05,
-      "loss": 0.9207,
+      "learning_rate": 1.0860112096094755e-05,
+      "loss": 0.8438,
       "step": 17212
     },
     {
-      "epoch": 0.4727417538655901,
+      "epoch": 0.4884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.1364693264316913e-05,
-      "loss": 0.9304,
+      "learning_rate": 1.085919642034259e-05,
+      "loss": 0.9494,
       "step": 17213
     },
     {
-      "epoch": 0.4727692180934333,
+      "epoch": 0.48847900113507375,
       "grad_norm": 0.0,
-      "learning_rate": 1.1363812064954136e-05,
-      "loss": 0.9182,
+      "learning_rate": 1.0858280737332756e-05,
+      "loss": 0.859,
       "step": 17214
     },
     {
-      "epoch": 0.47279668232127653,
+      "epoch": 0.48850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.1362930854800349e-05,
-      "loss": 0.8755,
+      "learning_rate": 1.0857365047072985e-05,
+      "loss": 0.9362,
       "step": 17215
     },
     {
-      "epoch": 0.47282414654911975,
+      "epoch": 0.4885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.1362049633862528e-05,
-      "loss": 0.9118,
+      "learning_rate": 1.0856449349571013e-05,
+      "loss": 0.8409,
       "step": 17216
     },
     {
-      "epoch": 0.472851610776963,
+      "epoch": 0.48856413166855844,
       "grad_norm": 0.0,
-      "learning_rate": 1.1361168402147641e-05,
-      "loss": 0.8797,
+      "learning_rate": 1.0855533644834573e-05,
+      "loss": 0.8876,
       "step": 17217
     },
     {
-      "epoch": 0.47287907500480625,
+      "epoch": 0.48859250851305336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1360287159662661e-05,
-      "loss": 0.8689,
+      "learning_rate": 1.0854617932871403e-05,
+      "loss": 0.8557,
       "step": 17218
     },
     {
-      "epoch": 0.47290653923264947,
+      "epoch": 0.4886208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.1359405906414565e-05,
-      "loss": 0.9017,
+      "learning_rate": 1.0853702213689236e-05,
+      "loss": 0.9505,
       "step": 17219
     },
     {
-      "epoch": 0.4729340034604927,
+      "epoch": 0.48864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.1358524642410318e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.085278648729581e-05,
+      "loss": 0.8338,
       "step": 17220
     },
     {
-      "epoch": 0.47296146768833597,
+      "epoch": 0.48867763904653805,
       "grad_norm": 0.0,
-      "learning_rate": 1.1357643367656903e-05,
-      "loss": 0.8004,
+      "learning_rate": 1.085187075369886e-05,
+      "loss": 0.9016,
       "step": 17221
     },
     {
-      "epoch": 0.4729889319161792,
+      "epoch": 0.4887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.1356762082161285e-05,
-      "loss": 0.968,
+      "learning_rate": 1.0850955012906114e-05,
+      "loss": 0.8855,
       "step": 17222
     },
     {
-      "epoch": 0.4730163961440224,
+      "epoch": 0.4887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.135588078593044e-05,
-      "loss": 0.9132,
+      "learning_rate": 1.0850039264925315e-05,
+      "loss": 0.8103,
       "step": 17223
     },
     {
-      "epoch": 0.47304386037186563,
+      "epoch": 0.4887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.135499947897134e-05,
-      "loss": 0.884,
+      "learning_rate": 1.08491235097642e-05,
+      "loss": 0.8695,
       "step": 17224
     },
     {
-      "epoch": 0.47307132459970885,
+      "epoch": 0.4887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.1354118161290966e-05,
-      "loss": 0.9217,
+      "learning_rate": 1.0848207747430495e-05,
+      "loss": 1.0085,
       "step": 17225
     },
     {
-      "epoch": 0.47309878882755213,
+      "epoch": 0.4888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.1353236832896278e-05,
-      "loss": 0.917,
+      "learning_rate": 1.0847291977931942e-05,
+      "loss": 0.8201,
       "step": 17226
     },
     {
-      "epoch": 0.47312625305539535,
+      "epoch": 0.48884790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 1.1352355493794255e-05,
-      "loss": 0.9034,
+      "learning_rate": 1.0846376201276278e-05,
+      "loss": 0.9501,
       "step": 17227
     },
     {
-      "epoch": 0.4731537172832386,
+      "epoch": 0.4888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.1351474143991874e-05,
-      "loss": 0.7986,
+      "learning_rate": 1.0845460417471237e-05,
+      "loss": 0.8907,
       "step": 17228
     },
     {
-      "epoch": 0.4731811815110818,
+      "epoch": 0.48890465380249715,
       "grad_norm": 0.0,
-      "learning_rate": 1.1350592783496109e-05,
-      "loss": 0.9247,
+      "learning_rate": 1.0844544626524551e-05,
+      "loss": 0.8944,
       "step": 17229
     },
     {
-      "epoch": 0.47320864573892507,
+      "epoch": 0.48893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.1349711412313923e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.084362882844396e-05,
+      "loss": 0.9859,
       "step": 17230
     },
     {
-      "epoch": 0.4732361099667683,
+      "epoch": 0.4889614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 1.1348830030452304e-05,
-      "loss": 0.8576,
+      "learning_rate": 1.08427130232372e-05,
+      "loss": 0.9372,
       "step": 17231
     },
     {
-      "epoch": 0.4732635741946115,
+      "epoch": 0.48898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.1347948637918215e-05,
-      "loss": 0.8906,
+      "learning_rate": 1.0841797210912e-05,
+      "loss": 0.7896,
       "step": 17232
     },
     {
-      "epoch": 0.47329103842245474,
+      "epoch": 0.48901816118047675,
       "grad_norm": 0.0,
-      "learning_rate": 1.1347067234718639e-05,
-      "loss": 0.8351,
+      "learning_rate": 1.0840881391476103e-05,
+      "loss": 0.9931,
       "step": 17233
     },
     {
-      "epoch": 0.473318502650298,
+      "epoch": 0.4890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.1346185820860545e-05,
-      "loss": 0.8268,
+      "learning_rate": 1.0839965564937244e-05,
+      "loss": 0.8411,
       "step": 17234
     },
     {
-      "epoch": 0.47334596687814123,
+      "epoch": 0.4890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.1345304396350903e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.0839049731303159e-05,
+      "loss": 0.8482,
       "step": 17235
     },
     {
-      "epoch": 0.47337343110598445,
+      "epoch": 0.4891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.1344422961196693e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.0838133890581583e-05,
+      "loss": 0.8473,
       "step": 17236
     },
     {
-      "epoch": 0.4734008953338277,
+      "epoch": 0.4891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.1343541515404893e-05,
-      "loss": 0.8983,
+      "learning_rate": 1.0837218042780254e-05,
+      "loss": 0.8761,
       "step": 17237
     },
     {
-      "epoch": 0.4734283595616709,
+      "epoch": 0.4891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.1342660058982465e-05,
-      "loss": 0.8364,
+      "learning_rate": 1.0836302187906904e-05,
+      "loss": 0.918,
       "step": 17238
     },
     {
-      "epoch": 0.4734558237895142,
+      "epoch": 0.4891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.1341778591936392e-05,
-      "loss": 0.9553,
+      "learning_rate": 1.083538632596927e-05,
+      "loss": 0.8867,
       "step": 17239
     },
     {
-      "epoch": 0.4734832880173574,
+      "epoch": 0.489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.1340897114273647e-05,
-      "loss": 0.7573,
+      "learning_rate": 1.0834470456975092e-05,
+      "loss": 0.8987,
       "step": 17240
     },
     {
-      "epoch": 0.4735107522452006,
+      "epoch": 0.48924517593643585,
       "grad_norm": 0.0,
-      "learning_rate": 1.1340015626001208e-05,
-      "loss": 0.8762,
+      "learning_rate": 1.0833554580932107e-05,
+      "loss": 0.7417,
       "step": 17241
     },
     {
-      "epoch": 0.47353821647304384,
+      "epoch": 0.48927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.1339134127126043e-05,
-      "loss": 0.8845,
+      "learning_rate": 1.0832638697848047e-05,
+      "loss": 0.8156,
       "step": 17242
     },
     {
-      "epoch": 0.4735656807008871,
+      "epoch": 0.4893019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 1.1338252617655132e-05,
-      "loss": 0.78,
+      "learning_rate": 1.083172280773065e-05,
+      "loss": 0.8847,
       "step": 17243
     },
     {
-      "epoch": 0.47359314492873034,
+      "epoch": 0.48933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.1337371097595444e-05,
-      "loss": 0.9848,
+      "learning_rate": 1.0830806910587655e-05,
+      "loss": 0.7776,
       "step": 17244
     },
     {
-      "epoch": 0.47362060915657356,
+      "epoch": 0.48935868331441545,
       "grad_norm": 0.0,
-      "learning_rate": 1.1336489566953964e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.0829891006426796e-05,
+      "loss": 0.9555,
       "step": 17245
     },
     {
-      "epoch": 0.4736480733844168,
+      "epoch": 0.4893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.1335608025737657e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.0828975095255806e-05,
+      "loss": 0.8976,
       "step": 17246
     },
     {
-      "epoch": 0.47367553761226006,
+      "epoch": 0.48941543700340523,
       "grad_norm": 0.0,
-      "learning_rate": 1.13347264739535e-05,
-      "loss": 0.975,
+      "learning_rate": 1.0828059177082433e-05,
+      "loss": 0.9043,
       "step": 17247
     },
     {
-      "epoch": 0.4737030018401033,
+      "epoch": 0.4894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.1333844911608474e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.0827143251914401e-05,
+      "loss": 0.7864,
       "step": 17248
     },
     {
-      "epoch": 0.4737304660679465,
+      "epoch": 0.489472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.1332963338709551e-05,
-      "loss": 0.965,
+      "learning_rate": 1.0826227319759458e-05,
+      "loss": 0.9776,
       "step": 17249
     },
     {
-      "epoch": 0.4737579302957897,
+      "epoch": 0.4895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.1332081755263701e-05,
-      "loss": 0.8792,
+      "learning_rate": 1.082531138062533e-05,
+      "loss": 1.0013,
       "step": 17250
     },
     {
-      "epoch": 0.47378539452363294,
+      "epoch": 0.4895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.1331200161277907e-05,
-      "loss": 0.8892,
+      "learning_rate": 1.0824395434519763e-05,
+      "loss": 0.955,
       "step": 17251
     },
     {
-      "epoch": 0.4738128587514762,
+      "epoch": 0.4895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.1330318556759144e-05,
-      "loss": 0.906,
+      "learning_rate": 1.0823479481450491e-05,
+      "loss": 0.9178,
       "step": 17252
     },
     {
-      "epoch": 0.47384032297931944,
+      "epoch": 0.48958569807037455,
       "grad_norm": 0.0,
-      "learning_rate": 1.1329436941714382e-05,
-      "loss": 0.7922,
+      "learning_rate": 1.0822563521425248e-05,
+      "loss": 0.8725,
       "step": 17253
     },
     {
-      "epoch": 0.47386778720716266,
+      "epoch": 0.48961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.13285553161506e-05,
-      "loss": 0.8605,
+      "learning_rate": 1.0821647554451779e-05,
+      "loss": 0.9711,
       "step": 17254
     },
     {
-      "epoch": 0.4738952514350059,
+      "epoch": 0.4896424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.1327673680074775e-05,
-      "loss": 0.9806,
+      "learning_rate": 1.0820731580537813e-05,
+      "loss": 0.8354,
       "step": 17255
     },
     {
-      "epoch": 0.47392271566284916,
+      "epoch": 0.48967082860385924,
       "grad_norm": 0.0,
-      "learning_rate": 1.1326792033493881e-05,
-      "loss": 1.0204,
+      "learning_rate": 1.081981559969109e-05,
+      "loss": 0.8271,
       "step": 17256
     },
     {
-      "epoch": 0.4739501798906924,
+      "epoch": 0.48969920544835416,
       "grad_norm": 0.0,
-      "learning_rate": 1.1325910376414894e-05,
-      "loss": 0.9185,
+      "learning_rate": 1.0818899611919349e-05,
+      "loss": 0.9211,
       "step": 17257
     },
     {
-      "epoch": 0.4739776441185356,
+      "epoch": 0.489727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.1325028708844791e-05,
-      "loss": 0.8935,
+      "learning_rate": 1.0817983617230326e-05,
+      "loss": 0.9062,
       "step": 17258
     },
     {
-      "epoch": 0.4740051083463788,
+      "epoch": 0.48975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.1324147030790548e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.0817067615631757e-05,
+      "loss": 0.8886,
       "step": 17259
     },
     {
-      "epoch": 0.4740325725742221,
+      "epoch": 0.4897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.1323265342259139e-05,
-      "loss": 0.8174,
+      "learning_rate": 1.0816151607131383e-05,
+      "loss": 0.8728,
       "step": 17260
     },
     {
-      "epoch": 0.4740600368020653,
+      "epoch": 0.4898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.1322383643257546e-05,
-      "loss": 0.9135,
+      "learning_rate": 1.081523559173694e-05,
+      "loss": 0.7822,
       "step": 17261
     },
     {
-      "epoch": 0.47408750102990854,
+      "epoch": 0.4898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.1321501933792737e-05,
-      "loss": 0.9143,
+      "learning_rate": 1.0814319569456164e-05,
+      "loss": 0.8638,
       "step": 17262
     },
     {
-      "epoch": 0.47411496525775176,
+      "epoch": 0.4898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.1320620213871693e-05,
-      "loss": 0.903,
+      "learning_rate": 1.0813403540296791e-05,
+      "loss": 1.0012,
       "step": 17263
     },
     {
-      "epoch": 0.474142429485595,
+      "epoch": 0.4898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.1319738483501395e-05,
-      "loss": 0.9355,
+      "learning_rate": 1.0812487504266567e-05,
+      "loss": 0.8905,
       "step": 17264
     },
     {
-      "epoch": 0.47416989371343826,
+      "epoch": 0.48992622020431326,
       "grad_norm": 0.0,
-      "learning_rate": 1.1318856742688812e-05,
-      "loss": 0.7848,
+      "learning_rate": 1.0811571461373224e-05,
+      "loss": 0.9338,
       "step": 17265
     },
     {
-      "epoch": 0.4741973579412815,
+      "epoch": 0.48995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.131797499144092e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.0810655411624497e-05,
+      "loss": 0.8872,
       "step": 17266
     },
     {
-      "epoch": 0.4742248221691247,
+      "epoch": 0.4899829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.1317093229764707e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.080973935502813e-05,
+      "loss": 0.9555,
       "step": 17267
     },
     {
-      "epoch": 0.4742522863969679,
+      "epoch": 0.49001135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.1316211457667135e-05,
-      "loss": 0.8547,
+      "learning_rate": 1.080882329159186e-05,
+      "loss": 0.8063,
       "step": 17268
     },
     {
-      "epoch": 0.4742797506248112,
+      "epoch": 0.49003972758229286,
       "grad_norm": 0.0,
-      "learning_rate": 1.1315329675155191e-05,
-      "loss": 0.9908,
+      "learning_rate": 1.0807907221323422e-05,
+      "loss": 0.974,
       "step": 17269
     },
     {
-      "epoch": 0.4743072148526544,
+      "epoch": 0.4900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.131444788223585e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.0806991144230551e-05,
+      "loss": 0.8292,
       "step": 17270
     },
     {
-      "epoch": 0.47433467908049765,
+      "epoch": 0.49009648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.1313566078916088e-05,
-      "loss": 0.8513,
+      "learning_rate": 1.0806075060320994e-05,
+      "loss": 0.8546,
       "step": 17271
     },
     {
-      "epoch": 0.47436214330834087,
+      "epoch": 0.49012485811577755,
       "grad_norm": 0.0,
-      "learning_rate": 1.131268426520288e-05,
-      "loss": 0.9398,
+      "learning_rate": 1.0805158969602484e-05,
+      "loss": 0.9237,
       "step": 17272
     },
     {
-      "epoch": 0.47438960753618414,
+      "epoch": 0.4901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.131180244110321e-05,
-      "loss": 0.8404,
+      "learning_rate": 1.0804242872082758e-05,
+      "loss": 0.8728,
       "step": 17273
     },
     {
-      "epoch": 0.47441707176402736,
+      "epoch": 0.4901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.1310920606624048e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.0803326767769559e-05,
+      "loss": 0.9263,
       "step": 17274
     },
     {
-      "epoch": 0.4744445359918706,
+      "epoch": 0.4902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.131003876177237e-05,
-      "loss": 0.955,
+      "learning_rate": 1.0802410656670623e-05,
+      "loss": 0.8932,
       "step": 17275
     },
     {
-      "epoch": 0.4744720002197138,
+      "epoch": 0.4902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.1309156906555165e-05,
-      "loss": 0.8152,
+      "learning_rate": 1.0801494538793684e-05,
+      "loss": 0.9898,
       "step": 17276
     },
     {
-      "epoch": 0.4744994644475571,
+      "epoch": 0.49026674233825196,
       "grad_norm": 0.0,
-      "learning_rate": 1.1308275040979401e-05,
-      "loss": 0.8545,
+      "learning_rate": 1.080057841414649e-05,
+      "loss": 0.8686,
       "step": 17277
     },
     {
-      "epoch": 0.4745269286754003,
+      "epoch": 0.4902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.1307393165052058e-05,
-      "loss": 0.8499,
+      "learning_rate": 1.079966228273677e-05,
+      "loss": 0.8631,
       "step": 17278
     },
     {
-      "epoch": 0.4745543929032435,
+      "epoch": 0.4903234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.1306511278780115e-05,
-      "loss": 0.9858,
+      "learning_rate": 1.0798746144572267e-05,
+      "loss": 0.8683,
       "step": 17279
     },
     {
-      "epoch": 0.47458185713108675,
+      "epoch": 0.49035187287173665,
       "grad_norm": 0.0,
-      "learning_rate": 1.1305629382170544e-05,
-      "loss": 0.9242,
+      "learning_rate": 1.079782999966072e-05,
+      "loss": 0.8879,
       "step": 17280
     },
     {
-      "epoch": 0.47460932135892997,
+      "epoch": 0.49038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.1304747475230336e-05,
-      "loss": 0.8749,
+      "learning_rate": 1.079691384800987e-05,
+      "loss": 0.9865,
       "step": 17281
     },
     {
-      "epoch": 0.47463678558677325,
+      "epoch": 0.4904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 1.1303865557966453e-05,
-      "loss": 0.9567,
+      "learning_rate": 1.079599768962745e-05,
+      "loss": 0.9196,
       "step": 17282
     },
     {
-      "epoch": 0.47466424981461647,
+      "epoch": 0.49043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.1302983630385882e-05,
-      "loss": 0.8926,
+      "learning_rate": 1.0795081524521201e-05,
+      "loss": 0.8571,
       "step": 17283
     },
     {
-      "epoch": 0.4746917140424597,
+      "epoch": 0.49046538024971625,
       "grad_norm": 0.0,
-      "learning_rate": 1.1302101692495602e-05,
-      "loss": 0.9336,
+      "learning_rate": 1.0794165352698864e-05,
+      "loss": 0.8496,
       "step": 17284
     },
     {
-      "epoch": 0.4747191782703029,
+      "epoch": 0.4904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.1301219744302589e-05,
-      "loss": 0.9294,
+      "learning_rate": 1.079324917416818e-05,
+      "loss": 0.8988,
       "step": 17285
     },
     {
-      "epoch": 0.4747466424981462,
+      "epoch": 0.49052213393870603,
       "grad_norm": 0.0,
-      "learning_rate": 1.130033778581382e-05,
-      "loss": 0.8731,
+      "learning_rate": 1.0792332988936878e-05,
+      "loss": 0.8727,
       "step": 17286
     },
     {
-      "epoch": 0.4747741067259894,
+      "epoch": 0.4905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.1299455817036276e-05,
-      "loss": 0.957,
+      "learning_rate": 1.0791416797012706e-05,
+      "loss": 0.8371,
       "step": 17287
     },
     {
-      "epoch": 0.47480157095383263,
+      "epoch": 0.4905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.1298573837976932e-05,
-      "loss": 0.9157,
+      "learning_rate": 1.0790500598403401e-05,
+      "loss": 0.8565,
       "step": 17288
     },
     {
-      "epoch": 0.47482903518167585,
+      "epoch": 0.4906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.129769184864277e-05,
-      "loss": 0.8547,
+      "learning_rate": 1.0789584393116706e-05,
+      "loss": 0.943,
       "step": 17289
     },
     {
-      "epoch": 0.47485649940951913,
+      "epoch": 0.4906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.1296809849040769e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.078866818116035e-05,
+      "loss": 0.8298,
       "step": 17290
     },
     {
-      "epoch": 0.47488396363736235,
+      "epoch": 0.4906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.1295927839177903e-05,
-      "loss": 0.9414,
+      "learning_rate": 1.0787751962542081e-05,
+      "loss": 1.0371,
       "step": 17291
     },
     {
-      "epoch": 0.47491142786520557,
+      "epoch": 0.49069239500567535,
       "grad_norm": 0.0,
-      "learning_rate": 1.1295045819061154e-05,
-      "loss": 0.8425,
+      "learning_rate": 1.0786835737269636e-05,
+      "loss": 0.9633,
       "step": 17292
     },
     {
-      "epoch": 0.4749388920930488,
+      "epoch": 0.49072077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.1294163788697505e-05,
-      "loss": 0.941,
+      "learning_rate": 1.0785919505350754e-05,
+      "loss": 0.7372,
       "step": 17293
     },
     {
-      "epoch": 0.474966356320892,
+      "epoch": 0.4907491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.1293281748093928e-05,
-      "loss": 0.8978,
+      "learning_rate": 1.0785003266793173e-05,
+      "loss": 0.8405,
       "step": 17294
     },
     {
-      "epoch": 0.4749938205487353,
+      "epoch": 0.49077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.1292399697257403e-05,
-      "loss": 0.9695,
+      "learning_rate": 1.0784087021604636e-05,
+      "loss": 0.8562,
       "step": 17295
     },
     {
-      "epoch": 0.4750212847765785,
+      "epoch": 0.49080590238365496,
       "grad_norm": 0.0,
-      "learning_rate": 1.1291517636194915e-05,
-      "loss": 0.7721,
+      "learning_rate": 1.0783170769792882e-05,
+      "loss": 0.9076,
       "step": 17296
     },
     {
-      "epoch": 0.47504874900442173,
+      "epoch": 0.4908342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.1290635564913438e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.0782254511365646e-05,
+      "loss": 0.9156,
       "step": 17297
     },
     {
-      "epoch": 0.47507621323226495,
+      "epoch": 0.49086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.128975348341995e-05,
-      "loss": 1.0773,
+      "learning_rate": 1.0781338246330673e-05,
+      "loss": 0.8283,
       "step": 17298
     },
     {
-      "epoch": 0.47510367746010823,
+      "epoch": 0.4908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.1288871391721434e-05,
-      "loss": 0.9668,
+      "learning_rate": 1.07804219746957e-05,
+      "loss": 0.8871,
       "step": 17299
     },
     {
-      "epoch": 0.47513114168795145,
+      "epoch": 0.4909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.1287989289824868e-05,
-      "loss": 0.8264,
+      "learning_rate": 1.0779505696468468e-05,
+      "loss": 0.8069,
       "step": 17300
     },
     {
-      "epoch": 0.4751586059157947,
+      "epoch": 0.4909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.1287107177737233e-05,
-      "loss": 0.8448,
+      "learning_rate": 1.0778589411656717e-05,
+      "loss": 0.877,
       "step": 17301
     },
     {
-      "epoch": 0.4751860701436379,
+      "epoch": 0.4909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.128622505546551e-05,
-      "loss": 0.8429,
+      "learning_rate": 1.0777673120268183e-05,
+      "loss": 0.8973,
       "step": 17302
     },
     {
-      "epoch": 0.47521353437148117,
+      "epoch": 0.4910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.1285342923016671e-05,
-      "loss": 0.927,
+      "learning_rate": 1.0776756822310615e-05,
+      "loss": 1.0618,
       "step": 17303
     },
     {
-      "epoch": 0.4752409985993244,
+      "epoch": 0.49103291713961406,
       "grad_norm": 0.0,
-      "learning_rate": 1.1284460780397701e-05,
-      "loss": 0.9998,
+      "learning_rate": 1.0775840517791741e-05,
+      "loss": 0.8569,
       "step": 17304
     },
     {
-      "epoch": 0.4752684628271676,
+      "epoch": 0.49106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.1283578627615586e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.077492420671931e-05,
+      "loss": 0.8656,
       "step": 17305
     },
     {
-      "epoch": 0.47529592705501084,
+      "epoch": 0.4910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.1282696464677295e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.0774007889101062e-05,
+      "loss": 0.92,
       "step": 17306
     },
     {
-      "epoch": 0.47532339128285406,
+      "epoch": 0.49111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.128181429158981e-05,
-      "loss": 0.8651,
+      "learning_rate": 1.077309156494473e-05,
+      "loss": 0.8425,
       "step": 17307
     },
     {
-      "epoch": 0.47535085551069733,
+      "epoch": 0.49114642451759366,
       "grad_norm": 0.0,
-      "learning_rate": 1.1280932108360118e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.0772175234258063e-05,
+      "loss": 0.9011,
       "step": 17308
     },
     {
-      "epoch": 0.47537831973854056,
+      "epoch": 0.4911748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.1280049914995194e-05,
-      "loss": 0.9726,
+      "learning_rate": 1.0771258897048796e-05,
+      "loss": 0.9124,
       "step": 17309
     },
     {
-      "epoch": 0.4754057839663838,
+      "epoch": 0.49120317820658344,
       "grad_norm": 0.0,
-      "learning_rate": 1.127916771150202e-05,
-      "loss": 0.9463,
+      "learning_rate": 1.0770342553324667e-05,
+      "loss": 1.0024,
       "step": 17310
     },
     {
-      "epoch": 0.475433248194227,
+      "epoch": 0.4912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.1278285497887575e-05,
-      "loss": 0.8485,
+      "learning_rate": 1.076942620309342e-05,
+      "loss": 0.9258,
       "step": 17311
     },
     {
-      "epoch": 0.4754607124220703,
+      "epoch": 0.4912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.1277403274158839e-05,
-      "loss": 0.901,
+      "learning_rate": 1.0768509846362798e-05,
+      "loss": 0.8387,
       "step": 17312
     },
     {
-      "epoch": 0.4754881766499135,
+      "epoch": 0.4912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.1276521040322794e-05,
-      "loss": 0.8946,
+      "learning_rate": 1.076759348314054e-05,
+      "loss": 0.8803,
       "step": 17313
     },
     {
-      "epoch": 0.4755156408777567,
+      "epoch": 0.491316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.1275638796386421e-05,
-      "loss": 0.8258,
+      "learning_rate": 1.0766677113434381e-05,
+      "loss": 0.8415,
       "step": 17314
     },
     {
-      "epoch": 0.47554310510559994,
+      "epoch": 0.4913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.1274756542356698e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.0765760737252072e-05,
+      "loss": 0.9374,
       "step": 17315
     },
     {
-      "epoch": 0.4755705693334432,
+      "epoch": 0.49137343927355276,
       "grad_norm": 0.0,
-      "learning_rate": 1.1273874278240607e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.0764844354601342e-05,
+      "loss": 0.7569,
       "step": 17316
     },
     {
-      "epoch": 0.47559803356128644,
+      "epoch": 0.4914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.1272992004045133e-05,
-      "loss": 0.9753,
+      "learning_rate": 1.0763927965489938e-05,
+      "loss": 0.8461,
       "step": 17317
     },
     {
-      "epoch": 0.47562549778912966,
+      "epoch": 0.4914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.1272109719777251e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.07630115699256e-05,
+      "loss": 0.87,
       "step": 17318
     },
     {
-      "epoch": 0.4756529620169729,
+      "epoch": 0.49145856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.1271227425443943e-05,
-      "loss": 0.9373,
+      "learning_rate": 1.0762095167916074e-05,
+      "loss": 0.8844,
       "step": 17319
     },
     {
-      "epoch": 0.4756804262448161,
+      "epoch": 0.49148694665153236,
       "grad_norm": 0.0,
-      "learning_rate": 1.1270345121052193e-05,
-      "loss": 1.0045,
+      "learning_rate": 1.0761178759469091e-05,
+      "loss": 0.8719,
       "step": 17320
     },
     {
-      "epoch": 0.4757078904726594,
+      "epoch": 0.4915153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.126946280660898e-05,
-      "loss": 0.9545,
+      "learning_rate": 1.0760262344592399e-05,
+      "loss": 0.8096,
       "step": 17321
     },
     {
-      "epoch": 0.4757353547005026,
+      "epoch": 0.49154370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.1268580482121283e-05,
-      "loss": 0.8763,
+      "learning_rate": 1.0759345923293736e-05,
+      "loss": 0.9627,
       "step": 17322
     },
     {
-      "epoch": 0.4757628189283458,
+      "epoch": 0.49157207718501705,
       "grad_norm": 0.0,
-      "learning_rate": 1.1267698147596089e-05,
-      "loss": 0.8892,
+      "learning_rate": 1.0758429495580844e-05,
+      "loss": 0.9264,
       "step": 17323
     },
     {
-      "epoch": 0.47579028315618904,
+      "epoch": 0.4916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.1266815803040375e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.0757513061461462e-05,
+      "loss": 0.8764,
       "step": 17324
     },
     {
-      "epoch": 0.4758177473840323,
+      "epoch": 0.49162883087400683,
       "grad_norm": 0.0,
-      "learning_rate": 1.1265933448461123e-05,
-      "loss": 0.894,
+      "learning_rate": 1.0756596620943335e-05,
+      "loss": 0.8598,
       "step": 17325
     },
     {
-      "epoch": 0.47584521161187554,
+      "epoch": 0.4916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.1265051083865317e-05,
-      "loss": 0.912,
+      "learning_rate": 1.0755680174034203e-05,
+      "loss": 0.8957,
       "step": 17326
     },
     {
-      "epoch": 0.47587267583971876,
+      "epoch": 0.4916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.1264168709259935e-05,
-      "loss": 0.8933,
+      "learning_rate": 1.0754763720741805e-05,
+      "loss": 0.9285,
       "step": 17327
     },
     {
-      "epoch": 0.475900140067562,
+      "epoch": 0.49171396140749146,
       "grad_norm": 0.0,
-      "learning_rate": 1.1263286324651958e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.0753847261073885e-05,
+      "loss": 0.9406,
       "step": 17328
     },
     {
-      "epoch": 0.47592760429540526,
+      "epoch": 0.4917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.1262403930048377e-05,
-      "loss": 0.8408,
+      "learning_rate": 1.0752930795038183e-05,
+      "loss": 0.8809,
       "step": 17329
     },
     {
-      "epoch": 0.4759550685232485,
+      "epoch": 0.4917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.1261521525456163e-05,
-      "loss": 1.0057,
+      "learning_rate": 1.075201432264244e-05,
+      "loss": 0.8947,
       "step": 17330
     },
     {
-      "epoch": 0.4759825327510917,
+      "epoch": 0.49179909194097615,
       "grad_norm": 0.0,
-      "learning_rate": 1.1260639110882302e-05,
-      "loss": 0.8922,
+      "learning_rate": 1.07510978438944e-05,
+      "loss": 0.9396,
       "step": 17331
     },
     {
-      "epoch": 0.4760099969789349,
+      "epoch": 0.49182746878547107,
       "grad_norm": 0.0,
-      "learning_rate": 1.1259756686333777e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.0750181358801803e-05,
+      "loss": 0.9315,
       "step": 17332
     },
     {
-      "epoch": 0.47603746120677815,
+      "epoch": 0.4918558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.1258874251817568e-05,
-      "loss": 0.9077,
+      "learning_rate": 1.0749264867372389e-05,
+      "loss": 0.9129,
       "step": 17333
     },
     {
-      "epoch": 0.4760649254346214,
+      "epoch": 0.49188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1257991807340658e-05,
-      "loss": 0.8347,
+      "learning_rate": 1.0748348369613899e-05,
+      "loss": 0.8378,
       "step": 17334
     },
     {
-      "epoch": 0.47609238966246464,
+      "epoch": 0.49191259931895576,
       "grad_norm": 0.0,
-      "learning_rate": 1.1257109352910032e-05,
-      "loss": 0.9555,
+      "learning_rate": 1.074743186553408e-05,
+      "loss": 0.8487,
       "step": 17335
     },
     {
-      "epoch": 0.47611985389030786,
+      "epoch": 0.4919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.1256226888532666e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.074651535514067e-05,
+      "loss": 0.8443,
       "step": 17336
     },
     {
-      "epoch": 0.4761473181181511,
+      "epoch": 0.49196935300794553,
       "grad_norm": 0.0,
-      "learning_rate": 1.1255344414215553e-05,
-      "loss": 1.004,
+      "learning_rate": 1.074559883844141e-05,
+      "loss": 0.8935,
       "step": 17337
     },
     {
-      "epoch": 0.47617478234599436,
+      "epoch": 0.4919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.125446192996566e-05,
-      "loss": 0.9666,
+      "learning_rate": 1.0744682315444044e-05,
+      "loss": 1.0414,
       "step": 17338
     },
     {
-      "epoch": 0.4762022465738376,
+      "epoch": 0.4920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.1253579435789984e-05,
-      "loss": 0.8246,
+      "learning_rate": 1.0743765786156313e-05,
+      "loss": 0.9195,
       "step": 17339
     },
     {
-      "epoch": 0.4762297108016808,
+      "epoch": 0.49205448354143017,
       "grad_norm": 0.0,
-      "learning_rate": 1.1252696931695502e-05,
-      "loss": 0.9041,
+      "learning_rate": 1.0742849250585957e-05,
+      "loss": 0.8606,
       "step": 17340
     },
     {
-      "epoch": 0.476257175029524,
+      "epoch": 0.4920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.1251814417689195e-05,
-      "loss": 0.8827,
+      "learning_rate": 1.074193270874072e-05,
+      "loss": 0.905,
       "step": 17341
     },
     {
-      "epoch": 0.4762846392573673,
+      "epoch": 0.49211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.1250931893778045e-05,
-      "loss": 0.9224,
+      "learning_rate": 1.0741016160628346e-05,
+      "loss": 1.0023,
       "step": 17342
     },
     {
-      "epoch": 0.4763121034852105,
+      "epoch": 0.49213961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 1.1250049359969041e-05,
-      "loss": 0.9693,
+      "learning_rate": 1.0740099606256576e-05,
+      "loss": 0.9565,
       "step": 17343
     },
     {
-      "epoch": 0.47633956771305375,
+      "epoch": 0.49216799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 1.1249166816269159e-05,
-      "loss": 0.9663,
+      "learning_rate": 1.0739183045633148e-05,
+      "loss": 1.1,
       "step": 17344
     },
     {
-      "epoch": 0.47636703194089697,
+      "epoch": 0.49219636776390463,
       "grad_norm": 0.0,
-      "learning_rate": 1.124828426268539e-05,
-      "loss": 0.8367,
+      "learning_rate": 1.0738266478765811e-05,
+      "loss": 0.8959,
       "step": 17345
     },
     {
-      "epoch": 0.4763944961687402,
+      "epoch": 0.49222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.124740169922471e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.0737349905662305e-05,
+      "loss": 0.932,
       "step": 17346
     },
     {
-      "epoch": 0.47642196039658347,
+      "epoch": 0.49225312145289446,
       "grad_norm": 0.0,
-      "learning_rate": 1.1246519125894102e-05,
-      "loss": 0.8396,
+      "learning_rate": 1.0736433326330366e-05,
+      "loss": 0.8869,
       "step": 17347
     },
     {
-      "epoch": 0.4764494246244267,
+      "epoch": 0.4922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.1245636542700553e-05,
-      "loss": 0.9354,
+      "learning_rate": 1.0735516740777742e-05,
+      "loss": 0.8511,
       "step": 17348
     },
     {
-      "epoch": 0.4764768888522699,
+      "epoch": 0.49230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.124475394965105e-05,
-      "loss": 0.9505,
+      "learning_rate": 1.073460014901218e-05,
+      "loss": 0.9692,
       "step": 17349
     },
     {
-      "epoch": 0.47650435308011313,
+      "epoch": 0.4923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.1243871346752566e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.0733683551041416e-05,
+      "loss": 1.0201,
       "step": 17350
     },
     {
-      "epoch": 0.4765318173079564,
+      "epoch": 0.492366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.124298873401209e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.073276694687319e-05,
+      "loss": 0.9275,
       "step": 17351
     },
     {
-      "epoch": 0.47655928153579963,
+      "epoch": 0.4923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.1242106111436608e-05,
-      "loss": 1.0255,
+      "learning_rate": 1.0731850336515253e-05,
+      "loss": 0.9904,
       "step": 17352
     },
     {
-      "epoch": 0.47658674576364285,
+      "epoch": 0.4924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.1241223479033103e-05,
-      "loss": 0.8767,
+      "learning_rate": 1.0730933719975343e-05,
+      "loss": 0.8881,
       "step": 17353
     },
     {
-      "epoch": 0.47661420999148607,
+      "epoch": 0.4924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.1240340836808554e-05,
-      "loss": 0.9109,
+      "learning_rate": 1.07300170972612e-05,
+      "loss": 0.9503,
       "step": 17354
     },
     {
-      "epoch": 0.47664167421932935,
+      "epoch": 0.49248013620885356,
       "grad_norm": 0.0,
-      "learning_rate": 1.123945818476995e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.0729100468380574e-05,
+      "loss": 0.8992,
       "step": 17355
     },
     {
-      "epoch": 0.47666913844717257,
+      "epoch": 0.4925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.1238575522924268e-05,
-      "loss": 0.9407,
+      "learning_rate": 1.0728183833341199e-05,
+      "loss": 0.967,
       "step": 17356
     },
     {
-      "epoch": 0.4766966026750158,
+      "epoch": 0.49253688989784333,
       "grad_norm": 0.0,
-      "learning_rate": 1.12376928512785e-05,
-      "loss": 0.9789,
+      "learning_rate": 1.0727267192150825e-05,
+      "loss": 0.9512,
       "step": 17357
     },
     {
-      "epoch": 0.476724066902859,
+      "epoch": 0.49256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.123681016983963e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.0726350544817192e-05,
+      "loss": 0.9176,
       "step": 17358
     },
     {
-      "epoch": 0.47675153113070223,
+      "epoch": 0.49259364358683316,
       "grad_norm": 0.0,
-      "learning_rate": 1.1235927478614634e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.0725433891348043e-05,
+      "loss": 0.8867,
       "step": 17359
     },
     {
-      "epoch": 0.4767789953585455,
+      "epoch": 0.492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.1235044777610502e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.0724517231751123e-05,
+      "loss": 0.8959,
       "step": 17360
     },
     {
-      "epoch": 0.47680645958638873,
+      "epoch": 0.49265039727582294,
       "grad_norm": 0.0,
-      "learning_rate": 1.123416206683422e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.0723600566034172e-05,
+      "loss": 0.7663,
       "step": 17361
     },
     {
-      "epoch": 0.47683392381423195,
+      "epoch": 0.4926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.1233279346292767e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.0722683894204935e-05,
+      "loss": 0.9257,
       "step": 17362
     },
     {
-      "epoch": 0.4768613880420752,
+      "epoch": 0.4927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.1232396615993129e-05,
-      "loss": 0.8604,
+      "learning_rate": 1.0721767216271153e-05,
+      "loss": 0.9673,
       "step": 17363
     },
     {
-      "epoch": 0.47688885226991845,
+      "epoch": 0.4927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.1231513875942295e-05,
-      "loss": 0.8554,
+      "learning_rate": 1.0720850532240572e-05,
+      "loss": 0.9957,
       "step": 17364
     },
     {
-      "epoch": 0.47691631649776167,
+      "epoch": 0.4927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.1230631126147245e-05,
-      "loss": 0.961,
+      "learning_rate": 1.0719933842120933e-05,
+      "loss": 0.9534,
       "step": 17365
     },
     {
-      "epoch": 0.4769437807256049,
+      "epoch": 0.4927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.1229748366614963e-05,
-      "loss": 0.918,
+      "learning_rate": 1.0719017145919984e-05,
+      "loss": 0.8293,
       "step": 17366
     },
     {
-      "epoch": 0.4769712449534481,
+      "epoch": 0.49282065834279226,
       "grad_norm": 0.0,
-      "learning_rate": 1.1228865597352438e-05,
-      "loss": 0.9642,
+      "learning_rate": 1.0718100443645461e-05,
+      "loss": 0.7501,
       "step": 17367
     },
     {
-      "epoch": 0.4769987091812914,
+      "epoch": 0.4928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.122798281836665e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.0717183735305112e-05,
+      "loss": 0.8624,
       "step": 17368
     },
     {
-      "epoch": 0.4770261734091346,
+      "epoch": 0.4928774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.1227100029664588e-05,
-      "loss": 0.8766,
+      "learning_rate": 1.0716267020906681e-05,
+      "loss": 0.9081,
       "step": 17369
     },
     {
-      "epoch": 0.47705363763697783,
+      "epoch": 0.49290578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.1226217231253236e-05,
-      "loss": 0.8582,
+      "learning_rate": 1.0715350300457912e-05,
+      "loss": 0.8433,
       "step": 17370
     },
     {
-      "epoch": 0.47708110186482106,
+      "epoch": 0.49293416572077187,
       "grad_norm": 0.0,
-      "learning_rate": 1.1225334423139575e-05,
-      "loss": 1.0295,
+      "learning_rate": 1.071443357396654e-05,
+      "loss": 0.8976,
       "step": 17371
     },
     {
-      "epoch": 0.47710856609266433,
+      "epoch": 0.4929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.1224451605330595e-05,
-      "loss": 0.8847,
+      "learning_rate": 1.0713516841440321e-05,
+      "loss": 0.9626,
       "step": 17372
     },
     {
-      "epoch": 0.47713603032050755,
+      "epoch": 0.49299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.1223568777833282e-05,
-      "loss": 0.8778,
+      "learning_rate": 1.0712600102886995e-05,
+      "loss": 0.817,
       "step": 17373
     },
     {
-      "epoch": 0.4771634945483508,
+      "epoch": 0.4930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.1222685940654617e-05,
-      "loss": 0.9427,
+      "learning_rate": 1.0711683358314298e-05,
+      "loss": 0.9301,
       "step": 17374
     },
     {
-      "epoch": 0.477190958776194,
+      "epoch": 0.4930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.1221803093801584e-05,
-      "loss": 0.8903,
+      "learning_rate": 1.0710766607729983e-05,
+      "loss": 0.8438,
       "step": 17375
     },
     {
-      "epoch": 0.4772184230040372,
+      "epoch": 0.49307604994324633,
       "grad_norm": 0.0,
-      "learning_rate": 1.1220920237281177e-05,
-      "loss": 0.9134,
+      "learning_rate": 1.0709849851141792e-05,
+      "loss": 0.9989,
       "step": 17376
     },
     {
-      "epoch": 0.4772458872318805,
+      "epoch": 0.4931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.1220037371100375e-05,
-      "loss": 0.7348,
+      "learning_rate": 1.0708933088557468e-05,
+      "loss": 0.9778,
       "step": 17377
     },
     {
-      "epoch": 0.4772733514597237,
+      "epoch": 0.4931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.121915449526616e-05,
-      "loss": 0.9302,
+      "learning_rate": 1.0708016319984751e-05,
+      "loss": 1.011,
       "step": 17378
     },
     {
-      "epoch": 0.47730081568756694,
+      "epoch": 0.49316118047673096,
       "grad_norm": 0.0,
-      "learning_rate": 1.1218271609785529e-05,
-      "loss": 0.8155,
+      "learning_rate": 1.0707099545431388e-05,
+      "loss": 0.8192,
       "step": 17379
     },
     {
-      "epoch": 0.47732827991541016,
+      "epoch": 0.4931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.1217388714665456e-05,
-      "loss": 0.9653,
+      "learning_rate": 1.0706182764905124e-05,
+      "loss": 0.8413,
       "step": 17380
     },
     {
-      "epoch": 0.47735574414325344,
+      "epoch": 0.4932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.1216505809912935e-05,
-      "loss": 0.9435,
+      "learning_rate": 1.0705265978413703e-05,
+      "loss": 0.9194,
       "step": 17381
     },
     {
-      "epoch": 0.47738320837109666,
+      "epoch": 0.49324631101021565,
       "grad_norm": 0.0,
-      "learning_rate": 1.1215622895534948e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.0704349185964868e-05,
+      "loss": 0.9533,
       "step": 17382
     },
     {
-      "epoch": 0.4774106725989399,
+      "epoch": 0.49327468785471057,
       "grad_norm": 0.0,
-      "learning_rate": 1.1214739971538479e-05,
-      "loss": 0.8979,
+      "learning_rate": 1.0703432387566367e-05,
+      "loss": 0.8894,
       "step": 17383
     },
     {
-      "epoch": 0.4774381368267831,
+      "epoch": 0.49330306469920543,
       "grad_norm": 0.0,
-      "learning_rate": 1.121385703793052e-05,
-      "loss": 0.8696,
+      "learning_rate": 1.0702515583225936e-05,
+      "loss": 0.8316,
       "step": 17384
     },
     {
-      "epoch": 0.4774656010546264,
+      "epoch": 0.49333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.1212974094718056e-05,
-      "loss": 0.9343,
+      "learning_rate": 1.070159877295133e-05,
+      "loss": 0.8463,
       "step": 17385
     },
     {
-      "epoch": 0.4774930652824696,
+      "epoch": 0.49335981838819526,
       "grad_norm": 0.0,
-      "learning_rate": 1.1212091141908067e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.0700681956750282e-05,
+      "loss": 0.9068,
       "step": 17386
     },
     {
-      "epoch": 0.4775205295103128,
+      "epoch": 0.4933881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.1211208179507544e-05,
-      "loss": 0.8098,
+      "learning_rate": 1.0699765134630544e-05,
+      "loss": 0.8876,
       "step": 17387
     },
     {
-      "epoch": 0.47754799373815604,
+      "epoch": 0.49341657207718503,
       "grad_norm": 0.0,
-      "learning_rate": 1.1210325207523474e-05,
-      "loss": 0.9423,
+      "learning_rate": 1.0698848306599858e-05,
+      "loss": 0.8811,
       "step": 17388
     },
     {
-      "epoch": 0.47757545796599926,
+      "epoch": 0.4934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.1209442225962843e-05,
-      "loss": 0.9518,
+      "learning_rate": 1.0697931472665972e-05,
+      "loss": 0.85,
       "step": 17389
     },
     {
-      "epoch": 0.47760292219384254,
+      "epoch": 0.4934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.1208559234832635e-05,
-      "loss": 0.8613,
+      "learning_rate": 1.0697014632836627e-05,
+      "loss": 0.9152,
       "step": 17390
     },
     {
-      "epoch": 0.47763038642168576,
+      "epoch": 0.49350170261066967,
       "grad_norm": 0.0,
-      "learning_rate": 1.1207676234139838e-05,
-      "loss": 0.8531,
+      "learning_rate": 1.0696097787119563e-05,
+      "loss": 0.9452,
       "step": 17391
     },
     {
-      "epoch": 0.477657850649529,
+      "epoch": 0.4935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.1206793223891437e-05,
-      "loss": 0.9287,
+      "learning_rate": 1.0695180935522536e-05,
+      "loss": 0.9033,
       "step": 17392
     },
     {
-      "epoch": 0.4776853148773722,
+      "epoch": 0.4935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.1205910204094423e-05,
-      "loss": 0.9809,
+      "learning_rate": 1.069426407805328e-05,
+      "loss": 1.02,
       "step": 17393
     },
     {
-      "epoch": 0.4777127791052155,
+      "epoch": 0.49358683314415436,
       "grad_norm": 0.0,
-      "learning_rate": 1.1205027174755783e-05,
-      "loss": 0.8722,
+      "learning_rate": 1.0693347214719545e-05,
+      "loss": 0.8619,
       "step": 17394
     },
     {
-      "epoch": 0.4777402433330587,
+      "epoch": 0.4936152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.1204144135882496e-05,
-      "loss": 0.8044,
+      "learning_rate": 1.0692430345529074e-05,
+      "loss": 0.8041,
       "step": 17395
     },
     {
-      "epoch": 0.4777677075609019,
+      "epoch": 0.49364358683314413,
       "grad_norm": 0.0,
-      "learning_rate": 1.1203261087481558e-05,
-      "loss": 0.9184,
+      "learning_rate": 1.0691513470489616e-05,
+      "loss": 0.9913,
       "step": 17396
     },
     {
-      "epoch": 0.47779517178874514,
+      "epoch": 0.49367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.120237802955995e-05,
-      "loss": 0.8902,
+      "learning_rate": 1.0690596589608912e-05,
+      "loss": 0.9246,
       "step": 17397
     },
     {
-      "epoch": 0.4778226360165884,
+      "epoch": 0.49370034052213396,
       "grad_norm": 0.0,
-      "learning_rate": 1.1201494962124662e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.0689679702894704e-05,
+      "loss": 0.9018,
       "step": 17398
     },
     {
-      "epoch": 0.47785010024443164,
+      "epoch": 0.4937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.1200611885182682e-05,
-      "loss": 0.9054,
+      "learning_rate": 1.0688762810354745e-05,
+      "loss": 0.9021,
       "step": 17399
     },
     {
-      "epoch": 0.47787756447227486,
+      "epoch": 0.49375709421112374,
       "grad_norm": 0.0,
-      "learning_rate": 1.1199728798740992e-05,
-      "loss": 0.924,
+      "learning_rate": 1.0687845911996775e-05,
+      "loss": 0.9115,
       "step": 17400
     },
     {
-      "epoch": 0.4779050287001181,
+      "epoch": 0.4937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.1198845702806587e-05,
-      "loss": 0.8168,
+      "learning_rate": 1.0686929007828536e-05,
+      "loss": 0.9444,
       "step": 17401
     },
     {
-      "epoch": 0.4779324929279613,
+      "epoch": 0.4938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.1197962597386449e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.068601209785778e-05,
+      "loss": 0.9143,
       "step": 17402
     },
     {
-      "epoch": 0.4779599571558046,
+      "epoch": 0.4938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1197079482487566e-05,
-      "loss": 0.9123,
+      "learning_rate": 1.0685095182092246e-05,
+      "loss": 0.9286,
       "step": 17403
     },
     {
-      "epoch": 0.4779874213836478,
+      "epoch": 0.4938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.1196196358116925e-05,
-      "loss": 0.9084,
+      "learning_rate": 1.0684178260539686e-05,
+      "loss": 0.9259,
       "step": 17404
     },
     {
-      "epoch": 0.478014885611491,
+      "epoch": 0.4938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.1195313224281519e-05,
-      "loss": 0.967,
+      "learning_rate": 1.0683261333207837e-05,
+      "loss": 0.9288,
       "step": 17405
     },
     {
-      "epoch": 0.47804234983933425,
+      "epoch": 0.49392735527809306,
       "grad_norm": 0.0,
-      "learning_rate": 1.1194430080988328e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.0682344400104454e-05,
+      "loss": 0.9109,
       "step": 17406
     },
     {
-      "epoch": 0.4780698140671775,
+      "epoch": 0.493955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.1193546928244343e-05,
-      "loss": 0.8716,
+      "learning_rate": 1.0681427461237275e-05,
+      "loss": 0.774,
       "step": 17407
     },
     {
-      "epoch": 0.47809727829502074,
+      "epoch": 0.49398410896708284,
       "grad_norm": 0.0,
-      "learning_rate": 1.1192663766056554e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.0680510516614047e-05,
+      "loss": 0.8892,
       "step": 17408
     },
     {
-      "epoch": 0.47812474252286397,
+      "epoch": 0.49401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.1191780594431951e-05,
-      "loss": 0.9495,
+      "learning_rate": 1.0679593566242514e-05,
+      "loss": 0.818,
       "step": 17409
     },
     {
-      "epoch": 0.4781522067507072,
+      "epoch": 0.49404086265607267,
       "grad_norm": 0.0,
-      "learning_rate": 1.1190897413377513e-05,
-      "loss": 0.973,
+      "learning_rate": 1.0678676610130427e-05,
+      "loss": 0.908,
       "step": 17410
     },
     {
-      "epoch": 0.47817967097855046,
+      "epoch": 0.4940692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.1190014222900237e-05,
-      "loss": 0.8341,
+      "learning_rate": 1.0677759648285529e-05,
+      "loss": 0.9073,
       "step": 17411
     },
     {
-      "epoch": 0.4782071352063937,
+      "epoch": 0.49409761634506244,
       "grad_norm": 0.0,
-      "learning_rate": 1.1189131023007105e-05,
-      "loss": 0.8322,
+      "learning_rate": 1.067684268071556e-05,
+      "loss": 1.0313,
       "step": 17412
     },
     {
-      "epoch": 0.4782345994342369,
+      "epoch": 0.4941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.1188247813705109e-05,
-      "loss": 0.8954,
+      "learning_rate": 1.0675925707428274e-05,
+      "loss": 0.7686,
       "step": 17413
     },
     {
-      "epoch": 0.47826206366208013,
+      "epoch": 0.4941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.1187364595001235e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.0675008728431415e-05,
+      "loss": 0.9192,
       "step": 17414
     },
     {
-      "epoch": 0.47828952788992335,
+      "epoch": 0.49418274687854713,
       "grad_norm": 0.0,
-      "learning_rate": 1.1186481366902474e-05,
-      "loss": 0.858,
+      "learning_rate": 1.0674091743732721e-05,
+      "loss": 0.9421,
       "step": 17415
     },
     {
-      "epoch": 0.4783169921177666,
+      "epoch": 0.494211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.1185598129415809e-05,
-      "loss": 1.013,
+      "learning_rate": 1.067317475333995e-05,
+      "loss": 0.9772,
       "step": 17416
     },
     {
-      "epoch": 0.47834445634560985,
+      "epoch": 0.4942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.1184714882548237e-05,
-      "loss": 0.8552,
+      "learning_rate": 1.0672257757260838e-05,
+      "loss": 0.813,
       "step": 17417
     },
     {
-      "epoch": 0.47837192057345307,
+      "epoch": 0.49426787741203176,
       "grad_norm": 0.0,
-      "learning_rate": 1.118383162630674e-05,
-      "loss": 0.8721,
+      "learning_rate": 1.0671340755503135e-05,
+      "loss": 0.9408,
       "step": 17418
     },
     {
-      "epoch": 0.4783993848012963,
+      "epoch": 0.4942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.1182948360698309e-05,
-      "loss": 0.8662,
+      "learning_rate": 1.0670423748074586e-05,
+      "loss": 0.8319,
       "step": 17419
     },
     {
-      "epoch": 0.47842684902913957,
+      "epoch": 0.49432463110102154,
       "grad_norm": 0.0,
-      "learning_rate": 1.1182065085729931e-05,
-      "loss": 0.9047,
+      "learning_rate": 1.066950673498294e-05,
+      "loss": 0.895,
       "step": 17420
     },
     {
-      "epoch": 0.4784543132569828,
+      "epoch": 0.49435300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 1.1181181801408599e-05,
-      "loss": 0.8925,
+      "learning_rate": 1.0668589716235937e-05,
+      "loss": 0.8434,
       "step": 17421
     },
     {
-      "epoch": 0.478481777484826,
+      "epoch": 0.49438138479001137,
       "grad_norm": 0.0,
-      "learning_rate": 1.1180298507741296e-05,
-      "loss": 0.9169,
+      "learning_rate": 1.0667672691841329e-05,
+      "loss": 0.9223,
       "step": 17422
     },
     {
-      "epoch": 0.47850924171266923,
+      "epoch": 0.49440976163450623,
       "grad_norm": 0.0,
-      "learning_rate": 1.1179415204735015e-05,
-      "loss": 0.994,
+      "learning_rate": 1.066675566180686e-05,
+      "loss": 0.7967,
       "step": 17423
     },
     {
-      "epoch": 0.4785367059405125,
+      "epoch": 0.49443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.1178531892396745e-05,
-      "loss": 0.8851,
+      "learning_rate": 1.0665838626140274e-05,
+      "loss": 0.8811,
       "step": 17424
     },
     {
-      "epoch": 0.47856417016835573,
+      "epoch": 0.494466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.1177648570733472e-05,
-      "loss": 0.993,
+      "learning_rate": 1.0664921584849317e-05,
+      "loss": 0.9686,
       "step": 17425
     },
     {
-      "epoch": 0.47859163439619895,
+      "epoch": 0.4944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.1176765239752192e-05,
-      "loss": 0.8666,
+      "learning_rate": 1.0664004537941744e-05,
+      "loss": 0.8712,
       "step": 17426
     },
     {
-      "epoch": 0.47861909862404217,
+      "epoch": 0.49452326901248583,
       "grad_norm": 0.0,
-      "learning_rate": 1.1175881899459884e-05,
-      "loss": 0.7865,
+      "learning_rate": 1.0663087485425293e-05,
+      "loss": 0.9066,
       "step": 17427
     },
     {
-      "epoch": 0.4786465628518854,
+      "epoch": 0.4945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.1174998549863545e-05,
-      "loss": 0.809,
+      "learning_rate": 1.0662170427307709e-05,
+      "loss": 0.9561,
       "step": 17428
     },
     {
-      "epoch": 0.47867402707972867,
+      "epoch": 0.4945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.1174115190970166e-05,
-      "loss": 0.804,
+      "learning_rate": 1.0661253363596742e-05,
+      "loss": 0.8249,
       "step": 17429
     },
     {
-      "epoch": 0.4787014913075719,
+      "epoch": 0.49460839954597047,
       "grad_norm": 0.0,
-      "learning_rate": 1.1173231822786727e-05,
-      "loss": 0.9086,
+      "learning_rate": 1.0660336294300143e-05,
+      "loss": 0.9385,
       "step": 17430
     },
     {
-      "epoch": 0.4787289555354151,
+      "epoch": 0.4946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.1172348445320224e-05,
-      "loss": 0.8312,
+      "learning_rate": 1.0659419219425649e-05,
+      "loss": 0.8125,
       "step": 17431
     },
     {
-      "epoch": 0.47875641976325833,
+      "epoch": 0.4946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.1171465058577648e-05,
-      "loss": 0.944,
+      "learning_rate": 1.0658502138981009e-05,
+      "loss": 0.7705,
       "step": 17432
     },
     {
-      "epoch": 0.4787838839911016,
+      "epoch": 0.49469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.1170581662565989e-05,
-      "loss": 0.968,
+      "learning_rate": 1.0657585052973976e-05,
+      "loss": 0.8751,
       "step": 17433
     },
     {
-      "epoch": 0.47881134821894483,
+      "epoch": 0.49472190692395007,
       "grad_norm": 0.0,
-      "learning_rate": 1.116969825729223e-05,
-      "loss": 0.9857,
+      "learning_rate": 1.0656667961412293e-05,
+      "loss": 0.8826,
       "step": 17434
     },
     {
-      "epoch": 0.47883881244678805,
+      "epoch": 0.49475028376844493,
       "grad_norm": 0.0,
-      "learning_rate": 1.1168814842763366e-05,
-      "loss": 0.802,
+      "learning_rate": 1.0655750864303702e-05,
+      "loss": 0.9128,
       "step": 17435
     },
     {
-      "epoch": 0.4788662766746313,
+      "epoch": 0.49477866061293985,
       "grad_norm": 0.0,
-      "learning_rate": 1.1167931418986389e-05,
-      "loss": 0.9312,
+      "learning_rate": 1.0654833761655958e-05,
+      "loss": 0.894,
       "step": 17436
     },
     {
-      "epoch": 0.47889374090247455,
+      "epoch": 0.4948070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.1167047985968284e-05,
-      "loss": 0.8342,
+      "learning_rate": 1.0653916653476803e-05,
+      "loss": 0.8831,
       "step": 17437
     },
     {
-      "epoch": 0.4789212051303178,
+      "epoch": 0.4948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.1166164543716044e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.0652999539773985e-05,
+      "loss": 0.7984,
       "step": 17438
     },
     {
-      "epoch": 0.478948669358161,
+      "epoch": 0.49486379114642454,
       "grad_norm": 0.0,
-      "learning_rate": 1.1165281092236658e-05,
-      "loss": 0.9399,
+      "learning_rate": 1.065208242055525e-05,
+      "loss": 0.9499,
       "step": 17439
     },
     {
-      "epoch": 0.4789761335860042,
+      "epoch": 0.4948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.1164397631537116e-05,
-      "loss": 0.9429,
+      "learning_rate": 1.0651165295828342e-05,
+      "loss": 0.8949,
       "step": 17440
     },
     {
-      "epoch": 0.47900359781384744,
+      "epoch": 0.4949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.1163514161624414e-05,
-      "loss": 0.9031,
+      "learning_rate": 1.0650248165601018e-05,
+      "loss": 0.8046,
       "step": 17441
     },
     {
-      "epoch": 0.4790310620416907,
+      "epoch": 0.49494892167990917,
       "grad_norm": 0.0,
-      "learning_rate": 1.1162630682505531e-05,
-      "loss": 0.8284,
+      "learning_rate": 1.0649331029881013e-05,
+      "loss": 0.9124,
       "step": 17442
     },
     {
-      "epoch": 0.47905852626953394,
+      "epoch": 0.4949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.1161747194187464e-05,
-      "loss": 0.9348,
+      "learning_rate": 1.0648413888676084e-05,
+      "loss": 0.8559,
       "step": 17443
     },
     {
-      "epoch": 0.47908599049737716,
+      "epoch": 0.495005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.1160863696677205e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.0647496741993972e-05,
+      "loss": 0.8252,
       "step": 17444
     },
     {
-      "epoch": 0.4791134547252204,
+      "epoch": 0.49503405221339386,
       "grad_norm": 0.0,
-      "learning_rate": 1.1159980189981743e-05,
-      "loss": 0.8632,
+      "learning_rate": 1.0646579589842423e-05,
+      "loss": 0.9025,
       "step": 17445
     },
     {
-      "epoch": 0.47914091895306365,
+      "epoch": 0.4950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.1159096674108068e-05,
-      "loss": 0.9066,
+      "learning_rate": 1.0645662432229192e-05,
+      "loss": 0.9508,
       "step": 17446
     },
     {
-      "epoch": 0.4791683831809069,
+      "epoch": 0.49509080590238363,
       "grad_norm": 0.0,
-      "learning_rate": 1.115821314906317e-05,
-      "loss": 0.8956,
+      "learning_rate": 1.064474526916202e-05,
+      "loss": 0.8962,
       "step": 17447
     },
     {
-      "epoch": 0.4791958474087501,
+      "epoch": 0.49511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.1157329614854042e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.0643828100648652e-05,
+      "loss": 0.9047,
       "step": 17448
     },
     {
-      "epoch": 0.4792233116365933,
+      "epoch": 0.49514755959137347,
       "grad_norm": 0.0,
-      "learning_rate": 1.1156446071487675e-05,
-      "loss": 0.8574,
+      "learning_rate": 1.0642910926696842e-05,
+      "loss": 0.9558,
       "step": 17449
     },
     {
-      "epoch": 0.4792507758644366,
+      "epoch": 0.4951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.1155562518971059e-05,
-      "loss": 0.8668,
+      "learning_rate": 1.0641993747314335e-05,
+      "loss": 0.8393,
       "step": 17450
     },
     {
-      "epoch": 0.4792782400922798,
+      "epoch": 0.49520431328036324,
       "grad_norm": 0.0,
-      "learning_rate": 1.115467895731118e-05,
-      "loss": 0.8912,
+      "learning_rate": 1.0641076562508878e-05,
+      "loss": 0.9547,
       "step": 17451
     },
     {
-      "epoch": 0.47930570432012304,
+      "epoch": 0.4952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.1153795386515037e-05,
-      "loss": 0.9216,
+      "learning_rate": 1.0640159372288217e-05,
+      "loss": 0.8514,
       "step": 17452
     },
     {
-      "epoch": 0.47933316854796626,
+      "epoch": 0.495261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.115291180658962e-05,
-      "loss": 1.0022,
+      "learning_rate": 1.0639242176660103e-05,
+      "loss": 0.9447,
       "step": 17453
     },
     {
-      "epoch": 0.4793606327758095,
+      "epoch": 0.4952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1152028217541918e-05,
-      "loss": 1.0054,
+      "learning_rate": 1.063832497563228e-05,
+      "loss": 0.9198,
       "step": 17454
     },
     {
-      "epoch": 0.47938809700365276,
+      "epoch": 0.4953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.1151144619378917e-05,
-      "loss": 0.8493,
+      "learning_rate": 1.0637407769212495e-05,
+      "loss": 0.9359,
       "step": 17455
     },
     {
-      "epoch": 0.479415561231496,
+      "epoch": 0.4953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.115026101210762e-05,
-      "loss": 0.8967,
+      "learning_rate": 1.0636490557408501e-05,
+      "loss": 0.9502,
       "step": 17456
     },
     {
-      "epoch": 0.4794430254593392,
+      "epoch": 0.49537457434733256,
       "grad_norm": 0.0,
-      "learning_rate": 1.114937739573501e-05,
-      "loss": 0.903,
+      "learning_rate": 1.0635573340228042e-05,
+      "loss": 0.8166,
       "step": 17457
     },
     {
-      "epoch": 0.4794704896871824,
+      "epoch": 0.4954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.114849377026808e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.0634656117678868e-05,
+      "loss": 0.8348,
       "step": 17458
     },
     {
-      "epoch": 0.4794979539150257,
+      "epoch": 0.49543132803632234,
       "grad_norm": 0.0,
-      "learning_rate": 1.1147610135713822e-05,
-      "loss": 0.7692,
+      "learning_rate": 1.0633738889768723e-05,
+      "loss": 0.9899,
       "step": 17459
     },
     {
-      "epoch": 0.4795254181428689,
+      "epoch": 0.49545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.114672649207923e-05,
-      "loss": 0.937,
+      "learning_rate": 1.063282165650536e-05,
+      "loss": 0.8425,
       "step": 17460
     },
     {
-      "epoch": 0.47955288237071214,
+      "epoch": 0.49548808172531217,
       "grad_norm": 0.0,
-      "learning_rate": 1.1145842839371293e-05,
-      "loss": 0.8448,
+      "learning_rate": 1.0631904417896522e-05,
+      "loss": 0.9355,
       "step": 17461
     },
     {
-      "epoch": 0.47958034659855536,
+      "epoch": 0.49551645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.1144959177597004e-05,
-      "loss": 1.0303,
+      "learning_rate": 1.0630987173949959e-05,
+      "loss": 0.9156,
       "step": 17462
     },
     {
-      "epoch": 0.47960781082639864,
+      "epoch": 0.49554483541430194,
       "grad_norm": 0.0,
-      "learning_rate": 1.1144075506763352e-05,
-      "loss": 0.9937,
+      "learning_rate": 1.0630069924673416e-05,
+      "loss": 0.9311,
       "step": 17463
     },
     {
-      "epoch": 0.47963527505424186,
+      "epoch": 0.4955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.1143191826877335e-05,
-      "loss": 0.8745,
+      "learning_rate": 1.0629152670074648e-05,
+      "loss": 0.8501,
       "step": 17464
     },
     {
-      "epoch": 0.4796627392820851,
+      "epoch": 0.4956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.1142308137945941e-05,
-      "loss": 0.8926,
+      "learning_rate": 1.0628235410161398e-05,
+      "loss": 0.8651,
       "step": 17465
     },
     {
-      "epoch": 0.4796902035099283,
+      "epoch": 0.49562996594778663,
       "grad_norm": 0.0,
-      "learning_rate": 1.114142443997616e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.0627318144941415e-05,
+      "loss": 0.9953,
       "step": 17466
     },
     {
-      "epoch": 0.4797176677377715,
+      "epoch": 0.4956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.114054073297499e-05,
-      "loss": 0.902,
+      "learning_rate": 1.062640087442245e-05,
+      "loss": 0.9477,
       "step": 17467
     },
     {
-      "epoch": 0.4797451319656148,
+      "epoch": 0.4956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.1139657016949417e-05,
-      "loss": 0.832,
+      "learning_rate": 1.0625483598612245e-05,
+      "loss": 0.9512,
       "step": 17468
     },
     {
-      "epoch": 0.479772596193458,
+      "epoch": 0.49571509648127127,
       "grad_norm": 0.0,
-      "learning_rate": 1.1138773291906438e-05,
-      "loss": 0.7757,
+      "learning_rate": 1.0624566317518552e-05,
+      "loss": 0.8416,
       "step": 17469
     },
     {
-      "epoch": 0.47980006042130124,
+      "epoch": 0.4957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.1137889557853043e-05,
-      "loss": 0.9263,
+      "learning_rate": 1.0623649031149122e-05,
+      "loss": 0.9006,
       "step": 17470
     },
     {
-      "epoch": 0.47982752464914447,
+      "epoch": 0.49577185017026104,
       "grad_norm": 0.0,
-      "learning_rate": 1.1137005814796224e-05,
-      "loss": 0.9273,
+      "learning_rate": 1.0622731739511699e-05,
+      "loss": 0.894,
       "step": 17471
     },
     {
-      "epoch": 0.47985498887698774,
+      "epoch": 0.49580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.1136122062742973e-05,
-      "loss": 0.8493,
+      "learning_rate": 1.0621814442614033e-05,
+      "loss": 0.9303,
       "step": 17472
     },
     {
-      "epoch": 0.47988245310483096,
+      "epoch": 0.49582860385925087,
       "grad_norm": 0.0,
-      "learning_rate": 1.113523830170029e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.062089714046387e-05,
+      "loss": 0.8883,
       "step": 17473
     },
     {
-      "epoch": 0.4799099173326742,
+      "epoch": 0.49585698070374573,
       "grad_norm": 0.0,
-      "learning_rate": 1.1134354531675158e-05,
-      "loss": 0.8544,
+      "learning_rate": 1.0619979833068965e-05,
+      "loss": 0.7952,
       "step": 17474
     },
     {
-      "epoch": 0.4799373815605174,
+      "epoch": 0.49588535754824065,
       "grad_norm": 0.0,
-      "learning_rate": 1.1133470752674572e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.0619062520437062e-05,
+      "loss": 0.8808,
       "step": 17475
     },
     {
-      "epoch": 0.4799648457883607,
+      "epoch": 0.4959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.113258696470553e-05,
-      "loss": 0.9067,
+      "learning_rate": 1.0618145202575909e-05,
+      "loss": 0.8537,
       "step": 17476
     },
     {
-      "epoch": 0.4799923100162039,
+      "epoch": 0.4959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.1131703167775018e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.0617227879493257e-05,
+      "loss": 0.8716,
       "step": 17477
     },
     {
-      "epoch": 0.4800197742440471,
+      "epoch": 0.49597048808172534,
       "grad_norm": 0.0,
-      "learning_rate": 1.1130819361890033e-05,
-      "loss": 0.9637,
+      "learning_rate": 1.061631055119685e-05,
+      "loss": 0.9143,
       "step": 17478
     },
     {
-      "epoch": 0.48004723847189035,
+      "epoch": 0.4959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.1129935547057568e-05,
-      "loss": 0.9479,
+      "learning_rate": 1.0615393217694441e-05,
+      "loss": 0.942,
       "step": 17479
     },
     {
-      "epoch": 0.4800747026997336,
+      "epoch": 0.4960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.1129051723284616e-05,
-      "loss": 0.9449,
+      "learning_rate": 1.061447587899378e-05,
+      "loss": 0.8954,
       "step": 17480
     },
     {
-      "epoch": 0.48010216692757685,
+      "epoch": 0.49605561861520997,
       "grad_norm": 0.0,
-      "learning_rate": 1.112816789057817e-05,
-      "loss": 0.9212,
+      "learning_rate": 1.0613558535102613e-05,
+      "loss": 0.8922,
       "step": 17481
     },
     {
-      "epoch": 0.48012963115542007,
+      "epoch": 0.4960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.1127284048945222e-05,
-      "loss": 0.8883,
+      "learning_rate": 1.0612641186028692e-05,
+      "loss": 0.9243,
       "step": 17482
     },
     {
-      "epoch": 0.4801570953832633,
+      "epoch": 0.4961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.1126400198392762e-05,
-      "loss": 0.7973,
+      "learning_rate": 1.0611723831779756e-05,
+      "loss": 0.7891,
       "step": 17483
     },
     {
-      "epoch": 0.4801845596111065,
+      "epoch": 0.49614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.1125516338927788e-05,
-      "loss": 0.8112,
+      "learning_rate": 1.0610806472363568e-05,
+      "loss": 0.7665,
       "step": 17484
     },
     {
-      "epoch": 0.4802120238389498,
+      "epoch": 0.4961691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.11246324705573e-05,
-      "loss": 0.8804,
+      "learning_rate": 1.060988910778787e-05,
+      "loss": 0.8056,
       "step": 17485
     },
     {
-      "epoch": 0.480239488066793,
+      "epoch": 0.49619750283768443,
       "grad_norm": 0.0,
-      "learning_rate": 1.1123748593288278e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.0608971738060405e-05,
+      "loss": 0.9095,
       "step": 17486
     },
     {
-      "epoch": 0.48026695229463623,
+      "epoch": 0.49622587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 1.112286470712772e-05,
-      "loss": 0.9498,
+      "learning_rate": 1.0608054363188934e-05,
+      "loss": 0.8696,
       "step": 17487
     },
     {
-      "epoch": 0.48029441652247945,
+      "epoch": 0.4962542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.1121980812082626e-05,
-      "loss": 0.8723,
+      "learning_rate": 1.06071369831812e-05,
+      "loss": 0.8684,
       "step": 17488
     },
     {
-      "epoch": 0.4803218807503227,
+      "epoch": 0.4962826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.1121096908159985e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.0606219598044952e-05,
+      "loss": 0.9961,
       "step": 17489
     },
     {
-      "epoch": 0.48034934497816595,
+      "epoch": 0.49631101021566404,
       "grad_norm": 0.0,
-      "learning_rate": 1.112021299536679e-05,
-      "loss": 0.9043,
+      "learning_rate": 1.060530220778794e-05,
+      "loss": 0.9049,
       "step": 17490
     },
     {
-      "epoch": 0.48037680920600917,
+      "epoch": 0.4963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.1119329073710037e-05,
-      "loss": 0.9158,
+      "learning_rate": 1.0604384812417912e-05,
+      "loss": 0.8524,
       "step": 17491
     },
     {
-      "epoch": 0.4804042734338524,
+      "epoch": 0.4963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.1118445143196715e-05,
-      "loss": 0.9658,
+      "learning_rate": 1.0603467411942618e-05,
+      "loss": 0.7971,
       "step": 17492
     },
     {
-      "epoch": 0.48043173766169567,
+      "epoch": 0.4963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.1117561203833827e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.0602550006369809e-05,
+      "loss": 1.0152,
       "step": 17493
     },
     {
-      "epoch": 0.4804592018895389,
+      "epoch": 0.4964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.1116677255628359e-05,
-      "loss": 0.9948,
+      "learning_rate": 1.0601632595707233e-05,
+      "loss": 0.8514,
       "step": 17494
     },
     {
-      "epoch": 0.4804866661173821,
+      "epoch": 0.4964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.1115793298587305e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.060071517996264e-05,
+      "loss": 0.9232,
       "step": 17495
     },
     {
-      "epoch": 0.48051413034522533,
+      "epoch": 0.49648127128263336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1114909332717664e-05,
-      "loss": 0.8976,
+      "learning_rate": 1.0599797759143775e-05,
+      "loss": 0.8208,
       "step": 17496
     },
     {
-      "epoch": 0.48054159457306855,
+      "epoch": 0.4965096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.1114025358026435e-05,
-      "loss": 0.9499,
+      "learning_rate": 1.0598880333258394e-05,
+      "loss": 0.8249,
       "step": 17497
     },
     {
-      "epoch": 0.48056905880091183,
+      "epoch": 0.49653802497162314,
       "grad_norm": 0.0,
-      "learning_rate": 1.1113141374520599e-05,
-      "loss": 1.0544,
+      "learning_rate": 1.0597962902314248e-05,
+      "loss": 0.7915,
       "step": 17498
     },
     {
-      "epoch": 0.48059652302875505,
+      "epoch": 0.49656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.1112257382207157e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.0597045466319076e-05,
+      "loss": 0.9446,
       "step": 17499
     },
     {
-      "epoch": 0.4806239872565983,
+      "epoch": 0.4965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 1.1111373381093107e-05,
-      "loss": 0.8703,
+      "learning_rate": 1.0596128025280637e-05,
+      "loss": 0.8212,
       "step": 17500
     },
     {
-      "epoch": 0.4806514514844415,
+      "epoch": 0.4966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.111048937118544e-05,
-      "loss": 0.8467,
+      "learning_rate": 1.0595210579206676e-05,
+      "loss": 0.9082,
       "step": 17501
     },
     {
-      "epoch": 0.48067891571228477,
+      "epoch": 0.49665153234960274,
       "grad_norm": 0.0,
-      "learning_rate": 1.1109605352491148e-05,
-      "loss": 0.8764,
+      "learning_rate": 1.0594293128104947e-05,
+      "loss": 0.9115,
       "step": 17502
     },
     {
-      "epoch": 0.480706379940128,
+      "epoch": 0.4966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.1108721325017231e-05,
-      "loss": 1.0196,
+      "learning_rate": 1.0593375671983195e-05,
+      "loss": 0.8291,
       "step": 17503
     },
     {
-      "epoch": 0.4807338441679712,
+      "epoch": 0.4967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.110783728877068e-05,
-      "loss": 0.9374,
+      "learning_rate": 1.0592458210849175e-05,
+      "loss": 0.8715,
       "step": 17504
     },
     {
-      "epoch": 0.48076130839581444,
+      "epoch": 0.4967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.1106953243758493e-05,
-      "loss": 1.0284,
+      "learning_rate": 1.0591540744710633e-05,
+      "loss": 1.0117,
       "step": 17505
     },
     {
-      "epoch": 0.4807887726236577,
+      "epoch": 0.4967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.1106069189987663e-05,
-      "loss": 0.868,
+      "learning_rate": 1.0590623273575317e-05,
+      "loss": 0.9682,
       "step": 17506
     },
     {
-      "epoch": 0.48081623685150093,
+      "epoch": 0.4967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.1105185127465183e-05,
-      "loss": 0.9426,
+      "learning_rate": 1.0589705797450983e-05,
+      "loss": 0.8602,
       "step": 17507
     },
     {
-      "epoch": 0.48084370107934415,
+      "epoch": 0.49682179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 1.110430105619805e-05,
-      "loss": 0.9624,
+      "learning_rate": 1.0588788316345377e-05,
+      "loss": 0.9501,
       "step": 17508
     },
     {
-      "epoch": 0.4808711653071874,
+      "epoch": 0.496850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.1103416976193264e-05,
-      "loss": 0.9153,
+      "learning_rate": 1.0587870830266247e-05,
+      "loss": 0.875,
       "step": 17509
     },
     {
-      "epoch": 0.4808986295350306,
+      "epoch": 0.49687854710556184,
       "grad_norm": 0.0,
-      "learning_rate": 1.1102532887457812e-05,
-      "loss": 0.8475,
+      "learning_rate": 1.0586953339221346e-05,
+      "loss": 0.8641,
       "step": 17510
     },
     {
-      "epoch": 0.4809260937628739,
+      "epoch": 0.49690692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.1101648789998692e-05,
-      "loss": 0.9515,
+      "learning_rate": 1.0586035843218427e-05,
+      "loss": 0.8058,
       "step": 17511
     },
     {
-      "epoch": 0.4809535579907171,
+      "epoch": 0.49693530079455167,
       "grad_norm": 0.0,
-      "learning_rate": 1.1100764683822902e-05,
-      "loss": 0.904,
+      "learning_rate": 1.0585118342265235e-05,
+      "loss": 0.9603,
       "step": 17512
     },
     {
-      "epoch": 0.4809810222185603,
+      "epoch": 0.49696367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 1.1099880568937435e-05,
-      "loss": 0.8011,
+      "learning_rate": 1.058420083636952e-05,
+      "loss": 0.842,
       "step": 17513
     },
     {
-      "epoch": 0.48100848644640354,
+      "epoch": 0.49699205448354145,
       "grad_norm": 0.0,
-      "learning_rate": 1.1098996445349286e-05,
-      "loss": 0.8697,
+      "learning_rate": 1.0583283325539037e-05,
+      "loss": 0.936,
       "step": 17514
     },
     {
-      "epoch": 0.4810359506742468,
+      "epoch": 0.4970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.1098112313065452e-05,
-      "loss": 0.8417,
+      "learning_rate": 1.0582365809781536e-05,
+      "loss": 0.8769,
       "step": 17515
     },
     {
-      "epoch": 0.48106341490209004,
+      "epoch": 0.4970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.1097228172092927e-05,
-      "loss": 0.9804,
+      "learning_rate": 1.0581448289104759e-05,
+      "loss": 0.874,
       "step": 17516
     },
     {
-      "epoch": 0.48109087912993326,
+      "epoch": 0.4970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.1096344022438707e-05,
-      "loss": 1.0244,
+      "learning_rate": 1.0580530763516466e-05,
+      "loss": 0.8123,
       "step": 17517
     },
     {
-      "epoch": 0.4811183433577765,
+      "epoch": 0.497105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.1095459864109792e-05,
-      "loss": 0.8568,
+      "learning_rate": 1.0579613233024402e-05,
+      "loss": 0.9305,
       "step": 17518
     },
     {
-      "epoch": 0.48114580758561976,
+      "epoch": 0.4971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.109457569711317e-05,
-      "loss": 0.8249,
+      "learning_rate": 1.057869569763632e-05,
+      "loss": 0.7893,
       "step": 17519
     },
     {
-      "epoch": 0.481173271813463,
+      "epoch": 0.49716231555051077,
       "grad_norm": 0.0,
-      "learning_rate": 1.1093691521455845e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.0577778157359969e-05,
+      "loss": 0.9519,
       "step": 17520
     },
     {
-      "epoch": 0.4812007360413062,
+      "epoch": 0.4971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.109280733714481e-05,
-      "loss": 0.8831,
+      "learning_rate": 1.05768606122031e-05,
+      "loss": 0.9591,
       "step": 17521
     },
     {
-      "epoch": 0.4812282002691494,
+      "epoch": 0.49721906923950054,
       "grad_norm": 0.0,
-      "learning_rate": 1.1091923144187053e-05,
-      "loss": 1.0063,
+      "learning_rate": 1.0575943062173464e-05,
+      "loss": 0.8288,
       "step": 17522
     },
     {
-      "epoch": 0.48125566449699264,
+      "epoch": 0.49724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.1091038942589582e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.057502550727881e-05,
+      "loss": 0.918,
       "step": 17523
     },
     {
-      "epoch": 0.4812831287248359,
+      "epoch": 0.4972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.1090154732359389e-05,
-      "loss": 0.9189,
+      "learning_rate": 1.0574107947526888e-05,
+      "loss": 0.9733,
       "step": 17524
     },
     {
-      "epoch": 0.48131059295267914,
+      "epoch": 0.49730419977298523,
       "grad_norm": 0.0,
-      "learning_rate": 1.1089270513503467e-05,
-      "loss": 0.8263,
+      "learning_rate": 1.0573190382925454e-05,
+      "loss": 0.8064,
       "step": 17525
     },
     {
-      "epoch": 0.48133805718052236,
+      "epoch": 0.49733257661748015,
       "grad_norm": 0.0,
-      "learning_rate": 1.1088386286028819e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.0572272813482254e-05,
+      "loss": 0.8976,
       "step": 17526
     },
     {
-      "epoch": 0.4813655214083656,
+      "epoch": 0.497360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.1087502049942432e-05,
-      "loss": 0.942,
+      "learning_rate": 1.0571355239205037e-05,
+      "loss": 0.8111,
       "step": 17527
     },
     {
-      "epoch": 0.48139298563620886,
+      "epoch": 0.4973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.1086617805251307e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.057043766010156e-05,
+      "loss": 0.8131,
       "step": 17528
     },
     {
-      "epoch": 0.4814204498640521,
+      "epoch": 0.49741770715096484,
       "grad_norm": 0.0,
-      "learning_rate": 1.1085733551962445e-05,
-      "loss": 0.9715,
+      "learning_rate": 1.0569520076179569e-05,
+      "loss": 0.9375,
       "step": 17529
     },
     {
-      "epoch": 0.4814479140918953,
+      "epoch": 0.4974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.1084849290082838e-05,
-      "loss": 0.9045,
+      "learning_rate": 1.0568602487446817e-05,
+      "loss": 0.8814,
       "step": 17530
     },
     {
-      "epoch": 0.4814753783197385,
+      "epoch": 0.4974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.108396501961948e-05,
-      "loss": 0.9144,
+      "learning_rate": 1.0567684893911054e-05,
+      "loss": 0.8952,
       "step": 17531
     },
     {
-      "epoch": 0.4815028425475818,
+      "epoch": 0.4975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.1083080740579375e-05,
-      "loss": 0.989,
+      "learning_rate": 1.056676729558003e-05,
+      "loss": 0.767,
       "step": 17532
     },
     {
-      "epoch": 0.481530306775425,
+      "epoch": 0.4975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.1082196452969513e-05,
-      "loss": 0.9099,
+      "learning_rate": 1.0565849692461497e-05,
+      "loss": 0.8863,
       "step": 17533
     },
     {
-      "epoch": 0.48155777100326824,
+      "epoch": 0.49755959137343925,
       "grad_norm": 0.0,
-      "learning_rate": 1.1081312156796894e-05,
-      "loss": 0.9388,
+      "learning_rate": 1.056493208456321e-05,
+      "loss": 0.8694,
       "step": 17534
     },
     {
-      "epoch": 0.48158523523111146,
+      "epoch": 0.49758796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.1080427852068516e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.0564014471892911e-05,
+      "loss": 0.9236,
       "step": 17535
     },
     {
-      "epoch": 0.4816126994589547,
+      "epoch": 0.4976163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.1079543538791372e-05,
-      "loss": 1.0034,
+      "learning_rate": 1.0563096854458361e-05,
+      "loss": 0.9843,
       "step": 17536
     },
     {
-      "epoch": 0.48164016368679796,
+      "epoch": 0.49764472190692394,
       "grad_norm": 0.0,
-      "learning_rate": 1.1078659216972462e-05,
-      "loss": 0.9301,
+      "learning_rate": 1.0562179232267304e-05,
+      "loss": 0.9036,
       "step": 17537
     },
     {
-      "epoch": 0.4816676279146412,
+      "epoch": 0.49767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.1077774886618784e-05,
-      "loss": 1.0634,
+      "learning_rate": 1.0561261605327495e-05,
+      "loss": 0.9176,
       "step": 17538
     },
     {
-      "epoch": 0.4816950921424844,
+      "epoch": 0.4977014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.1076890547737331e-05,
-      "loss": 0.9622,
+      "learning_rate": 1.0560343973646686e-05,
+      "loss": 0.7946,
       "step": 17539
     },
     {
-      "epoch": 0.4817225563703276,
+      "epoch": 0.4977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.1076006200335103e-05,
-      "loss": 0.806,
+      "learning_rate": 1.055942633723262e-05,
+      "loss": 1.0284,
       "step": 17540
     },
     {
-      "epoch": 0.4817500205981709,
+      "epoch": 0.49775822928490354,
       "grad_norm": 0.0,
-      "learning_rate": 1.10751218444191e-05,
-      "loss": 0.8207,
+      "learning_rate": 1.0558508696093058e-05,
+      "loss": 0.8321,
       "step": 17541
     },
     {
-      "epoch": 0.4817774848260141,
+      "epoch": 0.4977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.1074237479996315e-05,
-      "loss": 0.9315,
+      "learning_rate": 1.055759105023575e-05,
+      "loss": 0.8813,
       "step": 17542
     },
     {
-      "epoch": 0.48180494905385735,
+      "epoch": 0.4978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.1073353107073746e-05,
-      "loss": 1.0109,
+      "learning_rate": 1.0556673399668442e-05,
+      "loss": 0.9279,
       "step": 17543
     },
     {
-      "epoch": 0.48183241328170057,
+      "epoch": 0.4978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.1072468725658391e-05,
-      "loss": 0.8585,
+      "learning_rate": 1.055575574439889e-05,
+      "loss": 0.9679,
       "step": 17544
     },
     {
-      "epoch": 0.48185987750954384,
+      "epoch": 0.4978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.107158433575725e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.0554838084434846e-05,
+      "loss": 0.8612,
       "step": 17545
     },
     {
-      "epoch": 0.48188734173738706,
+      "epoch": 0.497900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.1070699937377318e-05,
-      "loss": 0.8887,
+      "learning_rate": 1.0553920419784056e-05,
+      "loss": 0.9428,
       "step": 17546
     },
     {
-      "epoch": 0.4819148059652303,
+      "epoch": 0.49792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.1069815530525595e-05,
-      "loss": 0.9787,
+      "learning_rate": 1.0553002750454277e-05,
+      "loss": 0.9833,
       "step": 17547
     },
     {
-      "epoch": 0.4819422701930735,
+      "epoch": 0.4979568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.1068931115209071e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.055208507645326e-05,
+      "loss": 0.9489,
       "step": 17548
     },
     {
-      "epoch": 0.48196973442091673,
+      "epoch": 0.49798524404086264,
       "grad_norm": 0.0,
-      "learning_rate": 1.1068046691434757e-05,
-      "loss": 0.8178,
+      "learning_rate": 1.0551167397788757e-05,
+      "loss": 0.9118,
       "step": 17549
     },
     {
-      "epoch": 0.48199719864876,
+      "epoch": 0.49801362088535756,
       "grad_norm": 0.0,
-      "learning_rate": 1.106716225920964e-05,
-      "loss": 0.8544,
+      "learning_rate": 1.0550249714468515e-05,
+      "loss": 0.9262,
       "step": 17550
     },
     {
-      "epoch": 0.4820246628766032,
+      "epoch": 0.4980419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.1066277818540722e-05,
-      "loss": 0.9606,
+      "learning_rate": 1.0549332026500291e-05,
+      "loss": 0.8842,
       "step": 17551
     },
     {
-      "epoch": 0.48205212710444645,
+      "epoch": 0.49807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.1065393369435002e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.0548414333891835e-05,
+      "loss": 0.979,
       "step": 17552
     },
     {
-      "epoch": 0.48207959133228967,
+      "epoch": 0.49809875141884224,
       "grad_norm": 0.0,
-      "learning_rate": 1.1064508911899478e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.0547496636650896e-05,
+      "loss": 0.8594,
       "step": 17553
     },
     {
-      "epoch": 0.48210705556013295,
+      "epoch": 0.4981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.1063624445941145e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.054657893478523e-05,
+      "loss": 0.8864,
       "step": 17554
     },
     {
-      "epoch": 0.48213451978797617,
+      "epoch": 0.498155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.1062739971567004e-05,
-      "loss": 0.812,
+      "learning_rate": 1.0545661228302586e-05,
+      "loss": 0.8715,
       "step": 17555
     },
     {
-      "epoch": 0.4821619840158194,
+      "epoch": 0.4981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.1061855488784054e-05,
-      "loss": 0.8674,
+      "learning_rate": 1.0544743517210718e-05,
+      "loss": 0.8594,
       "step": 17556
     },
     {
-      "epoch": 0.4821894482436626,
+      "epoch": 0.4982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.1060970997599292e-05,
-      "loss": 0.8564,
+      "learning_rate": 1.0543825801517375e-05,
+      "loss": 0.9692,
       "step": 17557
     },
     {
-      "epoch": 0.4822169124715059,
+      "epoch": 0.4982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.1060086498019712e-05,
-      "loss": 0.8763,
+      "learning_rate": 1.0542908081230314e-05,
+      "loss": 0.896,
       "step": 17558
     },
     {
-      "epoch": 0.4822443766993491,
+      "epoch": 0.49826901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 1.1059201990052321e-05,
-      "loss": 0.8436,
+      "learning_rate": 1.0541990356357285e-05,
+      "loss": 0.9982,
       "step": 17559
     },
     {
-      "epoch": 0.48227184092719233,
+      "epoch": 0.4982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.1058317473704115e-05,
-      "loss": 0.8311,
+      "learning_rate": 1.0541072626906035e-05,
+      "loss": 0.9108,
       "step": 17560
     },
     {
-      "epoch": 0.48229930515503555,
+      "epoch": 0.49832576617480134,
       "grad_norm": 0.0,
-      "learning_rate": 1.105743294898209e-05,
-      "loss": 0.8609,
+      "learning_rate": 1.0540154892884325e-05,
+      "loss": 0.9933,
       "step": 17561
     },
     {
-      "epoch": 0.4823267693828788,
+      "epoch": 0.49835414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.1056548415893244e-05,
-      "loss": 0.8487,
+      "learning_rate": 1.05392371542999e-05,
+      "loss": 0.788,
       "step": 17562
     },
     {
-      "epoch": 0.48235423361072205,
+      "epoch": 0.4983825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 1.1055663874444577e-05,
-      "loss": 0.9284,
+      "learning_rate": 1.0538319411160512e-05,
+      "loss": 0.8273,
       "step": 17563
     },
     {
-      "epoch": 0.48238169783856527,
+      "epoch": 0.49841089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.105477932464309e-05,
-      "loss": 0.804,
+      "learning_rate": 1.0537401663473916e-05,
+      "loss": 0.8457,
       "step": 17564
     },
     {
-      "epoch": 0.4824091620664085,
+      "epoch": 0.49843927355278095,
       "grad_norm": 0.0,
-      "learning_rate": 1.1053894766495784e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.0536483911247869e-05,
+      "loss": 0.8435,
       "step": 17565
     },
     {
-      "epoch": 0.4824366262942517,
+      "epoch": 0.4984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.1053010200009649e-05,
-      "loss": 0.9393,
+      "learning_rate": 1.0535566154490116e-05,
+      "loss": 0.9382,
       "step": 17566
     },
     {
-      "epoch": 0.482464090522095,
+      "epoch": 0.4984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.105212562519169e-05,
-      "loss": 0.9072,
+      "learning_rate": 1.0534648393208409e-05,
+      "loss": 0.848,
       "step": 17567
     },
     {
-      "epoch": 0.4824915547499382,
+      "epoch": 0.4985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.1051241042048908e-05,
-      "loss": 0.9226,
+      "learning_rate": 1.0533730627410505e-05,
+      "loss": 0.9169,
       "step": 17568
     },
     {
-      "epoch": 0.48251901897778143,
+      "epoch": 0.4985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.10503564505883e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.0532812857104155e-05,
+      "loss": 0.9089,
       "step": 17569
     },
     {
-      "epoch": 0.48254648320562465,
+      "epoch": 0.4985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.1049471850816862e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.0531895082297107e-05,
+      "loss": 0.7453,
       "step": 17570
     },
     {
-      "epoch": 0.48257394743346793,
+      "epoch": 0.49860953461975027,
       "grad_norm": 0.0,
-      "learning_rate": 1.10485872427416e-05,
-      "loss": 0.9609,
+      "learning_rate": 1.0530977302997121e-05,
+      "loss": 0.9679,
       "step": 17571
     },
     {
-      "epoch": 0.48260141166131115,
+      "epoch": 0.4986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.1047702626369506e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.0530059519211946e-05,
+      "loss": 0.9449,
       "step": 17572
     },
     {
-      "epoch": 0.4826288758891544,
+      "epoch": 0.49866628830874005,
       "grad_norm": 0.0,
-      "learning_rate": 1.1046818001707584e-05,
-      "loss": 0.9473,
+      "learning_rate": 1.0529141730949334e-05,
+      "loss": 0.9456,
       "step": 17573
     },
     {
-      "epoch": 0.4826563401169976,
+      "epoch": 0.49869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.1045933368762833e-05,
-      "loss": 0.8515,
+      "learning_rate": 1.0528223938217037e-05,
+      "loss": 1.0226,
       "step": 17574
     },
     {
-      "epoch": 0.48268380434484087,
+      "epoch": 0.4987230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.1045048727542251e-05,
-      "loss": 0.7984,
+      "learning_rate": 1.0527306141022808e-05,
+      "loss": 0.8057,
       "step": 17575
     },
     {
-      "epoch": 0.4827112685726841,
+      "epoch": 0.49875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 1.1044164078052837e-05,
-      "loss": 0.8488,
+      "learning_rate": 1.0526388339374403e-05,
+      "loss": 0.8502,
       "step": 17576
     },
     {
-      "epoch": 0.4827387328005273,
+      "epoch": 0.49877979568671965,
       "grad_norm": 0.0,
-      "learning_rate": 1.1043279420301598e-05,
-      "loss": 0.7602,
+      "learning_rate": 1.052547053327957e-05,
+      "loss": 0.9344,
       "step": 17577
     },
     {
-      "epoch": 0.48276619702837054,
+      "epoch": 0.4988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.1042394754295524e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.0524552722746063e-05,
+      "loss": 0.7621,
       "step": 17578
     },
     {
-      "epoch": 0.48279366125621376,
+      "epoch": 0.4988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.104151008004162e-05,
-      "loss": 0.9506,
+      "learning_rate": 1.0523634907781637e-05,
+      "loss": 0.8437,
       "step": 17579
     },
     {
-      "epoch": 0.48282112548405703,
+      "epoch": 0.4988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.1040625397546885e-05,
-      "loss": 0.9413,
+      "learning_rate": 1.0522717088394046e-05,
+      "loss": 0.9944,
       "step": 17580
     },
     {
-      "epoch": 0.48284858971190026,
+      "epoch": 0.4988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.1039740706818318e-05,
-      "loss": 0.892,
+      "learning_rate": 1.0521799264591035e-05,
+      "loss": 0.8463,
       "step": 17581
     },
     {
-      "epoch": 0.4828760539397435,
+      "epoch": 0.4989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.1038856007862922e-05,
-      "loss": 0.8511,
+      "learning_rate": 1.0520881436380366e-05,
+      "loss": 0.9875,
       "step": 17582
     },
     {
-      "epoch": 0.4829035181675867,
+      "epoch": 0.498950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.1037971300687691e-05,
-      "loss": 0.9957,
+      "learning_rate": 1.0519963603769787e-05,
+      "loss": 0.9923,
       "step": 17583
     },
     {
-      "epoch": 0.48293098239543,
+      "epoch": 0.4989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.1037086585299627e-05,
-      "loss": 0.8168,
+      "learning_rate": 1.0519045766767052e-05,
+      "loss": 0.9023,
       "step": 17584
     },
     {
-      "epoch": 0.4829584466232732,
+      "epoch": 0.49900681044267875,
       "grad_norm": 0.0,
-      "learning_rate": 1.103620186170574e-05,
-      "loss": 0.9257,
+      "learning_rate": 1.0518127925379914e-05,
+      "loss": 0.8259,
       "step": 17585
     },
     {
-      "epoch": 0.4829859108511164,
+      "epoch": 0.49903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.1035317129913018e-05,
-      "loss": 0.8622,
+      "learning_rate": 1.0517210079616126e-05,
+      "loss": 0.8416,
       "step": 17586
     },
     {
-      "epoch": 0.48301337507895964,
+      "epoch": 0.4990635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.1034432389928464e-05,
-      "loss": 1.0065,
+      "learning_rate": 1.0516292229483437e-05,
+      "loss": 1.0143,
       "step": 17587
     },
     {
-      "epoch": 0.4830408393068029,
+      "epoch": 0.49909194097616344,
       "grad_norm": 0.0,
-      "learning_rate": 1.1033547641759085e-05,
-      "loss": 0.9091,
+      "learning_rate": 1.051537437498961e-05,
+      "loss": 0.8994,
       "step": 17588
     },
     {
-      "epoch": 0.48306830353464614,
+      "epoch": 0.49912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.1032662885411876e-05,
-      "loss": 0.8931,
+      "learning_rate": 1.0514456516142393e-05,
+      "loss": 0.9307,
       "step": 17589
     },
     {
-      "epoch": 0.48309576776248936,
+      "epoch": 0.4991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.1031778120893834e-05,
-      "loss": 0.8356,
+      "learning_rate": 1.0513538652949538e-05,
+      "loss": 0.91,
       "step": 17590
     },
     {
-      "epoch": 0.4831232319903326,
+      "epoch": 0.49917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.1030893348211969e-05,
-      "loss": 0.8678,
+      "learning_rate": 1.0512620785418796e-05,
+      "loss": 0.89,
       "step": 17591
     },
     {
-      "epoch": 0.4831506962181758,
+      "epoch": 0.49920544835414304,
       "grad_norm": 0.0,
-      "learning_rate": 1.1030008567373271e-05,
-      "loss": 0.8255,
+      "learning_rate": 1.0511702913557926e-05,
+      "loss": 0.8425,
       "step": 17592
     },
     {
-      "epoch": 0.4831781604460191,
+      "epoch": 0.4992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.102912377838475e-05,
-      "loss": 0.937,
+      "learning_rate": 1.0510785037374675e-05,
+      "loss": 0.9183,
       "step": 17593
     },
     {
-      "epoch": 0.4832056246738623,
+      "epoch": 0.4992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.1028238981253403e-05,
-      "loss": 0.874,
+      "learning_rate": 1.0509867156876803e-05,
+      "loss": 0.8791,
       "step": 17594
     },
     {
-      "epoch": 0.4832330889017055,
+      "epoch": 0.4992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.102735417598623e-05,
-      "loss": 0.853,
+      "learning_rate": 1.0508949272072059e-05,
+      "loss": 0.9096,
       "step": 17595
     },
     {
-      "epoch": 0.48326055312954874,
+      "epoch": 0.4993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.1026469362590231e-05,
+      "learning_rate": 1.05080313829682e-05,
       "loss": 0.8855,
       "step": 17596
     },
     {
-      "epoch": 0.483288017357392,
+      "epoch": 0.49934733257661745,
       "grad_norm": 0.0,
-      "learning_rate": 1.1025584541072412e-05,
-      "loss": 0.9616,
+      "learning_rate": 1.0507113489572974e-05,
+      "loss": 0.9285,
       "step": 17597
     },
     {
-      "epoch": 0.48331548158523524,
+      "epoch": 0.49937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.1024699711439771e-05,
-      "loss": 0.8761,
+      "learning_rate": 1.0506195591894139e-05,
+      "loss": 0.9733,
       "step": 17598
     },
     {
-      "epoch": 0.48334294581307846,
+      "epoch": 0.4994040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.1023814873699306e-05,
-      "loss": 0.8204,
+      "learning_rate": 1.0505277689939448e-05,
+      "loss": 0.9233,
       "step": 17599
     },
     {
-      "epoch": 0.4833704100409217,
+      "epoch": 0.49943246311010214,
       "grad_norm": 0.0,
-      "learning_rate": 1.1022930027858024e-05,
-      "loss": 1.0172,
+      "learning_rate": 1.0504359783716652e-05,
+      "loss": 0.8736,
       "step": 17600
     },
     {
-      "epoch": 0.48339787426876496,
+      "epoch": 0.49946083995459706,
       "grad_norm": 0.0,
-      "learning_rate": 1.1022045173922922e-05,
-      "loss": 0.9451,
+      "learning_rate": 1.0503441873233505e-05,
+      "loss": 0.9289,
       "step": 17601
     },
     {
-      "epoch": 0.4834253384966082,
+      "epoch": 0.4994892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 1.1021160311901e-05,
-      "loss": 0.9732,
+      "learning_rate": 1.0502523958497763e-05,
+      "loss": 0.9047,
       "step": 17602
     },
     {
-      "epoch": 0.4834528027244514,
+      "epoch": 0.49951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.1020275441799267e-05,
-      "loss": 1.0397,
+      "learning_rate": 1.050160603951718e-05,
+      "loss": 0.866,
       "step": 17603
     },
     {
-      "epoch": 0.4834802669522946,
+      "epoch": 0.49954597048808175,
       "grad_norm": 0.0,
-      "learning_rate": 1.1019390563624714e-05,
-      "loss": 0.8825,
+      "learning_rate": 1.0500688116299507e-05,
+      "loss": 0.902,
       "step": 17604
     },
     {
-      "epoch": 0.48350773118013785,
+      "epoch": 0.4995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.1018505677384352e-05,
-      "loss": 0.9197,
+      "learning_rate": 1.0499770188852501e-05,
+      "loss": 0.7764,
       "step": 17605
     },
     {
-      "epoch": 0.4835351954079811,
+      "epoch": 0.4996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.101762078308518e-05,
-      "loss": 1.0104,
+      "learning_rate": 1.0498852257183912e-05,
+      "loss": 0.7925,
       "step": 17606
     },
     {
-      "epoch": 0.48356265963582434,
+      "epoch": 0.4996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.1016735880734193e-05,
-      "loss": 0.9761,
+      "learning_rate": 1.0497934321301492e-05,
+      "loss": 0.9339,
       "step": 17607
     },
     {
-      "epoch": 0.48359012386366756,
+      "epoch": 0.4996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.10158509703384e-05,
-      "loss": 0.9154,
+      "learning_rate": 1.0497016381213006e-05,
+      "loss": 0.9361,
       "step": 17608
     },
     {
-      "epoch": 0.4836175880915108,
+      "epoch": 0.4996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.1014966051904802e-05,
-      "loss": 0.9061,
+      "learning_rate": 1.0496098436926195e-05,
+      "loss": 0.8608,
       "step": 17609
     },
     {
-      "epoch": 0.48364505231935406,
+      "epoch": 0.49971623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.10140811254404e-05,
-      "loss": 0.7865,
+      "learning_rate": 1.049518048844882e-05,
+      "loss": 0.8496,
       "step": 17610
     },
     {
-      "epoch": 0.4836725165471973,
+      "epoch": 0.499744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.1013196190952193e-05,
-      "loss": 0.9346,
+      "learning_rate": 1.049426253578863e-05,
+      "loss": 0.8969,
       "step": 17611
     },
     {
-      "epoch": 0.4836999807750405,
+      "epoch": 0.49977298524404085,
       "grad_norm": 0.0,
-      "learning_rate": 1.1012311248447187e-05,
-      "loss": 0.8648,
+      "learning_rate": 1.0493344578953386e-05,
+      "loss": 0.7622,
       "step": 17612
     },
     {
-      "epoch": 0.4837274450028837,
+      "epoch": 0.49980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.1011426297932384e-05,
-      "loss": 0.9344,
+      "learning_rate": 1.0492426617950838e-05,
+      "loss": 0.7828,
       "step": 17613
     },
     {
-      "epoch": 0.483754909230727,
+      "epoch": 0.4998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.101054133941478e-05,
-      "loss": 0.9227,
+      "learning_rate": 1.0491508652788737e-05,
+      "loss": 0.893,
       "step": 17614
     },
     {
-      "epoch": 0.4837823734585702,
+      "epoch": 0.49985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.1009656372901383e-05,
-      "loss": 0.8839,
+      "learning_rate": 1.0490590683474844e-05,
+      "loss": 0.8463,
       "step": 17615
     },
     {
-      "epoch": 0.48380983768641345,
+      "epoch": 0.49988649262202045,
       "grad_norm": 0.0,
-      "learning_rate": 1.1008771398399193e-05,
-      "loss": 0.8146,
+      "learning_rate": 1.0489672710016907e-05,
+      "loss": 0.9417,
       "step": 17616
     },
     {
-      "epoch": 0.48383730191425667,
+      "epoch": 0.4999148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.1007886415915214e-05,
-      "loss": 0.9142,
+      "learning_rate": 1.0488754732422684e-05,
+      "loss": 0.8276,
       "step": 17617
     },
     {
-      "epoch": 0.4838647661420999,
+      "epoch": 0.4999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.100700142545645e-05,
-      "loss": 1.0508,
+      "learning_rate": 1.0487836750699924e-05,
+      "loss": 0.9465,
       "step": 17618
     },
     {
-      "epoch": 0.48389223036994317,
+      "epoch": 0.4999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.1006116427029895e-05,
-      "loss": 0.9366,
+      "learning_rate": 1.0486918764856391e-05,
+      "loss": 0.8995,
       "step": 17619
     },
     {
-      "epoch": 0.4839196945977864,
+      "epoch": 0.5,
       "grad_norm": 0.0,
-      "learning_rate": 1.100523142064256e-05,
-      "loss": 0.8618,
+      "learning_rate": 1.0486000774899832e-05,
+      "loss": 1.0087,
       "step": 17620
     },
     {
-      "epoch": 0.4839471588256296,
+      "epoch": 0.5000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.1004346406301445e-05,
-      "loss": 0.9496,
+      "learning_rate": 1.0485082780837998e-05,
+      "loss": 0.9397,
       "step": 17621
     },
     {
-      "epoch": 0.48397462305347283,
+      "epoch": 0.5000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.1003461384013555e-05,
-      "loss": 0.906,
+      "learning_rate": 1.0484164782678654e-05,
+      "loss": 0.9756,
       "step": 17622
     },
     {
-      "epoch": 0.4840020872813161,
+      "epoch": 0.5000851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.1002576353785884e-05,
-      "loss": 0.9006,
+      "learning_rate": 1.0483246780429546e-05,
+      "loss": 0.9634,
       "step": 17623
     },
     {
-      "epoch": 0.48402955150915933,
+      "epoch": 0.5001135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.1001691315625444e-05,
-      "loss": 0.8395,
+      "learning_rate": 1.048232877409843e-05,
+      "loss": 0.7934,
       "step": 17624
     },
     {
-      "epoch": 0.48405701573700255,
+      "epoch": 0.5001418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.1000806269539234e-05,
-      "loss": 0.9244,
+      "learning_rate": 1.0481410763693059e-05,
+      "loss": 1.0346,
       "step": 17625
     },
     {
-      "epoch": 0.48408447996484577,
+      "epoch": 0.5001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.0999921215534256e-05,
-      "loss": 0.8803,
+      "learning_rate": 1.0480492749221191e-05,
+      "loss": 0.8528,
       "step": 17626
     },
     {
-      "epoch": 0.48411194419268905,
+      "epoch": 0.5001986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.0999036153617515e-05,
-      "loss": 0.8695,
+      "learning_rate": 1.0479574730690583e-05,
+      "loss": 0.9987,
       "step": 17627
     },
     {
-      "epoch": 0.48413940842053227,
+      "epoch": 0.5002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.099815108379601e-05,
-      "loss": 0.9265,
+      "learning_rate": 1.0478656708108981e-05,
+      "loss": 0.7766,
       "step": 17628
     },
     {
-      "epoch": 0.4841668726483755,
+      "epoch": 0.5002553916004541,
       "grad_norm": 0.0,
-      "learning_rate": 1.0997266006076752e-05,
-      "loss": 0.9028,
+      "learning_rate": 1.0477738681484146e-05,
+      "loss": 0.8528,
       "step": 17629
     },
     {
-      "epoch": 0.4841943368762187,
+      "epoch": 0.5002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.0996380920466739e-05,
-      "loss": 0.9311,
+      "learning_rate": 1.0476820650823834e-05,
+      "loss": 0.9225,
       "step": 17630
     },
     {
-      "epoch": 0.48422180110406193,
+      "epoch": 0.5003121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.0995495826972969e-05,
-      "loss": 0.8614,
+      "learning_rate": 1.047590261613579e-05,
+      "loss": 0.9597,
       "step": 17631
     },
     {
-      "epoch": 0.4842492653319052,
+      "epoch": 0.5003405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 1.0994610725602452e-05,
-      "loss": 0.863,
+      "learning_rate": 1.0474984577427778e-05,
+      "loss": 0.9038,
       "step": 17632
     },
     {
-      "epoch": 0.48427672955974843,
+      "epoch": 0.5003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.0993725616362194e-05,
-      "loss": 0.8992,
+      "learning_rate": 1.0474066534707551e-05,
+      "loss": 0.9277,
       "step": 17633
     },
     {
-      "epoch": 0.48430419378759165,
+      "epoch": 0.5003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.099284049925919e-05,
-      "loss": 0.8752,
+      "learning_rate": 1.0473148487982865e-05,
+      "loss": 0.8277,
       "step": 17634
     },
     {
-      "epoch": 0.4843316580154349,
+      "epoch": 0.5004256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.0991955374300447e-05,
-      "loss": 0.8758,
+      "learning_rate": 1.0472230437261469e-05,
+      "loss": 0.9583,
       "step": 17635
     },
     {
-      "epoch": 0.48435912224327815,
+      "epoch": 0.5004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.099107024149297e-05,
-      "loss": 0.8453,
+      "learning_rate": 1.0471312382551121e-05,
+      "loss": 0.9757,
       "step": 17636
     },
     {
-      "epoch": 0.48438658647112137,
+      "epoch": 0.5004824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.0990185100843763e-05,
-      "loss": 0.8863,
+      "learning_rate": 1.0470394323859579e-05,
+      "loss": 0.9054,
       "step": 17637
     },
     {
-      "epoch": 0.4844140506989646,
+      "epoch": 0.500510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.0989299952359826e-05,
-      "loss": 0.9305,
+      "learning_rate": 1.0469476261194591e-05,
+      "loss": 0.8875,
       "step": 17638
     },
     {
-      "epoch": 0.4844415149268078,
+      "epoch": 0.500539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.0988414796048164e-05,
-      "loss": 0.992,
+      "learning_rate": 1.0468558194563919e-05,
+      "loss": 0.9479,
       "step": 17639
     },
     {
-      "epoch": 0.4844689791546511,
+      "epoch": 0.5005675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 1.0987529631915781e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.0467640123975314e-05,
+      "loss": 0.8224,
       "step": 17640
     },
     {
-      "epoch": 0.4844964433824943,
+      "epoch": 0.5005959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.0986644459969685e-05,
-      "loss": 0.9076,
+      "learning_rate": 1.046672204943653e-05,
+      "loss": 0.886,
       "step": 17641
     },
     {
-      "epoch": 0.48452390761033753,
+      "epoch": 0.5006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.0985759280216873e-05,
-      "loss": 0.9096,
+      "learning_rate": 1.0465803970955327e-05,
+      "loss": 0.8638,
       "step": 17642
     },
     {
-      "epoch": 0.48455137183818076,
+      "epoch": 0.5006526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.098487409266435e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.0464885888539455e-05,
+      "loss": 0.8441,
       "step": 17643
     },
     {
-      "epoch": 0.484578836066024,
+      "epoch": 0.5006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.0983988897319124e-05,
-      "loss": 0.9194,
+      "learning_rate": 1.0463967802196673e-05,
+      "loss": 0.8402,
       "step": 17644
     },
     {
-      "epoch": 0.48460630029386725,
+      "epoch": 0.5007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.0983103694188198e-05,
-      "loss": 1.0367,
+      "learning_rate": 1.0463049711934731e-05,
+      "loss": 0.9103,
       "step": 17645
     },
     {
-      "epoch": 0.4846337645217105,
+      "epoch": 0.5007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.098221848327857e-05,
-      "loss": 0.88,
+      "learning_rate": 1.046213161776139e-05,
+      "loss": 0.8945,
       "step": 17646
     },
     {
-      "epoch": 0.4846612287495537,
+      "epoch": 0.5007661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.0981333264597253e-05,
-      "loss": 0.8544,
+      "learning_rate": 1.0461213519684401e-05,
+      "loss": 0.9028,
       "step": 17647
     },
     {
-      "epoch": 0.4846886929773969,
+      "epoch": 0.500794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.0980448038151245e-05,
-      "loss": 0.9612,
+      "learning_rate": 1.046029541771152e-05,
+      "loss": 0.8369,
       "step": 17648
     },
     {
-      "epoch": 0.4847161572052402,
+      "epoch": 0.5008229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.0979562803947554e-05,
-      "loss": 0.9151,
+      "learning_rate": 1.0459377311850505e-05,
+      "loss": 0.9365,
       "step": 17649
     },
     {
-      "epoch": 0.4847436214330834,
+      "epoch": 0.5008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.0978677561993183e-05,
-      "loss": 0.8963,
+      "learning_rate": 1.0458459202109108e-05,
+      "loss": 0.9303,
       "step": 17650
     },
     {
-      "epoch": 0.48477108566092664,
+      "epoch": 0.5008796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.0977792312295132e-05,
-      "loss": 0.933,
+      "learning_rate": 1.0457541088495085e-05,
+      "loss": 0.8399,
       "step": 17651
     },
     {
-      "epoch": 0.48479854988876986,
+      "epoch": 0.5009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.0976907054860412e-05,
-      "loss": 0.9662,
+      "learning_rate": 1.0456622971016193e-05,
+      "loss": 0.9075,
       "step": 17652
     },
     {
-      "epoch": 0.48482601411661314,
+      "epoch": 0.5009364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.0976021789696025e-05,
-      "loss": 0.934,
+      "learning_rate": 1.0455704849680188e-05,
+      "loss": 0.9604,
       "step": 17653
     },
     {
-      "epoch": 0.48485347834445636,
+      "epoch": 0.5009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.0975136516808977e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.0454786724494819e-05,
+      "loss": 0.8621,
       "step": 17654
     },
     {
-      "epoch": 0.4848809425722996,
+      "epoch": 0.5009931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.0974251236206271e-05,
-      "loss": 0.8989,
+      "learning_rate": 1.0453868595467849e-05,
+      "loss": 0.9207,
       "step": 17655
     },
     {
-      "epoch": 0.4849084068001428,
+      "epoch": 0.5010215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 1.097336594789491e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.045295046260703e-05,
+      "loss": 0.9669,
       "step": 17656
     },
     {
-      "epoch": 0.484935871027986,
+      "epoch": 0.501049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.0972480651881903e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.0452032325920118e-05,
+      "loss": 0.978,
       "step": 17657
     },
     {
-      "epoch": 0.4849633352558293,
+      "epoch": 0.5010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.0971595348174248e-05,
-      "loss": 0.8202,
+      "learning_rate": 1.0451114185414867e-05,
+      "loss": 0.8861,
       "step": 17658
     },
     {
-      "epoch": 0.4849907994836725,
+      "epoch": 0.5011066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.097071003677896e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.045019604109904e-05,
+      "loss": 0.7601,
       "step": 17659
     },
     {
-      "epoch": 0.48501826371151574,
+      "epoch": 0.5011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.0969824717703032e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.0449277892980382e-05,
+      "loss": 0.8655,
       "step": 17660
     },
     {
-      "epoch": 0.48504572793935896,
+      "epoch": 0.5011634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.0968939390953479e-05,
-      "loss": 0.886,
+      "learning_rate": 1.0448359741066653e-05,
+      "loss": 0.9055,
       "step": 17661
     },
     {
-      "epoch": 0.48507319216720224,
+      "epoch": 0.5011918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.0968054056537303e-05,
-      "loss": 0.8875,
+      "learning_rate": 1.044744158536561e-05,
+      "loss": 0.9706,
       "step": 17662
     },
     {
-      "epoch": 0.48510065639504546,
+      "epoch": 0.5012202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.0967168714461506e-05,
-      "loss": 0.8548,
+      "learning_rate": 1.0446523425885008e-05,
+      "loss": 0.9925,
       "step": 17663
     },
     {
-      "epoch": 0.4851281206228887,
+      "epoch": 0.5012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.0966283364733094e-05,
-      "loss": 0.8706,
+      "learning_rate": 1.0445605262632603e-05,
+      "loss": 0.962,
       "step": 17664
     },
     {
-      "epoch": 0.4851555848507319,
+      "epoch": 0.5012769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.0965398007359078e-05,
-      "loss": 0.9384,
+      "learning_rate": 1.044468709561615e-05,
+      "loss": 0.8642,
       "step": 17665
     },
     {
-      "epoch": 0.4851830490785752,
+      "epoch": 0.5013053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.0964512642346459e-05,
-      "loss": 0.8326,
+      "learning_rate": 1.0443768924843405e-05,
+      "loss": 0.8238,
       "step": 17666
     },
     {
-      "epoch": 0.4852105133064184,
+      "epoch": 0.50133371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0963627269702238e-05,
-      "loss": 0.9406,
+      "learning_rate": 1.0442850750322126e-05,
+      "loss": 0.919,
       "step": 17667
     },
     {
-      "epoch": 0.4852379775342616,
+      "epoch": 0.5013620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.0962741889433427e-05,
-      "loss": 0.9316,
+      "learning_rate": 1.0441932572060062e-05,
+      "loss": 0.9322,
       "step": 17668
     },
     {
-      "epoch": 0.48526544176210484,
+      "epoch": 0.5013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.0961856501547033e-05,
-      "loss": 0.8897,
+      "learning_rate": 1.0441014390064978e-05,
+      "loss": 0.8849,
       "step": 17669
     },
     {
-      "epoch": 0.4852929059899481,
+      "epoch": 0.5014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.0960971106050051e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.0440096204344623e-05,
+      "loss": 1.0162,
       "step": 17670
     },
     {
-      "epoch": 0.48532037021779134,
+      "epoch": 0.5014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.0960085702949496e-05,
-      "loss": 0.9128,
+      "learning_rate": 1.043917801490676e-05,
+      "loss": 0.9631,
       "step": 17671
     },
     {
-      "epoch": 0.48534783444563456,
+      "epoch": 0.5014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.0959200292252372e-05,
-      "loss": 0.8269,
+      "learning_rate": 1.0438259821759133e-05,
+      "loss": 0.9195,
       "step": 17672
     },
     {
-      "epoch": 0.4853752986734778,
+      "epoch": 0.5015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.0958314873965683e-05,
-      "loss": 0.9093,
+      "learning_rate": 1.0437341624909512e-05,
+      "loss": 0.9159,
       "step": 17673
     },
     {
-      "epoch": 0.485402762901321,
+      "epoch": 0.5015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.0957429448096437e-05,
-      "loss": 0.8974,
+      "learning_rate": 1.0436423424365646e-05,
+      "loss": 0.8037,
       "step": 17674
     },
     {
-      "epoch": 0.4854302271291643,
+      "epoch": 0.501560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.0956544014651635e-05,
-      "loss": 0.8788,
+      "learning_rate": 1.0435505220135288e-05,
+      "loss": 0.9246,
       "step": 17675
     },
     {
-      "epoch": 0.4854576913570075,
+      "epoch": 0.5015891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 1.0955658573638289e-05,
-      "loss": 0.893,
+      "learning_rate": 1.0434587012226203e-05,
+      "loss": 0.9724,
       "step": 17676
     },
     {
-      "epoch": 0.4854851555848507,
+      "epoch": 0.5016174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.0954773125063403e-05,
-      "loss": 0.953,
+      "learning_rate": 1.0433668800646139e-05,
+      "loss": 0.8811,
       "step": 17677
     },
     {
-      "epoch": 0.48551261981269395,
+      "epoch": 0.5016458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.095388766893398e-05,
-      "loss": 0.8072,
+      "learning_rate": 1.0432750585402853e-05,
+      "loss": 0.9415,
       "step": 17678
     },
     {
-      "epoch": 0.4855400840405372,
+      "epoch": 0.5016742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.0953002205257029e-05,
-      "loss": 0.8866,
+      "learning_rate": 1.0431832366504104e-05,
+      "loss": 0.9113,
       "step": 17679
     },
     {
-      "epoch": 0.48556754826838044,
+      "epoch": 0.5017026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.0952116734039555e-05,
-      "loss": 1.0099,
+      "learning_rate": 1.0430914143957651e-05,
+      "loss": 0.9243,
       "step": 17680
     },
     {
-      "epoch": 0.48559501249622367,
+      "epoch": 0.5017309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.0951231255288565e-05,
-      "loss": 0.9105,
+      "learning_rate": 1.0429995917771247e-05,
+      "loss": 0.8589,
       "step": 17681
     },
     {
-      "epoch": 0.4856224767240669,
+      "epoch": 0.5017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.0950345769011066e-05,
-      "loss": 0.9027,
+      "learning_rate": 1.0429077687952645e-05,
+      "loss": 0.9325,
       "step": 17682
     },
     {
-      "epoch": 0.48564994095191016,
+      "epoch": 0.5017877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.094946027521406e-05,
-      "loss": 0.7926,
+      "learning_rate": 1.0428159454509605e-05,
+      "loss": 0.9713,
       "step": 17683
     },
     {
-      "epoch": 0.4856774051797534,
+      "epoch": 0.5018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.0948574773904559e-05,
-      "loss": 0.9261,
+      "learning_rate": 1.0427241217449886e-05,
+      "loss": 0.8698,
       "step": 17684
     },
     {
-      "epoch": 0.4857048694075966,
+      "epoch": 0.501844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.0947689265089568e-05,
-      "loss": 1.001,
+      "learning_rate": 1.0426322976781238e-05,
+      "loss": 0.8771,
       "step": 17685
     },
     {
-      "epoch": 0.48573233363543983,
+      "epoch": 0.5018728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.0946803748776092e-05,
-      "loss": 0.9546,
+      "learning_rate": 1.0425404732511419e-05,
+      "loss": 0.8343,
       "step": 17686
     },
     {
-      "epoch": 0.48575979786328305,
+      "epoch": 0.5019012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.0945918224971135e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.0424486484648192e-05,
+      "loss": 0.9074,
       "step": 17687
     },
     {
-      "epoch": 0.4857872620911263,
+      "epoch": 0.5019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.0945032693681708e-05,
-      "loss": 0.9161,
+      "learning_rate": 1.0423568233199306e-05,
+      "loss": 1.011,
       "step": 17688
     },
     {
-      "epoch": 0.48581472631896955,
+      "epoch": 0.5019580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.0944147154914817e-05,
-      "loss": 0.8133,
+      "learning_rate": 1.042264997817252e-05,
+      "loss": 0.8968,
       "step": 17689
     },
     {
-      "epoch": 0.48584219054681277,
+      "epoch": 0.5019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.0943261608677467e-05,
-      "loss": 0.789,
+      "learning_rate": 1.042173171957559e-05,
+      "loss": 0.825,
       "step": 17690
     },
     {
-      "epoch": 0.485869654774656,
+      "epoch": 0.5020147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.0942376054976666e-05,
-      "loss": 0.8345,
+      "learning_rate": 1.0420813457416275e-05,
+      "loss": 0.8893,
       "step": 17691
     },
     {
-      "epoch": 0.48589711900249927,
+      "epoch": 0.5020431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.0941490493819422e-05,
-      "loss": 0.8279,
+      "learning_rate": 1.0419895191702328e-05,
+      "loss": 0.8931,
       "step": 17692
     },
     {
-      "epoch": 0.4859245832303425,
+      "epoch": 0.5020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.094060492521274e-05,
-      "loss": 0.8456,
+      "learning_rate": 1.0418976922441506e-05,
+      "loss": 0.8634,
       "step": 17693
     },
     {
-      "epoch": 0.4859520474581857,
+      "epoch": 0.502099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.0939719349163627e-05,
-      "loss": 0.9213,
+      "learning_rate": 1.0418058649641571e-05,
+      "loss": 0.9543,
       "step": 17694
     },
     {
-      "epoch": 0.48597951168602893,
+      "epoch": 0.5021282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.0938833765679091e-05,
-      "loss": 0.7934,
+      "learning_rate": 1.0417140373310273e-05,
+      "loss": 0.9535,
       "step": 17695
     },
     {
-      "epoch": 0.4860069759138722,
+      "epoch": 0.5021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.0937948174766135e-05,
-      "loss": 0.9926,
+      "learning_rate": 1.0416222093455373e-05,
+      "loss": 0.9106,
       "step": 17696
     },
     {
-      "epoch": 0.48603444014171543,
+      "epoch": 0.5021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.0937062576431774e-05,
-      "loss": 0.9791,
+      "learning_rate": 1.0415303810084626e-05,
+      "loss": 0.8617,
       "step": 17697
     },
     {
-      "epoch": 0.48606190436955865,
+      "epoch": 0.5022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.0936176970683011e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.041438552320579e-05,
+      "loss": 0.9159,
       "step": 17698
     },
     {
-      "epoch": 0.48608936859740187,
+      "epoch": 0.5022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.0935291357526851e-05,
-      "loss": 0.9427,
+      "learning_rate": 1.041346723282662e-05,
+      "loss": 0.9764,
       "step": 17699
     },
     {
-      "epoch": 0.4861168328252451,
+      "epoch": 0.5022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.0934405736970306e-05,
-      "loss": 0.9392,
+      "learning_rate": 1.0412548938954874e-05,
+      "loss": 0.8589,
       "step": 17700
     },
     {
-      "epoch": 0.48614429705308837,
+      "epoch": 0.5022985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.0933520109020378e-05,
-      "loss": 0.8895,
+      "learning_rate": 1.0411630641598307e-05,
+      "loss": 0.9672,
       "step": 17701
     },
     {
-      "epoch": 0.4861717612809316,
+      "epoch": 0.5023269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.093263447368408e-05,
-      "loss": 0.9063,
+      "learning_rate": 1.0410712340764676e-05,
+      "loss": 1.0093,
       "step": 17702
     },
     {
-      "epoch": 0.4861992255087748,
+      "epoch": 0.502355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.0931748830968414e-05,
-      "loss": 0.8914,
+      "learning_rate": 1.0409794036461745e-05,
+      "loss": 0.8278,
       "step": 17703
     },
     {
-      "epoch": 0.48622668973661803,
+      "epoch": 0.502383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.093086318088039e-05,
-      "loss": 0.8147,
+      "learning_rate": 1.0408875728697264e-05,
+      "loss": 0.7892,
       "step": 17704
     },
     {
-      "epoch": 0.4862541539644613,
+      "epoch": 0.5024120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.092997752342702e-05,
-      "loss": 0.953,
+      "learning_rate": 1.0407957417478987e-05,
+      "loss": 0.841,
       "step": 17705
     },
     {
-      "epoch": 0.48628161819230453,
+      "epoch": 0.5024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.0929091858615307e-05,
-      "loss": 0.8259,
+      "learning_rate": 1.0407039102814677e-05,
+      "loss": 0.9373,
       "step": 17706
     },
     {
-      "epoch": 0.48630908242014775,
+      "epoch": 0.5024687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.0928206186452255e-05,
-      "loss": 0.9744,
+      "learning_rate": 1.0406120784712093e-05,
+      "loss": 0.8046,
       "step": 17707
     },
     {
-      "epoch": 0.486336546647991,
+      "epoch": 0.5024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.0927320506944876e-05,
-      "loss": 0.9062,
+      "learning_rate": 1.0405202463178985e-05,
+      "loss": 0.9108,
       "step": 17708
     },
     {
-      "epoch": 0.48636401087583425,
+      "epoch": 0.5025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.0926434820100182e-05,
-      "loss": 0.8898,
+      "learning_rate": 1.0404284138223113e-05,
+      "loss": 0.8741,
       "step": 17709
     },
     {
-      "epoch": 0.4863914751036775,
+      "epoch": 0.5025539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.0925549125925176e-05,
-      "loss": 0.9271,
+      "learning_rate": 1.0403365809852236e-05,
+      "loss": 0.8483,
       "step": 17710
     },
     {
-      "epoch": 0.4864189393315207,
+      "epoch": 0.5025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.0924663424426864e-05,
-      "loss": 0.9533,
+      "learning_rate": 1.0402447478074112e-05,
+      "loss": 0.8561,
       "step": 17711
     },
     {
-      "epoch": 0.4864464035593639,
+      "epoch": 0.5026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.0923777715612261e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.040152914289649e-05,
+      "loss": 0.9197,
       "step": 17712
     },
     {
-      "epoch": 0.48647386778720714,
+      "epoch": 0.5026390465380249,
       "grad_norm": 0.0,
-      "learning_rate": 1.092289199948837e-05,
-      "loss": 0.7749,
+      "learning_rate": 1.0400610804327141e-05,
+      "loss": 0.9368,
       "step": 17713
     },
     {
-      "epoch": 0.4865013320150504,
+      "epoch": 0.5026674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.0922006276062198e-05,
-      "loss": 0.9195,
+      "learning_rate": 1.0399692462373811e-05,
+      "loss": 0.8923,
       "step": 17714
     },
     {
-      "epoch": 0.48652879624289364,
+      "epoch": 0.5026958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.0921120545340758e-05,
-      "loss": 0.9361,
+      "learning_rate": 1.039877411704426e-05,
+      "loss": 0.9793,
       "step": 17715
     },
     {
-      "epoch": 0.48655626047073686,
+      "epoch": 0.5027241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.0920234807331052e-05,
-      "loss": 0.8745,
+      "learning_rate": 1.0397855768346246e-05,
+      "loss": 1.0094,
       "step": 17716
     },
     {
-      "epoch": 0.4865837246985801,
+      "epoch": 0.5027525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.0919349062040093e-05,
-      "loss": 0.9488,
+      "learning_rate": 1.0396937416287527e-05,
+      "loss": 0.8724,
       "step": 17717
     },
     {
-      "epoch": 0.48661118892642335,
+      "epoch": 0.5027809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.091846330947489e-05,
-      "loss": 0.9108,
+      "learning_rate": 1.0396019060875862e-05,
+      "loss": 0.8518,
       "step": 17718
     },
     {
-      "epoch": 0.4866386531542666,
+      "epoch": 0.5028093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.0917577549642447e-05,
-      "loss": 0.8786,
+      "learning_rate": 1.0395100702119003e-05,
+      "loss": 0.9765,
       "step": 17719
     },
     {
-      "epoch": 0.4866661173821098,
+      "epoch": 0.5028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.0916691782549773e-05,
-      "loss": 0.8637,
+      "learning_rate": 1.0394182340024712e-05,
+      "loss": 0.7856,
       "step": 17720
     },
     {
-      "epoch": 0.486693581609953,
+      "epoch": 0.5028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.0915806008203881e-05,
-      "loss": 0.8259,
+      "learning_rate": 1.0393263974600747e-05,
+      "loss": 0.8517,
       "step": 17721
     },
     {
-      "epoch": 0.4867210458377963,
+      "epoch": 0.502894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.091492022661178e-05,
-      "loss": 0.8995,
+      "learning_rate": 1.039234560585486e-05,
+      "loss": 0.8465,
       "step": 17722
     },
     {
-      "epoch": 0.4867485100656395,
+      "epoch": 0.5029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.0914034437780472e-05,
-      "loss": 0.8772,
+      "learning_rate": 1.0391427233794813e-05,
+      "loss": 0.9319,
       "step": 17723
     },
     {
-      "epoch": 0.48677597429348274,
+      "epoch": 0.5029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.091314864171697e-05,
-      "loss": 0.934,
+      "learning_rate": 1.0390508858428363e-05,
+      "loss": 0.7975,
       "step": 17724
     },
     {
-      "epoch": 0.48680343852132596,
+      "epoch": 0.5029795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.0912262838428285e-05,
-      "loss": 0.9394,
+      "learning_rate": 1.0389590479763267e-05,
+      "loss": 0.8778,
       "step": 17725
     },
     {
-      "epoch": 0.4868309027491692,
+      "epoch": 0.5030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.0911377027921419e-05,
-      "loss": 0.9519,
+      "learning_rate": 1.0388672097807282e-05,
+      "loss": 0.7959,
       "step": 17726
     },
     {
-      "epoch": 0.48685836697701246,
+      "epoch": 0.5030363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.091049121020339e-05,
-      "loss": 0.9737,
+      "learning_rate": 1.038775371256817e-05,
+      "loss": 0.9727,
       "step": 17727
     },
     {
-      "epoch": 0.4868858312048557,
+      "epoch": 0.5030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.0909605385281195e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.0386835324053682e-05,
+      "loss": 0.9275,
       "step": 17728
     },
     {
-      "epoch": 0.4869132954326989,
+      "epoch": 0.5030930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.0908719553161856e-05,
-      "loss": 0.8115,
+      "learning_rate": 1.0385916932271577e-05,
+      "loss": 0.9419,
       "step": 17729
     },
     {
-      "epoch": 0.4869407596605421,
+      "epoch": 0.5031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.0907833713852374e-05,
-      "loss": 0.9692,
+      "learning_rate": 1.0384998537229618e-05,
+      "loss": 0.8599,
       "step": 17730
     },
     {
-      "epoch": 0.4869682238883854,
+      "epoch": 0.5031498297389331,
       "grad_norm": 0.0,
-      "learning_rate": 1.0906947867359758e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.0384080138935555e-05,
+      "loss": 0.8624,
       "step": 17731
     },
     {
-      "epoch": 0.4869956881162286,
+      "epoch": 0.5031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.0906062013691019e-05,
-      "loss": 0.9023,
+      "learning_rate": 1.0383161737397154e-05,
+      "loss": 0.9467,
       "step": 17732
     },
     {
-      "epoch": 0.48702315234407184,
+      "epoch": 0.5032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.0905176152853168e-05,
-      "loss": 0.8646,
+      "learning_rate": 1.0382243332622164e-05,
+      "loss": 0.7986,
       "step": 17733
     },
     {
-      "epoch": 0.48705061657191506,
+      "epoch": 0.5032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.0904290284853214e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.038132492461835e-05,
+      "loss": 0.8186,
       "step": 17734
     },
     {
-      "epoch": 0.48707808079975834,
+      "epoch": 0.5032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0903404409698162e-05,
-      "loss": 0.9321,
+      "learning_rate": 1.0380406513393469e-05,
+      "loss": 0.8275,
       "step": 17735
     },
     {
-      "epoch": 0.48710554502760156,
+      "epoch": 0.5032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.0902518527395028e-05,
-      "loss": 0.9334,
+      "learning_rate": 1.0379488098955275e-05,
+      "loss": 0.9676,
       "step": 17736
     },
     {
-      "epoch": 0.4871330092554448,
+      "epoch": 0.5033200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 1.0901632637950816e-05,
-      "loss": 0.888,
+      "learning_rate": 1.0378569681311528e-05,
+      "loss": 1.0098,
       "step": 17737
     },
     {
-      "epoch": 0.487160473483288,
+      "epoch": 0.5033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.0900746741372536e-05,
-      "loss": 0.9123,
+      "learning_rate": 1.0377651260469987e-05,
+      "loss": 0.8723,
       "step": 17738
     },
     {
-      "epoch": 0.4871879377111312,
+      "epoch": 0.5033768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.08998608376672e-05,
-      "loss": 1.0027,
+      "learning_rate": 1.0376732836438406e-05,
+      "loss": 0.8192,
       "step": 17739
     },
     {
-      "epoch": 0.4872154019389745,
+      "epoch": 0.503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.0898974926841817e-05,
-      "loss": 0.8811,
+      "learning_rate": 1.0375814409224547e-05,
+      "loss": 0.9357,
       "step": 17740
     },
     {
-      "epoch": 0.4872428661668177,
+      "epoch": 0.503433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.0898089008903394e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.0374895978836169e-05,
+      "loss": 0.9536,
       "step": 17741
     },
     {
-      "epoch": 0.48727033039466094,
+      "epoch": 0.5034619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 1.0897203083858946e-05,
-      "loss": 0.8776,
+      "learning_rate": 1.0373977545281027e-05,
+      "loss": 0.7945,
       "step": 17742
     },
     {
-      "epoch": 0.48729779462250417,
+      "epoch": 0.5034903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.0896317151715477e-05,
-      "loss": 0.9464,
+      "learning_rate": 1.0373059108566878e-05,
+      "loss": 0.906,
       "step": 17743
     },
     {
-      "epoch": 0.48732525885034744,
+      "epoch": 0.5035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.089543121248e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.0372140668701483e-05,
+      "loss": 0.8094,
       "step": 17744
     },
     {
-      "epoch": 0.48735272307819066,
+      "epoch": 0.5035471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.0894545266159526e-05,
-      "loss": 0.9124,
+      "learning_rate": 1.0371222225692601e-05,
+      "loss": 0.8036,
       "step": 17745
     },
     {
-      "epoch": 0.4873801873060339,
+      "epoch": 0.5035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.0893659312761063e-05,
-      "loss": 0.9551,
+      "learning_rate": 1.0370303779547985e-05,
+      "loss": 0.9676,
       "step": 17746
     },
     {
-      "epoch": 0.4874076515338771,
+      "epoch": 0.5036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.0892773352291619e-05,
-      "loss": 0.8932,
+      "learning_rate": 1.0369385330275397e-05,
+      "loss": 0.8879,
       "step": 17747
     },
     {
-      "epoch": 0.4874351157617204,
+      "epoch": 0.5036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.0891887384758209e-05,
-      "loss": 0.8291,
+      "learning_rate": 1.0368466877882595e-05,
+      "loss": 0.7916,
       "step": 17748
     },
     {
-      "epoch": 0.4874625799895636,
+      "epoch": 0.5036606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.0891001410167842e-05,
-      "loss": 0.9207,
+      "learning_rate": 1.0367548422377336e-05,
+      "loss": 0.8186,
       "step": 17749
     },
     {
-      "epoch": 0.4874900442174068,
+      "epoch": 0.503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.0890115428527521e-05,
-      "loss": 0.9608,
+      "learning_rate": 1.036662996376738e-05,
+      "loss": 0.8943,
       "step": 17750
     },
     {
-      "epoch": 0.48751750844525005,
+      "epoch": 0.5037173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 1.0889229439844267e-05,
-      "loss": 0.8813,
+      "learning_rate": 1.0365711502060485e-05,
+      "loss": 0.8177,
       "step": 17751
     },
     {
-      "epoch": 0.48754497267309327,
+      "epoch": 0.5037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.0888343444125083e-05,
-      "loss": 0.8785,
+      "learning_rate": 1.0364793037264408e-05,
+      "loss": 0.8977,
       "step": 17752
     },
     {
-      "epoch": 0.48757243690093655,
+      "epoch": 0.5037741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.0887457441376981e-05,
-      "loss": 0.7888,
+      "learning_rate": 1.0363874569386907e-05,
+      "loss": 0.8737,
       "step": 17753
     },
     {
-      "epoch": 0.48759990112877977,
+      "epoch": 0.5038024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.0886571431606974e-05,
-      "loss": 0.9771,
+      "learning_rate": 1.0362956098435739e-05,
+      "loss": 0.946,
       "step": 17754
     },
     {
-      "epoch": 0.487627365356623,
+      "epoch": 0.5038308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.0885685414822069e-05,
-      "loss": 0.8899,
+      "learning_rate": 1.0362037624418668e-05,
+      "loss": 0.8761,
       "step": 17755
     },
     {
-      "epoch": 0.4876548295844662,
+      "epoch": 0.5038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.0884799391029277e-05,
-      "loss": 0.9719,
+      "learning_rate": 1.0361119147343448e-05,
+      "loss": 0.8754,
       "step": 17756
     },
     {
-      "epoch": 0.4876822938123095,
+      "epoch": 0.5038876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.0883913360235611e-05,
-      "loss": 0.9179,
+      "learning_rate": 1.0360200667217839e-05,
+      "loss": 0.871,
       "step": 17757
     },
     {
-      "epoch": 0.4877097580401527,
+      "epoch": 0.5039160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 1.0883027322448082e-05,
-      "loss": 0.9221,
+      "learning_rate": 1.0359282184049599e-05,
+      "loss": 0.9577,
       "step": 17758
     },
     {
-      "epoch": 0.48773722226799593,
+      "epoch": 0.50394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.0882141277673696e-05,
-      "loss": 1.0201,
+      "learning_rate": 1.0358363697846484e-05,
+      "loss": 0.933,
       "step": 17759
     },
     {
-      "epoch": 0.48776468649583915,
+      "epoch": 0.5039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.0881255225919468e-05,
-      "loss": 0.8552,
+      "learning_rate": 1.0357445208616256e-05,
+      "loss": 0.7959,
       "step": 17760
     },
     {
-      "epoch": 0.4877921507236824,
+      "epoch": 0.5040011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 1.0880369167192407e-05,
-      "loss": 0.8908,
+      "learning_rate": 1.0356526716366674e-05,
+      "loss": 0.9021,
       "step": 17761
     },
     {
-      "epoch": 0.48781961495152565,
+      "epoch": 0.5040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.0879483101499524e-05,
-      "loss": 0.8878,
+      "learning_rate": 1.035560822110549e-05,
+      "loss": 0.9115,
       "step": 17762
     },
     {
-      "epoch": 0.48784707917936887,
+      "epoch": 0.5040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.0878597028847828e-05,
-      "loss": 0.8431,
+      "learning_rate": 1.0354689722840473e-05,
+      "loss": 1.0048,
       "step": 17763
     },
     {
-      "epoch": 0.4878745434072121,
+      "epoch": 0.5040862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 1.0877710949244332e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.0353771221579372e-05,
+      "loss": 0.9594,
       "step": 17764
     },
     {
-      "epoch": 0.48790200763505537,
+      "epoch": 0.5041146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.087682486269605e-05,
-      "loss": 0.9279,
+      "learning_rate": 1.0352852717329952e-05,
+      "loss": 0.9658,
       "step": 17765
     },
     {
-      "epoch": 0.4879294718628986,
+      "epoch": 0.5041430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.0875938769209989e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.0351934210099968e-05,
+      "loss": 0.9192,
       "step": 17766
     },
     {
-      "epoch": 0.4879569360907418,
+      "epoch": 0.5041713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.087505266879316e-05,
-      "loss": 0.9633,
+      "learning_rate": 1.035101569989718e-05,
+      "loss": 0.8314,
       "step": 17767
     },
     {
-      "epoch": 0.48798440031858503,
+      "epoch": 0.504199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.0874166561452577e-05,
-      "loss": 0.9481,
+      "learning_rate": 1.035009718672935e-05,
+      "loss": 0.87,
       "step": 17768
     },
     {
-      "epoch": 0.48801186454642825,
+      "epoch": 0.504228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.0873280447195247e-05,
-      "loss": 0.9509,
+      "learning_rate": 1.0349178670604227e-05,
+      "loss": 0.8708,
       "step": 17769
     },
     {
-      "epoch": 0.48803932877427153,
+      "epoch": 0.5042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.0872394326028187e-05,
-      "loss": 0.8774,
+      "learning_rate": 1.0348260151529577e-05,
+      "loss": 0.9214,
       "step": 17770
     },
     {
-      "epoch": 0.48806679300211475,
+      "epoch": 0.5042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.0871508197958403e-05,
-      "loss": 0.9458,
+      "learning_rate": 1.034734162951316e-05,
+      "loss": 0.9713,
       "step": 17771
     },
     {
-      "epoch": 0.488094257229958,
+      "epoch": 0.5043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.0870622062992908e-05,
-      "loss": 0.9822,
+      "learning_rate": 1.0346423104562735e-05,
+      "loss": 0.8207,
       "step": 17772
     },
     {
-      "epoch": 0.4881217214578012,
+      "epoch": 0.5043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.0869735921138716e-05,
-      "loss": 0.8446,
+      "learning_rate": 1.0345504576686053e-05,
+      "loss": 0.9312,
       "step": 17773
     },
     {
-      "epoch": 0.48814918568564447,
+      "epoch": 0.5043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.0868849772402837e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.0344586045890882e-05,
+      "loss": 0.9186,
       "step": 17774
     },
     {
-      "epoch": 0.4881766499134877,
+      "epoch": 0.5043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.086796361679228e-05,
-      "loss": 0.8986,
+      "learning_rate": 1.034366751218498e-05,
+      "loss": 0.8274,
       "step": 17775
     },
     {
-      "epoch": 0.4882041141413309,
+      "epoch": 0.5044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.0867077454314056e-05,
-      "loss": 0.8223,
+      "learning_rate": 1.0342748975576097e-05,
+      "loss": 0.9791,
       "step": 17776
     },
     {
-      "epoch": 0.48823157836917414,
+      "epoch": 0.5044551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.0866191284975185e-05,
-      "loss": 0.9176,
+      "learning_rate": 1.0341830436072001e-05,
+      "loss": 0.9238,
       "step": 17777
     },
     {
-      "epoch": 0.4882590425970174,
+      "epoch": 0.5044835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 1.086530510878267e-05,
-      "loss": 0.9682,
+      "learning_rate": 1.0340911893680447e-05,
+      "loss": 0.9195,
       "step": 17778
     },
     {
-      "epoch": 0.48828650682486063,
+      "epoch": 0.5045119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.0864418925743524e-05,
-      "loss": 0.891,
+      "learning_rate": 1.0339993348409197e-05,
+      "loss": 0.9577,
       "step": 17779
     },
     {
-      "epoch": 0.48831397105270385,
+      "epoch": 0.5045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.0863532735864767e-05,
-      "loss": 0.833,
+      "learning_rate": 1.0339074800266005e-05,
+      "loss": 0.8399,
       "step": 17780
     },
     {
-      "epoch": 0.4883414352805471,
+      "epoch": 0.5045686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.08626465391534e-05,
-      "loss": 0.9372,
+      "learning_rate": 1.0338156249258635e-05,
+      "loss": 0.8581,
       "step": 17781
     },
     {
-      "epoch": 0.4883688995083903,
+      "epoch": 0.5045970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.0861760335616438e-05,
-      "loss": 0.8808,
+      "learning_rate": 1.0337237695394844e-05,
+      "loss": 0.9124,
       "step": 17782
     },
     {
-      "epoch": 0.4883963637362336,
+      "epoch": 0.5046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.0860874125260898e-05,
-      "loss": 0.905,
+      "learning_rate": 1.0336319138682389e-05,
+      "loss": 1.0091,
       "step": 17783
     },
     {
-      "epoch": 0.4884238279640768,
+      "epoch": 0.5046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.0859987908093786e-05,
-      "loss": 0.91,
+      "learning_rate": 1.0335400579129034e-05,
+      "loss": 0.8739,
       "step": 17784
     },
     {
-      "epoch": 0.48845129219192,
+      "epoch": 0.5046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.0859101684122118e-05,
-      "loss": 0.9326,
+      "learning_rate": 1.0334482016742533e-05,
+      "loss": 0.9254,
       "step": 17785
     },
     {
-      "epoch": 0.48847875641976324,
+      "epoch": 0.5047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.0858215453352904e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.0333563451530648e-05,
+      "loss": 0.9817,
       "step": 17786
     },
     {
-      "epoch": 0.4885062206476065,
+      "epoch": 0.504738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.0857329215793157e-05,
-      "loss": 0.8516,
+      "learning_rate": 1.0332644883501138e-05,
+      "loss": 0.9964,
       "step": 17787
     },
     {
-      "epoch": 0.48853368487544974,
+      "epoch": 0.5047673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 1.0856442971449888e-05,
-      "loss": 0.8796,
+      "learning_rate": 1.0331726312661761e-05,
+      "loss": 0.8076,
       "step": 17788
     },
     {
-      "epoch": 0.48856114910329296,
+      "epoch": 0.5047956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.0855556720330111e-05,
-      "loss": 0.8376,
+      "learning_rate": 1.0330807739020278e-05,
+      "loss": 0.872,
       "step": 17789
     },
     {
-      "epoch": 0.4885886133311362,
+      "epoch": 0.5048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.085467046244084e-05,
-      "loss": 0.9547,
+      "learning_rate": 1.0329889162584446e-05,
+      "loss": 0.9042,
       "step": 17790
     },
     {
-      "epoch": 0.48861607755897946,
+      "epoch": 0.5048524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.0853784197789082e-05,
-      "loss": 0.9258,
+      "learning_rate": 1.0328970583362026e-05,
+      "loss": 0.8775,
       "step": 17791
     },
     {
-      "epoch": 0.4886435417868227,
+      "epoch": 0.5048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.0852897926381856e-05,
-      "loss": 0.8751,
+      "learning_rate": 1.0328052001360778e-05,
+      "loss": 0.8427,
       "step": 17792
     },
     {
-      "epoch": 0.4886710060146659,
+      "epoch": 0.5049091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.085201164822617e-05,
-      "loss": 0.8618,
+      "learning_rate": 1.0327133416588457e-05,
+      "loss": 0.9014,
       "step": 17793
     },
     {
-      "epoch": 0.4886984702425091,
+      "epoch": 0.5049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.0851125363329037e-05,
-      "loss": 0.9892,
+      "learning_rate": 1.0326214829052826e-05,
+      "loss": 0.9946,
       "step": 17794
     },
     {
-      "epoch": 0.48872593447035234,
+      "epoch": 0.5049659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.085023907169747e-05,
-      "loss": 0.9467,
+      "learning_rate": 1.0325296238761644e-05,
+      "loss": 0.8772,
       "step": 17795
     },
     {
-      "epoch": 0.4887533986981956,
+      "epoch": 0.504994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.0849352773338482e-05,
-      "loss": 0.8038,
+      "learning_rate": 1.0324377645722671e-05,
+      "loss": 0.9061,
       "step": 17796
     },
     {
-      "epoch": 0.48878086292603884,
+      "epoch": 0.5050227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.0848466468259087e-05,
-      "loss": 0.863,
+      "learning_rate": 1.0323459049943665e-05,
+      "loss": 0.8116,
       "step": 17797
     },
     {
-      "epoch": 0.48880832715388206,
+      "epoch": 0.5050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.0847580156466297e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.0322540451432385e-05,
+      "loss": 0.8344,
       "step": 17798
     },
     {
-      "epoch": 0.4888357913817253,
+      "epoch": 0.5050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.0846693837967123e-05,
-      "loss": 0.8699,
+      "learning_rate": 1.0321621850196595e-05,
+      "loss": 0.8835,
       "step": 17799
     },
     {
-      "epoch": 0.48886325560956856,
+      "epoch": 0.5051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.084580751276858e-05,
-      "loss": 0.9266,
+      "learning_rate": 1.0320703246244044e-05,
+      "loss": 0.8471,
       "step": 17800
     },
     {
-      "epoch": 0.4888907198374118,
+      "epoch": 0.5051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.0844921180877681e-05,
-      "loss": 0.9647,
+      "learning_rate": 1.0319784639582502e-05,
+      "loss": 0.7441,
       "step": 17801
     },
     {
-      "epoch": 0.488918184065255,
+      "epoch": 0.5051645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.084403484230144e-05,
-      "loss": 0.9922,
+      "learning_rate": 1.0318866030219727e-05,
+      "loss": 0.7939,
       "step": 17802
     },
     {
-      "epoch": 0.4889456482930982,
+      "epoch": 0.5051929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.0843148497046863e-05,
-      "loss": 0.9085,
+      "learning_rate": 1.0317947418163471e-05,
+      "loss": 0.9181,
       "step": 17803
     },
     {
-      "epoch": 0.4889731125209415,
+      "epoch": 0.5052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.0842262145120975e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.0317028803421506e-05,
+      "loss": 0.8429,
       "step": 17804
     },
     {
-      "epoch": 0.4890005767487847,
+      "epoch": 0.505249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.0841375786530779e-05,
-      "loss": 0.966,
+      "learning_rate": 1.031611018600158e-05,
+      "loss": 0.8791,
       "step": 17805
     },
     {
-      "epoch": 0.48902804097662794,
+      "epoch": 0.50527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.084048942128329e-05,
-      "loss": 0.9016,
+      "learning_rate": 1.0315191565911458e-05,
+      "loss": 0.9157,
       "step": 17806
     },
     {
-      "epoch": 0.48905550520447116,
+      "epoch": 0.5053064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.0839603049385528e-05,
-      "loss": 0.8415,
+      "learning_rate": 1.0314272943158899e-05,
+      "loss": 0.7956,
       "step": 17807
     },
     {
-      "epoch": 0.4890829694323144,
+      "epoch": 0.5053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.0838716670844496e-05,
-      "loss": 0.9228,
+      "learning_rate": 1.0313354317751658e-05,
+      "loss": 0.8283,
       "step": 17808
     },
     {
-      "epoch": 0.48911043366015766,
+      "epoch": 0.5053632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.0837830285667217e-05,
-      "loss": 0.8806,
+      "learning_rate": 1.0312435689697504e-05,
+      "loss": 0.8284,
       "step": 17809
     },
     {
-      "epoch": 0.4891378978880009,
+      "epoch": 0.5053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.0836943893860698e-05,
-      "loss": 0.9383,
+      "learning_rate": 1.031151705900419e-05,
+      "loss": 0.9367,
       "step": 17810
     },
     {
-      "epoch": 0.4891653621158441,
+      "epoch": 0.5054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.0836057495431955e-05,
-      "loss": 0.9117,
+      "learning_rate": 1.031059842567948e-05,
+      "loss": 1.0543,
       "step": 17811
     },
     {
-      "epoch": 0.4891928263436873,
+      "epoch": 0.5054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.0835171090388002e-05,
-      "loss": 0.9251,
+      "learning_rate": 1.030967978973113e-05,
+      "loss": 0.875,
       "step": 17812
     },
     {
-      "epoch": 0.4892202905715306,
+      "epoch": 0.5054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.0834284678735851e-05,
-      "loss": 0.9148,
+      "learning_rate": 1.0308761151166896e-05,
+      "loss": 0.9466,
       "step": 17813
     },
     {
-      "epoch": 0.4892477547993738,
+      "epoch": 0.5055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.083339826048252e-05,
-      "loss": 0.9104,
+      "learning_rate": 1.0307842509994548e-05,
+      "loss": 0.8745,
       "step": 17814
     },
     {
-      "epoch": 0.48927521902721705,
+      "epoch": 0.5055334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 1.0832511835635013e-05,
-      "loss": 0.8861,
+      "learning_rate": 1.0306923866221839e-05,
+      "loss": 0.9207,
       "step": 17815
     },
     {
-      "epoch": 0.48930268325506027,
+      "epoch": 0.5055618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.0831625404200355e-05,
-      "loss": 0.9713,
+      "learning_rate": 1.030600521985653e-05,
+      "loss": 0.9423,
       "step": 17816
     },
     {
-      "epoch": 0.48933014748290354,
+      "epoch": 0.5055902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.0830738966185551e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.030508657090638e-05,
+      "loss": 0.8647,
       "step": 17817
     },
     {
-      "epoch": 0.48935761171074676,
+      "epoch": 0.5056186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.0829852521597624e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.0304167919379151e-05,
+      "loss": 0.9285,
       "step": 17818
     },
     {
-      "epoch": 0.48938507593859,
+      "epoch": 0.5056469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.0828966070443576e-05,
-      "loss": 0.8264,
+      "learning_rate": 1.0303249265282605e-05,
+      "loss": 0.9815,
       "step": 17819
     },
     {
-      "epoch": 0.4894125401664332,
+      "epoch": 0.5056753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.0828079612730429e-05,
-      "loss": 0.8651,
+      "learning_rate": 1.0302330608624494e-05,
+      "loss": 0.8737,
       "step": 17820
     },
     {
-      "epoch": 0.48944000439427643,
+      "epoch": 0.5057037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.0827193148465199e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.0301411949412586e-05,
+      "loss": 0.8835,
       "step": 17821
     },
     {
-      "epoch": 0.4894674686221197,
+      "epoch": 0.5057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.0826306677654894e-05,
-      "loss": 0.8578,
+      "learning_rate": 1.0300493287654635e-05,
+      "loss": 0.8343,
       "step": 17822
     },
     {
-      "epoch": 0.4894949328499629,
+      "epoch": 0.5057604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 1.0825420200306529e-05,
-      "loss": 0.8073,
+      "learning_rate": 1.0299574623358406e-05,
+      "loss": 0.9073,
       "step": 17823
     },
     {
-      "epoch": 0.48952239707780615,
+      "epoch": 0.505788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.082453371642712e-05,
-      "loss": 0.8418,
+      "learning_rate": 1.0298655956531653e-05,
+      "loss": 0.8469,
       "step": 17824
     },
     {
-      "epoch": 0.48954986130564937,
+      "epoch": 0.5058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.082364722602368e-05,
-      "loss": 0.9986,
+      "learning_rate": 1.0297737287182144e-05,
+      "loss": 0.8508,
       "step": 17825
     },
     {
-      "epoch": 0.48957732553349265,
+      "epoch": 0.5058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.0822760729103228e-05,
-      "loss": 0.8988,
+      "learning_rate": 1.0296818615317634e-05,
+      "loss": 0.8605,
       "step": 17826
     },
     {
-      "epoch": 0.48960478976133587,
+      "epoch": 0.5058740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.0821874225672773e-05,
-      "loss": 0.9175,
+      "learning_rate": 1.0295899940945884e-05,
+      "loss": 0.8937,
       "step": 17827
     },
     {
-      "epoch": 0.4896322539891791,
+      "epoch": 0.5059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.0820987715739328e-05,
-      "loss": 1.0377,
+      "learning_rate": 1.0294981264074653e-05,
+      "loss": 0.8209,
       "step": 17828
     },
     {
-      "epoch": 0.4896597182170223,
+      "epoch": 0.5059307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.0820101199309913e-05,
-      "loss": 0.9298,
+      "learning_rate": 1.0294062584711703e-05,
+      "loss": 0.8834,
       "step": 17829
     },
     {
-      "epoch": 0.4896871824448656,
+      "epoch": 0.5059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.0819214676391541e-05,
-      "loss": 0.8736,
+      "learning_rate": 1.0293143902864791e-05,
+      "loss": 0.7867,
       "step": 17830
     },
     {
-      "epoch": 0.4897146466727088,
+      "epoch": 0.5059875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.0818328146991222e-05,
-      "loss": 1.0497,
+      "learning_rate": 1.0292225218541678e-05,
+      "loss": 0.9518,
       "step": 17831
     },
     {
-      "epoch": 0.48974211090055203,
+      "epoch": 0.5060158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.081744161111597e-05,
-      "loss": 0.8704,
+      "learning_rate": 1.0291306531750129e-05,
+      "loss": 0.9514,
       "step": 17832
     },
     {
-      "epoch": 0.48976957512839525,
+      "epoch": 0.5060442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 1.0816555068772812e-05,
-      "loss": 0.9742,
+      "learning_rate": 1.0290387842497902e-05,
+      "loss": 0.8328,
       "step": 17833
     },
     {
-      "epoch": 0.4897970393562385,
+      "epoch": 0.5060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.081566851996875e-05,
-      "loss": 0.8019,
+      "learning_rate": 1.0289469150792753e-05,
+      "loss": 0.9856,
       "step": 17834
     },
     {
-      "epoch": 0.48982450358408175,
+      "epoch": 0.5061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.08147819647108e-05,
-      "loss": 1.0268,
+      "learning_rate": 1.0288550456642445e-05,
+      "loss": 0.917,
       "step": 17835
     },
     {
-      "epoch": 0.48985196781192497,
+      "epoch": 0.5061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.0813895403005982e-05,
-      "loss": 0.8029,
+      "learning_rate": 1.028763176005474e-05,
+      "loss": 0.885,
       "step": 17836
     },
     {
-      "epoch": 0.4898794320397682,
+      "epoch": 0.5061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.0813008834861309e-05,
-      "loss": 0.9222,
+      "learning_rate": 1.0286713061037394e-05,
+      "loss": 0.7394,
       "step": 17837
     },
     {
-      "epoch": 0.4899068962676114,
+      "epoch": 0.5061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.0812122260283795e-05,
-      "loss": 0.7912,
+      "learning_rate": 1.028579435959817e-05,
+      "loss": 0.8643,
       "step": 17838
     },
     {
-      "epoch": 0.4899343604954547,
+      "epoch": 0.5062145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 1.0811235679280455e-05,
-      "loss": 0.8228,
+      "learning_rate": 1.028487565574483e-05,
+      "loss": 0.9792,
       "step": 17839
     },
     {
-      "epoch": 0.4899618247232979,
+      "epoch": 0.5062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.0810349091858304e-05,
-      "loss": 0.8791,
+      "learning_rate": 1.0283956949485133e-05,
+      "loss": 0.8771,
       "step": 17840
     },
     {
-      "epoch": 0.48998928895114113,
+      "epoch": 0.5062712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.0809462498024356e-05,
-      "loss": 0.8538,
+      "learning_rate": 1.0283038240826837e-05,
+      "loss": 0.8992,
       "step": 17841
     },
     {
-      "epoch": 0.49001675317898435,
+      "epoch": 0.506299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.0808575897785628e-05,
-      "loss": 0.9708,
+      "learning_rate": 1.0282119529777705e-05,
+      "loss": 0.8966,
       "step": 17842
     },
     {
-      "epoch": 0.49004421740682763,
+      "epoch": 0.506328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.0807689291149135e-05,
-      "loss": 0.899,
+      "learning_rate": 1.0281200816345498e-05,
+      "loss": 1.0291,
       "step": 17843
     },
     {
-      "epoch": 0.49007168163467085,
+      "epoch": 0.5063564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 1.080680267812189e-05,
-      "loss": 0.877,
+      "learning_rate": 1.0280282100537972e-05,
+      "loss": 1.0016,
       "step": 17844
     },
     {
-      "epoch": 0.4900991458625141,
+      "epoch": 0.5063847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.080591605871091e-05,
-      "loss": 0.8337,
+      "learning_rate": 1.0279363382362894e-05,
+      "loss": 0.8967,
       "step": 17845
     },
     {
-      "epoch": 0.4901266100903573,
+      "epoch": 0.5064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.0805029432923212e-05,
-      "loss": 0.812,
+      "learning_rate": 1.0278444661828018e-05,
+      "loss": 0.9156,
       "step": 17846
     },
     {
-      "epoch": 0.4901540743182005,
+      "epoch": 0.5064415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.0804142800765807e-05,
-      "loss": 0.9419,
+      "learning_rate": 1.027752593894111e-05,
+      "loss": 0.8773,
       "step": 17847
     },
     {
-      "epoch": 0.4901815385460438,
+      "epoch": 0.5064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.0803256162245714e-05,
-      "loss": 0.8446,
+      "learning_rate": 1.0276607213709926e-05,
+      "loss": 0.8766,
       "step": 17848
     },
     {
-      "epoch": 0.490209002773887,
+      "epoch": 0.5064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.0802369517369944e-05,
-      "loss": 0.8454,
+      "learning_rate": 1.027568848614223e-05,
+      "loss": 0.985,
       "step": 17849
     },
     {
-      "epoch": 0.49023646700173024,
+      "epoch": 0.5065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.0801482866145518e-05,
-      "loss": 0.9328,
+      "learning_rate": 1.0274769756245781e-05,
+      "loss": 0.9735,
       "step": 17850
     },
     {
-      "epoch": 0.49026393122957346,
+      "epoch": 0.5065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.080059620857945e-05,
-      "loss": 0.9466,
+      "learning_rate": 1.0273851024028337e-05,
+      "loss": 0.8861,
       "step": 17851
     },
     {
-      "epoch": 0.49029139545741673,
+      "epoch": 0.506583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.0799709544678754e-05,
-      "loss": 0.9659,
+      "learning_rate": 1.0272932289497664e-05,
+      "loss": 0.9189,
       "step": 17852
     },
     {
-      "epoch": 0.49031885968525996,
+      "epoch": 0.5066118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.0798822874450447e-05,
-      "loss": 0.9437,
+      "learning_rate": 1.027201355266152e-05,
+      "loss": 0.8772,
       "step": 17853
     },
     {
-      "epoch": 0.4903463239131032,
+      "epoch": 0.5066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.0797936197901544e-05,
-      "loss": 0.9001,
+      "learning_rate": 1.0271094813527657e-05,
+      "loss": 0.8443,
       "step": 17854
     },
     {
-      "epoch": 0.4903737881409464,
+      "epoch": 0.5066685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.0797049515039057e-05,
-      "loss": 0.883,
+      "learning_rate": 1.0270176072103852e-05,
+      "loss": 0.9475,
       "step": 17855
     },
     {
-      "epoch": 0.4904012523687897,
+      "epoch": 0.5066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.0796162825870006e-05,
-      "loss": 0.8043,
+      "learning_rate": 1.026925732839786e-05,
+      "loss": 0.9214,
       "step": 17856
     },
     {
-      "epoch": 0.4904287165966329,
+      "epoch": 0.5067253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 1.079527613040141e-05,
-      "loss": 0.9166,
+      "learning_rate": 1.0268338582417432e-05,
+      "loss": 0.8916,
       "step": 17857
     },
     {
-      "epoch": 0.4904561808244761,
+      "epoch": 0.5067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.0794389428640282e-05,
-      "loss": 0.8614,
+      "learning_rate": 1.026741983417034e-05,
+      "loss": 0.8563,
       "step": 17858
     },
     {
-      "epoch": 0.49048364505231934,
+      "epoch": 0.5067820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.0793502720593634e-05,
-      "loss": 0.8856,
+      "learning_rate": 1.0266501083664341e-05,
+      "loss": 0.9049,
       "step": 17859
     },
     {
-      "epoch": 0.49051110928016256,
+      "epoch": 0.5068104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.0792616006268485e-05,
-      "loss": 0.8566,
+      "learning_rate": 1.0265582330907195e-05,
+      "loss": 0.7978,
       "step": 17860
     },
     {
-      "epoch": 0.49053857350800584,
+      "epoch": 0.506838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.0791729285671854e-05,
-      "loss": 0.9206,
+      "learning_rate": 1.0264663575906661e-05,
+      "loss": 0.9236,
       "step": 17861
     },
     {
-      "epoch": 0.49056603773584906,
+      "epoch": 0.5068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.0790842558810753e-05,
-      "loss": 0.8573,
+      "learning_rate": 1.0263744818670503e-05,
+      "loss": 0.8944,
       "step": 17862
     },
     {
-      "epoch": 0.4905935019636923,
+      "epoch": 0.5068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.0789955825692198e-05,
-      "loss": 0.7647,
+      "learning_rate": 1.026282605920648e-05,
+      "loss": 0.8594,
       "step": 17863
     },
     {
-      "epoch": 0.4906209661915355,
+      "epoch": 0.5069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.0789069086323205e-05,
-      "loss": 0.8112,
+      "learning_rate": 1.0261907297522356e-05,
+      "loss": 0.7997,
       "step": 17864
     },
     {
-      "epoch": 0.4906484304193788,
+      "epoch": 0.5069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.0788182340710796e-05,
-      "loss": 0.9849,
+      "learning_rate": 1.0260988533625889e-05,
+      "loss": 0.9661,
       "step": 17865
     },
     {
-      "epoch": 0.490675894647222,
+      "epoch": 0.5069807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.0787295588861981e-05,
-      "loss": 0.8759,
+      "learning_rate": 1.0260069767524838e-05,
+      "loss": 0.8771,
       "step": 17866
     },
     {
-      "epoch": 0.4907033588750652,
+      "epoch": 0.5070090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.0786408830783776e-05,
-      "loss": 0.974,
+      "learning_rate": 1.0259150999226965e-05,
+      "loss": 0.9571,
       "step": 17867
     },
     {
-      "epoch": 0.49073082310290844,
+      "epoch": 0.5070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.0785522066483203e-05,
-      "loss": 0.8612,
+      "learning_rate": 1.0258232228740035e-05,
+      "loss": 0.8244,
       "step": 17868
     },
     {
-      "epoch": 0.4907582873307517,
+      "epoch": 0.5070658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.0784635295967273e-05,
-      "loss": 0.7768,
+      "learning_rate": 1.0257313456071805e-05,
+      "loss": 0.9168,
       "step": 17869
     },
     {
-      "epoch": 0.49078575155859494,
+      "epoch": 0.507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.0783748519243007e-05,
-      "loss": 0.9669,
+      "learning_rate": 1.0256394681230035e-05,
+      "loss": 0.9535,
       "step": 17870
     },
     {
-      "epoch": 0.49081321578643816,
+      "epoch": 0.507122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.0782861736317416e-05,
-      "loss": 1.0153,
+      "learning_rate": 1.0255475904222488e-05,
+      "loss": 0.8918,
       "step": 17871
     },
     {
-      "epoch": 0.4908406800142814,
+      "epoch": 0.5071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.0781974947197524e-05,
-      "loss": 0.9821,
+      "learning_rate": 1.0254557125056926e-05,
+      "loss": 0.9125,
       "step": 17872
     },
     {
-      "epoch": 0.49086814424212466,
+      "epoch": 0.5071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.0781088151890339e-05,
-      "loss": 0.8113,
+      "learning_rate": 1.0253638343741108e-05,
+      "loss": 0.9611,
       "step": 17873
     },
     {
-      "epoch": 0.4908956084699679,
+      "epoch": 0.5072077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.0780201350402883e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.0252719560282795e-05,
+      "loss": 0.8812,
       "step": 17874
     },
     {
-      "epoch": 0.4909230726978111,
+      "epoch": 0.5072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.0779314542742174e-05,
-      "loss": 0.7904,
+      "learning_rate": 1.0251800774689748e-05,
+      "loss": 0.7569,
       "step": 17875
     },
     {
-      "epoch": 0.4909505369256543,
+      "epoch": 0.5072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.0778427728915223e-05,
-      "loss": 0.8803,
+      "learning_rate": 1.0250881986969733e-05,
+      "loss": 0.928,
       "step": 17876
     },
     {
-      "epoch": 0.49097800115349755,
+      "epoch": 0.5072928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 1.0777540908929054e-05,
-      "loss": 0.8267,
+      "learning_rate": 1.0249963197130502e-05,
+      "loss": 0.9272,
       "step": 17877
     },
     {
-      "epoch": 0.4910054653813408,
+      "epoch": 0.5073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.0776654082790681e-05,
-      "loss": 0.8638,
+      "learning_rate": 1.0249044405179818e-05,
+      "loss": 0.8414,
       "step": 17878
     },
     {
-      "epoch": 0.49103292960918404,
+      "epoch": 0.5073496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.0775767250507115e-05,
-      "loss": 0.9077,
+      "learning_rate": 1.024812561112545e-05,
+      "loss": 0.8964,
       "step": 17879
     },
     {
-      "epoch": 0.49106039383702726,
+      "epoch": 0.5073779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 1.077488041208538e-05,
-      "loss": 0.964,
+      "learning_rate": 1.0247206814975153e-05,
+      "loss": 0.8177,
       "step": 17880
     },
     {
-      "epoch": 0.4910878580648705,
+      "epoch": 0.5074063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.077399356753249e-05,
-      "loss": 0.9445,
+      "learning_rate": 1.0246288016736688e-05,
+      "loss": 0.858,
       "step": 17881
     },
     {
-      "epoch": 0.49111532229271376,
+      "epoch": 0.5074347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.0773106716855466e-05,
-      "loss": 0.9149,
+      "learning_rate": 1.0245369216417818e-05,
+      "loss": 0.9702,
       "step": 17882
     },
     {
-      "epoch": 0.491142786520557,
+      "epoch": 0.5074631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.0772219860061322e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.0244450414026301e-05,
+      "loss": 0.8376,
       "step": 17883
     },
     {
-      "epoch": 0.4911702507484002,
+      "epoch": 0.5074914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.0771332997157073e-05,
-      "loss": 0.8658,
+      "learning_rate": 1.02435316095699e-05,
+      "loss": 0.8784,
       "step": 17884
     },
     {
-      "epoch": 0.4911977149762434,
+      "epoch": 0.5075198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.0770446128149741e-05,
-      "loss": 0.8541,
+      "learning_rate": 1.0242612803056378e-05,
+      "loss": 1.0237,
       "step": 17885
     },
     {
-      "epoch": 0.4912251792040867,
+      "epoch": 0.5075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.076955925304634e-05,
-      "loss": 0.9646,
+      "learning_rate": 1.0241693994493496e-05,
+      "loss": 0.79,
       "step": 17886
     },
     {
-      "epoch": 0.4912526434319299,
+      "epoch": 0.5075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.0768672371853885e-05,
-      "loss": 0.9862,
+      "learning_rate": 1.0240775183889013e-05,
+      "loss": 0.9244,
       "step": 17887
     },
     {
-      "epoch": 0.49128010765977315,
+      "epoch": 0.5076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.0767785484579398e-05,
-      "loss": 0.9546,
+      "learning_rate": 1.023985637125069e-05,
+      "loss": 1.0004,
       "step": 17888
     },
     {
-      "epoch": 0.49130757188761637,
+      "epoch": 0.507633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0766898591229898e-05,
-      "loss": 0.9332,
+      "learning_rate": 1.0238937556586292e-05,
+      "loss": 0.924,
       "step": 17889
     },
     {
-      "epoch": 0.4913350361154596,
+      "epoch": 0.5076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.0766011691812396e-05,
-      "loss": 0.9455,
+      "learning_rate": 1.0238018739903576e-05,
+      "loss": 0.8238,
       "step": 17890
     },
     {
-      "epoch": 0.49136250034330287,
+      "epoch": 0.5076901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.0765124786333914e-05,
-      "loss": 0.9734,
+      "learning_rate": 1.0237099921210305e-05,
+      "loss": 0.8818,
       "step": 17891
     },
     {
-      "epoch": 0.4913899645711461,
+      "epoch": 0.5077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.0764237874801468e-05,
-      "loss": 0.8678,
+      "learning_rate": 1.0236181100514242e-05,
+      "loss": 0.8767,
       "step": 17892
     },
     {
-      "epoch": 0.4914174287989893,
+      "epoch": 0.5077468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.0763350957222073e-05,
-      "loss": 0.9897,
+      "learning_rate": 1.0235262277823144e-05,
+      "loss": 0.8725,
       "step": 17893
     },
     {
-      "epoch": 0.49144489302683253,
+      "epoch": 0.5077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.0762464033602755e-05,
-      "loss": 0.9026,
+      "learning_rate": 1.0234343453144777e-05,
+      "loss": 0.9657,
       "step": 17894
     },
     {
-      "epoch": 0.4914723572546758,
+      "epoch": 0.5078036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.0761577103950525e-05,
-      "loss": 0.9156,
+      "learning_rate": 1.0233424626486896e-05,
+      "loss": 0.9641,
       "step": 17895
     },
     {
-      "epoch": 0.49149982148251903,
+      "epoch": 0.5078320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.0760690168272398e-05,
-      "loss": 0.9075,
+      "learning_rate": 1.0232505797857273e-05,
+      "loss": 0.9567,
       "step": 17896
     },
     {
-      "epoch": 0.49152728571036225,
+      "epoch": 0.5078603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.0759803226575401e-05,
-      "loss": 1.0016,
+      "learning_rate": 1.023158696726366e-05,
+      "loss": 0.821,
       "step": 17897
     },
     {
-      "epoch": 0.49155474993820547,
+      "epoch": 0.5078887627695801,
       "grad_norm": 0.0,
-      "learning_rate": 1.0758916278866545e-05,
-      "loss": 0.8886,
+      "learning_rate": 1.0230668134713822e-05,
+      "loss": 1.0566,
       "step": 17898
     },
     {
-      "epoch": 0.49158221416604875,
+      "epoch": 0.5079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.0758029325152845e-05,
-      "loss": 0.8285,
+      "learning_rate": 1.022974930021552e-05,
+      "loss": 0.9499,
       "step": 17899
     },
     {
-      "epoch": 0.49160967839389197,
+      "epoch": 0.5079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.0757142365441327e-05,
-      "loss": 0.8215,
+      "learning_rate": 1.0228830463776514e-05,
+      "loss": 0.8796,
       "step": 17900
     },
     {
-      "epoch": 0.4916371426217352,
+      "epoch": 0.5079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.0756255399739006e-05,
-      "loss": 0.9289,
+      "learning_rate": 1.0227911625404567e-05,
+      "loss": 0.9355,
       "step": 17901
     },
     {
-      "epoch": 0.4916646068495784,
+      "epoch": 0.5080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.0755368428052899e-05,
-      "loss": 0.8517,
+      "learning_rate": 1.0226992785107441e-05,
+      "loss": 0.928,
       "step": 17902
     },
     {
-      "epoch": 0.49169207107742163,
+      "epoch": 0.5080306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.0754481450390022e-05,
-      "loss": 0.9,
+      "learning_rate": 1.0226073942892897e-05,
+      "loss": 0.861,
       "step": 17903
     },
     {
-      "epoch": 0.4917195353052649,
+      "epoch": 0.5080590238365493,
       "grad_norm": 0.0,
-      "learning_rate": 1.0753594466757397e-05,
-      "loss": 0.8694,
+      "learning_rate": 1.0225155098768697e-05,
+      "loss": 0.8835,
       "step": 17904
     },
     {
-      "epoch": 0.49174699953310813,
+      "epoch": 0.5080874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.075270747716204e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.0224236252742598e-05,
+      "loss": 1.0083,
       "step": 17905
     },
     {
-      "epoch": 0.49177446376095135,
+      "epoch": 0.5081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.0751820481610973e-05,
-      "loss": 0.9074,
+      "learning_rate": 1.022331740482237e-05,
+      "loss": 0.9103,
       "step": 17906
     },
     {
-      "epoch": 0.4918019279887946,
+      "epoch": 0.508144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.075093348011121e-05,
-      "loss": 0.9196,
+      "learning_rate": 1.0222398555015769e-05,
+      "loss": 1.0052,
       "step": 17907
     },
     {
-      "epoch": 0.49182939221663785,
+      "epoch": 0.508172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.0750046472669766e-05,
-      "loss": 0.8004,
+      "learning_rate": 1.0221479703330553e-05,
+      "loss": 0.8095,
       "step": 17908
     },
     {
-      "epoch": 0.49185685644448107,
+      "epoch": 0.5082009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.074915945929367e-05,
-      "loss": 0.8964,
+      "learning_rate": 1.0220560849774492e-05,
+      "loss": 0.9549,
       "step": 17909
     },
     {
-      "epoch": 0.4918843206723243,
+      "epoch": 0.5082292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.0748272439989931e-05,
-      "loss": 0.9412,
+      "learning_rate": 1.0219641994355343e-05,
+      "loss": 0.9895,
       "step": 17910
     },
     {
-      "epoch": 0.4919117849001675,
+      "epoch": 0.5082576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.074738541476557e-05,
-      "loss": 0.8218,
+      "learning_rate": 1.0218723137080868e-05,
+      "loss": 0.9496,
       "step": 17911
     },
     {
-      "epoch": 0.4919392491280108,
+      "epoch": 0.5082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.0746498383627606e-05,
-      "loss": 0.9066,
+      "learning_rate": 1.0217804277958829e-05,
+      "loss": 0.8769,
       "step": 17912
     },
     {
-      "epoch": 0.491966713355854,
+      "epoch": 0.5083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.0745611346583059e-05,
-      "loss": 0.8197,
+      "learning_rate": 1.0216885416996989e-05,
+      "loss": 0.9553,
       "step": 17913
     },
     {
-      "epoch": 0.49199417758369723,
+      "epoch": 0.5083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.0744724303638948e-05,
-      "loss": 0.9538,
+      "learning_rate": 1.0215966554203107e-05,
+      "loss": 0.8561,
       "step": 17914
     },
     {
-      "epoch": 0.49202164181154046,
+      "epoch": 0.5083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.0743837254802284e-05,
-      "loss": 0.8937,
+      "learning_rate": 1.0215047689584943e-05,
+      "loss": 0.8339,
       "step": 17915
     },
     {
-      "epoch": 0.4920491060393837,
+      "epoch": 0.5083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.0742950200080097e-05,
-      "loss": 0.9414,
+      "learning_rate": 1.0214128823150263e-05,
+      "loss": 0.918,
       "step": 17916
     },
     {
-      "epoch": 0.49207657026722695,
+      "epoch": 0.508427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.0742063139479398e-05,
-      "loss": 0.8532,
+      "learning_rate": 1.0213209954906829e-05,
+      "loss": 0.8129,
       "step": 17917
     },
     {
-      "epoch": 0.4921040344950702,
+      "epoch": 0.5084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.0741176073007209e-05,
-      "loss": 0.8176,
+      "learning_rate": 1.0212291084862397e-05,
+      "loss": 0.821,
       "step": 17918
     },
     {
-      "epoch": 0.4921314987229134,
+      "epoch": 0.5084846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.0740289000670548e-05,
-      "loss": 0.8511,
+      "learning_rate": 1.0211372213024738e-05,
+      "loss": 0.8298,
       "step": 17919
     },
     {
-      "epoch": 0.4921589629507566,
+      "epoch": 0.5085130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.073940192247643e-05,
-      "loss": 0.8994,
+      "learning_rate": 1.0210453339401607e-05,
+      "loss": 0.9952,
       "step": 17920
     },
     {
-      "epoch": 0.4921864271785999,
+      "epoch": 0.5085414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.0738514838431882e-05,
-      "loss": 0.9924,
+      "learning_rate": 1.0209534464000765e-05,
+      "loss": 0.9431,
       "step": 17921
     },
     {
-      "epoch": 0.4922138914064431,
+      "epoch": 0.5085698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.0737627748543918e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.0208615586829977e-05,
+      "loss": 0.8875,
       "step": 17922
     },
     {
-      "epoch": 0.49224135563428634,
+      "epoch": 0.5085981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.0736740652819556e-05,
-      "loss": 0.994,
+      "learning_rate": 1.0207696707897004e-05,
+      "loss": 0.7724,
       "step": 17923
     },
     {
-      "epoch": 0.49226881986212956,
+      "epoch": 0.5086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.0735853551265813e-05,
-      "loss": 0.9999,
+      "learning_rate": 1.0206777827209609e-05,
+      "loss": 0.9202,
       "step": 17924
     },
     {
-      "epoch": 0.49229628408997284,
+      "epoch": 0.5086549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 1.0734966443889718e-05,
-      "loss": 0.9505,
+      "learning_rate": 1.0205858944775549e-05,
+      "loss": 0.8509,
       "step": 17925
     },
     {
-      "epoch": 0.49232374831781606,
+      "epoch": 0.508683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.0734079330698282e-05,
-      "loss": 0.8028,
+      "learning_rate": 1.020494006060259e-05,
+      "loss": 0.8314,
       "step": 17926
     },
     {
-      "epoch": 0.4923512125456593,
+      "epoch": 0.5087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.0733192211698525e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.0204021174698497e-05,
+      "loss": 0.9349,
       "step": 17927
     },
     {
-      "epoch": 0.4923786767735025,
+      "epoch": 0.5087400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.0732305086897468e-05,
-      "loss": 0.8934,
+      "learning_rate": 1.0203102287071022e-05,
+      "loss": 0.9188,
       "step": 17928
     },
     {
-      "epoch": 0.4924061410013457,
+      "epoch": 0.5087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.0731417956302129e-05,
-      "loss": 0.9173,
+      "learning_rate": 1.0202183397727938e-05,
+      "loss": 0.9608,
       "step": 17929
     },
     {
-      "epoch": 0.492433605229189,
+      "epoch": 0.5087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.0730530819919526e-05,
-      "loss": 0.8362,
+      "learning_rate": 1.0201264506677e-05,
+      "loss": 0.9153,
       "step": 17930
     },
     {
-      "epoch": 0.4924610694570322,
+      "epoch": 0.5088251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.0729643677756685e-05,
-      "loss": 0.9411,
+      "learning_rate": 1.0200345613925971e-05,
+      "loss": 0.9687,
       "step": 17931
     },
     {
-      "epoch": 0.49248853368487544,
+      "epoch": 0.5088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.0728756529820617e-05,
-      "loss": 0.7675,
+      "learning_rate": 1.0199426719482612e-05,
+      "loss": 0.9541,
       "step": 17932
     },
     {
-      "epoch": 0.49251599791271866,
+      "epoch": 0.5088819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.0727869376118344e-05,
-      "loss": 0.8684,
+      "learning_rate": 1.0198507823354692e-05,
+      "loss": 0.8671,
       "step": 17933
     },
     {
-      "epoch": 0.49254346214056194,
+      "epoch": 0.5089103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.072698221665689e-05,
-      "loss": 0.8253,
+      "learning_rate": 1.0197588925549963e-05,
+      "loss": 0.8787,
       "step": 17934
     },
     {
-      "epoch": 0.49257092636840516,
+      "epoch": 0.5089387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 1.0726095051443269e-05,
-      "loss": 0.8211,
+      "learning_rate": 1.0196670026076192e-05,
+      "loss": 0.8426,
       "step": 17935
     },
     {
-      "epoch": 0.4925983905962484,
+      "epoch": 0.5089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.07252078804845e-05,
-      "loss": 0.8762,
+      "learning_rate": 1.0195751124941142e-05,
+      "loss": 0.9374,
       "step": 17936
     },
     {
-      "epoch": 0.4926258548240916,
+      "epoch": 0.5089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.0724320703787612e-05,
-      "loss": 0.894,
+      "learning_rate": 1.0194832222152573e-05,
+      "loss": 1.0037,
       "step": 17937
     },
     {
-      "epoch": 0.4926533190519349,
+      "epoch": 0.5090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.0723433521359614e-05,
-      "loss": 0.8629,
+      "learning_rate": 1.0193913317718245e-05,
+      "loss": 0.8895,
       "step": 17938
     },
     {
-      "epoch": 0.4926807832797781,
+      "epoch": 0.5090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.072254633320753e-05,
-      "loss": 0.9215,
+      "learning_rate": 1.0192994411645923e-05,
+      "loss": 0.8278,
       "step": 17939
     },
     {
-      "epoch": 0.4927082475076213,
+      "epoch": 0.5090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.072165913933838e-05,
-      "loss": 0.9424,
+      "learning_rate": 1.019207550394337e-05,
+      "loss": 1.0155,
       "step": 17940
     },
     {
-      "epoch": 0.49273571173546454,
+      "epoch": 0.5091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.0720771939759182e-05,
-      "loss": 0.933,
+      "learning_rate": 1.0191156594618348e-05,
+      "loss": 0.8425,
       "step": 17941
     },
     {
-      "epoch": 0.49276317596330776,
+      "epoch": 0.5091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.071988473447696e-05,
-      "loss": 0.798,
+      "learning_rate": 1.0190237683678614e-05,
+      "loss": 1.0131,
       "step": 17942
     },
     {
-      "epoch": 0.49279064019115104,
+      "epoch": 0.5091657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.0718997523498729e-05,
-      "loss": 0.835,
+      "learning_rate": 1.0189318771131938e-05,
+      "loss": 0.8899,
       "step": 17943
     },
     {
-      "epoch": 0.49281810441899426,
+      "epoch": 0.509194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.0718110306831511e-05,
-      "loss": 0.8667,
+      "learning_rate": 1.0188399856986077e-05,
+      "loss": 0.9414,
       "step": 17944
     },
     {
-      "epoch": 0.4928455686468375,
+      "epoch": 0.5092224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 1.0717223084482327e-05,
-      "loss": 1.0068,
+      "learning_rate": 1.018748094124879e-05,
+      "loss": 0.7706,
       "step": 17945
     },
     {
-      "epoch": 0.4928730328746807,
+      "epoch": 0.5092508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.0716335856458195e-05,
-      "loss": 0.8455,
+      "learning_rate": 1.0186562023927847e-05,
+      "loss": 0.9085,
       "step": 17946
     },
     {
-      "epoch": 0.492900497102524,
+      "epoch": 0.5092792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.0715448622766137e-05,
-      "loss": 0.9322,
+      "learning_rate": 1.0185643105031005e-05,
+      "loss": 0.9022,
       "step": 17947
     },
     {
-      "epoch": 0.4929279613303672,
+      "epoch": 0.5093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.0714561383413173e-05,
-      "loss": 0.876,
+      "learning_rate": 1.0184724184566028e-05,
+      "loss": 0.8247,
       "step": 17948
     },
     {
-      "epoch": 0.4929554255582104,
+      "epoch": 0.5093359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.071367413840632e-05,
-      "loss": 0.8918,
+      "learning_rate": 1.0183805262540673e-05,
+      "loss": 0.7897,
       "step": 17949
     },
     {
-      "epoch": 0.49298288978605365,
+      "epoch": 0.5093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.0712786887752605e-05,
-      "loss": 0.9297,
+      "learning_rate": 1.0182886338962711e-05,
+      "loss": 0.8762,
       "step": 17950
     },
     {
-      "epoch": 0.4930103540138969,
+      "epoch": 0.5093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.0711899631459041e-05,
-      "loss": 0.7971,
+      "learning_rate": 1.0181967413839901e-05,
+      "loss": 0.8561,
       "step": 17951
     },
     {
-      "epoch": 0.49303781824174014,
+      "epoch": 0.5094211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.0711012369532649e-05,
-      "loss": 0.8245,
+      "learning_rate": 1.0181048487180003e-05,
+      "loss": 0.9678,
       "step": 17952
     },
     {
-      "epoch": 0.49306528246958337,
+      "epoch": 0.5094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.0710125101980457e-05,
-      "loss": 0.9522,
+      "learning_rate": 1.018012955899078e-05,
+      "loss": 0.9381,
       "step": 17953
     },
     {
-      "epoch": 0.4930927466974266,
+      "epoch": 0.509477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.0709237828809478e-05,
-      "loss": 0.9204,
+      "learning_rate": 1.0179210629279992e-05,
+      "loss": 0.9083,
       "step": 17954
     },
     {
-      "epoch": 0.4931202109252698,
+      "epoch": 0.5095062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 1.0708350550026733e-05,
-      "loss": 0.918,
+      "learning_rate": 1.0178291698055408e-05,
+      "loss": 0.9966,
       "step": 17955
     },
     {
-      "epoch": 0.4931476751531131,
+      "epoch": 0.5095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.0707463265639244e-05,
-      "loss": 1.032,
+      "learning_rate": 1.0177372765324782e-05,
+      "loss": 0.9201,
       "step": 17956
     },
     {
-      "epoch": 0.4931751393809563,
+      "epoch": 0.5095629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.0706575975654035e-05,
-      "loss": 0.9614,
+      "learning_rate": 1.0176453831095883e-05,
+      "loss": 0.7977,
       "step": 17957
     },
     {
-      "epoch": 0.49320260360879953,
+      "epoch": 0.5095913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.070568868007812e-05,
-      "loss": 0.9219,
+      "learning_rate": 1.0175534895376471e-05,
+      "loss": 0.9291,
       "step": 17958
     },
     {
-      "epoch": 0.49323006783664275,
+      "epoch": 0.5096197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.0704801378918523e-05,
-      "loss": 0.973,
+      "learning_rate": 1.0174615958174305e-05,
+      "loss": 0.8663,
       "step": 17959
     },
     {
-      "epoch": 0.493257532064486,
+      "epoch": 0.5096481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.0703914072182267e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.0173697019497152e-05,
+      "loss": 0.8383,
       "step": 17960
     },
     {
-      "epoch": 0.49328499629232925,
+      "epoch": 0.5096765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.0703026759876367e-05,
-      "loss": 0.7937,
+      "learning_rate": 1.0172778079352774e-05,
+      "loss": 0.8988,
       "step": 17961
     },
     {
-      "epoch": 0.49331246052017247,
+      "epoch": 0.5097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.070213944200785e-05,
-      "loss": 0.8289,
+      "learning_rate": 1.0171859137748927e-05,
+      "loss": 0.9945,
       "step": 17962
     },
     {
-      "epoch": 0.4933399247480157,
+      "epoch": 0.509733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.0701252118583734e-05,
-      "loss": 0.936,
+      "learning_rate": 1.0170940194693381e-05,
+      "loss": 0.8102,
       "step": 17963
     },
     {
-      "epoch": 0.49336738897585897,
+      "epoch": 0.5097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.0700364789611035e-05,
-      "loss": 0.8896,
+      "learning_rate": 1.0170021250193898e-05,
+      "loss": 0.9615,
       "step": 17964
     },
     {
-      "epoch": 0.4933948532037022,
+      "epoch": 0.5097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.0699477455096783e-05,
-      "loss": 0.9234,
+      "learning_rate": 1.0169102304258234e-05,
+      "loss": 0.954,
       "step": 17965
     },
     {
-      "epoch": 0.4934223174315454,
+      "epoch": 0.5098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.0698590115047995e-05,
-      "loss": 0.9667,
+      "learning_rate": 1.0168183356894156e-05,
+      "loss": 0.9538,
       "step": 17966
     },
     {
-      "epoch": 0.49344978165938863,
+      "epoch": 0.5098467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.0697702769471687e-05,
-      "loss": 0.9617,
+      "learning_rate": 1.016726440810943e-05,
+      "loss": 0.8732,
       "step": 17967
     },
     {
-      "epoch": 0.4934772458872319,
+      "epoch": 0.5098751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.0696815418374887e-05,
-      "loss": 0.9552,
+      "learning_rate": 1.0166345457911811e-05,
+      "loss": 0.7964,
       "step": 17968
     },
     {
-      "epoch": 0.49350471011507513,
+      "epoch": 0.5099035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.0695928061764613e-05,
-      "loss": 0.8481,
+      "learning_rate": 1.0165426506309059e-05,
+      "loss": 0.9208,
       "step": 17969
     },
     {
-      "epoch": 0.49353217434291835,
+      "epoch": 0.5099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.0695040699647887e-05,
-      "loss": 0.9736,
+      "learning_rate": 1.0164507553308948e-05,
+      "loss": 0.8529,
       "step": 17970
     },
     {
-      "epoch": 0.49355963857076157,
+      "epoch": 0.5099602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.069415333203173e-05,
-      "loss": 0.9003,
+      "learning_rate": 1.0163588598919234e-05,
+      "loss": 0.8747,
       "step": 17971
     },
     {
-      "epoch": 0.4935871027986048,
+      "epoch": 0.509988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.0693265958923163e-05,
-      "loss": 0.8901,
+      "learning_rate": 1.0162669643147679e-05,
+      "loss": 0.9502,
       "step": 17972
     },
     {
-      "epoch": 0.49361456702644807,
+      "epoch": 0.510017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.0692378580329204e-05,
-      "loss": 0.9088,
+      "learning_rate": 1.0161750686002045e-05,
+      "loss": 0.8815,
       "step": 17973
     },
     {
-      "epoch": 0.4936420312542913,
+      "epoch": 0.5100454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.069149119625688e-05,
-      "loss": 0.9044,
+      "learning_rate": 1.0160831727490097e-05,
+      "loss": 0.942,
       "step": 17974
     },
     {
-      "epoch": 0.4936694954821345,
+      "epoch": 0.5100737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.069060380671321e-05,
-      "loss": 0.9783,
+      "learning_rate": 1.0159912767619595e-05,
+      "loss": 0.9092,
       "step": 17975
     },
     {
-      "epoch": 0.49369695970997773,
+      "epoch": 0.5101021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 1.0689716411705215e-05,
-      "loss": 0.8977,
+      "learning_rate": 1.0158993806398305e-05,
+      "loss": 0.9276,
       "step": 17976
     },
     {
-      "epoch": 0.493724423937821,
+      "epoch": 0.5101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.0688829011239918e-05,
-      "loss": 0.8836,
+      "learning_rate": 1.0158074843833984e-05,
+      "loss": 0.9245,
       "step": 17977
     },
     {
-      "epoch": 0.49375188816566423,
+      "epoch": 0.5101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.0687941605324339e-05,
-      "loss": 0.8705,
+      "learning_rate": 1.0157155879934399e-05,
+      "loss": 0.8678,
       "step": 17978
     },
     {
-      "epoch": 0.49377935239350745,
+      "epoch": 0.5101872871736662,
       "grad_norm": 0.0,
-      "learning_rate": 1.0687054193965495e-05,
-      "loss": 0.9168,
+      "learning_rate": 1.0156236914707312e-05,
+      "loss": 0.9382,
       "step": 17979
     },
     {
-      "epoch": 0.4938068166213507,
+      "epoch": 0.5102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.0686166777170415e-05,
-      "loss": 0.8141,
+      "learning_rate": 1.0155317948160486e-05,
+      "loss": 0.8238,
       "step": 17980
     },
     {
-      "epoch": 0.49383428084919395,
+      "epoch": 0.5102440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.0685279354946119e-05,
-      "loss": 0.9386,
+      "learning_rate": 1.015439898030168e-05,
+      "loss": 0.8763,
       "step": 17981
     },
     {
-      "epoch": 0.4938617450770372,
+      "epoch": 0.5102724177071509,
       "grad_norm": 0.0,
-      "learning_rate": 1.0684391927299628e-05,
-      "loss": 0.8648,
+      "learning_rate": 1.0153480011138658e-05,
+      "loss": 0.8159,
       "step": 17982
     },
     {
-      "epoch": 0.4938892093048804,
+      "epoch": 0.5103007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.068350449423796e-05,
-      "loss": 0.8939,
+      "learning_rate": 1.0152561040679188e-05,
+      "loss": 0.8832,
       "step": 17983
     },
     {
-      "epoch": 0.4939166735327236,
+      "epoch": 0.5103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.0682617055768141e-05,
-      "loss": 0.8311,
+      "learning_rate": 1.0151642068931024e-05,
+      "loss": 1.016,
       "step": 17984
     },
     {
-      "epoch": 0.49394413776056684,
+      "epoch": 0.5103575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.0681729611897191e-05,
-      "loss": 0.8592,
+      "learning_rate": 1.0150723095901933e-05,
+      "loss": 0.9758,
       "step": 17985
     },
     {
-      "epoch": 0.4939716019884101,
+      "epoch": 0.5103859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.0680842162632133e-05,
-      "loss": 0.8194,
+      "learning_rate": 1.0149804121599675e-05,
+      "loss": 0.7864,
       "step": 17986
     },
     {
-      "epoch": 0.49399906621625334,
+      "epoch": 0.5104143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.0679954707979989e-05,
-      "loss": 0.8253,
+      "learning_rate": 1.014888514603202e-05,
+      "loss": 0.9296,
       "step": 17987
     },
     {
-      "epoch": 0.49402653044409656,
+      "epoch": 0.5104426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.0679067247947776e-05,
-      "loss": 0.9485,
+      "learning_rate": 1.0147966169206724e-05,
+      "loss": 0.8212,
       "step": 17988
     },
     {
-      "epoch": 0.4940539946719398,
+      "epoch": 0.5104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.0678179782542523e-05,
-      "loss": 0.8594,
+      "learning_rate": 1.014704719113155e-05,
+      "loss": 0.8202,
       "step": 17989
     },
     {
-      "epoch": 0.49408145889978305,
+      "epoch": 0.5104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.067729231177125e-05,
-      "loss": 0.9029,
+      "learning_rate": 1.014612821181426e-05,
+      "loss": 0.7876,
       "step": 17990
     },
     {
-      "epoch": 0.4941089231276263,
+      "epoch": 0.510527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.0676404835640975e-05,
-      "loss": 0.8338,
+      "learning_rate": 1.0145209231262623e-05,
+      "loss": 0.9656,
       "step": 17991
     },
     {
-      "epoch": 0.4941363873554695,
+      "epoch": 0.5105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.0675517354158722e-05,
-      "loss": 0.8211,
+      "learning_rate": 1.0144290249484395e-05,
+      "loss": 0.8493,
       "step": 17992
     },
     {
-      "epoch": 0.4941638515833127,
+      "epoch": 0.5105845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.0674629867331518e-05,
-      "loss": 0.9209,
+      "learning_rate": 1.0143371266487339e-05,
+      "loss": 0.8776,
       "step": 17993
     },
     {
-      "epoch": 0.494191315811156,
+      "epoch": 0.5106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.067374237516638e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.014245228227922e-05,
+      "loss": 0.7842,
       "step": 17994
     },
     {
-      "epoch": 0.4942187800389992,
+      "epoch": 0.5106413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.0672854877670329e-05,
-      "loss": 0.8428,
+      "learning_rate": 1.0141533296867802e-05,
+      "loss": 0.8016,
       "step": 17995
     },
     {
-      "epoch": 0.49424624426684244,
+      "epoch": 0.5106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.0671967374850394e-05,
-      "loss": 0.8578,
+      "learning_rate": 1.0140614310260844e-05,
+      "loss": 0.8559,
       "step": 17996
     },
     {
-      "epoch": 0.49427370849468566,
+      "epoch": 0.5106980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.0671079866713588e-05,
-      "loss": 0.8947,
+      "learning_rate": 1.0139695322466112e-05,
+      "loss": 0.9132,
       "step": 17997
     },
     {
-      "epoch": 0.4943011727225289,
+      "epoch": 0.5107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.0670192353266942e-05,
-      "loss": 0.9409,
+      "learning_rate": 1.0138776333491369e-05,
+      "loss": 1.0073,
       "step": 17998
     },
     {
-      "epoch": 0.49432863695037216,
+      "epoch": 0.5107548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.0669304834517474e-05,
-      "loss": 0.9203,
+      "learning_rate": 1.0137857343344372e-05,
+      "loss": 0.8256,
       "step": 17999
     },
     {
-      "epoch": 0.4943561011782154,
+      "epoch": 0.5107832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 1.0668417310472203e-05,
-      "loss": 0.9272,
+      "learning_rate": 1.0136938352032888e-05,
+      "loss": 0.7901,
       "step": 18000
     },
     {
-      "epoch": 0.4943835654060586,
+      "epoch": 0.5108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.066752978113816e-05,
-      "loss": 0.9745,
+      "learning_rate": 1.0136019359564681e-05,
+      "loss": 1.0144,
       "step": 18001
     },
     {
-      "epoch": 0.4944110296339018,
+      "epoch": 0.5108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.066664224652236e-05,
-      "loss": 0.9852,
+      "learning_rate": 1.0135100365947514e-05,
+      "loss": 0.9183,
       "step": 18002
     },
     {
-      "epoch": 0.4944384938617451,
+      "epoch": 0.5108683314415438,
       "grad_norm": 0.0,
-      "learning_rate": 1.0665754706631828e-05,
-      "loss": 0.8687,
+      "learning_rate": 1.0134181371189146e-05,
+      "loss": 0.8738,
       "step": 18003
     },
     {
-      "epoch": 0.4944659580895883,
+      "epoch": 0.5108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.0664867161473589e-05,
-      "loss": 0.9081,
+      "learning_rate": 1.0133262375297344e-05,
+      "loss": 0.8289,
       "step": 18004
     },
     {
-      "epoch": 0.49449342231743154,
+      "epoch": 0.5109250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.066397961105466e-05,
-      "loss": 0.8962,
+      "learning_rate": 1.0132343378279868e-05,
+      "loss": 0.7979,
       "step": 18005
     },
     {
-      "epoch": 0.49452088654527476,
+      "epoch": 0.5109534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.0663092055382071e-05,
-      "loss": 0.8923,
+      "learning_rate": 1.0131424380144482e-05,
+      "loss": 0.8083,
       "step": 18006
     },
     {
-      "epoch": 0.49454835077311804,
+      "epoch": 0.5109818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.066220449446284e-05,
-      "loss": 0.8613,
+      "learning_rate": 1.0130505380898947e-05,
+      "loss": 0.863,
       "step": 18007
     },
     {
-      "epoch": 0.49457581500096126,
+      "epoch": 0.5110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.0661316928303988e-05,
-      "loss": 0.8011,
+      "learning_rate": 1.0129586380551026e-05,
+      "loss": 0.9553,
       "step": 18008
     },
     {
-      "epoch": 0.4946032792288045,
+      "epoch": 0.511038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.0660429356912543e-05,
-      "loss": 0.9474,
+      "learning_rate": 1.0128667379108486e-05,
+      "loss": 0.8688,
       "step": 18009
     },
     {
-      "epoch": 0.4946307434566477,
+      "epoch": 0.511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.0659541780295525e-05,
-      "loss": 0.9012,
+      "learning_rate": 1.0127748376579082e-05,
+      "loss": 0.7527,
       "step": 18010
     },
     {
-      "epoch": 0.4946582076844909,
+      "epoch": 0.5110953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 1.0658654198459956e-05,
-      "loss": 0.8969,
+      "learning_rate": 1.0126829372970588e-05,
+      "loss": 0.9532,
       "step": 18011
     },
     {
-      "epoch": 0.4946856719123342,
+      "epoch": 0.5111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.0657766611412859e-05,
-      "loss": 0.9422,
+      "learning_rate": 1.0125910368290757e-05,
+      "loss": 0.8458,
       "step": 18012
     },
     {
-      "epoch": 0.4947131361401774,
+      "epoch": 0.5111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.065687901916126e-05,
-      "loss": 0.9725,
+      "learning_rate": 1.0124991362547356e-05,
+      "loss": 0.8569,
       "step": 18013
     },
     {
-      "epoch": 0.49474060036802064,
+      "epoch": 0.5111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.065599142171218e-05,
-      "loss": 0.9878,
+      "learning_rate": 1.012407235574815e-05,
+      "loss": 0.8107,
       "step": 18014
     },
     {
-      "epoch": 0.49476806459586387,
+      "epoch": 0.5112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.065510381907264e-05,
-      "loss": 0.8898,
+      "learning_rate": 1.0123153347900893e-05,
+      "loss": 0.9334,
       "step": 18015
     },
     {
-      "epoch": 0.49479552882370714,
+      "epoch": 0.5112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.0654216211249664e-05,
-      "loss": 0.8998,
+      "learning_rate": 1.0122234339013357e-05,
+      "loss": 0.8643,
       "step": 18016
     },
     {
-      "epoch": 0.49482299305155036,
+      "epoch": 0.5112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.0653328598250277e-05,
-      "loss": 0.94,
+      "learning_rate": 1.0121315329093305e-05,
+      "loss": 0.8469,
       "step": 18017
     },
     {
-      "epoch": 0.4948504572793936,
+      "epoch": 0.5112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.06524409800815e-05,
-      "loss": 0.9377,
+      "learning_rate": 1.0120396318148495e-05,
+      "loss": 0.8539,
       "step": 18018
     },
     {
-      "epoch": 0.4948779215072368,
+      "epoch": 0.511322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.065155335675036e-05,
-      "loss": 0.8961,
+      "learning_rate": 1.0119477306186688e-05,
+      "loss": 0.9925,
       "step": 18019
     },
     {
-      "epoch": 0.4949053857350801,
+      "epoch": 0.5113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.0650665728263873e-05,
-      "loss": 0.9959,
+      "learning_rate": 1.0118558293215657e-05,
+      "loss": 0.9393,
       "step": 18020
     },
     {
-      "epoch": 0.4949328499629233,
+      "epoch": 0.5113791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.0649778094629071e-05,
-      "loss": 0.9019,
+      "learning_rate": 1.0117639279243159e-05,
+      "loss": 0.892,
       "step": 18021
     },
     {
-      "epoch": 0.4949603141907665,
+      "epoch": 0.5114074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.0648890455852974e-05,
-      "loss": 0.834,
+      "learning_rate": 1.0116720264276953e-05,
+      "loss": 0.8087,
       "step": 18022
     },
     {
-      "epoch": 0.49498777841860975,
+      "epoch": 0.5114358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.0648002811942599e-05,
-      "loss": 0.9575,
+      "learning_rate": 1.0115801248324803e-05,
+      "loss": 0.8523,
       "step": 18023
     },
     {
-      "epoch": 0.49501524264645297,
+      "epoch": 0.5114642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.0647115162904977e-05,
-      "loss": 0.92,
+      "learning_rate": 1.0114882231394481e-05,
+      "loss": 0.9163,
       "step": 18024
     },
     {
-      "epoch": 0.49504270687429625,
+      "epoch": 0.5114926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.0646227508747131e-05,
-      "loss": 0.8377,
+      "learning_rate": 1.011396321349374e-05,
+      "loss": 0.9652,
       "step": 18025
     },
     {
-      "epoch": 0.49507017110213947,
+      "epoch": 0.5115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.0645339849476081e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.0113044194630348e-05,
+      "loss": 0.9004,
       "step": 18026
     },
     {
-      "epoch": 0.4950976353299827,
+      "epoch": 0.5115493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 1.0644452185098852e-05,
-      "loss": 0.8551,
+      "learning_rate": 1.0112125174812067e-05,
+      "loss": 0.8723,
       "step": 18027
     },
     {
-      "epoch": 0.4951250995578259,
+      "epoch": 0.511577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.064356451562247e-05,
-      "loss": 0.8938,
+      "learning_rate": 1.011120615404666e-05,
+      "loss": 0.8568,
       "step": 18028
     },
     {
-      "epoch": 0.4951525637856692,
+      "epoch": 0.5116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.0642676841053952e-05,
-      "loss": 0.9102,
+      "learning_rate": 1.0110287132341888e-05,
+      "loss": 0.8118,
       "step": 18029
     },
     {
-      "epoch": 0.4951800280135124,
+      "epoch": 0.5116345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 1.064178916140033e-05,
-      "loss": 0.871,
+      "learning_rate": 1.0109368109705514e-05,
+      "loss": 0.9505,
       "step": 18030
     },
     {
-      "epoch": 0.49520749224135563,
+      "epoch": 0.5116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.064090147666862e-05,
-      "loss": 0.9125,
+      "learning_rate": 1.0108449086145304e-05,
+      "loss": 0.9075,
       "step": 18031
     },
     {
-      "epoch": 0.49523495646919885,
+      "epoch": 0.5116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.0640013786865853e-05,
-      "loss": 0.9605,
+      "learning_rate": 1.0107530061669021e-05,
+      "loss": 0.9565,
       "step": 18032
     },
     {
-      "epoch": 0.4952624206970421,
+      "epoch": 0.5117196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.0639126091999048e-05,
-      "loss": 0.9163,
+      "learning_rate": 1.0106611036284426e-05,
+      "loss": 0.8434,
       "step": 18033
     },
     {
-      "epoch": 0.49528988492488535,
+      "epoch": 0.5117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.0638238392075229e-05,
-      "loss": 0.8423,
+      "learning_rate": 1.0105692009999284e-05,
+      "loss": 0.8486,
       "step": 18034
     },
     {
-      "epoch": 0.49531734915272857,
+      "epoch": 0.5117763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.063735068710142e-05,
-      "loss": 0.877,
+      "learning_rate": 1.0104772982821357e-05,
+      "loss": 0.8669,
       "step": 18035
     },
     {
-      "epoch": 0.4953448133805718,
+      "epoch": 0.5118047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.0636462977084644e-05,
-      "loss": 0.8595,
+      "learning_rate": 1.0103853954758404e-05,
+      "loss": 0.8067,
       "step": 18036
     },
     {
-      "epoch": 0.495372277608415,
+      "epoch": 0.51183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.063557526203193e-05,
-      "loss": 0.8928,
+      "learning_rate": 1.0102934925818198e-05,
+      "loss": 0.924,
       "step": 18037
     },
     {
-      "epoch": 0.4953997418362583,
+      "epoch": 0.511861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.0634687541950297e-05,
-      "loss": 0.8889,
+      "learning_rate": 1.010201589600849e-05,
+      "loss": 0.8915,
       "step": 18038
     },
     {
-      "epoch": 0.4954272060641015,
+      "epoch": 0.5118898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.0633799816846769e-05,
-      "loss": 0.9438,
+      "learning_rate": 1.0101096865337053e-05,
+      "loss": 0.8747,
       "step": 18039
     },
     {
-      "epoch": 0.49545467029194473,
+      "epoch": 0.5119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.0632912086728373e-05,
-      "loss": 0.9525,
+      "learning_rate": 1.0100177833811645e-05,
+      "loss": 1.0014,
       "step": 18040
     },
     {
-      "epoch": 0.49548213451978795,
+      "epoch": 0.5119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.0632024351602131e-05,
-      "loss": 0.9138,
+      "learning_rate": 1.009925880144003e-05,
+      "loss": 0.8512,
       "step": 18041
     },
     {
-      "epoch": 0.49550959874763123,
+      "epoch": 0.5119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.0631136611475066e-05,
-      "loss": 0.9371,
+      "learning_rate": 1.0098339768229972e-05,
+      "loss": 0.9102,
       "step": 18042
     },
     {
-      "epoch": 0.49553706297547445,
+      "epoch": 0.5120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.0630248866354207e-05,
-      "loss": 0.9217,
+      "learning_rate": 1.0097420734189233e-05,
+      "loss": 0.862,
       "step": 18043
     },
     {
-      "epoch": 0.4955645272033177,
+      "epoch": 0.5120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.062936111624657e-05,
-      "loss": 0.8849,
+      "learning_rate": 1.0096501699325578e-05,
+      "loss": 0.8738,
       "step": 18044
     },
     {
-      "epoch": 0.4955919914311609,
+      "epoch": 0.5120601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.0628473361159188e-05,
-      "loss": 0.8316,
+      "learning_rate": 1.009558266364677e-05,
+      "loss": 0.9199,
       "step": 18045
     },
     {
-      "epoch": 0.49561945565900417,
+      "epoch": 0.512088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.0627585601099081e-05,
-      "loss": 0.8745,
+      "learning_rate": 1.0094663627160566e-05,
+      "loss": 0.7993,
       "step": 18046
     },
     {
-      "epoch": 0.4956469198868474,
+      "epoch": 0.5121169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 1.0626697836073272e-05,
-      "loss": 0.8236,
+      "learning_rate": 1.0093744589874737e-05,
+      "loss": 0.9631,
       "step": 18047
     },
     {
-      "epoch": 0.4956743841146906,
+      "epoch": 0.5121452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 1.0625810066088787e-05,
-      "loss": 0.9211,
+      "learning_rate": 1.0092825551797043e-05,
+      "loss": 0.9889,
       "step": 18048
     },
     {
-      "epoch": 0.49570184834253384,
+      "epoch": 0.5121736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.0624922291152653e-05,
-      "loss": 0.9596,
+      "learning_rate": 1.009190651293525e-05,
+      "loss": 0.8733,
       "step": 18049
     },
     {
-      "epoch": 0.49572931257037706,
+      "epoch": 0.5122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.062403451127189e-05,
-      "loss": 0.8492,
+      "learning_rate": 1.0090987473297113e-05,
+      "loss": 0.8918,
       "step": 18050
     },
     {
-      "epoch": 0.49575677679822033,
+      "epoch": 0.5122304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 1.0623146726453522e-05,
-      "loss": 0.9097,
+      "learning_rate": 1.0090068432890405e-05,
+      "loss": 0.8711,
       "step": 18051
     },
     {
-      "epoch": 0.49578424102606355,
+      "epoch": 0.5122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.062225893670458e-05,
-      "loss": 0.8441,
+      "learning_rate": 1.0089149391722886e-05,
+      "loss": 0.8819,
       "step": 18052
     },
     {
-      "epoch": 0.4958117052539068,
+      "epoch": 0.5122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.0621371142032079e-05,
-      "loss": 0.7348,
+      "learning_rate": 1.0088230349802315e-05,
+      "loss": 0.9167,
       "step": 18053
     },
     {
-      "epoch": 0.49583916948175,
+      "epoch": 0.5123155505107831,
       "grad_norm": 0.0,
-      "learning_rate": 1.0620483342443053e-05,
-      "loss": 0.839,
+      "learning_rate": 1.0087311307136457e-05,
+      "loss": 0.9593,
       "step": 18054
     },
     {
-      "epoch": 0.4958666337095933,
+      "epoch": 0.5123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.0619595537944523e-05,
-      "loss": 0.9779,
+      "learning_rate": 1.008639226373308e-05,
+      "loss": 0.867,
       "step": 18055
     },
     {
-      "epoch": 0.4958940979374365,
+      "epoch": 0.512372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.0618707728543509e-05,
-      "loss": 0.931,
+      "learning_rate": 1.0085473219599942e-05,
+      "loss": 0.8568,
       "step": 18056
     },
     {
-      "epoch": 0.4959215621652797,
+      "epoch": 0.5124006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 1.0617819914247045e-05,
-      "loss": 0.9146,
+      "learning_rate": 1.0084554174744807e-05,
+      "loss": 0.8483,
       "step": 18057
     },
     {
-      "epoch": 0.49594902639312294,
+      "epoch": 0.5124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.0616932095062148e-05,
-      "loss": 0.7918,
+      "learning_rate": 1.0083635129175442e-05,
+      "loss": 0.8723,
       "step": 18058
     },
     {
-      "epoch": 0.4959764906209662,
+      "epoch": 0.5124574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.0616044270995845e-05,
-      "loss": 0.8827,
+      "learning_rate": 1.0082716082899605e-05,
+      "loss": 0.808,
       "step": 18059
     },
     {
-      "epoch": 0.49600395484880944,
+      "epoch": 0.5124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.0615156442055161e-05,
-      "loss": 0.829,
+      "learning_rate": 1.0081797035925063e-05,
+      "loss": 0.864,
       "step": 18060
     },
     {
-      "epoch": 0.49603141907665266,
+      "epoch": 0.5125141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.061426860824712e-05,
-      "loss": 0.8372,
+      "learning_rate": 1.0080877988259575e-05,
+      "loss": 0.8051,
       "step": 18061
     },
     {
-      "epoch": 0.4960588833044959,
+      "epoch": 0.5125425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.061338076957875e-05,
-      "loss": 0.8985,
+      "learning_rate": 1.007995893991091e-05,
+      "loss": 0.9171,
       "step": 18062
     },
     {
-      "epoch": 0.49608634753233916,
+      "epoch": 0.5125709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.0612492926057073e-05,
-      "loss": 0.9259,
+      "learning_rate": 1.0079039890886826e-05,
+      "loss": 0.911,
       "step": 18063
     },
     {
-      "epoch": 0.4961138117601824,
+      "epoch": 0.5125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.0611605077689114e-05,
-      "loss": 0.7769,
+      "learning_rate": 1.0078120841195088e-05,
+      "loss": 0.9938,
       "step": 18064
     },
     {
-      "epoch": 0.4961412759880256,
+      "epoch": 0.512627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.0610717224481896e-05,
-      "loss": 0.829,
+      "learning_rate": 1.0077201790843463e-05,
+      "loss": 0.9756,
       "step": 18065
     },
     {
-      "epoch": 0.4961687402158688,
+      "epoch": 0.5126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.0609829366442452e-05,
-      "loss": 0.8755,
+      "learning_rate": 1.0076282739839711e-05,
+      "loss": 0.796,
       "step": 18066
     },
     {
-      "epoch": 0.49619620444371204,
+      "epoch": 0.5126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.0608941503577799e-05,
-      "loss": 0.9508,
+      "learning_rate": 1.0075363688191593e-05,
+      "loss": 0.8727,
       "step": 18067
     },
     {
-      "epoch": 0.4962236686715553,
+      "epoch": 0.5127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.0608053635894963e-05,
-      "loss": 0.9055,
+      "learning_rate": 1.0074444635906875e-05,
+      "loss": 0.8,
       "step": 18068
     },
     {
-      "epoch": 0.49625113289939854,
+      "epoch": 0.5127412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.0607165763400978e-05,
-      "loss": 0.9241,
+      "learning_rate": 1.0073525582993317e-05,
+      "loss": 0.8166,
       "step": 18069
     },
     {
-      "epoch": 0.49627859712724176,
+      "epoch": 0.5127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.0606277886102856e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.0072606529458688e-05,
+      "loss": 0.8911,
       "step": 18070
     },
     {
-      "epoch": 0.496306061355085,
+      "epoch": 0.5127979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.0605390004007629e-05,
-      "loss": 0.9212,
+      "learning_rate": 1.007168747531075e-05,
+      "loss": 0.9487,
       "step": 18071
     },
     {
-      "epoch": 0.49633352558292826,
+      "epoch": 0.5128263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.0604502117122326e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.0070768420557263e-05,
+      "loss": 0.876,
       "step": 18072
     },
     {
-      "epoch": 0.4963609898107715,
+      "epoch": 0.5128547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.0603614225453965e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.006984936520599e-05,
+      "loss": 0.8808,
       "step": 18073
     },
     {
-      "epoch": 0.4963884540386147,
+      "epoch": 0.512883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.0602726329009574e-05,
-      "loss": 0.9099,
+      "learning_rate": 1.00689303092647e-05,
+      "loss": 0.8869,
       "step": 18074
     },
     {
-      "epoch": 0.4964159182664579,
+      "epoch": 0.512911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.060183842779618e-05,
-      "loss": 0.9635,
+      "learning_rate": 1.0068011252741153e-05,
+      "loss": 0.8585,
       "step": 18075
     },
     {
-      "epoch": 0.4964433824943012,
+      "epoch": 0.5129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.0600950521820806e-05,
-      "loss": 0.8522,
+      "learning_rate": 1.0067092195643108e-05,
+      "loss": 0.838,
       "step": 18076
     },
     {
-      "epoch": 0.4964708467221444,
+      "epoch": 0.5129682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.0600062611090478e-05,
-      "loss": 0.9235,
+      "learning_rate": 1.0066173137978333e-05,
+      "loss": 0.891,
       "step": 18077
     },
     {
-      "epoch": 0.49649831094998764,
+      "epoch": 0.5129965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.0599174695612227e-05,
-      "loss": 0.9164,
+      "learning_rate": 1.0065254079754592e-05,
+      "loss": 0.837,
       "step": 18078
     },
     {
-      "epoch": 0.49652577517783086,
+      "epoch": 0.5130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.0598286775393069e-05,
-      "loss": 0.8819,
+      "learning_rate": 1.0064335020979648e-05,
+      "loss": 0.899,
       "step": 18079
     },
     {
-      "epoch": 0.4965532394056741,
+      "epoch": 0.5130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.0597398850440035e-05,
-      "loss": 0.8423,
+      "learning_rate": 1.0063415961661257e-05,
+      "loss": 0.8649,
       "step": 18080
     },
     {
-      "epoch": 0.49658070363351736,
+      "epoch": 0.5130817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.059651092076015e-05,
-      "loss": 0.9375,
+      "learning_rate": 1.0062496901807195e-05,
+      "loss": 0.8277,
       "step": 18081
     },
     {
-      "epoch": 0.4966081678613606,
+      "epoch": 0.5131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.059562298636044e-05,
-      "loss": 0.9233,
+      "learning_rate": 1.0061577841425219e-05,
+      "loss": 1.0065,
       "step": 18082
     },
     {
-      "epoch": 0.4966356320892038,
+      "epoch": 0.5131384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.0594735047247931e-05,
-      "loss": 0.9223,
+      "learning_rate": 1.0060658780523089e-05,
+      "loss": 0.9437,
       "step": 18083
     },
     {
-      "epoch": 0.496663096317047,
+      "epoch": 0.5131668558456299,
       "grad_norm": 0.0,
-      "learning_rate": 1.0593847103429649e-05,
-      "loss": 0.8085,
+      "learning_rate": 1.005973971910857e-05,
+      "loss": 0.9004,
       "step": 18084
     },
     {
-      "epoch": 0.4966905605448903,
+      "epoch": 0.5131952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.0592959154912614e-05,
-      "loss": 0.8201,
+      "learning_rate": 1.0058820657189432e-05,
+      "loss": 0.9533,
       "step": 18085
     },
     {
-      "epoch": 0.4967180247727335,
+      "epoch": 0.5132236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.059207120170386e-05,
-      "loss": 0.9078,
+      "learning_rate": 1.0057901594773432e-05,
+      "loss": 0.8754,
       "step": 18086
     },
     {
-      "epoch": 0.49674548900057675,
+      "epoch": 0.5132519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.059118324381041e-05,
-      "loss": 0.9209,
+      "learning_rate": 1.0056982531868331e-05,
+      "loss": 0.9239,
       "step": 18087
     },
     {
-      "epoch": 0.49677295322841997,
+      "epoch": 0.5132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.0590295281239286e-05,
-      "loss": 1.0067,
+      "learning_rate": 1.0056063468481901e-05,
+      "loss": 0.9039,
       "step": 18088
     },
     {
-      "epoch": 0.49680041745626324,
+      "epoch": 0.5133087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.0589407313997517e-05,
-      "loss": 0.8265,
+      "learning_rate": 1.0055144404621898e-05,
+      "loss": 0.9131,
       "step": 18089
     },
     {
-      "epoch": 0.49682788168410646,
+      "epoch": 0.5133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.0588519342092136e-05,
-      "loss": 0.9583,
+      "learning_rate": 1.0054225340296086e-05,
+      "loss": 0.879,
       "step": 18090
     },
     {
-      "epoch": 0.4968553459119497,
+      "epoch": 0.5133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.0587631365530155e-05,
-      "loss": 0.8531,
+      "learning_rate": 1.0053306275512234e-05,
+      "loss": 0.9073,
       "step": 18091
     },
     {
-      "epoch": 0.4968828101397929,
+      "epoch": 0.5133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.0586743384318607e-05,
-      "loss": 0.8849,
+      "learning_rate": 1.0052387210278096e-05,
+      "loss": 0.7966,
       "step": 18092
     },
     {
-      "epoch": 0.49691027436763613,
+      "epoch": 0.513422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.0585855398464522e-05,
-      "loss": 0.8809,
+      "learning_rate": 1.0051468144601446e-05,
+      "loss": 0.7853,
       "step": 18093
     },
     {
-      "epoch": 0.4969377385954794,
+      "epoch": 0.5134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.058496740797492e-05,
-      "loss": 0.8427,
+      "learning_rate": 1.0050549078490037e-05,
+      "loss": 0.7211,
       "step": 18094
     },
     {
-      "epoch": 0.4969652028233226,
+      "epoch": 0.5134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.0584079412856828e-05,
-      "loss": 0.9961,
+      "learning_rate": 1.0049630011951643e-05,
+      "loss": 0.8405,
       "step": 18095
     },
     {
-      "epoch": 0.49699266705116585,
+      "epoch": 0.5135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.0583191413117275e-05,
-      "loss": 0.8251,
+      "learning_rate": 1.004871094499402e-05,
+      "loss": 0.9228,
       "step": 18096
     },
     {
-      "epoch": 0.49702013127900907,
+      "epoch": 0.5135357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.0582303408763286e-05,
-      "loss": 0.9904,
+      "learning_rate": 1.0047791877624932e-05,
+      "loss": 0.8298,
       "step": 18097
     },
     {
-      "epoch": 0.49704759550685235,
+      "epoch": 0.5135641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 1.0581415399801887e-05,
-      "loss": 0.8847,
+      "learning_rate": 1.0046872809852148e-05,
+      "loss": 0.9351,
       "step": 18098
     },
     {
-      "epoch": 0.49707505973469557,
+      "epoch": 0.5135925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.0580527386240104e-05,
-      "loss": 0.973,
+      "learning_rate": 1.0045953741683426e-05,
+      "loss": 0.8969,
       "step": 18099
     },
     {
-      "epoch": 0.4971025239625388,
+      "epoch": 0.5136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.0579639368084962e-05,
-      "loss": 0.8501,
+      "learning_rate": 1.0045034673126528e-05,
+      "loss": 0.9085,
       "step": 18100
     },
     {
-      "epoch": 0.497129988190382,
+      "epoch": 0.5136492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.057875134534349e-05,
-      "loss": 0.8693,
+      "learning_rate": 1.004411560418922e-05,
+      "loss": 0.9059,
       "step": 18101
     },
     {
-      "epoch": 0.4971574524182253,
+      "epoch": 0.5136776390465381,
       "grad_norm": 0.0,
-      "learning_rate": 1.0577863318022714e-05,
-      "loss": 0.981,
+      "learning_rate": 1.0043196534879268e-05,
+      "loss": 0.8437,
       "step": 18102
     },
     {
-      "epoch": 0.4971849166460685,
+      "epoch": 0.5137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.0576975286129658e-05,
-      "loss": 0.8864,
+      "learning_rate": 1.004227746520443e-05,
+      "loss": 0.9007,
       "step": 18103
     },
     {
-      "epoch": 0.49721238087391173,
+      "epoch": 0.5137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.057608724967135e-05,
-      "loss": 0.8617,
+      "learning_rate": 1.0041358395172473e-05,
+      "loss": 0.7625,
       "step": 18104
     },
     {
-      "epoch": 0.49723984510175495,
+      "epoch": 0.5137627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.0575199208654815e-05,
-      "loss": 0.7945,
+      "learning_rate": 1.0040439324791163e-05,
+      "loss": 0.7614,
       "step": 18105
     },
     {
-      "epoch": 0.4972673093295982,
+      "epoch": 0.5137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.0574311163087085e-05,
-      "loss": 0.8491,
+      "learning_rate": 1.003952025406826e-05,
+      "loss": 0.9145,
       "step": 18106
     },
     {
-      "epoch": 0.49729477355744145,
+      "epoch": 0.5138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.057342311297518e-05,
-      "loss": 0.8957,
+      "learning_rate": 1.0038601183011524e-05,
+      "loss": 0.8351,
       "step": 18107
     },
     {
-      "epoch": 0.49732223778528467,
+      "epoch": 0.5138479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 1.057253505832613e-05,
-      "loss": 0.8359,
+      "learning_rate": 1.0037682111628723e-05,
+      "loss": 0.8482,
       "step": 18108
     },
     {
-      "epoch": 0.4973497020131279,
+      "epoch": 0.5138762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.057164699914696e-05,
-      "loss": 0.9421,
+      "learning_rate": 1.0036763039927621e-05,
+      "loss": 0.9223,
       "step": 18109
     },
     {
-      "epoch": 0.4973771662409711,
+      "epoch": 0.5139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.0570758935444699e-05,
-      "loss": 0.768,
+      "learning_rate": 1.003584396791598e-05,
+      "loss": 0.9117,
       "step": 18110
     },
     {
-      "epoch": 0.4974046304688144,
+      "epoch": 0.513933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.0569870867226373e-05,
-      "loss": 0.9136,
+      "learning_rate": 1.003492489560156e-05,
+      "loss": 0.8276,
       "step": 18111
     },
     {
-      "epoch": 0.4974320946966576,
+      "epoch": 0.513961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.0568982794499006e-05,
-      "loss": 0.9513,
+      "learning_rate": 1.0034005822992132e-05,
+      "loss": 0.9263,
       "step": 18112
     },
     {
-      "epoch": 0.49745955892450083,
+      "epoch": 0.5139897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 1.0568094717269626e-05,
-      "loss": 0.8698,
+      "learning_rate": 1.0033086750095455e-05,
+      "loss": 0.8519,
       "step": 18113
     },
     {
-      "epoch": 0.49748702315234405,
+      "epoch": 0.5140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.0567206635545264e-05,
-      "loss": 1.0478,
+      "learning_rate": 1.003216767691929e-05,
+      "loss": 0.9188,
       "step": 18114
     },
     {
-      "epoch": 0.49751448738018733,
+      "epoch": 0.5140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.0566318549332938e-05,
-      "loss": 0.8758,
+      "learning_rate": 1.0031248603471404e-05,
+      "loss": 0.987,
       "step": 18115
     },
     {
-      "epoch": 0.49754195160803055,
+      "epoch": 0.5140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.0565430458639683e-05,
-      "loss": 0.8738,
+      "learning_rate": 1.0030329529759561e-05,
+      "loss": 0.9089,
       "step": 18116
     },
     {
-      "epoch": 0.4975694158358738,
+      "epoch": 0.5141032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.0564542363472526e-05,
-      "loss": 0.9338,
+      "learning_rate": 1.0029410455791524e-05,
+      "loss": 0.9167,
       "step": 18117
     },
     {
-      "epoch": 0.497596880063717,
+      "epoch": 0.5141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.0563654263838487e-05,
-      "loss": 0.8245,
+      "learning_rate": 1.0028491381575054e-05,
+      "loss": 0.9245,
       "step": 18118
     },
     {
-      "epoch": 0.4976243442915602,
+      "epoch": 0.5141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.05627661597446e-05,
-      "loss": 0.8657,
+      "learning_rate": 1.0027572307117916e-05,
+      "loss": 0.7844,
       "step": 18119
     },
     {
-      "epoch": 0.4976518085194035,
+      "epoch": 0.5141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.0561878051197886e-05,
-      "loss": 0.949,
+      "learning_rate": 1.0026653232427873e-05,
+      "loss": 0.8593,
       "step": 18120
     },
     {
-      "epoch": 0.4976792727472467,
+      "epoch": 0.514216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.0560989938205377e-05,
-      "loss": 0.8793,
+      "learning_rate": 1.002573415751269e-05,
+      "loss": 0.8281,
       "step": 18121
     },
     {
-      "epoch": 0.49770673697508994,
+      "epoch": 0.5142451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 1.0560101820774103e-05,
-      "loss": 0.9277,
+      "learning_rate": 1.0024815082380131e-05,
+      "loss": 0.9172,
       "step": 18122
     },
     {
-      "epoch": 0.49773420120293316,
+      "epoch": 0.5142735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.0559213698911079e-05,
-      "loss": 0.9314,
+      "learning_rate": 1.0023896007037952e-05,
+      "loss": 0.9735,
       "step": 18123
     },
     {
-      "epoch": 0.49776166543077643,
+      "epoch": 0.5143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.0558325572623343e-05,
-      "loss": 0.8353,
+      "learning_rate": 1.0022976931493926e-05,
+      "loss": 0.8598,
       "step": 18124
     },
     {
-      "epoch": 0.49778912965861966,
+      "epoch": 0.5143303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.0557437441917917e-05,
-      "loss": 0.8088,
+      "learning_rate": 1.0022057855755814e-05,
+      "loss": 0.97,
       "step": 18125
     },
     {
-      "epoch": 0.4978165938864629,
+      "epoch": 0.5143586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.0556549306801834e-05,
-      "loss": 0.9309,
+      "learning_rate": 1.0021138779831378e-05,
+      "loss": 0.8024,
       "step": 18126
     },
     {
-      "epoch": 0.4978440581143061,
+      "epoch": 0.5143870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.0555661167282112e-05,
-      "loss": 0.9039,
+      "learning_rate": 1.0020219703728379e-05,
+      "loss": 0.7521,
       "step": 18127
     },
     {
-      "epoch": 0.4978715223421494,
+      "epoch": 0.5144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.0554773023365787e-05,
-      "loss": 0.9539,
+      "learning_rate": 1.0019300627454586e-05,
+      "loss": 0.9108,
       "step": 18128
     },
     {
-      "epoch": 0.4978989865699926,
+      "epoch": 0.5144438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 1.055388487505988e-05,
-      "loss": 0.8627,
+      "learning_rate": 1.0018381551017762e-05,
+      "loss": 0.7588,
       "step": 18129
     },
     {
-      "epoch": 0.4979264507978358,
+      "epoch": 0.514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.0552996722371425e-05,
-      "loss": 0.8965,
+      "learning_rate": 1.0017462474425664e-05,
+      "loss": 0.8722,
       "step": 18130
     },
     {
-      "epoch": 0.49795391502567904,
+      "epoch": 0.5145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.0552108565307446e-05,
-      "loss": 0.9076,
+      "learning_rate": 1.0016543397686059e-05,
+      "loss": 0.9748,
       "step": 18131
     },
     {
-      "epoch": 0.49798137925352226,
+      "epoch": 0.5145289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 1.0551220403874967e-05,
-      "loss": 0.879,
+      "learning_rate": 1.0015624320806714e-05,
+      "loss": 0.8023,
       "step": 18132
     },
     {
-      "epoch": 0.49800884348136554,
+      "epoch": 0.5145573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.0550332238081018e-05,
-      "loss": 0.9752,
+      "learning_rate": 1.0014705243795389e-05,
+      "loss": 0.8793,
       "step": 18133
     },
     {
-      "epoch": 0.49803630770920876,
+      "epoch": 0.5145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.0549444067932633e-05,
-      "loss": 1.0081,
+      "learning_rate": 1.0013786166659847e-05,
+      "loss": 0.8103,
       "step": 18134
     },
     {
-      "epoch": 0.498063771937052,
+      "epoch": 0.5146140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.0548555893436827e-05,
-      "loss": 0.9705,
+      "learning_rate": 1.0012867089407855e-05,
+      "loss": 0.8798,
       "step": 18135
     },
     {
-      "epoch": 0.4980912361648952,
+      "epoch": 0.5146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.0547667714600637e-05,
-      "loss": 0.8559,
+      "learning_rate": 1.0011948012047173e-05,
+      "loss": 0.8813,
       "step": 18136
     },
     {
-      "epoch": 0.4981187003927385,
+      "epoch": 0.5146708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 1.0546779531431088e-05,
-      "loss": 0.9615,
+      "learning_rate": 1.0011028934585565e-05,
+      "loss": 0.9045,
       "step": 18137
     },
     {
-      "epoch": 0.4981461646205817,
+      "epoch": 0.5146992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.0545891343935208e-05,
-      "loss": 0.9912,
+      "learning_rate": 1.0010109857030794e-05,
+      "loss": 0.7679,
       "step": 18138
     },
     {
-      "epoch": 0.4981736288484249,
+      "epoch": 0.514727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.0545003152120024e-05,
-      "loss": 0.9619,
+      "learning_rate": 1.0009190779390627e-05,
+      "loss": 0.9153,
       "step": 18139
     },
     {
-      "epoch": 0.49820109307626814,
+      "epoch": 0.514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.0544114955992563e-05,
-      "loss": 0.9119,
+      "learning_rate": 1.0008271701672825e-05,
+      "loss": 0.8054,
       "step": 18140
     },
     {
-      "epoch": 0.4982285573041114,
+      "epoch": 0.5147843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.0543226755559854e-05,
-      "loss": 0.9215,
+      "learning_rate": 1.0007352623885148e-05,
+      "loss": 0.8794,
       "step": 18141
     },
     {
-      "epoch": 0.49825602153195464,
+      "epoch": 0.5148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.0542338550828925e-05,
-      "loss": 0.9641,
+      "learning_rate": 1.0006433546035368e-05,
+      "loss": 0.9583,
       "step": 18142
     },
     {
-      "epoch": 0.49828348575979786,
+      "epoch": 0.5148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.0541450341806806e-05,
-      "loss": 0.8943,
+      "learning_rate": 1.0005514468131242e-05,
+      "loss": 0.9249,
       "step": 18143
     },
     {
-      "epoch": 0.4983109499876411,
+      "epoch": 0.5148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.0540562128500517e-05,
-      "loss": 0.8873,
+      "learning_rate": 1.0004595390180532e-05,
+      "loss": 0.9245,
       "step": 18144
     },
     {
-      "epoch": 0.4983384142154843,
+      "epoch": 0.5148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.0539673910917092e-05,
-      "loss": 0.8137,
+      "learning_rate": 1.0003676312191008e-05,
+      "loss": 0.8431,
       "step": 18145
     },
     {
-      "epoch": 0.4983658784433276,
+      "epoch": 0.5149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.0538785689063561e-05,
-      "loss": 0.8349,
+      "learning_rate": 1.0002757234170428e-05,
+      "loss": 0.8734,
       "step": 18146
     },
     {
-      "epoch": 0.4983933426711708,
+      "epoch": 0.5149545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.053789746294695e-05,
-      "loss": 0.8773,
+      "learning_rate": 1.0001838156126558e-05,
+      "loss": 0.87,
       "step": 18147
     },
     {
-      "epoch": 0.498420806899014,
+      "epoch": 0.514982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.053700923257428e-05,
-      "loss": 0.8323,
+      "learning_rate": 1.0000919078067162e-05,
+      "loss": 0.8582,
       "step": 18148
     },
     {
-      "epoch": 0.49844827112685725,
+      "epoch": 0.5150113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 1.0536120997952587e-05,
-      "loss": 0.9183,
+      "learning_rate": 1e-05,
+      "loss": 0.902,
       "step": 18149
     },
     {
-      "epoch": 0.4984757353547005,
+      "epoch": 0.5150397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 1.05352327590889e-05,
-      "loss": 0.8879,
+      "learning_rate": 9.999080921932843e-06,
+      "loss": 0.9224,
       "step": 18150
     },
     {
-      "epoch": 0.49850319958254374,
+      "epoch": 0.5150681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.0534344515990242e-05,
-      "loss": 0.9366,
+      "learning_rate": 9.998161843873443e-06,
+      "loss": 0.9545,
       "step": 18151
     },
     {
-      "epoch": 0.49853066381038696,
+      "epoch": 0.5150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.0533456268663645e-05,
-      "loss": 0.8809,
+      "learning_rate": 9.997242765829575e-06,
+      "loss": 0.8028,
       "step": 18152
     },
     {
-      "epoch": 0.4985581280382302,
+      "epoch": 0.5151248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.0532568017116131e-05,
-      "loss": 0.8748,
+      "learning_rate": 9.996323687808997e-06,
+      "loss": 0.8958,
       "step": 18153
     },
     {
-      "epoch": 0.49858559226607346,
+      "epoch": 0.5151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.0531679761354737e-05,
-      "loss": 0.9656,
+      "learning_rate": 9.99540460981947e-06,
+      "loss": 0.9162,
       "step": 18154
     },
     {
-      "epoch": 0.4986130564939167,
+      "epoch": 0.5151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.0530791501386485e-05,
-      "loss": 0.925,
+      "learning_rate": 9.994485531868763e-06,
+      "loss": 0.8465,
       "step": 18155
     },
     {
-      "epoch": 0.4986405207217599,
+      "epoch": 0.5152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.0529903237218404e-05,
-      "loss": 0.815,
+      "learning_rate": 9.993566453964639e-06,
+      "loss": 0.8863,
       "step": 18156
     },
     {
-      "epoch": 0.4986679849496031,
+      "epoch": 0.5152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.0529014968857526e-05,
-      "loss": 0.9086,
+      "learning_rate": 9.992647376114853e-06,
+      "loss": 1.052,
       "step": 18157
     },
     {
-      "epoch": 0.4986954491774464,
+      "epoch": 0.515266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.0528126696310876e-05,
-      "loss": 0.8495,
+      "learning_rate": 9.991728298327179e-06,
+      "loss": 0.8682,
       "step": 18158
     },
     {
-      "epoch": 0.4987229134052896,
+      "epoch": 0.5152951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 1.0527238419585482e-05,
-      "loss": 0.9011,
+      "learning_rate": 9.990809220609374e-06,
+      "loss": 0.8915,
       "step": 18159
     },
     {
-      "epoch": 0.49875037763313285,
+      "epoch": 0.5153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.052635013868837e-05,
-      "loss": 0.9265,
+      "learning_rate": 9.989890142969207e-06,
+      "loss": 0.741,
       "step": 18160
     },
     {
-      "epoch": 0.49877784186097607,
+      "epoch": 0.5153518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 1.0525461853626578e-05,
-      "loss": 0.8702,
+      "learning_rate": 9.98897106541444e-06,
+      "loss": 0.9214,
       "step": 18161
     },
     {
-      "epoch": 0.4988053060888193,
+      "epoch": 0.5153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.0524573564407125e-05,
-      "loss": 1.0104,
+      "learning_rate": 9.988051987952829e-06,
+      "loss": 0.9351,
       "step": 18162
     },
     {
-      "epoch": 0.49883277031666257,
+      "epoch": 0.5154086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.0523685271037043e-05,
-      "loss": 0.822,
+      "learning_rate": 9.987132910592147e-06,
+      "loss": 0.8313,
       "step": 18163
     },
     {
-      "epoch": 0.4988602345445058,
+      "epoch": 0.5154370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 1.0522796973523362e-05,
-      "loss": 0.8234,
+      "learning_rate": 9.986213833340156e-06,
+      "loss": 0.8477,
       "step": 18164
     },
     {
-      "epoch": 0.498887698772349,
+      "epoch": 0.5154653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.0521908671873107e-05,
-      "loss": 0.9403,
+      "learning_rate": 9.985294756204613e-06,
+      "loss": 0.8483,
       "step": 18165
     },
     {
-      "epoch": 0.49891516300019223,
+      "epoch": 0.5154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.052102036609331e-05,
-      "loss": 0.8918,
+      "learning_rate": 9.98437567919329e-06,
+      "loss": 0.891,
       "step": 18166
     },
     {
-      "epoch": 0.4989426272280355,
+      "epoch": 0.515522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.0520132056190997e-05,
-      "loss": 0.821,
+      "learning_rate": 9.983456602313944e-06,
+      "loss": 0.8639,
       "step": 18167
     },
     {
-      "epoch": 0.49897009145587873,
+      "epoch": 0.5155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.0519243742173196e-05,
-      "loss": 0.9143,
+      "learning_rate": 9.98253752557434e-06,
+      "loss": 0.9253,
       "step": 18168
     },
     {
-      "epoch": 0.49899755568372195,
+      "epoch": 0.5155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.051835542404694e-05,
-      "loss": 0.879,
+      "learning_rate": 9.981618448982245e-06,
+      "loss": 0.9974,
       "step": 18169
     },
     {
-      "epoch": 0.49902501991156517,
+      "epoch": 0.5156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.0517467101819254e-05,
-      "loss": 0.8228,
+      "learning_rate": 9.980699372545419e-06,
+      "loss": 0.8622,
       "step": 18170
     },
     {
-      "epoch": 0.49905248413940845,
+      "epoch": 0.5156356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.0516578775497166e-05,
-      "loss": 0.9342,
+      "learning_rate": 9.979780296271623e-06,
+      "loss": 0.7518,
       "step": 18171
     },
     {
-      "epoch": 0.49907994836725167,
+      "epoch": 0.5156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.0515690445087707e-05,
-      "loss": 0.9299,
+      "learning_rate": 9.978861220168627e-06,
+      "loss": 0.8156,
       "step": 18172
     },
     {
-      "epoch": 0.4991074125950949,
+      "epoch": 0.5156923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 1.0514802110597909e-05,
-      "loss": 0.8964,
+      "learning_rate": 9.977942144244186e-06,
+      "loss": 0.8354,
       "step": 18173
     },
     {
-      "epoch": 0.4991348768229381,
+      "epoch": 0.5157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.0513913772034797e-05,
-      "loss": 0.9395,
+      "learning_rate": 9.977023068506074e-06,
+      "loss": 0.7729,
       "step": 18174
     },
     {
-      "epoch": 0.49916234105078133,
+      "epoch": 0.5157491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.0513025429405397e-05,
-      "loss": 0.9102,
+      "learning_rate": 9.976103992962047e-06,
+      "loss": 1.0041,
       "step": 18175
     },
     {
-      "epoch": 0.4991898052786246,
+      "epoch": 0.51577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.0512137082716742e-05,
-      "loss": 0.9177,
+      "learning_rate": 9.975184917619874e-06,
+      "loss": 0.742,
       "step": 18176
     },
     {
-      "epoch": 0.49921726950646783,
+      "epoch": 0.515805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.051124873197586e-05,
-      "loss": 0.9426,
+      "learning_rate": 9.974265842487312e-06,
+      "loss": 0.8689,
       "step": 18177
     },
     {
-      "epoch": 0.49924473373431105,
+      "epoch": 0.5158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.0510360377189781e-05,
-      "loss": 0.8371,
+      "learning_rate": 9.973346767572128e-06,
+      "loss": 0.9681,
       "step": 18178
     },
     {
-      "epoch": 0.4992721979621543,
+      "epoch": 0.5158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.0509472018365534e-05,
-      "loss": 0.8961,
+      "learning_rate": 9.972427692882087e-06,
+      "loss": 0.8934,
       "step": 18179
     },
     {
-      "epoch": 0.49929966218999755,
+      "epoch": 0.5158910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 1.0508583655510144e-05,
-      "loss": 0.9018,
+      "learning_rate": 9.97150861842495e-06,
+      "loss": 0.9104,
       "step": 18180
     },
     {
-      "epoch": 0.49932712641784077,
+      "epoch": 0.5159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.0507695288630647e-05,
-      "loss": 0.9594,
+      "learning_rate": 9.97058954420848e-06,
+      "loss": 0.8076,
       "step": 18181
     },
     {
-      "epoch": 0.499354590645684,
+      "epoch": 0.5159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.0506806917734065e-05,
-      "loss": 0.976,
+      "learning_rate": 9.96967047024044e-06,
+      "loss": 0.9939,
       "step": 18182
     },
     {
-      "epoch": 0.4993820548735272,
+      "epoch": 0.5159761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 1.0505918542827432e-05,
-      "loss": 0.9166,
+      "learning_rate": 9.968751396528599e-06,
+      "loss": 0.8744,
       "step": 18183
     },
     {
-      "epoch": 0.4994095191013705,
+      "epoch": 0.5160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.0505030163917773e-05,
-      "loss": 0.8338,
+      "learning_rate": 9.967832323080715e-06,
+      "loss": 0.9579,
       "step": 18184
     },
     {
-      "epoch": 0.4994369833292137,
+      "epoch": 0.5160329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.050414178101212e-05,
-      "loss": 0.7551,
+      "learning_rate": 9.96691324990455e-06,
+      "loss": 0.9392,
       "step": 18185
     },
     {
-      "epoch": 0.49946444755705693,
+      "epoch": 0.5160612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 1.0503253394117505e-05,
-      "loss": 0.8721,
+      "learning_rate": 9.965994177007871e-06,
+      "loss": 0.791,
       "step": 18186
     },
     {
-      "epoch": 0.49949191178490016,
+      "epoch": 0.5160896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.0502365003240955e-05,
-      "loss": 0.9073,
+      "learning_rate": 9.965075104398444e-06,
+      "loss": 0.8812,
       "step": 18187
     },
     {
-      "epoch": 0.4995193760127434,
+      "epoch": 0.5161180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 1.0501476608389494e-05,
-      "loss": 1.0058,
+      "learning_rate": 9.964156032084022e-06,
+      "loss": 0.9492,
       "step": 18188
     },
     {
-      "epoch": 0.49954684024058665,
+      "epoch": 0.5161464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.0500588209570157e-05,
-      "loss": 1.0013,
+      "learning_rate": 9.96323696007238e-06,
+      "loss": 0.8602,
       "step": 18189
     },
     {
-      "epoch": 0.4995743044684299,
+      "epoch": 0.5161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 1.0499699806789976e-05,
-      "loss": 0.9907,
+      "learning_rate": 9.962317888371278e-06,
+      "loss": 0.9233,
       "step": 18190
     },
     {
-      "epoch": 0.4996017686962731,
+      "epoch": 0.5162031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 1.0498811400055972e-05,
-      "loss": 0.8775,
+      "learning_rate": 9.961398816988479e-06,
+      "loss": 0.8371,
       "step": 18191
     },
     {
-      "epoch": 0.4996292329241163,
+      "epoch": 0.5162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.049792298937518e-05,
-      "loss": 0.9616,
+      "learning_rate": 9.960479745931743e-06,
+      "loss": 0.7873,
       "step": 18192
     },
     {
-      "epoch": 0.4996566971519596,
+      "epoch": 0.5162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.049703457475463e-05,
-      "loss": 0.9759,
+      "learning_rate": 9.959560675208839e-06,
+      "loss": 0.9109,
       "step": 18193
     },
     {
-      "epoch": 0.4996841613798028,
+      "epoch": 0.5162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.049614615620135e-05,
-      "loss": 0.8691,
+      "learning_rate": 9.958641604827529e-06,
+      "loss": 0.9557,
       "step": 18194
     },
     {
-      "epoch": 0.49971162560764604,
+      "epoch": 0.516316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.0495257733722367e-05,
-      "loss": 0.8334,
+      "learning_rate": 9.95772253479557e-06,
+      "loss": 0.8969,
       "step": 18195
     },
     {
-      "epoch": 0.49973908983548926,
+      "epoch": 0.5163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.0494369307324715e-05,
-      "loss": 0.9464,
+      "learning_rate": 9.956803465120736e-06,
+      "loss": 0.9171,
       "step": 18196
     },
     {
-      "epoch": 0.49976655406333254,
+      "epoch": 0.5163734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 1.049348087701542e-05,
-      "loss": 0.8106,
+      "learning_rate": 9.955884395810783e-06,
+      "loss": 0.8993,
       "step": 18197
     },
     {
-      "epoch": 0.49979401829117576,
+      "epoch": 0.5164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.0492592442801515e-05,
-      "loss": 0.9186,
+      "learning_rate": 9.954965326873477e-06,
+      "loss": 0.8913,
       "step": 18198
     },
     {
-      "epoch": 0.499821482519019,
+      "epoch": 0.5164301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 1.0491704004690027e-05,
-      "loss": 0.9522,
+      "learning_rate": 9.95404625831658e-06,
+      "loss": 0.8793,
       "step": 18199
     },
     {
-      "epoch": 0.4998489467468622,
+      "epoch": 0.5164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 1.0490815562687985e-05,
-      "loss": 0.8457,
+      "learning_rate": 9.953127190147858e-06,
+      "loss": 0.9305,
       "step": 18200
     },
     {
-      "epoch": 0.4998764109747054,
+      "epoch": 0.5164869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.0489927116802421e-05,
-      "loss": 0.8734,
+      "learning_rate": 9.95220812237507e-06,
+      "loss": 0.8484,
       "step": 18201
     },
     {
-      "epoch": 0.4999038752025487,
+      "epoch": 0.5165153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.0489038667040368e-05,
-      "loss": 0.9399,
+      "learning_rate": 9.951289055005983e-06,
+      "loss": 0.845,
       "step": 18202
     },
     {
-      "epoch": 0.4999313394303919,
+      "epoch": 0.5165437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.0488150213408845e-05,
-      "loss": 0.9372,
+      "learning_rate": 9.950369988048357e-06,
+      "loss": 0.8291,
       "step": 18203
     },
     {
-      "epoch": 0.49995880365823514,
+      "epoch": 0.5165720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 1.0487261755914891e-05,
-      "loss": 0.8947,
+      "learning_rate": 9.949450921509962e-06,
+      "loss": 0.8879,
       "step": 18204
     },
     {
-      "epoch": 0.49998626788607836,
+      "epoch": 0.5166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.0486373294565534e-05,
-      "loss": 0.9717,
+      "learning_rate": 9.948531855398558e-06,
+      "loss": 0.8657,
       "step": 18205
     },
     {
-      "epoch": 0.5000137321139216,
+      "epoch": 0.5166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.0485484829367805e-05,
-      "loss": 0.9983,
+      "learning_rate": 9.947612789721904e-06,
+      "loss": 0.9212,
       "step": 18206
     },
     {
-      "epoch": 0.5000411963417648,
+      "epoch": 0.5166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.0484596360328727e-05,
-      "loss": 0.8177,
+      "learning_rate": 9.946693724487771e-06,
+      "loss": 0.86,
       "step": 18207
     },
     {
-      "epoch": 0.5000686605696081,
+      "epoch": 0.5166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.0483707887455338e-05,
-      "loss": 1.0118,
+      "learning_rate": 9.945774659703918e-06,
+      "loss": 0.8587,
       "step": 18208
     },
     {
-      "epoch": 0.5000961247974514,
+      "epoch": 0.5167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.0482819410754663e-05,
-      "loss": 0.8365,
+      "learning_rate": 9.944855595378106e-06,
+      "loss": 0.8071,
       "step": 18209
     },
     {
-      "epoch": 0.5001235890252945,
+      "epoch": 0.5167423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 1.0481930930233737e-05,
-      "loss": 0.9063,
+      "learning_rate": 9.943936531518104e-06,
+      "loss": 0.8035,
       "step": 18210
     },
     {
-      "epoch": 0.5001510532531378,
+      "epoch": 0.5167707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.0481042445899585e-05,
-      "loss": 0.9886,
+      "learning_rate": 9.943017468131672e-06,
+      "loss": 0.7725,
       "step": 18211
     },
     {
-      "epoch": 0.5001785174809811,
+      "epoch": 0.5167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.0480153957759239e-05,
-      "loss": 0.9111,
+      "learning_rate": 9.942098405226571e-06,
+      "loss": 0.9177,
       "step": 18212
     },
     {
-      "epoch": 0.5002059817088242,
+      "epoch": 0.516827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.0479265465819726e-05,
-      "loss": 0.9282,
+      "learning_rate": 9.941179342810571e-06,
+      "loss": 0.8585,
       "step": 18213
     },
     {
-      "epoch": 0.5002334459366675,
+      "epoch": 0.5168558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 1.0478376970088085e-05,
-      "loss": 0.8846,
+      "learning_rate": 9.940260280891432e-06,
+      "loss": 0.8952,
       "step": 18214
     },
     {
-      "epoch": 0.5002609101645107,
+      "epoch": 0.5168842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 1.0477488470571336e-05,
-      "loss": 0.9147,
+      "learning_rate": 9.939341219476915e-06,
+      "loss": 0.8524,
       "step": 18215
     },
     {
-      "epoch": 0.500288374392354,
+      "epoch": 0.5169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.0476599967276512e-05,
-      "loss": 0.8714,
+      "learning_rate": 9.938422158574786e-06,
+      "loss": 0.8195,
       "step": 18216
     },
     {
-      "epoch": 0.5003158386201972,
+      "epoch": 0.5169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.0475711460210646e-05,
-      "loss": 0.8457,
+      "learning_rate": 9.937503098192809e-06,
+      "loss": 0.8564,
       "step": 18217
     },
     {
-      "epoch": 0.5003433028480404,
+      "epoch": 0.5169693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 1.047482294938077e-05,
-      "loss": 0.8809,
+      "learning_rate": 9.936584038338744e-06,
+      "loss": 0.9553,
       "step": 18218
     },
     {
-      "epoch": 0.5003707670758837,
+      "epoch": 0.5169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.0473934434793907e-05,
-      "loss": 0.8191,
+      "learning_rate": 9.935664979020354e-06,
+      "loss": 0.8768,
       "step": 18219
     },
     {
-      "epoch": 0.5003982313037268,
+      "epoch": 0.5170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.0473045916457092e-05,
-      "loss": 0.8813,
+      "learning_rate": 9.93474592024541e-06,
+      "loss": 0.8015,
       "step": 18220
     },
     {
-      "epoch": 0.5004256955315701,
+      "epoch": 0.5170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.0472157394377353e-05,
-      "loss": 0.9389,
+      "learning_rate": 9.933826862021669e-06,
+      "loss": 0.9181,
       "step": 18221
     },
     {
-      "epoch": 0.5004531597594134,
+      "epoch": 0.5170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.0471268868561726e-05,
-      "loss": 0.9283,
+      "learning_rate": 9.932907804356893e-06,
+      "loss": 0.9062,
       "step": 18222
     },
     {
-      "epoch": 0.5004806239872566,
+      "epoch": 0.51711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.0470380339017235e-05,
-      "loss": 0.8618,
+      "learning_rate": 9.93198874725885e-06,
+      "loss": 0.8105,
       "step": 18223
     },
     {
-      "epoch": 0.5005080882150998,
+      "epoch": 0.5171396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 1.0469491805750912e-05,
-      "loss": 0.7952,
+      "learning_rate": 9.931069690735301e-06,
+      "loss": 0.8112,
       "step": 18224
     },
     {
-      "epoch": 0.5005355524429431,
+      "epoch": 0.5171679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.0468603268769787e-05,
-      "loss": 0.8822,
+      "learning_rate": 9.930150634794013e-06,
+      "loss": 0.9508,
       "step": 18225
     },
     {
-      "epoch": 0.5005630166707863,
+      "epoch": 0.5171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 1.0467714728080895e-05,
-      "loss": 0.9769,
+      "learning_rate": 9.92923157944274e-06,
+      "loss": 0.7864,
       "step": 18226
     },
     {
-      "epoch": 0.5005904808986296,
+      "epoch": 0.5172247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 1.0466826183691259e-05,
-      "loss": 0.8405,
+      "learning_rate": 9.928312524689252e-06,
+      "loss": 0.9004,
       "step": 18227
     },
     {
-      "epoch": 0.5006179451264727,
+      "epoch": 0.5172531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.0465937635607912e-05,
-      "loss": 0.9046,
+      "learning_rate": 9.927393470541314e-06,
+      "loss": 0.826,
       "step": 18228
     },
     {
-      "epoch": 0.500645409354316,
+      "epoch": 0.5172814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.0465049083837891e-05,
-      "loss": 0.8509,
+      "learning_rate": 9.926474417006684e-06,
+      "loss": 0.8406,
       "step": 18229
     },
     {
-      "epoch": 0.5006728735821593,
+      "epoch": 0.5173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.046416052838822e-05,
-      "loss": 0.8746,
+      "learning_rate": 9.92555536409313e-06,
+      "loss": 0.8946,
       "step": 18230
     },
     {
-      "epoch": 0.5007003378100024,
+      "epoch": 0.5173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.0463271969265928e-05,
-      "loss": 0.842,
+      "learning_rate": 9.924636311808413e-06,
+      "loss": 0.8133,
       "step": 18231
     },
     {
-      "epoch": 0.5007278020378457,
+      "epoch": 0.517366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.0462383406478054e-05,
-      "loss": 0.7703,
+      "learning_rate": 9.923717260160294e-06,
+      "loss": 0.8693,
       "step": 18232
     },
     {
-      "epoch": 0.5007552662656889,
+      "epoch": 0.5173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.0461494840031619e-05,
-      "loss": 0.8939,
+      "learning_rate": 9.922798209156542e-06,
+      "loss": 0.9223,
       "step": 18233
     },
     {
-      "epoch": 0.5007827304935322,
+      "epoch": 0.5174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.046060626993366e-05,
-      "loss": 0.9,
+      "learning_rate": 9.921879158804912e-06,
+      "loss": 0.9675,
       "step": 18234
     },
     {
-      "epoch": 0.5008101947213754,
+      "epoch": 0.5174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.0459717696191206e-05,
-      "loss": 0.8631,
+      "learning_rate": 9.920960109113176e-06,
+      "loss": 0.8681,
       "step": 18235
     },
     {
-      "epoch": 0.5008376589492186,
+      "epoch": 0.5174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.0458829118811287e-05,
-      "loss": 0.8873,
+      "learning_rate": 9.920041060089092e-06,
+      "loss": 0.885,
       "step": 18236
     },
     {
-      "epoch": 0.5008651231770619,
+      "epoch": 0.5175085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.0457940537800932e-05,
-      "loss": 0.7814,
+      "learning_rate": 9.919122011740427e-06,
+      "loss": 0.8297,
       "step": 18237
     },
     {
-      "epoch": 0.5008925874049052,
+      "epoch": 0.5175368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.045705195316718e-05,
-      "loss": 0.8897,
+      "learning_rate": 9.918202964074942e-06,
+      "loss": 0.9895,
       "step": 18238
     },
     {
-      "epoch": 0.5009200516327483,
+      "epoch": 0.5175652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 1.0456163364917049e-05,
-      "loss": 0.9673,
+      "learning_rate": 9.917283917100396e-06,
+      "loss": 0.8769,
       "step": 18239
     },
     {
-      "epoch": 0.5009475158605916,
+      "epoch": 0.5175936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 1.0455274773057583e-05,
-      "loss": 0.8654,
+      "learning_rate": 9.916364870824561e-06,
+      "loss": 0.9311,
       "step": 18240
     },
     {
-      "epoch": 0.5009749800884348,
+      "epoch": 0.517622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.0454386177595802e-05,
-      "loss": 1.0077,
+      "learning_rate": 9.915445825255196e-06,
+      "loss": 0.8857,
       "step": 18241
     },
     {
-      "epoch": 0.501002444316278,
+      "epoch": 0.517650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.0453497578538742e-05,
-      "loss": 0.9336,
+      "learning_rate": 9.91452678040006e-06,
+      "loss": 0.9232,
       "step": 18242
     },
     {
-      "epoch": 0.5010299085441213,
+      "epoch": 0.5176787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.0452608975893437e-05,
-      "loss": 0.9997,
+      "learning_rate": 9.913607736266923e-06,
+      "loss": 0.9193,
       "step": 18243
     },
     {
-      "epoch": 0.5010573727719645,
+      "epoch": 0.5177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.045172036966691e-05,
-      "loss": 0.9182,
+      "learning_rate": 9.912688692863545e-06,
+      "loss": 0.8089,
       "step": 18244
     },
     {
-      "epoch": 0.5010848369998078,
+      "epoch": 0.5177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.0450831759866198e-05,
-      "loss": 0.9383,
+      "learning_rate": 9.911769650197689e-06,
+      "loss": 0.9255,
       "step": 18245
     },
     {
-      "epoch": 0.5011123012276509,
+      "epoch": 0.5177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.0449943146498333e-05,
-      "loss": 0.8153,
+      "learning_rate": 9.91085060827712e-06,
+      "loss": 0.8635,
       "step": 18246
     },
     {
-      "epoch": 0.5011397654554942,
+      "epoch": 0.5177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.0449054529570339e-05,
-      "loss": 0.7769,
+      "learning_rate": 9.9099315671096e-06,
+      "loss": 0.9005,
       "step": 18247
     },
     {
-      "epoch": 0.5011672296833375,
+      "epoch": 0.5178206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 1.0448165909089254e-05,
-      "loss": 0.813,
+      "learning_rate": 9.909012526702889e-06,
+      "loss": 0.8911,
       "step": 18248
     },
     {
-      "epoch": 0.5011946939111807,
+      "epoch": 0.5178490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.0447277285062107e-05,
-      "loss": 0.9088,
+      "learning_rate": 9.908093487064755e-06,
+      "loss": 0.8831,
       "step": 18249
     },
     {
-      "epoch": 0.5012221581390239,
+      "epoch": 0.5178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 1.044638865749593e-05,
-      "loss": 0.9613,
+      "learning_rate": 9.907174448202957e-06,
+      "loss": 0.9708,
       "step": 18250
     },
     {
-      "epoch": 0.5012496223668672,
+      "epoch": 0.5179057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 1.044550002639775e-05,
-      "loss": 0.8377,
+      "learning_rate": 9.906255410125264e-06,
+      "loss": 0.8047,
       "step": 18251
     },
     {
-      "epoch": 0.5012770865947104,
+      "epoch": 0.5179341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 1.0444611391774603e-05,
-      "loss": 0.9286,
+      "learning_rate": 9.905336372839436e-06,
+      "loss": 0.9426,
       "step": 18252
     },
     {
-      "epoch": 0.5013045508225537,
+      "epoch": 0.5179625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.0443722753633519e-05,
-      "loss": 0.8673,
+      "learning_rate": 9.904417336353234e-06,
+      "loss": 0.8973,
       "step": 18253
     },
     {
-      "epoch": 0.5013320150503968,
+      "epoch": 0.5179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.0442834111981527e-05,
-      "loss": 0.9506,
+      "learning_rate": 9.903498300674425e-06,
+      "loss": 0.9449,
       "step": 18254
     },
     {
-      "epoch": 0.5013594792782401,
+      "epoch": 0.5180192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.0441945466825662e-05,
-      "loss": 0.8581,
+      "learning_rate": 9.90257926581077e-06,
+      "loss": 0.9716,
       "step": 18255
     },
     {
-      "epoch": 0.5013869435060834,
+      "epoch": 0.5180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.0441056818172953e-05,
-      "loss": 0.818,
+      "learning_rate": 9.90166023177003e-06,
+      "loss": 0.8476,
       "step": 18256
     },
     {
-      "epoch": 0.5014144077339265,
+      "epoch": 0.5180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.044016816603043e-05,
-      "loss": 0.8419,
+      "learning_rate": 9.900741198559971e-06,
+      "loss": 0.8843,
       "step": 18257
     },
     {
-      "epoch": 0.5014418719617698,
+      "epoch": 0.5181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.043927951040513e-05,
-      "loss": 0.9472,
+      "learning_rate": 9.899822166188359e-06,
+      "loss": 0.9144,
       "step": 18258
     },
     {
-      "epoch": 0.501469336189613,
+      "epoch": 0.5181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.0438390851304074e-05,
-      "loss": 0.9812,
+      "learning_rate": 9.898903134662949e-06,
+      "loss": 0.9321,
       "step": 18259
     },
     {
-      "epoch": 0.5014968004174563,
+      "epoch": 0.518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.0437502188734303e-05,
-      "loss": 0.9494,
+      "learning_rate": 9.897984103991513e-06,
+      "loss": 0.8259,
       "step": 18260
     },
     {
-      "epoch": 0.5015242646452995,
+      "epoch": 0.5181895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 1.0436613522702846e-05,
-      "loss": 0.8728,
+      "learning_rate": 9.897065074181809e-06,
+      "loss": 1.004,
       "step": 18261
     },
     {
-      "epoch": 0.5015517288731427,
+      "epoch": 0.5182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.0435724853216733e-05,
-      "loss": 0.8169,
+      "learning_rate": 9.896146045241598e-06,
+      "loss": 0.7812,
       "step": 18262
     },
     {
-      "epoch": 0.501579193100986,
+      "epoch": 0.5182463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.0434836180282995e-05,
-      "loss": 0.9544,
+      "learning_rate": 9.895227017178648e-06,
+      "loss": 0.7353,
       "step": 18263
     },
     {
-      "epoch": 0.5016066573288293,
+      "epoch": 0.5182746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.0433947503908668e-05,
-      "loss": 0.8705,
+      "learning_rate": 9.894307990000722e-06,
+      "loss": 0.9502,
       "step": 18264
     },
     {
-      "epoch": 0.5016341215566724,
+      "epoch": 0.5183030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 1.0433058824100774e-05,
-      "loss": 0.9135,
+      "learning_rate": 9.893388963715574e-06,
+      "loss": 0.8723,
       "step": 18265
     },
     {
-      "epoch": 0.5016615857845157,
+      "epoch": 0.5183314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 1.0432170140866358e-05,
-      "loss": 0.809,
+      "learning_rate": 9.89246993833098e-06,
+      "loss": 0.8621,
       "step": 18266
     },
     {
-      "epoch": 0.5016890500123589,
+      "epoch": 0.5183598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.043128145421244e-05,
-      "loss": 1.0042,
+      "learning_rate": 9.891550913854696e-06,
+      "loss": 0.8293,
       "step": 18267
     },
     {
-      "epoch": 0.5017165142402021,
+      "epoch": 0.5183881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 1.0430392764146058e-05,
-      "loss": 0.9204,
+      "learning_rate": 9.890631890294487e-06,
+      "loss": 0.8869,
       "step": 18268
     },
     {
-      "epoch": 0.5017439784680454,
+      "epoch": 0.5184165720771851,
       "grad_norm": 0.0,
-      "learning_rate": 1.0429504070674239e-05,
-      "loss": 0.9647,
+      "learning_rate": 9.889712867658117e-06,
+      "loss": 0.8671,
       "step": 18269
     },
     {
-      "epoch": 0.5017714426958886,
+      "epoch": 0.5184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.0428615373804023e-05,
-      "loss": 0.9301,
+      "learning_rate": 9.888793845953345e-06,
+      "loss": 0.8466,
       "step": 18270
     },
     {
-      "epoch": 0.5017989069237319,
+      "epoch": 0.5184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.042772667354243e-05,
-      "loss": 0.8582,
+      "learning_rate": 9.887874825187937e-06,
+      "loss": 0.9088,
       "step": 18271
     },
     {
-      "epoch": 0.501826371151575,
+      "epoch": 0.5185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.0426837969896502e-05,
-      "loss": 0.828,
+      "learning_rate": 9.886955805369655e-06,
+      "loss": 0.9004,
       "step": 18272
     },
     {
-      "epoch": 0.5018538353794183,
+      "epoch": 0.5185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.0425949262873265e-05,
-      "loss": 0.8171,
+      "learning_rate": 9.886036786506262e-06,
+      "loss": 0.8918,
       "step": 18273
     },
     {
-      "epoch": 0.5018812996072616,
+      "epoch": 0.5185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.0425060552479754e-05,
-      "loss": 0.7902,
+      "learning_rate": 9.885117768605522e-06,
+      "loss": 0.87,
       "step": 18274
     },
     {
-      "epoch": 0.5019087638351047,
+      "epoch": 0.5185868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 1.0424171838722997e-05,
-      "loss": 0.8666,
+      "learning_rate": 9.884198751675198e-06,
+      "loss": 0.8256,
       "step": 18275
     },
     {
-      "epoch": 0.501936228062948,
+      "epoch": 0.5186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.0423283121610029e-05,
-      "loss": 0.9041,
+      "learning_rate": 9.883279735723052e-06,
+      "loss": 0.9923,
       "step": 18276
     },
     {
-      "epoch": 0.5019636922907913,
+      "epoch": 0.5186435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 1.0422394401147879e-05,
-      "loss": 0.9135,
+      "learning_rate": 9.882360720756848e-06,
+      "loss": 0.9164,
       "step": 18277
     },
     {
-      "epoch": 0.5019911565186345,
+      "epoch": 0.518671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.0421505677343585e-05,
-      "loss": 0.8314,
+      "learning_rate": 9.881441706784348e-06,
+      "loss": 0.8893,
       "step": 18278
     },
     {
-      "epoch": 0.5020186207464777,
+      "epoch": 0.518700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.0420616950204173e-05,
-      "loss": 0.8556,
+      "learning_rate": 9.880522693813313e-06,
+      "loss": 0.955,
       "step": 18279
     },
     {
-      "epoch": 0.5020460849743209,
+      "epoch": 0.5187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.0419728219736675e-05,
-      "loss": 0.8873,
+      "learning_rate": 9.879603681851506e-06,
+      "loss": 0.8679,
       "step": 18280
     },
     {
-      "epoch": 0.5020735492021642,
+      "epoch": 0.5187570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.0418839485948124e-05,
-      "loss": 0.9268,
+      "learning_rate": 9.878684670906697e-06,
+      "loss": 0.9318,
       "step": 18281
     },
     {
-      "epoch": 0.5021010134300075,
+      "epoch": 0.5187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.0417950748845558e-05,
-      "loss": 0.9575,
+      "learning_rate": 9.877765660986643e-06,
+      "loss": 0.9097,
       "step": 18282
     },
     {
-      "epoch": 0.5021284776578506,
+      "epoch": 0.5188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.0417062008436001e-05,
-      "loss": 0.8245,
+      "learning_rate": 9.876846652099107e-06,
+      "loss": 0.8721,
       "step": 18283
     },
     {
-      "epoch": 0.5021559418856939,
+      "epoch": 0.5188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.0416173264726487e-05,
-      "loss": 0.8304,
+      "learning_rate": 9.875927644251856e-06,
+      "loss": 0.9064,
       "step": 18284
     },
     {
-      "epoch": 0.5021834061135371,
+      "epoch": 0.5188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.0415284517724049e-05,
-      "loss": 0.8383,
+      "learning_rate": 9.875008637452647e-06,
+      "loss": 0.8762,
       "step": 18285
     },
     {
-      "epoch": 0.5022108703413803,
+      "epoch": 0.5188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.0414395767435719e-05,
-      "loss": 0.9514,
+      "learning_rate": 9.874089631709245e-06,
+      "loss": 0.7245,
       "step": 18286
     },
     {
-      "epoch": 0.5022383345692236,
+      "epoch": 0.5189273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.0413507013868528e-05,
-      "loss": 1.0002,
+      "learning_rate": 9.873170627029416e-06,
+      "loss": 0.8515,
       "step": 18287
     },
     {
-      "epoch": 0.5022657987970668,
+      "epoch": 0.5189557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 1.0412618257029512e-05,
-      "loss": 0.9741,
+      "learning_rate": 9.87225162342092e-06,
+      "loss": 0.8476,
       "step": 18288
     },
     {
-      "epoch": 0.5022932630249101,
+      "epoch": 0.5189841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.0411729496925697e-05,
-      "loss": 0.9666,
+      "learning_rate": 9.871332620891519e-06,
+      "loss": 0.7605,
       "step": 18289
     },
     {
-      "epoch": 0.5023207272527533,
+      "epoch": 0.5190124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.0410840733564122e-05,
-      "loss": 0.9032,
+      "learning_rate": 9.870413619448977e-06,
+      "loss": 0.8935,
       "step": 18290
     },
     {
-      "epoch": 0.5023481914805965,
+      "epoch": 0.5190408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.0409951966951818e-05,
-      "loss": 0.9776,
+      "learning_rate": 9.869494619101058e-06,
+      "loss": 0.9523,
       "step": 18291
     },
     {
-      "epoch": 0.5023756557084398,
+      "epoch": 0.5190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 1.040906319709581e-05,
-      "loss": 0.9465,
+      "learning_rate": 9.868575619855525e-06,
+      "loss": 0.8636,
       "step": 18292
     },
     {
-      "epoch": 0.502403119936283,
+      "epoch": 0.5190976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 1.0408174424003137e-05,
-      "loss": 0.9678,
+      "learning_rate": 9.867656621720137e-06,
+      "loss": 0.9355,
       "step": 18293
     },
     {
-      "epoch": 0.5024305841641262,
+      "epoch": 0.5191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.0407285647680834e-05,
-      "loss": 0.9337,
+      "learning_rate": 9.866737624702661e-06,
+      "loss": 0.8552,
       "step": 18294
     },
     {
-      "epoch": 0.5024580483919695,
+      "epoch": 0.5191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.0406396868135924e-05,
-      "loss": 0.8741,
+      "learning_rate": 9.865818628810853e-06,
+      "loss": 0.7258,
       "step": 18295
     },
     {
-      "epoch": 0.5024855126198127,
+      "epoch": 0.5191827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 1.0405508085375447e-05,
-      "loss": 0.8643,
+      "learning_rate": 9.864899634052488e-06,
+      "loss": 0.8062,
       "step": 18296
     },
     {
-      "epoch": 0.502512976847656,
+      "epoch": 0.519211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.0404619299406434e-05,
-      "loss": 0.8446,
+      "learning_rate": 9.863980640435317e-06,
+      "loss": 0.9081,
       "step": 18297
     },
     {
-      "epoch": 0.5025404410754991,
+      "epoch": 0.5192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.0403730510235914e-05,
-      "loss": 0.8152,
+      "learning_rate": 9.863061647967113e-06,
+      "loss": 0.8659,
       "step": 18298
     },
     {
-      "epoch": 0.5025679053033424,
+      "epoch": 0.5192678774120317,
       "grad_norm": 0.0,
-      "learning_rate": 1.0402841717870921e-05,
-      "loss": 0.9695,
+      "learning_rate": 9.862142656655631e-06,
+      "loss": 0.7947,
       "step": 18299
     },
     {
-      "epoch": 0.5025953695311857,
+      "epoch": 0.5192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.0401952922318493e-05,
-      "loss": 0.8806,
+      "learning_rate": 9.861223666508635e-06,
+      "loss": 0.9537,
       "step": 18300
     },
     {
-      "epoch": 0.5026228337590288,
+      "epoch": 0.5193246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 1.0401064123585652e-05,
-      "loss": 0.9759,
+      "learning_rate": 9.860304677533891e-06,
+      "loss": 0.8788,
       "step": 18301
     },
     {
-      "epoch": 0.5026502979868721,
+      "epoch": 0.5193530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 1.0400175321679443e-05,
-      "loss": 0.9353,
+      "learning_rate": 9.859385689739157e-06,
+      "loss": 0.7781,
       "step": 18302
     },
     {
-      "epoch": 0.5026777622147154,
+      "epoch": 0.5193813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.0399286516606886e-05,
-      "loss": 0.8901,
+      "learning_rate": 9.8584667031322e-06,
+      "loss": 0.8626,
       "step": 18303
     },
     {
-      "epoch": 0.5027052264425586,
+      "epoch": 0.5194097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 1.0398397708375018e-05,
-      "loss": 0.8765,
+      "learning_rate": 9.857547717720783e-06,
+      "loss": 0.9631,
       "step": 18304
     },
     {
-      "epoch": 0.5027326906704018,
+      "epoch": 0.5194381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.0397508896990878e-05,
-      "loss": 0.9611,
+      "learning_rate": 9.856628733512665e-06,
+      "loss": 0.9157,
       "step": 18305
     },
     {
-      "epoch": 0.502760154898245,
+      "epoch": 0.5194665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 1.0396620082461493e-05,
-      "loss": 0.8995,
+      "learning_rate": 9.855709750515609e-06,
+      "loss": 0.9664,
       "step": 18306
     },
     {
-      "epoch": 0.5027876191260883,
+      "epoch": 0.5194948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.0395731264793893e-05,
-      "loss": 0.9382,
+      "learning_rate": 9.85479076873738e-06,
+      "loss": 0.8788,
       "step": 18307
     },
     {
-      "epoch": 0.5028150833539315,
+      "epoch": 0.5195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.0394842443995119e-05,
-      "loss": 0.9215,
+      "learning_rate": 9.853871788185743e-06,
+      "loss": 0.7856,
       "step": 18308
     },
     {
-      "epoch": 0.5028425475817747,
+      "epoch": 0.5195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.0393953620072192e-05,
-      "loss": 0.8578,
+      "learning_rate": 9.852952808868454e-06,
+      "loss": 0.8804,
       "step": 18309
     },
     {
-      "epoch": 0.502870011809618,
+      "epoch": 0.5195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.0393064793032158e-05,
-      "loss": 0.8496,
+      "learning_rate": 9.852033830793281e-06,
+      "loss": 0.8162,
       "step": 18310
     },
     {
-      "epoch": 0.5028974760374612,
+      "epoch": 0.5196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.039217596288204e-05,
-      "loss": 0.8408,
+      "learning_rate": 9.85111485396798e-06,
+      "loss": 0.8958,
       "step": 18311
     },
     {
-      "epoch": 0.5029249402653044,
+      "epoch": 0.5196367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 1.0391287129628873e-05,
-      "loss": 0.8929,
+      "learning_rate": 9.850195878400325e-06,
+      "loss": 0.7849,
       "step": 18312
     },
     {
-      "epoch": 0.5029524044931477,
+      "epoch": 0.5196651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.039039829327969e-05,
-      "loss": 0.8061,
+      "learning_rate": 9.84927690409807e-06,
+      "loss": 0.6904,
       "step": 18313
     },
     {
-      "epoch": 0.5029798687209909,
+      "epoch": 0.5196935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.0389509453841531e-05,
-      "loss": 0.9152,
+      "learning_rate": 9.848357931068978e-06,
+      "loss": 0.9006,
       "step": 18314
     },
     {
-      "epoch": 0.5030073329488342,
+      "epoch": 0.51972190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.0388620611321415e-05,
-      "loss": 0.9175,
+      "learning_rate": 9.847438959320816e-06,
+      "loss": 0.8685,
       "step": 18315
     },
     {
-      "epoch": 0.5030347971766774,
+      "epoch": 0.5197502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 1.0387731765726384e-05,
-      "loss": 0.8582,
+      "learning_rate": 9.846519988861343e-06,
+      "loss": 0.8557,
       "step": 18316
     },
     {
-      "epoch": 0.5030622614045206,
+      "epoch": 0.5197786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 1.0386842917063472e-05,
-      "loss": 0.8543,
+      "learning_rate": 9.845601019698323e-06,
+      "loss": 0.8515,
       "step": 18317
     },
     {
-      "epoch": 0.5030897256323639,
+      "epoch": 0.5198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.0385954065339706e-05,
-      "loss": 0.8413,
+      "learning_rate": 9.844682051839517e-06,
+      "loss": 0.8897,
       "step": 18318
     },
     {
-      "epoch": 0.503117189860207,
+      "epoch": 0.5198354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 1.0385065210562122e-05,
-      "loss": 0.829,
+      "learning_rate": 9.843763085292692e-06,
+      "loss": 0.9355,
       "step": 18319
     },
     {
-      "epoch": 0.5031446540880503,
+      "epoch": 0.5198637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 1.0384176352737754e-05,
-      "loss": 0.8718,
+      "learning_rate": 9.842844120065603e-06,
+      "loss": 0.7852,
       "step": 18320
     },
     {
-      "epoch": 0.5031721183158936,
+      "epoch": 0.5198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.0383287491873632e-05,
-      "loss": 0.8413,
+      "learning_rate": 9.84192515616602e-06,
+      "loss": 0.8533,
       "step": 18321
     },
     {
-      "epoch": 0.5031995825437368,
+      "epoch": 0.5199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.0382398627976793e-05,
-      "loss": 0.9848,
+      "learning_rate": 9.8410061936017e-06,
+      "loss": 0.8566,
       "step": 18322
     },
     {
-      "epoch": 0.50322704677158,
+      "epoch": 0.5199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.0381509761054267e-05,
-      "loss": 0.915,
+      "learning_rate": 9.840087232380408e-06,
+      "loss": 0.8772,
       "step": 18323
     },
     {
-      "epoch": 0.5032545109994232,
+      "epoch": 0.5199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.0380620891113089e-05,
-      "loss": 0.7998,
+      "learning_rate": 9.839168272509908e-06,
+      "loss": 0.8175,
       "step": 18324
     },
     {
-      "epoch": 0.5032819752272665,
+      "epoch": 0.520005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.0379732018160289e-05,
-      "loss": 0.813,
+      "learning_rate": 9.83824931399796e-06,
+      "loss": 1.0043,
       "step": 18325
     },
     {
-      "epoch": 0.5033094394551098,
+      "epoch": 0.5200340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 1.0378843142202906e-05,
-      "loss": 0.9818,
+      "learning_rate": 9.837330356852323e-06,
+      "loss": 0.8601,
       "step": 18326
     },
     {
-      "epoch": 0.5033369036829529,
+      "epoch": 0.5200624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.0377954263247964e-05,
-      "loss": 0.877,
+      "learning_rate": 9.836411401080766e-06,
+      "loss": 0.822,
       "step": 18327
     },
     {
-      "epoch": 0.5033643679107962,
+      "epoch": 0.5200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 1.0377065381302505e-05,
-      "loss": 0.9624,
+      "learning_rate": 9.835492446691054e-06,
+      "loss": 0.9293,
       "step": 18328
     },
     {
-      "epoch": 0.5033918321386395,
+      "epoch": 0.5201191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.0376176496373559e-05,
-      "loss": 0.849,
+      "learning_rate": 9.834573493690941e-06,
+      "loss": 0.9427,
       "step": 18329
     },
     {
-      "epoch": 0.5034192963664826,
+      "epoch": 0.5201475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.0375287608468159e-05,
-      "loss": 0.9268,
+      "learning_rate": 9.833654542088192e-06,
+      "loss": 0.9393,
       "step": 18330
     },
     {
-      "epoch": 0.5034467605943259,
+      "epoch": 0.5201759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.0374398717593334e-05,
-      "loss": 0.8868,
+      "learning_rate": 9.832735591890575e-06,
+      "loss": 0.8613,
       "step": 18331
     },
     {
-      "epoch": 0.5034742248221691,
+      "epoch": 0.5202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.0373509823756126e-05,
-      "loss": 0.9167,
+      "learning_rate": 9.831816643105845e-06,
+      "loss": 0.7927,
       "step": 18332
     },
     {
-      "epoch": 0.5035016890500124,
+      "epoch": 0.5202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.037262092696356e-05,
-      "loss": 0.8611,
+      "learning_rate": 9.83089769574177e-06,
+      "loss": 0.9006,
       "step": 18333
     },
     {
-      "epoch": 0.5035291532778556,
+      "epoch": 0.520261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.0371732027222677e-05,
-      "loss": 0.7924,
+      "learning_rate": 9.829978749806105e-06,
+      "loss": 0.8491,
       "step": 18334
     },
     {
-      "epoch": 0.5035566175056988,
+      "epoch": 0.5202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.0370843124540504e-05,
-      "loss": 0.9902,
+      "learning_rate": 9.82905980530662e-06,
+      "loss": 0.8417,
       "step": 18335
     },
     {
-      "epoch": 0.5035840817335421,
+      "epoch": 0.5203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.0369954218924074e-05,
-      "loss": 0.9236,
+      "learning_rate": 9.828140862251076e-06,
+      "loss": 0.8352,
       "step": 18336
     },
     {
-      "epoch": 0.5036115459613852,
+      "epoch": 0.5203461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.0369065310380424e-05,
-      "loss": 0.8197,
+      "learning_rate": 9.827221920647231e-06,
+      "loss": 0.8362,
       "step": 18337
     },
     {
-      "epoch": 0.5036390101892285,
+      "epoch": 0.5203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.0368176398916592e-05,
-      "loss": 0.7811,
+      "learning_rate": 9.826302980502853e-06,
+      "loss": 0.8983,
       "step": 18338
     },
     {
-      "epoch": 0.5036664744170718,
+      "epoch": 0.5204029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.03672874845396e-05,
-      "loss": 0.9783,
+      "learning_rate": 9.8253840418257e-06,
+      "loss": 0.8284,
       "step": 18339
     },
     {
-      "epoch": 0.503693938644915,
+      "epoch": 0.5204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.0366398567256487e-05,
-      "loss": 0.8863,
+      "learning_rate": 9.824465104623534e-06,
+      "loss": 0.8787,
       "step": 18340
     },
     {
-      "epoch": 0.5037214028727582,
+      "epoch": 0.5204597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.0365509647074288e-05,
-      "loss": 0.932,
+      "learning_rate": 9.823546168904117e-06,
+      "loss": 0.9191,
       "step": 18341
     },
     {
-      "epoch": 0.5037488671006015,
+      "epoch": 0.5204880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.0364620724000036e-05,
-      "loss": 0.8757,
+      "learning_rate": 9.822627234675218e-06,
+      "loss": 0.9383,
       "step": 18342
     },
     {
-      "epoch": 0.5037763313284447,
+      "epoch": 0.520516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.0363731798040762e-05,
-      "loss": 0.8804,
+      "learning_rate": 9.821708301944596e-06,
+      "loss": 0.9239,
       "step": 18343
     },
     {
-      "epoch": 0.503803795556288,
+      "epoch": 0.520544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.0362842869203503e-05,
-      "loss": 0.891,
+      "learning_rate": 9.820789370720008e-06,
+      "loss": 0.9498,
       "step": 18344
     },
     {
-      "epoch": 0.5038312597841311,
+      "epoch": 0.5205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.0361953937495288e-05,
-      "loss": 0.9176,
+      "learning_rate": 9.819870441009222e-06,
+      "loss": 0.9867,
       "step": 18345
     },
     {
-      "epoch": 0.5038587240119744,
+      "epoch": 0.5206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.0361065002923154e-05,
-      "loss": 0.9645,
+      "learning_rate": 9.81895151282e-06,
+      "loss": 0.9637,
       "step": 18346
     },
     {
-      "epoch": 0.5038861882398177,
+      "epoch": 0.5206299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 1.0360176065494136e-05,
-      "loss": 0.8855,
+      "learning_rate": 9.8180325861601e-06,
+      "loss": 0.7528,
       "step": 18347
     },
     {
-      "epoch": 0.5039136524676608,
+      "epoch": 0.5206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.0359287125215263e-05,
-      "loss": 0.8201,
+      "learning_rate": 9.81711366103729e-06,
+      "loss": 0.8036,
       "step": 18348
     },
     {
-      "epoch": 0.5039411166955041,
+      "epoch": 0.5206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.0358398182093573e-05,
-      "loss": 0.9447,
+      "learning_rate": 9.816194737459328e-06,
+      "loss": 0.9606,
       "step": 18349
     },
     {
-      "epoch": 0.5039685809233473,
+      "epoch": 0.5207150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 1.0357509236136099e-05,
-      "loss": 0.8987,
+      "learning_rate": 9.815275815433976e-06,
+      "loss": 0.9382,
       "step": 18350
     },
     {
-      "epoch": 0.5039960451511906,
+      "epoch": 0.5207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.0356620287349871e-05,
-      "loss": 0.9276,
+      "learning_rate": 9.814356894968998e-06,
+      "loss": 0.9367,
       "step": 18351
     },
     {
-      "epoch": 0.5040235093790338,
+      "epoch": 0.5207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 1.0355731335741923e-05,
-      "loss": 0.7244,
+      "learning_rate": 9.813437976072158e-06,
+      "loss": 0.8216,
       "step": 18352
     },
     {
-      "epoch": 0.504050973606877,
+      "epoch": 0.5208002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 1.0354842381319294e-05,
-      "loss": 0.9814,
+      "learning_rate": 9.812519058751211e-06,
+      "loss": 0.8745,
       "step": 18353
     },
     {
-      "epoch": 0.5040784378347203,
+      "epoch": 0.5208286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 1.0353953424089015e-05,
-      "loss": 0.9696,
+      "learning_rate": 9.811600143013928e-06,
+      "loss": 0.8716,
       "step": 18354
     },
     {
-      "epoch": 0.5041059020625636,
+      "epoch": 0.5208569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 1.0353064464058116e-05,
-      "loss": 0.9573,
+      "learning_rate": 9.810681228868067e-06,
+      "loss": 0.9013,
       "step": 18355
     },
     {
-      "epoch": 0.5041333662904067,
+      "epoch": 0.5208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.035217550123364e-05,
-      "loss": 0.9318,
+      "learning_rate": 9.809762316321388e-06,
+      "loss": 0.8627,
       "step": 18356
     },
     {
-      "epoch": 0.50416083051825,
+      "epoch": 0.5209137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.035128653562261e-05,
-      "loss": 0.8216,
+      "learning_rate": 9.808843405381653e-06,
+      "loss": 0.7993,
       "step": 18357
     },
     {
-      "epoch": 0.5041882947460932,
+      "epoch": 0.5209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.0350397567232066e-05,
-      "loss": 0.8875,
+      "learning_rate": 9.80792449605663e-06,
+      "loss": 0.8837,
       "step": 18358
     },
     {
-      "epoch": 0.5042157589739364,
+      "epoch": 0.5209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.0349508596069042e-05,
-      "loss": 0.8563,
+      "learning_rate": 9.807005588354077e-06,
+      "loss": 0.8181,
       "step": 18359
     },
     {
-      "epoch": 0.5042432232017797,
+      "epoch": 0.5209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.0348619622140568e-05,
-      "loss": 0.9961,
+      "learning_rate": 9.806086682281759e-06,
+      "loss": 0.847,
       "step": 18360
     },
     {
-      "epoch": 0.5042706874296229,
+      "epoch": 0.5210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.0347730645453685e-05,
-      "loss": 0.8442,
+      "learning_rate": 9.80516777784743e-06,
+      "loss": 0.8707,
       "step": 18361
     },
     {
-      "epoch": 0.5042981516574662,
+      "epoch": 0.52105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.034684166601542e-05,
-      "loss": 0.8911,
+      "learning_rate": 9.804248875058862e-06,
+      "loss": 0.8962,
       "step": 18362
     },
     {
-      "epoch": 0.5043256158853093,
+      "epoch": 0.5210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.0345952683832808e-05,
-      "loss": 0.933,
+      "learning_rate": 9.803329973923811e-06,
+      "loss": 0.8588,
       "step": 18363
     },
     {
-      "epoch": 0.5043530801131526,
+      "epoch": 0.5211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.0345063698912884e-05,
-      "loss": 0.9372,
+      "learning_rate": 9.80241107445004e-06,
+      "loss": 0.8588,
       "step": 18364
     },
     {
-      "epoch": 0.5043805443409959,
+      "epoch": 0.5211407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.0344174711262684e-05,
-      "loss": 0.883,
+      "learning_rate": 9.801492176645313e-06,
+      "loss": 0.9223,
       "step": 18365
     },
     {
-      "epoch": 0.504408008568839,
+      "epoch": 0.5211691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.034328572088924e-05,
-      "loss": 0.929,
+      "learning_rate": 9.80057328051739e-06,
+      "loss": 0.8344,
       "step": 18366
     },
     {
-      "epoch": 0.5044354727966823,
+      "epoch": 0.5211975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 1.0342396727799589e-05,
-      "loss": 0.8393,
+      "learning_rate": 9.799654386074032e-06,
+      "loss": 0.9504,
       "step": 18367
     },
     {
-      "epoch": 0.5044629370245256,
+      "epoch": 0.5212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 1.0341507732000757e-05,
-      "loss": 0.9308,
+      "learning_rate": 9.798735493323004e-06,
+      "loss": 0.8609,
       "step": 18368
     },
     {
-      "epoch": 0.5044904012523688,
+      "epoch": 0.5212542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.0340618733499785e-05,
-      "loss": 0.8707,
+      "learning_rate": 9.797816602272067e-06,
+      "loss": 0.9586,
       "step": 18369
     },
     {
-      "epoch": 0.504517865480212,
+      "epoch": 0.5212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 1.0339729732303711e-05,
-      "loss": 0.9005,
+      "learning_rate": 9.79689771292898e-06,
+      "loss": 0.9474,
       "step": 18370
     },
     {
-      "epoch": 0.5045453297080552,
+      "epoch": 0.5213110102156641,
       "grad_norm": 0.0,
-      "learning_rate": 1.0338840728419559e-05,
-      "loss": 0.9818,
+      "learning_rate": 9.79597882530151e-06,
+      "loss": 0.9009,
       "step": 18371
     },
     {
-      "epoch": 0.5045727939358985,
+      "epoch": 0.5213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.0337951721854367e-05,
-      "loss": 0.8718,
+      "learning_rate": 9.79505993939741e-06,
+      "loss": 0.8562,
       "step": 18372
     },
     {
-      "epoch": 0.5046002581637418,
+      "epoch": 0.5213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.0337062712615174e-05,
-      "loss": 0.7833,
+      "learning_rate": 9.794141055224451e-06,
+      "loss": 0.9456,
       "step": 18373
     },
     {
-      "epoch": 0.5046277223915849,
+      "epoch": 0.5213961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 1.0336173700709009e-05,
-      "loss": 0.9507,
+      "learning_rate": 9.793222172790396e-06,
+      "loss": 0.7801,
       "step": 18374
     },
     {
-      "epoch": 0.5046551866194282,
+      "epoch": 0.5214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.0335284686142905e-05,
-      "loss": 0.8338,
+      "learning_rate": 9.792303292102997e-06,
+      "loss": 0.816,
       "step": 18375
     },
     {
-      "epoch": 0.5046826508472714,
+      "epoch": 0.5214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 1.0334395668923903e-05,
-      "loss": 0.9075,
+      "learning_rate": 9.791384413170026e-06,
+      "loss": 0.7755,
       "step": 18376
     },
     {
-      "epoch": 0.5047101150751147,
+      "epoch": 0.5214812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 1.0333506649059026e-05,
-      "loss": 0.8291,
+      "learning_rate": 9.790465535999239e-06,
+      "loss": 0.9108,
       "step": 18377
     },
     {
-      "epoch": 0.5047375793029579,
+      "epoch": 0.5215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.0332617626555321e-05,
-      "loss": 0.9282,
+      "learning_rate": 9.789546660598396e-06,
+      "loss": 0.9195,
       "step": 18378
     },
     {
-      "epoch": 0.5047650435308011,
+      "epoch": 0.5215380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 1.0331728601419816e-05,
-      "loss": 0.8025,
+      "learning_rate": 9.788627786975264e-06,
+      "loss": 0.9489,
       "step": 18379
     },
     {
-      "epoch": 0.5047925077586444,
+      "epoch": 0.521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.0330839573659543e-05,
-      "loss": 0.9338,
+      "learning_rate": 9.787708915137604e-06,
+      "loss": 0.8878,
       "step": 18380
     },
     {
-      "epoch": 0.5048199719864876,
+      "epoch": 0.521594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.032995054328154e-05,
-      "loss": 0.9485,
+      "learning_rate": 9.786790045093175e-06,
+      "loss": 0.9722,
       "step": 18381
     },
     {
-      "epoch": 0.5048474362143308,
+      "epoch": 0.5216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.0329061510292847e-05,
-      "loss": 0.8412,
+      "learning_rate": 9.785871176849739e-06,
+      "loss": 0.9323,
       "step": 18382
     },
     {
-      "epoch": 0.5048749004421741,
+      "epoch": 0.5216515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.0328172474700481e-05,
-      "loss": 1.0362,
+      "learning_rate": 9.784952310415062e-06,
+      "loss": 0.8795,
       "step": 18383
     },
     {
-      "epoch": 0.5049023646700173,
+      "epoch": 0.5216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.0327283436511493e-05,
-      "loss": 0.8935,
+      "learning_rate": 9.7840334457969e-06,
+      "loss": 0.9181,
       "step": 18384
     },
     {
-      "epoch": 0.5049298288978605,
+      "epoch": 0.5217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.0326394395732912e-05,
-      "loss": 0.9985,
+      "learning_rate": 9.783114583003018e-06,
+      "loss": 0.9913,
       "step": 18385
     },
     {
-      "epoch": 0.5049572931257038,
+      "epoch": 0.5217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.0325505352371773e-05,
-      "loss": 0.9221,
+      "learning_rate": 9.782195722041176e-06,
+      "loss": 0.9266,
       "step": 18386
     },
     {
-      "epoch": 0.504984757353547,
+      "epoch": 0.5217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.0324616306435107e-05,
-      "loss": 0.9587,
+      "learning_rate": 9.781276862919133e-06,
+      "loss": 0.9327,
       "step": 18387
     },
     {
-      "epoch": 0.5050122215813903,
+      "epoch": 0.5217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.0323727257929952e-05,
-      "loss": 0.8403,
+      "learning_rate": 9.780358005644657e-06,
+      "loss": 0.8894,
       "step": 18388
     },
     {
-      "epoch": 0.5050396858092334,
+      "epoch": 0.5218217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.032283820686334e-05,
-      "loss": 0.9505,
+      "learning_rate": 9.77943915022551e-06,
+      "loss": 0.807,
       "step": 18389
     },
     {
-      "epoch": 0.5050671500370767,
+      "epoch": 0.521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.032194915324231e-05,
-      "loss": 0.8399,
+      "learning_rate": 9.778520296669449e-06,
+      "loss": 0.8576,
       "step": 18390
     },
     {
-      "epoch": 0.50509461426492,
+      "epoch": 0.5218785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 1.0321060097073891e-05,
-      "loss": 0.9753,
+      "learning_rate": 9.777601444984234e-06,
+      "loss": 0.8094,
       "step": 18391
     },
     {
-      "epoch": 0.5051220784927631,
+      "epoch": 0.5219069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.0320171038365121e-05,
-      "loss": 0.8909,
+      "learning_rate": 9.776682595177633e-06,
+      "loss": 0.8884,
       "step": 18392
     },
     {
-      "epoch": 0.5051495427206064,
+      "epoch": 0.5219353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.0319281977123033e-05,
-      "loss": 0.8628,
+      "learning_rate": 9.775763747257404e-06,
+      "loss": 0.7649,
       "step": 18393
     },
     {
-      "epoch": 0.5051770069484497,
+      "epoch": 0.5219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 1.0318392913354667e-05,
-      "loss": 0.9608,
+      "learning_rate": 9.774844901231306e-06,
+      "loss": 0.8853,
       "step": 18394
     },
     {
-      "epoch": 0.5052044711762929,
+      "epoch": 0.5219920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 1.0317503847067048e-05,
-      "loss": 0.7971,
+      "learning_rate": 9.773926057107106e-06,
+      "loss": 0.9014,
       "step": 18395
     },
     {
-      "epoch": 0.5052319354041361,
+      "epoch": 0.5220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.0316614778267215e-05,
-      "loss": 0.8632,
+      "learning_rate": 9.773007214892562e-06,
+      "loss": 0.8902,
       "step": 18396
     },
     {
-      "epoch": 0.5052593996319793,
+      "epoch": 0.5220488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.0315725706962206e-05,
-      "loss": 0.7924,
+      "learning_rate": 9.772088374595435e-06,
+      "loss": 0.7845,
       "step": 18397
     },
     {
-      "epoch": 0.5052868638598226,
+      "epoch": 0.5220771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 1.0314836633159053e-05,
-      "loss": 0.9428,
+      "learning_rate": 9.77116953622349e-06,
+      "loss": 0.9053,
       "step": 18398
     },
     {
-      "epoch": 0.5053143280876659,
+      "epoch": 0.522105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.0313947556864787e-05,
-      "loss": 0.8668,
+      "learning_rate": 9.770250699784486e-06,
+      "loss": 0.8398,
       "step": 18399
     },
     {
-      "epoch": 0.505341792315509,
+      "epoch": 0.5221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.031305847808645e-05,
-      "loss": 0.8385,
+      "learning_rate": 9.769331865286185e-06,
+      "loss": 0.9464,
       "step": 18400
     },
     {
-      "epoch": 0.5053692565433523,
+      "epoch": 0.5221623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 1.0312169396831071e-05,
-      "loss": 0.8835,
+      "learning_rate": 9.768413032736345e-06,
+      "loss": 0.899,
       "step": 18401
     },
     {
-      "epoch": 0.5053967207711955,
+      "epoch": 0.5221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.031128031310569e-05,
-      "loss": 0.8403,
+      "learning_rate": 9.767494202142734e-06,
+      "loss": 0.881,
       "step": 18402
     },
     {
-      "epoch": 0.5054241849990387,
+      "epoch": 0.5222190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 1.0310391226917336e-05,
-      "loss": 0.9586,
+      "learning_rate": 9.766575373513102e-06,
+      "loss": 0.8183,
       "step": 18403
     },
     {
-      "epoch": 0.505451649226882,
+      "epoch": 0.5222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 1.0309502138273044e-05,
-      "loss": 0.9344,
+      "learning_rate": 9.765656546855226e-06,
+      "loss": 0.9882,
       "step": 18404
     },
     {
-      "epoch": 0.5054791134547252,
+      "epoch": 0.5222758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.0308613047179854e-05,
-      "loss": 0.8615,
+      "learning_rate": 9.764737722176858e-06,
+      "loss": 0.8921,
       "step": 18405
     },
     {
-      "epoch": 0.5055065776825685,
+      "epoch": 0.5223041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 1.03077239536448e-05,
-      "loss": 0.8801,
+      "learning_rate": 9.763818899485761e-06,
+      "loss": 0.8555,
       "step": 18406
     },
     {
-      "epoch": 0.5055340419104117,
+      "epoch": 0.5223325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.0306834857674912e-05,
-      "loss": 0.8698,
+      "learning_rate": 9.762900078789698e-06,
+      "loss": 0.8608,
       "step": 18407
     },
     {
-      "epoch": 0.5055615061382549,
+      "epoch": 0.522360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.0305945759277227e-05,
-      "loss": 0.8423,
+      "learning_rate": 9.761981260096425e-06,
+      "loss": 0.9078,
       "step": 18408
     },
     {
-      "epoch": 0.5055889703660982,
+      "epoch": 0.52238933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.0305056658458783e-05,
-      "loss": 0.9875,
+      "learning_rate": 9.761062443413711e-06,
+      "loss": 0.8233,
       "step": 18409
     },
     {
-      "epoch": 0.5056164345939413,
+      "epoch": 0.5224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.0304167555226612e-05,
-      "loss": 0.9342,
+      "learning_rate": 9.760143628749312e-06,
+      "loss": 0.8926,
       "step": 18410
     },
     {
-      "epoch": 0.5056438988217846,
+      "epoch": 0.5224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.0303278449587748e-05,
-      "loss": 0.9485,
+      "learning_rate": 9.75922481611099e-06,
+      "loss": 0.87,
       "step": 18411
     },
     {
-      "epoch": 0.5056713630496279,
+      "epoch": 0.5224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.030238934154923e-05,
-      "loss": 0.8375,
+      "learning_rate": 9.758306005506508e-06,
+      "loss": 0.9109,
       "step": 18412
     },
     {
-      "epoch": 0.5056988272774711,
+      "epoch": 0.5225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.0301500231118087e-05,
-      "loss": 1.0045,
+      "learning_rate": 9.757387196943626e-06,
+      "loss": 0.7986,
       "step": 18413
     },
     {
-      "epoch": 0.5057262915053143,
+      "epoch": 0.5225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.030061111830136e-05,
-      "loss": 0.9181,
+      "learning_rate": 9.756468390430101e-06,
+      "loss": 0.8987,
       "step": 18414
     },
     {
-      "epoch": 0.5057537557331576,
+      "epoch": 0.5225595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.0299722003106083e-05,
-      "loss": 0.7709,
+      "learning_rate": 9.755549585973704e-06,
+      "loss": 0.9217,
       "step": 18415
     },
     {
-      "epoch": 0.5057812199610008,
+      "epoch": 0.5225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.0298832885539285e-05,
-      "loss": 0.818,
+      "learning_rate": 9.754630783582189e-06,
+      "loss": 0.9472,
       "step": 18416
     },
     {
-      "epoch": 0.5058086841888441,
+      "epoch": 0.522616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.0297943765608007e-05,
-      "loss": 0.923,
+      "learning_rate": 9.753711983263316e-06,
+      "loss": 0.7732,
       "step": 18417
     },
     {
-      "epoch": 0.5058361484166872,
+      "epoch": 0.5226447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 1.0297054643319286e-05,
-      "loss": 0.9637,
+      "learning_rate": 9.75279318502485e-06,
+      "loss": 0.9807,
       "step": 18418
     },
     {
-      "epoch": 0.5058636126445305,
+      "epoch": 0.5226730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 1.029616551868015e-05,
-      "loss": 0.8729,
+      "learning_rate": 9.75187438887455e-06,
+      "loss": 0.7318,
       "step": 18419
     },
     {
-      "epoch": 0.5058910768723738,
+      "epoch": 0.5227014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.0295276391697637e-05,
-      "loss": 0.937,
+      "learning_rate": 9.750955594820182e-06,
+      "loss": 0.9137,
       "step": 18420
     },
     {
-      "epoch": 0.505918541100217,
+      "epoch": 0.5227298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.0294387262378784e-05,
-      "loss": 0.8343,
+      "learning_rate": 9.750036802869503e-06,
+      "loss": 0.9619,
       "step": 18421
     },
     {
-      "epoch": 0.5059460053280602,
+      "epoch": 0.5227582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 1.0293498130730626e-05,
-      "loss": 0.8785,
+      "learning_rate": 9.74911801303027e-06,
+      "loss": 0.8776,
       "step": 18422
     },
     {
-      "epoch": 0.5059734695559034,
+      "epoch": 0.5227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.0292608996760199e-05,
-      "loss": 0.7916,
+      "learning_rate": 9.748199225310254e-06,
+      "loss": 0.9589,
       "step": 18423
     },
     {
-      "epoch": 0.5060009337837467,
+      "epoch": 0.5228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.0291719860474532e-05,
-      "loss": 0.9033,
+      "learning_rate": 9.747280439717208e-06,
+      "loss": 0.9189,
       "step": 18424
     },
     {
-      "epoch": 0.5060283980115899,
+      "epoch": 0.5228433598183881,
       "grad_norm": 0.0,
-      "learning_rate": 1.0290830721880665e-05,
-      "loss": 0.925,
+      "learning_rate": 9.746361656258893e-06,
+      "loss": 0.9008,
       "step": 18425
     },
     {
-      "epoch": 0.5060558622394331,
+      "epoch": 0.5228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.0289941580985635e-05,
-      "loss": 0.7696,
+      "learning_rate": 9.745442874943076e-06,
+      "loss": 0.9414,
       "step": 18426
     },
     {
-      "epoch": 0.5060833264672764,
+      "epoch": 0.522900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.0289052437796473e-05,
-      "loss": 0.9275,
+      "learning_rate": 9.744524095777514e-06,
+      "loss": 0.7492,
       "step": 18427
     },
     {
-      "epoch": 0.5061107906951197,
+      "epoch": 0.5229284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 1.0288163292320214e-05,
-      "loss": 0.8693,
+      "learning_rate": 9.743605318769967e-06,
+      "loss": 0.925,
       "step": 18428
     },
     {
-      "epoch": 0.5061382549229628,
+      "epoch": 0.5229568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.0287274144563902e-05,
-      "loss": 0.8473,
+      "learning_rate": 9.7426865439282e-06,
+      "loss": 0.8651,
       "step": 18429
     },
     {
-      "epoch": 0.5061657191508061,
+      "epoch": 0.5229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 1.0286384994534562e-05,
-      "loss": 0.8375,
+      "learning_rate": 9.74176777125997e-06,
+      "loss": 0.8659,
       "step": 18430
     },
     {
-      "epoch": 0.5061931833786493,
+      "epoch": 0.5230136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.028549584223923e-05,
-      "loss": 0.9481,
+      "learning_rate": 9.740849000773037e-06,
+      "loss": 0.9188,
       "step": 18431
     },
     {
-      "epoch": 0.5062206476064925,
+      "epoch": 0.5230419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 1.0284606687684948e-05,
-      "loss": 0.9037,
+      "learning_rate": 9.739930232475167e-06,
+      "loss": 0.802,
       "step": 18432
     },
     {
-      "epoch": 0.5062481118343358,
+      "epoch": 0.5230703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 1.0283717530878748e-05,
-      "loss": 0.9227,
+      "learning_rate": 9.739011466374113e-06,
+      "loss": 0.7975,
       "step": 18433
     },
     {
-      "epoch": 0.506275576062179,
+      "epoch": 0.5230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.0282828371827663e-05,
-      "loss": 0.9172,
+      "learning_rate": 9.738092702477646e-06,
+      "loss": 0.7908,
       "step": 18434
     },
     {
-      "epoch": 0.5063030402900223,
+      "epoch": 0.5231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.0281939210538732e-05,
-      "loss": 0.8489,
+      "learning_rate": 9.737173940793518e-06,
+      "loss": 0.9357,
       "step": 18435
     },
     {
-      "epoch": 0.5063305045178654,
+      "epoch": 0.523155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.0281050047018986e-05,
-      "loss": 0.878,
+      "learning_rate": 9.736255181329499e-06,
+      "loss": 0.836,
       "step": 18436
     },
     {
-      "epoch": 0.5063579687457087,
+      "epoch": 0.5231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.0280160881275465e-05,
-      "loss": 0.988,
+      "learning_rate": 9.735336424093342e-06,
+      "loss": 0.8747,
       "step": 18437
     },
     {
-      "epoch": 0.506385432973552,
+      "epoch": 0.5232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.0279271713315202e-05,
-      "loss": 0.7909,
+      "learning_rate": 9.734417669092807e-06,
+      "loss": 0.9264,
       "step": 18438
     },
     {
-      "epoch": 0.5064128972013952,
+      "epoch": 0.5232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.0278382543145236e-05,
-      "loss": 0.8325,
+      "learning_rate": 9.733498916335662e-06,
+      "loss": 0.8411,
       "step": 18439
     },
     {
-      "epoch": 0.5064403614292384,
+      "epoch": 0.5232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.0277493370772596e-05,
-      "loss": 0.9071,
+      "learning_rate": 9.732580165829663e-06,
+      "loss": 0.9378,
       "step": 18440
     },
     {
-      "epoch": 0.5064678256570817,
+      "epoch": 0.5232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.0276604196204323e-05,
-      "loss": 0.969,
+      "learning_rate": 9.731661417582571e-06,
+      "loss": 0.7687,
       "step": 18441
     },
     {
-      "epoch": 0.5064952898849249,
+      "epoch": 0.5233257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 1.027571501944745e-05,
-      "loss": 0.8728,
+      "learning_rate": 9.730742671602144e-06,
+      "loss": 0.7922,
       "step": 18442
     },
     {
-      "epoch": 0.5065227541127681,
+      "epoch": 0.5233541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.027482584050901e-05,
-      "loss": 0.9339,
+      "learning_rate": 9.72982392789615e-06,
+      "loss": 0.7329,
       "step": 18443
     },
     {
-      "epoch": 0.5065502183406113,
+      "epoch": 0.5233825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.0273936659396047e-05,
-      "loss": 0.957,
+      "learning_rate": 9.728905186472345e-06,
+      "loss": 0.8499,
       "step": 18444
     },
     {
-      "epoch": 0.5065776825684546,
+      "epoch": 0.523410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.0273047476115585e-05,
-      "loss": 0.8324,
+      "learning_rate": 9.727986447338487e-06,
+      "loss": 0.9369,
       "step": 18445
     },
     {
-      "epoch": 0.5066051467962979,
+      "epoch": 0.523439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.027215829067467e-05,
-      "loss": 0.8961,
+      "learning_rate": 9.727067710502341e-06,
+      "loss": 0.9478,
       "step": 18446
     },
     {
-      "epoch": 0.506632611024141,
+      "epoch": 0.5234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.0271269103080333e-05,
-      "loss": 0.9528,
+      "learning_rate": 9.726148975971668e-06,
+      "loss": 0.9974,
       "step": 18447
     },
     {
-      "epoch": 0.5066600752519843,
+      "epoch": 0.5234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.0270379913339608e-05,
-      "loss": 0.9766,
+      "learning_rate": 9.725230243754224e-06,
+      "loss": 0.9037,
       "step": 18448
     },
     {
-      "epoch": 0.5066875394798275,
+      "epoch": 0.5235244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 1.026949072145953e-05,
-      "loss": 0.8912,
+      "learning_rate": 9.724311513857772e-06,
+      "loss": 0.9457,
       "step": 18449
     },
     {
-      "epoch": 0.5067150037076708,
+      "epoch": 0.5235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.0268601527447145e-05,
-      "loss": 0.8757,
+      "learning_rate": 9.723392786290076e-06,
+      "loss": 0.8402,
       "step": 18450
     },
     {
-      "epoch": 0.506742467935514,
+      "epoch": 0.5235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.0267712331309474e-05,
-      "loss": 0.8891,
+      "learning_rate": 9.722474061058893e-06,
+      "loss": 0.932,
       "step": 18451
     },
     {
-      "epoch": 0.5067699321633572,
+      "epoch": 0.5236095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 1.0266823133053558e-05,
-      "loss": 0.8832,
+      "learning_rate": 9.721555338171982e-06,
+      "loss": 1.0179,
       "step": 18452
     },
     {
-      "epoch": 0.5067973963912005,
+      "epoch": 0.5236379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.0265933932686438e-05,
-      "loss": 0.8485,
+      "learning_rate": 9.720636617637108e-06,
+      "loss": 0.8917,
       "step": 18453
     },
     {
-      "epoch": 0.5068248606190437,
+      "epoch": 0.5236662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 1.0265044730215145e-05,
-      "loss": 0.7814,
+      "learning_rate": 9.71971789946203e-06,
+      "loss": 1.0358,
       "step": 18454
     },
     {
-      "epoch": 0.5068523248468869,
+      "epoch": 0.5236946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 1.0264155525646715e-05,
-      "loss": 0.7757,
+      "learning_rate": 9.718799183654505e-06,
+      "loss": 0.7701,
       "step": 18455
     },
     {
-      "epoch": 0.5068797890747302,
+      "epoch": 0.5237230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.0263266318988186e-05,
-      "loss": 0.9971,
+      "learning_rate": 9.717880470222298e-06,
+      "loss": 0.9081,
       "step": 18456
     },
     {
-      "epoch": 0.5069072533025734,
+      "epoch": 0.5237514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 1.0262377110246589e-05,
-      "loss": 0.8748,
+      "learning_rate": 9.716961759173166e-06,
+      "loss": 0.9325,
       "step": 18457
     },
     {
-      "epoch": 0.5069347175304166,
+      "epoch": 0.5237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.0261487899428963e-05,
-      "loss": 0.8734,
+      "learning_rate": 9.716043050514869e-06,
+      "loss": 0.9191,
       "step": 18458
     },
     {
-      "epoch": 0.5069621817582599,
+      "epoch": 0.5238081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.0260598686542346e-05,
-      "loss": 1.0431,
+      "learning_rate": 9.715124344255173e-06,
+      "loss": 0.9516,
       "step": 18459
     },
     {
-      "epoch": 0.5069896459861031,
+      "epoch": 0.5238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.0259709471593768e-05,
-      "loss": 0.836,
+      "learning_rate": 9.714205640401833e-06,
+      "loss": 0.8904,
       "step": 18460
     },
     {
-      "epoch": 0.5070171102139464,
+      "epoch": 0.5238649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.025882025459027e-05,
-      "loss": 0.887,
+      "learning_rate": 9.71328693896261e-06,
+      "loss": 0.935,
       "step": 18461
     },
     {
-      "epoch": 0.5070445744417895,
+      "epoch": 0.5238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.0257931035538889e-05,
-      "loss": 0.9018,
+      "learning_rate": 9.712368239945265e-06,
+      "loss": 1.0248,
       "step": 18462
     },
     {
-      "epoch": 0.5070720386696328,
+      "epoch": 0.5239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.025704181444665e-05,
-      "loss": 0.9557,
+      "learning_rate": 9.71144954335756e-06,
+      "loss": 0.943,
       "step": 18463
     },
     {
-      "epoch": 0.5070995028974761,
+      "epoch": 0.523950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.0256152591320601e-05,
-      "loss": 0.8963,
+      "learning_rate": 9.710530849207249e-06,
+      "loss": 0.933,
       "step": 18464
     },
     {
-      "epoch": 0.5071269671253192,
+      "epoch": 0.5239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.0255263366167773e-05,
-      "loss": 0.9383,
+      "learning_rate": 9.709612157502103e-06,
+      "loss": 0.9818,
       "step": 18465
     },
     {
-      "epoch": 0.5071544313531625,
+      "epoch": 0.5240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.0254374138995205e-05,
-      "loss": 0.9404,
+      "learning_rate": 9.70869346824987e-06,
+      "loss": 0.9786,
       "step": 18466
     },
     {
-      "epoch": 0.5071818955810058,
+      "epoch": 0.5240351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.0253484909809927e-05,
-      "loss": 0.8323,
+      "learning_rate": 9.70777478145832e-06,
+      "loss": 0.9313,
       "step": 18467
     },
     {
-      "epoch": 0.507209359808849,
+      "epoch": 0.5240635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.025259567861898e-05,
-      "loss": 0.8445,
+      "learning_rate": 9.706856097135212e-06,
+      "loss": 0.9065,
       "step": 18468
     },
     {
-      "epoch": 0.5072368240366922,
+      "epoch": 0.5240919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 1.0251706445429394e-05,
-      "loss": 0.8423,
+      "learning_rate": 9.7059374152883e-06,
+      "loss": 0.7469,
       "step": 18469
     },
     {
-      "epoch": 0.5072642882645354,
+      "epoch": 0.5241203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.0250817210248214e-05,
-      "loss": 0.806,
+      "learning_rate": 9.70501873592535e-06,
+      "loss": 0.8259,
       "step": 18470
     },
     {
-      "epoch": 0.5072917524923787,
+      "epoch": 0.5241486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.0249927973082472e-05,
-      "loss": 0.8916,
+      "learning_rate": 9.70410005905412e-06,
+      "loss": 0.7993,
       "step": 18471
     },
     {
-      "epoch": 0.507319216720222,
+      "epoch": 0.5241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.0249038733939199e-05,
-      "loss": 0.9263,
+      "learning_rate": 9.703181384682368e-06,
+      "loss": 0.8332,
       "step": 18472
     },
     {
-      "epoch": 0.5073466809480651,
+      "epoch": 0.5242054483541431,
       "grad_norm": 0.0,
-      "learning_rate": 1.0248149492825434e-05,
-      "loss": 0.8271,
+      "learning_rate": 9.702262712817857e-06,
+      "loss": 0.9486,
       "step": 18473
     },
     {
-      "epoch": 0.5073741451759084,
+      "epoch": 0.5242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.0247260249748222e-05,
-      "loss": 0.9054,
+      "learning_rate": 9.701344043468349e-06,
+      "loss": 0.916,
       "step": 18474
     },
     {
-      "epoch": 0.5074016094037516,
+      "epoch": 0.5242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.0246371004714583e-05,
-      "loss": 0.7372,
+      "learning_rate": 9.700425376641599e-06,
+      "loss": 0.8719,
       "step": 18475
     },
     {
-      "epoch": 0.5074290736315948,
+      "epoch": 0.5242905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 1.0245481757731563e-05,
-      "loss": 0.984,
+      "learning_rate": 9.699506712345368e-06,
+      "loss": 0.9215,
       "step": 18476
     },
     {
-      "epoch": 0.5074565378594381,
+      "epoch": 0.5243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.02445925088062e-05,
-      "loss": 0.9496,
+      "learning_rate": 9.69858805058742e-06,
+      "loss": 0.7822,
       "step": 18477
     },
     {
-      "epoch": 0.5074840020872813,
+      "epoch": 0.5243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.0243703257945524e-05,
-      "loss": 0.9316,
+      "learning_rate": 9.697669391375509e-06,
+      "loss": 0.8975,
       "step": 18478
     },
     {
-      "epoch": 0.5075114663151246,
+      "epoch": 0.5243757094211123,
       "grad_norm": 0.0,
-      "learning_rate": 1.024281400515657e-05,
-      "loss": 0.9128,
+      "learning_rate": 9.696750734717397e-06,
+      "loss": 0.815,
       "step": 18479
     },
     {
-      "epoch": 0.5075389305429678,
+      "epoch": 0.5244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.0241924750446382e-05,
-      "loss": 0.9239,
+      "learning_rate": 9.695832080620849e-06,
+      "loss": 0.9897,
       "step": 18480
     },
     {
-      "epoch": 0.507566394770811,
+      "epoch": 0.5244324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 1.0241035493821988e-05,
-      "loss": 0.8327,
+      "learning_rate": 9.694913429093621e-06,
+      "loss": 0.8671,
       "step": 18481
     },
     {
-      "epoch": 0.5075938589986543,
+      "epoch": 0.524460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.0240146235290431e-05,
-      "loss": 0.8921,
+      "learning_rate": 9.693994780143475e-06,
+      "loss": 0.7849,
       "step": 18482
     },
     {
-      "epoch": 0.5076213232264974,
+      "epoch": 0.524489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.0239256974858745e-05,
-      "loss": 0.9391,
+      "learning_rate": 9.693076133778164e-06,
+      "loss": 0.9747,
       "step": 18483
     },
     {
-      "epoch": 0.5076487874543407,
+      "epoch": 0.5245175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 1.0238367712533961e-05,
-      "loss": 0.889,
+      "learning_rate": 9.692157490005456e-06,
+      "loss": 0.9043,
       "step": 18484
     },
     {
-      "epoch": 0.507676251682184,
+      "epoch": 0.5245459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.0237478448323122e-05,
-      "loss": 0.8959,
+      "learning_rate": 9.691238848833106e-06,
+      "loss": 0.8195,
       "step": 18485
     },
     {
-      "epoch": 0.5077037159100272,
+      "epoch": 0.5245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.0236589182233263e-05,
-      "loss": 0.975,
+      "learning_rate": 9.690320210268876e-06,
+      "loss": 0.9544,
       "step": 18486
     },
     {
-      "epoch": 0.5077311801378704,
+      "epoch": 0.5246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.0235699914271415e-05,
-      "loss": 0.8231,
+      "learning_rate": 9.689401574320526e-06,
+      "loss": 0.8552,
       "step": 18487
     },
     {
-      "epoch": 0.5077586443657136,
+      "epoch": 0.5246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.0234810644444621e-05,
-      "loss": 0.9115,
+      "learning_rate": 9.688482940995814e-06,
+      "loss": 1.0516,
       "step": 18488
     },
     {
-      "epoch": 0.5077861085935569,
+      "epoch": 0.5246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.0233921372759913e-05,
-      "loss": 0.8819,
+      "learning_rate": 9.687564310302499e-06,
+      "loss": 0.7356,
       "step": 18489
     },
     {
-      "epoch": 0.5078135728214002,
+      "epoch": 0.5246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.0233032099224327e-05,
-      "loss": 0.8237,
+      "learning_rate": 9.686645682248345e-06,
+      "loss": 0.9023,
       "step": 18490
     },
     {
-      "epoch": 0.5078410370492433,
+      "epoch": 0.5247162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.0232142823844903e-05,
-      "loss": 0.8406,
+      "learning_rate": 9.685727056841108e-06,
+      "loss": 0.9425,
       "step": 18491
     },
     {
-      "epoch": 0.5078685012770866,
+      "epoch": 0.524744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.0231253546628672e-05,
-      "loss": 0.9632,
+      "learning_rate": 9.684808434088547e-06,
+      "loss": 0.92,
       "step": 18492
     },
     {
-      "epoch": 0.5078959655049299,
+      "epoch": 0.5247729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 1.0230364267582676e-05,
-      "loss": 0.8452,
+      "learning_rate": 9.683889813998426e-06,
+      "loss": 0.8678,
       "step": 18493
     },
     {
-      "epoch": 0.507923429732773,
+      "epoch": 0.5248013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 1.022947498671395e-05,
-      "loss": 0.8124,
+      "learning_rate": 9.682971196578501e-06,
+      "loss": 0.9389,
       "step": 18494
     },
     {
-      "epoch": 0.5079508939606163,
+      "epoch": 0.5248297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.0228585704029528e-05,
-      "loss": 0.9649,
+      "learning_rate": 9.682052581836529e-06,
+      "loss": 0.7892,
       "step": 18495
     },
     {
-      "epoch": 0.5079783581884595,
+      "epoch": 0.5248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 1.0227696419536443e-05,
-      "loss": 0.8851,
+      "learning_rate": 9.681133969780274e-06,
+      "loss": 0.8814,
       "step": 18496
     },
     {
-      "epoch": 0.5080058224163028,
+      "epoch": 0.5248864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.0226807133241741e-05,
-      "loss": 0.903,
+      "learning_rate": 9.680215360417498e-06,
+      "loss": 0.8412,
       "step": 18497
     },
     {
-      "epoch": 0.508033286644146,
+      "epoch": 0.5249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.0225917845152452e-05,
-      "loss": 0.8977,
+      "learning_rate": 9.679296753755958e-06,
+      "loss": 0.8696,
       "step": 18498
     },
     {
-      "epoch": 0.5080607508719892,
+      "epoch": 0.5249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.0225028555275614e-05,
-      "loss": 0.8862,
+      "learning_rate": 9.67837814980341e-06,
+      "loss": 0.9228,
       "step": 18499
     },
     {
-      "epoch": 0.5080882150998325,
+      "epoch": 0.5249716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 1.022413926361826e-05,
-      "loss": 0.9138,
+      "learning_rate": 9.677459548567617e-06,
+      "loss": 0.8714,
       "step": 18500
     },
     {
-      "epoch": 0.5081156793276757,
+      "epoch": 0.525,
       "grad_norm": 0.0,
-      "learning_rate": 1.0223249970187433e-05,
-      "loss": 0.8712,
+      "learning_rate": 9.676540950056339e-06,
+      "loss": 0.9515,
       "step": 18501
     },
     {
-      "epoch": 0.5081431435555189,
+      "epoch": 0.5250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.0222360674990164e-05,
-      "loss": 0.834,
+      "learning_rate": 9.67562235427733e-06,
+      "loss": 0.9873,
       "step": 18502
     },
     {
-      "epoch": 0.5081706077833622,
+      "epoch": 0.5250567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 1.0221471378033492e-05,
-      "loss": 0.9368,
+      "learning_rate": 9.674703761238358e-06,
+      "loss": 0.8462,
       "step": 18503
     },
     {
-      "epoch": 0.5081980720112054,
+      "epoch": 0.5250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.022058207932445e-05,
-      "loss": 0.9491,
+      "learning_rate": 9.673785170947177e-06,
+      "loss": 1.0433,
       "step": 18504
     },
     {
-      "epoch": 0.5082255362390486,
+      "epoch": 0.5251135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 1.0219692778870076e-05,
-      "loss": 0.8789,
+      "learning_rate": 9.672866583411546e-06,
+      "loss": 0.8385,
       "step": 18505
     },
     {
-      "epoch": 0.5082530004668919,
+      "epoch": 0.5251418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.0218803476677413e-05,
-      "loss": 0.7535,
+      "learning_rate": 9.671947998639229e-06,
+      "loss": 0.8948,
       "step": 18506
     },
     {
-      "epoch": 0.5082804646947351,
+      "epoch": 0.5251702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.0217914172753489e-05,
-      "loss": 0.9014,
+      "learning_rate": 9.671029416637979e-06,
+      "loss": 0.9905,
       "step": 18507
     },
     {
-      "epoch": 0.5083079289225784,
+      "epoch": 0.5251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 1.0217024867105346e-05,
-      "loss": 0.8605,
+      "learning_rate": 9.670110837415559e-06,
+      "loss": 0.9026,
       "step": 18508
     },
     {
-      "epoch": 0.5083353931504215,
+      "epoch": 0.5252270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.0216135559740015e-05,
-      "loss": 0.8666,
+      "learning_rate": 9.669192260979727e-06,
+      "loss": 0.7955,
       "step": 18509
     },
     {
-      "epoch": 0.5083628573782648,
+      "epoch": 0.525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.021524625066454e-05,
-      "loss": 0.9039,
+      "learning_rate": 9.668273687338239e-06,
+      "loss": 0.8012,
       "step": 18510
     },
     {
-      "epoch": 0.5083903216061081,
+      "epoch": 0.525283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.0214356939885947e-05,
-      "loss": 0.9418,
+      "learning_rate": 9.667355116498864e-06,
+      "loss": 0.9235,
       "step": 18511
     },
     {
-      "epoch": 0.5084177858339513,
+      "epoch": 0.5253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.0213467627411285e-05,
-      "loss": 0.8638,
+      "learning_rate": 9.666436548469354e-06,
+      "loss": 0.7898,
       "step": 18512
     },
     {
-      "epoch": 0.5084452500617945,
+      "epoch": 0.5253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.021257831324758e-05,
-      "loss": 0.9291,
+      "learning_rate": 9.665517983257469e-06,
+      "loss": 0.88,
       "step": 18513
     },
     {
-      "epoch": 0.5084727142896377,
+      "epoch": 0.5253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.0211688997401876e-05,
-      "loss": 0.9314,
+      "learning_rate": 9.664599420870967e-06,
+      "loss": 0.8767,
       "step": 18514
     },
     {
-      "epoch": 0.508500178517481,
+      "epoch": 0.5253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.0210799679881206e-05,
-      "loss": 0.8416,
+      "learning_rate": 9.663680861317613e-06,
+      "loss": 0.8875,
       "step": 18515
     },
     {
-      "epoch": 0.5085276427453242,
+      "epoch": 0.5254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.0209910360692605e-05,
-      "loss": 0.9186,
+      "learning_rate": 9.662762304605158e-06,
+      "loss": 0.9292,
       "step": 18516
     },
     {
-      "epoch": 0.5085551069731674,
+      "epoch": 0.5254540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.0209021039843114e-05,
-      "loss": 0.9963,
+      "learning_rate": 9.661843750741368e-06,
+      "loss": 0.8819,
       "step": 18517
     },
     {
-      "epoch": 0.5085825712010107,
+      "epoch": 0.5254824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.0208131717339768e-05,
-      "loss": 0.8679,
+      "learning_rate": 9.660925199733997e-06,
+      "loss": 0.9803,
       "step": 18518
     },
     {
-      "epoch": 0.508610035428854,
+      "epoch": 0.525510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.0207242393189602e-05,
-      "loss": 0.9861,
+      "learning_rate": 9.660006651590805e-06,
+      "loss": 0.9017,
       "step": 18519
     },
     {
-      "epoch": 0.5086374996566971,
+      "epoch": 0.5255391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 1.0206353067399654e-05,
-      "loss": 0.9596,
+      "learning_rate": 9.659088106319554e-06,
+      "loss": 0.8832,
       "step": 18520
     },
     {
-      "epoch": 0.5086649638845404,
+      "epoch": 0.5255675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 1.0205463739976963e-05,
-      "loss": 0.8579,
+      "learning_rate": 9.658169563928002e-06,
+      "loss": 0.8427,
       "step": 18521
     },
     {
-      "epoch": 0.5086924281123836,
+      "epoch": 0.5255959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.020457441092856e-05,
-      "loss": 0.8059,
+      "learning_rate": 9.657251024423904e-06,
+      "loss": 0.7617,
       "step": 18522
     },
     {
-      "epoch": 0.5087198923402269,
+      "epoch": 0.5256242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.0203685080261487e-05,
-      "loss": 0.9241,
+      "learning_rate": 9.656332487815026e-06,
+      "loss": 0.9081,
       "step": 18523
     },
     {
-      "epoch": 0.5087473565680701,
+      "epoch": 0.5256526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 1.020279574798278e-05,
-      "loss": 0.9506,
+      "learning_rate": 9.655413954109121e-06,
+      "loss": 0.9166,
       "step": 18524
     },
     {
-      "epoch": 0.5087748207959133,
+      "epoch": 0.5256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.0201906414099469e-05,
-      "loss": 0.8273,
+      "learning_rate": 9.654495423313945e-06,
+      "loss": 0.905,
       "step": 18525
     },
     {
-      "epoch": 0.5088022850237566,
+      "epoch": 0.5257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.0201017078618602e-05,
-      "loss": 0.9327,
+      "learning_rate": 9.653576895437268e-06,
+      "loss": 0.8318,
       "step": 18526
     },
     {
-      "epoch": 0.5088297492515997,
+      "epoch": 0.5257377979568671,
       "grad_norm": 0.0,
-      "learning_rate": 1.0200127741547208e-05,
-      "loss": 0.8636,
+      "learning_rate": 9.65265837048684e-06,
+      "loss": 0.8657,
       "step": 18527
     },
     {
-      "epoch": 0.508857213479443,
+      "epoch": 0.5257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.0199238402892324e-05,
-      "loss": 0.859,
+      "learning_rate": 9.651739848470423e-06,
+      "loss": 0.8503,
       "step": 18528
     },
     {
-      "epoch": 0.5088846777072863,
+      "epoch": 0.525794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.0198349062660991e-05,
-      "loss": 0.8975,
+      "learning_rate": 9.650821329395777e-06,
+      "loss": 0.9355,
       "step": 18529
     },
     {
-      "epoch": 0.5089121419351295,
+      "epoch": 0.5258229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 1.0197459720860244e-05,
-      "loss": 0.8864,
+      "learning_rate": 9.649902813270655e-06,
+      "loss": 0.8817,
       "step": 18530
     },
     {
-      "epoch": 0.5089396061629727,
+      "epoch": 0.5258513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.0196570377497117e-05,
-      "loss": 0.9381,
+      "learning_rate": 9.648984300102823e-06,
+      "loss": 0.9171,
       "step": 18531
     },
     {
-      "epoch": 0.508967070390816,
+      "epoch": 0.5258796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 1.0195681032578648e-05,
-      "loss": 0.8689,
+      "learning_rate": 9.648065789900035e-06,
+      "loss": 0.8213,
       "step": 18532
     },
     {
-      "epoch": 0.5089945346186592,
+      "epoch": 0.5259080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.0194791686111878e-05,
-      "loss": 0.9525,
+      "learning_rate": 9.647147282670051e-06,
+      "loss": 0.822,
       "step": 18533
     },
     {
-      "epoch": 0.5090219988465025,
+      "epoch": 0.5259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 1.0193902338103841e-05,
-      "loss": 0.9374,
+      "learning_rate": 9.646228778420632e-06,
+      "loss": 0.8966,
       "step": 18534
     },
     {
-      "epoch": 0.5090494630743456,
+      "epoch": 0.5259648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 1.019301298856157e-05,
-      "loss": 0.9005,
+      "learning_rate": 9.645310277159532e-06,
+      "loss": 0.8985,
       "step": 18535
     },
     {
-      "epoch": 0.5090769273021889,
+      "epoch": 0.5259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.0192123637492109e-05,
-      "loss": 0.98,
+      "learning_rate": 9.64439177889451e-06,
+      "loss": 0.7957,
       "step": 18536
     },
     {
-      "epoch": 0.5091043915300322,
+      "epoch": 0.5260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.0191234284902488e-05,
-      "loss": 0.9575,
+      "learning_rate": 9.64347328363333e-06,
+      "loss": 0.9116,
       "step": 18537
     },
     {
-      "epoch": 0.5091318557578753,
+      "epoch": 0.526049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.0190344930799749e-05,
-      "loss": 0.9752,
+      "learning_rate": 9.642554791383747e-06,
+      "loss": 0.8367,
       "step": 18538
     },
     {
-      "epoch": 0.5091593199857186,
+      "epoch": 0.5260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.018945557519093e-05,
-      "loss": 0.8322,
+      "learning_rate": 9.641636302153518e-06,
+      "loss": 0.9667,
       "step": 18539
     },
     {
-      "epoch": 0.5091867842135618,
+      "epoch": 0.5261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.018856621808306e-05,
-      "loss": 0.9355,
+      "learning_rate": 9.640717815950406e-06,
+      "loss": 0.9252,
       "step": 18540
     },
     {
-      "epoch": 0.5092142484414051,
+      "epoch": 0.5261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.0187676859483181e-05,
-      "loss": 0.9098,
+      "learning_rate": 9.639799332782161e-06,
+      "loss": 0.8495,
       "step": 18541
     },
     {
-      "epoch": 0.5092417126692483,
+      "epoch": 0.5261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.0186787499398337e-05,
-      "loss": 0.791,
+      "learning_rate": 9.638880852656552e-06,
+      "loss": 0.9232,
       "step": 18542
     },
     {
-      "epoch": 0.5092691768970915,
+      "epoch": 0.5261918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.0185898137835552e-05,
-      "loss": 0.7939,
+      "learning_rate": 9.637962375581332e-06,
+      "loss": 0.827,
       "step": 18543
     },
     {
-      "epoch": 0.5092966411249348,
+      "epoch": 0.5262202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 1.018500877480187e-05,
-      "loss": 0.9301,
+      "learning_rate": 9.63704390156426e-06,
+      "loss": 0.8559,
       "step": 18544
     },
     {
-      "epoch": 0.5093241053527781,
+      "epoch": 0.5262485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.0184119410304327e-05,
-      "loss": 0.825,
+      "learning_rate": 9.636125430613096e-06,
+      "loss": 0.8364,
       "step": 18545
     },
     {
-      "epoch": 0.5093515695806212,
+      "epoch": 0.5262769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.018323004434996e-05,
-      "loss": 0.8352,
+      "learning_rate": 9.635206962735593e-06,
+      "loss": 0.8773,
       "step": 18546
     },
     {
-      "epoch": 0.5093790338084645,
+      "epoch": 0.526305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.0182340676945808e-05,
-      "loss": 0.9903,
+      "learning_rate": 9.634288497939518e-06,
+      "loss": 0.873,
       "step": 18547
     },
     {
-      "epoch": 0.5094064980363077,
+      "epoch": 0.52633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0181451308098902e-05,
-      "loss": 0.9601,
+      "learning_rate": 9.633370036232622e-06,
+      "loss": 0.8372,
       "step": 18548
     },
     {
-      "epoch": 0.5094339622641509,
+      "epoch": 0.5263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.0180561937816284e-05,
-      "loss": 0.9666,
+      "learning_rate": 9.632451577622667e-06,
+      "loss": 0.7963,
       "step": 18549
     },
     {
-      "epoch": 0.5094614264919942,
+      "epoch": 0.5263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.0179672566104992e-05,
-      "loss": 0.8462,
+      "learning_rate": 9.631533122117409e-06,
+      "loss": 0.8279,
       "step": 18550
     },
     {
-      "epoch": 0.5094888907198374,
+      "epoch": 0.5264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.017878319297206e-05,
-      "loss": 0.8009,
+      "learning_rate": 9.630614669724607e-06,
+      "loss": 0.8531,
       "step": 18551
     },
     {
-      "epoch": 0.5095163549476807,
+      "epoch": 0.5264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.0177893818424525e-05,
-      "loss": 0.9229,
+      "learning_rate": 9.629696220452021e-06,
+      "loss": 1.0054,
       "step": 18552
     },
     {
-      "epoch": 0.5095438191755238,
+      "epoch": 0.5264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.0177004442469427e-05,
-      "loss": 0.8844,
+      "learning_rate": 9.628777774307404e-06,
+      "loss": 0.9925,
       "step": 18553
     },
     {
-      "epoch": 0.5095712834033671,
+      "epoch": 0.5265039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 1.0176115065113803e-05,
-      "loss": 0.9554,
+      "learning_rate": 9.627859331298522e-06,
+      "loss": 0.921,
       "step": 18554
     },
     {
-      "epoch": 0.5095987476312104,
+      "epoch": 0.5265323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.0175225686364686e-05,
-      "loss": 0.8046,
+      "learning_rate": 9.626940891433127e-06,
+      "loss": 0.8098,
       "step": 18555
     },
     {
-      "epoch": 0.5096262118590535,
+      "epoch": 0.5265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 1.0174336306229113e-05,
-      "loss": 0.945,
+      "learning_rate": 9.626022454718976e-06,
+      "loss": 0.8781,
       "step": 18556
     },
     {
-      "epoch": 0.5096536760868968,
+      "epoch": 0.5265891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 1.0173446924714126e-05,
-      "loss": 0.8148,
+      "learning_rate": 9.625104021163833e-06,
+      "loss": 0.8079,
       "step": 18557
     },
     {
-      "epoch": 0.5096811403147401,
+      "epoch": 0.5266174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.017255754182676e-05,
-      "loss": 0.8837,
+      "learning_rate": 9.624185590775454e-06,
+      "loss": 0.9802,
       "step": 18558
     },
     {
-      "epoch": 0.5097086045425833,
+      "epoch": 0.5266458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 1.0171668157574053e-05,
-      "loss": 0.8897,
+      "learning_rate": 9.623267163561596e-06,
+      "loss": 0.8821,
       "step": 18559
     },
     {
-      "epoch": 0.5097360687704265,
+      "epoch": 0.5266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.0170778771963039e-05,
-      "loss": 0.929,
+      "learning_rate": 9.622348739530017e-06,
+      "loss": 0.783,
       "step": 18560
     },
     {
-      "epoch": 0.5097635329982697,
+      "epoch": 0.5267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.0169889385000754e-05,
-      "loss": 0.8841,
+      "learning_rate": 9.621430318688475e-06,
+      "loss": 0.8461,
       "step": 18561
     },
     {
-      "epoch": 0.509790997226113,
+      "epoch": 0.5267309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 1.0168999996694246e-05,
-      "loss": 0.9388,
+      "learning_rate": 9.620511901044729e-06,
+      "loss": 0.9664,
       "step": 18562
     },
     {
-      "epoch": 0.5098184614539563,
+      "epoch": 0.5267593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.016811060705054e-05,
-      "loss": 0.9573,
+      "learning_rate": 9.619593486606533e-06,
+      "loss": 0.9869,
       "step": 18563
     },
     {
-      "epoch": 0.5098459256817994,
+      "epoch": 0.5267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.0167221216076679e-05,
-      "loss": 0.9018,
+      "learning_rate": 9.618675075381651e-06,
+      "loss": 0.9446,
       "step": 18564
     },
     {
-      "epoch": 0.5098733899096427,
+      "epoch": 0.5268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.0166331823779698e-05,
-      "loss": 0.9395,
+      "learning_rate": 9.61775666737784e-06,
+      "loss": 0.8359,
       "step": 18565
     },
     {
-      "epoch": 0.5099008541374859,
+      "epoch": 0.526844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.0165442430166639e-05,
-      "loss": 0.945,
+      "learning_rate": 9.61683826260285e-06,
+      "loss": 0.9503,
       "step": 18566
     },
     {
-      "epoch": 0.5099283183653291,
+      "epoch": 0.5268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.0164553035244528e-05,
-      "loss": 0.8759,
+      "learning_rate": 9.615919861064447e-06,
+      "loss": 0.9066,
       "step": 18567
     },
     {
-      "epoch": 0.5099557825931724,
+      "epoch": 0.5269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 1.0163663639020417e-05,
-      "loss": 0.9113,
+      "learning_rate": 9.615001462770387e-06,
+      "loss": 0.8576,
       "step": 18568
     },
     {
-      "epoch": 0.5099832468210156,
+      "epoch": 0.5269296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.016277424150133e-05,
-      "loss": 0.8194,
+      "learning_rate": 9.614083067728426e-06,
+      "loss": 0.8341,
       "step": 18569
     },
     {
-      "epoch": 0.5100107110488589,
+      "epoch": 0.5269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 1.0161884842694314e-05,
-      "loss": 1.0092,
+      "learning_rate": 9.613164675946323e-06,
+      "loss": 0.9213,
       "step": 18570
     },
     {
-      "epoch": 0.5100381752767021,
+      "epoch": 0.5269863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 1.0160995442606404e-05,
-      "loss": 0.9324,
+      "learning_rate": 9.612246287431832e-06,
+      "loss": 0.8843,
       "step": 18571
     },
     {
-      "epoch": 0.5100656395045453,
+      "epoch": 0.5270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.016010604124463e-05,
-      "loss": 0.8946,
+      "learning_rate": 9.611327902192718e-06,
+      "loss": 0.8939,
       "step": 18572
     },
     {
-      "epoch": 0.5100931037323886,
+      "epoch": 0.5270431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.0159216638616037e-05,
-      "loss": 0.9472,
+      "learning_rate": 9.610409520236735e-06,
+      "loss": 0.8524,
       "step": 18573
     },
     {
-      "epoch": 0.5101205679602318,
+      "epoch": 0.5270715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.0158327234727662e-05,
-      "loss": 0.934,
+      "learning_rate": 9.609491141571637e-06,
+      "loss": 0.8886,
       "step": 18574
     },
     {
-      "epoch": 0.510148032188075,
+      "epoch": 0.5270998864926221,
       "grad_norm": 0.0,
-      "learning_rate": 1.015743782958654e-05,
-      "loss": 1.0441,
+      "learning_rate": 9.608572766205189e-06,
+      "loss": 0.9706,
       "step": 18575
     },
     {
-      "epoch": 0.5101754964159183,
+      "epoch": 0.5271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.0156548423199708e-05,
-      "loss": 0.8252,
+      "learning_rate": 9.607654394145142e-06,
+      "loss": 0.9975,
       "step": 18576
     },
     {
-      "epoch": 0.5102029606437615,
+      "epoch": 0.5271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.0155659015574206e-05,
-      "loss": 0.8435,
+      "learning_rate": 9.606736025399256e-06,
+      "loss": 0.8194,
       "step": 18577
     },
     {
-      "epoch": 0.5102304248716047,
+      "epoch": 0.5271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.0154769606717069e-05,
-      "loss": 0.7979,
+      "learning_rate": 9.605817659975291e-06,
+      "loss": 0.697,
       "step": 18578
     },
     {
-      "epoch": 0.5102578890994479,
+      "epoch": 0.5272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.0153880196635332e-05,
-      "loss": 0.9652,
+      "learning_rate": 9.604899297881e-06,
+      "loss": 0.9135,
       "step": 18579
     },
     {
-      "epoch": 0.5102853533272912,
+      "epoch": 0.5272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.0152990785336042e-05,
-      "loss": 0.8949,
+      "learning_rate": 9.603980939124141e-06,
+      "loss": 0.8633,
       "step": 18580
     },
     {
-      "epoch": 0.5103128175551345,
+      "epoch": 0.5272701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 1.0152101372826222e-05,
-      "loss": 0.8971,
+      "learning_rate": 9.603062583712475e-06,
+      "loss": 0.9123,
       "step": 18581
     },
     {
-      "epoch": 0.5103402817829776,
+      "epoch": 0.5272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.0151211959112922e-05,
-      "loss": 0.7914,
+      "learning_rate": 9.602144231653758e-06,
+      "loss": 0.9114,
       "step": 18582
     },
     {
-      "epoch": 0.5103677460108209,
+      "epoch": 0.5273269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 1.0150322544203173e-05,
-      "loss": 0.8765,
+      "learning_rate": 9.601225882955742e-06,
+      "loss": 0.861,
       "step": 18583
     },
     {
-      "epoch": 0.5103952102386642,
+      "epoch": 0.527355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.0149433128104014e-05,
-      "loss": 0.8973,
+      "learning_rate": 9.600307537626194e-06,
+      "loss": 0.8607,
       "step": 18584
     },
     {
-      "epoch": 0.5104226744665074,
+      "epoch": 0.5273836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 1.014854371082248e-05,
-      "loss": 0.8696,
+      "learning_rate": 9.599389195672864e-06,
+      "loss": 0.8054,
       "step": 18585
     },
     {
-      "epoch": 0.5104501386943506,
+      "epoch": 0.5274120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 1.0147654292365613e-05,
-      "loss": 0.8475,
+      "learning_rate": 9.598470857103511e-06,
+      "loss": 0.9432,
       "step": 18586
     },
     {
-      "epoch": 0.5104776029221938,
+      "epoch": 0.5274404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.0146764872740449e-05,
-      "loss": 0.9632,
+      "learning_rate": 9.59755252192589e-06,
+      "loss": 0.8153,
       "step": 18587
     },
     {
-      "epoch": 0.5105050671500371,
+      "epoch": 0.5274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.0145875451954021e-05,
-      "loss": 0.9026,
+      "learning_rate": 9.596634190147765e-06,
+      "loss": 0.8846,
       "step": 18588
     },
     {
-      "epoch": 0.5105325313778803,
+      "epoch": 0.5274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.0144986030013373e-05,
-      "loss": 0.8401,
+      "learning_rate": 9.595715861776887e-06,
+      "loss": 0.7458,
       "step": 18589
     },
     {
-      "epoch": 0.5105599956057235,
+      "epoch": 0.5275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.014409660692554e-05,
-      "loss": 0.8756,
+      "learning_rate": 9.594797536821018e-06,
+      "loss": 0.8471,
       "step": 18590
     },
     {
-      "epoch": 0.5105874598335668,
+      "epoch": 0.5275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.0143207182697553e-05,
-      "loss": 0.8908,
+      "learning_rate": 9.59387921528791e-06,
+      "loss": 0.8804,
       "step": 18591
     },
     {
-      "epoch": 0.51061492406141,
+      "epoch": 0.5275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.0142317757336462e-05,
-      "loss": 0.9087,
+      "learning_rate": 9.592960897185324e-06,
+      "loss": 0.9037,
       "step": 18592
     },
     {
-      "epoch": 0.5106423882892532,
+      "epoch": 0.5276106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.0141428330849295e-05,
-      "loss": 0.9226,
+      "learning_rate": 9.592042582521016e-06,
+      "loss": 0.886,
       "step": 18593
     },
     {
-      "epoch": 0.5106698525170965,
+      "epoch": 0.527639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.0140538903243092e-05,
-      "loss": 0.8403,
+      "learning_rate": 9.591124271302741e-06,
+      "loss": 0.8953,
       "step": 18594
     },
     {
-      "epoch": 0.5106973167449397,
+      "epoch": 0.5276674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 1.013964947452489e-05,
-      "loss": 0.8653,
+      "learning_rate": 9.590205963538259e-06,
+      "loss": 0.8629,
       "step": 18595
     },
     {
-      "epoch": 0.510724780972783,
+      "epoch": 0.5276958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.013876004470173e-05,
-      "loss": 0.9012,
+      "learning_rate": 9.589287659235325e-06,
+      "loss": 0.89,
       "step": 18596
     },
     {
-      "epoch": 0.5107522452006262,
+      "epoch": 0.5277241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.0137870613780642e-05,
-      "loss": 0.9908,
+      "learning_rate": 9.588369358401696e-06,
+      "loss": 0.8188,
       "step": 18597
     },
     {
-      "epoch": 0.5107797094284694,
+      "epoch": 0.5277525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.0136981181768676e-05,
-      "loss": 0.9264,
+      "learning_rate": 9.587451061045131e-06,
+      "loss": 0.9151,
       "step": 18598
     },
     {
-      "epoch": 0.5108071736563127,
+      "epoch": 0.5277809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.0136091748672855e-05,
-      "loss": 0.9279,
+      "learning_rate": 9.586532767173385e-06,
+      "loss": 0.8776,
       "step": 18599
     },
     {
-      "epoch": 0.5108346378841558,
+      "epoch": 0.5278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.0135202314500226e-05,
-      "loss": 0.9484,
+      "learning_rate": 9.585614476794214e-06,
+      "loss": 0.8868,
       "step": 18600
     },
     {
-      "epoch": 0.5108621021119991,
+      "epoch": 0.5278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.0134312879257824e-05,
-      "loss": 0.7694,
+      "learning_rate": 9.584696189915379e-06,
+      "loss": 0.951,
       "step": 18601
     },
     {
-      "epoch": 0.5108895663398424,
+      "epoch": 0.5278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.0133423442952686e-05,
-      "loss": 0.8244,
+      "learning_rate": 9.583777906544628e-06,
+      "loss": 0.9617,
       "step": 18602
     },
     {
-      "epoch": 0.5109170305676856,
+      "epoch": 0.527894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.0132534005591852e-05,
-      "loss": 0.8348,
+      "learning_rate": 9.582859626689727e-06,
+      "loss": 0.8196,
       "step": 18603
     },
     {
-      "epoch": 0.5109444947955288,
+      "epoch": 0.5279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.0131644567182355e-05,
-      "loss": 0.87,
+      "learning_rate": 9.58194135035843e-06,
+      "loss": 0.9825,
       "step": 18604
     },
     {
-      "epoch": 0.5109719590233721,
+      "epoch": 0.5279511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 1.0130755127731233e-05,
-      "loss": 1.0229,
+      "learning_rate": 9.581023077558494e-06,
+      "loss": 0.7813,
       "step": 18605
     },
     {
-      "epoch": 0.5109994232512153,
+      "epoch": 0.5279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.012986568724553e-05,
-      "loss": 0.8956,
+      "learning_rate": 9.580104808297676e-06,
+      "loss": 0.9461,
       "step": 18606
     },
     {
-      "epoch": 0.5110268874790586,
+      "epoch": 0.5280079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.0128976245732278e-05,
-      "loss": 0.9034,
+      "learning_rate": 9.579186542583729e-06,
+      "loss": 0.8763,
       "step": 18607
     },
     {
-      "epoch": 0.5110543517069017,
+      "epoch": 0.5280363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.0128086803198515e-05,
-      "loss": 0.8134,
+      "learning_rate": 9.578268280424413e-06,
+      "loss": 0.9924,
       "step": 18608
     },
     {
-      "epoch": 0.511081815934745,
+      "epoch": 0.5280646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.0127197359651282e-05,
-      "loss": 0.8107,
+      "learning_rate": 9.577350021827486e-06,
+      "loss": 0.8908,
       "step": 18609
     },
     {
-      "epoch": 0.5111092801625883,
+      "epoch": 0.5280930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 1.0126307915097615e-05,
-      "loss": 1.0059,
+      "learning_rate": 9.576431766800697e-06,
+      "loss": 0.9817,
       "step": 18610
     },
     {
-      "epoch": 0.5111367443904314,
+      "epoch": 0.5281214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.0125418469544546e-05,
-      "loss": 0.9353,
+      "learning_rate": 9.575513515351813e-06,
+      "loss": 0.8917,
       "step": 18611
     },
     {
-      "epoch": 0.5111642086182747,
+      "epoch": 0.528149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.0124529022999122e-05,
-      "loss": 1.0255,
+      "learning_rate": 9.574595267488584e-06,
+      "loss": 0.8089,
       "step": 18612
     },
     {
-      "epoch": 0.5111916728461179,
+      "epoch": 0.528178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.0123639575468373e-05,
-      "loss": 0.9288,
+      "learning_rate": 9.573677023218765e-06,
+      "loss": 0.8338,
       "step": 18613
     },
     {
-      "epoch": 0.5112191370739612,
+      "epoch": 0.5282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.0122750126959341e-05,
-      "loss": 0.9683,
+      "learning_rate": 9.57275878255012e-06,
+      "loss": 0.915,
       "step": 18614
     },
     {
-      "epoch": 0.5112466013018044,
+      "epoch": 0.5282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.0121860677479065e-05,
-      "loss": 0.8546,
+      "learning_rate": 9.5718405454904e-06,
+      "loss": 0.9749,
       "step": 18615
     },
     {
-      "epoch": 0.5112740655296476,
+      "epoch": 0.5282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0120971227034575e-05,
-      "loss": 0.8434,
+      "learning_rate": 9.57092231204736e-06,
+      "loss": 0.8868,
       "step": 18616
     },
     {
-      "epoch": 0.5113015297574909,
+      "epoch": 0.5282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.0120081775632914e-05,
-      "loss": 0.8738,
+      "learning_rate": 9.570004082228757e-06,
+      "loss": 0.9218,
       "step": 18617
     },
     {
-      "epoch": 0.5113289939853342,
+      "epoch": 0.5283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.0119192323281123e-05,
-      "loss": 0.9229,
+      "learning_rate": 9.569085856042349e-06,
+      "loss": 0.9377,
       "step": 18618
     },
     {
-      "epoch": 0.5113564582131773,
+      "epoch": 0.5283484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 1.0118302869986239e-05,
-      "loss": 0.8545,
+      "learning_rate": 9.568167633495895e-06,
+      "loss": 0.897,
       "step": 18619
     },
     {
-      "epoch": 0.5113839224410206,
+      "epoch": 0.5283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.011741341575529e-05,
-      "loss": 0.9475,
+      "learning_rate": 9.567249414597148e-06,
+      "loss": 0.9045,
       "step": 18620
     },
     {
-      "epoch": 0.5114113866688638,
+      "epoch": 0.528405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.0116523960595325e-05,
-      "loss": 0.8032,
+      "learning_rate": 9.566331199353865e-06,
+      "loss": 0.9788,
       "step": 18621
     },
     {
-      "epoch": 0.511438850896707,
+      "epoch": 0.5284335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 1.0115634504513376e-05,
-      "loss": 0.8431,
+      "learning_rate": 9.565412987773802e-06,
+      "loss": 0.9243,
       "step": 18622
     },
     {
-      "epoch": 0.5114663151245503,
+      "epoch": 0.5284619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 1.0114745047516481e-05,
-      "loss": 0.8933,
+      "learning_rate": 9.564494779864713e-06,
+      "loss": 0.8755,
       "step": 18623
     },
     {
-      "epoch": 0.5114937793523935,
+      "epoch": 0.5284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.0113855589611681e-05,
-      "loss": 0.9185,
+      "learning_rate": 9.563576575634357e-06,
+      "loss": 0.8622,
       "step": 18624
     },
     {
-      "epoch": 0.5115212435802368,
+      "epoch": 0.5285187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.0112966130806007e-05,
-      "loss": 0.9451,
+      "learning_rate": 9.56265837509049e-06,
+      "loss": 0.9003,
       "step": 18625
     },
     {
-      "epoch": 0.5115487078080799,
+      "epoch": 0.5285471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.0112076671106505e-05,
-      "loss": 0.8404,
+      "learning_rate": 9.561740178240868e-06,
+      "loss": 0.7183,
       "step": 18626
     },
     {
-      "epoch": 0.5115761720359232,
+      "epoch": 0.5285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.011118721052021e-05,
-      "loss": 0.8459,
+      "learning_rate": 9.560821985093245e-06,
+      "loss": 0.8712,
       "step": 18627
     },
     {
-      "epoch": 0.5116036362637665,
+      "epoch": 0.5286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.0110297749054157e-05,
-      "loss": 0.8132,
+      "learning_rate": 9.559903795655379e-06,
+      "loss": 0.838,
       "step": 18628
     },
     {
-      "epoch": 0.5116311004916096,
+      "epoch": 0.5286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.0109408286715383e-05,
-      "loss": 0.8859,
+      "learning_rate": 9.558985609935027e-06,
+      "loss": 0.8824,
       "step": 18629
     },
     {
-      "epoch": 0.5116585647194529,
+      "epoch": 0.5286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.0108518823510931e-05,
-      "loss": 0.9293,
+      "learning_rate": 9.55806742793994e-06,
+      "loss": 0.9171,
       "step": 18630
     },
     {
-      "epoch": 0.5116860289472962,
+      "epoch": 0.528688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.0107629359447837e-05,
-      "loss": 0.8926,
+      "learning_rate": 9.55714924967788e-06,
+      "loss": 0.922,
       "step": 18631
     },
     {
-      "epoch": 0.5117134931751394,
+      "epoch": 0.5287173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 1.0106739894533135e-05,
-      "loss": 0.7854,
+      "learning_rate": 9.5562310751566e-06,
+      "loss": 0.8366,
       "step": 18632
     },
     {
-      "epoch": 0.5117409574029826,
+      "epoch": 0.5287457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.0105850428773869e-05,
-      "loss": 0.9493,
+      "learning_rate": 9.555312904383852e-06,
+      "loss": 0.9241,
       "step": 18633
     },
     {
-      "epoch": 0.5117684216308258,
+      "epoch": 0.5287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.0104960962177073e-05,
-      "loss": 0.8647,
+      "learning_rate": 9.554394737367398e-06,
+      "loss": 0.97,
       "step": 18634
     },
     {
-      "epoch": 0.5117958858586691,
+      "epoch": 0.5288024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.010407149474978e-05,
-      "loss": 0.9754,
+      "learning_rate": 9.553476574114993e-06,
+      "loss": 0.9435,
       "step": 18635
     },
     {
-      "epoch": 0.5118233500865124,
+      "epoch": 0.5288308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.0103182026499039e-05,
-      "loss": 0.8893,
+      "learning_rate": 9.552558414634392e-06,
+      "loss": 0.906,
       "step": 18636
     },
     {
-      "epoch": 0.5118508143143555,
+      "epoch": 0.5288592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 1.0102292557431879e-05,
-      "loss": 0.8381,
+      "learning_rate": 9.55164025893335e-06,
+      "loss": 0.9086,
       "step": 18637
     },
     {
-      "epoch": 0.5118782785421988,
+      "epoch": 0.5288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.0101403087555342e-05,
-      "loss": 0.9435,
+      "learning_rate": 9.550722107019621e-06,
+      "loss": 0.924,
       "step": 18638
     },
     {
-      "epoch": 0.511905742770042,
+      "epoch": 0.5289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.0100513616876463e-05,
-      "loss": 0.9172,
+      "learning_rate": 9.549803958900966e-06,
+      "loss": 0.8533,
       "step": 18639
     },
     {
-      "epoch": 0.5119332069978852,
+      "epoch": 0.52894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.0099624145402282e-05,
-      "loss": 0.9612,
+      "learning_rate": 9.548885814585134e-06,
+      "loss": 0.9297,
       "step": 18640
     },
     {
-      "epoch": 0.5119606712257285,
+      "epoch": 0.5289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.0098734673139836e-05,
-      "loss": 0.8951,
+      "learning_rate": 9.547967674079884e-06,
+      "loss": 0.8862,
       "step": 18641
     },
     {
-      "epoch": 0.5119881354535717,
+      "epoch": 0.5290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.0097845200096163e-05,
-      "loss": 0.9706,
+      "learning_rate": 9.547049537392973e-06,
+      "loss": 0.8773,
       "step": 18642
     },
     {
-      "epoch": 0.512015599681415,
+      "epoch": 0.5290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.0096955726278304e-05,
-      "loss": 0.8948,
+      "learning_rate": 9.546131404532154e-06,
+      "loss": 0.9032,
       "step": 18643
     },
     {
-      "epoch": 0.5120430639092582,
+      "epoch": 0.5290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.0096066251693289e-05,
-      "loss": 0.8728,
+      "learning_rate": 9.545213275505183e-06,
+      "loss": 0.773,
       "step": 18644
     },
     {
-      "epoch": 0.5120705281371014,
+      "epoch": 0.5290862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.0095176776348164e-05,
-      "loss": 0.916,
+      "learning_rate": 9.544295150319817e-06,
+      "loss": 0.9277,
       "step": 18645
     },
     {
-      "epoch": 0.5120979923649447,
+      "epoch": 0.5291146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 1.0094287300249961e-05,
-      "loss": 0.8013,
+      "learning_rate": 9.543377028983812e-06,
+      "loss": 0.7964,
       "step": 18646
     },
     {
-      "epoch": 0.5121254565927879,
+      "epoch": 0.5291430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.0093397823405721e-05,
-      "loss": 1.0045,
+      "learning_rate": 9.542458911504917e-06,
+      "loss": 0.87,
       "step": 18647
     },
     {
-      "epoch": 0.5121529208206311,
+      "epoch": 0.5291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.0092508345822482e-05,
-      "loss": 0.9513,
+      "learning_rate": 9.541540797890892e-06,
+      "loss": 0.9286,
       "step": 18648
     },
     {
-      "epoch": 0.5121803850484744,
+      "epoch": 0.529199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.009161886750728e-05,
-      "loss": 0.8548,
+      "learning_rate": 9.540622688149495e-06,
+      "loss": 1.0072,
       "step": 18649
     },
     {
-      "epoch": 0.5122078492763176,
+      "epoch": 0.529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.0090729388467156e-05,
-      "loss": 0.8943,
+      "learning_rate": 9.53970458228848e-06,
+      "loss": 0.8436,
       "step": 18650
     },
     {
-      "epoch": 0.5122353135041608,
+      "epoch": 0.5292565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.0089839908709144e-05,
-      "loss": 0.8769,
+      "learning_rate": 9.538786480315599e-06,
+      "loss": 0.8957,
       "step": 18651
     },
     {
-      "epoch": 0.512262777732004,
+      "epoch": 0.5292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.0088950428240284e-05,
-      "loss": 0.9264,
+      "learning_rate": 9.537868382238611e-06,
+      "loss": 0.968,
       "step": 18652
     },
     {
-      "epoch": 0.5122902419598473,
+      "epoch": 0.5293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.0088060947067613e-05,
-      "loss": 0.8627,
+      "learning_rate": 9.53695028806527e-06,
+      "loss": 0.9896,
       "step": 18653
     },
     {
-      "epoch": 0.5123177061876906,
+      "epoch": 0.5293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.0087171465198172e-05,
-      "loss": 0.9734,
+      "learning_rate": 9.536032197803329e-06,
+      "loss": 0.9742,
       "step": 18654
     },
     {
-      "epoch": 0.5123451704155337,
+      "epoch": 0.5293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.0086281982638997e-05,
-      "loss": 0.8346,
+      "learning_rate": 9.535114111460547e-06,
+      "loss": 0.8907,
       "step": 18655
     },
     {
-      "epoch": 0.512372634643377,
+      "epoch": 0.5293984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 1.0085392499397122e-05,
-      "loss": 0.8161,
+      "learning_rate": 9.534196029044677e-06,
+      "loss": 0.8742,
       "step": 18656
     },
     {
-      "epoch": 0.5124000988712203,
+      "epoch": 0.5294267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.0084503015479593e-05,
-      "loss": 0.942,
+      "learning_rate": 9.533277950563473e-06,
+      "loss": 0.7749,
       "step": 18657
     },
     {
-      "epoch": 0.5124275630990635,
+      "epoch": 0.5294551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.0083613530893441e-05,
-      "loss": 0.8873,
+      "learning_rate": 9.53235987602469e-06,
+      "loss": 0.8783,
       "step": 18658
     },
     {
-      "epoch": 0.5124550273269067,
+      "epoch": 0.5294835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 1.0082724045645704e-05,
-      "loss": 0.7651,
+      "learning_rate": 9.531441805436086e-06,
+      "loss": 0.9336,
       "step": 18659
     },
     {
-      "epoch": 0.5124824915547499,
+      "epoch": 0.5295119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.0081834559743426e-05,
-      "loss": 0.771,
+      "learning_rate": 9.530523738805412e-06,
+      "loss": 0.7603,
       "step": 18660
     },
     {
-      "epoch": 0.5125099557825932,
+      "epoch": 0.5295402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 1.0080945073193638e-05,
-      "loss": 0.8961,
+      "learning_rate": 9.529605676140426e-06,
+      "loss": 0.8298,
       "step": 18661
     },
     {
-      "epoch": 0.5125374200104365,
+      "epoch": 0.5295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.0080055586003385e-05,
-      "loss": 0.9374,
+      "learning_rate": 9.528687617448884e-06,
+      "loss": 0.8556,
       "step": 18662
     },
     {
-      "epoch": 0.5125648842382796,
+      "epoch": 0.5295970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.00791660981797e-05,
-      "loss": 0.8537,
+      "learning_rate": 9.527769562738533e-06,
+      "loss": 0.9199,
       "step": 18663
     },
     {
-      "epoch": 0.5125923484661229,
+      "epoch": 0.5296254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 1.007827660972962e-05,
-      "loss": 0.8789,
+      "learning_rate": 9.526851512017139e-06,
+      "loss": 0.8939,
       "step": 18664
     },
     {
-      "epoch": 0.5126198126939661,
+      "epoch": 0.5296538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.0077387120660188e-05,
-      "loss": 0.864,
+      "learning_rate": 9.525933465292449e-06,
+      "loss": 0.8511,
       "step": 18665
     },
     {
-      "epoch": 0.5126472769218093,
+      "epoch": 0.5296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.0076497630978437e-05,
-      "loss": 0.8659,
+      "learning_rate": 9.525015422572222e-06,
+      "loss": 0.9554,
       "step": 18666
     },
     {
-      "epoch": 0.5126747411496526,
+      "epoch": 0.5297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.007560814069141e-05,
-      "loss": 0.814,
+      "learning_rate": 9.524097383864211e-06,
+      "loss": 0.8786,
       "step": 18667
     },
     {
-      "epoch": 0.5127022053774958,
+      "epoch": 0.529738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.007471864980614e-05,
-      "loss": 0.8702,
+      "learning_rate": 9.52317934917617e-06,
+      "loss": 0.9705,
       "step": 18668
     },
     {
-      "epoch": 0.5127296696053391,
+      "epoch": 0.5297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.0073829158329664e-05,
-      "loss": 0.9069,
+      "learning_rate": 9.522261318515856e-06,
+      "loss": 0.8426,
       "step": 18669
     },
     {
-      "epoch": 0.5127571338331823,
+      "epoch": 0.5297956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 1.0072939666269028e-05,
-      "loss": 0.9163,
+      "learning_rate": 9.521343291891022e-06,
+      "loss": 0.9349,
       "step": 18670
     },
     {
-      "epoch": 0.5127845980610255,
+      "epoch": 0.5298240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.0072050173631262e-05,
-      "loss": 0.7829,
+      "learning_rate": 9.52042526930942e-06,
+      "loss": 0.8138,
       "step": 18671
     },
     {
-      "epoch": 0.5128120622888688,
+      "epoch": 0.5298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.0071160680423408e-05,
-      "loss": 0.9672,
+      "learning_rate": 9.51950725077881e-06,
+      "loss": 0.7675,
       "step": 18672
     },
     {
-      "epoch": 0.5128395265167119,
+      "epoch": 0.5298808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 1.0070271186652502e-05,
-      "loss": 0.8877,
+      "learning_rate": 9.518589236306945e-06,
+      "loss": 0.8356,
       "step": 18673
     },
     {
-      "epoch": 0.5128669907445552,
+      "epoch": 0.5299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.0069381692325587e-05,
-      "loss": 0.9119,
+      "learning_rate": 9.517671225901574e-06,
+      "loss": 0.8964,
       "step": 18674
     },
     {
-      "epoch": 0.5128944549723985,
+      "epoch": 0.5299375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.0068492197449696e-05,
-      "loss": 0.9022,
+      "learning_rate": 9.516753219570459e-06,
+      "loss": 0.8487,
       "step": 18675
     },
     {
-      "epoch": 0.5129219192002417,
+      "epoch": 0.5299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.0067602702031864e-05,
-      "loss": 0.9506,
+      "learning_rate": 9.515835217321351e-06,
+      "loss": 0.911,
       "step": 18676
     },
     {
-      "epoch": 0.5129493834280849,
+      "epoch": 0.5299943246311011,
       "grad_norm": 0.0,
-      "learning_rate": 1.0066713206079137e-05,
-      "loss": 1.0026,
+      "learning_rate": 9.514917219162004e-06,
+      "loss": 0.8476,
       "step": 18677
     },
     {
-      "epoch": 0.5129768476559281,
+      "epoch": 0.5300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.006582370959855e-05,
-      "loss": 0.9495,
+      "learning_rate": 9.513999225100173e-06,
+      "loss": 0.9198,
       "step": 18678
     },
     {
-      "epoch": 0.5130043118837714,
+      "epoch": 0.5300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.0064934212597137e-05,
-      "loss": 0.8498,
+      "learning_rate": 9.51308123514361e-06,
+      "loss": 0.8723,
       "step": 18679
     },
     {
-      "epoch": 0.5130317761116147,
+      "epoch": 0.5300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.0064044715081944e-05,
-      "loss": 0.8882,
+      "learning_rate": 9.512163249300074e-06,
+      "loss": 0.8384,
       "step": 18680
     },
     {
-      "epoch": 0.5130592403394578,
+      "epoch": 0.5301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.006315521706e-05,
-      "loss": 0.9066,
+      "learning_rate": 9.51124526757732e-06,
+      "loss": 0.9274,
       "step": 18681
     },
     {
-      "epoch": 0.5130867045673011,
+      "epoch": 0.5301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.0062265718538348e-05,
-      "loss": 0.8368,
+      "learning_rate": 9.510327289983095e-06,
+      "loss": 0.9294,
       "step": 18682
     },
     {
-      "epoch": 0.5131141687951444,
+      "epoch": 0.5301645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.0061376219524029e-05,
-      "loss": 0.8649,
+      "learning_rate": 9.509409316525159e-06,
+      "loss": 0.7343,
       "step": 18683
     },
     {
-      "epoch": 0.5131416330229875,
+      "epoch": 0.5301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.0060486720024073e-05,
-      "loss": 0.8656,
+      "learning_rate": 9.508491347211266e-06,
+      "loss": 0.8856,
       "step": 18684
     },
     {
-      "epoch": 0.5131690972508308,
+      "epoch": 0.5302213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.0059597220045525e-05,
-      "loss": 0.8129,
+      "learning_rate": 9.507573382049166e-06,
+      "loss": 0.763,
       "step": 18685
     },
     {
-      "epoch": 0.513196561478674,
+      "epoch": 0.530249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.0058707719595422e-05,
-      "loss": 0.8253,
+      "learning_rate": 9.506655421046617e-06,
+      "loss": 0.7407,
       "step": 18686
     },
     {
-      "epoch": 0.5132240257065173,
+      "epoch": 0.5302780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.00578182186808e-05,
-      "loss": 0.8202,
+      "learning_rate": 9.505737464211373e-06,
+      "loss": 0.9021,
       "step": 18687
     },
     {
-      "epoch": 0.5132514899343605,
+      "epoch": 0.5303064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 1.0056928717308694e-05,
-      "loss": 0.8864,
+      "learning_rate": 9.504819511551184e-06,
+      "loss": 0.8513,
       "step": 18688
     },
     {
-      "epoch": 0.5132789541622037,
+      "epoch": 0.5303348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.005603921548615e-05,
-      "loss": 0.9608,
+      "learning_rate": 9.50390156307381e-06,
+      "loss": 0.9028,
       "step": 18689
     },
     {
-      "epoch": 0.513306418390047,
+      "epoch": 0.5303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.0055149713220202e-05,
-      "loss": 0.8512,
+      "learning_rate": 9.502983618787001e-06,
+      "loss": 0.8755,
       "step": 18690
     },
     {
-      "epoch": 0.5133338826178901,
+      "epoch": 0.5303916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 1.0054260210517888e-05,
-      "loss": 1.0894,
+      "learning_rate": 9.50206567869851e-06,
+      "loss": 0.8987,
       "step": 18691
     },
     {
-      "epoch": 0.5133613468457334,
+      "epoch": 0.5304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.0053370707386244e-05,
-      "loss": 0.9415,
+      "learning_rate": 9.501147742816094e-06,
+      "loss": 0.8068,
       "step": 18692
     },
     {
-      "epoch": 0.5133888110735767,
+      "epoch": 0.5304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.0052481203832311e-05,
-      "loss": 0.83,
+      "learning_rate": 9.500229811147506e-06,
+      "loss": 0.9306,
       "step": 18693
     },
     {
-      "epoch": 0.5134162753014199,
+      "epoch": 0.5304767309875141,
       "grad_norm": 0.0,
-      "learning_rate": 1.0051591699863128e-05,
-      "loss": 0.8839,
+      "learning_rate": 9.499311883700494e-06,
+      "loss": 0.8975,
       "step": 18694
     },
     {
-      "epoch": 0.5134437395292631,
+      "epoch": 0.5305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.0050702195485733e-05,
-      "loss": 0.9839,
+      "learning_rate": 9.49839396048282e-06,
+      "loss": 0.8624,
       "step": 18695
     },
     {
-      "epoch": 0.5134712037571064,
+      "epoch": 0.530533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.0049812690707156e-05,
-      "loss": 0.8762,
+      "learning_rate": 9.497476041502237e-06,
+      "loss": 0.9981,
       "step": 18696
     },
     {
-      "epoch": 0.5134986679849496,
+      "epoch": 0.5305618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 1.0048923185534447e-05,
-      "loss": 0.8953,
+      "learning_rate": 9.496558126766497e-06,
+      "loss": 0.8748,
       "step": 18697
     },
     {
-      "epoch": 0.5135261322127929,
+      "epoch": 0.5305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.0048033679974638e-05,
-      "loss": 0.881,
+      "learning_rate": 9.495640216283352e-06,
+      "loss": 0.8572,
       "step": 18698
     },
     {
-      "epoch": 0.513553596440636,
+      "epoch": 0.5306186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.0047144174034768e-05,
-      "loss": 0.87,
+      "learning_rate": 9.494722310060555e-06,
+      "loss": 0.8924,
       "step": 18699
     },
     {
-      "epoch": 0.5135810606684793,
+      "epoch": 0.5306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.0046254667721873e-05,
-      "loss": 0.8602,
+      "learning_rate": 9.493804408105863e-06,
+      "loss": 0.9556,
       "step": 18700
     },
     {
-      "epoch": 0.5136085248963226,
+      "epoch": 0.5306753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.0045365161042998e-05,
-      "loss": 0.8805,
+      "learning_rate": 9.49288651042703e-06,
+      "loss": 0.9344,
       "step": 18701
     },
     {
-      "epoch": 0.5136359891241657,
+      "epoch": 0.5307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.0044475654005172e-05,
-      "loss": 0.91,
+      "learning_rate": 9.491968617031803e-06,
+      "loss": 0.8165,
       "step": 18702
     },
     {
-      "epoch": 0.513663453352009,
+      "epoch": 0.5307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.0043586146615438e-05,
-      "loss": 0.8379,
+      "learning_rate": 9.491050727927943e-06,
+      "loss": 0.93,
       "step": 18703
     },
     {
-      "epoch": 0.5136909175798522,
+      "epoch": 0.5307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.0042696638880833e-05,
-      "loss": 0.7903,
+      "learning_rate": 9.4901328431232e-06,
+      "loss": 0.9358,
       "step": 18704
     },
     {
-      "epoch": 0.5137183818076955,
+      "epoch": 0.530788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.0041807130808396e-05,
-      "loss": 0.9584,
+      "learning_rate": 9.489214962625327e-06,
+      "loss": 0.883,
       "step": 18705
     },
     {
-      "epoch": 0.5137458460355387,
+      "epoch": 0.5308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.0040917622405167e-05,
-      "loss": 0.9327,
+      "learning_rate": 9.48829708644208e-06,
+      "loss": 0.9268,
       "step": 18706
     },
     {
-      "epoch": 0.5137733102633819,
+      "epoch": 0.5308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.0040028113678181e-05,
-      "loss": 0.8549,
+      "learning_rate": 9.487379214581209e-06,
+      "loss": 0.7964,
       "step": 18707
     },
     {
-      "epoch": 0.5138007744912252,
+      "epoch": 0.5308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.0039138604634475e-05,
-      "loss": 0.9008,
+      "learning_rate": 9.486461347050466e-06,
+      "loss": 0.6948,
       "step": 18708
     },
     {
-      "epoch": 0.5138282387190685,
+      "epoch": 0.5309023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.003824909528109e-05,
-      "loss": 0.9262,
+      "learning_rate": 9.485543483857607e-06,
+      "loss": 0.863,
       "step": 18709
     },
     {
-      "epoch": 0.5138557029469116,
+      "epoch": 0.5309307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.0037359585625064e-05,
-      "loss": 0.979,
+      "learning_rate": 9.484625625010388e-06,
+      "loss": 0.8576,
       "step": 18710
     },
     {
-      "epoch": 0.5138831671747549,
+      "epoch": 0.5309591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.0036470075673436e-05,
-      "loss": 0.9063,
+      "learning_rate": 9.483707770516561e-06,
+      "loss": 0.7971,
       "step": 18711
     },
     {
-      "epoch": 0.5139106314025981,
+      "epoch": 0.5309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 1.003558056543324e-05,
-      "loss": 0.8697,
+      "learning_rate": 9.482789920383876e-06,
+      "loss": 0.9563,
       "step": 18712
     },
     {
-      "epoch": 0.5139380956304413,
+      "epoch": 0.5310158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.0034691054911519e-05,
-      "loss": 0.9457,
+      "learning_rate": 9.481872074620089e-06,
+      "loss": 0.9242,
       "step": 18713
     },
     {
-      "epoch": 0.5139655598582846,
+      "epoch": 0.531044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.0033801544115307e-05,
-      "loss": 0.8983,
+      "learning_rate": 9.480954233232952e-06,
+      "loss": 0.8511,
       "step": 18714
     },
     {
-      "epoch": 0.5139930240861278,
+      "epoch": 0.531072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.0032912033051645e-05,
-      "loss": 0.888,
+      "learning_rate": 9.480036396230216e-06,
+      "loss": 0.7632,
       "step": 18715
     },
     {
-      "epoch": 0.5140204883139711,
+      "epoch": 0.5311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.003202252172757e-05,
-      "loss": 0.8257,
+      "learning_rate": 9.479118563619638e-06,
+      "loss": 0.8846,
       "step": 18716
     },
     {
-      "epoch": 0.5140479525418142,
+      "epoch": 0.5311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.0031133010150121e-05,
-      "loss": 0.8864,
+      "learning_rate": 9.478200735408967e-06,
+      "loss": 0.8289,
       "step": 18717
     },
     {
-      "epoch": 0.5140754167696575,
+      "epoch": 0.5311577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 1.0030243498326336e-05,
-      "loss": 0.9062,
+      "learning_rate": 9.477282911605959e-06,
+      "loss": 0.832,
       "step": 18718
     },
     {
-      "epoch": 0.5141028809975008,
+      "epoch": 0.5311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.002935398626325e-05,
-      "loss": 0.8602,
+      "learning_rate": 9.476365092218365e-06,
+      "loss": 0.9509,
       "step": 18719
     },
     {
-      "epoch": 0.514130345225344,
+      "epoch": 0.5312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.0028464473967905e-05,
-      "loss": 0.843,
+      "learning_rate": 9.475447277253939e-06,
+      "loss": 0.8806,
       "step": 18720
     },
     {
-      "epoch": 0.5141578094531872,
+      "epoch": 0.5312429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 1.0027574961447338e-05,
-      "loss": 0.8654,
+      "learning_rate": 9.474529466720432e-06,
+      "loss": 1.0161,
       "step": 18721
     },
     {
-      "epoch": 0.5141852736810305,
+      "epoch": 0.5312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.0026685448708588e-05,
-      "loss": 0.8596,
+      "learning_rate": 9.4736116606256e-06,
+      "loss": 0.7302,
       "step": 18722
     },
     {
-      "epoch": 0.5142127379088737,
+      "epoch": 0.5312996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.0025795935758696e-05,
-      "loss": 0.7864,
+      "learning_rate": 9.472693858977195e-06,
+      "loss": 0.8754,
       "step": 18723
     },
     {
-      "epoch": 0.514240202136717,
+      "epoch": 0.5313280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 1.0024906422604693e-05,
-      "loss": 0.826,
+      "learning_rate": 9.471776061782968e-06,
+      "loss": 0.8988,
       "step": 18724
     },
     {
-      "epoch": 0.5142676663645601,
+      "epoch": 0.5313564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 1.0024016909253617e-05,
-      "loss": 0.8927,
+      "learning_rate": 9.470858269050669e-06,
+      "loss": 0.8251,
       "step": 18725
     },
     {
-      "epoch": 0.5142951305924034,
+      "epoch": 0.5313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.0023127395712515e-05,
-      "loss": 0.9923,
+      "learning_rate": 9.469940480788056e-06,
+      "loss": 0.8823,
       "step": 18726
     },
     {
-      "epoch": 0.5143225948202467,
+      "epoch": 0.5314131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.0022237881988421e-05,
-      "loss": 0.9352,
+      "learning_rate": 9.46902269700288e-06,
+      "loss": 0.9134,
       "step": 18727
     },
     {
-      "epoch": 0.5143500590480898,
+      "epoch": 0.5314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.0021348368088369e-05,
-      "loss": 0.9183,
+      "learning_rate": 9.468104917702894e-06,
+      "loss": 0.9697,
       "step": 18728
     },
     {
-      "epoch": 0.5143775232759331,
+      "epoch": 0.5314699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.0020458854019399e-05,
-      "loss": 0.8613,
+      "learning_rate": 9.467187142895848e-06,
+      "loss": 0.9284,
       "step": 18729
     },
     {
-      "epoch": 0.5144049875037763,
+      "epoch": 0.5314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.0019569339788556e-05,
-      "loss": 0.7481,
+      "learning_rate": 9.466269372589498e-06,
+      "loss": 0.9482,
       "step": 18730
     },
     {
-      "epoch": 0.5144324517316196,
+      "epoch": 0.5315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.0018679825402869e-05,
-      "loss": 0.7361,
+      "learning_rate": 9.465351606791595e-06,
+      "loss": 0.8927,
       "step": 18731
     },
     {
-      "epoch": 0.5144599159594628,
+      "epoch": 0.5315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.001779031086938e-05,
-      "loss": 0.9116,
+      "learning_rate": 9.464433845509888e-06,
+      "loss": 0.7615,
       "step": 18732
     },
     {
-      "epoch": 0.514487380187306,
+      "epoch": 0.531583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.0016900796195128e-05,
-      "loss": 0.877,
+      "learning_rate": 9.463516088752135e-06,
+      "loss": 0.9078,
       "step": 18733
     },
     {
-      "epoch": 0.5145148444151493,
+      "epoch": 0.5316118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.0016011281387151e-05,
-      "loss": 0.9146,
+      "learning_rate": 9.462598336526086e-06,
+      "loss": 0.9131,
       "step": 18734
     },
     {
-      "epoch": 0.5145423086429926,
+      "epoch": 0.5316401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.0015121766452485e-05,
-      "loss": 0.9791,
+      "learning_rate": 9.46168058883949e-06,
+      "loss": 0.8253,
       "step": 18735
     },
     {
-      "epoch": 0.5145697728708357,
+      "epoch": 0.5316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.001423225139817e-05,
-      "loss": 0.7772,
+      "learning_rate": 9.460762845700106e-06,
+      "loss": 0.8536,
       "step": 18736
     },
     {
-      "epoch": 0.514597237098679,
+      "epoch": 0.5316969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.0013342736231244e-05,
-      "loss": 0.9175,
+      "learning_rate": 9.459845107115682e-06,
+      "loss": 0.9332,
       "step": 18737
     },
     {
-      "epoch": 0.5146247013265222,
+      "epoch": 0.5317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.0012453220958747e-05,
-      "loss": 0.8776,
+      "learning_rate": 9.458927373093967e-06,
+      "loss": 0.8429,
       "step": 18738
     },
     {
-      "epoch": 0.5146521655543654,
+      "epoch": 0.5317536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 1.0011563705587715e-05,
-      "loss": 0.8253,
+      "learning_rate": 9.458009643642721e-06,
+      "loss": 0.877,
       "step": 18739
     },
     {
-      "epoch": 0.5146796297822087,
+      "epoch": 0.5317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.0010674190125184e-05,
-      "loss": 1.0053,
+      "learning_rate": 9.457091918769685e-06,
+      "loss": 0.9725,
       "step": 18740
     },
     {
-      "epoch": 0.5147070940100519,
+      "epoch": 0.5318104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.0009784674578193e-05,
-      "loss": 0.9166,
+      "learning_rate": 9.456174198482623e-06,
+      "loss": 0.8599,
       "step": 18741
     },
     {
-      "epoch": 0.5147345582378952,
+      "epoch": 0.5318388195232691,
       "grad_norm": 0.0,
-      "learning_rate": 1.0008895158953786e-05,
-      "loss": 0.979,
+      "learning_rate": 9.455256482789285e-06,
+      "loss": 0.8795,
       "step": 18742
     },
     {
-      "epoch": 0.5147620224657383,
+      "epoch": 0.5318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.0008005643258997e-05,
-      "loss": 1.0109,
+      "learning_rate": 9.454338771697414e-06,
+      "loss": 0.8784,
       "step": 18743
     },
     {
-      "epoch": 0.5147894866935816,
+      "epoch": 0.5318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.0007116127500862e-05,
-      "loss": 1.0033,
+      "learning_rate": 9.453421065214773e-06,
+      "loss": 0.7863,
       "step": 18744
     },
     {
-      "epoch": 0.5148169509214249,
+      "epoch": 0.5319239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 1.0006226611686425e-05,
-      "loss": 0.907,
+      "learning_rate": 9.452503363349107e-06,
+      "loss": 0.8207,
       "step": 18745
     },
     {
-      "epoch": 0.514844415149268,
+      "epoch": 0.5319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.0005337095822716e-05,
-      "loss": 0.9256,
+      "learning_rate": 9.451585666108168e-06,
+      "loss": 0.8234,
       "step": 18746
     },
     {
-      "epoch": 0.5148718793771113,
+      "epoch": 0.5319807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.0004447579916781e-05,
-      "loss": 0.8923,
+      "learning_rate": 9.450667973499712e-06,
+      "loss": 0.8861,
       "step": 18747
     },
     {
-      "epoch": 0.5148993436049546,
+      "epoch": 0.5320090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 1.0003558063975655e-05,
-      "loss": 0.8809,
+      "learning_rate": 9.449750285531488e-06,
+      "loss": 0.9239,
       "step": 18748
     },
     {
-      "epoch": 0.5149268078327978,
+      "epoch": 0.5320374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.0002668548006375e-05,
-      "loss": 1.0017,
+      "learning_rate": 9.448832602211246e-06,
+      "loss": 0.7606,
       "step": 18749
     },
     {
-      "epoch": 0.514954272060641,
+      "epoch": 0.5320658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.000177903201598e-05,
-      "loss": 0.8491,
+      "learning_rate": 9.447914923546741e-06,
+      "loss": 0.9082,
       "step": 18750
     },
     {
-      "epoch": 0.5149817362884842,
+      "epoch": 0.532094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.0000889516011512e-05,
-      "loss": 0.8701,
+      "learning_rate": 9.446997249545725e-06,
+      "loss": 0.9042,
       "step": 18751
     },
     {
-      "epoch": 0.5150092005163275,
+      "epoch": 0.532122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1e-05,
-      "loss": 0.9933,
+      "learning_rate": 9.446079580215946e-06,
+      "loss": 0.9689,
       "step": 18752
     },
     {
-      "epoch": 0.5150366647441708,
+      "epoch": 0.5321509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 9.999110483988491e-06,
-      "loss": 0.8959,
+      "learning_rate": 9.445161915565159e-06,
+      "loss": 0.9357,
       "step": 18753
     },
     {
-      "epoch": 0.5150641289720139,
+      "epoch": 0.5321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 9.99822096798402e-06,
-      "loss": 0.9578,
+      "learning_rate": 9.444244255601115e-06,
+      "loss": 0.774,
       "step": 18754
     },
     {
-      "epoch": 0.5150915931998572,
+      "epoch": 0.5322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 9.99733145199363e-06,
-      "loss": 0.8162,
+      "learning_rate": 9.443326600331562e-06,
+      "loss": 0.8575,
       "step": 18755
     },
     {
-      "epoch": 0.5151190574277004,
+      "epoch": 0.5322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 9.99644193602435e-06,
-      "loss": 0.8285,
+      "learning_rate": 9.44240894976425e-06,
+      "loss": 0.8917,
       "step": 18756
     },
     {
-      "epoch": 0.5151465216555436,
+      "epoch": 0.5322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 9.99555242008322e-06,
-      "loss": 0.9097,
+      "learning_rate": 9.441491303906942e-06,
+      "loss": 0.9226,
       "step": 18757
     },
     {
-      "epoch": 0.5151739858833869,
+      "epoch": 0.5322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 9.994662904177285e-06,
-      "loss": 1.0172,
+      "learning_rate": 9.440573662767382e-06,
+      "loss": 0.8099,
       "step": 18758
     },
     {
-      "epoch": 0.5152014501112301,
+      "epoch": 0.5323212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 9.993773388313578e-06,
-      "loss": 0.9265,
+      "learning_rate": 9.439656026353318e-06,
+      "loss": 0.8332,
       "step": 18759
     },
     {
-      "epoch": 0.5152289143390734,
+      "epoch": 0.5323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 9.992883872499137e-06,
-      "loss": 0.8806,
+      "learning_rate": 9.438738394672507e-06,
+      "loss": 0.9683,
       "step": 18760
     },
     {
-      "epoch": 0.5152563785669166,
+      "epoch": 0.5323779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 9.991994356741008e-06,
-      "loss": 0.9003,
+      "learning_rate": 9.437820767732699e-06,
+      "loss": 0.7871,
       "step": 18761
     },
     {
-      "epoch": 0.5152838427947598,
+      "epoch": 0.5324063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 9.991104841046216e-06,
-      "loss": 0.8775,
+      "learning_rate": 9.436903145541642e-06,
+      "loss": 0.9181,
       "step": 18762
     },
     {
-      "epoch": 0.5153113070226031,
+      "epoch": 0.5324347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 9.990215325421809e-06,
-      "loss": 0.9421,
+      "learning_rate": 9.43598552810709e-06,
+      "loss": 0.8481,
       "step": 18763
     },
     {
-      "epoch": 0.5153387712504462,
+      "epoch": 0.5324631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 9.98932580987482e-06,
-      "loss": 0.8745,
+      "learning_rate": 9.435067915436796e-06,
+      "loss": 0.8082,
       "step": 18764
     },
     {
-      "epoch": 0.5153662354782895,
+      "epoch": 0.5324914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 9.988436294412289e-06,
-      "loss": 0.8753,
+      "learning_rate": 9.434150307538507e-06,
+      "loss": 0.9184,
       "step": 18765
     },
     {
-      "epoch": 0.5153936997061328,
+      "epoch": 0.5325198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 9.987546779041256e-06,
-      "loss": 0.8956,
+      "learning_rate": 9.433232704419974e-06,
+      "loss": 0.7525,
       "step": 18766
     },
     {
-      "epoch": 0.515421163933976,
+      "epoch": 0.5325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 9.986657263768761e-06,
-      "loss": 0.8467,
+      "learning_rate": 9.432315106088951e-06,
+      "loss": 0.8082,
       "step": 18767
     },
     {
-      "epoch": 0.5154486281618192,
+      "epoch": 0.5325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 9.985767748601833e-06,
-      "loss": 0.8729,
+      "learning_rate": 9.431397512553188e-06,
+      "loss": 0.8714,
       "step": 18768
     },
     {
-      "epoch": 0.5154760923896624,
+      "epoch": 0.5326049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 9.984878233547519e-06,
-      "loss": 0.9449,
+      "learning_rate": 9.430479923820434e-06,
+      "loss": 0.9151,
       "step": 18769
     },
     {
-      "epoch": 0.5155035566175057,
+      "epoch": 0.532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 9.983988718612854e-06,
-      "loss": 0.9011,
+      "learning_rate": 9.429562339898446e-06,
+      "loss": 0.8275,
       "step": 18770
     },
     {
-      "epoch": 0.515531020845349,
+      "epoch": 0.5326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 9.983099203804875e-06,
-      "loss": 0.919,
+      "learning_rate": 9.428644760794965e-06,
+      "loss": 0.7993,
       "step": 18771
     },
     {
-      "epoch": 0.5155584850731921,
+      "epoch": 0.5326901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 9.98220968913062e-06,
-      "loss": 0.909,
+      "learning_rate": 9.42772718651775e-06,
+      "loss": 0.9077,
       "step": 18772
     },
     {
-      "epoch": 0.5155859493010354,
+      "epoch": 0.5327185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 9.981320174597136e-06,
-      "loss": 0.8473,
+      "learning_rate": 9.426809617074548e-06,
+      "loss": 0.8345,
       "step": 18773
     },
     {
-      "epoch": 0.5156134135288787,
+      "epoch": 0.5327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 9.980430660211449e-06,
-      "loss": 0.8111,
+      "learning_rate": 9.425892052473114e-06,
+      "loss": 0.8527,
       "step": 18774
     },
     {
-      "epoch": 0.5156408777567218,
+      "epoch": 0.5327752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 9.979541145980603e-06,
-      "loss": 0.9858,
+      "learning_rate": 9.424974492721194e-06,
+      "loss": 0.915,
       "step": 18775
     },
     {
-      "epoch": 0.5156683419845651,
+      "epoch": 0.5328036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 9.978651631911635e-06,
-      "loss": 0.9963,
+      "learning_rate": 9.424056937826539e-06,
+      "loss": 0.8425,
       "step": 18776
     },
     {
-      "epoch": 0.5156958062124083,
+      "epoch": 0.5328320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 9.977762118011582e-06,
-      "loss": 0.8831,
+      "learning_rate": 9.423139387796904e-06,
+      "loss": 0.7743,
       "step": 18777
     },
     {
-      "epoch": 0.5157232704402516,
+      "epoch": 0.5328603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 9.976872604287487e-06,
-      "loss": 0.9156,
+      "learning_rate": 9.422221842640035e-06,
+      "loss": 0.8466,
       "step": 18778
     },
     {
-      "epoch": 0.5157507346680948,
+      "epoch": 0.53288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 9.975983090746386e-06,
-      "loss": 0.9006,
+      "learning_rate": 9.421304302363683e-06,
+      "loss": 0.9115,
       "step": 18779
     },
     {
-      "epoch": 0.515778198895938,
+      "epoch": 0.5329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 9.975093577395312e-06,
-      "loss": 0.9764,
+      "learning_rate": 9.420386766975602e-06,
+      "loss": 0.8299,
       "step": 18780
     },
     {
-      "epoch": 0.5158056631237813,
+      "epoch": 0.5329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 9.97420406424131e-06,
-      "loss": 0.9674,
+      "learning_rate": 9.419469236483539e-06,
+      "loss": 0.7765,
       "step": 18781
     },
     {
-      "epoch": 0.5158331273516245,
+      "epoch": 0.5329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 9.973314551291414e-06,
-      "loss": 0.8582,
+      "learning_rate": 9.418551710895243e-06,
+      "loss": 0.8546,
       "step": 18782
     },
     {
-      "epoch": 0.5158605915794677,
+      "epoch": 0.5330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 9.972425038552662e-06,
-      "loss": 0.8712,
+      "learning_rate": 9.41763419021847e-06,
+      "loss": 0.918,
       "step": 18783
     },
     {
-      "epoch": 0.515888055807311,
+      "epoch": 0.5330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 9.971535526032094e-06,
-      "loss": 0.9774,
+      "learning_rate": 9.416716674460966e-06,
+      "loss": 0.9901,
       "step": 18784
     },
     {
-      "epoch": 0.5159155200351542,
+      "epoch": 0.5330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 9.970646013736753e-06,
-      "loss": 0.8764,
+      "learning_rate": 9.415799163630482e-06,
+      "loss": 0.9661,
       "step": 18785
     },
     {
-      "epoch": 0.5159429842629975,
+      "epoch": 0.5330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 9.96975650167367e-06,
-      "loss": 0.8478,
+      "learning_rate": 9.414881657734768e-06,
+      "loss": 0.871,
       "step": 18786
     },
     {
-      "epoch": 0.5159704484908407,
+      "epoch": 0.5331157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 9.968866989849884e-06,
-      "loss": 0.9064,
+      "learning_rate": 9.413964156781575e-06,
+      "loss": 0.8559,
       "step": 18787
     },
     {
-      "epoch": 0.5159979127186839,
+      "epoch": 0.533144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 9.967977478272433e-06,
-      "loss": 0.9544,
+      "learning_rate": 9.413046660778654e-06,
+      "loss": 0.8922,
       "step": 18788
     },
     {
-      "epoch": 0.5160253769465272,
+      "epoch": 0.5331725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 9.967087966948356e-06,
-      "loss": 0.8871,
+      "learning_rate": 9.412129169733756e-06,
+      "loss": 0.931,
       "step": 18789
     },
     {
-      "epoch": 0.5160528411743703,
+      "epoch": 0.5332009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 9.966198455884695e-06,
-      "loss": 0.9762,
+      "learning_rate": 9.411211683654626e-06,
+      "loss": 0.848,
       "step": 18790
     },
     {
-      "epoch": 0.5160803054022136,
+      "epoch": 0.5332292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 9.965308945088486e-06,
-      "loss": 0.8023,
+      "learning_rate": 9.410294202549019e-06,
+      "loss": 0.8336,
       "step": 18791
     },
     {
-      "epoch": 0.5161077696300569,
+      "epoch": 0.5332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 9.964419434566763e-06,
-      "loss": 0.903,
+      "learning_rate": 9.409376726424686e-06,
+      "loss": 0.7962,
       "step": 18792
     },
     {
-      "epoch": 0.5161352338579,
+      "epoch": 0.5332860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 9.963529924326567e-06,
-      "loss": 0.9314,
+      "learning_rate": 9.40845925528937e-06,
+      "loss": 0.9471,
       "step": 18793
     },
     {
-      "epoch": 0.5161626980857433,
+      "epoch": 0.5333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 9.96264041437494e-06,
-      "loss": 0.9098,
+      "learning_rate": 9.407541789150829e-06,
+      "loss": 0.8018,
       "step": 18794
     },
     {
-      "epoch": 0.5161901623135865,
+      "epoch": 0.5333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 9.961750904718911e-06,
-      "loss": 0.8339,
+      "learning_rate": 9.406624328016807e-06,
+      "loss": 0.8757,
       "step": 18795
     },
     {
-      "epoch": 0.5162176265414298,
+      "epoch": 0.5333711691259931,
       "grad_norm": 0.0,
-      "learning_rate": 9.960861395365525e-06,
-      "loss": 0.9079,
+      "learning_rate": 9.405706871895055e-06,
+      "loss": 0.9409,
       "step": 18796
     },
     {
-      "epoch": 0.516245090769273,
+      "epoch": 0.5333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 9.959971886321824e-06,
-      "loss": 0.8575,
+      "learning_rate": 9.404789420793327e-06,
+      "loss": 0.9562,
       "step": 18797
     },
     {
-      "epoch": 0.5162725549971162,
+      "epoch": 0.533427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 9.959082377594835e-06,
-      "loss": 0.8838,
+      "learning_rate": 9.403871974719368e-06,
+      "loss": 0.8625,
       "step": 18798
     },
     {
-      "epoch": 0.5163000192249595,
+      "epoch": 0.5334562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 9.958192869191606e-06,
-      "loss": 0.9509,
+      "learning_rate": 9.402954533680927e-06,
+      "loss": 0.8089,
       "step": 18799
     },
     {
-      "epoch": 0.5163274834528028,
+      "epoch": 0.5334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 9.957303361119168e-06,
-      "loss": 0.8399,
+      "learning_rate": 9.402037097685759e-06,
+      "loss": 0.8371,
       "step": 18800
     },
     {
-      "epoch": 0.5163549476806459,
+      "epoch": 0.5335130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 9.956413853384563e-06,
-      "loss": 0.8731,
+      "learning_rate": 9.40111966674161e-06,
+      "loss": 0.8194,
       "step": 18801
     },
     {
-      "epoch": 0.5163824119084892,
+      "epoch": 0.5335414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 9.95552434599483e-06,
-      "loss": 0.9075,
+      "learning_rate": 9.400202240856225e-06,
+      "loss": 0.9808,
       "step": 18802
     },
     {
-      "epoch": 0.5164098761363324,
+      "epoch": 0.5335698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 9.954634838957009e-06,
-      "loss": 0.8576,
+      "learning_rate": 9.399284820037361e-06,
+      "loss": 0.945,
       "step": 18803
     },
     {
-      "epoch": 0.5164373403641757,
+      "epoch": 0.5335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 9.953745332278129e-06,
-      "loss": 0.9247,
+      "learning_rate": 9.398367404292769e-06,
+      "loss": 0.8474,
       "step": 18804
     },
     {
-      "epoch": 0.5164648045920189,
+      "epoch": 0.5336265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 9.952855825965235e-06,
-      "loss": 0.9032,
+      "learning_rate": 9.397449993630193e-06,
+      "loss": 0.8386,
       "step": 18805
     },
     {
-      "epoch": 0.5164922688198621,
+      "epoch": 0.5336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 9.951966320025364e-06,
-      "loss": 1.0318,
+      "learning_rate": 9.396532588057384e-06,
+      "loss": 0.9619,
       "step": 18806
     },
     {
-      "epoch": 0.5165197330477054,
+      "epoch": 0.533683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.951076814465555e-06,
-      "loss": 0.8457,
+      "learning_rate": 9.39561518758209e-06,
+      "loss": 0.8281,
       "step": 18807
     },
     {
-      "epoch": 0.5165471972755487,
+      "epoch": 0.5337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 9.950187309292842e-06,
-      "loss": 0.8688,
+      "learning_rate": 9.394697792212064e-06,
+      "loss": 0.7737,
       "step": 18808
     },
     {
-      "epoch": 0.5165746615033918,
+      "epoch": 0.5337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 9.949297804514273e-06,
-      "loss": 0.9716,
+      "learning_rate": 9.393780401955053e-06,
+      "loss": 0.9036,
       "step": 18809
     },
     {
-      "epoch": 0.5166021257312351,
+      "epoch": 0.5337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 9.948408300136875e-06,
-      "loss": 0.8612,
+      "learning_rate": 9.392863016818803e-06,
+      "loss": 0.8912,
       "step": 18810
     },
     {
-      "epoch": 0.5166295899590783,
+      "epoch": 0.5337968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 9.94751879616769e-06,
-      "loss": 0.8519,
+      "learning_rate": 9.391945636811068e-06,
+      "loss": 0.8642,
       "step": 18811
     },
     {
-      "epoch": 0.5166570541869215,
+      "epoch": 0.5338251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 9.946629292613757e-06,
-      "loss": 0.9067,
+      "learning_rate": 9.391028261939598e-06,
+      "loss": 0.9692,
       "step": 18812
     },
     {
-      "epoch": 0.5166845184147648,
+      "epoch": 0.5338535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 9.945739789482114e-06,
-      "loss": 0.9349,
+      "learning_rate": 9.390110892212135e-06,
+      "loss": 0.8824,
       "step": 18813
     },
     {
-      "epoch": 0.516711982642608,
+      "epoch": 0.5338819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 9.944850286779803e-06,
-      "loss": 0.9917,
+      "learning_rate": 9.389193527636436e-06,
+      "loss": 0.778,
       "step": 18814
     },
     {
-      "epoch": 0.5167394468704513,
+      "epoch": 0.5339103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 9.943960784513854e-06,
-      "loss": 0.8934,
+      "learning_rate": 9.388276168220248e-06,
+      "loss": 0.9566,
       "step": 18815
     },
     {
-      "epoch": 0.5167669110982944,
+      "epoch": 0.533938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 9.943071282691307e-06,
-      "loss": 0.8655,
+      "learning_rate": 9.387358813971314e-06,
+      "loss": 0.955,
       "step": 18816
     },
     {
-      "epoch": 0.5167943753261377,
+      "epoch": 0.533967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 9.942181781319204e-06,
-      "loss": 0.9054,
+      "learning_rate": 9.386441464897387e-06,
+      "loss": 0.9772,
       "step": 18817
     },
     {
-      "epoch": 0.516821839553981,
+      "epoch": 0.5339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 9.941292280404583e-06,
-      "loss": 0.9437,
+      "learning_rate": 9.385524121006221e-06,
+      "loss": 0.8459,
       "step": 18818
     },
     {
-      "epoch": 0.5168493037818241,
+      "epoch": 0.5340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 9.940402779954476e-06,
-      "loss": 0.8682,
+      "learning_rate": 9.38460678230556e-06,
+      "loss": 0.8191,
       "step": 18819
     },
     {
-      "epoch": 0.5168767680096674,
+      "epoch": 0.5340522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 9.93951327997593e-06,
-      "loss": 0.8665,
+      "learning_rate": 9.383689448803151e-06,
+      "loss": 0.8171,
       "step": 18820
     },
     {
-      "epoch": 0.5169042322375107,
+      "epoch": 0.5340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 9.938623780475976e-06,
-      "loss": 0.9828,
+      "learning_rate": 9.382772120506747e-06,
+      "loss": 0.7345,
       "step": 18821
     },
     {
-      "epoch": 0.5169316964653539,
+      "epoch": 0.5341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 9.937734281461653e-06,
-      "loss": 0.8617,
+      "learning_rate": 9.381854797424094e-06,
+      "loss": 0.7916,
       "step": 18822
     },
     {
-      "epoch": 0.5169591606931971,
+      "epoch": 0.5341373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 9.936844782940002e-06,
-      "loss": 0.9727,
+      "learning_rate": 9.380937479562941e-06,
+      "loss": 0.8869,
       "step": 18823
     },
     {
-      "epoch": 0.5169866249210403,
+      "epoch": 0.5341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 9.93595528491806e-06,
-      "loss": 0.8775,
+      "learning_rate": 9.380020166931037e-06,
+      "loss": 0.8544,
       "step": 18824
     },
     {
-      "epoch": 0.5170140891488836,
+      "epoch": 0.5341940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 9.935065787402863e-06,
-      "loss": 0.9369,
+      "learning_rate": 9.379102859536131e-06,
+      "loss": 0.8158,
       "step": 18825
     },
     {
-      "epoch": 0.5170415533767269,
+      "epoch": 0.5342224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 9.934176290401455e-06,
-      "loss": 0.9736,
+      "learning_rate": 9.37818555738597e-06,
+      "loss": 0.9271,
       "step": 18826
     },
     {
-      "epoch": 0.51706901760457,
+      "epoch": 0.5342508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 9.933286793920866e-06,
-      "loss": 0.876,
+      "learning_rate": 9.377268260488305e-06,
+      "loss": 0.7924,
       "step": 18827
     },
     {
-      "epoch": 0.5170964818324133,
+      "epoch": 0.5342792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 9.932397297968139e-06,
-      "loss": 0.7874,
+      "learning_rate": 9.376350968850883e-06,
+      "loss": 0.8823,
       "step": 18828
     },
     {
-      "epoch": 0.5171239460602565,
+      "epoch": 0.5343076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 9.931507802550308e-06,
-      "loss": 0.8778,
+      "learning_rate": 9.375433682481451e-06,
+      "loss": 0.7688,
       "step": 18829
     },
     {
-      "epoch": 0.5171514102880997,
+      "epoch": 0.5343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 9.930618307674417e-06,
-      "loss": 0.8175,
+      "learning_rate": 9.37451640138776e-06,
+      "loss": 0.8975,
       "step": 18830
     },
     {
-      "epoch": 0.517178874515943,
+      "epoch": 0.5343643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 9.929728813347498e-06,
-      "loss": 0.9572,
+      "learning_rate": 9.373599125577557e-06,
+      "loss": 0.8086,
       "step": 18831
     },
     {
-      "epoch": 0.5172063387437862,
+      "epoch": 0.5343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 9.928839319576597e-06,
-      "loss": 0.9752,
+      "learning_rate": 9.372681855058587e-06,
+      "loss": 0.8561,
       "step": 18832
     },
     {
-      "epoch": 0.5172338029716295,
+      "epoch": 0.5344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.927949826368741e-06,
-      "loss": 0.9057,
+      "learning_rate": 9.371764589838605e-06,
+      "loss": 0.8992,
       "step": 18833
     },
     {
-      "epoch": 0.5172612671994727,
+      "epoch": 0.5344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 9.927060333730975e-06,
-      "loss": 0.8449,
+      "learning_rate": 9.370847329925354e-06,
+      "loss": 0.7994,
       "step": 18834
     },
     {
-      "epoch": 0.5172887314273159,
+      "epoch": 0.534477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 9.926170841670339e-06,
-      "loss": 0.8785,
+      "learning_rate": 9.369930075326586e-06,
+      "loss": 0.9971,
       "step": 18835
     },
     {
-      "epoch": 0.5173161956551592,
+      "epoch": 0.5345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 9.925281350193864e-06,
-      "loss": 0.8224,
+      "learning_rate": 9.369012826050046e-06,
+      "loss": 0.9794,
       "step": 18836
     },
     {
-      "epoch": 0.5173436598830023,
+      "epoch": 0.5345346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 9.924391859308592e-06,
-      "loss": 0.8833,
+      "learning_rate": 9.368095582103482e-06,
+      "loss": 0.9159,
       "step": 18837
     },
     {
-      "epoch": 0.5173711241108456,
+      "epoch": 0.5345629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 9.923502369021567e-06,
-      "loss": 0.9197,
+      "learning_rate": 9.367178343494644e-06,
+      "loss": 0.9811,
       "step": 18838
     },
     {
-      "epoch": 0.5173985883386889,
+      "epoch": 0.5345913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 9.922612879339817e-06,
-      "loss": 0.8964,
+      "learning_rate": 9.36626111023128e-06,
+      "loss": 0.867,
       "step": 18839
     },
     {
-      "epoch": 0.5174260525665321,
+      "epoch": 0.5346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 9.921723390270383e-06,
-      "loss": 0.7841,
+      "learning_rate": 9.365343882321134e-06,
+      "loss": 0.771,
       "step": 18840
     },
     {
-      "epoch": 0.5174535167943753,
+      "epoch": 0.5346481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 9.920833901820302e-06,
-      "loss": 0.9715,
+      "learning_rate": 9.36442665977196e-06,
+      "loss": 0.8448,
       "step": 18841
     },
     {
-      "epoch": 0.5174809810222185,
+      "epoch": 0.5346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 9.919944413996618e-06,
-      "loss": 0.7802,
+      "learning_rate": 9.363509442591502e-06,
+      "loss": 0.7228,
       "step": 18842
     },
     {
-      "epoch": 0.5175084452500618,
+      "epoch": 0.5347048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 9.919054926806362e-06,
-      "loss": 0.805,
+      "learning_rate": 9.362592230787507e-06,
+      "loss": 0.8186,
       "step": 18843
     },
     {
-      "epoch": 0.5175359094779051,
+      "epoch": 0.5347332576617481,
       "grad_norm": 0.0,
-      "learning_rate": 9.91816544025658e-06,
-      "loss": 0.928,
+      "learning_rate": 9.361675024367724e-06,
+      "loss": 0.8709,
       "step": 18844
     },
     {
-      "epoch": 0.5175633737057482,
+      "epoch": 0.5347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 9.9172759543543e-06,
-      "loss": 0.8823,
+      "learning_rate": 9.360757823339904e-06,
+      "loss": 0.8866,
       "step": 18845
     },
     {
-      "epoch": 0.5175908379335915,
+      "epoch": 0.5347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 9.916386469106562e-06,
-      "loss": 0.8382,
+      "learning_rate": 9.359840627711788e-06,
+      "loss": 0.8922,
       "step": 18846
     },
     {
-      "epoch": 0.5176183021614348,
+      "epoch": 0.5348183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 9.91549698452041e-06,
-      "loss": 0.8601,
+      "learning_rate": 9.358923437491127e-06,
+      "loss": 0.8518,
       "step": 18847
     },
     {
-      "epoch": 0.517645766389278,
+      "epoch": 0.5348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 9.914607500602878e-06,
-      "loss": 0.8856,
+      "learning_rate": 9.358006252685667e-06,
+      "loss": 0.8796,
       "step": 18848
     },
     {
-      "epoch": 0.5176732306171212,
+      "epoch": 0.5348751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 9.913718017361003e-06,
-      "loss": 0.949,
+      "learning_rate": 9.35708907330316e-06,
+      "loss": 0.8169,
       "step": 18849
     },
     {
-      "epoch": 0.5177006948449644,
+      "epoch": 0.5349035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 9.91282853480183e-06,
-      "loss": 0.8341,
+      "learning_rate": 9.356171899351351e-06,
+      "loss": 0.8351,
       "step": 18850
     },
     {
-      "epoch": 0.5177281590728077,
+      "epoch": 0.5349318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 9.91193905293239e-06,
-      "loss": 0.8644,
+      "learning_rate": 9.355254730837983e-06,
+      "loss": 0.8573,
       "step": 18851
     },
     {
-      "epoch": 0.517755623300651,
+      "epoch": 0.5349602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 9.911049571759719e-06,
-      "loss": 0.8079,
+      "learning_rate": 9.354337567770812e-06,
+      "loss": 0.844,
       "step": 18852
     },
     {
-      "epoch": 0.5177830875284941,
+      "epoch": 0.534988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 9.910160091290857e-06,
-      "loss": 0.9489,
+      "learning_rate": 9.353420410157579e-06,
+      "loss": 0.9974,
       "step": 18853
     },
     {
-      "epoch": 0.5178105517563374,
+      "epoch": 0.5350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 9.909270611532848e-06,
-      "loss": 0.8722,
+      "learning_rate": 9.352503258006031e-06,
+      "loss": 0.8398,
       "step": 18854
     },
     {
-      "epoch": 0.5178380159841806,
+      "epoch": 0.5350454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 9.908381132492722e-06,
-      "loss": 0.742,
+      "learning_rate": 9.351586111323921e-06,
+      "loss": 0.794,
       "step": 18855
     },
     {
-      "epoch": 0.5178654802120238,
+      "epoch": 0.5350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 9.907491654177523e-06,
-      "loss": 0.8982,
+      "learning_rate": 9.35066897011899e-06,
+      "loss": 0.8525,
       "step": 18856
     },
     {
-      "epoch": 0.5178929444398671,
+      "epoch": 0.5351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 9.906602176594282e-06,
-      "loss": 0.8187,
+      "learning_rate": 9.349751834398986e-06,
+      "loss": 0.9659,
       "step": 18857
     },
     {
-      "epoch": 0.5179204086677103,
+      "epoch": 0.5351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 9.90571269975004e-06,
-      "loss": 0.8877,
+      "learning_rate": 9.34883470417166e-06,
+      "loss": 0.8988,
       "step": 18858
     },
     {
-      "epoch": 0.5179478728955536,
+      "epoch": 0.5351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 9.90482322365184e-06,
-      "loss": 0.9897,
+      "learning_rate": 9.347917579444755e-06,
+      "loss": 0.8306,
       "step": 18859
     },
     {
-      "epoch": 0.5179753371233968,
+      "epoch": 0.5351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 9.903933748306713e-06,
-      "loss": 0.9069,
+      "learning_rate": 9.34700046022602e-06,
+      "loss": 0.8837,
       "step": 18860
     },
     {
-      "epoch": 0.51800280135124,
+      "epoch": 0.5352156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 9.903044273721697e-06,
-      "loss": 0.9642,
+      "learning_rate": 9.3460833465232e-06,
+      "loss": 0.7925,
       "step": 18861
     },
     {
-      "epoch": 0.5180302655790833,
+      "epoch": 0.5352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 9.90215479990384e-06,
-      "loss": 0.9369,
+      "learning_rate": 9.345166238344047e-06,
+      "loss": 0.9166,
       "step": 18862
     },
     {
-      "epoch": 0.5180577298069264,
+      "epoch": 0.535272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 9.901265326860167e-06,
-      "loss": 0.9145,
+      "learning_rate": 9.344249135696298e-06,
+      "loss": 0.9068,
       "step": 18863
     },
     {
-      "epoch": 0.5180851940347697,
+      "epoch": 0.5353007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 9.90037585459772e-06,
-      "loss": 0.9383,
+      "learning_rate": 9.343332038587707e-06,
+      "loss": 0.8889,
       "step": 18864
     },
     {
-      "epoch": 0.518112658262613,
+      "epoch": 0.5353291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 9.899486383123538e-06,
-      "loss": 0.8205,
+      "learning_rate": 9.342414947026026e-06,
+      "loss": 0.9598,
       "step": 18865
     },
     {
-      "epoch": 0.5181401224904562,
+      "epoch": 0.5353575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 9.898596912444661e-06,
-      "loss": 0.8448,
+      "learning_rate": 9.341497861018991e-06,
+      "loss": 0.9057,
       "step": 18866
     },
     {
-      "epoch": 0.5181675867182994,
+      "epoch": 0.5353859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 9.897707442568123e-06,
-      "loss": 0.8508,
+      "learning_rate": 9.340580780574355e-06,
+      "loss": 0.8765,
       "step": 18867
     },
     {
-      "epoch": 0.5181950509461426,
+      "epoch": 0.5354143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 9.896817973500965e-06,
-      "loss": 0.9194,
+      "learning_rate": 9.339663705699862e-06,
+      "loss": 0.8871,
       "step": 18868
     },
     {
-      "epoch": 0.5182225151739859,
+      "epoch": 0.5354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 9.895928505250221e-06,
-      "loss": 0.8702,
+      "learning_rate": 9.33874663640326e-06,
+      "loss": 0.9327,
       "step": 18869
     },
     {
-      "epoch": 0.5182499794018292,
+      "epoch": 0.5354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 9.89503903782293e-06,
-      "loss": 0.9315,
+      "learning_rate": 9.337829572692293e-06,
+      "loss": 0.8138,
       "step": 18870
     },
     {
-      "epoch": 0.5182774436296723,
+      "epoch": 0.53549943246311,
       "grad_norm": 0.0,
-      "learning_rate": 9.894149571226134e-06,
-      "loss": 0.8565,
+      "learning_rate": 9.336912514574712e-06,
+      "loss": 0.9031,
       "step": 18871
     },
     {
-      "epoch": 0.5183049078575156,
+      "epoch": 0.535527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 9.893260105466865e-06,
-      "loss": 0.9292,
+      "learning_rate": 9.33599546205826e-06,
+      "loss": 0.951,
       "step": 18872
     },
     {
-      "epoch": 0.5183323720853589,
+      "epoch": 0.5355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 9.892370640552168e-06,
-      "loss": 0.8552,
+      "learning_rate": 9.335078415150685e-06,
+      "loss": 0.9407,
       "step": 18873
     },
     {
-      "epoch": 0.518359836313202,
+      "epoch": 0.5355845629965947,
       "grad_norm": 0.0,
-      "learning_rate": 9.89148117648907e-06,
-      "loss": 0.8764,
+      "learning_rate": 9.334161373859729e-06,
+      "loss": 0.9152,
       "step": 18874
     },
     {
-      "epoch": 0.5183873005410453,
+      "epoch": 0.5356129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 9.890591713284619e-06,
-      "loss": 0.9395,
+      "learning_rate": 9.333244338193145e-06,
+      "loss": 0.9382,
       "step": 18875
     },
     {
-      "epoch": 0.5184147647688885,
+      "epoch": 0.5356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 9.889702250945847e-06,
-      "loss": 0.7763,
+      "learning_rate": 9.332327308158676e-06,
+      "loss": 0.8939,
       "step": 18876
     },
     {
-      "epoch": 0.5184422289967318,
+      "epoch": 0.5356696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 9.888812789479793e-06,
-      "loss": 0.9199,
+      "learning_rate": 9.331410283764066e-06,
+      "loss": 0.8524,
       "step": 18877
     },
     {
-      "epoch": 0.518469693224575,
+      "epoch": 0.5356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 9.887923328893497e-06,
-      "loss": 0.8417,
+      "learning_rate": 9.330493265017062e-06,
+      "loss": 0.8151,
       "step": 18878
     },
     {
-      "epoch": 0.5184971574524182,
+      "epoch": 0.5357264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 9.887033869193996e-06,
-      "loss": 0.841,
+      "learning_rate": 9.329576251925416e-06,
+      "loss": 0.8642,
       "step": 18879
     },
     {
-      "epoch": 0.5185246216802615,
+      "epoch": 0.5357548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 9.886144410388324e-06,
-      "loss": 0.9136,
+      "learning_rate": 9.328659244496867e-06,
+      "loss": 0.8675,
       "step": 18880
     },
     {
-      "epoch": 0.5185520859081046,
+      "epoch": 0.535783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 9.885254952483522e-06,
-      "loss": 0.8643,
+      "learning_rate": 9.327742242739164e-06,
+      "loss": 1.0232,
       "step": 18881
     },
     {
-      "epoch": 0.5185795501359479,
+      "epoch": 0.535811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 9.884365495486626e-06,
-      "loss": 0.8013,
+      "learning_rate": 9.326825246660053e-06,
+      "loss": 0.8118,
       "step": 18882
     },
     {
-      "epoch": 0.5186070143637912,
+      "epoch": 0.5358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 9.883476039404678e-06,
-      "loss": 0.8921,
+      "learning_rate": 9.32590825626728e-06,
+      "loss": 0.8843,
       "step": 18883
     },
     {
-      "epoch": 0.5186344785916344,
+      "epoch": 0.5358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.88258658424471e-06,
-      "loss": 0.9512,
+      "learning_rate": 9.324991271568589e-06,
+      "loss": 0.674,
       "step": 18884
     },
     {
-      "epoch": 0.5186619428194776,
+      "epoch": 0.5358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 9.881697130013768e-06,
-      "loss": 0.9603,
+      "learning_rate": 9.324074292571727e-06,
+      "loss": 0.9505,
       "step": 18885
     },
     {
-      "epoch": 0.5186894070473209,
+      "epoch": 0.5359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 9.880807676718878e-06,
-      "loss": 0.9293,
+      "learning_rate": 9.323157319284441e-06,
+      "loss": 0.8083,
       "step": 18886
     },
     {
-      "epoch": 0.5187168712751641,
+      "epoch": 0.5359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 9.879918224367087e-06,
-      "loss": 0.9074,
+      "learning_rate": 9.322240351714475e-06,
+      "loss": 0.935,
       "step": 18887
     },
     {
-      "epoch": 0.5187443355030074,
+      "epoch": 0.5359818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 9.879028772965428e-06,
-      "loss": 0.8638,
+      "learning_rate": 9.321323389869575e-06,
+      "loss": 0.963,
       "step": 18888
     },
     {
-      "epoch": 0.5187717997308505,
+      "epoch": 0.5360102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 9.878139322520938e-06,
-      "loss": 0.8788,
+      "learning_rate": 9.320406433757488e-06,
+      "loss": 0.8453,
       "step": 18889
     },
     {
-      "epoch": 0.5187992639586938,
+      "epoch": 0.536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 9.87724987304066e-06,
-      "loss": 0.8205,
+      "learning_rate": 9.319489483385956e-06,
+      "loss": 0.8744,
       "step": 18890
     },
     {
-      "epoch": 0.5188267281865371,
+      "epoch": 0.5360669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 9.876360424531632e-06,
-      "loss": 0.7977,
+      "learning_rate": 9.31857253876273e-06,
+      "loss": 0.9159,
       "step": 18891
     },
     {
-      "epoch": 0.5188541924143802,
+      "epoch": 0.5360953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 9.875470977000883e-06,
-      "loss": 0.9638,
+      "learning_rate": 9.317655599895551e-06,
+      "loss": 0.8266,
       "step": 18892
     },
     {
-      "epoch": 0.5188816566422235,
+      "epoch": 0.5361237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 9.874581530455456e-06,
-      "loss": 0.8257,
+      "learning_rate": 9.316738666792165e-06,
+      "loss": 0.8795,
       "step": 18893
     },
     {
-      "epoch": 0.5189091208700667,
+      "epoch": 0.5361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 9.873692084902388e-06,
-      "loss": 0.8822,
+      "learning_rate": 9.315821739460318e-06,
+      "loss": 0.8294,
       "step": 18894
     },
     {
-      "epoch": 0.51893658509791,
+      "epoch": 0.5361804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 9.872802640348721e-06,
-      "loss": 0.8512,
+      "learning_rate": 9.314904817907754e-06,
+      "loss": 0.7741,
       "step": 18895
     },
     {
-      "epoch": 0.5189640493257532,
+      "epoch": 0.5362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 9.871913196801485e-06,
-      "loss": 0.8619,
+      "learning_rate": 9.313987902142222e-06,
+      "loss": 0.8928,
       "step": 18896
     },
     {
-      "epoch": 0.5189915135535964,
+      "epoch": 0.5362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 9.871023754267726e-06,
-      "loss": 0.8201,
+      "learning_rate": 9.313070992171467e-06,
+      "loss": 0.9273,
       "step": 18897
     },
     {
-      "epoch": 0.5190189777814397,
+      "epoch": 0.5362656072644721,
       "grad_norm": 0.0,
-      "learning_rate": 9.870134312754473e-06,
-      "loss": 0.9349,
+      "learning_rate": 9.31215408800323e-06,
+      "loss": 0.901,
       "step": 18898
     },
     {
-      "epoch": 0.519046442009283,
+      "epoch": 0.5362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 9.86924487226877e-06,
-      "loss": 0.8879,
+      "learning_rate": 9.311237189645259e-06,
+      "loss": 0.7767,
       "step": 18899
     },
     {
-      "epoch": 0.5190739062371261,
+      "epoch": 0.536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 9.86835543281765e-06,
-      "loss": 0.7848,
+      "learning_rate": 9.310320297105298e-06,
+      "loss": 0.8875,
       "step": 18900
     },
     {
-      "epoch": 0.5191013704649694,
+      "epoch": 0.5363507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 9.867465994408151e-06,
-      "loss": 0.9821,
+      "learning_rate": 9.30940341039109e-06,
+      "loss": 0.9694,
       "step": 18901
     },
     {
-      "epoch": 0.5191288346928126,
+      "epoch": 0.5363791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 9.866576557047315e-06,
-      "loss": 0.8941,
+      "learning_rate": 9.308486529510386e-06,
+      "loss": 0.8824,
       "step": 18902
     },
     {
-      "epoch": 0.5191562989206558,
+      "epoch": 0.5364074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 9.86568712074218e-06,
-      "loss": 0.9807,
+      "learning_rate": 9.307569654470929e-06,
+      "loss": 0.903,
       "step": 18903
     },
     {
-      "epoch": 0.5191837631484991,
+      "epoch": 0.5364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 9.864797685499779e-06,
-      "loss": 0.9055,
+      "learning_rate": 9.306652785280456e-06,
+      "loss": 0.8986,
       "step": 18904
     },
     {
-      "epoch": 0.5192112273763423,
+      "epoch": 0.5364642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 9.863908251327148e-06,
-      "loss": 0.8445,
+      "learning_rate": 9.305735921946724e-06,
+      "loss": 0.9288,
       "step": 18905
     },
     {
-      "epoch": 0.5192386916041856,
+      "epoch": 0.5364926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 9.863018818231326e-06,
-      "loss": 0.8871,
+      "learning_rate": 9.30481906447747e-06,
+      "loss": 0.8685,
       "step": 18906
     },
     {
-      "epoch": 0.5192661558320287,
+      "epoch": 0.5365209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 9.862129386219357e-06,
-      "loss": 0.8581,
+      "learning_rate": 9.30390221288044e-06,
+      "loss": 0.8358,
       "step": 18907
     },
     {
-      "epoch": 0.519293620059872,
+      "epoch": 0.5365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 9.861239955298272e-06,
-      "loss": 0.9286,
+      "learning_rate": 9.30298536716338e-06,
+      "loss": 0.7988,
       "step": 18908
     },
     {
-      "epoch": 0.5193210842877153,
+      "epoch": 0.536577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 9.860350525475113e-06,
-      "loss": 0.9619,
+      "learning_rate": 9.30206852733403e-06,
+      "loss": 0.8162,
       "step": 18909
     },
     {
-      "epoch": 0.5193485485155585,
+      "epoch": 0.5366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 9.859461096756912e-06,
-      "loss": 1.0211,
+      "learning_rate": 9.301151693400142e-06,
+      "loss": 0.9294,
       "step": 18910
     },
     {
-      "epoch": 0.5193760127434017,
+      "epoch": 0.5366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 9.858571669150709e-06,
-      "loss": 0.8645,
+      "learning_rate": 9.300234865369455e-06,
+      "loss": 0.9087,
       "step": 18911
     },
     {
-      "epoch": 0.519403476971245,
+      "epoch": 0.5366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 9.857682242663541e-06,
-      "loss": 0.9044,
+      "learning_rate": 9.29931804324972e-06,
+      "loss": 0.8916,
       "step": 18912
     },
     {
-      "epoch": 0.5194309411990882,
+      "epoch": 0.5366912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 9.856792817302447e-06,
-      "loss": 0.9274,
+      "learning_rate": 9.298401227048674e-06,
+      "loss": 0.8756,
       "step": 18913
     },
     {
-      "epoch": 0.5194584054269314,
+      "epoch": 0.5367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 9.855903393074463e-06,
-      "loss": 0.8949,
+      "learning_rate": 9.297484416774066e-06,
+      "loss": 0.84,
       "step": 18914
     },
     {
-      "epoch": 0.5194858696547746,
+      "epoch": 0.5367480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 9.855013969986632e-06,
-      "loss": 0.9283,
+      "learning_rate": 9.296567612433635e-06,
+      "loss": 0.7975,
       "step": 18915
     },
     {
-      "epoch": 0.5195133338826179,
+      "epoch": 0.5367763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 9.854124548045982e-06,
-      "loss": 0.9096,
+      "learning_rate": 9.295650814035134e-06,
+      "loss": 0.8746,
       "step": 18916
     },
     {
-      "epoch": 0.5195407981104612,
+      "epoch": 0.5368047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 9.853235127259555e-06,
-      "loss": 0.7938,
+      "learning_rate": 9.2947340215863e-06,
+      "loss": 0.902,
       "step": 18917
     },
     {
-      "epoch": 0.5195682623383043,
+      "epoch": 0.53683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.852345707634389e-06,
-      "loss": 0.8702,
+      "learning_rate": 9.29381723509488e-06,
+      "loss": 0.9002,
       "step": 18918
     },
     {
-      "epoch": 0.5195957265661476,
+      "epoch": 0.536861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 9.851456289177521e-06,
-      "loss": 0.9939,
+      "learning_rate": 9.292900454568615e-06,
+      "loss": 0.8082,
       "step": 18919
     },
     {
-      "epoch": 0.5196231907939908,
+      "epoch": 0.5368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 9.850566871895987e-06,
-      "loss": 1.0241,
+      "learning_rate": 9.291983680015255e-06,
+      "loss": 0.8893,
       "step": 18920
     },
     {
-      "epoch": 0.519650655021834,
+      "epoch": 0.5369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 9.849677455796832e-06,
-      "loss": 0.9013,
+      "learning_rate": 9.291066911442537e-06,
+      "loss": 0.8304,
       "step": 18921
     },
     {
-      "epoch": 0.5196781192496773,
+      "epoch": 0.5369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 9.84878804088708e-06,
-      "loss": 0.9303,
+      "learning_rate": 9.290150148858212e-06,
+      "loss": 0.7903,
       "step": 18922
     },
     {
-      "epoch": 0.5197055834775205,
+      "epoch": 0.5369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 9.84789862717378e-06,
-      "loss": 0.8697,
+      "learning_rate": 9.28923339227002e-06,
+      "loss": 0.9127,
       "step": 18923
     },
     {
-      "epoch": 0.5197330477053638,
+      "epoch": 0.5370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 9.847009214663963e-06,
-      "loss": 0.8934,
+      "learning_rate": 9.288316641685702e-06,
+      "loss": 0.7555,
       "step": 18924
     },
     {
-      "epoch": 0.519760511933207,
+      "epoch": 0.5370317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 9.846119803364666e-06,
-      "loss": 0.9404,
+      "learning_rate": 9.287399897113007e-06,
+      "loss": 0.8984,
       "step": 18925
     },
     {
-      "epoch": 0.5197879761610502,
+      "epoch": 0.5370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 9.845230393282933e-06,
-      "loss": 0.9159,
+      "learning_rate": 9.286483158559679e-06,
+      "loss": 0.8486,
       "step": 18926
     },
     {
-      "epoch": 0.5198154403888935,
+      "epoch": 0.5370885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 9.844340984425799e-06,
-      "loss": 0.9543,
+      "learning_rate": 9.285566426033459e-06,
+      "loss": 0.8486,
       "step": 18927
     },
     {
-      "epoch": 0.5198429046167367,
+      "epoch": 0.5371169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 9.843451576800295e-06,
-      "loss": 0.8762,
+      "learning_rate": 9.284649699542092e-06,
+      "loss": 0.8652,
       "step": 18928
     },
     {
-      "epoch": 0.5198703688445799,
+      "epoch": 0.5371452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 9.842562170413462e-06,
-      "loss": 0.786,
+      "learning_rate": 9.283732979093322e-06,
+      "loss": 0.8064,
       "step": 18929
     },
     {
-      "epoch": 0.5198978330724232,
+      "epoch": 0.5371736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 9.841672765272341e-06,
-      "loss": 0.9064,
+      "learning_rate": 9.28281626469489e-06,
+      "loss": 0.8492,
       "step": 18930
     },
     {
-      "epoch": 0.5199252973002664,
+      "epoch": 0.5372020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 9.840783361383963e-06,
-      "loss": 0.9849,
+      "learning_rate": 9.28189955635454e-06,
+      "loss": 0.8028,
       "step": 18931
     },
     {
-      "epoch": 0.5199527615281097,
+      "epoch": 0.5372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 9.83989395875537e-06,
-      "loss": 0.843,
+      "learning_rate": 9.280982854080021e-06,
+      "loss": 0.9093,
       "step": 18932
     },
     {
-      "epoch": 0.5199802257559528,
+      "epoch": 0.5372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 9.839004557393601e-06,
-      "loss": 0.8988,
+      "learning_rate": 9.280066157879069e-06,
+      "loss": 0.8267,
       "step": 18933
     },
     {
-      "epoch": 0.5200076899837961,
+      "epoch": 0.5372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 9.838115157305688e-06,
-      "loss": 0.8583,
+      "learning_rate": 9.27914946775943e-06,
+      "loss": 0.9755,
       "step": 18934
     },
     {
-      "epoch": 0.5200351542116394,
+      "epoch": 0.5373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 9.837225758498671e-06,
-      "loss": 0.9787,
+      "learning_rate": 9.27823278372885e-06,
+      "loss": 0.8502,
       "step": 18935
     },
     {
-      "epoch": 0.5200626184394825,
+      "epoch": 0.5373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 9.836336360979587e-06,
-      "loss": 0.9572,
+      "learning_rate": 9.277316105795071e-06,
+      "loss": 0.954,
       "step": 18936
     },
     {
-      "epoch": 0.5200900826673258,
+      "epoch": 0.537372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 9.83544696475547e-06,
-      "loss": 0.8734,
+      "learning_rate": 9.276399433965835e-06,
+      "loss": 0.9427,
       "step": 18937
     },
     {
-      "epoch": 0.5201175468951691,
+      "epoch": 0.5374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 9.834557569833366e-06,
-      "loss": 0.9121,
+      "learning_rate": 9.275482768248882e-06,
+      "loss": 0.8099,
       "step": 18938
     },
     {
-      "epoch": 0.5201450111230123,
+      "epoch": 0.5374290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 9.833668176220305e-06,
-      "loss": 0.9015,
+      "learning_rate": 9.274566108651962e-06,
+      "loss": 0.9059,
       "step": 18939
     },
     {
-      "epoch": 0.5201724753508555,
+      "epoch": 0.5374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 9.832778783923324e-06,
-      "loss": 0.9201,
+      "learning_rate": 9.27364945518281e-06,
+      "loss": 0.9047,
       "step": 18940
     },
     {
-      "epoch": 0.5201999395786987,
+      "epoch": 0.5374858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 9.831889392949462e-06,
-      "loss": 0.9391,
+      "learning_rate": 9.272732807849177e-06,
+      "loss": 0.8946,
       "step": 18941
     },
     {
-      "epoch": 0.520227403806542,
+      "epoch": 0.5375141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 9.831000003305757e-06,
-      "loss": 0.9341,
+      "learning_rate": 9.271816166658801e-06,
+      "loss": 0.9216,
       "step": 18942
     },
     {
-      "epoch": 0.5202548680343853,
+      "epoch": 0.5375425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 9.830110614999244e-06,
-      "loss": 0.8536,
+      "learning_rate": 9.27089953161943e-06,
+      "loss": 0.9731,
       "step": 18943
     },
     {
-      "epoch": 0.5202823322622284,
+      "epoch": 0.5375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 9.829221228036966e-06,
-      "loss": 0.9727,
+      "learning_rate": 9.269982902738803e-06,
+      "loss": 0.8383,
       "step": 18944
     },
     {
-      "epoch": 0.5203097964900717,
+      "epoch": 0.5375993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 9.828331842425952e-06,
-      "loss": 0.8103,
+      "learning_rate": 9.26906628002466e-06,
+      "loss": 0.8072,
       "step": 18945
     },
     {
-      "epoch": 0.5203372607179149,
+      "epoch": 0.5376276958002271,
       "grad_norm": 0.0,
-      "learning_rate": 9.827442458173242e-06,
-      "loss": 0.9531,
+      "learning_rate": 9.268149663484749e-06,
+      "loss": 0.8477,
       "step": 18946
     },
     {
-      "epoch": 0.5203647249457581,
+      "epoch": 0.5376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 9.826553075285877e-06,
-      "loss": 0.9641,
+      "learning_rate": 9.267233053126811e-06,
+      "loss": 0.8019,
       "step": 18947
     },
     {
-      "epoch": 0.5203921891736014,
+      "epoch": 0.5376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 9.825663693770889e-06,
-      "loss": 0.977,
+      "learning_rate": 9.266316448958587e-06,
+      "loss": 0.8885,
       "step": 18948
     },
     {
-      "epoch": 0.5204196534014446,
+      "epoch": 0.5377128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 9.824774313635316e-06,
-      "loss": 0.9265,
+      "learning_rate": 9.265399850987824e-06,
+      "loss": 0.9408,
       "step": 18949
     },
     {
-      "epoch": 0.5204471176292879,
+      "epoch": 0.5377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 9.823884934886204e-06,
-      "loss": 0.97,
+      "learning_rate": 9.26448325922226e-06,
+      "loss": 0.9113,
       "step": 18950
     },
     {
-      "epoch": 0.5204745818571311,
+      "epoch": 0.5377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 9.822995557530576e-06,
-      "loss": 0.9191,
+      "learning_rate": 9.263566673669637e-06,
+      "loss": 0.927,
       "step": 18951
     },
     {
-      "epoch": 0.5205020460849743,
+      "epoch": 0.5377979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 9.822106181575477e-06,
-      "loss": 0.8104,
+      "learning_rate": 9.262650094337702e-06,
+      "loss": 0.8702,
       "step": 18952
     },
     {
-      "epoch": 0.5205295103128176,
+      "epoch": 0.5378263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 9.821216807027943e-06,
-      "loss": 0.882,
+      "learning_rate": 9.261733521234194e-06,
+      "loss": 0.9349,
       "step": 18953
     },
     {
-      "epoch": 0.5205569745406607,
+      "epoch": 0.5378547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 9.820327433895011e-06,
-      "loss": 0.8555,
+      "learning_rate": 9.260816954366853e-06,
+      "loss": 1.0676,
       "step": 18954
     },
     {
-      "epoch": 0.520584438768504,
+      "epoch": 0.537883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 9.819438062183718e-06,
-      "loss": 0.8222,
+      "learning_rate": 9.259900393743427e-06,
+      "loss": 0.7872,
       "step": 18955
     },
     {
-      "epoch": 0.5206119029963473,
+      "epoch": 0.5379114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 9.818548691901103e-06,
-      "loss": 0.8871,
+      "learning_rate": 9.258983839371655e-06,
+      "loss": 0.943,
       "step": 18956
     },
     {
-      "epoch": 0.5206393672241905,
+      "epoch": 0.5379398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 9.817659323054197e-06,
-      "loss": 0.8747,
+      "learning_rate": 9.25806729125928e-06,
+      "loss": 0.8956,
       "step": 18957
     },
     {
-      "epoch": 0.5206668314520337,
+      "epoch": 0.5379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 9.816769955650041e-06,
-      "loss": 0.8776,
+      "learning_rate": 9.257150749414045e-06,
+      "loss": 0.9065,
       "step": 18958
     },
     {
-      "epoch": 0.5206942956798769,
+      "epoch": 0.5379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 9.815880589695675e-06,
-      "loss": 0.946,
+      "learning_rate": 9.25623421384369e-06,
+      "loss": 0.8174,
       "step": 18959
     },
     {
-      "epoch": 0.5207217599077202,
+      "epoch": 0.5380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 9.814991225198131e-06,
-      "loss": 0.899,
+      "learning_rate": 9.25531768455596e-06,
+      "loss": 0.896,
       "step": 18960
     },
     {
-      "epoch": 0.5207492241355635,
+      "epoch": 0.5380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 9.81410186216445e-06,
-      "loss": 0.8877,
+      "learning_rate": 9.254401161558594e-06,
+      "loss": 0.7695,
       "step": 18961
     },
     {
-      "epoch": 0.5207766883634066,
+      "epoch": 0.5380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 9.813212500601668e-06,
-      "loss": 0.8605,
+      "learning_rate": 9.253484644859334e-06,
+      "loss": 0.865,
       "step": 18962
     },
     {
-      "epoch": 0.5208041525912499,
+      "epoch": 0.5381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 9.81232314051682e-06,
-      "loss": 0.835,
+      "learning_rate": 9.252568134465924e-06,
+      "loss": 0.8578,
       "step": 18963
     },
     {
-      "epoch": 0.5208316168190932,
+      "epoch": 0.5381384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 9.811433781916943e-06,
-      "loss": 1.0491,
+      "learning_rate": 9.251651630386103e-06,
+      "loss": 0.9774,
       "step": 18964
     },
     {
-      "epoch": 0.5208590810469363,
+      "epoch": 0.53816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.810544424809074e-06,
-      "loss": 0.8808,
+      "learning_rate": 9.250735132627614e-06,
+      "loss": 0.8918,
       "step": 18965
     },
     {
-      "epoch": 0.5208865452747796,
+      "epoch": 0.5381952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 9.809655069200253e-06,
-      "loss": 0.9129,
+      "learning_rate": 9.249818641198202e-06,
+      "loss": 0.9575,
       "step": 18966
     },
     {
-      "epoch": 0.5209140095026228,
+      "epoch": 0.5382236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 9.808765715097512e-06,
-      "loss": 0.9284,
+      "learning_rate": 9.248902156105605e-06,
+      "loss": 0.9254,
       "step": 18967
     },
     {
-      "epoch": 0.5209414737304661,
+      "epoch": 0.5382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 9.807876362507896e-06,
-      "loss": 0.8731,
+      "learning_rate": 9.247985677357563e-06,
+      "loss": 0.8549,
       "step": 18968
     },
     {
-      "epoch": 0.5209689379583093,
+      "epoch": 0.5382803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 9.806987011438432e-06,
-      "loss": 0.845,
+      "learning_rate": 9.24706920496182e-06,
+      "loss": 0.7987,
       "step": 18969
     },
     {
-      "epoch": 0.5209964021861525,
+      "epoch": 0.5383087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 9.806097661896162e-06,
-      "loss": 0.901,
+      "learning_rate": 9.246152738926115e-06,
+      "loss": 0.9274,
       "step": 18970
     },
     {
-      "epoch": 0.5210238664139958,
+      "epoch": 0.5383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 9.805208313888126e-06,
-      "loss": 0.8753,
+      "learning_rate": 9.245236279258197e-06,
+      "loss": 0.9366,
       "step": 18971
     },
     {
-      "epoch": 0.521051330641839,
+      "epoch": 0.5383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 9.804318967421352e-06,
-      "loss": 0.9692,
+      "learning_rate": 9.244319825965797e-06,
+      "loss": 0.8458,
       "step": 18972
     },
     {
-      "epoch": 0.5210787948696822,
+      "epoch": 0.538393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 9.803429622502883e-06,
-      "loss": 0.8627,
+      "learning_rate": 9.243403379056666e-06,
+      "loss": 0.9368,
       "step": 18973
     },
     {
-      "epoch": 0.5211062590975255,
+      "epoch": 0.538422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.802540279139761e-06,
-      "loss": 0.8242,
+      "learning_rate": 9.24248693853854e-06,
+      "loss": 0.8773,
       "step": 18974
     },
     {
-      "epoch": 0.5211337233253687,
+      "epoch": 0.5384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 9.801650937339014e-06,
-      "loss": 0.9238,
+      "learning_rate": 9.241570504419158e-06,
+      "loss": 0.8805,
       "step": 18975
     },
     {
-      "epoch": 0.521161187553212,
+      "epoch": 0.5384790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 9.800761597107678e-06,
-      "loss": 0.8207,
+      "learning_rate": 9.240654076706267e-06,
+      "loss": 0.7917,
       "step": 18976
     },
     {
-      "epoch": 0.5211886517810552,
+      "epoch": 0.5385073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 9.799872258452796e-06,
-      "loss": 0.8346,
+      "learning_rate": 9.239737655407605e-06,
+      "loss": 0.8056,
       "step": 18977
     },
     {
-      "epoch": 0.5212161160088984,
+      "epoch": 0.5385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 9.798982921381401e-06,
-      "loss": 0.9307,
+      "learning_rate": 9.238821240530914e-06,
+      "loss": 0.8764,
       "step": 18978
     },
     {
-      "epoch": 0.5212435802367417,
+      "epoch": 0.5385641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 9.798093585900531e-06,
-      "loss": 0.9832,
+      "learning_rate": 9.23790483208393e-06,
+      "loss": 0.865,
       "step": 18979
     },
     {
-      "epoch": 0.5212710444645848,
+      "epoch": 0.5385925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 9.797204252017226e-06,
-      "loss": 0.8261,
+      "learning_rate": 9.236988430074401e-06,
+      "loss": 0.8236,
       "step": 18980
     },
     {
-      "epoch": 0.5212985086924281,
+      "epoch": 0.5386208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 9.796314919738516e-06,
-      "loss": 0.8811,
+      "learning_rate": 9.236072034510067e-06,
+      "loss": 1.0663,
       "step": 18981
     },
     {
-      "epoch": 0.5213259729202714,
+      "epoch": 0.5386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 9.795425589071441e-06,
-      "loss": 0.7733,
+      "learning_rate": 9.235155645398663e-06,
+      "loss": 0.8608,
       "step": 18982
     },
     {
-      "epoch": 0.5213534371481146,
+      "epoch": 0.538677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 9.79453626002304e-06,
-      "loss": 0.9245,
+      "learning_rate": 9.234239262747935e-06,
+      "loss": 0.8858,
       "step": 18983
     },
     {
-      "epoch": 0.5213809013759578,
+      "epoch": 0.538706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 9.793646932600347e-06,
-      "loss": 0.8376,
+      "learning_rate": 9.233322886565624e-06,
+      "loss": 0.8559,
       "step": 18984
     },
     {
-      "epoch": 0.521408365603801,
+      "epoch": 0.5387343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 9.792757606810398e-06,
-      "loss": 0.9541,
+      "learning_rate": 9.232406516859465e-06,
+      "loss": 0.7684,
       "step": 18985
     },
     {
-      "epoch": 0.5214358298316443,
+      "epoch": 0.5387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 9.791868282660235e-06,
-      "loss": 0.8378,
+      "learning_rate": 9.231490153637203e-06,
+      "loss": 0.9146,
       "step": 18986
     },
     {
-      "epoch": 0.5214632940594875,
+      "epoch": 0.5387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 9.790978960156889e-06,
-      "loss": 0.9257,
+      "learning_rate": 9.23057379690658e-06,
+      "loss": 0.8306,
       "step": 18987
     },
     {
-      "epoch": 0.5214907582873307,
+      "epoch": 0.5388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 9.790089639307397e-06,
-      "loss": 0.7925,
+      "learning_rate": 9.229657446675337e-06,
+      "loss": 0.8591,
       "step": 18988
     },
     {
-      "epoch": 0.521518222515174,
+      "epoch": 0.5388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 9.789200320118797e-06,
-      "loss": 0.95,
+      "learning_rate": 9.22874110295121e-06,
+      "loss": 0.8907,
       "step": 18989
     },
     {
-      "epoch": 0.5215456867430173,
+      "epoch": 0.5388762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 9.788311002598127e-06,
-      "loss": 0.9131,
+      "learning_rate": 9.227824765741942e-06,
+      "loss": 0.8715,
       "step": 18990
     },
     {
-      "epoch": 0.5215731509708604,
+      "epoch": 0.5389046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 9.78742168675242e-06,
-      "loss": 0.9405,
+      "learning_rate": 9.226908435055273e-06,
+      "loss": 0.8062,
       "step": 18991
     },
     {
-      "epoch": 0.5216006151987037,
+      "epoch": 0.538933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 9.78653237258872e-06,
-      "loss": 0.7855,
+      "learning_rate": 9.225992110898941e-06,
+      "loss": 0.8982,
       "step": 18992
     },
     {
-      "epoch": 0.5216280794265469,
+      "epoch": 0.5389614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 9.785643060114055e-06,
-      "loss": 0.9568,
+      "learning_rate": 9.225075793280693e-06,
+      "loss": 0.8817,
       "step": 18993
     },
     {
-      "epoch": 0.5216555436543902,
+      "epoch": 0.5389897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 9.784753749335464e-06,
-      "loss": 0.8583,
+      "learning_rate": 9.22415948220826e-06,
+      "loss": 0.8459,
       "step": 18994
     },
     {
-      "epoch": 0.5216830078822334,
+      "epoch": 0.5390181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 9.783864440259987e-06,
-      "loss": 0.9779,
+      "learning_rate": 9.223243177689389e-06,
+      "loss": 0.849,
       "step": 18995
     },
     {
-      "epoch": 0.5217104721100766,
+      "epoch": 0.5390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 9.782975132894656e-06,
-      "loss": 0.9158,
+      "learning_rate": 9.222326879731818e-06,
+      "loss": 0.8884,
       "step": 18996
     },
     {
-      "epoch": 0.5217379363379199,
+      "epoch": 0.5390749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 9.782085827246515e-06,
-      "loss": 0.9224,
+      "learning_rate": 9.221410588343288e-06,
+      "loss": 0.8753,
       "step": 18997
     },
     {
-      "epoch": 0.5217654005657631,
+      "epoch": 0.5391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 9.78119652332259e-06,
-      "loss": 0.922,
+      "learning_rate": 9.220494303531534e-06,
+      "loss": 0.8848,
       "step": 18998
     },
     {
-      "epoch": 0.5217928647936063,
+      "epoch": 0.5391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.780307221129926e-06,
-      "loss": 0.8525,
+      "learning_rate": 9.219578025304303e-06,
+      "loss": 0.8575,
       "step": 18999
     },
     {
-      "epoch": 0.5218203290214496,
+      "epoch": 0.5391600454029511,
       "grad_norm": 0.0,
-      "learning_rate": 9.779417920675553e-06,
-      "loss": 0.9451,
+      "learning_rate": 9.21866175366933e-06,
+      "loss": 0.8784,
       "step": 19000
     },
     {
-      "epoch": 0.5218477932492928,
+      "epoch": 0.5391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 9.778528621966513e-06,
-      "loss": 0.9012,
+      "learning_rate": 9.217745488634354e-06,
+      "loss": 0.8799,
       "step": 19001
     },
     {
-      "epoch": 0.521875257477136,
+      "epoch": 0.539216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 9.77763932500984e-06,
-      "loss": 0.836,
+      "learning_rate": 9.21682923020712e-06,
+      "loss": 0.9063,
       "step": 19002
     },
     {
-      "epoch": 0.5219027217049793,
+      "epoch": 0.5392451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 9.776750029812572e-06,
-      "loss": 0.8936,
+      "learning_rate": 9.215912978395364e-06,
+      "loss": 0.909,
       "step": 19003
     },
     {
-      "epoch": 0.5219301859328225,
+      "epoch": 0.5392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 9.775860736381743e-06,
-      "loss": 0.8875,
+      "learning_rate": 9.214996733206827e-06,
+      "loss": 0.9062,
       "step": 19004
     },
     {
-      "epoch": 0.5219576501606658,
+      "epoch": 0.5393019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 9.77497144472439e-06,
-      "loss": 0.9103,
+      "learning_rate": 9.21408049464925e-06,
+      "loss": 0.8359,
       "step": 19005
     },
     {
-      "epoch": 0.5219851143885089,
+      "epoch": 0.5393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 9.774082154847551e-06,
-      "loss": 0.8792,
+      "learning_rate": 9.213164262730365e-06,
+      "loss": 0.9296,
       "step": 19006
     },
     {
-      "epoch": 0.5220125786163522,
+      "epoch": 0.5393586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 9.77319286675826e-06,
-      "loss": 0.9989,
+      "learning_rate": 9.21224803745792e-06,
+      "loss": 0.9093,
       "step": 19007
     },
     {
-      "epoch": 0.5220400428441955,
+      "epoch": 0.5393870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 9.772303580463557e-06,
-      "loss": 0.9356,
+      "learning_rate": 9.211331818839653e-06,
+      "loss": 0.8884,
       "step": 19008
     },
     {
-      "epoch": 0.5220675070720386,
+      "epoch": 0.5394154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 9.771414295970479e-06,
-      "loss": 0.7873,
+      "learning_rate": 9.210415606883298e-06,
+      "loss": 0.8065,
       "step": 19009
     },
     {
-      "epoch": 0.5220949712998819,
+      "epoch": 0.5394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 9.770525013286055e-06,
-      "loss": 1.0036,
+      "learning_rate": 9.2094994015966e-06,
+      "loss": 0.9412,
       "step": 19010
     },
     {
-      "epoch": 0.5221224355277252,
+      "epoch": 0.539472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 9.769635732417327e-06,
-      "loss": 0.8333,
+      "learning_rate": 9.208583202987297e-06,
+      "loss": 0.8578,
       "step": 19011
     },
     {
-      "epoch": 0.5221498997555684,
+      "epoch": 0.5395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 9.76874645337133e-06,
-      "loss": 0.9591,
+      "learning_rate": 9.207667011063124e-06,
+      "loss": 0.8313,
       "step": 19012
     },
     {
-      "epoch": 0.5221773639834116,
+      "epoch": 0.5395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 9.7678571761551e-06,
-      "loss": 0.9163,
+      "learning_rate": 9.206750825831826e-06,
+      "loss": 0.8835,
       "step": 19013
     },
     {
-      "epoch": 0.5222048282112548,
+      "epoch": 0.5395573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 9.766967900775676e-06,
-      "loss": 0.9087,
+      "learning_rate": 9.205834647301139e-06,
+      "loss": 0.8222,
       "step": 19014
     },
     {
-      "epoch": 0.5222322924390981,
+      "epoch": 0.5395856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 9.766078627240092e-06,
-      "loss": 0.8284,
+      "learning_rate": 9.2049184754788e-06,
+      "loss": 0.855,
       "step": 19015
     },
     {
-      "epoch": 0.5222597566669414,
+      "epoch": 0.5396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 9.765189355555384e-06,
-      "loss": 0.9265,
+      "learning_rate": 9.204002310372551e-06,
+      "loss": 0.8469,
       "step": 19016
     },
     {
-      "epoch": 0.5222872208947845,
+      "epoch": 0.5396424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 9.764300085728588e-06,
-      "loss": 0.8787,
+      "learning_rate": 9.203086151990132e-06,
+      "loss": 0.8203,
       "step": 19017
     },
     {
-      "epoch": 0.5223146851226278,
+      "epoch": 0.5396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 9.76341081776674e-06,
-      "loss": 0.9304,
+      "learning_rate": 9.20217000033928e-06,
+      "loss": 0.8763,
       "step": 19018
     },
     {
-      "epoch": 0.522342149350471,
+      "epoch": 0.5396992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 9.76252155167688e-06,
-      "loss": 0.8336,
+      "learning_rate": 9.201253855427734e-06,
+      "loss": 0.9607,
       "step": 19019
     },
     {
-      "epoch": 0.5223696135783142,
+      "epoch": 0.539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 9.761632287466039e-06,
-      "loss": 0.8718,
+      "learning_rate": 9.200337717263231e-06,
+      "loss": 0.8797,
       "step": 19020
     },
     {
-      "epoch": 0.5223970778061575,
+      "epoch": 0.539755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 9.76074302514126e-06,
-      "loss": 0.8849,
+      "learning_rate": 9.199421585853514e-06,
+      "loss": 0.9303,
       "step": 19021
     },
     {
-      "epoch": 0.5224245420340007,
+      "epoch": 0.5397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 9.759853764709572e-06,
-      "loss": 0.8593,
+      "learning_rate": 9.198505461206318e-06,
+      "loss": 0.9078,
       "step": 19022
     },
     {
-      "epoch": 0.522452006261844,
+      "epoch": 0.5398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 9.758964506178015e-06,
-      "loss": 1.0421,
+      "learning_rate": 9.19758934332938e-06,
+      "loss": 0.8566,
       "step": 19023
     },
     {
-      "epoch": 0.5224794704896872,
+      "epoch": 0.5398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 9.758075249553622e-06,
-      "loss": 0.9608,
+      "learning_rate": 9.196673232230444e-06,
+      "loss": 0.8727,
       "step": 19024
     },
     {
-      "epoch": 0.5225069347175304,
+      "epoch": 0.5398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 9.75718599484343e-06,
-      "loss": 0.8141,
+      "learning_rate": 9.195757127917245e-06,
+      "loss": 1.0159,
       "step": 19025
     },
     {
-      "epoch": 0.5225343989453737,
+      "epoch": 0.5398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 9.756296742054481e-06,
-      "loss": 0.8124,
+      "learning_rate": 9.194841030397518e-06,
+      "loss": 0.8075,
       "step": 19026
     },
     {
-      "epoch": 0.5225618631732168,
+      "epoch": 0.5399262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 9.755407491193807e-06,
-      "loss": 0.9278,
+      "learning_rate": 9.19392493967901e-06,
+      "loss": 0.8144,
       "step": 19027
     },
     {
-      "epoch": 0.5225893274010601,
+      "epoch": 0.5399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 9.75451824226844e-06,
-      "loss": 0.8615,
+      "learning_rate": 9.193008855769452e-06,
+      "loss": 0.7903,
       "step": 19028
     },
     {
-      "epoch": 0.5226167916289034,
+      "epoch": 0.5399829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 9.753628995285418e-06,
-      "loss": 0.8656,
+      "learning_rate": 9.192092778676583e-06,
+      "loss": 0.9615,
       "step": 19029
     },
     {
-      "epoch": 0.5226442558567466,
+      "epoch": 0.5400113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 9.752739750251782e-06,
-      "loss": 0.782,
+      "learning_rate": 9.191176708408145e-06,
+      "loss": 0.8595,
       "step": 19030
     },
     {
-      "epoch": 0.5226717200845898,
+      "epoch": 0.5400397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 9.751850507174566e-06,
-      "loss": 0.9966,
+      "learning_rate": 9.190260644971874e-06,
+      "loss": 0.869,
       "step": 19031
     },
     {
-      "epoch": 0.522699184312433,
+      "epoch": 0.5400681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 9.750961266060803e-06,
-      "loss": 0.8445,
+      "learning_rate": 9.189344588375503e-06,
+      "loss": 0.9708,
       "step": 19032
     },
     {
-      "epoch": 0.5227266485402763,
+      "epoch": 0.5400964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 9.750072026917535e-06,
-      "loss": 0.7804,
+      "learning_rate": 9.188428538626777e-06,
+      "loss": 0.8371,
       "step": 19033
     },
     {
-      "epoch": 0.5227541127681196,
+      "epoch": 0.5401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 9.749182789751787e-06,
-      "loss": 0.8705,
+      "learning_rate": 9.187512495733433e-06,
+      "loss": 0.8701,
       "step": 19034
     },
     {
-      "epoch": 0.5227815769959627,
+      "epoch": 0.5401532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 9.748293554570608e-06,
-      "loss": 0.818,
+      "learning_rate": 9.186596459703209e-06,
+      "loss": 0.9135,
       "step": 19035
     },
     {
-      "epoch": 0.522809041223806,
+      "epoch": 0.5401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 9.747404321381022e-06,
-      "loss": 0.8165,
+      "learning_rate": 9.185680430543838e-06,
+      "loss": 0.8733,
       "step": 19036
     },
     {
-      "epoch": 0.5228365054516493,
+      "epoch": 0.5402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 9.746515090190074e-06,
-      "loss": 0.7627,
+      "learning_rate": 9.184764408263062e-06,
+      "loss": 0.899,
       "step": 19037
     },
     {
-      "epoch": 0.5228639696794924,
+      "epoch": 0.5402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 9.745625861004798e-06,
-      "loss": 0.9305,
+      "learning_rate": 9.183848392868619e-06,
+      "loss": 0.9335,
       "step": 19038
     },
     {
-      "epoch": 0.5228914339073357,
+      "epoch": 0.540266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 9.74473663383223e-06,
-      "loss": 0.845,
+      "learning_rate": 9.182932384368245e-06,
+      "loss": 0.8724,
       "step": 19039
     },
     {
-      "epoch": 0.5229188981351789,
+      "epoch": 0.5402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 9.743847408679402e-06,
-      "loss": 0.8393,
+      "learning_rate": 9.182016382769678e-06,
+      "loss": 0.9158,
       "step": 19040
     },
     {
-      "epoch": 0.5229463623630222,
+      "epoch": 0.5403234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 9.742958185553351e-06,
-      "loss": 0.8783,
+      "learning_rate": 9.181100388080655e-06,
+      "loss": 0.9107,
       "step": 19041
     },
     {
-      "epoch": 0.5229738265908654,
+      "epoch": 0.5403518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 9.742068964461116e-06,
-      "loss": 0.9113,
+      "learning_rate": 9.180184400308913e-06,
+      "loss": 0.8482,
       "step": 19042
     },
     {
-      "epoch": 0.5230012908187086,
+      "epoch": 0.5403802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 9.74117974540973e-06,
-      "loss": 1.0361,
+      "learning_rate": 9.17926841946219e-06,
+      "loss": 0.8887,
       "step": 19043
     },
     {
-      "epoch": 0.5230287550465519,
+      "epoch": 0.5404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 9.740290528406232e-06,
-      "loss": 0.7756,
+      "learning_rate": 9.178352445548226e-06,
+      "loss": 0.9437,
       "step": 19044
     },
     {
-      "epoch": 0.523056219274395,
+      "epoch": 0.5404370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 9.739401313457659e-06,
-      "loss": 0.812,
+      "learning_rate": 9.177436478574755e-06,
+      "loss": 0.7646,
       "step": 19045
     },
     {
-      "epoch": 0.5230836835022383,
+      "epoch": 0.5404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 9.738512100571039e-06,
-      "loss": 0.8628,
+      "learning_rate": 9.176520518549514e-06,
+      "loss": 0.7206,
       "step": 19046
     },
     {
-      "epoch": 0.5231111477300816,
+      "epoch": 0.5404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 9.737622889753415e-06,
-      "loss": 0.8819,
+      "learning_rate": 9.175604565480237e-06,
+      "loss": 0.7976,
       "step": 19047
     },
     {
-      "epoch": 0.5231386119579248,
+      "epoch": 0.540522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 9.736733681011818e-06,
-      "loss": 0.8544,
+      "learning_rate": 9.174688619374671e-06,
+      "loss": 0.8637,
       "step": 19048
     },
     {
-      "epoch": 0.523166076185768,
+      "epoch": 0.5405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 9.735844474353285e-06,
-      "loss": 0.9679,
+      "learning_rate": 9.173772680240547e-06,
+      "loss": 0.8939,
       "step": 19049
     },
     {
-      "epoch": 0.5231935404136113,
+      "epoch": 0.5405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 9.734955269784857e-06,
-      "loss": 0.886,
+      "learning_rate": 9.1728567480856e-06,
+      "loss": 0.7755,
       "step": 19050
     },
     {
-      "epoch": 0.5232210046414545,
+      "epoch": 0.5406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 9.734066067313567e-06,
-      "loss": 0.946,
+      "learning_rate": 9.17194082291757e-06,
+      "loss": 0.7976,
       "step": 19051
     },
     {
-      "epoch": 0.5232484688692978,
+      "epoch": 0.5406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 9.733176866946444e-06,
-      "loss": 0.9685,
+      "learning_rate": 9.171024904744195e-06,
+      "loss": 0.8833,
       "step": 19052
     },
     {
-      "epoch": 0.5232759330971409,
+      "epoch": 0.5406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 9.73228766869053e-06,
-      "loss": 0.8865,
+      "learning_rate": 9.170108993573207e-06,
+      "loss": 0.855,
       "step": 19053
     },
     {
-      "epoch": 0.5233033973249842,
+      "epoch": 0.5406923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 9.731398472552858e-06,
-      "loss": 0.8651,
+      "learning_rate": 9.16919308941235e-06,
+      "loss": 0.8321,
       "step": 19054
     },
     {
-      "epoch": 0.5233308615528275,
+      "epoch": 0.5407207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 9.73050927854047e-06,
-      "loss": 0.875,
+      "learning_rate": 9.168277192269354e-06,
+      "loss": 0.9999,
       "step": 19055
     },
     {
-      "epoch": 0.5233583257806707,
+      "epoch": 0.5407491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 9.729620086660399e-06,
-      "loss": 0.8729,
+      "learning_rate": 9.167361302151956e-06,
+      "loss": 0.8717,
       "step": 19056
     },
     {
-      "epoch": 0.5233857900085139,
+      "epoch": 0.54077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 9.728730896919672e-06,
-      "loss": 0.9277,
+      "learning_rate": 9.166445419067896e-06,
+      "loss": 0.9366,
       "step": 19057
     },
     {
-      "epoch": 0.5234132542363571,
+      "epoch": 0.5408059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 9.727841709325333e-06,
-      "loss": 0.9489,
+      "learning_rate": 9.165529543024911e-06,
+      "loss": 0.9513,
       "step": 19058
     },
     {
-      "epoch": 0.5234407184642004,
+      "epoch": 0.5408342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 9.726952523884417e-06,
-      "loss": 0.9112,
+      "learning_rate": 9.164613674030731e-06,
+      "loss": 0.9039,
       "step": 19059
     },
     {
-      "epoch": 0.5234681826920436,
+      "epoch": 0.5408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 9.726063340603958e-06,
-      "loss": 0.9182,
+      "learning_rate": 9.163697812093101e-06,
+      "loss": 0.7799,
       "step": 19060
     },
     {
-      "epoch": 0.5234956469198868,
+      "epoch": 0.5408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 9.72517415949099e-06,
-      "loss": 0.8429,
+      "learning_rate": 9.162781957219753e-06,
+      "loss": 0.8425,
       "step": 19061
     },
     {
-      "epoch": 0.5235231111477301,
+      "epoch": 0.5409194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 9.724284980552556e-06,
-      "loss": 0.9128,
+      "learning_rate": 9.161866109418418e-06,
+      "loss": 0.9025,
       "step": 19062
     },
     {
-      "epoch": 0.5235505753755734,
+      "epoch": 0.5409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 9.723395803795683e-06,
-      "loss": 0.8753,
+      "learning_rate": 9.160950268696843e-06,
+      "loss": 0.8063,
       "step": 19063
     },
     {
-      "epoch": 0.5235780396034165,
+      "epoch": 0.5409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 9.722506629227407e-06,
-      "loss": 0.8891,
+      "learning_rate": 9.160034435062756e-06,
+      "loss": 0.8549,
       "step": 19064
     },
     {
-      "epoch": 0.5236055038312598,
+      "epoch": 0.5410045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 9.721617456854767e-06,
-      "loss": 0.8242,
+      "learning_rate": 9.159118608523898e-06,
+      "loss": 0.8418,
       "step": 19065
     },
     {
-      "epoch": 0.523632968059103,
+      "epoch": 0.5410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 9.7207282866848e-06,
-      "loss": 0.9455,
+      "learning_rate": 9.158202789088001e-06,
+      "loss": 0.9834,
       "step": 19066
     },
     {
-      "epoch": 0.5236604322869463,
+      "epoch": 0.541061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 9.719839118724537e-06,
-      "loss": 0.9146,
+      "learning_rate": 9.157286976762805e-06,
+      "loss": 0.9026,
       "step": 19067
     },
     {
-      "epoch": 0.5236878965147895,
+      "epoch": 0.5410896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 9.718949952981018e-06,
-      "loss": 0.8326,
+      "learning_rate": 9.156371171556044e-06,
+      "loss": 0.8502,
       "step": 19068
     },
     {
-      "epoch": 0.5237153607426327,
+      "epoch": 0.5411180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 9.718060789461273e-06,
-      "loss": 0.8533,
+      "learning_rate": 9.155455373475452e-06,
+      "loss": 0.7534,
       "step": 19069
     },
     {
-      "epoch": 0.523742824970476,
+      "epoch": 0.5411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 9.717171628172339e-06,
-      "loss": 0.9294,
+      "learning_rate": 9.154539582528766e-06,
+      "loss": 0.7884,
       "step": 19070
     },
     {
-      "epoch": 0.5237702891983191,
+      "epoch": 0.5411748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 9.716282469121257e-06,
-      "loss": 0.9576,
+      "learning_rate": 9.153623798723724e-06,
+      "loss": 0.8676,
       "step": 19071
     },
     {
-      "epoch": 0.5237977534261624,
+      "epoch": 0.5412031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 9.715393312315054e-06,
-      "loss": 0.822,
+      "learning_rate": 9.15270802206806e-06,
+      "loss": 0.8196,
       "step": 19072
     },
     {
-      "epoch": 0.5238252176540057,
+      "epoch": 0.5412315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 9.71450415776077e-06,
-      "loss": 0.8796,
+      "learning_rate": 9.151792252569506e-06,
+      "loss": 0.855,
       "step": 19073
     },
     {
-      "epoch": 0.5238526818818489,
+      "epoch": 0.5412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 9.713615005465444e-06,
-      "loss": 0.8972,
+      "learning_rate": 9.150876490235805e-06,
+      "loss": 0.8886,
       "step": 19074
     },
     {
-      "epoch": 0.5238801461096921,
+      "epoch": 0.5412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 9.712725855436105e-06,
-      "loss": 0.9398,
+      "learning_rate": 9.149960735074688e-06,
+      "loss": 0.8713,
       "step": 19075
     },
     {
-      "epoch": 0.5239076103375354,
+      "epoch": 0.541316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.711836707679787e-06,
-      "loss": 0.9592,
+      "learning_rate": 9.149044987093887e-06,
+      "loss": 0.8932,
       "step": 19076
     },
     {
-      "epoch": 0.5239350745653786,
+      "epoch": 0.5413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 9.71094756220353e-06,
-      "loss": 0.8441,
+      "learning_rate": 9.148129246301146e-06,
+      "loss": 0.92,
       "step": 19077
     },
     {
-      "epoch": 0.5239625387932219,
+      "epoch": 0.5413734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 9.710058419014369e-06,
-      "loss": 0.811,
+      "learning_rate": 9.14721351270419e-06,
+      "loss": 0.9411,
       "step": 19078
     },
     {
-      "epoch": 0.523990003021065,
+      "epoch": 0.5414018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 9.709169278119337e-06,
-      "loss": 0.9362,
+      "learning_rate": 9.146297786310764e-06,
+      "loss": 0.903,
       "step": 19079
     },
     {
-      "epoch": 0.5240174672489083,
+      "epoch": 0.5414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 9.708280139525473e-06,
-      "loss": 0.9634,
+      "learning_rate": 9.145382067128596e-06,
+      "loss": 0.8716,
       "step": 19080
     },
     {
-      "epoch": 0.5240449314767516,
+      "epoch": 0.5414585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 9.707391003239806e-06,
-      "loss": 0.9969,
+      "learning_rate": 9.144466355165428e-06,
+      "loss": 0.9564,
       "step": 19081
     },
     {
-      "epoch": 0.5240723957045947,
+      "epoch": 0.5414869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 9.706501869269375e-06,
-      "loss": 0.9746,
+      "learning_rate": 9.14355065042899e-06,
+      "loss": 0.8925,
       "step": 19082
     },
     {
-      "epoch": 0.524099859932438,
+      "epoch": 0.5415153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 9.705612737621219e-06,
-      "loss": 1.0021,
+      "learning_rate": 9.142634952927017e-06,
+      "loss": 0.8228,
       "step": 19083
     },
     {
-      "epoch": 0.5241273241602812,
+      "epoch": 0.5415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 9.704723608302364e-06,
-      "loss": 0.7801,
+      "learning_rate": 9.141719262667246e-06,
+      "loss": 0.9626,
       "step": 19084
     },
     {
-      "epoch": 0.5241547883881245,
+      "epoch": 0.541572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 9.703834481319852e-06,
-      "loss": 0.9125,
+      "learning_rate": 9.140803579657413e-06,
+      "loss": 0.8006,
       "step": 19085
     },
     {
-      "epoch": 0.5241822526159677,
+      "epoch": 0.541600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 9.702945356680719e-06,
-      "loss": 0.8644,
+      "learning_rate": 9.139887903905248e-06,
+      "loss": 0.8855,
       "step": 19086
     },
     {
-      "epoch": 0.5242097168438109,
+      "epoch": 0.5416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 9.702056234391995e-06,
-      "loss": 0.8897,
+      "learning_rate": 9.138972235418488e-06,
+      "loss": 0.8447,
       "step": 19087
     },
     {
-      "epoch": 0.5242371810716542,
+      "epoch": 0.5416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 9.701167114460717e-06,
-      "loss": 0.8958,
+      "learning_rate": 9.13805657420487e-06,
+      "loss": 0.8727,
       "step": 19088
     },
     {
-      "epoch": 0.5242646452994975,
+      "epoch": 0.5416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 9.700277996893919e-06,
-      "loss": 0.8751,
+      "learning_rate": 9.137140920272127e-06,
+      "loss": 0.9728,
       "step": 19089
     },
     {
-      "epoch": 0.5242921095273406,
+      "epoch": 0.5417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 9.699388881698641e-06,
-      "loss": 0.9492,
+      "learning_rate": 9.13622527362799e-06,
+      "loss": 0.9563,
       "step": 19090
     },
     {
-      "epoch": 0.5243195737551839,
+      "epoch": 0.5417423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 9.698499768881912e-06,
-      "loss": 0.9492,
+      "learning_rate": 9.135309634280202e-06,
+      "loss": 0.8516,
       "step": 19091
     },
     {
-      "epoch": 0.5243470379830271,
+      "epoch": 0.5417707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 9.697610658450774e-06,
-      "loss": 0.8448,
+      "learning_rate": 9.134394002236491e-06,
+      "loss": 0.8055,
       "step": 19092
     },
     {
-      "epoch": 0.5243745022108703,
+      "epoch": 0.5417990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 9.696721550412255e-06,
-      "loss": 0.9479,
+      "learning_rate": 9.133478377504588e-06,
+      "loss": 0.8595,
       "step": 19093
     },
     {
-      "epoch": 0.5244019664387136,
+      "epoch": 0.5418274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 9.69583244477339e-06,
-      "loss": 0.9459,
+      "learning_rate": 9.132562760092234e-06,
+      "loss": 0.915,
       "step": 19094
     },
     {
-      "epoch": 0.5244294306665568,
+      "epoch": 0.5418558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 9.69494334154122e-06,
-      "loss": 0.8556,
+      "learning_rate": 9.131647150007164e-06,
+      "loss": 0.7977,
       "step": 19095
     },
     {
-      "epoch": 0.5244568948944001,
+      "epoch": 0.5418842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 9.694054240722774e-06,
-      "loss": 0.8291,
+      "learning_rate": 9.13073154725711e-06,
+      "loss": 0.7327,
       "step": 19096
     },
     {
-      "epoch": 0.5244843591222432,
+      "epoch": 0.5419125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 9.693165142325088e-06,
-      "loss": 0.821,
+      "learning_rate": 9.129815951849803e-06,
+      "loss": 0.9688,
       "step": 19097
     },
     {
-      "epoch": 0.5245118233500865,
+      "epoch": 0.5419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 9.692276046355204e-06,
-      "loss": 0.913,
+      "learning_rate": 9.128900363792982e-06,
+      "loss": 0.9763,
       "step": 19098
     },
     {
-      "epoch": 0.5245392875779298,
+      "epoch": 0.5419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 9.691386952820149e-06,
-      "loss": 0.9162,
+      "learning_rate": 9.12798478309438e-06,
+      "loss": 0.7897,
       "step": 19099
     },
     {
-      "epoch": 0.524566751805773,
+      "epoch": 0.5419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 9.690497861726958e-06,
-      "loss": 0.9646,
+      "learning_rate": 9.127069209761726e-06,
+      "loss": 0.8336,
       "step": 19100
     },
     {
-      "epoch": 0.5245942160336162,
+      "epoch": 0.5420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 9.689608773082668e-06,
-      "loss": 0.8632,
+      "learning_rate": 9.126153643802759e-06,
+      "loss": 0.9284,
       "step": 19101
     },
     {
-      "epoch": 0.5246216802614595,
+      "epoch": 0.5420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 9.688719686894314e-06,
-      "loss": 1.0255,
+      "learning_rate": 9.125238085225214e-06,
+      "loss": 1.0554,
       "step": 19102
     },
     {
-      "epoch": 0.5246491444893027,
+      "epoch": 0.5420828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 9.687830603168929e-06,
-      "loss": 0.8259,
+      "learning_rate": 9.124322534036816e-06,
+      "loss": 0.8873,
       "step": 19103
     },
     {
-      "epoch": 0.5246766087171459,
+      "epoch": 0.54211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.686941521913554e-06,
-      "loss": 0.8338,
+      "learning_rate": 9.123406990245312e-06,
+      "loss": 0.9184,
       "step": 19104
     },
     {
-      "epoch": 0.5247040729449891,
+      "epoch": 0.5421396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 9.686052443135216e-06,
-      "loss": 0.8972,
+      "learning_rate": 9.122491453858425e-06,
+      "loss": 0.8265,
       "step": 19105
     },
     {
-      "epoch": 0.5247315371728324,
+      "epoch": 0.5421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 9.68516336684095e-06,
-      "loss": 1.0116,
+      "learning_rate": 9.121575924883892e-06,
+      "loss": 0.8978,
       "step": 19106
     },
     {
-      "epoch": 0.5247590014006757,
+      "epoch": 0.5421963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 9.684274293037798e-06,
-      "loss": 0.8282,
+      "learning_rate": 9.120660403329447e-06,
+      "loss": 0.9907,
       "step": 19107
     },
     {
-      "epoch": 0.5247864656285188,
+      "epoch": 0.5422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 9.683385221732787e-06,
-      "loss": 0.9265,
+      "learning_rate": 9.11974488920282e-06,
+      "loss": 0.8438,
       "step": 19108
     },
     {
-      "epoch": 0.5248139298563621,
+      "epoch": 0.5422531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 9.682496152932954e-06,
-      "loss": 0.8902,
+      "learning_rate": 9.118829382511752e-06,
+      "loss": 0.8441,
       "step": 19109
     },
     {
-      "epoch": 0.5248413940842053,
+      "epoch": 0.5422814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 9.68160708664534e-06,
-      "loss": 0.932,
+      "learning_rate": 9.117913883263972e-06,
+      "loss": 0.9654,
       "step": 19110
     },
     {
-      "epoch": 0.5248688583120485,
+      "epoch": 0.5423098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 9.68071802287697e-06,
-      "loss": 0.9122,
+      "learning_rate": 9.11699839146721e-06,
+      "loss": 0.9829,
       "step": 19111
     },
     {
-      "epoch": 0.5248963225398918,
+      "epoch": 0.5423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 9.679828961634882e-06,
-      "loss": 0.8024,
+      "learning_rate": 9.116082907129204e-06,
+      "loss": 0.9754,
       "step": 19112
     },
     {
-      "epoch": 0.524923786767735,
+      "epoch": 0.542366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 9.67893990292611e-06,
-      "loss": 0.8496,
+      "learning_rate": 9.115167430257687e-06,
+      "loss": 0.821,
       "step": 19113
     },
     {
-      "epoch": 0.5249512509955783,
+      "epoch": 0.5423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 9.678050846757693e-06,
-      "loss": 0.7768,
+      "learning_rate": 9.114251960860387e-06,
+      "loss": 0.8936,
       "step": 19114
     },
     {
-      "epoch": 0.5249787152234215,
+      "epoch": 0.5424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 9.677161793136662e-06,
-      "loss": 0.9001,
+      "learning_rate": 9.113336498945045e-06,
+      "loss": 0.8981,
       "step": 19115
     },
     {
-      "epoch": 0.5250061794512647,
+      "epoch": 0.5424517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 9.676272742070053e-06,
-      "loss": 0.9037,
+      "learning_rate": 9.112421044519388e-06,
+      "loss": 0.7289,
       "step": 19116
     },
     {
-      "epoch": 0.525033643679108,
+      "epoch": 0.5424801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 9.675383693564896e-06,
-      "loss": 0.9822,
+      "learning_rate": 9.111505597591147e-06,
+      "loss": 0.9229,
       "step": 19117
     },
     {
-      "epoch": 0.5250611079069512,
+      "epoch": 0.5425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 9.67449464762823e-06,
-      "loss": 0.7936,
+      "learning_rate": 9.110590158168061e-06,
+      "loss": 0.9721,
       "step": 19118
     },
     {
-      "epoch": 0.5250885721347944,
+      "epoch": 0.5425368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 9.673605604267091e-06,
-      "loss": 0.9575,
+      "learning_rate": 9.10967472625786e-06,
+      "loss": 0.7985,
       "step": 19119
     },
     {
-      "epoch": 0.5251160363626377,
+      "epoch": 0.5425652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 9.672716563488507e-06,
-      "loss": 0.8526,
+      "learning_rate": 9.108759301868274e-06,
+      "loss": 0.7576,
       "step": 19120
     },
     {
-      "epoch": 0.5251435005904809,
+      "epoch": 0.5425936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 9.671827525299522e-06,
-      "loss": 0.8525,
+      "learning_rate": 9.107843885007042e-06,
+      "loss": 0.9683,
       "step": 19121
     },
     {
-      "epoch": 0.5251709648183241,
+      "epoch": 0.542622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 9.67093848970716e-06,
-      "loss": 0.887,
+      "learning_rate": 9.106928475681893e-06,
+      "loss": 0.9393,
       "step": 19122
     },
     {
-      "epoch": 0.5251984290461673,
+      "epoch": 0.542650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 9.670049456718462e-06,
-      "loss": 0.8962,
+      "learning_rate": 9.106013073900554e-06,
+      "loss": 0.9173,
       "step": 19123
     },
     {
-      "epoch": 0.5252258932740106,
+      "epoch": 0.5426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 9.669160426340459e-06,
-      "loss": 0.8362,
+      "learning_rate": 9.105097679670764e-06,
+      "loss": 0.8495,
       "step": 19124
     },
     {
-      "epoch": 0.5252533575018539,
+      "epoch": 0.5427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 9.668271398580187e-06,
-      "loss": 0.8303,
+      "learning_rate": 9.104182293000256e-06,
+      "loss": 0.9763,
       "step": 19125
     },
     {
-      "epoch": 0.525280821729697,
+      "epoch": 0.5427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 9.66738237344468e-06,
-      "loss": 0.8411,
+      "learning_rate": 9.103266913896763e-06,
+      "loss": 0.7969,
       "step": 19126
     },
     {
-      "epoch": 0.5253082859575403,
+      "epoch": 0.5427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 9.666493350940975e-06,
-      "loss": 0.8698,
+      "learning_rate": 9.102351542368012e-06,
+      "loss": 0.9658,
       "step": 19127
     },
     {
-      "epoch": 0.5253357501853836,
+      "epoch": 0.5427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 9.665604331076104e-06,
-      "loss": 0.9539,
+      "learning_rate": 9.101436178421736e-06,
+      "loss": 0.8451,
       "step": 19128
     },
     {
-      "epoch": 0.5253632144132268,
+      "epoch": 0.5428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 9.664715313857098e-06,
-      "loss": 0.9093,
+      "learning_rate": 9.100520822065672e-06,
+      "loss": 0.8388,
       "step": 19129
     },
     {
-      "epoch": 0.52539067864107,
+      "epoch": 0.5428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 9.663826299290994e-06,
-      "loss": 0.8774,
+      "learning_rate": 9.09960547330755e-06,
+      "loss": 0.8024,
       "step": 19130
     },
     {
-      "epoch": 0.5254181428689132,
+      "epoch": 0.5428774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 9.66293728738483e-06,
-      "loss": 0.9172,
+      "learning_rate": 9.098690132155098e-06,
+      "loss": 0.8905,
       "step": 19131
     },
     {
-      "epoch": 0.5254456070967565,
+      "epoch": 0.5429057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 9.662048278145633e-06,
-      "loss": 0.9343,
+      "learning_rate": 9.097774798616053e-06,
+      "loss": 0.8684,
       "step": 19132
     },
     {
-      "epoch": 0.5254730713245997,
+      "epoch": 0.5429341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 9.661159271580446e-06,
-      "loss": 0.8819,
+      "learning_rate": 9.096859472698145e-06,
+      "loss": 0.8488,
       "step": 19133
     },
     {
-      "epoch": 0.5255005355524429,
+      "epoch": 0.5429625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 9.660270267696294e-06,
-      "loss": 0.9434,
+      "learning_rate": 9.095944154409103e-06,
+      "loss": 0.8751,
       "step": 19134
     },
     {
-      "epoch": 0.5255279997802862,
+      "epoch": 0.5429909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 9.659381266500217e-06,
-      "loss": 0.8252,
+      "learning_rate": 9.095028843756663e-06,
+      "loss": 0.9444,
       "step": 19135
     },
     {
-      "epoch": 0.5255554640081294,
+      "epoch": 0.5430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 9.658492267999245e-06,
-      "loss": 0.9466,
+      "learning_rate": 9.094113540748556e-06,
+      "loss": 0.8676,
       "step": 19136
     },
     {
-      "epoch": 0.5255829282359726,
+      "epoch": 0.5430476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 9.657603272200414e-06,
-      "loss": 0.871,
+      "learning_rate": 9.093198245392512e-06,
+      "loss": 0.8798,
       "step": 19137
     },
     {
-      "epoch": 0.5256103924638159,
+      "epoch": 0.5430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 9.656714279110761e-06,
-      "loss": 0.8708,
+      "learning_rate": 9.092282957696264e-06,
+      "loss": 0.8371,
       "step": 19138
     },
     {
-      "epoch": 0.5256378566916591,
+      "epoch": 0.5431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 9.65582528873732e-06,
-      "loss": 0.8401,
+      "learning_rate": 9.091367677667538e-06,
+      "loss": 0.8644,
       "step": 19139
     },
     {
-      "epoch": 0.5256653209195024,
+      "epoch": 0.5431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 9.65493630108712e-06,
-      "loss": 0.9014,
+      "learning_rate": 9.090452405314074e-06,
+      "loss": 0.7919,
       "step": 19140
     },
     {
-      "epoch": 0.5256927851473456,
+      "epoch": 0.543161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 9.654047316167195e-06,
-      "loss": 0.9814,
+      "learning_rate": 9.089537140643598e-06,
+      "loss": 0.8469,
       "step": 19141
     },
     {
-      "epoch": 0.5257202493751888,
+      "epoch": 0.5431895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 9.653158333984584e-06,
-      "loss": 0.8499,
+      "learning_rate": 9.088621883663845e-06,
+      "loss": 0.7928,
       "step": 19142
     },
     {
-      "epoch": 0.5257477136030321,
+      "epoch": 0.5432179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 9.65226935454632e-06,
-      "loss": 0.9164,
+      "learning_rate": 9.087706634382543e-06,
+      "loss": 0.7958,
       "step": 19143
     },
     {
-      "epoch": 0.5257751778308752,
+      "epoch": 0.5432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 9.651380377859433e-06,
-      "loss": 0.8715,
+      "learning_rate": 9.08679139280742e-06,
+      "loss": 0.9119,
       "step": 19144
     },
     {
-      "epoch": 0.5258026420587185,
+      "epoch": 0.5432746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 9.650491403930963e-06,
-      "loss": 0.8185,
+      "learning_rate": 9.085876158946217e-06,
+      "loss": 0.8098,
       "step": 19145
     },
     {
-      "epoch": 0.5258301062865618,
+      "epoch": 0.5433030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 9.649602432767937e-06,
-      "loss": 0.9156,
+      "learning_rate": 9.084960932806659e-06,
+      "loss": 0.8382,
       "step": 19146
     },
     {
-      "epoch": 0.525857570514405,
+      "epoch": 0.5433314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 9.648713464377396e-06,
-      "loss": 0.9895,
+      "learning_rate": 9.084045714396473e-06,
+      "loss": 1.0353,
       "step": 19147
     },
     {
-      "epoch": 0.5258850347422482,
+      "epoch": 0.5433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 9.647824498766366e-06,
-      "loss": 0.9109,
+      "learning_rate": 9.083130503723397e-06,
+      "loss": 0.8686,
       "step": 19148
     },
     {
-      "epoch": 0.5259124989700914,
+      "epoch": 0.5433881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 9.646935535941884e-06,
-      "loss": 0.8643,
+      "learning_rate": 9.082215300795159e-06,
+      "loss": 0.9493,
       "step": 19149
     },
     {
-      "epoch": 0.5259399631979347,
+      "epoch": 0.543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 9.646046575910987e-06,
-      "loss": 0.8344,
+      "learning_rate": 9.081300105619488e-06,
+      "loss": 0.9078,
       "step": 19150
     },
     {
-      "epoch": 0.525967427425778,
+      "epoch": 0.5434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 9.64515761868071e-06,
-      "loss": 0.826,
+      "learning_rate": 9.080384918204118e-06,
+      "loss": 0.9303,
       "step": 19151
     },
     {
-      "epoch": 0.5259948916536211,
+      "epoch": 0.5434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 9.64426866425808e-06,
-      "loss": 0.905,
+      "learning_rate": 9.079469738556777e-06,
+      "loss": 0.8534,
       "step": 19152
     },
     {
-      "epoch": 0.5260223558814644,
+      "epoch": 0.5435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 9.643379712650134e-06,
-      "loss": 0.8883,
+      "learning_rate": 9.078554566685198e-06,
+      "loss": 0.9335,
       "step": 19153
     },
     {
-      "epoch": 0.5260498201093077,
+      "epoch": 0.5435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 9.642490763863904e-06,
-      "loss": 0.882,
+      "learning_rate": 9.077639402597105e-06,
+      "loss": 0.7507,
       "step": 19154
     },
     {
-      "epoch": 0.5260772843371508,
+      "epoch": 0.5435584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 9.641601817906429e-06,
-      "loss": 0.8386,
+      "learning_rate": 9.076724246300237e-06,
+      "loss": 0.858,
       "step": 19155
     },
     {
-      "epoch": 0.5261047485649941,
+      "epoch": 0.5435868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 9.640712874784737e-06,
-      "loss": 0.9842,
+      "learning_rate": 9.075809097802319e-06,
+      "loss": 0.7967,
       "step": 19156
     },
     {
-      "epoch": 0.5261322127928373,
+      "epoch": 0.5436152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 9.639823934505869e-06,
-      "loss": 0.883,
+      "learning_rate": 9.074893957111087e-06,
+      "loss": 0.8538,
       "step": 19157
     },
     {
-      "epoch": 0.5261596770206806,
+      "epoch": 0.5436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 9.638934997076848e-06,
-      "loss": 0.8902,
+      "learning_rate": 9.073978824234262e-06,
+      "loss": 0.6791,
       "step": 19158
     },
     {
-      "epoch": 0.5261871412485238,
+      "epoch": 0.543671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 9.638046062504715e-06,
-      "loss": 0.9915,
+      "learning_rate": 9.073063699179586e-06,
+      "loss": 0.9633,
       "step": 19159
     },
     {
-      "epoch": 0.526214605476367,
+      "epoch": 0.5437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 9.6371571307965e-06,
-      "loss": 0.9757,
+      "learning_rate": 9.072148581954778e-06,
+      "loss": 0.8938,
       "step": 19160
     },
     {
-      "epoch": 0.5262420697042103,
+      "epoch": 0.5437287173666289,
       "grad_norm": 0.0,
-      "learning_rate": 9.636268201959238e-06,
-      "loss": 0.8867,
+      "learning_rate": 9.071233472567573e-06,
+      "loss": 0.8811,
       "step": 19161
     },
     {
-      "epoch": 0.5262695339320534,
+      "epoch": 0.5437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 9.635379275999967e-06,
-      "loss": 0.8864,
+      "learning_rate": 9.070318371025702e-06,
+      "loss": 1.0215,
       "step": 19162
     },
     {
-      "epoch": 0.5262969981598967,
+      "epoch": 0.5437854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 9.634490352925715e-06,
-      "loss": 0.8816,
+      "learning_rate": 9.069403277336895e-06,
+      "loss": 0.8431,
       "step": 19163
     },
     {
-      "epoch": 0.52632446238774,
+      "epoch": 0.5438138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 9.633601432743517e-06,
-      "loss": 1.058,
+      "learning_rate": 9.068488191508876e-06,
+      "loss": 0.8289,
       "step": 19164
     },
     {
-      "epoch": 0.5263519266155832,
+      "epoch": 0.5438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.632712515460404e-06,
-      "loss": 0.8325,
+      "learning_rate": 9.067573113549383e-06,
+      "loss": 0.8029,
       "step": 19165
     },
     {
-      "epoch": 0.5263793908434264,
+      "epoch": 0.5438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 9.631823601083412e-06,
-      "loss": 0.8499,
+      "learning_rate": 9.06665804346614e-06,
+      "loss": 0.7655,
       "step": 19166
     },
     {
-      "epoch": 0.5264068550712697,
+      "epoch": 0.5438989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 9.630934689619576e-06,
-      "loss": 0.8382,
+      "learning_rate": 9.065742981266878e-06,
+      "loss": 0.8951,
       "step": 19167
     },
     {
-      "epoch": 0.5264343192991129,
+      "epoch": 0.5439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 9.630045781075926e-06,
-      "loss": 0.8286,
+      "learning_rate": 9.064827926959329e-06,
+      "loss": 0.9877,
       "step": 19168
     },
     {
-      "epoch": 0.5264617835269562,
+      "epoch": 0.543955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 9.629156875459502e-06,
-      "loss": 0.913,
+      "learning_rate": 9.06391288055122e-06,
+      "loss": 0.9365,
       "step": 19169
     },
     {
-      "epoch": 0.5264892477547993,
+      "epoch": 0.5439841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 9.628267972777328e-06,
-      "loss": 0.9146,
+      "learning_rate": 9.062997842050276e-06,
+      "loss": 0.8614,
       "step": 19170
     },
     {
-      "epoch": 0.5265167119826426,
+      "epoch": 0.5440124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 9.627379073036444e-06,
-      "loss": 0.9167,
+      "learning_rate": 9.062082811464236e-06,
+      "loss": 0.8598,
       "step": 19171
     },
     {
-      "epoch": 0.5265441762104859,
+      "epoch": 0.5440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 9.626490176243878e-06,
-      "loss": 0.9036,
+      "learning_rate": 9.061167788800825e-06,
+      "loss": 0.9216,
       "step": 19172
     },
     {
-      "epoch": 0.526571640438329,
+      "epoch": 0.5440692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 9.625601282406667e-06,
-      "loss": 0.8518,
+      "learning_rate": 9.060252774067771e-06,
+      "loss": 0.8325,
       "step": 19173
     },
     {
-      "epoch": 0.5265991046661723,
+      "epoch": 0.5440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 9.624712391531848e-06,
-      "loss": 0.8752,
+      "learning_rate": 9.059337767272806e-06,
+      "loss": 0.9138,
       "step": 19174
     },
     {
-      "epoch": 0.5266265688940155,
+      "epoch": 0.5441259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 9.623823503626445e-06,
-      "loss": 0.8782,
+      "learning_rate": 9.058422768423653e-06,
+      "loss": 0.8152,
       "step": 19175
     },
     {
-      "epoch": 0.5266540331218588,
+      "epoch": 0.5441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 9.622934618697498e-06,
-      "loss": 0.8443,
+      "learning_rate": 9.057507777528048e-06,
+      "loss": 0.9669,
       "step": 19176
     },
     {
-      "epoch": 0.526681497349702,
+      "epoch": 0.5441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 9.622045736752038e-06,
-      "loss": 0.9548,
+      "learning_rate": 9.056592794593719e-06,
+      "loss": 0.8909,
       "step": 19177
     },
     {
-      "epoch": 0.5267089615775452,
+      "epoch": 0.544211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.621156857797098e-06,
-      "loss": 0.8634,
+      "learning_rate": 9.05567781962839e-06,
+      "loss": 0.8464,
       "step": 19178
     },
     {
-      "epoch": 0.5267364258053885,
+      "epoch": 0.5442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 9.620267981839713e-06,
-      "loss": 0.8737,
+      "learning_rate": 9.054762852639793e-06,
+      "loss": 0.845,
       "step": 19179
     },
     {
-      "epoch": 0.5267638900332318,
+      "epoch": 0.5442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 9.619379108886918e-06,
-      "loss": 0.8691,
+      "learning_rate": 9.053847893635658e-06,
+      "loss": 0.8257,
       "step": 19180
     },
     {
-      "epoch": 0.5267913542610749,
+      "epoch": 0.5442962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 9.618490238945737e-06,
-      "loss": 0.8987,
+      "learning_rate": 9.05293294262371e-06,
+      "loss": 0.8441,
       "step": 19181
     },
     {
-      "epoch": 0.5268188184889182,
+      "epoch": 0.5443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 9.61760137202321e-06,
-      "loss": 0.8295,
+      "learning_rate": 9.052017999611684e-06,
+      "loss": 0.9375,
       "step": 19182
     },
     {
-      "epoch": 0.5268462827167614,
+      "epoch": 0.5443530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 9.616712508126371e-06,
-      "loss": 0.8558,
+      "learning_rate": 9.051103064607303e-06,
+      "loss": 0.9018,
       "step": 19183
     },
     {
-      "epoch": 0.5268737469446046,
+      "epoch": 0.5443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 9.61582364726225e-06,
-      "loss": 0.917,
+      "learning_rate": 9.050188137618296e-06,
+      "loss": 0.9399,
       "step": 19184
     },
     {
-      "epoch": 0.5269012111724479,
+      "epoch": 0.5444097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 9.61493478943788e-06,
-      "loss": 0.9577,
+      "learning_rate": 9.04927321865239e-06,
+      "loss": 0.9546,
       "step": 19185
     },
     {
-      "epoch": 0.5269286754002911,
+      "epoch": 0.5444381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 9.6140459346603e-06,
-      "loss": 0.9419,
+      "learning_rate": 9.048358307717318e-06,
+      "loss": 0.8854,
       "step": 19186
     },
     {
-      "epoch": 0.5269561396281344,
+      "epoch": 0.544466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 9.613157082936535e-06,
-      "loss": 0.857,
+      "learning_rate": 9.047443404820808e-06,
+      "loss": 0.8989,
       "step": 19187
     },
     {
-      "epoch": 0.5269836038559775,
+      "epoch": 0.544494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 9.61226823427362e-06,
-      "loss": 0.9441,
+      "learning_rate": 9.046528509970584e-06,
+      "loss": 0.9343,
       "step": 19188
     },
     {
-      "epoch": 0.5270110680838208,
+      "epoch": 0.5445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 9.611379388678588e-06,
-      "loss": 0.8949,
+      "learning_rate": 9.045613623174378e-06,
+      "loss": 0.8012,
       "step": 19189
     },
     {
-      "epoch": 0.5270385323116641,
+      "epoch": 0.5445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 9.610490546158472e-06,
-      "loss": 0.9806,
+      "learning_rate": 9.044698744439918e-06,
+      "loss": 0.927,
       "step": 19190
     },
     {
-      "epoch": 0.5270659965395073,
+      "epoch": 0.5445800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 9.609601706720311e-06,
-      "loss": 0.8986,
+      "learning_rate": 9.04378387377493e-06,
+      "loss": 0.7549,
       "step": 19191
     },
     {
-      "epoch": 0.5270934607673505,
+      "epoch": 0.5446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 9.608712870371132e-06,
-      "loss": 0.9483,
+      "learning_rate": 9.042869011187142e-06,
+      "loss": 0.8592,
       "step": 19192
     },
     {
-      "epoch": 0.5271209249951938,
+      "epoch": 0.5446367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 9.607824037117965e-06,
-      "loss": 0.8973,
+      "learning_rate": 9.041954156684282e-06,
+      "loss": 0.8726,
       "step": 19193
     },
     {
-      "epoch": 0.527148389223037,
+      "epoch": 0.5446651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 9.606935206967846e-06,
-      "loss": 0.7576,
+      "learning_rate": 9.04103931027408e-06,
+      "loss": 0.9171,
       "step": 19194
     },
     {
-      "epoch": 0.5271758534508802,
+      "epoch": 0.5446935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 9.60604637992781e-06,
-      "loss": 0.8245,
+      "learning_rate": 9.040124471964258e-06,
+      "loss": 0.8691,
       "step": 19195
     },
     {
-      "epoch": 0.5272033176787234,
+      "epoch": 0.5447219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 9.605157556004885e-06,
-      "loss": 0.9099,
+      "learning_rate": 9.039209641762552e-06,
+      "loss": 0.8309,
       "step": 19196
     },
     {
-      "epoch": 0.5272307819065667,
+      "epoch": 0.5447502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 9.604268735206107e-06,
-      "loss": 0.8928,
+      "learning_rate": 9.038294819676685e-06,
+      "loss": 0.8291,
       "step": 19197
     },
     {
-      "epoch": 0.52725824613441,
+      "epoch": 0.5447786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 9.603379917538512e-06,
-      "loss": 0.9315,
+      "learning_rate": 9.037380005714382e-06,
+      "loss": 0.913,
       "step": 19198
     },
     {
-      "epoch": 0.5272857103622531,
+      "epoch": 0.5448070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 9.602491103009127e-06,
-      "loss": 0.88,
+      "learning_rate": 9.036465199883375e-06,
+      "loss": 0.9312,
       "step": 19199
     },
     {
-      "epoch": 0.5273131745900964,
+      "epoch": 0.5448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 9.601602291624984e-06,
-      "loss": 0.8568,
+      "learning_rate": 9.035550402191386e-06,
+      "loss": 0.7952,
       "step": 19200
     },
     {
-      "epoch": 0.5273406388179397,
+      "epoch": 0.5448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 9.600713483393117e-06,
-      "loss": 0.9718,
+      "learning_rate": 9.034635612646151e-06,
+      "loss": 0.9943,
       "step": 19201
     },
     {
-      "epoch": 0.5273681030457829,
+      "epoch": 0.5448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 9.59982467832056e-06,
-      "loss": 0.8994,
+      "learning_rate": 9.033720831255391e-06,
+      "loss": 0.8314,
       "step": 19202
     },
     {
-      "epoch": 0.5273955672736261,
+      "epoch": 0.5449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 9.59893587641435e-06,
-      "loss": 0.9335,
+      "learning_rate": 9.032806058026835e-06,
+      "loss": 0.8747,
       "step": 19203
     },
     {
-      "epoch": 0.5274230315014693,
+      "epoch": 0.5449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 9.598047077681514e-06,
-      "loss": 0.8681,
+      "learning_rate": 9.03189129296821e-06,
+      "loss": 0.8282,
       "step": 19204
     },
     {
-      "epoch": 0.5274504957293126,
+      "epoch": 0.5449772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 9.59715828212908e-06,
-      "loss": 0.9707,
+      "learning_rate": 9.030976536087242e-06,
+      "loss": 0.8795,
       "step": 19205
     },
     {
-      "epoch": 0.5274779599571559,
+      "epoch": 0.545005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 9.596269489764088e-06,
-      "loss": 0.9097,
+      "learning_rate": 9.03006178739166e-06,
+      "loss": 0.8564,
       "step": 19206
     },
     {
-      "epoch": 0.527505424184999,
+      "epoch": 0.5450340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 9.595380700593571e-06,
-      "loss": 0.9329,
+      "learning_rate": 9.02914704688919e-06,
+      "loss": 0.8494,
       "step": 19207
     },
     {
-      "epoch": 0.5275328884128423,
+      "epoch": 0.5450624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 9.594491914624553e-06,
-      "loss": 0.8333,
+      "learning_rate": 9.028232314587556e-06,
+      "loss": 0.865,
       "step": 19208
     },
     {
-      "epoch": 0.5275603526406855,
+      "epoch": 0.5450908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 9.593603131864076e-06,
-      "loss": 1.0142,
+      "learning_rate": 9.027317590494488e-06,
+      "loss": 0.8082,
       "step": 19209
     },
     {
-      "epoch": 0.5275878168685287,
+      "epoch": 0.5451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 9.59271435231917e-06,
-      "loss": 0.8395,
+      "learning_rate": 9.026402874617713e-06,
+      "loss": 0.7982,
       "step": 19210
     },
     {
-      "epoch": 0.527615281096372,
+      "epoch": 0.5451475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 9.591825575996867e-06,
-      "loss": 0.86,
+      "learning_rate": 9.025488166964955e-06,
+      "loss": 0.8539,
       "step": 19211
     },
     {
-      "epoch": 0.5276427453242152,
+      "epoch": 0.5451759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 9.590936802904191e-06,
-      "loss": 0.9939,
+      "learning_rate": 9.024573467543945e-06,
+      "loss": 0.9075,
       "step": 19212
     },
     {
-      "epoch": 0.5276702095520585,
+      "epoch": 0.5452043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 9.590048033048184e-06,
-      "loss": 0.9057,
+      "learning_rate": 9.023658776362406e-06,
+      "loss": 0.8153,
       "step": 19213
     },
     {
-      "epoch": 0.5276976737799017,
+      "epoch": 0.5452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 9.589159266435876e-06,
-      "loss": 0.8608,
+      "learning_rate": 9.022744093428062e-06,
+      "loss": 0.8484,
       "step": 19214
     },
     {
-      "epoch": 0.5277251380077449,
+      "epoch": 0.5452610669693531,
       "grad_norm": 0.0,
-      "learning_rate": 9.588270503074301e-06,
-      "loss": 0.8985,
+      "learning_rate": 9.021829418748647e-06,
+      "loss": 0.8574,
       "step": 19215
     },
     {
-      "epoch": 0.5277526022355882,
+      "epoch": 0.5452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 9.587381742970493e-06,
-      "loss": 0.9534,
+      "learning_rate": 9.020914752331877e-06,
+      "loss": 0.8716,
       "step": 19216
     },
     {
-      "epoch": 0.5277800664634313,
+      "epoch": 0.5453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 9.586492986131473e-06,
-      "loss": 0.8483,
+      "learning_rate": 9.02000009418549e-06,
+      "loss": 0.7579,
       "step": 19217
     },
     {
-      "epoch": 0.5278075306912746,
+      "epoch": 0.5453461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 9.585604232564283e-06,
-      "loss": 0.9067,
+      "learning_rate": 9.019085444317204e-06,
+      "loss": 0.9003,
       "step": 19218
     },
     {
-      "epoch": 0.5278349949191179,
+      "epoch": 0.5453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 9.584715482275955e-06,
-      "loss": 0.9068,
+      "learning_rate": 9.018170802734748e-06,
+      "loss": 0.9728,
       "step": 19219
     },
     {
-      "epoch": 0.5278624591469611,
+      "epoch": 0.5454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 9.583826735273515e-06,
-      "loss": 0.9477,
+      "learning_rate": 9.017256169445847e-06,
+      "loss": 0.8507,
       "step": 19220
     },
     {
-      "epoch": 0.5278899233748043,
+      "epoch": 0.5454313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 9.582937991564e-06,
-      "loss": 0.7888,
+      "learning_rate": 9.016341544458227e-06,
+      "loss": 0.8932,
       "step": 19221
     },
     {
-      "epoch": 0.5279173876026475,
+      "epoch": 0.5454597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 9.582049251154447e-06,
-      "loss": 0.8339,
+      "learning_rate": 9.015426927779613e-06,
+      "loss": 0.8718,
       "step": 19222
     },
     {
-      "epoch": 0.5279448518304908,
+      "epoch": 0.5454880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 9.581160514051878e-06,
-      "loss": 1.0325,
+      "learning_rate": 9.014512319417735e-06,
+      "loss": 0.8837,
       "step": 19223
     },
     {
-      "epoch": 0.5279723160583341,
+      "epoch": 0.545516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 9.580271780263328e-06,
-      "loss": 0.8405,
+      "learning_rate": 9.013597719380316e-06,
+      "loss": 0.8271,
       "step": 19224
     },
     {
-      "epoch": 0.5279997802861772,
+      "epoch": 0.5455448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 9.57938304979583e-06,
-      "loss": 0.9213,
+      "learning_rate": 9.012683127675078e-06,
+      "loss": 0.9717,
       "step": 19225
     },
     {
-      "epoch": 0.5280272445140205,
+      "epoch": 0.5455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 9.578494322656419e-06,
-      "loss": 0.89,
+      "learning_rate": 9.011768544309753e-06,
+      "loss": 0.8884,
       "step": 19226
     },
     {
-      "epoch": 0.5280547087418638,
+      "epoch": 0.5456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 9.57760559885212e-06,
-      "loss": 0.8069,
+      "learning_rate": 9.010853969292064e-06,
+      "loss": 0.9561,
       "step": 19227
     },
     {
-      "epoch": 0.5280821729697069,
+      "epoch": 0.5456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 9.576716878389976e-06,
-      "loss": 0.9329,
+      "learning_rate": 9.009939402629732e-06,
+      "loss": 0.8687,
       "step": 19228
     },
     {
-      "epoch": 0.5281096371975502,
+      "epoch": 0.5456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 9.575828161277008e-06,
-      "loss": 0.8183,
+      "learning_rate": 9.009024844330489e-06,
+      "loss": 0.8799,
       "step": 19229
     },
     {
-      "epoch": 0.5281371014253934,
+      "epoch": 0.5456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 9.57493944752025e-06,
-      "loss": 0.9577,
+      "learning_rate": 9.008110294402057e-06,
+      "loss": 0.9245,
       "step": 19230
     },
     {
-      "epoch": 0.5281645656532367,
+      "epoch": 0.5457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 9.574050737126738e-06,
-      "loss": 0.9159,
+      "learning_rate": 9.00719575285216e-06,
+      "loss": 0.8914,
       "step": 19231
     },
     {
-      "epoch": 0.5281920298810799,
+      "epoch": 0.5457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 9.5731620301035e-06,
-      "loss": 0.8433,
+      "learning_rate": 9.006281219688527e-06,
+      "loss": 0.921,
       "step": 19232
     },
     {
-      "epoch": 0.5282194941089231,
+      "epoch": 0.5457718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 9.57227332645757e-06,
-      "loss": 0.9382,
+      "learning_rate": 9.005366694918882e-06,
+      "loss": 0.8666,
       "step": 19233
     },
     {
-      "epoch": 0.5282469583367664,
+      "epoch": 0.545800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 9.571384626195982e-06,
-      "loss": 0.8821,
+      "learning_rate": 9.004452178550949e-06,
+      "loss": 0.9274,
       "step": 19234
     },
     {
-      "epoch": 0.5282744225646095,
+      "epoch": 0.5458286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 9.570495929325763e-06,
-      "loss": 0.9055,
+      "learning_rate": 9.003537670592455e-06,
+      "loss": 0.8861,
       "step": 19235
     },
     {
-      "epoch": 0.5283018867924528,
+      "epoch": 0.5458569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 9.569607235853945e-06,
-      "loss": 0.8485,
+      "learning_rate": 9.002623171051118e-06,
+      "loss": 0.8442,
       "step": 19236
     },
     {
-      "epoch": 0.5283293510202961,
+      "epoch": 0.5458853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 9.568718545787561e-06,
-      "loss": 0.8165,
+      "learning_rate": 9.001708679934672e-06,
+      "loss": 0.8033,
       "step": 19237
     },
     {
-      "epoch": 0.5283568152481393,
+      "epoch": 0.5459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 9.567829859133644e-06,
-      "loss": 1.0383,
+      "learning_rate": 9.000794197250836e-06,
+      "loss": 0.8568,
       "step": 19238
     },
     {
-      "epoch": 0.5283842794759825,
+      "epoch": 0.5459421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 9.566941175899228e-06,
-      "loss": 0.854,
+      "learning_rate": 8.999879723007338e-06,
+      "loss": 0.8805,
       "step": 19239
     },
     {
-      "epoch": 0.5284117437038258,
+      "epoch": 0.5459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 9.566052496091338e-06,
-      "loss": 0.8658,
+      "learning_rate": 8.9989652572119e-06,
+      "loss": 0.8413,
       "step": 19240
     },
     {
-      "epoch": 0.528439207931669,
+      "epoch": 0.5459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 9.565163819717007e-06,
-      "loss": 0.9057,
+      "learning_rate": 8.998050799872248e-06,
+      "loss": 0.8804,
       "step": 19241
     },
     {
-      "epoch": 0.5284666721595123,
+      "epoch": 0.546027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 9.56427514678327e-06,
-      "loss": 0.8139,
+      "learning_rate": 8.997136350996104e-06,
+      "loss": 0.9562,
       "step": 19242
     },
     {
-      "epoch": 0.5284941363873554,
+      "epoch": 0.54605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 9.563386477297158e-06,
-      "loss": 0.9797,
+      "learning_rate": 8.996221910591196e-06,
+      "loss": 0.866,
       "step": 19243
     },
     {
-      "epoch": 0.5285216006151987,
+      "epoch": 0.5460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 9.562497811265699e-06,
-      "loss": 0.9209,
+      "learning_rate": 8.995307478665246e-06,
+      "loss": 0.9385,
       "step": 19244
     },
     {
-      "epoch": 0.528549064843042,
+      "epoch": 0.5461123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 9.56160914869593e-06,
-      "loss": 0.9921,
+      "learning_rate": 8.99439305522598e-06,
+      "loss": 0.8493,
       "step": 19245
     },
     {
-      "epoch": 0.5285765290708851,
+      "epoch": 0.5461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 9.560720489594875e-06,
-      "loss": 0.8138,
+      "learning_rate": 8.993478640281117e-06,
+      "loss": 0.9023,
       "step": 19246
     },
     {
-      "epoch": 0.5286039932987284,
+      "epoch": 0.5461691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 9.559831833969574e-06,
-      "loss": 0.8744,
+      "learning_rate": 8.992564233838388e-06,
+      "loss": 0.8663,
       "step": 19247
     },
     {
-      "epoch": 0.5286314575265716,
+      "epoch": 0.5461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 9.55894318182705e-06,
-      "loss": 0.8725,
+      "learning_rate": 8.991649835905516e-06,
+      "loss": 0.8561,
       "step": 19248
     },
     {
-      "epoch": 0.5286589217544149,
+      "epoch": 0.5462258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 9.55805453317434e-06,
-      "loss": 0.8401,
+      "learning_rate": 8.990735446490219e-06,
+      "loss": 0.9911,
       "step": 19249
     },
     {
-      "epoch": 0.5286863859822581,
+      "epoch": 0.5462542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 9.557165888018474e-06,
-      "loss": 0.8506,
+      "learning_rate": 8.98982106560023e-06,
+      "loss": 0.8635,
       "step": 19250
     },
     {
-      "epoch": 0.5287138502101013,
+      "epoch": 0.5462826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 9.556277246366484e-06,
-      "loss": 0.8727,
+      "learning_rate": 8.988906693243266e-06,
+      "loss": 0.881,
       "step": 19251
     },
     {
-      "epoch": 0.5287413144379446,
+      "epoch": 0.546311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 9.555388608225402e-06,
-      "loss": 0.8498,
+      "learning_rate": 8.987992329427052e-06,
+      "loss": 0.9285,
       "step": 19252
     },
     {
-      "epoch": 0.5287687786657879,
+      "epoch": 0.5463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 9.554499973602252e-06,
-      "loss": 0.8182,
+      "learning_rate": 8.987077974159312e-06,
+      "loss": 0.8428,
       "step": 19253
     },
     {
-      "epoch": 0.528796242893631,
+      "epoch": 0.5463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 9.553611342504072e-06,
-      "loss": 0.7795,
+      "learning_rate": 8.986163627447773e-06,
+      "loss": 0.9141,
       "step": 19254
     },
     {
-      "epoch": 0.5288237071214743,
+      "epoch": 0.5463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 9.552722714937896e-06,
-      "loss": 0.8155,
+      "learning_rate": 8.985249289300152e-06,
+      "loss": 0.769,
       "step": 19255
     },
     {
-      "epoch": 0.5288511713493175,
+      "epoch": 0.5464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 9.551834090910747e-06,
-      "loss": 1.051,
+      "learning_rate": 8.984334959724177e-06,
+      "loss": 0.8918,
       "step": 19256
     },
     {
-      "epoch": 0.5288786355771607,
+      "epoch": 0.5464528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 9.550945470429665e-06,
-      "loss": 0.8065,
+      "learning_rate": 8.98342063872757e-06,
+      "loss": 0.853,
       "step": 19257
     },
     {
-      "epoch": 0.528906099805004,
+      "epoch": 0.5464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 9.550056853501672e-06,
-      "loss": 0.8662,
+      "learning_rate": 8.982506326318053e-06,
+      "loss": 0.8728,
       "step": 19258
     },
     {
-      "epoch": 0.5289335640328472,
+      "epoch": 0.5465096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 9.549168240133806e-06,
-      "loss": 0.8522,
+      "learning_rate": 8.981592022503353e-06,
+      "loss": 0.8625,
       "step": 19259
     },
     {
-      "epoch": 0.5289610282606905,
+      "epoch": 0.5465380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 9.548279630333092e-06,
-      "loss": 0.8825,
+      "learning_rate": 8.980677727291192e-06,
+      "loss": 0.9317,
       "step": 19260
     },
     {
-      "epoch": 0.5289884924885336,
+      "epoch": 0.546566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 9.547391024106568e-06,
-      "loss": 0.9702,
+      "learning_rate": 8.979763440689291e-06,
+      "loss": 0.7645,
       "step": 19261
     },
     {
-      "epoch": 0.5290159567163769,
+      "epoch": 0.5465947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 9.54650242146126e-06,
-      "loss": 0.9093,
+      "learning_rate": 8.97884916270537e-06,
+      "loss": 0.8373,
       "step": 19262
     },
     {
-      "epoch": 0.5290434209442202,
+      "epoch": 0.5466231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 9.545613822404203e-06,
-      "loss": 0.9089,
+      "learning_rate": 8.977934893347158e-06,
+      "loss": 0.8727,
       "step": 19263
     },
     {
-      "epoch": 0.5290708851720634,
+      "epoch": 0.5466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 9.544725226942424e-06,
-      "loss": 0.8166,
+      "learning_rate": 8.977020632622378e-06,
+      "loss": 0.9019,
       "step": 19264
     },
     {
-      "epoch": 0.5290983493999066,
+      "epoch": 0.5466799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 9.543836635082954e-06,
-      "loss": 0.8434,
+      "learning_rate": 8.976106380538752e-06,
+      "loss": 0.8339,
       "step": 19265
     },
     {
-      "epoch": 0.5291258136277499,
+      "epoch": 0.5467082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 9.542948046832824e-06,
-      "loss": 1.0457,
+      "learning_rate": 8.975192137103997e-06,
+      "loss": 0.9397,
       "step": 19266
     },
     {
-      "epoch": 0.5291532778555931,
+      "epoch": 0.5467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 9.54205946219907e-06,
-      "loss": 0.8433,
+      "learning_rate": 8.974277902325844e-06,
+      "loss": 0.9233,
       "step": 19267
     },
     {
-      "epoch": 0.5291807420834364,
+      "epoch": 0.5467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 9.541170881188715e-06,
-      "loss": 0.6924,
+      "learning_rate": 8.973363676212011e-06,
+      "loss": 0.8583,
       "step": 19268
     },
     {
-      "epoch": 0.5292082063112795,
+      "epoch": 0.5467934165720771,
       "grad_norm": 0.0,
-      "learning_rate": 9.540282303808799e-06,
-      "loss": 0.8718,
+      "learning_rate": 8.97244945877022e-06,
+      "loss": 0.9139,
       "step": 19269
     },
     {
-      "epoch": 0.5292356705391228,
+      "epoch": 0.5468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 9.539393730066343e-06,
-      "loss": 0.9049,
+      "learning_rate": 8.971535250008196e-06,
+      "loss": 0.8462,
       "step": 19270
     },
     {
-      "epoch": 0.5292631347669661,
+      "epoch": 0.546850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 9.538505159968385e-06,
-      "loss": 0.8693,
+      "learning_rate": 8.970621049933659e-06,
+      "loss": 0.9584,
       "step": 19271
     },
     {
-      "epoch": 0.5292905989948092,
+      "epoch": 0.5468785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 9.53761659352195e-06,
-      "loss": 0.86,
+      "learning_rate": 8.969706858554331e-06,
+      "loss": 0.8848,
       "step": 19272
     },
     {
-      "epoch": 0.5293180632226525,
+      "epoch": 0.5469069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 9.536728030734072e-06,
-      "loss": 0.8252,
+      "learning_rate": 8.968792675877937e-06,
+      "loss": 0.8692,
       "step": 19273
     },
     {
-      "epoch": 0.5293455274504957,
+      "epoch": 0.5469353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 9.535839471611783e-06,
-      "loss": 0.9527,
+      "learning_rate": 8.9678785019122e-06,
+      "loss": 0.9776,
       "step": 19274
     },
     {
-      "epoch": 0.529372991678339,
+      "epoch": 0.5469636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 9.534950916162112e-06,
-      "loss": 0.9373,
+      "learning_rate": 8.966964336664835e-06,
+      "loss": 0.7894,
       "step": 19275
     },
     {
-      "epoch": 0.5294004559061822,
+      "epoch": 0.5469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 9.53406236439209e-06,
-      "loss": 0.9076,
+      "learning_rate": 8.966050180143571e-06,
+      "loss": 0.8096,
       "step": 19276
     },
     {
-      "epoch": 0.5294279201340254,
+      "epoch": 0.5470204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 9.533173816308745e-06,
-      "loss": 0.8996,
+      "learning_rate": 8.965136032356125e-06,
+      "loss": 0.8318,
       "step": 19277
     },
     {
-      "epoch": 0.5294553843618687,
+      "epoch": 0.5470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 9.532285271919108e-06,
-      "loss": 0.904,
+      "learning_rate": 8.964221893310223e-06,
+      "loss": 0.8846,
       "step": 19278
     },
     {
-      "epoch": 0.529482848589712,
+      "epoch": 0.5470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 9.531396731230215e-06,
-      "loss": 0.8082,
+      "learning_rate": 8.963307763013588e-06,
+      "loss": 0.9034,
       "step": 19279
     },
     {
-      "epoch": 0.5295103128175551,
+      "epoch": 0.547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 9.53050819424909e-06,
-      "loss": 0.8916,
+      "learning_rate": 8.962393641473936e-06,
+      "loss": 0.9773,
       "step": 19280
     },
     {
-      "epoch": 0.5295377770453984,
+      "epoch": 0.5471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 9.52961966098277e-06,
-      "loss": 0.9088,
+      "learning_rate": 8.961479528698994e-06,
+      "loss": 0.8852,
       "step": 19281
     },
     {
-      "epoch": 0.5295652412732416,
+      "epoch": 0.5471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 9.528731131438277e-06,
-      "loss": 0.8746,
+      "learning_rate": 8.96056542469648e-06,
+      "loss": 0.8569,
       "step": 19282
     },
     {
-      "epoch": 0.5295927055010848,
+      "epoch": 0.5471906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 9.527842605622649e-06,
-      "loss": 0.9886,
+      "learning_rate": 8.959651329474115e-06,
+      "loss": 0.9369,
       "step": 19283
     },
     {
-      "epoch": 0.5296201697289281,
+      "epoch": 0.5472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 9.52695408354291e-06,
-      "loss": 0.9442,
+      "learning_rate": 8.958737243039623e-06,
+      "loss": 0.9047,
       "step": 19284
     },
     {
-      "epoch": 0.5296476339567713,
+      "epoch": 0.5472474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 9.526065565206095e-06,
-      "loss": 0.8319,
+      "learning_rate": 8.957823165400727e-06,
+      "loss": 0.8506,
       "step": 19285
     },
     {
-      "epoch": 0.5296750981846146,
+      "epoch": 0.5472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 9.525177050619233e-06,
-      "loss": 0.9479,
+      "learning_rate": 8.956909096565141e-06,
+      "loss": 0.7947,
       "step": 19286
     },
     {
-      "epoch": 0.5297025624124577,
+      "epoch": 0.5473041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 9.524288539789356e-06,
-      "loss": 0.8789,
+      "learning_rate": 8.955995036540593e-06,
+      "loss": 0.8152,
       "step": 19287
     },
     {
-      "epoch": 0.529730026640301,
+      "epoch": 0.5473325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 9.52340003272349e-06,
-      "loss": 0.8146,
+      "learning_rate": 8.955080985334803e-06,
+      "loss": 0.7698,
       "step": 19288
     },
     {
-      "epoch": 0.5297574908681443,
+      "epoch": 0.547360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 9.522511529428668e-06,
-      "loss": 0.8187,
+      "learning_rate": 8.954166942955488e-06,
+      "loss": 0.8356,
       "step": 19289
     },
     {
-      "epoch": 0.5297849550959874,
+      "epoch": 0.54738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 9.521623029911917e-06,
-      "loss": 0.947,
+      "learning_rate": 8.953252909410374e-06,
+      "loss": 0.8133,
       "step": 19290
     },
     {
-      "epoch": 0.5298124193238307,
+      "epoch": 0.5474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 9.520734534180274e-06,
-      "loss": 0.9342,
+      "learning_rate": 8.952338884707182e-06,
+      "loss": 0.8437,
       "step": 19291
     },
     {
-      "epoch": 0.529839883551674,
+      "epoch": 0.5474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 9.519846042240763e-06,
-      "loss": 0.7854,
+      "learning_rate": 8.951424868853623e-06,
+      "loss": 0.901,
       "step": 19292
     },
     {
-      "epoch": 0.5298673477795172,
+      "epoch": 0.5474744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 9.51895755410042e-06,
-      "loss": 0.8931,
+      "learning_rate": 8.950510861857429e-06,
+      "loss": 0.8284,
       "step": 19293
     },
     {
-      "epoch": 0.5298948120073604,
+      "epoch": 0.5475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 9.518069069766267e-06,
-      "loss": 0.8426,
+      "learning_rate": 8.949596863726319e-06,
+      "loss": 0.952,
       "step": 19294
     },
     {
-      "epoch": 0.5299222762352036,
+      "epoch": 0.5475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 9.517180589245339e-06,
-      "loss": 0.8578,
+      "learning_rate": 8.948682874468011e-06,
+      "loss": 0.8351,
       "step": 19295
     },
     {
-      "epoch": 0.5299497404630469,
+      "epoch": 0.5475595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 9.516292112544663e-06,
-      "loss": 0.8843,
+      "learning_rate": 8.947768894090224e-06,
+      "loss": 0.9402,
       "step": 19296
     },
     {
-      "epoch": 0.5299772046908902,
+      "epoch": 0.5475879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 9.515403639671274e-06,
-      "loss": 0.9247,
+      "learning_rate": 8.946854922600683e-06,
+      "loss": 0.9059,
       "step": 19297
     },
     {
-      "epoch": 0.5300046689187333,
+      "epoch": 0.5476163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 9.514515170632202e-06,
-      "loss": 0.8842,
+      "learning_rate": 8.945940960007106e-06,
+      "loss": 0.8542,
       "step": 19298
     },
     {
-      "epoch": 0.5300321331465766,
+      "epoch": 0.5476447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 9.513626705434469e-06,
-      "loss": 0.8745,
+      "learning_rate": 8.945027006317211e-06,
+      "loss": 0.9167,
       "step": 19299
     },
     {
-      "epoch": 0.5300595973744198,
+      "epoch": 0.5476730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 9.512738244085112e-06,
-      "loss": 0.822,
+      "learning_rate": 8.944113061538721e-06,
+      "loss": 0.8936,
       "step": 19300
     },
     {
-      "epoch": 0.530087061602263,
+      "epoch": 0.5477014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 9.511849786591158e-06,
-      "loss": 0.9581,
+      "learning_rate": 8.943199125679357e-06,
+      "loss": 0.8932,
       "step": 19301
     },
     {
-      "epoch": 0.5301145258301063,
+      "epoch": 0.5477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 9.510961332959637e-06,
-      "loss": 0.8567,
+      "learning_rate": 8.942285198746836e-06,
+      "loss": 0.9318,
       "step": 19302
     },
     {
-      "epoch": 0.5301419900579495,
+      "epoch": 0.5477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 9.51007288319758e-06,
-      "loss": 0.9628,
+      "learning_rate": 8.941371280748878e-06,
+      "loss": 0.9431,
       "step": 19303
     },
     {
-      "epoch": 0.5301694542857928,
+      "epoch": 0.5477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 9.50918443731202e-06,
-      "loss": 0.9073,
+      "learning_rate": 8.940457371693208e-06,
+      "loss": 0.8829,
       "step": 19304
     },
     {
-      "epoch": 0.530196918513636,
+      "epoch": 0.5478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 9.508295995309978e-06,
-      "loss": 0.9869,
+      "learning_rate": 8.93954347158754e-06,
+      "loss": 0.8886,
       "step": 19305
     },
     {
-      "epoch": 0.5302243827414792,
+      "epoch": 0.5478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 9.507407557198489e-06,
-      "loss": 0.9417,
+      "learning_rate": 8.938629580439595e-06,
+      "loss": 0.9369,
       "step": 19306
     },
     {
-      "epoch": 0.5302518469693225,
+      "epoch": 0.5478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 9.506519122984583e-06,
-      "loss": 0.8508,
+      "learning_rate": 8.937715698257096e-06,
+      "loss": 0.8761,
       "step": 19307
     },
     {
-      "epoch": 0.5302793111971656,
+      "epoch": 0.547900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 9.505630692675289e-06,
-      "loss": 0.9156,
+      "learning_rate": 8.936801825047755e-06,
+      "loss": 0.8468,
       "step": 19308
     },
     {
-      "epoch": 0.5303067754250089,
+      "epoch": 0.5479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 9.504742266277633e-06,
-      "loss": 1.006,
+      "learning_rate": 8.935887960819301e-06,
+      "loss": 0.9435,
       "step": 19309
     },
     {
-      "epoch": 0.5303342396528522,
+      "epoch": 0.5479568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 9.503853843798656e-06,
-      "loss": 0.9243,
+      "learning_rate": 8.934974105579449e-06,
+      "loss": 0.8627,
       "step": 19310
     },
     {
-      "epoch": 0.5303617038806954,
+      "epoch": 0.5479852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 9.502965425245372e-06,
-      "loss": 0.895,
+      "learning_rate": 8.934060259335919e-06,
+      "loss": 0.9064,
       "step": 19311
     },
     {
-      "epoch": 0.5303891681085386,
+      "epoch": 0.5480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 9.502077010624823e-06,
-      "loss": 0.8737,
+      "learning_rate": 8.93314642209643e-06,
+      "loss": 0.8775,
       "step": 19312
     },
     {
-      "epoch": 0.5304166323363818,
+      "epoch": 0.5480419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 9.50118859994403e-06,
-      "loss": 0.8969,
+      "learning_rate": 8.932232593868699e-06,
+      "loss": 0.7481,
       "step": 19313
     },
     {
-      "epoch": 0.5304440965642251,
+      "epoch": 0.5480703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 9.500300193210027e-06,
-      "loss": 0.9359,
+      "learning_rate": 8.93131877466045e-06,
+      "loss": 0.7797,
       "step": 19314
     },
     {
-      "epoch": 0.5304715607920684,
+      "epoch": 0.5480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 9.499411790429844e-06,
-      "loss": 0.9147,
+      "learning_rate": 8.9304049644794e-06,
+      "loss": 0.9434,
       "step": 19315
     },
     {
-      "epoch": 0.5304990250199115,
+      "epoch": 0.5481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 9.498523391610511e-06,
-      "loss": 0.8651,
+      "learning_rate": 8.929491163333263e-06,
+      "loss": 0.8517,
       "step": 19316
     },
     {
-      "epoch": 0.5305264892477548,
+      "epoch": 0.5481555051078321,
       "grad_norm": 0.0,
-      "learning_rate": 9.497634996759051e-06,
-      "loss": 0.8985,
+      "learning_rate": 8.928577371229767e-06,
+      "loss": 0.8674,
       "step": 19317
     },
     {
-      "epoch": 0.5305539534755981,
+      "epoch": 0.5481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 9.496746605882497e-06,
-      "loss": 0.8977,
+      "learning_rate": 8.927663588176626e-06,
+      "loss": 0.8688,
       "step": 19318
     },
     {
-      "epoch": 0.5305814177034412,
+      "epoch": 0.5482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 9.495858218987882e-06,
-      "loss": 0.8698,
+      "learning_rate": 8.926749814181555e-06,
+      "loss": 0.8461,
       "step": 19319
     },
     {
-      "epoch": 0.5306088819312845,
+      "epoch": 0.5482406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 9.494969836082229e-06,
-      "loss": 0.9131,
+      "learning_rate": 8.92583604925228e-06,
+      "loss": 0.9,
       "step": 19320
     },
     {
-      "epoch": 0.5306363461591277,
+      "epoch": 0.5482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 9.49408145717257e-06,
-      "loss": 0.7542,
+      "learning_rate": 8.924922293396514e-06,
+      "loss": 0.8752,
       "step": 19321
     },
     {
-      "epoch": 0.530663810386971,
+      "epoch": 0.5482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 9.493193082265939e-06,
-      "loss": 0.8212,
+      "learning_rate": 8.924008546621977e-06,
+      "loss": 0.9302,
       "step": 19322
     },
     {
-      "epoch": 0.5306912746148142,
+      "epoch": 0.5483257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 9.492304711369357e-06,
-      "loss": 0.8628,
+      "learning_rate": 8.923094808936388e-06,
+      "loss": 0.923,
       "step": 19323
     },
     {
-      "epoch": 0.5307187388426574,
+      "epoch": 0.5483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 9.491416344489859e-06,
-      "loss": 0.7747,
+      "learning_rate": 8.922181080347465e-06,
+      "loss": 0.8325,
       "step": 19324
     },
     {
-      "epoch": 0.5307462030705007,
+      "epoch": 0.5483825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 9.490527981634469e-06,
-      "loss": 0.8909,
+      "learning_rate": 8.92126736086293e-06,
+      "loss": 0.8349,
       "step": 19325
     },
     {
-      "epoch": 0.5307736672983439,
+      "epoch": 0.548410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 9.489639622810219e-06,
-      "loss": 0.8797,
+      "learning_rate": 8.920353650490499e-06,
+      "loss": 0.9031,
       "step": 19326
     },
     {
-      "epoch": 0.5308011315261871,
+      "epoch": 0.5484392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 9.48875126802414e-06,
-      "loss": 0.9689,
+      "learning_rate": 8.919439949237885e-06,
+      "loss": 0.908,
       "step": 19327
     },
     {
-      "epoch": 0.5308285957540304,
+      "epoch": 0.5484676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 9.487862917283263e-06,
-      "loss": 0.9132,
+      "learning_rate": 8.918526257112815e-06,
+      "loss": 0.9939,
       "step": 19328
     },
     {
-      "epoch": 0.5308560599818736,
+      "epoch": 0.5484960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 9.486974570594605e-06,
-      "loss": 0.7828,
+      "learning_rate": 8.917612574123e-06,
+      "loss": 0.8372,
       "step": 19329
     },
     {
-      "epoch": 0.5308835242097169,
+      "epoch": 0.5485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 9.486086227965207e-06,
-      "loss": 0.9895,
+      "learning_rate": 8.916698900276158e-06,
+      "loss": 1.0022,
       "step": 19330
     },
     {
-      "epoch": 0.5309109884375601,
+      "epoch": 0.5485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 9.485197889402093e-06,
-      "loss": 0.9223,
+      "learning_rate": 8.915785235580013e-06,
+      "loss": 0.817,
       "step": 19331
     },
     {
-      "epoch": 0.5309384526654033,
+      "epoch": 0.5485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 9.484309554912291e-06,
-      "loss": 0.9191,
+      "learning_rate": 8.914871580042278e-06,
+      "loss": 0.7761,
       "step": 19332
     },
     {
-      "epoch": 0.5309659168932466,
+      "epoch": 0.5486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 9.483421224502834e-06,
-      "loss": 0.8876,
+      "learning_rate": 8.913957933670669e-06,
+      "loss": 0.9235,
       "step": 19333
     },
     {
-      "epoch": 0.5309933811210897,
+      "epoch": 0.5486379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 9.482532898180751e-06,
-      "loss": 0.846,
+      "learning_rate": 8.913044296472908e-06,
+      "loss": 0.9623,
       "step": 19334
     },
     {
-      "epoch": 0.531020845348933,
+      "epoch": 0.5486662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 9.481644575953062e-06,
-      "loss": 0.902,
+      "learning_rate": 8.912130668456712e-06,
+      "loss": 0.7334,
       "step": 19335
     },
     {
-      "epoch": 0.5310483095767763,
+      "epoch": 0.548694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 9.480756257826807e-06,
-      "loss": 0.8416,
+      "learning_rate": 8.911217049629795e-06,
+      "loss": 0.9263,
       "step": 19336
     },
     {
-      "epoch": 0.5310757738046195,
+      "epoch": 0.5487230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 9.479867943809006e-06,
-      "loss": 0.919,
+      "learning_rate": 8.910303439999877e-06,
+      "loss": 0.7969,
       "step": 19337
     },
     {
-      "epoch": 0.5311032380324627,
+      "epoch": 0.5487514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 9.478979633906692e-06,
-      "loss": 0.8255,
+      "learning_rate": 8.909389839574676e-06,
+      "loss": 0.7548,
       "step": 19338
     },
     {
-      "epoch": 0.5311307022603059,
+      "epoch": 0.5487797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 9.478091328126895e-06,
-      "loss": 0.9097,
+      "learning_rate": 8.908476248361904e-06,
+      "loss": 0.8854,
       "step": 19339
     },
     {
-      "epoch": 0.5311581664881492,
+      "epoch": 0.5488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 9.477203026476643e-06,
-      "loss": 0.8895,
+      "learning_rate": 8.907562666369283e-06,
+      "loss": 0.8557,
       "step": 19340
     },
     {
-      "epoch": 0.5311856307159925,
+      "epoch": 0.5488365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 9.476314728962959e-06,
-      "loss": 0.9563,
+      "learning_rate": 8.906649093604533e-06,
+      "loss": 0.8012,
       "step": 19341
     },
     {
-      "epoch": 0.5312130949438356,
+      "epoch": 0.5488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 9.475426435592877e-06,
-      "loss": 0.9353,
+      "learning_rate": 8.905735530075367e-06,
+      "loss": 0.7417,
       "step": 19342
     },
     {
-      "epoch": 0.5312405591716789,
+      "epoch": 0.5488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 9.474538146373425e-06,
-      "loss": 0.9336,
+      "learning_rate": 8.904821975789501e-06,
+      "loss": 0.854,
       "step": 19343
     },
     {
-      "epoch": 0.5312680233995222,
+      "epoch": 0.548921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 9.473649861311628e-06,
-      "loss": 0.8875,
+      "learning_rate": 8.903908430754651e-06,
+      "loss": 0.8503,
       "step": 19344
     },
     {
-      "epoch": 0.5312954876273653,
+      "epoch": 0.548950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 9.47276158041452e-06,
-      "loss": 0.8692,
+      "learning_rate": 8.902994894978539e-06,
+      "loss": 0.7797,
       "step": 19345
     },
     {
-      "epoch": 0.5313229518552086,
+      "epoch": 0.5489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 9.47187330368913e-06,
-      "loss": 0.9072,
+      "learning_rate": 8.902081368468879e-06,
+      "loss": 0.7214,
       "step": 19346
     },
     {
-      "epoch": 0.5313504160830518,
+      "epoch": 0.5490068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 9.47098503114248e-06,
-      "loss": 0.8107,
+      "learning_rate": 8.901167851233384e-06,
+      "loss": 0.8075,
       "step": 19347
     },
     {
-      "epoch": 0.5313778803108951,
+      "epoch": 0.5490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 9.470096762781597e-06,
-      "loss": 0.9667,
+      "learning_rate": 8.900254343279774e-06,
+      "loss": 0.8996,
       "step": 19348
     },
     {
-      "epoch": 0.5314053445387383,
+      "epoch": 0.5490635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 9.469208498613516e-06,
-      "loss": 0.8987,
+      "learning_rate": 8.899340844615767e-06,
+      "loss": 0.9286,
       "step": 19349
     },
     {
-      "epoch": 0.5314328087665815,
+      "epoch": 0.5490919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 9.468320238645263e-06,
-      "loss": 0.921,
+      "learning_rate": 8.898427355249076e-06,
+      "loss": 0.9009,
       "step": 19350
     },
     {
-      "epoch": 0.5314602729944248,
+      "epoch": 0.5491203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 9.467431982883869e-06,
-      "loss": 0.9042,
+      "learning_rate": 8.897513875187419e-06,
+      "loss": 0.8497,
       "step": 19351
     },
     {
-      "epoch": 0.5314877372222679,
+      "epoch": 0.5491486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 9.46654373133636e-06,
-      "loss": 1.0161,
+      "learning_rate": 8.896600404438512e-06,
+      "loss": 0.8545,
       "step": 19352
     },
     {
-      "epoch": 0.5315152014501112,
+      "epoch": 0.5491770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 9.465655484009761e-06,
-      "loss": 0.8776,
+      "learning_rate": 8.89568694301007e-06,
+      "loss": 0.8658,
       "step": 19353
     },
     {
-      "epoch": 0.5315426656779545,
+      "epoch": 0.549205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 9.464767240911101e-06,
-      "loss": 0.9407,
+      "learning_rate": 8.894773490909806e-06,
+      "loss": 0.889,
       "step": 19354
     },
     {
-      "epoch": 0.5315701299057977,
+      "epoch": 0.549233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 9.463879002047414e-06,
-      "loss": 0.897,
+      "learning_rate": 8.893860048145446e-06,
+      "loss": 1.0027,
       "step": 19355
     },
     {
-      "epoch": 0.5315975941336409,
+      "epoch": 0.5492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 9.462990767425721e-06,
-      "loss": 0.7011,
+      "learning_rate": 8.892946614724698e-06,
+      "loss": 0.8431,
       "step": 19356
     },
     {
-      "epoch": 0.5316250583614842,
+      "epoch": 0.5492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 9.462102537053057e-06,
-      "loss": 0.9754,
+      "learning_rate": 8.892033190655281e-06,
+      "loss": 0.9206,
       "step": 19357
     },
     {
-      "epoch": 0.5316525225893274,
+      "epoch": 0.5493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 9.461214310936442e-06,
-      "loss": 0.8319,
+      "learning_rate": 8.891119775944908e-06,
+      "loss": 0.8974,
       "step": 19358
     },
     {
-      "epoch": 0.5316799868171707,
+      "epoch": 0.5493473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 9.46032608908291e-06,
-      "loss": 0.8852,
+      "learning_rate": 8.890206370601299e-06,
+      "loss": 0.8103,
       "step": 19359
     },
     {
-      "epoch": 0.5317074510450138,
+      "epoch": 0.5493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 9.459437871499485e-06,
-      "loss": 0.8775,
+      "learning_rate": 8.889292974632162e-06,
+      "loss": 0.9094,
       "step": 19360
     },
     {
-      "epoch": 0.5317349152728571,
+      "epoch": 0.5494040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 9.458549658193198e-06,
-      "loss": 0.907,
+      "learning_rate": 8.888379588045221e-06,
+      "loss": 0.888,
       "step": 19361
     },
     {
-      "epoch": 0.5317623795007004,
+      "epoch": 0.5494324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 9.457661449171075e-06,
-      "loss": 0.8096,
+      "learning_rate": 8.887466210848188e-06,
+      "loss": 0.8347,
       "step": 19362
     },
     {
-      "epoch": 0.5317898437285435,
+      "epoch": 0.549460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 9.456773244440148e-06,
-      "loss": 0.8187,
+      "learning_rate": 8.886552843048773e-06,
+      "loss": 0.9486,
       "step": 19363
     },
     {
-      "epoch": 0.5318173079563868,
+      "epoch": 0.5494892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 9.455885044007442e-06,
-      "loss": 0.815,
+      "learning_rate": 8.8856394846547e-06,
+      "loss": 0.7573,
       "step": 19364
     },
     {
-      "epoch": 0.53184477218423,
+      "epoch": 0.5495175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 9.45499684787998e-06,
-      "loss": 0.8526,
+      "learning_rate": 8.884726135673681e-06,
+      "loss": 0.8349,
       "step": 19365
     },
     {
-      "epoch": 0.5318722364120733,
+      "epoch": 0.5495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 9.454108656064794e-06,
-      "loss": 0.8681,
+      "learning_rate": 8.883812796113429e-06,
+      "loss": 0.8576,
       "step": 19366
     },
     {
-      "epoch": 0.5318997006399165,
+      "epoch": 0.5495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 9.453220468568914e-06,
-      "loss": 0.8929,
+      "learning_rate": 8.882899465981662e-06,
+      "loss": 0.9069,
       "step": 19367
     },
     {
-      "epoch": 0.5319271648677597,
+      "epoch": 0.5496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 9.452332285399364e-06,
-      "loss": 0.9008,
+      "learning_rate": 8.881986145286092e-06,
+      "loss": 0.8468,
       "step": 19368
     },
     {
-      "epoch": 0.531954629095603,
+      "epoch": 0.5496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 9.451444106563176e-06,
-      "loss": 0.8972,
+      "learning_rate": 8.881072834034433e-06,
+      "loss": 0.8075,
       "step": 19369
     },
     {
-      "epoch": 0.5319820933234463,
+      "epoch": 0.5496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 9.450555932067372e-06,
-      "loss": 0.8946,
+      "learning_rate": 8.880159532234404e-06,
+      "loss": 0.9739,
       "step": 19370
     },
     {
-      "epoch": 0.5320095575512894,
+      "epoch": 0.5496878547105561,
       "grad_norm": 0.0,
-      "learning_rate": 9.449667761918984e-06,
-      "loss": 0.7762,
+      "learning_rate": 8.879246239893717e-06,
+      "loss": 0.7743,
       "step": 19371
     },
     {
-      "epoch": 0.5320370217791327,
+      "epoch": 0.5497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 9.448779596125037e-06,
-      "loss": 0.8253,
+      "learning_rate": 8.87833295702009e-06,
+      "loss": 0.8541,
       "step": 19372
     },
     {
-      "epoch": 0.5320644860069759,
+      "epoch": 0.549744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 9.447891434692557e-06,
-      "loss": 0.9696,
+      "learning_rate": 8.877419683621233e-06,
+      "loss": 0.9797,
       "step": 19373
     },
     {
-      "epoch": 0.5320919502348191,
+      "epoch": 0.5497729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 9.447003277628575e-06,
-      "loss": 0.9717,
+      "learning_rate": 8.876506419704863e-06,
+      "loss": 0.8545,
       "step": 19374
     },
     {
-      "epoch": 0.5321194144626624,
+      "epoch": 0.5498013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 9.446115124940121e-06,
-      "loss": 0.8808,
+      "learning_rate": 8.875593165278694e-06,
+      "loss": 0.8317,
       "step": 19375
     },
     {
-      "epoch": 0.5321468786905056,
+      "epoch": 0.5498297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 9.445226976634218e-06,
-      "loss": 1.008,
+      "learning_rate": 8.874679920350441e-06,
+      "loss": 0.875,
       "step": 19376
     },
     {
-      "epoch": 0.5321743429183489,
+      "epoch": 0.5498581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 9.444338832717892e-06,
-      "loss": 0.8609,
+      "learning_rate": 8.873766684927814e-06,
+      "loss": 0.9749,
       "step": 19377
     },
     {
-      "epoch": 0.532201807146192,
+      "epoch": 0.5498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 9.44345069319817e-06,
-      "loss": 0.8834,
+      "learning_rate": 8.872853459018534e-06,
+      "loss": 0.8756,
       "step": 19378
     },
     {
-      "epoch": 0.5322292713740353,
+      "epoch": 0.5499148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 9.442562558082085e-06,
-      "loss": 0.8788,
+      "learning_rate": 8.871940242630311e-06,
+      "loss": 0.8355,
       "step": 19379
     },
     {
-      "epoch": 0.5322567356018786,
+      "epoch": 0.5499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 9.441674427376659e-06,
-      "loss": 0.791,
+      "learning_rate": 8.871027035770856e-06,
+      "loss": 0.8626,
       "step": 19380
     },
     {
-      "epoch": 0.5322841998297217,
+      "epoch": 0.5499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 9.440786301088926e-06,
-      "loss": 0.7552,
+      "learning_rate": 8.87011383844789e-06,
+      "loss": 0.8741,
       "step": 19381
     },
     {
-      "epoch": 0.532311664057565,
+      "epoch": 0.55,
       "grad_norm": 0.0,
-      "learning_rate": 9.439898179225904e-06,
-      "loss": 0.8804,
+      "learning_rate": 8.869200650669122e-06,
+      "loss": 0.8893,
       "step": 19382
     },
     {
-      "epoch": 0.5323391282854083,
+      "epoch": 0.5500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 9.439010061794626e-06,
-      "loss": 0.8936,
+      "learning_rate": 8.868287472442268e-06,
+      "loss": 0.9266,
       "step": 19383
     },
     {
-      "epoch": 0.5323665925132515,
+      "epoch": 0.5500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 9.438121948802116e-06,
-      "loss": 0.955,
+      "learning_rate": 8.867374303775039e-06,
+      "loss": 0.8613,
       "step": 19384
     },
     {
-      "epoch": 0.5323940567410947,
+      "epoch": 0.5500851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 9.437233840255402e-06,
-      "loss": 0.8662,
+      "learning_rate": 8.866461144675147e-06,
+      "loss": 0.9244,
       "step": 19385
     },
     {
-      "epoch": 0.5324215209689379,
+      "epoch": 0.5501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 9.436345736161514e-06,
-      "loss": 0.8625,
+      "learning_rate": 8.865547995150314e-06,
+      "loss": 0.9906,
       "step": 19386
     },
     {
-      "epoch": 0.5324489851967812,
+      "epoch": 0.5501418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 9.43545763652748e-06,
-      "loss": 0.8319,
+      "learning_rate": 8.864634855208248e-06,
+      "loss": 0.9178,
       "step": 19387
     },
     {
-      "epoch": 0.5324764494246245,
+      "epoch": 0.5501702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 9.43456954136032e-06,
-      "loss": 0.9055,
+      "learning_rate": 8.86372172485666e-06,
+      "loss": 0.8911,
       "step": 19388
     },
     {
-      "epoch": 0.5325039136524676,
+      "epoch": 0.5501986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 9.433681450667063e-06,
-      "loss": 0.8048,
+      "learning_rate": 8.862808604103267e-06,
+      "loss": 0.7946,
       "step": 19389
     },
     {
-      "epoch": 0.5325313778803109,
+      "epoch": 0.5502270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 9.43279336445474e-06,
-      "loss": 0.866,
+      "learning_rate": 8.861895492955781e-06,
+      "loss": 0.946,
       "step": 19390
     },
     {
-      "epoch": 0.5325588421081542,
+      "epoch": 0.550255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 9.431905282730376e-06,
-      "loss": 0.9927,
+      "learning_rate": 8.860982391421914e-06,
+      "loss": 0.8894,
       "step": 19391
     },
     {
-      "epoch": 0.5325863063359974,
+      "epoch": 0.550283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 9.431017205500996e-06,
-      "loss": 0.9571,
+      "learning_rate": 8.86006929950938e-06,
+      "loss": 0.8651,
       "step": 19392
     },
     {
-      "epoch": 0.5326137705638406,
+      "epoch": 0.5503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 9.430129132773632e-06,
-      "loss": 0.8367,
+      "learning_rate": 8.859156217225893e-06,
+      "loss": 0.884,
       "step": 19393
     },
     {
-      "epoch": 0.5326412347916838,
+      "epoch": 0.5503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 9.429241064555303e-06,
-      "loss": 0.8838,
+      "learning_rate": 8.858243144579163e-06,
+      "loss": 0.9277,
       "step": 19394
     },
     {
-      "epoch": 0.5326686990195271,
+      "epoch": 0.5503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 9.428353000853043e-06,
-      "loss": 0.9012,
+      "learning_rate": 8.857330081576906e-06,
+      "loss": 0.8349,
       "step": 19395
     },
     {
-      "epoch": 0.5326961632473703,
+      "epoch": 0.5503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 9.427464941673872e-06,
-      "loss": 0.8982,
+      "learning_rate": 8.856417028226833e-06,
+      "loss": 0.8487,
       "step": 19396
     },
     {
-      "epoch": 0.5327236274752135,
+      "epoch": 0.5504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 9.426576887024822e-06,
-      "loss": 0.8558,
+      "learning_rate": 8.855503984536656e-06,
+      "loss": 0.9048,
       "step": 19397
     },
     {
-      "epoch": 0.5327510917030568,
+      "epoch": 0.5504540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 9.425688836912917e-06,
-      "loss": 0.9055,
+      "learning_rate": 8.85459095051409e-06,
+      "loss": 0.8393,
       "step": 19398
     },
     {
-      "epoch": 0.5327785559309,
+      "epoch": 0.5504824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 9.424800791345186e-06,
-      "loss": 0.8938,
+      "learning_rate": 8.853677926166845e-06,
+      "loss": 0.8397,
       "step": 19399
     },
     {
-      "epoch": 0.5328060201587432,
+      "epoch": 0.5505107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 9.423912750328655e-06,
-      "loss": 0.9318,
+      "learning_rate": 8.85276491150263e-06,
+      "loss": 0.8461,
       "step": 19400
     },
     {
-      "epoch": 0.5328334843865865,
+      "epoch": 0.5505391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 9.423024713870345e-06,
-      "loss": 0.9481,
+      "learning_rate": 8.851851906529163e-06,
+      "loss": 0.8869,
       "step": 19401
     },
     {
-      "epoch": 0.5328609486144297,
+      "epoch": 0.5505675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 9.42213668197729e-06,
-      "loss": 0.8934,
+      "learning_rate": 8.850938911254156e-06,
+      "loss": 0.8878,
       "step": 19402
     },
     {
-      "epoch": 0.532888412842273,
+      "epoch": 0.5505959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 9.421248654656513e-06,
-      "loss": 0.8637,
+      "learning_rate": 8.850025925685322e-06,
+      "loss": 0.937,
       "step": 19403
     },
     {
-      "epoch": 0.5329158770701162,
+      "epoch": 0.5506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 9.420360631915038e-06,
-      "loss": 0.8431,
+      "learning_rate": 8.849112949830368e-06,
+      "loss": 0.7959,
       "step": 19404
     },
     {
-      "epoch": 0.5329433412979594,
+      "epoch": 0.5506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 9.419472613759901e-06,
-      "loss": 0.7591,
+      "learning_rate": 8.84819998369701e-06,
+      "loss": 0.968,
       "step": 19405
     },
     {
-      "epoch": 0.5329708055258027,
+      "epoch": 0.5506810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 9.418584600198117e-06,
-      "loss": 0.9005,
+      "learning_rate": 8.847287027292959e-06,
+      "loss": 0.8059,
       "step": 19406
     },
     {
-      "epoch": 0.5329982697536458,
+      "epoch": 0.5507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 9.417696591236717e-06,
-      "loss": 0.8336,
+      "learning_rate": 8.846374080625923e-06,
+      "loss": 0.858,
       "step": 19407
     },
     {
-      "epoch": 0.5330257339814891,
+      "epoch": 0.5507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 9.416808586882727e-06,
-      "loss": 0.8675,
+      "learning_rate": 8.84546114370362e-06,
+      "loss": 0.7984,
       "step": 19408
     },
     {
-      "epoch": 0.5330531982093324,
+      "epoch": 0.5507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 9.415920587143173e-06,
-      "loss": 0.9172,
+      "learning_rate": 8.844548216533759e-06,
+      "loss": 0.949,
       "step": 19409
     },
     {
-      "epoch": 0.5330806624371756,
+      "epoch": 0.550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 9.415032592025082e-06,
-      "loss": 0.8949,
+      "learning_rate": 8.843635299124051e-06,
+      "loss": 0.9501,
       "step": 19410
     },
     {
-      "epoch": 0.5331081266650188,
+      "epoch": 0.5508229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 9.414144601535481e-06,
-      "loss": 0.9019,
+      "learning_rate": 8.842722391482205e-06,
+      "loss": 0.7955,
       "step": 19411
     },
     {
-      "epoch": 0.533135590892862,
+      "epoch": 0.5508513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 9.413256615681395e-06,
-      "loss": 0.7769,
+      "learning_rate": 8.841809493615938e-06,
+      "loss": 0.7487,
       "step": 19412
     },
     {
-      "epoch": 0.5331630551207053,
+      "epoch": 0.5508796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 9.412368634469849e-06,
-      "loss": 0.8683,
+      "learning_rate": 8.840896605532957e-06,
+      "loss": 0.9536,
       "step": 19413
     },
     {
-      "epoch": 0.5331905193485486,
+      "epoch": 0.5509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 9.411480657907868e-06,
-      "loss": 0.9203,
+      "learning_rate": 8.839983727240972e-06,
+      "loss": 0.7586,
       "step": 19414
     },
     {
-      "epoch": 0.5332179835763917,
+      "epoch": 0.5509364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 9.410592686002483e-06,
-      "loss": 0.8697,
+      "learning_rate": 8.839070858747697e-06,
+      "loss": 0.8847,
       "step": 19415
     },
     {
-      "epoch": 0.533245447804235,
+      "epoch": 0.5509648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 9.409704718760719e-06,
-      "loss": 0.9259,
+      "learning_rate": 8.838158000060845e-06,
+      "loss": 0.8988,
       "step": 19416
     },
     {
-      "epoch": 0.5332729120320783,
+      "epoch": 0.5509931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 9.408816756189595e-06,
-      "loss": 0.8945,
+      "learning_rate": 8.837245151188126e-06,
+      "loss": 0.7838,
       "step": 19417
     },
     {
-      "epoch": 0.5333003762599214,
+      "epoch": 0.5510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 9.407928798296142e-06,
-      "loss": 0.9284,
+      "learning_rate": 8.836332312137246e-06,
+      "loss": 0.9364,
       "step": 19418
     },
     {
-      "epoch": 0.5333278404877647,
+      "epoch": 0.551049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 9.407040845087388e-06,
-      "loss": 0.9227,
+      "learning_rate": 8.835419482915922e-06,
+      "loss": 0.8747,
       "step": 19419
     },
     {
-      "epoch": 0.5333553047156079,
+      "epoch": 0.5510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 9.406152896570355e-06,
-      "loss": 0.9769,
+      "learning_rate": 8.834506663531861e-06,
+      "loss": 0.889,
       "step": 19420
     },
     {
-      "epoch": 0.5333827689434512,
+      "epoch": 0.5511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 9.40526495275207e-06,
-      "loss": 0.7832,
+      "learning_rate": 8.833593853992773e-06,
+      "loss": 0.813,
       "step": 19421
     },
     {
-      "epoch": 0.5334102331712944,
+      "epoch": 0.5511350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 9.404377013639563e-06,
-      "loss": 0.9162,
+      "learning_rate": 8.832681054306373e-06,
+      "loss": 0.8358,
       "step": 19422
     },
     {
-      "epoch": 0.5334376973991376,
+      "epoch": 0.5511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 9.403489079239853e-06,
-      "loss": 0.827,
+      "learning_rate": 8.831768264480368e-06,
+      "loss": 0.937,
       "step": 19423
     },
     {
-      "epoch": 0.5334651616269809,
+      "epoch": 0.5511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 9.402601149559969e-06,
-      "loss": 0.8818,
+      "learning_rate": 8.830855484522467e-06,
+      "loss": 0.9424,
       "step": 19424
     },
     {
-      "epoch": 0.533492625854824,
+      "epoch": 0.5512202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 9.401713224606934e-06,
-      "loss": 0.8813,
+      "learning_rate": 8.829942714440385e-06,
+      "loss": 0.8913,
       "step": 19425
     },
     {
-      "epoch": 0.5335200900826673,
+      "epoch": 0.5512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 9.400825304387776e-06,
-      "loss": 0.919,
+      "learning_rate": 8.829029954241828e-06,
+      "loss": 0.9424,
       "step": 19426
     },
     {
-      "epoch": 0.5335475543105106,
+      "epoch": 0.5512769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 9.399937388909522e-06,
-      "loss": 0.8311,
+      "learning_rate": 8.828117203934506e-06,
+      "loss": 0.9242,
       "step": 19427
     },
     {
-      "epoch": 0.5335750185383538,
+      "epoch": 0.551305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 9.3990494781792e-06,
-      "loss": 1.029,
+      "learning_rate": 8.827204463526133e-06,
+      "loss": 0.8162,
       "step": 19428
     },
     {
-      "epoch": 0.533602482766197,
+      "epoch": 0.5513337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 9.398161572203825e-06,
-      "loss": 0.8387,
+      "learning_rate": 8.826291733024417e-06,
+      "loss": 0.8233,
       "step": 19429
     },
     {
-      "epoch": 0.5336299469940403,
+      "epoch": 0.5513620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 9.397273670990429e-06,
-      "loss": 0.9609,
+      "learning_rate": 8.825379012437066e-06,
+      "loss": 0.8882,
       "step": 19430
     },
     {
-      "epoch": 0.5336574112218835,
+      "epoch": 0.5513904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 9.39638577454604e-06,
-      "loss": 0.8998,
+      "learning_rate": 8.82446630177179e-06,
+      "loss": 0.7865,
       "step": 19431
     },
     {
-      "epoch": 0.5336848754497268,
+      "epoch": 0.5514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 9.395497882877678e-06,
-      "loss": 0.9356,
+      "learning_rate": 8.8235536010363e-06,
+      "loss": 0.7842,
       "step": 19432
     },
     {
-      "epoch": 0.5337123396775699,
+      "epoch": 0.5514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 9.39460999599237e-06,
-      "loss": 0.8011,
+      "learning_rate": 8.822640910238309e-06,
+      "loss": 0.8916,
       "step": 19433
     },
     {
-      "epoch": 0.5337398039054132,
+      "epoch": 0.5514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 9.393722113897147e-06,
-      "loss": 0.8072,
+      "learning_rate": 8.821728229385522e-06,
+      "loss": 0.8525,
       "step": 19434
     },
     {
-      "epoch": 0.5337672681332565,
+      "epoch": 0.5515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 9.392834236599027e-06,
-      "loss": 0.8488,
+      "learning_rate": 8.820815558485647e-06,
+      "loss": 0.9256,
       "step": 19435
     },
     {
-      "epoch": 0.5337947323610996,
+      "epoch": 0.5515323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 9.391946364105039e-06,
-      "loss": 0.8043,
+      "learning_rate": 8.8199028975464e-06,
+      "loss": 0.8479,
       "step": 19436
     },
     {
-      "epoch": 0.5338221965889429,
+      "epoch": 0.5515607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 9.391058496422204e-06,
-      "loss": 0.826,
+      "learning_rate": 8.818990246575485e-06,
+      "loss": 0.8907,
       "step": 19437
     },
     {
-      "epoch": 0.5338496608167861,
+      "epoch": 0.551589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 9.39017063355755e-06,
-      "loss": 0.8874,
+      "learning_rate": 8.81807760558061e-06,
+      "loss": 0.877,
       "step": 19438
     },
     {
-      "epoch": 0.5338771250446294,
+      "epoch": 0.5516174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 9.389282775518105e-06,
-      "loss": 0.8833,
+      "learning_rate": 8.81716497456949e-06,
+      "loss": 0.8748,
       "step": 19439
     },
     {
-      "epoch": 0.5339045892724726,
+      "epoch": 0.5516458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 9.388394922310893e-06,
-      "loss": 0.8952,
+      "learning_rate": 8.816252353549829e-06,
+      "loss": 0.9903,
       "step": 19440
     },
     {
-      "epoch": 0.5339320535003158,
+      "epoch": 0.5516742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 9.387507073942932e-06,
-      "loss": 0.8748,
+      "learning_rate": 8.815339742529336e-06,
+      "loss": 0.9205,
       "step": 19441
     },
     {
-      "epoch": 0.5339595177281591,
+      "epoch": 0.5517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 9.386619230421254e-06,
-      "loss": 0.8806,
+      "learning_rate": 8.814427141515724e-06,
+      "loss": 0.8602,
       "step": 19442
     },
     {
-      "epoch": 0.5339869819560024,
+      "epoch": 0.5517309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 9.385731391752884e-06,
-      "loss": 0.9253,
+      "learning_rate": 8.8135145505167e-06,
+      "loss": 0.9395,
       "step": 19443
     },
     {
-      "epoch": 0.5340144461838455,
+      "epoch": 0.5517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 9.384843557944842e-06,
-      "loss": 0.9306,
+      "learning_rate": 8.812601969539968e-06,
+      "loss": 0.857,
       "step": 19444
     },
     {
-      "epoch": 0.5340419104116888,
+      "epoch": 0.5517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 9.383955729004156e-06,
-      "loss": 0.8644,
+      "learning_rate": 8.811689398593245e-06,
+      "loss": 0.8773,
       "step": 19445
     },
     {
-      "epoch": 0.534069374639532,
+      "epoch": 0.551816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 9.383067904937857e-06,
-      "loss": 0.9358,
+      "learning_rate": 8.810776837684229e-06,
+      "loss": 0.8515,
       "step": 19446
     },
     {
-      "epoch": 0.5340968388673752,
+      "epoch": 0.551844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 9.382180085752958e-06,
-      "loss": 0.7136,
+      "learning_rate": 8.809864286820638e-06,
+      "loss": 0.8876,
       "step": 19447
     },
     {
-      "epoch": 0.5341243030952185,
+      "epoch": 0.5518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 9.381292271456493e-06,
-      "loss": 0.8838,
+      "learning_rate": 8.808951746010176e-06,
+      "loss": 0.7905,
       "step": 19448
     },
     {
-      "epoch": 0.5341517673230617,
+      "epoch": 0.5519012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 9.38040446205548e-06,
-      "loss": 0.8839,
+      "learning_rate": 8.808039215260555e-06,
+      "loss": 0.9038,
       "step": 19449
     },
     {
-      "epoch": 0.534179231550905,
+      "epoch": 0.5519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 9.379516657556948e-06,
-      "loss": 1.0085,
+      "learning_rate": 8.80712669457948e-06,
+      "loss": 0.8455,
       "step": 19450
     },
     {
-      "epoch": 0.5342066957787481,
+      "epoch": 0.5519580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 9.378628857967921e-06,
-      "loss": 0.8684,
+      "learning_rate": 8.806214183974659e-06,
+      "loss": 0.8887,
       "step": 19451
     },
     {
-      "epoch": 0.5342341600065914,
+      "epoch": 0.5519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 9.377741063295425e-06,
-      "loss": 0.9128,
+      "learning_rate": 8.805301683453797e-06,
+      "loss": 0.8938,
       "step": 19452
     },
     {
-      "epoch": 0.5342616242344347,
+      "epoch": 0.5520147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 9.37685327354648e-06,
-      "loss": 0.9695,
+      "learning_rate": 8.804389193024609e-06,
+      "loss": 0.9216,
       "step": 19453
     },
     {
-      "epoch": 0.5342890884622779,
+      "epoch": 0.5520431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 9.375965488728113e-06,
-      "loss": 0.8808,
+      "learning_rate": 8.8034767126948e-06,
+      "loss": 0.8082,
       "step": 19454
     },
     {
-      "epoch": 0.5343165526901211,
+      "epoch": 0.5520715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 9.375077708847352e-06,
-      "loss": 0.8927,
+      "learning_rate": 8.802564242472074e-06,
+      "loss": 0.8424,
       "step": 19455
     },
     {
-      "epoch": 0.5343440169179644,
+      "epoch": 0.552099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 9.374189933911215e-06,
-      "loss": 0.8962,
+      "learning_rate": 8.801651782364143e-06,
+      "loss": 0.871,
       "step": 19456
     },
     {
-      "epoch": 0.5343714811458076,
+      "epoch": 0.552128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 9.373302163926728e-06,
-      "loss": 0.8386,
+      "learning_rate": 8.800739332378714e-06,
+      "loss": 0.8347,
       "step": 19457
     },
     {
-      "epoch": 0.5343989453736508,
+      "epoch": 0.5521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 9.372414398900924e-06,
-      "loss": 0.8524,
+      "learning_rate": 8.799826892523492e-06,
+      "loss": 0.9276,
       "step": 19458
     },
     {
-      "epoch": 0.534426409601494,
+      "epoch": 0.5521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 9.371526638840813e-06,
-      "loss": 0.8928,
+      "learning_rate": 8.798914462806188e-06,
+      "loss": 0.8515,
       "step": 19459
     },
     {
-      "epoch": 0.5344538738293373,
+      "epoch": 0.5522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 9.370638883753432e-06,
-      "loss": 0.9291,
+      "learning_rate": 8.798002043234506e-06,
+      "loss": 0.8029,
       "step": 19460
     },
     {
-      "epoch": 0.5344813380571806,
+      "epoch": 0.5522417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 9.369751133645796e-06,
-      "loss": 0.8197,
+      "learning_rate": 8.797089633816153e-06,
+      "loss": 0.8234,
       "step": 19461
     },
     {
-      "epoch": 0.5345088022850237,
+      "epoch": 0.5522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 9.368863388524934e-06,
-      "loss": 0.8624,
+      "learning_rate": 8.796177234558838e-06,
+      "loss": 0.9348,
       "step": 19462
     },
     {
-      "epoch": 0.534536266512867,
+      "epoch": 0.5522985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 9.36797564839787e-06,
-      "loss": 0.863,
+      "learning_rate": 8.795264845470272e-06,
+      "loss": 0.8202,
       "step": 19463
     },
     {
-      "epoch": 0.5345637307407102,
+      "epoch": 0.5523269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 9.367087913271632e-06,
-      "loss": 0.9185,
+      "learning_rate": 8.794352466558157e-06,
+      "loss": 0.8276,
       "step": 19464
     },
     {
-      "epoch": 0.5345911949685535,
+      "epoch": 0.552355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 9.366200183153233e-06,
-      "loss": 0.9512,
+      "learning_rate": 8.793440097830199e-06,
+      "loss": 0.9943,
       "step": 19465
     },
     {
-      "epoch": 0.5346186591963967,
+      "epoch": 0.5523836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 9.365312458049706e-06,
-      "loss": 0.8134,
+      "learning_rate": 8.792527739294109e-06,
+      "loss": 0.843,
       "step": 19466
     },
     {
-      "epoch": 0.5346461234242399,
+      "epoch": 0.5524120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 9.364424737968074e-06,
-      "loss": 0.9353,
+      "learning_rate": 8.791615390957591e-06,
+      "loss": 0.7955,
       "step": 19467
     },
     {
-      "epoch": 0.5346735876520832,
+      "epoch": 0.5524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 9.363537022915357e-06,
-      "loss": 0.8492,
+      "learning_rate": 8.79070305282835e-06,
+      "loss": 0.9273,
       "step": 19468
     },
     {
-      "epoch": 0.5347010518799264,
+      "epoch": 0.5524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 9.36264931289858e-06,
-      "loss": 0.8644,
+      "learning_rate": 8.789790724914098e-06,
+      "loss": 0.856,
       "step": 19469
     },
     {
-      "epoch": 0.5347285161077696,
+      "epoch": 0.5524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 9.361761607924775e-06,
-      "loss": 0.7389,
+      "learning_rate": 8.788878407222538e-06,
+      "loss": 0.8561,
       "step": 19470
     },
     {
-      "epoch": 0.5347559803356129,
+      "epoch": 0.5525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 9.360873908000956e-06,
-      "loss": 0.7678,
+      "learning_rate": 8.787966099761372e-06,
+      "loss": 0.7327,
       "step": 19471
     },
     {
-      "epoch": 0.5347834445634561,
+      "epoch": 0.5525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 9.35998621313415e-06,
-      "loss": 0.9391,
+      "learning_rate": 8.787053802538316e-06,
+      "loss": 0.9099,
       "step": 19472
     },
     {
-      "epoch": 0.5348109087912993,
+      "epoch": 0.5525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 9.359098523331381e-06,
-      "loss": 0.9026,
+      "learning_rate": 8.786141515561068e-06,
+      "loss": 0.7951,
       "step": 19473
     },
     {
-      "epoch": 0.5348383730191426,
+      "epoch": 0.5526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 9.35821083859967e-06,
-      "loss": 0.9102,
+      "learning_rate": 8.785229238837338e-06,
+      "loss": 0.8814,
       "step": 19474
     },
     {
-      "epoch": 0.5348658372469858,
+      "epoch": 0.552639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 9.357323158946048e-06,
-      "loss": 0.901,
+      "learning_rate": 8.784316972374832e-06,
+      "loss": 0.8471,
       "step": 19475
     },
     {
-      "epoch": 0.534893301474829,
+      "epoch": 0.5526674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 9.356435484377536e-06,
-      "loss": 0.7902,
+      "learning_rate": 8.783404716181255e-06,
+      "loss": 0.9711,
       "step": 19476
     },
     {
-      "epoch": 0.5349207657026722,
+      "epoch": 0.5526958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 9.35554781490115e-06,
-      "loss": 0.86,
+      "learning_rate": 8.78249247026431e-06,
+      "loss": 0.9028,
       "step": 19477
     },
     {
-      "epoch": 0.5349482299305155,
+      "epoch": 0.5527241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 9.35466015052392e-06,
-      "loss": 0.963,
+      "learning_rate": 8.781580234631709e-06,
+      "loss": 0.8508,
       "step": 19478
     },
     {
-      "epoch": 0.5349756941583588,
+      "epoch": 0.5527525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 9.353772491252872e-06,
-      "loss": 0.8033,
+      "learning_rate": 8.780668009291151e-06,
+      "loss": 0.8768,
       "step": 19479
     },
     {
-      "epoch": 0.5350031583862019,
+      "epoch": 0.5527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 9.352884837095024e-06,
-      "loss": 0.9025,
+      "learning_rate": 8.77975579425035e-06,
+      "loss": 1.0,
       "step": 19480
     },
     {
-      "epoch": 0.5350306226140452,
+      "epoch": 0.5528093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 9.351997188057406e-06,
-      "loss": 0.9452,
+      "learning_rate": 8.778843589517005e-06,
+      "loss": 0.7831,
       "step": 19481
     },
     {
-      "epoch": 0.5350580868418885,
+      "epoch": 0.5528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 9.351109544147032e-06,
-      "loss": 0.9168,
+      "learning_rate": 8.77793139509882e-06,
+      "loss": 0.8816,
       "step": 19482
     },
     {
-      "epoch": 0.5350855510697317,
+      "epoch": 0.5528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 9.35022190537093e-06,
-      "loss": 0.9276,
+      "learning_rate": 8.777019211003508e-06,
+      "loss": 0.8551,
       "step": 19483
     },
     {
-      "epoch": 0.5351130152975749,
+      "epoch": 0.552894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 9.349334271736128e-06,
-      "loss": 0.8817,
+      "learning_rate": 8.77610703723877e-06,
+      "loss": 0.8086,
       "step": 19484
     },
     {
-      "epoch": 0.5351404795254181,
+      "epoch": 0.5529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 9.348446643249644e-06,
-      "loss": 0.8575,
+      "learning_rate": 8.775194873812308e-06,
+      "loss": 0.8645,
       "step": 19485
     },
     {
-      "epoch": 0.5351679437532614,
+      "epoch": 0.5529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 9.3475590199185e-06,
-      "loss": 0.9469,
+      "learning_rate": 8.77428272073183e-06,
+      "loss": 0.8579,
       "step": 19486
     },
     {
-      "epoch": 0.5351954079811047,
+      "epoch": 0.5529795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 9.346671401749728e-06,
-      "loss": 0.8409,
+      "learning_rate": 8.773370578005042e-06,
+      "loss": 0.9158,
       "step": 19487
     },
     {
-      "epoch": 0.5352228722089478,
+      "epoch": 0.5530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 9.345783788750341e-06,
-      "loss": 0.7759,
+      "learning_rate": 8.772458445639647e-06,
+      "loss": 0.9431,
       "step": 19488
     },
     {
-      "epoch": 0.5352503364367911,
+      "epoch": 0.5530363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 9.344896180927366e-06,
-      "loss": 0.9347,
+      "learning_rate": 8.771546323643352e-06,
+      "loss": 0.9249,
       "step": 19489
     },
     {
-      "epoch": 0.5352778006646343,
+      "epoch": 0.5530646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 9.344008578287824e-06,
-      "loss": 0.8405,
+      "learning_rate": 8.770634212023861e-06,
+      "loss": 0.7697,
       "step": 19490
     },
     {
-      "epoch": 0.5353052648924775,
+      "epoch": 0.5530930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 9.343120980838744e-06,
-      "loss": 0.8846,
+      "learning_rate": 8.769722110788876e-06,
+      "loss": 0.8387,
       "step": 19491
     },
     {
-      "epoch": 0.5353327291203208,
+      "epoch": 0.5531214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 9.342233388587143e-06,
-      "loss": 0.8646,
+      "learning_rate": 8.768810019946102e-06,
+      "loss": 0.8918,
       "step": 19492
     },
     {
-      "epoch": 0.535360193348164,
+      "epoch": 0.553149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 9.341345801540049e-06,
-      "loss": 0.8963,
+      "learning_rate": 8.767897939503246e-06,
+      "loss": 0.8972,
       "step": 19493
     },
     {
-      "epoch": 0.5353876575760073,
+      "epoch": 0.553178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 9.34045821970448e-06,
-      "loss": 0.862,
+      "learning_rate": 8.766985869468014e-06,
+      "loss": 0.8867,
       "step": 19494
     },
     {
-      "epoch": 0.5354151218038505,
+      "epoch": 0.5532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 9.339570643087462e-06,
-      "loss": 0.9107,
+      "learning_rate": 8.766073809848108e-06,
+      "loss": 0.8594,
       "step": 19495
     },
     {
-      "epoch": 0.5354425860316937,
+      "epoch": 0.5532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 9.338683071696014e-06,
-      "loss": 0.8519,
+      "learning_rate": 8.765161760651228e-06,
+      "loss": 0.9193,
       "step": 19496
     },
     {
-      "epoch": 0.535470050259537,
+      "epoch": 0.5532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 9.337795505537163e-06,
-      "loss": 0.8787,
+      "learning_rate": 8.764249721885087e-06,
+      "loss": 0.835,
       "step": 19497
     },
     {
-      "epoch": 0.5354975144873801,
+      "epoch": 0.5532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 9.336907944617929e-06,
-      "loss": 0.8802,
+      "learning_rate": 8.763337693557383e-06,
+      "loss": 0.9465,
       "step": 19498
     },
     {
-      "epoch": 0.5355249787152234,
+      "epoch": 0.5533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 9.336020388945342e-06,
-      "loss": 0.8755,
+      "learning_rate": 8.762425675675818e-06,
+      "loss": 0.8129,
       "step": 19499
     },
     {
-      "epoch": 0.5355524429430667,
+      "epoch": 0.5533484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 9.335132838526416e-06,
-      "loss": 0.9139,
+      "learning_rate": 8.761513668248103e-06,
+      "loss": 0.9034,
       "step": 19500
     },
     {
-      "epoch": 0.5355799071709099,
+      "epoch": 0.5533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 9.334245293368175e-06,
-      "loss": 0.8835,
+      "learning_rate": 8.760601671281937e-06,
+      "loss": 0.8797,
       "step": 19501
     },
     {
-      "epoch": 0.5356073713987531,
+      "epoch": 0.5534052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 9.333357753477643e-06,
-      "loss": 0.7914,
+      "learning_rate": 8.759689684785021e-06,
+      "loss": 0.9532,
       "step": 19502
     },
     {
-      "epoch": 0.5356348356265963,
+      "epoch": 0.5534335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 9.332470218861844e-06,
-      "loss": 0.9409,
+      "learning_rate": 8.758777708765065e-06,
+      "loss": 0.9394,
       "step": 19503
     },
     {
-      "epoch": 0.5356622998544396,
+      "epoch": 0.5534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 9.331582689527799e-06,
-      "loss": 0.8705,
+      "learning_rate": 8.75786574322977e-06,
+      "loss": 0.9513,
       "step": 19504
     },
     {
-      "epoch": 0.5356897640822829,
+      "epoch": 0.5534903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 9.330695165482533e-06,
-      "loss": 0.83,
+      "learning_rate": 8.756953788186837e-06,
+      "loss": 0.8695,
       "step": 19505
     },
     {
-      "epoch": 0.535717228310126,
+      "epoch": 0.5535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 9.329807646733063e-06,
-      "loss": 0.9104,
+      "learning_rate": 8.756041843643971e-06,
+      "loss": 0.8514,
       "step": 19506
     },
     {
-      "epoch": 0.5357446925379693,
+      "epoch": 0.5535471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 9.328920133286416e-06,
-      "loss": 0.8698,
+      "learning_rate": 8.755129909608873e-06,
+      "loss": 0.8431,
       "step": 19507
     },
     {
-      "epoch": 0.5357721567658126,
+      "epoch": 0.5535754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 9.32803262514961e-06,
-      "loss": 0.8787,
+      "learning_rate": 8.754217986089253e-06,
+      "loss": 0.9209,
       "step": 19508
     },
     {
-      "epoch": 0.5357996209936557,
+      "epoch": 0.5536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 9.327145122329671e-06,
-      "loss": 0.838,
+      "learning_rate": 8.753306073092807e-06,
+      "loss": 0.9261,
       "step": 19509
     },
     {
-      "epoch": 0.535827085221499,
+      "epoch": 0.5536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 9.32625762483362e-06,
-      "loss": 0.8481,
+      "learning_rate": 8.752394170627243e-06,
+      "loss": 0.8102,
       "step": 19510
     },
     {
-      "epoch": 0.5358545494493422,
+      "epoch": 0.553660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 9.325370132668484e-06,
-      "loss": 0.9264,
+      "learning_rate": 8.751482278700262e-06,
+      "loss": 0.8707,
       "step": 19511
     },
     {
-      "epoch": 0.5358820136771855,
+      "epoch": 0.553688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 9.32448264584128e-06,
-      "loss": 0.9272,
+      "learning_rate": 8.750570397319562e-06,
+      "loss": 0.8496,
       "step": 19512
     },
     {
-      "epoch": 0.5359094779050287,
+      "epoch": 0.5537173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 9.323595164359027e-06,
-      "loss": 0.9413,
+      "learning_rate": 8.749658526492855e-06,
+      "loss": 1.0442,
       "step": 19513
     },
     {
-      "epoch": 0.5359369421328719,
+      "epoch": 0.5537457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 9.322707688228753e-06,
-      "loss": 0.9341,
+      "learning_rate": 8.748746666227838e-06,
+      "loss": 1.001,
       "step": 19514
     },
     {
-      "epoch": 0.5359644063607152,
+      "epoch": 0.5537741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 9.321820217457478e-06,
-      "loss": 0.8547,
+      "learning_rate": 8.747834816532212e-06,
+      "loss": 0.7853,
       "step": 19515
     },
     {
-      "epoch": 0.5359918705885584,
+      "epoch": 0.5538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 9.320932752052224e-06,
-      "loss": 0.8141,
+      "learning_rate": 8.746922977413684e-06,
+      "loss": 0.8769,
       "step": 19516
     },
     {
-      "epoch": 0.5360193348164016,
+      "epoch": 0.5538308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 9.320045292020016e-06,
-      "loss": 0.9097,
+      "learning_rate": 8.746011148879951e-06,
+      "loss": 0.9165,
       "step": 19517
     },
     {
-      "epoch": 0.5360467990442449,
+      "epoch": 0.5538592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 9.319157837367869e-06,
-      "loss": 0.9645,
+      "learning_rate": 8.745099330938721e-06,
+      "loss": 0.7956,
       "step": 19518
     },
     {
-      "epoch": 0.5360742632720881,
+      "epoch": 0.5538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 9.318270388102812e-06,
-      "loss": 0.8539,
+      "learning_rate": 8.744187523597692e-06,
+      "loss": 0.8507,
       "step": 19519
     },
     {
-      "epoch": 0.5361017274999313,
+      "epoch": 0.5539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 9.31738294423186e-06,
-      "loss": 0.855,
+      "learning_rate": 8.743275726864567e-06,
+      "loss": 0.8294,
       "step": 19520
     },
     {
-      "epoch": 0.5361291917277746,
+      "epoch": 0.55394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 9.31649550576204e-06,
-      "loss": 0.8507,
+      "learning_rate": 8.74236394074705e-06,
+      "loss": 0.8061,
       "step": 19521
     },
     {
-      "epoch": 0.5361566559556178,
+      "epoch": 0.5539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 9.315608072700373e-06,
-      "loss": 0.9443,
+      "learning_rate": 8.741452165252838e-06,
+      "loss": 0.8387,
       "step": 19522
     },
     {
-      "epoch": 0.5361841201834611,
+      "epoch": 0.5540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 9.314720645053882e-06,
-      "loss": 0.9478,
+      "learning_rate": 8.740540400389635e-06,
+      "loss": 0.859,
       "step": 19523
     },
     {
-      "epoch": 0.5362115844113042,
+      "epoch": 0.5540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 9.313833222829587e-06,
-      "loss": 0.8595,
+      "learning_rate": 8.739628646165149e-06,
+      "loss": 0.8051,
       "step": 19524
     },
     {
-      "epoch": 0.5362390486391475,
+      "epoch": 0.5540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 9.312945806034506e-06,
-      "loss": 0.9605,
+      "learning_rate": 8.738716902587074e-06,
+      "loss": 0.8661,
       "step": 19525
     },
     {
-      "epoch": 0.5362665128669908,
+      "epoch": 0.5540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 9.312058394675664e-06,
-      "loss": 0.892,
+      "learning_rate": 8.737805169663113e-06,
+      "loss": 0.7349,
       "step": 19526
     },
     {
-      "epoch": 0.536293977094834,
+      "epoch": 0.5541146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 9.311170988760085e-06,
-      "loss": 0.9255,
+      "learning_rate": 8.736893447400973e-06,
+      "loss": 0.9619,
       "step": 19527
     },
     {
-      "epoch": 0.5363214413226772,
+      "epoch": 0.5541430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 9.310283588294785e-06,
-      "loss": 0.8957,
+      "learning_rate": 8.735981735808348e-06,
+      "loss": 0.9231,
       "step": 19528
     },
     {
-      "epoch": 0.5363489055505204,
+      "epoch": 0.5541713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 9.309396193286793e-06,
-      "loss": 0.7945,
+      "learning_rate": 8.735070034892941e-06,
+      "loss": 0.93,
       "step": 19529
     },
     {
-      "epoch": 0.5363763697783637,
+      "epoch": 0.554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 9.308508803743122e-06,
-      "loss": 0.7222,
+      "learning_rate": 8.734158344662455e-06,
+      "loss": 0.9185,
       "step": 19530
     },
     {
-      "epoch": 0.536403834006207,
+      "epoch": 0.5542281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 9.307621419670798e-06,
-      "loss": 0.8423,
+      "learning_rate": 8.733246665124592e-06,
+      "loss": 0.8041,
       "step": 19531
     },
     {
-      "epoch": 0.5364312982340501,
+      "epoch": 0.5542565266742339,
       "grad_norm": 0.0,
-      "learning_rate": 9.30673404107684e-06,
-      "loss": 0.8936,
+      "learning_rate": 8.732334996287048e-06,
+      "loss": 0.8572,
       "step": 19532
     },
     {
-      "epoch": 0.5364587624618934,
+      "epoch": 0.5542849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 9.305846667968272e-06,
-      "loss": 0.8581,
+      "learning_rate": 8.731423338157532e-06,
+      "loss": 0.9395,
       "step": 19533
     },
     {
-      "epoch": 0.5364862266897367,
+      "epoch": 0.5543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 9.304959300352116e-06,
-      "loss": 0.8878,
+      "learning_rate": 8.730511690743738e-06,
+      "loss": 0.7961,
       "step": 19534
     },
     {
-      "epoch": 0.5365136909175798,
+      "epoch": 0.5543416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 9.30407193823539e-06,
-      "loss": 0.9095,
+      "learning_rate": 8.729600054053367e-06,
+      "loss": 0.984,
       "step": 19535
     },
     {
-      "epoch": 0.5365411551454231,
+      "epoch": 0.5543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 9.303184581625118e-06,
-      "loss": 0.8764,
+      "learning_rate": 8.728688428094123e-06,
+      "loss": 0.8672,
       "step": 19536
     },
     {
-      "epoch": 0.5365686193732663,
+      "epoch": 0.5543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 9.302297230528315e-06,
-      "loss": 0.936,
+      "learning_rate": 8.727776812873705e-06,
+      "loss": 0.9472,
       "step": 19537
     },
     {
-      "epoch": 0.5365960836011096,
+      "epoch": 0.5544267877412031,
       "grad_norm": 0.0,
-      "learning_rate": 9.301409884952009e-06,
-      "loss": 0.8393,
+      "learning_rate": 8.72686520839981e-06,
+      "loss": 0.8846,
       "step": 19538
     },
     {
-      "epoch": 0.5366235478289528,
+      "epoch": 0.5544551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 9.300522544903219e-06,
-      "loss": 1.0137,
+      "learning_rate": 8.725953614680147e-06,
+      "loss": 0.9589,
       "step": 19539
     },
     {
-      "epoch": 0.536651012056796,
+      "epoch": 0.554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 9.299635210388968e-06,
-      "loss": 0.8714,
+      "learning_rate": 8.725042031722406e-06,
+      "loss": 0.8945,
       "step": 19540
     },
     {
-      "epoch": 0.5366784762846393,
+      "epoch": 0.5545119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 9.298747881416271e-06,
-      "loss": 0.9505,
+      "learning_rate": 8.724130459534297e-06,
+      "loss": 0.8028,
       "step": 19541
     },
     {
-      "epoch": 0.5367059405124824,
+      "epoch": 0.5545402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 9.297860557992153e-06,
-      "loss": 0.8823,
+      "learning_rate": 8.723218898123514e-06,
+      "loss": 0.8504,
       "step": 19542
     },
     {
-      "epoch": 0.5367334047403257,
+      "epoch": 0.5545686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 9.296973240123634e-06,
-      "loss": 0.917,
+      "learning_rate": 8.722307347497757e-06,
+      "loss": 0.8547,
       "step": 19543
     },
     {
-      "epoch": 0.536760868968169,
+      "epoch": 0.5545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 9.296085927817735e-06,
-      "loss": 0.7527,
+      "learning_rate": 8.72139580766473e-06,
+      "loss": 0.8482,
       "step": 19544
     },
     {
-      "epoch": 0.5367883331960122,
+      "epoch": 0.5546254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 9.295198621081476e-06,
-      "loss": 0.9525,
+      "learning_rate": 8.720484278632129e-06,
+      "loss": 0.9151,
       "step": 19545
     },
     {
-      "epoch": 0.5368157974238554,
+      "epoch": 0.5546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 9.294311319921883e-06,
-      "loss": 0.8016,
+      "learning_rate": 8.719572760407652e-06,
+      "loss": 0.7136,
       "step": 19546
     },
     {
-      "epoch": 0.5368432616516987,
+      "epoch": 0.5546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 9.293424024345969e-06,
-      "loss": 0.7845,
+      "learning_rate": 8.718661252999004e-06,
+      "loss": 0.8641,
       "step": 19547
     },
     {
-      "epoch": 0.5368707258795419,
+      "epoch": 0.5547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 9.292536734360758e-06,
-      "loss": 0.8617,
+      "learning_rate": 8.717749756413883e-06,
+      "loss": 0.8158,
       "step": 19548
     },
     {
-      "epoch": 0.5368981901073852,
+      "epoch": 0.554738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 9.291649449973268e-06,
-      "loss": 0.8739,
+      "learning_rate": 8.716838270659985e-06,
+      "loss": 0.8089,
       "step": 19549
     },
     {
-      "epoch": 0.5369256543352283,
+      "epoch": 0.5547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 9.290762171190524e-06,
-      "loss": 0.9657,
+      "learning_rate": 8.715926795745013e-06,
+      "loss": 0.8174,
       "step": 19550
     },
     {
-      "epoch": 0.5369531185630716,
+      "epoch": 0.5547956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 9.289874898019546e-06,
-      "loss": 0.8427,
+      "learning_rate": 8.715015331676667e-06,
+      "loss": 0.8344,
       "step": 19551
     },
     {
-      "epoch": 0.5369805827909149,
+      "epoch": 0.5548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 9.288987630467354e-06,
-      "loss": 0.8965,
+      "learning_rate": 8.71410387846264e-06,
+      "loss": 0.8387,
       "step": 19552
     },
     {
-      "epoch": 0.537008047018758,
+      "epoch": 0.5548524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 9.288100368540964e-06,
-      "loss": 0.8898,
+      "learning_rate": 8.713192436110635e-06,
+      "loss": 0.7627,
       "step": 19553
     },
     {
-      "epoch": 0.5370355112466013,
+      "epoch": 0.5548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 9.2872131122474e-06,
-      "loss": 1.0143,
+      "learning_rate": 8.712281004628355e-06,
+      "loss": 0.8746,
       "step": 19554
     },
     {
-      "epoch": 0.5370629754744445,
+      "epoch": 0.5549091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 9.286325861593684e-06,
-      "loss": 0.8689,
+      "learning_rate": 8.711369584023494e-06,
+      "loss": 0.7925,
       "step": 19555
     },
     {
-      "epoch": 0.5370904397022878,
+      "epoch": 0.5549375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 9.28543861658683e-06,
-      "loss": 0.8408,
+      "learning_rate": 8.71045817430375e-06,
+      "loss": 0.8283,
       "step": 19556
     },
     {
-      "epoch": 0.537117903930131,
+      "epoch": 0.5549659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 9.284551377233865e-06,
-      "loss": 0.9527,
+      "learning_rate": 8.709546775476828e-06,
+      "loss": 0.9032,
       "step": 19557
     },
     {
-      "epoch": 0.5371453681579742,
+      "epoch": 0.554994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 9.28366414354181e-06,
-      "loss": 0.9532,
+      "learning_rate": 8.708635387550422e-06,
+      "loss": 0.8655,
       "step": 19558
     },
     {
-      "epoch": 0.5371728323858175,
+      "epoch": 0.555022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 9.282776915517677e-06,
-      "loss": 0.9713,
+      "learning_rate": 8.70772401053223e-06,
+      "loss": 0.818,
       "step": 19559
     },
     {
-      "epoch": 0.5372002966136608,
+      "epoch": 0.5550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 9.281889693168492e-06,
-      "loss": 0.8797,
+      "learning_rate": 8.70681264442995e-06,
+      "loss": 0.9539,
       "step": 19560
     },
     {
-      "epoch": 0.5372277608415039,
+      "epoch": 0.5550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 9.281002476501275e-06,
-      "loss": 0.9281,
+      "learning_rate": 8.705901289251282e-06,
+      "loss": 0.9478,
       "step": 19561
     },
     {
-      "epoch": 0.5372552250693472,
+      "epoch": 0.5551078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 9.280115265523042e-06,
-      "loss": 0.9028,
+      "learning_rate": 8.704989945003926e-06,
+      "loss": 0.8457,
       "step": 19562
     },
     {
-      "epoch": 0.5372826892971904,
+      "epoch": 0.5551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 9.27922806024082e-06,
-      "loss": 0.8357,
+      "learning_rate": 8.704078611695575e-06,
+      "loss": 0.8195,
       "step": 19563
     },
     {
-      "epoch": 0.5373101535250336,
+      "epoch": 0.5551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 9.278340860661626e-06,
-      "loss": 0.8247,
+      "learning_rate": 8.703167289333932e-06,
+      "loss": 0.8347,
       "step": 19564
     },
     {
-      "epoch": 0.5373376177528769,
+      "epoch": 0.5551929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 9.277453666792475e-06,
-      "loss": 0.9158,
+      "learning_rate": 8.702255977926695e-06,
+      "loss": 0.8997,
       "step": 19565
     },
     {
-      "epoch": 0.5373650819807201,
+      "epoch": 0.5552213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 9.276566478640389e-06,
-      "loss": 0.7932,
+      "learning_rate": 8.701344677481555e-06,
+      "loss": 0.8885,
       "step": 19566
     },
     {
-      "epoch": 0.5373925462085634,
+      "epoch": 0.5552497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 9.275679296212391e-06,
-      "loss": 0.7856,
+      "learning_rate": 8.700433388006219e-06,
+      "loss": 0.8911,
       "step": 19567
     },
     {
-      "epoch": 0.5374200104364065,
+      "epoch": 0.5552780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 9.2747921195155e-06,
-      "loss": 0.8758,
+      "learning_rate": 8.699522109508382e-06,
+      "loss": 0.8822,
       "step": 19568
     },
     {
-      "epoch": 0.5374474746642498,
+      "epoch": 0.5553064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 9.273904948556733e-06,
-      "loss": 0.811,
+      "learning_rate": 8.698610841995735e-06,
+      "loss": 0.93,
       "step": 19569
     },
     {
-      "epoch": 0.5374749388920931,
+      "epoch": 0.5553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 9.273017783343114e-06,
-      "loss": 0.8552,
+      "learning_rate": 8.69769958547598e-06,
+      "loss": 0.7557,
       "step": 19570
     },
     {
-      "epoch": 0.5375024031199362,
+      "epoch": 0.5553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 9.272130623881657e-06,
-      "loss": 0.7654,
+      "learning_rate": 8.696788339956819e-06,
+      "loss": 0.8637,
       "step": 19571
     },
     {
-      "epoch": 0.5375298673477795,
+      "epoch": 0.5553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 9.271243470179388e-06,
-      "loss": 0.8407,
+      "learning_rate": 8.695877105445947e-06,
+      "loss": 0.8382,
       "step": 19572
     },
     {
-      "epoch": 0.5375573315756228,
+      "epoch": 0.5554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 9.270356322243319e-06,
-      "loss": 0.7962,
+      "learning_rate": 8.694965881951057e-06,
+      "loss": 0.7915,
       "step": 19573
     },
     {
-      "epoch": 0.537584795803466,
+      "epoch": 0.5554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 9.269469180080472e-06,
-      "loss": 0.8596,
+      "learning_rate": 8.69405466947985e-06,
+      "loss": 0.8901,
       "step": 19574
     },
     {
-      "epoch": 0.5376122600313092,
+      "epoch": 0.5554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 9.268582043697873e-06,
-      "loss": 0.999,
+      "learning_rate": 8.693143468040023e-06,
+      "loss": 0.8703,
       "step": 19575
     },
     {
-      "epoch": 0.5376397242591524,
+      "epoch": 0.5555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 9.267694913102537e-06,
-      "loss": 0.8779,
+      "learning_rate": 8.692232277639269e-06,
+      "loss": 0.8344,
       "step": 19576
     },
     {
-      "epoch": 0.5376671884869957,
+      "epoch": 0.555533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 9.266807788301478e-06,
-      "loss": 0.9579,
+      "learning_rate": 8.69132109828529e-06,
+      "loss": 0.9222,
       "step": 19577
     },
     {
-      "epoch": 0.537694652714839,
+      "epoch": 0.5555618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 9.26592066930172e-06,
-      "loss": 0.998,
+      "learning_rate": 8.690409929985781e-06,
+      "loss": 0.9062,
       "step": 19578
     },
     {
-      "epoch": 0.5377221169426821,
+      "epoch": 0.5555902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 9.265033556110285e-06,
-      "loss": 0.9092,
+      "learning_rate": 8.689498772748436e-06,
+      "loss": 0.8453,
       "step": 19579
     },
     {
-      "epoch": 0.5377495811705254,
+      "epoch": 0.5556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 9.264146448734185e-06,
-      "loss": 0.7771,
+      "learning_rate": 8.688587626580954e-06,
+      "loss": 0.8466,
       "step": 19580
     },
     {
-      "epoch": 0.5377770453983686,
+      "epoch": 0.5556469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 9.263259347180446e-06,
-      "loss": 0.968,
+      "learning_rate": 8.687676491491033e-06,
+      "loss": 0.8376,
       "step": 19581
     },
     {
-      "epoch": 0.5378045096262118,
+      "epoch": 0.5556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 9.262372251456087e-06,
-      "loss": 0.9546,
+      "learning_rate": 8.686765367486364e-06,
+      "loss": 0.7915,
       "step": 19582
     },
     {
-      "epoch": 0.5378319738540551,
+      "epoch": 0.5557037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 9.261485161568121e-06,
-      "loss": 0.9447,
+      "learning_rate": 8.685854254574653e-06,
+      "loss": 0.9378,
       "step": 19583
     },
     {
-      "epoch": 0.5378594380818983,
+      "epoch": 0.5557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 9.260598077523573e-06,
-      "loss": 0.8918,
+      "learning_rate": 8.684943152763583e-06,
+      "loss": 0.8593,
       "step": 19584
     },
     {
-      "epoch": 0.5378869023097416,
+      "epoch": 0.5557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 9.259710999329455e-06,
-      "loss": 0.8779,
+      "learning_rate": 8.684032062060862e-06,
+      "loss": 0.9423,
       "step": 19585
     },
     {
-      "epoch": 0.5379143665375848,
+      "epoch": 0.555788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 9.258823926992793e-06,
-      "loss": 0.8561,
+      "learning_rate": 8.683120982474181e-06,
+      "loss": 0.8212,
       "step": 19586
     },
     {
-      "epoch": 0.537941830765428,
+      "epoch": 0.5558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 9.257936860520604e-06,
-      "loss": 0.8292,
+      "learning_rate": 8.682209914011233e-06,
+      "loss": 0.8889,
       "step": 19587
     },
     {
-      "epoch": 0.5379692949932713,
+      "epoch": 0.5558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 9.257049799919908e-06,
-      "loss": 0.9589,
+      "learning_rate": 8.681298856679721e-06,
+      "loss": 0.8146,
       "step": 19588
     },
     {
-      "epoch": 0.5379967592211145,
+      "epoch": 0.5558740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 9.256162745197717e-06,
-      "loss": 0.8587,
+      "learning_rate": 8.680387810487337e-06,
+      "loss": 0.9177,
       "step": 19589
     },
     {
-      "epoch": 0.5380242234489577,
+      "epoch": 0.5559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 9.255275696361057e-06,
-      "loss": 0.8725,
+      "learning_rate": 8.679476775441773e-06,
+      "loss": 0.8908,
       "step": 19590
     },
     {
-      "epoch": 0.538051687676801,
+      "epoch": 0.5559307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 9.254388653416943e-06,
-      "loss": 0.7927,
+      "learning_rate": 8.67856575155073e-06,
+      "loss": 0.7923,
       "step": 19591
     },
     {
-      "epoch": 0.5380791519046442,
+      "epoch": 0.5559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 9.253501616372396e-06,
-      "loss": 0.8349,
+      "learning_rate": 8.677654738821904e-06,
+      "loss": 0.9255,
       "step": 19592
     },
     {
-      "epoch": 0.5381066161324874,
+      "epoch": 0.5559875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 9.25261458523443e-06,
-      "loss": 0.8446,
+      "learning_rate": 8.676743737262985e-06,
+      "loss": 0.7881,
       "step": 19593
     },
     {
-      "epoch": 0.5381340803603307,
+      "epoch": 0.5560158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 9.251727560010074e-06,
-      "loss": 0.8582,
+      "learning_rate": 8.675832746881672e-06,
+      "loss": 0.9211,
       "step": 19594
     },
     {
-      "epoch": 0.5381615445881739,
+      "epoch": 0.556044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 9.250840540706334e-06,
-      "loss": 0.9193,
+      "learning_rate": 8.67492176768566e-06,
+      "loss": 0.8319,
       "step": 19595
     },
     {
-      "epoch": 0.5381890088160172,
+      "epoch": 0.5560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 9.249953527330235e-06,
-      "loss": 0.9125,
+      "learning_rate": 8.674010799682641e-06,
+      "loss": 0.8793,
       "step": 19596
     },
     {
-      "epoch": 0.5382164730438603,
+      "epoch": 0.5561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 9.249066519888794e-06,
-      "loss": 0.9213,
+      "learning_rate": 8.673099842880315e-06,
+      "loss": 0.8905,
       "step": 19597
     },
     {
-      "epoch": 0.5382439372717036,
+      "epoch": 0.5561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 9.248179518389029e-06,
-      "loss": 0.9897,
+      "learning_rate": 8.672188897286374e-06,
+      "loss": 0.7727,
       "step": 19598
     },
     {
-      "epoch": 0.5382714014995469,
+      "epoch": 0.5561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 9.247292522837961e-06,
-      "loss": 0.8489,
+      "learning_rate": 8.671277962908508e-06,
+      "loss": 0.8856,
       "step": 19599
     },
     {
-      "epoch": 0.53829886572739,
+      "epoch": 0.5561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 9.246405533242608e-06,
-      "loss": 0.8683,
+      "learning_rate": 8.670367039754421e-06,
+      "loss": 0.8502,
       "step": 19600
     },
     {
-      "epoch": 0.5383263299552333,
+      "epoch": 0.5562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 9.245518549609982e-06,
-      "loss": 0.8526,
+      "learning_rate": 8.669456127831802e-06,
+      "loss": 0.9182,
       "step": 19601
     },
     {
-      "epoch": 0.5383537941830765,
+      "epoch": 0.5562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 9.244631571947105e-06,
-      "loss": 0.9451,
+      "learning_rate": 8.66854522714835e-06,
+      "loss": 0.9554,
       "step": 19602
     },
     {
-      "epoch": 0.5383812584109198,
+      "epoch": 0.5562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 9.243744600260997e-06,
-      "loss": 1.0036,
+      "learning_rate": 8.667634337711754e-06,
+      "loss": 0.8701,
       "step": 19603
     },
     {
-      "epoch": 0.538408722638763,
+      "epoch": 0.5562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 9.242857634558674e-06,
-      "loss": 0.8327,
+      "learning_rate": 8.66672345952971e-06,
+      "loss": 0.8095,
       "step": 19604
     },
     {
-      "epoch": 0.5384361868666062,
+      "epoch": 0.5563280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 9.24197067484716e-06,
-      "loss": 0.9425,
+      "learning_rate": 8.665812592609914e-06,
+      "loss": 0.9363,
       "step": 19605
     },
     {
-      "epoch": 0.5384636510944495,
+      "epoch": 0.5563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 9.241083721133462e-06,
-      "loss": 0.8153,
+      "learning_rate": 8.66490173696006e-06,
+      "loss": 0.9165,
       "step": 19606
     },
     {
-      "epoch": 0.5384911153222928,
+      "epoch": 0.5563847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 9.240196773424602e-06,
-      "loss": 0.9357,
+      "learning_rate": 8.663990892587839e-06,
+      "loss": 0.7769,
       "step": 19607
     },
     {
-      "epoch": 0.5385185795501359,
+      "epoch": 0.5564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 9.239309831727603e-06,
-      "loss": 0.8827,
+      "learning_rate": 8.66308005950095e-06,
+      "loss": 0.8594,
       "step": 19608
     },
     {
-      "epoch": 0.5385460437779792,
+      "epoch": 0.5564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 9.238422896049478e-06,
-      "loss": 0.8679,
+      "learning_rate": 8.662169237707083e-06,
+      "loss": 0.8712,
       "step": 19609
     },
     {
-      "epoch": 0.5385735080058224,
+      "epoch": 0.5564699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 9.237535966397245e-06,
-      "loss": 0.8326,
+      "learning_rate": 8.66125842721393e-06,
+      "loss": 0.8195,
       "step": 19610
     },
     {
-      "epoch": 0.5386009722336657,
+      "epoch": 0.5564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 9.236649042777928e-06,
-      "loss": 0.8804,
+      "learning_rate": 8.660347628029188e-06,
+      "loss": 0.8619,
       "step": 19611
     },
     {
-      "epoch": 0.5386284364615089,
+      "epoch": 0.5565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 9.235762125198537e-06,
-      "loss": 0.8743,
+      "learning_rate": 8.659436840160553e-06,
+      "loss": 0.8636,
       "step": 19612
     },
     {
-      "epoch": 0.5386559006893521,
+      "epoch": 0.55655505107832,
       "grad_norm": 0.0,
-      "learning_rate": 9.23487521366609e-06,
-      "loss": 0.8494,
+      "learning_rate": 8.658526063615711e-06,
+      "loss": 0.7801,
       "step": 19613
     },
     {
-      "epoch": 0.5386833649171954,
+      "epoch": 0.556583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 9.233988308187607e-06,
-      "loss": 0.8749,
+      "learning_rate": 8.657615298402363e-06,
+      "loss": 1.0146,
       "step": 19614
     },
     {
-      "epoch": 0.5387108291450385,
+      "epoch": 0.5566118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 9.233101408770107e-06,
-      "loss": 0.9503,
+      "learning_rate": 8.656704544528192e-06,
+      "loss": 0.89,
       "step": 19615
     },
     {
-      "epoch": 0.5387382933728818,
+      "epoch": 0.5566401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 9.232214515420602e-06,
-      "loss": 0.9461,
+      "learning_rate": 8.655793802000905e-06,
+      "loss": 0.8994,
       "step": 19616
     },
     {
-      "epoch": 0.5387657576007251,
+      "epoch": 0.5566685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 9.23132762814612e-06,
-      "loss": 1.0282,
+      "learning_rate": 8.654883070828184e-06,
+      "loss": 0.9828,
       "step": 19617
     },
     {
-      "epoch": 0.5387932218285683,
+      "epoch": 0.5566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 9.230440746953667e-06,
-      "loss": 0.8027,
+      "learning_rate": 8.653972351017732e-06,
+      "loss": 0.8644,
       "step": 19618
     },
     {
-      "epoch": 0.5388206860564115,
+      "epoch": 0.5567253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 9.229553871850262e-06,
-      "loss": 0.9807,
+      "learning_rate": 8.653061642577232e-06,
+      "loss": 0.8209,
       "step": 19619
     },
     {
-      "epoch": 0.5388481502842548,
+      "epoch": 0.5567536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 9.22866700284293e-06,
-      "loss": 0.9383,
+      "learning_rate": 8.652150945514381e-06,
+      "loss": 0.8776,
       "step": 19620
     },
     {
-      "epoch": 0.538875614512098,
+      "epoch": 0.5567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 9.227780139938681e-06,
-      "loss": 0.9144,
+      "learning_rate": 8.651240259836873e-06,
+      "loss": 1.0142,
       "step": 19621
     },
     {
-      "epoch": 0.5389030787399413,
+      "epoch": 0.5568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 9.226893283144534e-06,
-      "loss": 0.9451,
+      "learning_rate": 8.6503295855524e-06,
+      "loss": 0.7823,
       "step": 19622
     },
     {
-      "epoch": 0.5389305429677844,
+      "epoch": 0.556838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 9.226006432467512e-06,
-      "loss": 0.8862,
+      "learning_rate": 8.649418922668649e-06,
+      "loss": 0.8523,
       "step": 19623
     },
     {
-      "epoch": 0.5389580071956277,
+      "epoch": 0.5568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 9.225119587914624e-06,
-      "loss": 0.8887,
+      "learning_rate": 8.648508271193323e-06,
+      "loss": 0.8893,
       "step": 19624
     },
     {
-      "epoch": 0.538985471423471,
+      "epoch": 0.5568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 9.224232749492888e-06,
-      "loss": 0.847,
+      "learning_rate": 8.647597631134105e-06,
+      "loss": 0.8614,
       "step": 19625
     },
     {
-      "epoch": 0.5390129356513141,
+      "epoch": 0.5569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 9.223345917209324e-06,
-      "loss": 0.8196,
+      "learning_rate": 8.646687002498692e-06,
+      "loss": 0.9216,
       "step": 19626
     },
     {
-      "epoch": 0.5390403998791574,
+      "epoch": 0.5569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 9.22245909107095e-06,
-      "loss": 0.8631,
+      "learning_rate": 8.645776385294773e-06,
+      "loss": 0.8928,
       "step": 19627
     },
     {
-      "epoch": 0.5390678641070006,
+      "epoch": 0.5569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 9.221572271084777e-06,
-      "loss": 0.9622,
+      "learning_rate": 8.644865779530042e-06,
+      "loss": 0.9486,
       "step": 19628
     },
     {
-      "epoch": 0.5390953283348439,
+      "epoch": 0.5570090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 9.22068545725783e-06,
-      "loss": 0.8947,
+      "learning_rate": 8.643955185212195e-06,
+      "loss": 0.9393,
       "step": 19629
     },
     {
-      "epoch": 0.5391227925626871,
+      "epoch": 0.5570374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 9.219798649597118e-06,
-      "loss": 0.976,
+      "learning_rate": 8.64304460234891e-06,
+      "loss": 0.9255,
       "step": 19630
     },
     {
-      "epoch": 0.5391502567905303,
+      "epoch": 0.5570658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 9.218911848109663e-06,
-      "loss": 0.9928,
+      "learning_rate": 8.642134030947893e-06,
+      "loss": 0.8802,
       "step": 19631
     },
     {
-      "epoch": 0.5391777210183736,
+      "epoch": 0.557094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 9.21802505280248e-06,
-      "loss": 0.868,
+      "learning_rate": 8.641223471016834e-06,
+      "loss": 0.8992,
       "step": 19632
     },
     {
-      "epoch": 0.5392051852462169,
+      "epoch": 0.5571225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 9.217138263682584e-06,
-      "loss": 1.0662,
+      "learning_rate": 8.64031292256342e-06,
+      "loss": 0.9656,
       "step": 19633
     },
     {
-      "epoch": 0.53923264947406,
+      "epoch": 0.5571509648127129,
       "grad_norm": 0.0,
-      "learning_rate": 9.216251480756993e-06,
-      "loss": 0.9359,
+      "learning_rate": 8.639402385595342e-06,
+      "loss": 0.8546,
       "step": 19634
     },
     {
-      "epoch": 0.5392601137019033,
+      "epoch": 0.5571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 9.215364704032728e-06,
-      "loss": 0.7998,
+      "learning_rate": 8.638491860120295e-06,
+      "loss": 0.8018,
       "step": 19635
     },
     {
-      "epoch": 0.5392875779297465,
+      "epoch": 0.5572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 9.214477933516802e-06,
-      "loss": 0.8569,
+      "learning_rate": 8.637581346145968e-06,
+      "loss": 0.9366,
       "step": 19636
     },
     {
-      "epoch": 0.5393150421575897,
+      "epoch": 0.5572360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 9.213591169216225e-06,
-      "loss": 0.8969,
+      "learning_rate": 8.636670843680051e-06,
+      "loss": 0.7515,
       "step": 19637
     },
     {
-      "epoch": 0.539342506385433,
+      "epoch": 0.5572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 9.212704411138022e-06,
-      "loss": 0.8836,
+      "learning_rate": 8.635760352730239e-06,
+      "loss": 0.8666,
       "step": 19638
     },
     {
-      "epoch": 0.5393699706132762,
+      "epoch": 0.5572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 9.211817659289208e-06,
-      "loss": 0.7607,
+      "learning_rate": 8.634849873304221e-06,
+      "loss": 0.8746,
       "step": 19639
     },
     {
-      "epoch": 0.5393974348411195,
+      "epoch": 0.5573212258796821,
       "grad_norm": 0.0,
-      "learning_rate": 9.210930913676794e-06,
-      "loss": 0.8903,
+      "learning_rate": 8.633939405409685e-06,
+      "loss": 0.767,
       "step": 19640
     },
     {
-      "epoch": 0.5394248990689626,
+      "epoch": 0.5573496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 9.210044174307807e-06,
-      "loss": 0.8084,
+      "learning_rate": 8.633028949054325e-06,
+      "loss": 0.7549,
       "step": 19641
     },
     {
-      "epoch": 0.5394523632968059,
+      "epoch": 0.557377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 9.20915744118925e-06,
-      "loss": 0.9034,
+      "learning_rate": 8.632118504245833e-06,
+      "loss": 0.9644,
       "step": 19642
     },
     {
-      "epoch": 0.5394798275246492,
+      "epoch": 0.5574063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 9.208270714328147e-06,
-      "loss": 0.8887,
+      "learning_rate": 8.631208070991896e-06,
+      "loss": 0.7953,
       "step": 19643
     },
     {
-      "epoch": 0.5395072917524923,
+      "epoch": 0.5574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 9.207383993731516e-06,
-      "loss": 0.8492,
+      "learning_rate": 8.630297649300205e-06,
+      "loss": 0.9016,
       "step": 19644
     },
     {
-      "epoch": 0.5395347559803356,
+      "epoch": 0.5574631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 9.206497279406366e-06,
-      "loss": 0.8337,
+      "learning_rate": 8.62938723917845e-06,
+      "loss": 0.8247,
       "step": 19645
     },
     {
-      "epoch": 0.5395622202081789,
+      "epoch": 0.5574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 9.20561057135972e-06,
-      "loss": 0.904,
+      "learning_rate": 8.628476840634326e-06,
+      "loss": 0.9053,
       "step": 19646
     },
     {
-      "epoch": 0.5395896844360221,
+      "epoch": 0.5575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 9.20472386959859e-06,
-      "loss": 0.8978,
+      "learning_rate": 8.62756645367552e-06,
+      "loss": 0.8666,
       "step": 19647
     },
     {
-      "epoch": 0.5396171486638653,
+      "epoch": 0.5575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 9.203837174129995e-06,
-      "loss": 0.8354,
+      "learning_rate": 8.626656078309718e-06,
+      "loss": 0.9041,
       "step": 19648
     },
     {
-      "epoch": 0.5396446128917085,
+      "epoch": 0.5575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 9.202950484960945e-06,
-      "loss": 0.8011,
+      "learning_rate": 8.625745714544618e-06,
+      "loss": 1.0588,
       "step": 19649
     },
     {
-      "epoch": 0.5396720771195518,
+      "epoch": 0.5576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 9.20206380209846e-06,
-      "loss": 0.9149,
+      "learning_rate": 8.624835362387905e-06,
+      "loss": 0.947,
       "step": 19650
     },
     {
-      "epoch": 0.5396995413473951,
+      "epoch": 0.557633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 9.201177125549556e-06,
-      "loss": 0.8467,
+      "learning_rate": 8.623925021847266e-06,
+      "loss": 0.8692,
       "step": 19651
     },
     {
-      "epoch": 0.5397270055752382,
+      "epoch": 0.5576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 9.200290455321246e-06,
-      "loss": 0.868,
+      "learning_rate": 8.623014692930399e-06,
+      "loss": 0.8179,
       "step": 19652
     },
     {
-      "epoch": 0.5397544698030815,
+      "epoch": 0.5576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 9.199403791420554e-06,
-      "loss": 0.8807,
+      "learning_rate": 8.622104375644986e-06,
+      "loss": 0.8592,
       "step": 19653
     },
     {
-      "epoch": 0.5397819340309247,
+      "epoch": 0.5577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 9.198517133854484e-06,
-      "loss": 0.921,
+      "learning_rate": 8.621194069998718e-06,
+      "loss": 0.7654,
       "step": 19654
     },
     {
-      "epoch": 0.539809398258768,
+      "epoch": 0.5577468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 9.197630482630058e-06,
-      "loss": 0.8685,
+      "learning_rate": 8.62028377599929e-06,
+      "loss": 0.8228,
       "step": 19655
     },
     {
-      "epoch": 0.5398368624866112,
+      "epoch": 0.5577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 9.19674383775429e-06,
-      "loss": 0.8406,
+      "learning_rate": 8.619373493654383e-06,
+      "loss": 0.8874,
       "step": 19656
     },
     {
-      "epoch": 0.5398643267144544,
+      "epoch": 0.5578036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 9.195857199234194e-06,
-      "loss": 0.9005,
+      "learning_rate": 8.618463222971692e-06,
+      "loss": 0.9047,
       "step": 19657
     },
     {
-      "epoch": 0.5398917909422977,
+      "epoch": 0.5578320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 9.194970567076793e-06,
-      "loss": 0.9074,
+      "learning_rate": 8.617552963958904e-06,
+      "loss": 0.9363,
       "step": 19658
     },
     {
-      "epoch": 0.5399192551701409,
+      "epoch": 0.5578603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 9.194083941289092e-06,
-      "loss": 0.8233,
+      "learning_rate": 8.616642716623709e-06,
+      "loss": 0.877,
       "step": 19659
     },
     {
-      "epoch": 0.5399467193979841,
+      "epoch": 0.55788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 9.193197321878114e-06,
-      "loss": 0.8462,
+      "learning_rate": 8.615732480973794e-06,
+      "loss": 0.9332,
       "step": 19660
     },
     {
-      "epoch": 0.5399741836258274,
+      "epoch": 0.557917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 9.192310708850868e-06,
-      "loss": 0.9326,
+      "learning_rate": 8.614822257016845e-06,
+      "loss": 0.8665,
       "step": 19661
     },
     {
-      "epoch": 0.5400016478536706,
+      "epoch": 0.5579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 9.191424102214373e-06,
-      "loss": 0.7485,
+      "learning_rate": 8.61391204476056e-06,
+      "loss": 0.785,
       "step": 19662
     },
     {
-      "epoch": 0.5400291120815138,
+      "epoch": 0.5579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 9.190537501975645e-06,
-      "loss": 0.8149,
+      "learning_rate": 8.613001844212622e-06,
+      "loss": 0.832,
       "step": 19663
     },
     {
-      "epoch": 0.5400565763093571,
+      "epoch": 0.5580022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 9.189650908141702e-06,
-      "loss": 0.906,
+      "learning_rate": 8.612091655380718e-06,
+      "loss": 0.894,
       "step": 19664
     },
     {
-      "epoch": 0.5400840405372003,
+      "epoch": 0.5580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 9.188764320719548e-06,
-      "loss": 0.879,
+      "learning_rate": 8.611181478272539e-06,
+      "loss": 0.868,
       "step": 19665
     },
     {
-      "epoch": 0.5401115047650435,
+      "epoch": 0.5580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 9.187877739716208e-06,
-      "loss": 0.8812,
+      "learning_rate": 8.610271312895773e-06,
+      "loss": 0.7769,
       "step": 19666
     },
     {
-      "epoch": 0.5401389689928867,
+      "epoch": 0.5580874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 9.186991165138694e-06,
-      "loss": 1.0289,
+      "learning_rate": 8.609361159258109e-06,
+      "loss": 1.0015,
       "step": 19667
     },
     {
-      "epoch": 0.54016643322073,
+      "epoch": 0.5581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 9.18610459699402e-06,
-      "loss": 0.8568,
+      "learning_rate": 8.608451017367231e-06,
+      "loss": 0.881,
       "step": 19668
     },
     {
-      "epoch": 0.5401938974485733,
+      "epoch": 0.5581441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 9.185218035289199e-06,
-      "loss": 0.8833,
+      "learning_rate": 8.607540887230832e-06,
+      "loss": 0.83,
       "step": 19669
     },
     {
-      "epoch": 0.5402213616764164,
+      "epoch": 0.5581725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 9.184331480031255e-06,
-      "loss": 0.8507,
+      "learning_rate": 8.606630768856596e-06,
+      "loss": 0.9262,
       "step": 19670
     },
     {
-      "epoch": 0.5402488259042597,
+      "epoch": 0.5582009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 9.183444931227192e-06,
-      "loss": 0.9383,
+      "learning_rate": 8.605720662252213e-06,
+      "loss": 0.8589,
       "step": 19671
     },
     {
-      "epoch": 0.540276290132103,
+      "epoch": 0.5582292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 9.182558388884032e-06,
-      "loss": 0.8513,
+      "learning_rate": 8.60481056742537e-06,
+      "loss": 0.8919,
       "step": 19672
     },
     {
-      "epoch": 0.5403037543599462,
+      "epoch": 0.5582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 9.181671853008783e-06,
-      "loss": 0.7676,
+      "learning_rate": 8.603900484383757e-06,
+      "loss": 0.8763,
       "step": 19673
     },
     {
-      "epoch": 0.5403312185877894,
+      "epoch": 0.5582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 9.180785323608462e-06,
-      "loss": 0.9757,
+      "learning_rate": 8.602990413135056e-06,
+      "loss": 0.8905,
       "step": 19674
     },
     {
-      "epoch": 0.5403586828156326,
+      "epoch": 0.5583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 9.179898800690089e-06,
-      "loss": 0.8229,
+      "learning_rate": 8.602080353686961e-06,
+      "loss": 0.9446,
       "step": 19675
     },
     {
-      "epoch": 0.5403861470434759,
+      "epoch": 0.5583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 9.179012284260677e-06,
-      "loss": 0.8595,
+      "learning_rate": 8.60117030604715e-06,
+      "loss": 0.8485,
       "step": 19676
     },
     {
-      "epoch": 0.5404136112713191,
+      "epoch": 0.5583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 9.178125774327232e-06,
-      "loss": 0.9306,
+      "learning_rate": 8.600260270223322e-06,
+      "loss": 0.7754,
       "step": 19677
     },
     {
-      "epoch": 0.5404410754991623,
+      "epoch": 0.5583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 9.177239270896775e-06,
-      "loss": 0.9339,
+      "learning_rate": 8.599350246223156e-06,
+      "loss": 0.8344,
       "step": 19678
     },
     {
-      "epoch": 0.5404685397270056,
+      "epoch": 0.558427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 9.176352773976321e-06,
-      "loss": 0.8683,
+      "learning_rate": 8.598440234054345e-06,
+      "loss": 0.8897,
       "step": 19679
     },
     {
-      "epoch": 0.5404960039548488,
+      "epoch": 0.5584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 9.175466283572883e-06,
-      "loss": 0.9561,
+      "learning_rate": 8.59753023372457e-06,
+      "loss": 0.9573,
       "step": 19680
     },
     {
-      "epoch": 0.540523468182692,
+      "epoch": 0.5584846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 9.174579799693473e-06,
-      "loss": 0.8279,
+      "learning_rate": 8.596620245241518e-06,
+      "loss": 0.8648,
       "step": 19681
     },
     {
-      "epoch": 0.5405509324105353,
+      "epoch": 0.5585130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 9.173693322345113e-06,
-      "loss": 1.0242,
+      "learning_rate": 8.595710268612881e-06,
+      "loss": 0.9361,
       "step": 19682
     },
     {
-      "epoch": 0.5405783966383785,
+      "epoch": 0.5585414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 9.172806851534806e-06,
-      "loss": 0.8895,
+      "learning_rate": 8.594800303846342e-06,
+      "loss": 0.8633,
       "step": 19683
     },
     {
-      "epoch": 0.5406058608662218,
+      "epoch": 0.5585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 9.171920387269573e-06,
-      "loss": 0.9035,
+      "learning_rate": 8.593890350949587e-06,
+      "loss": 0.9591,
       "step": 19684
     },
     {
-      "epoch": 0.540633325094065,
+      "epoch": 0.5585981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 9.171033929556426e-06,
-      "loss": 0.8749,
+      "learning_rate": 8.592980409930305e-06,
+      "loss": 0.778,
       "step": 19685
     },
     {
-      "epoch": 0.5406607893219082,
+      "epoch": 0.5586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 9.17014747840238e-06,
-      "loss": 1.0063,
+      "learning_rate": 8.59207048079618e-06,
+      "loss": 0.7578,
       "step": 19686
     },
     {
-      "epoch": 0.5406882535497515,
+      "epoch": 0.5586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 9.16926103381445e-06,
-      "loss": 0.9633,
+      "learning_rate": 8.591160563554898e-06,
+      "loss": 0.8404,
       "step": 19687
     },
     {
-      "epoch": 0.5407157177775946,
+      "epoch": 0.558683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.16837459579965e-06,
-      "loss": 1.0453,
+      "learning_rate": 8.590250658214148e-06,
+      "loss": 0.9672,
       "step": 19688
     },
     {
-      "epoch": 0.5407431820054379,
+      "epoch": 0.5587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 9.167488164364989e-06,
-      "loss": 0.9063,
+      "learning_rate": 8.589340764781613e-06,
+      "loss": 0.8436,
       "step": 19689
     },
     {
-      "epoch": 0.5407706462332812,
+      "epoch": 0.5587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 9.166601739517484e-06,
-      "loss": 0.915,
+      "learning_rate": 8.588430883264977e-06,
+      "loss": 0.9593,
       "step": 19690
     },
     {
-      "epoch": 0.5407981104611244,
+      "epoch": 0.5587684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 9.16571532126415e-06,
-      "loss": 0.8948,
+      "learning_rate": 8.587521013671931e-06,
+      "loss": 0.9273,
       "step": 19691
     },
     {
-      "epoch": 0.5408255746889676,
+      "epoch": 0.5587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 9.164828909612e-06,
-      "loss": 0.8421,
+      "learning_rate": 8.586611156010159e-06,
+      "loss": 0.8643,
       "step": 19692
     },
     {
-      "epoch": 0.5408530389168108,
+      "epoch": 0.5588251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 9.163942504568045e-06,
-      "loss": 0.8658,
+      "learning_rate": 8.585701310287346e-06,
+      "loss": 0.9116,
       "step": 19693
     },
     {
-      "epoch": 0.5408805031446541,
+      "epoch": 0.5588535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 9.163056106139305e-06,
-      "loss": 0.9528,
+      "learning_rate": 8.584791476511178e-06,
+      "loss": 0.7449,
       "step": 19694
     },
     {
-      "epoch": 0.5409079673724974,
+      "epoch": 0.5588819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 9.162169714332787e-06,
-      "loss": 0.9271,
+      "learning_rate": 8.583881654689338e-06,
+      "loss": 0.9044,
       "step": 19695
     },
     {
-      "epoch": 0.5409354316003405,
+      "epoch": 0.5589103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 9.161283329155507e-06,
-      "loss": 0.92,
+      "learning_rate": 8.582971844829518e-06,
+      "loss": 0.9501,
       "step": 19696
     },
     {
-      "epoch": 0.5409628958281838,
+      "epoch": 0.558938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 9.160396950614476e-06,
-      "loss": 0.8602,
+      "learning_rate": 8.582062046939396e-06,
+      "loss": 0.8102,
       "step": 19697
     },
     {
-      "epoch": 0.5409903600560271,
+      "epoch": 0.5589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 9.159510578716711e-06,
-      "loss": 0.9463,
+      "learning_rate": 8.581152261026657e-06,
+      "loss": 0.8457,
       "step": 19698
     },
     {
-      "epoch": 0.5410178242838702,
+      "epoch": 0.5589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 9.158624213469224e-06,
-      "loss": 0.8379,
+      "learning_rate": 8.580242487098993e-06,
+      "loss": 0.9282,
       "step": 19699
     },
     {
-      "epoch": 0.5410452885117135,
+      "epoch": 0.5590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 9.157737854879032e-06,
-      "loss": 0.7983,
+      "learning_rate": 8.579332725164081e-06,
+      "loss": 0.9496,
       "step": 19700
     },
     {
-      "epoch": 0.5410727527395567,
+      "epoch": 0.5590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 9.156851502953139e-06,
-      "loss": 0.7879,
+      "learning_rate": 8.57842297522961e-06,
+      "loss": 0.8716,
       "step": 19701
     },
     {
-      "epoch": 0.5411002169674,
+      "epoch": 0.5590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 9.155965157698564e-06,
-      "loss": 0.9285,
+      "learning_rate": 8.577513237303266e-06,
+      "loss": 0.7608,
       "step": 19702
     },
     {
-      "epoch": 0.5411276811952432,
+      "epoch": 0.5591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 9.155078819122322e-06,
-      "loss": 0.9336,
+      "learning_rate": 8.576603511392732e-06,
+      "loss": 0.8238,
       "step": 19703
     },
     {
-      "epoch": 0.5411551454230864,
+      "epoch": 0.5591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 9.154192487231421e-06,
-      "loss": 0.8546,
+      "learning_rate": 8.575693797505688e-06,
+      "loss": 0.9429,
       "step": 19704
     },
     {
-      "epoch": 0.5411826096509297,
+      "epoch": 0.5591657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 9.153306162032877e-06,
-      "loss": 1.0142,
+      "learning_rate": 8.574784095649826e-06,
+      "loss": 0.9139,
       "step": 19705
     },
     {
-      "epoch": 0.5412100738787728,
+      "epoch": 0.5591940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 9.152419843533706e-06,
-      "loss": 0.9056,
+      "learning_rate": 8.573874405832828e-06,
+      "loss": 0.9466,
       "step": 19706
     },
     {
-      "epoch": 0.5412375381066161,
+      "epoch": 0.55922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.151533531740915e-06,
-      "loss": 0.9506,
+      "learning_rate": 8.57296472806237e-06,
+      "loss": 0.8595,
       "step": 19707
     },
     {
-      "epoch": 0.5412650023344594,
+      "epoch": 0.5592508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 9.150647226661521e-06,
-      "loss": 0.7968,
+      "learning_rate": 8.572055062346149e-06,
+      "loss": 0.8753,
       "step": 19708
     },
     {
-      "epoch": 0.5412924665623026,
+      "epoch": 0.5592792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 9.149760928302532e-06,
-      "loss": 0.8925,
+      "learning_rate": 8.571145408691842e-06,
+      "loss": 0.891,
       "step": 19709
     },
     {
-      "epoch": 0.5413199307901458,
+      "epoch": 0.5593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 9.148874636670965e-06,
-      "loss": 0.9288,
+      "learning_rate": 8.570235767107134e-06,
+      "loss": 0.993,
       "step": 19710
     },
     {
-      "epoch": 0.5413473950179891,
+      "epoch": 0.5593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 9.147988351773833e-06,
-      "loss": 1.0159,
+      "learning_rate": 8.56932613759971e-06,
+      "loss": 0.8449,
       "step": 19711
     },
     {
-      "epoch": 0.5413748592458323,
+      "epoch": 0.5593643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 9.147102073618149e-06,
-      "loss": 0.8767,
+      "learning_rate": 8.568416520177249e-06,
+      "loss": 0.9619,
       "step": 19712
     },
     {
-      "epoch": 0.5414023234736756,
+      "epoch": 0.5593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 9.14621580221092e-06,
-      "loss": 0.8354,
+      "learning_rate": 8.56750691484744e-06,
+      "loss": 0.9083,
       "step": 19713
     },
     {
-      "epoch": 0.5414297877015187,
+      "epoch": 0.5594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.145329537559161e-06,
-      "loss": 0.8699,
+      "learning_rate": 8.566597321617967e-06,
+      "loss": 0.7984,
       "step": 19714
     },
     {
-      "epoch": 0.541457251929362,
+      "epoch": 0.559449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 9.14444327966989e-06,
-      "loss": 0.8865,
+      "learning_rate": 8.565687740496507e-06,
+      "loss": 0.9401,
       "step": 19715
     },
     {
-      "epoch": 0.5414847161572053,
+      "epoch": 0.559477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 9.143557028550114e-06,
-      "loss": 0.8555,
+      "learning_rate": 8.564778171490749e-06,
+      "loss": 0.8146,
       "step": 19716
     },
     {
-      "epoch": 0.5415121803850484,
+      "epoch": 0.5595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 9.14267078420685e-06,
-      "loss": 0.8673,
+      "learning_rate": 8.563868614608376e-06,
+      "loss": 0.8837,
       "step": 19717
     },
     {
-      "epoch": 0.5415396446128917,
+      "epoch": 0.5595346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 9.1417845466471e-06,
-      "loss": 0.8069,
+      "learning_rate": 8.562959069857064e-06,
+      "loss": 0.7965,
       "step": 19718
     },
     {
-      "epoch": 0.5415671088407349,
+      "epoch": 0.5595629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 9.140898315877885e-06,
-      "loss": 0.8342,
+      "learning_rate": 8.562049537244505e-06,
+      "loss": 0.8958,
       "step": 19719
     },
     {
-      "epoch": 0.5415945730685782,
+      "epoch": 0.5595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 9.140012091906217e-06,
-      "loss": 0.8184,
+      "learning_rate": 8.56114001677838e-06,
+      "loss": 0.7689,
       "step": 19720
     },
     {
-      "epoch": 0.5416220372964214,
+      "epoch": 0.5596197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 9.139125874739104e-06,
-      "loss": 0.9134,
+      "learning_rate": 8.560230508466367e-06,
+      "loss": 0.8762,
       "step": 19721
     },
     {
-      "epoch": 0.5416495015242646,
+      "epoch": 0.5596481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 9.138239664383562e-06,
-      "loss": 0.8252,
+      "learning_rate": 8.559321012316149e-06,
+      "loss": 0.8172,
       "step": 19722
     },
     {
-      "epoch": 0.5416769657521079,
+      "epoch": 0.5596765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 9.137353460846604e-06,
-      "loss": 0.7562,
+      "learning_rate": 8.558411528335414e-06,
+      "loss": 0.918,
       "step": 19723
     },
     {
-      "epoch": 0.5417044299799512,
+      "epoch": 0.5597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 9.136467264135238e-06,
-      "loss": 0.9486,
+      "learning_rate": 8.557502056531844e-06,
+      "loss": 0.8676,
       "step": 19724
     },
     {
-      "epoch": 0.5417318942077943,
+      "epoch": 0.559733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 9.135581074256477e-06,
-      "loss": 0.9223,
+      "learning_rate": 8.556592596913115e-06,
+      "loss": 0.8188,
       "step": 19725
     },
     {
-      "epoch": 0.5417593584356376,
+      "epoch": 0.5597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 9.134694891217331e-06,
-      "loss": 0.7778,
+      "learning_rate": 8.555683149486916e-06,
+      "loss": 0.8626,
       "step": 19726
     },
     {
-      "epoch": 0.5417868226634808,
+      "epoch": 0.5597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 9.133808715024818e-06,
-      "loss": 0.9157,
+      "learning_rate": 8.554773714260927e-06,
+      "loss": 0.8607,
       "step": 19727
     },
     {
-      "epoch": 0.541814286891324,
+      "epoch": 0.5598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 9.132922545685944e-06,
-      "loss": 1.0031,
+      "learning_rate": 8.553864291242827e-06,
+      "loss": 0.8007,
       "step": 19728
     },
     {
-      "epoch": 0.5418417511191673,
+      "epoch": 0.5598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 9.132036383207726e-06,
-      "loss": 0.8921,
+      "learning_rate": 8.552954880440303e-06,
+      "loss": 0.8333,
       "step": 19729
     },
     {
-      "epoch": 0.5418692153470105,
+      "epoch": 0.5598751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 9.131150227597168e-06,
-      "loss": 0.8994,
+      "learning_rate": 8.552045481861033e-06,
+      "loss": 0.8572,
       "step": 19730
     },
     {
-      "epoch": 0.5418966795748538,
+      "epoch": 0.5599035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 9.130264078861286e-06,
-      "loss": 0.8727,
+      "learning_rate": 8.551136095512698e-06,
+      "loss": 0.9197,
       "step": 19731
     },
     {
-      "epoch": 0.5419241438026969,
+      "epoch": 0.5599318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 9.129377937007094e-06,
-      "loss": 0.9404,
+      "learning_rate": 8.550226721402985e-06,
+      "loss": 0.9348,
       "step": 19732
     },
     {
-      "epoch": 0.5419516080305402,
+      "epoch": 0.5599602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 9.128491802041599e-06,
-      "loss": 0.9254,
+      "learning_rate": 8.549317359539571e-06,
+      "loss": 0.8363,
       "step": 19733
     },
     {
-      "epoch": 0.5419790722583835,
+      "epoch": 0.559988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 9.127605673971815e-06,
-      "loss": 0.8435,
+      "learning_rate": 8.548408009930138e-06,
+      "loss": 0.8768,
       "step": 19734
     },
     {
-      "epoch": 0.5420065364862267,
+      "epoch": 0.5600170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 9.126719552804758e-06,
-      "loss": 0.8534,
+      "learning_rate": 8.547498672582367e-06,
+      "loss": 0.8532,
       "step": 19735
     },
     {
-      "epoch": 0.5420340007140699,
+      "epoch": 0.5600454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 9.12583343854743e-06,
-      "loss": 0.8131,
+      "learning_rate": 8.546589347503939e-06,
+      "loss": 0.8808,
       "step": 19736
     },
     {
-      "epoch": 0.5420614649419132,
+      "epoch": 0.5600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 9.124947331206843e-06,
-      "loss": 0.985,
+      "learning_rate": 8.545680034702535e-06,
+      "loss": 0.8121,
       "step": 19737
     },
     {
-      "epoch": 0.5420889291697564,
+      "epoch": 0.5601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 9.124061230790015e-06,
-      "loss": 0.8954,
+      "learning_rate": 8.54477073418584e-06,
+      "loss": 0.8786,
       "step": 19738
     },
     {
-      "epoch": 0.5421163933975996,
+      "epoch": 0.5601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 9.123175137303954e-06,
-      "loss": 0.8872,
+      "learning_rate": 8.543861445961532e-06,
+      "loss": 0.9344,
       "step": 19739
     },
     {
-      "epoch": 0.5421438576254428,
+      "epoch": 0.5601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 9.122289050755668e-06,
-      "loss": 0.8964,
+      "learning_rate": 8.542952170037292e-06,
+      "loss": 0.8518,
       "step": 19740
     },
     {
-      "epoch": 0.5421713218532861,
+      "epoch": 0.5601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 9.121402971152177e-06,
-      "loss": 0.8527,
+      "learning_rate": 8.5420429064208e-06,
+      "loss": 0.7996,
       "step": 19741
     },
     {
-      "epoch": 0.5421987860811294,
+      "epoch": 0.5602156640181611,
       "grad_norm": 0.0,
-      "learning_rate": 9.120516898500482e-06,
-      "loss": 0.925,
+      "learning_rate": 8.541133655119736e-06,
+      "loss": 0.8873,
       "step": 19742
     },
     {
-      "epoch": 0.5422262503089725,
+      "epoch": 0.5602440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 9.119630832807597e-06,
-      "loss": 0.9908,
+      "learning_rate": 8.540224416141784e-06,
+      "loss": 0.8652,
       "step": 19743
     },
     {
-      "epoch": 0.5422537145368158,
+      "epoch": 0.560272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 9.118744774080537e-06,
-      "loss": 0.9398,
+      "learning_rate": 8.539315189494622e-06,
+      "loss": 0.801,
       "step": 19744
     },
     {
-      "epoch": 0.542281178764659,
+      "epoch": 0.5603007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 9.117858722326305e-06,
-      "loss": 0.8624,
+      "learning_rate": 8.538405975185927e-06,
+      "loss": 0.9385,
       "step": 19745
     },
     {
-      "epoch": 0.5423086429925023,
+      "epoch": 0.5603291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 9.11697267755192e-06,
-      "loss": 0.8744,
+      "learning_rate": 8.537496773223386e-06,
+      "loss": 0.8461,
       "step": 19746
     },
     {
-      "epoch": 0.5423361072203455,
+      "epoch": 0.5603575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 9.116086639764392e-06,
-      "loss": 0.927,
+      "learning_rate": 8.536587583614674e-06,
+      "loss": 0.9406,
       "step": 19747
     },
     {
-      "epoch": 0.5423635714481887,
+      "epoch": 0.5603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 9.115200608970726e-06,
-      "loss": 0.9576,
+      "learning_rate": 8.535678406367472e-06,
+      "loss": 0.8384,
       "step": 19748
     },
     {
-      "epoch": 0.542391035676032,
+      "epoch": 0.5604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 9.114314585177935e-06,
-      "loss": 0.9223,
+      "learning_rate": 8.53476924148946e-06,
+      "loss": 0.7628,
       "step": 19749
     },
     {
-      "epoch": 0.5424184999038753,
+      "epoch": 0.5604426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 9.113428568393028e-06,
-      "loss": 0.925,
+      "learning_rate": 8.533860088988321e-06,
+      "loss": 0.8678,
       "step": 19750
     },
     {
-      "epoch": 0.5424459641317184,
+      "epoch": 0.5604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 9.11254255862302e-06,
-      "loss": 0.895,
+      "learning_rate": 8.532950948871728e-06,
+      "loss": 0.8736,
       "step": 19751
     },
     {
-      "epoch": 0.5424734283595617,
+      "epoch": 0.5604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 9.111656555874919e-06,
-      "loss": 0.8443,
+      "learning_rate": 8.532041821147366e-06,
+      "loss": 0.9065,
       "step": 19752
     },
     {
-      "epoch": 0.5425008925874049,
+      "epoch": 0.560527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 9.110770560155738e-06,
-      "loss": 0.8364,
+      "learning_rate": 8.531132705822908e-06,
+      "loss": 0.8091,
       "step": 19753
     },
     {
-      "epoch": 0.5425283568152481,
+      "epoch": 0.5605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 9.10988457147248e-06,
-      "loss": 0.8452,
+      "learning_rate": 8.530223602906044e-06,
+      "loss": 0.7849,
       "step": 19754
     },
     {
-      "epoch": 0.5425558210430914,
+      "epoch": 0.5605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 9.108998589832162e-06,
-      "loss": 0.7245,
+      "learning_rate": 8.529314512404448e-06,
+      "loss": 0.9231,
       "step": 19755
     },
     {
-      "epoch": 0.5425832852709346,
+      "epoch": 0.5606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 9.108112615241795e-06,
-      "loss": 0.9455,
+      "learning_rate": 8.528405434325795e-06,
+      "loss": 0.773,
       "step": 19756
     },
     {
-      "epoch": 0.5426107494987779,
+      "epoch": 0.5606413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 9.107226647708381e-06,
-      "loss": 0.8375,
+      "learning_rate": 8.527496368677768e-06,
+      "loss": 0.8946,
       "step": 19757
     },
     {
-      "epoch": 0.542638213726621,
+      "epoch": 0.5606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 9.106340687238939e-06,
-      "loss": 0.9059,
+      "learning_rate": 8.526587315468048e-06,
+      "loss": 0.8746,
       "step": 19758
     },
     {
-      "epoch": 0.5426656779544643,
+      "epoch": 0.5606980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 9.105454733840477e-06,
-      "loss": 0.7859,
+      "learning_rate": 8.525678274704308e-06,
+      "loss": 0.8947,
       "step": 19759
     },
     {
-      "epoch": 0.5426931421823076,
+      "epoch": 0.5607264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 9.104568787520002e-06,
-      "loss": 0.8489,
+      "learning_rate": 8.524769246394233e-06,
+      "loss": 0.9676,
       "step": 19760
     },
     {
-      "epoch": 0.5427206064101507,
+      "epoch": 0.5607548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 9.103682848284525e-06,
-      "loss": 0.8306,
+      "learning_rate": 8.523860230545496e-06,
+      "loss": 0.8654,
       "step": 19761
     },
     {
-      "epoch": 0.542748070637994,
+      "epoch": 0.560783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 9.102796916141057e-06,
-      "loss": 0.9128,
+      "learning_rate": 8.522951227165779e-06,
+      "loss": 0.7633,
       "step": 19762
     },
     {
-      "epoch": 0.5427755348658373,
+      "epoch": 0.560811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 9.101910991096608e-06,
-      "loss": 0.8679,
+      "learning_rate": 8.52204223626276e-06,
+      "loss": 0.8397,
       "step": 19763
     },
     {
-      "epoch": 0.5428029990936805,
+      "epoch": 0.5608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 9.101025073158185e-06,
-      "loss": 0.8622,
+      "learning_rate": 8.521133257844116e-06,
+      "loss": 0.8551,
       "step": 19764
     },
     {
-      "epoch": 0.5428304633215237,
+      "epoch": 0.5608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.100139162332805e-06,
-      "loss": 1.0379,
+      "learning_rate": 8.520224291917524e-06,
+      "loss": 0.8907,
       "step": 19765
     },
     {
-      "epoch": 0.5428579275493669,
+      "epoch": 0.5608967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 9.099253258627468e-06,
-      "loss": 0.8692,
+      "learning_rate": 8.519315338490667e-06,
+      "loss": 0.8532,
       "step": 19766
     },
     {
-      "epoch": 0.5428853917772102,
+      "epoch": 0.5609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 9.098367362049187e-06,
-      "loss": 0.9765,
+      "learning_rate": 8.518406397571219e-06,
+      "loss": 0.796,
       "step": 19767
     },
     {
-      "epoch": 0.5429128560050535,
+      "epoch": 0.5609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 9.097481472604976e-06,
-      "loss": 0.8626,
+      "learning_rate": 8.517497469166855e-06,
+      "loss": 0.8224,
       "step": 19768
     },
     {
-      "epoch": 0.5429403202328966,
+      "epoch": 0.5609818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 9.096595590301838e-06,
-      "loss": 0.9728,
+      "learning_rate": 8.516588553285258e-06,
+      "loss": 0.9475,
       "step": 19769
     },
     {
-      "epoch": 0.5429677844607399,
+      "epoch": 0.5610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 9.095709715146786e-06,
-      "loss": 1.0402,
+      "learning_rate": 8.515679649934106e-06,
+      "loss": 0.8914,
       "step": 19770
     },
     {
-      "epoch": 0.5429952486885831,
+      "epoch": 0.5610385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 9.094823847146834e-06,
-      "loss": 0.8403,
+      "learning_rate": 8.514770759121076e-06,
+      "loss": 0.776,
       "step": 19771
     },
     {
-      "epoch": 0.5430227129164263,
+      "epoch": 0.5610669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 9.093937986308983e-06,
-      "loss": 0.9387,
+      "learning_rate": 8.513861880853843e-06,
+      "loss": 0.9432,
       "step": 19772
     },
     {
-      "epoch": 0.5430501771442696,
+      "epoch": 0.5610953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 9.093052132640245e-06,
-      "loss": 0.8651,
+      "learning_rate": 8.512953015140085e-06,
+      "loss": 0.8304,
       "step": 19773
     },
     {
-      "epoch": 0.5430776413721128,
+      "epoch": 0.5611237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 9.092166286147631e-06,
-      "loss": 0.8391,
+      "learning_rate": 8.51204416198748e-06,
+      "loss": 0.9437,
       "step": 19774
     },
     {
-      "epoch": 0.5431051055999561,
+      "epoch": 0.5611520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 9.091280446838148e-06,
-      "loss": 0.8047,
+      "learning_rate": 8.511135321403705e-06,
+      "loss": 0.8545,
       "step": 19775
     },
     {
-      "epoch": 0.5431325698277993,
+      "epoch": 0.5611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 9.090394614718805e-06,
-      "loss": 0.8953,
+      "learning_rate": 8.510226493396436e-06,
+      "loss": 0.9092,
       "step": 19776
     },
     {
-      "epoch": 0.5431600340556425,
+      "epoch": 0.5612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 9.089508789796617e-06,
-      "loss": 0.865,
+      "learning_rate": 8.509317677973351e-06,
+      "loss": 0.8788,
       "step": 19777
     },
     {
-      "epoch": 0.5431874982834858,
+      "epoch": 0.5612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 9.088622972078583e-06,
-      "loss": 0.9011,
+      "learning_rate": 8.50840887514213e-06,
+      "loss": 0.9969,
       "step": 19778
     },
     {
-      "epoch": 0.543214962511329,
+      "epoch": 0.5612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 9.087737161571718e-06,
-      "loss": 0.833,
+      "learning_rate": 8.507500084910441e-06,
+      "loss": 0.7974,
       "step": 19779
     },
     {
-      "epoch": 0.5432424267391722,
+      "epoch": 0.5612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 9.086851358283033e-06,
-      "loss": 0.9347,
+      "learning_rate": 8.50659130728597e-06,
+      "loss": 0.7878,
       "step": 19780
     },
     {
-      "epoch": 0.5432698909670155,
+      "epoch": 0.561322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 9.08596556221953e-06,
-      "loss": 0.7866,
+      "learning_rate": 8.505682542276388e-06,
+      "loss": 0.8099,
       "step": 19781
     },
     {
-      "epoch": 0.5432973551948587,
+      "epoch": 0.5613507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 9.085079773388226e-06,
-      "loss": 0.8491,
+      "learning_rate": 8.504773789889372e-06,
+      "loss": 0.9943,
       "step": 19782
     },
     {
-      "epoch": 0.5433248194227019,
+      "epoch": 0.5613791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 9.08419399179612e-06,
-      "loss": 0.8561,
+      "learning_rate": 8.503865050132596e-06,
+      "loss": 0.7934,
       "step": 19783
     },
     {
-      "epoch": 0.5433522836505452,
+      "epoch": 0.5614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 9.083308217450229e-06,
-      "loss": 0.878,
+      "learning_rate": 8.502956323013742e-06,
+      "loss": 0.8611,
       "step": 19784
     },
     {
-      "epoch": 0.5433797478783884,
+      "epoch": 0.5614358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 9.082422450357557e-06,
-      "loss": 0.857,
+      "learning_rate": 8.502047608540485e-06,
+      "loss": 0.8788,
       "step": 19785
     },
     {
-      "epoch": 0.5434072121062317,
+      "epoch": 0.5614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 9.081536690525114e-06,
-      "loss": 0.8433,
+      "learning_rate": 8.501138906720496e-06,
+      "loss": 0.923,
       "step": 19786
     },
     {
-      "epoch": 0.5434346763340748,
+      "epoch": 0.5614926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 9.08065093795991e-06,
-      "loss": 0.8118,
+      "learning_rate": 8.500230217561459e-06,
+      "loss": 0.8631,
       "step": 19787
     },
     {
-      "epoch": 0.5434621405619181,
+      "epoch": 0.5615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 9.079765192668955e-06,
-      "loss": 0.8743,
+      "learning_rate": 8.499321541071041e-06,
+      "loss": 0.895,
       "step": 19788
     },
     {
-      "epoch": 0.5434896047897614,
+      "epoch": 0.5615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 9.078879454659249e-06,
-      "loss": 0.9264,
+      "learning_rate": 8.49841287725692e-06,
+      "loss": 0.869,
       "step": 19789
     },
     {
-      "epoch": 0.5435170690176045,
+      "epoch": 0.561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 9.077993723937806e-06,
-      "loss": 0.8654,
+      "learning_rate": 8.497504226126777e-06,
+      "loss": 0.886,
       "step": 19790
     },
     {
-      "epoch": 0.5435445332454478,
+      "epoch": 0.5616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 9.077108000511633e-06,
-      "loss": 0.874,
+      "learning_rate": 8.496595587688281e-06,
+      "loss": 0.8309,
       "step": 19791
     },
     {
-      "epoch": 0.543571997473291,
+      "epoch": 0.5616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 9.07622228438774e-06,
-      "loss": 0.8136,
+      "learning_rate": 8.49568696194911e-06,
+      "loss": 0.7988,
       "step": 19792
     },
     {
-      "epoch": 0.5435994617011343,
+      "epoch": 0.5616628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 9.075336575573134e-06,
-      "loss": 0.9339,
+      "learning_rate": 8.49477834891694e-06,
+      "loss": 0.8547,
       "step": 19793
     },
     {
-      "epoch": 0.5436269259289775,
+      "epoch": 0.5616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 9.074450874074829e-06,
-      "loss": 0.8372,
+      "learning_rate": 8.493869748599445e-06,
+      "loss": 0.8859,
       "step": 19794
     },
     {
-      "epoch": 0.5436543901568207,
+      "epoch": 0.5617196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 9.07356517989982e-06,
-      "loss": 0.894,
+      "learning_rate": 8.492961161004297e-06,
+      "loss": 0.8727,
       "step": 19795
     },
     {
-      "epoch": 0.543681854384664,
+      "epoch": 0.5617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 9.072679493055125e-06,
-      "loss": 0.8696,
+      "learning_rate": 8.492052586139176e-06,
+      "loss": 0.9245,
       "step": 19796
     },
     {
-      "epoch": 0.5437093186125073,
+      "epoch": 0.5617763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 9.071793813547749e-06,
-      "loss": 0.9082,
+      "learning_rate": 8.491144024011755e-06,
+      "loss": 0.9084,
       "step": 19797
     },
     {
-      "epoch": 0.5437367828403504,
+      "epoch": 0.5618047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 9.070908141384698e-06,
-      "loss": 0.9755,
+      "learning_rate": 8.490235474629707e-06,
+      "loss": 0.9013,
       "step": 19798
     },
     {
-      "epoch": 0.5437642470681937,
+      "epoch": 0.56183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 9.070022476572982e-06,
-      "loss": 0.889,
+      "learning_rate": 8.489326938000708e-06,
+      "loss": 0.992,
       "step": 19799
     },
     {
-      "epoch": 0.5437917112960369,
+      "epoch": 0.5618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 9.069136819119615e-06,
-      "loss": 0.9205,
+      "learning_rate": 8.48841841413243e-06,
+      "loss": 0.9526,
       "step": 19800
     },
     {
-      "epoch": 0.5438191755238801,
+      "epoch": 0.5618898978433599,
       "grad_norm": 0.0,
-      "learning_rate": 9.06825116903159e-06,
-      "loss": 0.8308,
+      "learning_rate": 8.487509903032553e-06,
+      "loss": 0.8531,
       "step": 19801
     },
     {
-      "epoch": 0.5438466397517234,
+      "epoch": 0.5619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 9.067365526315924e-06,
-      "loss": 0.8731,
+      "learning_rate": 8.486601404708748e-06,
+      "loss": 0.9333,
       "step": 19802
     },
     {
-      "epoch": 0.5438741039795666,
+      "epoch": 0.5619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 9.066479890979623e-06,
-      "loss": 0.8627,
+      "learning_rate": 8.485692919168686e-06,
+      "loss": 0.9146,
       "step": 19803
     },
     {
-      "epoch": 0.5439015682074099,
+      "epoch": 0.5619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 9.065594263029697e-06,
-      "loss": 0.9201,
+      "learning_rate": 8.484784446420048e-06,
+      "loss": 0.8835,
       "step": 19804
     },
     {
-      "epoch": 0.543929032435253,
+      "epoch": 0.5620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 9.06470864247315e-06,
-      "loss": 0.8976,
+      "learning_rate": 8.483875986470502e-06,
+      "loss": 0.8526,
       "step": 19805
     },
     {
-      "epoch": 0.5439564966630963,
+      "epoch": 0.5620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 9.063823029316994e-06,
-      "loss": 0.9884,
+      "learning_rate": 8.482967539327722e-06,
+      "loss": 0.8557,
       "step": 19806
     },
     {
-      "epoch": 0.5439839608909396,
+      "epoch": 0.5620601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 9.062937423568228e-06,
-      "loss": 0.8298,
+      "learning_rate": 8.482059104999387e-06,
+      "loss": 0.8978,
       "step": 19807
     },
     {
-      "epoch": 0.5440114251187828,
+      "epoch": 0.5620885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 9.062051825233867e-06,
-      "loss": 0.8946,
+      "learning_rate": 8.481150683493166e-06,
+      "loss": 0.8737,
       "step": 19808
     },
     {
-      "epoch": 0.544038889346626,
+      "epoch": 0.562116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 9.061166234320914e-06,
-      "loss": 0.8873,
+      "learning_rate": 8.48024227481673e-06,
+      "loss": 0.783,
       "step": 19809
     },
     {
-      "epoch": 0.5440663535744693,
+      "epoch": 0.5621452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 9.060280650836375e-06,
-      "loss": 0.9672,
+      "learning_rate": 8.47933387897776e-06,
+      "loss": 0.8373,
       "step": 19810
     },
     {
-      "epoch": 0.5440938178023125,
+      "epoch": 0.5621736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 9.059395074787263e-06,
-      "loss": 0.8996,
+      "learning_rate": 8.478425495983923e-06,
+      "loss": 0.8556,
       "step": 19811
     },
     {
-      "epoch": 0.5441212820301558,
+      "epoch": 0.5622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 9.058509506180583e-06,
-      "loss": 0.8505,
+      "learning_rate": 8.477517125842893e-06,
+      "loss": 0.7558,
       "step": 19812
     },
     {
-      "epoch": 0.5441487462579989,
+      "epoch": 0.5622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 9.057623945023337e-06,
-      "loss": 0.8251,
+      "learning_rate": 8.476608768562348e-06,
+      "loss": 0.8472,
       "step": 19813
     },
     {
-      "epoch": 0.5441762104858422,
+      "epoch": 0.5622587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 9.056738391322535e-06,
-      "loss": 0.8588,
+      "learning_rate": 8.47570042414995e-06,
+      "loss": 0.7172,
       "step": 19814
     },
     {
-      "epoch": 0.5442036747136855,
+      "epoch": 0.5622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 9.055852845085186e-06,
-      "loss": 0.9191,
+      "learning_rate": 8.474792092613387e-06,
+      "loss": 0.8539,
       "step": 19815
     },
     {
-      "epoch": 0.5442311389415286,
+      "epoch": 0.5623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 9.054967306318293e-06,
-      "loss": 0.9405,
+      "learning_rate": 8.47388377396032e-06,
+      "loss": 0.8128,
       "step": 19816
     },
     {
-      "epoch": 0.5442586031693719,
+      "epoch": 0.562343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 9.054081775028867e-06,
-      "loss": 0.7486,
+      "learning_rate": 8.472975468198425e-06,
+      "loss": 0.8016,
       "step": 19817
     },
     {
-      "epoch": 0.5442860673972151,
+      "epoch": 0.562372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 9.053196251223915e-06,
-      "loss": 0.9524,
+      "learning_rate": 8.472067175335377e-06,
+      "loss": 0.864,
       "step": 19818
     },
     {
-      "epoch": 0.5443135316250584,
+      "epoch": 0.5624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 9.052310734910436e-06,
-      "loss": 0.8896,
+      "learning_rate": 8.471158895378846e-06,
+      "loss": 0.8561,
       "step": 19819
     },
     {
-      "epoch": 0.5443409958529016,
+      "epoch": 0.5624290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 9.051425226095445e-06,
-      "loss": 0.9504,
+      "learning_rate": 8.470250628336503e-06,
+      "loss": 0.7776,
       "step": 19820
     },
     {
-      "epoch": 0.5443684600807448,
+      "epoch": 0.5624574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 9.050539724785941e-06,
-      "loss": 0.8595,
+      "learning_rate": 8.469342374216022e-06,
+      "loss": 0.7507,
       "step": 19821
     },
     {
-      "epoch": 0.5443959243085881,
+      "epoch": 0.5624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 9.049654230988936e-06,
-      "loss": 0.8749,
+      "learning_rate": 8.468434133025077e-06,
+      "loss": 0.7964,
       "step": 19822
     },
     {
-      "epoch": 0.5444233885364314,
+      "epoch": 0.5625141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 9.048768744711436e-06,
-      "loss": 0.8973,
+      "learning_rate": 8.467525904771336e-06,
+      "loss": 0.7957,
       "step": 19823
     },
     {
-      "epoch": 0.5444508527642745,
+      "epoch": 0.5625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 9.04788326596045e-06,
-      "loss": 0.9453,
+      "learning_rate": 8.466617689462474e-06,
+      "loss": 0.8693,
       "step": 19824
     },
     {
-      "epoch": 0.5444783169921178,
+      "epoch": 0.5625709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 9.046997794742974e-06,
-      "loss": 0.884,
+      "learning_rate": 8.46570948710616e-06,
+      "loss": 0.9539,
       "step": 19825
     },
     {
-      "epoch": 0.544505781219961,
+      "epoch": 0.5625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 9.046112331066023e-06,
-      "loss": 0.8636,
+      "learning_rate": 8.464801297710065e-06,
+      "loss": 0.7131,
       "step": 19826
     },
     {
-      "epoch": 0.5445332454478042,
+      "epoch": 0.562627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 9.045226874936602e-06,
-      "loss": 0.9265,
+      "learning_rate": 8.463893121281865e-06,
+      "loss": 0.864,
       "step": 19827
     },
     {
-      "epoch": 0.5445607096756475,
+      "epoch": 0.562656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 9.044341426361713e-06,
-      "loss": 0.8225,
+      "learning_rate": 8.46298495782923e-06,
+      "loss": 0.926,
       "step": 19828
     },
     {
-      "epoch": 0.5445881739034907,
+      "epoch": 0.5626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 9.043455985348363e-06,
-      "loss": 0.8416,
+      "learning_rate": 8.462076807359824e-06,
+      "loss": 0.8138,
       "step": 19829
     },
     {
-      "epoch": 0.544615638131334,
+      "epoch": 0.5627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 9.042570551903568e-06,
-      "loss": 0.9722,
+      "learning_rate": 8.461168669881328e-06,
+      "loss": 0.7892,
       "step": 19830
     },
     {
-      "epoch": 0.5446431023591771,
+      "epoch": 0.5627412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 9.041685126034319e-06,
-      "loss": 0.9325,
+      "learning_rate": 8.460260545401411e-06,
+      "loss": 0.8935,
       "step": 19831
     },
     {
-      "epoch": 0.5446705665870204,
+      "epoch": 0.5627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 9.040799707747631e-06,
-      "loss": 0.814,
+      "learning_rate": 8.459352433927743e-06,
+      "loss": 0.8269,
       "step": 19832
     },
     {
-      "epoch": 0.5446980308148637,
+      "epoch": 0.5627979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 9.039914297050505e-06,
-      "loss": 0.8767,
+      "learning_rate": 8.45844433546799e-06,
+      "loss": 0.8772,
       "step": 19833
     },
     {
-      "epoch": 0.5447254950427068,
+      "epoch": 0.5628263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 9.039028893949949e-06,
-      "loss": 0.7829,
+      "learning_rate": 8.457536250029832e-06,
+      "loss": 0.9023,
       "step": 19834
     },
     {
-      "epoch": 0.5447529592705501,
+      "epoch": 0.5628547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 9.038143498452972e-06,
-      "loss": 0.8411,
+      "learning_rate": 8.456628177620933e-06,
+      "loss": 0.8639,
       "step": 19835
     },
     {
-      "epoch": 0.5447804234983934,
+      "epoch": 0.562883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 9.037258110566578e-06,
-      "loss": 0.9136,
+      "learning_rate": 8.455720118248963e-06,
+      "loss": 0.9303,
       "step": 19836
     },
     {
-      "epoch": 0.5448078877262366,
+      "epoch": 0.5629114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 9.036372730297764e-06,
-      "loss": 0.9518,
+      "learning_rate": 8.454812071921597e-06,
+      "loss": 0.9576,
       "step": 19837
     },
     {
-      "epoch": 0.5448353519540798,
+      "epoch": 0.5629398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 9.035487357653544e-06,
-      "loss": 0.9679,
+      "learning_rate": 8.453904038646501e-06,
+      "loss": 0.8823,
       "step": 19838
     },
     {
-      "epoch": 0.544862816181923,
+      "epoch": 0.5629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 9.034601992640924e-06,
-      "loss": 0.8807,
+      "learning_rate": 8.452996018431348e-06,
+      "loss": 0.8387,
       "step": 19839
     },
     {
-      "epoch": 0.5448902804097663,
+      "epoch": 0.5629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 9.033716635266906e-06,
-      "loss": 0.937,
+      "learning_rate": 8.452088011283807e-06,
+      "loss": 0.8776,
       "step": 19840
     },
     {
-      "epoch": 0.5449177446376096,
+      "epoch": 0.5630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 9.032831285538499e-06,
-      "loss": 0.8192,
+      "learning_rate": 8.45118001721155e-06,
+      "loss": 0.7779,
       "step": 19841
     },
     {
-      "epoch": 0.5449452088654527,
+      "epoch": 0.5630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 9.0319459434627e-06,
-      "loss": 0.9021,
+      "learning_rate": 8.450272036222245e-06,
+      "loss": 0.8405,
       "step": 19842
     },
     {
-      "epoch": 0.544972673093296,
+      "epoch": 0.5630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 9.031060609046523e-06,
-      "loss": 0.831,
+      "learning_rate": 8.449364068323559e-06,
+      "loss": 0.8416,
       "step": 19843
     },
     {
-      "epoch": 0.5450001373211392,
+      "epoch": 0.5631101021566401,
       "grad_norm": 0.0,
-      "learning_rate": 9.030175282296971e-06,
-      "loss": 0.8411,
+      "learning_rate": 8.448456113523166e-06,
+      "loss": 0.7819,
       "step": 19844
     },
     {
-      "epoch": 0.5450276015489824,
+      "epoch": 0.5631384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 9.029289963221044e-06,
-      "loss": 0.8466,
+      "learning_rate": 8.44754817182873e-06,
+      "loss": 0.8366,
       "step": 19845
     },
     {
-      "epoch": 0.5450550657768257,
+      "epoch": 0.56316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.028404651825752e-06,
-      "loss": 0.822,
+      "learning_rate": 8.44664024324793e-06,
+      "loss": 0.8827,
       "step": 19846
     },
     {
-      "epoch": 0.5450825300046689,
+      "epoch": 0.5631952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 9.027519348118102e-06,
-      "loss": 0.9016,
+      "learning_rate": 8.445732327788425e-06,
+      "loss": 0.8926,
       "step": 19847
     },
     {
-      "epoch": 0.5451099942325122,
+      "epoch": 0.5632236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 9.026634052105095e-06,
-      "loss": 0.8268,
+      "learning_rate": 8.444824425457893e-06,
+      "loss": 0.8144,
       "step": 19848
     },
     {
-      "epoch": 0.5451374584603554,
+      "epoch": 0.5632519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 9.025748763793734e-06,
-      "loss": 0.8811,
+      "learning_rate": 8.443916536263999e-06,
+      "loss": 0.8171,
       "step": 19849
     },
     {
-      "epoch": 0.5451649226881986,
+      "epoch": 0.5632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 9.024863483191025e-06,
-      "loss": 0.931,
+      "learning_rate": 8.443008660214409e-06,
+      "loss": 0.8343,
       "step": 19850
     },
     {
-      "epoch": 0.5451923869160419,
+      "epoch": 0.5633087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 9.023978210303976e-06,
-      "loss": 0.819,
+      "learning_rate": 8.442100797316797e-06,
+      "loss": 0.9446,
       "step": 19851
     },
     {
-      "epoch": 0.545219851143885,
+      "epoch": 0.5633371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 9.02309294513959e-06,
-      "loss": 0.9219,
+      "learning_rate": 8.441192947578829e-06,
+      "loss": 0.8028,
       "step": 19852
     },
     {
-      "epoch": 0.5452473153717283,
+      "epoch": 0.5633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 9.022207687704871e-06,
-      "loss": 0.9774,
+      "learning_rate": 8.440285111008172e-06,
+      "loss": 0.8004,
       "step": 19853
     },
     {
-      "epoch": 0.5452747795995716,
+      "epoch": 0.5633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 9.021322438006822e-06,
-      "loss": 0.8801,
+      "learning_rate": 8.4393772876125e-06,
+      "loss": 0.9549,
       "step": 19854
     },
     {
-      "epoch": 0.5453022438274148,
+      "epoch": 0.563422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.020437196052449e-06,
-      "loss": 0.9088,
+      "learning_rate": 8.43846947739948e-06,
+      "loss": 0.9068,
       "step": 19855
     },
     {
-      "epoch": 0.545329708055258,
+      "epoch": 0.5634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 9.019551961848758e-06,
-      "loss": 0.8875,
+      "learning_rate": 8.437561680376774e-06,
+      "loss": 1.0605,
       "step": 19856
     },
     {
-      "epoch": 0.5453571722831012,
+      "epoch": 0.5634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 9.018666735402749e-06,
-      "loss": 0.8942,
+      "learning_rate": 8.436653896552057e-06,
+      "loss": 0.8735,
       "step": 19857
     },
     {
-      "epoch": 0.5453846365109445,
+      "epoch": 0.5635073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 9.01778151672143e-06,
-      "loss": 0.8137,
+      "learning_rate": 8.435746125932995e-06,
+      "loss": 0.9928,
       "step": 19858
     },
     {
-      "epoch": 0.5454121007387878,
+      "epoch": 0.5635357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 9.016896305811807e-06,
-      "loss": 0.8562,
+      "learning_rate": 8.434838368527254e-06,
+      "loss": 0.8646,
       "step": 19859
     },
     {
-      "epoch": 0.5454395649666309,
+      "epoch": 0.5635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 9.016011102680881e-06,
-      "loss": 0.8522,
+      "learning_rate": 8.433930624342503e-06,
+      "loss": 0.9256,
       "step": 19860
     },
     {
-      "epoch": 0.5454670291944742,
+      "epoch": 0.5635925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 9.015125907335653e-06,
-      "loss": 0.8229,
+      "learning_rate": 8.43302289338641e-06,
+      "loss": 0.8656,
       "step": 19861
     },
     {
-      "epoch": 0.5454944934223175,
+      "epoch": 0.5636208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 9.01424071978313e-06,
-      "loss": 0.9447,
+      "learning_rate": 8.432115175666647e-06,
+      "loss": 0.8992,
       "step": 19862
     },
     {
-      "epoch": 0.5455219576501606,
+      "epoch": 0.5636492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 9.01335554003032e-06,
-      "loss": 0.8283,
+      "learning_rate": 8.431207471190877e-06,
+      "loss": 0.893,
       "step": 19863
     },
     {
-      "epoch": 0.5455494218780039,
+      "epoch": 0.563677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 9.01247036808422e-06,
-      "loss": 0.9307,
+      "learning_rate": 8.430299779966764e-06,
+      "loss": 0.844,
       "step": 19864
     },
     {
-      "epoch": 0.5455768861058471,
+      "epoch": 0.5637060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 9.01158520395184e-06,
-      "loss": 0.8044,
+      "learning_rate": 8.429392102001983e-06,
+      "loss": 0.9126,
       "step": 19865
     },
     {
-      "epoch": 0.5456043503336904,
+      "epoch": 0.5637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 9.010700047640179e-06,
-      "loss": 0.8436,
+      "learning_rate": 8.428484437304198e-06,
+      "loss": 0.8303,
       "step": 19866
     },
     {
-      "epoch": 0.5456318145615336,
+      "epoch": 0.5637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 9.00981489915624e-06,
-      "loss": 0.8107,
+      "learning_rate": 8.427576785881073e-06,
+      "loss": 0.8573,
       "step": 19867
     },
     {
-      "epoch": 0.5456592787893768,
+      "epoch": 0.5637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 9.008929758507032e-06,
-      "loss": 0.9105,
+      "learning_rate": 8.42666914774028e-06,
+      "loss": 0.8613,
       "step": 19868
     },
     {
-      "epoch": 0.5456867430172201,
+      "epoch": 0.5638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 9.008044625699555e-06,
-      "loss": 0.8806,
+      "learning_rate": 8.425761522889483e-06,
+      "loss": 0.9066,
       "step": 19869
     },
     {
-      "epoch": 0.5457142072450633,
+      "epoch": 0.5638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 9.007159500740812e-06,
-      "loss": 0.8372,
+      "learning_rate": 8.424853911336347e-06,
+      "loss": 0.9131,
       "step": 19870
     },
     {
-      "epoch": 0.5457416714729065,
+      "epoch": 0.5638762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 9.006274383637811e-06,
-      "loss": 0.9035,
+      "learning_rate": 8.423946313088543e-06,
+      "loss": 0.87,
       "step": 19871
     },
     {
-      "epoch": 0.5457691357007498,
+      "epoch": 0.5639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 9.005389274397552e-06,
-      "loss": 0.8224,
+      "learning_rate": 8.423038728153734e-06,
+      "loss": 0.823,
       "step": 19872
     },
     {
-      "epoch": 0.545796599928593,
+      "epoch": 0.5639330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 9.004504173027034e-06,
-      "loss": 0.8777,
+      "learning_rate": 8.422131156539588e-06,
+      "loss": 0.9033,
       "step": 19873
     },
     {
-      "epoch": 0.5458240641564363,
+      "epoch": 0.5639614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 9.003619079533266e-06,
-      "loss": 0.8161,
+      "learning_rate": 8.421223598253773e-06,
+      "loss": 0.8775,
       "step": 19874
     },
     {
-      "epoch": 0.5458515283842795,
+      "epoch": 0.5639897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 9.00273399392325e-06,
-      "loss": 0.8716,
+      "learning_rate": 8.420316053303948e-06,
+      "loss": 0.8117,
       "step": 19875
     },
     {
-      "epoch": 0.5458789926121227,
+      "epoch": 0.5640181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 9.00184891620399e-06,
-      "loss": 0.876,
+      "learning_rate": 8.419408521697788e-06,
+      "loss": 0.9188,
       "step": 19876
     },
     {
-      "epoch": 0.545906456839966,
+      "epoch": 0.5640465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 9.00096384638249e-06,
-      "loss": 0.8746,
+      "learning_rate": 8.418501003442954e-06,
+      "loss": 0.8893,
       "step": 19877
     },
     {
-      "epoch": 0.5459339210678091,
+      "epoch": 0.5640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 9.000078784465746e-06,
-      "loss": 0.9084,
+      "learning_rate": 8.417593498547116e-06,
+      "loss": 0.8691,
       "step": 19878
     },
     {
-      "epoch": 0.5459613852956524,
+      "epoch": 0.5641032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 8.999193730460769e-06,
-      "loss": 0.9974,
+      "learning_rate": 8.416686007017936e-06,
+      "loss": 0.8377,
       "step": 19879
     },
     {
-      "epoch": 0.5459888495234957,
+      "epoch": 0.5641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.99830868437456e-06,
-      "loss": 0.8566,
+      "learning_rate": 8.415778528863077e-06,
+      "loss": 0.8453,
       "step": 19880
     },
     {
-      "epoch": 0.5460163137513389,
+      "epoch": 0.5641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 8.997423646214118e-06,
-      "loss": 0.9258,
+      "learning_rate": 8.414871064090211e-06,
+      "loss": 0.909,
       "step": 19881
     },
     {
-      "epoch": 0.5460437779791821,
+      "epoch": 0.564188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 8.996538615986447e-06,
-      "loss": 0.925,
+      "learning_rate": 8.413963612707001e-06,
+      "loss": 0.8221,
       "step": 19882
     },
     {
-      "epoch": 0.5460712422070253,
+      "epoch": 0.564216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 8.995653593698556e-06,
-      "loss": 0.9716,
+      "learning_rate": 8.413056174721112e-06,
+      "loss": 0.8843,
       "step": 19883
     },
     {
-      "epoch": 0.5460987064348686,
+      "epoch": 0.5642451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 8.994768579357442e-06,
-      "loss": 0.7657,
+      "learning_rate": 8.412148750140206e-06,
+      "loss": 0.8856,
       "step": 19884
     },
     {
-      "epoch": 0.5461261706627119,
+      "epoch": 0.5642735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 8.993883572970107e-06,
-      "loss": 0.7738,
+      "learning_rate": 8.411241338971954e-06,
+      "loss": 0.7622,
       "step": 19885
     },
     {
-      "epoch": 0.546153634890555,
+      "epoch": 0.5643019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 8.992998574543554e-06,
-      "loss": 0.9315,
+      "learning_rate": 8.410333941224017e-06,
+      "loss": 0.8659,
       "step": 19886
     },
     {
-      "epoch": 0.5461810991183983,
+      "epoch": 0.5643303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 8.992113584084788e-06,
-      "loss": 0.8265,
+      "learning_rate": 8.409426556904058e-06,
+      "loss": 0.8849,
       "step": 19887
     },
     {
-      "epoch": 0.5462085633462416,
+      "epoch": 0.5643586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 8.991228601600808e-06,
-      "loss": 0.8459,
+      "learning_rate": 8.408519186019748e-06,
+      "loss": 0.9202,
       "step": 19888
     },
     {
-      "epoch": 0.5462360275740847,
+      "epoch": 0.5643870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 8.990343627098622e-06,
-      "loss": 0.7964,
+      "learning_rate": 8.407611828578746e-06,
+      "loss": 0.9243,
       "step": 19889
     },
     {
-      "epoch": 0.546263491801928,
+      "epoch": 0.5644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 8.989458660585225e-06,
-      "loss": 0.8762,
+      "learning_rate": 8.406704484588718e-06,
+      "loss": 0.8162,
       "step": 19890
     },
     {
-      "epoch": 0.5462909560297712,
+      "epoch": 0.5644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 8.98857370206762e-06,
-      "loss": 0.9253,
+      "learning_rate": 8.405797154057326e-06,
+      "loss": 0.8379,
       "step": 19891
     },
     {
-      "epoch": 0.5463184202576145,
+      "epoch": 0.564472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 8.987688751552816e-06,
-      "loss": 0.8057,
+      "learning_rate": 8.404889836992242e-06,
+      "loss": 0.8654,
       "step": 19892
     },
     {
-      "epoch": 0.5463458844854577,
+      "epoch": 0.5645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 8.986803809047809e-06,
-      "loss": 0.8513,
+      "learning_rate": 8.403982533401123e-06,
+      "loss": 0.8545,
       "step": 19893
     },
     {
-      "epoch": 0.5463733487133009,
+      "epoch": 0.5645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 8.985918874559601e-06,
-      "loss": 0.8971,
+      "learning_rate": 8.403075243291636e-06,
+      "loss": 0.8046,
       "step": 19894
     },
     {
-      "epoch": 0.5464008129411442,
+      "epoch": 0.5645573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 8.985033948095201e-06,
-      "loss": 0.9024,
+      "learning_rate": 8.402167966671445e-06,
+      "loss": 0.8818,
       "step": 19895
     },
     {
-      "epoch": 0.5464282771689873,
+      "epoch": 0.5645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 8.984149029661603e-06,
-      "loss": 0.826,
+      "learning_rate": 8.401260703548213e-06,
+      "loss": 0.8468,
       "step": 19896
     },
     {
-      "epoch": 0.5464557413968306,
+      "epoch": 0.5646140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 8.983264119265809e-06,
-      "loss": 0.8715,
+      "learning_rate": 8.400353453929601e-06,
+      "loss": 0.884,
       "step": 19897
     },
     {
-      "epoch": 0.5464832056246739,
+      "epoch": 0.5646424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 8.982379216914824e-06,
-      "loss": 0.7555,
+      "learning_rate": 8.39944621782328e-06,
+      "loss": 0.8641,
       "step": 19898
     },
     {
-      "epoch": 0.5465106698525171,
+      "epoch": 0.5646708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 8.98149432261565e-06,
-      "loss": 0.8836,
+      "learning_rate": 8.398538995236906e-06,
+      "loss": 0.8894,
       "step": 19899
     },
     {
-      "epoch": 0.5465381340803603,
+      "epoch": 0.5646992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 8.980609436375289e-06,
-      "loss": 0.8027,
+      "learning_rate": 8.397631786178143e-06,
+      "loss": 0.9414,
       "step": 19900
     },
     {
-      "epoch": 0.5465655983082036,
+      "epoch": 0.564727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 8.97972455820074e-06,
-      "loss": 0.7718,
+      "learning_rate": 8.39672459065466e-06,
+      "loss": 0.8551,
       "step": 19901
     },
     {
-      "epoch": 0.5465930625360468,
+      "epoch": 0.5647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 8.978839688099003e-06,
-      "loss": 0.8707,
+      "learning_rate": 8.395817408674117e-06,
+      "loss": 0.877,
       "step": 19902
     },
     {
-      "epoch": 0.5466205267638901,
+      "epoch": 0.5647843359818389,
       "grad_norm": 0.0,
-      "learning_rate": 8.977954826077081e-06,
-      "loss": 0.8913,
+      "learning_rate": 8.394910240244172e-06,
+      "loss": 0.7134,
       "step": 19903
     },
     {
-      "epoch": 0.5466479909917332,
+      "epoch": 0.5648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 8.977069972141981e-06,
-      "loss": 0.9684,
+      "learning_rate": 8.394003085372496e-06,
+      "loss": 0.9119,
       "step": 19904
     },
     {
-      "epoch": 0.5466754552195765,
+      "epoch": 0.5648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 8.976185126300696e-06,
-      "loss": 0.7626,
+      "learning_rate": 8.393095944066749e-06,
+      "loss": 0.9209,
       "step": 19905
     },
     {
-      "epoch": 0.5467029194474198,
+      "epoch": 0.5648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 8.975300288560236e-06,
-      "loss": 0.9748,
+      "learning_rate": 8.392188816334587e-06,
+      "loss": 0.93,
       "step": 19906
     },
     {
-      "epoch": 0.5467303836752629,
+      "epoch": 0.5648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 8.974415458927592e-06,
-      "loss": 0.8717,
+      "learning_rate": 8.391281702183683e-06,
+      "loss": 0.9215,
       "step": 19907
     },
     {
-      "epoch": 0.5467578479031062,
+      "epoch": 0.5649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 8.973530637409772e-06,
-      "loss": 0.9401,
+      "learning_rate": 8.39037460162169e-06,
+      "loss": 0.8487,
       "step": 19908
     },
     {
-      "epoch": 0.5467853121309494,
+      "epoch": 0.5649545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 8.972645824013774e-06,
-      "loss": 0.8887,
+      "learning_rate": 8.389467514656282e-06,
+      "loss": 0.7769,
       "step": 19909
     },
     {
-      "epoch": 0.5468127763587927,
+      "epoch": 0.5649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 8.9717610187466e-06,
-      "loss": 0.9022,
+      "learning_rate": 8.388560441295111e-06,
+      "loss": 0.9885,
       "step": 19910
     },
     {
-      "epoch": 0.5468402405866359,
+      "epoch": 0.565011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 8.970876221615253e-06,
-      "loss": 0.896,
+      "learning_rate": 8.387653381545841e-06,
+      "loss": 0.9109,
       "step": 19911
     },
     {
-      "epoch": 0.5468677048144791,
+      "epoch": 0.5650397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 8.96999143262673e-06,
-      "loss": 0.8409,
+      "learning_rate": 8.38674633541614e-06,
+      "loss": 0.7969,
       "step": 19912
     },
     {
-      "epoch": 0.5468951690423224,
+      "epoch": 0.5650681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 8.969106651788038e-06,
-      "loss": 0.8307,
+      "learning_rate": 8.385839302913661e-06,
+      "loss": 0.9092,
       "step": 19913
     },
     {
-      "epoch": 0.5469226332701657,
+      "epoch": 0.5650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 8.968221879106167e-06,
-      "loss": 0.886,
+      "learning_rate": 8.38493228404607e-06,
+      "loss": 0.8912,
       "step": 19914
     },
     {
-      "epoch": 0.5469500974980088,
+      "epoch": 0.5651248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 8.967337114588127e-06,
-      "loss": 0.8909,
+      "learning_rate": 8.384025278821032e-06,
+      "loss": 0.9272,
       "step": 19915
     },
     {
-      "epoch": 0.5469775617258521,
+      "epoch": 0.5651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 8.966452358240918e-06,
-      "loss": 0.8453,
+      "learning_rate": 8.383118287246203e-06,
+      "loss": 0.7954,
       "step": 19916
     },
     {
-      "epoch": 0.5470050259536953,
+      "epoch": 0.5651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 8.965567610071536e-06,
-      "loss": 0.8931,
+      "learning_rate": 8.382211309329245e-06,
+      "loss": 0.8774,
       "step": 19917
     },
     {
-      "epoch": 0.5470324901815385,
+      "epoch": 0.5652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 8.964682870086987e-06,
-      "loss": 0.7796,
+      "learning_rate": 8.381304345077823e-06,
+      "loss": 0.894,
       "step": 19918
     },
     {
-      "epoch": 0.5470599544093818,
+      "epoch": 0.5652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 8.963798138294265e-06,
-      "loss": 0.8137,
+      "learning_rate": 8.380397394499597e-06,
+      "loss": 0.8989,
       "step": 19919
     },
     {
-      "epoch": 0.547087418637225,
+      "epoch": 0.565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 8.962913414700374e-06,
-      "loss": 0.8367,
+      "learning_rate": 8.379490457602224e-06,
+      "loss": 0.9002,
       "step": 19920
     },
     {
-      "epoch": 0.5471148828650683,
+      "epoch": 0.5652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 8.962028699312314e-06,
-      "loss": 0.8896,
+      "learning_rate": 8.37858353439337e-06,
+      "loss": 0.9642,
       "step": 19921
     },
     {
-      "epoch": 0.5471423470929114,
+      "epoch": 0.5653234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 8.961143992137083e-06,
-      "loss": 0.8732,
+      "learning_rate": 8.377676624880688e-06,
+      "loss": 0.9666,
       "step": 19922
     },
     {
-      "epoch": 0.5471698113207547,
+      "epoch": 0.5653518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 8.960259293181685e-06,
-      "loss": 0.897,
+      "learning_rate": 8.376769729071852e-06,
+      "loss": 0.9099,
       "step": 19923
     },
     {
-      "epoch": 0.547197275548598,
+      "epoch": 0.5653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 8.959374602453122e-06,
-      "loss": 0.8597,
+      "learning_rate": 8.375862846974512e-06,
+      "loss": 0.9139,
       "step": 19924
     },
     {
-      "epoch": 0.5472247397764411,
+      "epoch": 0.5654086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 8.958489919958383e-06,
-      "loss": 0.8588,
+      "learning_rate": 8.374955978596331e-06,
+      "loss": 0.8281,
       "step": 19925
     },
     {
-      "epoch": 0.5472522040042844,
+      "epoch": 0.5654370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 8.957605245704479e-06,
-      "loss": 0.9567,
+      "learning_rate": 8.374049123944971e-06,
+      "loss": 0.8182,
       "step": 19926
     },
     {
-      "epoch": 0.5472796682321277,
+      "epoch": 0.5654653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 8.956720579698405e-06,
-      "loss": 0.891,
+      "learning_rate": 8.373142283028093e-06,
+      "loss": 0.9622,
       "step": 19927
     },
     {
-      "epoch": 0.5473071324599709,
+      "epoch": 0.5654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 8.955835921947164e-06,
-      "loss": 0.8441,
+      "learning_rate": 8.372235455853352e-06,
+      "loss": 0.8069,
       "step": 19928
     },
     {
-      "epoch": 0.5473345966878141,
+      "epoch": 0.565522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 8.95495127245775e-06,
-      "loss": 0.7851,
+      "learning_rate": 8.371328642428414e-06,
+      "loss": 0.8171,
       "step": 19929
     },
     {
-      "epoch": 0.5473620609156573,
+      "epoch": 0.565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 8.954066631237172e-06,
-      "loss": 0.8898,
+      "learning_rate": 8.370421842760934e-06,
+      "loss": 0.8738,
       "step": 19930
     },
     {
-      "epoch": 0.5473895251435006,
+      "epoch": 0.5655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 8.953181998292419e-06,
-      "loss": 0.9141,
+      "learning_rate": 8.369515056858575e-06,
+      "loss": 0.825,
       "step": 19931
     },
     {
-      "epoch": 0.5474169893713439,
+      "epoch": 0.5656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 8.952297373630498e-06,
-      "loss": 0.913,
+      "learning_rate": 8.368608284728997e-06,
+      "loss": 0.8676,
       "step": 19932
     },
     {
-      "epoch": 0.547444453599187,
+      "epoch": 0.5656356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 8.951412757258403e-06,
-      "loss": 0.8778,
+      "learning_rate": 8.367701526379857e-06,
+      "loss": 0.8926,
       "step": 19933
     },
     {
-      "epoch": 0.5474719178270303,
+      "epoch": 0.5656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 8.95052814918314e-06,
-      "loss": 0.9215,
+      "learning_rate": 8.366794781818813e-06,
+      "loss": 0.7923,
       "step": 19934
     },
     {
-      "epoch": 0.5474993820548735,
+      "epoch": 0.5656923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 8.949643549411703e-06,
-      "loss": 0.8893,
+      "learning_rate": 8.36588805105353e-06,
+      "loss": 0.8445,
       "step": 19935
     },
     {
-      "epoch": 0.5475268462827168,
+      "epoch": 0.5657207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 8.948758957951097e-06,
-      "loss": 0.8432,
+      "learning_rate": 8.364981334091665e-06,
+      "loss": 0.9124,
       "step": 19936
     },
     {
-      "epoch": 0.54755431051056,
+      "epoch": 0.5657491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 8.947874374808312e-06,
-      "loss": 0.9146,
+      "learning_rate": 8.364074630940869e-06,
+      "loss": 0.8511,
       "step": 19937
     },
     {
-      "epoch": 0.5475817747384032,
+      "epoch": 0.56577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 8.946989799990353e-06,
-      "loss": 0.8566,
+      "learning_rate": 8.363167941608814e-06,
+      "loss": 0.8529,
       "step": 19938
     },
     {
-      "epoch": 0.5476092389662465,
+      "epoch": 0.5658059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 8.946105233504218e-06,
-      "loss": 0.8742,
+      "learning_rate": 8.362261266103152e-06,
+      "loss": 0.9039,
       "step": 19939
     },
     {
-      "epoch": 0.5476367031940897,
+      "epoch": 0.5658342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 8.945220675356911e-06,
-      "loss": 0.9085,
+      "learning_rate": 8.361354604431544e-06,
+      "loss": 0.7566,
       "step": 19940
     },
     {
-      "epoch": 0.5476641674219329,
+      "epoch": 0.5658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 8.944336125555423e-06,
-      "loss": 0.8663,
+      "learning_rate": 8.360447956601645e-06,
+      "loss": 0.8686,
       "step": 19941
     },
     {
-      "epoch": 0.5476916316497762,
+      "epoch": 0.5658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 8.94345158410676e-06,
-      "loss": 0.8656,
+      "learning_rate": 8.359541322621118e-06,
+      "loss": 0.8919,
       "step": 19942
     },
     {
-      "epoch": 0.5477190958776194,
+      "epoch": 0.5659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 8.942567051017914e-06,
-      "loss": 0.9539,
+      "learning_rate": 8.358634702497618e-06,
+      "loss": 0.8405,
       "step": 19943
     },
     {
-      "epoch": 0.5477465601054626,
+      "epoch": 0.5659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 8.941682526295889e-06,
-      "loss": 0.9478,
+      "learning_rate": 8.357728096238804e-06,
+      "loss": 0.8572,
       "step": 19944
     },
     {
-      "epoch": 0.5477740243333059,
+      "epoch": 0.5659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 8.94079800994768e-06,
-      "loss": 0.8472,
+      "learning_rate": 8.356821503852338e-06,
+      "loss": 0.9198,
       "step": 19945
     },
     {
-      "epoch": 0.5478014885611491,
+      "epoch": 0.5660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 8.939913501980288e-06,
-      "loss": 0.8706,
+      "learning_rate": 8.355914925345872e-06,
+      "loss": 0.9045,
       "step": 19946
     },
     {
-      "epoch": 0.5478289527889924,
+      "epoch": 0.5660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 8.939029002400711e-06,
-      "loss": 0.858,
+      "learning_rate": 8.355008360727065e-06,
+      "loss": 0.8857,
       "step": 19947
     },
     {
-      "epoch": 0.5478564170168355,
+      "epoch": 0.566061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 8.938144511215951e-06,
-      "loss": 0.908,
+      "learning_rate": 8.354101810003578e-06,
+      "loss": 0.8644,
       "step": 19948
     },
     {
-      "epoch": 0.5478838812446788,
+      "epoch": 0.5660896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 8.937260028433e-06,
-      "loss": 0.9129,
+      "learning_rate": 8.353195273183068e-06,
+      "loss": 0.8066,
       "step": 19949
     },
     {
-      "epoch": 0.5479113454725221,
+      "epoch": 0.5661180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 8.936375554058857e-06,
-      "loss": 0.8351,
+      "learning_rate": 8.352288750273192e-06,
+      "loss": 0.9557,
       "step": 19950
     },
     {
-      "epoch": 0.5479388097003652,
+      "epoch": 0.5661464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 8.935491088100523e-06,
-      "loss": 0.8068,
+      "learning_rate": 8.351382241281604e-06,
+      "loss": 0.7772,
       "step": 19951
     },
     {
-      "epoch": 0.5479662739282085,
+      "epoch": 0.5661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 8.934606630565e-06,
-      "loss": 0.9725,
+      "learning_rate": 8.350475746215963e-06,
+      "loss": 0.8605,
       "step": 19952
     },
     {
-      "epoch": 0.5479937381560518,
+      "epoch": 0.5662031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 8.933722181459278e-06,
-      "loss": 0.7633,
+      "learning_rate": 8.34956926508393e-06,
+      "loss": 0.7649,
       "step": 19953
     },
     {
-      "epoch": 0.548021202383895,
+      "epoch": 0.5662315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 8.932837740790363e-06,
-      "loss": 0.9456,
+      "learning_rate": 8.34866279789316e-06,
+      "loss": 0.7937,
       "step": 19954
     },
     {
-      "epoch": 0.5480486666117382,
+      "epoch": 0.5662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 8.931953308565246e-06,
-      "loss": 0.8891,
+      "learning_rate": 8.34775634465131e-06,
+      "loss": 0.8639,
       "step": 19955
     },
     {
-      "epoch": 0.5480761308395814,
+      "epoch": 0.5662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 8.93106888479093e-06,
-      "loss": 0.8882,
+      "learning_rate": 8.346849905366036e-06,
+      "loss": 0.8669,
       "step": 19956
     },
     {
-      "epoch": 0.5481035950674247,
+      "epoch": 0.566316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.93018446947441e-06,
-      "loss": 0.8931,
+      "learning_rate": 8.345943480044996e-06,
+      "loss": 0.8065,
       "step": 19957
     },
     {
-      "epoch": 0.548131059295268,
+      "epoch": 0.5663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 8.929300062622683e-06,
-      "loss": 0.8479,
+      "learning_rate": 8.345037068695844e-06,
+      "loss": 0.8139,
       "step": 19958
     },
     {
-      "epoch": 0.5481585235231111,
+      "epoch": 0.5663734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 8.928415664242755e-06,
-      "loss": 0.8,
+      "learning_rate": 8.344130671326241e-06,
+      "loss": 0.8512,
       "step": 19959
     },
     {
-      "epoch": 0.5481859877509544,
+      "epoch": 0.5664018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 8.927531274341614e-06,
-      "loss": 0.8712,
+      "learning_rate": 8.343224287943838e-06,
+      "loss": 0.8489,
       "step": 19960
     },
     {
-      "epoch": 0.5482134519787976,
+      "epoch": 0.5664301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 8.926646892926259e-06,
-      "loss": 0.7642,
+      "learning_rate": 8.342317918556294e-06,
+      "loss": 0.7688,
       "step": 19961
     },
     {
-      "epoch": 0.5482409162066408,
+      "epoch": 0.5664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 8.925762520003689e-06,
-      "loss": 0.8025,
+      "learning_rate": 8.341411563171266e-06,
+      "loss": 0.844,
       "step": 19962
     },
     {
-      "epoch": 0.5482683804344841,
+      "epoch": 0.5664869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 8.924878155580904e-06,
-      "loss": 0.879,
+      "learning_rate": 8.34050522179641e-06,
+      "loss": 0.7928,
       "step": 19963
     },
     {
-      "epoch": 0.5482958446623273,
+      "epoch": 0.5665153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 8.923993799664898e-06,
-      "loss": 0.8831,
+      "learning_rate": 8.339598894439379e-06,
+      "loss": 0.8721,
       "step": 19964
     },
     {
-      "epoch": 0.5483233088901706,
+      "epoch": 0.5665437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 8.923109452262674e-06,
-      "loss": 0.9067,
+      "learning_rate": 8.338692581107833e-06,
+      "loss": 0.9645,
       "step": 19965
     },
     {
-      "epoch": 0.5483507731180138,
+      "epoch": 0.566572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.92222511338122e-06,
-      "loss": 0.8487,
+      "learning_rate": 8.337786281809425e-06,
+      "loss": 0.8783,
       "step": 19966
     },
     {
-      "epoch": 0.548378237345857,
+      "epoch": 0.5666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 8.921340783027541e-06,
-      "loss": 0.9113,
+      "learning_rate": 8.336879996551807e-06,
+      "loss": 0.8816,
       "step": 19967
     },
     {
-      "epoch": 0.5484057015737003,
+      "epoch": 0.5666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 8.920456461208632e-06,
-      "loss": 0.8979,
+      "learning_rate": 8.335973725342642e-06,
+      "loss": 0.7732,
       "step": 19968
     },
     {
-      "epoch": 0.5484331658015434,
+      "epoch": 0.5666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.919572147931487e-06,
-      "loss": 0.8344,
+      "learning_rate": 8.33506746818958e-06,
+      "loss": 0.8365,
       "step": 19969
     },
     {
-      "epoch": 0.5484606300293867,
+      "epoch": 0.5666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 8.918687843203107e-06,
-      "loss": 0.8028,
+      "learning_rate": 8.334161225100279e-06,
+      "loss": 0.8846,
       "step": 19970
     },
     {
-      "epoch": 0.54848809425723,
+      "epoch": 0.5667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 8.91780354703049e-06,
-      "loss": 0.8733,
+      "learning_rate": 8.333254996082394e-06,
+      "loss": 0.9567,
       "step": 19971
     },
     {
-      "epoch": 0.5485155584850732,
+      "epoch": 0.5667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 8.916919259420631e-06,
-      "loss": 0.8737,
+      "learning_rate": 8.332348781143579e-06,
+      "loss": 0.8184,
       "step": 19972
     },
     {
-      "epoch": 0.5485430227129164,
+      "epoch": 0.5667707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 8.916034980380523e-06,
-      "loss": 0.8166,
+      "learning_rate": 8.331442580291487e-06,
+      "loss": 0.9537,
       "step": 19973
     },
     {
-      "epoch": 0.5485704869407596,
+      "epoch": 0.5667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 8.915150709917165e-06,
-      "loss": 0.9072,
+      "learning_rate": 8.330536393533777e-06,
+      "loss": 0.8842,
       "step": 19974
     },
     {
-      "epoch": 0.5485979511686029,
+      "epoch": 0.5668274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 8.914266448037558e-06,
-      "loss": 0.8915,
+      "learning_rate": 8.329630220878098e-06,
+      "loss": 0.8582,
       "step": 19975
     },
     {
-      "epoch": 0.5486254153964462,
+      "epoch": 0.5668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 8.913382194748693e-06,
-      "loss": 0.8995,
+      "learning_rate": 8.32872406233211e-06,
+      "loss": 0.8296,
       "step": 19976
     },
     {
-      "epoch": 0.5486528796242893,
+      "epoch": 0.5668842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 8.912497950057573e-06,
-      "loss": 0.807,
+      "learning_rate": 8.327817917903464e-06,
+      "loss": 0.8771,
       "step": 19977
     },
     {
-      "epoch": 0.5486803438521326,
+      "epoch": 0.5669125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 8.911613713971188e-06,
-      "loss": 0.9807,
+      "learning_rate": 8.326911787599815e-06,
+      "loss": 0.9328,
       "step": 19978
     },
     {
-      "epoch": 0.5487078080799759,
+      "epoch": 0.5669409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 8.910729486496535e-06,
-      "loss": 0.7792,
+      "learning_rate": 8.32600567142882e-06,
+      "loss": 0.8569,
       "step": 19979
     },
     {
-      "epoch": 0.548735272307819,
+      "epoch": 0.5669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 8.909845267640616e-06,
-      "loss": 0.9008,
+      "learning_rate": 8.325099569398127e-06,
+      "loss": 0.8486,
       "step": 19980
     },
     {
-      "epoch": 0.5487627365356623,
+      "epoch": 0.5669977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 8.90896105741042e-06,
-      "loss": 0.857,
+      "learning_rate": 8.324193481515394e-06,
+      "loss": 0.8053,
       "step": 19981
     },
     {
-      "epoch": 0.5487902007635055,
+      "epoch": 0.5670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.908076855812947e-06,
-      "loss": 0.9142,
+      "learning_rate": 8.323287407788274e-06,
+      "loss": 0.838,
       "step": 19982
     },
     {
-      "epoch": 0.5488176649913488,
+      "epoch": 0.5670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 8.907192662855196e-06,
-      "loss": 0.8092,
+      "learning_rate": 8.32238134822442e-06,
+      "loss": 0.8872,
       "step": 19983
     },
     {
-      "epoch": 0.548845129219192,
+      "epoch": 0.567082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 8.90630847854416e-06,
-      "loss": 0.8675,
+      "learning_rate": 8.321475302831487e-06,
+      "loss": 0.879,
       "step": 19984
     },
     {
-      "epoch": 0.5488725934470352,
+      "epoch": 0.56711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 8.905424302886832e-06,
-      "loss": 0.8962,
+      "learning_rate": 8.320569271617127e-06,
+      "loss": 0.9255,
       "step": 19985
     },
     {
-      "epoch": 0.5489000576748785,
+      "epoch": 0.5671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 8.904540135890211e-06,
-      "loss": 0.9885,
+      "learning_rate": 8.319663254588995e-06,
+      "loss": 0.9389,
       "step": 19986
     },
     {
-      "epoch": 0.5489275219027218,
+      "epoch": 0.5671679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 8.903655977561295e-06,
-      "loss": 0.9993,
+      "learning_rate": 8.318757251754743e-06,
+      "loss": 0.8835,
       "step": 19987
     },
     {
-      "epoch": 0.5489549861305649,
+      "epoch": 0.5671963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 8.902771827907075e-06,
-      "loss": 0.8628,
+      "learning_rate": 8.317851263122023e-06,
+      "loss": 0.942,
       "step": 19988
     },
     {
-      "epoch": 0.5489824503584082,
+      "epoch": 0.5672247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 8.901887686934553e-06,
-      "loss": 0.8398,
+      "learning_rate": 8.31694528869849e-06,
+      "loss": 0.7413,
       "step": 19989
     },
     {
-      "epoch": 0.5490099145862514,
+      "epoch": 0.5672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 8.901003554650718e-06,
-      "loss": 0.8901,
+      "learning_rate": 8.316039328491796e-06,
+      "loss": 0.9742,
       "step": 19990
     },
     {
-      "epoch": 0.5490373788140946,
+      "epoch": 0.5672814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 8.900119431062568e-06,
-      "loss": 0.8846,
+      "learning_rate": 8.315133382509594e-06,
+      "loss": 0.8716,
       "step": 19991
     },
     {
-      "epoch": 0.5490648430419379,
+      "epoch": 0.5673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 8.899235316177101e-06,
-      "loss": 0.7892,
+      "learning_rate": 8.314227450759535e-06,
+      "loss": 0.9016,
       "step": 19992
     },
     {
-      "epoch": 0.5490923072697811,
+      "epoch": 0.5673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 8.89835121000131e-06,
-      "loss": 0.8469,
+      "learning_rate": 8.313321533249273e-06,
+      "loss": 0.8933,
       "step": 19993
     },
     {
-      "epoch": 0.5491197714976244,
+      "epoch": 0.567366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 8.897467112542188e-06,
-      "loss": 0.8748,
+      "learning_rate": 8.31241562998646e-06,
+      "loss": 0.8871,
       "step": 19994
     },
     {
-      "epoch": 0.5491472357254675,
+      "epoch": 0.5673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 8.896583023806741e-06,
-      "loss": 0.9049,
+      "learning_rate": 8.311509740978747e-06,
+      "loss": 1.0187,
       "step": 19995
     },
     {
-      "epoch": 0.5491746999533108,
+      "epoch": 0.5674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 8.895698943801953e-06,
-      "loss": 0.9996,
+      "learning_rate": 8.310603866233788e-06,
+      "loss": 0.9285,
       "step": 19996
     },
     {
-      "epoch": 0.5492021641811541,
+      "epoch": 0.5674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 8.894814872534819e-06,
-      "loss": 0.9283,
+      "learning_rate": 8.309698005759233e-06,
+      "loss": 0.7852,
       "step": 19997
     },
     {
-      "epoch": 0.5492296284089973,
+      "epoch": 0.5674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 8.893930810012339e-06,
-      "loss": 0.7719,
+      "learning_rate": 8.308792159562732e-06,
+      "loss": 0.7735,
       "step": 19998
     },
     {
-      "epoch": 0.5492570926368405,
+      "epoch": 0.5675085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 8.89304675624151e-06,
-      "loss": 0.8667,
+      "learning_rate": 8.307886327651943e-06,
+      "loss": 0.8926,
       "step": 19999
     },
     {
-      "epoch": 0.5492845568646838,
+      "epoch": 0.5675368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 8.89216271122932e-06,
-      "loss": 0.8289,
+      "learning_rate": 8.306980510034515e-06,
+      "loss": 0.8493,
       "step": 20000
     },
     {
-      "epoch": 0.549312021092527,
+      "epoch": 0.5675652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 8.891278674982774e-06,
-      "loss": 0.8702,
+      "learning_rate": 8.3060747067181e-06,
+      "loss": 0.8758,
       "step": 20001
     },
     {
-      "epoch": 0.5493394853203702,
+      "epoch": 0.5675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 8.890394647508854e-06,
-      "loss": 0.8886,
+      "learning_rate": 8.305168917710344e-06,
+      "loss": 0.8884,
       "step": 20002
     },
     {
-      "epoch": 0.5493669495482134,
+      "epoch": 0.567622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 8.889510628814564e-06,
-      "loss": 0.8391,
+      "learning_rate": 8.304263143018906e-06,
+      "loss": 0.8239,
       "step": 20003
     },
     {
-      "epoch": 0.5493944137760567,
+      "epoch": 0.5676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 8.888626618906896e-06,
-      "loss": 0.7872,
+      "learning_rate": 8.303357382651432e-06,
+      "loss": 0.9161,
       "step": 20004
     },
     {
-      "epoch": 0.5494218780039,
+      "epoch": 0.5676787741203179,
       "grad_norm": 0.0,
-      "learning_rate": 8.887742617792844e-06,
-      "loss": 0.9044,
+      "learning_rate": 8.302451636615574e-06,
+      "loss": 0.9279,
       "step": 20005
     },
     {
-      "epoch": 0.5494493422317431,
+      "epoch": 0.5677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 8.886858625479403e-06,
-      "loss": 0.7943,
+      "learning_rate": 8.301545904918986e-06,
+      "loss": 0.8919,
       "step": 20006
     },
     {
-      "epoch": 0.5494768064595864,
+      "epoch": 0.5677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 8.88597464197357e-06,
-      "loss": 0.8823,
+      "learning_rate": 8.300640187569313e-06,
+      "loss": 0.8434,
       "step": 20007
     },
     {
-      "epoch": 0.5495042706874296,
+      "epoch": 0.5677639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 8.885090667282337e-06,
-      "loss": 0.8887,
+      "learning_rate": 8.29973448457421e-06,
+      "loss": 0.875,
       "step": 20008
     },
     {
-      "epoch": 0.5495317349152729,
+      "epoch": 0.5677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 8.884206701412697e-06,
-      "loss": 0.8893,
+      "learning_rate": 8.298828795941327e-06,
+      "loss": 0.8843,
       "step": 20009
     },
     {
-      "epoch": 0.5495591991431161,
+      "epoch": 0.5678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 8.883322744371645e-06,
-      "loss": 0.9889,
+      "learning_rate": 8.297923121678313e-06,
+      "loss": 0.9752,
       "step": 20010
     },
     {
-      "epoch": 0.5495866633709593,
+      "epoch": 0.5678490351872871,
       "grad_norm": 0.0,
-      "learning_rate": 8.882438796166178e-06,
-      "loss": 0.8474,
+      "learning_rate": 8.297017461792817e-06,
+      "loss": 0.9785,
       "step": 20011
     },
     {
-      "epoch": 0.5496141275988026,
+      "epoch": 0.5678774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 8.881554856803285e-06,
-      "loss": 0.9433,
+      "learning_rate": 8.296111816292494e-06,
+      "loss": 0.9004,
       "step": 20012
     },
     {
-      "epoch": 0.5496415918266458,
+      "epoch": 0.567905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 8.88067092628997e-06,
-      "loss": 0.8489,
+      "learning_rate": 8.29520618518499e-06,
+      "loss": 0.9974,
       "step": 20013
     },
     {
-      "epoch": 0.549669056054489,
+      "epoch": 0.5679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 8.879787004633213e-06,
-      "loss": 0.9068,
+      "learning_rate": 8.294300568477951e-06,
+      "loss": 0.9089,
       "step": 20014
     },
     {
-      "epoch": 0.5496965202823323,
+      "epoch": 0.5679625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 8.878903091840017e-06,
-      "loss": 0.9222,
+      "learning_rate": 8.293394966179039e-06,
+      "loss": 0.8406,
       "step": 20015
     },
     {
-      "epoch": 0.5497239845101755,
+      "epoch": 0.5679909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 8.878019187917375e-06,
-      "loss": 0.834,
+      "learning_rate": 8.292489378295892e-06,
+      "loss": 0.9082,
       "step": 20016
     },
     {
-      "epoch": 0.5497514487380187,
+      "epoch": 0.5680192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 8.87713529287228e-06,
-      "loss": 0.8726,
+      "learning_rate": 8.291583804836167e-06,
+      "loss": 0.8114,
       "step": 20017
     },
     {
-      "epoch": 0.549778912965862,
+      "epoch": 0.5680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 8.876251406711727e-06,
-      "loss": 0.9109,
+      "learning_rate": 8.29067824580751e-06,
+      "loss": 0.8204,
       "step": 20018
     },
     {
-      "epoch": 0.5498063771937052,
+      "epoch": 0.5680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 8.875367529442705e-06,
-      "loss": 0.9128,
+      "learning_rate": 8.289772701217567e-06,
+      "loss": 0.8165,
       "step": 20019
     },
     {
-      "epoch": 0.5498338414215485,
+      "epoch": 0.5681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 8.874483661072213e-06,
-      "loss": 0.8111,
+      "learning_rate": 8.288867171073995e-06,
+      "loss": 0.8362,
       "step": 20020
     },
     {
-      "epoch": 0.5498613056493916,
+      "epoch": 0.5681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 8.87359980160724e-06,
-      "loss": 0.7961,
+      "learning_rate": 8.287961655384439e-06,
+      "loss": 0.998,
       "step": 20021
     },
     {
-      "epoch": 0.5498887698772349,
+      "epoch": 0.568161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 8.872715951054782e-06,
-      "loss": 0.8716,
+      "learning_rate": 8.287056154156546e-06,
+      "loss": 0.8272,
       "step": 20022
     },
     {
-      "epoch": 0.5499162341050782,
+      "epoch": 0.5681895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 8.871832109421833e-06,
-      "loss": 0.8589,
+      "learning_rate": 8.286150667397968e-06,
+      "loss": 0.9354,
       "step": 20023
     },
     {
-      "epoch": 0.5499436983329213,
+      "epoch": 0.5682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 8.870948276715387e-06,
-      "loss": 0.9361,
+      "learning_rate": 8.285245195116351e-06,
+      "loss": 0.8433,
       "step": 20024
     },
     {
-      "epoch": 0.5499711625607646,
+      "epoch": 0.5682463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 8.870064452942435e-06,
-      "loss": 0.9749,
+      "learning_rate": 8.284339737319344e-06,
+      "loss": 0.7349,
       "step": 20025
     },
     {
-      "epoch": 0.5499986267886079,
+      "epoch": 0.5682746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 8.869180638109969e-06,
-      "loss": 0.9071,
+      "learning_rate": 8.2834342940146e-06,
+      "loss": 0.8345,
       "step": 20026
     },
     {
-      "epoch": 0.5500260910164511,
+      "epoch": 0.5683030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 8.868296832224984e-06,
-      "loss": 0.9167,
+      "learning_rate": 8.282528865209762e-06,
+      "loss": 0.8341,
       "step": 20027
     },
     {
-      "epoch": 0.5500535552442943,
+      "epoch": 0.5683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 8.867413035294474e-06,
-      "loss": 0.9362,
+      "learning_rate": 8.281623450912479e-06,
+      "loss": 0.9253,
       "step": 20028
     },
     {
-      "epoch": 0.5500810194721375,
+      "epoch": 0.5683598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 8.866529247325428e-06,
-      "loss": 0.9337,
+      "learning_rate": 8.280718051130396e-06,
+      "loss": 0.875,
       "step": 20029
     },
     {
-      "epoch": 0.5501084836999808,
+      "epoch": 0.5683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 8.865645468324849e-06,
-      "loss": 0.8429,
+      "learning_rate": 8.279812665871171e-06,
+      "loss": 0.9682,
       "step": 20030
     },
     {
-      "epoch": 0.550135947927824,
+      "epoch": 0.568416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 8.864761698299715e-06,
-      "loss": 0.9225,
+      "learning_rate": 8.278907295142444e-06,
+      "loss": 0.8743,
       "step": 20031
     },
     {
-      "epoch": 0.5501634121556672,
+      "epoch": 0.56844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 8.863877937257029e-06,
-      "loss": 0.845,
+      "learning_rate": 8.278001938951865e-06,
+      "loss": 0.8647,
       "step": 20032
     },
     {
-      "epoch": 0.5501908763835105,
+      "epoch": 0.5684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 8.86299418520378e-06,
-      "loss": 0.9115,
+      "learning_rate": 8.27709659730708e-06,
+      "loss": 0.9445,
       "step": 20033
     },
     {
-      "epoch": 0.5502183406113537,
+      "epoch": 0.5685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 8.86211044214696e-06,
-      "loss": 0.8229,
+      "learning_rate": 8.276191270215739e-06,
+      "loss": 0.8042,
       "step": 20034
     },
     {
-      "epoch": 0.5502458048391969,
+      "epoch": 0.5685300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 8.861226708093566e-06,
-      "loss": 0.9581,
+      "learning_rate": 8.275285957685488e-06,
+      "loss": 0.871,
       "step": 20035
     },
     {
-      "epoch": 0.5502732690670402,
+      "epoch": 0.5685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 8.860342983050588e-06,
-      "loss": 0.8029,
+      "learning_rate": 8.274380659723971e-06,
+      "loss": 0.8389,
       "step": 20036
     },
     {
-      "epoch": 0.5503007332948834,
+      "epoch": 0.5685868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 8.859459267025016e-06,
-      "loss": 0.76,
+      "learning_rate": 8.27347537633884e-06,
+      "loss": 0.8382,
       "step": 20037
     },
     {
-      "epoch": 0.5503281975227267,
+      "epoch": 0.5686152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 8.858575560023843e-06,
-      "loss": 0.8524,
+      "learning_rate": 8.272570107537742e-06,
+      "loss": 0.8459,
       "step": 20038
     },
     {
-      "epoch": 0.5503556617505699,
+      "epoch": 0.5686435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 8.857691862054062e-06,
-      "loss": 0.8693,
+      "learning_rate": 8.271664853328319e-06,
+      "loss": 0.7855,
       "step": 20039
     },
     {
-      "epoch": 0.5503831259784131,
+      "epoch": 0.5686719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 8.856808173122667e-06,
-      "loss": 0.9401,
+      "learning_rate": 8.270759613718222e-06,
+      "loss": 0.9059,
       "step": 20040
     },
     {
-      "epoch": 0.5504105902062564,
+      "epoch": 0.5687003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 8.855924493236647e-06,
-      "loss": 0.8846,
+      "learning_rate": 8.269854388715098e-06,
+      "loss": 0.7835,
       "step": 20041
     },
     {
-      "epoch": 0.5504380544340995,
+      "epoch": 0.5687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 8.855040822403e-06,
-      "loss": 0.8405,
+      "learning_rate": 8.268949178326589e-06,
+      "loss": 0.7972,
       "step": 20042
     },
     {
-      "epoch": 0.5504655186619428,
+      "epoch": 0.5687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 8.85415716062871e-06,
-      "loss": 0.7429,
+      "learning_rate": 8.268043982560345e-06,
+      "loss": 0.8425,
       "step": 20043
     },
     {
-      "epoch": 0.5504929828897861,
+      "epoch": 0.5687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 8.853273507920774e-06,
-      "loss": 0.8686,
+      "learning_rate": 8.267138801424007e-06,
+      "loss": 0.8173,
       "step": 20044
     },
     {
-      "epoch": 0.5505204471176293,
+      "epoch": 0.5688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 8.85238986428618e-06,
-      "loss": 0.938,
+      "learning_rate": 8.266233634925232e-06,
+      "loss": 0.9437,
       "step": 20045
     },
     {
-      "epoch": 0.5505479113454725,
+      "epoch": 0.5688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 8.851506229731922e-06,
-      "loss": 0.8941,
+      "learning_rate": 8.265328483071655e-06,
+      "loss": 0.9696,
       "step": 20046
     },
     {
-      "epoch": 0.5505753755733157,
+      "epoch": 0.5688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 8.850622604264993e-06,
-      "loss": 0.7814,
+      "learning_rate": 8.264423345870928e-06,
+      "loss": 0.9821,
       "step": 20047
     },
     {
-      "epoch": 0.550602839801159,
+      "epoch": 0.5688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 8.849738987892385e-06,
-      "loss": 0.8438,
+      "learning_rate": 8.263518223330698e-06,
+      "loss": 0.9003,
       "step": 20048
     },
     {
-      "epoch": 0.5506303040290023,
+      "epoch": 0.5689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 8.848855380621085e-06,
-      "loss": 0.9373,
+      "learning_rate": 8.262613115458604e-06,
+      "loss": 0.8299,
       "step": 20049
     },
     {
-      "epoch": 0.5506577682568454,
+      "epoch": 0.568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 8.847971782458087e-06,
-      "loss": 0.8993,
+      "learning_rate": 8.261708022262295e-06,
+      "loss": 0.897,
       "step": 20050
     },
     {
-      "epoch": 0.5506852324846887,
+      "epoch": 0.5689841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 8.847088193410381e-06,
-      "loss": 0.8735,
+      "learning_rate": 8.26080294374942e-06,
+      "loss": 0.9781,
       "step": 20051
     },
     {
-      "epoch": 0.550712696712532,
+      "epoch": 0.5690124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 8.846204613484963e-06,
-      "loss": 0.911,
+      "learning_rate": 8.259897879927616e-06,
+      "loss": 0.8192,
       "step": 20052
     },
     {
-      "epoch": 0.5507401609403751,
+      "epoch": 0.5690408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 8.84532104268882e-06,
-      "loss": 0.9009,
+      "learning_rate": 8.258992830804538e-06,
+      "loss": 0.7958,
       "step": 20053
     },
     {
-      "epoch": 0.5507676251682184,
+      "epoch": 0.5690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 8.844437481028948e-06,
-      "loss": 0.8488,
+      "learning_rate": 8.258087796387823e-06,
+      "loss": 0.8515,
       "step": 20054
     },
     {
-      "epoch": 0.5507950893960616,
+      "epoch": 0.5690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 8.843553928512329e-06,
-      "loss": 0.8145,
+      "learning_rate": 8.257182776685118e-06,
+      "loss": 0.6908,
       "step": 20055
     },
     {
-      "epoch": 0.5508225536239049,
+      "epoch": 0.5691259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 8.842670385145961e-06,
-      "loss": 0.8876,
+      "learning_rate": 8.256277771704069e-06,
+      "loss": 0.8519,
       "step": 20056
     },
     {
-      "epoch": 0.5508500178517481,
+      "epoch": 0.5691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 8.841786850936831e-06,
-      "loss": 0.8766,
+      "learning_rate": 8.255372781452322e-06,
+      "loss": 0.8407,
       "step": 20057
     },
     {
-      "epoch": 0.5508774820795913,
+      "epoch": 0.5691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 8.840903325891934e-06,
-      "loss": 0.8995,
+      "learning_rate": 8.254467805937518e-06,
+      "loss": 0.8293,
       "step": 20058
     },
     {
-      "epoch": 0.5509049463074346,
+      "epoch": 0.569211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 8.840019810018258e-06,
-      "loss": 0.8995,
+      "learning_rate": 8.2535628451673e-06,
+      "loss": 0.812,
       "step": 20059
     },
     {
-      "epoch": 0.5509324105352778,
+      "epoch": 0.5692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 8.839136303322797e-06,
-      "loss": 0.858,
+      "learning_rate": 8.252657899149316e-06,
+      "loss": 0.9187,
       "step": 20060
     },
     {
-      "epoch": 0.550959874763121,
+      "epoch": 0.5692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 8.83825280581254e-06,
-      "loss": 0.8201,
+      "learning_rate": 8.251752967891211e-06,
+      "loss": 0.9028,
       "step": 20061
     },
     {
-      "epoch": 0.5509873389909643,
+      "epoch": 0.5692962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 8.837369317494472e-06,
-      "loss": 0.9008,
+      "learning_rate": 8.250848051400628e-06,
+      "loss": 0.9074,
       "step": 20062
     },
     {
-      "epoch": 0.5510148032188075,
+      "epoch": 0.5693246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 8.836485838375591e-06,
-      "loss": 0.9429,
+      "learning_rate": 8.249943149685209e-06,
+      "loss": 0.8756,
       "step": 20063
     },
     {
-      "epoch": 0.5510422674466507,
+      "epoch": 0.5693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 8.835602368462886e-06,
-      "loss": 0.8181,
+      "learning_rate": 8.2490382627526e-06,
+      "loss": 0.8967,
       "step": 20064
     },
     {
-      "epoch": 0.551069731674494,
+      "epoch": 0.5693813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 8.834718907763342e-06,
-      "loss": 0.8514,
+      "learning_rate": 8.248133390610445e-06,
+      "loss": 0.9267,
       "step": 20065
     },
     {
-      "epoch": 0.5510971959023372,
+      "epoch": 0.5694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 8.833835456283961e-06,
-      "loss": 0.872,
+      "learning_rate": 8.247228533266381e-06,
+      "loss": 0.9227,
       "step": 20066
     },
     {
-      "epoch": 0.5511246601301805,
+      "epoch": 0.5694381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 8.832952014031718e-06,
-      "loss": 0.9219,
+      "learning_rate": 8.24632369072806e-06,
+      "loss": 0.8578,
       "step": 20067
     },
     {
-      "epoch": 0.5511521243580236,
+      "epoch": 0.569466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 8.832068581013616e-06,
-      "loss": 0.8587,
+      "learning_rate": 8.245418863003123e-06,
+      "loss": 0.9749,
       "step": 20068
     },
     {
-      "epoch": 0.5511795885858669,
+      "epoch": 0.5694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 8.831185157236635e-06,
-      "loss": 0.8156,
+      "learning_rate": 8.244514050099207e-06,
+      "loss": 0.8823,
       "step": 20069
     },
     {
-      "epoch": 0.5512070528137102,
+      "epoch": 0.5695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 8.83030174270777e-06,
-      "loss": 0.8566,
+      "learning_rate": 8.243609252023964e-06,
+      "loss": 0.9039,
       "step": 20070
     },
     {
-      "epoch": 0.5512345170415534,
+      "epoch": 0.5695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 8.829418337434014e-06,
-      "loss": 0.9255,
+      "learning_rate": 8.242704468785032e-06,
+      "loss": 0.9978,
       "step": 20071
     },
     {
-      "epoch": 0.5512619812693966,
+      "epoch": 0.5695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 8.828534941422353e-06,
-      "loss": 0.9286,
+      "learning_rate": 8.241799700390052e-06,
+      "loss": 0.81,
       "step": 20072
     },
     {
-      "epoch": 0.5512894454972398,
+      "epoch": 0.5696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 8.827651554679777e-06,
-      "loss": 0.7726,
+      "learning_rate": 8.24089494684667e-06,
+      "loss": 0.8529,
       "step": 20073
     },
     {
-      "epoch": 0.5513169097250831,
+      "epoch": 0.5696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 8.826768177213276e-06,
-      "loss": 0.8718,
+      "learning_rate": 8.23999020816253e-06,
+      "loss": 0.806,
       "step": 20074
     },
     {
-      "epoch": 0.5513443739529263,
+      "epoch": 0.5696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 8.825884809029837e-06,
-      "loss": 0.9343,
+      "learning_rate": 8.239085484345265e-06,
+      "loss": 0.9036,
       "step": 20075
     },
     {
-      "epoch": 0.5513718381807695,
+      "epoch": 0.5696935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 8.825001450136456e-06,
-      "loss": 0.8581,
+      "learning_rate": 8.23818077540253e-06,
+      "loss": 0.8818,
       "step": 20076
     },
     {
-      "epoch": 0.5513993024086128,
+      "epoch": 0.5697219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 8.824118100540117e-06,
-      "loss": 1.0109,
+      "learning_rate": 8.237276081341959e-06,
+      "loss": 0.7996,
       "step": 20077
     },
     {
-      "epoch": 0.5514267666364561,
+      "epoch": 0.5697502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 8.823234760247813e-06,
-      "loss": 0.8503,
+      "learning_rate": 8.236371402171198e-06,
+      "loss": 0.8601,
       "step": 20078
     },
     {
-      "epoch": 0.5514542308642992,
+      "epoch": 0.5697786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 8.82235142926653e-06,
-      "loss": 0.8898,
+      "learning_rate": 8.235466737897886e-06,
+      "loss": 0.9178,
       "step": 20079
     },
     {
-      "epoch": 0.5514816950921425,
+      "epoch": 0.5698070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 8.821468107603258e-06,
-      "loss": 0.8948,
+      "learning_rate": 8.234562088529665e-06,
+      "loss": 0.8898,
       "step": 20080
     },
     {
-      "epoch": 0.5515091593199857,
+      "epoch": 0.5698354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 8.820584795264987e-06,
-      "loss": 0.813,
+      "learning_rate": 8.23365745407418e-06,
+      "loss": 0.8932,
       "step": 20081
     },
     {
-      "epoch": 0.551536623547829,
+      "epoch": 0.5698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 8.819701492258704e-06,
-      "loss": 0.8352,
+      "learning_rate": 8.232752834539068e-06,
+      "loss": 0.9774,
       "step": 20082
     },
     {
-      "epoch": 0.5515640877756722,
+      "epoch": 0.5698921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 8.818818198591406e-06,
-      "loss": 0.8064,
+      "learning_rate": 8.23184822993197e-06,
+      "loss": 0.9531,
       "step": 20083
     },
     {
-      "epoch": 0.5515915520035154,
+      "epoch": 0.5699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 8.817934914270074e-06,
-      "loss": 0.9446,
+      "learning_rate": 8.230943640260533e-06,
+      "loss": 0.935,
       "step": 20084
     },
     {
-      "epoch": 0.5516190162313587,
+      "epoch": 0.5699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 8.817051639301695e-06,
-      "loss": 0.7979,
+      "learning_rate": 8.230039065532395e-06,
+      "loss": 0.817,
       "step": 20085
     },
     {
-      "epoch": 0.5516464804592018,
+      "epoch": 0.569977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 8.816168373693262e-06,
-      "loss": 0.9294,
+      "learning_rate": 8.229134505755192e-06,
+      "loss": 0.9216,
       "step": 20086
     },
     {
-      "epoch": 0.5516739446870451,
+      "epoch": 0.570005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 8.815285117451764e-06,
-      "loss": 0.8451,
+      "learning_rate": 8.228229960936574e-06,
+      "loss": 0.9606,
       "step": 20087
     },
     {
-      "epoch": 0.5517014089148884,
+      "epoch": 0.5700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 8.814401870584191e-06,
-      "loss": 0.8335,
+      "learning_rate": 8.227325431084176e-06,
+      "loss": 0.9147,
       "step": 20088
     },
     {
-      "epoch": 0.5517288731427316,
+      "epoch": 0.5700624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 8.813518633097531e-06,
-      "loss": 0.9446,
+      "learning_rate": 8.22642091620564e-06,
+      "loss": 0.7551,
       "step": 20089
     },
     {
-      "epoch": 0.5517563373705748,
+      "epoch": 0.5700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 8.812635404998768e-06,
-      "loss": 0.9161,
+      "learning_rate": 8.225516416308601e-06,
+      "loss": 0.9154,
       "step": 20090
     },
     {
-      "epoch": 0.5517838015984181,
+      "epoch": 0.5701191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 8.811752186294895e-06,
-      "loss": 0.8907,
+      "learning_rate": 8.22461193140071e-06,
+      "loss": 0.7389,
       "step": 20091
     },
     {
-      "epoch": 0.5518112658262613,
+      "epoch": 0.5701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 8.810868976992899e-06,
-      "loss": 0.8795,
+      "learning_rate": 8.223707461489602e-06,
+      "loss": 0.7355,
       "step": 20092
     },
     {
-      "epoch": 0.5518387300541046,
+      "epoch": 0.5701759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 8.809985777099767e-06,
-      "loss": 0.8504,
+      "learning_rate": 8.222803006582915e-06,
+      "loss": 0.9361,
       "step": 20093
     },
     {
-      "epoch": 0.5518661942819477,
+      "epoch": 0.5702043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 8.809102586622487e-06,
-      "loss": 0.8616,
+      "learning_rate": 8.221898566688294e-06,
+      "loss": 0.8297,
       "step": 20094
     },
     {
-      "epoch": 0.551893658509791,
+      "epoch": 0.5702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 8.808219405568054e-06,
-      "loss": 0.818,
+      "learning_rate": 8.220994141813373e-06,
+      "loss": 0.9845,
       "step": 20095
     },
     {
-      "epoch": 0.5519211227376343,
+      "epoch": 0.570261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.807336233943449e-06,
-      "loss": 0.8864,
+      "learning_rate": 8.220089731965794e-06,
+      "loss": 0.7707,
       "step": 20096
     },
     {
-      "epoch": 0.5519485869654774,
+      "epoch": 0.5702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.80645307175566e-06,
-      "loss": 0.9814,
+      "learning_rate": 8.219185337153199e-06,
+      "loss": 0.8861,
       "step": 20097
     },
     {
-      "epoch": 0.5519760511933207,
+      "epoch": 0.5703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 8.805569919011675e-06,
-      "loss": 0.8533,
+      "learning_rate": 8.218280957383225e-06,
+      "loss": 0.8417,
       "step": 20098
     },
     {
-      "epoch": 0.5520035154211639,
+      "epoch": 0.5703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 8.804686775718484e-06,
-      "loss": 0.8967,
+      "learning_rate": 8.217376592663513e-06,
+      "loss": 0.8711,
       "step": 20099
     },
     {
-      "epoch": 0.5520309796490072,
+      "epoch": 0.5703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 8.803803641883077e-06,
-      "loss": 0.8634,
+      "learning_rate": 8.2164722430017e-06,
+      "loss": 0.7962,
       "step": 20100
     },
     {
-      "epoch": 0.5520584438768504,
+      "epoch": 0.5704029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 8.802920517512441e-06,
-      "loss": 0.9271,
+      "learning_rate": 8.215567908405424e-06,
+      "loss": 0.8359,
       "step": 20101
     },
     {
-      "epoch": 0.5520859081046936,
+      "epoch": 0.5704313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 8.802037402613556e-06,
-      "loss": 0.7767,
+      "learning_rate": 8.21466358888233e-06,
+      "loss": 0.8777,
       "step": 20102
     },
     {
-      "epoch": 0.5521133723325369,
+      "epoch": 0.5704597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 8.801154297193418e-06,
-      "loss": 0.8724,
+      "learning_rate": 8.213759284440049e-06,
+      "loss": 0.7977,
       "step": 20103
     },
     {
-      "epoch": 0.5521408365603802,
+      "epoch": 0.5704880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 8.800271201259011e-06,
-      "loss": 0.8648,
+      "learning_rate": 8.212854995086225e-06,
+      "loss": 0.826,
       "step": 20104
     },
     {
-      "epoch": 0.5521683007882233,
+      "epoch": 0.570516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 8.799388114817321e-06,
-      "loss": 0.876,
+      "learning_rate": 8.211950720828496e-06,
+      "loss": 0.8206,
       "step": 20105
     },
     {
-      "epoch": 0.5521957650160666,
+      "epoch": 0.5705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 8.798505037875338e-06,
-      "loss": 0.942,
+      "learning_rate": 8.211046461674497e-06,
+      "loss": 0.8243,
       "step": 20106
     },
     {
-      "epoch": 0.5522232292439098,
+      "epoch": 0.5705732122587969,
       "grad_norm": 0.0,
-      "learning_rate": 8.797621970440053e-06,
-      "loss": 0.9986,
+      "learning_rate": 8.21014221763187e-06,
+      "loss": 0.8226,
       "step": 20107
     },
     {
-      "epoch": 0.552250693471753,
+      "epoch": 0.5706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 8.796738912518447e-06,
-      "loss": 0.9526,
+      "learning_rate": 8.209237988708254e-06,
+      "loss": 0.7995,
       "step": 20108
     },
     {
-      "epoch": 0.5522781576995963,
+      "epoch": 0.5706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 8.795855864117505e-06,
-      "loss": 0.8785,
+      "learning_rate": 8.208333774911284e-06,
+      "loss": 0.8886,
       "step": 20109
     },
     {
-      "epoch": 0.5523056219274395,
+      "epoch": 0.5706583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 8.79497282524422e-06,
-      "loss": 0.8355,
+      "learning_rate": 8.2074295762486e-06,
+      "loss": 0.8375,
       "step": 20110
     },
     {
-      "epoch": 0.5523330861552828,
+      "epoch": 0.5706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 8.794089795905575e-06,
-      "loss": 0.8789,
+      "learning_rate": 8.20652539272784e-06,
+      "loss": 0.8236,
       "step": 20111
     },
     {
-      "epoch": 0.5523605503831259,
+      "epoch": 0.5707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 8.793206776108563e-06,
-      "loss": 0.952,
+      "learning_rate": 8.20562122435664e-06,
+      "loss": 0.9503,
       "step": 20112
     },
     {
-      "epoch": 0.5523880146109692,
+      "epoch": 0.5707434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 8.792323765860167e-06,
-      "loss": 0.8752,
+      "learning_rate": 8.204717071142635e-06,
+      "loss": 0.9234,
       "step": 20113
     },
     {
-      "epoch": 0.5524154788388125,
+      "epoch": 0.5707718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 8.79144076516737e-06,
-      "loss": 0.8348,
+      "learning_rate": 8.203812933093469e-06,
+      "loss": 0.9425,
       "step": 20114
     },
     {
-      "epoch": 0.5524429430666556,
+      "epoch": 0.570800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 8.79055777403716e-06,
-      "loss": 0.8464,
+      "learning_rate": 8.202908810216775e-06,
+      "loss": 0.9304,
       "step": 20115
     },
     {
-      "epoch": 0.5524704072944989,
+      "epoch": 0.5708286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 8.78967479247653e-06,
-      "loss": 0.9049,
+      "learning_rate": 8.20200470252019e-06,
+      "loss": 0.8814,
       "step": 20116
     },
     {
-      "epoch": 0.5524978715223422,
+      "epoch": 0.5708569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 8.788791820492457e-06,
-      "loss": 0.9656,
+      "learning_rate": 8.201100610011355e-06,
+      "loss": 0.746,
       "step": 20117
     },
     {
-      "epoch": 0.5525253357501854,
+      "epoch": 0.5708853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 8.787908858091934e-06,
-      "loss": 0.8452,
+      "learning_rate": 8.200196532697901e-06,
+      "loss": 0.8648,
       "step": 20118
     },
     {
-      "epoch": 0.5525527999780286,
+      "epoch": 0.5709137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 8.787025905281949e-06,
-      "loss": 0.9952,
+      "learning_rate": 8.199292470587468e-06,
+      "loss": 0.8534,
       "step": 20119
     },
     {
-      "epoch": 0.5525802642058718,
+      "epoch": 0.5709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 8.786142962069483e-06,
-      "loss": 0.808,
+      "learning_rate": 8.198388423687694e-06,
+      "loss": 0.8586,
       "step": 20120
     },
     {
-      "epoch": 0.5526077284337151,
+      "epoch": 0.5709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 8.785260028461523e-06,
-      "loss": 0.8407,
+      "learning_rate": 8.19748439200621e-06,
+      "loss": 0.7806,
       "step": 20121
     },
     {
-      "epoch": 0.5526351926615584,
+      "epoch": 0.5709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 8.784377104465053e-06,
-      "loss": 0.778,
+      "learning_rate": 8.19658037555066e-06,
+      "loss": 0.9643,
       "step": 20122
     },
     {
-      "epoch": 0.5526626568894015,
+      "epoch": 0.5710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 8.783494190087067e-06,
-      "loss": 0.9985,
+      "learning_rate": 8.195676374328679e-06,
+      "loss": 0.8096,
       "step": 20123
     },
     {
-      "epoch": 0.5526901211172448,
+      "epoch": 0.57105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.782611285334544e-06,
-      "loss": 0.9652,
+      "learning_rate": 8.194772388347896e-06,
+      "loss": 0.87,
       "step": 20124
     },
     {
-      "epoch": 0.552717585345088,
+      "epoch": 0.5710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 8.781728390214476e-06,
-      "loss": 0.9027,
+      "learning_rate": 8.193868417615954e-06,
+      "loss": 0.8695,
       "step": 20125
     },
     {
-      "epoch": 0.5527450495729312,
+      "epoch": 0.5711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 8.780845504733842e-06,
-      "loss": 0.9024,
+      "learning_rate": 8.192964462140487e-06,
+      "loss": 0.8813,
       "step": 20126
     },
     {
-      "epoch": 0.5527725138007745,
+      "epoch": 0.5711407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 8.779962628899629e-06,
-      "loss": 0.9022,
+      "learning_rate": 8.19206052192913e-06,
+      "loss": 0.8589,
       "step": 20127
     },
     {
-      "epoch": 0.5527999780286177,
+      "epoch": 0.5711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 8.779079762718826e-06,
-      "loss": 0.8815,
+      "learning_rate": 8.19115659698952e-06,
+      "loss": 0.8864,
       "step": 20128
     },
     {
-      "epoch": 0.552827442256461,
+      "epoch": 0.5711975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 8.778196906198416e-06,
-      "loss": 1.0436,
+      "learning_rate": 8.19025268732929e-06,
+      "loss": 1.0099,
       "step": 20129
     },
     {
-      "epoch": 0.5528549064843042,
+      "epoch": 0.5712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 8.777314059345385e-06,
-      "loss": 0.926,
+      "learning_rate": 8.189348792956076e-06,
+      "loss": 0.9211,
       "step": 20130
     },
     {
-      "epoch": 0.5528823707121474,
+      "epoch": 0.5712542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 8.776431222166723e-06,
-      "loss": 0.9393,
+      "learning_rate": 8.188444913877518e-06,
+      "loss": 1.0181,
       "step": 20131
     },
     {
-      "epoch": 0.5529098349399907,
+      "epoch": 0.5712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 8.775548394669409e-06,
-      "loss": 0.8881,
+      "learning_rate": 8.187541050101245e-06,
+      "loss": 0.8907,
       "step": 20132
     },
     {
-      "epoch": 0.5529372991678339,
+      "epoch": 0.571311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 8.774665576860428e-06,
-      "loss": 0.9283,
+      "learning_rate": 8.186637201634892e-06,
+      "loss": 0.8091,
       "step": 20133
     },
     {
-      "epoch": 0.5529647633956771,
+      "epoch": 0.571339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 8.773782768746769e-06,
-      "loss": 0.8849,
+      "learning_rate": 8.185733368486099e-06,
+      "loss": 0.8372,
       "step": 20134
     },
     {
-      "epoch": 0.5529922276235204,
+      "epoch": 0.5713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 8.772899970335416e-06,
-      "loss": 0.9106,
+      "learning_rate": 8.184829550662498e-06,
+      "loss": 1.0092,
       "step": 20135
     },
     {
-      "epoch": 0.5530196918513636,
+      "epoch": 0.5713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 8.77201718163335e-06,
-      "loss": 0.8796,
+      "learning_rate": 8.18392574817172e-06,
+      "loss": 0.8702,
       "step": 20136
     },
     {
-      "epoch": 0.5530471560792068,
+      "epoch": 0.5714245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 8.771134402647568e-06,
-      "loss": 0.8779,
+      "learning_rate": 8.183021961021402e-06,
+      "loss": 0.9171,
       "step": 20137
     },
     {
-      "epoch": 0.55307462030705,
+      "epoch": 0.5714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 8.77025163338504e-06,
-      "loss": 0.9483,
+      "learning_rate": 8.182118189219183e-06,
+      "loss": 0.8976,
       "step": 20138
     },
     {
-      "epoch": 0.5531020845348933,
+      "epoch": 0.5714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 8.769368873852758e-06,
-      "loss": 0.8002,
+      "learning_rate": 8.181214432772694e-06,
+      "loss": 0.8696,
       "step": 20139
     },
     {
-      "epoch": 0.5531295487627366,
+      "epoch": 0.5715096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 8.768486124057708e-06,
-      "loss": 0.8123,
+      "learning_rate": 8.180310691689569e-06,
+      "loss": 0.8896,
       "step": 20140
     },
     {
-      "epoch": 0.5531570129905797,
+      "epoch": 0.5715380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 8.767603384006871e-06,
-      "loss": 0.9327,
+      "learning_rate": 8.179406965977438e-06,
+      "loss": 0.89,
       "step": 20141
     },
     {
-      "epoch": 0.553184477218423,
+      "epoch": 0.5715664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 8.766720653707238e-06,
-      "loss": 0.8442,
+      "learning_rate": 8.17850325564394e-06,
+      "loss": 0.8839,
       "step": 20142
     },
     {
-      "epoch": 0.5532119414462663,
+      "epoch": 0.5715947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 8.765837933165786e-06,
-      "loss": 0.9382,
+      "learning_rate": 8.177599560696707e-06,
+      "loss": 0.8628,
       "step": 20143
     },
     {
-      "epoch": 0.5532394056741095,
+      "epoch": 0.5716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 8.764955222389502e-06,
-      "loss": 0.7746,
+      "learning_rate": 8.176695881143371e-06,
+      "loss": 0.8739,
       "step": 20144
     },
     {
-      "epoch": 0.5532668699019527,
+      "epoch": 0.5716515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 8.764072521385369e-06,
-      "loss": 0.8982,
+      "learning_rate": 8.17579221699157e-06,
+      "loss": 0.888,
       "step": 20145
     },
     {
-      "epoch": 0.5532943341297959,
+      "epoch": 0.5716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 8.763189830160374e-06,
-      "loss": 0.8814,
+      "learning_rate": 8.174888568248931e-06,
+      "loss": 0.8631,
       "step": 20146
     },
     {
-      "epoch": 0.5533217983576392,
+      "epoch": 0.5717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 8.762307148721501e-06,
-      "loss": 0.8371,
+      "learning_rate": 8.173984934923091e-06,
+      "loss": 0.9469,
       "step": 20147
     },
     {
-      "epoch": 0.5533492625854824,
+      "epoch": 0.5717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 8.761424477075733e-06,
-      "loss": 0.784,
+      "learning_rate": 8.173081317021684e-06,
+      "loss": 0.6925,
       "step": 20148
     },
     {
-      "epoch": 0.5533767268133256,
+      "epoch": 0.5717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 8.760541815230056e-06,
-      "loss": 0.8611,
+      "learning_rate": 8.17217771455234e-06,
+      "loss": 0.9021,
       "step": 20149
     },
     {
-      "epoch": 0.5534041910411689,
+      "epoch": 0.5717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 8.75965916319145e-06,
-      "loss": 0.8497,
+      "learning_rate": 8.171274127522692e-06,
+      "loss": 0.8752,
       "step": 20150
     },
     {
-      "epoch": 0.5534316552690121,
+      "epoch": 0.571821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 8.7587765209669e-06,
-      "loss": 0.9365,
+      "learning_rate": 8.170370555940375e-06,
+      "loss": 0.9186,
       "step": 20151
     },
     {
-      "epoch": 0.5534591194968553,
+      "epoch": 0.571850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 8.757893888563393e-06,
-      "loss": 0.8808,
+      "learning_rate": 8.169466999813016e-06,
+      "loss": 0.8066,
       "step": 20152
     },
     {
-      "epoch": 0.5534865837246986,
+      "epoch": 0.5718785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 8.75701126598791e-06,
-      "loss": 0.8409,
+      "learning_rate": 8.168563459148257e-06,
+      "loss": 0.7665,
       "step": 20153
     },
     {
-      "epoch": 0.5535140479525418,
+      "epoch": 0.5719069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 8.756128653247439e-06,
-      "loss": 0.8324,
+      "learning_rate": 8.167659933953723e-06,
+      "loss": 0.9104,
       "step": 20154
     },
     {
-      "epoch": 0.553541512180385,
+      "epoch": 0.5719353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 8.755246050348956e-06,
-      "loss": 0.8255,
+      "learning_rate": 8.166756424237047e-06,
+      "loss": 0.8349,
       "step": 20155
     },
     {
-      "epoch": 0.5535689764082283,
+      "epoch": 0.5719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 8.75436345729945e-06,
-      "loss": 0.8344,
+      "learning_rate": 8.165852930005865e-06,
+      "loss": 0.7948,
       "step": 20156
     },
     {
-      "epoch": 0.5535964406360715,
+      "epoch": 0.5719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 8.7534808741059e-06,
-      "loss": 0.8656,
+      "learning_rate": 8.1649494512678e-06,
+      "loss": 0.8519,
       "step": 20157
     },
     {
-      "epoch": 0.5536239048639148,
+      "epoch": 0.5720204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 8.752598300775294e-06,
-      "loss": 0.925,
+      "learning_rate": 8.164045988030495e-06,
+      "loss": 0.8513,
       "step": 20158
     },
     {
-      "epoch": 0.5536513690917579,
+      "epoch": 0.5720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 8.751715737314614e-06,
-      "loss": 0.883,
+      "learning_rate": 8.163142540301575e-06,
+      "loss": 0.8241,
       "step": 20159
     },
     {
-      "epoch": 0.5536788333196012,
+      "epoch": 0.5720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 8.750833183730843e-06,
-      "loss": 0.833,
+      "learning_rate": 8.162239108088671e-06,
+      "loss": 0.8732,
       "step": 20160
     },
     {
-      "epoch": 0.5537062975474445,
+      "epoch": 0.572105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.749950640030964e-06,
-      "loss": 0.842,
+      "learning_rate": 8.161335691399417e-06,
+      "loss": 0.7513,
       "step": 20161
     },
     {
-      "epoch": 0.5537337617752877,
+      "epoch": 0.5721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 8.749068106221958e-06,
-      "loss": 0.859,
+      "learning_rate": 8.160432290241443e-06,
+      "loss": 0.7617,
       "step": 20162
     },
     {
-      "epoch": 0.5537612260031309,
+      "epoch": 0.5721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 8.748185582310809e-06,
-      "loss": 0.8586,
+      "learning_rate": 8.159528904622379e-06,
+      "loss": 0.8395,
       "step": 20163
     },
     {
-      "epoch": 0.5537886902309741,
+      "epoch": 0.5721906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 8.747303068304503e-06,
-      "loss": 0.8739,
+      "learning_rate": 8.158625534549859e-06,
+      "loss": 0.904,
       "step": 20164
     },
     {
-      "epoch": 0.5538161544588174,
+      "epoch": 0.5722190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 8.746420564210018e-06,
-      "loss": 0.894,
+      "learning_rate": 8.157722180031512e-06,
+      "loss": 0.8971,
       "step": 20165
     },
     {
-      "epoch": 0.5538436186866607,
+      "epoch": 0.5722474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 8.745538070034343e-06,
-      "loss": 0.8684,
+      "learning_rate": 8.156818841074967e-06,
+      "loss": 0.8804,
       "step": 20166
     },
     {
-      "epoch": 0.5538710829145038,
+      "epoch": 0.5722758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 8.744655585784452e-06,
-      "loss": 0.8568,
+      "learning_rate": 8.155915517687854e-06,
+      "loss": 0.9313,
       "step": 20167
     },
     {
-      "epoch": 0.5538985471423471,
+      "epoch": 0.5723041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 8.743773111467336e-06,
-      "loss": 0.7866,
+      "learning_rate": 8.155012209877807e-06,
+      "loss": 0.8781,
       "step": 20168
     },
     {
-      "epoch": 0.5539260113701904,
+      "epoch": 0.5723325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 8.742890647089971e-06,
-      "loss": 0.8851,
+      "learning_rate": 8.154108917652456e-06,
+      "loss": 0.8781,
       "step": 20169
     },
     {
-      "epoch": 0.5539534755980335,
+      "epoch": 0.572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 8.742008192659342e-06,
-      "loss": 0.8252,
+      "learning_rate": 8.15320564101943e-06,
+      "loss": 0.8812,
       "step": 20170
     },
     {
-      "epoch": 0.5539809398258768,
+      "epoch": 0.5723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 8.741125748182434e-06,
-      "loss": 0.9668,
+      "learning_rate": 8.152302379986358e-06,
+      "loss": 0.7483,
       "step": 20171
     },
     {
-      "epoch": 0.55400840405372,
+      "epoch": 0.5724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 8.740243313666227e-06,
-      "loss": 0.9244,
+      "learning_rate": 8.151399134560872e-06,
+      "loss": 0.8851,
       "step": 20172
     },
     {
-      "epoch": 0.5540358682815633,
+      "epoch": 0.5724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 8.739360889117702e-06,
-      "loss": 0.9467,
+      "learning_rate": 8.150495904750601e-06,
+      "loss": 0.7915,
       "step": 20173
     },
     {
-      "epoch": 0.5540633325094065,
+      "epoch": 0.5724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 8.73847847454384e-06,
-      "loss": 0.9375,
+      "learning_rate": 8.149592690563172e-06,
+      "loss": 0.894,
       "step": 20174
     },
     {
-      "epoch": 0.5540907967372497,
+      "epoch": 0.5725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 8.737596069951626e-06,
-      "loss": 1.0272,
+      "learning_rate": 8.148689492006218e-06,
+      "loss": 0.9236,
       "step": 20175
     },
     {
-      "epoch": 0.554118260965093,
+      "epoch": 0.5725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 8.736713675348042e-06,
-      "loss": 0.9161,
+      "learning_rate": 8.14778630908737e-06,
+      "loss": 0.9329,
       "step": 20176
     },
     {
-      "epoch": 0.5541457251929363,
+      "epoch": 0.5725595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 8.735831290740066e-06,
-      "loss": 0.8679,
+      "learning_rate": 8.146883141814249e-06,
+      "loss": 0.9032,
       "step": 20177
     },
     {
-      "epoch": 0.5541731894207794,
+      "epoch": 0.5725879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 8.734948916134688e-06,
-      "loss": 0.8234,
+      "learning_rate": 8.145979990194492e-06,
+      "loss": 0.9201,
       "step": 20178
     },
     {
-      "epoch": 0.5542006536486227,
+      "epoch": 0.5726163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 8.73406655153888e-06,
-      "loss": 0.9161,
+      "learning_rate": 8.145076854235728e-06,
+      "loss": 0.8516,
       "step": 20179
     },
     {
-      "epoch": 0.5542281178764659,
+      "epoch": 0.572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 8.73318419695963e-06,
-      "loss": 0.8525,
+      "learning_rate": 8.144173733945579e-06,
+      "loss": 0.8123,
       "step": 20180
     },
     {
-      "epoch": 0.5542555821043091,
+      "epoch": 0.5726730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 8.732301852403914e-06,
-      "loss": 0.9614,
+      "learning_rate": 8.143270629331678e-06,
+      "loss": 0.8035,
       "step": 20181
     },
     {
-      "epoch": 0.5542830463321524,
+      "epoch": 0.5727014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 8.731419517878718e-06,
-      "loss": 0.879,
+      "learning_rate": 8.142367540401652e-06,
+      "loss": 0.8622,
       "step": 20182
     },
     {
-      "epoch": 0.5543105105599956,
+      "epoch": 0.5727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 8.730537193391024e-06,
-      "loss": 0.8847,
+      "learning_rate": 8.141464467163134e-06,
+      "loss": 0.8101,
       "step": 20183
     },
     {
-      "epoch": 0.5543379747878389,
+      "epoch": 0.5727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 8.72965487894781e-06,
-      "loss": 0.9399,
+      "learning_rate": 8.14056140962375e-06,
+      "loss": 0.9348,
       "step": 20184
     },
     {
-      "epoch": 0.554365439015682,
+      "epoch": 0.5727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 8.728772574556062e-06,
-      "loss": 0.8203,
+      "learning_rate": 8.139658367791126e-06,
+      "loss": 0.8362,
       "step": 20185
     },
     {
-      "epoch": 0.5543929032435253,
+      "epoch": 0.5728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 8.727890280222752e-06,
-      "loss": 0.9219,
+      "learning_rate": 8.138755341672892e-06,
+      "loss": 0.796,
       "step": 20186
     },
     {
-      "epoch": 0.5544203674713686,
+      "epoch": 0.5728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 8.727007995954869e-06,
-      "loss": 0.8155,
+      "learning_rate": 8.137852331276677e-06,
+      "loss": 0.8872,
       "step": 20187
     },
     {
-      "epoch": 0.5544478316992117,
+      "epoch": 0.572871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 8.726125721759393e-06,
-      "loss": 0.8517,
+      "learning_rate": 8.136949336610103e-06,
+      "loss": 0.8356,
       "step": 20188
     },
     {
-      "epoch": 0.554475295927055,
+      "epoch": 0.572900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 8.725243457643302e-06,
-      "loss": 0.9034,
+      "learning_rate": 8.136046357680804e-06,
+      "loss": 1.0266,
       "step": 20189
     },
     {
-      "epoch": 0.5545027601548983,
+      "epoch": 0.5729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 8.724361203613584e-06,
-      "loss": 0.7645,
+      "learning_rate": 8.135143394496407e-06,
+      "loss": 0.8691,
       "step": 20190
     },
     {
-      "epoch": 0.5545302243827415,
+      "epoch": 0.5729568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 8.723478959677208e-06,
-      "loss": 0.8681,
+      "learning_rate": 8.134240447064534e-06,
+      "loss": 0.9032,
       "step": 20191
     },
     {
-      "epoch": 0.5545576886105847,
+      "epoch": 0.5729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 8.722596725841165e-06,
-      "loss": 0.9147,
+      "learning_rate": 8.133337515392818e-06,
+      "loss": 0.8373,
       "step": 20192
     },
     {
-      "epoch": 0.5545851528384279,
+      "epoch": 0.5730136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 8.721714502112429e-06,
-      "loss": 0.8948,
+      "learning_rate": 8.132434599488883e-06,
+      "loss": 0.8913,
       "step": 20193
     },
     {
-      "epoch": 0.5546126170662712,
+      "epoch": 0.5730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 8.720832288497982e-06,
-      "loss": 0.9712,
+      "learning_rate": 8.131531699360357e-06,
+      "loss": 0.796,
       "step": 20194
     },
     {
-      "epoch": 0.5546400812941145,
+      "epoch": 0.5730703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 8.719950085004807e-06,
-      "loss": 0.8972,
+      "learning_rate": 8.130628815014868e-06,
+      "loss": 0.8968,
       "step": 20195
     },
     {
-      "epoch": 0.5546675455219576,
+      "epoch": 0.5730987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 8.719067891639885e-06,
-      "loss": 0.8752,
+      "learning_rate": 8.129725946460039e-06,
+      "loss": 0.905,
       "step": 20196
     },
     {
-      "epoch": 0.5546950097498009,
+      "epoch": 0.5731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 8.718185708410193e-06,
-      "loss": 0.869,
+      "learning_rate": 8.1288230937035e-06,
+      "loss": 0.9098,
       "step": 20197
     },
     {
-      "epoch": 0.5547224739776441,
+      "epoch": 0.573155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 8.71730353532271e-06,
-      "loss": 0.9243,
+      "learning_rate": 8.127920256752873e-06,
+      "loss": 0.8363,
       "step": 20198
     },
     {
-      "epoch": 0.5547499382054873,
+      "epoch": 0.573183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 8.716421372384418e-06,
-      "loss": 0.8195,
+      "learning_rate": 8.127017435615792e-06,
+      "loss": 0.8013,
       "step": 20199
     },
     {
-      "epoch": 0.5547774024333306,
+      "epoch": 0.5732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 8.715539219602299e-06,
-      "loss": 0.7315,
+      "learning_rate": 8.12611463029988e-06,
+      "loss": 0.8103,
       "step": 20200
     },
     {
-      "epoch": 0.5548048666611738,
+      "epoch": 0.5732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 8.714657076983335e-06,
-      "loss": 0.8269,
+      "learning_rate": 8.125211840812759e-06,
+      "loss": 0.8649,
       "step": 20201
     },
     {
-      "epoch": 0.5548323308890171,
+      "epoch": 0.5732690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 8.713774944534496e-06,
-      "loss": 0.8615,
+      "learning_rate": 8.12430906716206e-06,
+      "loss": 0.9514,
       "step": 20202
     },
     {
-      "epoch": 0.5548597951168603,
+      "epoch": 0.5732973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 8.712892822262769e-06,
-      "loss": 0.918,
+      "learning_rate": 8.123406309355406e-06,
+      "loss": 0.7747,
       "step": 20203
     },
     {
-      "epoch": 0.5548872593447035,
+      "epoch": 0.5733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 8.712010710175133e-06,
-      "loss": 0.742,
+      "learning_rate": 8.122503567400423e-06,
+      "loss": 0.7734,
       "step": 20204
     },
     {
-      "epoch": 0.5549147235725468,
+      "epoch": 0.5733541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 8.711128608278568e-06,
-      "loss": 0.885,
+      "learning_rate": 8.121600841304738e-06,
+      "loss": 0.9877,
       "step": 20205
     },
     {
-      "epoch": 0.55494218780039,
+      "epoch": 0.5733825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 8.71024651658005e-06,
-      "loss": 0.9091,
+      "learning_rate": 8.120698131075975e-06,
+      "loss": 0.9026,
       "step": 20206
     },
     {
-      "epoch": 0.5549696520282332,
+      "epoch": 0.573410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 8.709364435086567e-06,
-      "loss": 0.8936,
+      "learning_rate": 8.119795436721757e-06,
+      "loss": 0.761,
       "step": 20207
     },
     {
-      "epoch": 0.5549971162560765,
+      "epoch": 0.5734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 8.708482363805088e-06,
-      "loss": 0.9269,
+      "learning_rate": 8.118892758249713e-06,
+      "loss": 0.8307,
       "step": 20208
     },
     {
-      "epoch": 0.5550245804839197,
+      "epoch": 0.5734676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 8.707600302742599e-06,
-      "loss": 0.8738,
+      "learning_rate": 8.117990095667467e-06,
+      "loss": 0.9376,
       "step": 20209
     },
     {
-      "epoch": 0.555052044711763,
+      "epoch": 0.5734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 8.706718251906076e-06,
-      "loss": 0.8845,
+      "learning_rate": 8.117087448982643e-06,
+      "loss": 0.7968,
       "step": 20210
     },
     {
-      "epoch": 0.5550795089396061,
+      "epoch": 0.5735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 8.705836211302498e-06,
-      "loss": 0.9292,
+      "learning_rate": 8.116184818202864e-06,
+      "loss": 0.9187,
       "step": 20211
     },
     {
-      "epoch": 0.5551069731674494,
+      "epoch": 0.5735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 8.704954180938848e-06,
-      "loss": 0.8262,
+      "learning_rate": 8.115282203335759e-06,
+      "loss": 0.9812,
       "step": 20212
     },
     {
-      "epoch": 0.5551344373952927,
+      "epoch": 0.5735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 8.704072160822102e-06,
-      "loss": 0.8502,
+      "learning_rate": 8.114379604388946e-06,
+      "loss": 0.9149,
       "step": 20213
     },
     {
-      "epoch": 0.5551619016231358,
+      "epoch": 0.5736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 8.703190150959236e-06,
-      "loss": 0.8311,
+      "learning_rate": 8.113477021370057e-06,
+      "loss": 0.8294,
       "step": 20214
     },
     {
-      "epoch": 0.5551893658509791,
+      "epoch": 0.5736379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 8.702308151357232e-06,
-      "loss": 0.9308,
+      "learning_rate": 8.11257445428671e-06,
+      "loss": 0.8097,
       "step": 20215
     },
     {
-      "epoch": 0.5552168300788224,
+      "epoch": 0.5736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 8.701426162023072e-06,
-      "loss": 0.8676,
+      "learning_rate": 8.111671903146533e-06,
+      "loss": 0.9123,
       "step": 20216
     },
     {
-      "epoch": 0.5552442943066656,
+      "epoch": 0.573694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 8.700544182963729e-06,
-      "loss": 0.9304,
+      "learning_rate": 8.11076936795715e-06,
+      "loss": 0.86,
       "step": 20217
     },
     {
-      "epoch": 0.5552717585345088,
+      "epoch": 0.5737230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 8.699662214186182e-06,
-      "loss": 0.957,
+      "learning_rate": 8.10986684872618e-06,
+      "loss": 0.8356,
       "step": 20218
     },
     {
-      "epoch": 0.555299222762352,
+      "epoch": 0.5737514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 8.698780255697416e-06,
-      "loss": 0.9916,
+      "learning_rate": 8.108964345461252e-06,
+      "loss": 0.8224,
       "step": 20219
     },
     {
-      "epoch": 0.5553266869901953,
+      "epoch": 0.5737797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 8.697898307504401e-06,
-      "loss": 0.8908,
+      "learning_rate": 8.108061858169988e-06,
+      "loss": 0.8672,
       "step": 20220
     },
     {
-      "epoch": 0.5553541512180385,
+      "epoch": 0.5738081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 8.697016369614121e-06,
-      "loss": 0.923,
+      "learning_rate": 8.107159386860007e-06,
+      "loss": 0.8489,
       "step": 20221
     },
     {
-      "epoch": 0.5553816154458817,
+      "epoch": 0.5738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 8.69613444203355e-06,
-      "loss": 0.8585,
+      "learning_rate": 8.106256931538938e-06,
+      "loss": 0.8668,
       "step": 20222
     },
     {
-      "epoch": 0.555409079673725,
+      "epoch": 0.5738649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 8.695252524769669e-06,
-      "loss": 0.9098,
+      "learning_rate": 8.105354492214403e-06,
+      "loss": 0.9054,
       "step": 20223
     },
     {
-      "epoch": 0.5554365439015682,
+      "epoch": 0.5738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 8.694370617829456e-06,
-      "loss": 0.9224,
+      "learning_rate": 8.104452068894022e-06,
+      "loss": 0.7875,
       "step": 20224
     },
     {
-      "epoch": 0.5554640081294114,
+      "epoch": 0.5739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 8.693488721219892e-06,
-      "loss": 0.7892,
+      "learning_rate": 8.10354966158542e-06,
+      "loss": 0.8937,
       "step": 20225
     },
     {
-      "epoch": 0.5554914723572547,
+      "epoch": 0.573950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 8.692606834947946e-06,
-      "loss": 0.8755,
+      "learning_rate": 8.10264727029622e-06,
+      "loss": 0.8986,
       "step": 20226
     },
     {
-      "epoch": 0.5555189365850979,
+      "epoch": 0.5739784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 8.6917249590206e-06,
-      "loss": 0.9316,
+      "learning_rate": 8.101744895034041e-06,
+      "loss": 0.8572,
       "step": 20227
     },
     {
-      "epoch": 0.5555464008129412,
+      "epoch": 0.5740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 8.690843093444838e-06,
-      "loss": 0.8078,
+      "learning_rate": 8.100842535806509e-06,
+      "loss": 0.8803,
       "step": 20228
     },
     {
-      "epoch": 0.5555738650407844,
+      "epoch": 0.5740351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 8.689961238227628e-06,
-      "loss": 0.8572,
+      "learning_rate": 8.099940192621245e-06,
+      "loss": 0.8245,
       "step": 20229
     },
     {
-      "epoch": 0.5556013292686276,
+      "epoch": 0.5740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 8.689079393375954e-06,
-      "loss": 0.8486,
+      "learning_rate": 8.099037865485873e-06,
+      "loss": 0.8925,
       "step": 20230
     },
     {
-      "epoch": 0.5556287934964709,
+      "epoch": 0.5740919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 8.688197558896795e-06,
-      "loss": 0.8709,
+      "learning_rate": 8.098135554408015e-06,
+      "loss": 0.8547,
       "step": 20231
     },
     {
-      "epoch": 0.555656257724314,
+      "epoch": 0.5741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 8.687315734797121e-06,
-      "loss": 0.9302,
+      "learning_rate": 8.09723325939529e-06,
+      "loss": 0.9179,
       "step": 20232
     },
     {
-      "epoch": 0.5556837219521573,
+      "epoch": 0.5741486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 8.686433921083917e-06,
-      "loss": 0.9104,
+      "learning_rate": 8.096330980455322e-06,
+      "loss": 0.916,
       "step": 20233
     },
     {
-      "epoch": 0.5557111861800006,
+      "epoch": 0.5741770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 8.685552117764152e-06,
-      "loss": 0.9198,
+      "learning_rate": 8.095428717595732e-06,
+      "loss": 0.7593,
       "step": 20234
     },
     {
-      "epoch": 0.5557386504078438,
+      "epoch": 0.574205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 8.684670324844809e-06,
-      "loss": 0.8953,
+      "learning_rate": 8.09452647082414e-06,
+      "loss": 0.8735,
       "step": 20235
     },
     {
-      "epoch": 0.555766114635687,
+      "epoch": 0.5742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 8.683788542332865e-06,
-      "loss": 0.8524,
+      "learning_rate": 8.093624240148168e-06,
+      "loss": 0.9258,
       "step": 20236
     },
     {
-      "epoch": 0.5557935788635302,
+      "epoch": 0.5742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 8.6829067702353e-06,
-      "loss": 0.9534,
+      "learning_rate": 8.09272202557544e-06,
+      "loss": 0.8757,
       "step": 20237
     },
     {
-      "epoch": 0.5558210430913735,
+      "epoch": 0.5742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 8.68202500855908e-06,
-      "loss": 0.9181,
+      "learning_rate": 8.091819827113571e-06,
+      "loss": 0.8333,
       "step": 20238
     },
     {
-      "epoch": 0.5558485073192168,
+      "epoch": 0.5743189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 8.681143257311191e-06,
-      "loss": 0.9103,
+      "learning_rate": 8.090917644770191e-06,
+      "loss": 0.8902,
       "step": 20239
     },
     {
-      "epoch": 0.5558759715470599,
+      "epoch": 0.5743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 8.68026151649861e-06,
-      "loss": 0.8743,
+      "learning_rate": 8.090015478552914e-06,
+      "loss": 0.7787,
       "step": 20240
     },
     {
-      "epoch": 0.5559034357749032,
+      "epoch": 0.5743757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 8.679379786128306e-06,
-      "loss": 0.9658,
+      "learning_rate": 8.089113328469357e-06,
+      "loss": 0.9034,
       "step": 20241
     },
     {
-      "epoch": 0.5559309000027465,
+      "epoch": 0.5744040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 8.678498066207263e-06,
-      "loss": 0.798,
+      "learning_rate": 8.08821119452715e-06,
+      "loss": 0.955,
       "step": 20242
     },
     {
-      "epoch": 0.5559583642305896,
+      "epoch": 0.5744324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 8.67761635674246e-06,
-      "loss": 0.9757,
+      "learning_rate": 8.087309076733912e-06,
+      "loss": 0.9628,
       "step": 20243
     },
     {
-      "epoch": 0.5559858284584329,
+      "epoch": 0.5744608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 8.676734657740864e-06,
-      "loss": 0.8261,
+      "learning_rate": 8.086406975097252e-06,
+      "loss": 0.8324,
       "step": 20244
     },
     {
-      "epoch": 0.5560132926862761,
+      "epoch": 0.5744892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 8.675852969209455e-06,
-      "loss": 0.904,
+      "learning_rate": 8.085504889624803e-06,
+      "loss": 0.8791,
       "step": 20245
     },
     {
-      "epoch": 0.5560407569141194,
+      "epoch": 0.5745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 8.67497129115521e-06,
-      "loss": 0.8697,
+      "learning_rate": 8.08460282032418e-06,
+      "loss": 0.8959,
       "step": 20246
     },
     {
-      "epoch": 0.5560682211419626,
+      "epoch": 0.5745459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 8.674089623585106e-06,
-      "loss": 0.8874,
+      "learning_rate": 8.083700767203004e-06,
+      "loss": 0.7561,
       "step": 20247
     },
     {
-      "epoch": 0.5560956853698058,
+      "epoch": 0.5745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 8.67320796650612e-06,
-      "loss": 0.7731,
+      "learning_rate": 8.082798730268894e-06,
+      "loss": 0.7995,
       "step": 20248
     },
     {
-      "epoch": 0.5561231495976491,
+      "epoch": 0.5746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 8.67232631992523e-06,
-      "loss": 0.8783,
+      "learning_rate": 8.081896709529466e-06,
+      "loss": 0.8593,
       "step": 20249
     },
     {
-      "epoch": 0.5561506138254922,
+      "epoch": 0.5746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 8.671444683849402e-06,
-      "loss": 0.8926,
+      "learning_rate": 8.080994704992347e-06,
+      "loss": 0.782,
       "step": 20250
     },
     {
-      "epoch": 0.5561780780533355,
+      "epoch": 0.5746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 8.67056305828562e-06,
-      "loss": 0.9379,
+      "learning_rate": 8.080092716665151e-06,
+      "loss": 0.9934,
       "step": 20251
     },
     {
-      "epoch": 0.5562055422811788,
+      "epoch": 0.5746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 8.66968144324086e-06,
-      "loss": 0.8701,
+      "learning_rate": 8.079190744555494e-06,
+      "loss": 0.8324,
       "step": 20252
     },
     {
-      "epoch": 0.556233006509022,
+      "epoch": 0.574716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 8.668799838722093e-06,
-      "loss": 0.9338,
+      "learning_rate": 8.078288788671004e-06,
+      "loss": 0.9148,
       "step": 20253
     },
     {
-      "epoch": 0.5562604707368652,
+      "epoch": 0.574744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 8.667918244736297e-06,
-      "loss": 0.9798,
+      "learning_rate": 8.077386849019295e-06,
+      "loss": 0.895,
       "step": 20254
     },
     {
-      "epoch": 0.5562879349647085,
+      "epoch": 0.5747729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 8.667036661290454e-06,
-      "loss": 0.8625,
+      "learning_rate": 8.076484925607983e-06,
+      "loss": 0.9509,
       "step": 20255
     },
     {
-      "epoch": 0.5563153991925517,
+      "epoch": 0.5748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 8.66615508839153e-06,
-      "loss": 0.7834,
+      "learning_rate": 8.075583018444691e-06,
+      "loss": 0.8869,
       "step": 20256
     },
     {
-      "epoch": 0.556342863420395,
+      "epoch": 0.5748297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 8.665273526046501e-06,
-      "loss": 0.886,
+      "learning_rate": 8.074681127537035e-06,
+      "loss": 0.8802,
       "step": 20257
     },
     {
-      "epoch": 0.5563703276482381,
+      "epoch": 0.5748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 8.664391974262344e-06,
-      "loss": 0.8835,
+      "learning_rate": 8.073779252892633e-06,
+      "loss": 0.9035,
       "step": 20258
     },
     {
-      "epoch": 0.5563977918760814,
+      "epoch": 0.5748864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 8.663510433046038e-06,
-      "loss": 0.9394,
+      "learning_rate": 8.072877394519103e-06,
+      "loss": 0.8394,
       "step": 20259
     },
     {
-      "epoch": 0.5564252561039247,
+      "epoch": 0.5749148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 8.662628902404557e-06,
-      "loss": 0.9421,
+      "learning_rate": 8.071975552424066e-06,
+      "loss": 0.7764,
       "step": 20260
     },
     {
-      "epoch": 0.5564527203317678,
+      "epoch": 0.5749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 8.661747382344874e-06,
-      "loss": 0.8137,
+      "learning_rate": 8.07107372661514e-06,
+      "loss": 0.7901,
       "step": 20261
     },
     {
-      "epoch": 0.5564801845596111,
+      "epoch": 0.5749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 8.66086587287396e-06,
-      "loss": 0.827,
+      "learning_rate": 8.070171917099938e-06,
+      "loss": 0.9426,
       "step": 20262
     },
     {
-      "epoch": 0.5565076487874543,
+      "epoch": 0.575,
       "grad_norm": 0.0,
-      "learning_rate": 8.659984373998795e-06,
-      "loss": 0.8574,
+      "learning_rate": 8.069270123886084e-06,
+      "loss": 0.9412,
       "step": 20263
     },
     {
-      "epoch": 0.5565351130152976,
+      "epoch": 0.5750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 8.659102885726354e-06,
-      "loss": 0.8172,
+      "learning_rate": 8.068368346981192e-06,
+      "loss": 0.8607,
       "step": 20264
     },
     {
-      "epoch": 0.5565625772431408,
+      "epoch": 0.5750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 8.65822140806361e-06,
-      "loss": 0.8662,
+      "learning_rate": 8.067466586392875e-06,
+      "loss": 0.9323,
       "step": 20265
     },
     {
-      "epoch": 0.556590041470984,
+      "epoch": 0.5750851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 8.65733994101754e-06,
-      "loss": 0.972,
+      "learning_rate": 8.066564842128758e-06,
+      "loss": 0.8704,
       "step": 20266
     },
     {
-      "epoch": 0.5566175056988273,
+      "epoch": 0.5751135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 8.656458484595114e-06,
-      "loss": 0.8948,
+      "learning_rate": 8.065663114196455e-06,
+      "loss": 0.8752,
       "step": 20267
     },
     {
-      "epoch": 0.5566449699266706,
+      "epoch": 0.5751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 8.65557703880331e-06,
-      "loss": 0.9507,
+      "learning_rate": 8.064761402603582e-06,
+      "loss": 0.8701,
       "step": 20268
     },
     {
-      "epoch": 0.5566724341545137,
+      "epoch": 0.5751702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 8.654695603649099e-06,
-      "loss": 0.872,
+      "learning_rate": 8.063859707357757e-06,
+      "loss": 0.8613,
       "step": 20269
     },
     {
-      "epoch": 0.556699898382357,
+      "epoch": 0.5751986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 8.653814179139459e-06,
-      "loss": 0.9381,
+      "learning_rate": 8.062958028466596e-06,
+      "loss": 0.911,
       "step": 20270
     },
     {
-      "epoch": 0.5567273626102002,
+      "epoch": 0.5752270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 8.652932765281361e-06,
-      "loss": 0.8834,
+      "learning_rate": 8.062056365937713e-06,
+      "loss": 0.8755,
       "step": 20271
     },
     {
-      "epoch": 0.5567548268380434,
+      "epoch": 0.575255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 8.652051362081786e-06,
-      "loss": 0.8755,
+      "learning_rate": 8.06115471977873e-06,
+      "loss": 0.9802,
       "step": 20272
     },
     {
-      "epoch": 0.5567822910658867,
+      "epoch": 0.5752837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 8.651169969547701e-06,
-      "loss": 0.8819,
+      "learning_rate": 8.06025308999726e-06,
+      "loss": 0.9394,
       "step": 20273
     },
     {
-      "epoch": 0.5568097552937299,
+      "epoch": 0.5753121452894439,
       "grad_norm": 0.0,
-      "learning_rate": 8.650288587686078e-06,
-      "loss": 0.9381,
+      "learning_rate": 8.059351476600913e-06,
+      "loss": 0.9458,
       "step": 20274
     },
     {
-      "epoch": 0.5568372195215732,
+      "epoch": 0.5753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 8.649407216503896e-06,
-      "loss": 1.0286,
+      "learning_rate": 8.058449879597318e-06,
+      "loss": 0.8622,
       "step": 20275
     },
     {
-      "epoch": 0.5568646837494163,
+      "epoch": 0.5753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 8.648525856008127e-06,
-      "loss": 0.8443,
+      "learning_rate": 8.057548298994082e-06,
+      "loss": 0.8866,
       "step": 20276
     },
     {
-      "epoch": 0.5568921479772596,
+      "epoch": 0.5753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 8.647644506205746e-06,
-      "loss": 0.8744,
+      "learning_rate": 8.056646734798824e-06,
+      "loss": 0.8331,
       "step": 20277
     },
     {
-      "epoch": 0.5569196122051029,
+      "epoch": 0.5754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 8.646763167103728e-06,
-      "loss": 0.874,
+      "learning_rate": 8.05574518701916e-06,
+      "loss": 0.9035,
       "step": 20278
     },
     {
-      "epoch": 0.556947076432946,
+      "epoch": 0.5754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 8.64588183870904e-06,
-      "loss": 0.8896,
+      "learning_rate": 8.0548436556627e-06,
+      "loss": 0.8967,
       "step": 20279
     },
     {
-      "epoch": 0.5569745406607893,
+      "epoch": 0.5754824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 8.645000521028661e-06,
-      "loss": 1.0294,
+      "learning_rate": 8.053942140737066e-06,
+      "loss": 0.7774,
       "step": 20280
     },
     {
-      "epoch": 0.5570020048886326,
+      "epoch": 0.5755107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 8.644119214069562e-06,
-      "loss": 0.8921,
+      "learning_rate": 8.053040642249871e-06,
+      "loss": 0.8401,
       "step": 20281
     },
     {
-      "epoch": 0.5570294691164758,
+      "epoch": 0.575539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 8.643237917838717e-06,
-      "loss": 0.8862,
+      "learning_rate": 8.052139160208725e-06,
+      "loss": 0.7717,
       "step": 20282
     },
     {
-      "epoch": 0.557056933344319,
+      "epoch": 0.5755675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 8.642356632343099e-06,
-      "loss": 0.8611,
+      "learning_rate": 8.051237694621252e-06,
+      "loss": 0.9108,
       "step": 20283
     },
     {
-      "epoch": 0.5570843975721622,
+      "epoch": 0.5755959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 8.641475357589684e-06,
-      "loss": 0.7673,
+      "learning_rate": 8.05033624549506e-06,
+      "loss": 0.8416,
       "step": 20284
     },
     {
-      "epoch": 0.5571118618000055,
+      "epoch": 0.5756242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 8.640594093585442e-06,
-      "loss": 0.8744,
+      "learning_rate": 8.049434812837762e-06,
+      "loss": 0.9293,
       "step": 20285
     },
     {
-      "epoch": 0.5571393260278488,
+      "epoch": 0.5756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 8.639712840337342e-06,
-      "loss": 0.8691,
+      "learning_rate": 8.04853339665698e-06,
+      "loss": 0.8487,
       "step": 20286
     },
     {
-      "epoch": 0.5571667902556919,
+      "epoch": 0.5756810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 8.638831597852362e-06,
-      "loss": 0.8235,
+      "learning_rate": 8.047631996960324e-06,
+      "loss": 0.9078,
       "step": 20287
     },
     {
-      "epoch": 0.5571942544835352,
+      "epoch": 0.5757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 8.637950366137475e-06,
-      "loss": 0.9266,
+      "learning_rate": 8.046730613755404e-06,
+      "loss": 0.8046,
       "step": 20288
     },
     {
-      "epoch": 0.5572217187113784,
+      "epoch": 0.5757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 8.63706914519965e-06,
-      "loss": 0.8866,
+      "learning_rate": 8.045829247049844e-06,
+      "loss": 0.7823,
       "step": 20289
     },
     {
-      "epoch": 0.5572491829392217,
+      "epoch": 0.575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 8.636187935045869e-06,
-      "loss": 0.8878,
+      "learning_rate": 8.044927896851245e-06,
+      "loss": 0.7875,
       "step": 20290
     },
     {
-      "epoch": 0.5572766471670649,
+      "epoch": 0.575794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 8.63530673568309e-06,
-      "loss": 0.8968,
+      "learning_rate": 8.044026563167233e-06,
+      "loss": 0.8126,
       "step": 20291
     },
     {
-      "epoch": 0.5573041113949081,
+      "epoch": 0.5758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 8.634425547118298e-06,
-      "loss": 0.908,
+      "learning_rate": 8.043125246005416e-06,
+      "loss": 0.905,
       "step": 20292
     },
     {
-      "epoch": 0.5573315756227514,
+      "epoch": 0.5758513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 8.633544369358454e-06,
-      "loss": 0.8468,
+      "learning_rate": 8.042223945373407e-06,
+      "loss": 0.8305,
       "step": 20293
     },
     {
-      "epoch": 0.5573590398505947,
+      "epoch": 0.5758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 8.632663202410539e-06,
-      "loss": 0.8745,
+      "learning_rate": 8.041322661278822e-06,
+      "loss": 0.8751,
       "step": 20294
     },
     {
-      "epoch": 0.5573865040784378,
+      "epoch": 0.5759080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 8.631782046281523e-06,
-      "loss": 0.8732,
+      "learning_rate": 8.040421393729273e-06,
+      "loss": 0.9402,
       "step": 20295
     },
     {
-      "epoch": 0.5574139683062811,
+      "epoch": 0.5759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 8.630900900978379e-06,
-      "loss": 0.9025,
+      "learning_rate": 8.039520142732371e-06,
+      "loss": 0.7424,
       "step": 20296
     },
     {
-      "epoch": 0.5574414325341243,
+      "epoch": 0.5759648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 8.630019766508077e-06,
-      "loss": 0.7732,
+      "learning_rate": 8.03861890829573e-06,
+      "loss": 0.8575,
       "step": 20297
     },
     {
-      "epoch": 0.5574688967619675,
+      "epoch": 0.5759931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 8.629138642877586e-06,
-      "loss": 0.8692,
+      "learning_rate": 8.037717690426966e-06,
+      "loss": 0.951,
       "step": 20298
     },
     {
-      "epoch": 0.5574963609898108,
+      "epoch": 0.5760215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 8.628257530093882e-06,
-      "loss": 0.8645,
+      "learning_rate": 8.036816489133685e-06,
+      "loss": 0.8846,
       "step": 20299
     },
     {
-      "epoch": 0.557523825217654,
+      "epoch": 0.576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 8.627376428163938e-06,
-      "loss": 0.8655,
+      "learning_rate": 8.035915304423507e-06,
+      "loss": 0.8192,
       "step": 20300
     },
     {
-      "epoch": 0.5575512894454973,
+      "epoch": 0.576078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 8.626495337094721e-06,
-      "loss": 0.9089,
+      "learning_rate": 8.03501413630404e-06,
+      "loss": 0.8606,
       "step": 20301
     },
     {
-      "epoch": 0.5575787536733404,
+      "epoch": 0.5761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 8.625614256893208e-06,
-      "loss": 0.8522,
+      "learning_rate": 8.034112984782894e-06,
+      "loss": 0.9625,
       "step": 20302
     },
     {
-      "epoch": 0.5576062179011837,
+      "epoch": 0.5761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 8.624733187566365e-06,
-      "loss": 0.7632,
+      "learning_rate": 8.033211849867688e-06,
+      "loss": 0.8557,
       "step": 20303
     },
     {
-      "epoch": 0.557633682129027,
+      "epoch": 0.5761634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 8.623852129121168e-06,
-      "loss": 0.8147,
+      "learning_rate": 8.032310731566029e-06,
+      "loss": 0.8281,
       "step": 20304
     },
     {
-      "epoch": 0.5576611463568701,
+      "epoch": 0.5761918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 8.622971081564581e-06,
-      "loss": 0.9971,
+      "learning_rate": 8.031409629885525e-06,
+      "loss": 0.8009,
       "step": 20305
     },
     {
-      "epoch": 0.5576886105847134,
+      "epoch": 0.5762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 8.622090044903584e-06,
-      "loss": 0.8591,
+      "learning_rate": 8.030508544833793e-06,
+      "loss": 0.8048,
       "step": 20306
     },
     {
-      "epoch": 0.5577160748125567,
+      "epoch": 0.5762485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 8.621209019145145e-06,
-      "loss": 0.9422,
+      "learning_rate": 8.029607476418448e-06,
+      "loss": 0.8867,
       "step": 20307
     },
     {
-      "epoch": 0.5577435390403999,
+      "epoch": 0.5762769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 8.620328004296233e-06,
-      "loss": 0.8517,
+      "learning_rate": 8.028706424647098e-06,
+      "loss": 0.8692,
       "step": 20308
     },
     {
-      "epoch": 0.5577710032682431,
+      "epoch": 0.576305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 8.61944700036382e-06,
-      "loss": 0.7397,
+      "learning_rate": 8.02780538952735e-06,
+      "loss": 0.8413,
       "step": 20309
     },
     {
-      "epoch": 0.5577984674960863,
+      "epoch": 0.5763337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 8.618566007354878e-06,
-      "loss": 0.9028,
+      "learning_rate": 8.02690437106682e-06,
+      "loss": 0.8692,
       "step": 20310
     },
     {
-      "epoch": 0.5578259317239296,
+      "epoch": 0.5763620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 8.617685025276374e-06,
-      "loss": 0.8305,
+      "learning_rate": 8.02600336927312e-06,
+      "loss": 0.9427,
       "step": 20311
     },
     {
-      "epoch": 0.5578533959517729,
+      "epoch": 0.5763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 8.616804054135286e-06,
-      "loss": 0.8374,
+      "learning_rate": 8.025102384153854e-06,
+      "loss": 0.8345,
       "step": 20312
     },
     {
-      "epoch": 0.557880860179616,
+      "epoch": 0.5764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 8.615923093938576e-06,
-      "loss": 0.8562,
+      "learning_rate": 8.02420141571664e-06,
+      "loss": 0.8928,
       "step": 20313
     },
     {
-      "epoch": 0.5579083244074593,
+      "epoch": 0.5764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 8.615042144693224e-06,
-      "loss": 0.9744,
+      "learning_rate": 8.023300463969083e-06,
+      "loss": 0.8824,
       "step": 20314
     },
     {
-      "epoch": 0.5579357886353025,
+      "epoch": 0.5764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 8.614161206406191e-06,
-      "loss": 0.8712,
+      "learning_rate": 8.022399528918799e-06,
+      "loss": 0.8979,
       "step": 20315
     },
     {
-      "epoch": 0.5579632528631457,
+      "epoch": 0.5765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 8.613280279084452e-06,
-      "loss": 0.8709,
+      "learning_rate": 8.021498610573392e-06,
+      "loss": 0.7806,
       "step": 20316
     },
     {
-      "epoch": 0.557990717090989,
+      "epoch": 0.5765323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 8.612399362734976e-06,
-      "loss": 0.7549,
+      "learning_rate": 8.020597708940477e-06,
+      "loss": 0.86,
       "step": 20317
     },
     {
-      "epoch": 0.5580181813188322,
+      "epoch": 0.5765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 8.611518457364733e-06,
-      "loss": 0.8632,
+      "learning_rate": 8.019696824027662e-06,
+      "loss": 0.779,
       "step": 20318
     },
     {
-      "epoch": 0.5580456455466755,
+      "epoch": 0.576589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 8.610637562980697e-06,
-      "loss": 0.8408,
+      "learning_rate": 8.018795955842556e-06,
+      "loss": 0.7599,
       "step": 20319
     },
     {
-      "epoch": 0.5580731097745187,
+      "epoch": 0.5766174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 8.609756679589834e-06,
-      "loss": 0.8976,
+      "learning_rate": 8.017895104392765e-06,
+      "loss": 0.9,
       "step": 20320
     },
     {
-      "epoch": 0.5581005740023619,
+      "epoch": 0.5766458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 8.608875807199114e-06,
-      "loss": 0.9225,
+      "learning_rate": 8.016994269685909e-06,
+      "loss": 0.8196,
       "step": 20321
     },
     {
-      "epoch": 0.5581280382302052,
+      "epoch": 0.5766742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 8.607994945815505e-06,
-      "loss": 0.8784,
+      "learning_rate": 8.016093451729591e-06,
+      "loss": 0.8704,
       "step": 20322
     },
     {
-      "epoch": 0.5581555024580483,
+      "epoch": 0.5767026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 8.60711409544598e-06,
-      "loss": 0.9296,
+      "learning_rate": 8.015192650531421e-06,
+      "loss": 0.7377,
       "step": 20323
     },
     {
-      "epoch": 0.5581829666858916,
+      "epoch": 0.5767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 8.60623325609751e-06,
-      "loss": 0.8205,
+      "learning_rate": 8.014291866099008e-06,
+      "loss": 0.8941,
       "step": 20324
     },
     {
-      "epoch": 0.5582104309137349,
+      "epoch": 0.5767593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 8.605352427777066e-06,
-      "loss": 0.8085,
+      "learning_rate": 8.01339109843996e-06,
+      "loss": 0.8274,
       "step": 20325
     },
     {
-      "epoch": 0.5582378951415781,
+      "epoch": 0.5767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 8.604471610491608e-06,
-      "loss": 0.8861,
+      "learning_rate": 8.012490347561887e-06,
+      "loss": 0.7896,
       "step": 20326
     },
     {
-      "epoch": 0.5582653593694213,
+      "epoch": 0.5768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 8.603590804248111e-06,
-      "loss": 0.8445,
+      "learning_rate": 8.011589613472397e-06,
+      "loss": 0.8056,
       "step": 20327
     },
     {
-      "epoch": 0.5582928235972645,
+      "epoch": 0.5768444948921679,
       "grad_norm": 0.0,
-      "learning_rate": 8.602710009053548e-06,
-      "loss": 0.937,
+      "learning_rate": 8.010688896179101e-06,
+      "loss": 0.9501,
       "step": 20328
     },
     {
-      "epoch": 0.5583202878251078,
+      "epoch": 0.5768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 8.601829224914882e-06,
-      "loss": 0.9193,
+      "learning_rate": 8.009788195689601e-06,
+      "loss": 0.8667,
       "step": 20329
     },
     {
-      "epoch": 0.5583477520529511,
+      "epoch": 0.5769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 8.600948451839084e-06,
-      "loss": 0.8712,
+      "learning_rate": 8.008887512011515e-06,
+      "loss": 0.8654,
       "step": 20330
     },
     {
-      "epoch": 0.5583752162807942,
+      "epoch": 0.5769296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 8.600067689833128e-06,
-      "loss": 0.853,
+      "learning_rate": 8.007986845152443e-06,
+      "loss": 0.8423,
       "step": 20331
     },
     {
-      "epoch": 0.5584026805086375,
+      "epoch": 0.5769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 8.599186938903974e-06,
-      "loss": 0.9659,
+      "learning_rate": 8.007086195119994e-06,
+      "loss": 0.8337,
       "step": 20332
     },
     {
-      "epoch": 0.5584301447364808,
+      "epoch": 0.5769863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 8.598306199058599e-06,
-      "loss": 0.8854,
+      "learning_rate": 8.00618556192178e-06,
+      "loss": 0.7859,
       "step": 20333
     },
     {
-      "epoch": 0.558457608964324,
+      "epoch": 0.5770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 8.597425470303967e-06,
-      "loss": 1.037,
+      "learning_rate": 8.005284945565406e-06,
+      "loss": 0.9524,
       "step": 20334
     },
     {
-      "epoch": 0.5584850731921672,
+      "epoch": 0.5770431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 8.596544752647045e-06,
-      "loss": 0.7976,
+      "learning_rate": 8.004384346058477e-06,
+      "loss": 0.855,
       "step": 20335
     },
     {
-      "epoch": 0.5585125374200104,
+      "epoch": 0.5770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 8.59566404609481e-06,
-      "loss": 0.8817,
+      "learning_rate": 8.003483763408604e-06,
+      "loss": 0.8179,
       "step": 20336
     },
     {
-      "epoch": 0.5585400016478537,
+      "epoch": 0.577099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 8.594783350654225e-06,
-      "loss": 0.913,
+      "learning_rate": 8.002583197623392e-06,
+      "loss": 0.8402,
       "step": 20337
     },
     {
-      "epoch": 0.558567465875697,
+      "epoch": 0.5771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 8.593902666332254e-06,
-      "loss": 0.8375,
+      "learning_rate": 8.001682648710455e-06,
+      "loss": 0.8695,
       "step": 20338
     },
     {
-      "epoch": 0.5585949301035401,
+      "epoch": 0.5771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 8.593021993135871e-06,
-      "loss": 0.8967,
+      "learning_rate": 8.000782116677391e-06,
+      "loss": 0.8704,
       "step": 20339
     },
     {
-      "epoch": 0.5586223943313834,
+      "epoch": 0.5771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 8.592141331072044e-06,
-      "loss": 0.8222,
+      "learning_rate": 7.999881601531809e-06,
+      "loss": 0.8781,
       "step": 20340
     },
     {
-      "epoch": 0.5586498585592266,
+      "epoch": 0.5772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 8.59126068014774e-06,
-      "loss": 0.9261,
+      "learning_rate": 7.99898110328132e-06,
+      "loss": 0.8179,
       "step": 20341
     },
     {
-      "epoch": 0.5586773227870698,
+      "epoch": 0.5772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 8.590380040369924e-06,
-      "loss": 0.8374,
+      "learning_rate": 7.998080621933529e-06,
+      "loss": 0.8676,
       "step": 20342
     },
     {
-      "epoch": 0.5587047870149131,
+      "epoch": 0.5772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 8.589499411745574e-06,
-      "loss": 0.8141,
+      "learning_rate": 7.997180157496036e-06,
+      "loss": 0.8966,
       "step": 20343
     },
     {
-      "epoch": 0.5587322512427563,
+      "epoch": 0.5772985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 8.588618794281647e-06,
-      "loss": 0.919,
+      "learning_rate": 7.996279709976456e-06,
+      "loss": 0.8316,
       "step": 20344
     },
     {
-      "epoch": 0.5587597154705995,
+      "epoch": 0.5773269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 8.587738187985115e-06,
-      "loss": 0.9476,
+      "learning_rate": 7.99537927938239e-06,
+      "loss": 0.8874,
       "step": 20345
     },
     {
-      "epoch": 0.5587871796984428,
+      "epoch": 0.5773552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 8.586857592862946e-06,
-      "loss": 0.9334,
+      "learning_rate": 7.994478865721444e-06,
+      "loss": 0.9379,
       "step": 20346
     },
     {
-      "epoch": 0.558814643926286,
+      "epoch": 0.5773836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 8.585977008922104e-06,
-      "loss": 0.9194,
+      "learning_rate": 7.993578469001227e-06,
+      "loss": 0.8267,
       "step": 20347
     },
     {
-      "epoch": 0.5588421081541293,
+      "epoch": 0.5774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 8.585096436169563e-06,
-      "loss": 0.8764,
+      "learning_rate": 7.992678089229345e-06,
+      "loss": 0.8527,
       "step": 20348
     },
     {
-      "epoch": 0.5588695723819724,
+      "epoch": 0.5774404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 8.58421587461229e-06,
-      "loss": 0.9312,
+      "learning_rate": 7.991777726413396e-06,
+      "loss": 0.866,
       "step": 20349
     },
     {
-      "epoch": 0.5588970366098157,
+      "epoch": 0.5774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 8.583335324257245e-06,
-      "loss": 0.9256,
+      "learning_rate": 7.990877380560996e-06,
+      "loss": 0.8638,
       "step": 20350
     },
     {
-      "epoch": 0.558924500837659,
+      "epoch": 0.5774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 8.582454785111399e-06,
-      "loss": 0.8355,
+      "learning_rate": 7.989977051679738e-06,
+      "loss": 0.876,
       "step": 20351
     },
     {
-      "epoch": 0.5589519650655022,
+      "epoch": 0.5775255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 8.581574257181722e-06,
-      "loss": 0.808,
+      "learning_rate": 7.98907673977724e-06,
+      "loss": 0.9142,
       "step": 20352
     },
     {
-      "epoch": 0.5589794292933454,
+      "epoch": 0.5775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 8.580693740475177e-06,
-      "loss": 0.8944,
+      "learning_rate": 7.988176444861098e-06,
+      "loss": 0.8407,
       "step": 20353
     },
     {
-      "epoch": 0.5590068935211886,
+      "epoch": 0.5775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 8.579813234998733e-06,
-      "loss": 0.8729,
+      "learning_rate": 7.987276166938923e-06,
+      "loss": 0.9324,
       "step": 20354
     },
     {
-      "epoch": 0.5590343577490319,
+      "epoch": 0.57761066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.57893274075936e-06,
-      "loss": 0.9368,
+      "learning_rate": 7.986375906018316e-06,
+      "loss": 0.8806,
       "step": 20355
     },
     {
-      "epoch": 0.5590618219768752,
+      "epoch": 0.577639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 8.578052257764019e-06,
-      "loss": 0.8817,
+      "learning_rate": 7.985475662106885e-06,
+      "loss": 0.8498,
       "step": 20356
     },
     {
-      "epoch": 0.5590892862047183,
+      "epoch": 0.5776674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 8.57717178601968e-06,
-      "loss": 0.8857,
+      "learning_rate": 7.984575435212226e-06,
+      "loss": 0.8871,
       "step": 20357
     },
     {
-      "epoch": 0.5591167504325616,
+      "epoch": 0.5776958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 8.57629132553331e-06,
-      "loss": 0.8535,
+      "learning_rate": 7.983675225341953e-06,
+      "loss": 0.7791,
       "step": 20358
     },
     {
-      "epoch": 0.5591442146604049,
+      "epoch": 0.5777241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 8.575410876311871e-06,
-      "loss": 0.7841,
+      "learning_rate": 7.982775032503665e-06,
+      "loss": 0.8077,
       "step": 20359
     },
     {
-      "epoch": 0.559171678888248,
+      "epoch": 0.5777525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 8.574530438362336e-06,
-      "loss": 0.9105,
+      "learning_rate": 7.981874856704965e-06,
+      "loss": 0.8324,
       "step": 20360
     },
     {
-      "epoch": 0.5591991431160913,
+      "epoch": 0.5777809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 8.573650011691672e-06,
-      "loss": 0.8382,
+      "learning_rate": 7.98097469795346e-06,
+      "loss": 0.7534,
       "step": 20361
     },
     {
-      "epoch": 0.5592266073439345,
+      "epoch": 0.5778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 8.572769596306837e-06,
-      "loss": 0.8707,
+      "learning_rate": 7.980074556256756e-06,
+      "loss": 0.8987,
       "step": 20362
     },
     {
-      "epoch": 0.5592540715717778,
+      "epoch": 0.5778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 8.5718891922148e-06,
-      "loss": 0.9506,
+      "learning_rate": 7.979174431622447e-06,
+      "loss": 0.9591,
       "step": 20363
     },
     {
-      "epoch": 0.559281535799621,
+      "epoch": 0.5778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 8.571008799422531e-06,
-      "loss": 0.8991,
+      "learning_rate": 7.978274324058146e-06,
+      "loss": 0.9201,
       "step": 20364
     },
     {
-      "epoch": 0.5593090000274642,
+      "epoch": 0.577894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.570128417936994e-06,
-      "loss": 0.8747,
+      "learning_rate": 7.977374233571453e-06,
+      "loss": 0.7973,
       "step": 20365
     },
     {
-      "epoch": 0.5593364642553075,
+      "epoch": 0.5779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 8.569248047765152e-06,
-      "loss": 0.9371,
+      "learning_rate": 7.976474160169967e-06,
+      "loss": 0.8241,
       "step": 20366
     },
     {
-      "epoch": 0.5593639284831506,
+      "epoch": 0.5779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 8.56836768891398e-06,
-      "loss": 0.8043,
+      "learning_rate": 7.975574103861295e-06,
+      "loss": 0.8762,
       "step": 20367
     },
     {
-      "epoch": 0.5593913927109939,
+      "epoch": 0.5779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 8.567487341390434e-06,
-      "loss": 0.9298,
+      "learning_rate": 7.974674064653045e-06,
+      "loss": 0.7628,
       "step": 20368
     },
     {
-      "epoch": 0.5594188569388372,
+      "epoch": 0.5780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 8.566607005201483e-06,
-      "loss": 0.8737,
+      "learning_rate": 7.973774042552811e-06,
+      "loss": 0.7731,
       "step": 20369
     },
     {
-      "epoch": 0.5594463211666804,
+      "epoch": 0.5780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 8.565726680354091e-06,
-      "loss": 0.9115,
+      "learning_rate": 7.972874037568197e-06,
+      "loss": 0.7144,
       "step": 20370
     },
     {
-      "epoch": 0.5594737853945236,
+      "epoch": 0.5780646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 8.564846366855227e-06,
-      "loss": 0.8946,
+      "learning_rate": 7.97197404970681e-06,
+      "loss": 0.8449,
       "step": 20371
     },
     {
-      "epoch": 0.5595012496223669,
+      "epoch": 0.5780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 8.563966064711855e-06,
-      "loss": 0.8847,
+      "learning_rate": 7.97107407897625e-06,
+      "loss": 0.8702,
       "step": 20372
     },
     {
-      "epoch": 0.5595287138502101,
+      "epoch": 0.5781214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 8.563085773930941e-06,
-      "loss": 0.8918,
+      "learning_rate": 7.970174125384116e-06,
+      "loss": 0.7612,
       "step": 20373
     },
     {
-      "epoch": 0.5595561780780534,
+      "epoch": 0.578149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 8.562205494519447e-06,
-      "loss": 0.8808,
+      "learning_rate": 7.969274188938013e-06,
+      "loss": 0.9021,
       "step": 20374
     },
     {
-      "epoch": 0.5595836423058965,
+      "epoch": 0.5781782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 8.561325226484339e-06,
-      "loss": 0.9282,
+      "learning_rate": 7.968374269645545e-06,
+      "loss": 0.9194,
       "step": 20375
     },
     {
-      "epoch": 0.5596111065337398,
+      "epoch": 0.5782065834279229,
       "grad_norm": 0.0,
-      "learning_rate": 8.560444969832584e-06,
-      "loss": 0.8189,
+      "learning_rate": 7.967474367514306e-06,
+      "loss": 0.8855,
       "step": 20376
     },
     {
-      "epoch": 0.5596385707615831,
+      "epoch": 0.5782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 8.559564724571145e-06,
-      "loss": 0.8043,
+      "learning_rate": 7.966574482551905e-06,
+      "loss": 0.861,
       "step": 20377
     },
     {
-      "epoch": 0.5596660349894262,
+      "epoch": 0.5782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 8.558684490706987e-06,
-      "loss": 0.9786,
+      "learning_rate": 7.965674614765942e-06,
+      "loss": 0.7874,
       "step": 20378
     },
     {
-      "epoch": 0.5596934992172695,
+      "epoch": 0.5782917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 8.557804268247081e-06,
-      "loss": 1.0464,
+      "learning_rate": 7.964774764164013e-06,
+      "loss": 0.9459,
       "step": 20379
     },
     {
-      "epoch": 0.5597209634451128,
+      "epoch": 0.5783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 8.556924057198383e-06,
-      "loss": 0.9471,
+      "learning_rate": 7.963874930753728e-06,
+      "loss": 0.7858,
       "step": 20380
     },
     {
-      "epoch": 0.559748427672956,
+      "epoch": 0.5783484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 8.556043857567862e-06,
-      "loss": 0.9295,
+      "learning_rate": 7.962975114542681e-06,
+      "loss": 0.8117,
       "step": 20381
     },
     {
-      "epoch": 0.5597758919007992,
+      "epoch": 0.5783768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 8.555163669362479e-06,
-      "loss": 0.8099,
+      "learning_rate": 7.962075315538471e-06,
+      "loss": 0.9214,
       "step": 20382
     },
     {
-      "epoch": 0.5598033561286424,
+      "epoch": 0.5784052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 8.554283492589199e-06,
-      "loss": 0.8955,
+      "learning_rate": 7.961175533748707e-06,
+      "loss": 0.8633,
       "step": 20383
     },
     {
-      "epoch": 0.5598308203564857,
+      "epoch": 0.578433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 8.553403327254994e-06,
-      "loss": 0.9385,
+      "learning_rate": 7.960275769180982e-06,
+      "loss": 0.8758,
       "step": 20384
     },
     {
-      "epoch": 0.559858284584329,
+      "epoch": 0.5784619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 8.552523173366822e-06,
-      "loss": 0.9146,
+      "learning_rate": 7.959376021842903e-06,
+      "loss": 0.9141,
       "step": 20385
     },
     {
-      "epoch": 0.5598857488121721,
+      "epoch": 0.5784903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 8.551643030931643e-06,
-      "loss": 0.9138,
+      "learning_rate": 7.958476291742065e-06,
+      "loss": 0.7746,
       "step": 20386
     },
     {
-      "epoch": 0.5599132130400154,
+      "epoch": 0.5785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 8.550762899956424e-06,
-      "loss": 0.8775,
+      "learning_rate": 7.957576578886068e-06,
+      "loss": 0.9052,
       "step": 20387
     },
     {
-      "epoch": 0.5599406772678586,
+      "epoch": 0.5785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 8.549882780448136e-06,
-      "loss": 0.9161,
+      "learning_rate": 7.956676883282514e-06,
+      "loss": 0.9079,
       "step": 20388
     },
     {
-      "epoch": 0.5599681414957018,
+      "epoch": 0.5785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 8.549002672413732e-06,
-      "loss": 0.8793,
+      "learning_rate": 7.955777204939003e-06,
+      "loss": 0.8353,
       "step": 20389
     },
     {
-      "epoch": 0.5599956057235451,
+      "epoch": 0.5786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 8.548122575860186e-06,
-      "loss": 0.8581,
+      "learning_rate": 7.954877543863133e-06,
+      "loss": 0.942,
       "step": 20390
     },
     {
-      "epoch": 0.5600230699513883,
+      "epoch": 0.5786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 8.547242490794452e-06,
-      "loss": 0.8358,
+      "learning_rate": 7.953977900062506e-06,
+      "loss": 0.8486,
       "step": 20391
     },
     {
-      "epoch": 0.5600505341792316,
+      "epoch": 0.5786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 8.546362417223501e-06,
-      "loss": 0.8842,
+      "learning_rate": 7.953078273544718e-06,
+      "loss": 0.8202,
       "step": 20392
     },
     {
-      "epoch": 0.5600779984070748,
+      "epoch": 0.578688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 8.545482355154292e-06,
-      "loss": 0.8809,
+      "learning_rate": 7.952178664317371e-06,
+      "loss": 0.8244,
       "step": 20393
     },
     {
-      "epoch": 0.560105462634918,
+      "epoch": 0.5787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 8.54460230459379e-06,
-      "loss": 0.8545,
+      "learning_rate": 7.951279072388063e-06,
+      "loss": 0.8154,
       "step": 20394
     },
     {
-      "epoch": 0.5601329268627613,
+      "epoch": 0.5787457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 8.543722265548957e-06,
-      "loss": 0.8768,
+      "learning_rate": 7.950379497764392e-06,
+      "loss": 0.8604,
       "step": 20395
     },
     {
-      "epoch": 0.5601603910906044,
+      "epoch": 0.5787741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 8.542842238026764e-06,
-      "loss": 0.8873,
+      "learning_rate": 7.949479940453958e-06,
+      "loss": 0.9414,
       "step": 20396
     },
     {
-      "epoch": 0.5601878553184477,
+      "epoch": 0.5788024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 8.541962222034163e-06,
-      "loss": 0.9611,
+      "learning_rate": 7.948580400464356e-06,
+      "loss": 0.8789,
       "step": 20397
     },
     {
-      "epoch": 0.560215319546291,
+      "epoch": 0.5788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 8.54108221757812e-06,
-      "loss": 0.9441,
+      "learning_rate": 7.947680877803188e-06,
+      "loss": 0.7947,
       "step": 20398
     },
     {
-      "epoch": 0.5602427837741342,
+      "epoch": 0.5788592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 8.540202224665601e-06,
-      "loss": 0.9245,
+      "learning_rate": 7.946781372478056e-06,
+      "loss": 0.9802,
       "step": 20399
     },
     {
-      "epoch": 0.5602702480019774,
+      "epoch": 0.5788876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 8.539322243303568e-06,
-      "loss": 0.8698,
+      "learning_rate": 7.945881884496553e-06,
+      "loss": 0.8536,
       "step": 20400
     },
     {
-      "epoch": 0.5602977122298206,
+      "epoch": 0.5789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 8.538442273498983e-06,
-      "loss": 0.8447,
+      "learning_rate": 7.944982413866276e-06,
+      "loss": 0.9072,
       "step": 20401
     },
     {
-      "epoch": 0.5603251764576639,
+      "epoch": 0.57894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.537562315258811e-06,
-      "loss": 0.8125,
+      "learning_rate": 7.944082960594825e-06,
+      "loss": 0.8603,
       "step": 20402
     },
     {
-      "epoch": 0.5603526406855072,
+      "epoch": 0.578972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 8.53668236859001e-06,
+      "learning_rate": 7.943183524689801e-06,
       "loss": 0.877,
       "step": 20403
     },
     {
-      "epoch": 0.5603801049133503,
+      "epoch": 0.5790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 8.535802433499547e-06,
-      "loss": 0.8672,
+      "learning_rate": 7.942284106158795e-06,
+      "loss": 0.9605,
       "step": 20404
     },
     {
-      "epoch": 0.5604075691411936,
+      "epoch": 0.5790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 8.534922509994379e-06,
-      "loss": 0.8424,
+      "learning_rate": 7.94138470500941e-06,
+      "loss": 0.8574,
       "step": 20405
     },
     {
-      "epoch": 0.5604350333690369,
+      "epoch": 0.5790578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 8.534042598081472e-06,
-      "loss": 0.8914,
+      "learning_rate": 7.94048532124924e-06,
+      "loss": 0.873,
       "step": 20406
     },
     {
-      "epoch": 0.56046249759688,
+      "epoch": 0.5790862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 8.53316269776779e-06,
-      "loss": 0.9998,
+      "learning_rate": 7.93958595488588e-06,
+      "loss": 0.9494,
       "step": 20407
     },
     {
-      "epoch": 0.5604899618247233,
+      "epoch": 0.5791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 8.532282809060293e-06,
-      "loss": 0.7693,
+      "learning_rate": 7.938686605926934e-06,
+      "loss": 0.9167,
       "step": 20408
     },
     {
-      "epoch": 0.5605174260525665,
+      "epoch": 0.5791430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 8.531402931965944e-06,
-      "loss": 0.8091,
+      "learning_rate": 7.937787274379994e-06,
+      "loss": 0.8163,
       "step": 20409
     },
     {
-      "epoch": 0.5605448902804098,
+      "epoch": 0.5791713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 8.5305230664917e-06,
-      "loss": 0.8747,
+      "learning_rate": 7.936887960252658e-06,
+      "loss": 0.7696,
       "step": 20410
     },
     {
-      "epoch": 0.560572354508253,
+      "epoch": 0.579199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 8.529643212644528e-06,
-      "loss": 0.866,
+      "learning_rate": 7.935988663552521e-06,
+      "loss": 0.8704,
       "step": 20411
     },
     {
-      "epoch": 0.5605998187360962,
+      "epoch": 0.5792281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 8.52876337043139e-06,
-      "loss": 0.9595,
+      "learning_rate": 7.935089384287177e-06,
+      "loss": 0.9411,
       "step": 20412
     },
     {
-      "epoch": 0.5606272829639395,
+      "epoch": 0.5792565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 8.527883539859244e-06,
-      "loss": 0.8593,
+      "learning_rate": 7.934190122464232e-06,
+      "loss": 0.8611,
       "step": 20413
     },
     {
-      "epoch": 0.5606547471917827,
+      "epoch": 0.5792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 8.527003720935057e-06,
-      "loss": 0.8815,
+      "learning_rate": 7.93329087809127e-06,
+      "loss": 0.8381,
       "step": 20414
     },
     {
-      "epoch": 0.5606822114196259,
+      "epoch": 0.5793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 8.526123913665784e-06,
-      "loss": 0.769,
+      "learning_rate": 7.932391651175898e-06,
+      "loss": 0.7414,
       "step": 20415
     },
     {
-      "epoch": 0.5607096756474692,
+      "epoch": 0.5793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 8.525244118058393e-06,
-      "loss": 0.8971,
+      "learning_rate": 7.931492441725707e-06,
+      "loss": 0.859,
       "step": 20416
     },
     {
-      "epoch": 0.5607371398753124,
+      "epoch": 0.5793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 8.524364334119838e-06,
-      "loss": 0.9322,
+      "learning_rate": 7.930593249748289e-06,
+      "loss": 0.8323,
       "step": 20417
     },
     {
-      "epoch": 0.5607646041031557,
+      "epoch": 0.5793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 8.523484561857083e-06,
-      "loss": 0.8386,
+      "learning_rate": 7.929694075251244e-06,
+      "loss": 0.892,
       "step": 20418
     },
     {
-      "epoch": 0.5607920683309989,
+      "epoch": 0.5794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 8.522604801277093e-06,
-      "loss": 0.9342,
+      "learning_rate": 7.928794918242168e-06,
+      "loss": 0.7605,
       "step": 20419
     },
     {
-      "epoch": 0.5608195325588421,
+      "epoch": 0.5794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 8.521725052386828e-06,
-      "loss": 0.9176,
+      "learning_rate": 7.927895778728651e-06,
+      "loss": 0.8312,
       "step": 20420
     },
     {
-      "epoch": 0.5608469967866854,
+      "epoch": 0.579483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 8.520845315193245e-06,
-      "loss": 0.8489,
+      "learning_rate": 7.926996656718296e-06,
+      "loss": 0.6455,
       "step": 20421
     },
     {
-      "epoch": 0.5608744610145285,
+      "epoch": 0.5795119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 8.519965589703305e-06,
-      "loss": 0.7703,
+      "learning_rate": 7.926097552218692e-06,
+      "loss": 0.879,
       "step": 20422
     },
     {
-      "epoch": 0.5609019252423718,
+      "epoch": 0.5795402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 8.51908587592397e-06,
-      "loss": 0.9344,
+      "learning_rate": 7.925198465237433e-06,
+      "loss": 0.8868,
       "step": 20423
     },
     {
-      "epoch": 0.5609293894702151,
+      "epoch": 0.5795686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 8.518206173862205e-06,
-      "loss": 0.7566,
+      "learning_rate": 7.924299395782116e-06,
+      "loss": 0.8291,
       "step": 20424
     },
     {
-      "epoch": 0.5609568536980583,
+      "epoch": 0.5795970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 8.517326483524963e-06,
-      "loss": 0.8664,
+      "learning_rate": 7.923400343860338e-06,
+      "loss": 0.8065,
       "step": 20425
     },
     {
-      "epoch": 0.5609843179259015,
+      "epoch": 0.5796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 8.516446804919214e-06,
-      "loss": 0.8534,
+      "learning_rate": 7.92250130947969e-06,
+      "loss": 0.8772,
       "step": 20426
     },
     {
-      "epoch": 0.5610117821537447,
+      "epoch": 0.5796538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 8.515567138051908e-06,
-      "loss": 0.8025,
+      "learning_rate": 7.921602292647763e-06,
+      "loss": 0.8142,
       "step": 20427
     },
     {
-      "epoch": 0.561039246381588,
+      "epoch": 0.5796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 8.51468748293001e-06,
-      "loss": 0.9619,
+      "learning_rate": 7.920703293372153e-06,
+      "loss": 0.9304,
       "step": 20428
     },
     {
-      "epoch": 0.5610667106094313,
+      "epoch": 0.5797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.51380783956048e-06,
-      "loss": 0.9389,
+      "learning_rate": 7.919804311660463e-06,
+      "loss": 0.7099,
       "step": 20429
     },
     {
-      "epoch": 0.5610941748372744,
+      "epoch": 0.5797389330306469,
       "grad_norm": 0.0,
-      "learning_rate": 8.512928207950278e-06,
-      "loss": 0.8353,
+      "learning_rate": 7.918905347520278e-06,
+      "loss": 0.9284,
       "step": 20430
     },
     {
-      "epoch": 0.5611216390651177,
+      "epoch": 0.5797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 8.512048588106365e-06,
-      "loss": 0.8705,
+      "learning_rate": 7.918006400959191e-06,
+      "loss": 0.821,
       "step": 20431
     },
     {
-      "epoch": 0.561149103292961,
+      "epoch": 0.5797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 8.5111689800357e-06,
-      "loss": 0.9271,
+      "learning_rate": 7.917107471984798e-06,
+      "loss": 0.919,
       "step": 20432
     },
     {
-      "epoch": 0.5611765675208041,
+      "epoch": 0.5798240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 8.510289383745245e-06,
-      "loss": 0.9225,
+      "learning_rate": 7.916208560604693e-06,
+      "loss": 0.8577,
       "step": 20433
     },
     {
-      "epoch": 0.5612040317486474,
+      "epoch": 0.5798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 8.509409799241954e-06,
-      "loss": 0.8953,
+      "learning_rate": 7.915309666826466e-06,
+      "loss": 0.8473,
       "step": 20434
     },
     {
-      "epoch": 0.5612314959764906,
+      "epoch": 0.5798808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 8.508530226532789e-06,
-      "loss": 0.9189,
+      "learning_rate": 7.914410790657715e-06,
+      "loss": 0.8344,
       "step": 20435
     },
     {
-      "epoch": 0.5612589602043339,
+      "epoch": 0.5799091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 8.507650665624714e-06,
-      "loss": 0.8739,
+      "learning_rate": 7.913511932106028e-06,
+      "loss": 0.6851,
       "step": 20436
     },
     {
-      "epoch": 0.5612864244321771,
+      "epoch": 0.5799375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 8.506771116524681e-06,
-      "loss": 0.8172,
+      "learning_rate": 7.912613091178997e-06,
+      "loss": 0.813,
       "step": 20437
     },
     {
-      "epoch": 0.5613138886600203,
+      "epoch": 0.5799659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 8.505891579239659e-06,
-      "loss": 0.8067,
+      "learning_rate": 7.911714267884221e-06,
+      "loss": 0.8151,
       "step": 20438
     },
     {
-      "epoch": 0.5613413528878636,
+      "epoch": 0.579994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 8.505012053776597e-06,
-      "loss": 0.9349,
+      "learning_rate": 7.91081546222929e-06,
+      "loss": 0.8841,
       "step": 20439
     },
     {
-      "epoch": 0.5613688171157067,
+      "epoch": 0.5800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 8.504132540142461e-06,
-      "loss": 0.8594,
+      "learning_rate": 7.90991667422179e-06,
+      "loss": 0.8773,
       "step": 20440
     },
     {
-      "epoch": 0.56139628134355,
+      "epoch": 0.5800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 8.503253038344206e-06,
-      "loss": 0.8461,
+      "learning_rate": 7.909017903869319e-06,
+      "loss": 0.7977,
       "step": 20441
     },
     {
-      "epoch": 0.5614237455713933,
+      "epoch": 0.5800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 8.502373548388793e-06,
-      "loss": 0.8725,
+      "learning_rate": 7.90811915117947e-06,
+      "loss": 0.8834,
       "step": 20442
     },
     {
-      "epoch": 0.5614512097992365,
+      "epoch": 0.5801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 8.501494070283184e-06,
-      "loss": 0.9149,
+      "learning_rate": 7.907220416159827e-06,
+      "loss": 0.8677,
       "step": 20443
     },
     {
-      "epoch": 0.5614786740270797,
+      "epoch": 0.5801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 8.50061460403433e-06,
-      "loss": 0.9347,
+      "learning_rate": 7.906321698817992e-06,
+      "loss": 0.8262,
       "step": 20444
     },
     {
-      "epoch": 0.561506138254923,
+      "epoch": 0.5801645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 8.499735149649197e-06,
-      "loss": 0.8804,
+      "learning_rate": 7.90542299916155e-06,
+      "loss": 0.9297,
       "step": 20445
     },
     {
-      "epoch": 0.5615336024827662,
+      "epoch": 0.5801929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 8.498855707134739e-06,
-      "loss": 0.8777,
+      "learning_rate": 7.904524317198095e-06,
+      "loss": 1.0075,
       "step": 20446
     },
     {
-      "epoch": 0.5615610667106095,
+      "epoch": 0.5802213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 8.497976276497915e-06,
-      "loss": 1.0012,
+      "learning_rate": 7.903625652935217e-06,
+      "loss": 0.9464,
       "step": 20447
     },
     {
-      "epoch": 0.5615885309384526,
+      "epoch": 0.5802497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 8.497096857745687e-06,
-      "loss": 0.9006,
+      "learning_rate": 7.902727006380506e-06,
+      "loss": 0.8506,
       "step": 20448
     },
     {
-      "epoch": 0.5616159951662959,
+      "epoch": 0.5802780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 8.496217450885013e-06,
-      "loss": 0.9342,
+      "learning_rate": 7.901828377541558e-06,
+      "loss": 0.9095,
       "step": 20449
     },
     {
-      "epoch": 0.5616434593941392,
+      "epoch": 0.5803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 8.495338055922844e-06,
-      "loss": 0.7864,
+      "learning_rate": 7.900929766425958e-06,
+      "loss": 0.922,
       "step": 20450
     },
     {
-      "epoch": 0.5616709236219823,
+      "epoch": 0.5803348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 8.494458672866144e-06,
-      "loss": 0.9594,
+      "learning_rate": 7.900031173041296e-06,
+      "loss": 0.8779,
       "step": 20451
     },
     {
-      "epoch": 0.5616983878498256,
+      "epoch": 0.5803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 8.493579301721872e-06,
-      "loss": 0.8077,
+      "learning_rate": 7.89913259739517e-06,
+      "loss": 0.8561,
       "step": 20452
     },
     {
-      "epoch": 0.5617258520776688,
+      "epoch": 0.5803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 8.492699942496982e-06,
-      "loss": 0.9229,
+      "learning_rate": 7.898234039495162e-06,
+      "loss": 0.864,
       "step": 20453
     },
     {
-      "epoch": 0.5617533163055121,
+      "epoch": 0.5804199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 8.491820595198434e-06,
-      "loss": 0.8447,
+      "learning_rate": 7.897335499348866e-06,
+      "loss": 0.8493,
       "step": 20454
     },
     {
-      "epoch": 0.5617807805333553,
+      "epoch": 0.5804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 8.49094125983319e-06,
-      "loss": 0.8646,
+      "learning_rate": 7.896436976963872e-06,
+      "loss": 0.8592,
       "step": 20455
     },
     {
-      "epoch": 0.5618082447611985,
+      "epoch": 0.5804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 8.4900619364082e-06,
-      "loss": 0.7455,
+      "learning_rate": 7.895538472347772e-06,
+      "loss": 0.9069,
       "step": 20456
     },
     {
-      "epoch": 0.5618357089890418,
+      "epoch": 0.580505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 8.489182624930425e-06,
-      "loss": 0.9176,
+      "learning_rate": 7.894639985508147e-06,
+      "loss": 0.7862,
       "step": 20457
     },
     {
-      "epoch": 0.5618631732168851,
+      "epoch": 0.580533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 8.488303325406822e-06,
-      "loss": 0.8582,
+      "learning_rate": 7.893741516452594e-06,
+      "loss": 0.8285,
       "step": 20458
     },
     {
-      "epoch": 0.5618906374447282,
+      "epoch": 0.5805618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 8.487424037844348e-06,
-      "loss": 0.856,
+      "learning_rate": 7.892843065188705e-06,
+      "loss": 0.8695,
       "step": 20459
     },
     {
-      "epoch": 0.5619181016725715,
+      "epoch": 0.5805902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 8.486544762249961e-06,
-      "loss": 0.8994,
+      "learning_rate": 7.891944631724064e-06,
+      "loss": 0.8283,
       "step": 20460
     },
     {
-      "epoch": 0.5619455659004147,
+      "epoch": 0.5806186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 8.485665498630624e-06,
-      "loss": 0.87,
+      "learning_rate": 7.891046216066259e-06,
+      "loss": 0.8875,
       "step": 20461
     },
     {
-      "epoch": 0.5619730301282579,
+      "epoch": 0.5806469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 8.484786246993282e-06,
-      "loss": 0.9897,
+      "learning_rate": 7.890147818222885e-06,
+      "loss": 0.8945,
       "step": 20462
     },
     {
-      "epoch": 0.5620004943561012,
+      "epoch": 0.5806753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 8.483907007344898e-06,
-      "loss": 0.8366,
+      "learning_rate": 7.889249438201526e-06,
+      "loss": 0.8857,
       "step": 20463
     },
     {
-      "epoch": 0.5620279585839444,
+      "epoch": 0.5807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 8.48302777969243e-06,
-      "loss": 0.7926,
+      "learning_rate": 7.888351076009772e-06,
+      "loss": 0.863,
       "step": 20464
     },
     {
-      "epoch": 0.5620554228117877,
+      "epoch": 0.5807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 8.482148564042834e-06,
-      "loss": 0.9693,
+      "learning_rate": 7.88745273165521e-06,
+      "loss": 0.7932,
       "step": 20465
     },
     {
-      "epoch": 0.5620828870396308,
+      "epoch": 0.5807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 8.481269360403066e-06,
-      "loss": 0.9459,
+      "learning_rate": 7.886554405145433e-06,
+      "loss": 0.8401,
       "step": 20466
     },
     {
-      "epoch": 0.5621103512674741,
+      "epoch": 0.580788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 8.480390168780085e-06,
-      "loss": 0.8555,
+      "learning_rate": 7.885656096488023e-06,
+      "loss": 0.9511,
       "step": 20467
     },
     {
-      "epoch": 0.5621378154953174,
+      "epoch": 0.5808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 8.479510989180844e-06,
-      "loss": 0.9141,
+      "learning_rate": 7.884757805690572e-06,
+      "loss": 0.8232,
       "step": 20468
     },
     {
-      "epoch": 0.5621652797231605,
+      "epoch": 0.5808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 8.478631821612303e-06,
-      "loss": 0.9269,
+      "learning_rate": 7.883859532760666e-06,
+      "loss": 0.8253,
       "step": 20469
     },
     {
-      "epoch": 0.5621927439510038,
+      "epoch": 0.5808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 8.477752666081413e-06,
-      "loss": 0.7881,
+      "learning_rate": 7.882961277705897e-06,
+      "loss": 0.8486,
       "step": 20470
     },
     {
-      "epoch": 0.5622202081788471,
+      "epoch": 0.5809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 8.476873522595135e-06,
-      "loss": 0.7919,
+      "learning_rate": 7.882063040533844e-06,
+      "loss": 0.9529,
       "step": 20471
     },
     {
-      "epoch": 0.5622476724066903,
+      "epoch": 0.5809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 8.475994391160425e-06,
-      "loss": 0.843,
+      "learning_rate": 7.881164821252103e-06,
+      "loss": 0.7759,
       "step": 20472
     },
     {
-      "epoch": 0.5622751366345335,
+      "epoch": 0.5809591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 8.47511527178424e-06,
-      "loss": 0.9104,
+      "learning_rate": 7.880266619868257e-06,
+      "loss": 0.9476,
       "step": 20473
     },
     {
-      "epoch": 0.5623026008623767,
+      "epoch": 0.5809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 8.47423616447353e-06,
-      "loss": 0.9526,
+      "learning_rate": 7.87936843638989e-06,
+      "loss": 0.7982,
       "step": 20474
     },
     {
-      "epoch": 0.56233006509022,
+      "epoch": 0.5810158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 8.473357069235253e-06,
-      "loss": 0.8379,
+      "learning_rate": 7.8784702708246e-06,
+      "loss": 0.8846,
       "step": 20475
     },
     {
-      "epoch": 0.5623575293180633,
+      "epoch": 0.581044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 8.47247798607637e-06,
-      "loss": 0.9299,
+      "learning_rate": 7.877572123179964e-06,
+      "loss": 0.8721,
       "step": 20476
     },
     {
-      "epoch": 0.5623849935459064,
+      "epoch": 0.5810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 8.47159891500383e-06,
-      "loss": 0.9811,
+      "learning_rate": 7.876673993463573e-06,
+      "loss": 0.8563,
       "step": 20477
     },
     {
-      "epoch": 0.5624124577737497,
+      "epoch": 0.5811010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 8.47071985602459e-06,
-      "loss": 0.8443,
+      "learning_rate": 7.87577588168301e-06,
+      "loss": 0.9176,
       "step": 20478
     },
     {
-      "epoch": 0.5624399220015929,
+      "epoch": 0.5811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 8.469840809145613e-06,
-      "loss": 0.8371,
+      "learning_rate": 7.874877787845865e-06,
+      "loss": 0.8466,
       "step": 20479
     },
     {
-      "epoch": 0.5624673862294362,
+      "epoch": 0.5811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 8.468961774373844e-06,
-      "loss": 0.8277,
+      "learning_rate": 7.873979711959724e-06,
+      "loss": 0.9225,
       "step": 20480
     },
     {
-      "epoch": 0.5624948504572794,
+      "epoch": 0.5811861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 8.468082751716245e-06,
-      "loss": 0.7647,
+      "learning_rate": 7.87308165403217e-06,
+      "loss": 0.8636,
       "step": 20481
     },
     {
-      "epoch": 0.5625223146851226,
+      "epoch": 0.5812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 8.467203741179764e-06,
-      "loss": 0.9504,
+      "learning_rate": 7.872183614070791e-06,
+      "loss": 0.8902,
       "step": 20482
     },
     {
-      "epoch": 0.5625497789129659,
+      "epoch": 0.5812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 8.466324742771363e-06,
-      "loss": 0.8864,
+      "learning_rate": 7.871285592083175e-06,
+      "loss": 0.8265,
       "step": 20483
     },
     {
-      "epoch": 0.5625772431408091,
+      "epoch": 0.5812712826333711,
       "grad_norm": 0.0,
-      "learning_rate": 8.465445756497997e-06,
-      "loss": 0.9092,
+      "learning_rate": 7.870387588076902e-06,
+      "loss": 0.9123,
       "step": 20484
     },
     {
-      "epoch": 0.5626047073686523,
+      "epoch": 0.5812996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 8.464566782366619e-06,
-      "loss": 0.9202,
+      "learning_rate": 7.869489602059565e-06,
+      "loss": 0.8697,
       "step": 20485
     },
     {
-      "epoch": 0.5626321715964956,
+      "epoch": 0.581328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 8.46368782038418e-06,
-      "loss": 0.9399,
+      "learning_rate": 7.868591634038742e-06,
+      "loss": 0.8526,
       "step": 20486
     },
     {
-      "epoch": 0.5626596358243388,
+      "epoch": 0.5813564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 8.462808870557639e-06,
-      "loss": 0.9311,
+      "learning_rate": 7.867693684022022e-06,
+      "loss": 0.9309,
       "step": 20487
     },
     {
-      "epoch": 0.562687100052182,
+      "epoch": 0.5813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 8.46192993289395e-06,
-      "loss": 0.8804,
+      "learning_rate": 7.86679575201699e-06,
+      "loss": 0.8592,
       "step": 20488
     },
     {
-      "epoch": 0.5627145642800253,
+      "epoch": 0.5814131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 8.461051007400068e-06,
-      "loss": 0.8071,
+      "learning_rate": 7.865897838031225e-06,
+      "loss": 0.8849,
       "step": 20489
     },
     {
-      "epoch": 0.5627420285078685,
+      "epoch": 0.5814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 8.460172094082944e-06,
-      "loss": 0.8585,
+      "learning_rate": 7.864999942072321e-06,
+      "loss": 0.8382,
       "step": 20490
     },
     {
-      "epoch": 0.5627694927357118,
+      "epoch": 0.5814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 8.45929319294954e-06,
-      "loss": 0.9376,
+      "learning_rate": 7.86410206414786e-06,
+      "loss": 0.8449,
       "step": 20491
     },
     {
-      "epoch": 0.5627969569635549,
+      "epoch": 0.5814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 8.4584143040068e-06,
-      "loss": 0.9562,
+      "learning_rate": 7.863204204265423e-06,
+      "loss": 0.9323,
       "step": 20492
     },
     {
-      "epoch": 0.5628244211913982,
+      "epoch": 0.5815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 8.457535427261685e-06,
-      "loss": 0.7669,
+      "learning_rate": 7.862306362432597e-06,
+      "loss": 0.9717,
       "step": 20493
     },
     {
-      "epoch": 0.5628518854192415,
+      "epoch": 0.5815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 8.456656562721146e-06,
-      "loss": 0.9443,
+      "learning_rate": 7.861408538656966e-06,
+      "loss": 0.8113,
       "step": 20494
     },
     {
-      "epoch": 0.5628793496470846,
+      "epoch": 0.581583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 8.455777710392138e-06,
-      "loss": 0.9089,
+      "learning_rate": 7.860510732946111e-06,
+      "loss": 0.7923,
       "step": 20495
     },
     {
-      "epoch": 0.5629068138749279,
+      "epoch": 0.5816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 8.454898870281619e-06,
-      "loss": 0.8464,
+      "learning_rate": 7.85961294530762e-06,
+      "loss": 0.8912,
       "step": 20496
     },
     {
-      "epoch": 0.5629342781027712,
+      "epoch": 0.5816401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 8.454020042396537e-06,
-      "loss": 0.8195,
+      "learning_rate": 7.858715175749075e-06,
+      "loss": 0.8187,
       "step": 20497
     },
     {
-      "epoch": 0.5629617423306144,
+      "epoch": 0.5816685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 8.453141226743845e-06,
-      "loss": 0.9633,
+      "learning_rate": 7.857817424278056e-06,
+      "loss": 0.8578,
       "step": 20498
     },
     {
-      "epoch": 0.5629892065584576,
+      "epoch": 0.5816969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 8.452262423330499e-06,
-      "loss": 0.8173,
+      "learning_rate": 7.856919690902152e-06,
+      "loss": 0.9483,
       "step": 20499
     },
     {
-      "epoch": 0.5630166707863008,
+      "epoch": 0.5817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 8.451383632163455e-06,
-      "loss": 0.9425,
+      "learning_rate": 7.856021975628945e-06,
+      "loss": 0.8841,
       "step": 20500
     },
     {
-      "epoch": 0.5630441350141441,
+      "epoch": 0.5817536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 8.45050485324966e-06,
-      "loss": 0.9839,
+      "learning_rate": 7.855124278466013e-06,
+      "loss": 0.8773,
       "step": 20501
     },
     {
-      "epoch": 0.5630715992419874,
+      "epoch": 0.5817820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 8.449626086596075e-06,
-      "loss": 0.7944,
+      "learning_rate": 7.854226599420947e-06,
+      "loss": 0.8631,
       "step": 20502
     },
     {
-      "epoch": 0.5630990634698305,
+      "epoch": 0.5818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 8.448747332209645e-06,
-      "loss": 0.8195,
+      "learning_rate": 7.853328938501323e-06,
+      "loss": 0.7227,
       "step": 20503
     },
     {
-      "epoch": 0.5631265276976738,
+      "epoch": 0.581838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 8.447868590097327e-06,
-      "loss": 0.8942,
+      "learning_rate": 7.852431295714722e-06,
+      "loss": 0.8115,
       "step": 20504
     },
     {
-      "epoch": 0.563153991925517,
+      "epoch": 0.5818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 8.446989860266077e-06,
-      "loss": 0.8756,
+      "learning_rate": 7.851533671068737e-06,
+      "loss": 0.7968,
       "step": 20505
     },
     {
-      "epoch": 0.5631814561533602,
+      "epoch": 0.5818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 8.446111142722843e-06,
-      "loss": 0.9027,
+      "learning_rate": 7.850636064570939e-06,
+      "loss": 0.8645,
       "step": 20506
     },
     {
-      "epoch": 0.5632089203812035,
+      "epoch": 0.5819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 8.445232437474578e-06,
-      "loss": 0.8773,
+      "learning_rate": 7.849738476228916e-06,
+      "loss": 0.7883,
       "step": 20507
     },
     {
-      "epoch": 0.5632363846090467,
+      "epoch": 0.5819523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 8.44435374452824e-06,
-      "loss": 0.8358,
+      "learning_rate": 7.848840906050252e-06,
+      "loss": 0.8465,
       "step": 20508
     },
     {
-      "epoch": 0.56326384883689,
+      "epoch": 0.5819807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 8.443475063890778e-06,
-      "loss": 0.9459,
+      "learning_rate": 7.84794335404252e-06,
+      "loss": 0.7854,
       "step": 20509
     },
     {
-      "epoch": 0.5632913130647332,
+      "epoch": 0.5820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 8.44259639556914e-06,
-      "loss": 0.7925,
+      "learning_rate": 7.847045820213312e-06,
+      "loss": 0.8836,
       "step": 20510
     },
     {
-      "epoch": 0.5633187772925764,
+      "epoch": 0.5820374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 8.441717739570284e-06,
-      "loss": 0.8129,
+      "learning_rate": 7.846148304570204e-06,
+      "loss": 0.8013,
       "step": 20511
     },
     {
-      "epoch": 0.5633462415204197,
+      "epoch": 0.5820658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 8.440839095901162e-06,
-      "loss": 0.876,
+      "learning_rate": 7.845250807120776e-06,
+      "loss": 0.809,
       "step": 20512
     },
     {
-      "epoch": 0.5633737057482628,
+      "epoch": 0.5820942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 8.439960464568724e-06,
-      "loss": 0.9604,
+      "learning_rate": 7.844353327872614e-06,
+      "loss": 0.8216,
       "step": 20513
     },
     {
-      "epoch": 0.5634011699761061,
+      "epoch": 0.5821225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 8.439081845579927e-06,
-      "loss": 0.944,
+      "learning_rate": 7.843455866833295e-06,
+      "loss": 0.9932,
       "step": 20514
     },
     {
-      "epoch": 0.5634286342039494,
+      "epoch": 0.5821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 8.438203238941713e-06,
-      "loss": 0.8704,
+      "learning_rate": 7.8425584240104e-06,
+      "loss": 0.8821,
       "step": 20515
     },
     {
-      "epoch": 0.5634560984317926,
+      "epoch": 0.5821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 8.437324644661042e-06,
-      "loss": 0.7933,
+      "learning_rate": 7.841660999411513e-06,
+      "loss": 0.8551,
       "step": 20516
     },
     {
-      "epoch": 0.5634835626596358,
+      "epoch": 0.5822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 8.436446062744865e-06,
-      "loss": 0.858,
+      "learning_rate": 7.84076359304421e-06,
+      "loss": 0.8478,
       "step": 20517
     },
     {
-      "epoch": 0.563511026887479,
+      "epoch": 0.5822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 8.435567493200132e-06,
-      "loss": 0.8065,
+      "learning_rate": 7.839866204916077e-06,
+      "loss": 0.7141,
       "step": 20518
     },
     {
-      "epoch": 0.5635384911153223,
+      "epoch": 0.5822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 8.434688936033792e-06,
-      "loss": 0.8468,
+      "learning_rate": 7.83896883503469e-06,
+      "loss": 0.9181,
       "step": 20519
     },
     {
-      "epoch": 0.5635659553431656,
+      "epoch": 0.5822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 8.433810391252807e-06,
-      "loss": 0.9127,
+      "learning_rate": 7.838071483407627e-06,
+      "loss": 0.9892,
       "step": 20520
     },
     {
-      "epoch": 0.5635934195710087,
+      "epoch": 0.5823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 8.432931858864115e-06,
-      "loss": 0.877,
+      "learning_rate": 7.837174150042475e-06,
+      "loss": 0.9624,
       "step": 20521
     },
     {
-      "epoch": 0.563620883798852,
+      "epoch": 0.582349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 8.432053338874675e-06,
-      "loss": 0.908,
+      "learning_rate": 7.83627683494681e-06,
+      "loss": 0.8049,
       "step": 20522
     },
     {
-      "epoch": 0.5636483480266953,
+      "epoch": 0.582377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 8.431174831291435e-06,
-      "loss": 0.9269,
+      "learning_rate": 7.83537953812821e-06,
+      "loss": 0.8082,
       "step": 20523
     },
     {
-      "epoch": 0.5636758122545384,
+      "epoch": 0.5824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 8.43029633612135e-06,
-      "loss": 0.9555,
+      "learning_rate": 7.83448225959426e-06,
+      "loss": 0.8449,
       "step": 20524
     },
     {
-      "epoch": 0.5637032764823817,
+      "epoch": 0.5824347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 8.429417853371363e-06,
-      "loss": 0.9263,
+      "learning_rate": 7.833584999352533e-06,
+      "loss": 0.8913,
       "step": 20525
     },
     {
-      "epoch": 0.5637307407102249,
+      "epoch": 0.5824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 8.428539383048437e-06,
-      "loss": 0.8929,
+      "learning_rate": 7.832687757410612e-06,
+      "loss": 0.8581,
       "step": 20526
     },
     {
-      "epoch": 0.5637582049380682,
+      "epoch": 0.5824914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 8.427660925159512e-06,
-      "loss": 0.9095,
+      "learning_rate": 7.831790533776077e-06,
+      "loss": 0.8417,
       "step": 20527
     },
     {
-      "epoch": 0.5637856691659114,
+      "epoch": 0.5825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 8.426782479711542e-06,
-      "loss": 0.8958,
+      "learning_rate": 7.830893328456501e-06,
+      "loss": 0.9824,
       "step": 20528
     },
     {
-      "epoch": 0.5638131333937546,
+      "epoch": 0.5825482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 8.425904046711481e-06,
-      "loss": 0.9457,
+      "learning_rate": 7.829996141459468e-06,
+      "loss": 0.8613,
       "step": 20529
     },
     {
-      "epoch": 0.5638405976215979,
+      "epoch": 0.5825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 8.425025626166277e-06,
-      "loss": 0.9247,
+      "learning_rate": 7.829098972792555e-06,
+      "loss": 0.7548,
       "step": 20530
     },
     {
-      "epoch": 0.563868061849441,
+      "epoch": 0.5826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 8.424147218082877e-06,
-      "loss": 0.8601,
+      "learning_rate": 7.828201822463342e-06,
+      "loss": 0.7976,
       "step": 20531
     },
     {
-      "epoch": 0.5638955260772843,
+      "epoch": 0.582633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 8.42326882246824e-06,
-      "loss": 0.9178,
+      "learning_rate": 7.827304690479403e-06,
+      "loss": 0.8427,
       "step": 20532
     },
     {
-      "epoch": 0.5639229903051276,
+      "epoch": 0.5826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 8.42239043932931e-06,
-      "loss": 0.8838,
+      "learning_rate": 7.82640757684832e-06,
+      "loss": 0.8083,
       "step": 20533
     },
     {
-      "epoch": 0.5639504545329708,
+      "epoch": 0.5826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 8.421512068673035e-06,
-      "loss": 0.8474,
+      "learning_rate": 7.825510481577672e-06,
+      "loss": 0.7791,
       "step": 20534
     },
     {
-      "epoch": 0.563977918760814,
+      "epoch": 0.5827185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 8.420633710506368e-06,
-      "loss": 0.8728,
+      "learning_rate": 7.82461340467503e-06,
+      "loss": 0.8383,
       "step": 20535
     },
     {
-      "epoch": 0.5640053829886573,
+      "epoch": 0.5827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 8.41975536483626e-06,
-      "loss": 0.986,
+      "learning_rate": 7.823716346147977e-06,
+      "loss": 0.8662,
       "step": 20536
     },
     {
-      "epoch": 0.5640328472165005,
+      "epoch": 0.5827752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 8.418877031669658e-06,
-      "loss": 0.8155,
+      "learning_rate": 7.822819306004094e-06,
+      "loss": 0.7683,
       "step": 20537
     },
     {
-      "epoch": 0.5640603114443438,
+      "epoch": 0.5828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 8.417998711013518e-06,
-      "loss": 0.8073,
+      "learning_rate": 7.821922284250954e-06,
+      "loss": 0.9096,
       "step": 20538
     },
     {
-      "epoch": 0.5640877756721869,
+      "epoch": 0.5828320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 8.417120402874781e-06,
-      "loss": 0.8518,
+      "learning_rate": 7.82102528089613e-06,
+      "loss": 0.857,
       "step": 20539
     },
     {
-      "epoch": 0.5641152399000302,
+      "epoch": 0.5828603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 8.4162421072604e-06,
-      "loss": 0.8602,
+      "learning_rate": 7.820128295947206e-06,
+      "loss": 0.8341,
       "step": 20540
     },
     {
-      "epoch": 0.5641427041278735,
+      "epoch": 0.58288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 8.415363824177328e-06,
-      "loss": 0.8605,
+      "learning_rate": 7.819231329411758e-06,
+      "loss": 0.9157,
       "step": 20541
     },
     {
-      "epoch": 0.5641701683557167,
+      "epoch": 0.5829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 8.414485553632508e-06,
-      "loss": 0.8683,
+      "learning_rate": 7.818334381297359e-06,
+      "loss": 0.8804,
       "step": 20542
     },
     {
-      "epoch": 0.5641976325835599,
+      "epoch": 0.5829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 8.413607295632892e-06,
-      "loss": 0.9018,
+      "learning_rate": 7.81743745161159e-06,
+      "loss": 0.8737,
       "step": 20543
     },
     {
-      "epoch": 0.5642250968114031,
+      "epoch": 0.5829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 8.412729050185436e-06,
-      "loss": 0.875,
+      "learning_rate": 7.816540540362022e-06,
+      "loss": 0.8661,
       "step": 20544
     },
     {
-      "epoch": 0.5642525610392464,
+      "epoch": 0.5830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 8.411850817297078e-06,
-      "loss": 0.9799,
+      "learning_rate": 7.815643647556235e-06,
+      "loss": 0.9256,
       "step": 20545
     },
     {
-      "epoch": 0.5642800252670896,
+      "epoch": 0.5830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 8.410972596974772e-06,
-      "loss": 0.8623,
+      "learning_rate": 7.814746773201804e-06,
+      "loss": 0.8851,
       "step": 20546
     },
     {
-      "epoch": 0.5643074894949328,
+      "epoch": 0.5830590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 8.410094389225464e-06,
-      "loss": 0.7933,
+      "learning_rate": 7.813849917306308e-06,
+      "loss": 0.9232,
       "step": 20547
     },
     {
-      "epoch": 0.5643349537227761,
+      "epoch": 0.5830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 8.409216194056108e-06,
-      "loss": 0.9238,
+      "learning_rate": 7.812953079877317e-06,
+      "loss": 0.94,
       "step": 20548
     },
     {
-      "epoch": 0.5643624179506194,
+      "epoch": 0.5831157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 8.408338011473647e-06,
-      "loss": 0.9113,
+      "learning_rate": 7.81205626092241e-06,
+      "loss": 0.8519,
       "step": 20549
     },
     {
-      "epoch": 0.5643898821784625,
+      "epoch": 0.5831441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 8.407459841485037e-06,
-      "loss": 0.903,
+      "learning_rate": 7.811159460449164e-06,
+      "loss": 0.722,
       "step": 20550
     },
     {
-      "epoch": 0.5644173464063058,
+      "epoch": 0.5831725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 8.406581684097218e-06,
-      "loss": 0.9221,
+      "learning_rate": 7.810262678465147e-06,
+      "loss": 0.9491,
       "step": 20551
     },
     {
-      "epoch": 0.564444810634149,
+      "epoch": 0.5832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 8.405703539317144e-06,
-      "loss": 0.942,
+      "learning_rate": 7.809365914977945e-06,
+      "loss": 0.8839,
       "step": 20552
     },
     {
-      "epoch": 0.5644722748619923,
+      "epoch": 0.5832292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 8.404825407151758e-06,
-      "loss": 0.8543,
+      "learning_rate": 7.808469169995125e-06,
+      "loss": 0.7572,
       "step": 20553
     },
     {
-      "epoch": 0.5644997390898355,
+      "epoch": 0.5832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 8.403947287608012e-06,
-      "loss": 0.8915,
+      "learning_rate": 7.807572443524266e-06,
+      "loss": 0.8164,
       "step": 20554
     },
     {
-      "epoch": 0.5645272033176787,
+      "epoch": 0.5832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 8.403069180692854e-06,
-      "loss": 0.7875,
+      "learning_rate": 7.806675735572941e-06,
+      "loss": 0.8696,
       "step": 20555
     },
     {
-      "epoch": 0.564554667545522,
+      "epoch": 0.5833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 8.402191086413233e-06,
-      "loss": 0.7614,
+      "learning_rate": 7.805779046148723e-06,
+      "loss": 0.8987,
       "step": 20556
     },
     {
-      "epoch": 0.5645821317733651,
+      "epoch": 0.5833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 8.401313004776096e-06,
-      "loss": 0.7808,
+      "learning_rate": 7.80488237525919e-06,
+      "loss": 0.8587,
       "step": 20557
     },
     {
-      "epoch": 0.5646095960012084,
+      "epoch": 0.5833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 8.400434935788386e-06,
-      "loss": 0.8148,
+      "learning_rate": 7.803985722911915e-06,
+      "loss": 0.9725,
       "step": 20558
     },
     {
-      "epoch": 0.5646370602290517,
+      "epoch": 0.583399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 8.399556879457055e-06,
-      "loss": 0.7735,
+      "learning_rate": 7.80308908911447e-06,
+      "loss": 0.8967,
       "step": 20559
     },
     {
-      "epoch": 0.5646645244568949,
+      "epoch": 0.583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 8.398678835789052e-06,
-      "loss": 0.8163,
+      "learning_rate": 7.80219247387443e-06,
+      "loss": 0.8238,
       "step": 20560
     },
     {
-      "epoch": 0.5646919886847381,
+      "epoch": 0.5834562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 8.397800804791324e-06,
-      "loss": 0.7957,
+      "learning_rate": 7.80129587719937e-06,
+      "loss": 0.9679,
       "step": 20561
     },
     {
-      "epoch": 0.5647194529125814,
+      "epoch": 0.5834846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 8.396922786470813e-06,
-      "loss": 0.9039,
+      "learning_rate": 7.80039929909686e-06,
+      "loss": 0.848,
       "step": 20562
     },
     {
-      "epoch": 0.5647469171404246,
+      "epoch": 0.5835130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 8.396044780834469e-06,
-      "loss": 0.8417,
+      "learning_rate": 7.79950273957448e-06,
+      "loss": 0.9015,
       "step": 20563
     },
     {
-      "epoch": 0.5647743813682679,
+      "epoch": 0.5835414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 8.395166787889242e-06,
-      "loss": 0.8549,
+      "learning_rate": 7.798606198639797e-06,
+      "loss": 0.9289,
       "step": 20564
     },
     {
-      "epoch": 0.564801845596111,
+      "epoch": 0.5835698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 8.394288807642076e-06,
-      "loss": 0.9396,
+      "learning_rate": 7.797709676300385e-06,
+      "loss": 0.8531,
       "step": 20565
     },
     {
-      "epoch": 0.5648293098239543,
+      "epoch": 0.5835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 8.393410840099919e-06,
-      "loss": 0.8105,
+      "learning_rate": 7.796813172563817e-06,
+      "loss": 0.8227,
       "step": 20566
     },
     {
-      "epoch": 0.5648567740517976,
+      "epoch": 0.5836265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 8.392532885269721e-06,
-      "loss": 0.9719,
+      "learning_rate": 7.79591668743767e-06,
+      "loss": 0.8639,
       "step": 20567
     },
     {
-      "epoch": 0.5648842382796407,
+      "epoch": 0.5836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 8.39165494315842e-06,
-      "loss": 0.8575,
+      "learning_rate": 7.795020220929515e-06,
+      "loss": 0.903,
       "step": 20568
     },
     {
-      "epoch": 0.564911702507484,
+      "epoch": 0.583683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 8.390777013772974e-06,
-      "loss": 0.8684,
+      "learning_rate": 7.794123773046923e-06,
+      "loss": 0.9792,
       "step": 20569
     },
     {
-      "epoch": 0.5649391667353273,
+      "epoch": 0.5837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 8.389899097120318e-06,
-      "loss": 0.894,
+      "learning_rate": 7.793227343797465e-06,
+      "loss": 0.8349,
       "step": 20570
     },
     {
-      "epoch": 0.5649666309631705,
+      "epoch": 0.5837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 8.389021193207407e-06,
-      "loss": 0.9093,
+      "learning_rate": 7.792330933188716e-06,
+      "loss": 0.89,
       "step": 20571
     },
     {
-      "epoch": 0.5649940951910137,
+      "epoch": 0.5837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 8.388143302041184e-06,
-      "loss": 0.8457,
+      "learning_rate": 7.791434541228247e-06,
+      "loss": 0.9221,
       "step": 20572
     },
     {
-      "epoch": 0.5650215594188569,
+      "epoch": 0.5837968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 8.387265423628598e-06,
-      "loss": 0.9039,
+      "learning_rate": 7.790538167923628e-06,
+      "loss": 0.826,
       "step": 20573
     },
     {
-      "epoch": 0.5650490236467002,
+      "epoch": 0.5838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 8.38638755797659e-06,
-      "loss": 0.858,
+      "learning_rate": 7.789641813282433e-06,
+      "loss": 0.9189,
       "step": 20574
     },
     {
-      "epoch": 0.5650764878745435,
+      "epoch": 0.5838535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 8.385509705092107e-06,
-      "loss": 0.8387,
+      "learning_rate": 7.788745477312236e-06,
+      "loss": 0.8278,
       "step": 20575
     },
     {
-      "epoch": 0.5651039521023866,
+      "epoch": 0.5838819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 8.3846318649821e-06,
-      "loss": 0.8076,
+      "learning_rate": 7.787849160020601e-06,
+      "loss": 0.877,
       "step": 20576
     },
     {
-      "epoch": 0.5651314163302299,
+      "epoch": 0.5839103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 8.383754037653508e-06,
-      "loss": 0.8814,
+      "learning_rate": 7.786952861415106e-06,
+      "loss": 0.9238,
       "step": 20577
     },
     {
-      "epoch": 0.5651588805580731,
+      "epoch": 0.583938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 8.382876223113283e-06,
-      "loss": 0.9072,
+      "learning_rate": 7.78605658150332e-06,
+      "loss": 0.8933,
       "step": 20578
     },
     {
-      "epoch": 0.5651863447859163,
+      "epoch": 0.5839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 8.38199842136837e-06,
-      "loss": 0.7965,
+      "learning_rate": 7.785160320292812e-06,
+      "loss": 0.9211,
       "step": 20579
     },
     {
-      "epoch": 0.5652138090137596,
+      "epoch": 0.5839954597048809,
       "grad_norm": 0.0,
-      "learning_rate": 8.381120632425706e-06,
-      "loss": 0.8386,
+      "learning_rate": 7.784264077791156e-06,
+      "loss": 0.8736,
       "step": 20580
     },
     {
-      "epoch": 0.5652412732416028,
+      "epoch": 0.5840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 8.380242856292247e-06,
-      "loss": 0.9487,
+      "learning_rate": 7.783367854005916e-06,
+      "loss": 0.8741,
       "step": 20581
     },
     {
-      "epoch": 0.5652687374694461,
+      "epoch": 0.5840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 8.379365092974931e-06,
-      "loss": 0.9017,
+      "learning_rate": 7.782471648944673e-06,
+      "loss": 0.8621,
       "step": 20582
     },
     {
-      "epoch": 0.5652962016972893,
+      "epoch": 0.5840805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 8.378487342480709e-06,
-      "loss": 0.8114,
+      "learning_rate": 7.781575462614988e-06,
+      "loss": 0.9549,
       "step": 20583
     },
     {
-      "epoch": 0.5653236659251325,
+      "epoch": 0.5841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 8.377609604816521e-06,
-      "loss": 0.8742,
+      "learning_rate": 7.780679295024438e-06,
+      "loss": 0.9334,
       "step": 20584
     },
     {
-      "epoch": 0.5653511301529758,
+      "epoch": 0.5841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 8.376731879989319e-06,
-      "loss": 0.7623,
+      "learning_rate": 7.77978314618059e-06,
+      "loss": 0.9026,
       "step": 20585
     },
     {
-      "epoch": 0.5653785943808189,
+      "epoch": 0.5841657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 8.375854168006039e-06,
-      "loss": 0.8638,
+      "learning_rate": 7.77888701609101e-06,
+      "loss": 0.8934,
       "step": 20586
     },
     {
-      "epoch": 0.5654060586086622,
+      "epoch": 0.5841940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 8.374976468873629e-06,
-      "loss": 0.8841,
+      "learning_rate": 7.777990904763274e-06,
+      "loss": 0.8259,
       "step": 20587
     },
     {
-      "epoch": 0.5654335228365055,
+      "epoch": 0.58422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 8.374098782599037e-06,
-      "loss": 0.9895,
+      "learning_rate": 7.777094812204949e-06,
+      "loss": 0.8147,
       "step": 20588
     },
     {
-      "epoch": 0.5654609870643487,
+      "epoch": 0.5842508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 8.373221109189203e-06,
-      "loss": 0.9023,
+      "learning_rate": 7.7761987384236e-06,
+      "loss": 0.7515,
       "step": 20589
     },
     {
-      "epoch": 0.5654884512921919,
+      "epoch": 0.5842792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 8.372343448651074e-06,
-      "loss": 0.817,
+      "learning_rate": 7.775302683426806e-06,
+      "loss": 0.9382,
       "step": 20590
     },
     {
-      "epoch": 0.5655159155200351,
+      "epoch": 0.5843076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 8.371465800991598e-06,
-      "loss": 0.8849,
+      "learning_rate": 7.774406647222128e-06,
+      "loss": 0.8028,
       "step": 20591
     },
     {
-      "epoch": 0.5655433797478784,
+      "epoch": 0.5843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 8.370588166217712e-06,
-      "loss": 0.8801,
+      "learning_rate": 7.773510629817137e-06,
+      "loss": 0.8823,
       "step": 20592
     },
     {
-      "epoch": 0.5655708439757217,
+      "epoch": 0.5843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 8.369710544336365e-06,
-      "loss": 0.8702,
+      "learning_rate": 7.772614631219402e-06,
+      "loss": 0.8289,
       "step": 20593
     },
     {
-      "epoch": 0.5655983082035648,
+      "epoch": 0.5843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 8.368832935354498e-06,
-      "loss": 0.8438,
+      "learning_rate": 7.77171865143649e-06,
+      "loss": 0.8669,
       "step": 20594
     },
     {
-      "epoch": 0.5656257724314081,
+      "epoch": 0.5844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 8.367955339279058e-06,
-      "loss": 0.8472,
+      "learning_rate": 7.770822690475973e-06,
+      "loss": 0.9511,
       "step": 20595
     },
     {
-      "epoch": 0.5656532366592514,
+      "epoch": 0.5844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 8.367077756116986e-06,
-      "loss": 0.7681,
+      "learning_rate": 7.769926748345414e-06,
+      "loss": 0.8609,
       "step": 20596
     },
     {
-      "epoch": 0.5656807008870945,
+      "epoch": 0.584477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 8.366200185875232e-06,
-      "loss": 0.8429,
+      "learning_rate": 7.769030825052383e-06,
+      "loss": 0.7398,
       "step": 20597
     },
     {
-      "epoch": 0.5657081651149378,
+      "epoch": 0.5845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 8.36532262856073e-06,
-      "loss": 0.9189,
+      "learning_rate": 7.768134920604453e-06,
+      "loss": 0.8617,
       "step": 20598
     },
     {
-      "epoch": 0.565735629342781,
+      "epoch": 0.5845346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 8.364445084180429e-06,
-      "loss": 0.9912,
+      "learning_rate": 7.767239035009187e-06,
+      "loss": 0.9198,
       "step": 20599
     },
     {
-      "epoch": 0.5657630935706243,
+      "epoch": 0.5845629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 8.363567552741274e-06,
-      "loss": 0.8982,
+      "learning_rate": 7.76634316827415e-06,
+      "loss": 0.9769,
       "step": 20600
     },
     {
-      "epoch": 0.5657905577984675,
+      "epoch": 0.5845913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 8.362690034250206e-06,
-      "loss": 0.9484,
+      "learning_rate": 7.765447320406916e-06,
+      "loss": 0.806,
       "step": 20601
     },
     {
-      "epoch": 0.5658180220263107,
+      "epoch": 0.5846197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 8.361812528714166e-06,
-      "loss": 0.8835,
+      "learning_rate": 7.76455149141505e-06,
+      "loss": 0.9078,
       "step": 20602
     },
     {
-      "epoch": 0.565845486254154,
+      "epoch": 0.5846481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 8.360935036140105e-06,
-      "loss": 0.9541,
+      "learning_rate": 7.763655681306113e-06,
+      "loss": 0.8592,
       "step": 20603
     },
     {
-      "epoch": 0.5658729504819972,
+      "epoch": 0.5846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 8.360057556534957e-06,
-      "loss": 0.9,
+      "learning_rate": 7.762759890087682e-06,
+      "loss": 0.8215,
       "step": 20604
     },
     {
-      "epoch": 0.5659004147098404,
+      "epoch": 0.5847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 8.359180089905671e-06,
-      "loss": 0.8328,
+      "learning_rate": 7.761864117767316e-06,
+      "loss": 0.9222,
       "step": 20605
     },
     {
-      "epoch": 0.5659278789376837,
+      "epoch": 0.584733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 8.358302636259186e-06,
-      "loss": 0.7901,
+      "learning_rate": 7.760968364352584e-06,
+      "loss": 0.8903,
       "step": 20606
     },
     {
-      "epoch": 0.5659553431655269,
+      "epoch": 0.5847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 8.357425195602444e-06,
-      "loss": 0.7805,
+      "learning_rate": 7.760072629851056e-06,
+      "loss": 0.9681,
       "step": 20607
     },
     {
-      "epoch": 0.5659828073933701,
+      "epoch": 0.5847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 8.356547767942395e-06,
-      "loss": 0.8492,
+      "learning_rate": 7.759176914270293e-06,
+      "loss": 0.9519,
       "step": 20608
     },
     {
-      "epoch": 0.5660102716212134,
+      "epoch": 0.5848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 8.355670353285978e-06,
-      "loss": 0.8883,
+      "learning_rate": 7.758281217617863e-06,
+      "loss": 0.7813,
       "step": 20609
     },
     {
-      "epoch": 0.5660377358490566,
+      "epoch": 0.5848467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 8.354792951640129e-06,
-      "loss": 0.9079,
+      "learning_rate": 7.757385539901333e-06,
+      "loss": 0.9415,
       "step": 20610
     },
     {
-      "epoch": 0.5660652000768999,
+      "epoch": 0.5848751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 8.353915563011798e-06,
-      "loss": 0.8992,
+      "learning_rate": 7.756489881128269e-06,
+      "loss": 0.9231,
       "step": 20611
     },
     {
-      "epoch": 0.566092664304743,
+      "epoch": 0.5849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 8.353038187407925e-06,
-      "loss": 0.8738,
+      "learning_rate": 7.755594241306232e-06,
+      "loss": 0.8939,
       "step": 20612
     },
     {
-      "epoch": 0.5661201285325863,
+      "epoch": 0.5849318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 8.35216082483545e-06,
-      "loss": 0.8112,
+      "learning_rate": 7.754698620442794e-06,
+      "loss": 0.9205,
       "step": 20613
     },
     {
-      "epoch": 0.5661475927604296,
+      "epoch": 0.5849602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 8.351283475301314e-06,
-      "loss": 0.8102,
+      "learning_rate": 7.753803018545517e-06,
+      "loss": 0.8847,
       "step": 20614
     },
     {
-      "epoch": 0.5661750569882728,
+      "epoch": 0.5849886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 8.35040613881247e-06,
-      "loss": 0.8971,
+      "learning_rate": 7.752907435621968e-06,
+      "loss": 0.8136,
       "step": 20615
     },
     {
-      "epoch": 0.566202521216116,
+      "epoch": 0.5850170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 8.349528815375847e-06,
-      "loss": 0.8791,
+      "learning_rate": 7.752011871679712e-06,
+      "loss": 0.9323,
       "step": 20616
     },
     {
-      "epoch": 0.5662299854439592,
+      "epoch": 0.5850454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 8.348651504998391e-06,
-      "loss": 0.8141,
+      "learning_rate": 7.75111632672631e-06,
+      "loss": 0.8537,
       "step": 20617
     },
     {
-      "epoch": 0.5662574496718025,
+      "epoch": 0.5850737797956868,
       "grad_norm": 0.0,
-      "learning_rate": 8.347774207687045e-06,
-      "loss": 0.8332,
+      "learning_rate": 7.750220800769333e-06,
+      "loss": 0.8598,
       "step": 20618
     },
     {
-      "epoch": 0.5662849138996457,
+      "epoch": 0.5851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 8.346896923448747e-06,
-      "loss": 0.9175,
+      "learning_rate": 7.74932529381634e-06,
+      "loss": 0.8409,
       "step": 20619
     },
     {
-      "epoch": 0.5663123781274889,
+      "epoch": 0.5851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 8.346019652290443e-06,
-      "loss": 0.9509,
+      "learning_rate": 7.748429805874896e-06,
+      "loss": 0.7671,
       "step": 20620
     },
     {
-      "epoch": 0.5663398423553322,
+      "epoch": 0.5851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 8.345142394219077e-06,
-      "loss": 0.8903,
+      "learning_rate": 7.747534336952569e-06,
+      "loss": 0.9221,
       "step": 20621
     },
     {
-      "epoch": 0.5663673065831755,
+      "epoch": 0.5851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 8.344265149241578e-06,
-      "loss": 0.7326,
+      "learning_rate": 7.74663888705692e-06,
+      "loss": 0.8465,
       "step": 20622
     },
     {
-      "epoch": 0.5663947708110186,
+      "epoch": 0.5852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 8.343387917364896e-06,
-      "loss": 0.8512,
+      "learning_rate": 7.745743456195513e-06,
+      "loss": 0.941,
       "step": 20623
     },
     {
-      "epoch": 0.5664222350388619,
+      "epoch": 0.585244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 8.342510698595972e-06,
-      "loss": 0.859,
+      "learning_rate": 7.744848044375913e-06,
+      "loss": 0.8263,
       "step": 20624
     },
     {
-      "epoch": 0.5664496992667051,
+      "epoch": 0.585272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 8.341633492941744e-06,
-      "loss": 0.8493,
+      "learning_rate": 7.743952651605683e-06,
+      "loss": 0.7967,
       "step": 20625
     },
     {
-      "epoch": 0.5664771634945484,
+      "epoch": 0.5853007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 8.340756300409159e-06,
-      "loss": 0.7738,
+      "learning_rate": 7.743057277892385e-06,
+      "loss": 0.7754,
       "step": 20626
     },
     {
-      "epoch": 0.5665046277223916,
+      "epoch": 0.5853291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 8.339879121005147e-06,
-      "loss": 0.8496,
+      "learning_rate": 7.742161923243581e-06,
+      "loss": 0.9583,
       "step": 20627
     },
     {
-      "epoch": 0.5665320919502348,
+      "epoch": 0.5853575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 8.339001954736656e-06,
-      "loss": 0.9021,
+      "learning_rate": 7.741266587666841e-06,
+      "loss": 0.8561,
       "step": 20628
     },
     {
-      "epoch": 0.5665595561780781,
+      "epoch": 0.5853859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 8.338124801610627e-06,
-      "loss": 0.9241,
+      "learning_rate": 7.740371271169724e-06,
+      "loss": 0.9095,
       "step": 20629
     },
     {
-      "epoch": 0.5665870204059212,
+      "epoch": 0.5854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 8.337247661633995e-06,
-      "loss": 0.9291,
+      "learning_rate": 7.73947597375979e-06,
+      "loss": 0.8628,
       "step": 20630
     },
     {
-      "epoch": 0.5666144846337645,
+      "epoch": 0.5854426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 8.336370534813704e-06,
-      "loss": 0.8485,
+      "learning_rate": 7.738580695444607e-06,
+      "loss": 0.8704,
       "step": 20631
     },
     {
-      "epoch": 0.5666419488616078,
+      "epoch": 0.5854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 8.335493421156698e-06,
-      "loss": 0.8527,
+      "learning_rate": 7.737685436231734e-06,
+      "loss": 0.9052,
       "step": 20632
     },
     {
-      "epoch": 0.566669413089451,
+      "epoch": 0.5854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 8.334616320669911e-06,
-      "loss": 0.8864,
+      "learning_rate": 7.736790196128732e-06,
+      "loss": 0.7552,
       "step": 20633
     },
     {
-      "epoch": 0.5666968773172942,
+      "epoch": 0.585527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 8.333739233360284e-06,
-      "loss": 1.0225,
+      "learning_rate": 7.735894975143166e-06,
+      "loss": 0.807,
       "step": 20634
     },
     {
-      "epoch": 0.5667243415451375,
+      "epoch": 0.5855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 8.332862159234754e-06,
-      "loss": 0.8631,
+      "learning_rate": 7.734999773282597e-06,
+      "loss": 0.9009,
       "step": 20635
     },
     {
-      "epoch": 0.5667518057729807,
+      "epoch": 0.5855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 8.331985098300269e-06,
-      "loss": 0.9063,
+      "learning_rate": 7.734104590554588e-06,
+      "loss": 0.8668,
       "step": 20636
     },
     {
-      "epoch": 0.566779270000824,
+      "epoch": 0.5856129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 8.331108050563761e-06,
-      "loss": 0.8223,
+      "learning_rate": 7.733209426966697e-06,
+      "loss": 0.8829,
       "step": 20637
     },
     {
-      "epoch": 0.5668067342286671,
+      "epoch": 0.5856413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 8.330231016032179e-06,
-      "loss": 0.9003,
+      "learning_rate": 7.73231428252649e-06,
+      "loss": 0.8726,
       "step": 20638
     },
     {
-      "epoch": 0.5668341984565104,
+      "epoch": 0.5856696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 8.329353994712449e-06,
-      "loss": 0.9507,
+      "learning_rate": 7.731419157241525e-06,
+      "loss": 0.8631,
       "step": 20639
     },
     {
-      "epoch": 0.5668616626843537,
+      "epoch": 0.5856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 8.328476986611518e-06,
-      "loss": 0.869,
+      "learning_rate": 7.730524051119364e-06,
+      "loss": 0.924,
       "step": 20640
     },
     {
-      "epoch": 0.5668891269121968,
+      "epoch": 0.5857264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 8.327599991736328e-06,
-      "loss": 0.857,
+      "learning_rate": 7.729628964167568e-06,
+      "loss": 0.9215,
       "step": 20641
     },
     {
-      "epoch": 0.5669165911400401,
+      "epoch": 0.5857548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 8.326723010093811e-06,
-      "loss": 0.8516,
+      "learning_rate": 7.7287338963937e-06,
+      "loss": 0.9213,
       "step": 20642
     },
     {
-      "epoch": 0.5669440553678833,
+      "epoch": 0.585783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 8.32584604169091e-06,
-      "loss": 0.9224,
+      "learning_rate": 7.727838847805314e-06,
+      "loss": 0.8369,
       "step": 20643
     },
     {
-      "epoch": 0.5669715195957266,
+      "epoch": 0.5858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 8.32496908653457e-06,
-      "loss": 0.9525,
+      "learning_rate": 7.726943818409978e-06,
+      "loss": 0.8313,
       "step": 20644
     },
     {
-      "epoch": 0.5669989838235698,
+      "epoch": 0.5858399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 8.32409214463172e-06,
-      "loss": 0.8852,
+      "learning_rate": 7.726048808215252e-06,
+      "loss": 0.8493,
       "step": 20645
     },
     {
-      "epoch": 0.567026448051413,
+      "epoch": 0.5858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 8.323215215989298e-06,
-      "loss": 0.8907,
+      "learning_rate": 7.725153817228694e-06,
+      "loss": 0.9528,
       "step": 20646
     },
     {
-      "epoch": 0.5670539122792563,
+      "epoch": 0.5858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 8.322338300614248e-06,
-      "loss": 0.8797,
+      "learning_rate": 7.72425884545786e-06,
+      "loss": 0.9157,
       "step": 20647
     },
     {
-      "epoch": 0.5670813765070996,
+      "epoch": 0.5859250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 8.321461398513511e-06,
-      "loss": 0.8524,
+      "learning_rate": 7.723363892910319e-06,
+      "loss": 0.7974,
       "step": 20648
     },
     {
-      "epoch": 0.5671088407349427,
+      "epoch": 0.5859534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 8.320584509694018e-06,
-      "loss": 0.9656,
+      "learning_rate": 7.722468959593623e-06,
+      "loss": 0.8577,
       "step": 20649
     },
     {
-      "epoch": 0.567136304962786,
+      "epoch": 0.5859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 8.319707634162716e-06,
-      "loss": 1.0564,
+      "learning_rate": 7.721574045515334e-06,
+      "loss": 0.92,
       "step": 20650
     },
     {
-      "epoch": 0.5671637691906292,
+      "epoch": 0.5860102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 8.318830771926532e-06,
-      "loss": 0.9907,
+      "learning_rate": 7.720679150683012e-06,
+      "loss": 0.8323,
       "step": 20651
     },
     {
-      "epoch": 0.5671912334184724,
+      "epoch": 0.5860385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 8.317953922992413e-06,
-      "loss": 0.8405,
+      "learning_rate": 7.719784275104216e-06,
+      "loss": 0.8624,
       "step": 20652
     },
     {
-      "epoch": 0.5672186976463157,
+      "epoch": 0.586066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 8.317077087367294e-06,
-      "loss": 1.0048,
+      "learning_rate": 7.718889418786504e-06,
+      "loss": 0.8238,
       "step": 20653
     },
     {
-      "epoch": 0.5672461618741589,
+      "epoch": 0.5860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 8.31620026505811e-06,
-      "loss": 0.8041,
+      "learning_rate": 7.717994581737436e-06,
+      "loss": 0.759,
       "step": 20654
     },
     {
-      "epoch": 0.5672736261020022,
+      "epoch": 0.5861237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 8.315323456071805e-06,
-      "loss": 0.8717,
+      "learning_rate": 7.71709976396457e-06,
+      "loss": 0.9726,
       "step": 20655
     },
     {
-      "epoch": 0.5673010903298453,
+      "epoch": 0.5861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 8.314446660415314e-06,
-      "loss": 0.9148,
+      "learning_rate": 7.716204965475465e-06,
+      "loss": 0.7716,
       "step": 20656
     },
     {
-      "epoch": 0.5673285545576886,
+      "epoch": 0.5861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 8.313569878095575e-06,
-      "loss": 0.7775,
+      "learning_rate": 7.715310186277681e-06,
+      "loss": 0.845,
       "step": 20657
     },
     {
-      "epoch": 0.5673560187855319,
+      "epoch": 0.5862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 8.312693109119521e-06,
-      "loss": 0.8426,
+      "learning_rate": 7.714415426378767e-06,
+      "loss": 0.8265,
       "step": 20658
     },
     {
-      "epoch": 0.567383483013375,
+      "epoch": 0.5862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 8.311816353494094e-06,
-      "loss": 0.8818,
+      "learning_rate": 7.713520685786297e-06,
+      "loss": 0.7903,
       "step": 20659
     },
     {
-      "epoch": 0.5674109472412183,
+      "epoch": 0.5862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 8.310939611226229e-06,
-      "loss": 0.8782,
+      "learning_rate": 7.712625964507818e-06,
+      "loss": 0.9239,
       "step": 20660
     },
     {
-      "epoch": 0.5674384114690616,
+      "epoch": 0.586293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 8.310062882322865e-06,
-      "loss": 0.879,
+      "learning_rate": 7.71173126255089e-06,
+      "loss": 0.9889,
       "step": 20661
     },
     {
-      "epoch": 0.5674658756969048,
+      "epoch": 0.586322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 8.30918616679094e-06,
-      "loss": 0.9246,
+      "learning_rate": 7.710836579923072e-06,
+      "loss": 0.8296,
       "step": 20662
     },
     {
-      "epoch": 0.567493339924748,
+      "epoch": 0.5863507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 8.308309464637385e-06,
-      "loss": 0.8768,
+      "learning_rate": 7.70994191663192e-06,
+      "loss": 0.8961,
       "step": 20663
     },
     {
-      "epoch": 0.5675208041525912,
+      "epoch": 0.5863791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 8.307432775869142e-06,
-      "loss": 0.834,
+      "learning_rate": 7.70904727268499e-06,
+      "loss": 0.8578,
       "step": 20664
     },
     {
-      "epoch": 0.5675482683804345,
+      "epoch": 0.5864074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 8.306556100493148e-06,
-      "loss": 0.9346,
+      "learning_rate": 7.708152648089843e-06,
+      "loss": 0.7699,
       "step": 20665
     },
     {
-      "epoch": 0.5675757326082778,
+      "epoch": 0.5864358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 8.305679438516336e-06,
-      "loss": 0.875,
+      "learning_rate": 7.707258042854033e-06,
+      "loss": 0.8404,
       "step": 20666
     },
     {
-      "epoch": 0.5676031968361209,
+      "epoch": 0.5864642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 8.304802789945645e-06,
-      "loss": 0.9104,
+      "learning_rate": 7.706363456985115e-06,
+      "loss": 0.8472,
       "step": 20667
     },
     {
-      "epoch": 0.5676306610639642,
+      "epoch": 0.5864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 8.303926154788015e-06,
-      "loss": 0.8145,
+      "learning_rate": 7.70546889049065e-06,
+      "loss": 0.9245,
       "step": 20668
     },
     {
-      "epoch": 0.5676581252918074,
+      "epoch": 0.5865209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 8.303049533050377e-06,
-      "loss": 0.8554,
+      "learning_rate": 7.704574343378193e-06,
+      "loss": 0.882,
       "step": 20669
     },
     {
-      "epoch": 0.5676855895196506,
+      "epoch": 0.5865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 8.302172924739665e-06,
-      "loss": 0.8486,
+      "learning_rate": 7.703679815655297e-06,
+      "loss": 0.9363,
       "step": 20670
     },
     {
-      "epoch": 0.5677130537474939,
+      "epoch": 0.586577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 8.301296329862821e-06,
-      "loss": 0.8319,
+      "learning_rate": 7.702785307329526e-06,
+      "loss": 0.9634,
       "step": 20671
     },
     {
-      "epoch": 0.5677405179753371,
+      "epoch": 0.586606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 8.300419748426777e-06,
-      "loss": 0.9078,
+      "learning_rate": 7.701890818408427e-06,
+      "loss": 0.7543,
       "step": 20672
     },
     {
-      "epoch": 0.5677679822031804,
+      "epoch": 0.5866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 8.29954318043847e-06,
-      "loss": 0.8418,
+      "learning_rate": 7.700996348899558e-06,
+      "loss": 0.7591,
       "step": 20673
     },
     {
-      "epoch": 0.5677954464310236,
+      "epoch": 0.5866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 8.29866662590484e-06,
-      "loss": 0.897,
+      "learning_rate": 7.700101898810476e-06,
+      "loss": 0.8389,
       "step": 20674
     },
     {
-      "epoch": 0.5678229106588668,
+      "epoch": 0.5866912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 8.297790084832814e-06,
-      "loss": 0.9007,
+      "learning_rate": 7.69920746814874e-06,
+      "loss": 0.8237,
       "step": 20675
     },
     {
-      "epoch": 0.5678503748867101,
+      "epoch": 0.5867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 8.296913557229334e-06,
-      "loss": 0.8838,
+      "learning_rate": 7.698313056921903e-06,
+      "loss": 0.8579,
       "step": 20676
     },
     {
-      "epoch": 0.5678778391145533,
+      "epoch": 0.5867480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 8.296037043101334e-06,
-      "loss": 0.9456,
+      "learning_rate": 7.697418665137517e-06,
+      "loss": 0.81,
       "step": 20677
     },
     {
-      "epoch": 0.5679053033423965,
+      "epoch": 0.5867763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 8.295160542455746e-06,
-      "loss": 0.9318,
+      "learning_rate": 7.696524292803138e-06,
+      "loss": 0.8797,
       "step": 20678
     },
     {
-      "epoch": 0.5679327675702398,
+      "epoch": 0.5868047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 8.29428405529951e-06,
-      "loss": 0.9222,
+      "learning_rate": 7.695629939926323e-06,
+      "loss": 0.8975,
       "step": 20679
     },
     {
-      "epoch": 0.567960231798083,
+      "epoch": 0.58683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 8.293407581639563e-06,
-      "loss": 0.828,
+      "learning_rate": 7.694735606514627e-06,
+      "loss": 0.7729,
       "step": 20680
     },
     {
-      "epoch": 0.5679876960259262,
+      "epoch": 0.5868615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 8.292531121482835e-06,
-      "loss": 0.8626,
+      "learning_rate": 7.6938412925756e-06,
+      "loss": 0.7555,
       "step": 20681
     },
     {
-      "epoch": 0.5680151602537694,
+      "epoch": 0.5868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 8.29165467483626e-06,
-      "loss": 0.8039,
+      "learning_rate": 7.692946998116801e-06,
+      "loss": 0.8355,
       "step": 20682
     },
     {
-      "epoch": 0.5680426244816127,
+      "epoch": 0.5869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 8.290778241706772e-06,
-      "loss": 0.838,
+      "learning_rate": 7.692052723145782e-06,
+      "loss": 0.874,
       "step": 20683
     },
     {
-      "epoch": 0.568070088709456,
+      "epoch": 0.5869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 8.289901822101315e-06,
-      "loss": 0.8973,
+      "learning_rate": 7.691158467670097e-06,
+      "loss": 0.8413,
       "step": 20684
     },
     {
-      "epoch": 0.5680975529372991,
+      "epoch": 0.5869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 8.289025416026817e-06,
-      "loss": 0.7977,
+      "learning_rate": 7.6902642316973e-06,
+      "loss": 0.7984,
       "step": 20685
     },
     {
-      "epoch": 0.5681250171651424,
+      "epoch": 0.5870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 8.28814902349021e-06,
-      "loss": 0.827,
+      "learning_rate": 7.689370015234947e-06,
+      "loss": 0.7054,
       "step": 20686
     },
     {
-      "epoch": 0.5681524813929857,
+      "epoch": 0.5870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 8.287272644498428e-06,
-      "loss": 0.7719,
+      "learning_rate": 7.688475818290588e-06,
+      "loss": 0.8509,
       "step": 20687
     },
     {
-      "epoch": 0.5681799456208289,
+      "epoch": 0.5870601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 8.286396279058411e-06,
-      "loss": 0.877,
+      "learning_rate": 7.687581640871778e-06,
+      "loss": 0.8884,
       "step": 20688
     },
     {
-      "epoch": 0.5682074098486721,
+      "epoch": 0.5870885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 8.28551992717709e-06,
-      "loss": 0.7612,
+      "learning_rate": 7.686687482986066e-06,
+      "loss": 0.8278,
       "step": 20689
     },
     {
-      "epoch": 0.5682348740765153,
+      "epoch": 0.587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 8.284643588861399e-06,
-      "loss": 0.8575,
+      "learning_rate": 7.685793344641012e-06,
+      "loss": 0.7854,
       "step": 20690
     },
     {
-      "epoch": 0.5682623383043586,
+      "epoch": 0.5871452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 8.283767264118278e-06,
-      "loss": 0.8668,
+      "learning_rate": 7.684899225844165e-06,
+      "loss": 0.8812,
       "step": 20691
     },
     {
-      "epoch": 0.5682898025322018,
+      "epoch": 0.5871736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 8.282890952954649e-06,
-      "loss": 0.8268,
+      "learning_rate": 7.684005126603079e-06,
+      "loss": 0.8461,
       "step": 20692
     },
     {
-      "epoch": 0.568317266760045,
+      "epoch": 0.5872020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 8.282014655377455e-06,
-      "loss": 0.791,
+      "learning_rate": 7.683111046925306e-06,
+      "loss": 0.7788,
       "step": 20693
     },
     {
-      "epoch": 0.5683447309878883,
+      "epoch": 0.5872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 8.281138371393624e-06,
-      "loss": 0.9145,
+      "learning_rate": 7.682216986818398e-06,
+      "loss": 0.892,
       "step": 20694
     },
     {
-      "epoch": 0.5683721952157315,
+      "epoch": 0.5872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 8.280262101010092e-06,
-      "loss": 0.8114,
+      "learning_rate": 7.681322946289909e-06,
+      "loss": 0.8289,
       "step": 20695
     },
     {
-      "epoch": 0.5683996594435747,
+      "epoch": 0.5872871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 8.279385844233794e-06,
-      "loss": 0.8322,
+      "learning_rate": 7.680428925347388e-06,
+      "loss": 0.8925,
       "step": 20696
     },
     {
-      "epoch": 0.568427123671418,
+      "epoch": 0.5873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 8.278509601071664e-06,
-      "loss": 0.799,
+      "learning_rate": 7.679534923998386e-06,
+      "loss": 0.9033,
       "step": 20697
     },
     {
-      "epoch": 0.5684545878992612,
+      "epoch": 0.5873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 8.277633371530628e-06,
-      "loss": 0.9018,
+      "learning_rate": 7.67864094225046e-06,
+      "loss": 0.8646,
       "step": 20698
     },
     {
-      "epoch": 0.5684820521271045,
+      "epoch": 0.5873723041997729,
       "grad_norm": 0.0,
-      "learning_rate": 8.276757155617626e-06,
-      "loss": 0.8985,
+      "learning_rate": 7.677746980111158e-06,
+      "loss": 1.0038,
       "step": 20699
     },
     {
-      "epoch": 0.5685095163549477,
+      "epoch": 0.5874006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 8.275880953339587e-06,
-      "loss": 0.8665,
+      "learning_rate": 7.676853037588028e-06,
+      "loss": 0.8421,
       "step": 20700
     },
     {
-      "epoch": 0.5685369805827909,
+      "epoch": 0.5874290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 8.275004764703451e-06,
-      "loss": 0.936,
+      "learning_rate": 7.675959114688629e-06,
+      "loss": 0.9631,
       "step": 20701
     },
     {
-      "epoch": 0.5685644448106342,
+      "epoch": 0.5874574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 8.274128589716141e-06,
-      "loss": 0.8543,
+      "learning_rate": 7.675065211420508e-06,
+      "loss": 0.8925,
       "step": 20702
     },
     {
-      "epoch": 0.5685919090384773,
+      "epoch": 0.5874858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 8.273252428384599e-06,
-      "loss": 0.7081,
+      "learning_rate": 7.674171327791213e-06,
+      "loss": 0.8505,
       "step": 20703
     },
     {
-      "epoch": 0.5686193732663206,
+      "epoch": 0.5875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 8.272376280715746e-06,
-      "loss": 0.8334,
+      "learning_rate": 7.673277463808295e-06,
+      "loss": 0.9533,
       "step": 20704
     },
     {
-      "epoch": 0.5686468374941639,
+      "epoch": 0.5875425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 8.271500146716525e-06,
-      "loss": 0.8967,
+      "learning_rate": 7.67238361947931e-06,
+      "loss": 0.9582,
       "step": 20705
     },
     {
-      "epoch": 0.5686743017220071,
+      "epoch": 0.5875709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 8.270624026393862e-06,
-      "loss": 0.7998,
+      "learning_rate": 7.671489794811805e-06,
+      "loss": 0.9002,
       "step": 20706
     },
     {
-      "epoch": 0.5687017659498503,
+      "epoch": 0.5875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 8.269747919754693e-06,
-      "loss": 0.9005,
+      "learning_rate": 7.670595989813333e-06,
+      "loss": 1.0194,
       "step": 20707
     },
     {
-      "epoch": 0.5687292301776935,
+      "epoch": 0.587627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 8.268871826805952e-06,
-      "loss": 0.832,
+      "learning_rate": 7.669702204491435e-06,
+      "loss": 0.7722,
       "step": 20708
     },
     {
-      "epoch": 0.5687566944055368,
+      "epoch": 0.5876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 8.267995747554566e-06,
-      "loss": 0.9281,
+      "learning_rate": 7.668808438853674e-06,
+      "loss": 0.8944,
       "step": 20709
     },
     {
-      "epoch": 0.56878415863338,
+      "epoch": 0.5876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 8.267119682007466e-06,
-      "loss": 0.839,
+      "learning_rate": 7.667914692907591e-06,
+      "loss": 0.9013,
       "step": 20710
     },
     {
-      "epoch": 0.5688116228612232,
+      "epoch": 0.5877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 8.266243630171586e-06,
-      "loss": 0.8857,
+      "learning_rate": 7.667020966660736e-06,
+      "loss": 0.7423,
       "step": 20711
     },
     {
-      "epoch": 0.5688390870890665,
+      "epoch": 0.5877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 8.265367592053862e-06,
-      "loss": 0.8611,
+      "learning_rate": 7.66612726012066e-06,
+      "loss": 0.8067,
       "step": 20712
     },
     {
-      "epoch": 0.5688665513169098,
+      "epoch": 0.5877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 8.264491567661217e-06,
-      "loss": 0.9769,
+      "learning_rate": 7.665233573294914e-06,
+      "loss": 0.9231,
       "step": 20713
     },
     {
-      "epoch": 0.5688940155447529,
+      "epoch": 0.5877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 8.263615557000587e-06,
-      "loss": 0.8716,
+      "learning_rate": 7.664339906191042e-06,
+      "loss": 0.9174,
       "step": 20714
     },
     {
-      "epoch": 0.5689214797725962,
+      "epoch": 0.5878263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 8.262739560078908e-06,
-      "loss": 0.9249,
+      "learning_rate": 7.663446258816598e-06,
+      "loss": 0.7735,
       "step": 20715
     },
     {
-      "epoch": 0.5689489440004394,
+      "epoch": 0.5878547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 8.261863576903104e-06,
-      "loss": 0.8838,
+      "learning_rate": 7.662552631179127e-06,
+      "loss": 0.8355,
       "step": 20716
     },
     {
-      "epoch": 0.5689764082282827,
+      "epoch": 0.5878830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 8.260987607480109e-06,
-      "loss": 0.8503,
+      "learning_rate": 7.661659023286179e-06,
+      "loss": 0.9005,
       "step": 20717
     },
     {
-      "epoch": 0.5690038724561259,
+      "epoch": 0.5879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 8.260111651816853e-06,
-      "loss": 0.8832,
+      "learning_rate": 7.660765435145305e-06,
+      "loss": 0.8443,
       "step": 20718
     },
     {
-      "epoch": 0.5690313366839691,
+      "epoch": 0.5879398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 8.259235709920266e-06,
-      "loss": 0.8645,
+      "learning_rate": 7.659871866764044e-06,
+      "loss": 0.9252,
       "step": 20719
     },
     {
-      "epoch": 0.5690588009118124,
+      "epoch": 0.5879682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 8.258359781797283e-06,
-      "loss": 0.8541,
+      "learning_rate": 7.658978318149956e-06,
+      "loss": 0.7971,
       "step": 20720
     },
     {
-      "epoch": 0.5690862651396555,
+      "epoch": 0.5879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 8.257483867454834e-06,
-      "loss": 0.9125,
+      "learning_rate": 7.658084789310584e-06,
+      "loss": 0.8168,
       "step": 20721
     },
     {
-      "epoch": 0.5691137293674988,
+      "epoch": 0.5880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 8.256607966899843e-06,
-      "loss": 0.9313,
+      "learning_rate": 7.65719128025347e-06,
+      "loss": 0.8999,
       "step": 20722
     },
     {
-      "epoch": 0.5691411935953421,
+      "epoch": 0.5880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 8.255732080139248e-06,
-      "loss": 0.7949,
+      "learning_rate": 7.656297790986173e-06,
+      "loss": 0.7923,
       "step": 20723
     },
     {
-      "epoch": 0.5691686578231853,
+      "epoch": 0.5880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 8.254856207179975e-06,
-      "loss": 0.7619,
+      "learning_rate": 7.655404321516231e-06,
+      "loss": 0.6993,
       "step": 20724
     },
     {
-      "epoch": 0.5691961220510285,
+      "epoch": 0.5881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 8.253980348028955e-06,
-      "loss": 0.8132,
+      "learning_rate": 7.654510871851192e-06,
+      "loss": 0.9496,
       "step": 20725
     },
     {
-      "epoch": 0.5692235862788718,
+      "epoch": 0.588138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 8.253104502693119e-06,
-      "loss": 0.8825,
+      "learning_rate": 7.65361744199861e-06,
+      "loss": 1.0424,
       "step": 20726
     },
     {
-      "epoch": 0.569251050506715,
+      "epoch": 0.58816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.2522286711794e-06,
-      "loss": 0.9207,
+      "learning_rate": 7.652724031966023e-06,
+      "loss": 0.8684,
       "step": 20727
     },
     {
-      "epoch": 0.5692785147345583,
+      "epoch": 0.5881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 8.251352853494722e-06,
-      "loss": 0.879,
+      "learning_rate": 7.651830641760983e-06,
+      "loss": 0.833,
       "step": 20728
     },
     {
-      "epoch": 0.5693059789624014,
+      "epoch": 0.5882236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 8.250477049646018e-06,
-      "loss": 0.8978,
+      "learning_rate": 7.650937271391035e-06,
+      "loss": 0.7294,
       "step": 20729
     },
     {
-      "epoch": 0.5693334431902447,
+      "epoch": 0.5882519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 8.249601259640216e-06,
-      "loss": 0.8133,
+      "learning_rate": 7.650043920863726e-06,
+      "loss": 0.915,
       "step": 20730
     },
     {
-      "epoch": 0.569360907418088,
+      "epoch": 0.5882803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 8.248725483484248e-06,
-      "loss": 0.835,
+      "learning_rate": 7.6491505901866e-06,
+      "loss": 0.8556,
       "step": 20731
     },
     {
-      "epoch": 0.5693883716459311,
+      "epoch": 0.5883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 8.247849721185044e-06,
-      "loss": 0.78,
+      "learning_rate": 7.648257279367206e-06,
+      "loss": 0.894,
       "step": 20732
     },
     {
-      "epoch": 0.5694158358737744,
+      "epoch": 0.5883371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 8.246973972749534e-06,
-      "loss": 0.8329,
+      "learning_rate": 7.64736398841309e-06,
+      "loss": 0.8428,
       "step": 20733
     },
     {
-      "epoch": 0.5694433001016176,
+      "epoch": 0.5883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 8.246098238184638e-06,
-      "loss": 0.9347,
+      "learning_rate": 7.646470717331791e-06,
+      "loss": 0.9013,
       "step": 20734
     },
     {
-      "epoch": 0.5694707643294609,
+      "epoch": 0.5883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 8.245222517497295e-06,
-      "loss": 0.7427,
+      "learning_rate": 7.645577466130861e-06,
+      "loss": 0.7911,
       "step": 20735
     },
     {
-      "epoch": 0.5694982285573041,
+      "epoch": 0.588422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 8.244346810694433e-06,
-      "loss": 0.8724,
+      "learning_rate": 7.644684234817846e-06,
+      "loss": 0.833,
       "step": 20736
     },
     {
-      "epoch": 0.5695256927851473,
+      "epoch": 0.5884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 8.243471117782978e-06,
-      "loss": 0.7946,
+      "learning_rate": 7.64379102340029e-06,
+      "loss": 0.7298,
       "step": 20737
     },
     {
-      "epoch": 0.5695531570129906,
+      "epoch": 0.5884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 8.242595438769857e-06,
-      "loss": 0.8222,
+      "learning_rate": 7.642897831885734e-06,
+      "loss": 0.7033,
       "step": 20738
     },
     {
-      "epoch": 0.5695806212408339,
+      "epoch": 0.5885073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 8.241719773662008e-06,
-      "loss": 0.9179,
+      "learning_rate": 7.64200466028173e-06,
+      "loss": 0.8706,
       "step": 20739
     },
     {
-      "epoch": 0.569608085468677,
+      "epoch": 0.5885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 8.240844122466349e-06,
-      "loss": 0.7861,
+      "learning_rate": 7.641111508595818e-06,
+      "loss": 0.8864,
       "step": 20740
     },
     {
-      "epoch": 0.5696355496965203,
+      "epoch": 0.5885641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 8.239968485189813e-06,
-      "loss": 0.8238,
+      "learning_rate": 7.64021837683554e-06,
+      "loss": 0.8554,
       "step": 20741
     },
     {
-      "epoch": 0.5696630139243635,
+      "epoch": 0.5885925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 8.239092861839327e-06,
-      "loss": 0.898,
+      "learning_rate": 7.639325265008445e-06,
+      "loss": 0.9426,
       "step": 20742
     },
     {
-      "epoch": 0.5696904781522067,
+      "epoch": 0.5886208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 8.23821725242182e-06,
-      "loss": 0.8325,
+      "learning_rate": 7.638432173122078e-06,
+      "loss": 0.8846,
       "step": 20743
     },
     {
-      "epoch": 0.56971794238005,
+      "epoch": 0.5886492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 8.237341656944226e-06,
-      "loss": 0.9467,
+      "learning_rate": 7.63753910118398e-06,
+      "loss": 0.9348,
       "step": 20744
     },
     {
-      "epoch": 0.5697454066078932,
+      "epoch": 0.588677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 8.236466075413464e-06,
-      "loss": 0.8174,
+      "learning_rate": 7.636646049201692e-06,
+      "loss": 0.8136,
       "step": 20745
     },
     {
-      "epoch": 0.5697728708357365,
+      "epoch": 0.5887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 8.235590507836466e-06,
-      "loss": 0.833,
+      "learning_rate": 7.635753017182763e-06,
+      "loss": 0.8117,
       "step": 20746
     },
     {
-      "epoch": 0.5698003350635796,
+      "epoch": 0.5887343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 8.234714954220158e-06,
-      "loss": 0.7692,
+      "learning_rate": 7.634860005134737e-06,
+      "loss": 0.9208,
       "step": 20747
     },
     {
-      "epoch": 0.5698277992914229,
+      "epoch": 0.5887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 8.233839414571471e-06,
-      "loss": 0.8609,
+      "learning_rate": 7.63396701306515e-06,
+      "loss": 0.8945,
       "step": 20748
     },
     {
-      "epoch": 0.5698552635192662,
+      "epoch": 0.5887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 8.232963888897328e-06,
-      "loss": 0.8341,
+      "learning_rate": 7.633074040981553e-06,
+      "loss": 0.8572,
       "step": 20749
     },
     {
-      "epoch": 0.5698827277471094,
+      "epoch": 0.5888195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 8.232088377204665e-06,
-      "loss": 0.821,
+      "learning_rate": 7.632181088891482e-06,
+      "loss": 0.9326,
       "step": 20750
     },
     {
-      "epoch": 0.5699101919749526,
+      "epoch": 0.5888479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 8.231212879500397e-06,
-      "loss": 0.8228,
+      "learning_rate": 7.631288156802488e-06,
+      "loss": 0.9252,
       "step": 20751
     },
     {
-      "epoch": 0.5699376562027959,
+      "epoch": 0.5888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 8.230337395791458e-06,
-      "loss": 0.8208,
+      "learning_rate": 7.630395244722108e-06,
+      "loss": 0.8336,
       "step": 20752
     },
     {
-      "epoch": 0.5699651204306391,
+      "epoch": 0.5889046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 8.22946192608478e-06,
-      "loss": 0.808,
+      "learning_rate": 7.629502352657888e-06,
+      "loss": 0.8332,
       "step": 20753
     },
     {
-      "epoch": 0.5699925846584823,
+      "epoch": 0.5889330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 8.22858647038728e-06,
-      "loss": 0.8845,
+      "learning_rate": 7.628609480617366e-06,
+      "loss": 0.7161,
       "step": 20754
     },
     {
-      "epoch": 0.5700200488863255,
+      "epoch": 0.588961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 8.227711028705892e-06,
-      "loss": 0.9153,
+      "learning_rate": 7.627716628608086e-06,
+      "loss": 0.8579,
       "step": 20755
     },
     {
-      "epoch": 0.5700475131141688,
+      "epoch": 0.5889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 8.226835601047545e-06,
-      "loss": 0.8878,
+      "learning_rate": 7.626823796637592e-06,
+      "loss": 0.9052,
       "step": 20756
     },
     {
-      "epoch": 0.5700749773420121,
+      "epoch": 0.5890181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 8.225960187419158e-06,
-      "loss": 0.7847,
+      "learning_rate": 7.625930984713424e-06,
+      "loss": 0.7452,
       "step": 20757
     },
     {
-      "epoch": 0.5701024415698552,
+      "epoch": 0.5890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 8.22508478782766e-06,
-      "loss": 0.8324,
+      "learning_rate": 7.625038192843122e-06,
+      "loss": 0.781,
       "step": 20758
     },
     {
-      "epoch": 0.5701299057976985,
+      "epoch": 0.5890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 8.224209402279978e-06,
-      "loss": 0.8266,
+      "learning_rate": 7.624145421034231e-06,
+      "loss": 0.8607,
       "step": 20759
     },
     {
-      "epoch": 0.5701573700255417,
+      "epoch": 0.5891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 8.223334030783042e-06,
-      "loss": 0.9157,
+      "learning_rate": 7.623252669294291e-06,
+      "loss": 0.9167,
       "step": 20760
     },
     {
-      "epoch": 0.570184834253385,
+      "epoch": 0.5891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.222458673343774e-06,
-      "loss": 0.8121,
+      "learning_rate": 7.622359937630839e-06,
+      "loss": 0.8612,
       "step": 20761
     },
     {
-      "epoch": 0.5702122984812282,
+      "epoch": 0.5891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 8.221583329969106e-06,
-      "loss": 0.8486,
+      "learning_rate": 7.621467226051422e-06,
+      "loss": 0.8431,
       "step": 20762
     },
     {
-      "epoch": 0.5702397627090714,
+      "epoch": 0.589188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 8.220708000665952e-06,
-      "loss": 0.7916,
+      "learning_rate": 7.62057453456358e-06,
+      "loss": 0.7938,
       "step": 20763
     },
     {
-      "epoch": 0.5702672269369147,
+      "epoch": 0.589216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 8.219832685441248e-06,
-      "loss": 0.8436,
+      "learning_rate": 7.619681863174848e-06,
+      "loss": 0.8831,
       "step": 20764
     },
     {
-      "epoch": 0.570294691164758,
+      "epoch": 0.5892451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 8.21895738430192e-06,
-      "loss": 0.8522,
+      "learning_rate": 7.618789211892771e-06,
+      "loss": 0.9523,
       "step": 20765
     },
     {
-      "epoch": 0.5703221553926011,
+      "epoch": 0.5892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 8.218082097254887e-06,
-      "loss": 0.8604,
+      "learning_rate": 7.6178965807248885e-06,
+      "loss": 0.9559,
       "step": 20766
     },
     {
-      "epoch": 0.5703496196204444,
+      "epoch": 0.5893019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 8.21720682430708e-06,
-      "loss": 0.9364,
+      "learning_rate": 7.617003969678742e-06,
+      "loss": 0.948,
       "step": 20767
     },
     {
-      "epoch": 0.5703770838482876,
+      "epoch": 0.5893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 8.216331565465427e-06,
-      "loss": 0.9323,
+      "learning_rate": 7.616111378761872e-06,
+      "loss": 0.7966,
       "step": 20768
     },
     {
-      "epoch": 0.5704045480761308,
+      "epoch": 0.5893586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 8.21545632073685e-06,
-      "loss": 0.8499,
+      "learning_rate": 7.6152188079818125e-06,
+      "loss": 0.7903,
       "step": 20769
     },
     {
-      "epoch": 0.5704320123039741,
+      "epoch": 0.5893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 8.214581090128269e-06,
-      "loss": 1.0087,
+      "learning_rate": 7.6143262573461095e-06,
+      "loss": 0.8867,
       "step": 20770
     },
     {
-      "epoch": 0.5704594765318173,
+      "epoch": 0.5894154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 8.213705873646613e-06,
-      "loss": 0.7585,
+      "learning_rate": 7.613433726862299e-06,
+      "loss": 0.8854,
       "step": 20771
     },
     {
-      "epoch": 0.5704869407596606,
+      "epoch": 0.5894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 8.212830671298812e-06,
-      "loss": 0.8024,
+      "learning_rate": 7.612541216537921e-06,
+      "loss": 0.911,
       "step": 20772
     },
     {
-      "epoch": 0.5705144049875038,
+      "epoch": 0.589472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 8.211955483091784e-06,
-      "loss": 0.9322,
+      "learning_rate": 7.611648726380515e-06,
+      "loss": 0.9291,
       "step": 20773
     },
     {
-      "epoch": 0.570541869215347,
+      "epoch": 0.58950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 8.211080309032461e-06,
-      "loss": 0.831,
+      "learning_rate": 7.61075625639762e-06,
+      "loss": 0.8098,
       "step": 20774
     },
     {
-      "epoch": 0.5705693334431903,
+      "epoch": 0.5895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 8.210205149127762e-06,
-      "loss": 0.8458,
+      "learning_rate": 7.609863806596772e-06,
+      "loss": 0.8321,
       "step": 20775
     },
     {
-      "epoch": 0.5705967976710334,
+      "epoch": 0.5895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 8.20933000338461e-06,
-      "loss": 0.843,
+      "learning_rate": 7.608971376985514e-06,
+      "loss": 0.805,
       "step": 20776
     },
     {
-      "epoch": 0.5706242618988767,
+      "epoch": 0.5895856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 8.208454871809937e-06,
-      "loss": 0.932,
+      "learning_rate": 7.608078967571384e-06,
+      "loss": 0.8267,
       "step": 20777
     },
     {
-      "epoch": 0.57065172612672,
+      "epoch": 0.5896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 8.20757975441066e-06,
-      "loss": 1.0029,
+      "learning_rate": 7.607186578361915e-06,
+      "loss": 0.8344,
       "step": 20778
     },
     {
-      "epoch": 0.5706791903545632,
+      "epoch": 0.5896424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 8.206704651193705e-06,
-      "loss": 0.9165,
+      "learning_rate": 7.606294209364651e-06,
+      "loss": 0.9921,
       "step": 20779
     },
     {
-      "epoch": 0.5707066545824064,
+      "epoch": 0.5896708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 8.205829562166004e-06,
-      "loss": 1.0023,
+      "learning_rate": 7.6054018605871264e-06,
+      "loss": 0.7673,
       "step": 20780
     },
     {
-      "epoch": 0.5707341188102496,
+      "epoch": 0.5896992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 8.20495448733447e-06,
-      "loss": 0.7441,
+      "learning_rate": 7.6045095320368765e-06,
+      "loss": 0.8495,
       "step": 20781
     },
     {
-      "epoch": 0.5707615830380929,
+      "epoch": 0.589727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 8.204079426706031e-06,
-      "loss": 0.8525,
+      "learning_rate": 7.603617223721445e-06,
+      "loss": 0.8515,
       "step": 20782
     },
     {
-      "epoch": 0.5707890472659362,
+      "epoch": 0.5897559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 8.20320438028761e-06,
-      "loss": 0.8297,
+      "learning_rate": 7.602724935648367e-06,
+      "loss": 0.937,
       "step": 20783
     },
     {
-      "epoch": 0.5708165114937793,
+      "epoch": 0.5897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 8.202329348086135e-06,
-      "loss": 0.8911,
+      "learning_rate": 7.601832667825179e-06,
+      "loss": 0.8857,
       "step": 20784
     },
     {
-      "epoch": 0.5708439757216226,
+      "epoch": 0.5898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 8.201454330108525e-06,
-      "loss": 0.7997,
+      "learning_rate": 7.600940420259421e-06,
+      "loss": 0.9161,
       "step": 20785
     },
     {
-      "epoch": 0.5708714399494659,
+      "epoch": 0.5898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 8.200579326361708e-06,
-      "loss": 0.8332,
+      "learning_rate": 7.600048192958622e-06,
+      "loss": 0.9274,
       "step": 20786
     },
     {
-      "epoch": 0.570898904177309,
+      "epoch": 0.5898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 8.1997043368526e-06,
-      "loss": 0.8576,
+      "learning_rate": 7.599155985930328e-06,
+      "loss": 0.7952,
       "step": 20787
     },
     {
-      "epoch": 0.5709263684051523,
+      "epoch": 0.5898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 8.198829361588129e-06,
-      "loss": 0.7566,
+      "learning_rate": 7.598263799182071e-06,
+      "loss": 0.8862,
       "step": 20788
     },
     {
-      "epoch": 0.5709538326329955,
+      "epoch": 0.5899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 8.19795440057522e-06,
-      "loss": 0.8659,
+      "learning_rate": 7.597371632721384e-06,
+      "loss": 0.8998,
       "step": 20789
     },
     {
-      "epoch": 0.5709812968608388,
+      "epoch": 0.5899545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 8.197079453820792e-06,
-      "loss": 0.8279,
+      "learning_rate": 7.5964794865558115e-06,
+      "loss": 0.9453,
       "step": 20790
     },
     {
-      "epoch": 0.571008761088682,
+      "epoch": 0.5899829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 8.196204521331771e-06,
-      "loss": 0.8572,
+      "learning_rate": 7.595587360692883e-06,
+      "loss": 0.8091,
       "step": 20791
     },
     {
-      "epoch": 0.5710362253165252,
+      "epoch": 0.590011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 8.195329603115082e-06,
-      "loss": 0.8185,
+      "learning_rate": 7.5946952551401346e-06,
+      "loss": 0.8695,
       "step": 20792
     },
     {
-      "epoch": 0.5710636895443685,
+      "epoch": 0.5900397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 8.19445469917764e-06,
-      "loss": 0.8668,
+      "learning_rate": 7.5938031699051055e-06,
+      "loss": 0.8129,
       "step": 20793
     },
     {
-      "epoch": 0.5710911537722116,
+      "epoch": 0.5900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 8.193579809526372e-06,
-      "loss": 0.828,
+      "learning_rate": 7.592911104995328e-06,
+      "loss": 0.7762,
       "step": 20794
     },
     {
-      "epoch": 0.5711186180000549,
+      "epoch": 0.5900964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 8.1927049341682e-06,
-      "loss": 0.8198,
+      "learning_rate": 7.592019060418339e-06,
+      "loss": 0.839,
       "step": 20795
     },
     {
-      "epoch": 0.5711460822278982,
+      "epoch": 0.5901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 8.191830073110048e-06,
-      "loss": 0.8311,
+      "learning_rate": 7.591127036181669e-06,
+      "loss": 0.9293,
       "step": 20796
     },
     {
-      "epoch": 0.5711735464557414,
+      "epoch": 0.5901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 8.190955226358835e-06,
-      "loss": 0.8393,
+      "learning_rate": 7.590235032292863e-06,
+      "loss": 0.7899,
       "step": 20797
     },
     {
-      "epoch": 0.5712010106835846,
+      "epoch": 0.5901816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 8.19008039392149e-06,
-      "loss": 0.8362,
+      "learning_rate": 7.589343048759449e-06,
+      "loss": 0.7667,
       "step": 20798
     },
     {
-      "epoch": 0.5712284749114279,
+      "epoch": 0.5902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 8.189205575804925e-06,
-      "loss": 0.7627,
+      "learning_rate": 7.58845108558896e-06,
+      "loss": 0.7715,
       "step": 20799
     },
     {
-      "epoch": 0.5712559391392711,
+      "epoch": 0.5902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 8.188330772016067e-06,
-      "loss": 0.805,
+      "learning_rate": 7.587559142788936e-06,
+      "loss": 0.8532,
       "step": 20800
     },
     {
-      "epoch": 0.5712834033671144,
+      "epoch": 0.5902667423382519,
       "grad_norm": 0.0,
-      "learning_rate": 8.18745598256184e-06,
-      "loss": 0.9187,
+      "learning_rate": 7.586667220366909e-06,
+      "loss": 0.8395,
       "step": 20801
     },
     {
-      "epoch": 0.5713108675949575,
+      "epoch": 0.5902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 8.186581207449162e-06,
-      "loss": 0.8043,
+      "learning_rate": 7.58577531833041e-06,
+      "loss": 0.9548,
       "step": 20802
     },
     {
-      "epoch": 0.5713383318228008,
+      "epoch": 0.5903234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 8.18570644668496e-06,
-      "loss": 0.8394,
+      "learning_rate": 7.584883436686976e-06,
+      "loss": 0.8558,
       "step": 20803
     },
     {
-      "epoch": 0.5713657960506441,
+      "epoch": 0.5903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 8.184831700276146e-06,
-      "loss": 0.7984,
+      "learning_rate": 7.583991575444143e-06,
+      "loss": 0.7777,
       "step": 20804
     },
     {
-      "epoch": 0.5713932602784872,
+      "epoch": 0.5903802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 8.183956968229648e-06,
-      "loss": 0.7907,
+      "learning_rate": 7.583099734609437e-06,
+      "loss": 0.957,
       "step": 20805
     },
     {
-      "epoch": 0.5714207245063305,
+      "epoch": 0.5904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 8.183082250552385e-06,
-      "loss": 0.9002,
+      "learning_rate": 7.582207914190399e-06,
+      "loss": 0.773,
       "step": 20806
     },
     {
-      "epoch": 0.5714481887341737,
+      "epoch": 0.5904370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 8.182207547251279e-06,
-      "loss": 0.7784,
+      "learning_rate": 7.5813161141945605e-06,
+      "loss": 0.7786,
       "step": 20807
     },
     {
-      "epoch": 0.571475652962017,
+      "epoch": 0.5904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 8.181332858333254e-06,
-      "loss": 0.9156,
+      "learning_rate": 7.58042433462945e-06,
+      "loss": 0.9084,
       "step": 20808
     },
     {
-      "epoch": 0.5715031171898602,
+      "epoch": 0.5904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 8.180458183805227e-06,
-      "loss": 0.8922,
+      "learning_rate": 7.5795325755026075e-06,
+      "loss": 0.8975,
       "step": 20809
     },
     {
-      "epoch": 0.5715305814177034,
+      "epoch": 0.590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 8.17958352367412e-06,
-      "loss": 0.8033,
+      "learning_rate": 7.578640836821561e-06,
+      "loss": 0.8089,
       "step": 20810
     },
     {
-      "epoch": 0.5715580456455467,
+      "epoch": 0.5905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 8.178708877946852e-06,
-      "loss": 0.8784,
+      "learning_rate": 7.57774911859384e-06,
+      "loss": 0.8759,
       "step": 20811
     },
     {
-      "epoch": 0.57158550987339,
+      "epoch": 0.5905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 8.177834246630342e-06,
-      "loss": 0.8338,
+      "learning_rate": 7.576857420826987e-06,
+      "loss": 0.8652,
       "step": 20812
     },
     {
-      "epoch": 0.5716129741012331,
+      "epoch": 0.5906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 8.176959629731518e-06,
-      "loss": 0.9655,
+      "learning_rate": 7.575965743528524e-06,
+      "loss": 0.8363,
       "step": 20813
     },
     {
-      "epoch": 0.5716404383290764,
+      "epoch": 0.5906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 8.176085027257293e-06,
-      "loss": 0.8208,
+      "learning_rate": 7.57507408670599e-06,
+      "loss": 0.8278,
       "step": 20814
     },
     {
-      "epoch": 0.5716679025569196,
+      "epoch": 0.5906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 8.175210439214594e-06,
-      "loss": 0.845,
+      "learning_rate": 7.574182450366915e-06,
+      "loss": 0.882,
       "step": 20815
     },
     {
-      "epoch": 0.5716953667847628,
+      "epoch": 0.5906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 8.174335865610333e-06,
-      "loss": 0.9032,
+      "learning_rate": 7.5732908345188276e-06,
+      "loss": 0.7231,
       "step": 20816
     },
     {
-      "epoch": 0.5717228310126061,
+      "epoch": 0.5907207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 8.173461306451435e-06,
-      "loss": 0.8718,
+      "learning_rate": 7.572399239169263e-06,
+      "loss": 0.8848,
       "step": 20817
     },
     {
-      "epoch": 0.5717502952404493,
+      "epoch": 0.5907491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 8.172586761744817e-06,
-      "loss": 0.9029,
+      "learning_rate": 7.571507664325751e-06,
+      "loss": 0.8855,
       "step": 20818
     },
     {
-      "epoch": 0.5717777594682926,
+      "epoch": 0.5907775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 8.1717122314974e-06,
-      "loss": 0.8563,
+      "learning_rate": 7.570616109995821e-06,
+      "loss": 0.8632,
       "step": 20819
     },
     {
-      "epoch": 0.5718052236961357,
+      "epoch": 0.5908059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 8.170837715716107e-06,
-      "loss": 0.923,
+      "learning_rate": 7.569724576187009e-06,
+      "loss": 0.8293,
       "step": 20820
     },
     {
-      "epoch": 0.571832687923979,
+      "epoch": 0.5908342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 8.169963214407856e-06,
-      "loss": 0.7494,
+      "learning_rate": 7.568833062906842e-06,
+      "loss": 0.8539,
       "step": 20821
     },
     {
-      "epoch": 0.5718601521518223,
+      "epoch": 0.5908626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 8.169088727579562e-06,
-      "loss": 0.8624,
+      "learning_rate": 7.567941570162849e-06,
+      "loss": 0.8243,
       "step": 20822
     },
     {
-      "epoch": 0.5718876163796655,
+      "epoch": 0.5908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 8.168214255238145e-06,
-      "loss": 0.9192,
+      "learning_rate": 7.567050097962566e-06,
+      "loss": 0.8956,
       "step": 20823
     },
     {
-      "epoch": 0.5719150806075087,
+      "epoch": 0.5909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 8.167339797390527e-06,
-      "loss": 0.8779,
+      "learning_rate": 7.56615864631352e-06,
+      "loss": 0.935,
       "step": 20824
     },
     {
-      "epoch": 0.571942544835352,
+      "epoch": 0.5909477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 8.16646535404363e-06,
-      "loss": 0.9041,
+      "learning_rate": 7.565267215223238e-06,
+      "loss": 0.7227,
       "step": 20825
     },
     {
-      "epoch": 0.5719700090631952,
+      "epoch": 0.5909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 8.165590925204366e-06,
-      "loss": 0.828,
+      "learning_rate": 7.564375804699257e-06,
+      "loss": 0.8675,
       "step": 20826
     },
     {
-      "epoch": 0.5719974732910384,
+      "epoch": 0.5910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 8.164716510879662e-06,
-      "loss": 0.7876,
+      "learning_rate": 7.563484414749097e-06,
+      "loss": 0.9464,
       "step": 20827
     },
     {
-      "epoch": 0.5720249375188816,
+      "epoch": 0.591032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 8.163842111076427e-06,
-      "loss": 0.7823,
+      "learning_rate": 7.5625930453802995e-06,
+      "loss": 0.8721,
       "step": 20828
     },
     {
-      "epoch": 0.5720524017467249,
+      "epoch": 0.591061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 8.162967725801586e-06,
-      "loss": 0.892,
+      "learning_rate": 7.561701696600387e-06,
+      "loss": 0.8905,
       "step": 20829
     },
     {
-      "epoch": 0.5720798659745682,
+      "epoch": 0.5910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 8.162093355062056e-06,
-      "loss": 0.9136,
+      "learning_rate": 7.5608103684168885e-06,
+      "loss": 0.8239,
       "step": 20830
     },
     {
-      "epoch": 0.5721073302024113,
+      "epoch": 0.5911180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 8.161218998864753e-06,
-      "loss": 0.8356,
+      "learning_rate": 7.559919060837337e-06,
+      "loss": 0.7335,
       "step": 20831
     },
     {
-      "epoch": 0.5721347944302546,
+      "epoch": 0.5911464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 8.160344657216602e-06,
-      "loss": 0.9124,
+      "learning_rate": 7.559027773869259e-06,
+      "loss": 0.7597,
       "step": 20832
     },
     {
-      "epoch": 0.5721622586580978,
+      "epoch": 0.5911748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 8.159470330124517e-06,
-      "loss": 0.9213,
+      "learning_rate": 7.55813650752018e-06,
+      "loss": 0.8552,
       "step": 20833
     },
     {
-      "epoch": 0.572189722885941,
+      "epoch": 0.5912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 8.158596017595413e-06,
-      "loss": 0.8236,
+      "learning_rate": 7.557245261797633e-06,
+      "loss": 0.9158,
       "step": 20834
     },
     {
-      "epoch": 0.5722171871137843,
+      "epoch": 0.5912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 8.15772171963621e-06,
-      "loss": 0.8178,
+      "learning_rate": 7.556354036709147e-06,
+      "loss": 0.9702,
       "step": 20835
     },
     {
-      "epoch": 0.5722446513416275,
+      "epoch": 0.5912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 8.156847436253825e-06,
-      "loss": 0.8802,
+      "learning_rate": 7.5554628322622446e-06,
+      "loss": 0.9269,
       "step": 20836
     },
     {
-      "epoch": 0.5722721155694708,
+      "epoch": 0.5912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 8.15597316745518e-06,
-      "loss": 0.8197,
+      "learning_rate": 7.554571648464461e-06,
+      "loss": 0.8737,
       "step": 20837
     },
     {
-      "epoch": 0.572299579797314,
+      "epoch": 0.591316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.155098913247189e-06,
-      "loss": 0.8861,
+      "learning_rate": 7.5536804853233206e-06,
+      "loss": 0.872,
       "step": 20838
     },
     {
-      "epoch": 0.5723270440251572,
+      "epoch": 0.5913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 8.154224673636773e-06,
-      "loss": 0.8866,
+      "learning_rate": 7.552789342846347e-06,
+      "loss": 0.8032,
       "step": 20839
     },
     {
-      "epoch": 0.5723545082530005,
+      "epoch": 0.5913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 8.153350448630842e-06,
-      "loss": 0.8752,
+      "learning_rate": 7.551898221041076e-06,
+      "loss": 0.8933,
       "step": 20840
     },
     {
-      "epoch": 0.5723819724808437,
+      "epoch": 0.5914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 8.152476238236319e-06,
-      "loss": 0.8979,
+      "learning_rate": 7.5510071199150305e-06,
+      "loss": 0.9422,
       "step": 20841
     },
     {
-      "epoch": 0.5724094367086869,
+      "epoch": 0.5914301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 8.151602042460119e-06,
-      "loss": 0.8575,
+      "learning_rate": 7.550116039475734e-06,
+      "loss": 0.8059,
       "step": 20842
     },
     {
-      "epoch": 0.5724369009365302,
+      "epoch": 0.5914585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 8.150727861309158e-06,
-      "loss": 0.8938,
+      "learning_rate": 7.549224979730718e-06,
+      "loss": 0.9683,
       "step": 20843
     },
     {
-      "epoch": 0.5724643651643734,
+      "epoch": 0.5914869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 8.149853694790357e-06,
-      "loss": 0.7605,
+      "learning_rate": 7.548333940687512e-06,
+      "loss": 0.8167,
       "step": 20844
     },
     {
-      "epoch": 0.5724918293922167,
+      "epoch": 0.5915153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 8.148979542910633e-06,
-      "loss": 0.8693,
+      "learning_rate": 7.547442922353639e-06,
+      "loss": 0.7832,
       "step": 20845
     },
     {
-      "epoch": 0.5725192936200598,
+      "epoch": 0.5915437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 8.148105405676895e-06,
-      "loss": 0.8883,
+      "learning_rate": 7.546551924736625e-06,
+      "loss": 0.9702,
       "step": 20846
     },
     {
-      "epoch": 0.5725467578479031,
+      "epoch": 0.591572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.147231283096067e-06,
-      "loss": 0.8935,
+      "learning_rate": 7.5456609478439975e-06,
+      "loss": 0.7591,
       "step": 20847
     },
     {
-      "epoch": 0.5725742220757464,
+      "epoch": 0.5916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 8.14635717517506e-06,
-      "loss": 0.8318,
+      "learning_rate": 7.544769991683283e-06,
+      "loss": 0.867,
       "step": 20848
     },
     {
-      "epoch": 0.5726016863035895,
+      "epoch": 0.5916288308740069,
       "grad_norm": 0.0,
-      "learning_rate": 8.145483081920795e-06,
-      "loss": 0.947,
+      "learning_rate": 7.543879056262004e-06,
+      "loss": 0.8952,
       "step": 20849
     },
     {
-      "epoch": 0.5726291505314328,
+      "epoch": 0.5916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.144609003340186e-06,
-      "loss": 0.9033,
+      "learning_rate": 7.542988141587692e-06,
+      "loss": 0.9204,
       "step": 20850
     },
     {
-      "epoch": 0.5726566147592761,
+      "epoch": 0.5916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 8.143734939440153e-06,
-      "loss": 0.792,
+      "learning_rate": 7.54209724766787e-06,
+      "loss": 0.762,
       "step": 20851
     },
     {
-      "epoch": 0.5726840789871193,
+      "epoch": 0.5917139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 8.142860890227603e-06,
-      "loss": 0.8198,
+      "learning_rate": 7.5412063745100624e-06,
+      "loss": 0.8412,
       "step": 20852
     },
     {
-      "epoch": 0.5727115432149625,
+      "epoch": 0.5917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 8.141986855709459e-06,
-      "loss": 0.9365,
+      "learning_rate": 7.540315522121794e-06,
+      "loss": 0.9115,
       "step": 20853
     },
     {
-      "epoch": 0.5727390074428057,
+      "epoch": 0.5917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 8.141112835892635e-06,
-      "loss": 0.8947,
+      "learning_rate": 7.539424690510592e-06,
+      "loss": 0.8677,
       "step": 20854
     },
     {
-      "epoch": 0.572766471670649,
+      "epoch": 0.5917990919409761,
       "grad_norm": 0.0,
-      "learning_rate": 8.140238830784044e-06,
-      "loss": 0.8602,
+      "learning_rate": 7.5385338796839805e-06,
+      "loss": 0.9275,
       "step": 20855
     },
     {
-      "epoch": 0.5727939358984923,
+      "epoch": 0.5918274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 8.139364840390607e-06,
-      "loss": 0.9052,
+      "learning_rate": 7.537643089649482e-06,
+      "loss": 0.8121,
       "step": 20856
     },
     {
-      "epoch": 0.5728214001263354,
+      "epoch": 0.591855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 8.138490864719237e-06,
-      "loss": 0.8352,
+      "learning_rate": 7.5367523204146195e-06,
+      "loss": 0.8481,
       "step": 20857
     },
     {
-      "epoch": 0.5728488643541787,
+      "epoch": 0.5918842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 8.137616903776845e-06,
-      "loss": 0.9249,
+      "learning_rate": 7.535861571986927e-06,
+      "loss": 0.9692,
       "step": 20858
     },
     {
-      "epoch": 0.5728763285820219,
+      "epoch": 0.5919125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 8.13674295757035e-06,
-      "loss": 0.7963,
+      "learning_rate": 7.534970844373922e-06,
+      "loss": 0.8258,
       "step": 20859
     },
     {
-      "epoch": 0.5729037928098651,
+      "epoch": 0.5919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 8.135869026106666e-06,
-      "loss": 0.8164,
+      "learning_rate": 7.534080137583128e-06,
+      "loss": 0.8904,
       "step": 20860
     },
     {
-      "epoch": 0.5729312570377084,
+      "epoch": 0.5919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 8.13499510939271e-06,
-      "loss": 0.9078,
+      "learning_rate": 7.5331894516220696e-06,
+      "loss": 0.759,
       "step": 20861
     },
     {
-      "epoch": 0.5729587212655516,
+      "epoch": 0.5919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 8.134121207435397e-06,
-      "loss": 0.9167,
+      "learning_rate": 7.5322987864982725e-06,
+      "loss": 0.7676,
       "step": 20862
     },
     {
-      "epoch": 0.5729861854933949,
+      "epoch": 0.5920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.133247320241635e-06,
-      "loss": 0.9295,
+      "learning_rate": 7.531408142219257e-06,
+      "loss": 0.8097,
       "step": 20863
     },
     {
-      "epoch": 0.5730136497212381,
+      "epoch": 0.5920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 8.132373447818345e-06,
-      "loss": 0.8257,
+      "learning_rate": 7.530517518792548e-06,
+      "loss": 0.749,
       "step": 20864
     },
     {
-      "epoch": 0.5730411139490813,
+      "epoch": 0.5920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 8.131499590172441e-06,
-      "loss": 0.8593,
+      "learning_rate": 7.52962691622567e-06,
+      "loss": 0.8023,
       "step": 20865
     },
     {
-      "epoch": 0.5730685781769246,
+      "epoch": 0.59211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 8.130625747310835e-06,
-      "loss": 0.866,
+      "learning_rate": 7.528736334526141e-06,
+      "loss": 0.8226,
       "step": 20866
     },
     {
-      "epoch": 0.5730960424047677,
+      "epoch": 0.5921396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 8.12975191924044e-06,
-      "loss": 0.8805,
+      "learning_rate": 7.52784577370149e-06,
+      "loss": 0.7241,
       "step": 20867
     },
     {
-      "epoch": 0.573123506632611,
+      "epoch": 0.5921679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 8.128878105968178e-06,
-      "loss": 0.9576,
+      "learning_rate": 7.526955233759238e-06,
+      "loss": 0.9722,
       "step": 20868
     },
     {
-      "epoch": 0.5731509708604543,
+      "epoch": 0.5921963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 8.128004307500956e-06,
-      "loss": 0.7935,
+      "learning_rate": 7.5260647147069036e-06,
+      "loss": 1.0038,
       "step": 20869
     },
     {
-      "epoch": 0.5731784350882975,
+      "epoch": 0.5922247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 8.127130523845686e-06,
-      "loss": 0.8401,
+      "learning_rate": 7.525174216552014e-06,
+      "loss": 0.8048,
       "step": 20870
     },
     {
-      "epoch": 0.5732058993161407,
+      "epoch": 0.5922531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 8.126256755009284e-06,
-      "loss": 0.8425,
+      "learning_rate": 7.524283739302087e-06,
+      "loss": 0.8434,
       "step": 20871
     },
     {
-      "epoch": 0.5732333635439839,
+      "epoch": 0.5922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 8.125383000998667e-06,
-      "loss": 0.8493,
+      "learning_rate": 7.523393282964646e-06,
+      "loss": 0.8045,
       "step": 20872
     },
     {
-      "epoch": 0.5732608277718272,
+      "epoch": 0.5923098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 8.124509261820745e-06,
-      "loss": 0.8392,
+      "learning_rate": 7.522502847547214e-06,
+      "loss": 0.7911,
       "step": 20873
     },
     {
-      "epoch": 0.5732882919996705,
+      "epoch": 0.5923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 8.123635537482434e-06,
-      "loss": 0.7961,
+      "learning_rate": 7.521612433057312e-06,
+      "loss": 0.7433,
       "step": 20874
     },
     {
-      "epoch": 0.5733157562275136,
+      "epoch": 0.592366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 8.122761827990643e-06,
-      "loss": 0.8388,
+      "learning_rate": 7.5207220395024615e-06,
+      "loss": 0.9131,
       "step": 20875
     },
     {
-      "epoch": 0.5733432204553569,
+      "epoch": 0.5923950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 8.121888133352286e-06,
-      "loss": 0.8614,
+      "learning_rate": 7.519831666890185e-06,
+      "loss": 0.8573,
       "step": 20876
     },
     {
-      "epoch": 0.5733706846832002,
+      "epoch": 0.5924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 8.121014453574284e-06,
-      "loss": 0.8613,
+      "learning_rate": 7.518941315227999e-06,
+      "loss": 0.8642,
       "step": 20877
     },
     {
-      "epoch": 0.5733981489110433,
+      "epoch": 0.5924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 8.120140788663536e-06,
-      "loss": 0.8481,
+      "learning_rate": 7.518050984523429e-06,
+      "loss": 0.8768,
       "step": 20878
     },
     {
-      "epoch": 0.5734256131388866,
+      "epoch": 0.5924801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 8.119267138626966e-06,
-      "loss": 0.8445,
+      "learning_rate": 7.517160674783994e-06,
+      "loss": 0.8656,
       "step": 20879
     },
     {
-      "epoch": 0.5734530773667298,
+      "epoch": 0.5925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 8.118393503471484e-06,
-      "loss": 0.8578,
+      "learning_rate": 7.516270386017214e-06,
+      "loss": 0.9116,
       "step": 20880
     },
     {
-      "epoch": 0.5734805415945731,
+      "epoch": 0.5925368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 8.117519883204002e-06,
-      "loss": 0.9713,
+      "learning_rate": 7.51538011823061e-06,
+      "loss": 0.9028,
       "step": 20881
     },
     {
-      "epoch": 0.5735080058224163,
+      "epoch": 0.5925652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 8.116646277831428e-06,
-      "loss": 0.824,
+      "learning_rate": 7.514489871431703e-06,
+      "loss": 0.9087,
       "step": 20882
     },
     {
-      "epoch": 0.5735354700502595,
+      "epoch": 0.5925936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 8.11577268736068e-06,
-      "loss": 0.904,
+      "learning_rate": 7.513599645628008e-06,
+      "loss": 0.8121,
       "step": 20883
     },
     {
-      "epoch": 0.5735629342781028,
+      "epoch": 0.5926220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 8.114899111798669e-06,
-      "loss": 0.993,
+      "learning_rate": 7.512709440827052e-06,
+      "loss": 0.8709,
       "step": 20884
     },
     {
-      "epoch": 0.573590398505946,
+      "epoch": 0.5926503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 8.114025551152305e-06,
-      "loss": 0.8868,
+      "learning_rate": 7.51181925703635e-06,
+      "loss": 0.8362,
       "step": 20885
     },
     {
-      "epoch": 0.5736178627337892,
+      "epoch": 0.5926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 8.113152005428506e-06,
-      "loss": 0.8417,
+      "learning_rate": 7.5109290942634216e-06,
+      "loss": 0.8174,
       "step": 20886
     },
     {
-      "epoch": 0.5736453269616325,
+      "epoch": 0.5927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 8.112278474634173e-06,
-      "loss": 0.8195,
+      "learning_rate": 7.510038952515789e-06,
+      "loss": 0.8511,
       "step": 20887
     },
     {
-      "epoch": 0.5736727911894757,
+      "epoch": 0.5927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 8.111404958776226e-06,
-      "loss": 0.8779,
+      "learning_rate": 7.509148831800965e-06,
+      "loss": 0.8742,
       "step": 20888
     },
     {
-      "epoch": 0.573700255417319,
+      "epoch": 0.5927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 8.110531457861577e-06,
-      "loss": 0.8979,
+      "learning_rate": 7.508258732126477e-06,
+      "loss": 0.8158,
       "step": 20889
     },
     {
-      "epoch": 0.5737277196451622,
+      "epoch": 0.5927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 8.109657971897132e-06,
-      "loss": 0.8724,
+      "learning_rate": 7.507368653499838e-06,
+      "loss": 0.7519,
       "step": 20890
     },
     {
-      "epoch": 0.5737551838730054,
+      "epoch": 0.5928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 8.108784500889805e-06,
-      "loss": 0.8164,
+      "learning_rate": 7.506478595928569e-06,
+      "loss": 0.9208,
       "step": 20891
     },
     {
-      "epoch": 0.5737826481008487,
+      "epoch": 0.5928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 8.107911044846513e-06,
-      "loss": 0.8148,
+      "learning_rate": 7.505588559420188e-06,
+      "loss": 0.7386,
       "step": 20892
     },
     {
-      "epoch": 0.5738101123286918,
+      "epoch": 0.592877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 8.107037603774158e-06,
-      "loss": 0.8353,
+      "learning_rate": 7.504698543982213e-06,
+      "loss": 0.8918,
       "step": 20893
     },
     {
-      "epoch": 0.5738375765565351,
+      "epoch": 0.592905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 8.106164177679654e-06,
-      "loss": 0.9297,
+      "learning_rate": 7.503808549622158e-06,
+      "loss": 0.7762,
       "step": 20894
     },
     {
-      "epoch": 0.5738650407843784,
+      "epoch": 0.5929341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 8.10529076656991e-06,
-      "loss": 0.8769,
+      "learning_rate": 7.502918576347548e-06,
+      "loss": 0.9306,
       "step": 20895
     },
     {
-      "epoch": 0.5738925050122216,
+      "epoch": 0.5929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 8.104417370451845e-06,
-      "loss": 0.8554,
+      "learning_rate": 7.502028624165896e-06,
+      "loss": 0.9123,
       "step": 20896
     },
     {
-      "epoch": 0.5739199692400648,
+      "epoch": 0.5929909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 8.10354398933236e-06,
-      "loss": 0.8376,
+      "learning_rate": 7.50113869308472e-06,
+      "loss": 0.8835,
       "step": 20897
     },
     {
-      "epoch": 0.573947433467908,
+      "epoch": 0.5930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 8.102670623218375e-06,
-      "loss": 0.9036,
+      "learning_rate": 7.5002487831115375e-06,
+      "loss": 0.781,
       "step": 20898
     },
     {
-      "epoch": 0.5739748976957513,
+      "epoch": 0.5930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 8.101797272116791e-06,
-      "loss": 0.8663,
+      "learning_rate": 7.4993588942538675e-06,
+      "loss": 0.8309,
       "step": 20899
     },
     {
-      "epoch": 0.5740023619235946,
+      "epoch": 0.5930760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 8.100923936034521e-06,
-      "loss": 0.8969,
+      "learning_rate": 7.498469026519223e-06,
+      "loss": 0.8178,
       "step": 20900
     },
     {
-      "epoch": 0.5740298261514377,
+      "epoch": 0.5931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 8.100050614978478e-06,
-      "loss": 0.8706,
+      "learning_rate": 7.497579179915124e-06,
+      "loss": 0.832,
       "step": 20901
     },
     {
-      "epoch": 0.574057290379281,
+      "epoch": 0.5931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 8.09917730895557e-06,
-      "loss": 0.9951,
+      "learning_rate": 7.496689354449088e-06,
+      "loss": 0.8967,
       "step": 20902
     },
     {
-      "epoch": 0.5740847546071243,
+      "epoch": 0.5931611804767309,
       "grad_norm": 0.0,
-      "learning_rate": 8.098304017972707e-06,
-      "loss": 0.8662,
+      "learning_rate": 7.495799550128625e-06,
+      "loss": 0.8451,
       "step": 20903
     },
     {
-      "epoch": 0.5741122188349674,
+      "epoch": 0.5931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 8.097430742036803e-06,
-      "loss": 0.9214,
+      "learning_rate": 7.494909766961258e-06,
+      "loss": 0.9629,
       "step": 20904
     },
     {
-      "epoch": 0.5741396830628107,
+      "epoch": 0.5932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 8.096557481154762e-06,
-      "loss": 0.8107,
+      "learning_rate": 7.494020004954501e-06,
+      "loss": 0.7716,
       "step": 20905
     },
     {
-      "epoch": 0.5741671472906539,
+      "epoch": 0.5932463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 8.095684235333494e-06,
-      "loss": 0.8475,
+      "learning_rate": 7.493130264115871e-06,
+      "loss": 0.8198,
       "step": 20906
     },
     {
-      "epoch": 0.5741946115184972,
+      "epoch": 0.5932746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 8.094811004579908e-06,
-      "loss": 0.922,
+      "learning_rate": 7.492240544452881e-06,
+      "loss": 0.8392,
       "step": 20907
     },
     {
-      "epoch": 0.5742220757463404,
+      "epoch": 0.5933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 8.09393778890092e-06,
-      "loss": 0.7901,
+      "learning_rate": 7.491350845973049e-06,
+      "loss": 0.8736,
       "step": 20908
     },
     {
-      "epoch": 0.5742495399741836,
+      "epoch": 0.5933314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 8.09306458830343e-06,
-      "loss": 0.9067,
+      "learning_rate": 7.490461168683889e-06,
+      "loss": 0.8327,
       "step": 20909
     },
     {
-      "epoch": 0.5742770042020269,
+      "epoch": 0.5933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 8.092191402794359e-06,
-      "loss": 0.9395,
+      "learning_rate": 7.489571512592915e-06,
+      "loss": 0.9576,
       "step": 20910
     },
     {
-      "epoch": 0.57430446842987,
+      "epoch": 0.5933881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 8.091318232380602e-06,
-      "loss": 0.7838,
+      "learning_rate": 7.488681877707645e-06,
+      "loss": 0.844,
       "step": 20911
     },
     {
-      "epoch": 0.5743319326577133,
+      "epoch": 0.593416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 8.090445077069074e-06,
-      "loss": 0.8534,
+      "learning_rate": 7.487792264035593e-06,
+      "loss": 0.8908,
       "step": 20912
     },
     {
-      "epoch": 0.5743593968855566,
+      "epoch": 0.5934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 8.089571936866689e-06,
-      "loss": 0.9196,
+      "learning_rate": 7.486902671584268e-06,
+      "loss": 0.7574,
       "step": 20913
     },
     {
-      "epoch": 0.5743868611133998,
+      "epoch": 0.5934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 8.088698811780346e-06,
-      "loss": 0.9445,
+      "learning_rate": 7.486013100361193e-06,
+      "loss": 0.8467,
       "step": 20914
     },
     {
-      "epoch": 0.574414325341243,
+      "epoch": 0.5935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 8.08782570181696e-06,
-      "loss": 0.9644,
+      "learning_rate": 7.485123550373879e-06,
+      "loss": 0.9418,
       "step": 20915
     },
     {
-      "epoch": 0.5744417895690863,
+      "epoch": 0.5935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 8.086952606983442e-06,
-      "loss": 0.8798,
+      "learning_rate": 7.484234021629837e-06,
+      "loss": 0.8704,
       "step": 20916
     },
     {
-      "epoch": 0.5744692537969295,
+      "epoch": 0.5935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 8.086079527286693e-06,
-      "loss": 0.8435,
+      "learning_rate": 7.483344514136584e-06,
+      "loss": 0.7982,
       "step": 20917
     },
     {
-      "epoch": 0.5744967180247728,
+      "epoch": 0.5935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 8.085206462733623e-06,
-      "loss": 0.8196,
+      "learning_rate": 7.482455027901635e-06,
+      "loss": 0.9128,
       "step": 20918
     },
     {
-      "epoch": 0.5745241822526159,
+      "epoch": 0.5936152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 8.08433341333114e-06,
-      "loss": 0.9268,
+      "learning_rate": 7.481565562932496e-06,
+      "loss": 0.8603,
       "step": 20919
     },
     {
-      "epoch": 0.5745516464804592,
+      "epoch": 0.5936435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 8.083460379086157e-06,
-      "loss": 0.8976,
+      "learning_rate": 7.48067611923669e-06,
+      "loss": 0.8893,
       "step": 20920
     },
     {
-      "epoch": 0.5745791107083025,
+      "epoch": 0.5936719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 8.082587360005575e-06,
-      "loss": 0.9008,
+      "learning_rate": 7.479786696821724e-06,
+      "loss": 0.9282,
       "step": 20921
     },
     {
-      "epoch": 0.5746065749361456,
+      "epoch": 0.5937003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 8.081714356096308e-06,
-      "loss": 0.8696,
+      "learning_rate": 7.4788972956951164e-06,
+      "loss": 0.9231,
       "step": 20922
     },
     {
-      "epoch": 0.5746340391639889,
+      "epoch": 0.5937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 8.080841367365256e-06,
-      "loss": 0.8867,
+      "learning_rate": 7.478007915864376e-06,
+      "loss": 0.8376,
       "step": 20923
     },
     {
-      "epoch": 0.5746615033918321,
+      "epoch": 0.5937570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 8.07996839381933e-06,
-      "loss": 0.8415,
+      "learning_rate": 7.4771185573370135e-06,
+      "loss": 0.8916,
       "step": 20924
     },
     {
-      "epoch": 0.5746889676196754,
+      "epoch": 0.5937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 8.079095435465442e-06,
-      "loss": 0.8372,
+      "learning_rate": 7.4762292201205466e-06,
+      "loss": 0.8597,
       "step": 20925
     },
     {
-      "epoch": 0.5747164318475186,
+      "epoch": 0.5938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 8.07822249231049e-06,
-      "loss": 0.7796,
+      "learning_rate": 7.475339904222485e-06,
+      "loss": 0.8554,
       "step": 20926
     },
     {
-      "epoch": 0.5747438960753618,
+      "epoch": 0.5938422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 8.077349564361394e-06,
-      "loss": 0.7063,
+      "learning_rate": 7.474450609650339e-06,
+      "loss": 0.8421,
       "step": 20927
     },
     {
-      "epoch": 0.5747713603032051,
+      "epoch": 0.5938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 8.076476651625046e-06,
-      "loss": 0.8455,
+      "learning_rate": 7.4735613364116235e-06,
+      "loss": 0.8516,
       "step": 20928
     },
     {
-      "epoch": 0.5747988245310484,
+      "epoch": 0.5938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 8.075603754108364e-06,
-      "loss": 0.9026,
+      "learning_rate": 7.472672084513848e-06,
+      "loss": 0.912,
       "step": 20929
     },
     {
-      "epoch": 0.5748262887588915,
+      "epoch": 0.593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 8.074730871818246e-06,
-      "loss": 0.9105,
+      "learning_rate": 7.471782853964525e-06,
+      "loss": 0.821,
       "step": 20930
     },
     {
-      "epoch": 0.5748537529867348,
+      "epoch": 0.593955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 8.073858004761603e-06,
-      "loss": 0.8492,
+      "learning_rate": 7.470893644771167e-06,
+      "loss": 0.7538,
       "step": 20931
     },
     {
-      "epoch": 0.574881217214578,
+      "epoch": 0.5939841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 8.072985152945345e-06,
-      "loss": 0.8715,
+      "learning_rate": 7.4700044569412845e-06,
+      "loss": 0.8364,
       "step": 20932
     },
     {
-      "epoch": 0.5749086814424212,
+      "epoch": 0.5940124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 8.072112316376375e-06,
-      "loss": 0.9055,
+      "learning_rate": 7.469115290482385e-06,
+      "loss": 0.9086,
       "step": 20933
     },
     {
-      "epoch": 0.5749361456702645,
+      "epoch": 0.5940408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 8.071239495061601e-06,
-      "loss": 0.9434,
+      "learning_rate": 7.4682261454019826e-06,
+      "loss": 0.8193,
       "step": 20934
     },
     {
-      "epoch": 0.5749636098981077,
+      "epoch": 0.5940692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 8.070366689007923e-06,
-      "loss": 0.8925,
+      "learning_rate": 7.467337021707588e-06,
+      "loss": 0.8923,
       "step": 20935
     },
     {
-      "epoch": 0.574991074125951,
+      "epoch": 0.5940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 8.06949389822225e-06,
-      "loss": 0.8625,
+      "learning_rate": 7.466447919406714e-06,
+      "loss": 0.9046,
       "step": 20936
     },
     {
-      "epoch": 0.5750185383537941,
+      "epoch": 0.5941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 8.068621122711493e-06,
-      "loss": 0.8445,
+      "learning_rate": 7.465558838506868e-06,
+      "loss": 0.9825,
       "step": 20937
     },
     {
-      "epoch": 0.5750460025816374,
+      "epoch": 0.5941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 8.067748362482551e-06,
-      "loss": 0.808,
+      "learning_rate": 7.464669779015559e-06,
+      "loss": 0.8197,
       "step": 20938
     },
     {
-      "epoch": 0.5750734668094807,
+      "epoch": 0.5941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 8.066875617542337e-06,
-      "loss": 0.7761,
+      "learning_rate": 7.463780740940299e-06,
+      "loss": 0.7193,
       "step": 20939
     },
     {
-      "epoch": 0.5751009310373238,
+      "epoch": 0.594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 8.066002887897748e-06,
-      "loss": 0.8448,
+      "learning_rate": 7.462891724288599e-06,
+      "loss": 0.8285,
       "step": 20940
     },
     {
-      "epoch": 0.5751283952651671,
+      "epoch": 0.5942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 8.065130173555694e-06,
-      "loss": 0.8496,
+      "learning_rate": 7.462002729067963e-06,
+      "loss": 0.8834,
       "step": 20941
     },
     {
-      "epoch": 0.5751558594930104,
+      "epoch": 0.5942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 8.064257474523079e-06,
-      "loss": 0.9176,
+      "learning_rate": 7.461113755285907e-06,
+      "loss": 0.7433,
       "step": 20942
     },
     {
-      "epoch": 0.5751833237208536,
+      "epoch": 0.5942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 8.063384790806808e-06,
-      "loss": 0.9314,
+      "learning_rate": 7.460224802949936e-06,
+      "loss": 0.8969,
       "step": 20943
     },
     {
-      "epoch": 0.5752107879486968,
+      "epoch": 0.5943246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 8.06251212241379e-06,
-      "loss": 0.9669,
+      "learning_rate": 7.459335872067559e-06,
+      "loss": 0.9919,
       "step": 20944
     },
     {
-      "epoch": 0.57523825217654,
+      "epoch": 0.5943530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 8.061639469350925e-06,
-      "loss": 0.8771,
+      "learning_rate": 7.458446962646288e-06,
+      "loss": 0.8304,
       "step": 20945
     },
     {
-      "epoch": 0.5752657164043833,
+      "epoch": 0.5943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 8.06076683162512e-06,
-      "loss": 0.8903,
+      "learning_rate": 7.457558074693631e-06,
+      "loss": 0.809,
       "step": 20946
     },
     {
-      "epoch": 0.5752931806322266,
+      "epoch": 0.5944097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 8.059894209243278e-06,
-      "loss": 0.9157,
+      "learning_rate": 7.456669208217091e-06,
+      "loss": 0.8582,
       "step": 20947
     },
     {
-      "epoch": 0.5753206448600697,
+      "epoch": 0.5944381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 8.059021602212302e-06,
-      "loss": 0.7844,
+      "learning_rate": 7.455780363224184e-06,
+      "loss": 0.8869,
       "step": 20948
     },
     {
-      "epoch": 0.575348109087913,
+      "epoch": 0.594466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 8.058149010539103e-06,
-      "loss": 0.9412,
+      "learning_rate": 7.454891539722411e-06,
+      "loss": 1.0362,
       "step": 20949
     },
     {
-      "epoch": 0.5753755733157562,
+      "epoch": 0.5944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 8.057276434230579e-06,
-      "loss": 0.8166,
+      "learning_rate": 7.454002737719286e-06,
+      "loss": 0.8524,
       "step": 20950
     },
     {
-      "epoch": 0.5754030375435994,
+      "epoch": 0.5945232690124859,
       "grad_norm": 0.0,
-      "learning_rate": 8.05640387329364e-06,
-      "loss": 0.9065,
+      "learning_rate": 7.4531139572223135e-06,
+      "loss": 0.8473,
       "step": 20951
     },
     {
-      "epoch": 0.5754305017714427,
+      "epoch": 0.5945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 8.055531327735181e-06,
-      "loss": 0.8728,
+      "learning_rate": 7.452225198239004e-06,
+      "loss": 0.8952,
       "step": 20952
     },
     {
-      "epoch": 0.5754579659992859,
+      "epoch": 0.5945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 8.054658797562116e-06,
-      "loss": 0.8493,
+      "learning_rate": 7.451336460776861e-06,
+      "loss": 0.8352,
       "step": 20953
     },
     {
-      "epoch": 0.5754854302271292,
+      "epoch": 0.5946083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 8.05378628278134e-06,
-      "loss": 0.8845,
+      "learning_rate": 7.450447744843394e-06,
+      "loss": 0.8347,
       "step": 20954
     },
     {
-      "epoch": 0.5755128944549724,
+      "epoch": 0.5946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 8.052913783399762e-06,
-      "loss": 0.9027,
+      "learning_rate": 7.44955905044611e-06,
+      "loss": 0.7711,
       "step": 20955
     },
     {
-      "epoch": 0.5755403586828156,
+      "epoch": 0.5946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 8.052041299424287e-06,
-      "loss": 0.8002,
+      "learning_rate": 7.448670377592515e-06,
+      "loss": 0.8222,
       "step": 20956
     },
     {
-      "epoch": 0.5755678229106589,
+      "epoch": 0.5946935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 8.051168830861815e-06,
-      "loss": 0.8337,
+      "learning_rate": 7.447781726290116e-06,
+      "loss": 0.8159,
       "step": 20957
     },
     {
-      "epoch": 0.575595287138502,
+      "epoch": 0.5947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 8.05029637771925e-06,
-      "loss": 0.9159,
+      "learning_rate": 7.4468930965464194e-06,
+      "loss": 0.8575,
       "step": 20958
     },
     {
-      "epoch": 0.5756227513663453,
+      "epoch": 0.594750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 8.049423940003493e-06,
-      "loss": 0.856,
+      "learning_rate": 7.446004488368933e-06,
+      "loss": 0.8314,
       "step": 20959
     },
     {
-      "epoch": 0.5756502155941886,
+      "epoch": 0.5947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 8.04855151772145e-06,
-      "loss": 0.8416,
+      "learning_rate": 7.4451159017651605e-06,
+      "loss": 0.8713,
       "step": 20960
     },
     {
-      "epoch": 0.5756776798220318,
+      "epoch": 0.5948070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 8.047679110880026e-06,
-      "loss": 0.9034,
+      "learning_rate": 7.444227336742608e-06,
+      "loss": 0.9538,
       "step": 20961
     },
     {
-      "epoch": 0.575705144049875,
+      "epoch": 0.5948354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 8.046806719486118e-06,
-      "loss": 0.8595,
+      "learning_rate": 7.443338793308783e-06,
+      "loss": 0.8707,
       "step": 20962
     },
     {
-      "epoch": 0.5757326082777183,
+      "epoch": 0.5948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 8.045934343546634e-06,
-      "loss": 0.8815,
+      "learning_rate": 7.442450271471191e-06,
+      "loss": 0.887,
       "step": 20963
     },
     {
-      "epoch": 0.5757600725055615,
+      "epoch": 0.5948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 8.045061983068473e-06,
-      "loss": 0.8015,
+      "learning_rate": 7.441561771237333e-06,
+      "loss": 0.8658,
       "step": 20964
     },
     {
-      "epoch": 0.5757875367334048,
+      "epoch": 0.5949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 8.044189638058542e-06,
-      "loss": 0.8029,
+      "learning_rate": 7.440673292614717e-06,
+      "loss": 0.9253,
       "step": 20965
     },
     {
-      "epoch": 0.5758150009612479,
+      "epoch": 0.5949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 8.043317308523736e-06,
-      "loss": 0.9281,
+      "learning_rate": 7.439784835610853e-06,
+      "loss": 0.9403,
       "step": 20966
     },
     {
-      "epoch": 0.5758424651890912,
+      "epoch": 0.5949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 8.042444994470961e-06,
-      "loss": 0.8008,
+      "learning_rate": 7.4388964002332395e-06,
+      "loss": 0.8617,
       "step": 20967
     },
     {
-      "epoch": 0.5758699294169345,
+      "epoch": 0.5950056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 8.041572695907122e-06,
-      "loss": 0.8863,
+      "learning_rate": 7.438007986489384e-06,
+      "loss": 0.9445,
       "step": 20968
     },
     {
-      "epoch": 0.5758973936447777,
+      "epoch": 0.5950340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 8.04070041283912e-06,
-      "loss": 0.8423,
+      "learning_rate": 7.43711959438679e-06,
+      "loss": 0.7961,
       "step": 20969
     },
     {
-      "epoch": 0.5759248578726209,
+      "epoch": 0.5950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 8.039828145273856e-06,
-      "loss": 0.9192,
+      "learning_rate": 7.436231223932964e-06,
+      "loss": 0.9029,
       "step": 20970
     },
     {
-      "epoch": 0.5759523221004641,
+      "epoch": 0.5950908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 8.038955893218226e-06,
-      "loss": 0.8863,
+      "learning_rate": 7.435342875135405e-06,
+      "loss": 0.894,
       "step": 20971
     },
     {
-      "epoch": 0.5759797863283074,
+      "epoch": 0.5951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 8.038083656679137e-06,
-      "loss": 0.8293,
+      "learning_rate": 7.434454548001622e-06,
+      "loss": 0.8247,
       "step": 20972
     },
     {
-      "epoch": 0.5760072505561507,
+      "epoch": 0.5951475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 8.037211435663492e-06,
-      "loss": 0.8834,
+      "learning_rate": 7.433566242539117e-06,
+      "loss": 0.9065,
       "step": 20973
     },
     {
-      "epoch": 0.5760347147839938,
+      "epoch": 0.5951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 8.03633923017819e-06,
-      "loss": 0.9614,
+      "learning_rate": 7.432677958755391e-06,
+      "loss": 0.8362,
       "step": 20974
     },
     {
-      "epoch": 0.5760621790118371,
+      "epoch": 0.5952043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 8.035467040230136e-06,
-      "loss": 0.8989,
+      "learning_rate": 7.431789696657952e-06,
+      "loss": 0.8802,
       "step": 20975
     },
     {
-      "epoch": 0.5760896432396804,
+      "epoch": 0.5952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 8.034594865826224e-06,
-      "loss": 0.875,
+      "learning_rate": 7.4309014562543e-06,
+      "loss": 0.8882,
       "step": 20976
     },
     {
-      "epoch": 0.5761171074675235,
+      "epoch": 0.595261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.03372270697336e-06,
-      "loss": 0.9291,
+      "learning_rate": 7.430013237551938e-06,
+      "loss": 0.7493,
       "step": 20977
     },
     {
-      "epoch": 0.5761445716953668,
+      "epoch": 0.5952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.032850563678443e-06,
-      "loss": 0.7727,
+      "learning_rate": 7.429125040558372e-06,
+      "loss": 0.844,
       "step": 20978
     },
     {
-      "epoch": 0.57617203592321,
+      "epoch": 0.5953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 8.031978435948373e-06,
-      "loss": 0.8815,
+      "learning_rate": 7.428236865281102e-06,
+      "loss": 0.6879,
       "step": 20979
     },
     {
-      "epoch": 0.5761995001510533,
+      "epoch": 0.5953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 8.031106323790055e-06,
-      "loss": 0.8683,
+      "learning_rate": 7.427348711727627e-06,
+      "loss": 0.7707,
       "step": 20980
     },
     {
-      "epoch": 0.5762269643788965,
+      "epoch": 0.5953745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 8.030234227210387e-06,
-      "loss": 0.823,
+      "learning_rate": 7.4264605799054556e-06,
+      "loss": 0.9234,
       "step": 20981
     },
     {
-      "epoch": 0.5762544286067397,
+      "epoch": 0.5954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 8.029362146216266e-06,
-      "loss": 0.8834,
+      "learning_rate": 7.425572469822087e-06,
+      "loss": 0.91,
       "step": 20982
     },
     {
-      "epoch": 0.576281892834583,
+      "epoch": 0.5954313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 8.028490080814596e-06,
-      "loss": 1.0203,
+      "learning_rate": 7.424684381485023e-06,
+      "loss": 0.8552,
       "step": 20983
     },
     {
-      "epoch": 0.5763093570624261,
+      "epoch": 0.5954597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 8.027618031012273e-06,
-      "loss": 0.8916,
+      "learning_rate": 7.423796314901769e-06,
+      "loss": 0.8568,
       "step": 20984
     },
     {
-      "epoch": 0.5763368212902694,
+      "epoch": 0.5954880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 8.026745996816204e-06,
-      "loss": 0.8357,
+      "learning_rate": 7.4229082700798196e-06,
+      "loss": 0.8685,
       "step": 20985
     },
     {
-      "epoch": 0.5763642855181127,
+      "epoch": 0.5955164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 8.025873978233287e-06,
-      "loss": 0.8188,
+      "learning_rate": 7.422020247026682e-06,
+      "loss": 0.8534,
       "step": 20986
     },
     {
-      "epoch": 0.5763917497459559,
+      "epoch": 0.5955448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 8.025001975270416e-06,
-      "loss": 0.9067,
+      "learning_rate": 7.421132245749856e-06,
+      "loss": 0.9041,
       "step": 20987
     },
     {
-      "epoch": 0.5764192139737991,
+      "epoch": 0.5955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 8.024129987934495e-06,
-      "loss": 1.0186,
+      "learning_rate": 7.42024426625684e-06,
+      "loss": 0.9379,
       "step": 20988
     },
     {
-      "epoch": 0.5764466782016424,
+      "epoch": 0.5956015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 8.023258016232424e-06,
-      "loss": 0.8464,
+      "learning_rate": 7.419356308555137e-06,
+      "loss": 0.7881,
       "step": 20989
     },
     {
-      "epoch": 0.5764741424294856,
+      "epoch": 0.5956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 8.0223860601711e-06,
-      "loss": 0.8886,
+      "learning_rate": 7.418468372652249e-06,
+      "loss": 0.8698,
       "step": 20990
     },
     {
-      "epoch": 0.5765016066573289,
+      "epoch": 0.5956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 8.021514119757422e-06,
-      "loss": 0.7943,
+      "learning_rate": 7.417580458555671e-06,
+      "loss": 0.8489,
       "step": 20991
     },
     {
-      "epoch": 0.576529070885172,
+      "epoch": 0.5956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 8.020642194998296e-06,
-      "loss": 0.7972,
+      "learning_rate": 7.416692566272911e-06,
+      "loss": 0.901,
       "step": 20992
     },
     {
-      "epoch": 0.5765565351130153,
+      "epoch": 0.5957150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 8.019770285900613e-06,
-      "loss": 0.8239,
+      "learning_rate": 7.4158046958114635e-06,
+      "loss": 0.8755,
       "step": 20993
     },
     {
-      "epoch": 0.5765839993408586,
+      "epoch": 0.5957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 8.018898392471273e-06,
-      "loss": 0.8469,
+      "learning_rate": 7.414916847178828e-06,
+      "loss": 0.9541,
       "step": 20994
     },
     {
-      "epoch": 0.5766114635687017,
+      "epoch": 0.595771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 8.018026514717176e-06,
-      "loss": 0.8511,
+      "learning_rate": 7.414029020382505e-06,
+      "loss": 0.8213,
       "step": 20995
     },
     {
-      "epoch": 0.576638927796545,
+      "epoch": 0.595800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 8.017154652645221e-06,
-      "loss": 0.8433,
+      "learning_rate": 7.413141215429998e-06,
+      "loss": 0.8539,
       "step": 20996
     },
     {
-      "epoch": 0.5766663920243882,
+      "epoch": 0.5958286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 8.01628280626231e-06,
-      "loss": 0.8239,
+      "learning_rate": 7.4122534323288044e-06,
+      "loss": 0.914,
       "step": 20997
     },
     {
-      "epoch": 0.5766938562522315,
+      "epoch": 0.5958569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 8.015410975575337e-06,
-      "loss": 0.8578,
+      "learning_rate": 7.41136567108642e-06,
+      "loss": 0.8411,
       "step": 20998
     },
     {
-      "epoch": 0.5767213204800747,
+      "epoch": 0.5958853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 8.0145391605912e-06,
-      "loss": 0.8472,
+      "learning_rate": 7.410477931710348e-06,
+      "loss": 0.8358,
       "step": 20999
     },
     {
-      "epoch": 0.5767487847079179,
+      "epoch": 0.5959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 8.013667361316795e-06,
-      "loss": 0.9069,
+      "learning_rate": 7.409590214208087e-06,
+      "loss": 0.8481,
       "step": 21000
     },
     {
-      "epoch": 0.5767762489357612,
+      "epoch": 0.5959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 8.012795577759029e-06,
-      "loss": 0.7951,
+      "learning_rate": 7.408702518587132e-06,
+      "loss": 0.8689,
       "step": 21001
     },
     {
-      "epoch": 0.5768037131636045,
+      "epoch": 0.5959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 8.01192380992479e-06,
-      "loss": 0.8991,
+      "learning_rate": 7.407814844854981e-06,
+      "loss": 0.8039,
       "step": 21002
     },
     {
-      "epoch": 0.5768311773914476,
+      "epoch": 0.5959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 8.01105205782098e-06,
-      "loss": 0.8979,
+      "learning_rate": 7.406927193019138e-06,
+      "loss": 0.9202,
       "step": 21003
     },
     {
-      "epoch": 0.5768586416192909,
+      "epoch": 0.5960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 8.010180321454501e-06,
-      "loss": 0.898,
+      "learning_rate": 7.4060395630870965e-06,
+      "loss": 0.9106,
       "step": 21004
     },
     {
-      "epoch": 0.5768861058471341,
+      "epoch": 0.59605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.009308600832243e-06,
-      "loss": 0.9678,
+      "learning_rate": 7.405151955066353e-06,
+      "loss": 0.8987,
       "step": 21005
     },
     {
-      "epoch": 0.5769135700749773,
+      "epoch": 0.5960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 8.008436895961105e-06,
-      "loss": 0.8493,
+      "learning_rate": 7.404264368964411e-06,
+      "loss": 0.8525,
       "step": 21006
     },
     {
-      "epoch": 0.5769410343028206,
+      "epoch": 0.5961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 8.007565206847987e-06,
-      "loss": 0.8952,
+      "learning_rate": 7.403376804788764e-06,
+      "loss": 0.8427,
       "step": 21007
     },
     {
-      "epoch": 0.5769684985306638,
+      "epoch": 0.5961407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 8.006693533499783e-06,
-      "loss": 0.8189,
+      "learning_rate": 7.402489262546908e-06,
+      "loss": 0.9354,
       "step": 21008
     },
     {
-      "epoch": 0.5769959627585071,
+      "epoch": 0.5961691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 8.005821875923396e-06,
-      "loss": 0.9624,
+      "learning_rate": 7.401601742246343e-06,
+      "loss": 0.8809,
       "step": 21009
     },
     {
-      "epoch": 0.5770234269863502,
+      "epoch": 0.5961975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 8.00495023412572e-06,
-      "loss": 0.9554,
+      "learning_rate": 7.400714243894565e-06,
+      "loss": 0.799,
       "step": 21010
     },
     {
-      "epoch": 0.5770508912141935,
+      "epoch": 0.5962258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 8.004078608113645e-06,
-      "loss": 0.8512,
+      "learning_rate": 7.399826767499068e-06,
+      "loss": 0.8955,
       "step": 21011
     },
     {
-      "epoch": 0.5770783554420368,
+      "epoch": 0.5962542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 8.003206997894075e-06,
-      "loss": 0.9465,
+      "learning_rate": 7.398939313067353e-06,
+      "loss": 0.8341,
       "step": 21012
     },
     {
-      "epoch": 0.57710581966988,
+      "epoch": 0.5962826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 8.002335403473906e-06,
-      "loss": 0.8144,
+      "learning_rate": 7.3980518806069155e-06,
+      "loss": 0.9062,
       "step": 21013
     },
     {
-      "epoch": 0.5771332838977232,
+      "epoch": 0.596311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 8.001463824860032e-06,
-      "loss": 0.9333,
+      "learning_rate": 7.397164470125251e-06,
+      "loss": 0.8022,
       "step": 21014
     },
     {
-      "epoch": 0.5771607481255665,
+      "epoch": 0.5963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 8.000592262059348e-06,
-      "loss": 0.8604,
+      "learning_rate": 7.396277081629852e-06,
+      "loss": 0.9981,
       "step": 21015
     },
     {
-      "epoch": 0.5771882123534097,
+      "epoch": 0.5963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 7.999720715078759e-06,
-      "loss": 0.9179,
+      "learning_rate": 7.395389715128223e-06,
+      "loss": 0.8552,
       "step": 21016
     },
     {
-      "epoch": 0.577215676581253,
+      "epoch": 0.5963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 7.998849183925152e-06,
-      "loss": 0.8665,
+      "learning_rate": 7.394502370627852e-06,
+      "loss": 0.8787,
       "step": 21017
     },
     {
-      "epoch": 0.5772431408090961,
+      "epoch": 0.5964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 7.997977668605424e-06,
-      "loss": 0.9007,
+      "learning_rate": 7.393615048136234e-06,
+      "loss": 0.9266,
       "step": 21018
     },
     {
-      "epoch": 0.5772706050369394,
+      "epoch": 0.5964528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 7.99710616912647e-06,
-      "loss": 0.9068,
+      "learning_rate": 7.392727747660869e-06,
+      "loss": 0.9401,
       "step": 21019
     },
     {
-      "epoch": 0.5772980692647827,
+      "epoch": 0.5964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 7.996234685495192e-06,
-      "loss": 0.8333,
+      "learning_rate": 7.3918404692092504e-06,
+      "loss": 0.8739,
       "step": 21020
     },
     {
-      "epoch": 0.5773255334926258,
+      "epoch": 0.5965096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 7.995363217718477e-06,
-      "loss": 0.7995,
+      "learning_rate": 7.390953212788872e-06,
+      "loss": 0.8987,
       "step": 21021
     },
     {
-      "epoch": 0.5773529977204691,
+      "epoch": 0.5965380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 7.994491765803231e-06,
-      "loss": 0.9476,
+      "learning_rate": 7.390065978407228e-06,
+      "loss": 0.8746,
       "step": 21022
     },
     {
-      "epoch": 0.5773804619483123,
+      "epoch": 0.5965664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 7.993620329756337e-06,
-      "loss": 0.9471,
+      "learning_rate": 7.389178766071816e-06,
+      "loss": 0.9202,
       "step": 21023
     },
     {
-      "epoch": 0.5774079261761556,
+      "epoch": 0.596594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 7.992748909584698e-06,
-      "loss": 0.8825,
+      "learning_rate": 7.388291575790127e-06,
+      "loss": 0.8384,
       "step": 21024
     },
     {
-      "epoch": 0.5774353904039988,
+      "epoch": 0.5966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 7.991877505295208e-06,
-      "loss": 0.7496,
+      "learning_rate": 7.387404407569657e-06,
+      "loss": 0.8267,
       "step": 21025
     },
     {
-      "epoch": 0.577462854631842,
+      "epoch": 0.5966515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 7.99100611689476e-06,
-      "loss": 0.9076,
+      "learning_rate": 7.386517261417896e-06,
+      "loss": 0.8141,
       "step": 21026
     },
     {
-      "epoch": 0.5774903188596853,
+      "epoch": 0.5966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 7.99013474439025e-06,
-      "loss": 0.8192,
+      "learning_rate": 7.385630137342346e-06,
+      "loss": 0.7858,
       "step": 21027
     },
     {
-      "epoch": 0.5775177830875285,
+      "epoch": 0.5967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 7.989263387788577e-06,
-      "loss": 0.8665,
+      "learning_rate": 7.384743035350495e-06,
+      "loss": 0.9184,
       "step": 21028
     },
     {
-      "epoch": 0.5775452473153717,
+      "epoch": 0.5967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.988392047096627e-06,
-      "loss": 0.8785,
+      "learning_rate": 7.383855955449835e-06,
+      "loss": 0.909,
       "step": 21029
     },
     {
-      "epoch": 0.577572711543215,
+      "epoch": 0.5967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 7.987520722321297e-06,
-      "loss": 0.9542,
+      "learning_rate": 7.382968897647862e-06,
+      "loss": 0.8335,
       "step": 21030
     },
     {
-      "epoch": 0.5776001757710582,
+      "epoch": 0.5967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 7.986649413469484e-06,
-      "loss": 0.7817,
+      "learning_rate": 7.38208186195207e-06,
+      "loss": 0.7595,
       "step": 21031
     },
     {
-      "epoch": 0.5776276399989014,
+      "epoch": 0.596821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 7.98577812054808e-06,
-      "loss": 0.8123,
+      "learning_rate": 7.381194848369948e-06,
+      "loss": 0.883,
       "step": 21032
     },
     {
-      "epoch": 0.5776551042267447,
+      "epoch": 0.596850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 7.98490684356398e-06,
-      "loss": 0.8856,
+      "learning_rate": 7.3803078569089926e-06,
+      "loss": 0.933,
       "step": 21033
     },
     {
-      "epoch": 0.5776825684545879,
+      "epoch": 0.5968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 7.98403558252408e-06,
-      "loss": 0.8408,
+      "learning_rate": 7.379420887576696e-06,
+      "loss": 0.9257,
       "step": 21034
     },
     {
-      "epoch": 0.5777100326824312,
+      "epoch": 0.5969069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 7.983164337435268e-06,
-      "loss": 0.9686,
+      "learning_rate": 7.378533940380545e-06,
+      "loss": 0.9138,
       "step": 21035
     },
     {
-      "epoch": 0.5777374969102743,
+      "epoch": 0.5969353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 7.982293108304442e-06,
-      "loss": 0.8323,
+      "learning_rate": 7.3776470153280375e-06,
+      "loss": 0.8559,
       "step": 21036
     },
     {
-      "epoch": 0.5777649611381176,
+      "epoch": 0.5969636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 7.981421895138497e-06,
-      "loss": 0.9299,
+      "learning_rate": 7.376760112426665e-06,
+      "loss": 0.9651,
       "step": 21037
     },
     {
-      "epoch": 0.5777924253659609,
+      "epoch": 0.5969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 7.98055069794432e-06,
-      "loss": 0.943,
+      "learning_rate": 7.3758732316839155e-06,
+      "loss": 0.8026,
       "step": 21038
     },
     {
-      "epoch": 0.577819889593804,
+      "epoch": 0.5970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 7.979679516728807e-06,
-      "loss": 0.8827,
+      "learning_rate": 7.3749863731072845e-06,
+      "loss": 0.881,
       "step": 21039
     },
     {
-      "epoch": 0.5778473538216473,
+      "epoch": 0.5970488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 7.978808351498858e-06,
-      "loss": 0.9191,
+      "learning_rate": 7.374099536704261e-06,
+      "loss": 0.8934,
       "step": 21040
     },
     {
-      "epoch": 0.5778748180494906,
+      "epoch": 0.5970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 7.97793720226136e-06,
-      "loss": 0.8412,
+      "learning_rate": 7.373212722482334e-06,
+      "loss": 0.9094,
       "step": 21041
     },
     {
-      "epoch": 0.5779022822773338,
+      "epoch": 0.597105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 7.977066069023201e-06,
-      "loss": 0.8974,
+      "learning_rate": 7.372325930449e-06,
+      "loss": 0.9732,
       "step": 21042
     },
     {
-      "epoch": 0.577929746505177,
+      "epoch": 0.5971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 7.97619495179128e-06,
-      "loss": 0.8918,
+      "learning_rate": 7.371439160611745e-06,
+      "loss": 0.7283,
       "step": 21043
     },
     {
-      "epoch": 0.5779572107330202,
+      "epoch": 0.5971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 7.975323850572492e-06,
-      "loss": 0.8405,
+      "learning_rate": 7.370552412978065e-06,
+      "loss": 0.9147,
       "step": 21044
     },
     {
-      "epoch": 0.5779846749608635,
+      "epoch": 0.5971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 7.974452765373725e-06,
-      "loss": 0.8773,
+      "learning_rate": 7.369665687555445e-06,
+      "loss": 0.9657,
       "step": 21045
     },
     {
-      "epoch": 0.5780121391887068,
+      "epoch": 0.5972190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 7.973581696201872e-06,
-      "loss": 0.9664,
+      "learning_rate": 7.368778984351377e-06,
+      "loss": 0.864,
       "step": 21046
     },
     {
-      "epoch": 0.5780396034165499,
+      "epoch": 0.5972474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 7.972710643063824e-06,
-      "loss": 0.9458,
+      "learning_rate": 7.367892303373352e-06,
+      "loss": 0.8335,
       "step": 21047
     },
     {
-      "epoch": 0.5780670676443932,
+      "epoch": 0.5972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 7.971839605966473e-06,
-      "loss": 0.9229,
+      "learning_rate": 7.36700564462886e-06,
+      "loss": 0.9912,
       "step": 21048
     },
     {
-      "epoch": 0.5780945318722364,
+      "epoch": 0.5973041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 7.970968584916717e-06,
-      "loss": 0.9563,
+      "learning_rate": 7.366119008125386e-06,
+      "loss": 0.869,
       "step": 21049
     },
     {
-      "epoch": 0.5781219961000796,
+      "epoch": 0.5973325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 7.970097579921441e-06,
-      "loss": 0.8504,
+      "learning_rate": 7.3652323938704275e-06,
+      "loss": 0.982,
       "step": 21050
     },
     {
-      "epoch": 0.5781494603279229,
+      "epoch": 0.597360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 7.969226590987543e-06,
-      "loss": 0.805,
+      "learning_rate": 7.364345801871468e-06,
+      "loss": 0.7779,
       "step": 21051
     },
     {
-      "epoch": 0.5781769245557661,
+      "epoch": 0.5973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 7.968355618121905e-06,
-      "loss": 0.894,
+      "learning_rate": 7.363459232135996e-06,
+      "loss": 1.0058,
       "step": 21052
     },
     {
-      "epoch": 0.5782043887836094,
+      "epoch": 0.5974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 7.96748466133143e-06,
-      "loss": 0.9268,
+      "learning_rate": 7.362572684671504e-06,
+      "loss": 0.8296,
       "step": 21053
     },
     {
-      "epoch": 0.5782318530114526,
+      "epoch": 0.5974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 7.966613720622999e-06,
-      "loss": 0.8191,
+      "learning_rate": 7.361686159485479e-06,
+      "loss": 0.8203,
       "step": 21054
     },
     {
-      "epoch": 0.5782593172392958,
+      "epoch": 0.5974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 7.965742796003509e-06,
-      "loss": 0.8594,
+      "learning_rate": 7.360799656585409e-06,
+      "loss": 0.7727,
       "step": 21055
     },
     {
-      "epoch": 0.5782867814671391,
+      "epoch": 0.5975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 7.964871887479851e-06,
-      "loss": 0.958,
+      "learning_rate": 7.3599131759787835e-06,
+      "loss": 0.8243,
       "step": 21056
     },
     {
-      "epoch": 0.5783142456949822,
+      "epoch": 0.5975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 7.964000995058916e-06,
-      "loss": 0.8718,
+      "learning_rate": 7.3590267176730875e-06,
+      "loss": 0.8835,
       "step": 21057
     },
     {
-      "epoch": 0.5783417099228255,
+      "epoch": 0.5975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 7.963130118747592e-06,
-      "loss": 0.8319,
+      "learning_rate": 7.358140281675814e-06,
+      "loss": 0.8774,
       "step": 21058
     },
     {
-      "epoch": 0.5783691741506688,
+      "epoch": 0.5975879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 7.962259258552773e-06,
-      "loss": 0.9573,
+      "learning_rate": 7.3572538679944475e-06,
+      "loss": 0.9201,
       "step": 21059
     },
     {
-      "epoch": 0.578396638378512,
+      "epoch": 0.5976163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 7.961388414481344e-06,
-      "loss": 0.8304,
+      "learning_rate": 7.356367476636478e-06,
+      "loss": 0.8416,
       "step": 21060
     },
     {
-      "epoch": 0.5784241026063552,
+      "epoch": 0.597644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 7.960517586540205e-06,
-      "loss": 0.7943,
+      "learning_rate": 7.355481107609391e-06,
+      "loss": 0.7586,
       "step": 21061
     },
     {
-      "epoch": 0.5784515668341984,
+      "epoch": 0.5976730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 7.959646774736237e-06,
-      "loss": 0.8797,
+      "learning_rate": 7.354594760920673e-06,
+      "loss": 0.8672,
       "step": 21062
     },
     {
-      "epoch": 0.5784790310620417,
+      "epoch": 0.5977014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 7.958775979076338e-06,
-      "loss": 0.749,
+      "learning_rate": 7.353708436577813e-06,
+      "loss": 0.9131,
       "step": 21063
     },
     {
-      "epoch": 0.578506495289885,
+      "epoch": 0.5977298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 7.95790519956739e-06,
-      "loss": 0.8694,
+      "learning_rate": 7.352822134588297e-06,
+      "loss": 0.8806,
       "step": 21064
     },
     {
-      "epoch": 0.5785339595177281,
+      "epoch": 0.5977582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 7.957034436216291e-06,
-      "loss": 0.8839,
+      "learning_rate": 7.351935854959608e-06,
+      "loss": 0.8114,
       "step": 21065
     },
     {
-      "epoch": 0.5785614237455714,
+      "epoch": 0.5977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 7.956163689029924e-06,
-      "loss": 0.8536,
+      "learning_rate": 7.35104959769924e-06,
+      "loss": 0.9058,
       "step": 21066
     },
     {
-      "epoch": 0.5785888879734147,
+      "epoch": 0.5978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 7.955292958015181e-06,
-      "loss": 0.8353,
+      "learning_rate": 7.350163362814675e-06,
+      "loss": 0.7173,
       "step": 21067
     },
     {
-      "epoch": 0.5786163522012578,
+      "epoch": 0.5978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 7.954422243178954e-06,
-      "loss": 0.8572,
+      "learning_rate": 7.349277150313398e-06,
+      "loss": 0.8953,
       "step": 21068
     },
     {
-      "epoch": 0.5786438164291011,
+      "epoch": 0.5978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 7.953551544528131e-06,
-      "loss": 0.8344,
+      "learning_rate": 7.348390960202896e-06,
+      "loss": 0.8376,
       "step": 21069
     },
     {
-      "epoch": 0.5786712806569443,
+      "epoch": 0.5979001135073779,
       "grad_norm": 0.0,
-      "learning_rate": 7.952680862069601e-06,
-      "loss": 0.8472,
+      "learning_rate": 7.347504792490654e-06,
+      "loss": 0.9524,
       "step": 21070
     },
     {
-      "epoch": 0.5786987448847876,
+      "epoch": 0.5979284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 7.951810195810251e-06,
-      "loss": 0.93,
+      "learning_rate": 7.346618647184161e-06,
+      "loss": 0.934,
       "step": 21071
     },
     {
-      "epoch": 0.5787262091126308,
+      "epoch": 0.5979568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 7.950939545756972e-06,
-      "loss": 0.8758,
+      "learning_rate": 7.345732524290894e-06,
+      "loss": 0.8901,
       "step": 21072
     },
     {
-      "epoch": 0.578753673340474,
+      "epoch": 0.5979852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 7.950068911916655e-06,
-      "loss": 0.8571,
+      "learning_rate": 7.3448464238183455e-06,
+      "loss": 0.8495,
       "step": 21073
     },
     {
-      "epoch": 0.5787811375683173,
+      "epoch": 0.5980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 7.949198294296185e-06,
-      "loss": 0.925,
+      "learning_rate": 7.343960345774001e-06,
+      "loss": 0.9805,
       "step": 21074
     },
     {
-      "epoch": 0.5788086017961604,
+      "epoch": 0.5980419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 7.948327692902457e-06,
-      "loss": 0.8959,
+      "learning_rate": 7.343074290165343e-06,
+      "loss": 0.949,
       "step": 21075
     },
     {
-      "epoch": 0.5788360660240037,
+      "epoch": 0.5980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 7.94745710774235e-06,
-      "loss": 0.8493,
+      "learning_rate": 7.342188256999853e-06,
+      "loss": 0.7699,
       "step": 21076
     },
     {
-      "epoch": 0.578863530251847,
+      "epoch": 0.5980987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 7.94658653882276e-06,
-      "loss": 0.9822,
+      "learning_rate": 7.341302246285022e-06,
+      "loss": 0.8001,
       "step": 21077
     },
     {
-      "epoch": 0.5788909944796902,
+      "epoch": 0.5981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 7.94571598615057e-06,
-      "loss": 0.932,
+      "learning_rate": 7.3404162580283285e-06,
+      "loss": 0.8087,
       "step": 21078
     },
     {
-      "epoch": 0.5789184587075334,
+      "epoch": 0.598155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 7.944845449732672e-06,
-      "loss": 0.8884,
+      "learning_rate": 7.339530292237258e-06,
+      "loss": 0.8258,
       "step": 21079
     },
     {
-      "epoch": 0.5789459229353767,
+      "epoch": 0.5981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 7.943974929575955e-06,
-      "loss": 0.9898,
+      "learning_rate": 7.338644348919296e-06,
+      "loss": 0.856,
       "step": 21080
     },
     {
-      "epoch": 0.5789733871632199,
+      "epoch": 0.5982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 7.943104425687303e-06,
-      "loss": 0.8576,
+      "learning_rate": 7.337758428081925e-06,
+      "loss": 0.7743,
       "step": 21081
     },
     {
-      "epoch": 0.5790008513910632,
+      "epoch": 0.5982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 7.942233938073607e-06,
-      "loss": 0.9111,
+      "learning_rate": 7.336872529732626e-06,
+      "loss": 1.0128,
       "step": 21082
     },
     {
-      "epoch": 0.5790283156189063,
+      "epoch": 0.5982690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 7.941363466741751e-06,
-      "loss": 0.8932,
+      "learning_rate": 7.335986653878887e-06,
+      "loss": 0.94,
       "step": 21083
     },
     {
-      "epoch": 0.5790557798467496,
+      "epoch": 0.5982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 7.940493011698625e-06,
-      "loss": 0.8889,
+      "learning_rate": 7.3351008005281875e-06,
+      "loss": 0.766,
       "step": 21084
     },
     {
-      "epoch": 0.5790832440745929,
+      "epoch": 0.5983257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 7.939622572951118e-06,
-      "loss": 0.8402,
+      "learning_rate": 7.3342149696880106e-06,
+      "loss": 0.9022,
       "step": 21085
     },
     {
-      "epoch": 0.579110708302436,
+      "epoch": 0.5983541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 7.938752150506113e-06,
-      "loss": 0.9321,
+      "learning_rate": 7.333329161365841e-06,
+      "loss": 0.7765,
       "step": 21086
     },
     {
-      "epoch": 0.5791381725302793,
+      "epoch": 0.5983825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 7.937881744370503e-06,
-      "loss": 0.7483,
+      "learning_rate": 7.332443375569156e-06,
+      "loss": 0.7822,
       "step": 21087
     },
     {
-      "epoch": 0.5791656367581225,
+      "epoch": 0.5984108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 7.937011354551168e-06,
-      "loss": 0.8189,
+      "learning_rate": 7.331557612305445e-06,
+      "loss": 0.8618,
       "step": 21088
     },
     {
-      "epoch": 0.5791931009859658,
+      "epoch": 0.5984392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 7.936140981055e-06,
-      "loss": 0.9883,
+      "learning_rate": 7.330671871582188e-06,
+      "loss": 0.7782,
       "step": 21089
     },
     {
-      "epoch": 0.579220565213809,
+      "epoch": 0.5984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 7.935270623888881e-06,
-      "loss": 0.7929,
+      "learning_rate": 7.3297861534068636e-06,
+      "loss": 0.7588,
       "step": 21090
     },
     {
-      "epoch": 0.5792480294416522,
+      "epoch": 0.5984960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 7.934400283059703e-06,
-      "loss": 0.9504,
+      "learning_rate": 7.328900457786956e-06,
+      "loss": 0.7634,
       "step": 21091
     },
     {
-      "epoch": 0.5792754936694955,
+      "epoch": 0.5985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 7.933529958574351e-06,
-      "loss": 0.9417,
+      "learning_rate": 7.328014784729949e-06,
+      "loss": 0.8538,
       "step": 21092
     },
     {
-      "epoch": 0.5793029578973388,
+      "epoch": 0.5985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 7.932659650439711e-06,
-      "loss": 0.8059,
+      "learning_rate": 7.327129134243317e-06,
+      "loss": 0.9666,
       "step": 21093
     },
     {
-      "epoch": 0.5793304221251819,
+      "epoch": 0.5985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 7.931789358662668e-06,
-      "loss": 0.8217,
+      "learning_rate": 7.326243506334549e-06,
+      "loss": 0.8461,
       "step": 21094
     },
     {
-      "epoch": 0.5793578863530252,
+      "epoch": 0.5986095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 7.930919083250106e-06,
-      "loss": 0.8419,
+      "learning_rate": 7.3253579010111205e-06,
+      "loss": 0.8222,
       "step": 21095
     },
     {
-      "epoch": 0.5793853505808684,
+      "epoch": 0.5986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 7.930048824208914e-06,
-      "loss": 0.811,
+      "learning_rate": 7.324472318280513e-06,
+      "loss": 0.8793,
       "step": 21096
     },
     {
-      "epoch": 0.5794128148087117,
+      "epoch": 0.59866628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.929178581545981e-06,
-      "loss": 0.9336,
+      "learning_rate": 7.3235867581502095e-06,
+      "loss": 0.8083,
       "step": 21097
     },
     {
-      "epoch": 0.5794402790365549,
+      "epoch": 0.598694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 7.928308355268187e-06,
-      "loss": 0.8483,
+      "learning_rate": 7.32270122062769e-06,
+      "loss": 0.9134,
       "step": 21098
     },
     {
-      "epoch": 0.5794677432643981,
+      "epoch": 0.5987230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 7.927438145382423e-06,
-      "loss": 0.8917,
+      "learning_rate": 7.321815705720431e-06,
+      "loss": 0.91,
       "step": 21099
     },
     {
-      "epoch": 0.5794952074922414,
+      "epoch": 0.5987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 7.926567951895566e-06,
-      "loss": 0.8546,
+      "learning_rate": 7.320930213435917e-06,
+      "loss": 0.8846,
       "step": 21100
     },
     {
-      "epoch": 0.5795226717200845,
+      "epoch": 0.5987797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 7.925697774814511e-06,
-      "loss": 0.8507,
+      "learning_rate": 7.320044743781627e-06,
+      "loss": 0.7945,
       "step": 21101
     },
     {
-      "epoch": 0.5795501359479278,
+      "epoch": 0.5988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 7.924827614146134e-06,
-      "loss": 0.7957,
+      "learning_rate": 7.319159296765036e-06,
+      "loss": 0.7524,
       "step": 21102
     },
     {
-      "epoch": 0.5795776001757711,
+      "epoch": 0.5988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 7.923957469897326e-06,
-      "loss": 0.7943,
+      "learning_rate": 7.3182738723936255e-06,
+      "loss": 0.7951,
       "step": 21103
     },
     {
-      "epoch": 0.5796050644036143,
+      "epoch": 0.5988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 7.923087342074976e-06,
-      "loss": 0.9506,
+      "learning_rate": 7.31738847067488e-06,
+      "loss": 0.782,
       "step": 21104
     },
     {
-      "epoch": 0.5796325286314575,
+      "epoch": 0.5988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 7.922217230685957e-06,
-      "loss": 0.8557,
+      "learning_rate": 7.3165030916162745e-06,
+      "loss": 0.8727,
       "step": 21105
     },
     {
-      "epoch": 0.5796599928593008,
+      "epoch": 0.5989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 7.921347135737165e-06,
-      "loss": 0.7327,
+      "learning_rate": 7.315617735225287e-06,
+      "loss": 0.8523,
       "step": 21106
     },
     {
-      "epoch": 0.579687457087144,
+      "epoch": 0.598950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 7.920477057235477e-06,
-      "loss": 0.9567,
+      "learning_rate": 7.314732401509399e-06,
+      "loss": 0.8695,
       "step": 21107
     },
     {
-      "epoch": 0.5797149213149873,
+      "epoch": 0.5989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 7.919606995187777e-06,
-      "loss": 0.8228,
+      "learning_rate": 7.313847090476086e-06,
+      "loss": 0.8361,
       "step": 21108
     },
     {
-      "epoch": 0.5797423855428304,
+      "epoch": 0.5990068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 7.918736949600957e-06,
-      "loss": 0.8709,
+      "learning_rate": 7.3129618021328286e-06,
+      "loss": 0.7537,
       "step": 21109
     },
     {
-      "epoch": 0.5797698497706737,
+      "epoch": 0.5990351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 7.917866920481898e-06,
-      "loss": 0.8355,
+      "learning_rate": 7.3120765364871e-06,
+      "loss": 0.8471,
       "step": 21110
     },
     {
-      "epoch": 0.579797313998517,
+      "epoch": 0.5990635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 7.916996907837478e-06,
-      "loss": 0.8961,
+      "learning_rate": 7.3111912935463854e-06,
+      "loss": 0.9148,
       "step": 21111
     },
     {
-      "epoch": 0.5798247782263601,
+      "epoch": 0.5990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 7.916126911674587e-06,
-      "loss": 0.9746,
+      "learning_rate": 7.310306073318159e-06,
+      "loss": 0.8537,
       "step": 21112
     },
     {
-      "epoch": 0.5798522424542034,
+      "epoch": 0.5991203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 7.915256932000108e-06,
-      "loss": 0.9246,
+      "learning_rate": 7.309420875809895e-06,
+      "loss": 0.8574,
       "step": 21113
     },
     {
-      "epoch": 0.5798797066820466,
+      "epoch": 0.5991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 7.91438696882092e-06,
-      "loss": 0.8919,
+      "learning_rate": 7.3085357010290755e-06,
+      "loss": 0.9522,
       "step": 21114
     },
     {
-      "epoch": 0.5799071709098899,
+      "epoch": 0.5991770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 7.913517022143912e-06,
-      "loss": 0.8604,
+      "learning_rate": 7.3076505489831775e-06,
+      "loss": 0.9428,
       "step": 21115
     },
     {
-      "epoch": 0.5799346351377331,
+      "epoch": 0.599205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 7.91264709197597e-06,
-      "loss": 0.986,
+      "learning_rate": 7.306765419679674e-06,
+      "loss": 0.8979,
       "step": 21116
     },
     {
-      "epoch": 0.5799620993655763,
+      "epoch": 0.5992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 7.911777178323967e-06,
-      "loss": 0.8289,
+      "learning_rate": 7.305880313126045e-06,
+      "loss": 0.8057,
       "step": 21117
     },
     {
-      "epoch": 0.5799895635934196,
+      "epoch": 0.5992622020431329,
       "grad_norm": 0.0,
-      "learning_rate": 7.910907281194796e-06,
-      "loss": 0.7758,
+      "learning_rate": 7.3049952293297625e-06,
+      "loss": 0.8787,
       "step": 21118
     },
     {
-      "epoch": 0.5800170278212629,
+      "epoch": 0.5992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 7.910037400595334e-06,
-      "loss": 0.8154,
+      "learning_rate": 7.30411016829831e-06,
+      "loss": 0.7379,
       "step": 21119
     },
     {
-      "epoch": 0.580044492049106,
+      "epoch": 0.5993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 7.909167536532463e-06,
-      "loss": 0.8854,
+      "learning_rate": 7.30322513003916e-06,
+      "loss": 0.8953,
       "step": 21120
     },
     {
-      "epoch": 0.5800719562769493,
+      "epoch": 0.5993473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 7.908297689013073e-06,
-      "loss": 0.9073,
+      "learning_rate": 7.302340114559788e-06,
+      "loss": 0.8508,
       "step": 21121
     },
     {
-      "epoch": 0.5800994205047925,
+      "epoch": 0.5993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 7.907427858044042e-06,
-      "loss": 0.8994,
+      "learning_rate": 7.301455121867672e-06,
+      "loss": 0.7212,
       "step": 21122
     },
     {
-      "epoch": 0.5801268847326357,
+      "epoch": 0.5994040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 7.90655804363225e-06,
-      "loss": 0.8555,
+      "learning_rate": 7.300570151970283e-06,
+      "loss": 0.9375,
       "step": 21123
     },
     {
-      "epoch": 0.580154348960479,
+      "epoch": 0.5994324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 7.90568824578458e-06,
-      "loss": 0.8326,
+      "learning_rate": 7.299685204875102e-06,
+      "loss": 0.7833,
       "step": 21124
     },
     {
-      "epoch": 0.5801818131883222,
+      "epoch": 0.5994608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 7.904818464507918e-06,
-      "loss": 0.8125,
+      "learning_rate": 7.2988002805896015e-06,
+      "loss": 0.8205,
       "step": 21125
     },
     {
-      "epoch": 0.5802092774161655,
+      "epoch": 0.599489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 7.903948699809143e-06,
-      "loss": 0.9173,
+      "learning_rate": 7.2979153791212545e-06,
+      "loss": 0.8802,
       "step": 21126
     },
     {
-      "epoch": 0.5802367416440086,
+      "epoch": 0.5995175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 7.903078951695137e-06,
-      "loss": 0.8678,
+      "learning_rate": 7.297030500477538e-06,
+      "loss": 0.8168,
       "step": 21127
     },
     {
-      "epoch": 0.5802642058718519,
+      "epoch": 0.5995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 7.902209220172787e-06,
-      "loss": 0.9103,
+      "learning_rate": 7.296145644665928e-06,
+      "loss": 0.8755,
       "step": 21128
     },
     {
-      "epoch": 0.5802916700996952,
+      "epoch": 0.5995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 7.901339505248966e-06,
-      "loss": 0.8063,
+      "learning_rate": 7.295260811693896e-06,
+      "loss": 0.7581,
       "step": 21129
     },
     {
-      "epoch": 0.5803191343275383,
+      "epoch": 0.5996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 7.900469806930562e-06,
-      "loss": 0.8204,
+      "learning_rate": 7.2943760015689195e-06,
+      "loss": 0.9557,
       "step": 21130
     },
     {
-      "epoch": 0.5803465985553816,
+      "epoch": 0.5996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 7.899600125224453e-06,
-      "loss": 0.8395,
+      "learning_rate": 7.293491214298469e-06,
+      "loss": 0.8449,
       "step": 21131
     },
     {
-      "epoch": 0.5803740627832249,
+      "epoch": 0.5996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 7.898730460137521e-06,
-      "loss": 0.9411,
+      "learning_rate": 7.2926064498900184e-06,
+      "loss": 0.8464,
       "step": 21132
     },
     {
-      "epoch": 0.5804015270110681,
+      "epoch": 0.5996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 7.897860811676649e-06,
-      "loss": 0.8168,
+      "learning_rate": 7.291721708351045e-06,
+      "loss": 0.8626,
       "step": 21133
     },
     {
-      "epoch": 0.5804289912389113,
+      "epoch": 0.599716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 7.896991179848718e-06,
-      "loss": 0.8776,
+      "learning_rate": 7.2908369896890155e-06,
+      "loss": 0.7998,
       "step": 21134
     },
     {
-      "epoch": 0.5804564554667545,
+      "epoch": 0.599744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 7.896121564660606e-06,
-      "loss": 0.7916,
+      "learning_rate": 7.289952293911412e-06,
+      "loss": 0.8968,
       "step": 21135
     },
     {
-      "epoch": 0.5804839196945978,
+      "epoch": 0.5997729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 7.895251966119191e-06,
-      "loss": 0.8518,
+      "learning_rate": 7.289067621025704e-06,
+      "loss": 0.773,
       "step": 21136
     },
     {
-      "epoch": 0.5805113839224411,
+      "epoch": 0.5998013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 7.894382384231363e-06,
-      "loss": 0.824,
+      "learning_rate": 7.288182971039362e-06,
+      "loss": 0.8781,
       "step": 21137
     },
     {
-      "epoch": 0.5805388481502842,
+      "epoch": 0.5998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 7.893512819003995e-06,
-      "loss": 0.785,
+      "learning_rate": 7.287298343959861e-06,
+      "loss": 0.8994,
       "step": 21138
     },
     {
-      "epoch": 0.5805663123781275,
+      "epoch": 0.5998581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 7.892643270443968e-06,
-      "loss": 0.8728,
+      "learning_rate": 7.286413739794673e-06,
+      "loss": 0.9065,
       "step": 21139
     },
     {
-      "epoch": 0.5805937766059707,
+      "epoch": 0.5998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 7.891773738558168e-06,
-      "loss": 0.8966,
+      "learning_rate": 7.285529158551268e-06,
+      "loss": 0.8924,
       "step": 21140
     },
     {
-      "epoch": 0.580621240833814,
+      "epoch": 0.5999148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 7.890904223353471e-06,
-      "loss": 0.8889,
+      "learning_rate": 7.284644600237123e-06,
+      "loss": 0.8922,
       "step": 21141
     },
     {
-      "epoch": 0.5806487050616572,
+      "epoch": 0.5999432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 7.890034724836753e-06,
-      "loss": 0.7498,
+      "learning_rate": 7.2837600648597075e-06,
+      "loss": 0.8911,
       "step": 21142
     },
     {
-      "epoch": 0.5806761692895004,
+      "epoch": 0.5999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 7.889165243014898e-06,
-      "loss": 0.8944,
+      "learning_rate": 7.282875552426489e-06,
+      "loss": 0.7999,
       "step": 21143
     },
     {
-      "epoch": 0.5807036335173437,
+      "epoch": 0.6,
       "grad_norm": 0.0,
-      "learning_rate": 7.888295777894782e-06,
-      "loss": 0.8671,
+      "learning_rate": 7.2819910629449464e-06,
+      "loss": 0.7873,
       "step": 21144
     },
     {
-      "epoch": 0.5807310977451869,
+      "epoch": 0.6000283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 7.887426329483293e-06,
-      "loss": 0.897,
+      "learning_rate": 7.281106596422547e-06,
+      "loss": 0.9166,
       "step": 21145
     },
     {
-      "epoch": 0.5807585619730301,
+      "epoch": 0.6000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 7.886556897787308e-06,
-      "loss": 0.8547,
+      "learning_rate": 7.2802221528667604e-06,
+      "loss": 0.8091,
       "step": 21146
     },
     {
-      "epoch": 0.5807860262008734,
+      "epoch": 0.6000851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 7.8856874828137e-06,
-      "loss": 0.8119,
+      "learning_rate": 7.279337732285062e-06,
+      "loss": 0.9368,
       "step": 21147
     },
     {
-      "epoch": 0.5808134904287166,
+      "epoch": 0.6001135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 7.88481808456935e-06,
-      "loss": 0.9725,
+      "learning_rate": 7.27845333468492e-06,
+      "loss": 0.9756,
       "step": 21148
     },
     {
-      "epoch": 0.5808409546565598,
+      "epoch": 0.6001418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 7.88394870306114e-06,
-      "loss": 0.8714,
+      "learning_rate": 7.277568960073801e-06,
+      "loss": 0.8963,
       "step": 21149
     },
     {
-      "epoch": 0.5808684188844031,
+      "epoch": 0.6001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 7.883079338295947e-06,
-      "loss": 0.8635,
+      "learning_rate": 7.2766846084591835e-06,
+      "loss": 0.8415,
       "step": 21150
     },
     {
-      "epoch": 0.5808958831122463,
+      "epoch": 0.6001986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 7.882209990280649e-06,
-      "loss": 0.876,
+      "learning_rate": 7.27580027984853e-06,
+      "loss": 0.895,
       "step": 21151
     },
     {
-      "epoch": 0.5809233473400895,
+      "epoch": 0.6002270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 7.881340659022131e-06,
-      "loss": 0.7836,
+      "learning_rate": 7.274915974249317e-06,
+      "loss": 0.8767,
       "step": 21152
     },
     {
-      "epoch": 0.5809508115679327,
+      "epoch": 0.600255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 7.880471344527264e-06,
-      "loss": 0.9196,
+      "learning_rate": 7.274031691669012e-06,
+      "loss": 0.919,
       "step": 21153
     },
     {
-      "epoch": 0.580978275795776,
+      "epoch": 0.6002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 7.879602046802926e-06,
-      "loss": 0.8895,
+      "learning_rate": 7.273147432115082e-06,
+      "loss": 0.8601,
       "step": 21154
     },
     {
-      "epoch": 0.5810057400236193,
+      "epoch": 0.6003121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 7.878732765855999e-06,
-      "loss": 0.8688,
+      "learning_rate": 7.272263195594999e-06,
+      "loss": 0.7171,
       "step": 21155
     },
     {
-      "epoch": 0.5810332042514624,
+      "epoch": 0.6003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 7.87786350169336e-06,
-      "loss": 0.8584,
+      "learning_rate": 7.271378982116232e-06,
+      "loss": 0.7924,
       "step": 21156
     },
     {
-      "epoch": 0.5810606684793057,
+      "epoch": 0.6003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 7.876994254321888e-06,
-      "loss": 0.95,
+      "learning_rate": 7.270494791686247e-06,
+      "loss": 0.843,
       "step": 21157
     },
     {
-      "epoch": 0.581088132707149,
+      "epoch": 0.6003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 7.876125023748463e-06,
-      "loss": 0.8809,
+      "learning_rate": 7.269610624312517e-06,
+      "loss": 0.7198,
       "step": 21158
     },
     {
-      "epoch": 0.5811155969349922,
+      "epoch": 0.6004256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 7.875255809979954e-06,
-      "loss": 0.7998,
+      "learning_rate": 7.2687264800025105e-06,
+      "loss": 0.8486,
       "step": 21159
     },
     {
-      "epoch": 0.5811430611628354,
+      "epoch": 0.6004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 7.874386613023247e-06,
-      "loss": 0.8397,
+      "learning_rate": 7.267842358763691e-06,
+      "loss": 0.8672,
       "step": 21160
     },
     {
-      "epoch": 0.5811705253906786,
+      "epoch": 0.6004824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 7.873517432885215e-06,
-      "loss": 0.8217,
+      "learning_rate": 7.266958260603532e-06,
+      "loss": 0.8941,
       "step": 21161
     },
     {
-      "epoch": 0.5811979896185219,
+      "epoch": 0.600510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 7.872648269572736e-06,
-      "loss": 0.8029,
+      "learning_rate": 7.266074185529499e-06,
+      "loss": 0.9014,
       "step": 21162
     },
     {
-      "epoch": 0.5812254538463651,
+      "epoch": 0.600539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 7.871779123092694e-06,
-      "loss": 0.9493,
+      "learning_rate": 7.265190133549059e-06,
+      "loss": 0.9025,
       "step": 21163
     },
     {
-      "epoch": 0.5812529180742083,
+      "epoch": 0.6005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 7.870909993451954e-06,
-      "loss": 0.8237,
+      "learning_rate": 7.264306104669678e-06,
+      "loss": 0.9616,
       "step": 21164
     },
     {
-      "epoch": 0.5812803823020516,
+      "epoch": 0.6005959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 7.870040880657401e-06,
-      "loss": 0.8764,
+      "learning_rate": 7.26342209889883e-06,
+      "loss": 0.8903,
       "step": 21165
     },
     {
-      "epoch": 0.5813078465298949,
+      "epoch": 0.6006242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 7.869171784715908e-06,
-      "loss": 0.8826,
+      "learning_rate": 7.26253811624398e-06,
+      "loss": 0.9706,
       "step": 21166
     },
     {
-      "epoch": 0.581335310757738,
+      "epoch": 0.6006526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 7.868302705634354e-06,
-      "loss": 0.9099,
+      "learning_rate": 7.261654156712591e-06,
+      "loss": 0.7505,
       "step": 21167
     },
     {
-      "epoch": 0.5813627749855813,
+      "epoch": 0.6006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 7.867433643419614e-06,
-      "loss": 0.9192,
+      "learning_rate": 7.260770220312134e-06,
+      "loss": 0.8137,
       "step": 21168
     },
     {
-      "epoch": 0.5813902392134245,
+      "epoch": 0.6007094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 7.866564598078568e-06,
-      "loss": 0.8242,
+      "learning_rate": 7.259886307050075e-06,
+      "loss": 0.7732,
       "step": 21169
     },
     {
-      "epoch": 0.5814177034412678,
+      "epoch": 0.6007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 7.865695569618088e-06,
-      "loss": 0.9388,
+      "learning_rate": 7.259002416933876e-06,
+      "loss": 0.8921,
       "step": 21170
     },
     {
-      "epoch": 0.581445167669111,
+      "epoch": 0.6007661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 7.86482655804505e-06,
-      "loss": 0.8191,
+      "learning_rate": 7.258118549971011e-06,
+      "loss": 0.8734,
       "step": 21171
     },
     {
-      "epoch": 0.5814726318969542,
+      "epoch": 0.6007945516458569,
       "grad_norm": 0.0,
-      "learning_rate": 7.863957563366331e-06,
-      "loss": 0.8416,
+      "learning_rate": 7.257234706168941e-06,
+      "loss": 0.7867,
       "step": 21172
     },
     {
-      "epoch": 0.5815000961247975,
+      "epoch": 0.6008229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 7.86308858558881e-06,
-      "loss": 0.8456,
+      "learning_rate": 7.256350885535129e-06,
+      "loss": 0.854,
       "step": 21173
     },
     {
-      "epoch": 0.5815275603526406,
+      "epoch": 0.6008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 7.862219624719356e-06,
-      "loss": 0.8598,
+      "learning_rate": 7.255467088077048e-06,
+      "loss": 0.8592,
       "step": 21174
     },
     {
-      "epoch": 0.5815550245804839,
+      "epoch": 0.6008796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 7.861350680764854e-06,
-      "loss": 0.8541,
+      "learning_rate": 7.2545833138021596e-06,
+      "loss": 0.838,
       "step": 21175
     },
     {
-      "epoch": 0.5815824888083272,
+      "epoch": 0.6009080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 7.860481753732168e-06,
-      "loss": 0.9017,
+      "learning_rate": 7.2536995627179295e-06,
+      "loss": 0.937,
       "step": 21176
     },
     {
-      "epoch": 0.5816099530361704,
+      "epoch": 0.6009364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 7.859612843628184e-06,
-      "loss": 0.8906,
+      "learning_rate": 7.252815834831821e-06,
+      "loss": 0.9591,
       "step": 21177
     },
     {
-      "epoch": 0.5816374172640136,
+      "epoch": 0.6009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 7.858743950459769e-06,
-      "loss": 0.7912,
+      "learning_rate": 7.251932130151302e-06,
+      "loss": 0.8478,
       "step": 21178
     },
     {
-      "epoch": 0.5816648814918569,
+      "epoch": 0.6009931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 7.857875074233801e-06,
-      "loss": 0.8907,
+      "learning_rate": 7.251048448683836e-06,
+      "loss": 0.7682,
       "step": 21179
     },
     {
-      "epoch": 0.5816923457197001,
+      "epoch": 0.6010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 7.857006214957158e-06,
-      "loss": 0.8386,
+      "learning_rate": 7.2501647904368835e-06,
+      "loss": 0.8465,
       "step": 21180
     },
     {
-      "epoch": 0.5817198099475434,
+      "epoch": 0.601049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 7.85613737263671e-06,
-      "loss": 0.8173,
+      "learning_rate": 7.249281155417914e-06,
+      "loss": 0.846,
       "step": 21181
     },
     {
-      "epoch": 0.5817472741753865,
+      "epoch": 0.6010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 7.855268547279334e-06,
-      "loss": 0.8755,
+      "learning_rate": 7.248397543634394e-06,
+      "loss": 0.7777,
       "step": 21182
     },
     {
-      "epoch": 0.5817747384032298,
+      "epoch": 0.6011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 7.854399738891903e-06,
-      "loss": 0.9026,
+      "learning_rate": 7.247513955093781e-06,
+      "loss": 0.8004,
       "step": 21183
     },
     {
-      "epoch": 0.5818022026310731,
+      "epoch": 0.6011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 7.853530947481293e-06,
-      "loss": 0.9321,
+      "learning_rate": 7.246630389803541e-06,
+      "loss": 0.9038,
       "step": 21184
     },
     {
-      "epoch": 0.5818296668589162,
+      "epoch": 0.6011634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 7.852662173054379e-06,
+      "learning_rate": 7.24574684777114e-06,
       "loss": 0.909,
       "step": 21185
     },
     {
-      "epoch": 0.5818571310867595,
+      "epoch": 0.6011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 7.851793415618031e-06,
-      "loss": 0.8106,
+      "learning_rate": 7.24486332900404e-06,
+      "loss": 0.8389,
       "step": 21186
     },
     {
-      "epoch": 0.5818845953146027,
+      "epoch": 0.6012202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 7.850924675179131e-06,
-      "loss": 0.8494,
+      "learning_rate": 7.243979833509699e-06,
+      "loss": 0.8184,
       "step": 21187
     },
     {
-      "epoch": 0.581912059542446,
+      "epoch": 0.6012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 7.850055951744542e-06,
-      "loss": 0.8524,
+      "learning_rate": 7.243096361295587e-06,
+      "loss": 0.8061,
       "step": 21188
     },
     {
-      "epoch": 0.5819395237702892,
+      "epoch": 0.6012769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 7.849187245321145e-06,
-      "loss": 0.7433,
+      "learning_rate": 7.242212912369164e-06,
+      "loss": 0.9154,
       "step": 21189
     },
     {
-      "epoch": 0.5819669879981324,
+      "epoch": 0.6013053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 7.84831855591581e-06,
-      "loss": 0.836,
+      "learning_rate": 7.241329486737891e-06,
+      "loss": 0.9451,
       "step": 21190
     },
     {
-      "epoch": 0.5819944522259757,
+      "epoch": 0.6013337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 7.847449883535413e-06,
-      "loss": 0.8514,
+      "learning_rate": 7.240446084409234e-06,
+      "loss": 0.769,
       "step": 21191
     },
     {
-      "epoch": 0.582021916453819,
+      "epoch": 0.6013620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 7.846581228186827e-06,
-      "loss": 0.813,
+      "learning_rate": 7.239562705390652e-06,
+      "loss": 0.8432,
       "step": 21192
     },
     {
-      "epoch": 0.5820493806816621,
+      "epoch": 0.6013904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 7.845712589876926e-06,
-      "loss": 0.9254,
+      "learning_rate": 7.238679349689608e-06,
+      "loss": 0.9834,
       "step": 21193
     },
     {
-      "epoch": 0.5820768449095054,
+      "epoch": 0.6014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 7.844843968612581e-06,
-      "loss": 0.8528,
+      "learning_rate": 7.237796017313563e-06,
+      "loss": 0.7874,
       "step": 21194
     },
     {
-      "epoch": 0.5821043091373486,
+      "epoch": 0.6014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 7.843975364400662e-06,
-      "loss": 0.9192,
+      "learning_rate": 7.2369127082699766e-06,
+      "loss": 0.9206,
       "step": 21195
     },
     {
-      "epoch": 0.5821317733651918,
+      "epoch": 0.6014755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 7.843106777248047e-06,
-      "loss": 0.9572,
+      "learning_rate": 7.236029422566318e-06,
+      "loss": 0.7209,
       "step": 21196
     },
     {
-      "epoch": 0.5821592375930351,
+      "epoch": 0.6015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 7.842238207161607e-06,
-      "loss": 0.8546,
+      "learning_rate": 7.235146160210041e-06,
+      "loss": 0.7507,
       "step": 21197
     },
     {
-      "epoch": 0.5821867018208783,
+      "epoch": 0.6015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 7.841369654148213e-06,
-      "loss": 0.858,
+      "learning_rate": 7.234262921208607e-06,
+      "loss": 1.0271,
       "step": 21198
     },
     {
-      "epoch": 0.5822141660487216,
+      "epoch": 0.601560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 7.840501118214742e-06,
-      "loss": 0.8514,
+      "learning_rate": 7.233379705569481e-06,
+      "loss": 0.9118,
       "step": 21199
     },
     {
-      "epoch": 0.5822416302765647,
+      "epoch": 0.601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 7.83963259936806e-06,
-      "loss": 0.8549,
+      "learning_rate": 7.23249651330012e-06,
+      "loss": 0.9785,
       "step": 21200
     },
     {
-      "epoch": 0.582269094504408,
+      "epoch": 0.6016174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 7.838764097615043e-06,
-      "loss": 0.778,
+      "learning_rate": 7.231613344407984e-06,
+      "loss": 0.7809,
       "step": 21201
     },
     {
-      "epoch": 0.5822965587322513,
+      "epoch": 0.6016458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 7.837895612962558e-06,
-      "loss": 0.8603,
+      "learning_rate": 7.230730198900537e-06,
+      "loss": 0.9162,
       "step": 21202
     },
     {
-      "epoch": 0.5823240229600944,
+      "epoch": 0.6016742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 7.837027145417483e-06,
-      "loss": 0.9149,
+      "learning_rate": 7.229847076785237e-06,
+      "loss": 0.8361,
       "step": 21203
     },
     {
-      "epoch": 0.5823514871879377,
+      "epoch": 0.6017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 7.836158694986687e-06,
-      "loss": 0.8223,
+      "learning_rate": 7.22896397806954e-06,
+      "loss": 0.8326,
       "step": 21204
     },
     {
-      "epoch": 0.582378951415781,
+      "epoch": 0.6017309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 7.835290261677043e-06,
-      "loss": 0.9022,
+      "learning_rate": 7.228080902760911e-06,
+      "loss": 0.923,
       "step": 21205
     },
     {
-      "epoch": 0.5824064156436242,
+      "epoch": 0.6017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 7.834421845495421e-06,
-      "loss": 0.891,
+      "learning_rate": 7.227197850866807e-06,
+      "loss": 0.8599,
       "step": 21206
     },
     {
-      "epoch": 0.5824338798714674,
+      "epoch": 0.6017877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 7.83355344644869e-06,
-      "loss": 0.8255,
+      "learning_rate": 7.226314822394686e-06,
+      "loss": 0.8636,
       "step": 21207
     },
     {
-      "epoch": 0.5824613440993106,
+      "epoch": 0.6018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 7.832685064543722e-06,
-      "loss": 0.8637,
+      "learning_rate": 7.225431817352011e-06,
+      "loss": 0.8563,
       "step": 21208
     },
     {
-      "epoch": 0.5824888083271539,
+      "epoch": 0.601844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 7.831816699787393e-06,
-      "loss": 0.8935,
+      "learning_rate": 7.224548835746238e-06,
+      "loss": 0.8003,
       "step": 21209
     },
     {
-      "epoch": 0.5825162725549972,
+      "epoch": 0.6018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 7.830948352186567e-06,
-      "loss": 0.76,
+      "learning_rate": 7.22366587758482e-06,
+      "loss": 0.8195,
       "step": 21210
     },
     {
-      "epoch": 0.5825437367828403,
+      "epoch": 0.6019012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 7.830080021748123e-06,
-      "loss": 0.8542,
+      "learning_rate": 7.222782942875225e-06,
+      "loss": 0.7956,
       "step": 21211
     },
     {
-      "epoch": 0.5825712010106836,
+      "epoch": 0.6019296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 7.829211708478923e-06,
-      "loss": 0.8397,
+      "learning_rate": 7.221900031624909e-06,
+      "loss": 0.8064,
       "step": 21212
     },
     {
-      "epoch": 0.5825986652385268,
+      "epoch": 0.6019580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 7.828343412385841e-06,
-      "loss": 0.8635,
+      "learning_rate": 7.221017143841328e-06,
+      "loss": 0.7986,
       "step": 21213
     },
     {
-      "epoch": 0.58262612946637,
+      "epoch": 0.6019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 7.827475133475744e-06,
-      "loss": 0.8222,
+      "learning_rate": 7.2201342795319384e-06,
+      "loss": 0.8767,
       "step": 21214
     },
     {
-      "epoch": 0.5826535936942133,
+      "epoch": 0.6020147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 7.826606871755508e-06,
-      "loss": 0.8061,
+      "learning_rate": 7.2192514387042e-06,
+      "loss": 0.7388,
       "step": 21215
     },
     {
-      "epoch": 0.5826810579220565,
+      "epoch": 0.6020431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 7.825738627232001e-06,
-      "loss": 0.8326,
+      "learning_rate": 7.218368621365572e-06,
+      "loss": 0.9205,
       "step": 21216
     },
     {
-      "epoch": 0.5827085221498998,
+      "epoch": 0.6020715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 7.824870399912093e-06,
-      "loss": 0.915,
+      "learning_rate": 7.217485827523507e-06,
+      "loss": 0.8319,
       "step": 21217
     },
     {
-      "epoch": 0.582735986377743,
+      "epoch": 0.602099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 7.824002189802654e-06,
-      "loss": 0.8944,
+      "learning_rate": 7.216603057185465e-06,
+      "loss": 0.903,
       "step": 21218
     },
     {
-      "epoch": 0.5827634506055862,
+      "epoch": 0.6021282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 7.823133996910548e-06,
-      "loss": 0.9168,
+      "learning_rate": 7.215720310358903e-06,
+      "loss": 0.989,
       "step": 21219
     },
     {
-      "epoch": 0.5827909148334295,
+      "epoch": 0.6021566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 7.82226582124265e-06,
-      "loss": 0.8588,
+      "learning_rate": 7.214837587051275e-06,
+      "loss": 0.8794,
       "step": 21220
     },
     {
-      "epoch": 0.5828183790612727,
+      "epoch": 0.6021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 7.82139766280583e-06,
-      "loss": 0.9384,
+      "learning_rate": 7.213954887270039e-06,
+      "loss": 0.8331,
       "step": 21221
     },
     {
-      "epoch": 0.5828458432891159,
+      "epoch": 0.6022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 7.820529521606955e-06,
-      "loss": 0.8997,
+      "learning_rate": 7.213072211022653e-06,
+      "loss": 0.7623,
       "step": 21222
     },
     {
-      "epoch": 0.5828733075169592,
+      "epoch": 0.6022417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 7.819661397652898e-06,
-      "loss": 0.8683,
+      "learning_rate": 7.2121895583165716e-06,
+      "loss": 0.8834,
       "step": 21223
     },
     {
-      "epoch": 0.5829007717448024,
+      "epoch": 0.6022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 7.81879329095052e-06,
-      "loss": 0.8171,
+      "learning_rate": 7.211306929159247e-06,
+      "loss": 0.8581,
       "step": 21224
     },
     {
-      "epoch": 0.5829282359726456,
+      "epoch": 0.6022985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 7.817925201506697e-06,
-      "loss": 0.868,
+      "learning_rate": 7.21042432355814e-06,
+      "loss": 0.8463,
       "step": 21225
     },
     {
-      "epoch": 0.5829557002004888,
+      "epoch": 0.6023269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 7.817057129328293e-06,
-      "loss": 0.8359,
+      "learning_rate": 7.2095417415207015e-06,
+      "loss": 0.8024,
       "step": 21226
     },
     {
-      "epoch": 0.5829831644283321,
+      "epoch": 0.6023552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 7.816189074422178e-06,
-      "loss": 0.9217,
+      "learning_rate": 7.208659183054393e-06,
+      "loss": 0.9088,
       "step": 21227
     },
     {
-      "epoch": 0.5830106286561754,
+      "epoch": 0.602383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 7.815321036795227e-06,
-      "loss": 0.9662,
+      "learning_rate": 7.207776648166664e-06,
+      "loss": 0.7353,
       "step": 21228
     },
     {
-      "epoch": 0.5830380928840185,
+      "epoch": 0.6024120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 7.814453016454295e-06,
-      "loss": 0.9215,
+      "learning_rate": 7.2068941368649726e-06,
+      "loss": 0.8925,
       "step": 21229
     },
     {
-      "epoch": 0.5830655571118618,
+      "epoch": 0.6024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 7.813585013406261e-06,
-      "loss": 0.8606,
+      "learning_rate": 7.206011649156773e-06,
+      "loss": 0.8733,
       "step": 21230
     },
     {
-      "epoch": 0.5830930213397051,
+      "epoch": 0.6024687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 7.812717027657988e-06,
-      "loss": 0.9276,
+      "learning_rate": 7.205129185049515e-06,
+      "loss": 0.8575,
       "step": 21231
     },
     {
-      "epoch": 0.5831204855675483,
+      "epoch": 0.6024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 7.811849059216344e-06,
-      "loss": 0.8651,
+      "learning_rate": 7.20424674455066e-06,
+      "loss": 0.8327,
       "step": 21232
     },
     {
-      "epoch": 0.5831479497953915,
+      "epoch": 0.6025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 7.8109811080882e-06,
-      "loss": 0.8998,
+      "learning_rate": 7.203364327667657e-06,
+      "loss": 0.749,
       "step": 21233
     },
     {
-      "epoch": 0.5831754140232347,
+      "epoch": 0.6025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 7.810113174280424e-06,
-      "loss": 0.8206,
+      "learning_rate": 7.20248193440796e-06,
+      "loss": 0.778,
       "step": 21234
     },
     {
-      "epoch": 0.583202878251078,
+      "epoch": 0.6025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 7.809245257799878e-06,
-      "loss": 0.9512,
+      "learning_rate": 7.201599564779026e-06,
+      "loss": 0.7676,
       "step": 21235
     },
     {
-      "epoch": 0.5832303424789212,
+      "epoch": 0.60261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 7.80837735865343e-06,
-      "loss": 0.8695,
+      "learning_rate": 7.200717218788306e-06,
+      "loss": 0.9243,
       "step": 21236
     },
     {
-      "epoch": 0.5832578067067644,
+      "epoch": 0.602639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 7.807509476847952e-06,
-      "loss": 0.9099,
+      "learning_rate": 7.199834896443252e-06,
+      "loss": 0.8583,
       "step": 21237
     },
     {
-      "epoch": 0.5832852709346077,
+      "epoch": 0.6026674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 7.806641612390307e-06,
-      "loss": 0.7909,
+      "learning_rate": 7.19895259775132e-06,
+      "loss": 0.7879,
       "step": 21238
     },
     {
-      "epoch": 0.5833127351624509,
+      "epoch": 0.6026958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 7.805773765287363e-06,
-      "loss": 0.9024,
+      "learning_rate": 7.198070322719961e-06,
+      "loss": 0.8063,
       "step": 21239
     },
     {
-      "epoch": 0.5833401993902941,
+      "epoch": 0.6027241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 7.80490593554599e-06,
-      "loss": 0.8125,
+      "learning_rate": 7.197188071356627e-06,
+      "loss": 0.9085,
       "step": 21240
     },
     {
-      "epoch": 0.5833676636181374,
+      "epoch": 0.6027525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 7.804038123173048e-06,
-      "loss": 0.8571,
+      "learning_rate": 7.196305843668771e-06,
+      "loss": 0.8369,
       "step": 21241
     },
     {
-      "epoch": 0.5833951278459806,
+      "epoch": 0.6027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 7.803170328175409e-06,
-      "loss": 0.7478,
+      "learning_rate": 7.195423639663845e-06,
+      "loss": 0.8403,
       "step": 21242
     },
     {
-      "epoch": 0.5834225920738239,
+      "epoch": 0.6028093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 7.802302550559935e-06,
-      "loss": 0.9029,
+      "learning_rate": 7.194541459349304e-06,
+      "loss": 0.8521,
       "step": 21243
     },
     {
-      "epoch": 0.5834500563016671,
+      "epoch": 0.6028376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 7.801434790333496e-06,
-      "loss": 0.8174,
+      "learning_rate": 7.1936593027325965e-06,
+      "loss": 0.8782,
       "step": 21244
     },
     {
-      "epoch": 0.5834775205295103,
+      "epoch": 0.6028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 7.800567047502955e-06,
-      "loss": 0.8091,
+      "learning_rate": 7.192777169821174e-06,
+      "loss": 0.67,
       "step": 21245
     },
     {
-      "epoch": 0.5835049847573536,
+      "epoch": 0.602894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.799699322075184e-06,
-      "loss": 0.8719,
+      "learning_rate": 7.191895060622491e-06,
+      "loss": 0.8178,
       "step": 21246
     },
     {
-      "epoch": 0.5835324489851967,
+      "epoch": 0.6029228149829738,
       "grad_norm": 0.0,
-      "learning_rate": 7.798831614057042e-06,
-      "loss": 0.8292,
+      "learning_rate": 7.191012975143996e-06,
+      "loss": 0.9714,
       "step": 21247
     },
     {
-      "epoch": 0.58355991321304,
+      "epoch": 0.6029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 7.797963923455395e-06,
-      "loss": 0.8571,
+      "learning_rate": 7.19013091339314e-06,
+      "loss": 0.9531,
       "step": 21248
     },
     {
-      "epoch": 0.5835873774408833,
+      "epoch": 0.6029795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 7.797096250277112e-06,
-      "loss": 0.9225,
+      "learning_rate": 7.1892488753773744e-06,
+      "loss": 0.8558,
       "step": 21249
     },
     {
-      "epoch": 0.5836148416687265,
+      "epoch": 0.6030079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 7.796228594529056e-06,
-      "loss": 0.7944,
+      "learning_rate": 7.1883668611041525e-06,
+      "loss": 0.824,
       "step": 21250
     },
     {
-      "epoch": 0.5836423058965697,
+      "epoch": 0.6030363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 7.795360956218092e-06,
-      "loss": 0.838,
+      "learning_rate": 7.187484870580919e-06,
+      "loss": 0.8105,
       "step": 21251
     },
     {
-      "epoch": 0.5836697701244129,
+      "epoch": 0.6030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 7.794493335351091e-06,
-      "loss": 0.9704,
+      "learning_rate": 7.186602903815128e-06,
+      "loss": 0.8213,
       "step": 21252
     },
     {
-      "epoch": 0.5836972343522562,
+      "epoch": 0.6030930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 7.793625731934908e-06,
-      "loss": 0.8144,
+      "learning_rate": 7.185720960814229e-06,
+      "loss": 0.9028,
       "step": 21253
     },
     {
-      "epoch": 0.5837246985800995,
+      "epoch": 0.6031214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 7.792758145976416e-06,
-      "loss": 0.9196,
+      "learning_rate": 7.184839041585671e-06,
+      "loss": 0.8503,
       "step": 21254
     },
     {
-      "epoch": 0.5837521628079426,
+      "epoch": 0.603149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 7.791890577482474e-06,
-      "loss": 0.8589,
+      "learning_rate": 7.1839571461369054e-06,
+      "loss": 0.8607,
       "step": 21255
     },
     {
-      "epoch": 0.5837796270357859,
+      "epoch": 0.6031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 7.791023026459949e-06,
-      "loss": 0.8536,
+      "learning_rate": 7.183075274475376e-06,
+      "loss": 0.8687,
       "step": 21256
     },
     {
-      "epoch": 0.5838070912636292,
+      "epoch": 0.6032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 7.790155492915708e-06,
-      "loss": 0.8132,
+      "learning_rate": 7.182193426608542e-06,
+      "loss": 0.9248,
       "step": 21257
     },
     {
-      "epoch": 0.5838345554914723,
+      "epoch": 0.6032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 7.789287976856614e-06,
-      "loss": 0.9304,
+      "learning_rate": 7.181311602543845e-06,
+      "loss": 0.9283,
       "step": 21258
     },
     {
-      "epoch": 0.5838620197193156,
+      "epoch": 0.6032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 7.788420478289525e-06,
-      "loss": 0.8486,
+      "learning_rate": 7.180429802288733e-06,
+      "loss": 0.9468,
       "step": 21259
     },
     {
-      "epoch": 0.5838894839471588,
+      "epoch": 0.6032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 7.787552997221311e-06,
-      "loss": 0.9148,
+      "learning_rate": 7.17954802585066e-06,
+      "loss": 0.9947,
       "step": 21260
     },
     {
-      "epoch": 0.5839169481750021,
+      "epoch": 0.6033200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 7.786685533658838e-06,
-      "loss": 0.8052,
+      "learning_rate": 7.178666273237073e-06,
+      "loss": 0.7768,
       "step": 21261
     },
     {
-      "epoch": 0.5839444124028453,
+      "epoch": 0.6033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 7.785818087608963e-06,
-      "loss": 0.8756,
+      "learning_rate": 7.177784544455415e-06,
+      "loss": 0.9004,
       "step": 21262
     },
     {
-      "epoch": 0.5839718766306885,
+      "epoch": 0.6033768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 7.784950659078553e-06,
-      "loss": 0.8594,
+      "learning_rate": 7.1769028395131415e-06,
+      "loss": 0.9076,
       "step": 21263
     },
     {
-      "epoch": 0.5839993408585318,
+      "epoch": 0.603405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 7.784083248074475e-06,
-      "loss": 0.8479,
+      "learning_rate": 7.176021158417695e-06,
+      "loss": 0.9137,
       "step": 21264
     },
     {
-      "epoch": 0.584026805086375,
+      "epoch": 0.603433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 7.783215854603585e-06,
-      "loss": 0.9134,
+      "learning_rate": 7.175139501176524e-06,
+      "loss": 0.8085,
       "step": 21265
     },
     {
-      "epoch": 0.5840542693142182,
+      "epoch": 0.6034619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 7.782348478672753e-06,
-      "loss": 0.8531,
+      "learning_rate": 7.174257867797079e-06,
+      "loss": 0.8396,
       "step": 21266
     },
     {
-      "epoch": 0.5840817335420615,
+      "epoch": 0.6034903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 7.781481120288836e-06,
-      "loss": 0.752,
+      "learning_rate": 7.173376258286803e-06,
+      "loss": 0.8898,
       "step": 21267
     },
     {
-      "epoch": 0.5841091977699047,
+      "epoch": 0.6035187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 7.780613779458699e-06,
-      "loss": 0.9847,
+      "learning_rate": 7.172494672653143e-06,
+      "loss": 0.8261,
       "step": 21268
     },
     {
-      "epoch": 0.5841366619977479,
+      "epoch": 0.6035471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 7.77974645618921e-06,
-      "loss": 0.8805,
+      "learning_rate": 7.17161311090355e-06,
+      "loss": 0.9125,
       "step": 21269
     },
     {
-      "epoch": 0.5841641262255912,
+      "epoch": 0.6035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 7.778879150487226e-06,
-      "loss": 0.8322,
+      "learning_rate": 7.1707315730454695e-06,
+      "loss": 0.7594,
       "step": 21270
     },
     {
-      "epoch": 0.5841915904534344,
+      "epoch": 0.6036038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 7.778011862359608e-06,
-      "loss": 0.9277,
+      "learning_rate": 7.169850059086343e-06,
+      "loss": 0.7921,
       "step": 21271
     },
     {
-      "epoch": 0.5842190546812777,
+      "epoch": 0.6036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 7.777144591813222e-06,
-      "loss": 0.9399,
+      "learning_rate": 7.168968569033619e-06,
+      "loss": 0.7434,
       "step": 21272
     },
     {
-      "epoch": 0.5842465189091208,
+      "epoch": 0.6036606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 7.776277338854928e-06,
-      "loss": 0.9562,
+      "learning_rate": 7.16808710289475e-06,
+      "loss": 0.9014,
       "step": 21273
     },
     {
-      "epoch": 0.5842739831369641,
+      "epoch": 0.6036889897843359,
       "grad_norm": 0.0,
-      "learning_rate": 7.77541010349159e-06,
-      "loss": 0.8395,
+      "learning_rate": 7.167205660677174e-06,
+      "loss": 0.9541,
       "step": 21274
     },
     {
-      "epoch": 0.5843014473648074,
+      "epoch": 0.6037173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 7.774542885730067e-06,
-      "loss": 0.8882,
+      "learning_rate": 7.166324242388338e-06,
+      "loss": 0.8297,
       "step": 21275
     },
     {
-      "epoch": 0.5843289115926505,
+      "epoch": 0.6037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 7.773675685577227e-06,
-      "loss": 0.8381,
+      "learning_rate": 7.165442848035691e-06,
+      "loss": 0.9094,
       "step": 21276
     },
     {
-      "epoch": 0.5843563758204938,
+      "epoch": 0.6037741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 7.772808503039924e-06,
-      "loss": 0.7442,
+      "learning_rate": 7.1645614776266755e-06,
+      "loss": 0.7791,
       "step": 21277
     },
     {
-      "epoch": 0.584383840048337,
+      "epoch": 0.6038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 7.771941338125027e-06,
-      "loss": 0.8098,
+      "learning_rate": 7.1636801311687355e-06,
+      "loss": 0.9365,
       "step": 21278
     },
     {
-      "epoch": 0.5844113042761803,
+      "epoch": 0.6038308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 7.771074190839387e-06,
-      "loss": 0.8644,
+      "learning_rate": 7.162798808669318e-06,
+      "loss": 0.9394,
       "step": 21279
     },
     {
-      "epoch": 0.5844387685040235,
+      "epoch": 0.6038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 7.770207061189873e-06,
-      "loss": 0.9178,
+      "learning_rate": 7.161917510135866e-06,
+      "loss": 0.906,
       "step": 21280
     },
     {
-      "epoch": 0.5844662327318667,
+      "epoch": 0.6038876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 7.769339949183347e-06,
-      "loss": 0.8571,
+      "learning_rate": 7.161036235575822e-06,
+      "loss": 0.8247,
       "step": 21281
     },
     {
-      "epoch": 0.58449369695971,
+      "epoch": 0.6039160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 7.76847285482667e-06,
-      "loss": 0.7951,
+      "learning_rate": 7.160154984996637e-06,
+      "loss": 0.8709,
       "step": 21282
     },
     {
-      "epoch": 0.5845211611875533,
+      "epoch": 0.60394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.767605778126695e-06,
-      "loss": 0.922,
+      "learning_rate": 7.159273758405749e-06,
+      "loss": 0.9442,
       "step": 21283
     },
     {
-      "epoch": 0.5845486254153964,
+      "epoch": 0.6039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 7.76673871909029e-06,
-      "loss": 0.8975,
+      "learning_rate": 7.158392555810603e-06,
+      "loss": 0.9098,
       "step": 21284
     },
     {
-      "epoch": 0.5845760896432397,
+      "epoch": 0.6040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 7.765871677724312e-06,
-      "loss": 0.9065,
+      "learning_rate": 7.157511377218641e-06,
+      "loss": 0.8535,
       "step": 21285
     },
     {
-      "epoch": 0.5846035538710829,
+      "epoch": 0.6040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 7.765004654035623e-06,
-      "loss": 0.9218,
+      "learning_rate": 7.156630222637311e-06,
+      "loss": 0.8593,
       "step": 21286
     },
     {
-      "epoch": 0.5846310180989261,
+      "epoch": 0.6040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 7.764137648031086e-06,
-      "loss": 0.867,
+      "learning_rate": 7.1557490920740465e-06,
+      "loss": 0.9282,
       "step": 21287
     },
     {
-      "epoch": 0.5846584823267694,
+      "epoch": 0.6040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 7.763270659717554e-06,
-      "loss": 0.7829,
+      "learning_rate": 7.154867985536305e-06,
+      "loss": 0.8088,
       "step": 21288
     },
     {
-      "epoch": 0.5846859465546126,
+      "epoch": 0.6041146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 7.762403689101894e-06,
-      "loss": 0.9084,
+      "learning_rate": 7.153986903031518e-06,
+      "loss": 0.9367,
       "step": 21289
     },
     {
-      "epoch": 0.5847134107824559,
+      "epoch": 0.6041430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 7.76153673619096e-06,
-      "loss": 0.8012,
+      "learning_rate": 7.153105844567133e-06,
+      "loss": 0.921,
       "step": 21290
     },
     {
-      "epoch": 0.584740875010299,
+      "epoch": 0.6041713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 7.760669800991614e-06,
-      "loss": 0.9004,
+      "learning_rate": 7.152224810150592e-06,
+      "loss": 0.885,
       "step": 21291
     },
     {
-      "epoch": 0.5847683392381423,
+      "epoch": 0.6041997729852441,
       "grad_norm": 0.0,
-      "learning_rate": 7.759802883510716e-06,
-      "loss": 0.8596,
+      "learning_rate": 7.151343799789332e-06,
+      "loss": 0.9265,
       "step": 21292
     },
     {
-      "epoch": 0.5847958034659856,
+      "epoch": 0.6042281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 7.758935983755131e-06,
-      "loss": 0.7279,
+      "learning_rate": 7.150462813490804e-06,
+      "loss": 0.903,
       "step": 21293
     },
     {
-      "epoch": 0.5848232676938288,
+      "epoch": 0.6042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 7.75806910173171e-06,
-      "loss": 0.9088,
+      "learning_rate": 7.149581851262442e-06,
+      "loss": 0.8936,
       "step": 21294
     },
     {
-      "epoch": 0.584850731921672,
+      "epoch": 0.6042849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 7.757202237447311e-06,
-      "loss": 0.8834,
+      "learning_rate": 7.14870091311169e-06,
+      "loss": 0.8967,
       "step": 21295
     },
     {
-      "epoch": 0.5848781961495153,
+      "epoch": 0.6043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 7.756335390908798e-06,
-      "loss": 0.8842,
+      "learning_rate": 7.14781999904599e-06,
+      "loss": 0.9042,
       "step": 21296
     },
     {
-      "epoch": 0.5849056603773585,
+      "epoch": 0.6043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 7.75546856212303e-06,
-      "loss": 0.8814,
+      "learning_rate": 7.1469391090727834e-06,
+      "loss": 0.7987,
       "step": 21297
     },
     {
-      "epoch": 0.5849331246052017,
+      "epoch": 0.6043700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 7.754601751096864e-06,
-      "loss": 0.8518,
+      "learning_rate": 7.14605824319951e-06,
+      "loss": 0.7943,
       "step": 21298
     },
     {
-      "epoch": 0.5849605888330449,
+      "epoch": 0.6043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 7.75373495783716e-06,
-      "loss": 0.803,
+      "learning_rate": 7.145177401433611e-06,
+      "loss": 0.8652,
       "step": 21299
     },
     {
-      "epoch": 0.5849880530608882,
+      "epoch": 0.6044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 7.752868182350771e-06,
-      "loss": 0.8712,
+      "learning_rate": 7.1442965837825275e-06,
+      "loss": 0.8795,
       "step": 21300
     },
     {
-      "epoch": 0.5850155172887315,
+      "epoch": 0.604455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 7.752001424644565e-06,
-      "loss": 0.8152,
+      "learning_rate": 7.143415790253696e-06,
+      "loss": 0.6523,
       "step": 21301
     },
     {
-      "epoch": 0.5850429815165746,
+      "epoch": 0.604483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 7.751134684725389e-06,
-      "loss": 0.8196,
+      "learning_rate": 7.142535020854562e-06,
+      "loss": 0.8213,
       "step": 21302
     },
     {
-      "epoch": 0.5850704457444179,
+      "epoch": 0.6045119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 7.750267962600107e-06,
-      "loss": 0.8353,
+      "learning_rate": 7.141654275592561e-06,
+      "loss": 0.8232,
       "step": 21303
     },
     {
-      "epoch": 0.5850979099722611,
+      "epoch": 0.6045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 7.749401258275577e-06,
-      "loss": 0.9043,
+      "learning_rate": 7.140773554475138e-06,
+      "loss": 0.8616,
       "step": 21304
     },
     {
-      "epoch": 0.5851253742001044,
+      "epoch": 0.6045686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 7.748534571758661e-06,
-      "loss": 0.8652,
+      "learning_rate": 7.139892857509729e-06,
+      "loss": 0.8898,
       "step": 21305
     },
     {
-      "epoch": 0.5851528384279476,
+      "epoch": 0.6045970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 7.747667903056208e-06,
-      "loss": 0.8391,
+      "learning_rate": 7.139012184703771e-06,
+      "loss": 0.8494,
       "step": 21306
     },
     {
-      "epoch": 0.5851803026557908,
+      "epoch": 0.6046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 7.746801252175079e-06,
-      "loss": 0.7832,
+      "learning_rate": 7.138131536064709e-06,
+      "loss": 0.8903,
       "step": 21307
     },
     {
-      "epoch": 0.5852077668836341,
+      "epoch": 0.6046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 7.74593461912213e-06,
-      "loss": 0.767,
+      "learning_rate": 7.137250911599978e-06,
+      "loss": 0.8665,
       "step": 21308
     },
     {
-      "epoch": 0.5852352311114773,
+      "epoch": 0.6046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 7.745068003904221e-06,
-      "loss": 0.8508,
+      "learning_rate": 7.1363703113170135e-06,
+      "loss": 0.8648,
       "step": 21309
     },
     {
-      "epoch": 0.5852626953393205,
+      "epoch": 0.6047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 7.744201406528207e-06,
-      "loss": 0.852,
+      "learning_rate": 7.135489735223262e-06,
+      "loss": 0.837,
       "step": 21310
     },
     {
-      "epoch": 0.5852901595671638,
+      "epoch": 0.604738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 7.743334827000947e-06,
-      "loss": 0.9281,
+      "learning_rate": 7.134609183326156e-06,
+      "loss": 0.7659,
       "step": 21311
     },
     {
-      "epoch": 0.585317623795007,
+      "epoch": 0.6047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 7.742468265329293e-06,
-      "loss": 0.8774,
+      "learning_rate": 7.1337286556331345e-06,
+      "loss": 0.8533,
       "step": 21312
     },
     {
-      "epoch": 0.5853450880228502,
+      "epoch": 0.6047956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 7.741601721520108e-06,
-      "loss": 0.8134,
+      "learning_rate": 7.1328481521516366e-06,
+      "loss": 0.8034,
       "step": 21313
     },
     {
-      "epoch": 0.5853725522506935,
+      "epoch": 0.6048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 7.74073519558024e-06,
-      "loss": 0.6863,
+      "learning_rate": 7.131967672889101e-06,
+      "loss": 0.8889,
       "step": 21314
     },
     {
-      "epoch": 0.5854000164785367,
+      "epoch": 0.6048524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 7.739868687516553e-06,
-      "loss": 0.8176,
+      "learning_rate": 7.131087217852959e-06,
+      "loss": 0.8537,
       "step": 21315
     },
     {
-      "epoch": 0.58542748070638,
+      "epoch": 0.6048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 7.7390021973359e-06,
-      "loss": 1.0087,
+      "learning_rate": 7.130206787050656e-06,
+      "loss": 0.9971,
       "step": 21316
     },
     {
-      "epoch": 0.5854549449342231,
+      "epoch": 0.6049091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 7.738135725045141e-06,
-      "loss": 0.9542,
+      "learning_rate": 7.129326380489625e-06,
+      "loss": 0.7704,
       "step": 21317
     },
     {
-      "epoch": 0.5854824091620664,
+      "epoch": 0.6049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 7.737269270651126e-06,
-      "loss": 0.8196,
+      "learning_rate": 7.128445998177298e-06,
+      "loss": 0.8088,
       "step": 21318
     },
     {
-      "epoch": 0.5855098733899097,
+      "epoch": 0.6049659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 7.736402834160711e-06,
-      "loss": 0.8371,
+      "learning_rate": 7.12756564012112e-06,
+      "loss": 0.854,
       "step": 21319
     },
     {
-      "epoch": 0.5855373376177528,
+      "epoch": 0.604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 7.735536415580755e-06,
-      "loss": 0.8986,
+      "learning_rate": 7.126685306328526e-06,
+      "loss": 0.8322,
       "step": 21320
     },
     {
-      "epoch": 0.5855648018455961,
+      "epoch": 0.6050227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 7.734670014918112e-06,
-      "loss": 0.8221,
+      "learning_rate": 7.12580499680695e-06,
+      "loss": 0.7535,
       "step": 21321
     },
     {
-      "epoch": 0.5855922660734394,
+      "epoch": 0.6050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 7.733803632179635e-06,
-      "loss": 0.9156,
+      "learning_rate": 7.124924711563826e-06,
+      "loss": 0.8298,
       "step": 21322
     },
     {
-      "epoch": 0.5856197303012826,
+      "epoch": 0.6050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 7.732937267372189e-06,
-      "loss": 0.8216,
+      "learning_rate": 7.124044450606595e-06,
+      "loss": 0.9696,
       "step": 21323
     },
     {
-      "epoch": 0.5856471945291258,
+      "epoch": 0.6051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 7.732070920502615e-06,
-      "loss": 0.8002,
+      "learning_rate": 7.1231642139426904e-06,
+      "loss": 0.8212,
       "step": 21324
     },
     {
-      "epoch": 0.585674658756969,
+      "epoch": 0.6051362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 7.731204591577777e-06,
-      "loss": 0.8593,
+      "learning_rate": 7.122284001579546e-06,
+      "loss": 0.8362,
       "step": 21325
     },
     {
-      "epoch": 0.5857021229848123,
+      "epoch": 0.6051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 7.730338280604524e-06,
-      "loss": 0.8905,
+      "learning_rate": 7.121403813524596e-06,
+      "loss": 0.8478,
       "step": 21326
     },
     {
-      "epoch": 0.5857295872126556,
+      "epoch": 0.6051929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 7.729471987589713e-06,
-      "loss": 0.8433,
+      "learning_rate": 7.1205236497852796e-06,
+      "loss": 0.8035,
       "step": 21327
     },
     {
-      "epoch": 0.5857570514404987,
+      "epoch": 0.6052213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 7.728605712540205e-06,
-      "loss": 0.8075,
+      "learning_rate": 7.119643510369029e-06,
+      "loss": 0.9833,
       "step": 21328
     },
     {
-      "epoch": 0.585784515668342,
+      "epoch": 0.6052497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 7.727739455462845e-06,
-      "loss": 0.8453,
+      "learning_rate": 7.118763395283277e-06,
+      "loss": 0.8063,
       "step": 21329
     },
     {
-      "epoch": 0.5858119798961852,
+      "epoch": 0.60527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 7.726873216364492e-06,
-      "loss": 0.8804,
+      "learning_rate": 7.1178833045354625e-06,
+      "loss": 0.7915,
       "step": 21330
     },
     {
-      "epoch": 0.5858394441240284,
+      "epoch": 0.6053064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 7.726006995251997e-06,
-      "loss": 0.9076,
+      "learning_rate": 7.117003238133018e-06,
+      "loss": 0.8024,
       "step": 21331
     },
     {
-      "epoch": 0.5858669083518717,
+      "epoch": 0.6053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 7.725140792132215e-06,
-      "loss": 0.8039,
+      "learning_rate": 7.116123196083373e-06,
+      "loss": 0.8358,
       "step": 21332
     },
     {
-      "epoch": 0.5858943725797149,
+      "epoch": 0.6053632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 7.724274607012001e-06,
-      "loss": 0.8793,
+      "learning_rate": 7.115243178393965e-06,
+      "loss": 0.7629,
       "step": 21333
     },
     {
-      "epoch": 0.5859218368075582,
+      "epoch": 0.6053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 7.723408439898208e-06,
-      "loss": 0.8081,
+      "learning_rate": 7.1143631850722305e-06,
+      "loss": 0.7326,
       "step": 21334
     },
     {
-      "epoch": 0.5859493010354014,
+      "epoch": 0.6054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 7.722542290797692e-06,
-      "loss": 0.9476,
+      "learning_rate": 7.1134832161256004e-06,
+      "loss": 0.8188,
       "step": 21335
     },
     {
-      "epoch": 0.5859767652632446,
+      "epoch": 0.6054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 7.7216761597173e-06,
-      "loss": 0.828,
+      "learning_rate": 7.112603271561505e-06,
+      "loss": 0.8048,
       "step": 21336
     },
     {
-      "epoch": 0.5860042294910879,
+      "epoch": 0.6054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 7.72081004666389e-06,
-      "loss": 0.8465,
+      "learning_rate": 7.111723351387381e-06,
+      "loss": 0.7927,
       "step": 21337
     },
     {
-      "epoch": 0.586031693718931,
+      "epoch": 0.6055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 7.719943951644313e-06,
-      "loss": 0.8661,
+      "learning_rate": 7.110843455610661e-06,
+      "loss": 0.8793,
       "step": 21338
     },
     {
-      "epoch": 0.5860591579467743,
+      "epoch": 0.6055334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 7.719077874665421e-06,
-      "loss": 0.8766,
+      "learning_rate": 7.109963584238772e-06,
+      "loss": 0.7774,
       "step": 21339
     },
     {
-      "epoch": 0.5860866221746176,
+      "epoch": 0.6055618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 7.718211815734072e-06,
-      "loss": 0.8746,
+      "learning_rate": 7.109083737279154e-06,
+      "loss": 0.9086,
       "step": 21340
     },
     {
-      "epoch": 0.5861140864024608,
+      "epoch": 0.6055902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 7.717345774857115e-06,
-      "loss": 0.8369,
+      "learning_rate": 7.108203914739236e-06,
+      "loss": 0.9783,
       "step": 21341
     },
     {
-      "epoch": 0.586141550630304,
+      "epoch": 0.6056186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 7.716479752041401e-06,
-      "loss": 0.8167,
+      "learning_rate": 7.107324116626447e-06,
+      "loss": 0.7855,
       "step": 21342
     },
     {
-      "epoch": 0.5861690148581472,
+      "epoch": 0.6056469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 7.715613747293783e-06,
-      "loss": 0.8412,
+      "learning_rate": 7.106444342948224e-06,
+      "loss": 0.7524,
       "step": 21343
     },
     {
-      "epoch": 0.5861964790859905,
+      "epoch": 0.6056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 7.714747760621114e-06,
-      "loss": 0.8354,
+      "learning_rate": 7.1055645937119956e-06,
+      "loss": 0.8112,
       "step": 21344
     },
     {
-      "epoch": 0.5862239433138338,
+      "epoch": 0.6057037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 7.713881792030247e-06,
-      "loss": 0.9127,
+      "learning_rate": 7.10468486892519e-06,
+      "loss": 0.8678,
       "step": 21345
     },
     {
-      "epoch": 0.5862514075416769,
+      "epoch": 0.6057321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 7.713015841528036e-06,
-      "loss": 0.8688,
+      "learning_rate": 7.103805168595244e-06,
+      "loss": 0.708,
       "step": 21346
     },
     {
-      "epoch": 0.5862788717695202,
+      "epoch": 0.6057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 7.712149909121326e-06,
-      "loss": 0.936,
+      "learning_rate": 7.102925492729587e-06,
+      "loss": 0.9151,
       "step": 21347
     },
     {
-      "epoch": 0.5863063359973635,
+      "epoch": 0.605788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 7.71128399481697e-06,
-      "loss": 0.8198,
+      "learning_rate": 7.102045841335643e-06,
+      "loss": 0.7823,
       "step": 21348
     },
     {
-      "epoch": 0.5863338002252066,
+      "epoch": 0.6058172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 7.710418098621824e-06,
-      "loss": 0.9064,
+      "learning_rate": 7.101166214420851e-06,
+      "loss": 0.8808,
       "step": 21349
     },
     {
-      "epoch": 0.5863612644530499,
+      "epoch": 0.6058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 7.709552220542736e-06,
-      "loss": 0.8248,
+      "learning_rate": 7.10028661199264e-06,
+      "loss": 0.9509,
       "step": 21350
     },
     {
-      "epoch": 0.5863887286808931,
+      "epoch": 0.6058740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 7.708686360586557e-06,
-      "loss": 0.7991,
+      "learning_rate": 7.099407034058438e-06,
+      "loss": 0.9219,
       "step": 21351
     },
     {
-      "epoch": 0.5864161929087364,
+      "epoch": 0.6059023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 7.707820518760144e-06,
-      "loss": 0.9034,
+      "learning_rate": 7.098527480625676e-06,
+      "loss": 0.947,
       "step": 21352
     },
     {
-      "epoch": 0.5864436571365796,
+      "epoch": 0.6059307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 7.706954695070339e-06,
-      "loss": 0.9173,
+      "learning_rate": 7.097647951701779e-06,
+      "loss": 0.8027,
       "step": 21353
     },
     {
-      "epoch": 0.5864711213644228,
+      "epoch": 0.6059591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 7.706088889523999e-06,
-      "loss": 0.8794,
+      "learning_rate": 7.096768447294185e-06,
+      "loss": 0.8353,
       "step": 21354
     },
     {
-      "epoch": 0.5864985855922661,
+      "epoch": 0.6059875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 7.705223102127969e-06,
-      "loss": 0.8145,
+      "learning_rate": 7.095888967410316e-06,
+      "loss": 0.7769,
       "step": 21355
     },
     {
-      "epoch": 0.5865260498201094,
+      "epoch": 0.6060158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 7.704357332889103e-06,
-      "loss": 0.8991,
+      "learning_rate": 7.095009512057602e-06,
+      "loss": 0.9371,
       "step": 21356
     },
     {
-      "epoch": 0.5865535140479525,
+      "epoch": 0.6060442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 7.703491581814253e-06,
-      "loss": 0.8563,
+      "learning_rate": 7.094130081243475e-06,
+      "loss": 0.8512,
       "step": 21357
     },
     {
-      "epoch": 0.5865809782757958,
+      "epoch": 0.6060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 7.70262584891027e-06,
-      "loss": 0.8266,
+      "learning_rate": 7.093250674975363e-06,
+      "loss": 0.8736,
       "step": 21358
     },
     {
-      "epoch": 0.586608442503639,
+      "epoch": 0.6061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 7.701760134183997e-06,
-      "loss": 0.9081,
+      "learning_rate": 7.092371293260691e-06,
+      "loss": 0.8368,
       "step": 21359
     },
     {
-      "epoch": 0.5866359067314822,
+      "epoch": 0.6061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 7.700894437642288e-06,
-      "loss": 0.764,
+      "learning_rate": 7.0914919361068915e-06,
+      "loss": 0.8214,
       "step": 21360
     },
     {
-      "epoch": 0.5866633709593255,
+      "epoch": 0.6061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 7.700028759291995e-06,
-      "loss": 0.8091,
+      "learning_rate": 7.09061260352139e-06,
+      "loss": 0.8335,
       "step": 21361
     },
     {
-      "epoch": 0.5866908351871687,
+      "epoch": 0.6061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 7.699163099139963e-06,
-      "loss": 0.9391,
+      "learning_rate": 7.0897332955116115e-06,
+      "loss": 0.7861,
       "step": 21362
     },
     {
-      "epoch": 0.586718299415012,
+      "epoch": 0.6062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 7.698297457193044e-06,
-      "loss": 0.9377,
+      "learning_rate": 7.08885401208499e-06,
+      "loss": 0.7756,
       "step": 21363
     },
     {
-      "epoch": 0.5867457636428551,
+      "epoch": 0.6062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 7.69743183345809e-06,
-      "loss": 0.8882,
+      "learning_rate": 7.0879747532489455e-06,
+      "loss": 0.8982,
       "step": 21364
     },
     {
-      "epoch": 0.5867732278706984,
+      "epoch": 0.6062712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 7.696566227941944e-06,
-      "loss": 0.8495,
+      "learning_rate": 7.087095519010914e-06,
+      "loss": 0.792,
       "step": 21365
     },
     {
-      "epoch": 0.5868006920985417,
+      "epoch": 0.606299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 7.695700640651462e-06,
-      "loss": 0.8628,
+      "learning_rate": 7.086216309378315e-06,
+      "loss": 0.8905,
       "step": 21366
     },
     {
-      "epoch": 0.5868281563263849,
+      "epoch": 0.606328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 7.694835071593485e-06,
-      "loss": 0.9472,
+      "learning_rate": 7.085337124358578e-06,
+      "loss": 0.8217,
       "step": 21367
     },
     {
-      "epoch": 0.5868556205542281,
+      "epoch": 0.6063564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 7.693969520774867e-06,
-      "loss": 0.8994,
+      "learning_rate": 7.084457963959131e-06,
+      "loss": 0.9547,
       "step": 21368
     },
     {
-      "epoch": 0.5868830847820714,
+      "epoch": 0.6063847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 7.693103988202456e-06,
-      "loss": 0.8092,
+      "learning_rate": 7.0835788281873985e-06,
+      "loss": 0.8212,
       "step": 21369
     },
     {
-      "epoch": 0.5869105490099146,
+      "epoch": 0.6064131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 7.692238473883102e-06,
-      "loss": 0.8328,
+      "learning_rate": 7.082699717050804e-06,
+      "loss": 0.8744,
       "step": 21370
     },
     {
-      "epoch": 0.5869380132377578,
+      "epoch": 0.6064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 7.69137297782365e-06,
-      "loss": 0.9312,
+      "learning_rate": 7.0818206305567795e-06,
+      "loss": 0.8185,
       "step": 21371
     },
     {
-      "epoch": 0.586965477465601,
+      "epoch": 0.6064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 7.690507500030946e-06,
-      "loss": 0.8625,
+      "learning_rate": 7.0809415687127455e-06,
+      "loss": 0.8585,
       "step": 21372
     },
     {
-      "epoch": 0.5869929416934443,
+      "epoch": 0.6064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 7.689642040511844e-06,
-      "loss": 0.8616,
+      "learning_rate": 7.0800625315261285e-06,
+      "loss": 0.9144,
       "step": 21373
     },
     {
-      "epoch": 0.5870204059212876,
+      "epoch": 0.6065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 7.688776599273187e-06,
-      "loss": 0.927,
+      "learning_rate": 7.079183519004355e-06,
+      "loss": 0.8592,
       "step": 21374
     },
     {
-      "epoch": 0.5870478701491307,
+      "epoch": 0.6065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 7.687911176321826e-06,
-      "loss": 0.8136,
+      "learning_rate": 7.078304531154853e-06,
+      "loss": 0.8557,
       "step": 21375
     },
     {
-      "epoch": 0.587075334376974,
+      "epoch": 0.6065834279228149,
       "grad_norm": 0.0,
-      "learning_rate": 7.68704577166461e-06,
-      "loss": 0.8817,
+      "learning_rate": 7.077425567985039e-06,
+      "loss": 0.906,
       "step": 21376
     },
     {
-      "epoch": 0.5871027986048172,
+      "epoch": 0.6066118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 7.686180385308382e-06,
-      "loss": 0.8931,
+      "learning_rate": 7.076546629502345e-06,
+      "loss": 0.8928,
       "step": 21377
     },
     {
-      "epoch": 0.5871302628326605,
+      "epoch": 0.6066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 7.68531501725999e-06,
-      "loss": 0.9476,
+      "learning_rate": 7.075667715714194e-06,
+      "loss": 0.9694,
       "step": 21378
     },
     {
-      "epoch": 0.5871577270605037,
+      "epoch": 0.6066685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 7.684449667526284e-06,
-      "loss": 0.8946,
+      "learning_rate": 7.074788826628005e-06,
+      "loss": 0.8969,
       "step": 21379
     },
     {
-      "epoch": 0.5871851912883469,
+      "epoch": 0.6066969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 7.683584336114107e-06,
-      "loss": 0.7722,
+      "learning_rate": 7.0739099622512085e-06,
+      "loss": 0.8332,
       "step": 21380
     },
     {
-      "epoch": 0.5872126555161902,
+      "epoch": 0.6067253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 7.68271902303031e-06,
-      "loss": 0.798,
+      "learning_rate": 7.073031122591227e-06,
+      "loss": 0.8382,
       "step": 21381
     },
     {
-      "epoch": 0.5872401197440335,
+      "epoch": 0.6067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 7.68185372828174e-06,
-      "loss": 0.9102,
+      "learning_rate": 7.072152307655485e-06,
+      "loss": 0.8228,
       "step": 21382
     },
     {
-      "epoch": 0.5872675839718766,
+      "epoch": 0.6067820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 7.680988451875237e-06,
-      "loss": 0.9232,
+      "learning_rate": 7.0712735174514e-06,
+      "loss": 0.822,
       "step": 21383
     },
     {
-      "epoch": 0.5872950481997199,
+      "epoch": 0.6068104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 7.68012319381765e-06,
-      "loss": 0.7945,
+      "learning_rate": 7.070394751986402e-06,
+      "loss": 0.8498,
       "step": 21384
     },
     {
-      "epoch": 0.5873225124275631,
+      "epoch": 0.606838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 7.679257954115832e-06,
-      "loss": 0.8736,
+      "learning_rate": 7.069516011267912e-06,
+      "loss": 0.8364,
       "step": 21385
     },
     {
-      "epoch": 0.5873499766554063,
+      "epoch": 0.6068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 7.678392732776621e-06,
-      "loss": 0.983,
+      "learning_rate": 7.068637295303349e-06,
+      "loss": 0.7627,
       "step": 21386
     },
     {
-      "epoch": 0.5873774408832496,
+      "epoch": 0.6068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 7.677527529806865e-06,
-      "loss": 0.9283,
+      "learning_rate": 7.067758604100141e-06,
+      "loss": 0.9527,
       "step": 21387
     },
     {
-      "epoch": 0.5874049051110928,
+      "epoch": 0.6069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 7.676662345213416e-06,
-      "loss": 0.8587,
+      "learning_rate": 7.066879937665708e-06,
+      "loss": 0.776,
       "step": 21388
     },
     {
-      "epoch": 0.5874323693389361,
+      "epoch": 0.6069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 7.675797179003109e-06,
-      "loss": 0.8284,
+      "learning_rate": 7.066001296007469e-06,
+      "loss": 0.8446,
       "step": 21389
     },
     {
-      "epoch": 0.5874598335667792,
+      "epoch": 0.6069807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 7.674932031182797e-06,
-      "loss": 0.8617,
+      "learning_rate": 7.065122679132853e-06,
+      "loss": 0.8067,
       "step": 21390
     },
     {
-      "epoch": 0.5874872977946225,
+      "epoch": 0.6070090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 7.674066901759324e-06,
-      "loss": 0.8229,
+      "learning_rate": 7.0642440870492766e-06,
+      "loss": 0.8739,
       "step": 21391
     },
     {
-      "epoch": 0.5875147620224658,
+      "epoch": 0.6070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 7.673201790739533e-06,
-      "loss": 0.8831,
+      "learning_rate": 7.063365519764163e-06,
+      "loss": 0.9392,
       "step": 21392
     },
     {
-      "epoch": 0.5875422262503089,
+      "epoch": 0.6070658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 7.672336698130272e-06,
-      "loss": 0.8905,
+      "learning_rate": 7.062486977284929e-06,
+      "loss": 0.8183,
       "step": 21393
     },
     {
-      "epoch": 0.5875696904781522,
+      "epoch": 0.6070942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 7.671471623938389e-06,
-      "loss": 0.8594,
+      "learning_rate": 7.061608459618998e-06,
+      "loss": 0.9487,
       "step": 21394
     },
     {
-      "epoch": 0.5875971547059955,
+      "epoch": 0.6071225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 7.67060656817072e-06,
-      "loss": 0.8988,
+      "learning_rate": 7.060729966773798e-06,
+      "loss": 0.7769,
       "step": 21395
     },
     {
-      "epoch": 0.5876246189338387,
+      "epoch": 0.6071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 7.669741530834112e-06,
-      "loss": 0.9568,
+      "learning_rate": 7.059851498756743e-06,
+      "loss": 0.9258,
       "step": 21396
     },
     {
-      "epoch": 0.5876520831616819,
+      "epoch": 0.6071793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 7.668876511935415e-06,
-      "loss": 0.8929,
+      "learning_rate": 7.058973055575253e-06,
+      "loss": 0.9033,
       "step": 21397
     },
     {
-      "epoch": 0.5876795473895251,
+      "epoch": 0.6072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 7.668011511481467e-06,
-      "loss": 0.8885,
+      "learning_rate": 7.058094637236752e-06,
+      "loss": 0.8617,
       "step": 21398
     },
     {
-      "epoch": 0.5877070116173684,
+      "epoch": 0.6072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 7.667146529479117e-06,
-      "loss": 0.8444,
+      "learning_rate": 7.0572162437486565e-06,
+      "loss": 0.9134,
       "step": 21399
     },
     {
-      "epoch": 0.5877344758452117,
+      "epoch": 0.6072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 7.66628156593521e-06,
-      "loss": 0.8639,
+      "learning_rate": 7.056337875118386e-06,
+      "loss": 0.9118,
       "step": 21400
     },
     {
-      "epoch": 0.5877619400730548,
+      "epoch": 0.6072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 7.665416620856583e-06,
-      "loss": 0.9165,
+      "learning_rate": 7.0554595313533655e-06,
+      "loss": 0.7776,
       "step": 21401
     },
     {
-      "epoch": 0.5877894043008981,
+      "epoch": 0.6073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 7.664551694250088e-06,
-      "loss": 0.9578,
+      "learning_rate": 7.054581212461009e-06,
+      "loss": 0.7936,
       "step": 21402
     },
     {
-      "epoch": 0.5878168685287413,
+      "epoch": 0.607349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 7.663686786122562e-06,
-      "loss": 0.8217,
+      "learning_rate": 7.053702918448737e-06,
+      "loss": 0.8448,
       "step": 21403
     },
     {
-      "epoch": 0.5878443327565845,
+      "epoch": 0.607377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 7.66282189648085e-06,
-      "loss": 0.7212,
+      "learning_rate": 7.052824649323969e-06,
+      "loss": 0.7991,
       "step": 21404
     },
     {
-      "epoch": 0.5878717969844278,
+      "epoch": 0.6074063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 7.6619570253318e-06,
-      "loss": 0.8606,
+      "learning_rate": 7.0519464050941256e-06,
+      "loss": 0.9001,
       "step": 21405
     },
     {
-      "epoch": 0.587899261212271,
+      "epoch": 0.6074347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 7.661092172682252e-06,
-      "loss": 0.8163,
+      "learning_rate": 7.051068185766619e-06,
+      "loss": 0.7578,
       "step": 21406
     },
     {
-      "epoch": 0.5879267254401143,
+      "epoch": 0.6074631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 7.660227338539044e-06,
-      "loss": 0.9824,
+      "learning_rate": 7.050189991348877e-06,
+      "loss": 0.8686,
       "step": 21407
     },
     {
-      "epoch": 0.5879541896679575,
+      "epoch": 0.6074914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 7.659362522909027e-06,
-      "loss": 0.865,
+      "learning_rate": 7.04931182184831e-06,
+      "loss": 0.8129,
       "step": 21408
     },
     {
-      "epoch": 0.5879816538958007,
+      "epoch": 0.6075198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 7.658497725799039e-06,
-      "loss": 0.8414,
+      "learning_rate": 7.048433677272338e-06,
+      "loss": 0.8517,
       "step": 21409
     },
     {
-      "epoch": 0.588009118123644,
+      "epoch": 0.6075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 7.657632947215923e-06,
-      "loss": 0.8501,
+      "learning_rate": 7.0475555576283796e-06,
+      "loss": 0.9618,
       "step": 21410
     },
     {
-      "epoch": 0.5880365823514871,
+      "epoch": 0.6075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 7.656768187166528e-06,
-      "loss": 0.8275,
+      "learning_rate": 7.0466774629238525e-06,
+      "loss": 0.7693,
       "step": 21411
     },
     {
-      "epoch": 0.5880640465793304,
+      "epoch": 0.6076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 7.655903445657687e-06,
-      "loss": 0.888,
+      "learning_rate": 7.045799393166174e-06,
+      "loss": 0.8985,
       "step": 21412
     },
     {
-      "epoch": 0.5880915108071737,
+      "epoch": 0.607633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 7.655038722696243e-06,
-      "loss": 0.9042,
+      "learning_rate": 7.044921348362761e-06,
+      "loss": 0.8145,
       "step": 21413
     },
     {
-      "epoch": 0.5881189750350169,
+      "epoch": 0.6076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 7.654174018289045e-06,
-      "loss": 0.9823,
+      "learning_rate": 7.04404332852103e-06,
+      "loss": 0.9194,
       "step": 21414
     },
     {
-      "epoch": 0.5881464392628601,
+      "epoch": 0.6076901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 7.653309332442928e-06,
-      "loss": 0.9469,
+      "learning_rate": 7.043165333648399e-06,
+      "loss": 0.9042,
       "step": 21415
     },
     {
-      "epoch": 0.5881739034907033,
+      "epoch": 0.6077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 7.652444665164738e-06,
-      "loss": 0.7904,
+      "learning_rate": 7.0422873637522825e-06,
+      "loss": 0.8492,
       "step": 21416
     },
     {
-      "epoch": 0.5882013677185466,
+      "epoch": 0.6077468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 7.651580016461318e-06,
-      "loss": 0.8396,
+      "learning_rate": 7.041409418840096e-06,
+      "loss": 0.9067,
       "step": 21417
     },
     {
-      "epoch": 0.5882288319463899,
+      "epoch": 0.6077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 7.650715386339505e-06,
-      "loss": 0.9994,
+      "learning_rate": 7.04053149891926e-06,
+      "loss": 0.8027,
       "step": 21418
     },
     {
-      "epoch": 0.588256296174233,
+      "epoch": 0.6078036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 7.64985077480614e-06,
-      "loss": 0.8334,
+      "learning_rate": 7.039653603997187e-06,
+      "loss": 0.9025,
       "step": 21419
     },
     {
-      "epoch": 0.5882837604020763,
+      "epoch": 0.6078320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 7.648986181868066e-06,
-      "loss": 0.8702,
+      "learning_rate": 7.038775734081291e-06,
+      "loss": 0.8436,
       "step": 21420
     },
     {
-      "epoch": 0.5883112246299196,
+      "epoch": 0.6078603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 7.648121607532125e-06,
-      "loss": 0.8864,
+      "learning_rate": 7.037897889178991e-06,
+      "loss": 0.8836,
       "step": 21421
     },
     {
-      "epoch": 0.5883386888577627,
+      "epoch": 0.60788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 7.647257051805156e-06,
-      "loss": 0.812,
+      "learning_rate": 7.037020069297702e-06,
+      "loss": 0.9933,
       "step": 21422
     },
     {
-      "epoch": 0.588366153085606,
+      "epoch": 0.6079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 7.646392514694004e-06,
-      "loss": 0.8382,
+      "learning_rate": 7.036142274444834e-06,
+      "loss": 0.8848,
       "step": 21423
     },
     {
-      "epoch": 0.5883936173134492,
+      "epoch": 0.6079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 7.645527996205502e-06,
-      "loss": 0.7509,
+      "learning_rate": 7.0352645046278075e-06,
+      "loss": 0.7268,
       "step": 21424
     },
     {
-      "epoch": 0.5884210815412925,
+      "epoch": 0.6079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 7.644663496346496e-06,
-      "loss": 0.8483,
+      "learning_rate": 7.034386759854033e-06,
+      "loss": 0.9641,
       "step": 21425
     },
     {
-      "epoch": 0.5884485457691357,
+      "epoch": 0.6080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 7.643799015123824e-06,
-      "loss": 0.7674,
+      "learning_rate": 7.0335090401309295e-06,
+      "loss": 0.935,
       "step": 21426
     },
     {
-      "epoch": 0.5884760099969789,
+      "epoch": 0.6080306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 7.642934552544327e-06,
-      "loss": 0.8345,
+      "learning_rate": 7.032631345465905e-06,
+      "loss": 0.8532,
       "step": 21427
     },
     {
-      "epoch": 0.5885034742248222,
+      "epoch": 0.6080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 7.642070108614846e-06,
-      "loss": 0.8111,
+      "learning_rate": 7.031753675866382e-06,
+      "loss": 0.9177,
       "step": 21428
     },
     {
-      "epoch": 0.5885309384526654,
+      "epoch": 0.6080874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 7.641205683342222e-06,
-      "loss": 0.9527,
+      "learning_rate": 7.030876031339768e-06,
+      "loss": 0.9937,
       "step": 21429
     },
     {
-      "epoch": 0.5885584026805086,
+      "epoch": 0.6081157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 7.640341276733291e-06,
-      "loss": 0.7858,
+      "learning_rate": 7.029998411893475e-06,
+      "loss": 0.8699,
       "step": 21430
     },
     {
-      "epoch": 0.5885858669083519,
+      "epoch": 0.6081441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 7.639476888794892e-06,
-      "loss": 0.8508,
+      "learning_rate": 7.0291208175349206e-06,
+      "loss": 0.847,
       "step": 21431
     },
     {
-      "epoch": 0.5886133311361951,
+      "epoch": 0.608172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 7.638612519533865e-06,
-      "loss": 0.8171,
+      "learning_rate": 7.0282432482715165e-06,
+      "loss": 0.8461,
       "step": 21432
     },
     {
-      "epoch": 0.5886407953640383,
+      "epoch": 0.6082009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 7.637748168957051e-06,
-      "loss": 0.8254,
+      "learning_rate": 7.027365704110676e-06,
+      "loss": 0.9038,
       "step": 21433
     },
     {
-      "epoch": 0.5886682595918816,
+      "epoch": 0.6082292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 7.636883837071289e-06,
-      "loss": 0.8979,
+      "learning_rate": 7.026488185059808e-06,
+      "loss": 0.8497,
       "step": 21434
     },
     {
-      "epoch": 0.5886957238197248,
+      "epoch": 0.6082576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 7.636019523883422e-06,
-      "loss": 0.8656,
+      "learning_rate": 7.025610691126331e-06,
+      "loss": 0.9485,
       "step": 21435
     },
     {
-      "epoch": 0.5887231880475681,
+      "epoch": 0.6082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 7.63515522940028e-06,
-      "loss": 0.8967,
+      "learning_rate": 7.024733222317655e-06,
+      "loss": 0.9268,
       "step": 21436
     },
     {
-      "epoch": 0.5887506522754112,
+      "epoch": 0.6083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 7.634290953628705e-06,
-      "loss": 0.8699,
+      "learning_rate": 7.023855778641186e-06,
+      "loss": 0.7743,
       "step": 21437
     },
     {
-      "epoch": 0.5887781165032545,
+      "epoch": 0.6083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 7.633426696575538e-06,
-      "loss": 0.8328,
+      "learning_rate": 7.022978360104347e-06,
+      "loss": 0.8722,
       "step": 21438
     },
     {
-      "epoch": 0.5888055807310978,
+      "epoch": 0.6083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 7.632562458247614e-06,
-      "loss": 0.8081,
+      "learning_rate": 7.0221009667145405e-06,
+      "loss": 0.9257,
       "step": 21439
     },
     {
-      "epoch": 0.588833044958941,
+      "epoch": 0.6083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 7.631698238651773e-06,
-      "loss": 0.8044,
+      "learning_rate": 7.0212235984791785e-06,
+      "loss": 0.9016,
       "step": 21440
     },
     {
-      "epoch": 0.5888605091867842,
+      "epoch": 0.6084279228149829,
       "grad_norm": 0.0,
-      "learning_rate": 7.630834037794857e-06,
-      "loss": 0.8818,
+      "learning_rate": 7.0203462554056755e-06,
+      "loss": 0.8693,
       "step": 21441
     },
     {
-      "epoch": 0.5888879734146274,
+      "epoch": 0.6084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 7.629969855683698e-06,
-      "loss": 0.807,
+      "learning_rate": 7.019468937501445e-06,
+      "loss": 0.7343,
       "step": 21442
     },
     {
-      "epoch": 0.5889154376424707,
+      "epoch": 0.6084846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 7.629105692325134e-06,
-      "loss": 0.9457,
+      "learning_rate": 7.018591644773893e-06,
+      "loss": 0.8316,
       "step": 21443
     },
     {
-      "epoch": 0.588942901870314,
+      "epoch": 0.6085130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 7.628241547726003e-06,
-      "loss": 0.9588,
+      "learning_rate": 7.017714377230432e-06,
+      "loss": 0.8506,
       "step": 21444
     },
     {
-      "epoch": 0.5889703660981571,
+      "epoch": 0.6085414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 7.627377421893147e-06,
-      "loss": 0.7324,
+      "learning_rate": 7.016837134878471e-06,
+      "loss": 0.7858,
       "step": 21445
     },
     {
-      "epoch": 0.5889978303260004,
+      "epoch": 0.6085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 7.626513314833397e-06,
-      "loss": 0.7631,
+      "learning_rate": 7.0159599177254215e-06,
+      "loss": 0.8649,
       "step": 21446
     },
     {
-      "epoch": 0.5890252945538437,
+      "epoch": 0.6085981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 7.625649226553598e-06,
-      "loss": 0.8631,
+      "learning_rate": 7.01508272577869e-06,
+      "loss": 0.7576,
       "step": 21447
     },
     {
-      "epoch": 0.5890527587816868,
+      "epoch": 0.6086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 7.624785157060577e-06,
-      "loss": 0.8424,
+      "learning_rate": 7.014205559045692e-06,
+      "loss": 0.8705,
       "step": 21448
     },
     {
-      "epoch": 0.5890802230095301,
+      "epoch": 0.6086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 7.6239211063611775e-06,
-      "loss": 0.75,
+      "learning_rate": 7.013328417533834e-06,
+      "loss": 0.9221,
       "step": 21449
     },
     {
-      "epoch": 0.5891076872373733,
+      "epoch": 0.608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 7.623057074462233e-06,
-      "loss": 0.827,
+      "learning_rate": 7.012451301250523e-06,
+      "loss": 0.7613,
       "step": 21450
     },
     {
-      "epoch": 0.5891351514652166,
+      "epoch": 0.6087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 7.622193061370582e-06,
-      "loss": 0.7895,
+      "learning_rate": 7.011574210203173e-06,
+      "loss": 0.9301,
       "step": 21451
     },
     {
-      "epoch": 0.5891626156930598,
+      "epoch": 0.6087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 7.621329067093061e-06,
-      "loss": 0.889,
+      "learning_rate": 7.0106971443991874e-06,
+      "loss": 0.9135,
       "step": 21452
     },
     {
-      "epoch": 0.589190079920903,
+      "epoch": 0.6087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 7.620465091636509e-06,
-      "loss": 0.844,
+      "learning_rate": 7.0098201038459775e-06,
+      "loss": 0.878,
       "step": 21453
     },
     {
-      "epoch": 0.5892175441487463,
+      "epoch": 0.6087968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 7.619601135007757e-06,
-      "loss": 0.9015,
+      "learning_rate": 7.008943088550954e-06,
+      "loss": 0.946,
       "step": 21454
     },
     {
-      "epoch": 0.5892450083765894,
+      "epoch": 0.6088251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 7.618737197213642e-06,
-      "loss": 0.8594,
+      "learning_rate": 7.008066098521522e-06,
+      "loss": 0.9799,
       "step": 21455
     },
     {
-      "epoch": 0.5892724726044327,
+      "epoch": 0.6088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 7.617873278260999e-06,
-      "loss": 0.8249,
+      "learning_rate": 7.007189133765086e-06,
+      "loss": 0.8891,
       "step": 21456
     },
     {
-      "epoch": 0.589299936832276,
+      "epoch": 0.6088819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 7.617009378156668e-06,
-      "loss": 0.7459,
+      "learning_rate": 7.006312194289061e-06,
+      "loss": 0.8754,
       "step": 21457
     },
     {
-      "epoch": 0.5893274010601192,
+      "epoch": 0.6089103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 7.61614549690748e-06,
-      "loss": 0.9071,
+      "learning_rate": 7.005435280100849e-06,
+      "loss": 0.8727,
       "step": 21458
     },
     {
-      "epoch": 0.5893548652879624,
+      "epoch": 0.6089387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 7.615281634520277e-06,
-      "loss": 0.8758,
+      "learning_rate": 7.0045583912078655e-06,
+      "loss": 0.9204,
       "step": 21459
     },
     {
-      "epoch": 0.5893823295158057,
+      "epoch": 0.6089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 7.614417791001885e-06,
-      "loss": 0.917,
+      "learning_rate": 7.003681527617508e-06,
+      "loss": 0.9112,
       "step": 21460
     },
     {
-      "epoch": 0.5894097937436489,
+      "epoch": 0.6089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 7.613553966359145e-06,
-      "loss": 0.8334,
+      "learning_rate": 7.002804689337188e-06,
+      "loss": 0.8666,
       "step": 21461
     },
     {
-      "epoch": 0.5894372579714922,
+      "epoch": 0.6090238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 7.6126901605988886e-06,
-      "loss": 0.8624,
+      "learning_rate": 7.001927876374313e-06,
+      "loss": 0.891,
       "step": 21462
     },
     {
-      "epoch": 0.5894647221993353,
+      "epoch": 0.6090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 7.611826373727953e-06,
-      "loss": 0.8904,
+      "learning_rate": 7.001051088736286e-06,
+      "loss": 0.8149,
       "step": 21463
     },
     {
-      "epoch": 0.5894921864271786,
+      "epoch": 0.6090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 7.610962605753172e-06,
-      "loss": 0.9137,
+      "learning_rate": 7.000174326430515e-06,
+      "loss": 0.821,
       "step": 21464
     },
     {
-      "epoch": 0.5895196506550219,
+      "epoch": 0.6091089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 7.610098856681384e-06,
-      "loss": 0.8276,
+      "learning_rate": 6.999297589464409e-06,
+      "loss": 0.8346,
       "step": 21465
     },
     {
-      "epoch": 0.589547114882865,
+      "epoch": 0.6091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 7.609235126519417e-06,
-      "loss": 0.8833,
+      "learning_rate": 6.9984208778453685e-06,
+      "loss": 0.8101,
       "step": 21466
     },
     {
-      "epoch": 0.5895745791107083,
+      "epoch": 0.6091657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 7.6083714152741075e-06,
-      "loss": 0.7495,
+      "learning_rate": 6.997544191580803e-06,
+      "loss": 0.8327,
       "step": 21467
     },
     {
-      "epoch": 0.5896020433385515,
+      "epoch": 0.609194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 7.607507722952287e-06,
-      "loss": 0.8628,
+      "learning_rate": 6.996667530678116e-06,
+      "loss": 0.8116,
       "step": 21468
     },
     {
-      "epoch": 0.5896295075663948,
+      "epoch": 0.60922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 7.606644049560796e-06,
-      "loss": 0.7857,
+      "learning_rate": 6.9957908951447136e-06,
+      "loss": 0.8978,
       "step": 21469
     },
     {
-      "epoch": 0.589656971794238,
+      "epoch": 0.6092508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 7.605780395106468e-06,
-      "loss": 0.8143,
+      "learning_rate": 6.9949142849880015e-06,
+      "loss": 0.8051,
       "step": 21470
     },
     {
-      "epoch": 0.5896844360220812,
+      "epoch": 0.6092792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 7.604916759596126e-06,
-      "loss": 0.725,
+      "learning_rate": 6.994037700215378e-06,
+      "loss": 0.7544,
       "step": 21471
     },
     {
-      "epoch": 0.5897119002499245,
+      "epoch": 0.6093076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 7.604053143036612e-06,
-      "loss": 0.9244,
+      "learning_rate": 6.993161140834259e-06,
+      "loss": 0.7725,
       "step": 21472
     },
     {
-      "epoch": 0.5897393644777678,
+      "epoch": 0.6093359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 7.603189545434759e-06,
-      "loss": 0.805,
+      "learning_rate": 6.9922846068520425e-06,
+      "loss": 0.8087,
       "step": 21473
     },
     {
-      "epoch": 0.5897668287056109,
+      "epoch": 0.6093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 7.602325966797397e-06,
-      "loss": 0.9078,
+      "learning_rate": 6.991408098276134e-06,
+      "loss": 0.8629,
       "step": 21474
     },
     {
-      "epoch": 0.5897942929334542,
+      "epoch": 0.6093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 7.601462407131362e-06,
-      "loss": 0.949,
+      "learning_rate": 6.990531615113934e-06,
+      "loss": 0.8734,
       "step": 21475
     },
     {
-      "epoch": 0.5898217571612974,
+      "epoch": 0.6094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 7.6005988664434874e-06,
-      "loss": 0.8504,
+      "learning_rate": 6.9896551573728495e-06,
+      "loss": 0.8142,
       "step": 21476
     },
     {
-      "epoch": 0.5898492213891406,
+      "epoch": 0.6094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 7.5997353447406004e-06,
-      "loss": 0.8066,
+      "learning_rate": 6.988778725060285e-06,
+      "loss": 0.7237,
       "step": 21477
     },
     {
-      "epoch": 0.5898766856169839,
+      "epoch": 0.609477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 7.598871842029541e-06,
-      "loss": 0.7198,
+      "learning_rate": 6.987902318183639e-06,
+      "loss": 0.841,
       "step": 21478
     },
     {
-      "epoch": 0.5899041498448271,
+      "epoch": 0.6095062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 7.5980083583171325e-06,
-      "loss": 0.9074,
+      "learning_rate": 6.98702593675032e-06,
+      "loss": 0.849,
       "step": 21479
     },
     {
-      "epoch": 0.5899316140726704,
+      "epoch": 0.6095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 7.597144893610214e-06,
-      "loss": 0.8163,
+      "learning_rate": 6.986149580767727e-06,
+      "loss": 0.7331,
       "step": 21480
     },
     {
-      "epoch": 0.5899590783005135,
+      "epoch": 0.6095629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 7.596281447915619e-06,
-      "loss": 0.8434,
+      "learning_rate": 6.9852732502432645e-06,
+      "loss": 0.7743,
       "step": 21481
     },
     {
-      "epoch": 0.5899865425283568,
+      "epoch": 0.6095913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 7.595418021240177e-06,
-      "loss": 0.881,
+      "learning_rate": 6.984396945184335e-06,
+      "loss": 0.9094,
       "step": 21482
     },
     {
-      "epoch": 0.5900140067562001,
+      "epoch": 0.6096197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 7.594554613590715e-06,
-      "loss": 0.9305,
+      "learning_rate": 6.98352066559834e-06,
+      "loss": 0.8539,
       "step": 21483
     },
     {
-      "epoch": 0.5900414709840432,
+      "epoch": 0.6096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 7.5936912249740694e-06,
-      "loss": 0.9943,
+      "learning_rate": 6.9826444114926785e-06,
+      "loss": 0.8852,
       "step": 21484
     },
     {
-      "epoch": 0.5900689352118865,
+      "epoch": 0.6096765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 7.592827855397072e-06,
-      "loss": 0.8367,
+      "learning_rate": 6.9817681828747595e-06,
+      "loss": 0.9469,
       "step": 21485
     },
     {
-      "epoch": 0.5900963994397298,
+      "epoch": 0.6097048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 7.591964504866552e-06,
-      "loss": 0.8683,
+      "learning_rate": 6.980891979751974e-06,
+      "loss": 0.9614,
       "step": 21486
     },
     {
-      "epoch": 0.590123863667573,
+      "epoch": 0.609733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 7.591101173389341e-06,
-      "loss": 0.8332,
+      "learning_rate": 6.980015802131735e-06,
+      "loss": 0.7786,
       "step": 21487
     },
     {
-      "epoch": 0.5901513278954162,
+      "epoch": 0.6097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 7.590237860972275e-06,
-      "loss": 0.8793,
+      "learning_rate": 6.979139650021436e-06,
+      "loss": 0.9517,
       "step": 21488
     },
     {
-      "epoch": 0.5901787921232594,
+      "epoch": 0.6097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 7.5893745676221765e-06,
-      "loss": 0.8329,
+      "learning_rate": 6.978263523428483e-06,
+      "loss": 0.8544,
       "step": 21489
     },
     {
-      "epoch": 0.5902062563511027,
+      "epoch": 0.6098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 7.588511293345881e-06,
-      "loss": 0.8225,
+      "learning_rate": 6.977387422360273e-06,
+      "loss": 0.8748,
       "step": 21490
     },
     {
-      "epoch": 0.590233720578946,
+      "epoch": 0.6098467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 7.587648038150218e-06,
-      "loss": 0.9604,
+      "learning_rate": 6.976511346824205e-06,
+      "loss": 0.9496,
       "step": 21491
     },
     {
-      "epoch": 0.5902611848067891,
+      "epoch": 0.6098751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 7.5867848020420176e-06,
-      "loss": 0.7927,
+      "learning_rate": 6.975635296827685e-06,
+      "loss": 0.769,
       "step": 21492
     },
     {
-      "epoch": 0.5902886490346324,
+      "epoch": 0.6099035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 7.585921585028113e-06,
-      "loss": 0.9047,
+      "learning_rate": 6.974759272378109e-06,
+      "loss": 0.8872,
       "step": 21493
     },
     {
-      "epoch": 0.5903161132624756,
+      "epoch": 0.6099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 7.585058387115334e-06,
-      "loss": 0.8365,
+      "learning_rate": 6.973883273482874e-06,
+      "loss": 0.7651,
       "step": 21494
     },
     {
-      "epoch": 0.5903435774903188,
+      "epoch": 0.6099602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 7.584195208310503e-06,
-      "loss": 0.8374,
+      "learning_rate": 6.973007300149386e-06,
+      "loss": 0.8373,
       "step": 21495
     },
     {
-      "epoch": 0.5903710417181621,
+      "epoch": 0.6099886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 7.583332048620457e-06,
-      "loss": 0.8799,
+      "learning_rate": 6.972131352385042e-06,
+      "loss": 0.8526,
       "step": 21496
     },
     {
-      "epoch": 0.5903985059460053,
+      "epoch": 0.610017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 7.582468908052025e-06,
-      "loss": 0.8486,
+      "learning_rate": 6.971255430197238e-06,
+      "loss": 0.8986,
       "step": 21497
     },
     {
-      "epoch": 0.5904259701738486,
+      "epoch": 0.6100454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 7.581605786612034e-06,
-      "loss": 0.7468,
+      "learning_rate": 6.9703795335933775e-06,
+      "loss": 0.896,
       "step": 21498
     },
     {
-      "epoch": 0.5904534344016918,
+      "epoch": 0.6100737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 7.580742684307314e-06,
-      "loss": 0.8654,
+      "learning_rate": 6.969503662580858e-06,
+      "loss": 0.9045,
       "step": 21499
     },
     {
-      "epoch": 0.590480898629535,
+      "epoch": 0.6101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 7.5798796011446995e-06,
-      "loss": 0.8925,
+      "learning_rate": 6.9686278171670765e-06,
+      "loss": 0.8325,
       "step": 21500
     },
     {
-      "epoch": 0.5905083628573783,
+      "epoch": 0.6101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 7.5790165371310105e-06,
-      "loss": 0.8239,
+      "learning_rate": 6.96775199735943e-06,
+      "loss": 0.99,
       "step": 21501
     },
     {
-      "epoch": 0.5905358270852215,
+      "epoch": 0.6101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 7.578153492273083e-06,
-      "loss": 0.9044,
+      "learning_rate": 6.966876203165317e-06,
+      "loss": 0.814,
       "step": 21502
     },
     {
-      "epoch": 0.5905632913130647,
+      "epoch": 0.6101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 7.577290466577743e-06,
-      "loss": 0.7638,
+      "learning_rate": 6.966000434592142e-06,
+      "loss": 0.8884,
       "step": 21503
     },
     {
-      "epoch": 0.590590755540908,
+      "epoch": 0.6102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 7.5764274600518165e-06,
-      "loss": 0.8556,
+      "learning_rate": 6.965124691647296e-06,
+      "loss": 0.9712,
       "step": 21504
     },
     {
-      "epoch": 0.5906182197687512,
+      "epoch": 0.610244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 7.5755644727021385e-06,
-      "loss": 0.8418,
+      "learning_rate": 6.964248974338177e-06,
+      "loss": 0.8665,
       "step": 21505
     },
     {
-      "epoch": 0.5906456839965945,
+      "epoch": 0.610272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 7.574701504535534e-06,
-      "loss": 0.945,
+      "learning_rate": 6.963373282672185e-06,
+      "loss": 0.7423,
       "step": 21506
     },
     {
-      "epoch": 0.5906731482244376,
+      "epoch": 0.6103007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 7.573838555558828e-06,
-      "loss": 0.9699,
+      "learning_rate": 6.962497616656716e-06,
+      "loss": 0.8393,
       "step": 21507
     },
     {
-      "epoch": 0.5907006124522809,
+      "epoch": 0.6103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 7.57297562577885e-06,
-      "loss": 0.8134,
+      "learning_rate": 6.961621976299163e-06,
+      "loss": 0.7869,
       "step": 21508
     },
     {
-      "epoch": 0.5907280766801242,
+      "epoch": 0.6103575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 7.572112715202432e-06,
-      "loss": 0.8626,
+      "learning_rate": 6.96074636160693e-06,
+      "loss": 0.7604,
       "step": 21509
     },
     {
-      "epoch": 0.5907555409079673,
+      "epoch": 0.6103859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 7.571249823836396e-06,
-      "loss": 0.8169,
+      "learning_rate": 6.959870772587407e-06,
+      "loss": 0.8982,
       "step": 21510
     },
     {
-      "epoch": 0.5907830051358106,
+      "epoch": 0.6104143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 7.570386951687572e-06,
-      "loss": 0.8312,
+      "learning_rate": 6.9589952092479906e-06,
+      "loss": 0.8823,
       "step": 21511
     },
     {
-      "epoch": 0.5908104693636539,
+      "epoch": 0.6104426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 7.569524098762793e-06,
-      "loss": 0.9213,
+      "learning_rate": 6.958119671596081e-06,
+      "loss": 0.9207,
       "step": 21512
     },
     {
-      "epoch": 0.590837933591497,
+      "epoch": 0.6104710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 7.568661265068876e-06,
-      "loss": 0.9164,
+      "learning_rate": 6.957244159639072e-06,
+      "loss": 0.8079,
       "step": 21513
     },
     {
-      "epoch": 0.5908653978193403,
+      "epoch": 0.6104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 7.5677984506126546e-06,
-      "loss": 0.9114,
+      "learning_rate": 6.956368673384355e-06,
+      "loss": 0.9998,
       "step": 21514
     },
     {
-      "epoch": 0.5908928620471835,
+      "epoch": 0.610527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 7.566935655400952e-06,
-      "loss": 0.8582,
+      "learning_rate": 6.95549321283933e-06,
+      "loss": 0.8422,
       "step": 21515
     },
     {
-      "epoch": 0.5909203262750268,
+      "epoch": 0.6105561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 7.566072879440597e-06,
-      "loss": 0.9244,
+      "learning_rate": 6.954617778011392e-06,
+      "loss": 0.8752,
       "step": 21516
     },
     {
-      "epoch": 0.59094779050287,
+      "epoch": 0.6105845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 7.565210122738419e-06,
-      "loss": 0.9671,
+      "learning_rate": 6.95374236890793e-06,
+      "loss": 0.7756,
       "step": 21517
     },
     {
-      "epoch": 0.5909752547307132,
+      "epoch": 0.6106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 7.564347385301243e-06,
-      "loss": 0.8983,
+      "learning_rate": 6.952866985536347e-06,
+      "loss": 0.8356,
       "step": 21518
     },
     {
-      "epoch": 0.5910027189585565,
+      "epoch": 0.6106413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 7.5634846671358896e-06,
-      "loss": 0.9391,
+      "learning_rate": 6.951991627904032e-06,
+      "loss": 0.9219,
       "step": 21519
     },
     {
-      "epoch": 0.5910301831863997,
+      "epoch": 0.6106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 7.56262196824919e-06,
-      "loss": 0.7658,
+      "learning_rate": 6.951116296018383e-06,
+      "loss": 0.9769,
       "step": 21520
     },
     {
-      "epoch": 0.5910576474142429,
+      "epoch": 0.6106980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 7.561759288647971e-06,
-      "loss": 0.8286,
+      "learning_rate": 6.950240989886792e-06,
+      "loss": 0.8102,
       "step": 21521
     },
     {
-      "epoch": 0.5910851116420862,
+      "epoch": 0.6107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 7.560896628339055e-06,
-      "loss": 0.9361,
+      "learning_rate": 6.9493657095166485e-06,
+      "loss": 0.9641,
       "step": 21522
     },
     {
-      "epoch": 0.5911125758699294,
+      "epoch": 0.6107548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 7.560033987329269e-06,
-      "loss": 0.8256,
+      "learning_rate": 6.9484904549153546e-06,
+      "loss": 0.7446,
       "step": 21523
     },
     {
-      "epoch": 0.5911400400977727,
+      "epoch": 0.610783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 7.559171365625445e-06,
-      "loss": 0.8406,
+      "learning_rate": 6.947615226090297e-06,
+      "loss": 0.8967,
       "step": 21524
     },
     {
-      "epoch": 0.5911675043256159,
+      "epoch": 0.6108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 7.558308763234398e-06,
-      "loss": 0.8152,
+      "learning_rate": 6.94674002304887e-06,
+      "loss": 0.9534,
       "step": 21525
     },
     {
-      "epoch": 0.5911949685534591,
+      "epoch": 0.6108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 7.55744618016296e-06,
-      "loss": 0.7983,
+      "learning_rate": 6.945864845798469e-06,
+      "loss": 0.8827,
       "step": 21526
     },
     {
-      "epoch": 0.5912224327813024,
+      "epoch": 0.6108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 7.556583616417952e-06,
-      "loss": 0.9562,
+      "learning_rate": 6.9449896943464844e-06,
+      "loss": 0.8079,
       "step": 21527
     },
     {
-      "epoch": 0.5912498970091455,
+      "epoch": 0.6108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 7.5557210720062e-06,
-      "loss": 0.8568,
+      "learning_rate": 6.944114568700308e-06,
+      "loss": 0.7892,
       "step": 21528
     },
     {
-      "epoch": 0.5912773612369888,
+      "epoch": 0.6109250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 7.5548585469345345e-06,
-      "loss": 0.8763,
+      "learning_rate": 6.943239468867334e-06,
+      "loss": 0.7919,
       "step": 21529
     },
     {
-      "epoch": 0.5913048254648321,
+      "epoch": 0.6109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 7.553996041209773e-06,
-      "loss": 0.7453,
+      "learning_rate": 6.9423643948549544e-06,
+      "loss": 0.8544,
       "step": 21530
     },
     {
-      "epoch": 0.5913322896926753,
+      "epoch": 0.6109818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 7.553133554838741e-06,
-      "loss": 0.8549,
+      "learning_rate": 6.941489346670558e-06,
+      "loss": 0.8255,
       "step": 21531
     },
     {
-      "epoch": 0.5913597539205185,
+      "epoch": 0.6110102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 7.552271087828263e-06,
-      "loss": 0.7931,
+      "learning_rate": 6.940614324321537e-06,
+      "loss": 0.8627,
       "step": 21532
     },
     {
-      "epoch": 0.5913872181483617,
+      "epoch": 0.611038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 7.5514086401851676e-06,
-      "loss": 0.8348,
+      "learning_rate": 6.939739327815288e-06,
+      "loss": 0.8802,
       "step": 21533
     },
     {
-      "epoch": 0.591414682376205,
+      "epoch": 0.611066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 7.550546211916272e-06,
-      "loss": 0.7502,
+      "learning_rate": 6.938864357159198e-06,
+      "loss": 0.9267,
       "step": 21534
     },
     {
-      "epoch": 0.5914421466040483,
+      "epoch": 0.6110953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 7.549683803028409e-06,
-      "loss": 1.0038,
+      "learning_rate": 6.937989412360657e-06,
+      "loss": 0.8172,
       "step": 21535
     },
     {
-      "epoch": 0.5914696108318914,
+      "epoch": 0.6111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 7.548821413528392e-06,
-      "loss": 0.8432,
+      "learning_rate": 6.93711449342706e-06,
+      "loss": 0.7756,
       "step": 21536
     },
     {
-      "epoch": 0.5914970750597347,
+      "epoch": 0.6111520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 7.54795904342305e-06,
-      "loss": 0.8274,
+      "learning_rate": 6.936239600365793e-06,
+      "loss": 0.7947,
       "step": 21537
     },
     {
-      "epoch": 0.591524539287578,
+      "epoch": 0.6111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 7.547096692719207e-06,
-      "loss": 0.8382,
+      "learning_rate": 6.935364733184247e-06,
+      "loss": 0.7931,
       "step": 21538
     },
     {
-      "epoch": 0.5915520035154211,
+      "epoch": 0.6112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 7.5462343614236845e-06,
-      "loss": 0.8222,
+      "learning_rate": 6.934489891889813e-06,
+      "loss": 0.8264,
       "step": 21539
     },
     {
-      "epoch": 0.5915794677432644,
+      "epoch": 0.6112372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 7.545372049543305e-06,
-      "loss": 0.8115,
+      "learning_rate": 6.933615076489882e-06,
+      "loss": 0.8849,
       "step": 21540
     },
     {
-      "epoch": 0.5916069319711076,
+      "epoch": 0.6112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 7.544509757084898e-06,
-      "loss": 0.8568,
+      "learning_rate": 6.932740286991843e-06,
+      "loss": 0.7395,
       "step": 21541
     },
     {
-      "epoch": 0.5916343961989509,
+      "epoch": 0.6112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 7.543647484055277e-06,
-      "loss": 0.8289,
+      "learning_rate": 6.931865523403082e-06,
+      "loss": 0.885,
       "step": 21542
     },
     {
-      "epoch": 0.5916618604267941,
+      "epoch": 0.6113223609534619,
       "grad_norm": 0.0,
-      "learning_rate": 7.54278523046127e-06,
-      "loss": 0.9802,
+      "learning_rate": 6.930990785730992e-06,
+      "loss": 0.8394,
       "step": 21543
     },
     {
-      "epoch": 0.5916893246546373,
+      "epoch": 0.6113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 7.5419229963096945e-06,
-      "loss": 0.8484,
+      "learning_rate": 6.930116073982962e-06,
+      "loss": 0.8496,
       "step": 21544
     },
     {
-      "epoch": 0.5917167888824806,
+      "epoch": 0.6113791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 7.54106078160738e-06,
-      "loss": 0.9372,
+      "learning_rate": 6.929241388166377e-06,
+      "loss": 0.9869,
       "step": 21545
     },
     {
-      "epoch": 0.5917442531103237,
+      "epoch": 0.6114074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 7.540198586361143e-06,
-      "loss": 0.7878,
+      "learning_rate": 6.92836672828863e-06,
+      "loss": 0.833,
       "step": 21546
     },
     {
-      "epoch": 0.591771717338167,
+      "epoch": 0.6114358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 7.539336410577811e-06,
-      "loss": 0.7527,
+      "learning_rate": 6.927492094357108e-06,
+      "loss": 0.9192,
       "step": 21547
     },
     {
-      "epoch": 0.5917991815660103,
+      "epoch": 0.6114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 7.538474254264199e-06,
-      "loss": 0.7976,
+      "learning_rate": 6.926617486379194e-06,
+      "loss": 0.7984,
       "step": 21548
     },
     {
-      "epoch": 0.5918266457938535,
+      "epoch": 0.6114926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 7.537612117427132e-06,
-      "loss": 0.9505,
+      "learning_rate": 6.925742904362281e-06,
+      "loss": 0.8514,
       "step": 21549
     },
     {
-      "epoch": 0.5918541100216967,
+      "epoch": 0.6115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 7.536750000073434e-06,
-      "loss": 0.8989,
+      "learning_rate": 6.924868348313759e-06,
+      "loss": 0.8217,
       "step": 21550
     },
     {
-      "epoch": 0.59188157424954,
+      "epoch": 0.6115493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 7.5358879022099215e-06,
-      "loss": 0.847,
+      "learning_rate": 6.9239938182410126e-06,
+      "loss": 0.9314,
       "step": 21551
     },
     {
-      "epoch": 0.5919090384773832,
+      "epoch": 0.611577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 7.53502582384342e-06,
-      "loss": 0.7327,
+      "learning_rate": 6.923119314151426e-06,
+      "loss": 0.8603,
       "step": 21552
     },
     {
-      "epoch": 0.5919365027052265,
+      "epoch": 0.6116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 7.5341637649807516e-06,
-      "loss": 0.8345,
+      "learning_rate": 6.922244836052392e-06,
+      "loss": 0.9512,
       "step": 21553
     },
     {
-      "epoch": 0.5919639669330696,
+      "epoch": 0.6116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 7.5333017256287345e-06,
-      "loss": 0.8203,
+      "learning_rate": 6.9213703839512935e-06,
+      "loss": 0.7126,
       "step": 21554
     },
     {
-      "epoch": 0.5919914311609129,
+      "epoch": 0.6116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 7.532439705794185e-06,
-      "loss": 0.9042,
+      "learning_rate": 6.920495957855516e-06,
+      "loss": 0.8371,
       "step": 21555
     },
     {
-      "epoch": 0.5920188953887562,
+      "epoch": 0.6116912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 7.53157770548393e-06,
-      "loss": 0.7477,
+      "learning_rate": 6.919621557772449e-06,
+      "loss": 1.0488,
       "step": 21556
     },
     {
-      "epoch": 0.5920463596165993,
+      "epoch": 0.6117196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 7.53071572470479e-06,
-      "loss": 0.9894,
+      "learning_rate": 6.9187471837094774e-06,
+      "loss": 0.8683,
       "step": 21557
     },
     {
-      "epoch": 0.5920738238444426,
+      "epoch": 0.6117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 7.529853763463582e-06,
-      "loss": 0.822,
+      "learning_rate": 6.917872835673984e-06,
+      "loss": 0.9533,
       "step": 21558
     },
     {
-      "epoch": 0.5921012880722859,
+      "epoch": 0.6117763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 7.5289918217671325e-06,
-      "loss": 0.9204,
+      "learning_rate": 6.916998513673361e-06,
+      "loss": 0.8144,
       "step": 21559
     },
     {
-      "epoch": 0.5921287523001291,
+      "epoch": 0.6118047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 7.528129899622252e-06,
-      "loss": 0.8363,
+      "learning_rate": 6.916124217714989e-06,
+      "loss": 0.7463,
       "step": 21560
     },
     {
-      "epoch": 0.5921562165279723,
+      "epoch": 0.6118331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 7.527267997035764e-06,
-      "loss": 0.8497,
+      "learning_rate": 6.915249947806253e-06,
+      "loss": 0.8373,
       "step": 21561
     },
     {
-      "epoch": 0.5921836807558155,
+      "epoch": 0.6118615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 7.526406114014493e-06,
-      "loss": 0.8294,
+      "learning_rate": 6.914375703954541e-06,
+      "loss": 0.8905,
       "step": 21562
     },
     {
-      "epoch": 0.5922111449836588,
+      "epoch": 0.6118898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 7.5255442505652535e-06,
-      "loss": 0.9096,
+      "learning_rate": 6.913501486167231e-06,
+      "loss": 0.868,
       "step": 21563
     },
     {
-      "epoch": 0.5922386092115021,
+      "epoch": 0.6119182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 7.524682406694865e-06,
-      "loss": 0.8734,
+      "learning_rate": 6.912627294451719e-06,
+      "loss": 0.8889,
       "step": 21564
     },
     {
-      "epoch": 0.5922660734393452,
+      "epoch": 0.6119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 7.523820582410153e-06,
-      "loss": 0.9491,
+      "learning_rate": 6.911753128815382e-06,
+      "loss": 0.834,
       "step": 21565
     },
     {
-      "epoch": 0.5922935376671885,
+      "epoch": 0.6119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 7.52295877771793e-06,
-      "loss": 0.7925,
+      "learning_rate": 6.910878989265603e-06,
+      "loss": 0.8778,
       "step": 21566
     },
     {
-      "epoch": 0.5923210018950317,
+      "epoch": 0.6120034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 7.522096992625014e-06,
-      "loss": 0.8971,
+      "learning_rate": 6.91000487580977e-06,
+      "loss": 0.9139,
       "step": 21567
     },
     {
-      "epoch": 0.592348466122875,
+      "epoch": 0.6120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 7.521235227138226e-06,
-      "loss": 0.9165,
+      "learning_rate": 6.909130788455265e-06,
+      "loss": 0.7812,
       "step": 21568
     },
     {
-      "epoch": 0.5923759303507182,
+      "epoch": 0.6120601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 7.520373481264388e-06,
-      "loss": 0.8268,
+      "learning_rate": 6.908256727209469e-06,
+      "loss": 0.828,
       "step": 21569
     },
     {
-      "epoch": 0.5924033945785614,
+      "epoch": 0.612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 7.519511755010314e-06,
-      "loss": 0.98,
+      "learning_rate": 6.907382692079769e-06,
+      "loss": 0.7914,
       "step": 21570
     },
     {
-      "epoch": 0.5924308588064047,
+      "epoch": 0.612116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 7.5186500483828265e-06,
-      "loss": 0.796,
+      "learning_rate": 6.906508683073547e-06,
+      "loss": 0.9219,
       "step": 21571
     },
     {
-      "epoch": 0.592458323034248,
+      "epoch": 0.6121452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 7.517788361388737e-06,
-      "loss": 0.8503,
+      "learning_rate": 6.9056347001981825e-06,
+      "loss": 0.8656,
       "step": 21572
     },
     {
-      "epoch": 0.5924857872620911,
+      "epoch": 0.6121736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 7.5169266940348686e-06,
-      "loss": 0.909,
+      "learning_rate": 6.9047607434610645e-06,
+      "loss": 0.8865,
       "step": 21573
     },
     {
-      "epoch": 0.5925132514899344,
+      "epoch": 0.6122020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 7.516065046328039e-06,
-      "loss": 0.888,
+      "learning_rate": 6.90388681286957e-06,
+      "loss": 0.8133,
       "step": 21574
     },
     {
-      "epoch": 0.5925407157177776,
+      "epoch": 0.6122304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 7.515203418275063e-06,
-      "loss": 0.9298,
+      "learning_rate": 6.903012908431082e-06,
+      "loss": 0.8858,
       "step": 21575
     },
     {
-      "epoch": 0.5925681799456208,
+      "epoch": 0.6122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 7.514341809882761e-06,
-      "loss": 0.8601,
+      "learning_rate": 6.902139030152983e-06,
+      "loss": 0.8684,
       "step": 21576
     },
     {
-      "epoch": 0.5925956441734641,
+      "epoch": 0.6122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 7.5134802211579525e-06,
-      "loss": 0.9503,
+      "learning_rate": 6.901265178042657e-06,
+      "loss": 1.0041,
       "step": 21577
     },
     {
-      "epoch": 0.5926231084013073,
+      "epoch": 0.6123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 7.5126186521074506e-06,
-      "loss": 0.8701,
+      "learning_rate": 6.900391352107478e-06,
+      "loss": 0.8051,
       "step": 21578
     },
     {
-      "epoch": 0.5926505726291506,
+      "epoch": 0.6123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 7.511757102738071e-06,
-      "loss": 0.9031,
+      "learning_rate": 6.899517552354836e-06,
+      "loss": 0.8818,
       "step": 21579
     },
     {
-      "epoch": 0.5926780368569937,
+      "epoch": 0.612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 7.510895573056634e-06,
-      "loss": 0.7894,
+      "learning_rate": 6.89864377879211e-06,
+      "loss": 0.7497,
       "step": 21580
     },
     {
-      "epoch": 0.592705501084837,
+      "epoch": 0.6124006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 7.510034063069955e-06,
-      "loss": 0.9397,
+      "learning_rate": 6.897770031426679e-06,
+      "loss": 0.7986,
       "step": 21581
     },
     {
-      "epoch": 0.5927329653126803,
+      "epoch": 0.6124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 7.509172572784851e-06,
-      "loss": 0.9303,
+      "learning_rate": 6.896896310265925e-06,
+      "loss": 0.899,
       "step": 21582
     },
     {
-      "epoch": 0.5927604295405234,
+      "epoch": 0.6124574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 7.508311102208143e-06,
-      "loss": 0.9263,
+      "learning_rate": 6.896022615317224e-06,
+      "loss": 0.8756,
       "step": 21583
     },
     {
-      "epoch": 0.5927878937683667,
+      "epoch": 0.6124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 7.5074496513466365e-06,
-      "loss": 0.8676,
+      "learning_rate": 6.895148946587962e-06,
+      "loss": 0.7413,
       "step": 21584
     },
     {
-      "epoch": 0.59281535799621,
+      "epoch": 0.6125141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 7.506588220207155e-06,
-      "loss": 0.9461,
+      "learning_rate": 6.894275304085517e-06,
+      "loss": 0.8698,
       "step": 21585
     },
     {
-      "epoch": 0.5928428222240532,
+      "epoch": 0.6125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 7.5057268087965166e-06,
-      "loss": 0.9414,
+      "learning_rate": 6.893401687817266e-06,
+      "loss": 0.8634,
       "step": 21586
     },
     {
-      "epoch": 0.5928702864518964,
+      "epoch": 0.6125709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 7.50486541712153e-06,
-      "loss": 1.002,
+      "learning_rate": 6.892528097790592e-06,
+      "loss": 0.8616,
       "step": 21587
     },
     {
-      "epoch": 0.5928977506797396,
+      "epoch": 0.6125993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 7.5040040451890194e-06,
-      "loss": 0.8409,
+      "learning_rate": 6.891654534012872e-06,
+      "loss": 0.7838,
       "step": 21588
     },
     {
-      "epoch": 0.5929252149075829,
+      "epoch": 0.612627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 7.503142693005792e-06,
-      "loss": 0.8465,
+      "learning_rate": 6.890780996491486e-06,
+      "loss": 0.8192,
       "step": 21589
     },
     {
-      "epoch": 0.5929526791354262,
+      "epoch": 0.6126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 7.502281360578668e-06,
-      "loss": 0.8542,
+      "learning_rate": 6.889907485233813e-06,
+      "loss": 0.8657,
       "step": 21590
     },
     {
-      "epoch": 0.5929801433632693,
+      "epoch": 0.6126844494892169,
       "grad_norm": 0.0,
-      "learning_rate": 7.50142004791446e-06,
-      "loss": 0.9059,
+      "learning_rate": 6.889034000247232e-06,
+      "loss": 0.8092,
       "step": 21591
     },
     {
-      "epoch": 0.5930076075911126,
+      "epoch": 0.6127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 7.500558755019984e-06,
-      "loss": 0.8442,
+      "learning_rate": 6.88816054153912e-06,
+      "loss": 0.7754,
       "step": 21592
     },
     {
-      "epoch": 0.5930350718189558,
+      "epoch": 0.6127412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 7.499697481902057e-06,
-      "loss": 0.8236,
+      "learning_rate": 6.887287109116856e-06,
+      "loss": 0.8979,
       "step": 21593
     },
     {
-      "epoch": 0.593062536046799,
+      "epoch": 0.6127695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 7.498836228567496e-06,
-      "loss": 0.8749,
+      "learning_rate": 6.886413702987814e-06,
+      "loss": 0.8468,
       "step": 21594
     },
     {
-      "epoch": 0.5930900002746423,
+      "epoch": 0.6127979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 7.4979749950231054e-06,
-      "loss": 0.84,
+      "learning_rate": 6.88554032315938e-06,
+      "loss": 0.9541,
       "step": 21595
     },
     {
-      "epoch": 0.5931174645024855,
+      "epoch": 0.6128263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 7.497113781275706e-06,
-      "loss": 0.8836,
+      "learning_rate": 6.884666969638925e-06,
+      "loss": 0.8035,
       "step": 21596
     },
     {
-      "epoch": 0.5931449287303288,
+      "epoch": 0.6128547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 7.496252587332114e-06,
-      "loss": 0.843,
+      "learning_rate": 6.88379364243383e-06,
+      "loss": 0.7909,
       "step": 21597
     },
     {
-      "epoch": 0.593172392958172,
+      "epoch": 0.6128830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 7.495391413199142e-06,
-      "loss": 0.831,
+      "learning_rate": 6.882920341551469e-06,
+      "loss": 0.8599,
       "step": 21598
     },
     {
-      "epoch": 0.5931998571860152,
+      "epoch": 0.612911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 7.494530258883601e-06,
-      "loss": 0.9205,
+      "learning_rate": 6.882047066999218e-06,
+      "loss": 0.9262,
       "step": 21599
     },
     {
-      "epoch": 0.5932273214138585,
+      "epoch": 0.6129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 7.493669124392311e-06,
-      "loss": 0.9108,
+      "learning_rate": 6.881173818784459e-06,
+      "loss": 0.9066,
       "step": 21600
     },
     {
-      "epoch": 0.5932547856417016,
+      "epoch": 0.6129682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 7.492808009732079e-06,
-      "loss": 0.9317,
+      "learning_rate": 6.880300596914564e-06,
+      "loss": 0.8157,
       "step": 21601
     },
     {
-      "epoch": 0.5932822498695449,
+      "epoch": 0.6129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 7.491946914909723e-06,
-      "loss": 0.7736,
+      "learning_rate": 6.879427401396909e-06,
+      "loss": 0.9379,
       "step": 21602
     },
     {
-      "epoch": 0.5933097140973882,
+      "epoch": 0.6130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 7.491085839932051e-06,
-      "loss": 0.8746,
+      "learning_rate": 6.878554232238872e-06,
+      "loss": 0.7927,
       "step": 21603
     },
     {
-      "epoch": 0.5933371783252314,
+      "epoch": 0.6130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 7.490224784805881e-06,
-      "loss": 0.8778,
+      "learning_rate": 6.877681089447828e-06,
+      "loss": 0.8109,
       "step": 21604
     },
     {
-      "epoch": 0.5933646425530746,
+      "epoch": 0.6130817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 7.4893637495380255e-06,
-      "loss": 0.876,
+      "learning_rate": 6.87680797303115e-06,
+      "loss": 0.8605,
       "step": 21605
     },
     {
-      "epoch": 0.5933921067809178,
+      "epoch": 0.6131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 7.4885027341353e-06,
-      "loss": 0.9247,
+      "learning_rate": 6.875934882996218e-06,
+      "loss": 0.861,
       "step": 21606
     },
     {
-      "epoch": 0.5934195710087611,
+      "epoch": 0.613138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 7.487641738604509e-06,
-      "loss": 0.7653,
+      "learning_rate": 6.875061819350405e-06,
+      "loss": 0.8629,
       "step": 21607
     },
     {
-      "epoch": 0.5934470352366044,
+      "epoch": 0.61316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 7.48678076295247e-06,
-      "loss": 0.7871,
+      "learning_rate": 6.8741887821010844e-06,
+      "loss": 0.9906,
       "step": 21608
     },
     {
-      "epoch": 0.5934744994644475,
+      "epoch": 0.6131952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 7.4859198071859976e-06,
-      "loss": 0.8651,
+      "learning_rate": 6.873315771255627e-06,
+      "loss": 0.7856,
       "step": 21609
     },
     {
-      "epoch": 0.5935019636922908,
+      "epoch": 0.6132236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 7.485058871311899e-06,
-      "loss": 0.8882,
+      "learning_rate": 6.872442786821415e-06,
+      "loss": 0.9232,
       "step": 21610
     },
     {
-      "epoch": 0.5935294279201341,
+      "epoch": 0.6132519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 7.484197955336988e-06,
-      "loss": 0.8619,
+      "learning_rate": 6.871569828805822e-06,
+      "loss": 0.8402,
       "step": 21611
     },
     {
-      "epoch": 0.5935568921479772,
+      "epoch": 0.6132803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 7.483337059268083e-06,
-      "loss": 0.865,
+      "learning_rate": 6.870696897216218e-06,
+      "loss": 0.9071,
       "step": 21612
     },
     {
-      "epoch": 0.5935843563758205,
+      "epoch": 0.6133087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 7.482476183111984e-06,
-      "loss": 0.8377,
+      "learning_rate": 6.869823992059976e-06,
+      "loss": 0.8017,
       "step": 21613
     },
     {
-      "epoch": 0.5936118206036637,
+      "epoch": 0.6133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 7.481615326875512e-06,
-      "loss": 0.862,
+      "learning_rate": 6.868951113344473e-06,
+      "loss": 0.783,
       "step": 21614
     },
     {
-      "epoch": 0.593639284831507,
+      "epoch": 0.6133654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 7.480754490565473e-06,
-      "loss": 0.8884,
+      "learning_rate": 6.868078261077083e-06,
+      "loss": 0.8225,
       "step": 21615
     },
     {
-      "epoch": 0.5936667490593502,
+      "epoch": 0.6133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 7.479893674188681e-06,
-      "loss": 0.8024,
+      "learning_rate": 6.867205435265172e-06,
+      "loss": 0.8951,
       "step": 21616
     },
     {
-      "epoch": 0.5936942132871934,
+      "epoch": 0.613422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 7.479032877751946e-06,
-      "loss": 0.8582,
+      "learning_rate": 6.866332635916122e-06,
+      "loss": 0.8277,
       "step": 21617
     },
     {
-      "epoch": 0.5937216775150367,
+      "epoch": 0.6134506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 7.478172101262085e-06,
-      "loss": 0.8044,
+      "learning_rate": 6.8654598630372995e-06,
+      "loss": 0.819,
       "step": 21618
     },
     {
-      "epoch": 0.5937491417428798,
+      "epoch": 0.6134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 7.477311344725896e-06,
-      "loss": 0.7898,
+      "learning_rate": 6.864587116636077e-06,
+      "loss": 0.8503,
       "step": 21619
     },
     {
-      "epoch": 0.5937766059707231,
+      "epoch": 0.6135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 7.476450608150198e-06,
-      "loss": 0.8818,
+      "learning_rate": 6.863714396719829e-06,
+      "loss": 0.8385,
       "step": 21620
     },
     {
-      "epoch": 0.5938040701985664,
+      "epoch": 0.6135357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 7.475589891541802e-06,
-      "loss": 0.7678,
+      "learning_rate": 6.862841703295928e-06,
+      "loss": 0.9062,
       "step": 21621
     },
     {
-      "epoch": 0.5938315344264096,
+      "epoch": 0.6135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 7.474729194907515e-06,
-      "loss": 0.8531,
+      "learning_rate": 6.86196903637174e-06,
+      "loss": 0.806,
       "step": 21622
     },
     {
-      "epoch": 0.5938589986542528,
+      "epoch": 0.6135925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 7.473868518254149e-06,
-      "loss": 0.8849,
+      "learning_rate": 6.861096395954645e-06,
+      "loss": 0.7801,
       "step": 21623
     },
     {
-      "epoch": 0.5938864628820961,
+      "epoch": 0.6136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 7.473007861588517e-06,
-      "loss": 0.8688,
+      "learning_rate": 6.8602237820520045e-06,
+      "loss": 0.8999,
       "step": 21624
     },
     {
-      "epoch": 0.5939139271099393,
+      "epoch": 0.6136492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 7.472147224917423e-06,
-      "loss": 0.8953,
+      "learning_rate": 6.8593511946712e-06,
+      "loss": 0.8933,
       "step": 21625
     },
     {
-      "epoch": 0.5939413913377826,
+      "epoch": 0.613677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 7.4712866082476805e-06,
-      "loss": 0.8807,
+      "learning_rate": 6.858478633819596e-06,
+      "loss": 0.9169,
       "step": 21626
     },
     {
-      "epoch": 0.5939688555656257,
+      "epoch": 0.6137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 7.470426011586096e-06,
-      "loss": 0.9172,
+      "learning_rate": 6.857606099504563e-06,
+      "loss": 0.8533,
       "step": 21627
     },
     {
-      "epoch": 0.593996319793469,
+      "epoch": 0.6137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 7.469565434939481e-06,
-      "loss": 0.8842,
+      "learning_rate": 6.856733591733473e-06,
+      "loss": 0.8474,
       "step": 21628
     },
     {
-      "epoch": 0.5940237840213123,
+      "epoch": 0.6137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 7.468704878314645e-06,
-      "loss": 0.9143,
+      "learning_rate": 6.855861110513697e-06,
+      "loss": 0.7967,
       "step": 21629
     },
     {
-      "epoch": 0.5940512482491555,
+      "epoch": 0.6137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 7.467844341718401e-06,
-      "loss": 0.8491,
+      "learning_rate": 6.854988655852601e-06,
+      "loss": 0.9407,
       "step": 21630
     },
     {
-      "epoch": 0.5940787124769987,
+      "epoch": 0.6138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 7.466983825157548e-06,
-      "loss": 0.8321,
+      "learning_rate": 6.8541162277575615e-06,
+      "loss": 0.869,
       "step": 21631
     },
     {
-      "epoch": 0.5941061767048419,
+      "epoch": 0.6138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 7.466123328638901e-06,
-      "loss": 0.8179,
+      "learning_rate": 6.8532438262359404e-06,
+      "loss": 0.9136,
       "step": 21632
     },
     {
-      "epoch": 0.5941336409326852,
+      "epoch": 0.6138762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 7.465262852169268e-06,
-      "loss": 0.8551,
+      "learning_rate": 6.85237145129511e-06,
+      "loss": 0.8147,
       "step": 21633
     },
     {
-      "epoch": 0.5941611051605284,
+      "epoch": 0.6139046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 7.464402395755456e-06,
-      "loss": 0.8136,
+      "learning_rate": 6.85149910294244e-06,
+      "loss": 0.9608,
       "step": 21634
     },
     {
-      "epoch": 0.5941885693883716,
+      "epoch": 0.613933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 7.463541959404275e-06,
-      "loss": 0.865,
+      "learning_rate": 6.8506267811853e-06,
+      "loss": 0.8505,
       "step": 21635
     },
     {
-      "epoch": 0.5942160336162149,
+      "epoch": 0.613961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 7.462681543122537e-06,
-      "loss": 0.801,
+      "learning_rate": 6.849754486031055e-06,
+      "loss": 0.9139,
       "step": 21636
     },
     {
-      "epoch": 0.5942434978440582,
+      "epoch": 0.6139897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 7.461821146917041e-06,
-      "loss": 0.836,
+      "learning_rate": 6.848882217487077e-06,
+      "loss": 0.7708,
       "step": 21637
     },
     {
-      "epoch": 0.5942709620719013,
+      "epoch": 0.6140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 7.460960770794601e-06,
-      "loss": 0.7894,
+      "learning_rate": 6.848009975560732e-06,
+      "loss": 0.7945,
       "step": 21638
     },
     {
-      "epoch": 0.5942984262997446,
+      "epoch": 0.6140465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 7.460100414762021e-06,
-      "loss": 0.8012,
+      "learning_rate": 6.847137760259388e-06,
+      "loss": 0.9077,
       "step": 21639
     },
     {
-      "epoch": 0.5943258905275878,
+      "epoch": 0.6140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 7.459240078826111e-06,
-      "loss": 0.8949,
+      "learning_rate": 6.846265571590409e-06,
+      "loss": 0.9622,
       "step": 21640
     },
     {
-      "epoch": 0.594353354755431,
+      "epoch": 0.6141032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 7.458379762993679e-06,
-      "loss": 0.8616,
+      "learning_rate": 6.845393409561172e-06,
+      "loss": 0.933,
       "step": 21641
     },
     {
-      "epoch": 0.5943808189832743,
+      "epoch": 0.6141316685584562,
       "grad_norm": 0.0,
-      "learning_rate": 7.457519467271532e-06,
-      "loss": 0.8663,
+      "learning_rate": 6.844521274179039e-06,
+      "loss": 0.7786,
       "step": 21642
     },
     {
-      "epoch": 0.5944082832111175,
+      "epoch": 0.6141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 7.4566591916664735e-06,
-      "loss": 0.8967,
+      "learning_rate": 6.843649165451374e-06,
+      "loss": 0.8407,
       "step": 21643
     },
     {
-      "epoch": 0.5944357474389608,
+      "epoch": 0.6141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 7.455798936185312e-06,
-      "loss": 0.8909,
+      "learning_rate": 6.842777083385548e-06,
+      "loss": 0.8903,
       "step": 21644
     },
     {
-      "epoch": 0.5944632116668039,
+      "epoch": 0.6142167990919409,
       "grad_norm": 0.0,
-      "learning_rate": 7.454938700834856e-06,
-      "loss": 0.9391,
+      "learning_rate": 6.8419050279889264e-06,
+      "loss": 0.876,
       "step": 21645
     },
     {
-      "epoch": 0.5944906758946472,
+      "epoch": 0.6142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 7.4540784856219095e-06,
-      "loss": 0.8585,
+      "learning_rate": 6.841032999268873e-06,
+      "loss": 0.8382,
       "step": 21646
     },
     {
-      "epoch": 0.5945181401224905,
+      "epoch": 0.6142735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 7.4532182905532854e-06,
-      "loss": 0.8457,
+      "learning_rate": 6.840160997232758e-06,
+      "loss": 0.8696,
       "step": 21647
     },
     {
-      "epoch": 0.5945456043503337,
+      "epoch": 0.6143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 7.452358115635779e-06,
-      "loss": 0.9231,
+      "learning_rate": 6.839289021887945e-06,
+      "loss": 0.899,
       "step": 21648
     },
     {
-      "epoch": 0.5945730685781769,
+      "epoch": 0.6143303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 7.451497960876203e-06,
-      "loss": 0.9063,
+      "learning_rate": 6.838417073241798e-06,
+      "loss": 0.9136,
       "step": 21649
     },
     {
-      "epoch": 0.5946005328060202,
+      "epoch": 0.6143586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 7.450637826281364e-06,
-      "loss": 0.8855,
+      "learning_rate": 6.837545151301685e-06,
+      "loss": 0.8674,
       "step": 21650
     },
     {
-      "epoch": 0.5946279970338634,
+      "epoch": 0.6143870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 7.449777711858064e-06,
-      "loss": 0.7113,
+      "learning_rate": 6.8366732560749706e-06,
+      "loss": 0.7972,
       "step": 21651
     },
     {
-      "epoch": 0.5946554612617067,
+      "epoch": 0.6144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 7.44891761761311e-06,
-      "loss": 0.8214,
+      "learning_rate": 6.835801387569021e-06,
+      "loss": 0.914,
       "step": 21652
     },
     {
-      "epoch": 0.5946829254895498,
+      "epoch": 0.6144438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 7.448057543553313e-06,
-      "loss": 0.9732,
+      "learning_rate": 6.834929545791196e-06,
+      "loss": 0.8584,
       "step": 21653
     },
     {
-      "epoch": 0.5947103897173931,
+      "epoch": 0.614472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 7.447197489685472e-06,
-      "loss": 0.8608,
+      "learning_rate": 6.8340577307488645e-06,
+      "loss": 0.8889,
       "step": 21654
     },
     {
-      "epoch": 0.5947378539452364,
+      "epoch": 0.6145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 7.446337456016391e-06,
-      "loss": 0.8247,
+      "learning_rate": 6.8331859424493875e-06,
+      "loss": 0.8499,
       "step": 21655
     },
     {
-      "epoch": 0.5947653181730795,
+      "epoch": 0.6145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 7.445477442552876e-06,
-      "loss": 0.7974,
+      "learning_rate": 6.8323141809001334e-06,
+      "loss": 0.9184,
       "step": 21656
     },
     {
-      "epoch": 0.5947927824009228,
+      "epoch": 0.6145573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 7.444617449301734e-06,
-      "loss": 0.8276,
+      "learning_rate": 6.831442446108463e-06,
+      "loss": 0.9015,
       "step": 21657
     },
     {
-      "epoch": 0.594820246628766,
+      "epoch": 0.6145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 7.443757476269768e-06,
-      "loss": 0.877,
+      "learning_rate": 6.830570738081744e-06,
+      "loss": 0.9643,
       "step": 21658
     },
     {
-      "epoch": 0.5948477108566093,
+      "epoch": 0.6146140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 7.442897523463788e-06,
-      "loss": 0.8557,
+      "learning_rate": 6.8296990568273335e-06,
+      "loss": 0.8445,
       "step": 21659
     },
     {
-      "epoch": 0.5948751750844525,
+      "epoch": 0.6146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 7.4420375908905885e-06,
-      "loss": 0.895,
+      "learning_rate": 6.828827402352597e-06,
+      "loss": 0.7532,
       "step": 21660
     },
     {
-      "epoch": 0.5949026393122957,
+      "epoch": 0.6146708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 7.441177678556979e-06,
-      "loss": 0.8194,
+      "learning_rate": 6.827955774664901e-06,
+      "loss": 0.8854,
       "step": 21661
     },
     {
-      "epoch": 0.594930103540139,
+      "epoch": 0.6146992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 7.440317786469764e-06,
-      "loss": 0.8747,
+      "learning_rate": 6.827084173771603e-06,
+      "loss": 0.8908,
       "step": 21662
     },
     {
-      "epoch": 0.5949575677679823,
+      "epoch": 0.6147275822928491,
       "grad_norm": 0.0,
-      "learning_rate": 7.439457914635744e-06,
-      "loss": 0.8346,
+      "learning_rate": 6.826212599680066e-06,
+      "loss": 0.8222,
       "step": 21663
     },
     {
-      "epoch": 0.5949850319958254,
+      "epoch": 0.6147559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 7.438598063061725e-06,
-      "loss": 0.8262,
+      "learning_rate": 6.825341052397657e-06,
+      "loss": 0.8582,
       "step": 21664
     },
     {
-      "epoch": 0.5950124962236687,
+      "epoch": 0.6147843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 7.437738231754514e-06,
-      "loss": 0.8405,
+      "learning_rate": 6.824469531931735e-06,
+      "loss": 0.7489,
       "step": 21665
     },
     {
-      "epoch": 0.5950399604515119,
+      "epoch": 0.6148127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 7.43687842072091e-06,
-      "loss": 0.8446,
+      "learning_rate": 6.823598038289658e-06,
+      "loss": 0.8732,
       "step": 21666
     },
     {
-      "epoch": 0.5950674246793551,
+      "epoch": 0.6148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 7.436018629967714e-06,
-      "loss": 0.8569,
+      "learning_rate": 6.8227265714787945e-06,
+      "loss": 0.8395,
       "step": 21667
     },
     {
-      "epoch": 0.5950948889071984,
+      "epoch": 0.6148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 7.435158859501732e-06,
-      "loss": 0.8905,
+      "learning_rate": 6.821855131506503e-06,
+      "loss": 0.8886,
       "step": 21668
     },
     {
-      "epoch": 0.5951223531350416,
+      "epoch": 0.6148978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 7.434299109329769e-06,
-      "loss": 0.8749,
+      "learning_rate": 6.820983718380141e-06,
+      "loss": 0.8226,
       "step": 21669
     },
     {
-      "epoch": 0.5951498173628849,
+      "epoch": 0.6149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 7.433439379458622e-06,
-      "loss": 0.8854,
+      "learning_rate": 6.820112332107073e-06,
+      "loss": 0.832,
       "step": 21670
     },
     {
-      "epoch": 0.595177281590728,
+      "epoch": 0.6149545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 7.4325796698951025e-06,
-      "loss": 0.9041,
+      "learning_rate": 6.819240972694658e-06,
+      "loss": 0.927,
       "step": 21671
     },
     {
-      "epoch": 0.5952047458185713,
+      "epoch": 0.614982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 7.431719980646002e-06,
-      "loss": 0.7613,
+      "learning_rate": 6.818369640150261e-06,
+      "loss": 0.9421,
       "step": 21672
     },
     {
-      "epoch": 0.5952322100464146,
+      "epoch": 0.615011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 7.430860311718129e-06,
-      "loss": 0.8812,
+      "learning_rate": 6.817498334481239e-06,
+      "loss": 0.7523,
       "step": 21673
     },
     {
-      "epoch": 0.5952596742742577,
+      "epoch": 0.6150397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 7.430000663118284e-06,
-      "loss": 0.8798,
+      "learning_rate": 6.816627055694946e-06,
+      "loss": 0.9383,
       "step": 21674
     },
     {
-      "epoch": 0.595287138502101,
+      "epoch": 0.6150681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 7.429141034853269e-06,
-      "loss": 0.7902,
+      "learning_rate": 6.815755803798754e-06,
+      "loss": 0.7988,
       "step": 21675
     },
     {
-      "epoch": 0.5953146027299443,
+      "epoch": 0.6150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 7.428281426929886e-06,
-      "loss": 0.8398,
+      "learning_rate": 6.814884578800013e-06,
+      "loss": 0.8475,
       "step": 21676
     },
     {
-      "epoch": 0.5953420669577875,
+      "epoch": 0.6151248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 7.42742183935494e-06,
-      "loss": 0.8916,
+      "learning_rate": 6.8140133807060834e-06,
+      "loss": 0.8425,
       "step": 21677
     },
     {
-      "epoch": 0.5953695311856307,
+      "epoch": 0.6151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 7.4265622721352285e-06,
-      "loss": 0.85,
+      "learning_rate": 6.813142209524328e-06,
+      "loss": 0.8377,
       "step": 21678
     },
     {
-      "epoch": 0.5953969954134739,
+      "epoch": 0.6151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 7.42570272527755e-06,
-      "loss": 0.8246,
+      "learning_rate": 6.812271065262106e-06,
+      "loss": 0.8753,
       "step": 21679
     },
     {
-      "epoch": 0.5954244596413172,
+      "epoch": 0.6152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 7.424843198788708e-06,
-      "loss": 0.7031,
+      "learning_rate": 6.811399947926769e-06,
+      "loss": 0.8337,
       "step": 21680
     },
     {
-      "epoch": 0.5954519238691605,
+      "epoch": 0.6152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 7.423983692675507e-06,
-      "loss": 0.8846,
+      "learning_rate": 6.8105288575256824e-06,
+      "loss": 0.7702,
       "step": 21681
     },
     {
-      "epoch": 0.5954793880970036,
+      "epoch": 0.615266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 7.4231242069447425e-06,
-      "loss": 0.9056,
+      "learning_rate": 6.809657794066203e-06,
+      "loss": 0.8103,
       "step": 21682
     },
     {
-      "epoch": 0.5955068523248469,
+      "epoch": 0.6152951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 7.422264741603222e-06,
-      "loss": 0.8077,
+      "learning_rate": 6.808786757555684e-06,
+      "loss": 0.9079,
       "step": 21683
     },
     {
-      "epoch": 0.5955343165526901,
+      "epoch": 0.6153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 7.4214052966577375e-06,
-      "loss": 0.8113,
+      "learning_rate": 6.80791574800149e-06,
+      "loss": 0.815,
       "step": 21684
     },
     {
-      "epoch": 0.5955617807805333,
+      "epoch": 0.6153518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 7.420545872115092e-06,
-      "loss": 0.8981,
+      "learning_rate": 6.807044765410975e-06,
+      "loss": 0.7233,
       "step": 21685
     },
     {
-      "epoch": 0.5955892450083766,
+      "epoch": 0.6153802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 7.41968646798209e-06,
-      "loss": 0.8829,
+      "learning_rate": 6.806173809791492e-06,
+      "loss": 0.8658,
       "step": 21686
     },
     {
-      "epoch": 0.5956167092362198,
+      "epoch": 0.6154086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 7.418827084265525e-06,
-      "loss": 0.8904,
+      "learning_rate": 6.805302881150404e-06,
+      "loss": 0.8622,
       "step": 21687
     },
     {
-      "epoch": 0.5956441734640631,
+      "epoch": 0.6154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 7.417967720972201e-06,
-      "loss": 0.8208,
+      "learning_rate": 6.80443197949507e-06,
+      "loss": 0.8616,
       "step": 21688
     },
     {
-      "epoch": 0.5956716376919063,
+      "epoch": 0.6154653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 7.41710837810892e-06,
-      "loss": 0.8788,
+      "learning_rate": 6.803561104832841e-06,
+      "loss": 0.7868,
       "step": 21689
     },
     {
-      "epoch": 0.5956991019197495,
+      "epoch": 0.6154937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 7.4162490556824765e-06,
-      "loss": 0.8839,
+      "learning_rate": 6.802690257171075e-06,
+      "loss": 0.916,
       "step": 21690
     },
     {
-      "epoch": 0.5957265661475928,
+      "epoch": 0.615522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 7.4153897536996685e-06,
-      "loss": 0.9352,
+      "learning_rate": 6.801819436517128e-06,
+      "loss": 0.869,
       "step": 21691
     },
     {
-      "epoch": 0.595754030375436,
+      "epoch": 0.6155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 7.414530472167299e-06,
-      "loss": 0.8483,
+      "learning_rate": 6.800948642878356e-06,
+      "loss": 0.8312,
       "step": 21692
     },
     {
-      "epoch": 0.5957814946032792,
+      "epoch": 0.6155788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 7.413671211092167e-06,
-      "loss": 0.8414,
+      "learning_rate": 6.8000778762621145e-06,
+      "loss": 0.84,
       "step": 21693
     },
     {
-      "epoch": 0.5958089588311225,
+      "epoch": 0.6156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 7.4128119704810686e-06,
-      "loss": 0.8598,
+      "learning_rate": 6.799207136675758e-06,
+      "loss": 0.857,
       "step": 21694
     },
     {
-      "epoch": 0.5958364230589657,
+      "epoch": 0.6156356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 7.411952750340808e-06,
-      "loss": 0.7682,
+      "learning_rate": 6.798336424126645e-06,
+      "loss": 0.79,
       "step": 21695
     },
     {
-      "epoch": 0.595863887286809,
+      "epoch": 0.6156640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 7.411093550678177e-06,
-      "loss": 0.7401,
+      "learning_rate": 6.797465738622127e-06,
+      "loss": 0.8512,
       "step": 21696
     },
     {
-      "epoch": 0.5958913515146521,
+      "epoch": 0.6156923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 7.410234371499975e-06,
-      "loss": 0.9384,
+      "learning_rate": 6.796595080169559e-06,
+      "loss": 0.8549,
       "step": 21697
     },
     {
-      "epoch": 0.5959188157424954,
+      "epoch": 0.6157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 7.4093752128130034e-06,
-      "loss": 0.867,
+      "learning_rate": 6.795724448776298e-06,
+      "loss": 0.8511,
       "step": 21698
     },
     {
-      "epoch": 0.5959462799703387,
+      "epoch": 0.6157491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 7.408516074624058e-06,
-      "loss": 0.9786,
+      "learning_rate": 6.7948538444496955e-06,
+      "loss": 0.8559,
       "step": 21699
     },
     {
-      "epoch": 0.5959737441981818,
+      "epoch": 0.6157775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 7.407656956939937e-06,
-      "loss": 0.8096,
+      "learning_rate": 6.793983267197106e-06,
+      "loss": 0.8986,
       "step": 21700
     },
     {
-      "epoch": 0.5960012084260251,
+      "epoch": 0.615805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 7.406797859767443e-06,
-      "loss": 0.8147,
+      "learning_rate": 6.793112717025882e-06,
+      "loss": 0.9767,
       "step": 21701
     },
     {
-      "epoch": 0.5960286726538684,
+      "epoch": 0.6158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 7.4059387831133655e-06,
-      "loss": 0.8049,
+      "learning_rate": 6.792242193943382e-06,
+      "loss": 0.9325,
       "step": 21702
     },
     {
-      "epoch": 0.5960561368817116,
+      "epoch": 0.6158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 7.405079726984503e-06,
-      "loss": 0.9214,
+      "learning_rate": 6.791371697956957e-06,
+      "loss": 0.9371,
       "step": 21703
     },
     {
-      "epoch": 0.5960836011095548,
+      "epoch": 0.6158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 7.404220691387656e-06,
-      "loss": 0.8555,
+      "learning_rate": 6.790501229073958e-06,
+      "loss": 0.6561,
       "step": 21704
     },
     {
-      "epoch": 0.596111065337398,
+      "epoch": 0.6159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 7.403361676329622e-06,
-      "loss": 0.8372,
+      "learning_rate": 6.789630787301741e-06,
+      "loss": 0.8029,
       "step": 21705
     },
     {
-      "epoch": 0.5961385295652413,
+      "epoch": 0.6159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 7.402502681817196e-06,
-      "loss": 0.8408,
+      "learning_rate": 6.788760372647657e-06,
+      "loss": 0.887,
       "step": 21706
     },
     {
-      "epoch": 0.5961659937930845,
+      "epoch": 0.6159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 7.401643707857176e-06,
-      "loss": 0.8689,
+      "learning_rate": 6.787889985119056e-06,
+      "loss": 0.7422,
       "step": 21707
     },
     {
-      "epoch": 0.5961934580209277,
+      "epoch": 0.6160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 7.400784754456354e-06,
-      "loss": 0.791,
+      "learning_rate": 6.7870196247232954e-06,
+      "loss": 0.7759,
       "step": 21708
     },
     {
-      "epoch": 0.596220922248771,
+      "epoch": 0.6160329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 7.399925821621531e-06,
-      "loss": 0.8689,
+      "learning_rate": 6.786149291467724e-06,
+      "loss": 0.7945,
       "step": 21709
     },
     {
-      "epoch": 0.5962483864766142,
+      "epoch": 0.6160612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 7.399066909359502e-06,
-      "loss": 0.9154,
+      "learning_rate": 6.785278985359692e-06,
+      "loss": 0.8864,
       "step": 21710
     },
     {
-      "epoch": 0.5962758507044574,
+      "epoch": 0.6160896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 7.398208017677064e-06,
-      "loss": 0.8642,
+      "learning_rate": 6.784408706406555e-06,
+      "loss": 0.9153,
       "step": 21711
     },
     {
-      "epoch": 0.5963033149323007,
+      "epoch": 0.6161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 7.397349146581014e-06,
-      "loss": 0.8628,
+      "learning_rate": 6.783538454615664e-06,
+      "loss": 0.8265,
       "step": 21712
     },
     {
-      "epoch": 0.5963307791601439,
+      "epoch": 0.6161464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 7.396490296078143e-06,
-      "loss": 0.8701,
+      "learning_rate": 6.7826682299943635e-06,
+      "loss": 0.9571,
       "step": 21713
     },
     {
-      "epoch": 0.5963582433879872,
+      "epoch": 0.6161748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 7.395631466175251e-06,
-      "loss": 0.8677,
+      "learning_rate": 6.781798032550013e-06,
+      "loss": 0.7077,
       "step": 21714
     },
     {
-      "epoch": 0.5963857076158304,
+      "epoch": 0.6162031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 7.39477265687913e-06,
-      "loss": 0.882,
+      "learning_rate": 6.780927862289957e-06,
+      "loss": 0.7887,
       "step": 21715
     },
     {
-      "epoch": 0.5964131718436736,
+      "epoch": 0.6162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 7.393913868196576e-06,
-      "loss": 0.856,
+      "learning_rate": 6.780057719221551e-06,
+      "loss": 0.7607,
       "step": 21716
     },
     {
-      "epoch": 0.5964406360715169,
+      "epoch": 0.6162599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 7.3930551001343875e-06,
-      "loss": 0.9504,
+      "learning_rate": 6.779187603352136e-06,
+      "loss": 0.8354,
       "step": 21717
     },
     {
-      "epoch": 0.59646810029936,
+      "epoch": 0.6162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 7.392196352699359e-06,
-      "loss": 0.8862,
+      "learning_rate": 6.77831751468907e-06,
+      "loss": 0.8844,
       "step": 21718
     },
     {
-      "epoch": 0.5964955645272033,
+      "epoch": 0.616316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 7.391337625898278e-06,
-      "loss": 0.9205,
+      "learning_rate": 6.777447453239704e-06,
+      "loss": 0.952,
       "step": 21719
     },
     {
-      "epoch": 0.5965230287550466,
+      "epoch": 0.6163450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 7.3904789197379445e-06,
-      "loss": 0.911,
+      "learning_rate": 6.776577419011384e-06,
+      "loss": 0.8538,
       "step": 21720
     },
     {
-      "epoch": 0.5965504929828898,
+      "epoch": 0.6163734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 7.389620234225153e-06,
-      "loss": 0.8414,
+      "learning_rate": 6.7757074120114565e-06,
+      "loss": 0.888,
       "step": 21721
     },
     {
-      "epoch": 0.596577957210733,
+      "epoch": 0.6164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 7.3887615693667e-06,
-      "loss": 0.7775,
+      "learning_rate": 6.774837432247276e-06,
+      "loss": 0.8986,
       "step": 21722
     },
     {
-      "epoch": 0.5966054214385762,
+      "epoch": 0.6164301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 7.387902925169375e-06,
-      "loss": 0.8439,
+      "learning_rate": 6.773967479726189e-06,
+      "loss": 0.8589,
       "step": 21723
     },
     {
-      "epoch": 0.5966328856664195,
+      "epoch": 0.6164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 7.387044301639976e-06,
-      "loss": 0.8696,
+      "learning_rate": 6.773097554455543e-06,
+      "loss": 0.8125,
       "step": 21724
     },
     {
-      "epoch": 0.5966603498942628,
+      "epoch": 0.6164869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 7.386185698785293e-06,
-      "loss": 0.8449,
+      "learning_rate": 6.772227656442687e-06,
+      "loss": 0.8556,
       "step": 21725
     },
     {
-      "epoch": 0.5966878141221059,
+      "epoch": 0.6165153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 7.385327116612122e-06,
-      "loss": 0.8242,
+      "learning_rate": 6.77135778569497e-06,
+      "loss": 0.7918,
       "step": 21726
     },
     {
-      "epoch": 0.5967152783499492,
+      "epoch": 0.6165437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 7.384468555127253e-06,
-      "loss": 0.8413,
+      "learning_rate": 6.770487942219737e-06,
+      "loss": 0.936,
       "step": 21727
     },
     {
-      "epoch": 0.5967427425777925,
+      "epoch": 0.616572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 7.383610014337483e-06,
-      "loss": 0.7758,
+      "learning_rate": 6.769618126024338e-06,
+      "loss": 0.8525,
       "step": 21728
     },
     {
-      "epoch": 0.5967702068056356,
+      "epoch": 0.6166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 7.382751494249607e-06,
-      "loss": 0.882,
+      "learning_rate": 6.768748337116121e-06,
+      "loss": 0.6876,
       "step": 21729
     },
     {
-      "epoch": 0.5967976710334789,
+      "epoch": 0.6166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 7.381892994870416e-06,
-      "loss": 0.9027,
+      "learning_rate": 6.76787857550243e-06,
+      "loss": 0.9297,
       "step": 21730
     },
     {
-      "epoch": 0.5968251352613221,
+      "epoch": 0.6166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 7.381034516206698e-06,
-      "loss": 0.805,
+      "learning_rate": 6.767008841190616e-06,
+      "loss": 0.8934,
       "step": 21731
     },
     {
-      "epoch": 0.5968525994891654,
+      "epoch": 0.6166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 7.3801760582652495e-06,
-      "loss": 0.7994,
+      "learning_rate": 6.766139134188019e-06,
+      "loss": 0.9762,
       "step": 21732
     },
     {
-      "epoch": 0.5968800637170086,
+      "epoch": 0.6167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 7.379317621052862e-06,
-      "loss": 0.7791,
+      "learning_rate": 6.765269454501995e-06,
+      "loss": 0.7708,
       "step": 21733
     },
     {
-      "epoch": 0.5969075279448518,
+      "epoch": 0.6167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 7.37845920457633e-06,
-      "loss": 0.9189,
+      "learning_rate": 6.764399802139885e-06,
+      "loss": 0.8425,
       "step": 21734
     },
     {
-      "epoch": 0.5969349921726951,
+      "epoch": 0.6167707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 7.3776008088424444e-06,
-      "loss": 0.9236,
+      "learning_rate": 6.7635301771090345e-06,
+      "loss": 0.904,
       "step": 21735
     },
     {
-      "epoch": 0.5969624564005382,
+      "epoch": 0.6167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 7.376742433858e-06,
-      "loss": 0.9192,
+      "learning_rate": 6.762660579416791e-06,
+      "loss": 0.8302,
       "step": 21736
     },
     {
-      "epoch": 0.5969899206283815,
+      "epoch": 0.616827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 7.37588407962978e-06,
-      "loss": 0.7687,
+      "learning_rate": 6.761791009070499e-06,
+      "loss": 0.8495,
       "step": 21737
     },
     {
-      "epoch": 0.5970173848562248,
+      "epoch": 0.616855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 7.375025746164586e-06,
-      "loss": 0.8554,
+      "learning_rate": 6.7609214660775015e-06,
+      "loss": 0.8908,
       "step": 21738
     },
     {
-      "epoch": 0.597044849084068,
+      "epoch": 0.6168842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 7.3741674334692024e-06,
-      "loss": 0.8559,
+      "learning_rate": 6.760051950445149e-06,
+      "loss": 0.7386,
       "step": 21739
     },
     {
-      "epoch": 0.5970723133119112,
+      "epoch": 0.6169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 7.373309141550423e-06,
-      "loss": 0.86,
+      "learning_rate": 6.759182462180782e-06,
+      "loss": 0.8295,
       "step": 21740
     },
     {
-      "epoch": 0.5970997775397545,
+      "epoch": 0.6169409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 7.372450870415042e-06,
-      "loss": 0.9033,
+      "learning_rate": 6.7583130012917455e-06,
+      "loss": 0.8107,
       "step": 21741
     },
     {
-      "epoch": 0.5971272417675977,
+      "epoch": 0.6169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 7.3715926200698495e-06,
-      "loss": 0.8429,
+      "learning_rate": 6.757443567785387e-06,
+      "loss": 0.8655,
       "step": 21742
     },
     {
-      "epoch": 0.597154705995441,
+      "epoch": 0.6169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 7.370734390521629e-06,
-      "loss": 0.9072,
+      "learning_rate": 6.756574161669048e-06,
+      "loss": 0.8141,
       "step": 21743
     },
     {
-      "epoch": 0.5971821702232841,
+      "epoch": 0.6170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 7.369876181777177e-06,
-      "loss": 0.8525,
+      "learning_rate": 6.755704782950071e-06,
+      "loss": 0.8284,
       "step": 21744
     },
     {
-      "epoch": 0.5972096344511274,
+      "epoch": 0.6170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 7.369017993843284e-06,
-      "loss": 0.8673,
+      "learning_rate": 6.754835431635804e-06,
+      "loss": 0.8484,
       "step": 21745
     },
     {
-      "epoch": 0.5972370986789707,
+      "epoch": 0.6170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 7.368159826726742e-06,
-      "loss": 0.9029,
+      "learning_rate": 6.753966107733587e-06,
+      "loss": 0.7929,
       "step": 21746
     },
     {
-      "epoch": 0.5972645629068138,
+      "epoch": 0.6171112372304199,
       "grad_norm": 0.0,
-      "learning_rate": 7.3673016804343355e-06,
-      "loss": 0.8153,
+      "learning_rate": 6.753096811250761e-06,
+      "loss": 0.8843,
       "step": 21747
     },
     {
-      "epoch": 0.5972920271346571,
+      "epoch": 0.6171396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 7.366443554972863e-06,
-      "loss": 0.8615,
+      "learning_rate": 6.752227542194674e-06,
+      "loss": 0.892,
       "step": 21748
     },
     {
-      "epoch": 0.5973194913625004,
+      "epoch": 0.6171679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 7.365585450349104e-06,
-      "loss": 0.7321,
+      "learning_rate": 6.751358300572668e-06,
+      "loss": 0.8229,
       "step": 21749
     },
     {
-      "epoch": 0.5973469555903436,
+      "epoch": 0.6171963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 7.364727366569857e-06,
-      "loss": 0.9427,
+      "learning_rate": 6.750489086392084e-06,
+      "loss": 0.854,
       "step": 21750
     },
     {
-      "epoch": 0.5973744198181868,
+      "epoch": 0.6172247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 7.363869303641905e-06,
-      "loss": 0.8548,
+      "learning_rate": 6.749619899660265e-06,
+      "loss": 0.8112,
       "step": 21751
     },
     {
-      "epoch": 0.59740188404603,
+      "epoch": 0.6172531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 7.363011261572041e-06,
-      "loss": 0.8329,
+      "learning_rate": 6.748750740384553e-06,
+      "loss": 0.9041,
       "step": 21752
     },
     {
-      "epoch": 0.5974293482738733,
+      "epoch": 0.6172814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 7.362153240367055e-06,
-      "loss": 0.9242,
+      "learning_rate": 6.747881608572291e-06,
+      "loss": 0.9017,
       "step": 21753
     },
     {
-      "epoch": 0.5974568125017166,
+      "epoch": 0.6173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 7.361295240033737e-06,
-      "loss": 0.8586,
+      "learning_rate": 6.747012504230817e-06,
+      "loss": 0.8288,
       "step": 21754
     },
     {
-      "epoch": 0.5974842767295597,
+      "epoch": 0.6173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 7.360437260578867e-06,
-      "loss": 0.9122,
+      "learning_rate": 6.7461434273674765e-06,
+      "loss": 0.7885,
       "step": 21755
     },
     {
-      "epoch": 0.597511740957403,
+      "epoch": 0.617366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.359579302009241e-06,
-      "loss": 0.9046,
+      "learning_rate": 6.745274377989608e-06,
+      "loss": 0.8775,
       "step": 21756
     },
     {
-      "epoch": 0.5975392051852462,
+      "epoch": 0.6173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 7.358721364331649e-06,
-      "loss": 0.8266,
+      "learning_rate": 6.744405356104554e-06,
+      "loss": 0.8707,
       "step": 21757
     },
     {
-      "epoch": 0.5975666694130894,
+      "epoch": 0.6174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 7.357863447552874e-06,
-      "loss": 0.9207,
+      "learning_rate": 6.743536361719651e-06,
+      "loss": 0.8927,
       "step": 21758
     },
     {
-      "epoch": 0.5975941336409327,
+      "epoch": 0.6174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 7.357005551679708e-06,
-      "loss": 0.8136,
+      "learning_rate": 6.742667394842247e-06,
+      "loss": 0.8824,
       "step": 21759
     },
     {
-      "epoch": 0.5976215978687759,
+      "epoch": 0.6174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 7.356147676718941e-06,
-      "loss": 0.9436,
+      "learning_rate": 6.741798455479677e-06,
+      "loss": 0.9137,
       "step": 21760
     },
     {
-      "epoch": 0.5976490620966192,
+      "epoch": 0.6175085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 7.355289822677354e-06,
-      "loss": 0.9581,
+      "learning_rate": 6.740929543639279e-06,
+      "loss": 0.8259,
       "step": 21761
     },
     {
-      "epoch": 0.5976765263244624,
+      "epoch": 0.6175368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 7.35443198956174e-06,
-      "loss": 0.7716,
+      "learning_rate": 6.740060659328398e-06,
+      "loss": 0.8312,
       "step": 21762
     },
     {
-      "epoch": 0.5977039905523056,
+      "epoch": 0.6175652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 7.353574177378883e-06,
-      "loss": 0.7646,
+      "learning_rate": 6.739191802554368e-06,
+      "loss": 0.8054,
       "step": 21763
     },
     {
-      "epoch": 0.5977314547801489,
+      "epoch": 0.6175936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 7.3527163861355726e-06,
-      "loss": 0.8561,
+      "learning_rate": 6.7383229733245345e-06,
+      "loss": 0.8301,
       "step": 21764
     },
     {
-      "epoch": 0.597758919007992,
+      "epoch": 0.6176220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 7.351858615838597e-06,
-      "loss": 0.8259,
+      "learning_rate": 6.737454171646231e-06,
+      "loss": 0.9967,
       "step": 21765
     },
     {
-      "epoch": 0.5977863832358353,
+      "epoch": 0.6176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 7.351000866494744e-06,
-      "loss": 0.7806,
+      "learning_rate": 6.736585397526802e-06,
+      "loss": 0.8056,
       "step": 21766
     },
     {
-      "epoch": 0.5978138474636786,
+      "epoch": 0.6176787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 7.350143138110795e-06,
-      "loss": 0.7929,
+      "learning_rate": 6.735716650973583e-06,
+      "loss": 0.7994,
       "step": 21767
     },
     {
-      "epoch": 0.5978413116915218,
+      "epoch": 0.6177071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 7.349285430693538e-06,
-      "loss": 0.8672,
+      "learning_rate": 6.73484793199391e-06,
+      "loss": 0.9818,
       "step": 21768
     },
     {
-      "epoch": 0.597868775919365,
+      "epoch": 0.6177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 7.348427744249765e-06,
-      "loss": 0.944,
+      "learning_rate": 6.733979240595124e-06,
+      "loss": 0.7628,
       "step": 21769
     },
     {
-      "epoch": 0.5978962401472082,
+      "epoch": 0.6177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 7.347570078786257e-06,
-      "loss": 0.8549,
+      "learning_rate": 6.733110576784564e-06,
+      "loss": 0.7924,
       "step": 21770
     },
     {
-      "epoch": 0.5979237043750515,
+      "epoch": 0.6177922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 7.346712434309806e-06,
-      "loss": 0.8944,
+      "learning_rate": 6.7322419405695625e-06,
+      "loss": 0.8381,
       "step": 21771
     },
     {
-      "epoch": 0.5979511686028948,
+      "epoch": 0.6178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 7.345854810827189e-06,
-      "loss": 0.8844,
+      "learning_rate": 6.731373331957463e-06,
+      "loss": 0.8478,
       "step": 21772
     },
     {
-      "epoch": 0.5979786328307379,
+      "epoch": 0.6178490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 7.344997208345198e-06,
-      "loss": 0.9098,
+      "learning_rate": 6.7305047509555995e-06,
+      "loss": 0.7826,
       "step": 21773
     },
     {
-      "epoch": 0.5980060970585812,
+      "epoch": 0.617877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 7.344139626870619e-06,
-      "loss": 0.9911,
+      "learning_rate": 6.7296361975713074e-06,
+      "loss": 0.7965,
       "step": 21774
     },
     {
-      "epoch": 0.5980335612864245,
+      "epoch": 0.617905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 7.343282066410234e-06,
-      "loss": 0.8192,
+      "learning_rate": 6.728767671811927e-06,
+      "loss": 0.8416,
       "step": 21775
     },
     {
-      "epoch": 0.5980610255142677,
+      "epoch": 0.6179341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 7.342424526970831e-06,
-      "loss": 0.9282,
+      "learning_rate": 6.727899173684793e-06,
+      "loss": 0.7947,
       "step": 21776
     },
     {
-      "epoch": 0.5980884897421109,
+      "epoch": 0.6179625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 7.341567008559198e-06,
-      "loss": 0.8811,
+      "learning_rate": 6.727030703197241e-06,
+      "loss": 0.9538,
       "step": 21777
     },
     {
-      "epoch": 0.5981159539699541,
+      "epoch": 0.6179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 7.340709511182114e-06,
-      "loss": 0.8497,
+      "learning_rate": 6.726162260356607e-06,
+      "loss": 0.8752,
       "step": 21778
     },
     {
-      "epoch": 0.5981434181977974,
+      "epoch": 0.6180192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 7.3398520348463656e-06,
-      "loss": 0.8997,
+      "learning_rate": 6.725293845170227e-06,
+      "loss": 0.9504,
       "step": 21779
     },
     {
-      "epoch": 0.5981708824256406,
+      "epoch": 0.6180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 7.338994579558738e-06,
-      "loss": 0.7821,
+      "learning_rate": 6.724425457645441e-06,
+      "loss": 0.8848,
       "step": 21780
     },
     {
-      "epoch": 0.5981983466534838,
+      "epoch": 0.6180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 7.3381371453260175e-06,
-      "loss": 0.8666,
+      "learning_rate": 6.723557097789578e-06,
+      "loss": 0.9665,
       "step": 21781
     },
     {
-      "epoch": 0.5982258108813271,
+      "epoch": 0.6181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 7.337279732154985e-06,
-      "loss": 0.8679,
+      "learning_rate": 6.7226887656099745e-06,
+      "loss": 0.7786,
       "step": 21782
     },
     {
-      "epoch": 0.5982532751091703,
+      "epoch": 0.6181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 7.336422340052431e-06,
-      "loss": 0.9321,
+      "learning_rate": 6.7218204611139685e-06,
+      "loss": 0.79,
       "step": 21783
     },
     {
-      "epoch": 0.5982807393370135,
+      "epoch": 0.618161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 7.3355649690251315e-06,
-      "loss": 0.8272,
+      "learning_rate": 6.7209521843088925e-06,
+      "loss": 0.905,
       "step": 21784
     },
     {
-      "epoch": 0.5983082035648568,
+      "epoch": 0.6181895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 7.334707619079875e-06,
-      "loss": 0.8466,
+      "learning_rate": 6.720083935202078e-06,
+      "loss": 0.7908,
       "step": 21785
     },
     {
-      "epoch": 0.5983356677927,
+      "epoch": 0.6182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 7.333850290223444e-06,
-      "loss": 0.8834,
+      "learning_rate": 6.719215713800864e-06,
+      "loss": 0.7605,
       "step": 21786
     },
     {
-      "epoch": 0.5983631320205433,
+      "epoch": 0.6182463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 7.3329929824626215e-06,
-      "loss": 0.8543,
+      "learning_rate": 6.7183475201125835e-06,
+      "loss": 0.7424,
       "step": 21787
     },
     {
-      "epoch": 0.5983905962483865,
+      "epoch": 0.6182746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 7.332135695804193e-06,
-      "loss": 0.7905,
+      "learning_rate": 6.717479354144567e-06,
+      "loss": 0.9286,
       "step": 21788
     },
     {
-      "epoch": 0.5984180604762297,
+      "epoch": 0.6183030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 7.331278430254942e-06,
-      "loss": 0.8528,
+      "learning_rate": 6.716611215904151e-06,
+      "loss": 0.8433,
       "step": 21789
     },
     {
-      "epoch": 0.598445524704073,
+      "epoch": 0.6183314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 7.330421185821649e-06,
-      "loss": 0.904,
+      "learning_rate": 6.715743105398669e-06,
+      "loss": 0.823,
       "step": 21790
     },
     {
-      "epoch": 0.5984729889319161,
+      "epoch": 0.6183598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 7.329563962511095e-06,
-      "loss": 0.7377,
+      "learning_rate": 6.7148750226354485e-06,
+      "loss": 0.8434,
       "step": 21791
     },
     {
-      "epoch": 0.5985004531597594,
+      "epoch": 0.6183881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 7.328706760330066e-06,
-      "loss": 0.7709,
+      "learning_rate": 6.714006967621829e-06,
+      "loss": 0.8887,
       "step": 21792
     },
     {
-      "epoch": 0.5985279173876027,
+      "epoch": 0.618416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 7.327849579285346e-06,
-      "loss": 0.835,
+      "learning_rate": 6.7131389403651385e-06,
+      "loss": 0.9185,
       "step": 21793
     },
     {
-      "epoch": 0.5985553816154459,
+      "epoch": 0.6184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 7.326992419383712e-06,
-      "loss": 0.9183,
+      "learning_rate": 6.712270940872713e-06,
+      "loss": 0.9094,
       "step": 21794
     },
     {
-      "epoch": 0.5985828458432891,
+      "epoch": 0.6184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 7.3261352806319545e-06,
-      "loss": 0.997,
+      "learning_rate": 6.711402969151881e-06,
+      "loss": 0.7816,
       "step": 21795
     },
     {
-      "epoch": 0.5986103100711323,
+      "epoch": 0.6185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 7.3252781630368464e-06,
-      "loss": 0.7928,
+      "learning_rate": 6.710535025209978e-06,
+      "loss": 0.9648,
       "step": 21796
     },
     {
-      "epoch": 0.5986377742989756,
+      "epoch": 0.6185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 7.3244210666051725e-06,
-      "loss": 0.775,
+      "learning_rate": 6.709667109054334e-06,
+      "loss": 0.7929,
       "step": 21797
     },
     {
-      "epoch": 0.5986652385268189,
+      "epoch": 0.6185584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 7.323563991343718e-06,
-      "loss": 0.9007,
+      "learning_rate": 6.708799220692279e-06,
+      "loss": 0.7789,
       "step": 21798
     },
     {
-      "epoch": 0.598692702754662,
+      "epoch": 0.6185868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 7.322706937259261e-06,
-      "loss": 0.6923,
+      "learning_rate": 6.707931360131144e-06,
+      "loss": 0.7372,
       "step": 21799
     },
     {
-      "epoch": 0.5987201669825053,
+      "epoch": 0.6186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 7.3218499043585835e-06,
-      "loss": 0.8669,
+      "learning_rate": 6.707063527378261e-06,
+      "loss": 0.8544,
       "step": 21800
     },
     {
-      "epoch": 0.5987476312103486,
+      "epoch": 0.6186435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 7.3209928926484706e-06,
-      "loss": 0.7985,
+      "learning_rate": 6.706195722440963e-06,
+      "loss": 0.936,
       "step": 21801
     },
     {
-      "epoch": 0.5987750954381917,
+      "epoch": 0.618671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 7.320135902135698e-06,
-      "loss": 0.9454,
+      "learning_rate": 6.7053279453265744e-06,
+      "loss": 0.9677,
       "step": 21802
     },
     {
-      "epoch": 0.598802559666035,
+      "epoch": 0.618700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 7.319278932827045e-06,
-      "loss": 0.8896,
+      "learning_rate": 6.70446019604243e-06,
+      "loss": 0.8933,
       "step": 21803
     },
     {
-      "epoch": 0.5988300238938782,
+      "epoch": 0.6187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 7.318421984729296e-06,
-      "loss": 0.874,
+      "learning_rate": 6.703592474595859e-06,
+      "loss": 0.847,
       "step": 21804
     },
     {
-      "epoch": 0.5988574881217215,
+      "epoch": 0.6187570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 7.317565057849234e-06,
-      "loss": 0.8998,
+      "learning_rate": 6.702724780994189e-06,
+      "loss": 0.8133,
       "step": 21805
     },
     {
-      "epoch": 0.5988849523495647,
+      "epoch": 0.6187854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 7.316708152193633e-06,
-      "loss": 0.8723,
+      "learning_rate": 6.701857115244752e-06,
+      "loss": 0.8166,
       "step": 21806
     },
     {
-      "epoch": 0.5989124165774079,
+      "epoch": 0.6188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 7.315851267769283e-06,
-      "loss": 0.9407,
+      "learning_rate": 6.700989477354878e-06,
+      "loss": 0.9488,
       "step": 21807
     },
     {
-      "epoch": 0.5989398808052512,
+      "epoch": 0.6188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 7.314994404582952e-06,
-      "loss": 0.8475,
+      "learning_rate": 6.70012186733189e-06,
+      "loss": 0.8416,
       "step": 21808
     },
     {
-      "epoch": 0.5989673450330943,
+      "epoch": 0.6188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 7.314137562641424e-06,
-      "loss": 0.8176,
+      "learning_rate": 6.699254285183121e-06,
+      "loss": 0.9091,
       "step": 21809
     },
     {
-      "epoch": 0.5989948092609376,
+      "epoch": 0.6188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 7.3132807419514825e-06,
-      "loss": 0.8577,
+      "learning_rate": 6.698386730915903e-06,
+      "loss": 0.8333,
       "step": 21810
     },
     {
-      "epoch": 0.5990222734887809,
+      "epoch": 0.6189273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 7.312423942519904e-06,
-      "loss": 0.8202,
+      "learning_rate": 6.697519204537559e-06,
+      "loss": 0.9325,
       "step": 21811
     },
     {
-      "epoch": 0.5990497377166241,
+      "epoch": 0.6189557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 7.3115671643534655e-06,
-      "loss": 0.8517,
+      "learning_rate": 6.6966517060554185e-06,
+      "loss": 0.813,
       "step": 21812
     },
     {
-      "epoch": 0.5990772019444673,
+      "epoch": 0.6189841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 7.310710407458957e-06,
-      "loss": 0.8929,
+      "learning_rate": 6.695784235476811e-06,
+      "loss": 0.8438,
       "step": 21813
     },
     {
-      "epoch": 0.5991046661723106,
+      "epoch": 0.6190124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 7.3098536718431445e-06,
-      "loss": 0.886,
+      "learning_rate": 6.694916792809063e-06,
+      "loss": 0.8625,
       "step": 21814
     },
     {
-      "epoch": 0.5991321304001538,
+      "epoch": 0.6190408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 7.30899695751281e-06,
-      "loss": 0.8609,
+      "learning_rate": 6.694049378059498e-06,
+      "loss": 0.8224,
       "step": 21815
     },
     {
-      "epoch": 0.5991595946279971,
+      "epoch": 0.6190692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 7.308140264474735e-06,
-      "loss": 0.9211,
+      "learning_rate": 6.69318199123545e-06,
+      "loss": 0.7755,
       "step": 21816
     },
     {
-      "epoch": 0.5991870588558402,
+      "epoch": 0.6190976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 7.307283592735699e-06,
-      "loss": 0.8067,
+      "learning_rate": 6.692314632344241e-06,
+      "loss": 0.8492,
       "step": 21817
     },
     {
-      "epoch": 0.5992145230836835,
+      "epoch": 0.6191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 7.306426942302476e-06,
-      "loss": 0.8092,
+      "learning_rate": 6.691447301393199e-06,
+      "loss": 0.8649,
       "step": 21818
     },
     {
-      "epoch": 0.5992419873115268,
+      "epoch": 0.6191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 7.305570313181851e-06,
-      "loss": 0.9884,
+      "learning_rate": 6.69057999838965e-06,
+      "loss": 0.8091,
       "step": 21819
     },
     {
-      "epoch": 0.59926945153937,
+      "epoch": 0.6191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 7.304713705380593e-06,
-      "loss": 0.9303,
+      "learning_rate": 6.689712723340923e-06,
+      "loss": 0.8208,
       "step": 21820
     },
     {
-      "epoch": 0.5992969157672132,
+      "epoch": 0.619211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 7.303857118905483e-06,
-      "loss": 0.7657,
+      "learning_rate": 6.688845476254337e-06,
+      "loss": 1.0013,
       "step": 21821
     },
     {
-      "epoch": 0.5993243799950564,
+      "epoch": 0.6192395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 7.303000553763303e-06,
-      "loss": 0.7873,
+      "learning_rate": 6.6879782571372265e-06,
+      "loss": 0.8771,
       "step": 21822
     },
     {
-      "epoch": 0.5993518442228997,
+      "epoch": 0.6192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 7.302144009960826e-06,
-      "loss": 0.8289,
+      "learning_rate": 6.687111065996912e-06,
+      "loss": 0.7353,
       "step": 21823
     },
     {
-      "epoch": 0.5993793084507429,
+      "epoch": 0.6192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 7.301287487504829e-06,
-      "loss": 0.8157,
+      "learning_rate": 6.686243902840715e-06,
+      "loss": 0.8721,
       "step": 21824
     },
     {
-      "epoch": 0.5994067726785861,
+      "epoch": 0.6193246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 7.300430986402096e-06,
-      "loss": 0.888,
+      "learning_rate": 6.6853767676759675e-06,
+      "loss": 0.818,
       "step": 21825
     },
     {
-      "epoch": 0.5994342369064294,
+      "epoch": 0.6193530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 7.299574506659396e-06,
-      "loss": 0.9187,
+      "learning_rate": 6.68450966050999e-06,
+      "loss": 0.8019,
       "step": 21826
     },
     {
-      "epoch": 0.5994617011342727,
+      "epoch": 0.6193813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 7.2987180482835065e-06,
-      "loss": 0.8829,
+      "learning_rate": 6.68364258135011e-06,
+      "loss": 0.8452,
       "step": 21827
     },
     {
-      "epoch": 0.5994891653621158,
+      "epoch": 0.6194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 7.297861611281206e-06,
-      "loss": 0.8423,
+      "learning_rate": 6.682775530203651e-06,
+      "loss": 0.8079,
       "step": 21828
     },
     {
-      "epoch": 0.5995166295899591,
+      "epoch": 0.6194381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 7.297005195659275e-06,
-      "loss": 0.9677,
+      "learning_rate": 6.681908507077935e-06,
+      "loss": 0.9027,
       "step": 21829
     },
     {
-      "epoch": 0.5995440938178023,
+      "epoch": 0.6194665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 7.2961488014244844e-06,
-      "loss": 0.9272,
+      "learning_rate": 6.681041511980289e-06,
+      "loss": 0.9117,
       "step": 21830
     },
     {
-      "epoch": 0.5995715580456455,
+      "epoch": 0.6194948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 7.295292428583611e-06,
-      "loss": 0.8473,
+      "learning_rate": 6.680174544918032e-06,
+      "loss": 0.8896,
       "step": 21831
     },
     {
-      "epoch": 0.5995990222734888,
+      "epoch": 0.6195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 7.294436077143432e-06,
-      "loss": 0.7968,
+      "learning_rate": 6.67930760589849e-06,
+      "loss": 0.8987,
       "step": 21832
     },
     {
-      "epoch": 0.599626486501332,
+      "epoch": 0.6195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 7.29357974711072e-06,
-      "loss": 0.7524,
+      "learning_rate": 6.678440694928987e-06,
+      "loss": 0.7624,
       "step": 21833
     },
     {
-      "epoch": 0.5996539507291753,
+      "epoch": 0.6195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 7.292723438492258e-06,
-      "loss": 0.888,
+      "learning_rate": 6.677573812016846e-06,
+      "loss": 0.8248,
       "step": 21834
     },
     {
-      "epoch": 0.5996814149570184,
+      "epoch": 0.6196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 7.291867151294814e-06,
-      "loss": 0.8674,
+      "learning_rate": 6.676706957169385e-06,
+      "loss": 0.9006,
       "step": 21835
     },
     {
-      "epoch": 0.5997088791848617,
+      "epoch": 0.6196367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 7.2910108855251695e-06,
-      "loss": 0.8831,
+      "learning_rate": 6.675840130393933e-06,
+      "loss": 0.7712,
       "step": 21836
     },
     {
-      "epoch": 0.599736343412705,
+      "epoch": 0.6196651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 7.290154641190092e-06,
-      "loss": 0.7815,
+      "learning_rate": 6.674973331697808e-06,
+      "loss": 0.7288,
       "step": 21837
     },
     {
-      "epoch": 0.5997638076405482,
+      "epoch": 0.6196935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 7.289298418296363e-06,
-      "loss": 0.7893,
+      "learning_rate": 6.674106561088332e-06,
+      "loss": 0.877,
       "step": 21838
     },
     {
-      "epoch": 0.5997912718683914,
+      "epoch": 0.61972190692395,
       "grad_norm": 0.0,
-      "learning_rate": 7.288442216850753e-06,
-      "loss": 0.8593,
+      "learning_rate": 6.673239818572826e-06,
+      "loss": 0.9152,
       "step": 21839
     },
     {
-      "epoch": 0.5998187360962347,
+      "epoch": 0.619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 7.287586036860039e-06,
-      "loss": 0.8663,
+      "learning_rate": 6.672373104158612e-06,
+      "loss": 0.9267,
       "step": 21840
     },
     {
-      "epoch": 0.5998462003240779,
+      "epoch": 0.6197786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 7.286729878330997e-06,
-      "loss": 0.8898,
+      "learning_rate": 6.671506417853015e-06,
+      "loss": 0.8803,
       "step": 21841
     },
     {
-      "epoch": 0.5998736645519211,
+      "epoch": 0.6198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 7.2858737412703995e-06,
-      "loss": 0.9904,
+      "learning_rate": 6.670639759663353e-06,
+      "loss": 0.8581,
       "step": 21842
     },
     {
-      "epoch": 0.5999011287797643,
+      "epoch": 0.6198354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 7.2850176256850205e-06,
-      "loss": 0.9042,
+      "learning_rate": 6.669773129596945e-06,
+      "loss": 0.8199,
       "step": 21843
     },
     {
-      "epoch": 0.5999285930076076,
+      "epoch": 0.6198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 7.28416153158163e-06,
-      "loss": 0.9658,
+      "learning_rate": 6.668906527661115e-06,
+      "loss": 0.8551,
       "step": 21844
     },
     {
-      "epoch": 0.5999560572354509,
+      "epoch": 0.6198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 7.283305458967006e-06,
-      "loss": 0.881,
+      "learning_rate": 6.6680399538631816e-06,
+      "loss": 0.7591,
       "step": 21845
     },
     {
-      "epoch": 0.599983521463294,
+      "epoch": 0.6199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 7.282449407847925e-06,
-      "loss": 0.9059,
+      "learning_rate": 6.667173408210461e-06,
+      "loss": 0.8793,
       "step": 21846
     },
     {
-      "epoch": 0.6000109856911373,
+      "epoch": 0.6199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 7.281593378231153e-06,
-      "loss": 0.9205,
+      "learning_rate": 6.666306890710279e-06,
+      "loss": 0.89,
       "step": 21847
     },
     {
-      "epoch": 0.6000384499189805,
+      "epoch": 0.6199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 7.280737370123471e-06,
-      "loss": 0.8221,
+      "learning_rate": 6.665440401369954e-06,
+      "loss": 0.9237,
       "step": 21848
     },
     {
-      "epoch": 0.6000659141468238,
+      "epoch": 0.620005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 7.2798813835316465e-06,
-      "loss": 0.8174,
+      "learning_rate": 6.6645739401968e-06,
+      "loss": 0.9152,
       "step": 21849
     },
     {
-      "epoch": 0.600093378374667,
+      "epoch": 0.6200340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 7.279025418462455e-06,
-      "loss": 0.8411,
+      "learning_rate": 6.663707507198143e-06,
+      "loss": 0.9438,
       "step": 21850
     },
     {
-      "epoch": 0.6001208426025102,
+      "epoch": 0.6200624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 7.278169474922667e-06,
-      "loss": 0.8712,
+      "learning_rate": 6.662841102381298e-06,
+      "loss": 0.9296,
       "step": 21851
     },
     {
-      "epoch": 0.6001483068303535,
+      "epoch": 0.6200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 7.277313552919057e-06,
-      "loss": 0.8729,
+      "learning_rate": 6.661974725753582e-06,
+      "loss": 0.7682,
       "step": 21852
     },
     {
-      "epoch": 0.6001757710581967,
+      "epoch": 0.6201191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 7.276457652458397e-06,
-      "loss": 0.843,
+      "learning_rate": 6.661108377322319e-06,
+      "loss": 0.8877,
       "step": 21853
     },
     {
-      "epoch": 0.6002032352860399,
+      "epoch": 0.6201475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 7.275601773547459e-06,
-      "loss": 0.8123,
+      "learning_rate": 6.660242057094821e-06,
+      "loss": 0.8409,
       "step": 21854
     },
     {
-      "epoch": 0.6002306995138832,
+      "epoch": 0.6201759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 7.274745916193017e-06,
-      "loss": 0.7864,
+      "learning_rate": 6.659375765078407e-06,
+      "loss": 0.9195,
       "step": 21855
     },
     {
-      "epoch": 0.6002581637417264,
+      "epoch": 0.6202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 7.273890080401838e-06,
-      "loss": 0.94,
+      "learning_rate": 6.658509501280397e-06,
+      "loss": 0.9978,
       "step": 21856
     },
     {
-      "epoch": 0.6002856279695696,
+      "epoch": 0.6202326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 7.273034266180696e-06,
-      "loss": 0.8026,
+      "learning_rate": 6.6576432657081095e-06,
+      "loss": 0.8621,
       "step": 21857
     },
     {
-      "epoch": 0.6003130921974129,
+      "epoch": 0.620261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 7.272178473536366e-06,
-      "loss": 0.8522,
+      "learning_rate": 6.65677705836886e-06,
+      "loss": 0.867,
       "step": 21858
     },
     {
-      "epoch": 0.6003405564252561,
+      "epoch": 0.6202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.271322702475614e-06,
-      "loss": 0.84,
+      "learning_rate": 6.655910879269962e-06,
+      "loss": 0.8328,
       "step": 21859
     },
     {
-      "epoch": 0.6003680206530994,
+      "epoch": 0.6203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 7.2704669530052185e-06,
-      "loss": 0.9015,
+      "learning_rate": 6.655044728418738e-06,
+      "loss": 0.9285,
       "step": 21860
     },
     {
-      "epoch": 0.6003954848809425,
+      "epoch": 0.6203461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 7.269611225131941e-06,
-      "loss": 0.8276,
+      "learning_rate": 6.654178605822502e-06,
+      "loss": 0.7657,
       "step": 21861
     },
     {
-      "epoch": 0.6004229491087858,
+      "epoch": 0.6203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 7.26875551886256e-06,
-      "loss": 0.8821,
+      "learning_rate": 6.653312511488567e-06,
+      "loss": 0.8338,
       "step": 21862
     },
     {
-      "epoch": 0.6004504133366291,
+      "epoch": 0.6204029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 7.2678998342038405e-06,
-      "loss": 0.924,
+      "learning_rate": 6.652446445424254e-06,
+      "loss": 0.9123,
       "step": 21863
     },
     {
-      "epoch": 0.6004778775644722,
+      "epoch": 0.6204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 7.267044171162556e-06,
-      "loss": 0.9442,
+      "learning_rate": 6.651580407636876e-06,
+      "loss": 0.8001,
       "step": 21864
     },
     {
-      "epoch": 0.6005053417923155,
+      "epoch": 0.6204597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 7.266188529745479e-06,
-      "loss": 0.8351,
+      "learning_rate": 6.650714398133749e-06,
+      "loss": 0.8969,
       "step": 21865
     },
     {
-      "epoch": 0.6005328060201588,
+      "epoch": 0.6204880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 7.265332909959377e-06,
-      "loss": 0.8574,
+      "learning_rate": 6.6498484169221865e-06,
+      "loss": 0.8915,
       "step": 21866
     },
     {
-      "epoch": 0.600560270248002,
+      "epoch": 0.6205164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 7.264477311811021e-06,
-      "loss": 0.8063,
+      "learning_rate": 6.6489824640095055e-06,
+      "loss": 0.9219,
       "step": 21867
     },
     {
-      "epoch": 0.6005877344758452,
+      "epoch": 0.6205448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 7.263621735307178e-06,
-      "loss": 0.7982,
+      "learning_rate": 6.648116539403022e-06,
+      "loss": 0.8929,
       "step": 21868
     },
     {
-      "epoch": 0.6006151987036884,
+      "epoch": 0.6205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 7.262766180454619e-06,
-      "loss": 0.8846,
+      "learning_rate": 6.647250643110046e-06,
+      "loss": 0.8701,
       "step": 21869
     },
     {
-      "epoch": 0.6006426629315317,
+      "epoch": 0.6206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 7.261910647260117e-06,
-      "loss": 0.9298,
+      "learning_rate": 6.646384775137893e-06,
+      "loss": 0.8338,
       "step": 21870
     },
     {
-      "epoch": 0.600670127159375,
+      "epoch": 0.6206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 7.261055135730437e-06,
-      "loss": 0.8783,
+      "learning_rate": 6.645518935493882e-06,
+      "loss": 0.8836,
       "step": 21871
     },
     {
-      "epoch": 0.6006975913872181,
+      "epoch": 0.6206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 7.260199645872352e-06,
-      "loss": 0.871,
+      "learning_rate": 6.644653124185323e-06,
+      "loss": 0.8082,
       "step": 21872
     },
     {
-      "epoch": 0.6007250556150614,
+      "epoch": 0.6206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 7.259344177692626e-06,
-      "loss": 0.8851,
+      "learning_rate": 6.6437873412195284e-06,
+      "loss": 0.7879,
       "step": 21873
     },
     {
-      "epoch": 0.6007525198429046,
+      "epoch": 0.6207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 7.258488731198032e-06,
-      "loss": 0.8721,
+      "learning_rate": 6.642921586603815e-06,
+      "loss": 0.7525,
       "step": 21874
     },
     {
-      "epoch": 0.6007799840707478,
+      "epoch": 0.6207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 7.257633306395335e-06,
-      "loss": 0.8633,
+      "learning_rate": 6.642055860345494e-06,
+      "loss": 0.9025,
       "step": 21875
     },
     {
-      "epoch": 0.6008074482985911,
+      "epoch": 0.620771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 7.256777903291306e-06,
-      "loss": 0.8481,
+      "learning_rate": 6.641190162451876e-06,
+      "loss": 0.8716,
       "step": 21876
     },
     {
-      "epoch": 0.6008349125264343,
+      "epoch": 0.620800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 7.255922521892715e-06,
-      "loss": 0.7857,
+      "learning_rate": 6.640324492930277e-06,
+      "loss": 0.785,
       "step": 21877
     },
     {
-      "epoch": 0.6008623767542776,
+      "epoch": 0.6208286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 7.255067162206331e-06,
-      "loss": 0.9846,
+      "learning_rate": 6.63945885178801e-06,
+      "loss": 0.7694,
       "step": 21878
     },
     {
-      "epoch": 0.6008898409821208,
+      "epoch": 0.6208569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 7.254211824238913e-06,
-      "loss": 0.7602,
+      "learning_rate": 6.63859323903238e-06,
+      "loss": 0.8674,
       "step": 21879
     },
     {
-      "epoch": 0.600917305209964,
+      "epoch": 0.6208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 7.253356507997235e-06,
-      "loss": 0.915,
+      "learning_rate": 6.637727654670709e-06,
+      "loss": 0.901,
       "step": 21880
     },
     {
-      "epoch": 0.6009447694378073,
+      "epoch": 0.6209137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 7.252501213488066e-06,
-      "loss": 0.8906,
+      "learning_rate": 6.636862098710302e-06,
+      "loss": 0.8235,
       "step": 21881
     },
     {
-      "epoch": 0.6009722336656504,
+      "epoch": 0.6209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 7.251645940718172e-06,
-      "loss": 0.8424,
+      "learning_rate": 6.6359965711584706e-06,
+      "loss": 0.8701,
       "step": 21882
     },
     {
-      "epoch": 0.6009996978934937,
+      "epoch": 0.6209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 7.2507906896943185e-06,
-      "loss": 0.932,
+      "learning_rate": 6.63513107202253e-06,
+      "loss": 0.7714,
       "step": 21883
     },
     {
-      "epoch": 0.601027162121337,
+      "epoch": 0.6209988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 7.249935460423276e-06,
-      "loss": 0.912,
+      "learning_rate": 6.634265601309787e-06,
+      "loss": 0.8569,
       "step": 21884
     },
     {
-      "epoch": 0.6010546263491802,
+      "epoch": 0.6210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 7.249080252911809e-06,
-      "loss": 0.9204,
+      "learning_rate": 6.633400159027551e-06,
+      "loss": 0.8047,
       "step": 21885
     },
     {
-      "epoch": 0.6010820905770234,
+      "epoch": 0.62105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 7.248225067166683e-06,
-      "loss": 0.8963,
+      "learning_rate": 6.632534745183139e-06,
+      "loss": 0.7981,
       "step": 21886
     },
     {
-      "epoch": 0.6011095548048666,
+      "epoch": 0.6210839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 7.247369903194668e-06,
-      "loss": 0.8116,
+      "learning_rate": 6.631669359783854e-06,
+      "loss": 0.9154,
       "step": 21887
     },
     {
-      "epoch": 0.6011370190327099,
+      "epoch": 0.6211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 7.2465147610025255e-06,
-      "loss": 0.8814,
+      "learning_rate": 6.630804002837013e-06,
+      "loss": 0.8375,
       "step": 21888
     },
     {
-      "epoch": 0.6011644832605532,
+      "epoch": 0.6211407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 7.2456596405970295e-06,
-      "loss": 0.8398,
+      "learning_rate": 6.629938674349921e-06,
+      "loss": 0.7627,
       "step": 21889
     },
     {
-      "epoch": 0.6011919474883963,
+      "epoch": 0.6211691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 7.24480454198494e-06,
-      "loss": 0.876,
+      "learning_rate": 6.629073374329889e-06,
+      "loss": 0.9248,
       "step": 21890
     },
     {
-      "epoch": 0.6012194117162396,
+      "epoch": 0.6211975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 7.243949465173021e-06,
-      "loss": 0.916,
+      "learning_rate": 6.628208102784225e-06,
+      "loss": 0.8515,
       "step": 21891
     },
     {
-      "epoch": 0.6012468759440829,
+      "epoch": 0.6212258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 7.243094410168044e-06,
-      "loss": 0.8881,
+      "learning_rate": 6.6273428597202395e-06,
+      "loss": 0.7264,
       "step": 21892
     },
     {
-      "epoch": 0.601274340171926,
+      "epoch": 0.6212542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 7.24223937697677e-06,
-      "loss": 0.837,
+      "learning_rate": 6.626477645145238e-06,
+      "loss": 0.8512,
       "step": 21893
     },
     {
-      "epoch": 0.6013018043997693,
+      "epoch": 0.6212826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 7.241384365605968e-06,
-      "loss": 0.8476,
+      "learning_rate": 6.6256124590665346e-06,
+      "loss": 0.715,
       "step": 21894
     },
     {
-      "epoch": 0.6013292686276125,
+      "epoch": 0.621311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 7.240529376062404e-06,
-      "loss": 0.9013,
+      "learning_rate": 6.624747301491433e-06,
+      "loss": 0.8176,
       "step": 21895
     },
     {
-      "epoch": 0.6013567328554558,
+      "epoch": 0.6213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 7.239674408352837e-06,
-      "loss": 0.8809,
+      "learning_rate": 6.623882172427242e-06,
+      "loss": 0.8073,
       "step": 21896
     },
     {
-      "epoch": 0.601384197083299,
+      "epoch": 0.6213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 7.238819462484035e-06,
-      "loss": 0.9016,
+      "learning_rate": 6.6230170718812714e-06,
+      "loss": 0.7603,
       "step": 21897
     },
     {
-      "epoch": 0.6014116613111422,
+      "epoch": 0.6213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 7.237964538462765e-06,
-      "loss": 0.7859,
+      "learning_rate": 6.622151999860828e-06,
+      "loss": 1.0014,
       "step": 21898
     },
     {
-      "epoch": 0.6014391255389855,
+      "epoch": 0.6214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 7.237109636295787e-06,
-      "loss": 0.8835,
+      "learning_rate": 6.621286956373216e-06,
+      "loss": 0.8062,
       "step": 21899
     },
     {
-      "epoch": 0.6014665897668287,
+      "epoch": 0.6214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 7.236254755989869e-06,
-      "loss": 0.847,
+      "learning_rate": 6.620421941425747e-06,
+      "loss": 0.8747,
       "step": 21900
     },
     {
-      "epoch": 0.6014940539946719,
+      "epoch": 0.6214812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 7.235399897551777e-06,
-      "loss": 0.7919,
+      "learning_rate": 6.619556955025721e-06,
+      "loss": 0.8851,
       "step": 21901
     },
     {
-      "epoch": 0.6015215182225152,
+      "epoch": 0.6215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 7.234545060988269e-06,
-      "loss": 0.8411,
+      "learning_rate": 6.618691997180456e-06,
+      "loss": 0.9279,
       "step": 21902
     },
     {
-      "epoch": 0.6015489824503584,
+      "epoch": 0.6215380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 7.233690246306111e-06,
-      "loss": 0.7717,
+      "learning_rate": 6.617827067897246e-06,
+      "loss": 0.9361,
       "step": 21903
     },
     {
-      "epoch": 0.6015764466782016,
+      "epoch": 0.621566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 7.232835453512068e-06,
-      "loss": 0.922,
+      "learning_rate": 6.616962167183407e-06,
+      "loss": 0.8331,
       "step": 21904
     },
     {
-      "epoch": 0.6016039109060449,
+      "epoch": 0.621594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 7.231980682612902e-06,
-      "loss": 0.8675,
+      "learning_rate": 6.61609729504624e-06,
+      "loss": 0.8775,
       "step": 21905
     },
     {
-      "epoch": 0.6016313751338881,
+      "epoch": 0.6216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 7.231125933615378e-06,
-      "loss": 0.8654,
+      "learning_rate": 6.615232451493052e-06,
+      "loss": 0.8448,
       "step": 21906
     },
     {
-      "epoch": 0.6016588393617314,
+      "epoch": 0.6216515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 7.230271206526262e-06,
-      "loss": 0.8303,
+      "learning_rate": 6.614367636531145e-06,
+      "loss": 0.9127,
       "step": 21907
     },
     {
-      "epoch": 0.6016863035895745,
+      "epoch": 0.6216799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 7.229416501352309e-06,
-      "loss": 0.9016,
+      "learning_rate": 6.613502850167829e-06,
+      "loss": 0.8253,
       "step": 21908
     },
     {
-      "epoch": 0.6017137678174178,
+      "epoch": 0.6217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 7.2285618181002835e-06,
-      "loss": 0.9211,
+      "learning_rate": 6.612638092410406e-06,
+      "loss": 0.9014,
       "step": 21909
     },
     {
-      "epoch": 0.6017412320452611,
+      "epoch": 0.6217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.227707156776955e-06,
-      "loss": 0.9057,
+      "learning_rate": 6.611773363266181e-06,
+      "loss": 0.9005,
       "step": 21910
     },
     {
-      "epoch": 0.6017686962731043,
+      "epoch": 0.6217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 7.226852517389078e-06,
-      "loss": 0.8258,
+      "learning_rate": 6.610908662742459e-06,
+      "loss": 0.8542,
       "step": 21911
     },
     {
-      "epoch": 0.6017961605009475,
+      "epoch": 0.6217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 7.225997899943418e-06,
-      "loss": 0.8422,
+      "learning_rate": 6.610043990846545e-06,
+      "loss": 0.9145,
       "step": 21912
     },
     {
-      "epoch": 0.6018236247287907,
+      "epoch": 0.6218217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 7.225143304446742e-06,
-      "loss": 0.8993,
+      "learning_rate": 6.609179347585739e-06,
+      "loss": 0.8648,
       "step": 21913
     },
     {
-      "epoch": 0.601851088956634,
+      "epoch": 0.6218501702610669,
       "grad_norm": 0.0,
-      "learning_rate": 7.224288730905805e-06,
-      "loss": 0.868,
+      "learning_rate": 6.60831473296735e-06,
+      "loss": 0.753,
       "step": 21914
     },
     {
-      "epoch": 0.6018785531844772,
+      "epoch": 0.6218785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 7.223434179327369e-06,
-      "loss": 0.6976,
+      "learning_rate": 6.60745014699868e-06,
+      "loss": 0.8029,
       "step": 21915
     },
     {
-      "epoch": 0.6019060174123204,
+      "epoch": 0.6219069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 7.222579649718197e-06,
-      "loss": 0.9204,
+      "learning_rate": 6.606585589687025e-06,
+      "loss": 0.7819,
       "step": 21916
     },
     {
-      "epoch": 0.6019334816401637,
+      "epoch": 0.6219353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 7.221725142085052e-06,
-      "loss": 0.8617,
+      "learning_rate": 6.605721061039696e-06,
+      "loss": 0.7637,
       "step": 21917
     },
     {
-      "epoch": 0.601960945868007,
+      "epoch": 0.6219636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 7.220870656434692e-06,
-      "loss": 0.7931,
+      "learning_rate": 6.604856561063997e-06,
+      "loss": 0.8241,
       "step": 21918
     },
     {
-      "epoch": 0.6019884100958501,
+      "epoch": 0.6219920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 7.220016192773885e-06,
-      "loss": 0.8342,
+      "learning_rate": 6.603992089767227e-06,
+      "loss": 0.9087,
       "step": 21919
     },
     {
-      "epoch": 0.6020158743236934,
+      "epoch": 0.6220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 7.219161751109382e-06,
-      "loss": 0.8736,
+      "learning_rate": 6.603127647156686e-06,
+      "loss": 0.8828,
       "step": 21920
     },
     {
-      "epoch": 0.6020433385515366,
+      "epoch": 0.6220488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 7.218307331447948e-06,
-      "loss": 0.8181,
+      "learning_rate": 6.602263233239681e-06,
+      "loss": 0.8319,
       "step": 21921
     },
     {
-      "epoch": 0.6020708027793799,
+      "epoch": 0.6220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 7.217452933796346e-06,
-      "loss": 0.8878,
+      "learning_rate": 6.601398848023511e-06,
+      "loss": 0.7794,
       "step": 21922
     },
     {
-      "epoch": 0.6020982670072231,
+      "epoch": 0.622105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 7.216598558161333e-06,
-      "loss": 0.8981,
+      "learning_rate": 6.600534491515476e-06,
+      "loss": 0.98,
       "step": 21923
     },
     {
-      "epoch": 0.6021257312350663,
+      "epoch": 0.6221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 7.215744204549669e-06,
-      "loss": 0.9019,
+      "learning_rate": 6.599670163722881e-06,
+      "loss": 0.8722,
       "step": 21924
     },
     {
-      "epoch": 0.6021531954629096,
+      "epoch": 0.6221623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 7.21488987296812e-06,
-      "loss": 0.8613,
+      "learning_rate": 6.5988058646530255e-06,
+      "loss": 0.8359,
       "step": 21925
     },
     {
-      "epoch": 0.6021806596907527,
+      "epoch": 0.6221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 7.2140355634234405e-06,
-      "loss": 0.9138,
+      "learning_rate": 6.597941594313206e-06,
+      "loss": 0.7853,
       "step": 21926
     },
     {
-      "epoch": 0.602208123918596,
+      "epoch": 0.6222190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 7.213181275922387e-06,
-      "loss": 0.7756,
+      "learning_rate": 6.59707735271073e-06,
+      "loss": 0.8045,
       "step": 21927
     },
     {
-      "epoch": 0.6022355881464393,
+      "epoch": 0.6222474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 7.212327010471724e-06,
-      "loss": 0.9107,
+      "learning_rate": 6.5962131398528935e-06,
+      "loss": 0.7588,
       "step": 21928
     },
     {
-      "epoch": 0.6022630523742825,
+      "epoch": 0.6222758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 7.211472767078211e-06,
-      "loss": 0.8229,
+      "learning_rate": 6.5953489557469975e-06,
+      "loss": 0.7002,
       "step": 21929
     },
     {
-      "epoch": 0.6022905166021257,
+      "epoch": 0.6223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 7.210618545748604e-06,
-      "loss": 0.7817,
+      "learning_rate": 6.5944848004003426e-06,
+      "loss": 0.9127,
       "step": 21930
     },
     {
-      "epoch": 0.602317980829969,
+      "epoch": 0.6223325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 7.209764346489668e-06,
-      "loss": 0.7847,
+      "learning_rate": 6.593620673820225e-06,
+      "loss": 0.8165,
       "step": 21931
     },
     {
-      "epoch": 0.6023454450578122,
+      "epoch": 0.6223609534619751,
       "grad_norm": 0.0,
-      "learning_rate": 7.208910169308152e-06,
-      "loss": 0.9391,
+      "learning_rate": 6.592756576013949e-06,
+      "loss": 0.8403,
       "step": 21932
     },
     {
-      "epoch": 0.6023729092856555,
+      "epoch": 0.6223893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 7.208056014210821e-06,
-      "loss": 0.8298,
+      "learning_rate": 6.591892506988813e-06,
+      "loss": 0.866,
       "step": 21933
     },
     {
-      "epoch": 0.6024003735134986,
+      "epoch": 0.6224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 7.2072018812044345e-06,
-      "loss": 0.9156,
+      "learning_rate": 6.59102846675211e-06,
+      "loss": 0.9634,
       "step": 21934
     },
     {
-      "epoch": 0.6024278377413419,
+      "epoch": 0.6224460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 7.206347770295745e-06,
-      "loss": 0.8823,
+      "learning_rate": 6.590164455311147e-06,
+      "loss": 0.7262,
       "step": 21935
     },
     {
-      "epoch": 0.6024553019691852,
+      "epoch": 0.6224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 7.2054936814915156e-06,
-      "loss": 0.7818,
+      "learning_rate": 6.5893004726732165e-06,
+      "loss": 0.8379,
       "step": 21936
     },
     {
-      "epoch": 0.6024827661970283,
+      "epoch": 0.6225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 7.204639614798505e-06,
-      "loss": 0.8025,
+      "learning_rate": 6.588436518845617e-06,
+      "loss": 0.778,
       "step": 21937
     },
     {
-      "epoch": 0.6025102304248716,
+      "epoch": 0.6225312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 7.20378557022347e-06,
-      "loss": 0.7687,
+      "learning_rate": 6.587572593835649e-06,
+      "loss": 0.8159,
       "step": 21938
     },
     {
-      "epoch": 0.6025376946527148,
+      "epoch": 0.6225595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 7.202931547773161e-06,
-      "loss": 0.8127,
+      "learning_rate": 6.58670869765061e-06,
+      "loss": 0.9203,
       "step": 21939
     },
     {
-      "epoch": 0.6025651588805581,
+      "epoch": 0.6225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 7.2020775474543426e-06,
-      "loss": 0.8487,
+      "learning_rate": 6.585844830297793e-06,
+      "loss": 0.7311,
       "step": 21940
     },
     {
-      "epoch": 0.6025926231084013,
+      "epoch": 0.622616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 7.201223569273772e-06,
-      "loss": 0.902,
+      "learning_rate": 6.5849809917845e-06,
+      "loss": 0.7626,
       "step": 21941
     },
     {
-      "epoch": 0.6026200873362445,
+      "epoch": 0.622644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 7.200369613238203e-06,
-      "loss": 0.8105,
+      "learning_rate": 6.5841171821180265e-06,
+      "loss": 0.9216,
       "step": 21942
     },
     {
-      "epoch": 0.6026475515640878,
+      "epoch": 0.6226730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 7.199515679354399e-06,
-      "loss": 0.8538,
+      "learning_rate": 6.583253401305667e-06,
+      "loss": 1.0009,
       "step": 21943
     },
     {
-      "epoch": 0.6026750157919311,
+      "epoch": 0.6227014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 7.198661767629106e-06,
-      "loss": 0.9161,
+      "learning_rate": 6.582389649354721e-06,
+      "loss": 0.9331,
       "step": 21944
     },
     {
-      "epoch": 0.6027024800197742,
+      "epoch": 0.6227298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 7.197807878069087e-06,
-      "loss": 0.8156,
+      "learning_rate": 6.581525926272484e-06,
+      "loss": 0.8962,
       "step": 21945
     },
     {
-      "epoch": 0.6027299442476175,
+      "epoch": 0.6227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 7.1969540106811e-06,
-      "loss": 0.8971,
+      "learning_rate": 6.580662232066249e-06,
+      "loss": 0.7675,
       "step": 21946
     },
     {
-      "epoch": 0.6027574084754607,
+      "epoch": 0.6227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 7.196100165471896e-06,
-      "loss": 0.9056,
+      "learning_rate": 6.579798566743314e-06,
+      "loss": 0.8174,
       "step": 21947
     },
     {
-      "epoch": 0.6027848727033039,
+      "epoch": 0.6228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 7.195246342448239e-06,
-      "loss": 0.8909,
+      "learning_rate": 6.578934930310974e-06,
+      "loss": 0.8868,
       "step": 21948
     },
     {
-      "epoch": 0.6028123369311472,
+      "epoch": 0.6228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 7.194392541616872e-06,
-      "loss": 0.8185,
+      "learning_rate": 6.578071322776526e-06,
+      "loss": 0.7882,
       "step": 21949
     },
     {
-      "epoch": 0.6028398011589904,
+      "epoch": 0.6228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 7.1935387629845645e-06,
-      "loss": 0.8727,
+      "learning_rate": 6.577207744147262e-06,
+      "loss": 0.8222,
       "step": 21950
     },
     {
-      "epoch": 0.6028672653868337,
+      "epoch": 0.622900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 7.19268500655806e-06,
-      "loss": 0.903,
+      "learning_rate": 6.576344194430479e-06,
+      "loss": 0.8197,
       "step": 21951
     },
     {
-      "epoch": 0.6028947296146769,
+      "epoch": 0.6229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 7.191831272344121e-06,
-      "loss": 0.7736,
+      "learning_rate": 6.575480673633472e-06,
+      "loss": 0.9275,
       "step": 21952
     },
     {
-      "epoch": 0.6029221938425201,
+      "epoch": 0.6229568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 7.190977560349501e-06,
-      "loss": 0.841,
+      "learning_rate": 6.574617181763532e-06,
+      "loss": 0.9277,
       "step": 21953
     },
     {
-      "epoch": 0.6029496580703634,
+      "epoch": 0.6229852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 7.1901238705809566e-06,
-      "loss": 0.786,
+      "learning_rate": 6.573753718827953e-06,
+      "loss": 0.8925,
       "step": 21954
     },
     {
-      "epoch": 0.6029771222982065,
+      "epoch": 0.6230136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 7.1892702030452395e-06,
-      "loss": 0.8193,
+      "learning_rate": 6.572890284834034e-06,
+      "loss": 0.875,
       "step": 21955
     },
     {
-      "epoch": 0.6030045865260498,
+      "epoch": 0.6230419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 7.1884165577491035e-06,
-      "loss": 0.7852,
+      "learning_rate": 6.572026879789064e-06,
+      "loss": 0.988,
       "step": 21956
     },
     {
-      "epoch": 0.6030320507538931,
+      "epoch": 0.6230703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 7.187562934699306e-06,
-      "loss": 0.9456,
+      "learning_rate": 6.571163503700334e-06,
+      "loss": 0.8547,
       "step": 21957
     },
     {
-      "epoch": 0.6030595149817363,
+      "epoch": 0.6230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 7.186709333902599e-06,
-      "loss": 0.8308,
+      "learning_rate": 6.570300156575143e-06,
+      "loss": 0.9959,
       "step": 21958
     },
     {
-      "epoch": 0.6030869792095795,
+      "epoch": 0.6231271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 7.1858557553657385e-06,
-      "loss": 0.8473,
+      "learning_rate": 6.569436838420781e-06,
+      "loss": 0.8621,
       "step": 21959
     },
     {
-      "epoch": 0.6031144434374227,
+      "epoch": 0.623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 7.185002199095481e-06,
-      "loss": 0.9088,
+      "learning_rate": 6.5685735492445365e-06,
+      "loss": 0.8429,
       "step": 21960
     },
     {
-      "epoch": 0.603141907665266,
+      "epoch": 0.6231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 7.184148665098571e-06,
-      "loss": 0.87,
+      "learning_rate": 6.5677102890537105e-06,
+      "loss": 0.8409,
       "step": 21961
     },
     {
-      "epoch": 0.6031693718931093,
+      "epoch": 0.6232122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 7.1832951533817705e-06,
-      "loss": 0.9281,
+      "learning_rate": 6.566847057855583e-06,
+      "loss": 0.903,
       "step": 21962
     },
     {
-      "epoch": 0.6031968361209524,
+      "epoch": 0.6232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 7.182441663951827e-06,
-      "loss": 0.7057,
+      "learning_rate": 6.565983855657458e-06,
+      "loss": 0.8491,
       "step": 21963
     },
     {
-      "epoch": 0.6032243003487957,
+      "epoch": 0.6232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 7.181588196815498e-06,
-      "loss": 0.844,
+      "learning_rate": 6.565120682466621e-06,
+      "loss": 0.859,
       "step": 21964
     },
     {
-      "epoch": 0.603251764576639,
+      "epoch": 0.6232973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 7.180734751979535e-06,
-      "loss": 0.8781,
+      "learning_rate": 6.564257538290364e-06,
+      "loss": 0.7121,
       "step": 21965
     },
     {
-      "epoch": 0.6032792288044821,
+      "epoch": 0.6233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 7.179881329450693e-06,
-      "loss": 0.8464,
+      "learning_rate": 6.563394423135978e-06,
+      "loss": 0.9209,
       "step": 21966
     },
     {
-      "epoch": 0.6033066930323254,
+      "epoch": 0.6233541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 7.17902792923572e-06,
-      "loss": 0.8541,
+      "learning_rate": 6.562531337010754e-06,
+      "loss": 0.8604,
       "step": 21967
     },
     {
-      "epoch": 0.6033341572601686,
+      "epoch": 0.6233825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 7.17817455134137e-06,
-      "loss": 0.8878,
+      "learning_rate": 6.561668279921982e-06,
+      "loss": 0.8548,
       "step": 21968
     },
     {
-      "epoch": 0.6033616214880119,
+      "epoch": 0.6234108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 7.177321195774396e-06,
-      "loss": 0.9208,
+      "learning_rate": 6.560805251876954e-06,
+      "loss": 0.8094,
       "step": 21969
     },
     {
-      "epoch": 0.6033890857158551,
+      "epoch": 0.623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 7.176467862541552e-06,
-      "loss": 0.8747,
+      "learning_rate": 6.559942252882956e-06,
+      "loss": 0.932,
       "step": 21970
     },
     {
-      "epoch": 0.6034165499436983,
+      "epoch": 0.6234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 7.175614551649586e-06,
-      "loss": 0.8862,
+      "learning_rate": 6.559079282947282e-06,
+      "loss": 0.9187,
       "step": 21971
     },
     {
-      "epoch": 0.6034440141715416,
+      "epoch": 0.6234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 7.174761263105257e-06,
-      "loss": 0.9468,
+      "learning_rate": 6.558216342077222e-06,
+      "loss": 0.7619,
       "step": 21972
     },
     {
-      "epoch": 0.6034714783993848,
+      "epoch": 0.6235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 7.1739079969153045e-06,
-      "loss": 0.8357,
+      "learning_rate": 6.55735343028006e-06,
+      "loss": 0.8379,
       "step": 21973
     },
     {
-      "epoch": 0.603498942627228,
+      "epoch": 0.6235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 7.17305475308649e-06,
-      "loss": 0.7917,
+      "learning_rate": 6.556490547563089e-06,
+      "loss": 0.9478,
       "step": 21974
     },
     {
-      "epoch": 0.6035264068550713,
+      "epoch": 0.6235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 7.17220153162556e-06,
-      "loss": 0.8807,
+      "learning_rate": 6.555627693933598e-06,
+      "loss": 0.8565,
       "step": 21975
     },
     {
-      "epoch": 0.6035538710829145,
+      "epoch": 0.6236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 7.171348332539266e-06,
-      "loss": 0.8289,
+      "learning_rate": 6.554764869398875e-06,
+      "loss": 0.7547,
       "step": 21976
     },
     {
-      "epoch": 0.6035813353107577,
+      "epoch": 0.6236379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 7.170495155834363e-06,
-      "loss": 0.8909,
+      "learning_rate": 6.553902073966204e-06,
+      "loss": 0.8672,
       "step": 21977
     },
     {
-      "epoch": 0.603608799538601,
+      "epoch": 0.62366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 7.169642001517597e-06,
-      "loss": 0.8546,
+      "learning_rate": 6.553039307642879e-06,
+      "loss": 0.8603,
       "step": 21978
     },
     {
-      "epoch": 0.6036362637664442,
+      "epoch": 0.6236946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 7.16878886959572e-06,
-      "loss": 0.8523,
+      "learning_rate": 6.552176570436188e-06,
+      "loss": 0.8163,
       "step": 21979
     },
     {
-      "epoch": 0.6036637279942875,
+      "epoch": 0.6237230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 7.1679357600754805e-06,
-      "loss": 0.8067,
+      "learning_rate": 6.551313862353417e-06,
+      "loss": 0.8085,
       "step": 21980
     },
     {
-      "epoch": 0.6036911922221306,
+      "epoch": 0.6237514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 7.16708267296363e-06,
-      "loss": 0.9118,
+      "learning_rate": 6.55045118340185e-06,
+      "loss": 0.8051,
       "step": 21981
     },
     {
-      "epoch": 0.6037186564499739,
+      "epoch": 0.6237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 7.166229608266921e-06,
-      "loss": 0.809,
+      "learning_rate": 6.54958853358878e-06,
+      "loss": 0.8144,
       "step": 21982
     },
     {
-      "epoch": 0.6037461206778172,
+      "epoch": 0.6238081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 7.1653765659921e-06,
-      "loss": 0.8543,
+      "learning_rate": 6.548725912921489e-06,
+      "loss": 0.8455,
       "step": 21983
     },
     {
-      "epoch": 0.6037735849056604,
+      "epoch": 0.6238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 7.16452354614592e-06,
-      "loss": 0.7727,
+      "learning_rate": 6.547863321407265e-06,
+      "loss": 0.8251,
       "step": 21984
     },
     {
-      "epoch": 0.6038010491335036,
+      "epoch": 0.6238649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 7.163670548735125e-06,
-      "loss": 0.7739,
+      "learning_rate": 6.547000759053397e-06,
+      "loss": 0.8505,
       "step": 21985
     },
     {
-      "epoch": 0.6038285133613468,
+      "epoch": 0.6238933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 7.1628175737664695e-06,
-      "loss": 0.9083,
+      "learning_rate": 6.546138225867167e-06,
+      "loss": 0.892,
       "step": 21986
     },
     {
-      "epoch": 0.6038559775891901,
+      "epoch": 0.6239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 7.161964621246697e-06,
-      "loss": 0.8381,
+      "learning_rate": 6.545275721855862e-06,
+      "loss": 0.6456,
       "step": 21987
     },
     {
-      "epoch": 0.6038834418170334,
+      "epoch": 0.623950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 7.16111169118256e-06,
-      "loss": 0.8791,
+      "learning_rate": 6.5444132470267695e-06,
+      "loss": 0.8966,
       "step": 21988
     },
     {
-      "epoch": 0.6039109060448765,
+      "epoch": 0.6239784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 7.160258783580811e-06,
-      "loss": 0.9397,
+      "learning_rate": 6.543550801387174e-06,
+      "loss": 0.8629,
       "step": 21989
     },
     {
-      "epoch": 0.6039383702727198,
+      "epoch": 0.6240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 7.159405898448193e-06,
-      "loss": 0.8052,
+      "learning_rate": 6.542688384944358e-06,
+      "loss": 0.8671,
       "step": 21990
     },
     {
-      "epoch": 0.6039658345005631,
+      "epoch": 0.6240351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 7.158553035791455e-06,
-      "loss": 0.8471,
+      "learning_rate": 6.541825997705611e-06,
+      "loss": 0.8745,
       "step": 21991
     },
     {
-      "epoch": 0.6039932987284062,
+      "epoch": 0.6240635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 7.1577001956173465e-06,
-      "loss": 0.7508,
+      "learning_rate": 6.540963639678215e-06,
+      "loss": 0.8721,
       "step": 21992
     },
     {
-      "epoch": 0.6040207629562495,
+      "epoch": 0.6240919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 7.156847377932614e-06,
-      "loss": 0.9775,
+      "learning_rate": 6.540101310869451e-06,
+      "loss": 0.8615,
       "step": 21993
     },
     {
-      "epoch": 0.6040482271840927,
+      "epoch": 0.6241203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 7.155994582744008e-06,
-      "loss": 0.8198,
+      "learning_rate": 6.539239011286611e-06,
+      "loss": 0.8791,
       "step": 21994
     },
     {
-      "epoch": 0.604075691411936,
+      "epoch": 0.6241486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 7.155141810058273e-06,
-      "loss": 0.8819,
+      "learning_rate": 6.538376740936972e-06,
+      "loss": 0.7399,
       "step": 21995
     },
     {
-      "epoch": 0.6041031556397792,
+      "epoch": 0.6241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 7.15428905988216e-06,
-      "loss": 0.9026,
+      "learning_rate": 6.537514499827822e-06,
+      "loss": 0.8382,
       "step": 21996
     },
     {
-      "epoch": 0.6041306198676224,
+      "epoch": 0.624205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 7.153436332222412e-06,
-      "loss": 0.871,
+      "learning_rate": 6.536652287966443e-06,
+      "loss": 0.9365,
       "step": 21997
     },
     {
-      "epoch": 0.6041580840954657,
+      "epoch": 0.6242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 7.152583627085781e-06,
-      "loss": 0.8729,
+      "learning_rate": 6.535790105360116e-06,
+      "loss": 0.9037,
       "step": 21998
     },
     {
-      "epoch": 0.6041855483233088,
+      "epoch": 0.6242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 7.151730944479008e-06,
-      "loss": 0.8502,
+      "learning_rate": 6.534927952016128e-06,
+      "loss": 0.7987,
       "step": 21999
     },
     {
-      "epoch": 0.6042130125511521,
+      "epoch": 0.6242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 7.150878284408844e-06,
-      "loss": 0.9742,
+      "learning_rate": 6.534065827941759e-06,
+      "loss": 0.8,
       "step": 22000
     },
     {
-      "epoch": 0.6042404767789954,
+      "epoch": 0.6243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 7.150025646882038e-06,
-      "loss": 0.8779,
+      "learning_rate": 6.53320373314429e-06,
+      "loss": 0.9473,
       "step": 22001
     },
     {
-      "epoch": 0.6042679410068386,
+      "epoch": 0.6243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 7.14917303190533e-06,
-      "loss": 0.9062,
+      "learning_rate": 6.532341667631006e-06,
+      "loss": 0.7542,
       "step": 22002
     },
     {
-      "epoch": 0.6042954052346818,
+      "epoch": 0.6243757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 7.14832043948547e-06,
-      "loss": 0.9822,
+      "learning_rate": 6.5314796314091885e-06,
+      "loss": 0.7164,
       "step": 22003
     },
     {
-      "epoch": 0.6043228694625251,
+      "epoch": 0.6244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 7.147467869629203e-06,
-      "loss": 0.8657,
+      "learning_rate": 6.5306176244861175e-06,
+      "loss": 0.9421,
       "step": 22004
     },
     {
-      "epoch": 0.6043503336903683,
+      "epoch": 0.6244324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 7.1466153223432734e-06,
-      "loss": 0.8541,
+      "learning_rate": 6.529755646869076e-06,
+      "loss": 0.8183,
       "step": 22005
     },
     {
-      "epoch": 0.6043777979182116,
+      "epoch": 0.624460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 7.145762797634433e-06,
-      "loss": 0.8361,
+      "learning_rate": 6.5288936985653455e-06,
+      "loss": 0.7902,
       "step": 22006
     },
     {
-      "epoch": 0.6044052621460547,
+      "epoch": 0.624489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 7.1449102955094214e-06,
-      "loss": 0.9001,
+      "learning_rate": 6.528031779582202e-06,
+      "loss": 0.8992,
       "step": 22007
     },
     {
-      "epoch": 0.604432726373898,
+      "epoch": 0.6245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 7.144057815974984e-06,
-      "loss": 0.8856,
+      "learning_rate": 6.52716988992693e-06,
+      "loss": 0.9358,
       "step": 22008
     },
     {
-      "epoch": 0.6044601906017413,
+      "epoch": 0.6245459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 7.143205359037867e-06,
-      "loss": 0.8747,
+      "learning_rate": 6.5263080296068134e-06,
+      "loss": 0.7867,
       "step": 22009
     },
     {
-      "epoch": 0.6044876548295844,
+      "epoch": 0.6245743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 7.142352924704818e-06,
-      "loss": 0.935,
+      "learning_rate": 6.52544619862913e-06,
+      "loss": 0.8409,
       "step": 22010
     },
     {
-      "epoch": 0.6045151190574277,
+      "epoch": 0.6246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 7.141500512982578e-06,
-      "loss": 0.7832,
+      "learning_rate": 6.524584397001155e-06,
+      "loss": 0.8961,
       "step": 22011
     },
     {
-      "epoch": 0.6045425832852709,
+      "epoch": 0.6246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 7.140648123877891e-06,
-      "loss": 0.8506,
+      "learning_rate": 6.523722624730175e-06,
+      "loss": 0.9226,
       "step": 22012
     },
     {
-      "epoch": 0.6045700475131142,
+      "epoch": 0.6246594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 7.13979575739751e-06,
-      "loss": 0.8617,
+      "learning_rate": 6.5228608818234665e-06,
+      "loss": 0.9106,
       "step": 22013
     },
     {
-      "epoch": 0.6045975117409574,
+      "epoch": 0.6246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 7.138943413548169e-06,
-      "loss": 0.7822,
+      "learning_rate": 6.521999168288308e-06,
+      "loss": 0.8416,
       "step": 22014
     },
     {
-      "epoch": 0.6046249759688006,
+      "epoch": 0.6247162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 7.138091092336617e-06,
-      "loss": 0.8479,
+      "learning_rate": 6.521137484131976e-06,
+      "loss": 0.8085,
       "step": 22015
     },
     {
-      "epoch": 0.6046524401966439,
+      "epoch": 0.6247446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 7.137238793769594e-06,
-      "loss": 0.9023,
+      "learning_rate": 6.5202758293617554e-06,
+      "loss": 0.9467,
       "step": 22016
     },
     {
-      "epoch": 0.6046799044244872,
+      "epoch": 0.6247729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 7.136386517853848e-06,
-      "loss": 0.8278,
+      "learning_rate": 6.519414203984922e-06,
+      "loss": 0.8792,
       "step": 22017
     },
     {
-      "epoch": 0.6047073686523303,
+      "epoch": 0.6248013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 7.135534264596123e-06,
-      "loss": 0.7957,
+      "learning_rate": 6.518552608008749e-06,
+      "loss": 0.8654,
       "step": 22018
     },
     {
-      "epoch": 0.6047348328801736,
+      "epoch": 0.6248297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 7.134682034003162e-06,
-      "loss": 0.8482,
+      "learning_rate": 6.517691041440522e-06,
+      "loss": 0.9341,
       "step": 22019
     },
     {
-      "epoch": 0.6047622971080168,
+      "epoch": 0.6248581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 7.133829826081703e-06,
-      "loss": 0.8838,
+      "learning_rate": 6.516829504287514e-06,
+      "loss": 0.9102,
       "step": 22020
     },
     {
-      "epoch": 0.60478976133586,
+      "epoch": 0.6248864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 7.132977640838492e-06,
-      "loss": 0.8408,
+      "learning_rate": 6.515967996557003e-06,
+      "loss": 0.7811,
       "step": 22021
     },
     {
-      "epoch": 0.6048172255637033,
+      "epoch": 0.6249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 7.132125478280275e-06,
-      "loss": 0.9511,
+      "learning_rate": 6.515106518256269e-06,
+      "loss": 0.8358,
       "step": 22022
     },
     {
-      "epoch": 0.6048446897915465,
+      "epoch": 0.6249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 7.13127333841379e-06,
-      "loss": 0.9005,
+      "learning_rate": 6.514245069392583e-06,
+      "loss": 0.8622,
       "step": 22023
     },
     {
-      "epoch": 0.6048721540193898,
+      "epoch": 0.6249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 7.130421221245782e-06,
-      "loss": 0.9139,
+      "learning_rate": 6.513383649973229e-06,
+      "loss": 0.8458,
       "step": 22024
     },
     {
-      "epoch": 0.6048996182472329,
+      "epoch": 0.625,
       "grad_norm": 0.0,
-      "learning_rate": 7.129569126782996e-06,
-      "loss": 0.8198,
+      "learning_rate": 6.512522260005478e-06,
+      "loss": 0.8791,
       "step": 22025
     },
     {
-      "epoch": 0.6049270824750762,
+      "epoch": 0.6250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 7.128717055032168e-06,
-      "loss": 0.7221,
+      "learning_rate": 6.51166089949661e-06,
+      "loss": 0.8644,
       "step": 22026
     },
     {
-      "epoch": 0.6049545467029195,
+      "epoch": 0.6250567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 7.127865006000045e-06,
-      "loss": 0.9076,
+      "learning_rate": 6.5107995684539e-06,
+      "loss": 0.8516,
       "step": 22027
     },
     {
-      "epoch": 0.6049820109307626,
+      "epoch": 0.6250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 7.127012979693364e-06,
-      "loss": 0.7972,
+      "learning_rate": 6.5099382668846215e-06,
+      "loss": 0.8843,
       "step": 22028
     },
     {
-      "epoch": 0.6050094751586059,
+      "epoch": 0.6251135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 7.126160976118869e-06,
-      "loss": 0.7657,
+      "learning_rate": 6.509076994796052e-06,
+      "loss": 0.8737,
       "step": 22029
     },
     {
-      "epoch": 0.6050369393864492,
+      "epoch": 0.6251418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 7.1253089952833045e-06,
-      "loss": 0.9493,
+      "learning_rate": 6.508215752195466e-06,
+      "loss": 0.866,
       "step": 22030
     },
     {
-      "epoch": 0.6050644036142924,
+      "epoch": 0.6251702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 7.12445703719341e-06,
-      "loss": 0.8786,
+      "learning_rate": 6.507354539090138e-06,
+      "loss": 0.8495,
       "step": 22031
     },
     {
-      "epoch": 0.6050918678421356,
+      "epoch": 0.6251986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 7.123605101855921e-06,
-      "loss": 0.8318,
+      "learning_rate": 6.506493355487345e-06,
+      "loss": 0.872,
       "step": 22032
     },
     {
-      "epoch": 0.6051193320699788,
+      "epoch": 0.6252270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 7.122753189277582e-06,
-      "loss": 0.8868,
+      "learning_rate": 6.505632201394358e-06,
+      "loss": 0.8134,
       "step": 22033
     },
     {
-      "epoch": 0.6051467962978221,
+      "epoch": 0.6252553916004541,
       "grad_norm": 0.0,
-      "learning_rate": 7.121901299465138e-06,
-      "loss": 0.9373,
+      "learning_rate": 6.504771076818451e-06,
+      "loss": 0.8517,
       "step": 22034
     },
     {
-      "epoch": 0.6051742605256654,
+      "epoch": 0.6252837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 7.121049432425323e-06,
-      "loss": 0.8854,
+      "learning_rate": 6.503909981766903e-06,
+      "loss": 0.8552,
       "step": 22035
     },
     {
-      "epoch": 0.6052017247535085,
+      "epoch": 0.6253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 7.120197588164881e-06,
-      "loss": 0.7962,
+      "learning_rate": 6.503048916246983e-06,
+      "loss": 0.8196,
       "step": 22036
     },
     {
-      "epoch": 0.6052291889813518,
+      "epoch": 0.6253405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 7.119345766690554e-06,
-      "loss": 0.8443,
+      "learning_rate": 6.502187880265969e-06,
+      "loss": 0.6911,
       "step": 22037
     },
     {
-      "epoch": 0.605256653209195,
+      "epoch": 0.6253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 7.118493968009077e-06,
-      "loss": 0.9533,
+      "learning_rate": 6.501326873831126e-06,
+      "loss": 0.9079,
       "step": 22038
     },
     {
-      "epoch": 0.6052841174370382,
+      "epoch": 0.6253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 7.117642192127191e-06,
-      "loss": 0.8078,
+      "learning_rate": 6.500465896949732e-06,
+      "loss": 0.8582,
       "step": 22039
     },
     {
-      "epoch": 0.6053115816648815,
+      "epoch": 0.6254256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 7.116790439051636e-06,
-      "loss": 0.8051,
+      "learning_rate": 6.499604949629064e-06,
+      "loss": 0.9003,
       "step": 22040
     },
     {
-      "epoch": 0.6053390458927247,
+      "epoch": 0.6254540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 7.115938708789152e-06,
-      "loss": 0.8945,
+      "learning_rate": 6.49874403187639e-06,
+      "loss": 0.9602,
       "step": 22041
     },
     {
-      "epoch": 0.605366510120568,
+      "epoch": 0.6254824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 7.11508700134648e-06,
-      "loss": 0.7719,
+      "learning_rate": 6.49788314369898e-06,
+      "loss": 0.7602,
       "step": 22042
     },
     {
-      "epoch": 0.6053939743484112,
+      "epoch": 0.625510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 7.11423531673036e-06,
-      "loss": 0.8744,
+      "learning_rate": 6.4970222851041106e-06,
+      "loss": 0.7972,
       "step": 22043
     },
     {
-      "epoch": 0.6054214385762544,
+      "epoch": 0.625539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 7.113383654947523e-06,
-      "loss": 0.8819,
+      "learning_rate": 6.496161456099052e-06,
+      "loss": 0.7805,
       "step": 22044
     },
     {
-      "epoch": 0.6054489028040977,
+      "epoch": 0.6255675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 7.112532016004711e-06,
-      "loss": 0.9357,
+      "learning_rate": 6.495300656691072e-06,
+      "loss": 0.8552,
       "step": 22045
     },
     {
-      "epoch": 0.6054763670319409,
+      "epoch": 0.6255959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 7.11168039990867e-06,
-      "loss": 0.752,
+      "learning_rate": 6.494439886887448e-06,
+      "loss": 0.8336,
       "step": 22046
     },
     {
-      "epoch": 0.6055038312597841,
+      "epoch": 0.6256242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 7.110828806666128e-06,
-      "loss": 0.8807,
+      "learning_rate": 6.493579146695448e-06,
+      "loss": 0.8722,
       "step": 22047
     },
     {
-      "epoch": 0.6055312954876274,
+      "epoch": 0.6256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 7.109977236283829e-06,
-      "loss": 0.9572,
+      "learning_rate": 6.49271843612234e-06,
+      "loss": 0.9042,
       "step": 22048
     },
     {
-      "epoch": 0.6055587597154706,
+      "epoch": 0.6256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 7.109125688768512e-06,
-      "loss": 0.8157,
+      "learning_rate": 6.491857755175399e-06,
+      "loss": 0.8153,
       "step": 22049
     },
     {
-      "epoch": 0.6055862239433139,
+      "epoch": 0.6257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 7.108274164126912e-06,
-      "loss": 0.8842,
+      "learning_rate": 6.490997103861894e-06,
+      "loss": 0.7897,
       "step": 22050
     },
     {
-      "epoch": 0.605613688171157,
+      "epoch": 0.6257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 7.107422662365764e-06,
-      "loss": 0.8329,
+      "learning_rate": 6.490136482189091e-06,
+      "loss": 0.9839,
       "step": 22051
     },
     {
-      "epoch": 0.6056411523990003,
+      "epoch": 0.6257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 7.1065711834918085e-06,
-      "loss": 0.8442,
+      "learning_rate": 6.489275890164265e-06,
+      "loss": 0.8987,
       "step": 22052
     },
     {
-      "epoch": 0.6056686166268436,
+      "epoch": 0.625794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 7.105719727511782e-06,
-      "loss": 0.9077,
+      "learning_rate": 6.4884153277946836e-06,
+      "loss": 0.8486,
       "step": 22053
     },
     {
-      "epoch": 0.6056960808546867,
+      "epoch": 0.6258229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 7.104868294432425e-06,
-      "loss": 0.8556,
+      "learning_rate": 6.487554795087612e-06,
+      "loss": 0.9037,
       "step": 22054
     },
     {
-      "epoch": 0.60572354508253,
+      "epoch": 0.6258513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 7.104016884260473e-06,
-      "loss": 0.8836,
+      "learning_rate": 6.4866942920503274e-06,
+      "loss": 0.8293,
       "step": 22055
     },
     {
-      "epoch": 0.6057510093103733,
+      "epoch": 0.6258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 7.103165497002658e-06,
-      "loss": 0.8928,
+      "learning_rate": 6.485833818690092e-06,
+      "loss": 0.9354,
       "step": 22056
     },
     {
-      "epoch": 0.6057784735382165,
+      "epoch": 0.6259080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 7.1023141326657185e-06,
-      "loss": 0.8435,
+      "learning_rate": 6.484973375014177e-06,
+      "loss": 0.9177,
       "step": 22057
     },
     {
-      "epoch": 0.6058059377660597,
+      "epoch": 0.6259364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 7.101462791256395e-06,
-      "loss": 0.8037,
+      "learning_rate": 6.484112961029851e-06,
+      "loss": 0.869,
       "step": 22058
     },
     {
-      "epoch": 0.6058334019939029,
+      "epoch": 0.6259648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 7.100611472781418e-06,
-      "loss": 0.8717,
+      "learning_rate": 6.483252576744379e-06,
+      "loss": 0.9366,
       "step": 22059
     },
     {
-      "epoch": 0.6058608662217462,
+      "epoch": 0.6259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 7.099760177247527e-06,
-      "loss": 0.8301,
+      "learning_rate": 6.4823922221650324e-06,
+      "loss": 0.8288,
       "step": 22060
     },
     {
-      "epoch": 0.6058883304495895,
+      "epoch": 0.6260215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 7.098908904661462e-06,
-      "loss": 0.7847,
+      "learning_rate": 6.481531897299076e-06,
+      "loss": 0.884,
       "step": 22061
     },
     {
-      "epoch": 0.6059157946774326,
+      "epoch": 0.626049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 7.098057655029949e-06,
-      "loss": 0.7955,
+      "learning_rate": 6.480671602153778e-06,
+      "loss": 0.7447,
       "step": 22062
     },
     {
-      "epoch": 0.6059432589052759,
+      "epoch": 0.6260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 7.097206428359726e-06,
-      "loss": 0.8492,
+      "learning_rate": 6.479811336736406e-06,
+      "loss": 0.8781,
       "step": 22063
     },
     {
-      "epoch": 0.6059707231331191,
+      "epoch": 0.6261066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 7.096355224657532e-06,
-      "loss": 0.8887,
+      "learning_rate": 6.478951101054225e-06,
+      "loss": 0.8362,
       "step": 22064
     },
     {
-      "epoch": 0.6059981873609623,
+      "epoch": 0.6261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 7.095504043930099e-06,
-      "loss": 0.801,
+      "learning_rate": 6.478090895114501e-06,
+      "loss": 0.7631,
       "step": 22065
     },
     {
-      "epoch": 0.6060256515888056,
+      "epoch": 0.6261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 7.094652886184166e-06,
-      "loss": 0.9689,
+      "learning_rate": 6.477230718924503e-06,
+      "loss": 0.8213,
       "step": 22066
     },
     {
-      "epoch": 0.6060531158166488,
+      "epoch": 0.6261918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 7.093801751426466e-06,
-      "loss": 0.8116,
+      "learning_rate": 6.476370572491496e-06,
+      "loss": 0.7247,
       "step": 22067
     },
     {
-      "epoch": 0.6060805800444921,
+      "epoch": 0.6262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 7.092950639663728e-06,
-      "loss": 0.8513,
+      "learning_rate": 6.475510455822743e-06,
+      "loss": 0.9434,
       "step": 22068
     },
     {
-      "epoch": 0.6061080442723353,
+      "epoch": 0.6262485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 7.092099550902691e-06,
-      "loss": 0.8632,
+      "learning_rate": 6.47465036892551e-06,
+      "loss": 0.8447,
       "step": 22069
     },
     {
-      "epoch": 0.6061355085001785,
+      "epoch": 0.6262769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 7.09124848515009e-06,
-      "loss": 0.9172,
+      "learning_rate": 6.473790311807066e-06,
+      "loss": 0.8735,
       "step": 22070
     },
     {
-      "epoch": 0.6061629727280218,
+      "epoch": 0.6263053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 7.090397442412657e-06,
-      "loss": 0.8439,
+      "learning_rate": 6.472930284474677e-06,
+      "loss": 0.9474,
       "step": 22071
     },
     {
-      "epoch": 0.6061904369558649,
+      "epoch": 0.62633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 7.0895464226971315e-06,
-      "loss": 0.8542,
+      "learning_rate": 6.4720702869356015e-06,
+      "loss": 0.8097,
       "step": 22072
     },
     {
-      "epoch": 0.6062179011837082,
+      "epoch": 0.6263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 7.088695426010237e-06,
-      "loss": 0.9788,
+      "learning_rate": 6.471210319197108e-06,
+      "loss": 0.8951,
       "step": 22073
     },
     {
-      "epoch": 0.6062453654115515,
+      "epoch": 0.6263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 7.087844452358713e-06,
-      "loss": 0.8291,
+      "learning_rate": 6.470350381266459e-06,
+      "loss": 0.8428,
       "step": 22074
     },
     {
-      "epoch": 0.6062728296393947,
+      "epoch": 0.6264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 7.0869935017492905e-06,
-      "loss": 0.9349,
+      "learning_rate": 6.469490473150917e-06,
+      "loss": 0.9504,
       "step": 22075
     },
     {
-      "epoch": 0.6063002938672379,
+      "epoch": 0.6264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 7.086142574188703e-06,
-      "loss": 0.7666,
+      "learning_rate": 6.46863059485775e-06,
+      "loss": 0.8927,
       "step": 22076
     },
     {
-      "epoch": 0.6063277580950811,
+      "epoch": 0.6264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 7.085291669683687e-06,
-      "loss": 0.782,
+      "learning_rate": 6.46777074639422e-06,
+      "loss": 0.8062,
       "step": 22077
     },
     {
-      "epoch": 0.6063552223229244,
+      "epoch": 0.6265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 7.084440788240974e-06,
-      "loss": 0.8575,
+      "learning_rate": 6.466910927767589e-06,
+      "loss": 0.8218,
       "step": 22078
     },
     {
-      "epoch": 0.6063826865507677,
+      "epoch": 0.6265323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 7.083589929867293e-06,
-      "loss": 0.8486,
+      "learning_rate": 6.466051138985117e-06,
+      "loss": 0.8451,
       "step": 22079
     },
     {
-      "epoch": 0.6064101507786108,
+      "epoch": 0.626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 7.082739094569375e-06,
-      "loss": 0.9535,
+      "learning_rate": 6.465191380054075e-06,
+      "loss": 0.7443,
       "step": 22080
     },
     {
-      "epoch": 0.6064376150064541,
+      "epoch": 0.6265891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 7.081888282353957e-06,
-      "loss": 0.8246,
+      "learning_rate": 6.464331650981717e-06,
+      "loss": 0.837,
       "step": 22081
     },
     {
-      "epoch": 0.6064650792342974,
+      "epoch": 0.6266174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 7.081037493227771e-06,
-      "loss": 0.949,
+      "learning_rate": 6.4634719517753075e-06,
+      "loss": 0.8255,
       "step": 22082
     },
     {
-      "epoch": 0.6064925434621405,
+      "epoch": 0.6266458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 7.080186727197545e-06,
-      "loss": 0.7609,
+      "learning_rate": 6.4626122824421114e-06,
+      "loss": 0.8988,
       "step": 22083
     },
     {
-      "epoch": 0.6065200076899838,
+      "epoch": 0.6266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 7.079335984270017e-06,
-      "loss": 0.932,
+      "learning_rate": 6.461752642989389e-06,
+      "loss": 0.8489,
       "step": 22084
     },
     {
-      "epoch": 0.606547471917827,
+      "epoch": 0.6267026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 7.078485264451911e-06,
-      "loss": 0.8858,
+      "learning_rate": 6.4608930334243956e-06,
+      "loss": 0.8785,
       "step": 22085
     },
     {
-      "epoch": 0.6065749361456703,
+      "epoch": 0.6267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 7.077634567749962e-06,
-      "loss": 0.9148,
+      "learning_rate": 6.4600334537544015e-06,
+      "loss": 0.8077,
       "step": 22086
     },
     {
-      "epoch": 0.6066024003735135,
+      "epoch": 0.6267593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 7.076783894170897e-06,
-      "loss": 0.8891,
+      "learning_rate": 6.459173903986665e-06,
+      "loss": 0.8618,
       "step": 22087
     },
     {
-      "epoch": 0.6066298646013567,
+      "epoch": 0.6267877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 7.0759332437214526e-06,
-      "loss": 0.9629,
+      "learning_rate": 6.458314384128447e-06,
+      "loss": 0.8525,
       "step": 22088
     },
     {
-      "epoch": 0.6066573288292,
+      "epoch": 0.6268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 7.075082616408358e-06,
-      "loss": 0.9083,
+      "learning_rate": 6.457454894187003e-06,
+      "loss": 0.8883,
       "step": 22089
     },
     {
-      "epoch": 0.6066847930570431,
+      "epoch": 0.626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 7.074232012238342e-06,
-      "loss": 0.9308,
+      "learning_rate": 6.456595434169599e-06,
+      "loss": 0.8127,
       "step": 22090
     },
     {
-      "epoch": 0.6067122572848864,
+      "epoch": 0.6268728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 7.073381431218137e-06,
-      "loss": 0.8516,
+      "learning_rate": 6.455736004083494e-06,
+      "loss": 0.9352,
       "step": 22091
     },
     {
-      "epoch": 0.6067397215127297,
+      "epoch": 0.6269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 7.072530873354468e-06,
-      "loss": 0.9107,
+      "learning_rate": 6.454876603935942e-06,
+      "loss": 0.8081,
       "step": 22092
     },
     {
-      "epoch": 0.6067671857405729,
+      "epoch": 0.6269296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 7.07168033865407e-06,
-      "loss": 0.9018,
+      "learning_rate": 6.454017233734211e-06,
+      "loss": 0.7631,
       "step": 22093
     },
     {
-      "epoch": 0.6067946499684161,
+      "epoch": 0.6269580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 7.0708298271236744e-06,
-      "loss": 0.9532,
+      "learning_rate": 6.453157893485556e-06,
+      "loss": 0.924,
       "step": 22094
     },
     {
-      "epoch": 0.6068221141962594,
+      "epoch": 0.6269863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 7.069979338770004e-06,
-      "loss": 0.8424,
+      "learning_rate": 6.452298583197232e-06,
+      "loss": 0.8332,
       "step": 22095
     },
     {
-      "epoch": 0.6068495784241026,
+      "epoch": 0.6270147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 7.069128873599797e-06,
-      "loss": 0.8974,
+      "learning_rate": 6.451439302876503e-06,
+      "loss": 0.8832,
       "step": 22096
     },
     {
-      "epoch": 0.6068770426519459,
+      "epoch": 0.6270431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 7.068278431619774e-06,
-      "loss": 0.908,
+      "learning_rate": 6.450580052530626e-06,
+      "loss": 0.8557,
       "step": 22097
     },
     {
-      "epoch": 0.606904506879789,
+      "epoch": 0.6270715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 7.067428012836669e-06,
-      "loss": 0.8327,
+      "learning_rate": 6.449720832166859e-06,
+      "loss": 0.7948,
       "step": 22098
     },
     {
-      "epoch": 0.6069319711076323,
+      "epoch": 0.627099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 7.066577617257209e-06,
-      "loss": 0.8333,
+      "learning_rate": 6.44886164179246e-06,
+      "loss": 0.8953,
       "step": 22099
     },
     {
-      "epoch": 0.6069594353354756,
+      "epoch": 0.6271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 7.065727244888121e-06,
-      "loss": 0.7463,
+      "learning_rate": 6.448002481414681e-06,
+      "loss": 0.83,
       "step": 22100
     },
     {
-      "epoch": 0.6069868995633187,
+      "epoch": 0.6271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 7.064876895736139e-06,
-      "loss": 0.8335,
+      "learning_rate": 6.447143351040792e-06,
+      "loss": 0.7251,
       "step": 22101
     },
     {
-      "epoch": 0.607014363791162,
+      "epoch": 0.6271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 7.064026569807987e-06,
-      "loss": 0.9114,
+      "learning_rate": 6.446284250678041e-06,
+      "loss": 0.8189,
       "step": 22102
     },
     {
-      "epoch": 0.6070418280190052,
+      "epoch": 0.6272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 7.063176267110395e-06,
-      "loss": 0.9643,
+      "learning_rate": 6.445425180333684e-06,
+      "loss": 0.9162,
       "step": 22103
     },
     {
-      "epoch": 0.6070692922468485,
+      "epoch": 0.6272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 7.062325987650087e-06,
-      "loss": 0.8604,
+      "learning_rate": 6.444566140014983e-06,
+      "loss": 0.9149,
       "step": 22104
     },
     {
-      "epoch": 0.6070967564746917,
+      "epoch": 0.6272701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 7.061475731433795e-06,
-      "loss": 0.833,
+      "learning_rate": 6.443707129729192e-06,
+      "loss": 0.8082,
       "step": 22105
     },
     {
-      "epoch": 0.6071242207025349,
+      "epoch": 0.6272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 7.060625498468246e-06,
-      "loss": 0.9346,
+      "learning_rate": 6.442848149483565e-06,
+      "loss": 0.7504,
       "step": 22106
     },
     {
-      "epoch": 0.6071516849303782,
+      "epoch": 0.6273269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 7.059775288760165e-06,
-      "loss": 0.9107,
+      "learning_rate": 6.441989199285361e-06,
+      "loss": 0.893,
       "step": 22107
     },
     {
-      "epoch": 0.6071791491582215,
+      "epoch": 0.627355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 7.058925102316286e-06,
-      "loss": 0.7804,
+      "learning_rate": 6.441130279141834e-06,
+      "loss": 0.7832,
       "step": 22108
     },
     {
-      "epoch": 0.6072066133860646,
+      "epoch": 0.627383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 7.0580749391433265e-06,
-      "loss": 0.8396,
+      "learning_rate": 6.440271389060238e-06,
+      "loss": 0.849,
       "step": 22109
     },
     {
-      "epoch": 0.6072340776139079,
+      "epoch": 0.6274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 7.0572247992480195e-06,
-      "loss": 0.7843,
+      "learning_rate": 6.439412529047831e-06,
+      "loss": 0.9526,
       "step": 22110
     },
     {
-      "epoch": 0.6072615418417511,
+      "epoch": 0.6274404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 7.056374682637086e-06,
-      "loss": 0.8028,
+      "learning_rate": 6.438553699111867e-06,
+      "loss": 0.8603,
       "step": 22111
     },
     {
-      "epoch": 0.6072890060695944,
+      "epoch": 0.6274687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 7.0555245893172585e-06,
-      "loss": 0.8058,
+      "learning_rate": 6.437694899259597e-06,
+      "loss": 0.7393,
       "step": 22112
     },
     {
-      "epoch": 0.6073164702974376,
+      "epoch": 0.6274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 7.054674519295261e-06,
-      "loss": 0.8444,
+      "learning_rate": 6.4368361294982816e-06,
+      "loss": 0.8306,
       "step": 22113
     },
     {
-      "epoch": 0.6073439345252808,
+      "epoch": 0.6275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 7.053824472577821e-06,
-      "loss": 0.8876,
+      "learning_rate": 6.435977389835171e-06,
+      "loss": 0.8438,
       "step": 22114
     },
     {
-      "epoch": 0.6073713987531241,
+      "epoch": 0.6275539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 7.052974449171663e-06,
-      "loss": 0.8738,
+      "learning_rate": 6.4351186802775155e-06,
+      "loss": 0.9876,
       "step": 22115
     },
     {
-      "epoch": 0.6073988629809672,
+      "epoch": 0.6275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 7.052124449083511e-06,
-      "loss": 0.899,
+      "learning_rate": 6.434260000832575e-06,
+      "loss": 0.8833,
       "step": 22116
     },
     {
-      "epoch": 0.6074263272088105,
+      "epoch": 0.6276106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 7.051274472320093e-06,
-      "loss": 0.7766,
+      "learning_rate": 6.433401351507602e-06,
+      "loss": 0.9991,
       "step": 22117
     },
     {
-      "epoch": 0.6074537914366538,
+      "epoch": 0.6276390465380249,
       "grad_norm": 0.0,
-      "learning_rate": 7.050424518888135e-06,
-      "loss": 0.9327,
+      "learning_rate": 6.43254273230985e-06,
+      "loss": 0.8318,
       "step": 22118
     },
     {
-      "epoch": 0.607481255664497,
+      "epoch": 0.6276674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 7.049574588794357e-06,
-      "loss": 0.8658,
+      "learning_rate": 6.431684143246568e-06,
+      "loss": 0.8035,
       "step": 22119
     },
     {
-      "epoch": 0.6075087198923402,
+      "epoch": 0.6276958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 7.048724682045494e-06,
-      "loss": 0.8577,
+      "learning_rate": 6.430825584325009e-06,
+      "loss": 0.8388,
       "step": 22120
     },
     {
-      "epoch": 0.6075361841201835,
+      "epoch": 0.6277241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 7.04787479864826e-06,
-      "loss": 0.8332,
+      "learning_rate": 6.429967055552429e-06,
+      "loss": 0.8277,
       "step": 22121
     },
     {
-      "epoch": 0.6075636483480267,
+      "epoch": 0.6277525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 7.047024938609384e-06,
-      "loss": 0.8023,
+      "learning_rate": 6.4291085569360765e-06,
+      "loss": 0.7572,
       "step": 22122
     },
     {
-      "epoch": 0.60759111257587,
+      "epoch": 0.6277809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 7.046175101935591e-06,
-      "loss": 0.8543,
+      "learning_rate": 6.4282500884832045e-06,
+      "loss": 0.736,
       "step": 22123
     },
     {
-      "epoch": 0.6076185768037131,
+      "epoch": 0.6278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 7.045325288633605e-06,
-      "loss": 0.907,
+      "learning_rate": 6.427391650201065e-06,
+      "loss": 0.8538,
       "step": 22124
     },
     {
-      "epoch": 0.6076460410315564,
+      "epoch": 0.6278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 7.044475498710151e-06,
-      "loss": 0.786,
+      "learning_rate": 6.426533242096911e-06,
+      "loss": 0.912,
       "step": 22125
     },
     {
-      "epoch": 0.6076735052593997,
+      "epoch": 0.6278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 7.043625732171951e-06,
-      "loss": 0.9643,
+      "learning_rate": 6.425674864177987e-06,
+      "loss": 0.8525,
       "step": 22126
     },
     {
-      "epoch": 0.6077009694872428,
+      "epoch": 0.627894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.04277598902573e-06,
-      "loss": 0.8773,
+      "learning_rate": 6.424816516451551e-06,
+      "loss": 0.8309,
       "step": 22127
     },
     {
-      "epoch": 0.6077284337150861,
+      "epoch": 0.6279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 7.041926269278208e-06,
-      "loss": 0.9219,
+      "learning_rate": 6.42395819892485e-06,
+      "loss": 0.8102,
       "step": 22128
     },
     {
-      "epoch": 0.6077558979429293,
+      "epoch": 0.6279511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 7.0410765729361115e-06,
-      "loss": 0.8857,
+      "learning_rate": 6.423099911605133e-06,
+      "loss": 0.8703,
       "step": 22129
     },
     {
-      "epoch": 0.6077833621707726,
+      "epoch": 0.6279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 7.040226900006166e-06,
-      "loss": 0.8462,
+      "learning_rate": 6.422241654499654e-06,
+      "loss": 0.8011,
       "step": 22130
     },
     {
-      "epoch": 0.6078108263986158,
+      "epoch": 0.6280079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 7.039377250495094e-06,
-      "loss": 0.853,
+      "learning_rate": 6.421383427615658e-06,
+      "loss": 0.68,
       "step": 22131
     },
     {
-      "epoch": 0.607838290626459,
+      "epoch": 0.6280363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 7.038527624409612e-06,
-      "loss": 0.7581,
+      "learning_rate": 6.420525230960399e-06,
+      "loss": 0.7879,
       "step": 22132
     },
     {
-      "epoch": 0.6078657548543023,
+      "epoch": 0.6280646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 7.037678021756446e-06,
-      "loss": 0.8615,
+      "learning_rate": 6.419667064541123e-06,
+      "loss": 0.892,
       "step": 22133
     },
     {
-      "epoch": 0.6078932190821456,
+      "epoch": 0.6280930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 7.036828442542323e-06,
-      "loss": 0.9736,
+      "learning_rate": 6.418808928365083e-06,
+      "loss": 0.8525,
       "step": 22134
     },
     {
-      "epoch": 0.6079206833099887,
+      "epoch": 0.6281214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 7.035978886773957e-06,
-      "loss": 0.8471,
+      "learning_rate": 6.417950822439524e-06,
+      "loss": 0.8526,
       "step": 22135
     },
     {
-      "epoch": 0.607948147537832,
+      "epoch": 0.6281498297389331,
       "grad_norm": 0.0,
-      "learning_rate": 7.035129354458075e-06,
-      "loss": 0.8763,
+      "learning_rate": 6.417092746771693e-06,
+      "loss": 0.8351,
       "step": 22136
     },
     {
-      "epoch": 0.6079756117656752,
+      "epoch": 0.6281782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 7.034279845601402e-06,
-      "loss": 0.8382,
+      "learning_rate": 6.416234701368844e-06,
+      "loss": 0.8845,
       "step": 22137
     },
     {
-      "epoch": 0.6080030759935184,
+      "epoch": 0.6282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 7.03343036021065e-06,
-      "loss": 0.7748,
+      "learning_rate": 6.41537668623822e-06,
+      "loss": 0.9374,
       "step": 22138
     },
     {
-      "epoch": 0.6080305402213617,
+      "epoch": 0.6282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 7.032580898292549e-06,
-      "loss": 0.9285,
+      "learning_rate": 6.414518701387069e-06,
+      "loss": 0.9329,
       "step": 22139
     },
     {
-      "epoch": 0.6080580044492049,
+      "epoch": 0.6282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 7.031731459853817e-06,
-      "loss": 0.9142,
+      "learning_rate": 6.413660746822643e-06,
+      "loss": 0.7615,
       "step": 22140
     },
     {
-      "epoch": 0.6080854686770482,
+      "epoch": 0.6282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 7.030882044901174e-06,
-      "loss": 0.8423,
+      "learning_rate": 6.412802822552184e-06,
+      "loss": 0.7844,
       "step": 22141
     },
     {
-      "epoch": 0.6081129329048914,
+      "epoch": 0.6283200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 7.030032653441345e-06,
-      "loss": 0.9005,
+      "learning_rate": 6.411944928582941e-06,
+      "loss": 0.8631,
       "step": 22142
     },
     {
-      "epoch": 0.6081403971327346,
+      "epoch": 0.6283484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 7.029183285481049e-06,
-      "loss": 0.8854,
+      "learning_rate": 6.4110870649221605e-06,
+      "loss": 0.9153,
       "step": 22143
     },
     {
-      "epoch": 0.6081678613605779,
+      "epoch": 0.6283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 7.028333941027004e-06,
-      "loss": 0.8951,
+      "learning_rate": 6.410229231577089e-06,
+      "loss": 0.8927,
       "step": 22144
     },
     {
-      "epoch": 0.608195325588421,
+      "epoch": 0.628405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 7.0274846200859295e-06,
-      "loss": 0.893,
+      "learning_rate": 6.4093714285549736e-06,
+      "loss": 0.7762,
       "step": 22145
     },
     {
-      "epoch": 0.6082227898162643,
+      "epoch": 0.628433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 7.026635322664552e-06,
-      "loss": 0.9022,
+      "learning_rate": 6.408513655863054e-06,
+      "loss": 0.8963,
       "step": 22146
     },
     {
-      "epoch": 0.6082502540441076,
+      "epoch": 0.6284619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 7.025786048769585e-06,
-      "loss": 0.8845,
+      "learning_rate": 6.407655913508583e-06,
+      "loss": 0.874,
       "step": 22147
     },
     {
-      "epoch": 0.6082777182719508,
+      "epoch": 0.6284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 7.02493679840775e-06,
-      "loss": 0.8147,
+      "learning_rate": 6.4067982014988075e-06,
+      "loss": 0.6956,
       "step": 22148
     },
     {
-      "epoch": 0.608305182499794,
+      "epoch": 0.6285187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 7.024087571585772e-06,
-      "loss": 0.8629,
+      "learning_rate": 6.405940519840967e-06,
+      "loss": 0.9509,
       "step": 22149
     },
     {
-      "epoch": 0.6083326467276372,
+      "epoch": 0.6285471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 7.023238368310362e-06,
-      "loss": 0.8375,
+      "learning_rate": 6.4050828685423094e-06,
+      "loss": 0.8449,
       "step": 22150
     },
     {
-      "epoch": 0.6083601109554805,
+      "epoch": 0.6285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 7.022389188588246e-06,
-      "loss": 0.9048,
+      "learning_rate": 6.404225247610078e-06,
+      "loss": 0.9135,
       "step": 22151
     },
     {
-      "epoch": 0.6083875751833238,
+      "epoch": 0.6286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 7.021540032426136e-06,
-      "loss": 0.8294,
+      "learning_rate": 6.403367657051518e-06,
+      "loss": 0.8725,
       "step": 22152
     },
     {
-      "epoch": 0.6084150394111669,
+      "epoch": 0.6286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 7.020690899830757e-06,
-      "loss": 0.9216,
+      "learning_rate": 6.4025100968738715e-06,
+      "loss": 0.8676,
       "step": 22153
     },
     {
-      "epoch": 0.6084425036390102,
+      "epoch": 0.6286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 7.019841790808827e-06,
-      "loss": 0.8275,
+      "learning_rate": 6.401652567084386e-06,
+      "loss": 0.8593,
       "step": 22154
     },
     {
-      "epoch": 0.6084699678668535,
+      "epoch": 0.628688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 7.0189927053670644e-06,
-      "loss": 0.7291,
+      "learning_rate": 6.400795067690303e-06,
+      "loss": 0.7749,
       "step": 22155
     },
     {
-      "epoch": 0.6084974320946966,
+      "epoch": 0.6287173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 7.018143643512183e-06,
-      "loss": 0.7617,
+      "learning_rate": 6.399937598698865e-06,
+      "loss": 0.9548,
       "step": 22156
     },
     {
-      "epoch": 0.6085248963225399,
+      "epoch": 0.6287457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 7.017294605250904e-06,
-      "loss": 0.7994,
+      "learning_rate": 6.399080160117314e-06,
+      "loss": 0.8771,
       "step": 22157
     },
     {
-      "epoch": 0.6085523605503831,
+      "epoch": 0.6287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 7.016445590589949e-06,
-      "loss": 0.8569,
+      "learning_rate": 6.3982227519528986e-06,
+      "loss": 0.9594,
       "step": 22158
     },
     {
-      "epoch": 0.6085798247782264,
+      "epoch": 0.6288024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 7.015596599536027e-06,
-      "loss": 0.8106,
+      "learning_rate": 6.397365374212854e-06,
+      "loss": 0.7335,
       "step": 22159
     },
     {
-      "epoch": 0.6086072890060696,
+      "epoch": 0.6288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 7.014747632095863e-06,
-      "loss": 0.9294,
+      "learning_rate": 6.3965080269044285e-06,
+      "loss": 0.7917,
       "step": 22160
     },
     {
-      "epoch": 0.6086347532339128,
+      "epoch": 0.6288592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 7.013898688276176e-06,
-      "loss": 0.8358,
+      "learning_rate": 6.395650710034858e-06,
+      "loss": 0.8578,
       "step": 22161
     },
     {
-      "epoch": 0.6086622174617561,
+      "epoch": 0.6288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 7.013049768083675e-06,
-      "loss": 0.8631,
+      "learning_rate": 6.3947934236113915e-06,
+      "loss": 0.8864,
       "step": 22162
     },
     {
-      "epoch": 0.6086896816895992,
+      "epoch": 0.6289160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 7.012200871525083e-06,
-      "loss": 0.8748,
+      "learning_rate": 6.393936167641266e-06,
+      "loss": 0.8579,
       "step": 22163
     },
     {
-      "epoch": 0.6087171459174425,
+      "epoch": 0.62894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.0113519986071135e-06,
-      "loss": 0.7673,
+      "learning_rate": 6.393078942131723e-06,
+      "loss": 0.799,
       "step": 22164
     },
     {
-      "epoch": 0.6087446101452858,
+      "epoch": 0.6289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 7.010503149336484e-06,
-      "loss": 0.8567,
+      "learning_rate": 6.392221747090006e-06,
+      "loss": 0.8735,
       "step": 22165
     },
     {
-      "epoch": 0.608772074373129,
+      "epoch": 0.6290011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 7.009654323719914e-06,
-      "loss": 0.9052,
+      "learning_rate": 6.391364582523355e-06,
+      "loss": 0.763,
       "step": 22166
     },
     {
-      "epoch": 0.6087995386009722,
+      "epoch": 0.6290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 7.008805521764119e-06,
-      "loss": 0.9427,
+      "learning_rate": 6.390507448439007e-06,
+      "loss": 0.7325,
       "step": 22167
     },
     {
-      "epoch": 0.6088270028288155,
+      "epoch": 0.6290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 7.00795674347581e-06,
-      "loss": 0.8655,
+      "learning_rate": 6.389650344844206e-06,
+      "loss": 0.8667,
       "step": 22168
     },
     {
-      "epoch": 0.6088544670566587,
+      "epoch": 0.6290862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 7.007107988861707e-06,
-      "loss": 0.8493,
+      "learning_rate": 6.388793271746192e-06,
+      "loss": 0.8715,
       "step": 22169
     },
     {
-      "epoch": 0.608881931284502,
+      "epoch": 0.6291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 7.006259257928525e-06,
-      "loss": 0.8206,
+      "learning_rate": 6.3879362291522e-06,
+      "loss": 0.8365,
       "step": 22170
     },
     {
-      "epoch": 0.6089093955123451,
+      "epoch": 0.6291430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 7.005410550682978e-06,
-      "loss": 0.9046,
+      "learning_rate": 6.387079217069476e-06,
+      "loss": 0.889,
       "step": 22171
     },
     {
-      "epoch": 0.6089368597401884,
+      "epoch": 0.6291713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 7.004561867131784e-06,
-      "loss": 0.8194,
+      "learning_rate": 6.386222235505257e-06,
+      "loss": 0.9462,
       "step": 22172
     },
     {
-      "epoch": 0.6089643239680317,
+      "epoch": 0.629199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 7.003713207281659e-06,
-      "loss": 0.7846,
+      "learning_rate": 6.385365284466779e-06,
+      "loss": 0.9829,
       "step": 22173
     },
     {
-      "epoch": 0.6089917881958749,
+      "epoch": 0.629228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 7.002864571139314e-06,
-      "loss": 0.8617,
+      "learning_rate": 6.384508363961284e-06,
+      "loss": 0.8425,
       "step": 22174
     },
     {
-      "epoch": 0.6090192524237181,
+      "epoch": 0.6292565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 7.002015958711466e-06,
-      "loss": 0.9338,
+      "learning_rate": 6.383651473996011e-06,
+      "loss": 0.747,
       "step": 22175
     },
     {
-      "epoch": 0.6090467166515613,
+      "epoch": 0.6292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 7.001167370004826e-06,
-      "loss": 0.9004,
+      "learning_rate": 6.382794614578193e-06,
+      "loss": 0.8527,
       "step": 22176
     },
     {
-      "epoch": 0.6090741808794046,
+      "epoch": 0.6293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 7.000318805026114e-06,
-      "loss": 1.014,
+      "learning_rate": 6.381937785715069e-06,
+      "loss": 0.9043,
       "step": 22177
     },
     {
-      "epoch": 0.6091016451072478,
+      "epoch": 0.6293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 6.9994702637820425e-06,
-      "loss": 0.8384,
+      "learning_rate": 6.381080987413884e-06,
+      "loss": 0.8114,
       "step": 22178
     },
     {
-      "epoch": 0.609129109335091,
+      "epoch": 0.6293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 6.998621746279326e-06,
-      "loss": 0.8527,
+      "learning_rate": 6.380224219681872e-06,
+      "loss": 0.8454,
       "step": 22179
     },
     {
-      "epoch": 0.6091565735629343,
+      "epoch": 0.6293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 6.9977732525246735e-06,
-      "loss": 0.7924,
+      "learning_rate": 6.379367482526264e-06,
+      "loss": 0.7773,
       "step": 22180
     },
     {
-      "epoch": 0.6091840377907776,
+      "epoch": 0.6294267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 6.996924782524801e-06,
-      "loss": 0.8134,
+      "learning_rate": 6.378510775954305e-06,
+      "loss": 0.9052,
       "step": 22181
     },
     {
-      "epoch": 0.6092115020186207,
+      "epoch": 0.6294551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 6.996076336286426e-06,
-      "loss": 0.9099,
+      "learning_rate": 6.377654099973229e-06,
+      "loss": 0.885,
       "step": 22182
     },
     {
-      "epoch": 0.609238966246464,
+      "epoch": 0.6294835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 6.995227913816256e-06,
-      "loss": 0.8718,
+      "learning_rate": 6.376797454590269e-06,
+      "loss": 0.7931,
       "step": 22183
     },
     {
-      "epoch": 0.6092664304743072,
+      "epoch": 0.6295119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 6.994379515121006e-06,
-      "loss": 0.8967,
+      "learning_rate": 6.375940839812667e-06,
+      "loss": 0.901,
       "step": 22184
     },
     {
-      "epoch": 0.6092938947021505,
+      "epoch": 0.6295402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 6.993531140207394e-06,
-      "loss": 0.794,
+      "learning_rate": 6.375084255647654e-06,
+      "loss": 0.894,
       "step": 22185
     },
     {
-      "epoch": 0.6093213589299937,
+      "epoch": 0.6295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 6.992682789082127e-06,
-      "loss": 0.8468,
+      "learning_rate": 6.374227702102467e-06,
+      "loss": 0.844,
       "step": 22186
     },
     {
-      "epoch": 0.6093488231578369,
+      "epoch": 0.6295970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 6.991834461751916e-06,
-      "loss": 0.7966,
+      "learning_rate": 6.37337117918434e-06,
+      "loss": 0.8312,
       "step": 22187
     },
     {
-      "epoch": 0.6093762873856802,
+      "epoch": 0.6296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 6.9909861582234766e-06,
-      "loss": 0.8968,
+      "learning_rate": 6.372514686900513e-06,
+      "loss": 0.8874,
       "step": 22188
     },
     {
-      "epoch": 0.6094037516135233,
+      "epoch": 0.6296538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 6.99013787850352e-06,
-      "loss": 0.7998,
+      "learning_rate": 6.371658225258216e-06,
+      "loss": 0.8105,
       "step": 22189
     },
     {
-      "epoch": 0.6094312158413666,
+      "epoch": 0.6296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 6.9892896225987625e-06,
-      "loss": 0.7895,
+      "learning_rate": 6.370801794264683e-06,
+      "loss": 0.8303,
       "step": 22190
     },
     {
-      "epoch": 0.6094586800692099,
+      "epoch": 0.6297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.98844139051591e-06,
-      "loss": 0.849,
+      "learning_rate": 6.3699453939271525e-06,
+      "loss": 0.9252,
       "step": 22191
     },
     {
-      "epoch": 0.6094861442970531,
+      "epoch": 0.629738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 6.987593182261673e-06,
-      "loss": 0.8432,
+      "learning_rate": 6.3690890242528534e-06,
+      "loss": 0.8686,
       "step": 22192
     },
     {
-      "epoch": 0.6095136085248963,
+      "epoch": 0.6297673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 6.986744997842766e-06,
-      "loss": 0.8755,
+      "learning_rate": 6.368232685249026e-06,
+      "loss": 0.7939,
       "step": 22193
     },
     {
-      "epoch": 0.6095410727527396,
+      "epoch": 0.6297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 6.985896837265902e-06,
-      "loss": 0.9406,
+      "learning_rate": 6.367376376922898e-06,
+      "loss": 0.9327,
       "step": 22194
     },
     {
-      "epoch": 0.6095685369805828,
+      "epoch": 0.6298240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 6.985048700537786e-06,
-      "loss": 0.8479,
+      "learning_rate": 6.366520099281708e-06,
+      "loss": 0.7712,
       "step": 22195
     },
     {
-      "epoch": 0.609596001208426,
+      "epoch": 0.6298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 6.98420058766514e-06,
-      "loss": 0.8335,
+      "learning_rate": 6.365663852332685e-06,
+      "loss": 0.8973,
       "step": 22196
     },
     {
-      "epoch": 0.6096234654362692,
+      "epoch": 0.6298808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 6.983352498654661e-06,
-      "loss": 0.9192,
+      "learning_rate": 6.3648076360830615e-06,
+      "loss": 0.9028,
       "step": 22197
     },
     {
-      "epoch": 0.6096509296641125,
+      "epoch": 0.6299091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 6.982504433513067e-06,
-      "loss": 0.9208,
+      "learning_rate": 6.3639514505400735e-06,
+      "loss": 0.7932,
       "step": 22198
     },
     {
-      "epoch": 0.6096783938919558,
+      "epoch": 0.6299375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 6.9816563922470674e-06,
-      "loss": 0.8446,
+      "learning_rate": 6.36309529571095e-06,
+      "loss": 0.7864,
       "step": 22199
     },
     {
-      "epoch": 0.6097058581197989,
+      "epoch": 0.6299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 6.980808374863368e-06,
-      "loss": 0.8927,
+      "learning_rate": 6.362239171602923e-06,
+      "loss": 0.8865,
       "step": 22200
     },
     {
-      "epoch": 0.6097333223476422,
+      "epoch": 0.629994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 6.979960381368685e-06,
-      "loss": 0.8876,
+      "learning_rate": 6.361383078223227e-06,
+      "loss": 0.858,
       "step": 22201
     },
     {
-      "epoch": 0.6097607865754854,
+      "epoch": 0.6300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 6.9791124117697285e-06,
-      "loss": 1.0196,
+      "learning_rate": 6.3605270155790925e-06,
+      "loss": 0.8957,
       "step": 22202
     },
     {
-      "epoch": 0.6097882508033287,
+      "epoch": 0.6300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 6.978264466073203e-06,
-      "loss": 0.9066,
+      "learning_rate": 6.3596709836777455e-06,
+      "loss": 0.8607,
       "step": 22203
     },
     {
-      "epoch": 0.6098157150311719,
+      "epoch": 0.6300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 6.977416544285818e-06,
-      "loss": 0.9608,
+      "learning_rate": 6.3588149825264246e-06,
+      "loss": 0.9182,
       "step": 22204
     },
     {
-      "epoch": 0.6098431792590151,
+      "epoch": 0.6301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 6.976568646414282e-06,
-      "loss": 0.8904,
+      "learning_rate": 6.357959012132357e-06,
+      "loss": 0.8235,
       "step": 22205
     },
     {
-      "epoch": 0.6098706434868584,
+      "epoch": 0.6301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 6.975720772465311e-06,
-      "loss": 0.9186,
+      "learning_rate": 6.357103072502772e-06,
+      "loss": 0.7928,
       "step": 22206
     },
     {
-      "epoch": 0.6098981077147017,
+      "epoch": 0.6301645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 6.9748729224456034e-06,
-      "loss": 0.7424,
+      "learning_rate": 6.3562471636449016e-06,
+      "loss": 0.8156,
       "step": 22207
     },
     {
-      "epoch": 0.6099255719425448,
+      "epoch": 0.6301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 6.9740250963618805e-06,
-      "loss": 0.8818,
+      "learning_rate": 6.355391285565975e-06,
+      "loss": 0.7854,
       "step": 22208
     },
     {
-      "epoch": 0.6099530361703881,
+      "epoch": 0.6302213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 6.973177294220836e-06,
-      "loss": 0.8253,
+      "learning_rate": 6.354535438273224e-06,
+      "loss": 0.8448,
       "step": 22209
     },
     {
-      "epoch": 0.6099805003982313,
+      "epoch": 0.630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 6.9723295160291905e-06,
-      "loss": 0.8642,
+      "learning_rate": 6.353679621773875e-06,
+      "loss": 0.96,
       "step": 22210
     },
     {
-      "epoch": 0.6100079646260745,
+      "epoch": 0.63027809307605,
       "grad_norm": 0.0,
-      "learning_rate": 6.971481761793641e-06,
-      "loss": 0.8034,
+      "learning_rate": 6.352823836075157e-06,
+      "loss": 0.9124,
       "step": 22211
     },
     {
-      "epoch": 0.6100354288539178,
+      "epoch": 0.6303064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 6.970634031520904e-06,
-      "loss": 0.9091,
+      "learning_rate": 6.351968081184301e-06,
+      "loss": 0.7666,
       "step": 22212
     },
     {
-      "epoch": 0.610062893081761,
+      "epoch": 0.6303348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 6.969786325217682e-06,
-      "loss": 0.8661,
+      "learning_rate": 6.351112357108535e-06,
+      "loss": 0.9066,
       "step": 22213
     },
     {
-      "epoch": 0.6100903573096043,
+      "epoch": 0.6303632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 6.96893864289069e-06,
-      "loss": 0.8895,
+      "learning_rate": 6.350256663855086e-06,
+      "loss": 0.8039,
       "step": 22214
     },
     {
-      "epoch": 0.6101178215374474,
+      "epoch": 0.6303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 6.968090984546627e-06,
-      "loss": 0.8386,
+      "learning_rate": 6.349401001431185e-06,
+      "loss": 0.819,
       "step": 22215
     },
     {
-      "epoch": 0.6101452857652907,
+      "epoch": 0.6304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 6.967243350192201e-06,
-      "loss": 0.9031,
+      "learning_rate": 6.3485453698440565e-06,
+      "loss": 0.8717,
       "step": 22216
     },
     {
-      "epoch": 0.610172749993134,
+      "epoch": 0.6304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 6.96639573983412e-06,
-      "loss": 0.863,
+      "learning_rate": 6.347689769100929e-06,
+      "loss": 0.9197,
       "step": 22217
     },
     {
-      "epoch": 0.6102002142209771,
+      "epoch": 0.6304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 6.965548153479092e-06,
-      "loss": 0.8424,
+      "learning_rate": 6.346834199209031e-06,
+      "loss": 0.8308,
       "step": 22218
     },
     {
-      "epoch": 0.6102276784488204,
+      "epoch": 0.6305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 6.964700591133822e-06,
-      "loss": 0.801,
+      "learning_rate": 6.345978660175589e-06,
+      "loss": 0.7084,
       "step": 22219
     },
     {
-      "epoch": 0.6102551426766637,
+      "epoch": 0.6305334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 6.96385305280502e-06,
-      "loss": 0.882,
+      "learning_rate": 6.345123152007827e-06,
+      "loss": 0.7356,
       "step": 22220
     },
     {
-      "epoch": 0.6102826069045069,
+      "epoch": 0.6305618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 6.963005538499385e-06,
-      "loss": 0.8728,
+      "learning_rate": 6.344267674712976e-06,
+      "loss": 0.8578,
       "step": 22221
     },
     {
-      "epoch": 0.6103100711323501,
+      "epoch": 0.6305902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 6.962158048223628e-06,
-      "loss": 0.8156,
+      "learning_rate": 6.343412228298262e-06,
+      "loss": 0.7877,
       "step": 22222
     },
     {
-      "epoch": 0.6103375353601933,
+      "epoch": 0.6306186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 6.961310581984452e-06,
-      "loss": 0.9015,
+      "learning_rate": 6.342556812770904e-06,
+      "loss": 0.8898,
       "step": 22223
     },
     {
-      "epoch": 0.6103649995880366,
+      "epoch": 0.6306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 6.960463139788562e-06,
-      "loss": 0.7932,
+      "learning_rate": 6.341701428138135e-06,
+      "loss": 0.9014,
       "step": 22224
     },
     {
-      "epoch": 0.6103924638158799,
+      "epoch": 0.6306753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 6.959615721642669e-06,
-      "loss": 0.941,
+      "learning_rate": 6.340846074407181e-06,
+      "loss": 0.8575,
       "step": 22225
     },
     {
-      "epoch": 0.610419928043723,
+      "epoch": 0.6307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 6.9587683275534735e-06,
-      "loss": 0.834,
+      "learning_rate": 6.3399907515852645e-06,
+      "loss": 0.8442,
       "step": 22226
     },
     {
-      "epoch": 0.6104473922715663,
+      "epoch": 0.6307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 6.957920957527682e-06,
-      "loss": 0.7727,
+      "learning_rate": 6.339135459679611e-06,
+      "loss": 0.8222,
       "step": 22227
     },
     {
-      "epoch": 0.6104748564994095,
+      "epoch": 0.6307604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 6.957073611571995e-06,
-      "loss": 0.7213,
+      "learning_rate": 6.338280198697442e-06,
+      "loss": 0.9388,
       "step": 22228
     },
     {
-      "epoch": 0.6105023207272527,
+      "epoch": 0.630788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 6.95622628969312e-06,
-      "loss": 0.8777,
+      "learning_rate": 6.337424968645987e-06,
+      "loss": 0.7813,
       "step": 22229
     },
     {
-      "epoch": 0.610529784955096,
+      "epoch": 0.6308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 6.955378991897765e-06,
-      "loss": 0.8331,
+      "learning_rate": 6.336569769532468e-06,
+      "loss": 0.8235,
       "step": 22230
     },
     {
-      "epoch": 0.6105572491829392,
+      "epoch": 0.6308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 6.954531718192627e-06,
-      "loss": 0.8891,
+      "learning_rate": 6.3357146013641065e-06,
+      "loss": 0.7905,
       "step": 22231
     },
     {
-      "epoch": 0.6105847134107825,
+      "epoch": 0.6308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 6.95368446858442e-06,
-      "loss": 0.8341,
+      "learning_rate": 6.334859464148131e-06,
+      "loss": 0.9272,
       "step": 22232
     },
     {
-      "epoch": 0.6106121776386257,
+      "epoch": 0.6309023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 6.9528372430798365e-06,
-      "loss": 0.8775,
+      "learning_rate": 6.334004357891763e-06,
+      "loss": 0.9406,
       "step": 22233
     },
     {
-      "epoch": 0.6106396418664689,
+      "epoch": 0.6309307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 6.951990041685587e-06,
-      "loss": 0.8289,
+      "learning_rate": 6.333149282602222e-06,
+      "loss": 0.917,
       "step": 22234
     },
     {
-      "epoch": 0.6106671060943122,
+      "epoch": 0.6309591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 6.951142864408371e-06,
-      "loss": 0.8468,
+      "learning_rate": 6.332294238286736e-06,
+      "loss": 0.859,
       "step": 22235
     },
     {
-      "epoch": 0.6106945703221554,
+      "epoch": 0.6309875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 6.950295711254893e-06,
-      "loss": 0.8034,
+      "learning_rate": 6.331439224952525e-06,
+      "loss": 0.829,
       "step": 22236
     },
     {
-      "epoch": 0.6107220345499986,
+      "epoch": 0.6310158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 6.949448582231859e-06,
-      "loss": 0.8054,
+      "learning_rate": 6.3305842426068095e-06,
+      "loss": 0.9536,
       "step": 22237
     },
     {
-      "epoch": 0.6107494987778419,
+      "epoch": 0.6310442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 6.948601477345971e-06,
-      "loss": 1.0151,
+      "learning_rate": 6.329729291256814e-06,
+      "loss": 0.8813,
       "step": 22238
     },
     {
-      "epoch": 0.6107769630056851,
+      "epoch": 0.6310726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 6.947754396603928e-06,
-      "loss": 0.9058,
+      "learning_rate": 6.328874370909762e-06,
+      "loss": 0.9761,
       "step": 22239
     },
     {
-      "epoch": 0.6108044272335283,
+      "epoch": 0.6311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 6.946907340012434e-06,
-      "loss": 0.8315,
+      "learning_rate": 6.328019481572875e-06,
+      "loss": 0.8198,
       "step": 22240
     },
     {
-      "epoch": 0.6108318914613715,
+      "epoch": 0.6311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 6.946060307578191e-06,
-      "loss": 0.7525,
+      "learning_rate": 6.327164623253369e-06,
+      "loss": 0.7994,
       "step": 22241
     },
     {
-      "epoch": 0.6108593556892148,
+      "epoch": 0.6311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.945213299307903e-06,
-      "loss": 0.7652,
+      "learning_rate": 6.326309795958472e-06,
+      "loss": 0.8984,
       "step": 22242
     },
     {
-      "epoch": 0.6108868199170581,
+      "epoch": 0.6311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 6.944366315208271e-06,
-      "loss": 0.9465,
+      "learning_rate": 6.325454999695399e-06,
+      "loss": 0.775,
       "step": 22243
     },
     {
-      "epoch": 0.6109142841449012,
+      "epoch": 0.6312145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 6.9435193552859995e-06,
-      "loss": 0.8752,
+      "learning_rate": 6.324600234471372e-06,
+      "loss": 0.791,
       "step": 22244
     },
     {
-      "epoch": 0.6109417483727445,
+      "epoch": 0.6312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 6.942672419547781e-06,
-      "loss": 0.89,
+      "learning_rate": 6.323745500293614e-06,
+      "loss": 0.6732,
       "step": 22245
     },
     {
-      "epoch": 0.6109692126005878,
+      "epoch": 0.6312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 6.941825508000326e-06,
-      "loss": 0.8886,
+      "learning_rate": 6.322890797169344e-06,
+      "loss": 0.886,
       "step": 22246
     },
     {
-      "epoch": 0.610996676828431,
+      "epoch": 0.631299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 6.940978620650331e-06,
-      "loss": 0.8979,
+      "learning_rate": 6.322036125105779e-06,
+      "loss": 0.8795,
       "step": 22247
     },
     {
-      "epoch": 0.6110241410562742,
+      "epoch": 0.631328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 6.940131757504497e-06,
-      "loss": 0.9412,
+      "learning_rate": 6.3211814841101404e-06,
+      "loss": 0.8882,
       "step": 22248
     },
     {
-      "epoch": 0.6110516052841174,
+      "epoch": 0.6313564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 6.93928491856953e-06,
-      "loss": 0.8644,
+      "learning_rate": 6.320326874189648e-06,
+      "loss": 0.8614,
       "step": 22249
     },
     {
-      "epoch": 0.6110790695119607,
+      "epoch": 0.6313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 6.938438103852123e-06,
-      "loss": 0.8794,
+      "learning_rate": 6.319472295351517e-06,
+      "loss": 0.8184,
       "step": 22250
     },
     {
-      "epoch": 0.611106533739804,
+      "epoch": 0.6314131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 6.93759131335898e-06,
-      "loss": 0.7492,
+      "learning_rate": 6.318617747602973e-06,
+      "loss": 0.7575,
       "step": 22251
     },
     {
-      "epoch": 0.6111339979676471,
+      "epoch": 0.6314415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 6.9367445470968005e-06,
-      "loss": 0.8387,
+      "learning_rate": 6.317763230951228e-06,
+      "loss": 0.9316,
       "step": 22252
     },
     {
-      "epoch": 0.6111614621954904,
+      "epoch": 0.6314699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 6.9358978050722844e-06,
-      "loss": 0.8737,
+      "learning_rate": 6.3169087454035005e-06,
+      "loss": 0.8894,
       "step": 22253
     },
     {
-      "epoch": 0.6111889264233336,
+      "epoch": 0.6314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 6.935051087292133e-06,
-      "loss": 0.9154,
+      "learning_rate": 6.316054290967013e-06,
+      "loss": 0.7932,
       "step": 22254
     },
     {
-      "epoch": 0.6112163906511768,
+      "epoch": 0.6315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 6.934204393763048e-06,
-      "loss": 0.8536,
+      "learning_rate": 6.31519986764898e-06,
+      "loss": 0.879,
       "step": 22255
     },
     {
-      "epoch": 0.6112438548790201,
+      "epoch": 0.6315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 6.93335772449172e-06,
-      "loss": 0.815,
+      "learning_rate": 6.31434547545662e-06,
+      "loss": 0.9011,
       "step": 22256
     },
     {
-      "epoch": 0.6112713191068633,
+      "epoch": 0.631583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 6.932511079484854e-06,
-      "loss": 0.8438,
+      "learning_rate": 6.31349111439715e-06,
+      "loss": 0.7681,
       "step": 22257
     },
     {
-      "epoch": 0.6112987833347066,
+      "epoch": 0.6316118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 6.931664458749151e-06,
-      "loss": 0.9035,
+      "learning_rate": 6.312636784477783e-06,
+      "loss": 0.8327,
       "step": 22258
     },
     {
-      "epoch": 0.6113262475625498,
+      "epoch": 0.6316401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 6.930817862291306e-06,
-      "loss": 0.8591,
+      "learning_rate": 6.311782485705744e-06,
+      "loss": 0.8181,
       "step": 22259
     },
     {
-      "epoch": 0.611353711790393,
+      "epoch": 0.6316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 6.9299712901180175e-06,
-      "loss": 0.9186,
+      "learning_rate": 6.310928218088241e-06,
+      "loss": 0.7396,
       "step": 22260
     },
     {
-      "epoch": 0.6113811760182363,
+      "epoch": 0.6316969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 6.92912474223599e-06,
-      "loss": 0.9202,
+      "learning_rate": 6.310073981632493e-06,
+      "loss": 0.8927,
       "step": 22261
     },
     {
-      "epoch": 0.6114086402460794,
+      "epoch": 0.6317253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 6.928278218651913e-06,
-      "loss": 0.9716,
+      "learning_rate": 6.309219776345718e-06,
+      "loss": 0.8686,
       "step": 22262
     },
     {
-      "epoch": 0.6114361044739227,
+      "epoch": 0.6317536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 6.927431719372491e-06,
-      "loss": 0.9704,
+      "learning_rate": 6.308365602235128e-06,
+      "loss": 0.9436,
       "step": 22263
     },
     {
-      "epoch": 0.611463568701766,
+      "epoch": 0.6317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 6.926585244404416e-06,
-      "loss": 0.8322,
+      "learning_rate": 6.3075114593079375e-06,
+      "loss": 0.8765,
       "step": 22264
     },
     {
-      "epoch": 0.6114910329296092,
+      "epoch": 0.6318104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 6.9257387937543905e-06,
-      "loss": 0.9073,
+      "learning_rate": 6.306657347571367e-06,
+      "loss": 0.8997,
       "step": 22265
     },
     {
-      "epoch": 0.6115184971574524,
+      "epoch": 0.631838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 6.924892367429112e-06,
-      "loss": 0.9313,
+      "learning_rate": 6.3058032670326285e-06,
+      "loss": 0.7998,
       "step": 22266
     },
     {
-      "epoch": 0.6115459613852956,
+      "epoch": 0.6318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 6.924045965435279e-06,
-      "loss": 0.961,
+      "learning_rate": 6.304949217698932e-06,
+      "loss": 0.8043,
       "step": 22267
     },
     {
-      "epoch": 0.6115734256131389,
+      "epoch": 0.6318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 6.923199587779581e-06,
-      "loss": 0.8714,
+      "learning_rate": 6.3040951995775e-06,
+      "loss": 0.8702,
       "step": 22268
     },
     {
-      "epoch": 0.6116008898409822,
+      "epoch": 0.6319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 6.92235323446872e-06,
-      "loss": 0.8905,
+      "learning_rate": 6.303241212675536e-06,
+      "loss": 0.8519,
       "step": 22269
     },
     {
-      "epoch": 0.6116283540688253,
+      "epoch": 0.6319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 6.921506905509394e-06,
-      "loss": 0.7908,
+      "learning_rate": 6.302387257000264e-06,
+      "loss": 0.885,
       "step": 22270
     },
     {
-      "epoch": 0.6116558182966686,
+      "epoch": 0.6319807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 6.920660600908297e-06,
-      "loss": 0.9006,
+      "learning_rate": 6.3015333325588934e-06,
+      "loss": 0.7937,
       "step": 22271
     },
     {
-      "epoch": 0.6116832825245119,
+      "epoch": 0.6320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 6.9198143206721245e-06,
-      "loss": 0.8131,
+      "learning_rate": 6.300679439358635e-06,
+      "loss": 0.8032,
       "step": 22272
     },
     {
-      "epoch": 0.611710746752355,
+      "epoch": 0.6320374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 6.918968064807579e-06,
-      "loss": 0.8958,
+      "learning_rate": 6.2998255774067056e-06,
+      "loss": 0.8414,
       "step": 22273
     },
     {
-      "epoch": 0.6117382109801983,
+      "epoch": 0.6320658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 6.918121833321347e-06,
-      "loss": 0.8492,
+      "learning_rate": 6.298971746710316e-06,
+      "loss": 0.8444,
       "step": 22274
     },
     {
-      "epoch": 0.6117656752080415,
+      "epoch": 0.632094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 6.917275626220132e-06,
-      "loss": 0.9226,
+      "learning_rate": 6.298117947276676e-06,
+      "loss": 0.8255,
       "step": 22275
     },
     {
-      "epoch": 0.6117931394358848,
+      "epoch": 0.632122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 6.916429443510624e-06,
-      "loss": 0.8899,
+      "learning_rate": 6.297264179113001e-06,
+      "loss": 0.8847,
       "step": 22276
     },
     {
-      "epoch": 0.611820603663728,
+      "epoch": 0.6321509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 6.915583285199521e-06,
-      "loss": 0.9277,
+      "learning_rate": 6.296410442226504e-06,
+      "loss": 0.9059,
       "step": 22277
     },
     {
-      "epoch": 0.6118480678915712,
+      "epoch": 0.6321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 6.9147371512935205e-06,
-      "loss": 0.8153,
+      "learning_rate": 6.295556736624392e-06,
+      "loss": 0.9274,
       "step": 22278
     },
     {
-      "epoch": 0.6118755321194145,
+      "epoch": 0.6322077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 6.913891041799315e-06,
-      "loss": 0.8294,
+      "learning_rate": 6.29470306231388e-06,
+      "loss": 0.7677,
       "step": 22279
     },
     {
-      "epoch": 0.6119029963472576,
+      "epoch": 0.6322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 6.913044956723595e-06,
-      "loss": 0.8485,
+      "learning_rate": 6.293849419302179e-06,
+      "loss": 0.7995,
       "step": 22280
     },
     {
-      "epoch": 0.6119304605751009,
+      "epoch": 0.6322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 6.9121988960730616e-06,
-      "loss": 0.8287,
+      "learning_rate": 6.292995807596496e-06,
+      "loss": 0.9567,
       "step": 22281
     },
     {
-      "epoch": 0.6119579248029442,
+      "epoch": 0.6322928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 6.911352859854405e-06,
-      "loss": 0.8534,
+      "learning_rate": 6.292142227204045e-06,
+      "loss": 0.825,
       "step": 22282
     },
     {
-      "epoch": 0.6119853890307874,
+      "epoch": 0.6323212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 6.910506848074322e-06,
-      "loss": 0.7575,
+      "learning_rate": 6.291288678132037e-06,
+      "loss": 0.7341,
       "step": 22283
     },
     {
-      "epoch": 0.6120128532586306,
+      "epoch": 0.6323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 6.909660860739505e-06,
-      "loss": 0.8047,
+      "learning_rate": 6.2904351603876756e-06,
+      "loss": 0.9837,
       "step": 22284
     },
     {
-      "epoch": 0.6120403174864739,
+      "epoch": 0.6323779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 6.908814897856652e-06,
-      "loss": 0.8176,
+      "learning_rate": 6.289581673978177e-06,
+      "loss": 0.9197,
       "step": 22285
     },
     {
-      "epoch": 0.6120677817143171,
+      "epoch": 0.6324063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 6.907968959432449e-06,
-      "loss": 0.8074,
+      "learning_rate": 6.288728218910751e-06,
+      "loss": 0.8753,
       "step": 22286
     },
     {
-      "epoch": 0.6120952459421604,
+      "epoch": 0.6324347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 6.907123045473594e-06,
-      "loss": 0.836,
+      "learning_rate": 6.287874795192603e-06,
+      "loss": 0.7837,
       "step": 22287
     },
     {
-      "epoch": 0.6121227101700035,
+      "epoch": 0.6324631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 6.9062771559867795e-06,
-      "loss": 0.8348,
+      "learning_rate": 6.287021402830944e-06,
+      "loss": 0.684,
       "step": 22288
     },
     {
-      "epoch": 0.6121501743978468,
+      "epoch": 0.6324914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 6.905431290978697e-06,
-      "loss": 0.8728,
+      "learning_rate": 6.286168041832983e-06,
+      "loss": 0.7476,
       "step": 22289
     },
     {
-      "epoch": 0.6121776386256901,
+      "epoch": 0.6325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 6.904585450456044e-06,
-      "loss": 0.8661,
+      "learning_rate": 6.285314712205928e-06,
+      "loss": 0.8328,
       "step": 22290
     },
     {
-      "epoch": 0.6122051028535332,
+      "epoch": 0.6325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 6.903739634425511e-06,
-      "loss": 0.9395,
+      "learning_rate": 6.284461413956984e-06,
+      "loss": 0.78,
       "step": 22291
     },
     {
-      "epoch": 0.6122325670813765,
+      "epoch": 0.6325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 6.902893842893786e-06,
-      "loss": 0.8245,
+      "learning_rate": 6.283608147093363e-06,
+      "loss": 0.9043,
       "step": 22292
     },
     {
-      "epoch": 0.6122600313092197,
+      "epoch": 0.6326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 6.902048075867564e-06,
-      "loss": 0.889,
+      "learning_rate": 6.282754911622271e-06,
+      "loss": 0.7618,
       "step": 22293
     },
     {
-      "epoch": 0.612287495537063,
+      "epoch": 0.632633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 6.9012023333535405e-06,
-      "loss": 0.8623,
+      "learning_rate": 6.281901707550917e-06,
+      "loss": 0.7795,
       "step": 22294
     },
     {
-      "epoch": 0.6123149597649062,
+      "epoch": 0.6326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 6.900356615358402e-06,
-      "loss": 0.8091,
+      "learning_rate": 6.281048534886503e-06,
+      "loss": 0.8342,
       "step": 22295
     },
     {
-      "epoch": 0.6123424239927494,
+      "epoch": 0.6326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 6.899510921888843e-06,
-      "loss": 0.8629,
+      "learning_rate": 6.280195393636241e-06,
+      "loss": 0.8994,
       "step": 22296
     },
     {
-      "epoch": 0.6123698882205927,
+      "epoch": 0.6327185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 6.898665252951559e-06,
-      "loss": 0.8785,
+      "learning_rate": 6.279342283807337e-06,
+      "loss": 0.9462,
       "step": 22297
     },
     {
-      "epoch": 0.612397352448436,
+      "epoch": 0.6327468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 6.897819608553231e-06,
-      "loss": 0.7819,
+      "learning_rate": 6.278489205406992e-06,
+      "loss": 0.8893,
       "step": 22298
     },
     {
-      "epoch": 0.6124248166762791,
+      "epoch": 0.6327752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 6.896973988700559e-06,
-      "loss": 0.7099,
+      "learning_rate": 6.2776361584424205e-06,
+      "loss": 0.8999,
       "step": 22299
     },
     {
-      "epoch": 0.6124522809041224,
+      "epoch": 0.6328036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 6.89612839340023e-06,
-      "loss": 0.8597,
+      "learning_rate": 6.276783142920818e-06,
+      "loss": 0.9097,
       "step": 22300
     },
     {
-      "epoch": 0.6124797451319656,
+      "epoch": 0.6328320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 6.8952828226589345e-06,
-      "loss": 0.9418,
+      "learning_rate": 6.2759301588494e-06,
+      "loss": 0.8419,
       "step": 22301
     },
     {
-      "epoch": 0.6125072093598088,
+      "epoch": 0.6328603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 6.894437276483366e-06,
-      "loss": 0.8559,
+      "learning_rate": 6.275077206235366e-06,
+      "loss": 0.8675,
       "step": 22302
     },
     {
-      "epoch": 0.6125346735876521,
+      "epoch": 0.6328887627695801,
       "grad_norm": 0.0,
-      "learning_rate": 6.8935917548802145e-06,
-      "loss": 0.7871,
+      "learning_rate": 6.274224285085924e-06,
+      "loss": 0.8924,
       "step": 22303
     },
     {
-      "epoch": 0.6125621378154953,
+      "epoch": 0.6329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 6.8927462578561645e-06,
-      "loss": 0.8898,
+      "learning_rate": 6.273371395408276e-06,
+      "loss": 0.7766,
       "step": 22304
     },
     {
-      "epoch": 0.6125896020433386,
+      "epoch": 0.6329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 6.89190078541791e-06,
-      "loss": 0.7498,
+      "learning_rate": 6.2725185372096266e-06,
+      "loss": 0.8461,
       "step": 22305
     },
     {
-      "epoch": 0.6126170662711817,
+      "epoch": 0.6329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 6.891055337572142e-06,
-      "loss": 0.9472,
+      "learning_rate": 6.271665710497182e-06,
+      "loss": 0.851,
       "step": 22306
     },
     {
-      "epoch": 0.612644530499025,
+      "epoch": 0.6330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 6.890209914325548e-06,
-      "loss": 0.8384,
+      "learning_rate": 6.270812915278145e-06,
+      "loss": 0.9565,
       "step": 22307
     },
     {
-      "epoch": 0.6126719947268683,
+      "epoch": 0.6330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 6.88936451568482e-06,
-      "loss": 0.8421,
+      "learning_rate": 6.269960151559718e-06,
+      "loss": 0.8432,
       "step": 22308
     },
     {
-      "epoch": 0.6126994589547115,
+      "epoch": 0.6330590238365493,
       "grad_norm": 0.0,
-      "learning_rate": 6.88851914165664e-06,
-      "loss": 0.8499,
+      "learning_rate": 6.269107419349105e-06,
+      "loss": 0.8237,
       "step": 22309
     },
     {
-      "epoch": 0.6127269231825547,
+      "epoch": 0.6330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 6.8876737922477046e-06,
-      "loss": 0.7813,
+      "learning_rate": 6.2682547186535105e-06,
+      "loss": 0.9132,
       "step": 22310
     },
     {
-      "epoch": 0.612754387410398,
+      "epoch": 0.6331157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 6.886828467464699e-06,
-      "loss": 0.8871,
+      "learning_rate": 6.2674020494801335e-06,
+      "loss": 0.9189,
       "step": 22311
     },
     {
-      "epoch": 0.6127818516382412,
+      "epoch": 0.633144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 6.885983167314312e-06,
-      "loss": 0.8649,
+      "learning_rate": 6.266549411836182e-06,
+      "loss": 0.8053,
       "step": 22312
     },
     {
-      "epoch": 0.6128093158660844,
+      "epoch": 0.633172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 6.885137891803233e-06,
-      "loss": 0.8622,
+      "learning_rate": 6.265696805728856e-06,
+      "loss": 0.8555,
       "step": 22313
     },
     {
-      "epoch": 0.6128367800939276,
+      "epoch": 0.6332009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 6.884292640938152e-06,
-      "loss": 0.8338,
+      "learning_rate": 6.264844231165354e-06,
+      "loss": 0.7873,
       "step": 22314
     },
     {
-      "epoch": 0.6128642443217709,
+      "epoch": 0.6332292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 6.883447414725753e-06,
-      "loss": 0.833,
+      "learning_rate": 6.263991688152881e-06,
+      "loss": 0.8676,
       "step": 22315
     },
     {
-      "epoch": 0.6128917085496142,
+      "epoch": 0.6332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 6.882602213172725e-06,
-      "loss": 0.8072,
+      "learning_rate": 6.263139176698638e-06,
+      "loss": 0.9449,
       "step": 22316
     },
     {
-      "epoch": 0.6129191727774573,
+      "epoch": 0.6332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 6.8817570362857545e-06,
-      "loss": 0.8242,
+      "learning_rate": 6.262286696809827e-06,
+      "loss": 0.7717,
       "step": 22317
     },
     {
-      "epoch": 0.6129466370053006,
+      "epoch": 0.6333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 6.880911884071532e-06,
-      "loss": 0.839,
+      "learning_rate": 6.26143424849365e-06,
+      "loss": 0.9086,
       "step": 22318
     },
     {
-      "epoch": 0.6129741012331438,
+      "epoch": 0.6333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 6.880066756536741e-06,
-      "loss": 0.7983,
+      "learning_rate": 6.260581831757304e-06,
+      "loss": 0.9547,
       "step": 22319
     },
     {
-      "epoch": 0.613001565460987,
+      "epoch": 0.6333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 6.879221653688074e-06,
-      "loss": 0.8473,
+      "learning_rate": 6.259729446607994e-06,
+      "loss": 0.8962,
       "step": 22320
     },
     {
-      "epoch": 0.6130290296888303,
+      "epoch": 0.6333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 6.878376575532211e-06,
-      "loss": 0.8505,
+      "learning_rate": 6.258877093052915e-06,
+      "loss": 0.9138,
       "step": 22321
     },
     {
-      "epoch": 0.6130564939166735,
+      "epoch": 0.633427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 6.877531522075841e-06,
-      "loss": 0.8395,
+      "learning_rate": 6.258024771099269e-06,
+      "loss": 0.8507,
       "step": 22322
     },
     {
-      "epoch": 0.6130839581445168,
+      "epoch": 0.6334562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 6.8766864933256534e-06,
-      "loss": 0.8278,
+      "learning_rate": 6.257172480754259e-06,
+      "loss": 0.8559,
       "step": 22323
     },
     {
-      "epoch": 0.61311142237236,
+      "epoch": 0.6334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 6.87584148928833e-06,
-      "loss": 0.9346,
+      "learning_rate": 6.256320222025079e-06,
+      "loss": 0.953,
       "step": 22324
     },
     {
-      "epoch": 0.6131388866002032,
+      "epoch": 0.6335130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 6.874996509970559e-06,
-      "loss": 0.7787,
+      "learning_rate": 6.25546799491893e-06,
+      "loss": 0.91,
       "step": 22325
     },
     {
-      "epoch": 0.6131663508280465,
+      "epoch": 0.6335414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 6.874151555379032e-06,
-      "loss": 0.7965,
+      "learning_rate": 6.254615799443013e-06,
+      "loss": 0.8343,
       "step": 22326
     },
     {
-      "epoch": 0.6131938150558897,
+      "epoch": 0.6335698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 6.873306625520424e-06,
-      "loss": 0.9061,
+      "learning_rate": 6.253763635604524e-06,
+      "loss": 0.9492,
       "step": 22327
     },
     {
-      "epoch": 0.6132212792837329,
+      "epoch": 0.6335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 6.872461720401424e-06,
-      "loss": 0.8153,
+      "learning_rate": 6.252911503410662e-06,
+      "loss": 0.828,
       "step": 22328
     },
     {
-      "epoch": 0.6132487435115762,
+      "epoch": 0.6336265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 6.871616840028718e-06,
-      "loss": 0.8897,
+      "learning_rate": 6.252059402868624e-06,
+      "loss": 0.883,
       "step": 22329
     },
     {
-      "epoch": 0.6132762077394194,
+      "epoch": 0.6336549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 6.8707719844089945e-06,
-      "loss": 0.878,
+      "learning_rate": 6.2512073339856065e-06,
+      "loss": 0.8402,
       "step": 22330
     },
     {
-      "epoch": 0.6133036719672627,
+      "epoch": 0.633683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 6.869927153548932e-06,
-      "loss": 0.8156,
+      "learning_rate": 6.250355296768813e-06,
+      "loss": 0.8641,
       "step": 22331
     },
     {
-      "epoch": 0.6133311361951058,
+      "epoch": 0.6337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 6.869082347455224e-06,
-      "loss": 0.8962,
+      "learning_rate": 6.249503291225434e-06,
+      "loss": 0.8705,
       "step": 22332
     },
     {
-      "epoch": 0.6133586004229491,
+      "epoch": 0.6337400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 6.868237566134543e-06,
-      "loss": 0.7785,
+      "learning_rate": 6.248651317362673e-06,
+      "loss": 0.8322,
       "step": 22333
     },
     {
-      "epoch": 0.6133860646507924,
+      "epoch": 0.6337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 6.867392809593581e-06,
-      "loss": 0.8933,
+      "learning_rate": 6.247799375187721e-06,
+      "loss": 0.9561,
       "step": 22334
     },
     {
-      "epoch": 0.6134135288786355,
+      "epoch": 0.6337968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 6.866548077839023e-06,
-      "loss": 0.8743,
+      "learning_rate": 6.246947464707778e-06,
+      "loss": 0.9139,
       "step": 22335
     },
     {
-      "epoch": 0.6134409931064788,
+      "epoch": 0.6338251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 6.865703370877548e-06,
-      "loss": 0.8596,
+      "learning_rate": 6.246095585930034e-06,
+      "loss": 0.8534,
       "step": 22336
     },
     {
-      "epoch": 0.6134684573343221,
+      "epoch": 0.6338535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 6.8648586887158404e-06,
-      "loss": 0.8724,
+      "learning_rate": 6.245243738861693e-06,
+      "loss": 0.8099,
       "step": 22337
     },
     {
-      "epoch": 0.6134959215621653,
+      "epoch": 0.6338819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 6.864014031360591e-06,
-      "loss": 0.909,
+      "learning_rate": 6.244391923509948e-06,
+      "loss": 0.9053,
       "step": 22338
     },
     {
-      "epoch": 0.6135233857900085,
+      "epoch": 0.6339103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 6.863169398818474e-06,
-      "loss": 0.8443,
+      "learning_rate": 6.24354013988199e-06,
+      "loss": 0.8479,
       "step": 22339
     },
     {
-      "epoch": 0.6135508500178517,
+      "epoch": 0.6339387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 6.862324791096176e-06,
-      "loss": 0.7797,
+      "learning_rate": 6.2426883879850196e-06,
+      "loss": 0.8303,
       "step": 22340
     },
     {
-      "epoch": 0.613578314245695,
+      "epoch": 0.6339670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 6.861480208200377e-06,
-      "loss": 0.8639,
+      "learning_rate": 6.241836667826228e-06,
+      "loss": 0.8471,
       "step": 22341
     },
     {
-      "epoch": 0.6136057784735383,
+      "epoch": 0.6339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 6.860635650137767e-06,
-      "loss": 0.946,
+      "learning_rate": 6.2409849794128115e-06,
+      "loss": 0.9632,
       "step": 22342
     },
     {
-      "epoch": 0.6136332427013814,
+      "epoch": 0.6340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 6.85979111691502e-06,
-      "loss": 0.8585,
+      "learning_rate": 6.240133322751965e-06,
+      "loss": 0.8482,
       "step": 22343
     },
     {
-      "epoch": 0.6136607069292247,
+      "epoch": 0.6340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 6.8589466085388275e-06,
-      "loss": 0.8969,
+      "learning_rate": 6.239281697850881e-06,
+      "loss": 0.9033,
       "step": 22344
     },
     {
-      "epoch": 0.613688171157068,
+      "epoch": 0.6340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 6.858102125015861e-06,
-      "loss": 0.8736,
+      "learning_rate": 6.23843010471675e-06,
+      "loss": 0.6837,
       "step": 22345
     },
     {
-      "epoch": 0.6137156353849111,
+      "epoch": 0.6341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 6.85725766635281e-06,
-      "loss": 0.818,
+      "learning_rate": 6.237578543356769e-06,
+      "loss": 0.8434,
       "step": 22346
     },
     {
-      "epoch": 0.6137430996127544,
+      "epoch": 0.6341373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 6.856413232556351e-06,
-      "loss": 0.8329,
+      "learning_rate": 6.236727013778134e-06,
+      "loss": 0.852,
       "step": 22347
     },
     {
-      "epoch": 0.6137705638405976,
+      "epoch": 0.6341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 6.85556882363317e-06,
-      "loss": 0.9072,
+      "learning_rate": 6.235875515988037e-06,
+      "loss": 0.7368,
       "step": 22348
     },
     {
-      "epoch": 0.6137980280684409,
+      "epoch": 0.634194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 6.854724439589944e-06,
-      "loss": 0.8309,
+      "learning_rate": 6.235024049993665e-06,
+      "loss": 0.8298,
       "step": 22349
     },
     {
-      "epoch": 0.6138254922962841,
+      "epoch": 0.6342224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 6.853880080433362e-06,
-      "loss": 0.8413,
+      "learning_rate": 6.234172615802216e-06,
+      "loss": 0.699,
       "step": 22350
     },
     {
-      "epoch": 0.6138529565241273,
+      "epoch": 0.6342508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 6.853035746170098e-06,
-      "loss": 0.8631,
+      "learning_rate": 6.233321213420882e-06,
+      "loss": 0.8694,
       "step": 22351
     },
     {
-      "epoch": 0.6138804207519706,
+      "epoch": 0.6342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 6.852191436806831e-06,
-      "loss": 0.8556,
+      "learning_rate": 6.23246984285685e-06,
+      "loss": 0.7871,
       "step": 22352
     },
     {
-      "epoch": 0.6139078849798137,
+      "epoch": 0.6343076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 6.8513471523502455e-06,
-      "loss": 0.9662,
+      "learning_rate": 6.231618504117316e-06,
+      "loss": 0.9174,
       "step": 22353
     },
     {
-      "epoch": 0.613935349207657,
+      "epoch": 0.6343359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 6.850502892807023e-06,
-      "loss": 0.8402,
+      "learning_rate": 6.23076719720947e-06,
+      "loss": 0.7521,
       "step": 22354
     },
     {
-      "epoch": 0.6139628134355003,
+      "epoch": 0.6343643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 6.849658658183839e-06,
-      "loss": 0.8732,
+      "learning_rate": 6.229915922140502e-06,
+      "loss": 0.8665,
       "step": 22355
     },
     {
-      "epoch": 0.6139902776633435,
+      "epoch": 0.6343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 6.84881444848738e-06,
-      "loss": 0.8544,
+      "learning_rate": 6.2290646789176045e-06,
+      "loss": 0.8675,
       "step": 22356
     },
     {
-      "epoch": 0.6140177418911867,
+      "epoch": 0.6344211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 6.847970263724319e-06,
-      "loss": 0.938,
+      "learning_rate": 6.228213467547967e-06,
+      "loss": 0.8572,
       "step": 22357
     },
     {
-      "epoch": 0.61404520611903,
+      "epoch": 0.6344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 6.847126103901338e-06,
-      "loss": 0.9059,
+      "learning_rate": 6.227362288038778e-06,
+      "loss": 0.812,
       "step": 22358
     },
     {
-      "epoch": 0.6140726703468732,
+      "epoch": 0.634477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 6.846281969025117e-06,
-      "loss": 0.8535,
+      "learning_rate": 6.226511140397232e-06,
+      "loss": 0.868,
       "step": 22359
     },
     {
-      "epoch": 0.6141001345747165,
+      "epoch": 0.6345062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 6.8454378591023325e-06,
-      "loss": 0.7632,
+      "learning_rate": 6.2256600246305154e-06,
+      "loss": 0.868,
       "step": 22360
     },
     {
-      "epoch": 0.6141275988025596,
+      "epoch": 0.6345346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 6.844593774139667e-06,
-      "loss": 0.8842,
+      "learning_rate": 6.224808940745814e-06,
+      "loss": 0.7796,
       "step": 22361
     },
     {
-      "epoch": 0.6141550630304029,
+      "epoch": 0.6345629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 6.843749714143802e-06,
-      "loss": 0.7931,
+      "learning_rate": 6.223957888750327e-06,
+      "loss": 0.8184,
       "step": 22362
     },
     {
-      "epoch": 0.6141825272582462,
+      "epoch": 0.6345913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 6.84290567912141e-06,
-      "loss": 0.8862,
+      "learning_rate": 6.223106868651234e-06,
+      "loss": 0.8774,
       "step": 22363
     },
     {
-      "epoch": 0.6142099914860893,
+      "epoch": 0.6346197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 6.842061669079169e-06,
-      "loss": 0.8124,
+      "learning_rate": 6.22225588045573e-06,
+      "loss": 0.8289,
       "step": 22364
     },
     {
-      "epoch": 0.6142374557139326,
+      "epoch": 0.6346481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 6.8412176840237584e-06,
-      "loss": 0.8231,
+      "learning_rate": 6.221404924171e-06,
+      "loss": 0.856,
       "step": 22365
     },
     {
-      "epoch": 0.6142649199417758,
+      "epoch": 0.6346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 6.8403737239618595e-06,
-      "loss": 0.9072,
+      "learning_rate": 6.22055399980423e-06,
+      "loss": 0.8736,
       "step": 22366
     },
     {
-      "epoch": 0.6142923841696191,
+      "epoch": 0.6347048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 6.839529788900146e-06,
-      "loss": 0.7635,
+      "learning_rate": 6.2197031073626126e-06,
+      "loss": 0.8485,
       "step": 22367
     },
     {
-      "epoch": 0.6143198483974623,
+      "epoch": 0.634733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 6.838685878845302e-06,
-      "loss": 0.797,
+      "learning_rate": 6.218852246853333e-06,
+      "loss": 0.8154,
       "step": 22368
     },
     {
-      "epoch": 0.6143473126253055,
+      "epoch": 0.6347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 6.8378419938039945e-06,
-      "loss": 0.8269,
+      "learning_rate": 6.218001418283577e-06,
+      "loss": 0.8521,
       "step": 22369
     },
     {
-      "epoch": 0.6143747768531488,
+      "epoch": 0.6347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 6.836998133782908e-06,
-      "loss": 0.8739,
+      "learning_rate": 6.217150621660532e-06,
+      "loss": 0.8099,
       "step": 22370
     },
     {
-      "epoch": 0.6144022410809921,
+      "epoch": 0.6348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 6.8361542987887155e-06,
-      "loss": 0.9327,
+      "learning_rate": 6.21629985699139e-06,
+      "loss": 0.8166,
       "step": 22371
     },
     {
-      "epoch": 0.6144297053088352,
+      "epoch": 0.6348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 6.835310488828096e-06,
-      "loss": 0.7885,
+      "learning_rate": 6.215449124283328e-06,
+      "loss": 0.7437,
       "step": 22372
     },
     {
-      "epoch": 0.6144571695366785,
+      "epoch": 0.6348751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 6.834466703907729e-06,
-      "loss": 0.8407,
+      "learning_rate": 6.214598423543541e-06,
+      "loss": 0.7459,
       "step": 22373
     },
     {
-      "epoch": 0.6144846337645217,
+      "epoch": 0.6349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 6.833622944034286e-06,
-      "loss": 0.9272,
+      "learning_rate": 6.2137477547792116e-06,
+      "loss": 0.8632,
       "step": 22374
     },
     {
-      "epoch": 0.614512097992365,
+      "epoch": 0.6349318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 6.832779209214445e-06,
-      "loss": 0.8104,
+      "learning_rate": 6.212897117997523e-06,
+      "loss": 0.7712,
       "step": 22375
     },
     {
-      "epoch": 0.6145395622202082,
+      "epoch": 0.6349602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 6.83193549945488e-06,
-      "loss": 0.8181,
+      "learning_rate": 6.212046513205662e-06,
+      "loss": 0.912,
       "step": 22376
     },
     {
-      "epoch": 0.6145670264480514,
+      "epoch": 0.634988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 6.8310918147622685e-06,
-      "loss": 0.8461,
+      "learning_rate": 6.211195940410814e-06,
+      "loss": 0.872,
       "step": 22377
     },
     {
-      "epoch": 0.6145944906758947,
+      "epoch": 0.635017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 6.830248155143289e-06,
-      "loss": 0.8178,
+      "learning_rate": 6.210345399620167e-06,
+      "loss": 0.78,
       "step": 22378
     },
     {
-      "epoch": 0.6146219549037378,
+      "epoch": 0.6350454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 6.829404520604614e-06,
-      "loss": 0.9275,
+      "learning_rate": 6.2094948908409024e-06,
+      "loss": 1.0054,
       "step": 22379
     },
     {
-      "epoch": 0.6146494191315811,
+      "epoch": 0.6350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 6.828560911152917e-06,
-      "loss": 0.8428,
+      "learning_rate": 6.208644414080203e-06,
+      "loss": 0.8244,
       "step": 22380
     },
     {
-      "epoch": 0.6146768833594244,
+      "epoch": 0.6351021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 6.827717326794873e-06,
-      "loss": 0.8678,
+      "learning_rate": 6.207793969345257e-06,
+      "loss": 0.839,
       "step": 22381
     },
     {
-      "epoch": 0.6147043475872676,
+      "epoch": 0.6351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 6.826873767537161e-06,
-      "loss": 0.9093,
+      "learning_rate": 6.206943556643246e-06,
+      "loss": 0.817,
       "step": 22382
     },
     {
-      "epoch": 0.6147318118151108,
+      "epoch": 0.6351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 6.826030233386449e-06,
-      "loss": 0.8627,
+      "learning_rate": 6.20609317598135e-06,
+      "loss": 0.812,
       "step": 22383
     },
     {
-      "epoch": 0.6147592760429541,
+      "epoch": 0.6351872871736662,
       "grad_norm": 0.0,
-      "learning_rate": 6.825186724349417e-06,
-      "loss": 0.8946,
+      "learning_rate": 6.2052428273667576e-06,
+      "loss": 0.7986,
       "step": 22384
     },
     {
-      "epoch": 0.6147867402707973,
+      "epoch": 0.6352156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 6.8243432404327425e-06,
-      "loss": 0.71,
+      "learning_rate": 6.204392510806652e-06,
+      "loss": 0.8493,
       "step": 22385
     },
     {
-      "epoch": 0.6148142044986405,
+      "epoch": 0.6352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 6.8234997816430885e-06,
-      "loss": 0.8476,
+      "learning_rate": 6.203542226308208e-06,
+      "loss": 0.8297,
       "step": 22386
     },
     {
-      "epoch": 0.6148416687264837,
+      "epoch": 0.6352724177071509,
       "grad_norm": 0.0,
-      "learning_rate": 6.822656347987137e-06,
-      "loss": 0.8077,
+      "learning_rate": 6.202691973878618e-06,
+      "loss": 0.8558,
       "step": 22387
     },
     {
-      "epoch": 0.614869132954327,
+      "epoch": 0.6353007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 6.821812939471559e-06,
-      "loss": 0.7957,
+      "learning_rate": 6.201841753525059e-06,
+      "loss": 0.7636,
       "step": 22388
     },
     {
-      "epoch": 0.6148965971821703,
+      "epoch": 0.6353291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 6.820969556103025e-06,
-      "loss": 0.8857,
+      "learning_rate": 6.200991565254711e-06,
+      "loss": 0.8862,
       "step": 22389
     },
     {
-      "epoch": 0.6149240614100134,
+      "epoch": 0.6353575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 6.820126197888216e-06,
-      "loss": 0.8339,
+      "learning_rate": 6.20014140907476e-06,
+      "loss": 0.9001,
       "step": 22390
     },
     {
-      "epoch": 0.6149515256378567,
+      "epoch": 0.6353859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 6.819282864833802e-06,
-      "loss": 0.9639,
+      "learning_rate": 6.199291284992386e-06,
+      "loss": 0.7754,
       "step": 22391
     },
     {
-      "epoch": 0.6149789898656999,
+      "epoch": 0.6354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 6.818439556946449e-06,
-      "loss": 0.9525,
+      "learning_rate": 6.198441193014765e-06,
+      "loss": 0.8879,
       "step": 22392
     },
     {
-      "epoch": 0.6150064540935432,
+      "epoch": 0.6354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 6.817596274232835e-06,
-      "loss": 0.7408,
+      "learning_rate": 6.197591133149083e-06,
+      "loss": 0.9203,
       "step": 22393
     },
     {
-      "epoch": 0.6150339183213864,
+      "epoch": 0.6354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 6.8167530166996335e-06,
-      "loss": 0.806,
+      "learning_rate": 6.196741105402524e-06,
+      "loss": 0.8558,
       "step": 22394
     },
     {
-      "epoch": 0.6150613825492296,
+      "epoch": 0.6354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 6.815909784353514e-06,
-      "loss": 0.7954,
+      "learning_rate": 6.195891109782261e-06,
+      "loss": 0.8079,
       "step": 22395
     },
     {
-      "epoch": 0.6150888467770729,
+      "epoch": 0.635527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 6.8150665772011495e-06,
-      "loss": 0.9143,
+      "learning_rate": 6.195041146295477e-06,
+      "loss": 0.8329,
       "step": 22396
     },
     {
-      "epoch": 0.6151163110049161,
+      "epoch": 0.6355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 6.814223395249215e-06,
-      "loss": 0.8801,
+      "learning_rate": 6.194191214949354e-06,
+      "loss": 0.7489,
       "step": 22397
     },
     {
-      "epoch": 0.6151437752327593,
+      "epoch": 0.6355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 6.813380238504377e-06,
-      "loss": 0.9325,
+      "learning_rate": 6.193341315751067e-06,
+      "loss": 0.8001,
       "step": 22398
     },
     {
-      "epoch": 0.6151712394606026,
+      "epoch": 0.6356129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 6.81253710697331e-06,
-      "loss": 0.916,
+      "learning_rate": 6.192491448707796e-06,
+      "loss": 0.8766,
       "step": 22399
     },
     {
-      "epoch": 0.6151987036884458,
+      "epoch": 0.6356413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 6.811694000662682e-06,
-      "loss": 0.8553,
+      "learning_rate": 6.191641613826723e-06,
+      "loss": 0.7907,
       "step": 22400
     },
     {
-      "epoch": 0.615226167916289,
+      "epoch": 0.6356696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 6.810850919579166e-06,
-      "loss": 0.832,
+      "learning_rate": 6.190791811115025e-06,
+      "loss": 0.8657,
       "step": 22401
     },
     {
-      "epoch": 0.6152536321441323,
+      "epoch": 0.6356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 6.810007863729435e-06,
-      "loss": 0.8487,
+      "learning_rate": 6.189942040579879e-06,
+      "loss": 0.8439,
       "step": 22402
     },
     {
-      "epoch": 0.6152810963719755,
+      "epoch": 0.6357264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 6.80916483312016e-06,
-      "loss": 0.7756,
+      "learning_rate": 6.189092302228461e-06,
+      "loss": 0.9291,
       "step": 22403
     },
     {
-      "epoch": 0.6153085605998188,
+      "epoch": 0.6357548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 6.808321827758004e-06,
-      "loss": 0.7485,
+      "learning_rate": 6.188242596067955e-06,
+      "loss": 0.8246,
       "step": 22404
     },
     {
-      "epoch": 0.6153360248276619,
+      "epoch": 0.6357832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 6.807478847649643e-06,
-      "loss": 0.8165,
+      "learning_rate": 6.187392922105535e-06,
+      "loss": 0.8795,
       "step": 22405
     },
     {
-      "epoch": 0.6153634890555052,
+      "epoch": 0.6358115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 6.8066358928017475e-06,
-      "loss": 0.8375,
+      "learning_rate": 6.1865432803483765e-06,
+      "loss": 0.8298,
       "step": 22406
     },
     {
-      "epoch": 0.6153909532833485,
+      "epoch": 0.6358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 6.8057929632209825e-06,
-      "loss": 0.8898,
+      "learning_rate": 6.185693670803654e-06,
+      "loss": 0.8887,
       "step": 22407
     },
     {
-      "epoch": 0.6154184175111916,
+      "epoch": 0.6358683314415438,
       "grad_norm": 0.0,
-      "learning_rate": 6.804950058914023e-06,
-      "loss": 0.8248,
+      "learning_rate": 6.184844093478554e-06,
+      "loss": 0.861,
       "step": 22408
     },
     {
-      "epoch": 0.6154458817390349,
+      "epoch": 0.6358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 6.8041071798875394e-06,
-      "loss": 0.9098,
+      "learning_rate": 6.183994548380249e-06,
+      "loss": 0.8917,
       "step": 22409
     },
     {
-      "epoch": 0.6154733459668782,
+      "epoch": 0.6359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 6.803264326148194e-06,
-      "loss": 0.9785,
+      "learning_rate": 6.183145035515909e-06,
+      "loss": 0.7939,
       "step": 22410
     },
     {
-      "epoch": 0.6155008101947214,
+      "epoch": 0.6359534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 6.80242149770266e-06,
-      "loss": 0.8846,
+      "learning_rate": 6.182295554892719e-06,
+      "loss": 0.9211,
       "step": 22411
     },
     {
-      "epoch": 0.6155282744225646,
+      "epoch": 0.6359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 6.801578694557605e-06,
-      "loss": 0.7648,
+      "learning_rate": 6.18144610651785e-06,
+      "loss": 0.8679,
       "step": 22412
     },
     {
-      "epoch": 0.6155557386504078,
+      "epoch": 0.6360102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 6.800735916719699e-06,
-      "loss": 0.7483,
+      "learning_rate": 6.180596690398474e-06,
+      "loss": 0.7776,
       "step": 22413
     },
     {
-      "epoch": 0.6155832028782511,
+      "epoch": 0.636038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 6.799893164195611e-06,
-      "loss": 0.8414,
+      "learning_rate": 6.179747306541771e-06,
+      "loss": 0.8197,
       "step": 22414
     },
     {
-      "epoch": 0.6156106671060944,
+      "epoch": 0.636066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 6.7990504369920095e-06,
-      "loss": 0.9058,
+      "learning_rate": 6.1788979549549155e-06,
+      "loss": 0.8372,
       "step": 22415
     },
     {
-      "epoch": 0.6156381313339375,
+      "epoch": 0.6360953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 6.798207735115558e-06,
-      "loss": 0.9709,
+      "learning_rate": 6.178048635645078e-06,
+      "loss": 0.7858,
       "step": 22416
     },
     {
-      "epoch": 0.6156655955617808,
+      "epoch": 0.6361237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 6.797365058572927e-06,
-      "loss": 0.8462,
+      "learning_rate": 6.177199348619438e-06,
+      "loss": 0.9256,
       "step": 22417
     },
     {
-      "epoch": 0.615693059789624,
+      "epoch": 0.6361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 6.796522407370785e-06,
-      "loss": 0.8701,
+      "learning_rate": 6.176350093885166e-06,
+      "loss": 1.0205,
       "step": 22418
     },
     {
-      "epoch": 0.6157205240174672,
+      "epoch": 0.6361804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 6.795679781515797e-06,
-      "loss": 0.8329,
+      "learning_rate": 6.1755008714494356e-06,
+      "loss": 0.8804,
       "step": 22419
     },
     {
-      "epoch": 0.6157479882453105,
+      "epoch": 0.6362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 6.794837181014633e-06,
-      "loss": 0.8845,
+      "learning_rate": 6.174651681319424e-06,
+      "loss": 0.9155,
       "step": 22420
     },
     {
-      "epoch": 0.6157754524731537,
+      "epoch": 0.6362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 6.793994605873961e-06,
-      "loss": 0.8078,
+      "learning_rate": 6.1738025235023e-06,
+      "loss": 0.8161,
       "step": 22421
     },
     {
-      "epoch": 0.615802916700997,
+      "epoch": 0.6362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 6.793152056100443e-06,
-      "loss": 0.8362,
+      "learning_rate": 6.172953398005234e-06,
+      "loss": 0.8632,
       "step": 22422
     },
     {
-      "epoch": 0.6158303809288402,
+      "epoch": 0.6362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 6.79230953170075e-06,
-      "loss": 0.8905,
+      "learning_rate": 6.1721043048354075e-06,
+      "loss": 0.807,
       "step": 22423
     },
     {
-      "epoch": 0.6158578451566834,
+      "epoch": 0.636322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 6.791467032681544e-06,
+      "learning_rate": 6.171255243999988e-06,
       "loss": 0.8704,
       "step": 22424
     },
     {
-      "epoch": 0.6158853093845267,
+      "epoch": 0.6363507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 6.790624559049493e-06,
-      "loss": 0.9525,
+      "learning_rate": 6.170406215506146e-06,
+      "loss": 0.8918,
       "step": 22425
     },
     {
-      "epoch": 0.6159127736123698,
+      "epoch": 0.6363791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 6.789782110811267e-06,
-      "loss": 0.8061,
+      "learning_rate": 6.169557219361056e-06,
+      "loss": 0.8048,
       "step": 22426
     },
     {
-      "epoch": 0.6159402378402131,
+      "epoch": 0.6364074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 6.7889396879735305e-06,
-      "loss": 0.8739,
+      "learning_rate": 6.168708255571887e-06,
+      "loss": 0.7415,
       "step": 22427
     },
     {
-      "epoch": 0.6159677020680564,
+      "epoch": 0.6364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 6.788097290542942e-06,
-      "loss": 0.873,
+      "learning_rate": 6.167859324145813e-06,
+      "loss": 0.8584,
       "step": 22428
     },
     {
-      "epoch": 0.6159951662958996,
+      "epoch": 0.6364642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 6.787254918526171e-06,
-      "loss": 0.9089,
+      "learning_rate": 6.1670104250900045e-06,
+      "loss": 0.8141,
       "step": 22429
     },
     {
-      "epoch": 0.6160226305237428,
+      "epoch": 0.6364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 6.786412571929888e-06,
-      "loss": 0.8438,
+      "learning_rate": 6.1661615584116276e-06,
+      "loss": 0.8336,
       "step": 22430
     },
     {
-      "epoch": 0.616050094751586,
+      "epoch": 0.6365209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 6.785570250760751e-06,
-      "loss": 0.8352,
+      "learning_rate": 6.1653127241178605e-06,
+      "loss": 0.7585,
       "step": 22431
     },
     {
-      "epoch": 0.6160775589794293,
+      "epoch": 0.6365493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 6.784727955025431e-06,
-      "loss": 0.8258,
+      "learning_rate": 6.164463922215867e-06,
+      "loss": 0.7613,
       "step": 22432
     },
     {
-      "epoch": 0.6161050232072726,
+      "epoch": 0.636577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.783885684730584e-06,
-      "loss": 0.9249,
+      "learning_rate": 6.1636151527128185e-06,
+      "loss": 0.8752,
       "step": 22433
     },
     {
-      "epoch": 0.6161324874351157,
+      "epoch": 0.6366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 6.78304343988288e-06,
-      "loss": 0.8154,
+      "learning_rate": 6.162766415615887e-06,
+      "loss": 0.7701,
       "step": 22434
     },
     {
-      "epoch": 0.616159951662959,
+      "epoch": 0.6366345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 6.782201220488985e-06,
-      "loss": 0.838,
+      "learning_rate": 6.1619177109322406e-06,
+      "loss": 0.7826,
       "step": 22435
     },
     {
-      "epoch": 0.6161874158908023,
+      "epoch": 0.6366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 6.781359026555558e-06,
-      "loss": 0.8386,
+      "learning_rate": 6.161069038669045e-06,
+      "loss": 0.8583,
       "step": 22436
     },
     {
-      "epoch": 0.6162148801186454,
+      "epoch": 0.6366912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 6.780516858089266e-06,
-      "loss": 0.8856,
+      "learning_rate": 6.160220398833474e-06,
+      "loss": 0.7549,
       "step": 22437
     },
     {
-      "epoch": 0.6162423443464887,
+      "epoch": 0.6367196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 6.779674715096773e-06,
-      "loss": 0.8141,
+      "learning_rate": 6.15937179143269e-06,
+      "loss": 0.835,
       "step": 22438
     },
     {
-      "epoch": 0.6162698085743319,
+      "epoch": 0.6367480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 6.778832597584741e-06,
-      "loss": 0.8125,
+      "learning_rate": 6.158523216473868e-06,
+      "loss": 0.8389,
       "step": 22439
     },
     {
-      "epoch": 0.6162972728021752,
+      "epoch": 0.6367763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 6.777990505559832e-06,
-      "loss": 0.9581,
+      "learning_rate": 6.157674673964174e-06,
+      "loss": 0.8163,
       "step": 22440
     },
     {
-      "epoch": 0.6163247370300184,
+      "epoch": 0.6368047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 6.777148439028709e-06,
-      "loss": 0.9598,
+      "learning_rate": 6.156826163910774e-06,
+      "loss": 0.8471,
       "step": 22441
     },
     {
-      "epoch": 0.6163522012578616,
+      "epoch": 0.63683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 6.776306397998038e-06,
-      "loss": 0.7896,
+      "learning_rate": 6.155977686320837e-06,
+      "loss": 0.8275,
       "step": 22442
     },
     {
-      "epoch": 0.6163796654857049,
+      "epoch": 0.636861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 6.775464382474478e-06,
-      "loss": 0.7771,
+      "learning_rate": 6.15512924120153e-06,
+      "loss": 0.8338,
       "step": 22443
     },
     {
-      "epoch": 0.616407129713548,
+      "epoch": 0.6368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 6.774622392464697e-06,
-      "loss": 0.7946,
+      "learning_rate": 6.154280828560017e-06,
+      "loss": 0.8736,
       "step": 22444
     },
     {
-      "epoch": 0.6164345939413913,
+      "epoch": 0.6369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 6.77378042797535e-06,
-      "loss": 0.8839,
+      "learning_rate": 6.1534324484034666e-06,
+      "loss": 0.8416,
       "step": 22445
     },
     {
-      "epoch": 0.6164620581692346,
+      "epoch": 0.6369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 6.772938489013099e-06,
-      "loss": 0.841,
+      "learning_rate": 6.152584100739048e-06,
+      "loss": 0.7876,
       "step": 22446
     },
     {
-      "epoch": 0.6164895223970778,
+      "epoch": 0.6369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 6.772096575584613e-06,
-      "loss": 0.7364,
+      "learning_rate": 6.151735785573921e-06,
+      "loss": 0.8611,
       "step": 22447
     },
     {
-      "epoch": 0.616516986624921,
+      "epoch": 0.6370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 6.771254687696547e-06,
-      "loss": 0.8371,
+      "learning_rate": 6.150887502915257e-06,
+      "loss": 0.8289,
       "step": 22448
     },
     {
-      "epoch": 0.6165444508527643,
+      "epoch": 0.6370317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 6.770412825355563e-06,
-      "loss": 0.9053,
+      "learning_rate": 6.150039252770219e-06,
+      "loss": 0.8385,
       "step": 22449
     },
     {
-      "epoch": 0.6165719150806075,
+      "epoch": 0.6370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 6.76957098856833e-06,
-      "loss": 0.8559,
+      "learning_rate": 6.149191035145972e-06,
+      "loss": 0.8476,
       "step": 22450
     },
     {
-      "epoch": 0.6165993793084508,
+      "epoch": 0.637088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 6.7687291773414995e-06,
-      "loss": 0.85,
+      "learning_rate": 6.148342850049682e-06,
+      "loss": 0.7965,
       "step": 22451
     },
     {
-      "epoch": 0.6166268435362939,
+      "epoch": 0.6371169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 6.767887391681733e-06,
-      "loss": 0.7262,
+      "learning_rate": 6.147494697488514e-06,
+      "loss": 0.8331,
       "step": 22452
     },
     {
-      "epoch": 0.6166543077641372,
+      "epoch": 0.6371452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 6.767045631595694e-06,
-      "loss": 0.8795,
+      "learning_rate": 6.146646577469626e-06,
+      "loss": 0.8691,
       "step": 22453
     },
     {
-      "epoch": 0.6166817719919805,
+      "epoch": 0.6371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 6.7662038970900444e-06,
-      "loss": 0.9442,
+      "learning_rate": 6.145798490000191e-06,
+      "loss": 0.8051,
       "step": 22454
     },
     {
-      "epoch": 0.6167092362198237,
+      "epoch": 0.6372020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 6.7653621881714404e-06,
-      "loss": 0.8575,
+      "learning_rate": 6.144950435087371e-06,
+      "loss": 0.7708,
       "step": 22455
     },
     {
-      "epoch": 0.6167367004476669,
+      "epoch": 0.6372304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 6.764520504846548e-06,
-      "loss": 0.813,
+      "learning_rate": 6.144102412738327e-06,
+      "loss": 0.953,
       "step": 22456
     },
     {
-      "epoch": 0.6167641646755101,
+      "epoch": 0.6372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 6.763678847122019e-06,
-      "loss": 0.8331,
+      "learning_rate": 6.143254422960223e-06,
+      "loss": 0.8967,
       "step": 22457
     },
     {
-      "epoch": 0.6167916289033534,
+      "epoch": 0.6372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 6.762837215004514e-06,
-      "loss": 0.8167,
+      "learning_rate": 6.142406465760223e-06,
+      "loss": 0.8049,
       "step": 22458
     },
     {
-      "epoch": 0.6168190931311966,
+      "epoch": 0.6373155505107831,
       "grad_norm": 0.0,
-      "learning_rate": 6.7619956085007e-06,
-      "loss": 0.9055,
+      "learning_rate": 6.141558541145489e-06,
+      "loss": 0.7628,
       "step": 22459
     },
     {
-      "epoch": 0.6168465573590398,
+      "epoch": 0.6373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 6.761154027617227e-06,
-      "loss": 0.7706,
+      "learning_rate": 6.140710649123183e-06,
+      "loss": 0.8617,
       "step": 22460
     },
     {
-      "epoch": 0.6168740215868831,
+      "epoch": 0.637372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 6.76031247236076e-06,
-      "loss": 0.8357,
+      "learning_rate": 6.139862789700468e-06,
+      "loss": 0.9392,
       "step": 22461
     },
     {
-      "epoch": 0.6169014858147264,
+      "epoch": 0.6374006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 6.759470942737958e-06,
-      "loss": 0.9476,
+      "learning_rate": 6.139014962884506e-06,
+      "loss": 0.8982,
       "step": 22462
     },
     {
-      "epoch": 0.6169289500425695,
+      "epoch": 0.6374290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 6.758629438755475e-06,
-      "loss": 0.8106,
+      "learning_rate": 6.138167168682456e-06,
+      "loss": 0.781,
       "step": 22463
     },
     {
-      "epoch": 0.6169564142704128,
+      "epoch": 0.6374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 6.75778796041997e-06,
-      "loss": 0.7609,
+      "learning_rate": 6.137319407101484e-06,
+      "loss": 0.9218,
       "step": 22464
     },
     {
-      "epoch": 0.616983878498256,
+      "epoch": 0.6374858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 6.756946507738103e-06,
-      "loss": 0.8778,
+      "learning_rate": 6.136471678148748e-06,
+      "loss": 0.7121,
       "step": 22465
     },
     {
-      "epoch": 0.6170113427260993,
+      "epoch": 0.6375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 6.756105080716533e-06,
-      "loss": 0.9303,
+      "learning_rate": 6.1356239818314075e-06,
+      "loss": 0.8858,
       "step": 22466
     },
     {
-      "epoch": 0.6170388069539425,
+      "epoch": 0.6375425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 6.755263679361914e-06,
-      "loss": 0.7808,
+      "learning_rate": 6.134776318156627e-06,
+      "loss": 0.8281,
       "step": 22467
     },
     {
-      "epoch": 0.6170662711817857,
+      "epoch": 0.6375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 6.75442230368091e-06,
-      "loss": 0.8786,
+      "learning_rate": 6.1339286871315606e-06,
+      "loss": 0.8412,
       "step": 22468
     },
     {
-      "epoch": 0.617093735409629,
+      "epoch": 0.6375993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 6.75358095368017e-06,
-      "loss": 0.844,
+      "learning_rate": 6.133081088763377e-06,
+      "loss": 0.8727,
       "step": 22469
     },
     {
-      "epoch": 0.6171211996374721,
+      "epoch": 0.637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 6.752739629366354e-06,
-      "loss": 0.7795,
+      "learning_rate": 6.132233523059232e-06,
+      "loss": 0.9073,
       "step": 22470
     },
     {
-      "epoch": 0.6171486638653154,
+      "epoch": 0.6376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 6.751898330746123e-06,
-      "loss": 0.8891,
+      "learning_rate": 6.1313859900262805e-06,
+      "loss": 0.8202,
       "step": 22471
     },
     {
-      "epoch": 0.6171761280931587,
+      "epoch": 0.6376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 6.751057057826128e-06,
-      "loss": 0.8461,
+      "learning_rate": 6.130538489671689e-06,
+      "loss": 0.851,
       "step": 22472
     },
     {
-      "epoch": 0.6172035923210019,
+      "epoch": 0.6377128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 6.750215810613028e-06,
-      "loss": 0.8373,
+      "learning_rate": 6.12969102200261e-06,
+      "loss": 0.7993,
       "step": 22473
     },
     {
-      "epoch": 0.6172310565488451,
+      "epoch": 0.6377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 6.749374589113484e-06,
-      "loss": 0.8493,
+      "learning_rate": 6.128843587026207e-06,
+      "loss": 0.8268,
       "step": 22474
     },
     {
-      "epoch": 0.6172585207766884,
+      "epoch": 0.6377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 6.748533393334145e-06,
-      "loss": 0.856,
+      "learning_rate": 6.127996184749636e-06,
+      "loss": 0.8143,
       "step": 22475
     },
     {
-      "epoch": 0.6172859850045316,
+      "epoch": 0.6377979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 6.747692223281667e-06,
-      "loss": 0.9169,
+      "learning_rate": 6.127148815180055e-06,
+      "loss": 0.7884,
       "step": 22476
     },
     {
-      "epoch": 0.6173134492323749,
+      "epoch": 0.6378263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 6.746851078962709e-06,
-      "loss": 0.7681,
+      "learning_rate": 6.126301478324621e-06,
+      "loss": 0.8048,
       "step": 22477
     },
     {
-      "epoch": 0.617340913460218,
+      "epoch": 0.6378547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 6.746009960383928e-06,
-      "loss": 0.8637,
+      "learning_rate": 6.125454174190493e-06,
+      "loss": 0.8667,
       "step": 22478
     },
     {
-      "epoch": 0.6173683776880613,
+      "epoch": 0.637883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 6.7451688675519725e-06,
-      "loss": 0.923,
+      "learning_rate": 6.124606902784829e-06,
+      "loss": 0.7983,
       "step": 22479
     },
     {
-      "epoch": 0.6173958419159046,
+      "epoch": 0.637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 6.744327800473507e-06,
-      "loss": 0.8086,
+      "learning_rate": 6.123759664114783e-06,
+      "loss": 0.8739,
       "step": 22480
     },
     {
-      "epoch": 0.6174233061437477,
+      "epoch": 0.6379398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 6.743486759155177e-06,
-      "loss": 0.8441,
+      "learning_rate": 6.122912458187516e-06,
+      "loss": 0.7595,
       "step": 22481
     },
     {
-      "epoch": 0.617450770371591,
+      "epoch": 0.6379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 6.742645743603641e-06,
-      "loss": 0.8276,
+      "learning_rate": 6.122065285010179e-06,
+      "loss": 0.9049,
       "step": 22482
     },
     {
-      "epoch": 0.6174782345994342,
+      "epoch": 0.6379965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 6.741804753825556e-06,
-      "loss": 0.9091,
+      "learning_rate": 6.121218144589931e-06,
+      "loss": 0.7959,
       "step": 22483
     },
     {
-      "epoch": 0.6175056988272775,
+      "epoch": 0.6380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 6.74096378982757e-06,
-      "loss": 0.9026,
+      "learning_rate": 6.120371036933928e-06,
+      "loss": 0.7564,
       "step": 22484
     },
     {
-      "epoch": 0.6175331630551207,
+      "epoch": 0.6380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 6.740122851616344e-06,
-      "loss": 0.9533,
+      "learning_rate": 6.119523962049323e-06,
+      "loss": 0.8603,
       "step": 22485
     },
     {
-      "epoch": 0.6175606272829639,
+      "epoch": 0.6380817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 6.739281939198531e-06,
-      "loss": 0.9165,
+      "learning_rate": 6.118676919943277e-06,
+      "loss": 0.7702,
       "step": 22486
     },
     {
-      "epoch": 0.6175880915108072,
+      "epoch": 0.6381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 6.738441052580781e-06,
-      "loss": 0.8305,
+      "learning_rate": 6.117829910622942e-06,
+      "loss": 0.7907,
       "step": 22487
     },
     {
-      "epoch": 0.6176155557386505,
+      "epoch": 0.6381384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 6.737600191769746e-06,
-      "loss": 0.9115,
+      "learning_rate": 6.11698293409547e-06,
+      "loss": 0.8261,
       "step": 22488
     },
     {
-      "epoch": 0.6176430199664936,
+      "epoch": 0.6381668558456299,
       "grad_norm": 0.0,
-      "learning_rate": 6.736759356772082e-06,
-      "loss": 0.8296,
+      "learning_rate": 6.11613599036802e-06,
+      "loss": 0.8398,
       "step": 22489
     },
     {
-      "epoch": 0.6176704841943369,
+      "epoch": 0.6381952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 6.735918547594444e-06,
-      "loss": 0.9272,
+      "learning_rate": 6.115289079447743e-06,
+      "loss": 0.8611,
       "step": 22490
     },
     {
-      "epoch": 0.6176979484221801,
+      "epoch": 0.6382236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 6.735077764243486e-06,
-      "loss": 0.7829,
+      "learning_rate": 6.114442201341792e-06,
+      "loss": 0.8115,
       "step": 22491
     },
     {
-      "epoch": 0.6177254126500233,
+      "epoch": 0.6382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 6.734237006725853e-06,
-      "loss": 0.7857,
+      "learning_rate": 6.1135953560573245e-06,
+      "loss": 0.8109,
       "step": 22492
     },
     {
-      "epoch": 0.6177528768778666,
+      "epoch": 0.6382803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 6.7333962750482e-06,
-      "loss": 0.9046,
+      "learning_rate": 6.112748543601492e-06,
+      "loss": 0.8164,
       "step": 22493
     },
     {
-      "epoch": 0.6177803411057098,
+      "epoch": 0.6383087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 6.732555569217184e-06,
-      "loss": 0.8744,
+      "learning_rate": 6.111901763981445e-06,
+      "loss": 0.8705,
       "step": 22494
     },
     {
-      "epoch": 0.6178078053335531,
+      "epoch": 0.6383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 6.731714889239455e-06,
-      "loss": 0.8393,
+      "learning_rate": 6.11105501720434e-06,
+      "loss": 0.801,
       "step": 22495
     },
     {
-      "epoch": 0.6178352695613962,
+      "epoch": 0.6383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 6.73087423512166e-06,
-      "loss": 0.8266,
+      "learning_rate": 6.110208303277329e-06,
+      "loss": 0.8068,
       "step": 22496
     },
     {
-      "epoch": 0.6178627337892395,
+      "epoch": 0.6383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 6.730033606870461e-06,
-      "loss": 0.8818,
+      "learning_rate": 6.109361622207561e-06,
+      "loss": 0.8074,
       "step": 22497
     },
     {
-      "epoch": 0.6178901980170828,
+      "epoch": 0.638422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 6.729193004492498e-06,
-      "loss": 0.8838,
+      "learning_rate": 6.108514974002193e-06,
+      "loss": 0.8088,
       "step": 22498
     },
     {
-      "epoch": 0.617917662244926,
+      "epoch": 0.6384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 6.728352427994429e-06,
-      "loss": 0.8356,
+      "learning_rate": 6.107668358668368e-06,
+      "loss": 0.8864,
       "step": 22499
     },
     {
-      "epoch": 0.6179451264727692,
+      "epoch": 0.6384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 6.7275118773829e-06,
-      "loss": 0.699,
+      "learning_rate": 6.10682177621325e-06,
+      "loss": 0.7859,
       "step": 22500
     },
     {
-      "epoch": 0.6179725907006125,
+      "epoch": 0.6385073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 6.726671352664564e-06,
-      "loss": 0.8363,
+      "learning_rate": 6.105975226643979e-06,
+      "loss": 0.9326,
       "step": 22501
     },
     {
-      "epoch": 0.6180000549284557,
+      "epoch": 0.6385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 6.725830853846076e-06,
-      "loss": 0.9431,
+      "learning_rate": 6.105128709967714e-06,
+      "loss": 0.8448,
       "step": 22502
     },
     {
-      "epoch": 0.6180275191562989,
+      "epoch": 0.6385641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 6.7249903809340845e-06,
-      "loss": 0.9106,
+      "learning_rate": 6.104282226191601e-06,
+      "loss": 0.7061,
       "step": 22503
     },
     {
-      "epoch": 0.6180549833841421,
+      "epoch": 0.6385925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 6.724149933935233e-06,
-      "loss": 0.8206,
+      "learning_rate": 6.1034357753227904e-06,
+      "loss": 0.7853,
       "step": 22504
     },
     {
-      "epoch": 0.6180824476119854,
+      "epoch": 0.6386208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 6.723309512856175e-06,
-      "loss": 0.8487,
+      "learning_rate": 6.102589357368434e-06,
+      "loss": 0.7807,
       "step": 22505
     },
     {
-      "epoch": 0.6181099118398287,
+      "epoch": 0.6386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 6.722469117703565e-06,
-      "loss": 0.8422,
+      "learning_rate": 6.10174297233568e-06,
+      "loss": 0.8284,
       "step": 22506
     },
     {
-      "epoch": 0.6181373760676718,
+      "epoch": 0.6386776390465381,
       "grad_norm": 0.0,
-      "learning_rate": 6.721628748484048e-06,
-      "loss": 0.884,
+      "learning_rate": 6.1008966202316766e-06,
+      "loss": 0.8844,
       "step": 22507
     },
     {
-      "epoch": 0.6181648402955151,
+      "epoch": 0.6387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 6.7207884052042724e-06,
-      "loss": 0.8752,
+      "learning_rate": 6.100050301063577e-06,
+      "loss": 0.8172,
       "step": 22508
     },
     {
-      "epoch": 0.6181923045233583,
+      "epoch": 0.6387343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 6.719948087870893e-06,
-      "loss": 0.9061,
+      "learning_rate": 6.099204014838528e-06,
+      "loss": 0.7675,
       "step": 22509
     },
     {
-      "epoch": 0.6182197687512015,
+      "epoch": 0.6387627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 6.719107796490551e-06,
-      "loss": 0.7812,
+      "learning_rate": 6.098357761563678e-06,
+      "loss": 0.857,
       "step": 22510
     },
     {
-      "epoch": 0.6182472329790448,
+      "epoch": 0.6387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 6.718267531069903e-06,
-      "loss": 0.8467,
+      "learning_rate": 6.097511541246173e-06,
+      "loss": 0.8745,
       "step": 22511
     },
     {
-      "epoch": 0.618274697206888,
+      "epoch": 0.6388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 6.71742729161559e-06,
-      "loss": 0.8785,
+      "learning_rate": 6.096665353893165e-06,
+      "loss": 0.8519,
       "step": 22512
     },
     {
-      "epoch": 0.6183021614347313,
+      "epoch": 0.6388479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 6.716587078134263e-06,
-      "loss": 0.9199,
+      "learning_rate": 6.095819199511801e-06,
+      "loss": 0.8832,
       "step": 22513
     },
     {
-      "epoch": 0.6183296256625745,
+      "epoch": 0.6388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 6.715746890632575e-06,
-      "loss": 0.8609,
+      "learning_rate": 6.094973078109223e-06,
+      "loss": 0.8897,
       "step": 22514
     },
     {
-      "epoch": 0.6183570898904177,
+      "epoch": 0.6389046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 6.71490672911717e-06,
-      "loss": 0.9191,
+      "learning_rate": 6.094126989692586e-06,
+      "loss": 0.9424,
       "step": 22515
     },
     {
-      "epoch": 0.618384554118261,
+      "epoch": 0.638933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 6.714066593594693e-06,
-      "loss": 0.8923,
+      "learning_rate": 6.093280934269036e-06,
+      "loss": 0.6844,
       "step": 22516
     },
     {
-      "epoch": 0.6184120183461042,
+      "epoch": 0.638961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 6.713226484071793e-06,
-      "loss": 0.8662,
+      "learning_rate": 6.092434911845717e-06,
+      "loss": 0.8891,
       "step": 22517
     },
     {
-      "epoch": 0.6184394825739474,
+      "epoch": 0.6389897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 6.71238640055512e-06,
-      "loss": 0.8889,
+      "learning_rate": 6.091588922429774e-06,
+      "loss": 0.8698,
       "step": 22518
     },
     {
-      "epoch": 0.6184669468017907,
+      "epoch": 0.6390181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 6.711546343051318e-06,
-      "loss": 0.8321,
+      "learning_rate": 6.090742966028357e-06,
+      "loss": 0.8103,
       "step": 22519
     },
     {
-      "epoch": 0.6184944110296339,
+      "epoch": 0.6390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 6.710706311567035e-06,
-      "loss": 0.9342,
+      "learning_rate": 6.0898970426486095e-06,
+      "loss": 0.8348,
       "step": 22520
     },
     {
-      "epoch": 0.6185218752574771,
+      "epoch": 0.6390749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 6.709866306108921e-06,
-      "loss": 0.8492,
+      "learning_rate": 6.089051152297676e-06,
+      "loss": 0.7298,
       "step": 22521
     },
     {
-      "epoch": 0.6185493394853203,
+      "epoch": 0.6391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 6.709026326683615e-06,
-      "loss": 0.8736,
+      "learning_rate": 6.088205294982705e-06,
+      "loss": 0.8323,
       "step": 22522
     },
     {
-      "epoch": 0.6185768037131636,
+      "epoch": 0.6391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.708186373297771e-06,
-      "loss": 0.9413,
+      "learning_rate": 6.087359470710841e-06,
+      "loss": 0.8778,
       "step": 22523
     },
     {
-      "epoch": 0.6186042679410069,
+      "epoch": 0.6391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 6.70734644595803e-06,
-      "loss": 0.8917,
+      "learning_rate": 6.086513679489224e-06,
+      "loss": 0.7892,
       "step": 22524
     },
     {
-      "epoch": 0.61863173216885,
+      "epoch": 0.6391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 6.7065065446710365e-06,
-      "loss": 0.8612,
+      "learning_rate": 6.085667921325005e-06,
+      "loss": 0.8753,
       "step": 22525
     },
     {
-      "epoch": 0.6186591963966933,
+      "epoch": 0.639216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 6.705666669443444e-06,
-      "loss": 0.8773,
+      "learning_rate": 6.084822196225323e-06,
+      "loss": 0.8701,
       "step": 22526
     },
     {
-      "epoch": 0.6186866606245366,
+      "epoch": 0.6392451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 6.7048268202818935e-06,
-      "loss": 0.8417,
+      "learning_rate": 6.083976504197323e-06,
+      "loss": 0.9416,
       "step": 22527
     },
     {
-      "epoch": 0.6187141248523798,
+      "epoch": 0.6392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 6.703986997193026e-06,
-      "loss": 0.8597,
+      "learning_rate": 6.083130845248152e-06,
+      "loss": 0.7599,
       "step": 22528
     },
     {
-      "epoch": 0.618741589080223,
+      "epoch": 0.6393019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 6.703147200183489e-06,
-      "loss": 0.8406,
+      "learning_rate": 6.08228521938495e-06,
+      "loss": 0.8162,
       "step": 22529
     },
     {
-      "epoch": 0.6187690533080662,
+      "epoch": 0.6393303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 6.7023074292599315e-06,
-      "loss": 0.8373,
+      "learning_rate": 6.081439626614858e-06,
+      "loss": 0.7779,
       "step": 22530
     },
     {
-      "epoch": 0.6187965175359095,
+      "epoch": 0.6393586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 6.701467684428993e-06,
-      "loss": 0.9199,
+      "learning_rate": 6.080594066945025e-06,
+      "loss": 0.8957,
       "step": 22531
     },
     {
-      "epoch": 0.6188239817637528,
+      "epoch": 0.6393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 6.700627965697319e-06,
-      "loss": 0.8648,
+      "learning_rate": 6.079748540382588e-06,
+      "loss": 0.8291,
       "step": 22532
     },
     {
-      "epoch": 0.6188514459915959,
+      "epoch": 0.6394154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 6.69978827307156e-06,
-      "loss": 0.7589,
+      "learning_rate": 6.078903046934692e-06,
+      "loss": 0.7704,
       "step": 22533
     },
     {
-      "epoch": 0.6188789102194392,
+      "epoch": 0.6394438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 6.698948606558351e-06,
-      "loss": 0.7931,
+      "learning_rate": 6.078057586608481e-06,
+      "loss": 0.8566,
       "step": 22534
     },
     {
-      "epoch": 0.6189063744472824,
+      "epoch": 0.639472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 6.69810896616434e-06,
-      "loss": 0.8589,
+      "learning_rate": 6.07721215941109e-06,
+      "loss": 0.7609,
       "step": 22535
     },
     {
-      "epoch": 0.6189338386751256,
+      "epoch": 0.6395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 6.697269351896167e-06,
-      "loss": 0.9524,
+      "learning_rate": 6.076366765349666e-06,
+      "loss": 0.7949,
       "step": 22536
     },
     {
-      "epoch": 0.6189613029029689,
+      "epoch": 0.6395289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 6.696429763760479e-06,
-      "loss": 0.8215,
+      "learning_rate": 6.0755214044313505e-06,
+      "loss": 0.9227,
       "step": 22537
     },
     {
-      "epoch": 0.6189887671308121,
+      "epoch": 0.6395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 6.6955902017639215e-06,
-      "loss": 0.8377,
+      "learning_rate": 6.074676076663277e-06,
+      "loss": 0.7714,
       "step": 22538
     },
     {
-      "epoch": 0.6190162313586554,
+      "epoch": 0.6395856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 6.694750665913137e-06,
-      "loss": 0.9064,
+      "learning_rate": 6.073830782052595e-06,
+      "loss": 0.9054,
       "step": 22539
     },
     {
-      "epoch": 0.6190436955864986,
+      "epoch": 0.6396140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 6.693911156214758e-06,
-      "loss": 0.8486,
+      "learning_rate": 6.07298552060644e-06,
+      "loss": 0.9229,
       "step": 22540
     },
     {
-      "epoch": 0.6190711598143418,
+      "epoch": 0.6396424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 6.693071672675438e-06,
-      "loss": 0.7421,
+      "learning_rate": 6.072140292331951e-06,
+      "loss": 0.8667,
       "step": 22541
     },
     {
-      "epoch": 0.6190986240421851,
+      "epoch": 0.6396708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 6.692232215301817e-06,
-      "loss": 0.8131,
+      "learning_rate": 6.071295097236274e-06,
+      "loss": 0.8455,
       "step": 22542
     },
     {
-      "epoch": 0.6191260882700282,
+      "epoch": 0.6396992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 6.691392784100534e-06,
-      "loss": 0.8589,
+      "learning_rate": 6.070449935326541e-06,
+      "loss": 0.8182,
       "step": 22543
     },
     {
-      "epoch": 0.6191535524978715,
+      "epoch": 0.639727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 6.6905533790782315e-06,
-      "loss": 0.9251,
+      "learning_rate": 6.0696048066098935e-06,
+      "loss": 0.7535,
       "step": 22544
     },
     {
-      "epoch": 0.6191810167257148,
+      "epoch": 0.639755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 6.689714000241558e-06,
-      "loss": 0.8543,
+      "learning_rate": 6.068759711093468e-06,
+      "loss": 0.8565,
       "step": 22545
     },
     {
-      "epoch": 0.619208480953558,
+      "epoch": 0.6397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 6.688874647597145e-06,
-      "loss": 0.8654,
+      "learning_rate": 6.06791464878441e-06,
+      "loss": 0.9566,
       "step": 22546
     },
     {
-      "epoch": 0.6192359451814012,
+      "epoch": 0.6398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 6.688035321151642e-06,
-      "loss": 0.9508,
+      "learning_rate": 6.067069619689854e-06,
+      "loss": 0.9212,
       "step": 22547
     },
     {
-      "epoch": 0.6192634094092445,
+      "epoch": 0.6398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 6.687196020911682e-06,
-      "loss": 0.7793,
+      "learning_rate": 6.066224623816937e-06,
+      "loss": 0.9188,
       "step": 22548
     },
     {
-      "epoch": 0.6192908736370877,
+      "epoch": 0.6398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 6.6863567468839116e-06,
-      "loss": 0.8357,
+      "learning_rate": 6.065379661172797e-06,
+      "loss": 0.8584,
       "step": 22549
     },
     {
-      "epoch": 0.619318337864931,
+      "epoch": 0.6398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 6.685517499074975e-06,
-      "loss": 0.8092,
+      "learning_rate": 6.0645347317645735e-06,
+      "loss": 0.9199,
       "step": 22550
     },
     {
-      "epoch": 0.6193458020927741,
+      "epoch": 0.6399262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 6.684678277491506e-06,
-      "loss": 0.8607,
+      "learning_rate": 6.063689835599401e-06,
+      "loss": 0.7768,
       "step": 22551
     },
     {
-      "epoch": 0.6193732663206174,
+      "epoch": 0.6399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 6.683839082140142e-06,
-      "loss": 0.8172,
+      "learning_rate": 6.062844972684416e-06,
+      "loss": 0.8558,
       "step": 22552
     },
     {
-      "epoch": 0.6194007305484607,
+      "epoch": 0.639982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 6.6829999130275305e-06,
-      "loss": 0.8031,
+      "learning_rate": 6.0620001430267585e-06,
+      "loss": 0.7978,
       "step": 22553
     },
     {
-      "epoch": 0.6194281947763038,
+      "epoch": 0.6400113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 6.6821607701603086e-06,
-      "loss": 1.0292,
+      "learning_rate": 6.061155346633563e-06,
+      "loss": 0.9729,
       "step": 22554
     },
     {
-      "epoch": 0.6194556590041471,
+      "epoch": 0.6400397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 6.681321653545115e-06,
-      "loss": 0.8898,
+      "learning_rate": 6.060310583511964e-06,
+      "loss": 0.8365,
       "step": 22555
     },
     {
-      "epoch": 0.6194831232319903,
+      "epoch": 0.6400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 6.680482563188595e-06,
-      "loss": 0.8625,
+      "learning_rate": 6.059465853669098e-06,
+      "loss": 0.7974,
       "step": 22556
     },
     {
-      "epoch": 0.6195105874598336,
+      "epoch": 0.6400964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 6.6796434990973765e-06,
-      "loss": 0.8168,
+      "learning_rate": 6.058621157112103e-06,
+      "loss": 0.8024,
       "step": 22557
     },
     {
-      "epoch": 0.6195380516876768,
+      "epoch": 0.6401248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 6.678804461278106e-06,
-      "loss": 0.7277,
+      "learning_rate": 6.057776493848111e-06,
+      "loss": 0.8321,
       "step": 22558
     },
     {
-      "epoch": 0.61956551591552,
+      "epoch": 0.6401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 6.677965449737423e-06,
-      "loss": 0.7577,
+      "learning_rate": 6.056931863884259e-06,
+      "loss": 0.8799,
       "step": 22559
     },
     {
-      "epoch": 0.6195929801433633,
+      "epoch": 0.6401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 6.677126464481962e-06,
-      "loss": 0.7277,
+      "learning_rate": 6.056087267227677e-06,
+      "loss": 0.8485,
       "step": 22560
     },
     {
-      "epoch": 0.6196204443712066,
+      "epoch": 0.6402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 6.676287505518364e-06,
-      "loss": 0.7872,
+      "learning_rate": 6.0552427038855065e-06,
+      "loss": 0.8722,
       "step": 22561
     },
     {
-      "epoch": 0.6196479085990497,
+      "epoch": 0.6402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 6.6754485728532714e-06,
-      "loss": 0.7752,
+      "learning_rate": 6.0543981738648774e-06,
+      "loss": 0.8451,
       "step": 22562
     },
     {
-      "epoch": 0.619675372826893,
+      "epoch": 0.640266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 6.674609666493314e-06,
-      "loss": 0.9068,
+      "learning_rate": 6.0535536771729255e-06,
+      "loss": 0.8879,
       "step": 22563
     },
     {
-      "epoch": 0.6197028370547362,
+      "epoch": 0.6402951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 6.673770786445133e-06,
-      "loss": 0.799,
+      "learning_rate": 6.052709213816783e-06,
+      "loss": 0.8149,
       "step": 22564
     },
     {
-      "epoch": 0.6197303012825794,
+      "epoch": 0.6403234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 6.6729319327153645e-06,
-      "loss": 0.8638,
+      "learning_rate": 6.051864783803583e-06,
+      "loss": 0.8487,
       "step": 22565
     },
     {
-      "epoch": 0.6197577655104227,
+      "epoch": 0.6403518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 6.672093105310648e-06,
-      "loss": 0.8393,
+      "learning_rate": 6.051020387140458e-06,
+      "loss": 0.953,
       "step": 22566
     },
     {
-      "epoch": 0.6197852297382659,
+      "epoch": 0.6403802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 6.67125430423762e-06,
-      "loss": 0.7497,
+      "learning_rate": 6.050176023834544e-06,
+      "loss": 0.7773,
       "step": 22567
     },
     {
-      "epoch": 0.6198126939661092,
+      "epoch": 0.6404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 6.670415529502921e-06,
-      "loss": 0.8481,
+      "learning_rate": 6.049331693892966e-06,
+      "loss": 0.8942,
       "step": 22568
     },
     {
-      "epoch": 0.6198401581939523,
+      "epoch": 0.6404370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 6.6695767811131805e-06,
-      "loss": 0.8999,
+      "learning_rate": 6.048487397322864e-06,
+      "loss": 0.7854,
       "step": 22569
     },
     {
-      "epoch": 0.6198676224217956,
+      "epoch": 0.6404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 6.668738059075037e-06,
-      "loss": 0.7766,
+      "learning_rate": 6.047643134131367e-06,
+      "loss": 0.8733,
       "step": 22570
     },
     {
-      "epoch": 0.6198950866496389,
+      "epoch": 0.6404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 6.667899363395133e-06,
-      "loss": 0.8507,
+      "learning_rate": 6.046798904325603e-06,
+      "loss": 0.8473,
       "step": 22571
     },
     {
-      "epoch": 0.619922550877482,
+      "epoch": 0.640522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 6.667060694080096e-06,
-      "loss": 0.8702,
+      "learning_rate": 6.04595470791271e-06,
+      "loss": 0.8526,
       "step": 22572
     },
     {
-      "epoch": 0.6199500151053253,
+      "epoch": 0.6405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 6.666222051136566e-06,
-      "loss": 0.9196,
+      "learning_rate": 6.045110544899815e-06,
+      "loss": 0.8428,
       "step": 22573
     },
     {
-      "epoch": 0.6199774793331686,
+      "epoch": 0.6405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 6.6653834345711835e-06,
-      "loss": 0.8744,
+      "learning_rate": 6.044266415294046e-06,
+      "loss": 0.8827,
       "step": 22574
     },
     {
-      "epoch": 0.6200049435610118,
+      "epoch": 0.6406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 6.664544844390576e-06,
-      "loss": 0.6687,
+      "learning_rate": 6.043422319102539e-06,
+      "loss": 0.9241,
       "step": 22575
     },
     {
-      "epoch": 0.620032407788855,
+      "epoch": 0.6406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 6.66370628060138e-06,
-      "loss": 0.9589,
+      "learning_rate": 6.042578256332417e-06,
+      "loss": 0.8161,
       "step": 22576
     },
     {
-      "epoch": 0.6200598720166982,
+      "epoch": 0.6406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 6.6628677432102325e-06,
-      "loss": 0.8591,
+      "learning_rate": 6.041734226990819e-06,
+      "loss": 0.8232,
       "step": 22577
     },
     {
-      "epoch": 0.6200873362445415,
+      "epoch": 0.6406923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 6.66202923222377e-06,
-      "loss": 0.82,
+      "learning_rate": 6.040890231084869e-06,
+      "loss": 0.8253,
       "step": 22578
     },
     {
-      "epoch": 0.6201148004723848,
+      "epoch": 0.6407207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 6.6611907476486225e-06,
-      "loss": 0.7693,
+      "learning_rate": 6.040046268621696e-06,
+      "loss": 0.8205,
       "step": 22579
     },
     {
-      "epoch": 0.6201422647002279,
+      "epoch": 0.6407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 6.660352289491432e-06,
-      "loss": 0.8578,
+      "learning_rate": 6.039202339608432e-06,
+      "loss": 0.8576,
       "step": 22580
     },
     {
-      "epoch": 0.6201697289280712,
+      "epoch": 0.64077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.659513857758824e-06,
-      "loss": 0.9004,
+      "learning_rate": 6.038358444052204e-06,
+      "loss": 0.791,
       "step": 22581
     },
     {
-      "epoch": 0.6201971931559144,
+      "epoch": 0.640805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 6.658675452457435e-06,
-      "loss": 0.8085,
+      "learning_rate": 6.037514581960139e-06,
+      "loss": 0.7798,
       "step": 22582
     },
     {
-      "epoch": 0.6202246573837576,
+      "epoch": 0.6408342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 6.657837073593903e-06,
-      "loss": 0.8968,
+      "learning_rate": 6.036670753339367e-06,
+      "loss": 0.8577,
       "step": 22583
     },
     {
-      "epoch": 0.6202521216116009,
+      "epoch": 0.6408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 6.656998721174856e-06,
-      "loss": 0.7726,
+      "learning_rate": 6.035826958197016e-06,
+      "loss": 0.8774,
       "step": 22584
     },
     {
-      "epoch": 0.6202795858394441,
+      "epoch": 0.6408910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 6.656160395206932e-06,
-      "loss": 0.9247,
+      "learning_rate": 6.034983196540212e-06,
+      "loss": 0.8335,
       "step": 22585
     },
     {
-      "epoch": 0.6203070500672874,
+      "epoch": 0.6409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 6.655322095696765e-06,
-      "loss": 0.7818,
+      "learning_rate": 6.0341394683760835e-06,
+      "loss": 0.8559,
       "step": 22586
     },
     {
-      "epoch": 0.6203345142951306,
+      "epoch": 0.6409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 6.654483822650983e-06,
-      "loss": 0.8597,
+      "learning_rate": 6.0332957737117585e-06,
+      "loss": 0.8294,
       "step": 22587
     },
     {
-      "epoch": 0.6203619785229738,
+      "epoch": 0.6409761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 6.65364557607622e-06,
-      "loss": 0.8475,
+      "learning_rate": 6.032452112554359e-06,
+      "loss": 0.8967,
       "step": 22588
     },
     {
-      "epoch": 0.6203894427508171,
+      "epoch": 0.6410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 6.652807355979107e-06,
-      "loss": 0.8264,
+      "learning_rate": 6.031608484911018e-06,
+      "loss": 0.8971,
       "step": 22589
     },
     {
-      "epoch": 0.6204169069786603,
+      "epoch": 0.6410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 6.651969162366283e-06,
-      "loss": 0.9234,
+      "learning_rate": 6.030764890788857e-06,
+      "loss": 0.7183,
       "step": 22590
     },
     {
-      "epoch": 0.6204443712065035,
+      "epoch": 0.6410612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 6.651130995244373e-06,
-      "loss": 0.8391,
+      "learning_rate": 6.029921330195001e-06,
+      "loss": 0.8318,
       "step": 22591
     },
     {
-      "epoch": 0.6204718354343468,
+      "epoch": 0.6410896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 6.6502928546200165e-06,
-      "loss": 0.8474,
+      "learning_rate": 6.02907780313658e-06,
+      "loss": 0.7826,
       "step": 22592
     },
     {
-      "epoch": 0.62049929966219,
+      "epoch": 0.6411180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 6.649454740499837e-06,
-      "loss": 0.8561,
+      "learning_rate": 6.028234309620718e-06,
+      "loss": 0.8448,
       "step": 22593
     },
     {
-      "epoch": 0.6205267638900333,
+      "epoch": 0.6411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 6.648616652890467e-06,
-      "loss": 0.9344,
+      "learning_rate": 6.02739084965454e-06,
+      "loss": 0.9076,
       "step": 22594
     },
     {
-      "epoch": 0.6205542281178764,
+      "epoch": 0.6411748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 6.647778591798544e-06,
-      "loss": 0.7869,
+      "learning_rate": 6.026547423245171e-06,
+      "loss": 0.724,
       "step": 22595
     },
     {
-      "epoch": 0.6205816923457197,
+      "epoch": 0.6412031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 6.646940557230693e-06,
-      "loss": 0.8869,
+      "learning_rate": 6.02570403039973e-06,
+      "loss": 0.9382,
       "step": 22596
     },
     {
-      "epoch": 0.620609156573563,
+      "epoch": 0.6412315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 6.6461025491935475e-06,
-      "loss": 0.8468,
+      "learning_rate": 6.02486067112535e-06,
+      "loss": 0.8791,
       "step": 22597
     },
     {
-      "epoch": 0.6206366208014061,
+      "epoch": 0.6412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 6.645264567693741e-06,
-      "loss": 0.9524,
+      "learning_rate": 6.024017345429149e-06,
+      "loss": 0.8645,
       "step": 22598
     },
     {
-      "epoch": 0.6206640850292494,
+      "epoch": 0.6412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 6.6444266127379e-06,
-      "loss": 0.856,
+      "learning_rate": 6.023174053318252e-06,
+      "loss": 0.869,
       "step": 22599
     },
     {
-      "epoch": 0.6206915492570927,
+      "epoch": 0.641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.643588684332652e-06,
-      "loss": 0.8453,
+      "learning_rate": 6.022330794799784e-06,
+      "loss": 0.804,
       "step": 22600
     },
     {
-      "epoch": 0.6207190134849359,
+      "epoch": 0.6413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 6.642750782484631e-06,
-      "loss": 0.8688,
+      "learning_rate": 6.021487569880866e-06,
+      "loss": 0.8863,
       "step": 22601
     },
     {
-      "epoch": 0.6207464777127791,
+      "epoch": 0.6413734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 6.641912907200469e-06,
-      "loss": 0.7558,
+      "learning_rate": 6.020644378568621e-06,
+      "loss": 0.8538,
       "step": 22602
     },
     {
-      "epoch": 0.6207739419406223,
+      "epoch": 0.6414018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 6.64107505848679e-06,
-      "loss": 0.8694,
+      "learning_rate": 6.0198012208701715e-06,
+      "loss": 0.8179,
       "step": 22603
     },
     {
-      "epoch": 0.6208014061684656,
+      "epoch": 0.6414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 6.640237236350231e-06,
-      "loss": 0.9146,
+      "learning_rate": 6.018958096792642e-06,
+      "loss": 0.8455,
       "step": 22604
     },
     {
-      "epoch": 0.6208288703963089,
+      "epoch": 0.6414585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 6.639399440797413e-06,
-      "loss": 0.8954,
+      "learning_rate": 6.018115006343148e-06,
+      "loss": 0.9094,
       "step": 22605
     },
     {
-      "epoch": 0.620856334624152,
+      "epoch": 0.6414869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 6.638561671834967e-06,
-      "loss": 0.8764,
+      "learning_rate": 6.017271949528816e-06,
+      "loss": 0.9425,
       "step": 22606
     },
     {
-      "epoch": 0.6208837988519953,
+      "epoch": 0.6415153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 6.637723929469526e-06,
-      "loss": 0.9214,
+      "learning_rate": 6.01642892635677e-06,
+      "loss": 0.9013,
       "step": 22607
     },
     {
-      "epoch": 0.6209112630798385,
+      "epoch": 0.6415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 6.636886213707713e-06,
-      "loss": 0.8973,
+      "learning_rate": 6.015585936834127e-06,
+      "loss": 0.8851,
       "step": 22608
     },
     {
-      "epoch": 0.6209387273076817,
+      "epoch": 0.6415720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 6.636048524556159e-06,
-      "loss": 0.9268,
+      "learning_rate": 6.014742980968008e-06,
+      "loss": 0.8485,
       "step": 22609
     },
     {
-      "epoch": 0.620966191535525,
+      "epoch": 0.6416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 6.635210862021497e-06,
-      "loss": 0.8467,
+      "learning_rate": 6.013900058765536e-06,
+      "loss": 0.8228,
       "step": 22610
     },
     {
-      "epoch": 0.6209936557633682,
+      "epoch": 0.6416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 6.634373226110347e-06,
-      "loss": 0.8406,
+      "learning_rate": 6.013057170233829e-06,
+      "loss": 0.743,
       "step": 22611
     },
     {
-      "epoch": 0.6210211199912115,
+      "epoch": 0.6416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 6.633535616829338e-06,
-      "loss": 0.807,
+      "learning_rate": 6.012214315380005e-06,
+      "loss": 0.8268,
       "step": 22612
     },
     {
-      "epoch": 0.6210485842190547,
+      "epoch": 0.6416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 6.6326980341851e-06,
-      "loss": 0.7169,
+      "learning_rate": 6.011371494211189e-06,
+      "loss": 0.8695,
       "step": 22613
     },
     {
-      "epoch": 0.6210760484468979,
+      "epoch": 0.6417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 6.631860478184261e-06,
-      "loss": 0.815,
+      "learning_rate": 6.010528706734495e-06,
+      "loss": 0.7918,
       "step": 22614
     },
     {
-      "epoch": 0.6211035126747412,
+      "epoch": 0.6417423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 6.631022948833449e-06,
-      "loss": 0.8322,
+      "learning_rate": 6.009685952957045e-06,
+      "loss": 0.8756,
       "step": 22615
     },
     {
-      "epoch": 0.6211309769025843,
+      "epoch": 0.6417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 6.630185446139283e-06,
-      "loss": 0.9119,
+      "learning_rate": 6.0088432328859584e-06,
+      "loss": 0.7568,
       "step": 22616
     },
     {
-      "epoch": 0.6211584411304276,
+      "epoch": 0.6417990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 6.629347970108398e-06,
-      "loss": 0.8455,
+      "learning_rate": 6.008000546528352e-06,
+      "loss": 0.802,
       "step": 22617
     },
     {
-      "epoch": 0.6211859053582709,
+      "epoch": 0.641827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 6.628510520747415e-06,
-      "loss": 0.8265,
+      "learning_rate": 6.007157893891344e-06,
+      "loss": 0.8275,
       "step": 22618
     },
     {
-      "epoch": 0.6212133695861141,
+      "epoch": 0.6418558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 6.627673098062966e-06,
-      "loss": 0.8012,
+      "learning_rate": 6.0063152749820506e-06,
+      "loss": 0.8162,
       "step": 22619
     },
     {
-      "epoch": 0.6212408338139573,
+      "epoch": 0.6418842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 6.62683570206167e-06,
-      "loss": 0.8339,
+      "learning_rate": 6.005472689807596e-06,
+      "loss": 0.7769,
       "step": 22620
     },
     {
-      "epoch": 0.6212682980418005,
+      "epoch": 0.6419125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 6.625998332750162e-06,
-      "loss": 0.8607,
+      "learning_rate": 6.0046301383750915e-06,
+      "loss": 0.8896,
       "step": 22621
     },
     {
-      "epoch": 0.6212957622696438,
+      "epoch": 0.6419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 6.6251609901350576e-06,
-      "loss": 0.8553,
+      "learning_rate": 6.003787620691651e-06,
+      "loss": 0.9875,
       "step": 22622
     },
     {
-      "epoch": 0.6213232264974871,
+      "epoch": 0.6419693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 6.62432367422299e-06,
-      "loss": 0.7822,
+      "learning_rate": 6.002945136764399e-06,
+      "loss": 0.8343,
       "step": 22623
     },
     {
-      "epoch": 0.6213506907253302,
+      "epoch": 0.6419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 6.623486385020578e-06,
-      "loss": 0.7745,
+      "learning_rate": 6.002102686600451e-06,
+      "loss": 0.7955,
       "step": 22624
     },
     {
-      "epoch": 0.6213781549531735,
+      "epoch": 0.6420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 6.62264912253445e-06,
-      "loss": 0.7751,
+      "learning_rate": 6.001260270206921e-06,
+      "loss": 0.8394,
       "step": 22625
     },
     {
-      "epoch": 0.6214056191810168,
+      "epoch": 0.6420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 6.621811886771232e-06,
-      "loss": 0.9855,
+      "learning_rate": 6.000417887590924e-06,
+      "loss": 0.9341,
       "step": 22626
     },
     {
-      "epoch": 0.6214330834088599,
+      "epoch": 0.6420828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 6.620974677737548e-06,
-      "loss": 0.9029,
+      "learning_rate": 5.999575538759579e-06,
+      "loss": 0.9082,
       "step": 22627
     },
     {
-      "epoch": 0.6214605476367032,
+      "epoch": 0.64211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.62013749544002e-06,
-      "loss": 0.8303,
+      "learning_rate": 5.998733223719998e-06,
+      "loss": 0.8025,
       "step": 22628
     },
     {
-      "epoch": 0.6214880118645464,
+      "epoch": 0.6421396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 6.6193003398852705e-06,
-      "loss": 0.875,
+      "learning_rate": 5.997890942479296e-06,
+      "loss": 0.8094,
       "step": 22629
     },
     {
-      "epoch": 0.6215154760923897,
+      "epoch": 0.6421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 6.6184632110799285e-06,
-      "loss": 0.8753,
+      "learning_rate": 5.997048695044592e-06,
+      "loss": 0.8004,
       "step": 22630
     },
     {
-      "epoch": 0.6215429403202329,
+      "epoch": 0.6421963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 6.617626109030617e-06,
-      "loss": 0.8596,
+      "learning_rate": 5.996206481422997e-06,
+      "loss": 0.8403,
       "step": 22631
     },
     {
-      "epoch": 0.6215704045480761,
+      "epoch": 0.6422247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 6.616789033743954e-06,
-      "loss": 0.8574,
+      "learning_rate": 5.9953643016216225e-06,
+      "loss": 0.8861,
       "step": 22632
     },
     {
-      "epoch": 0.6215978687759194,
+      "epoch": 0.6422531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 6.615951985226573e-06,
-      "loss": 0.8716,
+      "learning_rate": 5.99452215564759e-06,
+      "loss": 0.86,
       "step": 22633
     },
     {
-      "epoch": 0.6216253330037625,
+      "epoch": 0.6422814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 6.615114963485085e-06,
-      "loss": 0.8799,
+      "learning_rate": 5.993680043508008e-06,
+      "loss": 0.7783,
       "step": 22634
     },
     {
-      "epoch": 0.6216527972316058,
+      "epoch": 0.6423098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 6.614277968526122e-06,
-      "loss": 0.8747,
+      "learning_rate": 5.992837965209988e-06,
+      "loss": 0.9112,
       "step": 22635
     },
     {
-      "epoch": 0.6216802614594491,
+      "epoch": 0.6423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 6.613441000356301e-06,
-      "loss": 0.9268,
+      "learning_rate": 5.991995920760648e-06,
+      "loss": 0.937,
       "step": 22636
     },
     {
-      "epoch": 0.6217077256872923,
+      "epoch": 0.642366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 6.612604058982246e-06,
-      "loss": 0.8248,
+      "learning_rate": 5.991153910167096e-06,
+      "loss": 0.8208,
       "step": 22637
     },
     {
-      "epoch": 0.6217351899151355,
+      "epoch": 0.6423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 6.6117671444105825e-06,
-      "loss": 0.8756,
+      "learning_rate": 5.99031193343645e-06,
+      "loss": 0.9305,
       "step": 22638
     },
     {
-      "epoch": 0.6217626541429788,
+      "epoch": 0.6424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 6.610930256647931e-06,
-      "loss": 0.7395,
+      "learning_rate": 5.9894699905758195e-06,
+      "loss": 0.8515,
       "step": 22639
     },
     {
-      "epoch": 0.621790118370822,
+      "epoch": 0.6424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 6.610093395700909e-06,
-      "loss": 0.8806,
+      "learning_rate": 5.9886280815923135e-06,
+      "loss": 0.9775,
       "step": 22640
     },
     {
-      "epoch": 0.6218175825986653,
+      "epoch": 0.6424801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 6.609256561576141e-06,
-      "loss": 0.9273,
+      "learning_rate": 5.987786206493049e-06,
+      "loss": 0.8965,
       "step": 22641
     },
     {
-      "epoch": 0.6218450468265084,
+      "epoch": 0.6425085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 6.608419754280249e-06,
-      "loss": 0.8745,
+      "learning_rate": 5.986944365285136e-06,
+      "loss": 0.8162,
       "step": 22642
     },
     {
-      "epoch": 0.6218725110543517,
+      "epoch": 0.6425368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 6.607582973819855e-06,
-      "loss": 0.758,
+      "learning_rate": 5.986102557975681e-06,
+      "loss": 0.9292,
       "step": 22643
     },
     {
-      "epoch": 0.621899975282195,
+      "epoch": 0.6425652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 6.606746220201576e-06,
-      "loss": 0.7937,
+      "learning_rate": 5.9852607845718e-06,
+      "loss": 0.7806,
       "step": 22644
     },
     {
-      "epoch": 0.6219274395100381,
+      "epoch": 0.6425936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 6.60590949343204e-06,
-      "loss": 0.824,
+      "learning_rate": 5.984419045080603e-06,
+      "loss": 0.8303,
       "step": 22645
     },
     {
-      "epoch": 0.6219549037378814,
+      "epoch": 0.642622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 6.605072793517859e-06,
-      "loss": 0.8385,
+      "learning_rate": 5.983577339509195e-06,
+      "loss": 0.887,
       "step": 22646
     },
     {
-      "epoch": 0.6219823679657246,
+      "epoch": 0.642650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.604236120465659e-06,
-      "loss": 0.9085,
+      "learning_rate": 5.982735667864694e-06,
+      "loss": 0.8851,
       "step": 22647
     },
     {
-      "epoch": 0.6220098321935679,
+      "epoch": 0.6426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 6.603399474282055e-06,
-      "loss": 0.9376,
+      "learning_rate": 5.981894030154203e-06,
+      "loss": 0.8407,
       "step": 22648
     },
     {
-      "epoch": 0.6220372964214111,
+      "epoch": 0.6427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 6.6025628549736706e-06,
-      "loss": 0.8575,
+      "learning_rate": 5.9810524263848335e-06,
+      "loss": 0.7805,
       "step": 22649
     },
     {
-      "epoch": 0.6220647606492543,
+      "epoch": 0.6427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 6.601726262547128e-06,
-      "loss": 0.9308,
+      "learning_rate": 5.9802108565636965e-06,
+      "loss": 0.8508,
       "step": 22650
     },
     {
-      "epoch": 0.6220922248770976,
+      "epoch": 0.6427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 6.6008896970090454e-06,
-      "loss": 0.8938,
+      "learning_rate": 5.979369320697899e-06,
+      "loss": 0.9327,
       "step": 22651
     },
     {
-      "epoch": 0.6221196891049409,
+      "epoch": 0.6427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 6.600053158366035e-06,
-      "loss": 0.873,
+      "learning_rate": 5.978527818794545e-06,
+      "loss": 0.8713,
       "step": 22652
     },
     {
-      "epoch": 0.622147153332784,
+      "epoch": 0.6428206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 6.599216646624723e-06,
-      "loss": 0.9091,
+      "learning_rate": 5.97768635086075e-06,
+      "loss": 0.832,
       "step": 22653
     },
     {
-      "epoch": 0.6221746175606273,
+      "epoch": 0.6428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 6.5983801617917265e-06,
-      "loss": 0.8458,
+      "learning_rate": 5.97684491690362e-06,
+      "loss": 0.7953,
       "step": 22654
     },
     {
-      "epoch": 0.6222020817884705,
+      "epoch": 0.6428774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 6.597543703873663e-06,
-      "loss": 0.8111,
+      "learning_rate": 5.976003516930263e-06,
+      "loss": 0.9447,
       "step": 22655
     },
     {
-      "epoch": 0.6222295460163138,
+      "epoch": 0.6429057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 6.596707272877152e-06,
-      "loss": 0.8821,
+      "learning_rate": 5.975162150947783e-06,
+      "loss": 0.7877,
       "step": 22656
     },
     {
-      "epoch": 0.622257010244157,
+      "epoch": 0.6429341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 6.5958708688088136e-06,
-      "loss": 0.7853,
+      "learning_rate": 5.97432081896329e-06,
+      "loss": 0.8447,
       "step": 22657
     },
     {
-      "epoch": 0.6222844744720002,
+      "epoch": 0.6429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 6.595034491675261e-06,
-      "loss": 0.8317,
+      "learning_rate": 5.973479520983892e-06,
+      "loss": 0.9699,
       "step": 22658
     },
     {
-      "epoch": 0.6223119386998435,
+      "epoch": 0.6429909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 6.594198141483116e-06,
-      "loss": 0.816,
+      "learning_rate": 5.972638257016693e-06,
+      "loss": 0.8817,
       "step": 22659
     },
     {
-      "epoch": 0.6223394029276866,
+      "epoch": 0.6430192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 6.5933618182389925e-06,
-      "loss": 0.851,
+      "learning_rate": 5.971797027068797e-06,
+      "loss": 0.8403,
       "step": 22660
     },
     {
-      "epoch": 0.6223668671555299,
+      "epoch": 0.6430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 6.592525521949508e-06,
-      "loss": 0.9529,
+      "learning_rate": 5.970955831147316e-06,
+      "loss": 0.8503,
       "step": 22661
     },
     {
-      "epoch": 0.6223943313833732,
+      "epoch": 0.6430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 6.591689252621287e-06,
-      "loss": 0.8768,
+      "learning_rate": 5.97011466925935e-06,
+      "loss": 0.8134,
       "step": 22662
     },
     {
-      "epoch": 0.6224217956112164,
+      "epoch": 0.6431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 6.5908530102609415e-06,
-      "loss": 0.8299,
+      "learning_rate": 5.969273541412006e-06,
+      "loss": 0.8615,
       "step": 22663
     },
     {
-      "epoch": 0.6224492598390596,
+      "epoch": 0.6431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 6.590016794875083e-06,
-      "loss": 0.8029,
+      "learning_rate": 5.968432447612391e-06,
+      "loss": 0.8867,
       "step": 22664
     },
     {
-      "epoch": 0.6224767240669029,
+      "epoch": 0.643161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 6.589180606470333e-06,
-      "loss": 0.8394,
+      "learning_rate": 5.967591387867609e-06,
+      "loss": 0.7966,
       "step": 22665
     },
     {
-      "epoch": 0.6225041882947461,
+      "epoch": 0.6431895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 6.58834444505331e-06,
-      "loss": 1.0068,
+      "learning_rate": 5.966750362184761e-06,
+      "loss": 0.869,
       "step": 22666
     },
     {
-      "epoch": 0.6225316525225894,
+      "epoch": 0.6432179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 6.5875083106306235e-06,
-      "loss": 0.8113,
+      "learning_rate": 5.965909370570957e-06,
+      "loss": 0.8941,
       "step": 22667
     },
     {
-      "epoch": 0.6225591167504325,
+      "epoch": 0.6432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 6.5866722032088945e-06,
-      "loss": 0.855,
+      "learning_rate": 5.965068413033292e-06,
+      "loss": 0.8924,
       "step": 22668
     },
     {
-      "epoch": 0.6225865809782758,
+      "epoch": 0.6432746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 6.58583612279474e-06,
-      "loss": 0.8228,
+      "learning_rate": 5.964227489578882e-06,
+      "loss": 0.8267,
       "step": 22669
     },
     {
-      "epoch": 0.6226140452061191,
+      "epoch": 0.6433030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 6.585000069394769e-06,
-      "loss": 0.9767,
+      "learning_rate": 5.96338660021482e-06,
+      "loss": 0.7508,
       "step": 22670
     },
     {
-      "epoch": 0.6226415094339622,
+      "epoch": 0.6433314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 6.584164043015604e-06,
-      "loss": 0.8741,
+      "learning_rate": 5.962545744948216e-06,
+      "loss": 0.8955,
       "step": 22671
     },
     {
-      "epoch": 0.6226689736618055,
+      "epoch": 0.6433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 6.583328043663852e-06,
-      "loss": 0.766,
+      "learning_rate": 5.961704923786169e-06,
+      "loss": 0.9201,
       "step": 22672
     },
     {
-      "epoch": 0.6226964378896487,
+      "epoch": 0.6433881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 6.582492071346134e-06,
-      "loss": 0.8228,
+      "learning_rate": 5.96086413673578e-06,
+      "loss": 0.8706,
       "step": 22673
     },
     {
-      "epoch": 0.622723902117492,
+      "epoch": 0.6434165720771851,
       "grad_norm": 0.0,
-      "learning_rate": 6.581656126069065e-06,
-      "loss": 0.8427,
+      "learning_rate": 5.9600233838041565e-06,
+      "loss": 0.8465,
       "step": 22674
     },
     {
-      "epoch": 0.6227513663453352,
+      "epoch": 0.6434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 6.580820207839255e-06,
-      "loss": 0.7146,
+      "learning_rate": 5.9591826649983955e-06,
+      "loss": 0.8712,
       "step": 22675
     },
     {
-      "epoch": 0.6227788305731784,
+      "epoch": 0.6434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 6.579984316663319e-06,
-      "loss": 0.8749,
+      "learning_rate": 5.958341980325598e-06,
+      "loss": 0.7838,
       "step": 22676
     },
     {
-      "epoch": 0.6228062948010217,
+      "epoch": 0.6435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 6.579148452547871e-06,
-      "loss": 0.7245,
+      "learning_rate": 5.95750132979287e-06,
+      "loss": 0.9062,
       "step": 22677
     },
     {
-      "epoch": 0.622833759028865,
+      "epoch": 0.6435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 6.5783126154995266e-06,
-      "loss": 0.84,
+      "learning_rate": 5.95666071340731e-06,
+      "loss": 0.8878,
       "step": 22678
     },
     {
-      "epoch": 0.6228612232567081,
+      "epoch": 0.6435584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 6.5774768055248984e-06,
-      "loss": 0.8556,
+      "learning_rate": 5.955820131176018e-06,
+      "loss": 0.8298,
       "step": 22679
     },
     {
-      "epoch": 0.6228886874845514,
+      "epoch": 0.6435868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 6.576641022630603e-06,
-      "loss": 0.8736,
+      "learning_rate": 5.954979583106095e-06,
+      "loss": 0.9327,
       "step": 22680
     },
     {
-      "epoch": 0.6229161517123946,
+      "epoch": 0.6436152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 6.575805266823245e-06,
-      "loss": 0.858,
+      "learning_rate": 5.954139069204643e-06,
+      "loss": 0.9721,
       "step": 22681
     },
     {
-      "epoch": 0.6229436159402378,
+      "epoch": 0.6436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 6.5749695381094426e-06,
-      "loss": 0.8032,
+      "learning_rate": 5.9532985894787575e-06,
+      "loss": 0.9151,
       "step": 22682
     },
     {
-      "epoch": 0.6229710801680811,
+      "epoch": 0.643671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 6.57413383649581e-06,
-      "loss": 0.7777,
+      "learning_rate": 5.95245814393554e-06,
+      "loss": 0.9443,
       "step": 22683
     },
     {
-      "epoch": 0.6229985443959243,
+      "epoch": 0.643700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 6.573298161988955e-06,
-      "loss": 0.9758,
+      "learning_rate": 5.9516177325820915e-06,
+      "loss": 0.8425,
       "step": 22684
     },
     {
-      "epoch": 0.6230260086237676,
+      "epoch": 0.6437287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 6.5724625145954925e-06,
-      "loss": 0.8316,
+      "learning_rate": 5.950777355425511e-06,
+      "loss": 0.8531,
       "step": 22685
     },
     {
-      "epoch": 0.6230534728516107,
+      "epoch": 0.6437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 6.571626894322039e-06,
-      "loss": 0.9407,
+      "learning_rate": 5.949937012472897e-06,
+      "loss": 0.8563,
       "step": 22686
     },
     {
-      "epoch": 0.623080937079454,
+      "epoch": 0.6437854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 6.570791301175199e-06,
-      "loss": 0.9044,
+      "learning_rate": 5.949096703731345e-06,
+      "loss": 0.8883,
       "step": 22687
     },
     {
-      "epoch": 0.6231084013072973,
+      "epoch": 0.6438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 6.569955735161584e-06,
-      "loss": 0.806,
+      "learning_rate": 5.948256429207957e-06,
+      "loss": 0.7917,
       "step": 22688
     },
     {
-      "epoch": 0.6231358655351404,
+      "epoch": 0.6438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 6.5691201962878094e-06,
-      "loss": 0.8801,
+      "learning_rate": 5.947416188909829e-06,
+      "loss": 0.7715,
       "step": 22689
     },
     {
-      "epoch": 0.6231633297629837,
+      "epoch": 0.6438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 6.568284684560485e-06,
-      "loss": 0.8188,
+      "learning_rate": 5.946575982844058e-06,
+      "loss": 0.7678,
       "step": 22690
     },
     {
-      "epoch": 0.623190793990827,
+      "epoch": 0.6438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 6.567449199986221e-06,
-      "loss": 0.7806,
+      "learning_rate": 5.945735811017742e-06,
+      "loss": 0.8916,
       "step": 22691
     },
     {
-      "epoch": 0.6232182582186702,
+      "epoch": 0.6439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 6.566613742571633e-06,
-      "loss": 0.9242,
+      "learning_rate": 5.944895673437978e-06,
+      "loss": 0.8324,
       "step": 22692
     },
     {
-      "epoch": 0.6232457224465134,
+      "epoch": 0.6439557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 6.565778312323321e-06,
-      "loss": 0.9269,
+      "learning_rate": 5.944055570111862e-06,
+      "loss": 0.7987,
       "step": 22693
     },
     {
-      "epoch": 0.6232731866743566,
+      "epoch": 0.6439841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 6.564942909247902e-06,
-      "loss": 0.8592,
+      "learning_rate": 5.943215501046492e-06,
+      "loss": 0.8732,
       "step": 22694
     },
     {
-      "epoch": 0.6233006509021999,
+      "epoch": 0.6440124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 6.564107533351988e-06,
-      "loss": 0.8745,
+      "learning_rate": 5.942375466248964e-06,
+      "loss": 0.8884,
       "step": 22695
     },
     {
-      "epoch": 0.6233281151300432,
+      "epoch": 0.6440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 6.563272184642184e-06,
-      "loss": 0.9215,
+      "learning_rate": 5.941535465726369e-06,
+      "loss": 0.8082,
       "step": 22696
     },
     {
-      "epoch": 0.6233555793578863,
+      "epoch": 0.6440692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 6.562436863125101e-06,
-      "loss": 0.8804,
+      "learning_rate": 5.940695499485809e-06,
+      "loss": 0.8354,
       "step": 22697
     },
     {
-      "epoch": 0.6233830435857296,
+      "epoch": 0.6440976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 6.561601568807354e-06,
-      "loss": 0.9157,
+      "learning_rate": 5.9398555675343736e-06,
+      "loss": 0.9352,
       "step": 22698
     },
     {
-      "epoch": 0.6234105078135728,
+      "epoch": 0.6441259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 6.560766301695545e-06,
-      "loss": 0.9076,
+      "learning_rate": 5.9390156698791645e-06,
+      "loss": 0.799,
       "step": 22699
     },
     {
-      "epoch": 0.623437972041416,
+      "epoch": 0.6441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 6.559931061796283e-06,
-      "loss": 0.912,
+      "learning_rate": 5.938175806527271e-06,
+      "loss": 0.9588,
       "step": 22700
     },
     {
-      "epoch": 0.6234654362692593,
+      "epoch": 0.6441827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 6.559095849116179e-06,
-      "loss": 0.7878,
+      "learning_rate": 5.937335977485789e-06,
+      "loss": 0.7784,
       "step": 22701
     },
     {
-      "epoch": 0.6234929004971025,
+      "epoch": 0.644211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.558260663661844e-06,
-      "loss": 0.9101,
+      "learning_rate": 5.9364961827618136e-06,
+      "loss": 0.8641,
       "step": 22702
     },
     {
-      "epoch": 0.6235203647249458,
+      "epoch": 0.6442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 6.557425505439881e-06,
-      "loss": 0.8141,
+      "learning_rate": 5.935656422362438e-06,
+      "loss": 0.8498,
       "step": 22703
     },
     {
-      "epoch": 0.623547828952789,
+      "epoch": 0.6442678774120317,
       "grad_norm": 0.0,
-      "learning_rate": 6.556590374456906e-06,
-      "loss": 0.9152,
+      "learning_rate": 5.934816696294753e-06,
+      "loss": 0.9305,
       "step": 22704
     },
     {
-      "epoch": 0.6235752931806322,
+      "epoch": 0.6442962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 6.555755270719517e-06,
-      "loss": 0.8411,
+      "learning_rate": 5.933977004565856e-06,
+      "loss": 0.9663,
       "step": 22705
     },
     {
-      "epoch": 0.6236027574084755,
+      "epoch": 0.6443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 6.5549201942343265e-06,
-      "loss": 0.911,
+      "learning_rate": 5.933137347182838e-06,
+      "loss": 0.8686,
       "step": 22706
     },
     {
-      "epoch": 0.6236302216363186,
+      "epoch": 0.6443530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 6.554085145007945e-06,
-      "loss": 0.8025,
+      "learning_rate": 5.932297724152791e-06,
+      "loss": 0.8478,
       "step": 22707
     },
     {
-      "epoch": 0.6236576858641619,
+      "epoch": 0.6443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 6.553250123046972e-06,
-      "loss": 0.9628,
+      "learning_rate": 5.931458135482808e-06,
+      "loss": 1.0003,
       "step": 22708
     },
     {
-      "epoch": 0.6236851500920052,
+      "epoch": 0.6444097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 6.552415128358022e-06,
-      "loss": 0.871,
+      "learning_rate": 5.9306185811799835e-06,
+      "loss": 0.8179,
       "step": 22709
     },
     {
-      "epoch": 0.6237126143198484,
+      "epoch": 0.6444381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 6.551580160947701e-06,
-      "loss": 0.8654,
+      "learning_rate": 5.929779061251402e-06,
+      "loss": 0.8833,
       "step": 22710
     },
     {
-      "epoch": 0.6237400785476916,
+      "epoch": 0.6444665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 6.550745220822614e-06,
-      "loss": 0.8732,
+      "learning_rate": 5.928939575704163e-06,
+      "loss": 0.99,
       "step": 22711
     },
     {
-      "epoch": 0.6237675427755348,
+      "epoch": 0.6444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 6.549910307989362e-06,
-      "loss": 0.8647,
+      "learning_rate": 5.9281001245453555e-06,
+      "loss": 0.7844,
       "step": 22712
     },
     {
-      "epoch": 0.6237950070033781,
+      "epoch": 0.6445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 6.549075422454558e-06,
-      "loss": 0.8257,
+      "learning_rate": 5.927260707782065e-06,
+      "loss": 0.8734,
       "step": 22713
     },
     {
-      "epoch": 0.6238224712312214,
+      "epoch": 0.6445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 6.548240564224808e-06,
-      "loss": 0.802,
+      "learning_rate": 5.926421325421385e-06,
+      "loss": 0.7433,
       "step": 22714
     },
     {
-      "epoch": 0.6238499354590645,
+      "epoch": 0.6445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 6.547405733306713e-06,
-      "loss": 0.929,
+      "learning_rate": 5.925581977470412e-06,
+      "loss": 0.9075,
       "step": 22715
     },
     {
-      "epoch": 0.6238773996869078,
+      "epoch": 0.6446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 6.5465709297068854e-06,
-      "loss": 0.8528,
+      "learning_rate": 5.924742663936232e-06,
+      "loss": 0.8409,
       "step": 22716
     },
     {
-      "epoch": 0.6239048639147511,
+      "epoch": 0.6446367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 6.545736153431923e-06,
-      "loss": 0.8036,
+      "learning_rate": 5.9239033848259305e-06,
+      "loss": 0.8963,
       "step": 22717
     },
     {
-      "epoch": 0.6239323281425943,
+      "epoch": 0.6446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 6.544901404488433e-06,
-      "loss": 0.8849,
+      "learning_rate": 5.923064140146603e-06,
+      "loss": 0.8644,
       "step": 22718
     },
     {
-      "epoch": 0.6239597923704375,
+      "epoch": 0.6446935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 6.544066682883025e-06,
-      "loss": 0.7838,
+      "learning_rate": 5.922224929905336e-06,
+      "loss": 0.9103,
       "step": 22719
     },
     {
-      "epoch": 0.6239872565982807,
+      "epoch": 0.64472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 6.543231988622296e-06,
-      "loss": 0.8717,
+      "learning_rate": 5.921385754109214e-06,
+      "loss": 0.7641,
       "step": 22720
     },
     {
-      "epoch": 0.624014720826124,
+      "epoch": 0.6447502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 6.542397321712856e-06,
-      "loss": 0.8446,
+      "learning_rate": 5.9205466127653345e-06,
+      "loss": 0.8038,
       "step": 22721
     },
     {
-      "epoch": 0.6240421850539672,
+      "epoch": 0.6447786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 6.5415626821613125e-06,
-      "loss": 0.7886,
+      "learning_rate": 5.91970750588078e-06,
+      "loss": 0.8529,
       "step": 22722
     },
     {
-      "epoch": 0.6240696492818104,
+      "epoch": 0.6448070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 6.540728069974261e-06,
-      "loss": 0.8713,
+      "learning_rate": 5.918868433462639e-06,
+      "loss": 0.872,
       "step": 22723
     },
     {
-      "epoch": 0.6240971135096537,
+      "epoch": 0.6448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 6.539893485158309e-06,
-      "loss": 0.8553,
+      "learning_rate": 5.918029395518001e-06,
+      "loss": 0.878,
       "step": 22724
     },
     {
-      "epoch": 0.6241245777374969,
+      "epoch": 0.6448637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 6.539058927720056e-06,
-      "loss": 0.8849,
+      "learning_rate": 5.917190392053953e-06,
+      "loss": 0.8034,
       "step": 22725
     },
     {
-      "epoch": 0.6241520419653401,
+      "epoch": 0.6448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 6.538224397666115e-06,
-      "loss": 0.9082,
+      "learning_rate": 5.91635142307758e-06,
+      "loss": 0.8135,
       "step": 22726
     },
     {
-      "epoch": 0.6241795061931834,
+      "epoch": 0.6449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 6.537389895003081e-06,
-      "loss": 0.9322,
+      "learning_rate": 5.915512488595968e-06,
+      "loss": 0.838,
       "step": 22727
     },
     {
-      "epoch": 0.6242069704210266,
+      "epoch": 0.6449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 6.536555419737564e-06,
-      "loss": 0.7848,
+      "learning_rate": 5.914673588616209e-06,
+      "loss": 0.8464,
       "step": 22728
     },
     {
-      "epoch": 0.6242344346488699,
+      "epoch": 0.6449772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 6.535720971876156e-06,
-      "loss": 0.8511,
+      "learning_rate": 5.9138347231453795e-06,
+      "loss": 0.8301,
       "step": 22729
     },
     {
-      "epoch": 0.6242618988767131,
+      "epoch": 0.645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 6.534886551425468e-06,
-      "loss": 0.7711,
+      "learning_rate": 5.912995892190578e-06,
+      "loss": 0.7507,
       "step": 22730
     },
     {
-      "epoch": 0.6242893631045563,
+      "epoch": 0.6450340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 6.534052158392101e-06,
-      "loss": 0.9094,
+      "learning_rate": 5.912157095758881e-06,
+      "loss": 0.7786,
       "step": 22731
     },
     {
-      "epoch": 0.6243168273323996,
+      "epoch": 0.6450624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 6.533217792782653e-06,
-      "loss": 0.7788,
+      "learning_rate": 5.9113183338573786e-06,
+      "loss": 0.777,
       "step": 22732
     },
     {
-      "epoch": 0.6243442915602427,
+      "epoch": 0.6450908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 6.5323834546037336e-06,
-      "loss": 0.7654,
+      "learning_rate": 5.910479606493156e-06,
+      "loss": 0.7812,
       "step": 22733
     },
     {
-      "epoch": 0.624371755788086,
+      "epoch": 0.6451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 6.531549143861935e-06,
-      "loss": 0.9486,
+      "learning_rate": 5.909640913673291e-06,
+      "loss": 0.8961,
       "step": 22734
     },
     {
-      "epoch": 0.6243992200159293,
+      "epoch": 0.6451475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 6.530714860563865e-06,
-      "loss": 0.8672,
+      "learning_rate": 5.908802255404877e-06,
+      "loss": 0.7137,
       "step": 22735
     },
     {
-      "epoch": 0.6244266842437725,
+      "epoch": 0.6451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 6.529880604716121e-06,
-      "loss": 0.9359,
+      "learning_rate": 5.907963631694994e-06,
+      "loss": 0.8022,
       "step": 22736
     },
     {
-      "epoch": 0.6244541484716157,
+      "epoch": 0.6452043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 6.529046376325307e-06,
-      "loss": 0.8159,
+      "learning_rate": 5.907125042550723e-06,
+      "loss": 0.8466,
       "step": 22737
     },
     {
-      "epoch": 0.624481612699459,
+      "epoch": 0.6452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 6.5282121753980236e-06,
-      "loss": 0.8024,
+      "learning_rate": 5.9062864879791535e-06,
+      "loss": 0.8698,
       "step": 22738
     },
     {
-      "epoch": 0.6245090769273022,
+      "epoch": 0.645261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 6.52737800194087e-06,
-      "loss": 0.8507,
+      "learning_rate": 5.905447967987368e-06,
+      "loss": 0.8527,
       "step": 22739
     },
     {
-      "epoch": 0.6245365411551455,
+      "epoch": 0.6452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 6.526543855960448e-06,
-      "loss": 0.8859,
+      "learning_rate": 5.904609482582443e-06,
+      "loss": 0.8413,
       "step": 22740
     },
     {
-      "epoch": 0.6245640053829886,
+      "epoch": 0.6453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 6.525709737463353e-06,
-      "loss": 0.8161,
+      "learning_rate": 5.903771031771468e-06,
+      "loss": 0.934,
       "step": 22741
     },
     {
-      "epoch": 0.6245914696108319,
+      "epoch": 0.6453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 6.52487564645619e-06,
-      "loss": 0.8715,
+      "learning_rate": 5.9029326155615245e-06,
+      "loss": 0.9124,
       "step": 22742
     },
     {
-      "epoch": 0.6246189338386752,
+      "epoch": 0.6453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 6.524041582945558e-06,
-      "loss": 0.8583,
+      "learning_rate": 5.9020942339596895e-06,
+      "loss": 0.8289,
       "step": 22743
     },
     {
-      "epoch": 0.6246463980665183,
+      "epoch": 0.6454029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 6.523207546938053e-06,
-      "loss": 0.9276,
+      "learning_rate": 5.901255886973051e-06,
+      "loss": 0.9,
       "step": 22744
     },
     {
-      "epoch": 0.6246738622943616,
+      "epoch": 0.6454313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 6.522373538440282e-06,
-      "loss": 0.8431,
+      "learning_rate": 5.900417574608687e-06,
+      "loss": 0.9222,
       "step": 22745
     },
     {
-      "epoch": 0.6247013265222048,
+      "epoch": 0.6454597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 6.521539557458834e-06,
-      "loss": 0.8146,
+      "learning_rate": 5.899579296873682e-06,
+      "loss": 0.9125,
       "step": 22746
     },
     {
-      "epoch": 0.6247287907500481,
+      "epoch": 0.6454880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 6.520705604000315e-06,
-      "loss": 0.7352,
+      "learning_rate": 5.898741053775115e-06,
+      "loss": 0.9691,
       "step": 22747
     },
     {
-      "epoch": 0.6247562549778913,
+      "epoch": 0.645516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 6.519871678071318e-06,
-      "loss": 0.8671,
+      "learning_rate": 5.897902845320065e-06,
+      "loss": 0.7818,
       "step": 22748
     },
     {
-      "epoch": 0.6247837192057345,
+      "epoch": 0.645544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 6.519037779678444e-06,
-      "loss": 0.7703,
+      "learning_rate": 5.897064671515616e-06,
+      "loss": 0.7422,
       "step": 22749
     },
     {
-      "epoch": 0.6248111834335778,
+      "epoch": 0.6455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 6.518203908828296e-06,
-      "loss": 0.7995,
+      "learning_rate": 5.896226532368847e-06,
+      "loss": 0.7868,
       "step": 22750
     },
     {
-      "epoch": 0.624838647661421,
+      "epoch": 0.6456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 6.517370065527464e-06,
-      "loss": 0.8195,
+      "learning_rate": 5.895388427886833e-06,
+      "loss": 0.8779,
       "step": 22751
     },
     {
-      "epoch": 0.6248661118892642,
+      "epoch": 0.6456299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 6.516536249782552e-06,
-      "loss": 0.7625,
+      "learning_rate": 5.894550358076661e-06,
+      "loss": 0.8754,
       "step": 22752
     },
     {
-      "epoch": 0.6248935761171075,
+      "epoch": 0.6456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 6.5157024616001504e-06,
-      "loss": 0.8313,
+      "learning_rate": 5.893712322945406e-06,
+      "loss": 0.8523,
       "step": 22753
     },
     {
-      "epoch": 0.6249210403449507,
+      "epoch": 0.6456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 6.514868700986862e-06,
-      "loss": 0.889,
+      "learning_rate": 5.8928743225001465e-06,
+      "loss": 0.8131,
       "step": 22754
     },
     {
-      "epoch": 0.6249485045727939,
+      "epoch": 0.6457150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 6.5140349679492835e-06,
-      "loss": 0.8945,
+      "learning_rate": 5.892036356747963e-06,
+      "loss": 0.7874,
       "step": 22755
     },
     {
-      "epoch": 0.6249759688006372,
+      "epoch": 0.6457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 6.5132012624940085e-06,
-      "loss": 0.7871,
+      "learning_rate": 5.891198425695934e-06,
+      "loss": 0.8792,
       "step": 22756
     },
     {
-      "epoch": 0.6250034330284804,
+      "epoch": 0.6457718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 6.512367584627641e-06,
-      "loss": 0.8247,
+      "learning_rate": 5.890360529351133e-06,
+      "loss": 0.8485,
       "step": 22757
     },
     {
-      "epoch": 0.6250308972563237,
+      "epoch": 0.6458002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 6.5115339343567685e-06,
-      "loss": 0.8515,
+      "learning_rate": 5.889522667720647e-06,
+      "loss": 0.8466,
       "step": 22758
     },
     {
-      "epoch": 0.6250583614841668,
+      "epoch": 0.6458286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 6.510700311687992e-06,
-      "loss": 0.855,
+      "learning_rate": 5.888684840811545e-06,
+      "loss": 0.8485,
       "step": 22759
     },
     {
-      "epoch": 0.6250858257120101,
+      "epoch": 0.6458569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 6.509866716627904e-06,
-      "loss": 0.7569,
+      "learning_rate": 5.8878470486309025e-06,
+      "loss": 0.8563,
       "step": 22760
     },
     {
-      "epoch": 0.6251132899398534,
+      "epoch": 0.6458853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 6.509033149183102e-06,
-      "loss": 0.9295,
+      "learning_rate": 5.887009291185803e-06,
+      "loss": 0.8274,
       "step": 22761
     },
     {
-      "epoch": 0.6251407541676965,
+      "epoch": 0.6459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 6.508199609360184e-06,
-      "loss": 0.9115,
+      "learning_rate": 5.8861715684833245e-06,
+      "loss": 0.8375,
       "step": 22762
     },
     {
-      "epoch": 0.6251682183955398,
+      "epoch": 0.6459421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 6.507366097165744e-06,
-      "loss": 0.8607,
+      "learning_rate": 5.885333880530539e-06,
+      "loss": 0.7975,
       "step": 22763
     },
     {
-      "epoch": 0.6251956826233831,
+      "epoch": 0.6459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 6.506532612606377e-06,
-      "loss": 0.9244,
+      "learning_rate": 5.8844962273345205e-06,
+      "loss": 0.9034,
       "step": 22764
     },
     {
-      "epoch": 0.6252231468512263,
+      "epoch": 0.6459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 6.505699155688674e-06,
-      "loss": 0.8775,
+      "learning_rate": 5.883658608902349e-06,
+      "loss": 0.8309,
       "step": 22765
     },
     {
-      "epoch": 0.6252506110790695,
+      "epoch": 0.6460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 6.504865726419234e-06,
-      "loss": 0.9019,
+      "learning_rate": 5.8828210252411e-06,
+      "loss": 0.8124,
       "step": 22766
     },
     {
-      "epoch": 0.6252780753069127,
+      "epoch": 0.64605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.50403232480465e-06,
-      "loss": 0.8405,
+      "learning_rate": 5.881983476357846e-06,
+      "loss": 0.8743,
       "step": 22767
     },
     {
-      "epoch": 0.625305539534756,
+      "epoch": 0.6460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 6.503198950851517e-06,
-      "loss": 0.8513,
+      "learning_rate": 5.8811459622596605e-06,
+      "loss": 0.8555,
       "step": 22768
     },
     {
-      "epoch": 0.6253330037625993,
+      "epoch": 0.6461123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 6.502365604566432e-06,
-      "loss": 0.8552,
+      "learning_rate": 5.880308482953623e-06,
+      "loss": 0.7437,
       "step": 22769
     },
     {
-      "epoch": 0.6253604679904424,
+      "epoch": 0.6461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 6.5015322859559805e-06,
-      "loss": 0.8655,
+      "learning_rate": 5.879471038446804e-06,
+      "loss": 0.811,
       "step": 22770
     },
     {
-      "epoch": 0.6253879322182857,
+      "epoch": 0.6461691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 6.500698995026762e-06,
-      "loss": 0.7527,
+      "learning_rate": 5.878633628746276e-06,
+      "loss": 0.9419,
       "step": 22771
     },
     {
-      "epoch": 0.6254153964461289,
+      "epoch": 0.6461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 6.499865731785368e-06,
-      "loss": 0.9726,
+      "learning_rate": 5.877796253859118e-06,
+      "loss": 0.7723,
       "step": 22772
     },
     {
-      "epoch": 0.6254428606739721,
+      "epoch": 0.6462258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 6.499032496238391e-06,
-      "loss": 0.8577,
+      "learning_rate": 5.876958913792401e-06,
+      "loss": 0.8994,
       "step": 22773
     },
     {
-      "epoch": 0.6254703249018154,
+      "epoch": 0.6462542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 6.4981992883924285e-06,
-      "loss": 0.8234,
+      "learning_rate": 5.876121608553194e-06,
+      "loss": 0.9119,
       "step": 22774
     },
     {
-      "epoch": 0.6254977891296586,
+      "epoch": 0.6462826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 6.497366108254071e-06,
-      "loss": 0.8549,
+      "learning_rate": 5.875284338148571e-06,
+      "loss": 0.8911,
       "step": 22775
     },
     {
-      "epoch": 0.6255252533575019,
+      "epoch": 0.6463110102156641,
       "grad_norm": 0.0,
-      "learning_rate": 6.496532955829904e-06,
-      "loss": 0.8282,
+      "learning_rate": 5.874447102585612e-06,
+      "loss": 0.7552,
       "step": 22776
     },
     {
-      "epoch": 0.6255527175853451,
+      "epoch": 0.6463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 6.4956998311265275e-06,
-      "loss": 0.8419,
+      "learning_rate": 5.873609901871382e-06,
+      "loss": 0.8364,
       "step": 22777
     },
     {
-      "epoch": 0.6255801818131883,
+      "epoch": 0.6463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 6.49486673415053e-06,
-      "loss": 0.9224,
+      "learning_rate": 5.872772736012955e-06,
+      "loss": 0.8356,
       "step": 22778
     },
     {
-      "epoch": 0.6256076460410316,
+      "epoch": 0.6463961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 6.494033664908508e-06,
-      "loss": 0.8429,
+      "learning_rate": 5.871935605017402e-06,
+      "loss": 0.8656,
       "step": 22779
     },
     {
-      "epoch": 0.6256351102688748,
+      "epoch": 0.6464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 6.493200623407047e-06,
-      "loss": 0.8189,
+      "learning_rate": 5.871098508891795e-06,
+      "loss": 0.8923,
       "step": 22780
     },
     {
-      "epoch": 0.625662574496718,
+      "epoch": 0.6464528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 6.492367609652745e-06,
-      "loss": 0.8457,
+      "learning_rate": 5.870261447643204e-06,
+      "loss": 0.7957,
       "step": 22781
     },
     {
-      "epoch": 0.6256900387245613,
+      "epoch": 0.6464812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 6.491534623652184e-06,
-      "loss": 0.8894,
+      "learning_rate": 5.8694244212787e-06,
+      "loss": 0.8238,
       "step": 22782
     },
     {
-      "epoch": 0.6257175029524045,
+      "epoch": 0.6465096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 6.490701665411965e-06,
-      "loss": 0.8512,
+      "learning_rate": 5.868587429805355e-06,
+      "loss": 0.7936,
       "step": 22783
     },
     {
-      "epoch": 0.6257449671802477,
+      "epoch": 0.6465380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 6.489868734938669e-06,
-      "loss": 0.8588,
+      "learning_rate": 5.867750473230236e-06,
+      "loss": 0.8994,
       "step": 22784
     },
     {
-      "epoch": 0.6257724314080909,
+      "epoch": 0.646566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 6.489035832238893e-06,
-      "loss": 0.7904,
+      "learning_rate": 5.866913551560416e-06,
+      "loss": 0.8843,
       "step": 22785
     },
     {
-      "epoch": 0.6257998956359342,
+      "epoch": 0.646594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 6.488202957319228e-06,
-      "loss": 0.7762,
+      "learning_rate": 5.866076664802962e-06,
+      "loss": 0.8086,
       "step": 22786
     },
     {
-      "epoch": 0.6258273598637775,
+      "epoch": 0.6466231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 6.487370110186262e-06,
-      "loss": 0.7995,
+      "learning_rate": 5.865239812964944e-06,
+      "loss": 0.8217,
       "step": 22787
     },
     {
-      "epoch": 0.6258548240916206,
+      "epoch": 0.6466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 6.486537290846581e-06,
-      "loss": 0.8733,
+      "learning_rate": 5.864402996053432e-06,
+      "loss": 0.7903,
       "step": 22788
     },
     {
-      "epoch": 0.6258822883194639,
+      "epoch": 0.6466799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 6.4857044993067775e-06,
-      "loss": 0.8298,
+      "learning_rate": 5.863566214075495e-06,
+      "loss": 0.8319,
       "step": 22789
     },
     {
-      "epoch": 0.6259097525473072,
+      "epoch": 0.6467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 6.484871735573441e-06,
-      "loss": 0.8263,
+      "learning_rate": 5.862729467038195e-06,
+      "loss": 0.7774,
       "step": 22790
     },
     {
-      "epoch": 0.6259372167751504,
+      "epoch": 0.6467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 6.484038999653164e-06,
-      "loss": 0.8184,
+      "learning_rate": 5.8618927549486095e-06,
+      "loss": 0.8859,
       "step": 22791
     },
     {
-      "epoch": 0.6259646810029936,
+      "epoch": 0.6467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.483206291552535e-06,
-      "loss": 0.8003,
+      "learning_rate": 5.861056077813799e-06,
+      "loss": 0.9761,
       "step": 22792
     },
     {
-      "epoch": 0.6259921452308368,
+      "epoch": 0.6467934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 6.482373611278134e-06,
-      "loss": 0.8728,
+      "learning_rate": 5.860219435640837e-06,
+      "loss": 0.7805,
       "step": 22793
     },
     {
-      "epoch": 0.6260196094586801,
+      "epoch": 0.6468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 6.481540958836556e-06,
-      "loss": 0.9754,
+      "learning_rate": 5.859382828436788e-06,
+      "loss": 0.8508,
       "step": 22794
     },
     {
-      "epoch": 0.6260470736865233,
+      "epoch": 0.646850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 6.480708334234393e-06,
-      "loss": 0.829,
+      "learning_rate": 5.858546256208715e-06,
+      "loss": 0.8134,
       "step": 22795
     },
     {
-      "epoch": 0.6260745379143665,
+      "epoch": 0.6468785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 6.4798757374782255e-06,
-      "loss": 0.9153,
+      "learning_rate": 5.857709718963691e-06,
+      "loss": 0.9216,
       "step": 22796
     },
     {
-      "epoch": 0.6261020021422098,
+      "epoch": 0.6469069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 6.479043168574643e-06,
-      "loss": 0.7893,
+      "learning_rate": 5.8568732167087786e-06,
+      "loss": 0.8506,
       "step": 22797
     },
     {
-      "epoch": 0.626129466370053,
+      "epoch": 0.6469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 6.4782106275302415e-06,
-      "loss": 0.8461,
+      "learning_rate": 5.856036749451043e-06,
+      "loss": 0.8203,
       "step": 22798
     },
     {
-      "epoch": 0.6261569305978962,
+      "epoch": 0.6469636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 6.477378114351599e-06,
-      "loss": 0.8611,
+      "learning_rate": 5.855200317197552e-06,
+      "loss": 0.8809,
       "step": 22799
     },
     {
-      "epoch": 0.6261843948257395,
+      "epoch": 0.6469920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 6.4765456290453035e-06,
-      "loss": 0.8195,
+      "learning_rate": 5.854363919955371e-06,
+      "loss": 0.831,
       "step": 22800
     },
     {
-      "epoch": 0.6262118590535827,
+      "epoch": 0.6470204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 6.475713171617944e-06,
-      "loss": 0.7475,
+      "learning_rate": 5.853527557731563e-06,
+      "loss": 0.9261,
       "step": 22801
     },
     {
-      "epoch": 0.626239323281426,
+      "epoch": 0.6470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 6.474880742076106e-06,
-      "loss": 0.7551,
+      "learning_rate": 5.852691230533196e-06,
+      "loss": 0.9099,
       "step": 22802
     },
     {
-      "epoch": 0.6262667875092692,
+      "epoch": 0.6470771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 6.47404834042638e-06,
-      "loss": 0.8704,
+      "learning_rate": 5.8518549383673316e-06,
+      "loss": 0.8016,
       "step": 22803
     },
     {
-      "epoch": 0.6262942517371124,
+      "epoch": 0.647105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.47321596667535e-06,
-      "loss": 0.7106,
+      "learning_rate": 5.851018681241034e-06,
+      "loss": 0.8732,
       "step": 22804
     },
     {
-      "epoch": 0.6263217159649557,
+      "epoch": 0.6471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 6.472383620829599e-06,
-      "loss": 0.9125,
+      "learning_rate": 5.850182459161369e-06,
+      "loss": 0.9015,
       "step": 22805
     },
     {
-      "epoch": 0.6263491801927988,
+      "epoch": 0.6471623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 6.471551302895713e-06,
-      "loss": 0.8924,
+      "learning_rate": 5.849346272135397e-06,
+      "loss": 0.8553,
       "step": 22806
     },
     {
-      "epoch": 0.6263766444206421,
+      "epoch": 0.6471906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 6.470719012880285e-06,
-      "loss": 0.8569,
+      "learning_rate": 5.848510120170185e-06,
+      "loss": 0.8821,
       "step": 22807
     },
     {
-      "epoch": 0.6264041086484854,
+      "epoch": 0.6472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 6.469886750789891e-06,
-      "loss": 0.7393,
+      "learning_rate": 5.847674003272798e-06,
+      "loss": 0.924,
       "step": 22808
     },
     {
-      "epoch": 0.6264315728763286,
+      "epoch": 0.6472474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 6.469054516631121e-06,
-      "loss": 0.7541,
+      "learning_rate": 5.84683792145029e-06,
+      "loss": 0.8509,
       "step": 22809
     },
     {
-      "epoch": 0.6264590371041718,
+      "epoch": 0.6472758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 6.4682223104105645e-06,
-      "loss": 0.8537,
+      "learning_rate": 5.846001874709733e-06,
+      "loss": 0.8868,
       "step": 22810
     },
     {
-      "epoch": 0.626486501332015,
+      "epoch": 0.6473041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 6.467390132134799e-06,
-      "loss": 0.7497,
+      "learning_rate": 5.845165863058184e-06,
+      "loss": 0.8455,
       "step": 22811
     },
     {
-      "epoch": 0.6265139655598583,
+      "epoch": 0.6473325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 6.466557981810408e-06,
-      "loss": 0.747,
+      "learning_rate": 5.844329886502704e-06,
+      "loss": 0.8338,
       "step": 22812
     },
     {
-      "epoch": 0.6265414297877016,
+      "epoch": 0.647360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 6.465725859443979e-06,
-      "loss": 0.8181,
+      "learning_rate": 5.843493945050359e-06,
+      "loss": 0.8479,
       "step": 22813
     },
     {
-      "epoch": 0.6265688940155447,
+      "epoch": 0.64738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 6.464893765042098e-06,
-      "loss": 0.7636,
+      "learning_rate": 5.842658038708206e-06,
+      "loss": 0.8402,
       "step": 22814
     },
     {
-      "epoch": 0.626596358243388,
+      "epoch": 0.6474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 6.4640616986113454e-06,
-      "loss": 0.7786,
+      "learning_rate": 5.841822167483306e-06,
+      "loss": 0.8128,
       "step": 22815
     },
     {
-      "epoch": 0.6266238224712313,
+      "epoch": 0.6474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 6.463229660158311e-06,
-      "loss": 0.8478,
+      "learning_rate": 5.840986331382724e-06,
+      "loss": 0.8785,
       "step": 22816
     },
     {
-      "epoch": 0.6266512866990744,
+      "epoch": 0.6474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 6.462397649689569e-06,
-      "loss": 0.8297,
+      "learning_rate": 5.840150530413518e-06,
+      "loss": 0.8736,
       "step": 22817
     },
     {
-      "epoch": 0.6266787509269177,
+      "epoch": 0.6475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 6.461565667211707e-06,
-      "loss": 0.7872,
+      "learning_rate": 5.839314764582743e-06,
+      "loss": 0.868,
       "step": 22818
     },
     {
-      "epoch": 0.6267062151547609,
+      "epoch": 0.6475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 6.460733712731311e-06,
-      "loss": 0.8908,
+      "learning_rate": 5.838479033897464e-06,
+      "loss": 0.9304,
       "step": 22819
     },
     {
-      "epoch": 0.6267336793826042,
+      "epoch": 0.6475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 6.459901786254957e-06,
-      "loss": 0.8222,
+      "learning_rate": 5.837643338364744e-06,
+      "loss": 0.9498,
       "step": 22820
     },
     {
-      "epoch": 0.6267611436104474,
+      "epoch": 0.6475879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 6.4590698877892325e-06,
-      "loss": 0.8453,
+      "learning_rate": 5.83680767799163e-06,
+      "loss": 0.7584,
       "step": 22821
     },
     {
-      "epoch": 0.6267886078382906,
+      "epoch": 0.647616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 6.458238017340724e-06,
-      "loss": 0.7961,
+      "learning_rate": 5.835972052785197e-06,
+      "loss": 0.7798,
       "step": 22822
     },
     {
-      "epoch": 0.6268160720661339,
+      "epoch": 0.6476447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 6.457406174916003e-06,
-      "loss": 0.8625,
+      "learning_rate": 5.835136462752491e-06,
+      "loss": 0.8827,
       "step": 22823
     },
     {
-      "epoch": 0.626843536293977,
+      "epoch": 0.6476730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 6.4565743605216566e-06,
-      "loss": 0.8829,
+      "learning_rate": 5.834300907900574e-06,
+      "loss": 0.8395,
       "step": 22824
     },
     {
-      "epoch": 0.6268710005218203,
+      "epoch": 0.6477014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 6.4557425741642655e-06,
-      "loss": 0.7563,
+      "learning_rate": 5.833465388236509e-06,
+      "loss": 0.8581,
       "step": 22825
     },
     {
-      "epoch": 0.6268984647496636,
+      "epoch": 0.6477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 6.454910815850415e-06,
-      "loss": 0.7575,
+      "learning_rate": 5.832629903767345e-06,
+      "loss": 0.8143,
       "step": 22826
     },
     {
-      "epoch": 0.6269259289775068,
+      "epoch": 0.6477582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 6.454079085586682e-06,
-      "loss": 0.8617,
+      "learning_rate": 5.8317944545001435e-06,
+      "loss": 0.7516,
       "step": 22827
     },
     {
-      "epoch": 0.62695339320535,
+      "epoch": 0.6477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 6.453247383379652e-06,
-      "loss": 0.8862,
+      "learning_rate": 5.830959040441966e-06,
+      "loss": 0.8085,
       "step": 22828
     },
     {
-      "epoch": 0.6269808574331933,
+      "epoch": 0.6478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 6.452415709235898e-06,
-      "loss": 0.8192,
+      "learning_rate": 5.83012366159986e-06,
+      "loss": 0.7662,
       "step": 22829
     },
     {
-      "epoch": 0.6270083216610365,
+      "epoch": 0.6478433598183881,
       "grad_norm": 0.0,
-      "learning_rate": 6.451584063162007e-06,
-      "loss": 0.7973,
+      "learning_rate": 5.829288317980888e-06,
+      "loss": 0.8085,
       "step": 22830
     },
     {
-      "epoch": 0.6270357858888798,
+      "epoch": 0.6478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 6.450752445164557e-06,
-      "loss": 0.9539,
+      "learning_rate": 5.828453009592108e-06,
+      "loss": 0.8669,
       "step": 22831
     },
     {
-      "epoch": 0.6270632501167229,
+      "epoch": 0.647900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 6.449920855250128e-06,
-      "loss": 0.7895,
+      "learning_rate": 5.82761773644057e-06,
+      "loss": 0.8123,
       "step": 22832
     },
     {
-      "epoch": 0.6270907143445662,
+      "epoch": 0.6479284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 6.449089293425299e-06,
-      "loss": 0.8195,
+      "learning_rate": 5.826782498533332e-06,
+      "loss": 0.8626,
       "step": 22833
     },
     {
-      "epoch": 0.6271181785724095,
+      "epoch": 0.6479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 6.448257759696657e-06,
-      "loss": 0.868,
+      "learning_rate": 5.825947295877455e-06,
+      "loss": 0.9269,
       "step": 22834
     },
     {
-      "epoch": 0.6271456428002526,
+      "epoch": 0.6479852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 6.447426254070772e-06,
-      "loss": 0.8347,
+      "learning_rate": 5.825112128479982e-06,
+      "loss": 0.9161,
       "step": 22835
     },
     {
-      "epoch": 0.6271731070280959,
+      "epoch": 0.6480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 6.446594776554225e-06,
-      "loss": 0.884,
+      "learning_rate": 5.8242769963479775e-06,
+      "loss": 0.8078,
       "step": 22836
     },
     {
-      "epoch": 0.6272005712559391,
+      "epoch": 0.6480419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 6.4457633271535965e-06,
-      "loss": 0.9045,
+      "learning_rate": 5.823441899488491e-06,
+      "loss": 0.8059,
       "step": 22837
     },
     {
-      "epoch": 0.6272280354837824,
+      "epoch": 0.6480703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 6.444931905875468e-06,
-      "loss": 0.8781,
+      "learning_rate": 5.8226068379085784e-06,
+      "loss": 0.8487,
       "step": 22838
     },
     {
-      "epoch": 0.6272554997116256,
+      "epoch": 0.6480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 6.444100512726412e-06,
-      "loss": 0.8404,
+      "learning_rate": 5.8217718116152985e-06,
+      "loss": 0.9036,
       "step": 22839
     },
     {
-      "epoch": 0.6272829639394688,
+      "epoch": 0.6481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 6.443269147713016e-06,
-      "loss": 0.7884,
+      "learning_rate": 5.820936820615697e-06,
+      "loss": 0.7987,
       "step": 22840
     },
     {
-      "epoch": 0.6273104281673121,
+      "epoch": 0.648155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 6.442437810841846e-06,
-      "loss": 0.9007,
+      "learning_rate": 5.820101864916827e-06,
+      "loss": 0.8936,
       "step": 22841
     },
     {
-      "epoch": 0.6273378923951554,
+      "epoch": 0.6481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 6.441606502119489e-06,
-      "loss": 0.8977,
+      "learning_rate": 5.819266944525752e-06,
+      "loss": 0.8513,
       "step": 22842
     },
     {
-      "epoch": 0.6273653566229985,
+      "epoch": 0.6482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 6.440775221552519e-06,
-      "loss": 0.8057,
+      "learning_rate": 5.818432059449511e-06,
+      "loss": 0.8467,
       "step": 22843
     },
     {
-      "epoch": 0.6273928208508418,
+      "epoch": 0.6482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 6.439943969147513e-06,
-      "loss": 0.8547,
+      "learning_rate": 5.817597209695163e-06,
+      "loss": 0.7819,
       "step": 22844
     },
     {
-      "epoch": 0.627420285078685,
+      "epoch": 0.6482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 6.439112744911049e-06,
-      "loss": 0.8589,
+      "learning_rate": 5.816762395269763e-06,
+      "loss": 0.8584,
       "step": 22845
     },
     {
-      "epoch": 0.6274477493065282,
+      "epoch": 0.6482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 6.43828154884971e-06,
-      "loss": 0.8794,
+      "learning_rate": 5.815927616180354e-06,
+      "loss": 0.8222,
       "step": 22846
     },
     {
-      "epoch": 0.6274752135343715,
+      "epoch": 0.6483257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 6.437450380970064e-06,
-      "loss": 0.9037,
+      "learning_rate": 5.815092872433994e-06,
+      "loss": 0.9135,
       "step": 22847
     },
     {
-      "epoch": 0.6275026777622147,
+      "epoch": 0.6483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 6.4366192412786875e-06,
-      "loss": 0.8279,
+      "learning_rate": 5.814258164037736e-06,
+      "loss": 0.8254,
       "step": 22848
     },
     {
-      "epoch": 0.627530141990058,
+      "epoch": 0.6483825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 6.435788129782161e-06,
-      "loss": 0.7471,
+      "learning_rate": 5.813423490998624e-06,
+      "loss": 1.0046,
       "step": 22849
     },
     {
-      "epoch": 0.6275576062179011,
+      "epoch": 0.648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 6.434957046487062e-06,
-      "loss": 0.8139,
+      "learning_rate": 5.812588853323713e-06,
+      "loss": 0.8512,
       "step": 22850
     },
     {
-      "epoch": 0.6275850704457444,
+      "epoch": 0.648439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 6.434125991399963e-06,
-      "loss": 0.8205,
+      "learning_rate": 5.8117542510200545e-06,
+      "loss": 0.8314,
       "step": 22851
     },
     {
-      "epoch": 0.6276125346735877,
+      "epoch": 0.6484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 6.433294964527441e-06,
-      "loss": 0.7945,
+      "learning_rate": 5.810919684094689e-06,
+      "loss": 0.9276,
       "step": 22852
     },
     {
-      "epoch": 0.6276399989014309,
+      "epoch": 0.6484960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 6.432463965876068e-06,
-      "loss": 0.8686,
+      "learning_rate": 5.810085152554681e-06,
+      "loss": 0.8615,
       "step": 22853
     },
     {
-      "epoch": 0.6276674631292741,
+      "epoch": 0.6485244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 6.431632995452422e-06,
-      "loss": 0.8467,
+      "learning_rate": 5.809250656407067e-06,
+      "loss": 0.7124,
       "step": 22854
     },
     {
-      "epoch": 0.6276949273571174,
+      "epoch": 0.6485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 6.430802053263081e-06,
-      "loss": 0.8444,
+      "learning_rate": 5.8084161956589016e-06,
+      "loss": 0.7984,
       "step": 22855
     },
     {
-      "epoch": 0.6277223915849606,
+      "epoch": 0.6485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 6.429971139314615e-06,
-      "loss": 0.8988,
+      "learning_rate": 5.807581770317237e-06,
+      "loss": 0.9153,
       "step": 22856
     },
     {
-      "epoch": 0.6277498558128038,
+      "epoch": 0.6486095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 6.429140253613604e-06,
-      "loss": 0.7473,
+      "learning_rate": 5.806747380389115e-06,
+      "loss": 0.9127,
       "step": 22857
     },
     {
-      "epoch": 0.627777320040647,
+      "epoch": 0.6486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 6.428309396166613e-06,
-      "loss": 0.9481,
+      "learning_rate": 5.805913025881586e-06,
+      "loss": 0.856,
       "step": 22858
     },
     {
-      "epoch": 0.6278047842684903,
+      "epoch": 0.6486662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 6.427478566980225e-06,
-      "loss": 0.8538,
+      "learning_rate": 5.805078706801701e-06,
+      "loss": 0.7581,
       "step": 22859
     },
     {
-      "epoch": 0.6278322484963336,
+      "epoch": 0.6486946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 6.426647766061007e-06,
-      "loss": 0.9019,
+      "learning_rate": 5.804244423156502e-06,
+      "loss": 0.9723,
       "step": 22860
     },
     {
-      "epoch": 0.6278597127241767,
+      "epoch": 0.6487230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 6.425816993415538e-06,
-      "loss": 0.8723,
+      "learning_rate": 5.803410174953037e-06,
+      "loss": 0.9231,
       "step": 22861
     },
     {
-      "epoch": 0.62788717695202,
+      "epoch": 0.6487514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 6.4249862490503916e-06,
-      "loss": 0.9003,
+      "learning_rate": 5.80257596219836e-06,
+      "loss": 0.8357,
       "step": 22862
     },
     {
-      "epoch": 0.6279146411798632,
+      "epoch": 0.6487797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 6.424155532972136e-06,
-      "loss": 0.8715,
+      "learning_rate": 5.80174178489951e-06,
+      "loss": 0.8093,
       "step": 22863
     },
     {
-      "epoch": 0.6279421054077065,
+      "epoch": 0.6488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 6.4233248451873495e-06,
-      "loss": 0.8476,
+      "learning_rate": 5.800907643063534e-06,
+      "loss": 0.8275,
       "step": 22864
     },
     {
-      "epoch": 0.6279695696355497,
+      "epoch": 0.6488365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 6.422494185702599e-06,
-      "loss": 0.8263,
+      "learning_rate": 5.8000735366974845e-06,
+      "loss": 0.8813,
       "step": 22865
     },
     {
-      "epoch": 0.6279970338633929,
+      "epoch": 0.6488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 6.421663554524462e-06,
-      "loss": 0.8835,
+      "learning_rate": 5.7992394658083996e-06,
+      "loss": 0.7861,
       "step": 22866
     },
     {
-      "epoch": 0.6280244980912362,
+      "epoch": 0.6488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 6.420832951659509e-06,
-      "loss": 0.8,
+      "learning_rate": 5.7984054304033264e-06,
+      "loss": 0.7688,
       "step": 22867
     },
     {
-      "epoch": 0.6280519623190794,
+      "epoch": 0.6489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 6.420002377114311e-06,
-      "loss": 0.9254,
+      "learning_rate": 5.7975714304893116e-06,
+      "loss": 0.8802,
       "step": 22868
     },
     {
-      "epoch": 0.6280794265469226,
+      "epoch": 0.648950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 6.419171830895445e-06,
-      "loss": 0.8831,
+      "learning_rate": 5.796737466073401e-06,
+      "loss": 1.0013,
       "step": 22869
     },
     {
-      "epoch": 0.6281068907747659,
+      "epoch": 0.6489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 6.418341313009475e-06,
-      "loss": 0.8508,
+      "learning_rate": 5.795903537162641e-06,
+      "loss": 0.8345,
       "step": 22870
     },
     {
-      "epoch": 0.6281343550026091,
+      "epoch": 0.6490068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 6.417510823462978e-06,
-      "loss": 0.7852,
+      "learning_rate": 5.7950696437640685e-06,
+      "loss": 0.7841,
       "step": 22871
     },
     {
-      "epoch": 0.6281618192304523,
+      "epoch": 0.6490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 6.41668036226252e-06,
-      "loss": 0.9375,
+      "learning_rate": 5.794235785884732e-06,
+      "loss": 0.9678,
       "step": 22872
     },
     {
-      "epoch": 0.6281892834582956,
+      "epoch": 0.6490635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 6.415849929414676e-06,
-      "loss": 0.873,
+      "learning_rate": 5.79340196353168e-06,
+      "loss": 0.9371,
       "step": 22873
     },
     {
-      "epoch": 0.6282167476861388,
+      "epoch": 0.6490919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 6.415019524926018e-06,
-      "loss": 0.8157,
+      "learning_rate": 5.792568176711945e-06,
+      "loss": 0.7872,
       "step": 22874
     },
     {
-      "epoch": 0.628244211913982,
+      "epoch": 0.6491203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 6.414189148803113e-06,
-      "loss": 0.8716,
+      "learning_rate": 5.791734425432576e-06,
+      "loss": 0.7201,
       "step": 22875
     },
     {
-      "epoch": 0.6282716761418252,
+      "epoch": 0.6491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 6.413358801052532e-06,
-      "loss": 0.9025,
+      "learning_rate": 5.79090070970062e-06,
+      "loss": 0.782,
       "step": 22876
     },
     {
-      "epoch": 0.6282991403696685,
+      "epoch": 0.6491770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 6.412528481680845e-06,
-      "loss": 0.7906,
+      "learning_rate": 5.790067029523111e-06,
+      "loss": 0.8587,
       "step": 22877
     },
     {
-      "epoch": 0.6283266045975118,
+      "epoch": 0.6492054483541431,
       "grad_norm": 0.0,
-      "learning_rate": 6.411698190694622e-06,
-      "loss": 0.8984,
+      "learning_rate": 5.789233384907095e-06,
+      "loss": 0.8169,
       "step": 22878
     },
     {
-      "epoch": 0.6283540688253549,
+      "epoch": 0.6492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 6.410867928100435e-06,
-      "loss": 0.9683,
+      "learning_rate": 5.788399775859617e-06,
+      "loss": 0.8663,
       "step": 22879
     },
     {
-      "epoch": 0.6283815330531982,
+      "epoch": 0.6492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 6.41003769390485e-06,
-      "loss": 0.9661,
+      "learning_rate": 5.7875662023877135e-06,
+      "loss": 0.8907,
       "step": 22880
     },
     {
-      "epoch": 0.6284089972810415,
+      "epoch": 0.6492905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 6.40920748811444e-06,
-      "loss": 0.7692,
+      "learning_rate": 5.7867326644984254e-06,
+      "loss": 0.8589,
       "step": 22881
     },
     {
-      "epoch": 0.6284364615088847,
+      "epoch": 0.6493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 6.408377310735767e-06,
-      "loss": 0.8455,
+      "learning_rate": 5.785899162198801e-06,
+      "loss": 0.85,
       "step": 22882
     },
     {
-      "epoch": 0.6284639257367279,
+      "epoch": 0.6493473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 6.407547161775406e-06,
-      "loss": 0.8467,
+      "learning_rate": 5.785065695495868e-06,
+      "loss": 0.7414,
       "step": 22883
     },
     {
-      "epoch": 0.6284913899645711,
+      "epoch": 0.6493757094211123,
       "grad_norm": 0.0,
-      "learning_rate": 6.406717041239922e-06,
-      "loss": 0.776,
+      "learning_rate": 5.784232264396682e-06,
+      "loss": 0.926,
       "step": 22884
     },
     {
-      "epoch": 0.6285188541924144,
+      "epoch": 0.6494040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 6.405886949135884e-06,
-      "loss": 0.8582,
+      "learning_rate": 5.783398868908272e-06,
+      "loss": 0.8122,
       "step": 22885
     },
     {
-      "epoch": 0.6285463184202577,
+      "epoch": 0.6494324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 6.4050568854698645e-06,
-      "loss": 0.8471,
+      "learning_rate": 5.782565509037681e-06,
+      "loss": 0.8623,
       "step": 22886
     },
     {
-      "epoch": 0.6285737826481008,
+      "epoch": 0.649460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 6.4042268502484265e-06,
-      "loss": 0.7845,
+      "learning_rate": 5.781732184791953e-06,
+      "loss": 0.8893,
       "step": 22887
     },
     {
-      "epoch": 0.6286012468759441,
+      "epoch": 0.649489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 6.403396843478138e-06,
-      "loss": 0.8162,
+      "learning_rate": 5.7808988961781186e-06,
+      "loss": 0.8924,
       "step": 22888
     },
     {
-      "epoch": 0.6286287111037873,
+      "epoch": 0.6495175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 6.402566865165565e-06,
-      "loss": 0.8736,
+      "learning_rate": 5.780065643203221e-06,
+      "loss": 0.9455,
       "step": 22889
     },
     {
-      "epoch": 0.6286561753316305,
+      "epoch": 0.6495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 6.401736915317277e-06,
-      "loss": 0.7806,
+      "learning_rate": 5.779232425874303e-06,
+      "loss": 0.8288,
       "step": 22890
     },
     {
-      "epoch": 0.6286836395594738,
+      "epoch": 0.6495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 6.400906993939842e-06,
-      "loss": 0.8157,
+      "learning_rate": 5.7783992441983936e-06,
+      "loss": 0.9302,
       "step": 22891
     },
     {
-      "epoch": 0.628711103787317,
+      "epoch": 0.6496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 6.400077101039823e-06,
-      "loss": 0.8511,
+      "learning_rate": 5.7775660981825365e-06,
+      "loss": 0.739,
       "step": 22892
     },
     {
-      "epoch": 0.6287385680151603,
+      "epoch": 0.6496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 6.3992472366237935e-06,
-      "loss": 0.7733,
+      "learning_rate": 5.7767329878337705e-06,
+      "loss": 0.8113,
       "step": 22893
     },
     {
-      "epoch": 0.6287660322430035,
+      "epoch": 0.6496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 6.398417400698309e-06,
-      "loss": 0.8294,
+      "learning_rate": 5.775899913159129e-06,
+      "loss": 0.7565,
       "step": 22894
     },
     {
-      "epoch": 0.6287934964708467,
+      "epoch": 0.6496878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 6.397587593269944e-06,
-      "loss": 0.849,
+      "learning_rate": 5.775066874165649e-06,
+      "loss": 0.7387,
       "step": 22895
     },
     {
-      "epoch": 0.62882096069869,
+      "epoch": 0.6497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 6.396757814345259e-06,
-      "loss": 0.8377,
+      "learning_rate": 5.774233870860375e-06,
+      "loss": 0.8873,
       "step": 22896
     },
     {
-      "epoch": 0.6288484249265331,
+      "epoch": 0.649744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 6.395928063930823e-06,
-      "loss": 0.9019,
+      "learning_rate": 5.773400903250332e-06,
+      "loss": 0.8955,
       "step": 22897
     },
     {
-      "epoch": 0.6288758891543764,
+      "epoch": 0.6497729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 6.395098342033201e-06,
-      "loss": 0.9084,
+      "learning_rate": 5.7725679713425575e-06,
+      "loss": 0.7718,
       "step": 22898
     },
     {
-      "epoch": 0.6289033533822197,
+      "epoch": 0.6498013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 6.3942686486589574e-06,
-      "loss": 0.8943,
+      "learning_rate": 5.7717350751441e-06,
+      "loss": 0.8861,
       "step": 22899
     },
     {
-      "epoch": 0.6289308176100629,
+      "epoch": 0.6498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 6.393438983814657e-06,
-      "loss": 0.872,
+      "learning_rate": 5.770902214661983e-06,
+      "loss": 0.8377,
       "step": 22900
     },
     {
-      "epoch": 0.6289582818379061,
+      "epoch": 0.6498581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 6.392609347506864e-06,
-      "loss": 0.9503,
+      "learning_rate": 5.7700693899032444e-06,
+      "loss": 0.8001,
       "step": 22901
     },
     {
-      "epoch": 0.6289857460657493,
+      "epoch": 0.6498864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 6.391779739742141e-06,
-      "loss": 0.7547,
+      "learning_rate": 5.769236600874924e-06,
+      "loss": 0.7772,
       "step": 22902
     },
     {
-      "epoch": 0.6290132102935926,
+      "epoch": 0.6499148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 6.3909501605270586e-06,
-      "loss": 0.8498,
+      "learning_rate": 5.768403847584048e-06,
+      "loss": 0.8884,
       "step": 22903
     },
     {
-      "epoch": 0.6290406745214359,
+      "epoch": 0.6499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 6.390120609868174e-06,
-      "loss": 0.9358,
+      "learning_rate": 5.767571130037655e-06,
+      "loss": 0.8248,
       "step": 22904
     },
     {
-      "epoch": 0.629068138749279,
+      "epoch": 0.6499716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 6.389291087772056e-06,
-      "loss": 0.8535,
+      "learning_rate": 5.766738448242783e-06,
+      "loss": 0.8969,
       "step": 22905
     },
     {
-      "epoch": 0.6290956029771223,
+      "epoch": 0.65,
       "grad_norm": 0.0,
-      "learning_rate": 6.388461594245261e-06,
-      "loss": 0.8243,
+      "learning_rate": 5.765905802206457e-06,
+      "loss": 0.8817,
       "step": 22906
     },
     {
-      "epoch": 0.6291230672049656,
+      "epoch": 0.6500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 6.387632129294361e-06,
-      "loss": 0.9317,
+      "learning_rate": 5.76507319193572e-06,
+      "loss": 0.8077,
       "step": 22907
     },
     {
-      "epoch": 0.6291505314328087,
+      "epoch": 0.6500567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 6.386802692925912e-06,
-      "loss": 0.876,
+      "learning_rate": 5.764240617437595e-06,
+      "loss": 0.8165,
       "step": 22908
     },
     {
-      "epoch": 0.629177995660652,
+      "epoch": 0.6500851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 6.385973285146481e-06,
-      "loss": 0.8276,
+      "learning_rate": 5.763408078719121e-06,
+      "loss": 0.8967,
       "step": 22909
     },
     {
-      "epoch": 0.6292054598884952,
+      "epoch": 0.6501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 6.38514390596263e-06,
-      "loss": 0.8969,
+      "learning_rate": 5.762575575787333e-06,
+      "loss": 0.9297,
       "step": 22910
     },
     {
-      "epoch": 0.6292329241163385,
+      "epoch": 0.6501418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 6.38431455538092e-06,
-      "loss": 0.883,
+      "learning_rate": 5.761743108649256e-06,
+      "loss": 0.7974,
       "step": 22911
     },
     {
-      "epoch": 0.6292603883441817,
+      "epoch": 0.6501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 6.383485233407914e-06,
-      "loss": 0.923,
+      "learning_rate": 5.760910677311925e-06,
+      "loss": 0.9186,
       "step": 22912
     },
     {
-      "epoch": 0.6292878525720249,
+      "epoch": 0.6501986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 6.382655940050173e-06,
-      "loss": 0.8555,
+      "learning_rate": 5.760078281782372e-06,
+      "loss": 0.7738,
       "step": 22913
     },
     {
-      "epoch": 0.6293153167998682,
+      "epoch": 0.6502270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 6.381826675314258e-06,
-      "loss": 0.8875,
+      "learning_rate": 5.759245922067629e-06,
+      "loss": 0.8266,
       "step": 22914
     },
     {
-      "epoch": 0.6293427810277114,
+      "epoch": 0.650255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 6.3809974392067354e-06,
-      "loss": 0.9189,
+      "learning_rate": 5.758413598174726e-06,
+      "loss": 0.8733,
       "step": 22915
     },
     {
-      "epoch": 0.6293702452555546,
+      "epoch": 0.650283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 6.380168231734164e-06,
-      "loss": 0.8797,
+      "learning_rate": 5.757581310110697e-06,
+      "loss": 0.8904,
       "step": 22916
     },
     {
-      "epoch": 0.6293977094833979,
+      "epoch": 0.6503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 6.3793390529031e-06,
-      "loss": 0.8511,
+      "learning_rate": 5.756749057882567e-06,
+      "loss": 0.8742,
       "step": 22917
     },
     {
-      "epoch": 0.6294251737112411,
+      "epoch": 0.6503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 6.378509902720107e-06,
-      "loss": 0.8182,
+      "learning_rate": 5.7559168414973665e-06,
+      "loss": 0.8857,
       "step": 22918
     },
     {
-      "epoch": 0.6294526379390843,
+      "epoch": 0.6503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 6.377680781191751e-06,
-      "loss": 0.9979,
+      "learning_rate": 5.755084660962134e-06,
+      "loss": 0.7559,
       "step": 22919
     },
     {
-      "epoch": 0.6294801021669276,
+      "epoch": 0.6503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 6.376851688324584e-06,
-      "loss": 0.9551,
+      "learning_rate": 5.754252516283886e-06,
+      "loss": 0.9029,
       "step": 22920
     },
     {
-      "epoch": 0.6295075663947708,
+      "epoch": 0.6504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 6.376022624125169e-06,
-      "loss": 0.7668,
+      "learning_rate": 5.7534204074696584e-06,
+      "loss": 0.9128,
       "step": 22921
     },
     {
-      "epoch": 0.6295350306226141,
+      "epoch": 0.6504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 6.3751935886000724e-06,
-      "loss": 0.9371,
+      "learning_rate": 5.7525883345264835e-06,
+      "loss": 0.9063,
       "step": 22922
     },
     {
-      "epoch": 0.6295624948504572,
+      "epoch": 0.6504824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 6.374364581755843e-06,
-      "loss": 0.8243,
+      "learning_rate": 5.751756297461381e-06,
+      "loss": 0.897,
       "step": 22923
     },
     {
-      "epoch": 0.6295899590783005,
+      "epoch": 0.650510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 6.3735356035990485e-06,
-      "loss": 0.7883,
+      "learning_rate": 5.750924296281386e-06,
+      "loss": 0.7817,
       "step": 22924
     },
     {
-      "epoch": 0.6296174233061438,
+      "epoch": 0.6505391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 6.3727066541362425e-06,
-      "loss": 0.8208,
+      "learning_rate": 5.750092330993527e-06,
+      "loss": 0.9395,
       "step": 22925
     },
     {
-      "epoch": 0.629644887533987,
+      "epoch": 0.6505675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 6.3718777333739874e-06,
-      "loss": 0.9019,
+      "learning_rate": 5.7492604016048256e-06,
+      "loss": 0.8975,
       "step": 22926
     },
     {
-      "epoch": 0.6296723517618302,
+      "epoch": 0.6505959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 6.371048841318842e-06,
-      "loss": 0.9032,
+      "learning_rate": 5.748428508122312e-06,
+      "loss": 0.9081,
       "step": 22927
     },
     {
-      "epoch": 0.6296998159896734,
+      "epoch": 0.6506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 6.370219977977366e-06,
-      "loss": 0.9207,
+      "learning_rate": 5.74759665055302e-06,
+      "loss": 0.7841,
       "step": 22928
     },
     {
-      "epoch": 0.6297272802175167,
+      "epoch": 0.6506526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 6.369391143356111e-06,
-      "loss": 0.8346,
+      "learning_rate": 5.74676482890396e-06,
+      "loss": 0.8844,
       "step": 22929
     },
     {
-      "epoch": 0.62975474444536,
+      "epoch": 0.6506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 6.368562337461641e-06,
-      "loss": 0.8246,
+      "learning_rate": 5.745933043182177e-06,
+      "loss": 0.7087,
       "step": 22930
     },
     {
-      "epoch": 0.6297822086732031,
+      "epoch": 0.6507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 6.367733560300515e-06,
-      "loss": 0.9939,
+      "learning_rate": 5.745101293394686e-06,
+      "loss": 0.8043,
       "step": 22931
     },
     {
-      "epoch": 0.6298096729010464,
+      "epoch": 0.6507377979568671,
       "grad_norm": 0.0,
-      "learning_rate": 6.366904811879284e-06,
-      "loss": 0.8462,
+      "learning_rate": 5.744269579548515e-06,
+      "loss": 0.7049,
       "step": 22932
     },
     {
-      "epoch": 0.6298371371288897,
+      "epoch": 0.6507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 6.366076092204512e-06,
-      "loss": 0.787,
+      "learning_rate": 5.743437901650695e-06,
+      "loss": 0.8675,
       "step": 22933
     },
     {
-      "epoch": 0.6298646013567328,
+      "epoch": 0.650794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 6.365247401282756e-06,
-      "loss": 0.8467,
+      "learning_rate": 5.742606259708241e-06,
+      "loss": 0.8636,
       "step": 22934
     },
     {
-      "epoch": 0.6298920655845761,
+      "epoch": 0.6508229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 6.364418739120568e-06,
-      "loss": 0.8515,
+      "learning_rate": 5.7417746537281825e-06,
+      "loss": 0.8256,
       "step": 22935
     },
     {
-      "epoch": 0.6299195298124193,
+      "epoch": 0.6508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 6.363590105724507e-06,
-      "loss": 0.8421,
+      "learning_rate": 5.740943083717551e-06,
+      "loss": 0.7616,
       "step": 22936
     },
     {
-      "epoch": 0.6299469940402626,
+      "epoch": 0.6508796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 6.3627615011011276e-06,
-      "loss": 0.9706,
+      "learning_rate": 5.7401115496833605e-06,
+      "loss": 0.8422,
       "step": 22937
     },
     {
-      "epoch": 0.6299744582681058,
+      "epoch": 0.6509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 6.36193292525699e-06,
-      "loss": 0.9785,
+      "learning_rate": 5.739280051632639e-06,
+      "loss": 0.8096,
       "step": 22938
     },
     {
-      "epoch": 0.630001922495949,
+      "epoch": 0.6509364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 6.361104378198648e-06,
-      "loss": 0.9245,
+      "learning_rate": 5.738448589572414e-06,
+      "loss": 0.9067,
       "step": 22939
     },
     {
-      "epoch": 0.6300293867237923,
+      "epoch": 0.6509648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 6.360275859932662e-06,
-      "loss": 0.859,
+      "learning_rate": 5.737617163509701e-06,
+      "loss": 0.7782,
       "step": 22940
     },
     {
-      "epoch": 0.6300568509516355,
+      "epoch": 0.6509931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 6.3594473704655794e-06,
-      "loss": 0.8619,
+      "learning_rate": 5.736785773451528e-06,
+      "loss": 0.8239,
       "step": 22941
     },
     {
-      "epoch": 0.6300843151794787,
+      "epoch": 0.6510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 6.358618909803957e-06,
-      "loss": 0.7615,
+      "learning_rate": 5.735954419404921e-06,
+      "loss": 0.8533,
       "step": 22942
     },
     {
-      "epoch": 0.630111779407322,
+      "epoch": 0.651049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 6.357790477954357e-06,
-      "loss": 0.7978,
+      "learning_rate": 5.735123101376895e-06,
+      "loss": 0.8941,
       "step": 22943
     },
     {
-      "epoch": 0.6301392436351652,
+      "epoch": 0.6510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 6.356962074923325e-06,
-      "loss": 0.8268,
+      "learning_rate": 5.734291819374475e-06,
+      "loss": 0.8118,
       "step": 22944
     },
     {
-      "epoch": 0.6301667078630084,
+      "epoch": 0.6511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 6.3561337007174215e-06,
-      "loss": 0.8235,
+      "learning_rate": 5.7334605734046855e-06,
+      "loss": 0.8061,
       "step": 22945
     },
     {
-      "epoch": 0.6301941720908517,
+      "epoch": 0.6511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 6.355305355343204e-06,
-      "loss": 0.9632,
+      "learning_rate": 5.732629363474544e-06,
+      "loss": 0.7721,
       "step": 22946
     },
     {
-      "epoch": 0.6302216363186949,
+      "epoch": 0.6511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 6.3544770388072205e-06,
-      "loss": 0.7634,
+      "learning_rate": 5.73179818959108e-06,
+      "loss": 0.8549,
       "step": 22947
     },
     {
-      "epoch": 0.6302491005465382,
+      "epoch": 0.6511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 6.353648751116025e-06,
-      "loss": 0.8972,
+      "learning_rate": 5.730967051761305e-06,
+      "loss": 0.8355,
       "step": 22948
     },
     {
-      "epoch": 0.6302765647743813,
+      "epoch": 0.6512202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 6.352820492276172e-06,
-      "loss": 0.8632,
+      "learning_rate": 5.730135949992242e-06,
+      "loss": 0.8842,
       "step": 22949
     },
     {
-      "epoch": 0.6303040290022246,
+      "epoch": 0.6512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 6.351992262294217e-06,
-      "loss": 0.7406,
+      "learning_rate": 5.729304884290916e-06,
+      "loss": 0.7839,
       "step": 22950
     },
     {
-      "epoch": 0.6303314932300679,
+      "epoch": 0.6512769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 6.351164061176713e-06,
-      "loss": 0.8396,
+      "learning_rate": 5.72847385466434e-06,
+      "loss": 0.8758,
       "step": 22951
     },
     {
-      "epoch": 0.630358957457911,
+      "epoch": 0.651305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 6.350335888930217e-06,
-      "loss": 0.8295,
+      "learning_rate": 5.727642861119538e-06,
+      "loss": 0.8692,
       "step": 22952
     },
     {
-      "epoch": 0.6303864216857543,
+      "epoch": 0.65133371169126,
       "grad_norm": 0.0,
-      "learning_rate": 6.349507745561271e-06,
-      "loss": 0.7756,
+      "learning_rate": 5.726811903663533e-06,
+      "loss": 0.7716,
       "step": 22953
     },
     {
-      "epoch": 0.6304138859135976,
+      "epoch": 0.6513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 6.348679631076434e-06,
-      "loss": 0.9166,
+      "learning_rate": 5.725980982303335e-06,
+      "loss": 0.907,
       "step": 22954
     },
     {
-      "epoch": 0.6304413501414408,
+      "epoch": 0.6513904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 6.34785154548226e-06,
-      "loss": 0.8058,
+      "learning_rate": 5.725150097045968e-06,
+      "loss": 0.8801,
       "step": 22955
     },
     {
-      "epoch": 0.630468814369284,
+      "epoch": 0.6514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 6.347023488785298e-06,
-      "loss": 0.8156,
+      "learning_rate": 5.724319247898456e-06,
+      "loss": 0.9284,
       "step": 22956
     },
     {
-      "epoch": 0.6304962785971272,
+      "epoch": 0.6514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 6.346195460992102e-06,
-      "loss": 0.8525,
+      "learning_rate": 5.723488434867805e-06,
+      "loss": 0.7834,
       "step": 22957
     },
     {
-      "epoch": 0.6305237428249705,
+      "epoch": 0.6514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 6.345367462109224e-06,
-      "loss": 0.7457,
+      "learning_rate": 5.722657657961041e-06,
+      "loss": 0.9018,
       "step": 22958
     },
     {
-      "epoch": 0.6305512070528138,
+      "epoch": 0.6515039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 6.344539492143214e-06,
-      "loss": 0.7661,
+      "learning_rate": 5.721826917185179e-06,
+      "loss": 0.9201,
       "step": 22959
     },
     {
-      "epoch": 0.6305786712806569,
+      "epoch": 0.6515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 6.343711551100622e-06,
-      "loss": 0.8394,
+      "learning_rate": 5.7209962125472385e-06,
+      "loss": 0.7391,
       "step": 22960
     },
     {
-      "epoch": 0.6306061355085002,
+      "epoch": 0.6515607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 6.342883638988e-06,
-      "loss": 0.87,
+      "learning_rate": 5.720165544054239e-06,
+      "loss": 0.7924,
       "step": 22961
     },
     {
-      "epoch": 0.6306335997363434,
+      "epoch": 0.6515891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 6.342055755811903e-06,
-      "loss": 0.7983,
+      "learning_rate": 5.719334911713189e-06,
+      "loss": 0.8139,
       "step": 22962
     },
     {
-      "epoch": 0.6306610639641866,
+      "epoch": 0.6516174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 6.341227901578872e-06,
-      "loss": 0.8417,
+      "learning_rate": 5.71850431553111e-06,
+      "loss": 0.8272,
       "step": 22963
     },
     {
-      "epoch": 0.6306885281920299,
+      "epoch": 0.6516458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 6.34040007629547e-06,
-      "loss": 0.8489,
+      "learning_rate": 5.717673755515021e-06,
+      "loss": 0.8769,
       "step": 22964
     },
     {
-      "epoch": 0.6307159924198731,
+      "epoch": 0.6516742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 6.339572279968236e-06,
-      "loss": 0.905,
+      "learning_rate": 5.71684323167193e-06,
+      "loss": 0.7729,
       "step": 22965
     },
     {
-      "epoch": 0.6307434566477164,
+      "epoch": 0.6517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 6.338744512603722e-06,
-      "loss": 0.818,
+      "learning_rate": 5.716012744008858e-06,
+      "loss": 0.8352,
       "step": 22966
     },
     {
-      "epoch": 0.6307709208755596,
+      "epoch": 0.6517309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 6.3379167742084844e-06,
-      "loss": 0.9493,
+      "learning_rate": 5.7151822925328215e-06,
+      "loss": 0.8576,
       "step": 22967
     },
     {
-      "epoch": 0.6307983851034028,
+      "epoch": 0.6517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 6.3370890647890645e-06,
-      "loss": 0.8391,
+      "learning_rate": 5.71435187725083e-06,
+      "loss": 0.8508,
       "step": 22968
     },
     {
-      "epoch": 0.6308258493312461,
+      "epoch": 0.6517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 6.3362613843520145e-06,
-      "loss": 0.7227,
+      "learning_rate": 5.713521498169901e-06,
+      "loss": 0.869,
       "step": 22969
     },
     {
-      "epoch": 0.6308533135590892,
+      "epoch": 0.6518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 6.335433732903889e-06,
-      "loss": 0.9121,
+      "learning_rate": 5.7126911552970525e-06,
+      "loss": 0.865,
       "step": 22970
     },
     {
-      "epoch": 0.6308807777869325,
+      "epoch": 0.651844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 6.334606110451228e-06,
-      "loss": 0.9099,
+      "learning_rate": 5.711860848639289e-06,
+      "loss": 0.8598,
       "step": 22971
     },
     {
-      "epoch": 0.6309082420147758,
+      "epoch": 0.6518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 6.333778517000582e-06,
-      "loss": 0.9016,
+      "learning_rate": 5.711030578203632e-06,
+      "loss": 0.733,
       "step": 22972
     },
     {
-      "epoch": 0.630935706242619,
+      "epoch": 0.6519012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 6.3329509525585e-06,
-      "loss": 0.7974,
+      "learning_rate": 5.710200343997094e-06,
+      "loss": 0.9398,
       "step": 22973
     },
     {
-      "epoch": 0.6309631704704622,
+      "epoch": 0.6519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 6.332123417131533e-06,
-      "loss": 0.8445,
+      "learning_rate": 5.709370146026684e-06,
+      "loss": 0.7581,
       "step": 22974
     },
     {
-      "epoch": 0.6309906346983054,
+      "epoch": 0.6519580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 6.331295910726225e-06,
-      "loss": 0.8254,
+      "learning_rate": 5.7085399842994175e-06,
+      "loss": 0.8077,
       "step": 22975
     },
     {
-      "epoch": 0.6310180989261487,
+      "epoch": 0.6519863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 6.330468433349127e-06,
-      "loss": 0.8682,
+      "learning_rate": 5.707709858822306e-06,
+      "loss": 0.8489,
       "step": 22976
     },
     {
-      "epoch": 0.631045563153992,
+      "epoch": 0.6520147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 6.32964098500678e-06,
-      "loss": 0.8391,
+      "learning_rate": 5.706879769602362e-06,
+      "loss": 0.9412,
       "step": 22977
     },
     {
-      "epoch": 0.6310730273818351,
+      "epoch": 0.6520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 6.328813565705737e-06,
-      "loss": 0.8266,
+      "learning_rate": 5.7060497166466025e-06,
+      "loss": 0.9082,
       "step": 22978
     },
     {
-      "epoch": 0.6311004916096784,
+      "epoch": 0.6520715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 6.327986175452545e-06,
-      "loss": 0.8271,
+      "learning_rate": 5.705219699962028e-06,
+      "loss": 0.8754,
       "step": 22979
     },
     {
-      "epoch": 0.6311279558375217,
+      "epoch": 0.6520998864926221,
       "grad_norm": 0.0,
-      "learning_rate": 6.327158814253746e-06,
-      "loss": 0.8291,
+      "learning_rate": 5.704389719555658e-06,
+      "loss": 0.8662,
       "step": 22980
     },
     {
-      "epoch": 0.6311554200653648,
+      "epoch": 0.6521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 6.326331482115893e-06,
-      "loss": 0.8937,
+      "learning_rate": 5.703559775434504e-06,
+      "loss": 0.8255,
       "step": 22981
     },
     {
-      "epoch": 0.6311828842932081,
+      "epoch": 0.6521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 6.325504179045525e-06,
-      "loss": 0.8436,
+      "learning_rate": 5.702729867605571e-06,
+      "loss": 0.9294,
       "step": 22982
     },
     {
-      "epoch": 0.6312103485210513,
+      "epoch": 0.6521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 6.324676905049193e-06,
-      "loss": 0.9083,
+      "learning_rate": 5.70189999607587e-06,
+      "loss": 0.7138,
       "step": 22983
     },
     {
-      "epoch": 0.6312378127488946,
+      "epoch": 0.6522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 6.323849660133439e-06,
-      "loss": 0.7757,
+      "learning_rate": 5.701070160852419e-06,
+      "loss": 0.8773,
       "step": 22984
     },
     {
-      "epoch": 0.6312652769767378,
+      "epoch": 0.6522417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 6.3230224443048115e-06,
-      "loss": 0.8911,
+      "learning_rate": 5.700240361942216e-06,
+      "loss": 0.9182,
       "step": 22985
     },
     {
-      "epoch": 0.631292741204581,
+      "epoch": 0.6522701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 6.322195257569855e-06,
-      "loss": 0.9751,
+      "learning_rate": 5.699410599352278e-06,
+      "loss": 0.8302,
       "step": 22986
     },
     {
-      "epoch": 0.6313202054324243,
+      "epoch": 0.6522985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 6.321368099935116e-06,
-      "loss": 0.923,
+      "learning_rate": 5.698580873089614e-06,
+      "loss": 0.8047,
       "step": 22987
     },
     {
-      "epoch": 0.6313476696602675,
+      "epoch": 0.6523269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 6.320540971407136e-06,
-      "loss": 0.8623,
+      "learning_rate": 5.697751183161227e-06,
+      "loss": 0.8532,
       "step": 22988
     },
     {
-      "epoch": 0.6313751338881107,
+      "epoch": 0.652355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 6.31971387199246e-06,
-      "loss": 0.947,
+      "learning_rate": 5.69692152957413e-06,
+      "loss": 0.8673,
       "step": 22989
     },
     {
-      "epoch": 0.631402598115954,
+      "epoch": 0.6523836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 6.318886801697634e-06,
-      "loss": 0.8103,
+      "learning_rate": 5.69609191233533e-06,
+      "loss": 0.8304,
       "step": 22990
     },
     {
-      "epoch": 0.6314300623437972,
+      "epoch": 0.6524120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 6.318059760529203e-06,
-      "loss": 0.8102,
+      "learning_rate": 5.695262331451834e-06,
+      "loss": 0.747,
       "step": 22991
     },
     {
-      "epoch": 0.6314575265716404,
+      "epoch": 0.6524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 6.317232748493707e-06,
-      "loss": 0.8566,
+      "learning_rate": 5.694432786930656e-06,
+      "loss": 0.8356,
       "step": 22992
     },
     {
-      "epoch": 0.6314849907994837,
+      "epoch": 0.6524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 6.316405765597696e-06,
-      "loss": 0.8607,
+      "learning_rate": 5.6936032787787945e-06,
+      "loss": 0.8076,
       "step": 22993
     },
     {
-      "epoch": 0.6315124550273269,
+      "epoch": 0.6524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 6.315578811847706e-06,
-      "loss": 0.8882,
+      "learning_rate": 5.692773807003258e-06,
+      "loss": 0.8367,
       "step": 22994
     },
     {
-      "epoch": 0.6315399192551702,
+      "epoch": 0.6525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 6.314751887250285e-06,
-      "loss": 0.7235,
+      "learning_rate": 5.691944371611059e-06,
+      "loss": 0.9833,
       "step": 22995
     },
     {
-      "epoch": 0.6315673834830133,
+      "epoch": 0.6525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 6.313924991811973e-06,
-      "loss": 0.8965,
+      "learning_rate": 5.6911149726091955e-06,
+      "loss": 0.8853,
       "step": 22996
     },
     {
-      "epoch": 0.6315948477108566,
+      "epoch": 0.6525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 6.313098125539314e-06,
-      "loss": 0.7873,
+      "learning_rate": 5.690285610004678e-06,
+      "loss": 0.86,
       "step": 22997
     },
     {
-      "epoch": 0.6316223119386999,
+      "epoch": 0.6526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 6.312271288438853e-06,
-      "loss": 0.912,
+      "learning_rate": 5.6894562838045155e-06,
+      "loss": 0.8846,
       "step": 22998
     },
     {
-      "epoch": 0.631649776166543,
+      "epoch": 0.652639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 6.31144448051713e-06,
-      "loss": 0.919,
+      "learning_rate": 5.688626994015706e-06,
+      "loss": 0.7998,
       "step": 22999
     },
     {
-      "epoch": 0.6316772403943863,
+      "epoch": 0.6526674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 6.310617701780688e-06,
-      "loss": 0.8816,
+      "learning_rate": 5.687797740645258e-06,
+      "loss": 0.9127,
       "step": 23000
     },
     {
-      "epoch": 0.6317047046222295,
+      "epoch": 0.6526958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 6.309790952236066e-06,
-      "loss": 0.8456,
+      "learning_rate": 5.68696852370018e-06,
+      "loss": 0.7982,
       "step": 23001
     },
     {
-      "epoch": 0.6317321688500728,
+      "epoch": 0.6527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 6.308964231889806e-06,
-      "loss": 0.873,
+      "learning_rate": 5.686139343187468e-06,
+      "loss": 0.8024,
       "step": 23002
     },
     {
-      "epoch": 0.631759633077916,
+      "epoch": 0.6527525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 6.308137540748454e-06,
-      "loss": 0.9753,
+      "learning_rate": 5.685310199114132e-06,
+      "loss": 0.7455,
       "step": 23003
     },
     {
-      "epoch": 0.6317870973057592,
+      "epoch": 0.6527809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 6.307310878818544e-06,
-      "loss": 0.9432,
+      "learning_rate": 5.6844810914871774e-06,
+      "loss": 0.8069,
       "step": 23004
     },
     {
-      "epoch": 0.6318145615336025,
+      "epoch": 0.6528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 6.3064842461066265e-06,
-      "loss": 0.7695,
+      "learning_rate": 5.683652020313599e-06,
+      "loss": 0.9141,
       "step": 23005
     },
     {
-      "epoch": 0.6318420257614458,
+      "epoch": 0.6528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 6.305657642619232e-06,
-      "loss": 0.8744,
+      "learning_rate": 5.682822985600409e-06,
+      "loss": 0.8457,
       "step": 23006
     },
     {
-      "epoch": 0.6318694899892889,
+      "epoch": 0.6528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 6.304831068362907e-06,
-      "loss": 0.9533,
+      "learning_rate": 5.681993987354611e-06,
+      "loss": 0.9523,
       "step": 23007
     },
     {
-      "epoch": 0.6318969542171322,
+      "epoch": 0.652894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 6.304004523344188e-06,
-      "loss": 0.8185,
+      "learning_rate": 5.6811650255831995e-06,
+      "loss": 0.8296,
       "step": 23008
     },
     {
-      "epoch": 0.6319244184449754,
+      "epoch": 0.6529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 6.3031780075696165e-06,
-      "loss": 0.8262,
+      "learning_rate": 5.680336100293182e-06,
+      "loss": 0.7873,
       "step": 23009
     },
     {
-      "epoch": 0.6319518826728187,
+      "epoch": 0.6529511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 6.302351521045735e-06,
-      "loss": 0.88,
+      "learning_rate": 5.679507211491564e-06,
+      "loss": 0.7647,
       "step": 23010
     },
     {
-      "epoch": 0.6319793469006619,
+      "epoch": 0.6529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 6.301525063779081e-06,
-      "loss": 0.825,
+      "learning_rate": 5.678678359185339e-06,
+      "loss": 0.8586,
       "step": 23011
     },
     {
-      "epoch": 0.6320068111285051,
+      "epoch": 0.6530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 6.300698635776195e-06,
-      "loss": 0.7896,
+      "learning_rate": 5.67784954338151e-06,
+      "loss": 0.8284,
       "step": 23012
     },
     {
-      "epoch": 0.6320342753563484,
+      "epoch": 0.6530363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 6.299872237043609e-06,
-      "loss": 0.856,
+      "learning_rate": 5.677020764087086e-06,
+      "loss": 0.8146,
       "step": 23013
     },
     {
-      "epoch": 0.6320617395841915,
+      "epoch": 0.6530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 6.2990458675878695e-06,
-      "loss": 0.8666,
+      "learning_rate": 5.676192021309058e-06,
+      "loss": 0.9294,
       "step": 23014
     },
     {
-      "epoch": 0.6320892038120348,
+      "epoch": 0.6530930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 6.298219527415513e-06,
-      "loss": 0.9009,
+      "learning_rate": 5.675363315054434e-06,
+      "loss": 0.811,
       "step": 23015
     },
     {
-      "epoch": 0.6321166680398781,
+      "epoch": 0.6531214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 6.297393216533077e-06,
-      "loss": 0.8632,
+      "learning_rate": 5.674534645330207e-06,
+      "loss": 0.8038,
       "step": 23016
     },
     {
-      "epoch": 0.6321441322677213,
+      "epoch": 0.653149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 6.296566934947104e-06,
-      "loss": 0.9893,
+      "learning_rate": 5.67370601214338e-06,
+      "loss": 0.7924,
       "step": 23017
     },
     {
-      "epoch": 0.6321715964955645,
+      "epoch": 0.653178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 6.295740682664122e-06,
-      "loss": 0.9399,
+      "learning_rate": 5.672877415500957e-06,
+      "loss": 0.7789,
       "step": 23018
     },
     {
-      "epoch": 0.6321990607234078,
+      "epoch": 0.6532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 6.294914459690679e-06,
-      "loss": 0.8711,
+      "learning_rate": 5.672048855409928e-06,
+      "loss": 0.7681,
       "step": 23019
     },
     {
-      "epoch": 0.632226524951251,
+      "epoch": 0.6532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 6.2940882660333045e-06,
-      "loss": 0.8117,
+      "learning_rate": 5.671220331877298e-06,
+      "loss": 0.7809,
       "step": 23020
     },
     {
-      "epoch": 0.6322539891790943,
+      "epoch": 0.6532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 6.29326210169854e-06,
-      "loss": 0.7902,
+      "learning_rate": 5.670391844910064e-06,
+      "loss": 0.8832,
       "step": 23021
     },
     {
-      "epoch": 0.6322814534069374,
+      "epoch": 0.6532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 6.292435966692922e-06,
-      "loss": 0.8371,
+      "learning_rate": 5.669563394515223e-06,
+      "loss": 0.8627,
       "step": 23022
     },
     {
-      "epoch": 0.6323089176347807,
+      "epoch": 0.6533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 6.2916098610229894e-06,
-      "loss": 0.7741,
+      "learning_rate": 5.668734980699777e-06,
+      "loss": 0.8144,
       "step": 23023
     },
     {
-      "epoch": 0.632336381862624,
+      "epoch": 0.6533484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 6.290783784695275e-06,
-      "loss": 0.8639,
+      "learning_rate": 5.667906603470724e-06,
+      "loss": 0.7915,
       "step": 23024
     },
     {
-      "epoch": 0.6323638460904671,
+      "epoch": 0.6533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 6.289957737716313e-06,
-      "loss": 0.8964,
+      "learning_rate": 5.6670782628350555e-06,
+      "loss": 0.8617,
       "step": 23025
     },
     {
-      "epoch": 0.6323913103183104,
+      "epoch": 0.653405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 6.289131720092644e-06,
-      "loss": 0.8571,
+      "learning_rate": 5.666249958799772e-06,
+      "loss": 0.8131,
       "step": 23026
     },
     {
-      "epoch": 0.6324187745461536,
+      "epoch": 0.6534335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 6.288305731830804e-06,
-      "loss": 0.8632,
+      "learning_rate": 5.665421691371873e-06,
+      "loss": 0.8469,
       "step": 23027
     },
     {
-      "epoch": 0.6324462387739969,
+      "epoch": 0.6534619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 6.2874797729373235e-06,
-      "loss": 0.9415,
+      "learning_rate": 5.6645934605583475e-06,
+      "loss": 0.7798,
       "step": 23028
     },
     {
-      "epoch": 0.6324737030018401,
+      "epoch": 0.6534903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 6.286653843418747e-06,
-      "loss": 0.8006,
+      "learning_rate": 5.663765266366197e-06,
+      "loss": 0.9162,
       "step": 23029
     },
     {
-      "epoch": 0.6325011672296833,
+      "epoch": 0.6535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 6.2858279432816e-06,
-      "loss": 0.7972,
+      "learning_rate": 5.66293710880242e-06,
+      "loss": 0.8223,
       "step": 23030
     },
     {
-      "epoch": 0.6325286314575266,
+      "epoch": 0.6535471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 6.28500207253242e-06,
-      "loss": 0.7591,
+      "learning_rate": 5.662108987874004e-06,
+      "loss": 0.758,
       "step": 23031
     },
     {
-      "epoch": 0.6325560956853699,
+      "epoch": 0.6535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 6.284176231177743e-06,
-      "loss": 0.7628,
+      "learning_rate": 5.661280903587949e-06,
+      "loss": 0.7556,
       "step": 23032
     },
     {
-      "epoch": 0.632583559913213,
+      "epoch": 0.6536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 6.283350419224103e-06,
-      "loss": 0.7914,
+      "learning_rate": 5.660452855951253e-06,
+      "loss": 0.8076,
       "step": 23033
     },
     {
-      "epoch": 0.6326110241410563,
+      "epoch": 0.6536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 6.282524636678038e-06,
-      "loss": 0.847,
+      "learning_rate": 5.659624844970901e-06,
+      "loss": 0.8921,
       "step": 23034
     },
     {
-      "epoch": 0.6326384883688995,
+      "epoch": 0.6536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 6.2816988835460735e-06,
-      "loss": 0.8291,
+      "learning_rate": 5.658796870653896e-06,
+      "loss": 0.8382,
       "step": 23035
     },
     {
-      "epoch": 0.6326659525967427,
+      "epoch": 0.653688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 6.28087315983475e-06,
-      "loss": 0.7824,
+      "learning_rate": 5.6579689330072275e-06,
+      "loss": 0.7361,
       "step": 23036
     },
     {
-      "epoch": 0.632693416824586,
+      "epoch": 0.6537173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 6.280047465550597e-06,
-      "loss": 0.8791,
+      "learning_rate": 5.65714103203789e-06,
+      "loss": 0.7688,
       "step": 23037
     },
     {
-      "epoch": 0.6327208810524292,
+      "epoch": 0.6537457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 6.279221800700149e-06,
-      "loss": 0.8102,
+      "learning_rate": 5.656313167752882e-06,
+      "loss": 0.7251,
       "step": 23038
     },
     {
-      "epoch": 0.6327483452802725,
+      "epoch": 0.6537741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 6.278396165289941e-06,
-      "loss": 0.9076,
+      "learning_rate": 5.655485340159188e-06,
+      "loss": 0.7868,
       "step": 23039
     },
     {
-      "epoch": 0.6327758095081156,
+      "epoch": 0.6538024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 6.277570559326508e-06,
-      "loss": 0.8765,
+      "learning_rate": 5.6546575492638045e-06,
+      "loss": 0.8444,
       "step": 23040
     },
     {
-      "epoch": 0.6328032737359589,
+      "epoch": 0.6538308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 6.276744982816374e-06,
-      "loss": 0.8456,
+      "learning_rate": 5.653829795073729e-06,
+      "loss": 0.9377,
       "step": 23041
     },
     {
-      "epoch": 0.6328307379638022,
+      "epoch": 0.6538592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 6.275919435766075e-06,
-      "loss": 0.8261,
+      "learning_rate": 5.653002077595944e-06,
+      "loss": 0.9454,
       "step": 23042
     },
     {
-      "epoch": 0.6328582021916453,
+      "epoch": 0.6538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 6.275093918182147e-06,
-      "loss": 0.9399,
+      "learning_rate": 5.652174396837444e-06,
+      "loss": 0.8172,
       "step": 23043
     },
     {
-      "epoch": 0.6328856664194886,
+      "epoch": 0.6539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 6.2742684300711155e-06,
-      "loss": 0.8964,
+      "learning_rate": 5.651346752805229e-06,
+      "loss": 0.8733,
       "step": 23044
     },
     {
-      "epoch": 0.6329131306473319,
+      "epoch": 0.65394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 6.273442971439516e-06,
-      "loss": 0.9623,
+      "learning_rate": 5.650519145506278e-06,
+      "loss": 0.9438,
       "step": 23045
     },
     {
-      "epoch": 0.6329405948751751,
+      "epoch": 0.6539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 6.272617542293884e-06,
-      "loss": 0.8665,
+      "learning_rate": 5.649691574947588e-06,
+      "loss": 0.8466,
       "step": 23046
     },
     {
-      "epoch": 0.6329680591030183,
+      "epoch": 0.6540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 6.27179214264074e-06,
-      "loss": 0.8324,
+      "learning_rate": 5.648864041136153e-06,
+      "loss": 0.744,
       "step": 23047
     },
     {
-      "epoch": 0.6329955233308615,
+      "epoch": 0.6540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 6.270966772486624e-06,
-      "loss": 0.7612,
+      "learning_rate": 5.648036544078954e-06,
+      "loss": 0.9188,
       "step": 23048
     },
     {
-      "epoch": 0.6330229875587048,
+      "epoch": 0.6540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 6.2701414318380615e-06,
-      "loss": 0.8135,
+      "learning_rate": 5.647209083782986e-06,
+      "loss": 0.8593,
       "step": 23049
     },
     {
-      "epoch": 0.6330504517865481,
+      "epoch": 0.6540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 6.269316120701585e-06,
-      "loss": 0.8611,
+      "learning_rate": 5.646381660255242e-06,
+      "loss": 0.8424,
       "step": 23050
     },
     {
-      "epoch": 0.6330779160143912,
+      "epoch": 0.6541146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 6.268490839083726e-06,
-      "loss": 0.917,
+      "learning_rate": 5.6455542735027024e-06,
+      "loss": 0.784,
       "step": 23051
     },
     {
-      "epoch": 0.6331053802422345,
+      "epoch": 0.6541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 6.2676655869910155e-06,
-      "loss": 0.8274,
+      "learning_rate": 5.644726923532368e-06,
+      "loss": 0.9277,
       "step": 23052
     },
     {
-      "epoch": 0.6331328444700777,
+      "epoch": 0.6541713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 6.266840364429976e-06,
-      "loss": 0.7944,
+      "learning_rate": 5.643899610351217e-06,
+      "loss": 0.8657,
       "step": 23053
     },
     {
-      "epoch": 0.633160308697921,
+      "epoch": 0.654199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 6.266015171407142e-06,
-      "loss": 0.9206,
+      "learning_rate": 5.6430723339662415e-06,
+      "loss": 0.8657,
       "step": 23054
     },
     {
-      "epoch": 0.6331877729257642,
+      "epoch": 0.654228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 6.265190007929044e-06,
-      "loss": 0.8552,
+      "learning_rate": 5.642245094384435e-06,
+      "loss": 0.7157,
       "step": 23055
     },
     {
-      "epoch": 0.6332152371536074,
+      "epoch": 0.6542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 6.264364874002208e-06,
-      "loss": 0.7933,
+      "learning_rate": 5.641417891612775e-06,
+      "loss": 0.8413,
       "step": 23056
     },
     {
-      "epoch": 0.6332427013814507,
+      "epoch": 0.6542849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 6.263539769633163e-06,
-      "loss": 0.7448,
+      "learning_rate": 5.640590725658255e-06,
+      "loss": 0.9262,
       "step": 23057
     },
     {
-      "epoch": 0.633270165609294,
+      "epoch": 0.6543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 6.262714694828444e-06,
-      "loss": 0.8917,
+      "learning_rate": 5.639763596527863e-06,
+      "loss": 0.8102,
       "step": 23058
     },
     {
-      "epoch": 0.6332976298371371,
+      "epoch": 0.6543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 6.261889649594569e-06,
-      "loss": 0.8218,
+      "learning_rate": 5.638936504228583e-06,
+      "loss": 0.8115,
       "step": 23059
     },
     {
-      "epoch": 0.6333250940649804,
+      "epoch": 0.6543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 6.261064633938075e-06,
-      "loss": 0.9091,
+      "learning_rate": 5.638109448767399e-06,
+      "loss": 0.9497,
       "step": 23060
     },
     {
-      "epoch": 0.6333525582928236,
+      "epoch": 0.6543984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 6.2602396478654815e-06,
-      "loss": 0.8733,
+      "learning_rate": 5.637282430151307e-06,
+      "loss": 0.9134,
       "step": 23061
     },
     {
-      "epoch": 0.6333800225206668,
+      "epoch": 0.6544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 6.259414691383321e-06,
-      "loss": 0.7749,
+      "learning_rate": 5.636455448387282e-06,
+      "loss": 0.9431,
       "step": 23062
     },
     {
-      "epoch": 0.6334074867485101,
+      "epoch": 0.6544551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 6.258589764498124e-06,
-      "loss": 0.8714,
+      "learning_rate": 5.635628503482313e-06,
+      "loss": 0.8102,
       "step": 23063
     },
     {
-      "epoch": 0.6334349509763533,
+      "epoch": 0.6544835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 6.2577648672164146e-06,
-      "loss": 0.8236,
+      "learning_rate": 5.634801595443391e-06,
+      "loss": 0.9539,
       "step": 23064
     },
     {
-      "epoch": 0.6334624152041965,
+      "epoch": 0.6545119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 6.256939999544715e-06,
-      "loss": 0.8675,
+      "learning_rate": 5.633974724277493e-06,
+      "loss": 0.772,
       "step": 23065
     },
     {
-      "epoch": 0.6334898794320397,
+      "epoch": 0.6545402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 6.2561151614895575e-06,
-      "loss": 0.871,
+      "learning_rate": 5.633147889991606e-06,
+      "loss": 0.8847,
       "step": 23066
     },
     {
-      "epoch": 0.633517343659883,
+      "epoch": 0.6545686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 6.255290353057468e-06,
-      "loss": 0.9366,
+      "learning_rate": 5.632321092592714e-06,
+      "loss": 0.9161,
       "step": 23067
     },
     {
-      "epoch": 0.6335448078877263,
+      "epoch": 0.6545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 6.2544655742549695e-06,
-      "loss": 0.8142,
+      "learning_rate": 5.631494332087804e-06,
+      "loss": 0.8668,
       "step": 23068
     },
     {
-      "epoch": 0.6335722721155694,
+      "epoch": 0.6546254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 6.253640825088592e-06,
-      "loss": 0.8075,
+      "learning_rate": 5.63066760848386e-06,
+      "loss": 0.8462,
       "step": 23069
     },
     {
-      "epoch": 0.6335997363434127,
+      "epoch": 0.6546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 6.252816105564862e-06,
-      "loss": 0.8126,
+      "learning_rate": 5.6298409217878605e-06,
+      "loss": 0.8665,
       "step": 23070
     },
     {
-      "epoch": 0.633627200571256,
+      "epoch": 0.6546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 6.251991415690298e-06,
-      "loss": 0.9213,
+      "learning_rate": 5.62901427200679e-06,
+      "loss": 0.8321,
       "step": 23071
     },
     {
-      "epoch": 0.6336546647990992,
+      "epoch": 0.6547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.2511667554714325e-06,
-      "loss": 0.8384,
+      "learning_rate": 5.628187659147637e-06,
+      "loss": 0.8833,
       "step": 23072
     },
     {
-      "epoch": 0.6336821290269424,
+      "epoch": 0.654738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 6.250342124914786e-06,
-      "loss": 0.8163,
+      "learning_rate": 5.627361083217375e-06,
+      "loss": 0.7435,
       "step": 23073
     },
     {
-      "epoch": 0.6337095932547856,
+      "epoch": 0.6547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 6.2495175240268845e-06,
-      "loss": 0.8303,
+      "learning_rate": 5.626534544222993e-06,
+      "loss": 0.8707,
       "step": 23074
     },
     {
-      "epoch": 0.6337370574826289,
+      "epoch": 0.6547956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 6.248692952814257e-06,
-      "loss": 0.7664,
+      "learning_rate": 5.625708042171472e-06,
+      "loss": 0.9257,
       "step": 23075
     },
     {
-      "epoch": 0.6337645217104722,
+      "epoch": 0.6548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 6.247868411283425e-06,
-      "loss": 0.9125,
+      "learning_rate": 5.624881577069788e-06,
+      "loss": 0.8516,
       "step": 23076
     },
     {
-      "epoch": 0.6337919859383153,
+      "epoch": 0.6548524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 6.247043899440906e-06,
-      "loss": 0.8672,
+      "learning_rate": 5.624055148924925e-06,
+      "loss": 0.9374,
       "step": 23077
     },
     {
-      "epoch": 0.6338194501661586,
+      "epoch": 0.6548808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 6.246219417293231e-06,
-      "loss": 0.7972,
+      "learning_rate": 5.62322875774387e-06,
+      "loss": 0.8942,
       "step": 23078
     },
     {
-      "epoch": 0.6338469143940018,
+      "epoch": 0.6549091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 6.245394964846922e-06,
-      "loss": 0.8633,
+      "learning_rate": 5.622402403533594e-06,
+      "loss": 0.9496,
       "step": 23079
     },
     {
-      "epoch": 0.633874378621845,
+      "epoch": 0.6549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 6.244570542108504e-06,
-      "loss": 0.9417,
+      "learning_rate": 5.6215760863010825e-06,
+      "loss": 0.8294,
       "step": 23080
     },
     {
-      "epoch": 0.6339018428496883,
+      "epoch": 0.6549659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 6.243746149084495e-06,
-      "loss": 0.8313,
+      "learning_rate": 5.620749806053318e-06,
+      "loss": 0.7496,
       "step": 23081
     },
     {
-      "epoch": 0.6339293070775315,
+      "epoch": 0.6549943246311011,
       "grad_norm": 0.0,
-      "learning_rate": 6.242921785781428e-06,
-      "loss": 0.8038,
+      "learning_rate": 5.61992356279727e-06,
+      "loss": 0.7905,
       "step": 23082
     },
     {
-      "epoch": 0.6339567713053748,
+      "epoch": 0.6550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 6.242097452205817e-06,
-      "loss": 0.888,
+      "learning_rate": 5.619097356539931e-06,
+      "loss": 0.956,
       "step": 23083
     },
     {
-      "epoch": 0.633984235533218,
+      "epoch": 0.6550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 6.241273148364183e-06,
-      "loss": 0.8433,
+      "learning_rate": 5.618271187288269e-06,
+      "loss": 0.9422,
       "step": 23084
     },
     {
-      "epoch": 0.6340116997610612,
+      "epoch": 0.6550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 6.240448874263053e-06,
-      "loss": 0.9313,
+      "learning_rate": 5.617445055049269e-06,
+      "loss": 0.7958,
       "step": 23085
     },
     {
-      "epoch": 0.6340391639889045,
+      "epoch": 0.6551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 6.239624629908947e-06,
-      "loss": 0.8116,
+      "learning_rate": 5.616618959829911e-06,
+      "loss": 0.8466,
       "step": 23086
     },
     {
-      "epoch": 0.6340666282167476,
+      "epoch": 0.6551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 6.23880041530839e-06,
-      "loss": 0.8568,
+      "learning_rate": 5.6157929016371645e-06,
+      "loss": 0.7709,
       "step": 23087
     },
     {
-      "epoch": 0.6340940924445909,
+      "epoch": 0.6551645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 6.237976230467903e-06,
-      "loss": 0.8301,
+      "learning_rate": 5.614966880478014e-06,
+      "loss": 0.7476,
       "step": 23088
     },
     {
-      "epoch": 0.6341215566724342,
+      "epoch": 0.6551929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 6.237152075394002e-06,
-      "loss": 0.9017,
+      "learning_rate": 5.614140896359439e-06,
+      "loss": 0.8837,
       "step": 23089
     },
     {
-      "epoch": 0.6341490209002774,
+      "epoch": 0.6552213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 6.236327950093211e-06,
-      "loss": 0.8196,
+      "learning_rate": 5.613314949288409e-06,
+      "loss": 0.8557,
       "step": 23090
     },
     {
-      "epoch": 0.6341764851281206,
+      "epoch": 0.655249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 6.2355038545720535e-06,
-      "loss": 0.8222,
+      "learning_rate": 5.612489039271904e-06,
+      "loss": 0.9433,
       "step": 23091
     },
     {
-      "epoch": 0.6342039493559638,
+      "epoch": 0.6552780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 6.234679788837046e-06,
-      "loss": 0.8013,
+      "learning_rate": 5.611663166316908e-06,
+      "loss": 0.8189,
       "step": 23092
     },
     {
-      "epoch": 0.6342314135838071,
+      "epoch": 0.6553064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 6.2338557528947134e-06,
-      "loss": 0.8955,
+      "learning_rate": 5.610837330430383e-06,
+      "loss": 0.7764,
       "step": 23093
     },
     {
-      "epoch": 0.6342588778116504,
+      "epoch": 0.6553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 6.23303174675157e-06,
-      "loss": 0.9592,
+      "learning_rate": 5.610011531619315e-06,
+      "loss": 0.7304,
       "step": 23094
     },
     {
-      "epoch": 0.6342863420394935,
+      "epoch": 0.6553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 6.232207770414139e-06,
-      "loss": 0.7943,
+      "learning_rate": 5.60918576989068e-06,
+      "loss": 0.864,
       "step": 23095
     },
     {
-      "epoch": 0.6343138062673368,
+      "epoch": 0.6553916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 6.231383823888939e-06,
-      "loss": 0.8852,
+      "learning_rate": 5.608360045251445e-06,
+      "loss": 0.9005,
       "step": 23096
     },
     {
-      "epoch": 0.6343412704951801,
+      "epoch": 0.6554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 6.2305599071824895e-06,
-      "loss": 0.8655,
+      "learning_rate": 5.607534357708591e-06,
+      "loss": 0.8665,
       "step": 23097
     },
     {
-      "epoch": 0.6343687347230232,
+      "epoch": 0.6554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 6.2297360203013104e-06,
-      "loss": 0.8025,
+      "learning_rate": 5.606708707269093e-06,
+      "loss": 0.7354,
       "step": 23098
     },
     {
-      "epoch": 0.6343961989508665,
+      "epoch": 0.6554767309875141,
       "grad_norm": 0.0,
-      "learning_rate": 6.228912163251924e-06,
-      "loss": 0.8236,
+      "learning_rate": 5.605883093939922e-06,
+      "loss": 0.7336,
       "step": 23099
     },
     {
-      "epoch": 0.6344236631787097,
+      "epoch": 0.6555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 6.228088336040844e-06,
-      "loss": 0.8417,
+      "learning_rate": 5.6050575177280586e-06,
+      "loss": 0.812,
       "step": 23100
     },
     {
-      "epoch": 0.634451127406553,
+      "epoch": 0.655533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 6.227264538674587e-06,
-      "loss": 0.7028,
+      "learning_rate": 5.604231978640469e-06,
+      "loss": 0.8748,
       "step": 23101
     },
     {
-      "epoch": 0.6344785916343962,
+      "epoch": 0.6555618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 6.226440771159677e-06,
-      "loss": 0.8033,
+      "learning_rate": 5.603406476684129e-06,
+      "loss": 0.7516,
       "step": 23102
     },
     {
-      "epoch": 0.6345060558622394,
+      "epoch": 0.6555902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 6.225617033502629e-06,
-      "loss": 0.8257,
+      "learning_rate": 5.602581011866015e-06,
+      "loss": 0.82,
       "step": 23103
     },
     {
-      "epoch": 0.6345335200900827,
+      "epoch": 0.6556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 6.224793325709959e-06,
-      "loss": 0.8954,
+      "learning_rate": 5.601755584193094e-06,
+      "loss": 0.9094,
       "step": 23104
     },
     {
-      "epoch": 0.6345609843179258,
+      "epoch": 0.6556469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 6.223969647788191e-06,
-      "loss": 0.8475,
+      "learning_rate": 5.6009301936723415e-06,
+      "loss": 0.7922,
       "step": 23105
     },
     {
-      "epoch": 0.6345884485457691,
+      "epoch": 0.6556753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 6.223145999743834e-06,
-      "loss": 0.83,
+      "learning_rate": 5.600104840310733e-06,
+      "loss": 0.8448,
       "step": 23106
     },
     {
-      "epoch": 0.6346159127736124,
+      "epoch": 0.6557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 6.222322381583411e-06,
-      "loss": 0.9325,
+      "learning_rate": 5.5992795241152335e-06,
+      "loss": 0.8294,
       "step": 23107
     },
     {
-      "epoch": 0.6346433770014556,
+      "epoch": 0.6557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 6.221498793313434e-06,
-      "loss": 0.8918,
+      "learning_rate": 5.598454245092816e-06,
+      "loss": 0.8749,
       "step": 23108
     },
     {
-      "epoch": 0.6346708412292988,
+      "epoch": 0.6557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 6.220675234940423e-06,
-      "loss": 0.9079,
+      "learning_rate": 5.597629003250457e-06,
+      "loss": 0.8482,
       "step": 23109
     },
     {
-      "epoch": 0.6346983054571421,
+      "epoch": 0.655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 6.219851706470894e-06,
-      "loss": 0.863,
+      "learning_rate": 5.596803798595123e-06,
+      "loss": 0.7824,
       "step": 23110
     },
     {
-      "epoch": 0.6347257696849853,
+      "epoch": 0.6558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 6.219028207911366e-06,
-      "loss": 0.8842,
+      "learning_rate": 5.5959786311337825e-06,
+      "loss": 0.9406,
       "step": 23111
     },
     {
-      "epoch": 0.6347532339128286,
+      "epoch": 0.6558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 6.218204739268349e-06,
-      "loss": 0.7714,
+      "learning_rate": 5.5951535008734135e-06,
+      "loss": 0.6753,
       "step": 23112
     },
     {
-      "epoch": 0.6347806981406717,
+      "epoch": 0.6558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 6.217381300548359e-06,
-      "loss": 0.9036,
+      "learning_rate": 5.5943284078209725e-06,
+      "loss": 0.8263,
       "step": 23113
     },
     {
-      "epoch": 0.634808162368515,
+      "epoch": 0.6559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 6.216557891757914e-06,
-      "loss": 0.9039,
+      "learning_rate": 5.5935033519834415e-06,
+      "loss": 0.6933,
       "step": 23114
     },
     {
-      "epoch": 0.6348356265963583,
+      "epoch": 0.6559307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 6.21573451290353e-06,
-      "loss": 0.8522,
+      "learning_rate": 5.592678333367789e-06,
+      "loss": 0.8845,
       "step": 23115
     },
     {
-      "epoch": 0.6348630908242014,
+      "epoch": 0.6559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 6.214911163991719e-06,
-      "loss": 0.7821,
+      "learning_rate": 5.591853351980978e-06,
+      "loss": 0.884,
       "step": 23116
     },
     {
-      "epoch": 0.6348905550520447,
+      "epoch": 0.6559875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 6.214087845029e-06,
-      "loss": 0.8547,
+      "learning_rate": 5.591028407829979e-06,
+      "loss": 0.7694,
       "step": 23117
     },
     {
-      "epoch": 0.6349180192798879,
+      "epoch": 0.6560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 6.2132645560218814e-06,
-      "loss": 0.8743,
+      "learning_rate": 5.590203500921767e-06,
+      "loss": 0.8447,
       "step": 23118
     },
     {
-      "epoch": 0.6349454835077312,
+      "epoch": 0.656044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 6.212441296976883e-06,
-      "loss": 0.807,
+      "learning_rate": 5.589378631263298e-06,
+      "loss": 0.8097,
       "step": 23119
     },
     {
-      "epoch": 0.6349729477355744,
+      "epoch": 0.656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 6.211618067900513e-06,
-      "loss": 0.9101,
+      "learning_rate": 5.5885537988615465e-06,
+      "loss": 0.7938,
       "step": 23120
     },
     {
-      "epoch": 0.6350004119634176,
+      "epoch": 0.6561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 6.210794868799288e-06,
-      "loss": 0.9363,
+      "learning_rate": 5.587729003723484e-06,
+      "loss": 0.8471,
       "step": 23121
     },
     {
-      "epoch": 0.6350278761912609,
+      "epoch": 0.6561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 6.209971699679725e-06,
-      "loss": 0.8785,
+      "learning_rate": 5.586904245856068e-06,
+      "loss": 0.8065,
       "step": 23122
     },
     {
-      "epoch": 0.6350553404191042,
+      "epoch": 0.6561577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 6.209148560548335e-06,
-      "loss": 0.8276,
+      "learning_rate": 5.586079525266276e-06,
+      "loss": 1.0267,
       "step": 23123
     },
     {
-      "epoch": 0.6350828046469473,
+      "epoch": 0.6561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 6.208325451411628e-06,
-      "loss": 0.8244,
+      "learning_rate": 5.585254841961063e-06,
+      "loss": 0.8212,
       "step": 23124
     },
     {
-      "epoch": 0.6351102688747906,
+      "epoch": 0.6562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 6.2075023722761175e-06,
-      "loss": 0.7773,
+      "learning_rate": 5.584430195947402e-06,
+      "loss": 0.7854,
       "step": 23125
     },
     {
-      "epoch": 0.6351377331026338,
+      "epoch": 0.6562429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 6.206679323148317e-06,
-      "loss": 0.9285,
+      "learning_rate": 5.583605587232262e-06,
+      "loss": 0.8302,
       "step": 23126
     },
     {
-      "epoch": 0.635165197330477,
+      "epoch": 0.6562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 6.205856304034741e-06,
-      "loss": 0.9151,
+      "learning_rate": 5.582781015822598e-06,
+      "loss": 0.8235,
       "step": 23127
     },
     {
-      "epoch": 0.6351926615583203,
+      "epoch": 0.6562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 6.205033314941898e-06,
-      "loss": 0.9123,
+      "learning_rate": 5.581956481725379e-06,
+      "loss": 0.8782,
       "step": 23128
     },
     {
-      "epoch": 0.6352201257861635,
+      "epoch": 0.6563280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 6.204210355876306e-06,
-      "loss": 0.8413,
+      "learning_rate": 5.58113198494758e-06,
+      "loss": 0.8962,
       "step": 23129
     },
     {
-      "epoch": 0.6352475900140068,
+      "epoch": 0.6563564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 6.203387426844467e-06,
-      "loss": 0.7841,
+      "learning_rate": 5.580307525496154e-06,
+      "loss": 0.7436,
       "step": 23130
     },
     {
-      "epoch": 0.63527505424185,
+      "epoch": 0.6563847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 6.2025645278529e-06,
-      "loss": 0.9087,
+      "learning_rate": 5.579483103378067e-06,
+      "loss": 0.901,
       "step": 23131
     },
     {
-      "epoch": 0.6353025184696932,
+      "epoch": 0.6564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 6.201741658908111e-06,
-      "loss": 0.8419,
+      "learning_rate": 5.578658718600291e-06,
+      "loss": 0.8361,
       "step": 23132
     },
     {
-      "epoch": 0.6353299826975365,
+      "epoch": 0.6564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 6.200918820016614e-06,
-      "loss": 0.749,
+      "learning_rate": 5.577834371169779e-06,
+      "loss": 0.837,
       "step": 23133
     },
     {
-      "epoch": 0.6353574469253797,
+      "epoch": 0.6564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 6.200096011184921e-06,
-      "loss": 0.8579,
+      "learning_rate": 5.5770100610935e-06,
+      "loss": 0.9694,
       "step": 23134
     },
     {
-      "epoch": 0.6353849111532229,
+      "epoch": 0.6564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 6.199273232419539e-06,
-      "loss": 0.9475,
+      "learning_rate": 5.576185788378418e-06,
+      "loss": 0.7487,
       "step": 23135
     },
     {
-      "epoch": 0.6354123753810662,
+      "epoch": 0.6565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 6.198450483726981e-06,
-      "loss": 0.9651,
+      "learning_rate": 5.57536155303149e-06,
+      "loss": 0.8847,
       "step": 23136
     },
     {
-      "epoch": 0.6354398396089094,
+      "epoch": 0.6565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 6.197627765113752e-06,
-      "loss": 0.8929,
+      "learning_rate": 5.574537355059682e-06,
+      "loss": 0.7939,
       "step": 23137
     },
     {
-      "epoch": 0.6354673038367527,
+      "epoch": 0.656583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 6.196805076586365e-06,
-      "loss": 0.8723,
+      "learning_rate": 5.573713194469961e-06,
+      "loss": 0.7648,
       "step": 23138
     },
     {
-      "epoch": 0.6354947680645958,
+      "epoch": 0.6566118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 6.195982418151332e-06,
-      "loss": 0.8488,
+      "learning_rate": 5.572889071269279e-06,
+      "loss": 0.8782,
       "step": 23139
     },
     {
-      "epoch": 0.6355222322924391,
+      "epoch": 0.6566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 6.195159789815158e-06,
-      "loss": 0.8845,
+      "learning_rate": 5.5720649854646e-06,
+      "loss": 0.8937,
       "step": 23140
     },
     {
-      "epoch": 0.6355496965202824,
+      "epoch": 0.6566685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 6.194337191584358e-06,
-      "loss": 0.8979,
+      "learning_rate": 5.571240937062893e-06,
+      "loss": 0.8658,
       "step": 23141
     },
     {
-      "epoch": 0.6355771607481255,
+      "epoch": 0.6566969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 6.1935146234654305e-06,
-      "loss": 0.8331,
+      "learning_rate": 5.570416926071109e-06,
+      "loss": 0.9265,
       "step": 23142
     },
     {
-      "epoch": 0.6356046249759688,
+      "epoch": 0.6567253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 6.192692085464893e-06,
-      "loss": 0.886,
+      "learning_rate": 5.569592952496212e-06,
+      "loss": 0.7489,
       "step": 23143
     },
     {
-      "epoch": 0.6356320892038121,
+      "epoch": 0.6567536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 6.191869577589249e-06,
-      "loss": 0.8682,
+      "learning_rate": 5.568769016345163e-06,
+      "loss": 0.8845,
       "step": 23144
     },
     {
-      "epoch": 0.6356595534316553,
+      "epoch": 0.6567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 6.191047099845006e-06,
-      "loss": 0.7748,
+      "learning_rate": 5.56794511762492e-06,
+      "loss": 0.7537,
       "step": 23145
     },
     {
-      "epoch": 0.6356870176594985,
+      "epoch": 0.6568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 6.190224652238679e-06,
-      "loss": 0.8451,
+      "learning_rate": 5.567121256342449e-06,
+      "loss": 0.7813,
       "step": 23146
     },
     {
-      "epoch": 0.6357144818873417,
+      "epoch": 0.6568388195232691,
       "grad_norm": 0.0,
-      "learning_rate": 6.189402234776768e-06,
-      "loss": 0.8094,
+      "learning_rate": 5.566297432504699e-06,
+      "loss": 0.8355,
       "step": 23147
     },
     {
-      "epoch": 0.635741946115185,
+      "epoch": 0.6568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 6.188579847465783e-06,
-      "loss": 0.9606,
+      "learning_rate": 5.565473646118635e-06,
+      "loss": 0.8256,
       "step": 23148
     },
     {
-      "epoch": 0.6357694103430283,
+      "epoch": 0.6568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 6.187757490312229e-06,
-      "loss": 0.9888,
+      "learning_rate": 5.564649897191218e-06,
+      "loss": 0.7393,
       "step": 23149
     },
     {
-      "epoch": 0.6357968745708714,
+      "epoch": 0.6569239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 6.1869351633226134e-06,
-      "loss": 0.8471,
+      "learning_rate": 5.563826185729398e-06,
+      "loss": 0.6977,
       "step": 23150
     },
     {
-      "epoch": 0.6358243387987147,
+      "epoch": 0.6569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 6.186112866503447e-06,
-      "loss": 0.819,
+      "learning_rate": 5.5630025117401385e-06,
+      "loss": 0.9395,
       "step": 23151
     },
     {
-      "epoch": 0.6358518030265579,
+      "epoch": 0.6569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 6.185290599861235e-06,
-      "loss": 0.894,
+      "learning_rate": 5.562178875230401e-06,
+      "loss": 0.8932,
       "step": 23152
     },
     {
-      "epoch": 0.6358792672544011,
+      "epoch": 0.6570090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 6.184468363402477e-06,
-      "loss": 0.9192,
+      "learning_rate": 5.561355276207132e-06,
+      "loss": 0.9515,
       "step": 23153
     },
     {
-      "epoch": 0.6359067314822444,
+      "epoch": 0.6570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 6.183646157133684e-06,
-      "loss": 0.8048,
+      "learning_rate": 5.560531714677296e-06,
+      "loss": 0.8303,
       "step": 23154
     },
     {
-      "epoch": 0.6359341957100876,
+      "epoch": 0.6570658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 6.1828239810613624e-06,
-      "loss": 0.7825,
+      "learning_rate": 5.559708190647852e-06,
+      "loss": 0.8294,
       "step": 23155
     },
     {
-      "epoch": 0.6359616599379309,
+      "epoch": 0.657094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 6.182001835192013e-06,
-      "loss": 0.8562,
+      "learning_rate": 5.558884704125749e-06,
+      "loss": 0.8998,
       "step": 23156
     },
     {
-      "epoch": 0.6359891241657741,
+      "epoch": 0.657122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 6.181179719532145e-06,
-      "loss": 0.9531,
+      "learning_rate": 5.558061255117946e-06,
+      "loss": 0.8887,
       "step": 23157
     },
     {
-      "epoch": 0.6360165883936173,
+      "epoch": 0.6571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 6.180357634088268e-06,
-      "loss": 0.796,
+      "learning_rate": 5.557237843631402e-06,
+      "loss": 0.845,
       "step": 23158
     },
     {
-      "epoch": 0.6360440526214606,
+      "epoch": 0.6571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 6.179535578866875e-06,
-      "loss": 0.9255,
+      "learning_rate": 5.556414469673064e-06,
+      "loss": 0.876,
       "step": 23159
     },
     {
-      "epoch": 0.6360715168493037,
+      "epoch": 0.6572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 6.178713553874478e-06,
-      "loss": 0.7923,
+      "learning_rate": 5.555591133249898e-06,
+      "loss": 0.8512,
       "step": 23160
     },
     {
-      "epoch": 0.636098981077147,
+      "epoch": 0.6572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 6.177891559117579e-06,
-      "loss": 0.8343,
+      "learning_rate": 5.554767834368851e-06,
+      "loss": 0.789,
       "step": 23161
     },
     {
-      "epoch": 0.6361264453049903,
+      "epoch": 0.6572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 6.177069594602682e-06,
-      "loss": 0.8721,
+      "learning_rate": 5.55394457303688e-06,
+      "loss": 0.8524,
       "step": 23162
     },
     {
-      "epoch": 0.6361539095328335,
+      "epoch": 0.6572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 6.176247660336292e-06,
-      "loss": 0.8335,
+      "learning_rate": 5.553121349260943e-06,
+      "loss": 0.8708,
       "step": 23163
     },
     {
-      "epoch": 0.6361813737606767,
+      "epoch": 0.6573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 6.175425756324915e-06,
-      "loss": 0.9064,
+      "learning_rate": 5.5522981630479865e-06,
+      "loss": 0.895,
       "step": 23164
     },
     {
-      "epoch": 0.6362088379885199,
+      "epoch": 0.6573496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 6.174603882575048e-06,
-      "loss": 0.8394,
+      "learning_rate": 5.551475014404966e-06,
+      "loss": 0.8634,
       "step": 23165
     },
     {
-      "epoch": 0.6362363022163632,
+      "epoch": 0.6573779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 6.173782039093195e-06,
-      "loss": 0.881,
+      "learning_rate": 5.550651903338842e-06,
+      "loss": 0.9642,
       "step": 23166
     },
     {
-      "epoch": 0.6362637664442065,
+      "epoch": 0.6574063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 6.172960225885863e-06,
-      "loss": 0.8,
+      "learning_rate": 5.549828829856555e-06,
+      "loss": 0.7637,
       "step": 23167
     },
     {
-      "epoch": 0.6362912306720496,
+      "epoch": 0.6574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 6.1721384429595495e-06,
-      "loss": 0.8415,
+      "learning_rate": 5.549005793965065e-06,
+      "loss": 0.9725,
       "step": 23168
     },
     {
-      "epoch": 0.6363186948998929,
+      "epoch": 0.6574631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 6.171316690320761e-06,
-      "loss": 0.8447,
+      "learning_rate": 5.548182795671328e-06,
+      "loss": 0.8324,
       "step": 23169
     },
     {
-      "epoch": 0.6363461591277362,
+      "epoch": 0.6574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 6.170494967976002e-06,
-      "loss": 0.799,
+      "learning_rate": 5.547359834982286e-06,
+      "loss": 0.7942,
       "step": 23170
     },
     {
-      "epoch": 0.6363736233555793,
+      "epoch": 0.6575198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 6.169673275931764e-06,
-      "loss": 0.9278,
+      "learning_rate": 5.546536911904896e-06,
+      "loss": 0.8253,
       "step": 23171
     },
     {
-      "epoch": 0.6364010875834226,
+      "epoch": 0.6575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 6.168851614194559e-06,
-      "loss": 0.9089,
+      "learning_rate": 5.545714026446112e-06,
+      "loss": 0.8732,
       "step": 23172
     },
     {
-      "epoch": 0.6364285518112658,
+      "epoch": 0.6575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 6.168029982770881e-06,
-      "loss": 0.9135,
+      "learning_rate": 5.544891178612879e-06,
+      "loss": 0.8623,
       "step": 23173
     },
     {
-      "epoch": 0.6364560160391091,
+      "epoch": 0.6576049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 6.167208381667235e-06,
-      "loss": 0.8302,
+      "learning_rate": 5.54406836841215e-06,
+      "loss": 0.8814,
       "step": 23174
     },
     {
-      "epoch": 0.6364834802669523,
+      "epoch": 0.657633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 6.166386810890122e-06,
-      "loss": 0.8122,
+      "learning_rate": 5.543245595850876e-06,
+      "loss": 0.8238,
       "step": 23175
     },
     {
-      "epoch": 0.6365109444947955,
+      "epoch": 0.6576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 6.165565270446044e-06,
-      "loss": 0.9411,
+      "learning_rate": 5.542422860936006e-06,
+      "loss": 0.7987,
       "step": 23176
     },
     {
-      "epoch": 0.6365384087226388,
+      "epoch": 0.6576901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 6.164743760341495e-06,
-      "loss": 0.9404,
+      "learning_rate": 5.541600163674497e-06,
+      "loss": 0.751,
       "step": 23177
     },
     {
-      "epoch": 0.636565872950482,
+      "epoch": 0.6577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 6.163922280582979e-06,
-      "loss": 0.8559,
+      "learning_rate": 5.5407775040732845e-06,
+      "loss": 0.7821,
       "step": 23178
     },
     {
-      "epoch": 0.6365933371783252,
+      "epoch": 0.6577468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 6.163100831176998e-06,
-      "loss": 0.8741,
+      "learning_rate": 5.539954882139327e-06,
+      "loss": 0.9067,
       "step": 23179
     },
     {
-      "epoch": 0.6366208014061685,
+      "epoch": 0.6577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 6.162279412130046e-06,
-      "loss": 0.927,
+      "learning_rate": 5.539132297879575e-06,
+      "loss": 0.9527,
       "step": 23180
     },
     {
-      "epoch": 0.6366482656340117,
+      "epoch": 0.6578036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 6.161458023448628e-06,
-      "loss": 0.7684,
+      "learning_rate": 5.538309751300967e-06,
+      "loss": 0.9691,
       "step": 23181
     },
     {
-      "epoch": 0.6366757298618549,
+      "epoch": 0.6578320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 6.160636665139243e-06,
-      "loss": 0.8831,
+      "learning_rate": 5.5374872424104595e-06,
+      "loss": 0.8539,
       "step": 23182
     },
     {
-      "epoch": 0.6367031940896982,
+      "epoch": 0.6578603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 6.159815337208386e-06,
-      "loss": 0.9098,
+      "learning_rate": 5.536664771215002e-06,
+      "loss": 0.8567,
       "step": 23183
     },
     {
-      "epoch": 0.6367306583175414,
+      "epoch": 0.65788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 6.158994039662558e-06,
-      "loss": 0.795,
+      "learning_rate": 5.535842337721531e-06,
+      "loss": 0.7775,
       "step": 23184
     },
     {
-      "epoch": 0.6367581225453847,
+      "epoch": 0.6579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 6.158172772508257e-06,
-      "loss": 0.8754,
+      "learning_rate": 5.535019941937003e-06,
+      "loss": 0.8828,
       "step": 23185
     },
     {
-      "epoch": 0.6367855867732278,
+      "epoch": 0.6579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 6.157351535751979e-06,
-      "loss": 0.8196,
+      "learning_rate": 5.534197583868366e-06,
+      "loss": 0.7823,
       "step": 23186
     },
     {
-      "epoch": 0.6368130510010711,
+      "epoch": 0.6579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 6.156530329400228e-06,
-      "loss": 0.9562,
+      "learning_rate": 5.533375263522558e-06,
+      "loss": 0.8758,
       "step": 23187
     },
     {
-      "epoch": 0.6368405152289144,
+      "epoch": 0.6580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 6.155709153459498e-06,
-      "loss": 0.9712,
+      "learning_rate": 5.532552980906529e-06,
+      "loss": 0.7599,
       "step": 23188
     },
     {
-      "epoch": 0.6368679794567575,
+      "epoch": 0.6580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 6.1548880079362835e-06,
-      "loss": 0.893,
+      "learning_rate": 5.5317307360272275e-06,
+      "loss": 0.7431,
       "step": 23189
     },
     {
-      "epoch": 0.6368954436846008,
+      "epoch": 0.6580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 6.154066892837085e-06,
-      "loss": 0.9311,
+      "learning_rate": 5.5309085288915965e-06,
+      "loss": 0.8134,
       "step": 23190
     },
     {
-      "epoch": 0.636922907912444,
+      "epoch": 0.6580874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 6.1532458081684e-06,
-      "loss": 0.8538,
+      "learning_rate": 5.530086359506587e-06,
+      "loss": 0.8077,
       "step": 23191
     },
     {
-      "epoch": 0.6369503721402873,
+      "epoch": 0.6581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 6.152424753936722e-06,
-      "loss": 0.8047,
+      "learning_rate": 5.529264227879134e-06,
+      "loss": 0.8443,
       "step": 23192
     },
     {
-      "epoch": 0.6369778363681305,
+      "epoch": 0.658144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 6.1516037301485495e-06,
-      "loss": 0.7824,
+      "learning_rate": 5.528442134016188e-06,
+      "loss": 0.8318,
       "step": 23193
     },
     {
-      "epoch": 0.6370053005959737,
+      "epoch": 0.6581725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 6.150782736810385e-06,
-      "loss": 0.9795,
+      "learning_rate": 5.527620077924696e-06,
+      "loss": 0.7833,
       "step": 23194
     },
     {
-      "epoch": 0.637032764823817,
+      "epoch": 0.6582009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 6.149961773928713e-06,
-      "loss": 0.871,
+      "learning_rate": 5.526798059611593e-06,
+      "loss": 0.8751,
       "step": 23195
     },
     {
-      "epoch": 0.6370602290516603,
+      "epoch": 0.6582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 6.149140841510037e-06,
-      "loss": 0.8133,
+      "learning_rate": 5.5259760790838294e-06,
+      "loss": 0.9629,
       "step": 23196
     },
     {
-      "epoch": 0.6370876932795034,
+      "epoch": 0.6582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 6.148319939560848e-06,
-      "loss": 0.8608,
+      "learning_rate": 5.52515413634835e-06,
+      "loss": 0.6862,
       "step": 23197
     },
     {
-      "epoch": 0.6371151575073467,
+      "epoch": 0.6582860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 6.147499068087645e-06,
-      "loss": 0.8322,
+      "learning_rate": 5.52433223141209e-06,
+      "loss": 0.8814,
       "step": 23198
     },
     {
-      "epoch": 0.6371426217351899,
+      "epoch": 0.6583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 6.1466782270969224e-06,
-      "loss": 0.8339,
+      "learning_rate": 5.523510364281999e-06,
+      "loss": 0.8609,
       "step": 23199
     },
     {
-      "epoch": 0.6371700859630332,
+      "epoch": 0.6583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 6.145857416595177e-06,
-      "loss": 0.8002,
+      "learning_rate": 5.52268853496502e-06,
+      "loss": 0.839,
       "step": 23200
     },
     {
-      "epoch": 0.6371975501908764,
+      "epoch": 0.6583711691259931,
       "grad_norm": 0.0,
-      "learning_rate": 6.145036636588897e-06,
-      "loss": 0.849,
+      "learning_rate": 5.52186674346809e-06,
+      "loss": 0.8906,
       "step": 23201
     },
     {
-      "epoch": 0.6372250144187196,
+      "epoch": 0.6583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 6.144215887084579e-06,
-      "loss": 0.9025,
+      "learning_rate": 5.52104498979815e-06,
+      "loss": 0.8653,
       "step": 23202
     },
     {
-      "epoch": 0.6372524786465629,
+      "epoch": 0.658427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 6.143395168088721e-06,
-      "loss": 0.8493,
+      "learning_rate": 5.52022327396215e-06,
+      "loss": 0.9359,
       "step": 23203
     },
     {
-      "epoch": 0.637279942874406,
+      "epoch": 0.6584562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 6.142574479607813e-06,
-      "loss": 0.8851,
+      "learning_rate": 5.519401595967021e-06,
+      "loss": 0.9284,
       "step": 23204
     },
     {
-      "epoch": 0.6373074071022493,
+      "epoch": 0.6584846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 6.141753821648348e-06,
-      "loss": 0.8652,
+      "learning_rate": 5.518579955819707e-06,
+      "loss": 0.7828,
       "step": 23205
     },
     {
-      "epoch": 0.6373348713300926,
+      "epoch": 0.6585130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 6.1409331942168284e-06,
-      "loss": 0.8047,
+      "learning_rate": 5.517758353527152e-06,
+      "loss": 0.8615,
       "step": 23206
     },
     {
-      "epoch": 0.6373623355579358,
+      "epoch": 0.6585414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 6.140112597319734e-06,
-      "loss": 0.928,
+      "learning_rate": 5.516936789096292e-06,
+      "loss": 0.8117,
       "step": 23207
     },
     {
-      "epoch": 0.637389799785779,
+      "epoch": 0.6585698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 6.139292030963566e-06,
-      "loss": 0.821,
+      "learning_rate": 5.516115262534073e-06,
+      "loss": 0.8958,
       "step": 23208
     },
     {
-      "epoch": 0.6374172640136223,
+      "epoch": 0.6585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 6.138471495154814e-06,
-      "loss": 0.9409,
+      "learning_rate": 5.515293773847426e-06,
+      "loss": 0.8684,
       "step": 23209
     },
     {
-      "epoch": 0.6374447282414655,
+      "epoch": 0.6586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 6.137650989899971e-06,
-      "loss": 0.8716,
+      "learning_rate": 5.5144723230432935e-06,
+      "loss": 0.7311,
       "step": 23210
     },
     {
-      "epoch": 0.6374721924693088,
+      "epoch": 0.6586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 6.136830515205532e-06,
-      "loss": 0.8037,
+      "learning_rate": 5.51365091012862e-06,
+      "loss": 0.8837,
       "step": 23211
     },
     {
-      "epoch": 0.6374996566971519,
+      "epoch": 0.658683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 6.136010071077987e-06,
-      "loss": 0.8236,
+      "learning_rate": 5.512829535110334e-06,
+      "loss": 0.8369,
       "step": 23212
     },
     {
-      "epoch": 0.6375271209249952,
+      "epoch": 0.6587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 6.135189657523825e-06,
-      "loss": 0.7764,
+      "learning_rate": 5.512008197995379e-06,
+      "loss": 0.7863,
       "step": 23213
     },
     {
-      "epoch": 0.6375545851528385,
+      "epoch": 0.6587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 6.134369274549537e-06,
-      "loss": 0.8251,
+      "learning_rate": 5.511186898790697e-06,
+      "loss": 0.7865,
       "step": 23214
     },
     {
-      "epoch": 0.6375820493806816,
+      "epoch": 0.6587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 6.133548922161621e-06,
-      "loss": 0.8908,
+      "learning_rate": 5.510365637503216e-06,
+      "loss": 0.8059,
       "step": 23215
     },
     {
-      "epoch": 0.6376095136085249,
+      "epoch": 0.6587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 6.13272860036656e-06,
-      "loss": 0.8379,
+      "learning_rate": 5.509544414139878e-06,
+      "loss": 0.808,
       "step": 23216
     },
     {
-      "epoch": 0.6376369778363681,
+      "epoch": 0.6588251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 6.131908309170853e-06,
-      "loss": 0.8405,
+      "learning_rate": 5.508723228707625e-06,
+      "loss": 0.7288,
       "step": 23217
     },
     {
-      "epoch": 0.6376644420642114,
+      "epoch": 0.6588535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 6.131088048580982e-06,
-      "loss": 0.7714,
+      "learning_rate": 5.507902081213385e-06,
+      "loss": 0.8045,
       "step": 23218
     },
     {
-      "epoch": 0.6376919062920546,
+      "epoch": 0.6588819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 6.130267818603441e-06,
-      "loss": 0.6918,
+      "learning_rate": 5.507080971664098e-06,
+      "loss": 0.8106,
       "step": 23219
     },
     {
-      "epoch": 0.6377193705198978,
+      "epoch": 0.6589103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 6.1294476192447216e-06,
-      "loss": 0.8944,
+      "learning_rate": 5.5062599000666995e-06,
+      "loss": 0.8308,
       "step": 23220
     },
     {
-      "epoch": 0.6377468347477411,
+      "epoch": 0.658938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 6.128627450511311e-06,
-      "loss": 0.8902,
+      "learning_rate": 5.505438866428125e-06,
+      "loss": 0.8468,
       "step": 23221
     },
     {
-      "epoch": 0.6377742989755844,
+      "epoch": 0.658967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 6.127807312409699e-06,
-      "loss": 0.9599,
+      "learning_rate": 5.504617870755313e-06,
+      "loss": 0.8062,
       "step": 23222
     },
     {
-      "epoch": 0.6378017632034275,
+      "epoch": 0.6589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 6.1269872049463794e-06,
-      "loss": 0.8024,
+      "learning_rate": 5.503796913055197e-06,
+      "loss": 0.8196,
       "step": 23223
     },
     {
-      "epoch": 0.6378292274312708,
+      "epoch": 0.6590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 6.126167128127836e-06,
-      "loss": 0.7929,
+      "learning_rate": 5.5029759933347094e-06,
+      "loss": 0.8318,
       "step": 23224
     },
     {
-      "epoch": 0.637856691659114,
+      "epoch": 0.6590522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 6.125347081960555e-06,
-      "loss": 0.7358,
+      "learning_rate": 5.5021551116007835e-06,
+      "loss": 0.8786,
       "step": 23225
     },
     {
-      "epoch": 0.6378841558869572,
+      "epoch": 0.6590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 6.124527066451031e-06,
-      "loss": 0.821,
+      "learning_rate": 5.501334267860361e-06,
+      "loss": 0.8752,
       "step": 23226
     },
     {
-      "epoch": 0.6379116201148005,
+      "epoch": 0.6591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 6.12370708160575e-06,
-      "loss": 0.928,
+      "learning_rate": 5.500513462120365e-06,
+      "loss": 0.7933,
       "step": 23227
     },
     {
-      "epoch": 0.6379390843426437,
+      "epoch": 0.6591373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 6.1228871274312e-06,
-      "loss": 0.8006,
+      "learning_rate": 5.499692694387735e-06,
+      "loss": 0.8688,
       "step": 23228
     },
     {
-      "epoch": 0.637966548570487,
+      "epoch": 0.6591657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 6.1220672039338716e-06,
-      "loss": 0.816,
+      "learning_rate": 5.498871964669405e-06,
+      "loss": 0.7535,
       "step": 23229
     },
     {
-      "epoch": 0.6379940127983301,
+      "epoch": 0.6591940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 6.121247311120247e-06,
-      "loss": 0.8609,
+      "learning_rate": 5.498051272972304e-06,
+      "loss": 0.8217,
       "step": 23230
     },
     {
-      "epoch": 0.6380214770261734,
+      "epoch": 0.6592224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 6.120427448996814e-06,
-      "loss": 0.8603,
+      "learning_rate": 5.49723061930337e-06,
+      "loss": 0.9169,
       "step": 23231
     },
     {
-      "epoch": 0.6380489412540167,
+      "epoch": 0.6592508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 6.119607617570065e-06,
-      "loss": 0.8088,
+      "learning_rate": 5.4964100036695274e-06,
+      "loss": 0.7943,
       "step": 23232
     },
     {
-      "epoch": 0.6380764054818598,
+      "epoch": 0.6592792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 6.118787816846482e-06,
-      "loss": 0.8494,
+      "learning_rate": 5.495589426077712e-06,
+      "loss": 0.9485,
       "step": 23233
     },
     {
-      "epoch": 0.6381038697097031,
+      "epoch": 0.6593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 6.117968046832552e-06,
-      "loss": 0.8563,
+      "learning_rate": 5.494768886534858e-06,
+      "loss": 0.8702,
       "step": 23234
     },
     {
-      "epoch": 0.6381313339375464,
+      "epoch": 0.6593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 6.117148307534769e-06,
-      "loss": 0.8906,
+      "learning_rate": 5.493948385047887e-06,
+      "loss": 0.7466,
       "step": 23235
     },
     {
-      "epoch": 0.6381587981653896,
+      "epoch": 0.6593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 6.116328598959609e-06,
-      "loss": 0.7803,
+      "learning_rate": 5.49312792162374e-06,
+      "loss": 0.7292,
       "step": 23236
     },
     {
-      "epoch": 0.6381862623932328,
+      "epoch": 0.6593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 6.11550892111356e-06,
-      "loss": 0.8661,
+      "learning_rate": 5.492307496269349e-06,
+      "loss": 0.81,
       "step": 23237
     },
     {
-      "epoch": 0.638213726621076,
+      "epoch": 0.6594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.114689274003107e-06,
-      "loss": 0.89,
+      "learning_rate": 5.4914871089916334e-06,
+      "loss": 0.9491,
       "step": 23238
     },
     {
-      "epoch": 0.6382411908489193,
+      "epoch": 0.6594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 6.113869657634742e-06,
-      "loss": 0.8103,
+      "learning_rate": 5.490666759797529e-06,
+      "loss": 0.8751,
       "step": 23239
     },
     {
-      "epoch": 0.6382686550767626,
+      "epoch": 0.659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 6.1130500720149435e-06,
-      "loss": 0.8008,
+      "learning_rate": 5.489846448693972e-06,
+      "loss": 0.8458,
       "step": 23240
     },
     {
-      "epoch": 0.6382961193046057,
+      "epoch": 0.6595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 6.112230517150202e-06,
-      "loss": 0.7933,
+      "learning_rate": 5.489026175687878e-06,
+      "loss": 0.8426,
       "step": 23241
     },
     {
-      "epoch": 0.638323583532449,
+      "epoch": 0.6595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 6.111410993046993e-06,
-      "loss": 0.7663,
+      "learning_rate": 5.488205940786184e-06,
+      "loss": 0.8286,
       "step": 23242
     },
     {
-      "epoch": 0.6383510477602922,
+      "epoch": 0.6595629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 6.110591499711811e-06,
-      "loss": 0.8705,
+      "learning_rate": 5.487385743995821e-06,
+      "loss": 0.7582,
       "step": 23243
     },
     {
-      "epoch": 0.6383785119881354,
+      "epoch": 0.6595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 6.109772037151132e-06,
-      "loss": 0.8913,
+      "learning_rate": 5.486565585323709e-06,
+      "loss": 0.8205,
       "step": 23244
     },
     {
-      "epoch": 0.6384059762159787,
+      "epoch": 0.6596197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 6.108952605371445e-06,
-      "loss": 0.8387,
+      "learning_rate": 5.4857454647767805e-06,
+      "loss": 0.8405,
       "step": 23245
     },
     {
-      "epoch": 0.6384334404438219,
+      "epoch": 0.6596481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 6.108133204379232e-06,
-      "loss": 0.7684,
+      "learning_rate": 5.484925382361967e-06,
+      "loss": 0.6484,
       "step": 23246
     },
     {
-      "epoch": 0.6384609046716652,
+      "epoch": 0.6596765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 6.107313834180979e-06,
-      "loss": 0.9158,
+      "learning_rate": 5.4841053380861875e-06,
+      "loss": 0.8131,
       "step": 23247
     },
     {
-      "epoch": 0.6384883688995084,
+      "epoch": 0.6597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 6.106494494783166e-06,
-      "loss": 0.9304,
+      "learning_rate": 5.483285331956374e-06,
+      "loss": 0.859,
       "step": 23248
     },
     {
-      "epoch": 0.6385158331273516,
+      "epoch": 0.6597332576617481,
       "grad_norm": 0.0,
-      "learning_rate": 6.105675186192275e-06,
-      "loss": 0.8204,
+      "learning_rate": 5.482465363979457e-06,
+      "loss": 0.7975,
       "step": 23249
     },
     {
-      "epoch": 0.6385432973551949,
+      "epoch": 0.6597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 6.104855908414791e-06,
-      "loss": 0.8529,
+      "learning_rate": 5.481645434162353e-06,
+      "loss": 0.841,
       "step": 23250
     },
     {
-      "epoch": 0.638570761583038,
+      "epoch": 0.6597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 6.1040366614571964e-06,
-      "loss": 0.794,
+      "learning_rate": 5.480825542511991e-06,
+      "loss": 0.7891,
       "step": 23251
     },
     {
-      "epoch": 0.6385982258108813,
+      "epoch": 0.6598183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 6.103217445325972e-06,
-      "loss": 0.8093,
+      "learning_rate": 5.4800056890353025e-06,
+      "loss": 0.9199,
       "step": 23252
     },
     {
-      "epoch": 0.6386256900387246,
+      "epoch": 0.6598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 6.102398260027604e-06,
-      "loss": 0.8382,
+      "learning_rate": 5.479185873739208e-06,
+      "loss": 0.7986,
       "step": 23253
     },
     {
-      "epoch": 0.6386531542665678,
+      "epoch": 0.6598751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 6.101579105568567e-06,
-      "loss": 0.8513,
+      "learning_rate": 5.478366096630636e-06,
+      "loss": 0.8994,
       "step": 23254
     },
     {
-      "epoch": 0.638680618494411,
+      "epoch": 0.6599035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 6.100759981955349e-06,
-      "loss": 0.871,
+      "learning_rate": 5.477546357716508e-06,
+      "loss": 0.9705,
       "step": 23255
     },
     {
-      "epoch": 0.6387080827222542,
+      "epoch": 0.6599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 6.099940889194425e-06,
-      "loss": 0.9125,
+      "learning_rate": 5.476726657003746e-06,
+      "loss": 0.8134,
       "step": 23256
     },
     {
-      "epoch": 0.6387355469500975,
+      "epoch": 0.6599602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 6.09912182729228e-06,
-      "loss": 0.8648,
+      "learning_rate": 5.475906994499283e-06,
+      "loss": 0.7606,
       "step": 23257
     },
     {
-      "epoch": 0.6387630111779408,
+      "epoch": 0.659988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 6.098302796255394e-06,
-      "loss": 0.9642,
+      "learning_rate": 5.475087370210033e-06,
+      "loss": 0.8022,
       "step": 23258
     },
     {
-      "epoch": 0.6387904754057839,
+      "epoch": 0.6600170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 6.097483796090252e-06,
-      "loss": 0.8236,
+      "learning_rate": 5.474267784142923e-06,
+      "loss": 0.8536,
       "step": 23259
     },
     {
-      "epoch": 0.6388179396336272,
+      "epoch": 0.6600454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 6.096664826803328e-06,
-      "loss": 0.8121,
+      "learning_rate": 5.47344823630488e-06,
+      "loss": 0.813,
       "step": 23260
     },
     {
-      "epoch": 0.6388454038614705,
+      "epoch": 0.6600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 6.095845888401102e-06,
-      "loss": 0.8602,
+      "learning_rate": 5.47262872670282e-06,
+      "loss": 0.833,
       "step": 23261
     },
     {
-      "epoch": 0.6388728680893137,
+      "epoch": 0.6601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 6.0950269808900556e-06,
-      "loss": 0.8109,
+      "learning_rate": 5.4718092553436684e-06,
+      "loss": 0.8784,
       "step": 23262
     },
     {
-      "epoch": 0.6389003323171569,
+      "epoch": 0.6601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 6.094208104276669e-06,
-      "loss": 0.8484,
+      "learning_rate": 5.4709898222343515e-06,
+      "loss": 0.9144,
       "step": 23263
     },
     {
-      "epoch": 0.6389277965450001,
+      "epoch": 0.6601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 6.0933892585674195e-06,
-      "loss": 0.9294,
+      "learning_rate": 5.4701704273817825e-06,
+      "loss": 0.8745,
       "step": 23264
     },
     {
-      "epoch": 0.6389552607728434,
+      "epoch": 0.6601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 6.092570443768793e-06,
-      "loss": 0.8218,
+      "learning_rate": 5.469351070792888e-06,
+      "loss": 0.7761,
       "step": 23265
     },
     {
-      "epoch": 0.6389827250006866,
+      "epoch": 0.6602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 6.091751659887258e-06,
-      "loss": 0.8621,
+      "learning_rate": 5.468531752474591e-06,
+      "loss": 0.7957,
       "step": 23266
     },
     {
-      "epoch": 0.6390101892285298,
+      "epoch": 0.6602440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 6.090932906929299e-06,
-      "loss": 0.8707,
+      "learning_rate": 5.467712472433807e-06,
+      "loss": 0.8888,
       "step": 23267
     },
     {
-      "epoch": 0.6390376534563731,
+      "epoch": 0.660272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 6.09011418490139e-06,
-      "loss": 0.8133,
+      "learning_rate": 5.466893230677463e-06,
+      "loss": 0.8364,
       "step": 23268
     },
     {
-      "epoch": 0.6390651176842163,
+      "epoch": 0.6603007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 6.089295493810015e-06,
-      "loss": 0.8852,
+      "learning_rate": 5.466074027212475e-06,
+      "loss": 0.8611,
       "step": 23269
     },
     {
-      "epoch": 0.6390925819120595,
+      "epoch": 0.6603291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 6.0884768336616475e-06,
-      "loss": 0.8998,
+      "learning_rate": 5.465254862045761e-06,
+      "loss": 0.7588,
       "step": 23270
     },
     {
-      "epoch": 0.6391200461399028,
+      "epoch": 0.6603575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 6.0876582044627694e-06,
-      "loss": 0.8893,
+      "learning_rate": 5.464435735184248e-06,
+      "loss": 0.7821,
       "step": 23271
     },
     {
-      "epoch": 0.639147510367746,
+      "epoch": 0.6603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 6.086839606219854e-06,
-      "loss": 0.9502,
+      "learning_rate": 5.463616646634846e-06,
+      "loss": 0.8861,
       "step": 23272
     },
     {
-      "epoch": 0.6391749745955893,
+      "epoch": 0.6604143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 6.086021038939378e-06,
-      "loss": 0.8206,
+      "learning_rate": 5.462797596404477e-06,
+      "loss": 0.8581,
       "step": 23273
     },
     {
-      "epoch": 0.6392024388234325,
+      "epoch": 0.6604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 6.08520250262782e-06,
-      "loss": 0.888,
+      "learning_rate": 5.461978584500066e-06,
+      "loss": 0.7791,
       "step": 23274
     },
     {
-      "epoch": 0.6392299030512757,
+      "epoch": 0.6604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 6.084383997291656e-06,
-      "loss": 0.8044,
+      "learning_rate": 5.461159610928521e-06,
+      "loss": 0.8391,
       "step": 23275
     },
     {
-      "epoch": 0.639257367279119,
+      "epoch": 0.66049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 6.083565522937367e-06,
-      "loss": 0.8598,
+      "learning_rate": 5.460340675696766e-06,
+      "loss": 0.9114,
       "step": 23276
     },
     {
-      "epoch": 0.6392848315069621,
+      "epoch": 0.660527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 6.082747079571421e-06,
-      "loss": 0.9075,
+      "learning_rate": 5.45952177881172e-06,
+      "loss": 0.7905,
       "step": 23277
     },
     {
-      "epoch": 0.6393122957348054,
+      "epoch": 0.6605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 6.081928667200295e-06,
-      "loss": 0.8518,
+      "learning_rate": 5.4587029202802946e-06,
+      "loss": 0.8424,
       "step": 23278
     },
     {
-      "epoch": 0.6393397599626487,
+      "epoch": 0.6605845629965947,
       "grad_norm": 0.0,
-      "learning_rate": 6.08111028583047e-06,
-      "loss": 0.8386,
+      "learning_rate": 5.45788410010941e-06,
+      "loss": 0.8396,
       "step": 23279
     },
     {
-      "epoch": 0.6393672241904919,
+      "epoch": 0.6606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 6.080291935468416e-06,
-      "loss": 0.8648,
+      "learning_rate": 5.457065318305985e-06,
+      "loss": 0.8344,
       "step": 23280
     },
     {
-      "epoch": 0.6393946884183351,
+      "epoch": 0.6606413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 6.079473616120612e-06,
-      "loss": 0.8626,
+      "learning_rate": 5.4562465748769266e-06,
+      "loss": 0.7803,
       "step": 23281
     },
     {
-      "epoch": 0.6394221526461783,
+      "epoch": 0.6606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 6.078655327793535e-06,
-      "loss": 0.8918,
+      "learning_rate": 5.455427869829166e-06,
+      "loss": 0.8279,
       "step": 23282
     },
     {
-      "epoch": 0.6394496168740216,
+      "epoch": 0.6606980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 6.077837070493653e-06,
-      "loss": 0.8377,
+      "learning_rate": 5.454609203169606e-06,
+      "loss": 0.8506,
       "step": 23283
     },
     {
-      "epoch": 0.6394770811018649,
+      "epoch": 0.6607264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 6.077018844227443e-06,
-      "loss": 0.8439,
+      "learning_rate": 5.453790574905167e-06,
+      "loss": 0.8819,
       "step": 23284
     },
     {
-      "epoch": 0.639504545329708,
+      "epoch": 0.6607548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 6.0762006490013795e-06,
-      "loss": 0.7388,
+      "learning_rate": 5.452971985042767e-06,
+      "loss": 0.8774,
       "step": 23285
     },
     {
-      "epoch": 0.6395320095575513,
+      "epoch": 0.660783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 6.075382484821934e-06,
-      "loss": 0.78,
+      "learning_rate": 5.452153433589313e-06,
+      "loss": 0.8792,
       "step": 23286
     },
     {
-      "epoch": 0.6395594737853946,
+      "epoch": 0.660811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 6.074564351695588e-06,
-      "loss": 0.6946,
+      "learning_rate": 5.451334920551723e-06,
+      "loss": 0.8016,
       "step": 23287
     },
     {
-      "epoch": 0.6395869380132377,
+      "epoch": 0.6608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 6.073746249628809e-06,
-      "loss": 0.9054,
+      "learning_rate": 5.450516445936915e-06,
+      "loss": 0.9156,
       "step": 23288
     },
     {
-      "epoch": 0.639614402241081,
+      "epoch": 0.6608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 6.072928178628067e-06,
-      "loss": 0.8156,
+      "learning_rate": 5.449698009751796e-06,
+      "loss": 0.8651,
       "step": 23289
     },
     {
-      "epoch": 0.6396418664689242,
+      "epoch": 0.6608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 6.072110138699838e-06,
-      "loss": 0.9084,
+      "learning_rate": 5.448879612003283e-06,
+      "loss": 0.8666,
       "step": 23290
     },
     {
-      "epoch": 0.6396693306967675,
+      "epoch": 0.6609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 6.071292129850599e-06,
-      "loss": 0.8157,
+      "learning_rate": 5.4480612526982915e-06,
+      "loss": 0.7929,
       "step": 23291
     },
     {
-      "epoch": 0.6396967949246107,
+      "epoch": 0.6609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 6.070474152086814e-06,
-      "loss": 0.8414,
+      "learning_rate": 5.447242931843728e-06,
+      "loss": 0.806,
       "step": 23292
     },
     {
-      "epoch": 0.6397242591524539,
+      "epoch": 0.6609818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 6.069656205414962e-06,
-      "loss": 0.916,
+      "learning_rate": 5.446424649446507e-06,
+      "loss": 0.9184,
       "step": 23293
     },
     {
-      "epoch": 0.6397517233802972,
+      "epoch": 0.6610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 6.068838289841514e-06,
-      "loss": 0.815,
+      "learning_rate": 5.445606405513546e-06,
+      "loss": 0.7726,
       "step": 23294
     },
     {
-      "epoch": 0.6397791876081403,
+      "epoch": 0.661038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 6.068020405372939e-06,
-      "loss": 0.7794,
+      "learning_rate": 5.444788200051749e-06,
+      "loss": 0.8369,
       "step": 23295
     },
     {
-      "epoch": 0.6398066518359836,
+      "epoch": 0.6610669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 6.06720255201571e-06,
-      "loss": 0.8156,
+      "learning_rate": 5.443970033068031e-06,
+      "loss": 0.8171,
       "step": 23296
     },
     {
-      "epoch": 0.6398341160638269,
+      "epoch": 0.6610953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 6.0663847297762956e-06,
-      "loss": 0.8441,
+      "learning_rate": 5.443151904569302e-06,
+      "loss": 0.8098,
       "step": 23297
     },
     {
-      "epoch": 0.6398615802916701,
+      "epoch": 0.6611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 6.0655669386611695e-06,
-      "loss": 0.8309,
+      "learning_rate": 5.442333814562475e-06,
+      "loss": 0.8132,
       "step": 23298
     },
     {
-      "epoch": 0.6398890445195133,
+      "epoch": 0.6611520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 6.064749178676804e-06,
-      "loss": 0.842,
+      "learning_rate": 5.441515763054462e-06,
+      "loss": 0.7907,
       "step": 23299
     },
     {
-      "epoch": 0.6399165087473566,
+      "epoch": 0.6611804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 6.063931449829669e-06,
-      "loss": 0.9071,
+      "learning_rate": 5.440697750052166e-06,
+      "loss": 0.8757,
       "step": 23300
     },
     {
-      "epoch": 0.6399439729751998,
+      "epoch": 0.6612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 6.0631137521262294e-06,
-      "loss": 0.8713,
+      "learning_rate": 5.439879775562502e-06,
+      "loss": 0.8507,
       "step": 23301
     },
     {
-      "epoch": 0.6399714372030431,
+      "epoch": 0.6612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 6.062296085572958e-06,
-      "loss": 0.7427,
+      "learning_rate": 5.439061839592382e-06,
+      "loss": 0.8794,
       "step": 23302
     },
     {
-      "epoch": 0.6399989014308862,
+      "epoch": 0.6612656072644721,
       "grad_norm": 0.0,
-      "learning_rate": 6.061478450176329e-06,
-      "loss": 0.809,
+      "learning_rate": 5.438243942148709e-06,
+      "loss": 0.8473,
       "step": 23303
     },
     {
-      "epoch": 0.6400263656587295,
+      "epoch": 0.6612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 6.060660845942805e-06,
-      "loss": 0.8253,
+      "learning_rate": 5.437426083238393e-06,
+      "loss": 1.0122,
       "step": 23304
     },
     {
-      "epoch": 0.6400538298865728,
+      "epoch": 0.661322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 6.059843272878858e-06,
-      "loss": 0.7599,
+      "learning_rate": 5.436608262868348e-06,
+      "loss": 0.8208,
       "step": 23305
     },
     {
-      "epoch": 0.6400812941144159,
+      "epoch": 0.6613507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 6.0590257309909615e-06,
-      "loss": 0.7822,
+      "learning_rate": 5.435790481045474e-06,
+      "loss": 0.8976,
       "step": 23306
     },
     {
-      "epoch": 0.6401087583422592,
+      "epoch": 0.6613791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 6.058208220285578e-06,
-      "loss": 0.8418,
+      "learning_rate": 5.4349727377766845e-06,
+      "loss": 0.8163,
       "step": 23307
     },
     {
-      "epoch": 0.6401362225701024,
+      "epoch": 0.6614074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 6.057390740769178e-06,
-      "loss": 0.8898,
+      "learning_rate": 5.434155033068889e-06,
+      "loss": 0.8721,
       "step": 23308
     },
     {
-      "epoch": 0.6401636867979457,
+      "epoch": 0.6614358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 6.056573292448227e-06,
-      "loss": 0.7534,
+      "learning_rate": 5.4333373669289855e-06,
+      "loss": 0.8148,
       "step": 23309
     },
     {
-      "epoch": 0.6401911510257889,
+      "epoch": 0.6614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 6.055755875329197e-06,
-      "loss": 0.7635,
+      "learning_rate": 5.4325197393638885e-06,
+      "loss": 0.9265,
       "step": 23310
     },
     {
-      "epoch": 0.6402186152536321,
+      "epoch": 0.6614926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 6.054938489418556e-06,
-      "loss": 0.8046,
+      "learning_rate": 5.4317021503805045e-06,
+      "loss": 0.7874,
       "step": 23311
     },
     {
-      "epoch": 0.6402460794814754,
+      "epoch": 0.6615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 6.0541211347227715e-06,
-      "loss": 0.7731,
+      "learning_rate": 5.430884599985732e-06,
+      "loss": 0.869,
       "step": 23312
     },
     {
-      "epoch": 0.6402735437093187,
+      "epoch": 0.6615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 6.053303811248306e-06,
-      "loss": 0.8318,
+      "learning_rate": 5.4300670881864895e-06,
+      "loss": 0.8651,
       "step": 23313
     },
     {
-      "epoch": 0.6403010079371618,
+      "epoch": 0.661577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.052486519001628e-06,
-      "loss": 0.8199,
+      "learning_rate": 5.429249614989671e-06,
+      "loss": 0.7969,
       "step": 23314
     },
     {
-      "epoch": 0.6403284721650051,
+      "epoch": 0.6616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 6.051669257989208e-06,
-      "loss": 0.8322,
+      "learning_rate": 5.428432180402188e-06,
+      "loss": 0.7478,
       "step": 23315
     },
     {
-      "epoch": 0.6403559363928483,
+      "epoch": 0.6616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 6.050852028217509e-06,
-      "loss": 0.8437,
+      "learning_rate": 5.427614784430948e-06,
+      "loss": 0.981,
       "step": 23316
     },
     {
-      "epoch": 0.6403834006206915,
+      "epoch": 0.6616628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 6.050034829692999e-06,
-      "loss": 0.7573,
+      "learning_rate": 5.426797427082846e-06,
+      "loss": 0.8726,
       "step": 23317
     },
     {
-      "epoch": 0.6404108648485348,
+      "epoch": 0.6616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 6.049217662422146e-06,
-      "loss": 0.8315,
+      "learning_rate": 5.425980108364793e-06,
+      "loss": 0.7819,
       "step": 23318
     },
     {
-      "epoch": 0.640438329076378,
+      "epoch": 0.6617196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 6.048400526411411e-06,
-      "loss": 0.7233,
+      "learning_rate": 5.4251628282836955e-06,
+      "loss": 0.8975,
       "step": 23319
     },
     {
-      "epoch": 0.6404657933042213,
+      "epoch": 0.6617480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 6.047583421667262e-06,
-      "loss": 0.9624,
+      "learning_rate": 5.424345586846449e-06,
+      "loss": 0.8643,
       "step": 23320
     },
     {
-      "epoch": 0.6404932575320644,
+      "epoch": 0.6617763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 6.046766348196163e-06,
-      "loss": 0.8607,
+      "learning_rate": 5.423528384059961e-06,
+      "loss": 0.8189,
       "step": 23321
     },
     {
-      "epoch": 0.6405207217599077,
+      "epoch": 0.6618047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 6.045949306004581e-06,
-      "loss": 0.8419,
+      "learning_rate": 5.422711219931138e-06,
+      "loss": 0.8166,
       "step": 23322
     },
     {
-      "epoch": 0.640548185987751,
+      "epoch": 0.66183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 6.04513229509898e-06,
-      "loss": 0.8335,
+      "learning_rate": 5.421894094466875e-06,
+      "loss": 0.8149,
       "step": 23323
     },
     {
-      "epoch": 0.6405756502155942,
+      "epoch": 0.661861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 6.044315315485827e-06,
-      "loss": 0.7433,
+      "learning_rate": 5.42107700767408e-06,
+      "loss": 0.8252,
       "step": 23324
     },
     {
-      "epoch": 0.6406031144434374,
+      "epoch": 0.6618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 6.04349836717158e-06,
-      "loss": 0.7906,
+      "learning_rate": 5.420259959559655e-06,
+      "loss": 0.8286,
       "step": 23325
     },
     {
-      "epoch": 0.6406305786712807,
+      "epoch": 0.6619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 6.042681450162705e-06,
-      "loss": 0.8507,
+      "learning_rate": 5.419442950130498e-06,
+      "loss": 0.7693,
       "step": 23326
     },
     {
-      "epoch": 0.6406580428991239,
+      "epoch": 0.6619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 6.04186456446567e-06,
-      "loss": 0.8851,
+      "learning_rate": 5.418625979393508e-06,
+      "loss": 0.8511,
       "step": 23327
     },
     {
-      "epoch": 0.6406855071269671,
+      "epoch": 0.6619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 6.041047710086934e-06,
-      "loss": 0.8722,
+      "learning_rate": 5.4178090473555975e-06,
+      "loss": 0.8159,
       "step": 23328
     },
     {
-      "epoch": 0.6407129713548103,
+      "epoch": 0.6620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 6.040230887032963e-06,
-      "loss": 0.9392,
+      "learning_rate": 5.416992154023656e-06,
+      "loss": 0.8376,
       "step": 23329
     },
     {
-      "epoch": 0.6407404355826536,
+      "epoch": 0.6620317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 6.039414095310221e-06,
-      "loss": 0.8717,
+      "learning_rate": 5.416175299404588e-06,
+      "loss": 0.847,
       "step": 23330
     },
     {
-      "epoch": 0.6407678998104969,
+      "epoch": 0.6620601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 6.038597334925167e-06,
-      "loss": 0.8638,
+      "learning_rate": 5.415358483505298e-06,
+      "loss": 0.9901,
       "step": 23331
     },
     {
-      "epoch": 0.64079536403834,
+      "epoch": 0.6620885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 6.037780605884266e-06,
-      "loss": 0.9129,
+      "learning_rate": 5.4145417063326755e-06,
+      "loss": 0.7943,
       "step": 23332
     },
     {
-      "epoch": 0.6408228282661833,
+      "epoch": 0.6621169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 6.036963908193978e-06,
-      "loss": 0.8559,
+      "learning_rate": 5.4137249678936265e-06,
+      "loss": 0.8687,
       "step": 23333
     },
     {
-      "epoch": 0.6408502924940266,
+      "epoch": 0.6621452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 6.036147241860767e-06,
-      "loss": 0.7569,
+      "learning_rate": 5.412908268195052e-06,
+      "loss": 0.8445,
       "step": 23334
     },
     {
-      "epoch": 0.6408777567218698,
+      "epoch": 0.6621736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 6.035330606891099e-06,
-      "loss": 0.8603,
+      "learning_rate": 5.412091607243845e-06,
+      "loss": 0.8261,
       "step": 23335
     },
     {
-      "epoch": 0.640905220949713,
+      "epoch": 0.6622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 6.034514003291427e-06,
-      "loss": 0.8446,
+      "learning_rate": 5.411274985046906e-06,
+      "loss": 0.8385,
       "step": 23336
     },
     {
-      "epoch": 0.6409326851775562,
+      "epoch": 0.6622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 6.033697431068215e-06,
-      "loss": 0.7776,
+      "learning_rate": 5.410458401611136e-06,
+      "loss": 0.8512,
       "step": 23337
     },
     {
-      "epoch": 0.6409601494053995,
+      "epoch": 0.6622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 6.032880890227927e-06,
-      "loss": 0.9792,
+      "learning_rate": 5.409641856943429e-06,
+      "loss": 0.9017,
       "step": 23338
     },
     {
-      "epoch": 0.6409876136332427,
+      "epoch": 0.6622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 6.032064380777023e-06,
-      "loss": 0.7921,
+      "learning_rate": 5.408825351050684e-06,
+      "loss": 0.8563,
       "step": 23339
     },
     {
-      "epoch": 0.6410150778610859,
+      "epoch": 0.6623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 6.0312479027219595e-06,
-      "loss": 0.8254,
+      "learning_rate": 5.408008883939796e-06,
+      "loss": 0.845,
       "step": 23340
     },
     {
-      "epoch": 0.6410425420889292,
+      "epoch": 0.6623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 6.030431456069206e-06,
-      "loss": 0.8616,
+      "learning_rate": 5.407192455617662e-06,
+      "loss": 0.778,
       "step": 23341
     },
     {
-      "epoch": 0.6410700063167724,
+      "epoch": 0.662372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 6.029615040825212e-06,
-      "loss": 0.8225,
+      "learning_rate": 5.4063760660911855e-06,
+      "loss": 0.8496,
       "step": 23342
     },
     {
-      "epoch": 0.6410974705446156,
+      "epoch": 0.6624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 6.028798656996441e-06,
-      "loss": 0.7984,
+      "learning_rate": 5.405559715367248e-06,
+      "loss": 0.8284,
       "step": 23343
     },
     {
-      "epoch": 0.6411249347724589,
+      "epoch": 0.6624290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 6.027982304589355e-06,
-      "loss": 0.8853,
+      "learning_rate": 5.404743403452758e-06,
+      "loss": 0.8203,
       "step": 23344
     },
     {
-      "epoch": 0.6411523990003021,
+      "epoch": 0.6624574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 6.0271659836104105e-06,
-      "loss": 0.773,
+      "learning_rate": 5.403927130354611e-06,
+      "loss": 0.8189,
       "step": 23345
     },
     {
-      "epoch": 0.6411798632281454,
+      "epoch": 0.6624858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 6.026349694066067e-06,
-      "loss": 0.7998,
+      "learning_rate": 5.403110896079695e-06,
+      "loss": 0.8959,
       "step": 23346
     },
     {
-      "epoch": 0.6412073274559886,
+      "epoch": 0.6625141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 6.025533435962789e-06,
-      "loss": 0.9328,
+      "learning_rate": 5.402294700634907e-06,
+      "loss": 0.8347,
       "step": 23347
     },
     {
-      "epoch": 0.6412347916838318,
+      "epoch": 0.6625425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 6.024717209307028e-06,
-      "loss": 0.9277,
+      "learning_rate": 5.401478544027146e-06,
+      "loss": 0.8248,
       "step": 23348
     },
     {
-      "epoch": 0.6412622559116751,
+      "epoch": 0.6625709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 6.02390101410524e-06,
-      "loss": 0.8433,
+      "learning_rate": 5.400662426263299e-06,
+      "loss": 0.9256,
       "step": 23349
     },
     {
-      "epoch": 0.6412897201395182,
+      "epoch": 0.6625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 6.0230848503638895e-06,
-      "loss": 0.8808,
+      "learning_rate": 5.399846347350263e-06,
+      "loss": 0.8333,
       "step": 23350
     },
     {
-      "epoch": 0.6413171843673615,
+      "epoch": 0.6626276958002271,
       "grad_norm": 0.0,
-      "learning_rate": 6.022268718089435e-06,
-      "loss": 0.7874,
+      "learning_rate": 5.399030307294937e-06,
+      "loss": 0.8567,
       "step": 23351
     },
     {
-      "epoch": 0.6413446485952048,
+      "epoch": 0.6626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 6.021452617288327e-06,
-      "loss": 0.8755,
+      "learning_rate": 5.398214306104204e-06,
+      "loss": 0.858,
       "step": 23352
     },
     {
-      "epoch": 0.641372112823048,
+      "epoch": 0.6626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 6.020636547967031e-06,
-      "loss": 0.9235,
+      "learning_rate": 5.3973983437849605e-06,
+      "loss": 0.8388,
       "step": 23353
     },
     {
-      "epoch": 0.6413995770508912,
+      "epoch": 0.6627128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 6.019820510131997e-06,
-      "loss": 0.847,
+      "learning_rate": 5.396582420344105e-06,
+      "loss": 0.7433,
       "step": 23354
     },
     {
-      "epoch": 0.6414270412787344,
+      "epoch": 0.6627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 6.019004503789686e-06,
-      "loss": 0.8838,
+      "learning_rate": 5.39576653578852e-06,
+      "loss": 0.7738,
       "step": 23355
     },
     {
-      "epoch": 0.6414545055065777,
+      "epoch": 0.6627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 6.018188528946554e-06,
-      "loss": 0.8074,
+      "learning_rate": 5.394950690125104e-06,
+      "loss": 0.8045,
       "step": 23356
     },
     {
-      "epoch": 0.641481969734421,
+      "epoch": 0.6627979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 6.017372585609056e-06,
-      "loss": 0.7777,
+      "learning_rate": 5.394134883360749e-06,
+      "loss": 0.8984,
       "step": 23357
     },
     {
-      "epoch": 0.6415094339622641,
+      "epoch": 0.6628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 6.016556673783648e-06,
-      "loss": 0.8974,
+      "learning_rate": 5.393319115502336e-06,
+      "loss": 0.9272,
       "step": 23358
     },
     {
-      "epoch": 0.6415368981901074,
+      "epoch": 0.6628547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 6.01574079347679e-06,
-      "loss": 0.8071,
+      "learning_rate": 5.392503386556771e-06,
+      "loss": 0.8769,
       "step": 23359
     },
     {
-      "epoch": 0.6415643624179507,
+      "epoch": 0.662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 6.014924944694933e-06,
-      "loss": 0.8103,
+      "learning_rate": 5.391687696530934e-06,
+      "loss": 0.8428,
       "step": 23360
     },
     {
-      "epoch": 0.6415918266457938,
+      "epoch": 0.6629114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 6.014109127444532e-06,
-      "loss": 0.8465,
+      "learning_rate": 5.390872045431718e-06,
+      "loss": 0.8962,
       "step": 23361
     },
     {
-      "epoch": 0.6416192908736371,
+      "epoch": 0.6629398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 6.013293341732043e-06,
-      "loss": 0.9071,
+      "learning_rate": 5.390056433266017e-06,
+      "loss": 0.9384,
       "step": 23362
     },
     {
-      "epoch": 0.6416467551014803,
+      "epoch": 0.6629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 6.012477587563923e-06,
-      "loss": 0.9139,
+      "learning_rate": 5.389240860040712e-06,
+      "loss": 0.8892,
       "step": 23363
     },
     {
-      "epoch": 0.6416742193293236,
+      "epoch": 0.6629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 6.011661864946623e-06,
-      "loss": 0.8846,
+      "learning_rate": 5.388425325762696e-06,
+      "loss": 0.8568,
       "step": 23364
     },
     {
-      "epoch": 0.6417016835571668,
+      "epoch": 0.6630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 6.010846173886604e-06,
-      "loss": 0.9153,
+      "learning_rate": 5.387609830438865e-06,
+      "loss": 0.9018,
       "step": 23365
     },
     {
-      "epoch": 0.64172914778501,
+      "epoch": 0.6630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 6.010030514390311e-06,
-      "loss": 0.82,
+      "learning_rate": 5.386794374076096e-06,
+      "loss": 0.9013,
       "step": 23366
     },
     {
-      "epoch": 0.6417566120128533,
+      "epoch": 0.6630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 6.009214886464202e-06,
-      "loss": 0.9415,
+      "learning_rate": 5.385978956681281e-06,
+      "loss": 0.8021,
       "step": 23367
     },
     {
-      "epoch": 0.6417840762406964,
+      "epoch": 0.6631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 6.008399290114733e-06,
-      "loss": 0.9384,
+      "learning_rate": 5.385163578261314e-06,
+      "loss": 0.8218,
       "step": 23368
     },
     {
-      "epoch": 0.6418115404685397,
+      "epoch": 0.6631384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 6.007583725348352e-06,
-      "loss": 0.9346,
+      "learning_rate": 5.384348238823074e-06,
+      "loss": 0.8263,
       "step": 23369
     },
     {
-      "epoch": 0.641839004696383,
+      "epoch": 0.66316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.006768192171516e-06,
-      "loss": 0.844,
+      "learning_rate": 5.383532938373453e-06,
+      "loss": 1.0063,
       "step": 23370
     },
     {
-      "epoch": 0.6418664689242262,
+      "epoch": 0.6631952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 6.005952690590681e-06,
-      "loss": 0.7816,
+      "learning_rate": 5.382717676919338e-06,
+      "loss": 0.8708,
       "step": 23371
     },
     {
-      "epoch": 0.6418939331520694,
+      "epoch": 0.6632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 6.005137220612292e-06,
-      "loss": 0.7901,
+      "learning_rate": 5.381902454467612e-06,
+      "loss": 0.8569,
       "step": 23372
     },
     {
-      "epoch": 0.6419213973799127,
+      "epoch": 0.6632519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 6.004321782242804e-06,
-      "loss": 0.8663,
+      "learning_rate": 5.381087271025164e-06,
+      "loss": 0.8683,
       "step": 23373
     },
     {
-      "epoch": 0.6419488616077559,
+      "epoch": 0.6632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 6.00350637548867e-06,
-      "loss": 0.8488,
+      "learning_rate": 5.380272126598879e-06,
+      "loss": 0.7867,
       "step": 23374
     },
     {
-      "epoch": 0.6419763258355992,
+      "epoch": 0.6633087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 6.002691000356345e-06,
-      "loss": 0.822,
+      "learning_rate": 5.379457021195642e-06,
+      "loss": 0.8553,
       "step": 23375
     },
     {
-      "epoch": 0.6420037900634423,
+      "epoch": 0.6633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 6.001875656852272e-06,
-      "loss": 0.8452,
+      "learning_rate": 5.378641954822345e-06,
+      "loss": 0.8885,
       "step": 23376
     },
     {
-      "epoch": 0.6420312542912856,
+      "epoch": 0.6633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 6.001060344982914e-06,
-      "loss": 0.8016,
+      "learning_rate": 5.377826927485863e-06,
+      "loss": 0.9333,
       "step": 23377
     },
     {
-      "epoch": 0.6420587185191289,
+      "epoch": 0.663393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 6.00024506475471e-06,
-      "loss": 0.8484,
+      "learning_rate": 5.377011939193084e-06,
+      "loss": 0.9035,
       "step": 23378
     },
     {
-      "epoch": 0.642086182746972,
+      "epoch": 0.663422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 5.999429816174117e-06,
-      "loss": 0.8331,
+      "learning_rate": 5.376196989950897e-06,
+      "loss": 0.842,
       "step": 23379
     },
     {
-      "epoch": 0.6421136469748153,
+      "epoch": 0.6634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 5.9986145992475875e-06,
-      "loss": 0.9247,
+      "learning_rate": 5.375382079766177e-06,
+      "loss": 0.9102,
       "step": 23380
     },
     {
-      "epoch": 0.6421411112026585,
+      "epoch": 0.6634790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 5.997799413981566e-06,
-      "loss": 0.8348,
+      "learning_rate": 5.374567208645813e-06,
+      "loss": 0.785,
       "step": 23381
     },
     {
-      "epoch": 0.6421685754305018,
+      "epoch": 0.6635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 5.996984260382508e-06,
-      "loss": 0.793,
+      "learning_rate": 5.3737523765966925e-06,
+      "loss": 0.8593,
       "step": 23382
     },
     {
-      "epoch": 0.642196039658345,
+      "epoch": 0.6635357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 5.996169138456864e-06,
-      "loss": 0.8998,
+      "learning_rate": 5.37293758362569e-06,
+      "loss": 0.7912,
       "step": 23383
     },
     {
-      "epoch": 0.6422235038861882,
+      "epoch": 0.6635641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 5.9953540482110785e-06,
-      "loss": 0.7698,
+      "learning_rate": 5.37212282973969e-06,
+      "loss": 0.8473,
       "step": 23384
     },
     {
-      "epoch": 0.6422509681140315,
+      "epoch": 0.6635925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 5.994538989651602e-06,
-      "loss": 0.8078,
+      "learning_rate": 5.371308114945581e-06,
+      "loss": 0.6895,
       "step": 23385
     },
     {
-      "epoch": 0.6422784323418748,
+      "epoch": 0.6636208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 5.993723962784883e-06,
-      "loss": 0.8247,
+      "learning_rate": 5.370493439250237e-06,
+      "loss": 0.8733,
       "step": 23386
     },
     {
-      "epoch": 0.6423058965697179,
+      "epoch": 0.6636492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 5.992908967617374e-06,
-      "loss": 0.8725,
+      "learning_rate": 5.369678802660544e-06,
+      "loss": 0.8501,
       "step": 23387
     },
     {
-      "epoch": 0.6423333607975612,
+      "epoch": 0.663677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 5.99209400415552e-06,
-      "loss": 0.9111,
+      "learning_rate": 5.368864205183385e-06,
+      "loss": 0.8757,
       "step": 23388
     },
     {
-      "epoch": 0.6423608250254044,
+      "epoch": 0.663706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 5.991279072405775e-06,
-      "loss": 0.7969,
+      "learning_rate": 5.36804964682563e-06,
+      "loss": 0.8329,
       "step": 23389
     },
     {
-      "epoch": 0.6423882892532476,
+      "epoch": 0.6637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 5.990464172374578e-06,
-      "loss": 0.9941,
+      "learning_rate": 5.367235127594177e-06,
+      "loss": 0.7396,
       "step": 23390
     },
     {
-      "epoch": 0.6424157534810909,
+      "epoch": 0.6637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 5.989649304068385e-06,
-      "loss": 0.7912,
+      "learning_rate": 5.366420647495892e-06,
+      "loss": 0.833,
       "step": 23391
     },
     {
-      "epoch": 0.6424432177089341,
+      "epoch": 0.6637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 5.988834467493637e-06,
-      "loss": 0.832,
+      "learning_rate": 5.3656062065376616e-06,
+      "loss": 0.9665,
       "step": 23392
     },
     {
-      "epoch": 0.6424706819367774,
+      "epoch": 0.6638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 5.988019662656786e-06,
-      "loss": 0.9095,
+      "learning_rate": 5.364791804726366e-06,
+      "loss": 0.9064,
       "step": 23393
     },
     {
-      "epoch": 0.6424981461646205,
+      "epoch": 0.6638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 5.98720488956428e-06,
-      "loss": 0.8109,
+      "learning_rate": 5.36397744206888e-06,
+      "loss": 0.773,
       "step": 23394
     },
     {
-      "epoch": 0.6425256103924638,
+      "epoch": 0.6638762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 5.986390148222558e-06,
-      "loss": 0.7795,
+      "learning_rate": 5.3631631185720835e-06,
+      "loss": 0.9017,
       "step": 23395
     },
     {
-      "epoch": 0.6425530746203071,
+      "epoch": 0.6639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 5.985575438638076e-06,
-      "loss": 0.8747,
+      "learning_rate": 5.362348834242861e-06,
+      "loss": 0.7392,
       "step": 23396
     },
     {
-      "epoch": 0.6425805388481503,
+      "epoch": 0.663933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 5.984760760817274e-06,
-      "loss": 0.8424,
+      "learning_rate": 5.361534589088083e-06,
+      "loss": 0.8293,
       "step": 23397
     },
     {
-      "epoch": 0.6426080030759935,
+      "epoch": 0.6639614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 5.9839461147666e-06,
-      "loss": 0.8314,
+      "learning_rate": 5.360720383114631e-06,
+      "loss": 0.888,
       "step": 23398
     },
     {
-      "epoch": 0.6426354673038368,
+      "epoch": 0.6639897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 5.983131500492503e-06,
-      "loss": 0.85,
+      "learning_rate": 5.3599062163293855e-06,
+      "loss": 0.8945,
       "step": 23399
     },
     {
-      "epoch": 0.64266293153168,
+      "epoch": 0.6640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 5.982316918001425e-06,
-      "loss": 0.9598,
+      "learning_rate": 5.359092088739218e-06,
+      "loss": 0.9214,
       "step": 23400
     },
     {
-      "epoch": 0.6426903957595232,
+      "epoch": 0.6640465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 5.98150236729981e-06,
-      "loss": 0.8722,
+      "learning_rate": 5.358278000351007e-06,
+      "loss": 0.8715,
       "step": 23401
     },
     {
-      "epoch": 0.6427178599873664,
+      "epoch": 0.6640749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 5.980687848394104e-06,
-      "loss": 0.826,
+      "learning_rate": 5.357463951171635e-06,
+      "loss": 0.8377,
       "step": 23402
     },
     {
-      "epoch": 0.6427453242152097,
+      "epoch": 0.6641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 5.9798733612907554e-06,
-      "loss": 0.8311,
+      "learning_rate": 5.35664994120797e-06,
+      "loss": 0.7698,
       "step": 23403
     },
     {
-      "epoch": 0.642772788443053,
+      "epoch": 0.6641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 5.9790589059962035e-06,
-      "loss": 0.9034,
+      "learning_rate": 5.355835970466891e-06,
+      "loss": 0.8427,
       "step": 23404
     },
     {
-      "epoch": 0.6428002526708961,
+      "epoch": 0.6641600454029511,
       "grad_norm": 0.0,
-      "learning_rate": 5.978244482516895e-06,
-      "loss": 0.8871,
+      "learning_rate": 5.3550220389552745e-06,
+      "loss": 0.7957,
       "step": 23405
     },
     {
-      "epoch": 0.6428277168987394,
+      "epoch": 0.6641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 5.977430090859278e-06,
-      "loss": 0.8803,
+      "learning_rate": 5.3542081466799955e-06,
+      "loss": 0.8273,
       "step": 23406
     },
     {
-      "epoch": 0.6428551811265826,
+      "epoch": 0.664216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 5.976615731029789e-06,
-      "loss": 0.8802,
+      "learning_rate": 5.353394293647933e-06,
+      "loss": 0.8639,
       "step": 23407
     },
     {
-      "epoch": 0.6428826453544259,
+      "epoch": 0.6642451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 5.9758014030348755e-06,
-      "loss": 0.7999,
+      "learning_rate": 5.352580479865954e-06,
+      "loss": 0.8428,
       "step": 23408
     },
     {
-      "epoch": 0.6429101095822691,
+      "epoch": 0.6642735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 5.9749871068809795e-06,
-      "loss": 0.7568,
+      "learning_rate": 5.3517667053409375e-06,
+      "loss": 0.8459,
       "step": 23409
     },
     {
-      "epoch": 0.6429375738101123,
+      "epoch": 0.6643019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 5.974172842574544e-06,
-      "loss": 0.6804,
+      "learning_rate": 5.350952970079759e-06,
+      "loss": 0.7951,
       "step": 23410
     },
     {
-      "epoch": 0.6429650380379556,
+      "epoch": 0.6643303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 5.973358610122014e-06,
-      "loss": 0.9027,
+      "learning_rate": 5.350139274089288e-06,
+      "loss": 0.8503,
       "step": 23411
     },
     {
-      "epoch": 0.6429925022657988,
+      "epoch": 0.6643586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 5.972544409529833e-06,
-      "loss": 0.7856,
+      "learning_rate": 5.3493256173763976e-06,
+      "loss": 0.7296,
       "step": 23412
     },
     {
-      "epoch": 0.643019966493642,
+      "epoch": 0.6643870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 5.971730240804436e-06,
-      "loss": 0.8857,
+      "learning_rate": 5.348511999947968e-06,
+      "loss": 0.8829,
       "step": 23413
     },
     {
-      "epoch": 0.6430474307214853,
+      "epoch": 0.6644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 5.970916103952271e-06,
-      "loss": 0.7468,
+      "learning_rate": 5.347698421810861e-06,
+      "loss": 0.7579,
       "step": 23414
     },
     {
-      "epoch": 0.6430748949493285,
+      "epoch": 0.6644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 5.9701019989797804e-06,
-      "loss": 0.9416,
+      "learning_rate": 5.3468848829719555e-06,
+      "loss": 0.842,
       "step": 23415
     },
     {
-      "epoch": 0.6431023591771717,
+      "epoch": 0.664472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 5.969287925893401e-06,
-      "loss": 0.8809,
+      "learning_rate": 5.346071383438127e-06,
+      "loss": 0.8282,
       "step": 23416
     },
     {
-      "epoch": 0.643129823405015,
+      "epoch": 0.6645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 5.9684738846995775e-06,
-      "loss": 0.9323,
+      "learning_rate": 5.3452579232162384e-06,
+      "loss": 0.7068,
       "step": 23417
     },
     {
-      "epoch": 0.6431572876328582,
+      "epoch": 0.6645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 5.9676598754047546e-06,
-      "loss": 0.9099,
+      "learning_rate": 5.344444502313164e-06,
+      "loss": 0.7573,
       "step": 23418
     },
     {
-      "epoch": 0.6431847518607015,
+      "epoch": 0.6645573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 5.966845898015363e-06,
-      "loss": 0.7324,
+      "learning_rate": 5.343631120735778e-06,
+      "loss": 0.7081,
       "step": 23419
     },
     {
-      "epoch": 0.6432122160885446,
+      "epoch": 0.6645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 5.966031952537853e-06,
-      "loss": 0.8083,
+      "learning_rate": 5.342817778490949e-06,
+      "loss": 0.7942,
       "step": 23420
     },
     {
-      "epoch": 0.6432396803163879,
+      "epoch": 0.6646140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 5.965218038978657e-06,
-      "loss": 0.8592,
+      "learning_rate": 5.3420044755855514e-06,
+      "loss": 0.9412,
       "step": 23421
     },
     {
-      "epoch": 0.6432671445442312,
+      "epoch": 0.6646424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 5.964404157344221e-06,
-      "loss": 0.9001,
+      "learning_rate": 5.3411912120264474e-06,
+      "loss": 0.9532,
       "step": 23422
     },
     {
-      "epoch": 0.6432946087720743,
+      "epoch": 0.6646708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 5.9635903076409825e-06,
-      "loss": 0.917,
+      "learning_rate": 5.340377987820511e-06,
+      "loss": 0.8372,
       "step": 23423
     },
     {
-      "epoch": 0.6433220729999176,
+      "epoch": 0.6646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 5.962776489875384e-06,
-      "loss": 0.9069,
+      "learning_rate": 5.339564802974615e-06,
+      "loss": 0.7484,
       "step": 23424
     },
     {
-      "epoch": 0.6433495372277609,
+      "epoch": 0.664727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 5.961962704053858e-06,
-      "loss": 0.877,
+      "learning_rate": 5.33875165749562e-06,
+      "loss": 0.8363,
       "step": 23425
     },
     {
-      "epoch": 0.6433770014556041,
+      "epoch": 0.664755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 5.961148950182847e-06,
-      "loss": 0.8123,
+      "learning_rate": 5.337938551390399e-06,
+      "loss": 0.8187,
       "step": 23426
     },
     {
-      "epoch": 0.6434044656834473,
+      "epoch": 0.6647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 5.960335228268792e-06,
-      "loss": 0.927,
+      "learning_rate": 5.337125484665826e-06,
+      "loss": 0.6669,
       "step": 23427
     },
     {
-      "epoch": 0.6434319299112905,
+      "epoch": 0.6648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 5.959521538318127e-06,
-      "loss": 0.8135,
+      "learning_rate": 5.336312457328759e-06,
+      "loss": 0.8405,
       "step": 23428
     },
     {
-      "epoch": 0.6434593941391338,
+      "epoch": 0.6648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 5.958707880337294e-06,
-      "loss": 0.8134,
+      "learning_rate": 5.33549946938607e-06,
+      "loss": 0.8953,
       "step": 23429
     },
     {
-      "epoch": 0.643486858366977,
+      "epoch": 0.6648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 5.957894254332732e-06,
-      "loss": 0.9001,
+      "learning_rate": 5.334686520844631e-06,
+      "loss": 0.8338,
       "step": 23430
     },
     {
-      "epoch": 0.6435143225948202,
+      "epoch": 0.6648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 5.957080660310872e-06,
-      "loss": 0.8674,
+      "learning_rate": 5.3338736117113e-06,
+      "loss": 0.7547,
       "step": 23431
     },
     {
-      "epoch": 0.6435417868226635,
+      "epoch": 0.6649262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 5.9562670982781605e-06,
-      "loss": 0.8714,
+      "learning_rate": 5.333060741992949e-06,
+      "loss": 0.9165,
       "step": 23432
     },
     {
-      "epoch": 0.6435692510505067,
+      "epoch": 0.6649545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 5.955453568241027e-06,
-      "loss": 0.9387,
+      "learning_rate": 5.332247911696448e-06,
+      "loss": 0.7409,
       "step": 23433
     },
     {
-      "epoch": 0.6435967152783499,
+      "epoch": 0.6649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 5.954640070205911e-06,
-      "loss": 0.8531,
+      "learning_rate": 5.331435120828654e-06,
+      "loss": 0.8332,
       "step": 23434
     },
     {
-      "epoch": 0.6436241795061932,
+      "epoch": 0.6650113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 5.953826604179252e-06,
-      "loss": 0.8001,
+      "learning_rate": 5.330622369396435e-06,
+      "loss": 0.7419,
       "step": 23435
     },
     {
-      "epoch": 0.6436516437340364,
+      "epoch": 0.6650397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 5.953013170167485e-06,
-      "loss": 0.8684,
+      "learning_rate": 5.329809657406664e-06,
+      "loss": 0.7571,
       "step": 23436
     },
     {
-      "epoch": 0.6436791079618797,
+      "epoch": 0.6650681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 5.952199768177044e-06,
-      "loss": 0.8481,
+      "learning_rate": 5.328996984866199e-06,
+      "loss": 0.8026,
       "step": 23437
     },
     {
-      "epoch": 0.6437065721897229,
+      "epoch": 0.6650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 5.951386398214365e-06,
-      "loss": 0.8666,
+      "learning_rate": 5.328184351781905e-06,
+      "loss": 0.6869,
       "step": 23438
     },
     {
-      "epoch": 0.6437340364175661,
+      "epoch": 0.6651248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 5.950573060285887e-06,
-      "loss": 0.8478,
+      "learning_rate": 5.327371758160654e-06,
+      "loss": 0.6095,
       "step": 23439
     },
     {
-      "epoch": 0.6437615006454094,
+      "epoch": 0.6651532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 5.949759754398041e-06,
-      "loss": 0.8527,
+      "learning_rate": 5.326559204009298e-06,
+      "loss": 0.8556,
       "step": 23440
     },
     {
-      "epoch": 0.6437889648732525,
+      "epoch": 0.6651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 5.948946480557266e-06,
-      "loss": 0.8033,
+      "learning_rate": 5.325746689334707e-06,
+      "loss": 0.792,
       "step": 23441
     },
     {
-      "epoch": 0.6438164291010958,
+      "epoch": 0.6652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 5.94813323877e-06,
-      "loss": 0.8578,
+      "learning_rate": 5.324934214143747e-06,
+      "loss": 0.7927,
       "step": 23442
     },
     {
-      "epoch": 0.6438438933289391,
+      "epoch": 0.6652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 5.947320029042668e-06,
-      "loss": 0.9175,
+      "learning_rate": 5.324121778443274e-06,
+      "loss": 0.7566,
       "step": 23443
     },
     {
-      "epoch": 0.6438713575567823,
+      "epoch": 0.665266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 5.946506851381713e-06,
-      "loss": 0.8361,
+      "learning_rate": 5.323309382240156e-06,
+      "loss": 0.8121,
       "step": 23444
     },
     {
-      "epoch": 0.6438988217846255,
+      "epoch": 0.6652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 5.945693705793562e-06,
-      "loss": 0.8338,
+      "learning_rate": 5.3224970255412566e-06,
+      "loss": 0.8728,
       "step": 23445
     },
     {
-      "epoch": 0.6439262860124687,
+      "epoch": 0.6653234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 5.944880592284655e-06,
-      "loss": 0.8393,
+      "learning_rate": 5.3216847083534315e-06,
+      "loss": 0.8006,
       "step": 23446
     },
     {
-      "epoch": 0.643953750240312,
+      "epoch": 0.6653518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 5.9440675108614245e-06,
-      "loss": 0.7396,
+      "learning_rate": 5.32087243068355e-06,
+      "loss": 0.8331,
       "step": 23447
     },
     {
-      "epoch": 0.6439812144681553,
+      "epoch": 0.6653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 5.9432544615303055e-06,
-      "loss": 0.9319,
+      "learning_rate": 5.320060192538465e-06,
+      "loss": 0.8164,
       "step": 23448
     },
     {
-      "epoch": 0.6440086786959984,
+      "epoch": 0.6654086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 5.942441444297724e-06,
-      "loss": 0.9144,
+      "learning_rate": 5.319247993925043e-06,
+      "loss": 0.7165,
       "step": 23449
     },
     {
-      "epoch": 0.6440361429238417,
+      "epoch": 0.6654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 5.941628459170119e-06,
-      "loss": 0.8318,
+      "learning_rate": 5.3184358348501415e-06,
+      "loss": 0.8942,
       "step": 23450
     },
     {
-      "epoch": 0.644063607151685,
+      "epoch": 0.6654653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 5.940815506153922e-06,
-      "loss": 0.7924,
+      "learning_rate": 5.317623715320626e-06,
+      "loss": 0.7505,
       "step": 23451
     },
     {
-      "epoch": 0.6440910713795281,
+      "epoch": 0.6654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 5.9400025852555645e-06,
-      "loss": 0.8761,
+      "learning_rate": 5.31681163534335e-06,
+      "loss": 0.875,
       "step": 23452
     },
     {
-      "epoch": 0.6441185356073714,
+      "epoch": 0.665522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 5.939189696481478e-06,
-      "loss": 0.9294,
+      "learning_rate": 5.315999594925183e-06,
+      "loss": 0.7559,
       "step": 23453
     },
     {
-      "epoch": 0.6441459998352146,
+      "epoch": 0.6655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 5.9383768398381e-06,
-      "loss": 0.9099,
+      "learning_rate": 5.315187594072974e-06,
+      "loss": 0.8406,
       "step": 23454
     },
     {
-      "epoch": 0.6441734640630579,
+      "epoch": 0.6655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 5.937564015331854e-06,
-      "loss": 0.9717,
+      "learning_rate": 5.314375632793584e-06,
+      "loss": 0.776,
       "step": 23455
     },
     {
-      "epoch": 0.6442009282909011,
+      "epoch": 0.6656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 5.936751222969177e-06,
-      "loss": 0.8444,
+      "learning_rate": 5.31356371109388e-06,
+      "loss": 1.0055,
       "step": 23456
     },
     {
-      "epoch": 0.6442283925187443,
+      "epoch": 0.6656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 5.935938462756496e-06,
-      "loss": 0.8575,
+      "learning_rate": 5.312751828980709e-06,
+      "loss": 1.0386,
       "step": 23457
     },
     {
-      "epoch": 0.6442558567465876,
+      "epoch": 0.6656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 5.935125734700244e-06,
-      "loss": 0.883,
+      "learning_rate": 5.311939986460934e-06,
+      "loss": 0.7793,
       "step": 23458
     },
     {
-      "epoch": 0.6442833209744308,
+      "epoch": 0.6656923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 5.934313038806856e-06,
-      "loss": 0.8255,
+      "learning_rate": 5.311128183541416e-06,
+      "loss": 0.8021,
       "step": 23459
     },
     {
-      "epoch": 0.644310785202274,
+      "epoch": 0.6657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 5.933500375082757e-06,
-      "loss": 0.7912,
+      "learning_rate": 5.310316420229007e-06,
+      "loss": 0.8206,
       "step": 23460
     },
     {
-      "epoch": 0.6443382494301173,
+      "epoch": 0.6657491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 5.932687743534375e-06,
-      "loss": 0.8007,
+      "learning_rate": 5.309504696530565e-06,
+      "loss": 0.8969,
       "step": 23461
     },
     {
-      "epoch": 0.6443657136579605,
+      "epoch": 0.66577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.931875144168144e-06,
-      "loss": 0.9153,
+      "learning_rate": 5.30869301245295e-06,
+      "loss": 0.8471,
       "step": 23462
     },
     {
-      "epoch": 0.6443931778858037,
+      "epoch": 0.6658059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 5.931062576990494e-06,
-      "loss": 0.7583,
+      "learning_rate": 5.3078813680030135e-06,
+      "loss": 0.8106,
       "step": 23463
     },
     {
-      "epoch": 0.644420642113647,
+      "epoch": 0.6658342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 5.9302500420078525e-06,
-      "loss": 0.825,
+      "learning_rate": 5.307069763187615e-06,
+      "loss": 0.887,
       "step": 23464
     },
     {
-      "epoch": 0.6444481063414902,
+      "epoch": 0.6658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 5.929437539226651e-06,
-      "loss": 0.9307,
+      "learning_rate": 5.3062581980136115e-06,
+      "loss": 0.9604,
       "step": 23465
     },
     {
-      "epoch": 0.6444755705693335,
+      "epoch": 0.6658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 5.928625068653313e-06,
-      "loss": 0.9352,
+      "learning_rate": 5.305446672487849e-06,
+      "loss": 0.8928,
       "step": 23466
     },
     {
-      "epoch": 0.6445030347971766,
+      "epoch": 0.6659194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 5.927812630294272e-06,
-      "loss": 0.8075,
+      "learning_rate": 5.304635186617198e-06,
+      "loss": 0.8365,
       "step": 23467
     },
     {
-      "epoch": 0.6445304990250199,
+      "epoch": 0.6659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 5.927000224155957e-06,
-      "loss": 0.8323,
+      "learning_rate": 5.3038237404084995e-06,
+      "loss": 0.918,
       "step": 23468
     },
     {
-      "epoch": 0.6445579632528632,
+      "epoch": 0.6659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 5.926187850244789e-06,
-      "loss": 0.8758,
+      "learning_rate": 5.303012333868613e-06,
+      "loss": 0.712,
       "step": 23469
     },
     {
-      "epoch": 0.6445854274807064,
+      "epoch": 0.6660045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 5.9253755085672035e-06,
-      "loss": 0.8237,
+      "learning_rate": 5.302200967004396e-06,
+      "loss": 0.771,
       "step": 23470
     },
     {
-      "epoch": 0.6446128917085496,
+      "epoch": 0.6660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 5.924563199129629e-06,
-      "loss": 0.8491,
+      "learning_rate": 5.301389639822696e-06,
+      "loss": 0.8624,
       "step": 23471
     },
     {
-      "epoch": 0.6446403559363928,
+      "epoch": 0.666061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 5.9237509219384855e-06,
-      "loss": 0.8015,
+      "learning_rate": 5.300578352330367e-06,
+      "loss": 0.8331,
       "step": 23472
     },
     {
-      "epoch": 0.6446678201642361,
+      "epoch": 0.6660896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 5.9229386770002025e-06,
-      "loss": 0.868,
+      "learning_rate": 5.29976710453427e-06,
+      "loss": 0.9041,
       "step": 23473
     },
     {
-      "epoch": 0.6446952843920793,
+      "epoch": 0.6661180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 5.922126464321209e-06,
-      "loss": 0.8126,
+      "learning_rate": 5.298955896441247e-06,
+      "loss": 0.9068,
       "step": 23474
     },
     {
-      "epoch": 0.6447227486199225,
+      "epoch": 0.6661464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 5.921314283907933e-06,
-      "loss": 0.8947,
+      "learning_rate": 5.2981447280581545e-06,
+      "loss": 0.9069,
       "step": 23475
     },
     {
-      "epoch": 0.6447502128477658,
+      "epoch": 0.6661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 5.9205021357667945e-06,
-      "loss": 0.8613,
+      "learning_rate": 5.29733359939185e-06,
+      "loss": 0.8053,
       "step": 23476
     },
     {
-      "epoch": 0.6447776770756091,
+      "epoch": 0.6662031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 5.919690019904228e-06,
-      "loss": 0.7237,
+      "learning_rate": 5.296522510449174e-06,
+      "loss": 0.8699,
       "step": 23477
     },
     {
-      "epoch": 0.6448051413034522,
+      "epoch": 0.6662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 5.91887793632665e-06,
-      "loss": 0.805,
+      "learning_rate": 5.295711461236985e-06,
+      "loss": 0.7173,
       "step": 23478
     },
     {
-      "epoch": 0.6448326055312955,
+      "epoch": 0.6662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 5.9180658850404915e-06,
-      "loss": 0.9022,
+      "learning_rate": 5.2949004517621365e-06,
+      "loss": 0.7578,
       "step": 23479
     },
     {
-      "epoch": 0.6448600697591387,
+      "epoch": 0.6662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 5.917253866052179e-06,
-      "loss": 0.8488,
+      "learning_rate": 5.294089482031471e-06,
+      "loss": 0.8233,
       "step": 23480
     },
     {
-      "epoch": 0.644887533986982,
+      "epoch": 0.666316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 5.916441879368133e-06,
-      "loss": 0.7904,
+      "learning_rate": 5.293278552051843e-06,
+      "loss": 0.8198,
       "step": 23481
     },
     {
-      "epoch": 0.6449149982148252,
+      "epoch": 0.6663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 5.915629924994781e-06,
-      "loss": 0.8813,
+      "learning_rate": 5.292467661830104e-06,
+      "loss": 0.9634,
       "step": 23482
     },
     {
-      "epoch": 0.6449424624426684,
+      "epoch": 0.6663734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 5.914818002938552e-06,
-      "loss": 0.7841,
+      "learning_rate": 5.291656811373102e-06,
+      "loss": 0.8583,
       "step": 23483
     },
     {
-      "epoch": 0.6449699266705117,
+      "epoch": 0.6664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 5.914006113205861e-06,
-      "loss": 0.8079,
+      "learning_rate": 5.290846000687689e-06,
+      "loss": 0.848,
       "step": 23484
     },
     {
-      "epoch": 0.6449973908983548,
+      "epoch": 0.6664301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 5.913194255803137e-06,
-      "loss": 0.8971,
+      "learning_rate": 5.2900352297807085e-06,
+      "loss": 0.7696,
       "step": 23485
     },
     {
-      "epoch": 0.6450248551261981,
+      "epoch": 0.6664585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 5.9123824307368004e-06,
-      "loss": 0.9027,
+      "learning_rate": 5.2892244986590135e-06,
+      "loss": 0.8221,
       "step": 23486
     },
     {
-      "epoch": 0.6450523193540414,
+      "epoch": 0.6664869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 5.911570638013282e-06,
-      "loss": 1.0243,
+      "learning_rate": 5.288413807329453e-06,
+      "loss": 0.8896,
       "step": 23487
     },
     {
-      "epoch": 0.6450797835818846,
+      "epoch": 0.6665153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 5.910758877638997e-06,
-      "loss": 0.8573,
+      "learning_rate": 5.2876031557988695e-06,
+      "loss": 0.8101,
       "step": 23488
     },
     {
-      "epoch": 0.6451072478097278,
+      "epoch": 0.6665437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 5.909947149620376e-06,
-      "loss": 0.8669,
+      "learning_rate": 5.2867925440741145e-06,
+      "loss": 0.7863,
       "step": 23489
     },
     {
-      "epoch": 0.6451347120375711,
+      "epoch": 0.666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 5.909135453963833e-06,
-      "loss": 0.8383,
+      "learning_rate": 5.2859819721620375e-06,
+      "loss": 0.8277,
       "step": 23490
     },
     {
-      "epoch": 0.6451621762654143,
+      "epoch": 0.666600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 5.908323790675796e-06,
-      "loss": 0.7867,
+      "learning_rate": 5.28517144006948e-06,
+      "loss": 0.759,
       "step": 23491
     },
     {
-      "epoch": 0.6451896404932576,
+      "epoch": 0.6666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 5.9075121597626875e-06,
-      "loss": 0.9122,
+      "learning_rate": 5.284360947803291e-06,
+      "loss": 0.7749,
       "step": 23492
     },
     {
-      "epoch": 0.6452171047211007,
+      "epoch": 0.6666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 5.906700561230927e-06,
-      "loss": 0.933,
+      "learning_rate": 5.28355049537032e-06,
+      "loss": 0.7425,
       "step": 23493
     },
     {
-      "epoch": 0.645244568948944,
+      "epoch": 0.6666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 5.905888995086936e-06,
-      "loss": 0.7968,
+      "learning_rate": 5.282740082777407e-06,
+      "loss": 0.9719,
       "step": 23494
     },
     {
-      "epoch": 0.6452720331767873,
+      "epoch": 0.6667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 5.905077461337141e-06,
-      "loss": 0.9022,
+      "learning_rate": 5.2819297100314e-06,
+      "loss": 0.766,
       "step": 23495
     },
     {
-      "epoch": 0.6452994974046304,
+      "epoch": 0.6667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 5.904265959987958e-06,
-      "loss": 0.8556,
+      "learning_rate": 5.281119377139146e-06,
+      "loss": 0.7395,
       "step": 23496
     },
     {
-      "epoch": 0.6453269616324737,
+      "epoch": 0.6667707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 5.903454491045808e-06,
-      "loss": 0.7637,
+      "learning_rate": 5.280309084107488e-06,
+      "loss": 0.8923,
       "step": 23497
     },
     {
-      "epoch": 0.6453544258603169,
+      "epoch": 0.6667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 5.902643054517112e-06,
-      "loss": 0.8046,
+      "learning_rate": 5.279498830943275e-06,
+      "loss": 0.825,
       "step": 23498
     },
     {
-      "epoch": 0.6453818900881602,
+      "epoch": 0.6668274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 5.901831650408294e-06,
-      "loss": 0.8453,
+      "learning_rate": 5.278688617653345e-06,
+      "loss": 0.8633,
       "step": 23499
     },
     {
-      "epoch": 0.6454093543160034,
+      "epoch": 0.6668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 5.901020278725768e-06,
-      "loss": 0.7756,
+      "learning_rate": 5.277878444244542e-06,
+      "loss": 0.8511,
       "step": 23500
     },
     {
-      "epoch": 0.6454368185438466,
+      "epoch": 0.6668842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 5.900208939475963e-06,
-      "loss": 0.8694,
+      "learning_rate": 5.277068310723717e-06,
+      "loss": 0.7997,
       "step": 23501
     },
     {
-      "epoch": 0.6454642827716899,
+      "epoch": 0.6669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 5.899397632665288e-06,
-      "loss": 0.818,
+      "learning_rate": 5.276258217097705e-06,
+      "loss": 0.9385,
       "step": 23502
     },
     {
-      "epoch": 0.6454917469995332,
+      "epoch": 0.6669409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 5.898586358300168e-06,
-      "loss": 0.9157,
+      "learning_rate": 5.2754481633733525e-06,
+      "loss": 0.7237,
       "step": 23503
     },
     {
-      "epoch": 0.6455192112273763,
+      "epoch": 0.6669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 5.897775116387023e-06,
-      "loss": 0.8582,
+      "learning_rate": 5.274638149557505e-06,
+      "loss": 0.8114,
       "step": 23504
     },
     {
-      "epoch": 0.6455466754552196,
+      "epoch": 0.6669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 5.896963906932268e-06,
-      "loss": 0.7684,
+      "learning_rate": 5.273828175656997e-06,
+      "loss": 0.7202,
       "step": 23505
     },
     {
-      "epoch": 0.6455741396830628,
+      "epoch": 0.6670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 5.8961527299423245e-06,
-      "loss": 0.7693,
+      "learning_rate": 5.2730182416786756e-06,
+      "loss": 0.9573,
       "step": 23506
     },
     {
-      "epoch": 0.645601603910906,
+      "epoch": 0.6670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 5.895341585423615e-06,
-      "loss": 0.9294,
+      "learning_rate": 5.272208347629386e-06,
+      "loss": 0.8277,
       "step": 23507
     },
     {
-      "epoch": 0.6456290681387493,
+      "epoch": 0.6670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 5.894530473382549e-06,
-      "loss": 0.9418,
+      "learning_rate": 5.27139849351596e-06,
+      "loss": 0.723,
       "step": 23508
     },
     {
-      "epoch": 0.6456565323665925,
+      "epoch": 0.66711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 5.8937193938255475e-06,
-      "loss": 0.8611,
+      "learning_rate": 5.270588679345244e-06,
+      "loss": 0.7543,
       "step": 23509
     },
     {
-      "epoch": 0.6456839965944358,
+      "epoch": 0.6671396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 5.892908346759028e-06,
-      "loss": 0.757,
+      "learning_rate": 5.269778905124082e-06,
+      "loss": 0.8571,
       "step": 23510
     },
     {
-      "epoch": 0.6457114608222789,
+      "epoch": 0.6671679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 5.892097332189412e-06,
-      "loss": 0.8355,
+      "learning_rate": 5.268969170859306e-06,
+      "loss": 0.8174,
       "step": 23511
     },
     {
-      "epoch": 0.6457389250501222,
+      "epoch": 0.6671963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 5.89128635012311e-06,
-      "loss": 0.809,
+      "learning_rate": 5.268159476557762e-06,
+      "loss": 0.8107,
       "step": 23512
     },
     {
-      "epoch": 0.6457663892779655,
+      "epoch": 0.6672247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 5.890475400566547e-06,
-      "loss": 0.8319,
+      "learning_rate": 5.267349822226284e-06,
+      "loss": 0.873,
       "step": 23513
     },
     {
-      "epoch": 0.6457938535058086,
+      "epoch": 0.6672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 5.88966448352613e-06,
-      "loss": 0.7708,
+      "learning_rate": 5.266540207871717e-06,
+      "loss": 0.902,
       "step": 23514
     },
     {
-      "epoch": 0.6458213177336519,
+      "epoch": 0.6672814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 5.8888535990082785e-06,
-      "loss": 0.8538,
+      "learning_rate": 5.265730633500902e-06,
+      "loss": 0.8715,
       "step": 23515
     },
     {
-      "epoch": 0.6458487819614952,
+      "epoch": 0.6673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 5.8880427470194135e-06,
-      "loss": 0.8897,
+      "learning_rate": 5.264921099120668e-06,
+      "loss": 0.7896,
       "step": 23516
     },
     {
-      "epoch": 0.6458762461893384,
+      "epoch": 0.6673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 5.887231927565944e-06,
-      "loss": 0.8487,
+      "learning_rate": 5.264111604737859e-06,
+      "loss": 0.7962,
       "step": 23517
     },
     {
-      "epoch": 0.6459037104171816,
+      "epoch": 0.667366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 5.886421140654292e-06,
-      "loss": 0.8846,
+      "learning_rate": 5.263302150359316e-06,
+      "loss": 0.8736,
       "step": 23518
     },
     {
-      "epoch": 0.6459311746450248,
+      "epoch": 0.6673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.885610386290868e-06,
-      "loss": 0.907,
+      "learning_rate": 5.262492735991868e-06,
+      "loss": 0.9385,
       "step": 23519
     },
     {
-      "epoch": 0.6459586388728681,
+      "epoch": 0.6674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 5.884799664482088e-06,
-      "loss": 0.8892,
+      "learning_rate": 5.261683361642358e-06,
+      "loss": 0.7918,
       "step": 23520
     },
     {
-      "epoch": 0.6459861031007114,
+      "epoch": 0.6674517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 5.883988975234366e-06,
-      "loss": 0.8565,
+      "learning_rate": 5.260874027317626e-06,
+      "loss": 0.8657,
       "step": 23521
     },
     {
-      "epoch": 0.6460135673285545,
+      "epoch": 0.6674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 5.883178318554117e-06,
-      "loss": 0.8093,
+      "learning_rate": 5.260064733024499e-06,
+      "loss": 0.9034,
       "step": 23522
     },
     {
-      "epoch": 0.6460410315563978,
+      "epoch": 0.6675085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 5.882367694447759e-06,
-      "loss": 0.8232,
+      "learning_rate": 5.259255478769818e-06,
+      "loss": 0.7534,
       "step": 23523
     },
     {
-      "epoch": 0.6460684957842411,
+      "epoch": 0.6675368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 5.881557102921704e-06,
-      "loss": 0.8675,
+      "learning_rate": 5.258446264560424e-06,
+      "loss": 0.8812,
       "step": 23524
     },
     {
-      "epoch": 0.6460959600120842,
+      "epoch": 0.6675652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 5.880746543982361e-06,
-      "loss": 0.832,
+      "learning_rate": 5.257637090403143e-06,
+      "loss": 0.7948,
       "step": 23525
     },
     {
-      "epoch": 0.6461234242399275,
+      "epoch": 0.6675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 5.879936017636148e-06,
-      "loss": 0.783,
+      "learning_rate": 5.256827956304815e-06,
+      "loss": 0.7643,
       "step": 23526
     },
     {
-      "epoch": 0.6461508884677707,
+      "epoch": 0.667622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 5.879125523889475e-06,
-      "loss": 0.8378,
+      "learning_rate": 5.256018862272274e-06,
+      "loss": 0.7836,
       "step": 23527
     },
     {
-      "epoch": 0.646178352695614,
+      "epoch": 0.667650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 5.878315062748761e-06,
-      "loss": 0.812,
+      "learning_rate": 5.255209808312356e-06,
+      "loss": 0.9055,
       "step": 23528
     },
     {
-      "epoch": 0.6462058169234572,
+      "epoch": 0.6676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 5.87750463422041e-06,
-      "loss": 0.8432,
+      "learning_rate": 5.254400794431897e-06,
+      "loss": 0.9042,
       "step": 23529
     },
     {
-      "epoch": 0.6462332811513004,
+      "epoch": 0.6677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 5.876694238310846e-06,
-      "loss": 0.819,
+      "learning_rate": 5.253591820637725e-06,
+      "loss": 0.8842,
       "step": 23530
     },
     {
-      "epoch": 0.6462607453791437,
+      "epoch": 0.6677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 5.875883875026469e-06,
-      "loss": 0.8846,
+      "learning_rate": 5.252782886936675e-06,
+      "loss": 0.8416,
       "step": 23531
     },
     {
-      "epoch": 0.6462882096069869,
+      "epoch": 0.6677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 5.8750735443736995e-06,
-      "loss": 0.8318,
+      "learning_rate": 5.251973993335586e-06,
+      "loss": 0.7511,
       "step": 23532
     },
     {
-      "epoch": 0.6463156738348301,
+      "epoch": 0.6677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 5.8742632463589434e-06,
-      "loss": 0.7697,
+      "learning_rate": 5.251165139841281e-06,
+      "loss": 0.9773,
       "step": 23533
     },
     {
-      "epoch": 0.6463431380626734,
+      "epoch": 0.6678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 5.873452980988614e-06,
-      "loss": 0.934,
+      "learning_rate": 5.2503563264605995e-06,
+      "loss": 0.8447,
       "step": 23534
     },
     {
-      "epoch": 0.6463706022905166,
+      "epoch": 0.6678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 5.872642748269126e-06,
-      "loss": 0.9033,
+      "learning_rate": 5.249547553200374e-06,
+      "loss": 0.829,
       "step": 23535
     },
     {
-      "epoch": 0.6463980665183598,
+      "epoch": 0.6678774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 5.871832548206888e-06,
-      "loss": 0.9113,
+      "learning_rate": 5.248738820067431e-06,
+      "loss": 0.8686,
       "step": 23536
     },
     {
-      "epoch": 0.6464255307462031,
+      "epoch": 0.6679057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 5.871022380808309e-06,
-      "loss": 0.9264,
+      "learning_rate": 5.247930127068604e-06,
+      "loss": 0.8663,
       "step": 23537
     },
     {
-      "epoch": 0.6464529949740463,
+      "epoch": 0.6679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 5.870212246079797e-06,
-      "loss": 0.7737,
+      "learning_rate": 5.247121474210728e-06,
+      "loss": 0.814,
       "step": 23538
     },
     {
-      "epoch": 0.6464804592018896,
+      "epoch": 0.6679625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 5.869402144027767e-06,
-      "loss": 0.9104,
+      "learning_rate": 5.246312861500629e-06,
+      "loss": 0.865,
       "step": 23539
     },
     {
-      "epoch": 0.6465079234297327,
+      "epoch": 0.6679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 5.86859207465863e-06,
-      "loss": 0.8749,
+      "learning_rate": 5.245504288945137e-06,
+      "loss": 0.7403,
       "step": 23540
     },
     {
-      "epoch": 0.646535387657576,
+      "epoch": 0.6680192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 5.867782037978789e-06,
-      "loss": 0.7692,
+      "learning_rate": 5.244695756551088e-06,
+      "loss": 0.775,
       "step": 23541
     },
     {
-      "epoch": 0.6465628518854193,
+      "epoch": 0.6680476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 5.866972033994662e-06,
-      "loss": 0.8335,
+      "learning_rate": 5.2438872643253e-06,
+      "loss": 0.7762,
       "step": 23542
     },
     {
-      "epoch": 0.6465903161132625,
+      "epoch": 0.6680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 5.86616206271265e-06,
-      "loss": 0.8123,
+      "learning_rate": 5.243078812274615e-06,
+      "loss": 0.7492,
       "step": 23543
     },
     {
-      "epoch": 0.6466177803411057,
+      "epoch": 0.6681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 5.865352124139165e-06,
-      "loss": 0.899,
+      "learning_rate": 5.24227040040586e-06,
+      "loss": 0.7792,
       "step": 23544
     },
     {
-      "epoch": 0.6466452445689489,
+      "epoch": 0.6681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 5.864542218280615e-06,
-      "loss": 0.8972,
+      "learning_rate": 5.241462028725858e-06,
+      "loss": 0.8945,
       "step": 23545
     },
     {
-      "epoch": 0.6466727087967922,
+      "epoch": 0.668161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 5.863732345143409e-06,
-      "loss": 0.881,
+      "learning_rate": 5.240653697241439e-06,
+      "loss": 0.8089,
       "step": 23546
     },
     {
-      "epoch": 0.6467001730246354,
+      "epoch": 0.6681895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 5.862922504733958e-06,
-      "loss": 0.8654,
+      "learning_rate": 5.239845405959436e-06,
+      "loss": 0.8364,
       "step": 23547
     },
     {
-      "epoch": 0.6467276372524786,
+      "epoch": 0.6682179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 5.862112697058667e-06,
-      "loss": 0.8787,
+      "learning_rate": 5.23903715488667e-06,
+      "loss": 0.8234,
       "step": 23548
     },
     {
-      "epoch": 0.6467551014803219,
+      "epoch": 0.6682463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 5.861302922123939e-06,
-      "loss": 0.7734,
+      "learning_rate": 5.238228944029971e-06,
+      "loss": 0.8571,
       "step": 23549
     },
     {
-      "epoch": 0.6467825657081652,
+      "epoch": 0.6682746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 5.8604931799361865e-06,
-      "loss": 0.7667,
+      "learning_rate": 5.237420773396169e-06,
+      "loss": 0.7653,
       "step": 23550
     },
     {
-      "epoch": 0.6468100299360083,
+      "epoch": 0.6683030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 5.859683470501818e-06,
-      "loss": 0.8087,
+      "learning_rate": 5.236612642992085e-06,
+      "loss": 0.8765,
       "step": 23551
     },
     {
-      "epoch": 0.6468374941638516,
+      "epoch": 0.6683314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 5.8588737938272344e-06,
-      "loss": 0.8013,
+      "learning_rate": 5.235804552824548e-06,
+      "loss": 0.7989,
       "step": 23552
     },
     {
-      "epoch": 0.6468649583916948,
+      "epoch": 0.6683598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 5.858064149918845e-06,
-      "loss": 0.8398,
+      "learning_rate": 5.234996502900387e-06,
+      "loss": 0.9182,
       "step": 23553
     },
     {
-      "epoch": 0.646892422619538,
+      "epoch": 0.6683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 5.8572545387830615e-06,
-      "loss": 0.8438,
+      "learning_rate": 5.234188493226422e-06,
+      "loss": 0.8865,
       "step": 23554
     },
     {
-      "epoch": 0.6469198868473813,
+      "epoch": 0.668416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 5.856444960426282e-06,
-      "loss": 0.705,
+      "learning_rate": 5.2333805238094835e-06,
+      "loss": 0.9138,
       "step": 23555
     },
     {
-      "epoch": 0.6469473510752245,
+      "epoch": 0.6684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 5.855635414854914e-06,
-      "loss": 0.8871,
+      "learning_rate": 5.232572594656392e-06,
+      "loss": 0.7136,
       "step": 23556
     },
     {
-      "epoch": 0.6469748153030678,
+      "epoch": 0.6684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 5.854825902075365e-06,
-      "loss": 0.8793,
+      "learning_rate": 5.231764705773973e-06,
+      "loss": 0.8262,
       "step": 23557
     },
     {
-      "epoch": 0.6470022795309109,
+      "epoch": 0.6685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 5.8540164220940374e-06,
-      "loss": 0.897,
+      "learning_rate": 5.230956857169052e-06,
+      "loss": 0.852,
       "step": 23558
     },
     {
-      "epoch": 0.6470297437587542,
+      "epoch": 0.6685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 5.85320697491734e-06,
-      "loss": 0.8077,
+      "learning_rate": 5.230149048848453e-06,
+      "loss": 0.8658,
       "step": 23559
     },
     {
-      "epoch": 0.6470572079865975,
+      "epoch": 0.6685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 5.852397560551678e-06,
-      "loss": 0.9038,
+      "learning_rate": 5.229341280818999e-06,
+      "loss": 0.9817,
       "step": 23560
     },
     {
-      "epoch": 0.6470846722144407,
+      "epoch": 0.6685868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 5.851588179003449e-06,
-      "loss": 0.8554,
+      "learning_rate": 5.228533553087518e-06,
+      "loss": 0.8883,
       "step": 23561
     },
     {
-      "epoch": 0.6471121364422839,
+      "epoch": 0.6686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 5.8507788302790605e-06,
-      "loss": 0.8473,
+      "learning_rate": 5.227725865660825e-06,
+      "loss": 0.7903,
       "step": 23562
     },
     {
-      "epoch": 0.6471396006701272,
+      "epoch": 0.6686435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 5.84996951438492e-06,
-      "loss": 0.8072,
+      "learning_rate": 5.2269182185457456e-06,
+      "loss": 0.8203,
       "step": 23563
     },
     {
-      "epoch": 0.6471670648979704,
+      "epoch": 0.668671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 5.849160231327427e-06,
-      "loss": 0.8324,
+      "learning_rate": 5.226110611749107e-06,
+      "loss": 0.898,
       "step": 23564
     },
     {
-      "epoch": 0.6471945291258137,
+      "epoch": 0.6687003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 5.848350981112985e-06,
-      "loss": 0.8897,
+      "learning_rate": 5.225303045277721e-06,
+      "loss": 0.7175,
       "step": 23565
     },
     {
-      "epoch": 0.6472219933536568,
+      "epoch": 0.6687287173666289,
       "grad_norm": 0.0,
-      "learning_rate": 5.8475417637480016e-06,
-      "loss": 0.8565,
+      "learning_rate": 5.224495519138417e-06,
+      "loss": 0.9117,
       "step": 23566
     },
     {
-      "epoch": 0.6472494575815001,
+      "epoch": 0.6687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 5.846732579238873e-06,
-      "loss": 0.7926,
+      "learning_rate": 5.223688033338017e-06,
+      "loss": 0.872,
       "step": 23567
     },
     {
-      "epoch": 0.6472769218093434,
+      "epoch": 0.6687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 5.845923427592006e-06,
-      "loss": 0.8748,
+      "learning_rate": 5.222880587883335e-06,
+      "loss": 0.833,
       "step": 23568
     },
     {
-      "epoch": 0.6473043860371865,
+      "epoch": 0.6688138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 5.845114308813802e-06,
-      "loss": 0.8222,
+      "learning_rate": 5.2220731827811955e-06,
+      "loss": 0.7766,
       "step": 23569
     },
     {
-      "epoch": 0.6473318502650298,
+      "epoch": 0.6688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 5.844305222910663e-06,
-      "loss": 0.8184,
+      "learning_rate": 5.221265818038422e-06,
+      "loss": 0.8368,
       "step": 23570
     },
     {
-      "epoch": 0.647359314492873,
+      "epoch": 0.6688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 5.843496169888992e-06,
-      "loss": 0.7666,
+      "learning_rate": 5.220458493661829e-06,
+      "loss": 0.9729,
       "step": 23571
     },
     {
-      "epoch": 0.6473867787207163,
+      "epoch": 0.6688989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 5.842687149755191e-06,
-      "loss": 0.9343,
+      "learning_rate": 5.219651209658235e-06,
+      "loss": 0.8012,
       "step": 23572
     },
     {
-      "epoch": 0.6474142429485595,
+      "epoch": 0.6689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 5.841878162515655e-06,
-      "loss": 0.7832,
+      "learning_rate": 5.218843966034464e-06,
+      "loss": 0.7401,
       "step": 23573
     },
     {
-      "epoch": 0.6474417071764027,
+      "epoch": 0.668955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 5.841069208176791e-06,
-      "loss": 0.8468,
+      "learning_rate": 5.2180367627973324e-06,
+      "loss": 0.8134,
       "step": 23574
     },
     {
-      "epoch": 0.647469171404246,
+      "epoch": 0.6689841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 5.840260286745e-06,
-      "loss": 0.9189,
+      "learning_rate": 5.2172295999536635e-06,
+      "loss": 0.8592,
       "step": 23575
     },
     {
-      "epoch": 0.6474966356320893,
+      "epoch": 0.6690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 5.8394513982266786e-06,
-      "loss": 0.8352,
+      "learning_rate": 5.216422477510267e-06,
+      "loss": 0.8763,
       "step": 23576
     },
     {
-      "epoch": 0.6475240998599324,
+      "epoch": 0.6690408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 5.838642542628235e-06,
-      "loss": 0.8016,
+      "learning_rate": 5.2156153954739655e-06,
+      "loss": 0.8539,
       "step": 23577
     },
     {
-      "epoch": 0.6475515640877757,
+      "epoch": 0.6690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 5.837833719956058e-06,
-      "loss": 0.9012,
+      "learning_rate": 5.2148083538515796e-06,
+      "loss": 0.8524,
       "step": 23578
     },
     {
-      "epoch": 0.6475790283156189,
+      "epoch": 0.6690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 5.837024930216555e-06,
-      "loss": 0.7736,
+      "learning_rate": 5.214001352649918e-06,
+      "loss": 0.9075,
       "step": 23579
     },
     {
-      "epoch": 0.6476064925434621,
+      "epoch": 0.6691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 5.8362161734161204e-06,
-      "loss": 0.9739,
+      "learning_rate": 5.213194391875802e-06,
+      "loss": 0.7653,
       "step": 23580
     },
     {
-      "epoch": 0.6476339567713054,
+      "epoch": 0.6691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 5.835407449561158e-06,
-      "loss": 0.9454,
+      "learning_rate": 5.212387471536053e-06,
+      "loss": 0.8388,
       "step": 23581
     },
     {
-      "epoch": 0.6476614209991486,
+      "epoch": 0.6691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 5.834598758658066e-06,
-      "loss": 0.8654,
+      "learning_rate": 5.211580591637478e-06,
+      "loss": 0.733,
       "step": 23582
     },
     {
-      "epoch": 0.6476888852269919,
+      "epoch": 0.669211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 5.833790100713243e-06,
-      "loss": 0.7334,
+      "learning_rate": 5.210773752186895e-06,
+      "loss": 0.7982,
       "step": 23583
     },
     {
-      "epoch": 0.647716349454835,
+      "epoch": 0.6692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 5.832981475733085e-06,
-      "loss": 0.9142,
+      "learning_rate": 5.209966953191128e-06,
+      "loss": 0.9126,
       "step": 23584
     },
     {
-      "epoch": 0.6477438136826783,
+      "epoch": 0.6692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 5.83217288372399e-06,
-      "loss": 0.9122,
+      "learning_rate": 5.20916019465698e-06,
+      "loss": 0.8473,
       "step": 23585
     },
     {
-      "epoch": 0.6477712779105216,
+      "epoch": 0.6692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 5.831364324692363e-06,
-      "loss": 0.8912,
+      "learning_rate": 5.208353476591272e-06,
+      "loss": 0.827,
       "step": 23586
     },
     {
-      "epoch": 0.6477987421383647,
+      "epoch": 0.6693246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 5.830555798644593e-06,
-      "loss": 0.8438,
+      "learning_rate": 5.207546799000822e-06,
+      "loss": 0.7996,
       "step": 23587
     },
     {
-      "epoch": 0.647826206366208,
+      "epoch": 0.6693530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 5.829747305587079e-06,
-      "loss": 0.7606,
+      "learning_rate": 5.206740161892432e-06,
+      "loss": 0.7912,
       "step": 23588
     },
     {
-      "epoch": 0.6478536705940513,
+      "epoch": 0.6693813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 5.828938845526224e-06,
-      "loss": 0.7704,
+      "learning_rate": 5.205933565272929e-06,
+      "loss": 0.8461,
       "step": 23589
     },
     {
-      "epoch": 0.6478811348218945,
+      "epoch": 0.6694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 5.828130418468416e-06,
-      "loss": 0.8988,
+      "learning_rate": 5.205127009149118e-06,
+      "loss": 0.8777,
       "step": 23590
     },
     {
-      "epoch": 0.6479085990497377,
+      "epoch": 0.6694381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 5.827322024420057e-06,
-      "loss": 0.8621,
+      "learning_rate": 5.204320493527815e-06,
+      "loss": 0.9042,
       "step": 23591
     },
     {
-      "epoch": 0.6479360632775809,
+      "epoch": 0.669466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 5.826513663387543e-06,
-      "loss": 0.8554,
+      "learning_rate": 5.2035140184158365e-06,
+      "loss": 0.7756,
       "step": 23592
     },
     {
-      "epoch": 0.6479635275054242,
+      "epoch": 0.669494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 5.825705335377272e-06,
-      "loss": 0.8661,
+      "learning_rate": 5.202707583819986e-06,
+      "loss": 0.7926,
       "step": 23593
     },
     {
-      "epoch": 0.6479909917332675,
+      "epoch": 0.6695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 5.824897040395634e-06,
-      "loss": 0.7342,
+      "learning_rate": 5.201901189747081e-06,
+      "loss": 0.868,
       "step": 23594
     },
     {
-      "epoch": 0.6480184559611106,
+      "epoch": 0.6695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.824088778449033e-06,
-      "loss": 0.784,
+      "learning_rate": 5.201094836203935e-06,
+      "loss": 0.8756,
       "step": 23595
     },
     {
-      "epoch": 0.6480459201889539,
+      "epoch": 0.6695800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 5.823280549543854e-06,
-      "loss": 0.8,
+      "learning_rate": 5.200288523197354e-06,
+      "loss": 0.7624,
       "step": 23596
     },
     {
-      "epoch": 0.6480733844167971,
+      "epoch": 0.6696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 5.822472353686497e-06,
-      "loss": 0.8236,
+      "learning_rate": 5.199482250734151e-06,
+      "loss": 0.8109,
       "step": 23597
     },
     {
-      "epoch": 0.6481008486446403,
+      "epoch": 0.6696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 5.821664190883358e-06,
-      "loss": 0.8972,
+      "learning_rate": 5.198676018821142e-06,
+      "loss": 0.7811,
       "step": 23598
     },
     {
-      "epoch": 0.6481283128724836,
+      "epoch": 0.6696651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 5.82085606114083e-06,
-      "loss": 0.8357,
+      "learning_rate": 5.1978698274651276e-06,
+      "loss": 0.8859,
       "step": 23599
     },
     {
-      "epoch": 0.6481557771003268,
+      "epoch": 0.6696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 5.8200479644653075e-06,
-      "loss": 0.7825,
+      "learning_rate": 5.1970636766729225e-06,
+      "loss": 0.9515,
       "step": 23600
     },
     {
-      "epoch": 0.6481832413281701,
+      "epoch": 0.6697219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 5.81923990086319e-06,
-      "loss": 0.8107,
+      "learning_rate": 5.196257566451342e-06,
+      "loss": 0.7103,
       "step": 23601
     },
     {
-      "epoch": 0.6482107055560133,
+      "epoch": 0.6697502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 5.818431870340864e-06,
-      "loss": 0.9489,
+      "learning_rate": 5.195451496807184e-06,
+      "loss": 0.7103,
       "step": 23602
     },
     {
-      "epoch": 0.6482381697838565,
+      "epoch": 0.6697786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 5.81762387290472e-06,
-      "loss": 0.7976,
+      "learning_rate": 5.1946454677472635e-06,
+      "loss": 0.9054,
       "step": 23603
     },
     {
-      "epoch": 0.6482656340116998,
+      "epoch": 0.6698070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 5.816815908561159e-06,
-      "loss": 0.8774,
+      "learning_rate": 5.193839479278388e-06,
+      "loss": 0.8771,
       "step": 23604
     },
     {
-      "epoch": 0.648293098239543,
+      "epoch": 0.6698354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 5.816007977316569e-06,
-      "loss": 0.8884,
+      "learning_rate": 5.1930335314073685e-06,
+      "loss": 0.8892,
       "step": 23605
     },
     {
-      "epoch": 0.6483205624673862,
+      "epoch": 0.6698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 5.815200079177346e-06,
-      "loss": 0.8441,
+      "learning_rate": 5.192227624141014e-06,
+      "loss": 0.85,
       "step": 23606
     },
     {
-      "epoch": 0.6483480266952295,
+      "epoch": 0.6698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 5.814392214149883e-06,
-      "loss": 0.8762,
+      "learning_rate": 5.191421757486125e-06,
+      "loss": 0.9474,
       "step": 23607
     },
     {
-      "epoch": 0.6483754909230727,
+      "epoch": 0.6699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 5.813584382240567e-06,
-      "loss": 0.8675,
+      "learning_rate": 5.190615931449512e-06,
+      "loss": 0.8403,
       "step": 23608
     },
     {
-      "epoch": 0.648402955150916,
+      "epoch": 0.6699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 5.8127765834557934e-06,
-      "loss": 0.8322,
+      "learning_rate": 5.189810146037987e-06,
+      "loss": 0.8329,
       "step": 23609
     },
     {
-      "epoch": 0.6484304193787591,
+      "epoch": 0.6699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 5.8119688178019575e-06,
-      "loss": 0.7731,
+      "learning_rate": 5.1890044012583476e-06,
+      "loss": 0.8391,
       "step": 23610
     },
     {
-      "epoch": 0.6484578836066024,
+      "epoch": 0.670005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 5.811161085285441e-06,
-      "loss": 0.8047,
+      "learning_rate": 5.188198697117403e-06,
+      "loss": 0.9051,
       "step": 23611
     },
     {
-      "epoch": 0.6484853478344457,
+      "epoch": 0.6700340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 5.810353385912643e-06,
-      "loss": 0.8793,
+      "learning_rate": 5.187393033621967e-06,
+      "loss": 0.8118,
       "step": 23612
     },
     {
-      "epoch": 0.6485128120622888,
+      "epoch": 0.6700624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 5.809545719689955e-06,
-      "loss": 0.9055,
+      "learning_rate": 5.1865874107788336e-06,
+      "loss": 0.7512,
       "step": 23613
     },
     {
-      "epoch": 0.6485402762901321,
+      "epoch": 0.6700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 5.80873808662376e-06,
-      "loss": 0.7745,
+      "learning_rate": 5.185781828594811e-06,
+      "loss": 0.886,
       "step": 23614
     },
     {
-      "epoch": 0.6485677405179754,
+      "epoch": 0.6701191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 5.807930486720455e-06,
-      "loss": 0.8666,
+      "learning_rate": 5.184976287076712e-06,
+      "loss": 0.8412,
       "step": 23615
     },
     {
-      "epoch": 0.6485952047458186,
+      "epoch": 0.6701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 5.8071229199864254e-06,
-      "loss": 0.7603,
+      "learning_rate": 5.184170786231328e-06,
+      "loss": 0.8605,
       "step": 23616
     },
     {
-      "epoch": 0.6486226689736618,
+      "epoch": 0.6701759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 5.806315386428067e-06,
-      "loss": 0.9146,
+      "learning_rate": 5.183365326065472e-06,
+      "loss": 0.8742,
       "step": 23617
     },
     {
-      "epoch": 0.648650133201505,
+      "epoch": 0.6702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 5.805507886051764e-06,
-      "loss": 0.8395,
+      "learning_rate": 5.18255990658595e-06,
+      "loss": 0.8434,
       "step": 23618
     },
     {
-      "epoch": 0.6486775974293483,
+      "epoch": 0.6702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 5.8047004188639085e-06,
-      "loss": 0.9335,
+      "learning_rate": 5.181754527799552e-06,
+      "loss": 0.9024,
       "step": 23619
     },
     {
-      "epoch": 0.6487050616571916,
+      "epoch": 0.6702610669693531,
       "grad_norm": 0.0,
-      "learning_rate": 5.803892984870887e-06,
-      "loss": 0.891,
+      "learning_rate": 5.180949189713098e-06,
+      "loss": 0.9323,
       "step": 23620
     },
     {
-      "epoch": 0.6487325258850347,
+      "epoch": 0.6702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.803085584079089e-06,
-      "loss": 0.7184,
+      "learning_rate": 5.1801438923333805e-06,
+      "loss": 0.8898,
       "step": 23621
     },
     {
-      "epoch": 0.648759990112878,
+      "epoch": 0.6703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 5.802278216494903e-06,
-      "loss": 0.7976,
+      "learning_rate": 5.179338635667201e-06,
+      "loss": 0.8032,
       "step": 23622
     },
     {
-      "epoch": 0.6487874543407212,
+      "epoch": 0.6703461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 5.801470882124718e-06,
-      "loss": 0.7535,
+      "learning_rate": 5.178533419721371e-06,
+      "loss": 0.7794,
       "step": 23623
     },
     {
-      "epoch": 0.6488149185685644,
+      "epoch": 0.6703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 5.800663580974921e-06,
-      "loss": 0.9549,
+      "learning_rate": 5.177728244502682e-06,
+      "loss": 0.9131,
       "step": 23624
     },
     {
-      "epoch": 0.6488423827964077,
+      "epoch": 0.6704029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 5.799856313051906e-06,
-      "loss": 0.7313,
+      "learning_rate": 5.176923110017938e-06,
+      "loss": 0.8162,
       "step": 23625
     },
     {
-      "epoch": 0.6488698470242509,
+      "epoch": 0.6704313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 5.799049078362054e-06,
-      "loss": 0.8612,
+      "learning_rate": 5.176118016273947e-06,
+      "loss": 0.8331,
       "step": 23626
     },
     {
-      "epoch": 0.6488973112520942,
+      "epoch": 0.6704597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 5.79824187691175e-06,
-      "loss": 0.8461,
+      "learning_rate": 5.175312963277499e-06,
+      "loss": 0.9911,
       "step": 23627
     },
     {
-      "epoch": 0.6489247754799374,
+      "epoch": 0.6704880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 5.7974347087073825e-06,
-      "loss": 0.8313,
+      "learning_rate": 5.174507951035399e-06,
+      "loss": 0.814,
       "step": 23628
     },
     {
-      "epoch": 0.6489522397077806,
+      "epoch": 0.670516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.7966275737553404e-06,
-      "loss": 0.8047,
+      "learning_rate": 5.173702979554452e-06,
+      "loss": 0.8305,
       "step": 23629
     },
     {
-      "epoch": 0.6489797039356239,
+      "epoch": 0.6705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 5.795820472062008e-06,
-      "loss": 0.9265,
+      "learning_rate": 5.172898048841448e-06,
+      "loss": 0.8704,
       "step": 23630
     },
     {
-      "epoch": 0.649007168163467,
+      "epoch": 0.6705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 5.795013403633778e-06,
-      "loss": 0.8783,
+      "learning_rate": 5.172093158903191e-06,
+      "loss": 0.841,
       "step": 23631
     },
     {
-      "epoch": 0.6490346323913103,
+      "epoch": 0.6706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 5.794206368477026e-06,
-      "loss": 0.8374,
+      "learning_rate": 5.171288309746485e-06,
+      "loss": 0.8203,
       "step": 23632
     },
     {
-      "epoch": 0.6490620966191536,
+      "epoch": 0.6706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 5.793399366598142e-06,
-      "loss": 0.8609,
+      "learning_rate": 5.17048350137812e-06,
+      "loss": 0.8043,
       "step": 23633
     },
     {
-      "epoch": 0.6490895608469968,
+      "epoch": 0.6706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 5.792592398003516e-06,
-      "loss": 0.8172,
+      "learning_rate": 5.169678733804893e-06,
+      "loss": 0.7727,
       "step": 23634
     },
     {
-      "epoch": 0.64911702507484,
+      "epoch": 0.6706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 5.791785462699525e-06,
-      "loss": 0.8265,
+      "learning_rate": 5.168874007033615e-06,
+      "loss": 0.8447,
       "step": 23635
     },
     {
-      "epoch": 0.6491444893026832,
+      "epoch": 0.6707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 5.7909785606925615e-06,
-      "loss": 0.8424,
+      "learning_rate": 5.168069321071072e-06,
+      "loss": 0.7327,
       "step": 23636
     },
     {
-      "epoch": 0.6491719535305265,
+      "epoch": 0.6707434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 5.7901716919890016e-06,
-      "loss": 0.8336,
+      "learning_rate": 5.167264675924068e-06,
+      "loss": 0.8083,
       "step": 23637
     },
     {
-      "epoch": 0.6491994177583698,
+      "epoch": 0.6707718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 5.789364856595234e-06,
-      "loss": 0.8174,
+      "learning_rate": 5.166460071599394e-06,
+      "loss": 0.8907,
       "step": 23638
     },
     {
-      "epoch": 0.6492268819862129,
+      "epoch": 0.670800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 5.7885580545176435e-06,
-      "loss": 0.8616,
+      "learning_rate": 5.165655508103848e-06,
+      "loss": 0.8237,
       "step": 23639
     },
     {
-      "epoch": 0.6492543462140562,
+      "epoch": 0.6708286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 5.787751285762613e-06,
-      "loss": 0.828,
+      "learning_rate": 5.1648509854442305e-06,
+      "loss": 0.8507,
       "step": 23640
     },
     {
-      "epoch": 0.6492818104418995,
+      "epoch": 0.6708569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 5.786944550336529e-06,
-      "loss": 0.8454,
+      "learning_rate": 5.164046503627331e-06,
+      "loss": 0.851,
       "step": 23641
     },
     {
-      "epoch": 0.6493092746697426,
+      "epoch": 0.6708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 5.786137848245769e-06,
-      "loss": 0.7825,
+      "learning_rate": 5.163242062659947e-06,
+      "loss": 0.874,
       "step": 23642
     },
     {
-      "epoch": 0.6493367388975859,
+      "epoch": 0.6709137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 5.785331179496721e-06,
-      "loss": 0.8892,
+      "learning_rate": 5.16243766254888e-06,
+      "loss": 0.8995,
       "step": 23643
     },
     {
-      "epoch": 0.6493642031254291,
+      "epoch": 0.6709421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 5.784524544095762e-06,
-      "loss": 0.8757,
+      "learning_rate": 5.161633303300916e-06,
+      "loss": 0.7488,
       "step": 23644
     },
     {
-      "epoch": 0.6493916673532724,
+      "epoch": 0.6709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 5.783717942049278e-06,
-      "loss": 0.8566,
+      "learning_rate": 5.1608289849228525e-06,
+      "loss": 0.797,
       "step": 23645
     },
     {
-      "epoch": 0.6494191315811156,
+      "epoch": 0.6709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 5.782911373363651e-06,
-      "loss": 0.7924,
+      "learning_rate": 5.1600247074214875e-06,
+      "loss": 0.8419,
       "step": 23646
     },
     {
-      "epoch": 0.6494465958089588,
+      "epoch": 0.671027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 5.782104838045264e-06,
-      "loss": 0.7935,
+      "learning_rate": 5.159220470803608e-06,
+      "loss": 0.8064,
       "step": 23647
     },
     {
-      "epoch": 0.6494740600368021,
+      "epoch": 0.67105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.7812983361004995e-06,
-      "loss": 0.8307,
+      "learning_rate": 5.1584162750760105e-06,
+      "loss": 0.7853,
       "step": 23648
     },
     {
-      "epoch": 0.6495015242646452,
+      "epoch": 0.6710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 5.780491867535733e-06,
-      "loss": 0.794,
+      "learning_rate": 5.1576121202454925e-06,
+      "loss": 0.8147,
       "step": 23649
     },
     {
-      "epoch": 0.6495289884924885,
+      "epoch": 0.6711123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 5.7796854323573535e-06,
-      "loss": 0.8166,
+      "learning_rate": 5.156808006318834e-06,
+      "loss": 0.9078,
       "step": 23650
     },
     {
-      "epoch": 0.6495564527203318,
+      "epoch": 0.6711407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 5.778879030571734e-06,
-      "loss": 0.9596,
+      "learning_rate": 5.15600393330284e-06,
+      "loss": 0.8575,
       "step": 23651
     },
     {
-      "epoch": 0.649583916948175,
+      "epoch": 0.6711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 5.778072662185258e-06,
-      "loss": 0.928,
+      "learning_rate": 5.155199901204303e-06,
+      "loss": 0.8737,
       "step": 23652
     },
     {
-      "epoch": 0.6496113811760182,
+      "epoch": 0.6711975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 5.777266327204307e-06,
-      "loss": 0.833,
+      "learning_rate": 5.154395910030005e-06,
+      "loss": 0.8806,
       "step": 23653
     },
     {
-      "epoch": 0.6496388454038615,
+      "epoch": 0.6712258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 5.776460025635264e-06,
-      "loss": 0.9199,
+      "learning_rate": 5.153591959786745e-06,
+      "loss": 0.9112,
       "step": 23654
     },
     {
-      "epoch": 0.6496663096317047,
+      "epoch": 0.6712542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 5.775653757484503e-06,
-      "loss": 0.8211,
+      "learning_rate": 5.152788050481314e-06,
+      "loss": 0.9117,
       "step": 23655
     },
     {
-      "epoch": 0.649693773859548,
+      "epoch": 0.6712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 5.774847522758404e-06,
-      "loss": 0.8081,
+      "learning_rate": 5.151984182120496e-06,
+      "loss": 0.8222,
       "step": 23656
     },
     {
-      "epoch": 0.6497212380873911,
+      "epoch": 0.671311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 5.774041321463349e-06,
-      "loss": 0.8849,
+      "learning_rate": 5.151180354711087e-06,
+      "loss": 0.7693,
       "step": 23657
     },
     {
-      "epoch": 0.6497487023152344,
+      "epoch": 0.6713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 5.773235153605718e-06,
-      "loss": 0.775,
+      "learning_rate": 5.150376568259879e-06,
+      "loss": 0.8096,
       "step": 23658
     },
     {
-      "epoch": 0.6497761665430777,
+      "epoch": 0.6713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 5.772429019191887e-06,
-      "loss": 0.9043,
+      "learning_rate": 5.149572822773652e-06,
+      "loss": 0.8841,
       "step": 23659
     },
     {
-      "epoch": 0.6498036307709208,
+      "epoch": 0.6713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 5.771622918228237e-06,
-      "loss": 0.8148,
+      "learning_rate": 5.148769118259204e-06,
+      "loss": 0.8185,
       "step": 23660
     },
     {
-      "epoch": 0.6498310949987641,
+      "epoch": 0.6714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 5.770816850721142e-06,
-      "loss": 0.7797,
+      "learning_rate": 5.1479654547233246e-06,
+      "loss": 0.7502,
       "step": 23661
     },
     {
-      "epoch": 0.6498585592266073,
+      "epoch": 0.6714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 5.770010816676981e-06,
-      "loss": 0.8632,
+      "learning_rate": 5.147161832172795e-06,
+      "loss": 0.8371,
       "step": 23662
     },
     {
-      "epoch": 0.6498860234544506,
+      "epoch": 0.6714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 5.769204816102133e-06,
-      "loss": 0.9154,
+      "learning_rate": 5.146358250614412e-06,
+      "loss": 0.8679,
       "step": 23663
     },
     {
-      "epoch": 0.6499134876822938,
+      "epoch": 0.6715096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 5.7683988490029765e-06,
-      "loss": 0.8494,
+      "learning_rate": 5.145554710054954e-06,
+      "loss": 0.859,
       "step": 23664
     },
     {
-      "epoch": 0.649940951910137,
+      "epoch": 0.6715380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 5.76759291538589e-06,
-      "loss": 0.8404,
+      "learning_rate": 5.14475121050121e-06,
+      "loss": 0.8531,
       "step": 23665
     },
     {
-      "epoch": 0.6499684161379803,
+      "epoch": 0.671566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 5.766787015257249e-06,
-      "loss": 0.8744,
+      "learning_rate": 5.143947751959979e-06,
+      "loss": 0.8892,
       "step": 23666
     },
     {
-      "epoch": 0.6499958803658236,
+      "epoch": 0.6715947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 5.765981148623424e-06,
-      "loss": 0.7999,
+      "learning_rate": 5.143144334438035e-06,
+      "loss": 0.8964,
       "step": 23667
     },
     {
-      "epoch": 0.6500233445936667,
+      "epoch": 0.6716231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 5.765175315490796e-06,
-      "loss": 0.8888,
+      "learning_rate": 5.14234095794217e-06,
+      "loss": 0.8127,
       "step": 23668
     },
     {
-      "epoch": 0.65005080882151,
+      "epoch": 0.6716515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 5.764369515865742e-06,
-      "loss": 0.8813,
+      "learning_rate": 5.1415376224791734e-06,
+      "loss": 0.8514,
       "step": 23669
     },
     {
-      "epoch": 0.6500782730493532,
+      "epoch": 0.6716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 5.763563749754637e-06,
-      "loss": 0.9143,
+      "learning_rate": 5.140734328055822e-06,
+      "loss": 0.8279,
       "step": 23670
     },
     {
-      "epoch": 0.6501057372771964,
+      "epoch": 0.6717082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 5.7627580171638574e-06,
-      "loss": 0.8462,
+      "learning_rate": 5.139931074678907e-06,
+      "loss": 0.8813,
       "step": 23671
     },
     {
-      "epoch": 0.6501332015050397,
+      "epoch": 0.6717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 5.7619523180997814e-06,
-      "loss": 0.8808,
+      "learning_rate": 5.139127862355215e-06,
+      "loss": 0.8216,
       "step": 23672
     },
     {
-      "epoch": 0.6501606657328829,
+      "epoch": 0.6717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 5.761146652568777e-06,
-      "loss": 0.9519,
+      "learning_rate": 5.138324691091527e-06,
+      "loss": 0.7129,
       "step": 23673
     },
     {
-      "epoch": 0.6501881299607262,
+      "epoch": 0.6717934165720771,
       "grad_norm": 0.0,
-      "learning_rate": 5.760341020577225e-06,
-      "loss": 0.8444,
+      "learning_rate": 5.1375215608946275e-06,
+      "loss": 0.8909,
       "step": 23674
     },
     {
-      "epoch": 0.6502155941885693,
+      "epoch": 0.6718217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 5.759535422131492e-06,
-      "loss": 0.8546,
+      "learning_rate": 5.136718471771307e-06,
+      "loss": 0.7866,
       "step": 23675
     },
     {
-      "epoch": 0.6502430584164126,
+      "epoch": 0.671850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 5.758729857237959e-06,
-      "loss": 0.8334,
+      "learning_rate": 5.135915423728339e-06,
+      "loss": 0.7991,
       "step": 23676
     },
     {
-      "epoch": 0.6502705226442559,
+      "epoch": 0.6718785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 5.7579243259029985e-06,
-      "loss": 0.8034,
+      "learning_rate": 5.135112416772512e-06,
+      "loss": 0.7838,
       "step": 23677
     },
     {
-      "epoch": 0.650297986872099,
+      "epoch": 0.6719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 5.757118828132987e-06,
-      "loss": 0.845,
+      "learning_rate": 5.134309450910613e-06,
+      "loss": 0.8842,
       "step": 23678
     },
     {
-      "epoch": 0.6503254510999423,
+      "epoch": 0.6719353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 5.7563133639342915e-06,
-      "loss": 0.7964,
+      "learning_rate": 5.133506526149418e-06,
+      "loss": 0.8447,
       "step": 23679
     },
     {
-      "epoch": 0.6503529153277856,
+      "epoch": 0.6719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 5.755507933313289e-06,
-      "loss": 0.8973,
+      "learning_rate": 5.132703642495711e-06,
+      "loss": 0.8811,
       "step": 23680
     },
     {
-      "epoch": 0.6503803795556288,
+      "epoch": 0.6719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 5.754702536276351e-06,
-      "loss": 0.8937,
+      "learning_rate": 5.131900799956276e-06,
+      "loss": 0.8298,
       "step": 23681
     },
     {
-      "epoch": 0.650407843783472,
+      "epoch": 0.6720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 5.753897172829855e-06,
-      "loss": 0.8116,
+      "learning_rate": 5.131097998537893e-06,
+      "loss": 0.7886,
       "step": 23682
     },
     {
-      "epoch": 0.6504353080113152,
+      "epoch": 0.6720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 5.753091842980165e-06,
-      "loss": 0.7765,
+      "learning_rate": 5.130295238247349e-06,
+      "loss": 0.8911,
       "step": 23683
     },
     {
-      "epoch": 0.6504627722391585,
+      "epoch": 0.6720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 5.752286546733663e-06,
-      "loss": 0.8931,
+      "learning_rate": 5.129492519091415e-06,
+      "loss": 0.9789,
       "step": 23684
     },
     {
-      "epoch": 0.6504902364670018,
+      "epoch": 0.672105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.751481284096709e-06,
-      "loss": 0.7939,
+      "learning_rate": 5.128689841076877e-06,
+      "loss": 0.8597,
       "step": 23685
     },
     {
-      "epoch": 0.6505177006948449,
+      "epoch": 0.6721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 5.750676055075683e-06,
-      "loss": 0.8014,
+      "learning_rate": 5.12788720421052e-06,
+      "loss": 0.765,
       "step": 23686
     },
     {
-      "epoch": 0.6505451649226882,
+      "epoch": 0.6721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 5.749870859676954e-06,
-      "loss": 0.8748,
+      "learning_rate": 5.127084608499113e-06,
+      "loss": 0.7917,
       "step": 23687
     },
     {
-      "epoch": 0.6505726291505314,
+      "epoch": 0.6721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 5.749065697906891e-06,
-      "loss": 0.8133,
+      "learning_rate": 5.126282053949443e-06,
+      "loss": 0.9498,
       "step": 23688
     },
     {
-      "epoch": 0.6506000933783747,
+      "epoch": 0.6722190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 5.748260569771873e-06,
-      "loss": 0.9148,
+      "learning_rate": 5.12547954056829e-06,
+      "loss": 0.883,
       "step": 23689
     },
     {
-      "epoch": 0.6506275576062179,
+      "epoch": 0.6722474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 5.747455475278263e-06,
-      "loss": 0.8511,
+      "learning_rate": 5.124677068362427e-06,
+      "loss": 0.8419,
       "step": 23690
     },
     {
-      "epoch": 0.6506550218340611,
+      "epoch": 0.6722758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 5.746650414432428e-06,
-      "loss": 0.8013,
+      "learning_rate": 5.123874637338637e-06,
+      "loss": 0.8912,
       "step": 23691
     },
     {
-      "epoch": 0.6506824860619044,
+      "epoch": 0.6723041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 5.745845387240745e-06,
-      "loss": 0.9065,
+      "learning_rate": 5.1230722475037e-06,
+      "loss": 0.8881,
       "step": 23692
     },
     {
-      "epoch": 0.6507099502897477,
+      "epoch": 0.6723325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 5.7450403937095785e-06,
-      "loss": 0.7727,
+      "learning_rate": 5.122269898864387e-06,
+      "loss": 0.7569,
       "step": 23693
     },
     {
-      "epoch": 0.6507374145175908,
+      "epoch": 0.672360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 5.7442354338453026e-06,
-      "loss": 0.8795,
+      "learning_rate": 5.1214675914274805e-06,
+      "loss": 0.8192,
       "step": 23694
     },
     {
-      "epoch": 0.6507648787454341,
+      "epoch": 0.67238933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.743430507654287e-06,
-      "loss": 0.8339,
+      "learning_rate": 5.120665325199759e-06,
+      "loss": 0.8718,
       "step": 23695
     },
     {
-      "epoch": 0.6507923429732773,
+      "epoch": 0.6724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 5.742625615142896e-06,
-      "loss": 0.797,
+      "learning_rate": 5.119863100187989e-06,
+      "loss": 0.7985,
       "step": 23696
     },
     {
-      "epoch": 0.6508198072011205,
+      "epoch": 0.6724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.741820756317498e-06,
-      "loss": 0.822,
+      "learning_rate": 5.1190609163989635e-06,
+      "loss": 0.9054,
       "step": 23697
     },
     {
-      "epoch": 0.6508472714289638,
+      "epoch": 0.6724744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 5.7410159311844694e-06,
-      "loss": 0.8537,
+      "learning_rate": 5.118258773839446e-06,
+      "loss": 0.8506,
       "step": 23698
     },
     {
-      "epoch": 0.650874735656807,
+      "epoch": 0.6725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 5.740211139750167e-06,
-      "loss": 0.829,
+      "learning_rate": 5.117456672516217e-06,
+      "loss": 0.897,
       "step": 23699
     },
     {
-      "epoch": 0.6509021998846503,
+      "epoch": 0.6725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 5.739406382020966e-06,
-      "loss": 0.8478,
+      "learning_rate": 5.116654612436053e-06,
+      "loss": 0.8047,
       "step": 23700
     },
     {
-      "epoch": 0.6509296641124934,
+      "epoch": 0.6725595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 5.738601658003234e-06,
-      "loss": 0.8217,
+      "learning_rate": 5.115852593605723e-06,
+      "loss": 0.8014,
       "step": 23701
     },
     {
-      "epoch": 0.6509571283403367,
+      "epoch": 0.6725879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 5.737796967703334e-06,
-      "loss": 0.8285,
+      "learning_rate": 5.115050616032007e-06,
+      "loss": 0.871,
       "step": 23702
     },
     {
-      "epoch": 0.65098459256818,
+      "epoch": 0.6726163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 5.736992311127635e-06,
-      "loss": 0.8536,
+      "learning_rate": 5.1142486797216816e-06,
+      "loss": 0.903,
       "step": 23703
     },
     {
-      "epoch": 0.6510120567960231,
+      "epoch": 0.6726447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 5.736187688282504e-06,
-      "loss": 0.861,
+      "learning_rate": 5.113446784681513e-06,
+      "loss": 0.972,
       "step": 23704
     },
     {
-      "epoch": 0.6510395210238664,
+      "epoch": 0.6726730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 5.735383099174312e-06,
-      "loss": 0.8891,
+      "learning_rate": 5.11264493091828e-06,
+      "loss": 0.7807,
       "step": 23705
     },
     {
-      "epoch": 0.6510669852517097,
+      "epoch": 0.6727014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 5.734578543809416e-06,
-      "loss": 0.8767,
+      "learning_rate": 5.111843118438757e-06,
+      "loss": 0.9317,
       "step": 23706
     },
     {
-      "epoch": 0.6510944494795529,
+      "epoch": 0.6727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 5.733774022194192e-06,
-      "loss": 0.8217,
+      "learning_rate": 5.111041347249712e-06,
+      "loss": 0.8609,
       "step": 23707
     },
     {
-      "epoch": 0.6511219137073961,
+      "epoch": 0.6727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 5.732969534334995e-06,
-      "loss": 0.8024,
+      "learning_rate": 5.110239617357921e-06,
+      "loss": 0.8554,
       "step": 23708
     },
     {
-      "epoch": 0.6511493779352393,
+      "epoch": 0.6727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 5.732165080238197e-06,
-      "loss": 0.8844,
+      "learning_rate": 5.10943792877016e-06,
+      "loss": 0.8553,
       "step": 23709
     },
     {
-      "epoch": 0.6511768421630826,
+      "epoch": 0.6728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 5.731360659910163e-06,
-      "loss": 0.9038,
+      "learning_rate": 5.108636281493191e-06,
+      "loss": 0.8687,
       "step": 23710
     },
     {
-      "epoch": 0.6512043063909259,
+      "epoch": 0.6728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 5.730556273357255e-06,
-      "loss": 0.7681,
+      "learning_rate": 5.1078346755337935e-06,
+      "loss": 0.7935,
       "step": 23711
     },
     {
-      "epoch": 0.651231770618769,
+      "epoch": 0.6728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 5.72975192058584e-06,
-      "loss": 0.963,
+      "learning_rate": 5.107033110898735e-06,
+      "loss": 0.8519,
       "step": 23712
     },
     {
-      "epoch": 0.6512592348466123,
+      "epoch": 0.672900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 5.7289476016022875e-06,
-      "loss": 0.8666,
+      "learning_rate": 5.106231587594788e-06,
+      "loss": 0.851,
       "step": 23713
     },
     {
-      "epoch": 0.6512866990744555,
+      "epoch": 0.6729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 5.728143316412955e-06,
-      "loss": 0.9095,
+      "learning_rate": 5.105430105628726e-06,
+      "loss": 0.7863,
       "step": 23714
     },
     {
-      "epoch": 0.6513141633022987,
+      "epoch": 0.6729568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 5.727339065024202e-06,
-      "loss": 0.8865,
+      "learning_rate": 5.104628665007313e-06,
+      "loss": 0.7769,
       "step": 23715
     },
     {
-      "epoch": 0.651341627530142,
+      "epoch": 0.6729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 5.7265348474424e-06,
-      "loss": 0.8087,
+      "learning_rate": 5.103827265737322e-06,
+      "loss": 0.771,
       "step": 23716
     },
     {
-      "epoch": 0.6513690917579852,
+      "epoch": 0.6730136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 5.725730663673907e-06,
-      "loss": 0.9254,
+      "learning_rate": 5.103025907825525e-06,
+      "loss": 0.7711,
       "step": 23717
     },
     {
-      "epoch": 0.6513965559858285,
+      "epoch": 0.6730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 5.724926513725091e-06,
-      "loss": 0.8596,
+      "learning_rate": 5.102224591278685e-06,
+      "loss": 0.7329,
       "step": 23718
     },
     {
-      "epoch": 0.6514240202136717,
+      "epoch": 0.6730703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 5.724122397602316e-06,
-      "loss": 0.8404,
+      "learning_rate": 5.101423316103573e-06,
+      "loss": 0.791,
       "step": 23719
     },
     {
-      "epoch": 0.6514514844415149,
+      "epoch": 0.6730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 5.7233183153119355e-06,
-      "loss": 0.8719,
+      "learning_rate": 5.100622082306964e-06,
+      "loss": 0.7924,
       "step": 23720
     },
     {
-      "epoch": 0.6514789486693582,
+      "epoch": 0.6731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 5.7225142668603175e-06,
-      "loss": 0.8488,
+      "learning_rate": 5.099820889895616e-06,
+      "loss": 0.7959,
       "step": 23721
     },
     {
-      "epoch": 0.6515064128972013,
+      "epoch": 0.6731555051078321,
       "grad_norm": 0.0,
-      "learning_rate": 5.72171025225383e-06,
-      "loss": 0.7517,
+      "learning_rate": 5.0990197388763005e-06,
+      "loss": 0.8723,
       "step": 23722
     },
     {
-      "epoch": 0.6515338771250446,
+      "epoch": 0.6731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 5.72090627149882e-06,
-      "loss": 0.9261,
+      "learning_rate": 5.0982186292557915e-06,
+      "loss": 0.7873,
       "step": 23723
     },
     {
-      "epoch": 0.6515613413528879,
+      "epoch": 0.6732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 5.720102324601659e-06,
-      "loss": 0.8616,
+      "learning_rate": 5.097417561040844e-06,
+      "loss": 0.7515,
       "step": 23724
     },
     {
-      "epoch": 0.6515888055807311,
+      "epoch": 0.6732406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 5.719298411568712e-06,
-      "loss": 0.8087,
+      "learning_rate": 5.096616534238232e-06,
+      "loss": 0.8246,
       "step": 23725
     },
     {
-      "epoch": 0.6516162698085743,
+      "epoch": 0.6732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 5.718494532406329e-06,
-      "loss": 0.9094,
+      "learning_rate": 5.0958155488547186e-06,
+      "loss": 0.803,
       "step": 23726
     },
     {
-      "epoch": 0.6516437340364176,
+      "epoch": 0.6732973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 5.717690687120876e-06,
-      "loss": 0.9171,
+      "learning_rate": 5.095014604897075e-06,
+      "loss": 0.84,
       "step": 23727
     },
     {
-      "epoch": 0.6516711982642608,
+      "epoch": 0.6733257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 5.716886875718713e-06,
-      "loss": 0.9107,
+      "learning_rate": 5.094213702372065e-06,
+      "loss": 0.8549,
       "step": 23728
     },
     {
-      "epoch": 0.6516986624921041,
+      "epoch": 0.6733541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 5.716083098206203e-06,
-      "loss": 0.8868,
+      "learning_rate": 5.093412841286449e-06,
+      "loss": 0.9103,
       "step": 23729
     },
     {
-      "epoch": 0.6517261267199472,
+      "epoch": 0.6733825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 5.7152793545897e-06,
-      "loss": 0.853,
+      "learning_rate": 5.0926120216469956e-06,
+      "loss": 0.7629,
       "step": 23730
     },
     {
-      "epoch": 0.6517535909477905,
+      "epoch": 0.673410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 5.71447564487557e-06,
-      "loss": 0.8911,
+      "learning_rate": 5.091811243460473e-06,
+      "loss": 0.8178,
       "step": 23731
     },
     {
-      "epoch": 0.6517810551756338,
+      "epoch": 0.6734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 5.713671969070165e-06,
-      "loss": 0.9242,
+      "learning_rate": 5.091010506733637e-06,
+      "loss": 0.8074,
       "step": 23732
     },
     {
-      "epoch": 0.651808519403477,
+      "epoch": 0.6734676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 5.712868327179848e-06,
-      "loss": 0.8753,
+      "learning_rate": 5.090209811473257e-06,
+      "loss": 0.8576,
       "step": 23733
     },
     {
-      "epoch": 0.6518359836313202,
+      "epoch": 0.6734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 5.712064719210977e-06,
-      "loss": 0.9175,
+      "learning_rate": 5.089409157686098e-06,
+      "loss": 0.9092,
       "step": 23734
     },
     {
-      "epoch": 0.6518634478591634,
+      "epoch": 0.6735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 5.71126114516991e-06,
-      "loss": 0.9475,
+      "learning_rate": 5.088608545378917e-06,
+      "loss": 0.861,
       "step": 23735
     },
     {
-      "epoch": 0.6518909120870067,
+      "epoch": 0.6735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 5.710457605063009e-06,
-      "loss": 0.7866,
+      "learning_rate": 5.087807974558481e-06,
+      "loss": 0.8418,
       "step": 23736
     },
     {
-      "epoch": 0.65191837631485,
+      "epoch": 0.6735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 5.709654098896631e-06,
-      "loss": 0.9477,
+      "learning_rate": 5.0870074452315555e-06,
+      "loss": 0.8074,
       "step": 23737
     },
     {
-      "epoch": 0.6519458405426931,
+      "epoch": 0.6736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 5.708850626677132e-06,
-      "loss": 0.8842,
+      "learning_rate": 5.0862069574048956e-06,
+      "loss": 0.7659,
       "step": 23738
     },
     {
-      "epoch": 0.6519733047705364,
+      "epoch": 0.6736379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 5.708047188410866e-06,
-      "loss": 0.8618,
+      "learning_rate": 5.085406511085265e-06,
+      "loss": 0.8237,
       "step": 23739
     },
     {
-      "epoch": 0.6520007689983797,
+      "epoch": 0.6736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 5.707243784104192e-06,
-      "loss": 0.8515,
+      "learning_rate": 5.084606106279431e-06,
+      "loss": 0.913,
       "step": 23740
     },
     {
-      "epoch": 0.6520282332262228,
+      "epoch": 0.673694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 5.706440413763467e-06,
-      "loss": 0.8701,
+      "learning_rate": 5.083805742994147e-06,
+      "loss": 0.8466,
       "step": 23741
     },
     {
-      "epoch": 0.6520556974540661,
+      "epoch": 0.6737230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 5.705637077395051e-06,
-      "loss": 0.9501,
+      "learning_rate": 5.083005421236173e-06,
+      "loss": 0.8282,
       "step": 23742
     },
     {
-      "epoch": 0.6520831616819093,
+      "epoch": 0.6737514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 5.704833775005301e-06,
-      "loss": 0.9045,
+      "learning_rate": 5.08220514101228e-06,
+      "loss": 0.7394,
       "step": 23743
     },
     {
-      "epoch": 0.6521106259097526,
+      "epoch": 0.6737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 5.704030506600566e-06,
-      "loss": 0.8866,
+      "learning_rate": 5.081404902329219e-06,
+      "loss": 0.8926,
       "step": 23744
     },
     {
-      "epoch": 0.6521380901375958,
+      "epoch": 0.6738081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 5.703227272187205e-06,
-      "loss": 0.91,
+      "learning_rate": 5.080604705193754e-06,
+      "loss": 0.9991,
       "step": 23745
     },
     {
-      "epoch": 0.652165554365439,
+      "epoch": 0.6738365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 5.70242407177158e-06,
-      "loss": 0.9098,
+      "learning_rate": 5.079804549612639e-06,
+      "loss": 0.8223,
       "step": 23746
     },
     {
-      "epoch": 0.6521930185932823,
+      "epoch": 0.6738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 5.701620905360034e-06,
-      "loss": 0.84,
+      "learning_rate": 5.079004435592636e-06,
+      "loss": 0.9263,
       "step": 23747
     },
     {
-      "epoch": 0.6522204828211254,
+      "epoch": 0.6738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 5.70081777295893e-06,
-      "loss": 0.7364,
+      "learning_rate": 5.078204363140507e-06,
+      "loss": 0.8792,
       "step": 23748
     },
     {
-      "epoch": 0.6522479470489687,
+      "epoch": 0.673921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 5.700014674574626e-06,
-      "loss": 0.8276,
+      "learning_rate": 5.077404332263005e-06,
+      "loss": 0.81,
       "step": 23749
     },
     {
-      "epoch": 0.652275411276812,
+      "epoch": 0.673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.699211610213464e-06,
-      "loss": 0.8052,
+      "learning_rate": 5.076604342966889e-06,
+      "loss": 0.8853,
       "step": 23750
     },
     {
-      "epoch": 0.6523028755046552,
+      "epoch": 0.6739784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 5.698408579881809e-06,
-      "loss": 0.8743,
+      "learning_rate": 5.075804395258921e-06,
+      "loss": 0.7675,
       "step": 23751
     },
     {
-      "epoch": 0.6523303397324984,
+      "epoch": 0.6740068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 5.69760558358601e-06,
-      "loss": 0.8425,
+      "learning_rate": 5.075004489145851e-06,
+      "loss": 0.8118,
       "step": 23752
     },
     {
-      "epoch": 0.6523578039603417,
+      "epoch": 0.6740351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 5.696802621332428e-06,
-      "loss": 0.8281,
+      "learning_rate": 5.074204624634439e-06,
+      "loss": 0.8565,
       "step": 23753
     },
     {
-      "epoch": 0.6523852681881849,
+      "epoch": 0.6740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 5.6959996931274054e-06,
-      "loss": 0.7856,
+      "learning_rate": 5.073404801731446e-06,
+      "loss": 0.8061,
       "step": 23754
     },
     {
-      "epoch": 0.6524127324160282,
+      "epoch": 0.6740919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 5.695196798977305e-06,
-      "loss": 0.8107,
+      "learning_rate": 5.07260502044362e-06,
+      "loss": 0.7482,
       "step": 23755
     },
     {
-      "epoch": 0.6524401966438713,
+      "epoch": 0.6741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 5.69439393888847e-06,
-      "loss": 0.8576,
+      "learning_rate": 5.071805280777722e-06,
+      "loss": 0.8912,
       "step": 23756
     },
     {
-      "epoch": 0.6524676608717146,
+      "epoch": 0.6741486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 5.693591112867258e-06,
-      "loss": 0.8092,
+      "learning_rate": 5.0710055827405045e-06,
+      "loss": 0.7436,
       "step": 23757
     },
     {
-      "epoch": 0.6524951250995579,
+      "epoch": 0.6741770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 5.692788320920022e-06,
-      "loss": 0.8258,
+      "learning_rate": 5.0702059263387275e-06,
+      "loss": 0.7355,
       "step": 23758
     },
     {
-      "epoch": 0.652522589327401,
+      "epoch": 0.674205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 5.6919855630531125e-06,
-      "loss": 0.7822,
+      "learning_rate": 5.069406311579139e-06,
+      "loss": 0.9204,
       "step": 23759
     },
     {
-      "epoch": 0.6525500535552443,
+      "epoch": 0.674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 5.691182839272886e-06,
-      "loss": 0.8563,
+      "learning_rate": 5.068606738468503e-06,
+      "loss": 0.8514,
       "step": 23760
     },
     {
-      "epoch": 0.6525775177830875,
+      "epoch": 0.6742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 5.690380149585688e-06,
-      "loss": 0.8444,
+      "learning_rate": 5.067807207013564e-06,
+      "loss": 0.9054,
       "step": 23761
     },
     {
-      "epoch": 0.6526049820109308,
+      "epoch": 0.6742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 5.6895774939978735e-06,
-      "loss": 0.7997,
+      "learning_rate": 5.067007717221079e-06,
+      "loss": 0.8273,
       "step": 23762
     },
     {
-      "epoch": 0.652632446238774,
+      "epoch": 0.6743189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 5.688774872515788e-06,
-      "loss": 1.0216,
+      "learning_rate": 5.0662082690978045e-06,
+      "loss": 0.8763,
       "step": 23763
     },
     {
-      "epoch": 0.6526599104666172,
+      "epoch": 0.6743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 5.687972285145786e-06,
-      "loss": 0.978,
+      "learning_rate": 5.065408862650487e-06,
+      "loss": 0.8468,
       "step": 23764
     },
     {
-      "epoch": 0.6526873746944605,
+      "epoch": 0.6743757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 5.687169731894218e-06,
-      "loss": 0.8075,
+      "learning_rate": 5.064609497885883e-06,
+      "loss": 0.8454,
       "step": 23765
     },
     {
-      "epoch": 0.6527148389223038,
+      "epoch": 0.6744040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 5.686367212767436e-06,
-      "loss": 0.8081,
+      "learning_rate": 5.063810174810748e-06,
+      "loss": 0.7878,
       "step": 23766
     },
     {
-      "epoch": 0.6527423031501469,
+      "epoch": 0.6744324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 5.685564727771787e-06,
-      "loss": 0.9587,
+      "learning_rate": 5.063010893431827e-06,
+      "loss": 0.7511,
       "step": 23767
     },
     {
-      "epoch": 0.6527697673779902,
+      "epoch": 0.674460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.684762276913618e-06,
-      "loss": 0.8758,
+      "learning_rate": 5.062211653755874e-06,
+      "loss": 0.7321,
       "step": 23768
     },
     {
-      "epoch": 0.6527972316058334,
+      "epoch": 0.6744892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 5.683959860199282e-06,
-      "loss": 0.8841,
+      "learning_rate": 5.061412455789645e-06,
+      "loss": 0.9576,
       "step": 23769
     },
     {
-      "epoch": 0.6528246958336766,
+      "epoch": 0.6745175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 5.6831574776351314e-06,
-      "loss": 0.9263,
+      "learning_rate": 5.060613299539885e-06,
+      "loss": 0.8989,
       "step": 23770
     },
     {
-      "epoch": 0.6528521600615199,
+      "epoch": 0.6745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 5.682355129227508e-06,
-      "loss": 0.749,
+      "learning_rate": 5.059814185013348e-06,
+      "loss": 0.8297,
       "step": 23771
     },
     {
-      "epoch": 0.6528796242893631,
+      "epoch": 0.6745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 5.681552814982767e-06,
-      "loss": 0.9114,
+      "learning_rate": 5.059015112216776e-06,
+      "loss": 0.8802,
       "step": 23772
     },
     {
-      "epoch": 0.6529070885172064,
+      "epoch": 0.6746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 5.680750534907248e-06,
-      "loss": 0.8282,
+      "learning_rate": 5.058216081156929e-06,
+      "loss": 0.8581,
       "step": 23773
     },
     {
-      "epoch": 0.6529345527450495,
+      "epoch": 0.6746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 5.6799482890073046e-06,
-      "loss": 0.7481,
+      "learning_rate": 5.057417091840558e-06,
+      "loss": 0.7391,
       "step": 23774
     },
     {
-      "epoch": 0.6529620169728928,
+      "epoch": 0.6746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 5.6791460772892835e-06,
-      "loss": 0.891,
+      "learning_rate": 5.056618144274401e-06,
+      "loss": 0.8099,
       "step": 23775
     },
     {
-      "epoch": 0.6529894812007361,
+      "epoch": 0.6746878547105561,
       "grad_norm": 0.0,
-      "learning_rate": 5.678343899759533e-06,
-      "loss": 0.9167,
+      "learning_rate": 5.055819238465214e-06,
+      "loss": 0.8511,
       "step": 23776
     },
     {
-      "epoch": 0.6530169454285792,
+      "epoch": 0.6747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 5.677541756424403e-06,
-      "loss": 0.8688,
+      "learning_rate": 5.055020374419748e-06,
+      "loss": 0.8813,
       "step": 23777
     },
     {
-      "epoch": 0.6530444096564225,
+      "epoch": 0.674744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 5.676739647290232e-06,
-      "loss": 0.8245,
+      "learning_rate": 5.0542215521447445e-06,
+      "loss": 0.9001,
       "step": 23778
     },
     {
-      "epoch": 0.6530718738842658,
+      "epoch": 0.6747729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 5.6759375723633785e-06,
-      "loss": 0.8619,
+      "learning_rate": 5.053422771646952e-06,
+      "loss": 0.8074,
       "step": 23779
     },
     {
-      "epoch": 0.653099338112109,
+      "epoch": 0.6748013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 5.675135531650175e-06,
-      "loss": 0.8204,
+      "learning_rate": 5.052624032933124e-06,
+      "loss": 0.862,
       "step": 23780
     },
     {
-      "epoch": 0.6531268023399522,
+      "epoch": 0.6748297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 5.674333525156977e-06,
-      "loss": 0.87,
+      "learning_rate": 5.05182533601e-06,
+      "loss": 0.8387,
       "step": 23781
     },
     {
-      "epoch": 0.6531542665677954,
+      "epoch": 0.6748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 5.673531552890126e-06,
-      "loss": 0.9131,
+      "learning_rate": 5.05102668088433e-06,
+      "loss": 0.8973,
       "step": 23782
     },
     {
-      "epoch": 0.6531817307956387,
+      "epoch": 0.6748864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 5.67272961485597e-06,
-      "loss": 0.7851,
+      "learning_rate": 5.0502280675628625e-06,
+      "loss": 0.8526,
       "step": 23783
     },
     {
-      "epoch": 0.653209195023482,
+      "epoch": 0.6749148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 5.671927711060856e-06,
-      "loss": 0.7735,
+      "learning_rate": 5.049429496052338e-06,
+      "loss": 0.8123,
       "step": 23784
     },
     {
-      "epoch": 0.6532366592513251,
+      "epoch": 0.6749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 5.671125841511125e-06,
-      "loss": 0.9153,
+      "learning_rate": 5.048630966359505e-06,
+      "loss": 0.8505,
       "step": 23785
     },
     {
-      "epoch": 0.6532641234791684,
+      "epoch": 0.6749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 5.670324006213125e-06,
-      "loss": 0.8025,
+      "learning_rate": 5.047832478491113e-06,
+      "loss": 0.8225,
       "step": 23786
     },
     {
-      "epoch": 0.6532915877070116,
+      "epoch": 0.675,
       "grad_norm": 0.0,
-      "learning_rate": 5.669522205173195e-06,
-      "loss": 0.8647,
+      "learning_rate": 5.0470340324538995e-06,
+      "loss": 0.7423,
       "step": 23787
     },
     {
-      "epoch": 0.6533190519348548,
+      "epoch": 0.6750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 5.668720438397683e-06,
-      "loss": 0.7928,
+      "learning_rate": 5.04623562825461e-06,
+      "loss": 0.8984,
       "step": 23788
     },
     {
-      "epoch": 0.6533465161626981,
+      "epoch": 0.6750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 5.6679187058929345e-06,
-      "loss": 0.876,
+      "learning_rate": 5.045437265899993e-06,
+      "loss": 0.7851,
       "step": 23789
     },
     {
-      "epoch": 0.6533739803905413,
+      "epoch": 0.6750851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 5.667117007665293e-06,
-      "loss": 0.8097,
+      "learning_rate": 5.044638945396789e-06,
+      "loss": 0.8659,
       "step": 23790
     },
     {
-      "epoch": 0.6534014446183846,
+      "epoch": 0.6751135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 5.666315343721096e-06,
-      "loss": 0.9762,
+      "learning_rate": 5.043840666751747e-06,
+      "loss": 0.8409,
       "step": 23791
     },
     {
-      "epoch": 0.6534289088462278,
+      "epoch": 0.6751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 5.665513714066693e-06,
-      "loss": 0.8122,
+      "learning_rate": 5.0430424299716005e-06,
+      "loss": 0.8529,
       "step": 23792
     },
     {
-      "epoch": 0.653456373074071,
+      "epoch": 0.6751702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 5.664712118708422e-06,
-      "loss": 0.8735,
+      "learning_rate": 5.042244235063098e-06,
+      "loss": 0.8716,
       "step": 23793
     },
     {
-      "epoch": 0.6534838373019143,
+      "epoch": 0.6751986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 5.663910557652633e-06,
-      "loss": 0.8463,
+      "learning_rate": 5.041446082032986e-06,
+      "loss": 0.8172,
       "step": 23794
     },
     {
-      "epoch": 0.6535113015297574,
+      "epoch": 0.6752270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 5.66310903090566e-06,
-      "loss": 0.7897,
+      "learning_rate": 5.040647970887998e-06,
+      "loss": 0.8062,
       "step": 23795
     },
     {
-      "epoch": 0.6535387657576007,
+      "epoch": 0.675255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 5.662307538473851e-06,
-      "loss": 0.8187,
+      "learning_rate": 5.039849901634879e-06,
+      "loss": 0.8671,
       "step": 23796
     },
     {
-      "epoch": 0.653566229985444,
+      "epoch": 0.675283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 5.661506080363541e-06,
-      "loss": 0.8164,
+      "learning_rate": 5.039051874280374e-06,
+      "loss": 0.8515,
       "step": 23797
     },
     {
-      "epoch": 0.6535936942132872,
+      "epoch": 0.6753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 5.660704656581077e-06,
-      "loss": 0.844,
+      "learning_rate": 5.038253888831219e-06,
+      "loss": 0.8509,
       "step": 23798
     },
     {
-      "epoch": 0.6536211584411304,
+      "epoch": 0.6753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 5.659903267132798e-06,
-      "loss": 0.8563,
+      "learning_rate": 5.0374559452941544e-06,
+      "loss": 0.8891,
       "step": 23799
     },
     {
-      "epoch": 0.6536486226689736,
+      "epoch": 0.6753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 5.659101912025045e-06,
-      "loss": 0.8258,
+      "learning_rate": 5.036658043675928e-06,
+      "loss": 0.907,
       "step": 23800
     },
     {
-      "epoch": 0.6536760868968169,
+      "epoch": 0.6753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 5.658300591264163e-06,
-      "loss": 0.796,
+      "learning_rate": 5.03586018398327e-06,
+      "loss": 0.808,
       "step": 23801
     },
     {
-      "epoch": 0.6537035511246602,
+      "epoch": 0.6754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 5.657499304856486e-06,
-      "loss": 0.8407,
+      "learning_rate": 5.0350623662229225e-06,
+      "loss": 0.8745,
       "step": 23802
     },
     {
-      "epoch": 0.6537310153525033,
+      "epoch": 0.6754540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 5.65669805280836e-06,
-      "loss": 0.8785,
+      "learning_rate": 5.034264590401627e-06,
+      "loss": 0.8736,
       "step": 23803
     },
     {
-      "epoch": 0.6537584795803466,
+      "epoch": 0.6754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 5.655896835126116e-06,
-      "loss": 0.814,
+      "learning_rate": 5.033466856526124e-06,
+      "loss": 0.8402,
       "step": 23804
     },
     {
-      "epoch": 0.6537859438081899,
+      "epoch": 0.6755107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 5.6550956518161e-06,
-      "loss": 0.809,
+      "learning_rate": 5.032669164603152e-06,
+      "loss": 0.762,
       "step": 23805
     },
     {
-      "epoch": 0.653813408036033,
+      "epoch": 0.6755391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 5.654294502884652e-06,
-      "loss": 0.8474,
+      "learning_rate": 5.031871514639443e-06,
+      "loss": 0.8497,
       "step": 23806
     },
     {
-      "epoch": 0.6538408722638763,
+      "epoch": 0.6755675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 5.653493388338107e-06,
-      "loss": 0.9024,
+      "learning_rate": 5.031073906641738e-06,
+      "loss": 0.8243,
       "step": 23807
     },
     {
-      "epoch": 0.6538683364917195,
+      "epoch": 0.6755959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 5.65269230818281e-06,
-      "loss": 0.7581,
+      "learning_rate": 5.03027634061678e-06,
+      "loss": 0.8261,
       "step": 23808
     },
     {
-      "epoch": 0.6538958007195628,
+      "epoch": 0.6756242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 5.6518912624250935e-06,
-      "loss": 0.8974,
+      "learning_rate": 5.029478816571296e-06,
+      "loss": 0.7689,
       "step": 23809
     },
     {
-      "epoch": 0.653923264947406,
+      "epoch": 0.6756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 5.6510902510712985e-06,
-      "loss": 0.7999,
+      "learning_rate": 5.028681334512029e-06,
+      "loss": 0.896,
       "step": 23810
     },
     {
-      "epoch": 0.6539507291752492,
+      "epoch": 0.6756810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 5.65028927412776e-06,
-      "loss": 0.8469,
+      "learning_rate": 5.027883894445717e-06,
+      "loss": 0.8026,
       "step": 23811
     },
     {
-      "epoch": 0.6539781934030925,
+      "epoch": 0.6757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 5.649488331600818e-06,
-      "loss": 0.8944,
+      "learning_rate": 5.027086496379091e-06,
+      "loss": 0.888,
       "step": 23812
     },
     {
-      "epoch": 0.6540056576309357,
+      "epoch": 0.6757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 5.648687423496808e-06,
-      "loss": 0.8796,
+      "learning_rate": 5.026289140318889e-06,
+      "loss": 0.869,
       "step": 23813
     },
     {
-      "epoch": 0.6540331218587789,
+      "epoch": 0.6757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 5.647886549822075e-06,
-      "loss": 0.6798,
+      "learning_rate": 5.025491826271849e-06,
+      "loss": 0.8315,
       "step": 23814
     },
     {
-      "epoch": 0.6540605860866222,
+      "epoch": 0.675794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 5.647085710582944e-06,
-      "loss": 0.8213,
+      "learning_rate": 5.024694554244699e-06,
+      "loss": 0.8192,
       "step": 23815
     },
     {
-      "epoch": 0.6540880503144654,
+      "epoch": 0.6758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 5.646284905785757e-06,
-      "loss": 0.8561,
+      "learning_rate": 5.023897324244178e-06,
+      "loss": 0.7905,
       "step": 23816
     },
     {
-      "epoch": 0.6541155145423087,
+      "epoch": 0.6758513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 5.64548413543685e-06,
-      "loss": 0.8591,
+      "learning_rate": 5.023100136277024e-06,
+      "loss": 0.8328,
       "step": 23817
     },
     {
-      "epoch": 0.6541429787701519,
+      "epoch": 0.6758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 5.6446833995425635e-06,
-      "loss": 0.8718,
+      "learning_rate": 5.0223029903499595e-06,
+      "loss": 0.902,
       "step": 23818
     },
     {
-      "epoch": 0.6541704429979951,
+      "epoch": 0.6759080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 5.643882698109229e-06,
-      "loss": 0.9095,
+      "learning_rate": 5.021505886469733e-06,
+      "loss": 0.8749,
       "step": 23819
     },
     {
-      "epoch": 0.6541979072258384,
+      "epoch": 0.6759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 5.643082031143179e-06,
-      "loss": 0.9361,
+      "learning_rate": 5.020708824643066e-06,
+      "loss": 0.9185,
       "step": 23820
     },
     {
-      "epoch": 0.6542253714536815,
+      "epoch": 0.6759648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 5.642281398650751e-06,
-      "loss": 0.9064,
+      "learning_rate": 5.019911804876696e-06,
+      "loss": 0.8863,
       "step": 23821
     },
     {
-      "epoch": 0.6542528356815248,
+      "epoch": 0.6759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 5.641480800638279e-06,
-      "loss": 0.8359,
+      "learning_rate": 5.019114827177358e-06,
+      "loss": 0.7816,
       "step": 23822
     },
     {
-      "epoch": 0.6542802999093681,
+      "epoch": 0.6760215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 5.640680237112102e-06,
-      "loss": 0.8717,
+      "learning_rate": 5.018317891551777e-06,
+      "loss": 0.8681,
       "step": 23823
     },
     {
-      "epoch": 0.6543077641372113,
+      "epoch": 0.676049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 5.639879708078551e-06,
-      "loss": 0.8529,
+      "learning_rate": 5.01752099800669e-06,
+      "loss": 0.8,
       "step": 23824
     },
     {
-      "epoch": 0.6543352283650545,
+      "epoch": 0.6760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 5.6390792135439635e-06,
-      "loss": 0.9485,
+      "learning_rate": 5.016724146548829e-06,
+      "loss": 0.7757,
       "step": 23825
     },
     {
-      "epoch": 0.6543626925928977,
+      "epoch": 0.6761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 5.63827875351467e-06,
-      "loss": 0.9426,
+      "learning_rate": 5.01592733718492e-06,
+      "loss": 0.7443,
       "step": 23826
     },
     {
-      "epoch": 0.654390156820741,
+      "epoch": 0.6761350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 5.6374783279970015e-06,
-      "loss": 0.7828,
+      "learning_rate": 5.015130569921698e-06,
+      "loss": 0.9156,
       "step": 23827
     },
     {
-      "epoch": 0.6544176210485843,
+      "epoch": 0.6761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 5.636677936997294e-06,
-      "loss": 0.8086,
+      "learning_rate": 5.014333844765895e-06,
+      "loss": 0.9457,
       "step": 23828
     },
     {
-      "epoch": 0.6544450852764274,
+      "epoch": 0.6761918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 5.6358775805218825e-06,
-      "loss": 0.7821,
+      "learning_rate": 5.013537161724234e-06,
+      "loss": 0.7917,
       "step": 23829
     },
     {
-      "epoch": 0.6544725495042707,
+      "epoch": 0.6762202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 5.635077258577097e-06,
-      "loss": 0.8976,
+      "learning_rate": 5.01274052080345e-06,
+      "loss": 0.9008,
       "step": 23830
     },
     {
-      "epoch": 0.654500013732114,
+      "epoch": 0.6762485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 5.6342769711692744e-06,
-      "loss": 0.7588,
+      "learning_rate": 5.011943922010274e-06,
+      "loss": 0.8549,
       "step": 23831
     },
     {
-      "epoch": 0.6545274779599571,
+      "epoch": 0.6762769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 5.63347671830474e-06,
-      "loss": 0.8301,
+      "learning_rate": 5.011147365351427e-06,
+      "loss": 0.8673,
       "step": 23832
     },
     {
-      "epoch": 0.6545549421878004,
+      "epoch": 0.676305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 5.63267649998983e-06,
-      "loss": 0.8241,
+      "learning_rate": 5.010350850833645e-06,
+      "loss": 0.9442,
       "step": 23833
     },
     {
-      "epoch": 0.6545824064156436,
+      "epoch": 0.6763337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 5.631876316230878e-06,
-      "loss": 0.8404,
+      "learning_rate": 5.009554378463652e-06,
+      "loss": 0.8607,
       "step": 23834
     },
     {
-      "epoch": 0.6546098706434869,
+      "epoch": 0.6763620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 5.631076167034208e-06,
-      "loss": 0.8489,
+      "learning_rate": 5.008757948248177e-06,
+      "loss": 0.8736,
       "step": 23835
     },
     {
-      "epoch": 0.6546373348713301,
+      "epoch": 0.6763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 5.630276052406158e-06,
-      "loss": 0.7628,
+      "learning_rate": 5.007961560193952e-06,
+      "loss": 0.7517,
       "step": 23836
     },
     {
-      "epoch": 0.6546647990991733,
+      "epoch": 0.6764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 5.62947597235306e-06,
-      "loss": 0.7894,
+      "learning_rate": 5.007165214307699e-06,
+      "loss": 0.8498,
       "step": 23837
     },
     {
-      "epoch": 0.6546922633270166,
+      "epoch": 0.6764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 5.628675926881236e-06,
-      "loss": 0.8303,
+      "learning_rate": 5.006368910596142e-06,
+      "loss": 0.8514,
       "step": 23838
     },
     {
-      "epoch": 0.6547197275548597,
+      "epoch": 0.6764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 5.627875915997022e-06,
-      "loss": 0.8037,
+      "learning_rate": 5.005572649066017e-06,
+      "loss": 0.8353,
       "step": 23839
     },
     {
-      "epoch": 0.654747191782703,
+      "epoch": 0.6765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 5.627075939706747e-06,
-      "loss": 0.8083,
+      "learning_rate": 5.004776429724041e-06,
+      "loss": 0.9066,
       "step": 23840
     },
     {
-      "epoch": 0.6547746560105463,
+      "epoch": 0.6765323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 5.626275998016742e-06,
-      "loss": 0.8555,
+      "learning_rate": 5.003980252576942e-06,
+      "loss": 0.7499,
       "step": 23841
     },
     {
-      "epoch": 0.6548021202383895,
+      "epoch": 0.6765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 5.625476090933338e-06,
-      "loss": 0.8574,
+      "learning_rate": 5.003184117631451e-06,
+      "loss": 0.8356,
       "step": 23842
     },
     {
-      "epoch": 0.6548295844662327,
+      "epoch": 0.676589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 5.624676218462863e-06,
-      "loss": 0.8736,
+      "learning_rate": 5.002388024894285e-06,
+      "loss": 0.7999,
       "step": 23843
     },
     {
-      "epoch": 0.654857048694076,
+      "epoch": 0.6766174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 5.623876380611641e-06,
-      "loss": 0.8983,
+      "learning_rate": 5.001591974372171e-06,
+      "loss": 0.7255,
       "step": 23844
     },
     {
-      "epoch": 0.6548845129219192,
+      "epoch": 0.6766458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 5.623076577386002e-06,
-      "loss": 0.8314,
+      "learning_rate": 5.000795966071837e-06,
+      "loss": 0.8762,
       "step": 23845
     },
     {
-      "epoch": 0.6549119771497625,
+      "epoch": 0.6766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 5.622276808792278e-06,
-      "loss": 0.7937,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.871,
       "step": 23846
     },
     {
-      "epoch": 0.6549394413776056,
+      "epoch": 0.6767026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 5.6214770748367965e-06,
-      "loss": 0.8302,
+      "learning_rate": 4.999204076163392e-06,
+      "loss": 0.7698,
       "step": 23847
     },
     {
-      "epoch": 0.6549669056054489,
+      "epoch": 0.6767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 5.620677375525883e-06,
-      "loss": 0.8657,
+      "learning_rate": 4.998408194568734e-06,
+      "loss": 0.8016,
       "step": 23848
     },
     {
-      "epoch": 0.6549943698332922,
+      "epoch": 0.6767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 5.619877710865871e-06,
-      "loss": 0.8459,
+      "learning_rate": 4.997612355222738e-06,
+      "loss": 0.7672,
       "step": 23849
     },
     {
-      "epoch": 0.6550218340611353,
+      "epoch": 0.6767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 5.619078080863083e-06,
-      "loss": 0.7754,
+      "learning_rate": 4.996816558132139e-06,
+      "loss": 0.8046,
       "step": 23850
     },
     {
-      "epoch": 0.6550492982889786,
+      "epoch": 0.676816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 5.618278485523842e-06,
-      "loss": 0.758,
+      "learning_rate": 4.996020803303659e-06,
+      "loss": 0.8351,
       "step": 23851
     },
     {
-      "epoch": 0.6550767625168218,
+      "epoch": 0.676844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 5.61747892485448e-06,
-      "loss": 0.8335,
+      "learning_rate": 4.995225090744014e-06,
+      "loss": 0.8246,
       "step": 23852
     },
     {
-      "epoch": 0.6551042267446651,
+      "epoch": 0.6768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 5.616679398861322e-06,
-      "loss": 0.9013,
+      "learning_rate": 4.994429420459931e-06,
+      "loss": 0.8277,
       "step": 23853
     },
     {
-      "epoch": 0.6551316909725083,
+      "epoch": 0.6769012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 5.615879907550694e-06,
-      "loss": 0.826,
+      "learning_rate": 4.993633792458123e-06,
+      "loss": 0.833,
       "step": 23854
     },
     {
-      "epoch": 0.6551591552003515,
+      "epoch": 0.6769296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 5.615080450928927e-06,
-      "loss": 0.8073,
+      "learning_rate": 4.992838206745316e-06,
+      "loss": 0.8465,
       "step": 23855
     },
     {
-      "epoch": 0.6551866194281948,
+      "epoch": 0.6769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 5.614281029002337e-06,
-      "loss": 0.8623,
+      "learning_rate": 4.992042663328234e-06,
+      "loss": 0.7373,
       "step": 23856
     },
     {
-      "epoch": 0.6552140836560381,
+      "epoch": 0.6769863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 5.613481641777256e-06,
-      "loss": 0.9022,
+      "learning_rate": 4.991247162213588e-06,
+      "loss": 0.9135,
       "step": 23857
     },
     {
-      "epoch": 0.6552415478838812,
+      "epoch": 0.6770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 5.612682289260012e-06,
-      "loss": 0.8824,
+      "learning_rate": 4.990451703408104e-06,
+      "loss": 0.8677,
       "step": 23858
     },
     {
-      "epoch": 0.6552690121117245,
+      "epoch": 0.6770431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 5.611882971456919e-06,
-      "loss": 0.8408,
+      "learning_rate": 4.989656286918502e-06,
+      "loss": 0.8861,
       "step": 23859
     },
     {
-      "epoch": 0.6552964763395677,
+      "epoch": 0.6770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 5.611083688374308e-06,
-      "loss": 0.8551,
+      "learning_rate": 4.988860912751494e-06,
+      "loss": 0.8144,
       "step": 23860
     },
     {
-      "epoch": 0.655323940567411,
+      "epoch": 0.677099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 5.6102844400185074e-06,
-      "loss": 0.8349,
+      "learning_rate": 4.988065580913804e-06,
+      "loss": 0.7518,
       "step": 23861
     },
     {
-      "epoch": 0.6553514047952542,
+      "epoch": 0.677128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 5.6094852263958336e-06,
-      "loss": 0.8366,
+      "learning_rate": 4.987270291412154e-06,
+      "loss": 0.8728,
       "step": 23862
     },
     {
-      "epoch": 0.6553788690230974,
+      "epoch": 0.6771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 5.6086860475126126e-06,
-      "loss": 0.8659,
+      "learning_rate": 4.986475044253253e-06,
+      "loss": 0.8013,
       "step": 23863
     },
     {
-      "epoch": 0.6554063332509407,
+      "epoch": 0.6771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 5.607886903375168e-06,
-      "loss": 0.8027,
+      "learning_rate": 4.985679839443819e-06,
+      "loss": 0.8428,
       "step": 23864
     },
     {
-      "epoch": 0.6554337974787838,
+      "epoch": 0.6772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 5.607087793989829e-06,
-      "loss": 0.8114,
+      "learning_rate": 4.98488467699058e-06,
+      "loss": 0.8931,
       "step": 23865
     },
     {
-      "epoch": 0.6554612617066271,
+      "epoch": 0.6772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 5.606288719362908e-06,
-      "loss": 0.9886,
+      "learning_rate": 4.984089556900241e-06,
+      "loss": 0.8259,
       "step": 23866
     },
     {
-      "epoch": 0.6554887259344704,
+      "epoch": 0.6772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 5.605489679500737e-06,
-      "loss": 0.8549,
+      "learning_rate": 4.983294479179525e-06,
+      "loss": 0.8504,
       "step": 23867
     },
     {
-      "epoch": 0.6555161901623136,
+      "epoch": 0.6772985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 5.60469067440963e-06,
-      "loss": 0.799,
+      "learning_rate": 4.982499443835148e-06,
+      "loss": 0.8342,
       "step": 23868
     },
     {
-      "epoch": 0.6555436543901568,
+      "epoch": 0.6773269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 5.6038917040959115e-06,
-      "loss": 0.9097,
+      "learning_rate": 4.981704450873821e-06,
+      "loss": 0.7015,
       "step": 23869
     },
     {
-      "epoch": 0.6555711186180001,
+      "epoch": 0.677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 5.603092768565908e-06,
-      "loss": 0.9285,
+      "learning_rate": 4.980909500302261e-06,
+      "loss": 0.8713,
       "step": 23870
     },
     {
-      "epoch": 0.6555985828458433,
+      "epoch": 0.6773836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 5.602293867825934e-06,
-      "loss": 0.8384,
+      "learning_rate": 4.980114592127188e-06,
+      "loss": 0.8015,
       "step": 23871
     },
     {
-      "epoch": 0.6556260470736865,
+      "epoch": 0.6774120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 5.6014950018823166e-06,
-      "loss": 0.8676,
+      "learning_rate": 4.9793197263553105e-06,
+      "loss": 0.8378,
       "step": 23872
     },
     {
-      "epoch": 0.6556535113015297,
+      "epoch": 0.6774404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 5.600696170741377e-06,
-      "loss": 0.8595,
+      "learning_rate": 4.978524902993343e-06,
+      "loss": 0.8518,
       "step": 23873
     },
     {
-      "epoch": 0.655680975529373,
+      "epoch": 0.6774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 5.599897374409436e-06,
-      "loss": 0.7751,
+      "learning_rate": 4.977730122048007e-06,
+      "loss": 0.8181,
       "step": 23874
     },
     {
-      "epoch": 0.6557084397572163,
+      "epoch": 0.6774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 5.599098612892806e-06,
-      "loss": 0.8287,
+      "learning_rate": 4.976935383526006e-06,
+      "loss": 0.8254,
       "step": 23875
     },
     {
-      "epoch": 0.6557359039850594,
+      "epoch": 0.6775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 5.598299886197812e-06,
-      "loss": 0.9271,
+      "learning_rate": 4.9761406874340565e-06,
+      "loss": 0.8663,
       "step": 23876
     },
     {
-      "epoch": 0.6557633682129027,
+      "epoch": 0.6775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 5.597501194330774e-06,
-      "loss": 0.7932,
+      "learning_rate": 4.975346033778877e-06,
+      "loss": 0.8137,
       "step": 23877
     },
     {
-      "epoch": 0.6557908324407459,
+      "epoch": 0.6775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 5.596702537298018e-06,
-      "loss": 0.8277,
+      "learning_rate": 4.974551422567172e-06,
+      "loss": 0.8473,
       "step": 23878
     },
     {
-      "epoch": 0.6558182966685892,
+      "epoch": 0.6776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 5.595903915105851e-06,
-      "loss": 0.8168,
+      "learning_rate": 4.973756853805661e-06,
+      "loss": 0.9009,
       "step": 23879
     },
     {
-      "epoch": 0.6558457608964324,
+      "epoch": 0.677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 5.5951053277605995e-06,
-      "loss": 0.8986,
+      "learning_rate": 4.9729623275010444e-06,
+      "loss": 0.848,
       "step": 23880
     },
     {
-      "epoch": 0.6558732251242756,
+      "epoch": 0.6776674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 5.594306775268581e-06,
-      "loss": 0.8307,
+      "learning_rate": 4.9721678436600444e-06,
+      "loss": 0.8439,
       "step": 23881
     },
     {
-      "epoch": 0.6559006893521189,
+      "epoch": 0.6776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 5.593508257636118e-06,
-      "loss": 0.8245,
+      "learning_rate": 4.971373402289371e-06,
+      "loss": 0.7664,
       "step": 23882
     },
     {
-      "epoch": 0.6559281535799621,
+      "epoch": 0.6777241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 5.592709774869519e-06,
-      "loss": 0.8011,
+      "learning_rate": 4.97057900339573e-06,
+      "loss": 0.827,
       "step": 23883
     },
     {
-      "epoch": 0.6559556178078053,
+      "epoch": 0.6777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 5.591911326975112e-06,
-      "loss": 0.7488,
+      "learning_rate": 4.969784646985834e-06,
+      "loss": 0.7586,
       "step": 23884
     },
     {
-      "epoch": 0.6559830820356486,
+      "epoch": 0.6777809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 5.591112913959207e-06,
-      "loss": 0.9382,
+      "learning_rate": 4.968990333066398e-06,
+      "loss": 0.8526,
       "step": 23885
     },
     {
-      "epoch": 0.6560105462634918,
+      "epoch": 0.6778093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 5.590314535828125e-06,
-      "loss": 0.803,
+      "learning_rate": 4.968196061644123e-06,
+      "loss": 0.8571,
       "step": 23886
     },
     {
-      "epoch": 0.656038010491335,
+      "epoch": 0.6778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 5.58951619258818e-06,
-      "loss": 0.8277,
+      "learning_rate": 4.96740183272572e-06,
+      "loss": 0.7934,
       "step": 23887
     },
     {
-      "epoch": 0.6560654747191783,
+      "epoch": 0.6778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 5.588717884245693e-06,
-      "loss": 0.7912,
+      "learning_rate": 4.966607646317906e-06,
+      "loss": 0.8574,
       "step": 23888
     },
     {
-      "epoch": 0.6560929389470215,
+      "epoch": 0.677894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.5879196108069845e-06,
-      "loss": 0.9601,
+      "learning_rate": 4.965813502427378e-06,
+      "loss": 0.8533,
       "step": 23889
     },
     {
-      "epoch": 0.6561204031748648,
+      "epoch": 0.6779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 5.587121372278359e-06,
-      "loss": 0.8145,
+      "learning_rate": 4.965019401060851e-06,
+      "loss": 0.8401,
       "step": 23890
     },
     {
-      "epoch": 0.6561478674027079,
+      "epoch": 0.6779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 5.5863231686661435e-06,
-      "loss": 0.8508,
+      "learning_rate": 4.964225342225034e-06,
+      "loss": 0.9224,
       "step": 23891
     },
     {
-      "epoch": 0.6561753316305512,
+      "epoch": 0.6779795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 5.585524999976643e-06,
-      "loss": 0.8371,
+      "learning_rate": 4.9634313259266284e-06,
+      "loss": 0.8269,
       "step": 23892
     },
     {
-      "epoch": 0.6562027958583945,
+      "epoch": 0.6780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 5.584726866216181e-06,
-      "loss": 0.829,
+      "learning_rate": 4.9626373521723455e-06,
+      "loss": 0.784,
       "step": 23893
     },
     {
-      "epoch": 0.6562302600862376,
+      "epoch": 0.6780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 5.583928767391069e-06,
-      "loss": 0.8468,
+      "learning_rate": 4.961843420968895e-06,
+      "loss": 0.8447,
       "step": 23894
     },
     {
-      "epoch": 0.6562577243140809,
+      "epoch": 0.6780646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 5.583130703507624e-06,
-      "loss": 0.8282,
+      "learning_rate": 4.961049532322971e-06,
+      "loss": 0.9069,
       "step": 23895
     },
     {
-      "epoch": 0.6562851885419242,
+      "epoch": 0.6780930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 5.582332674572163e-06,
-      "loss": 0.8373,
+      "learning_rate": 4.960255686241298e-06,
+      "loss": 0.8072,
       "step": 23896
     },
     {
-      "epoch": 0.6563126527697674,
+      "epoch": 0.6781214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 5.581534680590995e-06,
-      "loss": 0.8837,
+      "learning_rate": 4.959461882730566e-06,
+      "loss": 0.9069,
       "step": 23897
     },
     {
-      "epoch": 0.6563401169976106,
+      "epoch": 0.678149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 5.580736721570436e-06,
-      "loss": 0.8629,
+      "learning_rate": 4.958668121797486e-06,
+      "loss": 0.7395,
       "step": 23898
     },
     {
-      "epoch": 0.6563675812254538,
+      "epoch": 0.678178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 5.5799387975168e-06,
-      "loss": 0.8269,
+      "learning_rate": 4.957874403448768e-06,
+      "loss": 0.8581,
       "step": 23899
     },
     {
-      "epoch": 0.6563950454532971,
+      "epoch": 0.6782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 5.579140908436399e-06,
-      "loss": 0.8594,
+      "learning_rate": 4.957080727691107e-06,
+      "loss": 0.8395,
       "step": 23900
     },
     {
-      "epoch": 0.6564225096811404,
+      "epoch": 0.6782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 5.578343054335548e-06,
-      "loss": 0.7822,
+      "learning_rate": 4.956287094531211e-06,
+      "loss": 0.7485,
       "step": 23901
     },
     {
-      "epoch": 0.6564499739089835,
+      "epoch": 0.6782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 5.5775452352205626e-06,
-      "loss": 0.8468,
+      "learning_rate": 4.9554935039757905e-06,
+      "loss": 0.8384,
       "step": 23902
     },
     {
-      "epoch": 0.6564774381368268,
+      "epoch": 0.6782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 5.576747451097748e-06,
-      "loss": 0.8578,
+      "learning_rate": 4.954699956031538e-06,
+      "loss": 0.8432,
       "step": 23903
     },
     {
-      "epoch": 0.65650490236467,
+      "epoch": 0.6783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 5.575949701973422e-06,
-      "loss": 0.853,
+      "learning_rate": 4.953906450705161e-06,
+      "loss": 0.8853,
       "step": 23904
     },
     {
-      "epoch": 0.6565323665925132,
+      "epoch": 0.6783484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 5.575151987853897e-06,
-      "loss": 0.7583,
+      "learning_rate": 4.953112988003368e-06,
+      "loss": 0.8038,
       "step": 23905
     },
     {
-      "epoch": 0.6565598308203565,
+      "epoch": 0.6783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 5.574354308745486e-06,
-      "loss": 0.8764,
+      "learning_rate": 4.952319567932853e-06,
+      "loss": 0.8613,
       "step": 23906
     },
     {
-      "epoch": 0.6565872950481997,
+      "epoch": 0.6784052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 5.573556664654494e-06,
-      "loss": 0.7635,
+      "learning_rate": 4.951526190500321e-06,
+      "loss": 0.8395,
       "step": 23907
     },
     {
-      "epoch": 0.656614759276043,
+      "epoch": 0.6784335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 5.57275905558724e-06,
-      "loss": 0.882,
+      "learning_rate": 4.950732855712478e-06,
+      "loss": 0.7348,
       "step": 23908
     },
     {
-      "epoch": 0.6566422235038862,
+      "epoch": 0.6784619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 5.571961481550028e-06,
-      "loss": 0.7984,
+      "learning_rate": 4.9499395635760174e-06,
+      "loss": 0.9005,
       "step": 23909
     },
     {
-      "epoch": 0.6566696877317294,
+      "epoch": 0.6784903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 5.571163942549171e-06,
-      "loss": 0.8166,
+      "learning_rate": 4.949146314097645e-06,
+      "loss": 0.8155,
       "step": 23910
     },
     {
-      "epoch": 0.6566971519595727,
+      "epoch": 0.6785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 5.570366438590981e-06,
-      "loss": 0.8912,
+      "learning_rate": 4.94835310728406e-06,
+      "loss": 0.8196,
       "step": 23911
     },
     {
-      "epoch": 0.6567246161874158,
+      "epoch": 0.6785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 5.569568969681769e-06,
-      "loss": 0.8521,
+      "learning_rate": 4.947559943141964e-06,
+      "loss": 0.7896,
       "step": 23912
     },
     {
-      "epoch": 0.6567520804152591,
+      "epoch": 0.6785754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 5.568771535827846e-06,
-      "loss": 0.8252,
+      "learning_rate": 4.94676682167806e-06,
+      "loss": 0.7996,
       "step": 23913
     },
     {
-      "epoch": 0.6567795446431024,
+      "epoch": 0.6786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 5.567974137035516e-06,
-      "loss": 0.8765,
+      "learning_rate": 4.9459737428990395e-06,
+      "loss": 0.8713,
       "step": 23914
     },
     {
-      "epoch": 0.6568070088709456,
+      "epoch": 0.6786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 5.567176773311093e-06,
-      "loss": 0.888,
+      "learning_rate": 4.9451807068116064e-06,
+      "loss": 0.8707,
       "step": 23915
     },
     {
-      "epoch": 0.6568344730987888,
+      "epoch": 0.678660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 5.566379444660883e-06,
-      "loss": 0.7876,
+      "learning_rate": 4.944387713422463e-06,
+      "loss": 0.864,
       "step": 23916
     },
     {
-      "epoch": 0.6568619373266321,
+      "epoch": 0.678688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 5.565582151091196e-06,
-      "loss": 0.8302,
+      "learning_rate": 4.9435947627383e-06,
+      "loss": 0.7876,
       "step": 23917
     },
     {
-      "epoch": 0.6568894015544753,
+      "epoch": 0.6787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 5.56478489260834e-06,
-      "loss": 0.9038,
+      "learning_rate": 4.94280185476582e-06,
+      "loss": 0.8082,
       "step": 23918
     },
     {
-      "epoch": 0.6569168657823186,
+      "epoch": 0.6787457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 5.563987669218625e-06,
-      "loss": 0.9002,
+      "learning_rate": 4.942008989511723e-06,
+      "loss": 0.7047,
       "step": 23919
     },
     {
-      "epoch": 0.6569443300101617,
+      "epoch": 0.6787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 5.563190480928361e-06,
-      "loss": 0.8207,
+      "learning_rate": 4.9412161669826995e-06,
+      "loss": 0.7767,
       "step": 23920
     },
     {
-      "epoch": 0.656971794238005,
+      "epoch": 0.6788024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 5.5623933277438494e-06,
-      "loss": 0.7536,
+      "learning_rate": 4.94042338718545e-06,
+      "loss": 0.8085,
       "step": 23921
     },
     {
-      "epoch": 0.6569992584658483,
+      "epoch": 0.6788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 5.561596209671403e-06,
-      "loss": 0.8625,
+      "learning_rate": 4.9396306501266764e-06,
+      "loss": 0.819,
       "step": 23922
     },
     {
-      "epoch": 0.6570267226936914,
+      "epoch": 0.6788592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 5.560799126717325e-06,
-      "loss": 0.7966,
+      "learning_rate": 4.938837955813066e-06,
+      "loss": 0.7859,
       "step": 23923
     },
     {
-      "epoch": 0.6570541869215347,
+      "epoch": 0.6788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 5.560002078887922e-06,
-      "loss": 0.6946,
+      "learning_rate": 4.938045304251319e-06,
+      "loss": 0.8704,
       "step": 23924
     },
     {
-      "epoch": 0.6570816511493779,
+      "epoch": 0.6789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 5.559205066189504e-06,
-      "loss": 0.8506,
+      "learning_rate": 4.937252695448135e-06,
+      "loss": 0.9087,
       "step": 23925
     },
     {
-      "epoch": 0.6571091153772212,
+      "epoch": 0.67894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.558408088628378e-06,
-      "loss": 0.8834,
+      "learning_rate": 4.936460129410197e-06,
+      "loss": 0.8982,
       "step": 23926
     },
     {
-      "epoch": 0.6571365796050644,
+      "epoch": 0.6789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 5.557611146210844e-06,
-      "loss": 0.9362,
+      "learning_rate": 4.9356676061442154e-06,
+      "loss": 0.7449,
       "step": 23927
     },
     {
-      "epoch": 0.6571640438329076,
+      "epoch": 0.6790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 5.556814238943212e-06,
-      "loss": 0.8177,
+      "learning_rate": 4.9348751256568735e-06,
+      "loss": 0.8231,
       "step": 23928
     },
     {
-      "epoch": 0.6571915080607509,
+      "epoch": 0.6790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 5.5560173668317876e-06,
-      "loss": 0.8328,
+      "learning_rate": 4.93408268795487e-06,
+      "loss": 0.8276,
       "step": 23929
     },
     {
-      "epoch": 0.6572189722885942,
+      "epoch": 0.6790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 5.5552205298828786e-06,
-      "loss": 0.754,
+      "learning_rate": 4.9332902930449e-06,
+      "loss": 0.8685,
       "step": 23930
     },
     {
-      "epoch": 0.6572464365164373,
+      "epoch": 0.6790862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 5.554423728102783e-06,
-      "loss": 0.8867,
+      "learning_rate": 4.932497940933653e-06,
+      "loss": 0.8289,
       "step": 23931
     },
     {
-      "epoch": 0.6572739007442806,
+      "epoch": 0.6791146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 5.553626961497811e-06,
-      "loss": 0.8972,
+      "learning_rate": 4.931705631627821e-06,
+      "loss": 0.7536,
       "step": 23932
     },
     {
-      "epoch": 0.6573013649721238,
+      "epoch": 0.6791430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 5.552830230074262e-06,
-      "loss": 0.8693,
+      "learning_rate": 4.9309133651341046e-06,
+      "loss": 0.8182,
       "step": 23933
     },
     {
-      "epoch": 0.657328829199967,
+      "epoch": 0.6791713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 5.552033533838443e-06,
-      "loss": 0.9183,
+      "learning_rate": 4.9301211414591875e-06,
+      "loss": 0.8555,
       "step": 23934
     },
     {
-      "epoch": 0.6573562934278103,
+      "epoch": 0.679199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 5.5512368727966584e-06,
-      "loss": 0.8118,
+      "learning_rate": 4.929328960609764e-06,
+      "loss": 0.8343,
       "step": 23935
     },
     {
-      "epoch": 0.6573837576556535,
+      "epoch": 0.6792281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 5.55044024695521e-06,
-      "loss": 0.8404,
+      "learning_rate": 4.928536822592531e-06,
+      "loss": 0.9438,
       "step": 23936
     },
     {
-      "epoch": 0.6574112218834968,
+      "epoch": 0.6792565266742339,
       "grad_norm": 0.0,
-      "learning_rate": 5.549643656320406e-06,
-      "loss": 0.8761,
+      "learning_rate": 4.927744727414172e-06,
+      "loss": 0.7287,
       "step": 23937
     },
     {
-      "epoch": 0.6574386861113399,
+      "epoch": 0.6792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 5.548847100898542e-06,
-      "loss": 0.9109,
+      "learning_rate": 4.926952675081381e-06,
+      "loss": 0.8457,
       "step": 23938
     },
     {
-      "epoch": 0.6574661503391832,
+      "epoch": 0.6793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 5.5480505806959275e-06,
-      "loss": 0.8081,
+      "learning_rate": 4.926160665600853e-06,
+      "loss": 0.7692,
       "step": 23939
     },
     {
-      "epoch": 0.6574936145670265,
+      "epoch": 0.6793416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 5.547254095718857e-06,
-      "loss": 0.9033,
+      "learning_rate": 4.925368698979269e-06,
+      "loss": 0.7421,
       "step": 23940
     },
     {
-      "epoch": 0.6575210787948697,
+      "epoch": 0.6793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 5.546457645973638e-06,
-      "loss": 0.8547,
+      "learning_rate": 4.924576775223326e-06,
+      "loss": 0.9438,
       "step": 23941
     },
     {
-      "epoch": 0.6575485430227129,
+      "epoch": 0.6793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 5.54566123146657e-06,
-      "loss": 0.8788,
+      "learning_rate": 4.923784894339708e-06,
+      "loss": 0.7565,
       "step": 23942
     },
     {
-      "epoch": 0.6575760072505562,
+      "epoch": 0.6794267877412031,
       "grad_norm": 0.0,
-      "learning_rate": 5.54486485220396e-06,
-      "loss": 0.9037,
+      "learning_rate": 4.9229930563351095e-06,
+      "loss": 0.8852,
       "step": 23943
     },
     {
-      "epoch": 0.6576034714783994,
+      "epoch": 0.6794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 5.544068508192101e-06,
-      "loss": 0.8857,
+      "learning_rate": 4.92220126121622e-06,
+      "loss": 0.8732,
       "step": 23944
     },
     {
-      "epoch": 0.6576309357062426,
+      "epoch": 0.679483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 5.543272199437298e-06,
-      "loss": 0.8846,
+      "learning_rate": 4.921409508989721e-06,
+      "loss": 0.958,
       "step": 23945
     },
     {
-      "epoch": 0.6576583999340858,
+      "epoch": 0.6795119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 5.542475925945856e-06,
-      "loss": 0.7797,
+      "learning_rate": 4.920617799662304e-06,
+      "loss": 0.7738,
       "step": 23946
     },
     {
-      "epoch": 0.6576858641619291,
+      "epoch": 0.6795402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 5.541679687724068e-06,
-      "loss": 0.8174,
+      "learning_rate": 4.91982613324066e-06,
+      "loss": 0.8776,
       "step": 23947
     },
     {
-      "epoch": 0.6577133283897724,
+      "epoch": 0.6795686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 5.5408834847782365e-06,
-      "loss": 0.8433,
+      "learning_rate": 4.919034509731469e-06,
+      "loss": 0.8743,
       "step": 23948
     },
     {
-      "epoch": 0.6577407926176155,
+      "epoch": 0.6795970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 5.540087317114666e-06,
-      "loss": 0.8677,
+      "learning_rate": 4.918242929141423e-06,
+      "loss": 0.7738,
       "step": 23949
     },
     {
-      "epoch": 0.6577682568454588,
+      "epoch": 0.6796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 5.539291184739648e-06,
-      "loss": 0.8216,
+      "learning_rate": 4.917451391477212e-06,
+      "loss": 0.7701,
       "step": 23950
     },
     {
-      "epoch": 0.657795721073302,
+      "epoch": 0.6796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 5.5384950876594855e-06,
-      "loss": 0.8573,
+      "learning_rate": 4.916659896745513e-06,
+      "loss": 0.8357,
       "step": 23951
     },
     {
-      "epoch": 0.6578231853011453,
+      "epoch": 0.6796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 5.537699025880479e-06,
-      "loss": 0.8647,
+      "learning_rate": 4.9158684449530155e-06,
+      "loss": 0.7835,
       "step": 23952
     },
     {
-      "epoch": 0.6578506495289885,
+      "epoch": 0.6797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.536902999408927e-06,
-      "loss": 0.8897,
+      "learning_rate": 4.91507703610641e-06,
+      "loss": 0.8189,
       "step": 23953
     },
     {
-      "epoch": 0.6578781137568317,
+      "epoch": 0.679738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.53610700825113e-06,
-      "loss": 0.8779,
+      "learning_rate": 4.9142856702123745e-06,
+      "loss": 0.7577,
       "step": 23954
     },
     {
-      "epoch": 0.657905577984675,
+      "epoch": 0.6797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 5.535311052413386e-06,
-      "loss": 0.9011,
+      "learning_rate": 4.913494347277595e-06,
+      "loss": 0.8112,
       "step": 23955
     },
     {
-      "epoch": 0.6579330422125182,
+      "epoch": 0.6797956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 5.5345151319019826e-06,
-      "loss": 0.8813,
+      "learning_rate": 4.9127030673087585e-06,
+      "loss": 0.8071,
       "step": 23956
     },
     {
-      "epoch": 0.6579605064403614,
+      "epoch": 0.6798240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 5.533719246723228e-06,
-      "loss": 0.8171,
+      "learning_rate": 4.911911830312548e-06,
+      "loss": 0.8659,
       "step": 23957
     },
     {
-      "epoch": 0.6579879706682047,
+      "epoch": 0.6798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 5.532923396883416e-06,
-      "loss": 0.8798,
+      "learning_rate": 4.911120636295647e-06,
+      "loss": 0.6601,
       "step": 23958
     },
     {
-      "epoch": 0.6580154348960479,
+      "epoch": 0.6798808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 5.532127582388844e-06,
-      "loss": 0.8383,
+      "learning_rate": 4.910329485264741e-06,
+      "loss": 0.7847,
       "step": 23959
     },
     {
-      "epoch": 0.6580428991238911,
+      "epoch": 0.6799091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 5.531331803245811e-06,
-      "loss": 0.6954,
+      "learning_rate": 4.909538377226508e-06,
+      "loss": 0.7713,
       "step": 23960
     },
     {
-      "epoch": 0.6580703633517344,
+      "epoch": 0.6799375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 5.530536059460614e-06,
-      "loss": 0.8715,
+      "learning_rate": 4.908747312187637e-06,
+      "loss": 0.9011,
       "step": 23961
     },
     {
-      "epoch": 0.6580978275795776,
+      "epoch": 0.6799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 5.529740351039543e-06,
-      "loss": 0.8701,
+      "learning_rate": 4.907956290154803e-06,
+      "loss": 0.8597,
       "step": 23962
     },
     {
-      "epoch": 0.6581252918074209,
+      "epoch": 0.679994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 5.528944677988903e-06,
-      "loss": 0.8503,
+      "learning_rate": 4.90716531113469e-06,
+      "loss": 0.8889,
       "step": 23963
     },
     {
-      "epoch": 0.658152756035264,
+      "epoch": 0.680022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 5.528149040314981e-06,
-      "loss": 0.8551,
+      "learning_rate": 4.906374375133985e-06,
+      "loss": 0.9119,
       "step": 23964
     },
     {
-      "epoch": 0.6581802202631073,
+      "epoch": 0.6800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 5.527353438024075e-06,
-      "loss": 0.8022,
+      "learning_rate": 4.905583482159361e-06,
+      "loss": 0.8665,
       "step": 23965
     },
     {
-      "epoch": 0.6582076844909506,
+      "epoch": 0.6800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 5.52655787112248e-06,
-      "loss": 0.7447,
+      "learning_rate": 4.904792632217502e-06,
+      "loss": 0.8865,
       "step": 23966
     },
     {
-      "epoch": 0.6582351487187937,
+      "epoch": 0.6801078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 5.525762339616498e-06,
-      "loss": 0.8418,
+      "learning_rate": 4.904001825315092e-06,
+      "loss": 0.8492,
       "step": 23967
     },
     {
-      "epoch": 0.658262612946637,
+      "epoch": 0.6801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 5.524966843512412e-06,
-      "loss": 0.7912,
+      "learning_rate": 4.903211061458805e-06,
+      "loss": 0.7888,
       "step": 23968
     },
     {
-      "epoch": 0.6582900771744803,
+      "epoch": 0.6801645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 5.524171382816523e-06,
-      "loss": 0.8282,
+      "learning_rate": 4.902420340655323e-06,
+      "loss": 0.802,
       "step": 23969
     },
     {
-      "epoch": 0.6583175414023235,
+      "epoch": 0.6801929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 5.523375957535128e-06,
-      "loss": 0.8424,
+      "learning_rate": 4.9016296629113296e-06,
+      "loss": 0.9424,
       "step": 23970
     },
     {
-      "epoch": 0.6583450056301667,
+      "epoch": 0.6802213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 5.522580567674512e-06,
-      "loss": 0.9325,
+      "learning_rate": 4.900839028233494e-06,
+      "loss": 0.8203,
       "step": 23971
     },
     {
-      "epoch": 0.6583724698580099,
+      "epoch": 0.6802497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 5.521785213240972e-06,
-      "loss": 0.8266,
+      "learning_rate": 4.900048436628498e-06,
+      "loss": 0.907,
       "step": 23972
     },
     {
-      "epoch": 0.6583999340858532,
+      "epoch": 0.6802780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 5.5209898942408065e-06,
-      "loss": 0.8232,
+      "learning_rate": 4.899257888103029e-06,
+      "loss": 0.9477,
       "step": 23973
     },
     {
-      "epoch": 0.6584273983136965,
+      "epoch": 0.6803064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 5.5201946106803e-06,
-      "loss": 0.7935,
+      "learning_rate": 4.898467382663756e-06,
+      "loss": 0.9076,
       "step": 23974
     },
     {
-      "epoch": 0.6584548625415396,
+      "epoch": 0.6803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 5.5193993625657495e-06,
-      "loss": 0.8965,
+      "learning_rate": 4.897676920317357e-06,
+      "loss": 0.7551,
       "step": 23975
     },
     {
-      "epoch": 0.6584823267693829,
+      "epoch": 0.6803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 5.518604149903447e-06,
-      "loss": 0.8689,
+      "learning_rate": 4.8968865010705135e-06,
+      "loss": 0.8585,
       "step": 23976
     },
     {
-      "epoch": 0.6585097909972261,
+      "epoch": 0.6803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 5.517808972699684e-06,
-      "loss": 0.9109,
+      "learning_rate": 4.896096124929897e-06,
+      "loss": 0.9225,
       "step": 23977
     },
     {
-      "epoch": 0.6585372552250693,
+      "epoch": 0.6804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 5.517013830960757e-06,
-      "loss": 0.764,
+      "learning_rate": 4.895305791902184e-06,
+      "loss": 0.7509,
       "step": 23978
     },
     {
-      "epoch": 0.6585647194529126,
+      "epoch": 0.6804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 5.516218724692952e-06,
-      "loss": 0.8965,
+      "learning_rate": 4.894515501994057e-06,
+      "loss": 0.9226,
       "step": 23979
     },
     {
-      "epoch": 0.6585921836807558,
+      "epoch": 0.6804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 5.515423653902558e-06,
-      "loss": 0.8156,
+      "learning_rate": 4.8937252552121835e-06,
+      "loss": 0.9074,
       "step": 23980
     },
     {
-      "epoch": 0.6586196479085991,
+      "epoch": 0.6805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 5.51462861859587e-06,
-      "loss": 0.9124,
+      "learning_rate": 4.892935051563243e-06,
+      "loss": 0.8031,
       "step": 23981
     },
     {
-      "epoch": 0.6586471121364423,
+      "epoch": 0.680533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 5.513833618779177e-06,
-      "loss": 0.8767,
+      "learning_rate": 4.892144891053912e-06,
+      "loss": 0.746,
       "step": 23982
     },
     {
-      "epoch": 0.6586745763642855,
+      "epoch": 0.6805618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 5.513038654458771e-06,
-      "loss": 0.8597,
+      "learning_rate": 4.891354773690861e-06,
+      "loss": 0.8511,
       "step": 23983
     },
     {
-      "epoch": 0.6587020405921288,
+      "epoch": 0.6805902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 5.51224372564094e-06,
-      "loss": 0.7831,
+      "learning_rate": 4.890564699480764e-06,
+      "loss": 0.8709,
       "step": 23984
     },
     {
-      "epoch": 0.658729504819972,
+      "epoch": 0.6806186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 5.5114488323319806e-06,
-      "loss": 0.8071,
+      "learning_rate": 4.889774668430301e-06,
+      "loss": 0.9465,
       "step": 23985
     },
     {
-      "epoch": 0.6587569690478152,
+      "epoch": 0.6806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 5.510653974538176e-06,
-      "loss": 0.8288,
+      "learning_rate": 4.888984680546137e-06,
+      "loss": 0.9562,
       "step": 23986
     },
     {
-      "epoch": 0.6587844332756585,
+      "epoch": 0.6806753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 5.509859152265812e-06,
-      "loss": 0.8928,
+      "learning_rate": 4.88819473583495e-06,
+      "loss": 0.7712,
       "step": 23987
     },
     {
-      "epoch": 0.6588118975035017,
+      "epoch": 0.6807037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 5.509064365521182e-06,
-      "loss": 0.821,
+      "learning_rate": 4.887404834303412e-06,
+      "loss": 0.8564,
       "step": 23988
     },
     {
-      "epoch": 0.6588393617313449,
+      "epoch": 0.6807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 5.508269614310573e-06,
-      "loss": 0.837,
+      "learning_rate": 4.886614975958194e-06,
+      "loss": 0.8529,
       "step": 23989
     },
     {
-      "epoch": 0.6588668259591881,
+      "epoch": 0.6807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 5.507474898640276e-06,
-      "loss": 0.8557,
+      "learning_rate": 4.8858251608059735e-06,
+      "loss": 0.7801,
       "step": 23990
     },
     {
-      "epoch": 0.6588942901870314,
+      "epoch": 0.680788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 5.506680218516583e-06,
-      "loss": 0.7943,
+      "learning_rate": 4.885035388853414e-06,
+      "loss": 0.7812,
       "step": 23991
     },
     {
-      "epoch": 0.6589217544148747,
+      "epoch": 0.6808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 5.505885573945771e-06,
-      "loss": 0.8434,
+      "learning_rate": 4.884245660107191e-06,
+      "loss": 0.902,
       "step": 23992
     },
     {
-      "epoch": 0.6589492186427178,
+      "epoch": 0.6808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 5.505090964934132e-06,
-      "loss": 0.8698,
+      "learning_rate": 4.883455974573979e-06,
+      "loss": 0.886,
       "step": 23993
     },
     {
-      "epoch": 0.6589766828705611,
+      "epoch": 0.6808740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 5.50429639148796e-06,
-      "loss": 0.8399,
+      "learning_rate": 4.882666332260438e-06,
+      "loss": 0.931,
       "step": 23994
     },
     {
-      "epoch": 0.6590041470984044,
+      "epoch": 0.6809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 5.503501853613533e-06,
-      "loss": 0.8307,
+      "learning_rate": 4.881876733173248e-06,
+      "loss": 0.9155,
       "step": 23995
     },
     {
-      "epoch": 0.6590316113262475,
+      "epoch": 0.6809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 5.502707351317144e-06,
-      "loss": 0.8162,
+      "learning_rate": 4.8810871773190774e-06,
+      "loss": 0.7317,
       "step": 23996
     },
     {
-      "epoch": 0.6590590755540908,
+      "epoch": 0.6809591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 5.501912884605073e-06,
-      "loss": 0.8799,
+      "learning_rate": 4.880297664704589e-06,
+      "loss": 0.7905,
       "step": 23997
     },
     {
-      "epoch": 0.659086539781934,
+      "epoch": 0.6809875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 5.501118453483609e-06,
-      "loss": 0.8646,
+      "learning_rate": 4.879508195336457e-06,
+      "loss": 0.781,
       "step": 23998
     },
     {
-      "epoch": 0.6591140040097773,
+      "epoch": 0.6810158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 5.50032405795904e-06,
-      "loss": 0.7831,
+      "learning_rate": 4.878718769221354e-06,
+      "loss": 0.8555,
       "step": 23999
     },
     {
-      "epoch": 0.6591414682376205,
+      "epoch": 0.681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 5.499529698037649e-06,
-      "loss": 0.904,
+      "learning_rate": 4.877929386365939e-06,
+      "loss": 0.8019,
       "step": 24000
     },
     {
-      "epoch": 0.6591689324654637,
+      "epoch": 0.6810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 5.498735373725723e-06,
-      "loss": 0.8353,
+      "learning_rate": 4.877140046776885e-06,
+      "loss": 0.9298,
       "step": 24001
     },
     {
-      "epoch": 0.659196396693307,
+      "epoch": 0.6811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 5.49794108502955e-06,
-      "loss": 0.8555,
+      "learning_rate": 4.8763507504608595e-06,
+      "loss": 0.972,
       "step": 24002
     },
     {
-      "epoch": 0.6592238609211502,
+      "epoch": 0.6811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 5.497146831955411e-06,
-      "loss": 0.8193,
+      "learning_rate": 4.8755614974245295e-06,
+      "loss": 0.9087,
       "step": 24003
     },
     {
-      "epoch": 0.6592513251489934,
+      "epoch": 0.6811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.496352614509586e-06,
-      "loss": 0.9129,
+      "learning_rate": 4.874772287674566e-06,
+      "loss": 0.873,
       "step": 24004
     },
     {
-      "epoch": 0.6592787893768367,
+      "epoch": 0.6811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 5.495558432698365e-06,
-      "loss": 0.8724,
+      "learning_rate": 4.873983121217627e-06,
+      "loss": 0.8539,
       "step": 24005
     },
     {
-      "epoch": 0.6593062536046799,
+      "epoch": 0.6812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 5.494764286528029e-06,
-      "loss": 0.8296,
+      "learning_rate": 4.873193998060382e-06,
+      "loss": 0.8004,
       "step": 24006
     },
     {
-      "epoch": 0.6593337178325231,
+      "epoch": 0.6812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 5.493970176004863e-06,
-      "loss": 0.859,
+      "learning_rate": 4.872404918209504e-06,
+      "loss": 0.8187,
       "step": 24007
     },
     {
-      "epoch": 0.6593611820603664,
+      "epoch": 0.6812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 5.493176101135156e-06,
-      "loss": 0.8493,
+      "learning_rate": 4.8716158816716476e-06,
+      "loss": 0.845,
       "step": 24008
     },
     {
-      "epoch": 0.6593886462882096,
+      "epoch": 0.6812996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 5.49238206192518e-06,
-      "loss": 0.8134,
+      "learning_rate": 4.870826888453481e-06,
+      "loss": 0.8703,
       "step": 24009
     },
     {
-      "epoch": 0.6594161105160529,
+      "epoch": 0.6813280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 5.491588058381227e-06,
-      "loss": 0.815,
+      "learning_rate": 4.870037938561676e-06,
+      "loss": 0.8686,
       "step": 24010
     },
     {
-      "epoch": 0.659443574743896,
+      "epoch": 0.6813564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 5.490794090509574e-06,
-      "loss": 0.8743,
+      "learning_rate": 4.8692490320028866e-06,
+      "loss": 0.7912,
       "step": 24011
     },
     {
-      "epoch": 0.6594710389717393,
+      "epoch": 0.6813847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 5.490000158316503e-06,
-      "loss": 0.9306,
+      "learning_rate": 4.86846016878378e-06,
+      "loss": 0.9021,
       "step": 24012
     },
     {
-      "epoch": 0.6594985031995826,
+      "epoch": 0.6814131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 5.489206261808298e-06,
-      "loss": 0.8778,
+      "learning_rate": 4.867671348911026e-06,
+      "loss": 0.8897,
       "step": 24013
     },
     {
-      "epoch": 0.6595259674274258,
+      "epoch": 0.6814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 5.488412400991245e-06,
-      "loss": 0.7282,
+      "learning_rate": 4.866882572391279e-06,
+      "loss": 0.8923,
       "step": 24014
     },
     {
-      "epoch": 0.659553431655269,
+      "epoch": 0.6814699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 5.487618575871616e-06,
-      "loss": 0.8269,
+      "learning_rate": 4.866093839231205e-06,
+      "loss": 0.7966,
       "step": 24015
     },
     {
-      "epoch": 0.6595808958831122,
+      "epoch": 0.6814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 5.4868247864556966e-06,
-      "loss": 0.9022,
+      "learning_rate": 4.865305149437471e-06,
+      "loss": 0.8882,
       "step": 24016
     },
     {
-      "epoch": 0.6596083601109555,
+      "epoch": 0.6815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 5.486031032749768e-06,
-      "loss": 0.6921,
+      "learning_rate": 4.864516503016733e-06,
+      "loss": 0.7567,
       "step": 24017
     },
     {
-      "epoch": 0.6596358243387987,
+      "epoch": 0.68155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 5.485237314760115e-06,
-      "loss": 0.9032,
+      "learning_rate": 4.863727899975653e-06,
+      "loss": 0.884,
       "step": 24018
     },
     {
-      "epoch": 0.6596632885666419,
+      "epoch": 0.681583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 5.484443632493008e-06,
-      "loss": 0.8454,
+      "learning_rate": 4.8629393403208955e-06,
+      "loss": 0.8486,
       "step": 24019
     },
     {
-      "epoch": 0.6596907527944852,
+      "epoch": 0.6816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 5.483649985954737e-06,
-      "loss": 0.8456,
+      "learning_rate": 4.86215082405912e-06,
+      "loss": 0.8331,
       "step": 24020
     },
     {
-      "epoch": 0.6597182170223285,
+      "epoch": 0.6816401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 5.482856375151573e-06,
-      "loss": 0.809,
+      "learning_rate": 4.861362351196991e-06,
+      "loss": 0.8162,
       "step": 24021
     },
     {
-      "epoch": 0.6597456812501716,
+      "epoch": 0.6816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 5.482062800089799e-06,
-      "loss": 0.8378,
+      "learning_rate": 4.860573921741162e-06,
+      "loss": 0.8248,
       "step": 24022
     },
     {
-      "epoch": 0.6597731454780149,
+      "epoch": 0.6816969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 5.481269260775694e-06,
-      "loss": 0.8185,
+      "learning_rate": 4.859785535698296e-06,
+      "loss": 0.8305,
       "step": 24023
     },
     {
-      "epoch": 0.6598006097058581,
+      "epoch": 0.6817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 5.480475757215537e-06,
-      "loss": 0.9426,
+      "learning_rate": 4.8589971930750554e-06,
+      "loss": 0.9451,
       "step": 24024
     },
     {
-      "epoch": 0.6598280739337014,
+      "epoch": 0.6817536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 5.479682289415611e-06,
-      "loss": 0.8549,
+      "learning_rate": 4.858208893878095e-06,
+      "loss": 0.9411,
       "step": 24025
     },
     {
-      "epoch": 0.6598555381615446,
+      "epoch": 0.6817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 5.478888857382185e-06,
-      "loss": 0.6932,
+      "learning_rate": 4.857420638114073e-06,
+      "loss": 0.8359,
       "step": 24026
     },
     {
-      "epoch": 0.6598830023893878,
+      "epoch": 0.6818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 5.478095461121546e-06,
-      "loss": 0.9641,
+      "learning_rate": 4.8566324257896545e-06,
+      "loss": 0.8981,
       "step": 24027
     },
     {
-      "epoch": 0.6599104666172311,
+      "epoch": 0.681838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 5.477302100639963e-06,
-      "loss": 0.7991,
+      "learning_rate": 4.855844256911489e-06,
+      "loss": 0.858,
       "step": 24028
     },
     {
-      "epoch": 0.6599379308450742,
+      "epoch": 0.6818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 5.4765087759437185e-06,
-      "loss": 0.8011,
+      "learning_rate": 4.85505613148624e-06,
+      "loss": 0.7618,
       "step": 24029
     },
     {
-      "epoch": 0.6599653950729175,
+      "epoch": 0.6818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 5.475715487039088e-06,
-      "loss": 0.965,
+      "learning_rate": 4.854268049520565e-06,
+      "loss": 0.7906,
       "step": 24030
     },
     {
-      "epoch": 0.6599928593007608,
+      "epoch": 0.6819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 5.474922233932349e-06,
-      "loss": 0.8315,
+      "learning_rate": 4.853480011021116e-06,
+      "loss": 0.8678,
       "step": 24031
     },
     {
-      "epoch": 0.660020323528604,
+      "epoch": 0.6819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 5.474129016629783e-06,
-      "loss": 0.7749,
+      "learning_rate": 4.8526920159945536e-06,
+      "loss": 0.747,
       "step": 24032
     },
     {
-      "epoch": 0.6600477877564472,
+      "epoch": 0.6819807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 5.473335835137657e-06,
-      "loss": 0.8736,
+      "learning_rate": 4.851904064447531e-06,
+      "loss": 0.8899,
       "step": 24033
     },
     {
-      "epoch": 0.6600752519842905,
+      "epoch": 0.6820090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 5.4725426894622545e-06,
-      "loss": 0.8443,
+      "learning_rate": 4.851116156386708e-06,
+      "loss": 0.7891,
       "step": 24034
     },
     {
-      "epoch": 0.6601027162121337,
+      "epoch": 0.6820374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 5.471749579609844e-06,
-      "loss": 0.9482,
+      "learning_rate": 4.850328291818742e-06,
+      "loss": 0.8651,
       "step": 24035
     },
     {
-      "epoch": 0.660130180439977,
+      "epoch": 0.6820658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 5.4709565055867066e-06,
-      "loss": 0.8759,
+      "learning_rate": 4.84954047075028e-06,
+      "loss": 0.836,
       "step": 24036
     },
     {
-      "epoch": 0.6601576446678201,
+      "epoch": 0.682094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 5.470163467399115e-06,
-      "loss": 0.7769,
+      "learning_rate": 4.8487526931879815e-06,
+      "loss": 0.8702,
       "step": 24037
     },
     {
-      "epoch": 0.6601851088956634,
+      "epoch": 0.6821225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 5.469370465053347e-06,
-      "loss": 0.7259,
+      "learning_rate": 4.847964959138504e-06,
+      "loss": 0.7987,
       "step": 24038
     },
     {
-      "epoch": 0.6602125731235067,
+      "epoch": 0.6821509648127129,
       "grad_norm": 0.0,
-      "learning_rate": 5.468577498555674e-06,
-      "loss": 0.8062,
+      "learning_rate": 4.847177268608494e-06,
+      "loss": 0.8107,
       "step": 24039
     },
     {
-      "epoch": 0.6602400373513498,
+      "epoch": 0.6821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 5.467784567912369e-06,
-      "loss": 0.8959,
+      "learning_rate": 4.84638962160461e-06,
+      "loss": 0.8491,
       "step": 24040
     },
     {
-      "epoch": 0.6602675015791931,
+      "epoch": 0.6822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 5.466991673129709e-06,
-      "loss": 0.7817,
+      "learning_rate": 4.845602018133509e-06,
+      "loss": 0.9258,
       "step": 24041
     },
     {
-      "epoch": 0.6602949658070363,
+      "epoch": 0.6822360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 5.46619881421397e-06,
-      "loss": 0.86,
+      "learning_rate": 4.844814458201834e-06,
+      "loss": 0.8366,
       "step": 24042
     },
     {
-      "epoch": 0.6603224300348796,
+      "epoch": 0.6822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 5.465405991171417e-06,
-      "loss": 0.8835,
+      "learning_rate": 4.844026941816243e-06,
+      "loss": 0.7732,
       "step": 24043
     },
     {
-      "epoch": 0.6603498942627228,
+      "epoch": 0.6822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 5.464613204008336e-06,
-      "loss": 0.8082,
+      "learning_rate": 4.843239468983394e-06,
+      "loss": 0.7964,
       "step": 24044
     },
     {
-      "epoch": 0.660377358490566,
+      "epoch": 0.6823212258796821,
       "grad_norm": 0.0,
-      "learning_rate": 5.463820452730985e-06,
-      "loss": 0.8366,
+      "learning_rate": 4.842452039709927e-06,
+      "loss": 0.7888,
       "step": 24045
     },
     {
-      "epoch": 0.6604048227184093,
+      "epoch": 0.6823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 5.4630277373456455e-06,
-      "loss": 0.784,
+      "learning_rate": 4.8416646540025e-06,
+      "loss": 0.8596,
       "step": 24046
     },
     {
-      "epoch": 0.6604322869462526,
+      "epoch": 0.682377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 5.462235057858587e-06,
-      "loss": 0.8535,
+      "learning_rate": 4.840877311867768e-06,
+      "loss": 0.7705,
       "step": 24047
     },
     {
-      "epoch": 0.6604597511740957,
+      "epoch": 0.6824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 5.461442414276083e-06,
-      "loss": 0.8411,
+      "learning_rate": 4.8400900133123744e-06,
+      "loss": 0.7083,
       "step": 24048
     },
     {
-      "epoch": 0.660487215401939,
+      "epoch": 0.6824347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 5.4606498066044075e-06,
-      "loss": 0.8158,
+      "learning_rate": 4.839302758342971e-06,
+      "loss": 0.7454,
       "step": 24049
     },
     {
-      "epoch": 0.6605146796297822,
+      "epoch": 0.6824631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 5.459857234849827e-06,
-      "loss": 0.977,
+      "learning_rate": 4.838515546966209e-06,
+      "loss": 0.7754,
       "step": 24050
     },
     {
-      "epoch": 0.6605421438576254,
+      "epoch": 0.6824914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 5.459064699018617e-06,
-      "loss": 0.8396,
+      "learning_rate": 4.837728379188738e-06,
+      "loss": 0.7735,
       "step": 24051
     },
     {
-      "epoch": 0.6605696080854687,
+      "epoch": 0.6825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 5.4582721991170425e-06,
-      "loss": 0.8905,
+      "learning_rate": 4.836941255017212e-06,
+      "loss": 0.8694,
       "step": 24052
     },
     {
-      "epoch": 0.6605970723133119,
+      "epoch": 0.6825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 5.457479735151379e-06,
-      "loss": 0.9068,
+      "learning_rate": 4.8361541744582715e-06,
+      "loss": 0.7831,
       "step": 24053
     },
     {
-      "epoch": 0.6606245365411552,
+      "epoch": 0.6825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 5.456687307127895e-06,
-      "loss": 0.8642,
+      "learning_rate": 4.83536713751857e-06,
+      "loss": 0.8234,
       "step": 24054
     },
     {
-      "epoch": 0.6606520007689983,
+      "epoch": 0.6826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 5.455894915052859e-06,
-      "loss": 0.8856,
+      "learning_rate": 4.834580144204757e-06,
+      "loss": 0.8303,
       "step": 24055
     },
     {
-      "epoch": 0.6606794649968416,
+      "epoch": 0.682633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 5.455102558932549e-06,
-      "loss": 0.8315,
+      "learning_rate": 4.833793194523473e-06,
+      "loss": 0.709,
       "step": 24056
     },
     {
-      "epoch": 0.6607069292246849,
+      "epoch": 0.6826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 5.454310238773224e-06,
-      "loss": 0.8541,
+      "learning_rate": 4.8330062884813714e-06,
+      "loss": 0.9254,
       "step": 24057
     },
     {
-      "epoch": 0.660734393452528,
+      "epoch": 0.6826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 5.45351795458116e-06,
-      "loss": 0.8905,
+      "learning_rate": 4.832219426085101e-06,
+      "loss": 0.8201,
       "step": 24058
     },
     {
-      "epoch": 0.6607618576803713,
+      "epoch": 0.6827185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 5.452725706362619e-06,
-      "loss": 0.825,
+      "learning_rate": 4.831432607341302e-06,
+      "loss": 0.8524,
       "step": 24059
     },
     {
-      "epoch": 0.6607893219082146,
+      "epoch": 0.6827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 5.451933494123874e-06,
-      "loss": 0.7829,
+      "learning_rate": 4.830645832256624e-06,
+      "loss": 0.7066,
       "step": 24060
     },
     {
-      "epoch": 0.6608167861360578,
+      "epoch": 0.6827752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 5.451141317871198e-06,
-      "loss": 0.8125,
+      "learning_rate": 4.829859100837717e-06,
+      "loss": 0.8419,
       "step": 24061
     },
     {
-      "epoch": 0.660844250363901,
+      "epoch": 0.6828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 5.450349177610849e-06,
-      "loss": 0.8303,
+      "learning_rate": 4.829072413091219e-06,
+      "loss": 0.8118,
       "step": 24062
     },
     {
-      "epoch": 0.6608717145917442,
+      "epoch": 0.6828320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 5.449557073349099e-06,
-      "loss": 0.8721,
+      "learning_rate": 4.828285769023778e-06,
+      "loss": 0.7042,
       "step": 24063
     },
     {
-      "epoch": 0.6608991788195875,
+      "epoch": 0.6828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 5.448765005092217e-06,
-      "loss": 0.8962,
+      "learning_rate": 4.82749916864204e-06,
+      "loss": 0.8531,
       "step": 24064
     },
     {
-      "epoch": 0.6609266430474308,
+      "epoch": 0.68288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 5.447972972846469e-06,
-      "loss": 0.8708,
+      "learning_rate": 4.8267126119526495e-06,
+      "loss": 0.8695,
       "step": 24065
     },
     {
-      "epoch": 0.6609541072752739,
+      "epoch": 0.682917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 5.447180976618127e-06,
-      "loss": 0.899,
+      "learning_rate": 4.82592609896225e-06,
+      "loss": 0.746,
       "step": 24066
     },
     {
-      "epoch": 0.6609815715031172,
+      "epoch": 0.6829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 5.446389016413451e-06,
-      "loss": 0.8349,
+      "learning_rate": 4.8251396296774886e-06,
+      "loss": 0.808,
       "step": 24067
     },
     {
-      "epoch": 0.6610090357309604,
+      "epoch": 0.6829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.445597092238706e-06,
-      "loss": 0.8063,
+      "learning_rate": 4.8243532041050024e-06,
+      "loss": 0.8863,
       "step": 24068
     },
     {
-      "epoch": 0.6610364999588036,
+      "epoch": 0.6830022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 5.4448052041001606e-06,
-      "loss": 0.8604,
+      "learning_rate": 4.823566822251441e-06,
+      "loss": 0.8835,
       "step": 24069
     },
     {
-      "epoch": 0.6610639641866469,
+      "epoch": 0.6830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 5.444013352004082e-06,
-      "loss": 0.942,
+      "learning_rate": 4.822780484123438e-06,
+      "loss": 0.8758,
       "step": 24070
     },
     {
-      "epoch": 0.6610914284144901,
+      "epoch": 0.6830590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 5.443221535956733e-06,
-      "loss": 0.7635,
+      "learning_rate": 4.8219941897276425e-06,
+      "loss": 0.751,
       "step": 24071
     },
     {
-      "epoch": 0.6611188926423334,
+      "epoch": 0.6830874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 5.442429755964382e-06,
-      "loss": 0.8744,
+      "learning_rate": 4.821207939070699e-06,
+      "loss": 0.9182,
       "step": 24072
     },
     {
-      "epoch": 0.6611463568701766,
+      "epoch": 0.6831157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 5.441638012033294e-06,
-      "loss": 0.7388,
+      "learning_rate": 4.82042173215924e-06,
+      "loss": 0.8255,
       "step": 24073
     },
     {
-      "epoch": 0.6611738210980198,
+      "epoch": 0.6831441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 5.440846304169731e-06,
-      "loss": 0.8248,
+      "learning_rate": 4.819635568999912e-06,
+      "loss": 0.9062,
       "step": 24074
     },
     {
-      "epoch": 0.6612012853258631,
+      "epoch": 0.6831725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 5.44005463237996e-06,
-      "loss": 0.7283,
+      "learning_rate": 4.8188494495993584e-06,
+      "loss": 0.7957,
       "step": 24075
     },
     {
-      "epoch": 0.6612287495537063,
+      "epoch": 0.6832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 5.439262996670239e-06,
-      "loss": 0.8506,
+      "learning_rate": 4.818063373964214e-06,
+      "loss": 0.8438,
       "step": 24076
     },
     {
-      "epoch": 0.6612562137815495,
+      "epoch": 0.6832292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 5.438471397046836e-06,
-      "loss": 0.8301,
+      "learning_rate": 4.81727734210112e-06,
+      "loss": 0.9398,
       "step": 24077
     },
     {
-      "epoch": 0.6612836780093928,
+      "epoch": 0.6832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 5.437679833516014e-06,
-      "loss": 0.9251,
+      "learning_rate": 4.8164913540167214e-06,
+      "loss": 0.8351,
       "step": 24078
     },
     {
-      "epoch": 0.661311142237236,
+      "epoch": 0.6832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 5.436888306084042e-06,
-      "loss": 0.874,
+      "learning_rate": 4.815705409717644e-06,
+      "loss": 0.8562,
       "step": 24079
     },
     {
-      "epoch": 0.6613386064650792,
+      "epoch": 0.6833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 5.436096814757172e-06,
-      "loss": 0.9265,
+      "learning_rate": 4.8149195092105425e-06,
+      "loss": 0.7858,
       "step": 24080
     },
     {
-      "epoch": 0.6613660706929224,
+      "epoch": 0.6833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 5.4353053595416716e-06,
-      "loss": 0.8628,
+      "learning_rate": 4.814133652502052e-06,
+      "loss": 0.8471,
       "step": 24081
     },
     {
-      "epoch": 0.6613935349207657,
+      "epoch": 0.6833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 5.434513940443808e-06,
-      "loss": 0.8252,
+      "learning_rate": 4.813347839598805e-06,
+      "loss": 0.8566,
       "step": 24082
     },
     {
-      "epoch": 0.661420999148609,
+      "epoch": 0.6833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 5.4337225574698345e-06,
-      "loss": 0.7381,
+      "learning_rate": 4.81256207050744e-06,
+      "loss": 0.7817,
       "step": 24083
     },
     {
-      "epoch": 0.6614484633764521,
+      "epoch": 0.683427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 5.432931210626018e-06,
-      "loss": 0.9107,
+      "learning_rate": 4.8117763452346e-06,
+      "loss": 0.8525,
       "step": 24084
     },
     {
-      "epoch": 0.6614759276042954,
+      "epoch": 0.6834562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 5.43213989991862e-06,
-      "loss": 0.9054,
+      "learning_rate": 4.8109906637869155e-06,
+      "loss": 0.8942,
       "step": 24085
     },
     {
-      "epoch": 0.6615033918321387,
+      "epoch": 0.6834846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 5.431348625353899e-06,
-      "loss": 0.8065,
+      "learning_rate": 4.8102050261710265e-06,
+      "loss": 0.8494,
       "step": 24086
     },
     {
-      "epoch": 0.6615308560599819,
+      "epoch": 0.6835130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 5.430557386938118e-06,
-      "loss": 0.8549,
+      "learning_rate": 4.8094194323935716e-06,
+      "loss": 0.849,
       "step": 24087
     },
     {
-      "epoch": 0.6615583202878251,
+      "epoch": 0.6835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 5.429766184677534e-06,
-      "loss": 0.7699,
+      "learning_rate": 4.808633882461181e-06,
+      "loss": 0.911,
       "step": 24088
     },
     {
-      "epoch": 0.6615857845156683,
+      "epoch": 0.6835698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 5.428975018578413e-06,
-      "loss": 0.8571,
+      "learning_rate": 4.807848376380494e-06,
+      "loss": 0.8474,
       "step": 24089
     },
     {
-      "epoch": 0.6616132487435116,
+      "epoch": 0.6835981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 5.428183888647015e-06,
-      "loss": 0.8201,
+      "learning_rate": 4.807062914158148e-06,
+      "loss": 0.8189,
       "step": 24090
     },
     {
-      "epoch": 0.6616407129713548,
+      "epoch": 0.6836265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 5.427392794889598e-06,
-      "loss": 0.9091,
+      "learning_rate": 4.806277495800772e-06,
+      "loss": 0.9261,
       "step": 24091
     },
     {
-      "epoch": 0.661668177199198,
+      "epoch": 0.6836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 5.426601737312416e-06,
-      "loss": 0.9091,
+      "learning_rate": 4.805492121315003e-06,
+      "loss": 0.8639,
       "step": 24092
     },
     {
-      "epoch": 0.6616956414270413,
+      "epoch": 0.683683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 5.425810715921731e-06,
-      "loss": 0.8153,
+      "learning_rate": 4.804706790707479e-06,
+      "loss": 0.8829,
       "step": 24093
     },
     {
-      "epoch": 0.6617231056548845,
+      "epoch": 0.6837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 5.4250197307238065e-06,
-      "loss": 0.8169,
+      "learning_rate": 4.803921503984828e-06,
+      "loss": 0.807,
       "step": 24094
     },
     {
-      "epoch": 0.6617505698827277,
+      "epoch": 0.6837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 5.4242287817248965e-06,
-      "loss": 0.7502,
+      "learning_rate": 4.803136261153684e-06,
+      "loss": 0.744,
       "step": 24095
     },
     {
-      "epoch": 0.661778034110571,
+      "epoch": 0.6837684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 5.42343786893126e-06,
-      "loss": 0.8921,
+      "learning_rate": 4.802351062220681e-06,
+      "loss": 0.8894,
       "step": 24096
     },
     {
-      "epoch": 0.6618054983384142,
+      "epoch": 0.6837968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 5.422646992349161e-06,
-      "loss": 0.797,
+      "learning_rate": 4.8015659071924535e-06,
+      "loss": 0.8966,
       "step": 24097
     },
     {
-      "epoch": 0.6618329625662575,
+      "epoch": 0.6838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 5.421856151984848e-06,
-      "loss": 0.8172,
+      "learning_rate": 4.800780796075637e-06,
+      "loss": 0.8887,
       "step": 24098
     },
     {
-      "epoch": 0.6618604267941007,
+      "epoch": 0.6838535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 5.421065347844585e-06,
-      "loss": 0.824,
+      "learning_rate": 4.799995728876854e-06,
+      "loss": 0.8298,
       "step": 24099
     },
     {
-      "epoch": 0.6618878910219439,
+      "epoch": 0.6838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 5.4202745799346244e-06,
-      "loss": 0.8736,
+      "learning_rate": 4.79921070560274e-06,
+      "loss": 0.96,
       "step": 24100
     },
     {
-      "epoch": 0.6619153552497872,
+      "epoch": 0.6839103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 5.419483848261225e-06,
-      "loss": 0.7935,
+      "learning_rate": 4.798425726259933e-06,
+      "loss": 0.8313,
       "step": 24101
     },
     {
-      "epoch": 0.6619428194776303,
+      "epoch": 0.683938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 5.418693152830644e-06,
-      "loss": 1.0245,
+      "learning_rate": 4.797640790855053e-06,
+      "loss": 0.7635,
       "step": 24102
     },
     {
-      "epoch": 0.6619702837054736,
+      "epoch": 0.6839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 5.4179024936491405e-06,
-      "loss": 0.7933,
+      "learning_rate": 4.796855899394734e-06,
+      "loss": 0.8326,
       "step": 24103
     },
     {
-      "epoch": 0.6619977479333169,
+      "epoch": 0.6839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 5.417111870722964e-06,
-      "loss": 0.879,
+      "learning_rate": 4.796071051885611e-06,
+      "loss": 0.8119,
       "step": 24104
     },
     {
-      "epoch": 0.6620252121611601,
+      "epoch": 0.6840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 5.416321284058375e-06,
-      "loss": 0.7874,
+      "learning_rate": 4.7952862483343075e-06,
+      "loss": 0.8437,
       "step": 24105
     },
     {
-      "epoch": 0.6620526763890033,
+      "epoch": 0.6840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 5.415530733661629e-06,
-      "loss": 0.8355,
+      "learning_rate": 4.794501488747454e-06,
+      "loss": 0.8857,
       "step": 24106
     },
     {
-      "epoch": 0.6620801406168465,
+      "epoch": 0.6840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 5.414740219538977e-06,
-      "loss": 0.7888,
+      "learning_rate": 4.793716773131685e-06,
+      "loss": 0.8252,
       "step": 24107
     },
     {
-      "epoch": 0.6621076048446898,
+      "epoch": 0.6841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 5.413949741696676e-06,
-      "loss": 0.9071,
+      "learning_rate": 4.792932101493619e-06,
+      "loss": 0.8014,
       "step": 24108
     },
     {
-      "epoch": 0.6621350690725331,
+      "epoch": 0.6841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 5.413159300140986e-06,
-      "loss": 0.8206,
+      "learning_rate": 4.792147473839891e-06,
+      "loss": 0.8024,
       "step": 24109
     },
     {
-      "epoch": 0.6621625333003762,
+      "epoch": 0.6841657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 5.412368894878151e-06,
-      "loss": 0.853,
+      "learning_rate": 4.791362890177127e-06,
+      "loss": 0.8638,
       "step": 24110
     },
     {
-      "epoch": 0.6621899975282195,
+      "epoch": 0.6841940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 5.411578525914431e-06,
-      "loss": 0.9093,
+      "learning_rate": 4.790578350511955e-06,
+      "loss": 0.6911,
       "step": 24111
     },
     {
-      "epoch": 0.6622174617560628,
+      "epoch": 0.68422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 5.41078819325608e-06,
-      "loss": 0.8722,
+      "learning_rate": 4.789793854851006e-06,
+      "loss": 0.7966,
       "step": 24112
     },
     {
-      "epoch": 0.6622449259839059,
+      "epoch": 0.6842508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 5.40999789690935e-06,
-      "loss": 0.8474,
+      "learning_rate": 4.789009403200898e-06,
+      "loss": 0.8753,
       "step": 24113
     },
     {
-      "epoch": 0.6622723902117492,
+      "epoch": 0.6842792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 5.409207636880498e-06,
-      "loss": 0.8503,
+      "learning_rate": 4.788224995568263e-06,
+      "loss": 0.7505,
       "step": 24114
     },
     {
-      "epoch": 0.6622998544395924,
+      "epoch": 0.6843076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 5.408417413175775e-06,
-      "loss": 0.8918,
+      "learning_rate": 4.787440631959728e-06,
+      "loss": 0.9443,
       "step": 24115
     },
     {
-      "epoch": 0.6623273186674357,
+      "epoch": 0.6843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 5.407627225801426e-06,
-      "loss": 0.8835,
+      "learning_rate": 4.786656312381913e-06,
+      "loss": 0.7936,
       "step": 24116
     },
     {
-      "epoch": 0.6623547828952789,
+      "epoch": 0.6843643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 5.406837074763711e-06,
-      "loss": 0.8151,
+      "learning_rate": 4.785872036841447e-06,
+      "loss": 0.8493,
       "step": 24117
     },
     {
-      "epoch": 0.6623822471231221,
+      "epoch": 0.6843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 5.40604696006888e-06,
-      "loss": 0.8876,
+      "learning_rate": 4.7850878053449566e-06,
+      "loss": 0.8306,
       "step": 24118
     },
     {
-      "epoch": 0.6624097113509654,
+      "epoch": 0.6844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 5.405256881723184e-06,
-      "loss": 0.8458,
+      "learning_rate": 4.784303617899062e-06,
+      "loss": 0.8186,
       "step": 24119
     },
     {
-      "epoch": 0.6624371755788087,
+      "epoch": 0.684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 5.40446683973288e-06,
-      "loss": 0.7667,
+      "learning_rate": 4.783519474510388e-06,
+      "loss": 0.8595,
       "step": 24120
     },
     {
-      "epoch": 0.6624646398066518,
+      "epoch": 0.684477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 5.40367683410421e-06,
-      "loss": 0.848,
+      "learning_rate": 4.782735375185565e-06,
+      "loss": 0.9005,
       "step": 24121
     },
     {
-      "epoch": 0.6624921040344951,
+      "epoch": 0.6845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 5.402886864843431e-06,
-      "loss": 0.8172,
+      "learning_rate": 4.781951319931205e-06,
+      "loss": 0.7891,
       "step": 24122
     },
     {
-      "epoch": 0.6625195682623383,
+      "epoch": 0.6845346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 5.402096931956794e-06,
-      "loss": 0.9575,
+      "learning_rate": 4.781167308753938e-06,
+      "loss": 0.8733,
       "step": 24123
     },
     {
-      "epoch": 0.6625470324901815,
+      "epoch": 0.6845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 5.401307035450544e-06,
-      "loss": 0.8232,
+      "learning_rate": 4.780383341660389e-06,
+      "loss": 0.7963,
       "step": 24124
     },
     {
-      "epoch": 0.6625744967180248,
+      "epoch": 0.6845913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 5.400517175330934e-06,
-      "loss": 0.7854,
+      "learning_rate": 4.779599418657168e-06,
+      "loss": 0.788,
       "step": 24125
     },
     {
-      "epoch": 0.662601960945868,
+      "epoch": 0.6846197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 5.399727351604217e-06,
-      "loss": 0.795,
+      "learning_rate": 4.778815539750913e-06,
+      "loss": 0.8387,
       "step": 24126
     },
     {
-      "epoch": 0.6626294251737113,
+      "epoch": 0.6846481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 5.398937564276636e-06,
-      "loss": 0.79,
+      "learning_rate": 4.778031704948235e-06,
+      "loss": 0.8246,
       "step": 24127
     },
     {
-      "epoch": 0.6626568894015544,
+      "epoch": 0.6846765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 5.398147813354444e-06,
-      "loss": 0.8536,
+      "learning_rate": 4.777247914255757e-06,
+      "loss": 0.7695,
       "step": 24128
     },
     {
-      "epoch": 0.6626843536293977,
+      "epoch": 0.6847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 5.3973580988438875e-06,
-      "loss": 0.8422,
+      "learning_rate": 4.776464167680106e-06,
+      "loss": 0.8148,
       "step": 24129
     },
     {
-      "epoch": 0.662711817857241,
+      "epoch": 0.684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 5.396568420751221e-06,
-      "loss": 0.9261,
+      "learning_rate": 4.775680465227892e-06,
+      "loss": 0.8288,
       "step": 24130
     },
     {
-      "epoch": 0.6627392820850841,
+      "epoch": 0.6847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 5.395778779082685e-06,
-      "loss": 0.8273,
+      "learning_rate": 4.77489680690574e-06,
+      "loss": 0.9236,
       "step": 24131
     },
     {
-      "epoch": 0.6627667463129274,
+      "epoch": 0.6847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 5.394989173844534e-06,
-      "loss": 0.8818,
+      "learning_rate": 4.774113192720273e-06,
+      "loss": 0.8598,
       "step": 24132
     },
     {
-      "epoch": 0.6627942105407707,
+      "epoch": 0.6848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 5.394199605043009e-06,
-      "loss": 0.7948,
+      "learning_rate": 4.773329622678105e-06,
+      "loss": 0.8577,
       "step": 24133
     },
     {
-      "epoch": 0.6628216747686139,
+      "epoch": 0.6848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 5.393410072684362e-06,
-      "loss": 0.8588,
+      "learning_rate": 4.772546096785854e-06,
+      "loss": 0.8792,
       "step": 24134
     },
     {
-      "epoch": 0.6628491389964571,
+      "epoch": 0.6848751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 5.392620576774836e-06,
-      "loss": 0.8235,
+      "learning_rate": 4.771762615050146e-06,
+      "loss": 0.891,
       "step": 24135
     },
     {
-      "epoch": 0.6628766032243003,
+      "epoch": 0.6849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 5.391831117320683e-06,
-      "loss": 0.8828,
+      "learning_rate": 4.77097917747759e-06,
+      "loss": 0.8637,
       "step": 24136
     },
     {
-      "epoch": 0.6629040674521436,
+      "epoch": 0.6849318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 5.391041694328148e-06,
-      "loss": 0.7737,
+      "learning_rate": 4.7701957840748096e-06,
+      "loss": 0.8525,
       "step": 24137
     },
     {
-      "epoch": 0.6629315316799869,
+      "epoch": 0.6849602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 5.390252307803478e-06,
-      "loss": 0.9456,
+      "learning_rate": 4.7694124348484225e-06,
+      "loss": 0.7862,
       "step": 24138
     },
     {
-      "epoch": 0.66295899590783,
+      "epoch": 0.684988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 5.3894629577529174e-06,
-      "loss": 0.8021,
+      "learning_rate": 4.768629129805041e-06,
+      "loss": 0.9348,
       "step": 24139
     },
     {
-      "epoch": 0.6629864601356733,
+      "epoch": 0.6850170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 5.38867364418271e-06,
-      "loss": 0.784,
+      "learning_rate": 4.767845868951284e-06,
+      "loss": 0.9711,
       "step": 24140
     },
     {
-      "epoch": 0.6630139243635165,
+      "epoch": 0.6850454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 5.387884367099102e-06,
-      "loss": 0.9451,
+      "learning_rate": 4.767062652293768e-06,
+      "loss": 0.9073,
       "step": 24141
     },
     {
-      "epoch": 0.6630413885913597,
+      "epoch": 0.6850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 5.387095126508339e-06,
-      "loss": 0.8768,
+      "learning_rate": 4.766279479839109e-06,
+      "loss": 0.9061,
       "step": 24142
     },
     {
-      "epoch": 0.663068852819203,
+      "epoch": 0.6851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 5.386305922416667e-06,
-      "loss": 0.7547,
+      "learning_rate": 4.765496351593927e-06,
+      "loss": 0.7869,
       "step": 24143
     },
     {
-      "epoch": 0.6630963170470462,
+      "epoch": 0.6851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 5.385516754830334e-06,
-      "loss": 0.8228,
+      "learning_rate": 4.764713267564828e-06,
+      "loss": 0.8046,
       "step": 24144
     },
     {
-      "epoch": 0.6631237812748895,
+      "epoch": 0.6851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 5.384727623755575e-06,
-      "loss": 0.799,
+      "learning_rate": 4.763930227758431e-06,
+      "loss": 0.8285,
       "step": 24145
     },
     {
-      "epoch": 0.6631512455027327,
+      "epoch": 0.6851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 5.3839385291986425e-06,
-      "loss": 0.919,
+      "learning_rate": 4.763147232181355e-06,
+      "loss": 0.7836,
       "step": 24146
     },
     {
-      "epoch": 0.6631787097305759,
+      "epoch": 0.6852156640181611,
       "grad_norm": 0.0,
-      "learning_rate": 5.383149471165774e-06,
-      "loss": 0.8444,
+      "learning_rate": 4.762364280840207e-06,
+      "loss": 0.882,
       "step": 24147
     },
     {
-      "epoch": 0.6632061739584192,
+      "epoch": 0.6852440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 5.382360449663213e-06,
-      "loss": 0.8239,
+      "learning_rate": 4.7615813737416014e-06,
+      "loss": 0.8145,
       "step": 24148
     },
     {
-      "epoch": 0.6632336381862624,
+      "epoch": 0.685272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 5.3815714646972065e-06,
-      "loss": 0.9059,
+      "learning_rate": 4.76079851089216e-06,
+      "loss": 0.8415,
       "step": 24149
     },
     {
-      "epoch": 0.6632611024141056,
+      "epoch": 0.6853007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 5.380782516273999e-06,
-      "loss": 0.8774,
+      "learning_rate": 4.760015692298483e-06,
+      "loss": 0.8046,
       "step": 24150
     },
     {
-      "epoch": 0.6632885666419489,
+      "epoch": 0.6853291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 5.379993604399826e-06,
-      "loss": 0.7962,
+      "learning_rate": 4.75923291796719e-06,
+      "loss": 0.8024,
       "step": 24151
     },
     {
-      "epoch": 0.6633160308697921,
+      "epoch": 0.6853575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 5.379204729080934e-06,
-      "loss": 0.8161,
+      "learning_rate": 4.758450187904895e-06,
+      "loss": 0.8682,
       "step": 24152
     },
     {
-      "epoch": 0.6633434950976353,
+      "epoch": 0.6853859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 5.378415890323563e-06,
-      "loss": 0.8207,
+      "learning_rate": 4.757667502118203e-06,
+      "loss": 0.8312,
       "step": 24153
     },
     {
-      "epoch": 0.6633709593254785,
+      "epoch": 0.6854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 5.377627088133959e-06,
-      "loss": 0.8537,
+      "learning_rate": 4.7568848606137294e-06,
+      "loss": 0.8454,
       "step": 24154
     },
     {
-      "epoch": 0.6633984235533218,
+      "epoch": 0.6854426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 5.376838322518356e-06,
-      "loss": 0.8126,
+      "learning_rate": 4.756102263398091e-06,
+      "loss": 0.8774,
       "step": 24155
     },
     {
-      "epoch": 0.6634258877811651,
+      "epoch": 0.6854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 5.376049593483004e-06,
-      "loss": 0.8795,
+      "learning_rate": 4.755319710477882e-06,
+      "loss": 0.8788,
       "step": 24156
     },
     {
-      "epoch": 0.6634533520090082,
+      "epoch": 0.6854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 5.375260901034135e-06,
-      "loss": 0.8787,
+      "learning_rate": 4.754537201859732e-06,
+      "loss": 0.8079,
       "step": 24157
     },
     {
-      "epoch": 0.6634808162368515,
+      "epoch": 0.685527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 5.374472245177993e-06,
-      "loss": 0.8332,
+      "learning_rate": 4.753754737550239e-06,
+      "loss": 0.8917,
       "step": 24158
     },
     {
-      "epoch": 0.6635082804646948,
+      "epoch": 0.6855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 5.373683625920818e-06,
-      "loss": 0.8111,
+      "learning_rate": 4.752972317556015e-06,
+      "loss": 0.8587,
       "step": 24159
     },
     {
-      "epoch": 0.663535744692538,
+      "epoch": 0.6855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 5.3728950432688505e-06,
-      "loss": 0.7987,
+      "learning_rate": 4.752189941883673e-06,
+      "loss": 0.9322,
       "step": 24160
     },
     {
-      "epoch": 0.6635632089203812,
+      "epoch": 0.6856129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 5.372106497228329e-06,
-      "loss": 0.8226,
+      "learning_rate": 4.751407610539815e-06,
+      "loss": 0.7917,
       "step": 24161
     },
     {
-      "epoch": 0.6635906731482244,
+      "epoch": 0.6856413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 5.3713179878054985e-06,
-      "loss": 0.7713,
+      "learning_rate": 4.750625323531053e-06,
+      "loss": 0.9932,
       "step": 24162
     },
     {
-      "epoch": 0.6636181373760677,
+      "epoch": 0.6856696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 5.3705295150065926e-06,
-      "loss": 0.8949,
+      "learning_rate": 4.749843080863998e-06,
+      "loss": 0.7599,
       "step": 24163
     },
     {
-      "epoch": 0.663645601603911,
+      "epoch": 0.6856980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 5.369741078837847e-06,
-      "loss": 0.8788,
+      "learning_rate": 4.749060882545251e-06,
+      "loss": 0.859,
       "step": 24164
     },
     {
-      "epoch": 0.6636730658317541,
+      "epoch": 0.6857264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 5.368952679305503e-06,
-      "loss": 0.8482,
+      "learning_rate": 4.748278728581424e-06,
+      "loss": 0.7824,
       "step": 24165
     },
     {
-      "epoch": 0.6637005300595974,
+      "epoch": 0.6857548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 5.3681643164158e-06,
-      "loss": 0.9884,
+      "learning_rate": 4.747496618979125e-06,
+      "loss": 0.7673,
       "step": 24166
     },
     {
-      "epoch": 0.6637279942874406,
+      "epoch": 0.685783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 5.367375990174974e-06,
-      "loss": 0.8581,
+      "learning_rate": 4.746714553744956e-06,
+      "loss": 0.8741,
       "step": 24167
     },
     {
-      "epoch": 0.6637554585152838,
+      "epoch": 0.685811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 5.366587700589268e-06,
-      "loss": 0.8284,
+      "learning_rate": 4.745932532885523e-06,
+      "loss": 0.8219,
       "step": 24168
     },
     {
-      "epoch": 0.6637829227431271,
+      "epoch": 0.6858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 5.365799447664912e-06,
-      "loss": 0.8972,
+      "learning_rate": 4.7451505564074395e-06,
+      "loss": 0.8402,
       "step": 24169
     },
     {
-      "epoch": 0.6638103869709703,
+      "epoch": 0.6858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 5.36501123140815e-06,
-      "loss": 0.8365,
+      "learning_rate": 4.7443686243173015e-06,
+      "loss": 0.8679,
       "step": 24170
     },
     {
-      "epoch": 0.6638378511988136,
+      "epoch": 0.6858967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 5.364223051825208e-06,
-      "loss": 0.9704,
+      "learning_rate": 4.743586736621714e-06,
+      "loss": 0.7867,
       "step": 24171
     },
     {
-      "epoch": 0.6638653154266568,
+      "epoch": 0.6859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 5.363434908922332e-06,
-      "loss": 0.9812,
+      "learning_rate": 4.742804893327293e-06,
+      "loss": 0.8337,
       "step": 24172
     },
     {
-      "epoch": 0.6638927796545,
+      "epoch": 0.6859534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 5.362646802705753e-06,
-      "loss": 0.9789,
+      "learning_rate": 4.7420230944406306e-06,
+      "loss": 0.9249,
       "step": 24173
     },
     {
-      "epoch": 0.6639202438823433,
+      "epoch": 0.6859818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 5.361858733181713e-06,
-      "loss": 0.8185,
+      "learning_rate": 4.741241339968338e-06,
+      "loss": 0.807,
       "step": 24174
     },
     {
-      "epoch": 0.6639477081101864,
+      "epoch": 0.6860102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 5.361070700356441e-06,
-      "loss": 0.9502,
+      "learning_rate": 4.740459629917018e-06,
+      "loss": 0.8282,
       "step": 24175
     },
     {
-      "epoch": 0.6639751723380297,
+      "epoch": 0.6860385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 5.360282704236172e-06,
-      "loss": 0.8468,
+      "learning_rate": 4.739677964293269e-06,
+      "loss": 0.8905,
       "step": 24176
     },
     {
-      "epoch": 0.664002636565873,
+      "epoch": 0.6860669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 5.359494744827145e-06,
-      "loss": 0.8791,
+      "learning_rate": 4.7388963431037e-06,
+      "loss": 0.9263,
       "step": 24177
     },
     {
-      "epoch": 0.6640301007937162,
+      "epoch": 0.6860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 5.358706822135598e-06,
-      "loss": 0.8932,
+      "learning_rate": 4.738114766354907e-06,
+      "loss": 0.775,
       "step": 24178
     },
     {
-      "epoch": 0.6640575650215594,
+      "epoch": 0.6861237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 5.357918936167758e-06,
-      "loss": 0.8084,
+      "learning_rate": 4.737333234053494e-06,
+      "loss": 0.685,
       "step": 24179
     },
     {
-      "epoch": 0.6640850292494026,
+      "epoch": 0.6861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 5.357131086929861e-06,
-      "loss": 0.8698,
+      "learning_rate": 4.7365517462060685e-06,
+      "loss": 0.7821,
       "step": 24180
     },
     {
-      "epoch": 0.6641124934772459,
+      "epoch": 0.6861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 5.356343274428139e-06,
-      "loss": 0.8517,
+      "learning_rate": 4.735770302819223e-06,
+      "loss": 0.9268,
       "step": 24181
     },
     {
-      "epoch": 0.6641399577050892,
+      "epoch": 0.6862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 5.3555554986688275e-06,
-      "loss": 0.7989,
+      "learning_rate": 4.734988903899562e-06,
+      "loss": 0.8595,
       "step": 24182
     },
     {
-      "epoch": 0.6641674219329323,
+      "epoch": 0.6862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 5.354767759658161e-06,
-      "loss": 0.8849,
+      "learning_rate": 4.734207549453691e-06,
+      "loss": 0.9205,
       "step": 24183
     },
     {
-      "epoch": 0.6641948861607756,
+      "epoch": 0.6862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 5.353980057402372e-06,
-      "loss": 0.8239,
+      "learning_rate": 4.733426239488201e-06,
+      "loss": 0.7666,
       "step": 24184
     },
     {
-      "epoch": 0.6642223503886189,
+      "epoch": 0.6862939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 5.353192391907695e-06,
-      "loss": 0.9031,
+      "learning_rate": 4.732644974009697e-06,
+      "loss": 0.7633,
       "step": 24185
     },
     {
-      "epoch": 0.664249814616462,
+      "epoch": 0.686322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 5.352404763180356e-06,
-      "loss": 0.8641,
+      "learning_rate": 4.7318637530247805e-06,
+      "loss": 0.6691,
       "step": 24186
     },
     {
-      "epoch": 0.6642772788443053,
+      "epoch": 0.6863507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 5.351617171226595e-06,
-      "loss": 0.8396,
+      "learning_rate": 4.731082576540042e-06,
+      "loss": 0.8706,
       "step": 24187
     },
     {
-      "epoch": 0.6643047430721485,
+      "epoch": 0.6863791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 5.350829616052635e-06,
-      "loss": 0.827,
+      "learning_rate": 4.730301444562088e-06,
+      "loss": 0.8469,
       "step": 24188
     },
     {
-      "epoch": 0.6643322072999918,
+      "epoch": 0.6864074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 5.3500420976647135e-06,
-      "loss": 0.825,
+      "learning_rate": 4.729520357097518e-06,
+      "loss": 0.8353,
       "step": 24189
     },
     {
-      "epoch": 0.664359671527835,
+      "epoch": 0.6864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 5.34925461606906e-06,
-      "loss": 0.8892,
+      "learning_rate": 4.7287393141529236e-06,
+      "loss": 0.839,
       "step": 24190
     },
     {
-      "epoch": 0.6643871357556782,
+      "epoch": 0.6864642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 5.348467171271908e-06,
-      "loss": 0.8099,
+      "learning_rate": 4.727958315734904e-06,
+      "loss": 0.8948,
       "step": 24191
     },
     {
-      "epoch": 0.6644145999835215,
+      "epoch": 0.6864926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 5.347679763279483e-06,
-      "loss": 0.8385,
+      "learning_rate": 4.7271773618500625e-06,
+      "loss": 0.8422,
       "step": 24192
     },
     {
-      "epoch": 0.6644420642113646,
+      "epoch": 0.6865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 5.346892392098018e-06,
-      "loss": 0.7667,
+      "learning_rate": 4.726396452504986e-06,
+      "loss": 0.8532,
       "step": 24193
     },
     {
-      "epoch": 0.6644695284392079,
+      "epoch": 0.6865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 5.346105057733746e-06,
-      "loss": 0.8707,
+      "learning_rate": 4.725615587706278e-06,
+      "loss": 0.7395,
       "step": 24194
     },
     {
-      "epoch": 0.6644969926670512,
+      "epoch": 0.686577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.345317760192888e-06,
-      "loss": 0.9146,
+      "learning_rate": 4.724834767460534e-06,
+      "loss": 0.872,
       "step": 24195
     },
     {
-      "epoch": 0.6645244568948944,
+      "epoch": 0.6866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 5.344530499481681e-06,
-      "loss": 0.9239,
+      "learning_rate": 4.724053991774345e-06,
+      "loss": 0.7767,
       "step": 24196
     },
     {
-      "epoch": 0.6645519211227376,
+      "epoch": 0.6866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 5.343743275606356e-06,
-      "loss": 0.8374,
+      "learning_rate": 4.7232732606543085e-06,
+      "loss": 0.833,
       "step": 24197
     },
     {
-      "epoch": 0.6645793853505809,
+      "epoch": 0.6866628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 5.342956088573133e-06,
-      "loss": 0.878,
+      "learning_rate": 4.722492574107024e-06,
+      "loss": 0.8548,
       "step": 24198
     },
     {
-      "epoch": 0.6646068495784241,
+      "epoch": 0.6866912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 5.342168938388246e-06,
-      "loss": 0.9135,
+      "learning_rate": 4.721711932139078e-06,
+      "loss": 0.8533,
       "step": 24199
     },
     {
-      "epoch": 0.6646343138062674,
+      "epoch": 0.6867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 5.341381825057923e-06,
-      "loss": 0.7663,
+      "learning_rate": 4.720931334757068e-06,
+      "loss": 0.8865,
       "step": 24200
     },
     {
-      "epoch": 0.6646617780341105,
+      "epoch": 0.6867480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 5.34059474858839e-06,
-      "loss": 0.8649,
+      "learning_rate": 4.720150781967594e-06,
+      "loss": 0.8557,
       "step": 24201
     },
     {
-      "epoch": 0.6646892422619538,
+      "epoch": 0.6867763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 5.3398077089858825e-06,
-      "loss": 0.8956,
+      "learning_rate": 4.719370273777235e-06,
+      "loss": 0.8924,
       "step": 24202
     },
     {
-      "epoch": 0.6647167064897971,
+      "epoch": 0.6868047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 5.33902070625662e-06,
-      "loss": 0.8121,
+      "learning_rate": 4.7185898101926e-06,
+      "loss": 0.7809,
       "step": 24203
     },
     {
-      "epoch": 0.6647441707176402,
+      "epoch": 0.68683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 5.338233740406827e-06,
-      "loss": 0.9172,
+      "learning_rate": 4.717809391220271e-06,
+      "loss": 0.8109,
       "step": 24204
     },
     {
-      "epoch": 0.6647716349454835,
+      "epoch": 0.6868615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 5.337446811442736e-06,
-      "loss": 0.8906,
+      "learning_rate": 4.7170290168668435e-06,
+      "loss": 0.9051,
       "step": 24205
     },
     {
-      "epoch": 0.6647990991733267,
+      "epoch": 0.6868898978433599,
       "grad_norm": 0.0,
-      "learning_rate": 5.33665991937057e-06,
-      "loss": 0.8415,
+      "learning_rate": 4.7162486871389125e-06,
+      "loss": 0.8057,
       "step": 24206
     },
     {
-      "epoch": 0.66482656340117,
+      "epoch": 0.6869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 5.335873064196558e-06,
-      "loss": 0.8448,
+      "learning_rate": 4.715468402043063e-06,
+      "loss": 0.9261,
       "step": 24207
     },
     {
-      "epoch": 0.6648540276290132,
+      "epoch": 0.6869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 5.335086245926926e-06,
-      "loss": 0.8241,
+      "learning_rate": 4.71468816158589e-06,
+      "loss": 0.8485,
       "step": 24208
     },
     {
-      "epoch": 0.6648814918568564,
+      "epoch": 0.6869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 5.334299464567901e-06,
-      "loss": 0.8126,
+      "learning_rate": 4.713907965773986e-06,
+      "loss": 0.8708,
       "step": 24209
     },
     {
-      "epoch": 0.6649089560846997,
+      "epoch": 0.6870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 5.333512720125703e-06,
-      "loss": 0.844,
+      "learning_rate": 4.7131278146139355e-06,
+      "loss": 0.8546,
       "step": 24210
     },
     {
-      "epoch": 0.664936420312543,
+      "epoch": 0.6870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 5.332726012606564e-06,
-      "loss": 0.8083,
+      "learning_rate": 4.712347708112334e-06,
+      "loss": 0.867,
       "step": 24211
     },
     {
-      "epoch": 0.6649638845403861,
+      "epoch": 0.6870601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 5.3319393420167e-06,
-      "loss": 0.7985,
+      "learning_rate": 4.711567646275771e-06,
+      "loss": 0.773,
       "step": 24212
     },
     {
-      "epoch": 0.6649913487682294,
+      "epoch": 0.6870885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 5.33115270836234e-06,
-      "loss": 0.7646,
+      "learning_rate": 4.7107876291108315e-06,
+      "loss": 0.8275,
       "step": 24213
     },
     {
-      "epoch": 0.6650188129960726,
+      "epoch": 0.687116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 5.330366111649709e-06,
-      "loss": 0.9753,
+      "learning_rate": 4.7100076566241045e-06,
+      "loss": 0.8342,
       "step": 24214
     },
     {
-      "epoch": 0.6650462772239158,
+      "epoch": 0.6871452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 5.329579551885035e-06,
-      "loss": 0.8052,
+      "learning_rate": 4.7092277288221865e-06,
+      "loss": 0.69,
       "step": 24215
     },
     {
-      "epoch": 0.6650737414517591,
+      "epoch": 0.6871736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 5.32879302907453e-06,
-      "loss": 0.9282,
+      "learning_rate": 4.7084478457116545e-06,
+      "loss": 0.9053,
       "step": 24216
     },
     {
-      "epoch": 0.6651012056796023,
+      "epoch": 0.6872020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 5.328006543224427e-06,
-      "loss": 0.8647,
+      "learning_rate": 4.707668007299102e-06,
+      "loss": 0.8405,
       "step": 24217
     },
     {
-      "epoch": 0.6651286699074456,
+      "epoch": 0.6872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 5.327220094340948e-06,
-      "loss": 0.9326,
+      "learning_rate": 4.7068882135911165e-06,
+      "loss": 0.7963,
       "step": 24218
     },
     {
-      "epoch": 0.6651561341352887,
+      "epoch": 0.6872587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 5.326433682430311e-06,
-      "loss": 0.8148,
+      "learning_rate": 4.706108464594283e-06,
+      "loss": 0.8799,
       "step": 24219
     },
     {
-      "epoch": 0.665183598363132,
+      "epoch": 0.6872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 5.32564730749874e-06,
-      "loss": 0.8042,
+      "learning_rate": 4.7053287603151935e-06,
+      "loss": 0.7707,
       "step": 24220
     },
     {
-      "epoch": 0.6652110625909753,
+      "epoch": 0.6873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 5.3248609695524635e-06,
-      "loss": 0.8466,
+      "learning_rate": 4.704549100760426e-06,
+      "loss": 0.7807,
       "step": 24221
     },
     {
-      "epoch": 0.6652385268188185,
+      "epoch": 0.687343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 5.324074668597693e-06,
-      "loss": 0.8785,
+      "learning_rate": 4.703769485936571e-06,
+      "loss": 0.8336,
       "step": 24222
     },
     {
-      "epoch": 0.6652659910466617,
+      "epoch": 0.687372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 5.323288404640655e-06,
-      "loss": 0.8237,
+      "learning_rate": 4.702989915850217e-06,
+      "loss": 0.8181,
       "step": 24223
     },
     {
-      "epoch": 0.665293455274505,
+      "epoch": 0.6874006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 5.3225021776875706e-06,
-      "loss": 0.7861,
+      "learning_rate": 4.7022103905079405e-06,
+      "loss": 0.8651,
       "step": 24224
     },
     {
-      "epoch": 0.6653209195023482,
+      "epoch": 0.6874290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 5.321715987744662e-06,
-      "loss": 0.8094,
+      "learning_rate": 4.701430909916331e-06,
+      "loss": 0.868,
       "step": 24225
     },
     {
-      "epoch": 0.6653483837301915,
+      "epoch": 0.6874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 5.320929834818151e-06,
-      "loss": 0.8515,
+      "learning_rate": 4.700651474081977e-06,
+      "loss": 0.7698,
       "step": 24226
     },
     {
-      "epoch": 0.6653758479580346,
+      "epoch": 0.6874858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 5.3201437189142565e-06,
-      "loss": 0.8648,
+      "learning_rate": 4.6998720830114554e-06,
+      "loss": 0.8431,
       "step": 24227
     },
     {
-      "epoch": 0.6654033121858779,
+      "epoch": 0.6875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 5.319357640039193e-06,
-      "loss": 0.8329,
+      "learning_rate": 4.699092736711351e-06,
+      "loss": 0.7903,
       "step": 24228
     },
     {
-      "epoch": 0.6654307764137212,
+      "epoch": 0.6875425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 5.318571598199187e-06,
-      "loss": 0.9262,
+      "learning_rate": 4.698313435188254e-06,
+      "loss": 0.7959,
       "step": 24229
     },
     {
-      "epoch": 0.6654582406415643,
+      "epoch": 0.6875709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 5.317785593400454e-06,
-      "loss": 0.8084,
+      "learning_rate": 4.697534178448737e-06,
+      "loss": 0.8242,
       "step": 24230
     },
     {
-      "epoch": 0.6654857048694076,
+      "epoch": 0.6875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 5.316999625649215e-06,
-      "loss": 0.7659,
+      "learning_rate": 4.696754966499387e-06,
+      "loss": 0.828,
       "step": 24231
     },
     {
-      "epoch": 0.6655131690972508,
+      "epoch": 0.687627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 5.316213694951691e-06,
-      "loss": 0.8518,
+      "learning_rate": 4.69597579934679e-06,
+      "loss": 0.9081,
       "step": 24232
     },
     {
-      "epoch": 0.665540633325094,
+      "epoch": 0.687656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 5.315427801314101e-06,
-      "loss": 0.8589,
+      "learning_rate": 4.695196676997517e-06,
+      "loss": 0.8763,
       "step": 24233
     },
     {
-      "epoch": 0.6655680975529373,
+      "epoch": 0.6876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 5.314641944742657e-06,
-      "loss": 0.8599,
+      "learning_rate": 4.694417599458163e-06,
+      "loss": 0.894,
       "step": 24234
     },
     {
-      "epoch": 0.6655955617807805,
+      "epoch": 0.6877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 5.313856125243585e-06,
-      "loss": 0.8706,
+      "learning_rate": 4.693638566735298e-06,
+      "loss": 0.7652,
       "step": 24235
     },
     {
-      "epoch": 0.6656230260086238,
+      "epoch": 0.6877412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 5.313070342823094e-06,
-      "loss": 0.7769,
+      "learning_rate": 4.692859578835507e-06,
+      "loss": 0.7284,
       "step": 24236
     },
     {
-      "epoch": 0.665650490236467,
+      "epoch": 0.6877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 5.312284597487407e-06,
-      "loss": 0.7234,
+      "learning_rate": 4.6920806357653736e-06,
+      "loss": 0.8068,
       "step": 24237
     },
     {
-      "epoch": 0.6656779544643102,
+      "epoch": 0.6877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 5.311498889242743e-06,
-      "loss": 0.8382,
+      "learning_rate": 4.691301737531469e-06,
+      "loss": 0.9149,
       "step": 24238
     },
     {
-      "epoch": 0.6657054186921535,
+      "epoch": 0.6878263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 5.310713218095313e-06,
-      "loss": 0.9455,
+      "learning_rate": 4.690522884140379e-06,
+      "loss": 0.7652,
       "step": 24239
     },
     {
-      "epoch": 0.6657328829199967,
+      "epoch": 0.6878547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 5.309927584051335e-06,
-      "loss": 0.7691,
+      "learning_rate": 4.689744075598684e-06,
+      "loss": 0.7539,
       "step": 24240
     },
     {
-      "epoch": 0.6657603471478399,
+      "epoch": 0.687883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 5.309141987117028e-06,
-      "loss": 0.8312,
+      "learning_rate": 4.688965311912955e-06,
+      "loss": 0.8197,
       "step": 24241
     },
     {
-      "epoch": 0.6657878113756832,
+      "epoch": 0.6879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 5.308356427298611e-06,
-      "loss": 0.8333,
+      "learning_rate": 4.6881865930897756e-06,
+      "loss": 0.8546,
       "step": 24242
     },
     {
-      "epoch": 0.6658152756035264,
+      "epoch": 0.6879398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 5.3075709046022905e-06,
-      "loss": 0.8891,
+      "learning_rate": 4.687407919135726e-06,
+      "loss": 0.7869,
       "step": 24243
     },
     {
-      "epoch": 0.6658427398313697,
+      "epoch": 0.6879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 5.306785419034291e-06,
-      "loss": 0.8476,
+      "learning_rate": 4.686629290057377e-06,
+      "loss": 0.7508,
       "step": 24244
     },
     {
-      "epoch": 0.6658702040592128,
+      "epoch": 0.6879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 5.305999970600819e-06,
-      "loss": 0.808,
+      "learning_rate": 4.685850705861309e-06,
+      "loss": 0.7769,
       "step": 24245
     },
     {
-      "epoch": 0.6658976682870561,
+      "epoch": 0.6880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 5.305214559308094e-06,
-      "loss": 0.8524,
+      "learning_rate": 4.685072166554102e-06,
+      "loss": 0.7557,
       "step": 24246
     },
     {
-      "epoch": 0.6659251325148994,
+      "epoch": 0.6880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 5.3044291851623295e-06,
-      "loss": 0.8557,
+      "learning_rate": 4.684293672142327e-06,
+      "loss": 0.8953,
       "step": 24247
     },
     {
-      "epoch": 0.6659525967427425,
+      "epoch": 0.6880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 5.3036438481697405e-06,
-      "loss": 0.901,
+      "learning_rate": 4.683515222632562e-06,
+      "loss": 0.8795,
       "step": 24248
     },
     {
-      "epoch": 0.6659800609705858,
+      "epoch": 0.6881101021566401,
       "grad_norm": 0.0,
-      "learning_rate": 5.30285854833654e-06,
-      "loss": 0.8543,
+      "learning_rate": 4.682736818031382e-06,
+      "loss": 0.8963,
       "step": 24249
     },
     {
-      "epoch": 0.6660075251984291,
+      "epoch": 0.6881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 5.302073285668947e-06,
-      "loss": 0.808,
+      "learning_rate": 4.681958458345365e-06,
+      "loss": 0.8152,
       "step": 24250
     },
     {
-      "epoch": 0.6660349894262723,
+      "epoch": 0.68816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 5.30128806017317e-06,
-      "loss": 0.8828,
+      "learning_rate": 4.681180143581086e-06,
+      "loss": 0.8012,
       "step": 24251
     },
     {
-      "epoch": 0.6660624536541155,
+      "epoch": 0.6881952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 5.300502871855417e-06,
-      "loss": 0.8225,
+      "learning_rate": 4.680401873745114e-06,
+      "loss": 0.7954,
       "step": 24252
     },
     {
-      "epoch": 0.6660899178819587,
+      "epoch": 0.6882236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 5.299717720721906e-06,
-      "loss": 0.8216,
+      "learning_rate": 4.679623648844027e-06,
+      "loss": 0.7565,
       "step": 24253
     },
     {
-      "epoch": 0.666117382109802,
+      "epoch": 0.6882519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 5.29893260677885e-06,
-      "loss": 0.8504,
+      "learning_rate": 4.678845468884402e-06,
+      "loss": 0.8512,
       "step": 24254
     },
     {
-      "epoch": 0.6661448463376453,
+      "epoch": 0.6882803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 5.2981475300324596e-06,
-      "loss": 0.8672,
+      "learning_rate": 4.678067333872804e-06,
+      "loss": 0.9457,
       "step": 24255
     },
     {
-      "epoch": 0.6661723105654884,
+      "epoch": 0.6883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 5.297362490488951e-06,
-      "loss": 0.8436,
+      "learning_rate": 4.6772892438158115e-06,
+      "loss": 0.838,
       "step": 24256
     },
     {
-      "epoch": 0.6661997747933317,
+      "epoch": 0.6883371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 5.296577488154528e-06,
-      "loss": 0.8251,
+      "learning_rate": 4.6765111987199985e-06,
+      "loss": 0.9105,
       "step": 24257
     },
     {
-      "epoch": 0.6662272390211749,
+      "epoch": 0.6883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 5.295792523035405e-06,
-      "loss": 0.8476,
+      "learning_rate": 4.6757331985919315e-06,
+      "loss": 0.8977,
       "step": 24258
     },
     {
-      "epoch": 0.6662547032490181,
+      "epoch": 0.6883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 5.2950075951378e-06,
-      "loss": 0.8706,
+      "learning_rate": 4.674955243438186e-06,
+      "loss": 0.9675,
       "step": 24259
     },
     {
-      "epoch": 0.6662821674768614,
+      "epoch": 0.688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 5.294222704467913e-06,
-      "loss": 0.7319,
+      "learning_rate": 4.674177333265336e-06,
+      "loss": 0.8146,
       "step": 24260
     },
     {
-      "epoch": 0.6663096317047046,
+      "epoch": 0.6884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 5.293437851031957e-06,
-      "loss": 0.8422,
+      "learning_rate": 4.6733994680799466e-06,
+      "loss": 0.7207,
       "step": 24261
     },
     {
-      "epoch": 0.6663370959325479,
+      "epoch": 0.6884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 5.292653034836149e-06,
-      "loss": 0.8345,
+      "learning_rate": 4.672621647888591e-06,
+      "loss": 0.8395,
       "step": 24262
     },
     {
-      "epoch": 0.6663645601603911,
+      "epoch": 0.6885073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 5.29186825588669e-06,
-      "loss": 0.8826,
+      "learning_rate": 4.67184387269784e-06,
+      "loss": 0.8878,
       "step": 24263
     },
     {
-      "epoch": 0.6663920243882343,
+      "epoch": 0.6885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 5.291083514189792e-06,
-      "loss": 0.7743,
+      "learning_rate": 4.671066142514262e-06,
+      "loss": 0.873,
       "step": 24264
     },
     {
-      "epoch": 0.6664194886160776,
+      "epoch": 0.6885641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 5.290298809751666e-06,
-      "loss": 0.8616,
+      "learning_rate": 4.670288457344433e-06,
+      "loss": 0.7715,
       "step": 24265
     },
     {
-      "epoch": 0.6664469528439207,
+      "epoch": 0.6885925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 5.289514142578525e-06,
-      "loss": 0.8376,
+      "learning_rate": 4.669510817194913e-06,
+      "loss": 0.8116,
       "step": 24266
     },
     {
-      "epoch": 0.666474417071764,
+      "epoch": 0.6886208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 5.288729512676568e-06,
-      "loss": 0.7987,
+      "learning_rate": 4.668733222072275e-06,
+      "loss": 0.8907,
       "step": 24267
     },
     {
-      "epoch": 0.6665018812996073,
+      "epoch": 0.6886492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 5.287944920052011e-06,
-      "loss": 0.8914,
+      "learning_rate": 4.66795567198309e-06,
+      "loss": 0.8536,
       "step": 24268
     },
     {
-      "epoch": 0.6665293455274505,
+      "epoch": 0.688677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 5.287160364711057e-06,
-      "loss": 0.807,
+      "learning_rate": 4.66717816693392e-06,
+      "loss": 0.8836,
       "step": 24269
     },
     {
-      "epoch": 0.6665568097552937,
+      "epoch": 0.6887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 5.286375846659915e-06,
-      "loss": 0.7683,
+      "learning_rate": 4.666400706931335e-06,
+      "loss": 0.8797,
       "step": 24270
     },
     {
-      "epoch": 0.6665842739831369,
+      "epoch": 0.6887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 5.285591365904793e-06,
-      "loss": 0.8415,
+      "learning_rate": 4.6656232919819074e-06,
+      "loss": 0.7445,
       "step": 24271
     },
     {
-      "epoch": 0.6666117382109802,
+      "epoch": 0.6887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 5.284806922451898e-06,
-      "loss": 0.7799,
+      "learning_rate": 4.664845922092196e-06,
+      "loss": 0.7738,
       "step": 24272
     },
     {
-      "epoch": 0.6666392024388235,
+      "epoch": 0.6887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 5.284022516307439e-06,
-      "loss": 0.7563,
+      "learning_rate": 4.664068597268771e-06,
+      "loss": 0.8442,
       "step": 24273
     },
     {
-      "epoch": 0.6666666666666666,
+      "epoch": 0.6888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 5.2832381474776225e-06,
-      "loss": 0.978,
+      "learning_rate": 4.663291317518202e-06,
+      "loss": 0.831,
       "step": 24274
     },
     {
-      "epoch": 0.6666941308945099,
+      "epoch": 0.6888479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 5.282453815968653e-06,
-      "loss": 0.7947,
+      "learning_rate": 4.662514082847047e-06,
+      "loss": 0.8378,
       "step": 24275
     },
     {
-      "epoch": 0.6667215951223532,
+      "epoch": 0.6888762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 5.281669521786733e-06,
-      "loss": 0.8013,
+      "learning_rate": 4.661736893261876e-06,
+      "loss": 0.8158,
       "step": 24276
     },
     {
-      "epoch": 0.6667490593501963,
+      "epoch": 0.6889046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 5.280885264938072e-06,
-      "loss": 0.8191,
+      "learning_rate": 4.660959748769257e-06,
+      "loss": 0.8974,
       "step": 24277
     },
     {
-      "epoch": 0.6667765235780396,
+      "epoch": 0.6889330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 5.280101045428874e-06,
-      "loss": 0.8614,
+      "learning_rate": 4.660182649375747e-06,
+      "loss": 0.9133,
       "step": 24278
     },
     {
-      "epoch": 0.6668039878058828,
+      "epoch": 0.6889614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 5.279316863265344e-06,
-      "loss": 0.8567,
+      "learning_rate": 4.65940559508791e-06,
+      "loss": 0.8252,
       "step": 24279
     },
     {
-      "epoch": 0.6668314520337261,
+      "epoch": 0.6889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 5.278532718453693e-06,
-      "loss": 0.7909,
+      "learning_rate": 4.658628585912323e-06,
+      "loss": 0.8643,
       "step": 24280
     },
     {
-      "epoch": 0.6668589162615693,
+      "epoch": 0.6890181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 5.277748611000115e-06,
-      "loss": 0.8037,
+      "learning_rate": 4.657851621855536e-06,
+      "loss": 0.9136,
       "step": 24281
     },
     {
-      "epoch": 0.6668863804894125,
+      "epoch": 0.6890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 5.276964540910821e-06,
-      "loss": 0.8148,
+      "learning_rate": 4.657074702924116e-06,
+      "loss": 0.7288,
       "step": 24282
     },
     {
-      "epoch": 0.6669138447172558,
+      "epoch": 0.6890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 5.276180508192015e-06,
-      "loss": 0.9029,
+      "learning_rate": 4.656297829124631e-06,
+      "loss": 0.8925,
       "step": 24283
     },
     {
-      "epoch": 0.666941308945099,
+      "epoch": 0.6891032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 5.2753965128498955e-06,
-      "loss": 0.7473,
+      "learning_rate": 4.655521000463633e-06,
+      "loss": 0.8541,
       "step": 24284
     },
     {
-      "epoch": 0.6669687731729422,
+      "epoch": 0.6891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 5.274612554890669e-06,
-      "loss": 0.8297,
+      "learning_rate": 4.654744216947695e-06,
+      "loss": 0.8833,
       "step": 24285
     },
     {
-      "epoch": 0.6669962374007855,
+      "epoch": 0.6891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 5.2738286343205415e-06,
-      "loss": 0.7933,
+      "learning_rate": 4.6539674785833675e-06,
+      "loss": 0.8349,
       "step": 24286
     },
     {
-      "epoch": 0.6670237016286287,
+      "epoch": 0.689188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 5.27304475114571e-06,
-      "loss": 0.8124,
+      "learning_rate": 4.653190785377218e-06,
+      "loss": 0.7977,
       "step": 24287
     },
     {
-      "epoch": 0.667051165856472,
+      "epoch": 0.689216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 5.27226090537238e-06,
-      "loss": 0.8266,
+      "learning_rate": 4.65241413733581e-06,
+      "loss": 0.7903,
       "step": 24288
     },
     {
-      "epoch": 0.6670786300843152,
+      "epoch": 0.6892451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 5.2714770970067505e-06,
-      "loss": 0.7795,
+      "learning_rate": 4.651637534465696e-06,
+      "loss": 0.8264,
       "step": 24289
     },
     {
-      "epoch": 0.6671060943121584,
+      "epoch": 0.6892735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 5.270693326055032e-06,
-      "loss": 0.9232,
+      "learning_rate": 4.650860976773442e-06,
+      "loss": 0.8895,
       "step": 24290
     },
     {
-      "epoch": 0.6671335585400017,
+      "epoch": 0.6893019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 5.269909592523415e-06,
-      "loss": 0.886,
+      "learning_rate": 4.650084464265608e-06,
+      "loss": 0.7936,
       "step": 24291
     },
     {
-      "epoch": 0.6671610227678448,
+      "epoch": 0.6893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 5.26912589641811e-06,
-      "loss": 0.8457,
+      "learning_rate": 4.649307996948747e-06,
+      "loss": 0.7474,
       "step": 24292
     },
     {
-      "epoch": 0.6671884869956881,
+      "epoch": 0.6893586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 5.26834223774531e-06,
-      "loss": 0.7789,
+      "learning_rate": 4.6485315748294215e-06,
+      "loss": 0.8392,
       "step": 24293
     },
     {
-      "epoch": 0.6672159512235314,
+      "epoch": 0.6893870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 5.2675586165112166e-06,
-      "loss": 0.9431,
+      "learning_rate": 4.6477551979141914e-06,
+      "loss": 0.8096,
       "step": 24294
     },
     {
-      "epoch": 0.6672434154513746,
+      "epoch": 0.6894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 5.266775032722036e-06,
-      "loss": 0.8588,
+      "learning_rate": 4.646978866209613e-06,
+      "loss": 0.7904,
       "step": 24295
     },
     {
-      "epoch": 0.6672708796792178,
+      "epoch": 0.6894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 5.2659914863839625e-06,
-      "loss": 0.9028,
+      "learning_rate": 4.6462025797222445e-06,
+      "loss": 0.8667,
       "step": 24296
     },
     {
-      "epoch": 0.667298343907061,
+      "epoch": 0.689472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 5.2652079775032016e-06,
-      "loss": 0.7586,
+      "learning_rate": 4.645426338458648e-06,
+      "loss": 0.8227,
       "step": 24297
     },
     {
-      "epoch": 0.6673258081349043,
+      "epoch": 0.6895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 5.264424506085948e-06,
-      "loss": 0.9156,
+      "learning_rate": 4.644650142425372e-06,
+      "loss": 0.8597,
       "step": 24298
     },
     {
-      "epoch": 0.6673532723627476,
+      "epoch": 0.6895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 5.263641072138403e-06,
-      "loss": 0.7805,
+      "learning_rate": 4.643873991628977e-06,
+      "loss": 0.8835,
       "step": 24299
     },
     {
-      "epoch": 0.6673807365905907,
+      "epoch": 0.6895573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 5.262857675666762e-06,
-      "loss": 0.7858,
+      "learning_rate": 4.6430978860760236e-06,
+      "loss": 0.7703,
       "step": 24300
     },
     {
-      "epoch": 0.667408200818434,
+      "epoch": 0.6895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 5.262074316677227e-06,
-      "loss": 0.8751,
+      "learning_rate": 4.6423218257730585e-06,
+      "loss": 0.7694,
       "step": 24301
     },
     {
-      "epoch": 0.6674356650462773,
+      "epoch": 0.6896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 5.2612909951759935e-06,
-      "loss": 0.942,
+      "learning_rate": 4.641545810726642e-06,
+      "loss": 0.7766,
       "step": 24302
     },
     {
-      "epoch": 0.6674631292741204,
+      "epoch": 0.6896424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 5.260507711169267e-06,
-      "loss": 0.8773,
+      "learning_rate": 4.6407698409433325e-06,
+      "loss": 0.7259,
       "step": 24303
     },
     {
-      "epoch": 0.6674905935019637,
+      "epoch": 0.6896708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 5.259724464663234e-06,
-      "loss": 0.8004,
+      "learning_rate": 4.639993916429677e-06,
+      "loss": 0.9052,
       "step": 24304
     },
     {
-      "epoch": 0.6675180577298069,
+      "epoch": 0.6896992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 5.258941255664098e-06,
-      "loss": 0.7769,
+      "learning_rate": 4.639218037192235e-06,
+      "loss": 0.8891,
       "step": 24305
     },
     {
-      "epoch": 0.6675455219576502,
+      "epoch": 0.689727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 5.25815808417806e-06,
-      "loss": 0.91,
+      "learning_rate": 4.638442203237562e-06,
+      "loss": 0.8902,
       "step": 24306
     },
     {
-      "epoch": 0.6675729861854934,
+      "epoch": 0.6897559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 5.257374950211307e-06,
-      "loss": 0.8341,
+      "learning_rate": 4.637666414572205e-06,
+      "loss": 0.9772,
       "step": 24307
     },
     {
-      "epoch": 0.6676004504133366,
+      "epoch": 0.6897843359818389,
       "grad_norm": 0.0,
-      "learning_rate": 5.256591853770041e-06,
-      "loss": 0.8145,
+      "learning_rate": 4.636890671202725e-06,
+      "loss": 0.8117,
       "step": 24308
     },
     {
-      "epoch": 0.6676279146411799,
+      "epoch": 0.6898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 5.255808794860461e-06,
-      "loss": 0.8485,
+      "learning_rate": 4.636114973135663e-06,
+      "loss": 0.8992,
       "step": 24309
     },
     {
-      "epoch": 0.6676553788690232,
+      "epoch": 0.6898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 5.255025773488756e-06,
-      "loss": 0.7604,
+      "learning_rate": 4.635339320377582e-06,
+      "loss": 0.7975,
       "step": 24310
     },
     {
-      "epoch": 0.6676828430968663,
+      "epoch": 0.6898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 5.254242789661125e-06,
-      "loss": 0.8383,
+      "learning_rate": 4.634563712935036e-06,
+      "loss": 0.8696,
       "step": 24311
     },
     {
-      "epoch": 0.6677103073247096,
+      "epoch": 0.6898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 5.253459843383763e-06,
-      "loss": 0.809,
+      "learning_rate": 4.633788150814566e-06,
+      "loss": 0.6868,
       "step": 24312
     },
     {
-      "epoch": 0.6677377715525528,
+      "epoch": 0.6899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 5.252676934662866e-06,
-      "loss": 0.8681,
+      "learning_rate": 4.633012634022731e-06,
+      "loss": 0.8413,
       "step": 24313
     },
     {
-      "epoch": 0.667765235780396,
+      "epoch": 0.6899545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 5.251894063504632e-06,
-      "loss": 0.7876,
+      "learning_rate": 4.6322371625660825e-06,
+      "loss": 0.8156,
       "step": 24314
     },
     {
-      "epoch": 0.6677927000082393,
+      "epoch": 0.6899829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 5.251111229915251e-06,
-      "loss": 0.8506,
+      "learning_rate": 4.6314617364511625e-06,
+      "loss": 0.9067,
       "step": 24315
     },
     {
-      "epoch": 0.6678201642360825,
+      "epoch": 0.690011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 5.250328433900913e-06,
-      "loss": 0.8602,
+      "learning_rate": 4.630686355684528e-06,
+      "loss": 0.8559,
       "step": 24316
     },
     {
-      "epoch": 0.6678476284639258,
+      "epoch": 0.6900397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 5.249545675467817e-06,
-      "loss": 0.8239,
+      "learning_rate": 4.629911020272731e-06,
+      "loss": 0.7659,
       "step": 24317
     },
     {
-      "epoch": 0.6678750926917689,
+      "epoch": 0.6900681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 5.2487629546221556e-06,
-      "loss": 0.9934,
+      "learning_rate": 4.629135730222314e-06,
+      "loss": 0.9131,
       "step": 24318
     },
     {
-      "epoch": 0.6679025569196122,
+      "epoch": 0.6900964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 5.2479802713701234e-06,
-      "loss": 0.8674,
+      "learning_rate": 4.628360485539828e-06,
+      "loss": 0.9184,
       "step": 24319
     },
     {
-      "epoch": 0.6679300211474555,
+      "epoch": 0.6901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 5.24719762571791e-06,
-      "loss": 0.7707,
+      "learning_rate": 4.627585286231826e-06,
+      "loss": 0.8614,
       "step": 24320
     },
     {
-      "epoch": 0.6679574853752986,
+      "epoch": 0.6901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 5.246415017671716e-06,
-      "loss": 0.796,
+      "learning_rate": 4.626810132304848e-06,
+      "loss": 0.8427,
       "step": 24321
     },
     {
-      "epoch": 0.6679849496031419,
+      "epoch": 0.6901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 5.245632447237722e-06,
-      "loss": 0.8562,
+      "learning_rate": 4.626035023765448e-06,
+      "loss": 0.7521,
       "step": 24322
     },
     {
-      "epoch": 0.6680124138309852,
+      "epoch": 0.6902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 5.24484991442213e-06,
-      "loss": 0.887,
+      "learning_rate": 4.6252599606201755e-06,
+      "loss": 0.7317,
       "step": 24323
     },
     {
-      "epoch": 0.6680398780588284,
+      "epoch": 0.6902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 5.244067419231125e-06,
-      "loss": 0.822,
+      "learning_rate": 4.624484942875569e-06,
+      "loss": 0.8704,
       "step": 24324
     },
     {
-      "epoch": 0.6680673422866716,
+      "epoch": 0.690266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 5.2432849616708995e-06,
-      "loss": 0.7863,
+      "learning_rate": 4.623709970538181e-06,
+      "loss": 0.9418,
       "step": 24325
     },
     {
-      "epoch": 0.6680948065145148,
+      "epoch": 0.6902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 5.242502541747645e-06,
-      "loss": 0.8786,
+      "learning_rate": 4.622935043614555e-06,
+      "loss": 0.7611,
       "step": 24326
     },
     {
-      "epoch": 0.6681222707423581,
+      "epoch": 0.6903234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 5.241720159467559e-06,
-      "loss": 0.8292,
+      "learning_rate": 4.622160162111239e-06,
+      "loss": 0.9066,
       "step": 24327
     },
     {
-      "epoch": 0.6681497349702014,
+      "epoch": 0.6903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 5.240937814836821e-06,
-      "loss": 0.895,
+      "learning_rate": 4.6213853260347816e-06,
+      "loss": 0.8243,
       "step": 24328
     },
     {
-      "epoch": 0.6681771991980445,
+      "epoch": 0.6903802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 5.240155507861628e-06,
-      "loss": 0.8704,
+      "learning_rate": 4.620610535391721e-06,
+      "loss": 0.7875,
       "step": 24329
     },
     {
-      "epoch": 0.6682046634258878,
+      "epoch": 0.6904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 5.239373238548171e-06,
-      "loss": 0.7817,
+      "learning_rate": 4.619835790188605e-06,
+      "loss": 0.7596,
       "step": 24330
     },
     {
-      "epoch": 0.668232127653731,
+      "epoch": 0.6904370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 5.238591006902634e-06,
-      "loss": 0.8245,
+      "learning_rate": 4.619061090431981e-06,
+      "loss": 0.7799,
       "step": 24331
     },
     {
-      "epoch": 0.6682595918815742,
+      "epoch": 0.6904653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 5.237808812931209e-06,
-      "loss": 0.815,
+      "learning_rate": 4.618286436128386e-06,
+      "loss": 0.8789,
       "step": 24332
     },
     {
-      "epoch": 0.6682870561094175,
+      "epoch": 0.6904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 5.237026656640089e-06,
-      "loss": 0.7573,
+      "learning_rate": 4.617511827284368e-06,
+      "loss": 0.8221,
       "step": 24333
     },
     {
-      "epoch": 0.6683145203372607,
+      "epoch": 0.690522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 5.236244538035456e-06,
-      "loss": 0.6907,
+      "learning_rate": 4.616737263906473e-06,
+      "loss": 0.8486,
       "step": 24334
     },
     {
-      "epoch": 0.668341984565104,
+      "epoch": 0.690550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 5.235462457123501e-06,
-      "loss": 0.8548,
+      "learning_rate": 4.615962746001237e-06,
+      "loss": 0.8449,
       "step": 24335
     },
     {
-      "epoch": 0.6683694487929472,
+      "epoch": 0.6905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 5.234680413910413e-06,
-      "loss": 0.9024,
+      "learning_rate": 4.615188273575205e-06,
+      "loss": 0.7157,
       "step": 24336
     },
     {
-      "epoch": 0.6683969130207904,
+      "epoch": 0.6906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 5.233898408402379e-06,
-      "loss": 0.798,
+      "learning_rate": 4.614413846634924e-06,
+      "loss": 0.7895,
       "step": 24337
     },
     {
-      "epoch": 0.6684243772486337,
+      "epoch": 0.6906356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 5.2331164406055916e-06,
-      "loss": 0.8122,
+      "learning_rate": 4.613639465186928e-06,
+      "loss": 0.9937,
       "step": 24338
     },
     {
-      "epoch": 0.6684518414764768,
+      "epoch": 0.6906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 5.232334510526233e-06,
-      "loss": 0.8053,
+      "learning_rate": 4.61286512923776e-06,
+      "loss": 0.8522,
       "step": 24339
     },
     {
-      "epoch": 0.6684793057043201,
+      "epoch": 0.6906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 5.231552618170486e-06,
-      "loss": 0.817,
+      "learning_rate": 4.612090838793964e-06,
+      "loss": 0.8907,
       "step": 24340
     },
     {
-      "epoch": 0.6685067699321634,
+      "epoch": 0.6907207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 5.230770763544541e-06,
-      "loss": 0.8451,
+      "learning_rate": 4.6113165938620785e-06,
+      "loss": 0.7831,
       "step": 24341
     },
     {
-      "epoch": 0.6685342341600066,
+      "epoch": 0.6907491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 5.229988946654587e-06,
-      "loss": 0.8015,
+      "learning_rate": 4.6105423944486475e-06,
+      "loss": 0.8131,
       "step": 24342
     },
     {
-      "epoch": 0.6685616983878498,
+      "epoch": 0.69077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.229207167506809e-06,
-      "loss": 0.8932,
+      "learning_rate": 4.609768240560204e-06,
+      "loss": 0.7277,
       "step": 24343
     },
     {
-      "epoch": 0.668589162615693,
+      "epoch": 0.6908059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 5.2284254261073905e-06,
-      "loss": 0.7571,
+      "learning_rate": 4.608994132203289e-06,
+      "loss": 0.8275,
       "step": 24344
     },
     {
-      "epoch": 0.6686166268435363,
+      "epoch": 0.6908342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 5.227643722462522e-06,
-      "loss": 0.7588,
+      "learning_rate": 4.608220069384448e-06,
+      "loss": 0.8874,
       "step": 24345
     },
     {
-      "epoch": 0.6686440910713796,
+      "epoch": 0.6908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 5.2268620565783815e-06,
-      "loss": 0.8355,
+      "learning_rate": 4.607446052110211e-06,
+      "loss": 0.6726,
       "step": 24346
     },
     {
-      "epoch": 0.6686715552992227,
+      "epoch": 0.6908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 5.226080428461161e-06,
-      "loss": 0.921,
+      "learning_rate": 4.606672080387118e-06,
+      "loss": 0.7673,
       "step": 24347
     },
     {
-      "epoch": 0.668699019527066,
+      "epoch": 0.6909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 5.225298838117038e-06,
-      "loss": 0.8586,
+      "learning_rate": 4.605898154221713e-06,
+      "loss": 0.7489,
       "step": 24348
     },
     {
-      "epoch": 0.6687264837549093,
+      "epoch": 0.6909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 5.2245172855522e-06,
-      "loss": 0.9173,
+      "learning_rate": 4.605124273620526e-06,
+      "loss": 0.7696,
       "step": 24349
     },
     {
-      "epoch": 0.6687539479827525,
+      "epoch": 0.6909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 5.223735770772831e-06,
-      "loss": 0.8959,
+      "learning_rate": 4.604350438590095e-06,
+      "loss": 0.8122,
       "step": 24350
     },
     {
-      "epoch": 0.6687814122105957,
+      "epoch": 0.6910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 5.222954293785118e-06,
-      "loss": 0.7599,
+      "learning_rate": 4.603576649136964e-06,
+      "loss": 0.7394,
       "step": 24351
     },
     {
-      "epoch": 0.6688088764384389,
+      "epoch": 0.6910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 5.222172854595238e-06,
-      "loss": 0.8155,
+      "learning_rate": 4.602802905267658e-06,
+      "loss": 0.7427,
       "step": 24352
     },
     {
-      "epoch": 0.6688363406662822,
+      "epoch": 0.691061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 5.221391453209377e-06,
-      "loss": 0.7999,
+      "learning_rate": 4.60202920698872e-06,
+      "loss": 0.9177,
       "step": 24353
     },
     {
-      "epoch": 0.6688638048941254,
+      "epoch": 0.6910896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 5.220610089633722e-06,
-      "loss": 0.9122,
+      "learning_rate": 4.601255554306686e-06,
+      "loss": 0.8435,
       "step": 24354
     },
     {
-      "epoch": 0.6688912691219686,
+      "epoch": 0.6911180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 5.219828763874446e-06,
-      "loss": 0.896,
+      "learning_rate": 4.600481947228084e-06,
+      "loss": 0.8192,
       "step": 24355
     },
     {
-      "epoch": 0.6689187333498119,
+      "epoch": 0.6911464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 5.219047475937737e-06,
-      "loss": 0.8768,
+      "learning_rate": 4.5997083857594595e-06,
+      "loss": 0.8119,
       "step": 24356
     },
     {
-      "epoch": 0.668946197577655,
+      "epoch": 0.6911748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 5.2182662258297826e-06,
-      "loss": 0.8622,
+      "learning_rate": 4.598934869907337e-06,
+      "loss": 0.7984,
       "step": 24357
     },
     {
-      "epoch": 0.6689736618054983,
+      "epoch": 0.6912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 5.2174850135567516e-06,
-      "loss": 0.8028,
+      "learning_rate": 4.5981613996782546e-06,
+      "loss": 0.7974,
       "step": 24358
     },
     {
-      "epoch": 0.6690011260333416,
+      "epoch": 0.6912315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 5.216703839124833e-06,
-      "loss": 0.8524,
+      "learning_rate": 4.597387975078751e-06,
+      "loss": 0.8704,
       "step": 24359
     },
     {
-      "epoch": 0.6690285902611848,
+      "epoch": 0.6912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 5.215922702540206e-06,
-      "loss": 0.9636,
+      "learning_rate": 4.596614596115348e-06,
+      "loss": 0.8401,
       "step": 24360
     },
     {
-      "epoch": 0.669056054489028,
+      "epoch": 0.6912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 5.215141603809051e-06,
-      "loss": 0.8182,
+      "learning_rate": 4.595841262794586e-06,
+      "loss": 0.9305,
       "step": 24361
     },
     {
-      "epoch": 0.6690835187168713,
+      "epoch": 0.691316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 5.2143605429375535e-06,
-      "loss": 0.8949,
+      "learning_rate": 4.595067975122998e-06,
+      "loss": 0.8496,
       "step": 24362
     },
     {
-      "epoch": 0.6691109829447145,
+      "epoch": 0.6913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 5.21357951993189e-06,
-      "loss": 0.794,
+      "learning_rate": 4.594294733107112e-06,
+      "loss": 0.7664,
       "step": 24363
     },
     {
-      "epoch": 0.6691384471725578,
+      "epoch": 0.6913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 5.212798534798235e-06,
-      "loss": 0.8302,
+      "learning_rate": 4.5935215367534616e-06,
+      "loss": 0.8935,
       "step": 24364
     },
     {
-      "epoch": 0.6691659114004009,
+      "epoch": 0.6914018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 5.2120175875427726e-06,
-      "loss": 0.8277,
+      "learning_rate": 4.592748386068579e-06,
+      "loss": 0.7684,
       "step": 24365
     },
     {
-      "epoch": 0.6691933756282442,
+      "epoch": 0.6914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 5.2112366781716805e-06,
-      "loss": 0.7338,
+      "learning_rate": 4.591975281058992e-06,
+      "loss": 0.8017,
       "step": 24366
     },
     {
-      "epoch": 0.6692208398560875,
+      "epoch": 0.6914585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 5.21045580669114e-06,
-      "loss": 0.8423,
+      "learning_rate": 4.591202221731232e-06,
+      "loss": 0.851,
       "step": 24367
     },
     {
-      "epoch": 0.6692483040839307,
+      "epoch": 0.6914869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 5.209674973107333e-06,
-      "loss": 0.7932,
+      "learning_rate": 4.5904292080918355e-06,
+      "loss": 0.7464,
       "step": 24368
     },
     {
-      "epoch": 0.6692757683117739,
+      "epoch": 0.6915153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 5.208894177426427e-06,
-      "loss": 0.7794,
+      "learning_rate": 4.589656240147321e-06,
+      "loss": 0.7827,
       "step": 24369
     },
     {
-      "epoch": 0.6693032325396171,
+      "epoch": 0.6915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 5.208113419654608e-06,
-      "loss": 0.8698,
+      "learning_rate": 4.588883317904224e-06,
+      "loss": 0.8081,
       "step": 24370
     },
     {
-      "epoch": 0.6693306967674604,
+      "epoch": 0.691572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 5.207332699798055e-06,
-      "loss": 0.9226,
+      "learning_rate": 4.588110441369074e-06,
+      "loss": 0.8595,
       "step": 24371
     },
     {
-      "epoch": 0.6693581609953037,
+      "epoch": 0.6916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 5.2065520178629395e-06,
-      "loss": 0.8256,
+      "learning_rate": 4.587337610548397e-06,
+      "loss": 0.8294,
       "step": 24372
     },
     {
-      "epoch": 0.6693856252231468,
+      "epoch": 0.6916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 5.205771373855439e-06,
-      "loss": 0.792,
+      "learning_rate": 4.586564825448726e-06,
+      "loss": 0.8238,
       "step": 24373
     },
     {
-      "epoch": 0.6694130894509901,
+      "epoch": 0.6916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 5.204990767781739e-06,
-      "loss": 0.924,
+      "learning_rate": 4.5857920860765825e-06,
+      "loss": 0.7737,
       "step": 24374
     },
     {
-      "epoch": 0.6694405536788334,
+      "epoch": 0.6916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 5.204210199648003e-06,
-      "loss": 0.8865,
+      "learning_rate": 4.585019392438496e-06,
+      "loss": 0.8418,
       "step": 24375
     },
     {
-      "epoch": 0.6694680179066765,
+      "epoch": 0.6917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 5.203429669460416e-06,
-      "loss": 0.878,
+      "learning_rate": 4.584246744540998e-06,
+      "loss": 0.8697,
       "step": 24376
     },
     {
-      "epoch": 0.6694954821345198,
+      "epoch": 0.6917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 5.20264917722515e-06,
-      "loss": 0.903,
+      "learning_rate": 4.583474142390608e-06,
+      "loss": 0.8895,
       "step": 24377
     },
     {
-      "epoch": 0.669522946362363,
+      "epoch": 0.6917707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 5.201868722948386e-06,
-      "loss": 0.8968,
+      "learning_rate": 4.582701585993855e-06,
+      "loss": 0.8264,
       "step": 24378
     },
     {
-      "epoch": 0.6695504105902063,
+      "epoch": 0.6917990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 5.201088306636293e-06,
-      "loss": 0.7956,
+      "learning_rate": 4.581929075357269e-06,
+      "loss": 0.7989,
       "step": 24379
     },
     {
-      "epoch": 0.6695778748180495,
+      "epoch": 0.6918274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 5.2003079282950495e-06,
-      "loss": 0.7433,
+      "learning_rate": 4.581156610487367e-06,
+      "loss": 0.8775,
       "step": 24380
     },
     {
-      "epoch": 0.6696053390458927,
+      "epoch": 0.6918558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 5.199527587930826e-06,
-      "loss": 0.8513,
+      "learning_rate": 4.58038419139068e-06,
+      "loss": 0.8412,
       "step": 24381
     },
     {
-      "epoch": 0.669632803273736,
+      "epoch": 0.6918842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 5.198747285549801e-06,
-      "loss": 0.7914,
+      "learning_rate": 4.579611818073735e-06,
+      "loss": 0.7499,
       "step": 24382
     },
     {
-      "epoch": 0.6696602675015791,
+      "epoch": 0.6919125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 5.197967021158146e-06,
-      "loss": 0.8379,
+      "learning_rate": 4.578839490543049e-06,
+      "loss": 0.7641,
       "step": 24383
     },
     {
-      "epoch": 0.6696877317294224,
+      "epoch": 0.6919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 5.197186794762037e-06,
-      "loss": 0.9076,
+      "learning_rate": 4.578067208805148e-06,
+      "loss": 0.7593,
       "step": 24384
     },
     {
-      "epoch": 0.6697151959572657,
+      "epoch": 0.6919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 5.196406606367646e-06,
-      "loss": 0.8549,
+      "learning_rate": 4.577294972866563e-06,
+      "loss": 0.8444,
       "step": 24385
     },
     {
-      "epoch": 0.6697426601851089,
+      "epoch": 0.6919977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 5.195626455981149e-06,
-      "loss": 0.7742,
+      "learning_rate": 4.576522782733802e-06,
+      "loss": 0.8126,
       "step": 24386
     },
     {
-      "epoch": 0.6697701244129521,
+      "epoch": 0.6920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 5.194846343608719e-06,
-      "loss": 0.8537,
+      "learning_rate": 4.5757506384134005e-06,
+      "loss": 0.9033,
       "step": 24387
     },
     {
-      "epoch": 0.6697975886407954,
+      "epoch": 0.6920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 5.194066269256522e-06,
-      "loss": 0.8911,
+      "learning_rate": 4.574978539911881e-06,
+      "loss": 0.7732,
       "step": 24388
     },
     {
-      "epoch": 0.6698250528686386,
+      "epoch": 0.692082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 5.1932862329307324e-06,
-      "loss": 0.8054,
+      "learning_rate": 4.5742064872357594e-06,
+      "loss": 0.8048,
       "step": 24389
     },
     {
-      "epoch": 0.6698525170964819,
+      "epoch": 0.69211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 5.192506234637525e-06,
-      "loss": 0.853,
+      "learning_rate": 4.573434480391559e-06,
+      "loss": 0.8249,
       "step": 24390
     },
     {
-      "epoch": 0.669879981324325,
+      "epoch": 0.6921396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 5.191726274383071e-06,
-      "loss": 0.8584,
+      "learning_rate": 4.572662519385804e-06,
+      "loss": 0.8611,
       "step": 24391
     },
     {
-      "epoch": 0.6699074455521683,
+      "epoch": 0.6921679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 5.1909463521735445e-06,
-      "loss": 0.7687,
+      "learning_rate": 4.57189060422501e-06,
+      "loss": 0.8614,
       "step": 24392
     },
     {
-      "epoch": 0.6699349097800116,
+      "epoch": 0.6921963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 5.19016646801511e-06,
-      "loss": 0.8368,
+      "learning_rate": 4.571118734915704e-06,
+      "loss": 0.8539,
       "step": 24393
     },
     {
-      "epoch": 0.6699623740078547,
+      "epoch": 0.6922247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 5.189386621913941e-06,
-      "loss": 0.8502,
+      "learning_rate": 4.570346911464397e-06,
+      "loss": 0.8394,
       "step": 24394
     },
     {
-      "epoch": 0.669989838235698,
+      "epoch": 0.6922531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 5.188606813876212e-06,
-      "loss": 0.8676,
+      "learning_rate": 4.569575133877615e-06,
+      "loss": 0.8328,
       "step": 24395
     },
     {
-      "epoch": 0.6700173024635412,
+      "epoch": 0.6922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 5.187827043908087e-06,
-      "loss": 0.9104,
+      "learning_rate": 4.5688034021618795e-06,
+      "loss": 0.9449,
       "step": 24396
     },
     {
-      "epoch": 0.6700447666913845,
+      "epoch": 0.6923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 5.187047312015738e-06,
-      "loss": 0.8766,
+      "learning_rate": 4.568031716323702e-06,
+      "loss": 0.7913,
       "step": 24397
     },
     {
-      "epoch": 0.6700722309192277,
+      "epoch": 0.6923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 5.1862676182053384e-06,
-      "loss": 0.7983,
+      "learning_rate": 4.567260076369605e-06,
+      "loss": 0.9145,
       "step": 24398
     },
     {
-      "epoch": 0.6700996951470709,
+      "epoch": 0.692366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 5.185487962483051e-06,
-      "loss": 0.8349,
+      "learning_rate": 4.56648848230611e-06,
+      "loss": 0.7998,
       "step": 24399
     },
     {
-      "epoch": 0.6701271593749142,
+      "epoch": 0.6923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.184708344855047e-06,
-      "loss": 0.8533,
+      "learning_rate": 4.5657169341397265e-06,
+      "loss": 0.8862,
       "step": 24400
     },
     {
-      "epoch": 0.6701546236027575,
+      "epoch": 0.6924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 5.183928765327497e-06,
-      "loss": 0.7288,
+      "learning_rate": 4.564945431876973e-06,
+      "loss": 0.9194,
       "step": 24401
     },
     {
-      "epoch": 0.6701820878306006,
+      "epoch": 0.6924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 5.183149223906572e-06,
-      "loss": 0.7829,
+      "learning_rate": 4.564173975524377e-06,
+      "loss": 0.8455,
       "step": 24402
     },
     {
-      "epoch": 0.6702095520584439,
+      "epoch": 0.6924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 5.1823697205984305e-06,
-      "loss": 0.8882,
+      "learning_rate": 4.563402565088445e-06,
+      "loss": 0.8862,
       "step": 24403
     },
     {
-      "epoch": 0.6702370162862871,
+      "epoch": 0.6925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 5.181590255409251e-06,
-      "loss": 0.8511,
+      "learning_rate": 4.562631200575696e-06,
+      "loss": 0.8411,
       "step": 24404
     },
     {
-      "epoch": 0.6702644805141303,
+      "epoch": 0.6925368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 5.180810828345193e-06,
-      "loss": 0.8665,
+      "learning_rate": 4.561859881992649e-06,
+      "loss": 0.8013,
       "step": 24405
     },
     {
-      "epoch": 0.6702919447419736,
+      "epoch": 0.6925652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 5.180031439412425e-06,
-      "loss": 0.8802,
+      "learning_rate": 4.561088609345812e-06,
+      "loss": 0.7323,
       "step": 24406
     },
     {
-      "epoch": 0.6703194089698168,
+      "epoch": 0.6925936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 5.179252088617115e-06,
-      "loss": 0.8099,
+      "learning_rate": 4.560317382641704e-06,
+      "loss": 0.8537,
       "step": 24407
     },
     {
-      "epoch": 0.6703468731976601,
+      "epoch": 0.692622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 5.178472775965431e-06,
-      "loss": 0.8504,
+      "learning_rate": 4.559546201886843e-06,
+      "loss": 0.8818,
       "step": 24408
     },
     {
-      "epoch": 0.6703743374255032,
+      "epoch": 0.6926503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 5.1776935014635375e-06,
-      "loss": 0.8572,
+      "learning_rate": 4.558775067087736e-06,
+      "loss": 0.8904,
       "step": 24409
     },
     {
-      "epoch": 0.6704018016533465,
+      "epoch": 0.6926787741203179,
       "grad_norm": 0.0,
-      "learning_rate": 5.176914265117605e-06,
-      "loss": 0.7832,
+      "learning_rate": 4.558003978250901e-06,
+      "loss": 0.8557,
       "step": 24410
     },
     {
-      "epoch": 0.6704292658811898,
+      "epoch": 0.6927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 5.1761350669337965e-06,
-      "loss": 0.8873,
+      "learning_rate": 4.5572329353828546e-06,
+      "loss": 0.8774,
       "step": 24411
     },
     {
-      "epoch": 0.670456730109033,
+      "epoch": 0.6927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 5.1753559069182705e-06,
-      "loss": 0.8518,
+      "learning_rate": 4.5564619384901035e-06,
+      "loss": 0.8756,
       "step": 24412
     },
     {
-      "epoch": 0.6704841943368762,
+      "epoch": 0.6927639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 5.174576785077196e-06,
-      "loss": 0.8111,
+      "learning_rate": 4.555690987579162e-06,
+      "loss": 0.8257,
       "step": 24413
     },
     {
-      "epoch": 0.6705116585647195,
+      "epoch": 0.6927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 5.173797701416743e-06,
-      "loss": 0.828,
+      "learning_rate": 4.554920082656548e-06,
+      "loss": 0.951,
       "step": 24414
     },
     {
-      "epoch": 0.6705391227925627,
+      "epoch": 0.6928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 5.173018655943069e-06,
-      "loss": 0.8388,
+      "learning_rate": 4.554149223728764e-06,
+      "loss": 0.8599,
       "step": 24415
     },
     {
-      "epoch": 0.670566587020406,
+      "epoch": 0.6928490351872871,
       "grad_norm": 0.0,
-      "learning_rate": 5.172239648662346e-06,
-      "loss": 0.8151,
+      "learning_rate": 4.553378410802331e-06,
+      "loss": 0.8266,
       "step": 24416
     },
     {
-      "epoch": 0.6705940512482491,
+      "epoch": 0.6928774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 5.171460679580729e-06,
-      "loss": 0.8937,
+      "learning_rate": 4.5526076438837486e-06,
+      "loss": 0.7777,
       "step": 24417
     },
     {
-      "epoch": 0.6706215154760924,
+      "epoch": 0.692905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 5.1706817487043856e-06,
-      "loss": 0.82,
+      "learning_rate": 4.551836922979537e-06,
+      "loss": 0.8059,
       "step": 24418
     },
     {
-      "epoch": 0.6706489797039357,
+      "epoch": 0.6929341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 5.169902856039484e-06,
-      "loss": 0.8576,
+      "learning_rate": 4.551066248096208e-06,
+      "loss": 0.8177,
       "step": 24419
     },
     {
-      "epoch": 0.6706764439317788,
+      "epoch": 0.6929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 5.169124001592176e-06,
-      "loss": 0.8857,
+      "learning_rate": 4.550295619240262e-06,
+      "loss": 0.8975,
       "step": 24420
     },
     {
-      "epoch": 0.6707039081596221,
+      "epoch": 0.6929909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 5.168345185368635e-06,
-      "loss": 0.9073,
+      "learning_rate": 4.549525036418217e-06,
+      "loss": 0.7323,
       "step": 24421
     },
     {
-      "epoch": 0.6707313723874653,
+      "epoch": 0.6930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 5.1675664073750155e-06,
-      "loss": 0.8307,
+      "learning_rate": 4.54875449963658e-06,
+      "loss": 0.8222,
       "step": 24422
     },
     {
-      "epoch": 0.6707588366153086,
+      "epoch": 0.6930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 5.166787667617481e-06,
-      "loss": 0.8833,
+      "learning_rate": 4.547984008901855e-06,
+      "loss": 0.8169,
       "step": 24423
     },
     {
-      "epoch": 0.6707863008431518,
+      "epoch": 0.6930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 5.166008966102196e-06,
-      "loss": 0.8188,
+      "learning_rate": 4.547213564220556e-06,
+      "loss": 0.6929,
       "step": 24424
     },
     {
-      "epoch": 0.670813765070995,
+      "epoch": 0.6931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 5.165230302835319e-06,
-      "loss": 0.843,
+      "learning_rate": 4.546443165599193e-06,
+      "loss": 0.8123,
       "step": 24425
     },
     {
-      "epoch": 0.6708412292988383,
+      "epoch": 0.6931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 5.164451677823017e-06,
-      "loss": 0.8727,
+      "learning_rate": 4.545672813044264e-06,
+      "loss": 0.8244,
       "step": 24426
     },
     {
-      "epoch": 0.6708686935266815,
+      "epoch": 0.693161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 5.163673091071448e-06,
-      "loss": 0.8659,
+      "learning_rate": 4.544902506562283e-06,
+      "loss": 0.879,
       "step": 24427
     },
     {
-      "epoch": 0.6708961577545247,
+      "epoch": 0.6931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 5.162894542586765e-06,
-      "loss": 0.7893,
+      "learning_rate": 4.54413224615976e-06,
+      "loss": 0.9296,
       "step": 24428
     },
     {
-      "epoch": 0.670923621982368,
+      "epoch": 0.6932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 5.1621160323751355e-06,
-      "loss": 0.9391,
+      "learning_rate": 4.543362031843193e-06,
+      "loss": 0.9436,
       "step": 24429
     },
     {
-      "epoch": 0.6709510862102112,
+      "epoch": 0.6932463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 5.161337560442718e-06,
-      "loss": 0.8787,
+      "learning_rate": 4.542591863619092e-06,
+      "loss": 0.849,
       "step": 24430
     },
     {
-      "epoch": 0.6709785504380544,
+      "epoch": 0.6932746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 5.160559126795671e-06,
-      "loss": 0.8128,
+      "learning_rate": 4.541821741493967e-06,
+      "loss": 0.7616,
       "step": 24431
     },
     {
-      "epoch": 0.6710060146658977,
+      "epoch": 0.6933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 5.159780731440156e-06,
-      "loss": 0.8535,
+      "learning_rate": 4.541051665474311e-06,
+      "loss": 0.8444,
       "step": 24432
     },
     {
-      "epoch": 0.6710334788937409,
+      "epoch": 0.6933314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 5.159002374382333e-06,
-      "loss": 0.8228,
+      "learning_rate": 4.540281635566645e-06,
+      "loss": 0.9787,
       "step": 24433
     },
     {
-      "epoch": 0.6710609431215842,
+      "epoch": 0.6933598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 5.158224055628356e-06,
-      "loss": 0.7936,
+      "learning_rate": 4.539511651777461e-06,
+      "loss": 0.6743,
       "step": 24434
     },
     {
-      "epoch": 0.6710884073494273,
+      "epoch": 0.6933881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 5.1574457751843886e-06,
-      "loss": 0.9166,
+      "learning_rate": 4.538741714113268e-06,
+      "loss": 0.868,
       "step": 24435
     },
     {
-      "epoch": 0.6711158715772706,
+      "epoch": 0.693416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 5.156667533056583e-06,
-      "loss": 0.8864,
+      "learning_rate": 4.537971822580573e-06,
+      "loss": 0.7874,
       "step": 24436
     },
     {
-      "epoch": 0.6711433358051139,
+      "epoch": 0.69344494892168,
       "grad_norm": 0.0,
-      "learning_rate": 5.1558893292511e-06,
-      "loss": 0.8804,
+      "learning_rate": 4.537201977185872e-06,
+      "loss": 0.8689,
       "step": 24437
     },
     {
-      "epoch": 0.671170800032957,
+      "epoch": 0.6934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 5.1551111637740966e-06,
-      "loss": 0.9355,
+      "learning_rate": 4.536432177935672e-06,
+      "loss": 0.8045,
       "step": 24438
     },
     {
-      "epoch": 0.6711982642608003,
+      "epoch": 0.6935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 5.154333036631736e-06,
-      "loss": 0.8062,
+      "learning_rate": 4.535662424836478e-06,
+      "loss": 0.787,
       "step": 24439
     },
     {
-      "epoch": 0.6712257284886436,
+      "epoch": 0.6935300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 5.153554947830164e-06,
-      "loss": 0.8449,
+      "learning_rate": 4.534892717894785e-06,
+      "loss": 0.8185,
       "step": 24440
     },
     {
-      "epoch": 0.6712531927164868,
+      "epoch": 0.6935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 5.152776897375542e-06,
-      "loss": 0.8112,
+      "learning_rate": 4.5341230571171e-06,
+      "loss": 0.7905,
       "step": 24441
     },
     {
-      "epoch": 0.67128065694433,
+      "epoch": 0.6935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 5.151998885274028e-06,
-      "loss": 0.8963,
+      "learning_rate": 4.5333534425099265e-06,
+      "loss": 0.8249,
       "step": 24442
     },
     {
-      "epoch": 0.6713081211721732,
+      "epoch": 0.6936152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 5.15122091153178e-06,
-      "loss": 0.8359,
+      "learning_rate": 4.532583874079758e-06,
+      "loss": 0.8335,
       "step": 24443
     },
     {
-      "epoch": 0.6713355854000165,
+      "epoch": 0.6936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 5.150442976154946e-06,
-      "loss": 0.8285,
+      "learning_rate": 4.531814351833099e-06,
+      "loss": 0.9423,
       "step": 24444
     },
     {
-      "epoch": 0.6713630496278598,
+      "epoch": 0.6936719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 5.14966507914969e-06,
-      "loss": 0.9212,
+      "learning_rate": 4.5310448757764545e-06,
+      "loss": 0.7805,
       "step": 24445
     },
     {
-      "epoch": 0.6713905138557029,
+      "epoch": 0.6937003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 5.14888722052216e-06,
-      "loss": 0.8978,
+      "learning_rate": 4.530275445916317e-06,
+      "loss": 0.8772,
       "step": 24446
     },
     {
-      "epoch": 0.6714179780835462,
+      "epoch": 0.6937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 5.148109400278511e-06,
-      "loss": 0.8074,
+      "learning_rate": 4.529506062259187e-06,
+      "loss": 0.8464,
       "step": 24447
     },
     {
-      "epoch": 0.6714454423113894,
+      "epoch": 0.6937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 5.147331618424902e-06,
-      "loss": 0.8091,
+      "learning_rate": 4.528736724811565e-06,
+      "loss": 0.8745,
       "step": 24448
     },
     {
-      "epoch": 0.6714729065392326,
+      "epoch": 0.6937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 5.146553874967485e-06,
-      "loss": 0.8449,
+      "learning_rate": 4.527967433579952e-06,
+      "loss": 0.8876,
       "step": 24449
     },
     {
-      "epoch": 0.6715003707670759,
+      "epoch": 0.6938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 5.145776169912417e-06,
-      "loss": 0.8354,
+      "learning_rate": 4.527198188570846e-06,
+      "loss": 0.8629,
       "step": 24450
     },
     {
-      "epoch": 0.6715278349949191,
+      "epoch": 0.6938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 5.144998503265848e-06,
-      "loss": 0.8934,
+      "learning_rate": 4.5264289897907396e-06,
+      "loss": 0.8666,
       "step": 24451
     },
     {
-      "epoch": 0.6715552992227624,
+      "epoch": 0.6938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 5.144220875033928e-06,
-      "loss": 0.895,
+      "learning_rate": 4.525659837246134e-06,
+      "loss": 0.7817,
       "step": 24452
     },
     {
-      "epoch": 0.6715827634506056,
+      "epoch": 0.6938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 5.143443285222814e-06,
-      "loss": 0.8633,
+      "learning_rate": 4.524890730943528e-06,
+      "loss": 0.8021,
       "step": 24453
     },
     {
-      "epoch": 0.6716102276784488,
+      "epoch": 0.6939273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 5.142665733838657e-06,
-      "loss": 0.8679,
+      "learning_rate": 4.524121670889414e-06,
+      "loss": 0.9449,
       "step": 24454
     },
     {
-      "epoch": 0.6716376919062921,
+      "epoch": 0.693955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 5.14188822088761e-06,
-      "loss": 0.9017,
+      "learning_rate": 4.523352657090291e-06,
+      "loss": 0.7597,
       "step": 24455
     },
     {
-      "epoch": 0.6716651561341352,
+      "epoch": 0.6939841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 5.1411107463758256e-06,
-      "loss": 0.8282,
+      "learning_rate": 4.522583689552656e-06,
+      "loss": 0.8236,
       "step": 24456
     },
     {
-      "epoch": 0.6716926203619785,
+      "epoch": 0.6940124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 5.14033331030946e-06,
-      "loss": 0.8139,
+      "learning_rate": 4.521814768283e-06,
+      "loss": 0.8132,
       "step": 24457
     },
     {
-      "epoch": 0.6717200845898218,
+      "epoch": 0.6940408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 5.139555912694655e-06,
-      "loss": 0.9114,
+      "learning_rate": 4.521045893287821e-06,
+      "loss": 0.8601,
       "step": 24458
     },
     {
-      "epoch": 0.671747548817665,
+      "epoch": 0.6940692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 5.1387785535375714e-06,
-      "loss": 0.8024,
+      "learning_rate": 4.520277064573617e-06,
+      "loss": 0.8408,
       "step": 24459
     },
     {
-      "epoch": 0.6717750130455082,
+      "epoch": 0.6940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 5.1380012328443495e-06,
-      "loss": 0.8593,
+      "learning_rate": 4.519508282146876e-06,
+      "loss": 0.9135,
       "step": 24460
     },
     {
-      "epoch": 0.6718024772733514,
+      "epoch": 0.6941259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 5.137223950621145e-06,
-      "loss": 0.8542,
+      "learning_rate": 4.518739546014095e-06,
+      "loss": 0.7995,
       "step": 24461
     },
     {
-      "epoch": 0.6718299415011947,
+      "epoch": 0.6941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 5.1364467068741095e-06,
-      "loss": 0.9117,
+      "learning_rate": 4.517970856181771e-06,
+      "loss": 0.7827,
       "step": 24462
     },
     {
-      "epoch": 0.671857405729038,
+      "epoch": 0.6941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 5.135669501609395e-06,
-      "loss": 0.8078,
+      "learning_rate": 4.517202212656385e-06,
+      "loss": 0.8127,
       "step": 24463
     },
     {
-      "epoch": 0.6718848699568811,
+      "epoch": 0.694211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 5.134892334833143e-06,
-      "loss": 0.8714,
+      "learning_rate": 4.516433615444447e-06,
+      "loss": 0.7937,
       "step": 24464
     },
     {
-      "epoch": 0.6719123341847244,
+      "epoch": 0.6942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 5.134115206551509e-06,
-      "loss": 0.847,
+      "learning_rate": 4.515665064552437e-06,
+      "loss": 0.9203,
       "step": 24465
     },
     {
-      "epoch": 0.6719397984125677,
+      "epoch": 0.6942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 5.133338116770643e-06,
-      "loss": 0.76,
+      "learning_rate": 4.51489655998685e-06,
+      "loss": 0.9509,
       "step": 24466
     },
     {
-      "epoch": 0.6719672626404108,
+      "epoch": 0.6942962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 5.132561065496688e-06,
-      "loss": 0.8283,
+      "learning_rate": 4.514128101754183e-06,
+      "loss": 0.8124,
       "step": 24467
     },
     {
-      "epoch": 0.6719947268682541,
+      "epoch": 0.6943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 5.1317840527357945e-06,
-      "loss": 0.9177,
+      "learning_rate": 4.513359689860918e-06,
+      "loss": 0.8089,
       "step": 24468
     },
     {
-      "epoch": 0.6720221910960973,
+      "epoch": 0.6943530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 5.131007078494114e-06,
-      "loss": 0.868,
+      "learning_rate": 4.51259132431355e-06,
+      "loss": 0.8096,
       "step": 24469
     },
     {
-      "epoch": 0.6720496553239406,
+      "epoch": 0.6943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 5.1302301427777904e-06,
-      "loss": 0.8197,
+      "learning_rate": 4.511823005118574e-06,
+      "loss": 0.8342,
       "step": 24470
     },
     {
-      "epoch": 0.6720771195517838,
+      "epoch": 0.6944097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 5.12945324559297e-06,
-      "loss": 0.8959,
+      "learning_rate": 4.511054732282472e-06,
+      "loss": 0.6955,
       "step": 24471
     },
     {
-      "epoch": 0.672104583779627,
+      "epoch": 0.6944381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 5.128676386945805e-06,
-      "loss": 0.8011,
+      "learning_rate": 4.510286505811738e-06,
+      "loss": 0.9353,
       "step": 24472
     },
     {
-      "epoch": 0.6721320480074703,
+      "epoch": 0.694466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 5.127899566842437e-06,
-      "loss": 1.0026,
+      "learning_rate": 4.509518325712865e-06,
+      "loss": 0.874,
       "step": 24473
     },
     {
-      "epoch": 0.6721595122353134,
+      "epoch": 0.6944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 5.127122785289019e-06,
-      "loss": 0.899,
+      "learning_rate": 4.508750191992332e-06,
+      "loss": 0.88,
       "step": 24474
     },
     {
-      "epoch": 0.6721869764631567,
+      "epoch": 0.6945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 5.126346042291695e-06,
-      "loss": 0.8789,
+      "learning_rate": 4.5079821046566345e-06,
+      "loss": 0.816,
       "step": 24475
     },
     {
-      "epoch": 0.672214440691,
+      "epoch": 0.6945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.125569337856602e-06,
-      "loss": 0.8031,
+      "learning_rate": 4.507214063712262e-06,
+      "loss": 0.8455,
       "step": 24476
     },
     {
-      "epoch": 0.6722419049188432,
+      "epoch": 0.6945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 5.124792671989896e-06,
-      "loss": 0.8304,
+      "learning_rate": 4.506446069165696e-06,
+      "loss": 0.8824,
       "step": 24477
     },
     {
-      "epoch": 0.6722693691466864,
+      "epoch": 0.6946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 5.124016044697716e-06,
-      "loss": 0.8437,
+      "learning_rate": 4.505678121023426e-06,
+      "loss": 0.6996,
       "step": 24478
     },
     {
-      "epoch": 0.6722968333745297,
+      "epoch": 0.6946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 5.123239455986211e-06,
-      "loss": 0.8363,
+      "learning_rate": 4.504910219291941e-06,
+      "loss": 0.7949,
       "step": 24479
     },
     {
-      "epoch": 0.6723242976023729,
+      "epoch": 0.6946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 5.122462905861528e-06,
-      "loss": 0.8263,
+      "learning_rate": 4.5041423639777245e-06,
+      "loss": 0.861,
       "step": 24480
     },
     {
-      "epoch": 0.6723517618302162,
+      "epoch": 0.6946935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 5.121686394329804e-06,
-      "loss": 0.8668,
+      "learning_rate": 4.50337455508727e-06,
+      "loss": 0.9327,
       "step": 24481
     },
     {
-      "epoch": 0.6723792260580593,
+      "epoch": 0.6947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 5.120909921397188e-06,
-      "loss": 0.8757,
+      "learning_rate": 4.502606792627053e-06,
+      "loss": 0.8269,
       "step": 24482
     },
     {
-      "epoch": 0.6724066902859026,
+      "epoch": 0.6947502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 5.120133487069825e-06,
-      "loss": 0.8738,
+      "learning_rate": 4.501839076603565e-06,
+      "loss": 0.9028,
       "step": 24483
     },
     {
-      "epoch": 0.6724341545137459,
+      "epoch": 0.6947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 5.119357091353854e-06,
-      "loss": 0.7991,
+      "learning_rate": 4.501071407023292e-06,
+      "loss": 0.8774,
       "step": 24484
     },
     {
-      "epoch": 0.672461618741589,
+      "epoch": 0.6948070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 5.118580734255419e-06,
-      "loss": 0.8748,
+      "learning_rate": 4.500303783892712e-06,
+      "loss": 0.8599,
       "step": 24485
     },
     {
-      "epoch": 0.6724890829694323,
+      "epoch": 0.6948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 5.117804415780667e-06,
-      "loss": 0.8165,
+      "learning_rate": 4.499536207218315e-06,
+      "loss": 0.8085,
       "step": 24486
     },
     {
-      "epoch": 0.6725165471972755,
+      "epoch": 0.6948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 5.1170281359357355e-06,
-      "loss": 0.8962,
+      "learning_rate": 4.4987686770065855e-06,
+      "loss": 0.8689,
       "step": 24487
     },
     {
-      "epoch": 0.6725440114251188,
+      "epoch": 0.6948921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 5.116251894726769e-06,
-      "loss": 0.8573,
+      "learning_rate": 4.498001193264e-06,
+      "loss": 0.797,
       "step": 24488
     },
     {
-      "epoch": 0.672571475652962,
+      "epoch": 0.6949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 5.115475692159908e-06,
-      "loss": 0.8647,
+      "learning_rate": 4.4972337559970455e-06,
+      "loss": 0.7678,
       "step": 24489
     },
     {
-      "epoch": 0.6725989398808052,
+      "epoch": 0.6949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 5.1146995282413005e-06,
-      "loss": 0.9282,
+      "learning_rate": 4.49646636521221e-06,
+      "loss": 0.8046,
       "step": 24490
     },
     {
-      "epoch": 0.6726264041086485,
+      "epoch": 0.694977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 5.113923402977079e-06,
-      "loss": 0.865,
+      "learning_rate": 4.495699020915966e-06,
+      "loss": 0.8966,
       "step": 24491
     },
     {
-      "epoch": 0.6726538683364918,
+      "epoch": 0.695005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 5.11314731637339e-06,
-      "loss": 0.8139,
+      "learning_rate": 4.494931723114799e-06,
+      "loss": 0.8445,
       "step": 24492
     },
     {
-      "epoch": 0.6726813325643349,
+      "epoch": 0.6950340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 5.112371268436371e-06,
-      "loss": 0.7504,
+      "learning_rate": 4.4941644718151935e-06,
+      "loss": 0.7987,
       "step": 24493
     },
     {
-      "epoch": 0.6727087967921782,
+      "epoch": 0.6950624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 5.111595259172162e-06,
-      "loss": 0.8654,
+      "learning_rate": 4.493397267023626e-06,
+      "loss": 0.8957,
       "step": 24494
     },
     {
-      "epoch": 0.6727362610200214,
+      "epoch": 0.6950908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 5.110819288586906e-06,
-      "loss": 0.8959,
+      "learning_rate": 4.49263010874658e-06,
+      "loss": 0.8783,
       "step": 24495
     },
     {
-      "epoch": 0.6727637252478647,
+      "epoch": 0.6951191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 5.110043356686742e-06,
-      "loss": 0.8343,
+      "learning_rate": 4.491862996990539e-06,
+      "loss": 0.8494,
       "step": 24496
     },
     {
-      "epoch": 0.6727911894757079,
+      "epoch": 0.6951475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 5.109267463477807e-06,
-      "loss": 0.9157,
+      "learning_rate": 4.491095931761975e-06,
+      "loss": 0.7955,
       "step": 24497
     },
     {
-      "epoch": 0.6728186537035511,
+      "epoch": 0.6951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 5.108491608966248e-06,
-      "loss": 0.7955,
+      "learning_rate": 4.490328913067376e-06,
+      "loss": 0.8374,
       "step": 24498
     },
     {
-      "epoch": 0.6728461179313944,
+      "epoch": 0.6952043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 5.107715793158199e-06,
-      "loss": 0.8985,
+      "learning_rate": 4.4895619409132105e-06,
+      "loss": 0.7818,
       "step": 24499
     },
     {
-      "epoch": 0.6728735821592375,
+      "epoch": 0.6952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 5.106940016059792e-06,
-      "loss": 0.8563,
+      "learning_rate": 4.488795015305965e-06,
+      "loss": 0.8346,
       "step": 24500
     },
     {
-      "epoch": 0.6729010463870808,
+      "epoch": 0.695261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 5.106164277677173e-06,
-      "loss": 0.8352,
+      "learning_rate": 4.488028136252118e-06,
+      "loss": 0.8715,
       "step": 24501
     },
     {
-      "epoch": 0.6729285106149241,
+      "epoch": 0.6952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.105388578016476e-06,
-      "loss": 0.7979,
+      "learning_rate": 4.487261303758143e-06,
+      "loss": 0.803,
       "step": 24502
     },
     {
-      "epoch": 0.6729559748427673,
+      "epoch": 0.6953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 5.104612917083841e-06,
-      "loss": 0.8015,
+      "learning_rate": 4.486494517830519e-06,
+      "loss": 0.8199,
       "step": 24503
     },
     {
-      "epoch": 0.6729834390706105,
+      "epoch": 0.6953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 5.103837294885409e-06,
-      "loss": 0.9677,
+      "learning_rate": 4.485727778475726e-06,
+      "loss": 0.7972,
       "step": 24504
     },
     {
-      "epoch": 0.6730109032984538,
+      "epoch": 0.6953745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 5.103061711427309e-06,
-      "loss": 0.8125,
+      "learning_rate": 4.484961085700236e-06,
+      "loss": 0.926,
       "step": 24505
     },
     {
-      "epoch": 0.673038367526297,
+      "epoch": 0.6954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 5.102286166715682e-06,
-      "loss": 0.9082,
+      "learning_rate": 4.484194439510527e-06,
+      "loss": 0.8151,
       "step": 24506
     },
     {
-      "epoch": 0.6730658317541403,
+      "epoch": 0.6954313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 5.101510660756667e-06,
-      "loss": 0.834,
+      "learning_rate": 4.4834278399130795e-06,
+      "loss": 0.8493,
       "step": 24507
     },
     {
-      "epoch": 0.6730932959819834,
+      "epoch": 0.6954597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 5.100735193556394e-06,
-      "loss": 0.8426,
+      "learning_rate": 4.482661286914362e-06,
+      "loss": 0.8485,
       "step": 24508
     },
     {
-      "epoch": 0.6731207602098267,
+      "epoch": 0.6954880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 5.099959765121001e-06,
-      "loss": 0.7598,
+      "learning_rate": 4.481894780520848e-06,
+      "loss": 0.8624,
       "step": 24509
     },
     {
-      "epoch": 0.67314822443767,
+      "epoch": 0.695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 5.0991843754566295e-06,
-      "loss": 0.8264,
+      "learning_rate": 4.481128320739024e-06,
+      "loss": 0.8226,
       "step": 24510
     },
     {
-      "epoch": 0.6731756886655131,
+      "epoch": 0.6955448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 5.098409024569405e-06,
-      "loss": 0.9261,
+      "learning_rate": 4.480361907575354e-06,
+      "loss": 0.8853,
       "step": 24511
     },
     {
-      "epoch": 0.6732031528933564,
+      "epoch": 0.6955732122587969,
       "grad_norm": 0.0,
-      "learning_rate": 5.097633712465469e-06,
-      "loss": 0.844,
+      "learning_rate": 4.479595541036316e-06,
+      "loss": 0.9114,
       "step": 24512
     },
     {
-      "epoch": 0.6732306171211997,
+      "epoch": 0.6956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 5.096858439150951e-06,
-      "loss": 0.8994,
+      "learning_rate": 4.478829221128384e-06,
+      "loss": 0.8302,
       "step": 24513
     },
     {
-      "epoch": 0.6732580813490429,
+      "epoch": 0.6956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 5.096083204631994e-06,
-      "loss": 0.937,
+      "learning_rate": 4.478062947858028e-06,
+      "loss": 0.918,
       "step": 24514
     },
     {
-      "epoch": 0.6732855455768861,
+      "epoch": 0.6956583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 5.0953080089147215e-06,
-      "loss": 0.8148,
+      "learning_rate": 4.477296721231722e-06,
+      "loss": 0.8701,
       "step": 24515
     },
     {
-      "epoch": 0.6733130098047293,
+      "epoch": 0.6956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 5.0945328520052764e-06,
-      "loss": 0.9165,
+      "learning_rate": 4.476530541255942e-06,
+      "loss": 0.7974,
       "step": 24516
     },
     {
-      "epoch": 0.6733404740325726,
+      "epoch": 0.6957150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 5.093757733909783e-06,
-      "loss": 0.9366,
+      "learning_rate": 4.4757644079371544e-06,
+      "loss": 0.8703,
       "step": 24517
     },
     {
-      "epoch": 0.6733679382604159,
+      "epoch": 0.6957434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 5.0929826546343796e-06,
-      "loss": 0.763,
+      "learning_rate": 4.474998321281833e-06,
+      "loss": 0.956,
       "step": 24518
     },
     {
-      "epoch": 0.673395402488259,
+      "epoch": 0.6957718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 5.092207614185199e-06,
-      "loss": 0.9418,
+      "learning_rate": 4.474232281296451e-06,
+      "loss": 0.7368,
       "step": 24519
     },
     {
-      "epoch": 0.6734228667161023,
+      "epoch": 0.695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 5.091432612568372e-06,
-      "loss": 0.8318,
+      "learning_rate": 4.473466287987476e-06,
+      "loss": 0.7393,
       "step": 24520
     },
     {
-      "epoch": 0.6734503309439455,
+      "epoch": 0.6958286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 5.090657649790032e-06,
-      "loss": 0.9146,
+      "learning_rate": 4.472700341361378e-06,
+      "loss": 0.8727,
       "step": 24521
     },
     {
-      "epoch": 0.6734777951717887,
+      "epoch": 0.6958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 5.089882725856312e-06,
-      "loss": 0.8343,
+      "learning_rate": 4.471934441424634e-06,
+      "loss": 0.8548,
       "step": 24522
     },
     {
-      "epoch": 0.673505259399632,
+      "epoch": 0.6958853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 5.089107840773344e-06,
-      "loss": 0.7428,
+      "learning_rate": 4.471168588183703e-06,
+      "loss": 0.8346,
       "step": 24523
     },
     {
-      "epoch": 0.6735327236274752,
+      "epoch": 0.6959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 5.088332994547254e-06,
-      "loss": 0.8002,
+      "learning_rate": 4.470402781645059e-06,
+      "loss": 0.8155,
       "step": 24524
     },
     {
-      "epoch": 0.6735601878553185,
+      "epoch": 0.6959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 5.087558187184174e-06,
-      "loss": 0.7091,
+      "learning_rate": 4.469637021815173e-06,
+      "loss": 0.8916,
       "step": 24525
     },
     {
-      "epoch": 0.6735876520831617,
+      "epoch": 0.6959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 5.086783418690238e-06,
-      "loss": 0.8169,
+      "learning_rate": 4.468871308700511e-06,
+      "loss": 0.7562,
       "step": 24526
     },
     {
-      "epoch": 0.6736151163110049,
+      "epoch": 0.6959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 5.086008689071573e-06,
-      "loss": 0.7659,
+      "learning_rate": 4.468105642307545e-06,
+      "loss": 0.823,
       "step": 24527
     },
     {
-      "epoch": 0.6736425805388482,
+      "epoch": 0.6960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 5.085233998334315e-06,
-      "loss": 0.9119,
+      "learning_rate": 4.467340022642736e-06,
+      "loss": 0.8558,
       "step": 24528
     },
     {
-      "epoch": 0.6736700447666913,
+      "epoch": 0.69605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.0844593464845874e-06,
-      "loss": 0.943,
+      "learning_rate": 4.466574449712555e-06,
+      "loss": 0.9101,
       "step": 24529
     },
     {
-      "epoch": 0.6736975089945346,
+      "epoch": 0.6960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 5.08368473352852e-06,
-      "loss": 0.8655,
+      "learning_rate": 4.465808923523471e-06,
+      "loss": 0.9674,
       "step": 24530
     },
     {
-      "epoch": 0.6737249732223779,
+      "epoch": 0.6961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 5.082910159472247e-06,
-      "loss": 0.8491,
+      "learning_rate": 4.465043444081945e-06,
+      "loss": 0.8639,
       "step": 24531
     },
     {
-      "epoch": 0.6737524374502211,
+      "epoch": 0.6961407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 5.082135624321891e-06,
-      "loss": 0.8105,
+      "learning_rate": 4.464278011394445e-06,
+      "loss": 0.7336,
       "step": 24532
     },
     {
-      "epoch": 0.6737799016780643,
+      "epoch": 0.6961691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 5.0813611280835824e-06,
-      "loss": 0.7932,
+      "learning_rate": 4.463512625467442e-06,
+      "loss": 0.7607,
       "step": 24533
     },
     {
-      "epoch": 0.6738073659059075,
+      "epoch": 0.6961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 5.080586670763455e-06,
-      "loss": 0.7964,
+      "learning_rate": 4.462747286307393e-06,
+      "loss": 0.8563,
       "step": 24534
     },
     {
-      "epoch": 0.6738348301337508,
+      "epoch": 0.6962258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 5.079812252367625e-06,
-      "loss": 0.8158,
+      "learning_rate": 4.461981993920766e-06,
+      "loss": 0.9151,
       "step": 24535
     },
     {
-      "epoch": 0.6738622943615941,
+      "epoch": 0.6962542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 5.079037872902228e-06,
-      "loss": 0.8162,
+      "learning_rate": 4.46121674831403e-06,
+      "loss": 0.8999,
       "step": 24536
     },
     {
-      "epoch": 0.6738897585894372,
+      "epoch": 0.6962826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 5.078263532373389e-06,
-      "loss": 0.8843,
+      "learning_rate": 4.460451549493642e-06,
+      "loss": 0.8165,
       "step": 24537
     },
     {
-      "epoch": 0.6739172228172805,
+      "epoch": 0.696311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 5.077489230787238e-06,
-      "loss": 0.9171,
+      "learning_rate": 4.459686397466068e-06,
+      "loss": 0.7058,
       "step": 24538
     },
     {
-      "epoch": 0.6739446870451238,
+      "epoch": 0.696339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 5.076714968149897e-06,
-      "loss": 0.8783,
+      "learning_rate": 4.458921292237773e-06,
+      "loss": 0.8294,
       "step": 24539
     },
     {
-      "epoch": 0.673972151272967,
+      "epoch": 0.6963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 5.075940744467498e-06,
-      "loss": 0.8686,
+      "learning_rate": 4.458156233815218e-06,
+      "loss": 0.7716,
       "step": 24540
     },
     {
-      "epoch": 0.6739996155008102,
+      "epoch": 0.6963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 5.075166559746159e-06,
-      "loss": 0.7974,
+      "learning_rate": 4.45739122220487e-06,
+      "loss": 0.9002,
       "step": 24541
     },
     {
-      "epoch": 0.6740270797286534,
+      "epoch": 0.6964245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 5.07439241399201e-06,
-      "loss": 0.7906,
+      "learning_rate": 4.456626257413185e-06,
+      "loss": 0.784,
       "step": 24542
     },
     {
-      "epoch": 0.6740545439564967,
+      "epoch": 0.6964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 5.073618307211177e-06,
-      "loss": 0.843,
+      "learning_rate": 4.4558613394466265e-06,
+      "loss": 0.6948,
       "step": 24543
     },
     {
-      "epoch": 0.6740820081843399,
+      "epoch": 0.6964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 5.072844239409782e-06,
-      "loss": 0.9019,
+      "learning_rate": 4.455096468311659e-06,
+      "loss": 0.7554,
       "step": 24544
     },
     {
-      "epoch": 0.6741094724121831,
+      "epoch": 0.6965096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 5.0720702105939575e-06,
-      "loss": 0.7714,
+      "learning_rate": 4.454331644014739e-06,
+      "loss": 0.7063,
       "step": 24545
     },
     {
-      "epoch": 0.6741369366400264,
+      "epoch": 0.6965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 5.071296220769818e-06,
-      "loss": 0.8745,
+      "learning_rate": 4.453566866562328e-06,
+      "loss": 0.8858,
       "step": 24546
     },
     {
-      "epoch": 0.6741644008678696,
+      "epoch": 0.6965664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 5.070522269943497e-06,
-      "loss": 0.8311,
+      "learning_rate": 4.452802135960891e-06,
+      "loss": 0.768,
       "step": 24547
     },
     {
-      "epoch": 0.6741918650957128,
+      "epoch": 0.6965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 5.069748358121107e-06,
-      "loss": 0.8527,
+      "learning_rate": 4.45203745221688e-06,
+      "loss": 0.7877,
       "step": 24548
     },
     {
-      "epoch": 0.6742193293235561,
+      "epoch": 0.6966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 5.06897448530878e-06,
-      "loss": 0.7925,
+      "learning_rate": 4.451272815336758e-06,
+      "loss": 0.876,
       "step": 24549
     },
     {
-      "epoch": 0.6742467935513993,
+      "epoch": 0.6966515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 5.068200651512636e-06,
-      "loss": 0.9152,
+      "learning_rate": 4.450508225326987e-06,
+      "loss": 0.8973,
       "step": 24550
     },
     {
-      "epoch": 0.6742742577792425,
+      "epoch": 0.6966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 5.067426856738804e-06,
-      "loss": 0.8497,
+      "learning_rate": 4.4497436821940195e-06,
+      "loss": 0.7611,
       "step": 24551
     },
     {
-      "epoch": 0.6743017220070858,
+      "epoch": 0.6967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 5.066653100993395e-06,
-      "loss": 0.7971,
+      "learning_rate": 4.448979185944317e-06,
+      "loss": 0.7723,
       "step": 24552
     },
     {
-      "epoch": 0.674329186234929,
+      "epoch": 0.6967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 5.0658793842825395e-06,
-      "loss": 0.9801,
+      "learning_rate": 4.448214736584339e-06,
+      "loss": 0.7502,
       "step": 24553
     },
     {
-      "epoch": 0.6743566504627723,
+      "epoch": 0.6967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 5.0651057066123576e-06,
-      "loss": 0.859,
+      "learning_rate": 4.4474503341205385e-06,
+      "loss": 0.9936,
       "step": 24554
     },
     {
-      "epoch": 0.6743841146906154,
+      "epoch": 0.6967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 5.064332067988975e-06,
-      "loss": 0.8584,
+      "learning_rate": 4.446685978559375e-06,
+      "loss": 0.8228,
       "step": 24555
     },
     {
-      "epoch": 0.6744115789184587,
+      "epoch": 0.696821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 5.0635584684185035e-06,
-      "loss": 0.8426,
+      "learning_rate": 4.445921669907303e-06,
+      "loss": 0.8339,
       "step": 24556
     },
     {
-      "epoch": 0.674439043146302,
+      "epoch": 0.696850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 5.0627849079070745e-06,
-      "loss": 0.9145,
+      "learning_rate": 4.445157408170781e-06,
+      "loss": 0.8812,
       "step": 24557
     },
     {
-      "epoch": 0.6744665073741452,
+      "epoch": 0.6968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 5.062011386460801e-06,
-      "loss": 0.7873,
+      "learning_rate": 4.444393193356269e-06,
+      "loss": 0.8986,
       "step": 24558
     },
     {
-      "epoch": 0.6744939716019884,
+      "epoch": 0.6969069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 5.0612379040858055e-06,
-      "loss": 0.8546,
+      "learning_rate": 4.443629025470213e-06,
+      "loss": 0.8216,
       "step": 24559
     },
     {
-      "epoch": 0.6745214358298316,
+      "epoch": 0.6969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 5.060464460788208e-06,
-      "loss": 0.8654,
+      "learning_rate": 4.442864904519072e-06,
+      "loss": 0.8317,
       "step": 24560
     },
     {
-      "epoch": 0.6745489000576749,
+      "epoch": 0.6969636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 5.05969105657413e-06,
-      "loss": 0.7362,
+      "learning_rate": 4.442100830509305e-06,
+      "loss": 0.9721,
       "step": 24561
     },
     {
-      "epoch": 0.6745763642855181,
+      "epoch": 0.6969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 5.058917691449693e-06,
-      "loss": 0.8402,
+      "learning_rate": 4.441336803447358e-06,
+      "loss": 0.8317,
       "step": 24562
     },
     {
-      "epoch": 0.6746038285133613,
+      "epoch": 0.6970204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 5.058144365421014e-06,
-      "loss": 0.8147,
+      "learning_rate": 4.440572823339689e-06,
+      "loss": 0.9218,
       "step": 24563
     },
     {
-      "epoch": 0.6746312927412046,
+      "epoch": 0.6970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 5.057371078494207e-06,
-      "loss": 0.8332,
+      "learning_rate": 4.439808890192755e-06,
+      "loss": 0.8501,
       "step": 24564
     },
     {
-      "epoch": 0.6746587569690479,
+      "epoch": 0.6970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 5.0565978306753935e-06,
-      "loss": 0.7993,
+      "learning_rate": 4.4390450040130015e-06,
+      "loss": 0.757,
       "step": 24565
     },
     {
-      "epoch": 0.674686221196891,
+      "epoch": 0.697105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.055824621970695e-06,
-      "loss": 0.8327,
+      "learning_rate": 4.438281164806885e-06,
+      "loss": 0.8777,
       "step": 24566
     },
     {
-      "epoch": 0.6747136854247343,
+      "epoch": 0.6971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 5.055051452386226e-06,
-      "loss": 0.8653,
+      "learning_rate": 4.4375173725808615e-06,
+      "loss": 0.8314,
       "step": 24567
     },
     {
-      "epoch": 0.6747411496525775,
+      "epoch": 0.6971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 5.0542783219281055e-06,
-      "loss": 0.8783,
+      "learning_rate": 4.436753627341375e-06,
+      "loss": 0.8046,
       "step": 24568
     },
     {
-      "epoch": 0.6747686138804208,
+      "epoch": 0.6971906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 5.053505230602454e-06,
-      "loss": 0.8083,
+      "learning_rate": 4.43598992909488e-06,
+      "loss": 0.6321,
       "step": 24569
     },
     {
-      "epoch": 0.674796078108264,
+      "epoch": 0.6972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 5.052732178415383e-06,
-      "loss": 0.9282,
+      "learning_rate": 4.435226277847828e-06,
+      "loss": 0.7631,
       "step": 24570
     },
     {
-      "epoch": 0.6748235423361072,
+      "epoch": 0.6972474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 5.051959165373014e-06,
-      "loss": 0.8156,
+      "learning_rate": 4.434462673606671e-06,
+      "loss": 0.9323,
       "step": 24571
     },
     {
-      "epoch": 0.6748510065639505,
+      "epoch": 0.6972758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 5.0511861914814565e-06,
-      "loss": 0.8926,
+      "learning_rate": 4.433699116377861e-06,
+      "loss": 0.8151,
       "step": 24572
     },
     {
-      "epoch": 0.6748784707917936,
+      "epoch": 0.6973041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 5.050413256746831e-06,
-      "loss": 0.8548,
+      "learning_rate": 4.432935606167842e-06,
+      "loss": 0.8422,
       "step": 24573
     },
     {
-      "epoch": 0.6749059350196369,
+      "epoch": 0.6973325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 5.049640361175255e-06,
-      "loss": 0.9257,
+      "learning_rate": 4.432172142983065e-06,
+      "loss": 0.8207,
       "step": 24574
     },
     {
-      "epoch": 0.6749333992474802,
+      "epoch": 0.697360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 5.048867504772843e-06,
-      "loss": 0.7777,
+      "learning_rate": 4.431408726829984e-06,
+      "loss": 0.9038,
       "step": 24575
     },
     {
-      "epoch": 0.6749608634753234,
+      "epoch": 0.6973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 5.0480946875457085e-06,
-      "loss": 0.8007,
+      "learning_rate": 4.430645357715041e-06,
+      "loss": 0.9134,
       "step": 24576
     },
     {
-      "epoch": 0.6749883277031666,
+      "epoch": 0.6974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 5.047321909499965e-06,
-      "loss": 0.8013,
+      "learning_rate": 4.429882035644686e-06,
+      "loss": 0.8374,
       "step": 24577
     },
     {
-      "epoch": 0.6750157919310099,
+      "epoch": 0.6974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.046549170641728e-06,
-      "loss": 0.7706,
+      "learning_rate": 4.429118760625373e-06,
+      "loss": 0.8154,
       "step": 24578
     },
     {
-      "epoch": 0.6750432561588531,
+      "epoch": 0.6974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 5.045776470977118e-06,
-      "loss": 0.9243,
+      "learning_rate": 4.428355532663538e-06,
+      "loss": 0.8968,
       "step": 24579
     },
     {
-      "epoch": 0.6750707203866964,
+      "epoch": 0.6975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 5.0450038105122415e-06,
-      "loss": 0.898,
+      "learning_rate": 4.427592351765637e-06,
+      "loss": 0.8948,
       "step": 24580
     },
     {
-      "epoch": 0.6750981846145395,
+      "epoch": 0.6975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 5.044231189253215e-06,
-      "loss": 0.7516,
+      "learning_rate": 4.426829217938117e-06,
+      "loss": 0.6896,
       "step": 24581
     },
     {
-      "epoch": 0.6751256488423828,
+      "epoch": 0.6975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 5.043458607206149e-06,
-      "loss": 0.9075,
+      "learning_rate": 4.426066131187416e-06,
+      "loss": 0.7888,
       "step": 24582
     },
     {
-      "epoch": 0.6751531130702261,
+      "epoch": 0.6975879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 5.042686064377158e-06,
-      "loss": 0.8383,
+      "learning_rate": 4.425303091519986e-06,
+      "loss": 0.9174,
       "step": 24583
     },
     {
-      "epoch": 0.6751805772980692,
+      "epoch": 0.6976163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 5.041913560772354e-06,
-      "loss": 0.7563,
+      "learning_rate": 4.424540098942275e-06,
+      "loss": 0.8548,
       "step": 24584
     },
     {
-      "epoch": 0.6752080415259125,
+      "epoch": 0.697644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 5.041141096397852e-06,
-      "loss": 0.9205,
+      "learning_rate": 4.423777153460717e-06,
+      "loss": 0.8335,
       "step": 24585
     },
     {
-      "epoch": 0.6752355057537557,
+      "epoch": 0.6976730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 5.040368671259767e-06,
-      "loss": 0.8139,
+      "learning_rate": 4.423014255081771e-06,
+      "loss": 0.7575,
       "step": 24586
     },
     {
-      "epoch": 0.675262969981599,
+      "epoch": 0.6977014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 5.0395962853642055e-06,
-      "loss": 0.9039,
+      "learning_rate": 4.422251403811872e-06,
+      "loss": 0.9205,
       "step": 24587
     },
     {
-      "epoch": 0.6752904342094422,
+      "epoch": 0.6977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 5.0388239387172744e-06,
-      "loss": 0.8435,
+      "learning_rate": 4.421488599657464e-06,
+      "loss": 0.7898,
       "step": 24588
     },
     {
-      "epoch": 0.6753178984372854,
+      "epoch": 0.6977582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 5.038051631325092e-06,
-      "loss": 0.7648,
+      "learning_rate": 4.420725842624998e-06,
+      "loss": 0.896,
       "step": 24589
     },
     {
-      "epoch": 0.6753453626651287,
+      "epoch": 0.6977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 5.037279363193765e-06,
-      "loss": 0.8605,
+      "learning_rate": 4.419963132720907e-06,
+      "loss": 0.8784,
       "step": 24590
     },
     {
-      "epoch": 0.675372826892972,
+      "epoch": 0.6978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 5.036507134329409e-06,
-      "loss": 0.8455,
+      "learning_rate": 4.419200469951639e-06,
+      "loss": 0.7543,
       "step": 24591
     },
     {
-      "epoch": 0.6754002911208151,
+      "epoch": 0.6978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 5.035734944738129e-06,
-      "loss": 0.8914,
+      "learning_rate": 4.418437854323638e-06,
+      "loss": 0.8211,
       "step": 24592
     },
     {
-      "epoch": 0.6754277553486584,
+      "epoch": 0.697871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 5.034962794426042e-06,
-      "loss": 0.7577,
+      "learning_rate": 4.41767528584334e-06,
+      "loss": 0.8828,
       "step": 24593
     },
     {
-      "epoch": 0.6754552195765016,
+      "epoch": 0.697900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 5.034190683399248e-06,
-      "loss": 0.8463,
+      "learning_rate": 4.416912764517191e-06,
+      "loss": 0.9041,
       "step": 24594
     },
     {
-      "epoch": 0.6754826838043448,
+      "epoch": 0.6979284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 5.033418611663865e-06,
-      "loss": 0.7812,
+      "learning_rate": 4.416150290351634e-06,
+      "loss": 0.7702,
       "step": 24595
     },
     {
-      "epoch": 0.6755101480321881,
+      "epoch": 0.6979568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 5.032646579225995e-06,
-      "loss": 0.9105,
+      "learning_rate": 4.4153878633531024e-06,
+      "loss": 0.7717,
       "step": 24596
     },
     {
-      "epoch": 0.6755376122600313,
+      "epoch": 0.6979852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 5.031874586091747e-06,
-      "loss": 0.8366,
+      "learning_rate": 4.4146254835280415e-06,
+      "loss": 0.8586,
       "step": 24597
     },
     {
-      "epoch": 0.6755650764878746,
+      "epoch": 0.6980136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 5.031102632267234e-06,
-      "loss": 0.9521,
+      "learning_rate": 4.413863150882892e-06,
+      "loss": 0.7728,
       "step": 24598
     },
     {
-      "epoch": 0.6755925407157177,
+      "epoch": 0.6980419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 5.030330717758565e-06,
-      "loss": 0.8193,
+      "learning_rate": 4.413100865424089e-06,
+      "loss": 0.912,
       "step": 24599
     },
     {
-      "epoch": 0.675620004943561,
+      "epoch": 0.6980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 5.02955884257184e-06,
-      "loss": 0.7958,
+      "learning_rate": 4.412338627158075e-06,
+      "loss": 0.728,
       "step": 24600
     },
     {
-      "epoch": 0.6756474691714043,
+      "epoch": 0.6980987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 5.028787006713172e-06,
-      "loss": 0.8604,
+      "learning_rate": 4.411576436091286e-06,
+      "loss": 0.7742,
       "step": 24601
     },
     {
-      "epoch": 0.6756749333992474,
+      "epoch": 0.6981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 5.0280152101886705e-06,
-      "loss": 0.8109,
+      "learning_rate": 4.410814292230163e-06,
+      "loss": 0.872,
       "step": 24602
     },
     {
-      "epoch": 0.6757023976270907,
+      "epoch": 0.698155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 5.027243453004433e-06,
-      "loss": 0.9269,
+      "learning_rate": 4.410052195581144e-06,
+      "loss": 0.8421,
       "step": 24603
     },
     {
-      "epoch": 0.675729861854934,
+      "epoch": 0.698183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 5.026471735166577e-06,
-      "loss": 0.9069,
+      "learning_rate": 4.409290146150668e-06,
+      "loss": 0.7837,
       "step": 24604
     },
     {
-      "epoch": 0.6757573260827772,
+      "epoch": 0.6982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 5.025700056681199e-06,
-      "loss": 0.8383,
+      "learning_rate": 4.4085281439451655e-06,
+      "loss": 0.816,
       "step": 24605
     },
     {
-      "epoch": 0.6757847903106204,
+      "epoch": 0.6982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 5.0249284175544085e-06,
-      "loss": 0.8496,
+      "learning_rate": 4.407766188971081e-06,
+      "loss": 0.8428,
       "step": 24606
     },
     {
-      "epoch": 0.6758122545384636,
+      "epoch": 0.6982690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 5.024156817792312e-06,
-      "loss": 0.8321,
+      "learning_rate": 4.407004281234843e-06,
+      "loss": 0.7712,
       "step": 24607
     },
     {
-      "epoch": 0.6758397187663069,
+      "epoch": 0.6982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 5.0233852574010144e-06,
-      "loss": 0.9036,
+      "learning_rate": 4.406242420742892e-06,
+      "loss": 0.8063,
       "step": 24608
     },
     {
-      "epoch": 0.6758671829941502,
+      "epoch": 0.6983257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 5.022613736386621e-06,
-      "loss": 0.8445,
+      "learning_rate": 4.405480607501666e-06,
+      "loss": 0.9367,
       "step": 24609
     },
     {
-      "epoch": 0.6758946472219933,
+      "epoch": 0.6983541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 5.021842254755238e-06,
-      "loss": 0.8056,
+      "learning_rate": 4.4047188415175935e-06,
+      "loss": 0.8784,
       "step": 24610
     },
     {
-      "epoch": 0.6759221114498366,
+      "epoch": 0.6983825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 5.02107081251297e-06,
-      "loss": 0.8657,
+      "learning_rate": 4.403957122797111e-06,
+      "loss": 0.8784,
       "step": 24611
     },
     {
-      "epoch": 0.6759495756776798,
+      "epoch": 0.698410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 5.020299409665912e-06,
-      "loss": 0.8648,
+      "learning_rate": 4.40319545134666e-06,
+      "loss": 0.8726,
       "step": 24612
     },
     {
-      "epoch": 0.675977039905523,
+      "epoch": 0.6984392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 5.019528046220175e-06,
-      "loss": 0.8192,
+      "learning_rate": 4.402433827172664e-06,
+      "loss": 0.8826,
       "step": 24613
     },
     {
-      "epoch": 0.6760045041333663,
+      "epoch": 0.6984676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 5.018756722181861e-06,
-      "loss": 0.9124,
+      "learning_rate": 4.401672250281561e-06,
+      "loss": 0.8644,
       "step": 24614
     },
     {
-      "epoch": 0.6760319683612095,
+      "epoch": 0.6984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 5.017985437557074e-06,
-      "loss": 0.9038,
+      "learning_rate": 4.4009107206797876e-06,
+      "loss": 0.9023,
       "step": 24615
     },
     {
-      "epoch": 0.6760594325890528,
+      "epoch": 0.6985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 5.017214192351918e-06,
-      "loss": 0.864,
+      "learning_rate": 4.4001492383737665e-06,
+      "loss": 0.9016,
       "step": 24616
     },
     {
-      "epoch": 0.676086896816896,
+      "epoch": 0.6985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 5.016442986572492e-06,
-      "loss": 0.8959,
+      "learning_rate": 4.399387803369939e-06,
+      "loss": 0.8348,
       "step": 24617
     },
     {
-      "epoch": 0.6761143610447392,
+      "epoch": 0.6985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 5.015671820224899e-06,
-      "loss": 0.8597,
+      "learning_rate": 4.398626415674739e-06,
+      "loss": 0.8391,
       "step": 24618
     },
     {
-      "epoch": 0.6761418252725825,
+      "epoch": 0.6986095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 5.0149006933152436e-06,
-      "loss": 0.8744,
+      "learning_rate": 4.39786507529459e-06,
+      "loss": 0.8717,
       "step": 24619
     },
     {
-      "epoch": 0.6761692895004257,
+      "epoch": 0.6986379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 5.014129605849623e-06,
-      "loss": 0.8386,
+      "learning_rate": 4.397103782235925e-06,
+      "loss": 0.9046,
       "step": 24620
     },
     {
-      "epoch": 0.6761967537282689,
+      "epoch": 0.6986662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 5.013358557834139e-06,
-      "loss": 0.8251,
+      "learning_rate": 4.396342536505181e-06,
+      "loss": 0.8119,
       "step": 24621
     },
     {
-      "epoch": 0.6762242179561122,
+      "epoch": 0.698694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 5.012587549274899e-06,
-      "loss": 0.9091,
+      "learning_rate": 4.395581338108781e-06,
+      "loss": 0.8915,
       "step": 24622
     },
     {
-      "epoch": 0.6762516821839554,
+      "epoch": 0.6987230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 5.0118165801779926e-06,
-      "loss": 0.8004,
+      "learning_rate": 4.394820187053156e-06,
+      "loss": 0.7803,
       "step": 24623
     },
     {
-      "epoch": 0.6762791464117986,
+      "epoch": 0.6987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 5.011045650549526e-06,
-      "loss": 0.8415,
+      "learning_rate": 4.394059083344741e-06,
+      "loss": 0.874,
       "step": 24624
     },
     {
-      "epoch": 0.6763066106396418,
+      "epoch": 0.6987797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 5.0102747603956e-06,
-      "loss": 0.7398,
+      "learning_rate": 4.393298026989957e-06,
+      "loss": 0.8001,
       "step": 24625
     },
     {
-      "epoch": 0.6763340748674851,
+      "epoch": 0.6988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 5.009503909722316e-06,
-      "loss": 0.7543,
+      "learning_rate": 4.392537017995236e-06,
+      "loss": 0.8395,
       "step": 24626
     },
     {
-      "epoch": 0.6763615390953284,
+      "epoch": 0.6988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 5.008733098535767e-06,
-      "loss": 0.9152,
+      "learning_rate": 4.391776056367012e-06,
+      "loss": 0.8581,
       "step": 24627
     },
     {
-      "epoch": 0.6763890033231715,
+      "epoch": 0.6988649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 5.007962326842058e-06,
-      "loss": 0.8186,
+      "learning_rate": 4.391015142111703e-06,
+      "loss": 0.6972,
       "step": 24628
     },
     {
-      "epoch": 0.6764164675510148,
+      "epoch": 0.6988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 5.007191594647283e-06,
-      "loss": 0.8895,
+      "learning_rate": 4.3902542752357415e-06,
+      "loss": 0.7326,
       "step": 24629
     },
     {
-      "epoch": 0.6764439317788581,
+      "epoch": 0.6989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 5.00642090195754e-06,
-      "loss": 0.9212,
+      "learning_rate": 4.389493455745558e-06,
+      "loss": 0.8339,
       "step": 24630
     },
     {
-      "epoch": 0.6764713960067013,
+      "epoch": 0.698950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.005650248778929e-06,
-      "loss": 0.8333,
+      "learning_rate": 4.388732683647569e-06,
+      "loss": 0.8596,
       "step": 24631
     },
     {
-      "epoch": 0.6764988602345445,
+      "epoch": 0.6989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 5.004879635117549e-06,
-      "loss": 0.8742,
+      "learning_rate": 4.387971958948213e-06,
+      "loss": 0.9109,
       "step": 24632
     },
     {
-      "epoch": 0.6765263244623877,
+      "epoch": 0.6990068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 5.0041090609794965e-06,
-      "loss": 0.8218,
+      "learning_rate": 4.387211281653907e-06,
+      "loss": 0.8448,
       "step": 24633
     },
     {
-      "epoch": 0.676553788690231,
+      "epoch": 0.6990351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 5.003338526370871e-06,
-      "loss": 0.8398,
+      "learning_rate": 4.3864506517710804e-06,
+      "loss": 0.8313,
       "step": 24634
     },
     {
-      "epoch": 0.6765812529180742,
+      "epoch": 0.6990635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 5.002568031297767e-06,
-      "loss": 0.8129,
+      "learning_rate": 4.38569006930616e-06,
+      "loss": 0.753,
       "step": 24635
     },
     {
-      "epoch": 0.6766087171459174,
+      "epoch": 0.6990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 5.001797575766276e-06,
-      "loss": 0.9021,
+      "learning_rate": 4.384929534265565e-06,
+      "loss": 0.7812,
       "step": 24636
     },
     {
-      "epoch": 0.6766361813737607,
+      "epoch": 0.6991203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 5.0010271597825e-06,
-      "loss": 0.799,
+      "learning_rate": 4.384169046655723e-06,
+      "loss": 0.8305,
       "step": 24637
     },
     {
-      "epoch": 0.6766636456016039,
+      "epoch": 0.6991486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 5.000256783352533e-06,
-      "loss": 0.9193,
+      "learning_rate": 4.383408606483061e-06,
+      "loss": 0.8398,
       "step": 24638
     },
     {
-      "epoch": 0.6766911098294471,
+      "epoch": 0.6991770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 4.999486446482471e-06,
-      "loss": 0.9016,
+      "learning_rate": 4.382648213753994e-06,
+      "loss": 0.8907,
       "step": 24639
     },
     {
-      "epoch": 0.6767185740572904,
+      "epoch": 0.699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 4.998716149178413e-06,
-      "loss": 0.7766,
+      "learning_rate": 4.381887868474951e-06,
+      "loss": 0.8146,
       "step": 24640
     },
     {
-      "epoch": 0.6767460382851336,
+      "epoch": 0.6992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 4.997945891446447e-06,
-      "loss": 0.8323,
+      "learning_rate": 4.381127570652358e-06,
+      "loss": 0.8505,
       "step": 24641
     },
     {
-      "epoch": 0.6767735025129769,
+      "epoch": 0.6992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 4.9971756732926706e-06,
-      "loss": 0.8232,
+      "learning_rate": 4.380367320292629e-06,
+      "loss": 0.8612,
       "step": 24642
     },
     {
-      "epoch": 0.6768009667408201,
+      "epoch": 0.6992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 4.9964054947231814e-06,
-      "loss": 0.8746,
+      "learning_rate": 4.379607117402189e-06,
+      "loss": 0.7893,
       "step": 24643
     },
     {
-      "epoch": 0.6768284309686633,
+      "epoch": 0.6993189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 4.9956353557440665e-06,
-      "loss": 0.7719,
+      "learning_rate": 4.378846961987465e-06,
+      "loss": 0.796,
       "step": 24644
     },
     {
-      "epoch": 0.6768558951965066,
+      "epoch": 0.6993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 4.9948652563614234e-06,
-      "loss": 0.879,
+      "learning_rate": 4.3780868540548695e-06,
+      "loss": 0.7549,
       "step": 24645
     },
     {
-      "epoch": 0.6768833594243497,
+      "epoch": 0.6993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 4.9940951965813485e-06,
-      "loss": 0.8649,
+      "learning_rate": 4.3773267936108265e-06,
+      "loss": 0.9508,
       "step": 24646
     },
     {
-      "epoch": 0.676910823652193,
+      "epoch": 0.6994040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 4.9933251764099275e-06,
-      "loss": 0.9479,
+      "learning_rate": 4.376566780661757e-06,
+      "loss": 0.7754,
       "step": 24647
     },
     {
-      "epoch": 0.6769382878800363,
+      "epoch": 0.6994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 4.992555195853258e-06,
-      "loss": 0.8048,
+      "learning_rate": 4.375806815214082e-06,
+      "loss": 0.8644,
       "step": 24648
     },
     {
-      "epoch": 0.6769657521078795,
+      "epoch": 0.6994608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 4.991785254917431e-06,
-      "loss": 0.8764,
+      "learning_rate": 4.3750468972742225e-06,
+      "loss": 0.7809,
       "step": 24649
     },
     {
-      "epoch": 0.6769932163357227,
+      "epoch": 0.6994892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 4.991015353608543e-06,
-      "loss": 0.7511,
+      "learning_rate": 4.3742870268485906e-06,
+      "loss": 0.8932,
       "step": 24650
     },
     {
-      "epoch": 0.6770206805635659,
+      "epoch": 0.6995175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 4.990245491932678e-06,
-      "loss": 0.7108,
+      "learning_rate": 4.373527203943609e-06,
+      "loss": 0.9189,
       "step": 24651
     },
     {
-      "epoch": 0.6770481447914092,
+      "epoch": 0.6995459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 4.989475669895935e-06,
-      "loss": 0.8884,
+      "learning_rate": 4.372767428565701e-06,
+      "loss": 0.7742,
       "step": 24652
     },
     {
-      "epoch": 0.6770756090192525,
+      "epoch": 0.6995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 4.988705887504398e-06,
-      "loss": 0.8321,
+      "learning_rate": 4.372007700721275e-06,
+      "loss": 0.8286,
       "step": 24653
     },
     {
-      "epoch": 0.6771030732470956,
+      "epoch": 0.6996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 4.987936144764161e-06,
-      "loss": 0.9382,
+      "learning_rate": 4.371248020416752e-06,
+      "loss": 0.8757,
       "step": 24654
     },
     {
-      "epoch": 0.6771305374749389,
+      "epoch": 0.6996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 4.9871664416813135e-06,
-      "loss": 0.8923,
+      "learning_rate": 4.370488387658555e-06,
+      "loss": 0.8627,
       "step": 24655
     },
     {
-      "epoch": 0.6771580017027822,
+      "epoch": 0.6996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 4.9863967782619485e-06,
-      "loss": 0.7978,
+      "learning_rate": 4.369728802453091e-06,
+      "loss": 0.8186,
       "step": 24656
     },
     {
-      "epoch": 0.6771854659306253,
+      "epoch": 0.6996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 4.985627154512154e-06,
-      "loss": 0.8951,
+      "learning_rate": 4.368969264806781e-06,
+      "loss": 0.7877,
       "step": 24657
     },
     {
-      "epoch": 0.6772129301584686,
+      "epoch": 0.699716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 4.984857570438023e-06,
-      "loss": 0.9341,
+      "learning_rate": 4.368209774726045e-06,
+      "loss": 0.9369,
       "step": 24658
     },
     {
-      "epoch": 0.6772403943863118,
+      "epoch": 0.699744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 4.984088026045642e-06,
-      "loss": 0.8022,
+      "learning_rate": 4.367450332217291e-06,
+      "loss": 0.7722,
       "step": 24659
     },
     {
-      "epoch": 0.6772678586141551,
+      "epoch": 0.6997729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 4.983318521341096e-06,
-      "loss": 0.8119,
+      "learning_rate": 4.366690937286936e-06,
+      "loss": 0.8128,
       "step": 24660
     },
     {
-      "epoch": 0.6772953228419983,
+      "epoch": 0.6998013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 4.982549056330477e-06,
-      "loss": 0.8009,
+      "learning_rate": 4.3659315899414e-06,
+      "loss": 0.9592,
       "step": 24661
     },
     {
-      "epoch": 0.6773227870698415,
+      "epoch": 0.6998297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 4.981779631019873e-06,
-      "loss": 0.8421,
+      "learning_rate": 4.365172290187086e-06,
+      "loss": 0.9352,
       "step": 24662
     },
     {
-      "epoch": 0.6773502512976848,
+      "epoch": 0.6998581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 4.9810102454153765e-06,
-      "loss": 0.8024,
+      "learning_rate": 4.364413038030423e-06,
+      "loss": 0.842,
       "step": 24663
     },
     {
-      "epoch": 0.677377715525528,
+      "epoch": 0.6998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 4.980240899523067e-06,
-      "loss": 0.9813,
+      "learning_rate": 4.363653833477812e-06,
+      "loss": 0.8501,
       "step": 24664
     },
     {
-      "epoch": 0.6774051797533712,
+      "epoch": 0.6999148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 4.979471593349037e-06,
-      "loss": 0.9172,
+      "learning_rate": 4.36289467653567e-06,
+      "loss": 0.8722,
       "step": 24665
     },
     {
-      "epoch": 0.6774326439812145,
+      "epoch": 0.6999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 4.978702326899371e-06,
-      "loss": 0.8954,
+      "learning_rate": 4.362135567210415e-06,
+      "loss": 0.9048,
       "step": 24666
     },
     {
-      "epoch": 0.6774601082090577,
+      "epoch": 0.6999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 4.977933100180163e-06,
-      "loss": 0.8806,
+      "learning_rate": 4.3613765055084485e-06,
+      "loss": 0.8504,
       "step": 24667
     },
     {
-      "epoch": 0.6774875724369009,
+      "epoch": 0.7,
       "grad_norm": 0.0,
-      "learning_rate": 4.977163913197488e-06,
-      "loss": 0.8766,
+      "learning_rate": 4.36061749143619e-06,
+      "loss": 0.7967,
       "step": 24668
     },
     {
-      "epoch": 0.6775150366647442,
+      "epoch": 0.7000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 4.976394765957442e-06,
-      "loss": 0.8033,
+      "learning_rate": 4.359858525000052e-06,
+      "loss": 0.8144,
       "step": 24669
     },
     {
-      "epoch": 0.6775425008925874,
+      "epoch": 0.7000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 4.9756256584661036e-06,
-      "loss": 0.8675,
+      "learning_rate": 4.35909960620644e-06,
+      "loss": 0.8864,
       "step": 24670
     },
     {
-      "epoch": 0.6775699651204307,
+      "epoch": 0.7000851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 4.9748565907295595e-06,
-      "loss": 0.809,
+      "learning_rate": 4.358340735061766e-06,
+      "loss": 0.8333,
       "step": 24671
     },
     {
-      "epoch": 0.6775974293482738,
+      "epoch": 0.7001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 4.974087562753899e-06,
-      "loss": 0.8353,
+      "learning_rate": 4.357581911572445e-06,
+      "loss": 0.7327,
       "step": 24672
     },
     {
-      "epoch": 0.6776248935761171,
+      "epoch": 0.7001418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 4.973318574545203e-06,
-      "loss": 0.844,
+      "learning_rate": 4.35682313574488e-06,
+      "loss": 0.9391,
       "step": 24673
     },
     {
-      "epoch": 0.6776523578039604,
+      "epoch": 0.7001702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 4.972549626109562e-06,
-      "loss": 0.9042,
+      "learning_rate": 4.356064407585484e-06,
+      "loss": 0.8226,
       "step": 24674
     },
     {
-      "epoch": 0.6776798220318035,
+      "epoch": 0.7001986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 4.971780717453051e-06,
-      "loss": 0.8075,
+      "learning_rate": 4.355305727100668e-06,
+      "loss": 0.961,
       "step": 24675
     },
     {
-      "epoch": 0.6777072862596468,
+      "epoch": 0.7002270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 4.9710118485817635e-06,
-      "loss": 0.9093,
+      "learning_rate": 4.354547094296836e-06,
+      "loss": 0.7304,
       "step": 24676
     },
     {
-      "epoch": 0.67773475048749,
+      "epoch": 0.700255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 4.970243019501774e-06,
-      "loss": 0.7804,
+      "learning_rate": 4.353788509180398e-06,
+      "loss": 0.8344,
       "step": 24677
     },
     {
-      "epoch": 0.6777622147153333,
+      "epoch": 0.7002837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 4.96947423021917e-06,
-      "loss": 0.8119,
+      "learning_rate": 4.353029971757762e-06,
+      "loss": 0.8344,
       "step": 24678
     },
     {
-      "epoch": 0.6777896789431765,
+      "epoch": 0.7003121452894439,
       "grad_norm": 0.0,
-      "learning_rate": 4.968705480740035e-06,
-      "loss": 0.8635,
+      "learning_rate": 4.352271482035336e-06,
+      "loss": 0.9312,
       "step": 24679
     },
     {
-      "epoch": 0.6778171431710197,
+      "epoch": 0.7003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 4.967936771070452e-06,
-      "loss": 0.8361,
+      "learning_rate": 4.3515130400195305e-06,
+      "loss": 0.8436,
       "step": 24680
     },
     {
-      "epoch": 0.677844607398863,
+      "epoch": 0.7003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 4.967168101216505e-06,
-      "loss": 0.7016,
+      "learning_rate": 4.3507546457167445e-06,
+      "loss": 0.746,
       "step": 24681
     },
     {
-      "epoch": 0.6778720716267063,
+      "epoch": 0.7003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 4.96639947118427e-06,
-      "loss": 0.8772,
+      "learning_rate": 4.3499962991333874e-06,
+      "loss": 0.8048,
       "step": 24682
     },
     {
-      "epoch": 0.6778995358545494,
+      "epoch": 0.7004256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 4.965630880979836e-06,
-      "loss": 0.796,
+      "learning_rate": 4.349238000275871e-06,
+      "loss": 0.7961,
       "step": 24683
     },
     {
-      "epoch": 0.6779270000823927,
+      "epoch": 0.7004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 4.964862330609277e-06,
-      "loss": 0.8287,
+      "learning_rate": 4.348479749150589e-06,
+      "loss": 0.8039,
       "step": 24684
     },
     {
-      "epoch": 0.6779544643102359,
+      "epoch": 0.7004824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 4.964093820078677e-06,
-      "loss": 0.8203,
+      "learning_rate": 4.347721545763955e-06,
+      "loss": 0.8082,
       "step": 24685
     },
     {
-      "epoch": 0.6779819285380791,
+      "epoch": 0.7005107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 4.9633253493941184e-06,
-      "loss": 0.8309,
+      "learning_rate": 4.346963390122373e-06,
+      "loss": 0.8329,
       "step": 24686
     },
     {
-      "epoch": 0.6780093927659224,
+      "epoch": 0.700539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 4.962556918561684e-06,
-      "loss": 0.8803,
+      "learning_rate": 4.346205282232242e-06,
+      "loss": 0.733,
       "step": 24687
     },
     {
-      "epoch": 0.6780368569937656,
+      "epoch": 0.7005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 4.961788527587447e-06,
-      "loss": 0.8651,
+      "learning_rate": 4.34544722209997e-06,
+      "loss": 0.8129,
       "step": 24688
     },
     {
-      "epoch": 0.6780643212216089,
+      "epoch": 0.7005959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 4.96102017647749e-06,
-      "loss": 0.8265,
+      "learning_rate": 4.344689209731963e-06,
+      "loss": 0.8329,
       "step": 24689
     },
     {
-      "epoch": 0.678091785449452,
+      "epoch": 0.7006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 4.960251865237894e-06,
-      "loss": 0.765,
+      "learning_rate": 4.343931245134616e-06,
+      "loss": 0.8298,
       "step": 24690
     },
     {
-      "epoch": 0.6781192496772953,
+      "epoch": 0.7006526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 4.959483593874741e-06,
-      "loss": 0.8073,
+      "learning_rate": 4.3431733283143365e-06,
+      "loss": 0.8414,
       "step": 24691
     },
     {
-      "epoch": 0.6781467139051386,
+      "epoch": 0.7006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 4.958715362394103e-06,
-      "loss": 0.9269,
+      "learning_rate": 4.34241545927753e-06,
+      "loss": 0.8267,
       "step": 24692
     },
     {
-      "epoch": 0.6781741781329818,
+      "epoch": 0.7007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 4.957947170802065e-06,
-      "loss": 0.7775,
+      "learning_rate": 4.341657638030587e-06,
+      "loss": 0.8172,
       "step": 24693
     },
     {
-      "epoch": 0.678201642360825,
+      "epoch": 0.7007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 4.957179019104698e-06,
-      "loss": 0.8143,
+      "learning_rate": 4.340899864579924e-06,
+      "loss": 0.8276,
       "step": 24694
     },
     {
-      "epoch": 0.6782291065886683,
+      "epoch": 0.700766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 4.956410907308085e-06,
-      "loss": 0.9638,
+      "learning_rate": 4.34014213893193e-06,
+      "loss": 0.8629,
       "step": 24695
     },
     {
-      "epoch": 0.6782565708165115,
+      "epoch": 0.700794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 4.955642835418301e-06,
-      "loss": 0.7918,
+      "learning_rate": 4.33938446109301e-06,
+      "loss": 0.7901,
       "step": 24696
     },
     {
-      "epoch": 0.6782840350443547,
+      "epoch": 0.7008229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 4.954874803441427e-06,
-      "loss": 0.8038,
+      "learning_rate": 4.3386268310695675e-06,
+      "loss": 0.8572,
       "step": 24697
     },
     {
-      "epoch": 0.6783114992721979,
+      "epoch": 0.7008513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 4.954106811383541e-06,
-      "loss": 0.8398,
+      "learning_rate": 4.337869248867995e-06,
+      "loss": 0.8784,
       "step": 24698
     },
     {
-      "epoch": 0.6783389635000412,
+      "epoch": 0.7008796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 4.9533388592507125e-06,
-      "loss": 0.8469,
+      "learning_rate": 4.337111714494696e-06,
+      "loss": 0.8506,
       "step": 24699
     },
     {
-      "epoch": 0.6783664277278845,
+      "epoch": 0.7009080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 4.952570947049026e-06,
-      "loss": 0.7991,
+      "learning_rate": 4.336354227956072e-06,
+      "loss": 0.8776,
       "step": 24700
     },
     {
-      "epoch": 0.6783938919557276,
+      "epoch": 0.7009364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 4.951803074784548e-06,
-      "loss": 0.7663,
+      "learning_rate": 4.335596789258515e-06,
+      "loss": 0.8533,
       "step": 24701
     },
     {
-      "epoch": 0.6784213561835709,
+      "epoch": 0.7009648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 4.951035242463361e-06,
-      "loss": 0.9066,
+      "learning_rate": 4.334839398408426e-06,
+      "loss": 0.96,
       "step": 24702
     },
     {
-      "epoch": 0.6784488204114142,
+      "epoch": 0.7009931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 4.950267450091539e-06,
-      "loss": 0.8571,
+      "learning_rate": 4.334082055412207e-06,
+      "loss": 0.9534,
       "step": 24703
     },
     {
-      "epoch": 0.6784762846392574,
+      "epoch": 0.7010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 4.949499697675156e-06,
-      "loss": 0.9128,
+      "learning_rate": 4.3333247602762485e-06,
+      "loss": 0.8842,
       "step": 24704
     },
     {
-      "epoch": 0.6785037488671006,
+      "epoch": 0.701049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 4.948731985220293e-06,
-      "loss": 0.8138,
+      "learning_rate": 4.332567513006951e-06,
+      "loss": 0.8137,
       "step": 24705
     },
     {
-      "epoch": 0.6785312130949438,
+      "epoch": 0.701078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 4.947964312733015e-06,
-      "loss": 0.8307,
+      "learning_rate": 4.331810313610713e-06,
+      "loss": 0.7428,
       "step": 24706
     },
     {
-      "epoch": 0.6785586773227871,
+      "epoch": 0.7011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 4.947196680219402e-06,
-      "loss": 0.8522,
+      "learning_rate": 4.331053162093924e-06,
+      "loss": 0.8613,
       "step": 24707
     },
     {
-      "epoch": 0.6785861415506304,
+      "epoch": 0.7011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 4.946429087685525e-06,
-      "loss": 0.801,
+      "learning_rate": 4.330296058462982e-06,
+      "loss": 0.8881,
       "step": 24708
     },
     {
-      "epoch": 0.6786136057784735,
+      "epoch": 0.7011634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 4.945661535137457e-06,
-      "loss": 0.8596,
+      "learning_rate": 4.32953900272429e-06,
+      "loss": 0.8657,
       "step": 24709
     },
     {
-      "epoch": 0.6786410700063168,
+      "epoch": 0.7011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 4.944894022581273e-06,
-      "loss": 0.8234,
+      "learning_rate": 4.3287819948842334e-06,
+      "loss": 0.9183,
       "step": 24710
     },
     {
-      "epoch": 0.67866853423416,
+      "epoch": 0.7012202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 4.944126550023051e-06,
-      "loss": 0.8891,
+      "learning_rate": 4.328025034949211e-06,
+      "loss": 0.8133,
       "step": 24711
     },
     {
-      "epoch": 0.6786959984620032,
+      "epoch": 0.7012485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 4.943359117468854e-06,
-      "loss": 0.8767,
+      "learning_rate": 4.32726812292562e-06,
+      "loss": 0.8626,
       "step": 24712
     },
     {
-      "epoch": 0.6787234626898465,
+      "epoch": 0.7012769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 4.942591724924757e-06,
-      "loss": 0.8702,
+      "learning_rate": 4.326511258819846e-06,
+      "loss": 0.7636,
       "step": 24713
     },
     {
-      "epoch": 0.6787509269176897,
+      "epoch": 0.701305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 4.941824372396834e-06,
-      "loss": 0.8782,
+      "learning_rate": 4.325754442638289e-06,
+      "loss": 0.8256,
       "step": 24714
     },
     {
-      "epoch": 0.678778391145533,
+      "epoch": 0.7013337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 4.94105705989116e-06,
-      "loss": 0.7777,
+      "learning_rate": 4.324997674387337e-06,
+      "loss": 0.9475,
       "step": 24715
     },
     {
-      "epoch": 0.6788058553733762,
+      "epoch": 0.7013620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 4.940289787413798e-06,
-      "loss": 0.8049,
+      "learning_rate": 4.324240954073383e-06,
+      "loss": 0.8175,
       "step": 24716
     },
     {
-      "epoch": 0.6788333196012194,
+      "epoch": 0.7013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 4.939522554970827e-06,
-      "loss": 0.8248,
+      "learning_rate": 4.323484281702827e-06,
+      "loss": 0.8831,
       "step": 24717
     },
     {
-      "epoch": 0.6788607838290627,
+      "epoch": 0.7014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 4.9387553625683105e-06,
-      "loss": 0.9119,
+      "learning_rate": 4.322727657282048e-06,
+      "loss": 0.8914,
       "step": 24718
     },
     {
-      "epoch": 0.6788882480569058,
+      "epoch": 0.7014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 4.937988210212321e-06,
-      "loss": 0.7791,
+      "learning_rate": 4.3219710808174465e-06,
+      "loss": 0.7773,
       "step": 24719
     },
     {
-      "epoch": 0.6789157122847491,
+      "epoch": 0.7014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 4.93722109790893e-06,
-      "loss": 0.8889,
+      "learning_rate": 4.321214552315413e-06,
+      "loss": 0.7747,
       "step": 24720
     },
     {
-      "epoch": 0.6789431765125924,
+      "epoch": 0.7015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 4.936454025664208e-06,
-      "loss": 0.8643,
+      "learning_rate": 4.320458071782331e-06,
+      "loss": 0.8088,
       "step": 24721
     },
     {
-      "epoch": 0.6789706407404356,
+      "epoch": 0.7015323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 4.935686993484226e-06,
-      "loss": 0.7914,
+      "learning_rate": 4.319701639224596e-06,
+      "loss": 0.787,
       "step": 24722
     },
     {
-      "epoch": 0.6789981049682788,
+      "epoch": 0.7015607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 4.934920001375051e-06,
-      "loss": 0.8123,
+      "learning_rate": 4.3189452546486e-06,
+      "loss": 0.8452,
       "step": 24723
     },
     {
-      "epoch": 0.679025569196122,
+      "epoch": 0.701589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 4.9341530493427455e-06,
-      "loss": 0.777,
+      "learning_rate": 4.318188918060721e-06,
+      "loss": 0.754,
       "step": 24724
     },
     {
-      "epoch": 0.6790530334239653,
+      "epoch": 0.7016174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 4.933386137393384e-06,
-      "loss": 0.9041,
+      "learning_rate": 4.31743262946736e-06,
+      "loss": 0.6851,
       "step": 24725
     },
     {
-      "epoch": 0.6790804976518086,
+      "epoch": 0.7016458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 4.932619265533035e-06,
-      "loss": 0.8448,
+      "learning_rate": 4.316676388874904e-06,
+      "loss": 0.9111,
       "step": 24726
     },
     {
-      "epoch": 0.6791079618796517,
+      "epoch": 0.7016742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 4.931852433767765e-06,
-      "loss": 0.843,
+      "learning_rate": 4.315920196289735e-06,
+      "loss": 0.8405,
       "step": 24727
     },
     {
-      "epoch": 0.679135426107495,
+      "epoch": 0.7017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 4.931085642103645e-06,
-      "loss": 0.9036,
+      "learning_rate": 4.315164051718243e-06,
+      "loss": 0.8559,
       "step": 24728
     },
     {
-      "epoch": 0.6791628903353383,
+      "epoch": 0.7017309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 4.930318890546736e-06,
-      "loss": 0.8747,
+      "learning_rate": 4.3144079551668205e-06,
+      "loss": 0.844,
       "step": 24729
     },
     {
-      "epoch": 0.6791903545631814,
+      "epoch": 0.7017593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 4.929552179103106e-06,
-      "loss": 0.8458,
+      "learning_rate": 4.313651906641845e-06,
+      "loss": 0.8258,
       "step": 24730
     },
     {
-      "epoch": 0.6792178187910247,
+      "epoch": 0.7017877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 4.928785507778828e-06,
-      "loss": 0.8279,
+      "learning_rate": 4.312895906149708e-06,
+      "loss": 0.7691,
       "step": 24731
     },
     {
-      "epoch": 0.6792452830188679,
+      "epoch": 0.7018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 4.9280188765799604e-06,
-      "loss": 0.7471,
+      "learning_rate": 4.312139953696797e-06,
+      "loss": 0.866,
       "step": 24732
     },
     {
-      "epoch": 0.6792727472467112,
+      "epoch": 0.7018444948921679,
       "grad_norm": 0.0,
-      "learning_rate": 4.927252285512572e-06,
-      "loss": 0.8651,
+      "learning_rate": 4.311384049289495e-06,
+      "loss": 0.8041,
       "step": 24733
     },
     {
-      "epoch": 0.6793002114745544,
+      "epoch": 0.7018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 4.926485734582732e-06,
-      "loss": 0.8835,
+      "learning_rate": 4.3106281929341855e-06,
+      "loss": 0.7866,
       "step": 24734
     },
     {
-      "epoch": 0.6793276757023976,
+      "epoch": 0.7019012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 4.925719223796498e-06,
-      "loss": 0.8398,
+      "learning_rate": 4.309872384637259e-06,
+      "loss": 0.905,
       "step": 24735
     },
     {
-      "epoch": 0.6793551399302409,
+      "epoch": 0.7019296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 4.92495275315994e-06,
-      "loss": 0.8344,
+      "learning_rate": 4.309116624405093e-06,
+      "loss": 0.8184,
       "step": 24736
     },
     {
-      "epoch": 0.679382604158084,
+      "epoch": 0.7019580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 4.92418632267912e-06,
-      "loss": 1.0033,
+      "learning_rate": 4.308360912244074e-06,
+      "loss": 0.7883,
       "step": 24737
     },
     {
-      "epoch": 0.6794100683859273,
+      "epoch": 0.7019863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 4.923419932360105e-06,
-      "loss": 0.7725,
+      "learning_rate": 4.307605248160591e-06,
+      "loss": 0.8121,
       "step": 24738
     },
     {
-      "epoch": 0.6794375326137706,
+      "epoch": 0.7020147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 4.922653582208963e-06,
-      "loss": 0.8688,
+      "learning_rate": 4.306849632161015e-06,
+      "loss": 0.8872,
       "step": 24739
     },
     {
-      "epoch": 0.6794649968416138,
+      "epoch": 0.7020431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 4.92188727223175e-06,
-      "loss": 0.8678,
+      "learning_rate": 4.306094064251742e-06,
+      "loss": 0.7977,
       "step": 24740
     },
     {
-      "epoch": 0.679492461069457,
+      "epoch": 0.7020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 4.92112100243453e-06,
-      "loss": 0.9474,
+      "learning_rate": 4.305338544439146e-06,
+      "loss": 0.7802,
       "step": 24741
     },
     {
-      "epoch": 0.6795199252973003,
+      "epoch": 0.702099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 4.920354772823368e-06,
-      "loss": 0.8023,
+      "learning_rate": 4.304583072729611e-06,
+      "loss": 0.8788,
       "step": 24742
     },
     {
-      "epoch": 0.6795473895251435,
+      "epoch": 0.7021282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 4.919588583404327e-06,
-      "loss": 0.7992,
+      "learning_rate": 4.303827649129522e-06,
+      "loss": 0.759,
       "step": 24743
     },
     {
-      "epoch": 0.6795748537529868,
+      "epoch": 0.7021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 4.9188224341834674e-06,
-      "loss": 0.8561,
+      "learning_rate": 4.3030722736452545e-06,
+      "loss": 0.9512,
       "step": 24744
     },
     {
-      "epoch": 0.6796023179808299,
+      "epoch": 0.7021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 4.918056325166855e-06,
-      "loss": 0.8849,
+      "learning_rate": 4.302316946283192e-06,
+      "loss": 0.8904,
       "step": 24745
     },
     {
-      "epoch": 0.6796297822086732,
+      "epoch": 0.7022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 4.917290256360552e-06,
-      "loss": 0.9626,
+      "learning_rate": 4.301561667049716e-06,
+      "loss": 0.8293,
       "step": 24746
     },
     {
-      "epoch": 0.6796572464365165,
+      "epoch": 0.7022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 4.916524227770617e-06,
-      "loss": 0.7855,
+      "learning_rate": 4.300806435951203e-06,
+      "loss": 0.9927,
       "step": 24747
     },
     {
-      "epoch": 0.6796847106643596,
+      "epoch": 0.7022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 4.915758239403109e-06,
-      "loss": 0.8491,
+      "learning_rate": 4.300051252994032e-06,
+      "loss": 0.933,
       "step": 24748
     },
     {
-      "epoch": 0.6797121748922029,
+      "epoch": 0.7022985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 4.91499229126409e-06,
-      "loss": 0.7961,
+      "learning_rate": 4.299296118184589e-06,
+      "loss": 0.8892,
       "step": 24749
     },
     {
-      "epoch": 0.6797396391200461,
+      "epoch": 0.7023269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 4.914226383359623e-06,
-      "loss": 0.865,
+      "learning_rate": 4.2985410315292455e-06,
+      "loss": 0.9164,
       "step": 24750
     },
     {
-      "epoch": 0.6797671033478894,
+      "epoch": 0.7023552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 4.913460515695766e-06,
-      "loss": 0.9091,
+      "learning_rate": 4.297785993034381e-06,
+      "loss": 0.9053,
       "step": 24751
     },
     {
-      "epoch": 0.6797945675757326,
+      "epoch": 0.7023836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 4.912694688278583e-06,
-      "loss": 0.868,
+      "learning_rate": 4.297031002706377e-06,
+      "loss": 0.7591,
       "step": 24752
     },
     {
-      "epoch": 0.6798220318035758,
+      "epoch": 0.7024120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 4.911928901114127e-06,
-      "loss": 0.8293,
+      "learning_rate": 4.296276060551607e-06,
+      "loss": 0.8099,
       "step": 24753
     },
     {
-      "epoch": 0.6798494960314191,
+      "epoch": 0.7024404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 4.911163154208461e-06,
-      "loss": 0.8155,
+      "learning_rate": 4.29552116657645e-06,
+      "loss": 0.7871,
       "step": 24754
     },
     {
-      "epoch": 0.6798769602592624,
+      "epoch": 0.7024687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 4.910397447567646e-06,
-      "loss": 0.8751,
+      "learning_rate": 4.2947663207872804e-06,
+      "loss": 0.7962,
       "step": 24755
     },
     {
-      "epoch": 0.6799044244871055,
+      "epoch": 0.7024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 4.909631781197736e-06,
-      "loss": 0.8654,
+      "learning_rate": 4.294011523190477e-06,
+      "loss": 0.8552,
       "step": 24756
     },
     {
-      "epoch": 0.6799318887149488,
+      "epoch": 0.7025255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 4.90886615510479e-06,
-      "loss": 0.8785,
+      "learning_rate": 4.293256773792418e-06,
+      "loss": 0.8033,
       "step": 24757
     },
     {
-      "epoch": 0.679959352942792,
+      "epoch": 0.7025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 4.908100569294871e-06,
-      "loss": 0.8683,
+      "learning_rate": 4.292502072599471e-06,
+      "loss": 0.915,
       "step": 24758
     },
     {
-      "epoch": 0.6799868171706352,
+      "epoch": 0.7025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 4.907335023774029e-06,
-      "loss": 0.8142,
+      "learning_rate": 4.291747419618017e-06,
+      "loss": 0.8728,
       "step": 24759
     },
     {
-      "epoch": 0.6800142813984785,
+      "epoch": 0.70261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 4.906569518548325e-06,
-      "loss": 0.7894,
+      "learning_rate": 4.290992814854432e-06,
+      "loss": 0.8915,
       "step": 24760
     },
     {
-      "epoch": 0.6800417456263217,
+      "epoch": 0.702639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 4.905804053623815e-06,
-      "loss": 0.8653,
+      "learning_rate": 4.290238258315085e-06,
+      "loss": 0.8623,
       "step": 24761
     },
     {
-      "epoch": 0.680069209854165,
+      "epoch": 0.7026674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 4.905038629006562e-06,
-      "loss": 0.8469,
+      "learning_rate": 4.28948375000635e-06,
+      "loss": 0.7996,
       "step": 24762
     },
     {
-      "epoch": 0.6800966740820081,
+      "epoch": 0.7026958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 4.904273244702612e-06,
-      "loss": 0.8638,
+      "learning_rate": 4.288729289934608e-06,
+      "loss": 0.8275,
       "step": 24763
     },
     {
-      "epoch": 0.6801241383098514,
+      "epoch": 0.7027241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 4.903507900718031e-06,
-      "loss": 0.7663,
+      "learning_rate": 4.287974878106222e-06,
+      "loss": 0.8493,
       "step": 24764
     },
     {
-      "epoch": 0.6801516025376947,
+      "epoch": 0.7027525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 4.902742597058865e-06,
-      "loss": 0.8074,
+      "learning_rate": 4.287220514527569e-06,
+      "loss": 0.7458,
       "step": 24765
     },
     {
-      "epoch": 0.6801790667655379,
+      "epoch": 0.7027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 4.901977333731174e-06,
-      "loss": 0.8739,
+      "learning_rate": 4.286466199205025e-06,
+      "loss": 0.7979,
       "step": 24766
     },
     {
-      "epoch": 0.6802065309933811,
+      "epoch": 0.7028093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 4.901212110741013e-06,
-      "loss": 0.8553,
+      "learning_rate": 4.2857119321449536e-06,
+      "loss": 0.9006,
       "step": 24767
     },
     {
-      "epoch": 0.6802339952212244,
+      "epoch": 0.7028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 4.900446928094437e-06,
-      "loss": 0.8729,
+      "learning_rate": 4.284957713353731e-06,
+      "loss": 0.8017,
       "step": 24768
     },
     {
-      "epoch": 0.6802614594490676,
+      "epoch": 0.7028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 4.899681785797501e-06,
-      "loss": 0.7371,
+      "learning_rate": 4.284203542837732e-06,
+      "loss": 0.9023,
       "step": 24769
     },
     {
-      "epoch": 0.6802889236769109,
+      "epoch": 0.702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 4.898916683856262e-06,
-      "loss": 0.7859,
+      "learning_rate": 4.283449420603312e-06,
+      "loss": 0.8898,
       "step": 24770
     },
     {
-      "epoch": 0.680316387904754,
+      "epoch": 0.7029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 4.898151622276769e-06,
-      "loss": 0.8484,
+      "learning_rate": 4.2826953466568626e-06,
+      "loss": 0.8231,
       "step": 24771
     },
     {
-      "epoch": 0.6803438521325973,
+      "epoch": 0.7029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 4.897386601065074e-06,
-      "loss": 0.8202,
+      "learning_rate": 4.281941321004738e-06,
+      "loss": 0.8453,
       "step": 24772
     },
     {
-      "epoch": 0.6803713163604406,
+      "epoch": 0.7029795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 4.896621620227232e-06,
-      "loss": 0.8688,
+      "learning_rate": 4.2811873436533116e-06,
+      "loss": 0.7674,
       "step": 24773
     },
     {
-      "epoch": 0.6803987805882837,
+      "epoch": 0.7030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 4.8958566797692974e-06,
-      "loss": 0.7222,
+      "learning_rate": 4.2804334146089566e-06,
+      "loss": 0.7903,
       "step": 24774
     },
     {
-      "epoch": 0.680426244816127,
+      "epoch": 0.7030363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 4.895091779697321e-06,
-      "loss": 0.8423,
+      "learning_rate": 4.2796795338780336e-06,
+      "loss": 0.8443,
       "step": 24775
     },
     {
-      "epoch": 0.6804537090439702,
+      "epoch": 0.7030646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 4.89432692001736e-06,
-      "loss": 0.9102,
+      "learning_rate": 4.278925701466915e-06,
+      "loss": 0.7853,
       "step": 24776
     },
     {
-      "epoch": 0.6804811732718135,
+      "epoch": 0.7030930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 4.8935621007354594e-06,
-      "loss": 0.8504,
+      "learning_rate": 4.2781719173819725e-06,
+      "loss": 0.8109,
       "step": 24777
     },
     {
-      "epoch": 0.6805086374996567,
+      "epoch": 0.7031214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 4.892797321857672e-06,
-      "loss": 0.8479,
+      "learning_rate": 4.2774181816295645e-06,
+      "loss": 0.7489,
       "step": 24778
     },
     {
-      "epoch": 0.6805361017274999,
+      "epoch": 0.703149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 4.892032583390055e-06,
-      "loss": 0.8511,
+      "learning_rate": 4.276664494216063e-06,
+      "loss": 0.8493,
       "step": 24779
     },
     {
-      "epoch": 0.6805635659553432,
+      "epoch": 0.7031782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 4.891267885338652e-06,
-      "loss": 0.8574,
+      "learning_rate": 4.275910855147837e-06,
+      "loss": 0.8453,
       "step": 24780
     },
     {
-      "epoch": 0.6805910301831865,
+      "epoch": 0.7032065834279229,
       "grad_norm": 0.0,
-      "learning_rate": 4.890503227709519e-06,
-      "loss": 0.8379,
+      "learning_rate": 4.275157264431246e-06,
+      "loss": 0.8415,
       "step": 24781
     },
     {
-      "epoch": 0.6806184944110296,
+      "epoch": 0.7032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 4.8897386105087e-06,
-      "loss": 0.9209,
+      "learning_rate": 4.274403722072658e-06,
+      "loss": 0.8193,
       "step": 24782
     },
     {
-      "epoch": 0.6806459586388729,
+      "epoch": 0.7032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 4.8889740337422485e-06,
-      "loss": 0.8259,
+      "learning_rate": 4.273650228078444e-06,
+      "loss": 0.8791,
       "step": 24783
     },
     {
-      "epoch": 0.6806734228667161,
+      "epoch": 0.7032917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 4.888209497416214e-06,
-      "loss": 0.8802,
+      "learning_rate": 4.27289678245496e-06,
+      "loss": 0.8417,
       "step": 24784
     },
     {
-      "epoch": 0.6807008870945593,
+      "epoch": 0.7033200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 4.887445001536647e-06,
-      "loss": 0.7671,
+      "learning_rate": 4.272143385208574e-06,
+      "loss": 0.9188,
       "step": 24785
     },
     {
-      "epoch": 0.6807283513224026,
+      "epoch": 0.7033484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 4.886680546109599e-06,
-      "loss": 0.7736,
+      "learning_rate": 4.271390036345651e-06,
+      "loss": 0.8309,
       "step": 24786
     },
     {
-      "epoch": 0.6807558155502458,
+      "epoch": 0.7033768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 4.885916131141112e-06,
-      "loss": 0.9135,
+      "learning_rate": 4.270636735872553e-06,
+      "loss": 0.8011,
       "step": 24787
     },
     {
-      "epoch": 0.6807832797780891,
+      "epoch": 0.7034052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 4.885151756637241e-06,
-      "loss": 0.8546,
+      "learning_rate": 4.2698834837956484e-06,
+      "loss": 0.9146,
       "step": 24788
     },
     {
-      "epoch": 0.6808107440059322,
+      "epoch": 0.703433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 4.8843874226040265e-06,
-      "loss": 0.8284,
+      "learning_rate": 4.269130280121291e-06,
+      "loss": 0.8353,
       "step": 24789
     },
     {
-      "epoch": 0.6808382082337755,
+      "epoch": 0.7034619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 4.88362312904752e-06,
-      "loss": 0.889,
+      "learning_rate": 4.268377124855849e-06,
+      "loss": 0.9107,
       "step": 24790
     },
     {
-      "epoch": 0.6808656724616188,
+      "epoch": 0.7034903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 4.88285887597377e-06,
-      "loss": 0.8629,
+      "learning_rate": 4.267624018005686e-06,
+      "loss": 0.7966,
       "step": 24791
     },
     {
-      "epoch": 0.6808931366894619,
+      "epoch": 0.7035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 4.882094663388822e-06,
-      "loss": 0.805,
+      "learning_rate": 4.266870959577157e-06,
+      "loss": 0.8712,
       "step": 24792
     },
     {
-      "epoch": 0.6809206009173052,
+      "epoch": 0.7035471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 4.881330491298728e-06,
-      "loss": 0.8713,
+      "learning_rate": 4.266117949576627e-06,
+      "loss": 0.8501,
       "step": 24793
     },
     {
-      "epoch": 0.6809480651451485,
+      "epoch": 0.7035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 4.8805663597095256e-06,
-      "loss": 1.0116,
+      "learning_rate": 4.2653649880104595e-06,
+      "loss": 0.8966,
       "step": 24794
     },
     {
-      "epoch": 0.6809755293729917,
+      "epoch": 0.7036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 4.87980226862727e-06,
-      "loss": 0.8049,
+      "learning_rate": 4.264612074885008e-06,
+      "loss": 0.8168,
       "step": 24795
     },
     {
-      "epoch": 0.6810029936008349,
+      "epoch": 0.7036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 4.879038218057996e-06,
-      "loss": 0.8377,
+      "learning_rate": 4.263859210206637e-06,
+      "loss": 0.83,
       "step": 24796
     },
     {
-      "epoch": 0.6810304578286781,
+      "epoch": 0.7036606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 4.878274208007758e-06,
-      "loss": 0.9736,
+      "learning_rate": 4.263106393981708e-06,
+      "loss": 0.8477,
       "step": 24797
     },
     {
-      "epoch": 0.6810579220565214,
+      "epoch": 0.703688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 4.877510238482598e-06,
-      "loss": 0.8356,
+      "learning_rate": 4.262353626216575e-06,
+      "loss": 0.8787,
       "step": 24798
     },
     {
-      "epoch": 0.6810853862843647,
+      "epoch": 0.7037173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 4.876746309488564e-06,
-      "loss": 0.8467,
+      "learning_rate": 4.261600906917596e-06,
+      "loss": 0.8133,
       "step": 24799
     },
     {
-      "epoch": 0.6811128505122078,
+      "epoch": 0.7037457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 4.875982421031694e-06,
-      "loss": 0.8453,
+      "learning_rate": 4.260848236091135e-06,
+      "loss": 0.8557,
       "step": 24800
     },
     {
-      "epoch": 0.6811403147400511,
+      "epoch": 0.7037741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 4.8752185731180364e-06,
-      "loss": 0.7828,
+      "learning_rate": 4.260095613743546e-06,
+      "loss": 0.8428,
       "step": 24801
     },
     {
-      "epoch": 0.6811677789678943,
+      "epoch": 0.7038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 4.874454765753635e-06,
-      "loss": 0.8521,
+      "learning_rate": 4.25934303988119e-06,
+      "loss": 0.8896,
       "step": 24802
     },
     {
-      "epoch": 0.6811952431957375,
+      "epoch": 0.7038308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 4.873690998944536e-06,
-      "loss": 0.7848,
+      "learning_rate": 4.258590514510419e-06,
+      "loss": 0.8691,
       "step": 24803
     },
     {
-      "epoch": 0.6812227074235808,
+      "epoch": 0.7038592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 4.8729272726967754e-06,
-      "loss": 0.8546,
+      "learning_rate": 4.257838037637591e-06,
+      "loss": 0.8141,
       "step": 24804
     },
     {
-      "epoch": 0.681250171651424,
+      "epoch": 0.7038876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 4.872163587016405e-06,
-      "loss": 0.8032,
+      "learning_rate": 4.2570856092690686e-06,
+      "loss": 0.7505,
       "step": 24805
     },
     {
-      "epoch": 0.6812776358792673,
+      "epoch": 0.7039160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 4.8713999419094594e-06,
-      "loss": 0.8573,
+      "learning_rate": 4.256333229411197e-06,
+      "loss": 0.7638,
       "step": 24806
     },
     {
-      "epoch": 0.6813051001071105,
+      "epoch": 0.70394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 4.8706363373819835e-06,
-      "loss": 0.871,
+      "learning_rate": 4.255580898070337e-06,
+      "loss": 0.7949,
       "step": 24807
     },
     {
-      "epoch": 0.6813325643349537,
+      "epoch": 0.703972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 4.869872773440018e-06,
-      "loss": 0.756,
+      "learning_rate": 4.2548286152528474e-06,
+      "loss": 0.716,
       "step": 24808
     },
     {
-      "epoch": 0.681360028562797,
+      "epoch": 0.7040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 4.8691092500896085e-06,
-      "loss": 0.8812,
+      "learning_rate": 4.2540763809650745e-06,
+      "loss": 0.7796,
       "step": 24809
     },
     {
-      "epoch": 0.6813874927906401,
+      "epoch": 0.7040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 4.868345767336797e-06,
-      "loss": 0.8181,
+      "learning_rate": 4.253324195213377e-06,
+      "loss": 0.7687,
       "step": 24810
     },
     {
-      "epoch": 0.6814149570184834,
+      "epoch": 0.7040578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 4.867582325187619e-06,
-      "loss": 0.8665,
+      "learning_rate": 4.252572058004112e-06,
+      "loss": 0.8872,
       "step": 24811
     },
     {
-      "epoch": 0.6814424212463267,
+      "epoch": 0.7040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 4.8668189236481215e-06,
-      "loss": 0.8623,
+      "learning_rate": 4.251819969343626e-06,
+      "loss": 0.9587,
       "step": 24812
     },
     {
-      "epoch": 0.6814698854741699,
+      "epoch": 0.7041146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 4.866055562724335e-06,
-      "loss": 0.9122,
+      "learning_rate": 4.251067929238275e-06,
+      "loss": 0.8067,
       "step": 24813
     },
     {
-      "epoch": 0.6814973497020131,
+      "epoch": 0.7041430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 4.865292242422308e-06,
-      "loss": 0.8451,
+      "learning_rate": 4.250315937694415e-06,
+      "loss": 0.8564,
       "step": 24814
     },
     {
-      "epoch": 0.6815248139298563,
+      "epoch": 0.7041713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 4.864528962748076e-06,
-      "loss": 0.8685,
+      "learning_rate": 4.249563994718391e-06,
+      "loss": 0.9404,
       "step": 24815
     },
     {
-      "epoch": 0.6815522781576996,
+      "epoch": 0.704199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 4.863765723707682e-06,
-      "loss": 0.8671,
+      "learning_rate": 4.248812100316555e-06,
+      "loss": 0.8499,
       "step": 24816
     },
     {
-      "epoch": 0.6815797423855429,
+      "epoch": 0.7042281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 4.863002525307166e-06,
-      "loss": 0.9405,
+      "learning_rate": 4.248060254495269e-06,
+      "loss": 0.8423,
       "step": 24817
     },
     {
-      "epoch": 0.681607206613386,
+      "epoch": 0.7042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 4.862239367552559e-06,
-      "loss": 0.9367,
+      "learning_rate": 4.247308457260874e-06,
+      "loss": 0.8063,
       "step": 24818
     },
     {
-      "epoch": 0.6816346708412293,
+      "epoch": 0.7042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 4.86147625044991e-06,
-      "loss": 0.8162,
+      "learning_rate": 4.246556708619721e-06,
+      "loss": 0.7664,
       "step": 24819
     },
     {
-      "epoch": 0.6816621350690726,
+      "epoch": 0.7043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 4.860713174005246e-06,
-      "loss": 0.7609,
+      "learning_rate": 4.2458050085781665e-06,
+      "loss": 0.8201,
       "step": 24820
     },
     {
-      "epoch": 0.6816895992969157,
+      "epoch": 0.7043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 4.859950138224611e-06,
-      "loss": 0.7719,
+      "learning_rate": 4.2450533571425534e-06,
+      "loss": 0.8587,
       "step": 24821
     },
     {
-      "epoch": 0.681717063524759,
+      "epoch": 0.7043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 4.859187143114042e-06,
-      "loss": 0.9094,
+      "learning_rate": 4.244301754319235e-06,
+      "loss": 0.876,
       "step": 24822
     },
     {
-      "epoch": 0.6817445277526022,
+      "epoch": 0.7043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 4.858424188679579e-06,
-      "loss": 0.8669,
+      "learning_rate": 4.243550200114555e-06,
+      "loss": 0.7728,
       "step": 24823
     },
     {
-      "epoch": 0.6817719919804455,
+      "epoch": 0.7044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 4.857661274927251e-06,
-      "loss": 0.7975,
+      "learning_rate": 4.2427986945348666e-06,
+      "loss": 0.8392,
       "step": 24824
     },
     {
-      "epoch": 0.6817994562082887,
+      "epoch": 0.7044551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 4.856898401863099e-06,
-      "loss": 0.9593,
+      "learning_rate": 4.24204723758652e-06,
+      "loss": 0.8182,
       "step": 24825
     },
     {
-      "epoch": 0.6818269204361319,
+      "epoch": 0.704483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 4.85613556949316e-06,
-      "loss": 0.7733,
+      "learning_rate": 4.2412958292758544e-06,
+      "loss": 0.9066,
       "step": 24826
     },
     {
-      "epoch": 0.6818543846639752,
+      "epoch": 0.7045119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 4.855372777823473e-06,
-      "loss": 0.8505,
+      "learning_rate": 4.240544469609222e-06,
+      "loss": 0.8713,
       "step": 24827
     },
     {
-      "epoch": 0.6818818488918184,
+      "epoch": 0.7045402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 4.854610026860065e-06,
-      "loss": 0.91,
+      "learning_rate": 4.239793158592974e-06,
+      "loss": 0.7767,
       "step": 24828
     },
     {
-      "epoch": 0.6819093131196616,
+      "epoch": 0.7045686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 4.8538473166089795e-06,
-      "loss": 0.8126,
+      "learning_rate": 4.239041896233448e-06,
+      "loss": 0.8398,
       "step": 24829
     },
     {
-      "epoch": 0.6819367773475049,
+      "epoch": 0.7045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 4.853084647076242e-06,
-      "loss": 0.795,
+      "learning_rate": 4.238290682536994e-06,
+      "loss": 0.8007,
       "step": 24830
     },
     {
-      "epoch": 0.6819642415753481,
+      "epoch": 0.7046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 4.8523220182678955e-06,
-      "loss": 0.7861,
+      "learning_rate": 4.237539517509958e-06,
+      "loss": 0.8651,
       "step": 24831
     },
     {
-      "epoch": 0.6819917058031914,
+      "epoch": 0.7046538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 4.851559430189969e-06,
-      "loss": 0.793,
+      "learning_rate": 4.2367884011586836e-06,
+      "loss": 0.86,
       "step": 24832
     },
     {
-      "epoch": 0.6820191700310346,
+      "epoch": 0.7046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 4.8507968828485e-06,
-      "loss": 0.8105,
+      "learning_rate": 4.236037333489518e-06,
+      "loss": 0.9024,
       "step": 24833
     },
     {
-      "epoch": 0.6820466342588778,
+      "epoch": 0.7047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 4.8500343762495235e-06,
-      "loss": 0.8061,
+      "learning_rate": 4.235286314508808e-06,
+      "loss": 0.8617,
       "step": 24834
     },
     {
-      "epoch": 0.6820740984867211,
+      "epoch": 0.7047389330306469,
       "grad_norm": 0.0,
-      "learning_rate": 4.849271910399067e-06,
-      "loss": 0.8986,
+      "learning_rate": 4.2345353442228876e-06,
+      "loss": 0.7605,
       "step": 24835
     },
     {
-      "epoch": 0.6821015627145642,
+      "epoch": 0.7047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 4.848509485303169e-06,
-      "loss": 0.7851,
+      "learning_rate": 4.2337844226381085e-06,
+      "loss": 0.8418,
       "step": 24836
     },
     {
-      "epoch": 0.6821290269424075,
+      "epoch": 0.7047956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 4.847747100967857e-06,
-      "loss": 0.798,
+      "learning_rate": 4.233033549760815e-06,
+      "loss": 0.8129,
       "step": 24837
     },
     {
-      "epoch": 0.6821564911702508,
+      "epoch": 0.7048240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 4.846984757399164e-06,
-      "loss": 0.7447,
+      "learning_rate": 4.232282725597342e-06,
+      "loss": 0.8273,
       "step": 24838
     },
     {
-      "epoch": 0.682183955398094,
+      "epoch": 0.7048524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 4.846222454603125e-06,
-      "loss": 0.8557,
+      "learning_rate": 4.2315319501540365e-06,
+      "loss": 0.9304,
       "step": 24839
     },
     {
-      "epoch": 0.6822114196259372,
+      "epoch": 0.7048808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 4.845460192585773e-06,
-      "loss": 0.7561,
+      "learning_rate": 4.230781223437244e-06,
+      "loss": 0.9097,
       "step": 24840
     },
     {
-      "epoch": 0.6822388838537804,
+      "epoch": 0.7049091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 4.844697971353135e-06,
-      "loss": 0.7729,
+      "learning_rate": 4.230030545453298e-06,
+      "loss": 0.8294,
       "step": 24841
     },
     {
-      "epoch": 0.6822663480816237,
+      "epoch": 0.7049375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 4.843935790911241e-06,
-      "loss": 0.872,
+      "learning_rate": 4.229279916208542e-06,
+      "loss": 0.7429,
       "step": 24842
     },
     {
-      "epoch": 0.682293812309467,
+      "epoch": 0.7049659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 4.84317365126613e-06,
-      "loss": 0.854,
+      "learning_rate": 4.228529335709323e-06,
+      "loss": 0.8531,
       "step": 24843
     },
     {
-      "epoch": 0.6823212765373101,
+      "epoch": 0.704994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.842411552423822e-06,
-      "loss": 0.8193,
+      "learning_rate": 4.227778803961972e-06,
+      "loss": 0.898,
       "step": 24844
     },
     {
-      "epoch": 0.6823487407651534,
+      "epoch": 0.7050227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 4.841649494390352e-06,
-      "loss": 0.8087,
+      "learning_rate": 4.227028320972832e-06,
+      "loss": 0.7675,
       "step": 24845
     },
     {
-      "epoch": 0.6823762049929967,
+      "epoch": 0.7050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 4.840887477171753e-06,
-      "loss": 0.9374,
+      "learning_rate": 4.226277886748245e-06,
+      "loss": 0.8352,
       "step": 24846
     },
     {
-      "epoch": 0.6824036692208398,
+      "epoch": 0.7050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 4.840125500774047e-06,
-      "loss": 0.9246,
+      "learning_rate": 4.225527501294548e-06,
+      "loss": 0.7336,
       "step": 24847
     },
     {
-      "epoch": 0.6824311334486831,
+      "epoch": 0.7051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 4.839363565203266e-06,
-      "loss": 0.899,
+      "learning_rate": 4.224777164618084e-06,
+      "loss": 0.723,
       "step": 24848
     },
     {
-      "epoch": 0.6824585976765263,
+      "epoch": 0.7051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 4.8386016704654404e-06,
-      "loss": 0.8088,
+      "learning_rate": 4.2240268767251815e-06,
+      "loss": 0.824,
       "step": 24849
     },
     {
-      "epoch": 0.6824860619043696,
+      "epoch": 0.7051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 4.837839816566598e-06,
-      "loss": 0.7909,
+      "learning_rate": 4.223276637622184e-06,
+      "loss": 0.8238,
       "step": 24850
     },
     {
-      "epoch": 0.6825135261322128,
+      "epoch": 0.7051929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 4.83707800351277e-06,
-      "loss": 0.8753,
+      "learning_rate": 4.222526447315432e-06,
+      "loss": 0.8593,
       "step": 24851
     },
     {
-      "epoch": 0.682540990360056,
+      "epoch": 0.7052213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 4.836316231309981e-06,
-      "loss": 0.7377,
+      "learning_rate": 4.221776305811256e-06,
+      "loss": 0.8347,
       "step": 24852
     },
     {
-      "epoch": 0.6825684545878993,
+      "epoch": 0.7052497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 4.835554499964253e-06,
-      "loss": 0.8709,
+      "learning_rate": 4.221026213115995e-06,
+      "loss": 0.8464,
       "step": 24853
     },
     {
-      "epoch": 0.6825959188157424,
+      "epoch": 0.7052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 4.83479280948162e-06,
-      "loss": 0.8717,
+      "learning_rate": 4.22027616923599e-06,
+      "loss": 0.7955,
       "step": 24854
     },
     {
-      "epoch": 0.6826233830435857,
+      "epoch": 0.7053064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 4.834031159868107e-06,
-      "loss": 0.7876,
+      "learning_rate": 4.219526174177566e-06,
+      "loss": 0.8325,
       "step": 24855
     },
     {
-      "epoch": 0.682650847271429,
+      "epoch": 0.7053348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 4.83326955112974e-06,
-      "loss": 0.7048,
+      "learning_rate": 4.2187762279470654e-06,
+      "loss": 0.8123,
       "step": 24856
     },
     {
-      "epoch": 0.6826783114992722,
+      "epoch": 0.7053632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 4.832507983272545e-06,
-      "loss": 0.8162,
+      "learning_rate": 4.218026330550826e-06,
+      "loss": 0.7885,
       "step": 24857
     },
     {
-      "epoch": 0.6827057757271154,
+      "epoch": 0.7053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 4.831746456302553e-06,
-      "loss": 0.9086,
+      "learning_rate": 4.217276481995175e-06,
+      "loss": 0.7289,
       "step": 24858
     },
     {
-      "epoch": 0.6827332399549587,
+      "epoch": 0.7054199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 4.830984970225782e-06,
-      "loss": 0.8635,
+      "learning_rate": 4.216526682286448e-06,
+      "loss": 0.749,
       "step": 24859
     },
     {
-      "epoch": 0.6827607041828019,
+      "epoch": 0.7054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 4.8302235250482636e-06,
-      "loss": 0.772,
+      "learning_rate": 4.2157769314309846e-06,
+      "loss": 0.7544,
       "step": 24860
     },
     {
-      "epoch": 0.6827881684106452,
+      "epoch": 0.7054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 4.829462120776015e-06,
-      "loss": 0.8737,
+      "learning_rate": 4.2150272294351105e-06,
+      "loss": 0.8022,
       "step": 24861
     },
     {
-      "epoch": 0.6828156326384883,
+      "epoch": 0.705505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 4.828700757415065e-06,
-      "loss": 0.8409,
+      "learning_rate": 4.21427757630516e-06,
+      "loss": 0.8228,
       "step": 24862
     },
     {
-      "epoch": 0.6828430968663316,
+      "epoch": 0.705533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 4.827939434971438e-06,
-      "loss": 0.8238,
+      "learning_rate": 4.2135279720474675e-06,
+      "loss": 0.7549,
       "step": 24863
     },
     {
-      "epoch": 0.6828705610941749,
+      "epoch": 0.7055618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 4.827178153451161e-06,
-      "loss": 0.8141,
+      "learning_rate": 4.212778416668364e-06,
+      "loss": 0.8918,
       "step": 24864
     },
     {
-      "epoch": 0.682898025322018,
+      "epoch": 0.7055902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 4.826416912860251e-06,
-      "loss": 0.8236,
+      "learning_rate": 4.212028910174186e-06,
+      "loss": 0.8555,
       "step": 24865
     },
     {
-      "epoch": 0.6829254895498613,
+      "epoch": 0.7056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 4.825655713204735e-06,
-      "loss": 0.8332,
+      "learning_rate": 4.211279452571255e-06,
+      "loss": 0.7519,
       "step": 24866
     },
     {
-      "epoch": 0.6829529537777045,
+      "epoch": 0.7056469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 4.824894554490638e-06,
-      "loss": 0.8076,
+      "learning_rate": 4.210530043865908e-06,
+      "loss": 0.7874,
       "step": 24867
     },
     {
-      "epoch": 0.6829804180055478,
+      "epoch": 0.7056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 4.824133436723976e-06,
-      "loss": 0.7114,
+      "learning_rate": 4.2097806840644776e-06,
+      "loss": 0.8879,
       "step": 24868
     },
     {
-      "epoch": 0.683007882233391,
+      "epoch": 0.7057037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 4.823372359910776e-06,
-      "loss": 0.8426,
+      "learning_rate": 4.209031373173284e-06,
+      "loss": 0.8129,
       "step": 24869
     },
     {
-      "epoch": 0.6830353464612342,
+      "epoch": 0.7057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 4.822611324057062e-06,
-      "loss": 0.8264,
+      "learning_rate": 4.208282111198666e-06,
+      "loss": 0.8369,
       "step": 24870
     },
     {
-      "epoch": 0.6830628106890775,
+      "epoch": 0.7057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 4.82185032916885e-06,
-      "loss": 0.8709,
+      "learning_rate": 4.207532898146951e-06,
+      "loss": 0.7519,
       "step": 24871
     },
     {
-      "epoch": 0.6830902749169208,
+      "epoch": 0.705788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.821089375252164e-06,
-      "loss": 0.7612,
+      "learning_rate": 4.206783734024463e-06,
+      "loss": 0.7334,
       "step": 24872
     },
     {
-      "epoch": 0.6831177391447639,
+      "epoch": 0.7058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 4.820328462313024e-06,
-      "loss": 0.8085,
+      "learning_rate": 4.2060346188375335e-06,
+      "loss": 0.8439,
       "step": 24873
     },
     {
-      "epoch": 0.6831452033726072,
+      "epoch": 0.7058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 4.819567590357452e-06,
-      "loss": 0.7597,
+      "learning_rate": 4.205285552592493e-06,
+      "loss": 0.7225,
       "step": 24874
     },
     {
-      "epoch": 0.6831726676004504,
+      "epoch": 0.7058740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 4.818806759391472e-06,
-      "loss": 0.8779,
+      "learning_rate": 4.204536535295662e-06,
+      "loss": 0.8448,
       "step": 24875
     },
     {
-      "epoch": 0.6832001318282936,
+      "epoch": 0.7059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 4.8180459694211e-06,
-      "loss": 0.8179,
+      "learning_rate": 4.203787566953372e-06,
+      "loss": 0.8283,
       "step": 24876
     },
     {
-      "epoch": 0.6832275960561369,
+      "epoch": 0.7059307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 4.817285220452351e-06,
-      "loss": 0.8621,
+      "learning_rate": 4.20303864757195e-06,
+      "loss": 0.901,
       "step": 24877
     },
     {
-      "epoch": 0.6832550602839801,
+      "epoch": 0.7059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 4.81652451249125e-06,
-      "loss": 0.8645,
+      "learning_rate": 4.20228977715772e-06,
+      "loss": 0.861,
       "step": 24878
     },
     {
-      "epoch": 0.6832825245118234,
+      "epoch": 0.7059875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 4.815763845543815e-06,
-      "loss": 0.9395,
+      "learning_rate": 4.201540955717012e-06,
+      "loss": 0.8726,
       "step": 24879
     },
     {
-      "epoch": 0.6833099887396665,
+      "epoch": 0.7060158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 4.8150032196160636e-06,
-      "loss": 0.8079,
+      "learning_rate": 4.200792183256145e-06,
+      "loss": 0.8374,
       "step": 24880
     },
     {
-      "epoch": 0.6833374529675098,
+      "epoch": 0.706044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 4.814242634714016e-06,
-      "loss": 0.8658,
+      "learning_rate": 4.200043459781448e-06,
+      "loss": 0.8159,
       "step": 24881
     },
     {
-      "epoch": 0.6833649171953531,
+      "epoch": 0.7060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 4.813482090843693e-06,
-      "loss": 0.8445,
+      "learning_rate": 4.199294785299247e-06,
+      "loss": 0.8303,
       "step": 24882
     },
     {
-      "epoch": 0.6833923814231962,
+      "epoch": 0.7061010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 4.8127215880111085e-06,
-      "loss": 0.9011,
+      "learning_rate": 4.19854615981586e-06,
+      "loss": 0.8046,
       "step": 24883
     },
     {
-      "epoch": 0.6834198456510395,
+      "epoch": 0.7061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 4.811961126222276e-06,
-      "loss": 0.9602,
+      "learning_rate": 4.197797583337616e-06,
+      "loss": 0.811,
       "step": 24884
     },
     {
-      "epoch": 0.6834473098788828,
+      "epoch": 0.7061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.811200705483218e-06,
-      "loss": 0.7636,
+      "learning_rate": 4.197049055870838e-06,
+      "loss": 0.8981,
       "step": 24885
     },
     {
-      "epoch": 0.683474774106726,
+      "epoch": 0.7061861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 4.810440325799949e-06,
-      "loss": 0.8303,
+      "learning_rate": 4.196300577421847e-06,
+      "loss": 0.6969,
       "step": 24886
     },
     {
-      "epoch": 0.6835022383345692,
+      "epoch": 0.7062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 4.809679987178487e-06,
-      "loss": 0.7876,
+      "learning_rate": 4.195552147996963e-06,
+      "loss": 0.7833,
       "step": 24887
     },
     {
-      "epoch": 0.6835297025624124,
+      "epoch": 0.7062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 4.80891968962485e-06,
-      "loss": 0.8115,
+      "learning_rate": 4.1948037676025156e-06,
+      "loss": 0.7277,
       "step": 24888
     },
     {
-      "epoch": 0.6835571667902557,
+      "epoch": 0.7062712826333711,
       "grad_norm": 0.0,
-      "learning_rate": 4.808159433145049e-06,
-      "loss": 0.8209,
+      "learning_rate": 4.194055436244818e-06,
+      "loss": 0.8657,
       "step": 24889
     },
     {
-      "epoch": 0.683584631018099,
+      "epoch": 0.7062996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 4.807399217745101e-06,
-      "loss": 0.847,
+      "learning_rate": 4.1933071539301965e-06,
+      "loss": 0.7457,
       "step": 24890
     },
     {
-      "epoch": 0.6836120952459421,
+      "epoch": 0.706328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 4.806639043431026e-06,
-      "loss": 0.8417,
+      "learning_rate": 4.192558920664972e-06,
+      "loss": 0.7506,
       "step": 24891
     },
     {
-      "epoch": 0.6836395594737854,
+      "epoch": 0.7063564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 4.80587891020883e-06,
-      "loss": 0.9247,
+      "learning_rate": 4.1918107364554575e-06,
+      "loss": 0.7736,
       "step": 24892
     },
     {
-      "epoch": 0.6836670237016286,
+      "epoch": 0.7063847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 4.805118818084533e-06,
-      "loss": 0.8583,
+      "learning_rate": 4.191062601307984e-06,
+      "loss": 0.8056,
       "step": 24893
     },
     {
-      "epoch": 0.6836944879294719,
+      "epoch": 0.7064131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 4.804358767064151e-06,
-      "loss": 0.8467,
+      "learning_rate": 4.190314515228865e-06,
+      "loss": 0.9357,
       "step": 24894
     },
     {
-      "epoch": 0.6837219521573151,
+      "epoch": 0.7064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 4.803598757153692e-06,
-      "loss": 0.7414,
+      "learning_rate": 4.189566478224419e-06,
+      "loss": 0.83,
       "step": 24895
     },
     {
-      "epoch": 0.6837494163851583,
+      "epoch": 0.7064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 4.802838788359173e-06,
-      "loss": 0.8604,
+      "learning_rate": 4.18881849030097e-06,
+      "loss": 0.7952,
       "step": 24896
     },
     {
-      "epoch": 0.6837768806130016,
+      "epoch": 0.7064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 4.802078860686605e-06,
-      "loss": 0.8022,
+      "learning_rate": 4.188070551464829e-06,
+      "loss": 0.8801,
       "step": 24897
     },
     {
-      "epoch": 0.6838043448408448,
+      "epoch": 0.7065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 4.801318974142004e-06,
-      "loss": 0.8673,
+      "learning_rate": 4.187322661722317e-06,
+      "loss": 0.938,
       "step": 24898
     },
     {
-      "epoch": 0.683831809068688,
+      "epoch": 0.7065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 4.800559128731384e-06,
-      "loss": 0.9436,
+      "learning_rate": 4.186574821079755e-06,
+      "loss": 0.7955,
       "step": 24899
     },
     {
-      "epoch": 0.6838592732965313,
+      "epoch": 0.706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 4.799799324460755e-06,
-      "loss": 0.9553,
+      "learning_rate": 4.185827029543454e-06,
+      "loss": 0.8246,
       "step": 24900
     },
     {
-      "epoch": 0.6838867375243745,
+      "epoch": 0.7066118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 4.799039561336124e-06,
-      "loss": 0.8847,
+      "learning_rate": 4.185079287119733e-06,
+      "loss": 0.9884,
       "step": 24901
     },
     {
-      "epoch": 0.6839142017522177,
+      "epoch": 0.7066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 4.798279839363506e-06,
-      "loss": 0.8282,
+      "learning_rate": 4.184331593814913e-06,
+      "loss": 0.8181,
       "step": 24902
     },
     {
-      "epoch": 0.683941665980061,
+      "epoch": 0.7066685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 4.797520158548914e-06,
-      "loss": 0.8163,
+      "learning_rate": 4.183583949635301e-06,
+      "loss": 0.8907,
       "step": 24903
     },
     {
-      "epoch": 0.6839691302079042,
+      "epoch": 0.7066969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 4.796760518898358e-06,
-      "loss": 0.9365,
+      "learning_rate": 4.182836354587218e-06,
+      "loss": 0.9069,
       "step": 24904
     },
     {
-      "epoch": 0.6839965944357475,
+      "epoch": 0.7067253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 4.796000920417852e-06,
-      "loss": 0.8884,
+      "learning_rate": 4.18208880867698e-06,
+      "loss": 0.8701,
       "step": 24905
     },
     {
-      "epoch": 0.6840240586635907,
+      "epoch": 0.7067536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 4.795241363113398e-06,
-      "loss": 0.8025,
+      "learning_rate": 4.181341311910897e-06,
+      "loss": 0.7074,
       "step": 24906
     },
     {
-      "epoch": 0.6840515228914339,
+      "epoch": 0.7067820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 4.794481846991013e-06,
-      "loss": 0.9309,
+      "learning_rate": 4.180593864295285e-06,
+      "loss": 0.8113,
       "step": 24907
     },
     {
-      "epoch": 0.6840789871192772,
+      "epoch": 0.7068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 4.793722372056702e-06,
-      "loss": 0.822,
+      "learning_rate": 4.1798464658364566e-06,
+      "loss": 0.771,
       "step": 24908
     },
     {
-      "epoch": 0.6841064513471203,
+      "epoch": 0.706838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 4.792962938316477e-06,
-      "loss": 0.8144,
+      "learning_rate": 4.179099116540729e-06,
+      "loss": 0.8236,
       "step": 24909
     },
     {
-      "epoch": 0.6841339155749636,
+      "epoch": 0.7068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 4.792203545776344e-06,
-      "loss": 0.8623,
+      "learning_rate": 4.178351816414415e-06,
+      "loss": 0.7687,
       "step": 24910
     },
     {
-      "epoch": 0.6841613798028069,
+      "epoch": 0.7068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 4.791444194442319e-06,
-      "loss": 0.9067,
+      "learning_rate": 4.177604565463822e-06,
+      "loss": 0.7803,
       "step": 24911
     },
     {
-      "epoch": 0.6841888440306501,
+      "epoch": 0.7069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 4.790684884320402e-06,
-      "loss": 0.7681,
+      "learning_rate": 4.176857363695266e-06,
+      "loss": 0.7634,
       "step": 24912
     },
     {
-      "epoch": 0.6842163082584933,
+      "epoch": 0.7069523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 4.789925615416603e-06,
-      "loss": 0.8209,
+      "learning_rate": 4.17611021111506e-06,
+      "loss": 0.7208,
       "step": 24913
     },
     {
-      "epoch": 0.6842437724863365,
+      "epoch": 0.7069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 4.789166387736931e-06,
-      "loss": 0.7681,
+      "learning_rate": 4.175363107729509e-06,
+      "loss": 0.8798,
       "step": 24914
     },
     {
-      "epoch": 0.6842712367141798,
+      "epoch": 0.7070090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 4.7884072012873965e-06,
-      "loss": 0.8768,
+      "learning_rate": 4.174616053544928e-06,
+      "loss": 0.7421,
       "step": 24915
     },
     {
-      "epoch": 0.684298700942023,
+      "epoch": 0.7070374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 4.787648056073999e-06,
-      "loss": 0.8022,
+      "learning_rate": 4.17386904856763e-06,
+      "loss": 0.863,
       "step": 24916
     },
     {
-      "epoch": 0.6843261651698662,
+      "epoch": 0.7070658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 4.786888952102754e-06,
-      "loss": 0.8195,
+      "learning_rate": 4.17312209280392e-06,
+      "loss": 0.742,
       "step": 24917
     },
     {
-      "epoch": 0.6843536293977095,
+      "epoch": 0.7070942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 4.786129889379658e-06,
-      "loss": 0.8166,
+      "learning_rate": 4.172375186260108e-06,
+      "loss": 0.9367,
       "step": 24918
     },
     {
-      "epoch": 0.6843810936255528,
+      "epoch": 0.7071225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 4.785370867910723e-06,
-      "loss": 0.8878,
+      "learning_rate": 4.17162832894251e-06,
+      "loss": 0.8701,
       "step": 24919
     },
     {
-      "epoch": 0.6844085578533959,
+      "epoch": 0.7071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 4.7846118877019535e-06,
-      "loss": 0.9279,
+      "learning_rate": 4.170881520857425e-06,
+      "loss": 0.8361,
       "step": 24920
     },
     {
-      "epoch": 0.6844360220812392,
+      "epoch": 0.7071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 4.783852948759354e-06,
-      "loss": 0.8616,
+      "learning_rate": 4.170134762011165e-06,
+      "loss": 0.8498,
       "step": 24921
     },
     {
-      "epoch": 0.6844634863090824,
+      "epoch": 0.7072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 4.783094051088936e-06,
-      "loss": 0.8179,
+      "learning_rate": 4.169388052410044e-06,
+      "loss": 0.9369,
       "step": 24922
     },
     {
-      "epoch": 0.6844909505369257,
+      "epoch": 0.7072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 4.782335194696693e-06,
-      "loss": 0.8068,
+      "learning_rate": 4.168641392060357e-06,
+      "loss": 0.8333,
       "step": 24923
     },
     {
-      "epoch": 0.6845184147647689,
+      "epoch": 0.7072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 4.7815763795886405e-06,
-      "loss": 0.8125,
+      "learning_rate": 4.16789478096842e-06,
+      "loss": 0.7742,
       "step": 24924
     },
     {
-      "epoch": 0.6845458789926121,
+      "epoch": 0.7072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 4.780817605770772e-06,
-      "loss": 0.8101,
+      "learning_rate": 4.167148219140543e-06,
+      "loss": 0.8277,
       "step": 24925
     },
     {
-      "epoch": 0.6845733432204554,
+      "epoch": 0.7073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 4.780058873249098e-06,
-      "loss": 0.8627,
+      "learning_rate": 4.1664017065830235e-06,
+      "loss": 0.8214,
       "step": 24926
     },
     {
-      "epoch": 0.6846008074482985,
+      "epoch": 0.707349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 4.779300182029619e-06,
-      "loss": 0.8619,
+      "learning_rate": 4.16565524330217e-06,
+      "loss": 0.842,
       "step": 24927
     },
     {
-      "epoch": 0.6846282716761418,
+      "epoch": 0.707377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 4.778541532118339e-06,
-      "loss": 0.9272,
+      "learning_rate": 4.1649088293042935e-06,
+      "loss": 0.8428,
       "step": 24928
     },
     {
-      "epoch": 0.6846557359039851,
+      "epoch": 0.7074063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 4.777782923521266e-06,
-      "loss": 0.8906,
+      "learning_rate": 4.164162464595691e-06,
+      "loss": 0.9395,
       "step": 24929
     },
     {
-      "epoch": 0.6846832001318283,
+      "epoch": 0.7074347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 4.777024356244393e-06,
-      "loss": 0.7367,
+      "learning_rate": 4.163416149182674e-06,
+      "loss": 0.9489,
       "step": 24930
     },
     {
-      "epoch": 0.6847106643596715,
+      "epoch": 0.7074631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 4.776265830293729e-06,
-      "loss": 0.7724,
+      "learning_rate": 4.16266988307154e-06,
+      "loss": 0.9139,
       "step": 24931
     },
     {
-      "epoch": 0.6847381285875148,
+      "epoch": 0.7074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 4.775507345675271e-06,
-      "loss": 0.784,
+      "learning_rate": 4.161923666268595e-06,
+      "loss": 0.8734,
       "step": 24932
     },
     {
-      "epoch": 0.684765592815358,
+      "epoch": 0.7075198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 4.774748902395022e-06,
-      "loss": 0.7693,
+      "learning_rate": 4.1611774987801465e-06,
+      "loss": 0.9036,
       "step": 24933
     },
     {
-      "epoch": 0.6847930570432013,
+      "epoch": 0.7075482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 4.7739905004589845e-06,
-      "loss": 0.7493,
+      "learning_rate": 4.160431380612492e-06,
+      "loss": 0.7336,
       "step": 24934
     },
     {
-      "epoch": 0.6848205212710444,
+      "epoch": 0.7075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 4.7732321398731615e-06,
-      "loss": 0.7825,
+      "learning_rate": 4.159685311771935e-06,
+      "loss": 0.8104,
       "step": 24935
     },
     {
-      "epoch": 0.6848479854988877,
+      "epoch": 0.7076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 4.772473820643548e-06,
-      "loss": 0.8933,
+      "learning_rate": 4.1589392922647816e-06,
+      "loss": 0.9355,
       "step": 24936
     },
     {
-      "epoch": 0.684875449726731,
+      "epoch": 0.707633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 4.771715542776144e-06,
-      "loss": 0.7877,
+      "learning_rate": 4.158193322097328e-06,
+      "loss": 0.7704,
       "step": 24937
     },
     {
-      "epoch": 0.6849029139545741,
+      "epoch": 0.7076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 4.770957306276954e-06,
-      "loss": 0.8763,
+      "learning_rate": 4.157447401275875e-06,
+      "loss": 0.8474,
       "step": 24938
     },
     {
-      "epoch": 0.6849303781824174,
+      "epoch": 0.7076901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 4.770199111151978e-06,
-      "loss": 0.8634,
+      "learning_rate": 4.156701529806732e-06,
+      "loss": 0.8632,
       "step": 24939
     },
     {
-      "epoch": 0.6849578424102606,
+      "epoch": 0.7077185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 4.76944095740721e-06,
-      "loss": 0.7688,
+      "learning_rate": 4.155955707696192e-06,
+      "loss": 0.8551,
       "step": 24940
     },
     {
-      "epoch": 0.6849853066381039,
+      "epoch": 0.7077468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 4.7686828450486545e-06,
-      "loss": 0.7972,
+      "learning_rate": 4.155209934950556e-06,
+      "loss": 0.8226,
       "step": 24941
     },
     {
-      "epoch": 0.6850127708659471,
+      "epoch": 0.7077752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 4.767924774082302e-06,
-      "loss": 0.838,
+      "learning_rate": 4.154464211576128e-06,
+      "loss": 0.8504,
       "step": 24942
     },
     {
-      "epoch": 0.6850402350937903,
+      "epoch": 0.7078036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 4.767166744514157e-06,
-      "loss": 0.864,
+      "learning_rate": 4.1537185375792e-06,
+      "loss": 0.8326,
       "step": 24943
     },
     {
-      "epoch": 0.6850676993216336,
+      "epoch": 0.7078320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 4.766408756350214e-06,
-      "loss": 0.8549,
+      "learning_rate": 4.152972912966075e-06,
+      "loss": 0.7945,
       "step": 24944
     },
     {
-      "epoch": 0.6850951635494769,
+      "epoch": 0.7078603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 4.7656508095964735e-06,
-      "loss": 0.7849,
+      "learning_rate": 4.152227337743053e-06,
+      "loss": 0.8759,
       "step": 24945
     },
     {
-      "epoch": 0.68512262777732,
+      "epoch": 0.70788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.764892904258934e-06,
-      "loss": 0.8148,
+      "learning_rate": 4.151481811916427e-06,
+      "loss": 0.9427,
       "step": 24946
     },
     {
-      "epoch": 0.6851500920051633,
+      "epoch": 0.7079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 4.764135040343587e-06,
-      "loss": 0.8464,
+      "learning_rate": 4.150736335492496e-06,
+      "loss": 0.8049,
       "step": 24947
     },
     {
-      "epoch": 0.6851775562330065,
+      "epoch": 0.7079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 4.763377217856435e-06,
-      "loss": 0.9023,
+      "learning_rate": 4.149990908477564e-06,
+      "loss": 0.816,
       "step": 24948
     },
     {
-      "epoch": 0.6852050204608497,
+      "epoch": 0.7079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 4.762619436803467e-06,
-      "loss": 0.8815,
+      "learning_rate": 4.1492455308779156e-06,
+      "loss": 0.9193,
       "step": 24949
     },
     {
-      "epoch": 0.685232484688693,
+      "epoch": 0.7080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 4.761861697190684e-06,
-      "loss": 0.8155,
+      "learning_rate": 4.148500202699854e-06,
+      "loss": 0.7766,
       "step": 24950
     },
     {
-      "epoch": 0.6852599489165362,
+      "epoch": 0.7080306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 4.761103999024079e-06,
-      "loss": 0.8213,
+      "learning_rate": 4.1477549239496785e-06,
+      "loss": 0.8555,
       "step": 24951
     },
     {
-      "epoch": 0.6852874131443795,
+      "epoch": 0.7080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 4.7603463423096496e-06,
-      "loss": 0.8871,
+      "learning_rate": 4.147009694633676e-06,
+      "loss": 0.7796,
       "step": 24952
     },
     {
-      "epoch": 0.6853148773722226,
+      "epoch": 0.7080874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 4.759588727053392e-06,
-      "loss": 0.9236,
+      "learning_rate": 4.1462645147581456e-06,
+      "loss": 0.8418,
       "step": 24953
     },
     {
-      "epoch": 0.6853423416000659,
+      "epoch": 0.7081157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 4.758831153261295e-06,
-      "loss": 0.8454,
+      "learning_rate": 4.145519384329383e-06,
+      "loss": 0.889,
       "step": 24954
     },
     {
-      "epoch": 0.6853698058279092,
+      "epoch": 0.7081441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 4.758073620939361e-06,
-      "loss": 0.7012,
+      "learning_rate": 4.1447743033536805e-06,
+      "loss": 0.829,
       "step": 24955
     },
     {
-      "epoch": 0.6853972700557524,
+      "epoch": 0.7081725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 4.757316130093574e-06,
-      "loss": 0.885,
+      "learning_rate": 4.1440292718373366e-06,
+      "loss": 0.8264,
       "step": 24956
     },
     {
-      "epoch": 0.6854247342835956,
+      "epoch": 0.7082009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 4.7565586807299336e-06,
-      "loss": 0.7341,
+      "learning_rate": 4.143284289786637e-06,
+      "loss": 0.7478,
       "step": 24957
     },
     {
-      "epoch": 0.6854521985114389,
+      "epoch": 0.7082292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 4.755801272854431e-06,
-      "loss": 0.844,
+      "learning_rate": 4.142539357207877e-06,
+      "loss": 0.9053,
       "step": 24958
     },
     {
-      "epoch": 0.6854796627392821,
+      "epoch": 0.7082576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 4.7550439064730645e-06,
-      "loss": 0.8616,
+      "learning_rate": 4.1417944741073555e-06,
+      "loss": 0.8149,
       "step": 24959
     },
     {
-      "epoch": 0.6855071269671253,
+      "epoch": 0.7082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 4.754286581591818e-06,
-      "loss": 0.9367,
+      "learning_rate": 4.141049640491356e-06,
+      "loss": 0.8606,
       "step": 24960
     },
     {
-      "epoch": 0.6855345911949685,
+      "epoch": 0.7083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 4.753529298216689e-06,
-      "loss": 0.8589,
+      "learning_rate": 4.140304856366172e-06,
+      "loss": 0.8357,
       "step": 24961
     },
     {
-      "epoch": 0.6855620554228118,
+      "epoch": 0.7083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.752772056353668e-06,
-      "loss": 0.802,
+      "learning_rate": 4.139560121738101e-06,
+      "loss": 0.7704,
       "step": 24962
     },
     {
-      "epoch": 0.6855895196506551,
+      "epoch": 0.7083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 4.752014856008751e-06,
-      "loss": 0.8041,
+      "learning_rate": 4.138815436613426e-06,
+      "loss": 0.847,
       "step": 24963
     },
     {
-      "epoch": 0.6856169838784982,
+      "epoch": 0.708399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 4.751257697187925e-06,
-      "loss": 0.7921,
+      "learning_rate": 4.1380708009984394e-06,
+      "loss": 0.8876,
       "step": 24964
     },
     {
-      "epoch": 0.6856444481063415,
+      "epoch": 0.708427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.750500579897178e-06,
-      "loss": 0.7562,
+      "learning_rate": 4.137326214899436e-06,
+      "loss": 0.7992,
       "step": 24965
     },
     {
-      "epoch": 0.6856719123341847,
+      "epoch": 0.7084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 4.749743504142503e-06,
-      "loss": 0.8278,
+      "learning_rate": 4.1365816783226985e-06,
+      "loss": 0.8831,
       "step": 24966
     },
     {
-      "epoch": 0.685699376562028,
+      "epoch": 0.7084846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 4.748986469929893e-06,
-      "loss": 0.8259,
+      "learning_rate": 4.1358371912745185e-06,
+      "loss": 0.7676,
       "step": 24967
     },
     {
-      "epoch": 0.6857268407898712,
+      "epoch": 0.7085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 4.748229477265334e-06,
-      "loss": 0.8401,
+      "learning_rate": 4.1350927537611894e-06,
+      "loss": 0.9089,
       "step": 24968
     },
     {
-      "epoch": 0.6857543050177144,
+      "epoch": 0.7085414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 4.7474725261548185e-06,
-      "loss": 0.7798,
+      "learning_rate": 4.134348365788988e-06,
+      "loss": 0.9698,
       "step": 24969
     },
     {
-      "epoch": 0.6857817692455577,
+      "epoch": 0.7085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 4.746715616604339e-06,
-      "loss": 0.8,
+      "learning_rate": 4.133604027364217e-06,
+      "loss": 0.8127,
       "step": 24970
     },
     {
-      "epoch": 0.685809233473401,
+      "epoch": 0.7085981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 4.745958748619876e-06,
-      "loss": 0.8266,
+      "learning_rate": 4.132859738493154e-06,
+      "loss": 0.817,
       "step": 24971
     },
     {
-      "epoch": 0.6858366977012441,
+      "epoch": 0.7086265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 4.745201922207427e-06,
-      "loss": 0.9044,
+      "learning_rate": 4.132115499182088e-06,
+      "loss": 0.8345,
       "step": 24972
     },
     {
-      "epoch": 0.6858641619290874,
+      "epoch": 0.7086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 4.744445137372972e-06,
-      "loss": 0.8616,
+      "learning_rate": 4.131371309437309e-06,
+      "loss": 0.7084,
       "step": 24973
     },
     {
-      "epoch": 0.6858916261569306,
+      "epoch": 0.708683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.743688394122503e-06,
-      "loss": 0.8622,
+      "learning_rate": 4.130627169265096e-06,
+      "loss": 0.8259,
       "step": 24974
     },
     {
-      "epoch": 0.6859190903847738,
+      "epoch": 0.7087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.7429316924620075e-06,
-      "loss": 0.8007,
+      "learning_rate": 4.129883078671741e-06,
+      "loss": 0.8481,
       "step": 24975
     },
     {
-      "epoch": 0.6859465546126171,
+      "epoch": 0.7087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 4.742175032397477e-06,
-      "loss": 0.9507,
+      "learning_rate": 4.12913903766353e-06,
+      "loss": 0.8958,
       "step": 24976
     },
     {
-      "epoch": 0.6859740188404603,
+      "epoch": 0.7087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 4.7414184139348896e-06,
-      "loss": 0.8662,
+      "learning_rate": 4.1283950462467426e-06,
+      "loss": 0.8887,
       "step": 24977
     },
     {
-      "epoch": 0.6860014830683036,
+      "epoch": 0.7087968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 4.7406618370802375e-06,
-      "loss": 0.7704,
+      "learning_rate": 4.127651104427666e-06,
+      "loss": 0.9391,
       "step": 24978
     },
     {
-      "epoch": 0.6860289472961467,
+      "epoch": 0.7088251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 4.739905301839509e-06,
-      "loss": 0.8052,
+      "learning_rate": 4.126907212212587e-06,
+      "loss": 0.8145,
       "step": 24979
     },
     {
-      "epoch": 0.68605641152399,
+      "epoch": 0.7088535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 4.7391488082186845e-06,
-      "loss": 0.7524,
+      "learning_rate": 4.126163369607784e-06,
+      "loss": 0.9265,
       "step": 24980
     },
     {
-      "epoch": 0.6860838757518333,
+      "epoch": 0.7088819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 4.738392356223752e-06,
-      "loss": 0.9083,
+      "learning_rate": 4.125419576619544e-06,
+      "loss": 0.7635,
       "step": 24981
     },
     {
-      "epoch": 0.6861113399796764,
+      "epoch": 0.7089103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 4.737635945860701e-06,
-      "loss": 0.9073,
+      "learning_rate": 4.124675833254152e-06,
+      "loss": 0.8701,
       "step": 24982
     },
     {
-      "epoch": 0.6861388042075197,
+      "epoch": 0.708938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 4.73687957713551e-06,
-      "loss": 0.7896,
+      "learning_rate": 4.123932139517882e-06,
+      "loss": 0.8071,
       "step": 24983
     },
     {
-      "epoch": 0.686166268435363,
+      "epoch": 0.7089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 4.736123250054166e-06,
-      "loss": 0.8795,
+      "learning_rate": 4.123188495417023e-06,
+      "loss": 0.8021,
       "step": 24984
     },
     {
-      "epoch": 0.6861937326632062,
+      "epoch": 0.7089954597048809,
       "grad_norm": 0.0,
-      "learning_rate": 4.735366964622653e-06,
-      "loss": 0.8261,
+      "learning_rate": 4.122444900957855e-06,
+      "loss": 0.822,
       "step": 24985
     },
     {
-      "epoch": 0.6862211968910494,
+      "epoch": 0.7090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 4.734610720846957e-06,
-      "loss": 0.823,
+      "learning_rate": 4.121701356146659e-06,
+      "loss": 0.8543,
       "step": 24986
     },
     {
-      "epoch": 0.6862486611188926,
+      "epoch": 0.7090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 4.733854518733064e-06,
-      "loss": 0.7761,
+      "learning_rate": 4.120957860989719e-06,
+      "loss": 0.7998,
       "step": 24987
     },
     {
-      "epoch": 0.6862761253467359,
+      "epoch": 0.7090805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 4.733098358286954e-06,
-      "loss": 0.7909,
+      "learning_rate": 4.120214415493309e-06,
+      "loss": 0.815,
       "step": 24988
     },
     {
-      "epoch": 0.6863035895745792,
+      "epoch": 0.7091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 4.732342239514606e-06,
-      "loss": 0.7775,
+      "learning_rate": 4.119471019663712e-06,
+      "loss": 0.8285,
       "step": 24989
     },
     {
-      "epoch": 0.6863310538024223,
+      "epoch": 0.7091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 4.731586162422007e-06,
-      "loss": 0.916,
+      "learning_rate": 4.118727673507213e-06,
+      "loss": 0.8381,
       "step": 24990
     },
     {
-      "epoch": 0.6863585180302656,
+      "epoch": 0.7091657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 4.730830127015139e-06,
-      "loss": 0.9096,
+      "learning_rate": 4.11798437703008e-06,
+      "loss": 0.8866,
       "step": 24991
     },
     {
-      "epoch": 0.6863859822581088,
+      "epoch": 0.7091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 4.730074133299984e-06,
-      "loss": 0.764,
+      "learning_rate": 4.117241130238597e-06,
+      "loss": 0.7555,
       "step": 24992
     },
     {
-      "epoch": 0.686413446485952,
+      "epoch": 0.70922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 4.729318181282524e-06,
-      "loss": 0.8557,
+      "learning_rate": 4.1164979331390476e-06,
+      "loss": 0.7908,
       "step": 24993
     },
     {
-      "epoch": 0.6864409107137953,
+      "epoch": 0.7092508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 4.728562270968744e-06,
-      "loss": 0.8796,
+      "learning_rate": 4.1157547857377e-06,
+      "loss": 0.7733,
       "step": 24994
     },
     {
-      "epoch": 0.6864683749416385,
+      "epoch": 0.7092792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 4.727806402364617e-06,
-      "loss": 0.8614,
+      "learning_rate": 4.115011688040838e-06,
+      "loss": 0.8806,
       "step": 24995
     },
     {
-      "epoch": 0.6864958391694818,
+      "epoch": 0.7093076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 4.727050575476133e-06,
-      "loss": 0.8163,
+      "learning_rate": 4.114268640054738e-06,
+      "loss": 0.6926,
       "step": 24996
     },
     {
-      "epoch": 0.686523303397325,
+      "epoch": 0.7093359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 4.726294790309263e-06,
-      "loss": 0.8196,
+      "learning_rate": 4.113525641785673e-06,
+      "loss": 0.7792,
       "step": 24997
     },
     {
-      "epoch": 0.6865507676251682,
+      "epoch": 0.7093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 4.72553904686999e-06,
-      "loss": 0.7875,
+      "learning_rate": 4.112782693239922e-06,
+      "loss": 0.9111,
       "step": 24998
     },
     {
-      "epoch": 0.6865782318530115,
+      "epoch": 0.7093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 4.7247833451642965e-06,
-      "loss": 0.7324,
+      "learning_rate": 4.112039794423764e-06,
+      "loss": 0.8698,
       "step": 24999
     },
     {
-      "epoch": 0.6866056960808546,
+      "epoch": 0.7094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.724027685198165e-06,
-      "loss": 0.846,
+      "learning_rate": 4.111296945343462e-06,
+      "loss": 0.871,
       "step": 25000
     },
     {
-      "epoch": 0.6866331603086979,
+      "epoch": 0.7094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 4.723272066977565e-06,
-      "loss": 0.8647,
+      "learning_rate": 4.110554146005307e-06,
+      "loss": 0.8495,
       "step": 25001
     },
     {
-      "epoch": 0.6866606245365412,
+      "epoch": 0.709477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 4.722516490508482e-06,
-      "loss": 0.8766,
+      "learning_rate": 4.109811396415563e-06,
+      "loss": 0.84,
       "step": 25002
     },
     {
-      "epoch": 0.6866880887643844,
+      "epoch": 0.7095062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 4.7217609557968956e-06,
-      "loss": 0.8756,
+      "learning_rate": 4.109068696580506e-06,
+      "loss": 0.8145,
       "step": 25003
     },
     {
-      "epoch": 0.6867155529922276,
+      "epoch": 0.7095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 4.721005462848778e-06,
-      "loss": 0.8535,
+      "learning_rate": 4.108326046506415e-06,
+      "loss": 0.8193,
       "step": 25004
     },
     {
-      "epoch": 0.6867430172200708,
+      "epoch": 0.7095629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 4.7202500116701085e-06,
-      "loss": 0.905,
+      "learning_rate": 4.107583446199555e-06,
+      "loss": 0.8616,
       "step": 25005
     },
     {
-      "epoch": 0.6867704814479141,
+      "epoch": 0.7095913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 4.7194946022668704e-06,
-      "loss": 0.9427,
+      "learning_rate": 4.1068408956662024e-06,
+      "loss": 0.8307,
       "step": 25006
     },
     {
-      "epoch": 0.6867979456757574,
+      "epoch": 0.7096197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 4.718739234645034e-06,
-      "loss": 0.758,
+      "learning_rate": 4.106098394912632e-06,
+      "loss": 0.9084,
       "step": 25007
     },
     {
-      "epoch": 0.6868254099036005,
+      "epoch": 0.7096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 4.717983908810578e-06,
-      "loss": 0.8079,
+      "learning_rate": 4.105355943945112e-06,
+      "loss": 0.8344,
       "step": 25008
     },
     {
-      "epoch": 0.6868528741314438,
+      "epoch": 0.7096765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 4.717228624769479e-06,
-      "loss": 0.7559,
+      "learning_rate": 4.104613542769914e-06,
+      "loss": 0.9044,
       "step": 25009
     },
     {
-      "epoch": 0.6868803383592871,
+      "epoch": 0.7097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 4.716473382527713e-06,
-      "loss": 0.891,
+      "learning_rate": 4.103871191393314e-06,
+      "loss": 0.8746,
       "step": 25010
     },
     {
-      "epoch": 0.6869078025871302,
+      "epoch": 0.709733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 4.7157181820912616e-06,
-      "loss": 0.8693,
+      "learning_rate": 4.103128889821576e-06,
+      "loss": 0.7872,
       "step": 25011
     },
     {
-      "epoch": 0.6869352668149735,
+      "epoch": 0.7097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 4.714963023466093e-06,
-      "loss": 0.8177,
+      "learning_rate": 4.102386638060974e-06,
+      "loss": 0.7577,
       "step": 25012
     },
     {
-      "epoch": 0.6869627310428167,
+      "epoch": 0.7097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 4.714207906658181e-06,
-      "loss": 0.876,
+      "learning_rate": 4.101644436117779e-06,
+      "loss": 0.8409,
       "step": 25013
     },
     {
-      "epoch": 0.68699019527066,
+      "epoch": 0.7098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 4.713452831673504e-06,
-      "loss": 0.8372,
+      "learning_rate": 4.100902283998256e-06,
+      "loss": 0.8752,
       "step": 25014
     },
     {
-      "epoch": 0.6870176594985032,
+      "epoch": 0.7098467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 4.7126977985180365e-06,
-      "loss": 0.9138,
+      "learning_rate": 4.1001601817086765e-06,
+      "loss": 0.8092,
       "step": 25015
     },
     {
-      "epoch": 0.6870451237263464,
+      "epoch": 0.7098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 4.711942807197751e-06,
-      "loss": 0.8184,
+      "learning_rate": 4.099418129255309e-06,
+      "loss": 0.8465,
       "step": 25016
     },
     {
-      "epoch": 0.6870725879541897,
+      "epoch": 0.7099035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 4.711187857718623e-06,
-      "loss": 0.7835,
+      "learning_rate": 4.098676126644422e-06,
+      "loss": 0.8301,
       "step": 25017
     },
     {
-      "epoch": 0.6871000521820328,
+      "epoch": 0.7099318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 4.71043295008663e-06,
-      "loss": 0.9023,
+      "learning_rate": 4.097934173882287e-06,
+      "loss": 0.8137,
       "step": 25018
     },
     {
-      "epoch": 0.6871275164098761,
+      "epoch": 0.7099602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 4.709678084307735e-06,
-      "loss": 0.9438,
+      "learning_rate": 4.097192270975163e-06,
+      "loss": 0.8657,
       "step": 25019
     },
     {
-      "epoch": 0.6871549806377194,
+      "epoch": 0.7099886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 4.708923260387921e-06,
-      "loss": 0.823,
+      "learning_rate": 4.096450417929322e-06,
+      "loss": 0.8166,
       "step": 25020
     },
     {
-      "epoch": 0.6871824448655626,
+      "epoch": 0.7100170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 4.708168478333152e-06,
-      "loss": 0.8147,
+      "learning_rate": 4.0957086147510325e-06,
+      "loss": 0.8995,
       "step": 25021
     },
     {
-      "epoch": 0.6872099090934058,
+      "epoch": 0.7100454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 4.707413738149404e-06,
-      "loss": 0.7722,
+      "learning_rate": 4.094966861446555e-06,
+      "loss": 0.9897,
       "step": 25022
     },
     {
-      "epoch": 0.6872373733212491,
+      "epoch": 0.7100737797956868,
       "grad_norm": 0.0,
-      "learning_rate": 4.706659039842652e-06,
-      "loss": 0.8674,
+      "learning_rate": 4.094225158022157e-06,
+      "loss": 0.8005,
       "step": 25023
     },
     {
-      "epoch": 0.6872648375490923,
+      "epoch": 0.7101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 4.705904383418859e-06,
-      "loss": 0.8051,
+      "learning_rate": 4.093483504484109e-06,
+      "loss": 0.7685,
       "step": 25024
     },
     {
-      "epoch": 0.6872923017769356,
+      "epoch": 0.7101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 4.705149768884003e-06,
-      "loss": 0.8405,
+      "learning_rate": 4.092741900838667e-06,
+      "loss": 0.8885,
       "step": 25025
     },
     {
-      "epoch": 0.6873197660047787,
+      "epoch": 0.7101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 4.704395196244052e-06,
-      "loss": 0.8656,
+      "learning_rate": 4.0920003470921e-06,
+      "loss": 0.8173,
       "step": 25026
     },
     {
-      "epoch": 0.687347230232622,
+      "epoch": 0.7101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 4.703640665504983e-06,
-      "loss": 0.7847,
+      "learning_rate": 4.091258843250675e-06,
+      "loss": 0.7865,
       "step": 25027
     },
     {
-      "epoch": 0.6873746944604653,
+      "epoch": 0.7102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 4.702886176672755e-06,
-      "loss": 0.8199,
+      "learning_rate": 4.09051738932065e-06,
+      "loss": 0.8946,
       "step": 25028
     },
     {
-      "epoch": 0.6874021586883085,
+      "epoch": 0.710244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 4.702131729753349e-06,
-      "loss": 0.8405,
+      "learning_rate": 4.0897759853082885e-06,
+      "loss": 0.856,
       "step": 25029
     },
     {
-      "epoch": 0.6874296229161517,
+      "epoch": 0.710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 4.701377324752724e-06,
-      "loss": 0.7946,
+      "learning_rate": 4.089034631219856e-06,
+      "loss": 0.9408,
       "step": 25030
     },
     {
-      "epoch": 0.6874570871439949,
+      "epoch": 0.7103007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 4.700622961676854e-06,
-      "loss": 0.8738,
+      "learning_rate": 4.0882933270616135e-06,
+      "loss": 0.8546,
       "step": 25031
     },
     {
-      "epoch": 0.6874845513718382,
+      "epoch": 0.7103291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 4.699868640531708e-06,
-      "loss": 0.8298,
+      "learning_rate": 4.087552072839822e-06,
+      "loss": 0.8521,
       "step": 25032
     },
     {
-      "epoch": 0.6875120155996814,
+      "epoch": 0.7103575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 4.699114361323256e-06,
-      "loss": 0.8691,
+      "learning_rate": 4.086810868560749e-06,
+      "loss": 0.8624,
       "step": 25033
     },
     {
-      "epoch": 0.6875394798275246,
+      "epoch": 0.7103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 4.698360124057463e-06,
-      "loss": 0.786,
+      "learning_rate": 4.086069714230646e-06,
+      "loss": 0.9126,
       "step": 25034
     },
     {
-      "epoch": 0.6875669440553679,
+      "epoch": 0.7104143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 4.697605928740302e-06,
-      "loss": 0.9417,
+      "learning_rate": 4.085328609855777e-06,
+      "loss": 0.7354,
       "step": 25035
     },
     {
-      "epoch": 0.6875944082832112,
+      "epoch": 0.7104426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 4.696851775377738e-06,
-      "loss": 0.7706,
+      "learning_rate": 4.084587555442408e-06,
+      "loss": 0.9144,
       "step": 25036
     },
     {
-      "epoch": 0.6876218725110543,
+      "epoch": 0.7104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 4.6960976639757315e-06,
-      "loss": 0.8733,
+      "learning_rate": 4.08384655099679e-06,
+      "loss": 0.8441,
       "step": 25037
     },
     {
-      "epoch": 0.6876493367388976,
+      "epoch": 0.7104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.6953435945402555e-06,
-      "loss": 0.8405,
+      "learning_rate": 4.083105596525191e-06,
+      "loss": 0.7558,
       "step": 25038
     },
     {
-      "epoch": 0.6876768009667408,
+      "epoch": 0.710527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 4.694589567077276e-06,
-      "loss": 0.8276,
+      "learning_rate": 4.08236469203386e-06,
+      "loss": 0.848,
       "step": 25039
     },
     {
-      "epoch": 0.687704265194584,
+      "epoch": 0.7105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 4.69383558159276e-06,
-      "loss": 0.8845,
+      "learning_rate": 4.081623837529061e-06,
+      "loss": 0.9035,
       "step": 25040
     },
     {
-      "epoch": 0.6877317294224273,
+      "epoch": 0.7105845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 4.693081638092674e-06,
-      "loss": 0.7761,
+      "learning_rate": 4.080883033017055e-06,
+      "loss": 0.8059,
       "step": 25041
     },
     {
-      "epoch": 0.6877591936502705,
+      "epoch": 0.7106129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 4.69232773658298e-06,
-      "loss": 0.9334,
+      "learning_rate": 4.080142278504091e-06,
+      "loss": 0.8228,
       "step": 25042
     },
     {
-      "epoch": 0.6877866578781138,
+      "epoch": 0.7106413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 4.691573877069647e-06,
-      "loss": 0.8716,
+      "learning_rate": 4.079401573996433e-06,
+      "loss": 0.848,
       "step": 25043
     },
     {
-      "epoch": 0.6878141221059569,
+      "epoch": 0.7106696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 4.690820059558635e-06,
-      "loss": 0.8208,
+      "learning_rate": 4.078660919500339e-06,
+      "loss": 0.8895,
       "step": 25044
     },
     {
-      "epoch": 0.6878415863338002,
+      "epoch": 0.7106980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 4.690066284055912e-06,
-      "loss": 0.9056,
+      "learning_rate": 4.077920315022059e-06,
+      "loss": 0.7785,
       "step": 25045
     },
     {
-      "epoch": 0.6878690505616435,
+      "epoch": 0.7107264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 4.689312550567439e-06,
-      "loss": 0.8522,
+      "learning_rate": 4.077179760567848e-06,
+      "loss": 0.8332,
       "step": 25046
     },
     {
-      "epoch": 0.6878965147894867,
+      "epoch": 0.7107548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 4.688558859099188e-06,
-      "loss": 0.8509,
+      "learning_rate": 4.076439256143975e-06,
+      "loss": 0.8132,
       "step": 25047
     },
     {
-      "epoch": 0.6879239790173299,
+      "epoch": 0.710783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 4.687805209657113e-06,
-      "loss": 0.7918,
+      "learning_rate": 4.075698801756681e-06,
+      "loss": 0.8192,
       "step": 25048
     },
     {
-      "epoch": 0.6879514432451732,
+      "epoch": 0.7108115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 4.68705160224718e-06,
-      "loss": 0.8343,
+      "learning_rate": 4.074958397412225e-06,
+      "loss": 0.8303,
       "step": 25049
     },
     {
-      "epoch": 0.6879789074730164,
+      "epoch": 0.7108399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 4.686298036875354e-06,
-      "loss": 0.8516,
+      "learning_rate": 4.074218043116866e-06,
+      "loss": 0.8387,
       "step": 25050
     },
     {
-      "epoch": 0.6880063717008597,
+      "epoch": 0.7108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.685544513547599e-06,
-      "loss": 0.775,
+      "learning_rate": 4.07347773887685e-06,
+      "loss": 0.8103,
       "step": 25051
     },
     {
-      "epoch": 0.6880338359287028,
+      "epoch": 0.7108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 4.684791032269871e-06,
-      "loss": 0.8719,
+      "learning_rate": 4.072737484698435e-06,
+      "loss": 0.8107,
       "step": 25052
     },
     {
-      "epoch": 0.6880613001565461,
+      "epoch": 0.7109250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 4.6840375930481384e-06,
-      "loss": 0.8036,
+      "learning_rate": 4.071997280587875e-06,
+      "loss": 0.7479,
       "step": 25053
     },
     {
-      "epoch": 0.6880887643843894,
+      "epoch": 0.7109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 4.683284195888357e-06,
-      "loss": 0.8556,
+      "learning_rate": 4.071257126551417e-06,
+      "loss": 0.7209,
       "step": 25054
     },
     {
-      "epoch": 0.6881162286122325,
+      "epoch": 0.7109818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 4.68253084079649e-06,
-      "loss": 0.8473,
+      "learning_rate": 4.070517022595316e-06,
+      "loss": 0.8438,
       "step": 25055
     },
     {
-      "epoch": 0.6881436928400758,
+      "epoch": 0.7110102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 4.681777527778498e-06,
-      "loss": 0.771,
+      "learning_rate": 4.069776968725829e-06,
+      "loss": 0.8491,
       "step": 25056
     },
     {
-      "epoch": 0.688171157067919,
+      "epoch": 0.7110385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 4.681024256840343e-06,
-      "loss": 0.7269,
+      "learning_rate": 4.069036964949199e-06,
+      "loss": 0.923,
       "step": 25057
     },
     {
-      "epoch": 0.6881986212957623,
+      "epoch": 0.711066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 4.680271027987985e-06,
-      "loss": 0.7367,
+      "learning_rate": 4.06829701127168e-06,
+      "loss": 0.8665,
       "step": 25058
     },
     {
-      "epoch": 0.6882260855236055,
+      "epoch": 0.7110953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 4.6795178412273865e-06,
-      "loss": 0.8449,
+      "learning_rate": 4.067557107699526e-06,
+      "loss": 0.8187,
       "step": 25059
     },
     {
-      "epoch": 0.6882535497514487,
+      "epoch": 0.7111237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 4.678764696564504e-06,
-      "loss": 0.8385,
+      "learning_rate": 4.066817254238981e-06,
+      "loss": 0.8789,
       "step": 25060
     },
     {
-      "epoch": 0.688281013979292,
+      "epoch": 0.7111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 4.678011594005292e-06,
-      "loss": 0.7817,
+      "learning_rate": 4.0660774508962964e-06,
+      "loss": 0.8126,
       "step": 25061
     },
     {
-      "epoch": 0.6883084782071353,
+      "epoch": 0.7111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 4.677258533555714e-06,
-      "loss": 0.86,
+      "learning_rate": 4.065337697677723e-06,
+      "loss": 0.7914,
       "step": 25062
     },
     {
-      "epoch": 0.6883359424349784,
+      "epoch": 0.7112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 4.6765055152217285e-06,
-      "loss": 0.9576,
+      "learning_rate": 4.064597994589509e-06,
+      "loss": 0.8026,
       "step": 25063
     },
     {
-      "epoch": 0.6883634066628217,
+      "epoch": 0.7112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 4.675752539009294e-06,
-      "loss": 0.8304,
+      "learning_rate": 4.063858341637905e-06,
+      "loss": 0.7674,
       "step": 25064
     },
     {
-      "epoch": 0.6883908708906649,
+      "epoch": 0.7112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 4.674999604924372e-06,
-      "loss": 0.8216,
+      "learning_rate": 4.063118738829154e-06,
+      "loss": 0.8644,
       "step": 25065
     },
     {
-      "epoch": 0.6884183351185081,
+      "epoch": 0.711293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 4.674246712972911e-06,
-      "loss": 0.9061,
+      "learning_rate": 4.0623791861695065e-06,
+      "loss": 0.8011,
       "step": 25066
     },
     {
-      "epoch": 0.6884457993463514,
+      "epoch": 0.711322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 4.673493863160876e-06,
-      "loss": 0.9265,
+      "learning_rate": 4.061639683665211e-06,
+      "loss": 0.849,
       "step": 25067
     },
     {
-      "epoch": 0.6884732635741946,
+      "epoch": 0.7113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 4.672741055494219e-06,
-      "loss": 0.9421,
+      "learning_rate": 4.060900231322509e-06,
+      "loss": 0.7663,
       "step": 25068
     },
     {
-      "epoch": 0.6885007278020379,
+      "epoch": 0.7113791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 4.671988289978897e-06,
-      "loss": 0.8745,
+      "learning_rate": 4.0601608291476495e-06,
+      "loss": 0.8008,
       "step": 25069
     },
     {
-      "epoch": 0.688528192029881,
+      "epoch": 0.7114074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 4.6712355666208674e-06,
-      "loss": 0.853,
+      "learning_rate": 4.059421477146882e-06,
+      "loss": 0.8708,
       "step": 25070
     },
     {
-      "epoch": 0.6885556562577243,
+      "epoch": 0.7114358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 4.670482885426091e-06,
-      "loss": 0.9018,
+      "learning_rate": 4.058682175326446e-06,
+      "loss": 1.0414,
       "step": 25071
     },
     {
-      "epoch": 0.6885831204855676,
+      "epoch": 0.7114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 4.669730246400515e-06,
-      "loss": 0.8049,
+      "learning_rate": 4.057942923692587e-06,
+      "loss": 0.7423,
       "step": 25072
     },
     {
-      "epoch": 0.6886105847134107,
+      "epoch": 0.7114926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 4.668977649550097e-06,
-      "loss": 0.9003,
+      "learning_rate": 4.057203722251556e-06,
+      "loss": 0.9587,
       "step": 25073
     },
     {
-      "epoch": 0.688638048941254,
+      "epoch": 0.7115209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 4.6682250948807925e-06,
-      "loss": 0.8433,
+      "learning_rate": 4.056464571009589e-06,
+      "loss": 0.8287,
       "step": 25074
     },
     {
-      "epoch": 0.6886655131690973,
+      "epoch": 0.7115493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 4.667472582398561e-06,
-      "loss": 0.7355,
+      "learning_rate": 4.055725469972932e-06,
+      "loss": 0.8024,
       "step": 25075
     },
     {
-      "epoch": 0.6886929773969405,
+      "epoch": 0.711577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.666720112109348e-06,
-      "loss": 0.7773,
+      "learning_rate": 4.054986419147829e-06,
+      "loss": 0.8858,
       "step": 25076
     },
     {
-      "epoch": 0.6887204416247837,
+      "epoch": 0.711606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 4.665967684019115e-06,
-      "loss": 0.7691,
+      "learning_rate": 4.054247418540523e-06,
+      "loss": 0.8431,
       "step": 25077
     },
     {
-      "epoch": 0.6887479058526269,
+      "epoch": 0.7116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 4.665215298133808e-06,
-      "loss": 0.8249,
+      "learning_rate": 4.053508468157261e-06,
+      "loss": 0.8381,
       "step": 25078
     },
     {
-      "epoch": 0.6887753700804702,
+      "epoch": 0.7116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 4.664462954459386e-06,
-      "loss": 0.9016,
+      "learning_rate": 4.052769568004277e-06,
+      "loss": 0.8834,
       "step": 25079
     },
     {
-      "epoch": 0.6888028343083135,
+      "epoch": 0.7116912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 4.663710653001797e-06,
-      "loss": 0.8949,
+      "learning_rate": 4.052030718087814e-06,
+      "loss": 0.8175,
       "step": 25080
     },
     {
-      "epoch": 0.6888302985361566,
+      "epoch": 0.7117196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 4.662958393766998e-06,
-      "loss": 0.7504,
+      "learning_rate": 4.05129191841412e-06,
+      "loss": 0.8468,
       "step": 25081
     },
     {
-      "epoch": 0.6888577627639999,
+      "epoch": 0.7117480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 4.662206176760943e-06,
-      "loss": 0.8625,
+      "learning_rate": 4.050553168989426e-06,
+      "loss": 0.8906,
       "step": 25082
     },
     {
-      "epoch": 0.6888852269918431,
+      "epoch": 0.7117763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 4.661454001989577e-06,
-      "loss": 0.8485,
+      "learning_rate": 4.049814469819977e-06,
+      "loss": 0.8383,
       "step": 25083
     },
     {
-      "epoch": 0.6889126912196863,
+      "epoch": 0.7118047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 4.660701869458858e-06,
-      "loss": 0.8036,
+      "learning_rate": 4.049075820912016e-06,
+      "loss": 0.7393,
       "step": 25084
     },
     {
-      "epoch": 0.6889401554475296,
+      "epoch": 0.71183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.6599497791747305e-06,
-      "loss": 0.8561,
+      "learning_rate": 4.048337222271777e-06,
+      "loss": 0.8739,
       "step": 25085
     },
     {
-      "epoch": 0.6889676196753728,
+      "epoch": 0.7118615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 4.659197731143149e-06,
-      "loss": 0.8824,
+      "learning_rate": 4.047598673905499e-06,
+      "loss": 0.9004,
       "step": 25086
     },
     {
-      "epoch": 0.6889950839032161,
+      "epoch": 0.7118898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 4.658445725370064e-06,
-      "loss": 0.8729,
+      "learning_rate": 4.046860175819427e-06,
+      "loss": 0.871,
       "step": 25087
     },
     {
-      "epoch": 0.6890225481310593,
+      "epoch": 0.7119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 4.657693761861429e-06,
-      "loss": 0.8707,
+      "learning_rate": 4.046121728019792e-06,
+      "loss": 0.8117,
       "step": 25088
     },
     {
-      "epoch": 0.6890500123589025,
+      "epoch": 0.7119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 4.656941840623187e-06,
-      "loss": 0.7777,
+      "learning_rate": 4.045383330512833e-06,
+      "loss": 0.8866,
       "step": 25089
     },
     {
-      "epoch": 0.6890774765867458,
+      "epoch": 0.7119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 4.65618996166129e-06,
-      "loss": 0.9553,
+      "learning_rate": 4.0446449833047925e-06,
+      "loss": 0.9099,
       "step": 25090
     },
     {
-      "epoch": 0.689104940814589,
+      "epoch": 0.7120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 4.655438124981693e-06,
-      "loss": 0.8322,
+      "learning_rate": 4.0439066864019005e-06,
+      "loss": 0.7223,
       "step": 25091
     },
     {
-      "epoch": 0.6891324050424322,
+      "epoch": 0.7120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 4.654686330590335e-06,
-      "loss": 0.8705,
+      "learning_rate": 4.043168439810396e-06,
+      "loss": 0.7024,
       "step": 25092
     },
     {
-      "epoch": 0.6891598692702755,
+      "epoch": 0.7120601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 4.65393457849317e-06,
-      "loss": 0.7953,
+      "learning_rate": 4.042430243536517e-06,
+      "loss": 0.8858,
       "step": 25093
     },
     {
-      "epoch": 0.6891873334981187,
+      "epoch": 0.7120885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 4.653182868696149e-06,
-      "loss": 0.8065,
+      "learning_rate": 4.041692097586496e-06,
+      "loss": 0.8406,
       "step": 25094
     },
     {
-      "epoch": 0.689214797725962,
+      "epoch": 0.712116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.652431201205213e-06,
-      "loss": 0.8455,
+      "learning_rate": 4.040954001966574e-06,
+      "loss": 0.8754,
       "step": 25095
     },
     {
-      "epoch": 0.6892422619538052,
+      "epoch": 0.7121452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 4.6516795760263125e-06,
-      "loss": 0.7964,
+      "learning_rate": 4.040215956682977e-06,
+      "loss": 0.8561,
       "step": 25096
     },
     {
-      "epoch": 0.6892697261816484,
+      "epoch": 0.7121736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 4.650927993165395e-06,
-      "loss": 0.9007,
+      "learning_rate": 4.039477961741944e-06,
+      "loss": 0.7405,
       "step": 25097
     },
     {
-      "epoch": 0.6892971904094917,
+      "epoch": 0.7122020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 4.650176452628407e-06,
-      "loss": 0.8766,
+      "learning_rate": 4.038740017149713e-06,
+      "loss": 0.8844,
       "step": 25098
     },
     {
-      "epoch": 0.6893246546373348,
+      "epoch": 0.7122304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 4.6494249544212985e-06,
-      "loss": 0.8238,
+      "learning_rate": 4.038002122912509e-06,
+      "loss": 0.8354,
       "step": 25099
     },
     {
-      "epoch": 0.6893521188651781,
+      "epoch": 0.7122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 4.6486734985500135e-06,
-      "loss": 0.7747,
+      "learning_rate": 4.037264279036568e-06,
+      "loss": 0.8368,
       "step": 25100
     },
     {
-      "epoch": 0.6893795830930214,
+      "epoch": 0.7122871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 4.647922085020492e-06,
-      "loss": 0.902,
+      "learning_rate": 4.036526485528128e-06,
+      "loss": 0.9451,
       "step": 25101
     },
     {
-      "epoch": 0.6894070473208646,
+      "epoch": 0.7123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 4.647170713838685e-06,
-      "loss": 0.8306,
+      "learning_rate": 4.035788742393414e-06,
+      "loss": 0.7999,
       "step": 25102
     },
     {
-      "epoch": 0.6894345115487078,
+      "epoch": 0.7123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 4.646419385010536e-06,
-      "loss": 0.8075,
+      "learning_rate": 4.03505104963866e-06,
+      "loss": 0.7925,
       "step": 25103
     },
     {
-      "epoch": 0.689461975776551,
+      "epoch": 0.7123723041997729,
       "grad_norm": 0.0,
-      "learning_rate": 4.645668098541991e-06,
-      "loss": 0.8253,
+      "learning_rate": 4.0343134072701015e-06,
+      "loss": 0.7746,
       "step": 25104
     },
     {
-      "epoch": 0.6894894400043943,
+      "epoch": 0.7124006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 4.644916854438993e-06,
-      "loss": 0.8289,
+      "learning_rate": 4.033575815293962e-06,
+      "loss": 0.7188,
       "step": 25105
     },
     {
-      "epoch": 0.6895169042322375,
+      "epoch": 0.7124290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 4.644165652707493e-06,
-      "loss": 0.844,
+      "learning_rate": 4.032838273716476e-06,
+      "loss": 0.7766,
       "step": 25106
     },
     {
-      "epoch": 0.6895443684600807,
+      "epoch": 0.7124574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 4.643414493353424e-06,
-      "loss": 0.8734,
+      "learning_rate": 4.032100782543874e-06,
+      "loss": 0.7993,
       "step": 25107
     },
     {
-      "epoch": 0.689571832687924,
+      "epoch": 0.7124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 4.642663376382739e-06,
-      "loss": 0.8593,
+      "learning_rate": 4.031363341782385e-06,
+      "loss": 0.7372,
       "step": 25108
     },
     {
-      "epoch": 0.6895992969157673,
+      "epoch": 0.7125141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 4.641912301801371e-06,
-      "loss": 0.8261,
+      "learning_rate": 4.03062595143824e-06,
+      "loss": 0.9232,
       "step": 25109
     },
     {
-      "epoch": 0.6896267611436104,
+      "epoch": 0.7125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 4.641161269615271e-06,
-      "loss": 0.8335,
+      "learning_rate": 4.029888611517664e-06,
+      "loss": 0.9024,
       "step": 25110
     },
     {
-      "epoch": 0.6896542253714537,
+      "epoch": 0.7125709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 4.640410279830377e-06,
-      "loss": 0.8648,
+      "learning_rate": 4.029151322026887e-06,
+      "loss": 0.7658,
       "step": 25111
     },
     {
-      "epoch": 0.6896816895992969,
+      "epoch": 0.7125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 4.6396593324526376e-06,
-      "loss": 0.8204,
+      "learning_rate": 4.028414082972141e-06,
+      "loss": 0.8574,
       "step": 25112
     },
     {
-      "epoch": 0.6897091538271402,
+      "epoch": 0.712627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 4.638908427487986e-06,
-      "loss": 0.8374,
+      "learning_rate": 4.027676894359646e-06,
+      "loss": 0.862,
       "step": 25113
     },
     {
-      "epoch": 0.6897366180549834,
+      "epoch": 0.7126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 4.638157564942367e-06,
-      "loss": 0.8513,
+      "learning_rate": 4.026939756195632e-06,
+      "loss": 0.8022,
       "step": 25114
     },
     {
-      "epoch": 0.6897640822828266,
+      "epoch": 0.7126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 4.637406744821727e-06,
-      "loss": 0.8424,
+      "learning_rate": 4.0262026684863295e-06,
+      "loss": 0.7246,
       "step": 25115
     },
     {
-      "epoch": 0.6897915465106699,
+      "epoch": 0.7127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 4.6366559671319974e-06,
-      "loss": 0.8554,
+      "learning_rate": 4.025465631237959e-06,
+      "loss": 0.7937,
       "step": 25116
     },
     {
-      "epoch": 0.689819010738513,
+      "epoch": 0.7127412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 4.6359052318791245e-06,
-      "loss": 0.8135,
+      "learning_rate": 4.024728644456749e-06,
+      "loss": 0.8695,
       "step": 25117
     },
     {
-      "epoch": 0.6898464749663563,
+      "epoch": 0.7127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 4.63515453906905e-06,
-      "loss": 0.8659,
+      "learning_rate": 4.023991708148928e-06,
+      "loss": 0.8639,
       "step": 25118
     },
     {
-      "epoch": 0.6898739391941996,
+      "epoch": 0.7127979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 4.634403888707707e-06,
-      "loss": 0.8585,
+      "learning_rate": 4.0232548223207145e-06,
+      "loss": 0.8173,
       "step": 25119
     },
     {
-      "epoch": 0.6899014034220428,
+      "epoch": 0.7128263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 4.633653280801039e-06,
-      "loss": 0.8605,
+      "learning_rate": 4.022517986978336e-06,
+      "loss": 0.7706,
       "step": 25120
     },
     {
-      "epoch": 0.689928867649886,
+      "epoch": 0.7128547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 4.632902715354986e-06,
-      "loss": 0.8155,
+      "learning_rate": 4.02178120212802e-06,
+      "loss": 0.9403,
       "step": 25121
     },
     {
-      "epoch": 0.6899563318777293,
+      "epoch": 0.7128830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 4.632152192375485e-06,
-      "loss": 0.8854,
+      "learning_rate": 4.021044467775979e-06,
+      "loss": 0.884,
       "step": 25122
     },
     {
-      "epoch": 0.6899837961055725,
+      "epoch": 0.7129114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 4.631401711868479e-06,
-      "loss": 0.8801,
+      "learning_rate": 4.020307783928453e-06,
+      "loss": 0.8557,
       "step": 25123
     },
     {
-      "epoch": 0.6900112603334158,
+      "epoch": 0.7129398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 4.630651273839902e-06,
-      "loss": 0.8715,
+      "learning_rate": 4.019571150591653e-06,
+      "loss": 0.8105,
       "step": 25124
     },
     {
-      "epoch": 0.6900387245612589,
+      "epoch": 0.7129682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 4.6299008782956885e-06,
-      "loss": 0.8351,
+      "learning_rate": 4.018834567771802e-06,
+      "loss": 0.853,
       "step": 25125
     },
     {
-      "epoch": 0.6900661887891022,
+      "epoch": 0.7129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 4.62915052524178e-06,
-      "loss": 0.9158,
+      "learning_rate": 4.018098035475129e-06,
+      "loss": 0.7693,
       "step": 25126
     },
     {
-      "epoch": 0.6900936530169455,
+      "epoch": 0.7130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 4.628400214684113e-06,
-      "loss": 0.8429,
+      "learning_rate": 4.017361553707847e-06,
+      "loss": 0.9135,
       "step": 25127
     },
     {
-      "epoch": 0.6901211172447886,
+      "epoch": 0.7130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 4.627649946628623e-06,
-      "loss": 0.7743,
+      "learning_rate": 4.016625122476181e-06,
+      "loss": 0.6693,
       "step": 25128
     },
     {
-      "epoch": 0.6901485814726319,
+      "epoch": 0.7130817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 4.626899721081248e-06,
-      "loss": 0.7153,
+      "learning_rate": 4.015888741786355e-06,
+      "loss": 0.8734,
       "step": 25129
     },
     {
-      "epoch": 0.6901760457004751,
+      "epoch": 0.7131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 4.626149538047928e-06,
-      "loss": 0.9181,
+      "learning_rate": 4.015152411644583e-06,
+      "loss": 0.7366,
       "step": 25130
     },
     {
-      "epoch": 0.6902035099283184,
+      "epoch": 0.713138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 4.625399397534591e-06,
-      "loss": 0.8273,
+      "learning_rate": 4.014416132057086e-06,
+      "loss": 0.7929,
       "step": 25131
     },
     {
-      "epoch": 0.6902309741561616,
+      "epoch": 0.71316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.624649299547179e-06,
-      "loss": 0.7796,
+      "learning_rate": 4.01367990303009e-06,
+      "loss": 0.8354,
       "step": 25132
     },
     {
-      "epoch": 0.6902584383840048,
+      "epoch": 0.7131952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 4.623899244091621e-06,
-      "loss": 0.8947,
+      "learning_rate": 4.012943724569806e-06,
+      "loss": 0.8119,
       "step": 25133
     },
     {
-      "epoch": 0.6902859026118481,
+      "epoch": 0.7132236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 4.623149231173854e-06,
-      "loss": 0.8234,
+      "learning_rate": 4.012207596682454e-06,
+      "loss": 0.7793,
       "step": 25134
     },
     {
-      "epoch": 0.6903133668396914,
+      "epoch": 0.7132519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 4.622399260799814e-06,
-      "loss": 0.8068,
+      "learning_rate": 4.0114715193742574e-06,
+      "loss": 0.8181,
       "step": 25135
     },
     {
-      "epoch": 0.6903408310675345,
+      "epoch": 0.7132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 4.621649332975436e-06,
-      "loss": 0.8858,
+      "learning_rate": 4.010735492651427e-06,
+      "loss": 0.823,
       "step": 25136
     },
     {
-      "epoch": 0.6903682952953778,
+      "epoch": 0.7133087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 4.620899447706649e-06,
-      "loss": 0.7571,
+      "learning_rate": 4.0099995165201825e-06,
+      "loss": 0.9111,
       "step": 25137
     },
     {
-      "epoch": 0.690395759523221,
+      "epoch": 0.7133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 4.62014960499939e-06,
-      "loss": 0.8554,
+      "learning_rate": 4.009263590986742e-06,
+      "loss": 0.8845,
       "step": 25138
     },
     {
-      "epoch": 0.6904232237510642,
+      "epoch": 0.7133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 4.619399804859594e-06,
-      "loss": 0.876,
+      "learning_rate": 4.008527716057321e-06,
+      "loss": 0.7607,
       "step": 25139
     },
     {
-      "epoch": 0.6904506879789075,
+      "epoch": 0.7133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 4.618650047293189e-06,
-      "loss": 0.8434,
+      "learning_rate": 4.007791891738135e-06,
+      "loss": 0.9296,
       "step": 25140
     },
     {
-      "epoch": 0.6904781522067507,
+      "epoch": 0.713422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 4.617900332306111e-06,
-      "loss": 0.8371,
+      "learning_rate": 4.007056118035405e-06,
+      "loss": 0.8656,
       "step": 25141
     },
     {
-      "epoch": 0.690505616434594,
+      "epoch": 0.7134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 4.617150659904287e-06,
-      "loss": 0.8039,
+      "learning_rate": 4.006320394955337e-06,
+      "loss": 0.7912,
       "step": 25142
     },
     {
-      "epoch": 0.6905330806624371,
+      "epoch": 0.7134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 4.616401030093653e-06,
-      "loss": 0.8954,
+      "learning_rate": 4.00558472250415e-06,
+      "loss": 0.8477,
       "step": 25143
     },
     {
-      "epoch": 0.6905605448902804,
+      "epoch": 0.7135073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 4.615651442880139e-06,
-      "loss": 0.8922,
+      "learning_rate": 4.004849100688063e-06,
+      "loss": 0.6605,
       "step": 25144
     },
     {
-      "epoch": 0.6905880091181237,
+      "epoch": 0.7135357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 4.614901898269677e-06,
-      "loss": 0.8335,
+      "learning_rate": 4.0041135295132805e-06,
+      "loss": 0.8931,
       "step": 25145
     },
     {
-      "epoch": 0.6906154733459668,
+      "epoch": 0.7135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 4.614152396268196e-06,
-      "loss": 0.7472,
+      "learning_rate": 4.003378008986024e-06,
+      "loss": 0.8046,
       "step": 25146
     },
     {
-      "epoch": 0.6906429375738101,
+      "epoch": 0.7135925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 4.613402936881633e-06,
-      "loss": 0.7526,
+      "learning_rate": 4.0026425391125e-06,
+      "loss": 0.8295,
       "step": 25147
     },
     {
-      "epoch": 0.6906704018016534,
+      "epoch": 0.7136208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 4.61265352011591e-06,
-      "loss": 0.7494,
+      "learning_rate": 4.0019071198989246e-06,
+      "loss": 0.8263,
       "step": 25148
     },
     {
-      "epoch": 0.6906978660294966,
+      "epoch": 0.7136492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 4.611904145976958e-06,
-      "loss": 0.8988,
+      "learning_rate": 4.001171751351512e-06,
+      "loss": 0.8493,
       "step": 25149
     },
     {
-      "epoch": 0.6907253302573398,
+      "epoch": 0.713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 4.611154814470706e-06,
-      "loss": 0.8755,
+      "learning_rate": 4.000436433476468e-06,
+      "loss": 0.8467,
       "step": 25150
     },
     {
-      "epoch": 0.690752794485183,
+      "epoch": 0.7137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 4.610405525603085e-06,
-      "loss": 0.9513,
+      "learning_rate": 3.999701166280008e-06,
+      "loss": 0.8119,
       "step": 25151
     },
     {
-      "epoch": 0.6907802587130263,
+      "epoch": 0.7137343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 4.6096562793800234e-06,
-      "loss": 0.8825,
+      "learning_rate": 3.998965949768344e-06,
+      "loss": 0.8415,
       "step": 25152
     },
     {
-      "epoch": 0.6908077229408696,
+      "epoch": 0.7137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 4.608907075807453e-06,
-      "loss": 0.8613,
+      "learning_rate": 3.998230783947679e-06,
+      "loss": 0.8081,
       "step": 25153
     },
     {
-      "epoch": 0.6908351871687127,
+      "epoch": 0.7137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 4.608157914891294e-06,
-      "loss": 0.8125,
+      "learning_rate": 3.99749566882423e-06,
+      "loss": 0.8392,
       "step": 25154
     },
     {
-      "epoch": 0.690862651396556,
+      "epoch": 0.7138195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 4.607408796637478e-06,
-      "loss": 0.8317,
+      "learning_rate": 3.9967606044042094e-06,
+      "loss": 0.828,
       "step": 25155
     },
     {
-      "epoch": 0.6908901156243992,
+      "epoch": 0.7138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 4.606659721051937e-06,
-      "loss": 0.8567,
+      "learning_rate": 3.9960255906938185e-06,
+      "loss": 0.7581,
       "step": 25156
     },
     {
-      "epoch": 0.6909175798522424,
+      "epoch": 0.7138762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 4.6059106881405906e-06,
-      "loss": 0.8148,
+      "learning_rate": 3.9952906276992686e-06,
+      "loss": 0.8922,
       "step": 25157
     },
     {
-      "epoch": 0.6909450440800857,
+      "epoch": 0.7139046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 4.605161697909366e-06,
-      "loss": 0.8033,
+      "learning_rate": 3.9945557154267736e-06,
+      "loss": 0.7979,
       "step": 25158
     },
     {
-      "epoch": 0.6909725083079289,
+      "epoch": 0.7139330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 4.6044127503641964e-06,
-      "loss": 0.8745,
+      "learning_rate": 3.993820853882532e-06,
+      "loss": 0.7693,
       "step": 25159
     },
     {
-      "epoch": 0.6909999725357722,
+      "epoch": 0.713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 4.603663845511e-06,
-      "loss": 0.7976,
+      "learning_rate": 3.993086043072756e-06,
+      "loss": 0.7158,
       "step": 25160
     },
     {
-      "epoch": 0.6910274367636154,
+      "epoch": 0.7139897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 4.602914983355705e-06,
-      "loss": 0.7772,
+      "learning_rate": 3.992351283003655e-06,
+      "loss": 0.8416,
       "step": 25161
     },
     {
-      "epoch": 0.6910549009914586,
+      "epoch": 0.7140181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 4.602166163904238e-06,
-      "loss": 0.9148,
+      "learning_rate": 3.9916165736814294e-06,
+      "loss": 0.9009,
       "step": 25162
     },
     {
-      "epoch": 0.6910823652193019,
+      "epoch": 0.7140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 4.601417387162526e-06,
-      "loss": 0.9504,
+      "learning_rate": 3.990881915112289e-06,
+      "loss": 0.7649,
       "step": 25163
     },
     {
-      "epoch": 0.691109829447145,
+      "epoch": 0.7140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 4.600668653136489e-06,
-      "loss": 0.8829,
+      "learning_rate": 3.990147307302443e-06,
+      "loss": 0.7009,
       "step": 25164
     },
     {
-      "epoch": 0.6911372936749883,
+      "epoch": 0.7141032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 4.5999199618320545e-06,
-      "loss": 0.7997,
+      "learning_rate": 3.989412750258089e-06,
+      "loss": 0.7803,
       "step": 25165
     },
     {
-      "epoch": 0.6911647579028316,
+      "epoch": 0.7141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.599171313255143e-06,
-      "loss": 0.7999,
+      "learning_rate": 3.988678243985437e-06,
+      "loss": 0.8477,
       "step": 25166
     },
     {
-      "epoch": 0.6911922221306748,
+      "epoch": 0.7141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 4.598422707411679e-06,
-      "loss": 0.8355,
+      "learning_rate": 3.987943788490692e-06,
+      "loss": 0.9101,
       "step": 25167
     },
     {
-      "epoch": 0.691219686358518,
+      "epoch": 0.714188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 4.597674144307588e-06,
-      "loss": 0.7997,
+      "learning_rate": 3.987209383780048e-06,
+      "loss": 0.7694,
       "step": 25168
     },
     {
-      "epoch": 0.6912471505863612,
+      "epoch": 0.714216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 4.5969256239487915e-06,
-      "loss": 0.8526,
+      "learning_rate": 3.9864750298597255e-06,
+      "loss": 0.8063,
       "step": 25169
     },
     {
-      "epoch": 0.6912746148142045,
+      "epoch": 0.7142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 4.596177146341212e-06,
-      "loss": 0.8908,
+      "learning_rate": 3.985740726735915e-06,
+      "loss": 0.8345,
       "step": 25170
     },
     {
-      "epoch": 0.6913020790420478,
+      "epoch": 0.7142735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 4.595428711490776e-06,
-      "loss": 0.8219,
+      "learning_rate": 3.985006474414823e-06,
+      "loss": 0.8343,
       "step": 25171
     },
     {
-      "epoch": 0.6913295432698909,
+      "epoch": 0.7143019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 4.594680319403401e-06,
-      "loss": 0.7178,
+      "learning_rate": 3.984272272902655e-06,
+      "loss": 0.8813,
       "step": 25172
     },
     {
-      "epoch": 0.6913570074977342,
+      "epoch": 0.7143303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 4.5939319700850064e-06,
-      "loss": 0.8404,
+      "learning_rate": 3.9835381222056055e-06,
+      "loss": 0.902,
       "step": 25173
     },
     {
-      "epoch": 0.6913844717255775,
+      "epoch": 0.7143586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 4.593183663541516e-06,
-      "loss": 0.7824,
+      "learning_rate": 3.982804022329881e-06,
+      "loss": 0.7118,
       "step": 25174
     },
     {
-      "epoch": 0.6914119359534207,
+      "epoch": 0.7143870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 4.592435399778851e-06,
-      "loss": 0.8641,
+      "learning_rate": 3.982069973281685e-06,
+      "loss": 0.8214,
       "step": 25175
     },
     {
-      "epoch": 0.6914394001812639,
+      "epoch": 0.7144154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 4.591687178802931e-06,
-      "loss": 0.7533,
+      "learning_rate": 3.98133597506721e-06,
+      "loss": 0.8681,
       "step": 25176
     },
     {
-      "epoch": 0.6914668644091071,
+      "epoch": 0.7144438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 4.59093900061968e-06,
-      "loss": 0.8065,
+      "learning_rate": 3.980602027692661e-06,
+      "loss": 0.8637,
       "step": 25177
     },
     {
-      "epoch": 0.6914943286369504,
+      "epoch": 0.714472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 4.5901908652350134e-06,
-      "loss": 0.8094,
+      "learning_rate": 3.979868131164241e-06,
+      "loss": 0.833,
       "step": 25178
     },
     {
-      "epoch": 0.6915217928647936,
+      "epoch": 0.71450056753689,
       "grad_norm": 0.0,
-      "learning_rate": 4.589442772654851e-06,
-      "loss": 0.7628,
+      "learning_rate": 3.979134285488141e-06,
+      "loss": 0.8228,
       "step": 25179
     },
     {
-      "epoch": 0.6915492570926368,
+      "epoch": 0.7145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 4.588694722885117e-06,
-      "loss": 0.953,
+      "learning_rate": 3.978400490670565e-06,
+      "loss": 0.7511,
       "step": 25180
     },
     {
-      "epoch": 0.6915767213204801,
+      "epoch": 0.7145573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 4.587946715931723e-06,
-      "loss": 0.8325,
+      "learning_rate": 3.977666746717714e-06,
+      "loss": 0.878,
       "step": 25181
     },
     {
-      "epoch": 0.6916041855483233,
+      "epoch": 0.7145856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 4.587198751800591e-06,
-      "loss": 0.7254,
+      "learning_rate": 3.97693305363578e-06,
+      "loss": 0.7244,
       "step": 25182
     },
     {
-      "epoch": 0.6916316497761665,
+      "epoch": 0.7146140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 4.586450830497643e-06,
-      "loss": 0.8428,
+      "learning_rate": 3.976199411430962e-06,
+      "loss": 0.9389,
       "step": 25183
     },
     {
-      "epoch": 0.6916591140040098,
+      "epoch": 0.7146424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 4.585702952028789e-06,
-      "loss": 0.7306,
+      "learning_rate": 3.97546582010946e-06,
+      "loss": 0.7372,
       "step": 25184
     },
     {
-      "epoch": 0.691686578231853,
+      "epoch": 0.7146708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 4.584955116399951e-06,
-      "loss": 0.7814,
+      "learning_rate": 3.974732279677468e-06,
+      "loss": 0.887,
       "step": 25185
     },
     {
-      "epoch": 0.6917140424596963,
+      "epoch": 0.7146992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 4.5842073236170445e-06,
-      "loss": 0.7799,
+      "learning_rate": 3.973998790141187e-06,
+      "loss": 0.7823,
       "step": 25186
     },
     {
-      "epoch": 0.6917415066875395,
+      "epoch": 0.714727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 4.583459573685993e-06,
-      "loss": 0.867,
+      "learning_rate": 3.973265351506805e-06,
+      "loss": 0.8965,
       "step": 25187
     },
     {
-      "epoch": 0.6917689709153827,
+      "epoch": 0.7147559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 4.582711866612703e-06,
-      "loss": 0.9473,
+      "learning_rate": 3.972531963780523e-06,
+      "loss": 0.8834,
       "step": 25188
     },
     {
-      "epoch": 0.691796435143226,
+      "epoch": 0.7147843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 4.581964202403099e-06,
-      "loss": 0.8666,
+      "learning_rate": 3.971798626968536e-06,
+      "loss": 0.8266,
       "step": 25189
     },
     {
-      "epoch": 0.6918238993710691,
+      "epoch": 0.7148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 4.581216581063088e-06,
-      "loss": 0.8005,
+      "learning_rate": 3.971065341077035e-06,
+      "loss": 0.7808,
       "step": 25190
     },
     {
-      "epoch": 0.6918513635989124,
+      "epoch": 0.7148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 4.580469002598592e-06,
-      "loss": 0.833,
+      "learning_rate": 3.970332106112216e-06,
+      "loss": 0.8784,
       "step": 25191
     },
     {
-      "epoch": 0.6918788278267557,
+      "epoch": 0.7148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 4.579721467015524e-06,
-      "loss": 0.8864,
+      "learning_rate": 3.969598922080275e-06,
+      "loss": 0.742,
       "step": 25192
     },
     {
-      "epoch": 0.6919062920545989,
+      "epoch": 0.7148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 4.5789739743198e-06,
-      "loss": 0.8882,
+      "learning_rate": 3.9688657889874e-06,
+      "loss": 0.8686,
       "step": 25193
     },
     {
-      "epoch": 0.6919337562824421,
+      "epoch": 0.7149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 4.578226524517333e-06,
-      "loss": 0.8898,
+      "learning_rate": 3.968132706839788e-06,
+      "loss": 0.8053,
       "step": 25194
     },
     {
-      "epoch": 0.6919612205102853,
+      "epoch": 0.7149545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 4.577479117614042e-06,
-      "loss": 0.8491,
+      "learning_rate": 3.9673996756436325e-06,
+      "loss": 1.017,
       "step": 25195
     },
     {
-      "epoch": 0.6919886847381286,
+      "epoch": 0.7149829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 4.576731753615836e-06,
-      "loss": 0.8113,
+      "learning_rate": 3.96666669540512e-06,
+      "loss": 0.8528,
       "step": 25196
     },
     {
-      "epoch": 0.6920161489659719,
+      "epoch": 0.715011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 4.5759844325286275e-06,
-      "loss": 0.8863,
+      "learning_rate": 3.965933766130445e-06,
+      "loss": 0.9469,
       "step": 25197
     },
     {
-      "epoch": 0.692043613193815,
+      "epoch": 0.7150397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 4.575237154358329e-06,
-      "loss": 0.8438,
+      "learning_rate": 3.965200887825802e-06,
+      "loss": 0.7795,
       "step": 25198
     },
     {
-      "epoch": 0.6920710774216583,
+      "epoch": 0.7150681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 4.574489919110856e-06,
-      "loss": 0.771,
+      "learning_rate": 3.9644680604973715e-06,
+      "loss": 0.8525,
       "step": 25199
     },
     {
-      "epoch": 0.6920985416495016,
+      "epoch": 0.7150964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 4.573742726792121e-06,
-      "loss": 0.8701,
+      "learning_rate": 3.963735284151358e-06,
+      "loss": 0.8517,
       "step": 25200
     },
     {
-      "epoch": 0.6921260058773447,
+      "epoch": 0.7151248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 4.572995577408039e-06,
-      "loss": 0.7477,
+      "learning_rate": 3.96300255879394e-06,
+      "loss": 0.7611,
       "step": 25201
     },
     {
-      "epoch": 0.692153470105188,
+      "epoch": 0.7151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 4.572248470964515e-06,
-      "loss": 0.875,
+      "learning_rate": 3.9622698844313115e-06,
+      "loss": 0.9531,
       "step": 25202
     },
     {
-      "epoch": 0.6921809343330312,
+      "epoch": 0.7151816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 4.5715014074674655e-06,
-      "loss": 0.8949,
+      "learning_rate": 3.961537261069663e-06,
+      "loss": 0.843,
       "step": 25203
     },
     {
-      "epoch": 0.6922083985608745,
+      "epoch": 0.7152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 4.570754386922797e-06,
-      "loss": 0.8236,
+      "learning_rate": 3.960804688715178e-06,
+      "loss": 0.941,
       "step": 25204
     },
     {
-      "epoch": 0.6922358627887177,
+      "epoch": 0.7152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 4.570007409336424e-06,
-      "loss": 0.7753,
+      "learning_rate": 3.960072167374047e-06,
+      "loss": 0.8699,
       "step": 25205
     },
     {
-      "epoch": 0.6922633270165609,
+      "epoch": 0.7152667423382519,
       "grad_norm": 0.0,
-      "learning_rate": 4.569260474714259e-06,
-      "loss": 0.9035,
+      "learning_rate": 3.9593396970524625e-06,
+      "loss": 0.8165,
       "step": 25206
     },
     {
-      "epoch": 0.6922907912444042,
+      "epoch": 0.7152951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 4.568513583062204e-06,
-      "loss": 0.7902,
+      "learning_rate": 3.958607277756602e-06,
+      "loss": 0.876,
       "step": 25207
     },
     {
-      "epoch": 0.6923182554722473,
+      "epoch": 0.7153234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 4.5677667343861745e-06,
-      "loss": 0.8594,
+      "learning_rate": 3.957874909492658e-06,
+      "loss": 0.851,
       "step": 25208
     },
     {
-      "epoch": 0.6923457197000906,
+      "epoch": 0.7153518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 4.567019928692078e-06,
-      "loss": 0.9328,
+      "learning_rate": 3.9571425922668206e-06,
+      "loss": 0.8138,
       "step": 25209
     },
     {
-      "epoch": 0.6923731839279339,
+      "epoch": 0.7153802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 4.566273165985825e-06,
-      "loss": 0.8614,
+      "learning_rate": 3.956410326085267e-06,
+      "loss": 0.7497,
       "step": 25210
     },
     {
-      "epoch": 0.6924006481557771,
+      "epoch": 0.7154086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 4.5655264462733275e-06,
-      "loss": 0.9501,
+      "learning_rate": 3.9556781109541865e-06,
+      "loss": 0.9859,
       "step": 25211
     },
     {
-      "epoch": 0.6924281123836203,
+      "epoch": 0.7154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 4.564779769560489e-06,
-      "loss": 0.8241,
+      "learning_rate": 3.954945946879769e-06,
+      "loss": 0.7762,
       "step": 25212
     },
     {
-      "epoch": 0.6924555766114636,
+      "epoch": 0.7154653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 4.564033135853215e-06,
-      "loss": 0.7831,
+      "learning_rate": 3.954213833868191e-06,
+      "loss": 0.8129,
       "step": 25213
     },
     {
-      "epoch": 0.6924830408393068,
+      "epoch": 0.7154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 4.563286545157416e-06,
-      "loss": 0.8473,
+      "learning_rate": 3.953481771925641e-06,
+      "loss": 0.8972,
       "step": 25214
     },
     {
-      "epoch": 0.6925105050671501,
+      "epoch": 0.715522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 4.562539997479001e-06,
-      "loss": 0.8286,
+      "learning_rate": 3.952749761058302e-06,
+      "loss": 0.8886,
       "step": 25215
     },
     {
-      "epoch": 0.6925379692949932,
+      "epoch": 0.7155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 4.561793492823875e-06,
-      "loss": 0.9422,
+      "learning_rate": 3.952017801272358e-06,
+      "loss": 0.9041,
       "step": 25216
     },
     {
-      "epoch": 0.6925654335228365,
+      "epoch": 0.7155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.561047031197946e-06,
-      "loss": 0.8623,
+      "learning_rate": 3.951285892573994e-06,
+      "loss": 0.7469,
       "step": 25217
     },
     {
-      "epoch": 0.6925928977506798,
+      "epoch": 0.7156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 4.560300612607122e-06,
-      "loss": 0.8907,
+      "learning_rate": 3.9505540349693875e-06,
+      "loss": 0.8033,
       "step": 25218
     },
     {
-      "epoch": 0.692620361978523,
+      "epoch": 0.7156356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 4.5595542370573054e-06,
-      "loss": 0.835,
+      "learning_rate": 3.9498222284647225e-06,
+      "loss": 0.8786,
       "step": 25219
     },
     {
-      "epoch": 0.6926478262063662,
+      "epoch": 0.7156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 4.558807904554405e-06,
-      "loss": 0.8257,
+      "learning_rate": 3.949090473066185e-06,
+      "loss": 0.8046,
       "step": 25220
     },
     {
-      "epoch": 0.6926752904342094,
+      "epoch": 0.7156923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 4.558061615104321e-06,
-      "loss": 0.9907,
+      "learning_rate": 3.9483587687799485e-06,
+      "loss": 0.7987,
       "step": 25221
     },
     {
-      "epoch": 0.6927027546620527,
+      "epoch": 0.7157207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 4.557315368712961e-06,
-      "loss": 0.9155,
+      "learning_rate": 3.947627115612197e-06,
+      "loss": 0.8441,
       "step": 25222
     },
     {
-      "epoch": 0.6927302188898959,
+      "epoch": 0.7157491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 4.556569165386231e-06,
-      "loss": 0.8264,
+      "learning_rate": 3.946895513569116e-06,
+      "loss": 0.802,
       "step": 25223
     },
     {
-      "epoch": 0.6927576831177391,
+      "epoch": 0.7157775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 4.555823005130038e-06,
-      "loss": 0.8473,
+      "learning_rate": 3.946163962656876e-06,
+      "loss": 0.7275,
       "step": 25224
     },
     {
-      "epoch": 0.6927851473455824,
+      "epoch": 0.7158059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 4.555076887950278e-06,
-      "loss": 0.7393,
+      "learning_rate": 3.945432462881662e-06,
+      "loss": 0.8584,
       "step": 25225
     },
     {
-      "epoch": 0.6928126115734257,
+      "epoch": 0.7158342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 4.5543308138528584e-06,
-      "loss": 0.847,
+      "learning_rate": 3.9447010142496555e-06,
+      "loss": 0.8556,
       "step": 25226
     },
     {
-      "epoch": 0.6928400758012688,
+      "epoch": 0.7158626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 4.553584782843687e-06,
-      "loss": 0.8538,
+      "learning_rate": 3.943969616767027e-06,
+      "loss": 0.9222,
       "step": 25227
     },
     {
-      "epoch": 0.6928675400291121,
+      "epoch": 0.7158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 4.55283879492866e-06,
-      "loss": 0.7799,
+      "learning_rate": 3.943238270439961e-06,
+      "loss": 0.826,
       "step": 25228
     },
     {
-      "epoch": 0.6928950042569553,
+      "epoch": 0.7159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 4.552092850113682e-06,
-      "loss": 0.9062,
+      "learning_rate": 3.942506975274637e-06,
+      "loss": 0.7981,
       "step": 25229
     },
     {
-      "epoch": 0.6929224684847985,
+      "epoch": 0.7159477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 4.551346948404658e-06,
-      "loss": 0.8362,
+      "learning_rate": 3.941775731277221e-06,
+      "loss": 0.8837,
       "step": 25230
     },
     {
-      "epoch": 0.6929499327126418,
+      "epoch": 0.7159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 4.550601089807485e-06,
-      "loss": 0.8434,
+      "learning_rate": 3.941044538453905e-06,
+      "loss": 0.6938,
       "step": 25231
     },
     {
-      "epoch": 0.692977396940485,
+      "epoch": 0.7160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 4.5498552743280665e-06,
-      "loss": 0.7766,
+      "learning_rate": 3.940313396810855e-06,
+      "loss": 0.7471,
       "step": 25232
     },
     {
-      "epoch": 0.6930048611683283,
+      "epoch": 0.716032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 4.5491095019723056e-06,
-      "loss": 0.8221,
+      "learning_rate": 3.939582306354251e-06,
+      "loss": 0.8688,
       "step": 25233
     },
     {
-      "epoch": 0.6930323253961714,
+      "epoch": 0.716061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 4.5483637727460995e-06,
-      "loss": 0.8511,
+      "learning_rate": 3.938851267090269e-06,
+      "loss": 0.7667,
       "step": 25234
     },
     {
-      "epoch": 0.6930597896240147,
+      "epoch": 0.7160896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 4.547618086655356e-06,
-      "loss": 0.7747,
+      "learning_rate": 3.938120279025081e-06,
+      "loss": 0.7205,
       "step": 25235
     },
     {
-      "epoch": 0.693087253851858,
+      "epoch": 0.7161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 4.546872443705971e-06,
-      "loss": 0.8233,
+      "learning_rate": 3.9373893421648625e-06,
+      "loss": 0.9573,
       "step": 25236
     },
     {
-      "epoch": 0.6931147180797012,
+      "epoch": 0.7161464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 4.546126843903839e-06,
-      "loss": 0.9416,
+      "learning_rate": 3.936658456515793e-06,
+      "loss": 0.8881,
       "step": 25237
     },
     {
-      "epoch": 0.6931421823075444,
+      "epoch": 0.7161748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 4.545381287254864e-06,
-      "loss": 0.9114,
+      "learning_rate": 3.935927622084038e-06,
+      "loss": 0.803,
       "step": 25238
     },
     {
-      "epoch": 0.6931696465353877,
+      "epoch": 0.7162031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 4.544635773764945e-06,
-      "loss": 0.8787,
+      "learning_rate": 3.935196838875776e-06,
+      "loss": 0.811,
       "step": 25239
     },
     {
-      "epoch": 0.6931971107632309,
+      "epoch": 0.7162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 4.543890303439982e-06,
-      "loss": 0.9216,
+      "learning_rate": 3.934466106897181e-06,
+      "loss": 0.9278,
       "step": 25240
     },
     {
-      "epoch": 0.6932245749910741,
+      "epoch": 0.7162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 4.5431448762858706e-06,
-      "loss": 0.809,
+      "learning_rate": 3.933735426154421e-06,
+      "loss": 0.8925,
       "step": 25241
     },
     {
-      "epoch": 0.6932520392189173,
+      "epoch": 0.7162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 4.542399492308516e-06,
-      "loss": 0.7467,
+      "learning_rate": 3.933004796653671e-06,
+      "loss": 0.8005,
       "step": 25242
     },
     {
-      "epoch": 0.6932795034467606,
+      "epoch": 0.716316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.541654151513805e-06,
-      "loss": 0.7687,
+      "learning_rate": 3.932274218401104e-06,
+      "loss": 0.9656,
       "step": 25243
     },
     {
-      "epoch": 0.6933069676746039,
+      "epoch": 0.7163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 4.540908853907647e-06,
-      "loss": 0.7516,
+      "learning_rate": 3.931543691402887e-06,
+      "loss": 0.8565,
       "step": 25244
     },
     {
-      "epoch": 0.693334431902447,
+      "epoch": 0.7163734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 4.540163599495928e-06,
-      "loss": 0.8189,
+      "learning_rate": 3.93081321566519e-06,
+      "loss": 0.871,
       "step": 25245
     },
     {
-      "epoch": 0.6933618961302903,
+      "epoch": 0.7164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 4.53941838828455e-06,
-      "loss": 0.9724,
+      "learning_rate": 3.930082791194193e-06,
+      "loss": 0.8786,
       "step": 25246
     },
     {
-      "epoch": 0.6933893603581335,
+      "epoch": 0.7164301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 4.538673220279408e-06,
-      "loss": 0.7856,
+      "learning_rate": 3.929352417996056e-06,
+      "loss": 0.8149,
       "step": 25247
     },
     {
-      "epoch": 0.6934168245859768,
+      "epoch": 0.7164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 4.537928095486404e-06,
-      "loss": 0.7947,
+      "learning_rate": 3.928622096076953e-06,
+      "loss": 0.7961,
       "step": 25248
     },
     {
-      "epoch": 0.69344428881382,
+      "epoch": 0.7164869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 4.537183013911424e-06,
-      "loss": 0.7383,
+      "learning_rate": 3.9278918254430546e-06,
+      "loss": 0.7898,
       "step": 25249
     },
     {
-      "epoch": 0.6934717530416632,
+      "epoch": 0.7165153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 4.536437975560367e-06,
-      "loss": 0.8397,
+      "learning_rate": 3.9271616061005235e-06,
+      "loss": 0.8612,
       "step": 25250
     },
     {
-      "epoch": 0.6934992172695065,
+      "epoch": 0.7165437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 4.535692980439135e-06,
-      "loss": 0.7989,
+      "learning_rate": 3.926431438055532e-06,
+      "loss": 0.8961,
       "step": 25251
     },
     {
-      "epoch": 0.6935266814973498,
+      "epoch": 0.716572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 4.534948028553612e-06,
-      "loss": 0.8438,
+      "learning_rate": 3.9257013213142505e-06,
+      "loss": 0.9271,
       "step": 25252
     },
     {
-      "epoch": 0.6935541457251929,
+      "epoch": 0.7166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 4.534203119909697e-06,
-      "loss": 0.8359,
+      "learning_rate": 3.924971255882839e-06,
+      "loss": 0.8691,
       "step": 25253
     },
     {
-      "epoch": 0.6935816099530362,
+      "epoch": 0.7166288308740069,
       "grad_norm": 0.0,
-      "learning_rate": 4.533458254513289e-06,
-      "loss": 0.9283,
+      "learning_rate": 3.924241241767472e-06,
+      "loss": 0.875,
       "step": 25254
     },
     {
-      "epoch": 0.6936090741808794,
+      "epoch": 0.7166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 4.532713432370271e-06,
-      "loss": 0.8499,
+      "learning_rate": 3.923511278974309e-06,
+      "loss": 0.859,
       "step": 25255
     },
     {
-      "epoch": 0.6936365384087226,
+      "epoch": 0.7166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 4.531968653486544e-06,
-      "loss": 0.9439,
+      "learning_rate": 3.92278136750952e-06,
+      "loss": 0.8366,
       "step": 25256
     },
     {
-      "epoch": 0.6936640026365659,
+      "epoch": 0.7167139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 4.531223917867998e-06,
-      "loss": 0.8514,
+      "learning_rate": 3.922051507379272e-06,
+      "loss": 0.8886,
       "step": 25257
     },
     {
-      "epoch": 0.6936914668644091,
+      "epoch": 0.7167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 4.5304792255205276e-06,
-      "loss": 0.8812,
+      "learning_rate": 3.921321698589725e-06,
+      "loss": 0.8568,
       "step": 25258
     },
     {
-      "epoch": 0.6937189310922524,
+      "epoch": 0.7167707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 4.529734576450028e-06,
-      "loss": 0.7707,
+      "learning_rate": 3.920591941147047e-06,
+      "loss": 0.9045,
       "step": 25259
     },
     {
-      "epoch": 0.6937463953200955,
+      "epoch": 0.7167990919409761,
       "grad_norm": 0.0,
-      "learning_rate": 4.528989970662387e-06,
-      "loss": 0.7782,
+      "learning_rate": 3.919862235057407e-06,
+      "loss": 0.864,
       "step": 25260
     },
     {
-      "epoch": 0.6937738595479388,
+      "epoch": 0.7168274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 4.528245408163493e-06,
-      "loss": 0.8319,
+      "learning_rate": 3.919132580326955e-06,
+      "loss": 0.7862,
       "step": 25261
     },
     {
-      "epoch": 0.6938013237757821,
+      "epoch": 0.716855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 4.527500888959241e-06,
-      "loss": 0.8485,
+      "learning_rate": 3.918402976961868e-06,
+      "loss": 0.7779,
       "step": 25262
     },
     {
-      "epoch": 0.6938287880036252,
+      "epoch": 0.7168842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 4.526756413055522e-06,
-      "loss": 0.8508,
+      "learning_rate": 3.917673424968308e-06,
+      "loss": 0.8302,
       "step": 25263
     },
     {
-      "epoch": 0.6938562522314685,
+      "epoch": 0.7169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 4.5260119804582256e-06,
-      "loss": 0.8155,
+      "learning_rate": 3.916943924352431e-06,
+      "loss": 0.7524,
       "step": 25264
     },
     {
-      "epoch": 0.6938837164593118,
+      "epoch": 0.7169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 4.525267591173248e-06,
-      "loss": 0.8368,
+      "learning_rate": 3.9162144751204015e-06,
+      "loss": 0.843,
       "step": 25265
     },
     {
-      "epoch": 0.693911180687155,
+      "epoch": 0.7169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 4.524523245206469e-06,
-      "loss": 0.8278,
+      "learning_rate": 3.915485077278385e-06,
+      "loss": 0.8116,
       "step": 25266
     },
     {
-      "epoch": 0.6939386449149982,
+      "epoch": 0.7169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 4.523778942563783e-06,
-      "loss": 0.8817,
+      "learning_rate": 3.9147557308325355e-06,
+      "loss": 0.7487,
       "step": 25267
     },
     {
-      "epoch": 0.6939661091428414,
+      "epoch": 0.7170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 4.523034683251083e-06,
-      "loss": 0.8917,
+      "learning_rate": 3.914026435789019e-06,
+      "loss": 0.8095,
       "step": 25268
     },
     {
-      "epoch": 0.6939935733706847,
+      "epoch": 0.7170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 4.522290467274252e-06,
-      "loss": 0.8452,
+      "learning_rate": 3.9132971921539985e-06,
+      "loss": 0.783,
       "step": 25269
     },
     {
-      "epoch": 0.694021037598528,
+      "epoch": 0.7170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 4.521546294639179e-06,
-      "loss": 0.7591,
+      "learning_rate": 3.912567999933626e-06,
+      "loss": 0.9013,
       "step": 25270
     },
     {
-      "epoch": 0.6940485018263711,
+      "epoch": 0.71711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.520802165351758e-06,
-      "loss": 0.7513,
+      "learning_rate": 3.911838859134066e-06,
+      "loss": 0.7396,
       "step": 25271
     },
     {
-      "epoch": 0.6940759660542144,
+      "epoch": 0.7171396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 4.5200580794178695e-06,
-      "loss": 0.7709,
+      "learning_rate": 3.91110976976148e-06,
+      "loss": 0.8431,
       "step": 25272
     },
     {
-      "epoch": 0.6941034302820576,
+      "epoch": 0.7171679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 4.519314036843403e-06,
-      "loss": 0.8752,
+      "learning_rate": 3.91038073182202e-06,
+      "loss": 0.7603,
       "step": 25273
     },
     {
-      "epoch": 0.6941308945099008,
+      "epoch": 0.7171963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 4.518570037634248e-06,
-      "loss": 0.8715,
+      "learning_rate": 3.909651745321848e-06,
+      "loss": 0.8985,
       "step": 25274
     },
     {
-      "epoch": 0.6941583587377441,
+      "epoch": 0.7172247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 4.517826081796293e-06,
-      "loss": 0.816,
+      "learning_rate": 3.908922810267124e-06,
+      "loss": 0.7992,
       "step": 25275
     },
     {
-      "epoch": 0.6941858229655873,
+      "epoch": 0.7172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 4.517082169335419e-06,
-      "loss": 0.8298,
+      "learning_rate": 3.908193926663995e-06,
+      "loss": 0.8365,
       "step": 25276
     },
     {
-      "epoch": 0.6942132871934306,
+      "epoch": 0.7172814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 4.516338300257518e-06,
-      "loss": 0.8736,
+      "learning_rate": 3.907465094518636e-06,
+      "loss": 0.8076,
       "step": 25277
     },
     {
-      "epoch": 0.6942407514212738,
+      "epoch": 0.7173098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 4.515594474568469e-06,
-      "loss": 0.8518,
+      "learning_rate": 3.906736313837187e-06,
+      "loss": 0.8325,
       "step": 25278
     },
     {
-      "epoch": 0.694268215649117,
+      "epoch": 0.7173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 4.51485069227416e-06,
-      "loss": 0.8988,
+      "learning_rate": 3.906007584625811e-06,
+      "loss": 0.779,
       "step": 25279
     },
     {
-      "epoch": 0.6942956798769603,
+      "epoch": 0.717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 4.5141069533804785e-06,
-      "loss": 0.8501,
+      "learning_rate": 3.905278906890666e-06,
+      "loss": 0.7949,
       "step": 25280
     },
     {
-      "epoch": 0.6943231441048034,
+      "epoch": 0.7173950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 4.513363257893306e-06,
-      "loss": 0.8142,
+      "learning_rate": 3.904550280637901e-06,
+      "loss": 0.7863,
       "step": 25281
     },
     {
-      "epoch": 0.6943506083326467,
+      "epoch": 0.7174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 4.51261960581853e-06,
-      "loss": 0.8683,
+      "learning_rate": 3.903821705873674e-06,
+      "loss": 0.846,
       "step": 25282
     },
     {
-      "epoch": 0.69437807256049,
+      "epoch": 0.7174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 4.511875997162037e-06,
-      "loss": 0.857,
+      "learning_rate": 3.903093182604143e-06,
+      "loss": 0.8276,
       "step": 25283
     },
     {
-      "epoch": 0.6944055367883332,
+      "epoch": 0.7174801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 4.511132431929707e-06,
-      "loss": 0.7326,
+      "learning_rate": 3.902364710835453e-06,
+      "loss": 0.834,
       "step": 25284
     },
     {
-      "epoch": 0.6944330010161764,
+      "epoch": 0.7175085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 4.510388910127419e-06,
-      "loss": 0.7787,
+      "learning_rate": 3.901636290573763e-06,
+      "loss": 0.9386,
       "step": 25285
     },
     {
-      "epoch": 0.6944604652440196,
+      "epoch": 0.7175368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 4.509645431761062e-06,
-      "loss": 0.8453,
+      "learning_rate": 3.90090792182523e-06,
+      "loss": 0.773,
       "step": 25286
     },
     {
-      "epoch": 0.6944879294718629,
+      "epoch": 0.7175652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 4.5089019968365155e-06,
-      "loss": 0.7701,
+      "learning_rate": 3.900179604595998e-06,
+      "loss": 0.8171,
       "step": 25287
     },
     {
-      "epoch": 0.6945153936997062,
+      "epoch": 0.7175936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 4.508158605359664e-06,
-      "loss": 0.7718,
+      "learning_rate": 3.899451338892223e-06,
+      "loss": 0.8561,
       "step": 25288
     },
     {
-      "epoch": 0.6945428579275493,
+      "epoch": 0.7176220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 4.5074152573363915e-06,
-      "loss": 0.8264,
+      "learning_rate": 3.898723124720059e-06,
+      "loss": 0.7695,
       "step": 25289
     },
     {
-      "epoch": 0.6945703221553926,
+      "epoch": 0.7176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 4.506671952772576e-06,
-      "loss": 0.8359,
+      "learning_rate": 3.897994962085653e-06,
+      "loss": 0.7538,
       "step": 25290
     },
     {
-      "epoch": 0.6945977863832359,
+      "epoch": 0.7176787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 4.505928691674097e-06,
-      "loss": 0.8574,
+      "learning_rate": 3.897266850995158e-06,
+      "loss": 0.7658,
       "step": 25291
     },
     {
-      "epoch": 0.694625250611079,
+      "epoch": 0.7177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 4.505185474046844e-06,
-      "loss": 0.7915,
+      "learning_rate": 3.896538791454723e-06,
+      "loss": 0.8168,
       "step": 25292
     },
     {
-      "epoch": 0.6946527148389223,
+      "epoch": 0.7177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 4.5044422998966875e-06,
-      "loss": 0.7907,
+      "learning_rate": 3.8958107834705e-06,
+      "loss": 0.9921,
       "step": 25293
     },
     {
-      "epoch": 0.6946801790667655,
+      "epoch": 0.7177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 4.5036991692295125e-06,
-      "loss": 0.8211,
+      "learning_rate": 3.895082827048641e-06,
+      "loss": 0.845,
       "step": 25294
     },
     {
-      "epoch": 0.6947076432946088,
+      "epoch": 0.7177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 4.502956082051203e-06,
-      "loss": 0.7733,
+      "learning_rate": 3.894354922195288e-06,
+      "loss": 0.9273,
       "step": 25295
     },
     {
-      "epoch": 0.694735107522452,
+      "epoch": 0.7178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 4.50221303836763e-06,
-      "loss": 0.8338,
+      "learning_rate": 3.8936270689165945e-06,
+      "loss": 0.9013,
       "step": 25296
     },
     {
-      "epoch": 0.6947625717502952,
+      "epoch": 0.7178490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 4.501470038184678e-06,
-      "loss": 0.8388,
+      "learning_rate": 3.89289926721871e-06,
+      "loss": 0.9023,
       "step": 25297
     },
     {
-      "epoch": 0.6947900359781385,
+      "epoch": 0.717877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 4.500727081508223e-06,
-      "loss": 0.8774,
+      "learning_rate": 3.892171517107777e-06,
+      "loss": 0.8507,
       "step": 25298
     },
     {
-      "epoch": 0.6948175002059818,
+      "epoch": 0.717905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 4.4999841683441515e-06,
-      "loss": 0.8311,
+      "learning_rate": 3.891443818589946e-06,
+      "loss": 0.7839,
       "step": 25299
     },
     {
-      "epoch": 0.6948449644338249,
+      "epoch": 0.7179341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 4.499241298698331e-06,
-      "loss": 0.8069,
+      "learning_rate": 3.890716171671367e-06,
+      "loss": 0.8179,
       "step": 25300
     },
     {
-      "epoch": 0.6948724286616682,
+      "epoch": 0.7179625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 4.498498472576648e-06,
-      "loss": 0.8728,
+      "learning_rate": 3.889988576358179e-06,
+      "loss": 0.7959,
       "step": 25301
     },
     {
-      "epoch": 0.6948998928895114,
+      "epoch": 0.7179909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 4.4977556899849715e-06,
-      "loss": 0.9137,
+      "learning_rate": 3.889261032656533e-06,
+      "loss": 0.7565,
       "step": 25302
     },
     {
-      "epoch": 0.6949273571173546,
+      "epoch": 0.7180192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 4.497012950929185e-06,
-      "loss": 0.8547,
+      "learning_rate": 3.888533540572577e-06,
+      "loss": 0.7995,
       "step": 25303
     },
     {
-      "epoch": 0.6949548213451979,
+      "epoch": 0.7180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 4.496270255415163e-06,
-      "loss": 0.96,
+      "learning_rate": 3.887806100112449e-06,
+      "loss": 0.9327,
       "step": 25304
     },
     {
-      "epoch": 0.6949822855730411,
+      "epoch": 0.7180760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 4.4955276034487825e-06,
-      "loss": 0.8326,
+      "learning_rate": 3.8870787112822974e-06,
+      "loss": 0.8463,
       "step": 25305
     },
     {
-      "epoch": 0.6950097498008844,
+      "epoch": 0.7181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 4.49478499503592e-06,
-      "loss": 0.8121,
+      "learning_rate": 3.886351374088271e-06,
+      "loss": 0.7232,
       "step": 25306
     },
     {
-      "epoch": 0.6950372140287275,
+      "epoch": 0.7181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 4.494042430182455e-06,
-      "loss": 0.9159,
+      "learning_rate": 3.885624088536501e-06,
+      "loss": 0.8679,
       "step": 25307
     },
     {
-      "epoch": 0.6950646782565708,
+      "epoch": 0.7181611804767309,
       "grad_norm": 0.0,
-      "learning_rate": 4.493299908894258e-06,
-      "loss": 0.8293,
+      "learning_rate": 3.8848968546331475e-06,
+      "loss": 0.7796,
       "step": 25308
     },
     {
-      "epoch": 0.6950921424844141,
+      "epoch": 0.7181895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 4.492557431177202e-06,
-      "loss": 0.9164,
+      "learning_rate": 3.884169672384342e-06,
+      "loss": 0.8734,
       "step": 25309
     },
     {
-      "epoch": 0.6951196067122573,
+      "epoch": 0.7182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 4.491814997037166e-06,
-      "loss": 0.8385,
+      "learning_rate": 3.88344254179623e-06,
+      "loss": 0.9233,
       "step": 25310
     },
     {
-      "epoch": 0.6951470709401005,
+      "epoch": 0.7182463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 4.491072606480023e-06,
-      "loss": 0.7314,
+      "learning_rate": 3.882715462874957e-06,
+      "loss": 0.8718,
       "step": 25311
     },
     {
-      "epoch": 0.6951745351679438,
+      "epoch": 0.7182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 4.4903302595116475e-06,
-      "loss": 0.8097,
+      "learning_rate": 3.881988435626658e-06,
+      "loss": 0.9598,
       "step": 25312
     },
     {
-      "epoch": 0.695201999395787,
+      "epoch": 0.7183030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 4.4895879561379166e-06,
-      "loss": 0.8046,
+      "learning_rate": 3.881261460057477e-06,
+      "loss": 0.8085,
       "step": 25313
     },
     {
-      "epoch": 0.6952294636236303,
+      "epoch": 0.7183314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 4.488845696364697e-06,
-      "loss": 0.7744,
+      "learning_rate": 3.88053453617356e-06,
+      "loss": 0.8652,
       "step": 25314
     },
     {
-      "epoch": 0.6952569278514734,
+      "epoch": 0.7183598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 4.488103480197864e-06,
-      "loss": 0.8876,
+      "learning_rate": 3.879807663981039e-06,
+      "loss": 0.9068,
       "step": 25315
     },
     {
-      "epoch": 0.6952843920793167,
+      "epoch": 0.7183881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 4.487361307643295e-06,
-      "loss": 0.8339,
+      "learning_rate": 3.879080843486057e-06,
+      "loss": 0.9095,
       "step": 25316
     },
     {
-      "epoch": 0.69531185630716,
+      "epoch": 0.718416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 4.486619178706857e-06,
-      "loss": 0.8723,
+      "learning_rate": 3.878354074694759e-06,
+      "loss": 0.8729,
       "step": 25317
     },
     {
-      "epoch": 0.6953393205350031,
+      "epoch": 0.7184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 4.485877093394422e-06,
-      "loss": 0.8645,
+      "learning_rate": 3.877627357613276e-06,
+      "loss": 0.7679,
       "step": 25318
     },
     {
-      "epoch": 0.6953667847628464,
+      "epoch": 0.7184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 4.4851350517118675e-06,
-      "loss": 0.8439,
+      "learning_rate": 3.876900692247749e-06,
+      "loss": 0.7149,
       "step": 25319
     },
     {
-      "epoch": 0.6953942489906896,
+      "epoch": 0.7185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 4.484393053665058e-06,
-      "loss": 0.8512,
+      "learning_rate": 3.876174078604321e-06,
+      "loss": 0.8078,
       "step": 25320
     },
     {
-      "epoch": 0.6954217132185329,
+      "epoch": 0.7185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 4.483651099259866e-06,
-      "loss": 0.8388,
+      "learning_rate": 3.8754475166891225e-06,
+      "loss": 0.6939,
       "step": 25321
     },
     {
-      "epoch": 0.6954491774463761,
+      "epoch": 0.7185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 4.482909188502163e-06,
-      "loss": 0.8942,
+      "learning_rate": 3.874721006508293e-06,
+      "loss": 0.7739,
       "step": 25322
     },
     {
-      "epoch": 0.6954766416742193,
+      "epoch": 0.7185868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 4.482167321397825e-06,
-      "loss": 0.7902,
+      "learning_rate": 3.873994548067972e-06,
+      "loss": 0.8407,
       "step": 25323
     },
     {
-      "epoch": 0.6955041059020626,
+      "epoch": 0.7186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 4.481425497952716e-06,
-      "loss": 0.7509,
+      "learning_rate": 3.873268141374293e-06,
+      "loss": 0.6963,
       "step": 25324
     },
     {
-      "epoch": 0.6955315701299059,
+      "epoch": 0.7186435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 4.480683718172701e-06,
-      "loss": 0.866,
+      "learning_rate": 3.872541786433398e-06,
+      "loss": 0.9436,
       "step": 25325
     },
     {
-      "epoch": 0.695559034357749,
+      "epoch": 0.7186719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 4.4799419820636555e-06,
-      "loss": 0.8928,
+      "learning_rate": 3.871815483251414e-06,
+      "loss": 0.7693,
       "step": 25326
     },
     {
-      "epoch": 0.6955864985855923,
+      "epoch": 0.7187003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 4.479200289631448e-06,
-      "loss": 0.8014,
+      "learning_rate": 3.8710892318344804e-06,
+      "loss": 0.7878,
       "step": 25327
     },
     {
-      "epoch": 0.6956139628134355,
+      "epoch": 0.7187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 4.478458640881945e-06,
-      "loss": 0.7365,
+      "learning_rate": 3.870363032188735e-06,
+      "loss": 0.8222,
       "step": 25328
     },
     {
-      "epoch": 0.6956414270412787,
+      "epoch": 0.7187570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 4.477717035821018e-06,
-      "loss": 0.7283,
+      "learning_rate": 3.869636884320306e-06,
+      "loss": 0.8405,
       "step": 25329
     },
     {
-      "epoch": 0.695668891269122,
+      "epoch": 0.7187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 4.476975474454534e-06,
-      "loss": 0.7993,
+      "learning_rate": 3.868910788235328e-06,
+      "loss": 0.7901,
       "step": 25330
     },
     {
-      "epoch": 0.6956963554969652,
+      "epoch": 0.7188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 4.476233956788358e-06,
-      "loss": 0.9392,
+      "learning_rate": 3.868184743939941e-06,
+      "loss": 0.7673,
       "step": 25331
     },
     {
-      "epoch": 0.6957238197248085,
+      "epoch": 0.7188422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 4.475492482828361e-06,
-      "loss": 0.818,
+      "learning_rate": 3.86745875144027e-06,
+      "loss": 0.931,
       "step": 25332
     },
     {
-      "epoch": 0.6957512839526516,
+      "epoch": 0.7188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 4.474751052580404e-06,
-      "loss": 0.8583,
+      "learning_rate": 3.86673281074245e-06,
+      "loss": 0.8325,
       "step": 25333
     },
     {
-      "epoch": 0.6957787481804949,
+      "epoch": 0.7188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 4.474009666050358e-06,
-      "loss": 0.8496,
+      "learning_rate": 3.8660069218526165e-06,
+      "loss": 0.742,
       "step": 25334
     },
     {
-      "epoch": 0.6958062124083382,
+      "epoch": 0.718927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 4.4732683232440865e-06,
-      "loss": 0.8548,
+      "learning_rate": 3.865281084776895e-06,
+      "loss": 0.7761,
       "step": 25335
     },
     {
-      "epoch": 0.6958336766361813,
+      "epoch": 0.718955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 4.4725270241674615e-06,
-      "loss": 0.8759,
+      "learning_rate": 3.864555299521421e-06,
+      "loss": 0.7412,
       "step": 25336
     },
     {
-      "epoch": 0.6958611408640246,
+      "epoch": 0.7189841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 4.47178576882634e-06,
-      "loss": 0.7841,
+      "learning_rate": 3.863829566092323e-06,
+      "loss": 0.8517,
       "step": 25337
     },
     {
-      "epoch": 0.6958886050918679,
+      "epoch": 0.7190124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 4.471044557226591e-06,
-      "loss": 0.8792,
+      "learning_rate": 3.863103884495732e-06,
+      "loss": 0.8424,
       "step": 25338
     },
     {
-      "epoch": 0.6959160693197111,
+      "epoch": 0.7190408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 4.470303389374079e-06,
-      "loss": 0.8561,
+      "learning_rate": 3.862378254737783e-06,
+      "loss": 0.8241,
       "step": 25339
     },
     {
-      "epoch": 0.6959435335475543,
+      "epoch": 0.7190692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 4.469562265274673e-06,
-      "loss": 0.8969,
+      "learning_rate": 3.861652676824597e-06,
+      "loss": 0.7425,
       "step": 25340
     },
     {
-      "epoch": 0.6959709977753975,
+      "epoch": 0.7190976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 4.4688211849342274e-06,
-      "loss": 0.8255,
+      "learning_rate": 3.860927150762307e-06,
+      "loss": 0.8264,
       "step": 25341
     },
     {
-      "epoch": 0.6959984620032408,
+      "epoch": 0.7191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 4.468080148358616e-06,
-      "loss": 0.7281,
+      "learning_rate": 3.860201676557045e-06,
+      "loss": 0.8074,
       "step": 25342
     },
     {
-      "epoch": 0.6960259262310841,
+      "epoch": 0.7191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 4.467339155553694e-06,
-      "loss": 0.8433,
+      "learning_rate": 3.859476254214933e-06,
+      "loss": 0.9444,
       "step": 25343
     },
     {
-      "epoch": 0.6960533904589272,
+      "epoch": 0.7191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 4.466598206525327e-06,
-      "loss": 0.8334,
+      "learning_rate": 3.858750883742098e-06,
+      "loss": 0.9549,
       "step": 25344
     },
     {
-      "epoch": 0.6960808546867705,
+      "epoch": 0.719211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.4658573012793785e-06,
-      "loss": 0.7963,
+      "learning_rate": 3.858025565144676e-06,
+      "loss": 0.8806,
       "step": 25345
     },
     {
-      "epoch": 0.6961083189146137,
+      "epoch": 0.7192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 4.465116439821711e-06,
-      "loss": 0.7959,
+      "learning_rate": 3.8573002984287845e-06,
+      "loss": 0.8863,
       "step": 25346
     },
     {
-      "epoch": 0.6961357831424569,
+      "epoch": 0.7192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 4.464375622158189e-06,
-      "loss": 0.8692,
+      "learning_rate": 3.856575083600553e-06,
+      "loss": 0.8355,
       "step": 25347
     },
     {
-      "epoch": 0.6961632473703002,
+      "epoch": 0.7192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 4.463634848294673e-06,
-      "loss": 0.8784,
+      "learning_rate": 3.855849920666111e-06,
+      "loss": 0.9126,
       "step": 25348
     },
     {
-      "epoch": 0.6961907115981434,
+      "epoch": 0.7193246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 4.462894118237017e-06,
-      "loss": 0.8153,
+      "learning_rate": 3.8551248096315785e-06,
+      "loss": 0.7867,
       "step": 25349
     },
     {
-      "epoch": 0.6962181758259867,
+      "epoch": 0.7193530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 4.462153431991088e-06,
-      "loss": 0.7695,
+      "learning_rate": 3.8543997505030826e-06,
+      "loss": 0.9125,
       "step": 25350
     },
     {
-      "epoch": 0.6962456400538299,
+      "epoch": 0.7193813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 4.461412789562748e-06,
-      "loss": 0.8226,
+      "learning_rate": 3.853674743286752e-06,
+      "loss": 0.8604,
       "step": 25351
     },
     {
-      "epoch": 0.6962731042816731,
+      "epoch": 0.7194097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 4.460672190957854e-06,
-      "loss": 0.9004,
+      "learning_rate": 3.852949787988704e-06,
+      "loss": 0.8495,
       "step": 25352
     },
     {
-      "epoch": 0.6963005685095164,
+      "epoch": 0.7194381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 4.459931636182267e-06,
-      "loss": 0.8177,
+      "learning_rate": 3.852224884615061e-06,
+      "loss": 0.8326,
       "step": 25353
     },
     {
-      "epoch": 0.6963280327373595,
+      "epoch": 0.719466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 4.459191125241852e-06,
-      "loss": 0.7485,
+      "learning_rate": 3.851500033171958e-06,
+      "loss": 0.8169,
       "step": 25354
     },
     {
-      "epoch": 0.6963554969652028,
+      "epoch": 0.7194948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 4.458450658142459e-06,
-      "loss": 0.8017,
+      "learning_rate": 3.850775233665507e-06,
+      "loss": 0.8631,
       "step": 25355
     },
     {
-      "epoch": 0.6963829611930461,
+      "epoch": 0.7195232690124859,
       "grad_norm": 0.0,
-      "learning_rate": 4.457710234889954e-06,
-      "loss": 0.8107,
+      "learning_rate": 3.850050486101834e-06,
+      "loss": 0.6999,
       "step": 25356
     },
     {
-      "epoch": 0.6964104254208893,
+      "epoch": 0.7195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 4.456969855490189e-06,
-      "loss": 0.8239,
+      "learning_rate": 3.8493257904870654e-06,
+      "loss": 0.962,
       "step": 25357
     },
     {
-      "epoch": 0.6964378896487325,
+      "epoch": 0.7195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 4.456229519949026e-06,
-      "loss": 0.8658,
+      "learning_rate": 3.8486011468273145e-06,
+      "loss": 0.7861,
       "step": 25358
     },
     {
-      "epoch": 0.6964653538765757,
+      "epoch": 0.7196083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 4.455489228272323e-06,
-      "loss": 0.85,
+      "learning_rate": 3.847876555128706e-06,
+      "loss": 0.9084,
       "step": 25359
     },
     {
-      "epoch": 0.696492818104419,
+      "epoch": 0.7196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 4.454748980465939e-06,
-      "loss": 0.8486,
+      "learning_rate": 3.847152015397363e-06,
+      "loss": 0.7371,
       "step": 25360
     },
     {
-      "epoch": 0.6965202823322623,
+      "epoch": 0.7196651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 4.454008776535727e-06,
-      "loss": 0.7943,
+      "learning_rate": 3.846427527639401e-06,
+      "loss": 0.9031,
       "step": 25361
     },
     {
-      "epoch": 0.6965477465601054,
+      "epoch": 0.7196935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 4.453268616487545e-06,
-      "loss": 0.8951,
+      "learning_rate": 3.845703091860947e-06,
+      "loss": 0.7877,
       "step": 25362
     },
     {
-      "epoch": 0.6965752107879487,
+      "epoch": 0.7197219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 4.452528500327254e-06,
-      "loss": 0.7807,
+      "learning_rate": 3.8449787080681115e-06,
+      "loss": 0.846,
       "step": 25363
     },
     {
-      "epoch": 0.696602675015792,
+      "epoch": 0.719750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 4.451788428060704e-06,
-      "loss": 0.8498,
+      "learning_rate": 3.844254376267017e-06,
+      "loss": 0.8186,
       "step": 25364
     },
     {
-      "epoch": 0.6966301392436351,
+      "epoch": 0.7197786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 4.4510483996937515e-06,
-      "loss": 0.8189,
+      "learning_rate": 3.843530096463786e-06,
+      "loss": 0.8643,
       "step": 25365
     },
     {
-      "epoch": 0.6966576034714784,
+      "epoch": 0.7198070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 4.450308415232258e-06,
-      "loss": 0.8027,
+      "learning_rate": 3.84280586866453e-06,
+      "loss": 0.7289,
       "step": 25366
     },
     {
-      "epoch": 0.6966850676993216,
+      "epoch": 0.7198354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 4.44956847468207e-06,
-      "loss": 0.8601,
+      "learning_rate": 3.842081692875369e-06,
+      "loss": 0.8043,
       "step": 25367
     },
     {
-      "epoch": 0.6967125319271649,
+      "epoch": 0.7198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 4.448828578049047e-06,
-      "loss": 0.8108,
+      "learning_rate": 3.841357569102421e-06,
+      "loss": 0.9029,
       "step": 25368
     },
     {
-      "epoch": 0.6967399961550081,
+      "epoch": 0.7198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 4.448088725339041e-06,
-      "loss": 0.8528,
+      "learning_rate": 3.840633497351802e-06,
+      "loss": 0.8248,
       "step": 25369
     },
     {
-      "epoch": 0.6967674603828513,
+      "epoch": 0.7199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 4.44734891655791e-06,
-      "loss": 0.7439,
+      "learning_rate": 3.83990947762963e-06,
+      "loss": 0.9258,
       "step": 25370
     },
     {
-      "epoch": 0.6967949246106946,
+      "epoch": 0.7199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 4.4466091517115074e-06,
-      "loss": 0.9408,
+      "learning_rate": 3.839185509942023e-06,
+      "loss": 0.8736,
       "step": 25371
     },
     {
-      "epoch": 0.6968223888385378,
+      "epoch": 0.7199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 4.445869430805684e-06,
-      "loss": 0.7984,
+      "learning_rate": 3.8384615942950885e-06,
+      "loss": 0.7949,
       "step": 25372
     },
     {
-      "epoch": 0.696849853066381,
+      "epoch": 0.7200056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 4.44512975384629e-06,
-      "loss": 0.865,
+      "learning_rate": 3.837737730694946e-06,
+      "loss": 0.8441,
       "step": 25373
     },
     {
-      "epoch": 0.6968773172942243,
+      "epoch": 0.7200340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 4.444390120839182e-06,
-      "loss": 0.7767,
+      "learning_rate": 3.837013919147714e-06,
+      "loss": 0.8639,
       "step": 25374
     },
     {
-      "epoch": 0.6969047815220675,
+      "epoch": 0.7200624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 4.44365053179021e-06,
-      "loss": 0.8689,
+      "learning_rate": 3.836290159659498e-06,
+      "loss": 0.8333,
       "step": 25375
     },
     {
-      "epoch": 0.6969322457499108,
+      "epoch": 0.7200908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 4.442910986705228e-06,
-      "loss": 0.7865,
+      "learning_rate": 3.835566452236416e-06,
+      "loss": 0.8177,
       "step": 25376
     },
     {
-      "epoch": 0.696959709977754,
+      "epoch": 0.7201191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 4.4421714855900866e-06,
-      "loss": 0.8137,
+      "learning_rate": 3.834842796884586e-06,
+      "loss": 0.8101,
       "step": 25377
     },
     {
-      "epoch": 0.6969871742055972,
+      "epoch": 0.7201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 4.441432028450641e-06,
-      "loss": 0.7862,
+      "learning_rate": 3.834119193610112e-06,
+      "loss": 0.7855,
       "step": 25378
     },
     {
-      "epoch": 0.6970146384334405,
+      "epoch": 0.7201759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 4.440692615292735e-06,
-      "loss": 0.7475,
+      "learning_rate": 3.833395642419111e-06,
+      "loss": 0.7764,
       "step": 25379
     },
     {
-      "epoch": 0.6970421026612836,
+      "epoch": 0.7202043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 4.439953246122226e-06,
-      "loss": 0.9618,
+      "learning_rate": 3.8326721433176975e-06,
+      "loss": 0.9037,
       "step": 25380
     },
     {
-      "epoch": 0.6970695668891269,
+      "epoch": 0.7202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 4.439213920944957e-06,
-      "loss": 0.7803,
+      "learning_rate": 3.831948696311974e-06,
+      "loss": 0.8238,
       "step": 25381
     },
     {
-      "epoch": 0.6970970311169702,
+      "epoch": 0.720261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 4.438474639766781e-06,
-      "loss": 0.9035,
+      "learning_rate": 3.83122530140806e-06,
+      "loss": 0.8806,
       "step": 25382
     },
     {
-      "epoch": 0.6971244953448134,
+      "epoch": 0.7202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 4.437735402593553e-06,
-      "loss": 0.7889,
+      "learning_rate": 3.8305019586120615e-06,
+      "loss": 0.778,
       "step": 25383
     },
     {
-      "epoch": 0.6971519595726566,
+      "epoch": 0.7203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 4.436996209431113e-06,
-      "loss": 0.8773,
+      "learning_rate": 3.82977866793009e-06,
+      "loss": 0.8747,
       "step": 25384
     },
     {
-      "epoch": 0.6971794238004998,
+      "epoch": 0.7203461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 4.436257060285313e-06,
-      "loss": 0.7014,
+      "learning_rate": 3.8290554293682614e-06,
+      "loss": 0.754,
       "step": 25385
     },
     {
-      "epoch": 0.6972068880283431,
+      "epoch": 0.7203745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 4.435517955162003e-06,
-      "loss": 0.7977,
+      "learning_rate": 3.828332242932673e-06,
+      "loss": 0.8888,
       "step": 25386
     },
     {
-      "epoch": 0.6972343522561864,
+      "epoch": 0.7204029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 4.4347788940670335e-06,
-      "loss": 0.8015,
+      "learning_rate": 3.827609108629441e-06,
+      "loss": 0.8067,
       "step": 25387
     },
     {
-      "epoch": 0.6972618164840295,
+      "epoch": 0.7204313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 4.434039877006247e-06,
-      "loss": 0.908,
+      "learning_rate": 3.826886026464676e-06,
+      "loss": 0.8584,
       "step": 25388
     },
     {
-      "epoch": 0.6972892807118728,
+      "epoch": 0.7204597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 4.433300903985495e-06,
-      "loss": 0.8149,
+      "learning_rate": 3.826162996444477e-06,
+      "loss": 0.8275,
       "step": 25389
     },
     {
-      "epoch": 0.6973167449397161,
+      "epoch": 0.7204880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 4.43256197501062e-06,
-      "loss": 0.7954,
+      "learning_rate": 3.825440018574958e-06,
+      "loss": 0.7658,
       "step": 25390
     },
     {
-      "epoch": 0.6973442091675592,
+      "epoch": 0.7205164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 4.431823090087471e-06,
-      "loss": 0.8529,
+      "learning_rate": 3.824717092862228e-06,
+      "loss": 0.793,
       "step": 25391
     },
     {
-      "epoch": 0.6973716733954025,
+      "epoch": 0.7205448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 4.431084249221893e-06,
-      "loss": 0.9,
+      "learning_rate": 3.823994219312387e-06,
+      "loss": 0.7652,
       "step": 25392
     },
     {
-      "epoch": 0.6973991376232457,
+      "epoch": 0.7205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 4.430345452419735e-06,
-      "loss": 0.8151,
+      "learning_rate": 3.823271397931544e-06,
+      "loss": 0.9409,
       "step": 25393
     },
     {
-      "epoch": 0.697426601851089,
+      "epoch": 0.7206015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 4.429606699686842e-06,
-      "loss": 0.779,
+      "learning_rate": 3.82254862872581e-06,
+      "loss": 0.9435,
       "step": 25394
     },
     {
-      "epoch": 0.6974540660789322,
+      "epoch": 0.7206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 4.428867991029061e-06,
-      "loss": 0.8145,
+      "learning_rate": 3.82182591170128e-06,
+      "loss": 0.8818,
       "step": 25395
     },
     {
-      "epoch": 0.6974815303067754,
+      "epoch": 0.7206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 4.428129326452235e-06,
-      "loss": 0.7938,
+      "learning_rate": 3.821103246864065e-06,
+      "loss": 0.8035,
       "step": 25396
     },
     {
-      "epoch": 0.6975089945346187,
+      "epoch": 0.7206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 4.427390705962205e-06,
-      "loss": 0.8876,
+      "learning_rate": 3.820380634220272e-06,
+      "loss": 0.862,
       "step": 25397
     },
     {
-      "epoch": 0.6975364587624618,
+      "epoch": 0.7207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 4.426652129564818e-06,
-      "loss": 0.8418,
+      "learning_rate": 3.819658073775998e-06,
+      "loss": 0.8205,
       "step": 25398
     },
     {
-      "epoch": 0.6975639229903051,
+      "epoch": 0.7207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 4.425913597265917e-06,
-      "loss": 0.8194,
+      "learning_rate": 3.81893556553735e-06,
+      "loss": 0.8755,
       "step": 25399
     },
     {
-      "epoch": 0.6975913872181484,
+      "epoch": 0.720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 4.425175109071348e-06,
-      "loss": 0.7291,
+      "learning_rate": 3.818213109510432e-06,
+      "loss": 0.7698,
       "step": 25400
     },
     {
-      "epoch": 0.6976188514459916,
+      "epoch": 0.720800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 4.424436664986957e-06,
-      "loss": 0.8784,
+      "learning_rate": 3.817490705701346e-06,
+      "loss": 0.8004,
       "step": 25401
     },
     {
-      "epoch": 0.6976463156738348,
+      "epoch": 0.7208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 4.423698265018579e-06,
-      "loss": 0.8826,
+      "learning_rate": 3.816768354116196e-06,
+      "loss": 0.8611,
       "step": 25402
     },
     {
-      "epoch": 0.6976737799016781,
+      "epoch": 0.7208569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 4.4229599091720595e-06,
-      "loss": 0.86,
+      "learning_rate": 3.8160460547610785e-06,
+      "loss": 0.8779,
       "step": 25403
     },
     {
-      "epoch": 0.6977012441295213,
+      "epoch": 0.7208853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 4.4222215974532466e-06,
-      "loss": 0.8229,
+      "learning_rate": 3.8153238076420995e-06,
+      "loss": 0.8561,
       "step": 25404
     },
     {
-      "epoch": 0.6977287083573646,
+      "epoch": 0.7209137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 4.421483329867972e-06,
-      "loss": 0.8246,
+      "learning_rate": 3.8146016127653605e-06,
+      "loss": 0.8655,
       "step": 25405
     },
     {
-      "epoch": 0.6977561725852077,
+      "epoch": 0.7209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 4.420745106422083e-06,
-      "loss": 0.8276,
+      "learning_rate": 3.8138794701369562e-06,
+      "loss": 0.8322,
       "step": 25406
     },
     {
-      "epoch": 0.697783636813051,
+      "epoch": 0.7209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 4.420006927121423e-06,
-      "loss": 0.8312,
+      "learning_rate": 3.8131573797629906e-06,
+      "loss": 0.8502,
       "step": 25407
     },
     {
-      "epoch": 0.6978111010408943,
+      "epoch": 0.7209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 4.4192687919718265e-06,
-      "loss": 0.836,
+      "learning_rate": 3.8124353416495653e-06,
+      "loss": 0.7471,
       "step": 25408
     },
     {
-      "epoch": 0.6978385652687374,
+      "epoch": 0.7210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 4.418530700979138e-06,
-      "loss": 0.7975,
+      "learning_rate": 3.8117133558027743e-06,
+      "loss": 0.8066,
       "step": 25409
     },
     {
-      "epoch": 0.6978660294965807,
+      "epoch": 0.72105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 4.417792654149196e-06,
-      "loss": 0.8972,
+      "learning_rate": 3.810991422228718e-06,
+      "loss": 0.8997,
       "step": 25410
     },
     {
-      "epoch": 0.6978934937244239,
+      "epoch": 0.7210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 4.417054651487843e-06,
-      "loss": 0.7816,
+      "learning_rate": 3.8102695409335e-06,
+      "loss": 0.7708,
       "step": 25411
     },
     {
-      "epoch": 0.6979209579522672,
+      "epoch": 0.7211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 4.416316693000914e-06,
-      "loss": 0.8649,
+      "learning_rate": 3.8095477119232092e-06,
+      "loss": 0.9176,
       "step": 25412
     },
     {
-      "epoch": 0.6979484221801104,
+      "epoch": 0.7211407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 4.415578778694254e-06,
-      "loss": 0.7839,
+      "learning_rate": 3.808825935203947e-06,
+      "loss": 0.8525,
       "step": 25413
     },
     {
-      "epoch": 0.6979758864079536,
+      "epoch": 0.7211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 4.414840908573693e-06,
-      "loss": 0.8506,
+      "learning_rate": 3.80810421078181e-06,
+      "loss": 0.9028,
       "step": 25414
     },
     {
-      "epoch": 0.6980033506357969,
+      "epoch": 0.7211975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 4.414103082645076e-06,
-      "loss": 0.9106,
+      "learning_rate": 3.807382538662896e-06,
+      "loss": 0.8686,
       "step": 25415
     },
     {
-      "epoch": 0.6980308148636402,
+      "epoch": 0.7212258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 4.413365300914237e-06,
-      "loss": 0.763,
+      "learning_rate": 3.8066609188533033e-06,
+      "loss": 0.8374,
       "step": 25416
     },
     {
-      "epoch": 0.6980582790914833,
+      "epoch": 0.7212542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 4.412627563387018e-06,
-      "loss": 0.7703,
+      "learning_rate": 3.8059393513591203e-06,
+      "loss": 0.8454,
       "step": 25417
     },
     {
-      "epoch": 0.6980857433193266,
+      "epoch": 0.7212826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 4.411889870069253e-06,
-      "loss": 0.7656,
+      "learning_rate": 3.8052178361864467e-06,
+      "loss": 0.8446,
       "step": 25418
     },
     {
-      "epoch": 0.6981132075471698,
+      "epoch": 0.721311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 4.411152220966783e-06,
-      "loss": 0.8545,
+      "learning_rate": 3.8044963733413797e-06,
+      "loss": 0.8345,
       "step": 25419
     },
     {
-      "epoch": 0.698140671775013,
+      "epoch": 0.7213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 4.410414616085443e-06,
-      "loss": 0.9019,
+      "learning_rate": 3.803774962830007e-06,
+      "loss": 0.8116,
       "step": 25420
     },
     {
-      "epoch": 0.6981681360028563,
+      "epoch": 0.7213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 4.409677055431063e-06,
-      "loss": 0.8075,
+      "learning_rate": 3.8030536046584253e-06,
+      "loss": 0.9447,
       "step": 25421
     },
     {
-      "epoch": 0.6981956002306995,
+      "epoch": 0.7213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 4.408939539009485e-06,
-      "loss": 0.8859,
+      "learning_rate": 3.8023322988327325e-06,
+      "loss": 0.7523,
       "step": 25422
     },
     {
-      "epoch": 0.6982230644585428,
+      "epoch": 0.7214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 4.408202066826542e-06,
-      "loss": 0.8356,
+      "learning_rate": 3.801611045359014e-06,
+      "loss": 0.8134,
       "step": 25423
     },
     {
-      "epoch": 0.6982505286863859,
+      "epoch": 0.7214528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 4.40746463888807e-06,
-      "loss": 0.8406,
+      "learning_rate": 3.800889844243365e-06,
+      "loss": 0.8263,
       "step": 25424
     },
     {
-      "epoch": 0.6982779929142292,
+      "epoch": 0.7214812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 4.406727255199908e-06,
-      "loss": 0.801,
+      "learning_rate": 3.8001686954918815e-06,
+      "loss": 0.8946,
       "step": 25425
     },
     {
-      "epoch": 0.6983054571420725,
+      "epoch": 0.7215096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 4.405989915767883e-06,
-      "loss": 0.882,
+      "learning_rate": 3.7994475991106496e-06,
+      "loss": 0.8389,
       "step": 25426
     },
     {
-      "epoch": 0.6983329213699156,
+      "epoch": 0.7215380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 4.405252620597833e-06,
-      "loss": 0.912,
+      "learning_rate": 3.7987265551057606e-06,
+      "loss": 0.8978,
       "step": 25427
     },
     {
-      "epoch": 0.6983603855977589,
+      "epoch": 0.7215664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 4.404515369695595e-06,
-      "loss": 0.7825,
+      "learning_rate": 3.798005563483312e-06,
+      "loss": 0.8049,
       "step": 25428
     },
     {
-      "epoch": 0.6983878498256022,
+      "epoch": 0.721594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 4.403778163066996e-06,
-      "loss": 0.7066,
+      "learning_rate": 3.7972846242493823e-06,
+      "loss": 0.9158,
       "step": 25429
     },
     {
-      "epoch": 0.6984153140534454,
+      "epoch": 0.7216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 4.40304100071787e-06,
-      "loss": 0.8259,
+      "learning_rate": 3.796563737410074e-06,
+      "loss": 0.8361,
       "step": 25430
     },
     {
-      "epoch": 0.6984427782812886,
+      "epoch": 0.7216515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 4.402303882654057e-06,
-      "loss": 0.8196,
+      "learning_rate": 3.795842902971468e-06,
+      "loss": 0.9321,
       "step": 25431
     },
     {
-      "epoch": 0.6984702425091318,
+      "epoch": 0.7216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 4.40156680888138e-06,
-      "loss": 0.7677,
+      "learning_rate": 3.795122120939656e-06,
+      "loss": 0.799,
       "step": 25432
     },
     {
-      "epoch": 0.6984977067369751,
+      "epoch": 0.7217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 4.400829779405675e-06,
-      "loss": 0.9242,
+      "learning_rate": 3.79440139132073e-06,
+      "loss": 0.7635,
       "step": 25433
     },
     {
-      "epoch": 0.6985251709648184,
+      "epoch": 0.7217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 4.4000927942327735e-06,
-      "loss": 0.9499,
+      "learning_rate": 3.79368071412077e-06,
+      "loss": 0.8005,
       "step": 25434
     },
     {
-      "epoch": 0.6985526351926615,
+      "epoch": 0.7217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 4.399355853368511e-06,
-      "loss": 0.8866,
+      "learning_rate": 3.7929600893458684e-06,
+      "loss": 0.7845,
       "step": 25435
     },
     {
-      "epoch": 0.6985800994205048,
+      "epoch": 0.7217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 4.39861895681871e-06,
-      "loss": 0.8007,
+      "learning_rate": 3.7922395170021163e-06,
+      "loss": 0.8175,
       "step": 25436
     },
     {
-      "epoch": 0.698607563648348,
+      "epoch": 0.721821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 4.397882104589211e-06,
-      "loss": 0.796,
+      "learning_rate": 3.7915189970955913e-06,
+      "loss": 0.7751,
       "step": 25437
     },
     {
-      "epoch": 0.6986350278761913,
+      "epoch": 0.721850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 4.397145296685833e-06,
-      "loss": 0.8555,
+      "learning_rate": 3.7907985296323857e-06,
+      "loss": 0.8562,
       "step": 25438
     },
     {
-      "epoch": 0.6986624921040345,
+      "epoch": 0.7218785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 4.3964085331144124e-06,
-      "loss": 0.8021,
+      "learning_rate": 3.790078114618586e-06,
+      "loss": 0.8151,
       "step": 25439
     },
     {
-      "epoch": 0.6986899563318777,
+      "epoch": 0.7219069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 4.395671813880779e-06,
-      "loss": 0.7543,
+      "learning_rate": 3.7893577520602733e-06,
+      "loss": 0.793,
       "step": 25440
     },
     {
-      "epoch": 0.698717420559721,
+      "epoch": 0.7219353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 4.39493513899076e-06,
-      "loss": 0.8465,
+      "learning_rate": 3.788637441963534e-06,
+      "loss": 0.8346,
       "step": 25441
     },
     {
-      "epoch": 0.6987448847875642,
+      "epoch": 0.7219636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 4.394198508450191e-06,
-      "loss": 0.8543,
+      "learning_rate": 3.7879171843344576e-06,
+      "loss": 0.7983,
       "step": 25442
     },
     {
-      "epoch": 0.6987723490154074,
+      "epoch": 0.7219920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 4.3934619222648885e-06,
-      "loss": 0.8219,
+      "learning_rate": 3.787196979179121e-06,
+      "loss": 0.8746,
       "step": 25443
     },
     {
-      "epoch": 0.6987998132432507,
+      "epoch": 0.7220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 4.392725380440692e-06,
-      "loss": 0.8394,
+      "learning_rate": 3.78647682650361e-06,
+      "loss": 0.7422,
       "step": 25444
     },
     {
-      "epoch": 0.6988272774710939,
+      "epoch": 0.7220488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 4.391988882983421e-06,
-      "loss": 0.8155,
+      "learning_rate": 3.7857567263140084e-06,
+      "loss": 0.7918,
       "step": 25445
     },
     {
-      "epoch": 0.6988547416989371,
+      "epoch": 0.7220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 4.391252429898906e-06,
-      "loss": 0.9193,
+      "learning_rate": 3.7850366786163995e-06,
+      "loss": 0.8452,
       "step": 25446
     },
     {
-      "epoch": 0.6988822059267804,
+      "epoch": 0.722105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 4.3905160211929746e-06,
-      "loss": 0.8392,
+      "learning_rate": 3.784316683416869e-06,
+      "loss": 0.8644,
       "step": 25447
     },
     {
-      "epoch": 0.6989096701546236,
+      "epoch": 0.7221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 4.389779656871457e-06,
-      "loss": 0.8554,
+      "learning_rate": 3.7835967407214913e-06,
+      "loss": 0.8234,
       "step": 25448
     },
     {
-      "epoch": 0.6989371343824669,
+      "epoch": 0.7221623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 4.389043336940172e-06,
-      "loss": 0.9016,
+      "learning_rate": 3.782876850536351e-06,
+      "loss": 0.8561,
       "step": 25449
     },
     {
-      "epoch": 0.69896459861031,
+      "epoch": 0.7221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 4.38830706140495e-06,
-      "loss": 0.801,
+      "learning_rate": 3.7821570128675323e-06,
+      "loss": 0.7919,
       "step": 25450
     },
     {
-      "epoch": 0.6989920628381533,
+      "epoch": 0.7222190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 4.387570830271615e-06,
-      "loss": 0.7754,
+      "learning_rate": 3.781437227721111e-06,
+      "loss": 0.7308,
       "step": 25451
     },
     {
-      "epoch": 0.6990195270659966,
+      "epoch": 0.7222474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 4.386834643545999e-06,
-      "loss": 0.8396,
+      "learning_rate": 3.780717495103168e-06,
+      "loss": 0.8889,
       "step": 25452
     },
     {
-      "epoch": 0.6990469912938397,
+      "epoch": 0.7222758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 4.386098501233916e-06,
-      "loss": 0.845,
+      "learning_rate": 3.7799978150197867e-06,
+      "loss": 0.8182,
       "step": 25453
     },
     {
-      "epoch": 0.699074455521683,
+      "epoch": 0.7223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 4.3853624033412015e-06,
-      "loss": 0.8656,
+      "learning_rate": 3.77927818747704e-06,
+      "loss": 0.757,
       "step": 25454
     },
     {
-      "epoch": 0.6991019197495263,
+      "epoch": 0.7223325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 4.3846263498736705e-06,
-      "loss": 0.8693,
+      "learning_rate": 3.7785586124810113e-06,
+      "loss": 0.949,
       "step": 25455
     },
     {
-      "epoch": 0.6991293839773695,
+      "epoch": 0.722360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 4.383890340837151e-06,
-      "loss": 0.8061,
+      "learning_rate": 3.7778390900377804e-06,
+      "loss": 0.8183,
       "step": 25456
     },
     {
-      "epoch": 0.6991568482052127,
+      "epoch": 0.7223893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 4.383154376237468e-06,
-      "loss": 0.8747,
+      "learning_rate": 3.77711962015342e-06,
+      "loss": 0.85,
       "step": 25457
     },
     {
-      "epoch": 0.6991843124330559,
+      "epoch": 0.7224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 4.382418456080441e-06,
-      "loss": 0.8603,
+      "learning_rate": 3.7764002028340086e-06,
+      "loss": 0.7765,
       "step": 25458
     },
     {
-      "epoch": 0.6992117766608992,
+      "epoch": 0.7224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 4.3816825803719e-06,
-      "loss": 0.778,
+      "learning_rate": 3.775680838085628e-06,
+      "loss": 0.8144,
       "step": 25459
     },
     {
-      "epoch": 0.6992392408887425,
+      "epoch": 0.7224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 4.380946749117663e-06,
-      "loss": 0.7931,
+      "learning_rate": 3.7749615259143445e-06,
+      "loss": 0.7186,
       "step": 25460
     },
     {
-      "epoch": 0.6992667051165856,
+      "epoch": 0.7225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 4.380210962323547e-06,
-      "loss": 0.7939,
+      "learning_rate": 3.7742422663262445e-06,
+      "loss": 0.8745,
       "step": 25461
     },
     {
-      "epoch": 0.6992941693444289,
+      "epoch": 0.7225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 4.379475219995379e-06,
-      "loss": 0.7938,
+      "learning_rate": 3.773523059327403e-06,
+      "loss": 0.8628,
       "step": 25462
     },
     {
-      "epoch": 0.6993216335722721,
+      "epoch": 0.7225595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 4.378739522138981e-06,
-      "loss": 0.907,
+      "learning_rate": 3.7728039049238896e-06,
+      "loss": 0.8097,
       "step": 25463
     },
     {
-      "epoch": 0.6993490978001153,
+      "epoch": 0.7225879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 4.378003868760174e-06,
-      "loss": 0.8009,
+      "learning_rate": 3.7720848031217804e-06,
+      "loss": 0.8365,
       "step": 25464
     },
     {
-      "epoch": 0.6993765620279586,
+      "epoch": 0.7226163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 4.377268259864777e-06,
-      "loss": 0.8542,
+      "learning_rate": 3.7713657539271553e-06,
+      "loss": 0.7262,
       "step": 25465
     },
     {
-      "epoch": 0.6994040262558018,
+      "epoch": 0.722644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 4.376532695458616e-06,
-      "loss": 0.8035,
+      "learning_rate": 3.770646757346079e-06,
+      "loss": 0.8112,
       "step": 25466
     },
     {
-      "epoch": 0.6994314904836451,
+      "epoch": 0.7226730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 4.375797175547503e-06,
-      "loss": 0.7996,
+      "learning_rate": 3.7699278133846306e-06,
+      "loss": 0.9166,
       "step": 25467
     },
     {
-      "epoch": 0.6994589547114883,
+      "epoch": 0.7227014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 4.3750617001372645e-06,
-      "loss": 0.7854,
+      "learning_rate": 3.769208922048886e-06,
+      "loss": 0.7807,
       "step": 25468
     },
     {
-      "epoch": 0.6994864189393315,
+      "epoch": 0.7227298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 4.374326269233713e-06,
-      "loss": 0.7629,
+      "learning_rate": 3.768490083344909e-06,
+      "loss": 0.8814,
       "step": 25469
     },
     {
-      "epoch": 0.6995138831671748,
+      "epoch": 0.7227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 4.373590882842671e-06,
-      "loss": 0.8458,
+      "learning_rate": 3.7677712972787807e-06,
+      "loss": 0.9192,
       "step": 25470
     },
     {
-      "epoch": 0.6995413473950179,
+      "epoch": 0.7227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 4.372855540969958e-06,
-      "loss": 0.8876,
+      "learning_rate": 3.7670525638565636e-06,
+      "loss": 0.7991,
       "step": 25471
     },
     {
-      "epoch": 0.6995688116228612,
+      "epoch": 0.7228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 4.372120243621395e-06,
-      "loss": 0.8214,
+      "learning_rate": 3.7663338830843353e-06,
+      "loss": 0.8321,
       "step": 25472
     },
     {
-      "epoch": 0.6995962758507045,
+      "epoch": 0.7228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 4.371384990802794e-06,
-      "loss": 0.9131,
+      "learning_rate": 3.7656152549681668e-06,
+      "loss": 0.9136,
       "step": 25473
     },
     {
-      "epoch": 0.6996237400785477,
+      "epoch": 0.7228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 4.3706497825199746e-06,
-      "loss": 0.8994,
+      "learning_rate": 3.764896679514124e-06,
+      "loss": 0.917,
       "step": 25474
     },
     {
-      "epoch": 0.6996512043063909,
+      "epoch": 0.7229001135073779,
       "grad_norm": 0.0,
-      "learning_rate": 4.369914618778754e-06,
-      "loss": 0.8901,
+      "learning_rate": 3.7641781567282767e-06,
+      "loss": 0.7642,
       "step": 25475
     },
     {
-      "epoch": 0.6996786685342341,
+      "epoch": 0.7229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 4.369179499584955e-06,
-      "loss": 0.8885,
+      "learning_rate": 3.7634596866167015e-06,
+      "loss": 0.8276,
       "step": 25476
     },
     {
-      "epoch": 0.6997061327620774,
+      "epoch": 0.7229568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 4.368444424944384e-06,
-      "loss": 0.8776,
+      "learning_rate": 3.7627412691854613e-06,
+      "loss": 0.833,
       "step": 25477
     },
     {
-      "epoch": 0.6997335969899207,
+      "epoch": 0.7229852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 4.367709394862865e-06,
-      "loss": 0.7616,
+      "learning_rate": 3.7620229044406255e-06,
+      "loss": 0.9004,
       "step": 25478
     },
     {
-      "epoch": 0.6997610612177638,
+      "epoch": 0.7230136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 4.36697440934621e-06,
-      "loss": 0.8877,
+      "learning_rate": 3.761304592388265e-06,
+      "loss": 0.8518,
       "step": 25479
     },
     {
-      "epoch": 0.6997885254456071,
+      "epoch": 0.7230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 4.366239468400233e-06,
-      "loss": 0.8501,
+      "learning_rate": 3.760586333034443e-06,
+      "loss": 0.8917,
       "step": 25480
     },
     {
-      "epoch": 0.6998159896734504,
+      "epoch": 0.7230703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 4.365504572030753e-06,
-      "loss": 0.7939,
+      "learning_rate": 3.7598681263852276e-06,
+      "loss": 0.8375,
       "step": 25481
     },
     {
-      "epoch": 0.6998434539012935,
+      "epoch": 0.7230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 4.364769720243583e-06,
-      "loss": 0.8502,
+      "learning_rate": 3.7591499724466906e-06,
+      "loss": 0.8416,
       "step": 25482
     },
     {
-      "epoch": 0.6998709181291368,
+      "epoch": 0.7231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 4.364034913044541e-06,
-      "loss": 0.9095,
+      "learning_rate": 3.7584318712248915e-06,
+      "loss": 0.8394,
       "step": 25483
     },
     {
-      "epoch": 0.69989838235698,
+      "epoch": 0.723155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 4.363300150439439e-06,
-      "loss": 0.7888,
+      "learning_rate": 3.757713822725898e-06,
+      "loss": 0.8691,
       "step": 25484
     },
     {
-      "epoch": 0.6999258465848233,
+      "epoch": 0.7231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 4.362565432434086e-06,
-      "loss": 0.739,
+      "learning_rate": 3.7569958269557806e-06,
+      "loss": 0.9334,
       "step": 25485
     },
     {
-      "epoch": 0.6999533108126665,
+      "epoch": 0.7232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 4.361830759034299e-06,
-      "loss": 0.7967,
+      "learning_rate": 3.756277883920596e-06,
+      "loss": 0.8469,
       "step": 25486
     },
     {
-      "epoch": 0.6999807750405097,
+      "epoch": 0.7232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 4.36109613024589e-06,
-      "loss": 0.8756,
+      "learning_rate": 3.755559993626413e-06,
+      "loss": 0.8536,
       "step": 25487
     },
     {
-      "epoch": 0.700008239268353,
+      "epoch": 0.7232690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 4.360361546074674e-06,
-      "loss": 0.7714,
+      "learning_rate": 3.7548421560792993e-06,
+      "loss": 0.7047,
       "step": 25488
     },
     {
-      "epoch": 0.7000357034961963,
+      "epoch": 0.7232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 4.359627006526461e-06,
-      "loss": 0.8745,
+      "learning_rate": 3.7541243712853105e-06,
+      "loss": 0.7857,
       "step": 25489
     },
     {
-      "epoch": 0.7000631677240394,
+      "epoch": 0.7233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 4.358892511607068e-06,
-      "loss": 0.9783,
+      "learning_rate": 3.753406639250514e-06,
+      "loss": 0.7817,
       "step": 25490
     },
     {
-      "epoch": 0.7000906319518827,
+      "epoch": 0.7233541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 4.358158061322299e-06,
-      "loss": 0.828,
+      "learning_rate": 3.7526889599809725e-06,
+      "loss": 0.8794,
       "step": 25491
     },
     {
-      "epoch": 0.7001180961797259,
+      "epoch": 0.7233825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 4.357423655677973e-06,
-      "loss": 0.8382,
+      "learning_rate": 3.751971333482749e-06,
+      "loss": 0.9122,
       "step": 25492
     },
     {
-      "epoch": 0.7001455604075691,
+      "epoch": 0.7234108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 4.356689294679892e-06,
-      "loss": 0.8643,
+      "learning_rate": 3.751253759761907e-06,
+      "loss": 0.764,
       "step": 25493
     },
     {
-      "epoch": 0.7001730246354124,
+      "epoch": 0.7234392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 4.355954978333872e-06,
-      "loss": 0.7731,
+      "learning_rate": 3.7505362388245026e-06,
+      "loss": 0.806,
       "step": 25494
     },
     {
-      "epoch": 0.7002004888632556,
+      "epoch": 0.7234676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 4.355220706645722e-06,
-      "loss": 0.831,
+      "learning_rate": 3.7498187706765987e-06,
+      "loss": 0.882,
       "step": 25495
     },
     {
-      "epoch": 0.7002279530910989,
+      "epoch": 0.7234960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 4.354486479621257e-06,
-      "loss": 0.8012,
+      "learning_rate": 3.7491013553242605e-06,
+      "loss": 0.9651,
       "step": 25496
     },
     {
-      "epoch": 0.700255417318942,
+      "epoch": 0.7235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 4.353752297266277e-06,
-      "loss": 0.8208,
+      "learning_rate": 3.748383992773541e-06,
+      "loss": 0.8041,
       "step": 25497
     },
     {
-      "epoch": 0.7002828815467853,
+      "epoch": 0.7235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 4.353018159586597e-06,
-      "loss": 0.7884,
+      "learning_rate": 3.747666683030503e-06,
+      "loss": 0.8912,
       "step": 25498
     },
     {
-      "epoch": 0.7003103457746286,
+      "epoch": 0.7235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 4.352284066588028e-06,
-      "loss": 0.8153,
+      "learning_rate": 3.7469494261012086e-06,
+      "loss": 0.9097,
       "step": 25499
     },
     {
-      "epoch": 0.7003378100024718,
+      "epoch": 0.7236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 4.351550018276371e-06,
-      "loss": 0.9103,
+      "learning_rate": 3.74623222199171e-06,
+      "loss": 0.9041,
       "step": 25500
     },
     {
-      "epoch": 0.700365274230315,
+      "epoch": 0.7236379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 4.35081601465744e-06,
-      "loss": 0.8669,
+      "learning_rate": 3.745515070708069e-06,
+      "loss": 0.876,
       "step": 25501
     },
     {
-      "epoch": 0.7003927384581583,
+      "epoch": 0.72366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 4.350082055737043e-06,
-      "loss": 0.8536,
+      "learning_rate": 3.7447979722563464e-06,
+      "loss": 0.8946,
       "step": 25502
     },
     {
-      "epoch": 0.7004202026860015,
+      "epoch": 0.723694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 4.349348141520981e-06,
-      "loss": 1.0078,
+      "learning_rate": 3.744080926642594e-06,
+      "loss": 0.8454,
       "step": 25503
     },
     {
-      "epoch": 0.7004476669138447,
+      "epoch": 0.7237230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 4.348614272015067e-06,
-      "loss": 0.8034,
+      "learning_rate": 3.7433639338728712e-06,
+      "loss": 0.7555,
       "step": 25504
     },
     {
-      "epoch": 0.7004751311416879,
+      "epoch": 0.7237514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 4.347880447225106e-06,
-      "loss": 0.8541,
+      "learning_rate": 3.7426469939532374e-06,
+      "loss": 0.7565,
       "step": 25505
     },
     {
-      "epoch": 0.7005025953695312,
+      "epoch": 0.7237797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 4.347146667156904e-06,
-      "loss": 0.8569,
+      "learning_rate": 3.7419301068897398e-06,
+      "loss": 0.8708,
       "step": 25506
     },
     {
-      "epoch": 0.7005300595973745,
+      "epoch": 0.7238081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 4.34641293181627e-06,
-      "loss": 0.7267,
+      "learning_rate": 3.7412132726884452e-06,
+      "loss": 0.8557,
       "step": 25507
     },
     {
-      "epoch": 0.7005575238252176,
+      "epoch": 0.7238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 4.345679241209007e-06,
-      "loss": 0.831,
+      "learning_rate": 3.740496491355401e-06,
+      "loss": 0.8187,
       "step": 25508
     },
     {
-      "epoch": 0.7005849880530609,
+      "epoch": 0.7238649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 4.344945595340918e-06,
-      "loss": 0.7333,
+      "learning_rate": 3.739779762896664e-06,
+      "loss": 0.7972,
       "step": 25509
     },
     {
-      "epoch": 0.7006124522809041,
+      "epoch": 0.7238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 4.344211994217809e-06,
-      "loss": 0.9208,
+      "learning_rate": 3.739063087318292e-06,
+      "loss": 0.9059,
       "step": 25510
     },
     {
-      "epoch": 0.7006399165087474,
+      "epoch": 0.7239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 4.343478437845485e-06,
-      "loss": 0.8922,
+      "learning_rate": 3.7383464646263324e-06,
+      "loss": 0.881,
       "step": 25511
     },
     {
-      "epoch": 0.7006673807365906,
+      "epoch": 0.723950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 4.342744926229751e-06,
-      "loss": 0.7842,
+      "learning_rate": 3.737629894826842e-06,
+      "loss": 0.8707,
       "step": 25512
     },
     {
-      "epoch": 0.7006948449644338,
+      "epoch": 0.7239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 4.342011459376415e-06,
-      "loss": 0.8079,
+      "learning_rate": 3.736913377925877e-06,
+      "loss": 0.7188,
       "step": 25513
     },
     {
-      "epoch": 0.7007223091922771,
+      "epoch": 0.7240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 4.341278037291272e-06,
-      "loss": 0.8529,
+      "learning_rate": 3.736196913929482e-06,
+      "loss": 0.7973,
       "step": 25514
     },
     {
-      "epoch": 0.7007497734201203,
+      "epoch": 0.7240351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 4.3405446599801295e-06,
-      "loss": 0.8565,
+      "learning_rate": 3.7354805028437137e-06,
+      "loss": 0.7623,
       "step": 25515
     },
     {
-      "epoch": 0.7007772376479635,
+      "epoch": 0.7240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 4.339811327448793e-06,
-      "loss": 0.8735,
+      "learning_rate": 3.7347641446746263e-06,
+      "loss": 0.7191,
       "step": 25516
     },
     {
-      "epoch": 0.7008047018758068,
+      "epoch": 0.7240919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 4.339078039703059e-06,
-      "loss": 0.735,
+      "learning_rate": 3.7340478394282655e-06,
+      "loss": 0.8607,
       "step": 25517
     },
     {
-      "epoch": 0.70083216610365,
+      "epoch": 0.7241203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 4.338344796748731e-06,
-      "loss": 0.8351,
+      "learning_rate": 3.7333315871106833e-06,
+      "loss": 0.8317,
       "step": 25518
     },
     {
-      "epoch": 0.7008596303314932,
+      "epoch": 0.7241486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 4.337611598591617e-06,
-      "loss": 0.8294,
+      "learning_rate": 3.732615387727934e-06,
+      "loss": 0.8729,
       "step": 25519
     },
     {
-      "epoch": 0.7008870945593365,
+      "epoch": 0.7241770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 4.336878445237509e-06,
-      "loss": 0.8045,
+      "learning_rate": 3.731899241286061e-06,
+      "loss": 0.7741,
       "step": 25520
     },
     {
-      "epoch": 0.7009145587871797,
+      "epoch": 0.724205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 4.336145336692212e-06,
-      "loss": 0.844,
+      "learning_rate": 3.7311831477911177e-06,
+      "loss": 0.8259,
       "step": 25521
     },
     {
-      "epoch": 0.700942023015023,
+      "epoch": 0.7242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 4.335412272961528e-06,
-      "loss": 0.8884,
+      "learning_rate": 3.7304671072491517e-06,
+      "loss": 0.7681,
       "step": 25522
     },
     {
-      "epoch": 0.7009694872428661,
+      "epoch": 0.7242622020431329,
       "grad_norm": 0.0,
-      "learning_rate": 4.3346792540512596e-06,
-      "loss": 0.8481,
+      "learning_rate": 3.729751119666212e-06,
+      "loss": 0.8356,
       "step": 25523
     },
     {
-      "epoch": 0.7009969514707094,
+      "epoch": 0.7242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 4.333946279967201e-06,
-      "loss": 0.9084,
+      "learning_rate": 3.7290351850483497e-06,
+      "loss": 0.858,
       "step": 25524
     },
     {
-      "epoch": 0.7010244156985527,
+      "epoch": 0.7243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 4.3332133507151565e-06,
-      "loss": 0.7717,
+      "learning_rate": 3.728319303401605e-06,
+      "loss": 0.8499,
       "step": 25525
     },
     {
-      "epoch": 0.7010518799263958,
+      "epoch": 0.7243473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 4.332480466300919e-06,
-      "loss": 0.9102,
+      "learning_rate": 3.72760347473203e-06,
+      "loss": 0.8158,
       "step": 25526
     },
     {
-      "epoch": 0.7010793441542391,
+      "epoch": 0.7243757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 4.3317476267302925e-06,
-      "loss": 0.9849,
+      "learning_rate": 3.7268876990456735e-06,
+      "loss": 0.7705,
       "step": 25527
     },
     {
-      "epoch": 0.7011068083820824,
+      "epoch": 0.7244040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 4.3310148320090736e-06,
-      "loss": 0.8703,
+      "learning_rate": 3.726171976348575e-06,
+      "loss": 0.884,
       "step": 25528
     },
     {
-      "epoch": 0.7011342726099256,
+      "epoch": 0.7244324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 4.330282082143062e-06,
-      "loss": 0.9108,
+      "learning_rate": 3.725456306646784e-06,
+      "loss": 0.8708,
       "step": 25529
     },
     {
-      "epoch": 0.7011617368377688,
+      "epoch": 0.7244608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 4.329549377138055e-06,
-      "loss": 0.8114,
+      "learning_rate": 3.724740689946349e-06,
+      "loss": 0.8355,
       "step": 25530
     },
     {
-      "epoch": 0.701189201065612,
+      "epoch": 0.724489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 4.328816716999853e-06,
-      "loss": 0.8628,
+      "learning_rate": 3.7240251262533087e-06,
+      "loss": 0.861,
       "step": 25531
     },
     {
-      "epoch": 0.7012166652934553,
+      "epoch": 0.7245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 4.3280841017342504e-06,
-      "loss": 0.819,
+      "learning_rate": 3.7233096155737092e-06,
+      "loss": 0.8034,
       "step": 25532
     },
     {
-      "epoch": 0.7012441295212986,
+      "epoch": 0.7245459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 4.3273515313470395e-06,
-      "loss": 0.7305,
+      "learning_rate": 3.7225941579136004e-06,
+      "loss": 0.8597,
       "step": 25533
     },
     {
-      "epoch": 0.7012715937491417,
+      "epoch": 0.7245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 4.3266190058440205e-06,
-      "loss": 0.7943,
+      "learning_rate": 3.7218787532790167e-06,
+      "loss": 0.8176,
       "step": 25534
     },
     {
-      "epoch": 0.701299057976985,
+      "epoch": 0.7246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 4.32588652523099e-06,
-      "loss": 0.894,
+      "learning_rate": 3.7211634016760045e-06,
+      "loss": 0.7951,
       "step": 25535
     },
     {
-      "epoch": 0.7013265222048282,
+      "epoch": 0.7246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 4.325154089513742e-06,
-      "loss": 0.7556,
+      "learning_rate": 3.720448103110612e-06,
+      "loss": 0.8433,
       "step": 25536
     },
     {
-      "epoch": 0.7013539864326714,
+      "epoch": 0.7246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 4.324421698698078e-06,
-      "loss": 0.8319,
+      "learning_rate": 3.719732857588869e-06,
+      "loss": 0.8429,
       "step": 25537
     },
     {
-      "epoch": 0.7013814506605147,
+      "epoch": 0.7246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 4.323689352789785e-06,
-      "loss": 0.9076,
+      "learning_rate": 3.719017665116833e-06,
+      "loss": 0.8353,
       "step": 25538
     },
     {
-      "epoch": 0.7014089148883579,
+      "epoch": 0.724716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 4.322957051794659e-06,
-      "loss": 0.9227,
+      "learning_rate": 3.7183025257005324e-06,
+      "loss": 0.9055,
       "step": 25539
     },
     {
-      "epoch": 0.7014363791162012,
+      "epoch": 0.724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 4.3222247957185e-06,
-      "loss": 0.7334,
+      "learning_rate": 3.717587439346013e-06,
+      "loss": 0.8192,
       "step": 25540
     },
     {
-      "epoch": 0.7014638433440444,
+      "epoch": 0.7247729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 4.321492584567096e-06,
-      "loss": 0.798,
+      "learning_rate": 3.7168724060593186e-06,
+      "loss": 0.7605,
       "step": 25541
     },
     {
-      "epoch": 0.7014913075718876,
+      "epoch": 0.7248013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 4.3207604183462395e-06,
-      "loss": 0.8532,
+      "learning_rate": 3.7161574258464817e-06,
+      "loss": 0.8816,
       "step": 25542
     },
     {
-      "epoch": 0.7015187717997309,
+      "epoch": 0.7248297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 4.3200282970617315e-06,
-      "loss": 0.9457,
+      "learning_rate": 3.7154424987135463e-06,
+      "loss": 0.7397,
       "step": 25543
     },
     {
-      "epoch": 0.701546236027574,
+      "epoch": 0.7248581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 4.319296220719356e-06,
-      "loss": 0.9265,
+      "learning_rate": 3.714727624666553e-06,
+      "loss": 0.752,
       "step": 25544
     },
     {
-      "epoch": 0.7015737002554173,
+      "epoch": 0.7248864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 4.31856418932491e-06,
-      "loss": 0.9384,
+      "learning_rate": 3.714012803711535e-06,
+      "loss": 0.7585,
       "step": 25545
     },
     {
-      "epoch": 0.7016011644832606,
+      "epoch": 0.7249148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 4.317832202884183e-06,
-      "loss": 0.8865,
+      "learning_rate": 3.7132980358545333e-06,
+      "loss": 0.7966,
       "step": 25546
     },
     {
-      "epoch": 0.7016286287111038,
+      "epoch": 0.7249432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 4.317100261402975e-06,
-      "loss": 0.7957,
+      "learning_rate": 3.7125833211015895e-06,
+      "loss": 0.8606,
       "step": 25547
     },
     {
-      "epoch": 0.701656092938947,
+      "epoch": 0.7249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 4.3163683648870645e-06,
-      "loss": 0.8399,
+      "learning_rate": 3.711868659458734e-06,
+      "loss": 0.8127,
       "step": 25548
     },
     {
-      "epoch": 0.7016835571667902,
+      "epoch": 0.725,
       "grad_norm": 0.0,
-      "learning_rate": 4.315636513342255e-06,
-      "loss": 0.7827,
+      "learning_rate": 3.7111540509320053e-06,
+      "loss": 0.8062,
       "step": 25549
     },
     {
-      "epoch": 0.7017110213946335,
+      "epoch": 0.7250283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 4.314904706774327e-06,
-      "loss": 0.8684,
+      "learning_rate": 3.710439495527446e-06,
+      "loss": 0.8076,
       "step": 25550
     },
     {
-      "epoch": 0.7017384856224768,
+      "epoch": 0.7250567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 4.314172945189074e-06,
-      "loss": 0.8882,
+      "learning_rate": 3.7097249932510828e-06,
+      "loss": 0.7963,
       "step": 25551
     },
     {
-      "epoch": 0.7017659498503199,
+      "epoch": 0.7250851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 4.313441228592289e-06,
-      "loss": 0.8613,
+      "learning_rate": 3.709010544108954e-06,
+      "loss": 0.8741,
       "step": 25552
     },
     {
-      "epoch": 0.7017934140781632,
+      "epoch": 0.7251135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 4.3127095569897594e-06,
-      "loss": 0.7583,
+      "learning_rate": 3.7082961481070977e-06,
+      "loss": 0.8351,
       "step": 25553
     },
     {
-      "epoch": 0.7018208783060065,
+      "epoch": 0.7251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 4.311977930387274e-06,
-      "loss": 0.8509,
+      "learning_rate": 3.7075818052515466e-06,
+      "loss": 0.8656,
       "step": 25554
     },
     {
-      "epoch": 0.7018483425338496,
+      "epoch": 0.7251702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 4.311246348790627e-06,
-      "loss": 0.8778,
+      "learning_rate": 3.706867515548338e-06,
+      "loss": 0.8845,
       "step": 25555
     },
     {
-      "epoch": 0.7018758067616929,
+      "epoch": 0.7251986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 4.310514812205602e-06,
-      "loss": 0.7829,
+      "learning_rate": 3.706153279003498e-06,
+      "loss": 0.9102,
       "step": 25556
     },
     {
-      "epoch": 0.7019032709895361,
+      "epoch": 0.7252270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 4.309783320637985e-06,
-      "loss": 0.8451,
+      "learning_rate": 3.7054390956230658e-06,
+      "loss": 0.854,
       "step": 25557
     },
     {
-      "epoch": 0.7019307352173794,
+      "epoch": 0.725255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 4.309051874093565e-06,
-      "loss": 0.6754,
+      "learning_rate": 3.704724965413075e-06,
+      "loss": 0.7562,
       "step": 25558
     },
     {
-      "epoch": 0.7019581994452226,
+      "epoch": 0.7252837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 4.308320472578131e-06,
-      "loss": 0.8894,
+      "learning_rate": 3.7040108883795522e-06,
+      "loss": 0.7141,
       "step": 25559
     },
     {
-      "epoch": 0.7019856636730658,
+      "epoch": 0.7253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 4.3075891160974714e-06,
-      "loss": 0.8961,
+      "learning_rate": 3.703296864528532e-06,
+      "loss": 0.898,
       "step": 25560
     },
     {
-      "epoch": 0.7020131279009091,
+      "epoch": 0.7253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 4.306857804657375e-06,
-      "loss": 0.8357,
+      "learning_rate": 3.70258289386605e-06,
+      "loss": 0.8872,
       "step": 25561
     },
     {
-      "epoch": 0.7020405921287522,
+      "epoch": 0.7253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 4.306126538263621e-06,
-      "loss": 0.8034,
+      "learning_rate": 3.70186897639813e-06,
+      "loss": 0.7741,
       "step": 25562
     },
     {
-      "epoch": 0.7020680563565955,
+      "epoch": 0.7253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 4.3053953169219994e-06,
-      "loss": 0.8198,
+      "learning_rate": 3.7011551121308065e-06,
+      "loss": 0.8863,
       "step": 25563
     },
     {
-      "epoch": 0.7020955205844388,
+      "epoch": 0.7254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 4.304664140638301e-06,
-      "loss": 0.8066,
+      "learning_rate": 3.7004413010701114e-06,
+      "loss": 0.861,
       "step": 25564
     },
     {
-      "epoch": 0.702122984812282,
+      "epoch": 0.7254540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 4.303933009418301e-06,
-      "loss": 0.8545,
+      "learning_rate": 3.699727543222069e-06,
+      "loss": 0.7999,
       "step": 25565
     },
     {
-      "epoch": 0.7021504490401252,
+      "epoch": 0.7254824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 4.303201923267792e-06,
-      "loss": 0.8325,
+      "learning_rate": 3.699013838592711e-06,
+      "loss": 0.8342,
       "step": 25566
     },
     {
-      "epoch": 0.7021779132679685,
+      "epoch": 0.725510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 4.302470882192554e-06,
-      "loss": 0.9371,
+      "learning_rate": 3.6983001871880665e-06,
+      "loss": 0.8675,
       "step": 25567
     },
     {
-      "epoch": 0.7022053774958117,
+      "epoch": 0.725539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 4.301739886198374e-06,
-      "loss": 0.79,
+      "learning_rate": 3.6975865890141627e-06,
+      "loss": 0.8566,
       "step": 25568
     },
     {
-      "epoch": 0.702232841723655,
+      "epoch": 0.7255675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 4.3010089352910345e-06,
-      "loss": 0.9045,
+      "learning_rate": 3.6968730440770294e-06,
+      "loss": 0.8417,
       "step": 25569
     },
     {
-      "epoch": 0.7022603059514981,
+      "epoch": 0.7255959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 4.3002780294763205e-06,
-      "loss": 0.8189,
+      "learning_rate": 3.6961595523826955e-06,
+      "loss": 0.7403,
       "step": 25570
     },
     {
-      "epoch": 0.7022877701793414,
+      "epoch": 0.7256242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 4.2995471687600165e-06,
-      "loss": 0.776,
+      "learning_rate": 3.6954461139371823e-06,
+      "loss": 0.8014,
       "step": 25571
     },
     {
-      "epoch": 0.7023152344071847,
+      "epoch": 0.7256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 4.298816353147901e-06,
-      "loss": 0.879,
+      "learning_rate": 3.6947327287465195e-06,
+      "loss": 0.9447,
       "step": 25572
     },
     {
-      "epoch": 0.7023426986350279,
+      "epoch": 0.7256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 4.2980855826457635e-06,
-      "loss": 0.8423,
+      "learning_rate": 3.694019396816736e-06,
+      "loss": 0.8667,
       "step": 25573
     },
     {
-      "epoch": 0.7023701628628711,
+      "epoch": 0.7257094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 4.297354857259377e-06,
-      "loss": 0.9484,
+      "learning_rate": 3.69330611815385e-06,
+      "loss": 0.8698,
       "step": 25574
     },
     {
-      "epoch": 0.7023976270907143,
+      "epoch": 0.7257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 4.296624176994527e-06,
-      "loss": 0.8058,
+      "learning_rate": 3.692592892763891e-06,
+      "loss": 0.8101,
       "step": 25575
     },
     {
-      "epoch": 0.7024250913185576,
+      "epoch": 0.7257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 4.295893541856997e-06,
-      "loss": 0.9079,
+      "learning_rate": 3.691879720652888e-06,
+      "loss": 0.8556,
       "step": 25576
     },
     {
-      "epoch": 0.7024525555464008,
+      "epoch": 0.7257945516458569,
       "grad_norm": 0.0,
-      "learning_rate": 4.2951629518525666e-06,
-      "loss": 0.896,
+      "learning_rate": 3.6911666018268556e-06,
+      "loss": 0.7301,
       "step": 25577
     },
     {
-      "epoch": 0.702480019774244,
+      "epoch": 0.7258229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 4.29443240698702e-06,
-      "loss": 0.8185,
+      "learning_rate": 3.690453536291827e-06,
+      "loss": 0.8347,
       "step": 25578
     },
     {
-      "epoch": 0.7025074840020873,
+      "epoch": 0.7258513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 4.293701907266131e-06,
-      "loss": 0.8285,
+      "learning_rate": 3.6897405240538163e-06,
+      "loss": 0.8292,
       "step": 25579
     },
     {
-      "epoch": 0.7025349482299306,
+      "epoch": 0.7258796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 4.292971452695686e-06,
-      "loss": 0.7619,
+      "learning_rate": 3.6890275651188524e-06,
+      "loss": 0.7069,
       "step": 25580
     },
     {
-      "epoch": 0.7025624124577737,
+      "epoch": 0.7259080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 4.292241043281459e-06,
-      "loss": 0.8383,
+      "learning_rate": 3.68831465949296e-06,
+      "loss": 0.8147,
       "step": 25581
     },
     {
-      "epoch": 0.702589876685617,
+      "epoch": 0.7259364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 4.291510679029232e-06,
-      "loss": 0.8212,
+      "learning_rate": 3.6876018071821527e-06,
+      "loss": 0.8499,
       "step": 25582
     },
     {
-      "epoch": 0.7026173409134602,
+      "epoch": 0.7259648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 4.290780359944783e-06,
-      "loss": 0.8369,
+      "learning_rate": 3.6868890081924537e-06,
+      "loss": 0.765,
       "step": 25583
     },
     {
-      "epoch": 0.7026448051413035,
+      "epoch": 0.7259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 4.290050086033894e-06,
-      "loss": 0.7118,
+      "learning_rate": 3.6861762625298937e-06,
+      "loss": 0.9778,
       "step": 25584
     },
     {
-      "epoch": 0.7026722693691467,
+      "epoch": 0.7260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 4.289319857302338e-06,
-      "loss": 0.8382,
+      "learning_rate": 3.6854635702004837e-06,
+      "loss": 0.8372,
       "step": 25585
     },
     {
-      "epoch": 0.7026997335969899,
+      "epoch": 0.726049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 4.2885896737558965e-06,
-      "loss": 0.9283,
+      "learning_rate": 3.684750931210247e-06,
+      "loss": 0.8114,
       "step": 25586
     },
     {
-      "epoch": 0.7027271978248332,
+      "epoch": 0.7260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 4.287859535400344e-06,
-      "loss": 0.8747,
+      "learning_rate": 3.6840383455652063e-06,
+      "loss": 0.7799,
       "step": 25587
     },
     {
-      "epoch": 0.7027546620526763,
+      "epoch": 0.7261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 4.287129442241464e-06,
-      "loss": 0.8261,
+      "learning_rate": 3.6833258132713736e-06,
+      "loss": 0.769,
       "step": 25588
     },
     {
-      "epoch": 0.7027821262805196,
+      "epoch": 0.7261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 4.286399394285024e-06,
-      "loss": 0.7961,
+      "learning_rate": 3.6826133343347716e-06,
+      "loss": 0.8428,
       "step": 25589
     },
     {
-      "epoch": 0.7028095905083629,
+      "epoch": 0.7261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 4.2856693915368104e-06,
-      "loss": 0.8285,
+      "learning_rate": 3.681900908761423e-06,
+      "loss": 0.8303,
       "step": 25590
     },
     {
-      "epoch": 0.7028370547362061,
+      "epoch": 0.7261918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 4.284939434002591e-06,
-      "loss": 0.9196,
+      "learning_rate": 3.6811885365573374e-06,
+      "loss": 0.7485,
       "step": 25591
     },
     {
-      "epoch": 0.7028645189640493,
+      "epoch": 0.7262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 4.284209521688143e-06,
-      "loss": 0.8549,
+      "learning_rate": 3.680476217728537e-06,
+      "loss": 0.8571,
       "step": 25592
     },
     {
-      "epoch": 0.7028919831918926,
+      "epoch": 0.7262485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 4.283479654599244e-06,
-      "loss": 0.9148,
+      "learning_rate": 3.6797639522810413e-06,
+      "loss": 0.7429,
       "step": 25593
     },
     {
-      "epoch": 0.7029194474197358,
+      "epoch": 0.7262769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 4.282749832741668e-06,
-      "loss": 0.7972,
+      "learning_rate": 3.679051740220859e-06,
+      "loss": 0.8516,
       "step": 25594
     },
     {
-      "epoch": 0.702946911647579,
+      "epoch": 0.7263053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 4.282020056121194e-06,
-      "loss": 0.8435,
+      "learning_rate": 3.6783395815540124e-06,
+      "loss": 0.9957,
       "step": 25595
     },
     {
-      "epoch": 0.7029743758754222,
+      "epoch": 0.7263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 4.281290324743589e-06,
-      "loss": 0.8818,
+      "learning_rate": 3.677627476286518e-06,
+      "loss": 0.8201,
       "step": 25596
     },
     {
-      "epoch": 0.7030018401032655,
+      "epoch": 0.7263620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 4.280560638614634e-06,
-      "loss": 0.891,
+      "learning_rate": 3.6769154244243855e-06,
+      "loss": 0.796,
       "step": 25597
     },
     {
-      "epoch": 0.7030293043311088,
+      "epoch": 0.7263904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 4.279830997740093e-06,
-      "loss": 0.8243,
+      "learning_rate": 3.6762034259736323e-06,
+      "loss": 0.8056,
       "step": 25598
     },
     {
-      "epoch": 0.7030567685589519,
+      "epoch": 0.7264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 4.279101402125747e-06,
-      "loss": 0.7398,
+      "learning_rate": 3.675491480940274e-06,
+      "loss": 0.698,
       "step": 25599
     },
     {
-      "epoch": 0.7030842327867952,
+      "epoch": 0.7264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 4.278371851777366e-06,
-      "loss": 0.7403,
+      "learning_rate": 3.674779589330323e-06,
+      "loss": 0.8085,
       "step": 25600
     },
     {
-      "epoch": 0.7031116970146384,
+      "epoch": 0.7264755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 4.277642346700724e-06,
-      "loss": 0.8928,
+      "learning_rate": 3.6740677511497958e-06,
+      "loss": 0.8412,
       "step": 25601
     },
     {
-      "epoch": 0.7031391612424817,
+      "epoch": 0.7265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 4.2769128869015954e-06,
-      "loss": 0.7882,
+      "learning_rate": 3.6733559664047005e-06,
+      "loss": 0.8607,
       "step": 25602
     },
     {
-      "epoch": 0.7031666254703249,
+      "epoch": 0.7265323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 4.276183472385744e-06,
-      "loss": 0.7692,
+      "learning_rate": 3.6726442351010504e-06,
+      "loss": 0.8657,
       "step": 25603
     },
     {
-      "epoch": 0.7031940896981681,
+      "epoch": 0.726560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 4.275454103158952e-06,
-      "loss": 0.817,
+      "learning_rate": 3.6719325572448626e-06,
+      "loss": 0.8167,
       "step": 25604
     },
     {
-      "epoch": 0.7032215539260114,
+      "epoch": 0.726589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 4.2747247792269795e-06,
-      "loss": 0.7661,
+      "learning_rate": 3.6712209328421424e-06,
+      "loss": 0.8173,
       "step": 25605
     },
     {
-      "epoch": 0.7032490181538547,
+      "epoch": 0.7266174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 4.273995500595603e-06,
-      "loss": 0.8641,
+      "learning_rate": 3.670509361898903e-06,
+      "loss": 0.7629,
       "step": 25606
     },
     {
-      "epoch": 0.7032764823816978,
+      "epoch": 0.7266458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 4.273266267270591e-06,
-      "loss": 0.8301,
+      "learning_rate": 3.669797844421159e-06,
+      "loss": 0.7861,
       "step": 25607
     },
     {
-      "epoch": 0.7033039466095411,
+      "epoch": 0.7266742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 4.2725370792577205e-06,
-      "loss": 0.8437,
+      "learning_rate": 3.6690863804149135e-06,
+      "loss": 0.8489,
       "step": 25608
     },
     {
-      "epoch": 0.7033314108373843,
+      "epoch": 0.7267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 4.27180793656275e-06,
-      "loss": 0.7954,
+      "learning_rate": 3.6683749698861803e-06,
+      "loss": 0.807,
       "step": 25609
     },
     {
-      "epoch": 0.7033588750652275,
+      "epoch": 0.7267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 4.2710788391914545e-06,
-      "loss": 0.807,
+      "learning_rate": 3.6676636128409713e-06,
+      "loss": 0.7926,
       "step": 25610
     },
     {
-      "epoch": 0.7033863392930708,
+      "epoch": 0.7267593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 4.270349787149603e-06,
-      "loss": 0.8492,
+      "learning_rate": 3.66695230928529e-06,
+      "loss": 0.8072,
       "step": 25611
     },
     {
-      "epoch": 0.703413803520914,
+      "epoch": 0.7267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 4.269620780442965e-06,
-      "loss": 0.8607,
+      "learning_rate": 3.666241059225146e-06,
+      "loss": 0.8132,
       "step": 25612
     },
     {
-      "epoch": 0.7034412677487573,
+      "epoch": 0.7268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 4.268891819077305e-06,
-      "loss": 0.8225,
+      "learning_rate": 3.6655298626665493e-06,
+      "loss": 0.8064,
       "step": 25613
     },
     {
-      "epoch": 0.7034687319766004,
+      "epoch": 0.726844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 4.268162903058397e-06,
-      "loss": 0.8998,
+      "learning_rate": 3.664818719615506e-06,
+      "loss": 0.7856,
       "step": 25614
     },
     {
-      "epoch": 0.7034961962044437,
+      "epoch": 0.7268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 4.267434032392001e-06,
-      "loss": 0.8446,
+      "learning_rate": 3.6641076300780275e-06,
+      "loss": 0.8226,
       "step": 25615
     },
     {
-      "epoch": 0.703523660432287,
+      "epoch": 0.7269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 4.266705207083885e-06,
-      "loss": 0.8175,
+      "learning_rate": 3.6633965940601136e-06,
+      "loss": 0.8414,
       "step": 25616
     },
     {
-      "epoch": 0.7035511246601301,
+      "epoch": 0.7269296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 4.265976427139821e-06,
-      "loss": 0.8885,
+      "learning_rate": 3.6626856115677722e-06,
+      "loss": 0.8424,
       "step": 25617
     },
     {
-      "epoch": 0.7035785888879734,
+      "epoch": 0.7269580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 4.2652476925655716e-06,
-      "loss": 0.8627,
+      "learning_rate": 3.6619746826070157e-06,
+      "loss": 0.8194,
       "step": 25618
     },
     {
-      "epoch": 0.7036060531158167,
+      "epoch": 0.7269863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 4.264519003366907e-06,
-      "loss": 0.8266,
+      "learning_rate": 3.6612638071838393e-06,
+      "loss": 0.7529,
       "step": 25619
     },
     {
-      "epoch": 0.7036335173436599,
+      "epoch": 0.7270147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 4.26379035954959e-06,
-      "loss": 0.7924,
+      "learning_rate": 3.660552985304253e-06,
+      "loss": 0.8553,
       "step": 25620
     },
     {
-      "epoch": 0.7036609815715031,
+      "epoch": 0.7270431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 4.263061761119383e-06,
-      "loss": 0.8246,
+      "learning_rate": 3.659842216974263e-06,
+      "loss": 0.7726,
       "step": 25621
     },
     {
-      "epoch": 0.7036884457993463,
+      "epoch": 0.7270715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 4.262333208082052e-06,
-      "loss": 0.8469,
+      "learning_rate": 3.6591315021998684e-06,
+      "loss": 0.8721,
       "step": 25622
     },
     {
-      "epoch": 0.7037159100271896,
+      "epoch": 0.727099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 4.261604700443365e-06,
-      "loss": 0.8141,
+      "learning_rate": 3.658420840987075e-06,
+      "loss": 0.7597,
       "step": 25623
     },
     {
-      "epoch": 0.7037433742550329,
+      "epoch": 0.7271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 4.260876238209084e-06,
-      "loss": 0.8743,
+      "learning_rate": 3.657710233341889e-06,
+      "loss": 0.7972,
       "step": 25624
     },
     {
-      "epoch": 0.703770838482876,
+      "epoch": 0.7271566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 4.2601478213849754e-06,
-      "loss": 0.8637,
+      "learning_rate": 3.6569996792703065e-06,
+      "loss": 0.8197,
       "step": 25625
     },
     {
-      "epoch": 0.7037983027107193,
+      "epoch": 0.7271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 4.259419449976799e-06,
-      "loss": 0.7808,
+      "learning_rate": 3.6562891787783326e-06,
+      "loss": 0.8053,
       "step": 25626
     },
     {
-      "epoch": 0.7038257669385625,
+      "epoch": 0.7272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 4.258691123990318e-06,
-      "loss": 0.7529,
+      "learning_rate": 3.655578731871973e-06,
+      "loss": 0.8132,
       "step": 25627
     },
     {
-      "epoch": 0.7038532311664057,
+      "epoch": 0.7272417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 4.257962843431301e-06,
-      "loss": 0.7704,
+      "learning_rate": 3.6548683385572215e-06,
+      "loss": 0.8173,
       "step": 25628
     },
     {
-      "epoch": 0.703880695394249,
+      "epoch": 0.7272701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 4.2572346083055034e-06,
-      "loss": 0.8575,
+      "learning_rate": 3.654157998840082e-06,
+      "loss": 0.7716,
       "step": 25629
     },
     {
-      "epoch": 0.7039081596220922,
+      "epoch": 0.7272985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 4.25650641861869e-06,
-      "loss": 0.8153,
+      "learning_rate": 3.6534477127265555e-06,
+      "loss": 0.7709,
       "step": 25630
     },
     {
-      "epoch": 0.7039356238499355,
+      "epoch": 0.7273269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 4.255778274376625e-06,
-      "loss": 0.7922,
+      "learning_rate": 3.652737480222641e-06,
+      "loss": 0.9015,
       "step": 25631
     },
     {
-      "epoch": 0.7039630880777787,
+      "epoch": 0.7273552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 4.255050175585065e-06,
-      "loss": 0.8582,
+      "learning_rate": 3.6520273013343423e-06,
+      "loss": 0.8358,
       "step": 25632
     },
     {
-      "epoch": 0.7039905523056219,
+      "epoch": 0.727383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 4.2543221222497736e-06,
-      "loss": 0.8924,
+      "learning_rate": 3.6513171760676514e-06,
+      "loss": 0.8304,
       "step": 25633
     },
     {
-      "epoch": 0.7040180165334652,
+      "epoch": 0.7274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 4.25359411437651e-06,
-      "loss": 0.8832,
+      "learning_rate": 3.6506071044285684e-06,
+      "loss": 0.8199,
       "step": 25634
     },
     {
-      "epoch": 0.7040454807613084,
+      "epoch": 0.7274404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 4.252866151971037e-06,
-      "loss": 0.8135,
+      "learning_rate": 3.6498970864230966e-06,
+      "loss": 0.7303,
       "step": 25635
     },
     {
-      "epoch": 0.7040729449891516,
+      "epoch": 0.7274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 4.252138235039116e-06,
-      "loss": 0.8233,
+      "learning_rate": 3.649187122057226e-06,
+      "loss": 0.7787,
       "step": 25636
     },
     {
-      "epoch": 0.7041004092169949,
+      "epoch": 0.7274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 4.2514103635865035e-06,
-      "loss": 0.857,
+      "learning_rate": 3.648477211336958e-06,
+      "loss": 0.8922,
       "step": 25637
     },
     {
-      "epoch": 0.7041278734448381,
+      "epoch": 0.7275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 4.250682537618956e-06,
-      "loss": 0.8652,
+      "learning_rate": 3.6477673542682913e-06,
+      "loss": 0.9139,
       "step": 25638
     },
     {
-      "epoch": 0.7041553376726813,
+      "epoch": 0.7275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 4.249954757142235e-06,
-      "loss": 0.7808,
+      "learning_rate": 3.6470575508572157e-06,
+      "loss": 0.7715,
       "step": 25639
     },
     {
-      "epoch": 0.7041828019005245,
+      "epoch": 0.7275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 4.249227022162099e-06,
-      "loss": 0.7159,
+      "learning_rate": 3.646347801109731e-06,
+      "loss": 0.6953,
       "step": 25640
     },
     {
-      "epoch": 0.7042102661283678,
+      "epoch": 0.72761066969353,
       "grad_norm": 0.0,
-      "learning_rate": 4.2484993326843075e-06,
-      "loss": 0.7676,
+      "learning_rate": 3.6456381050318357e-06,
+      "loss": 0.779,
       "step": 25641
     },
     {
-      "epoch": 0.7042377303562111,
+      "epoch": 0.727639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 4.247771688714617e-06,
-      "loss": 0.8527,
+      "learning_rate": 3.6449284626295167e-06,
+      "loss": 0.7208,
       "step": 25642
     },
     {
-      "epoch": 0.7042651945840542,
+      "epoch": 0.7276674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 4.247044090258789e-06,
-      "loss": 0.7367,
+      "learning_rate": 3.6442188739087735e-06,
+      "loss": 0.8565,
       "step": 25643
     },
     {
-      "epoch": 0.7042926588118975,
+      "epoch": 0.7276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 4.246316537322574e-06,
-      "loss": 0.9397,
+      "learning_rate": 3.6435093388755993e-06,
+      "loss": 0.8129,
       "step": 25644
     },
     {
-      "epoch": 0.7043201230397408,
+      "epoch": 0.7277241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 4.2455890299117306e-06,
-      "loss": 0.8138,
+      "learning_rate": 3.6427998575359867e-06,
+      "loss": 0.8886,
       "step": 25645
     },
     {
-      "epoch": 0.704347587267584,
+      "epoch": 0.7277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 4.244861568032015e-06,
-      "loss": 0.8536,
+      "learning_rate": 3.6420904298959336e-06,
+      "loss": 0.8557,
       "step": 25646
     },
     {
-      "epoch": 0.7043750514954272,
+      "epoch": 0.7277809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 4.244134151689183e-06,
-      "loss": 0.8783,
+      "learning_rate": 3.641381055961424e-06,
+      "loss": 0.8494,
       "step": 25647
     },
     {
-      "epoch": 0.7044025157232704,
+      "epoch": 0.7278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 4.243406780888992e-06,
-      "loss": 0.8922,
+      "learning_rate": 3.640671735738456e-06,
+      "loss": 0.8246,
       "step": 25648
     },
     {
-      "epoch": 0.7044299799511137,
+      "epoch": 0.7278376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 4.242679455637201e-06,
-      "loss": 0.8944,
+      "learning_rate": 3.6399624692330217e-06,
+      "loss": 0.8396,
       "step": 25649
     },
     {
-      "epoch": 0.704457444178957,
+      "epoch": 0.7278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 4.241952175939555e-06,
-      "loss": 0.8549,
+      "learning_rate": 3.639253256451107e-06,
+      "loss": 0.7625,
       "step": 25650
     },
     {
-      "epoch": 0.7044849084068001,
+      "epoch": 0.727894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 4.241224941801816e-06,
-      "loss": 0.7592,
+      "learning_rate": 3.638544097398706e-06,
+      "loss": 0.7995,
       "step": 25651
     },
     {
-      "epoch": 0.7045123726346434,
+      "epoch": 0.7279228149829738,
       "grad_norm": 0.0,
-      "learning_rate": 4.2404977532297375e-06,
-      "loss": 0.8352,
+      "learning_rate": 3.6378349920818137e-06,
+      "loss": 0.7413,
       "step": 25652
     },
     {
-      "epoch": 0.7045398368624866,
+      "epoch": 0.7279511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 4.239770610229069e-06,
-      "loss": 0.8802,
+      "learning_rate": 3.637125940506411e-06,
+      "loss": 0.7852,
       "step": 25653
     },
     {
-      "epoch": 0.7045673010903298,
+      "epoch": 0.7279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 4.239043512805567e-06,
-      "loss": 0.7681,
+      "learning_rate": 3.636416942678491e-06,
+      "loss": 0.8363,
       "step": 25654
     },
     {
-      "epoch": 0.7045947653181731,
+      "epoch": 0.7280079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 4.238316460964987e-06,
-      "loss": 0.8241,
+      "learning_rate": 3.6357079986040465e-06,
+      "loss": 0.8716,
       "step": 25655
     },
     {
-      "epoch": 0.7046222295460163,
+      "epoch": 0.7280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 4.237589454713077e-06,
-      "loss": 0.7656,
+      "learning_rate": 3.6349991082890602e-06,
+      "loss": 0.9382,
       "step": 25656
     },
     {
-      "epoch": 0.7046496937738596,
+      "epoch": 0.7280646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 4.2368624940555905e-06,
-      "loss": 0.8125,
+      "learning_rate": 3.6342902717395225e-06,
+      "loss": 0.7499,
       "step": 25657
     },
     {
-      "epoch": 0.7046771580017028,
+      "epoch": 0.7280930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 4.236135578998281e-06,
-      "loss": 0.8073,
+      "learning_rate": 3.633581488961424e-06,
+      "loss": 0.8375,
       "step": 25658
     },
     {
-      "epoch": 0.704704622229546,
+      "epoch": 0.7281214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 4.235408709546901e-06,
-      "loss": 0.8765,
+      "learning_rate": 3.6328727599607416e-06,
+      "loss": 0.8441,
       "step": 25659
     },
     {
-      "epoch": 0.7047320864573893,
+      "epoch": 0.728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 4.234681885707198e-06,
-      "loss": 0.8673,
+      "learning_rate": 3.632164084743477e-06,
+      "loss": 0.7648,
       "step": 25660
     },
     {
-      "epoch": 0.7047595506852324,
+      "epoch": 0.7281782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 4.233955107484929e-06,
-      "loss": 0.9223,
+      "learning_rate": 3.6314554633156043e-06,
+      "loss": 0.8043,
       "step": 25661
     },
     {
-      "epoch": 0.7047870149130757,
+      "epoch": 0.7282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 4.233228374885837e-06,
-      "loss": 0.8283,
+      "learning_rate": 3.6307468956831136e-06,
+      "loss": 0.8307,
       "step": 25662
     },
     {
-      "epoch": 0.704814479140919,
+      "epoch": 0.7282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 4.232501687915677e-06,
-      "loss": 0.7194,
+      "learning_rate": 3.6300383818519946e-06,
+      "loss": 0.8151,
       "step": 25663
     },
     {
-      "epoch": 0.7048419433687622,
+      "epoch": 0.7282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 4.231775046580197e-06,
-      "loss": 0.8663,
+      "learning_rate": 3.629329921828224e-06,
+      "loss": 0.784,
       "step": 25664
     },
     {
-      "epoch": 0.7048694075966054,
+      "epoch": 0.7282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 4.231048450885146e-06,
-      "loss": 0.8982,
+      "learning_rate": 3.6286215156177895e-06,
+      "loss": 0.903,
       "step": 25665
     },
     {
-      "epoch": 0.7048968718244486,
+      "epoch": 0.7283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 4.230321900836275e-06,
-      "loss": 0.8619,
+      "learning_rate": 3.6279131632266783e-06,
+      "loss": 0.8097,
       "step": 25666
     },
     {
-      "epoch": 0.7049243360522919,
+      "epoch": 0.7283484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 4.2295953964393365e-06,
-      "loss": 0.7749,
+      "learning_rate": 3.627204864660869e-06,
+      "loss": 0.9209,
       "step": 25667
     },
     {
-      "epoch": 0.7049518002801352,
+      "epoch": 0.7283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 4.228868937700073e-06,
-      "loss": 0.9154,
+      "learning_rate": 3.626496619926346e-06,
+      "loss": 0.7882,
       "step": 25668
     },
     {
-      "epoch": 0.7049792645079783,
+      "epoch": 0.728405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 4.228142524624233e-06,
-      "loss": 0.8912,
+      "learning_rate": 3.6257884290290955e-06,
+      "loss": 0.9157,
       "step": 25669
     },
     {
-      "epoch": 0.7050067287358216,
+      "epoch": 0.728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 4.227416157217563e-06,
-      "loss": 0.8703,
+      "learning_rate": 3.625080291975095e-06,
+      "loss": 0.803,
       "step": 25670
     },
     {
-      "epoch": 0.7050341929636649,
+      "epoch": 0.7284619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 4.226689835485812e-06,
-      "loss": 0.8358,
+      "learning_rate": 3.624372208770326e-06,
+      "loss": 0.7831,
       "step": 25671
     },
     {
-      "epoch": 0.705061657191508,
+      "epoch": 0.7284903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 4.225963559434729e-06,
-      "loss": 0.7521,
+      "learning_rate": 3.6236641794207762e-06,
+      "loss": 0.7827,
       "step": 25672
     },
     {
-      "epoch": 0.7050891214193513,
+      "epoch": 0.7285187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 4.225237329070062e-06,
-      "loss": 0.7979,
+      "learning_rate": 3.622956203932417e-06,
+      "loss": 0.8236,
       "step": 25673
     },
     {
-      "epoch": 0.7051165856471945,
+      "epoch": 0.7285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 4.22451114439755e-06,
-      "loss": 0.8362,
+      "learning_rate": 3.622248282311235e-06,
+      "loss": 0.8313,
       "step": 25674
     },
     {
-      "epoch": 0.7051440498750378,
+      "epoch": 0.7285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 4.223785005422944e-06,
-      "loss": 0.8609,
+      "learning_rate": 3.6215404145632073e-06,
+      "loss": 0.7497,
       "step": 25675
     },
     {
-      "epoch": 0.705171514102881,
+      "epoch": 0.7286038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 4.223058912151992e-06,
-      "loss": 0.8866,
+      "learning_rate": 3.620832600694314e-06,
+      "loss": 0.7448,
       "step": 25676
     },
     {
-      "epoch": 0.7051989783307242,
+      "epoch": 0.7286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 4.222332864590432e-06,
-      "loss": 0.818,
+      "learning_rate": 3.620124840710536e-06,
+      "loss": 0.7248,
       "step": 25677
     },
     {
-      "epoch": 0.7052264425585675,
+      "epoch": 0.7286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 4.221606862744011e-06,
-      "loss": 0.8857,
+      "learning_rate": 3.619417134617853e-06,
+      "loss": 0.813,
       "step": 25678
     },
     {
-      "epoch": 0.7052539067864106,
+      "epoch": 0.7286889897843359,
       "grad_norm": 0.0,
-      "learning_rate": 4.220880906618479e-06,
-      "loss": 0.774,
+      "learning_rate": 3.6187094824222367e-06,
+      "loss": 0.7135,
       "step": 25679
     },
     {
-      "epoch": 0.7052813710142539,
+      "epoch": 0.7287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 4.2201549962195735e-06,
-      "loss": 0.8064,
+      "learning_rate": 3.6180018841296684e-06,
+      "loss": 0.8183,
       "step": 25680
     },
     {
-      "epoch": 0.7053088352420972,
+      "epoch": 0.7287457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 4.21942913155304e-06,
-      "loss": 0.8788,
+      "learning_rate": 3.617294339746128e-06,
+      "loss": 0.6773,
       "step": 25681
     },
     {
-      "epoch": 0.7053362994699404,
+      "epoch": 0.7287741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 4.2187033126246225e-06,
-      "loss": 0.8469,
+      "learning_rate": 3.616586849277587e-06,
+      "loss": 0.8448,
       "step": 25682
     },
     {
-      "epoch": 0.7053637636977836,
+      "epoch": 0.7288024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 4.217977539440067e-06,
-      "loss": 0.8385,
+      "learning_rate": 3.6158794127300224e-06,
+      "loss": 0.7271,
       "step": 25683
     },
     {
-      "epoch": 0.7053912279256269,
+      "epoch": 0.7288308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 4.21725181200511e-06,
-      "loss": 0.7958,
+      "learning_rate": 3.615172030109415e-06,
+      "loss": 0.8232,
       "step": 25684
     },
     {
-      "epoch": 0.7054186921534701,
+      "epoch": 0.7288592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 4.216526130325499e-06,
-      "loss": 0.9257,
+      "learning_rate": 3.614464701421734e-06,
+      "loss": 0.84,
       "step": 25685
     },
     {
-      "epoch": 0.7054461563813134,
+      "epoch": 0.7288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 4.215800494406971e-06,
-      "loss": 0.9238,
+      "learning_rate": 3.6137574266729604e-06,
+      "loss": 0.7738,
       "step": 25686
     },
     {
-      "epoch": 0.7054736206091565,
+      "epoch": 0.7289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 4.21507490425527e-06,
-      "loss": 0.8146,
+      "learning_rate": 3.61305020586906e-06,
+      "loss": 0.7633,
       "step": 25687
     },
     {
-      "epoch": 0.7055010848369998,
+      "epoch": 0.72894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 4.2143493598761375e-06,
-      "loss": 0.7544,
+      "learning_rate": 3.612343039016013e-06,
+      "loss": 0.7815,
       "step": 25688
     },
     {
-      "epoch": 0.7055285490648431,
+      "epoch": 0.7289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 4.2136238612753135e-06,
-      "loss": 0.8089,
+      "learning_rate": 3.611635926119794e-06,
+      "loss": 0.7691,
       "step": 25689
     },
     {
-      "epoch": 0.7055560132926862,
+      "epoch": 0.7290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 4.212898408458542e-06,
-      "loss": 0.816,
+      "learning_rate": 3.6109288671863674e-06,
+      "loss": 0.9419,
       "step": 25690
     },
     {
-      "epoch": 0.7055834775205295,
+      "epoch": 0.7290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 4.212173001431557e-06,
-      "loss": 0.704,
+      "learning_rate": 3.6102218622217143e-06,
+      "loss": 0.8209,
       "step": 25691
     },
     {
-      "epoch": 0.7056109417483728,
+      "epoch": 0.7290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 4.211447640200106e-06,
-      "loss": 0.779,
+      "learning_rate": 3.6095149112318083e-06,
+      "loss": 0.7825,
       "step": 25692
     },
     {
-      "epoch": 0.705638405976216,
+      "epoch": 0.7290862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 4.210722324769918e-06,
-      "loss": 0.8346,
+      "learning_rate": 3.6088080142226144e-06,
+      "loss": 0.778,
       "step": 25693
     },
     {
-      "epoch": 0.7056658702040592,
+      "epoch": 0.7291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 4.209997055146739e-06,
-      "loss": 0.8014,
+      "learning_rate": 3.6081011712001055e-06,
+      "loss": 0.7986,
       "step": 25694
     },
     {
-      "epoch": 0.7056933344319024,
+      "epoch": 0.7291430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 4.209271831336306e-06,
-      "loss": 0.7645,
+      "learning_rate": 3.607394382170257e-06,
+      "loss": 0.8123,
       "step": 25695
     },
     {
-      "epoch": 0.7057207986597457,
+      "epoch": 0.7291713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 4.20854665334436e-06,
-      "loss": 0.8165,
+      "learning_rate": 3.6066876471390323e-06,
+      "loss": 0.8785,
       "step": 25696
     },
     {
-      "epoch": 0.705748262887589,
+      "epoch": 0.7291997729852441,
       "grad_norm": 0.0,
-      "learning_rate": 4.207821521176633e-06,
-      "loss": 0.8683,
+      "learning_rate": 3.6059809661124047e-06,
+      "loss": 0.7823,
       "step": 25697
     },
     {
-      "epoch": 0.7057757271154321,
+      "epoch": 0.7292281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 4.207096434838865e-06,
-      "loss": 0.9014,
+      "learning_rate": 3.6052743390963464e-06,
+      "loss": 0.8266,
       "step": 25698
     },
     {
-      "epoch": 0.7058031913432754,
+      "epoch": 0.7292565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 4.206371394336796e-06,
-      "loss": 0.8386,
+      "learning_rate": 3.604567766096819e-06,
+      "loss": 0.7143,
       "step": 25699
     },
     {
-      "epoch": 0.7058306555711186,
+      "epoch": 0.7292849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 4.205646399676163e-06,
-      "loss": 0.8477,
+      "learning_rate": 3.6038612471197963e-06,
+      "loss": 0.8492,
       "step": 25700
     },
     {
-      "epoch": 0.7058581197989618,
+      "epoch": 0.7293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 4.204921450862697e-06,
-      "loss": 0.8706,
+      "learning_rate": 3.6031547821712486e-06,
+      "loss": 0.8034,
       "step": 25701
     },
     {
-      "epoch": 0.7058855840268051,
+      "epoch": 0.7293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 4.204196547902141e-06,
-      "loss": 0.8334,
+      "learning_rate": 3.6024483712571366e-06,
+      "loss": 0.7815,
       "step": 25702
     },
     {
-      "epoch": 0.7059130482546483,
+      "epoch": 0.7293700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 4.203471690800223e-06,
-      "loss": 0.7406,
+      "learning_rate": 3.60174201438343e-06,
+      "loss": 0.891,
       "step": 25703
     },
     {
-      "epoch": 0.7059405124824916,
+      "epoch": 0.7293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 4.202746879562682e-06,
-      "loss": 0.7924,
+      "learning_rate": 3.6010357115561e-06,
+      "loss": 0.7284,
       "step": 25704
     },
     {
-      "epoch": 0.7059679767103348,
+      "epoch": 0.7294267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 4.202022114195255e-06,
-      "loss": 0.8368,
+      "learning_rate": 3.6003294627811025e-06,
+      "loss": 0.8345,
       "step": 25705
     },
     {
-      "epoch": 0.705995440938178,
+      "epoch": 0.729455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 4.201297394703673e-06,
-      "loss": 0.8672,
+      "learning_rate": 3.599623268064416e-06,
+      "loss": 0.8047,
       "step": 25706
     },
     {
-      "epoch": 0.7060229051660213,
+      "epoch": 0.729483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 4.2005727210936764e-06,
-      "loss": 0.8275,
+      "learning_rate": 3.598917127411997e-06,
+      "loss": 0.8029,
       "step": 25707
     },
     {
-      "epoch": 0.7060503693938645,
+      "epoch": 0.7295119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 4.199848093370992e-06,
-      "loss": 0.9337,
+      "learning_rate": 3.5982110408298123e-06,
+      "loss": 0.8054,
       "step": 25708
     },
     {
-      "epoch": 0.7060778336217077,
+      "epoch": 0.7295402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 4.199123511541359e-06,
-      "loss": 0.8511,
+      "learning_rate": 3.5975050083238294e-06,
+      "loss": 0.7879,
       "step": 25709
     },
     {
-      "epoch": 0.706105297849551,
+      "epoch": 0.7295686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 4.198398975610505e-06,
-      "loss": 0.716,
+      "learning_rate": 3.5967990299000066e-06,
+      "loss": 0.7944,
       "step": 25710
     },
     {
-      "epoch": 0.7061327620773942,
+      "epoch": 0.7295970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 4.197674485584164e-06,
-      "loss": 0.8862,
+      "learning_rate": 3.596093105564311e-06,
+      "loss": 0.9217,
       "step": 25711
     },
     {
-      "epoch": 0.7061602263052374,
+      "epoch": 0.7296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 4.196950041468072e-06,
-      "loss": 0.8354,
+      "learning_rate": 3.595387235322707e-06,
+      "loss": 0.8395,
       "step": 25712
     },
     {
-      "epoch": 0.7061876905330806,
+      "epoch": 0.7296538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 4.1962256432679586e-06,
-      "loss": 0.7933,
+      "learning_rate": 3.594681419181153e-06,
+      "loss": 0.8568,
       "step": 25713
     },
     {
-      "epoch": 0.7062151547609239,
+      "epoch": 0.7296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 4.195501290989559e-06,
-      "loss": 0.8385,
+      "learning_rate": 3.593975657145612e-06,
+      "loss": 0.8133,
       "step": 25714
     },
     {
-      "epoch": 0.7062426189887672,
+      "epoch": 0.7297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 4.194776984638599e-06,
-      "loss": 0.7342,
+      "learning_rate": 3.5932699492220502e-06,
+      "loss": 0.7228,
       "step": 25715
     },
     {
-      "epoch": 0.7062700832166103,
+      "epoch": 0.729738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 4.194052724220815e-06,
-      "loss": 0.7423,
+      "learning_rate": 3.5925642954164208e-06,
+      "loss": 0.9444,
       "step": 25716
     },
     {
-      "epoch": 0.7062975474444536,
+      "epoch": 0.7297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 4.1933285097419315e-06,
-      "loss": 0.9562,
+      "learning_rate": 3.59185869573469e-06,
+      "loss": 0.7585,
       "step": 25717
     },
     {
-      "epoch": 0.7063250116722969,
+      "epoch": 0.7297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 4.192604341207682e-06,
-      "loss": 0.8612,
+      "learning_rate": 3.59115315018282e-06,
+      "loss": 0.8844,
       "step": 25718
     },
     {
-      "epoch": 0.70635247590014,
+      "epoch": 0.7298240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 4.191880218623797e-06,
-      "loss": 0.8174,
+      "learning_rate": 3.5904476587667637e-06,
+      "loss": 0.77,
       "step": 25719
     },
     {
-      "epoch": 0.7063799401279833,
+      "epoch": 0.7298524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 4.191156141996009e-06,
-      "loss": 0.8128,
+      "learning_rate": 3.5897422214924847e-06,
+      "loss": 0.8316,
       "step": 25720
     },
     {
-      "epoch": 0.7064074043558265,
+      "epoch": 0.7298808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 4.19043211133004e-06,
-      "loss": 0.7639,
+      "learning_rate": 3.5890368383659414e-06,
+      "loss": 0.7681,
       "step": 25721
     },
     {
-      "epoch": 0.7064348685836698,
+      "epoch": 0.7299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 4.189708126631623e-06,
-      "loss": 0.823,
+      "learning_rate": 3.5883315093930913e-06,
+      "loss": 0.8878,
       "step": 25722
     },
     {
-      "epoch": 0.706462332811513,
+      "epoch": 0.7299375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 4.188984187906485e-06,
-      "loss": 0.8428,
+      "learning_rate": 3.5876262345798972e-06,
+      "loss": 0.8162,
       "step": 25723
     },
     {
-      "epoch": 0.7064897970393562,
+      "epoch": 0.7299659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 4.18826029516036e-06,
-      "loss": 0.823,
+      "learning_rate": 3.5869210139323086e-06,
+      "loss": 0.9321,
       "step": 25724
     },
     {
-      "epoch": 0.7065172612671995,
+      "epoch": 0.729994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.187536448398966e-06,
-      "loss": 0.8589,
+      "learning_rate": 3.586215847456287e-06,
+      "loss": 0.8474,
       "step": 25725
     },
     {
-      "epoch": 0.7065447254950427,
+      "epoch": 0.7300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 4.1868126476280395e-06,
-      "loss": 0.8234,
+      "learning_rate": 3.585510735157791e-06,
+      "loss": 0.8107,
       "step": 25726
     },
     {
-      "epoch": 0.7065721897228859,
+      "epoch": 0.7300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 4.1860888928533e-06,
-      "loss": 0.8184,
+      "learning_rate": 3.584805677042772e-06,
+      "loss": 0.8942,
       "step": 25727
     },
     {
-      "epoch": 0.7065996539507292,
+      "epoch": 0.7300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 4.1853651840804754e-06,
-      "loss": 0.8552,
+      "learning_rate": 3.584100673117187e-06,
+      "loss": 0.9887,
       "step": 25728
     },
     {
-      "epoch": 0.7066271181785724,
+      "epoch": 0.7301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 4.184641521315296e-06,
-      "loss": 0.8361,
+      "learning_rate": 3.5833957233869963e-06,
+      "loss": 0.8116,
       "step": 25729
     },
     {
-      "epoch": 0.7066545824064157,
+      "epoch": 0.7301362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 4.183917904563484e-06,
-      "loss": 0.8418,
+      "learning_rate": 3.5826908278581463e-06,
+      "loss": 0.8476,
       "step": 25730
     },
     {
-      "epoch": 0.7066820466342589,
+      "epoch": 0.7301645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 4.1831943338307705e-06,
-      "loss": 0.8107,
+      "learning_rate": 3.5819859865365957e-06,
+      "loss": 0.7614,
       "step": 25731
     },
     {
-      "epoch": 0.7067095108621021,
+      "epoch": 0.7301929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 4.182470809122873e-06,
-      "loss": 0.8072,
+      "learning_rate": 3.5812811994283005e-06,
+      "loss": 0.7341,
       "step": 25732
     },
     {
-      "epoch": 0.7067369750899454,
+      "epoch": 0.7302213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 4.181747330445524e-06,
-      "loss": 0.8439,
+      "learning_rate": 3.5805764665392094e-06,
+      "loss": 0.8613,
       "step": 25733
     },
     {
-      "epoch": 0.7067644393177885,
+      "epoch": 0.7302497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 4.181023897804439e-06,
-      "loss": 0.9329,
+      "learning_rate": 3.5798717878752765e-06,
+      "loss": 0.7393,
       "step": 25734
     },
     {
-      "epoch": 0.7067919035456318,
+      "epoch": 0.73027809307605,
       "grad_norm": 0.0,
-      "learning_rate": 4.180300511205348e-06,
-      "loss": 0.8182,
+      "learning_rate": 3.5791671634424586e-06,
+      "loss": 0.6941,
       "step": 25735
     },
     {
-      "epoch": 0.7068193677734751,
+      "epoch": 0.7303064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 4.179577170653974e-06,
-      "loss": 0.8379,
+      "learning_rate": 3.5784625932466978e-06,
+      "loss": 0.8725,
       "step": 25736
     },
     {
-      "epoch": 0.7068468320013183,
+      "epoch": 0.7303348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 4.17885387615604e-06,
-      "loss": 0.8297,
+      "learning_rate": 3.577758077293959e-06,
+      "loss": 0.8534,
       "step": 25737
     },
     {
-      "epoch": 0.7068742962291615,
+      "epoch": 0.7303632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 4.178130627717273e-06,
-      "loss": 0.7357,
+      "learning_rate": 3.5770536155901838e-06,
+      "loss": 0.7543,
       "step": 25738
     },
     {
-      "epoch": 0.7069017604570047,
+      "epoch": 0.7303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 4.177407425343387e-06,
-      "loss": 0.8086,
+      "learning_rate": 3.5763492081413242e-06,
+      "loss": 0.7945,
       "step": 25739
     },
     {
-      "epoch": 0.706929224684848,
+      "epoch": 0.7304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 4.1766842690401135e-06,
-      "loss": 0.8837,
+      "learning_rate": 3.575644854953335e-06,
+      "loss": 0.8057,
       "step": 25740
     },
     {
-      "epoch": 0.7069566889126913,
+      "epoch": 0.7304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 4.175961158813165e-06,
-      "loss": 0.8643,
+      "learning_rate": 3.5749405560321606e-06,
+      "loss": 0.8781,
       "step": 25741
     },
     {
-      "epoch": 0.7069841531405344,
+      "epoch": 0.7304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 4.175238094668268e-06,
-      "loss": 0.8372,
+      "learning_rate": 3.574236311383751e-06,
+      "loss": 0.7694,
       "step": 25742
     },
     {
-      "epoch": 0.7070116173683777,
+      "epoch": 0.7305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 4.174515076611144e-06,
-      "loss": 0.8463,
+      "learning_rate": 3.5735321210140596e-06,
+      "loss": 0.8175,
       "step": 25743
     },
     {
-      "epoch": 0.707039081596221,
+      "epoch": 0.7305334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 4.173792104647517e-06,
-      "loss": 0.8118,
+      "learning_rate": 3.572827984929028e-06,
+      "loss": 0.8049,
       "step": 25744
     },
     {
-      "epoch": 0.7070665458240641,
+      "epoch": 0.7305618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 4.1730691787831e-06,
-      "loss": 0.8715,
+      "learning_rate": 3.5721239031346067e-06,
+      "loss": 0.8047,
       "step": 25745
     },
     {
-      "epoch": 0.7070940100519074,
+      "epoch": 0.7305902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 4.172346299023616e-06,
-      "loss": 0.8945,
+      "learning_rate": 3.571419875636748e-06,
+      "loss": 0.7794,
       "step": 25746
     },
     {
-      "epoch": 0.7071214742797506,
+      "epoch": 0.7306186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 4.1716234653747866e-06,
-      "loss": 0.8207,
+      "learning_rate": 3.570715902441391e-06,
+      "loss": 0.6967,
       "step": 25747
     },
     {
-      "epoch": 0.7071489385075939,
+      "epoch": 0.7306469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 4.170900677842334e-06,
-      "loss": 0.8723,
+      "learning_rate": 3.5700119835544856e-06,
+      "loss": 0.7296,
       "step": 25748
     },
     {
-      "epoch": 0.7071764027354371,
+      "epoch": 0.7306753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 4.1701779364319715e-06,
-      "loss": 0.8778,
+      "learning_rate": 3.5693081189819813e-06,
+      "loss": 0.8377,
       "step": 25749
     },
     {
-      "epoch": 0.7072038669632803,
+      "epoch": 0.7307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 4.169455241149417e-06,
-      "loss": 0.809,
+      "learning_rate": 3.568604308729817e-06,
+      "loss": 0.72,
       "step": 25750
     },
     {
-      "epoch": 0.7072313311911236,
+      "epoch": 0.7307321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 4.1687325920003905e-06,
-      "loss": 0.8378,
+      "learning_rate": 3.5679005528039423e-06,
+      "loss": 0.7724,
       "step": 25751
     },
     {
-      "epoch": 0.7072587954189667,
+      "epoch": 0.7307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 4.168009988990613e-06,
-      "loss": 0.8971,
+      "learning_rate": 3.5671968512102993e-06,
+      "loss": 0.8563,
       "step": 25752
     },
     {
-      "epoch": 0.70728625964681,
+      "epoch": 0.730788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.167287432125796e-06,
-      "loss": 0.755,
+      "learning_rate": 3.566493203954834e-06,
+      "loss": 0.8084,
       "step": 25753
     },
     {
-      "epoch": 0.7073137238746533,
+      "epoch": 0.7308172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 4.1665649214116624e-06,
-      "loss": 0.8928,
+      "learning_rate": 3.5657896110434942e-06,
+      "loss": 0.8785,
       "step": 25754
     },
     {
-      "epoch": 0.7073411881024965,
+      "epoch": 0.7308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 4.165842456853929e-06,
-      "loss": 0.836,
+      "learning_rate": 3.5650860724822144e-06,
+      "loss": 0.7968,
       "step": 25755
     },
     {
-      "epoch": 0.7073686523303397,
+      "epoch": 0.7308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 4.165120038458308e-06,
-      "loss": 0.7994,
+      "learning_rate": 3.564382588276942e-06,
+      "loss": 0.735,
       "step": 25756
     },
     {
-      "epoch": 0.707396116558183,
+      "epoch": 0.7309023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 4.164397666230519e-06,
-      "loss": 0.8774,
+      "learning_rate": 3.5636791584336227e-06,
+      "loss": 0.7628,
       "step": 25757
     },
     {
-      "epoch": 0.7074235807860262,
+      "epoch": 0.7309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 4.1636753401762756e-06,
-      "loss": 0.7609,
+      "learning_rate": 3.5629757829581912e-06,
+      "loss": 0.7875,
       "step": 25758
     },
     {
-      "epoch": 0.7074510450138695,
+      "epoch": 0.7309591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 4.1629530603012915e-06,
-      "loss": 0.7364,
+      "learning_rate": 3.562272461856593e-06,
+      "loss": 0.8559,
       "step": 25759
     },
     {
-      "epoch": 0.7074785092417126,
+      "epoch": 0.7309875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 4.1622308266112846e-06,
-      "loss": 0.8176,
+      "learning_rate": 3.561569195134772e-06,
+      "loss": 0.7475,
       "step": 25760
     },
     {
-      "epoch": 0.7075059734695559,
+      "epoch": 0.7310158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 4.161508639111972e-06,
-      "loss": 0.8437,
+      "learning_rate": 3.5608659827986624e-06,
+      "loss": 0.7799,
       "step": 25761
     },
     {
-      "epoch": 0.7075334376973992,
+      "epoch": 0.7310442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 4.160786497809061e-06,
-      "loss": 0.9157,
+      "learning_rate": 3.5601628248542065e-06,
+      "loss": 0.8652,
       "step": 25762
     },
     {
-      "epoch": 0.7075609019252423,
+      "epoch": 0.7310726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 4.160064402708268e-06,
-      "loss": 0.8585,
+      "learning_rate": 3.559459721307349e-06,
+      "loss": 0.899,
       "step": 25763
     },
     {
-      "epoch": 0.7075883661530856,
+      "epoch": 0.7311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 4.159342353815313e-06,
-      "loss": 0.9004,
+      "learning_rate": 3.558756672164021e-06,
+      "loss": 0.8884,
       "step": 25764
     },
     {
-      "epoch": 0.7076158303809288,
+      "epoch": 0.7311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 4.158620351135899e-06,
-      "loss": 0.8811,
+      "learning_rate": 3.5580536774301643e-06,
+      "loss": 0.8119,
       "step": 25765
     },
     {
-      "epoch": 0.7076432946087721,
+      "epoch": 0.7311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.157898394675745e-06,
-      "loss": 0.8167,
+      "learning_rate": 3.5573507371117223e-06,
+      "loss": 0.8198,
       "step": 25766
     },
     {
-      "epoch": 0.7076707588366153,
+      "epoch": 0.7311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 4.1571764844405635e-06,
-      "loss": 0.9408,
+      "learning_rate": 3.556647851214621e-06,
+      "loss": 0.8593,
       "step": 25767
     },
     {
-      "epoch": 0.7076982230644585,
+      "epoch": 0.7312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 4.156454620436064e-06,
-      "loss": 0.7536,
+      "learning_rate": 3.555945019744811e-06,
+      "loss": 0.8029,
       "step": 25768
     },
     {
-      "epoch": 0.7077256872923018,
+      "epoch": 0.7312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 4.155732802667957e-06,
-      "loss": 0.7534,
+      "learning_rate": 3.55524224270822e-06,
+      "loss": 0.8431,
       "step": 25769
     },
     {
-      "epoch": 0.7077531515201451,
+      "epoch": 0.7312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 4.155011031141957e-06,
-      "loss": 0.8885,
+      "learning_rate": 3.554539520110788e-06,
+      "loss": 0.8231,
       "step": 25770
     },
     {
-      "epoch": 0.7077806157479882,
+      "epoch": 0.731299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 4.154289305863773e-06,
-      "loss": 0.8676,
+      "learning_rate": 3.5538368519584543e-06,
+      "loss": 0.8671,
       "step": 25771
     },
     {
-      "epoch": 0.7078080799758315,
+      "epoch": 0.731328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 4.15356762683912e-06,
-      "loss": 0.8966,
+      "learning_rate": 3.5531342382571467e-06,
+      "loss": 0.9106,
       "step": 25772
     },
     {
-      "epoch": 0.7078355442036747,
+      "epoch": 0.7313564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 4.1528459940737055e-06,
-      "loss": 0.8314,
+      "learning_rate": 3.5524316790128032e-06,
+      "loss": 0.8569,
       "step": 25773
     },
     {
-      "epoch": 0.707863008431518,
+      "epoch": 0.7313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 4.1521244075732335e-06,
-      "loss": 0.8426,
+      "learning_rate": 3.551729174231363e-06,
+      "loss": 0.8385,
       "step": 25774
     },
     {
-      "epoch": 0.7078904726593612,
+      "epoch": 0.7314131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 4.1514028673434194e-06,
-      "loss": 0.7512,
+      "learning_rate": 3.5510267239187525e-06,
+      "loss": 0.7703,
       "step": 25775
     },
     {
-      "epoch": 0.7079179368872044,
+      "epoch": 0.7314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 4.150681373389971e-06,
-      "loss": 0.8239,
+      "learning_rate": 3.5503243280809097e-06,
+      "loss": 0.8301,
       "step": 25776
     },
     {
-      "epoch": 0.7079454011150477,
+      "epoch": 0.7314699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 4.1499599257185985e-06,
-      "loss": 0.7784,
+      "learning_rate": 3.5496219867237712e-06,
+      "loss": 0.8335,
       "step": 25777
     },
     {
-      "epoch": 0.7079728653428908,
+      "epoch": 0.7314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 4.149238524335008e-06,
-      "loss": 0.823,
+      "learning_rate": 3.5489196998532616e-06,
+      "loss": 0.8494,
       "step": 25778
     },
     {
-      "epoch": 0.7080003295707341,
+      "epoch": 0.7315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 4.148517169244912e-06,
-      "loss": 0.8383,
+      "learning_rate": 3.548217467475317e-06,
+      "loss": 0.755,
       "step": 25779
     },
     {
-      "epoch": 0.7080277937985774,
+      "epoch": 0.7315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 4.147795860454016e-06,
-      "loss": 0.8201,
+      "learning_rate": 3.547515289595873e-06,
+      "loss": 0.8358,
       "step": 25780
     },
     {
-      "epoch": 0.7080552580264206,
+      "epoch": 0.7315834279228149,
       "grad_norm": 0.0,
-      "learning_rate": 4.147074597968022e-06,
-      "loss": 0.9486,
+      "learning_rate": 3.5468131662208547e-06,
+      "loss": 0.9581,
       "step": 25781
     },
     {
-      "epoch": 0.7080827222542638,
+      "epoch": 0.7316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 4.146353381792641e-06,
-      "loss": 0.8433,
+      "learning_rate": 3.546111097356192e-06,
+      "loss": 0.822,
       "step": 25782
     },
     {
-      "epoch": 0.7081101864821071,
+      "epoch": 0.7316401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 4.145632211933579e-06,
-      "loss": 0.8668,
+      "learning_rate": 3.5454090830078257e-06,
+      "loss": 0.8702,
       "step": 25783
     },
     {
-      "epoch": 0.7081376507099503,
+      "epoch": 0.7316685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 4.144911088396543e-06,
-      "loss": 0.8655,
+      "learning_rate": 3.5447071231816753e-06,
+      "loss": 0.7927,
       "step": 25784
     },
     {
-      "epoch": 0.7081651149377935,
+      "epoch": 0.7316969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 4.144190011187241e-06,
-      "loss": 0.8046,
+      "learning_rate": 3.544005217883675e-06,
+      "loss": 0.8221,
       "step": 25785
     },
     {
-      "epoch": 0.7081925791656367,
+      "epoch": 0.7317253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 4.143468980311375e-06,
-      "loss": 0.8722,
+      "learning_rate": 3.543303367119755e-06,
+      "loss": 0.8164,
       "step": 25786
     },
     {
-      "epoch": 0.70822004339348,
+      "epoch": 0.7317536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 4.142747995774649e-06,
-      "loss": 0.8583,
+      "learning_rate": 3.5426015708958396e-06,
+      "loss": 0.8639,
       "step": 25787
     },
     {
-      "epoch": 0.7082475076213233,
+      "epoch": 0.7317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 4.142027057582774e-06,
-      "loss": 0.791,
+      "learning_rate": 3.5418998292178573e-06,
+      "loss": 0.8023,
       "step": 25788
     },
     {
-      "epoch": 0.7082749718491664,
+      "epoch": 0.7318104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 4.141306165741447e-06,
-      "loss": 0.8885,
+      "learning_rate": 3.541198142091742e-06,
+      "loss": 0.7972,
       "step": 25789
     },
     {
-      "epoch": 0.7083024360770097,
+      "epoch": 0.731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 4.140585320256374e-06,
-      "loss": 0.7941,
+      "learning_rate": 3.5404965095234136e-06,
+      "loss": 0.876,
       "step": 25790
     },
     {
-      "epoch": 0.7083299003048529,
+      "epoch": 0.7318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 4.139864521133263e-06,
-      "loss": 0.8008,
+      "learning_rate": 3.5397949315187995e-06,
+      "loss": 0.7978,
       "step": 25791
     },
     {
-      "epoch": 0.7083573645326962,
+      "epoch": 0.7318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 4.139143768377812e-06,
-      "loss": 0.8571,
+      "learning_rate": 3.5390934080838333e-06,
+      "loss": 0.885,
       "step": 25792
     },
     {
-      "epoch": 0.7083848287605394,
+      "epoch": 0.7319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 4.138423061995724e-06,
-      "loss": 0.8507,
+      "learning_rate": 3.5383919392244315e-06,
+      "loss": 0.8362,
       "step": 25793
     },
     {
-      "epoch": 0.7084122929883826,
+      "epoch": 0.7319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 4.137702401992704e-06,
-      "loss": 0.7683,
+      "learning_rate": 3.5376905249465267e-06,
+      "loss": 0.8934,
       "step": 25794
     },
     {
-      "epoch": 0.7084397572162259,
+      "epoch": 0.7319807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 4.136981788374454e-06,
-      "loss": 0.7976,
+      "learning_rate": 3.5369891652560375e-06,
+      "loss": 0.8991,
       "step": 25795
     },
     {
-      "epoch": 0.7084672214440692,
+      "epoch": 0.7320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 4.136261221146678e-06,
-      "loss": 0.8675,
+      "learning_rate": 3.5362878601588915e-06,
+      "loss": 0.8585,
       "step": 25796
     },
     {
-      "epoch": 0.7084946856719123,
+      "epoch": 0.7320374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 4.1355407003150745e-06,
-      "loss": 0.9057,
+      "learning_rate": 3.5355866096610115e-06,
+      "loss": 0.8244,
       "step": 25797
     },
     {
-      "epoch": 0.7085221498997556,
+      "epoch": 0.7320658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 4.134820225885342e-06,
-      "loss": 0.787,
+      "learning_rate": 3.534885413768323e-06,
+      "loss": 0.8345,
       "step": 25798
     },
     {
-      "epoch": 0.7085496141275988,
+      "epoch": 0.7320942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 4.134099797863182e-06,
-      "loss": 0.8457,
+      "learning_rate": 3.534184272486747e-06,
+      "loss": 0.8191,
       "step": 25799
     },
     {
-      "epoch": 0.708577078355442,
+      "epoch": 0.7321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 4.133379416254299e-06,
-      "loss": 0.889,
+      "learning_rate": 3.533483185822212e-06,
+      "loss": 0.8533,
       "step": 25800
     },
     {
-      "epoch": 0.7086045425832853,
+      "epoch": 0.7321509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 4.132659081064389e-06,
-      "loss": 0.877,
+      "learning_rate": 3.5327821537806305e-06,
+      "loss": 0.8166,
       "step": 25801
     },
     {
-      "epoch": 0.7086320068111285,
+      "epoch": 0.7321793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 4.131938792299154e-06,
-      "loss": 0.7743,
+      "learning_rate": 3.53208117636793e-06,
+      "loss": 0.9557,
       "step": 25802
     },
     {
-      "epoch": 0.7086594710389718,
+      "epoch": 0.7322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 4.131218549964295e-06,
-      "loss": 0.7932,
+      "learning_rate": 3.5313802535900334e-06,
+      "loss": 0.939,
       "step": 25803
     },
     {
-      "epoch": 0.7086869352668149,
+      "epoch": 0.7322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 4.13049835406551e-06,
-      "loss": 0.9387,
+      "learning_rate": 3.5306793854528543e-06,
+      "loss": 0.6973,
       "step": 25804
     },
     {
-      "epoch": 0.7087143994946582,
+      "epoch": 0.7322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 4.12977820460849e-06,
-      "loss": 0.7691,
+      "learning_rate": 3.529978571962318e-06,
+      "loss": 0.8763,
       "step": 25805
     },
     {
-      "epoch": 0.7087418637225015,
+      "epoch": 0.7322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 4.12905810159894e-06,
-      "loss": 0.8437,
+      "learning_rate": 3.5292778131243475e-06,
+      "loss": 0.8249,
       "step": 25806
     },
     {
-      "epoch": 0.7087693279503446,
+      "epoch": 0.7323212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 4.1283380450425555e-06,
-      "loss": 0.8169,
+      "learning_rate": 3.5285771089448538e-06,
+      "loss": 0.8893,
       "step": 25807
     },
     {
-      "epoch": 0.7087967921781879,
+      "epoch": 0.732349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 4.127618034945039e-06,
-      "loss": 0.7019,
+      "learning_rate": 3.5278764594297608e-06,
+      "loss": 0.7897,
       "step": 25808
     },
     {
-      "epoch": 0.7088242564060312,
+      "epoch": 0.732377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 4.12689807131208e-06,
-      "loss": 0.8431,
+      "learning_rate": 3.5271758645849897e-06,
+      "loss": 0.8014,
       "step": 25809
     },
     {
-      "epoch": 0.7088517206338744,
+      "epoch": 0.7324063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 4.126178154149378e-06,
-      "loss": 0.8019,
+      "learning_rate": 3.5264753244164517e-06,
+      "loss": 0.8518,
       "step": 25810
     },
     {
-      "epoch": 0.7088791848617176,
+      "epoch": 0.7324347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 4.12545828346263e-06,
-      "loss": 0.8142,
+      "learning_rate": 3.5257748389300673e-06,
+      "loss": 0.7736,
       "step": 25811
     },
     {
-      "epoch": 0.7089066490895608,
+      "epoch": 0.7324631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 4.124738459257536e-06,
-      "loss": 0.8336,
+      "learning_rate": 3.5250744081317567e-06,
+      "loss": 0.8479,
       "step": 25812
     },
     {
-      "epoch": 0.7089341133174041,
+      "epoch": 0.7324914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 4.124018681539782e-06,
-      "loss": 0.75,
+      "learning_rate": 3.5243740320274277e-06,
+      "loss": 0.7913,
       "step": 25813
     },
     {
-      "epoch": 0.7089615775452474,
+      "epoch": 0.7325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 4.123298950315073e-06,
-      "loss": 0.8773,
+      "learning_rate": 3.523673710623009e-06,
+      "loss": 0.8277,
       "step": 25814
     },
     {
-      "epoch": 0.7089890417730905,
+      "epoch": 0.7325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 4.122579265589095e-06,
-      "loss": 0.942,
+      "learning_rate": 3.5229734439244068e-06,
+      "loss": 0.8877,
       "step": 25815
     },
     {
-      "epoch": 0.7090165060009338,
+      "epoch": 0.7325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 4.121859627367547e-06,
-      "loss": 0.7299,
+      "learning_rate": 3.5222732319375385e-06,
+      "loss": 0.828,
       "step": 25816
     },
     {
-      "epoch": 0.709043970228777,
+      "epoch": 0.7326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 4.121140035656123e-06,
-      "loss": 0.8224,
+      "learning_rate": 3.521573074668323e-06,
+      "loss": 0.8669,
       "step": 25817
     },
     {
-      "epoch": 0.7090714344566202,
+      "epoch": 0.732633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 4.120420490460515e-06,
-      "loss": 0.8664,
+      "learning_rate": 3.5208729721226663e-06,
+      "loss": 0.7274,
       "step": 25818
     },
     {
-      "epoch": 0.7090988986844635,
+      "epoch": 0.7326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 4.119700991786421e-06,
-      "loss": 0.798,
+      "learning_rate": 3.5201729243064885e-06,
+      "loss": 0.8565,
       "step": 25819
     },
     {
-      "epoch": 0.7091263629123067,
+      "epoch": 0.7326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 4.118981539639529e-06,
-      "loss": 0.8509,
+      "learning_rate": 3.5194729312257035e-06,
+      "loss": 0.8158,
       "step": 25820
     },
     {
-      "epoch": 0.70915382714015,
+      "epoch": 0.7327185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 4.118262134025536e-06,
-      "loss": 0.8042,
+      "learning_rate": 3.51877299288622e-06,
+      "loss": 0.8571,
       "step": 25821
     },
     {
-      "epoch": 0.7091812913679932,
+      "epoch": 0.7327468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 4.117542774950127e-06,
-      "loss": 0.7747,
+      "learning_rate": 3.5180731092939514e-06,
+      "loss": 0.8974,
       "step": 25822
     },
     {
-      "epoch": 0.7092087555958364,
+      "epoch": 0.7327752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 4.116823462418998e-06,
-      "loss": 0.768,
+      "learning_rate": 3.5173732804548144e-06,
+      "loss": 0.7795,
       "step": 25823
     },
     {
-      "epoch": 0.7092362198236797,
+      "epoch": 0.7328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 4.116104196437842e-06,
-      "loss": 0.6877,
+      "learning_rate": 3.5166735063747125e-06,
+      "loss": 0.8244,
       "step": 25824
     },
     {
-      "epoch": 0.7092636840515228,
+      "epoch": 0.7328320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 4.115384977012348e-06,
-      "loss": 0.8669,
+      "learning_rate": 3.5159737870595623e-06,
+      "loss": 0.8177,
       "step": 25825
     },
     {
-      "epoch": 0.7092911482793661,
+      "epoch": 0.7328603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 4.114665804148211e-06,
-      "loss": 0.7872,
+      "learning_rate": 3.5152741225152755e-06,
+      "loss": 0.9066,
       "step": 25826
     },
     {
-      "epoch": 0.7093186125072094,
+      "epoch": 0.73288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.113946677851115e-06,
-      "loss": 0.8009,
+      "learning_rate": 3.514574512747756e-06,
+      "loss": 0.8496,
       "step": 25827
     },
     {
-      "epoch": 0.7093460767350526,
+      "epoch": 0.7329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 4.113227598126755e-06,
-      "loss": 0.8245,
+      "learning_rate": 3.5138749577629184e-06,
+      "loss": 0.7777,
       "step": 25828
     },
     {
-      "epoch": 0.7093735409628958,
+      "epoch": 0.7329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 4.112508564980817e-06,
-      "loss": 0.9012,
+      "learning_rate": 3.5131754575666687e-06,
+      "loss": 0.9174,
       "step": 25829
     },
     {
-      "epoch": 0.709401005190739,
+      "epoch": 0.7329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 4.111789578418991e-06,
-      "loss": 0.8184,
+      "learning_rate": 3.5124760121649195e-06,
+      "loss": 0.8329,
       "step": 25830
     },
     {
-      "epoch": 0.7094284694185823,
+      "epoch": 0.7330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 4.111070638446966e-06,
-      "loss": 0.7993,
+      "learning_rate": 3.5117766215635795e-06,
+      "loss": 0.7276,
       "step": 25831
     },
     {
-      "epoch": 0.7094559336464256,
+      "epoch": 0.7330306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 4.110351745070435e-06,
-      "loss": 0.8029,
+      "learning_rate": 3.5110772857685505e-06,
+      "loss": 0.9026,
       "step": 25832
     },
     {
-      "epoch": 0.7094833978742687,
+      "epoch": 0.7330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 4.109632898295079e-06,
-      "loss": 0.8896,
+      "learning_rate": 3.510378004785744e-06,
+      "loss": 0.8273,
       "step": 25833
     },
     {
-      "epoch": 0.709510862102112,
+      "epoch": 0.7330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 4.108914098126591e-06,
-      "loss": 0.7744,
+      "learning_rate": 3.5096787786210686e-06,
+      "loss": 0.7614,
       "step": 25834
     },
     {
-      "epoch": 0.7095383263299553,
+      "epoch": 0.7331157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 4.1081953445706545e-06,
-      "loss": 0.7948,
+      "learning_rate": 3.508979607280426e-06,
+      "loss": 0.882,
       "step": 25835
     },
     {
-      "epoch": 0.7095657905577984,
+      "epoch": 0.7331441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 4.107476637632963e-06,
-      "loss": 0.7602,
+      "learning_rate": 3.508280490769723e-06,
+      "loss": 0.8601,
       "step": 25836
     },
     {
-      "epoch": 0.7095932547856417,
+      "epoch": 0.733172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 4.106757977319195e-06,
-      "loss": 0.8699,
+      "learning_rate": 3.507581429094872e-06,
+      "loss": 0.7929,
       "step": 25837
     },
     {
-      "epoch": 0.7096207190134849,
+      "epoch": 0.7332009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 4.106039363635045e-06,
-      "loss": 0.7752,
+      "learning_rate": 3.506882422261767e-06,
+      "loss": 0.9541,
       "step": 25838
     },
     {
-      "epoch": 0.7096481832413282,
+      "epoch": 0.7332292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 4.10532079658619e-06,
-      "loss": 0.7736,
+      "learning_rate": 3.50618347027632e-06,
+      "loss": 0.8878,
       "step": 25839
     },
     {
-      "epoch": 0.7096756474691714,
+      "epoch": 0.7332576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 4.1046022761783215e-06,
-      "loss": 0.8347,
+      "learning_rate": 3.505484573144435e-06,
+      "loss": 0.7949,
       "step": 25840
     },
     {
-      "epoch": 0.7097031116970146,
+      "epoch": 0.7332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 4.103883802417122e-06,
-      "loss": 0.8647,
+      "learning_rate": 3.504785730872011e-06,
+      "loss": 0.7886,
       "step": 25841
     },
     {
-      "epoch": 0.7097305759248579,
+      "epoch": 0.7333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 4.103165375308279e-06,
-      "loss": 0.7641,
+      "learning_rate": 3.504086943464954e-06,
+      "loss": 0.8642,
       "step": 25842
     },
     {
-      "epoch": 0.709758040152701,
+      "epoch": 0.7333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.1024469948574785e-06,
-      "loss": 0.8554,
+      "learning_rate": 3.5033882109291693e-06,
+      "loss": 0.8181,
       "step": 25843
     },
     {
-      "epoch": 0.7097855043805443,
+      "epoch": 0.7333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 4.101728661070399e-06,
-      "loss": 0.8667,
+      "learning_rate": 3.5026895332705504e-06,
+      "loss": 0.8829,
       "step": 25844
     },
     {
-      "epoch": 0.7098129686083876,
+      "epoch": 0.7333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 4.10101037395273e-06,
-      "loss": 0.7743,
+      "learning_rate": 3.5019909104950124e-06,
+      "loss": 0.7758,
       "step": 25845
     },
     {
-      "epoch": 0.7098404328362308,
+      "epoch": 0.7334279228149829,
       "grad_norm": 0.0,
-      "learning_rate": 4.100292133510149e-06,
-      "loss": 0.8488,
+      "learning_rate": 3.5012923426084443e-06,
+      "loss": 0.7906,
       "step": 25846
     },
     {
-      "epoch": 0.709867897064074,
+      "epoch": 0.7334562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 4.099573939748341e-06,
-      "loss": 0.8575,
+      "learning_rate": 3.500593829616753e-06,
+      "loss": 0.8471,
       "step": 25847
     },
     {
-      "epoch": 0.7098953612919173,
+      "epoch": 0.7334846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 4.098855792672991e-06,
-      "loss": 0.9439,
+      "learning_rate": 3.499895371525841e-06,
+      "loss": 0.7176,
       "step": 25848
     },
     {
-      "epoch": 0.7099228255197605,
+      "epoch": 0.7335130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 4.09813769228978e-06,
-      "loss": 0.8816,
+      "learning_rate": 3.4991969683416015e-06,
+      "loss": 0.8539,
       "step": 25849
     },
     {
-      "epoch": 0.7099502897476038,
+      "epoch": 0.7335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 4.097419638604392e-06,
-      "loss": 0.851,
+      "learning_rate": 3.498498620069938e-06,
+      "loss": 0.7727,
       "step": 25850
     },
     {
-      "epoch": 0.7099777539754469,
+      "epoch": 0.7335698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 4.096701631622503e-06,
-      "loss": 0.82,
+      "learning_rate": 3.4978003267167516e-06,
+      "loss": 0.938,
       "step": 25851
     },
     {
-      "epoch": 0.7100052182032902,
+      "epoch": 0.7335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 4.095983671349802e-06,
-      "loss": 0.7796,
+      "learning_rate": 3.4971020882879348e-06,
+      "loss": 0.8665,
       "step": 25852
     },
     {
-      "epoch": 0.7100326824311335,
+      "epoch": 0.7336265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 4.0952657577919595e-06,
-      "loss": 0.8689,
+      "learning_rate": 3.4964039047893893e-06,
+      "loss": 0.8591,
       "step": 25853
     },
     {
-      "epoch": 0.7100601466589767,
+      "epoch": 0.7336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 4.094547890954663e-06,
-      "loss": 0.7303,
+      "learning_rate": 3.495705776227015e-06,
+      "loss": 0.8517,
       "step": 25854
     },
     {
-      "epoch": 0.7100876108868199,
+      "epoch": 0.733683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.093830070843592e-06,
-      "loss": 0.7885,
+      "learning_rate": 3.495007702606704e-06,
+      "loss": 0.8407,
       "step": 25855
     },
     {
-      "epoch": 0.7101150751146631,
+      "epoch": 0.7337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.093112297464428e-06,
-      "loss": 0.833,
+      "learning_rate": 3.494309683934355e-06,
+      "loss": 0.7791,
       "step": 25856
     },
     {
-      "epoch": 0.7101425393425064,
+      "epoch": 0.7337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 4.092394570822843e-06,
-      "loss": 0.641,
+      "learning_rate": 3.493611720215868e-06,
+      "loss": 0.8385,
       "step": 25857
     },
     {
-      "epoch": 0.7101700035703497,
+      "epoch": 0.7337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 4.091676890924522e-06,
-      "loss": 0.959,
+      "learning_rate": 3.4929138114571325e-06,
+      "loss": 0.9118,
       "step": 25858
     },
     {
-      "epoch": 0.7101974677981928,
+      "epoch": 0.7337968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 4.090959257775141e-06,
-      "loss": 0.8456,
+      "learning_rate": 3.492215957664046e-06,
+      "loss": 0.8186,
       "step": 25859
     },
     {
-      "epoch": 0.7102249320260361,
+      "epoch": 0.7338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 4.090241671380384e-06,
-      "loss": 0.883,
+      "learning_rate": 3.491518158842503e-06,
+      "loss": 0.8054,
       "step": 25860
     },
     {
-      "epoch": 0.7102523962538794,
+      "epoch": 0.7338535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 4.08952413174592e-06,
-      "loss": 0.892,
+      "learning_rate": 3.4908204149983993e-06,
+      "loss": 0.8098,
       "step": 25861
     },
     {
-      "epoch": 0.7102798604817225,
+      "epoch": 0.7338819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 4.088806638877434e-06,
-      "loss": 0.8439,
+      "learning_rate": 3.4901227261376326e-06,
+      "loss": 0.6579,
       "step": 25862
     },
     {
-      "epoch": 0.7103073247095658,
+      "epoch": 0.7339103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 4.088089192780596e-06,
-      "loss": 0.7938,
+      "learning_rate": 3.4894250922660865e-06,
+      "loss": 0.7321,
       "step": 25863
     },
     {
-      "epoch": 0.710334788937409,
+      "epoch": 0.7339387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 4.087371793461087e-06,
-      "loss": 0.8607,
+      "learning_rate": 3.4887275133896605e-06,
+      "loss": 0.9133,
       "step": 25864
     },
     {
-      "epoch": 0.7103622531652523,
+      "epoch": 0.7339670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 4.086654440924582e-06,
-      "loss": 0.7854,
+      "learning_rate": 3.48802998951425e-06,
+      "loss": 0.8855,
       "step": 25865
     },
     {
-      "epoch": 0.7103897173930955,
+      "epoch": 0.7339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 4.085937135176759e-06,
-      "loss": 0.831,
+      "learning_rate": 3.487332520645739e-06,
+      "loss": 0.8071,
       "step": 25866
     },
     {
-      "epoch": 0.7104171816209387,
+      "epoch": 0.7340238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 4.085219876223294e-06,
-      "loss": 0.8491,
+      "learning_rate": 3.4866351067900216e-06,
+      "loss": 0.8444,
       "step": 25867
     },
     {
-      "epoch": 0.710444645848782,
+      "epoch": 0.7340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 4.084502664069858e-06,
-      "loss": 0.7768,
+      "learning_rate": 3.4859377479529944e-06,
+      "loss": 0.7224,
       "step": 25868
     },
     {
-      "epoch": 0.7104721100766251,
+      "epoch": 0.7340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 4.0837854987221315e-06,
-      "loss": 0.879,
+      "learning_rate": 3.485240444140541e-06,
+      "loss": 0.7109,
       "step": 25869
     },
     {
-      "epoch": 0.7104995743044684,
+      "epoch": 0.7341089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 4.083068380185783e-06,
-      "loss": 0.9391,
+      "learning_rate": 3.484543195358554e-06,
+      "loss": 0.7838,
       "step": 25870
     },
     {
-      "epoch": 0.7105270385323117,
+      "epoch": 0.7341373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 4.082351308466489e-06,
-      "loss": 0.8435,
+      "learning_rate": 3.4838460016129273e-06,
+      "loss": 0.8314,
       "step": 25871
     },
     {
-      "epoch": 0.7105545027601549,
+      "epoch": 0.7341657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 4.081634283569923e-06,
-      "loss": 0.7789,
+      "learning_rate": 3.4831488629095433e-06,
+      "loss": 0.7311,
       "step": 25872
     },
     {
-      "epoch": 0.7105819669879981,
+      "epoch": 0.734194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 4.0809173055017634e-06,
-      "loss": 0.81,
+      "learning_rate": 3.4824517792542935e-06,
+      "loss": 0.8028,
       "step": 25873
     },
     {
-      "epoch": 0.7106094312158414,
+      "epoch": 0.73422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 4.080200374267675e-06,
-      "loss": 0.7812,
+      "learning_rate": 3.4817547506530656e-06,
+      "loss": 0.7758,
       "step": 25874
     },
     {
-      "epoch": 0.7106368954436846,
+      "epoch": 0.7342508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 4.079483489873335e-06,
-      "loss": 0.9175,
+      "learning_rate": 3.481057777111749e-06,
+      "loss": 0.7686,
       "step": 25875
     },
     {
-      "epoch": 0.7106643596715279,
+      "epoch": 0.7342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 4.078766652324418e-06,
-      "loss": 0.8761,
+      "learning_rate": 3.480360858636234e-06,
+      "loss": 0.8636,
       "step": 25876
     },
     {
-      "epoch": 0.710691823899371,
+      "epoch": 0.7343076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 4.078049861626589e-06,
-      "loss": 0.8281,
+      "learning_rate": 3.4796639952324007e-06,
+      "loss": 0.8305,
       "step": 25877
     },
     {
-      "epoch": 0.7107192881272143,
+      "epoch": 0.7343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 4.0773331177855245e-06,
-      "loss": 0.8196,
+      "learning_rate": 3.4789671869061383e-06,
+      "loss": 0.8301,
       "step": 25878
     },
     {
-      "epoch": 0.7107467523550576,
+      "epoch": 0.7343643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 4.076616420806898e-06,
-      "loss": 0.7802,
+      "learning_rate": 3.4782704336633365e-06,
+      "loss": 0.7524,
       "step": 25879
     },
     {
-      "epoch": 0.7107742165829007,
+      "epoch": 0.7343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 4.075899770696373e-06,
-      "loss": 0.776,
+      "learning_rate": 3.4775737355098737e-06,
+      "loss": 0.8666,
       "step": 25880
     },
     {
-      "epoch": 0.710801680810744,
+      "epoch": 0.7344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.075183167459625e-06,
-      "loss": 0.8342,
+      "learning_rate": 3.4768770924516393e-06,
+      "loss": 0.7736,
       "step": 25881
     },
     {
-      "epoch": 0.7108291450385873,
+      "epoch": 0.7344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 4.07446661110232e-06,
-      "loss": 0.8387,
+      "learning_rate": 3.4761805044945206e-06,
+      "loss": 0.7909,
       "step": 25882
     },
     {
-      "epoch": 0.7108566092664305,
+      "epoch": 0.734477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 4.073750101630133e-06,
-      "loss": 0.8131,
+      "learning_rate": 3.4754839716443956e-06,
+      "loss": 0.7528,
       "step": 25883
     },
     {
-      "epoch": 0.7108840734942737,
+      "epoch": 0.7345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 4.0730336390487336e-06,
-      "loss": 0.7821,
+      "learning_rate": 3.4747874939071503e-06,
+      "loss": 0.8647,
       "step": 25884
     },
     {
-      "epoch": 0.7109115377221169,
+      "epoch": 0.7345346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 4.072317223363787e-06,
-      "loss": 0.7892,
+      "learning_rate": 3.474091071288672e-06,
+      "loss": 0.9,
       "step": 25885
     },
     {
-      "epoch": 0.7109390019499602,
+      "epoch": 0.7345629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 4.0716008545809605e-06,
-      "loss": 0.935,
+      "learning_rate": 3.4733947037948367e-06,
+      "loss": 0.8804,
       "step": 25886
     },
     {
-      "epoch": 0.7109664661778035,
+      "epoch": 0.7345913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 4.070884532705923e-06,
-      "loss": 0.8708,
+      "learning_rate": 3.4726983914315294e-06,
+      "loss": 0.7332,
       "step": 25887
     },
     {
-      "epoch": 0.7109939304056466,
+      "epoch": 0.7346197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 4.070168257744345e-06,
-      "loss": 0.8668,
+      "learning_rate": 3.472002134204635e-06,
+      "loss": 0.8235,
       "step": 25888
     },
     {
-      "epoch": 0.7110213946334899,
+      "epoch": 0.7346481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 4.069452029701892e-06,
-      "loss": 0.8629,
+      "learning_rate": 3.4713059321200293e-06,
+      "loss": 0.8656,
       "step": 25889
     },
     {
-      "epoch": 0.7110488588613331,
+      "epoch": 0.7346765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 4.0687358485842335e-06,
-      "loss": 0.7744,
+      "learning_rate": 3.470609785183592e-06,
+      "loss": 0.7643,
       "step": 25890
     },
     {
-      "epoch": 0.7110763230891763,
+      "epoch": 0.7347048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 4.0680197143970365e-06,
-      "loss": 0.8986,
+      "learning_rate": 3.4699136934012145e-06,
+      "loss": 0.8361,
       "step": 25891
     },
     {
-      "epoch": 0.7111037873170196,
+      "epoch": 0.734733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 4.067303627145962e-06,
-      "loss": 0.8722,
+      "learning_rate": 3.469217656778766e-06,
+      "loss": 0.7938,
       "step": 25892
     },
     {
-      "epoch": 0.7111312515448628,
+      "epoch": 0.7347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 4.066587586836684e-06,
-      "loss": 0.8192,
+      "learning_rate": 3.4685216753221295e-06,
+      "loss": 0.7911,
       "step": 25893
     },
     {
-      "epoch": 0.7111587157727061,
+      "epoch": 0.7347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 4.065871593474859e-06,
-      "loss": 0.8393,
+      "learning_rate": 3.467825749037188e-06,
+      "loss": 0.8927,
       "step": 25894
     },
     {
-      "epoch": 0.7111861800005493,
+      "epoch": 0.7348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 4.065155647066156e-06,
-      "loss": 0.8994,
+      "learning_rate": 3.4671298779298123e-06,
+      "loss": 0.829,
       "step": 25895
     },
     {
-      "epoch": 0.7112136442283925,
+      "epoch": 0.7348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 4.064439747616241e-06,
-      "loss": 0.8837,
+      "learning_rate": 3.4664340620058844e-06,
+      "loss": 0.7851,
       "step": 25896
     },
     {
-      "epoch": 0.7112411084562358,
+      "epoch": 0.7348751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 4.063723895130781e-06,
-      "loss": 0.828,
+      "learning_rate": 3.4657383012712844e-06,
+      "loss": 0.7832,
       "step": 25897
     },
     {
-      "epoch": 0.711268572684079,
+      "epoch": 0.7349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 4.0630080896154346e-06,
-      "loss": 0.8693,
+      "learning_rate": 3.4650425957318846e-06,
+      "loss": 0.8512,
       "step": 25898
     },
     {
-      "epoch": 0.7112960369119222,
+      "epoch": 0.7349318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 4.0622923310758665e-06,
-      "loss": 0.8738,
+      "learning_rate": 3.4643469453935653e-06,
+      "loss": 0.9693,
       "step": 25899
     },
     {
-      "epoch": 0.7113235011397655,
+      "epoch": 0.7349602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 4.0615766195177466e-06,
-      "loss": 0.8558,
+      "learning_rate": 3.4636513502621994e-06,
+      "loss": 0.7858,
       "step": 25900
     },
     {
-      "epoch": 0.7113509653676087,
+      "epoch": 0.7349886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 4.060860954946728e-06,
-      "loss": 0.8714,
+      "learning_rate": 3.4629558103436634e-06,
+      "loss": 0.7724,
       "step": 25901
     },
     {
-      "epoch": 0.7113784295954519,
+      "epoch": 0.735017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 4.060145337368479e-06,
-      "loss": 0.808,
+      "learning_rate": 3.4622603256438368e-06,
+      "loss": 0.8056,
       "step": 25902
     },
     {
-      "epoch": 0.7114058938232951,
+      "epoch": 0.7350454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 4.059429766788664e-06,
-      "loss": 0.7773,
+      "learning_rate": 3.461564896168589e-06,
+      "loss": 0.8724,
       "step": 25903
     },
     {
-      "epoch": 0.7114333580511384,
+      "epoch": 0.7350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 4.058714243212938e-06,
-      "loss": 0.7386,
+      "learning_rate": 3.4608695219237954e-06,
+      "loss": 0.7657,
       "step": 25904
     },
     {
-      "epoch": 0.7114608222789817,
+      "epoch": 0.7351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 4.057998766646967e-06,
-      "loss": 0.9011,
+      "learning_rate": 3.4601742029153306e-06,
+      "loss": 0.8093,
       "step": 25905
     },
     {
-      "epoch": 0.7114882865068248,
+      "epoch": 0.7351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 4.057283337096412e-06,
-      "loss": 0.7484,
+      "learning_rate": 3.459478939149069e-06,
+      "loss": 0.8617,
       "step": 25906
     },
     {
-      "epoch": 0.7115157507346681,
+      "epoch": 0.7351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 4.056567954566931e-06,
-      "loss": 0.808,
+      "learning_rate": 3.4587837306308823e-06,
+      "loss": 0.9177,
       "step": 25907
     },
     {
-      "epoch": 0.7115432149625114,
+      "epoch": 0.7351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 4.055852619064192e-06,
-      "loss": 0.6907,
+      "learning_rate": 3.4580885773666475e-06,
+      "loss": 0.8471,
       "step": 25908
     },
     {
-      "epoch": 0.7115706791903545,
+      "epoch": 0.7352156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 4.05513733059385e-06,
-      "loss": 0.9084,
+      "learning_rate": 3.457393479362229e-06,
+      "loss": 0.841,
       "step": 25909
     },
     {
-      "epoch": 0.7115981434181978,
+      "epoch": 0.735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 4.054422089161559e-06,
-      "loss": 0.8191,
+      "learning_rate": 3.4566984366235023e-06,
+      "loss": 0.7367,
       "step": 25910
     },
     {
-      "epoch": 0.711625607646041,
+      "epoch": 0.735272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 4.053706894772985e-06,
-      "loss": 0.8577,
+      "learning_rate": 3.45600344915634e-06,
+      "loss": 0.8755,
       "step": 25911
     },
     {
-      "epoch": 0.7116530718738843,
+      "epoch": 0.7353007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 4.052991747433786e-06,
-      "loss": 0.9981,
+      "learning_rate": 3.4553085169666088e-06,
+      "loss": 0.8166,
       "step": 25912
     },
     {
-      "epoch": 0.7116805361017275,
+      "epoch": 0.7353291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 4.052276647149618e-06,
-      "loss": 0.8649,
+      "learning_rate": 3.4546136400601803e-06,
+      "loss": 0.805,
       "step": 25913
     },
     {
-      "epoch": 0.7117080003295707,
+      "epoch": 0.7353575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 4.051561593926142e-06,
-      "loss": 0.9175,
+      "learning_rate": 3.4539188184429275e-06,
+      "loss": 0.8872,
       "step": 25914
     },
     {
-      "epoch": 0.711735464557414,
+      "epoch": 0.7353859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 4.050846587769019e-06,
-      "loss": 0.7717,
+      "learning_rate": 3.453224052120714e-06,
+      "loss": 0.7935,
       "step": 25915
     },
     {
-      "epoch": 0.7117629287852572,
+      "epoch": 0.7354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 4.050131628683897e-06,
-      "loss": 0.7907,
+      "learning_rate": 3.4525293410994108e-06,
+      "loss": 0.7697,
       "step": 25916
     },
     {
-      "epoch": 0.7117903930131004,
+      "epoch": 0.7354426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 4.049416716676443e-06,
-      "loss": 0.7825,
+      "learning_rate": 3.4518346853848906e-06,
+      "loss": 0.7645,
       "step": 25917
     },
     {
-      "epoch": 0.7118178572409437,
+      "epoch": 0.7354710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 4.048701851752306e-06,
-      "loss": 0.8483,
+      "learning_rate": 3.4511400849830135e-06,
+      "loss": 0.9471,
       "step": 25918
     },
     {
-      "epoch": 0.7118453214687869,
+      "epoch": 0.7354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.047987033917144e-06,
-      "loss": 0.8408,
+      "learning_rate": 3.4504455398996495e-06,
+      "loss": 0.8698,
       "step": 25919
     },
     {
-      "epoch": 0.7118727856966302,
+      "epoch": 0.735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 4.047272263176615e-06,
-      "loss": 0.8658,
+      "learning_rate": 3.4497510501406673e-06,
+      "loss": 0.9215,
       "step": 25920
     },
     {
-      "epoch": 0.7119002499244734,
+      "epoch": 0.7355561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 4.0465575395363774e-06,
-      "loss": 0.8874,
+      "learning_rate": 3.449056615711932e-06,
+      "loss": 0.8848,
       "step": 25921
     },
     {
-      "epoch": 0.7119277141523166,
+      "epoch": 0.7355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 4.045842863002078e-06,
-      "loss": 0.764,
+      "learning_rate": 3.4483622366193146e-06,
+      "loss": 0.7391,
       "step": 25922
     },
     {
-      "epoch": 0.7119551783801599,
+      "epoch": 0.7356129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 4.045128233579379e-06,
-      "loss": 0.8527,
+      "learning_rate": 3.4476679128686718e-06,
+      "loss": 0.7298,
       "step": 25923
     },
     {
-      "epoch": 0.711982642608003,
+      "epoch": 0.7356413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 4.044413651273933e-06,
-      "loss": 0.9039,
+      "learning_rate": 3.446973644465872e-06,
+      "loss": 0.8087,
       "step": 25924
     },
     {
-      "epoch": 0.7120101068358463,
+      "epoch": 0.7356696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 4.043699116091391e-06,
-      "loss": 0.883,
+      "learning_rate": 3.4462794314167846e-06,
+      "loss": 0.8091,
       "step": 25925
     },
     {
-      "epoch": 0.7120375710636896,
+      "epoch": 0.7356980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 4.042984628037411e-06,
-      "loss": 0.8982,
+      "learning_rate": 3.4455852737272654e-06,
+      "loss": 0.7833,
       "step": 25926
     },
     {
-      "epoch": 0.7120650352915328,
+      "epoch": 0.7357264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 4.042270187117642e-06,
-      "loss": 0.8571,
+      "learning_rate": 3.444891171403183e-06,
+      "loss": 0.8565,
       "step": 25927
     },
     {
-      "epoch": 0.712092499519376,
+      "epoch": 0.7357548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 4.041555793337738e-06,
-      "loss": 0.8308,
+      "learning_rate": 3.4441971244504023e-06,
+      "loss": 0.7887,
       "step": 25928
     },
     {
-      "epoch": 0.7121199637472192,
+      "epoch": 0.735783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 4.0408414467033526e-06,
-      "loss": 0.7987,
+      "learning_rate": 3.4435031328747793e-06,
+      "loss": 0.851,
       "step": 25929
     },
     {
-      "epoch": 0.7121474279750625,
+      "epoch": 0.7358115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 4.040127147220139e-06,
-      "loss": 0.8067,
+      "learning_rate": 3.4428091966821806e-06,
+      "loss": 0.7951,
       "step": 25930
     },
     {
-      "epoch": 0.7121748922029058,
+      "epoch": 0.7358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 4.039412894893747e-06,
-      "loss": 0.7605,
+      "learning_rate": 3.4421153158784716e-06,
+      "loss": 0.7853,
       "step": 25931
     },
     {
-      "epoch": 0.7122023564307489,
+      "epoch": 0.7358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.038698689729834e-06,
-      "loss": 0.7585,
+      "learning_rate": 3.4414214904695045e-06,
+      "loss": 0.9489,
       "step": 25932
     },
     {
-      "epoch": 0.7122298206585922,
+      "epoch": 0.7358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 4.037984531734045e-06,
-      "loss": 0.7755,
+      "learning_rate": 3.4407277204611456e-06,
+      "loss": 0.709,
       "step": 25933
     },
     {
-      "epoch": 0.7122572848864355,
+      "epoch": 0.7359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 4.037270420912029e-06,
-      "loss": 0.8762,
+      "learning_rate": 3.4400340058592587e-06,
+      "loss": 0.8476,
       "step": 25934
     },
     {
-      "epoch": 0.7122847491142786,
+      "epoch": 0.7359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 4.036556357269439e-06,
-      "loss": 0.8742,
+      "learning_rate": 3.439340346669695e-06,
+      "loss": 0.7719,
       "step": 25935
     },
     {
-      "epoch": 0.7123122133421219,
+      "epoch": 0.7359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 4.035842340811925e-06,
-      "loss": 0.9073,
+      "learning_rate": 3.438646742898318e-06,
+      "loss": 0.8978,
       "step": 25936
     },
     {
-      "epoch": 0.7123396775699651,
+      "epoch": 0.7360102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 4.035128371545136e-06,
-      "loss": 0.8119,
+      "learning_rate": 3.437953194550988e-06,
+      "loss": 0.8167,
       "step": 25937
     },
     {
-      "epoch": 0.7123671417978084,
+      "epoch": 0.736038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 4.034414449474727e-06,
-      "loss": 0.8513,
+      "learning_rate": 3.437259701633562e-06,
+      "loss": 0.9026,
       "step": 25938
     },
     {
-      "epoch": 0.7123946060256516,
+      "epoch": 0.736066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 4.033700574606337e-06,
-      "loss": 0.7846,
+      "learning_rate": 3.4365662641519004e-06,
+      "loss": 0.9208,
       "step": 25939
     },
     {
-      "epoch": 0.7124220702534948,
+      "epoch": 0.7360953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 4.032986746945623e-06,
-      "loss": 0.7394,
+      "learning_rate": 3.435872882111857e-06,
+      "loss": 0.8072,
       "step": 25940
     },
     {
-      "epoch": 0.7124495344813381,
+      "epoch": 0.7361237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 4.032272966498226e-06,
-      "loss": 0.7371,
+      "learning_rate": 3.435179555519289e-06,
+      "loss": 0.7649,
       "step": 25941
     },
     {
-      "epoch": 0.7124769987091812,
+      "epoch": 0.7361520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 4.031559233269796e-06,
-      "loss": 0.7555,
+      "learning_rate": 3.4344862843800587e-06,
+      "loss": 0.7643,
       "step": 25942
     },
     {
-      "epoch": 0.7125044629370245,
+      "epoch": 0.7361804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 4.0308455472659825e-06,
-      "loss": 0.8461,
+      "learning_rate": 3.433793068700014e-06,
+      "loss": 0.7949,
       "step": 25943
     },
     {
-      "epoch": 0.7125319271648678,
+      "epoch": 0.7362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 4.030131908492434e-06,
-      "loss": 0.8743,
+      "learning_rate": 3.433099908485016e-06,
+      "loss": 0.8095,
       "step": 25944
     },
     {
-      "epoch": 0.712559391392711,
+      "epoch": 0.7362372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 4.0294183169547915e-06,
-      "loss": 0.7957,
+      "learning_rate": 3.43240680374092e-06,
+      "loss": 0.7477,
       "step": 25945
     },
     {
-      "epoch": 0.7125868556205542,
+      "epoch": 0.7362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 4.028704772658704e-06,
-      "loss": 0.7866,
+      "learning_rate": 3.4317137544735755e-06,
+      "loss": 0.6721,
       "step": 25946
     },
     {
-      "epoch": 0.7126143198483975,
+      "epoch": 0.7362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 4.027991275609817e-06,
-      "loss": 0.818,
+      "learning_rate": 3.4310207606888413e-06,
+      "loss": 0.7711,
       "step": 25947
     },
     {
-      "epoch": 0.7126417840762407,
+      "epoch": 0.7363223609534619,
       "grad_norm": 0.0,
-      "learning_rate": 4.02727782581378e-06,
-      "loss": 0.8394,
+      "learning_rate": 3.430327822392573e-06,
+      "loss": 0.7371,
       "step": 25948
     },
     {
-      "epoch": 0.712669248304084,
+      "epoch": 0.7363507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 4.02656442327623e-06,
-      "loss": 0.8209,
+      "learning_rate": 3.429634939590617e-06,
+      "loss": 0.7376,
       "step": 25949
     },
     {
-      "epoch": 0.7126967125319271,
+      "epoch": 0.7363791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 4.025851068002822e-06,
-      "loss": 0.7646,
+      "learning_rate": 3.42894211228883e-06,
+      "loss": 0.8651,
       "step": 25950
     },
     {
-      "epoch": 0.7127241767597704,
+      "epoch": 0.7364074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 4.025137759999189e-06,
-      "loss": 0.9112,
+      "learning_rate": 3.428249340493065e-06,
+      "loss": 0.8254,
       "step": 25951
     },
     {
-      "epoch": 0.7127516409876137,
+      "epoch": 0.7364358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 4.024424499270981e-06,
-      "loss": 0.9179,
+      "learning_rate": 3.427556624209173e-06,
+      "loss": 0.7034,
       "step": 25952
     },
     {
-      "epoch": 0.7127791052154568,
+      "epoch": 0.7364642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 4.0237112858238415e-06,
-      "loss": 0.8286,
+      "learning_rate": 3.426863963443009e-06,
+      "loss": 0.8272,
       "step": 25953
     },
     {
-      "epoch": 0.7128065694433001,
+      "epoch": 0.7364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 4.022998119663413e-06,
-      "loss": 0.8164,
+      "learning_rate": 3.4261713582004173e-06,
+      "loss": 0.7747,
       "step": 25954
     },
     {
-      "epoch": 0.7128340336711433,
+      "epoch": 0.7365209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 4.022285000795339e-06,
-      "loss": 0.8206,
+      "learning_rate": 3.4254788084872513e-06,
+      "loss": 0.7429,
       "step": 25955
     },
     {
-      "epoch": 0.7128614978989866,
+      "epoch": 0.7365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 4.021571929225264e-06,
-      "loss": 0.8007,
+      "learning_rate": 3.424786314309365e-06,
+      "loss": 0.8613,
       "step": 25956
     },
     {
-      "epoch": 0.7128889621268298,
+      "epoch": 0.736577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.0208589049588285e-06,
-      "loss": 0.8188,
+      "learning_rate": 3.4240938756726016e-06,
+      "loss": 0.8687,
       "step": 25957
     },
     {
-      "epoch": 0.712916426354673,
+      "epoch": 0.7366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 4.02014592800167e-06,
-      "loss": 0.8513,
+      "learning_rate": 3.4234014925828116e-06,
+      "loss": 0.8895,
       "step": 25958
     },
     {
-      "epoch": 0.7129438905825163,
+      "epoch": 0.7366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 4.019432998359433e-06,
-      "loss": 0.8599,
+      "learning_rate": 3.422709165045849e-06,
+      "loss": 0.8457,
       "step": 25959
     },
     {
-      "epoch": 0.7129713548103596,
+      "epoch": 0.7366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 4.018720116037759e-06,
-      "loss": 0.9413,
+      "learning_rate": 3.4220168930675536e-06,
+      "loss": 0.8267,
       "step": 25960
     },
     {
-      "epoch": 0.7129988190382027,
+      "epoch": 0.7366912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 4.018007281042288e-06,
-      "loss": 0.836,
+      "learning_rate": 3.4213246766537776e-06,
+      "loss": 0.8789,
       "step": 25961
     },
     {
-      "epoch": 0.713026283266046,
+      "epoch": 0.7367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 4.017294493378664e-06,
-      "loss": 0.7794,
+      "learning_rate": 3.4206325158103715e-06,
+      "loss": 0.686,
       "step": 25962
     },
     {
-      "epoch": 0.7130537474938892,
+      "epoch": 0.7367480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 4.016581753052521e-06,
-      "loss": 0.8077,
+      "learning_rate": 3.4199404105431755e-06,
+      "loss": 0.8447,
       "step": 25963
     },
     {
-      "epoch": 0.7130812117217324,
+      "epoch": 0.7367763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 4.015869060069504e-06,
-      "loss": 0.8247,
+      "learning_rate": 3.4192483608580375e-06,
+      "loss": 0.7986,
       "step": 25964
     },
     {
-      "epoch": 0.7131086759495757,
+      "epoch": 0.7368047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 4.015156414435245e-06,
-      "loss": 0.8992,
+      "learning_rate": 3.418556366760808e-06,
+      "loss": 0.88,
       "step": 25965
     },
     {
-      "epoch": 0.7131361401774189,
+      "epoch": 0.7368331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 4.014443816155387e-06,
-      "loss": 0.884,
+      "learning_rate": 3.4178644282573227e-06,
+      "loss": 0.715,
       "step": 25966
     },
     {
-      "epoch": 0.7131636044052622,
+      "epoch": 0.7368615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 4.013731265235568e-06,
-      "loss": 0.7575,
+      "learning_rate": 3.4171725453534375e-06,
+      "loss": 0.757,
       "step": 25967
     },
     {
-      "epoch": 0.7131910686331053,
+      "epoch": 0.7368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 4.01301876168143e-06,
-      "loss": 0.8103,
+      "learning_rate": 3.4164807180549898e-06,
+      "loss": 0.8422,
       "step": 25968
     },
     {
-      "epoch": 0.7132185328609486,
+      "epoch": 0.7369182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 4.0123063054986026e-06,
-      "loss": 0.6815,
+      "learning_rate": 3.4157889463678252e-06,
+      "loss": 0.9041,
       "step": 25969
     },
     {
-      "epoch": 0.7132459970887919,
+      "epoch": 0.7369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 4.011593896692727e-06,
-      "loss": 0.9429,
+      "learning_rate": 3.4150972302977914e-06,
+      "loss": 0.8363,
       "step": 25970
     },
     {
-      "epoch": 0.713273461316635,
+      "epoch": 0.7369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 4.010881535269441e-06,
-      "loss": 0.8461,
+      "learning_rate": 3.4144055698507227e-06,
+      "loss": 0.819,
       "step": 25971
     },
     {
-      "epoch": 0.7133009255444783,
+      "epoch": 0.7370034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 4.0101692212343835e-06,
-      "loss": 0.8539,
+      "learning_rate": 3.413713965032467e-06,
+      "loss": 0.856,
       "step": 25972
     },
     {
-      "epoch": 0.7133283897723216,
+      "epoch": 0.7370317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 4.009456954593184e-06,
-      "loss": 0.8401,
+      "learning_rate": 3.413022415848868e-06,
+      "loss": 0.8104,
       "step": 25973
     },
     {
-      "epoch": 0.7133558540001648,
+      "epoch": 0.7370601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 4.008744735351484e-06,
-      "loss": 0.8575,
+      "learning_rate": 3.412330922305762e-06,
+      "loss": 0.7335,
       "step": 25974
     },
     {
-      "epoch": 0.713383318228008,
+      "epoch": 0.737088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 4.008032563514914e-06,
-      "loss": 0.7927,
+      "learning_rate": 3.411639484408993e-06,
+      "loss": 0.7954,
       "step": 25975
     },
     {
-      "epoch": 0.7134107824558512,
+      "epoch": 0.737116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.007320439089112e-06,
-      "loss": 0.8031,
+      "learning_rate": 3.410948102164404e-06,
+      "loss": 0.835,
       "step": 25976
     },
     {
-      "epoch": 0.7134382466836945,
+      "epoch": 0.7371452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 4.00660836207971e-06,
-      "loss": 0.7661,
+      "learning_rate": 3.4102567755778304e-06,
+      "loss": 0.8441,
       "step": 25977
     },
     {
-      "epoch": 0.7134657109115378,
+      "epoch": 0.7371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 4.005896332492346e-06,
-      "loss": 0.7724,
+      "learning_rate": 3.4095655046551137e-06,
+      "loss": 0.8432,
       "step": 25978
     },
     {
-      "epoch": 0.7134931751393809,
+      "epoch": 0.7372020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 4.005184350332655e-06,
-      "loss": 0.8654,
+      "learning_rate": 3.4088742894020966e-06,
+      "loss": 0.8759,
       "step": 25979
     },
     {
-      "epoch": 0.7135206393672242,
+      "epoch": 0.7372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 4.004472415606265e-06,
-      "loss": 0.8094,
+      "learning_rate": 3.4081831298246115e-06,
+      "loss": 0.7245,
       "step": 25980
     },
     {
-      "epoch": 0.7135481035950674,
+      "epoch": 0.7372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 4.003760528318814e-06,
-      "loss": 0.929,
+      "learning_rate": 3.4074920259284995e-06,
+      "loss": 0.8068,
       "step": 25981
     },
     {
-      "epoch": 0.7135755678229107,
+      "epoch": 0.7372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 4.003048688475929e-06,
-      "loss": 0.7967,
+      "learning_rate": 3.4068009777195987e-06,
+      "loss": 0.8405,
       "step": 25982
     },
     {
-      "epoch": 0.7136030320507539,
+      "epoch": 0.7373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 4.002336896083246e-06,
-      "loss": 0.8812,
+      "learning_rate": 3.406109985203746e-06,
+      "loss": 0.7476,
       "step": 25983
     },
     {
-      "epoch": 0.7136304962785971,
+      "epoch": 0.7373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 4.001625151146396e-06,
-      "loss": 0.7786,
+      "learning_rate": 3.4054190483867833e-06,
+      "loss": 0.8914,
       "step": 25984
     },
     {
-      "epoch": 0.7136579605064404,
+      "epoch": 0.737372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 4.000913453671017e-06,
-      "loss": 0.8615,
+      "learning_rate": 3.404728167274538e-06,
+      "loss": 0.7889,
       "step": 25985
     },
     {
-      "epoch": 0.7136854247342836,
+      "epoch": 0.7374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 4.00020180366273e-06,
-      "loss": 0.8055,
+      "learning_rate": 3.4040373418728503e-06,
+      "loss": 0.8567,
       "step": 25986
     },
     {
-      "epoch": 0.7137128889621268,
+      "epoch": 0.7374290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 3.99949020112717e-06,
-      "loss": 0.8446,
+      "learning_rate": 3.403346572187559e-06,
+      "loss": 0.8417,
       "step": 25987
     },
     {
-      "epoch": 0.7137403531899701,
+      "epoch": 0.7374574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 3.998778646069972e-06,
-      "loss": 0.7439,
+      "learning_rate": 3.4026558582244928e-06,
+      "loss": 0.8149,
       "step": 25988
     },
     {
-      "epoch": 0.7137678174178133,
+      "epoch": 0.7374858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 3.998067138496758e-06,
-      "loss": 0.7506,
+      "learning_rate": 3.4019651999894898e-06,
+      "loss": 0.7348,
       "step": 25989
     },
     {
-      "epoch": 0.7137952816456565,
+      "epoch": 0.7375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 3.997355678413162e-06,
-      "loss": 0.869,
+      "learning_rate": 3.4012745974883852e-06,
+      "loss": 0.8018,
       "step": 25990
     },
     {
-      "epoch": 0.7138227458734998,
+      "epoch": 0.7375425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 3.996644265824816e-06,
-      "loss": 0.7948,
+      "learning_rate": 3.4005840507270084e-06,
+      "loss": 0.7655,
       "step": 25991
     },
     {
-      "epoch": 0.713850210101343,
+      "epoch": 0.7375709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 3.995932900737343e-06,
-      "loss": 0.8456,
+      "learning_rate": 3.399893559711195e-06,
+      "loss": 0.8176,
       "step": 25992
     },
     {
-      "epoch": 0.7138776743291863,
+      "epoch": 0.7375993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 3.9952215831563745e-06,
-      "loss": 0.8811,
+      "learning_rate": 3.3992031244467806e-06,
+      "loss": 0.7223,
       "step": 25993
     },
     {
-      "epoch": 0.7139051385570294,
+      "epoch": 0.737627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 3.994510313087538e-06,
-      "loss": 0.8019,
+      "learning_rate": 3.3985127449395893e-06,
+      "loss": 0.8637,
       "step": 25994
     },
     {
-      "epoch": 0.7139326027848727,
+      "epoch": 0.7376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 3.993799090536465e-06,
-      "loss": 0.739,
+      "learning_rate": 3.397822421195459e-06,
+      "loss": 0.8784,
       "step": 25995
     },
     {
-      "epoch": 0.713960067012716,
+      "epoch": 0.7376844494892169,
       "grad_norm": 0.0,
-      "learning_rate": 3.993087915508781e-06,
-      "loss": 0.8255,
+      "learning_rate": 3.3971321532202228e-06,
+      "loss": 0.8859,
       "step": 25996
     },
     {
-      "epoch": 0.7139875312405591,
+      "epoch": 0.7377128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 3.992376788010113e-06,
-      "loss": 0.7333,
+      "learning_rate": 3.3964419410197013e-06,
+      "loss": 0.8209,
       "step": 25997
     },
     {
-      "epoch": 0.7140149954684024,
+      "epoch": 0.7377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 3.991665708046083e-06,
-      "loss": 0.8895,
+      "learning_rate": 3.3957517845997347e-06,
+      "loss": 0.9488,
       "step": 25998
     },
     {
-      "epoch": 0.7140424596962457,
+      "epoch": 0.7377695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 3.990954675622323e-06,
-      "loss": 0.8851,
+      "learning_rate": 3.3950616839661532e-06,
+      "loss": 0.7685,
       "step": 25999
     },
     {
-      "epoch": 0.7140699239240889,
+      "epoch": 0.7377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 3.990243690744455e-06,
-      "loss": 0.825,
+      "learning_rate": 3.3943716391247796e-06,
+      "loss": 0.8137,
       "step": 26000
     },
     {
-      "epoch": 0.7140973881519321,
+      "epoch": 0.7378263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 3.989532753418108e-06,
-      "loss": 0.7412,
+      "learning_rate": 3.393681650081445e-06,
+      "loss": 0.8641,
       "step": 26001
     },
     {
-      "epoch": 0.7141248523797753,
+      "epoch": 0.7378547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 3.988821863648906e-06,
-      "loss": 0.726,
+      "learning_rate": 3.3929917168419825e-06,
+      "loss": 0.792,
       "step": 26002
     },
     {
-      "epoch": 0.7141523166076186,
+      "epoch": 0.7378830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 3.988111021442478e-06,
-      "loss": 0.795,
+      "learning_rate": 3.392301839412212e-06,
+      "loss": 0.8175,
       "step": 26003
     },
     {
-      "epoch": 0.7141797808354619,
+      "epoch": 0.737911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 3.987400226804441e-06,
-      "loss": 0.8048,
+      "learning_rate": 3.391612017797965e-06,
+      "loss": 0.856,
       "step": 26004
     },
     {
-      "epoch": 0.714207245063305,
+      "epoch": 0.7379398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 3.986689479740425e-06,
-      "loss": 0.8418,
+      "learning_rate": 3.3909222520050723e-06,
+      "loss": 0.8401,
       "step": 26005
     },
     {
-      "epoch": 0.7142347092911483,
+      "epoch": 0.7379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 3.985978780256048e-06,
-      "loss": 0.873,
+      "learning_rate": 3.390232542039352e-06,
+      "loss": 0.7745,
       "step": 26006
     },
     {
-      "epoch": 0.7142621735189915,
+      "epoch": 0.7379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 3.985268128356937e-06,
-      "loss": 0.8191,
+      "learning_rate": 3.3895428879066384e-06,
+      "loss": 0.8538,
       "step": 26007
     },
     {
-      "epoch": 0.7142896377468347,
+      "epoch": 0.7380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 3.984557524048714e-06,
-      "loss": 0.7294,
+      "learning_rate": 3.38885328961275e-06,
+      "loss": 0.8448,
       "step": 26008
     },
     {
-      "epoch": 0.714317101974678,
+      "epoch": 0.7380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 3.983846967337005e-06,
-      "loss": 0.8366,
+      "learning_rate": 3.3881637471635154e-06,
+      "loss": 0.994,
       "step": 26009
     },
     {
-      "epoch": 0.7143445662025212,
+      "epoch": 0.7380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 3.983136458227427e-06,
-      "loss": 0.9136,
+      "learning_rate": 3.3874742605647613e-06,
+      "loss": 0.8802,
       "step": 26010
     },
     {
-      "epoch": 0.7143720304303645,
+      "epoch": 0.7381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 3.982425996725603e-06,
-      "loss": 0.895,
+      "learning_rate": 3.386784829822307e-06,
+      "loss": 0.7772,
       "step": 26011
     },
     {
-      "epoch": 0.7143994946582077,
+      "epoch": 0.738138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 3.98171558283716e-06,
-      "loss": 0.929,
+      "learning_rate": 3.3860954549419746e-06,
+      "loss": 0.7852,
       "step": 26012
     },
     {
-      "epoch": 0.7144269588860509,
+      "epoch": 0.73816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 3.98100521656771e-06,
-      "loss": 0.8497,
+      "learning_rate": 3.3854061359295985e-06,
+      "loss": 0.7874,
       "step": 26013
     },
     {
-      "epoch": 0.7144544231138942,
+      "epoch": 0.7381952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 3.980294897922878e-06,
-      "loss": 0.861,
+      "learning_rate": 3.3847168727909896e-06,
+      "loss": 0.8182,
       "step": 26014
     },
     {
-      "epoch": 0.7144818873417373,
+      "epoch": 0.7382236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 3.9795846269082894e-06,
-      "loss": 0.7563,
+      "learning_rate": 3.384027665531976e-06,
+      "loss": 0.8126,
       "step": 26015
     },
     {
-      "epoch": 0.7145093515695806,
+      "epoch": 0.7382519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 3.978874403529555e-06,
-      "loss": 0.8643,
+      "learning_rate": 3.3833385141583808e-06,
+      "loss": 0.8795,
       "step": 26016
     },
     {
-      "epoch": 0.7145368157974239,
+      "epoch": 0.7382803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 3.9781642277923e-06,
-      "loss": 0.8106,
+      "learning_rate": 3.3826494186760195e-06,
+      "loss": 0.8432,
       "step": 26017
     },
     {
-      "epoch": 0.7145642800252671,
+      "epoch": 0.7383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 3.97745409970214e-06,
-      "loss": 0.8832,
+      "learning_rate": 3.381960379090715e-06,
+      "loss": 0.7857,
       "step": 26018
     },
     {
-      "epoch": 0.7145917442531103,
+      "epoch": 0.7383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 3.976744019264698e-06,
-      "loss": 0.8512,
+      "learning_rate": 3.381271395408293e-06,
+      "loss": 0.8122,
       "step": 26019
     },
     {
-      "epoch": 0.7146192084809535,
+      "epoch": 0.7383654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 3.976033986485592e-06,
-      "loss": 0.865,
+      "learning_rate": 3.3805824676345656e-06,
+      "loss": 0.7438,
       "step": 26020
     },
     {
-      "epoch": 0.7146466727087968,
+      "epoch": 0.7383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 3.9753240013704405e-06,
-      "loss": 0.8048,
+      "learning_rate": 3.3798935957753555e-06,
+      "loss": 0.7259,
       "step": 26021
     },
     {
-      "epoch": 0.7146741369366401,
+      "epoch": 0.738422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 3.974614063924856e-06,
-      "loss": 0.8406,
+      "learning_rate": 3.379204779836486e-06,
+      "loss": 0.8201,
       "step": 26022
     },
     {
-      "epoch": 0.7147016011644832,
+      "epoch": 0.7384506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 3.973904174154457e-06,
-      "loss": 0.8081,
+      "learning_rate": 3.3785160198237663e-06,
+      "loss": 0.875,
       "step": 26023
     },
     {
-      "epoch": 0.7147290653923265,
+      "epoch": 0.7384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 3.973194332064864e-06,
-      "loss": 0.7979,
+      "learning_rate": 3.377827315743021e-06,
+      "loss": 0.8328,
       "step": 26024
     },
     {
-      "epoch": 0.7147565296201698,
+      "epoch": 0.7385073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 3.972484537661691e-06,
-      "loss": 0.684,
+      "learning_rate": 3.3771386676000682e-06,
+      "loss": 0.8794,
       "step": 26025
     },
     {
-      "epoch": 0.7147839938480129,
+      "epoch": 0.7385357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 3.971774790950556e-06,
-      "loss": 0.8848,
+      "learning_rate": 3.3764500754007203e-06,
+      "loss": 0.8149,
       "step": 26026
     },
     {
-      "epoch": 0.7148114580758562,
+      "epoch": 0.7385641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 3.9710650919370764e-06,
-      "loss": 0.9052,
+      "learning_rate": 3.375761539150796e-06,
+      "loss": 0.8008,
       "step": 26027
     },
     {
-      "epoch": 0.7148389223036994,
+      "epoch": 0.7385925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 3.970355440626863e-06,
-      "loss": 0.8945,
+      "learning_rate": 3.375073058856112e-06,
+      "loss": 0.8178,
       "step": 26028
     },
     {
-      "epoch": 0.7148663865315427,
+      "epoch": 0.7386208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 3.969645837025535e-06,
-      "loss": 0.8368,
+      "learning_rate": 3.374384634522484e-06,
+      "loss": 0.7751,
       "step": 26029
     },
     {
-      "epoch": 0.7148938507593859,
+      "epoch": 0.7386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 3.968936281138703e-06,
-      "loss": 0.8132,
+      "learning_rate": 3.3736962661557293e-06,
+      "loss": 0.8844,
       "step": 26030
     },
     {
-      "epoch": 0.7149213149872291,
+      "epoch": 0.738677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 3.9682267729719835e-06,
-      "loss": 0.7733,
+      "learning_rate": 3.3730079537616576e-06,
+      "loss": 0.8016,
       "step": 26031
     },
     {
-      "epoch": 0.7149487792150724,
+      "epoch": 0.7387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 3.9675173125309895e-06,
-      "loss": 0.8951,
+      "learning_rate": 3.3723196973460846e-06,
+      "loss": 0.8365,
       "step": 26032
     },
     {
-      "epoch": 0.7149762434429155,
+      "epoch": 0.7387343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 3.966807899821339e-06,
-      "loss": 0.7866,
+      "learning_rate": 3.371631496914829e-06,
+      "loss": 0.8649,
       "step": 26033
     },
     {
-      "epoch": 0.7150037076707588,
+      "epoch": 0.7387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 3.966098534848638e-06,
-      "loss": 0.7829,
+      "learning_rate": 3.3709433524736967e-06,
+      "loss": 0.8274,
       "step": 26034
     },
     {
-      "epoch": 0.7150311718986021,
+      "epoch": 0.7387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 3.965389217618501e-06,
-      "loss": 0.7813,
+      "learning_rate": 3.370255264028505e-06,
+      "loss": 0.778,
       "step": 26035
     },
     {
-      "epoch": 0.7150586361264453,
+      "epoch": 0.7388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 3.9646799481365485e-06,
-      "loss": 0.9133,
+      "learning_rate": 3.369567231585067e-06,
+      "loss": 0.8275,
       "step": 26036
     },
     {
-      "epoch": 0.7150861003542885,
+      "epoch": 0.7388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 3.963970726408381e-06,
-      "loss": 0.8503,
+      "learning_rate": 3.36887925514919e-06,
+      "loss": 0.8128,
       "step": 26037
     },
     {
-      "epoch": 0.7151135645821318,
+      "epoch": 0.7388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 3.963261552439615e-06,
-      "loss": 0.8468,
+      "learning_rate": 3.368191334726687e-06,
+      "loss": 0.8908,
       "step": 26038
     },
     {
-      "epoch": 0.715141028809975,
+      "epoch": 0.7389046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 3.9625524262358665e-06,
-      "loss": 0.826,
+      "learning_rate": 3.3675034703233743e-06,
+      "loss": 0.7917,
       "step": 26039
     },
     {
-      "epoch": 0.7151684930378183,
+      "epoch": 0.738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 3.961843347802739e-06,
-      "loss": 0.8041,
+      "learning_rate": 3.366815661945054e-06,
+      "loss": 0.7289,
       "step": 26040
     },
     {
-      "epoch": 0.7151959572656614,
+      "epoch": 0.738961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 3.961134317145844e-06,
-      "loss": 0.7836,
+      "learning_rate": 3.366127909597541e-06,
+      "loss": 0.8628,
       "step": 26041
     },
     {
-      "epoch": 0.7152234214935047,
+      "epoch": 0.7389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 3.960425334270795e-06,
-      "loss": 0.945,
+      "learning_rate": 3.3654402132866458e-06,
+      "loss": 0.7908,
       "step": 26042
     },
     {
-      "epoch": 0.715250885721348,
+      "epoch": 0.7390181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 3.959716399183199e-06,
-      "loss": 0.7524,
+      "learning_rate": 3.3647525730181695e-06,
+      "loss": 0.7751,
       "step": 26043
     },
     {
-      "epoch": 0.7152783499491912,
+      "epoch": 0.7390465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 3.959007511888671e-06,
-      "loss": 0.7953,
+      "learning_rate": 3.3640649887979327e-06,
+      "loss": 0.8969,
       "step": 26044
     },
     {
-      "epoch": 0.7153058141770344,
+      "epoch": 0.7390749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 3.958298672392815e-06,
-      "loss": 0.7778,
+      "learning_rate": 3.3633774606317348e-06,
+      "loss": 0.7889,
       "step": 26045
     },
     {
-      "epoch": 0.7153332784048776,
+      "epoch": 0.7391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 3.957589880701237e-06,
-      "loss": 0.8468,
+      "learning_rate": 3.362689988525385e-06,
+      "loss": 0.7429,
       "step": 26046
     },
     {
-      "epoch": 0.7153607426327209,
+      "epoch": 0.7391316685584562,
       "grad_norm": 0.0,
-      "learning_rate": 3.956881136819548e-06,
-      "loss": 0.7975,
+      "learning_rate": 3.362002572484695e-06,
+      "loss": 0.901,
       "step": 26047
     },
     {
-      "epoch": 0.7153882068605641,
+      "epoch": 0.7391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 3.956172440753357e-06,
-      "loss": 0.841,
+      "learning_rate": 3.3613152125154636e-06,
+      "loss": 0.8985,
       "step": 26048
     },
     {
-      "epoch": 0.7154156710884073,
+      "epoch": 0.7391884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 3.955463792508271e-06,
-      "loss": 0.8528,
+      "learning_rate": 3.3606279086235015e-06,
+      "loss": 0.7848,
       "step": 26049
     },
     {
-      "epoch": 0.7154431353162506,
+      "epoch": 0.7392167990919409,
       "grad_norm": 0.0,
-      "learning_rate": 3.9547551920899e-06,
-      "loss": 0.8171,
+      "learning_rate": 3.3599406608146178e-06,
+      "loss": 0.7113,
       "step": 26050
     },
     {
-      "epoch": 0.7154705995440939,
+      "epoch": 0.7392451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 3.954046639503843e-06,
-      "loss": 0.8815,
+      "learning_rate": 3.35925346909461e-06,
+      "loss": 0.7578,
       "step": 26051
     },
     {
-      "epoch": 0.715498063771937,
+      "epoch": 0.7392735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 3.953338134755713e-06,
-      "loss": 0.8055,
+      "learning_rate": 3.3585663334692864e-06,
+      "loss": 0.7825,
       "step": 26052
     },
     {
-      "epoch": 0.7155255279997803,
+      "epoch": 0.7393019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 3.9526296778511165e-06,
-      "loss": 0.8764,
+      "learning_rate": 3.357879253944455e-06,
+      "loss": 0.8974,
       "step": 26053
     },
     {
-      "epoch": 0.7155529922276235,
+      "epoch": 0.7393303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 3.9519212687956535e-06,
-      "loss": 0.8119,
+      "learning_rate": 3.3571922305259132e-06,
+      "loss": 0.8923,
       "step": 26054
     },
     {
-      "epoch": 0.7155804564554668,
+      "epoch": 0.7393586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 3.951212907594931e-06,
-      "loss": 0.8412,
+      "learning_rate": 3.356505263219466e-06,
+      "loss": 0.8714,
       "step": 26055
     },
     {
-      "epoch": 0.71560792068331,
+      "epoch": 0.7393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 3.95050459425456e-06,
-      "loss": 0.9086,
+      "learning_rate": 3.355818352030923e-06,
+      "loss": 0.8006,
       "step": 26056
     },
     {
-      "epoch": 0.7156353849111532,
+      "epoch": 0.7394154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 3.949796328780135e-06,
-      "loss": 1.0137,
+      "learning_rate": 3.355131496966075e-06,
+      "loss": 0.748,
       "step": 26057
     },
     {
-      "epoch": 0.7156628491389965,
+      "epoch": 0.7394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 3.949088111177267e-06,
-      "loss": 0.7802,
+      "learning_rate": 3.3544446980307322e-06,
+      "loss": 0.8314,
       "step": 26058
     },
     {
-      "epoch": 0.7156903133668396,
+      "epoch": 0.739472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.948379941451557e-06,
-      "loss": 0.889,
+      "learning_rate": 3.3537579552306933e-06,
+      "loss": 0.7448,
       "step": 26059
     },
     {
-      "epoch": 0.7157177775946829,
+      "epoch": 0.7395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 3.947671819608612e-06,
-      "loss": 0.8578,
+      "learning_rate": 3.353071268571759e-06,
+      "loss": 0.8351,
       "step": 26060
     },
     {
-      "epoch": 0.7157452418225262,
+      "epoch": 0.7395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 3.946963745654029e-06,
-      "loss": 0.8143,
+      "learning_rate": 3.3523846380597347e-06,
+      "loss": 0.8818,
       "step": 26061
     },
     {
-      "epoch": 0.7157727060503694,
+      "epoch": 0.7395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 3.946255719593416e-06,
-      "loss": 0.8314,
+      "learning_rate": 3.3516980637004127e-06,
+      "loss": 0.8707,
       "step": 26062
     },
     {
-      "epoch": 0.7158001702782126,
+      "epoch": 0.7395856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 3.9455477414323705e-06,
-      "loss": 0.7782,
+      "learning_rate": 3.351011545499595e-06,
+      "loss": 0.8808,
       "step": 26063
     },
     {
-      "epoch": 0.7158276345060559,
+      "epoch": 0.7396140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 3.944839811176497e-06,
-      "loss": 0.7875,
+      "learning_rate": 3.3503250834630864e-06,
+      "loss": 0.7667,
       "step": 26064
     },
     {
-      "epoch": 0.7158550987338991,
+      "epoch": 0.7396424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 3.944131928831395e-06,
-      "loss": 0.8788,
+      "learning_rate": 3.349638677596676e-06,
+      "loss": 0.8242,
       "step": 26065
     },
     {
-      "epoch": 0.7158825629617424,
+      "epoch": 0.7396708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 3.943424094402667e-06,
-      "loss": 0.8123,
+      "learning_rate": 3.3489523279061677e-06,
+      "loss": 0.827,
       "step": 26066
     },
     {
-      "epoch": 0.7159100271895855,
+      "epoch": 0.7396992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 3.942716307895914e-06,
-      "loss": 0.9254,
+      "learning_rate": 3.348266034397362e-06,
+      "loss": 0.8744,
       "step": 26067
     },
     {
-      "epoch": 0.7159374914174288,
+      "epoch": 0.7397275822928491,
       "grad_norm": 0.0,
-      "learning_rate": 3.94200856931674e-06,
-      "loss": 0.891,
+      "learning_rate": 3.3475797970760473e-06,
+      "loss": 0.7775,
       "step": 26068
     },
     {
-      "epoch": 0.7159649556452721,
+      "epoch": 0.7397559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 3.941300878670741e-06,
-      "loss": 0.8526,
+      "learning_rate": 3.346893615948026e-06,
+      "loss": 0.9059,
       "step": 26069
     },
     {
-      "epoch": 0.7159924198731152,
+      "epoch": 0.7397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 3.940593235963512e-06,
-      "loss": 0.8037,
+      "learning_rate": 3.3462074910190955e-06,
+      "loss": 0.8536,
       "step": 26070
     },
     {
-      "epoch": 0.7160198841009585,
+      "epoch": 0.7398127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 3.939885641200657e-06,
-      "loss": 0.9012,
+      "learning_rate": 3.345521422295047e-06,
+      "loss": 0.9263,
       "step": 26071
     },
     {
-      "epoch": 0.7160473483288017,
+      "epoch": 0.7398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 3.939178094387774e-06,
-      "loss": 0.7658,
+      "learning_rate": 3.3448354097816794e-06,
+      "loss": 0.8663,
       "step": 26072
     },
     {
-      "epoch": 0.716074812556645,
+      "epoch": 0.7398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 3.938470595530463e-06,
-      "loss": 0.7784,
+      "learning_rate": 3.3441494534847875e-06,
+      "loss": 0.8622,
       "step": 26073
     },
     {
-      "epoch": 0.7161022767844882,
+      "epoch": 0.7398978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 3.937763144634323e-06,
-      "loss": 0.8178,
+      "learning_rate": 3.343463553410159e-06,
+      "loss": 0.8397,
       "step": 26074
     },
     {
-      "epoch": 0.7161297410123314,
+      "epoch": 0.7399262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 3.937055741704945e-06,
-      "loss": 0.8058,
+      "learning_rate": 3.3427777095635996e-06,
+      "loss": 0.8065,
       "step": 26075
     },
     {
-      "epoch": 0.7161572052401747,
+      "epoch": 0.7399545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 3.936348386747932e-06,
-      "loss": 0.8388,
+      "learning_rate": 3.342091921950892e-06,
+      "loss": 0.7934,
       "step": 26076
     },
     {
-      "epoch": 0.716184669468018,
+      "epoch": 0.739982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 3.935641079768881e-06,
-      "loss": 0.8347,
+      "learning_rate": 3.3414061905778337e-06,
+      "loss": 0.8736,
       "step": 26077
     },
     {
-      "epoch": 0.7162121336958611,
+      "epoch": 0.740011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 3.934933820773385e-06,
-      "loss": 0.9095,
+      "learning_rate": 3.340720515450221e-06,
+      "loss": 0.7987,
       "step": 26078
     },
     {
-      "epoch": 0.7162395979237044,
+      "epoch": 0.7400397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 3.9342266097670415e-06,
-      "loss": 0.8682,
+      "learning_rate": 3.340034896573837e-06,
+      "loss": 0.812,
       "step": 26079
     },
     {
-      "epoch": 0.7162670621515476,
+      "epoch": 0.7400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 3.93351944675545e-06,
-      "loss": 0.86,
+      "learning_rate": 3.339349333954478e-06,
+      "loss": 0.8242,
       "step": 26080
     },
     {
-      "epoch": 0.7162945263793908,
+      "epoch": 0.7400964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 3.9328123317442e-06,
-      "loss": 0.871,
+      "learning_rate": 3.3386638275979376e-06,
+      "loss": 0.7756,
       "step": 26081
     },
     {
-      "epoch": 0.7163219906072341,
+      "epoch": 0.7401248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 3.932105264738888e-06,
-      "loss": 0.8782,
+      "learning_rate": 3.3379783775100005e-06,
+      "loss": 0.8389,
       "step": 26082
     },
     {
-      "epoch": 0.7163494548350773,
+      "epoch": 0.7401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 3.931398245745111e-06,
-      "loss": 0.8523,
+      "learning_rate": 3.3372929836964583e-06,
+      "loss": 0.8474,
       "step": 26083
     },
     {
-      "epoch": 0.7163769190629206,
+      "epoch": 0.7401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 3.930691274768464e-06,
-      "loss": 0.9652,
+      "learning_rate": 3.336607646163106e-06,
+      "loss": 0.7637,
       "step": 26084
     },
     {
-      "epoch": 0.7164043832907638,
+      "epoch": 0.7402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 3.929984351814535e-06,
-      "loss": 0.826,
+      "learning_rate": 3.3359223649157255e-06,
+      "loss": 0.7868,
       "step": 26085
     },
     {
-      "epoch": 0.716431847518607,
+      "epoch": 0.7402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 3.929277476888925e-06,
-      "loss": 0.8235,
+      "learning_rate": 3.335237139960108e-06,
+      "loss": 0.8811,
       "step": 26086
     },
     {
-      "epoch": 0.7164593117464503,
+      "epoch": 0.740266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 3.92857064999722e-06,
-      "loss": 0.9266,
+      "learning_rate": 3.3345519713020445e-06,
+      "loss": 0.8896,
       "step": 26087
     },
     {
-      "epoch": 0.7164867759742934,
+      "epoch": 0.7402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 3.9278638711450155e-06,
-      "loss": 0.912,
+      "learning_rate": 3.3338668589473176e-06,
+      "loss": 0.8096,
       "step": 26088
     },
     {
-      "epoch": 0.7165142402021367,
+      "epoch": 0.7403234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 3.9271571403379045e-06,
-      "loss": 0.811,
+      "learning_rate": 3.333181802901717e-06,
+      "loss": 0.8246,
       "step": 26089
     },
     {
-      "epoch": 0.71654170442998,
+      "epoch": 0.7403518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 3.9264504575814786e-06,
-      "loss": 0.8848,
+      "learning_rate": 3.33249680317103e-06,
+      "loss": 0.9209,
       "step": 26090
     },
     {
-      "epoch": 0.7165691686578232,
+      "epoch": 0.7403802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 3.92574382288133e-06,
-      "loss": 0.8238,
+      "learning_rate": 3.3318118597610417e-06,
+      "loss": 0.7435,
       "step": 26091
     },
     {
-      "epoch": 0.7165966328856664,
+      "epoch": 0.7404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 3.9250372362430515e-06,
-      "loss": 0.9362,
+      "learning_rate": 3.331126972677542e-06,
+      "loss": 0.8364,
       "step": 26092
     },
     {
-      "epoch": 0.7166240971135096,
+      "epoch": 0.7404370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 3.924330697672231e-06,
-      "loss": 0.9037,
+      "learning_rate": 3.330442141926308e-06,
+      "loss": 0.7618,
       "step": 26093
     },
     {
-      "epoch": 0.7166515613413529,
+      "epoch": 0.7404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 3.923624207174458e-06,
-      "loss": 0.7672,
+      "learning_rate": 3.32975736751313e-06,
+      "loss": 0.8018,
       "step": 26094
     },
     {
-      "epoch": 0.7166790255691962,
+      "epoch": 0.7404937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 3.922917764755323e-06,
-      "loss": 0.9175,
+      "learning_rate": 3.329072649443795e-06,
+      "loss": 0.7968,
       "step": 26095
     },
     {
-      "epoch": 0.7167064897970393,
+      "epoch": 0.740522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 3.922211370420417e-06,
-      "loss": 0.8397,
+      "learning_rate": 3.328387987724079e-06,
+      "loss": 0.8088,
       "step": 26096
     },
     {
-      "epoch": 0.7167339540248826,
+      "epoch": 0.7405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 3.921505024175328e-06,
-      "loss": 0.7881,
+      "learning_rate": 3.3277033823597706e-06,
+      "loss": 0.8027,
       "step": 26097
     },
     {
-      "epoch": 0.7167614182527259,
+      "epoch": 0.7405788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 3.9207987260256495e-06,
-      "loss": 0.7647,
+      "learning_rate": 3.3270188333566543e-06,
+      "loss": 0.7285,
       "step": 26098
     },
     {
-      "epoch": 0.716788882480569,
+      "epoch": 0.7406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 3.920092475976964e-06,
-      "loss": 0.8029,
+      "learning_rate": 3.326334340720506e-06,
+      "loss": 0.8925,
       "step": 26099
     },
     {
-      "epoch": 0.7168163467084123,
+      "epoch": 0.7406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 3.919386274034864e-06,
-      "loss": 0.7511,
+      "learning_rate": 3.325649904457112e-06,
+      "loss": 0.8077,
       "step": 26100
     },
     {
-      "epoch": 0.7168438109362555,
+      "epoch": 0.7406640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 3.918680120204933e-06,
-      "loss": 0.8495,
+      "learning_rate": 3.324965524572258e-06,
+      "loss": 0.707,
       "step": 26101
     },
     {
-      "epoch": 0.7168712751640988,
+      "epoch": 0.7406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 3.91797401449276e-06,
-      "loss": 0.8623,
+      "learning_rate": 3.3242812010717153e-06,
+      "loss": 0.9134,
       "step": 26102
     },
     {
-      "epoch": 0.716898739391942,
+      "epoch": 0.7407207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 3.917267956903933e-06,
-      "loss": 0.7428,
+      "learning_rate": 3.3235969339612693e-06,
+      "loss": 0.7334,
       "step": 26103
     },
     {
-      "epoch": 0.7169262036197852,
+      "epoch": 0.7407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 3.916561947444041e-06,
-      "loss": 0.7325,
+      "learning_rate": 3.3229127232467005e-06,
+      "loss": 0.8664,
       "step": 26104
     },
     {
-      "epoch": 0.7169536678476285,
+      "epoch": 0.7407775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 3.915855986118664e-06,
-      "loss": 0.8757,
+      "learning_rate": 3.3222285689337886e-06,
+      "loss": 0.8262,
       "step": 26105
     },
     {
-      "epoch": 0.7169811320754716,
+      "epoch": 0.740805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 3.915150072933392e-06,
-      "loss": 0.8176,
+      "learning_rate": 3.3215444710283116e-06,
+      "loss": 0.8698,
       "step": 26106
     },
     {
-      "epoch": 0.7170085963033149,
+      "epoch": 0.7408342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 3.914444207893808e-06,
-      "loss": 0.801,
+      "learning_rate": 3.3208604295360526e-06,
+      "loss": 0.8588,
       "step": 26107
     },
     {
-      "epoch": 0.7170360605311582,
+      "epoch": 0.7408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 3.9137383910055035e-06,
-      "loss": 0.8318,
+      "learning_rate": 3.3201764444627825e-06,
+      "loss": 0.8144,
       "step": 26108
     },
     {
-      "epoch": 0.7170635247590014,
+      "epoch": 0.7408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 3.913032622274059e-06,
-      "loss": 0.8382,
+      "learning_rate": 3.319492515814282e-06,
+      "loss": 0.7339,
       "step": 26109
     },
     {
-      "epoch": 0.7170909889868446,
+      "epoch": 0.7409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 3.912326901705054e-06,
-      "loss": 0.8197,
+      "learning_rate": 3.318808643596332e-06,
+      "loss": 0.8826,
       "step": 26110
     },
     {
-      "epoch": 0.7171184532146879,
+      "epoch": 0.7409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 3.911621229304077e-06,
-      "loss": 0.7238,
+      "learning_rate": 3.3181248278147017e-06,
+      "loss": 0.8309,
       "step": 26111
     },
     {
-      "epoch": 0.7171459174425311,
+      "epoch": 0.7409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 3.9109156050767106e-06,
-      "loss": 0.8057,
+      "learning_rate": 3.317441068475171e-06,
+      "loss": 0.8223,
       "step": 26112
     },
     {
-      "epoch": 0.7171733816703744,
+      "epoch": 0.7410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 3.910210029028538e-06,
-      "loss": 0.8284,
+      "learning_rate": 3.3167573655835206e-06,
+      "loss": 0.8367,
       "step": 26113
     },
     {
-      "epoch": 0.7172008458982175,
+      "epoch": 0.7410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 3.909504501165145e-06,
-      "loss": 0.8148,
+      "learning_rate": 3.3160737191455173e-06,
+      "loss": 0.8194,
       "step": 26114
     },
     {
-      "epoch": 0.7172283101260608,
+      "epoch": 0.7410612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 3.908799021492112e-06,
-      "loss": 0.8099,
+      "learning_rate": 3.315390129166943e-06,
+      "loss": 0.8732,
       "step": 26115
     },
     {
-      "epoch": 0.7172557743539041,
+      "epoch": 0.7410896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 3.908093590015019e-06,
-      "loss": 0.8238,
+      "learning_rate": 3.314706595653566e-06,
+      "loss": 0.8404,
       "step": 26116
     },
     {
-      "epoch": 0.7172832385817473,
+      "epoch": 0.7411180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 3.907388206739453e-06,
-      "loss": 0.7293,
+      "learning_rate": 3.314023118611163e-06,
+      "loss": 0.7852,
       "step": 26117
     },
     {
-      "epoch": 0.7173107028095905,
+      "epoch": 0.7411464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 3.906682871670988e-06,
-      "loss": 0.8786,
+      "learning_rate": 3.313339698045509e-06,
+      "loss": 0.8159,
       "step": 26118
     },
     {
-      "epoch": 0.7173381670374337,
+      "epoch": 0.7411748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 3.905977584815208e-06,
-      "loss": 0.8021,
+      "learning_rate": 3.312656333962373e-06,
+      "loss": 0.7818,
       "step": 26119
     },
     {
-      "epoch": 0.717365631265277,
+      "epoch": 0.7412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 3.905272346177694e-06,
-      "loss": 0.7654,
+      "learning_rate": 3.3119730263675263e-06,
+      "loss": 0.8152,
       "step": 26120
     },
     {
-      "epoch": 0.7173930954931202,
+      "epoch": 0.7412315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 3.90456715576403e-06,
-      "loss": 0.8602,
+      "learning_rate": 3.3112897752667504e-06,
+      "loss": 0.7634,
       "step": 26121
     },
     {
-      "epoch": 0.7174205597209634,
+      "epoch": 0.7412599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 3.9038620135797874e-06,
-      "loss": 0.8116,
+      "learning_rate": 3.310606580665807e-06,
+      "loss": 0.8329,
       "step": 26122
     },
     {
-      "epoch": 0.7174480239488067,
+      "epoch": 0.7412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 3.903156919630551e-06,
-      "loss": 0.7875,
+      "learning_rate": 3.309923442570472e-06,
+      "loss": 0.92,
       "step": 26123
     },
     {
-      "epoch": 0.71747548817665,
+      "epoch": 0.741316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 3.902451873921901e-06,
-      "loss": 0.7972,
+      "learning_rate": 3.3092403609865163e-06,
+      "loss": 0.8218,
       "step": 26124
     },
     {
-      "epoch": 0.7175029524044931,
+      "epoch": 0.7413450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 3.901746876459412e-06,
-      "loss": 0.8636,
+      "learning_rate": 3.3085573359197045e-06,
+      "loss": 0.9051,
       "step": 26125
     },
     {
-      "epoch": 0.7175304166323364,
+      "epoch": 0.7413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 3.901041927248662e-06,
-      "loss": 0.8277,
+      "learning_rate": 3.3078743673758107e-06,
+      "loss": 0.833,
       "step": 26126
     },
     {
-      "epoch": 0.7175578808601796,
+      "epoch": 0.7414018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 3.900337026295236e-06,
-      "loss": 0.6548,
+      "learning_rate": 3.3071914553606055e-06,
+      "loss": 0.9207,
       "step": 26127
     },
     {
-      "epoch": 0.7175853450880229,
+      "epoch": 0.7414301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 3.8996321736047015e-06,
-      "loss": 0.8385,
+      "learning_rate": 3.3065085998798516e-06,
+      "loss": 0.8045,
       "step": 26128
     },
     {
-      "epoch": 0.7176128093158661,
+      "epoch": 0.7414585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 3.89892736918264e-06,
-      "loss": 0.8718,
+      "learning_rate": 3.3058258009393207e-06,
+      "loss": 0.8103,
       "step": 26129
     },
     {
-      "epoch": 0.7176402735437093,
+      "epoch": 0.7414869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 3.89822261303463e-06,
-      "loss": 0.8473,
+      "learning_rate": 3.305143058544784e-06,
+      "loss": 0.8184,
       "step": 26130
     },
     {
-      "epoch": 0.7176677377715526,
+      "epoch": 0.7415153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 3.897517905166246e-06,
-      "loss": 0.843,
+      "learning_rate": 3.3044603727020007e-06,
+      "loss": 0.8659,
       "step": 26131
     },
     {
-      "epoch": 0.7176952019993957,
+      "epoch": 0.7415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 3.896813245583066e-06,
-      "loss": 0.7773,
+      "learning_rate": 3.3037777434167418e-06,
+      "loss": 0.8069,
       "step": 26132
     },
     {
-      "epoch": 0.717722666227239,
+      "epoch": 0.741572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 3.896108634290665e-06,
-      "loss": 0.8441,
+      "learning_rate": 3.3030951706947777e-06,
+      "loss": 0.7696,
       "step": 26133
     },
     {
-      "epoch": 0.7177501304550823,
+      "epoch": 0.7416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 3.895404071294614e-06,
-      "loss": 0.8153,
+      "learning_rate": 3.3024126545418655e-06,
+      "loss": 0.8847,
       "step": 26134
     },
     {
-      "epoch": 0.7177775946829255,
+      "epoch": 0.7416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 3.8946995566004894e-06,
-      "loss": 0.8981,
+      "learning_rate": 3.301730194963775e-06,
+      "loss": 0.7599,
       "step": 26135
     },
     {
-      "epoch": 0.7178050589107687,
+      "epoch": 0.7416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 3.893995090213868e-06,
-      "loss": 0.7921,
+      "learning_rate": 3.30104779196627e-06,
+      "loss": 0.7539,
       "step": 26136
     },
     {
-      "epoch": 0.717832523138612,
+      "epoch": 0.7416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 3.8932906721403225e-06,
-      "loss": 0.907,
+      "learning_rate": 3.300365445555116e-06,
+      "loss": 0.84,
       "step": 26137
     },
     {
-      "epoch": 0.7178599873664552,
+      "epoch": 0.7417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 3.892586302385426e-06,
-      "loss": 0.7518,
+      "learning_rate": 3.2996831557360788e-06,
+      "loss": 0.8279,
       "step": 26138
     },
     {
-      "epoch": 0.7178874515942985,
+      "epoch": 0.7417423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 3.891881980954757e-06,
-      "loss": 0.8797,
+      "learning_rate": 3.2990009225149167e-06,
+      "loss": 0.8327,
       "step": 26139
     },
     {
-      "epoch": 0.7179149158221416,
+      "epoch": 0.7417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 3.891177707853879e-06,
-      "loss": 0.8271,
+      "learning_rate": 3.298318745897394e-06,
+      "loss": 0.8341,
       "step": 26140
     },
     {
-      "epoch": 0.7179423800499849,
+      "epoch": 0.7417990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 3.890473483088374e-06,
-      "loss": 0.7905,
+      "learning_rate": 3.297636625889279e-06,
+      "loss": 0.8869,
       "step": 26141
     },
     {
-      "epoch": 0.7179698442778282,
+      "epoch": 0.741827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 3.889769306663805e-06,
-      "loss": 0.8883,
+      "learning_rate": 3.296954562496324e-06,
+      "loss": 0.9338,
       "step": 26142
     },
     {
-      "epoch": 0.7179973085056713,
+      "epoch": 0.741855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 3.889065178585749e-06,
-      "loss": 0.803,
+      "learning_rate": 3.296272555724296e-06,
+      "loss": 0.7819,
       "step": 26143
     },
     {
-      "epoch": 0.7180247727335146,
+      "epoch": 0.7418842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 3.888361098859777e-06,
-      "loss": 0.772,
+      "learning_rate": 3.295590605578959e-06,
+      "loss": 0.7501,
       "step": 26144
     },
     {
-      "epoch": 0.7180522369613578,
+      "epoch": 0.7419125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 3.8876570674914615e-06,
-      "loss": 0.8308,
+      "learning_rate": 3.294908712066065e-06,
+      "loss": 0.8063,
       "step": 26145
     },
     {
-      "epoch": 0.7180797011892011,
+      "epoch": 0.7419409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 3.886953084486368e-06,
-      "loss": 0.8944,
+      "learning_rate": 3.2942268751913785e-06,
+      "loss": 0.8285,
       "step": 26146
     },
     {
-      "epoch": 0.7181071654170443,
+      "epoch": 0.7419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 3.886249149850068e-06,
-      "loss": 0.8286,
+      "learning_rate": 3.293545094960663e-06,
+      "loss": 0.8368,
       "step": 26147
     },
     {
-      "epoch": 0.7181346296448875,
+      "epoch": 0.7419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 3.885545263588139e-06,
-      "loss": 0.7499,
+      "learning_rate": 3.29286337137967e-06,
+      "loss": 0.7989,
       "step": 26148
     },
     {
-      "epoch": 0.7181620938727308,
+      "epoch": 0.7420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.8848414257061395e-06,
-      "loss": 0.8389,
+      "learning_rate": 3.2921817044541614e-06,
+      "loss": 0.8143,
       "step": 26149
     },
     {
-      "epoch": 0.718189558100574,
+      "epoch": 0.7420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 3.8841376362096436e-06,
-      "loss": 0.875,
+      "learning_rate": 3.2915000941898958e-06,
+      "loss": 0.7589,
       "step": 26150
     },
     {
-      "epoch": 0.7182170223284172,
+      "epoch": 0.7420828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 3.883433895104224e-06,
-      "loss": 0.7092,
+      "learning_rate": 3.29081854059263e-06,
+      "loss": 0.8161,
       "step": 26151
     },
     {
-      "epoch": 0.7182444865562605,
+      "epoch": 0.7421112372304199,
       "grad_norm": 0.0,
-      "learning_rate": 3.882730202395441e-06,
-      "loss": 0.8083,
+      "learning_rate": 3.290137043668126e-06,
+      "loss": 0.7545,
       "step": 26152
     },
     {
-      "epoch": 0.7182719507841037,
+      "epoch": 0.7421396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 3.882026558088866e-06,
-      "loss": 0.7737,
+      "learning_rate": 3.2894556034221316e-06,
+      "loss": 0.8286,
       "step": 26153
     },
     {
-      "epoch": 0.7182994150119469,
+      "epoch": 0.7421679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 3.881322962190067e-06,
-      "loss": 0.7815,
+      "learning_rate": 3.288774219860408e-06,
+      "loss": 0.9023,
       "step": 26154
     },
     {
-      "epoch": 0.7183268792397902,
+      "epoch": 0.7421963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 3.88061941470461e-06,
-      "loss": 0.8248,
+      "learning_rate": 3.2880928929887123e-06,
+      "loss": 0.8088,
       "step": 26155
     },
     {
-      "epoch": 0.7183543434676334,
+      "epoch": 0.7422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 3.8799159156380675e-06,
-      "loss": 0.8113,
+      "learning_rate": 3.287411622812796e-06,
+      "loss": 0.8735,
       "step": 26156
     },
     {
-      "epoch": 0.7183818076954767,
+      "epoch": 0.7422531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 3.879212464995999e-06,
-      "loss": 0.841,
+      "learning_rate": 3.2867304093384145e-06,
+      "loss": 0.8817,
       "step": 26157
     },
     {
-      "epoch": 0.7184092719233198,
+      "epoch": 0.7422814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 3.878509062783972e-06,
-      "loss": 0.833,
+      "learning_rate": 3.2860492525713274e-06,
+      "loss": 0.8748,
       "step": 26158
     },
     {
-      "epoch": 0.7184367361511631,
+      "epoch": 0.7423098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 3.877805709007549e-06,
-      "loss": 0.8283,
+      "learning_rate": 3.2853681525172796e-06,
+      "loss": 0.847,
       "step": 26159
     },
     {
-      "epoch": 0.7184642003790064,
+      "epoch": 0.7423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 3.877102403672301e-06,
-      "loss": 0.9351,
+      "learning_rate": 3.284687109182029e-06,
+      "loss": 0.8197,
       "step": 26160
     },
     {
-      "epoch": 0.7184916646068495,
+      "epoch": 0.742366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.87639914678379e-06,
-      "loss": 0.8012,
+      "learning_rate": 3.2840061225713316e-06,
+      "loss": 0.8588,
       "step": 26161
     },
     {
-      "epoch": 0.7185191288346928,
+      "epoch": 0.7423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 3.875695938347581e-06,
-      "loss": 0.8108,
+      "learning_rate": 3.2833251926909335e-06,
+      "loss": 0.9187,
       "step": 26162
     },
     {
-      "epoch": 0.7185465930625361,
+      "epoch": 0.7424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 3.874992778369242e-06,
-      "loss": 0.9495,
+      "learning_rate": 3.2826443195465896e-06,
+      "loss": 0.8553,
       "step": 26163
     },
     {
-      "epoch": 0.7185740572903793,
+      "epoch": 0.7424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 3.874289666854328e-06,
-      "loss": 0.8074,
+      "learning_rate": 3.281963503144053e-06,
+      "loss": 0.8456,
       "step": 26164
     },
     {
-      "epoch": 0.7186015215182225,
+      "epoch": 0.7424801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 3.873586603808413e-06,
-      "loss": 0.7504,
+      "learning_rate": 3.2812827434890713e-06,
+      "loss": 0.8052,
       "step": 26165
     },
     {
-      "epoch": 0.7186289857460657,
+      "epoch": 0.7425085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 3.8728835892370476e-06,
-      "loss": 0.8466,
+      "learning_rate": 3.280602040587395e-06,
+      "loss": 0.8178,
       "step": 26166
     },
     {
-      "epoch": 0.718656449973909,
+      "epoch": 0.7425368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 3.872180623145803e-06,
-      "loss": 0.8311,
+      "learning_rate": 3.279921394444776e-06,
+      "loss": 0.8366,
       "step": 26167
     },
     {
-      "epoch": 0.7186839142017523,
+      "epoch": 0.7425652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 3.871477705540241e-06,
-      "loss": 0.8541,
+      "learning_rate": 3.2792408050669634e-06,
+      "loss": 0.7316,
       "step": 26168
     },
     {
-      "epoch": 0.7187113784295954,
+      "epoch": 0.7425936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 3.870774836425918e-06,
-      "loss": 0.9257,
+      "learning_rate": 3.2785602724597078e-06,
+      "loss": 0.7886,
       "step": 26169
     },
     {
-      "epoch": 0.7187388426574387,
+      "epoch": 0.7426220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 3.870072015808398e-06,
-      "loss": 0.8793,
+      "learning_rate": 3.2778797966287534e-06,
+      "loss": 0.8406,
       "step": 26170
     },
     {
-      "epoch": 0.7187663068852819,
+      "epoch": 0.7426503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 3.869369243693244e-06,
-      "loss": 0.9066,
+      "learning_rate": 3.2771993775798507e-06,
+      "loss": 0.7409,
       "step": 26171
     },
     {
-      "epoch": 0.7187937711131251,
+      "epoch": 0.7426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 3.868666520086018e-06,
-      "loss": 0.9072,
+      "learning_rate": 3.27651901531875e-06,
+      "loss": 0.7926,
       "step": 26172
     },
     {
-      "epoch": 0.7188212353409684,
+      "epoch": 0.7427071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 3.867963844992275e-06,
-      "loss": 0.8874,
+      "learning_rate": 3.275838709851191e-06,
+      "loss": 0.8049,
       "step": 26173
     },
     {
-      "epoch": 0.7188486995688116,
+      "epoch": 0.7427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 3.86726121841758e-06,
-      "loss": 0.9575,
+      "learning_rate": 3.275158461182927e-06,
+      "loss": 0.7798,
       "step": 26174
     },
     {
-      "epoch": 0.7188761637966549,
+      "epoch": 0.7427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 3.866558640367487e-06,
-      "loss": 0.8216,
+      "learning_rate": 3.274478269319703e-06,
+      "loss": 0.7572,
       "step": 26175
     },
     {
-      "epoch": 0.7189036280244981,
+      "epoch": 0.7427922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 3.865856110847558e-06,
-      "loss": 0.8218,
+      "learning_rate": 3.273798134267261e-06,
+      "loss": 0.8415,
       "step": 26176
     },
     {
-      "epoch": 0.7189310922523413,
+      "epoch": 0.7428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 3.865153629863351e-06,
-      "loss": 0.8937,
+      "learning_rate": 3.273118056031349e-06,
+      "loss": 0.7889,
       "step": 26177
     },
     {
-      "epoch": 0.7189585564801846,
+      "epoch": 0.7428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 3.8644511974204256e-06,
-      "loss": 0.88,
+      "learning_rate": 3.2724380346177143e-06,
+      "loss": 0.8445,
       "step": 26178
     },
     {
-      "epoch": 0.7189860207080278,
+      "epoch": 0.742877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 3.863748813524339e-06,
-      "loss": 0.9421,
+      "learning_rate": 3.271758070032094e-06,
+      "loss": 0.7976,
       "step": 26179
     },
     {
-      "epoch": 0.719013484935871,
+      "epoch": 0.742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 3.863046478180651e-06,
-      "loss": 0.8534,
+      "learning_rate": 3.271078162280236e-06,
+      "loss": 0.7642,
       "step": 26180
     },
     {
-      "epoch": 0.7190409491637143,
+      "epoch": 0.7429341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 3.862344191394917e-06,
-      "loss": 0.9373,
+      "learning_rate": 3.2703983113678837e-06,
+      "loss": 0.9387,
       "step": 26181
     },
     {
-      "epoch": 0.7190684133915575,
+      "epoch": 0.7429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 3.86164195317269e-06,
-      "loss": 0.8842,
+      "learning_rate": 3.2697185173007784e-06,
+      "loss": 0.8578,
       "step": 26182
     },
     {
-      "epoch": 0.7190958776194007,
+      "epoch": 0.7429909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 3.86093976351953e-06,
-      "loss": 0.8728,
+      "learning_rate": 3.2690387800846677e-06,
+      "loss": 0.8102,
       "step": 26183
     },
     {
-      "epoch": 0.7191233418472439,
+      "epoch": 0.7430192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 3.860237622440992e-06,
-      "loss": 0.8714,
+      "learning_rate": 3.2683590997252845e-06,
+      "loss": 0.8773,
       "step": 26184
     },
     {
-      "epoch": 0.7191508060750872,
+      "epoch": 0.7430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 3.859535529942633e-06,
-      "loss": 0.905,
+      "learning_rate": 3.267679476228376e-06,
+      "loss": 0.7878,
       "step": 26185
     },
     {
-      "epoch": 0.7191782703029305,
+      "epoch": 0.7430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 3.85883348603001e-06,
-      "loss": 0.8323,
+      "learning_rate": 3.266999909599684e-06,
+      "loss": 0.8699,
       "step": 26186
     },
     {
-      "epoch": 0.7192057345307736,
+      "epoch": 0.7431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 3.858131490708673e-06,
-      "loss": 0.844,
+      "learning_rate": 3.2663203998449433e-06,
+      "loss": 0.8505,
       "step": 26187
     },
     {
-      "epoch": 0.7192331987586169,
+      "epoch": 0.7431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 3.857429543984178e-06,
-      "loss": 0.8515,
+      "learning_rate": 3.265640946969897e-06,
+      "loss": 0.8742,
       "step": 26188
     },
     {
-      "epoch": 0.7192606629864602,
+      "epoch": 0.743161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.856727645862084e-06,
-      "loss": 0.8784,
+      "learning_rate": 3.264961550980288e-06,
+      "loss": 0.8658,
       "step": 26189
     },
     {
-      "epoch": 0.7192881272143034,
+      "epoch": 0.7431895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 3.8560257963479366e-06,
-      "loss": 0.8507,
+      "learning_rate": 3.2642822118818475e-06,
+      "loss": 0.8311,
       "step": 26190
     },
     {
-      "epoch": 0.7193155914421466,
+      "epoch": 0.7432179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 3.855323995447292e-06,
-      "loss": 0.7556,
+      "learning_rate": 3.2636029296803186e-06,
+      "loss": 0.8854,
       "step": 26191
     },
     {
-      "epoch": 0.7193430556699898,
+      "epoch": 0.7432463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 3.854622243165709e-06,
-      "loss": 0.7606,
+      "learning_rate": 3.262923704381441e-06,
+      "loss": 0.7743,
       "step": 26192
     },
     {
-      "epoch": 0.7193705198978331,
+      "epoch": 0.7432746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 3.853920539508731e-06,
-      "loss": 0.8981,
+      "learning_rate": 3.262244535990947e-06,
+      "loss": 0.8957,
       "step": 26193
     },
     {
-      "epoch": 0.7193979841256763,
+      "epoch": 0.7433030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 3.853218884481915e-06,
-      "loss": 0.9037,
+      "learning_rate": 3.2615654245145765e-06,
+      "loss": 0.786,
       "step": 26194
     },
     {
-      "epoch": 0.7194254483535195,
+      "epoch": 0.7433314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 3.852517278090812e-06,
-      "loss": 0.8596,
+      "learning_rate": 3.2608863699580684e-06,
+      "loss": 0.9054,
       "step": 26195
     },
     {
-      "epoch": 0.7194529125813628,
+      "epoch": 0.7433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 3.851815720340978e-06,
-      "loss": 0.8162,
+      "learning_rate": 3.2602073723271497e-06,
+      "loss": 0.8652,
       "step": 26196
     },
     {
-      "epoch": 0.719480376809206,
+      "epoch": 0.7433881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 3.8511142112379555e-06,
-      "loss": 0.8034,
+      "learning_rate": 3.25952843162757e-06,
+      "loss": 0.9111,
       "step": 26197
     },
     {
-      "epoch": 0.7195078410370492,
+      "epoch": 0.743416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 3.850412750787302e-06,
-      "loss": 0.8664,
+      "learning_rate": 3.2588495478650518e-06,
+      "loss": 0.781,
       "step": 26198
     },
     {
-      "epoch": 0.7195353052648925,
+      "epoch": 0.7434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 3.8497113389945615e-06,
-      "loss": 0.7483,
+      "learning_rate": 3.258170721045335e-06,
+      "loss": 0.8086,
       "step": 26199
     },
     {
-      "epoch": 0.7195627694927357,
+      "epoch": 0.7434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 3.8490099758652876e-06,
-      "loss": 0.7856,
+      "learning_rate": 3.2574919511741575e-06,
+      "loss": 0.8557,
       "step": 26200
     },
     {
-      "epoch": 0.719590233720579,
+      "epoch": 0.7435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 3.84830866140503e-06,
-      "loss": 0.8849,
+      "learning_rate": 3.2568132382572437e-06,
+      "loss": 0.8481,
       "step": 26201
     },
     {
-      "epoch": 0.7196176979484222,
+      "epoch": 0.7435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 3.847607395619337e-06,
-      "loss": 0.7628,
+      "learning_rate": 3.2561345823003332e-06,
+      "loss": 0.8501,
       "step": 26202
     },
     {
-      "epoch": 0.7196451621762654,
+      "epoch": 0.7435584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 3.846906178513757e-06,
-      "loss": 0.8027,
+      "learning_rate": 3.255455983309159e-06,
+      "loss": 0.9121,
       "step": 26203
     },
     {
-      "epoch": 0.7196726264041087,
+      "epoch": 0.7435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 3.846205010093843e-06,
-      "loss": 0.8124,
+      "learning_rate": 3.2547774412894485e-06,
+      "loss": 0.8318,
       "step": 26204
     },
     {
-      "epoch": 0.7197000906319518,
+      "epoch": 0.7436152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 3.845503890365138e-06,
-      "loss": 0.8744,
+      "learning_rate": 3.254098956246936e-06,
+      "loss": 0.8462,
       "step": 26205
     },
     {
-      "epoch": 0.7197275548597951,
+      "epoch": 0.7436435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 3.844802819333187e-06,
-      "loss": 0.9647,
+      "learning_rate": 3.2534205281873565e-06,
+      "loss": 0.8324,
       "step": 26206
     },
     {
-      "epoch": 0.7197550190876384,
+      "epoch": 0.743671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 3.844101797003541e-06,
-      "loss": 0.8251,
+      "learning_rate": 3.252742157116434e-06,
+      "loss": 0.8799,
       "step": 26207
     },
     {
-      "epoch": 0.7197824833154816,
+      "epoch": 0.743700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 3.8434008233817455e-06,
-      "loss": 0.8382,
+      "learning_rate": 3.252063843039901e-06,
+      "loss": 0.9026,
       "step": 26208
     },
     {
-      "epoch": 0.7198099475433248,
+      "epoch": 0.7437287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 3.842699898473347e-06,
-      "loss": 0.7128,
+      "learning_rate": 3.2513855859634924e-06,
+      "loss": 0.8844,
       "step": 26209
     },
     {
-      "epoch": 0.719837411771168,
+      "epoch": 0.7437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 3.8419990222838965e-06,
-      "loss": 0.8545,
+      "learning_rate": 3.2507073858929282e-06,
+      "loss": 0.8097,
       "step": 26210
     },
     {
-      "epoch": 0.7198648759990113,
+      "epoch": 0.7437854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 3.841298194818931e-06,
-      "loss": 0.8176,
+      "learning_rate": 3.250029242833943e-06,
+      "loss": 0.8356,
       "step": 26211
     },
     {
-      "epoch": 0.7198923402268546,
+      "epoch": 0.7438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 3.8405974160839995e-06,
-      "loss": 0.8525,
+      "learning_rate": 3.2493511567922655e-06,
+      "loss": 0.8601,
       "step": 26212
     },
     {
-      "epoch": 0.7199198044546977,
+      "epoch": 0.7438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 3.839896686084652e-06,
-      "loss": 0.8455,
+      "learning_rate": 3.2486731277736205e-06,
+      "loss": 0.8716,
       "step": 26213
     },
     {
-      "epoch": 0.719947268682541,
+      "epoch": 0.7438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 3.8391960048264246e-06,
-      "loss": 0.8656,
+      "learning_rate": 3.2479951557837375e-06,
+      "loss": 0.8115,
       "step": 26214
     },
     {
-      "epoch": 0.7199747329103843,
+      "epoch": 0.7438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 3.838495372314864e-06,
-      "loss": 0.8886,
+      "learning_rate": 3.247317240828347e-06,
+      "loss": 0.8974,
       "step": 26215
     },
     {
-      "epoch": 0.7200021971382274,
+      "epoch": 0.7439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 3.837794788555518e-06,
-      "loss": 0.8102,
+      "learning_rate": 3.246639382913167e-06,
+      "loss": 0.8384,
       "step": 26216
     },
     {
-      "epoch": 0.7200296613660707,
+      "epoch": 0.7439557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 3.837094253553923e-06,
-      "loss": 0.7785,
+      "learning_rate": 3.2459615820439285e-06,
+      "loss": 0.8536,
       "step": 26217
     },
     {
-      "epoch": 0.7200571255939139,
+      "epoch": 0.7439841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 3.836393767315626e-06,
-      "loss": 0.7657,
+      "learning_rate": 3.2452838382263587e-06,
+      "loss": 0.7542,
       "step": 26218
     },
     {
-      "epoch": 0.7200845898217572,
+      "epoch": 0.7440124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 3.835693329846169e-06,
-      "loss": 0.8362,
+      "learning_rate": 3.2446061514661775e-06,
+      "loss": 0.8408,
       "step": 26219
     },
     {
-      "epoch": 0.7201120540496004,
+      "epoch": 0.7440408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 3.834992941151097e-06,
-      "loss": 0.8442,
+      "learning_rate": 3.2439285217691106e-06,
+      "loss": 0.9437,
       "step": 26220
     },
     {
-      "epoch": 0.7201395182774436,
+      "epoch": 0.7440692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 3.834292601235947e-06,
-      "loss": 0.8844,
+      "learning_rate": 3.243250949140887e-06,
+      "loss": 0.8338,
       "step": 26221
     },
     {
-      "epoch": 0.7201669825052869,
+      "epoch": 0.7440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 3.833592310106266e-06,
-      "loss": 0.8973,
+      "learning_rate": 3.242573433587224e-06,
+      "loss": 0.8515,
       "step": 26222
     },
     {
-      "epoch": 0.72019444673313,
+      "epoch": 0.7441259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 3.832892067767587e-06,
-      "loss": 0.8359,
+      "learning_rate": 3.241895975113849e-06,
+      "loss": 0.8353,
       "step": 26223
     },
     {
-      "epoch": 0.7202219109609733,
+      "epoch": 0.7441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 3.832191874225454e-06,
-      "loss": 0.7648,
+      "learning_rate": 3.24121857372648e-06,
+      "loss": 0.8127,
       "step": 26224
     },
     {
-      "epoch": 0.7202493751888166,
+      "epoch": 0.7441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 3.83149172948541e-06,
-      "loss": 0.7207,
+      "learning_rate": 3.2405412294308413e-06,
+      "loss": 0.8483,
       "step": 26225
     },
     {
-      "epoch": 0.7202768394166598,
+      "epoch": 0.744211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 3.830791633552991e-06,
-      "loss": 0.8889,
+      "learning_rate": 3.2398639422326583e-06,
+      "loss": 0.8035,
       "step": 26226
     },
     {
-      "epoch": 0.720304303644503,
+      "epoch": 0.7442395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 3.830091586433742e-06,
-      "loss": 0.8751,
+      "learning_rate": 3.239186712137642e-06,
+      "loss": 0.8003,
       "step": 26227
     },
     {
-      "epoch": 0.7203317678723463,
+      "epoch": 0.7442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 3.829391588133196e-06,
-      "loss": 0.8444,
+      "learning_rate": 3.238509539151522e-06,
+      "loss": 0.7952,
       "step": 26228
     },
     {
-      "epoch": 0.7203592321001895,
+      "epoch": 0.7442962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 3.828691638656896e-06,
-      "loss": 0.8533,
+      "learning_rate": 3.2378324232800195e-06,
+      "loss": 0.8846,
       "step": 26229
     },
     {
-      "epoch": 0.7203866963280328,
+      "epoch": 0.7443246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 3.827991738010376e-06,
-      "loss": 0.8222,
+      "learning_rate": 3.2371553645288467e-06,
+      "loss": 0.8056,
       "step": 26230
     },
     {
-      "epoch": 0.7204141605558759,
+      "epoch": 0.7443530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 3.827291886199176e-06,
-      "loss": 0.7623,
+      "learning_rate": 3.2364783629037277e-06,
+      "loss": 0.8139,
       "step": 26231
     },
     {
-      "epoch": 0.7204416247837192,
+      "epoch": 0.7443813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 3.826592083228834e-06,
-      "loss": 0.825,
+      "learning_rate": 3.2358014184103827e-06,
+      "loss": 0.8237,
       "step": 26232
     },
     {
-      "epoch": 0.7204690890115625,
+      "epoch": 0.7444097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 3.82589232910489e-06,
-      "loss": 0.8146,
+      "learning_rate": 3.235124531054523e-06,
+      "loss": 0.8612,
       "step": 26233
     },
     {
-      "epoch": 0.7204965532394056,
+      "epoch": 0.7444381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 3.8251926238328725e-06,
-      "loss": 0.8666,
+      "learning_rate": 3.2344477008418717e-06,
+      "loss": 0.8165,
       "step": 26234
     },
     {
-      "epoch": 0.7205240174672489,
+      "epoch": 0.7444665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 3.824492967418325e-06,
-      "loss": 0.7527,
+      "learning_rate": 3.2337709277781484e-06,
+      "loss": 0.7128,
       "step": 26235
     },
     {
-      "epoch": 0.7205514816950921,
+      "epoch": 0.7444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 3.8237933598667805e-06,
-      "loss": 0.8522,
+      "learning_rate": 3.2330942118690625e-06,
+      "loss": 0.7692,
       "step": 26236
     },
     {
-      "epoch": 0.7205789459229354,
+      "epoch": 0.7445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 3.82309380118378e-06,
-      "loss": 0.8679,
+      "learning_rate": 3.2324175531203327e-06,
+      "loss": 0.7645,
       "step": 26237
     },
     {
-      "epoch": 0.7206064101507786,
+      "epoch": 0.7445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 3.822394291374848e-06,
-      "loss": 0.8743,
+      "learning_rate": 3.2317409515376808e-06,
+      "loss": 0.8363,
       "step": 26238
     },
     {
-      "epoch": 0.7206338743786218,
+      "epoch": 0.7445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 3.82169483044553e-06,
-      "loss": 0.8475,
+      "learning_rate": 3.231064407126814e-06,
+      "loss": 0.8606,
       "step": 26239
     },
     {
-      "epoch": 0.7206613386064651,
+      "epoch": 0.7446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 3.8209954184013535e-06,
-      "loss": 0.8878,
+      "learning_rate": 3.2303879198934494e-06,
+      "loss": 0.7972,
       "step": 26240
     },
     {
-      "epoch": 0.7206888028343084,
+      "epoch": 0.7446367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 3.820296055247853e-06,
-      "loss": 0.8963,
+      "learning_rate": 3.2297114898433067e-06,
+      "loss": 0.7998,
       "step": 26241
     },
     {
-      "epoch": 0.7207162670621515,
+      "epoch": 0.7446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 3.8195967409905655e-06,
-      "loss": 0.8174,
+      "learning_rate": 3.2290351169820888e-06,
+      "loss": 0.7687,
       "step": 26242
     },
     {
-      "epoch": 0.7207437312899948,
+      "epoch": 0.7446935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 3.818897475635022e-06,
-      "loss": 0.8124,
+      "learning_rate": 3.228358801315522e-06,
+      "loss": 0.9002,
       "step": 26243
     },
     {
-      "epoch": 0.720771195517838,
+      "epoch": 0.74472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.818198259186759e-06,
-      "loss": 0.8586,
+      "learning_rate": 3.2276825428493085e-06,
+      "loss": 0.9277,
       "step": 26244
     },
     {
-      "epoch": 0.7207986597456812,
+      "epoch": 0.744750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 3.817499091651305e-06,
-      "loss": 0.7928,
+      "learning_rate": 3.2270063415891663e-06,
+      "loss": 0.731,
       "step": 26245
     },
     {
-      "epoch": 0.7208261239735245,
+      "epoch": 0.7447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 3.8167999730341924e-06,
-      "loss": 0.8806,
+      "learning_rate": 3.226330197540809e-06,
+      "loss": 0.6864,
       "step": 26246
     },
     {
-      "epoch": 0.7208535882013677,
+      "epoch": 0.7448070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 3.816100903340951e-06,
-      "loss": 0.7824,
+      "learning_rate": 3.2256541107099416e-06,
+      "loss": 0.746,
       "step": 26247
     },
     {
-      "epoch": 0.720881052429211,
+      "epoch": 0.7448354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 3.8154018825771155e-06,
-      "loss": 0.748,
+      "learning_rate": 3.224978081102279e-06,
+      "loss": 0.8068,
       "step": 26248
     },
     {
-      "epoch": 0.7209085166570541,
+      "epoch": 0.7448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 3.814702910748217e-06,
-      "loss": 0.8559,
+      "learning_rate": 3.2243021087235336e-06,
+      "loss": 0.7185,
       "step": 26249
     },
     {
-      "epoch": 0.7209359808848974,
+      "epoch": 0.7448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 3.8140039878597833e-06,
-      "loss": 0.8384,
+      "learning_rate": 3.22362619357941e-06,
+      "loss": 0.7812,
       "step": 26250
     },
     {
-      "epoch": 0.7209634451127407,
+      "epoch": 0.7449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 3.8133051139173517e-06,
-      "loss": 0.7977,
+      "learning_rate": 3.2229503356756196e-06,
+      "loss": 0.7336,
       "step": 26251
     },
     {
-      "epoch": 0.7209909093405839,
+      "epoch": 0.7449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 3.8126062889264415e-06,
-      "loss": 0.766,
+      "learning_rate": 3.2222745350178776e-06,
+      "loss": 0.7434,
       "step": 26252
     },
     {
-      "epoch": 0.7210183735684271,
+      "epoch": 0.7449772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 3.811907512892591e-06,
-      "loss": 0.7339,
+      "learning_rate": 3.221598791611882e-06,
+      "loss": 0.7769,
       "step": 26253
     },
     {
-      "epoch": 0.7210458377962704,
+      "epoch": 0.745005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 3.811208785821322e-06,
-      "loss": 0.8085,
+      "learning_rate": 3.220923105463347e-06,
+      "loss": 0.9579,
       "step": 26254
     },
     {
-      "epoch": 0.7210733020241136,
+      "epoch": 0.7450340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 3.810510107718166e-06,
-      "loss": 0.8453,
+      "learning_rate": 3.220247476577982e-06,
+      "loss": 0.7777,
       "step": 26255
     },
     {
-      "epoch": 0.7211007662519568,
+      "epoch": 0.7450624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 3.809811478588652e-06,
-      "loss": 0.8316,
+      "learning_rate": 3.2195719049614893e-06,
+      "loss": 0.7685,
       "step": 26256
     },
     {
-      "epoch": 0.7211282304798,
+      "epoch": 0.7450908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 3.809112898438311e-06,
-      "loss": 0.829,
+      "learning_rate": 3.218896390619577e-06,
+      "loss": 0.8353,
       "step": 26257
     },
     {
-      "epoch": 0.7211556947076433,
+      "epoch": 0.7451191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 3.8084143672726637e-06,
-      "loss": 0.6954,
+      "learning_rate": 3.2182209335579516e-06,
+      "loss": 0.8424,
       "step": 26258
     },
     {
-      "epoch": 0.7211831589354866,
+      "epoch": 0.7451475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 3.8077158850972397e-06,
-      "loss": 0.8546,
+      "learning_rate": 3.217545533782319e-06,
+      "loss": 0.7576,
       "step": 26259
     },
     {
-      "epoch": 0.7212106231633297,
+      "epoch": 0.7451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 3.8070174519175696e-06,
-      "loss": 0.8822,
+      "learning_rate": 3.216870191298387e-06,
+      "loss": 0.8036,
       "step": 26260
     },
     {
-      "epoch": 0.721238087391173,
+      "epoch": 0.7452043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 3.8063190677391727e-06,
-      "loss": 0.8591,
+      "learning_rate": 3.2161949061118547e-06,
+      "loss": 0.8326,
       "step": 26261
     },
     {
-      "epoch": 0.7212655516190162,
+      "epoch": 0.7452326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 3.8056207325675797e-06,
-      "loss": 0.8213,
+      "learning_rate": 3.2155196782284282e-06,
+      "loss": 0.8957,
       "step": 26262
     },
     {
-      "epoch": 0.7212930158468595,
+      "epoch": 0.745261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.804922446408318e-06,
-      "loss": 0.8085,
+      "learning_rate": 3.214844507653816e-06,
+      "loss": 0.8967,
       "step": 26263
     },
     {
-      "epoch": 0.7213204800747027,
+      "epoch": 0.7452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.8042242092669056e-06,
-      "loss": 0.8606,
+      "learning_rate": 3.2141693943937134e-06,
+      "loss": 0.8509,
       "step": 26264
     },
     {
-      "epoch": 0.7213479443025459,
+      "epoch": 0.7453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 3.803526021148871e-06,
-      "loss": 0.8546,
+      "learning_rate": 3.2134943384538274e-06,
+      "loss": 0.7884,
       "step": 26265
     },
     {
-      "epoch": 0.7213754085303892,
+      "epoch": 0.7453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 3.80282788205974e-06,
-      "loss": 0.7759,
+      "learning_rate": 3.212819339839863e-06,
+      "loss": 0.7819,
       "step": 26266
     },
     {
-      "epoch": 0.7214028727582324,
+      "epoch": 0.7453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 3.802129792005035e-06,
-      "loss": 0.8285,
+      "learning_rate": 3.212144398557515e-06,
+      "loss": 0.7802,
       "step": 26267
     },
     {
-      "epoch": 0.7214303369860756,
+      "epoch": 0.7454029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 3.8014317509902822e-06,
-      "loss": 0.7828,
+      "learning_rate": 3.2114695146124885e-06,
+      "loss": 0.7511,
       "step": 26268
     },
     {
-      "epoch": 0.7214578012139189,
+      "epoch": 0.7454313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 3.8007337590210035e-06,
-      "loss": 0.8282,
+      "learning_rate": 3.210794688010488e-06,
+      "loss": 0.8334,
       "step": 26269
     },
     {
-      "epoch": 0.7214852654417621,
+      "epoch": 0.7454597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 3.8000358161027175e-06,
-      "loss": 0.8699,
+      "learning_rate": 3.2101199187572064e-06,
+      "loss": 0.8182,
       "step": 26270
     },
     {
-      "epoch": 0.7215127296696053,
+      "epoch": 0.7454880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 3.7993379222409486e-06,
-      "loss": 0.777,
+      "learning_rate": 3.209445206858347e-06,
+      "loss": 0.8534,
       "step": 26271
     },
     {
-      "epoch": 0.7215401938974486,
+      "epoch": 0.7455164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 3.7986400774412215e-06,
-      "loss": 0.8177,
+      "learning_rate": 3.2087705523196135e-06,
+      "loss": 0.7766,
       "step": 26272
     },
     {
-      "epoch": 0.7215676581252918,
+      "epoch": 0.7455448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 3.797942281709055e-06,
-      "loss": 0.8477,
+      "learning_rate": 3.2080959551466926e-06,
+      "loss": 0.8035,
       "step": 26273
     },
     {
-      "epoch": 0.721595122353135,
+      "epoch": 0.7455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 3.797244535049972e-06,
-      "loss": 0.6984,
+      "learning_rate": 3.2074214153452975e-06,
+      "loss": 0.766,
       "step": 26274
     },
     {
-      "epoch": 0.7216225865809783,
+      "epoch": 0.7456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 3.7965468374694968e-06,
-      "loss": 0.7838,
+      "learning_rate": 3.2067469329211154e-06,
+      "loss": 0.9624,
       "step": 26275
     },
     {
-      "epoch": 0.7216500508088215,
+      "epoch": 0.7456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 3.795849188973142e-06,
-      "loss": 0.7989,
+      "learning_rate": 3.206072507879847e-06,
+      "loss": 0.7561,
       "step": 26276
     },
     {
-      "epoch": 0.7216775150366648,
+      "epoch": 0.7456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 3.795151589566436e-06,
-      "loss": 0.7214,
+      "learning_rate": 3.205398140227194e-06,
+      "loss": 0.8033,
       "step": 26277
     },
     {
-      "epoch": 0.7217049792645079,
+      "epoch": 0.7456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 3.79445403925489e-06,
-      "loss": 0.8825,
+      "learning_rate": 3.2047238299688443e-06,
+      "loss": 0.7093,
       "step": 26278
     },
     {
-      "epoch": 0.7217324434923512,
+      "epoch": 0.7457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 3.7937565380440268e-06,
-      "loss": 0.8324,
+      "learning_rate": 3.2040495771104983e-06,
+      "loss": 0.8244,
       "step": 26279
     },
     {
-      "epoch": 0.7217599077201945,
+      "epoch": 0.7457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 3.7930590859393666e-06,
-      "loss": 0.8367,
+      "learning_rate": 3.2033753816578537e-06,
+      "loss": 0.7847,
       "step": 26280
     },
     {
-      "epoch": 0.7217873719480377,
+      "epoch": 0.745771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 3.7923616829464314e-06,
-      "loss": 0.891,
+      "learning_rate": 3.2027012436166004e-06,
+      "loss": 0.7898,
       "step": 26281
     },
     {
-      "epoch": 0.7218148361758809,
+      "epoch": 0.745800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 3.791664329070731e-06,
-      "loss": 0.8916,
+      "learning_rate": 3.2020271629924347e-06,
+      "loss": 0.8431,
       "step": 26282
     },
     {
-      "epoch": 0.7218423004037241,
+      "epoch": 0.7458286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 3.7909670243177877e-06,
-      "loss": 0.7459,
+      "learning_rate": 3.2013531397910546e-06,
+      "loss": 0.7986,
       "step": 26283
     },
     {
-      "epoch": 0.7218697646315674,
+      "epoch": 0.7458569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 3.790269768693121e-06,
-      "loss": 0.832,
+      "learning_rate": 3.2006791740181466e-06,
+      "loss": 0.8125,
       "step": 26284
     },
     {
-      "epoch": 0.7218972288594107,
+      "epoch": 0.7458853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 3.789572562202243e-06,
-      "loss": 0.8239,
+      "learning_rate": 3.2000052656794066e-06,
+      "loss": 0.8374,
       "step": 26285
     },
     {
-      "epoch": 0.7219246930872538,
+      "epoch": 0.7459137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 3.7888754048506747e-06,
-      "loss": 0.8452,
+      "learning_rate": 3.1993314147805322e-06,
+      "loss": 0.8412,
       "step": 26286
     },
     {
-      "epoch": 0.7219521573150971,
+      "epoch": 0.7459421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 3.7881782966439285e-06,
-      "loss": 0.7822,
+      "learning_rate": 3.198657621327208e-06,
+      "loss": 0.8715,
       "step": 26287
     },
     {
-      "epoch": 0.7219796215429404,
+      "epoch": 0.7459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 3.7874812375875204e-06,
-      "loss": 0.8764,
+      "learning_rate": 3.1979838853251277e-06,
+      "loss": 0.7561,
       "step": 26288
     },
     {
-      "epoch": 0.7220070857707835,
+      "epoch": 0.7459988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 3.786784227686967e-06,
-      "loss": 0.7851,
+      "learning_rate": 3.1973102067799833e-06,
+      "loss": 0.8327,
       "step": 26289
     },
     {
-      "epoch": 0.7220345499986268,
+      "epoch": 0.7460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 3.7860872669477844e-06,
-      "loss": 0.9456,
+      "learning_rate": 3.1966365856974656e-06,
+      "loss": 0.8879,
       "step": 26290
     },
     {
-      "epoch": 0.72206201422647,
+      "epoch": 0.74605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.785390355375487e-06,
-      "loss": 0.8371,
+      "learning_rate": 3.1959630220832683e-06,
+      "loss": 0.8772,
       "step": 26291
     },
     {
-      "epoch": 0.7220894784543133,
+      "epoch": 0.7460839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 3.7846934929755906e-06,
-      "loss": 0.7922,
+      "learning_rate": 3.195289515943073e-06,
+      "loss": 0.8665,
       "step": 26292
     },
     {
-      "epoch": 0.7221169426821565,
+      "epoch": 0.7461123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 3.7839966797536078e-06,
-      "loss": 0.87,
+      "learning_rate": 3.1946160672825742e-06,
+      "loss": 0.8373,
       "step": 26293
     },
     {
-      "epoch": 0.7221444069099997,
+      "epoch": 0.7461407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 3.7832999157150475e-06,
-      "loss": 0.7805,
+      "learning_rate": 3.193942676107462e-06,
+      "loss": 0.8923,
       "step": 26294
     },
     {
-      "epoch": 0.722171871137843,
+      "epoch": 0.7461691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 3.7826032008654256e-06,
-      "loss": 0.7815,
+      "learning_rate": 3.1932693424234186e-06,
+      "loss": 0.7994,
       "step": 26295
     },
     {
-      "epoch": 0.7221993353656861,
+      "epoch": 0.7461975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 3.781906535210256e-06,
-      "loss": 0.8013,
+      "learning_rate": 3.1925960662361355e-06,
+      "loss": 0.7851,
       "step": 26296
     },
     {
-      "epoch": 0.7222267995935294,
+      "epoch": 0.7462258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 3.781209918755051e-06,
-      "loss": 0.9178,
+      "learning_rate": 3.191922847551302e-06,
+      "loss": 0.8637,
       "step": 26297
     },
     {
-      "epoch": 0.7222542638213727,
+      "epoch": 0.7462542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 3.780513351505325e-06,
-      "loss": 0.7976,
+      "learning_rate": 3.1912496863746e-06,
+      "loss": 0.8747,
       "step": 26298
     },
     {
-      "epoch": 0.7222817280492159,
+      "epoch": 0.7462826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 3.779816833466583e-06,
-      "loss": 0.8941,
+      "learning_rate": 3.1905765827117173e-06,
+      "loss": 0.8222,
       "step": 26299
     },
     {
-      "epoch": 0.7223091922770591,
+      "epoch": 0.746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 3.7791203646443398e-06,
-      "loss": 0.886,
+      "learning_rate": 3.189903536568343e-06,
+      "loss": 0.8073,
       "step": 26300
     },
     {
-      "epoch": 0.7223366565049024,
+      "epoch": 0.7463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 3.7784239450441097e-06,
-      "loss": 0.7752,
+      "learning_rate": 3.1892305479501574e-06,
+      "loss": 0.8938,
       "step": 26301
     },
     {
-      "epoch": 0.7223641207327456,
+      "epoch": 0.7463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 3.777727574671396e-06,
-      "loss": 0.8701,
+      "learning_rate": 3.1885576168628473e-06,
+      "loss": 0.806,
       "step": 26302
     },
     {
-      "epoch": 0.7223915849605889,
+      "epoch": 0.7463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 3.777031253531712e-06,
-      "loss": 0.767,
+      "learning_rate": 3.1878847433121005e-06,
+      "loss": 0.8175,
       "step": 26303
     },
     {
-      "epoch": 0.722419049188432,
+      "epoch": 0.7464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 3.7763349816305706e-06,
-      "loss": 0.8186,
+      "learning_rate": 3.1872119273035905e-06,
+      "loss": 0.952,
       "step": 26304
     },
     {
-      "epoch": 0.7224465134162753,
+      "epoch": 0.7464528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 3.7756387589734735e-06,
-      "loss": 0.8485,
+      "learning_rate": 3.1865391688430147e-06,
+      "loss": 0.7627,
       "step": 26305
     },
     {
-      "epoch": 0.7224739776441186,
+      "epoch": 0.7464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 3.774942585565935e-06,
-      "loss": 0.7874,
+      "learning_rate": 3.185866467936045e-06,
+      "loss": 0.7404,
       "step": 26306
     },
     {
-      "epoch": 0.7225014418719617,
+      "epoch": 0.7465096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 3.774246461413461e-06,
-      "loss": 0.8448,
+      "learning_rate": 3.1851938245883686e-06,
+      "loss": 0.8472,
       "step": 26307
     },
     {
-      "epoch": 0.722528906099805,
+      "epoch": 0.7465380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 3.773550386521564e-06,
-      "loss": 0.8057,
+      "learning_rate": 3.184521238805668e-06,
+      "loss": 0.7607,
       "step": 26308
     },
     {
-      "epoch": 0.7225563703276482,
+      "epoch": 0.746566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 3.7728543608957458e-06,
-      "loss": 0.8188,
+      "learning_rate": 3.1838487105936212e-06,
+      "loss": 0.925,
       "step": 26309
     },
     {
-      "epoch": 0.7225838345554915,
+      "epoch": 0.746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 3.7721583845415186e-06,
-      "loss": 0.8175,
+      "learning_rate": 3.1831762399579093e-06,
+      "loss": 0.7825,
       "step": 26310
     },
     {
-      "epoch": 0.7226112987833347,
+      "epoch": 0.7466231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 3.771462457464383e-06,
-      "loss": 0.8583,
+      "learning_rate": 3.1825038269042184e-06,
+      "loss": 0.8647,
       "step": 26311
     },
     {
-      "epoch": 0.7226387630111779,
+      "epoch": 0.7466515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 3.77076657966985e-06,
-      "loss": 0.795,
+      "learning_rate": 3.1818314714382202e-06,
+      "loss": 0.8954,
       "step": 26312
     },
     {
-      "epoch": 0.7226662272390212,
+      "epoch": 0.7466799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 3.770070751163425e-06,
-      "loss": 0.7938,
+      "learning_rate": 3.181159173565599e-06,
+      "loss": 0.8253,
       "step": 26313
     },
     {
-      "epoch": 0.7226936914668645,
+      "epoch": 0.7467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 3.7693749719506134e-06,
-      "loss": 0.7526,
+      "learning_rate": 3.180486933292035e-06,
+      "loss": 0.8238,
       "step": 26314
     },
     {
-      "epoch": 0.7227211556947076,
+      "epoch": 0.7467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.76867924203692e-06,
-      "loss": 0.8517,
+      "learning_rate": 3.1798147506232023e-06,
+      "loss": 0.8812,
       "step": 26315
     },
     {
-      "epoch": 0.7227486199225509,
+      "epoch": 0.7467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 3.767983561427854e-06,
-      "loss": 0.8297,
+      "learning_rate": 3.1791426255647805e-06,
+      "loss": 0.8659,
       "step": 26316
     },
     {
-      "epoch": 0.7227760841503941,
+      "epoch": 0.7467934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 3.7672879301289154e-06,
-      "loss": 0.9174,
+      "learning_rate": 3.1784705581224508e-06,
+      "loss": 0.8369,
       "step": 26317
     },
     {
-      "epoch": 0.7228035483782373,
+      "epoch": 0.7468217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 3.7665923481456057e-06,
-      "loss": 0.88,
+      "learning_rate": 3.1777985483018835e-06,
+      "loss": 0.8086,
       "step": 26318
     },
     {
-      "epoch": 0.7228310126060806,
+      "epoch": 0.7468501702610669,
       "grad_norm": 0.0,
-      "learning_rate": 3.765896815483433e-06,
-      "loss": 0.8027,
+      "learning_rate": 3.1771265961087548e-06,
+      "loss": 0.8383,
       "step": 26319
     },
     {
-      "epoch": 0.7228584768339238,
+      "epoch": 0.7468785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 3.7652013321478984e-06,
-      "loss": 0.8612,
+      "learning_rate": 3.176454701548751e-06,
+      "loss": 0.848,
       "step": 26320
     },
     {
-      "epoch": 0.7228859410617671,
+      "epoch": 0.7469069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 3.764505898144507e-06,
-      "loss": 0.8315,
+      "learning_rate": 3.175782864627539e-06,
+      "loss": 0.7825,
       "step": 26321
     },
     {
-      "epoch": 0.7229134052896102,
+      "epoch": 0.7469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 3.7638105134787627e-06,
-      "loss": 0.7653,
+      "learning_rate": 3.175111085350795e-06,
+      "loss": 0.8287,
       "step": 26322
     },
     {
-      "epoch": 0.7229408695174535,
+      "epoch": 0.7469636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 3.7631151781561626e-06,
-      "loss": 0.8534,
+      "learning_rate": 3.174439363724199e-06,
+      "loss": 0.8706,
       "step": 26323
     },
     {
-      "epoch": 0.7229683337452968,
+      "epoch": 0.7469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 3.7624198921822107e-06,
-      "loss": 0.7948,
+      "learning_rate": 3.173767699753416e-06,
+      "loss": 0.8994,
       "step": 26324
     },
     {
-      "epoch": 0.72299579797314,
+      "epoch": 0.7470204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 3.761724655562413e-06,
-      "loss": 0.8579,
+      "learning_rate": 3.1730960934441234e-06,
+      "loss": 0.8941,
       "step": 26325
     },
     {
-      "epoch": 0.7230232622009832,
+      "epoch": 0.7470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 3.761029468302262e-06,
-      "loss": 0.7708,
+      "learning_rate": 3.1724245448019996e-06,
+      "loss": 0.8326,
       "step": 26326
     },
     {
-      "epoch": 0.7230507264288265,
+      "epoch": 0.7470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 3.760334330407263e-06,
-      "loss": 0.8448,
+      "learning_rate": 3.171753053832709e-06,
+      "loss": 0.8236,
       "step": 26327
     },
     {
-      "epoch": 0.7230781906566697,
+      "epoch": 0.747105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.7596392418829187e-06,
-      "loss": 0.7296,
+      "learning_rate": 3.171081620541927e-06,
+      "loss": 0.832,
       "step": 26328
     },
     {
-      "epoch": 0.723105654884513,
+      "epoch": 0.7471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 3.7589442027347233e-06,
-      "loss": 0.796,
+      "learning_rate": 3.170410244935329e-06,
+      "loss": 0.7683,
       "step": 26329
     },
     {
-      "epoch": 0.7231331191123561,
+      "epoch": 0.7471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 3.758249212968179e-06,
-      "loss": 0.8196,
+      "learning_rate": 3.16973892701858e-06,
+      "loss": 0.8582,
       "step": 26330
     },
     {
-      "epoch": 0.7231605833401994,
+      "epoch": 0.7471906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 3.7575542725887836e-06,
-      "loss": 0.8512,
+      "learning_rate": 3.1690676667973563e-06,
+      "loss": 0.7669,
       "step": 26331
     },
     {
-      "epoch": 0.7231880475680427,
+      "epoch": 0.7472190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 3.7568593816020417e-06,
-      "loss": 0.7563,
+      "learning_rate": 3.1683964642773215e-06,
+      "loss": 0.8346,
       "step": 26332
     },
     {
-      "epoch": 0.7232155117958858,
+      "epoch": 0.7472474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 3.7561645400134427e-06,
-      "loss": 0.7018,
+      "learning_rate": 3.167725319464149e-06,
+      "loss": 0.8098,
       "step": 26333
     },
     {
-      "epoch": 0.7232429760237291,
+      "epoch": 0.7472758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 3.7554697478284917e-06,
-      "loss": 0.7038,
+      "learning_rate": 3.1670542323635077e-06,
+      "loss": 0.9046,
       "step": 26334
     },
     {
-      "epoch": 0.7232704402515723,
+      "epoch": 0.7473041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 3.75477500505268e-06,
-      "loss": 0.972,
+      "learning_rate": 3.166383202981066e-06,
+      "loss": 0.8513,
       "step": 26335
     },
     {
-      "epoch": 0.7232979044794156,
+      "epoch": 0.7473325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 3.754080311691507e-06,
-      "loss": 0.8117,
+      "learning_rate": 3.165712231322493e-06,
+      "loss": 0.7909,
       "step": 26336
     },
     {
-      "epoch": 0.7233253687072588,
+      "epoch": 0.7473609534619751,
       "grad_norm": 0.0,
-      "learning_rate": 3.753385667750471e-06,
-      "loss": 0.8421,
+      "learning_rate": 3.1650413173934604e-06,
+      "loss": 0.8374,
       "step": 26337
     },
     {
-      "epoch": 0.723352832935102,
+      "epoch": 0.7473893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 3.7526910732350663e-06,
-      "loss": 0.7725,
+      "learning_rate": 3.1643704611996262e-06,
+      "loss": 0.8859,
       "step": 26338
     },
     {
-      "epoch": 0.7233802971629453,
+      "epoch": 0.7474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 3.751996528150791e-06,
-      "loss": 0.8693,
+      "learning_rate": 3.1636996627466622e-06,
+      "loss": 0.8412,
       "step": 26339
     },
     {
-      "epoch": 0.7234077613907886,
+      "epoch": 0.7474460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 3.751302032503141e-06,
-      "loss": 0.9533,
+      "learning_rate": 3.163028922040239e-06,
+      "loss": 0.8038,
       "step": 26340
     },
     {
-      "epoch": 0.7234352256186317,
+      "epoch": 0.7474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 3.750607586297612e-06,
-      "loss": 0.7518,
+      "learning_rate": 3.162358239086013e-06,
+      "loss": 0.8786,
       "step": 26341
     },
     {
-      "epoch": 0.723462689846475,
+      "epoch": 0.7475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 3.749913189539691e-06,
-      "loss": 0.9291,
+      "learning_rate": 3.161687613889655e-06,
+      "loss": 0.7886,
       "step": 26342
     },
     {
-      "epoch": 0.7234901540743182,
+      "epoch": 0.7475312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 3.7492188422348797e-06,
-      "loss": 0.7859,
+      "learning_rate": 3.1610170464568325e-06,
+      "loss": 0.7258,
       "step": 26343
     },
     {
-      "epoch": 0.7235176183021614,
+      "epoch": 0.7475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 3.74852454438867e-06,
-      "loss": 0.8487,
+      "learning_rate": 3.160346536793203e-06,
+      "loss": 0.8402,
       "step": 26344
     },
     {
-      "epoch": 0.7235450825300047,
+      "epoch": 0.7475879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 3.747830296006555e-06,
-      "loss": 0.9012,
+      "learning_rate": 3.159676084904434e-06,
+      "loss": 0.7585,
       "step": 26345
     },
     {
-      "epoch": 0.7235725467578479,
+      "epoch": 0.747616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 3.7471360970940317e-06,
-      "loss": 0.7852,
+      "learning_rate": 3.1590056907961918e-06,
+      "loss": 0.8353,
       "step": 26346
     },
     {
-      "epoch": 0.7236000109856912,
+      "epoch": 0.747644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 3.7464419476565874e-06,
-      "loss": 0.8631,
+      "learning_rate": 3.1583353544741322e-06,
+      "loss": 0.8326,
       "step": 26347
     },
     {
-      "epoch": 0.7236274752135343,
+      "epoch": 0.7476730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 3.7457478476997157e-06,
-      "loss": 0.8174,
+      "learning_rate": 3.157665075943922e-06,
+      "loss": 0.7802,
       "step": 26348
     },
     {
-      "epoch": 0.7236549394413776,
+      "epoch": 0.7477014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 3.7450537972289136e-06,
-      "loss": 0.8222,
+      "learning_rate": 3.156994855211226e-06,
+      "loss": 0.8159,
       "step": 26349
     },
     {
-      "epoch": 0.7236824036692209,
+      "epoch": 0.7477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 3.7443597962496646e-06,
-      "loss": 0.9067,
+      "learning_rate": 3.1563246922816947e-06,
+      "loss": 0.882,
       "step": 26350
     },
     {
-      "epoch": 0.723709867897064,
+      "epoch": 0.7477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 3.743665844767469e-06,
-      "loss": 0.7782,
+      "learning_rate": 3.1556545871610034e-06,
+      "loss": 0.8721,
       "step": 26351
     },
     {
-      "epoch": 0.7237373321249073,
+      "epoch": 0.7477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 3.7429719427878085e-06,
-      "loss": 0.863,
+      "learning_rate": 3.154984539854803e-06,
+      "loss": 0.8494,
       "step": 26352
     },
     {
-      "epoch": 0.7237647963527506,
+      "epoch": 0.7478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 3.742278090316177e-06,
-      "loss": 0.8208,
+      "learning_rate": 3.1543145503687546e-06,
+      "loss": 0.6747,
       "step": 26353
     },
     {
-      "epoch": 0.7237922605805938,
+      "epoch": 0.7478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 3.7415842873580664e-06,
-      "loss": 0.7806,
+      "learning_rate": 3.1536446187085236e-06,
+      "loss": 0.8276,
       "step": 26354
     },
     {
-      "epoch": 0.723819724808437,
+      "epoch": 0.7478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 3.740890533918965e-06,
-      "loss": 0.8299,
+      "learning_rate": 3.1529747448797598e-06,
+      "loss": 0.7904,
       "step": 26355
     },
     {
-      "epoch": 0.7238471890362802,
+      "epoch": 0.747900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 3.7401968300043655e-06,
-      "loss": 0.9201,
+      "learning_rate": 3.152304928888126e-06,
+      "loss": 0.7101,
       "step": 26356
     },
     {
-      "epoch": 0.7238746532641235,
+      "epoch": 0.7479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 3.739503175619753e-06,
-      "loss": 0.8358,
+      "learning_rate": 3.1516351707392843e-06,
+      "loss": 0.8998,
       "step": 26357
     },
     {
-      "epoch": 0.7239021174919668,
+      "epoch": 0.7479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 3.738809570770614e-06,
-      "loss": 0.8139,
+      "learning_rate": 3.150965470438885e-06,
+      "loss": 0.8059,
       "step": 26358
     },
     {
-      "epoch": 0.7239295817198099,
+      "epoch": 0.7479852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 3.7381160154624385e-06,
-      "loss": 0.9067,
+      "learning_rate": 3.150295827992588e-06,
+      "loss": 0.8364,
       "step": 26359
     },
     {
-      "epoch": 0.7239570459476532,
+      "epoch": 0.7480136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 3.7374225097007146e-06,
-      "loss": 0.748,
+      "learning_rate": 3.1496262434060522e-06,
+      "loss": 0.7401,
       "step": 26360
     },
     {
-      "epoch": 0.7239845101754964,
+      "epoch": 0.7480419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 3.7367290534909305e-06,
-      "loss": 0.8953,
+      "learning_rate": 3.14895671668493e-06,
+      "loss": 0.8256,
       "step": 26361
     },
     {
-      "epoch": 0.7240119744033396,
+      "epoch": 0.7480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 3.7360356468385717e-06,
-      "loss": 0.7726,
+      "learning_rate": 3.1482872478348768e-06,
+      "loss": 0.8109,
       "step": 26362
     },
     {
-      "epoch": 0.7240394386311829,
+      "epoch": 0.7480987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 3.7353422897491287e-06,
-      "loss": 0.8138,
+      "learning_rate": 3.147617836861554e-06,
+      "loss": 0.8349,
       "step": 26363
     },
     {
-      "epoch": 0.7240669028590261,
+      "epoch": 0.7481271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 3.7346489822280808e-06,
-      "loss": 0.8506,
+      "learning_rate": 3.1469484837706065e-06,
+      "loss": 0.8709,
       "step": 26364
     },
     {
-      "epoch": 0.7240943670868694,
+      "epoch": 0.748155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 3.73395572428092e-06,
-      "loss": 0.7801,
+      "learning_rate": 3.1462791885676948e-06,
+      "loss": 0.9258,
       "step": 26365
     },
     {
-      "epoch": 0.7241218313147126,
+      "epoch": 0.7481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 3.733262515913125e-06,
-      "loss": 0.8231,
+      "learning_rate": 3.14560995125847e-06,
+      "loss": 0.7668,
       "step": 26366
     },
     {
-      "epoch": 0.7241492955425558,
+      "epoch": 0.7482122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 3.7325693571301858e-06,
-      "loss": 0.8828,
+      "learning_rate": 3.144940771848586e-06,
+      "loss": 0.796,
       "step": 26367
     },
     {
-      "epoch": 0.7241767597703991,
+      "epoch": 0.7482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 3.731876247937586e-06,
-      "loss": 0.8176,
+      "learning_rate": 3.144271650343699e-06,
+      "loss": 0.8766,
       "step": 26368
     },
     {
-      "epoch": 0.7242042239982422,
+      "epoch": 0.7482690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 3.7311831883408113e-06,
-      "loss": 0.913,
+      "learning_rate": 3.1436025867494545e-06,
+      "loss": 0.8717,
       "step": 26369
     },
     {
-      "epoch": 0.7242316882260855,
+      "epoch": 0.7482973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 3.7304901783453407e-06,
-      "loss": 0.7461,
+      "learning_rate": 3.1429335810715067e-06,
+      "loss": 0.8653,
       "step": 26370
     },
     {
-      "epoch": 0.7242591524539288,
+      "epoch": 0.7483257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 3.72979721795666e-06,
-      "loss": 0.7751,
+      "learning_rate": 3.1422646333155103e-06,
+      "loss": 0.8804,
       "step": 26371
     },
     {
-      "epoch": 0.724286616681772,
+      "epoch": 0.7483541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 3.7291043071802533e-06,
-      "loss": 0.8405,
+      "learning_rate": 3.1415957434871105e-06,
+      "loss": 0.7904,
       "step": 26372
     },
     {
-      "epoch": 0.7243140809096152,
+      "epoch": 0.7483825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 3.7284114460216057e-06,
-      "loss": 0.8003,
+      "learning_rate": 3.1409269115919593e-06,
+      "loss": 0.7702,
       "step": 26373
     },
     {
-      "epoch": 0.7243415451374584,
+      "epoch": 0.7484108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 3.7277186344861926e-06,
-      "loss": 0.8989,
+      "learning_rate": 3.14025813763571e-06,
+      "loss": 0.8281,
       "step": 26374
     },
     {
-      "epoch": 0.7243690093653017,
+      "epoch": 0.748439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 3.7270258725795017e-06,
-      "loss": 0.8091,
+      "learning_rate": 3.1395894216240054e-06,
+      "loss": 0.6728,
       "step": 26375
     },
     {
-      "epoch": 0.724396473593145,
+      "epoch": 0.7484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 3.7263331603070094e-06,
-      "loss": 0.834,
+      "learning_rate": 3.1389207635624974e-06,
+      "loss": 0.9048,
       "step": 26376
     },
     {
-      "epoch": 0.7244239378209881,
+      "epoch": 0.7484960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 3.7256404976742e-06,
-      "loss": 0.7802,
+      "learning_rate": 3.1382521634568365e-06,
+      "loss": 0.8,
       "step": 26377
     },
     {
-      "epoch": 0.7244514020488314,
+      "epoch": 0.7485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 3.7249478846865526e-06,
-      "loss": 0.7567,
+      "learning_rate": 3.1375836213126653e-06,
+      "loss": 0.8628,
       "step": 26378
     },
     {
-      "epoch": 0.7244788662766747,
+      "epoch": 0.7485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 3.724255321349549e-06,
-      "loss": 0.864,
+      "learning_rate": 3.1369151371356343e-06,
+      "loss": 0.8134,
       "step": 26379
     },
     {
-      "epoch": 0.7245063305045178,
+      "epoch": 0.7485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 3.723562807668671e-06,
-      "loss": 0.7588,
+      "learning_rate": 3.1362467109313898e-06,
+      "loss": 0.8239,
       "step": 26380
     },
     {
-      "epoch": 0.7245337947323611,
+      "epoch": 0.7486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 3.7228703436493952e-06,
-      "loss": 0.8254,
+      "learning_rate": 3.1355783427055773e-06,
+      "loss": 0.7785,
       "step": 26381
     },
     {
-      "epoch": 0.7245612589602043,
+      "epoch": 0.7486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 3.722177929297198e-06,
-      "loss": 0.8012,
+      "learning_rate": 3.134910032463846e-06,
+      "loss": 0.762,
       "step": 26382
     },
     {
-      "epoch": 0.7245887231880476,
+      "epoch": 0.74866628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.7214855646175607e-06,
-      "loss": 0.8055,
+      "learning_rate": 3.1342417802118365e-06,
+      "loss": 0.8149,
       "step": 26383
     },
     {
-      "epoch": 0.7246161874158908,
+      "epoch": 0.7486946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 3.7207932496159615e-06,
-      "loss": 0.9343,
+      "learning_rate": 3.133573585955194e-06,
+      "loss": 0.9041,
       "step": 26384
     },
     {
-      "epoch": 0.724643651643734,
+      "epoch": 0.7487230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 3.7201009842978786e-06,
-      "loss": 0.8306,
+      "learning_rate": 3.132905449699567e-06,
+      "loss": 0.8137,
       "step": 26385
     },
     {
-      "epoch": 0.7246711158715773,
+      "epoch": 0.7487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 3.7194087686687884e-06,
-      "loss": 0.8978,
+      "learning_rate": 3.132237371450594e-06,
+      "loss": 0.8282,
       "step": 26386
     },
     {
-      "epoch": 0.7246985800994205,
+      "epoch": 0.7487797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 3.718716602734174e-06,
-      "loss": 0.855,
+      "learning_rate": 3.13156935121392e-06,
+      "loss": 0.7822,
       "step": 26387
     },
     {
-      "epoch": 0.7247260443272637,
+      "epoch": 0.7488081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 3.718024486499502e-06,
-      "loss": 0.7773,
+      "learning_rate": 3.130901388995192e-06,
+      "loss": 0.8076,
       "step": 26388
     },
     {
-      "epoch": 0.724753508555107,
+      "epoch": 0.7488365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 3.7173324199702575e-06,
-      "loss": 0.9325,
+      "learning_rate": 3.130233484800046e-06,
+      "loss": 0.8082,
       "step": 26389
     },
     {
-      "epoch": 0.7247809727829502,
+      "epoch": 0.7488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 3.7166404031519097e-06,
-      "loss": 0.8388,
+      "learning_rate": 3.1295656386341267e-06,
+      "loss": 0.8095,
       "step": 26390
     },
     {
-      "epoch": 0.7248084370107934,
+      "epoch": 0.7488933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 3.7159484360499366e-06,
-      "loss": 0.8695,
+      "learning_rate": 3.1288978505030777e-06,
+      "loss": 0.7326,
       "step": 26391
     },
     {
-      "epoch": 0.7248359012386367,
+      "epoch": 0.7489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 3.715256518669814e-06,
-      "loss": 0.9314,
+      "learning_rate": 3.1282301204125342e-06,
+      "loss": 0.7857,
       "step": 26392
     },
     {
-      "epoch": 0.7248633654664799,
+      "epoch": 0.748950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 3.7145646510170196e-06,
-      "loss": 0.9597,
+      "learning_rate": 3.12756244836814e-06,
+      "loss": 0.7939,
       "step": 26393
     },
     {
-      "epoch": 0.7248908296943232,
+      "epoch": 0.7489784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 3.713872833097021e-06,
-      "loss": 0.8743,
+      "learning_rate": 3.126894834375539e-06,
+      "loss": 0.8559,
       "step": 26394
     },
     {
-      "epoch": 0.7249182939221663,
+      "epoch": 0.7490068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 3.7131810649152943e-06,
-      "loss": 0.7835,
+      "learning_rate": 3.1262272784403613e-06,
+      "loss": 0.871,
       "step": 26395
     },
     {
-      "epoch": 0.7249457581500096,
+      "epoch": 0.7490351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 3.7124893464773193e-06,
-      "loss": 0.8946,
+      "learning_rate": 3.1255597805682515e-06,
+      "loss": 0.8151,
       "step": 26396
     },
     {
-      "epoch": 0.7249732223778529,
+      "epoch": 0.7490635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 3.71179767778856e-06,
-      "loss": 0.8265,
+      "learning_rate": 3.124892340764848e-06,
+      "loss": 0.8741,
       "step": 26397
     },
     {
-      "epoch": 0.725000686605696,
+      "epoch": 0.7490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 3.7111060588544933e-06,
-      "loss": 0.7889,
+      "learning_rate": 3.1242249590357877e-06,
+      "loss": 0.8159,
       "step": 26398
     },
     {
-      "epoch": 0.7250281508335393,
+      "epoch": 0.7491203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 3.710414489680596e-06,
-      "loss": 0.8351,
+      "learning_rate": 3.12355763538671e-06,
+      "loss": 0.7676,
       "step": 26399
     },
     {
-      "epoch": 0.7250556150613825,
+      "epoch": 0.7491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 3.7097229702723316e-06,
-      "loss": 0.8944,
+      "learning_rate": 3.122890369823248e-06,
+      "loss": 0.7603,
       "step": 26400
     },
     {
-      "epoch": 0.7250830792892258,
+      "epoch": 0.7491770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 3.709031500635175e-06,
-      "loss": 0.9043,
+      "learning_rate": 3.12222316235104e-06,
+      "loss": 0.8843,
       "step": 26401
     },
     {
-      "epoch": 0.725110543517069,
+      "epoch": 0.749205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 3.7083400807745983e-06,
-      "loss": 0.7889,
+      "learning_rate": 3.121556012975726e-06,
+      "loss": 0.7987,
       "step": 26402
     },
     {
-      "epoch": 0.7251380077449122,
+      "epoch": 0.7492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 3.7076487106960724e-06,
-      "loss": 0.9447,
+      "learning_rate": 3.1208889217029336e-06,
+      "loss": 0.7124,
       "step": 26403
     },
     {
-      "epoch": 0.7251654719727555,
+      "epoch": 0.7492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 3.7069573904050703e-06,
-      "loss": 0.8123,
+      "learning_rate": 3.120221888538303e-06,
+      "loss": 0.9607,
       "step": 26404
     },
     {
-      "epoch": 0.7251929362005988,
+      "epoch": 0.7492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 3.7062661199070592e-06,
-      "loss": 0.833,
+      "learning_rate": 3.119554913487469e-06,
+      "loss": 0.8155,
       "step": 26405
     },
     {
-      "epoch": 0.7252204004284419,
+      "epoch": 0.7493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 3.705574899207506e-06,
-      "loss": 0.8079,
+      "learning_rate": 3.1188879965560614e-06,
+      "loss": 0.9381,
       "step": 26406
     },
     {
-      "epoch": 0.7252478646562852,
+      "epoch": 0.7493473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 3.704883728311881e-06,
-      "loss": 0.924,
+      "learning_rate": 3.118221137749716e-06,
+      "loss": 0.8076,
       "step": 26407
     },
     {
-      "epoch": 0.7252753288841284,
+      "epoch": 0.7493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 3.704192607225655e-06,
-      "loss": 0.8424,
+      "learning_rate": 3.117554337074069e-06,
+      "loss": 0.8197,
       "step": 26408
     },
     {
-      "epoch": 0.7253027931119717,
+      "epoch": 0.7494040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 3.7035015359542956e-06,
-      "loss": 0.8975,
+      "learning_rate": 3.116887594534748e-06,
+      "loss": 0.7543,
       "step": 26409
     },
     {
-      "epoch": 0.7253302573398149,
+      "epoch": 0.7494324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 3.7028105145032754e-06,
-      "loss": 0.8285,
+      "learning_rate": 3.116220910137385e-06,
+      "loss": 0.7994,
       "step": 26410
     },
     {
-      "epoch": 0.7253577215676581,
+      "epoch": 0.749460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 3.702119542878053e-06,
-      "loss": 0.8517,
+      "learning_rate": 3.115554283887614e-06,
+      "loss": 0.8172,
       "step": 26411
     },
     {
-      "epoch": 0.7253851857955014,
+      "epoch": 0.749489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 3.701428621084102e-06,
-      "loss": 0.7743,
+      "learning_rate": 3.114887715791066e-06,
+      "loss": 0.7918,
       "step": 26412
     },
     {
-      "epoch": 0.7254126500233445,
+      "epoch": 0.7495175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 3.7007377491268902e-06,
-      "loss": 0.8241,
+      "learning_rate": 3.1142212058533738e-06,
+      "loss": 0.748,
       "step": 26413
     },
     {
-      "epoch": 0.7254401142511878,
+      "epoch": 0.7495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 3.7000469270118777e-06,
-      "loss": 0.8059,
+      "learning_rate": 3.1135547540801625e-06,
+      "loss": 0.8632,
       "step": 26414
     },
     {
-      "epoch": 0.7254675784790311,
+      "epoch": 0.7495743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 3.6993561547445344e-06,
-      "loss": 0.7627,
+      "learning_rate": 3.112888360477062e-06,
+      "loss": 0.7155,
       "step": 26415
     },
     {
-      "epoch": 0.7254950427068743,
+      "epoch": 0.7496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 3.6986654323303295e-06,
-      "loss": 0.7492,
+      "learning_rate": 3.112222025049707e-06,
+      "loss": 0.8603,
       "step": 26416
     },
     {
-      "epoch": 0.7255225069347175,
+      "epoch": 0.7496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 3.6979747597747205e-06,
-      "loss": 0.7707,
+      "learning_rate": 3.1115557478037196e-06,
+      "loss": 0.8072,
       "step": 26417
     },
     {
-      "epoch": 0.7255499711625608,
+      "epoch": 0.7496594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 3.697284137083177e-06,
-      "loss": 0.9403,
+      "learning_rate": 3.1108895287447303e-06,
+      "loss": 0.789,
       "step": 26418
     },
     {
-      "epoch": 0.725577435390404,
+      "epoch": 0.7496878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 3.696593564261163e-06,
-      "loss": 0.8016,
+      "learning_rate": 3.11022336787837e-06,
+      "loss": 0.8193,
       "step": 26419
     },
     {
-      "epoch": 0.7256048996182473,
+      "epoch": 0.7497162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 3.695903041314145e-06,
-      "loss": 0.8419,
+      "learning_rate": 3.109557265210259e-06,
+      "loss": 0.8514,
       "step": 26420
     },
     {
-      "epoch": 0.7256323638460904,
+      "epoch": 0.7497446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 3.6952125682475814e-06,
-      "loss": 0.7846,
+      "learning_rate": 3.1088912207460276e-06,
+      "loss": 0.8309,
       "step": 26421
     },
     {
-      "epoch": 0.7256598280739337,
+      "epoch": 0.7497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 3.694522145066941e-06,
-      "loss": 0.7133,
+      "learning_rate": 3.1082252344913045e-06,
+      "loss": 0.8112,
       "step": 26422
     },
     {
-      "epoch": 0.725687292301777,
+      "epoch": 0.7498013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 3.6938317717776815e-06,
-      "loss": 0.8051,
+      "learning_rate": 3.1075593064517105e-06,
+      "loss": 0.7551,
       "step": 26423
     },
     {
-      "epoch": 0.7257147565296201,
+      "epoch": 0.7498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 3.693141448385267e-06,
-      "loss": 0.8177,
+      "learning_rate": 3.1068934366328727e-06,
+      "loss": 0.8665,
       "step": 26424
     },
     {
-      "epoch": 0.7257422207574634,
+      "epoch": 0.7498581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 3.692451174895161e-06,
-      "loss": 0.8789,
+      "learning_rate": 3.106227625040419e-06,
+      "loss": 0.9057,
       "step": 26425
     },
     {
-      "epoch": 0.7257696849853066,
+      "epoch": 0.7498864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 3.691760951312824e-06,
-      "loss": 0.778,
+      "learning_rate": 3.105561871679966e-06,
+      "loss": 0.8691,
       "step": 26426
     },
     {
-      "epoch": 0.7257971492131499,
+      "epoch": 0.7499148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 3.6910707776437182e-06,
-      "loss": 0.7856,
+      "learning_rate": 3.104896176557141e-06,
+      "loss": 0.8053,
       "step": 26427
     },
     {
-      "epoch": 0.7258246134409931,
+      "epoch": 0.7499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 3.6903806538933075e-06,
-      "loss": 0.8037,
+      "learning_rate": 3.1042305396775728e-06,
+      "loss": 0.9173,
       "step": 26428
     },
     {
-      "epoch": 0.7258520776688363,
+      "epoch": 0.7499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 3.68969058006705e-06,
-      "loss": 0.7357,
+      "learning_rate": 3.1035649610468753e-06,
+      "loss": 0.8286,
       "step": 26429
     },
     {
-      "epoch": 0.7258795418966796,
+      "epoch": 0.75,
       "grad_norm": 0.0,
-      "learning_rate": 3.6890005561704002e-06,
-      "loss": 0.7832,
+      "learning_rate": 3.1028994406706757e-06,
+      "loss": 0.8642,
       "step": 26430
     },
     {
-      "epoch": 0.7259070061245229,
+      "epoch": 0.7500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 3.688310582208824e-06,
-      "loss": 0.824,
+      "learning_rate": 3.1022339785545986e-06,
+      "loss": 0.8142,
       "step": 26431
     },
     {
-      "epoch": 0.725934470352366,
+      "epoch": 0.7500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 3.687620658187778e-06,
-      "loss": 0.8229,
+      "learning_rate": 3.101568574704257e-06,
+      "loss": 0.8024,
       "step": 26432
     },
     {
-      "epoch": 0.7259619345802093,
+      "epoch": 0.7500851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 3.6869307841127232e-06,
-      "loss": 0.7866,
+      "learning_rate": 3.100903229125275e-06,
+      "loss": 0.706,
       "step": 26433
     },
     {
-      "epoch": 0.7259893988080525,
+      "epoch": 0.7501135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 3.6862409599891214e-06,
-      "loss": 0.8245,
+      "learning_rate": 3.1002379418232773e-06,
+      "loss": 0.8231,
       "step": 26434
     },
     {
-      "epoch": 0.7260168630358957,
+      "epoch": 0.7501418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 3.6855511858224234e-06,
-      "loss": 0.9413,
+      "learning_rate": 3.0995727128038775e-06,
+      "loss": 0.8209,
       "step": 26435
     },
     {
-      "epoch": 0.726044327263739,
+      "epoch": 0.7501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 3.68486146161809e-06,
-      "loss": 0.8301,
+      "learning_rate": 3.0989075420726967e-06,
+      "loss": 0.7974,
       "step": 26436
     },
     {
-      "epoch": 0.7260717914915822,
+      "epoch": 0.7501986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 3.684171787381582e-06,
-      "loss": 0.8253,
+      "learning_rate": 3.0982424296353576e-06,
+      "loss": 0.7863,
       "step": 26437
     },
     {
-      "epoch": 0.7260992557194255,
+      "epoch": 0.7502270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 3.6834821631183506e-06,
-      "loss": 0.9348,
+      "learning_rate": 3.0975773754974705e-06,
+      "loss": 0.9217,
       "step": 26438
     },
     {
-      "epoch": 0.7261267199472686,
+      "epoch": 0.7502553916004541,
       "grad_norm": 0.0,
-      "learning_rate": 3.6827925888338543e-06,
-      "loss": 0.8465,
+      "learning_rate": 3.0969123796646617e-06,
+      "loss": 0.9422,
       "step": 26439
     },
     {
-      "epoch": 0.7261541841751119,
+      "epoch": 0.7502837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 3.682103064533554e-06,
-      "loss": 0.8018,
+      "learning_rate": 3.0962474421425413e-06,
+      "loss": 0.7675,
       "step": 26440
     },
     {
-      "epoch": 0.7261816484029552,
+      "epoch": 0.7503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 3.681413590222899e-06,
-      "loss": 0.8411,
+      "learning_rate": 3.0955825629367288e-06,
+      "loss": 0.8298,
       "step": 26441
     },
     {
-      "epoch": 0.7262091126307983,
+      "epoch": 0.7503405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 3.680724165907347e-06,
-      "loss": 0.8282,
+      "learning_rate": 3.094917742052841e-06,
+      "loss": 0.8341,
       "step": 26442
     },
     {
-      "epoch": 0.7262365768586416,
+      "epoch": 0.7503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.6800347915923518e-06,
-      "loss": 0.8246,
+      "learning_rate": 3.0942529794964926e-06,
+      "loss": 0.7798,
       "step": 26443
     },
     {
-      "epoch": 0.7262640410864849,
+      "epoch": 0.7503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 3.679345467283374e-06,
-      "loss": 0.841,
+      "learning_rate": 3.0935882752733005e-06,
+      "loss": 0.8249,
       "step": 26444
     },
     {
-      "epoch": 0.7262915053143281,
+      "epoch": 0.7504256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 3.67865619298586e-06,
-      "loss": 0.8063,
+      "learning_rate": 3.092923629388882e-06,
+      "loss": 0.7832,
       "step": 26445
     },
     {
-      "epoch": 0.7263189695421713,
+      "epoch": 0.7504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 3.67796696870527e-06,
-      "loss": 0.7978,
+      "learning_rate": 3.092259041848845e-06,
+      "loss": 0.8572,
       "step": 26446
     },
     {
-      "epoch": 0.7263464337700145,
+      "epoch": 0.7504824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 3.6772777944470505e-06,
-      "loss": 0.8234,
+      "learning_rate": 3.0915945126588063e-06,
+      "loss": 0.8484,
       "step": 26447
     },
     {
-      "epoch": 0.7263738979978578,
+      "epoch": 0.750510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 3.6765886702166586e-06,
-      "loss": 0.812,
+      "learning_rate": 3.090930041824383e-06,
+      "loss": 0.7743,
       "step": 26448
     },
     {
-      "epoch": 0.7264013622257011,
+      "epoch": 0.750539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 3.6758995960195466e-06,
-      "loss": 0.8186,
+      "learning_rate": 3.090265629351179e-06,
+      "loss": 0.7241,
       "step": 26449
     },
     {
-      "epoch": 0.7264288264535442,
+      "epoch": 0.7505675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 3.6752105718611664e-06,
-      "loss": 0.8359,
+      "learning_rate": 3.0896012752448134e-06,
+      "loss": 0.7696,
       "step": 26450
     },
     {
-      "epoch": 0.7264562906813875,
+      "epoch": 0.7505959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 3.6745215977469704e-06,
-      "loss": 0.7925,
+      "learning_rate": 3.088936979510899e-06,
+      "loss": 0.8244,
       "step": 26451
     },
     {
-      "epoch": 0.7264837549092307,
+      "epoch": 0.7506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 3.6738326736824126e-06,
-      "loss": 0.8803,
+      "learning_rate": 3.0882727421550428e-06,
+      "loss": 0.8245,
       "step": 26452
     },
     {
-      "epoch": 0.726511219137074,
+      "epoch": 0.7506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 3.6731437996729425e-06,
-      "loss": 0.8106,
+      "learning_rate": 3.0876085631828568e-06,
+      "loss": 0.8887,
       "step": 26453
     },
     {
-      "epoch": 0.7265386833649172,
+      "epoch": 0.7506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 3.6724549757240046e-06,
-      "loss": 0.7986,
+      "learning_rate": 3.086944442599954e-06,
+      "loss": 0.8218,
       "step": 26454
     },
     {
-      "epoch": 0.7265661475927604,
+      "epoch": 0.7507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 3.671766201841055e-06,
-      "loss": 0.8254,
+      "learning_rate": 3.08628038041194e-06,
+      "loss": 0.8075,
       "step": 26455
     },
     {
-      "epoch": 0.7265936118206037,
+      "epoch": 0.7507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 3.671077478029542e-06,
-      "loss": 0.7748,
+      "learning_rate": 3.085616376624426e-06,
+      "loss": 0.7523,
       "step": 26456
     },
     {
-      "epoch": 0.726621076048447,
+      "epoch": 0.7507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 3.6703888042949166e-06,
-      "loss": 0.8768,
+      "learning_rate": 3.0849524312430223e-06,
+      "loss": 0.8047,
       "step": 26457
     },
     {
-      "epoch": 0.7266485402762901,
+      "epoch": 0.750794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 3.669700180642629e-06,
-      "loss": 0.7811,
+      "learning_rate": 3.084288544273336e-06,
+      "loss": 0.8073,
       "step": 26458
     },
     {
-      "epoch": 0.7266760045041334,
+      "epoch": 0.7508229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 3.6690116070781225e-06,
-      "loss": 0.7965,
+      "learning_rate": 3.0836247157209775e-06,
+      "loss": 0.8615,
       "step": 26459
     },
     {
-      "epoch": 0.7267034687319766,
+      "epoch": 0.7508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 3.6683230836068484e-06,
-      "loss": 0.8358,
+      "learning_rate": 3.08296094559155e-06,
+      "loss": 0.8445,
       "step": 26460
     },
     {
-      "epoch": 0.7267309329598198,
+      "epoch": 0.7508796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 3.6676346102342586e-06,
-      "loss": 0.9621,
+      "learning_rate": 3.082297233890661e-06,
+      "loss": 0.8721,
       "step": 26461
     },
     {
-      "epoch": 0.7267583971876631,
+      "epoch": 0.7509080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 3.666946186965794e-06,
-      "loss": 0.8307,
+      "learning_rate": 3.081633580623923e-06,
+      "loss": 0.8781,
       "step": 26462
     },
     {
-      "epoch": 0.7267858614155063,
+      "epoch": 0.7509364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 3.6662578138069028e-06,
-      "loss": 0.8257,
+      "learning_rate": 3.0809699857969323e-06,
+      "loss": 0.9177,
       "step": 26463
     },
     {
-      "epoch": 0.7268133256433496,
+      "epoch": 0.7509648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 3.665569490763038e-06,
-      "loss": 0.8705,
+      "learning_rate": 3.0803064494153002e-06,
+      "loss": 0.7953,
       "step": 26464
     },
     {
-      "epoch": 0.7268407898711927,
+      "epoch": 0.7509931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 3.6648812178396364e-06,
-      "loss": 0.8852,
+      "learning_rate": 3.079642971484633e-06,
+      "loss": 0.8918,
       "step": 26465
     },
     {
-      "epoch": 0.726868254099036,
+      "epoch": 0.7510215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 3.6641929950421497e-06,
-      "loss": 0.8642,
+      "learning_rate": 3.0789795520105303e-06,
+      "loss": 0.8409,
       "step": 26466
     },
     {
-      "epoch": 0.7268957183268793,
+      "epoch": 0.751049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 3.6635048223760217e-06,
-      "loss": 0.7818,
+      "learning_rate": 3.078316190998598e-06,
+      "loss": 0.7729,
       "step": 26467
     },
     {
-      "epoch": 0.7269231825547224,
+      "epoch": 0.7510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 3.6628166998467018e-06,
-      "loss": 0.8183,
+      "learning_rate": 3.0776528884544433e-06,
+      "loss": 0.7699,
       "step": 26468
     },
     {
-      "epoch": 0.7269506467825657,
+      "epoch": 0.7511066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 3.662128627459628e-06,
-      "loss": 0.7262,
+      "learning_rate": 3.076989644383663e-06,
+      "loss": 0.8183,
       "step": 26469
     },
     {
-      "epoch": 0.726978111010409,
+      "epoch": 0.7511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 3.66144060522025e-06,
-      "loss": 0.8866,
+      "learning_rate": 3.0763264587918616e-06,
+      "loss": 0.7751,
       "step": 26470
     },
     {
-      "epoch": 0.7270055752382522,
+      "epoch": 0.7511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 3.660752633134006e-06,
-      "loss": 0.8698,
+      "learning_rate": 3.0756633316846463e-06,
+      "loss": 0.8914,
       "step": 26471
     },
     {
-      "epoch": 0.7270330394660954,
+      "epoch": 0.7511918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 3.6600647112063435e-06,
-      "loss": 0.9138,
+      "learning_rate": 3.07500026306761e-06,
+      "loss": 0.8767,
       "step": 26472
     },
     {
-      "epoch": 0.7270605036939386,
+      "epoch": 0.7512202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 3.6593768394427043e-06,
-      "loss": 0.8331,
+      "learning_rate": 3.0743372529463577e-06,
+      "loss": 0.8775,
       "step": 26473
     },
     {
-      "epoch": 0.7270879679217819,
+      "epoch": 0.7512485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 3.6586890178485314e-06,
-      "loss": 0.9154,
+      "learning_rate": 3.0736743013264903e-06,
+      "loss": 0.8813,
       "step": 26474
     },
     {
-      "epoch": 0.7271154321496252,
+      "epoch": 0.7512769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 3.6580012464292712e-06,
-      "loss": 0.8597,
+      "learning_rate": 3.0730114082136077e-06,
+      "loss": 0.75,
       "step": 26475
     },
     {
-      "epoch": 0.7271428963774683,
+      "epoch": 0.7513053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 3.657313525190358e-06,
-      "loss": 0.8367,
+      "learning_rate": 3.0723485736133117e-06,
+      "loss": 0.8332,
       "step": 26476
     },
     {
-      "epoch": 0.7271703606053116,
+      "epoch": 0.75133371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.656625854137241e-06,
-      "loss": 0.7987,
+      "learning_rate": 3.0716857975311953e-06,
+      "loss": 0.8874,
       "step": 26477
     },
     {
-      "epoch": 0.7271978248331549,
+      "epoch": 0.7513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 3.6559382332753536e-06,
-      "loss": 0.8557,
+      "learning_rate": 3.0710230799728613e-06,
+      "loss": 0.8055,
       "step": 26478
     },
     {
-      "epoch": 0.727225289060998,
+      "epoch": 0.7513904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 3.6552506626101415e-06,
-      "loss": 0.8616,
+      "learning_rate": 3.0703604209439097e-06,
+      "loss": 0.7652,
       "step": 26479
     },
     {
-      "epoch": 0.7272527532888413,
+      "epoch": 0.7514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 3.654563142147043e-06,
-      "loss": 0.7825,
+      "learning_rate": 3.0696978204499318e-06,
+      "loss": 0.8062,
       "step": 26480
     },
     {
-      "epoch": 0.7272802175166845,
+      "epoch": 0.7514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.6538756718915016e-06,
-      "loss": 0.757,
+      "learning_rate": 3.0690352784965274e-06,
+      "loss": 0.9251,
       "step": 26481
     },
     {
-      "epoch": 0.7273076817445278,
+      "epoch": 0.7514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 3.6531882518489513e-06,
-      "loss": 0.8514,
+      "learning_rate": 3.068372795089297e-06,
+      "loss": 0.782,
       "step": 26482
     },
     {
-      "epoch": 0.727335145972371,
+      "epoch": 0.7515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 3.6525008820248344e-06,
-      "loss": 0.8906,
+      "learning_rate": 3.0677103702338306e-06,
+      "loss": 0.8082,
       "step": 26483
     },
     {
-      "epoch": 0.7273626102002142,
+      "epoch": 0.7515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 3.6518135624245886e-06,
-      "loss": 0.9044,
+      "learning_rate": 3.0670480039357264e-06,
+      "loss": 0.7542,
       "step": 26484
     },
     {
-      "epoch": 0.7273900744280575,
+      "epoch": 0.751560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 3.6511262930536575e-06,
-      "loss": 0.7556,
+      "learning_rate": 3.0663856962005822e-06,
+      "loss": 0.8415,
       "step": 26485
     },
     {
-      "epoch": 0.7274175386559006,
+      "epoch": 0.7515891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 3.65043907391747e-06,
-      "loss": 0.8375,
+      "learning_rate": 3.065723447033987e-06,
+      "loss": 0.8507,
       "step": 26486
     },
     {
-      "epoch": 0.7274450028837439,
+      "epoch": 0.7516174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 3.6497519050214724e-06,
-      "loss": 0.7899,
+      "learning_rate": 3.065061256441536e-06,
+      "loss": 0.9136,
       "step": 26487
     },
     {
-      "epoch": 0.7274724671115872,
+      "epoch": 0.7516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 3.649064786371094e-06,
-      "loss": 0.8908,
+      "learning_rate": 3.064399124428826e-06,
+      "loss": 0.9674,
       "step": 26488
     },
     {
-      "epoch": 0.7274999313394304,
+      "epoch": 0.7516742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 3.648377717971774e-06,
-      "loss": 0.9088,
+      "learning_rate": 3.0637370510014474e-06,
+      "loss": 0.6662,
       "step": 26489
     },
     {
-      "epoch": 0.7275273955672736,
+      "epoch": 0.7517026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 3.6476906998289506e-06,
-      "loss": 0.9586,
+      "learning_rate": 3.0630750361649974e-06,
+      "loss": 0.8803,
       "step": 26490
     },
     {
-      "epoch": 0.7275548597951169,
+      "epoch": 0.7517309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 3.64700373194806e-06,
-      "loss": 0.7476,
+      "learning_rate": 3.062413079925062e-06,
+      "loss": 0.7691,
       "step": 26491
     },
     {
-      "epoch": 0.7275823240229601,
+      "epoch": 0.7517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 3.6463168143345394e-06,
-      "loss": 0.7994,
+      "learning_rate": 3.0617511822872337e-06,
+      "loss": 0.7426,
       "step": 26492
     },
     {
-      "epoch": 0.7276097882508034,
+      "epoch": 0.7517877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 3.6456299469938173e-06,
-      "loss": 0.8006,
+      "learning_rate": 3.0610893432571086e-06,
+      "loss": 0.8405,
       "step": 26493
     },
     {
-      "epoch": 0.7276372524786465,
+      "epoch": 0.7518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.6449431299313365e-06,
-      "loss": 0.8568,
+      "learning_rate": 3.0604275628402713e-06,
+      "loss": 0.885,
       "step": 26494
     },
     {
-      "epoch": 0.7276647167064898,
+      "epoch": 0.751844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 3.644256363152524e-06,
-      "loss": 0.744,
+      "learning_rate": 3.0597658410423135e-06,
+      "loss": 0.736,
       "step": 26495
     },
     {
-      "epoch": 0.7276921809343331,
+      "epoch": 0.7518728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 3.643569646662818e-06,
-      "loss": 0.8935,
+      "learning_rate": 3.059104177868829e-06,
+      "loss": 0.7454,
       "step": 26496
     },
     {
-      "epoch": 0.7277196451621762,
+      "epoch": 0.7519012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 3.642882980467649e-06,
-      "loss": 0.8499,
+      "learning_rate": 3.058442573325401e-06,
+      "loss": 0.7538,
       "step": 26497
     },
     {
-      "epoch": 0.7277471093900195,
+      "epoch": 0.7519296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 3.6421963645724545e-06,
-      "loss": 0.844,
+      "learning_rate": 3.05778102741762e-06,
+      "loss": 0.9208,
       "step": 26498
     },
     {
-      "epoch": 0.7277745736178627,
+      "epoch": 0.7519580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 3.6415097989826674e-06,
-      "loss": 0.8108,
+      "learning_rate": 3.0571195401510777e-06,
+      "loss": 0.8629,
       "step": 26499
     },
     {
-      "epoch": 0.727802037845706,
+      "epoch": 0.7519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 3.6408232837037148e-06,
-      "loss": 0.924,
+      "learning_rate": 3.056458111531356e-06,
+      "loss": 0.9111,
       "step": 26500
     },
     {
-      "epoch": 0.7278295020735492,
+      "epoch": 0.7520147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 3.6401368187410344e-06,
-      "loss": 0.8362,
+      "learning_rate": 3.0557967415640456e-06,
+      "loss": 0.7792,
       "step": 26501
     },
     {
-      "epoch": 0.7278569663013924,
+      "epoch": 0.7520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 3.6394504041000534e-06,
-      "loss": 0.873,
+      "learning_rate": 3.0551354302547343e-06,
+      "loss": 0.7464,
       "step": 26502
     },
     {
-      "epoch": 0.7278844305292357,
+      "epoch": 0.7520715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 3.6387640397862036e-06,
-      "loss": 0.8346,
+      "learning_rate": 3.0544741776089993e-06,
+      "loss": 0.9159,
       "step": 26503
     },
     {
-      "epoch": 0.727911894757079,
+      "epoch": 0.752099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 3.638077725804916e-06,
-      "loss": 0.8585,
+      "learning_rate": 3.05381298363244e-06,
+      "loss": 0.8254,
       "step": 26504
     },
     {
-      "epoch": 0.7279393589849221,
+      "epoch": 0.7521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 3.637391462161627e-06,
-      "loss": 0.8174,
+      "learning_rate": 3.053151848330632e-06,
+      "loss": 0.8672,
       "step": 26505
     },
     {
-      "epoch": 0.7279668232127654,
+      "epoch": 0.7521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 3.636705248861756e-06,
-      "loss": 0.8396,
+      "learning_rate": 3.0524907717091636e-06,
+      "loss": 0.781,
       "step": 26506
     },
     {
-      "epoch": 0.7279942874406086,
+      "epoch": 0.7521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 3.6360190859107392e-06,
-      "loss": 0.7922,
+      "learning_rate": 3.0518297537736197e-06,
+      "loss": 0.7557,
       "step": 26507
     },
     {
-      "epoch": 0.7280217516684518,
+      "epoch": 0.7522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 3.6353329733140042e-06,
-      "loss": 0.9138,
+      "learning_rate": 3.0511687945295788e-06,
+      "loss": 0.7734,
       "step": 26508
     },
     {
-      "epoch": 0.7280492158962951,
+      "epoch": 0.7522417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 3.6346469110769834e-06,
-      "loss": 0.8235,
+      "learning_rate": 3.050507893982628e-06,
+      "loss": 0.7842,
       "step": 26509
     },
     {
-      "epoch": 0.7280766801241383,
+      "epoch": 0.7522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 3.633960899205099e-06,
-      "loss": 0.8821,
+      "learning_rate": 3.0498470521383527e-06,
+      "loss": 0.8872,
       "step": 26510
     },
     {
-      "epoch": 0.7281041443519816,
+      "epoch": 0.7522985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 3.6332749377037857e-06,
-      "loss": 0.8699,
+      "learning_rate": 3.049186269002329e-06,
+      "loss": 0.8406,
       "step": 26511
     },
     {
-      "epoch": 0.7281316085798247,
+      "epoch": 0.7523269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 3.6325890265784625e-06,
-      "loss": 0.7219,
+      "learning_rate": 3.0485255445801397e-06,
+      "loss": 0.7309,
       "step": 26512
     },
     {
-      "epoch": 0.728159072807668,
+      "epoch": 0.752355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 3.6319031658345628e-06,
-      "loss": 0.8425,
+      "learning_rate": 3.0478648788773723e-06,
+      "loss": 0.9653,
       "step": 26513
     },
     {
-      "epoch": 0.7281865370355113,
+      "epoch": 0.752383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 3.631217355477511e-06,
-      "loss": 0.8806,
+      "learning_rate": 3.0472042718996e-06,
+      "loss": 0.8484,
       "step": 26514
     },
     {
-      "epoch": 0.7282140012633544,
+      "epoch": 0.7524120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 3.630531595512735e-06,
-      "loss": 0.864,
+      "learning_rate": 3.046543723652404e-06,
+      "loss": 0.7841,
       "step": 26515
     },
     {
-      "epoch": 0.7282414654911977,
+      "epoch": 0.7524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 3.6298458859456643e-06,
-      "loss": 0.9121,
+      "learning_rate": 3.04588323414137e-06,
+      "loss": 1.0076,
       "step": 26516
     },
     {
-      "epoch": 0.728268929719041,
+      "epoch": 0.7524687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 3.6291602267817193e-06,
-      "loss": 0.7815,
+      "learning_rate": 3.0452228033720697e-06,
+      "loss": 0.8499,
       "step": 26517
     },
     {
-      "epoch": 0.7282963939468842,
+      "epoch": 0.7524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 3.628474618026324e-06,
-      "loss": 0.8762,
+      "learning_rate": 3.0445624313500853e-06,
+      "loss": 0.8122,
       "step": 26518
     },
     {
-      "epoch": 0.7283238581747274,
+      "epoch": 0.7525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 3.6277890596849054e-06,
-      "loss": 0.8172,
+      "learning_rate": 3.0439021180809946e-06,
+      "loss": 0.8078,
       "step": 26519
     },
     {
-      "epoch": 0.7283513224025706,
+      "epoch": 0.7525539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 3.6271035517628885e-06,
-      "loss": 0.8702,
+      "learning_rate": 3.0432418635703753e-06,
+      "loss": 0.8397,
       "step": 26520
     },
     {
-      "epoch": 0.7283787866304139,
+      "epoch": 0.7525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 3.6264180942656955e-06,
-      "loss": 0.7648,
+      "learning_rate": 3.0425816678238073e-06,
+      "loss": 0.7811,
       "step": 26521
     },
     {
-      "epoch": 0.7284062508582572,
+      "epoch": 0.7526106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 3.625732687198753e-06,
-      "loss": 0.7978,
+      "learning_rate": 3.041921530846862e-06,
+      "loss": 0.8433,
       "step": 26522
     },
     {
-      "epoch": 0.7284337150861003,
+      "epoch": 0.7526390465380249,
       "grad_norm": 0.0,
-      "learning_rate": 3.625047330567485e-06,
-      "loss": 0.7202,
+      "learning_rate": 3.041261452645119e-06,
+      "loss": 0.8448,
       "step": 26523
     },
     {
-      "epoch": 0.7284611793139436,
+      "epoch": 0.7526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 3.6243620243773094e-06,
-      "loss": 0.8643,
+      "learning_rate": 3.040601433224155e-06,
+      "loss": 0.826,
       "step": 26524
     },
     {
-      "epoch": 0.7284886435417868,
+      "epoch": 0.7526958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 3.6236767686336536e-06,
-      "loss": 0.7909,
+      "learning_rate": 3.0399414725895417e-06,
+      "loss": 0.7486,
       "step": 26525
     },
     {
-      "epoch": 0.72851610776963,
+      "epoch": 0.7527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 3.6229915633419344e-06,
-      "loss": 0.8613,
+      "learning_rate": 3.039281570746855e-06,
+      "loss": 0.6954,
       "step": 26526
     },
     {
-      "epoch": 0.7285435719974733,
+      "epoch": 0.7527525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 3.6223064085075756e-06,
-      "loss": 0.8217,
+      "learning_rate": 3.0386217277016738e-06,
+      "loss": 0.9082,
       "step": 26527
     },
     {
-      "epoch": 0.7285710362253165,
+      "epoch": 0.7527809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 3.6216213041360025e-06,
-      "loss": 0.9196,
+      "learning_rate": 3.037961943459563e-06,
+      "loss": 0.8443,
       "step": 26528
     },
     {
-      "epoch": 0.7285985004531598,
+      "epoch": 0.7528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 3.6209362502326295e-06,
-      "loss": 0.8272,
+      "learning_rate": 3.0373022180261025e-06,
+      "loss": 0.8507,
       "step": 26529
     },
     {
-      "epoch": 0.728625964681003,
+      "epoch": 0.7528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 3.6202512468028795e-06,
-      "loss": 0.7979,
+      "learning_rate": 3.0366425514068652e-06,
+      "loss": 0.7796,
       "step": 26530
     },
     {
-      "epoch": 0.7286534289088462,
+      "epoch": 0.7528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 3.6195662938521736e-06,
-      "loss": 0.8068,
+      "learning_rate": 3.0359829436074182e-06,
+      "loss": 0.9148,
       "step": 26531
     },
     {
-      "epoch": 0.7286808931366895,
+      "epoch": 0.752894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.61888139138593e-06,
-      "loss": 0.8239,
+      "learning_rate": 3.035323394633336e-06,
+      "loss": 0.8362,
       "step": 26532
     },
     {
-      "epoch": 0.7287083573645327,
+      "epoch": 0.7529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 3.618196539409572e-06,
-      "loss": 0.8889,
+      "learning_rate": 3.034663904490194e-06,
+      "loss": 0.7689,
       "step": 26533
     },
     {
-      "epoch": 0.7287358215923759,
+      "epoch": 0.7529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 3.617511737928515e-06,
-      "loss": 0.8918,
+      "learning_rate": 3.0340044731835526e-06,
+      "loss": 0.8528,
       "step": 26534
     },
     {
-      "epoch": 0.7287632858202192,
+      "epoch": 0.7529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 3.6168269869481743e-06,
-      "loss": 0.8413,
+      "learning_rate": 3.033345100718992e-06,
+      "loss": 0.7626,
       "step": 26535
     },
     {
-      "epoch": 0.7287907500480624,
+      "epoch": 0.7530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 3.6161422864739715e-06,
-      "loss": 0.8565,
+      "learning_rate": 3.032685787102081e-06,
+      "loss": 0.7903,
       "step": 26536
     },
     {
-      "epoch": 0.7288182142759057,
+      "epoch": 0.7530363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 3.6154576365113237e-06,
-      "loss": 0.8326,
+      "learning_rate": 3.0320265323383834e-06,
+      "loss": 0.7753,
       "step": 26537
     },
     {
-      "epoch": 0.7288456785037488,
+      "epoch": 0.7530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 3.6147730370656476e-06,
-      "loss": 0.8104,
+      "learning_rate": 3.0313673364334716e-06,
+      "loss": 0.8571,
       "step": 26538
     },
     {
-      "epoch": 0.7288731427315921,
+      "epoch": 0.7530930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 3.614088488142361e-06,
-      "loss": 0.7988,
+      "learning_rate": 3.0307081993929155e-06,
+      "loss": 0.8191,
       "step": 26539
     },
     {
-      "epoch": 0.7289006069594354,
+      "epoch": 0.7531214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 3.613403989746883e-06,
-      "loss": 0.8458,
+      "learning_rate": 3.0300491212222784e-06,
+      "loss": 0.8346,
       "step": 26540
     },
     {
-      "epoch": 0.7289280711872785,
+      "epoch": 0.7531498297389331,
       "grad_norm": 0.0,
-      "learning_rate": 3.6127195418846274e-06,
-      "loss": 0.7067,
+      "learning_rate": 3.029390101927129e-06,
+      "loss": 0.788,
       "step": 26541
     },
     {
-      "epoch": 0.7289555354151218,
+      "epoch": 0.7531782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 3.6120351445610046e-06,
-      "loss": 0.7912,
+      "learning_rate": 3.028731141513037e-06,
+      "loss": 0.8139,
       "step": 26542
     },
     {
-      "epoch": 0.7289829996429651,
+      "epoch": 0.7532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 3.6113507977814353e-06,
-      "loss": 0.8203,
+      "learning_rate": 3.0280722399855644e-06,
+      "loss": 0.7508,
       "step": 26543
     },
     {
-      "epoch": 0.7290104638708083,
+      "epoch": 0.7532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 3.610666501551332e-06,
-      "loss": 0.8327,
+      "learning_rate": 3.0274133973502797e-06,
+      "loss": 0.873,
       "step": 26544
     },
     {
-      "epoch": 0.7290379280986515,
+      "epoch": 0.7532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.609982255876111e-06,
-      "loss": 0.8232,
+      "learning_rate": 3.026754613612749e-06,
+      "loss": 0.8699,
       "step": 26545
     },
     {
-      "epoch": 0.7290653923264947,
+      "epoch": 0.7532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 3.609298060761189e-06,
-      "loss": 0.9291,
+      "learning_rate": 3.026095888778533e-06,
+      "loss": 0.7159,
       "step": 26546
     },
     {
-      "epoch": 0.729092856554338,
+      "epoch": 0.7533200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 3.6086139162119727e-06,
-      "loss": 0.8421,
+      "learning_rate": 3.0254372228532015e-06,
+      "loss": 0.8666,
       "step": 26547
     },
     {
-      "epoch": 0.7291203207821813,
+      "epoch": 0.7533484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 3.60792982223388e-06,
-      "loss": 0.9169,
+      "learning_rate": 3.024778615842313e-06,
+      "loss": 0.8126,
       "step": 26548
     },
     {
-      "epoch": 0.7291477850100244,
+      "epoch": 0.7533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 3.6072457788323244e-06,
-      "loss": 0.7976,
+      "learning_rate": 3.0241200677514292e-06,
+      "loss": 0.8062,
       "step": 26549
     },
     {
-      "epoch": 0.7291752492378677,
+      "epoch": 0.753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 3.6065617860127135e-06,
-      "loss": 0.8344,
+      "learning_rate": 3.023461578586123e-06,
+      "loss": 0.8528,
       "step": 26550
     },
     {
-      "epoch": 0.7292027134657109,
+      "epoch": 0.753433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 3.605877843780463e-06,
-      "loss": 0.8636,
+      "learning_rate": 3.022803148351948e-06,
+      "loss": 0.7773,
       "step": 26551
     },
     {
-      "epoch": 0.7292301776935541,
+      "epoch": 0.7534619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 3.6051939521409864e-06,
-      "loss": 0.8642,
+      "learning_rate": 3.0221447770544676e-06,
+      "loss": 0.7812,
       "step": 26552
     },
     {
-      "epoch": 0.7292576419213974,
+      "epoch": 0.7534903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 3.6045101110996893e-06,
-      "loss": 0.7913,
+      "learning_rate": 3.021486464699248e-06,
+      "loss": 0.9339,
       "step": 26553
     },
     {
-      "epoch": 0.7292851061492406,
+      "epoch": 0.7535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 3.603826320661986e-06,
-      "loss": 0.7927,
+      "learning_rate": 3.0208282112918417e-06,
+      "loss": 0.8461,
       "step": 26554
     },
     {
-      "epoch": 0.7293125703770839,
+      "epoch": 0.7535471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 3.6031425808332853e-06,
-      "loss": 0.9324,
+      "learning_rate": 3.0201700168378147e-06,
+      "loss": 0.7745,
       "step": 26555
     },
     {
-      "epoch": 0.7293400346049271,
+      "epoch": 0.7535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 3.6024588916190027e-06,
-      "loss": 0.7837,
+      "learning_rate": 3.019511881342727e-06,
+      "loss": 0.9434,
       "step": 26556
     },
     {
-      "epoch": 0.7293674988327703,
+      "epoch": 0.7536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 3.601775253024541e-06,
-      "loss": 0.8184,
+      "learning_rate": 3.018853804812134e-06,
+      "loss": 0.8444,
       "step": 26557
     },
     {
-      "epoch": 0.7293949630606136,
+      "epoch": 0.7536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 3.601091665055314e-06,
-      "loss": 0.8723,
+      "learning_rate": 3.0181957872515966e-06,
+      "loss": 0.7816,
       "step": 26558
     },
     {
-      "epoch": 0.7294224272884567,
+      "epoch": 0.7536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 3.6004081277167256e-06,
-      "loss": 0.8408,
+      "learning_rate": 3.017537828666676e-06,
+      "loss": 0.7679,
       "step": 26559
     },
     {
-      "epoch": 0.7294498915163,
+      "epoch": 0.753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.599724641014186e-06,
-      "loss": 0.8607,
+      "learning_rate": 3.0168799290629246e-06,
+      "loss": 0.7303,
       "step": 26560
     },
     {
-      "epoch": 0.7294773557441433,
+      "epoch": 0.7537173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 3.5990412049531053e-06,
-      "loss": 0.8085,
+      "learning_rate": 3.0162220884459015e-06,
+      "loss": 0.7743,
       "step": 26561
     },
     {
-      "epoch": 0.7295048199719865,
+      "epoch": 0.7537457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 3.59835781953889e-06,
-      "loss": 0.7209,
+      "learning_rate": 3.0155643068211683e-06,
+      "loss": 0.8517,
       "step": 26562
     },
     {
-      "epoch": 0.7295322841998297,
+      "epoch": 0.7537741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 3.5976744847769474e-06,
-      "loss": 0.925,
+      "learning_rate": 3.0149065841942737e-06,
+      "loss": 0.729,
       "step": 26563
     },
     {
-      "epoch": 0.7295597484276729,
+      "epoch": 0.7538024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 3.596991200672688e-06,
-      "loss": 0.8538,
+      "learning_rate": 3.014248920570778e-06,
+      "loss": 0.8714,
       "step": 26564
     },
     {
-      "epoch": 0.7295872126555162,
+      "epoch": 0.7538308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 3.5963079672315136e-06,
-      "loss": 0.8798,
+      "learning_rate": 3.0135913159562347e-06,
+      "loss": 0.9473,
       "step": 26565
     },
     {
-      "epoch": 0.7296146768833595,
+      "epoch": 0.7538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 3.595624784458829e-06,
-      "loss": 0.8447,
+      "learning_rate": 3.0129337703562e-06,
+      "loss": 0.8128,
       "step": 26566
     },
     {
-      "epoch": 0.7296421411112026,
+      "epoch": 0.7538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 3.594941652360041e-06,
-      "loss": 0.7774,
+      "learning_rate": 3.012276283776231e-06,
+      "loss": 0.7259,
       "step": 26567
     },
     {
-      "epoch": 0.7296696053390459,
+      "epoch": 0.7539160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 3.5942585709405563e-06,
-      "loss": 0.9911,
+      "learning_rate": 3.0116188562218764e-06,
+      "loss": 0.8404,
       "step": 26568
     },
     {
-      "epoch": 0.7296970695668892,
+      "epoch": 0.75394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.593575540205778e-06,
-      "loss": 0.8378,
+      "learning_rate": 3.0109614876986893e-06,
+      "loss": 0.8521,
       "step": 26569
     },
     {
-      "epoch": 0.7297245337947323,
+      "epoch": 0.7539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 3.5928925601611163e-06,
-      "loss": 0.812,
+      "learning_rate": 3.0103041782122287e-06,
+      "loss": 0.8064,
       "step": 26570
     },
     {
-      "epoch": 0.7297519980225756,
+      "epoch": 0.7540011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 3.592209630811967e-06,
-      "loss": 0.9083,
+      "learning_rate": 3.009646927768041e-06,
+      "loss": 0.8928,
       "step": 26571
     },
     {
-      "epoch": 0.7297794622504188,
+      "epoch": 0.7540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 3.5915267521637364e-06,
-      "loss": 0.7896,
+      "learning_rate": 3.008989736371678e-06,
+      "loss": 0.8283,
       "step": 26572
     },
     {
-      "epoch": 0.7298069264782621,
+      "epoch": 0.7540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 3.5908439242218317e-06,
-      "loss": 0.7644,
+      "learning_rate": 3.0083326040286977e-06,
+      "loss": 0.8679,
       "step": 26573
     },
     {
-      "epoch": 0.7298343907061053,
+      "epoch": 0.7540862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 3.590161146991651e-06,
-      "loss": 0.8085,
+      "learning_rate": 3.0076755307446427e-06,
+      "loss": 0.8361,
       "step": 26574
     },
     {
-      "epoch": 0.7298618549339485,
+      "epoch": 0.7541146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 3.589478420478597e-06,
-      "loss": 0.8231,
+      "learning_rate": 3.0070185165250664e-06,
+      "loss": 0.8173,
       "step": 26575
     },
     {
-      "epoch": 0.7298893191617918,
+      "epoch": 0.7541430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 3.5887957446880763e-06,
-      "loss": 0.8665,
+      "learning_rate": 3.0063615613755214e-06,
+      "loss": 0.7095,
       "step": 26576
     },
     {
-      "epoch": 0.729916783389635,
+      "epoch": 0.7541713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 3.5881131196254837e-06,
-      "loss": 0.7785,
+      "learning_rate": 3.0057046653015532e-06,
+      "loss": 0.7834,
       "step": 26577
     },
     {
-      "epoch": 0.7299442476174782,
+      "epoch": 0.754199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 3.5874305452962234e-06,
-      "loss": 0.7753,
+      "learning_rate": 3.0050478283087116e-06,
+      "loss": 0.7218,
       "step": 26578
     },
     {
-      "epoch": 0.7299717118453215,
+      "epoch": 0.754228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 3.586748021705697e-06,
-      "loss": 0.9153,
+      "learning_rate": 3.0043910504025477e-06,
+      "loss": 0.77,
       "step": 26579
     },
     {
-      "epoch": 0.7299991760731647,
+      "epoch": 0.7542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 3.586065548859309e-06,
-      "loss": 0.8117,
+      "learning_rate": 3.0037343315886013e-06,
+      "loss": 0.7565,
       "step": 26580
     },
     {
-      "epoch": 0.730026640301008,
+      "epoch": 0.7542849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 3.5853831267624508e-06,
-      "loss": 0.8813,
+      "learning_rate": 3.0030776718724318e-06,
+      "loss": 0.849,
       "step": 26581
     },
     {
-      "epoch": 0.7300541045288512,
+      "epoch": 0.7543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 3.584700755420528e-06,
-      "loss": 0.7522,
+      "learning_rate": 3.0024210712595768e-06,
+      "loss": 0.8066,
       "step": 26582
     },
     {
-      "epoch": 0.7300815687566944,
+      "epoch": 0.7543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 3.5840184348389362e-06,
-      "loss": 0.8741,
+      "learning_rate": 3.0017645297555853e-06,
+      "loss": 0.8589,
       "step": 26583
     },
     {
-      "epoch": 0.7301090329845377,
+      "epoch": 0.7543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 3.583336165023076e-06,
-      "loss": 0.8203,
+      "learning_rate": 3.001108047366007e-06,
+      "loss": 0.8712,
       "step": 26584
     },
     {
-      "epoch": 0.7301364972123808,
+      "epoch": 0.7543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 3.582653945978345e-06,
-      "loss": 0.8126,
+      "learning_rate": 3.00045162409638e-06,
+      "loss": 0.8352,
       "step": 26585
     },
     {
-      "epoch": 0.7301639614402241,
+      "epoch": 0.7544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 3.5819717777101416e-06,
-      "loss": 0.8583,
+      "learning_rate": 2.999795259952253e-06,
+      "loss": 0.937,
       "step": 26586
     },
     {
-      "epoch": 0.7301914256680674,
+      "epoch": 0.7544551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 3.5812896602238633e-06,
-      "loss": 0.8875,
+      "learning_rate": 2.9991389549391735e-06,
+      "loss": 0.8007,
       "step": 26587
     },
     {
-      "epoch": 0.7302188898959106,
+      "epoch": 0.7544835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 3.580607593524912e-06,
-      "loss": 0.8065,
+      "learning_rate": 2.9984827090626788e-06,
+      "loss": 0.878,
       "step": 26588
     },
     {
-      "epoch": 0.7302463541237538,
+      "epoch": 0.7545119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 3.5799255776186802e-06,
-      "loss": 0.7778,
+      "learning_rate": 2.9978265223283152e-06,
+      "loss": 0.8002,
       "step": 26589
     },
     {
-      "epoch": 0.730273818351597,
+      "epoch": 0.7545402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 3.57924361251056e-06,
-      "loss": 0.8577,
+      "learning_rate": 2.9971703947416284e-06,
+      "loss": 0.6408,
       "step": 26590
     },
     {
-      "epoch": 0.7303012825794403,
+      "epoch": 0.7545686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 3.578561698205951e-06,
-      "loss": 0.8341,
+      "learning_rate": 2.9965143263081554e-06,
+      "loss": 0.8523,
       "step": 26591
     },
     {
-      "epoch": 0.7303287468072835,
+      "epoch": 0.7545970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 3.5778798347102507e-06,
-      "loss": 0.8914,
+      "learning_rate": 2.995858317033441e-06,
+      "loss": 0.8467,
       "step": 26592
     },
     {
-      "epoch": 0.7303562110351267,
+      "epoch": 0.7546254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 3.577198022028855e-06,
-      "loss": 0.7926,
+      "learning_rate": 2.995202366923029e-06,
+      "loss": 0.8453,
       "step": 26593
     },
     {
-      "epoch": 0.73038367526297,
+      "epoch": 0.7546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 3.576516260167153e-06,
-      "loss": 0.9359,
+      "learning_rate": 2.994546475982455e-06,
+      "loss": 0.7991,
       "step": 26594
     },
     {
-      "epoch": 0.7304111394908133,
+      "epoch": 0.7546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 3.5758345491305434e-06,
-      "loss": 0.8416,
+      "learning_rate": 2.993890644217261e-06,
+      "loss": 0.7405,
       "step": 26595
     },
     {
-      "epoch": 0.7304386037186564,
+      "epoch": 0.7547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.575152888924419e-06,
-      "loss": 0.7937,
+      "learning_rate": 2.9932348716329873e-06,
+      "loss": 0.9657,
       "step": 26596
     },
     {
-      "epoch": 0.7304660679464997,
+      "epoch": 0.754738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 3.5744712795541767e-06,
-      "loss": 0.8403,
+      "learning_rate": 2.992579158235175e-06,
+      "loss": 0.8401,
       "step": 26597
     },
     {
-      "epoch": 0.7304935321743429,
+      "epoch": 0.7547673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 3.573789721025205e-06,
-      "loss": 0.8011,
+      "learning_rate": 2.991923504029364e-06,
+      "loss": 0.7579,
       "step": 26598
     },
     {
-      "epoch": 0.7305209964021862,
+      "epoch": 0.7547956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 3.5731082133428995e-06,
-      "loss": 0.8755,
+      "learning_rate": 2.9912679090210873e-06,
+      "loss": 0.9191,
       "step": 26599
     },
     {
-      "epoch": 0.7305484606300294,
+      "epoch": 0.7548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 3.5724267565126493e-06,
-      "loss": 0.8642,
+      "learning_rate": 2.9906123732158842e-06,
+      "loss": 0.8368,
       "step": 26600
     },
     {
-      "epoch": 0.7305759248578726,
+      "epoch": 0.7548524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 3.571745350539848e-06,
-      "loss": 0.8456,
+      "learning_rate": 2.989956896619298e-06,
+      "loss": 0.9034,
       "step": 26601
     },
     {
-      "epoch": 0.7306033890857159,
+      "epoch": 0.7548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 3.5710639954298874e-06,
-      "loss": 0.8503,
+      "learning_rate": 2.989301479236858e-06,
+      "loss": 0.6929,
       "step": 26602
     },
     {
-      "epoch": 0.730630853313559,
+      "epoch": 0.7549091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 3.5703826911881588e-06,
-      "loss": 0.8109,
+      "learning_rate": 2.9886461210741035e-06,
+      "loss": 0.8142,
       "step": 26603
     },
     {
-      "epoch": 0.7306583175414023,
+      "epoch": 0.7549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 3.5697014378200566e-06,
-      "loss": 0.9419,
+      "learning_rate": 2.9879908221365726e-06,
+      "loss": 0.8134,
       "step": 26604
     },
     {
-      "epoch": 0.7306857817692456,
+      "epoch": 0.7549659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 3.5690202353309644e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.9873355824297967e-06,
+      "loss": 0.8935,
       "step": 26605
     },
     {
-      "epoch": 0.7307132459970888,
+      "epoch": 0.754994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 3.5683390837262776e-06,
-      "loss": 0.8709,
+      "learning_rate": 2.986680401959311e-06,
+      "loss": 0.8355,
       "step": 26606
     },
     {
-      "epoch": 0.730740710224932,
+      "epoch": 0.7550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 3.567657983011381e-06,
-      "loss": 0.6941,
+      "learning_rate": 2.9860252807306554e-06,
+      "loss": 0.8933,
       "step": 26607
     },
     {
-      "epoch": 0.7307681744527753,
+      "epoch": 0.7550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 3.566976933191665e-06,
-      "loss": 0.9007,
+      "learning_rate": 2.985370218749356e-06,
+      "loss": 0.7917,
       "step": 26608
     },
     {
-      "epoch": 0.7307956386806185,
+      "epoch": 0.7550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 3.56629593427252e-06,
-      "loss": 0.7956,
+      "learning_rate": 2.9847152160209494e-06,
+      "loss": 0.749,
       "step": 26609
     },
     {
-      "epoch": 0.7308231029084618,
+      "epoch": 0.7551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 3.5656149862593337e-06,
-      "loss": 0.851,
+      "learning_rate": 2.9840602725509727e-06,
+      "loss": 0.8614,
       "step": 26610
     },
     {
-      "epoch": 0.7308505671363049,
+      "epoch": 0.7551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 3.564934089157498e-06,
-      "loss": 0.8078,
+      "learning_rate": 2.9834053883449467e-06,
+      "loss": 0.8248,
       "step": 26611
     },
     {
-      "epoch": 0.7308780313641482,
+      "epoch": 0.7551645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 3.564253242972392e-06,
-      "loss": 0.8264,
+      "learning_rate": 2.9827505634084185e-06,
+      "loss": 0.7973,
       "step": 26612
     },
     {
-      "epoch": 0.7309054955919915,
+      "epoch": 0.7551929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 3.5635724477094114e-06,
-      "loss": 0.799,
+      "learning_rate": 2.982095797746908e-06,
+      "loss": 0.8084,
       "step": 26613
     },
     {
-      "epoch": 0.7309329598198346,
+      "epoch": 0.7552213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 3.562891703373936e-06,
-      "loss": 0.7449,
+      "learning_rate": 2.9814410913659496e-06,
+      "loss": 0.7967,
       "step": 26614
     },
     {
-      "epoch": 0.7309604240476779,
+      "epoch": 0.755249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 3.5622110099713536e-06,
-      "loss": 0.9326,
+      "learning_rate": 2.9807864442710767e-06,
+      "loss": 0.7759,
       "step": 26615
     },
     {
-      "epoch": 0.7309878882755211,
+      "epoch": 0.75527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 3.5615303675070524e-06,
-      "loss": 0.8386,
+      "learning_rate": 2.9801318564678138e-06,
+      "loss": 0.8961,
       "step": 26616
     },
     {
-      "epoch": 0.7310153525033644,
+      "epoch": 0.7553064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 3.5608497759864204e-06,
-      "loss": 0.7563,
+      "learning_rate": 2.9794773279616916e-06,
+      "loss": 0.7972,
       "step": 26617
     },
     {
-      "epoch": 0.7310428167312076,
+      "epoch": 0.7553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 3.5601692354148354e-06,
-      "loss": 0.8273,
+      "learning_rate": 2.9788228587582447e-06,
+      "loss": 0.7897,
       "step": 26618
     },
     {
-      "epoch": 0.7310702809590508,
+      "epoch": 0.7553632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 3.5594887457976866e-06,
-      "loss": 0.9016,
+      "learning_rate": 2.9781684488629923e-06,
+      "loss": 0.8122,
       "step": 26619
     },
     {
-      "epoch": 0.7310977451868941,
+      "epoch": 0.7553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 3.5588083071403566e-06,
-      "loss": 0.8051,
+      "learning_rate": 2.9775140982814678e-06,
+      "loss": 0.8501,
       "step": 26620
     },
     {
-      "epoch": 0.7311252094147374,
+      "epoch": 0.7554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 3.558127919448234e-06,
-      "loss": 0.8459,
+      "learning_rate": 2.976859807019199e-06,
+      "loss": 0.7623,
       "step": 26621
     },
     {
-      "epoch": 0.7311526736425805,
+      "epoch": 0.7554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 3.5574475827266963e-06,
-      "loss": 0.783,
+      "learning_rate": 2.9762055750817086e-06,
+      "loss": 0.8221,
       "step": 26622
     },
     {
-      "epoch": 0.7311801378704238,
+      "epoch": 0.7554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 3.5567672969811317e-06,
-      "loss": 0.8281,
+      "learning_rate": 2.9755514024745257e-06,
+      "loss": 0.798,
       "step": 26623
     },
     {
-      "epoch": 0.731207602098267,
+      "epoch": 0.7555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 3.5560870622169162e-06,
-      "loss": 0.7998,
+      "learning_rate": 2.9748972892031802e-06,
+      "loss": 0.8699,
       "step": 26624
     },
     {
-      "epoch": 0.7312350663261102,
+      "epoch": 0.7555334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 3.555406878439437e-06,
-      "loss": 0.9439,
+      "learning_rate": 2.9742432352731885e-06,
+      "loss": 0.837,
       "step": 26625
     },
     {
-      "epoch": 0.7312625305539535,
+      "epoch": 0.7555618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 3.5547267456540746e-06,
-      "loss": 0.8458,
+      "learning_rate": 2.9735892406900766e-06,
+      "loss": 0.7124,
       "step": 26626
     },
     {
-      "epoch": 0.7312899947817967,
+      "epoch": 0.7555902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 3.5540466638662108e-06,
-      "loss": 0.9107,
+      "learning_rate": 2.972935305459379e-06,
+      "loss": 0.8844,
       "step": 26627
     },
     {
-      "epoch": 0.73131745900964,
+      "epoch": 0.7556186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 3.55336663308123e-06,
-      "loss": 0.8459,
+      "learning_rate": 2.97228142958661e-06,
+      "loss": 0.9391,
       "step": 26628
     },
     {
-      "epoch": 0.7313449232374831,
+      "epoch": 0.7556469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 3.552686653304506e-06,
-      "loss": 0.8619,
+      "learning_rate": 2.9716276130772983e-06,
+      "loss": 0.8662,
       "step": 26629
     },
     {
-      "epoch": 0.7313723874653264,
+      "epoch": 0.7556753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 3.552006724541426e-06,
-      "loss": 0.8575,
+      "learning_rate": 2.9709738559369617e-06,
+      "loss": 0.8082,
       "step": 26630
     },
     {
-      "epoch": 0.7313998516931697,
+      "epoch": 0.7557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 3.551326846797363e-06,
-      "loss": 0.8525,
+      "learning_rate": 2.9703201581711238e-06,
+      "loss": 0.8642,
       "step": 26631
     },
     {
-      "epoch": 0.7314273159210128,
+      "epoch": 0.7557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 3.550647020077701e-06,
-      "loss": 0.9208,
+      "learning_rate": 2.9696665197853113e-06,
+      "loss": 0.8413,
       "step": 26632
     },
     {
-      "epoch": 0.7314547801488561,
+      "epoch": 0.7557604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 3.5499672443878162e-06,
-      "loss": 0.8383,
+      "learning_rate": 2.969012940785039e-06,
+      "loss": 0.7969,
       "step": 26633
     },
     {
-      "epoch": 0.7314822443766994,
+      "epoch": 0.755788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 3.549287519733091e-06,
-      "loss": 0.8447,
+      "learning_rate": 2.9683594211758304e-06,
+      "loss": 0.7462,
       "step": 26634
     },
     {
-      "epoch": 0.7315097086045426,
+      "epoch": 0.7558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 3.5486078461189045e-06,
-      "loss": 0.8568,
+      "learning_rate": 2.9677059609632086e-06,
+      "loss": 0.7733,
       "step": 26635
     },
     {
-      "epoch": 0.7315371728323858,
+      "epoch": 0.7558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 3.5479282235506273e-06,
-      "loss": 0.8609,
+      "learning_rate": 2.967052560152687e-06,
+      "loss": 0.785,
       "step": 26636
     },
     {
-      "epoch": 0.731564637060229,
+      "epoch": 0.7558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 3.547248652033646e-06,
-      "loss": 0.787,
+      "learning_rate": 2.9663992187497893e-06,
+      "loss": 0.7594,
       "step": 26637
     },
     {
-      "epoch": 0.7315921012880723,
+      "epoch": 0.7559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 3.5465691315733286e-06,
-      "loss": 0.7356,
+      "learning_rate": 2.965745936760037e-06,
+      "loss": 0.7417,
       "step": 26638
     },
     {
-      "epoch": 0.7316195655159156,
+      "epoch": 0.7559307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 3.5458896621750573e-06,
-      "loss": 0.9016,
+      "learning_rate": 2.9650927141889407e-06,
+      "loss": 0.7577,
       "step": 26639
     },
     {
-      "epoch": 0.7316470297437587,
+      "epoch": 0.7559591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 3.545210243844206e-06,
-      "loss": 0.7823,
+      "learning_rate": 2.9644395510420234e-06,
+      "loss": 0.8172,
       "step": 26640
     },
     {
-      "epoch": 0.731674493971602,
+      "epoch": 0.7559875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 3.544530876586155e-06,
-      "loss": 0.8437,
+      "learning_rate": 2.963786447324801e-06,
+      "loss": 0.7169,
       "step": 26641
     },
     {
-      "epoch": 0.7317019581994452,
+      "epoch": 0.7560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 3.5438515604062727e-06,
-      "loss": 0.7966,
+      "learning_rate": 2.9631334030427915e-06,
+      "loss": 0.77,
       "step": 26642
     },
     {
-      "epoch": 0.7317294224272884,
+      "epoch": 0.7560442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 3.543172295309939e-06,
-      "loss": 0.8427,
+      "learning_rate": 2.96248041820151e-06,
+      "loss": 0.8299,
       "step": 26643
     },
     {
-      "epoch": 0.7317568866551317,
+      "epoch": 0.7560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 3.542493081302526e-06,
-      "loss": 0.7118,
+      "learning_rate": 2.961827492806476e-06,
+      "loss": 0.7785,
       "step": 26644
     },
     {
-      "epoch": 0.7317843508829749,
+      "epoch": 0.7561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 3.541813918389413e-06,
-      "loss": 0.8105,
+      "learning_rate": 2.961174626863198e-06,
+      "loss": 0.7254,
       "step": 26645
     },
     {
-      "epoch": 0.7318118151108182,
+      "epoch": 0.7561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 3.5411348065759664e-06,
-      "loss": 0.734,
+      "learning_rate": 2.9605218203771946e-06,
+      "loss": 0.8017,
       "step": 26646
     },
     {
-      "epoch": 0.7318392793386614,
+      "epoch": 0.7561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 3.540455745867567e-06,
-      "loss": 0.8461,
+      "learning_rate": 2.9598690733539837e-06,
+      "loss": 0.8756,
       "step": 26647
     },
     {
-      "epoch": 0.7318667435665046,
+      "epoch": 0.7561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 3.53977673626958e-06,
-      "loss": 0.8446,
+      "learning_rate": 2.9592163857990706e-06,
+      "loss": 0.7329,
       "step": 26648
     },
     {
-      "epoch": 0.7318942077943479,
+      "epoch": 0.7562145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 3.5390977777873824e-06,
-      "loss": 0.8253,
+      "learning_rate": 2.958563757717975e-06,
+      "loss": 0.774,
       "step": 26649
     },
     {
-      "epoch": 0.731921672022191,
+      "epoch": 0.7562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 3.5384188704263455e-06,
-      "loss": 0.7158,
+      "learning_rate": 2.9579111891162094e-06,
+      "loss": 0.8613,
       "step": 26650
     },
     {
-      "epoch": 0.7319491362500343,
+      "epoch": 0.7562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 3.5377400141918417e-06,
-      "loss": 0.7762,
+      "learning_rate": 2.957258679999282e-06,
+      "loss": 0.9105,
       "step": 26651
     },
     {
-      "epoch": 0.7319766004778776,
+      "epoch": 0.756299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 3.5370612090892453e-06,
-      "loss": 0.8679,
+      "learning_rate": 2.9566062303727073e-06,
+      "loss": 0.8603,
       "step": 26652
     },
     {
-      "epoch": 0.7320040647057208,
+      "epoch": 0.756328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 3.536382455123921e-06,
-      "loss": 0.8787,
+      "learning_rate": 2.9559538402419994e-06,
+      "loss": 0.8164,
       "step": 26653
     },
     {
-      "epoch": 0.732031528933564,
+      "epoch": 0.7563564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 3.535703752301246e-06,
-      "loss": 0.8615,
+      "learning_rate": 2.9553015096126638e-06,
+      "loss": 0.7772,
       "step": 26654
     },
     {
-      "epoch": 0.7320589931614072,
+      "epoch": 0.7563847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 3.5350251006265833e-06,
-      "loss": 0.8173,
+      "learning_rate": 2.954649238490215e-06,
+      "loss": 0.8115,
       "step": 26655
     },
     {
-      "epoch": 0.7320864573892505,
+      "epoch": 0.7564131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 3.5343465001053056e-06,
-      "loss": 0.7343,
+      "learning_rate": 2.9539970268801575e-06,
+      "loss": 0.8378,
       "step": 26656
     },
     {
-      "epoch": 0.7321139216170938,
+      "epoch": 0.7564415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 3.533667950742784e-06,
-      "loss": 0.844,
+      "learning_rate": 2.953344874788001e-06,
+      "loss": 0.7207,
       "step": 26657
     },
     {
-      "epoch": 0.7321413858449369,
+      "epoch": 0.7564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 3.53298945254439e-06,
-      "loss": 0.7767,
+      "learning_rate": 2.952692782219264e-06,
+      "loss": 0.8442,
       "step": 26658
     },
     {
-      "epoch": 0.7321688500727802,
+      "epoch": 0.7564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 3.532311005515484e-06,
-      "loss": 0.9543,
+      "learning_rate": 2.9520407491794445e-06,
+      "loss": 0.9066,
       "step": 26659
     },
     {
-      "epoch": 0.7321963143006235,
+      "epoch": 0.7565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 3.5316326096614394e-06,
-      "loss": 0.8421,
+      "learning_rate": 2.9513887756740534e-06,
+      "loss": 0.8249,
       "step": 26660
     },
     {
-      "epoch": 0.7322237785284667,
+      "epoch": 0.7565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 3.530954264987626e-06,
-      "loss": 0.8525,
+      "learning_rate": 2.950736861708601e-06,
+      "loss": 0.8587,
       "step": 26661
     },
     {
-      "epoch": 0.7322512427563099,
+      "epoch": 0.756583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 3.530275971499405e-06,
-      "loss": 0.8867,
+      "learning_rate": 2.950085007288589e-06,
+      "loss": 0.7887,
       "step": 26662
     },
     {
-      "epoch": 0.7322787069841531,
+      "epoch": 0.7566118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 3.5295977292021476e-06,
-      "loss": 0.8363,
+      "learning_rate": 2.9494332124195247e-06,
+      "loss": 0.8272,
       "step": 26663
     },
     {
-      "epoch": 0.7323061712119964,
+      "epoch": 0.7566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 3.528919538101222e-06,
-      "loss": 0.8354,
+      "learning_rate": 2.948781477106919e-06,
+      "loss": 0.8199,
       "step": 26664
     },
     {
-      "epoch": 0.7323336354398396,
+      "epoch": 0.7566685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 3.528241398201988e-06,
-      "loss": 0.8267,
+      "learning_rate": 2.94812980135627e-06,
+      "loss": 0.8889,
       "step": 26665
     },
     {
-      "epoch": 0.7323610996676828,
+      "epoch": 0.7566969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 3.527563309509816e-06,
-      "loss": 0.8477,
+      "learning_rate": 2.947478185173085e-06,
+      "loss": 0.8465,
       "step": 26666
     },
     {
-      "epoch": 0.7323885638955261,
+      "epoch": 0.7567253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 3.5268852720300697e-06,
-      "loss": 0.6863,
+      "learning_rate": 2.946826628562872e-06,
+      "loss": 0.7678,
       "step": 26667
     },
     {
-      "epoch": 0.7324160281233693,
+      "epoch": 0.7567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 3.526207285768114e-06,
-      "loss": 0.843,
+      "learning_rate": 2.946175131531128e-06,
+      "loss": 0.8724,
       "step": 26668
     },
     {
-      "epoch": 0.7324434923512125,
+      "epoch": 0.7567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 3.5255293507293177e-06,
-      "loss": 0.9133,
+      "learning_rate": 2.945523694083361e-06,
+      "loss": 0.9465,
       "step": 26669
     },
     {
-      "epoch": 0.7324709565790558,
+      "epoch": 0.7568104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 3.5248514669190416e-06,
-      "loss": 0.7728,
+      "learning_rate": 2.9448723162250748e-06,
+      "loss": 0.7409,
       "step": 26670
     },
     {
-      "epoch": 0.732498420806899,
+      "epoch": 0.756838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 3.524173634342646e-06,
-      "loss": 0.8661,
+      "learning_rate": 2.9442209979617665e-06,
+      "loss": 0.8098,
       "step": 26671
     },
     {
-      "epoch": 0.7325258850347423,
+      "epoch": 0.7568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 3.523495853005495e-06,
-      "loss": 0.8808,
+      "learning_rate": 2.9435697392989405e-06,
+      "loss": 0.8268,
       "step": 26672
     },
     {
-      "epoch": 0.7325533492625855,
+      "epoch": 0.7568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 3.5228181229129553e-06,
-      "loss": 0.7983,
+      "learning_rate": 2.942918540242098e-06,
+      "loss": 0.8738,
       "step": 26673
     },
     {
-      "epoch": 0.7325808134904287,
+      "epoch": 0.7569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 3.522140444070388e-06,
-      "loss": 0.8044,
+      "learning_rate": 2.94226740079674e-06,
+      "loss": 0.7948,
       "step": 26674
     },
     {
-      "epoch": 0.732608277718272,
+      "epoch": 0.7569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 3.5214628164831534e-06,
-      "loss": 0.7564,
+      "learning_rate": 2.9416163209683702e-06,
+      "loss": 0.8234,
       "step": 26675
     },
     {
-      "epoch": 0.7326357419461151,
+      "epoch": 0.7569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 3.520785240156619e-06,
-      "loss": 0.8364,
+      "learning_rate": 2.9409653007624806e-06,
+      "loss": 0.8896,
       "step": 26676
     },
     {
-      "epoch": 0.7326632061739584,
+      "epoch": 0.7570090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 3.5201077150961405e-06,
-      "loss": 0.8153,
+      "learning_rate": 2.940314340184575e-06,
+      "loss": 0.8551,
       "step": 26677
     },
     {
-      "epoch": 0.7326906704018017,
+      "epoch": 0.7570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 3.519430241307078e-06,
-      "loss": 0.7477,
+      "learning_rate": 2.9396634392401534e-06,
+      "loss": 0.7886,
       "step": 26678
     },
     {
-      "epoch": 0.7327181346296449,
+      "epoch": 0.7570658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 3.518752818794793e-06,
-      "loss": 0.8728,
+      "learning_rate": 2.9390125979347106e-06,
+      "loss": 0.7983,
       "step": 26679
     },
     {
-      "epoch": 0.7327455988574881,
+      "epoch": 0.757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 3.518075447564646e-06,
-      "loss": 0.8461,
+      "learning_rate": 2.938361816273745e-06,
+      "loss": 0.7544,
       "step": 26680
     },
     {
-      "epoch": 0.7327730630853314,
+      "epoch": 0.757122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 3.517398127621997e-06,
-      "loss": 0.6843,
+      "learning_rate": 2.9377110942627573e-06,
+      "loss": 0.8497,
       "step": 26681
     },
     {
-      "epoch": 0.7328005273131746,
+      "epoch": 0.7571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 3.516720858972208e-06,
-      "loss": 0.9727,
+      "learning_rate": 2.937060431907239e-06,
+      "loss": 0.8662,
       "step": 26682
     },
     {
-      "epoch": 0.7328279915410179,
+      "epoch": 0.7571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 3.516043641620632e-06,
-      "loss": 0.7974,
+      "learning_rate": 2.9364098292126886e-06,
+      "loss": 0.7526,
       "step": 26683
     },
     {
-      "epoch": 0.732855455768861,
+      "epoch": 0.7572077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 3.5153664755726314e-06,
-      "loss": 0.79,
+      "learning_rate": 2.935759286184605e-06,
+      "loss": 0.8644,
       "step": 26684
     },
     {
-      "epoch": 0.7328829199967043,
+      "epoch": 0.7572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 3.514689360833565e-06,
-      "loss": 0.9038,
+      "learning_rate": 2.935108802828478e-06,
+      "loss": 0.8076,
       "step": 26685
     },
     {
-      "epoch": 0.7329103842245476,
+      "epoch": 0.7572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 3.5140122974087863e-06,
-      "loss": 0.784,
+      "learning_rate": 2.9344583791498028e-06,
+      "loss": 0.9047,
       "step": 26686
     },
     {
-      "epoch": 0.7329378484523907,
+      "epoch": 0.7572928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 3.513335285303654e-06,
-      "loss": 0.8376,
+      "learning_rate": 2.933808015154077e-06,
+      "loss": 0.8126,
       "step": 26687
     },
     {
-      "epoch": 0.732965312680234,
+      "epoch": 0.7573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 3.512658324523529e-06,
-      "loss": 0.8394,
+      "learning_rate": 2.9331577108467914e-06,
+      "loss": 0.8209,
       "step": 26688
     },
     {
-      "epoch": 0.7329927769080772,
+      "epoch": 0.7573496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 3.5119814150737618e-06,
-      "loss": 0.8004,
+      "learning_rate": 2.932507466233444e-06,
+      "loss": 0.8206,
       "step": 26689
     },
     {
-      "epoch": 0.7330202411359205,
+      "epoch": 0.7573779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 3.5113045569597105e-06,
-      "loss": 0.7879,
+      "learning_rate": 2.93185728131952e-06,
+      "loss": 0.7816,
       "step": 26690
     },
     {
-      "epoch": 0.7330477053637637,
+      "epoch": 0.7574063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 3.5106277501867305e-06,
-      "loss": 0.8318,
+      "learning_rate": 2.931207156110516e-06,
+      "loss": 0.7133,
       "step": 26691
     },
     {
-      "epoch": 0.7330751695916069,
+      "epoch": 0.7574347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 3.5099509947601775e-06,
-      "loss": 0.8331,
+      "learning_rate": 2.9305570906119253e-06,
+      "loss": 0.8481,
       "step": 26692
     },
     {
-      "epoch": 0.7331026338194502,
+      "epoch": 0.7574631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 3.5092742906854095e-06,
-      "loss": 0.8788,
+      "learning_rate": 2.929907084829234e-06,
+      "loss": 0.8988,
       "step": 26693
     },
     {
-      "epoch": 0.7331300980472935,
+      "epoch": 0.7574914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 3.508597637967778e-06,
-      "loss": 0.8561,
+      "learning_rate": 2.929257138767936e-06,
+      "loss": 0.823,
       "step": 26694
     },
     {
-      "epoch": 0.7331575622751366,
+      "epoch": 0.7575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 3.5079210366126327e-06,
-      "loss": 0.8639,
+      "learning_rate": 2.9286072524335227e-06,
+      "loss": 0.8478,
       "step": 26695
     },
     {
-      "epoch": 0.7331850265029799,
+      "epoch": 0.7575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 3.507244486625332e-06,
-      "loss": 0.8902,
+      "learning_rate": 2.927957425831479e-06,
+      "loss": 0.8542,
       "step": 26696
     },
     {
-      "epoch": 0.7332124907308231,
+      "epoch": 0.7575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 3.506567988011227e-06,
-      "loss": 0.8328,
+      "learning_rate": 2.9273076589672976e-06,
+      "loss": 0.6804,
       "step": 26697
     },
     {
-      "epoch": 0.7332399549586663,
+      "epoch": 0.7576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 3.505891540775672e-06,
-      "loss": 0.7013,
+      "learning_rate": 2.9266579518464687e-06,
+      "loss": 0.7722,
       "step": 26698
     },
     {
-      "epoch": 0.7332674191865096,
+      "epoch": 0.757633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.505215144924018e-06,
-      "loss": 0.8475,
+      "learning_rate": 2.926008304474475e-06,
+      "loss": 0.8291,
       "step": 26699
     },
     {
-      "epoch": 0.7332948834143528,
+      "epoch": 0.7576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 3.504538800461621e-06,
-      "loss": 0.8851,
+      "learning_rate": 2.9253587168568074e-06,
+      "loss": 0.8245,
       "step": 26700
     },
     {
-      "epoch": 0.7333223476421961,
+      "epoch": 0.7576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 3.5038625073938306e-06,
-      "loss": 0.8469,
+      "learning_rate": 2.9247091889989555e-06,
+      "loss": 0.7778,
       "step": 26701
     },
     {
-      "epoch": 0.7333498118700392,
+      "epoch": 0.7577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 3.503186265725991e-06,
-      "loss": 0.8052,
+      "learning_rate": 2.9240597209064003e-06,
+      "loss": 0.8919,
       "step": 26702
     },
     {
-      "epoch": 0.7333772760978825,
+      "epoch": 0.7577468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 3.50251007546346e-06,
-      "loss": 0.8131,
+      "learning_rate": 2.9234103125846314e-06,
+      "loss": 0.7612,
       "step": 26703
     },
     {
-      "epoch": 0.7334047403257258,
+      "epoch": 0.7577752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 3.501833936611585e-06,
-      "loss": 0.932,
+      "learning_rate": 2.922760964039133e-06,
+      "loss": 0.8796,
       "step": 26704
     },
     {
-      "epoch": 0.733432204553569,
+      "epoch": 0.7578036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 3.501157849175717e-06,
-      "loss": 0.8469,
+      "learning_rate": 2.9221116752753908e-06,
+      "loss": 0.9133,
       "step": 26705
     },
     {
-      "epoch": 0.7334596687814122,
+      "epoch": 0.7578320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 3.5004818131612096e-06,
-      "loss": 0.9378,
+      "learning_rate": 2.9214624462988928e-06,
+      "loss": 0.8077,
       "step": 26706
     },
     {
-      "epoch": 0.7334871330092555,
+      "epoch": 0.7578603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 3.499805828573404e-06,
-      "loss": 0.7987,
+      "learning_rate": 2.9208132771151167e-06,
+      "loss": 0.8448,
       "step": 26707
     },
     {
-      "epoch": 0.7335145972370987,
+      "epoch": 0.7578887627695801,
       "grad_norm": 0.0,
-      "learning_rate": 3.4991298954176523e-06,
-      "loss": 0.8152,
+      "learning_rate": 2.920164167729548e-06,
+      "loss": 0.8994,
       "step": 26708
     },
     {
-      "epoch": 0.7335420614649419,
+      "epoch": 0.7579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 3.4984540136993074e-06,
-      "loss": 0.9219,
+      "learning_rate": 2.9195151181476757e-06,
+      "loss": 0.7884,
       "step": 26709
     },
     {
-      "epoch": 0.7335695256927851,
+      "epoch": 0.7579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 3.497778183423709e-06,
-      "loss": 0.8586,
+      "learning_rate": 2.918866128374973e-06,
+      "loss": 0.8693,
       "step": 26710
     },
     {
-      "epoch": 0.7335969899206284,
+      "epoch": 0.7579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 3.497102404596211e-06,
-      "loss": 0.8408,
+      "learning_rate": 2.918217198416927e-06,
+      "loss": 0.7912,
       "step": 26711
     },
     {
-      "epoch": 0.7336244541484717,
+      "epoch": 0.7580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 3.4964266772221558e-06,
-      "loss": 0.813,
+      "learning_rate": 2.9175683282790212e-06,
+      "loss": 0.84,
       "step": 26712
     },
     {
-      "epoch": 0.7336519183763148,
+      "epoch": 0.7580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 3.4957510013068907e-06,
-      "loss": 0.8387,
+      "learning_rate": 2.916919517966732e-06,
+      "loss": 0.8407,
       "step": 26713
     },
     {
-      "epoch": 0.7336793826041581,
+      "epoch": 0.7580590238365493,
       "grad_norm": 0.0,
-      "learning_rate": 3.495075376855763e-06,
-      "loss": 0.9152,
+      "learning_rate": 2.9162707674855416e-06,
+      "loss": 0.786,
       "step": 26714
     },
     {
-      "epoch": 0.7337068468320013,
+      "epoch": 0.7580874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 3.49439980387412e-06,
-      "loss": 0.8557,
+      "learning_rate": 2.9156220768409336e-06,
+      "loss": 0.7179,
       "step": 26715
     },
     {
-      "epoch": 0.7337343110598445,
+      "epoch": 0.7581157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 3.4937242823673077e-06,
-      "loss": 0.8183,
+      "learning_rate": 2.9149734460383804e-06,
+      "loss": 0.6374,
       "step": 26716
     },
     {
-      "epoch": 0.7337617752876878,
+      "epoch": 0.758144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 3.493048812340667e-06,
-      "loss": 0.8589,
+      "learning_rate": 2.914324875083366e-06,
+      "loss": 0.8856,
       "step": 26717
     },
     {
-      "epoch": 0.733789239515531,
+      "epoch": 0.758172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 3.4923733937995476e-06,
-      "loss": 0.8662,
+      "learning_rate": 2.913676363981368e-06,
+      "loss": 0.7894,
       "step": 26718
     },
     {
-      "epoch": 0.7338167037433743,
+      "epoch": 0.7582009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 3.4916980267492872e-06,
-      "loss": 0.9279,
+      "learning_rate": 2.9130279127378635e-06,
+      "loss": 0.7537,
       "step": 26719
     },
     {
-      "epoch": 0.7338441679712175,
+      "epoch": 0.7582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 3.4910227111952335e-06,
-      "loss": 0.7814,
+      "learning_rate": 2.912379521358335e-06,
+      "loss": 0.874,
       "step": 26720
     },
     {
-      "epoch": 0.7338716321990607,
+      "epoch": 0.7582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 3.490347447142729e-06,
-      "loss": 0.9135,
+      "learning_rate": 2.9117311898482514e-06,
+      "loss": 0.8847,
       "step": 26721
     },
     {
-      "epoch": 0.733899096426904,
+      "epoch": 0.7582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 3.489672234597118e-06,
-      "loss": 0.8268,
+      "learning_rate": 2.9110829182130928e-06,
+      "loss": 0.8794,
       "step": 26722
     },
     {
-      "epoch": 0.7339265606547472,
+      "epoch": 0.7583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 3.4889970735637447e-06,
-      "loss": 0.7951,
+      "learning_rate": 2.9104347064583393e-06,
+      "loss": 0.7425,
       "step": 26723
     },
     {
-      "epoch": 0.7339540248825904,
+      "epoch": 0.7583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 3.4883219640479448e-06,
-      "loss": 0.7937,
+      "learning_rate": 2.9097865545894598e-06,
+      "loss": 0.8505,
       "step": 26724
     },
     {
-      "epoch": 0.7339814891104337,
+      "epoch": 0.7583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 3.4876469060550677e-06,
-      "loss": 0.8497,
+      "learning_rate": 2.909138462611931e-06,
+      "loss": 0.9396,
       "step": 26725
     },
     {
-      "epoch": 0.7340089533382769,
+      "epoch": 0.7583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 3.4869718995904477e-06,
-      "loss": 0.7819,
+      "learning_rate": 2.9084904305312324e-06,
+      "loss": 0.8262,
       "step": 26726
     },
     {
-      "epoch": 0.7340364175661201,
+      "epoch": 0.758427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 3.48629694465943e-06,
-      "loss": 0.8097,
+      "learning_rate": 2.90784245835283e-06,
+      "loss": 0.8208,
       "step": 26727
     },
     {
-      "epoch": 0.7340638817939633,
+      "epoch": 0.7584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 3.4856220412673523e-06,
-      "loss": 0.8493,
+      "learning_rate": 2.9071945460822014e-06,
+      "loss": 0.8405,
       "step": 26728
     },
     {
-      "epoch": 0.7340913460218066,
+      "epoch": 0.7584846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 3.4849471894195607e-06,
-      "loss": 0.8219,
+      "learning_rate": 2.906546693724822e-06,
+      "loss": 0.8466,
       "step": 26729
     },
     {
-      "epoch": 0.7341188102496499,
+      "epoch": 0.7585130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 3.484272389121387e-06,
-      "loss": 0.8079,
+      "learning_rate": 2.905898901286159e-06,
+      "loss": 0.7706,
       "step": 26730
     },
     {
-      "epoch": 0.734146274477493,
+      "epoch": 0.7585414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 3.4835976403781734e-06,
-      "loss": 0.7619,
+      "learning_rate": 2.9052511687716867e-06,
+      "loss": 0.9373,
       "step": 26731
     },
     {
-      "epoch": 0.7341737387053363,
+      "epoch": 0.7585698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 3.4829229431952606e-06,
-      "loss": 0.785,
+      "learning_rate": 2.9046034961868797e-06,
+      "loss": 0.8537,
       "step": 26732
     },
     {
-      "epoch": 0.7342012029331796,
+      "epoch": 0.7585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 3.482248297577989e-06,
-      "loss": 0.7843,
+      "learning_rate": 2.903955883537201e-06,
+      "loss": 0.8042,
       "step": 26733
     },
     {
-      "epoch": 0.7342286671610228,
+      "epoch": 0.7586265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 3.48157370353169e-06,
-      "loss": 0.8497,
+      "learning_rate": 2.9033083308281273e-06,
+      "loss": 0.7321,
       "step": 26734
     },
     {
-      "epoch": 0.734256131388866,
+      "epoch": 0.7586549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 3.4808991610617083e-06,
-      "loss": 0.7993,
+      "learning_rate": 2.902660838065131e-06,
+      "loss": 0.7768,
       "step": 26735
     },
     {
-      "epoch": 0.7342835956167092,
+      "epoch": 0.758683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 3.4802246701733746e-06,
-      "loss": 0.8795,
+      "learning_rate": 2.9020134052536742e-06,
+      "loss": 0.8208,
       "step": 26736
     },
     {
-      "epoch": 0.7343110598445525,
+      "epoch": 0.7587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 3.4795502308720287e-06,
-      "loss": 0.7381,
+      "learning_rate": 2.901366032399232e-06,
+      "loss": 0.7687,
       "step": 26737
     },
     {
-      "epoch": 0.7343385240723957,
+      "epoch": 0.7587400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 3.4788758431630077e-06,
-      "loss": 0.8307,
+      "learning_rate": 2.9007187195072684e-06,
+      "loss": 0.8546,
       "step": 26738
     },
     {
-      "epoch": 0.7343659883002389,
+      "epoch": 0.7587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 3.478201507051646e-06,
-      "loss": 0.9292,
+      "learning_rate": 2.900071466583252e-06,
+      "loss": 0.8829,
       "step": 26739
     },
     {
-      "epoch": 0.7343934525280822,
+      "epoch": 0.7587968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 3.4775272225432844e-06,
-      "loss": 0.8157,
+      "learning_rate": 2.899424273632654e-06,
+      "loss": 0.8542,
       "step": 26740
     },
     {
-      "epoch": 0.7344209167559254,
+      "epoch": 0.7588251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 3.47685298964325e-06,
-      "loss": 0.8416,
+      "learning_rate": 2.8987771406609353e-06,
+      "loss": 0.8418,
       "step": 26741
     },
     {
-      "epoch": 0.7344483809837686,
+      "epoch": 0.7588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 3.476178808356886e-06,
-      "loss": 0.8429,
+      "learning_rate": 2.898130067673566e-06,
+      "loss": 0.7521,
       "step": 26742
     },
     {
-      "epoch": 0.7344758452116119,
+      "epoch": 0.7588819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 3.4755046786895173e-06,
-      "loss": 0.8336,
+      "learning_rate": 2.8974830546760135e-06,
+      "loss": 0.8436,
       "step": 26743
     },
     {
-      "epoch": 0.7345033094394551,
+      "epoch": 0.7589103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 3.4748306006464837e-06,
-      "loss": 0.8531,
+      "learning_rate": 2.896836101673738e-06,
+      "loss": 0.7698,
       "step": 26744
     },
     {
-      "epoch": 0.7345307736672984,
+      "epoch": 0.7589387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 3.4741565742331186e-06,
-      "loss": 0.8241,
+      "learning_rate": 2.8961892086722076e-06,
+      "loss": 0.7623,
       "step": 26745
     },
     {
-      "epoch": 0.7345582378951416,
+      "epoch": 0.7589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 3.4734825994547537e-06,
-      "loss": 0.7923,
+      "learning_rate": 2.8955423756768887e-06,
+      "loss": 0.8299,
       "step": 26746
     },
     {
-      "epoch": 0.7345857021229848,
+      "epoch": 0.7589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 3.472808676316727e-06,
-      "loss": 0.9312,
+      "learning_rate": 2.8948956026932405e-06,
+      "loss": 0.739,
       "step": 26747
     },
     {
-      "epoch": 0.7346131663508281,
+      "epoch": 0.7590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 3.472134804824363e-06,
-      "loss": 0.8995,
+      "learning_rate": 2.8942488897267284e-06,
+      "loss": 0.728,
       "step": 26748
     },
     {
-      "epoch": 0.7346406305786712,
+      "epoch": 0.7590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 3.4714609849829996e-06,
-      "loss": 0.729,
+      "learning_rate": 2.8936022367828153e-06,
+      "loss": 0.7667,
       "step": 26749
     },
     {
-      "epoch": 0.7346680948065145,
+      "epoch": 0.7590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 3.4707872167979627e-06,
-      "loss": 0.9248,
+      "learning_rate": 2.8929556438669626e-06,
+      "loss": 0.7483,
       "step": 26750
     },
     {
-      "epoch": 0.7346955590343578,
+      "epoch": 0.7591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 3.4701135002745868e-06,
-      "loss": 0.8871,
+      "learning_rate": 2.8923091109846348e-06,
+      "loss": 0.7774,
       "step": 26751
     },
     {
-      "epoch": 0.734723023262201,
+      "epoch": 0.7591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 3.4694398354182026e-06,
-      "loss": 0.813,
+      "learning_rate": 2.8916626381412927e-06,
+      "loss": 0.7493,
       "step": 26752
     },
     {
-      "epoch": 0.7347504874900442,
+      "epoch": 0.7591657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 3.468766222234142e-06,
-      "loss": 0.8523,
+      "learning_rate": 2.8910162253423947e-06,
+      "loss": 0.7646,
       "step": 26753
     },
     {
-      "epoch": 0.7347779517178874,
+      "epoch": 0.759194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 3.468092660727731e-06,
-      "loss": 0.7215,
+      "learning_rate": 2.8903698725934002e-06,
+      "loss": 0.8743,
       "step": 26754
     },
     {
-      "epoch": 0.7348054159457307,
+      "epoch": 0.7592224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 3.467419150904301e-06,
-      "loss": 0.7803,
+      "learning_rate": 2.8897235798997757e-06,
+      "loss": 0.7899,
       "step": 26755
     },
     {
-      "epoch": 0.734832880173574,
+      "epoch": 0.7592508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 3.4667456927691812e-06,
-      "loss": 0.8173,
+      "learning_rate": 2.889077347266972e-06,
+      "loss": 0.7851,
       "step": 26756
     },
     {
-      "epoch": 0.7348603444014171,
+      "epoch": 0.7592792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 3.466072286327703e-06,
-      "loss": 0.9049,
+      "learning_rate": 2.8884311747004513e-06,
+      "loss": 0.827,
       "step": 26757
     },
     {
-      "epoch": 0.7348878086292604,
+      "epoch": 0.7593076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 3.4653989315851887e-06,
-      "loss": 0.8717,
+      "learning_rate": 2.8877850622056756e-06,
+      "loss": 0.7856,
       "step": 26758
     },
     {
-      "epoch": 0.7349152728571037,
+      "epoch": 0.7593359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 3.4647256285469722e-06,
-      "loss": 0.7307,
+      "learning_rate": 2.8871390097880958e-06,
+      "loss": 0.8297,
       "step": 26759
     },
     {
-      "epoch": 0.7349427370849468,
+      "epoch": 0.7593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 3.464052377218375e-06,
-      "loss": 0.8419,
+      "learning_rate": 2.8864930174531726e-06,
+      "loss": 0.8243,
       "step": 26760
     },
     {
-      "epoch": 0.7349702013127901,
+      "epoch": 0.7593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 3.4633791776047276e-06,
-      "loss": 0.7922,
+      "learning_rate": 2.885847085206366e-06,
+      "loss": 0.7756,
       "step": 26761
     },
     {
-      "epoch": 0.7349976655406333,
+      "epoch": 0.7594211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 3.462706029711357e-06,
-      "loss": 0.805,
+      "learning_rate": 2.885201213053126e-06,
+      "loss": 0.8946,
       "step": 26762
     },
     {
-      "epoch": 0.7350251297684766,
+      "epoch": 0.7594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 3.4620329335435876e-06,
-      "loss": 0.8042,
+      "learning_rate": 2.884555400998914e-06,
+      "loss": 0.9069,
       "step": 26763
     },
     {
-      "epoch": 0.7350525939963198,
+      "epoch": 0.759477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 3.461359889106749e-06,
-      "loss": 0.7429,
+      "learning_rate": 2.883909649049176e-06,
+      "loss": 0.8457,
       "step": 26764
     },
     {
-      "epoch": 0.735080058224163,
+      "epoch": 0.7595062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 3.4606868964061613e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.883263957209377e-06,
+      "loss": 0.8186,
       "step": 26765
     },
     {
-      "epoch": 0.7351075224520063,
+      "epoch": 0.7595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 3.4600139554471544e-06,
-      "loss": 0.7765,
+      "learning_rate": 2.882618325484969e-06,
+      "loss": 0.9751,
       "step": 26766
     },
     {
-      "epoch": 0.7351349866798494,
+      "epoch": 0.7595629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 3.459341066235048e-06,
-      "loss": 0.8112,
+      "learning_rate": 2.8819727538814024e-06,
+      "loss": 0.8305,
       "step": 26767
     },
     {
-      "epoch": 0.7351624509076927,
+      "epoch": 0.7595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 3.458668228775168e-06,
-      "loss": 0.7631,
+      "learning_rate": 2.8813272424041306e-06,
+      "loss": 0.8285,
       "step": 26768
     },
     {
-      "epoch": 0.735189915135536,
+      "epoch": 0.7596197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 3.45799544307284e-06,
-      "loss": 0.8579,
+      "learning_rate": 2.8806817910586116e-06,
+      "loss": 0.8705,
       "step": 26769
     },
     {
-      "epoch": 0.7352173793633792,
+      "epoch": 0.7596481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 3.4573227091333882e-06,
-      "loss": 0.7535,
+      "learning_rate": 2.88003639985029e-06,
+      "loss": 0.7217,
       "step": 26770
     },
     {
-      "epoch": 0.7352448435912224,
+      "epoch": 0.7596765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 3.4566500269621304e-06,
-      "loss": 0.9587,
+      "learning_rate": 2.8793910687846203e-06,
+      "loss": 0.7811,
       "step": 26771
     },
     {
-      "epoch": 0.7352723078190657,
+      "epoch": 0.7597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 3.4559773965643926e-06,
-      "loss": 0.8267,
+      "learning_rate": 2.8787457978670586e-06,
+      "loss": 0.8237,
       "step": 26772
     },
     {
-      "epoch": 0.7352997720469089,
+      "epoch": 0.759733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 3.4553048179455007e-06,
-      "loss": 0.7497,
+      "learning_rate": 2.8781005871030476e-06,
+      "loss": 0.8383,
       "step": 26773
     },
     {
-      "epoch": 0.7353272362747522,
+      "epoch": 0.7597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 3.4546322911107676e-06,
-      "loss": 0.7778,
+      "learning_rate": 2.877455436498041e-06,
+      "loss": 0.8618,
       "step": 26774
     },
     {
-      "epoch": 0.7353547005025953,
+      "epoch": 0.7597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 3.4539598160655195e-06,
-      "loss": 0.8755,
+      "learning_rate": 2.8768103460574925e-06,
+      "loss": 0.776,
       "step": 26775
     },
     {
-      "epoch": 0.7353821647304386,
+      "epoch": 0.7598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 3.4532873928150812e-06,
-      "loss": 0.8286,
+      "learning_rate": 2.8761653157868442e-06,
+      "loss": 0.7675,
       "step": 26776
     },
     {
-      "epoch": 0.7354096289582819,
+      "epoch": 0.7598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 3.4526150213647645e-06,
-      "loss": 0.8561,
+      "learning_rate": 2.875520345691546e-06,
+      "loss": 0.7993,
       "step": 26777
     },
     {
-      "epoch": 0.735437093186125,
+      "epoch": 0.7598751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 3.4519427017198947e-06,
-      "loss": 0.7677,
+      "learning_rate": 2.8748754357770525e-06,
+      "loss": 0.8799,
       "step": 26778
     },
     {
-      "epoch": 0.7354645574139683,
+      "epoch": 0.7599035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 3.4512704338857907e-06,
-      "loss": 0.7279,
+      "learning_rate": 2.8742305860487994e-06,
+      "loss": 0.778,
       "step": 26779
     },
     {
-      "epoch": 0.7354920216418115,
+      "epoch": 0.7599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 3.450598217867772e-06,
-      "loss": 0.7983,
+      "learning_rate": 2.873585796512247e-06,
+      "loss": 0.8336,
       "step": 26780
     },
     {
-      "epoch": 0.7355194858696548,
+      "epoch": 0.7599602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 3.44992605367116e-06,
-      "loss": 0.8922,
+      "learning_rate": 2.872941067172833e-06,
+      "loss": 0.7548,
       "step": 26781
     },
     {
-      "epoch": 0.735546950097498,
+      "epoch": 0.759988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 3.4492539413012704e-06,
-      "loss": 0.8379,
+      "learning_rate": 2.8722963980360064e-06,
+      "loss": 0.7872,
       "step": 26782
     },
     {
-      "epoch": 0.7355744143253412,
+      "epoch": 0.760017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 3.4485818807634175e-06,
-      "loss": 0.965,
+      "learning_rate": 2.8716517891072148e-06,
+      "loss": 0.8295,
       "step": 26783
     },
     {
-      "epoch": 0.7356018785531845,
+      "epoch": 0.7600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 3.4479098720629234e-06,
-      "loss": 0.8053,
+      "learning_rate": 2.8710072403918986e-06,
+      "loss": 0.7719,
       "step": 26784
     },
     {
-      "epoch": 0.7356293427810278,
+      "epoch": 0.7600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 3.4472379152051037e-06,
-      "loss": 0.7348,
+      "learning_rate": 2.8703627518955046e-06,
+      "loss": 0.8844,
       "step": 26785
     },
     {
-      "epoch": 0.7356568070088709,
+      "epoch": 0.7601021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 3.4465660101952757e-06,
-      "loss": 0.8371,
+      "learning_rate": 2.8697183236234805e-06,
+      "loss": 0.6954,
       "step": 26786
     },
     {
-      "epoch": 0.7356842712367142,
+      "epoch": 0.7601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 3.445894157038757e-06,
-      "loss": 0.739,
+      "learning_rate": 2.869073955581262e-06,
+      "loss": 0.7847,
       "step": 26787
     },
     {
-      "epoch": 0.7357117354645574,
+      "epoch": 0.7601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 3.445222355740865e-06,
-      "loss": 0.7808,
+      "learning_rate": 2.868429647774297e-06,
+      "loss": 0.8604,
       "step": 26788
     },
     {
-      "epoch": 0.7357391996924006,
+      "epoch": 0.7601872871736662,
       "grad_norm": 0.0,
-      "learning_rate": 3.44455060630691e-06,
-      "loss": 0.8383,
+      "learning_rate": 2.8677854002080307e-06,
+      "loss": 0.7894,
       "step": 26789
     },
     {
-      "epoch": 0.7357666639202439,
+      "epoch": 0.7602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 3.443878908742213e-06,
-      "loss": 0.9604,
+      "learning_rate": 2.8671412128878985e-06,
+      "loss": 0.8091,
       "step": 26790
     },
     {
-      "epoch": 0.7357941281480871,
+      "epoch": 0.7602440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 3.443207263052082e-06,
-      "loss": 0.7581,
+      "learning_rate": 2.866497085819344e-06,
+      "loss": 0.8536,
       "step": 26791
     },
     {
-      "epoch": 0.7358215923759304,
+      "epoch": 0.7602724177071509,
       "grad_norm": 0.0,
-      "learning_rate": 3.4425356692418353e-06,
-      "loss": 0.9066,
+      "learning_rate": 2.8658530190078138e-06,
+      "loss": 0.8045,
       "step": 26792
     },
     {
-      "epoch": 0.7358490566037735,
+      "epoch": 0.7603007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 3.4418641273167874e-06,
-      "loss": 0.7249,
+      "learning_rate": 2.8652090124587405e-06,
+      "loss": 0.7756,
       "step": 26793
     },
     {
-      "epoch": 0.7358765208316168,
+      "epoch": 0.7603291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 3.4411926372822535e-06,
-      "loss": 0.8276,
+      "learning_rate": 2.864565066177567e-06,
+      "loss": 0.7658,
       "step": 26794
     },
     {
-      "epoch": 0.7359039850594601,
+      "epoch": 0.7603575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 3.4405211991435406e-06,
-      "loss": 0.7352,
+      "learning_rate": 2.863921180169733e-06,
+      "loss": 0.8151,
       "step": 26795
     },
     {
-      "epoch": 0.7359314492873033,
+      "epoch": 0.7603859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 3.4398498129059644e-06,
-      "loss": 0.8668,
+      "learning_rate": 2.863277354440679e-06,
+      "loss": 0.8162,
       "step": 26796
     },
     {
-      "epoch": 0.7359589135151465,
+      "epoch": 0.7604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 3.4391784785748416e-06,
-      "loss": 0.9275,
+      "learning_rate": 2.8626335889958433e-06,
+      "loss": 0.8803,
       "step": 26797
     },
     {
-      "epoch": 0.7359863777429898,
+      "epoch": 0.7604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 3.438507196155477e-06,
-      "loss": 0.7264,
+      "learning_rate": 2.86198988384066e-06,
+      "loss": 0.9233,
       "step": 26798
     },
     {
-      "epoch": 0.736013841970833,
+      "epoch": 0.7604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 3.4378359656531846e-06,
-      "loss": 0.8134,
+      "learning_rate": 2.86134623898057e-06,
+      "loss": 0.7732,
       "step": 26799
     },
     {
-      "epoch": 0.7360413061986762,
+      "epoch": 0.7604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 3.4371647870732795e-06,
-      "loss": 0.8156,
+      "learning_rate": 2.8607026544210115e-06,
+      "loss": 0.8679,
       "step": 26800
     },
     {
-      "epoch": 0.7360687704265194,
+      "epoch": 0.760527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 3.436493660421065e-06,
-      "loss": 0.8675,
+      "learning_rate": 2.8600591301674153e-06,
+      "loss": 0.7973,
       "step": 26801
     },
     {
-      "epoch": 0.7360962346543627,
+      "epoch": 0.7605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 3.4358225857018556e-06,
-      "loss": 0.8591,
+      "learning_rate": 2.8594156662252205e-06,
+      "loss": 0.8046,
       "step": 26802
     },
     {
-      "epoch": 0.736123698882206,
+      "epoch": 0.7605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 3.4351515629209607e-06,
-      "loss": 0.7423,
+      "learning_rate": 2.858772262599866e-06,
+      "loss": 0.8687,
       "step": 26803
     },
     {
-      "epoch": 0.7361511631100491,
+      "epoch": 0.7606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 3.43448059208369e-06,
-      "loss": 0.8318,
+      "learning_rate": 2.858128919296781e-06,
+      "loss": 0.7932,
       "step": 26804
     },
     {
-      "epoch": 0.7361786273378924,
+      "epoch": 0.7606413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 3.4338096731953553e-06,
-      "loss": 0.7816,
+      "learning_rate": 2.857485636321401e-06,
+      "loss": 0.8303,
       "step": 26805
     },
     {
-      "epoch": 0.7362060915657356,
+      "epoch": 0.7606696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 3.4331388062612615e-06,
-      "loss": 0.8635,
+      "learning_rate": 2.856842413679164e-06,
+      "loss": 0.7875,
       "step": 26806
     },
     {
-      "epoch": 0.7362335557935789,
+      "epoch": 0.7606980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 3.432467991286714e-06,
-      "loss": 0.7907,
+      "learning_rate": 2.8561992513754967e-06,
+      "loss": 0.8057,
       "step": 26807
     },
     {
-      "epoch": 0.7362610200214221,
+      "epoch": 0.7607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 3.431797228277024e-06,
-      "loss": 0.8443,
+      "learning_rate": 2.8555561494158367e-06,
+      "loss": 0.844,
       "step": 26808
     },
     {
-      "epoch": 0.7362884842492653,
+      "epoch": 0.7607548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 3.4311265172374987e-06,
-      "loss": 0.7579,
+      "learning_rate": 2.854913107805617e-06,
+      "loss": 0.7656,
       "step": 26809
     },
     {
-      "epoch": 0.7363159484771086,
+      "epoch": 0.7607832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 3.4304558581734437e-06,
-      "loss": 0.8475,
+      "learning_rate": 2.8542701265502624e-06,
+      "loss": 0.949,
       "step": 26810
     },
     {
-      "epoch": 0.7363434127049518,
+      "epoch": 0.7608115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 3.4297852510901685e-06,
-      "loss": 0.7778,
+      "learning_rate": 2.853627205655215e-06,
+      "loss": 0.8012,
       "step": 26811
     },
     {
-      "epoch": 0.736370876932795,
+      "epoch": 0.7608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 3.42911469599298e-06,
-      "loss": 0.832,
+      "learning_rate": 2.8529843451258965e-06,
+      "loss": 0.7569,
       "step": 26812
     },
     {
-      "epoch": 0.7363983411606383,
+      "epoch": 0.7608683314415438,
       "grad_norm": 0.0,
-      "learning_rate": 3.428444192887177e-06,
-      "loss": 0.8168,
+      "learning_rate": 2.8523415449677404e-06,
+      "loss": 0.7701,
       "step": 26813
     },
     {
-      "epoch": 0.7364258053884815,
+      "epoch": 0.7608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 3.4277737417780743e-06,
-      "loss": 0.7603,
+      "learning_rate": 2.8516988051861803e-06,
+      "loss": 0.8393,
       "step": 26814
     },
     {
-      "epoch": 0.7364532696163247,
+      "epoch": 0.7609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 3.4271033426709664e-06,
-      "loss": 0.7289,
+      "learning_rate": 2.8510561257866375e-06,
+      "loss": 0.7722,
       "step": 26815
     },
     {
-      "epoch": 0.736480733844168,
+      "epoch": 0.7609534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 3.426432995571165e-06,
-      "loss": 0.7851,
+      "learning_rate": 2.8504135067745463e-06,
+      "loss": 0.8879,
       "step": 26816
     },
     {
-      "epoch": 0.7365081980720112,
+      "epoch": 0.7609818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 3.425762700483971e-06,
-      "loss": 0.8836,
+      "learning_rate": 2.849770948155336e-06,
+      "loss": 0.8947,
       "step": 26817
     },
     {
-      "epoch": 0.7365356622998545,
+      "epoch": 0.7610102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 3.4250924574146926e-06,
-      "loss": 0.7798,
+      "learning_rate": 2.8491284499344287e-06,
+      "loss": 0.8291,
       "step": 26818
     },
     {
-      "epoch": 0.7365631265276976,
+      "epoch": 0.761038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 3.4244222663686256e-06,
-      "loss": 0.8111,
+      "learning_rate": 2.848486012117255e-06,
+      "loss": 0.8482,
       "step": 26819
     },
     {
-      "epoch": 0.7365905907555409,
+      "epoch": 0.761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 3.4237521273510776e-06,
-      "loss": 0.8665,
+      "learning_rate": 2.8478436347092454e-06,
+      "loss": 0.8296,
       "step": 26820
     },
     {
-      "epoch": 0.7366180549833842,
+      "epoch": 0.7610953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 3.4230820403673526e-06,
-      "loss": 0.9445,
+      "learning_rate": 2.8472013177158174e-06,
+      "loss": 0.8719,
       "step": 26821
     },
     {
-      "epoch": 0.7366455192112273,
+      "epoch": 0.7611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 3.4224120054227485e-06,
-      "loss": 0.7381,
+      "learning_rate": 2.846559061142403e-06,
+      "loss": 0.8185,
       "step": 26822
     },
     {
-      "epoch": 0.7366729834390706,
+      "epoch": 0.7611520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 3.421742022522567e-06,
-      "loss": 0.7614,
+      "learning_rate": 2.845916864994428e-06,
+      "loss": 0.8835,
       "step": 26823
     },
     {
-      "epoch": 0.7367004476669139,
+      "epoch": 0.7611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 3.421072091672114e-06,
-      "loss": 0.8022,
+      "learning_rate": 2.845274729277312e-06,
+      "loss": 0.7951,
       "step": 26824
     },
     {
-      "epoch": 0.7367279118947571,
+      "epoch": 0.7612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 3.4204022128766845e-06,
-      "loss": 0.7138,
+      "learning_rate": 2.8446326539964818e-06,
+      "loss": 0.8084,
       "step": 26825
     },
     {
-      "epoch": 0.7367553761226003,
+      "epoch": 0.7612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 3.419732386141581e-06,
-      "loss": 0.886,
+      "learning_rate": 2.843990639157361e-06,
+      "loss": 0.8383,
       "step": 26826
     },
     {
-      "epoch": 0.7367828403504435,
+      "epoch": 0.7612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 3.419062611472105e-06,
-      "loss": 0.8409,
+      "learning_rate": 2.8433486847653734e-06,
+      "loss": 0.8974,
       "step": 26827
     },
     {
-      "epoch": 0.7368103045782868,
+      "epoch": 0.7612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 3.418392888873553e-06,
-      "loss": 0.8614,
+      "learning_rate": 2.842706790825944e-06,
+      "loss": 0.885,
       "step": 26828
     },
     {
-      "epoch": 0.7368377688061301,
+      "epoch": 0.761322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 3.41772321835123e-06,
-      "loss": 0.7925,
+      "learning_rate": 2.8420649573444893e-06,
+      "loss": 0.7747,
       "step": 26829
     },
     {
-      "epoch": 0.7368652330339732,
+      "epoch": 0.7613507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 3.4170535999104306e-06,
-      "loss": 0.9695,
+      "learning_rate": 2.841423184326434e-06,
+      "loss": 0.8879,
       "step": 26830
     },
     {
-      "epoch": 0.7368926972618165,
+      "epoch": 0.7613791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 3.4163840335564503e-06,
-      "loss": 0.8454,
+      "learning_rate": 2.840781471777201e-06,
+      "loss": 0.7713,
       "step": 26831
     },
     {
-      "epoch": 0.7369201614896597,
+      "epoch": 0.7614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 3.4157145192945896e-06,
-      "loss": 0.8345,
+      "learning_rate": 2.8401398197022067e-06,
+      "loss": 0.7995,
       "step": 26832
     },
     {
-      "epoch": 0.7369476257175029,
+      "epoch": 0.7614358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 3.415045057130145e-06,
-      "loss": 0.8932,
+      "learning_rate": 2.8394982281068727e-06,
+      "loss": 0.8446,
       "step": 26833
     },
     {
-      "epoch": 0.7369750899453462,
+      "epoch": 0.7614642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 3.4143756470684163e-06,
-      "loss": 0.8172,
+      "learning_rate": 2.838856696996621e-06,
+      "loss": 0.9771,
       "step": 26834
     },
     {
-      "epoch": 0.7370025541731894,
+      "epoch": 0.7614926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 3.413706289114701e-06,
-      "loss": 0.8167,
+      "learning_rate": 2.838215226376867e-06,
+      "loss": 0.7708,
       "step": 26835
     },
     {
-      "epoch": 0.7370300184010327,
+      "epoch": 0.7615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 3.4130369832742894e-06,
-      "loss": 0.7661,
+      "learning_rate": 2.837573816253031e-06,
+      "loss": 0.7672,
       "step": 26836
     },
     {
-      "epoch": 0.7370574826288759,
+      "epoch": 0.7615493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 3.412367729552484e-06,
-      "loss": 0.8363,
+      "learning_rate": 2.836932466630533e-06,
+      "loss": 0.8624,
       "step": 26837
     },
     {
-      "epoch": 0.7370849468567191,
+      "epoch": 0.761577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 3.4116985279545745e-06,
-      "loss": 0.8562,
+      "learning_rate": 2.8362911775147863e-06,
+      "loss": 0.9012,
       "step": 26838
     },
     {
-      "epoch": 0.7371124110845624,
+      "epoch": 0.7616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 3.4110293784858573e-06,
-      "loss": 0.825,
+      "learning_rate": 2.83564994891121e-06,
+      "loss": 0.8044,
       "step": 26839
     },
     {
-      "epoch": 0.7371398753124055,
+      "epoch": 0.7616345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 3.410360281151628e-06,
-      "loss": 0.833,
+      "learning_rate": 2.8350087808252236e-06,
+      "loss": 0.776,
       "step": 26840
     },
     {
-      "epoch": 0.7371673395402488,
+      "epoch": 0.7616628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 3.4096912359571843e-06,
-      "loss": 0.8241,
+      "learning_rate": 2.8343676732622336e-06,
+      "loss": 0.8832,
       "step": 26841
     },
     {
-      "epoch": 0.7371948037680921,
+      "epoch": 0.7616912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 3.4090222429078134e-06,
-      "loss": 0.8228,
+      "learning_rate": 2.8337266262276654e-06,
+      "loss": 0.7384,
       "step": 26842
     },
     {
-      "epoch": 0.7372222679959353,
+      "epoch": 0.7617196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 3.408353302008811e-06,
-      "loss": 0.8823,
+      "learning_rate": 2.8330856397269336e-06,
+      "loss": 0.8274,
       "step": 26843
     },
     {
-      "epoch": 0.7372497322237785,
+      "epoch": 0.7617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 3.4076844132654717e-06,
-      "loss": 0.8456,
+      "learning_rate": 2.8324447137654464e-06,
+      "loss": 0.8136,
       "step": 26844
     },
     {
-      "epoch": 0.7372771964516217,
+      "epoch": 0.7617763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 3.407015576683089e-06,
-      "loss": 0.8336,
+      "learning_rate": 2.8318038483486245e-06,
+      "loss": 0.7258,
       "step": 26845
     },
     {
-      "epoch": 0.737304660679465,
+      "epoch": 0.7618047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 3.4063467922669503e-06,
-      "loss": 0.7227,
+      "learning_rate": 2.8311630434818736e-06,
+      "loss": 0.9802,
       "step": 26846
     },
     {
-      "epoch": 0.7373321249073083,
+      "epoch": 0.76183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 3.405678060022354e-06,
-      "loss": 0.7859,
+      "learning_rate": 2.8305222991706114e-06,
+      "loss": 0.8298,
       "step": 26847
     },
     {
-      "epoch": 0.7373595891351514,
+      "epoch": 0.761861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 3.4050093799545827e-06,
-      "loss": 0.8492,
+      "learning_rate": 2.8298816154202526e-06,
+      "loss": 0.8295,
       "step": 26848
     },
     {
-      "epoch": 0.7373870533629947,
+      "epoch": 0.7618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 3.404340752068932e-06,
-      "loss": 0.8253,
+      "learning_rate": 2.8292409922362028e-06,
+      "loss": 0.7379,
       "step": 26849
     },
     {
-      "epoch": 0.737414517590838,
+      "epoch": 0.7619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 3.4036721763706925e-06,
-      "loss": 0.9098,
+      "learning_rate": 2.8286004296238767e-06,
+      "loss": 0.8618,
       "step": 26850
     },
     {
-      "epoch": 0.7374419818186811,
+      "epoch": 0.7619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 3.4030036528651543e-06,
-      "loss": 0.7656,
+      "learning_rate": 2.8279599275886895e-06,
+      "loss": 0.7612,
       "step": 26851
     },
     {
-      "epoch": 0.7374694460465244,
+      "epoch": 0.7619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 3.4023351815576057e-06,
-      "loss": 0.8467,
+      "learning_rate": 2.827319486136042e-06,
+      "loss": 0.827,
       "step": 26852
     },
     {
-      "epoch": 0.7374969102743676,
+      "epoch": 0.7620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 3.4016667624533404e-06,
-      "loss": 0.7924,
+      "learning_rate": 2.8266791052713503e-06,
+      "loss": 0.743,
       "step": 26853
     },
     {
-      "epoch": 0.7375243745022109,
+      "epoch": 0.7620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 3.400998395557644e-06,
-      "loss": 0.9237,
+      "learning_rate": 2.826038785000026e-06,
+      "loss": 0.7464,
       "step": 26854
     },
     {
-      "epoch": 0.7375518387300541,
+      "epoch": 0.7620601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 3.400330080875801e-06,
-      "loss": 0.8115,
+      "learning_rate": 2.825398525327472e-06,
+      "loss": 0.7578,
       "step": 26855
     },
     {
-      "epoch": 0.7375793029578973,
+      "epoch": 0.762088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 3.3996618184131026e-06,
-      "loss": 0.8658,
+      "learning_rate": 2.824758326259095e-06,
+      "loss": 0.9122,
       "step": 26856
     },
     {
-      "epoch": 0.7376067671857406,
+      "epoch": 0.7621169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 3.3989936081748374e-06,
-      "loss": 0.762,
+      "learning_rate": 2.8241181878003134e-06,
+      "loss": 0.6902,
       "step": 26857
     },
     {
-      "epoch": 0.7376342314135838,
+      "epoch": 0.7621452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 3.398325450166291e-06,
-      "loss": 0.8378,
+      "learning_rate": 2.8234781099565245e-06,
+      "loss": 0.8574,
       "step": 26858
     },
     {
-      "epoch": 0.737661695641427,
+      "epoch": 0.7621736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 3.397657344392755e-06,
-      "loss": 0.8778,
+      "learning_rate": 2.8228380927331388e-06,
+      "loss": 0.8783,
       "step": 26859
     },
     {
-      "epoch": 0.7376891598692703,
+      "epoch": 0.7622020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 3.396989290859507e-06,
-      "loss": 0.7618,
+      "learning_rate": 2.8221981361355666e-06,
+      "loss": 0.8397,
       "step": 26860
     },
     {
-      "epoch": 0.7377166240971135,
+      "epoch": 0.7622304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 3.396321289571842e-06,
-      "loss": 0.8115,
+      "learning_rate": 2.8215582401692054e-06,
+      "loss": 0.7552,
       "step": 26861
     },
     {
-      "epoch": 0.7377440883249567,
+      "epoch": 0.7622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 3.3956533405350365e-06,
-      "loss": 0.8715,
+      "learning_rate": 2.8209184048394645e-06,
+      "loss": 0.844,
       "step": 26862
     },
     {
-      "epoch": 0.7377715525528,
+      "epoch": 0.7622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 3.394985443754382e-06,
-      "loss": 0.8729,
+      "learning_rate": 2.8202786301517516e-06,
+      "loss": 0.7234,
       "step": 26863
     },
     {
-      "epoch": 0.7377990167806432,
+      "epoch": 0.7623155505107831,
       "grad_norm": 0.0,
-      "learning_rate": 3.3943175992351595e-06,
-      "loss": 0.8186,
+      "learning_rate": 2.8196389161114647e-06,
+      "loss": 0.8923,
       "step": 26864
     },
     {
-      "epoch": 0.7378264810084865,
+      "epoch": 0.7623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 3.3936498069826586e-06,
-      "loss": 0.9221,
+      "learning_rate": 2.8189992627240117e-06,
+      "loss": 0.8097,
       "step": 26865
     },
     {
-      "epoch": 0.7378539452363296,
+      "epoch": 0.762372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 3.3929820670021564e-06,
-      "loss": 0.9531,
+      "learning_rate": 2.8183596699947967e-06,
+      "loss": 0.8487,
       "step": 26866
     },
     {
-      "epoch": 0.7378814094641729,
+      "epoch": 0.7624006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 3.392314379298939e-06,
-      "loss": 0.854,
+      "learning_rate": 2.8177201379292174e-06,
+      "loss": 0.7781,
       "step": 26867
     },
     {
-      "epoch": 0.7379088736920162,
+      "epoch": 0.7624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 3.3916467438782907e-06,
-      "loss": 0.8257,
+      "learning_rate": 2.8170806665326787e-06,
+      "loss": 0.7798,
       "step": 26868
     },
     {
-      "epoch": 0.7379363379198594,
+      "epoch": 0.7624574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 3.3909791607454957e-06,
-      "loss": 0.7316,
+      "learning_rate": 2.8164412558105856e-06,
+      "loss": 0.7888,
       "step": 26869
     },
     {
-      "epoch": 0.7379638021477026,
+      "epoch": 0.7624858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 3.3903116299058303e-06,
-      "loss": 0.7538,
+      "learning_rate": 2.815801905768334e-06,
+      "loss": 0.9004,
       "step": 26870
     },
     {
-      "epoch": 0.7379912663755459,
+      "epoch": 0.7625141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 3.389644151364584e-06,
-      "loss": 0.7816,
+      "learning_rate": 2.8151626164113265e-06,
+      "loss": 0.7858,
       "step": 26871
     },
     {
-      "epoch": 0.7380187306033891,
+      "epoch": 0.7625425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 3.388976725127029e-06,
-      "loss": 0.7842,
+      "learning_rate": 2.814523387744963e-06,
+      "loss": 0.814,
       "step": 26872
     },
     {
-      "epoch": 0.7380461948312323,
+      "epoch": 0.7625709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 3.388309351198452e-06,
-      "loss": 0.8625,
+      "learning_rate": 2.813884219774643e-06,
+      "loss": 0.7769,
       "step": 26873
     },
     {
-      "epoch": 0.7380736590590755,
+      "epoch": 0.7625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 3.3876420295841316e-06,
-      "loss": 0.7749,
+      "learning_rate": 2.8132451125057703e-06,
+      "loss": 0.8728,
       "step": 26874
     },
     {
-      "epoch": 0.7381011232869188,
+      "epoch": 0.762627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 3.38697476028935e-06,
-      "loss": 0.7507,
+      "learning_rate": 2.8126060659437347e-06,
+      "loss": 0.7982,
       "step": 26875
     },
     {
-      "epoch": 0.7381285875147621,
+      "epoch": 0.7626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 3.386307543319388e-06,
-      "loss": 0.8976,
+      "learning_rate": 2.8119670800939393e-06,
+      "loss": 0.892,
       "step": 26876
     },
     {
-      "epoch": 0.7381560517426052,
+      "epoch": 0.7626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 3.38564037867952e-06,
-      "loss": 0.8599,
+      "learning_rate": 2.8113281549617842e-06,
+      "loss": 0.7635,
       "step": 26877
     },
     {
-      "epoch": 0.7381835159704485,
+      "epoch": 0.7627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 3.384973266375031e-06,
-      "loss": 0.8512,
+      "learning_rate": 2.8106892905526606e-06,
+      "loss": 0.7491,
       "step": 26878
     },
     {
-      "epoch": 0.7382109801982917,
+      "epoch": 0.7627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 3.384306206411192e-06,
-      "loss": 0.8057,
+      "learning_rate": 2.810050486871968e-06,
+      "loss": 0.8039,
       "step": 26879
     },
     {
-      "epoch": 0.738238444426135,
+      "epoch": 0.7627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 3.383639198793286e-06,
-      "loss": 0.7579,
+      "learning_rate": 2.8094117439251045e-06,
+      "loss": 0.7306,
       "step": 26880
     },
     {
-      "epoch": 0.7382659086539782,
+      "epoch": 0.7627979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 3.382972243526589e-06,
-      "loss": 0.8961,
+      "learning_rate": 2.8087730617174603e-06,
+      "loss": 0.8798,
       "step": 26881
     },
     {
-      "epoch": 0.7382933728818214,
+      "epoch": 0.7628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 3.3823053406163775e-06,
-      "loss": 0.7782,
+      "learning_rate": 2.808134440254433e-06,
+      "loss": 0.8984,
       "step": 26882
     },
     {
-      "epoch": 0.7383208371096647,
+      "epoch": 0.7628547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 3.3816384900679346e-06,
-      "loss": 0.8776,
+      "learning_rate": 2.8074958795414207e-06,
+      "loss": 0.9624,
       "step": 26883
     },
     {
-      "epoch": 0.738348301337508,
+      "epoch": 0.762883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 3.3809716918865278e-06,
-      "loss": 0.8867,
+      "learning_rate": 2.806857379583812e-06,
+      "loss": 0.8769,
       "step": 26884
     },
     {
-      "epoch": 0.7383757655653511,
+      "epoch": 0.762911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 3.3803049460774396e-06,
-      "loss": 0.9557,
+      "learning_rate": 2.806218940387001e-06,
+      "loss": 0.6356,
       "step": 26885
     },
     {
-      "epoch": 0.7384032297931944,
+      "epoch": 0.7629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 3.37963825264594e-06,
-      "loss": 0.7957,
+      "learning_rate": 2.8055805619563857e-06,
+      "loss": 0.7783,
       "step": 26886
     },
     {
-      "epoch": 0.7384306940210376,
+      "epoch": 0.7629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 3.378971611597308e-06,
-      "loss": 0.8985,
+      "learning_rate": 2.8049422442973483e-06,
+      "loss": 0.787,
       "step": 26887
     },
     {
-      "epoch": 0.7384581582488808,
+      "epoch": 0.7629965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 3.3783050229368164e-06,
-      "loss": 0.8373,
+      "learning_rate": 2.8043039874152945e-06,
+      "loss": 0.8489,
       "step": 26888
     },
     {
-      "epoch": 0.7384856224767241,
+      "epoch": 0.7630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 3.3776384866697444e-06,
-      "loss": 0.826,
+      "learning_rate": 2.803665791315604e-06,
+      "loss": 0.8475,
       "step": 26889
     },
     {
-      "epoch": 0.7385130867045673,
+      "epoch": 0.7630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 3.3769720028013574e-06,
-      "loss": 0.8743,
+      "learning_rate": 2.803027656003672e-06,
+      "loss": 0.7482,
       "step": 26890
     },
     {
-      "epoch": 0.7385405509324106,
+      "epoch": 0.7630817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 3.376305571336935e-06,
-      "loss": 0.8404,
+      "learning_rate": 2.8023895814848923e-06,
+      "loss": 0.6296,
       "step": 26891
     },
     {
-      "epoch": 0.7385680151602537,
+      "epoch": 0.7631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 3.375639192281748e-06,
-      "loss": 0.7878,
+      "learning_rate": 2.801751567764649e-06,
+      "loss": 0.7938,
       "step": 26892
     },
     {
-      "epoch": 0.738595479388097,
+      "epoch": 0.7631384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 3.374972865641073e-06,
-      "loss": 0.6515,
+      "learning_rate": 2.801113614848333e-06,
+      "loss": 0.8376,
       "step": 26893
     },
     {
-      "epoch": 0.7386229436159403,
+      "epoch": 0.7631668558456299,
       "grad_norm": 0.0,
-      "learning_rate": 3.374306591420179e-06,
-      "loss": 0.8708,
+      "learning_rate": 2.800475722741337e-06,
+      "loss": 0.8546,
       "step": 26894
     },
     {
-      "epoch": 0.7386504078437834,
+      "epoch": 0.7631952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 3.3736403696243336e-06,
-      "loss": 0.8333,
+      "learning_rate": 2.7998378914490433e-06,
+      "loss": 0.9331,
       "step": 26895
     },
     {
-      "epoch": 0.7386778720716267,
+      "epoch": 0.7632236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 3.372974200258814e-06,
-      "loss": 0.7829,
+      "learning_rate": 2.7992001209768427e-06,
+      "loss": 0.749,
       "step": 26896
     },
     {
-      "epoch": 0.73870533629947,
+      "epoch": 0.7632519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 3.372308083328889e-06,
-      "loss": 0.7528,
+      "learning_rate": 2.798562411330126e-06,
+      "loss": 0.7929,
       "step": 26897
     },
     {
-      "epoch": 0.7387328005273132,
+      "epoch": 0.7632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 3.371642018839829e-06,
-      "loss": 0.7361,
+      "learning_rate": 2.7979247625142724e-06,
+      "loss": 0.8845,
       "step": 26898
     },
     {
-      "epoch": 0.7387602647551564,
+      "epoch": 0.7633087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 3.3709760067969055e-06,
-      "loss": 0.8054,
+      "learning_rate": 2.797287174534672e-06,
+      "loss": 0.8805,
       "step": 26899
     },
     {
-      "epoch": 0.7387877289829996,
+      "epoch": 0.7633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 3.370310047205392e-06,
-      "loss": 0.8806,
+      "learning_rate": 2.796649647396714e-06,
+      "loss": 0.8264,
       "step": 26900
     },
     {
-      "epoch": 0.7388151932108429,
+      "epoch": 0.7633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 3.369644140070548e-06,
-      "loss": 0.8098,
+      "learning_rate": 2.796012181105777e-06,
+      "loss": 0.7736,
       "step": 26901
     },
     {
-      "epoch": 0.7388426574386862,
+      "epoch": 0.7633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 3.368978285397653e-06,
-      "loss": 0.7538,
+      "learning_rate": 2.79537477566725e-06,
+      "loss": 0.7863,
       "step": 26902
     },
     {
-      "epoch": 0.7388701216665293,
+      "epoch": 0.763422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 3.3683124831919665e-06,
-      "loss": 0.8298,
+      "learning_rate": 2.794737431086515e-06,
+      "loss": 0.7513,
       "step": 26903
     },
     {
-      "epoch": 0.7388975858943726,
+      "epoch": 0.7634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 3.3676467334587604e-06,
-      "loss": 0.8974,
+      "learning_rate": 2.794100147368957e-06,
+      "loss": 0.8282,
       "step": 26904
     },
     {
-      "epoch": 0.7389250501222158,
+      "epoch": 0.7634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 3.3669810362033027e-06,
-      "loss": 0.8345,
+      "learning_rate": 2.793462924519962e-06,
+      "loss": 0.7968,
       "step": 26905
     },
     {
-      "epoch": 0.738952514350059,
+      "epoch": 0.7635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 3.366315391430863e-06,
-      "loss": 0.8568,
+      "learning_rate": 2.792825762544907e-06,
+      "loss": 0.746,
       "step": 26906
     },
     {
-      "epoch": 0.7389799785779023,
+      "epoch": 0.7635357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 3.365649799146704e-06,
-      "loss": 0.856,
+      "learning_rate": 2.7921886614491765e-06,
+      "loss": 0.7699,
       "step": 26907
     },
     {
-      "epoch": 0.7390074428057455,
+      "epoch": 0.7635641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 3.3649842593560924e-06,
-      "loss": 0.8087,
+      "learning_rate": 2.7915516212381554e-06,
+      "loss": 0.8692,
       "step": 26908
     },
     {
-      "epoch": 0.7390349070335888,
+      "epoch": 0.7635925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 3.3643187720642987e-06,
-      "loss": 0.9477,
+      "learning_rate": 2.7909146419172184e-06,
+      "loss": 0.7933,
       "step": 26909
     },
     {
-      "epoch": 0.739062371261432,
+      "epoch": 0.7636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 3.363653337276582e-06,
-      "loss": 0.8008,
+      "learning_rate": 2.79027772349175e-06,
+      "loss": 0.8393,
       "step": 26910
     },
     {
-      "epoch": 0.7390898354892752,
+      "epoch": 0.7636492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 3.3629879549982113e-06,
-      "loss": 0.7944,
+      "learning_rate": 2.7896408659671327e-06,
+      "loss": 0.7851,
       "step": 26911
     },
     {
-      "epoch": 0.7391172997171185,
+      "epoch": 0.7636776390465381,
       "grad_norm": 0.0,
-      "learning_rate": 3.3623226252344543e-06,
-      "loss": 0.9541,
+      "learning_rate": 2.789004069348741e-06,
+      "loss": 0.7219,
       "step": 26912
     },
     {
-      "epoch": 0.7391447639449616,
+      "epoch": 0.7637060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 3.3616573479905677e-06,
-      "loss": 0.7965,
+      "learning_rate": 2.7883673336419547e-06,
+      "loss": 0.9103,
       "step": 26913
     },
     {
-      "epoch": 0.7391722281728049,
+      "epoch": 0.7637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 3.36099212327182e-06,
-      "loss": 0.9621,
+      "learning_rate": 2.7877306588521567e-06,
+      "loss": 0.8396,
       "step": 26914
     },
     {
-      "epoch": 0.7391996924006482,
+      "epoch": 0.7637627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 3.3603269510834745e-06,
-      "loss": 0.7657,
+      "learning_rate": 2.7870940449847194e-06,
+      "loss": 0.7073,
       "step": 26915
     },
     {
-      "epoch": 0.7392271566284914,
+      "epoch": 0.7637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 3.3596618314307937e-06,
-      "loss": 0.734,
+      "learning_rate": 2.786457492045024e-06,
+      "loss": 0.851,
       "step": 26916
     },
     {
-      "epoch": 0.7392546208563346,
+      "epoch": 0.7638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 3.3589967643190448e-06,
-      "loss": 0.8453,
+      "learning_rate": 2.7858210000384443e-06,
+      "loss": 0.9055,
       "step": 26917
     },
     {
-      "epoch": 0.7392820850841778,
+      "epoch": 0.7638479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 3.358331749753486e-06,
-      "loss": 0.7543,
+      "learning_rate": 2.7851845689703605e-06,
+      "loss": 0.8585,
       "step": 26918
     },
     {
-      "epoch": 0.7393095493120211,
+      "epoch": 0.7638762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 3.3576667877393754e-06,
-      "loss": 0.8334,
+      "learning_rate": 2.78454819884615e-06,
+      "loss": 0.7521,
       "step": 26919
     },
     {
-      "epoch": 0.7393370135398644,
+      "epoch": 0.7639046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 3.357001878281978e-06,
-      "loss": 0.8835,
+      "learning_rate": 2.7839118896711813e-06,
+      "loss": 0.8349,
       "step": 26920
     },
     {
-      "epoch": 0.7393644777677075,
+      "epoch": 0.763933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 3.3563370213865552e-06,
-      "loss": 0.8828,
+      "learning_rate": 2.7832756414508343e-06,
+      "loss": 0.8244,
       "step": 26921
     },
     {
-      "epoch": 0.7393919419955508,
+      "epoch": 0.763961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 3.3556722170583667e-06,
-      "loss": 0.8307,
+      "learning_rate": 2.7826394541904846e-06,
+      "loss": 0.8127,
       "step": 26922
     },
     {
-      "epoch": 0.7394194062233941,
+      "epoch": 0.7639897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 3.3550074653026745e-06,
-      "loss": 0.875,
+      "learning_rate": 2.7820033278955016e-06,
+      "loss": 0.8339,
       "step": 26923
     },
     {
-      "epoch": 0.7394468704512372,
+      "epoch": 0.7640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 3.3543427661247397e-06,
-      "loss": 0.8188,
+      "learning_rate": 2.781367262571261e-06,
+      "loss": 0.8023,
       "step": 26924
     },
     {
-      "epoch": 0.7394743346790805,
+      "epoch": 0.7640465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 3.3536781195298163e-06,
-      "loss": 0.8506,
+      "learning_rate": 2.7807312582231373e-06,
+      "loss": 0.764,
       "step": 26925
     },
     {
-      "epoch": 0.7395017989069237,
+      "epoch": 0.7640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 3.35301352552317e-06,
-      "loss": 0.7581,
+      "learning_rate": 2.780095314856499e-06,
+      "loss": 0.92,
       "step": 26926
     },
     {
-      "epoch": 0.739529263134767,
+      "epoch": 0.7641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 3.352348984110051e-06,
-      "loss": 0.79,
+      "learning_rate": 2.77945943247672e-06,
+      "loss": 0.7685,
       "step": 26927
     },
     {
-      "epoch": 0.7395567273626102,
+      "epoch": 0.7641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 3.3516844952957227e-06,
-      "loss": 0.7433,
+      "learning_rate": 2.7788236110891754e-06,
+      "loss": 0.8506,
       "step": 26928
     },
     {
-      "epoch": 0.7395841915904534,
+      "epoch": 0.7641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 3.3510200590854425e-06,
-      "loss": 0.8209,
+      "learning_rate": 2.7781878506992288e-06,
+      "loss": 0.6986,
       "step": 26929
     },
     {
-      "epoch": 0.7396116558182967,
+      "epoch": 0.7641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 3.350355675484469e-06,
-      "loss": 0.731,
+      "learning_rate": 2.7775521513122537e-06,
+      "loss": 0.7609,
       "step": 26930
     },
     {
-      "epoch": 0.7396391200461399,
+      "epoch": 0.764216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 3.349691344498055e-06,
-      "loss": 0.7715,
+      "learning_rate": 2.776916512933624e-06,
+      "loss": 0.8518,
       "step": 26931
     },
     {
-      "epoch": 0.7396665842739831,
+      "epoch": 0.7642451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 3.3490270661314584e-06,
-      "loss": 0.8654,
+      "learning_rate": 2.7762809355687013e-06,
+      "loss": 0.8583,
       "step": 26932
     },
     {
-      "epoch": 0.7396940485018264,
+      "epoch": 0.7642735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 3.3483628403899404e-06,
-      "loss": 0.773,
+      "learning_rate": 2.7756454192228597e-06,
+      "loss": 0.8114,
       "step": 26933
     },
     {
-      "epoch": 0.7397215127296696,
+      "epoch": 0.7643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 3.347698667278748e-06,
-      "loss": 0.8322,
+      "learning_rate": 2.775009963901465e-06,
+      "loss": 0.7521,
       "step": 26934
     },
     {
-      "epoch": 0.7397489769575128,
+      "epoch": 0.7643303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 3.3470345468031406e-06,
-      "loss": 0.7444,
+      "learning_rate": 2.7743745696098858e-06,
+      "loss": 0.7612,
       "step": 26935
     },
     {
-      "epoch": 0.7397764411853561,
+      "epoch": 0.7643586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 3.346370478968377e-06,
-      "loss": 0.8963,
+      "learning_rate": 2.7737392363534934e-06,
+      "loss": 0.8027,
       "step": 26936
     },
     {
-      "epoch": 0.7398039054131993,
+      "epoch": 0.7643870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 3.3457064637797034e-06,
-      "loss": 0.641,
+      "learning_rate": 2.773103964137647e-06,
+      "loss": 0.9995,
       "step": 26937
     },
     {
-      "epoch": 0.7398313696410426,
+      "epoch": 0.7644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 3.345042501242378e-06,
-      "loss": 0.8457,
+      "learning_rate": 2.772468752967715e-06,
+      "loss": 0.8619,
       "step": 26938
     },
     {
-      "epoch": 0.7398588338688857,
+      "epoch": 0.7644438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 3.3443785913616543e-06,
-      "loss": 0.8076,
+      "learning_rate": 2.771833602849069e-06,
+      "loss": 0.7278,
       "step": 26939
     },
     {
-      "epoch": 0.739886298096729,
+      "epoch": 0.764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.3437147341427858e-06,
-      "loss": 0.8932,
+      "learning_rate": 2.771198513787066e-06,
+      "loss": 0.938,
       "step": 26940
     },
     {
-      "epoch": 0.7399137623245723,
+      "epoch": 0.7645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 3.3430509295910274e-06,
-      "loss": 0.7458,
+      "learning_rate": 2.7705634857870747e-06,
+      "loss": 0.7708,
       "step": 26941
     },
     {
-      "epoch": 0.7399412265524155,
+      "epoch": 0.7645289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 3.342387177711629e-06,
-      "loss": 0.7995,
+      "learning_rate": 2.76992851885446e-06,
+      "loss": 0.8241,
       "step": 26942
     },
     {
-      "epoch": 0.7399686907802587,
+      "epoch": 0.7645573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 3.3417234785098397e-06,
-      "loss": 0.808,
+      "learning_rate": 2.7692936129945823e-06,
+      "loss": 0.8193,
       "step": 26943
     },
     {
-      "epoch": 0.7399961550081019,
+      "epoch": 0.7645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 3.3410598319909127e-06,
-      "loss": 0.8608,
+      "learning_rate": 2.7686587682128062e-06,
+      "loss": 0.785,
       "step": 26944
     },
     {
-      "epoch": 0.7400236192359452,
+      "epoch": 0.7646140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 3.3403962381600997e-06,
-      "loss": 0.8097,
+      "learning_rate": 2.7680239845144986e-06,
+      "loss": 0.8586,
       "step": 26945
     },
     {
-      "epoch": 0.7400510834637885,
+      "epoch": 0.7646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 3.339732697022651e-06,
-      "loss": 0.8967,
+      "learning_rate": 2.7673892619050135e-06,
+      "loss": 0.8009,
       "step": 26946
     },
     {
-      "epoch": 0.7400785476916316,
+      "epoch": 0.7646708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 3.339069208583817e-06,
-      "loss": 0.8547,
+      "learning_rate": 2.766754600389716e-06,
+      "loss": 0.7226,
       "step": 26947
     },
     {
-      "epoch": 0.7401060119194749,
+      "epoch": 0.7646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 3.338405772848852e-06,
-      "loss": 0.8327,
+      "learning_rate": 2.7661199999739683e-06,
+      "loss": 0.7578,
       "step": 26948
     },
     {
-      "epoch": 0.7401334761473182,
+      "epoch": 0.764727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 3.3377423898229964e-06,
-      "loss": 0.9939,
+      "learning_rate": 2.76548546066313e-06,
+      "loss": 0.8104,
       "step": 26949
     },
     {
-      "epoch": 0.7401609403751613,
+      "epoch": 0.764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 3.3370790595115077e-06,
-      "loss": 0.8304,
+      "learning_rate": 2.7648509824625603e-06,
+      "loss": 0.8316,
       "step": 26950
     },
     {
-      "epoch": 0.7401884046030046,
+      "epoch": 0.7647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 3.3364157819196274e-06,
-      "loss": 0.8972,
+      "learning_rate": 2.7642165653776242e-06,
+      "loss": 0.7984,
       "step": 26951
     },
     {
-      "epoch": 0.7402158688308478,
+      "epoch": 0.7648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 3.3357525570526063e-06,
-      "loss": 0.8354,
+      "learning_rate": 2.763582209413672e-06,
+      "loss": 0.8308,
       "step": 26952
     },
     {
-      "epoch": 0.740243333058691,
+      "epoch": 0.7648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 3.3350893849156952e-06,
-      "loss": 0.825,
+      "learning_rate": 2.7629479145760694e-06,
+      "loss": 0.9337,
       "step": 26953
     },
     {
-      "epoch": 0.7402707972865343,
+      "epoch": 0.7648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 3.334426265514137e-06,
-      "loss": 0.8559,
+      "learning_rate": 2.762313680870168e-06,
+      "loss": 0.881,
       "step": 26954
     },
     {
-      "epoch": 0.7402982615143775,
+      "epoch": 0.7648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 3.333763198853178e-06,
-      "loss": 0.7861,
+      "learning_rate": 2.761679508301328e-06,
+      "loss": 0.7945,
       "step": 26955
     },
     {
-      "epoch": 0.7403257257422208,
+      "epoch": 0.7649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 3.3331001849380684e-06,
-      "loss": 0.845,
+      "learning_rate": 2.7610453968749108e-06,
+      "loss": 0.8098,
       "step": 26956
     },
     {
-      "epoch": 0.7403531899700639,
+      "epoch": 0.7649545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 3.3324372237740552e-06,
-      "loss": 0.8231,
+      "learning_rate": 2.7604113465962643e-06,
+      "loss": 0.6985,
       "step": 26957
     },
     {
-      "epoch": 0.7403806541979072,
+      "epoch": 0.764982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 3.3317743153663783e-06,
-      "loss": 0.7777,
+      "learning_rate": 2.759777357470749e-06,
+      "loss": 0.7807,
       "step": 26958
     },
     {
-      "epoch": 0.7404081184257505,
+      "epoch": 0.7650113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 3.3311114597202896e-06,
-      "loss": 0.9429,
+      "learning_rate": 2.7591434295037236e-06,
+      "loss": 0.8499,
       "step": 26959
     },
     {
-      "epoch": 0.7404355826535937,
+      "epoch": 0.7650397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 3.3304486568410264e-06,
-      "loss": 0.875,
+      "learning_rate": 2.7585095627005353e-06,
+      "loss": 0.8213,
       "step": 26960
     },
     {
-      "epoch": 0.7404630468814369,
+      "epoch": 0.7650681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 3.3297859067338378e-06,
-      "loss": 0.9004,
+      "learning_rate": 2.7578757570665416e-06,
+      "loss": 0.8395,
       "step": 26961
     },
     {
-      "epoch": 0.7404905111092802,
+      "epoch": 0.7650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 3.329123209403967e-06,
-      "loss": 0.8405,
+      "learning_rate": 2.757242012607101e-06,
+      "loss": 0.8022,
       "step": 26962
     },
     {
-      "epoch": 0.7405179753371234,
+      "epoch": 0.7651248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 3.328460564856657e-06,
-      "loss": 0.8531,
+      "learning_rate": 2.756608329327557e-06,
+      "loss": 0.8668,
       "step": 26963
     },
     {
-      "epoch": 0.7405454395649667,
+      "epoch": 0.7651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 3.327797973097151e-06,
-      "loss": 0.906,
+      "learning_rate": 2.7559747072332697e-06,
+      "loss": 0.7721,
       "step": 26964
     },
     {
-      "epoch": 0.7405729037928098,
+      "epoch": 0.7651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 3.327135434130696e-06,
-      "loss": 0.7788,
+      "learning_rate": 2.755341146329594e-06,
+      "loss": 0.8406,
       "step": 26965
     },
     {
-      "epoch": 0.7406003680206531,
+      "epoch": 0.7652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 3.326472947962529e-06,
-      "loss": 0.8662,
+      "learning_rate": 2.7547076466218734e-06,
+      "loss": 0.8838,
       "step": 26966
     },
     {
-      "epoch": 0.7406278322484964,
+      "epoch": 0.7652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 3.3258105145978916e-06,
-      "loss": 0.8433,
+      "learning_rate": 2.7540742081154638e-06,
+      "loss": 0.7775,
       "step": 26967
     },
     {
-      "epoch": 0.7406552964763395,
+      "epoch": 0.765266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 3.3251481340420256e-06,
-      "loss": 0.872,
+      "learning_rate": 2.753440830815718e-06,
+      "loss": 0.8571,
       "step": 26968
     },
     {
-      "epoch": 0.7406827607041828,
+      "epoch": 0.7652951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 3.324485806300173e-06,
-      "loss": 0.8443,
+      "learning_rate": 2.75280751472798e-06,
+      "loss": 0.7406,
       "step": 26969
     },
     {
-      "epoch": 0.740710224932026,
+      "epoch": 0.7653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 3.3238235313775744e-06,
-      "loss": 0.8258,
+      "learning_rate": 2.752174259857602e-06,
+      "loss": 0.819,
       "step": 26970
     },
     {
-      "epoch": 0.7407376891598693,
+      "epoch": 0.7653518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 3.3231613092794747e-06,
-      "loss": 0.8069,
+      "learning_rate": 2.7515410662099375e-06,
+      "loss": 0.7959,
       "step": 26971
     },
     {
-      "epoch": 0.7407651533877125,
+      "epoch": 0.7653802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 3.3224991400111052e-06,
-      "loss": 0.8324,
+      "learning_rate": 2.750907933790329e-06,
+      "loss": 0.8415,
       "step": 26972
     },
     {
-      "epoch": 0.7407926176155557,
+      "epoch": 0.7654086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 3.3218370235777085e-06,
-      "loss": 0.7636,
+      "learning_rate": 2.7502748626041266e-06,
+      "loss": 0.8074,
       "step": 26973
     },
     {
-      "epoch": 0.740820081843399,
+      "epoch": 0.7654370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 3.321174959984528e-06,
-      "loss": 0.7881,
+      "learning_rate": 2.74964185265668e-06,
+      "loss": 0.8496,
       "step": 26974
     },
     {
-      "epoch": 0.7408475460712423,
+      "epoch": 0.7654653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 3.320512949236795e-06,
-      "loss": 0.8732,
+      "learning_rate": 2.749008903953333e-06,
+      "loss": 0.7859,
       "step": 26975
     },
     {
-      "epoch": 0.7408750102990854,
+      "epoch": 0.7654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 3.3198509913397504e-06,
-      "loss": 0.7378,
+      "learning_rate": 2.7483760164994344e-06,
+      "loss": 0.8239,
       "step": 26976
     },
     {
-      "epoch": 0.7409024745269287,
+      "epoch": 0.765522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 3.3191890862986366e-06,
-      "loss": 0.7595,
+      "learning_rate": 2.7477431903003314e-06,
+      "loss": 0.844,
       "step": 26977
     },
     {
-      "epoch": 0.7409299387547719,
+      "epoch": 0.7655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 3.3185272341186824e-06,
-      "loss": 0.7917,
+      "learning_rate": 2.7471104253613645e-06,
+      "loss": 0.8224,
       "step": 26978
     },
     {
-      "epoch": 0.7409574029826151,
+      "epoch": 0.7655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 3.3178654348051296e-06,
-      "loss": 0.8104,
+      "learning_rate": 2.7464777216878825e-06,
+      "loss": 0.8941,
       "step": 26979
     },
     {
-      "epoch": 0.7409848672104584,
+      "epoch": 0.7656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 3.317203688363214e-06,
-      "loss": 0.8243,
+      "learning_rate": 2.7458450792852296e-06,
+      "loss": 0.8168,
       "step": 26980
     },
     {
-      "epoch": 0.7410123314383016,
+      "epoch": 0.7656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 3.316541994798175e-06,
-      "loss": 0.8429,
+      "learning_rate": 2.745212498158749e-06,
+      "loss": 0.8879,
       "step": 26981
     },
     {
-      "epoch": 0.7410397956661449,
+      "epoch": 0.7656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 3.3158803541152397e-06,
-      "loss": 0.8535,
+      "learning_rate": 2.744579978313787e-06,
+      "loss": 0.8229,
       "step": 26982
     },
     {
-      "epoch": 0.741067259893988,
+      "epoch": 0.7656923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 3.3152187663196534e-06,
-      "loss": 0.9523,
+      "learning_rate": 2.743947519755682e-06,
+      "loss": 0.7864,
       "step": 26983
     },
     {
-      "epoch": 0.7410947241218313,
+      "epoch": 0.7657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 3.3145572314166406e-06,
-      "loss": 0.7424,
+      "learning_rate": 2.7433151224897782e-06,
+      "loss": 0.9086,
       "step": 26984
     },
     {
-      "epoch": 0.7411221883496746,
+      "epoch": 0.7657491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 3.313895749411442e-06,
-      "loss": 0.8909,
+      "learning_rate": 2.7426827865214212e-06,
+      "loss": 0.7885,
       "step": 26985
     },
     {
-      "epoch": 0.7411496525775177,
+      "epoch": 0.76577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 3.3132343203092897e-06,
-      "loss": 0.8568,
+      "learning_rate": 2.7420505118559447e-06,
+      "loss": 0.8361,
       "step": 26986
     },
     {
-      "epoch": 0.741177116805361,
+      "epoch": 0.765805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 3.3125729441154188e-06,
-      "loss": 0.8625,
+      "learning_rate": 2.7414182984986947e-06,
+      "loss": 0.8351,
       "step": 26987
     },
     {
-      "epoch": 0.7412045810332043,
+      "epoch": 0.7658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 3.31191162083506e-06,
-      "loss": 0.9326,
+      "learning_rate": 2.7407861464550134e-06,
+      "loss": 0.7031,
       "step": 26988
     },
     {
-      "epoch": 0.7412320452610475,
+      "epoch": 0.7658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 3.3112503504734504e-06,
-      "loss": 0.8855,
+      "learning_rate": 2.7401540557302355e-06,
+      "loss": 0.7607,
       "step": 26989
     },
     {
-      "epoch": 0.7412595094888907,
+      "epoch": 0.7658910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 3.3105891330358185e-06,
-      "loss": 0.8139,
+      "learning_rate": 2.739522026329702e-06,
+      "loss": 0.7565,
       "step": 26990
     },
     {
-      "epoch": 0.7412869737167339,
+      "epoch": 0.7659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 3.3099279685273946e-06,
-      "loss": 0.8179,
+      "learning_rate": 2.7388900582587553e-06,
+      "loss": 0.851,
       "step": 26991
     },
     {
-      "epoch": 0.7413144379445772,
+      "epoch": 0.7659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 3.3092668569534112e-06,
-      "loss": 0.8162,
+      "learning_rate": 2.738258151522727e-06,
+      "loss": 0.7998,
       "step": 26992
     },
     {
-      "epoch": 0.7413419021724205,
+      "epoch": 0.7659761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 3.308605798319101e-06,
-      "loss": 0.8141,
+      "learning_rate": 2.7376263061269594e-06,
+      "loss": 0.846,
       "step": 26993
     },
     {
-      "epoch": 0.7413693664002636,
+      "epoch": 0.7660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 3.3079447926296936e-06,
-      "loss": 0.7734,
+      "learning_rate": 2.736994522076789e-06,
+      "loss": 0.8281,
       "step": 26994
     },
     {
-      "epoch": 0.7413968306281069,
+      "epoch": 0.7660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 3.307283839890422e-06,
-      "loss": 0.7494,
+      "learning_rate": 2.7363627993775522e-06,
+      "loss": 0.7607,
       "step": 26995
     },
     {
-      "epoch": 0.7414242948559501,
+      "epoch": 0.7660612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 3.3066229401065098e-06,
-      "loss": 0.7545,
+      "learning_rate": 2.7357311380345873e-06,
+      "loss": 0.8834,
       "step": 26996
     },
     {
-      "epoch": 0.7414517590837933,
+      "epoch": 0.7660896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 3.305962093283194e-06,
-      "loss": 0.7635,
+      "learning_rate": 2.7350995380532264e-06,
+      "loss": 0.7576,
       "step": 26997
     },
     {
-      "epoch": 0.7414792233116366,
+      "epoch": 0.7661180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 3.305301299425695e-06,
-      "loss": 0.8134,
+      "learning_rate": 2.7344679994388057e-06,
+      "loss": 0.8608,
       "step": 26998
     },
     {
-      "epoch": 0.7415066875394798,
+      "epoch": 0.7661464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 3.304640558539245e-06,
-      "loss": 0.788,
+      "learning_rate": 2.7338365221966634e-06,
+      "loss": 0.8036,
       "step": 26999
     },
     {
-      "epoch": 0.7415341517673231,
+      "epoch": 0.7661748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 3.3039798706290737e-06,
-      "loss": 0.7831,
+      "learning_rate": 2.7332051063321284e-06,
+      "loss": 0.7675,
       "step": 27000
     },
     {
-      "epoch": 0.7415616159951663,
+      "epoch": 0.7662031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 3.3033192357004096e-06,
-      "loss": 0.8433,
+      "learning_rate": 2.732573751850536e-06,
+      "loss": 0.806,
       "step": 27001
     },
     {
-      "epoch": 0.7415890802230095,
+      "epoch": 0.7662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 3.302658653758475e-06,
-      "loss": 0.7831,
+      "learning_rate": 2.731942458757223e-06,
+      "loss": 0.8765,
       "step": 27002
     },
     {
-      "epoch": 0.7416165444508528,
+      "epoch": 0.7662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 3.3019981248084996e-06,
-      "loss": 0.8624,
+      "learning_rate": 2.731311227057515e-06,
+      "loss": 0.8396,
       "step": 27003
     },
     {
-      "epoch": 0.741644008678696,
+      "epoch": 0.7662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 3.3013376488557094e-06,
-      "loss": 0.8246,
+      "learning_rate": 2.730680056756748e-06,
+      "loss": 0.8286,
       "step": 27004
     },
     {
-      "epoch": 0.7416714729065392,
+      "epoch": 0.766316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 3.3006772259053333e-06,
-      "loss": 0.8777,
+      "learning_rate": 2.7300489478602567e-06,
+      "loss": 0.8415,
       "step": 27005
     },
     {
-      "epoch": 0.7416989371343825,
+      "epoch": 0.7663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 3.300016855962591e-06,
-      "loss": 0.8822,
+      "learning_rate": 2.7294179003733656e-06,
+      "loss": 0.7499,
       "step": 27006
     },
     {
-      "epoch": 0.7417264013622257,
+      "epoch": 0.7663734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 3.2993565390327142e-06,
-      "loss": 0.8409,
+      "learning_rate": 2.728786914301409e-06,
+      "loss": 0.8442,
       "step": 27007
     },
     {
-      "epoch": 0.741753865590069,
+      "epoch": 0.7664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 3.2986962751209207e-06,
-      "loss": 0.8108,
+      "learning_rate": 2.728155989649719e-06,
+      "loss": 0.7564,
       "step": 27008
     },
     {
-      "epoch": 0.7417813298179121,
+      "epoch": 0.7664301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 3.298036064232438e-06,
-      "loss": 0.9118,
+      "learning_rate": 2.727525126423618e-06,
+      "loss": 0.8572,
       "step": 27009
     },
     {
-      "epoch": 0.7418087940457554,
+      "epoch": 0.7664585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 3.297375906372491e-06,
-      "loss": 0.8672,
+      "learning_rate": 2.7268943246284407e-06,
+      "loss": 0.8332,
       "step": 27010
     },
     {
-      "epoch": 0.7418362582735987,
+      "epoch": 0.7664869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 3.296715801546301e-06,
-      "loss": 0.7356,
+      "learning_rate": 2.726263584269513e-06,
+      "loss": 0.9026,
       "step": 27011
     },
     {
-      "epoch": 0.7418637225014418,
+      "epoch": 0.7665153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 3.296055749759096e-06,
-      "loss": 0.8206,
+      "learning_rate": 2.7256329053521646e-06,
+      "loss": 0.793,
       "step": 27012
     },
     {
-      "epoch": 0.7418911867292851,
+      "epoch": 0.7665437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 3.2953957510160904e-06,
-      "loss": 0.879,
+      "learning_rate": 2.725002287881724e-06,
+      "loss": 0.7629,
       "step": 27013
     },
     {
-      "epoch": 0.7419186509571284,
+      "epoch": 0.7665720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 3.294735805322515e-06,
-      "loss": 0.9169,
+      "learning_rate": 2.7243717318635143e-06,
+      "loss": 0.7748,
       "step": 27014
     },
     {
-      "epoch": 0.7419461151849716,
+      "epoch": 0.7666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 3.294075912683584e-06,
-      "loss": 0.8189,
+      "learning_rate": 2.723741237302863e-06,
+      "loss": 0.8246,
       "step": 27015
     },
     {
-      "epoch": 0.7419735794128148,
+      "epoch": 0.7666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 3.2934160731045216e-06,
-      "loss": 0.8047,
+      "learning_rate": 2.723110804205099e-06,
+      "loss": 0.7797,
       "step": 27016
     },
     {
-      "epoch": 0.742001043640658,
+      "epoch": 0.7666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 3.2927562865905506e-06,
-      "loss": 0.8322,
+      "learning_rate": 2.7224804325755427e-06,
+      "loss": 0.8037,
       "step": 27017
     },
     {
-      "epoch": 0.7420285078685013,
+      "epoch": 0.7666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 3.292096553146892e-06,
-      "loss": 0.8056,
+      "learning_rate": 2.7218501224195217e-06,
+      "loss": 0.6979,
       "step": 27018
     },
     {
-      "epoch": 0.7420559720963446,
+      "epoch": 0.7667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 3.2914368727787603e-06,
-      "loss": 0.8655,
+      "learning_rate": 2.7212198737423624e-06,
+      "loss": 0.8588,
       "step": 27019
     },
     {
-      "epoch": 0.7420834363241877,
+      "epoch": 0.7667423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 3.290777245491379e-06,
-      "loss": 0.9004,
+      "learning_rate": 2.720589686549383e-06,
+      "loss": 0.8245,
       "step": 27020
     },
     {
-      "epoch": 0.742110900552031,
+      "epoch": 0.7667707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 3.2901176712899706e-06,
-      "loss": 0.814,
+      "learning_rate": 2.7199595608459107e-06,
+      "loss": 0.7872,
       "step": 27021
     },
     {
-      "epoch": 0.7421383647798742,
+      "epoch": 0.7667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 3.2894581501797464e-06,
-      "loss": 0.823,
+      "learning_rate": 2.7193294966372697e-06,
+      "loss": 0.8735,
       "step": 27022
     },
     {
-      "epoch": 0.7421658290077174,
+      "epoch": 0.766827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 3.28879868216593e-06,
-      "loss": 0.7837,
+      "learning_rate": 2.718699493928776e-06,
+      "loss": 0.7957,
       "step": 27023
     },
     {
-      "epoch": 0.7421932932355607,
+      "epoch": 0.7668558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 3.2881392672537403e-06,
-      "loss": 0.8721,
+      "learning_rate": 2.718069552725756e-06,
+      "loss": 0.9076,
       "step": 27024
     },
     {
-      "epoch": 0.7422207574634039,
+      "epoch": 0.7668842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 3.2874799054483897e-06,
-      "loss": 0.8318,
+      "learning_rate": 2.7174396730335296e-06,
+      "loss": 0.8037,
       "step": 27025
     },
     {
-      "epoch": 0.7422482216912472,
+      "epoch": 0.7669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 3.2868205967550983e-06,
-      "loss": 0.8021,
+      "learning_rate": 2.7168098548574175e-06,
+      "loss": 0.8169,
       "step": 27026
     },
     {
-      "epoch": 0.7422756859190904,
+      "epoch": 0.7669409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 3.2861613411790825e-06,
-      "loss": 0.8645,
+      "learning_rate": 2.716180098202742e-06,
+      "loss": 0.7763,
       "step": 27027
     },
     {
-      "epoch": 0.7423031501469336,
+      "epoch": 0.7669693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 3.285502138725559e-06,
-      "loss": 0.8017,
+      "learning_rate": 2.7155504030748193e-06,
+      "loss": 0.7596,
       "step": 27028
     },
     {
-      "epoch": 0.7423306143747769,
+      "epoch": 0.7669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 3.284842989399747e-06,
-      "loss": 0.8031,
+      "learning_rate": 2.714920769478969e-06,
+      "loss": 0.7406,
       "step": 27029
     },
     {
-      "epoch": 0.74235807860262,
+      "epoch": 0.7670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.2841838932068583e-06,
-      "loss": 0.8291,
+      "learning_rate": 2.7142911974205135e-06,
+      "loss": 0.9037,
       "step": 27030
     },
     {
-      "epoch": 0.7423855428304633,
+      "epoch": 0.7670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 3.2835248501521044e-06,
-      "loss": 0.8733,
+      "learning_rate": 2.713661686904765e-06,
+      "loss": 0.7309,
       "step": 27031
     },
     {
-      "epoch": 0.7424130070583066,
+      "epoch": 0.7670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 3.282865860240705e-06,
-      "loss": 0.8405,
+      "learning_rate": 2.7130322379370434e-06,
+      "loss": 0.8065,
       "step": 27032
     },
     {
-      "epoch": 0.7424404712861498,
+      "epoch": 0.76711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 3.282206923477871e-06,
-      "loss": 0.7953,
+      "learning_rate": 2.7124028505226685e-06,
+      "loss": 0.7455,
       "step": 27033
     },
     {
-      "epoch": 0.742467935513993,
+      "epoch": 0.7671396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 3.281548039868819e-06,
-      "loss": 0.8297,
+      "learning_rate": 2.7117735246669517e-06,
+      "loss": 0.8083,
       "step": 27034
     },
     {
-      "epoch": 0.7424953997418362,
+      "epoch": 0.7671679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 3.2808892094187616e-06,
-      "loss": 0.8123,
+      "learning_rate": 2.7111442603752125e-06,
+      "loss": 0.7621,
       "step": 27035
     },
     {
-      "epoch": 0.7425228639696795,
+      "epoch": 0.7671963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 3.280230432132915e-06,
-      "loss": 0.8712,
+      "learning_rate": 2.7105150576527672e-06,
+      "loss": 0.8235,
       "step": 27036
     },
     {
-      "epoch": 0.7425503281975228,
+      "epoch": 0.7672247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 3.2795717080164847e-06,
-      "loss": 0.7771,
+      "learning_rate": 2.709885916504927e-06,
+      "loss": 0.7789,
       "step": 27037
     },
     {
-      "epoch": 0.7425777924253659,
+      "epoch": 0.7672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 3.278913037074689e-06,
-      "loss": 0.7854,
+      "learning_rate": 2.7092568369370075e-06,
+      "loss": 0.8421,
       "step": 27038
     },
     {
-      "epoch": 0.7426052566532092,
+      "epoch": 0.7672814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 3.2782544193127343e-06,
-      "loss": 0.8158,
+      "learning_rate": 2.7086278189543267e-06,
+      "loss": 0.8564,
       "step": 27039
     },
     {
-      "epoch": 0.7426327208810525,
+      "epoch": 0.7673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 3.2775958547358334e-06,
-      "loss": 0.9127,
+      "learning_rate": 2.7079988625621877e-06,
+      "loss": 0.7464,
       "step": 27040
     },
     {
-      "epoch": 0.7426601851088956,
+      "epoch": 0.7673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 3.276937343349199e-06,
-      "loss": 0.8561,
+      "learning_rate": 2.707369967765917e-06,
+      "loss": 0.8763,
       "step": 27041
     },
     {
-      "epoch": 0.7426876493367389,
+      "epoch": 0.767366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.276278885158043e-06,
-      "loss": 0.7864,
+      "learning_rate": 2.706741134570816e-06,
+      "loss": 0.8447,
       "step": 27042
     },
     {
-      "epoch": 0.7427151135645821,
+      "epoch": 0.7673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 3.2756204801675704e-06,
-      "loss": 0.7569,
+      "learning_rate": 2.7061123629822016e-06,
+      "loss": 0.8309,
       "step": 27043
     },
     {
-      "epoch": 0.7427425777924254,
+      "epoch": 0.7674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 3.2749621283829934e-06,
-      "loss": 0.8246,
+      "learning_rate": 2.7054836530053864e-06,
+      "loss": 0.7458,
       "step": 27044
     },
     {
-      "epoch": 0.7427700420202686,
+      "epoch": 0.7674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 3.274303829809524e-06,
-      "loss": 0.8254,
+      "learning_rate": 2.704855004645676e-06,
+      "loss": 0.8937,
       "step": 27045
     },
     {
-      "epoch": 0.7427975062481118,
+      "epoch": 0.7674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 3.273645584452365e-06,
-      "loss": 0.8883,
+      "learning_rate": 2.7042264179083834e-06,
+      "loss": 0.7813,
       "step": 27046
     },
     {
-      "epoch": 0.7428249704759551,
+      "epoch": 0.7675085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 3.2729873923167275e-06,
-      "loss": 0.7455,
+      "learning_rate": 2.7035978927988205e-06,
+      "loss": 0.8589,
       "step": 27047
     },
     {
-      "epoch": 0.7428524347037982,
+      "epoch": 0.7675368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 3.272329253407823e-06,
-      "loss": 0.8412,
+      "learning_rate": 2.7029694293222907e-06,
+      "loss": 0.8725,
       "step": 27048
     },
     {
-      "epoch": 0.7428798989316415,
+      "epoch": 0.7675652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 3.2716711677308523e-06,
-      "loss": 0.7856,
+      "learning_rate": 2.702341027484108e-06,
+      "loss": 0.7843,
       "step": 27049
     },
     {
-      "epoch": 0.7429073631594848,
+      "epoch": 0.7675936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 3.2710131352910257e-06,
-      "loss": 0.9538,
+      "learning_rate": 2.7017126872895805e-06,
+      "loss": 0.7337,
       "step": 27050
     },
     {
-      "epoch": 0.742934827387328,
+      "epoch": 0.767622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 3.2703551560935497e-06,
-      "loss": 0.9045,
+      "learning_rate": 2.7010844087440113e-06,
+      "loss": 0.7615,
       "step": 27051
     },
     {
-      "epoch": 0.7429622916151712,
+      "epoch": 0.767650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 3.2696972301436302e-06,
-      "loss": 0.9392,
+      "learning_rate": 2.7004561918527096e-06,
+      "loss": 0.8245,
       "step": 27052
     },
     {
-      "epoch": 0.7429897558430145,
+      "epoch": 0.7676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 3.2690393574464765e-06,
-      "loss": 0.7832,
+      "learning_rate": 2.699828036620986e-06,
+      "loss": 0.7901,
       "step": 27053
     },
     {
-      "epoch": 0.7430172200708577,
+      "epoch": 0.7677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 3.2683815380072905e-06,
-      "loss": 0.7668,
+      "learning_rate": 2.6991999430541405e-06,
+      "loss": 0.7905,
       "step": 27054
     },
     {
-      "epoch": 0.743044684298701,
+      "epoch": 0.7677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 3.2677237718312748e-06,
-      "loss": 0.8522,
+      "learning_rate": 2.6985719111574814e-06,
+      "loss": 0.7891,
       "step": 27055
     },
     {
-      "epoch": 0.7430721485265441,
+      "epoch": 0.7677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 3.267066058923636e-06,
-      "loss": 0.8928,
+      "learning_rate": 2.697943940936313e-06,
+      "loss": 0.8175,
       "step": 27056
     },
     {
-      "epoch": 0.7430996127543874,
+      "epoch": 0.7677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 3.266408399289579e-06,
-      "loss": 0.8721,
+      "learning_rate": 2.69731603239594e-06,
+      "loss": 0.8268,
       "step": 27057
     },
     {
-      "epoch": 0.7431270769822307,
+      "epoch": 0.7678206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 3.265750792934307e-06,
-      "loss": 0.8534,
+      "learning_rate": 2.6966881855416684e-06,
+      "loss": 0.8898,
       "step": 27058
     },
     {
-      "epoch": 0.7431545412100738,
+      "epoch": 0.7678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 3.2650932398630233e-06,
-      "loss": 0.6961,
+      "learning_rate": 2.6960604003788014e-06,
+      "loss": 0.8306,
       "step": 27059
     },
     {
-      "epoch": 0.7431820054379171,
+      "epoch": 0.7678774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 3.2644357400809332e-06,
-      "loss": 0.8242,
+      "learning_rate": 2.695432676912638e-06,
+      "loss": 0.7986,
       "step": 27060
     },
     {
-      "epoch": 0.7432094696657603,
+      "epoch": 0.7679057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 3.263778293593234e-06,
-      "loss": 0.7487,
+      "learning_rate": 2.6948050151484862e-06,
+      "loss": 0.8189,
       "step": 27061
     },
     {
-      "epoch": 0.7432369338936036,
+      "epoch": 0.7679341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 3.2631209004051335e-06,
-      "loss": 0.6999,
+      "learning_rate": 2.694177415091642e-06,
+      "loss": 0.82,
       "step": 27062
     },
     {
-      "epoch": 0.7432643981214468,
+      "epoch": 0.7679625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 3.262463560521827e-06,
-      "loss": 0.8099,
+      "learning_rate": 2.69354987674741e-06,
+      "loss": 0.7923,
       "step": 27063
     },
     {
-      "epoch": 0.74329186234929,
+      "epoch": 0.7679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 3.261806273948519e-06,
-      "loss": 0.8007,
+      "learning_rate": 2.692922400121093e-06,
+      "loss": 0.8463,
       "step": 27064
     },
     {
-      "epoch": 0.7433193265771333,
+      "epoch": 0.7680192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 3.261149040690409e-06,
-      "loss": 0.7526,
+      "learning_rate": 2.692294985217986e-06,
+      "loss": 0.8687,
       "step": 27065
     },
     {
-      "epoch": 0.7433467908049766,
+      "epoch": 0.7680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 3.2604918607527023e-06,
-      "loss": 0.8626,
+      "learning_rate": 2.691667632043391e-06,
+      "loss": 0.8854,
       "step": 27066
     },
     {
-      "epoch": 0.7433742550328197,
+      "epoch": 0.7680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 3.259834734140591e-06,
-      "loss": 0.8532,
+      "learning_rate": 2.691040340602612e-06,
+      "loss": 0.8362,
       "step": 27067
     },
     {
-      "epoch": 0.743401719260663,
+      "epoch": 0.7681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 3.2591776608592784e-06,
-      "loss": 0.8554,
+      "learning_rate": 2.690413110900941e-06,
+      "loss": 0.8279,
       "step": 27068
     },
     {
-      "epoch": 0.7434291834885062,
+      "epoch": 0.7681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 3.2585206409139656e-06,
-      "loss": 0.7911,
+      "learning_rate": 2.689785942943679e-06,
+      "loss": 0.8553,
       "step": 27069
     },
     {
-      "epoch": 0.7434566477163495,
+      "epoch": 0.768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.2578636743098467e-06,
-      "loss": 0.7991,
+      "learning_rate": 2.6891588367361265e-06,
+      "loss": 0.8215,
       "step": 27070
     },
     {
-      "epoch": 0.7434841119441927,
+      "epoch": 0.7681895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 3.2572067610521253e-06,
-      "loss": 0.9107,
+      "learning_rate": 2.6885317922835717e-06,
+      "loss": 0.8167,
       "step": 27071
     },
     {
-      "epoch": 0.7435115761720359,
+      "epoch": 0.7682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 3.256549901145992e-06,
-      "loss": 0.7666,
+      "learning_rate": 2.6879048095913206e-06,
+      "loss": 0.841,
       "step": 27072
     },
     {
-      "epoch": 0.7435390403998792,
+      "epoch": 0.7682463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 3.255893094596648e-06,
-      "loss": 0.762,
+      "learning_rate": 2.68727788866467e-06,
+      "loss": 0.7621,
       "step": 27073
     },
     {
-      "epoch": 0.7435665046277224,
+      "epoch": 0.7682746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 3.255236341409291e-06,
-      "loss": 0.8294,
+      "learning_rate": 2.6866510295089077e-06,
+      "loss": 0.8842,
       "step": 27074
     },
     {
-      "epoch": 0.7435939688555656,
+      "epoch": 0.7683030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 3.2545796415891164e-06,
-      "loss": 0.8492,
+      "learning_rate": 2.686024232129334e-06,
+      "loss": 0.8448,
       "step": 27075
     },
     {
-      "epoch": 0.7436214330834089,
+      "epoch": 0.7683314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 3.253922995141321e-06,
-      "loss": 0.778,
+      "learning_rate": 2.6853974965312446e-06,
+      "loss": 0.7312,
       "step": 27076
     },
     {
-      "epoch": 0.743648897311252,
+      "epoch": 0.7683598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 3.2532664020711026e-06,
-      "loss": 0.902,
+      "learning_rate": 2.684770822719929e-06,
+      "loss": 0.7688,
       "step": 27077
     },
     {
-      "epoch": 0.7436763615390953,
+      "epoch": 0.7683881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 3.2526098623836542e-06,
-      "loss": 0.7555,
+      "learning_rate": 2.684144210700682e-06,
+      "loss": 0.8245,
       "step": 27078
     },
     {
-      "epoch": 0.7437038257669386,
+      "epoch": 0.7684165720771851,
       "grad_norm": 0.0,
-      "learning_rate": 3.251953376084167e-06,
-      "loss": 0.8785,
+      "learning_rate": 2.6835176604788014e-06,
+      "loss": 0.84,
       "step": 27079
     },
     {
-      "epoch": 0.7437312899947818,
+      "epoch": 0.7684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 3.2512969431778385e-06,
-      "loss": 0.7641,
+      "learning_rate": 2.682891172059573e-06,
+      "loss": 0.9411,
       "step": 27080
     },
     {
-      "epoch": 0.743758754222625,
+      "epoch": 0.7684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 3.250640563669862e-06,
-      "loss": 0.75,
+      "learning_rate": 2.682264745448292e-06,
+      "loss": 0.727,
       "step": 27081
     },
     {
-      "epoch": 0.7437862184504682,
+      "epoch": 0.7685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 3.249984237565433e-06,
-      "loss": 0.8203,
+      "learning_rate": 2.681638380650252e-06,
+      "loss": 0.7444,
       "step": 27082
     },
     {
-      "epoch": 0.7438136826783115,
+      "epoch": 0.7685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 3.2493279648697463e-06,
-      "loss": 0.8204,
+      "learning_rate": 2.6810120776707395e-06,
+      "loss": 0.8182,
       "step": 27083
     },
     {
-      "epoch": 0.7438411469061548,
+      "epoch": 0.7685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 3.248671745587989e-06,
-      "loss": 0.8939,
+      "learning_rate": 2.680385836515046e-06,
+      "loss": 0.7007,
       "step": 27084
     },
     {
-      "epoch": 0.7438686111339979,
+      "epoch": 0.7685868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 3.2480155797253553e-06,
-      "loss": 0.7615,
+      "learning_rate": 2.6797596571884663e-06,
+      "loss": 0.7534,
       "step": 27085
     },
     {
-      "epoch": 0.7438960753618412,
+      "epoch": 0.7686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 3.2473594672870413e-06,
-      "loss": 0.9093,
+      "learning_rate": 2.6791335396962794e-06,
+      "loss": 0.8722,
       "step": 27086
     },
     {
-      "epoch": 0.7439235395896845,
+      "epoch": 0.7686435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 3.246703408278231e-06,
-      "loss": 0.8535,
+      "learning_rate": 2.6785074840437864e-06,
+      "loss": 0.8576,
       "step": 27087
     },
     {
-      "epoch": 0.7439510038175277,
+      "epoch": 0.768671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 3.24604740270412e-06,
-      "loss": 0.8314,
+      "learning_rate": 2.677881490236267e-06,
+      "loss": 0.7956,
       "step": 27088
     },
     {
-      "epoch": 0.7439784680453709,
+      "epoch": 0.768700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 3.2453914505699015e-06,
-      "loss": 0.7887,
+      "learning_rate": 2.6772555582790128e-06,
+      "loss": 0.7951,
       "step": 27089
     },
     {
-      "epoch": 0.7440059322732141,
+      "epoch": 0.7687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 3.2447355518807588e-06,
-      "loss": 0.8117,
+      "learning_rate": 2.676629688177311e-06,
+      "loss": 0.9308,
       "step": 27090
     },
     {
-      "epoch": 0.7440333965010574,
+      "epoch": 0.7687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 3.244079706641886e-06,
-      "loss": 0.6866,
+      "learning_rate": 2.6760038799364462e-06,
+      "loss": 0.909,
       "step": 27091
     },
     {
-      "epoch": 0.7440608607289007,
+      "epoch": 0.7687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 3.24342391485847e-06,
-      "loss": 0.8639,
+      "learning_rate": 2.6753781335617057e-06,
+      "loss": 0.75,
       "step": 27092
     },
     {
-      "epoch": 0.7440883249567438,
+      "epoch": 0.7688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 3.2427681765357054e-06,
-      "loss": 0.8404,
+      "learning_rate": 2.6747524490583775e-06,
+      "loss": 0.793,
       "step": 27093
     },
     {
-      "epoch": 0.7441157891845871,
+      "epoch": 0.7688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 3.242112491678774e-06,
-      "loss": 0.884,
+      "learning_rate": 2.674126826431742e-06,
+      "loss": 0.8277,
       "step": 27094
     },
     {
-      "epoch": 0.7441432534124303,
+      "epoch": 0.7688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 3.241456860292869e-06,
-      "loss": 0.7659,
+      "learning_rate": 2.6735012656870874e-06,
+      "loss": 0.8655,
       "step": 27095
     },
     {
-      "epoch": 0.7441707176402735,
+      "epoch": 0.7688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 3.240801282383171e-06,
-      "loss": 0.8217,
+      "learning_rate": 2.672875766829699e-06,
+      "loss": 0.884,
       "step": 27096
     },
     {
-      "epoch": 0.7441981818681168,
+      "epoch": 0.7689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 3.2401457579548725e-06,
-      "loss": 0.8399,
+      "learning_rate": 2.672250329864855e-06,
+      "loss": 0.9211,
       "step": 27097
     },
     {
-      "epoch": 0.74422564609596,
+      "epoch": 0.7689557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 3.2394902870131596e-06,
-      "loss": 0.8157,
+      "learning_rate": 2.6716249547978424e-06,
+      "loss": 0.7671,
       "step": 27098
     },
     {
-      "epoch": 0.7442531103238033,
+      "epoch": 0.7689841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 3.2388348695632176e-06,
-      "loss": 0.9013,
+      "learning_rate": 2.6709996416339468e-06,
+      "loss": 0.8434,
       "step": 27099
     },
     {
-      "epoch": 0.7442805745516465,
+      "epoch": 0.7690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 3.2381795056102337e-06,
-      "loss": 0.8411,
+      "learning_rate": 2.670374390378443e-06,
+      "loss": 0.6745,
       "step": 27100
     },
     {
-      "epoch": 0.7443080387794897,
+      "epoch": 0.7690408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 3.2375241951593948e-06,
-      "loss": 0.9162,
+      "learning_rate": 2.6697492010366165e-06,
+      "loss": 0.8179,
       "step": 27101
     },
     {
-      "epoch": 0.744335503007333,
+      "epoch": 0.7690692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 3.236868938215886e-06,
-      "loss": 0.858,
+      "learning_rate": 2.669124073613748e-06,
+      "loss": 0.7952,
       "step": 27102
     },
     {
-      "epoch": 0.7443629672351761,
+      "epoch": 0.7690976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 3.2362137347848854e-06,
-      "loss": 0.7864,
+      "learning_rate": 2.6684990081151174e-06,
+      "loss": 0.7286,
       "step": 27103
     },
     {
-      "epoch": 0.7443904314630194,
+      "epoch": 0.7691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 3.235558584871582e-06,
-      "loss": 0.7458,
+      "learning_rate": 2.6678740045460085e-06,
+      "loss": 0.8235,
       "step": 27104
     },
     {
-      "epoch": 0.7444178956908627,
+      "epoch": 0.7691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 3.2349034884811593e-06,
-      "loss": 0.8812,
+      "learning_rate": 2.6672490629116954e-06,
+      "loss": 0.8176,
       "step": 27105
     },
     {
-      "epoch": 0.7444453599187059,
+      "epoch": 0.7691827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 3.2342484456188005e-06,
-      "loss": 0.7854,
+      "learning_rate": 2.6666241832174577e-06,
+      "loss": 0.7782,
       "step": 27106
     },
     {
-      "epoch": 0.7444728241465491,
+      "epoch": 0.769211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 3.233593456289693e-06,
-      "loss": 0.836,
+      "learning_rate": 2.665999365468579e-06,
+      "loss": 0.7151,
       "step": 27107
     },
     {
-      "epoch": 0.7445002883743923,
+      "epoch": 0.7692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 3.2329385204990116e-06,
-      "loss": 0.7824,
+      "learning_rate": 2.6653746096703305e-06,
+      "loss": 0.8196,
       "step": 27108
     },
     {
-      "epoch": 0.7445277526022356,
+      "epoch": 0.7692678774120317,
       "grad_norm": 0.0,
-      "learning_rate": 3.2322836382519407e-06,
-      "loss": 0.822,
+      "learning_rate": 2.664749915827992e-06,
+      "loss": 0.7088,
       "step": 27109
     },
     {
-      "epoch": 0.7445552168300789,
+      "epoch": 0.7692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 3.2316288095536672e-06,
-      "loss": 0.806,
+      "learning_rate": 2.6641252839468435e-06,
+      "loss": 0.8242,
       "step": 27110
     },
     {
-      "epoch": 0.744582681057922,
+      "epoch": 0.7693246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 3.230974034409365e-06,
-      "loss": 0.8649,
+      "learning_rate": 2.6635007140321557e-06,
+      "loss": 0.8295,
       "step": 27111
     },
     {
-      "epoch": 0.7446101452857653,
+      "epoch": 0.7693530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 3.2303193128242194e-06,
-      "loss": 0.8954,
+      "learning_rate": 2.6628762060892076e-06,
+      "loss": 0.8818,
       "step": 27112
     },
     {
-      "epoch": 0.7446376095136086,
+      "epoch": 0.7693813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 3.229664644803412e-06,
-      "loss": 0.8081,
+      "learning_rate": 2.6622517601232766e-06,
+      "loss": 0.7572,
       "step": 27113
     },
     {
-      "epoch": 0.7446650737414517,
+      "epoch": 0.7694097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 3.2290100303521176e-06,
-      "loss": 0.8716,
+      "learning_rate": 2.6616273761396315e-06,
+      "loss": 0.8951,
       "step": 27114
     },
     {
-      "epoch": 0.744692537969295,
+      "epoch": 0.7694381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 3.2283554694755192e-06,
-      "loss": 0.805,
+      "learning_rate": 2.6610030541435504e-06,
+      "loss": 0.7898,
       "step": 27115
     },
     {
-      "epoch": 0.7447200021971382,
+      "epoch": 0.7694665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 3.2277009621787946e-06,
-      "loss": 0.8851,
+      "learning_rate": 2.6603787941403092e-06,
+      "loss": 0.812,
       "step": 27116
     },
     {
-      "epoch": 0.7447474664249815,
+      "epoch": 0.7694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 3.2270465084671267e-06,
-      "loss": 0.8748,
+      "learning_rate": 2.6597545961351733e-06,
+      "loss": 0.7844,
       "step": 27117
     },
     {
-      "epoch": 0.7447749306528247,
+      "epoch": 0.7695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 3.226392108345687e-06,
-      "loss": 0.9047,
+      "learning_rate": 2.6591304601334247e-06,
+      "loss": 0.6546,
       "step": 27118
     },
     {
-      "epoch": 0.7448023948806679,
+      "epoch": 0.7695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 3.2257377618196606e-06,
-      "loss": 0.7815,
+      "learning_rate": 2.6585063861403293e-06,
+      "loss": 0.8402,
       "step": 27119
     },
     {
-      "epoch": 0.7448298591085112,
+      "epoch": 0.7695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 3.225083468894217e-06,
-      "loss": 0.7992,
+      "learning_rate": 2.6578823741611593e-06,
+      "loss": 0.7909,
       "step": 27120
     },
     {
-      "epoch": 0.7448573233363543,
+      "epoch": 0.7696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 3.2244292295745383e-06,
-      "loss": 0.8859,
+      "learning_rate": 2.657258424201191e-06,
+      "loss": 0.8094,
       "step": 27121
     },
     {
-      "epoch": 0.7448847875641976,
+      "epoch": 0.7696367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 3.2237750438657987e-06,
-      "loss": 0.8203,
+      "learning_rate": 2.6566345362656876e-06,
+      "loss": 0.8309,
       "step": 27122
     },
     {
-      "epoch": 0.7449122517920409,
+      "epoch": 0.7696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 3.223120911773178e-06,
-      "loss": 0.9229,
+      "learning_rate": 2.656010710359922e-06,
+      "loss": 0.7928,
       "step": 27123
     },
     {
-      "epoch": 0.7449397160198841,
+      "epoch": 0.7696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 3.222466833301848e-06,
-      "loss": 0.8222,
+      "learning_rate": 2.655386946489167e-06,
+      "loss": 0.9464,
       "step": 27124
     },
     {
-      "epoch": 0.7449671802477273,
+      "epoch": 0.76972190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.2218128084569888e-06,
-      "loss": 0.7523,
+      "learning_rate": 2.654763244658686e-06,
+      "loss": 0.8681,
       "step": 27125
     },
     {
-      "epoch": 0.7449946444755706,
+      "epoch": 0.7697502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 3.2211588372437742e-06,
-      "loss": 0.8668,
+      "learning_rate": 2.65413960487375e-06,
+      "loss": 0.7304,
       "step": 27126
     },
     {
-      "epoch": 0.7450221087034138,
+      "epoch": 0.7697786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 3.2205049196673723e-06,
-      "loss": 0.8468,
+      "learning_rate": 2.65351602713963e-06,
+      "loss": 0.8241,
       "step": 27127
     },
     {
-      "epoch": 0.7450495729312571,
+      "epoch": 0.7698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 3.2198510557329608e-06,
-      "loss": 0.8122,
+      "learning_rate": 2.6528925114615877e-06,
+      "loss": 0.7967,
       "step": 27128
     },
     {
-      "epoch": 0.7450770371591002,
+      "epoch": 0.7698354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 3.2191972454457156e-06,
-      "loss": 0.8362,
+      "learning_rate": 2.652269057844892e-06,
+      "loss": 0.8688,
       "step": 27129
     },
     {
-      "epoch": 0.7451045013869435,
+      "epoch": 0.7698637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 3.2185434888108104e-06,
-      "loss": 0.85,
+      "learning_rate": 2.651645666294813e-06,
+      "loss": 0.8008,
       "step": 27130
     },
     {
-      "epoch": 0.7451319656147868,
+      "epoch": 0.7698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 3.2178897858334133e-06,
-      "loss": 0.8126,
+      "learning_rate": 2.651022336816611e-06,
+      "loss": 0.7011,
       "step": 27131
     },
     {
-      "epoch": 0.74515942984263,
+      "epoch": 0.7699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 3.2172361365186998e-06,
-      "loss": 0.8709,
+      "learning_rate": 2.6503990694155522e-06,
+      "loss": 0.8064,
       "step": 27132
     },
     {
-      "epoch": 0.7451868940704732,
+      "epoch": 0.7699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 3.216582540871841e-06,
-      "loss": 0.8409,
+      "learning_rate": 2.649775864096904e-06,
+      "loss": 0.8174,
       "step": 27133
     },
     {
-      "epoch": 0.7452143582983164,
+      "epoch": 0.7699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 3.2159289988980124e-06,
-      "loss": 0.8015,
+      "learning_rate": 2.6491527208659294e-06,
+      "loss": 0.8197,
       "step": 27134
     },
     {
-      "epoch": 0.7452418225261597,
+      "epoch": 0.770005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 3.215275510602378e-06,
-      "loss": 0.8026,
+      "learning_rate": 2.648529639727896e-06,
+      "loss": 0.819,
       "step": 27135
     },
     {
-      "epoch": 0.745269286754003,
+      "epoch": 0.7700340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 3.214622075990115e-06,
-      "loss": 0.8443,
+      "learning_rate": 2.647906620688059e-06,
+      "loss": 0.6771,
       "step": 27136
     },
     {
-      "epoch": 0.7452967509818461,
+      "epoch": 0.7700624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 3.213968695066387e-06,
-      "loss": 0.7816,
+      "learning_rate": 2.647283663751685e-06,
+      "loss": 0.8058,
       "step": 27137
     },
     {
-      "epoch": 0.7453242152096894,
+      "epoch": 0.7700908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 3.213315367836368e-06,
-      "loss": 0.8509,
+      "learning_rate": 2.646660768924041e-06,
+      "loss": 0.9134,
       "step": 27138
     },
     {
-      "epoch": 0.7453516794375327,
+      "epoch": 0.7701191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 3.2126620943052267e-06,
-      "loss": 0.7447,
+      "learning_rate": 2.6460379362103794e-06,
+      "loss": 0.8322,
       "step": 27139
     },
     {
-      "epoch": 0.7453791436653758,
+      "epoch": 0.7701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 3.2120088744781307e-06,
-      "loss": 0.8707,
+      "learning_rate": 2.6454151656159666e-06,
+      "loss": 0.8004,
       "step": 27140
     },
     {
-      "epoch": 0.7454066078932191,
+      "epoch": 0.7701759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 3.2113557083602543e-06,
-      "loss": 0.824,
+      "learning_rate": 2.644792457146066e-06,
+      "loss": 0.8912,
       "step": 27141
     },
     {
-      "epoch": 0.7454340721210623,
+      "epoch": 0.7702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 3.2107025959567616e-06,
-      "loss": 0.7042,
+      "learning_rate": 2.64416981080593e-06,
+      "loss": 0.7582,
       "step": 27142
     },
     {
-      "epoch": 0.7454615363489056,
+      "epoch": 0.7702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 3.210049537272816e-06,
-      "loss": 0.8824,
+      "learning_rate": 2.643547226600823e-06,
+      "loss": 0.7104,
       "step": 27143
     },
     {
-      "epoch": 0.7454890005767488,
+      "epoch": 0.770261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.209396532313589e-06,
-      "loss": 0.7998,
+      "learning_rate": 2.6429247045360062e-06,
+      "loss": 0.7479,
       "step": 27144
     },
     {
-      "epoch": 0.745516464804592,
+      "epoch": 0.7702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.208743581084247e-06,
-      "loss": 0.8646,
+      "learning_rate": 2.6423022446167325e-06,
+      "loss": 0.828,
       "step": 27145
     },
     {
-      "epoch": 0.7455439290324353,
+      "epoch": 0.7703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 3.208090683589956e-06,
-      "loss": 0.8169,
+      "learning_rate": 2.641679846848262e-06,
+      "loss": 0.8878,
       "step": 27146
     },
     {
-      "epoch": 0.7455713932602784,
+      "epoch": 0.7703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 3.207437839835883e-06,
-      "loss": 0.7909,
+      "learning_rate": 2.641057511235856e-06,
+      "loss": 0.7299,
       "step": 27147
     },
     {
-      "epoch": 0.7455988574881217,
+      "epoch": 0.7703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 3.2067850498271957e-06,
-      "loss": 0.8679,
+      "learning_rate": 2.640435237784762e-06,
+      "loss": 0.8321,
       "step": 27148
     },
     {
-      "epoch": 0.745626321715965,
+      "epoch": 0.7704029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 3.2061323135690536e-06,
-      "loss": 0.8219,
+      "learning_rate": 2.6398130265002476e-06,
+      "loss": 0.7754,
       "step": 27149
     },
     {
-      "epoch": 0.7456537859438082,
+      "epoch": 0.7704313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 3.205479631066628e-06,
-      "loss": 0.8957,
+      "learning_rate": 2.6391908773875607e-06,
+      "loss": 0.8629,
       "step": 27150
     },
     {
-      "epoch": 0.7456812501716514,
+      "epoch": 0.7704597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 3.204827002325076e-06,
-      "loss": 0.8782,
+      "learning_rate": 2.6385687904519595e-06,
+      "loss": 0.8216,
       "step": 27151
     },
     {
-      "epoch": 0.7457087143994947,
+      "epoch": 0.7704880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 3.2041744273495644e-06,
-      "loss": 0.7686,
+      "learning_rate": 2.6379467656987025e-06,
+      "loss": 0.8189,
       "step": 27152
     },
     {
-      "epoch": 0.7457361786273379,
+      "epoch": 0.770516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 3.2035219061452572e-06,
-      "loss": 0.8353,
+      "learning_rate": 2.6373248031330367e-06,
+      "loss": 0.7701,
       "step": 27153
     },
     {
-      "epoch": 0.7457636428551812,
+      "epoch": 0.770544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 3.202869438717321e-06,
-      "loss": 0.7938,
+      "learning_rate": 2.6367029027602187e-06,
+      "loss": 0.8334,
       "step": 27154
     },
     {
-      "epoch": 0.7457911070830243,
+      "epoch": 0.7705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 3.2022170250709117e-06,
-      "loss": 0.8724,
+      "learning_rate": 2.636081064585505e-06,
+      "loss": 0.9233,
       "step": 27155
     },
     {
-      "epoch": 0.7458185713108676,
+      "epoch": 0.7706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 3.201564665211194e-06,
-      "loss": 0.7447,
+      "learning_rate": 2.635459288614144e-06,
+      "loss": 0.8252,
       "step": 27156
     },
     {
-      "epoch": 0.7458460355387109,
+      "epoch": 0.7706299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 3.2009123591433342e-06,
-      "loss": 0.6865,
+      "learning_rate": 2.6348375748513887e-06,
+      "loss": 0.7732,
       "step": 27157
     },
     {
-      "epoch": 0.745873499766554,
+      "epoch": 0.7706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 3.200260106872486e-06,
-      "loss": 0.8969,
+      "learning_rate": 2.6342159233024945e-06,
+      "loss": 0.802,
       "step": 27158
     },
     {
-      "epoch": 0.7459009639943973,
+      "epoch": 0.7706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 3.1996079084038134e-06,
-      "loss": 0.8433,
+      "learning_rate": 2.6335943339727053e-06,
+      "loss": 0.8268,
       "step": 27159
     },
     {
-      "epoch": 0.7459284282222405,
+      "epoch": 0.7707150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 3.1989557637424806e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.6329728068672777e-06,
+      "loss": 0.8485,
       "step": 27160
     },
     {
-      "epoch": 0.7459558924500838,
+      "epoch": 0.7707434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 3.198303672893641e-06,
-      "loss": 0.8532,
+      "learning_rate": 2.6323513419914626e-06,
+      "loss": 0.8684,
       "step": 27161
     },
     {
-      "epoch": 0.745983356677927,
+      "epoch": 0.7707718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 3.1976516358624577e-06,
-      "loss": 0.7656,
+      "learning_rate": 2.6317299393505035e-06,
+      "loss": 0.7707,
       "step": 27162
     },
     {
-      "epoch": 0.7460108209057702,
+      "epoch": 0.7708002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 3.196999652654089e-06,
-      "loss": 0.7992,
+      "learning_rate": 2.63110859894965e-06,
+      "loss": 0.8657,
       "step": 27163
     },
     {
-      "epoch": 0.7460382851336135,
+      "epoch": 0.7708286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 3.196347723273695e-06,
-      "loss": 0.8666,
+      "learning_rate": 2.6304873207941584e-06,
+      "loss": 0.8538,
       "step": 27164
     },
     {
-      "epoch": 0.7460657493614568,
+      "epoch": 0.7708569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 3.1956958477264367e-06,
-      "loss": 0.8333,
+      "learning_rate": 2.6298661048892705e-06,
+      "loss": 0.7897,
       "step": 27165
     },
     {
-      "epoch": 0.7460932135892999,
+      "epoch": 0.7708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 3.1950440260174677e-06,
-      "loss": 0.8489,
+      "learning_rate": 2.629244951240234e-06,
+      "loss": 0.8631,
       "step": 27166
     },
     {
-      "epoch": 0.7461206778171432,
+      "epoch": 0.7709137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 3.1943922581519437e-06,
-      "loss": 0.8263,
+      "learning_rate": 2.6286238598522993e-06,
+      "loss": 0.711,
       "step": 27167
     },
     {
-      "epoch": 0.7461481420449864,
+      "epoch": 0.7709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 3.1937405441350244e-06,
-      "loss": 0.8088,
+      "learning_rate": 2.628002830730708e-06,
+      "loss": 0.8413,
       "step": 27168
     },
     {
-      "epoch": 0.7461756062728296,
+      "epoch": 0.7709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 3.1930888839718665e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.62738186388071e-06,
+      "loss": 0.8334,
       "step": 27169
     },
     {
-      "epoch": 0.7462030705006729,
+      "epoch": 0.7709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 3.192437277667626e-06,
-      "loss": 0.8543,
+      "learning_rate": 2.626760959307547e-06,
+      "loss": 0.8353,
       "step": 27170
     },
     {
-      "epoch": 0.7462305347285161,
+      "epoch": 0.7710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 3.1917857252274587e-06,
-      "loss": 0.7197,
+      "learning_rate": 2.6261401170164658e-06,
+      "loss": 0.8169,
       "step": 27171
     },
     {
-      "epoch": 0.7462579989563594,
+      "epoch": 0.77105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.1911342266565226e-06,
-      "loss": 0.8593,
+      "learning_rate": 2.625519337012713e-06,
+      "loss": 0.8423,
       "step": 27172
     },
     {
-      "epoch": 0.7462854631842025,
+      "epoch": 0.7710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 3.190482781959967e-06,
-      "loss": 0.7744,
+      "learning_rate": 2.624898619301527e-06,
+      "loss": 0.8547,
       "step": 27173
     },
     {
-      "epoch": 0.7463129274120458,
+      "epoch": 0.7711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 3.1898313911429533e-06,
-      "loss": 0.754,
+      "learning_rate": 2.6242779638881532e-06,
+      "loss": 0.7989,
       "step": 27174
     },
     {
-      "epoch": 0.7463403916398891,
+      "epoch": 0.7711407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 3.189180054210629e-06,
-      "loss": 0.8442,
+      "learning_rate": 2.623657370777839e-06,
+      "loss": 0.8319,
       "step": 27175
     },
     {
-      "epoch": 0.7463678558677322,
+      "epoch": 0.7711691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 3.188528771168149e-06,
-      "loss": 0.8387,
+      "learning_rate": 2.623036839975819e-06,
+      "loss": 0.7908,
       "step": 27176
     },
     {
-      "epoch": 0.7463953200955755,
+      "epoch": 0.7711975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 3.1878775420206685e-06,
-      "loss": 0.7795,
+      "learning_rate": 2.6224163714873386e-06,
+      "loss": 0.821,
       "step": 27177
     },
     {
-      "epoch": 0.7464227843234188,
+      "epoch": 0.7712258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 3.1872263667733427e-06,
-      "loss": 0.7959,
+      "learning_rate": 2.6217959653176396e-06,
+      "loss": 0.7313,
       "step": 27178
     },
     {
-      "epoch": 0.746450248551262,
+      "epoch": 0.7712542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 3.1865752454313172e-06,
-      "loss": 0.8869,
+      "learning_rate": 2.621175621471961e-06,
+      "loss": 0.8063,
       "step": 27179
     },
     {
-      "epoch": 0.7464777127791052,
+      "epoch": 0.7712826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 3.1859241779997485e-06,
-      "loss": 0.8292,
+      "learning_rate": 2.6205553399555428e-06,
+      "loss": 0.8337,
       "step": 27180
     },
     {
-      "epoch": 0.7465051770069484,
+      "epoch": 0.7713110102156641,
       "grad_norm": 0.0,
-      "learning_rate": 3.18527316448379e-06,
-      "loss": 0.877,
+      "learning_rate": 2.619935120773629e-06,
+      "loss": 0.8228,
       "step": 27181
     },
     {
-      "epoch": 0.7465326412347917,
+      "epoch": 0.7713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 3.1846222048885867e-06,
-      "loss": 0.9272,
+      "learning_rate": 2.619314963931452e-06,
+      "loss": 0.762,
       "step": 27182
     },
     {
-      "epoch": 0.746560105462635,
+      "epoch": 0.7713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 3.1839712992192927e-06,
-      "loss": 0.8336,
+      "learning_rate": 2.618694869434254e-06,
+      "loss": 0.8497,
       "step": 27183
     },
     {
-      "epoch": 0.7465875696904781,
+      "epoch": 0.7713961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 3.1833204474810607e-06,
-      "loss": 0.6886,
+      "learning_rate": 2.6180748372872743e-06,
+      "loss": 0.8253,
       "step": 27184
     },
     {
-      "epoch": 0.7466150339183214,
+      "epoch": 0.7714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 3.1826696496790353e-06,
-      "loss": 0.8772,
+      "learning_rate": 2.6174548674957457e-06,
+      "loss": 0.868,
       "step": 27185
     },
     {
-      "epoch": 0.7466424981461646,
+      "epoch": 0.7714528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 3.182018905818368e-06,
-      "loss": 0.8037,
+      "learning_rate": 2.616834960064908e-06,
+      "loss": 0.7862,
       "step": 27186
     },
     {
-      "epoch": 0.7466699623740078,
+      "epoch": 0.7714812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 3.181368215904209e-06,
-      "loss": 0.8374,
+      "learning_rate": 2.616215114999999e-06,
+      "loss": 0.9896,
       "step": 27187
     },
     {
-      "epoch": 0.7466974266018511,
+      "epoch": 0.7715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 3.1807175799417056e-06,
-      "loss": 0.8717,
+      "learning_rate": 2.615595332306251e-06,
+      "loss": 0.8488,
       "step": 27188
     },
     {
-      "epoch": 0.7467248908296943,
+      "epoch": 0.7715380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 3.180066997936009e-06,
-      "loss": 0.7935,
+      "learning_rate": 2.6149756119889013e-06,
+      "loss": 0.8521,
       "step": 27189
     },
     {
-      "epoch": 0.7467523550575376,
+      "epoch": 0.771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 3.179416469892265e-06,
-      "loss": 0.8842,
+      "learning_rate": 2.614355954053187e-06,
+      "loss": 0.8013,
       "step": 27190
     },
     {
-      "epoch": 0.7467798192853808,
+      "epoch": 0.771594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 3.178765995815616e-06,
-      "loss": 0.8267,
+      "learning_rate": 2.6137363585043374e-06,
+      "loss": 0.7583,
       "step": 27191
     },
     {
-      "epoch": 0.746807283513224,
+      "epoch": 0.7716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 3.1781155757112125e-06,
-      "loss": 0.8623,
+      "learning_rate": 2.6131168253475892e-06,
+      "loss": 0.7315,
       "step": 27192
     },
     {
-      "epoch": 0.7468347477410673,
+      "epoch": 0.7716515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 3.177465209584202e-06,
-      "loss": 0.8759,
+      "learning_rate": 2.612497354588177e-06,
+      "loss": 0.8946,
       "step": 27193
     },
     {
-      "epoch": 0.7468622119689104,
+      "epoch": 0.7716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 3.1768148974397282e-06,
-      "loss": 0.8827,
+      "learning_rate": 2.611877946231327e-06,
+      "loss": 0.7252,
       "step": 27194
     },
     {
-      "epoch": 0.7468896761967537,
+      "epoch": 0.7717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 3.176164639282943e-06,
-      "loss": 0.7835,
+      "learning_rate": 2.6112586002822814e-06,
+      "loss": 0.8285,
       "step": 27195
     },
     {
-      "epoch": 0.746917140424597,
+      "epoch": 0.7717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.1755144351189813e-06,
-      "loss": 0.7906,
+      "learning_rate": 2.610639316746263e-06,
+      "loss": 0.8317,
       "step": 27196
     },
     {
-      "epoch": 0.7469446046524402,
+      "epoch": 0.7717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 3.174864284952994e-06,
-      "loss": 0.8184,
+      "learning_rate": 2.610020095628507e-06,
+      "loss": 1.0211,
       "step": 27197
     },
     {
-      "epoch": 0.7469720688802834,
+      "epoch": 0.7717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 3.1742141887901266e-06,
-      "loss": 0.8837,
+      "learning_rate": 2.6094009369342477e-06,
+      "loss": 0.7877,
       "step": 27198
     },
     {
-      "epoch": 0.7469995331081266,
+      "epoch": 0.7718217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 3.1735641466355192e-06,
-      "loss": 0.8717,
+      "learning_rate": 2.608781840668706e-06,
+      "loss": 0.7776,
       "step": 27199
     },
     {
-      "epoch": 0.7470269973359699,
+      "epoch": 0.771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 3.1729141584943158e-06,
-      "loss": 0.9444,
+      "learning_rate": 2.608162806837118e-06,
+      "loss": 0.8496,
       "step": 27200
     },
     {
-      "epoch": 0.7470544615638132,
+      "epoch": 0.7718785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 3.1722642243716627e-06,
-      "loss": 0.7486,
+      "learning_rate": 2.6075438354447125e-06,
+      "loss": 0.8629,
       "step": 27201
     },
     {
-      "epoch": 0.7470819257916563,
+      "epoch": 0.7719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 3.1716143442726965e-06,
-      "loss": 0.9133,
+      "learning_rate": 2.6069249264967145e-06,
+      "loss": 0.8516,
       "step": 27202
     },
     {
-      "epoch": 0.7471093900194996,
+      "epoch": 0.7719353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 3.1709645182025637e-06,
-      "loss": 0.8119,
+      "learning_rate": 2.606306079998354e-06,
+      "loss": 0.769,
       "step": 27203
     },
     {
-      "epoch": 0.7471368542473429,
+      "epoch": 0.7719636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 3.170314746166404e-06,
-      "loss": 0.7964,
+      "learning_rate": 2.6056872959548607e-06,
+      "loss": 0.8009,
       "step": 27204
     },
     {
-      "epoch": 0.747164318475186,
+      "epoch": 0.7719920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 3.169665028169363e-06,
-      "loss": 0.8182,
+      "learning_rate": 2.6050685743714565e-06,
+      "loss": 0.8627,
       "step": 27205
     },
     {
-      "epoch": 0.7471917827030293,
+      "epoch": 0.7720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 3.1690153642165756e-06,
-      "loss": 0.9347,
+      "learning_rate": 2.6044499152533708e-06,
+      "loss": 0.7733,
       "step": 27206
     },
     {
-      "epoch": 0.7472192469308725,
+      "epoch": 0.7720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 3.1683657543131876e-06,
-      "loss": 0.7999,
+      "learning_rate": 2.6038313186058317e-06,
+      "loss": 0.6891,
       "step": 27207
     },
     {
-      "epoch": 0.7472467111587158,
+      "epoch": 0.7720771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 3.1677161984643323e-06,
-      "loss": 0.7886,
+      "learning_rate": 2.603212784434059e-06,
+      "loss": 0.8109,
       "step": 27208
     },
     {
-      "epoch": 0.747274175386559,
+      "epoch": 0.772105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.167066696675154e-06,
-      "loss": 0.7357,
+      "learning_rate": 2.6025943127432807e-06,
+      "loss": 0.757,
       "step": 27209
     },
     {
-      "epoch": 0.7473016396144022,
+      "epoch": 0.7721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 3.1664172489507904e-06,
-      "loss": 0.8022,
+      "learning_rate": 2.6019759035387216e-06,
+      "loss": 0.7387,
       "step": 27210
     },
     {
-      "epoch": 0.7473291038422455,
+      "epoch": 0.7721623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 3.16576785529638e-06,
-      "loss": 0.8771,
+      "learning_rate": 2.6013575568256034e-06,
+      "loss": 0.7802,
       "step": 27211
     },
     {
-      "epoch": 0.7473565680700887,
+      "epoch": 0.7721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 3.165118515717063e-06,
-      "loss": 0.8113,
+      "learning_rate": 2.600739272609154e-06,
+      "loss": 0.8686,
       "step": 27212
     },
     {
-      "epoch": 0.7473840322979319,
+      "epoch": 0.7722190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 3.164469230217978e-06,
-      "loss": 0.9621,
+      "learning_rate": 2.6001210508945894e-06,
+      "loss": 0.8383,
       "step": 27213
     },
     {
-      "epoch": 0.7474114965257752,
+      "epoch": 0.7722474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 3.1638199988042608e-06,
-      "loss": 0.7453,
+      "learning_rate": 2.5995028916871346e-06,
+      "loss": 0.9129,
       "step": 27214
     },
     {
-      "epoch": 0.7474389607536184,
+      "epoch": 0.7722758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 3.163170821481044e-06,
-      "loss": 0.8786,
+      "learning_rate": 2.598884794992015e-06,
+      "loss": 0.8632,
       "step": 27215
     },
     {
-      "epoch": 0.7474664249814617,
+      "epoch": 0.7723041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 3.1625216982534678e-06,
-      "loss": 0.8898,
+      "learning_rate": 2.5982667608144454e-06,
+      "loss": 0.875,
       "step": 27216
     },
     {
-      "epoch": 0.7474938892093049,
+      "epoch": 0.7723325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 3.16187262912667e-06,
-      "loss": 0.7869,
+      "learning_rate": 2.597648789159648e-06,
+      "loss": 0.7911,
       "step": 27217
     },
     {
-      "epoch": 0.7475213534371481,
+      "epoch": 0.772360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 3.1612236141057828e-06,
-      "loss": 0.8281,
+      "learning_rate": 2.597030880032848e-06,
+      "loss": 0.804,
       "step": 27218
     },
     {
-      "epoch": 0.7475488176649914,
+      "epoch": 0.77238933030647,
       "grad_norm": 0.0,
-      "learning_rate": 3.160574653195948e-06,
-      "loss": 0.8137,
+      "learning_rate": 2.5964130334392577e-06,
+      "loss": 0.8277,
       "step": 27219
     },
     {
-      "epoch": 0.7475762818928345,
+      "epoch": 0.7724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 3.1599257464022924e-06,
-      "loss": 0.7879,
+      "learning_rate": 2.5957952493841e-06,
+      "loss": 0.7563,
       "step": 27220
     },
     {
-      "epoch": 0.7476037461206778,
+      "epoch": 0.7724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 3.1592768937299547e-06,
-      "loss": 0.8657,
+      "learning_rate": 2.5951775278725956e-06,
+      "loss": 0.9776,
       "step": 27221
     },
     {
-      "epoch": 0.7476312103485211,
+      "epoch": 0.7724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 3.1586280951840697e-06,
-      "loss": 0.7446,
+      "learning_rate": 2.594559868909956e-06,
+      "loss": 0.8005,
       "step": 27222
     },
     {
-      "epoch": 0.7476586745763643,
+      "epoch": 0.7725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 3.1579793507697674e-06,
-      "loss": 0.8619,
+      "learning_rate": 2.593942272501402e-06,
+      "loss": 0.8807,
       "step": 27223
     },
     {
-      "epoch": 0.7476861388042075,
+      "epoch": 0.7725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 3.1573306604921817e-06,
-      "loss": 0.83,
+      "learning_rate": 2.593324738652151e-06,
+      "loss": 0.8559,
       "step": 27224
     },
     {
-      "epoch": 0.7477136030320507,
+      "epoch": 0.7725595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 3.15668202435645e-06,
-      "loss": 0.9183,
+      "learning_rate": 2.592707267367418e-06,
+      "loss": 0.8107,
       "step": 27225
     },
     {
-      "epoch": 0.747741067259894,
+      "epoch": 0.7725879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 3.1560334423676986e-06,
-      "loss": 0.7445,
+      "learning_rate": 2.5920898586524233e-06,
+      "loss": 0.7212,
       "step": 27226
     },
     {
-      "epoch": 0.7477685314877373,
+      "epoch": 0.772616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 3.1553849145310613e-06,
-      "loss": 0.8243,
+      "learning_rate": 2.5914725125123765e-06,
+      "loss": 0.8502,
       "step": 27227
     },
     {
-      "epoch": 0.7477959957155804,
+      "epoch": 0.7726447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 3.154736440851669e-06,
-      "loss": 0.9442,
+      "learning_rate": 2.5908552289524926e-06,
+      "loss": 0.817,
       "step": 27228
     },
     {
-      "epoch": 0.7478234599434237,
+      "epoch": 0.7726730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 3.154088021334658e-06,
-      "loss": 0.819,
+      "learning_rate": 2.590238007977992e-06,
+      "loss": 0.7779,
       "step": 27229
     },
     {
-      "epoch": 0.747850924171267,
+      "epoch": 0.7727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 3.153439655985151e-06,
-      "loss": 0.8074,
+      "learning_rate": 2.58962084959408e-06,
+      "loss": 0.8183,
       "step": 27230
     },
     {
-      "epoch": 0.7478783883991101,
+      "epoch": 0.7727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 3.152791344808284e-06,
-      "loss": 0.8642,
+      "learning_rate": 2.5890037538059744e-06,
+      "loss": 0.8815,
       "step": 27231
     },
     {
-      "epoch": 0.7479058526269534,
+      "epoch": 0.7727582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 3.152143087809182e-06,
-      "loss": 0.8797,
+      "learning_rate": 2.5883867206188907e-06,
+      "loss": 0.8499,
       "step": 27232
     },
     {
-      "epoch": 0.7479333168547966,
+      "epoch": 0.7727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 3.151494884992975e-06,
-      "loss": 0.8604,
+      "learning_rate": 2.5877697500380337e-06,
+      "loss": 0.7957,
       "step": 27233
     },
     {
-      "epoch": 0.7479607810826399,
+      "epoch": 0.7728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 3.1508467363647943e-06,
-      "loss": 0.8186,
+      "learning_rate": 2.5871528420686196e-06,
+      "loss": 0.8114,
       "step": 27234
     },
     {
-      "epoch": 0.7479882453104831,
+      "epoch": 0.7728433598183881,
       "grad_norm": 0.0,
-      "learning_rate": 3.150198641929768e-06,
-      "loss": 0.7066,
+      "learning_rate": 2.586535996715861e-06,
+      "loss": 0.902,
       "step": 27235
     },
     {
-      "epoch": 0.7480157095383263,
+      "epoch": 0.7728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 3.149550601693022e-06,
-      "loss": 0.7446,
+      "learning_rate": 2.585919213984963e-06,
+      "loss": 0.9553,
       "step": 27236
     },
     {
-      "epoch": 0.7480431737661696,
+      "epoch": 0.772900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 3.148902615659689e-06,
-      "loss": 0.8449,
+      "learning_rate": 2.58530249388114e-06,
+      "loss": 0.804,
       "step": 27237
     },
     {
-      "epoch": 0.7480706379940127,
+      "epoch": 0.7729284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 3.1482546838348925e-06,
-      "loss": 0.8239,
+      "learning_rate": 2.5846858364096017e-06,
+      "loss": 0.7819,
       "step": 27238
     },
     {
-      "epoch": 0.748098102221856,
+      "epoch": 0.7729568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 3.1476068062237565e-06,
-      "loss": 0.7743,
+      "learning_rate": 2.584069241575553e-06,
+      "loss": 0.7981,
       "step": 27239
     },
     {
-      "epoch": 0.7481255664496993,
+      "epoch": 0.7729852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 3.1469589828314096e-06,
-      "loss": 0.8427,
+      "learning_rate": 2.583452709384204e-06,
+      "loss": 0.7789,
       "step": 27240
     },
     {
-      "epoch": 0.7481530306775425,
+      "epoch": 0.7730136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 3.146311213662977e-06,
-      "loss": 0.7743,
+      "learning_rate": 2.582836239840765e-06,
+      "loss": 0.8091,
       "step": 27241
     },
     {
-      "epoch": 0.7481804949053857,
+      "epoch": 0.7730419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 3.1456634987235857e-06,
-      "loss": 0.8067,
+      "learning_rate": 2.5822198329504412e-06,
+      "loss": 0.8266,
       "step": 27242
     },
     {
-      "epoch": 0.748207959133229,
+      "epoch": 0.7730703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 3.1450158380183624e-06,
-      "loss": 0.7932,
+      "learning_rate": 2.5816034887184414e-06,
+      "loss": 0.8001,
       "step": 27243
     },
     {
-      "epoch": 0.7482354233610722,
+      "epoch": 0.7730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 3.144368231552426e-06,
-      "loss": 0.855,
+      "learning_rate": 2.5809872071499687e-06,
+      "loss": 0.8986,
       "step": 27244
     },
     {
-      "epoch": 0.7482628875889155,
+      "epoch": 0.7731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 3.143720679330905e-06,
-      "loss": 0.7723,
+      "learning_rate": 2.58037098825023e-06,
+      "loss": 0.8534,
       "step": 27245
     },
     {
-      "epoch": 0.7482903518167586,
+      "epoch": 0.773155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 3.143073181358923e-06,
-      "loss": 0.7452,
+      "learning_rate": 2.579754832024436e-06,
+      "loss": 0.748,
       "step": 27246
     },
     {
-      "epoch": 0.7483178160446019,
+      "epoch": 0.7731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 3.142425737641601e-06,
-      "loss": 0.8631,
+      "learning_rate": 2.579138738477781e-06,
+      "loss": 0.7811,
       "step": 27247
     },
     {
-      "epoch": 0.7483452802724452,
+      "epoch": 0.7732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 3.1417783481840615e-06,
-      "loss": 0.869,
+      "learning_rate": 2.578522707615476e-06,
+      "loss": 0.8419,
       "step": 27248
     },
     {
-      "epoch": 0.7483727445002883,
+      "epoch": 0.7732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 3.141131012991433e-06,
-      "loss": 0.8325,
+      "learning_rate": 2.577906739442726e-06,
+      "loss": 0.8197,
       "step": 27249
     },
     {
-      "epoch": 0.7484002087281316,
+      "epoch": 0.7732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 3.1404837320688287e-06,
-      "loss": 0.8048,
+      "learning_rate": 2.5772908339647286e-06,
+      "loss": 0.8024,
       "step": 27250
     },
     {
-      "epoch": 0.7484276729559748,
+      "epoch": 0.7732973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 3.1398365054213732e-06,
-      "loss": 0.9618,
+      "learning_rate": 2.5766749911866897e-06,
+      "loss": 0.8326,
       "step": 27251
     },
     {
-      "epoch": 0.7484551371838181,
+      "epoch": 0.7733257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 3.13918933305419e-06,
-      "loss": 0.8644,
+      "learning_rate": 2.5760592111138126e-06,
+      "loss": 0.769,
       "step": 27252
     },
     {
-      "epoch": 0.7484826014116613,
+      "epoch": 0.7733541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 3.138542214972401e-06,
-      "loss": 0.8105,
+      "learning_rate": 2.575443493751294e-06,
+      "loss": 0.7719,
       "step": 27253
     },
     {
-      "epoch": 0.7485100656395045,
+      "epoch": 0.7733825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 3.137895151181124e-06,
-      "loss": 0.7646,
+      "learning_rate": 2.574827839104339e-06,
+      "loss": 0.8403,
       "step": 27254
     },
     {
-      "epoch": 0.7485375298673478,
+      "epoch": 0.773410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 3.137248141685476e-06,
-      "loss": 0.7999,
+      "learning_rate": 2.574212247178147e-06,
+      "loss": 0.8408,
       "step": 27255
     },
     {
-      "epoch": 0.7485649940951911,
+      "epoch": 0.773439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 3.1366011864905786e-06,
-      "loss": 0.7525,
+      "learning_rate": 2.5735967179779174e-06,
+      "loss": 0.9781,
       "step": 27256
     },
     {
-      "epoch": 0.7485924583230342,
+      "epoch": 0.7734676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 3.135954285601551e-06,
-      "loss": 0.8115,
+      "learning_rate": 2.572981251508854e-06,
+      "loss": 0.8123,
       "step": 27257
     },
     {
-      "epoch": 0.7486199225508775,
+      "epoch": 0.7734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 3.1353074390235126e-06,
-      "loss": 0.8569,
+      "learning_rate": 2.572365847776147e-06,
+      "loss": 0.8828,
       "step": 27258
     },
     {
-      "epoch": 0.7486473867787207,
+      "epoch": 0.7735244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 3.1346606467615805e-06,
-      "loss": 0.7564,
+      "learning_rate": 2.5717505067850012e-06,
+      "loss": 0.8857,
       "step": 27259
     },
     {
-      "epoch": 0.748674851006564,
+      "epoch": 0.7735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 3.1340139088208766e-06,
-      "loss": 0.9143,
+      "learning_rate": 2.5711352285406155e-06,
+      "loss": 0.7966,
       "step": 27260
     },
     {
-      "epoch": 0.7487023152344072,
+      "epoch": 0.7735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 3.1333672252065104e-06,
-      "loss": 0.7165,
+      "learning_rate": 2.570520013048181e-06,
+      "loss": 0.7223,
       "step": 27261
     },
     {
-      "epoch": 0.7487297794622504,
+      "epoch": 0.7736095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 3.132720595923605e-06,
-      "loss": 0.847,
+      "learning_rate": 2.5699048603128986e-06,
+      "loss": 0.8299,
       "step": 27262
     },
     {
-      "epoch": 0.7487572436900937,
+      "epoch": 0.7736379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 3.1320740209772726e-06,
-      "loss": 0.7536,
+      "learning_rate": 2.5692897703399665e-06,
+      "loss": 0.8049,
       "step": 27263
     },
     {
-      "epoch": 0.7487847079179369,
+      "epoch": 0.7736662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 3.13142750037263e-06,
-      "loss": 0.8085,
+      "learning_rate": 2.568674743134575e-06,
+      "loss": 0.8332,
       "step": 27264
     },
     {
-      "epoch": 0.7488121721457801,
+      "epoch": 0.7736946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 3.1307810341147947e-06,
-      "loss": 0.7292,
+      "learning_rate": 2.5680597787019214e-06,
+      "loss": 0.8245,
       "step": 27265
     },
     {
-      "epoch": 0.7488396363736234,
+      "epoch": 0.7737230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 3.1301346222088833e-06,
-      "loss": 0.8188,
+      "learning_rate": 2.567444877047205e-06,
+      "loss": 0.829,
       "step": 27266
     },
     {
-      "epoch": 0.7488671006014666,
+      "epoch": 0.7737514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 3.1294882646600043e-06,
-      "loss": 0.84,
+      "learning_rate": 2.5668300381756116e-06,
+      "loss": 0.8681,
       "step": 27267
     },
     {
-      "epoch": 0.7488945648293098,
+      "epoch": 0.7737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 3.1288419614732755e-06,
-      "loss": 0.8352,
+      "learning_rate": 2.56621526209234e-06,
+      "loss": 0.8402,
       "step": 27268
     },
     {
-      "epoch": 0.7489220290571531,
+      "epoch": 0.7738081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 3.128195712653812e-06,
-      "loss": 0.8117,
+      "learning_rate": 2.565600548802585e-06,
+      "loss": 0.928,
       "step": 27269
     },
     {
-      "epoch": 0.7489494932849963,
+      "epoch": 0.7738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 3.1275495182067285e-06,
-      "loss": 0.8924,
+      "learning_rate": 2.5649858983115293e-06,
+      "loss": 0.8237,
       "step": 27270
     },
     {
-      "epoch": 0.7489769575128395,
+      "epoch": 0.7738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 3.1269033781371315e-06,
-      "loss": 0.8199,
+      "learning_rate": 2.5643713106243762e-06,
+      "loss": 0.854,
       "step": 27271
     },
     {
-      "epoch": 0.7490044217406827,
+      "epoch": 0.7738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 3.126257292450141e-06,
-      "loss": 0.7769,
+      "learning_rate": 2.5637567857463154e-06,
+      "loss": 0.8331,
       "step": 27272
     },
     {
-      "epoch": 0.749031885968526,
+      "epoch": 0.7739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 3.1256112611508615e-06,
-      "loss": 0.8447,
+      "learning_rate": 2.5631423236825328e-06,
+      "loss": 0.7534,
       "step": 27273
     },
     {
-      "epoch": 0.7490593501963693,
+      "epoch": 0.773950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 3.12496528424441e-06,
-      "loss": 0.9131,
+      "learning_rate": 2.5625279244382206e-06,
+      "loss": 0.7828,
       "step": 27274
     },
     {
-      "epoch": 0.7490868144242124,
+      "epoch": 0.7739784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 3.1243193617358947e-06,
-      "loss": 0.8697,
+      "learning_rate": 2.561913588018573e-06,
+      "loss": 0.8203,
       "step": 27275
     },
     {
-      "epoch": 0.7491142786520557,
+      "epoch": 0.7740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 3.123673493630429e-06,
-      "loss": 0.8656,
+      "learning_rate": 2.5612993144287725e-06,
+      "loss": 0.7902,
       "step": 27276
     },
     {
-      "epoch": 0.749141742879899,
+      "epoch": 0.7740351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 3.123027679933126e-06,
-      "loss": 0.8274,
+      "learning_rate": 2.560685103674013e-06,
+      "loss": 0.8714,
       "step": 27277
     },
     {
-      "epoch": 0.7491692071077422,
+      "epoch": 0.7740635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 3.122381920649091e-06,
-      "loss": 0.7635,
+      "learning_rate": 2.5600709557594794e-06,
+      "loss": 0.7766,
       "step": 27278
     },
     {
-      "epoch": 0.7491966713355854,
+      "epoch": 0.7740919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 3.121736215783431e-06,
-      "loss": 0.8297,
+      "learning_rate": 2.55945687069036e-06,
+      "loss": 0.8669,
       "step": 27279
     },
     {
-      "epoch": 0.7492241355634286,
+      "epoch": 0.7741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 3.1210905653412595e-06,
-      "loss": 0.7913,
+      "learning_rate": 2.5588428484718464e-06,
+      "loss": 0.7491,
       "step": 27280
     },
     {
-      "epoch": 0.7492515997912719,
+      "epoch": 0.7741486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 3.120444969327683e-06,
-      "loss": 0.8269,
+      "learning_rate": 2.558228889109119e-06,
+      "loss": 0.797,
       "step": 27281
     },
     {
-      "epoch": 0.7492790640191151,
+      "epoch": 0.7741770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 3.119799427747812e-06,
-      "loss": 0.8026,
+      "learning_rate": 2.557614992607366e-06,
+      "loss": 0.8979,
       "step": 27282
     },
     {
-      "epoch": 0.7493065282469583,
+      "epoch": 0.7742054483541431,
       "grad_norm": 0.0,
-      "learning_rate": 3.1191539406067516e-06,
-      "loss": 0.798,
+      "learning_rate": 2.5570011589717768e-06,
+      "loss": 0.8578,
       "step": 27283
     },
     {
-      "epoch": 0.7493339924748016,
+      "epoch": 0.7742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 3.1185085079096146e-06,
-      "loss": 0.7975,
+      "learning_rate": 2.556387388207531e-06,
+      "loss": 0.9291,
       "step": 27284
     },
     {
-      "epoch": 0.7493614567026448,
+      "epoch": 0.7742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 3.1178631296615003e-06,
-      "loss": 0.9003,
+      "learning_rate": 2.555773680319815e-06,
+      "loss": 0.9015,
       "step": 27285
     },
     {
-      "epoch": 0.749388920930488,
+      "epoch": 0.7742905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 3.117217805867522e-06,
-      "loss": 0.7384,
+      "learning_rate": 2.555160035313814e-06,
+      "loss": 0.8319,
       "step": 27286
     },
     {
-      "epoch": 0.7494163851583313,
+      "epoch": 0.7743189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 3.11657253653278e-06,
-      "loss": 0.8189,
+      "learning_rate": 2.5545464531947097e-06,
+      "loss": 0.8338,
       "step": 27287
     },
     {
-      "epoch": 0.7494438493861745,
+      "epoch": 0.7743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 3.115927321662382e-06,
-      "loss": 0.8338,
+      "learning_rate": 2.5539329339676866e-06,
+      "loss": 0.8144,
       "step": 27288
     },
     {
-      "epoch": 0.7494713136140178,
+      "epoch": 0.7743757094211123,
       "grad_norm": 0.0,
-      "learning_rate": 3.115282161261434e-06,
-      "loss": 0.8635,
+      "learning_rate": 2.55331947763793e-06,
+      "loss": 0.8762,
       "step": 27289
     },
     {
-      "epoch": 0.749498777841861,
+      "epoch": 0.7744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 3.1146370553350435e-06,
-      "loss": 0.7669,
+      "learning_rate": 2.5527060842106154e-06,
+      "loss": 0.9209,
       "step": 27290
     },
     {
-      "epoch": 0.7495262420697042,
+      "epoch": 0.7744324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 3.1139920038883077e-06,
-      "loss": 0.8949,
+      "learning_rate": 2.5520927536909277e-06,
+      "loss": 0.7883,
       "step": 27291
     },
     {
-      "epoch": 0.7495537062975475,
+      "epoch": 0.774460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 3.113347006926336e-06,
-      "loss": 0.8406,
+      "learning_rate": 2.5514794860840498e-06,
+      "loss": 0.8117,
       "step": 27292
     },
     {
-      "epoch": 0.7495811705253906,
+      "epoch": 0.774489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 3.1127020644542326e-06,
-      "loss": 0.8337,
+      "learning_rate": 2.550866281395157e-06,
+      "loss": 0.97,
       "step": 27293
     },
     {
-      "epoch": 0.7496086347532339,
+      "epoch": 0.7745175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 3.112057176477096e-06,
-      "loss": 0.8384,
+      "learning_rate": 2.5502531396294318e-06,
+      "loss": 0.8203,
       "step": 27294
     },
     {
-      "epoch": 0.7496360989810772,
+      "epoch": 0.7745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 3.1114123430000296e-06,
-      "loss": 0.7699,
+      "learning_rate": 2.5496400607920556e-06,
+      "loss": 0.9292,
       "step": 27295
     },
     {
-      "epoch": 0.7496635632089204,
+      "epoch": 0.7745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 3.1107675640281423e-06,
-      "loss": 0.8488,
+      "learning_rate": 2.5490270448882016e-06,
+      "loss": 0.8009,
       "step": 27296
     },
     {
-      "epoch": 0.7496910274367636,
+      "epoch": 0.7746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 3.110122839566526e-06,
-      "loss": 0.8141,
+      "learning_rate": 2.5484140919230516e-06,
+      "loss": 0.8,
       "step": 27297
     },
     {
-      "epoch": 0.7497184916646068,
+      "epoch": 0.7746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 3.109478169620287e-06,
-      "loss": 0.7949,
+      "learning_rate": 2.547801201901785e-06,
+      "loss": 0.8591,
       "step": 27298
     },
     {
-      "epoch": 0.7497459558924501,
+      "epoch": 0.7746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 3.108833554194526e-06,
-      "loss": 0.828,
+      "learning_rate": 2.5471883748295747e-06,
+      "loss": 0.7981,
       "step": 27299
     },
     {
-      "epoch": 0.7497734201202934,
+      "epoch": 0.7746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 3.108188993294343e-06,
-      "loss": 0.8233,
+      "learning_rate": 2.546575610711598e-06,
+      "loss": 0.7997,
       "step": 27300
     },
     {
-      "epoch": 0.7498008843481365,
+      "epoch": 0.7747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 3.1075444869248428e-06,
-      "loss": 0.8184,
+      "learning_rate": 2.5459629095530336e-06,
+      "loss": 0.756,
       "step": 27301
     },
     {
-      "epoch": 0.7498283485759798,
+      "epoch": 0.774744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 3.1069000350911192e-06,
-      "loss": 0.7602,
+      "learning_rate": 2.545350271359055e-06,
+      "loss": 0.7789,
       "step": 27302
     },
     {
-      "epoch": 0.7498558128038231,
+      "epoch": 0.7747729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 3.10625563779827e-06,
-      "loss": 0.7935,
+      "learning_rate": 2.5447376961348404e-06,
+      "loss": 0.7964,
       "step": 27303
     },
     {
-      "epoch": 0.7498832770316662,
+      "epoch": 0.7748013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 3.1056112950513962e-06,
-      "loss": 0.8573,
+      "learning_rate": 2.5441251838855584e-06,
+      "loss": 0.8704,
       "step": 27304
     },
     {
-      "epoch": 0.7499107412595095,
+      "epoch": 0.7748297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 3.104967006855597e-06,
-      "loss": 0.8422,
+      "learning_rate": 2.543512734616387e-06,
+      "loss": 0.861,
       "step": 27305
     },
     {
-      "epoch": 0.7499382054873527,
+      "epoch": 0.7748581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 3.104322773215969e-06,
-      "loss": 0.7737,
+      "learning_rate": 2.542900348332501e-06,
+      "loss": 0.7905,
       "step": 27306
     },
     {
-      "epoch": 0.749965669715196,
+      "epoch": 0.7748864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 3.103678594137611e-06,
-      "loss": 0.8156,
+      "learning_rate": 2.542288025039068e-06,
+      "loss": 0.8271,
       "step": 27307
     },
     {
-      "epoch": 0.7499931339430392,
+      "epoch": 0.7749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 3.103034469625622e-06,
-      "loss": 0.8138,
+      "learning_rate": 2.541675764741264e-06,
+      "loss": 0.8649,
       "step": 27308
     },
     {
-      "epoch": 0.7500205981708824,
+      "epoch": 0.7749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 3.1023903996850935e-06,
-      "loss": 0.8128,
+      "learning_rate": 2.5410635674442618e-06,
+      "loss": 0.7948,
       "step": 27309
     },
     {
-      "epoch": 0.7500480623987257,
+      "epoch": 0.7749716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 3.101746384321126e-06,
-      "loss": 0.8155,
+      "learning_rate": 2.540451433153229e-06,
+      "loss": 0.8147,
       "step": 27310
     },
     {
-      "epoch": 0.7500755266265688,
+      "epoch": 0.775,
       "grad_norm": 0.0,
-      "learning_rate": 3.101102423538811e-06,
-      "loss": 0.9385,
+      "learning_rate": 2.5398393618733386e-06,
+      "loss": 0.7438,
       "step": 27311
     },
     {
-      "epoch": 0.7501029908544121,
+      "epoch": 0.7750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 3.1004585173432457e-06,
-      "loss": 0.7705,
+      "learning_rate": 2.5392273536097623e-06,
+      "loss": 0.8601,
       "step": 27312
     },
     {
-      "epoch": 0.7501304550822554,
+      "epoch": 0.7750567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 3.099814665739529e-06,
-      "loss": 0.8128,
+      "learning_rate": 2.5386154083676662e-06,
+      "loss": 0.7029,
       "step": 27313
     },
     {
-      "epoch": 0.7501579193100986,
+      "epoch": 0.7750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 3.099170868732747e-06,
-      "loss": 0.8007,
+      "learning_rate": 2.5380035261522207e-06,
+      "loss": 0.8181,
       "step": 27314
     },
     {
-      "epoch": 0.7501853835379418,
+      "epoch": 0.7751135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 3.098527126327999e-06,
-      "loss": 0.8501,
+      "learning_rate": 2.5373917069685972e-06,
+      "loss": 0.9202,
       "step": 27315
     },
     {
-      "epoch": 0.7502128477657851,
+      "epoch": 0.7751418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 3.097883438530378e-06,
-      "loss": 0.8423,
+      "learning_rate": 2.536779950821956e-06,
+      "loss": 0.8121,
       "step": 27316
     },
     {
-      "epoch": 0.7502403119936283,
+      "epoch": 0.7751702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 3.0972398053449805e-06,
-      "loss": 0.8321,
+      "learning_rate": 2.536168257717476e-06,
+      "loss": 0.8805,
       "step": 27317
     },
     {
-      "epoch": 0.7502677762214716,
+      "epoch": 0.7751986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 3.096596226776891e-06,
-      "loss": 0.8056,
+      "learning_rate": 2.535556627660315e-06,
+      "loss": 0.7292,
       "step": 27318
     },
     {
-      "epoch": 0.7502952404493147,
+      "epoch": 0.7752270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 3.095952702831211e-06,
-      "loss": 0.7735,
+      "learning_rate": 2.5349450606556425e-06,
+      "loss": 0.8135,
       "step": 27319
     },
     {
-      "epoch": 0.750322704677158,
+      "epoch": 0.775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 3.095309233513023e-06,
-      "loss": 0.7978,
+      "learning_rate": 2.534333556708628e-06,
+      "loss": 0.8583,
       "step": 27320
     },
     {
-      "epoch": 0.7503501689050013,
+      "epoch": 0.775283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 3.094665818827424e-06,
-      "loss": 0.7973,
+      "learning_rate": 2.5337221158244306e-06,
+      "loss": 0.724,
       "step": 27321
     },
     {
-      "epoch": 0.7503776331328444,
+      "epoch": 0.7753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 3.094022458779503e-06,
-      "loss": 0.8972,
+      "learning_rate": 2.533110738008219e-06,
+      "loss": 0.7728,
       "step": 27322
     },
     {
-      "epoch": 0.7504050973606877,
+      "epoch": 0.7753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 3.0933791533743517e-06,
-      "loss": 0.8383,
+      "learning_rate": 2.532499423265159e-06,
+      "loss": 0.8618,
       "step": 27323
     },
     {
-      "epoch": 0.7504325615885309,
+      "epoch": 0.7753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.09273590261706e-06,
-      "loss": 0.7673,
+      "learning_rate": 2.5318881716004085e-06,
+      "loss": 0.7449,
       "step": 27324
     },
     {
-      "epoch": 0.7504600258163742,
+      "epoch": 0.7753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 3.0920927065127205e-06,
-      "loss": 0.8195,
+      "learning_rate": 2.5312769830191343e-06,
+      "loss": 0.9393,
       "step": 27325
     },
     {
-      "epoch": 0.7504874900442174,
+      "epoch": 0.7754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 3.09144956506642e-06,
-      "loss": 0.8272,
+      "learning_rate": 2.530665857526503e-06,
+      "loss": 0.8293,
       "step": 27326
     },
     {
-      "epoch": 0.7505149542720606,
+      "epoch": 0.7754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 3.0908064782832424e-06,
-      "loss": 0.7523,
+      "learning_rate": 2.5300547951276688e-06,
+      "loss": 0.7575,
       "step": 27327
     },
     {
-      "epoch": 0.7505424184999039,
+      "epoch": 0.7754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 3.0901634461682816e-06,
-      "loss": 0.7419,
+      "learning_rate": 2.5294437958277985e-06,
+      "loss": 0.9161,
       "step": 27328
     },
     {
-      "epoch": 0.7505698827277472,
+      "epoch": 0.775510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 3.0895204687266233e-06,
-      "loss": 0.8833,
+      "learning_rate": 2.5288328596320557e-06,
+      "loss": 0.8569,
       "step": 27329
     },
     {
-      "epoch": 0.7505973469555903,
+      "epoch": 0.7755391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 3.0888775459633568e-06,
-      "loss": 0.8078,
+      "learning_rate": 2.5282219865455935e-06,
+      "loss": 0.8695,
       "step": 27330
     },
     {
-      "epoch": 0.7506248111834336,
+      "epoch": 0.7755675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 3.0882346778835704e-06,
-      "loss": 0.7221,
+      "learning_rate": 2.527611176573577e-06,
+      "loss": 0.9368,
       "step": 27331
     },
     {
-      "epoch": 0.7506522754112768,
+      "epoch": 0.7755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 3.0875918644923464e-06,
-      "loss": 0.8301,
+      "learning_rate": 2.5270004297211637e-06,
+      "loss": 0.7476,
       "step": 27332
     },
     {
-      "epoch": 0.75067973963912,
+      "epoch": 0.7756242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 3.086949105794772e-06,
-      "loss": 0.837,
+      "learning_rate": 2.5263897459935138e-06,
+      "loss": 0.9504,
       "step": 27333
     },
     {
-      "epoch": 0.7507072038669633,
+      "epoch": 0.7756526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 3.0863064017959376e-06,
-      "loss": 0.7358,
+      "learning_rate": 2.5257791253957897e-06,
+      "loss": 0.7202,
       "step": 27334
     },
     {
-      "epoch": 0.7507346680948065,
+      "epoch": 0.7756810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 3.0856637525009223e-06,
-      "loss": 0.7968,
+      "learning_rate": 2.525168567933142e-06,
+      "loss": 0.7977,
       "step": 27335
     },
     {
-      "epoch": 0.7507621323226498,
+      "epoch": 0.7757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 3.085021157914814e-06,
-      "loss": 0.8034,
+      "learning_rate": 2.5245580736107313e-06,
+      "loss": 0.8791,
       "step": 27336
     },
     {
-      "epoch": 0.7507895965504929,
+      "epoch": 0.7757377979568671,
       "grad_norm": 0.0,
-      "learning_rate": 3.0843786180427004e-06,
-      "loss": 0.8497,
+      "learning_rate": 2.523947642433717e-06,
+      "loss": 0.8145,
       "step": 27337
     },
     {
-      "epoch": 0.7508170607783362,
+      "epoch": 0.7757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 3.0837361328896578e-06,
-      "loss": 0.824,
+      "learning_rate": 2.5233372744072504e-06,
+      "loss": 0.8265,
       "step": 27338
     },
     {
-      "epoch": 0.7508445250061795,
+      "epoch": 0.775794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 3.0830937024607745e-06,
-      "loss": 0.7439,
+      "learning_rate": 2.522726969536491e-06,
+      "loss": 0.7919,
       "step": 27339
     },
     {
-      "epoch": 0.7508719892340227,
+      "epoch": 0.7758229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 3.082451326761133e-06,
-      "loss": 0.8732,
+      "learning_rate": 2.5221167278265946e-06,
+      "loss": 0.9254,
       "step": 27340
     },
     {
-      "epoch": 0.7508994534618659,
+      "epoch": 0.7758513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 3.0818090057958195e-06,
-      "loss": 0.8546,
+      "learning_rate": 2.521506549282712e-06,
+      "loss": 0.7724,
       "step": 27341
     },
     {
-      "epoch": 0.7509269176897092,
+      "epoch": 0.7758796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 3.0811667395699096e-06,
-      "loss": 0.7974,
+      "learning_rate": 2.5208964339099996e-06,
+      "loss": 0.884,
       "step": 27342
     },
     {
-      "epoch": 0.7509543819175524,
+      "epoch": 0.7759080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 3.080524528088492e-06,
-      "loss": 0.8116,
+      "learning_rate": 2.5202863817136137e-06,
+      "loss": 0.7358,
       "step": 27343
     },
     {
-      "epoch": 0.7509818461453956,
+      "epoch": 0.7759364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 3.0798823713566416e-06,
-      "loss": 0.8392,
+      "learning_rate": 2.519676392698703e-06,
+      "loss": 0.8664,
       "step": 27344
     },
     {
-      "epoch": 0.7510093103732388,
+      "epoch": 0.7759648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 3.079240269379442e-06,
-      "loss": 0.8575,
+      "learning_rate": 2.519066466870421e-06,
+      "loss": 0.7854,
       "step": 27345
     },
     {
-      "epoch": 0.7510367746010821,
+      "epoch": 0.7759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 3.078598222161975e-06,
-      "loss": 0.7604,
+      "learning_rate": 2.5184566042339253e-06,
+      "loss": 0.8029,
       "step": 27346
     },
     {
-      "epoch": 0.7510642388289254,
+      "epoch": 0.7760215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 3.077956229709319e-06,
-      "loss": 0.7502,
+      "learning_rate": 2.517846804794356e-06,
+      "loss": 0.9194,
       "step": 27347
     },
     {
-      "epoch": 0.7510917030567685,
+      "epoch": 0.776049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 3.0773142920265552e-06,
-      "loss": 0.777,
+      "learning_rate": 2.517237068556877e-06,
+      "loss": 0.8403,
       "step": 27348
     },
     {
-      "epoch": 0.7511191672846118,
+      "epoch": 0.7760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 3.076672409118766e-06,
-      "loss": 0.9134,
+      "learning_rate": 2.5166273955266297e-06,
+      "loss": 0.8285,
       "step": 27349
     },
     {
-      "epoch": 0.751146631512455,
+      "epoch": 0.7761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 3.0760305809910263e-06,
-      "loss": 0.9205,
+      "learning_rate": 2.5160177857087677e-06,
+      "loss": 0.7669,
       "step": 27350
     },
     {
-      "epoch": 0.7511740957402983,
+      "epoch": 0.7761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 3.0753888076484105e-06,
-      "loss": 0.8503,
+      "learning_rate": 2.5154082391084435e-06,
+      "loss": 0.8416,
       "step": 27351
     },
     {
-      "epoch": 0.7512015599681415,
+      "epoch": 0.7761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 3.074747089096002e-06,
-      "loss": 0.8145,
+      "learning_rate": 2.5147987557307983e-06,
+      "loss": 0.8742,
       "step": 27352
     },
     {
-      "epoch": 0.7512290241959847,
+      "epoch": 0.7761918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 3.0741054253388767e-06,
-      "loss": 0.8593,
+      "learning_rate": 2.5141893355809845e-06,
+      "loss": 0.801,
       "step": 27353
     },
     {
-      "epoch": 0.751256488423828,
+      "epoch": 0.7762202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 3.0734638163821117e-06,
-      "loss": 0.7711,
+      "learning_rate": 2.5135799786641534e-06,
+      "loss": 0.7745,
       "step": 27354
     },
     {
-      "epoch": 0.7512839526516712,
+      "epoch": 0.7762485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 3.072822262230788e-06,
-      "loss": 0.7364,
+      "learning_rate": 2.512970684985445e-06,
+      "loss": 0.6505,
       "step": 27355
     },
     {
-      "epoch": 0.7513114168795144,
+      "epoch": 0.7762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 3.072180762889975e-06,
-      "loss": 0.7406,
+      "learning_rate": 2.5123614545500107e-06,
+      "loss": 0.7181,
       "step": 27356
     },
     {
-      "epoch": 0.7513388811073577,
+      "epoch": 0.776305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 3.07153931836475e-06,
-      "loss": 0.8149,
+      "learning_rate": 2.5117522873629986e-06,
+      "loss": 0.8577,
       "step": 27357
     },
     {
-      "epoch": 0.7513663453352009,
+      "epoch": 0.77633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.070897928660194e-06,
-      "loss": 0.7666,
+      "learning_rate": 2.5111431834295486e-06,
+      "loss": 0.8782,
       "step": 27358
     },
     {
-      "epoch": 0.7513938095630441,
+      "epoch": 0.7763620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 3.0702565937813744e-06,
-      "loss": 0.7904,
+      "learning_rate": 2.510534142754808e-06,
+      "loss": 0.7754,
       "step": 27359
     },
     {
-      "epoch": 0.7514212737908874,
+      "epoch": 0.7763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 3.0696153137333685e-06,
-      "loss": 0.8088,
+      "learning_rate": 2.509925165343926e-06,
+      "loss": 0.7737,
       "step": 27360
     },
     {
-      "epoch": 0.7514487380187306,
+      "epoch": 0.7764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 3.0689740885212537e-06,
-      "loss": 0.8759,
+      "learning_rate": 2.50931625120204e-06,
+      "loss": 0.7109,
       "step": 27361
     },
     {
-      "epoch": 0.7514762022465739,
+      "epoch": 0.7764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.0683329181500986e-06,
-      "loss": 0.7625,
+      "learning_rate": 2.508707400334296e-06,
+      "loss": 0.8688,
       "step": 27362
     },
     {
-      "epoch": 0.751503666474417,
+      "epoch": 0.7764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 3.067691802624978e-06,
-      "loss": 0.89,
+      "learning_rate": 2.5080986127458373e-06,
+      "loss": 0.7428,
       "step": 27363
     },
     {
-      "epoch": 0.7515311307022603,
+      "epoch": 0.7765039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 3.0670507419509645e-06,
-      "loss": 0.7759,
+      "learning_rate": 2.5074898884418063e-06,
+      "loss": 0.8352,
       "step": 27364
     },
     {
-      "epoch": 0.7515585949301036,
+      "epoch": 0.7765323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 3.066409736133136e-06,
-      "loss": 0.7978,
+      "learning_rate": 2.506881227427348e-06,
+      "loss": 0.8994,
       "step": 27365
     },
     {
-      "epoch": 0.7515860591579467,
+      "epoch": 0.7765607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 3.065768785176555e-06,
-      "loss": 0.8427,
+      "learning_rate": 2.506272629707598e-06,
+      "loss": 0.9462,
       "step": 27366
     },
     {
-      "epoch": 0.75161352338579,
+      "epoch": 0.7765891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 3.0651278890863003e-06,
-      "loss": 0.8152,
+      "learning_rate": 2.5056640952876997e-06,
+      "loss": 0.8214,
       "step": 27367
     },
     {
-      "epoch": 0.7516409876136333,
+      "epoch": 0.7766174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 3.0644870478674373e-06,
-      "loss": 0.8213,
+      "learning_rate": 2.5050556241727963e-06,
+      "loss": 1.015,
       "step": 27368
     },
     {
-      "epoch": 0.7516684518414765,
+      "epoch": 0.7766458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 3.06384626152504e-06,
-      "loss": 0.9111,
+      "learning_rate": 2.5044472163680234e-06,
+      "loss": 0.8439,
       "step": 27369
     },
     {
-      "epoch": 0.7516959160693197,
+      "epoch": 0.7766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 3.0632055300641773e-06,
-      "loss": 0.8566,
+      "learning_rate": 2.50383887187852e-06,
+      "loss": 0.8589,
       "step": 27370
     },
     {
-      "epoch": 0.7517233802971629,
+      "epoch": 0.7767026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 3.0625648534899188e-06,
-      "loss": 0.8417,
+      "learning_rate": 2.503230590709431e-06,
+      "loss": 0.8269,
       "step": 27371
     },
     {
-      "epoch": 0.7517508445250062,
+      "epoch": 0.7767309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 3.061924231807338e-06,
-      "loss": 0.8387,
+      "learning_rate": 2.5026223728658873e-06,
+      "loss": 0.8698,
       "step": 27372
     },
     {
-      "epoch": 0.7517783087528495,
+      "epoch": 0.7767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 3.0612836650214983e-06,
-      "loss": 0.8549,
+      "learning_rate": 2.502014218353028e-06,
+      "loss": 0.8612,
       "step": 27373
     },
     {
-      "epoch": 0.7518057729806926,
+      "epoch": 0.7767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 3.0606431531374723e-06,
-      "loss": 0.7965,
+      "learning_rate": 2.501406127175996e-06,
+      "loss": 0.828,
       "step": 27374
     },
     {
-      "epoch": 0.7518332372085359,
+      "epoch": 0.7768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.0600026961603225e-06,
-      "loss": 0.869,
+      "learning_rate": 2.5007980993399197e-06,
+      "loss": 0.8826,
       "step": 27375
     },
     {
-      "epoch": 0.7518607014363791,
+      "epoch": 0.776844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 3.0593622940951194e-06,
-      "loss": 0.853,
+      "learning_rate": 2.500190134849939e-06,
+      "loss": 0.7339,
       "step": 27376
     },
     {
-      "epoch": 0.7518881656642223,
+      "epoch": 0.7768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 3.058721946946931e-06,
-      "loss": 0.8406,
+      "learning_rate": 2.499582233711193e-06,
+      "loss": 0.7319,
       "step": 27377
     },
     {
-      "epoch": 0.7519156298920656,
+      "epoch": 0.7769012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 3.058081654720827e-06,
-      "loss": 0.8224,
+      "learning_rate": 2.498974395928807e-06,
+      "loss": 0.8377,
       "step": 27378
     },
     {
-      "epoch": 0.7519430941199088,
+      "epoch": 0.7769296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 3.057441417421866e-06,
-      "loss": 0.8989,
+      "learning_rate": 2.498366621507925e-06,
+      "loss": 0.8913,
       "step": 27379
     },
     {
-      "epoch": 0.7519705583477521,
+      "epoch": 0.7769580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 3.0568012350551177e-06,
-      "loss": 0.8613,
+      "learning_rate": 2.497758910453679e-06,
+      "loss": 0.8461,
       "step": 27380
     },
     {
-      "epoch": 0.7519980225755953,
+      "epoch": 0.7769863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 3.0561611076256483e-06,
-      "loss": 0.8265,
+      "learning_rate": 2.497151262771198e-06,
+      "loss": 0.8152,
       "step": 27381
     },
     {
-      "epoch": 0.7520254868034385,
+      "epoch": 0.7770147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 3.0555210351385246e-06,
-      "loss": 0.876,
+      "learning_rate": 2.4965436784656182e-06,
+      "loss": 0.8712,
       "step": 27382
     },
     {
-      "epoch": 0.7520529510312818,
+      "epoch": 0.7770431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 3.054881017598805e-06,
-      "loss": 0.8878,
+      "learning_rate": 2.495936157542074e-06,
+      "loss": 0.8122,
       "step": 27383
     },
     {
-      "epoch": 0.752080415259125,
+      "epoch": 0.7770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 3.054241055011561e-06,
-      "loss": 0.8654,
+      "learning_rate": 2.495328700005691e-06,
+      "loss": 0.8625,
       "step": 27384
     },
     {
-      "epoch": 0.7521078794869682,
+      "epoch": 0.7770998864926221,
       "grad_norm": 0.0,
-      "learning_rate": 3.0536011473818483e-06,
-      "loss": 0.8615,
+      "learning_rate": 2.494721305861607e-06,
+      "loss": 0.8951,
       "step": 27385
     },
     {
-      "epoch": 0.7521353437148115,
+      "epoch": 0.7771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 3.0529612947147347e-06,
-      "loss": 0.7418,
+      "learning_rate": 2.494113975114947e-06,
+      "loss": 0.7995,
       "step": 27386
     },
     {
-      "epoch": 0.7521628079426547,
+      "epoch": 0.7771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 3.0523214970152825e-06,
-      "loss": 0.7768,
+      "learning_rate": 2.4935067077708443e-06,
+      "loss": 0.7566,
       "step": 27387
     },
     {
-      "epoch": 0.7521902721704979,
+      "epoch": 0.7771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 3.0516817542885537e-06,
-      "loss": 0.8532,
+      "learning_rate": 2.4928995038344294e-06,
+      "loss": 0.7663,
       "step": 27388
     },
     {
-      "epoch": 0.7522177363983411,
+      "epoch": 0.7772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 3.0510420665396134e-06,
-      "loss": 0.7148,
+      "learning_rate": 2.4922923633108274e-06,
+      "loss": 0.8538,
       "step": 27389
     },
     {
-      "epoch": 0.7522452006261844,
+      "epoch": 0.7772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 3.050402433773516e-06,
-      "loss": 0.7453,
+      "learning_rate": 2.4916852862051687e-06,
+      "loss": 0.8003,
       "step": 27390
     },
     {
-      "epoch": 0.7522726648540277,
+      "epoch": 0.7772701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 3.0497628559953307e-06,
-      "loss": 0.8984,
+      "learning_rate": 2.4910782725225856e-06,
+      "loss": 0.8904,
       "step": 27391
     },
     {
-      "epoch": 0.7523001290818708,
+      "epoch": 0.7772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 3.049123333210111e-06,
-      "loss": 0.7141,
+      "learning_rate": 2.4904713222682e-06,
+      "loss": 0.6858,
       "step": 27392
     },
     {
-      "epoch": 0.7523275933097141,
+      "epoch": 0.7773269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 3.0484838654229197e-06,
-      "loss": 0.8362,
+      "learning_rate": 2.489864435447136e-06,
+      "loss": 0.8552,
       "step": 27393
     },
     {
-      "epoch": 0.7523550575375574,
+      "epoch": 0.777355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 3.047844452638817e-06,
-      "loss": 0.848,
+      "learning_rate": 2.489257612064531e-06,
+      "loss": 0.7856,
       "step": 27394
     },
     {
-      "epoch": 0.7523825217654005,
+      "epoch": 0.7773836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 3.0472050948628617e-06,
-      "loss": 0.9348,
+      "learning_rate": 2.4886508521255015e-06,
+      "loss": 0.7894,
       "step": 27395
     },
     {
-      "epoch": 0.7524099859932438,
+      "epoch": 0.7774120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 3.046565792100117e-06,
-      "loss": 0.8542,
+      "learning_rate": 2.4880441556351764e-06,
+      "loss": 0.862,
       "step": 27396
     },
     {
-      "epoch": 0.752437450221087,
+      "epoch": 0.7774404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 3.0459265443556328e-06,
-      "loss": 0.7997,
+      "learning_rate": 2.4874375225986834e-06,
+      "loss": 0.8119,
       "step": 27397
     },
     {
-      "epoch": 0.7524649144489303,
+      "epoch": 0.7774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 3.045287351634475e-06,
-      "loss": 0.8258,
+      "learning_rate": 2.48683095302114e-06,
+      "loss": 0.847,
       "step": 27398
     },
     {
-      "epoch": 0.7524923786767735,
+      "epoch": 0.7774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 3.0446482139416956e-06,
-      "loss": 0.8295,
+      "learning_rate": 2.4862244469076724e-06,
+      "loss": 0.7478,
       "step": 27399
     },
     {
-      "epoch": 0.7525198429046167,
+      "epoch": 0.7775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 3.0440091312823527e-06,
-      "loss": 0.7763,
+      "learning_rate": 2.4856180042634082e-06,
+      "loss": 0.842,
       "step": 27400
     },
     {
-      "epoch": 0.75254730713246,
+      "epoch": 0.7775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 3.043370103661504e-06,
-      "loss": 0.7863,
+      "learning_rate": 2.4850116250934642e-06,
+      "loss": 0.7153,
       "step": 27401
     },
     {
-      "epoch": 0.7525747713603032,
+      "epoch": 0.7775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 3.042731131084209e-06,
-      "loss": 0.7321,
+      "learning_rate": 2.484405309402964e-06,
+      "loss": 0.8568,
       "step": 27402
     },
     {
-      "epoch": 0.7526022355881464,
+      "epoch": 0.7776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 3.0420922135555173e-06,
-      "loss": 0.7723,
+      "learning_rate": 2.483799057197034e-06,
+      "loss": 0.8461,
       "step": 27403
     },
     {
-      "epoch": 0.7526296998159897,
+      "epoch": 0.777639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 3.041453351080488e-06,
-      "loss": 0.8373,
+      "learning_rate": 2.483192868480787e-06,
+      "loss": 0.8514,
       "step": 27404
     },
     {
-      "epoch": 0.7526571640438329,
+      "epoch": 0.7776674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 3.0408145436641743e-06,
-      "loss": 0.9046,
+      "learning_rate": 2.482586743259349e-06,
+      "loss": 0.9535,
       "step": 27405
     },
     {
-      "epoch": 0.7526846282716761,
+      "epoch": 0.7776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 3.040175791311635e-06,
-      "loss": 0.8027,
+      "learning_rate": 2.48198068153784e-06,
+      "loss": 0.8294,
       "step": 27406
     },
     {
-      "epoch": 0.7527120924995194,
+      "epoch": 0.7777241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 3.0395370940279177e-06,
-      "loss": 0.9094,
+      "learning_rate": 2.481374683321376e-06,
+      "loss": 0.8038,
       "step": 27407
     },
     {
-      "epoch": 0.7527395567273626,
+      "epoch": 0.7777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 3.038898451818082e-06,
-      "loss": 0.8316,
+      "learning_rate": 2.480768748615079e-06,
+      "loss": 0.7553,
       "step": 27408
     },
     {
-      "epoch": 0.7527670209552059,
+      "epoch": 0.7777809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 3.0382598646871753e-06,
-      "loss": 0.8621,
+      "learning_rate": 2.480162877424065e-06,
+      "loss": 0.8644,
       "step": 27409
     },
     {
-      "epoch": 0.752794485183049,
+      "epoch": 0.7778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 3.037621332640254e-06,
-      "loss": 0.9087,
+      "learning_rate": 2.4795570697534544e-06,
+      "loss": 0.7628,
       "step": 27410
     },
     {
-      "epoch": 0.7528219494108923,
+      "epoch": 0.7778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 3.036982855682369e-06,
-      "loss": 0.729,
+      "learning_rate": 2.478951325608365e-06,
+      "loss": 0.7887,
       "step": 27411
     },
     {
-      "epoch": 0.7528494136387356,
+      "epoch": 0.7778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 3.036344433818572e-06,
-      "loss": 0.836,
+      "learning_rate": 2.47834564499391e-06,
+      "loss": 0.8279,
       "step": 27412
     },
     {
-      "epoch": 0.7528768778665788,
+      "epoch": 0.777894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.0357060670539197e-06,
-      "loss": 0.7923,
+      "learning_rate": 2.4777400279152064e-06,
+      "loss": 0.7864,
       "step": 27413
     },
     {
-      "epoch": 0.752904342094422,
+      "epoch": 0.7779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 3.0350677553934582e-06,
-      "loss": 0.8208,
+      "learning_rate": 2.477134474377375e-06,
+      "loss": 0.9189,
       "step": 27414
     },
     {
-      "epoch": 0.7529318063222652,
+      "epoch": 0.7779511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 3.0344294988422364e-06,
-      "loss": 0.8203,
+      "learning_rate": 2.4765289843855233e-06,
+      "loss": 0.8206,
       "step": 27415
     },
     {
-      "epoch": 0.7529592705501085,
+      "epoch": 0.7779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 3.033791297405305e-06,
-      "loss": 0.8076,
+      "learning_rate": 2.475923557944769e-06,
+      "loss": 0.8513,
       "step": 27416
     },
     {
-      "epoch": 0.7529867347779517,
+      "epoch": 0.7780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 3.0331531510877166e-06,
-      "loss": 0.8644,
+      "learning_rate": 2.475318195060231e-06,
+      "loss": 0.7208,
       "step": 27417
     },
     {
-      "epoch": 0.7530141990057949,
+      "epoch": 0.7780363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 3.0325150598945187e-06,
-      "loss": 0.8201,
+      "learning_rate": 2.474712895737015e-06,
+      "loss": 0.8563,
       "step": 27418
     },
     {
-      "epoch": 0.7530416632336382,
+      "epoch": 0.7780646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 3.0318770238307605e-06,
-      "loss": 0.798,
+      "learning_rate": 2.474107659980236e-06,
+      "loss": 0.8654,
       "step": 27419
     },
     {
-      "epoch": 0.7530691274614815,
+      "epoch": 0.7780930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 3.0312390429014936e-06,
-      "loss": 0.8903,
+      "learning_rate": 2.473502487795012e-06,
+      "loss": 0.767,
       "step": 27420
     },
     {
-      "epoch": 0.7530965916893246,
+      "epoch": 0.7781214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 3.03060111711176e-06,
-      "loss": 0.8201,
+      "learning_rate": 2.4728973791864485e-06,
+      "loss": 0.8081,
       "step": 27421
     },
     {
-      "epoch": 0.7531240559171679,
+      "epoch": 0.778149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 3.0299632464666127e-06,
-      "loss": 0.8714,
+      "learning_rate": 2.4722923341596585e-06,
+      "loss": 0.8466,
       "step": 27422
     },
     {
-      "epoch": 0.7531515201450111,
+      "epoch": 0.778178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 3.029325430971093e-06,
-      "loss": 0.8131,
+      "learning_rate": 2.471687352719756e-06,
+      "loss": 0.8413,
       "step": 27423
     },
     {
-      "epoch": 0.7531789843728544,
+      "epoch": 0.7782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 3.0286876706302517e-06,
-      "loss": 0.8085,
+      "learning_rate": 2.471082434871844e-06,
+      "loss": 0.8693,
       "step": 27424
     },
     {
-      "epoch": 0.7532064486006976,
+      "epoch": 0.7782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 3.0280499654491324e-06,
-      "loss": 0.8001,
+      "learning_rate": 2.4704775806210423e-06,
+      "loss": 0.8075,
       "step": 27425
     },
     {
-      "epoch": 0.7532339128285408,
+      "epoch": 0.7782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.027412315432787e-06,
-      "loss": 0.8009,
+      "learning_rate": 2.469872789972453e-06,
+      "loss": 0.8655,
       "step": 27426
     },
     {
-      "epoch": 0.7532613770563841,
+      "epoch": 0.7782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 3.026774720586252e-06,
-      "loss": 0.8248,
+      "learning_rate": 2.469268062931186e-06,
+      "loss": 0.8369,
       "step": 27427
     },
     {
-      "epoch": 0.7532888412842272,
+      "epoch": 0.7783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 3.026137180914578e-06,
-      "loss": 0.9021,
+      "learning_rate": 2.4686633995023525e-06,
+      "loss": 0.8242,
       "step": 27428
     },
     {
-      "epoch": 0.7533163055120705,
+      "epoch": 0.7783484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 3.025499696422807e-06,
-      "loss": 0.9055,
+      "learning_rate": 2.4680587996910555e-06,
+      "loss": 0.9178,
       "step": 27429
     },
     {
-      "epoch": 0.7533437697399138,
+      "epoch": 0.7783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 3.024862267115987e-06,
-      "loss": 0.7144,
+      "learning_rate": 2.4674542635024048e-06,
+      "loss": 0.8296,
       "step": 27430
     },
     {
-      "epoch": 0.753371233967757,
+      "epoch": 0.778405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 3.024224892999159e-06,
-      "loss": 0.8897,
+      "learning_rate": 2.466849790941509e-06,
+      "loss": 0.8018,
       "step": 27431
     },
     {
-      "epoch": 0.7533986981956002,
+      "epoch": 0.7784335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 3.0235875740773622e-06,
-      "loss": 0.8703,
+      "learning_rate": 2.466245382013469e-06,
+      "loss": 0.7868,
       "step": 27432
     },
     {
-      "epoch": 0.7534261624234435,
+      "epoch": 0.7784619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 3.0229503103556434e-06,
-      "loss": 0.7262,
+      "learning_rate": 2.4656410367233928e-06,
+      "loss": 0.9031,
       "step": 27433
     },
     {
-      "epoch": 0.7534536266512867,
+      "epoch": 0.7784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 3.022313101839045e-06,
-      "loss": 0.8674,
+      "learning_rate": 2.4650367550763877e-06,
+      "loss": 0.7445,
       "step": 27434
     },
     {
-      "epoch": 0.75348109087913,
+      "epoch": 0.7785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 3.0216759485326073e-06,
-      "loss": 0.7479,
+      "learning_rate": 2.4644325370775533e-06,
+      "loss": 0.7609,
       "step": 27435
     },
     {
-      "epoch": 0.7535085551069731,
+      "epoch": 0.7785471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 3.0210388504413733e-06,
-      "loss": 0.7692,
+      "learning_rate": 2.4638283827319964e-06,
+      "loss": 0.7868,
       "step": 27436
     },
     {
-      "epoch": 0.7535360193348164,
+      "epoch": 0.7785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 3.0204018075703854e-06,
-      "loss": 0.8003,
+      "learning_rate": 2.4632242920448233e-06,
+      "loss": 0.7484,
       "step": 27437
     },
     {
-      "epoch": 0.7535634835626597,
+      "epoch": 0.7786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 3.0197648199246833e-06,
-      "loss": 0.8087,
+      "learning_rate": 2.4626202650211293e-06,
+      "loss": 0.8391,
       "step": 27438
     },
     {
-      "epoch": 0.7535909477905028,
+      "epoch": 0.7786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 3.0191278875093024e-06,
-      "loss": 0.8421,
+      "learning_rate": 2.4620163016660216e-06,
+      "loss": 0.7798,
       "step": 27439
     },
     {
-      "epoch": 0.7536184120183461,
+      "epoch": 0.7786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 3.018491010329285e-06,
-      "loss": 0.8495,
+      "learning_rate": 2.4614124019846007e-06,
+      "loss": 0.9446,
       "step": 27440
     },
     {
-      "epoch": 0.7536458762461893,
+      "epoch": 0.778688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.0178541883896718e-06,
-      "loss": 0.7695,
+      "learning_rate": 2.460808565981969e-06,
+      "loss": 0.7184,
       "step": 27441
     },
     {
-      "epoch": 0.7536733404740326,
+      "epoch": 0.7787173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 3.017217421695501e-06,
-      "loss": 0.8881,
+      "learning_rate": 2.4602047936632277e-06,
+      "loss": 0.8201,
       "step": 27442
     },
     {
-      "epoch": 0.7537008047018758,
+      "epoch": 0.7787457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 3.016580710251813e-06,
-      "loss": 0.7479,
+      "learning_rate": 2.459601085033474e-06,
+      "loss": 0.8572,
       "step": 27443
     },
     {
-      "epoch": 0.753728268929719,
+      "epoch": 0.7787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 3.015944054063642e-06,
-      "loss": 0.8998,
+      "learning_rate": 2.458997440097808e-06,
+      "loss": 0.7744,
       "step": 27444
     },
     {
-      "epoch": 0.7537557331575623,
+      "epoch": 0.7788024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 3.015307453136026e-06,
-      "loss": 0.7394,
+      "learning_rate": 2.458393858861332e-06,
+      "loss": 0.7655,
       "step": 27445
     },
     {
-      "epoch": 0.7537831973854056,
+      "epoch": 0.7788308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 3.0146709074740053e-06,
-      "loss": 0.8321,
+      "learning_rate": 2.4577903413291405e-06,
+      "loss": 0.8322,
       "step": 27446
     },
     {
-      "epoch": 0.7538106616132487,
+      "epoch": 0.7788592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 3.014034417082613e-06,
-      "loss": 0.8463,
+      "learning_rate": 2.457186887506332e-06,
+      "loss": 0.932,
       "step": 27447
     },
     {
-      "epoch": 0.753838125841092,
+      "epoch": 0.7788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 3.0133979819668856e-06,
-      "loss": 0.8561,
+      "learning_rate": 2.4565834973980086e-06,
+      "loss": 0.7778,
       "step": 27448
     },
     {
-      "epoch": 0.7538655900689352,
+      "epoch": 0.7789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 3.0127616021318627e-06,
-      "loss": 0.7629,
+      "learning_rate": 2.4559801710092602e-06,
+      "loss": 0.8215,
       "step": 27449
     },
     {
-      "epoch": 0.7538930542967784,
+      "epoch": 0.77894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.0121252775825736e-06,
-      "loss": 0.7926,
+      "learning_rate": 2.455376908345186e-06,
+      "loss": 0.7675,
       "step": 27450
     },
     {
-      "epoch": 0.7539205185246217,
+      "epoch": 0.7789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 3.0114890083240556e-06,
-      "loss": 0.7767,
+      "learning_rate": 2.454773709410886e-06,
+      "loss": 0.8405,
       "step": 27451
     },
     {
-      "epoch": 0.7539479827524649,
+      "epoch": 0.7790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 3.010852794361344e-06,
-      "loss": 0.805,
+      "learning_rate": 2.4541705742114486e-06,
+      "loss": 0.8646,
       "step": 27452
     },
     {
-      "epoch": 0.7539754469803082,
+      "epoch": 0.7790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 3.0102166356994767e-06,
-      "loss": 0.7938,
+      "learning_rate": 2.453567502751971e-06,
+      "loss": 1.0068,
       "step": 27453
     },
     {
-      "epoch": 0.7540029112081513,
+      "epoch": 0.7790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 3.009580532343478e-06,
-      "loss": 0.795,
+      "learning_rate": 2.4529644950375474e-06,
+      "loss": 0.8542,
       "step": 27454
     },
     {
-      "epoch": 0.7540303754359946,
+      "epoch": 0.7790862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 3.0089444842983908e-06,
-      "loss": 0.747,
+      "learning_rate": 2.4523615510732734e-06,
+      "loss": 0.7959,
       "step": 27455
     },
     {
-      "epoch": 0.7540578396638379,
+      "epoch": 0.7791146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 3.0083084915692384e-06,
-      "loss": 0.8367,
+      "learning_rate": 2.4517586708642427e-06,
+      "loss": 0.7569,
       "step": 27456
     },
     {
-      "epoch": 0.754085303891681,
+      "epoch": 0.7791430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 3.0076725541610584e-06,
-      "loss": 0.76,
+      "learning_rate": 2.4511558544155434e-06,
+      "loss": 0.9086,
       "step": 27457
     },
     {
-      "epoch": 0.7541127681195243,
+      "epoch": 0.7791713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 3.0070366720788814e-06,
-      "loss": 0.8691,
+      "learning_rate": 2.4505531017322705e-06,
+      "loss": 0.8408,
       "step": 27458
     },
     {
-      "epoch": 0.7541402323473676,
+      "epoch": 0.779199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 3.006400845327739e-06,
-      "loss": 0.7695,
+      "learning_rate": 2.449950412819516e-06,
+      "loss": 0.8616,
       "step": 27459
     },
     {
-      "epoch": 0.7541676965752108,
+      "epoch": 0.779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 3.005765073912662e-06,
-      "loss": 0.7932,
+      "learning_rate": 2.449347787682368e-06,
+      "loss": 0.8647,
       "step": 27460
     },
     {
-      "epoch": 0.754195160803054,
+      "epoch": 0.7792565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 3.005129357838684e-06,
-      "loss": 0.8318,
+      "learning_rate": 2.448745226325919e-06,
+      "loss": 0.8783,
       "step": 27461
     },
     {
-      "epoch": 0.7542226250308972,
+      "epoch": 0.7792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 3.004493697110833e-06,
-      "loss": 0.7832,
+      "learning_rate": 2.4481427287552606e-06,
+      "loss": 0.7701,
       "step": 27462
     },
     {
-      "epoch": 0.7542500892587405,
+      "epoch": 0.7793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 3.0038580917341343e-06,
-      "loss": 0.8558,
+      "learning_rate": 2.447540294975478e-06,
+      "loss": 0.7599,
       "step": 27463
     },
     {
-      "epoch": 0.7542775534865838,
+      "epoch": 0.7793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 3.0032225417136196e-06,
-      "loss": 0.6993,
+      "learning_rate": 2.4469379249916614e-06,
+      "loss": 0.8557,
       "step": 27464
     },
     {
-      "epoch": 0.7543050177144269,
+      "epoch": 0.7793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 3.0025870470543194e-06,
-      "loss": 0.8047,
+      "learning_rate": 2.4463356188089025e-06,
+      "loss": 0.9209,
       "step": 27465
     },
     {
-      "epoch": 0.7543324819422702,
+      "epoch": 0.7793984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 3.001951607761261e-06,
-      "loss": 0.8468,
+      "learning_rate": 2.445733376432284e-06,
+      "loss": 0.8372,
       "step": 27466
     },
     {
-      "epoch": 0.7543599461701135,
+      "epoch": 0.7794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 3.0013162238394745e-06,
-      "loss": 0.7634,
+      "learning_rate": 2.4451311978668947e-06,
+      "loss": 0.6985,
       "step": 27467
     },
     {
-      "epoch": 0.7543874103979566,
+      "epoch": 0.7794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 3.0006808952939824e-06,
-      "loss": 0.8099,
+      "learning_rate": 2.4445290831178246e-06,
+      "loss": 0.8313,
       "step": 27468
     },
     {
-      "epoch": 0.7544148746257999,
+      "epoch": 0.7794835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 3.0000456221298148e-06,
-      "loss": 0.9891,
+      "learning_rate": 2.443927032190154e-06,
+      "loss": 0.8324,
       "step": 27469
     },
     {
-      "epoch": 0.7544423388536431,
+      "epoch": 0.7795119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 2.999410404352e-06,
-      "loss": 0.9573,
+      "learning_rate": 2.4433250450889724e-06,
+      "loss": 0.8911,
       "step": 27470
     },
     {
-      "epoch": 0.7544698030814864,
+      "epoch": 0.7795402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 2.998775241965559e-06,
-      "loss": 0.8509,
+      "learning_rate": 2.4427231218193628e-06,
+      "loss": 0.743,
       "step": 27471
     },
     {
-      "epoch": 0.7544972673093296,
+      "epoch": 0.7795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 2.9981401349755213e-06,
-      "loss": 0.904,
+      "learning_rate": 2.442121262386411e-06,
+      "loss": 0.7963,
       "step": 27472
     },
     {
-      "epoch": 0.7545247315371728,
+      "epoch": 0.7795970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 2.9975050833869136e-06,
-      "loss": 0.7902,
+      "learning_rate": 2.4415194667952038e-06,
+      "loss": 0.7627,
       "step": 27473
     },
     {
-      "epoch": 0.7545521957650161,
+      "epoch": 0.7796254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 2.996870087204754e-06,
-      "loss": 0.7442,
+      "learning_rate": 2.4409177350508184e-06,
+      "loss": 0.7916,
       "step": 27474
     },
     {
-      "epoch": 0.7545796599928593,
+      "epoch": 0.7796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 2.9962351464340722e-06,
-      "loss": 0.8277,
+      "learning_rate": 2.4403160671583402e-06,
+      "loss": 0.9249,
       "step": 27475
     },
     {
-      "epoch": 0.7546071242207025,
+      "epoch": 0.7796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 2.9956002610798906e-06,
-      "loss": 0.7823,
+      "learning_rate": 2.439714463122855e-06,
+      "loss": 0.8108,
       "step": 27476
     },
     {
-      "epoch": 0.7546345884485458,
+      "epoch": 0.7797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.9949654311472365e-06,
-      "loss": 0.7976,
+      "learning_rate": 2.4391129229494402e-06,
+      "loss": 0.9347,
       "step": 27477
     },
     {
-      "epoch": 0.754662052676389,
+      "epoch": 0.779738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.9943306566411257e-06,
-      "loss": 0.8298,
+      "learning_rate": 2.438511446643177e-06,
+      "loss": 0.7884,
       "step": 27478
     },
     {
-      "epoch": 0.7546895169042322,
+      "epoch": 0.7797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 2.9936959375665873e-06,
-      "loss": 0.8543,
+      "learning_rate": 2.4379100342091512e-06,
+      "loss": 0.7951,
       "step": 27479
     },
     {
-      "epoch": 0.7547169811320755,
+      "epoch": 0.7797956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 2.9930612739286393e-06,
-      "loss": 0.8147,
+      "learning_rate": 2.437308685652435e-06,
+      "loss": 0.7465,
       "step": 27480
     },
     {
-      "epoch": 0.7547444453599187,
+      "epoch": 0.7798240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 2.992426665732303e-06,
-      "loss": 0.7561,
+      "learning_rate": 2.4367074009781135e-06,
+      "loss": 0.8725,
       "step": 27481
     },
     {
-      "epoch": 0.754771909587762,
+      "epoch": 0.7798524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 2.991792112982601e-06,
-      "loss": 0.7854,
+      "learning_rate": 2.4361061801912666e-06,
+      "loss": 0.8197,
       "step": 27482
     },
     {
-      "epoch": 0.7547993738156051,
+      "epoch": 0.7798808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 2.9911576156845554e-06,
-      "loss": 0.7781,
+      "learning_rate": 2.435505023296969e-06,
+      "loss": 0.7955,
       "step": 27483
     },
     {
-      "epoch": 0.7548268380434484,
+      "epoch": 0.7799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 2.9905231738431862e-06,
-      "loss": 0.8032,
+      "learning_rate": 2.4349039303002997e-06,
+      "loss": 0.6739,
       "step": 27484
     },
     {
-      "epoch": 0.7548543022712917,
+      "epoch": 0.7799375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 2.989888787463515e-06,
-      "loss": 0.8465,
+      "learning_rate": 2.434302901206338e-06,
+      "loss": 0.8372,
       "step": 27485
     },
     {
-      "epoch": 0.7548817664991349,
+      "epoch": 0.7799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 2.98925445655056e-06,
-      "loss": 0.7836,
+      "learning_rate": 2.4337019360201585e-06,
+      "loss": 0.8117,
       "step": 27486
     },
     {
-      "epoch": 0.7549092307269781,
+      "epoch": 0.7799943246311011,
       "grad_norm": 0.0,
-      "learning_rate": 2.988620181109336e-06,
-      "loss": 0.6665,
+      "learning_rate": 2.4331010347468397e-06,
+      "loss": 0.7798,
       "step": 27487
     },
     {
-      "epoch": 0.7549366949548213,
+      "epoch": 0.7800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 2.987985961144865e-06,
-      "loss": 0.8502,
+      "learning_rate": 2.4325001973914586e-06,
+      "loss": 0.8712,
       "step": 27488
     },
     {
-      "epoch": 0.7549641591826646,
+      "epoch": 0.7800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 2.9873517966621655e-06,
-      "loss": 0.8296,
+      "learning_rate": 2.4318994239590866e-06,
+      "loss": 0.8396,
       "step": 27489
     },
     {
-      "epoch": 0.7549916234105079,
+      "epoch": 0.7800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 2.986717687666255e-06,
-      "loss": 0.7938,
+      "learning_rate": 2.4312987144548005e-06,
+      "loss": 0.7719,
       "step": 27490
     },
     {
-      "epoch": 0.755019087638351,
+      "epoch": 0.7801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 2.9860836341621537e-06,
-      "loss": 0.805,
+      "learning_rate": 2.430698068883677e-06,
+      "loss": 0.7347,
       "step": 27491
     },
     {
-      "epoch": 0.7550465518661943,
+      "epoch": 0.7801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 2.9854496361548722e-06,
-      "loss": 0.7662,
+      "learning_rate": 2.4300974872507833e-06,
+      "loss": 0.8513,
       "step": 27492
     },
     {
-      "epoch": 0.7550740160940376,
+      "epoch": 0.7801645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 2.984815693649431e-06,
-      "loss": 0.7169,
+      "learning_rate": 2.4294969695612e-06,
+      "loss": 0.8054,
       "step": 27493
     },
     {
-      "epoch": 0.7551014803218807,
+      "epoch": 0.7801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 2.9841818066508487e-06,
-      "loss": 0.8635,
+      "learning_rate": 2.428896515819994e-06,
+      "loss": 0.8,
       "step": 27494
     },
     {
-      "epoch": 0.755128944549724,
+      "epoch": 0.7802213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 2.9835479751641338e-06,
-      "loss": 0.8249,
+      "learning_rate": 2.428296126032239e-06,
+      "loss": 0.7174,
       "step": 27495
     },
     {
-      "epoch": 0.7551564087775672,
+      "epoch": 0.780249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 2.9829141991943087e-06,
-      "loss": 0.8013,
+      "learning_rate": 2.4276958002030093e-06,
+      "loss": 0.7118,
       "step": 27496
     },
     {
-      "epoch": 0.7551838730054105,
+      "epoch": 0.7802780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 2.9822804787463812e-06,
-      "loss": 0.7582,
+      "learning_rate": 2.4270955383373717e-06,
+      "loss": 0.8947,
       "step": 27497
     },
     {
-      "epoch": 0.7552113372332537,
+      "epoch": 0.7803064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 2.9816468138253695e-06,
-      "loss": 0.8355,
+      "learning_rate": 2.426495340440398e-06,
+      "loss": 0.8293,
       "step": 27498
     },
     {
-      "epoch": 0.7552388014610969,
+      "epoch": 0.7803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 2.981013204436287e-06,
-      "loss": 0.7743,
+      "learning_rate": 2.4258952065171603e-06,
+      "loss": 0.8757,
       "step": 27499
     },
     {
-      "epoch": 0.7552662656889402,
+      "epoch": 0.7803632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 2.9803796505841465e-06,
-      "loss": 0.8495,
+      "learning_rate": 2.4252951365727218e-06,
+      "loss": 0.889,
       "step": 27500
     },
     {
-      "epoch": 0.7552937299167833,
+      "epoch": 0.7803916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 2.9797461522739645e-06,
-      "loss": 0.7483,
+      "learning_rate": 2.424695130612159e-06,
+      "loss": 0.784,
       "step": 27501
     },
     {
-      "epoch": 0.7553211941446266,
+      "epoch": 0.7804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 2.9791127095107463e-06,
-      "loss": 0.8234,
+      "learning_rate": 2.4240951886405396e-06,
+      "loss": 0.8265,
       "step": 27502
     },
     {
-      "epoch": 0.7553486583724699,
+      "epoch": 0.7804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 2.9784793222995122e-06,
-      "loss": 0.8652,
+      "learning_rate": 2.4234953106629256e-06,
+      "loss": 0.7767,
       "step": 27503
     },
     {
-      "epoch": 0.7553761226003131,
+      "epoch": 0.7804767309875141,
       "grad_norm": 0.0,
-      "learning_rate": 2.977845990645266e-06,
-      "loss": 0.8329,
+      "learning_rate": 2.4228954966843864e-06,
+      "loss": 0.8055,
       "step": 27504
     },
     {
-      "epoch": 0.7554035868281563,
+      "epoch": 0.7805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 2.9772127145530225e-06,
-      "loss": 0.8929,
+      "learning_rate": 2.4222957467099937e-06,
+      "loss": 0.7767,
       "step": 27505
     },
     {
-      "epoch": 0.7554310510559996,
+      "epoch": 0.780533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.9765794940277936e-06,
-      "loss": 0.8726,
+      "learning_rate": 2.421696060744806e-06,
+      "loss": 0.9547,
       "step": 27506
     },
     {
-      "epoch": 0.7554585152838428,
+      "epoch": 0.7805618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 2.975946329074587e-06,
-      "loss": 0.8345,
+      "learning_rate": 2.421096438793893e-06,
+      "loss": 0.7995,
       "step": 27507
     },
     {
-      "epoch": 0.7554859795116861,
+      "epoch": 0.7805902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 2.975313219698417e-06,
-      "loss": 0.9006,
+      "learning_rate": 2.4204968808623207e-06,
+      "loss": 0.7224,
       "step": 27508
     },
     {
-      "epoch": 0.7555134437395292,
+      "epoch": 0.7806186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 2.9746801659042877e-06,
-      "loss": 0.9138,
+      "learning_rate": 2.41989738695515e-06,
+      "loss": 0.8677,
       "step": 27509
     },
     {
-      "epoch": 0.7555409079673725,
+      "epoch": 0.7806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 2.9740471676972126e-06,
-      "loss": 0.8132,
+      "learning_rate": 2.4192979570774467e-06,
+      "loss": 0.74,
       "step": 27510
     },
     {
-      "epoch": 0.7555683721952158,
+      "epoch": 0.7806753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 2.9734142250821953e-06,
-      "loss": 0.846,
+      "learning_rate": 2.4186985912342773e-06,
+      "loss": 0.791,
       "step": 27511
     },
     {
-      "epoch": 0.7555958364230589,
+      "epoch": 0.7807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 2.9727813380642458e-06,
-      "loss": 0.767,
+      "learning_rate": 2.4180992894306985e-06,
+      "loss": 0.7564,
       "step": 27512
     },
     {
-      "epoch": 0.7556233006509022,
+      "epoch": 0.7807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 2.9721485066483723e-06,
-      "loss": 0.8647,
+      "learning_rate": 2.4175000516717763e-06,
+      "loss": 0.892,
       "step": 27513
     },
     {
-      "epoch": 0.7556507648787454,
+      "epoch": 0.7807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 2.9715157308395858e-06,
-      "loss": 0.8056,
+      "learning_rate": 2.416900877962575e-06,
+      "loss": 0.8417,
       "step": 27514
     },
     {
-      "epoch": 0.7556782291065887,
+      "epoch": 0.780788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.9708830106428864e-06,
-      "loss": 0.7411,
+      "learning_rate": 2.4163017683081503e-06,
+      "loss": 0.7145,
       "step": 27515
     },
     {
-      "epoch": 0.7557056933344319,
+      "epoch": 0.7808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 2.9702503460632838e-06,
-      "loss": 0.8353,
+      "learning_rate": 2.415702722713564e-06,
+      "loss": 0.7684,
       "step": 27516
     },
     {
-      "epoch": 0.7557331575622751,
+      "epoch": 0.7808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 2.969617737105783e-06,
-      "loss": 0.8682,
+      "learning_rate": 2.4151037411838784e-06,
+      "loss": 0.7252,
       "step": 27517
     },
     {
-      "epoch": 0.7557606217901184,
+      "epoch": 0.7808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 2.9689851837753935e-06,
-      "loss": 0.8141,
+      "learning_rate": 2.414504823724153e-06,
+      "loss": 0.8244,
       "step": 27518
     },
     {
-      "epoch": 0.7557880860179617,
+      "epoch": 0.7809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 2.968352686077113e-06,
-      "loss": 0.841,
+      "learning_rate": 2.4139059703394486e-06,
+      "loss": 0.8143,
       "step": 27519
     },
     {
-      "epoch": 0.7558155502458048,
+      "epoch": 0.7809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 2.9677202440159536e-06,
-      "loss": 0.846,
+      "learning_rate": 2.4133071810348198e-06,
+      "loss": 0.7704,
       "step": 27520
     },
     {
-      "epoch": 0.7558430144736481,
+      "epoch": 0.7809591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 2.967087857596912e-06,
-      "loss": 0.8341,
+      "learning_rate": 2.4127084558153248e-06,
+      "loss": 0.8251,
       "step": 27521
     },
     {
-      "epoch": 0.7558704787014913,
+      "epoch": 0.7809875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 2.9664555268249962e-06,
-      "loss": 0.8278,
+      "learning_rate": 2.4121097946860263e-06,
+      "loss": 0.796,
       "step": 27522
     },
     {
-      "epoch": 0.7558979429293345,
+      "epoch": 0.7810158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 2.9658232517052077e-06,
-      "loss": 0.79,
+      "learning_rate": 2.4115111976519735e-06,
+      "loss": 0.7093,
       "step": 27523
     },
     {
-      "epoch": 0.7559254071571778,
+      "epoch": 0.781044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.9651910322425514e-06,
-      "loss": 0.7664,
+      "learning_rate": 2.410912664718228e-06,
+      "loss": 0.8153,
       "step": 27524
     },
     {
-      "epoch": 0.755952871385021,
+      "epoch": 0.781072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 2.964558868442031e-06,
-      "loss": 0.8654,
+      "learning_rate": 2.410314195889847e-06,
+      "loss": 0.7803,
       "step": 27525
     },
     {
-      "epoch": 0.7559803356128643,
+      "epoch": 0.7811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 2.963926760308644e-06,
-      "loss": 0.7839,
+      "learning_rate": 2.4097157911718793e-06,
+      "loss": 0.8488,
       "step": 27526
     },
     {
-      "epoch": 0.7560077998407074,
+      "epoch": 0.7811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 2.963294707847396e-06,
-      "loss": 0.7741,
+      "learning_rate": 2.409117450569385e-06,
+      "loss": 0.8626,
       "step": 27527
     },
     {
-      "epoch": 0.7560352640685507,
+      "epoch": 0.7811577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 2.9626627110632843e-06,
-      "loss": 0.8401,
+      "learning_rate": 2.4085191740874182e-06,
+      "loss": 0.811,
       "step": 27528
     },
     {
-      "epoch": 0.756062728296394,
+      "epoch": 0.7811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 2.9620307699613103e-06,
-      "loss": 0.7877,
+      "learning_rate": 2.4079209617310296e-06,
+      "loss": 0.7262,
       "step": 27529
     },
     {
-      "epoch": 0.7560901925242371,
+      "epoch": 0.7812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 2.9613988845464746e-06,
-      "loss": 0.8513,
+      "learning_rate": 2.407322813505273e-06,
+      "loss": 0.8274,
       "step": 27530
     },
     {
-      "epoch": 0.7561176567520804,
+      "epoch": 0.7812429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 2.9607670548237787e-06,
-      "loss": 0.8604,
+      "learning_rate": 2.406724729415203e-06,
+      "loss": 0.8796,
       "step": 27531
     },
     {
-      "epoch": 0.7561451209799237,
+      "epoch": 0.7812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 2.960135280798223e-06,
-      "loss": 0.8045,
+      "learning_rate": 2.4061267094658693e-06,
+      "loss": 0.8259,
       "step": 27532
     },
     {
-      "epoch": 0.7561725852077669,
+      "epoch": 0.7812996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 2.9595035624748003e-06,
-      "loss": 0.8366,
+      "learning_rate": 2.405528753662328e-06,
+      "loss": 0.769,
       "step": 27533
     },
     {
-      "epoch": 0.7562000494356101,
+      "epoch": 0.7813280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 2.9588718998585166e-06,
-      "loss": 0.7926,
+      "learning_rate": 2.4049308620096244e-06,
+      "loss": 0.9118,
       "step": 27534
     },
     {
-      "epoch": 0.7562275136634533,
+      "epoch": 0.7813564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 2.958240292954363e-06,
-      "loss": 0.8763,
+      "learning_rate": 2.40433303451281e-06,
+      "loss": 0.8746,
       "step": 27535
     },
     {
-      "epoch": 0.7562549778912966,
+      "epoch": 0.7813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 2.9576087417673395e-06,
-      "loss": 0.7937,
+      "learning_rate": 2.40373527117694e-06,
+      "loss": 0.7445,
       "step": 27536
     },
     {
-      "epoch": 0.7562824421191399,
+      "epoch": 0.7814131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 2.9569772463024425e-06,
-      "loss": 0.845,
+      "learning_rate": 2.403137572007056e-06,
+      "loss": 0.8535,
       "step": 27537
     },
     {
-      "epoch": 0.756309906346983,
+      "epoch": 0.7814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 2.9563458065646744e-06,
-      "loss": 0.8497,
+      "learning_rate": 2.4025399370082114e-06,
+      "loss": 0.8383,
       "step": 27538
     },
     {
-      "epoch": 0.7563373705748263,
+      "epoch": 0.7814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 2.955714422559023e-06,
-      "loss": 0.8473,
+      "learning_rate": 2.4019423661854557e-06,
+      "loss": 0.8137,
       "step": 27539
     },
     {
-      "epoch": 0.7563648348026695,
+      "epoch": 0.7814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 2.9550830942904884e-06,
-      "loss": 0.7861,
+      "learning_rate": 2.4013448595438316e-06,
+      "loss": 0.7654,
       "step": 27540
     },
     {
-      "epoch": 0.7563922990305127,
+      "epoch": 0.7815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 2.954451821764065e-06,
-      "loss": 0.8623,
+      "learning_rate": 2.4007474170883893e-06,
+      "loss": 0.7693,
       "step": 27541
     },
     {
-      "epoch": 0.756419763258356,
+      "epoch": 0.7815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 2.9538206049847505e-06,
-      "loss": 0.7134,
+      "learning_rate": 2.400150038824177e-06,
+      "loss": 0.7634,
       "step": 27542
     },
     {
-      "epoch": 0.7564472274861992,
+      "epoch": 0.781583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 2.9531894439575338e-06,
-      "loss": 0.7825,
+      "learning_rate": 2.399552724756237e-06,
+      "loss": 0.8715,
       "step": 27543
     },
     {
-      "epoch": 0.7564746917140425,
+      "epoch": 0.7816118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 2.9525583386874157e-06,
-      "loss": 0.9092,
+      "learning_rate": 2.398955474889617e-06,
+      "loss": 0.7338,
       "step": 27544
     },
     {
-      "epoch": 0.7565021559418857,
+      "epoch": 0.7816401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 2.9519272891793828e-06,
-      "loss": 0.866,
+      "learning_rate": 2.3983582892293642e-06,
+      "loss": 0.7808,
       "step": 27545
     },
     {
-      "epoch": 0.7565296201697289,
+      "epoch": 0.7816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 2.9512962954384304e-06,
-      "loss": 0.8338,
+      "learning_rate": 2.3977611677805146e-06,
+      "loss": 0.7773,
       "step": 27546
     },
     {
-      "epoch": 0.7565570843975722,
+      "epoch": 0.7816969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 2.9506653574695523e-06,
-      "loss": 0.8192,
+      "learning_rate": 2.397164110548125e-06,
+      "loss": 0.7991,
       "step": 27547
     },
     {
-      "epoch": 0.7565845486254154,
+      "epoch": 0.7817253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 2.950034475277741e-06,
-      "loss": 0.9006,
+      "learning_rate": 2.3965671175372274e-06,
+      "loss": 0.8501,
       "step": 27548
     },
     {
-      "epoch": 0.7566120128532586,
+      "epoch": 0.7817536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 2.9494036488679913e-06,
-      "loss": 0.8345,
+      "learning_rate": 2.3959701887528707e-06,
+      "loss": 0.887,
       "step": 27549
     },
     {
-      "epoch": 0.7566394770811019,
+      "epoch": 0.7817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 2.9487728782452873e-06,
-      "loss": 0.82,
+      "learning_rate": 2.3953733242000977e-06,
+      "loss": 0.8336,
       "step": 27550
     },
     {
-      "epoch": 0.7566669413089451,
+      "epoch": 0.7818104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 2.948142163414627e-06,
-      "loss": 0.8281,
+      "learning_rate": 2.394776523883946e-06,
+      "loss": 0.8802,
       "step": 27551
     },
     {
-      "epoch": 0.7566944055367884,
+      "epoch": 0.7818388195232691,
       "grad_norm": 0.0,
-      "learning_rate": 2.947511504380994e-06,
-      "loss": 0.8377,
+      "learning_rate": 2.3941797878094576e-06,
+      "loss": 0.8407,
       "step": 27552
     },
     {
-      "epoch": 0.7567218697646315,
+      "epoch": 0.7818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 2.9468809011493825e-06,
-      "loss": 0.8379,
+      "learning_rate": 2.393583115981678e-06,
+      "loss": 0.7924,
       "step": 27553
     },
     {
-      "epoch": 0.7567493339924748,
+      "epoch": 0.7818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.946250353724781e-06,
-      "loss": 0.7989,
+      "learning_rate": 2.3929865084056415e-06,
+      "loss": 0.7857,
       "step": 27554
     },
     {
-      "epoch": 0.7567767982203181,
+      "epoch": 0.7819239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 2.9456198621121824e-06,
-      "loss": 0.887,
+      "learning_rate": 2.392389965086389e-06,
+      "loss": 0.8277,
       "step": 27555
     },
     {
-      "epoch": 0.7568042624481612,
+      "epoch": 0.7819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 2.944989426316569e-06,
-      "loss": 0.8463,
+      "learning_rate": 2.391793486028964e-06,
+      "loss": 0.9047,
       "step": 27556
     },
     {
-      "epoch": 0.7568317266760045,
+      "epoch": 0.7819807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 2.944359046342932e-06,
-      "loss": 0.8315,
+      "learning_rate": 2.391197071238398e-06,
+      "loss": 0.7626,
       "step": 27557
     },
     {
-      "epoch": 0.7568591909038478,
+      "epoch": 0.7820090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 2.943728722196263e-06,
-      "loss": 0.819,
+      "learning_rate": 2.3906007207197325e-06,
+      "loss": 0.8416,
       "step": 27558
     },
     {
-      "epoch": 0.756886655131691,
+      "epoch": 0.7820374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 2.943098453881542e-06,
-      "loss": 0.8089,
+      "learning_rate": 2.3900044344780084e-06,
+      "loss": 0.844,
       "step": 27559
     },
     {
-      "epoch": 0.7569141193595342,
+      "epoch": 0.7820658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 2.94246824140376e-06,
-      "loss": 0.8334,
+      "learning_rate": 2.389408212518255e-06,
+      "loss": 0.7634,
       "step": 27560
     },
     {
-      "epoch": 0.7569415835873774,
+      "epoch": 0.782094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 2.9418380847679053e-06,
-      "loss": 0.8046,
+      "learning_rate": 2.3888120548455127e-06,
+      "loss": 0.8728,
       "step": 27561
     },
     {
-      "epoch": 0.7569690478152207,
+      "epoch": 0.782122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 2.9412079839789585e-06,
-      "loss": 0.7894,
+      "learning_rate": 2.388215961464817e-06,
+      "loss": 0.7502,
       "step": 27562
     },
     {
-      "epoch": 0.756996512043064,
+      "epoch": 0.7821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 2.94057793904191e-06,
-      "loss": 0.8274,
+      "learning_rate": 2.3876199323812043e-06,
+      "loss": 0.756,
       "step": 27563
     },
     {
-      "epoch": 0.7570239762709071,
+      "epoch": 0.7821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 2.939947949961742e-06,
-      "loss": 0.8391,
+      "learning_rate": 2.38702396759971e-06,
+      "loss": 0.8145,
       "step": 27564
     },
     {
-      "epoch": 0.7570514404987504,
+      "epoch": 0.7822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 2.939318016743441e-06,
-      "loss": 0.8405,
+      "learning_rate": 2.3864280671253647e-06,
+      "loss": 0.9206,
       "step": 27565
     },
     {
-      "epoch": 0.7570789047265936,
+      "epoch": 0.7822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 2.9386881393919943e-06,
-      "loss": 0.8385,
+      "learning_rate": 2.385832230963203e-06,
+      "loss": 0.8768,
       "step": 27566
     },
     {
-      "epoch": 0.7571063689544368,
+      "epoch": 0.7822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 2.938058317912382e-06,
-      "loss": 0.8779,
+      "learning_rate": 2.385236459118262e-06,
+      "loss": 0.6989,
       "step": 27567
     },
     {
-      "epoch": 0.7571338331822801,
+      "epoch": 0.7822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 2.9374285523095857e-06,
-      "loss": 0.8658,
+      "learning_rate": 2.384640751595567e-06,
+      "loss": 0.7576,
       "step": 27568
     },
     {
-      "epoch": 0.7571612974101233,
+      "epoch": 0.7823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 2.9367988425885895e-06,
-      "loss": 0.8082,
+      "learning_rate": 2.384045108400155e-06,
+      "loss": 0.7151,
       "step": 27569
     },
     {
-      "epoch": 0.7571887616379666,
+      "epoch": 0.7823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 2.9361691887543764e-06,
-      "loss": 0.8181,
+      "learning_rate": 2.3834495295370586e-06,
+      "loss": 0.7782,
       "step": 27570
     },
     {
-      "epoch": 0.7572162258658098,
+      "epoch": 0.7823779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 2.93553959081193e-06,
-      "loss": 0.8152,
+      "learning_rate": 2.3828540150113043e-06,
+      "loss": 0.7952,
       "step": 27571
     },
     {
-      "epoch": 0.757243690093653,
+      "epoch": 0.7824063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 2.9349100487662307e-06,
-      "loss": 0.8303,
+      "learning_rate": 2.382258564827924e-06,
+      "loss": 0.8745,
       "step": 27572
     },
     {
-      "epoch": 0.7572711543214963,
+      "epoch": 0.7824347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 2.9342805626222626e-06,
-      "loss": 0.7793,
+      "learning_rate": 2.381663178991951e-06,
+      "loss": 0.7855,
       "step": 27573
     },
     {
-      "epoch": 0.7572986185493394,
+      "epoch": 0.7824631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 2.9336511323850037e-06,
-      "loss": 0.814,
+      "learning_rate": 2.381067857508409e-06,
+      "loss": 0.7903,
       "step": 27574
     },
     {
-      "epoch": 0.7573260827771827,
+      "epoch": 0.7824914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 2.9330217580594313e-06,
-      "loss": 0.8733,
+      "learning_rate": 2.3804726003823287e-06,
+      "loss": 0.8437,
       "step": 27575
     },
     {
-      "epoch": 0.757353547005026,
+      "epoch": 0.7825198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 2.932392439650529e-06,
-      "loss": 0.7976,
+      "learning_rate": 2.3798774076187412e-06,
+      "loss": 0.7059,
       "step": 27576
     },
     {
-      "epoch": 0.7573810112328692,
+      "epoch": 0.7825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 2.931763177163274e-06,
-      "loss": 0.8814,
+      "learning_rate": 2.3792822792226676e-06,
+      "loss": 0.8068,
       "step": 27577
     },
     {
-      "epoch": 0.7574084754607124,
+      "epoch": 0.7825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 2.9311339706026474e-06,
-      "loss": 0.7632,
+      "learning_rate": 2.3786872151991435e-06,
+      "loss": 0.7927,
       "step": 27578
     },
     {
-      "epoch": 0.7574359396885556,
+      "epoch": 0.7826049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 2.9305048199736296e-06,
-      "loss": 0.821,
+      "learning_rate": 2.3780922155531896e-06,
+      "loss": 0.8153,
       "step": 27579
     },
     {
-      "epoch": 0.7574634039163989,
+      "epoch": 0.782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 2.929875725281194e-06,
-      "loss": 0.8115,
+      "learning_rate": 2.377497280289832e-06,
+      "loss": 0.785,
       "step": 27580
     },
     {
-      "epoch": 0.7574908681442422,
+      "epoch": 0.7826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 2.9292466865303205e-06,
-      "loss": 0.9404,
+      "learning_rate": 2.3769024094141015e-06,
+      "loss": 0.8503,
       "step": 27581
     },
     {
-      "epoch": 0.7575183323720853,
+      "epoch": 0.7826901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 2.928617703725988e-06,
-      "loss": 0.9026,
+      "learning_rate": 2.376307602931016e-06,
+      "loss": 0.8555,
       "step": 27582
     },
     {
-      "epoch": 0.7575457965999286,
+      "epoch": 0.7827185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 2.9279887768731695e-06,
-      "loss": 0.7679,
+      "learning_rate": 2.3757128608456028e-06,
+      "loss": 0.7725,
       "step": 27583
     },
     {
-      "epoch": 0.7575732608277719,
+      "epoch": 0.7827468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 2.927359905976842e-06,
-      "loss": 0.799,
+      "learning_rate": 2.375118183162889e-06,
+      "loss": 0.7789,
       "step": 27584
     },
     {
-      "epoch": 0.757600725055615,
+      "epoch": 0.7827752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 2.9267310910419867e-06,
-      "loss": 0.808,
+      "learning_rate": 2.374523569887891e-06,
+      "loss": 0.7931,
       "step": 27585
     },
     {
-      "epoch": 0.7576281892834583,
+      "epoch": 0.7828036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 2.9261023320735714e-06,
-      "loss": 0.823,
+      "learning_rate": 2.373929021025635e-06,
+      "loss": 0.7587,
       "step": 27586
     },
     {
-      "epoch": 0.7576556535113015,
+      "epoch": 0.7828320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 2.925473629076575e-06,
-      "loss": 0.7239,
+      "learning_rate": 2.3733345365811465e-06,
+      "loss": 0.7836,
       "step": 27587
     },
     {
-      "epoch": 0.7576831177391448,
+      "epoch": 0.7828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 2.924844982055972e-06,
-      "loss": 0.8125,
+      "learning_rate": 2.372740116559442e-06,
+      "loss": 0.8496,
       "step": 27588
     },
     {
-      "epoch": 0.757710581966988,
+      "epoch": 0.78288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.9242163910167354e-06,
-      "loss": 0.7178,
+      "learning_rate": 2.372145760965544e-06,
+      "loss": 0.9076,
       "step": 27589
     },
     {
-      "epoch": 0.7577380461948312,
+      "epoch": 0.7829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 2.923587855963843e-06,
-      "loss": 0.9081,
+      "learning_rate": 2.3715514698044762e-06,
+      "loss": 0.7965,
       "step": 27590
     },
     {
-      "epoch": 0.7577655104226745,
+      "epoch": 0.7829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 2.922959376902265e-06,
-      "loss": 0.9313,
+      "learning_rate": 2.370957243081253e-06,
+      "loss": 0.8714,
       "step": 27591
     },
     {
-      "epoch": 0.7577929746505176,
+      "epoch": 0.7829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.9223309538369692e-06,
-      "loss": 0.9143,
+      "learning_rate": 2.370363080800898e-06,
+      "loss": 0.8611,
       "step": 27592
     },
     {
-      "epoch": 0.7578204388783609,
+      "epoch": 0.7830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 2.9217025867729343e-06,
-      "loss": 0.7993,
+      "learning_rate": 2.369768982968429e-06,
+      "loss": 0.7718,
       "step": 27593
     },
     {
-      "epoch": 0.7578479031062042,
+      "epoch": 0.7830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 2.921074275715129e-06,
-      "loss": 0.8662,
+      "learning_rate": 2.369174949588864e-06,
+      "loss": 0.8116,
       "step": 27594
     },
     {
-      "epoch": 0.7578753673340474,
+      "epoch": 0.7830590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 2.920446020668527e-06,
-      "loss": 0.8079,
+      "learning_rate": 2.368580980667221e-06,
+      "loss": 0.8237,
       "step": 27595
     },
     {
-      "epoch": 0.7579028315618906,
+      "epoch": 0.7830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 2.919817821638098e-06,
-      "loss": 0.8115,
+      "learning_rate": 2.3679870762085198e-06,
+      "loss": 0.8474,
       "step": 27596
     },
     {
-      "epoch": 0.7579302957897339,
+      "epoch": 0.7831157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 2.9191896786288165e-06,
-      "loss": 0.8426,
+      "learning_rate": 2.367393236217773e-06,
+      "loss": 0.7947,
       "step": 27597
     },
     {
-      "epoch": 0.7579577600175771,
+      "epoch": 0.783144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 2.9185615916456488e-06,
-      "loss": 0.8087,
+      "learning_rate": 2.3667994607000013e-06,
+      "loss": 0.9276,
       "step": 27598
     },
     {
-      "epoch": 0.7579852242454204,
+      "epoch": 0.7831725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 2.9179335606935623e-06,
-      "loss": 0.8515,
+      "learning_rate": 2.3662057496602142e-06,
+      "loss": 0.8121,
       "step": 27599
     },
     {
-      "epoch": 0.7580126884732635,
+      "epoch": 0.7832009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 2.917305585777528e-06,
-      "loss": 0.8758,
+      "learning_rate": 2.365612103103432e-06,
+      "loss": 0.7814,
       "step": 27600
     },
     {
-      "epoch": 0.7580401527011068,
+      "epoch": 0.7832292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 2.9166776669025166e-06,
-      "loss": 0.9335,
+      "learning_rate": 2.3650185210346686e-06,
+      "loss": 0.8612,
       "step": 27601
     },
     {
-      "epoch": 0.7580676169289501,
+      "epoch": 0.7832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 2.916049804073494e-06,
-      "loss": 0.828,
+      "learning_rate": 2.3644250034589342e-06,
+      "loss": 0.7408,
       "step": 27602
     },
     {
-      "epoch": 0.7580950811567932,
+      "epoch": 0.7832860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 2.9154219972954334e-06,
-      "loss": 0.8797,
+      "learning_rate": 2.3638315503812457e-06,
+      "loss": 0.8016,
       "step": 27603
     },
     {
-      "epoch": 0.7581225453846365,
+      "epoch": 0.7833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 2.914794246573295e-06,
-      "loss": 0.8281,
+      "learning_rate": 2.363238161806618e-06,
+      "loss": 0.7655,
       "step": 27604
     },
     {
-      "epoch": 0.7581500096124797,
+      "epoch": 0.7833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.914166551912049e-06,
-      "loss": 0.7114,
+      "learning_rate": 2.362644837740059e-06,
+      "loss": 0.7998,
       "step": 27605
     },
     {
-      "epoch": 0.758177473840323,
+      "epoch": 0.7833711691259931,
       "grad_norm": 0.0,
-      "learning_rate": 2.9135389133166647e-06,
-      "loss": 0.748,
+      "learning_rate": 2.3620515781865803e-06,
+      "loss": 0.8592,
       "step": 27606
     },
     {
-      "epoch": 0.7582049380681662,
+      "epoch": 0.7833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 2.912911330792103e-06,
-      "loss": 0.854,
+      "learning_rate": 2.3614583831512006e-06,
+      "loss": 0.8704,
       "step": 27607
     },
     {
-      "epoch": 0.7582324022960094,
+      "epoch": 0.783427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.9122838043433323e-06,
-      "loss": 0.7802,
+      "learning_rate": 2.3608652526389177e-06,
+      "loss": 0.7385,
       "step": 27608
     },
     {
-      "epoch": 0.7582598665238527,
+      "epoch": 0.7834562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 2.9116563339753202e-06,
-      "loss": 0.7421,
+      "learning_rate": 2.360272186654753e-06,
+      "loss": 0.8209,
       "step": 27609
     },
     {
-      "epoch": 0.758287330751696,
+      "epoch": 0.7834846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 2.911028919693025e-06,
-      "loss": 0.8881,
+      "learning_rate": 2.359679185203715e-06,
+      "loss": 0.7538,
       "step": 27610
     },
     {
-      "epoch": 0.7583147949795391,
+      "epoch": 0.7835130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 2.910401561501417e-06,
-      "loss": 0.7589,
+      "learning_rate": 2.3590862482908085e-06,
+      "loss": 0.9343,
       "step": 27611
     },
     {
-      "epoch": 0.7583422592073824,
+      "epoch": 0.7835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 2.909774259405457e-06,
-      "loss": 0.7133,
+      "learning_rate": 2.3584933759210426e-06,
+      "loss": 0.784,
       "step": 27612
     },
     {
-      "epoch": 0.7583697234352256,
+      "epoch": 0.7835698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 2.909147013410113e-06,
-      "loss": 0.8333,
+      "learning_rate": 2.3579005680994305e-06,
+      "loss": 0.7528,
       "step": 27613
     },
     {
-      "epoch": 0.7583971876630689,
+      "epoch": 0.7835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 2.9085198235203406e-06,
-      "loss": 0.7764,
+      "learning_rate": 2.3573078248309722e-06,
+      "loss": 0.828,
       "step": 27614
     },
     {
-      "epoch": 0.7584246518909121,
+      "epoch": 0.7836265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 2.907892689741111e-06,
-      "loss": 0.9188,
+      "learning_rate": 2.3567151461206792e-06,
+      "loss": 0.8519,
       "step": 27615
     },
     {
-      "epoch": 0.7584521161187553,
+      "epoch": 0.7836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 2.9072656120773768e-06,
-      "loss": 0.942,
+      "learning_rate": 2.356122531973559e-06,
+      "loss": 0.8255,
       "step": 27616
     },
     {
-      "epoch": 0.7584795803465986,
+      "epoch": 0.783683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.9066385905341054e-06,
-      "loss": 0.8124,
+      "learning_rate": 2.355529982394613e-06,
+      "loss": 0.804,
       "step": 27617
     },
     {
-      "epoch": 0.7585070445744417,
+      "epoch": 0.7837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 2.9060116251162573e-06,
-      "loss": 0.8959,
+      "learning_rate": 2.3549374973888483e-06,
+      "loss": 0.7765,
       "step": 27618
     },
     {
-      "epoch": 0.758534508802285,
+      "epoch": 0.7837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 2.905384715828793e-06,
-      "loss": 0.9329,
+      "learning_rate": 2.354345076961274e-06,
+      "loss": 0.8103,
       "step": 27619
     },
     {
-      "epoch": 0.7585619730301283,
+      "epoch": 0.7837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 2.904757862676676e-06,
-      "loss": 0.8172,
+      "learning_rate": 2.3537527211168877e-06,
+      "loss": 0.8684,
       "step": 27620
     },
     {
-      "epoch": 0.7585894372579715,
+      "epoch": 0.7837968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 2.90413106566486e-06,
-      "loss": 0.8327,
+      "learning_rate": 2.353160429860695e-06,
+      "loss": 0.8977,
       "step": 27621
     },
     {
-      "epoch": 0.7586169014858147,
+      "epoch": 0.7838251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 2.903504324798312e-06,
-      "loss": 0.8271,
+      "learning_rate": 2.3525682031977027e-06,
+      "loss": 0.7946,
       "step": 27622
     },
     {
-      "epoch": 0.758644365713658,
+      "epoch": 0.7838535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 2.9028776400819826e-06,
-      "loss": 0.7918,
+      "learning_rate": 2.3519760411329053e-06,
+      "loss": 0.874,
       "step": 27623
     },
     {
-      "epoch": 0.7586718299415012,
+      "epoch": 0.7838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 2.9022510115208346e-06,
-      "loss": 0.8277,
+      "learning_rate": 2.351383943671316e-06,
+      "loss": 0.8394,
       "step": 27624
     },
     {
-      "epoch": 0.7586992941693445,
+      "epoch": 0.7839103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 2.9016244391198256e-06,
-      "loss": 0.805,
+      "learning_rate": 2.3507919108179277e-06,
+      "loss": 0.8408,
       "step": 27625
     },
     {
-      "epoch": 0.7587267583971876,
+      "epoch": 0.783938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 2.9009979228839167e-06,
-      "loss": 0.791,
+      "learning_rate": 2.3501999425777433e-06,
+      "loss": 0.7593,
       "step": 27626
     },
     {
-      "epoch": 0.7587542226250309,
+      "epoch": 0.783967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 2.90037146281806e-06,
-      "loss": 0.8127,
+      "learning_rate": 2.3496080389557675e-06,
+      "loss": 0.8268,
       "step": 27627
     },
     {
-      "epoch": 0.7587816868528742,
+      "epoch": 0.7839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 2.8997450589272136e-06,
-      "loss": 0.8983,
+      "learning_rate": 2.3490161999569925e-06,
+      "loss": 0.8033,
       "step": 27628
     },
     {
-      "epoch": 0.7588091510807173,
+      "epoch": 0.7840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 2.8991187112163353e-06,
-      "loss": 0.8188,
+      "learning_rate": 2.348424425586422e-06,
+      "loss": 0.7753,
       "step": 27629
     },
     {
-      "epoch": 0.7588366153085606,
+      "epoch": 0.7840522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 2.898492419690383e-06,
-      "loss": 0.8363,
+      "learning_rate": 2.3478327158490578e-06,
+      "loss": 0.9019,
       "step": 27630
     },
     {
-      "epoch": 0.7588640795364038,
+      "epoch": 0.7840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 2.8978661843543067e-06,
-      "loss": 0.8754,
+      "learning_rate": 2.3472410707498917e-06,
+      "loss": 0.785,
       "step": 27631
     },
     {
-      "epoch": 0.7588915437642471,
+      "epoch": 0.7841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 2.8972400052130666e-06,
-      "loss": 0.8679,
+      "learning_rate": 2.346649490293924e-06,
+      "loss": 0.7982,
       "step": 27632
     },
     {
-      "epoch": 0.7589190079920903,
+      "epoch": 0.7841373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 2.8966138822716127e-06,
-      "loss": 0.7635,
+      "learning_rate": 2.3460579744861547e-06,
+      "loss": 0.9077,
       "step": 27633
     },
     {
-      "epoch": 0.7589464722199335,
+      "epoch": 0.7841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 2.8959878155349007e-06,
-      "loss": 0.7458,
+      "learning_rate": 2.3454665233315753e-06,
+      "loss": 0.8653,
       "step": 27634
     },
     {
-      "epoch": 0.7589739364477768,
+      "epoch": 0.7841940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 2.895361805007886e-06,
-      "loss": 0.985,
+      "learning_rate": 2.344875136835184e-06,
+      "loss": 0.7516,
       "step": 27635
     },
     {
-      "epoch": 0.75900140067562,
+      "epoch": 0.7842224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 2.89473585069552e-06,
-      "loss": 0.794,
+      "learning_rate": 2.34428381500198e-06,
+      "loss": 0.9759,
       "step": 27636
     },
     {
-      "epoch": 0.7590288649034632,
+      "epoch": 0.7842508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 2.8941099526027593e-06,
-      "loss": 0.8063,
+      "learning_rate": 2.343692557836951e-06,
+      "loss": 0.8254,
       "step": 27637
     },
     {
-      "epoch": 0.7590563291313065,
+      "epoch": 0.7842792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 2.89348411073455e-06,
-      "loss": 0.7917,
+      "learning_rate": 2.343101365345095e-06,
+      "loss": 0.8015,
       "step": 27638
     },
     {
-      "epoch": 0.7590837933591497,
+      "epoch": 0.7843076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 2.892858325095851e-06,
-      "loss": 0.7894,
+      "learning_rate": 2.3425102375314066e-06,
+      "loss": 0.7611,
       "step": 27639
     },
     {
-      "epoch": 0.7591112575869929,
+      "epoch": 0.7843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 2.8922325956916064e-06,
-      "loss": 0.7939,
+      "learning_rate": 2.3419191744008794e-06,
+      "loss": 0.7115,
       "step": 27640
     },
     {
-      "epoch": 0.7591387218148362,
+      "epoch": 0.7843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 2.8916069225267705e-06,
-      "loss": 0.8519,
+      "learning_rate": 2.3413281759585073e-06,
+      "loss": 0.7456,
       "step": 27641
     },
     {
-      "epoch": 0.7591661860426794,
+      "epoch": 0.7843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 2.8909813056062954e-06,
-      "loss": 0.8021,
+      "learning_rate": 2.3407372422092765e-06,
+      "loss": 0.8618,
       "step": 27642
     },
     {
-      "epoch": 0.7591936502705227,
+      "epoch": 0.7844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.8903557449351285e-06,
-      "loss": 0.8138,
+      "learning_rate": 2.3401463731581832e-06,
+      "loss": 0.7267,
       "step": 27643
     },
     {
-      "epoch": 0.7592211144983658,
+      "epoch": 0.7844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 2.889730240518225e-06,
-      "loss": 0.83,
+      "learning_rate": 2.339555568810221e-06,
+      "loss": 0.8365,
       "step": 27644
     },
     {
-      "epoch": 0.7592485787262091,
+      "epoch": 0.784477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 2.8891047923605265e-06,
-      "loss": 0.8103,
+      "learning_rate": 2.338964829170375e-06,
+      "loss": 0.8058,
       "step": 27645
     },
     {
-      "epoch": 0.7592760429540524,
+      "epoch": 0.7845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 2.888479400466988e-06,
-      "loss": 0.8299,
+      "learning_rate": 2.3383741542436354e-06,
+      "loss": 0.8665,
       "step": 27646
     },
     {
-      "epoch": 0.7593035071818955,
+      "epoch": 0.7845346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 2.8878540648425536e-06,
-      "loss": 0.7758,
+      "learning_rate": 2.3377835440349985e-06,
+      "loss": 0.8464,
       "step": 27647
     },
     {
-      "epoch": 0.7593309714097388,
+      "epoch": 0.7845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 2.887228785492172e-06,
-      "loss": 0.8198,
+      "learning_rate": 2.3371929985494445e-06,
+      "loss": 0.8506,
       "step": 27648
     },
     {
-      "epoch": 0.7593584356375821,
+      "epoch": 0.7845913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 2.886603562420791e-06,
-      "loss": 0.8984,
+      "learning_rate": 2.336602517791966e-06,
+      "loss": 0.8405,
       "step": 27649
     },
     {
-      "epoch": 0.7593858998654253,
+      "epoch": 0.7846197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 2.885978395633362e-06,
-      "loss": 0.8523,
+      "learning_rate": 2.336012101767554e-06,
+      "loss": 0.7728,
       "step": 27650
     },
     {
-      "epoch": 0.7594133640932685,
+      "epoch": 0.7846481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 2.885353285134824e-06,
-      "loss": 0.8009,
+      "learning_rate": 2.3354217504811893e-06,
+      "loss": 0.9036,
       "step": 27651
     },
     {
-      "epoch": 0.7594408283211117,
+      "epoch": 0.7846765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 2.884728230930127e-06,
-      "loss": 0.8223,
+      "learning_rate": 2.334831463937861e-06,
+      "loss": 0.78,
       "step": 27652
     },
     {
-      "epoch": 0.759468292548955,
+      "epoch": 0.7847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 2.8841032330242156e-06,
-      "loss": 0.7837,
+      "learning_rate": 2.3342412421425585e-06,
+      "loss": 0.85,
       "step": 27653
     },
     {
-      "epoch": 0.7594957567767983,
+      "epoch": 0.7847332576617481,
       "grad_norm": 0.0,
-      "learning_rate": 2.8834782914220407e-06,
-      "loss": 0.7309,
+      "learning_rate": 2.3336510851002604e-06,
+      "loss": 0.9516,
       "step": 27654
     },
     {
-      "epoch": 0.7595232210046414,
+      "epoch": 0.7847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 2.8828534061285374e-06,
-      "loss": 0.7593,
+      "learning_rate": 2.333060992815961e-06,
+      "loss": 0.8115,
       "step": 27655
     },
     {
-      "epoch": 0.7595506852324847,
+      "epoch": 0.7847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 2.8822285771486593e-06,
-      "loss": 0.8139,
+      "learning_rate": 2.3324709652946376e-06,
+      "loss": 0.8081,
       "step": 27656
     },
     {
-      "epoch": 0.759578149460328,
+      "epoch": 0.7848183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 2.8816038044873416e-06,
-      "loss": 0.9471,
+      "learning_rate": 2.3318810025412765e-06,
+      "loss": 0.7123,
       "step": 27657
     },
     {
-      "epoch": 0.7596056136881711,
+      "epoch": 0.7848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 2.8809790881495326e-06,
-      "loss": 0.8015,
+      "learning_rate": 2.3312911045608645e-06,
+      "loss": 0.7263,
       "step": 27658
     },
     {
-      "epoch": 0.7596330779160144,
+      "epoch": 0.7848751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 2.8803544281401753e-06,
-      "loss": 0.8218,
+      "learning_rate": 2.3307012713583776e-06,
+      "loss": 0.9203,
       "step": 27659
     },
     {
-      "epoch": 0.7596605421438576,
+      "epoch": 0.7849035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 2.8797298244642104e-06,
-      "loss": 0.8302,
+      "learning_rate": 2.330111502938802e-06,
+      "loss": 0.8109,
       "step": 27660
     },
     {
-      "epoch": 0.7596880063717009,
+      "epoch": 0.7849318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 2.879105277126585e-06,
-      "loss": 0.8405,
+      "learning_rate": 2.329521799307123e-06,
+      "loss": 0.8094,
       "step": 27661
     },
     {
-      "epoch": 0.7597154705995441,
+      "epoch": 0.7849602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 2.8784807861322337e-06,
-      "loss": 0.8038,
+      "learning_rate": 2.3289321604683135e-06,
+      "loss": 0.9184,
       "step": 27662
     },
     {
-      "epoch": 0.7597429348273873,
+      "epoch": 0.784988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 2.8778563514861048e-06,
-      "loss": 0.7145,
+      "learning_rate": 2.3283425864273602e-06,
+      "loss": 0.7679,
       "step": 27663
     },
     {
-      "epoch": 0.7597703990552306,
+      "epoch": 0.7850170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 2.877231973193132e-06,
-      "loss": 0.8936,
+      "learning_rate": 2.3277530771892453e-06,
+      "loss": 0.8082,
       "step": 27664
     },
     {
-      "epoch": 0.7597978632830737,
+      "epoch": 0.7850454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 2.8766076512582587e-06,
-      "loss": 0.7752,
+      "learning_rate": 2.327163632758941e-06,
+      "loss": 0.794,
       "step": 27665
     },
     {
-      "epoch": 0.759825327510917,
+      "epoch": 0.7850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 2.875983385686425e-06,
-      "loss": 0.8086,
+      "learning_rate": 2.3265742531414316e-06,
+      "loss": 0.8514,
       "step": 27666
     },
     {
-      "epoch": 0.7598527917387603,
+      "epoch": 0.7851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 2.8753591764825697e-06,
-      "loss": 0.8074,
+      "learning_rate": 2.3259849383416965e-06,
+      "loss": 0.9447,
       "step": 27667
     },
     {
-      "epoch": 0.7598802559666035,
+      "epoch": 0.7851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 2.874735023651636e-06,
-      "loss": 0.8602,
+      "learning_rate": 2.325395688364709e-06,
+      "loss": 0.8928,
       "step": 27668
     },
     {
-      "epoch": 0.7599077201944467,
+      "epoch": 0.7851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 2.8741109271985556e-06,
-      "loss": 0.8945,
+      "learning_rate": 2.324806503215449e-06,
+      "loss": 0.771,
       "step": 27669
     },
     {
-      "epoch": 0.75993518442229,
+      "epoch": 0.7851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 2.873486887128274e-06,
-      "loss": 0.8156,
+      "learning_rate": 2.3242173828988932e-06,
+      "loss": 0.8211,
       "step": 27670
     },
     {
-      "epoch": 0.7599626486501332,
+      "epoch": 0.7852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 2.87286290344572e-06,
-      "loss": 0.8189,
+      "learning_rate": 2.3236283274200187e-06,
+      "loss": 0.7669,
       "step": 27671
     },
     {
-      "epoch": 0.7599901128779765,
+      "epoch": 0.7852440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 2.872238976155837e-06,
-      "loss": 0.6889,
+      "learning_rate": 2.323039336783803e-06,
+      "loss": 0.7382,
       "step": 27672
     },
     {
-      "epoch": 0.7600175771058196,
+      "epoch": 0.785272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 2.871615105263562e-06,
-      "loss": 0.7216,
+      "learning_rate": 2.3224504109952172e-06,
+      "loss": 0.7458,
       "step": 27673
     },
     {
-      "epoch": 0.7600450413336629,
+      "epoch": 0.7853007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 2.8709912907738278e-06,
-      "loss": 0.8498,
+      "learning_rate": 2.321861550059238e-06,
+      "loss": 0.7977,
       "step": 27674
     },
     {
-      "epoch": 0.7600725055615062,
+      "epoch": 0.7853291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 2.8703675326915715e-06,
-      "loss": 0.8549,
+      "learning_rate": 2.321272753980841e-06,
+      "loss": 0.7846,
       "step": 27675
     },
     {
-      "epoch": 0.7600999697893494,
+      "epoch": 0.7853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 2.8697438310217286e-06,
-      "loss": 0.858,
+      "learning_rate": 2.3206840227649965e-06,
+      "loss": 0.8246,
       "step": 27676
     },
     {
-      "epoch": 0.7601274340171926,
+      "epoch": 0.7853859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 2.8691201857692343e-06,
-      "loss": 0.853,
+      "learning_rate": 2.3200953564166785e-06,
+      "loss": 0.8015,
       "step": 27677
     },
     {
-      "epoch": 0.7601548982450358,
+      "epoch": 0.7854143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 2.8684965969390276e-06,
-      "loss": 0.784,
+      "learning_rate": 2.3195067549408636e-06,
+      "loss": 0.8185,
       "step": 27678
     },
     {
-      "epoch": 0.7601823624728791,
+      "epoch": 0.7854426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 2.8678730645360374e-06,
-      "loss": 0.9064,
+      "learning_rate": 2.3189182183425174e-06,
+      "loss": 0.8904,
       "step": 27679
     },
     {
-      "epoch": 0.7602098267007223,
+      "epoch": 0.7854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 2.867249588565195e-06,
-      "loss": 0.9232,
+      "learning_rate": 2.3183297466266142e-06,
+      "loss": 0.8295,
       "step": 27680
     },
     {
-      "epoch": 0.7602372909285655,
+      "epoch": 0.78549943246311,
       "grad_norm": 0.0,
-      "learning_rate": 2.866626169031438e-06,
-      "loss": 0.7942,
+      "learning_rate": 2.3177413397981286e-06,
+      "loss": 0.831,
       "step": 27681
     },
     {
-      "epoch": 0.7602647551564088,
+      "epoch": 0.785527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 2.866002805939696e-06,
-      "loss": 0.869,
+      "learning_rate": 2.3171529978620234e-06,
+      "loss": 0.7941,
       "step": 27682
     },
     {
-      "epoch": 0.7602922193842521,
+      "epoch": 0.7855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 2.8653794992949047e-06,
-      "loss": 0.8785,
+      "learning_rate": 2.3165647208232723e-06,
+      "loss": 0.7909,
       "step": 27683
     },
     {
-      "epoch": 0.7603196836120952,
+      "epoch": 0.7855845629965947,
       "grad_norm": 0.0,
-      "learning_rate": 2.8647562491019943e-06,
-      "loss": 0.7624,
+      "learning_rate": 2.315976508686848e-06,
+      "loss": 0.9534,
       "step": 27684
     },
     {
-      "epoch": 0.7603471478399385,
+      "epoch": 0.7856129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 2.8641330553658996e-06,
-      "loss": 0.9255,
+      "learning_rate": 2.315388361457709e-06,
+      "loss": 0.809,
       "step": 27685
     },
     {
-      "epoch": 0.7603746120677817,
+      "epoch": 0.7856413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 2.8635099180915447e-06,
-      "loss": 0.9555,
+      "learning_rate": 2.3148002791408363e-06,
+      "loss": 0.7557,
       "step": 27686
     },
     {
-      "epoch": 0.760402076295625,
+      "epoch": 0.7856696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 2.862886837283867e-06,
-      "loss": 0.8206,
+      "learning_rate": 2.314212261741188e-06,
+      "loss": 0.7527,
       "step": 27687
     },
     {
-      "epoch": 0.7604295405234682,
+      "epoch": 0.7856980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 2.862263812947791e-06,
-      "loss": 0.7895,
+      "learning_rate": 2.3136243092637345e-06,
+      "loss": 0.8401,
       "step": 27688
     },
     {
-      "epoch": 0.7604570047513114,
+      "epoch": 0.7857264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 2.861640845088247e-06,
-      "loss": 0.7269,
+      "learning_rate": 2.3130364217134436e-06,
+      "loss": 0.8043,
       "step": 27689
     },
     {
-      "epoch": 0.7604844689791547,
+      "epoch": 0.7857548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 2.861017933710167e-06,
-      "loss": 0.8265,
+      "learning_rate": 2.3124485990952784e-06,
+      "loss": 0.8286,
       "step": 27690
     },
     {
-      "epoch": 0.7605119332069978,
+      "epoch": 0.785783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 2.860395078818481e-06,
-      "loss": 0.6732,
+      "learning_rate": 2.311860841414204e-06,
+      "loss": 0.6899,
       "step": 27691
     },
     {
-      "epoch": 0.7605393974348411,
+      "epoch": 0.785811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 2.8597722804181128e-06,
-      "loss": 0.8045,
+      "learning_rate": 2.3112731486751905e-06,
+      "loss": 0.8272,
       "step": 27692
     },
     {
-      "epoch": 0.7605668616626844,
+      "epoch": 0.7858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 2.8591495385139912e-06,
-      "loss": 0.7844,
+      "learning_rate": 2.310685520883196e-06,
+      "loss": 0.7574,
       "step": 27693
     },
     {
-      "epoch": 0.7605943258905276,
+      "epoch": 0.7858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.858526853111048e-06,
-      "loss": 0.9191,
+      "learning_rate": 2.310097958043185e-06,
+      "loss": 0.8363,
       "step": 27694
     },
     {
-      "epoch": 0.7606217901183708,
+      "epoch": 0.7858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 2.857904224214203e-06,
-      "loss": 0.8713,
+      "learning_rate": 2.309510460160126e-06,
+      "loss": 0.8835,
       "step": 27695
     },
     {
-      "epoch": 0.7606492543462141,
+      "epoch": 0.7859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 2.8572816518283866e-06,
-      "loss": 0.7811,
+      "learning_rate": 2.308923027238975e-06,
+      "loss": 0.9025,
       "step": 27696
     },
     {
-      "epoch": 0.7606767185740573,
+      "epoch": 0.7859534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 2.8566591359585273e-06,
-      "loss": 0.8305,
+      "learning_rate": 2.3083356592846974e-06,
+      "loss": 0.7514,
       "step": 27697
     },
     {
-      "epoch": 0.7607041828019006,
+      "epoch": 0.7859818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 2.856036676609546e-06,
-      "loss": 0.8568,
+      "learning_rate": 2.307748356302256e-06,
+      "loss": 0.9778,
       "step": 27698
     },
     {
-      "epoch": 0.7607316470297437,
+      "epoch": 0.7860102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 2.8554142737863687e-06,
-      "loss": 0.8411,
+      "learning_rate": 2.307161118296608e-06,
+      "loss": 0.8808,
       "step": 27699
     },
     {
-      "epoch": 0.760759111257587,
+      "epoch": 0.786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 2.8547919274939217e-06,
-      "loss": 0.77,
+      "learning_rate": 2.306573945272713e-06,
+      "loss": 0.8946,
       "step": 27700
     },
     {
-      "epoch": 0.7607865754854303,
+      "epoch": 0.7860669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 2.8541696377371287e-06,
-      "loss": 0.8217,
+      "learning_rate": 2.3059868372355387e-06,
+      "loss": 0.872,
       "step": 27701
     },
     {
-      "epoch": 0.7608140397132734,
+      "epoch": 0.7860953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 2.8535474045209167e-06,
-      "loss": 0.8243,
+      "learning_rate": 2.3053997941900376e-06,
+      "loss": 0.8522,
       "step": 27702
     },
     {
-      "epoch": 0.7608415039411167,
+      "epoch": 0.7861237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 2.8529252278502063e-06,
-      "loss": 0.8298,
+      "learning_rate": 2.3048128161411697e-06,
+      "loss": 0.724,
       "step": 27703
     },
     {
-      "epoch": 0.7608689681689599,
+      "epoch": 0.7861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 2.852303107729917e-06,
-      "loss": 0.867,
+      "learning_rate": 2.304225903093896e-06,
+      "loss": 0.8282,
       "step": 27704
     },
     {
-      "epoch": 0.7608964323968032,
+      "epoch": 0.7861804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 2.8516810441649732e-06,
-      "loss": 0.8008,
+      "learning_rate": 2.3036390550531705e-06,
+      "loss": 0.8484,
       "step": 27705
     },
     {
-      "epoch": 0.7609238966246464,
+      "epoch": 0.7862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 2.8510590371602985e-06,
-      "loss": 0.884,
+      "learning_rate": 2.3030522720239546e-06,
+      "loss": 0.7764,
       "step": 27706
     },
     {
-      "epoch": 0.7609513608524896,
+      "epoch": 0.7862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 2.8504370867208144e-06,
-      "loss": 0.7469,
+      "learning_rate": 2.3024655540111984e-06,
+      "loss": 0.8518,
       "step": 27707
     },
     {
-      "epoch": 0.7609788250803329,
+      "epoch": 0.7862656072644721,
       "grad_norm": 0.0,
-      "learning_rate": 2.8498151928514408e-06,
-      "loss": 0.7576,
+      "learning_rate": 2.301878901019863e-06,
+      "loss": 0.756,
       "step": 27708
     },
     {
-      "epoch": 0.7610062893081762,
+      "epoch": 0.7862939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 2.8491933555571028e-06,
-      "loss": 0.8331,
+      "learning_rate": 2.3012923130549037e-06,
+      "loss": 0.8118,
       "step": 27709
     },
     {
-      "epoch": 0.7610337535360193,
+      "epoch": 0.786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 2.8485715748427132e-06,
-      "loss": 0.9287,
+      "learning_rate": 2.3007057901212726e-06,
+      "loss": 0.881,
       "step": 27710
     },
     {
-      "epoch": 0.7610612177638626,
+      "epoch": 0.7863507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 2.8479498507131997e-06,
-      "loss": 0.8278,
+      "learning_rate": 2.3001193322239255e-06,
+      "loss": 0.8183,
       "step": 27711
     },
     {
-      "epoch": 0.7610886819917058,
+      "epoch": 0.7863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 2.8473281831734733e-06,
-      "loss": 0.9374,
+      "learning_rate": 2.29953293936782e-06,
+      "loss": 0.6979,
       "step": 27712
     },
     {
-      "epoch": 0.761116146219549,
+      "epoch": 0.7864074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 2.846706572228457e-06,
-      "loss": 0.8792,
+      "learning_rate": 2.298946611557903e-06,
+      "loss": 0.9088,
       "step": 27713
     },
     {
-      "epoch": 0.7611436104473923,
+      "epoch": 0.7864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 2.8460850178830703e-06,
-      "loss": 0.7547,
+      "learning_rate": 2.298360348799129e-06,
+      "loss": 0.8038,
       "step": 27714
     },
     {
-      "epoch": 0.7611710746752355,
+      "epoch": 0.7864642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 2.845463520142232e-06,
-      "loss": 0.7873,
+      "learning_rate": 2.2977741510964523e-06,
+      "loss": 0.7709,
       "step": 27715
     },
     {
-      "epoch": 0.7611985389030788,
+      "epoch": 0.7864926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 2.8448420790108557e-06,
-      "loss": 0.8712,
+      "learning_rate": 2.2971880184548234e-06,
+      "loss": 0.8907,
       "step": 27716
     },
     {
-      "epoch": 0.7612260031309219,
+      "epoch": 0.7865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 2.844220694493859e-06,
-      "loss": 0.9418,
+      "learning_rate": 2.2966019508791938e-06,
+      "loss": 0.8371,
       "step": 27717
     },
     {
-      "epoch": 0.7612534673587652,
+      "epoch": 0.7865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 2.843599366596165e-06,
-      "loss": 0.7253,
+      "learning_rate": 2.296015948374516e-06,
+      "loss": 0.7518,
       "step": 27718
     },
     {
-      "epoch": 0.7612809315866085,
+      "epoch": 0.786577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.8429780953226806e-06,
-      "loss": 0.8319,
+      "learning_rate": 2.2954300109457363e-06,
+      "loss": 0.8381,
       "step": 27719
     },
     {
-      "epoch": 0.7613083958144516,
+      "epoch": 0.7866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 2.842356880678325e-06,
-      "loss": 0.7513,
+      "learning_rate": 2.294844138597804e-06,
+      "loss": 0.7982,
       "step": 27720
     },
     {
-      "epoch": 0.7613358600422949,
+      "epoch": 0.7866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 2.8417357226680198e-06,
-      "loss": 0.7677,
+      "learning_rate": 2.2942583313356736e-06,
+      "loss": 0.772,
       "step": 27721
     },
     {
-      "epoch": 0.7613633242701382,
+      "epoch": 0.7866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 2.84111462129667e-06,
-      "loss": 0.7471,
+      "learning_rate": 2.2936725891642864e-06,
+      "loss": 0.7742,
       "step": 27722
     },
     {
-      "epoch": 0.7613907884979814,
+      "epoch": 0.7866912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 2.8404935765691944e-06,
-      "loss": 0.835,
+      "learning_rate": 2.2930869120885925e-06,
+      "loss": 0.7008,
       "step": 27723
     },
     {
-      "epoch": 0.7614182527258246,
+      "epoch": 0.7867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 2.8398725884905064e-06,
-      "loss": 0.8744,
+      "learning_rate": 2.292501300113543e-06,
+      "loss": 0.7397,
       "step": 27724
     },
     {
-      "epoch": 0.7614457169536678,
+      "epoch": 0.7867480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 2.8392516570655213e-06,
-      "loss": 0.8852,
+      "learning_rate": 2.2919157532440796e-06,
+      "loss": 0.8661,
       "step": 27725
     },
     {
-      "epoch": 0.7614731811815111,
+      "epoch": 0.7867763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 2.8386307822991533e-06,
-      "loss": 0.8103,
+      "learning_rate": 2.2913302714851493e-06,
+      "loss": 0.7402,
       "step": 27726
     },
     {
-      "epoch": 0.7615006454093544,
+      "epoch": 0.7868047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 2.838009964196312e-06,
-      "loss": 0.8585,
+      "learning_rate": 2.2907448548417024e-06,
+      "loss": 0.8121,
       "step": 27727
     },
     {
-      "epoch": 0.7615281096371975,
+      "epoch": 0.78683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.837389202761909e-06,
-      "loss": 0.8001,
+      "learning_rate": 2.2901595033186765e-06,
+      "loss": 0.7519,
       "step": 27728
     },
     {
-      "epoch": 0.7615555738650408,
+      "epoch": 0.786861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 2.836768498000856e-06,
-      "loss": 0.7741,
+      "learning_rate": 2.289574216921021e-06,
+      "loss": 0.8021,
       "step": 27729
     },
     {
-      "epoch": 0.761583038092884,
+      "epoch": 0.7868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 2.836147849918065e-06,
-      "loss": 0.7981,
+      "learning_rate": 2.2889889956536803e-06,
+      "loss": 0.8408,
       "step": 27730
     },
     {
-      "epoch": 0.7616105023207272,
+      "epoch": 0.7869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 2.8355272585184466e-06,
-      "loss": 0.8632,
+      "learning_rate": 2.2884038395215915e-06,
+      "loss": 0.7938,
       "step": 27731
     },
     {
-      "epoch": 0.7616379665485705,
+      "epoch": 0.7869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 2.8349067238069138e-06,
-      "loss": 0.8932,
+      "learning_rate": 2.2878187485297075e-06,
+      "loss": 0.8346,
       "step": 27732
     },
     {
-      "epoch": 0.7616654307764137,
+      "epoch": 0.7869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 2.8342862457883756e-06,
-      "loss": 0.7913,
+      "learning_rate": 2.2872337226829623e-06,
+      "loss": 0.7786,
       "step": 27733
     },
     {
-      "epoch": 0.761692895004257,
+      "epoch": 0.7870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 2.833665824467742e-06,
-      "loss": 0.8341,
+      "learning_rate": 2.2866487619863e-06,
+      "loss": 0.8314,
       "step": 27734
     },
     {
-      "epoch": 0.7617203592321002,
+      "epoch": 0.7870317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 2.8330454598499167e-06,
-      "loss": 0.718,
+      "learning_rate": 2.2860638664446665e-06,
+      "loss": 0.8943,
       "step": 27735
     },
     {
-      "epoch": 0.7617478234599434,
+      "epoch": 0.7870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 2.8324251519398115e-06,
-      "loss": 0.8013,
+      "learning_rate": 2.285479036062994e-06,
+      "loss": 0.8012,
       "step": 27736
     },
     {
-      "epoch": 0.7617752876877867,
+      "epoch": 0.7870885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 2.8318049007423356e-06,
-      "loss": 0.7496,
+      "learning_rate": 2.2848942708462275e-06,
+      "loss": 0.8954,
       "step": 27737
     },
     {
-      "epoch": 0.7618027519156298,
+      "epoch": 0.7871169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 2.831184706262399e-06,
-      "loss": 0.819,
+      "learning_rate": 2.284309570799309e-06,
+      "loss": 0.7519,
       "step": 27738
     },
     {
-      "epoch": 0.7618302161434731,
+      "epoch": 0.7871452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 2.8305645685049022e-06,
-      "loss": 0.8449,
+      "learning_rate": 2.2837249359271718e-06,
+      "loss": 0.7789,
       "step": 27739
     },
     {
-      "epoch": 0.7618576803713164,
+      "epoch": 0.7871736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 2.829944487474756e-06,
+      "learning_rate": 2.2831403662347563e-06,
       "loss": 0.8394,
       "step": 27740
     },
     {
-      "epoch": 0.7618851445991596,
+      "epoch": 0.7872020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 2.829324463176867e-06,
-      "loss": 0.7603,
+      "learning_rate": 2.282555861727004e-06,
+      "loss": 0.7812,
       "step": 27741
     },
     {
-      "epoch": 0.7619126088270028,
+      "epoch": 0.7872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 2.8287044956161424e-06,
-      "loss": 0.6887,
+      "learning_rate": 2.281971422408846e-06,
+      "loss": 0.9066,
       "step": 27742
     },
     {
-      "epoch": 0.761940073054846,
+      "epoch": 0.7872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 2.8280845847974837e-06,
-      "loss": 0.7876,
+      "learning_rate": 2.281387048285223e-06,
+      "loss": 0.8159,
       "step": 27743
     },
     {
-      "epoch": 0.7619675372826893,
+      "epoch": 0.7872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 2.8274647307258007e-06,
-      "loss": 0.8584,
+      "learning_rate": 2.2808027393610733e-06,
+      "loss": 0.8792,
       "step": 27744
     },
     {
-      "epoch": 0.7619950015105326,
+      "epoch": 0.7873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 2.8268449334059922e-06,
-      "loss": 0.7842,
+      "learning_rate": 2.2802184956413277e-06,
+      "loss": 0.8209,
       "step": 27745
     },
     {
-      "epoch": 0.7620224657383757,
+      "epoch": 0.7873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 2.826225192842965e-06,
-      "loss": 0.8427,
+      "learning_rate": 2.279634317130922e-06,
+      "loss": 0.7006,
       "step": 27746
     },
     {
-      "epoch": 0.762049929966219,
+      "epoch": 0.787372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 2.8256055090416224e-06,
-      "loss": 0.7896,
+      "learning_rate": 2.279050203834794e-06,
+      "loss": 0.8237,
       "step": 27747
     },
     {
-      "epoch": 0.7620773941940623,
+      "epoch": 0.7874006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 2.824985882006869e-06,
-      "loss": 0.8276,
+      "learning_rate": 2.2784661557578747e-06,
+      "loss": 0.7712,
       "step": 27748
     },
     {
-      "epoch": 0.7621048584219055,
+      "epoch": 0.7874290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 2.824366311743606e-06,
-      "loss": 0.8992,
+      "learning_rate": 2.2778821729051017e-06,
+      "loss": 0.8431,
       "step": 27749
     },
     {
-      "epoch": 0.7621323226497487,
+      "epoch": 0.7874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 2.8237467982567402e-06,
-      "loss": 0.795,
+      "learning_rate": 2.277298255281403e-06,
+      "loss": 0.8457,
       "step": 27750
     },
     {
-      "epoch": 0.7621597868775919,
+      "epoch": 0.7874858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 2.8231273415511682e-06,
-      "loss": 0.9674,
+      "learning_rate": 2.276714402891713e-06,
+      "loss": 0.8241,
       "step": 27751
     },
     {
-      "epoch": 0.7621872511054352,
+      "epoch": 0.7875141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 2.8225079416317913e-06,
-      "loss": 0.9125,
+      "learning_rate": 2.2761306157409657e-06,
+      "loss": 0.9194,
       "step": 27752
     },
     {
-      "epoch": 0.7622147153332784,
+      "epoch": 0.7875425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 2.8218885985035107e-06,
-      "loss": 0.8404,
+      "learning_rate": 2.2755468938340884e-06,
+      "loss": 0.8593,
       "step": 27753
     },
     {
-      "epoch": 0.7622421795611216,
+      "epoch": 0.7875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 2.8212693121712286e-06,
-      "loss": 0.7599,
+      "learning_rate": 2.274963237176013e-06,
+      "loss": 0.7116,
       "step": 27754
     },
     {
-      "epoch": 0.7622696437889649,
+      "epoch": 0.7875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 2.8206500826398444e-06,
-      "loss": 0.8583,
+      "learning_rate": 2.2743796457716736e-06,
+      "loss": 0.813,
       "step": 27755
     },
     {
-      "epoch": 0.7622971080168081,
+      "epoch": 0.7876276958002271,
       "grad_norm": 0.0,
-      "learning_rate": 2.8200309099142597e-06,
-      "loss": 0.7975,
+      "learning_rate": 2.273796119625994e-06,
+      "loss": 0.7659,
       "step": 27756
     },
     {
-      "epoch": 0.7623245722446513,
+      "epoch": 0.7876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 2.8194117939993705e-06,
-      "loss": 0.7551,
+      "learning_rate": 2.273212658743905e-06,
+      "loss": 0.8613,
       "step": 27757
     },
     {
-      "epoch": 0.7623520364724946,
+      "epoch": 0.7876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.818792734900079e-06,
-      "loss": 0.7922,
+      "learning_rate": 2.27262926313034e-06,
+      "loss": 0.8402,
       "step": 27758
     },
     {
-      "epoch": 0.7623795007003378,
+      "epoch": 0.7877128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 2.818173732621278e-06,
-      "loss": 0.8084,
+      "learning_rate": 2.27204593279022e-06,
+      "loss": 0.8266,
       "step": 27759
     },
     {
-      "epoch": 0.762406964928181,
+      "epoch": 0.7877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 2.817554787167868e-06,
-      "loss": 0.8365,
+      "learning_rate": 2.2714626677284746e-06,
+      "loss": 0.785,
       "step": 27760
     },
     {
-      "epoch": 0.7624344291560243,
+      "epoch": 0.7877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 2.816935898544747e-06,
-      "loss": 0.7526,
+      "learning_rate": 2.2708794679500325e-06,
+      "loss": 0.7951,
       "step": 27761
     },
     {
-      "epoch": 0.7624618933838675,
+      "epoch": 0.7877979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 2.816317066756814e-06,
-      "loss": 0.8372,
+      "learning_rate": 2.2702963334598184e-06,
+      "loss": 0.7919,
       "step": 27762
     },
     {
-      "epoch": 0.7624893576117108,
+      "epoch": 0.7878263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 2.8156982918089616e-06,
-      "loss": 0.7598,
+      "learning_rate": 2.269713264262762e-06,
+      "loss": 0.9326,
       "step": 27763
     },
     {
-      "epoch": 0.7625168218395539,
+      "epoch": 0.7878547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 2.8150795737060853e-06,
-      "loss": 0.9178,
+      "learning_rate": 2.269130260363781e-06,
+      "loss": 0.756,
       "step": 27764
     },
     {
-      "epoch": 0.7625442860673972,
+      "epoch": 0.787883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 2.8144609124530843e-06,
-      "loss": 0.8759,
+      "learning_rate": 2.2685473217678057e-06,
+      "loss": 0.839,
       "step": 27765
     },
     {
-      "epoch": 0.7625717502952405,
+      "epoch": 0.7879114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 2.8138423080548537e-06,
-      "loss": 0.8289,
+      "learning_rate": 2.2679644484797602e-06,
+      "loss": 0.8671,
       "step": 27766
     },
     {
-      "epoch": 0.7625992145230837,
+      "epoch": 0.7879398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 2.813223760516285e-06,
-      "loss": 0.8065,
+      "learning_rate": 2.2673816405045644e-06,
+      "loss": 0.7638,
       "step": 27767
     },
     {
-      "epoch": 0.7626266787509269,
+      "epoch": 0.7879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 2.8126052698422755e-06,
-      "loss": 0.7376,
+      "learning_rate": 2.2667988978471432e-06,
+      "loss": 0.7873,
       "step": 27768
     },
     {
-      "epoch": 0.7626541429787701,
+      "epoch": 0.7879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 2.8119868360377146e-06,
-      "loss": 0.9191,
+      "learning_rate": 2.2662162205124227e-06,
+      "loss": 0.7356,
       "step": 27769
     },
     {
-      "epoch": 0.7626816072066134,
+      "epoch": 0.7880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 2.8113684591074986e-06,
-      "loss": 0.8957,
+      "learning_rate": 2.2656336085053186e-06,
+      "loss": 0.8861,
       "step": 27770
     },
     {
-      "epoch": 0.7627090714344567,
+      "epoch": 0.7880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.8107501390565195e-06,
-      "loss": 0.8946,
+      "learning_rate": 2.2650510618307555e-06,
+      "loss": 0.8637,
       "step": 27771
     },
     {
-      "epoch": 0.7627365356622998,
+      "epoch": 0.7880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 2.81013187588967e-06,
-      "loss": 0.7457,
+      "learning_rate": 2.264468580493655e-06,
+      "loss": 0.8405,
       "step": 27772
     },
     {
-      "epoch": 0.7627639998901431,
+      "epoch": 0.7881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 2.8095136696118452e-06,
-      "loss": 0.8688,
+      "learning_rate": 2.2638861644989353e-06,
+      "loss": 0.7426,
       "step": 27773
     },
     {
-      "epoch": 0.7627914641179864,
+      "epoch": 0.7881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 2.8088955202279312e-06,
-      "loss": 0.798,
+      "learning_rate": 2.263303813851515e-06,
+      "loss": 0.7687,
       "step": 27774
     },
     {
-      "epoch": 0.7628189283458295,
+      "epoch": 0.78816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.8082774277428236e-06,
-      "loss": 0.8426,
+      "learning_rate": 2.26272152855632e-06,
+      "loss": 0.9241,
       "step": 27775
     },
     {
-      "epoch": 0.7628463925736728,
+      "epoch": 0.7881952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 2.807659392161408e-06,
-      "loss": 0.8674,
+      "learning_rate": 2.2621393086182597e-06,
+      "loss": 0.7956,
       "step": 27776
     },
     {
-      "epoch": 0.762873856801516,
+      "epoch": 0.7882236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 2.807041413488578e-06,
-      "loss": 0.7668,
+      "learning_rate": 2.2615571540422567e-06,
+      "loss": 0.8425,
       "step": 27777
     },
     {
-      "epoch": 0.7629013210293593,
+      "epoch": 0.7882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 2.8064234917292222e-06,
-      "loss": 0.8117,
+      "learning_rate": 2.2609750648332286e-06,
+      "loss": 0.8536,
       "step": 27778
     },
     {
-      "epoch": 0.7629287852572025,
+      "epoch": 0.7882803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 2.8058056268882294e-06,
-      "loss": 0.8046,
+      "learning_rate": 2.2603930409960927e-06,
+      "loss": 0.8129,
       "step": 27779
     },
     {
-      "epoch": 0.7629562494850457,
+      "epoch": 0.7883087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 2.805187818970494e-06,
-      "loss": 0.8076,
+      "learning_rate": 2.2598110825357667e-06,
+      "loss": 0.7499,
       "step": 27780
     },
     {
-      "epoch": 0.762983713712889,
+      "epoch": 0.7883371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 2.8045700679808952e-06,
-      "loss": 0.8268,
+      "learning_rate": 2.2592291894571617e-06,
+      "loss": 0.7492,
       "step": 27781
     },
     {
-      "epoch": 0.7630111779407321,
+      "epoch": 0.7883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 2.803952373924329e-06,
-      "loss": 0.7012,
+      "learning_rate": 2.258647361765196e-06,
+      "loss": 0.8865,
       "step": 27782
     },
     {
-      "epoch": 0.7630386421685754,
+      "epoch": 0.788393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 2.803334736805675e-06,
-      "loss": 0.8271,
+      "learning_rate": 2.258065599464786e-06,
+      "loss": 0.7702,
       "step": 27783
     },
     {
-      "epoch": 0.7630661063964187,
+      "epoch": 0.788422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.802717156629825e-06,
-      "loss": 0.8963,
+      "learning_rate": 2.257483902560842e-06,
+      "loss": 0.9125,
       "step": 27784
     },
     {
-      "epoch": 0.7630935706242619,
+      "epoch": 0.7884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 2.802099633401666e-06,
-      "loss": 0.8006,
+      "learning_rate": 2.256902271058279e-06,
+      "loss": 0.8761,
       "step": 27785
     },
     {
-      "epoch": 0.7631210348521051,
+      "epoch": 0.7884790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 2.801482167126084e-06,
-      "loss": 0.761,
+      "learning_rate": 2.256320704962014e-06,
+      "loss": 0.9013,
       "step": 27786
     },
     {
-      "epoch": 0.7631484990799484,
+      "epoch": 0.7885073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 2.8008647578079615e-06,
-      "loss": 0.7881,
+      "learning_rate": 2.255739204276952e-06,
+      "loss": 0.9221,
       "step": 27787
     },
     {
-      "epoch": 0.7631759633077916,
+      "epoch": 0.7885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 2.8002474054521854e-06,
-      "loss": 0.9339,
+      "learning_rate": 2.255157769008011e-06,
+      "loss": 0.8748,
       "step": 27788
     },
     {
-      "epoch": 0.7632034275356349,
+      "epoch": 0.7885641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 2.7996301100636414e-06,
-      "loss": 0.8473,
+      "learning_rate": 2.2545763991601024e-06,
+      "loss": 0.7984,
       "step": 27789
     },
     {
-      "epoch": 0.763230891763478,
+      "epoch": 0.7885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 2.7990128716472153e-06,
-      "loss": 0.8606,
+      "learning_rate": 2.2539950947381318e-06,
+      "loss": 0.7965,
       "step": 27790
     },
     {
-      "epoch": 0.7632583559913213,
+      "epoch": 0.7886208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 2.7983956902077857e-06,
-      "loss": 0.7571,
+      "learning_rate": 2.253413855747013e-06,
+      "loss": 0.8426,
       "step": 27791
     },
     {
-      "epoch": 0.7632858202191646,
+      "epoch": 0.7886492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 2.797778565750243e-06,
-      "loss": 0.9166,
+      "learning_rate": 2.2528326821916556e-06,
+      "loss": 0.9473,
       "step": 27792
     },
     {
-      "epoch": 0.7633132844470077,
+      "epoch": 0.788677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 2.7971614982794624e-06,
-      "loss": 0.8779,
+      "learning_rate": 2.252251574076969e-06,
+      "loss": 0.7852,
       "step": 27793
     },
     {
-      "epoch": 0.763340748674851,
+      "epoch": 0.788706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 2.79654448780033e-06,
-      "loss": 0.7873,
+      "learning_rate": 2.2516705314078645e-06,
+      "loss": 0.7882,
       "step": 27794
     },
     {
-      "epoch": 0.7633682129026942,
+      "epoch": 0.7887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 2.7959275343177283e-06,
-      "loss": 0.7667,
+      "learning_rate": 2.2510895541892453e-06,
+      "loss": 0.7788,
       "step": 27795
     },
     {
-      "epoch": 0.7633956771305375,
+      "epoch": 0.7887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.795310637836539e-06,
-      "loss": 0.8937,
+      "learning_rate": 2.2505086424260204e-06,
+      "loss": 0.8721,
       "step": 27796
     },
     {
-      "epoch": 0.7634231413583807,
+      "epoch": 0.7887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 2.7946937983616452e-06,
-      "loss": 0.8261,
+      "learning_rate": 2.249927796123099e-06,
+      "loss": 0.814,
       "step": 27797
     },
     {
-      "epoch": 0.7634506055862239,
+      "epoch": 0.7888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 2.7940770158979225e-06,
-      "loss": 0.787,
+      "learning_rate": 2.2493470152853846e-06,
+      "loss": 0.864,
       "step": 27798
     },
     {
-      "epoch": 0.7634780698140672,
+      "epoch": 0.7888479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 2.7934602904502562e-06,
-      "loss": 0.6921,
+      "learning_rate": 2.248766299917784e-06,
+      "loss": 0.7656,
       "step": 27799
     },
     {
-      "epoch": 0.7635055340419105,
+      "epoch": 0.7888762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 2.792843622023521e-06,
-      "loss": 0.7729,
+      "learning_rate": 2.2481856500252053e-06,
+      "loss": 0.8556,
       "step": 27800
     },
     {
-      "epoch": 0.7635329982697536,
+      "epoch": 0.7889046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 2.792227010622598e-06,
-      "loss": 0.8438,
+      "learning_rate": 2.247605065612548e-06,
+      "loss": 0.8001,
       "step": 27801
     },
     {
-      "epoch": 0.7635604624975969,
+      "epoch": 0.788933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 2.7916104562523683e-06,
-      "loss": 0.8033,
+      "learning_rate": 2.2470245466847196e-06,
+      "loss": 0.8455,
       "step": 27802
     },
     {
-      "epoch": 0.7635879267254401,
+      "epoch": 0.7889614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 2.7909939589177114e-06,
-      "loss": 0.7184,
+      "learning_rate": 2.2464440932466267e-06,
+      "loss": 0.7606,
       "step": 27803
     },
     {
-      "epoch": 0.7636153909532833,
+      "epoch": 0.7889897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 2.7903775186234992e-06,
-      "loss": 0.8257,
+      "learning_rate": 2.2458637053031653e-06,
+      "loss": 0.7805,
       "step": 27804
     },
     {
-      "epoch": 0.7636428551811266,
+      "epoch": 0.7890181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 2.7897611353746136e-06,
-      "loss": 0.8504,
+      "learning_rate": 2.2452833828592425e-06,
+      "loss": 0.898,
       "step": 27805
     },
     {
-      "epoch": 0.7636703194089698,
+      "epoch": 0.7890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 2.7891448091759333e-06,
-      "loss": 0.8765,
+      "learning_rate": 2.2447031259197614e-06,
+      "loss": 0.8407,
       "step": 27806
     },
     {
-      "epoch": 0.7636977836368131,
+      "epoch": 0.7890749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 2.7885285400323302e-06,
-      "loss": 0.7694,
+      "learning_rate": 2.2441229344896175e-06,
+      "loss": 0.9026,
       "step": 27807
     },
     {
-      "epoch": 0.7637252478646562,
+      "epoch": 0.7891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 2.787912327948682e-06,
-      "loss": 0.7968,
+      "learning_rate": 2.2435428085737175e-06,
+      "loss": 0.8928,
       "step": 27808
     },
     {
-      "epoch": 0.7637527120924995,
+      "epoch": 0.7891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.7872961729298686e-06,
-      "loss": 0.8846,
+      "learning_rate": 2.2429627481769633e-06,
+      "loss": 0.7342,
       "step": 27809
     },
     {
-      "epoch": 0.7637801763203428,
+      "epoch": 0.7891600454029511,
       "grad_norm": 0.0,
-      "learning_rate": 2.7866800749807586e-06,
-      "loss": 0.8597,
+      "learning_rate": 2.2423827533042486e-06,
+      "loss": 0.7493,
       "step": 27810
     },
     {
-      "epoch": 0.763807640548186,
+      "epoch": 0.7891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 2.7860640341062307e-06,
-      "loss": 0.8613,
+      "learning_rate": 2.241802823960475e-06,
+      "loss": 0.9021,
       "step": 27811
     },
     {
-      "epoch": 0.7638351047760292,
+      "epoch": 0.789216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 2.7854480503111568e-06,
-      "loss": 0.7733,
+      "learning_rate": 2.2412229601505454e-06,
+      "loss": 0.8722,
       "step": 27812
     },
     {
-      "epoch": 0.7638625690038725,
+      "epoch": 0.7892451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 2.784832123600414e-06,
-      "loss": 0.8134,
+      "learning_rate": 2.24064316187935e-06,
+      "loss": 0.75,
       "step": 27813
     },
     {
-      "epoch": 0.7638900332317157,
+      "epoch": 0.7892735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 2.7842162539788774e-06,
-      "loss": 0.8811,
+      "learning_rate": 2.2400634291517952e-06,
+      "loss": 0.7601,
       "step": 27814
     },
     {
-      "epoch": 0.763917497459559,
+      "epoch": 0.7893019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 2.7836004414514175e-06,
-      "loss": 0.7446,
+      "learning_rate": 2.2394837619727692e-06,
+      "loss": 0.8101,
       "step": 27815
     },
     {
-      "epoch": 0.7639449616874021,
+      "epoch": 0.7893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 2.7829846860229017e-06,
-      "loss": 0.8189,
+      "learning_rate": 2.238904160347174e-06,
+      "loss": 0.8352,
       "step": 27816
     },
     {
-      "epoch": 0.7639724259152454,
+      "epoch": 0.7893586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 2.7823689876982072e-06,
-      "loss": 0.7277,
+      "learning_rate": 2.238324624279906e-06,
+      "loss": 0.8302,
       "step": 27817
     },
     {
-      "epoch": 0.7639998901430887,
+      "epoch": 0.7893870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 2.7817533464822054e-06,
-      "loss": 0.7284,
+      "learning_rate": 2.2377451537758567e-06,
+      "loss": 0.7888,
       "step": 27818
     },
     {
-      "epoch": 0.7640273543709318,
+      "epoch": 0.7894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 2.7811377623797663e-06,
-      "loss": 0.7692,
+      "learning_rate": 2.2371657488399223e-06,
+      "loss": 0.7602,
       "step": 27819
     },
     {
-      "epoch": 0.7640548185987751,
+      "epoch": 0.7894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 2.7805222353957628e-06,
-      "loss": 0.7779,
+      "learning_rate": 2.2365864094770007e-06,
+      "loss": 0.8483,
       "step": 27820
     },
     {
-      "epoch": 0.7640822828266183,
+      "epoch": 0.789472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 2.7799067655350666e-06,
-      "loss": 0.752,
+      "learning_rate": 2.236007135691981e-06,
+      "loss": 0.8457,
       "step": 27821
     },
     {
-      "epoch": 0.7641097470544616,
+      "epoch": 0.7895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 2.779291352802541e-06,
-      "loss": 0.8392,
+      "learning_rate": 2.2354279274897574e-06,
+      "loss": 0.7822,
       "step": 27822
     },
     {
-      "epoch": 0.7641372112823048,
+      "epoch": 0.7895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 2.778675997203064e-06,
-      "loss": 0.7939,
+      "learning_rate": 2.234848784875222e-06,
+      "loss": 0.8543,
       "step": 27823
     },
     {
-      "epoch": 0.764164675510148,
+      "epoch": 0.7895573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 2.778060698741496e-06,
-      "loss": 0.8204,
+      "learning_rate": 2.2342697078532695e-06,
+      "loss": 0.6795,
       "step": 27824
     },
     {
-      "epoch": 0.7641921397379913,
+      "epoch": 0.7895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 2.7774454574227083e-06,
-      "loss": 0.7683,
+      "learning_rate": 2.2336906964287897e-06,
+      "loss": 0.7699,
       "step": 27825
     },
     {
-      "epoch": 0.7642196039658345,
+      "epoch": 0.7896140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 2.7768302732515707e-06,
-      "loss": 0.7938,
+      "learning_rate": 2.233111750606676e-06,
+      "loss": 0.8643,
       "step": 27826
     },
     {
-      "epoch": 0.7642470681936777,
+      "epoch": 0.7896424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 2.7762151462329534e-06,
-      "loss": 0.8293,
+      "learning_rate": 2.2325328703918135e-06,
+      "loss": 0.7017,
       "step": 27827
     },
     {
-      "epoch": 0.764274532421521,
+      "epoch": 0.7896708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 2.775600076371716e-06,
-      "loss": 0.8938,
+      "learning_rate": 2.231954055789095e-06,
+      "loss": 0.864,
       "step": 27828
     },
     {
-      "epoch": 0.7643019966493642,
+      "epoch": 0.7896992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 2.7749850636727295e-06,
-      "loss": 0.9542,
+      "learning_rate": 2.231375306803414e-06,
+      "loss": 0.7552,
       "step": 27829
     },
     {
-      "epoch": 0.7643294608772074,
+      "epoch": 0.789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 2.7743701081408638e-06,
-      "loss": 0.8624,
+      "learning_rate": 2.2307966234396504e-06,
+      "loss": 0.9255,
       "step": 27830
     },
     {
-      "epoch": 0.7643569251050507,
+      "epoch": 0.789755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 2.773755209780977e-06,
-      "loss": 0.8358,
+      "learning_rate": 2.230218005702698e-06,
+      "loss": 0.7682,
       "step": 27831
     },
     {
-      "epoch": 0.7643843893328939,
+      "epoch": 0.7897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 2.7731403685979384e-06,
-      "loss": 0.7983,
+      "learning_rate": 2.2296394535974455e-06,
+      "loss": 0.7489,
       "step": 27832
     },
     {
-      "epoch": 0.7644118535607372,
+      "epoch": 0.7898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 2.772525584596616e-06,
-      "loss": 0.8204,
+      "learning_rate": 2.2290609671287756e-06,
+      "loss": 0.8476,
       "step": 27833
     },
     {
-      "epoch": 0.7644393177885803,
+      "epoch": 0.7898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.7719108577818664e-06,
-      "loss": 0.8567,
+      "learning_rate": 2.2284825463015767e-06,
+      "loss": 0.704,
       "step": 27834
     },
     {
-      "epoch": 0.7644667820164236,
+      "epoch": 0.7898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 2.771296188158559e-06,
-      "loss": 0.7701,
+      "learning_rate": 2.2279041911207377e-06,
+      "loss": 0.7423,
       "step": 27835
     },
     {
-      "epoch": 0.7644942462442669,
+      "epoch": 0.7898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 2.7706815757315563e-06,
-      "loss": 0.8325,
+      "learning_rate": 2.2273259015911397e-06,
+      "loss": 0.7254,
       "step": 27836
     },
     {
-      "epoch": 0.76452171047211,
+      "epoch": 0.7899262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 2.770067020505721e-06,
-      "loss": 0.664,
+      "learning_rate": 2.2267476777176676e-06,
+      "loss": 0.8473,
       "step": 27837
     },
     {
-      "epoch": 0.7645491746999533,
+      "epoch": 0.7899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 2.7694525224859194e-06,
-      "loss": 0.7973,
+      "learning_rate": 2.226169519505209e-06,
+      "loss": 0.7789,
       "step": 27838
     },
     {
-      "epoch": 0.7645766389277966,
+      "epoch": 0.7899829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 2.768838081677011e-06,
-      "loss": 0.8207,
+      "learning_rate": 2.2255914269586456e-06,
+      "loss": 0.8038,
       "step": 27839
     },
     {
-      "epoch": 0.7646041031556398,
+      "epoch": 0.7900113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 2.7682236980838527e-06,
-      "loss": 0.927,
+      "learning_rate": 2.225013400082863e-06,
+      "loss": 0.8704,
       "step": 27840
     },
     {
-      "epoch": 0.764631567383483,
+      "epoch": 0.7900397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 2.7676093717113097e-06,
-      "loss": 0.8631,
+      "learning_rate": 2.224435438882738e-06,
+      "loss": 0.8531,
       "step": 27841
     },
     {
-      "epoch": 0.7646590316113262,
+      "epoch": 0.7900681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 2.766995102564244e-06,
-      "loss": 0.7557,
+      "learning_rate": 2.223857543363158e-06,
+      "loss": 0.872,
       "step": 27842
     },
     {
-      "epoch": 0.7646864958391695,
+      "epoch": 0.7900964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 2.766380890647514e-06,
-      "loss": 0.8386,
+      "learning_rate": 2.223279713529005e-06,
+      "loss": 0.7883,
       "step": 27843
     },
     {
-      "epoch": 0.7647139600670128,
+      "epoch": 0.7901248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 2.76576673596598e-06,
-      "loss": 0.7801,
+      "learning_rate": 2.2227019493851553e-06,
+      "loss": 0.8343,
       "step": 27844
     },
     {
-      "epoch": 0.7647414242948559,
+      "epoch": 0.7901532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 2.7651526385245054e-06,
-      "loss": 0.8607,
+      "learning_rate": 2.2221242509364914e-06,
+      "loss": 0.774,
       "step": 27845
     },
     {
-      "epoch": 0.7647688885226992,
+      "epoch": 0.7901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 2.7645385983279428e-06,
-      "loss": 0.8251,
+      "learning_rate": 2.2215466181878975e-06,
+      "loss": 0.8678,
       "step": 27846
     },
     {
-      "epoch": 0.7647963527505424,
+      "epoch": 0.7902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 2.763924615381157e-06,
-      "loss": 0.8186,
+      "learning_rate": 2.220969051144245e-06,
+      "loss": 0.644,
       "step": 27847
     },
     {
-      "epoch": 0.7648238169783856,
+      "epoch": 0.7902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 2.7633106896889995e-06,
-      "loss": 0.8539,
+      "learning_rate": 2.220391549810418e-06,
+      "loss": 0.8313,
       "step": 27848
     },
     {
-      "epoch": 0.7648512812062289,
+      "epoch": 0.790266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 2.76269682125633e-06,
-      "loss": 0.7804,
+      "learning_rate": 2.2198141141912944e-06,
+      "loss": 0.8638,
       "step": 27849
     },
     {
-      "epoch": 0.7648787454340721,
+      "epoch": 0.7902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 2.762083010088007e-06,
-      "loss": 0.8285,
+      "learning_rate": 2.2192367442917484e-06,
+      "loss": 0.864,
       "step": 27850
     },
     {
-      "epoch": 0.7649062096619154,
+      "epoch": 0.7903234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 2.76146925618889e-06,
-      "loss": 0.8186,
+      "learning_rate": 2.2186594401166604e-06,
+      "loss": 0.8184,
       "step": 27851
     },
     {
-      "epoch": 0.7649336738897586,
+      "epoch": 0.7903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 2.76085555956383e-06,
-      "loss": 0.7592,
+      "learning_rate": 2.2180822016709082e-06,
+      "loss": 0.8188,
       "step": 27852
     },
     {
-      "epoch": 0.7649611381176018,
+      "epoch": 0.7903802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 2.7602419202176844e-06,
-      "loss": 0.8333,
+      "learning_rate": 2.2175050289593592e-06,
+      "loss": 0.8783,
       "step": 27853
     },
     {
-      "epoch": 0.7649886023454451,
+      "epoch": 0.7904086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 2.7596283381553123e-06,
-      "loss": 0.8445,
+      "learning_rate": 2.2169279219869012e-06,
+      "loss": 0.8094,
       "step": 27854
     },
     {
-      "epoch": 0.7650160665732882,
+      "epoch": 0.7904370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 2.759014813381562e-06,
-      "loss": 0.9339,
+      "learning_rate": 2.2163508807584e-06,
+      "loss": 0.8708,
       "step": 27855
     },
     {
-      "epoch": 0.7650435308011315,
+      "epoch": 0.7904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 2.758401345901295e-06,
-      "loss": 0.7494,
+      "learning_rate": 2.215773905278733e-06,
+      "loss": 0.698,
       "step": 27856
     },
     {
-      "epoch": 0.7650709950289748,
+      "epoch": 0.7904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 2.757787935719358e-06,
-      "loss": 0.8714,
+      "learning_rate": 2.2151969955527764e-06,
+      "loss": 0.7957,
       "step": 27857
     },
     {
-      "epoch": 0.765098459256818,
+      "epoch": 0.790522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.7571745828406094e-06,
-      "loss": 0.8676,
+      "learning_rate": 2.214620151585397e-06,
+      "loss": 0.8725,
       "step": 27858
     },
     {
-      "epoch": 0.7651259234846612,
+      "epoch": 0.7905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 2.7565612872698997e-06,
-      "loss": 0.8759,
+      "learning_rate": 2.2140433733814727e-06,
+      "loss": 0.8019,
       "step": 27859
     },
     {
-      "epoch": 0.7651533877125045,
+      "epoch": 0.7905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.7559480490120836e-06,
-      "loss": 0.8761,
+      "learning_rate": 2.2134666609458764e-06,
+      "loss": 0.8522,
       "step": 27860
     },
     {
-      "epoch": 0.7651808519403477,
+      "epoch": 0.7906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 2.7553348680720126e-06,
-      "loss": 0.8729,
+      "learning_rate": 2.212890014283474e-06,
+      "loss": 0.7788,
       "step": 27861
     },
     {
-      "epoch": 0.765208316168191,
+      "epoch": 0.7906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 2.7547217444545413e-06,
-      "loss": 0.959,
+      "learning_rate": 2.2123134333991405e-06,
+      "loss": 0.7804,
       "step": 27862
     },
     {
-      "epoch": 0.7652357803960341,
+      "epoch": 0.7906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 2.7541086781645187e-06,
-      "loss": 0.7861,
+      "learning_rate": 2.2117369182977476e-06,
+      "loss": 0.7431,
       "step": 27863
     },
     {
-      "epoch": 0.7652632446238774,
+      "epoch": 0.7906923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 2.7534956692067915e-06,
-      "loss": 0.8416,
+      "learning_rate": 2.2111604689841602e-06,
+      "loss": 0.7342,
       "step": 27864
     },
     {
-      "epoch": 0.7652907088517207,
+      "epoch": 0.7907207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 2.752882717586214e-06,
-      "loss": 0.9175,
+      "learning_rate": 2.210584085463251e-06,
+      "loss": 0.8754,
       "step": 27865
     },
     {
-      "epoch": 0.7653181730795638,
+      "epoch": 0.7907491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 2.752269823307635e-06,
-      "loss": 0.8186,
+      "learning_rate": 2.21000776773989e-06,
+      "loss": 0.806,
       "step": 27866
     },
     {
-      "epoch": 0.7653456373074071,
+      "epoch": 0.79077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.7516569863759046e-06,
-      "loss": 0.7599,
+      "learning_rate": 2.2094315158189416e-06,
+      "loss": 0.8622,
       "step": 27867
     },
     {
-      "epoch": 0.7653731015352503,
+      "epoch": 0.7908059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 2.751044206795875e-06,
-      "loss": 0.873,
+      "learning_rate": 2.208855329705275e-06,
+      "loss": 0.8165,
       "step": 27868
     },
     {
-      "epoch": 0.7654005657630936,
+      "epoch": 0.7908342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 2.750431484572389e-06,
-      "loss": 0.917,
+      "learning_rate": 2.2082792094037585e-06,
+      "loss": 0.8546,
       "step": 27869
     },
     {
-      "epoch": 0.7654280299909368,
+      "epoch": 0.7908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 2.749818819710297e-06,
-      "loss": 0.8196,
+      "learning_rate": 2.207703154919257e-06,
+      "loss": 0.7801,
       "step": 27870
     },
     {
-      "epoch": 0.76545549421878,
+      "epoch": 0.7908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 2.7492062122144505e-06,
-      "loss": 0.9092,
+      "learning_rate": 2.2071271662566396e-06,
+      "loss": 0.7816,
       "step": 27871
     },
     {
-      "epoch": 0.7654829584466233,
+      "epoch": 0.7909194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 2.74859366208969e-06,
-      "loss": 0.765,
+      "learning_rate": 2.2065512434207683e-06,
+      "loss": 0.8424,
       "step": 27872
     },
     {
-      "epoch": 0.7655104226744666,
+      "epoch": 0.7909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 2.7479811693408655e-06,
-      "loss": 0.8641,
+      "learning_rate": 2.205975386416507e-06,
+      "loss": 0.7793,
       "step": 27873
     },
     {
-      "epoch": 0.7655378869023097,
+      "epoch": 0.7909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 2.747368733972825e-06,
-      "loss": 0.7144,
+      "learning_rate": 2.205399595248726e-06,
+      "loss": 0.8113,
       "step": 27874
     },
     {
-      "epoch": 0.765565351130153,
+      "epoch": 0.7910045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 2.746756355990411e-06,
-      "loss": 0.7933,
+      "learning_rate": 2.2048238699222814e-06,
+      "loss": 0.7973,
       "step": 27875
     },
     {
-      "epoch": 0.7655928153579962,
+      "epoch": 0.7910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 2.7461440353984693e-06,
-      "loss": 0.8106,
+      "learning_rate": 2.2042482104420404e-06,
+      "loss": 0.7593,
       "step": 27876
     },
     {
-      "epoch": 0.7656202795858394,
+      "epoch": 0.791061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 2.745531772201846e-06,
-      "loss": 0.841,
+      "learning_rate": 2.2036726168128687e-06,
+      "loss": 0.7167,
       "step": 27877
     },
     {
-      "epoch": 0.7656477438136827,
+      "epoch": 0.7910896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 2.7449195664053885e-06,
-      "loss": 0.7819,
+      "learning_rate": 2.203097089039621e-06,
+      "loss": 0.8314,
       "step": 27878
     },
     {
-      "epoch": 0.7656752080415259,
+      "epoch": 0.7911180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 2.744307418013935e-06,
-      "loss": 0.8651,
+      "learning_rate": 2.2025216271271632e-06,
+      "loss": 0.7186,
       "step": 27879
     },
     {
-      "epoch": 0.7657026722693692,
+      "epoch": 0.7911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 2.743695327032334e-06,
-      "loss": 0.8612,
+      "learning_rate": 2.2019462310803586e-06,
+      "loss": 0.7602,
       "step": 27880
     },
     {
-      "epoch": 0.7657301364972123,
+      "epoch": 0.7911748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 2.743083293465425e-06,
-      "loss": 0.7845,
+      "learning_rate": 2.2013709009040617e-06,
+      "loss": 0.7185,
       "step": 27881
     },
     {
-      "epoch": 0.7657576007250556,
+      "epoch": 0.7912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 2.7424713173180507e-06,
-      "loss": 0.8085,
+      "learning_rate": 2.2007956366031358e-06,
+      "loss": 0.7602,
       "step": 27882
     },
     {
-      "epoch": 0.7657850649528989,
+      "epoch": 0.7912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 2.741859398595055e-06,
-      "loss": 0.768,
+      "learning_rate": 2.200220438182442e-06,
+      "loss": 0.6907,
       "step": 27883
     },
     {
-      "epoch": 0.765812529180742,
+      "epoch": 0.7912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 2.741247537301279e-06,
-      "loss": 0.818,
+      "learning_rate": 2.1996453056468316e-06,
+      "loss": 0.752,
       "step": 27884
     },
     {
-      "epoch": 0.7658399934085853,
+      "epoch": 0.7912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 2.7406357334415646e-06,
-      "loss": 0.8567,
+      "learning_rate": 2.199070239001173e-06,
+      "loss": 0.8402,
       "step": 27885
     },
     {
-      "epoch": 0.7658674576364286,
+      "epoch": 0.791316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.7400239870207556e-06,
-      "loss": 0.8521,
+      "learning_rate": 2.1984952382503165e-06,
+      "loss": 0.8687,
       "step": 27886
     },
     {
-      "epoch": 0.7658949218642718,
+      "epoch": 0.7913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 2.7394122980436886e-06,
-      "loss": 0.7591,
+      "learning_rate": 2.1979203033991205e-06,
+      "loss": 0.8783,
       "step": 27887
     },
     {
-      "epoch": 0.765922386092115,
+      "epoch": 0.7913734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 2.738800666515201e-06,
-      "loss": 0.7042,
+      "learning_rate": 2.197345434452446e-06,
+      "loss": 0.8953,
       "step": 27888
     },
     {
-      "epoch": 0.7659498503199582,
+      "epoch": 0.7914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 2.7381890924401354e-06,
-      "loss": 0.8213,
+      "learning_rate": 2.196770631415143e-06,
+      "loss": 0.7441,
       "step": 27889
     },
     {
-      "epoch": 0.7659773145478015,
+      "epoch": 0.7914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 2.7375775758233303e-06,
-      "loss": 0.7443,
+      "learning_rate": 2.196195894292068e-06,
+      "loss": 0.9143,
       "step": 27890
     },
     {
-      "epoch": 0.7660047787756448,
+      "epoch": 0.7914585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 2.736966116669625e-06,
-      "loss": 0.8782,
+      "learning_rate": 2.1956212230880815e-06,
+      "loss": 0.8761,
       "step": 27891
     },
     {
-      "epoch": 0.7660322430034879,
+      "epoch": 0.7914869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 2.7363547149838597e-06,
-      "loss": 0.8188,
+      "learning_rate": 2.19504661780803e-06,
+      "loss": 0.7731,
       "step": 27892
     },
     {
-      "epoch": 0.7660597072313312,
+      "epoch": 0.7915153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 2.735743370770868e-06,
-      "loss": 0.8237,
+      "learning_rate": 2.1944720784567698e-06,
+      "loss": 0.9314,
       "step": 27893
     },
     {
-      "epoch": 0.7660871714591744,
+      "epoch": 0.7915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 2.73513208403549e-06,
-      "loss": 0.8446,
+      "learning_rate": 2.193897605039158e-06,
+      "loss": 0.8818,
       "step": 27894
     },
     {
-      "epoch": 0.7661146356870177,
+      "epoch": 0.791572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.7345208547825584e-06,
-      "loss": 0.8348,
+      "learning_rate": 2.1933231975600412e-06,
+      "loss": 0.8509,
       "step": 27895
     },
     {
-      "epoch": 0.7661420999148609,
+      "epoch": 0.791600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 2.733909683016913e-06,
-      "loss": 0.8729,
+      "learning_rate": 2.192748856024275e-06,
+      "loss": 0.7965,
       "step": 27896
     },
     {
-      "epoch": 0.7661695641427041,
+      "epoch": 0.7916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 2.7332985687433888e-06,
-      "loss": 0.8072,
+      "learning_rate": 2.1921745804367125e-06,
+      "loss": 0.8141,
       "step": 27897
     },
     {
-      "epoch": 0.7661970283705474,
+      "epoch": 0.7916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.7326875119668227e-06,
-      "loss": 0.7918,
+      "learning_rate": 2.1916003708021993e-06,
+      "loss": 0.6858,
       "step": 27898
     },
     {
-      "epoch": 0.7662244925983906,
+      "epoch": 0.7916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 2.7320765126920467e-06,
-      "loss": 0.8174,
+      "learning_rate": 2.191026227125589e-06,
+      "loss": 0.8831,
       "step": 27899
     },
     {
-      "epoch": 0.7662519568262338,
+      "epoch": 0.7917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 2.731465570923896e-06,
-      "loss": 0.8508,
+      "learning_rate": 2.1904521494117304e-06,
+      "loss": 0.7945,
       "step": 27900
     },
     {
-      "epoch": 0.7662794210540771,
+      "epoch": 0.7917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 2.7308546866672046e-06,
-      "loss": 0.8566,
+      "learning_rate": 2.189878137665474e-06,
+      "loss": 0.7465,
       "step": 27901
     },
     {
-      "epoch": 0.7663068852819203,
+      "epoch": 0.7917707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 2.7302438599268113e-06,
-      "loss": 0.8246,
+      "learning_rate": 2.189304191891671e-06,
+      "loss": 0.9045,
       "step": 27902
     },
     {
-      "epoch": 0.7663343495097635,
+      "epoch": 0.7917990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 2.7296330907075407e-06,
-      "loss": 0.7623,
+      "learning_rate": 2.188730312095165e-06,
+      "loss": 0.8739,
       "step": 27903
     },
     {
-      "epoch": 0.7663618137376068,
+      "epoch": 0.7918274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 2.7290223790142323e-06,
-      "loss": 0.8477,
+      "learning_rate": 2.1881564982808033e-06,
+      "loss": 0.8546,
       "step": 27904
     },
     {
-      "epoch": 0.76638927796545,
+      "epoch": 0.7918558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 2.7284117248517117e-06,
-      "loss": 0.8067,
+      "learning_rate": 2.1875827504534387e-06,
+      "loss": 0.8496,
       "step": 27905
     },
     {
-      "epoch": 0.7664167421932933,
+      "epoch": 0.7918842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 2.727801128224815e-06,
-      "loss": 0.8283,
+      "learning_rate": 2.1870090686179112e-06,
+      "loss": 0.7439,
       "step": 27906
     },
     {
-      "epoch": 0.7664442064211364,
+      "epoch": 0.7919125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 2.7271905891383734e-06,
-      "loss": 0.7135,
+      "learning_rate": 2.186435452779069e-06,
+      "loss": 0.7932,
       "step": 27907
     },
     {
-      "epoch": 0.7664716706489797,
+      "epoch": 0.7919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 2.7265801075972165e-06,
-      "loss": 0.7275,
+      "learning_rate": 2.1858619029417606e-06,
+      "loss": 0.7209,
       "step": 27908
     },
     {
-      "epoch": 0.766499134876823,
+      "epoch": 0.7919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 2.7259696836061745e-06,
-      "loss": 0.7783,
+      "learning_rate": 2.1852884191108246e-06,
+      "loss": 0.8166,
       "step": 27909
     },
     {
-      "epoch": 0.7665265991046661,
+      "epoch": 0.7919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 2.725359317170082e-06,
-      "loss": 0.7917,
+      "learning_rate": 2.184715001291109e-06,
+      "loss": 0.8239,
       "step": 27910
     },
     {
-      "epoch": 0.7665540633325094,
+      "epoch": 0.7920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 2.7247490082937644e-06,
-      "loss": 0.8591,
+      "learning_rate": 2.18414164948746e-06,
+      "loss": 0.807,
       "step": 27911
     },
     {
-      "epoch": 0.7665815275603527,
+      "epoch": 0.7920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 2.7241387569820477e-06,
-      "loss": 0.7999,
+      "learning_rate": 2.1835683637047156e-06,
+      "loss": 0.7823,
       "step": 27912
     },
     {
-      "epoch": 0.7666089917881959,
+      "epoch": 0.7920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 2.7235285632397635e-06,
-      "loss": 0.8342,
+      "learning_rate": 2.1829951439477193e-06,
+      "loss": 0.7821,
       "step": 27913
     },
     {
-      "epoch": 0.7666364560160391,
+      "epoch": 0.79211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.72291842707174e-06,
-      "loss": 0.7488,
+      "learning_rate": 2.1824219902213183e-06,
+      "loss": 0.8123,
       "step": 27914
     },
     {
-      "epoch": 0.7666639202438823,
+      "epoch": 0.7921396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 2.722308348482807e-06,
-      "loss": 0.8232,
+      "learning_rate": 2.181848902530345e-06,
+      "loss": 0.8134,
       "step": 27915
     },
     {
-      "epoch": 0.7666913844717256,
+      "epoch": 0.7921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 2.7216983274777877e-06,
-      "loss": 0.7698,
+      "learning_rate": 2.181275880879646e-06,
+      "loss": 0.7359,
       "step": 27916
     },
     {
-      "epoch": 0.7667188486995689,
+      "epoch": 0.7921963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 2.7210883640615105e-06,
-      "loss": 0.8363,
+      "learning_rate": 2.1807029252740665e-06,
+      "loss": 0.7766,
       "step": 27917
     },
     {
-      "epoch": 0.766746312927412,
+      "epoch": 0.7922247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 2.720478458238803e-06,
-      "loss": 0.7842,
+      "learning_rate": 2.1801300357184363e-06,
+      "loss": 0.7926,
       "step": 27918
     },
     {
-      "epoch": 0.7667737771552553,
+      "epoch": 0.7922531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 2.7198686100144888e-06,
-      "loss": 0.7885,
+      "learning_rate": 2.1795572122176003e-06,
+      "loss": 0.8169,
       "step": 27919
     },
     {
-      "epoch": 0.7668012413830985,
+      "epoch": 0.7922814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 2.719258819393393e-06,
-      "loss": 0.8288,
+      "learning_rate": 2.1789844547763984e-06,
+      "loss": 0.8545,
       "step": 27920
     },
     {
-      "epoch": 0.7668287056109417,
+      "epoch": 0.7923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 2.7186490863803462e-06,
-      "loss": 0.8531,
+      "learning_rate": 2.1784117633996638e-06,
+      "loss": 0.8611,
       "step": 27921
     },
     {
-      "epoch": 0.766856169838785,
+      "epoch": 0.7923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 2.7180394109801633e-06,
-      "loss": 0.8379,
+      "learning_rate": 2.1778391380922394e-06,
+      "loss": 1.0455,
       "step": 27922
     },
     {
-      "epoch": 0.7668836340666282,
+      "epoch": 0.792366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.7174297931976746e-06,
-      "loss": 0.862,
+      "learning_rate": 2.1772665788589566e-06,
+      "loss": 0.822,
       "step": 27923
     },
     {
-      "epoch": 0.7669110982944715,
+      "epoch": 0.7923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 2.7168202330377023e-06,
-      "loss": 0.8668,
+      "learning_rate": 2.176694085704654e-06,
+      "loss": 0.8355,
       "step": 27924
     },
     {
-      "epoch": 0.7669385625223147,
+      "epoch": 0.7924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 2.716210730505069e-06,
-      "loss": 0.7683,
+      "learning_rate": 2.1761216586341717e-06,
+      "loss": 0.8295,
       "step": 27925
     },
     {
-      "epoch": 0.7669660267501579,
+      "epoch": 0.7924517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 2.715601285604602e-06,
-      "loss": 0.7939,
+      "learning_rate": 2.1755492976523384e-06,
+      "loss": 0.844,
       "step": 27926
     },
     {
-      "epoch": 0.7669934909780012,
+      "epoch": 0.7924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 2.7149918983411183e-06,
-      "loss": 0.9644,
+      "learning_rate": 2.1749770027639917e-06,
+      "loss": 0.8235,
       "step": 27927
     },
     {
-      "epoch": 0.7670209552058443,
+      "epoch": 0.7925085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 2.714382568719438e-06,
-      "loss": 0.8262,
+      "learning_rate": 2.1744047739739695e-06,
+      "loss": 0.8059,
       "step": 27928
     },
     {
-      "epoch": 0.7670484194336876,
+      "epoch": 0.7925368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 2.713773296744385e-06,
-      "loss": 0.7226,
+      "learning_rate": 2.173832611287099e-06,
+      "loss": 0.826,
       "step": 27929
     },
     {
-      "epoch": 0.7670758836615309,
+      "epoch": 0.7925652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 2.7131640824207794e-06,
-      "loss": 0.8331,
+      "learning_rate": 2.1732605147082133e-06,
+      "loss": 0.8327,
       "step": 27930
     },
     {
-      "epoch": 0.7671033478893741,
+      "epoch": 0.7925936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 2.7125549257534434e-06,
-      "loss": 0.9025,
+      "learning_rate": 2.1726884842421524e-06,
+      "loss": 0.7752,
       "step": 27931
     },
     {
-      "epoch": 0.7671308121172173,
+      "epoch": 0.792622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 2.7119458267471955e-06,
-      "loss": 0.7365,
+      "learning_rate": 2.172116519893742e-06,
+      "loss": 0.897,
       "step": 27932
     },
     {
-      "epoch": 0.7671582763450605,
+      "epoch": 0.792650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 2.711336785406857e-06,
-      "loss": 0.8988,
+      "learning_rate": 2.171544621667816e-06,
+      "loss": 0.812,
       "step": 27933
     },
     {
-      "epoch": 0.7671857405729038,
+      "epoch": 0.7926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 2.710727801737244e-06,
-      "loss": 0.7862,
+      "learning_rate": 2.1709727895692056e-06,
+      "loss": 0.858,
       "step": 27934
     },
     {
-      "epoch": 0.7672132048007471,
+      "epoch": 0.7927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 2.7101188757431783e-06,
-      "loss": 0.8315,
+      "learning_rate": 2.1704010236027385e-06,
+      "loss": 0.7684,
       "step": 27935
     },
     {
-      "epoch": 0.7672406690285902,
+      "epoch": 0.7927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 2.7095100074294733e-06,
-      "loss": 0.8786,
+      "learning_rate": 2.169829323773246e-06,
+      "loss": 0.8952,
       "step": 27936
     },
     {
-      "epoch": 0.7672681332564335,
+      "epoch": 0.7927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 2.708901196800948e-06,
-      "loss": 0.8211,
+      "learning_rate": 2.169257690085559e-06,
+      "loss": 0.7898,
       "step": 27937
     },
     {
-      "epoch": 0.7672955974842768,
+      "epoch": 0.7927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 2.7082924438624225e-06,
-      "loss": 0.9128,
+      "learning_rate": 2.1686861225445024e-06,
+      "loss": 0.827,
       "step": 27938
     },
     {
-      "epoch": 0.76732306171212,
+      "epoch": 0.7928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 2.7076837486187134e-06,
-      "loss": 0.6716,
+      "learning_rate": 2.1681146211549054e-06,
+      "loss": 0.8034,
       "step": 27939
     },
     {
-      "epoch": 0.7673505259399632,
+      "epoch": 0.7928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 2.707075111074633e-06,
-      "loss": 0.7405,
+      "learning_rate": 2.167543185921599e-06,
+      "loss": 0.8806,
       "step": 27940
     },
     {
-      "epoch": 0.7673779901678064,
+      "epoch": 0.7928774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 2.706466531234999e-06,
-      "loss": 0.7727,
+      "learning_rate": 2.166971816849406e-06,
+      "loss": 0.9624,
       "step": 27941
     },
     {
-      "epoch": 0.7674054543956497,
+      "epoch": 0.7929057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 2.7058580091046304e-06,
-      "loss": 0.8535,
+      "learning_rate": 2.166400513943152e-06,
+      "loss": 0.7841,
       "step": 27942
     },
     {
-      "epoch": 0.7674329186234929,
+      "epoch": 0.7929341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 2.7052495446883355e-06,
-      "loss": 0.7568,
+      "learning_rate": 2.1658292772076695e-06,
+      "loss": 0.8958,
       "step": 27943
     },
     {
-      "epoch": 0.7674603828513361,
+      "epoch": 0.7929625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 2.704641137990931e-06,
-      "loss": 0.7679,
+      "learning_rate": 2.165258106647776e-06,
+      "loss": 0.7919,
       "step": 27944
     },
     {
-      "epoch": 0.7674878470791794,
+      "epoch": 0.7929909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 2.704032789017236e-06,
-      "loss": 0.7674,
+      "learning_rate": 2.1646870022682997e-06,
+      "loss": 0.8881,
       "step": 27945
     },
     {
-      "epoch": 0.7675153113070226,
+      "epoch": 0.7930192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 2.7034244977720558e-06,
-      "loss": 0.8004,
+      "learning_rate": 2.1641159640740638e-06,
+      "loss": 0.7542,
       "step": 27946
     },
     {
-      "epoch": 0.7675427755348658,
+      "epoch": 0.7930476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 2.7028162642602084e-06,
-      "loss": 0.7426,
+      "learning_rate": 2.1635449920698936e-06,
+      "loss": 0.8547,
       "step": 27947
     },
     {
-      "epoch": 0.7675702397627091,
+      "epoch": 0.7930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 2.7022080884865055e-06,
-      "loss": 0.856,
+      "learning_rate": 2.162974086260613e-06,
+      "loss": 0.8508,
       "step": 27948
     },
     {
-      "epoch": 0.7675977039905523,
+      "epoch": 0.7931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.701599970455758e-06,
-      "loss": 0.8244,
+      "learning_rate": 2.1624032466510405e-06,
+      "loss": 0.7796,
       "step": 27949
     },
     {
-      "epoch": 0.7676251682183955,
+      "epoch": 0.7931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 2.700991910172782e-06,
-      "loss": 0.8093,
+      "learning_rate": 2.1618324732459993e-06,
+      "loss": 0.834,
       "step": 27950
     },
     {
-      "epoch": 0.7676526324462388,
+      "epoch": 0.793161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 2.7003839076423853e-06,
-      "loss": 0.8002,
+      "learning_rate": 2.1612617660503154e-06,
+      "loss": 0.7683,
       "step": 27951
     },
     {
-      "epoch": 0.767680096674082,
+      "epoch": 0.7931895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 2.6997759628693766e-06,
-      "loss": 0.9266,
+      "learning_rate": 2.160691125068802e-06,
+      "loss": 0.7383,
       "step": 27952
     },
     {
-      "epoch": 0.7677075609019253,
+      "epoch": 0.7932179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 2.699168075858568e-06,
-      "loss": 0.8586,
+      "learning_rate": 2.160120550306284e-06,
+      "loss": 0.8544,
       "step": 27953
     },
     {
-      "epoch": 0.7677350251297684,
+      "epoch": 0.7932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 2.6985602466147688e-06,
-      "loss": 0.6983,
+      "learning_rate": 2.159550041767582e-06,
+      "loss": 0.874,
       "step": 27954
     },
     {
-      "epoch": 0.7677624893576117,
+      "epoch": 0.7932746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 2.6979524751427898e-06,
-      "loss": 0.8125,
+      "learning_rate": 2.1589795994575102e-06,
+      "loss": 0.8377,
       "step": 27955
     },
     {
-      "epoch": 0.767789953585455,
+      "epoch": 0.7933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 2.6973447614474393e-06,
-      "loss": 0.7899,
+      "learning_rate": 2.1584092233808904e-06,
+      "loss": 0.7005,
       "step": 27956
     },
     {
-      "epoch": 0.7678174178132982,
+      "epoch": 0.7933314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 2.696737105533529e-06,
-      "loss": 0.7755,
+      "learning_rate": 2.157838913542545e-06,
+      "loss": 0.875,
       "step": 27957
     },
     {
-      "epoch": 0.7678448820411414,
+      "epoch": 0.7933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 2.6961295074058604e-06,
-      "loss": 0.7963,
+      "learning_rate": 2.157268669947282e-06,
+      "loss": 0.8059,
       "step": 27958
     },
     {
-      "epoch": 0.7678723462689846,
+      "epoch": 0.7933881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 2.695521967069248e-06,
-      "loss": 0.7555,
+      "learning_rate": 2.1566984925999234e-06,
+      "loss": 0.8279,
       "step": 27959
     },
     {
-      "epoch": 0.7678998104968279,
+      "epoch": 0.793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 2.6949144845284912e-06,
-      "loss": 0.8096,
+      "learning_rate": 2.156128381505289e-06,
+      "loss": 0.7502,
       "step": 27960
     },
     {
-      "epoch": 0.7679272747246711,
+      "epoch": 0.7934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 2.6943070597884027e-06,
-      "loss": 0.6792,
+      "learning_rate": 2.1555583366681843e-06,
+      "loss": 0.8273,
       "step": 27961
     },
     {
-      "epoch": 0.7679547389525143,
+      "epoch": 0.7934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 2.693699692853785e-06,
-      "loss": 0.8799,
+      "learning_rate": 2.154988358093437e-06,
+      "loss": 0.7992,
       "step": 27962
     },
     {
-      "epoch": 0.7679822031803576,
+      "epoch": 0.7935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 2.6930923837294488e-06,
-      "loss": 0.7247,
+      "learning_rate": 2.1544184457858517e-06,
+      "loss": 0.8531,
       "step": 27963
     },
     {
-      "epoch": 0.7680096674082009,
+      "epoch": 0.7935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 2.6924851324201926e-06,
-      "loss": 0.7537,
+      "learning_rate": 2.153848599750248e-06,
+      "loss": 0.7672,
       "step": 27964
     },
     {
-      "epoch": 0.768037131636044,
+      "epoch": 0.7935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 2.691877938930826e-06,
-      "loss": 0.8041,
+      "learning_rate": 2.15327881999144e-06,
+      "loss": 0.8652,
       "step": 27965
     },
     {
-      "epoch": 0.7680645958638873,
+      "epoch": 0.7935868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 2.691270803266154e-06,
-      "loss": 0.8589,
+      "learning_rate": 2.1527091065142358e-06,
+      "loss": 0.8182,
       "step": 27966
     },
     {
-      "epoch": 0.7680920600917305,
+      "epoch": 0.7936152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 2.6906637254309762e-06,
-      "loss": 0.8019,
+      "learning_rate": 2.152139459323451e-06,
+      "loss": 0.7155,
       "step": 27967
     },
     {
-      "epoch": 0.7681195243195738,
+      "epoch": 0.7936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 2.6900567054300973e-06,
-      "loss": 0.8065,
+      "learning_rate": 2.151569878423899e-06,
+      "loss": 0.8325,
       "step": 27968
     },
     {
-      "epoch": 0.768146988547417,
+      "epoch": 0.793671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 2.6894497432683254e-06,
-      "loss": 0.8812,
+      "learning_rate": 2.151000363820387e-06,
+      "loss": 0.7749,
       "step": 27969
     },
     {
-      "epoch": 0.7681744527752602,
+      "epoch": 0.7937003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 2.6888428389504563e-06,
-      "loss": 0.8504,
+      "learning_rate": 2.150430915517727e-06,
+      "loss": 0.8856,
       "step": 27970
     },
     {
-      "epoch": 0.7682019170031035,
+      "epoch": 0.7937287173666289,
       "grad_norm": 0.0,
-      "learning_rate": 2.688235992481294e-06,
-      "loss": 0.8844,
+      "learning_rate": 2.149861533520734e-06,
+      "loss": 0.8567,
       "step": 27971
     },
     {
-      "epoch": 0.7682293812309466,
+      "epoch": 0.7937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 2.68762920386564e-06,
-      "loss": 0.8487,
+      "learning_rate": 2.1492922178342092e-06,
+      "loss": 0.8324,
       "step": 27972
     },
     {
-      "epoch": 0.7682568454587899,
+      "epoch": 0.7937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 2.687022473108298e-06,
-      "loss": 0.7554,
+      "learning_rate": 2.148722968462967e-06,
+      "loss": 0.8601,
       "step": 27973
     },
     {
-      "epoch": 0.7682843096866332,
+      "epoch": 0.7938138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 2.6864158002140693e-06,
-      "loss": 0.8684,
+      "learning_rate": 2.1481537854118172e-06,
+      "loss": 0.7646,
       "step": 27974
     },
     {
-      "epoch": 0.7683117739144764,
+      "epoch": 0.7938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.685809185187751e-06,
-      "loss": 0.7975,
+      "learning_rate": 2.147584668685563e-06,
+      "loss": 0.8418,
       "step": 27975
     },
     {
-      "epoch": 0.7683392381423196,
+      "epoch": 0.7938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 2.6852026280341413e-06,
-      "loss": 0.7802,
+      "learning_rate": 2.147015618289013e-06,
+      "loss": 0.7826,
       "step": 27976
     },
     {
-      "epoch": 0.7683667023701629,
+      "epoch": 0.7938989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 2.6845961287580424e-06,
-      "loss": 0.8457,
+      "learning_rate": 2.1464466342269763e-06,
+      "loss": 0.7221,
       "step": 27977
     },
     {
-      "epoch": 0.7683941665980061,
+      "epoch": 0.7939273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 2.6839896873642526e-06,
-      "loss": 0.7596,
+      "learning_rate": 2.1458777165042566e-06,
+      "loss": 0.8259,
       "step": 27978
     },
     {
-      "epoch": 0.7684216308258494,
+      "epoch": 0.793955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.6833833038575697e-06,
-      "loss": 0.8194,
+      "learning_rate": 2.145308865125665e-06,
+      "loss": 0.8202,
       "step": 27979
     },
     {
-      "epoch": 0.7684490950536925,
+      "epoch": 0.7939841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 2.682776978242795e-06,
-      "loss": 0.8309,
+      "learning_rate": 2.1447400800959994e-06,
+      "loss": 0.8408,
       "step": 27980
     },
     {
-      "epoch": 0.7684765592815358,
+      "epoch": 0.7940124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 2.6821707105247206e-06,
-      "loss": 0.8171,
+      "learning_rate": 2.1441713614200664e-06,
+      "loss": 0.7983,
       "step": 27981
     },
     {
-      "epoch": 0.7685040235093791,
+      "epoch": 0.7940408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 2.681564500708147e-06,
-      "loss": 0.8136,
+      "learning_rate": 2.1436027091026745e-06,
+      "loss": 0.83,
       "step": 27982
     },
     {
-      "epoch": 0.7685314877372222,
+      "epoch": 0.7940692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 2.680958348797872e-06,
-      "loss": 0.8761,
+      "learning_rate": 2.143034123148622e-06,
+      "loss": 0.8505,
       "step": 27983
     },
     {
-      "epoch": 0.7685589519650655,
+      "epoch": 0.7940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 2.6803522547986873e-06,
-      "loss": 0.7812,
+      "learning_rate": 2.1424656035627124e-06,
+      "loss": 0.8664,
       "step": 27984
     },
     {
-      "epoch": 0.7685864161929087,
+      "epoch": 0.7941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 2.679746218715391e-06,
-      "loss": 0.8182,
+      "learning_rate": 2.141897150349752e-06,
+      "loss": 0.7388,
       "step": 27985
     },
     {
-      "epoch": 0.768613880420752,
+      "epoch": 0.7941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 2.67914024055278e-06,
-      "loss": 0.884,
+      "learning_rate": 2.1413287635145365e-06,
+      "loss": 0.7571,
       "step": 27986
     },
     {
-      "epoch": 0.7686413446485952,
+      "epoch": 0.7941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 2.6785343203156454e-06,
-      "loss": 0.8768,
+      "learning_rate": 2.14076044306187e-06,
+      "loss": 0.9072,
       "step": 27987
     },
     {
-      "epoch": 0.7686688088764384,
+      "epoch": 0.794211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.677928458008783e-06,
-      "loss": 0.8043,
+      "learning_rate": 2.140192188996557e-06,
+      "loss": 0.7317,
       "step": 27988
     },
     {
-      "epoch": 0.7686962731042817,
+      "epoch": 0.7942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 2.6773226536369867e-06,
-      "loss": 0.7403,
+      "learning_rate": 2.1396240013233903e-06,
+      "loss": 0.8087,
       "step": 27989
     },
     {
-      "epoch": 0.768723737332125,
+      "epoch": 0.7942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 2.6767169072050537e-06,
-      "loss": 0.737,
+      "learning_rate": 2.139055880047174e-06,
+      "loss": 0.8426,
       "step": 27990
     },
     {
-      "epoch": 0.7687512015599681,
+      "epoch": 0.7942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 2.6761112187177697e-06,
-      "loss": 0.9019,
+      "learning_rate": 2.138487825172706e-06,
+      "loss": 0.7904,
       "step": 27991
     },
     {
-      "epoch": 0.7687786657878114,
+      "epoch": 0.7943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 2.675505588179934e-06,
-      "loss": 0.8464,
+      "learning_rate": 2.1379198367047838e-06,
+      "loss": 0.7792,
       "step": 27992
     },
     {
-      "epoch": 0.7688061300156546,
+      "epoch": 0.7943530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 2.674900015596332e-06,
-      "loss": 0.7593,
+      "learning_rate": 2.1373519146482092e-06,
+      "loss": 0.7539,
       "step": 27993
     },
     {
-      "epoch": 0.7688335942434978,
+      "epoch": 0.7943813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 2.6742945009717593e-06,
-      "loss": 0.8209,
+      "learning_rate": 2.1367840590077735e-06,
+      "loss": 0.7743,
       "step": 27994
     },
     {
-      "epoch": 0.7688610584713411,
+      "epoch": 0.7944097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 2.6736890443110063e-06,
-      "loss": 0.855,
+      "learning_rate": 2.1362162697882773e-06,
+      "loss": 0.8275,
       "step": 27995
     },
     {
-      "epoch": 0.7688885226991843,
+      "epoch": 0.7944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 2.6730836456188623e-06,
-      "loss": 0.8464,
+      "learning_rate": 2.1356485469945175e-06,
+      "loss": 0.8074,
       "step": 27996
     },
     {
-      "epoch": 0.7689159869270276,
+      "epoch": 0.794466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 2.67247830490012e-06,
-      "loss": 0.8168,
+      "learning_rate": 2.135080890631286e-06,
+      "loss": 0.7688,
       "step": 27997
     },
     {
-      "epoch": 0.7689434511548707,
+      "epoch": 0.794494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 2.67187302215957e-06,
-      "loss": 0.7791,
+      "learning_rate": 2.134513300703379e-06,
+      "loss": 0.8053,
       "step": 27998
     },
     {
-      "epoch": 0.768970915382714,
+      "epoch": 0.7945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 2.6712677974019994e-06,
-      "loss": 0.7954,
+      "learning_rate": 2.1339457772155957e-06,
+      "loss": 0.8484,
       "step": 27999
     },
     {
-      "epoch": 0.7689983796105573,
+      "epoch": 0.7945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 2.6706626306321935e-06,
-      "loss": 0.804,
+      "learning_rate": 2.133378320172722e-06,
+      "loss": 0.8196,
       "step": 28000
     },
     {
-      "epoch": 0.7690258438384004,
+      "epoch": 0.7945800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 2.670057521854945e-06,
-      "loss": 0.842,
+      "learning_rate": 2.1328109295795563e-06,
+      "loss": 0.7378,
       "step": 28001
     },
     {
-      "epoch": 0.7690533080662437,
+      "epoch": 0.7946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 2.6694524710750404e-06,
-      "loss": 0.8376,
+      "learning_rate": 2.1322436054408923e-06,
+      "loss": 0.8456,
       "step": 28002
     },
     {
-      "epoch": 0.769080772294087,
+      "epoch": 0.7946367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 2.668847478297266e-06,
-      "loss": 0.9367,
+      "learning_rate": 2.1316763477615175e-06,
+      "loss": 0.8169,
       "step": 28003
     },
     {
-      "epoch": 0.7691082365219302,
+      "epoch": 0.7946651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 2.668242543526415e-06,
-      "loss": 0.8456,
+      "learning_rate": 2.131109156546226e-06,
+      "loss": 0.7251,
       "step": 28004
     },
     {
-      "epoch": 0.7691357007497734,
+      "epoch": 0.7946935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 2.6676376667672655e-06,
-      "loss": 0.8022,
+      "learning_rate": 2.130542031799813e-06,
+      "loss": 0.7467,
       "step": 28005
     },
     {
-      "epoch": 0.7691631649776166,
+      "epoch": 0.7947219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 2.6670328480246065e-06,
-      "loss": 0.9194,
+      "learning_rate": 2.1299749735270613e-06,
+      "loss": 0.8536,
       "step": 28006
     },
     {
-      "epoch": 0.7691906292054599,
+      "epoch": 0.7947502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 2.666428087303228e-06,
-      "loss": 0.767,
+      "learning_rate": 2.129407981732766e-06,
+      "loss": 0.8356,
       "step": 28007
     },
     {
-      "epoch": 0.7692180934333032,
+      "epoch": 0.7947786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 2.665823384607907e-06,
-      "loss": 0.7804,
+      "learning_rate": 2.1288410564217145e-06,
+      "loss": 0.7962,
       "step": 28008
     },
     {
-      "epoch": 0.7692455576611463,
+      "epoch": 0.7948070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 2.6652187399434336e-06,
-      "loss": 0.8317,
+      "learning_rate": 2.1282741975986964e-06,
+      "loss": 0.8664,
       "step": 28009
     },
     {
-      "epoch": 0.7692730218889896,
+      "epoch": 0.7948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 2.664614153314593e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.127707405268503e-06,
+      "loss": 0.8542,
       "step": 28010
     },
     {
-      "epoch": 0.7693004861168328,
+      "epoch": 0.7948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 2.664009624726164e-06,
-      "loss": 0.8235,
+      "learning_rate": 2.127140679435916e-06,
+      "loss": 0.7982,
       "step": 28011
     },
     {
-      "epoch": 0.769327950344676,
+      "epoch": 0.7948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 2.663405154182933e-06,
-      "loss": 0.8086,
+      "learning_rate": 2.126574020105726e-06,
+      "loss": 0.7806,
       "step": 28012
     },
     {
-      "epoch": 0.7693554145725193,
+      "epoch": 0.7949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 2.6628007416896815e-06,
-      "loss": 0.856,
+      "learning_rate": 2.1260074272827227e-06,
+      "loss": 0.8553,
       "step": 28013
     },
     {
-      "epoch": 0.7693828788003625,
+      "epoch": 0.7949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 2.662196387251196e-06,
-      "loss": 0.6968,
+      "learning_rate": 2.1254409009716837e-06,
+      "loss": 0.7828,
       "step": 28014
     },
     {
-      "epoch": 0.7694103430282058,
+      "epoch": 0.7949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 2.6615920908722525e-06,
-      "loss": 0.8442,
+      "learning_rate": 2.1248744411774014e-06,
+      "loss": 0.7774,
       "step": 28015
     },
     {
-      "epoch": 0.769437807256049,
+      "epoch": 0.795005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 2.660987852557637e-06,
-      "loss": 0.7725,
+      "learning_rate": 2.1243080479046606e-06,
+      "loss": 0.8355,
       "step": 28016
     },
     {
-      "epoch": 0.7694652714838922,
+      "epoch": 0.7950340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 2.6603836723121266e-06,
-      "loss": 0.8609,
+      "learning_rate": 2.123741721158241e-06,
+      "loss": 0.8107,
       "step": 28017
     },
     {
-      "epoch": 0.7694927357117355,
+      "epoch": 0.7950624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 2.6597795501405034e-06,
-      "loss": 0.7949,
+      "learning_rate": 2.123175460942929e-06,
+      "loss": 0.7677,
       "step": 28018
     },
     {
-      "epoch": 0.7695201999395787,
+      "epoch": 0.7950908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 2.659175486047547e-06,
-      "loss": 0.852,
+      "learning_rate": 2.122609267263511e-06,
+      "loss": 0.7904,
       "step": 28019
     },
     {
-      "epoch": 0.7695476641674219,
+      "epoch": 0.7951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 2.658571480038038e-06,
-      "loss": 0.9631,
+      "learning_rate": 2.122043140124764e-06,
+      "loss": 0.8893,
       "step": 28020
     },
     {
-      "epoch": 0.7695751283952652,
+      "epoch": 0.7951475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 2.6579675321167565e-06,
-      "loss": 0.8217,
+      "learning_rate": 2.121477079531472e-06,
+      "loss": 0.727,
       "step": 28021
     },
     {
-      "epoch": 0.7696025926231084,
+      "epoch": 0.7951759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 2.657363642288482e-06,
-      "loss": 0.7773,
+      "learning_rate": 2.1209110854884184e-06,
+      "loss": 0.7927,
       "step": 28022
     },
     {
-      "epoch": 0.7696300568509516,
+      "epoch": 0.7952043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 2.6567598105579895e-06,
-      "loss": 0.7576,
+      "learning_rate": 2.120345158000382e-06,
+      "loss": 0.8681,
       "step": 28023
     },
     {
-      "epoch": 0.7696575210787948,
+      "epoch": 0.7952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 2.6561560369300564e-06,
-      "loss": 0.7623,
+      "learning_rate": 2.1197792970721454e-06,
+      "loss": 0.7431,
       "step": 28024
     },
     {
-      "epoch": 0.7696849853066381,
+      "epoch": 0.7952610669693531,
       "grad_norm": 0.0,
-      "learning_rate": 2.6555523214094614e-06,
-      "loss": 0.9023,
+      "learning_rate": 2.1192135027084893e-06,
+      "loss": 0.8414,
       "step": 28025
     },
     {
-      "epoch": 0.7697124495344814,
+      "epoch": 0.7952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 2.65494866400098e-06,
-      "loss": 0.7753,
+      "learning_rate": 2.118647774914189e-06,
+      "loss": 0.8222,
       "step": 28026
     },
     {
-      "epoch": 0.7697399137623245,
+      "epoch": 0.7953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 2.6543450647093915e-06,
-      "loss": 0.7259,
+      "learning_rate": 2.1180821136940244e-06,
+      "loss": 0.8484,
       "step": 28027
     },
     {
-      "epoch": 0.7697673779901678,
+      "epoch": 0.7953461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 2.6537415235394724e-06,
-      "loss": 0.8339,
+      "learning_rate": 2.117516519052777e-06,
+      "loss": 0.9273,
       "step": 28028
     },
     {
-      "epoch": 0.7697948422180111,
+      "epoch": 0.7953745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 2.6531380404959937e-06,
-      "loss": 0.8568,
+      "learning_rate": 2.11695099099522e-06,
+      "loss": 0.8832,
       "step": 28029
     },
     {
-      "epoch": 0.7698223064458543,
+      "epoch": 0.7954029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 2.652534615583735e-06,
-      "loss": 0.8673,
+      "learning_rate": 2.116385529526135e-06,
+      "loss": 0.8226,
       "step": 28030
     },
     {
-      "epoch": 0.7698497706736975,
+      "epoch": 0.7954313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 2.6519312488074666e-06,
-      "loss": 0.8511,
+      "learning_rate": 2.1158201346502927e-06,
+      "loss": 0.8211,
       "step": 28031
     },
     {
-      "epoch": 0.7698772349015407,
+      "epoch": 0.7954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 2.651327940171964e-06,
-      "loss": 0.8308,
+      "learning_rate": 2.1152548063724722e-06,
+      "loss": 0.7286,
       "step": 28032
     },
     {
-      "epoch": 0.769904699129384,
+      "epoch": 0.7954880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 2.6507246896820004e-06,
-      "loss": 0.7822,
+      "learning_rate": 2.114689544697451e-06,
+      "loss": 0.8586,
       "step": 28033
     },
     {
-      "epoch": 0.7699321633572273,
+      "epoch": 0.795516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 2.650121497342354e-06,
-      "loss": 0.8872,
+      "learning_rate": 2.11412434963e-06,
+      "loss": 0.8387,
       "step": 28034
     },
     {
-      "epoch": 0.7699596275850704,
+      "epoch": 0.7955448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 2.6495183631577904e-06,
-      "loss": 0.7832,
+      "learning_rate": 2.1135592211748947e-06,
+      "loss": 0.7379,
       "step": 28035
     },
     {
-      "epoch": 0.7699870918129137,
+      "epoch": 0.7955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 2.648915287133085e-06,
-      "loss": 0.806,
+      "learning_rate": 2.112994159336912e-06,
+      "loss": 0.8265,
       "step": 28036
     },
     {
-      "epoch": 0.7700145560407569,
+      "epoch": 0.7956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 2.648312269273008e-06,
-      "loss": 0.8163,
+      "learning_rate": 2.112429164120816e-06,
+      "loss": 0.8206,
       "step": 28037
     },
     {
-      "epoch": 0.7700420202686001,
+      "epoch": 0.7956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 2.647709309582336e-06,
-      "loss": 0.8909,
+      "learning_rate": 2.111864235531389e-06,
+      "loss": 0.8069,
       "step": 28038
     },
     {
-      "epoch": 0.7700694844964434,
+      "epoch": 0.7956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 2.647106408065836e-06,
-      "loss": 0.8979,
+      "learning_rate": 2.111299373573402e-06,
+      "loss": 0.6778,
       "step": 28039
     },
     {
-      "epoch": 0.7700969487242866,
+      "epoch": 0.7956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 2.6465035647282753e-06,
-      "loss": 0.7369,
+      "learning_rate": 2.1107345782516207e-06,
+      "loss": 0.7874,
       "step": 28040
     },
     {
-      "epoch": 0.7701244129521299,
+      "epoch": 0.7957150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 2.645900779574427e-06,
-      "loss": 0.8279,
+      "learning_rate": 2.110169849570819e-06,
+      "loss": 0.7967,
       "step": 28041
     },
     {
-      "epoch": 0.7701518771799731,
+      "epoch": 0.7957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 2.64529805260906e-06,
-      "loss": 0.806,
+      "learning_rate": 2.1096051875357703e-06,
+      "loss": 0.8579,
       "step": 28042
     },
     {
-      "epoch": 0.7701793414078163,
+      "epoch": 0.7957718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 2.6446953838369437e-06,
-      "loss": 0.7417,
+      "learning_rate": 2.109040592151239e-06,
+      "loss": 0.7252,
       "step": 28043
     },
     {
-      "epoch": 0.7702068056356596,
+      "epoch": 0.795800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 2.6440927732628476e-06,
-      "loss": 0.7768,
+      "learning_rate": 2.108476063421996e-06,
+      "loss": 0.7036,
       "step": 28044
     },
     {
-      "epoch": 0.7702342698635027,
+      "epoch": 0.7958286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 2.6434902208915414e-06,
-      "loss": 0.8857,
+      "learning_rate": 2.107911601352813e-06,
+      "loss": 0.7614,
       "step": 28045
     },
     {
-      "epoch": 0.770261734091346,
+      "epoch": 0.7958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 2.6428877267277863e-06,
-      "loss": 0.8005,
+      "learning_rate": 2.1073472059484534e-06,
+      "loss": 0.7868,
       "step": 28046
     },
     {
-      "epoch": 0.7702891983191893,
+      "epoch": 0.7958853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 2.6422852907763573e-06,
-      "loss": 0.745,
+      "learning_rate": 2.1067828772136856e-06,
+      "loss": 1.0014,
       "step": 28047
     },
     {
-      "epoch": 0.7703166625470325,
+      "epoch": 0.7959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 2.641682913042014e-06,
-      "loss": 0.9249,
+      "learning_rate": 2.1062186151532813e-06,
+      "loss": 0.8203,
       "step": 28048
     },
     {
-      "epoch": 0.7703441267748757,
+      "epoch": 0.7959421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 2.641080593529527e-06,
-      "loss": 0.8332,
+      "learning_rate": 2.1056544197720006e-06,
+      "loss": 0.8329,
       "step": 28049
     },
     {
-      "epoch": 0.770371591002719,
+      "epoch": 0.7959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 2.640478332243659e-06,
-      "loss": 0.8188,
+      "learning_rate": 2.1050902910746118e-06,
+      "loss": 0.9742,
       "step": 28050
     },
     {
-      "epoch": 0.7703990552305622,
+      "epoch": 0.7959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 2.639876129189182e-06,
-      "loss": 0.7897,
+      "learning_rate": 2.104526229065883e-06,
+      "loss": 0.6776,
       "step": 28051
     },
     {
-      "epoch": 0.7704265194584055,
+      "epoch": 0.796027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 2.6392739843708513e-06,
-      "loss": 0.7852,
+      "learning_rate": 2.1039622337505726e-06,
+      "loss": 0.9295,
       "step": 28052
     },
     {
-      "epoch": 0.7704539836862486,
+      "epoch": 0.79605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.6386718977934368e-06,
-      "loss": 0.773,
+      "learning_rate": 2.1033983051334485e-06,
+      "loss": 0.8938,
       "step": 28053
     },
     {
-      "epoch": 0.7704814479140919,
+      "epoch": 0.7960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 2.638069869461706e-06,
-      "loss": 0.7413,
+      "learning_rate": 2.102834443219275e-06,
+      "loss": 0.9515,
       "step": 28054
     },
     {
-      "epoch": 0.7705089121419352,
+      "epoch": 0.7961123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 2.637467899380414e-06,
-      "loss": 0.7889,
+      "learning_rate": 2.102270648012812e-06,
+      "loss": 0.8186,
       "step": 28055
     },
     {
-      "epoch": 0.7705363763697783,
+      "epoch": 0.7961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 2.636865987554328e-06,
-      "loss": 0.8494,
+      "learning_rate": 2.101706919518828e-06,
+      "loss": 0.8318,
       "step": 28056
     },
     {
-      "epoch": 0.7705638405976216,
+      "epoch": 0.7961691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 2.6362641339882133e-06,
-      "loss": 0.8415,
+      "learning_rate": 2.1011432577420777e-06,
+      "loss": 0.7952,
       "step": 28057
     },
     {
-      "epoch": 0.7705913048254648,
+      "epoch": 0.7961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 2.6356623386868264e-06,
-      "loss": 0.781,
+      "learning_rate": 2.1005796626873253e-06,
+      "loss": 0.8855,
       "step": 28058
     },
     {
-      "epoch": 0.7706187690533081,
+      "epoch": 0.7962258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 2.635060601654932e-06,
-      "loss": 0.844,
+      "learning_rate": 2.100016134359334e-06,
+      "loss": 0.8647,
       "step": 28059
     },
     {
-      "epoch": 0.7706462332811513,
+      "epoch": 0.7962542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 2.6344589228972906e-06,
-      "loss": 0.9494,
+      "learning_rate": 2.0994526727628593e-06,
+      "loss": 0.8015,
       "step": 28060
     },
     {
-      "epoch": 0.7706736975089945,
+      "epoch": 0.7962826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 2.633857302418663e-06,
-      "loss": 0.6859,
+      "learning_rate": 2.0988892779026626e-06,
+      "loss": 0.9025,
       "step": 28061
     },
     {
-      "epoch": 0.7707011617368378,
+      "epoch": 0.796311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 2.6332557402238134e-06,
-      "loss": 0.8223,
+      "learning_rate": 2.098325949783506e-06,
+      "loss": 0.8825,
       "step": 28062
     },
     {
-      "epoch": 0.7707286259646811,
+      "epoch": 0.7963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 2.632654236317498e-06,
-      "loss": 0.7776,
+      "learning_rate": 2.0977626884101433e-06,
+      "loss": 0.7247,
       "step": 28063
     },
     {
-      "epoch": 0.7707560901925242,
+      "epoch": 0.7963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 2.6320527907044733e-06,
-      "loss": 0.7422,
+      "learning_rate": 2.0971994937873342e-06,
+      "loss": 0.785,
       "step": 28064
     },
     {
-      "epoch": 0.7707835544203675,
+      "epoch": 0.7963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 2.631451403389501e-06,
-      "loss": 0.7353,
+      "learning_rate": 2.0966363659198385e-06,
+      "loss": 0.7773,
       "step": 28065
     },
     {
-      "epoch": 0.7708110186482107,
+      "epoch": 0.7964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 2.6308500743773392e-06,
-      "loss": 0.8653,
+      "learning_rate": 2.0960733048124082e-06,
+      "loss": 0.8859,
       "step": 28066
     },
     {
-      "epoch": 0.7708384828760539,
+      "epoch": 0.7964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 2.6302488036727457e-06,
-      "loss": 0.8413,
+      "learning_rate": 2.0955103104698026e-06,
+      "loss": 0.6983,
       "step": 28067
     },
     {
-      "epoch": 0.7708659471038972,
+      "epoch": 0.7964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 2.6296475912804786e-06,
-      "loss": 0.8347,
+      "learning_rate": 2.094947382896776e-06,
+      "loss": 0.6953,
       "step": 28068
     },
     {
-      "epoch": 0.7708934113317404,
+      "epoch": 0.7965096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 2.6290464372052983e-06,
-      "loss": 0.8369,
+      "learning_rate": 2.094384522098084e-06,
+      "loss": 0.891,
       "step": 28069
     },
     {
-      "epoch": 0.7709208755595837,
+      "epoch": 0.7965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 2.628445341451955e-06,
-      "loss": 0.7858,
+      "learning_rate": 2.0938217280784846e-06,
+      "loss": 0.759,
       "step": 28070
     },
     {
-      "epoch": 0.7709483397874268,
+      "epoch": 0.796566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 2.62784430402521e-06,
-      "loss": 0.8419,
+      "learning_rate": 2.0932590008427256e-06,
+      "loss": 0.788,
       "step": 28071
     },
     {
-      "epoch": 0.7709758040152701,
+      "epoch": 0.7965947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 2.6272433249298135e-06,
-      "loss": 0.9664,
+      "learning_rate": 2.0926963403955637e-06,
+      "loss": 0.8058,
       "step": 28072
     },
     {
-      "epoch": 0.7710032682431134,
+      "epoch": 0.7966231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 2.626642404170524e-06,
-      "loss": 0.7903,
+      "learning_rate": 2.0921337467417536e-06,
+      "loss": 0.7253,
       "step": 28073
     },
     {
-      "epoch": 0.7710307324709565,
+      "epoch": 0.7966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 2.626041541752097e-06,
-      "loss": 0.8602,
+      "learning_rate": 2.0915712198860417e-06,
+      "loss": 0.8843,
       "step": 28074
     },
     {
-      "epoch": 0.7710581966987998,
+      "epoch": 0.7966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 2.6254407376792867e-06,
-      "loss": 0.7297,
+      "learning_rate": 2.091008759833184e-06,
+      "loss": 0.7838,
       "step": 28075
     },
     {
-      "epoch": 0.7710856609266431,
+      "epoch": 0.7967082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 2.6248399919568446e-06,
-      "loss": 0.8899,
+      "learning_rate": 2.090446366587934e-06,
+      "loss": 0.8925,
       "step": 28076
     },
     {
-      "epoch": 0.7711131251544863,
+      "epoch": 0.7967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.6242393045895244e-06,
-      "loss": 0.8225,
+      "learning_rate": 2.0898840401550356e-06,
+      "loss": 0.7518,
       "step": 28077
     },
     {
-      "epoch": 0.7711405893823295,
+      "epoch": 0.7967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 2.623638675582083e-06,
-      "loss": 0.9033,
+      "learning_rate": 2.089321780539243e-06,
+      "loss": 0.7919,
       "step": 28078
     },
     {
-      "epoch": 0.7711680536101727,
+      "epoch": 0.7967934165720771,
       "grad_norm": 0.0,
-      "learning_rate": 2.623038104939267e-06,
-      "loss": 0.8468,
+      "learning_rate": 2.088759587745307e-06,
+      "loss": 0.7871,
       "step": 28079
     },
     {
-      "epoch": 0.771195517838016,
+      "epoch": 0.7968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 2.6224375926658307e-06,
-      "loss": 0.8673,
+      "learning_rate": 2.088197461777972e-06,
+      "loss": 0.8213,
       "step": 28080
     },
     {
-      "epoch": 0.7712229820658593,
+      "epoch": 0.796850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 2.62183713876653e-06,
-      "loss": 0.7257,
+      "learning_rate": 2.0876354026419887e-06,
+      "loss": 0.7914,
       "step": 28081
     },
     {
-      "epoch": 0.7712504462937024,
+      "epoch": 0.7968785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 2.6212367432461084e-06,
-      "loss": 0.8417,
+      "learning_rate": 2.0870734103421075e-06,
+      "loss": 0.8437,
       "step": 28082
     },
     {
-      "epoch": 0.7712779105215457,
+      "epoch": 0.7969069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 2.6206364061093194e-06,
-      "loss": 0.7444,
+      "learning_rate": 2.086511484883067e-06,
+      "loss": 0.7804,
       "step": 28083
     },
     {
-      "epoch": 0.7713053747493889,
+      "epoch": 0.7969353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 2.620036127360914e-06,
-      "loss": 0.8076,
+      "learning_rate": 2.0859496262696254e-06,
+      "loss": 0.7683,
       "step": 28084
     },
     {
-      "epoch": 0.7713328389772321,
+      "epoch": 0.7969636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 2.619435907005642e-06,
-      "loss": 0.8052,
+      "learning_rate": 2.0853878345065203e-06,
+      "loss": 0.762,
       "step": 28085
     },
     {
-      "epoch": 0.7713603032050754,
+      "epoch": 0.7969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 2.618835745048254e-06,
-      "loss": 0.7717,
+      "learning_rate": 2.0848261095985e-06,
+      "loss": 0.7509,
       "step": 28086
     },
     {
-      "epoch": 0.7713877674329186,
+      "epoch": 0.7970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 2.6182356414934973e-06,
-      "loss": 0.927,
+      "learning_rate": 2.084264451550313e-06,
+      "loss": 0.7758,
       "step": 28087
     },
     {
-      "epoch": 0.7714152316607619,
+      "epoch": 0.7970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 2.6176355963461164e-06,
-      "loss": 0.7551,
+      "learning_rate": 2.0837028603666965e-06,
+      "loss": 0.8207,
       "step": 28088
     },
     {
-      "epoch": 0.7714426958886051,
+      "epoch": 0.7970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 2.617035609610863e-06,
-      "loss": 0.8394,
+      "learning_rate": 2.083141336052398e-06,
+      "loss": 0.8813,
       "step": 28089
     },
     {
-      "epoch": 0.7714701601164483,
+      "epoch": 0.797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.616435681292483e-06,
-      "loss": 0.7869,
+      "learning_rate": 2.0825798786121632e-06,
+      "loss": 0.7508,
       "step": 28090
     },
     {
-      "epoch": 0.7714976243442916,
+      "epoch": 0.7971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 2.615835811395724e-06,
-      "loss": 0.7349,
+      "learning_rate": 2.0820184880507287e-06,
+      "loss": 0.7859,
       "step": 28091
     },
     {
-      "epoch": 0.7715250885721348,
+      "epoch": 0.7971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 2.615235999925333e-06,
-      "loss": 0.8087,
+      "learning_rate": 2.081457164372841e-06,
+      "loss": 0.9124,
       "step": 28092
     },
     {
-      "epoch": 0.771552552799978,
+      "epoch": 0.7971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 2.6146362468860573e-06,
-      "loss": 0.8759,
+      "learning_rate": 2.0808959075832423e-06,
+      "loss": 0.9303,
       "step": 28093
     },
     {
-      "epoch": 0.7715800170278213,
+      "epoch": 0.7972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 2.6140365522826383e-06,
-      "loss": 0.8949,
+      "learning_rate": 2.080334717686671e-06,
+      "loss": 0.7578,
       "step": 28094
     },
     {
-      "epoch": 0.7716074812556645,
+      "epoch": 0.7972474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 2.613436916119826e-06,
-      "loss": 0.8286,
+      "learning_rate": 2.0797735946878663e-06,
+      "loss": 0.9327,
       "step": 28095
     },
     {
-      "epoch": 0.7716349454835078,
+      "epoch": 0.7972758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 2.6128373384023596e-06,
-      "loss": 0.8129,
+      "learning_rate": 2.079212538591574e-06,
+      "loss": 0.9008,
       "step": 28096
     },
     {
-      "epoch": 0.7716624097113509,
+      "epoch": 0.7973041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 2.612237819134985e-06,
-      "loss": 0.6896,
+      "learning_rate": 2.078651549402526e-06,
+      "loss": 0.7726,
       "step": 28097
     },
     {
-      "epoch": 0.7716898739391942,
+      "epoch": 0.7973325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 2.6116383583224503e-06,
-      "loss": 0.9207,
+      "learning_rate": 2.0780906271254643e-06,
+      "loss": 0.827,
       "step": 28098
     },
     {
-      "epoch": 0.7717173381670375,
+      "epoch": 0.797360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 2.6110389559694902e-06,
-      "loss": 0.8534,
+      "learning_rate": 2.0775297717651275e-06,
+      "loss": 0.814,
       "step": 28099
     },
     {
-      "epoch": 0.7717448023948806,
+      "epoch": 0.79738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.610439612080854e-06,
-      "loss": 0.8392,
+      "learning_rate": 2.076968983326253e-06,
+      "loss": 0.8335,
       "step": 28100
     },
     {
-      "epoch": 0.7717722666227239,
+      "epoch": 0.7974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 2.6098403266612815e-06,
-      "loss": 0.8566,
+      "learning_rate": 2.0764082618135806e-06,
+      "loss": 0.777,
       "step": 28101
     },
     {
-      "epoch": 0.7717997308505672,
+      "epoch": 0.7974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 2.609241099715517e-06,
-      "loss": 0.9015,
+      "learning_rate": 2.075847607231841e-06,
+      "loss": 0.8844,
       "step": 28102
     },
     {
-      "epoch": 0.7718271950784104,
+      "epoch": 0.7974744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 2.608641931248298e-06,
-      "loss": 0.778,
+      "learning_rate": 2.075287019585773e-06,
+      "loss": 0.8071,
       "step": 28103
     },
     {
-      "epoch": 0.7718546593062536,
+      "epoch": 0.7975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 2.6080428212643692e-06,
-      "loss": 0.7998,
+      "learning_rate": 2.0747264988801142e-06,
+      "loss": 0.7609,
       "step": 28104
     },
     {
-      "epoch": 0.7718821235340968,
+      "epoch": 0.7975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 2.6074437697684673e-06,
-      "loss": 0.8506,
+      "learning_rate": 2.0741660451195942e-06,
+      "loss": 0.8577,
       "step": 28105
     },
     {
-      "epoch": 0.7719095877619401,
+      "epoch": 0.7975595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 2.6068447767653328e-06,
-      "loss": 0.8585,
+      "learning_rate": 2.0736056583089505e-06,
+      "loss": 0.7584,
       "step": 28106
     },
     {
-      "epoch": 0.7719370519897834,
+      "epoch": 0.7975879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 2.606245842259707e-06,
-      "loss": 0.8958,
+      "learning_rate": 2.073045338452918e-06,
+      "loss": 0.8415,
       "step": 28107
     },
     {
-      "epoch": 0.7719645162176265,
+      "epoch": 0.7976163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 2.605646966256328e-06,
-      "loss": 0.8125,
+      "learning_rate": 2.072485085556225e-06,
+      "loss": 0.7102,
       "step": 28108
     },
     {
-      "epoch": 0.7719919804454698,
+      "epoch": 0.7976447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 2.605048148759933e-06,
-      "loss": 0.8333,
+      "learning_rate": 2.0719248996236075e-06,
+      "loss": 0.8514,
       "step": 28109
     },
     {
-      "epoch": 0.772019444673313,
+      "epoch": 0.7976730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 2.6044493897752667e-06,
-      "loss": 0.8911,
+      "learning_rate": 2.071364780659799e-06,
+      "loss": 0.757,
       "step": 28110
     },
     {
-      "epoch": 0.7720469089011562,
+      "epoch": 0.7977014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 2.6038506893070604e-06,
-      "loss": 0.8699,
+      "learning_rate": 2.0708047286695265e-06,
+      "loss": 0.7997,
       "step": 28111
     },
     {
-      "epoch": 0.7720743731289995,
+      "epoch": 0.7977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 2.60325204736005e-06,
-      "loss": 0.7643,
+      "learning_rate": 2.0702447436575223e-06,
+      "loss": 0.8235,
       "step": 28112
     },
     {
-      "epoch": 0.7721018373568427,
+      "epoch": 0.7977582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 2.6026534639389745e-06,
-      "loss": 0.8263,
+      "learning_rate": 2.0696848256285206e-06,
+      "loss": 0.8605,
       "step": 28113
     },
     {
-      "epoch": 0.772129301584686,
+      "epoch": 0.7977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 2.6020549390485695e-06,
-      "loss": 0.7823,
+      "learning_rate": 2.069124974587242e-06,
+      "loss": 0.8441,
       "step": 28114
     },
     {
-      "epoch": 0.7721567658125292,
+      "epoch": 0.7978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 2.6014564726935722e-06,
-      "loss": 0.9094,
+      "learning_rate": 2.0685651905384264e-06,
+      "loss": 0.8462,
       "step": 28115
     },
     {
-      "epoch": 0.7721842300403724,
+      "epoch": 0.7978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 2.6008580648787195e-06,
-      "loss": 0.8125,
+      "learning_rate": 2.068005473486795e-06,
+      "loss": 0.8734,
       "step": 28116
     },
     {
-      "epoch": 0.7722116942682157,
+      "epoch": 0.7978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 2.600259715608742e-06,
-      "loss": 0.7461,
+      "learning_rate": 2.0674458234370777e-06,
+      "loss": 0.8097,
       "step": 28117
     },
     {
-      "epoch": 0.7722391584960588,
+      "epoch": 0.797900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 2.5996614248883755e-06,
-      "loss": 0.8055,
+      "learning_rate": 2.066886240394004e-06,
+      "loss": 0.8307,
       "step": 28118
     },
     {
-      "epoch": 0.7722666227239021,
+      "epoch": 0.7979284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 2.5990631927223574e-06,
-      "loss": 0.8474,
+      "learning_rate": 2.0663267243622963e-06,
+      "loss": 0.7994,
       "step": 28119
     },
     {
-      "epoch": 0.7722940869517454,
+      "epoch": 0.7979568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 2.598465019115416e-06,
-      "loss": 0.8322,
+      "learning_rate": 2.065767275346684e-06,
+      "loss": 0.8316,
       "step": 28120
     },
     {
-      "epoch": 0.7723215511795886,
+      "epoch": 0.7979852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 2.5978669040722857e-06,
-      "loss": 0.7989,
+      "learning_rate": 2.065207893351895e-06,
+      "loss": 0.6845,
       "step": 28121
     },
     {
-      "epoch": 0.7723490154074318,
+      "epoch": 0.7980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 2.5972688475977027e-06,
-      "loss": 0.8876,
+      "learning_rate": 2.0646485783826488e-06,
+      "loss": 0.8457,
       "step": 28122
     },
     {
-      "epoch": 0.772376479635275,
+      "epoch": 0.7980419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 2.5966708496963933e-06,
-      "loss": 0.7533,
+      "learning_rate": 2.064089330443673e-06,
+      "loss": 0.9491,
       "step": 28123
     },
     {
-      "epoch": 0.7724039438631183,
+      "epoch": 0.7980703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 2.596072910373092e-06,
-      "loss": 0.7536,
+      "learning_rate": 2.063530149539694e-06,
+      "loss": 0.8669,
       "step": 28124
     },
     {
-      "epoch": 0.7724314080909616,
+      "epoch": 0.7980987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 2.5954750296325294e-06,
-      "loss": 0.7718,
+      "learning_rate": 2.0629710356754296e-06,
+      "loss": 0.6781,
       "step": 28125
     },
     {
-      "epoch": 0.7724588723188047,
+      "epoch": 0.7981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 2.5948772074794405e-06,
-      "loss": 0.829,
+      "learning_rate": 2.062411988855606e-06,
+      "loss": 0.8762,
       "step": 28126
     },
     {
-      "epoch": 0.772486336546648,
+      "epoch": 0.7981555051078321,
       "grad_norm": 0.0,
-      "learning_rate": 2.594279443918547e-06,
-      "loss": 0.8563,
+      "learning_rate": 2.0618530090849475e-06,
+      "loss": 0.9263,
       "step": 28127
     },
     {
-      "epoch": 0.7725138007744913,
+      "epoch": 0.7981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 2.5936817389545878e-06,
-      "loss": 0.8063,
+      "learning_rate": 2.0612940963681703e-06,
+      "loss": 0.8432,
       "step": 28128
     },
     {
-      "epoch": 0.7725412650023344,
+      "epoch": 0.7982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 2.593084092592284e-06,
-      "loss": 0.833,
+      "learning_rate": 2.0607352507099963e-06,
+      "loss": 0.8727,
       "step": 28129
     },
     {
-      "epoch": 0.7725687292301777,
+      "epoch": 0.7982406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 2.592486504836368e-06,
-      "loss": 0.7924,
+      "learning_rate": 2.060176472115155e-06,
+      "loss": 0.8395,
       "step": 28130
     },
     {
-      "epoch": 0.7725961934580209,
+      "epoch": 0.7982690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 2.5918889756915677e-06,
-      "loss": 0.8862,
+      "learning_rate": 2.059617760588356e-06,
+      "loss": 0.8218,
       "step": 28131
     },
     {
-      "epoch": 0.7726236576858642,
+      "epoch": 0.7982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 2.5912915051626118e-06,
-      "loss": 0.8223,
+      "learning_rate": 2.059059116134323e-06,
+      "loss": 0.8971,
       "step": 28132
     },
     {
-      "epoch": 0.7726511219137074,
+      "epoch": 0.7983257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 2.590694093254228e-06,
-      "loss": 0.7812,
+      "learning_rate": 2.0585005387577784e-06,
+      "loss": 0.75,
       "step": 28133
     },
     {
-      "epoch": 0.7726785861415506,
+      "epoch": 0.7983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 2.590096739971145e-06,
-      "loss": 0.8938,
+      "learning_rate": 2.057942028463433e-06,
+      "loss": 0.8324,
       "step": 28134
     },
     {
-      "epoch": 0.7727060503693939,
+      "epoch": 0.7983825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 2.589499445318088e-06,
-      "loss": 0.8765,
+      "learning_rate": 2.0573835852560086e-06,
+      "loss": 0.8337,
       "step": 28135
     },
     {
-      "epoch": 0.772733514597237,
+      "epoch": 0.798410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.5889022092997774e-06,
-      "loss": 0.8424,
+      "learning_rate": 2.056825209140224e-06,
+      "loss": 0.8413,
       "step": 28136
     },
     {
-      "epoch": 0.7727609788250803,
+      "epoch": 0.7984392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 2.5883050319209446e-06,
-      "loss": 0.7737,
+      "learning_rate": 2.0562669001207923e-06,
+      "loss": 0.7867,
       "step": 28137
     },
     {
-      "epoch": 0.7727884430529236,
+      "epoch": 0.7984676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 2.5877079131863125e-06,
-      "loss": 0.7806,
+      "learning_rate": 2.055708658202433e-06,
+      "loss": 0.7985,
       "step": 28138
     },
     {
-      "epoch": 0.7728159072807668,
+      "epoch": 0.7984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 2.587110853100606e-06,
-      "loss": 0.8642,
+      "learning_rate": 2.0551504833898582e-06,
+      "loss": 0.885,
       "step": 28139
     },
     {
-      "epoch": 0.77284337150861,
+      "epoch": 0.7985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 2.586513851668554e-06,
-      "loss": 0.8086,
+      "learning_rate": 2.0545923756877837e-06,
+      "loss": 0.8052,
       "step": 28140
     },
     {
-      "epoch": 0.7728708357364533,
+      "epoch": 0.7985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 2.5859169088948734e-06,
-      "loss": 0.7516,
+      "learning_rate": 2.0540343351009272e-06,
+      "loss": 0.8732,
       "step": 28141
     },
     {
-      "epoch": 0.7728982999642965,
+      "epoch": 0.7985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 2.58532002478429e-06,
-      "loss": 0.7892,
+      "learning_rate": 2.053476361633997e-06,
+      "loss": 0.8627,
       "step": 28142
     },
     {
-      "epoch": 0.7729257641921398,
+      "epoch": 0.7986095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 2.5847231993415288e-06,
-      "loss": 0.8802,
+      "learning_rate": 2.05291845529171e-06,
+      "loss": 0.9613,
       "step": 28143
     },
     {
-      "epoch": 0.7729532284199829,
+      "epoch": 0.7986379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 2.5841264325713078e-06,
-      "loss": 0.8832,
+      "learning_rate": 2.0523606160787803e-06,
+      "loss": 0.8018,
       "step": 28144
     },
     {
-      "epoch": 0.7729806926478262,
+      "epoch": 0.7986662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 2.583529724478351e-06,
-      "loss": 0.8291,
+      "learning_rate": 2.051802843999913e-06,
+      "loss": 0.8622,
       "step": 28145
     },
     {
-      "epoch": 0.7730081568756695,
+      "epoch": 0.798694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 2.5829330750673833e-06,
-      "loss": 0.8943,
+      "learning_rate": 2.0512451390598254e-06,
+      "loss": 0.8751,
       "step": 28146
     },
     {
-      "epoch": 0.7730356211035126,
+      "epoch": 0.7987230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 2.5823364843431187e-06,
-      "loss": 0.875,
+      "learning_rate": 2.050687501263231e-06,
+      "loss": 0.7453,
       "step": 28147
     },
     {
-      "epoch": 0.7730630853313559,
+      "epoch": 0.7987514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 2.5817399523102803e-06,
-      "loss": 0.9633,
+      "learning_rate": 2.050129930614835e-06,
+      "loss": 0.8897,
       "step": 28148
     },
     {
-      "epoch": 0.7730905495591991,
+      "epoch": 0.7987797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 2.58114347897359e-06,
-      "loss": 0.7016,
+      "learning_rate": 2.0495724271193472e-06,
+      "loss": 0.8199,
       "step": 28149
     },
     {
-      "epoch": 0.7731180137870424,
+      "epoch": 0.7988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 2.5805470643377694e-06,
-      "loss": 0.8615,
+      "learning_rate": 2.0490149907814825e-06,
+      "loss": 0.8038,
       "step": 28150
     },
     {
-      "epoch": 0.7731454780148856,
+      "epoch": 0.7988365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 2.5799507084075315e-06,
-      "loss": 0.8937,
+      "learning_rate": 2.0484576216059414e-06,
+      "loss": 0.8394,
       "step": 28151
     },
     {
-      "epoch": 0.7731729422427288,
+      "epoch": 0.7988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 2.5793544111876e-06,
-      "loss": 0.9008,
+      "learning_rate": 2.0479003195974376e-06,
+      "loss": 0.8398,
       "step": 28152
     },
     {
-      "epoch": 0.7732004064705721,
+      "epoch": 0.7988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 2.578758172682689e-06,
-      "loss": 0.778,
+      "learning_rate": 2.04734308476068e-06,
+      "loss": 0.7646,
       "step": 28153
     },
     {
-      "epoch": 0.7732278706984154,
+      "epoch": 0.798921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 2.5781619928975177e-06,
-      "loss": 0.7952,
+      "learning_rate": 2.046785917100369e-06,
+      "loss": 0.6956,
       "step": 28154
     },
     {
-      "epoch": 0.7732553349262585,
+      "epoch": 0.798950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 2.5775658718368034e-06,
-      "loss": 0.7928,
+      "learning_rate": 2.0462288166212164e-06,
+      "loss": 0.8389,
       "step": 28155
     },
     {
-      "epoch": 0.7732827991541018,
+      "epoch": 0.7989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 2.576969809505263e-06,
-      "loss": 0.7824,
+      "learning_rate": 2.0456717833279284e-06,
+      "loss": 0.8366,
       "step": 28156
     },
     {
-      "epoch": 0.773310263381945,
+      "epoch": 0.7990068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 2.576373805907616e-06,
-      "loss": 0.8089,
+      "learning_rate": 2.0451148172252057e-06,
+      "loss": 0.7478,
       "step": 28157
     },
     {
-      "epoch": 0.7733377276097883,
+      "epoch": 0.7990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 2.575777861048572e-06,
-      "loss": 0.927,
+      "learning_rate": 2.0445579183177554e-06,
+      "loss": 0.8315,
       "step": 28158
     },
     {
-      "epoch": 0.7733651918376315,
+      "epoch": 0.7990635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 2.575181974932852e-06,
-      "loss": 0.8367,
+      "learning_rate": 2.0440010866102844e-06,
+      "loss": 0.7617,
       "step": 28159
     },
     {
-      "epoch": 0.7733926560654747,
+      "epoch": 0.7990919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 2.5745861475651668e-06,
-      "loss": 0.7778,
+      "learning_rate": 2.0434443221074896e-06,
+      "loss": 0.8612,
       "step": 28160
     },
     {
-      "epoch": 0.773420120293318,
+      "epoch": 0.7991203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 2.5739903789502307e-06,
-      "loss": 0.8779,
+      "learning_rate": 2.042887624814083e-06,
+      "loss": 0.845,
       "step": 28161
     },
     {
-      "epoch": 0.7734475845211611,
+      "epoch": 0.7991486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 2.57339466909276e-06,
-      "loss": 0.774,
+      "learning_rate": 2.0423309947347603e-06,
+      "loss": 0.8556,
       "step": 28162
     },
     {
-      "epoch": 0.7734750487490044,
+      "epoch": 0.7991770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 2.5727990179974705e-06,
-      "loss": 0.7382,
+      "learning_rate": 2.041774431874225e-06,
+      "loss": 0.843,
       "step": 28163
     },
     {
-      "epoch": 0.7735025129768477,
+      "epoch": 0.799205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 2.57220342566907e-06,
-      "loss": 0.7714,
+      "learning_rate": 2.041217936237181e-06,
+      "loss": 0.7334,
       "step": 28164
     },
     {
-      "epoch": 0.7735299772046909,
+      "epoch": 0.799233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 2.5716078921122724e-06,
-      "loss": 0.8281,
+      "learning_rate": 2.040661507828324e-06,
+      "loss": 0.7235,
       "step": 28165
     },
     {
-      "epoch": 0.7735574414325341,
+      "epoch": 0.7992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 2.5710124173317897e-06,
-      "loss": 0.7837,
+      "learning_rate": 2.040105146652358e-06,
+      "loss": 0.8637,
       "step": 28166
     },
     {
-      "epoch": 0.7735849056603774,
+      "epoch": 0.7992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 2.570417001332338e-06,
-      "loss": 0.894,
+      "learning_rate": 2.039548852713984e-06,
+      "loss": 0.7504,
       "step": 28167
     },
     {
-      "epoch": 0.7736123698882206,
+      "epoch": 0.7993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 2.569821644118623e-06,
-      "loss": 0.8049,
+      "learning_rate": 2.038992626017896e-06,
+      "loss": 0.7665,
       "step": 28168
     },
     {
-      "epoch": 0.7736398341160639,
+      "epoch": 0.7993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 2.569226345695359e-06,
-      "loss": 0.8277,
+      "learning_rate": 2.038436466568795e-06,
+      "loss": 0.7969,
       "step": 28169
     },
     {
-      "epoch": 0.773667298343907,
+      "epoch": 0.7993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 2.568631106067253e-06,
-      "loss": 0.8741,
+      "learning_rate": 2.0378803743713826e-06,
+      "loss": 0.8026,
       "step": 28170
     },
     {
-      "epoch": 0.7736947625717503,
+      "epoch": 0.7994040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 2.5680359252390153e-06,
-      "loss": 0.886,
+      "learning_rate": 2.0373243494303483e-06,
+      "loss": 0.8712,
       "step": 28171
     },
     {
-      "epoch": 0.7737222267995936,
+      "epoch": 0.7994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 2.5674408032153565e-06,
-      "loss": 0.7758,
+      "learning_rate": 2.0367683917503945e-06,
+      "loss": 0.7891,
       "step": 28172
     },
     {
-      "epoch": 0.7737496910274367,
+      "epoch": 0.799460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 2.566845740000985e-06,
-      "loss": 0.8201,
+      "learning_rate": 2.0362125013362187e-06,
+      "loss": 0.7601,
       "step": 28173
     },
     {
-      "epoch": 0.77377715525528,
+      "epoch": 0.7994892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 2.566250735600613e-06,
-      "loss": 0.8239,
+      "learning_rate": 2.0356566781925103e-06,
+      "loss": 0.8078,
       "step": 28174
     },
     {
-      "epoch": 0.7738046194831232,
+      "epoch": 0.7995175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 2.5656557900189437e-06,
-      "loss": 0.7992,
+      "learning_rate": 2.0351009223239693e-06,
+      "loss": 0.892,
       "step": 28175
     },
     {
-      "epoch": 0.7738320837109665,
+      "epoch": 0.7995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 2.5650609032606832e-06,
-      "loss": 0.8187,
+      "learning_rate": 2.034545233735288e-06,
+      "loss": 0.7358,
       "step": 28176
     },
     {
-      "epoch": 0.7738595479388097,
+      "epoch": 0.7995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 2.5644660753305407e-06,
-      "loss": 0.8314,
+      "learning_rate": 2.033989612431161e-06,
+      "loss": 0.8436,
       "step": 28177
     },
     {
-      "epoch": 0.7738870121666529,
+      "epoch": 0.7996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 2.5638713062332233e-06,
-      "loss": 0.7794,
+      "learning_rate": 2.033434058416285e-06,
+      "loss": 0.8746,
       "step": 28178
     },
     {
-      "epoch": 0.7739144763944962,
+      "epoch": 0.7996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 2.563276595973436e-06,
-      "loss": 0.8517,
+      "learning_rate": 2.032878571695347e-06,
+      "loss": 0.9033,
       "step": 28179
     },
     {
-      "epoch": 0.7739419406223395,
+      "epoch": 0.7996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 2.5626819445558846e-06,
-      "loss": 0.8242,
+      "learning_rate": 2.032323152273041e-06,
+      "loss": 0.8784,
       "step": 28180
     },
     {
-      "epoch": 0.7739694048501826,
+      "epoch": 0.7996878547105561,
       "grad_norm": 0.0,
-      "learning_rate": 2.5620873519852784e-06,
-      "loss": 0.7983,
+      "learning_rate": 2.031767800154063e-06,
+      "loss": 0.7133,
       "step": 28181
     },
     {
-      "epoch": 0.7739968690780259,
+      "epoch": 0.7997162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 2.5614928182663156e-06,
-      "loss": 0.865,
+      "learning_rate": 2.0312125153430974e-06,
+      "loss": 0.8566,
       "step": 28182
     },
     {
-      "epoch": 0.7740243333058691,
+      "epoch": 0.799744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 2.5608983434037047e-06,
-      "loss": 0.9156,
+      "learning_rate": 2.030657297844838e-06,
+      "loss": 0.7455,
       "step": 28183
     },
     {
-      "epoch": 0.7740517975337123,
+      "epoch": 0.7997729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 2.5603039274021456e-06,
-      "loss": 0.8887,
+      "learning_rate": 2.0301021476639782e-06,
+      "loss": 0.8058,
       "step": 28184
     },
     {
-      "epoch": 0.7740792617615556,
+      "epoch": 0.7998013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 2.559709570266343e-06,
-      "loss": 0.9138,
+      "learning_rate": 2.0295470648051995e-06,
+      "loss": 0.8438,
       "step": 28185
     },
     {
-      "epoch": 0.7741067259893988,
+      "epoch": 0.7998297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 2.5591152720010013e-06,
-      "loss": 0.9172,
+      "learning_rate": 2.0289920492731963e-06,
+      "loss": 0.7278,
       "step": 28186
     },
     {
-      "epoch": 0.7741341902172421,
+      "epoch": 0.7998581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 2.5585210326108245e-06,
-      "loss": 0.7924,
+      "learning_rate": 2.0284371010726578e-06,
+      "loss": 0.8385,
       "step": 28187
     },
     {
-      "epoch": 0.7741616544450852,
+      "epoch": 0.7998864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 2.5579268521005095e-06,
-      "loss": 0.8047,
+      "learning_rate": 2.027882220208267e-06,
+      "loss": 0.7923,
       "step": 28188
     },
     {
-      "epoch": 0.7741891186729285,
+      "epoch": 0.7999148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 2.55733273047476e-06,
-      "loss": 0.77,
+      "learning_rate": 2.0273274066847136e-06,
+      "loss": 0.8201,
       "step": 28189
     },
     {
-      "epoch": 0.7742165829007718,
+      "epoch": 0.7999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 2.5567386677382788e-06,
-      "loss": 0.7779,
+      "learning_rate": 2.0267726605066864e-06,
+      "loss": 0.759,
       "step": 28190
     },
     {
-      "epoch": 0.7742440471286149,
+      "epoch": 0.7999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 2.5561446638957622e-06,
-      "loss": 0.7827,
+      "learning_rate": 2.026217981678864e-06,
+      "loss": 0.8291,
       "step": 28191
     },
     {
-      "epoch": 0.7742715113564582,
+      "epoch": 0.8,
       "grad_norm": 0.0,
-      "learning_rate": 2.5555507189519125e-06,
-      "loss": 0.7712,
+      "learning_rate": 2.0256633702059415e-06,
+      "loss": 0.8047,
       "step": 28192
     },
     {
-      "epoch": 0.7742989755843015,
+      "epoch": 0.8000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 2.554956832911432e-06,
-      "loss": 0.8035,
+      "learning_rate": 2.0251088260925967e-06,
+      "loss": 0.7194,
       "step": 28193
     },
     {
-      "epoch": 0.7743264398121447,
+      "epoch": 0.8000567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 2.5543630057790137e-06,
-      "loss": 0.8654,
+      "learning_rate": 2.024554349343516e-06,
+      "loss": 0.8768,
       "step": 28194
     },
     {
-      "epoch": 0.7743539040399879,
+      "epoch": 0.8000851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 2.553769237559359e-06,
-      "loss": 0.7821,
+      "learning_rate": 2.023999939963386e-06,
+      "loss": 0.7004,
       "step": 28195
     },
     {
-      "epoch": 0.7743813682678311,
+      "epoch": 0.8001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 2.553175528257168e-06,
-      "loss": 0.8956,
+      "learning_rate": 2.0234455979568847e-06,
+      "loss": 0.7628,
       "step": 28196
     },
     {
-      "epoch": 0.7744088324956744,
+      "epoch": 0.8001418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 2.5525818778771362e-06,
-      "loss": 0.8712,
+      "learning_rate": 2.0228913233286963e-06,
+      "loss": 0.7894,
       "step": 28197
     },
     {
-      "epoch": 0.7744362967235177,
+      "epoch": 0.8001702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 2.551988286423964e-06,
-      "loss": 0.7838,
+      "learning_rate": 2.022337116083507e-06,
+      "loss": 0.8619,
       "step": 28198
     },
     {
-      "epoch": 0.7744637609513608,
+      "epoch": 0.8001986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 2.5513947539023465e-06,
-      "loss": 0.8075,
+      "learning_rate": 2.021782976225991e-06,
+      "loss": 0.8398,
       "step": 28199
     },
     {
-      "epoch": 0.7744912251792041,
+      "epoch": 0.8002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 2.5508012803169756e-06,
-      "loss": 0.8194,
+      "learning_rate": 2.0212289037608335e-06,
+      "loss": 0.7888,
       "step": 28200
     },
     {
-      "epoch": 0.7745186894070473,
+      "epoch": 0.800255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 2.5502078656725503e-06,
-      "loss": 0.7543,
+      "learning_rate": 2.020674898692716e-06,
+      "loss": 0.7026,
       "step": 28201
     },
     {
-      "epoch": 0.7745461536348905,
+      "epoch": 0.800283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 2.5496145099737657e-06,
-      "loss": 0.7632,
+      "learning_rate": 2.0201209610263153e-06,
+      "loss": 0.7733,
       "step": 28202
     },
     {
-      "epoch": 0.7745736178627338,
+      "epoch": 0.8003121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 2.549021213225318e-06,
-      "loss": 0.7937,
+      "learning_rate": 2.0195670907663102e-06,
+      "loss": 0.7694,
       "step": 28203
     },
     {
-      "epoch": 0.774601082090577,
+      "epoch": 0.8003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 2.5484279754318995e-06,
-      "loss": 0.8601,
+      "learning_rate": 2.019013287917384e-06,
+      "loss": 0.7461,
       "step": 28204
     },
     {
-      "epoch": 0.7746285463184203,
+      "epoch": 0.8003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 2.54783479659821e-06,
-      "loss": 0.8161,
+      "learning_rate": 2.0184595524842077e-06,
+      "loss": 0.7274,
       "step": 28205
     },
     {
-      "epoch": 0.7746560105462635,
+      "epoch": 0.8003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 2.5472416767289345e-06,
-      "loss": 0.7068,
+      "learning_rate": 2.0179058844714627e-06,
+      "loss": 0.8725,
       "step": 28206
     },
     {
-      "epoch": 0.7746834747741067,
+      "epoch": 0.8004256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 2.546648615828772e-06,
-      "loss": 0.7569,
+      "learning_rate": 2.0173522838838265e-06,
+      "loss": 0.9259,
       "step": 28207
     },
     {
-      "epoch": 0.77471093900195,
+      "epoch": 0.8004540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 2.5460556139024106e-06,
-      "loss": 0.8553,
+      "learning_rate": 2.0167987507259735e-06,
+      "loss": 0.8276,
       "step": 28208
     },
     {
-      "epoch": 0.7747384032297931,
+      "epoch": 0.8004824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 2.545462670954545e-06,
-      "loss": 0.846,
+      "learning_rate": 2.0162452850025827e-06,
+      "loss": 0.7975,
       "step": 28209
     },
     {
-      "epoch": 0.7747658674576364,
+      "epoch": 0.8005107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 2.5448697869898654e-06,
-      "loss": 0.816,
+      "learning_rate": 2.015691886718325e-06,
+      "loss": 0.7948,
       "step": 28210
     },
     {
-      "epoch": 0.7747933316854797,
+      "epoch": 0.8005391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 2.5442769620130668e-06,
-      "loss": 0.8311,
+      "learning_rate": 2.015138555877877e-06,
+      "loss": 0.8011,
       "step": 28211
     },
     {
-      "epoch": 0.7748207959133229,
+      "epoch": 0.8005675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 2.543684196028835e-06,
-      "loss": 0.8497,
+      "learning_rate": 2.0145852924859146e-06,
+      "loss": 0.8547,
       "step": 28212
     },
     {
-      "epoch": 0.7748482601411661,
+      "epoch": 0.8005959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 2.543091489041861e-06,
-      "loss": 0.7314,
+      "learning_rate": 2.014032096547107e-06,
+      "loss": 0.9126,
       "step": 28213
     },
     {
-      "epoch": 0.7748757243690093,
+      "epoch": 0.8006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 2.5424988410568394e-06,
-      "loss": 0.8352,
+      "learning_rate": 2.013478968066128e-06,
+      "loss": 0.7729,
       "step": 28214
     },
     {
-      "epoch": 0.7749031885968526,
+      "epoch": 0.8006526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 2.541906252078453e-06,
-      "loss": 0.8147,
+      "learning_rate": 2.012925907047656e-06,
+      "loss": 0.8805,
       "step": 28215
     },
     {
-      "epoch": 0.7749306528246959,
+      "epoch": 0.8006810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 2.541313722111396e-06,
-      "loss": 0.8063,
+      "learning_rate": 2.012372913496353e-06,
+      "loss": 0.7643,
       "step": 28216
     },
     {
-      "epoch": 0.774958117052539,
+      "epoch": 0.8007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 2.5407212511603507e-06,
-      "loss": 0.8598,
+      "learning_rate": 2.0118199874168966e-06,
+      "loss": 0.8506,
       "step": 28217
     },
     {
-      "epoch": 0.7749855812803823,
+      "epoch": 0.8007377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 2.540128839230008e-06,
-      "loss": 0.7558,
+      "learning_rate": 2.0112671288139583e-06,
+      "loss": 0.8684,
       "step": 28218
     },
     {
-      "epoch": 0.7750130455082256,
+      "epoch": 0.8007661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 2.539536486325057e-06,
-      "loss": 0.9163,
+      "learning_rate": 2.0107143376922024e-06,
+      "loss": 0.8085,
       "step": 28219
     },
     {
-      "epoch": 0.7750405097360688,
+      "epoch": 0.800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 2.5389441924501825e-06,
-      "loss": 0.8689,
+      "learning_rate": 2.010161614056302e-06,
+      "loss": 0.8309,
       "step": 28220
     },
     {
-      "epoch": 0.775067973963912,
+      "epoch": 0.8008229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 2.538351957610071e-06,
-      "loss": 0.7947,
+      "learning_rate": 2.009608957910929e-06,
+      "loss": 0.7286,
       "step": 28221
     },
     {
-      "epoch": 0.7750954381917552,
+      "epoch": 0.8008513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 2.5377597818094134e-06,
-      "loss": 0.7876,
+      "learning_rate": 2.009056369260742e-06,
+      "loss": 0.8369,
       "step": 28222
     },
     {
-      "epoch": 0.7751229024195985,
+      "epoch": 0.8008796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 2.53716766505289e-06,
-      "loss": 0.8203,
+      "learning_rate": 2.0085038481104204e-06,
+      "loss": 0.75,
       "step": 28223
     },
     {
-      "epoch": 0.7751503666474417,
+      "epoch": 0.8009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 2.5365756073451843e-06,
-      "loss": 0.7794,
+      "learning_rate": 2.0079513944646236e-06,
+      "loss": 0.6989,
       "step": 28224
     },
     {
-      "epoch": 0.7751778308752849,
+      "epoch": 0.8009364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 2.535983608690984e-06,
-      "loss": 0.8609,
+      "learning_rate": 2.0073990083280214e-06,
+      "loss": 0.8657,
       "step": 28225
     },
     {
-      "epoch": 0.7752052951031282,
+      "epoch": 0.8009648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 2.535391669094971e-06,
-      "loss": 0.7924,
+      "learning_rate": 2.0068466897052806e-06,
+      "loss": 0.8586,
       "step": 28226
     },
     {
-      "epoch": 0.7752327593309714,
+      "epoch": 0.8009931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 2.5347997885618327e-06,
-      "loss": 0.7785,
+      "learning_rate": 2.006294438601063e-06,
+      "loss": 0.8293,
       "step": 28227
     },
     {
-      "epoch": 0.7752602235588146,
+      "epoch": 0.8010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 2.5342079670962518e-06,
-      "loss": 0.8479,
+      "learning_rate": 2.0057422550200355e-06,
+      "loss": 0.7447,
       "step": 28228
     },
     {
-      "epoch": 0.7752876877866579,
+      "epoch": 0.801049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 2.5336162047029066e-06,
-      "loss": 0.7704,
+      "learning_rate": 2.0051901389668647e-06,
+      "loss": 0.8622,
       "step": 28229
     },
     {
-      "epoch": 0.7753151520145011,
+      "epoch": 0.8010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 2.533024501386483e-06,
-      "loss": 0.8389,
+      "learning_rate": 2.0046380904462103e-06,
+      "loss": 0.7515,
       "step": 28230
     },
     {
-      "epoch": 0.7753426162423444,
+      "epoch": 0.8011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 2.532432857151663e-06,
-      "loss": 0.732,
+      "learning_rate": 2.0040861094627374e-06,
+      "loss": 0.8096,
       "step": 28231
     },
     {
-      "epoch": 0.7753700804701876,
+      "epoch": 0.8011350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 2.531841272003126e-06,
-      "loss": 0.8653,
+      "learning_rate": 2.003534196021111e-06,
+      "loss": 0.723,
       "step": 28232
     },
     {
-      "epoch": 0.7753975446980308,
+      "epoch": 0.8011634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 2.531249745945553e-06,
-      "loss": 0.7571,
+      "learning_rate": 2.0029823501259884e-06,
+      "loss": 0.8505,
       "step": 28233
     },
     {
-      "epoch": 0.7754250089258741,
+      "epoch": 0.8011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 2.5306582789836287e-06,
-      "loss": 0.811,
+      "learning_rate": 2.0024305717820326e-06,
+      "loss": 0.8176,
       "step": 28234
     },
     {
-      "epoch": 0.7754524731537172,
+      "epoch": 0.8012202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 2.5300668711220256e-06,
-      "loss": 0.8552,
+      "learning_rate": 2.0018788609939087e-06,
+      "loss": 0.7864,
       "step": 28235
     },
     {
-      "epoch": 0.7754799373815605,
+      "epoch": 0.8012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 2.5294755223654266e-06,
-      "loss": 0.8536,
+      "learning_rate": 2.0013272177662702e-06,
+      "loss": 0.7569,
       "step": 28236
     },
     {
-      "epoch": 0.7755074016094038,
+      "epoch": 0.8012769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 2.5288842327185127e-06,
-      "loss": 0.8227,
+      "learning_rate": 2.0007756421037784e-06,
+      "loss": 0.8979,
       "step": 28237
     },
     {
-      "epoch": 0.775534865837247,
+      "epoch": 0.801305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 2.528293002185962e-06,
-      "loss": 0.8769,
+      "learning_rate": 2.0002241340110983e-06,
+      "loss": 0.7587,
       "step": 28238
     },
     {
-      "epoch": 0.7755623300650902,
+      "epoch": 0.8013337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 2.52770183077245e-06,
-      "loss": 0.8396,
+      "learning_rate": 1.9996726934928834e-06,
+      "loss": 0.84,
       "step": 28239
     },
     {
-      "epoch": 0.7755897942929334,
+      "epoch": 0.8013620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 2.527110718482657e-06,
-      "loss": 0.8115,
+      "learning_rate": 1.9991213205537917e-06,
+      "loss": 0.7737,
       "step": 28240
     },
     {
-      "epoch": 0.7756172585207767,
+      "epoch": 0.8013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 2.5265196653212567e-06,
-      "loss": 0.8573,
+      "learning_rate": 1.9985700151984856e-06,
+      "loss": 0.721,
       "step": 28241
     },
     {
-      "epoch": 0.77564472274862,
+      "epoch": 0.8014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 2.5259286712929275e-06,
-      "loss": 0.8127,
+      "learning_rate": 1.998018777431615e-06,
+      "loss": 0.8646,
       "step": 28242
     },
     {
-      "epoch": 0.7756721869764631,
+      "epoch": 0.8014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 2.5253377364023456e-06,
-      "loss": 0.8414,
+      "learning_rate": 1.9974676072578405e-06,
+      "loss": 0.761,
       "step": 28243
     },
     {
-      "epoch": 0.7756996512043064,
+      "epoch": 0.8014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 2.524746860654188e-06,
-      "loss": 0.8049,
+      "learning_rate": 1.9969165046818184e-06,
+      "loss": 0.7226,
       "step": 28244
     },
     {
-      "epoch": 0.7757271154321497,
+      "epoch": 0.8015039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 2.5241560440531277e-06,
-      "loss": 0.863,
+      "learning_rate": 1.9963654697082e-06,
+      "loss": 0.8469,
       "step": 28245
     },
     {
-      "epoch": 0.7757545796599928,
+      "epoch": 0.8015323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 2.5235652866038442e-06,
-      "loss": 0.9006,
+      "learning_rate": 1.9958145023416443e-06,
+      "loss": 0.7163,
       "step": 28246
     },
     {
-      "epoch": 0.7757820438878361,
+      "epoch": 0.8015607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 2.522974588311008e-06,
-      "loss": 0.7906,
+      "learning_rate": 1.9952636025868012e-06,
+      "loss": 0.7665,
       "step": 28247
     },
     {
-      "epoch": 0.7758095081156793,
+      "epoch": 0.801589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 2.52238394917929e-06,
-      "loss": 0.8291,
+      "learning_rate": 1.994712770448327e-06,
+      "loss": 0.7884,
       "step": 28248
     },
     {
-      "epoch": 0.7758369723435226,
+      "epoch": 0.8016174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 2.521793369213367e-06,
-      "loss": 0.7463,
+      "learning_rate": 1.9941620059308753e-06,
+      "loss": 0.8087,
       "step": 28249
     },
     {
-      "epoch": 0.7758644365713658,
+      "epoch": 0.8016458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 2.521202848417912e-06,
-      "loss": 0.8631,
+      "learning_rate": 1.9936113090390953e-06,
+      "loss": 0.7946,
       "step": 28250
     },
     {
-      "epoch": 0.775891900799209,
+      "epoch": 0.8016742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 2.520612386797596e-06,
-      "loss": 0.7928,
+      "learning_rate": 1.9930606797776396e-06,
+      "loss": 0.8376,
       "step": 28251
     },
     {
-      "epoch": 0.7759193650270523,
+      "epoch": 0.8017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 2.520021984357096e-06,
-      "loss": 0.8647,
+      "learning_rate": 1.992510118151162e-06,
+      "loss": 0.8622,
       "step": 28252
     },
     {
-      "epoch": 0.7759468292548956,
+      "epoch": 0.8017309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 2.5194316411010767e-06,
-      "loss": 0.8434,
+      "learning_rate": 1.991959624164309e-06,
+      "loss": 0.8746,
       "step": 28253
     },
     {
-      "epoch": 0.7759742934827387,
+      "epoch": 0.8017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 2.518841357034212e-06,
-      "loss": 0.8432,
+      "learning_rate": 1.9914091978217354e-06,
+      "loss": 0.8061,
       "step": 28254
     },
     {
-      "epoch": 0.776001757710582,
+      "epoch": 0.8017877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 2.5182511321611746e-06,
-      "loss": 0.8371,
+      "learning_rate": 1.9908588391280893e-06,
+      "loss": 0.7234,
       "step": 28255
     },
     {
-      "epoch": 0.7760292219384252,
+      "epoch": 0.801816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 2.517660966486629e-06,
-      "loss": 0.8524,
+      "learning_rate": 1.990308548088017e-06,
+      "loss": 0.8158,
       "step": 28256
     },
     {
-      "epoch": 0.7760566861662684,
+      "epoch": 0.801844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.5170708600152493e-06,
-      "loss": 0.7969,
+      "learning_rate": 1.9897583247061667e-06,
+      "loss": 0.8067,
       "step": 28257
     },
     {
-      "epoch": 0.7760841503941117,
+      "epoch": 0.8018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 2.516480812751706e-06,
-      "loss": 0.8683,
+      "learning_rate": 1.9892081689871923e-06,
+      "loss": 0.8354,
       "step": 28258
     },
     {
-      "epoch": 0.7761116146219549,
+      "epoch": 0.8019012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 2.515890824700663e-06,
-      "loss": 0.7809,
+      "learning_rate": 1.9886580809357335e-06,
+      "loss": 0.8334,
       "step": 28259
     },
     {
-      "epoch": 0.7761390788497982,
+      "epoch": 0.8019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 2.5153008958667903e-06,
-      "loss": 0.7922,
+      "learning_rate": 1.9881080605564406e-06,
+      "loss": 0.6847,
       "step": 28260
     },
     {
-      "epoch": 0.7761665430776413,
+      "epoch": 0.8019580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 2.514711026254756e-06,
-      "loss": 0.7807,
+      "learning_rate": 1.987558107853961e-06,
+      "loss": 0.8396,
       "step": 28261
     },
     {
-      "epoch": 0.7761940073054846,
+      "epoch": 0.8019863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 2.514121215869231e-06,
-      "loss": 0.7742,
+      "learning_rate": 1.9870082228329355e-06,
+      "loss": 0.8438,
       "step": 28262
     },
     {
-      "epoch": 0.7762214715333279,
+      "epoch": 0.8020147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 2.5135314647148754e-06,
-      "loss": 0.8219,
+      "learning_rate": 1.986458405498012e-06,
+      "loss": 0.7503,
       "step": 28263
     },
     {
-      "epoch": 0.776248935761171,
+      "epoch": 0.8020431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 2.512941772796362e-06,
-      "loss": 0.8624,
+      "learning_rate": 1.9859086558538377e-06,
+      "loss": 0.7804,
       "step": 28264
     },
     {
-      "epoch": 0.7762763999890143,
+      "epoch": 0.8020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 2.5123521401183503e-06,
-      "loss": 0.8407,
+      "learning_rate": 1.9853589739050494e-06,
+      "loss": 0.7085,
       "step": 28265
     },
     {
-      "epoch": 0.7763038642168576,
+      "epoch": 0.802099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 2.5117625666855095e-06,
-      "loss": 0.8818,
+      "learning_rate": 1.9848093596562945e-06,
+      "loss": 0.8141,
       "step": 28266
     },
     {
-      "epoch": 0.7763313284447008,
+      "epoch": 0.802128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 2.5111730525025034e-06,
-      "loss": 0.8736,
+      "learning_rate": 1.984259813112219e-06,
+      "loss": 0.6905,
       "step": 28267
     },
     {
-      "epoch": 0.776358792672544,
+      "epoch": 0.8021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 2.5105835975739976e-06,
-      "loss": 0.7838,
+      "learning_rate": 1.9837103342774543e-06,
+      "loss": 0.8422,
       "step": 28268
     },
     {
-      "epoch": 0.7763862569003872,
+      "epoch": 0.8021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 2.509994201904655e-06,
-      "loss": 0.975,
+      "learning_rate": 1.983160923156655e-06,
+      "loss": 0.8794,
       "step": 28269
     },
     {
-      "epoch": 0.7764137211282305,
+      "epoch": 0.8022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 2.509404865499143e-06,
-      "loss": 0.8829,
+      "learning_rate": 1.982611579754453e-06,
+      "loss": 0.871,
       "step": 28270
     },
     {
-      "epoch": 0.7764411853560738,
+      "epoch": 0.8022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 2.5088155883621212e-06,
-      "loss": 0.8487,
+      "learning_rate": 1.9820623040754915e-06,
+      "loss": 0.8439,
       "step": 28271
     },
     {
-      "epoch": 0.7764686495839169,
+      "epoch": 0.8022701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 2.508226370498249e-06,
-      "loss": 0.7409,
+      "learning_rate": 1.9815130961244135e-06,
+      "loss": 0.8263,
       "step": 28272
     },
     {
-      "epoch": 0.7764961138117602,
+      "epoch": 0.8022985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 2.5076372119121917e-06,
-      "loss": 0.915,
+      "learning_rate": 1.9809639559058513e-06,
+      "loss": 0.8385,
       "step": 28273
     },
     {
-      "epoch": 0.7765235780396034,
+      "epoch": 0.8023269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 2.507048112608611e-06,
-      "loss": 0.8066,
+      "learning_rate": 1.9804148834244465e-06,
+      "loss": 0.7914,
       "step": 28274
     },
     {
-      "epoch": 0.7765510422674466,
+      "epoch": 0.802355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 2.506459072592171e-06,
-      "loss": 0.812,
+      "learning_rate": 1.9798658786848414e-06,
+      "loss": 0.8269,
       "step": 28275
     },
     {
-      "epoch": 0.7765785064952899,
+      "epoch": 0.8023836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 2.5058700918675274e-06,
-      "loss": 0.7994,
+      "learning_rate": 1.979316941691667e-06,
+      "loss": 0.767,
       "step": 28276
     },
     {
-      "epoch": 0.7766059707231331,
+      "epoch": 0.8024120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 2.505281170439342e-06,
-      "loss": 0.8978,
+      "learning_rate": 1.9787680724495617e-06,
+      "loss": 0.8667,
       "step": 28277
     },
     {
-      "epoch": 0.7766334349509764,
+      "epoch": 0.8024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 2.504692308312275e-06,
-      "loss": 0.8308,
+      "learning_rate": 1.9782192709631677e-06,
+      "loss": 0.7616,
       "step": 28278
     },
     {
-      "epoch": 0.7766608991788196,
+      "epoch": 0.8024687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 2.504103505490989e-06,
-      "loss": 0.827,
+      "learning_rate": 1.977670537237112e-06,
+      "loss": 0.8263,
       "step": 28279
     },
     {
-      "epoch": 0.7766883634066628,
+      "epoch": 0.8024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 2.503514761980137e-06,
-      "loss": 0.7832,
+      "learning_rate": 1.9771218712760344e-06,
+      "loss": 0.8495,
       "step": 28280
     },
     {
-      "epoch": 0.7767158276345061,
+      "epoch": 0.8025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 2.5029260777843824e-06,
-      "loss": 0.8255,
+      "learning_rate": 1.976573273084571e-06,
+      "loss": 0.8228,
       "step": 28281
     },
     {
-      "epoch": 0.7767432918623492,
+      "epoch": 0.8025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 2.502337452908379e-06,
-      "loss": 0.8356,
+      "learning_rate": 1.976024742667352e-06,
+      "loss": 0.9527,
       "step": 28282
     },
     {
-      "epoch": 0.7767707560901925,
+      "epoch": 0.8025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 2.501748887356784e-06,
-      "loss": 0.9479,
+      "learning_rate": 1.975476280029013e-06,
+      "loss": 0.9011,
       "step": 28283
     },
     {
-      "epoch": 0.7767982203180358,
+      "epoch": 0.8026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 2.5011603811342577e-06,
-      "loss": 0.7674,
+      "learning_rate": 1.974927885174186e-06,
+      "loss": 0.8233,
       "step": 28284
     },
     {
-      "epoch": 0.776825684545879,
+      "epoch": 0.802639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 2.500571934245456e-06,
-      "loss": 0.8727,
+      "learning_rate": 1.974379558107503e-06,
+      "loss": 0.7814,
       "step": 28285
     },
     {
-      "epoch": 0.7768531487737222,
+      "epoch": 0.8026674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 2.4999835466950363e-06,
-      "loss": 0.7005,
+      "learning_rate": 1.9738312988336006e-06,
+      "loss": 0.8143,
       "step": 28286
     },
     {
-      "epoch": 0.7768806130015654,
+      "epoch": 0.8026958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 2.499395218487648e-06,
-      "loss": 0.8053,
+      "learning_rate": 1.9732831073571036e-06,
+      "loss": 0.6979,
       "step": 28287
     },
     {
-      "epoch": 0.7769080772294087,
+      "epoch": 0.8027241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 2.4988069496279545e-06,
-      "loss": 0.8007,
+      "learning_rate": 1.972734983682644e-06,
+      "loss": 0.8828,
       "step": 28288
     },
     {
-      "epoch": 0.776935541457252,
+      "epoch": 0.8027525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 2.4982187401206025e-06,
-      "loss": 0.7474,
+      "learning_rate": 1.9721869278148563e-06,
+      "loss": 0.8702,
       "step": 28289
     },
     {
-      "epoch": 0.7769630056850951,
+      "epoch": 0.8027809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 2.4976305899702503e-06,
-      "loss": 0.7819,
+      "learning_rate": 1.9716389397583636e-06,
+      "loss": 0.7965,
       "step": 28290
     },
     {
-      "epoch": 0.7769904699129384,
+      "epoch": 0.8028093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 2.49704249918155e-06,
-      "loss": 0.8572,
+      "learning_rate": 1.971091019517797e-06,
+      "loss": 0.8712,
       "step": 28291
     },
     {
-      "epoch": 0.7770179341407817,
+      "epoch": 0.8028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 2.4964544677591574e-06,
-      "loss": 0.7738,
+      "learning_rate": 1.970543167097789e-06,
+      "loss": 0.9088,
       "step": 28292
     },
     {
-      "epoch": 0.7770453983686249,
+      "epoch": 0.8028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 2.495866495707725e-06,
-      "loss": 0.8324,
+      "learning_rate": 1.9699953825029607e-06,
+      "loss": 0.7793,
       "step": 28293
     },
     {
-      "epoch": 0.7770728625964681,
+      "epoch": 0.802894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 2.4952785830319014e-06,
-      "loss": 0.7828,
+      "learning_rate": 1.9694476657379425e-06,
+      "loss": 0.8291,
       "step": 28294
     },
     {
-      "epoch": 0.7771003268243113,
+      "epoch": 0.8029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 2.494690729736344e-06,
-      "loss": 0.87,
+      "learning_rate": 1.9689000168073634e-06,
+      "loss": 0.7697,
       "step": 28295
     },
     {
-      "epoch": 0.7771277910521546,
+      "epoch": 0.8029511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 2.4941029358256976e-06,
-      "loss": 0.814,
+      "learning_rate": 1.9683524357158445e-06,
+      "loss": 0.7923,
       "step": 28296
     },
     {
-      "epoch": 0.7771552552799978,
+      "epoch": 0.8029795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 2.493515201304616e-06,
-      "loss": 0.8304,
+      "learning_rate": 1.967804922468014e-06,
+      "loss": 0.7838,
       "step": 28297
     },
     {
-      "epoch": 0.777182719507841,
+      "epoch": 0.8030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 2.4929275261777507e-06,
-      "loss": 0.8471,
+      "learning_rate": 1.967257477068495e-06,
+      "loss": 0.8791,
       "step": 28298
     },
     {
-      "epoch": 0.7772101837356843,
+      "epoch": 0.8030363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 2.4923399104497525e-06,
-      "loss": 0.7352,
+      "learning_rate": 1.9667100995219147e-06,
+      "loss": 0.7813,
       "step": 28299
     },
     {
-      "epoch": 0.7772376479635275,
+      "epoch": 0.8030646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 2.4917523541252675e-06,
-      "loss": 0.8016,
+      "learning_rate": 1.966162789832897e-06,
+      "loss": 0.772,
       "step": 28300
     },
     {
-      "epoch": 0.7772651121913707,
+      "epoch": 0.8030930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 2.491164857208945e-06,
-      "loss": 0.8971,
+      "learning_rate": 1.9656155480060613e-06,
+      "loss": 0.8652,
       "step": 28301
     },
     {
-      "epoch": 0.777292576419214,
+      "epoch": 0.8031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 2.4905774197054356e-06,
-      "loss": 0.7565,
+      "learning_rate": 1.965068374046032e-06,
+      "loss": 0.8086,
       "step": 28302
     },
     {
-      "epoch": 0.7773200406470572,
+      "epoch": 0.803149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 2.489990041619389e-06,
-      "loss": 0.9149,
+      "learning_rate": 1.9645212679574345e-06,
+      "loss": 0.7845,
       "step": 28303
     },
     {
-      "epoch": 0.7773475048749005,
+      "epoch": 0.803178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 2.4894027229554474e-06,
-      "loss": 0.7483,
+      "learning_rate": 1.963974229744884e-06,
+      "loss": 0.788,
       "step": 28304
     },
     {
-      "epoch": 0.7773749691027437,
+      "epoch": 0.8032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 2.4888154637182637e-06,
-      "loss": 0.8568,
+      "learning_rate": 1.963427259413003e-06,
+      "loss": 0.7694,
       "step": 28305
     },
     {
-      "epoch": 0.7774024333305869,
+      "epoch": 0.8032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 2.48822826391248e-06,
-      "loss": 0.8429,
+      "learning_rate": 1.9628803569664178e-06,
+      "loss": 0.8,
       "step": 28306
     },
     {
-      "epoch": 0.7774298975584302,
+      "epoch": 0.8032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 2.487641123542742e-06,
-      "loss": 0.8411,
+      "learning_rate": 1.9623335224097404e-06,
+      "loss": 0.8514,
       "step": 28307
     },
     {
-      "epoch": 0.7774573617862733,
+      "epoch": 0.8032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 2.487054042613699e-06,
-      "loss": 0.8074,
+      "learning_rate": 1.9617867557475933e-06,
+      "loss": 0.8063,
       "step": 28308
     },
     {
-      "epoch": 0.7774848260141166,
+      "epoch": 0.8033200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 2.4864670211299937e-06,
-      "loss": 0.8311,
+      "learning_rate": 1.961240056984597e-06,
+      "loss": 0.7997,
       "step": 28309
     },
     {
-      "epoch": 0.7775122902419599,
+      "epoch": 0.8033484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 2.485880059096275e-06,
-      "loss": 0.7658,
+      "learning_rate": 1.960693426125364e-06,
+      "loss": 0.7519,
       "step": 28310
     },
     {
-      "epoch": 0.7775397544698031,
+      "epoch": 0.8033768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 2.4852931565171847e-06,
-      "loss": 0.897,
+      "learning_rate": 1.960146863174516e-06,
+      "loss": 0.7844,
       "step": 28311
     },
     {
-      "epoch": 0.7775672186976463,
+      "epoch": 0.8034052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 2.4847063133973614e-06,
-      "loss": 0.792,
+      "learning_rate": 1.9596003681366716e-06,
+      "loss": 0.8421,
       "step": 28312
     },
     {
-      "epoch": 0.7775946829254895,
+      "epoch": 0.8034335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 2.4841195297414546e-06,
-      "loss": 0.8528,
+      "learning_rate": 1.959053941016441e-06,
+      "loss": 0.7446,
       "step": 28313
     },
     {
-      "epoch": 0.7776221471533328,
+      "epoch": 0.8034619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 2.4835328055541043e-06,
-      "loss": 0.7957,
+      "learning_rate": 1.958507581818444e-06,
+      "loss": 0.8226,
       "step": 28314
     },
     {
-      "epoch": 0.777649611381176,
+      "epoch": 0.8034903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 2.4829461408399548e-06,
-      "loss": 0.8065,
+      "learning_rate": 1.9579612905472944e-06,
+      "loss": 0.8708,
       "step": 28315
     },
     {
-      "epoch": 0.7776770756090192,
+      "epoch": 0.8035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 2.482359535603647e-06,
-      "loss": 0.8098,
+      "learning_rate": 1.9574150672076076e-06,
+      "loss": 0.767,
       "step": 28316
     },
     {
-      "epoch": 0.7777045398368625,
+      "epoch": 0.8035471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 2.481772989849824e-06,
-      "loss": 0.8095,
+      "learning_rate": 1.9568689118039997e-06,
+      "loss": 0.8082,
       "step": 28317
     },
     {
-      "epoch": 0.7777320040647058,
+      "epoch": 0.8035754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 2.481186503583124e-06,
-      "loss": 0.8006,
+      "learning_rate": 1.9563228243410794e-06,
+      "loss": 0.7885,
       "step": 28318
     },
     {
-      "epoch": 0.7777594682925489,
+      "epoch": 0.8036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 2.4806000768081917e-06,
-      "loss": 0.7877,
+      "learning_rate": 1.9557768048234606e-06,
+      "loss": 0.746,
       "step": 28319
     },
     {
-      "epoch": 0.7777869325203922,
+      "epoch": 0.8036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 2.4800137095296616e-06,
-      "loss": 0.8326,
+      "learning_rate": 1.9552308532557607e-06,
+      "loss": 0.8171,
       "step": 28320
     },
     {
-      "epoch": 0.7778143967482354,
+      "epoch": 0.803660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 2.479427401752176e-06,
-      "loss": 0.8552,
+      "learning_rate": 1.954684969642584e-06,
+      "loss": 0.7519,
       "step": 28321
     },
     {
-      "epoch": 0.7778418609760787,
+      "epoch": 0.803688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 2.4788411534803735e-06,
-      "loss": 0.8909,
+      "learning_rate": 1.9541391539885456e-06,
+      "loss": 0.7892,
       "step": 28322
     },
     {
-      "epoch": 0.7778693252039219,
+      "epoch": 0.8037173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 2.478254964718897e-06,
-      "loss": 0.8508,
+      "learning_rate": 1.9535934062982575e-06,
+      "loss": 0.8204,
       "step": 28323
     },
     {
-      "epoch": 0.7778967894317651,
+      "epoch": 0.8037457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 2.4776688354723764e-06,
-      "loss": 0.8535,
+      "learning_rate": 1.9530477265763258e-06,
+      "loss": 0.8042,
       "step": 28324
     },
     {
-      "epoch": 0.7779242536596084,
+      "epoch": 0.8037741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 2.4770827657454555e-06,
-      "loss": 0.8389,
+      "learning_rate": 1.9525021148273604e-06,
+      "loss": 0.7813,
       "step": 28325
     },
     {
-      "epoch": 0.7779517178874515,
+      "epoch": 0.8038024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 2.4764967555427687e-06,
-      "loss": 0.7625,
+      "learning_rate": 1.951956571055975e-06,
+      "loss": 0.7207,
       "step": 28326
     },
     {
-      "epoch": 0.7779791821152948,
+      "epoch": 0.8038308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 2.4759108048689572e-06,
-      "loss": 0.8501,
+      "learning_rate": 1.951411095266772e-06,
+      "loss": 0.8729,
       "step": 28327
     },
     {
-      "epoch": 0.7780066463431381,
+      "epoch": 0.8038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 2.4753249137286515e-06,
-      "loss": 0.7458,
+      "learning_rate": 1.9508656874643604e-06,
+      "loss": 0.8593,
       "step": 28328
     },
     {
-      "epoch": 0.7780341105709813,
+      "epoch": 0.8038876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 2.4747390821264916e-06,
-      "loss": 0.8004,
+      "learning_rate": 1.9503203476533483e-06,
+      "loss": 0.7731,
       "step": 28329
     },
     {
-      "epoch": 0.7780615747988245,
+      "epoch": 0.8039160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 2.474153310067109e-06,
-      "loss": 0.775,
+      "learning_rate": 1.9497750758383427e-06,
+      "loss": 0.8042,
       "step": 28330
     },
     {
-      "epoch": 0.7780890390266678,
+      "epoch": 0.80394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 2.473567597555141e-06,
-      "loss": 0.859,
+      "learning_rate": 1.949229872023951e-06,
+      "loss": 0.8718,
       "step": 28331
     },
     {
-      "epoch": 0.778116503254511,
+      "epoch": 0.8039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 2.4729819445952207e-06,
-      "loss": 0.8383,
+      "learning_rate": 1.9486847362147743e-06,
+      "loss": 0.9501,
       "step": 28332
     },
     {
-      "epoch": 0.7781439674823543,
+      "epoch": 0.8040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 2.4723963511919837e-06,
-      "loss": 0.8478,
+      "learning_rate": 1.9481396684154188e-06,
+      "loss": 0.7988,
       "step": 28333
     },
     {
-      "epoch": 0.7781714317101974,
+      "epoch": 0.8040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 2.4718108173500655e-06,
-      "loss": 0.8816,
+      "learning_rate": 1.9475946686304927e-06,
+      "loss": 0.7325,
       "step": 28334
     },
     {
-      "epoch": 0.7781988959380407,
+      "epoch": 0.8040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 2.471225343074096e-06,
-      "loss": 0.9128,
+      "learning_rate": 1.9470497368645934e-06,
+      "loss": 0.7753,
       "step": 28335
     },
     {
-      "epoch": 0.778226360165884,
+      "epoch": 0.8040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 2.4706399283687055e-06,
-      "loss": 0.7812,
+      "learning_rate": 1.946504873122327e-06,
+      "loss": 0.5911,
       "step": 28336
     },
     {
-      "epoch": 0.7782538243937271,
+      "epoch": 0.8041146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 2.4700545732385284e-06,
-      "loss": 0.7527,
+      "learning_rate": 1.9459600774082987e-06,
+      "loss": 0.7887,
       "step": 28337
     },
     {
-      "epoch": 0.7782812886215704,
+      "epoch": 0.8041430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 2.4694692776881957e-06,
-      "loss": 0.8818,
+      "learning_rate": 1.9454153497271055e-06,
+      "loss": 0.8739,
       "step": 28338
     },
     {
-      "epoch": 0.7783087528494136,
+      "epoch": 0.8041713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 2.4688840417223393e-06,
-      "loss": 0.8379,
+      "learning_rate": 1.94487069008335e-06,
+      "loss": 0.7355,
       "step": 28339
     },
     {
-      "epoch": 0.7783362170772569,
+      "epoch": 0.804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 2.468298865345593e-06,
-      "loss": 0.8164,
+      "learning_rate": 1.944326098481638e-06,
+      "loss": 0.7527,
       "step": 28340
     },
     {
-      "epoch": 0.7783636813051001,
+      "epoch": 0.8042281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 2.4677137485625802e-06,
-      "loss": 0.8415,
+      "learning_rate": 1.943781574926562e-06,
+      "loss": 0.8094,
       "step": 28341
     },
     {
-      "epoch": 0.7783911455329433,
+      "epoch": 0.8042565266742339,
       "grad_norm": 0.0,
-      "learning_rate": 2.4671286913779347e-06,
-      "loss": 0.8286,
+      "learning_rate": 1.943237119422725e-06,
+      "loss": 0.7401,
       "step": 28342
     },
     {
-      "epoch": 0.7784186097607866,
+      "epoch": 0.8042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 2.4665436937962872e-06,
-      "loss": 0.8834,
+      "learning_rate": 1.942692731974729e-06,
+      "loss": 0.8464,
       "step": 28343
     },
     {
-      "epoch": 0.7784460739886299,
+      "epoch": 0.8043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 2.465958755822261e-06,
-      "loss": 0.8571,
+      "learning_rate": 1.942148412587165e-06,
+      "loss": 0.706,
       "step": 28344
     },
     {
-      "epoch": 0.778473538216473,
+      "epoch": 0.8043416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 2.4653738774604875e-06,
-      "loss": 0.8128,
+      "learning_rate": 1.9416041612646376e-06,
+      "loss": 0.9081,
       "step": 28345
     },
     {
-      "epoch": 0.7785010024443163,
+      "epoch": 0.8043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 2.4647890587155976e-06,
-      "loss": 0.8294,
+      "learning_rate": 1.9410599780117443e-06,
+      "loss": 0.8746,
       "step": 28346
     },
     {
-      "epoch": 0.7785284666721595,
+      "epoch": 0.8043984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 2.4642042995922133e-06,
-      "loss": 0.81,
+      "learning_rate": 1.9405158628330777e-06,
+      "loss": 0.9046,
       "step": 28347
     },
     {
-      "epoch": 0.7785559309000027,
+      "epoch": 0.8044267877412031,
       "grad_norm": 0.0,
-      "learning_rate": 2.463619600094963e-06,
-      "loss": 0.7573,
+      "learning_rate": 1.9399718157332358e-06,
+      "loss": 0.7686,
       "step": 28348
     },
     {
-      "epoch": 0.778583395127846,
+      "epoch": 0.8044551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 2.463034960228474e-06,
-      "loss": 0.7702,
+      "learning_rate": 1.9394278367168175e-06,
+      "loss": 0.7908,
       "step": 28349
     },
     {
-      "epoch": 0.7786108593556892,
+      "epoch": 0.804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 2.462450379997375e-06,
-      "loss": 0.7096,
+      "learning_rate": 1.938883925788411e-06,
+      "loss": 0.8315,
       "step": 28350
     },
     {
-      "epoch": 0.7786383235835325,
+      "epoch": 0.8045119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 2.461865859406285e-06,
-      "loss": 0.854,
+      "learning_rate": 1.9383400829526144e-06,
+      "loss": 0.8454,
       "step": 28351
     },
     {
-      "epoch": 0.7786657878113756,
+      "epoch": 0.8045402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 2.461281398459836e-06,
-      "loss": 0.8638,
+      "learning_rate": 1.937796308214025e-06,
+      "loss": 0.7716,
       "step": 28352
     },
     {
-      "epoch": 0.7786932520392189,
+      "epoch": 0.8045686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 2.4606969971626447e-06,
-      "loss": 0.885,
+      "learning_rate": 1.9372526015772296e-06,
+      "loss": 0.7305,
       "step": 28353
     },
     {
-      "epoch": 0.7787207162670622,
+      "epoch": 0.8045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 2.46011265551934e-06,
-      "loss": 0.8173,
+      "learning_rate": 1.936708963046826e-06,
+      "loss": 0.8169,
       "step": 28354
     },
     {
-      "epoch": 0.7787481804949054,
+      "epoch": 0.8046254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 2.4595283735345445e-06,
-      "loss": 0.77,
+      "learning_rate": 1.9361653926274016e-06,
+      "loss": 0.8695,
       "step": 28355
     },
     {
-      "epoch": 0.7787756447227486,
+      "epoch": 0.8046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 2.458944151212882e-06,
-      "loss": 0.7365,
+      "learning_rate": 1.9356218903235515e-06,
+      "loss": 0.7559,
       "step": 28356
     },
     {
-      "epoch": 0.7788031089505919,
+      "epoch": 0.8046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 2.4583599885589737e-06,
-      "loss": 0.8433,
+      "learning_rate": 1.9350784561398685e-06,
+      "loss": 0.8021,
       "step": 28357
     },
     {
-      "epoch": 0.7788305731784351,
+      "epoch": 0.8047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.4577758855774457e-06,
-      "loss": 0.876,
+      "learning_rate": 1.934535090080937e-06,
+      "loss": 0.7853,
       "step": 28358
     },
     {
-      "epoch": 0.7788580374062783,
+      "epoch": 0.804738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.4571918422729168e-06,
-      "loss": 0.8366,
+      "learning_rate": 1.9339917921513494e-06,
+      "loss": 0.7955,
       "step": 28359
     },
     {
-      "epoch": 0.7788855016341215,
+      "epoch": 0.8047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 2.456607858650004e-06,
-      "loss": 0.7784,
+      "learning_rate": 1.9334485623556977e-06,
+      "loss": 0.8368,
       "step": 28360
     },
     {
-      "epoch": 0.7789129658619648,
+      "epoch": 0.8047956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 2.456023934713333e-06,
-      "loss": 0.8242,
+      "learning_rate": 1.932905400698567e-06,
+      "loss": 0.8009,
       "step": 28361
     },
     {
-      "epoch": 0.7789404300898081,
+      "epoch": 0.8048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 2.455440070467522e-06,
-      "loss": 0.7785,
+      "learning_rate": 1.932362307184548e-06,
+      "loss": 0.8131,
       "step": 28362
     },
     {
-      "epoch": 0.7789678943176512,
+      "epoch": 0.8048524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 2.4548562659171915e-06,
-      "loss": 0.7614,
+      "learning_rate": 1.931819281818229e-06,
+      "loss": 0.8371,
       "step": 28363
     },
     {
-      "epoch": 0.7789953585454945,
+      "epoch": 0.8048808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 2.454272521066965e-06,
-      "loss": 0.8193,
+      "learning_rate": 1.9312763246041934e-06,
+      "loss": 0.7793,
       "step": 28364
     },
     {
-      "epoch": 0.7790228227733377,
+      "epoch": 0.8049091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 2.453688835921453e-06,
-      "loss": 0.7744,
+      "learning_rate": 1.9307334355470285e-06,
+      "loss": 0.7682,
       "step": 28365
     },
     {
-      "epoch": 0.779050287001181,
+      "epoch": 0.8049375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 2.4531052104852795e-06,
-      "loss": 0.7817,
+      "learning_rate": 1.930190614651325e-06,
+      "loss": 0.8564,
       "step": 28366
     },
     {
-      "epoch": 0.7790777512290242,
+      "epoch": 0.8049659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 2.4525216447630627e-06,
-      "loss": 0.818,
+      "learning_rate": 1.9296478619216608e-06,
+      "loss": 0.7962,
       "step": 28367
     },
     {
-      "epoch": 0.7791052154568674,
+      "epoch": 0.804994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 2.4519381387594153e-06,
-      "loss": 0.7841,
+      "learning_rate": 1.9291051773626236e-06,
+      "loss": 0.7649,
       "step": 28368
     },
     {
-      "epoch": 0.7791326796847107,
+      "epoch": 0.805022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 2.4513546924789577e-06,
-      "loss": 0.7872,
+      "learning_rate": 1.9285625609788005e-06,
+      "loss": 0.8343,
       "step": 28369
     },
     {
-      "epoch": 0.779160143912554,
+      "epoch": 0.8050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 2.4507713059263096e-06,
-      "loss": 0.8399,
+      "learning_rate": 1.9280200127747706e-06,
+      "loss": 0.7987,
       "step": 28370
     },
     {
-      "epoch": 0.7791876081403971,
+      "epoch": 0.8050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 2.450187979106079e-06,
-      "loss": 0.7574,
+      "learning_rate": 1.9274775327551176e-06,
+      "loss": 0.8013,
       "step": 28371
     },
     {
-      "epoch": 0.7792150723682404,
+      "epoch": 0.8051078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 2.449604712022886e-06,
-      "loss": 0.7997,
+      "learning_rate": 1.926935120924427e-06,
+      "loss": 0.7952,
       "step": 28372
     },
     {
-      "epoch": 0.7792425365960836,
+      "epoch": 0.8051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 2.4490215046813447e-06,
-      "loss": 0.8018,
+      "learning_rate": 1.9263927772872757e-06,
+      "loss": 0.7396,
       "step": 28373
     },
     {
-      "epoch": 0.7792700008239268,
+      "epoch": 0.8051645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 2.4484383570860736e-06,
-      "loss": 0.7481,
+      "learning_rate": 1.925850501848249e-06,
+      "loss": 0.8996,
       "step": 28374
     },
     {
-      "epoch": 0.7792974650517701,
+      "epoch": 0.8051929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 2.447855269241681e-06,
-      "loss": 0.8646,
+      "learning_rate": 1.9253082946119252e-06,
+      "loss": 0.7275,
       "step": 28375
     },
     {
-      "epoch": 0.7793249292796133,
+      "epoch": 0.8052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 2.447272241152785e-06,
-      "loss": 0.8107,
+      "learning_rate": 1.9247661555828844e-06,
+      "loss": 0.8087,
       "step": 28376
     },
     {
-      "epoch": 0.7793523935074566,
+      "epoch": 0.8052497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 2.4466892728239943e-06,
-      "loss": 0.8168,
+      "learning_rate": 1.9242240847657113e-06,
+      "loss": 0.8684,
       "step": 28377
     },
     {
-      "epoch": 0.7793798577352997,
+      "epoch": 0.8052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 2.4461063642599247e-06,
-      "loss": 0.9071,
+      "learning_rate": 1.923682082164976e-06,
+      "loss": 0.6896,
       "step": 28378
     },
     {
-      "epoch": 0.779407321963143,
+      "epoch": 0.8053064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 2.4455235154651867e-06,
-      "loss": 0.8746,
+      "learning_rate": 1.9231401477852617e-06,
+      "loss": 0.8017,
       "step": 28379
     },
     {
-      "epoch": 0.7794347861909863,
+      "epoch": 0.8053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 2.4449407264443927e-06,
-      "loss": 0.8355,
+      "learning_rate": 1.9225982816311484e-06,
+      "loss": 0.7514,
       "step": 28380
     },
     {
-      "epoch": 0.7794622504188294,
+      "epoch": 0.8053632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 2.444357997202155e-06,
-      "loss": 0.8046,
+      "learning_rate": 1.9220564837072086e-06,
+      "loss": 0.8227,
       "step": 28381
     },
     {
-      "epoch": 0.7794897146466727,
+      "epoch": 0.8053916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 2.4437753277430852e-06,
-      "loss": 0.8172,
+      "learning_rate": 1.9215147540180203e-06,
+      "loss": 0.7982,
       "step": 28382
     },
     {
-      "epoch": 0.779517178874516,
+      "epoch": 0.8054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 2.443192718071794e-06,
-      "loss": 0.8041,
+      "learning_rate": 1.9209730925681623e-06,
+      "loss": 0.8592,
       "step": 28383
     },
     {
-      "epoch": 0.7795446431023592,
+      "epoch": 0.8054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 2.442610168192885e-06,
-      "loss": 0.7549,
+      "learning_rate": 1.920431499362205e-06,
+      "loss": 0.8997,
       "step": 28384
     },
     {
-      "epoch": 0.7795721073302024,
+      "epoch": 0.8054767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 2.442027678110972e-06,
-      "loss": 0.8237,
+      "learning_rate": 1.919889974404727e-06,
+      "loss": 0.7743,
       "step": 28385
     },
     {
-      "epoch": 0.7795995715580456,
+      "epoch": 0.8055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 2.441445247830664e-06,
-      "loss": 0.8981,
+      "learning_rate": 1.9193485177003037e-06,
+      "loss": 0.8297,
       "step": 28386
     },
     {
-      "epoch": 0.7796270357858889,
+      "epoch": 0.805533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.440862877356568e-06,
-      "loss": 0.8292,
+      "learning_rate": 1.918807129253504e-06,
+      "loss": 0.8935,
       "step": 28387
     },
     {
-      "epoch": 0.7796545000137322,
+      "epoch": 0.8055618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 2.440280566693297e-06,
-      "loss": 0.8913,
+      "learning_rate": 1.9182658090689043e-06,
+      "loss": 0.8513,
       "step": 28388
     },
     {
-      "epoch": 0.7796819642415753,
+      "epoch": 0.8055902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 2.439698315845451e-06,
-      "loss": 0.8649,
+      "learning_rate": 1.9177245571510784e-06,
+      "loss": 0.7709,
       "step": 28389
     },
     {
-      "epoch": 0.7797094284694186,
+      "epoch": 0.8056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 2.439116124817642e-06,
-      "loss": 0.773,
+      "learning_rate": 1.917183373504592e-06,
+      "loss": 0.7971,
       "step": 28390
     },
     {
-      "epoch": 0.7797368926972618,
+      "epoch": 0.8056469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 2.438533993614477e-06,
-      "loss": 0.7782,
+      "learning_rate": 1.916642258134026e-06,
+      "loss": 0.8261,
       "step": 28391
     },
     {
-      "epoch": 0.779764356925105,
+      "epoch": 0.8056753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 2.4379519222405566e-06,
-      "loss": 0.8394,
+      "learning_rate": 1.9161012110439424e-06,
+      "loss": 0.9006,
       "step": 28392
     },
     {
-      "epoch": 0.7797918211529483,
+      "epoch": 0.8057037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 2.43736991070049e-06,
-      "loss": 0.8002,
+      "learning_rate": 1.9155602322389167e-06,
+      "loss": 0.8347,
       "step": 28393
     },
     {
-      "epoch": 0.7798192853807915,
+      "epoch": 0.8057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 2.436787958998885e-06,
-      "loss": 0.7604,
+      "learning_rate": 1.915019321723519e-06,
+      "loss": 0.8166,
       "step": 28394
     },
     {
-      "epoch": 0.7798467496086348,
+      "epoch": 0.8057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 2.436206067140341e-06,
-      "loss": 0.8531,
+      "learning_rate": 1.914478479502315e-06,
+      "loss": 0.8437,
       "step": 28395
     },
     {
-      "epoch": 0.779874213836478,
+      "epoch": 0.805788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.4356242351294647e-06,
-      "loss": 0.9069,
+      "learning_rate": 1.9139377055798736e-06,
+      "loss": 0.8032,
       "step": 28396
     },
     {
-      "epoch": 0.7799016780643212,
+      "epoch": 0.8058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 2.435042462970859e-06,
-      "loss": 0.763,
+      "learning_rate": 1.913396999960767e-06,
+      "loss": 0.842,
       "step": 28397
     },
     {
-      "epoch": 0.7799291422921645,
+      "epoch": 0.8058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 2.434460750669131e-06,
-      "loss": 0.8192,
+      "learning_rate": 1.9128563626495568e-06,
+      "loss": 0.9083,
       "step": 28398
     },
     {
-      "epoch": 0.7799566065200076,
+      "epoch": 0.8058740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 2.4338790982288775e-06,
-      "loss": 0.7719,
+      "learning_rate": 1.912315793650813e-06,
+      "loss": 0.7935,
       "step": 28399
     },
     {
-      "epoch": 0.7799840707478509,
+      "epoch": 0.8059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 2.433297505654706e-06,
-      "loss": 0.877,
+      "learning_rate": 1.9117752929691037e-06,
+      "loss": 0.7944,
       "step": 28400
     },
     {
-      "epoch": 0.7800115349756942,
+      "epoch": 0.8059307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 2.4327159729512116e-06,
-      "loss": 0.7745,
+      "learning_rate": 1.911234860608988e-06,
+      "loss": 0.9423,
       "step": 28401
     },
     {
-      "epoch": 0.7800389992035374,
+      "epoch": 0.8059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 2.4321345001230012e-06,
-      "loss": 0.8707,
+      "learning_rate": 1.9106944965750364e-06,
+      "loss": 0.8023,
       "step": 28402
     },
     {
-      "epoch": 0.7800664634313806,
+      "epoch": 0.8059875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 2.4315530871746738e-06,
-      "loss": 0.8628,
+      "learning_rate": 1.9101542008718144e-06,
+      "loss": 0.8336,
       "step": 28403
     },
     {
-      "epoch": 0.7800939276592238,
+      "epoch": 0.8060158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 2.4309717341108294e-06,
-      "loss": 0.7373,
+      "learning_rate": 1.909613973503881e-06,
+      "loss": 0.8469,
       "step": 28404
     },
     {
-      "epoch": 0.7801213918870671,
+      "epoch": 0.806044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.4303904409360736e-06,
-      "loss": 0.8551,
+      "learning_rate": 1.9090738144758027e-06,
+      "loss": 0.8691,
       "step": 28405
     },
     {
-      "epoch": 0.7801488561149104,
+      "epoch": 0.8060726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 2.4298092076549962e-06,
-      "loss": 0.8279,
+      "learning_rate": 1.9085337237921398e-06,
+      "loss": 0.8799,
       "step": 28406
     },
     {
-      "epoch": 0.7801763203427535,
+      "epoch": 0.8061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 2.429228034272204e-06,
-      "loss": 0.8961,
+      "learning_rate": 1.907993701457458e-06,
+      "loss": 0.801,
       "step": 28407
     },
     {
-      "epoch": 0.7802037845705968,
+      "epoch": 0.8061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 2.428646920792289e-06,
-      "loss": 0.8336,
+      "learning_rate": 1.907453747476319e-06,
+      "loss": 0.7741,
       "step": 28408
     },
     {
-      "epoch": 0.7802312487984401,
+      "epoch": 0.8061577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.428065867219852e-06,
-      "loss": 0.8773,
+      "learning_rate": 1.906913861853279e-06,
+      "loss": 0.9736,
       "step": 28409
     },
     {
-      "epoch": 0.7802587130262832,
+      "epoch": 0.8061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 2.4274848735594925e-06,
-      "loss": 0.8271,
+      "learning_rate": 1.9063740445929025e-06,
+      "loss": 0.9189,
       "step": 28410
     },
     {
-      "epoch": 0.7802861772541265,
+      "epoch": 0.8062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 2.426903939815808e-06,
-      "loss": 0.8051,
+      "learning_rate": 1.90583429569975e-06,
+      "loss": 0.8704,
       "step": 28411
     },
     {
-      "epoch": 0.7803136414819697,
+      "epoch": 0.8062429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 2.42632306599339e-06,
-      "loss": 0.8187,
+      "learning_rate": 1.905294615178377e-06,
+      "loss": 0.7907,
       "step": 28412
     },
     {
-      "epoch": 0.780341105709813,
+      "epoch": 0.8062712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 2.4257422520968378e-06,
-      "loss": 0.8302,
+      "learning_rate": 1.904755003033344e-06,
+      "loss": 0.6819,
       "step": 28413
     },
     {
-      "epoch": 0.7803685699376562,
+      "epoch": 0.8062996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 2.425161498130747e-06,
-      "loss": 0.8344,
+      "learning_rate": 1.9042154592692118e-06,
+      "loss": 0.7981,
       "step": 28414
     },
     {
-      "epoch": 0.7803960341654994,
+      "epoch": 0.8063280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 2.4245808040997156e-06,
-      "loss": 0.8571,
+      "learning_rate": 1.9036759838905329e-06,
+      "loss": 0.746,
       "step": 28415
     },
     {
-      "epoch": 0.7804234983933427,
+      "epoch": 0.8063564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 2.4240001700083327e-06,
-      "loss": 0.8433,
+      "learning_rate": 1.9031365769018673e-06,
+      "loss": 0.7727,
       "step": 28416
     },
     {
-      "epoch": 0.7804509626211859,
+      "epoch": 0.8063847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 2.4234195958611974e-06,
-      "loss": 0.9224,
+      "learning_rate": 1.902597238307774e-06,
+      "loss": 0.7661,
       "step": 28417
     },
     {
-      "epoch": 0.7804784268490291,
+      "epoch": 0.8064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 2.422839081662898e-06,
-      "loss": 0.8278,
+      "learning_rate": 1.9020579681128027e-06,
+      "loss": 0.8552,
       "step": 28418
     },
     {
-      "epoch": 0.7805058910768724,
+      "epoch": 0.8064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 2.4222586274180315e-06,
-      "loss": 0.8395,
+      "learning_rate": 1.9015187663215117e-06,
+      "loss": 0.9178,
       "step": 28419
     },
     {
-      "epoch": 0.7805333553047156,
+      "epoch": 0.8064699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 2.4216782331311908e-06,
-      "loss": 0.8235,
+      "learning_rate": 1.9009796329384578e-06,
+      "loss": 0.7872,
       "step": 28420
     },
     {
-      "epoch": 0.7805608195325588,
+      "epoch": 0.8064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 2.4210978988069655e-06,
-      "loss": 0.8712,
+      "learning_rate": 1.9004405679681893e-06,
+      "loss": 0.7833,
       "step": 28421
     },
     {
-      "epoch": 0.7805882837604021,
+      "epoch": 0.8065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 2.420517624449954e-06,
-      "loss": 0.8024,
+      "learning_rate": 1.899901571415268e-06,
+      "loss": 0.7548,
       "step": 28422
     },
     {
-      "epoch": 0.7806157479882453,
+      "epoch": 0.80655505107832,
       "grad_norm": 0.0,
-      "learning_rate": 2.41993741006474e-06,
-      "loss": 0.8149,
+      "learning_rate": 1.8993626432842394e-06,
+      "loss": 0.7731,
       "step": 28423
     },
     {
-      "epoch": 0.7806432122160886,
+      "epoch": 0.806583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 2.4193572556559196e-06,
-      "loss": 0.7041,
+      "learning_rate": 1.8988237835796586e-06,
+      "loss": 0.8263,
       "step": 28424
     },
     {
-      "epoch": 0.7806706764439317,
+      "epoch": 0.8066118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 2.4187771612280785e-06,
-      "loss": 0.8584,
+      "learning_rate": 1.8982849923060797e-06,
+      "loss": 0.713,
       "step": 28425
     },
     {
-      "epoch": 0.780698140671775,
+      "epoch": 0.8066401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 2.4181971267858085e-06,
-      "loss": 0.8357,
+      "learning_rate": 1.8977462694680493e-06,
+      "loss": 0.6675,
       "step": 28426
     },
     {
-      "epoch": 0.7807256048996183,
+      "epoch": 0.8066685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 2.4176171523337012e-06,
-      "loss": 0.7792,
+      "learning_rate": 1.8972076150701212e-06,
+      "loss": 0.7801,
       "step": 28427
     },
     {
-      "epoch": 0.7807530691274615,
+      "epoch": 0.8066969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 2.417037237876343e-06,
-      "loss": 0.7843,
+      "learning_rate": 1.8966690291168467e-06,
+      "loss": 0.7284,
       "step": 28428
     },
     {
-      "epoch": 0.7807805333553047,
+      "epoch": 0.8067253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 2.416457383418327e-06,
-      "loss": 0.9233,
+      "learning_rate": 1.8961305116127705e-06,
+      "loss": 0.8225,
       "step": 28429
     },
     {
-      "epoch": 0.7808079975831479,
+      "epoch": 0.8067536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 2.415877588964235e-06,
-      "loss": 0.8276,
+      "learning_rate": 1.8955920625624435e-06,
+      "loss": 0.7871,
       "step": 28430
     },
     {
-      "epoch": 0.7808354618109912,
+      "epoch": 0.8067820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 2.415297854518659e-06,
-      "loss": 0.8402,
+      "learning_rate": 1.895053681970419e-06,
+      "loss": 0.8055,
       "step": 28431
     },
     {
-      "epoch": 0.7808629260388344,
+      "epoch": 0.8068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 2.414718180086183e-06,
-      "loss": 0.8982,
+      "learning_rate": 1.8945153698412367e-06,
+      "loss": 0.8816,
       "step": 28432
     },
     {
-      "epoch": 0.7808903902666776,
+      "epoch": 0.806838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 2.4141385656713947e-06,
-      "loss": 0.8229,
+      "learning_rate": 1.8939771261794471e-06,
+      "loss": 0.769,
       "step": 28433
     },
     {
-      "epoch": 0.7809178544945209,
+      "epoch": 0.8068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 2.41355901127888e-06,
-      "loss": 0.8315,
+      "learning_rate": 1.8934389509896e-06,
+      "loss": 0.8602,
       "step": 28434
     },
     {
-      "epoch": 0.7809453187223642,
+      "epoch": 0.8068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.4129795169132276e-06,
-      "loss": 0.8737,
+      "learning_rate": 1.8929008442762365e-06,
+      "loss": 0.9251,
       "step": 28435
     },
     {
-      "epoch": 0.7809727829502073,
+      "epoch": 0.8069239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 2.4124000825790182e-06,
-      "loss": 0.8274,
+      "learning_rate": 1.8923628060439037e-06,
+      "loss": 0.8784,
       "step": 28436
     },
     {
-      "epoch": 0.7810002471780506,
+      "epoch": 0.8069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 2.4118207082808385e-06,
-      "loss": 0.8363,
+      "learning_rate": 1.8918248362971459e-06,
+      "loss": 0.7883,
       "step": 28437
     },
     {
-      "epoch": 0.7810277114058938,
+      "epoch": 0.8069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 2.4112413940232726e-06,
-      "loss": 0.8854,
+      "learning_rate": 1.8912869350405095e-06,
+      "loss": 0.8581,
       "step": 28438
     },
     {
-      "epoch": 0.781055175633737,
+      "epoch": 0.8070090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 2.4106621398109066e-06,
-      "loss": 0.8295,
+      "learning_rate": 1.8907491022785385e-06,
+      "loss": 0.8131,
       "step": 28439
     },
     {
-      "epoch": 0.7810826398615803,
+      "epoch": 0.8070374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 2.4100829456483197e-06,
-      "loss": 0.8514,
+      "learning_rate": 1.8902113380157715e-06,
+      "loss": 0.7904,
       "step": 28440
     },
     {
-      "epoch": 0.7811101040894235,
+      "epoch": 0.8070658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 2.4095038115400995e-06,
-      "loss": 0.8612,
+      "learning_rate": 1.8896736422567552e-06,
+      "loss": 0.7719,
       "step": 28441
     },
     {
-      "epoch": 0.7811375683172668,
+      "epoch": 0.807094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 2.408924737490822e-06,
-      "loss": 0.8191,
+      "learning_rate": 1.8891360150060323e-06,
+      "loss": 0.8795,
       "step": 28442
     },
     {
-      "epoch": 0.78116503254511,
+      "epoch": 0.8071225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 2.408345723505072e-06,
-      "loss": 0.8048,
+      "learning_rate": 1.8885984562681392e-06,
+      "loss": 0.7653,
       "step": 28443
     },
     {
-      "epoch": 0.7811924967729532,
+      "epoch": 0.8071509648127129,
       "grad_norm": 0.0,
-      "learning_rate": 2.407766769587432e-06,
-      "loss": 0.8228,
+      "learning_rate": 1.888060966047619e-06,
+      "loss": 0.7522,
       "step": 28444
     },
     {
-      "epoch": 0.7812199610007965,
+      "epoch": 0.8071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 2.4071878757424826e-06,
-      "loss": 0.9135,
+      "learning_rate": 1.887523544349016e-06,
+      "loss": 0.8265,
       "step": 28445
     },
     {
-      "epoch": 0.7812474252286397,
+      "epoch": 0.8072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 2.406609041974806e-06,
-      "loss": 0.8278,
+      "learning_rate": 1.886986191176864e-06,
+      "loss": 0.7693,
       "step": 28446
     },
     {
-      "epoch": 0.7812748894564829,
+      "epoch": 0.8072360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 2.406030268288978e-06,
-      "loss": 0.844,
+      "learning_rate": 1.8864489065357039e-06,
+      "loss": 0.8237,
       "step": 28447
     },
     {
-      "epoch": 0.7813023536843262,
+      "epoch": 0.8072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 2.4054515546895817e-06,
-      "loss": 0.8703,
+      "learning_rate": 1.8859116904300768e-06,
+      "loss": 0.7328,
       "step": 28448
     },
     {
-      "epoch": 0.7813298179121694,
+      "epoch": 0.8072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 2.4048729011811934e-06,
-      "loss": 0.8159,
+      "learning_rate": 1.8853745428645165e-06,
+      "loss": 0.7045,
       "step": 28449
     },
     {
-      "epoch": 0.7813572821400127,
+      "epoch": 0.8073212258796821,
       "grad_norm": 0.0,
-      "learning_rate": 2.4042943077683912e-06,
-      "loss": 0.8384,
+      "learning_rate": 1.8848374638435618e-06,
+      "loss": 0.9406,
       "step": 28450
     },
     {
-      "epoch": 0.7813847463678558,
+      "epoch": 0.8073496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 2.403715774455756e-06,
-      "loss": 0.9329,
+      "learning_rate": 1.8843004533717523e-06,
+      "loss": 0.8266,
       "step": 28451
     },
     {
-      "epoch": 0.7814122105956991,
+      "epoch": 0.807377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 2.403137301247863e-06,
-      "loss": 0.8726,
+      "learning_rate": 1.8837635114536168e-06,
+      "loss": 0.822,
       "step": 28452
     },
     {
-      "epoch": 0.7814396748235424,
+      "epoch": 0.8074063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 2.4025588881492947e-06,
-      "loss": 0.8157,
+      "learning_rate": 1.883226638093698e-06,
+      "loss": 0.9123,
       "step": 28453
     },
     {
-      "epoch": 0.7814671390513855,
+      "epoch": 0.8074347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 2.4019805351646186e-06,
-      "loss": 0.7903,
+      "learning_rate": 1.8826898332965315e-06,
+      "loss": 0.7941,
       "step": 28454
     },
     {
-      "epoch": 0.7814946032792288,
+      "epoch": 0.8074631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 2.4014022422984206e-06,
-      "loss": 0.876,
+      "learning_rate": 1.8821530970666468e-06,
+      "loss": 0.7808,
       "step": 28455
     },
     {
-      "epoch": 0.7815220675070721,
+      "epoch": 0.8074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 2.400824009555267e-06,
-      "loss": 0.9278,
+      "learning_rate": 1.8816164294085792e-06,
+      "loss": 0.8346,
       "step": 28456
     },
     {
-      "epoch": 0.7815495317349153,
+      "epoch": 0.8075198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 2.4002458369397387e-06,
-      "loss": 0.8676,
+      "learning_rate": 1.8810798303268651e-06,
+      "loss": 0.8574,
       "step": 28457
     },
     {
-      "epoch": 0.7815769959627585,
+      "epoch": 0.8075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 2.3996677244564113e-06,
-      "loss": 0.7821,
+      "learning_rate": 1.880543299826032e-06,
+      "loss": 0.8424,
       "step": 28458
     },
     {
-      "epoch": 0.7816044601906017,
+      "epoch": 0.8075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 2.3990896721098545e-06,
-      "loss": 0.8558,
+      "learning_rate": 1.8800068379106152e-06,
+      "loss": 0.7363,
       "step": 28459
     },
     {
-      "epoch": 0.781631924418445,
+      "epoch": 0.8076049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 2.3985116799046436e-06,
-      "loss": 0.7889,
+      "learning_rate": 1.8794704445851474e-06,
+      "loss": 0.6885,
       "step": 28460
     },
     {
-      "epoch": 0.7816593886462883,
+      "epoch": 0.807633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 2.3979337478453534e-06,
-      "loss": 0.7594,
+      "learning_rate": 1.8789341198541556e-06,
+      "loss": 0.8099,
       "step": 28461
     },
     {
-      "epoch": 0.7816868528741314,
+      "epoch": 0.8076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 2.3973558759365565e-06,
-      "loss": 0.7889,
+      "learning_rate": 1.8783978637221755e-06,
+      "loss": 0.8567,
       "step": 28462
     },
     {
-      "epoch": 0.7817143171019747,
+      "epoch": 0.8076901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 2.3967780641828275e-06,
-      "loss": 0.7555,
+      "learning_rate": 1.8778616761937308e-06,
+      "loss": 0.8425,
       "step": 28463
     },
     {
-      "epoch": 0.7817417813298179,
+      "epoch": 0.8077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 2.396200312588737e-06,
-      "loss": 0.7028,
+      "learning_rate": 1.877325557273354e-06,
+      "loss": 0.8627,
       "step": 28464
     },
     {
-      "epoch": 0.7817692455576611,
+      "epoch": 0.8077468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 2.3956226211588506e-06,
-      "loss": 0.7335,
+      "learning_rate": 1.876789506965575e-06,
+      "loss": 0.877,
       "step": 28465
     },
     {
-      "epoch": 0.7817967097855044,
+      "epoch": 0.8077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 2.3950449898977435e-06,
-      "loss": 0.8105,
+      "learning_rate": 1.8762535252749181e-06,
+      "loss": 0.8118,
       "step": 28466
     },
     {
-      "epoch": 0.7818241740133476,
+      "epoch": 0.8078036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 2.394467418809987e-06,
-      "loss": 0.8584,
+      "learning_rate": 1.8757176122059106e-06,
+      "loss": 0.8911,
       "step": 28467
     },
     {
-      "epoch": 0.7818516382411909,
+      "epoch": 0.8078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 2.39388990790015e-06,
-      "loss": 0.8489,
+      "learning_rate": 1.8751817677630857e-06,
+      "loss": 0.757,
       "step": 28468
     },
     {
-      "epoch": 0.7818791024690341,
+      "epoch": 0.8078603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 2.393312457172803e-06,
-      "loss": 0.7168,
+      "learning_rate": 1.874645991950964e-06,
+      "loss": 0.7889,
       "step": 28469
     },
     {
-      "epoch": 0.7819065666968773,
+      "epoch": 0.80788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.3927350666325167e-06,
-      "loss": 0.8246,
+      "learning_rate": 1.8741102847740734e-06,
+      "loss": 0.7717,
       "step": 28470
     },
     {
-      "epoch": 0.7819340309247206,
+      "epoch": 0.807917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 2.392157736283858e-06,
-      "loss": 0.8597,
+      "learning_rate": 1.8735746462369398e-06,
+      "loss": 0.7262,
       "step": 28471
     },
     {
-      "epoch": 0.7819614951525637,
+      "epoch": 0.8079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 2.3915804661313902e-06,
-      "loss": 0.8018,
+      "learning_rate": 1.8730390763440853e-06,
+      "loss": 0.8152,
       "step": 28472
     },
     {
-      "epoch": 0.781988959380407,
+      "epoch": 0.8079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.3910032561796858e-06,
-      "loss": 0.91,
+      "learning_rate": 1.8725035751000342e-06,
+      "loss": 0.7587,
       "step": 28473
     },
     {
-      "epoch": 0.7820164236082503,
+      "epoch": 0.8080022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 2.390426106433311e-06,
-      "loss": 0.9038,
+      "learning_rate": 1.8719681425093127e-06,
+      "loss": 0.7244,
       "step": 28474
     },
     {
-      "epoch": 0.7820438878360935,
+      "epoch": 0.8080306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 2.3898490168968325e-06,
-      "loss": 0.8546,
+      "learning_rate": 1.8714327785764397e-06,
+      "loss": 0.8487,
       "step": 28475
     },
     {
-      "epoch": 0.7820713520639367,
+      "epoch": 0.8080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 2.3892719875748193e-06,
-      "loss": 0.7916,
+      "learning_rate": 1.87089748330594e-06,
+      "loss": 0.8441,
       "step": 28476
     },
     {
-      "epoch": 0.7820988162917799,
+      "epoch": 0.8080874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 2.3886950184718304e-06,
-      "loss": 0.8847,
+      "learning_rate": 1.8703622567023373e-06,
+      "loss": 0.823,
       "step": 28477
     },
     {
-      "epoch": 0.7821262805196232,
+      "epoch": 0.8081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 2.3881181095924364e-06,
-      "loss": 0.7818,
+      "learning_rate": 1.8698270987701462e-06,
+      "loss": 0.681,
       "step": 28478
     },
     {
-      "epoch": 0.7821537447474665,
+      "epoch": 0.8081441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 2.387541260941203e-06,
-      "loss": 0.7738,
+      "learning_rate": 1.8692920095138923e-06,
+      "loss": 0.8923,
       "step": 28479
     },
     {
-      "epoch": 0.7821812089753096,
+      "epoch": 0.8081725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 2.386964472522688e-06,
-      "loss": 0.8835,
+      "learning_rate": 1.8687569889380962e-06,
+      "loss": 0.7967,
       "step": 28480
     },
     {
-      "epoch": 0.7822086732031529,
+      "epoch": 0.8082009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 2.3863877443414607e-06,
-      "loss": 0.8979,
+      "learning_rate": 1.8682220370472737e-06,
+      "loss": 0.8456,
       "step": 28481
     },
     {
-      "epoch": 0.7822361374309962,
+      "epoch": 0.8082292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 2.385811076402085e-06,
-      "loss": 0.8234,
+      "learning_rate": 1.867687153845944e-06,
+      "loss": 0.7803,
       "step": 28482
     },
     {
-      "epoch": 0.7822636016588393,
+      "epoch": 0.8082576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 2.385234468709118e-06,
-      "loss": 0.8637,
+      "learning_rate": 1.8671523393386272e-06,
+      "loss": 0.8784,
       "step": 28483
     },
     {
-      "epoch": 0.7822910658866826,
+      "epoch": 0.8082860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 2.384657921267126e-06,
-      "loss": 0.866,
+      "learning_rate": 1.8666175935298393e-06,
+      "loss": 0.8527,
       "step": 28484
     },
     {
-      "epoch": 0.7823185301145258,
+      "epoch": 0.8083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 2.3840814340806705e-06,
-      "loss": 0.8387,
+      "learning_rate": 1.8660829164241023e-06,
+      "loss": 0.8917,
       "step": 28485
     },
     {
-      "epoch": 0.7823459943423691,
+      "epoch": 0.8083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.383505007154313e-06,
-      "loss": 0.8433,
+      "learning_rate": 1.8655483080259252e-06,
+      "loss": 0.8616,
       "step": 28486
     },
     {
-      "epoch": 0.7823734585702123,
+      "epoch": 0.8083711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 2.382928640492617e-06,
-      "loss": 0.7946,
+      "learning_rate": 1.8650137683398261e-06,
+      "loss": 0.7874,
       "step": 28487
     },
     {
-      "epoch": 0.7824009227980555,
+      "epoch": 0.8083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 2.38235233410014e-06,
-      "loss": 0.8423,
+      "learning_rate": 1.8644792973703252e-06,
+      "loss": 0.7945,
       "step": 28488
     },
     {
-      "epoch": 0.7824283870258988,
+      "epoch": 0.808427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.38177608798144e-06,
-      "loss": 0.872,
+      "learning_rate": 1.8639448951219308e-06,
+      "loss": 0.751,
       "step": 28489
     },
     {
-      "epoch": 0.782455851253742,
+      "epoch": 0.8084562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 2.381199902141077e-06,
-      "loss": 0.7439,
+      "learning_rate": 1.8634105615991592e-06,
+      "loss": 0.8166,
       "step": 28490
     },
     {
-      "epoch": 0.7824833154815852,
+      "epoch": 0.8084846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 2.3806237765836125e-06,
-      "loss": 0.8558,
+      "learning_rate": 1.8628762968065272e-06,
+      "loss": 0.9382,
       "step": 28491
     },
     {
-      "epoch": 0.7825107797094285,
+      "epoch": 0.8085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 2.3800477113136042e-06,
-      "loss": 0.8098,
+      "learning_rate": 1.8623421007485431e-06,
+      "loss": 0.7435,
       "step": 28492
     },
     {
-      "epoch": 0.7825382439372717,
+      "epoch": 0.8085414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 2.379471706335611e-06,
-      "loss": 0.8447,
+      "learning_rate": 1.86180797342972e-06,
+      "loss": 0.7484,
       "step": 28493
     },
     {
-      "epoch": 0.782565708165115,
+      "epoch": 0.8085698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 2.3788957616541906e-06,
-      "loss": 0.8165,
+      "learning_rate": 1.8612739148545733e-06,
+      "loss": 0.8698,
       "step": 28494
     },
     {
-      "epoch": 0.7825931723929582,
+      "epoch": 0.8085981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 2.3783198772739e-06,
-      "loss": 0.7353,
+      "learning_rate": 1.8607399250276104e-06,
+      "loss": 0.8137,
       "step": 28495
     },
     {
-      "epoch": 0.7826206366208014,
+      "epoch": 0.8086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 2.377744053199292e-06,
-      "loss": 0.7948,
+      "learning_rate": 1.860206003953342e-06,
+      "loss": 0.9097,
       "step": 28496
     },
     {
-      "epoch": 0.7826481008486447,
+      "epoch": 0.8086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 2.3771682894349268e-06,
-      "loss": 0.7844,
+      "learning_rate": 1.8596721516362825e-06,
+      "loss": 0.9088,
       "step": 28497
     },
     {
-      "epoch": 0.7826755650764878,
+      "epoch": 0.808683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.3765925859853576e-06,
-      "loss": 0.8196,
+      "learning_rate": 1.8591383680809327e-06,
+      "loss": 0.684,
       "step": 28498
     },
     {
-      "epoch": 0.7827030293043311,
+      "epoch": 0.8087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 2.3760169428551404e-06,
-      "loss": 0.759,
+      "learning_rate": 1.8586046532918123e-06,
+      "loss": 0.7933,
       "step": 28499
     },
     {
-      "epoch": 0.7827304935321744,
+      "epoch": 0.8087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 2.3754413600488347e-06,
-      "loss": 0.8235,
+      "learning_rate": 1.858071007273422e-06,
+      "loss": 0.6893,
       "step": 28500
     },
     {
-      "epoch": 0.7827579577600176,
+      "epoch": 0.8087684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 2.3748658375709866e-06,
-      "loss": 0.8956,
+      "learning_rate": 1.8575374300302717e-06,
+      "loss": 0.8233,
       "step": 28501
     },
     {
-      "epoch": 0.7827854219878608,
+      "epoch": 0.8087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 2.3742903754261535e-06,
-      "loss": 0.8573,
+      "learning_rate": 1.8570039215668712e-06,
+      "loss": 0.8203,
       "step": 28502
     },
     {
-      "epoch": 0.782812886215704,
+      "epoch": 0.8088251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 2.3737149736188925e-06,
-      "loss": 0.7919,
+      "learning_rate": 1.8564704818877222e-06,
+      "loss": 0.6585,
       "step": 28503
     },
     {
-      "epoch": 0.7828403504435473,
+      "epoch": 0.8088535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 2.373139632153749e-06,
-      "loss": 0.8042,
+      "learning_rate": 1.8559371109973335e-06,
+      "loss": 0.7342,
       "step": 28504
     },
     {
-      "epoch": 0.7828678146713905,
+      "epoch": 0.8088819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 2.37256435103528e-06,
-      "loss": 0.8061,
+      "learning_rate": 1.8554038089002125e-06,
+      "loss": 0.8412,
       "step": 28505
     },
     {
-      "epoch": 0.7828952788992337,
+      "epoch": 0.8089103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 2.3719891302680386e-06,
-      "loss": 0.8231,
+      "learning_rate": 1.854870575600859e-06,
+      "loss": 0.9006,
       "step": 28506
     },
     {
-      "epoch": 0.782922743127077,
+      "epoch": 0.808938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 2.371413969856572e-06,
-      "loss": 0.822,
+      "learning_rate": 1.8543374111037804e-06,
+      "loss": 0.8184,
       "step": 28507
     },
     {
-      "epoch": 0.7829502073549203,
+      "epoch": 0.8089670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 2.370838869805433e-06,
-      "loss": 0.8306,
+      "learning_rate": 1.853804315413481e-06,
+      "loss": 0.7124,
       "step": 28508
     },
     {
-      "epoch": 0.7829776715827634,
+      "epoch": 0.8089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 2.3702638301191715e-06,
-      "loss": 0.8084,
+      "learning_rate": 1.8532712885344616e-06,
+      "loss": 0.8624,
       "step": 28509
     },
     {
-      "epoch": 0.7830051358106067,
+      "epoch": 0.8090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 2.369688850802341e-06,
-      "loss": 0.7934,
+      "learning_rate": 1.8527383304712254e-06,
+      "loss": 0.8652,
       "step": 28510
     },
     {
-      "epoch": 0.7830326000384499,
+      "epoch": 0.8090522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.3691139318594845e-06,
-      "loss": 0.8752,
+      "learning_rate": 1.8522054412282775e-06,
+      "loss": 0.9023,
       "step": 28511
     },
     {
-      "epoch": 0.7830600642662932,
+      "epoch": 0.8090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 2.3685390732951584e-06,
-      "loss": 0.788,
+      "learning_rate": 1.8516726208101133e-06,
+      "loss": 0.7691,
       "step": 28512
     },
     {
-      "epoch": 0.7830875284941364,
+      "epoch": 0.8091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 2.367964275113904e-06,
-      "loss": 0.8525,
+      "learning_rate": 1.8511398692212379e-06,
+      "loss": 0.8327,
       "step": 28513
     },
     {
-      "epoch": 0.7831149927219796,
+      "epoch": 0.8091373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 2.3673895373202725e-06,
-      "loss": 0.7877,
+      "learning_rate": 1.850607186466149e-06,
+      "loss": 0.8799,
       "step": 28514
     },
     {
-      "epoch": 0.7831424569498229,
+      "epoch": 0.8091657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 2.3668148599188124e-06,
-      "loss": 0.7313,
+      "learning_rate": 1.8500745725493485e-06,
+      "loss": 0.8098,
       "step": 28515
     },
     {
-      "epoch": 0.783169921177666,
+      "epoch": 0.8091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 2.3662402429140696e-06,
-      "loss": 0.718,
+      "learning_rate": 1.8495420274753362e-06,
+      "loss": 0.848,
       "step": 28516
     },
     {
-      "epoch": 0.7831973854055093,
+      "epoch": 0.80922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.3656656863105933e-06,
-      "loss": 0.8519,
+      "learning_rate": 1.8490095512486072e-06,
+      "loss": 0.8673,
       "step": 28517
     },
     {
-      "epoch": 0.7832248496333526,
+      "epoch": 0.8092508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 2.3650911901129248e-06,
-      "loss": 0.777,
+      "learning_rate": 1.8484771438736604e-06,
+      "loss": 0.8039,
       "step": 28518
     },
     {
-      "epoch": 0.7832523138611958,
+      "epoch": 0.8092792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 2.3645167543256155e-06,
-      "loss": 0.8402,
+      "learning_rate": 1.8479448053549965e-06,
+      "loss": 0.7653,
       "step": 28519
     },
     {
-      "epoch": 0.783279778089039,
+      "epoch": 0.8093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 2.3639423789532044e-06,
-      "loss": 0.8881,
+      "learning_rate": 1.8474125356971061e-06,
+      "loss": 0.8393,
       "step": 28520
     },
     {
-      "epoch": 0.7833072423168823,
+      "epoch": 0.8093359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 2.3633680640002387e-06,
-      "loss": 0.8592,
+      "learning_rate": 1.8468803349044894e-06,
+      "loss": 0.854,
       "step": 28521
     },
     {
-      "epoch": 0.7833347065447255,
+      "epoch": 0.8093643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 2.3627938094712632e-06,
-      "loss": 0.8306,
+      "learning_rate": 1.8463482029816427e-06,
+      "loss": 0.8133,
       "step": 28522
     },
     {
-      "epoch": 0.7833621707725688,
+      "epoch": 0.8093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 2.3622196153708244e-06,
-      "loss": 0.7497,
+      "learning_rate": 1.8458161399330565e-06,
+      "loss": 0.6878,
       "step": 28523
     },
     {
-      "epoch": 0.7833896350004119,
+      "epoch": 0.8094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.3616454817034597e-06,
-      "loss": 0.7988,
+      "learning_rate": 1.8452841457632287e-06,
+      "loss": 0.7716,
       "step": 28524
     },
     {
-      "epoch": 0.7834170992282552,
+      "epoch": 0.809449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 2.3610714084737152e-06,
-      "loss": 0.7596,
+      "learning_rate": 1.8447522204766545e-06,
+      "loss": 0.6482,
       "step": 28525
     },
     {
-      "epoch": 0.7834445634560985,
+      "epoch": 0.809477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 2.360497395686132e-06,
-      "loss": 0.796,
+      "learning_rate": 1.8442203640778222e-06,
+      "loss": 0.8826,
       "step": 28526
     },
     {
-      "epoch": 0.7834720276839416,
+      "epoch": 0.8095062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 2.3599234433452558e-06,
-      "loss": 0.8102,
+      "learning_rate": 1.8436885765712277e-06,
+      "loss": 0.7833,
       "step": 28527
     },
     {
-      "epoch": 0.7834994919117849,
+      "epoch": 0.8095346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 2.359349551455622e-06,
-      "loss": 0.7709,
+      "learning_rate": 1.8431568579613624e-06,
+      "loss": 0.8098,
       "step": 28528
     },
     {
-      "epoch": 0.7835269561396281,
+      "epoch": 0.8095629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 2.3587757200217766e-06,
-      "loss": 0.9033,
+      "learning_rate": 1.8426252082527175e-06,
+      "loss": 0.8568,
       "step": 28529
     },
     {
-      "epoch": 0.7835544203674714,
+      "epoch": 0.8095913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 2.3582019490482554e-06,
-      "loss": 0.7961,
+      "learning_rate": 1.842093627449787e-06,
+      "loss": 0.8909,
       "step": 28530
     },
     {
-      "epoch": 0.7835818845953146,
+      "epoch": 0.8096197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 2.357628238539601e-06,
-      "loss": 0.8757,
+      "learning_rate": 1.841562115557055e-06,
+      "loss": 0.8388,
       "step": 28531
     },
     {
-      "epoch": 0.7836093488231578,
+      "epoch": 0.8096481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 2.3570545885003515e-06,
-      "loss": 0.8618,
+      "learning_rate": 1.8410306725790151e-06,
+      "loss": 0.8279,
       "step": 28532
     },
     {
-      "epoch": 0.7836368130510011,
+      "epoch": 0.8096765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 2.356480998935047e-06,
-      "loss": 0.9076,
+      "learning_rate": 1.8404992985201587e-06,
+      "loss": 0.7871,
       "step": 28533
     },
     {
-      "epoch": 0.7836642772788444,
+      "epoch": 0.8097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 2.3559074698482277e-06,
-      "loss": 0.7794,
+      "learning_rate": 1.839967993384969e-06,
+      "loss": 0.7056,
       "step": 28534
     },
     {
-      "epoch": 0.7836917415066875,
+      "epoch": 0.809733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 2.355334001244428e-06,
-      "loss": 0.8078,
+      "learning_rate": 1.839436757177936e-06,
+      "loss": 0.7133,
       "step": 28535
     },
     {
-      "epoch": 0.7837192057345308,
+      "epoch": 0.8097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 2.354760593128189e-06,
-      "loss": 0.7828,
+      "learning_rate": 1.83890558990355e-06,
+      "loss": 0.8429,
       "step": 28536
     },
     {
-      "epoch": 0.783746669962374,
+      "epoch": 0.8097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 2.3541872455040427e-06,
-      "loss": 0.7928,
+      "learning_rate": 1.8383744915662927e-06,
+      "loss": 0.8392,
       "step": 28537
     },
     {
-      "epoch": 0.7837741341902172,
+      "epoch": 0.8098183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 2.3536139583765293e-06,
-      "loss": 0.7611,
+      "learning_rate": 1.8378434621706542e-06,
+      "loss": 0.8404,
       "step": 28538
     },
     {
-      "epoch": 0.7838015984180605,
+      "epoch": 0.8098467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 2.3530407317501834e-06,
-      "loss": 0.7894,
+      "learning_rate": 1.8373125017211201e-06,
+      "loss": 0.8462,
       "step": 28539
     },
     {
-      "epoch": 0.7838290626459037,
+      "epoch": 0.8098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 2.3524675656295427e-06,
-      "loss": 0.8867,
+      "learning_rate": 1.8367816102221724e-06,
+      "loss": 0.7194,
       "step": 28540
     },
     {
-      "epoch": 0.783856526873747,
+      "epoch": 0.8099035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 2.351894460019143e-06,
-      "loss": 0.8324,
+      "learning_rate": 1.8362507876782964e-06,
+      "loss": 0.87,
       "step": 28541
     },
     {
-      "epoch": 0.7838839911015901,
+      "epoch": 0.8099318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 2.351321414923513e-06,
-      "loss": 0.8965,
+      "learning_rate": 1.8357200340939807e-06,
+      "loss": 0.8234,
       "step": 28542
     },
     {
-      "epoch": 0.7839114553294334,
+      "epoch": 0.8099602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 2.350748430347195e-06,
-      "loss": 0.804,
+      "learning_rate": 1.8351893494737017e-06,
+      "loss": 0.8348,
       "step": 28543
     },
     {
-      "epoch": 0.7839389195572767,
+      "epoch": 0.809988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 2.350175506294715e-06,
-      "loss": 0.8694,
+      "learning_rate": 1.8346587338219456e-06,
+      "loss": 0.8615,
       "step": 28544
     },
     {
-      "epoch": 0.7839663837851198,
+      "epoch": 0.8100170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 2.34960264277061e-06,
-      "loss": 0.8431,
+      "learning_rate": 1.8341281871431947e-06,
+      "loss": 0.8092,
       "step": 28545
     },
     {
-      "epoch": 0.7839938480129631,
+      "epoch": 0.8100454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 2.349029839779412e-06,
-      "loss": 0.8134,
+      "learning_rate": 1.8335977094419288e-06,
+      "loss": 0.7958,
       "step": 28546
     },
     {
-      "epoch": 0.7840213122408064,
+      "epoch": 0.8100737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 2.3484570973256572e-06,
-      "loss": 0.6877,
+      "learning_rate": 1.8330673007226341e-06,
+      "loss": 0.7917,
       "step": 28547
     },
     {
-      "epoch": 0.7840487764686496,
+      "epoch": 0.8101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 2.3478844154138704e-06,
-      "loss": 0.8832,
+      "learning_rate": 1.8325369609897837e-06,
+      "loss": 0.8015,
       "step": 28548
     },
     {
-      "epoch": 0.7840762406964928,
+      "epoch": 0.8101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 2.347311794048587e-06,
-      "loss": 0.7237,
+      "learning_rate": 1.8320066902478606e-06,
+      "loss": 0.8567,
       "step": 28549
     },
     {
-      "epoch": 0.784103704924336,
+      "epoch": 0.8101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 2.3467392332343353e-06,
-      "loss": 0.7219,
+      "learning_rate": 1.831476488501347e-06,
+      "loss": 0.8151,
       "step": 28550
     },
     {
-      "epoch": 0.7841311691521793,
+      "epoch": 0.8101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 2.346166732975652e-06,
-      "loss": 0.842,
+      "learning_rate": 1.830946355754717e-06,
+      "loss": 0.8007,
       "step": 28551
     },
     {
-      "epoch": 0.7841586333800226,
+      "epoch": 0.8102156640181611,
       "grad_norm": 0.0,
-      "learning_rate": 2.345594293277058e-06,
-      "loss": 0.8164,
+      "learning_rate": 1.8304162920124492e-06,
+      "loss": 0.8779,
       "step": 28552
     },
     {
-      "epoch": 0.7841860976078657,
+      "epoch": 0.8102440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 2.3450219141430906e-06,
-      "loss": 0.7352,
+      "learning_rate": 1.8298862972790243e-06,
+      "loss": 0.8124,
       "step": 28553
     },
     {
-      "epoch": 0.784213561835709,
+      "epoch": 0.810272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 2.344449595578273e-06,
-      "loss": 0.8031,
+      "learning_rate": 1.8293563715589158e-06,
+      "loss": 0.6941,
       "step": 28554
     },
     {
-      "epoch": 0.7842410260635522,
+      "epoch": 0.8103007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 2.343877337587135e-06,
-      "loss": 0.7951,
+      "learning_rate": 1.8288265148566008e-06,
+      "loss": 0.8202,
       "step": 28555
     },
     {
-      "epoch": 0.7842684902913954,
+      "epoch": 0.8103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 2.3433051401742047e-06,
-      "loss": 0.846,
+      "learning_rate": 1.8282967271765583e-06,
+      "loss": 0.8063,
       "step": 28556
     },
     {
-      "epoch": 0.7842959545192387,
+      "epoch": 0.8103575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 2.342733003344011e-06,
-      "loss": 0.8223,
+      "learning_rate": 1.8277670085232578e-06,
+      "loss": 0.8438,
       "step": 28557
     },
     {
-      "epoch": 0.7843234187470819,
+      "epoch": 0.8103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 2.3421609271010816e-06,
-      "loss": 0.8664,
+      "learning_rate": 1.8272373589011772e-06,
+      "loss": 0.8025,
       "step": 28558
     },
     {
-      "epoch": 0.7843508829749252,
+      "epoch": 0.8104143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 2.341588911449939e-06,
-      "loss": 0.7512,
+      "learning_rate": 1.8267077783147901e-06,
+      "loss": 0.7991,
       "step": 28559
     },
     {
-      "epoch": 0.7843783472027684,
+      "epoch": 0.8104426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 2.3410169563951144e-06,
-      "loss": 0.9051,
+      "learning_rate": 1.826178266768569e-06,
+      "loss": 0.8333,
       "step": 28560
     },
     {
-      "epoch": 0.7844058114306116,
+      "epoch": 0.8104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 2.3404450619411277e-06,
-      "loss": 0.802,
+      "learning_rate": 1.8256488242669878e-06,
+      "loss": 0.8326,
       "step": 28561
     },
     {
-      "epoch": 0.7844332756584549,
+      "epoch": 0.8104994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 2.3398732280925063e-06,
-      "loss": 0.7615,
+      "learning_rate": 1.825119450814522e-06,
+      "loss": 0.8199,
       "step": 28562
     },
     {
-      "epoch": 0.784460739886298,
+      "epoch": 0.810527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 2.3393014548537764e-06,
-      "loss": 0.7238,
+      "learning_rate": 1.8245901464156358e-06,
+      "loss": 0.8287,
       "step": 28563
     },
     {
-      "epoch": 0.7844882041141413,
+      "epoch": 0.8105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 2.33872974222946e-06,
-      "loss": 0.7875,
+      "learning_rate": 1.8240609110748053e-06,
+      "loss": 0.8107,
       "step": 28564
     },
     {
-      "epoch": 0.7845156683419846,
+      "epoch": 0.8105845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 2.338158090224084e-06,
-      "loss": 0.8309,
+      "learning_rate": 1.8235317447965017e-06,
+      "loss": 0.8759,
       "step": 28565
     },
     {
-      "epoch": 0.7845431325698278,
+      "epoch": 0.8106129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 2.337586498842167e-06,
-      "loss": 0.7383,
+      "learning_rate": 1.8230026475851902e-06,
+      "loss": 0.7203,
       "step": 28566
     },
     {
-      "epoch": 0.784570596797671,
+      "epoch": 0.8106413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 2.3370149680882357e-06,
-      "loss": 0.7862,
+      "learning_rate": 1.8224736194453429e-06,
+      "loss": 0.7696,
       "step": 28567
     },
     {
-      "epoch": 0.7845980610255142,
+      "epoch": 0.8106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 2.3364434979668084e-06,
-      "loss": 0.7886,
+      "learning_rate": 1.8219446603814317e-06,
+      "loss": 0.7902,
       "step": 28568
     },
     {
-      "epoch": 0.7846255252533575,
+      "epoch": 0.8106980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 2.335872088482408e-06,
-      "loss": 0.8212,
+      "learning_rate": 1.8214157703979186e-06,
+      "loss": 0.8483,
       "step": 28569
     },
     {
-      "epoch": 0.7846529894812008,
+      "epoch": 0.8107264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 2.3353007396395566e-06,
-      "loss": 0.8174,
+      "learning_rate": 1.8208869494992776e-06,
+      "loss": 0.8125,
       "step": 28570
     },
     {
-      "epoch": 0.7846804537090439,
+      "epoch": 0.8107548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 2.3347294514427776e-06,
+      "learning_rate": 1.8203581976899687e-06,
       "loss": 0.7444,
       "step": 28571
     },
     {
-      "epoch": 0.7847079179368872,
+      "epoch": 0.810783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 2.3341582238965867e-06,
-      "loss": 0.8739,
+      "learning_rate": 1.8198295149744617e-06,
+      "loss": 0.8511,
       "step": 28572
     },
     {
-      "epoch": 0.7847353821647305,
+      "epoch": 0.810811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 2.333587057005504e-06,
-      "loss": 0.8069,
+      "learning_rate": 1.819300901357226e-06,
+      "loss": 0.7966,
       "step": 28573
     },
     {
-      "epoch": 0.7847628463925737,
+      "epoch": 0.8108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 2.3330159507740513e-06,
-      "loss": 0.8073,
+      "learning_rate": 1.8187723568427173e-06,
+      "loss": 0.7885,
       "step": 28574
     },
     {
-      "epoch": 0.7847903106204169,
+      "epoch": 0.8108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.3324449052067477e-06,
-      "loss": 0.7376,
+      "learning_rate": 1.8182438814354087e-06,
+      "loss": 0.8389,
       "step": 28575
     },
     {
-      "epoch": 0.7848177748482601,
+      "epoch": 0.8108967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 2.3318739203081107e-06,
-      "loss": 0.7517,
+      "learning_rate": 1.8177154751397652e-06,
+      "loss": 0.7888,
       "step": 28576
     },
     {
-      "epoch": 0.7848452390761034,
+      "epoch": 0.8109250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 2.331302996082655e-06,
-      "loss": 0.7715,
+      "learning_rate": 1.8171871379602435e-06,
+      "loss": 0.7885,
       "step": 28577
     },
     {
-      "epoch": 0.7848727033039467,
+      "epoch": 0.8109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 2.330732132534901e-06,
-      "loss": 0.8656,
+      "learning_rate": 1.816658869901311e-06,
+      "loss": 0.911,
       "step": 28578
     },
     {
-      "epoch": 0.7849001675317898,
+      "epoch": 0.8109818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 2.3301613296693648e-06,
-      "loss": 0.8488,
+      "learning_rate": 1.816130670967431e-06,
+      "loss": 0.913,
       "step": 28579
     },
     {
-      "epoch": 0.7849276317596331,
+      "epoch": 0.8110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 2.3295905874905633e-06,
-      "loss": 0.7707,
+      "learning_rate": 1.8156025411630596e-06,
+      "loss": 0.8761,
       "step": 28580
     },
     {
-      "epoch": 0.7849550959874763,
+      "epoch": 0.8110385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 2.3290199060030117e-06,
-      "loss": 0.8452,
+      "learning_rate": 1.8150744804926623e-06,
+      "loss": 0.8707,
       "step": 28581
     },
     {
-      "epoch": 0.7849825602153195,
+      "epoch": 0.8110669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 2.3284492852112297e-06,
-      "loss": 0.7681,
+      "learning_rate": 1.8145464889607012e-06,
+      "loss": 0.8776,
       "step": 28582
     },
     {
-      "epoch": 0.7850100244431628,
+      "epoch": 0.8110953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 2.3278787251197256e-06,
-      "loss": 0.8768,
+      "learning_rate": 1.8140185665716315e-06,
+      "loss": 0.7984,
       "step": 28583
     },
     {
-      "epoch": 0.785037488671006,
+      "epoch": 0.8111237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 2.327308225733019e-06,
-      "loss": 0.7739,
+      "learning_rate": 1.813490713329915e-06,
+      "loss": 0.8954,
       "step": 28584
     },
     {
-      "epoch": 0.7850649528988493,
+      "epoch": 0.8111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 2.3267377870556196e-06,
-      "loss": 0.8104,
+      "learning_rate": 1.812962929240013e-06,
+      "loss": 0.8128,
       "step": 28585
     },
     {
-      "epoch": 0.7850924171266925,
+      "epoch": 0.8111804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 2.326167409092043e-06,
-      "loss": 0.8687,
+      "learning_rate": 1.8124352143063784e-06,
+      "loss": 0.8035,
       "step": 28586
     },
     {
-      "epoch": 0.7851198813545357,
+      "epoch": 0.8112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 2.325597091846803e-06,
-      "loss": 0.7205,
+      "learning_rate": 1.811907568533472e-06,
+      "loss": 0.8552,
       "step": 28587
     },
     {
-      "epoch": 0.785147345582379,
+      "epoch": 0.8112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 2.325026835324413e-06,
-      "loss": 0.8151,
+      "learning_rate": 1.8113799919257525e-06,
+      "loss": 0.8253,
       "step": 28588
     },
     {
-      "epoch": 0.7851748098102221,
+      "epoch": 0.8112656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 2.3244566395293823e-06,
-      "loss": 0.7764,
+      "learning_rate": 1.810852484487672e-06,
+      "loss": 0.8749,
       "step": 28589
     },
     {
-      "epoch": 0.7852022740380654,
+      "epoch": 0.8112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 2.323886504466224e-06,
-      "loss": 0.7182,
+      "learning_rate": 1.8103250462236888e-06,
+      "loss": 0.7762,
       "step": 28590
     },
     {
-      "epoch": 0.7852297382659087,
+      "epoch": 0.811322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 2.3233164301394508e-06,
-      "loss": 0.8342,
+      "learning_rate": 1.8097976771382575e-06,
+      "loss": 0.9039,
       "step": 28591
     },
     {
-      "epoch": 0.7852572024937519,
+      "epoch": 0.8113507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 2.3227464165535695e-06,
-      "loss": 0.7167,
+      "learning_rate": 1.8092703772358345e-06,
+      "loss": 0.8114,
       "step": 28592
     },
     {
-      "epoch": 0.7852846667215951,
+      "epoch": 0.8113791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 2.322176463713092e-06,
-      "loss": 0.8263,
+      "learning_rate": 1.8087431465208738e-06,
+      "loss": 0.9059,
       "step": 28593
     },
     {
-      "epoch": 0.7853121309494383,
+      "epoch": 0.8114074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 2.321606571622532e-06,
-      "loss": 0.888,
+      "learning_rate": 1.8082159849978265e-06,
+      "loss": 0.7675,
       "step": 28594
     },
     {
-      "epoch": 0.7853395951772816,
+      "epoch": 0.8114358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 2.321036740286392e-06,
-      "loss": 0.8524,
+      "learning_rate": 1.8076888926711466e-06,
+      "loss": 0.8047,
       "step": 28595
     },
     {
-      "epoch": 0.7853670594051249,
+      "epoch": 0.8114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 2.320466969709183e-06,
-      "loss": 0.8084,
+      "learning_rate": 1.8071618695452896e-06,
+      "loss": 0.8044,
       "step": 28596
     },
     {
-      "epoch": 0.785394523632968,
+      "epoch": 0.8114926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 2.319897259895415e-06,
-      "loss": 0.7078,
+      "learning_rate": 1.806634915624702e-06,
+      "loss": 0.7602,
       "step": 28597
     },
     {
-      "epoch": 0.7854219878608113,
+      "epoch": 0.8115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 2.3193276108495943e-06,
-      "loss": 0.8211,
+      "learning_rate": 1.8061080309138379e-06,
+      "loss": 0.7529,
       "step": 28598
     },
     {
-      "epoch": 0.7854494520886546,
+      "epoch": 0.8115493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 2.318758022576233e-06,
-      "loss": 0.7258,
+      "learning_rate": 1.8055812154171505e-06,
+      "loss": 0.7212,
       "step": 28599
     },
     {
-      "epoch": 0.7854769163164977,
+      "epoch": 0.811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.3181884950798316e-06,
-      "loss": 0.7753,
+      "learning_rate": 1.8050544691390836e-06,
+      "loss": 0.7892,
       "step": 28600
     },
     {
-      "epoch": 0.785504380544341,
+      "epoch": 0.8116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 2.3176190283648968e-06,
-      "loss": 0.7229,
+      "learning_rate": 1.80452779208409e-06,
+      "loss": 0.809,
       "step": 28601
     },
     {
-      "epoch": 0.7855318447721842,
+      "epoch": 0.8116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 2.3170496224359363e-06,
-      "loss": 0.9344,
+      "learning_rate": 1.8040011842566218e-06,
+      "loss": 0.8653,
       "step": 28602
     },
     {
-      "epoch": 0.7855593090000275,
+      "epoch": 0.8116628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 2.3164802772974548e-06,
-      "loss": 0.8586,
+      "learning_rate": 1.8034746456611207e-06,
+      "loss": 0.792,
       "step": 28603
     },
     {
-      "epoch": 0.7855867732278707,
+      "epoch": 0.8116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 2.315910992953958e-06,
-      "loss": 0.8587,
+      "learning_rate": 1.8029481763020384e-06,
+      "loss": 0.8267,
       "step": 28604
     },
     {
-      "epoch": 0.7856142374557139,
+      "epoch": 0.8117196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 2.31534176940995e-06,
-      "loss": 0.8734,
+      "learning_rate": 1.8024217761838202e-06,
+      "loss": 0.813,
       "step": 28605
     },
     {
-      "epoch": 0.7856417016835572,
+      "epoch": 0.8117480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 2.314772606669937e-06,
-      "loss": 0.8517,
+      "learning_rate": 1.801895445310915e-06,
+      "loss": 0.6711,
       "step": 28606
     },
     {
-      "epoch": 0.7856691659114003,
+      "epoch": 0.8117763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 2.314203504738418e-06,
-      "loss": 0.7977,
+      "learning_rate": 1.80136918368777e-06,
+      "loss": 0.7664,
       "step": 28607
     },
     {
-      "epoch": 0.7856966301392436,
+      "epoch": 0.8118047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 2.3136344636199006e-06,
-      "loss": 0.875,
+      "learning_rate": 1.800842991318824e-06,
+      "loss": 0.8569,
       "step": 28608
     },
     {
-      "epoch": 0.7857240943670869,
+      "epoch": 0.81183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.3130654833188814e-06,
-      "loss": 0.826,
+      "learning_rate": 1.8003168682085282e-06,
+      "loss": 0.8642,
       "step": 28609
     },
     {
-      "epoch": 0.7857515585949301,
+      "epoch": 0.8118615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 2.3124965638398656e-06,
-      "loss": 0.765,
+      "learning_rate": 1.7997908143613252e-06,
+      "loss": 0.7483,
       "step": 28610
     },
     {
-      "epoch": 0.7857790228227733,
+      "epoch": 0.8118898978433599,
       "grad_norm": 0.0,
-      "learning_rate": 2.3119277051873555e-06,
-      "loss": 0.7912,
+      "learning_rate": 1.7992648297816563e-06,
+      "loss": 0.8139,
       "step": 28611
     },
     {
-      "epoch": 0.7858064870506166,
+      "epoch": 0.8119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 2.3113589073658536e-06,
-      "loss": 0.8969,
+      "learning_rate": 1.7987389144739653e-06,
+      "loss": 0.7744,
       "step": 28612
     },
     {
-      "epoch": 0.7858339512784598,
+      "epoch": 0.8119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 2.3107901703798563e-06,
-      "loss": 0.73,
+      "learning_rate": 1.798213068442698e-06,
+      "loss": 0.7751,
       "step": 28613
     },
     {
-      "epoch": 0.7858614155063031,
+      "epoch": 0.8119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 2.310221494233865e-06,
-      "loss": 0.7098,
+      "learning_rate": 1.7976872916922905e-06,
+      "loss": 0.8485,
       "step": 28614
     },
     {
-      "epoch": 0.7858888797341462,
+      "epoch": 0.8120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 2.3096528789323834e-06,
-      "loss": 0.8601,
+      "learning_rate": 1.7971615842271883e-06,
+      "loss": 0.9004,
       "step": 28615
     },
     {
-      "epoch": 0.7859163439619895,
+      "epoch": 0.8120317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 2.3090843244799042e-06,
-      "loss": 0.8127,
+      "learning_rate": 1.7966359460518323e-06,
+      "loss": 0.8547,
       "step": 28616
     },
     {
-      "epoch": 0.7859438081898328,
+      "epoch": 0.8120601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 2.3085158308809296e-06,
-      "loss": 0.7203,
+      "learning_rate": 1.79611037717066e-06,
+      "loss": 0.8469,
       "step": 28617
     },
     {
-      "epoch": 0.785971272417676,
+      "epoch": 0.8120885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 2.3079473981399592e-06,
-      "loss": 0.9435,
+      "learning_rate": 1.7955848775881114e-06,
+      "loss": 0.8719,
       "step": 28618
     },
     {
-      "epoch": 0.7859987366455192,
+      "epoch": 0.812116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 2.307379026261486e-06,
-      "loss": 0.7235,
+      "learning_rate": 1.795059447308629e-06,
+      "loss": 0.7386,
       "step": 28619
     },
     {
-      "epoch": 0.7860262008733624,
+      "epoch": 0.8121452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 2.3068107152500095e-06,
-      "loss": 0.8635,
+      "learning_rate": 1.7945340863366434e-06,
+      "loss": 0.805,
       "step": 28620
     },
     {
-      "epoch": 0.7860536651012057,
+      "epoch": 0.8121736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 2.306242465110027e-06,
-      "loss": 0.8391,
+      "learning_rate": 1.7940087946766017e-06,
+      "loss": 0.8765,
       "step": 28621
     },
     {
-      "epoch": 0.7860811293290489,
+      "epoch": 0.8122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 2.3056742758460336e-06,
-      "loss": 0.8002,
+      "learning_rate": 1.7934835723329347e-06,
+      "loss": 0.8405,
       "step": 28622
     },
     {
-      "epoch": 0.7861085935568921,
+      "epoch": 0.8122304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 2.30510614746253e-06,
-      "loss": 0.7784,
+      "learning_rate": 1.7929584193100802e-06,
+      "loss": 0.8185,
       "step": 28623
     },
     {
-      "epoch": 0.7861360577847354,
+      "epoch": 0.8122587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 2.304538079964006e-06,
-      "loss": 0.8301,
+      "learning_rate": 1.792433335612478e-06,
+      "loss": 0.6992,
       "step": 28624
     },
     {
-      "epoch": 0.7861635220125787,
+      "epoch": 0.8122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 2.3039700733549564e-06,
-      "loss": 0.774,
+      "learning_rate": 1.7919083212445566e-06,
+      "loss": 0.7076,
       "step": 28625
     },
     {
-      "epoch": 0.7861909862404218,
+      "epoch": 0.8123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 2.3034021276398754e-06,
-      "loss": 0.8565,
+      "learning_rate": 1.7913833762107557e-06,
+      "loss": 0.8207,
       "step": 28626
     },
     {
-      "epoch": 0.7862184504682651,
+      "epoch": 0.812343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 2.3028342428232585e-06,
-      "loss": 0.7805,
+      "learning_rate": 1.790858500515511e-06,
+      "loss": 0.7883,
       "step": 28627
     },
     {
-      "epoch": 0.7862459146961083,
+      "epoch": 0.812372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 2.302266418909599e-06,
-      "loss": 0.7552,
+      "learning_rate": 1.790333694163251e-06,
+      "loss": 0.8306,
       "step": 28628
     },
     {
-      "epoch": 0.7862733789239515,
+      "epoch": 0.8124006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 2.3016986559033884e-06,
-      "loss": 0.7732,
+      "learning_rate": 1.789808957158411e-06,
+      "loss": 0.8243,
       "step": 28629
     },
     {
-      "epoch": 0.7863008431517948,
+      "epoch": 0.8124290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 2.301130953809123e-06,
-      "loss": 0.8248,
+      "learning_rate": 1.7892842895054263e-06,
+      "loss": 0.7466,
       "step": 28630
     },
     {
-      "epoch": 0.786328307379638,
+      "epoch": 0.8124574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 2.300563312631293e-06,
-      "loss": 0.813,
+      "learning_rate": 1.7887596912087245e-06,
+      "loss": 0.8074,
       "step": 28631
     },
     {
-      "epoch": 0.7863557716074813,
+      "epoch": 0.8124858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 2.299995732374385e-06,
-      "loss": 0.7515,
+      "learning_rate": 1.7882351622727378e-06,
+      "loss": 0.7763,
       "step": 28632
     },
     {
-      "epoch": 0.7863832358353244,
+      "epoch": 0.8125141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 2.2994282130428935e-06,
-      "loss": 0.8493,
+      "learning_rate": 1.7877107027019003e-06,
+      "loss": 0.8282,
       "step": 28633
     },
     {
-      "epoch": 0.7864107000631677,
+      "epoch": 0.8125425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 2.298860754641309e-06,
-      "loss": 0.8493,
+      "learning_rate": 1.7871863125006383e-06,
+      "loss": 0.8658,
       "step": 28634
     },
     {
-      "epoch": 0.786438164291011,
+      "epoch": 0.8125709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 2.2982933571741208e-06,
-      "loss": 0.7868,
+      "learning_rate": 1.786661991673382e-06,
+      "loss": 0.689,
       "step": 28635
     },
     {
-      "epoch": 0.7864656285188542,
+      "epoch": 0.8125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 2.297726020645823e-06,
-      "loss": 0.7426,
+      "learning_rate": 1.7861377402245605e-06,
+      "loss": 0.8186,
       "step": 28636
     },
     {
-      "epoch": 0.7864930927466974,
+      "epoch": 0.812627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.2971587450608965e-06,
-      "loss": 0.7576,
+      "learning_rate": 1.785613558158603e-06,
+      "loss": 0.8846,
       "step": 28637
     },
     {
-      "epoch": 0.7865205569745407,
+      "epoch": 0.812656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 2.296591530423834e-06,
-      "loss": 0.8519,
+      "learning_rate": 1.7850894454799405e-06,
+      "loss": 0.7828,
       "step": 28638
     },
     {
-      "epoch": 0.7865480212023839,
+      "epoch": 0.8126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.2960243767391265e-06,
-      "loss": 0.7559,
+      "learning_rate": 1.7845654021929936e-06,
+      "loss": 0.87,
       "step": 28639
     },
     {
-      "epoch": 0.7865754854302272,
+      "epoch": 0.8127128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 2.2954572840112553e-06,
-      "loss": 0.7841,
+      "learning_rate": 1.7840414283021923e-06,
+      "loss": 0.7443,
       "step": 28640
     },
     {
-      "epoch": 0.7866029496580703,
+      "epoch": 0.8127412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 2.294890252244714e-06,
-      "loss": 0.9486,
+      "learning_rate": 1.783517523811964e-06,
+      "loss": 0.7712,
       "step": 28641
     },
     {
-      "epoch": 0.7866304138859136,
+      "epoch": 0.8127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 2.294323281443981e-06,
-      "loss": 0.8333,
+      "learning_rate": 1.7829936887267306e-06,
+      "loss": 0.8959,
       "step": 28642
     },
     {
-      "epoch": 0.7866578781137569,
+      "epoch": 0.8127979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 2.293756371613549e-06,
-      "loss": 0.8163,
+      "learning_rate": 1.782469923050919e-06,
+      "loss": 0.8844,
       "step": 28643
     },
     {
-      "epoch": 0.7866853423416,
+      "epoch": 0.8128263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 2.2931895227579003e-06,
-      "loss": 0.8047,
+      "learning_rate": 1.7819462267889564e-06,
+      "loss": 0.8031,
       "step": 28644
     },
     {
-      "epoch": 0.7867128065694433,
+      "epoch": 0.8128547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 2.2926227348815224e-06,
-      "loss": 0.715,
+      "learning_rate": 1.7814225999452605e-06,
+      "loss": 0.9039,
       "step": 28645
     },
     {
-      "epoch": 0.7867402707972866,
+      "epoch": 0.812883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 2.292056007988899e-06,
-      "loss": 0.7535,
+      "learning_rate": 1.7808990425242567e-06,
+      "loss": 0.7456,
       "step": 28646
     },
     {
-      "epoch": 0.7867677350251298,
+      "epoch": 0.8129114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 2.291489342084516e-06,
-      "loss": 0.742,
+      "learning_rate": 1.7803755545303714e-06,
+      "loss": 0.9313,
       "step": 28647
     },
     {
-      "epoch": 0.786795199252973,
+      "epoch": 0.8129398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 2.290922737172857e-06,
-      "loss": 0.8318,
+      "learning_rate": 1.77985213596802e-06,
+      "loss": 0.9164,
       "step": 28648
     },
     {
-      "epoch": 0.7868226634808162,
+      "epoch": 0.8129682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 2.2903561932583996e-06,
-      "loss": 0.7827,
+      "learning_rate": 1.7793287868416275e-06,
+      "loss": 0.9063,
       "step": 28649
     },
     {
-      "epoch": 0.7868501277086595,
+      "epoch": 0.8129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 2.289789710345631e-06,
-      "loss": 0.8722,
+      "learning_rate": 1.7788055071556175e-06,
+      "loss": 0.8171,
       "step": 28650
     },
     {
-      "epoch": 0.7868775919365028,
+      "epoch": 0.8130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 2.289223288439032e-06,
-      "loss": 0.7585,
+      "learning_rate": 1.778282296914402e-06,
+      "loss": 0.7142,
       "step": 28651
     },
     {
-      "epoch": 0.7869050561643459,
+      "epoch": 0.8130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.2886569275430857e-06,
-      "loss": 0.879,
+      "learning_rate": 1.7777591561224094e-06,
+      "loss": 0.7791,
       "step": 28652
     },
     {
-      "epoch": 0.7869325203921892,
+      "epoch": 0.8130817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 2.288090627662276e-06,
-      "loss": 0.7418,
+      "learning_rate": 1.777236084784053e-06,
+      "loss": 0.7656,
       "step": 28653
     },
     {
-      "epoch": 0.7869599846200324,
+      "epoch": 0.8131101021566401,
       "grad_norm": 0.0,
-      "learning_rate": 2.2875243888010777e-06,
-      "loss": 0.7765,
+      "learning_rate": 1.7767130829037527e-06,
+      "loss": 0.7611,
       "step": 28654
     },
     {
-      "epoch": 0.7869874488478756,
+      "epoch": 0.8131384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 2.2869582109639765e-06,
-      "loss": 0.8038,
+      "learning_rate": 1.7761901504859291e-06,
+      "loss": 0.9168,
       "step": 28655
     },
     {
-      "epoch": 0.7870149130757189,
+      "epoch": 0.81316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.2863920941554472e-06,
-      "loss": 0.8947,
+      "learning_rate": 1.7756672875349956e-06,
+      "loss": 0.7582,
       "step": 28656
     },
     {
-      "epoch": 0.7870423773035621,
+      "epoch": 0.8131952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 2.2858260383799714e-06,
-      "loss": 0.7528,
+      "learning_rate": 1.7751444940553686e-06,
+      "loss": 0.8145,
       "step": 28657
     },
     {
-      "epoch": 0.7870698415314054,
+      "epoch": 0.8132236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 2.285260043642028e-06,
-      "loss": 0.764,
+      "learning_rate": 1.77462177005147e-06,
+      "loss": 0.8569,
       "step": 28658
     },
     {
-      "epoch": 0.7870973057592486,
+      "epoch": 0.8132519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 2.2846941099460986e-06,
-      "loss": 0.7738,
+      "learning_rate": 1.7740991155277076e-06,
+      "loss": 0.8255,
       "step": 28659
     },
     {
-      "epoch": 0.7871247699870918,
+      "epoch": 0.8132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 2.2841282372966544e-06,
-      "loss": 0.7823,
+      "learning_rate": 1.7735765304884988e-06,
+      "loss": 0.7549,
       "step": 28660
     },
     {
-      "epoch": 0.7871522342149351,
+      "epoch": 0.8133087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 2.283562425698177e-06,
-      "loss": 0.8373,
+      "learning_rate": 1.7730540149382625e-06,
+      "loss": 0.8252,
       "step": 28661
     },
     {
-      "epoch": 0.7871796984427782,
+      "epoch": 0.8133371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 2.2829966751551437e-06,
-      "loss": 0.9133,
+      "learning_rate": 1.7725315688814059e-06,
+      "loss": 0.9313,
       "step": 28662
     },
     {
-      "epoch": 0.7872071626706215,
+      "epoch": 0.8133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 2.2824309856720317e-06,
-      "loss": 0.7791,
+      "learning_rate": 1.7720091923223458e-06,
+      "loss": 0.7741,
       "step": 28663
     },
     {
-      "epoch": 0.7872346268984648,
+      "epoch": 0.8133938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 2.281865357253312e-06,
-      "loss": 0.7966,
+      "learning_rate": 1.7714868852654953e-06,
+      "loss": 0.8616,
       "step": 28664
     },
     {
-      "epoch": 0.787262091126308,
+      "epoch": 0.813422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.2812997899034674e-06,
-      "loss": 0.816,
+      "learning_rate": 1.770964647715263e-06,
+      "loss": 0.7913,
       "step": 28665
     },
     {
-      "epoch": 0.7872895553541512,
+      "epoch": 0.8134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 2.2807342836269653e-06,
-      "loss": 0.7012,
+      "learning_rate": 1.77044247967606e-06,
+      "loss": 0.9026,
       "step": 28666
     },
     {
-      "epoch": 0.7873170195819944,
+      "epoch": 0.8134790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 2.280168838428284e-06,
-      "loss": 0.8815,
+      "learning_rate": 1.7699203811523047e-06,
+      "loss": 0.826,
       "step": 28667
     },
     {
-      "epoch": 0.7873444838098377,
+      "epoch": 0.8135073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 2.2796034543118972e-06,
-      "loss": 0.9106,
+      "learning_rate": 1.7693983521483982e-06,
+      "loss": 0.7056,
       "step": 28668
     },
     {
-      "epoch": 0.787371948037681,
+      "epoch": 0.8135357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 2.279038131282277e-06,
-      "loss": 0.8284,
+      "learning_rate": 1.7688763926687546e-06,
+      "loss": 0.8731,
       "step": 28669
     },
     {
-      "epoch": 0.7873994122655241,
+      "epoch": 0.8135641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 2.278472869343903e-06,
-      "loss": 0.8173,
+      "learning_rate": 1.768354502717784e-06,
+      "loss": 0.7025,
       "step": 28670
     },
     {
-      "epoch": 0.7874268764933674,
+      "epoch": 0.8135925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 2.277907668501239e-06,
-      "loss": 0.785,
+      "learning_rate": 1.7678326822998914e-06,
+      "loss": 0.8011,
       "step": 28671
     },
     {
-      "epoch": 0.7874543407212107,
+      "epoch": 0.8136208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 2.2773425287587635e-06,
-      "loss": 0.8446,
+      "learning_rate": 1.7673109314194858e-06,
+      "loss": 0.74,
       "step": 28672
     },
     {
-      "epoch": 0.7874818049490538,
+      "epoch": 0.8136492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 2.276777450120943e-06,
-      "loss": 0.8612,
+      "learning_rate": 1.766789250080977e-06,
+      "loss": 0.795,
       "step": 28673
     },
     {
-      "epoch": 0.7875092691768971,
+      "epoch": 0.813677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 2.2762124325922498e-06,
-      "loss": 0.8593,
+      "learning_rate": 1.7662676382887667e-06,
+      "loss": 0.8328,
       "step": 28674
     },
     {
-      "epoch": 0.7875367334047403,
+      "epoch": 0.8137060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 2.2756474761771562e-06,
-      "loss": 0.8323,
+      "learning_rate": 1.765746096047265e-06,
+      "loss": 0.8511,
       "step": 28675
     },
     {
-      "epoch": 0.7875641976325836,
+      "epoch": 0.8137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 2.2750825808801324e-06,
-      "loss": 0.8371,
+      "learning_rate": 1.7652246233608783e-06,
+      "loss": 0.7955,
       "step": 28676
     },
     {
-      "epoch": 0.7875916618604268,
+      "epoch": 0.8137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.2745177467056502e-06,
-      "loss": 0.8313,
+      "learning_rate": 1.7647032202340065e-06,
+      "loss": 0.8694,
       "step": 28677
     },
     {
-      "epoch": 0.78761912608827,
+      "epoch": 0.8137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 2.2739529736581735e-06,
-      "loss": 0.8514,
+      "learning_rate": 1.7641818866710592e-06,
+      "loss": 0.7445,
       "step": 28678
     },
     {
-      "epoch": 0.7876465903161133,
+      "epoch": 0.8138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 2.2733882617421764e-06,
-      "loss": 0.7683,
+      "learning_rate": 1.7636606226764353e-06,
+      "loss": 0.7511,
       "step": 28679
     },
     {
-      "epoch": 0.7876740545439564,
+      "epoch": 0.8138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 2.2728236109621214e-06,
-      "loss": 0.7344,
+      "learning_rate": 1.76313942825454e-06,
+      "loss": 0.8615,
       "step": 28680
     },
     {
-      "epoch": 0.7877015187717997,
+      "epoch": 0.8138762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 2.272259021322479e-06,
-      "loss": 0.9094,
+      "learning_rate": 1.76261830340978e-06,
+      "loss": 0.7655,
       "step": 28681
     },
     {
-      "epoch": 0.787728982999643,
+      "epoch": 0.8139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 2.2716944928277173e-06,
-      "loss": 0.8271,
+      "learning_rate": 1.762097248146547e-06,
+      "loss": 0.7441,
       "step": 28682
     },
     {
-      "epoch": 0.7877564472274862,
+      "epoch": 0.8139330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 2.2711300254823045e-06,
-      "loss": 0.808,
+      "learning_rate": 1.7615762624692523e-06,
+      "loss": 0.8289,
       "step": 28683
     },
     {
-      "epoch": 0.7877839114553294,
+      "epoch": 0.8139614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 2.2705656192907024e-06,
-      "loss": 0.8307,
+      "learning_rate": 1.7610553463822954e-06,
+      "loss": 0.8843,
       "step": 28684
     },
     {
-      "epoch": 0.7878113756831727,
+      "epoch": 0.8139897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 2.2700012742573795e-06,
-      "loss": 0.7919,
+      "learning_rate": 1.760534499890072e-06,
+      "loss": 0.8586,
       "step": 28685
     },
     {
-      "epoch": 0.7878388399110159,
+      "epoch": 0.8140181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 2.269436990386801e-06,
-      "loss": 0.6399,
+      "learning_rate": 1.7600137229969837e-06,
+      "loss": 0.8555,
       "step": 28686
     },
     {
-      "epoch": 0.7878663041388592,
+      "epoch": 0.8140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 2.268872767683433e-06,
-      "loss": 0.8547,
+      "learning_rate": 1.759493015707433e-06,
+      "loss": 0.8651,
       "step": 28687
     },
     {
-      "epoch": 0.7878937683667023,
+      "epoch": 0.8140749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 2.2683086061517367e-06,
-      "loss": 0.8215,
+      "learning_rate": 1.7589723780258127e-06,
+      "loss": 0.7833,
       "step": 28688
     },
     {
-      "epoch": 0.7879212325945456,
+      "epoch": 0.8141032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 2.26774450579618e-06,
-      "loss": 0.7193,
+      "learning_rate": 1.758451809956523e-06,
+      "loss": 0.8325,
       "step": 28689
     },
     {
-      "epoch": 0.7879486968223889,
+      "epoch": 0.8141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.2671804666212217e-06,
-      "loss": 0.8488,
+      "learning_rate": 1.7579313115039642e-06,
+      "loss": 0.8263,
       "step": 28690
     },
     {
-      "epoch": 0.787976161050232,
+      "epoch": 0.8141600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 2.266616488631327e-06,
-      "loss": 0.8514,
+      "learning_rate": 1.7574108826725268e-06,
+      "loss": 0.8271,
       "step": 28691
     },
     {
-      "epoch": 0.7880036252780753,
+      "epoch": 0.814188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 2.2660525718309577e-06,
-      "loss": 0.8257,
+      "learning_rate": 1.756890523466611e-06,
+      "loss": 0.7824,
       "step": 28692
     },
     {
-      "epoch": 0.7880310895059185,
+      "epoch": 0.814216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 2.2654887162245763e-06,
-      "loss": 0.8445,
+      "learning_rate": 1.756370233890613e-06,
+      "loss": 0.8002,
       "step": 28693
     },
     {
-      "epoch": 0.7880585537337618,
+      "epoch": 0.8142451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 2.264924921816648e-06,
-      "loss": 0.8235,
+      "learning_rate": 1.7558500139489243e-06,
+      "loss": 0.705,
       "step": 28694
     },
     {
-      "epoch": 0.788086017961605,
+      "epoch": 0.8142735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 2.264361188611627e-06,
-      "loss": 0.8088,
+      "learning_rate": 1.7553298636459416e-06,
+      "loss": 0.8763,
       "step": 28695
     },
     {
-      "epoch": 0.7881134821894482,
+      "epoch": 0.8143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 2.263797516613978e-06,
-      "loss": 0.8797,
+      "learning_rate": 1.7548097829860599e-06,
+      "loss": 0.8821,
       "step": 28696
     },
     {
-      "epoch": 0.7881409464172915,
+      "epoch": 0.8143303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 2.2632339058281584e-06,
-      "loss": 0.7559,
+      "learning_rate": 1.7542897719736662e-06,
+      "loss": 0.7668,
       "step": 28697
     },
     {
-      "epoch": 0.7881684106451348,
+      "epoch": 0.8143586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 2.2626703562586284e-06,
-      "loss": 0.739,
+      "learning_rate": 1.7537698306131624e-06,
+      "loss": 0.7936,
       "step": 28698
     },
     {
-      "epoch": 0.7881958748729779,
+      "epoch": 0.8143870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 2.262106867909848e-06,
-      "loss": 0.876,
+      "learning_rate": 1.7532499589089324e-06,
+      "loss": 0.7509,
       "step": 28699
     },
     {
-      "epoch": 0.7882233391008212,
+      "epoch": 0.8144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 2.261543440786279e-06,
-      "loss": 0.6914,
+      "learning_rate": 1.7527301568653709e-06,
+      "loss": 0.807,
       "step": 28700
     },
     {
-      "epoch": 0.7882508033286644,
+      "epoch": 0.8144438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 2.2609800748923716e-06,
-      "loss": 0.9662,
+      "learning_rate": 1.752210424486872e-06,
+      "loss": 0.7594,
       "step": 28701
     },
     {
-      "epoch": 0.7882782675565077,
+      "epoch": 0.814472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 2.2604167702325885e-06,
-      "loss": 0.6567,
+      "learning_rate": 1.7516907617778189e-06,
+      "loss": 0.7812,
       "step": 28702
     },
     {
-      "epoch": 0.7883057317843509,
+      "epoch": 0.8145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 2.2598535268113885e-06,
-      "loss": 0.7916,
+      "learning_rate": 1.7511711687426047e-06,
+      "loss": 0.7796,
       "step": 28703
     },
     {
-      "epoch": 0.7883331960121941,
+      "epoch": 0.8145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 2.259290344633224e-06,
-      "loss": 0.8082,
+      "learning_rate": 1.7506516453856216e-06,
+      "loss": 0.7681,
       "step": 28704
     },
     {
-      "epoch": 0.7883606602400374,
+      "epoch": 0.8145573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 2.2587272237025514e-06,
-      "loss": 0.7668,
+      "learning_rate": 1.7501321917112525e-06,
+      "loss": 0.8032,
       "step": 28705
     },
     {
-      "epoch": 0.7883881244678805,
+      "epoch": 0.8145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 2.2581641640238307e-06,
-      "loss": 0.8251,
+      "learning_rate": 1.7496128077238872e-06,
+      "loss": 0.7531,
       "step": 28706
     },
     {
-      "epoch": 0.7884155886957238,
+      "epoch": 0.8146140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 2.2576011656015116e-06,
-      "loss": 0.7474,
+      "learning_rate": 1.7490934934279159e-06,
+      "loss": 0.7529,
       "step": 28707
     },
     {
-      "epoch": 0.7884430529235671,
+      "epoch": 0.8146424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 2.2570382284400505e-06,
-      "loss": 0.8596,
+      "learning_rate": 1.7485742488277202e-06,
+      "loss": 0.7811,
       "step": 28708
     },
     {
-      "epoch": 0.7884705171514103,
+      "epoch": 0.8146708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 2.256475352543903e-06,
-      "loss": 0.8058,
+      "learning_rate": 1.748055073927688e-06,
+      "loss": 0.7959,
       "step": 28709
     },
     {
-      "epoch": 0.7884979813792535,
+      "epoch": 0.8146992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 2.2559125379175217e-06,
-      "loss": 0.8163,
+      "learning_rate": 1.7475359687322092e-06,
+      "loss": 0.9049,
       "step": 28710
     },
     {
-      "epoch": 0.7885254456070968,
+      "epoch": 0.814727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 2.2553497845653627e-06,
-      "loss": 0.8635,
+      "learning_rate": 1.747016933245662e-06,
+      "loss": 0.7566,
       "step": 28711
     },
     {
-      "epoch": 0.78855290983494,
+      "epoch": 0.8147559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 2.254787092491877e-06,
-      "loss": 0.8083,
+      "learning_rate": 1.7464979674724335e-06,
+      "loss": 0.8267,
       "step": 28712
     },
     {
-      "epoch": 0.7885803740627833,
+      "epoch": 0.8147843359818389,
       "grad_norm": 0.0,
-      "learning_rate": 2.254224461701513e-06,
-      "loss": 0.8652,
+      "learning_rate": 1.7459790714169089e-06,
+      "loss": 0.8307,
       "step": 28713
     },
     {
-      "epoch": 0.7886078382906264,
+      "epoch": 0.8148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 2.2536618921987253e-06,
-      "loss": 0.7395,
+      "learning_rate": 1.745460245083469e-06,
+      "loss": 0.711,
       "step": 28714
     },
     {
-      "epoch": 0.7886353025184697,
+      "epoch": 0.8148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.2530993839879657e-06,
-      "loss": 0.7966,
+      "learning_rate": 1.7449414884765005e-06,
+      "loss": 0.8221,
       "step": 28715
     },
     {
-      "epoch": 0.788662766746313,
+      "epoch": 0.8148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 2.252536937073684e-06,
-      "loss": 0.8428,
+      "learning_rate": 1.7444228016003795e-06,
+      "loss": 0.8383,
       "step": 28716
     },
     {
-      "epoch": 0.7886902309741561,
+      "epoch": 0.8148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 2.251974551460331e-06,
-      "loss": 0.7778,
+      "learning_rate": 1.7439041844594907e-06,
+      "loss": 0.8086,
       "step": 28717
     },
     {
-      "epoch": 0.7887176952019994,
+      "epoch": 0.8149262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 2.2514122271523598e-06,
-      "loss": 0.6971,
+      "learning_rate": 1.743385637058216e-06,
+      "loss": 0.8482,
       "step": 28718
     },
     {
-      "epoch": 0.7887451594298426,
+      "epoch": 0.8149545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 2.2508499641542135e-06,
-      "loss": 0.7896,
+      "learning_rate": 1.7428671594009327e-06,
+      "loss": 0.8539,
       "step": 28719
     },
     {
-      "epoch": 0.7887726236576859,
+      "epoch": 0.8149829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 2.250287762470348e-06,
-      "loss": 0.8612,
+      "learning_rate": 1.74234875149202e-06,
+      "loss": 0.813,
       "step": 28720
     },
     {
-      "epoch": 0.7888000878855291,
+      "epoch": 0.815011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 2.2497256221052034e-06,
-      "loss": 0.8568,
+      "learning_rate": 1.7418304133358633e-06,
+      "loss": 0.7028,
       "step": 28721
     },
     {
-      "epoch": 0.7888275521133723,
+      "epoch": 0.8150397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 2.249163543063233e-06,
-      "loss": 0.9356,
+      "learning_rate": 1.741312144936832e-06,
+      "loss": 0.9066,
       "step": 28722
     },
     {
-      "epoch": 0.7888550163412156,
+      "epoch": 0.8150681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 2.2486015253488826e-06,
-      "loss": 0.8869,
+      "learning_rate": 1.7407939462993094e-06,
+      "loss": 0.8019,
       "step": 28723
     },
     {
-      "epoch": 0.7888824805690589,
+      "epoch": 0.8150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 2.2480395689666025e-06,
-      "loss": 0.8729,
+      "learning_rate": 1.7402758174276734e-06,
+      "loss": 0.9114,
       "step": 28724
     },
     {
-      "epoch": 0.788909944796902,
+      "epoch": 0.8151248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 2.2474776739208327e-06,
-      "loss": 0.791,
+      "learning_rate": 1.739757758326297e-06,
+      "loss": 0.87,
       "step": 28725
     },
     {
-      "epoch": 0.7889374090247453,
+      "epoch": 0.8151532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 2.246915840216023e-06,
-      "loss": 0.8772,
+      "learning_rate": 1.7392397689995578e-06,
+      "loss": 0.9003,
       "step": 28726
     },
     {
-      "epoch": 0.7889648732525885,
+      "epoch": 0.8151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 2.246354067856621e-06,
-      "loss": 0.7351,
+      "learning_rate": 1.7387218494518331e-06,
+      "loss": 0.8405,
       "step": 28727
     },
     {
-      "epoch": 0.7889923374804317,
+      "epoch": 0.8152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 2.2457923568470664e-06,
-      "loss": 0.7814,
+      "learning_rate": 1.7382039996874934e-06,
+      "loss": 0.8478,
       "step": 28728
     },
     {
-      "epoch": 0.789019801708275,
+      "epoch": 0.8152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 2.2452307071918066e-06,
-      "loss": 0.707,
+      "learning_rate": 1.737686219710919e-06,
+      "loss": 0.7514,
       "step": 28729
     },
     {
-      "epoch": 0.7890472659361182,
+      "epoch": 0.815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 2.2446691188952883e-06,
-      "loss": 0.8313,
+      "learning_rate": 1.737168509526479e-06,
+      "loss": 0.7424,
       "step": 28730
     },
     {
-      "epoch": 0.7890747301639615,
+      "epoch": 0.8152951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 2.2441075919619494e-06,
-      "loss": 0.8463,
+      "learning_rate": 1.7366508691385475e-06,
+      "loss": 0.7902,
       "step": 28731
     },
     {
-      "epoch": 0.7891021943918046,
+      "epoch": 0.8153234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 2.2435461263962344e-06,
-      "loss": 0.7668,
+      "learning_rate": 1.7361332985515011e-06,
+      "loss": 0.803,
       "step": 28732
     },
     {
-      "epoch": 0.7891296586196479,
+      "epoch": 0.8153518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 2.242984722202588e-06,
-      "loss": 0.7503,
+      "learning_rate": 1.7356157977697052e-06,
+      "loss": 0.798,
       "step": 28733
     },
     {
-      "epoch": 0.7891571228474912,
+      "epoch": 0.8153802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 2.242423379385451e-06,
-      "loss": 0.8107,
+      "learning_rate": 1.7350983667975342e-06,
+      "loss": 0.7361,
       "step": 28734
     },
     {
-      "epoch": 0.7891845870753343,
+      "epoch": 0.8154086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 2.241862097949268e-06,
-      "loss": 0.808,
+      "learning_rate": 1.734581005639361e-06,
+      "loss": 0.8181,
       "step": 28735
     },
     {
-      "epoch": 0.7892120513031776,
+      "epoch": 0.8154370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 2.241300877898477e-06,
-      "loss": 0.7755,
+      "learning_rate": 1.7340637142995508e-06,
+      "loss": 0.869,
       "step": 28736
     },
     {
-      "epoch": 0.7892395155310209,
+      "epoch": 0.8154653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 2.2407397192375167e-06,
-      "loss": 0.7806,
+      "learning_rate": 1.7335464927824764e-06,
+      "loss": 0.7466,
       "step": 28737
     },
     {
-      "epoch": 0.7892669797588641,
+      "epoch": 0.8154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 2.240178621970829e-06,
-      "loss": 0.8604,
+      "learning_rate": 1.7330293410925091e-06,
+      "loss": 0.8153,
       "step": 28738
     },
     {
-      "epoch": 0.7892944439867073,
+      "epoch": 0.815522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.2396175861028545e-06,
-      "loss": 0.8974,
+      "learning_rate": 1.7325122592340115e-06,
+      "loss": 0.8198,
       "step": 28739
     },
     {
-      "epoch": 0.7893219082145505,
+      "epoch": 0.815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 2.2390566116380307e-06,
-      "loss": 0.905,
+      "learning_rate": 1.7319952472113554e-06,
+      "loss": 0.8027,
       "step": 28740
     },
     {
-      "epoch": 0.7893493724423938,
+      "epoch": 0.8155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.2384956985807983e-06,
-      "loss": 0.762,
+      "learning_rate": 1.7314783050289086e-06,
+      "loss": 0.6793,
       "step": 28741
     },
     {
-      "epoch": 0.7893768366702371,
+      "epoch": 0.8156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 2.2379348469355968e-06,
-      "loss": 0.8374,
+      "learning_rate": 1.730961432691034e-06,
+      "loss": 0.7462,
       "step": 28742
     },
     {
-      "epoch": 0.7894043008980802,
+      "epoch": 0.8156356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 2.237374056706859e-06,
-      "loss": 0.7829,
+      "learning_rate": 1.7304446302021005e-06,
+      "loss": 0.8334,
       "step": 28743
     },
     {
-      "epoch": 0.7894317651259235,
+      "epoch": 0.8156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 2.236813327899027e-06,
-      "loss": 0.7892,
+      "learning_rate": 1.7299278975664724e-06,
+      "loss": 0.7228,
       "step": 28744
     },
     {
-      "epoch": 0.7894592293537667,
+      "epoch": 0.8156923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 2.2362526605165324e-06,
-      "loss": 0.8018,
+      "learning_rate": 1.729411234788516e-06,
+      "loss": 0.6773,
       "step": 28745
     },
     {
-      "epoch": 0.7894866935816099,
+      "epoch": 0.8157207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 2.2356920545638137e-06,
-      "loss": 0.815,
+      "learning_rate": 1.728894641872596e-06,
+      "loss": 0.7586,
       "step": 28746
     },
     {
-      "epoch": 0.7895141578094532,
+      "epoch": 0.8157491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 2.235131510045309e-06,
-      "loss": 0.8025,
+      "learning_rate": 1.728378118823073e-06,
+      "loss": 0.8137,
       "step": 28747
     },
     {
-      "epoch": 0.7895416220372964,
+      "epoch": 0.81577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.234571026965452e-06,
-      "loss": 0.8232,
+      "learning_rate": 1.7278616656443115e-06,
+      "loss": 0.7489,
       "step": 28748
     },
     {
-      "epoch": 0.7895690862651397,
+      "epoch": 0.8158059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 2.234010605328676e-06,
-      "loss": 0.8569,
+      "learning_rate": 1.7273452823406768e-06,
+      "loss": 0.7492,
       "step": 28749
     },
     {
-      "epoch": 0.7895965504929829,
+      "epoch": 0.8158342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 2.2334502451394145e-06,
-      "loss": 0.7493,
+      "learning_rate": 1.7268289689165263e-06,
+      "loss": 0.8991,
       "step": 28750
     },
     {
-      "epoch": 0.7896240147208261,
+      "epoch": 0.8158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 2.232889946402107e-06,
-      "loss": 0.8012,
+      "learning_rate": 1.7263127253762234e-06,
+      "loss": 0.8115,
       "step": 28751
     },
     {
-      "epoch": 0.7896514789486694,
+      "epoch": 0.8158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 2.2323297091211802e-06,
-      "loss": 0.7355,
+      "learning_rate": 1.7257965517241316e-06,
+      "loss": 0.8674,
       "step": 28752
     },
     {
-      "epoch": 0.7896789431765125,
+      "epoch": 0.8159194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 2.23176953330107e-06,
-      "loss": 0.8923,
+      "learning_rate": 1.7252804479646046e-06,
+      "loss": 0.7159,
       "step": 28753
     },
     {
-      "epoch": 0.7897064074043558,
+      "epoch": 0.8159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 2.2312094189462108e-06,
-      "loss": 0.8482,
+      "learning_rate": 1.724764414102007e-06,
+      "loss": 0.7831,
       "step": 28754
     },
     {
-      "epoch": 0.7897338716321991,
+      "epoch": 0.8159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 2.2306493660610296e-06,
-      "loss": 0.8782,
+      "learning_rate": 1.7242484501406976e-06,
+      "loss": 0.8071,
       "step": 28755
     },
     {
-      "epoch": 0.7897613358600423,
+      "epoch": 0.8160045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 2.2300893746499595e-06,
-      "loss": 0.7933,
+      "learning_rate": 1.7237325560850316e-06,
+      "loss": 0.7764,
       "step": 28756
     },
     {
-      "epoch": 0.7897888000878855,
+      "epoch": 0.8160329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 2.229529444717432e-06,
-      "loss": 0.8642,
+      "learning_rate": 1.723216731939369e-06,
+      "loss": 0.8195,
       "step": 28757
     },
     {
-      "epoch": 0.7898162643157287,
+      "epoch": 0.816061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 2.2289695762678788e-06,
-      "loss": 0.77,
+      "learning_rate": 1.7227009777080694e-06,
+      "loss": 0.6754,
       "step": 28758
     },
     {
-      "epoch": 0.789843728543572,
+      "epoch": 0.8160896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 2.2284097693057305e-06,
-      "loss": 0.9567,
+      "learning_rate": 1.7221852933954809e-06,
+      "loss": 0.7911,
       "step": 28759
     },
     {
-      "epoch": 0.7898711927714153,
+      "epoch": 0.8161180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 2.2278500238354137e-06,
-      "loss": 0.8194,
+      "learning_rate": 1.7216696790059718e-06,
+      "loss": 0.8133,
       "step": 28760
     },
     {
-      "epoch": 0.7898986569992584,
+      "epoch": 0.8161464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 2.2272903398613567e-06,
-      "loss": 0.8266,
+      "learning_rate": 1.7211541345438864e-06,
+      "loss": 0.7724,
       "step": 28761
     },
     {
-      "epoch": 0.7899261212271017,
+      "epoch": 0.8161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 2.2267307173879882e-06,
-      "loss": 0.8563,
+      "learning_rate": 1.7206386600135861e-06,
+      "loss": 0.7988,
       "step": 28762
     },
     {
-      "epoch": 0.789953585454945,
+      "epoch": 0.8162031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 2.2261711564197386e-06,
-      "loss": 0.8751,
+      "learning_rate": 1.7201232554194247e-06,
+      "loss": 0.8253,
       "step": 28763
     },
     {
-      "epoch": 0.7899810496827882,
+      "epoch": 0.8162315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 2.2256116569610333e-06,
-      "loss": 0.8585,
+      "learning_rate": 1.7196079207657523e-06,
+      "loss": 0.9286,
       "step": 28764
     },
     {
-      "epoch": 0.7900085139106314,
+      "epoch": 0.8162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 2.2250522190163025e-06,
-      "loss": 0.7601,
+      "learning_rate": 1.7190926560569242e-06,
+      "loss": 0.8626,
       "step": 28765
     },
     {
-      "epoch": 0.7900359781384746,
+      "epoch": 0.8162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 2.2244928425899683e-06,
-      "loss": 0.7808,
+      "learning_rate": 1.718577461297295e-06,
+      "loss": 0.7686,
       "step": 28766
     },
     {
-      "epoch": 0.7900634423663179,
+      "epoch": 0.816316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.223933527686458e-06,
-      "loss": 0.8695,
+      "learning_rate": 1.7180623364912118e-06,
+      "loss": 0.8082,
       "step": 28767
     },
     {
-      "epoch": 0.7900909065941611,
+      "epoch": 0.8163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 2.2233742743102015e-06,
-      "loss": 0.8941,
+      "learning_rate": 1.7175472816430284e-06,
+      "loss": 0.8393,
       "step": 28768
     },
     {
-      "epoch": 0.7901183708220043,
+      "epoch": 0.8163734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 2.222815082465617e-06,
-      "loss": 0.9246,
+      "learning_rate": 1.7170322967570974e-06,
+      "loss": 0.7838,
       "step": 28769
     },
     {
-      "epoch": 0.7901458350498476,
+      "epoch": 0.8164018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 2.222255952157132e-06,
-      "loss": 0.745,
+      "learning_rate": 1.7165173818377646e-06,
+      "loss": 0.7784,
       "step": 28770
     },
     {
-      "epoch": 0.7901732992776908,
+      "epoch": 0.8164301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 2.2216968833891727e-06,
-      "loss": 0.8034,
+      "learning_rate": 1.7160025368893817e-06,
+      "loss": 0.8133,
       "step": 28771
     },
     {
-      "epoch": 0.790200763505534,
+      "epoch": 0.8164585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 2.221137876166158e-06,
-      "loss": 0.7835,
+      "learning_rate": 1.7154877619163013e-06,
+      "loss": 0.7725,
       "step": 28772
     },
     {
-      "epoch": 0.7902282277333773,
+      "epoch": 0.8164869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 2.2205789304925153e-06,
-      "loss": 0.8763,
+      "learning_rate": 1.714973056922865e-06,
+      "loss": 0.8213,
       "step": 28773
     },
     {
-      "epoch": 0.7902556919612205,
+      "epoch": 0.8165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 2.220020046372664e-06,
-      "loss": 0.7648,
+      "learning_rate": 1.7144584219134207e-06,
+      "loss": 0.8406,
       "step": 28774
     },
     {
-      "epoch": 0.7902831561890638,
+      "epoch": 0.8165437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 2.2194612238110312e-06,
-      "loss": 0.7819,
+      "learning_rate": 1.7139438568923239e-06,
+      "loss": 0.8962,
       "step": 28775
     },
     {
-      "epoch": 0.790310620416907,
+      "epoch": 0.816572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.2189024628120326e-06,
-      "loss": 0.7963,
+      "learning_rate": 1.7134293618639142e-06,
+      "loss": 0.8947,
       "step": 28776
     },
     {
-      "epoch": 0.7903380846447502,
+      "epoch": 0.8166004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 2.2183437633800942e-06,
-      "loss": 0.8677,
+      "learning_rate": 1.7129149368325382e-06,
+      "loss": 0.8894,
       "step": 28777
     },
     {
-      "epoch": 0.7903655488725935,
+      "epoch": 0.8166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 2.2177851255196313e-06,
-      "loss": 0.9063,
+      "learning_rate": 1.7124005818025447e-06,
+      "loss": 0.7708,
       "step": 28778
     },
     {
-      "epoch": 0.7903930131004366,
+      "epoch": 0.8166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.2172265492350675e-06,
-      "loss": 0.865,
+      "learning_rate": 1.7118862967782745e-06,
+      "loss": 0.7422,
       "step": 28779
     },
     {
-      "epoch": 0.7904204773282799,
+      "epoch": 0.8166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 2.2166680345308213e-06,
-      "loss": 0.892,
+      "learning_rate": 1.7113720817640723e-06,
+      "loss": 0.7889,
       "step": 28780
     },
     {
-      "epoch": 0.7904479415561232,
+      "epoch": 0.8167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 2.216109581411313e-06,
-      "loss": 0.8043,
+      "learning_rate": 1.7108579367642853e-06,
+      "loss": 0.9082,
       "step": 28781
     },
     {
-      "epoch": 0.7904754057839664,
+      "epoch": 0.8167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 2.215551189880961e-06,
-      "loss": 0.7654,
+      "learning_rate": 1.7103438617832524e-06,
+      "loss": 0.8559,
       "step": 28782
     },
     {
-      "epoch": 0.7905028700118096,
+      "epoch": 0.8167707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 2.214992859944186e-06,
-      "loss": 0.7694,
+      "learning_rate": 1.709829856825317e-06,
+      "loss": 0.8068,
       "step": 28783
     },
     {
-      "epoch": 0.7905303342396528,
+      "epoch": 0.8167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 2.2144345916054034e-06,
-      "loss": 0.7815,
+      "learning_rate": 1.709315921894823e-06,
+      "loss": 0.6693,
       "step": 28784
     },
     {
-      "epoch": 0.7905577984674961,
+      "epoch": 0.8168274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 2.213876384869027e-06,
-      "loss": 0.8329,
+      "learning_rate": 1.7088020569961094e-06,
+      "loss": 0.8105,
       "step": 28785
     },
     {
-      "epoch": 0.7905852626953394,
+      "epoch": 0.8168558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 2.213318239739477e-06,
-      "loss": 0.7984,
+      "learning_rate": 1.7082882621335184e-06,
+      "loss": 0.7527,
       "step": 28786
     },
     {
-      "epoch": 0.7906127269231825,
+      "epoch": 0.8168842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 2.2127601562211697e-06,
-      "loss": 0.8481,
+      "learning_rate": 1.7077745373113874e-06,
+      "loss": 0.7924,
       "step": 28787
     },
     {
-      "epoch": 0.7906401911510258,
+      "epoch": 0.8169125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 2.2122021343185196e-06,
-      "loss": 0.7739,
+      "learning_rate": 1.7072608825340576e-06,
+      "loss": 0.76,
       "step": 28788
     },
     {
-      "epoch": 0.7906676553788691,
+      "epoch": 0.8169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 2.211644174035946e-06,
-      "loss": 0.7054,
+      "learning_rate": 1.7067472978058685e-06,
+      "loss": 0.889,
       "step": 28789
     },
     {
-      "epoch": 0.7906951196067122,
+      "epoch": 0.8169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 2.211086275377858e-06,
-      "loss": 0.683,
+      "learning_rate": 1.706233783131157e-06,
+      "loss": 0.6926,
       "step": 28790
     },
     {
-      "epoch": 0.7907225838345555,
+      "epoch": 0.8169977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 2.210528438348676e-06,
-      "loss": 0.8712,
+      "learning_rate": 1.7057203385142618e-06,
+      "loss": 0.8905,
       "step": 28791
     },
     {
-      "epoch": 0.7907500480623987,
+      "epoch": 0.8170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 2.2099706629528063e-06,
-      "loss": 0.8174,
+      "learning_rate": 1.7052069639595225e-06,
+      "loss": 0.833,
       "step": 28792
     },
     {
-      "epoch": 0.790777512290242,
+      "epoch": 0.8170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 2.209412949194667e-06,
-      "loss": 0.7784,
+      "learning_rate": 1.7046936594712704e-06,
+      "loss": 0.8529,
       "step": 28793
     },
     {
-      "epoch": 0.7908049765180852,
+      "epoch": 0.817082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 2.2088552970786704e-06,
-      "loss": 0.9177,
+      "learning_rate": 1.7041804250538429e-06,
+      "loss": 0.779,
       "step": 28794
     },
     {
-      "epoch": 0.7908324407459284,
+      "epoch": 0.81711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.20829770660923e-06,
-      "loss": 0.7761,
+      "learning_rate": 1.70366726071158e-06,
+      "loss": 0.7727,
       "step": 28795
     },
     {
-      "epoch": 0.7908599049737717,
+      "epoch": 0.8171396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 2.207740177790756e-06,
-      "loss": 0.8525,
+      "learning_rate": 1.7031541664488093e-06,
+      "loss": 0.8409,
       "step": 28796
     },
     {
-      "epoch": 0.7908873692016148,
+      "epoch": 0.8171679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 2.207182710627659e-06,
-      "loss": 0.7838,
+      "learning_rate": 1.7026411422698685e-06,
+      "loss": 0.8178,
       "step": 28797
     },
     {
-      "epoch": 0.7909148334294581,
+      "epoch": 0.8171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 2.2066253051243503e-06,
-      "loss": 0.7161,
+      "learning_rate": 1.702128188179094e-06,
+      "loss": 0.7861,
       "step": 28798
     },
     {
-      "epoch": 0.7909422976573014,
+      "epoch": 0.8172247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 2.2060679612852443e-06,
-      "loss": 0.8782,
+      "learning_rate": 1.7016153041808125e-06,
+      "loss": 0.8607,
       "step": 28799
     },
     {
-      "epoch": 0.7909697618851446,
+      "epoch": 0.8172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 2.205510679114745e-06,
-      "loss": 0.8172,
+      "learning_rate": 1.7011024902793604e-06,
+      "loss": 0.8848,
       "step": 28800
     },
     {
-      "epoch": 0.7909972261129878,
+      "epoch": 0.8172814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 2.2049534586172672e-06,
-      "loss": 0.7433,
+      "learning_rate": 1.700589746479071e-06,
+      "loss": 0.8064,
       "step": 28801
     },
     {
-      "epoch": 0.7910246903408311,
+      "epoch": 0.8173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 2.2043962997972145e-06,
-      "loss": 0.8187,
+      "learning_rate": 1.7000770727842698e-06,
+      "loss": 0.7606,
       "step": 28802
     },
     {
-      "epoch": 0.7910521545686743,
+      "epoch": 0.8173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 2.2038392026589985e-06,
-      "loss": 0.6945,
+      "learning_rate": 1.6995644691992907e-06,
+      "loss": 0.776,
       "step": 28803
     },
     {
-      "epoch": 0.7910796187965176,
+      "epoch": 0.817366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.203282167207026e-06,
-      "loss": 0.8087,
+      "learning_rate": 1.6990519357284674e-06,
+      "loss": 0.8018,
       "step": 28804
     },
     {
-      "epoch": 0.7911070830243607,
+      "epoch": 0.8173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 2.202725193445705e-06,
-      "loss": 0.8393,
+      "learning_rate": 1.6985394723761194e-06,
+      "loss": 0.7912,
       "step": 28805
     },
     {
-      "epoch": 0.791134547252204,
+      "epoch": 0.8174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 2.202168281379443e-06,
-      "loss": 0.7738,
+      "learning_rate": 1.6980270791465868e-06,
+      "loss": 0.7783,
       "step": 28806
     },
     {
-      "epoch": 0.7911620114800473,
+      "epoch": 0.8174517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 2.201611431012649e-06,
-      "loss": 0.7975,
+      "learning_rate": 1.6975147560441908e-06,
+      "loss": 0.666,
       "step": 28807
     },
     {
-      "epoch": 0.7911894757078904,
+      "epoch": 0.8174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 2.201054642349726e-06,
-      "loss": 0.7755,
+      "learning_rate": 1.6970025030732606e-06,
+      "loss": 0.8334,
       "step": 28808
     },
     {
-      "epoch": 0.7912169399357337,
+      "epoch": 0.8175085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 2.2004979153950766e-06,
-      "loss": 0.7687,
+      "learning_rate": 1.6964903202381255e-06,
+      "loss": 0.7619,
       "step": 28809
     },
     {
-      "epoch": 0.7912444041635769,
+      "epoch": 0.8175368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 2.1999412501531103e-06,
-      "loss": 0.8771,
+      "learning_rate": 1.6959782075431076e-06,
+      "loss": 0.8234,
       "step": 28810
     },
     {
-      "epoch": 0.7912718683914202,
+      "epoch": 0.8175652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 2.1993846466282287e-06,
-      "loss": 0.8162,
+      "learning_rate": 1.6954661649925352e-06,
+      "loss": 0.7732,
       "step": 28811
     },
     {
-      "epoch": 0.7912993326192634,
+      "epoch": 0.8175936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 2.1988281048248385e-06,
-      "loss": 0.7858,
+      "learning_rate": 1.6949541925907364e-06,
+      "loss": 0.8592,
       "step": 28812
     },
     {
-      "epoch": 0.7913267968471066,
+      "epoch": 0.817622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 2.1982716247473445e-06,
-      "loss": 0.7748,
+      "learning_rate": 1.69444229034203e-06,
+      "loss": 0.836,
       "step": 28813
     },
     {
-      "epoch": 0.7913542610749499,
+      "epoch": 0.8176503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 2.197715206400145e-06,
-      "loss": 0.6696,
+      "learning_rate": 1.693930458250742e-06,
+      "loss": 0.8157,
       "step": 28814
     },
     {
-      "epoch": 0.7913817253027932,
+      "epoch": 0.8176787741203179,
       "grad_norm": 0.0,
-      "learning_rate": 2.197158849787647e-06,
-      "loss": 0.7332,
+      "learning_rate": 1.6934186963212006e-06,
+      "loss": 0.8531,
       "step": 28815
     },
     {
-      "epoch": 0.7914091895306363,
+      "epoch": 0.8177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 2.196602554914249e-06,
-      "loss": 0.6839,
+      "learning_rate": 1.6929070045577222e-06,
+      "loss": 0.8406,
       "step": 28816
     },
     {
-      "epoch": 0.7914366537584796,
+      "epoch": 0.8177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 2.1960463217843542e-06,
-      "loss": 0.847,
+      "learning_rate": 1.692395382964631e-06,
+      "loss": 0.8104,
       "step": 28817
     },
     {
-      "epoch": 0.7914641179863228,
+      "epoch": 0.8177639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 2.195490150402366e-06,
-      "loss": 0.7584,
+      "learning_rate": 1.6918838315462528e-06,
+      "loss": 0.8188,
       "step": 28818
     },
     {
-      "epoch": 0.791491582214166,
+      "epoch": 0.8177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 2.1949340407726806e-06,
-      "loss": 0.8737,
+      "learning_rate": 1.6913723503069024e-06,
+      "loss": 0.8039,
       "step": 28819
     },
     {
-      "epoch": 0.7915190464420093,
+      "epoch": 0.8178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 2.1943779928997e-06,
-      "loss": 0.7974,
+      "learning_rate": 1.6908609392509035e-06,
+      "loss": 0.9014,
       "step": 28820
     },
     {
-      "epoch": 0.7915465106698525,
+      "epoch": 0.8178490351872871,
       "grad_norm": 0.0,
-      "learning_rate": 2.1938220067878246e-06,
-      "loss": 0.8445,
+      "learning_rate": 1.6903495983825756e-06,
+      "loss": 0.7941,
       "step": 28821
     },
     {
-      "epoch": 0.7915739748976958,
+      "epoch": 0.8178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 2.193266082441453e-06,
-      "loss": 0.8174,
+      "learning_rate": 1.689838327706238e-06,
+      "loss": 0.8544,
       "step": 28822
     },
     {
-      "epoch": 0.7916014391255389,
+      "epoch": 0.817905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 2.1927102198649876e-06,
-      "loss": 0.8244,
+      "learning_rate": 1.6893271272262123e-06,
+      "loss": 0.8802,
       "step": 28823
     },
     {
-      "epoch": 0.7916289033533822,
+      "epoch": 0.8179341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 2.192154419062823e-06,
-      "loss": 0.8184,
+      "learning_rate": 1.688815996946812e-06,
+      "loss": 0.8101,
       "step": 28824
     },
     {
-      "epoch": 0.7916563675812255,
+      "epoch": 0.8179625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 2.191598680039354e-06,
-      "loss": 0.8662,
+      "learning_rate": 1.688304936872356e-06,
+      "loss": 0.8124,
       "step": 28825
     },
     {
-      "epoch": 0.7916838318090686,
+      "epoch": 0.8179909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 2.1910430027989814e-06,
-      "loss": 0.8673,
+      "learning_rate": 1.6877939470071648e-06,
+      "loss": 0.8255,
       "step": 28826
     },
     {
-      "epoch": 0.7917112960369119,
+      "epoch": 0.8180192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 2.190487387346102e-06,
-      "loss": 0.8839,
+      "learning_rate": 1.6872830273555485e-06,
+      "loss": 0.7982,
       "step": 28827
     },
     {
-      "epoch": 0.7917387602647552,
+      "epoch": 0.8180476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 2.1899318336851105e-06,
-      "loss": 0.7428,
+      "learning_rate": 1.686772177921826e-06,
+      "loss": 0.7743,
       "step": 28828
     },
     {
-      "epoch": 0.7917662244925984,
+      "epoch": 0.8180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 2.189376341820404e-06,
-      "loss": 0.8056,
+      "learning_rate": 1.6862613987103161e-06,
+      "loss": 0.8476,
       "step": 28829
     },
     {
-      "epoch": 0.7917936887204416,
+      "epoch": 0.8181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.1888209117563795e-06,
-      "loss": 0.8114,
+      "learning_rate": 1.6857506897253272e-06,
+      "loss": 0.7983,
       "step": 28830
     },
     {
-      "epoch": 0.7918211529482848,
+      "epoch": 0.8181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 2.1882655434974276e-06,
-      "loss": 0.7441,
+      "learning_rate": 1.6852400509711754e-06,
+      "loss": 0.7176,
       "step": 28831
     },
     {
-      "epoch": 0.7918486171761281,
+      "epoch": 0.818161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 2.1877102370479476e-06,
-      "loss": 0.7789,
+      "learning_rate": 1.6847294824521777e-06,
+      "loss": 0.8805,
       "step": 28832
     },
     {
-      "epoch": 0.7918760814039714,
+      "epoch": 0.8181895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 2.1871549924123282e-06,
-      "loss": 0.8449,
+      "learning_rate": 1.684218984172641e-06,
+      "loss": 0.8474,
       "step": 28833
     },
     {
-      "epoch": 0.7919035456318145,
+      "epoch": 0.8182179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 2.1865998095949636e-06,
-      "loss": 0.7881,
+      "learning_rate": 1.6837085561368805e-06,
+      "loss": 0.7689,
       "step": 28834
     },
     {
-      "epoch": 0.7919310098596578,
+      "epoch": 0.8182463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 2.1860446886002484e-06,
-      "loss": 0.792,
+      "learning_rate": 1.6831981983492074e-06,
+      "loss": 0.8593,
       "step": 28835
     },
     {
-      "epoch": 0.7919584740875011,
+      "epoch": 0.8182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 2.1854896294325777e-06,
-      "loss": 0.8731,
+      "learning_rate": 1.6826879108139338e-06,
+      "loss": 0.7733,
       "step": 28836
     },
     {
-      "epoch": 0.7919859383153443,
+      "epoch": 0.8183030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 2.184934632096337e-06,
-      "loss": 0.7701,
+      "learning_rate": 1.6821776935353706e-06,
+      "loss": 0.813,
       "step": 28837
     },
     {
-      "epoch": 0.7920134025431875,
+      "epoch": 0.8183314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 2.184379696595921e-06,
-      "loss": 0.8683,
+      "learning_rate": 1.6816675465178257e-06,
+      "loss": 0.9406,
       "step": 28838
     },
     {
-      "epoch": 0.7920408667710307,
+      "epoch": 0.8183598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 2.1838248229357217e-06,
-      "loss": 0.789,
+      "learning_rate": 1.6811574697656075e-06,
+      "loss": 0.8654,
       "step": 28839
     },
     {
-      "epoch": 0.792068330998874,
+      "epoch": 0.8183881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 2.1832700111201267e-06,
-      "loss": 0.8452,
+      "learning_rate": 1.6806474632830284e-06,
+      "loss": 0.7558,
       "step": 28840
     },
     {
-      "epoch": 0.7920957952267172,
+      "epoch": 0.818416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 2.1827152611535273e-06,
-      "loss": 0.8076,
+      "learning_rate": 1.6801375270743925e-06,
+      "loss": 0.7871,
       "step": 28841
     },
     {
-      "epoch": 0.7921232594545604,
+      "epoch": 0.81844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.182160573040315e-06,
-      "loss": 0.751,
+      "learning_rate": 1.6796276611440088e-06,
+      "loss": 0.738,
       "step": 28842
     },
     {
-      "epoch": 0.7921507236824037,
+      "epoch": 0.8184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 2.181605946784875e-06,
-      "loss": 0.8018,
+      "learning_rate": 1.679117865496186e-06,
+      "loss": 0.875,
       "step": 28843
     },
     {
-      "epoch": 0.7921781879102469,
+      "epoch": 0.8185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 2.181051382391596e-06,
-      "loss": 0.8869,
+      "learning_rate": 1.6786081401352271e-06,
+      "loss": 0.6652,
       "step": 28844
     },
     {
-      "epoch": 0.7922056521380901,
+      "epoch": 0.8185300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 2.180496879864867e-06,
-      "loss": 0.8734,
+      "learning_rate": 1.6780984850654392e-06,
+      "loss": 0.8408,
       "step": 28845
     },
     {
-      "epoch": 0.7922331163659334,
+      "epoch": 0.8185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 2.1799424392090774e-06,
-      "loss": 0.9113,
+      "learning_rate": 1.6775889002911295e-06,
+      "loss": 0.8068,
       "step": 28846
     },
     {
-      "epoch": 0.7922605805937766,
+      "epoch": 0.8185868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 2.1793880604286145e-06,
-      "loss": 0.8979,
+      "learning_rate": 1.6770793858165979e-06,
+      "loss": 0.9044,
       "step": 28847
     },
     {
-      "epoch": 0.7922880448216199,
+      "epoch": 0.8186152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 2.1788337435278615e-06,
-      "loss": 0.8625,
+      "learning_rate": 1.6765699416461511e-06,
+      "loss": 0.8088,
       "step": 28848
     },
     {
-      "epoch": 0.7923155090494631,
+      "epoch": 0.8186435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 2.178279488511205e-06,
-      "loss": 0.7355,
+      "learning_rate": 1.6760605677840947e-06,
+      "loss": 0.7766,
       "step": 28849
     },
     {
-      "epoch": 0.7923429732773063,
+      "epoch": 0.8186719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 2.1777252953830286e-06,
-      "loss": 0.8759,
+      "learning_rate": 1.675551264234726e-06,
+      "loss": 0.7191,
       "step": 28850
     },
     {
-      "epoch": 0.7923704375051496,
+      "epoch": 0.8187003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 2.177171164147721e-06,
-      "loss": 0.6939,
+      "learning_rate": 1.6750420310023507e-06,
+      "loss": 0.8225,
       "step": 28851
     },
     {
-      "epoch": 0.7923979017329927,
+      "epoch": 0.8187287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 2.1766170948096653e-06,
-      "loss": 0.8511,
+      "learning_rate": 1.6745328680912697e-06,
+      "loss": 0.8265,
       "step": 28852
     },
     {
-      "epoch": 0.792425365960836,
+      "epoch": 0.8187570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 2.1760630873732457e-06,
-      "loss": 0.7894,
+      "learning_rate": 1.674023775505783e-06,
+      "loss": 0.9151,
       "step": 28853
     },
     {
-      "epoch": 0.7924528301886793,
+      "epoch": 0.8187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 2.175509141842849e-06,
-      "loss": 0.8906,
+      "learning_rate": 1.6735147532501949e-06,
+      "loss": 0.8334,
       "step": 28854
     },
     {
-      "epoch": 0.7924802944165225,
+      "epoch": 0.8188138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 2.1749552582228506e-06,
-      "loss": 0.8813,
+      "learning_rate": 1.6730058013287986e-06,
+      "loss": 0.83,
       "step": 28855
     },
     {
-      "epoch": 0.7925077586443657,
+      "epoch": 0.8188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.1744014365176413e-06,
-      "loss": 0.8148,
+      "learning_rate": 1.6724969197458972e-06,
+      "loss": 0.7891,
       "step": 28856
     },
     {
-      "epoch": 0.7925352228722089,
+      "epoch": 0.8188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 2.173847676731595e-06,
-      "loss": 0.8018,
+      "learning_rate": 1.671988108505792e-06,
+      "loss": 0.9021,
       "step": 28857
     },
     {
-      "epoch": 0.7925626871000522,
+      "epoch": 0.8188989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 2.1732939788690986e-06,
-      "loss": 0.8985,
+      "learning_rate": 1.671479367612774e-06,
+      "loss": 0.8219,
       "step": 28858
     },
     {
-      "epoch": 0.7925901513278955,
+      "epoch": 0.8189273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 2.1727403429345327e-06,
-      "loss": 0.7939,
+      "learning_rate": 1.6709706970711447e-06,
+      "loss": 0.7821,
       "step": 28859
     },
     {
-      "epoch": 0.7926176155557386,
+      "epoch": 0.818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.1721867689322783e-06,
-      "loss": 0.8245,
+      "learning_rate": 1.6704620968852027e-06,
+      "loss": 0.6814,
       "step": 28860
     },
     {
-      "epoch": 0.7926450797835819,
+      "epoch": 0.8189841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 2.171633256866713e-06,
-      "loss": 0.8272,
+      "learning_rate": 1.6699535670592393e-06,
+      "loss": 0.7544,
       "step": 28861
     },
     {
-      "epoch": 0.7926725440114252,
+      "epoch": 0.8190124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 2.171079806742217e-06,
-      "loss": 0.8487,
+      "learning_rate": 1.6694451075975526e-06,
+      "loss": 0.7027,
       "step": 28862
     },
     {
-      "epoch": 0.7927000082392683,
+      "epoch": 0.8190408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 2.1705264185631737e-06,
-      "loss": 0.7252,
+      "learning_rate": 1.6689367185044393e-06,
+      "loss": 0.8777,
       "step": 28863
     },
     {
-      "epoch": 0.7927274724671116,
+      "epoch": 0.8190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 2.169973092333955e-06,
-      "loss": 0.769,
+      "learning_rate": 1.6684283997841899e-06,
+      "loss": 0.726,
       "step": 28864
     },
     {
-      "epoch": 0.7927549366949548,
+      "epoch": 0.8190976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 2.169419828058942e-06,
-      "loss": 0.7389,
+      "learning_rate": 1.6679201514411014e-06,
+      "loss": 0.6855,
       "step": 28865
     },
     {
-      "epoch": 0.7927824009227981,
+      "epoch": 0.8191259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 2.1688666257425163e-06,
-      "loss": 0.8595,
+      "learning_rate": 1.6674119734794647e-06,
+      "loss": 0.8708,
       "step": 28866
     },
     {
-      "epoch": 0.7928098651506413,
+      "epoch": 0.8191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 2.1683134853890475e-06,
-      "loss": 0.7427,
+      "learning_rate": 1.6669038659035741e-06,
+      "loss": 0.7579,
       "step": 28867
     },
     {
-      "epoch": 0.7928373293784845,
+      "epoch": 0.8191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 2.1677604070029166e-06,
-      "loss": 0.8649,
+      "learning_rate": 1.666395828717724e-06,
+      "loss": 0.8847,
       "step": 28868
     },
     {
-      "epoch": 0.7928647936063278,
+      "epoch": 0.819211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.1672073905884983e-06,
-      "loss": 0.8191,
+      "learning_rate": 1.6658878619261997e-06,
+      "loss": 0.8089,
       "step": 28869
     },
     {
-      "epoch": 0.7928922578341709,
+      "epoch": 0.8192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 2.1666544361501707e-06,
-      "loss": 0.8398,
+      "learning_rate": 1.6653799655332958e-06,
+      "loss": 0.8087,
       "step": 28870
     },
     {
-      "epoch": 0.7929197220620142,
+      "epoch": 0.8192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 2.1661015436923093e-06,
-      "loss": 0.8449,
+      "learning_rate": 1.6648721395433043e-06,
+      "loss": 0.8572,
       "step": 28871
     },
     {
-      "epoch": 0.7929471862898575,
+      "epoch": 0.8192962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 2.1655487132192878e-06,
-      "loss": 0.8069,
+      "learning_rate": 1.6643643839605106e-06,
+      "loss": 0.7725,
       "step": 28872
     },
     {
-      "epoch": 0.7929746505177007,
+      "epoch": 0.8193246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 2.1649959447354763e-06,
-      "loss": 0.7952,
+      "learning_rate": 1.6638566987892057e-06,
+      "loss": 0.7593,
       "step": 28873
     },
     {
-      "epoch": 0.7930021147455439,
+      "epoch": 0.8193530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 2.164443238245253e-06,
-      "loss": 0.9033,
+      "learning_rate": 1.6633490840336797e-06,
+      "loss": 0.7855,
       "step": 28874
     },
     {
-      "epoch": 0.7930295789733872,
+      "epoch": 0.8193813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 2.163890593752989e-06,
-      "loss": 0.824,
+      "learning_rate": 1.6628415396982167e-06,
+      "loss": 0.8302,
       "step": 28875
     },
     {
-      "epoch": 0.7930570432012304,
+      "epoch": 0.8194097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 2.1633380112630575e-06,
-      "loss": 0.82,
+      "learning_rate": 1.6623340657871057e-06,
+      "loss": 0.8619,
       "step": 28876
     },
     {
-      "epoch": 0.7930845074290737,
+      "epoch": 0.8194381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 2.162785490779833e-06,
-      "loss": 0.8965,
+      "learning_rate": 1.6618266623046363e-06,
+      "loss": 0.7785,
       "step": 28877
     },
     {
-      "epoch": 0.7931119716569168,
+      "epoch": 0.819466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 2.1622330323076866e-06,
-      "loss": 0.7681,
+      "learning_rate": 1.6613193292550888e-06,
+      "loss": 0.7882,
       "step": 28878
     },
     {
-      "epoch": 0.7931394358847601,
+      "epoch": 0.8194948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 2.1616806358509866e-06,
-      "loss": 0.8033,
+      "learning_rate": 1.660812066642753e-06,
+      "loss": 0.8967,
       "step": 28879
     },
     {
-      "epoch": 0.7931669001126034,
+      "epoch": 0.8195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 2.161128301414108e-06,
-      "loss": 0.7743,
+      "learning_rate": 1.6603048744719141e-06,
+      "loss": 0.8016,
       "step": 28880
     },
     {
-      "epoch": 0.7931943643404465,
+      "epoch": 0.8195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 2.1605760290014165e-06,
-      "loss": 0.8151,
+      "learning_rate": 1.65979775274685e-06,
+      "loss": 0.758,
       "step": 28881
     },
     {
-      "epoch": 0.7932218285682898,
+      "epoch": 0.8195800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 2.160023818617284e-06,
-      "loss": 0.7802,
+      "learning_rate": 1.6592907014718518e-06,
+      "loss": 0.7207,
       "step": 28882
     },
     {
-      "epoch": 0.793249292796133,
+      "epoch": 0.8196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 2.1594716702660825e-06,
-      "loss": 0.7716,
+      "learning_rate": 1.6587837206512025e-06,
+      "loss": 0.9282,
       "step": 28883
     },
     {
-      "epoch": 0.7932767570239763,
+      "epoch": 0.8196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 2.1589195839521758e-06,
-      "loss": 0.8711,
+      "learning_rate": 1.65827681028918e-06,
+      "loss": 0.8029,
       "step": 28884
     },
     {
-      "epoch": 0.7933042212518195,
+      "epoch": 0.8196651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 2.1583675596799346e-06,
-      "loss": 0.7468,
+      "learning_rate": 1.6577699703900686e-06,
+      "loss": 0.8986,
       "step": 28885
     },
     {
-      "epoch": 0.7933316854796627,
+      "epoch": 0.8196935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 2.1578155974537272e-06,
-      "loss": 0.7659,
+      "learning_rate": 1.6572632009581502e-06,
+      "loss": 0.8625,
       "step": 28886
     },
     {
-      "epoch": 0.793359149707506,
+      "epoch": 0.8197219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 2.1572636972779225e-06,
-      "loss": 0.7944,
+      "learning_rate": 1.6567565019977028e-06,
+      "loss": 0.7989,
       "step": 28887
     },
     {
-      "epoch": 0.7933866139353493,
+      "epoch": 0.8197502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 2.1567118591568836e-06,
-      "loss": 0.8446,
+      "learning_rate": 1.6562498735130084e-06,
+      "loss": 0.8698,
       "step": 28888
     },
     {
-      "epoch": 0.7934140781631924,
+      "epoch": 0.8197786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 2.1561600830949803e-06,
-      "loss": 0.8071,
+      "learning_rate": 1.6557433155083491e-06,
+      "loss": 0.8143,
       "step": 28889
     },
     {
-      "epoch": 0.7934415423910357,
+      "epoch": 0.8198070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 2.1556083690965757e-06,
-      "loss": 0.7606,
+      "learning_rate": 1.6552368279879982e-06,
+      "loss": 0.8221,
       "step": 28890
     },
     {
-      "epoch": 0.7934690066188789,
+      "epoch": 0.8198354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 2.1550567171660353e-06,
-      "loss": 0.8494,
+      "learning_rate": 1.654730410956238e-06,
+      "loss": 0.8582,
       "step": 28891
     },
     {
-      "epoch": 0.7934964708467221,
+      "epoch": 0.8198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 2.1545051273077245e-06,
-      "loss": 0.8053,
+      "learning_rate": 1.654224064417347e-06,
+      "loss": 0.7686,
       "step": 28892
     },
     {
-      "epoch": 0.7935239350745654,
+      "epoch": 0.8198921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 2.1539535995260087e-06,
-      "loss": 0.8049,
+      "learning_rate": 1.6537177883755974e-06,
+      "loss": 0.735,
       "step": 28893
     },
     {
-      "epoch": 0.7935513993024086,
+      "epoch": 0.8199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 2.153402133825251e-06,
-      "loss": 0.7817,
+      "learning_rate": 1.6532115828352712e-06,
+      "loss": 0.802,
       "step": 28894
     },
     {
-      "epoch": 0.7935788635302519,
+      "epoch": 0.8199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 2.152850730209819e-06,
-      "loss": 0.8335,
+      "learning_rate": 1.65270544780064e-06,
+      "loss": 0.8196,
       "step": 28895
     },
     {
-      "epoch": 0.793606327758095,
+      "epoch": 0.819977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 2.1522993886840706e-06,
-      "loss": 0.864,
+      "learning_rate": 1.652199383275982e-06,
+      "loss": 0.8208,
       "step": 28896
     },
     {
-      "epoch": 0.7936337919859383,
+      "epoch": 0.820005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 2.1517481092523663e-06,
-      "loss": 0.819,
+      "learning_rate": 1.6516933892655694e-06,
+      "loss": 0.8264,
       "step": 28897
     },
     {
-      "epoch": 0.7936612562137816,
+      "epoch": 0.8200340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 2.151196891919072e-06,
-      "loss": 0.8092,
+      "learning_rate": 1.6511874657736792e-06,
+      "loss": 0.8351,
       "step": 28898
     },
     {
-      "epoch": 0.7936887204416248,
+      "epoch": 0.8200624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 2.150645736688548e-06,
-      "loss": 0.8447,
+      "learning_rate": 1.6506816128045832e-06,
+      "loss": 0.6967,
       "step": 28899
     },
     {
-      "epoch": 0.793716184669468,
+      "epoch": 0.8200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 2.150094643565155e-06,
-      "loss": 0.797,
+      "learning_rate": 1.6501758303625581e-06,
+      "loss": 0.7883,
       "step": 28900
     },
     {
-      "epoch": 0.7937436488973113,
+      "epoch": 0.8201191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 2.1495436125532575e-06,
-      "loss": 0.8217,
+      "learning_rate": 1.6496701184518704e-06,
+      "loss": 0.8074,
       "step": 28901
     },
     {
-      "epoch": 0.7937711131251545,
+      "epoch": 0.8201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 2.1489926436572084e-06,
-      "loss": 0.8238,
+      "learning_rate": 1.6491644770767946e-06,
+      "loss": 0.8628,
       "step": 28902
     },
     {
-      "epoch": 0.7937985773529977,
+      "epoch": 0.8201759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 2.1484417368813715e-06,
-      "loss": 0.8446,
+      "learning_rate": 1.6486589062416037e-06,
+      "loss": 0.848,
       "step": 28903
     },
     {
-      "epoch": 0.7938260415808409,
+      "epoch": 0.8202043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 2.147890892230107e-06,
-      "loss": 0.8297,
+      "learning_rate": 1.6481534059505643e-06,
+      "loss": 0.8073,
       "step": 28904
     },
     {
-      "epoch": 0.7938535058086842,
+      "epoch": 0.8202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 2.1473401097077683e-06,
-      "loss": 0.9061,
+      "learning_rate": 1.647647976207949e-06,
+      "loss": 0.7482,
       "step": 28905
     },
     {
-      "epoch": 0.7938809700365275,
+      "epoch": 0.820261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 2.1467893893187176e-06,
-      "loss": 0.8585,
+      "learning_rate": 1.6471426170180282e-06,
+      "loss": 0.8673,
       "step": 28906
     },
     {
-      "epoch": 0.7939084342643706,
+      "epoch": 0.8202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 2.1462387310673127e-06,
-      "loss": 0.8044,
+      "learning_rate": 1.6466373283850679e-06,
+      "loss": 0.8219,
       "step": 28907
     },
     {
-      "epoch": 0.7939358984922139,
+      "epoch": 0.8203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 2.1456881349579074e-06,
-      "loss": 0.7729,
+      "learning_rate": 1.6461321103133366e-06,
+      "loss": 0.7483,
       "step": 28908
     },
     {
-      "epoch": 0.7939633627200571,
+      "epoch": 0.8203461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 2.1451376009948597e-06,
-      "loss": 0.8996,
+      "learning_rate": 1.6456269628071052e-06,
+      "loss": 0.8094,
       "step": 28909
     },
     {
-      "epoch": 0.7939908269479004,
+      "epoch": 0.8203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 2.144587129182527e-06,
-      "loss": 0.9028,
+      "learning_rate": 1.6451218858706374e-06,
+      "loss": 0.8767,
       "step": 28910
     },
     {
-      "epoch": 0.7940182911757436,
+      "epoch": 0.8204029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 2.1440367195252655e-06,
-      "loss": 0.8493,
+      "learning_rate": 1.6446168795081997e-06,
+      "loss": 0.7249,
       "step": 28911
     },
     {
-      "epoch": 0.7940457554035868,
+      "epoch": 0.8204313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 2.1434863720274257e-06,
-      "loss": 0.8525,
+      "learning_rate": 1.6441119437240582e-06,
+      "loss": 0.9157,
       "step": 28912
     },
     {
-      "epoch": 0.7940732196314301,
+      "epoch": 0.8204597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 2.142936086693368e-06,
-      "loss": 0.7166,
+      "learning_rate": 1.64360707852248e-06,
+      "loss": 0.8975,
       "step": 28913
     },
     {
-      "epoch": 0.7941006838592733,
+      "epoch": 0.8204880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 2.142385863527442e-06,
-      "loss": 0.868,
+      "learning_rate": 1.6431022839077293e-06,
+      "loss": 0.8563,
       "step": 28914
     },
     {
-      "epoch": 0.7941281480871165,
+      "epoch": 0.820516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 2.141835702534001e-06,
-      "loss": 0.8381,
+      "learning_rate": 1.6425975598840683e-06,
+      "loss": 0.8034,
       "step": 28915
     },
     {
-      "epoch": 0.7941556123149598,
+      "epoch": 0.8205448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 2.141285603717401e-06,
-      "loss": 0.8004,
+      "learning_rate": 1.6420929064557611e-06,
+      "loss": 0.7115,
       "step": 28916
     },
     {
-      "epoch": 0.794183076542803,
+      "epoch": 0.8205732122587969,
       "grad_norm": 0.0,
-      "learning_rate": 2.140735567081993e-06,
-      "loss": 0.8596,
+      "learning_rate": 1.6415883236270724e-06,
+      "loss": 0.7402,
       "step": 28917
     },
     {
-      "epoch": 0.7942105407706462,
+      "epoch": 0.8206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 2.140185592632129e-06,
-      "loss": 0.7984,
+      "learning_rate": 1.641083811402262e-06,
+      "loss": 0.8365,
       "step": 28918
     },
     {
-      "epoch": 0.7942380049984895,
+      "epoch": 0.8206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 2.139635680372165e-06,
-      "loss": 0.8843,
+      "learning_rate": 1.640579369785591e-06,
+      "loss": 0.9041,
       "step": 28919
     },
     {
-      "epoch": 0.7942654692263327,
+      "epoch": 0.8206583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 2.1390858303064477e-06,
-      "loss": 0.7653,
+      "learning_rate": 1.6400749987813247e-06,
+      "loss": 0.8162,
       "step": 28920
     },
     {
-      "epoch": 0.794292933454176,
+      "epoch": 0.8206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 2.1385360424393264e-06,
-      "loss": 0.7977,
+      "learning_rate": 1.6395706983937186e-06,
+      "loss": 0.8105,
       "step": 28921
     },
     {
-      "epoch": 0.7943203976820191,
+      "epoch": 0.8207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 2.1379863167751525e-06,
-      "loss": 0.878,
+      "learning_rate": 1.6390664686270342e-06,
+      "loss": 0.9146,
       "step": 28922
     },
     {
-      "epoch": 0.7943478619098624,
+      "epoch": 0.8207434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 2.137436653318277e-06,
-      "loss": 0.8333,
+      "learning_rate": 1.6385623094855341e-06,
+      "loss": 0.9146,
       "step": 28923
     },
     {
-      "epoch": 0.7943753261377057,
+      "epoch": 0.8207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 2.1368870520730467e-06,
-      "loss": 0.874,
+      "learning_rate": 1.6380582209734707e-06,
+      "loss": 0.7717,
       "step": 28924
     },
     {
-      "epoch": 0.7944027903655488,
+      "epoch": 0.820800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 2.136337513043816e-06,
-      "loss": 0.9091,
+      "learning_rate": 1.6375542030951063e-06,
+      "loss": 0.7461,
       "step": 28925
     },
     {
-      "epoch": 0.7944302545933921,
+      "epoch": 0.8208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 2.135788036234926e-06,
-      "loss": 0.7398,
+      "learning_rate": 1.6370502558546997e-06,
+      "loss": 0.8097,
       "step": 28926
     },
     {
-      "epoch": 0.7944577188212354,
+      "epoch": 0.8208569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 2.1352386216507302e-06,
-      "loss": 0.8301,
+      "learning_rate": 1.6365463792565005e-06,
+      "loss": 0.8799,
       "step": 28927
     },
     {
-      "epoch": 0.7944851830490786,
+      "epoch": 0.8208853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 2.13468926929557e-06,
-      "loss": 0.7373,
+      "learning_rate": 1.6360425733047758e-06,
+      "loss": 0.8374,
       "step": 28928
     },
     {
-      "epoch": 0.7945126472769218,
+      "epoch": 0.8209137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 2.134139979173795e-06,
-      "loss": 0.7984,
+      "learning_rate": 1.6355388380037717e-06,
+      "loss": 0.7515,
       "step": 28929
     },
     {
-      "epoch": 0.794540111504765,
+      "epoch": 0.8209421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 2.1335907512897524e-06,
-      "loss": 0.7459,
+      "learning_rate": 1.6350351733577474e-06,
+      "loss": 0.8645,
       "step": 28930
     },
     {
-      "epoch": 0.7945675757326083,
+      "epoch": 0.8209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 2.133041585647788e-06,
-      "loss": 0.8091,
+      "learning_rate": 1.6345315793709592e-06,
+      "loss": 0.7822,
       "step": 28931
     },
     {
-      "epoch": 0.7945950399604516,
+      "epoch": 0.8209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 2.1324924822522443e-06,
-      "loss": 0.9167,
+      "learning_rate": 1.6340280560476573e-06,
+      "loss": 0.8202,
       "step": 28932
     },
     {
-      "epoch": 0.7946225041882947,
+      "epoch": 0.8210272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 2.131943441107468e-06,
-      "loss": 0.8655,
+      "learning_rate": 1.6335246033920959e-06,
+      "loss": 0.8544,
       "step": 28933
     },
     {
-      "epoch": 0.794649968416138,
+      "epoch": 0.82105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.131394462217802e-06,
-      "loss": 0.83,
+      "learning_rate": 1.6330212214085306e-06,
+      "loss": 0.8014,
       "step": 28934
     },
     {
-      "epoch": 0.7946774326439812,
+      "epoch": 0.8210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 2.1308455455875943e-06,
-      "loss": 0.8602,
+      "learning_rate": 1.632517910101209e-06,
+      "loss": 0.8298,
       "step": 28935
     },
     {
-      "epoch": 0.7947048968718244,
+      "epoch": 0.8211123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 2.130296691221182e-06,
-      "loss": 0.7478,
+      "learning_rate": 1.6320146694743844e-06,
+      "loss": 0.9295,
       "step": 28936
     },
     {
-      "epoch": 0.7947323610996677,
+      "epoch": 0.8211407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 2.1297478991229137e-06,
-      "loss": 0.7002,
+      "learning_rate": 1.6315114995323111e-06,
+      "loss": 0.7741,
       "step": 28937
     },
     {
-      "epoch": 0.7947598253275109,
+      "epoch": 0.8211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 2.129199169297125e-06,
-      "loss": 0.831,
+      "learning_rate": 1.6310084002792336e-06,
+      "loss": 0.767,
       "step": 28938
     },
     {
-      "epoch": 0.7947872895553542,
+      "epoch": 0.8211975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 2.1286505017481617e-06,
-      "loss": 0.8601,
+      "learning_rate": 1.6305053717194041e-06,
+      "loss": 0.7537,
       "step": 28939
     },
     {
-      "epoch": 0.7948147537831974,
+      "epoch": 0.8212258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 2.1281018964803656e-06,
-      "loss": 0.8449,
+      "learning_rate": 1.6300024138570748e-06,
+      "loss": 0.794,
       "step": 28940
     },
     {
-      "epoch": 0.7948422180110406,
+      "epoch": 0.8212542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 2.127553353498075e-06,
-      "loss": 0.791,
+      "learning_rate": 1.6294995266964885e-06,
+      "loss": 0.7827,
       "step": 28941
     },
     {
-      "epoch": 0.7948696822388839,
+      "epoch": 0.8212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 2.127004872805636e-06,
-      "loss": 0.7807,
+      "learning_rate": 1.6289967102418969e-06,
+      "loss": 0.847,
       "step": 28942
     },
     {
-      "epoch": 0.794897146466727,
+      "epoch": 0.821311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 2.1264564544073806e-06,
-      "loss": 0.8041,
+      "learning_rate": 1.6284939644975451e-06,
+      "loss": 0.8771,
       "step": 28943
     },
     {
-      "epoch": 0.7949246106945703,
+      "epoch": 0.821339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 2.1259080983076553e-06,
-      "loss": 0.8595,
+      "learning_rate": 1.627991289467682e-06,
+      "loss": 0.8144,
       "step": 28944
     },
     {
-      "epoch": 0.7949520749224136,
+      "epoch": 0.8213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 2.1253598045107917e-06,
-      "loss": 0.7883,
+      "learning_rate": 1.6274886851565552e-06,
+      "loss": 0.872,
       "step": 28945
     },
     {
-      "epoch": 0.7949795391502568,
+      "epoch": 0.8213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 2.1248115730211317e-06,
-      "loss": 0.8252,
+      "learning_rate": 1.626986151568405e-06,
+      "loss": 0.8341,
       "step": 28946
     },
     {
-      "epoch": 0.7950070033781,
+      "epoch": 0.8214245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 2.1242634038430134e-06,
-      "loss": 0.7879,
+      "learning_rate": 1.6264836887074797e-06,
+      "loss": 0.838,
       "step": 28947
     },
     {
-      "epoch": 0.7950344676059432,
+      "epoch": 0.8214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 2.123715296980776e-06,
-      "loss": 0.6795,
+      "learning_rate": 1.6259812965780253e-06,
+      "loss": 0.9161,
       "step": 28948
     },
     {
-      "epoch": 0.7950619318337865,
+      "epoch": 0.8214812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 2.1231672524387525e-06,
-      "loss": 0.9348,
+      "learning_rate": 1.6254789751842803e-06,
+      "loss": 0.8172,
       "step": 28949
     },
     {
-      "epoch": 0.7950893960616298,
+      "epoch": 0.8215096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 2.12261927022128e-06,
-      "loss": 0.8093,
+      "learning_rate": 1.6249767245304914e-06,
+      "loss": 0.7733,
       "step": 28950
     },
     {
-      "epoch": 0.7951168602894729,
+      "epoch": 0.8215380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 2.1220713503326985e-06,
-      "loss": 0.9398,
+      "learning_rate": 1.6244745446209032e-06,
+      "loss": 0.7946,
       "step": 28951
     },
     {
-      "epoch": 0.7951443245173162,
+      "epoch": 0.8215664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 2.1215234927773366e-06,
-      "loss": 0.835,
+      "learning_rate": 1.623972435459752e-06,
+      "loss": 0.8065,
       "step": 28952
     },
     {
-      "epoch": 0.7951717887451595,
+      "epoch": 0.8215947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 2.120975697559533e-06,
-      "loss": 0.7754,
+      "learning_rate": 1.6234703970512822e-06,
+      "loss": 0.7663,
       "step": 28953
     },
     {
-      "epoch": 0.7951992529730026,
+      "epoch": 0.8216231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 2.1204279646836245e-06,
-      "loss": 0.7154,
+      "learning_rate": 1.6229684293997372e-06,
+      "loss": 0.7116,
       "step": 28954
     },
     {
-      "epoch": 0.7952267172008459,
+      "epoch": 0.8216515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 2.11988029415394e-06,
-      "loss": 0.8596,
+      "learning_rate": 1.6224665325093525e-06,
+      "loss": 0.8336,
       "step": 28955
     },
     {
-      "epoch": 0.7952541814286891,
+      "epoch": 0.8216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 2.1193326859748143e-06,
-      "loss": 0.8073,
+      "learning_rate": 1.6219647063843691e-06,
+      "loss": 0.8703,
       "step": 28956
     },
     {
-      "epoch": 0.7952816456565324,
+      "epoch": 0.8217082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 2.118785140150581e-06,
-      "loss": 0.8062,
+      "learning_rate": 1.6214629510290291e-06,
+      "loss": 0.8745,
       "step": 28957
     },
     {
-      "epoch": 0.7953091098843756,
+      "epoch": 0.8217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.1182376566855733e-06,
-      "loss": 0.7559,
+      "learning_rate": 1.6209612664475639e-06,
+      "loss": 0.807,
       "step": 28958
     },
     {
-      "epoch": 0.7953365741122188,
+      "epoch": 0.8217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 2.1176902355841244e-06,
-      "loss": 0.7737,
+      "learning_rate": 1.6204596526442195e-06,
+      "loss": 0.8484,
       "step": 28959
     },
     {
-      "epoch": 0.7953640383400621,
+      "epoch": 0.8217934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 2.1171428768505644e-06,
-      "loss": 0.7984,
+      "learning_rate": 1.6199581096232264e-06,
+      "loss": 0.7628,
       "step": 28960
     },
     {
-      "epoch": 0.7953915025679053,
+      "epoch": 0.821821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 2.1165955804892202e-06,
-      "loss": 0.8703,
+      "learning_rate": 1.6194566373888244e-06,
+      "loss": 0.7875,
       "step": 28961
     },
     {
-      "epoch": 0.7954189667957485,
+      "epoch": 0.821850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 2.116048346504426e-06,
-      "loss": 0.8475,
+      "learning_rate": 1.6189552359452509e-06,
+      "loss": 0.8295,
       "step": 28962
     },
     {
-      "epoch": 0.7954464310235918,
+      "epoch": 0.8218785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 2.1155011749005105e-06,
-      "loss": 0.8897,
+      "learning_rate": 1.6184539052967375e-06,
+      "loss": 0.7843,
       "step": 28963
     },
     {
-      "epoch": 0.795473895251435,
+      "epoch": 0.8219069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 2.1149540656818037e-06,
-      "loss": 0.9277,
+      "learning_rate": 1.6179526454475202e-06,
+      "loss": 0.8379,
       "step": 28964
     },
     {
-      "epoch": 0.7955013594792782,
+      "epoch": 0.8219353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 2.1144070188526355e-06,
-      "loss": 0.8304,
+      "learning_rate": 1.6174514564018374e-06,
+      "loss": 0.8215,
       "step": 28965
     },
     {
-      "epoch": 0.7955288237071215,
+      "epoch": 0.8219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 2.1138600344173364e-06,
-      "loss": 0.8235,
+      "learning_rate": 1.616950338163915e-06,
+      "loss": 0.8071,
       "step": 28966
     },
     {
-      "epoch": 0.7955562879349647,
+      "epoch": 0.8219920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 2.1133131123802287e-06,
-      "loss": 0.7635,
+      "learning_rate": 1.6164492907379903e-06,
+      "loss": 0.9129,
       "step": 28967
     },
     {
-      "epoch": 0.795583752162808,
+      "epoch": 0.8220204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 2.112766252745645e-06,
-      "loss": 0.8497,
+      "learning_rate": 1.6159483141282984e-06,
+      "loss": 0.845,
       "step": 28968
     },
     {
-      "epoch": 0.7956112163906511,
+      "epoch": 0.8220488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 2.1122194555179067e-06,
-      "loss": 0.7133,
+      "learning_rate": 1.6154474083390648e-06,
+      "loss": 0.8012,
       "step": 28969
     },
     {
-      "epoch": 0.7956386806184944,
+      "epoch": 0.8220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 2.1116727207013443e-06,
-      "loss": 0.8309,
+      "learning_rate": 1.614946573374524e-06,
+      "loss": 0.8487,
       "step": 28970
     },
     {
-      "epoch": 0.7956661448463377,
+      "epoch": 0.822105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 2.1111260483002825e-06,
-      "loss": 0.834,
+      "learning_rate": 1.6144458092389092e-06,
+      "loss": 0.7899,
       "step": 28971
     },
     {
-      "epoch": 0.7956936090741809,
+      "epoch": 0.8221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 2.11057943831905e-06,
-      "loss": 0.7645,
+      "learning_rate": 1.6139451159364449e-06,
+      "loss": 0.7518,
       "step": 28972
     },
     {
-      "epoch": 0.7957210733020241,
+      "epoch": 0.8221623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 2.1100328907619673e-06,
-      "loss": 0.8476,
+      "learning_rate": 1.6134444934713611e-06,
+      "loss": 0.7118,
       "step": 28973
     },
     {
-      "epoch": 0.7957485375298673,
+      "epoch": 0.8221906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 2.1094864056333596e-06,
-      "loss": 0.8737,
+      "learning_rate": 1.6129439418478932e-06,
+      "loss": 0.8754,
       "step": 28974
     },
     {
-      "epoch": 0.7957760017577106,
+      "epoch": 0.8222190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 2.1089399829375556e-06,
-      "loss": 0.7407,
+      "learning_rate": 1.6124434610702622e-06,
+      "loss": 0.8995,
       "step": 28975
     },
     {
-      "epoch": 0.7958034659855538,
+      "epoch": 0.8222474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 2.1083936226788714e-06,
-      "loss": 0.794,
+      "learning_rate": 1.6119430511427015e-06,
+      "loss": 0.7523,
       "step": 28976
     },
     {
-      "epoch": 0.795830930213397,
+      "epoch": 0.8222758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 2.1078473248616338e-06,
-      "loss": 0.8535,
+      "learning_rate": 1.611442712069431e-06,
+      "loss": 0.7529,
       "step": 28977
     },
     {
-      "epoch": 0.7958583944412403,
+      "epoch": 0.8223041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 2.1073010894901692e-06,
-      "loss": 0.7689,
+      "learning_rate": 1.6109424438546827e-06,
+      "loss": 0.7986,
       "step": 28978
     },
     {
-      "epoch": 0.7958858586690836,
+      "epoch": 0.8223325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 2.1067549165687918e-06,
-      "loss": 0.8574,
+      "learning_rate": 1.6104422465026825e-06,
+      "loss": 0.8482,
       "step": 28979
     },
     {
-      "epoch": 0.7959133228969267,
+      "epoch": 0.822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 2.1062088061018284e-06,
-      "loss": 0.8699,
+      "learning_rate": 1.6099421200176513e-06,
+      "loss": 0.8081,
       "step": 28980
     },
     {
-      "epoch": 0.79594078712477,
+      "epoch": 0.8223893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 2.1056627580935974e-06,
-      "loss": 0.7962,
+      "learning_rate": 1.6094420644038157e-06,
+      "loss": 0.7638,
       "step": 28981
     },
     {
-      "epoch": 0.7959682513526132,
+      "epoch": 0.8224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 2.105116772548421e-06,
-      "loss": 0.853,
+      "learning_rate": 1.608942079665403e-06,
+      "loss": 0.7189,
       "step": 28982
     },
     {
-      "epoch": 0.7959957155804565,
+      "epoch": 0.8224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 2.104570849470622e-06,
-      "loss": 0.7766,
+      "learning_rate": 1.608442165806632e-06,
+      "loss": 0.8063,
       "step": 28983
     },
     {
-      "epoch": 0.7960231798082997,
+      "epoch": 0.8224744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 2.104024988864516e-06,
-      "loss": 0.8054,
+      "learning_rate": 1.6079423228317259e-06,
+      "loss": 0.7153,
       "step": 28984
     },
     {
-      "epoch": 0.7960506440361429,
+      "epoch": 0.8225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 2.10347919073442e-06,
-      "loss": 0.8931,
+      "learning_rate": 1.6074425507449122e-06,
+      "loss": 0.8552,
       "step": 28985
     },
     {
-      "epoch": 0.7960781082639862,
+      "epoch": 0.8225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 2.102933455084657e-06,
-      "loss": 0.7827,
+      "learning_rate": 1.6069428495504047e-06,
+      "loss": 0.9139,
       "step": 28986
     },
     {
-      "epoch": 0.7961055724918293,
+      "epoch": 0.8225595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 2.102387781919543e-06,
-      "loss": 0.8749,
+      "learning_rate": 1.6064432192524293e-06,
+      "loss": 0.7852,
       "step": 28987
     },
     {
-      "epoch": 0.7961330367196726,
+      "epoch": 0.8225879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 2.1018421712433946e-06,
-      "loss": 0.7297,
+      "learning_rate": 1.6059436598552068e-06,
+      "loss": 0.8166,
       "step": 28988
     },
     {
-      "epoch": 0.7961605009475159,
+      "epoch": 0.8226163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 2.1012966230605313e-06,
-      "loss": 0.8391,
+      "learning_rate": 1.605444171362952e-06,
+      "loss": 0.8536,
       "step": 28989
     },
     {
-      "epoch": 0.7961879651753591,
+      "epoch": 0.822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 2.100751137375272e-06,
-      "loss": 0.8572,
+      "learning_rate": 1.604944753779889e-06,
+      "loss": 0.838,
       "step": 28990
     },
     {
-      "epoch": 0.7962154294032023,
+      "epoch": 0.8226730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 2.1002057141919262e-06,
-      "loss": 0.8184,
+      "learning_rate": 1.6044454071102379e-06,
+      "loss": 0.8081,
       "step": 28991
     },
     {
-      "epoch": 0.7962428936310456,
+      "epoch": 0.8227014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 2.0996603535148165e-06,
-      "loss": 0.8033,
+      "learning_rate": 1.6039461313582106e-06,
+      "loss": 0.8476,
       "step": 28992
     },
     {
-      "epoch": 0.7962703578588888,
+      "epoch": 0.8227298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 2.099115055348251e-06,
-      "loss": 0.7166,
+      "learning_rate": 1.6034469265280295e-06,
+      "loss": 0.8508,
       "step": 28993
     },
     {
-      "epoch": 0.796297822086732,
+      "epoch": 0.8227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 2.098569819696548e-06,
-      "loss": 0.7738,
+      "learning_rate": 1.6029477926239102e-06,
+      "loss": 0.705,
       "step": 28994
     },
     {
-      "epoch": 0.7963252863145752,
+      "epoch": 0.8227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 2.098024646564021e-06,
-      "loss": 0.6955,
+      "learning_rate": 1.6024487296500668e-06,
+      "loss": 0.7782,
       "step": 28995
     },
     {
-      "epoch": 0.7963527505424185,
+      "epoch": 0.8228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 2.097479535954987e-06,
-      "loss": 0.8601,
+      "learning_rate": 1.6019497376107162e-06,
+      "loss": 0.8443,
       "step": 28996
     },
     {
-      "epoch": 0.7963802147702618,
+      "epoch": 0.8228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 2.096934487873753e-06,
-      "loss": 0.768,
+      "learning_rate": 1.6014508165100772e-06,
+      "loss": 0.8294,
       "step": 28997
     },
     {
-      "epoch": 0.7964076789981049,
+      "epoch": 0.822871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 2.096389502324635e-06,
-      "loss": 0.8884,
+      "learning_rate": 1.6009519663523577e-06,
+      "loss": 0.8308,
       "step": 28998
     },
     {
-      "epoch": 0.7964351432259482,
+      "epoch": 0.822900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 2.0958445793119474e-06,
-      "loss": 0.8463,
+      "learning_rate": 1.600453187141776e-06,
+      "loss": 0.849,
       "step": 28999
     },
     {
-      "epoch": 0.7964626074537914,
+      "epoch": 0.8229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 2.095299718839997e-06,
-      "loss": 0.8724,
+      "learning_rate": 1.5999544788825427e-06,
+      "loss": 0.82,
       "step": 29000
     },
     {
-      "epoch": 0.7964900716816347,
+      "epoch": 0.8229568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 2.094754920913099e-06,
-      "loss": 0.911,
+      "learning_rate": 1.599455841578871e-06,
+      "loss": 0.8669,
       "step": 29001
     },
     {
-      "epoch": 0.7965175359094779,
+      "epoch": 0.8229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 2.09421018553556e-06,
-      "loss": 0.7873,
+      "learning_rate": 1.5989572752349758e-06,
+      "loss": 0.8872,
       "step": 29002
     },
     {
-      "epoch": 0.7965450001373211,
+      "epoch": 0.8230136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 2.093665512711692e-06,
-      "loss": 0.7901,
+      "learning_rate": 1.5984587798550633e-06,
+      "loss": 0.7955,
       "step": 29003
     },
     {
-      "epoch": 0.7965724643651644,
+      "epoch": 0.8230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 2.0931209024458054e-06,
-      "loss": 0.7274,
+      "learning_rate": 1.5979603554433454e-06,
+      "loss": 0.9014,
       "step": 29004
     },
     {
-      "epoch": 0.7965999285930077,
+      "epoch": 0.8230703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 2.0925763547422097e-06,
-      "loss": 0.7933,
+      "learning_rate": 1.5974620020040388e-06,
+      "loss": 0.8502,
       "step": 29005
     },
     {
-      "epoch": 0.7966273928208508,
+      "epoch": 0.8230987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 2.092031869605212e-06,
-      "loss": 0.8905,
+      "learning_rate": 1.5969637195413457e-06,
+      "loss": 0.803,
       "step": 29006
     },
     {
-      "epoch": 0.7966548570486941,
+      "epoch": 0.8231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 2.0914874470391246e-06,
-      "loss": 0.8046,
+      "learning_rate": 1.596465508059477e-06,
+      "loss": 0.888,
       "step": 29007
     },
     {
-      "epoch": 0.7966823212765373,
+      "epoch": 0.823155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 2.0909430870482515e-06,
-      "loss": 0.7729,
+      "learning_rate": 1.5959673675626452e-06,
+      "loss": 0.968,
       "step": 29008
     },
     {
-      "epoch": 0.7967097855043805,
+      "epoch": 0.823183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 2.090398789636898e-06,
-      "loss": 0.8532,
+      "learning_rate": 1.5954692980550534e-06,
+      "loss": 0.8132,
       "step": 29009
     },
     {
-      "epoch": 0.7967372497322238,
+      "epoch": 0.8232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 2.089854554809374e-06,
-      "loss": 0.8116,
+      "learning_rate": 1.5949712995409095e-06,
+      "loss": 0.8026,
       "step": 29010
     },
     {
-      "epoch": 0.796764713960067,
+      "epoch": 0.8232406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 2.0893103825699835e-06,
-      "loss": 0.797,
+      "learning_rate": 1.5944733720244222e-06,
+      "loss": 0.7498,
       "step": 29011
     },
     {
-      "epoch": 0.7967921781879103,
+      "epoch": 0.8232690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 2.088766272923034e-06,
-      "loss": 0.7615,
+      "learning_rate": 1.5939755155097947e-06,
+      "loss": 0.8002,
       "step": 29012
     },
     {
-      "epoch": 0.7968196424157534,
+      "epoch": 0.8232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 2.0882222258728334e-06,
-      "loss": 0.8509,
+      "learning_rate": 1.5934777300012339e-06,
+      "loss": 0.8074,
       "step": 29013
     },
     {
-      "epoch": 0.7968471066435967,
+      "epoch": 0.8233257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 2.087678241423681e-06,
-      "loss": 0.8361,
+      "learning_rate": 1.5929800155029473e-06,
+      "loss": 0.7702,
       "step": 29014
     },
     {
-      "epoch": 0.79687457087144,
+      "epoch": 0.8233541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 2.087134319579882e-06,
-      "loss": 0.8548,
+      "learning_rate": 1.592482372019134e-06,
+      "loss": 0.7656,
       "step": 29015
     },
     {
-      "epoch": 0.7969020350992831,
+      "epoch": 0.8233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 2.086590460345744e-06,
-      "loss": 0.8285,
+      "learning_rate": 1.5919847995539993e-06,
+      "loss": 0.8131,
       "step": 29016
     },
     {
-      "epoch": 0.7969294993271264,
+      "epoch": 0.823410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.086046663725566e-06,
-      "loss": 0.8822,
+      "learning_rate": 1.5914872981117501e-06,
+      "loss": 0.8239,
       "step": 29017
     },
     {
-      "epoch": 0.7969569635549697,
+      "epoch": 0.8234392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 2.085502929723653e-06,
-      "loss": 0.7592,
+      "learning_rate": 1.5909898676965828e-06,
+      "loss": 0.8653,
       "step": 29018
     },
     {
-      "epoch": 0.7969844277828129,
+      "epoch": 0.8234676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 2.084959258344308e-06,
-      "loss": 0.7514,
+      "learning_rate": 1.5904925083127021e-06,
+      "loss": 0.9415,
       "step": 29019
     },
     {
-      "epoch": 0.7970118920106561,
+      "epoch": 0.8234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 2.084415649591829e-06,
-      "loss": 0.8812,
+      "learning_rate": 1.5899952199643099e-06,
+      "loss": 0.7639,
       "step": 29020
     },
     {
-      "epoch": 0.7970393562384993,
+      "epoch": 0.8235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 2.083872103470519e-06,
-      "loss": 0.8242,
+      "learning_rate": 1.5894980026556062e-06,
+      "loss": 0.846,
       "step": 29021
     },
     {
-      "epoch": 0.7970668204663426,
+      "epoch": 0.8235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 2.0833286199846802e-06,
-      "loss": 0.8008,
+      "learning_rate": 1.589000856390792e-06,
+      "loss": 0.6944,
       "step": 29022
     },
     {
-      "epoch": 0.7970942846941859,
+      "epoch": 0.8235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 2.082785199138614e-06,
-      "loss": 0.7173,
+      "learning_rate": 1.5885037811740644e-06,
+      "loss": 0.8826,
       "step": 29023
     },
     {
-      "epoch": 0.797121748922029,
+      "epoch": 0.8236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 2.0822418409366163e-06,
-      "loss": 0.8625,
+      "learning_rate": 1.588006777009623e-06,
+      "loss": 0.7427,
       "step": 29024
     },
     {
-      "epoch": 0.7971492131498723,
+      "epoch": 0.8236379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 2.0816985453829898e-06,
-      "loss": 0.8896,
+      "learning_rate": 1.587509843901669e-06,
+      "loss": 0.8641,
       "step": 29025
     },
     {
-      "epoch": 0.7971766773777155,
+      "epoch": 0.8236662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 2.0811553124820293e-06,
-      "loss": 0.8325,
+      "learning_rate": 1.5870129818543956e-06,
+      "loss": 0.798,
       "step": 29026
     },
     {
-      "epoch": 0.7972041416055587,
+      "epoch": 0.823694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 2.0806121422380364e-06,
-      "loss": 0.8846,
+      "learning_rate": 1.5865161908720006e-06,
+      "loss": 0.9046,
       "step": 29027
     },
     {
-      "epoch": 0.797231605833402,
+      "epoch": 0.8237230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 2.0800690346553065e-06,
-      "loss": 0.7365,
+      "learning_rate": 1.586019470958685e-06,
+      "loss": 0.729,
       "step": 29028
     },
     {
-      "epoch": 0.7972590700612452,
+      "epoch": 0.8237514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 2.079525989738139e-06,
-      "loss": 0.8281,
+      "learning_rate": 1.5855228221186391e-06,
+      "loss": 0.7949,
       "step": 29029
     },
     {
-      "epoch": 0.7972865342890885,
+      "epoch": 0.8237797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 2.07898300749083e-06,
-      "loss": 0.7633,
+      "learning_rate": 1.5850262443560594e-06,
+      "loss": 0.7293,
       "step": 29030
     },
     {
-      "epoch": 0.7973139985169317,
+      "epoch": 0.8238081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 2.078440087917679e-06,
-      "loss": 0.846,
+      "learning_rate": 1.5845297376751433e-06,
+      "loss": 0.9186,
       "step": 29031
     },
     {
-      "epoch": 0.7973414627447749,
+      "epoch": 0.8238365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 2.0778972310229773e-06,
-      "loss": 0.8029,
+      "learning_rate": 1.5840333020800813e-06,
+      "loss": 0.8,
       "step": 29032
     },
     {
-      "epoch": 0.7973689269726182,
+      "epoch": 0.8238649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 2.0773544368110198e-06,
-      "loss": 0.7725,
+      "learning_rate": 1.5835369375750676e-06,
+      "loss": 0.8284,
       "step": 29033
     },
     {
-      "epoch": 0.7973963912004614,
+      "epoch": 0.8238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 2.0768117052861025e-06,
-      "loss": 0.7521,
+      "learning_rate": 1.5830406441642987e-06,
+      "loss": 0.8075,
       "step": 29034
     },
     {
-      "epoch": 0.7974238554283046,
+      "epoch": 0.8239216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 2.0762690364525196e-06,
-      "loss": 0.7782,
+      "learning_rate": 1.5825444218519593e-06,
+      "loss": 0.7989,
       "step": 29035
     },
     {
-      "epoch": 0.7974513196561479,
+      "epoch": 0.823950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 2.0757264303145663e-06,
-      "loss": 0.8438,
+      "learning_rate": 1.582048270642249e-06,
+      "loss": 0.7388,
       "step": 29036
     },
     {
-      "epoch": 0.7974787838839911,
+      "epoch": 0.8239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 2.0751838868765375e-06,
-      "loss": 0.8031,
+      "learning_rate": 1.5815521905393538e-06,
+      "loss": 0.7978,
       "step": 29037
     },
     {
-      "epoch": 0.7975062481118343,
+      "epoch": 0.8240068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 2.0746414061427213e-06,
-      "loss": 0.7896,
+      "learning_rate": 1.5810561815474657e-06,
+      "loss": 0.8815,
       "step": 29038
     },
     {
-      "epoch": 0.7975337123396776,
+      "epoch": 0.8240351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 2.0740989881174114e-06,
-      "loss": 0.7936,
+      "learning_rate": 1.580560243670778e-06,
+      "loss": 0.8065,
       "step": 29039
     },
     {
-      "epoch": 0.7975611765675208,
+      "epoch": 0.8240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 2.073556632804904e-06,
-      "loss": 0.7947,
+      "learning_rate": 1.5800643769134728e-06,
+      "loss": 0.9437,
       "step": 29040
     },
     {
-      "epoch": 0.7975886407953641,
+      "epoch": 0.8240919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 2.0730143402094838e-06,
-      "loss": 0.8284,
+      "learning_rate": 1.5795685812797423e-06,
+      "loss": 0.8242,
       "step": 29041
     },
     {
-      "epoch": 0.7976161050232072,
+      "epoch": 0.8241203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 2.072472110335444e-06,
-      "loss": 0.8222,
+      "learning_rate": 1.5790728567737767e-06,
+      "loss": 0.8117,
       "step": 29042
     },
     {
-      "epoch": 0.7976435692510505,
+      "epoch": 0.8241486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 2.07192994318708e-06,
-      "loss": 0.8304,
+      "learning_rate": 1.5785772033997593e-06,
+      "loss": 0.7218,
       "step": 29043
     },
     {
-      "epoch": 0.7976710334788938,
+      "epoch": 0.8241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 2.071387838768674e-06,
-      "loss": 0.8204,
+      "learning_rate": 1.5780816211618787e-06,
+      "loss": 0.8718,
       "step": 29044
     },
     {
-      "epoch": 0.797698497706737,
+      "epoch": 0.824205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 2.0708457970845175e-06,
-      "loss": 0.8802,
+      "learning_rate": 1.577586110064323e-06,
+      "loss": 0.8319,
       "step": 29045
     },
     {
-      "epoch": 0.7977259619345802,
+      "epoch": 0.8242338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 2.070303818138901e-06,
-      "loss": 0.8394,
+      "learning_rate": 1.5770906701112753e-06,
+      "loss": 0.8006,
       "step": 29046
     },
     {
-      "epoch": 0.7977534261624234,
+      "epoch": 0.8242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 2.069761901936116e-06,
-      "loss": 0.8009,
+      "learning_rate": 1.5765953013069201e-06,
+      "loss": 0.8919,
       "step": 29047
     },
     {
-      "epoch": 0.7977808903902667,
+      "epoch": 0.8242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 2.0692200484804438e-06,
-      "loss": 0.7792,
+      "learning_rate": 1.5761000036554453e-06,
+      "loss": 0.8355,
       "step": 29048
     },
     {
-      "epoch": 0.79780835461811,
+      "epoch": 0.8243189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 2.068678257776178e-06,
-      "loss": 0.8943,
+      "learning_rate": 1.5756047771610306e-06,
+      "loss": 0.8095,
       "step": 29049
     },
     {
-      "epoch": 0.7978358188459531,
+      "epoch": 0.8243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 2.0681365298275978e-06,
-      "loss": 0.669,
+      "learning_rate": 1.5751096218278606e-06,
+      "loss": 0.8007,
       "step": 29050
     },
     {
-      "epoch": 0.7978632830737964,
+      "epoch": 0.8243757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 2.0675948646389944e-06,
-      "loss": 0.799,
+      "learning_rate": 1.5746145376601184e-06,
+      "loss": 0.8337,
       "step": 29051
     },
     {
-      "epoch": 0.7978907473016397,
+      "epoch": 0.8244040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 2.067053262214653e-06,
-      "loss": 0.8694,
+      "learning_rate": 1.5741195246619867e-06,
+      "loss": 0.8225,
       "step": 29052
     },
     {
-      "epoch": 0.7979182115294828,
+      "epoch": 0.8244324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 2.0665117225588605e-06,
-      "loss": 0.8031,
+      "learning_rate": 1.5736245828376483e-06,
+      "loss": 0.8087,
       "step": 29053
     },
     {
-      "epoch": 0.7979456757573261,
+      "epoch": 0.8244608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 2.065970245675899e-06,
-      "loss": 0.8429,
+      "learning_rate": 1.5731297121912791e-06,
+      "loss": 0.832,
       "step": 29054
     },
     {
-      "epoch": 0.7979731399851693,
+      "epoch": 0.8244892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 2.0654288315700577e-06,
-      "loss": 0.6923,
+      "learning_rate": 1.5726349127270624e-06,
+      "loss": 0.896,
       "step": 29055
     },
     {
-      "epoch": 0.7980006042130126,
+      "epoch": 0.8245175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 2.0648874802456165e-06,
-      "loss": 0.8353,
+      "learning_rate": 1.572140184449179e-06,
+      "loss": 0.7661,
       "step": 29056
     },
     {
-      "epoch": 0.7980280684408558,
+      "epoch": 0.8245459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 2.0643461917068564e-06,
-      "loss": 0.8194,
+      "learning_rate": 1.5716455273618048e-06,
+      "loss": 0.7031,
       "step": 29057
     },
     {
-      "epoch": 0.798055532668699,
+      "epoch": 0.8245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 2.063804965958064e-06,
-      "loss": 0.8393,
+      "learning_rate": 1.5711509414691196e-06,
+      "loss": 0.7681,
       "step": 29058
     },
     {
-      "epoch": 0.7980829968965423,
+      "epoch": 0.8246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 2.06326380300352e-06,
-      "loss": 0.832,
+      "learning_rate": 1.5706564267753032e-06,
+      "loss": 0.8166,
       "step": 29059
     },
     {
-      "epoch": 0.7981104611243854,
+      "epoch": 0.8246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 2.0627227028475104e-06,
-      "loss": 0.8838,
+      "learning_rate": 1.5701619832845284e-06,
+      "loss": 0.8474,
       "step": 29060
     },
     {
-      "epoch": 0.7981379253522287,
+      "epoch": 0.8246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 2.0621816654943117e-06,
-      "loss": 0.7604,
+      "learning_rate": 1.5696676110009746e-06,
+      "loss": 0.8615,
       "step": 29061
     },
     {
-      "epoch": 0.798165389580072,
+      "epoch": 0.8246878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 2.061640690948206e-06,
-      "loss": 0.8983,
+      "learning_rate": 1.5691733099288197e-06,
+      "loss": 0.8054,
       "step": 29062
     },
     {
-      "epoch": 0.7981928538079152,
+      "epoch": 0.824716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 2.0610997792134724e-06,
-      "loss": 0.7887,
+      "learning_rate": 1.5686790800722352e-06,
+      "loss": 0.8217,
       "step": 29063
     },
     {
-      "epoch": 0.7982203180357584,
+      "epoch": 0.824744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 2.0605589302943974e-06,
-      "loss": 0.8687,
+      "learning_rate": 1.5681849214353973e-06,
+      "loss": 0.8547,
       "step": 29064
     },
     {
-      "epoch": 0.7982477822636017,
+      "epoch": 0.8247729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 2.0600181441952516e-06,
-      "loss": 0.7865,
+      "learning_rate": 1.5676908340224794e-06,
+      "loss": 0.9324,
       "step": 29065
     },
     {
-      "epoch": 0.7982752464914449,
+      "epoch": 0.8248013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 2.059477420920322e-06,
-      "loss": 0.7633,
+      "learning_rate": 1.5671968178376574e-06,
+      "loss": 0.7606,
       "step": 29066
     },
     {
-      "epoch": 0.7983027107192882,
+      "epoch": 0.8248297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 2.0589367604738788e-06,
-      "loss": 0.8365,
+      "learning_rate": 1.5667028728851041e-06,
+      "loss": 0.7457,
       "step": 29067
     },
     {
-      "epoch": 0.7983301749471313,
+      "epoch": 0.8248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 2.0583961628602046e-06,
-      "loss": 0.9144,
+      "learning_rate": 1.5662089991689889e-06,
+      "loss": 0.8018,
       "step": 29068
     },
     {
-      "epoch": 0.7983576391749746,
+      "epoch": 0.8248864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 2.057855628083576e-06,
-      "loss": 0.7894,
+      "learning_rate": 1.5657151966934857e-06,
+      "loss": 0.8416,
       "step": 29069
     },
     {
-      "epoch": 0.7983851034028179,
+      "epoch": 0.8249148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 2.0573151561482694e-06,
-      "loss": 0.9166,
+      "learning_rate": 1.5652214654627672e-06,
+      "loss": 0.7917,
       "step": 29070
     },
     {
-      "epoch": 0.798412567630661,
+      "epoch": 0.8249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 2.0567747470585663e-06,
-      "loss": 0.8625,
+      "learning_rate": 1.5647278054810012e-06,
+      "loss": 0.8081,
       "step": 29071
     },
     {
-      "epoch": 0.7984400318585043,
+      "epoch": 0.8249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 2.0562344008187376e-06,
-      "loss": 0.8517,
+      "learning_rate": 1.5642342167523571e-06,
+      "loss": 0.899,
       "step": 29072
     },
     {
-      "epoch": 0.7984674960863475,
+      "epoch": 0.825,
       "grad_norm": 0.0,
-      "learning_rate": 2.055694117433057e-06,
-      "loss": 0.7172,
+      "learning_rate": 1.563740699281009e-06,
+      "loss": 0.7506,
       "step": 29073
     },
     {
-      "epoch": 0.7984949603141908,
+      "epoch": 0.8250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 2.0551538969058017e-06,
-      "loss": 0.765,
+      "learning_rate": 1.5632472530711195e-06,
+      "loss": 0.797,
       "step": 29074
     },
     {
-      "epoch": 0.798522424542034,
+      "epoch": 0.8250567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 2.0546137392412457e-06,
-      "loss": 0.8021,
+      "learning_rate": 1.5627538781268591e-06,
+      "loss": 0.7981,
       "step": 29075
     },
     {
-      "epoch": 0.7985498887698772,
+      "epoch": 0.8250851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 2.0540736444436636e-06,
-      "loss": 0.8571,
+      "learning_rate": 1.5622605744524e-06,
+      "loss": 0.7732,
       "step": 29076
     },
     {
-      "epoch": 0.7985773529977205,
+      "epoch": 0.8251135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 2.05353361251733e-06,
-      "loss": 0.7231,
+      "learning_rate": 1.5617673420519019e-06,
+      "loss": 0.8301,
       "step": 29077
     },
     {
-      "epoch": 0.7986048172255638,
+      "epoch": 0.8251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 2.052993643466518e-06,
-      "loss": 0.7805,
+      "learning_rate": 1.5612741809295339e-06,
+      "loss": 0.8193,
       "step": 29078
     },
     {
-      "epoch": 0.7986322814534069,
+      "epoch": 0.8251702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 2.052453737295497e-06,
-      "loss": 0.7625,
+      "learning_rate": 1.560781091089465e-06,
+      "loss": 0.8282,
       "step": 29079
     },
     {
-      "epoch": 0.7986597456812502,
+      "epoch": 0.8251986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 2.0519138940085427e-06,
-      "loss": 0.8196,
+      "learning_rate": 1.560288072535855e-06,
+      "loss": 0.7621,
       "step": 29080
     },
     {
-      "epoch": 0.7986872099090934,
+      "epoch": 0.8252270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 2.0513741136099217e-06,
-      "loss": 0.7846,
+      "learning_rate": 1.5597951252728694e-06,
+      "loss": 0.8084,
       "step": 29081
     },
     {
-      "epoch": 0.7987146741369366,
+      "epoch": 0.825255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 2.050834396103908e-06,
-      "loss": 0.7626,
+      "learning_rate": 1.5593022493046782e-06,
+      "loss": 0.6946,
       "step": 29082
     },
     {
-      "epoch": 0.7987421383647799,
+      "epoch": 0.8252837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 2.0502947414947714e-06,
-      "loss": 0.7649,
+      "learning_rate": 1.5588094446354373e-06,
+      "loss": 0.7908,
       "step": 29083
     },
     {
-      "epoch": 0.7987696025926231,
+      "epoch": 0.8253121452894439,
       "grad_norm": 0.0,
-      "learning_rate": 2.0497551497867853e-06,
-      "loss": 0.8565,
+      "learning_rate": 1.5583167112693153e-06,
+      "loss": 0.8534,
       "step": 29084
     },
     {
-      "epoch": 0.7987970668204664,
+      "epoch": 0.8253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 2.049215620984214e-06,
-      "loss": 0.764,
+      "learning_rate": 1.5578240492104701e-06,
+      "loss": 0.7928,
       "step": 29085
     },
     {
-      "epoch": 0.7988245310483095,
+      "epoch": 0.8253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 2.048676155091327e-06,
-      "loss": 0.7909,
+      "learning_rate": 1.5573314584630639e-06,
+      "loss": 0.7349,
       "step": 29086
     },
     {
-      "epoch": 0.7988519952761528,
+      "epoch": 0.8253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 2.0481367521123986e-06,
-      "loss": 0.7953,
+      "learning_rate": 1.55683893903126e-06,
+      "loss": 0.8172,
       "step": 29087
     },
     {
-      "epoch": 0.7988794595039961,
+      "epoch": 0.8254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 2.0475974120516885e-06,
-      "loss": 0.982,
+      "learning_rate": 1.5563464909192162e-06,
+      "loss": 0.8467,
       "step": 29088
     },
     {
-      "epoch": 0.7989069237318392,
+      "epoch": 0.8254540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 2.047058134913469e-06,
-      "loss": 0.8812,
+      "learning_rate": 1.5558541141310923e-06,
+      "loss": 0.8457,
       "step": 29089
     },
     {
-      "epoch": 0.7989343879596825,
+      "epoch": 0.8254824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 2.0465189207020085e-06,
-      "loss": 0.7731,
+      "learning_rate": 1.555361808671051e-06,
+      "loss": 0.7258,
       "step": 29090
     },
     {
-      "epoch": 0.7989618521875258,
+      "epoch": 0.8255107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 2.045979769421569e-06,
-      "loss": 0.7376,
+      "learning_rate": 1.554869574543245e-06,
+      "loss": 0.8103,
       "step": 29091
     },
     {
-      "epoch": 0.798989316415369,
+      "epoch": 0.825539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 2.045440681076418e-06,
-      "loss": 0.7241,
+      "learning_rate": 1.5543774117518362e-06,
+      "loss": 0.7628,
       "step": 29092
     },
     {
-      "epoch": 0.7990167806432122,
+      "epoch": 0.8255675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 2.044901655670821e-06,
-      "loss": 0.8577,
+      "learning_rate": 1.553885320300983e-06,
+      "loss": 0.7133,
       "step": 29093
     },
     {
-      "epoch": 0.7990442448710554,
+      "epoch": 0.8255959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 2.0443626932090456e-06,
-      "loss": 0.8434,
+      "learning_rate": 1.5533933001948376e-06,
+      "loss": 0.8996,
       "step": 29094
     },
     {
-      "epoch": 0.7990717090988987,
+      "epoch": 0.8256242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 2.0438237936953552e-06,
-      "loss": 0.8834,
+      "learning_rate": 1.5529013514375591e-06,
+      "loss": 0.734,
       "step": 29095
     },
     {
-      "epoch": 0.799099173326742,
+      "epoch": 0.8256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 2.043284957134014e-06,
-      "loss": 0.8245,
+      "learning_rate": 1.5524094740333028e-06,
+      "loss": 0.7472,
       "step": 29096
     },
     {
-      "epoch": 0.7991266375545851,
+      "epoch": 0.8256810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 2.0427461835292807e-06,
-      "loss": 0.7247,
+      "learning_rate": 1.5519176679862224e-06,
+      "loss": 0.9652,
       "step": 29097
     },
     {
-      "epoch": 0.7991541017824284,
+      "epoch": 0.8257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 2.0422074728854223e-06,
-      "loss": 0.8016,
+      "learning_rate": 1.5514259333004744e-06,
+      "loss": 0.8575,
       "step": 29098
     },
     {
-      "epoch": 0.7991815660102716,
+      "epoch": 0.8257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 2.041668825206702e-06,
-      "loss": 0.7455,
+      "learning_rate": 1.5509342699802132e-06,
+      "loss": 0.8064,
       "step": 29099
     },
     {
-      "epoch": 0.7992090302381148,
+      "epoch": 0.825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 2.041130240497379e-06,
-      "loss": 0.8748,
+      "learning_rate": 1.5504426780295877e-06,
+      "loss": 0.8168,
       "step": 29100
     },
     {
-      "epoch": 0.7992364944659581,
+      "epoch": 0.825794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 2.0405917187617175e-06,
-      "loss": 0.8147,
+      "learning_rate": 1.5499511574527526e-06,
+      "loss": 0.8555,
       "step": 29101
     },
     {
-      "epoch": 0.7992639586938013,
+      "epoch": 0.8258229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 2.0400532600039802e-06,
-      "loss": 0.7497,
+      "learning_rate": 1.549459708253863e-06,
+      "loss": 0.7931,
       "step": 29102
     },
     {
-      "epoch": 0.7992914229216446,
+      "epoch": 0.8258513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 2.0395148642284213e-06,
-      "loss": 0.8016,
+      "learning_rate": 1.548968330437064e-06,
+      "loss": 0.8012,
       "step": 29103
     },
     {
-      "epoch": 0.7993188871494878,
+      "epoch": 0.8258796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 2.0389765314393085e-06,
-      "loss": 0.8875,
+      "learning_rate": 1.5484770240065094e-06,
+      "loss": 0.8778,
       "step": 29104
     },
     {
-      "epoch": 0.799346351377331,
+      "epoch": 0.8259080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 2.038438261640894e-06,
-      "loss": 0.7881,
+      "learning_rate": 1.5479857889663519e-06,
+      "loss": 0.8313,
       "step": 29105
     },
     {
-      "epoch": 0.7993738156051743,
+      "epoch": 0.8259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 2.0379000548374394e-06,
-      "loss": 0.7954,
+      "learning_rate": 1.5474946253207357e-06,
+      "loss": 0.7848,
       "step": 29106
     },
     {
-      "epoch": 0.7994012798330175,
+      "epoch": 0.8259648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 2.0373619110332043e-06,
-      "loss": 0.9612,
+      "learning_rate": 1.5470035330738153e-06,
+      "loss": 0.8329,
       "step": 29107
     },
     {
-      "epoch": 0.7994287440608607,
+      "epoch": 0.8259931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 2.0368238302324483e-06,
-      "loss": 0.8338,
+      "learning_rate": 1.5465125122297342e-06,
+      "loss": 0.6738,
       "step": 29108
     },
     {
-      "epoch": 0.799456208288704,
+      "epoch": 0.8260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 2.0362858124394247e-06,
-      "loss": 0.7575,
+      "learning_rate": 1.5460215627926411e-06,
+      "loss": 0.832,
       "step": 29109
     },
     {
-      "epoch": 0.7994836725165472,
+      "epoch": 0.826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 2.035747857658392e-06,
-      "loss": 0.8148,
+      "learning_rate": 1.5455306847666861e-06,
+      "loss": 0.818,
       "step": 29110
     },
     {
-      "epoch": 0.7995111367443904,
+      "epoch": 0.826078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 2.035209965893611e-06,
-      "loss": 0.8092,
+      "learning_rate": 1.5450398781560083e-06,
+      "loss": 0.825,
       "step": 29111
     },
     {
-      "epoch": 0.7995386009722336,
+      "epoch": 0.8261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 2.0346721371493304e-06,
-      "loss": 0.8577,
+      "learning_rate": 1.5445491429647618e-06,
+      "loss": 0.8584,
       "step": 29112
     },
     {
-      "epoch": 0.7995660652000769,
+      "epoch": 0.8261350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 2.03413437142981e-06,
-      "loss": 0.7935,
+      "learning_rate": 1.5440584791970914e-06,
+      "loss": 0.8386,
       "step": 29113
     },
     {
-      "epoch": 0.7995935294279202,
+      "epoch": 0.8261634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 2.033596668739306e-06,
-      "loss": 0.8437,
+      "learning_rate": 1.5435678868571369e-06,
+      "loss": 0.8187,
       "step": 29114
     },
     {
-      "epoch": 0.7996209936557633,
+      "epoch": 0.8261918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 2.0330590290820694e-06,
-      "loss": 0.8312,
+      "learning_rate": 1.543077365949044e-06,
+      "loss": 0.7763,
       "step": 29115
     },
     {
-      "epoch": 0.7996484578836066,
+      "epoch": 0.8262202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 2.032521452462355e-06,
-      "loss": 0.8666,
+      "learning_rate": 1.5425869164769592e-06,
+      "loss": 0.7815,
       "step": 29116
     },
     {
-      "epoch": 0.7996759221114499,
+      "epoch": 0.8262485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 2.0319839388844178e-06,
-      "loss": 0.8997,
+      "learning_rate": 1.5420965384450205e-06,
+      "loss": 0.7619,
       "step": 29117
     },
     {
-      "epoch": 0.799703386339293,
+      "epoch": 0.8262769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 2.03144648835251e-06,
-      "loss": 0.8751,
+      "learning_rate": 1.541606231857372e-06,
+      "loss": 0.9126,
       "step": 29118
     },
     {
-      "epoch": 0.7997308505671363,
+      "epoch": 0.826305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 2.030909100870887e-06,
-      "loss": 0.7332,
+      "learning_rate": 1.54111599671816e-06,
+      "loss": 0.7117,
       "step": 29119
     },
     {
-      "epoch": 0.7997583147949795,
+      "epoch": 0.8263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 2.030371776443798e-06,
-      "loss": 0.9062,
+      "learning_rate": 1.5406258330315172e-06,
+      "loss": 0.7691,
       "step": 29120
     },
     {
-      "epoch": 0.7997857790228228,
+      "epoch": 0.8263620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 2.029834515075492e-06,
-      "loss": 0.7648,
+      "learning_rate": 1.5401357408015893e-06,
+      "loss": 0.8795,
       "step": 29121
     },
     {
-      "epoch": 0.799813243250666,
+      "epoch": 0.8263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 2.0292973167702225e-06,
-      "loss": 0.869,
+      "learning_rate": 1.5396457200325177e-06,
+      "loss": 0.8372,
       "step": 29122
     },
     {
-      "epoch": 0.7998407074785092,
+      "epoch": 0.8264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 2.02876018153224e-06,
-      "loss": 0.904,
+      "learning_rate": 1.5391557707284355e-06,
+      "loss": 0.8819,
       "step": 29123
     },
     {
-      "epoch": 0.7998681717063525,
+      "epoch": 0.8264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 2.0282231093657943e-06,
-      "loss": 0.8571,
+      "learning_rate": 1.538665892893486e-06,
+      "loss": 0.895,
       "step": 29124
     },
     {
-      "epoch": 0.7998956359341957,
+      "epoch": 0.8264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 2.0276861002751392e-06,
-      "loss": 0.8858,
+      "learning_rate": 1.5381760865318073e-06,
+      "loss": 0.8488,
       "step": 29125
     },
     {
-      "epoch": 0.7999231001620389,
+      "epoch": 0.8265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 2.027149154264517e-06,
-      "loss": 0.8548,
+      "learning_rate": 1.5376863516475339e-06,
+      "loss": 0.7115,
       "step": 29126
     },
     {
-      "epoch": 0.7999505643898822,
+      "epoch": 0.8265323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 2.0266122713381787e-06,
-      "loss": 0.8122,
+      "learning_rate": 1.537196688244804e-06,
+      "loss": 0.7165,
       "step": 29127
     },
     {
-      "epoch": 0.7999780286177254,
+      "epoch": 0.8265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 2.0260754515003754e-06,
-      "loss": 0.7999,
+      "learning_rate": 1.5367070963277553e-06,
+      "loss": 0.7131,
       "step": 29128
     },
     {
-      "epoch": 0.8000054928455687,
+      "epoch": 0.826589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 2.025538694755349e-06,
-      "loss": 0.8756,
+      "learning_rate": 1.5362175759005204e-06,
+      "loss": 0.8243,
       "step": 29129
     },
     {
-      "epoch": 0.8000329570734119,
+      "epoch": 0.8266174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 2.02500200110735e-06,
-      "loss": 0.8233,
+      "learning_rate": 1.5357281269672398e-06,
+      "loss": 0.801,
       "step": 29130
     },
     {
-      "epoch": 0.8000604213012551,
+      "epoch": 0.8266458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 2.0244653705606263e-06,
-      "loss": 0.855,
+      "learning_rate": 1.535238749532042e-06,
+      "loss": 0.836,
       "step": 29131
     },
     {
-      "epoch": 0.8000878855290984,
+      "epoch": 0.8266742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 2.0239288031194204e-06,
-      "loss": 0.7574,
+      "learning_rate": 1.5347494435990617e-06,
+      "loss": 0.7942,
       "step": 29132
     },
     {
-      "epoch": 0.8001153497569415,
+      "epoch": 0.8267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 2.0233922987879797e-06,
-      "loss": 0.7955,
+      "learning_rate": 1.5342602091724367e-06,
+      "loss": 0.8047,
       "step": 29133
     },
     {
-      "epoch": 0.8001428139847848,
+      "epoch": 0.8267309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 2.022855857570547e-06,
-      "loss": 0.7951,
+      "learning_rate": 1.5337710462562938e-06,
+      "loss": 0.7996,
       "step": 29134
     },
     {
-      "epoch": 0.8001702782126281,
+      "epoch": 0.8267593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 2.022319479471373e-06,
-      "loss": 0.8576,
+      "learning_rate": 1.5332819548547672e-06,
+      "loss": 0.7778,
       "step": 29135
     },
     {
-      "epoch": 0.8001977424404713,
+      "epoch": 0.8267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 2.021783164494695e-06,
-      "loss": 0.7609,
+      "learning_rate": 1.5327929349719906e-06,
+      "loss": 0.7712,
       "step": 29136
     },
     {
-      "epoch": 0.8002252066683145,
+      "epoch": 0.8268161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 2.0212469126447608e-06,
-      "loss": 0.8114,
+      "learning_rate": 1.5323039866120914e-06,
+      "loss": 0.7829,
       "step": 29137
     },
     {
-      "epoch": 0.8002526708961577,
+      "epoch": 0.8268444948921679,
       "grad_norm": 0.0,
-      "learning_rate": 2.0207107239258096e-06,
-      "loss": 0.806,
+      "learning_rate": 1.5318151097792e-06,
+      "loss": 0.8108,
       "step": 29138
     },
     {
-      "epoch": 0.800280135124001,
+      "epoch": 0.8268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 2.0201745983420862e-06,
-      "loss": 0.9049,
+      "learning_rate": 1.5313263044774496e-06,
+      "loss": 0.8273,
       "step": 29139
     },
     {
-      "epoch": 0.8003075993518443,
+      "epoch": 0.8269012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 2.0196385358978323e-06,
-      "loss": 0.882,
+      "learning_rate": 1.5308375707109647e-06,
+      "loss": 0.8038,
       "step": 29140
     },
     {
-      "epoch": 0.8003350635796874,
+      "epoch": 0.8269296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 2.0191025365972893e-06,
-      "loss": 0.747,
+      "learning_rate": 1.5303489084838751e-06,
+      "loss": 0.8676,
       "step": 29141
     },
     {
-      "epoch": 0.8003625278075307,
+      "epoch": 0.8269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 2.018566600444699e-06,
-      "loss": 0.8249,
+      "learning_rate": 1.529860317800309e-06,
+      "loss": 0.8727,
       "step": 29142
     },
     {
-      "epoch": 0.800389992035374,
+      "epoch": 0.8269863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 2.0180307274443033e-06,
-      "loss": 0.781,
+      "learning_rate": 1.5293717986643941e-06,
+      "loss": 0.8002,
       "step": 29143
     },
     {
-      "epoch": 0.8004174562632171,
+      "epoch": 0.8270147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 2.017494917600341e-06,
-      "loss": 0.7828,
+      "learning_rate": 1.5288833510802593e-06,
+      "loss": 0.7576,
       "step": 29144
     },
     {
-      "epoch": 0.8004449204910604,
+      "epoch": 0.8270431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 2.016959170917049e-06,
-      "loss": 0.757,
+      "learning_rate": 1.5283949750520245e-06,
+      "loss": 0.8036,
       "step": 29145
     },
     {
-      "epoch": 0.8004723847189036,
+      "epoch": 0.8270715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 2.0164234873986675e-06,
-      "loss": 0.8315,
+      "learning_rate": 1.5279066705838175e-06,
+      "loss": 0.7793,
       "step": 29146
     },
     {
-      "epoch": 0.8004998489467469,
+      "epoch": 0.827099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 2.0158878670494364e-06,
-      "loss": 0.8234,
+      "learning_rate": 1.5274184376797662e-06,
+      "loss": 0.7839,
       "step": 29147
     },
     {
-      "epoch": 0.8005273131745901,
+      "epoch": 0.8271282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 2.015352309873593e-06,
-      "loss": 0.6372,
+      "learning_rate": 1.5269302763439907e-06,
+      "loss": 0.8159,
       "step": 29148
     },
     {
-      "epoch": 0.8005547774024333,
+      "epoch": 0.8271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 2.0148168158753767e-06,
-      "loss": 0.8134,
+      "learning_rate": 1.5264421865806155e-06,
+      "loss": 0.833,
       "step": 29149
     },
     {
-      "epoch": 0.8005822416302766,
+      "epoch": 0.8271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 2.0142813850590213e-06,
-      "loss": 0.8237,
+      "learning_rate": 1.5259541683937673e-06,
+      "loss": 0.8456,
       "step": 29150
     },
     {
-      "epoch": 0.8006097058581197,
+      "epoch": 0.8272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 2.013746017428765e-06,
-      "loss": 0.8004,
+      "learning_rate": 1.5254662217875615e-06,
+      "loss": 0.839,
       "step": 29151
     },
     {
-      "epoch": 0.800637170085963,
+      "epoch": 0.8272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 2.0132107129888465e-06,
-      "loss": 0.8624,
+      "learning_rate": 1.524978346766124e-06,
+      "loss": 0.8639,
       "step": 29152
     },
     {
-      "epoch": 0.8006646343138063,
+      "epoch": 0.8272701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 2.012675471743496e-06,
-      "loss": 0.7722,
+      "learning_rate": 1.5244905433335777e-06,
+      "loss": 0.6985,
       "step": 29153
     },
     {
-      "epoch": 0.8006920985416495,
+      "epoch": 0.8272985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 2.0121402936969515e-06,
-      "loss": 0.7734,
+      "learning_rate": 1.5240028114940386e-06,
+      "loss": 0.7558,
       "step": 29154
     },
     {
-      "epoch": 0.8007195627694927,
+      "epoch": 0.8273269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 2.0116051788534487e-06,
-      "loss": 0.869,
+      "learning_rate": 1.5235151512516288e-06,
+      "loss": 0.8657,
       "step": 29155
     },
     {
-      "epoch": 0.800747026997336,
+      "epoch": 0.8273552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 2.0110701272172185e-06,
-      "loss": 0.8027,
+      "learning_rate": 1.5230275626104707e-06,
+      "loss": 0.933,
       "step": 29156
     },
     {
-      "epoch": 0.8007744912251792,
+      "epoch": 0.8273836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 2.0105351387924953e-06,
-      "loss": 0.7727,
+      "learning_rate": 1.522540045574674e-06,
+      "loss": 0.756,
       "step": 29157
     },
     {
-      "epoch": 0.8008019554530225,
+      "epoch": 0.8274120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 2.010000213583514e-06,
-      "loss": 0.7411,
+      "learning_rate": 1.5220526001483671e-06,
+      "loss": 0.8681,
       "step": 29158
     },
     {
-      "epoch": 0.8008294196808656,
+      "epoch": 0.8274404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 2.009465351594507e-06,
-      "loss": 0.7193,
+      "learning_rate": 1.5215652263356618e-06,
+      "loss": 0.7711,
       "step": 29159
     },
     {
-      "epoch": 0.8008568839087089,
+      "epoch": 0.8274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 2.008930552829704e-06,
-      "loss": 0.8324,
+      "learning_rate": 1.5210779241406747e-06,
+      "loss": 0.7861,
       "step": 29160
     },
     {
-      "epoch": 0.8008843481365522,
+      "epoch": 0.8274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 2.0083958172933403e-06,
-      "loss": 0.7889,
+      "learning_rate": 1.5205906935675274e-06,
+      "loss": 0.8875,
       "step": 29161
     },
     {
-      "epoch": 0.8009118123643953,
+      "epoch": 0.8275255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 2.0078611449896423e-06,
-      "loss": 0.8441,
+      "learning_rate": 1.5201035346203286e-06,
+      "loss": 0.7959,
       "step": 29162
     },
     {
-      "epoch": 0.8009392765922386,
+      "epoch": 0.8275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 2.007326535922842e-06,
-      "loss": 0.8264,
+      "learning_rate": 1.519616447303196e-06,
+      "loss": 0.8427,
       "step": 29163
     },
     {
-      "epoch": 0.8009667408200818,
+      "epoch": 0.8275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 2.006791990097171e-06,
-      "loss": 0.9952,
+      "learning_rate": 1.5191294316202476e-06,
+      "loss": 0.8673,
       "step": 29164
     },
     {
-      "epoch": 0.8009942050479251,
+      "epoch": 0.82761066969353,
       "grad_norm": 0.0,
-      "learning_rate": 2.0062575075168577e-06,
-      "loss": 0.8159,
+      "learning_rate": 1.518642487575591e-06,
+      "loss": 0.7657,
       "step": 29165
     },
     {
-      "epoch": 0.8010216692757683,
+      "epoch": 0.827639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 2.0057230881861313e-06,
-      "loss": 0.7822,
+      "learning_rate": 1.5181556151733433e-06,
+      "loss": 0.7619,
       "step": 29166
     },
     {
-      "epoch": 0.8010491335036115,
+      "epoch": 0.8276674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 2.0051887321092233e-06,
-      "loss": 0.7657,
+      "learning_rate": 1.5176688144176188e-06,
+      "loss": 0.7523,
       "step": 29167
     },
     {
-      "epoch": 0.8010765977314548,
+      "epoch": 0.8276958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 2.0046544392903587e-06,
-      "loss": 0.8445,
+      "learning_rate": 1.517182085312524e-06,
+      "loss": 0.7746,
       "step": 29168
     },
     {
-      "epoch": 0.8011040619592981,
+      "epoch": 0.8277241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 2.004120209733763e-06,
-      "loss": 0.8953,
+      "learning_rate": 1.5166954278621748e-06,
+      "loss": 0.8684,
       "step": 29169
     },
     {
-      "epoch": 0.8011315261871412,
+      "epoch": 0.8277525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 2.003586043443665e-06,
-      "loss": 0.8832,
+      "learning_rate": 1.5162088420706822e-06,
+      "loss": 0.8205,
       "step": 29170
     },
     {
-      "epoch": 0.8011589904149845,
+      "epoch": 0.8277809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 2.003051940424292e-06,
-      "loss": 0.8467,
+      "learning_rate": 1.5157223279421517e-06,
+      "loss": 0.7921,
       "step": 29171
     },
     {
-      "epoch": 0.8011864546428277,
+      "epoch": 0.8278093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 2.00251790067987e-06,
-      "loss": 0.7731,
+      "learning_rate": 1.515235885480697e-06,
+      "loss": 0.751,
       "step": 29172
     },
     {
-      "epoch": 0.801213918870671,
+      "epoch": 0.8278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 2.0019839242146265e-06,
-      "loss": 0.7189,
+      "learning_rate": 1.5147495146904268e-06,
+      "loss": 0.8762,
       "step": 29173
     },
     {
-      "epoch": 0.8012413830985142,
+      "epoch": 0.8278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 2.001450011032783e-06,
-      "loss": 0.719,
+      "learning_rate": 1.5142632155754478e-06,
+      "loss": 0.8101,
       "step": 29174
     },
     {
-      "epoch": 0.8012688473263574,
+      "epoch": 0.827894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 2.0009161611385643e-06,
-      "loss": 0.7864,
+      "learning_rate": 1.5137769881398722e-06,
+      "loss": 0.9117,
       "step": 29175
     },
     {
-      "epoch": 0.8012963115542007,
+      "epoch": 0.8279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 2.000382374536197e-06,
-      "loss": 0.8109,
+      "learning_rate": 1.5132908323878004e-06,
+      "loss": 0.9326,
       "step": 29176
     },
     {
-      "epoch": 0.8013237757820438,
+      "epoch": 0.8279511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.999848651229902e-06,
-      "loss": 0.8239,
+      "learning_rate": 1.5128047483233432e-06,
+      "loss": 0.975,
       "step": 29177
     },
     {
-      "epoch": 0.8013512400098871,
+      "epoch": 0.8279795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.999314991223903e-06,
-      "loss": 0.8335,
+      "learning_rate": 1.5123187359506075e-06,
+      "loss": 0.791,
       "step": 29178
     },
     {
-      "epoch": 0.8013787042377304,
+      "epoch": 0.8280079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.9987813945224265e-06,
-      "loss": 0.8062,
+      "learning_rate": 1.5118327952736955e-06,
+      "loss": 0.9395,
       "step": 29179
     },
     {
-      "epoch": 0.8014061684655736,
+      "epoch": 0.8280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.9982478611296876e-06,
-      "loss": 0.8593,
+      "learning_rate": 1.511346926296713e-06,
+      "loss": 0.8955,
       "step": 29180
     },
     {
-      "epoch": 0.8014336326934168,
+      "epoch": 0.8280646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 1.9977143910499107e-06,
-      "loss": 0.7245,
+      "learning_rate": 1.5108611290237675e-06,
+      "loss": 0.8838,
       "step": 29181
     },
     {
-      "epoch": 0.8014610969212601,
+      "epoch": 0.8280930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.997180984287318e-06,
-      "loss": 0.9376,
+      "learning_rate": 1.5103754034589568e-06,
+      "loss": 0.8473,
       "step": 29182
     },
     {
-      "epoch": 0.8014885611491033,
+      "epoch": 0.8281214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.996647640846131e-06,
-      "loss": 0.7347,
+      "learning_rate": 1.509889749606387e-06,
+      "loss": 0.7682,
       "step": 29183
     },
     {
-      "epoch": 0.8015160253769466,
+      "epoch": 0.828149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.9961143607305666e-06,
-      "loss": 0.7597,
+      "learning_rate": 1.509404167470162e-06,
+      "loss": 0.7976,
       "step": 29184
     },
     {
-      "epoch": 0.8015434896047897,
+      "epoch": 0.8281782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.995581143944848e-06,
-      "loss": 0.7721,
+      "learning_rate": 1.508918657054379e-06,
+      "loss": 0.8232,
       "step": 29185
     },
     {
-      "epoch": 0.801570953832633,
+      "epoch": 0.8282065834279229,
       "grad_norm": 0.0,
-      "learning_rate": 1.9950479904931898e-06,
-      "loss": 0.769,
+      "learning_rate": 1.5084332183631423e-06,
+      "loss": 0.8797,
       "step": 29186
     },
     {
-      "epoch": 0.8015984180604763,
+      "epoch": 0.8282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.9945149003798112e-06,
-      "loss": 0.8698,
+      "learning_rate": 1.507947851400554e-06,
+      "loss": 0.8142,
       "step": 29187
     },
     {
-      "epoch": 0.8016258822883194,
+      "epoch": 0.8282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.9939818736089333e-06,
-      "loss": 0.7906,
+      "learning_rate": 1.507462556170708e-06,
+      "loss": 0.7945,
       "step": 29188
     },
     {
-      "epoch": 0.8016533465161627,
+      "epoch": 0.8282917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 1.993448910184771e-06,
-      "loss": 0.7784,
+      "learning_rate": 1.5069773326777116e-06,
+      "loss": 0.8617,
       "step": 29189
     },
     {
-      "epoch": 0.8016808107440059,
+      "epoch": 0.8283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.9929160101115465e-06,
-      "loss": 0.8436,
+      "learning_rate": 1.5064921809256572e-06,
+      "loss": 0.8076,
       "step": 29190
     },
     {
-      "epoch": 0.8017082749718492,
+      "epoch": 0.8283484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.9923831733934684e-06,
-      "loss": 0.8262,
+      "learning_rate": 1.5060071009186462e-06,
+      "loss": 0.8448,
       "step": 29191
     },
     {
-      "epoch": 0.8017357391996924,
+      "epoch": 0.8283768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 1.9918504000347593e-06,
-      "loss": 0.8171,
+      "learning_rate": 1.505522092660776e-06,
+      "loss": 0.844,
       "step": 29192
     },
     {
-      "epoch": 0.8017632034275356,
+      "epoch": 0.8284052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 1.991317690039629e-06,
-      "loss": 0.8546,
+      "learning_rate": 1.5050371561561405e-06,
+      "loss": 0.6641,
       "step": 29193
     },
     {
-      "epoch": 0.8017906676553789,
+      "epoch": 0.828433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.9907850434122955e-06,
-      "loss": 0.8545,
+      "learning_rate": 1.5045522914088385e-06,
+      "loss": 0.7356,
       "step": 29194
     },
     {
-      "epoch": 0.8018181318832222,
+      "epoch": 0.8284619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.990252460156974e-06,
-      "loss": 0.8705,
+      "learning_rate": 1.5040674984229676e-06,
+      "loss": 0.7806,
       "step": 29195
     },
     {
-      "epoch": 0.8018455961110653,
+      "epoch": 0.8284903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 1.989719940277879e-06,
-      "loss": 0.8201,
+      "learning_rate": 1.5035827772026168e-06,
+      "loss": 0.7786,
       "step": 29196
     },
     {
-      "epoch": 0.8018730603389086,
+      "epoch": 0.8285187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.989187483779221e-06,
-      "loss": 0.803,
+      "learning_rate": 1.5030981277518852e-06,
+      "loss": 0.8182,
       "step": 29197
     },
     {
-      "epoch": 0.8019005245667518,
+      "epoch": 0.8285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.988655090665215e-06,
-      "loss": 0.7704,
+      "learning_rate": 1.5026135500748684e-06,
+      "loss": 0.8146,
       "step": 29198
     },
     {
-      "epoch": 0.801927988794595,
+      "epoch": 0.8285754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.9881227609400724e-06,
-      "loss": 0.8108,
+      "learning_rate": 1.5021290441756542e-06,
+      "loss": 0.8619,
       "step": 29199
     },
     {
-      "epoch": 0.8019554530224383,
+      "epoch": 0.8286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.9875904946080094e-06,
-      "loss": 0.7955,
+      "learning_rate": 1.5016446100583381e-06,
+      "loss": 0.8061,
       "step": 29200
     },
     {
-      "epoch": 0.8019829172502815,
+      "epoch": 0.8286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.9870582916732307e-06,
-      "loss": 0.7865,
+      "learning_rate": 1.5011602477270137e-06,
+      "loss": 0.7399,
       "step": 29201
     },
     {
-      "epoch": 0.8020103814781248,
+      "epoch": 0.8286606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.9865261521399535e-06,
-      "loss": 0.7982,
+      "learning_rate": 1.5006759571857687e-06,
+      "loss": 0.8914,
       "step": 29202
     },
     {
-      "epoch": 0.8020378457059679,
+      "epoch": 0.828688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.985994076012383e-06,
-      "loss": 0.8537,
+      "learning_rate": 1.5001917384386944e-06,
+      "loss": 0.7932,
       "step": 29203
     },
     {
-      "epoch": 0.8020653099338112,
+      "epoch": 0.8287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.985462063294732e-06,
-      "loss": 0.8056,
+      "learning_rate": 1.4997075914898863e-06,
+      "loss": 0.8943,
       "step": 29204
     },
     {
-      "epoch": 0.8020927741616545,
+      "epoch": 0.8287457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 1.9849301139912104e-06,
-      "loss": 0.7891,
+      "learning_rate": 1.4992235163434288e-06,
+      "loss": 0.6832,
       "step": 29205
     },
     {
-      "epoch": 0.8021202383894976,
+      "epoch": 0.8287741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.9843982281060247e-06,
-      "loss": 0.7784,
+      "learning_rate": 1.498739513003411e-06,
+      "loss": 0.7694,
       "step": 29206
     },
     {
-      "epoch": 0.8021477026173409,
+      "epoch": 0.8288024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.9838664056433886e-06,
-      "loss": 0.8975,
+      "learning_rate": 1.4982555814739264e-06,
+      "loss": 0.8041,
       "step": 29207
     },
     {
-      "epoch": 0.8021751668451842,
+      "epoch": 0.8288308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 1.9833346466075067e-06,
-      "loss": 0.817,
+      "learning_rate": 1.497771721759056e-06,
+      "loss": 0.7494,
       "step": 29208
     },
     {
-      "epoch": 0.8022026310730274,
+      "epoch": 0.8288592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.9828029510025836e-06,
-      "loss": 0.7871,
+      "learning_rate": 1.4972879338628909e-06,
+      "loss": 0.8268,
       "step": 29209
     },
     {
-      "epoch": 0.8022300953008706,
+      "epoch": 0.8288876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 1.9822713188328303e-06,
-      "loss": 0.7891,
+      "learning_rate": 1.496804217789518e-06,
+      "loss": 0.8832,
       "step": 29210
     },
     {
-      "epoch": 0.8022575595287138,
+      "epoch": 0.8289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.9817397501024503e-06,
-      "loss": 0.789,
+      "learning_rate": 1.496320573543021e-06,
+      "loss": 0.805,
       "step": 29211
     },
     {
-      "epoch": 0.8022850237565571,
+      "epoch": 0.82894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.981208244815652e-06,
-      "loss": 0.7372,
+      "learning_rate": 1.4958370011274859e-06,
+      "loss": 0.8732,
       "step": 29212
     },
     {
-      "epoch": 0.8023124879844004,
+      "epoch": 0.828972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.980676802976641e-06,
-      "loss": 0.8224,
+      "learning_rate": 1.495353500547e-06,
+      "loss": 0.7659,
       "step": 29213
     },
     {
-      "epoch": 0.8023399522122435,
+      "epoch": 0.8290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.9801454245896234e-06,
-      "loss": 0.8502,
+      "learning_rate": 1.494870071805643e-06,
+      "loss": 0.8674,
       "step": 29214
     },
     {
-      "epoch": 0.8023674164400868,
+      "epoch": 0.8290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.9796141096587997e-06,
-      "loss": 0.8123,
+      "learning_rate": 1.4943867149075032e-06,
+      "loss": 0.8346,
       "step": 29215
     },
     {
-      "epoch": 0.80239488066793,
+      "epoch": 0.8290578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 1.979082858188378e-06,
-      "loss": 0.8295,
+      "learning_rate": 1.4939034298566591e-06,
+      "loss": 0.7918,
       "step": 29216
     },
     {
-      "epoch": 0.8024223448957732,
+      "epoch": 0.8290862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.978551670182558e-06,
-      "loss": 0.7809,
+      "learning_rate": 1.4934202166571953e-06,
+      "loss": 0.8076,
       "step": 29217
     },
     {
-      "epoch": 0.8024498091236165,
+      "epoch": 0.8291146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.9780205456455435e-06,
-      "loss": 0.8761,
+      "learning_rate": 1.4929370753131956e-06,
+      "loss": 0.7627,
       "step": 29218
     },
     {
-      "epoch": 0.8024772733514597,
+      "epoch": 0.8291430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.9774894845815386e-06,
-      "loss": 0.8149,
+      "learning_rate": 1.492454005828734e-06,
+      "loss": 0.7982,
       "step": 29219
     },
     {
-      "epoch": 0.802504737579303,
+      "epoch": 0.8291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 1.976958486994747e-06,
-      "loss": 0.8239,
+      "learning_rate": 1.4919710082078976e-06,
+      "loss": 0.8287,
       "step": 29220
     },
     {
-      "epoch": 0.8025322018071462,
+      "epoch": 0.829199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.976427552889364e-06,
-      "loss": 0.7114,
+      "learning_rate": 1.4914880824547673e-06,
+      "loss": 0.8197,
       "step": 29221
     },
     {
-      "epoch": 0.8025596660349894,
+      "epoch": 0.8292281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.9758966822695945e-06,
-      "loss": 0.8343,
+      "learning_rate": 1.4910052285734177e-06,
+      "loss": 0.6845,
       "step": 29222
     },
     {
-      "epoch": 0.8025871302628327,
+      "epoch": 0.8292565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.97536587513964e-06,
-      "loss": 0.714,
+      "learning_rate": 1.490522446567929e-06,
+      "loss": 0.7986,
       "step": 29223
     },
     {
-      "epoch": 0.8026145944906758,
+      "epoch": 0.8292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.9748351315037006e-06,
-      "loss": 0.8166,
+      "learning_rate": 1.4900397364423825e-06,
+      "loss": 0.9353,
       "step": 29224
     },
     {
-      "epoch": 0.8026420587185191,
+      "epoch": 0.8293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.9743044513659725e-06,
-      "loss": 0.8106,
+      "learning_rate": 1.4895570982008511e-06,
+      "loss": 0.8008,
       "step": 29225
     },
     {
-      "epoch": 0.8026695229463624,
+      "epoch": 0.8293416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.973773834730658e-06,
-      "loss": 0.7766,
+      "learning_rate": 1.489074531847412e-06,
+      "loss": 0.7327,
       "step": 29226
     },
     {
-      "epoch": 0.8026969871742056,
+      "epoch": 0.8293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.973243281601952e-06,
-      "loss": 0.7925,
+      "learning_rate": 1.4885920373861463e-06,
+      "loss": 0.8394,
       "step": 29227
     },
     {
-      "epoch": 0.8027244514020488,
+      "epoch": 0.8293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.972712791984055e-06,
-      "loss": 0.7786,
+      "learning_rate": 1.4881096148211238e-06,
+      "loss": 0.7036,
       "step": 29228
     },
     {
-      "epoch": 0.8027519156298921,
+      "epoch": 0.8294267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.9721823658811623e-06,
-      "loss": 0.8524,
+      "learning_rate": 1.4876272641564215e-06,
+      "loss": 0.8692,
       "step": 29229
     },
     {
-      "epoch": 0.8027793798577353,
+      "epoch": 0.8294551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.971652003297474e-06,
-      "loss": 0.8154,
+      "learning_rate": 1.4871449853961172e-06,
+      "loss": 0.7644,
       "step": 29230
     },
     {
-      "epoch": 0.8028068440855786,
+      "epoch": 0.829483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.9711217042371866e-06,
-      "loss": 0.8157,
+      "learning_rate": 1.4866627785442788e-06,
+      "loss": 0.8317,
       "step": 29231
     },
     {
-      "epoch": 0.8028343083134217,
+      "epoch": 0.8295119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 1.970591468704495e-06,
-      "loss": 0.822,
+      "learning_rate": 1.486180643604983e-06,
+      "loss": 0.7725,
       "step": 29232
     },
     {
-      "epoch": 0.802861772541265,
+      "epoch": 0.8295402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.9700612967035902e-06,
-      "loss": 0.7282,
+      "learning_rate": 1.4856985805823055e-06,
+      "loss": 0.7886,
       "step": 29233
     },
     {
-      "epoch": 0.8028892367691083,
+      "epoch": 0.8295686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 1.9695311882386727e-06,
-      "loss": 0.7062,
+      "learning_rate": 1.4852165894803083e-06,
+      "loss": 0.9008,
       "step": 29234
     },
     {
-      "epoch": 0.8029167009969514,
+      "epoch": 0.8295970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 1.9690011433139334e-06,
-      "loss": 0.8669,
+      "learning_rate": 1.484734670303075e-06,
+      "loss": 0.7793,
       "step": 29235
     },
     {
-      "epoch": 0.8029441652247947,
+      "epoch": 0.8296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.9684711619335694e-06,
-      "loss": 0.8195,
+      "learning_rate": 1.4842528230546681e-06,
+      "loss": 0.8347,
       "step": 29236
     },
     {
-      "epoch": 0.8029716294526379,
+      "epoch": 0.8296538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679412441017722e-06,
-      "loss": 0.9267,
+      "learning_rate": 1.483771047739161e-06,
+      "loss": 0.7517,
       "step": 29237
     },
     {
-      "epoch": 0.8029990936804812,
+      "epoch": 0.8296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.9674113898227376e-06,
-      "loss": 0.7069,
+      "learning_rate": 1.4832893443606244e-06,
+      "loss": 0.736,
       "step": 29238
     },
     {
-      "epoch": 0.8030265579083244,
+      "epoch": 0.8297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.966881599100653e-06,
-      "loss": 0.7681,
+      "learning_rate": 1.4828077129231245e-06,
+      "loss": 0.6394,
       "step": 29239
     },
     {
-      "epoch": 0.8030540221361676,
+      "epoch": 0.8297389330306469,
       "grad_norm": 0.0,
-      "learning_rate": 1.9663518719397156e-06,
-      "loss": 0.7878,
+      "learning_rate": 1.4823261534307286e-06,
+      "loss": 0.8342,
       "step": 29240
     },
     {
-      "epoch": 0.8030814863640109,
+      "epoch": 0.8297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.9658222083441113e-06,
-      "loss": 0.8762,
+      "learning_rate": 1.4818446658875108e-06,
+      "loss": 0.9032,
       "step": 29241
     },
     {
-      "epoch": 0.8031089505918542,
+      "epoch": 0.8297956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.9652926083180337e-06,
-      "loss": 0.719,
+      "learning_rate": 1.4813632502975305e-06,
+      "loss": 0.9419,
       "step": 29242
     },
     {
-      "epoch": 0.8031364148196973,
+      "epoch": 0.8298240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 1.9647630718656763e-06,
-      "loss": 0.8735,
+      "learning_rate": 1.4808819066648583e-06,
+      "loss": 0.7744,
       "step": 29243
     },
     {
-      "epoch": 0.8031638790475406,
+      "epoch": 0.8298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 1.9642335989912232e-06,
-      "loss": 0.8127,
+      "learning_rate": 1.4804006349935618e-06,
+      "loss": 0.8801,
       "step": 29244
     },
     {
-      "epoch": 0.8031913432753838,
+      "epoch": 0.8298808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.963704189698867e-06,
-      "loss": 0.7324,
+      "learning_rate": 1.4799194352877023e-06,
+      "loss": 0.7926,
       "step": 29245
     },
     {
-      "epoch": 0.803218807503227,
+      "epoch": 0.8299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.9631748439927967e-06,
-      "loss": 0.8435,
+      "learning_rate": 1.4794383075513453e-06,
+      "loss": 0.7788,
       "step": 29246
     },
     {
-      "epoch": 0.8032462717310703,
+      "epoch": 0.8299375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 1.962645561877201e-06,
-      "loss": 0.8376,
+      "learning_rate": 1.4789572517885586e-06,
+      "loss": 0.8583,
       "step": 29247
     },
     {
-      "epoch": 0.8032737359589135,
+      "epoch": 0.8299659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 1.962116343356266e-06,
-      "loss": 0.8918,
+      "learning_rate": 1.4784762680034015e-06,
+      "loss": 0.8179,
       "step": 29248
     },
     {
-      "epoch": 0.8033012001867568,
+      "epoch": 0.829994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.961587188434182e-06,
-      "loss": 0.8134,
+      "learning_rate": 1.4779953561999372e-06,
+      "loss": 0.7907,
       "step": 29249
     },
     {
-      "epoch": 0.8033286644145999,
+      "epoch": 0.8300227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.9610580971151318e-06,
-      "loss": 0.7798,
+      "learning_rate": 1.4775145163822302e-06,
+      "loss": 0.806,
       "step": 29250
     },
     {
-      "epoch": 0.8033561286424432,
+      "epoch": 0.8300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.960529069403304e-06,
-      "loss": 0.7195,
+      "learning_rate": 1.4770337485543407e-06,
+      "loss": 0.7764,
       "step": 29251
     },
     {
-      "epoch": 0.8033835928702865,
+      "epoch": 0.8300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.960000105302884e-06,
-      "loss": 0.8073,
+      "learning_rate": 1.476553052720333e-06,
+      "loss": 0.8272,
       "step": 29252
     },
     {
-      "epoch": 0.8034110570981297,
+      "epoch": 0.8301078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.9594712048180574e-06,
-      "loss": 0.8567,
+      "learning_rate": 1.476072428884262e-06,
+      "loss": 0.8484,
       "step": 29253
     },
     {
-      "epoch": 0.8034385213259729,
+      "epoch": 0.8301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.9589423679530095e-06,
-      "loss": 0.7358,
+      "learning_rate": 1.475591877050191e-06,
+      "loss": 0.7674,
       "step": 29254
     },
     {
-      "epoch": 0.8034659855538162,
+      "epoch": 0.8301645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.9584135947119267e-06,
-      "loss": 0.7996,
+      "learning_rate": 1.4751113972221797e-06,
+      "loss": 0.8598,
       "step": 29255
     },
     {
-      "epoch": 0.8034934497816594,
+      "epoch": 0.8301929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 1.9578848850989907e-06,
-      "loss": 0.7836,
+      "learning_rate": 1.4746309894042843e-06,
+      "loss": 0.7953,
       "step": 29256
     },
     {
-      "epoch": 0.8035209140095027,
+      "epoch": 0.8302213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 1.957356239118382e-06,
-      "loss": 0.78,
+      "learning_rate": 1.4741506536005646e-06,
+      "loss": 0.8178,
       "step": 29257
     },
     {
-      "epoch": 0.8035483782373458,
+      "epoch": 0.8302497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 1.9568276567742872e-06,
-      "loss": 0.8284,
+      "learning_rate": 1.4736703898150795e-06,
+      "loss": 0.7628,
       "step": 29258
     },
     {
-      "epoch": 0.8035758424651891,
+      "epoch": 0.8302780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 1.9562991380708863e-06,
-      "loss": 0.7458,
+      "learning_rate": 1.4731901980518816e-06,
+      "loss": 0.782,
       "step": 29259
     },
     {
-      "epoch": 0.8036033066930324,
+      "epoch": 0.8303064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.9557706830123625e-06,
-      "loss": 0.7798,
+      "learning_rate": 1.4727100783150283e-06,
+      "loss": 0.8397,
       "step": 29260
     },
     {
-      "epoch": 0.8036307709208755,
+      "epoch": 0.8303348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 1.9552422916029e-06,
-      "loss": 0.8462,
+      "learning_rate": 1.4722300306085802e-06,
+      "loss": 0.8536,
       "step": 29261
     },
     {
-      "epoch": 0.8036582351487188,
+      "epoch": 0.8303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 1.954713963846675e-06,
-      "loss": 0.9251,
+      "learning_rate": 1.4717500549365848e-06,
+      "loss": 0.8073,
       "step": 29262
     },
     {
-      "epoch": 0.803685699376562,
+      "epoch": 0.8303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.954185699747868e-06,
-      "loss": 0.7705,
+      "learning_rate": 1.4712701513031015e-06,
+      "loss": 0.9244,
       "step": 29263
     },
     {
-      "epoch": 0.8037131636044053,
+      "epoch": 0.8304199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 1.953657499310665e-06,
-      "loss": 0.72,
+      "learning_rate": 1.470790319712183e-06,
+      "loss": 0.8706,
       "step": 29264
     },
     {
-      "epoch": 0.8037406278322485,
+      "epoch": 0.8304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.953129362539237e-06,
-      "loss": 0.8442,
+      "learning_rate": 1.470310560167879e-06,
+      "loss": 0.8921,
       "step": 29265
     },
     {
-      "epoch": 0.8037680920600917,
+      "epoch": 0.8304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.9526012894377653e-06,
-      "loss": 0.8642,
+      "learning_rate": 1.4698308726742493e-06,
+      "loss": 0.8258,
       "step": 29266
     },
     {
-      "epoch": 0.803795556287935,
+      "epoch": 0.830505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.9520732800104337e-06,
-      "loss": 0.7478,
+      "learning_rate": 1.4693512572353396e-06,
+      "loss": 0.8695,
       "step": 29267
     },
     {
-      "epoch": 0.8038230205157783,
+      "epoch": 0.830533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.951545334261413e-06,
-      "loss": 0.7942,
+      "learning_rate": 1.4688717138552032e-06,
+      "loss": 0.793,
       "step": 29268
     },
     {
-      "epoch": 0.8038504847436214,
+      "epoch": 0.8305618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.951017452194882e-06,
-      "loss": 0.6789,
+      "learning_rate": 1.4683922425378926e-06,
+      "loss": 0.8095,
       "step": 29269
     },
     {
-      "epoch": 0.8038779489714647,
+      "epoch": 0.8305902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.9504896338150204e-06,
-      "loss": 0.7595,
+      "learning_rate": 1.4679128432874545e-06,
+      "loss": 0.8499,
       "step": 29270
     },
     {
-      "epoch": 0.8039054131993079,
+      "epoch": 0.8306186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 1.9499618791260045e-06,
-      "loss": 0.7836,
+      "learning_rate": 1.4674335161079401e-06,
+      "loss": 0.7407,
       "step": 29271
     },
     {
-      "epoch": 0.8039328774271511,
+      "epoch": 0.8306469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.9494341881320065e-06,
-      "loss": 0.9366,
+      "learning_rate": 1.4669542610034016e-06,
+      "loss": 0.8009,
       "step": 29272
     },
     {
-      "epoch": 0.8039603416549944,
+      "epoch": 0.8306753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.948906560837206e-06,
-      "loss": 0.7382,
+      "learning_rate": 1.4664750779778802e-06,
+      "loss": 0.7857,
       "step": 29273
     },
     {
-      "epoch": 0.8039878058828376,
+      "epoch": 0.8307037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.9483789972457736e-06,
-      "loss": 0.7284,
+      "learning_rate": 1.4659959670354285e-06,
+      "loss": 0.8186,
       "step": 29274
     },
     {
-      "epoch": 0.8040152701106809,
+      "epoch": 0.8307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.9478514973618844e-06,
-      "loss": 0.8847,
+      "learning_rate": 1.465516928180094e-06,
+      "loss": 0.8132,
       "step": 29275
     },
     {
-      "epoch": 0.804042734338524,
+      "epoch": 0.8307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.9473240611897137e-06,
-      "loss": 0.8224,
+      "learning_rate": 1.4650379614159193e-06,
+      "loss": 0.823,
       "step": 29276
     },
     {
-      "epoch": 0.8040701985663673,
+      "epoch": 0.830788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.9467966887334344e-06,
-      "loss": 0.7887,
+      "learning_rate": 1.4645590667469533e-06,
+      "loss": 0.691,
       "step": 29277
     },
     {
-      "epoch": 0.8040976627942106,
+      "epoch": 0.8308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.946269379997219e-06,
-      "loss": 0.7245,
+      "learning_rate": 1.464080244177243e-06,
+      "loss": 0.8671,
       "step": 29278
     },
     {
-      "epoch": 0.8041251270220537,
+      "epoch": 0.8308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.9457421349852433e-06,
-      "loss": 0.8088,
+      "learning_rate": 1.4636014937108278e-06,
+      "loss": 0.7085,
       "step": 29279
     },
     {
-      "epoch": 0.804152591249897,
+      "epoch": 0.8308740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.9452149537016753e-06,
-      "loss": 0.767,
+      "learning_rate": 1.463122815351755e-06,
+      "loss": 0.7118,
       "step": 29280
     },
     {
-      "epoch": 0.8041800554777403,
+      "epoch": 0.8309023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.944687836150684e-06,
-      "loss": 0.8139,
+      "learning_rate": 1.462644209104067e-06,
+      "loss": 0.7978,
       "step": 29281
     },
     {
-      "epoch": 0.8042075197055835,
+      "epoch": 0.8309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.944160782336443e-06,
-      "loss": 0.765,
+      "learning_rate": 1.4621656749718072e-06,
+      "loss": 0.8301,
       "step": 29282
     },
     {
-      "epoch": 0.8042349839334267,
+      "epoch": 0.8309591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 1.943633792263122e-06,
-      "loss": 0.7408,
+      "learning_rate": 1.46168721295902e-06,
+      "loss": 0.9693,
       "step": 29283
     },
     {
-      "epoch": 0.8042624481612699,
+      "epoch": 0.8309875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.9431068659348905e-06,
-      "loss": 0.8564,
+      "learning_rate": 1.461208823069743e-06,
+      "loss": 0.8328,
       "step": 29284
     },
     {
-      "epoch": 0.8042899123891132,
+      "epoch": 0.8310158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 1.9425800033559226e-06,
-      "loss": 0.7864,
+      "learning_rate": 1.4607305053080179e-06,
+      "loss": 0.8111,
       "step": 29285
     },
     {
-      "epoch": 0.8043173766169565,
+      "epoch": 0.831044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.9420532045303795e-06,
-      "loss": 0.8119,
+      "learning_rate": 1.4602522596778889e-06,
+      "loss": 0.773,
       "step": 29286
     },
     {
-      "epoch": 0.8043448408447996,
+      "epoch": 0.8310726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.9415264694624327e-06,
-      "loss": 0.9382,
+      "learning_rate": 1.4597740861833909e-06,
+      "loss": 0.8232,
       "step": 29287
     },
     {
-      "epoch": 0.8043723050726429,
+      "epoch": 0.8311010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 1.9409997981562535e-06,
-      "loss": 0.7334,
+      "learning_rate": 1.4592959848285649e-06,
+      "loss": 0.8847,
       "step": 29288
     },
     {
-      "epoch": 0.8043997693004861,
+      "epoch": 0.8311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.9404731906160024e-06,
-      "loss": 0.8226,
+      "learning_rate": 1.458817955617452e-06,
+      "loss": 0.9457,
       "step": 29289
     },
     {
-      "epoch": 0.8044272335283293,
+      "epoch": 0.8311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.9399466468458495e-06,
-      "loss": 0.788,
+      "learning_rate": 1.4583399985540859e-06,
+      "loss": 0.7829,
       "step": 29290
     },
     {
-      "epoch": 0.8044546977561726,
+      "epoch": 0.8311861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 1.9394201668499634e-06,
-      "loss": 0.7308,
+      "learning_rate": 1.4578621136425053e-06,
+      "loss": 0.9142,
       "step": 29291
     },
     {
-      "epoch": 0.8044821619840158,
+      "epoch": 0.8312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.9388937506325047e-06,
-      "loss": 0.7663,
+      "learning_rate": 1.4573843008867495e-06,
+      "loss": 0.7511,
       "step": 29292
     },
     {
-      "epoch": 0.8045096262118591,
+      "epoch": 0.8312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.9383673981976404e-06,
-      "loss": 0.8807,
+      "learning_rate": 1.4569065602908506e-06,
+      "loss": 0.8209,
       "step": 29293
     },
     {
-      "epoch": 0.8045370904397023,
+      "epoch": 0.8312712826333711,
       "grad_norm": 0.0,
-      "learning_rate": 1.9378411095495374e-06,
-      "loss": 0.763,
+      "learning_rate": 1.4564288918588465e-06,
+      "loss": 0.7423,
       "step": 29294
     },
     {
-      "epoch": 0.8045645546675455,
+      "epoch": 0.8312996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 1.9373148846923608e-06,
-      "loss": 0.8013,
+      "learning_rate": 1.4559512955947708e-06,
+      "loss": 0.8434,
       "step": 29295
     },
     {
-      "epoch": 0.8045920188953888,
+      "epoch": 0.831328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.9367887236302684e-06,
-      "loss": 0.9513,
+      "learning_rate": 1.455473771502659e-06,
+      "loss": 0.8957,
       "step": 29296
     },
     {
-      "epoch": 0.804619483123232,
+      "epoch": 0.8313564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.936262626367431e-06,
-      "loss": 0.799,
+      "learning_rate": 1.454996319586547e-06,
+      "loss": 0.7962,
       "step": 29297
     },
     {
-      "epoch": 0.8046469473510752,
+      "epoch": 0.8313847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.935736592908004e-06,
-      "loss": 0.8151,
+      "learning_rate": 1.4545189398504622e-06,
+      "loss": 0.8168,
       "step": 29298
     },
     {
-      "epoch": 0.8046744115789185,
+      "epoch": 0.8314131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.935210623256153e-06,
-      "loss": 0.7822,
+      "learning_rate": 1.4540416322984396e-06,
+      "loss": 0.8021,
       "step": 29299
     },
     {
-      "epoch": 0.8047018758067617,
+      "epoch": 0.8314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.9346847174160387e-06,
-      "loss": 0.8245,
+      "learning_rate": 1.4535643969345149e-06,
+      "loss": 0.9205,
       "step": 29300
     },
     {
-      "epoch": 0.804729340034605,
+      "epoch": 0.8314699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.934158875391824e-06,
-      "loss": 0.7873,
+      "learning_rate": 1.4530872337627132e-06,
+      "loss": 0.8126,
       "step": 29301
     },
     {
-      "epoch": 0.8047568042624481,
+      "epoch": 0.8314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.9336330971876717e-06,
-      "loss": 0.8076,
+      "learning_rate": 1.4526101427870675e-06,
+      "loss": 0.8159,
       "step": 29302
     },
     {
-      "epoch": 0.8047842684902914,
+      "epoch": 0.8315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.933107382807736e-06,
-      "loss": 0.7953,
+      "learning_rate": 1.452133124011611e-06,
+      "loss": 0.8326,
       "step": 29303
     },
     {
-      "epoch": 0.8048117327181347,
+      "epoch": 0.8315550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.9325817322561815e-06,
-      "loss": 0.8837,
+      "learning_rate": 1.4516561774403682e-06,
+      "loss": 0.8353,
       "step": 29304
     },
     {
-      "epoch": 0.8048391969459778,
+      "epoch": 0.831583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9320561455371635e-06,
-      "loss": 0.7689,
+      "learning_rate": 1.451179303077369e-06,
+      "loss": 0.8407,
       "step": 29305
     },
     {
-      "epoch": 0.8048666611738211,
+      "epoch": 0.8316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.9315306226548424e-06,
-      "loss": 0.807,
+      "learning_rate": 1.450702500926645e-06,
+      "loss": 0.7386,
       "step": 29306
     },
     {
-      "epoch": 0.8048941254016644,
+      "epoch": 0.8316401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.9310051636133764e-06,
-      "loss": 0.763,
+      "learning_rate": 1.4502257709922197e-06,
+      "loss": 0.7662,
       "step": 29307
     },
     {
-      "epoch": 0.8049215896295076,
+      "epoch": 0.8316685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.930479768416925e-06,
-      "loss": 0.7106,
+      "learning_rate": 1.4497491132781216e-06,
+      "loss": 0.8603,
       "step": 29308
     },
     {
-      "epoch": 0.8049490538573508,
+      "epoch": 0.8316969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.929954437069642e-06,
-      "loss": 0.8159,
+      "learning_rate": 1.4492725277883791e-06,
+      "loss": 0.9175,
       "step": 29309
     },
     {
-      "epoch": 0.804976518085194,
+      "epoch": 0.8317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 1.929429169575685e-06,
-      "loss": 0.815,
+      "learning_rate": 1.448796014527013e-06,
+      "loss": 0.8099,
       "step": 29310
     },
     {
-      "epoch": 0.8050039823130373,
+      "epoch": 0.8317536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.9289039659392106e-06,
-      "loss": 0.8038,
+      "learning_rate": 1.4483195734980504e-06,
+      "loss": 0.8335,
       "step": 29311
     },
     {
-      "epoch": 0.8050314465408805,
+      "epoch": 0.8317820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 1.928378826164378e-06,
-      "loss": 0.7756,
+      "learning_rate": 1.4478432047055202e-06,
+      "loss": 0.8372,
       "step": 29312
     },
     {
-      "epoch": 0.8050589107687237,
+      "epoch": 0.8318104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 1.927853750255334e-06,
-      "loss": 0.8428,
+      "learning_rate": 1.4473669081534414e-06,
+      "loss": 0.8638,
       "step": 29313
     },
     {
-      "epoch": 0.805086374996567,
+      "epoch": 0.831838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.927328738216242e-06,
-      "loss": 0.8036,
+      "learning_rate": 1.4468906838458385e-06,
+      "loss": 0.8816,
       "step": 29314
     },
     {
-      "epoch": 0.8051138392244102,
+      "epoch": 0.8318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.926803790051249e-06,
-      "loss": 0.8354,
+      "learning_rate": 1.4464145317867372e-06,
+      "loss": 0.8131,
       "step": 29315
     },
     {
-      "epoch": 0.8051413034522534,
+      "epoch": 0.8318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.9262789057645126e-06,
-      "loss": 0.7598,
+      "learning_rate": 1.4459384519801533e-06,
+      "loss": 0.8522,
       "step": 29316
     },
     {
-      "epoch": 0.8051687676800967,
+      "epoch": 0.8319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.9257540853601843e-06,
-      "loss": 0.8315,
+      "learning_rate": 1.4454624444301135e-06,
+      "loss": 0.8404,
       "step": 29317
     },
     {
-      "epoch": 0.8051962319079399,
+      "epoch": 0.8319523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 1.9252293288424186e-06,
-      "loss": 0.8097,
+      "learning_rate": 1.444986509140638e-06,
+      "loss": 0.7387,
       "step": 29318
     },
     {
-      "epoch": 0.8052236961357832,
+      "epoch": 0.8319807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 1.9247046362153677e-06,
-      "loss": 0.8378,
+      "learning_rate": 1.4445106461157443e-06,
+      "loss": 0.8555,
       "step": 29319
     },
     {
-      "epoch": 0.8052511603636264,
+      "epoch": 0.8320090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.9241800074831787e-06,
-      "loss": 0.8711,
+      "learning_rate": 1.444034855359453e-06,
+      "loss": 0.7638,
       "step": 29320
     },
     {
-      "epoch": 0.8052786245914696,
+      "epoch": 0.8320374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 1.9236554426500087e-06,
-      "loss": 0.7862,
+      "learning_rate": 1.4435591368757872e-06,
+      "loss": 0.8374,
       "step": 29321
     },
     {
-      "epoch": 0.8053060888193129,
+      "epoch": 0.8320658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 1.9231309417200027e-06,
-      "loss": 0.8596,
+      "learning_rate": 1.4430834906687597e-06,
+      "loss": 0.7609,
       "step": 29322
     },
     {
-      "epoch": 0.805333553047156,
+      "epoch": 0.8320942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 1.9226065046973143e-06,
-      "loss": 0.7076,
+      "learning_rate": 1.4426079167423923e-06,
+      "loss": 0.8342,
       "step": 29323
     },
     {
-      "epoch": 0.8053610172749993,
+      "epoch": 0.8321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.9220821315860916e-06,
-      "loss": 0.8572,
+      "learning_rate": 1.4421324151006988e-06,
+      "loss": 0.834,
       "step": 29324
     },
     {
-      "epoch": 0.8053884815028426,
+      "epoch": 0.8321509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.921557822390484e-06,
-      "loss": 0.7961,
+      "learning_rate": 1.441656985747697e-06,
+      "loss": 0.8311,
       "step": 29325
     },
     {
-      "epoch": 0.8054159457306858,
+      "epoch": 0.8321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.9210335771146426e-06,
-      "loss": 0.9391,
+      "learning_rate": 1.4411816286874036e-06,
+      "loss": 0.8676,
       "step": 29326
     },
     {
-      "epoch": 0.805443409958529,
+      "epoch": 0.8322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.9205093957627108e-06,
-      "loss": 0.8468,
+      "learning_rate": 1.4407063439238333e-06,
+      "loss": 0.7919,
       "step": 29327
     },
     {
-      "epoch": 0.8054708741863722,
+      "epoch": 0.8322360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.9199852783388407e-06,
-      "loss": 0.9179,
+      "learning_rate": 1.4402311314610019e-06,
+      "loss": 0.7479,
       "step": 29328
     },
     {
-      "epoch": 0.8054983384142155,
+      "epoch": 0.8322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.9194612248471744e-06,
-      "loss": 0.8607,
+      "learning_rate": 1.4397559913029247e-06,
+      "loss": 0.7584,
       "step": 29329
     },
     {
-      "epoch": 0.8055258026420588,
+      "epoch": 0.8322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.9189372352918613e-06,
-      "loss": 0.7726,
+      "learning_rate": 1.439280923453612e-06,
+      "loss": 0.8411,
       "step": 29330
     },
     {
-      "epoch": 0.8055532668699019,
+      "epoch": 0.8323212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.918413309677046e-06,
-      "loss": 0.8083,
+      "learning_rate": 1.4388059279170774e-06,
+      "loss": 0.7973,
       "step": 29331
     },
     {
-      "epoch": 0.8055807310977452,
+      "epoch": 0.832349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.917889448006879e-06,
-      "loss": 0.7646,
+      "learning_rate": 1.4383310046973365e-06,
+      "loss": 0.7964,
       "step": 29332
     },
     {
-      "epoch": 0.8056081953255885,
+      "epoch": 0.832377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.9173656502854986e-06,
-      "loss": 0.8583,
+      "learning_rate": 1.437856153798396e-06,
+      "loss": 0.8374,
       "step": 29333
     },
     {
-      "epoch": 0.8056356595534316,
+      "epoch": 0.8324063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 1.9168419165170514e-06,
-      "loss": 0.7298,
+      "learning_rate": 1.4373813752242694e-06,
+      "loss": 0.9408,
       "step": 29334
     },
     {
-      "epoch": 0.8056631237812749,
+      "epoch": 0.8324347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.9163182467056827e-06,
-      "loss": 0.7507,
+      "learning_rate": 1.4369066689789702e-06,
+      "loss": 0.8055,
       "step": 29335
     },
     {
-      "epoch": 0.8056905880091181,
+      "epoch": 0.8324631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 1.915794640855537e-06,
-      "loss": 0.76,
+      "learning_rate": 1.4364320350665018e-06,
+      "loss": 0.7576,
       "step": 29336
     },
     {
-      "epoch": 0.8057180522369614,
+      "epoch": 0.8324914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 1.9152710989707545e-06,
-      "loss": 0.8221,
+      "learning_rate": 1.4359574734908777e-06,
+      "loss": 0.8828,
       "step": 29337
     },
     {
-      "epoch": 0.8057455164648046,
+      "epoch": 0.8325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.9147476210554806e-06,
-      "loss": 0.807,
+      "learning_rate": 1.435482984256108e-06,
+      "loss": 0.9105,
       "step": 29338
     },
     {
-      "epoch": 0.8057729806926478,
+      "epoch": 0.8325482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 1.9142242071138527e-06,
-      "loss": 0.8442,
+      "learning_rate": 1.4350085673661974e-06,
+      "loss": 0.9063,
       "step": 29339
     },
     {
-      "epoch": 0.8058004449204911,
+      "epoch": 0.8325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.913700857150016e-06,
-      "loss": 0.7798,
+      "learning_rate": 1.4345342228251524e-06,
+      "loss": 0.7696,
       "step": 29340
     },
     {
-      "epoch": 0.8058279091483342,
+      "epoch": 0.8326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.9131775711681087e-06,
-      "loss": 0.9026,
+      "learning_rate": 1.4340599506369835e-06,
+      "loss": 0.8122,
       "step": 29341
     },
     {
-      "epoch": 0.8058553733761775,
+      "epoch": 0.832633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.912654349172275e-06,
-      "loss": 0.8452,
+      "learning_rate": 1.4335857508056949e-06,
+      "loss": 0.7718,
       "step": 29342
     },
     {
-      "epoch": 0.8058828376040208,
+      "epoch": 0.8326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.912131191166654e-06,
-      "loss": 0.7873,
+      "learning_rate": 1.4331116233352938e-06,
+      "loss": 0.8122,
       "step": 29343
     },
     {
-      "epoch": 0.805910301831864,
+      "epoch": 0.8326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.911608097155382e-06,
-      "loss": 0.7997,
+      "learning_rate": 1.4326375682297833e-06,
+      "loss": 0.7485,
       "step": 29344
     },
     {
-      "epoch": 0.8059377660597072,
+      "epoch": 0.8327185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 1.9110850671426017e-06,
-      "loss": 0.8474,
+      "learning_rate": 1.4321635854931671e-06,
+      "loss": 0.8456,
       "step": 29345
     },
     {
-      "epoch": 0.8059652302875505,
+      "epoch": 0.8327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 1.910562101132447e-06,
-      "loss": 0.7671,
+      "learning_rate": 1.4316896751294528e-06,
+      "loss": 0.7649,
       "step": 29346
     },
     {
-      "epoch": 0.8059926945153937,
+      "epoch": 0.8327752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.9100391991290602e-06,
-      "loss": 0.8608,
+      "learning_rate": 1.4312158371426388e-06,
+      "loss": 0.9178,
       "step": 29347
     },
     {
-      "epoch": 0.806020158743237,
+      "epoch": 0.8328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.9095163611365752e-06,
-      "loss": 0.7978,
+      "learning_rate": 1.4307420715367304e-06,
+      "loss": 0.7775,
       "step": 29348
     },
     {
-      "epoch": 0.8060476229710801,
+      "epoch": 0.8328320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.9089935871591312e-06,
-      "loss": 0.7684,
+      "learning_rate": 1.43026837831573e-06,
+      "loss": 0.826,
       "step": 29349
     },
     {
-      "epoch": 0.8060750871989234,
+      "epoch": 0.8328603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.9084708772008677e-06,
-      "loss": 0.9174,
+      "learning_rate": 1.4297947574836367e-06,
+      "loss": 0.717,
       "step": 29350
     },
     {
-      "epoch": 0.8061025514267667,
+      "epoch": 0.83288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.907948231265914e-06,
-      "loss": 0.9796,
+      "learning_rate": 1.4293212090444518e-06,
+      "loss": 0.8933,
       "step": 29351
     },
     {
-      "epoch": 0.8061300156546098,
+      "epoch": 0.8329171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.9074256493584096e-06,
-      "loss": 0.8194,
+      "learning_rate": 1.4288477330021777e-06,
+      "loss": 0.8338,
       "step": 29352
     },
     {
-      "epoch": 0.8061574798824531,
+      "epoch": 0.8329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.906903131482487e-06,
-      "loss": 0.7868,
+      "learning_rate": 1.428374329360811e-06,
+      "loss": 0.8693,
       "step": 29353
     },
     {
-      "epoch": 0.8061849441102963,
+      "epoch": 0.8329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.9063806776422822e-06,
-      "loss": 0.9313,
+      "learning_rate": 1.4279009981243507e-06,
+      "loss": 0.7809,
       "step": 29354
     },
     {
-      "epoch": 0.8062124083381396,
+      "epoch": 0.8330022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.9058582878419275e-06,
-      "loss": 0.9286,
+      "learning_rate": 1.4274277392967982e-06,
+      "loss": 0.8001,
       "step": 29355
     },
     {
-      "epoch": 0.8062398725659828,
+      "epoch": 0.8330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.9053359620855605e-06,
-      "loss": 0.8342,
+      "learning_rate": 1.4269545528821483e-06,
+      "loss": 0.7441,
       "step": 29356
     },
     {
-      "epoch": 0.806267336793826,
+      "epoch": 0.8330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.9048137003773082e-06,
-      "loss": 0.8052,
+      "learning_rate": 1.4264814388843973e-06,
+      "loss": 0.9287,
       "step": 29357
     },
     {
-      "epoch": 0.8062948010216693,
+      "epoch": 0.8330874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.904291502721306e-06,
-      "loss": 0.8245,
+      "learning_rate": 1.4260083973075433e-06,
+      "loss": 0.7875,
       "step": 29358
     },
     {
-      "epoch": 0.8063222652495126,
+      "epoch": 0.8331157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 1.9037693691216852e-06,
-      "loss": 0.8281,
+      "learning_rate": 1.425535428155581e-06,
+      "loss": 0.8026,
       "step": 29359
     },
     {
-      "epoch": 0.8063497294773557,
+      "epoch": 0.8331441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.903247299582579e-06,
-      "loss": 0.7363,
+      "learning_rate": 1.4250625314325095e-06,
+      "loss": 0.8318,
       "step": 29360
     },
     {
-      "epoch": 0.806377193705199,
+      "epoch": 0.8331725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 1.9027252941081175e-06,
-      "loss": 0.7455,
+      "learning_rate": 1.4245897071423175e-06,
+      "loss": 0.8907,
       "step": 29361
     },
     {
-      "epoch": 0.8064046579330422,
+      "epoch": 0.8332009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.9022033527024275e-06,
-      "loss": 0.7789,
+      "learning_rate": 1.4241169552890022e-06,
+      "loss": 0.7908,
       "step": 29362
     },
     {
-      "epoch": 0.8064321221608854,
+      "epoch": 0.8332292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 1.9016814753696411e-06,
-      "loss": 0.8369,
+      "learning_rate": 1.423644275876559e-06,
+      "loss": 0.7645,
       "step": 29363
     },
     {
-      "epoch": 0.8064595863887287,
+      "epoch": 0.8332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.9011596621138883e-06,
-      "loss": 0.8746,
+      "learning_rate": 1.4231716689089757e-06,
+      "loss": 0.8275,
       "step": 29364
     },
     {
-      "epoch": 0.8064870506165719,
+      "epoch": 0.8332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.900637912939297e-06,
-      "loss": 0.7295,
+      "learning_rate": 1.4226991343902463e-06,
+      "loss": 0.7332,
       "step": 29365
     },
     {
-      "epoch": 0.8065145148444152,
+      "epoch": 0.8333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.9001162278499962e-06,
-      "loss": 0.7993,
+      "learning_rate": 1.422226672324366e-06,
+      "loss": 0.7899,
       "step": 29366
     },
     {
-      "epoch": 0.8065419790722583,
+      "epoch": 0.8333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8995946068501158e-06,
-      "loss": 0.7609,
+      "learning_rate": 1.4217542827153196e-06,
+      "loss": 0.8014,
       "step": 29367
     },
     {
-      "epoch": 0.8065694433001016,
+      "epoch": 0.8333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.8990730499437816e-06,
-      "loss": 0.7094,
+      "learning_rate": 1.4212819655671006e-06,
+      "loss": 0.8236,
       "step": 29368
     },
     {
-      "epoch": 0.8065969075279449,
+      "epoch": 0.833399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.8985515571351154e-06,
-      "loss": 0.7484,
+      "learning_rate": 1.4208097208837013e-06,
+      "loss": 0.6689,
       "step": 29369
     },
     {
-      "epoch": 0.806624371755788,
+      "epoch": 0.833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.8980301284282477e-06,
-      "loss": 0.8802,
+      "learning_rate": 1.420337548669105e-06,
+      "loss": 0.7874,
       "step": 29370
     },
     {
-      "epoch": 0.8066518359836313,
+      "epoch": 0.8334562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.8975087638273049e-06,
-      "loss": 0.8777,
+      "learning_rate": 1.4198654489273035e-06,
+      "loss": 0.7998,
       "step": 29371
     },
     {
-      "epoch": 0.8066793002114746,
+      "epoch": 0.8334846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 1.8969874633364105e-06,
-      "loss": 0.8645,
+      "learning_rate": 1.419393421662284e-06,
+      "loss": 0.7936,
       "step": 29372
     },
     {
-      "epoch": 0.8067067644393178,
+      "epoch": 0.8335130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 1.8964662269596935e-06,
-      "loss": 0.7913,
+      "learning_rate": 1.4189214668780337e-06,
+      "loss": 0.712,
       "step": 29373
     },
     {
-      "epoch": 0.806734228667161,
+      "epoch": 0.8335414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.8959450547012714e-06,
-      "loss": 0.747,
+      "learning_rate": 1.4184495845785418e-06,
+      "loss": 0.8137,
       "step": 29374
     },
     {
-      "epoch": 0.8067616928950042,
+      "epoch": 0.8335698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.8954239465652714e-06,
-      "loss": 0.8248,
+      "learning_rate": 1.4179777747677903e-06,
+      "loss": 0.7686,
       "step": 29375
     },
     {
-      "epoch": 0.8067891571228475,
+      "epoch": 0.8335981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.8949029025558196e-06,
-      "loss": 0.7996,
+      "learning_rate": 1.4175060374497662e-06,
+      "loss": 0.8372,
       "step": 29376
     },
     {
-      "epoch": 0.8068166213506908,
+      "epoch": 0.8336265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.894381922677032e-06,
-      "loss": 0.7596,
+      "learning_rate": 1.4170343726284564e-06,
+      "loss": 0.8499,
       "step": 29377
     },
     {
-      "epoch": 0.8068440855785339,
+      "epoch": 0.8336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.8938610069330354e-06,
-      "loss": 0.7023,
+      "learning_rate": 1.4165627803078418e-06,
+      "loss": 0.7038,
       "step": 29378
     },
     {
-      "epoch": 0.8068715498063772,
+      "epoch": 0.833683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.893340155327953e-06,
-      "loss": 0.8537,
+      "learning_rate": 1.4160912604919063e-06,
+      "loss": 0.799,
       "step": 29379
     },
     {
-      "epoch": 0.8068990140342204,
+      "epoch": 0.8337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.8928193678659012e-06,
-      "loss": 0.7769,
+      "learning_rate": 1.4156198131846367e-06,
+      "loss": 0.7906,
       "step": 29380
     },
     {
-      "epoch": 0.8069264782620637,
+      "epoch": 0.8337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.8922986445510017e-06,
-      "loss": 0.8328,
+      "learning_rate": 1.4151484383900093e-06,
+      "loss": 0.7353,
       "step": 29381
     },
     {
-      "epoch": 0.8069539424899069,
+      "epoch": 0.8337684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.8917779853873775e-06,
-      "loss": 0.7422,
+      "learning_rate": 1.4146771361120094e-06,
+      "loss": 0.9139,
       "step": 29382
     },
     {
-      "epoch": 0.8069814067177501,
+      "epoch": 0.8337968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.8912573903791453e-06,
-      "loss": 0.6628,
+      "learning_rate": 1.4142059063546199e-06,
+      "loss": 0.7812,
       "step": 29383
     },
     {
-      "epoch": 0.8070088709455934,
+      "epoch": 0.8338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.8907368595304298e-06,
-      "loss": 0.7921,
+      "learning_rate": 1.4137347491218167e-06,
+      "loss": 0.7484,
       "step": 29384
     },
     {
-      "epoch": 0.8070363351734366,
+      "epoch": 0.8338535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.8902163928453443e-06,
-      "loss": 0.9007,
+      "learning_rate": 1.4132636644175824e-06,
+      "loss": 0.808,
       "step": 29385
     },
     {
-      "epoch": 0.8070637994012798,
+      "epoch": 0.8338819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.8896959903280054e-06,
-      "loss": 0.8538,
+      "learning_rate": 1.412792652245898e-06,
+      "loss": 0.7942,
       "step": 29386
     },
     {
-      "epoch": 0.8070912636291231,
+      "epoch": 0.8339103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 1.8891756519825344e-06,
-      "loss": 0.7405,
+      "learning_rate": 1.412321712610737e-06,
+      "loss": 0.9127,
       "step": 29387
     },
     {
-      "epoch": 0.8071187278569663,
+      "epoch": 0.833938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.8886553778130468e-06,
-      "loss": 0.7931,
+      "learning_rate": 1.4118508455160818e-06,
+      "loss": 0.8405,
       "step": 29388
     },
     {
-      "epoch": 0.8071461920848095,
+      "epoch": 0.8339670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.8881351678236603e-06,
-      "loss": 0.8073,
+      "learning_rate": 1.411380050965908e-06,
+      "loss": 0.8564,
       "step": 29389
     },
     {
-      "epoch": 0.8071736563126528,
+      "epoch": 0.8339954597048809,
       "grad_norm": 0.0,
-      "learning_rate": 1.8876150220184907e-06,
-      "loss": 0.8357,
+      "learning_rate": 1.410909328964193e-06,
+      "loss": 0.6507,
       "step": 29390
     },
     {
-      "epoch": 0.807201120540496,
+      "epoch": 0.8340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.8870949404016558e-06,
-      "loss": 0.7911,
+      "learning_rate": 1.410438679514915e-06,
+      "loss": 0.7671,
       "step": 29391
     },
     {
-      "epoch": 0.8072285847683393,
+      "epoch": 0.8340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.8865749229772678e-06,
-      "loss": 0.7538,
+      "learning_rate": 1.4099681026220457e-06,
+      "loss": 0.7831,
       "step": 29392
     },
     {
-      "epoch": 0.8072560489961824,
+      "epoch": 0.8340805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 1.8860549697494402e-06,
-      "loss": 0.7556,
+      "learning_rate": 1.409497598289562e-06,
+      "loss": 0.7684,
       "step": 29393
     },
     {
-      "epoch": 0.8072835132240257,
+      "epoch": 0.8341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.8855350807222883e-06,
-      "loss": 0.9075,
+      "learning_rate": 1.4090271665214405e-06,
+      "loss": 0.7215,
       "step": 29394
     },
     {
-      "epoch": 0.807310977451869,
+      "epoch": 0.8341373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.8850152558999258e-06,
-      "loss": 0.8799,
+      "learning_rate": 1.4085568073216505e-06,
+      "loss": 0.8255,
       "step": 29395
     },
     {
-      "epoch": 0.8073384416797121,
+      "epoch": 0.8341657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 1.884495495286467e-06,
-      "loss": 0.7379,
+      "learning_rate": 1.4080865206941674e-06,
+      "loss": 0.6745,
       "step": 29396
     },
     {
-      "epoch": 0.8073659059075554,
+      "epoch": 0.8341940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 1.8839757988860252e-06,
-      "loss": 0.743,
+      "learning_rate": 1.407616306642966e-06,
+      "loss": 0.8313,
       "step": 29397
     },
     {
-      "epoch": 0.8073933701353987,
+      "epoch": 0.83422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.8834561667027086e-06,
-      "loss": 0.788,
+      "learning_rate": 1.407146165172013e-06,
+      "loss": 0.7999,
       "step": 29398
     },
     {
-      "epoch": 0.8074208343632419,
+      "epoch": 0.8342508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 1.8829365987406301e-06,
-      "loss": 0.8261,
+      "learning_rate": 1.4066760962852832e-06,
+      "loss": 0.8404,
       "step": 29399
     },
     {
-      "epoch": 0.8074482985910851,
+      "epoch": 0.8342792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.8824170950039044e-06,
-      "loss": 0.8375,
+      "learning_rate": 1.4062060999867489e-06,
+      "loss": 0.7866,
       "step": 29400
     },
     {
-      "epoch": 0.8074757628189283,
+      "epoch": 0.8343076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.8818976554966372e-06,
-      "loss": 0.8831,
+      "learning_rate": 1.4057361762803756e-06,
+      "loss": 0.7825,
       "step": 29401
     },
     {
-      "epoch": 0.8075032270467716,
+      "epoch": 0.8343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.8813782802229396e-06,
-      "loss": 0.8777,
+      "learning_rate": 1.405266325170136e-06,
+      "loss": 0.6745,
       "step": 29402
     },
     {
-      "epoch": 0.8075306912746149,
+      "epoch": 0.8343643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.880858969186925e-06,
-      "loss": 0.9859,
+      "learning_rate": 1.4047965466599967e-06,
+      "loss": 0.8428,
       "step": 29403
     },
     {
-      "epoch": 0.807558155502458,
+      "epoch": 0.8343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.8803397223926967e-06,
-      "loss": 0.8356,
+      "learning_rate": 1.404326840753929e-06,
+      "loss": 0.9674,
       "step": 29404
     },
     {
-      "epoch": 0.8075856197303013,
+      "epoch": 0.8344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.879820539844366e-06,
-      "loss": 0.7168,
+      "learning_rate": 1.4038572074558997e-06,
+      "loss": 0.8277,
       "step": 29405
     },
     {
-      "epoch": 0.8076130839581445,
+      "epoch": 0.8344494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.8793014215460393e-06,
-      "loss": 0.8005,
+      "learning_rate": 1.4033876467698738e-06,
+      "loss": 0.781,
       "step": 29406
     },
     {
-      "epoch": 0.8076405481859877,
+      "epoch": 0.834477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.8787823675018291e-06,
-      "loss": 0.9141,
+      "learning_rate": 1.4029181586998175e-06,
+      "loss": 0.7418,
       "step": 29407
     },
     {
-      "epoch": 0.807668012413831,
+      "epoch": 0.8345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.8782633777158355e-06,
-      "loss": 0.8007,
+      "learning_rate": 1.4024487432497013e-06,
+      "loss": 0.714,
       "step": 29408
     },
     {
-      "epoch": 0.8076954766416742,
+      "epoch": 0.8345346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.8777444521921718e-06,
-      "loss": 0.9556,
+      "learning_rate": 1.4019794004234844e-06,
+      "loss": 0.7961,
       "step": 29409
     },
     {
-      "epoch": 0.8077229408695175,
+      "epoch": 0.8345629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.8772255909349368e-06,
-      "loss": 0.8105,
+      "learning_rate": 1.4015101302251343e-06,
+      "loss": 0.8519,
       "step": 29410
     },
     {
-      "epoch": 0.8077504050973607,
+      "epoch": 0.8345913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 1.8767067939482398e-06,
-      "loss": 0.7001,
+      "learning_rate": 1.401040932658617e-06,
+      "loss": 0.6943,
       "step": 29411
     },
     {
-      "epoch": 0.8077778693252039,
+      "epoch": 0.8346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 1.876188061236185e-06,
-      "loss": 0.8487,
+      "learning_rate": 1.4005718077278907e-06,
+      "loss": 0.9004,
       "step": 29412
     },
     {
-      "epoch": 0.8078053335530472,
+      "epoch": 0.8346481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.875669392802877e-06,
-      "loss": 0.8499,
+      "learning_rate": 1.4001027554369217e-06,
+      "loss": 0.8764,
       "step": 29413
     },
     {
-      "epoch": 0.8078327977808903,
+      "epoch": 0.8346765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 1.87515078865242e-06,
-      "loss": 0.7439,
+      "learning_rate": 1.3996337757896726e-06,
+      "loss": 0.7701,
       "step": 29414
     },
     {
-      "epoch": 0.8078602620087336,
+      "epoch": 0.8347048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.8746322487889202e-06,
-      "loss": 0.8104,
+      "learning_rate": 1.3991648687901017e-06,
+      "loss": 0.8339,
       "step": 29415
     },
     {
-      "epoch": 0.8078877262365769,
+      "epoch": 0.834733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.8741137732164771e-06,
-      "loss": 0.8198,
+      "learning_rate": 1.3986960344421718e-06,
+      "loss": 0.7451,
       "step": 29416
     },
     {
-      "epoch": 0.8079151904644201,
+      "epoch": 0.8347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.87359536193919e-06,
-      "loss": 0.8648,
+      "learning_rate": 1.398227272749846e-06,
+      "loss": 0.8047,
       "step": 29417
     },
     {
-      "epoch": 0.8079426546922633,
+      "epoch": 0.8347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.8730770149611655e-06,
-      "loss": 0.8261,
+      "learning_rate": 1.397758583717076e-06,
+      "loss": 0.8678,
       "step": 29418
     },
     {
-      "epoch": 0.8079701189201065,
+      "epoch": 0.8348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.8725587322865035e-06,
-      "loss": 0.8252,
+      "learning_rate": 1.397289967347828e-06,
+      "loss": 0.818,
       "step": 29419
     },
     {
-      "epoch": 0.8079975831479498,
+      "epoch": 0.8348467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 1.8720405139193077e-06,
-      "loss": 0.8364,
+      "learning_rate": 1.396821423646062e-06,
+      "loss": 0.6978,
       "step": 29420
     },
     {
-      "epoch": 0.8080250473757931,
+      "epoch": 0.8348751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 1.8715223598636734e-06,
-      "loss": 0.7795,
+      "learning_rate": 1.3963529526157283e-06,
+      "loss": 0.7525,
       "step": 29421
     },
     {
-      "epoch": 0.8080525116036362,
+      "epoch": 0.8349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.871004270123702e-06,
-      "loss": 0.7503,
+      "learning_rate": 1.395884554260789e-06,
+      "loss": 0.7666,
       "step": 29422
     },
     {
-      "epoch": 0.8080799758314795,
+      "epoch": 0.8349318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 1.8704862447034943e-06,
-      "loss": 0.7874,
+      "learning_rate": 1.3954162285852013e-06,
+      "loss": 0.834,
       "step": 29423
     },
     {
-      "epoch": 0.8081074400593228,
+      "epoch": 0.8349602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 1.8699682836071508e-06,
-      "loss": 0.8413,
+      "learning_rate": 1.3949479755929174e-06,
+      "loss": 0.8059,
       "step": 29424
     },
     {
-      "epoch": 0.808134904287166,
+      "epoch": 0.8349886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 1.869450386838766e-06,
-      "loss": 0.8416,
+      "learning_rate": 1.394479795287894e-06,
+      "loss": 0.7963,
       "step": 29425
     },
     {
-      "epoch": 0.8081623685150092,
+      "epoch": 0.8350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 1.8689325544024405e-06,
-      "loss": 0.7272,
+      "learning_rate": 1.3940116876740905e-06,
+      "loss": 0.8307,
       "step": 29426
     },
     {
-      "epoch": 0.8081898327428524,
+      "epoch": 0.8350454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.8684147863022684e-06,
-      "loss": 0.8392,
+      "learning_rate": 1.3935436527554535e-06,
+      "loss": 0.8183,
       "step": 29427
     },
     {
-      "epoch": 0.8082172969706957,
+      "epoch": 0.8350737797956868,
       "grad_norm": 0.0,
-      "learning_rate": 1.8678970825423481e-06,
-      "loss": 0.7939,
+      "learning_rate": 1.393075690535941e-06,
+      "loss": 0.7567,
       "step": 29428
     },
     {
-      "epoch": 0.8082447611985389,
+      "epoch": 0.8351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.8673794431267767e-06,
-      "loss": 0.7338,
+      "learning_rate": 1.3926078010195066e-06,
+      "loss": 0.8467,
       "step": 29429
     },
     {
-      "epoch": 0.8082722254263821,
+      "epoch": 0.8351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.8668618680596485e-06,
-      "loss": 0.8068,
+      "learning_rate": 1.392139984210099e-06,
+      "loss": 0.7962,
       "step": 29430
     },
     {
-      "epoch": 0.8082996896542254,
+      "epoch": 0.8351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.8663443573450635e-06,
-      "loss": 0.7682,
+      "learning_rate": 1.3916722401116756e-06,
+      "loss": 0.6983,
       "step": 29431
     },
     {
-      "epoch": 0.8083271538820687,
+      "epoch": 0.8351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.8658269109871085e-06,
-      "loss": 0.867,
+      "learning_rate": 1.3912045687281794e-06,
+      "loss": 0.7507,
       "step": 29432
     },
     {
-      "epoch": 0.8083546181099118,
+      "epoch": 0.8352156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.865309528989886e-06,
-      "loss": 0.833,
+      "learning_rate": 1.3907369700635664e-06,
+      "loss": 0.7459,
       "step": 29433
     },
     {
-      "epoch": 0.8083820823377551,
+      "epoch": 0.835244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.8647922113574813e-06,
-      "loss": 0.8101,
+      "learning_rate": 1.3902694441217845e-06,
+      "loss": 0.78,
       "step": 29434
     },
     {
-      "epoch": 0.8084095465655983,
+      "epoch": 0.835272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.8642749580939934e-06,
-      "loss": 0.786,
+      "learning_rate": 1.389801990906784e-06,
+      "loss": 0.9526,
       "step": 29435
     },
     {
-      "epoch": 0.8084370107934415,
+      "epoch": 0.8353007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.8637577692035124e-06,
-      "loss": 0.8167,
+      "learning_rate": 1.3893346104225135e-06,
+      "loss": 0.8059,
       "step": 29436
     },
     {
-      "epoch": 0.8084644750212848,
+      "epoch": 0.8353291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.863240644690133e-06,
-      "loss": 0.8613,
+      "learning_rate": 1.3888673026729227e-06,
+      "loss": 0.9319,
       "step": 29437
     },
     {
-      "epoch": 0.808491939249128,
+      "epoch": 0.8353575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 1.8627235845579471e-06,
-      "loss": 0.8089,
+      "learning_rate": 1.3884000676619546e-06,
+      "loss": 0.74,
       "step": 29438
     },
     {
-      "epoch": 0.8085194034769713,
+      "epoch": 0.8353859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 1.8622065888110418e-06,
-      "loss": 0.7809,
+      "learning_rate": 1.3879329053935575e-06,
+      "loss": 0.8033,
       "step": 29439
     },
     {
-      "epoch": 0.8085468677048144,
+      "epoch": 0.8354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.8616896574535126e-06,
-      "loss": 0.7769,
+      "learning_rate": 1.3874658158716814e-06,
+      "loss": 0.8737,
       "step": 29440
     },
     {
-      "epoch": 0.8085743319326577,
+      "epoch": 0.8354426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 1.861172790489445e-06,
-      "loss": 0.8104,
+      "learning_rate": 1.3869987991002665e-06,
+      "loss": 0.819,
       "step": 29441
     },
     {
-      "epoch": 0.808601796160501,
+      "epoch": 0.8354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.8606559879229313e-06,
-      "loss": 0.7956,
+      "learning_rate": 1.3865318550832595e-06,
+      "loss": 0.8174,
       "step": 29442
     },
     {
-      "epoch": 0.8086292603883442,
+      "epoch": 0.8354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.8601392497580595e-06,
-      "loss": 0.8064,
+      "learning_rate": 1.3860649838246076e-06,
+      "loss": 0.8059,
       "step": 29443
     },
     {
-      "epoch": 0.8086567246161874,
+      "epoch": 0.835527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.8596225759989216e-06,
-      "loss": 0.7858,
+      "learning_rate": 1.3855981853282496e-06,
+      "loss": 0.7828,
       "step": 29444
     },
     {
-      "epoch": 0.8086841888440307,
+      "epoch": 0.8355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.8591059666496014e-06,
-      "loss": 0.7326,
+      "learning_rate": 1.3851314595981314e-06,
+      "loss": 0.8954,
       "step": 29445
     },
     {
-      "epoch": 0.8087116530718739,
+      "epoch": 0.8355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.8585894217141876e-06,
-      "loss": 0.8308,
+      "learning_rate": 1.3846648066381962e-06,
+      "loss": 0.7912,
       "step": 29446
     },
     {
-      "epoch": 0.8087391172997171,
+      "epoch": 0.8356129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 1.858072941196769e-06,
-      "loss": 0.8901,
+      "learning_rate": 1.3841982264523835e-06,
+      "loss": 0.8051,
       "step": 29447
     },
     {
-      "epoch": 0.8087665815275603,
+      "epoch": 0.8356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 1.8575565251014328e-06,
-      "loss": 0.9275,
+      "learning_rate": 1.3837317190446354e-06,
+      "loss": 0.817,
       "step": 29448
     },
     {
-      "epoch": 0.8087940457554036,
+      "epoch": 0.8356696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.8570401734322607e-06,
-      "loss": 0.8496,
+      "learning_rate": 1.3832652844188932e-06,
+      "loss": 0.7742,
       "step": 29449
     },
     {
-      "epoch": 0.8088215099832469,
+      "epoch": 0.8356980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 1.8565238861933443e-06,
-      "loss": 0.8022,
+      "learning_rate": 1.3827989225790961e-06,
+      "loss": 0.7657,
       "step": 29450
     },
     {
-      "epoch": 0.80884897421109,
+      "epoch": 0.8357264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.8560076633887626e-06,
-      "loss": 0.8944,
+      "learning_rate": 1.3823326335291853e-06,
+      "loss": 0.8032,
       "step": 29451
     },
     {
-      "epoch": 0.8088764384389333,
+      "epoch": 0.8357548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.855491505022603e-06,
-      "loss": 0.8357,
+      "learning_rate": 1.3818664172730967e-06,
+      "loss": 0.8842,
       "step": 29452
     },
     {
-      "epoch": 0.8089039026667765,
+      "epoch": 0.835783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.8549754110989492e-06,
-      "loss": 0.8295,
+      "learning_rate": 1.3814002738147704e-06,
+      "loss": 0.8408,
       "step": 29453
     },
     {
-      "epoch": 0.8089313668946198,
+      "epoch": 0.8358115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.8544593816218847e-06,
-      "loss": 0.7257,
+      "learning_rate": 1.3809342031581451e-06,
+      "loss": 0.8175,
       "step": 29454
     },
     {
-      "epoch": 0.808958831122463,
+      "epoch": 0.8358399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 1.8539434165954962e-06,
-      "loss": 0.8865,
+      "learning_rate": 1.3804682053071538e-06,
+      "loss": 0.8709,
       "step": 29455
     },
     {
-      "epoch": 0.8089862953503062,
+      "epoch": 0.8358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8534275160238603e-06,
-      "loss": 0.8199,
+      "learning_rate": 1.3800022802657343e-06,
+      "loss": 0.7276,
       "step": 29456
     },
     {
-      "epoch": 0.8090137595781495,
+      "epoch": 0.8358967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.8529116799110635e-06,
-      "loss": 0.7998,
+      "learning_rate": 1.3795364280378265e-06,
+      "loss": 0.8688,
       "step": 29457
     },
     {
-      "epoch": 0.8090412238059927,
+      "epoch": 0.8359250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 1.852395908261183e-06,
-      "loss": 0.9227,
+      "learning_rate": 1.3790706486273587e-06,
+      "loss": 0.6719,
       "step": 29458
     },
     {
-      "epoch": 0.8090686880338359,
+      "epoch": 0.8359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.8518802010783032e-06,
-      "loss": 0.7759,
+      "learning_rate": 1.3786049420382697e-06,
+      "loss": 0.7971,
       "step": 29459
     },
     {
-      "epoch": 0.8090961522616792,
+      "epoch": 0.8359818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.8513645583665018e-06,
-      "loss": 0.8155,
+      "learning_rate": 1.3781393082744942e-06,
+      "loss": 0.8923,
       "step": 29460
     },
     {
-      "epoch": 0.8091236164895224,
+      "epoch": 0.8360102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 1.8508489801298624e-06,
-      "loss": 0.8113,
+      "learning_rate": 1.3776737473399615e-06,
+      "loss": 0.9541,
       "step": 29461
     },
     {
-      "epoch": 0.8091510807173656,
+      "epoch": 0.8360385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 1.8503334663724636e-06,
-      "loss": 0.7621,
+      "learning_rate": 1.377208259238606e-06,
+      "loss": 0.7718,
       "step": 29462
     },
     {
-      "epoch": 0.8091785449452089,
+      "epoch": 0.836066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.8498180170983815e-06,
-      "loss": 0.696,
+      "learning_rate": 1.3767428439743612e-06,
+      "loss": 0.7,
       "step": 29463
     },
     {
-      "epoch": 0.8092060091730521,
+      "epoch": 0.8360953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.849302632311698e-06,
-      "loss": 0.666,
+      "learning_rate": 1.3762775015511543e-06,
+      "loss": 0.803,
       "step": 29464
     },
     {
-      "epoch": 0.8092334734008954,
+      "epoch": 0.8361237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 1.8487873120164878e-06,
-      "loss": 0.7619,
+      "learning_rate": 1.3758122319729228e-06,
+      "loss": 0.7658,
       "step": 29465
     },
     {
-      "epoch": 0.8092609376287385,
+      "epoch": 0.8361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.8482720562168288e-06,
-      "loss": 0.8358,
+      "learning_rate": 1.3753470352435904e-06,
+      "loss": 0.9274,
       "step": 29466
     },
     {
-      "epoch": 0.8092884018565818,
+      "epoch": 0.8361804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.8477568649167998e-06,
-      "loss": 0.7696,
+      "learning_rate": 1.3748819113670908e-06,
+      "loss": 0.8495,
       "step": 29467
     },
     {
-      "epoch": 0.8093158660844251,
+      "epoch": 0.8362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.8472417381204789e-06,
-      "loss": 0.8247,
+      "learning_rate": 1.3744168603473518e-06,
+      "loss": 0.7028,
       "step": 29468
     },
     {
-      "epoch": 0.8093433303122682,
+      "epoch": 0.8362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.8467266758319368e-06,
-      "loss": 0.7985,
+      "learning_rate": 1.3739518821883001e-06,
+      "loss": 0.8384,
       "step": 29469
     },
     {
-      "epoch": 0.8093707945401115,
+      "epoch": 0.8362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.846211678055252e-06,
-      "loss": 0.8559,
+      "learning_rate": 1.3734869768938652e-06,
+      "loss": 0.8146,
       "step": 29470
     },
     {
-      "epoch": 0.8093982587679548,
+      "epoch": 0.836293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.8456967447944974e-06,
-      "loss": 0.824,
+      "learning_rate": 1.3730221444679748e-06,
+      "loss": 0.8932,
       "step": 29471
     },
     {
-      "epoch": 0.809425722995798,
+      "epoch": 0.836322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.8451818760537521e-06,
-      "loss": 0.7971,
+      "learning_rate": 1.3725573849145525e-06,
+      "loss": 0.7911,
       "step": 29472
     },
     {
-      "epoch": 0.8094531872236412,
+      "epoch": 0.8363507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.8446670718370852e-06,
-      "loss": 0.7889,
+      "learning_rate": 1.372092698237525e-06,
+      "loss": 0.8901,
       "step": 29473
     },
     {
-      "epoch": 0.8094806514514844,
+      "epoch": 0.8363791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 1.8441523321485733e-06,
-      "loss": 0.8324,
+      "learning_rate": 1.3716280844408215e-06,
+      "loss": 0.7168,
       "step": 29474
     },
     {
-      "epoch": 0.8095081156793277,
+      "epoch": 0.8364074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.8436376569922843e-06,
-      "loss": 0.9426,
+      "learning_rate": 1.371163543528361e-06,
+      "loss": 0.8831,
       "step": 29475
     },
     {
-      "epoch": 0.809535579907171,
+      "epoch": 0.8364358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.8431230463722938e-06,
-      "loss": 0.917,
+      "learning_rate": 1.3706990755040695e-06,
+      "loss": 0.7919,
       "step": 29476
     },
     {
-      "epoch": 0.8095630441350141,
+      "epoch": 0.8364642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.842608500292674e-06,
-      "loss": 0.812,
+      "learning_rate": 1.3702346803718736e-06,
+      "loss": 0.7616,
       "step": 29477
     },
     {
-      "epoch": 0.8095905083628574,
+      "epoch": 0.8364926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.8420940187574954e-06,
-      "loss": 0.7523,
+      "learning_rate": 1.3697703581356903e-06,
+      "loss": 0.8207,
       "step": 29478
     },
     {
-      "epoch": 0.8096179725907006,
+      "epoch": 0.8365209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 1.841579601770831e-06,
-      "loss": 0.789,
+      "learning_rate": 1.3693061087994452e-06,
+      "loss": 0.8724,
       "step": 29479
     },
     {
-      "epoch": 0.8096454368185438,
+      "epoch": 0.8365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.8410652493367476e-06,
-      "loss": 0.8085,
+      "learning_rate": 1.3688419323670598e-06,
+      "loss": 0.8161,
       "step": 29480
     },
     {
-      "epoch": 0.8096729010463871,
+      "epoch": 0.836577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.8405509614593175e-06,
-      "loss": 0.8148,
+      "learning_rate": 1.3683778288424542e-06,
+      "loss": 0.8483,
       "step": 29481
     },
     {
-      "epoch": 0.8097003652742303,
+      "epoch": 0.836606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.8400367381426076e-06,
-      "loss": 0.8468,
+      "learning_rate": 1.367913798229551e-06,
+      "loss": 0.7208,
       "step": 29482
     },
     {
-      "epoch": 0.8097278295020736,
+      "epoch": 0.8366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.839522579390688e-06,
-      "loss": 0.838,
+      "learning_rate": 1.3674498405322668e-06,
+      "loss": 0.8343,
       "step": 29483
     },
     {
-      "epoch": 0.8097552937299168,
+      "epoch": 0.8366628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.8390084852076263e-06,
-      "loss": 0.743,
+      "learning_rate": 1.3669859557545206e-06,
+      "loss": 0.824,
       "step": 29484
     },
     {
-      "epoch": 0.80978275795776,
+      "epoch": 0.8366912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.8384944555974925e-06,
-      "loss": 0.7665,
+      "learning_rate": 1.3665221439002352e-06,
+      "loss": 0.8684,
       "step": 29485
     },
     {
-      "epoch": 0.8098102221856033,
+      "epoch": 0.8367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.8379804905643505e-06,
-      "loss": 0.7835,
+      "learning_rate": 1.3660584049733228e-06,
+      "loss": 0.811,
       "step": 29486
     },
     {
-      "epoch": 0.8098376864134464,
+      "epoch": 0.8367480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.837466590112268e-06,
-      "loss": 0.853,
+      "learning_rate": 1.3655947389777025e-06,
+      "loss": 0.8487,
       "step": 29487
     },
     {
-      "epoch": 0.8098651506412897,
+      "epoch": 0.8367763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.8369527542453148e-06,
-      "loss": 0.7838,
+      "learning_rate": 1.3651311459172944e-06,
+      "loss": 0.8077,
       "step": 29488
     },
     {
-      "epoch": 0.809892614869133,
+      "epoch": 0.8368047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 1.83643898296755e-06,
-      "loss": 0.8094,
+      "learning_rate": 1.36466762579601e-06,
+      "loss": 0.831,
       "step": 29489
     },
     {
-      "epoch": 0.8099200790969762,
+      "epoch": 0.83683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.8359252762830438e-06,
-      "loss": 0.7624,
+      "learning_rate": 1.3642041786177662e-06,
+      "loss": 0.8322,
       "step": 29490
     },
     {
-      "epoch": 0.8099475433248194,
+      "epoch": 0.8368615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.8354116341958606e-06,
-      "loss": 0.8228,
+      "learning_rate": 1.3637408043864797e-06,
+      "loss": 0.7492,
       "step": 29491
     },
     {
-      "epoch": 0.8099750075526626,
+      "epoch": 0.8368898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.8348980567100605e-06,
-      "loss": 0.8568,
+      "learning_rate": 1.3632775031060608e-06,
+      "loss": 0.8327,
       "step": 29492
     },
     {
-      "epoch": 0.8100024717805059,
+      "epoch": 0.8369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.8343845438297104e-06,
-      "loss": 0.7824,
+      "learning_rate": 1.3628142747804252e-06,
+      "loss": 0.8443,
       "step": 29493
     },
     {
-      "epoch": 0.8100299360083492,
+      "epoch": 0.8369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.8338710955588723e-06,
-      "loss": 0.8003,
+      "learning_rate": 1.3623511194134875e-06,
+      "loss": 0.7584,
       "step": 29494
     },
     {
-      "epoch": 0.8100574002361923,
+      "epoch": 0.8369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.83335771190161e-06,
-      "loss": 0.7651,
+      "learning_rate": 1.3618880370091536e-06,
+      "loss": 0.8446,
       "step": 29495
     },
     {
-      "epoch": 0.8100848644640356,
+      "epoch": 0.8370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.8328443928619877e-06,
-      "loss": 0.8268,
+      "learning_rate": 1.3614250275713447e-06,
+      "loss": 0.7909,
       "step": 29496
     },
     {
-      "epoch": 0.8101123286918789,
+      "epoch": 0.8370317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.8323311384440635e-06,
-      "loss": 0.8103,
+      "learning_rate": 1.3609620911039644e-06,
+      "loss": 0.8398,
       "step": 29497
     },
     {
-      "epoch": 0.810139792919722,
+      "epoch": 0.8370601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 1.8318179486518972e-06,
-      "loss": 0.791,
+      "learning_rate": 1.3604992276109263e-06,
+      "loss": 0.7374,
       "step": 29498
     },
     {
-      "epoch": 0.8101672571475653,
+      "epoch": 0.8370885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 1.8313048234895503e-06,
-      "loss": 0.8638,
+      "learning_rate": 1.3600364370961417e-06,
+      "loss": 0.8499,
       "step": 29499
     },
     {
-      "epoch": 0.8101947213754085,
+      "epoch": 0.837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.830791762961085e-06,
-      "loss": 0.8993,
+      "learning_rate": 1.359573719563515e-06,
+      "loss": 0.7321,
       "step": 29500
     },
     {
-      "epoch": 0.8102221856032518,
+      "epoch": 0.8371452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 1.8302787670705591e-06,
-      "loss": 0.8444,
+      "learning_rate": 1.359111075016958e-06,
+      "loss": 0.7939,
       "step": 29501
     },
     {
-      "epoch": 0.810249649831095,
+      "epoch": 0.8371736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.8297658358220326e-06,
-      "loss": 0.8105,
+      "learning_rate": 1.3586485034603792e-06,
+      "loss": 0.8138,
       "step": 29502
     },
     {
-      "epoch": 0.8102771140589382,
+      "epoch": 0.8372020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.829252969219566e-06,
-      "loss": 0.7884,
+      "learning_rate": 1.3581860048976836e-06,
+      "loss": 0.6968,
       "step": 29503
     },
     {
-      "epoch": 0.8103045782867815,
+      "epoch": 0.8372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.8287401672672134e-06,
-      "loss": 0.7768,
+      "learning_rate": 1.3577235793327792e-06,
+      "loss": 0.8364,
       "step": 29504
     },
     {
-      "epoch": 0.8103320425146247,
+      "epoch": 0.8372587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.828227429969035e-06,
-      "loss": 0.7941,
+      "learning_rate": 1.357261226769575e-06,
+      "loss": 0.7972,
       "step": 29505
     },
     {
-      "epoch": 0.8103595067424679,
+      "epoch": 0.8372871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 1.8277147573290843e-06,
-      "loss": 0.809,
+      "learning_rate": 1.3567989472119714e-06,
+      "loss": 0.8555,
       "step": 29506
     },
     {
-      "epoch": 0.8103869709703112,
+      "epoch": 0.8373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.8272021493514203e-06,
-      "loss": 0.8401,
+      "learning_rate": 1.356336740663875e-06,
+      "loss": 0.9111,
       "step": 29507
     },
     {
-      "epoch": 0.8104144351981544,
+      "epoch": 0.8373439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.8266896060400985e-06,
+      "learning_rate": 1.3558746071291928e-06,
       "loss": 0.8558,
       "step": 29508
     },
     {
-      "epoch": 0.8104418994259976,
+      "epoch": 0.8373723041997729,
       "grad_norm": 0.0,
-      "learning_rate": 1.826177127399178e-06,
-      "loss": 0.8211,
+      "learning_rate": 1.355412546611825e-06,
+      "loss": 0.8434,
       "step": 29509
     },
     {
-      "epoch": 0.8104693636538409,
+      "epoch": 0.8374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.8256647134327065e-06,
-      "loss": 0.8557,
+      "learning_rate": 1.354950559115673e-06,
+      "loss": 0.7953,
       "step": 29510
     },
     {
-      "epoch": 0.8104968278816841,
+      "epoch": 0.8374290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.8251523641447432e-06,
-      "loss": 0.857,
+      "learning_rate": 1.354488644644647e-06,
+      "loss": 0.9151,
       "step": 29511
     },
     {
-      "epoch": 0.8105242921095274,
+      "epoch": 0.8374574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.8246400795393437e-06,
-      "loss": 0.8775,
+      "learning_rate": 1.3540268032026427e-06,
+      "loss": 0.7523,
       "step": 29512
     },
     {
-      "epoch": 0.8105517563373705,
+      "epoch": 0.8374858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 1.8241278596205558e-06,
-      "loss": 0.8181,
+      "learning_rate": 1.3535650347935636e-06,
+      "loss": 0.9322,
       "step": 29513
     },
     {
-      "epoch": 0.8105792205652138,
+      "epoch": 0.8375141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.8236157043924352e-06,
-      "loss": 0.6917,
+      "learning_rate": 1.3531033394213078e-06,
+      "loss": 0.7072,
       "step": 29514
     },
     {
-      "epoch": 0.8106066847930571,
+      "epoch": 0.8375425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.8231036138590363e-06,
-      "loss": 0.8319,
+      "learning_rate": 1.3526417170897766e-06,
+      "loss": 0.7918,
       "step": 29515
     },
     {
-      "epoch": 0.8106341490209003,
+      "epoch": 0.8375709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 1.8225915880244072e-06,
-      "loss": 0.8387,
+      "learning_rate": 1.3521801678028713e-06,
+      "loss": 0.7924,
       "step": 29516
     },
     {
-      "epoch": 0.8106616132487435,
+      "epoch": 0.8375993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.822079626892601e-06,
-      "loss": 0.7376,
+      "learning_rate": 1.3517186915644876e-06,
+      "loss": 0.792,
       "step": 29517
     },
     {
-      "epoch": 0.8106890774765867,
+      "epoch": 0.837627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8215677304676692e-06,
-      "loss": 0.878,
+      "learning_rate": 1.3512572883785235e-06,
+      "loss": 0.819,
       "step": 29518
     },
     {
-      "epoch": 0.81071654170443,
+      "epoch": 0.8376560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.8210558987536608e-06,
-      "loss": 0.8507,
+      "learning_rate": 1.3507959582488816e-06,
+      "loss": 0.8338,
       "step": 29519
     },
     {
-      "epoch": 0.8107440059322732,
+      "epoch": 0.8376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.8205441317546291e-06,
-      "loss": 0.9305,
+      "learning_rate": 1.3503347011794522e-06,
+      "loss": 0.7605,
       "step": 29520
     },
     {
-      "epoch": 0.8107714701601164,
+      "epoch": 0.8377128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.8200324294746209e-06,
-      "loss": 0.8615,
+      "learning_rate": 1.349873517174134e-06,
+      "loss": 0.874,
       "step": 29521
     },
     {
-      "epoch": 0.8107989343879597,
+      "epoch": 0.8377412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.8195207919176817e-06,
-      "loss": 0.9132,
+      "learning_rate": 1.3494124062368264e-06,
+      "loss": 0.8828,
       "step": 29522
     },
     {
-      "epoch": 0.810826398615803,
+      "epoch": 0.8377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.8190092190878627e-06,
-      "loss": 0.802,
+      "learning_rate": 1.3489513683714183e-06,
+      "loss": 0.7993,
       "step": 29523
     },
     {
-      "epoch": 0.8108538628436461,
+      "epoch": 0.8377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.818497710989212e-06,
-      "loss": 0.8586,
+      "learning_rate": 1.3484904035818059e-06,
+      "loss": 0.7822,
       "step": 29524
     },
     {
-      "epoch": 0.8108813270714894,
+      "epoch": 0.8378263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 1.8179862676257766e-06,
-      "loss": 0.783,
+      "learning_rate": 1.3480295118718877e-06,
+      "loss": 0.8159,
       "step": 29525
     },
     {
-      "epoch": 0.8109087912993326,
+      "epoch": 0.8378547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.8174748890016037e-06,
-      "loss": 0.8337,
+      "learning_rate": 1.347568693245548e-06,
+      "loss": 0.8138,
       "step": 29526
     },
     {
-      "epoch": 0.8109362555271759,
+      "epoch": 0.8378830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.8169635751207414e-06,
-      "loss": 0.8403,
+      "learning_rate": 1.3471079477066863e-06,
+      "loss": 0.8224,
       "step": 29527
     },
     {
-      "epoch": 0.8109637197550191,
+      "epoch": 0.8379114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 1.8164523259872324e-06,
-      "loss": 0.7699,
+      "learning_rate": 1.3466472752591953e-06,
+      "loss": 0.876,
       "step": 29528
     },
     {
-      "epoch": 0.8109911839828623,
+      "epoch": 0.8379398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.8159411416051209e-06,
-      "loss": 0.8691,
+      "learning_rate": 1.346186675906962e-06,
+      "loss": 0.89,
       "step": 29529
     },
     {
-      "epoch": 0.8110186482107056,
+      "epoch": 0.8379682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 1.8154300219784526e-06,
-      "loss": 0.8167,
+      "learning_rate": 1.3457261496538787e-06,
+      "loss": 0.7398,
       "step": 29530
     },
     {
-      "epoch": 0.8110461124385487,
+      "epoch": 0.8379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.8149189671112733e-06,
-      "loss": 0.7698,
+      "learning_rate": 1.345265696503838e-06,
+      "loss": 0.804,
       "step": 29531
     },
     {
-      "epoch": 0.811073576666392,
+      "epoch": 0.8380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.814407977007625e-06,
-      "loss": 0.8044,
+      "learning_rate": 1.3448053164607243e-06,
+      "loss": 0.7138,
       "step": 29532
     },
     {
-      "epoch": 0.8111010408942353,
+      "epoch": 0.8380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.8138970516715547e-06,
-      "loss": 0.8507,
+      "learning_rate": 1.3443450095284294e-06,
+      "loss": 0.8469,
       "step": 29533
     },
     {
-      "epoch": 0.8111285051220785,
+      "epoch": 0.8380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.8133861911070993e-06,
-      "loss": 0.864,
+      "learning_rate": 1.343884775710843e-06,
+      "loss": 0.7706,
       "step": 29534
     },
     {
-      "epoch": 0.8111559693499217,
+      "epoch": 0.8381101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.8128753953183032e-06,
-      "loss": 0.8518,
+      "learning_rate": 1.343424615011849e-06,
+      "loss": 0.8917,
       "step": 29535
     },
     {
-      "epoch": 0.811183433577765,
+      "epoch": 0.838138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.8123646643092108e-06,
-      "loss": 0.8243,
+      "learning_rate": 1.3429645274353375e-06,
+      "loss": 0.8184,
       "step": 29536
     },
     {
-      "epoch": 0.8112108978056082,
+      "epoch": 0.83816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.811853998083858e-06,
-      "loss": 0.8362,
+      "learning_rate": 1.3425045129851954e-06,
+      "loss": 0.8616,
       "step": 29537
     },
     {
-      "epoch": 0.8112383620334515,
+      "epoch": 0.8381952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.811343396646289e-06,
-      "loss": 0.845,
+      "learning_rate": 1.3420445716653042e-06,
+      "loss": 0.911,
       "step": 29538
     },
     {
-      "epoch": 0.8112658262612946,
+      "epoch": 0.8382236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 1.8108328600005454e-06,
-      "loss": 0.8966,
+      "learning_rate": 1.3415847034795537e-06,
+      "loss": 0.7003,
       "step": 29539
     },
     {
-      "epoch": 0.8112932904891379,
+      "epoch": 0.8382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 1.8103223881506615e-06,
-      "loss": 0.7447,
+      "learning_rate": 1.3411249084318245e-06,
+      "loss": 0.9179,
       "step": 29540
     },
     {
-      "epoch": 0.8113207547169812,
+      "epoch": 0.8382803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.8098119811006787e-06,
-      "loss": 0.851,
+      "learning_rate": 1.3406651865260012e-06,
+      "loss": 0.8384,
       "step": 29541
     },
     {
-      "epoch": 0.8113482189448243,
+      "epoch": 0.8383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.809301638854637e-06,
-      "loss": 0.7609,
+      "learning_rate": 1.3402055377659716e-06,
+      "loss": 0.8087,
       "step": 29542
     },
     {
-      "epoch": 0.8113756831726676,
+      "epoch": 0.8383371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.808791361416573e-06,
-      "loss": 0.8591,
+      "learning_rate": 1.339745962155613e-06,
+      "loss": 0.7833,
       "step": 29543
     },
     {
-      "epoch": 0.8114031474005108,
+      "epoch": 0.8383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.808281148790527e-06,
-      "loss": 0.8798,
+      "learning_rate": 1.3392864596988097e-06,
+      "loss": 0.7984,
       "step": 29544
     },
     {
-      "epoch": 0.8114306116283541,
+      "epoch": 0.8383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.807771000980535e-06,
-      "loss": 0.8493,
+      "learning_rate": 1.3388270303994455e-06,
+      "loss": 0.7957,
       "step": 29545
     },
     {
-      "epoch": 0.8114580758561973,
+      "epoch": 0.838422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.8072609179906286e-06,
-      "loss": 0.7899,
+      "learning_rate": 1.3383676742613971e-06,
+      "loss": 0.8737,
       "step": 29546
     },
     {
-      "epoch": 0.8114855400840405,
+      "epoch": 0.8384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.8067508998248473e-06,
-      "loss": 0.8327,
+      "learning_rate": 1.337908391288545e-06,
+      "loss": 0.8606,
       "step": 29547
     },
     {
-      "epoch": 0.8115130043118838,
+      "epoch": 0.8384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.8062409464872278e-06,
-      "loss": 0.784,
+      "learning_rate": 1.3374491814847734e-06,
+      "loss": 0.7307,
       "step": 29548
     },
     {
-      "epoch": 0.8115404685397271,
+      "epoch": 0.8385073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 1.805731057981802e-06,
-      "loss": 0.8234,
+      "learning_rate": 1.3369900448539553e-06,
+      "loss": 0.8079,
       "step": 29549
     },
     {
-      "epoch": 0.8115679327675702,
+      "epoch": 0.8385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.8052212343126107e-06,
-      "loss": 0.7566,
+      "learning_rate": 1.3365309813999716e-06,
+      "loss": 0.7138,
       "step": 29550
     },
     {
-      "epoch": 0.8115953969954135,
+      "epoch": 0.8385641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.8047114754836792e-06,
-      "loss": 0.8337,
+      "learning_rate": 1.3360719911267027e-06,
+      "loss": 0.8384,
       "step": 29551
     },
     {
-      "epoch": 0.8116228612232567,
+      "epoch": 0.8385925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 1.8042017814990488e-06,
-      "loss": 0.8457,
+      "learning_rate": 1.3356130740380203e-06,
+      "loss": 0.83,
       "step": 29552
     },
     {
-      "epoch": 0.8116503254510999,
+      "epoch": 0.8386208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.803692152362746e-06,
-      "loss": 0.8095,
+      "learning_rate": 1.335154230137804e-06,
+      "loss": 0.7746,
       "step": 29553
     },
     {
-      "epoch": 0.8116777896789432,
+      "epoch": 0.8386492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 1.8031825880788044e-06,
-      "loss": 0.8131,
+      "learning_rate": 1.3346954594299322e-06,
+      "loss": 0.7807,
       "step": 29554
     },
     {
-      "epoch": 0.8117052539067864,
+      "epoch": 0.838677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.802673088651259e-06,
-      "loss": 0.8198,
+      "learning_rate": 1.3342367619182751e-06,
+      "loss": 0.74,
       "step": 29555
     },
     {
-      "epoch": 0.8117327181346297,
+      "epoch": 0.8387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.8021636540841415e-06,
-      "loss": 0.7566,
+      "learning_rate": 1.3337781376067094e-06,
+      "loss": 0.7594,
       "step": 29556
     },
     {
-      "epoch": 0.8117601823624728,
+      "epoch": 0.8387343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 1.8016542843814776e-06,
-      "loss": 0.8497,
+      "learning_rate": 1.3333195864991088e-06,
+      "loss": 0.8842,
       "step": 29557
     },
     {
-      "epoch": 0.8117876465903161,
+      "epoch": 0.8387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.8011449795473012e-06,
-      "loss": 0.7692,
+      "learning_rate": 1.3328611085993482e-06,
+      "loss": 0.9163,
       "step": 29558
     },
     {
-      "epoch": 0.8118151108181594,
+      "epoch": 0.8387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.8006357395856423e-06,
-      "loss": 0.7352,
+      "learning_rate": 1.3324027039113008e-06,
+      "loss": 0.7342,
       "step": 29559
     },
     {
-      "epoch": 0.8118425750460025,
+      "epoch": 0.8388195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 1.8001265645005306e-06,
-      "loss": 0.813,
+      "learning_rate": 1.3319443724388358e-06,
+      "loss": 0.7413,
       "step": 29560
     },
     {
-      "epoch": 0.8118700392738458,
+      "epoch": 0.8388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.7996174542959922e-06,
-      "loss": 0.7716,
+      "learning_rate": 1.331486114185826e-06,
+      "loss": 0.8178,
       "step": 29561
     },
     {
-      "epoch": 0.8118975035016891,
+      "epoch": 0.8388762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.7991084089760603e-06,
-      "loss": 0.8473,
+      "learning_rate": 1.3310279291561446e-06,
+      "loss": 0.8462,
       "step": 29562
     },
     {
-      "epoch": 0.8119249677295323,
+      "epoch": 0.8389046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 1.7985994285447562e-06,
-      "loss": 0.7937,
+      "learning_rate": 1.3305698173536586e-06,
+      "loss": 0.8107,
       "step": 29563
     },
     {
-      "epoch": 0.8119524319573755,
+      "epoch": 0.8389330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 1.7980905130061098e-06,
-      "loss": 0.8074,
+      "learning_rate": 1.3301117787822382e-06,
+      "loss": 0.7722,
       "step": 29564
     },
     {
-      "epoch": 0.8119798961852187,
+      "epoch": 0.838961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.7975816623641485e-06,
-      "loss": 0.8801,
+      "learning_rate": 1.329653813445756e-06,
+      "loss": 0.7845,
       "step": 29565
     },
     {
-      "epoch": 0.812007360413062,
+      "epoch": 0.8389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.7970728766228985e-06,
-      "loss": 0.8618,
+      "learning_rate": 1.3291959213480753e-06,
+      "loss": 0.8503,
       "step": 29566
     },
     {
-      "epoch": 0.8120348246409053,
+      "epoch": 0.8390181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 1.796564155786389e-06,
-      "loss": 0.8247,
+      "learning_rate": 1.3287381024930668e-06,
+      "loss": 0.8511,
       "step": 29567
     },
     {
-      "epoch": 0.8120622888687484,
+      "epoch": 0.8390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.7960554998586376e-06,
-      "loss": 0.9074,
+      "learning_rate": 1.3282803568845992e-06,
+      "loss": 0.6626,
       "step": 29568
     },
     {
-      "epoch": 0.8120897530965917,
+      "epoch": 0.8390749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.795546908843676e-06,
-      "loss": 0.8771,
+      "learning_rate": 1.327822684526535e-06,
+      "loss": 0.7807,
       "step": 29569
     },
     {
-      "epoch": 0.8121172173244349,
+      "epoch": 0.8391032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7950383827455231e-06,
-      "loss": 0.8114,
+      "learning_rate": 1.3273650854227438e-06,
+      "loss": 0.8049,
       "step": 29570
     },
     {
-      "epoch": 0.8121446815522781,
+      "epoch": 0.8391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.7945299215682045e-06,
-      "loss": 0.7044,
+      "learning_rate": 1.3269075595770896e-06,
+      "loss": 0.7373,
       "step": 29571
     },
     {
-      "epoch": 0.8121721457801214,
+      "epoch": 0.8391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.7940215253157444e-06,
-      "loss": 0.7852,
+      "learning_rate": 1.3264501069934343e-06,
+      "loss": 0.8166,
       "step": 29572
     },
     {
-      "epoch": 0.8121996100079646,
+      "epoch": 0.839188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.7935131939921635e-06,
-      "loss": 0.8568,
+      "learning_rate": 1.325992727675649e-06,
+      "loss": 0.7581,
       "step": 29573
     },
     {
-      "epoch": 0.8122270742358079,
+      "epoch": 0.839216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.7930049276014883e-06,
-      "loss": 0.8478,
+      "learning_rate": 1.3255354216275905e-06,
+      "loss": 0.7517,
       "step": 29574
     },
     {
-      "epoch": 0.8122545384636511,
+      "epoch": 0.8392451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.7924967261477344e-06,
-      "loss": 0.9055,
+      "learning_rate": 1.3250781888531239e-06,
+      "loss": 0.7478,
       "step": 29575
     },
     {
-      "epoch": 0.8122820026914943,
+      "epoch": 0.8392735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 1.7919885896349287e-06,
-      "loss": 0.8181,
+      "learning_rate": 1.3246210293561146e-06,
+      "loss": 0.8073,
       "step": 29576
     },
     {
-      "epoch": 0.8123094669193376,
+      "epoch": 0.8393019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.7914805180670858e-06,
-      "loss": 0.8267,
+      "learning_rate": 1.3241639431404196e-06,
+      "loss": 0.776,
       "step": 29577
     },
     {
-      "epoch": 0.8123369311471808,
+      "epoch": 0.8393303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.790972511448229e-06,
-      "loss": 0.8292,
+      "learning_rate": 1.3237069302099014e-06,
+      "loss": 0.8972,
       "step": 29578
     },
     {
-      "epoch": 0.812364395375024,
+      "epoch": 0.8393586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 1.7904645697823774e-06,
-      "loss": 0.8481,
+      "learning_rate": 1.3232499905684226e-06,
+      "loss": 0.7756,
       "step": 29579
     },
     {
-      "epoch": 0.8123918596028673,
+      "epoch": 0.8393870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.7899566930735535e-06,
-      "loss": 0.8207,
+      "learning_rate": 1.3227931242198388e-06,
+      "loss": 0.8732,
       "step": 29580
     },
     {
-      "epoch": 0.8124193238307105,
+      "epoch": 0.8394154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 1.7894488813257682e-06,
-      "loss": 0.8573,
+      "learning_rate": 1.3223363311680126e-06,
+      "loss": 0.8323,
       "step": 29581
     },
     {
-      "epoch": 0.8124467880585537,
+      "epoch": 0.8394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.7889411345430452e-06,
-      "loss": 0.8228,
+      "learning_rate": 1.3218796114168031e-06,
+      "loss": 0.7079,
       "step": 29582
     },
     {
-      "epoch": 0.8124742522863969,
+      "epoch": 0.839472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.7884334527294e-06,
-      "loss": 0.8193,
+      "learning_rate": 1.3214229649700639e-06,
+      "loss": 0.7099,
       "step": 29583
     },
     {
-      "epoch": 0.8125017165142402,
+      "epoch": 0.83950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.787925835888853e-06,
-      "loss": 0.8408,
+      "learning_rate": 1.3209663918316562e-06,
+      "loss": 0.7437,
       "step": 29584
     },
     {
-      "epoch": 0.8125291807420835,
+      "epoch": 0.8395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.787418284025415e-06,
-      "loss": 0.8184,
+      "learning_rate": 1.320509892005437e-06,
+      "loss": 0.7855,
       "step": 29585
     },
     {
-      "epoch": 0.8125566449699266,
+      "epoch": 0.8395573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.7869107971431076e-06,
-      "loss": 0.8532,
+      "learning_rate": 1.320053465495259e-06,
+      "loss": 0.7776,
       "step": 29586
     },
     {
-      "epoch": 0.8125841091977699,
+      "epoch": 0.8395856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 1.786403375245942e-06,
-      "loss": 0.8927,
+      "learning_rate": 1.3195971123049788e-06,
+      "loss": 0.8581,
       "step": 29587
     },
     {
-      "epoch": 0.8126115734256132,
+      "epoch": 0.8396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 1.7858960183379326e-06,
-      "loss": 0.7521,
+      "learning_rate": 1.3191408324384525e-06,
+      "loss": 0.8259,
       "step": 29588
     },
     {
-      "epoch": 0.8126390376534564,
+      "epoch": 0.8396424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.7853887264230973e-06,
-      "loss": 0.8491,
+      "learning_rate": 1.3186846258995346e-06,
+      "loss": 0.8102,
       "step": 29589
     },
     {
-      "epoch": 0.8126665018812996,
+      "epoch": 0.8396708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 1.7848814995054476e-06,
-      "loss": 0.7726,
+      "learning_rate": 1.3182284926920797e-06,
+      "loss": 0.7161,
       "step": 29590
     },
     {
-      "epoch": 0.8126939661091428,
+      "epoch": 0.8396992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 1.7843743375889998e-06,
-      "loss": 0.7876,
+      "learning_rate": 1.317772432819936e-06,
+      "loss": 0.9015,
       "step": 29591
     },
     {
-      "epoch": 0.8127214303369861,
+      "epoch": 0.839727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.7838672406777625e-06,
-      "loss": 0.8332,
+      "learning_rate": 1.3173164462869591e-06,
+      "loss": 0.7343,
       "step": 29592
     },
     {
-      "epoch": 0.8127488945648293,
+      "epoch": 0.8397559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.7833602087757528e-06,
-      "loss": 0.7769,
+      "learning_rate": 1.316860533097003e-06,
+      "loss": 0.8127,
       "step": 29593
     },
     {
-      "epoch": 0.8127763587926725,
+      "epoch": 0.8397843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.7828532418869782e-06,
-      "loss": 0.7777,
+      "learning_rate": 1.3164046932539142e-06,
+      "loss": 0.7197,
       "step": 29594
     },
     {
-      "epoch": 0.8128038230205158,
+      "epoch": 0.8398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 1.7823463400154505e-06,
-      "loss": 0.7935,
+      "learning_rate": 1.3159489267615455e-06,
+      "loss": 0.89,
       "step": 29595
     },
     {
-      "epoch": 0.812831287248359,
+      "epoch": 0.8398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.7818395031651825e-06,
-      "loss": 0.8375,
+      "learning_rate": 1.3154932336237481e-06,
+      "loss": 0.8116,
       "step": 29596
     },
     {
-      "epoch": 0.8128587514762022,
+      "epoch": 0.8398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.781332731340184e-06,
-      "loss": 0.7506,
+      "learning_rate": 1.3150376138443676e-06,
+      "loss": 0.737,
       "step": 29597
     },
     {
-      "epoch": 0.8128862157040455,
+      "epoch": 0.8398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 1.7808260245444663e-06,
-      "loss": 0.7729,
+      "learning_rate": 1.3145820674272558e-06,
+      "loss": 0.823,
       "step": 29598
     },
     {
-      "epoch": 0.8129136799318887,
+      "epoch": 0.8399262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.7803193827820354e-06,
-      "loss": 0.8297,
+      "learning_rate": 1.3141265943762615e-06,
+      "loss": 0.8521,
       "step": 29599
     },
     {
-      "epoch": 0.812941144159732,
+      "epoch": 0.8399545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 1.779812806056903e-06,
-      "loss": 0.7652,
+      "learning_rate": 1.3136711946952274e-06,
+      "loss": 0.8555,
       "step": 29600
     },
     {
-      "epoch": 0.8129686083875752,
+      "epoch": 0.8399829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 1.7793062943730744e-06,
-      "loss": 0.7305,
+      "learning_rate": 1.3132158683880037e-06,
+      "loss": 0.8014,
       "step": 29601
     },
     {
-      "epoch": 0.8129960726154184,
+      "epoch": 0.840011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.778799847734558e-06,
-      "loss": 0.8075,
+      "learning_rate": 1.312760615458436e-06,
+      "loss": 0.8889,
       "step": 29602
     },
     {
-      "epoch": 0.8130235368432617,
+      "epoch": 0.8400397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 1.7782934661453644e-06,
-      "loss": 0.8616,
+      "learning_rate": 1.3123054359103705e-06,
+      "loss": 0.8307,
       "step": 29603
     },
     {
-      "epoch": 0.8130510010711048,
+      "epoch": 0.8400681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.7777871496094944e-06,
-      "loss": 0.7653,
+      "learning_rate": 1.3118503297476525e-06,
+      "loss": 0.7955,
       "step": 29604
     },
     {
-      "epoch": 0.8130784652989481,
+      "epoch": 0.8400964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 1.7772808981309586e-06,
-      "loss": 0.8039,
+      "learning_rate": 1.311395296974124e-06,
+      "loss": 0.8045,
       "step": 29605
     },
     {
-      "epoch": 0.8131059295267914,
+      "epoch": 0.8401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 1.7767747117137613e-06,
-      "loss": 0.744,
+      "learning_rate": 1.3109403375936302e-06,
+      "loss": 0.8697,
       "step": 29606
     },
     {
-      "epoch": 0.8131333937546346,
+      "epoch": 0.8401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.7762685903619082e-06,
-      "loss": 0.7649,
+      "learning_rate": 1.3104854516100152e-06,
+      "loss": 0.8871,
       "step": 29607
     },
     {
-      "epoch": 0.8131608579824778,
+      "epoch": 0.8401816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 1.775762534079406e-06,
-      "loss": 0.8346,
+      "learning_rate": 1.3100306390271188e-06,
+      "loss": 0.9156,
       "step": 29608
     },
     {
-      "epoch": 0.813188322210321,
+      "epoch": 0.8402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.775256542870255e-06,
-      "loss": 0.7608,
+      "learning_rate": 1.309575899848784e-06,
+      "loss": 0.8091,
       "step": 29609
     },
     {
-      "epoch": 0.8132157864381643,
+      "epoch": 0.8402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.7747506167384598e-06,
-      "loss": 0.7094,
+      "learning_rate": 1.309121234078854e-06,
+      "loss": 0.72,
       "step": 29610
     },
     {
-      "epoch": 0.8132432506660076,
+      "epoch": 0.8402667423382519,
       "grad_norm": 0.0,
-      "learning_rate": 1.7742447556880228e-06,
-      "loss": 0.8272,
+      "learning_rate": 1.3086666417211658e-06,
+      "loss": 0.8155,
       "step": 29611
     },
     {
-      "epoch": 0.8132707148938507,
+      "epoch": 0.8402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.773738959722947e-06,
-      "loss": 0.7518,
+      "learning_rate": 1.3082121227795619e-06,
+      "loss": 0.8987,
       "step": 29612
     },
     {
-      "epoch": 0.813298179121694,
+      "epoch": 0.8403234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.7732332288472365e-06,
-      "loss": 0.7588,
+      "learning_rate": 1.3077576772578827e-06,
+      "loss": 0.7992,
       "step": 29613
     },
     {
-      "epoch": 0.8133256433495373,
+      "epoch": 0.8403518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.7727275630648898e-06,
-      "loss": 0.9491,
+      "learning_rate": 1.3073033051599626e-06,
+      "loss": 0.6708,
       "step": 29614
     },
     {
-      "epoch": 0.8133531075773804,
+      "epoch": 0.8403802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 1.7722219623799131e-06,
-      "loss": 0.811,
+      "learning_rate": 1.3068490064896422e-06,
+      "loss": 0.825,
       "step": 29615
     },
     {
-      "epoch": 0.8133805718052237,
+      "epoch": 0.8404086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.7717164267963005e-06,
-      "loss": 0.8341,
+      "learning_rate": 1.3063947812507615e-06,
+      "loss": 0.7653,
       "step": 29616
     },
     {
-      "epoch": 0.8134080360330669,
+      "epoch": 0.8404370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.771210956318058e-06,
-      "loss": 0.743,
+      "learning_rate": 1.3059406294471523e-06,
+      "loss": 0.8858,
       "step": 29617
     },
     {
-      "epoch": 0.8134355002609102,
+      "epoch": 0.8404653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.7707055509491788e-06,
-      "loss": 0.8487,
+      "learning_rate": 1.3054865510826508e-06,
+      "loss": 0.7793,
       "step": 29618
     },
     {
-      "epoch": 0.8134629644887534,
+      "epoch": 0.8404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 1.7702002106936655e-06,
-      "loss": 0.8472,
+      "learning_rate": 1.3050325461611013e-06,
+      "loss": 0.8317,
       "step": 29619
     },
     {
-      "epoch": 0.8134904287165966,
+      "epoch": 0.840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.7696949355555159e-06,
-      "loss": 0.7793,
+      "learning_rate": 1.3045786146863293e-06,
+      "loss": 0.8454,
       "step": 29620
     },
     {
-      "epoch": 0.8135178929444399,
+      "epoch": 0.8405505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.7691897255387304e-06,
-      "loss": 0.8452,
+      "learning_rate": 1.3041247566621752e-06,
+      "loss": 0.7506,
       "step": 29621
     },
     {
-      "epoch": 0.8135453571722832,
+      "epoch": 0.8405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.7686845806473018e-06,
-      "loss": 0.9148,
+      "learning_rate": 1.3036709720924679e-06,
+      "loss": 0.8296,
       "step": 29622
     },
     {
-      "epoch": 0.8135728214001263,
+      "epoch": 0.8406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.7681795008852287e-06,
-      "loss": 0.8349,
+      "learning_rate": 1.3032172609810422e-06,
+      "loss": 0.7038,
       "step": 29623
     },
     {
-      "epoch": 0.8136002856279696,
+      "epoch": 0.8406356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.767674486256512e-06,
-      "loss": 0.747,
+      "learning_rate": 1.3027636233317342e-06,
+      "loss": 0.8354,
       "step": 29624
     },
     {
-      "epoch": 0.8136277498558128,
+      "epoch": 0.8406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 1.7671695367651397e-06,
-      "loss": 0.8325,
+      "learning_rate": 1.3023100591483707e-06,
+      "loss": 0.7029,
       "step": 29625
     },
     {
-      "epoch": 0.813655214083656,
+      "epoch": 0.8406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.766664652415112e-06,
-      "loss": 0.8853,
+      "learning_rate": 1.301856568434785e-06,
+      "loss": 0.7558,
       "step": 29626
     },
     {
-      "epoch": 0.8136826783114993,
+      "epoch": 0.8407207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 1.7661598332104246e-06,
-      "loss": 0.8747,
+      "learning_rate": 1.30140315119481e-06,
+      "loss": 0.8165,
       "step": 29627
     },
     {
-      "epoch": 0.8137101425393425,
+      "epoch": 0.8407491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.7656550791550685e-06,
+      "learning_rate": 1.3009498074322712e-06,
       "loss": 0.8291,
       "step": 29628
     },
     {
-      "epoch": 0.8137376067671858,
+      "epoch": 0.8407775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 1.7651503902530387e-06,
-      "loss": 0.9596,
+      "learning_rate": 1.3004965371510003e-06,
+      "loss": 0.838,
       "step": 29629
     },
     {
-      "epoch": 0.8137650709950289,
+      "epoch": 0.8408059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 1.7646457665083284e-06,
-      "loss": 0.7042,
+      "learning_rate": 1.3000433403548297e-06,
+      "loss": 0.7999,
       "step": 29630
     },
     {
-      "epoch": 0.8137925352228722,
+      "epoch": 0.8408342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.764141207924931e-06,
-      "loss": 0.9524,
+      "learning_rate": 1.2995902170475804e-06,
+      "loss": 0.855,
       "step": 29631
     },
     {
-      "epoch": 0.8138199994507155,
+      "epoch": 0.8408626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 1.7636367145068422e-06,
-      "loss": 0.8156,
+      "learning_rate": 1.2991371672330844e-06,
+      "loss": 0.8292,
       "step": 29632
     },
     {
-      "epoch": 0.8138474636785586,
+      "epoch": 0.8408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.7631322862580491e-06,
-      "loss": 0.8634,
+      "learning_rate": 1.2986841909151683e-06,
+      "loss": 0.8334,
       "step": 29633
     },
     {
-      "epoch": 0.8138749279064019,
+      "epoch": 0.8409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.7626279231825416e-06,
-      "loss": 0.7427,
+      "learning_rate": 1.2982312880976567e-06,
+      "loss": 0.8965,
       "step": 29634
     },
     {
-      "epoch": 0.8139023921342452,
+      "epoch": 0.8409477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 1.7621236252843133e-06,
-      "loss": 0.7459,
+      "learning_rate": 1.2977784587843778e-06,
+      "loss": 0.8661,
       "step": 29635
     },
     {
-      "epoch": 0.8139298563620884,
+      "epoch": 0.8409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.761619392567353e-06,
-      "loss": 0.8296,
+      "learning_rate": 1.2973257029791564e-06,
+      "loss": 0.7602,
       "step": 29636
     },
     {
-      "epoch": 0.8139573205899316,
+      "epoch": 0.8410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.7611152250356523e-06,
-      "loss": 0.8411,
+      "learning_rate": 1.2968730206858148e-06,
+      "loss": 0.7975,
       "step": 29637
     },
     {
-      "epoch": 0.8139847848177748,
+      "epoch": 0.841032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.7606111226931989e-06,
-      "loss": 0.7723,
+      "learning_rate": 1.296420411908178e-06,
+      "loss": 0.8016,
       "step": 29638
     },
     {
-      "epoch": 0.8140122490456181,
+      "epoch": 0.841061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.760107085543985e-06,
-      "loss": 0.775,
+      "learning_rate": 1.2959678766500706e-06,
+      "loss": 0.7955,
       "step": 29639
     },
     {
-      "epoch": 0.8140397132734614,
+      "epoch": 0.8410896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.7596031135919922e-06,
-      "loss": 0.774,
+      "learning_rate": 1.2955154149153126e-06,
+      "loss": 0.8784,
       "step": 29640
     },
     {
-      "epoch": 0.8140671775013045,
+      "epoch": 0.8411180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.759099206841215e-06,
-      "loss": 0.8433,
+      "learning_rate": 1.295063026707727e-06,
+      "loss": 0.7681,
       "step": 29641
     },
     {
-      "epoch": 0.8140946417291478,
+      "epoch": 0.8411464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 1.7585953652956344e-06,
-      "loss": 0.7898,
+      "learning_rate": 1.2946107120311369e-06,
+      "loss": 0.7555,
       "step": 29642
     },
     {
-      "epoch": 0.814122105956991,
+      "epoch": 0.8411748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.7580915889592387e-06,
-      "loss": 0.7598,
+      "learning_rate": 1.2941584708893596e-06,
+      "loss": 0.7713,
       "step": 29643
     },
     {
-      "epoch": 0.8141495701848342,
+      "epoch": 0.8412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.7575878778360157e-06,
-      "loss": 0.8445,
+      "learning_rate": 1.2937063032862174e-06,
+      "loss": 0.8138,
       "step": 29644
     },
     {
-      "epoch": 0.8141770344126775,
+      "epoch": 0.8412315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.757084231929953e-06,
-      "loss": 0.7294,
+      "learning_rate": 1.2932542092255308e-06,
+      "loss": 0.8692,
       "step": 29645
     },
     {
-      "epoch": 0.8142044986405207,
+      "epoch": 0.8412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 1.7565806512450302e-06,
-      "loss": 0.7599,
+      "learning_rate": 1.2928021887111164e-06,
+      "loss": 0.7622,
       "step": 29646
     },
     {
-      "epoch": 0.814231962868364,
+      "epoch": 0.8412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.7560771357852347e-06,
-      "loss": 0.7934,
+      "learning_rate": 1.2923502417467948e-06,
+      "loss": 0.8107,
       "step": 29647
     },
     {
-      "epoch": 0.8142594270962072,
+      "epoch": 0.841316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.7555736855545524e-06,
-      "loss": 0.7112,
+      "learning_rate": 1.2918983683363772e-06,
+      "loss": 0.8102,
       "step": 29648
     },
     {
-      "epoch": 0.8142868913240504,
+      "epoch": 0.8413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 1.7550703005569625e-06,
-      "loss": 0.8927,
+      "learning_rate": 1.2914465684836886e-06,
+      "loss": 0.7069,
       "step": 29649
     },
     {
-      "epoch": 0.8143143555518937,
+      "epoch": 0.8413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.7545669807964494e-06,
-      "loss": 0.7953,
+      "learning_rate": 1.2909948421925445e-06,
+      "loss": 0.8383,
       "step": 29650
     },
     {
-      "epoch": 0.8143418197797369,
+      "epoch": 0.8414018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.7540637262769999e-06,
-      "loss": 0.8558,
+      "learning_rate": 1.2905431894667552e-06,
+      "loss": 0.8071,
       "step": 29651
     },
     {
-      "epoch": 0.8143692840075801,
+      "epoch": 0.8414301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.7535605370025887e-06,
-      "loss": 0.8324,
+      "learning_rate": 1.2900916103101403e-06,
+      "loss": 0.8055,
       "step": 29652
     },
     {
-      "epoch": 0.8143967482354234,
+      "epoch": 0.8414585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.7530574129772015e-06,
-      "loss": 0.7688,
+      "learning_rate": 1.289640104726515e-06,
+      "loss": 0.8209,
       "step": 29653
     },
     {
-      "epoch": 0.8144242124632666,
+      "epoch": 0.8414869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 1.7525543542048185e-06,
-      "loss": 0.7962,
+      "learning_rate": 1.289188672719689e-06,
+      "loss": 0.7025,
       "step": 29654
     },
     {
-      "epoch": 0.8144516766911098,
+      "epoch": 0.8415153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.7520513606894207e-06,
-      "loss": 0.7426,
+      "learning_rate": 1.2887373142934767e-06,
+      "loss": 0.7956,
       "step": 29655
     },
     {
-      "epoch": 0.814479140918953,
+      "epoch": 0.8415437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 1.7515484324349895e-06,
-      "loss": 0.7764,
+      "learning_rate": 1.2882860294516952e-06,
+      "loss": 0.7915,
       "step": 29656
     },
     {
-      "epoch": 0.8145066051467963,
+      "epoch": 0.841572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.751045569445502e-06,
-      "loss": 0.7983,
+      "learning_rate": 1.2878348181981516e-06,
+      "loss": 0.865,
       "step": 29657
     },
     {
-      "epoch": 0.8145340693746396,
+      "epoch": 0.8416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.7505427717249357e-06,
-      "loss": 0.7234,
+      "learning_rate": 1.2873836805366579e-06,
+      "loss": 0.7742,
       "step": 29658
     },
     {
-      "epoch": 0.8145615336024827,
+      "epoch": 0.8416288308740069,
       "grad_norm": 0.0,
-      "learning_rate": 1.75004003927727e-06,
-      "loss": 0.8511,
+      "learning_rate": 1.2869326164710294e-06,
+      "loss": 0.83,
       "step": 29659
     },
     {
-      "epoch": 0.814588997830326,
+      "epoch": 0.8416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.7495373721064835e-06,
-      "loss": 0.7789,
+      "learning_rate": 1.2864816260050694e-06,
+      "loss": 0.8788,
       "step": 29660
     },
     {
-      "epoch": 0.8146164620581693,
+      "epoch": 0.8416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.7490347702165523e-06,
-      "loss": 0.8398,
+      "learning_rate": 1.2860307091425905e-06,
+      "loss": 0.7978,
       "step": 29661
     },
     {
-      "epoch": 0.8146439262860125,
+      "epoch": 0.8417139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 1.7485322336114574e-06,
-      "loss": 0.7867,
+      "learning_rate": 1.2855798658874052e-06,
+      "loss": 0.7449,
       "step": 29662
     },
     {
-      "epoch": 0.8146713905138557,
+      "epoch": 0.8417423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.7480297622951692e-06,
-      "loss": 0.8741,
+      "learning_rate": 1.285129096243316e-06,
+      "loss": 0.8071,
       "step": 29663
     },
     {
-      "epoch": 0.8146988547416989,
+      "epoch": 0.8417707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.7475273562716667e-06,
-      "loss": 0.7656,
+      "learning_rate": 1.2846784002141333e-06,
+      "loss": 0.8091,
       "step": 29664
     },
     {
-      "epoch": 0.8147263189695422,
+      "epoch": 0.8417990919409761,
       "grad_norm": 0.0,
-      "learning_rate": 1.7470250155449253e-06,
-      "loss": 0.7807,
+      "learning_rate": 1.2842277778036627e-06,
+      "loss": 0.8426,
       "step": 29665
     },
     {
-      "epoch": 0.8147537831973855,
+      "epoch": 0.8418274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 1.7465227401189177e-06,
-      "loss": 0.8308,
+      "learning_rate": 1.2837772290157135e-06,
+      "loss": 0.8438,
       "step": 29666
     },
     {
-      "epoch": 0.8147812474252286,
+      "epoch": 0.841855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.7460205299976186e-06,
-      "loss": 0.8662,
+      "learning_rate": 1.2833267538540907e-06,
+      "loss": 0.8638,
       "step": 29667
     },
     {
-      "epoch": 0.8148087116530719,
+      "epoch": 0.8418842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.7455183851850055e-06,
-      "loss": 0.8161,
+      "learning_rate": 1.2828763523225974e-06,
+      "loss": 0.8244,
       "step": 29668
     },
     {
-      "epoch": 0.8148361758809151,
+      "epoch": 0.8419125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.7450163056850455e-06,
-      "loss": 0.8561,
+      "learning_rate": 1.2824260244250385e-06,
+      "loss": 0.883,
       "step": 29669
     },
     {
-      "epoch": 0.8148636401087583,
+      "epoch": 0.8419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 1.7445142915017144e-06,
-      "loss": 0.8602,
+      "learning_rate": 1.2819757701652224e-06,
+      "loss": 0.7674,
       "step": 29670
     },
     {
-      "epoch": 0.8148911043366016,
+      "epoch": 0.8419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.7440123426389833e-06,
-      "loss": 0.8414,
+      "learning_rate": 1.2815255895469458e-06,
+      "loss": 0.8444,
       "step": 29671
     },
     {
-      "epoch": 0.8149185685644448,
+      "epoch": 0.8419977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.7435104591008279e-06,
-      "loss": 0.8049,
+      "learning_rate": 1.2810754825740146e-06,
+      "loss": 0.8069,
       "step": 29672
     },
     {
-      "epoch": 0.814946032792288,
+      "epoch": 0.8420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.7430086408912138e-06,
-      "loss": 0.89,
+      "learning_rate": 1.2806254492502323e-06,
+      "loss": 0.8293,
       "step": 29673
     },
     {
-      "epoch": 0.8149734970201313,
+      "epoch": 0.8420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.742506888014116e-06,
-      "loss": 0.7781,
+      "learning_rate": 1.2801754895793972e-06,
+      "loss": 0.8294,
       "step": 29674
     },
     {
-      "epoch": 0.8150009612479745,
+      "epoch": 0.8420828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.7420052004734989e-06,
-      "loss": 0.8159,
+      "learning_rate": 1.2797256035653128e-06,
+      "loss": 0.8141,
       "step": 29675
     },
     {
-      "epoch": 0.8150284254758178,
+      "epoch": 0.84211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.741503578273337e-06,
-      "loss": 0.811,
+      "learning_rate": 1.2792757912117793e-06,
+      "loss": 0.7814,
       "step": 29676
     },
     {
-      "epoch": 0.8150558897036609,
+      "epoch": 0.8421396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.7410020214175972e-06,
-      "loss": 0.8462,
+      "learning_rate": 1.2788260525225926e-06,
+      "loss": 0.8212,
       "step": 29677
     },
     {
-      "epoch": 0.8150833539315042,
+      "epoch": 0.8421679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 1.7405005299102485e-06,
-      "loss": 0.7638,
+      "learning_rate": 1.2783763875015542e-06,
+      "loss": 0.8692,
       "step": 29678
     },
     {
-      "epoch": 0.8151108181593475,
+      "epoch": 0.8421963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.7399991037552587e-06,
-      "loss": 0.87,
+      "learning_rate": 1.2779267961524633e-06,
+      "loss": 0.8667,
       "step": 29679
     },
     {
-      "epoch": 0.8151382823871907,
+      "epoch": 0.8422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.7394977429565985e-06,
-      "loss": 0.7958,
+      "learning_rate": 1.2774772784791168e-06,
+      "loss": 0.7224,
       "step": 29680
     },
     {
-      "epoch": 0.8151657466150339,
+      "epoch": 0.8422531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 1.738996447518232e-06,
-      "loss": 0.8828,
+      "learning_rate": 1.2770278344853139e-06,
+      "loss": 0.865,
       "step": 29681
     },
     {
-      "epoch": 0.8151932108428771,
+      "epoch": 0.8422814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.7384952174441227e-06,
-      "loss": 0.7901,
+      "learning_rate": 1.2765784641748469e-06,
+      "loss": 0.8582,
       "step": 29682
     },
     {
-      "epoch": 0.8152206750707204,
+      "epoch": 0.8423098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 1.73799405273824e-06,
-      "loss": 0.7527,
+      "learning_rate": 1.276129167551513e-06,
+      "loss": 0.817,
       "step": 29683
     },
     {
-      "epoch": 0.8152481392985637,
+      "epoch": 0.8423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.737492953404548e-06,
-      "loss": 0.7718,
+      "learning_rate": 1.2756799446191114e-06,
+      "loss": 0.9648,
       "step": 29684
     },
     {
-      "epoch": 0.8152756035264068,
+      "epoch": 0.842366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.736991919447013e-06,
-      "loss": 0.8275,
+      "learning_rate": 1.2752307953814313e-06,
+      "loss": 0.7846,
       "step": 29685
     },
     {
-      "epoch": 0.8153030677542501,
+      "epoch": 0.8423950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 1.7364909508696015e-06,
-      "loss": 0.8363,
+      "learning_rate": 1.2747817198422685e-06,
+      "loss": 0.9028,
       "step": 29686
     },
     {
-      "epoch": 0.8153305319820934,
+      "epoch": 0.8424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.7359900476762715e-06,
-      "loss": 0.7327,
+      "learning_rate": 1.2743327180054188e-06,
+      "loss": 0.8257,
       "step": 29687
     },
     {
-      "epoch": 0.8153579962099365,
+      "epoch": 0.8424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.735489209870992e-06,
-      "loss": 0.805,
+      "learning_rate": 1.2738837898746703e-06,
+      "loss": 0.7889,
       "step": 29688
     },
     {
-      "epoch": 0.8153854604377798,
+      "epoch": 0.8424801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 1.734988437457721e-06,
-      "loss": 0.8072,
+      "learning_rate": 1.2734349354538188e-06,
+      "loss": 0.7358,
       "step": 29689
     },
     {
-      "epoch": 0.815412924665623,
+      "epoch": 0.8425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 1.7344877304404229e-06,
-      "loss": 0.9151,
+      "learning_rate": 1.272986154746656e-06,
+      "loss": 0.8267,
       "step": 29690
     },
     {
-      "epoch": 0.8154403888934663,
+      "epoch": 0.8425368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.7339870888230592e-06,
-      "loss": 0.896,
+      "learning_rate": 1.2725374477569697e-06,
+      "loss": 0.7207,
       "step": 29691
     },
     {
-      "epoch": 0.8154678531213095,
+      "epoch": 0.8425652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.7334865126095946e-06,
-      "loss": 0.808,
+      "learning_rate": 1.2720888144885512e-06,
+      "loss": 0.7135,
       "step": 29692
     },
     {
-      "epoch": 0.8154953173491527,
+      "epoch": 0.8425936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 1.7329860018039845e-06,
-      "loss": 0.864,
+      "learning_rate": 1.2716402549451945e-06,
+      "loss": 0.7631,
       "step": 29693
     },
     {
-      "epoch": 0.815522781576996,
+      "epoch": 0.8426220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 1.7324855564101916e-06,
-      "loss": 0.8939,
+      "learning_rate": 1.2711917691306797e-06,
+      "loss": 0.7788,
       "step": 29694
     },
     {
-      "epoch": 0.8155502458048391,
+      "epoch": 0.8426503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.731985176432175e-06,
-      "loss": 0.8779,
+      "learning_rate": 1.2707433570488037e-06,
+      "loss": 0.8487,
       "step": 29695
     },
     {
-      "epoch": 0.8155777100326824,
+      "epoch": 0.8426787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.7314848618738967e-06,
-      "loss": 0.7897,
+      "learning_rate": 1.2702950187033502e-06,
+      "loss": 0.8292,
       "step": 29696
     },
     {
-      "epoch": 0.8156051742605257,
+      "epoch": 0.8427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 1.7309846127393114e-06,
-      "loss": 0.7779,
+      "learning_rate": 1.2698467540981064e-06,
+      "loss": 0.7413,
       "step": 29697
     },
     {
-      "epoch": 0.8156326384883689,
+      "epoch": 0.8427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.7304844290323807e-06,
-      "loss": 0.8437,
+      "learning_rate": 1.2693985632368622e-06,
+      "loss": 0.7832,
       "step": 29698
     },
     {
-      "epoch": 0.8156601027162121,
+      "epoch": 0.8427639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.729984310757058e-06,
-      "loss": 0.7975,
+      "learning_rate": 1.2689504461233982e-06,
+      "loss": 0.647,
       "step": 29699
     },
     {
-      "epoch": 0.8156875669440554,
+      "epoch": 0.8427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 1.7294842579173033e-06,
-      "loss": 0.8068,
+      "learning_rate": 1.2685024027615034e-06,
+      "loss": 0.8098,
       "step": 29700
     },
     {
-      "epoch": 0.8157150311718986,
+      "epoch": 0.8428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.7289842705170734e-06,
-      "loss": 0.8326,
+      "learning_rate": 1.2680544331549617e-06,
+      "loss": 0.8829,
       "step": 29701
     },
     {
-      "epoch": 0.8157424953997419,
+      "epoch": 0.8428490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.7284843485603231e-06,
-      "loss": 0.8806,
+      "learning_rate": 1.2676065373075552e-06,
+      "loss": 0.7465,
       "step": 29702
     },
     {
-      "epoch": 0.815769959627585,
+      "epoch": 0.842877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.7279844920510093e-06,
-      "loss": 0.8246,
+      "learning_rate": 1.267158715223069e-06,
+      "loss": 0.7909,
       "step": 29703
     },
     {
-      "epoch": 0.8157974238554283,
+      "epoch": 0.842905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7274847009930872e-06,
-      "loss": 0.8508,
+      "learning_rate": 1.266710966905289e-06,
+      "loss": 0.873,
       "step": 29704
     },
     {
-      "epoch": 0.8158248880832716,
+      "epoch": 0.8429341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 1.7269849753905122e-06,
-      "loss": 0.9174,
+      "learning_rate": 1.2662632923579899e-06,
+      "loss": 0.8599,
       "step": 29705
     },
     {
-      "epoch": 0.8158523523111147,
+      "epoch": 0.8429625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.7264853152472338e-06,
-      "loss": 0.8167,
+      "learning_rate": 1.2658156915849585e-06,
+      "loss": 0.7571,
       "step": 29706
     },
     {
-      "epoch": 0.815879816538958,
+      "epoch": 0.8429909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 1.7259857205672092e-06,
-      "loss": 0.853,
+      "learning_rate": 1.2653681645899774e-06,
+      "loss": 0.7411,
       "step": 29707
     },
     {
-      "epoch": 0.8159072807668012,
+      "epoch": 0.8430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 1.7254861913543897e-06,
-      "loss": 0.7907,
+      "learning_rate": 1.2649207113768202e-06,
+      "loss": 0.8552,
       "step": 29708
     },
     {
-      "epoch": 0.8159347449946445,
+      "epoch": 0.8430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.7249867276127297e-06,
-      "loss": 0.804,
+      "learning_rate": 1.264473331949272e-06,
+      "loss": 0.8071,
       "step": 29709
     },
     {
-      "epoch": 0.8159622092224877,
+      "epoch": 0.8430760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 1.7244873293461818e-06,
-      "loss": 0.8222,
+      "learning_rate": 1.2640260263111093e-06,
+      "loss": 0.7426,
       "step": 29710
     },
     {
-      "epoch": 0.8159896734503309,
+      "epoch": 0.8431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.7239879965586926e-06,
-      "loss": 0.7725,
+      "learning_rate": 1.2635787944661114e-06,
+      "loss": 0.8077,
       "step": 29711
     },
     {
-      "epoch": 0.8160171376781742,
+      "epoch": 0.8431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.7234887292542202e-06,
-      "loss": 0.7392,
+      "learning_rate": 1.263131636418059e-06,
+      "loss": 0.6895,
       "step": 29712
     },
     {
-      "epoch": 0.8160446019060175,
+      "epoch": 0.8431611804767309,
       "grad_norm": 0.0,
-      "learning_rate": 1.722989527436707e-06,
-      "loss": 0.7992,
+      "learning_rate": 1.2626845521707232e-06,
+      "loss": 0.7522,
       "step": 29713
     },
     {
-      "epoch": 0.8160720661338606,
+      "epoch": 0.8431895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 1.7224903911101088e-06,
-      "loss": 0.7114,
+      "learning_rate": 1.2622375417278843e-06,
+      "loss": 0.8532,
       "step": 29714
     },
     {
-      "epoch": 0.8160995303617039,
+      "epoch": 0.8432179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.721991320278371e-06,
-      "loss": 0.8701,
+      "learning_rate": 1.2617906050933204e-06,
+      "loss": 0.771,
       "step": 29715
     },
     {
-      "epoch": 0.8161269945895471,
+      "epoch": 0.8432463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 1.7214923149454476e-06,
-      "loss": 0.809,
+      "learning_rate": 1.261343742270802e-06,
+      "loss": 0.962,
       "step": 29716
     },
     {
-      "epoch": 0.8161544588173903,
+      "epoch": 0.8432746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 1.720993375115282e-06,
-      "loss": 0.7425,
+      "learning_rate": 1.2608969532641046e-06,
+      "loss": 0.9327,
       "step": 29717
     },
     {
-      "epoch": 0.8161819230452336,
+      "epoch": 0.8433030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.7204945007918238e-06,
-      "loss": 0.8788,
+      "learning_rate": 1.2604502380770067e-06,
+      "loss": 0.8697,
       "step": 29718
     },
     {
-      "epoch": 0.8162093872730768,
+      "epoch": 0.8433314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.7199956919790195e-06,
-      "loss": 0.8622,
+      "learning_rate": 1.2600035967132762e-06,
+      "loss": 0.8095,
       "step": 29719
     },
     {
-      "epoch": 0.8162368515009201,
+      "epoch": 0.8433598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.71949694868082e-06,
-      "loss": 0.8776,
+      "learning_rate": 1.259557029176688e-06,
+      "loss": 0.7022,
       "step": 29720
     },
     {
-      "epoch": 0.8162643157287632,
+      "epoch": 0.8433881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 1.7189982709011643e-06,
-      "loss": 0.8607,
+      "learning_rate": 1.2591105354710165e-06,
+      "loss": 0.8298,
       "step": 29721
     },
     {
-      "epoch": 0.8162917799566065,
+      "epoch": 0.843416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.718499658644006e-06,
-      "loss": 0.9027,
+      "learning_rate": 1.258664115600029e-06,
+      "loss": 0.7427,
       "step": 29722
     },
     {
-      "epoch": 0.8163192441844498,
+      "epoch": 0.8434449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.7180011119132834e-06,
-      "loss": 0.7904,
+      "learning_rate": 1.258217769567499e-06,
+      "loss": 0.7323,
       "step": 29723
     },
     {
-      "epoch": 0.816346708412293,
+      "epoch": 0.8434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.717502630712944e-06,
-      "loss": 0.7719,
+      "learning_rate": 1.2577714973771983e-06,
+      "loss": 0.775,
       "step": 29724
     },
     {
-      "epoch": 0.8163741726401362,
+      "epoch": 0.8435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.7170042150469313e-06,
-      "loss": 0.8473,
+      "learning_rate": 1.2573252990328911e-06,
+      "loss": 0.8309,
       "step": 29725
     },
     {
-      "epoch": 0.8164016368679795,
+      "epoch": 0.8435300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.7165058649191911e-06,
-      "loss": 0.8425,
+      "learning_rate": 1.2568791745383513e-06,
+      "loss": 0.7858,
       "step": 29726
     },
     {
-      "epoch": 0.8164291010958227,
+      "epoch": 0.8435584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 1.7160075803336674e-06,
-      "loss": 0.818,
+      "learning_rate": 1.2564331238973494e-06,
+      "loss": 0.886,
       "step": 29727
     },
     {
-      "epoch": 0.816456565323666,
+      "epoch": 0.8435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.7155093612942975e-06,
-      "loss": 0.7347,
+      "learning_rate": 1.2559871471136464e-06,
+      "loss": 0.8496,
       "step": 29728
     },
     {
-      "epoch": 0.8164840295515091,
+      "epoch": 0.8436152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.7150112078050296e-06,
-      "loss": 0.8777,
+      "learning_rate": 1.2555412441910165e-06,
+      "loss": 0.8954,
       "step": 29729
     },
     {
-      "epoch": 0.8165114937793524,
+      "epoch": 0.8436435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 1.7145131198697996e-06,
-      "loss": 0.8687,
+      "learning_rate": 1.2550954151332205e-06,
+      "loss": 0.8201,
       "step": 29730
     },
     {
-      "epoch": 0.8165389580071957,
+      "epoch": 0.8436719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 1.7140150974925518e-06,
-      "loss": 0.7745,
+      "learning_rate": 1.2546496599440261e-06,
+      "loss": 0.7777,
       "step": 29731
     },
     {
-      "epoch": 0.8165664222350388,
+      "epoch": 0.8437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 1.7135171406772267e-06,
-      "loss": 0.9035,
+      "learning_rate": 1.2542039786272008e-06,
+      "loss": 0.9089,
       "step": 29732
     },
     {
-      "epoch": 0.8165938864628821,
+      "epoch": 0.8437287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.7130192494277654e-06,
-      "loss": 0.838,
+      "learning_rate": 1.2537583711865065e-06,
+      "loss": 0.8239,
       "step": 29733
     },
     {
-      "epoch": 0.8166213506907253,
+      "epoch": 0.8437570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 1.7125214237481035e-06,
-      "loss": 0.8685,
+      "learning_rate": 1.2533128376257088e-06,
+      "loss": 0.8735,
       "step": 29734
     },
     {
-      "epoch": 0.8166488149185686,
+      "epoch": 0.8437854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 1.7120236636421817e-06,
-      "loss": 0.7623,
+      "learning_rate": 1.2528673779485723e-06,
+      "loss": 0.8236,
       "step": 29735
     },
     {
-      "epoch": 0.8166762791464118,
+      "epoch": 0.8438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.711525969113943e-06,
-      "loss": 0.8136,
+      "learning_rate": 1.2524219921588553e-06,
+      "loss": 0.8293,
       "step": 29736
     },
     {
-      "epoch": 0.816703743374255,
+      "epoch": 0.8438422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 1.7110283401673177e-06,
-      "loss": 0.8484,
+      "learning_rate": 1.2519766802603239e-06,
+      "loss": 0.854,
       "step": 29737
     },
     {
-      "epoch": 0.8167312076020983,
+      "epoch": 0.8438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.7105307768062485e-06,
-      "loss": 0.6883,
+      "learning_rate": 1.2515314422567405e-06,
+      "loss": 0.9057,
       "step": 29738
     },
     {
-      "epoch": 0.8167586718299416,
+      "epoch": 0.8438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.7100332790346719e-06,
-      "loss": 0.7851,
+      "learning_rate": 1.2510862781518628e-06,
+      "loss": 0.6796,
       "step": 29739
     },
     {
-      "epoch": 0.8167861360577847,
+      "epoch": 0.843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.7095358468565215e-06,
-      "loss": 0.8424,
+      "learning_rate": 1.2506411879494496e-06,
+      "loss": 0.7247,
       "step": 29740
     },
     {
-      "epoch": 0.816813600285628,
+      "epoch": 0.843955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.7090384802757343e-06,
-      "loss": 0.7355,
+      "learning_rate": 1.2501961716532685e-06,
+      "loss": 0.7943,
       "step": 29741
     },
     {
-      "epoch": 0.8168410645134712,
+      "epoch": 0.8439841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.708541179296247e-06,
-      "loss": 0.8198,
+      "learning_rate": 1.249751229267071e-06,
+      "loss": 0.8858,
       "step": 29742
     },
     {
-      "epoch": 0.8168685287413144,
+      "epoch": 0.8440124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 1.708043943921993e-06,
-      "loss": 0.8202,
+      "learning_rate": 1.2493063607946188e-06,
+      "loss": 0.7716,
       "step": 29743
     },
     {
-      "epoch": 0.8168959929691577,
+      "epoch": 0.8440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 1.7075467741569107e-06,
-      "loss": 0.8415,
+      "learning_rate": 1.2488615662396709e-06,
+      "loss": 0.7812,
       "step": 29744
     },
     {
-      "epoch": 0.8169234571970009,
+      "epoch": 0.8440692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.7070496700049298e-06,
-      "loss": 0.8577,
+      "learning_rate": 1.248416845605981e-06,
+      "loss": 0.864,
       "step": 29745
     },
     {
-      "epoch": 0.8169509214248442,
+      "epoch": 0.8440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.706552631469982e-06,
-      "loss": 0.8135,
+      "learning_rate": 1.247972198897307e-06,
+      "loss": 0.8689,
       "step": 29746
     },
     {
-      "epoch": 0.8169783856526873,
+      "epoch": 0.8441259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.7060556585560017e-06,
-      "loss": 0.8448,
+      "learning_rate": 1.2475276261174074e-06,
+      "loss": 0.8249,
       "step": 29747
     },
     {
-      "epoch": 0.8170058498805306,
+      "epoch": 0.8441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.7055587512669225e-06,
-      "loss": 0.7465,
+      "learning_rate": 1.2470831272700334e-06,
+      "loss": 0.852,
       "step": 29748
     },
     {
-      "epoch": 0.8170333141083739,
+      "epoch": 0.8441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.7050619096066757e-06,
-      "loss": 0.8506,
+      "learning_rate": 1.246638702358941e-06,
+      "loss": 0.7739,
       "step": 29749
     },
     {
-      "epoch": 0.817060778336217,
+      "epoch": 0.844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.7045651335791914e-06,
-      "loss": 0.8503,
+      "learning_rate": 1.2461943513878883e-06,
+      "loss": 0.759,
       "step": 29750
     },
     {
-      "epoch": 0.8170882425640603,
+      "epoch": 0.8442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 1.704068423188404e-06,
-      "loss": 0.7063,
+      "learning_rate": 1.2457500743606232e-06,
+      "loss": 0.7708,
       "step": 29751
     },
     {
-      "epoch": 0.8171157067919036,
+      "epoch": 0.8442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.7035717784382387e-06,
-      "loss": 0.7881,
+      "learning_rate": 1.2453058712808996e-06,
+      "loss": 0.9415,
       "step": 29752
     },
     {
-      "epoch": 0.8171431710197468,
+      "epoch": 0.8442962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.703075199332629e-06,
-      "loss": 0.8174,
+      "learning_rate": 1.2448617421524745e-06,
+      "loss": 0.7883,
       "step": 29753
     },
     {
-      "epoch": 0.81717063524759,
+      "epoch": 0.8443246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.7025786858755e-06,
-      "loss": 0.8616,
+      "learning_rate": 1.2444176869790925e-06,
+      "loss": 0.7737,
       "step": 29754
     },
     {
-      "epoch": 0.8171980994754332,
+      "epoch": 0.8443530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.7020822380707835e-06,
-      "loss": 0.9045,
+      "learning_rate": 1.2439737057645106e-06,
+      "loss": 0.8257,
       "step": 29755
     },
     {
-      "epoch": 0.8172255637032765,
+      "epoch": 0.8443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.7015858559224053e-06,
-      "loss": 0.7647,
+      "learning_rate": 1.2435297985124716e-06,
+      "loss": 0.7763,
       "step": 29756
     },
     {
-      "epoch": 0.8172530279311198,
+      "epoch": 0.8444097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.7010895394342975e-06,
-      "loss": 0.7941,
+      "learning_rate": 1.2430859652267323e-06,
+      "loss": 0.8289,
       "step": 29757
     },
     {
-      "epoch": 0.8172804921589629,
+      "epoch": 0.8444381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 1.7005932886103815e-06,
-      "loss": 0.873,
+      "learning_rate": 1.242642205911042e-06,
+      "loss": 0.8719,
       "step": 29758
     },
     {
-      "epoch": 0.8173079563868062,
+      "epoch": 0.844466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.7000971034545854e-06,
-      "loss": 0.7858,
+      "learning_rate": 1.2421985205691444e-06,
+      "loss": 0.7926,
       "step": 29759
     },
     {
-      "epoch": 0.8173354206146494,
+      "epoch": 0.8444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.6996009839708394e-06,
-      "loss": 0.8581,
+      "learning_rate": 1.2417549092047898e-06,
+      "loss": 0.8014,
       "step": 29760
     },
     {
-      "epoch": 0.8173628848424926,
+      "epoch": 0.8445232690124859,
       "grad_norm": 0.0,
-      "learning_rate": 1.6991049301630636e-06,
-      "loss": 0.7401,
+      "learning_rate": 1.2413113718217273e-06,
+      "loss": 0.844,
       "step": 29761
     },
     {
-      "epoch": 0.8173903490703359,
+      "epoch": 0.8445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.6986089420351836e-06,
-      "loss": 0.8664,
+      "learning_rate": 1.2408679084236985e-06,
+      "loss": 0.7687,
       "step": 29762
     },
     {
-      "epoch": 0.8174178132981791,
+      "epoch": 0.8445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.6981130195911288e-06,
-      "loss": 0.8313,
+      "learning_rate": 1.2404245190144537e-06,
+      "loss": 0.7192,
       "step": 29763
     },
     {
-      "epoch": 0.8174452775260224,
+      "epoch": 0.8446083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.697617162834817e-06,
-      "loss": 0.8435,
+      "learning_rate": 1.2399812035977377e-06,
+      "loss": 0.8085,
       "step": 29764
     },
     {
-      "epoch": 0.8174727417538656,
+      "epoch": 0.8446367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.697121371770174e-06,
-      "loss": 0.7612,
+      "learning_rate": 1.2395379621772918e-06,
+      "loss": 0.7922,
       "step": 29765
     },
     {
-      "epoch": 0.8175002059817088,
+      "epoch": 0.8446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.6966256464011233e-06,
-      "loss": 0.8905,
+      "learning_rate": 1.2390947947568632e-06,
+      "loss": 0.8136,
       "step": 29766
     },
     {
-      "epoch": 0.8175276702095521,
+      "epoch": 0.8446935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 1.6961299867315873e-06,
-      "loss": 0.8195,
+      "learning_rate": 1.2386517013401967e-06,
+      "loss": 0.9113,
       "step": 29767
     },
     {
-      "epoch": 0.8175551344373952,
+      "epoch": 0.8447219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.6956343927654895e-06,
-      "loss": 0.8546,
+      "learning_rate": 1.2382086819310312e-06,
+      "loss": 0.7621,
       "step": 29768
     },
     {
-      "epoch": 0.8175825986652385,
+      "epoch": 0.844750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.695138864506749e-06,
-      "loss": 0.7734,
+      "learning_rate": 1.237765736533111e-06,
+      "loss": 0.7482,
       "step": 29769
     },
     {
-      "epoch": 0.8176100628930818,
+      "epoch": 0.8447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.694643401959284e-06,
-      "loss": 0.8033,
+      "learning_rate": 1.237322865150179e-06,
+      "loss": 0.8853,
       "step": 29770
     },
     {
-      "epoch": 0.817637527120925,
+      "epoch": 0.8448070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.6941480051270176e-06,
-      "loss": 0.6577,
+      "learning_rate": 1.2368800677859716e-06,
+      "loss": 0.8199,
       "step": 29771
     },
     {
-      "epoch": 0.8176649913487682,
+      "epoch": 0.8448354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.6936526740138704e-06,
-      "loss": 0.8513,
+      "learning_rate": 1.2364373444442356e-06,
+      "loss": 0.7741,
       "step": 29772
     },
     {
-      "epoch": 0.8176924555766114,
+      "epoch": 0.8448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.6931574086237602e-06,
-      "loss": 0.7514,
+      "learning_rate": 1.235994695128705e-06,
+      "loss": 0.7553,
       "step": 29773
     },
     {
-      "epoch": 0.8177199198044547,
+      "epoch": 0.8448921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.692662208960607e-06,
-      "loss": 0.8605,
+      "learning_rate": 1.2355521198431209e-06,
+      "loss": 0.828,
       "step": 29774
     },
     {
-      "epoch": 0.817747384032298,
+      "epoch": 0.8449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6921670750283303e-06,
-      "loss": 0.7988,
+      "learning_rate": 1.2351096185912248e-06,
+      "loss": 0.9626,
       "step": 29775
     },
     {
-      "epoch": 0.8177748482601411,
+      "epoch": 0.8449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.6916720068308435e-06,
-      "loss": 0.8586,
+      "learning_rate": 1.2346671913767494e-06,
+      "loss": 0.8751,
       "step": 29776
     },
     {
-      "epoch": 0.8178023124879844,
+      "epoch": 0.8449772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.6911770043720688e-06,
-      "loss": 0.8741,
+      "learning_rate": 1.234224838203435e-06,
+      "loss": 0.7491,
       "step": 29777
     },
     {
-      "epoch": 0.8178297767158277,
+      "epoch": 0.8450056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 1.6906820676559165e-06,
-      "loss": 0.7772,
+      "learning_rate": 1.2337825590750186e-06,
+      "loss": 0.827,
       "step": 29778
     },
     {
-      "epoch": 0.8178572409436708,
+      "epoch": 0.8450340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.690187196686308e-06,
-      "loss": 0.7983,
+      "learning_rate": 1.2333403539952327e-06,
+      "loss": 0.8025,
       "step": 29779
     },
     {
-      "epoch": 0.8178847051715141,
+      "epoch": 0.8450624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.689692391467156e-06,
-      "loss": 0.7922,
+      "learning_rate": 1.2328982229678154e-06,
+      "loss": 0.8485,
       "step": 29780
     },
     {
-      "epoch": 0.8179121693993573,
+      "epoch": 0.8450908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.68919765200238e-06,
-      "loss": 0.8541,
+      "learning_rate": 1.2324561659965029e-06,
+      "loss": 0.8259,
       "step": 29781
     },
     {
-      "epoch": 0.8179396336272006,
+      "epoch": 0.8451191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.6887029782958896e-06,
-      "loss": 0.7986,
+      "learning_rate": 1.2320141830850253e-06,
+      "loss": 0.763,
       "step": 29782
     },
     {
-      "epoch": 0.8179670978550438,
+      "epoch": 0.8451475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.6882083703515994e-06,
-      "loss": 0.6662,
+      "learning_rate": 1.2315722742371184e-06,
+      "loss": 0.8567,
       "step": 29783
     },
     {
-      "epoch": 0.817994562082887,
+      "epoch": 0.8451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.6877138281734284e-06,
-      "loss": 0.8168,
+      "learning_rate": 1.231130439456516e-06,
+      "loss": 0.7539,
       "step": 29784
     },
     {
-      "epoch": 0.8180220263107303,
+      "epoch": 0.8452043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 1.6872193517652813e-06,
-      "loss": 0.8375,
+      "learning_rate": 1.2306886787469475e-06,
+      "loss": 0.8157,
       "step": 29785
     },
     {
-      "epoch": 0.8180494905385735,
+      "epoch": 0.8452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.6867249411310782e-06,
-      "loss": 0.7456,
+      "learning_rate": 1.2302469921121462e-06,
+      "loss": 0.8575,
       "step": 29786
     },
     {
-      "epoch": 0.8180769547664167,
+      "epoch": 0.845261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.686230596274725e-06,
-      "loss": 0.8555,
+      "learning_rate": 1.2298053795558419e-06,
+      "loss": 0.8382,
       "step": 29787
     },
     {
-      "epoch": 0.81810441899426,
+      "epoch": 0.8452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.6857363172001362e-06,
-      "loss": 0.6992,
+      "learning_rate": 1.2293638410817665e-06,
+      "loss": 0.8059,
       "step": 29788
     },
     {
-      "epoch": 0.8181318832221032,
+      "epoch": 0.8453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.685242103911221e-06,
-      "loss": 0.7232,
+      "learning_rate": 1.2289223766936508e-06,
+      "loss": 0.7677,
       "step": 29789
     },
     {
-      "epoch": 0.8181593474499465,
+      "epoch": 0.8453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.684747956411893e-06,
-      "loss": 0.7563,
+      "learning_rate": 1.2284809863952197e-06,
+      "loss": 0.7887,
       "step": 29790
     },
     {
-      "epoch": 0.8181868116777897,
+      "epoch": 0.8453745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.6842538747060588e-06,
-      "loss": 0.831,
+      "learning_rate": 1.2280396701902043e-06,
+      "loss": 0.8357,
       "step": 29791
     },
     {
-      "epoch": 0.8182142759056329,
+      "epoch": 0.8454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.6837598587976312e-06,
-      "loss": 0.8966,
+      "learning_rate": 1.227598428082335e-06,
+      "loss": 0.72,
       "step": 29792
     },
     {
-      "epoch": 0.8182417401334762,
+      "epoch": 0.8454313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.6832659086905178e-06,
-      "loss": 0.7551,
+      "learning_rate": 1.2271572600753324e-06,
+      "loss": 0.801,
       "step": 29793
     },
     {
-      "epoch": 0.8182692043613193,
+      "epoch": 0.8454597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 1.6827720243886235e-06,
-      "loss": 0.8992,
+      "learning_rate": 1.2267161661729277e-06,
+      "loss": 0.7374,
       "step": 29794
     },
     {
-      "epoch": 0.8182966685891626,
+      "epoch": 0.8454880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.6822782058958587e-06,
-      "loss": 0.8054,
+      "learning_rate": 1.226275146378848e-06,
+      "loss": 0.7241,
       "step": 29795
     },
     {
-      "epoch": 0.8183241328170059,
+      "epoch": 0.8455164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 1.681784453216131e-06,
-      "loss": 0.7182,
+      "learning_rate": 1.225834200696815e-06,
+      "loss": 0.8384,
       "step": 29796
     },
     {
-      "epoch": 0.818351597044849,
+      "epoch": 0.8455448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.6812907663533473e-06,
-      "loss": 0.8597,
+      "learning_rate": 1.2253933291305542e-06,
+      "loss": 0.7768,
       "step": 29797
     },
     {
-      "epoch": 0.8183790612726923,
+      "epoch": 0.8455732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.6807971453114146e-06,
-      "loss": 0.7473,
+      "learning_rate": 1.224952531683793e-06,
+      "loss": 0.9258,
       "step": 29798
     },
     {
-      "epoch": 0.8184065255005355,
+      "epoch": 0.8456015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 1.6803035900942355e-06,
-      "loss": 0.8501,
+      "learning_rate": 1.2245118083602492e-06,
+      "loss": 0.7333,
       "step": 29799
     },
     {
-      "epoch": 0.8184339897283788,
+      "epoch": 0.8456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.6798101007057178e-06,
-      "loss": 0.7506,
+      "learning_rate": 1.224071159163649e-06,
+      "loss": 0.8426,
       "step": 29800
     },
     {
-      "epoch": 0.818461453956222,
+      "epoch": 0.8456583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.679316677149767e-06,
-      "loss": 0.7466,
+      "learning_rate": 1.223630584097717e-06,
+      "loss": 0.8274,
       "step": 29801
     },
     {
-      "epoch": 0.8184889181840652,
+      "epoch": 0.8456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.678823319430284e-06,
-      "loss": 0.7995,
+      "learning_rate": 1.223190083166167e-06,
+      "loss": 0.7593,
       "step": 29802
     },
     {
-      "epoch": 0.8185163824119085,
+      "epoch": 0.8457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.6783300275511737e-06,
-      "loss": 0.7114,
+      "learning_rate": 1.2227496563727293e-06,
+      "loss": 0.7925,
       "step": 29803
     },
     {
-      "epoch": 0.8185438466397518,
+      "epoch": 0.8457434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.6778368015163426e-06,
-      "loss": 0.8413,
+      "learning_rate": 1.2223093037211186e-06,
+      "loss": 0.7491,
       "step": 29804
     },
     {
-      "epoch": 0.8185713108675949,
+      "epoch": 0.845771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.6773436413296883e-06,
-      "loss": 0.8455,
+      "learning_rate": 1.2218690252150546e-06,
+      "loss": 0.8291,
       "step": 29805
     },
     {
-      "epoch": 0.8185987750954382,
+      "epoch": 0.845800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.6768505469951146e-06,
-      "loss": 0.8364,
+      "learning_rate": 1.2214288208582604e-06,
+      "loss": 0.8542,
       "step": 29806
     },
     {
-      "epoch": 0.8186262393232814,
+      "epoch": 0.8458286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.6763575185165237e-06,
-      "loss": 0.7638,
+      "learning_rate": 1.2209886906544489e-06,
+      "loss": 0.865,
       "step": 29807
     },
     {
-      "epoch": 0.8186537035511247,
+      "epoch": 0.8458569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.675864555897818e-06,
-      "loss": 0.8552,
+      "learning_rate": 1.2205486346073415e-06,
+      "loss": 0.806,
       "step": 29808
     },
     {
-      "epoch": 0.8186811677789679,
+      "epoch": 0.8458853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.6753716591428948e-06,
-      "loss": 0.875,
+      "learning_rate": 1.2201086527206552e-06,
+      "loss": 0.7423,
       "step": 29809
     },
     {
-      "epoch": 0.8187086320068111,
+      "epoch": 0.8459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.674878828255657e-06,
-      "loss": 0.7984,
+      "learning_rate": 1.219668744998105e-06,
+      "loss": 0.8463,
       "step": 29810
     },
     {
-      "epoch": 0.8187360962346544,
+      "epoch": 0.8459421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.674386063240002e-06,
-      "loss": 0.7581,
+      "learning_rate": 1.2192289114434064e-06,
+      "loss": 0.825,
       "step": 29811
     },
     {
-      "epoch": 0.8187635604624975,
+      "epoch": 0.8459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.673893364099829e-06,
-      "loss": 0.7636,
+      "learning_rate": 1.218789152060279e-06,
+      "loss": 0.7739,
       "step": 29812
     },
     {
-      "epoch": 0.8187910246903408,
+      "epoch": 0.8459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.6734007308390355e-06,
-      "loss": 0.7653,
+      "learning_rate": 1.218349466852432e-06,
+      "loss": 0.7783,
       "step": 29813
     },
     {
-      "epoch": 0.8188184889181841,
+      "epoch": 0.8460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.6729081634615218e-06,
-      "loss": 0.8024,
+      "learning_rate": 1.2179098558235814e-06,
+      "loss": 0.753,
       "step": 29814
     },
     {
-      "epoch": 0.8188459531460273,
+      "epoch": 0.84605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.6724156619711851e-06,
-      "loss": 0.8016,
+      "learning_rate": 1.217470318977444e-06,
+      "loss": 0.7771,
       "step": 29815
     },
     {
-      "epoch": 0.8188734173738705,
+      "epoch": 0.8460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.671923226371923e-06,
-      "loss": 0.7822,
+      "learning_rate": 1.2170308563177268e-06,
+      "loss": 0.8551,
       "step": 29816
     },
     {
-      "epoch": 0.8189008816017138,
+      "epoch": 0.8461123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.6714308566676306e-06,
-      "loss": 0.7654,
+      "learning_rate": 1.2165914678481449e-06,
+      "loss": 0.8079,
       "step": 29817
     },
     {
-      "epoch": 0.818928345829557,
+      "epoch": 0.8461407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.6709385528622012e-06,
-      "loss": 0.7094,
+      "learning_rate": 1.2161521535724097e-06,
+      "loss": 0.7865,
       "step": 29818
     },
     {
-      "epoch": 0.8189558100574003,
+      "epoch": 0.8461691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.6704463149595318e-06,
-      "loss": 0.7403,
+      "learning_rate": 1.2157129134942325e-06,
+      "loss": 0.8679,
       "step": 29819
     },
     {
-      "epoch": 0.8189832742852434,
+      "epoch": 0.8461975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.669954142963518e-06,
-      "loss": 0.7143,
+      "learning_rate": 1.2152737476173248e-06,
+      "loss": 0.7487,
       "step": 29820
     },
     {
-      "epoch": 0.8190107385130867,
+      "epoch": 0.8462258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.6694620368780534e-06,
-      "loss": 0.7712,
+      "learning_rate": 1.2148346559453928e-06,
+      "loss": 0.7284,
       "step": 29821
     },
     {
-      "epoch": 0.81903820274093,
+      "epoch": 0.8462542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.668969996707035e-06,
-      "loss": 0.8043,
+      "learning_rate": 1.2143956384821475e-06,
+      "loss": 0.8359,
       "step": 29822
     },
     {
-      "epoch": 0.8190656669687731,
+      "epoch": 0.8462826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 1.6684780224543507e-06,
-      "loss": 0.8427,
+      "learning_rate": 1.2139566952312987e-06,
+      "loss": 0.7722,
       "step": 29823
     },
     {
-      "epoch": 0.8190931311966164,
+      "epoch": 0.846311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6679861141238985e-06,
-      "loss": 0.9485,
+      "learning_rate": 1.2135178261965508e-06,
+      "loss": 0.7095,
       "step": 29824
     },
     {
-      "epoch": 0.8191205954244597,
+      "epoch": 0.8463393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.6674942717195642e-06,
-      "loss": 0.7773,
+      "learning_rate": 1.2130790313816133e-06,
+      "loss": 0.862,
       "step": 29825
     },
     {
-      "epoch": 0.8191480596523029,
+      "epoch": 0.8463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.6670024952452435e-06,
-      "loss": 0.7907,
+      "learning_rate": 1.212640310790194e-06,
+      "loss": 0.8451,
       "step": 29826
     },
     {
-      "epoch": 0.8191755238801461,
+      "epoch": 0.8463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.6665107847048267e-06,
-      "loss": 0.8278,
+      "learning_rate": 1.2122016644259948e-06,
+      "loss": 0.8445,
       "step": 29827
     },
     {
-      "epoch": 0.8192029881079893,
+      "epoch": 0.8464245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.6660191401022074e-06,
-      "loss": 0.8151,
+      "learning_rate": 1.2117630922927237e-06,
+      "loss": 0.8449,
       "step": 29828
     },
     {
-      "epoch": 0.8192304523358326,
+      "epoch": 0.8464528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.6655275614412702e-06,
-      "loss": 0.7901,
+      "learning_rate": 1.2113245943940865e-06,
+      "loss": 0.7462,
       "step": 29829
     },
     {
-      "epoch": 0.8192579165636759,
+      "epoch": 0.8464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.6650360487259077e-06,
-      "loss": 0.8325,
+      "learning_rate": 1.2108861707337827e-06,
+      "loss": 0.8224,
       "step": 29830
     },
     {
-      "epoch": 0.819285380791519,
+      "epoch": 0.8465096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.6645446019600087e-06,
-      "loss": 0.7827,
+      "learning_rate": 1.2104478213155202e-06,
+      "loss": 0.9013,
       "step": 29831
     },
     {
-      "epoch": 0.8193128450193623,
+      "epoch": 0.8465380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 1.6640532211474637e-06,
-      "loss": 0.791,
+      "learning_rate": 1.2100095461429984e-06,
+      "loss": 0.7207,
       "step": 29832
     },
     {
-      "epoch": 0.8193403092472055,
+      "epoch": 0.8465664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 1.663561906292156e-06,
-      "loss": 0.77,
+      "learning_rate": 1.209571345219922e-06,
+      "loss": 0.7697,
       "step": 29833
     },
     {
-      "epoch": 0.8193677734750487,
+      "epoch": 0.846594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.6630706573979793e-06,
-      "loss": 0.7385,
+      "learning_rate": 1.2091332185499916e-06,
+      "loss": 0.8249,
       "step": 29834
     },
     {
-      "epoch": 0.819395237702892,
+      "epoch": 0.8466231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.662579474468814e-06,
-      "loss": 0.8049,
+      "learning_rate": 1.2086951661369095e-06,
+      "loss": 0.7698,
       "step": 29835
     },
     {
-      "epoch": 0.8194227019307352,
+      "epoch": 0.8466515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 1.662088357508549e-06,
-      "loss": 0.8051,
+      "learning_rate": 1.208257187984373e-06,
+      "loss": 0.8424,
       "step": 29836
     },
     {
-      "epoch": 0.8194501661585785,
+      "epoch": 0.8466799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.661597306521071e-06,
-      "loss": 0.7482,
+      "learning_rate": 1.2078192840960857e-06,
+      "loss": 0.7512,
       "step": 29837
     },
     {
-      "epoch": 0.8194776303864217,
+      "epoch": 0.8467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.6611063215102652e-06,
-      "loss": 0.8065,
+      "learning_rate": 1.2073814544757413e-06,
+      "loss": 0.7643,
       "step": 29838
     },
     {
-      "epoch": 0.8195050946142649,
+      "epoch": 0.8467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.660615402480016e-06,
-      "loss": 0.7735,
+      "learning_rate": 1.206943699127041e-06,
+      "loss": 0.7768,
       "step": 29839
     },
     {
-      "epoch": 0.8195325588421082,
+      "epoch": 0.8467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.6601245494342111e-06,
-      "loss": 0.7753,
+      "learning_rate": 1.2065060180536858e-06,
+      "loss": 0.6969,
       "step": 29840
     },
     {
-      "epoch": 0.8195600230699513,
+      "epoch": 0.8467934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.659633762376731e-06,
-      "loss": 0.8405,
+      "learning_rate": 1.2060684112593668e-06,
+      "loss": 0.8108,
       "step": 29841
     },
     {
-      "epoch": 0.8195874872977946,
+      "epoch": 0.846821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.6591430413114573e-06,
-      "loss": 0.8127,
+      "learning_rate": 1.2056308787477822e-06,
+      "loss": 0.8698,
       "step": 29842
     },
     {
-      "epoch": 0.8196149515256379,
+      "epoch": 0.846850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.6586523862422743e-06,
-      "loss": 0.8434,
+      "learning_rate": 1.2051934205226324e-06,
+      "loss": 0.8153,
       "step": 29843
     },
     {
-      "epoch": 0.8196424157534811,
+      "epoch": 0.8468785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.6581617971730646e-06,
-      "loss": 0.8861,
+      "learning_rate": 1.2047560365876065e-06,
+      "loss": 0.9056,
       "step": 29844
     },
     {
-      "epoch": 0.8196698799813243,
+      "epoch": 0.8469069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.6576712741077127e-06,
-      "loss": 0.8334,
+      "learning_rate": 1.2043187269464018e-06,
+      "loss": 0.8747,
       "step": 29845
     },
     {
-      "epoch": 0.8196973442091675,
+      "epoch": 0.8469353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.6571808170500948e-06,
-      "loss": 0.79,
+      "learning_rate": 1.2038814916027142e-06,
+      "loss": 0.829,
       "step": 29846
     },
     {
-      "epoch": 0.8197248084370108,
+      "epoch": 0.8469636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.6566904260040939e-06,
-      "loss": 0.8675,
+      "learning_rate": 1.2034443305602327e-06,
+      "loss": 0.8501,
       "step": 29847
     },
     {
-      "epoch": 0.8197522726648541,
+      "epoch": 0.8469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.6562001009735929e-06,
-      "loss": 0.7548,
+      "learning_rate": 1.2030072438226514e-06,
+      "loss": 0.8205,
       "step": 29848
     },
     {
-      "epoch": 0.8197797368926972,
+      "epoch": 0.8470204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.6557098419624663e-06,
-      "loss": 0.7305,
+      "learning_rate": 1.202570231393666e-06,
+      "loss": 0.8591,
       "step": 29849
     },
     {
-      "epoch": 0.8198072011205405,
+      "epoch": 0.8470488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 1.6552196489745952e-06,
-      "loss": 0.7379,
+      "learning_rate": 1.2021332932769648e-06,
+      "loss": 0.7012,
       "step": 29850
     },
     {
-      "epoch": 0.8198346653483838,
+      "epoch": 0.8470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.6547295220138627e-06,
-      "loss": 0.7006,
+      "learning_rate": 1.2016964294762379e-06,
+      "loss": 0.783,
       "step": 29851
     },
     {
-      "epoch": 0.819862129576227,
+      "epoch": 0.847105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.6542394610841384e-06,
-      "loss": 0.7817,
+      "learning_rate": 1.2012596399951793e-06,
+      "loss": 0.8608,
       "step": 29852
     },
     {
-      "epoch": 0.8198895938040702,
+      "epoch": 0.8471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.6537494661893062e-06,
-      "loss": 0.7583,
+      "learning_rate": 1.2008229248374737e-06,
+      "loss": 0.7377,
       "step": 29853
     },
     {
-      "epoch": 0.8199170580319134,
+      "epoch": 0.8471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.6532595373332395e-06,
-      "loss": 0.8349,
+      "learning_rate": 1.2003862840068125e-06,
+      "loss": 0.7521,
       "step": 29854
     },
     {
-      "epoch": 0.8199445222597567,
+      "epoch": 0.8471906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.6527696745198173e-06,
-      "loss": 0.8364,
+      "learning_rate": 1.1999497175068863e-06,
+      "loss": 0.7593,
       "step": 29855
     },
     {
-      "epoch": 0.8199719864876,
+      "epoch": 0.8472190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.652279877752917e-06,
-      "loss": 0.8657,
+      "learning_rate": 1.1995132253413788e-06,
+      "loss": 0.7533,
       "step": 29856
     },
     {
-      "epoch": 0.8199994507154431,
+      "epoch": 0.8472474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.6517901470364128e-06,
-      "loss": 0.7504,
+      "learning_rate": 1.199076807513978e-06,
+      "loss": 0.806,
       "step": 29857
     },
     {
-      "epoch": 0.8200269149432864,
+      "epoch": 0.8472758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.6513004823741751e-06,
-      "loss": 0.8424,
+      "learning_rate": 1.1986404640283734e-06,
+      "loss": 0.9158,
       "step": 29858
     },
     {
-      "epoch": 0.8200543791711296,
+      "epoch": 0.8473041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.650810883770083e-06,
-      "loss": 0.8242,
+      "learning_rate": 1.1982041948882462e-06,
+      "loss": 0.7171,
       "step": 29859
     },
     {
-      "epoch": 0.8200818433989728,
+      "epoch": 0.8473325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 1.6503213512280091e-06,
-      "loss": 0.7553,
+      "learning_rate": 1.1977680000972836e-06,
+      "loss": 0.8712,
       "step": 29860
     },
     {
-      "epoch": 0.8201093076268161,
+      "epoch": 0.847360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.649831884751827e-06,
-      "loss": 0.766,
+      "learning_rate": 1.1973318796591726e-06,
+      "loss": 0.7014,
       "step": 29861
     },
     {
-      "epoch": 0.8201367718546593,
+      "epoch": 0.8473893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.6493424843454099e-06,
-      "loss": 0.835,
+      "learning_rate": 1.1968958335775938e-06,
+      "loss": 0.7515,
       "step": 29862
     },
     {
-      "epoch": 0.8201642360825026,
+      "epoch": 0.8474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.6488531500126315e-06,
-      "loss": 0.7774,
+      "learning_rate": 1.1964598618562305e-06,
+      "loss": 0.7531,
       "step": 29863
     },
     {
-      "epoch": 0.8201917003103458,
+      "epoch": 0.8474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.64836388175736e-06,
-      "loss": 0.8315,
+      "learning_rate": 1.196023964498767e-06,
+      "loss": 0.8527,
       "step": 29864
     },
     {
-      "epoch": 0.820219164538189,
+      "epoch": 0.8474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6478746795834698e-06,
-      "loss": 0.8428,
+      "learning_rate": 1.1955881415088844e-06,
+      "loss": 0.8731,
       "step": 29865
     },
     {
-      "epoch": 0.8202466287660323,
+      "epoch": 0.8475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.6473855434948293e-06,
-      "loss": 0.8848,
+      "learning_rate": 1.1951523928902665e-06,
+      "loss": 0.7656,
       "step": 29866
     },
     {
-      "epoch": 0.8202740929938754,
+      "epoch": 0.8475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.6468964734953096e-06,
-      "loss": 0.7983,
+      "learning_rate": 1.1947167186465903e-06,
+      "loss": 0.8619,
       "step": 29867
     },
     {
-      "epoch": 0.8203015572217187,
+      "epoch": 0.8475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.646407469588781e-06,
-      "loss": 0.9188,
+      "learning_rate": 1.1942811187815385e-06,
+      "loss": 0.8522,
       "step": 29868
     },
     {
-      "epoch": 0.820329021449562,
+      "epoch": 0.8475879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 1.6459185317791138e-06,
-      "loss": 0.8568,
+      "learning_rate": 1.1938455932987913e-06,
+      "loss": 0.8418,
       "step": 29869
     },
     {
-      "epoch": 0.8203564856774052,
+      "epoch": 0.8476163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.6454296600701735e-06,
-      "loss": 0.7885,
+      "learning_rate": 1.193410142202024e-06,
+      "loss": 0.7763,
       "step": 29870
     },
     {
-      "epoch": 0.8203839499052484,
+      "epoch": 0.847644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.6449408544658307e-06,
-      "loss": 0.7559,
+      "learning_rate": 1.1929747654949174e-06,
+      "loss": 0.8855,
       "step": 29871
     },
     {
-      "epoch": 0.8204114141330916,
+      "epoch": 0.8476730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.6444521149699533e-06,
-      "loss": 0.791,
+      "learning_rate": 1.1925394631811515e-06,
+      "loss": 0.7714,
       "step": 29872
     },
     {
-      "epoch": 0.8204388783609349,
+      "epoch": 0.8477014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.6439634415864059e-06,
-      "loss": 0.8431,
+      "learning_rate": 1.1921042352643975e-06,
+      "loss": 0.8163,
       "step": 29873
     },
     {
-      "epoch": 0.8204663425887782,
+      "epoch": 0.8477298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.6434748343190566e-06,
-      "loss": 0.7466,
+      "learning_rate": 1.1916690817483355e-06,
+      "loss": 0.7568,
       "step": 29874
     },
     {
-      "epoch": 0.8204938068166213,
+      "epoch": 0.8477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.6429862931717733e-06,
-      "loss": 0.8063,
+      "learning_rate": 1.1912340026366431e-06,
+      "loss": 0.8568,
       "step": 29875
     },
     {
-      "epoch": 0.8205212710444646,
+      "epoch": 0.8477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.6424978181484174e-06,
-      "loss": 0.8316,
+      "learning_rate": 1.1907989979329904e-06,
+      "loss": 0.7955,
       "step": 29876
     },
     {
-      "epoch": 0.8205487352723079,
+      "epoch": 0.8478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.6420094092528561e-06,
-      "loss": 0.8228,
+      "learning_rate": 1.1903640676410556e-06,
+      "loss": 0.7472,
       "step": 29877
     },
     {
-      "epoch": 0.820576199500151,
+      "epoch": 0.8478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.6415210664889536e-06,
-      "loss": 0.7968,
+      "learning_rate": 1.1899292117645101e-06,
+      "loss": 0.8984,
       "step": 29878
     },
     {
-      "epoch": 0.8206036637279943,
+      "epoch": 0.8478717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.641032789860575e-06,
-      "loss": 0.9252,
+      "learning_rate": 1.1894944303070299e-06,
+      "loss": 0.8335,
       "step": 29879
     },
     {
-      "epoch": 0.8206311279558375,
+      "epoch": 0.8479001135073779,
       "grad_norm": 0.0,
-      "learning_rate": 1.6405445793715847e-06,
-      "loss": 0.7508,
+      "learning_rate": 1.1890597232722877e-06,
+      "loss": 0.788,
       "step": 29880
     },
     {
-      "epoch": 0.8206585921836808,
+      "epoch": 0.8479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.640056435025844e-06,
-      "loss": 0.6655,
+      "learning_rate": 1.1886250906639518e-06,
+      "loss": 0.7534,
       "step": 29881
     },
     {
-      "epoch": 0.820686056411524,
+      "epoch": 0.8479568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.6395683568272136e-06,
-      "loss": 0.7425,
+      "learning_rate": 1.1881905324856968e-06,
+      "loss": 0.8711,
       "step": 29882
     },
     {
-      "epoch": 0.8207135206393672,
+      "epoch": 0.8479852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.6390803447795555e-06,
-      "loss": 0.8156,
+      "learning_rate": 1.1877560487411931e-06,
+      "loss": 0.7359,
       "step": 29883
     },
     {
-      "epoch": 0.8207409848672105,
+      "epoch": 0.8480136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.6385923988867336e-06,
-      "loss": 0.7417,
+      "learning_rate": 1.1873216394341091e-06,
+      "loss": 0.7613,
       "step": 29884
     },
     {
-      "epoch": 0.8207684490950536,
+      "epoch": 0.8480419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.6381045191526068e-06,
-      "loss": 0.8472,
+      "learning_rate": 1.186887304568115e-06,
+      "loss": 0.8443,
       "step": 29885
     },
     {
-      "epoch": 0.8207959133228969,
+      "epoch": 0.8480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.6376167055810355e-06,
-      "loss": 0.7886,
+      "learning_rate": 1.1864530441468824e-06,
+      "loss": 0.765,
       "step": 29886
     },
     {
-      "epoch": 0.8208233775507402,
+      "epoch": 0.8480987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.6371289581758842e-06,
-      "loss": 0.7097,
+      "learning_rate": 1.186018858174075e-06,
+      "loss": 0.767,
       "step": 29887
     },
     {
-      "epoch": 0.8208508417785834,
+      "epoch": 0.8481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.6366412769410044e-06,
-      "loss": 0.7967,
+      "learning_rate": 1.1855847466533632e-06,
+      "loss": 0.7779,
       "step": 29888
     },
     {
-      "epoch": 0.8208783060064266,
+      "epoch": 0.848155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.6361536618802608e-06,
-      "loss": 0.8936,
+      "learning_rate": 1.185150709588414e-06,
+      "loss": 0.8204,
       "step": 29889
     },
     {
-      "epoch": 0.8209057702342699,
+      "epoch": 0.8481838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.635666112997507e-06,
-      "loss": 0.8263,
+      "learning_rate": 1.1847167469828924e-06,
+      "loss": 0.7602,
       "step": 29890
     },
     {
-      "epoch": 0.8209332344621131,
+      "epoch": 0.8482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.6351786302966032e-06,
-      "loss": 0.8591,
+      "learning_rate": 1.184282858840463e-06,
+      "loss": 0.7503,
       "step": 29891
     },
     {
-      "epoch": 0.8209606986899564,
+      "epoch": 0.8482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.6346912137814063e-06,
-      "loss": 0.7905,
+      "learning_rate": 1.1838490451647965e-06,
+      "loss": 0.8781,
       "step": 29892
     },
     {
-      "epoch": 0.8209881629177995,
+      "epoch": 0.8482690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.6342038634557756e-06,
-      "loss": 0.7967,
+      "learning_rate": 1.18341530595955e-06,
+      "loss": 0.8596,
       "step": 29893
     },
     {
-      "epoch": 0.8210156271456428,
+      "epoch": 0.8482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.6337165793235622e-06,
-      "loss": 0.7942,
+      "learning_rate": 1.1829816412283912e-06,
+      "loss": 0.7072,
       "step": 29894
     },
     {
-      "epoch": 0.8210430913734861,
+      "epoch": 0.8483257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.633229361388624e-06,
-      "loss": 0.7079,
+      "learning_rate": 1.1825480509749831e-06,
+      "loss": 0.8434,
       "step": 29895
     },
     {
-      "epoch": 0.8210705556013292,
+      "epoch": 0.8483541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.6327422096548174e-06,
-      "loss": 0.9027,
+      "learning_rate": 1.1821145352029872e-06,
+      "loss": 0.7729,
       "step": 29896
     },
     {
-      "epoch": 0.8210980198291725,
+      "epoch": 0.8483825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.6322551241259943e-06,
-      "loss": 0.8146,
+      "learning_rate": 1.1816810939160684e-06,
+      "loss": 0.8417,
       "step": 29897
     },
     {
-      "epoch": 0.8211254840570157,
+      "epoch": 0.8484108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 1.6317681048060096e-06,
-      "loss": 0.87,
+      "learning_rate": 1.1812477271178846e-06,
+      "loss": 0.8233,
       "step": 29898
     },
     {
-      "epoch": 0.821152948284859,
+      "epoch": 0.8484392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.6312811516987203e-06,
-      "loss": 0.7692,
+      "learning_rate": 1.1808144348120975e-06,
+      "loss": 0.8109,
       "step": 29899
     },
     {
-      "epoch": 0.8211804125127022,
+      "epoch": 0.8484676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.6307942648079734e-06,
-      "loss": 0.885,
+      "learning_rate": 1.1803812170023688e-06,
+      "loss": 0.794,
       "step": 29900
     },
     {
-      "epoch": 0.8212078767405454,
+      "epoch": 0.8484960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 1.6303074441376231e-06,
-      "loss": 0.9211,
+      "learning_rate": 1.179948073692354e-06,
+      "loss": 0.9086,
       "step": 29901
     },
     {
-      "epoch": 0.8212353409683887,
+      "epoch": 0.8485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.6298206896915237e-06,
-      "loss": 0.8033,
+      "learning_rate": 1.1795150048857151e-06,
+      "loss": 0.7884,
       "step": 29902
     },
     {
-      "epoch": 0.821262805196232,
+      "epoch": 0.8485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.6293340014735249e-06,
-      "loss": 0.82,
+      "learning_rate": 1.1790820105861123e-06,
+      "loss": 0.8535,
       "step": 29903
     },
     {
-      "epoch": 0.8212902694240751,
+      "epoch": 0.8485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.6288473794874804e-06,
-      "loss": 0.8269,
+      "learning_rate": 1.178649090797197e-06,
+      "loss": 0.8665,
       "step": 29904
     },
     {
-      "epoch": 0.8213177336519184,
+      "epoch": 0.8486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.6283608237372384e-06,
-      "loss": 0.7481,
+      "learning_rate": 1.1782162455226299e-06,
+      "loss": 0.7978,
       "step": 29905
     },
     {
-      "epoch": 0.8213451978797616,
+      "epoch": 0.8486379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.6278743342266457e-06,
-      "loss": 0.8145,
+      "learning_rate": 1.1777834747660678e-06,
+      "loss": 0.8111,
       "step": 29906
     },
     {
-      "epoch": 0.8213726621076048,
+      "epoch": 0.84866628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.6273879109595537e-06,
-      "loss": 0.8576,
+      "learning_rate": 1.1773507785311645e-06,
+      "loss": 0.9698,
       "step": 29907
     },
     {
-      "epoch": 0.8214001263354481,
+      "epoch": 0.848694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.6269015539398126e-06,
-      "loss": 0.8044,
+      "learning_rate": 1.1769181568215749e-06,
+      "loss": 0.7578,
       "step": 29908
     },
     {
-      "epoch": 0.8214275905632913,
+      "epoch": 0.8487230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.6264152631712705e-06,
-      "loss": 0.7766,
+      "learning_rate": 1.176485609640955e-06,
+      "loss": 0.7918,
       "step": 29909
     },
     {
-      "epoch": 0.8214550547911346,
+      "epoch": 0.8487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.6259290386577754e-06,
-      "loss": 0.7838,
+      "learning_rate": 1.1760531369929573e-06,
+      "loss": 0.762,
       "step": 29910
     },
     {
-      "epoch": 0.8214825190189777,
+      "epoch": 0.8487797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.6254428804031718e-06,
-      "loss": 0.9449,
+      "learning_rate": 1.175620738881238e-06,
+      "loss": 0.7686,
       "step": 29911
     },
     {
-      "epoch": 0.821509983246821,
+      "epoch": 0.8488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.6249567884113082e-06,
-      "loss": 0.7997,
+      "learning_rate": 1.175188415309444e-06,
+      "loss": 0.8655,
       "step": 29912
     },
     {
-      "epoch": 0.8215374474746643,
+      "epoch": 0.8488365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.6244707626860335e-06,
-      "loss": 0.778,
+      "learning_rate": 1.1747561662812312e-06,
+      "loss": 0.8634,
       "step": 29913
     },
     {
-      "epoch": 0.8215649117025074,
+      "epoch": 0.8488649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.6239848032311867e-06,
-      "loss": 0.8708,
+      "learning_rate": 1.1743239918002513e-06,
+      "loss": 0.9096,
       "step": 29914
     },
     {
-      "epoch": 0.8215923759303507,
+      "epoch": 0.8488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.6234989100506181e-06,
-      "loss": 0.8226,
+      "learning_rate": 1.1738918918701515e-06,
+      "loss": 0.9243,
       "step": 29915
     },
     {
-      "epoch": 0.821619840158194,
+      "epoch": 0.8489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.6230130831481728e-06,
-      "loss": 0.7883,
+      "learning_rate": 1.173459866494583e-06,
+      "loss": 0.8118,
       "step": 29916
     },
     {
-      "epoch": 0.8216473043860372,
+      "epoch": 0.848950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.6225273225276904e-06,
-      "loss": 0.784,
+      "learning_rate": 1.1730279156771984e-06,
+      "loss": 0.7956,
       "step": 29917
     },
     {
-      "epoch": 0.8216747686138804,
+      "epoch": 0.8489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.6220416281930173e-06,
-      "loss": 0.8483,
+      "learning_rate": 1.1725960394216417e-06,
+      "loss": 0.6795,
       "step": 29918
     },
     {
-      "epoch": 0.8217022328417236,
+      "epoch": 0.8490068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.6215560001479958e-06,
-      "loss": 0.8214,
+      "learning_rate": 1.1721642377315635e-06,
+      "loss": 0.7954,
       "step": 29919
     },
     {
-      "epoch": 0.8217296970695669,
+      "epoch": 0.8490351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.6210704383964704e-06,
-      "loss": 0.7173,
+      "learning_rate": 1.1717325106106126e-06,
+      "loss": 0.7671,
       "step": 29920
     },
     {
-      "epoch": 0.8217571612974102,
+      "epoch": 0.8490635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.6205849429422803e-06,
-      "loss": 0.8212,
+      "learning_rate": 1.1713008580624319e-06,
+      "loss": 0.8286,
       "step": 29921
     },
     {
-      "epoch": 0.8217846255252533,
+      "epoch": 0.8490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.6200995137892695e-06,
-      "loss": 0.8666,
+      "learning_rate": 1.1708692800906696e-06,
+      "loss": 0.8535,
       "step": 29922
     },
     {
-      "epoch": 0.8218120897530966,
+      "epoch": 0.8491203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.6196141509412755e-06,
-      "loss": 0.786,
+      "learning_rate": 1.1704377766989727e-06,
+      "loss": 0.7405,
       "step": 29923
     },
     {
-      "epoch": 0.8218395539809398,
+      "epoch": 0.8491486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.6191288544021411e-06,
-      "loss": 0.8884,
+      "learning_rate": 1.1700063478909817e-06,
+      "loss": 0.7661,
       "step": 29924
     },
     {
-      "epoch": 0.821867018208783,
+      "epoch": 0.8491770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 1.6186436241757042e-06,
-      "loss": 0.8151,
+      "learning_rate": 1.1695749936703482e-06,
+      "loss": 0.8593,
       "step": 29925
     },
     {
-      "epoch": 0.8218944824366263,
+      "epoch": 0.849205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.6181584602658062e-06,
-      "loss": 0.8255,
+      "learning_rate": 1.1691437140407092e-06,
+      "loss": 0.8796,
       "step": 29926
     },
     {
-      "epoch": 0.8219219466644695,
+      "epoch": 0.8492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.6176733626762865e-06,
-      "loss": 0.8306,
+      "learning_rate": 1.1687125090057106e-06,
+      "loss": 0.9077,
       "step": 29927
     },
     {
-      "epoch": 0.8219494108923128,
+      "epoch": 0.8492622020431329,
       "grad_norm": 0.0,
-      "learning_rate": 1.6171883314109826e-06,
-      "loss": 0.9011,
+      "learning_rate": 1.1682813785689962e-06,
+      "loss": 0.8191,
       "step": 29928
     },
     {
-      "epoch": 0.821976875120156,
+      "epoch": 0.8492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.616703366473733e-06,
-      "loss": 0.7926,
+      "learning_rate": 1.1678503227342042e-06,
+      "loss": 0.8332,
       "step": 29929
     },
     {
-      "epoch": 0.8220043393479992,
+      "epoch": 0.8493189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.6162184678683712e-06,
-      "loss": 0.8889,
+      "learning_rate": 1.1674193415049773e-06,
+      "loss": 0.8463,
       "step": 29930
     },
     {
-      "epoch": 0.8220318035758425,
+      "epoch": 0.8493473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.6157336355987362e-06,
-      "loss": 0.781,
+      "learning_rate": 1.1669884348849593e-06,
+      "loss": 0.6398,
       "step": 29931
     },
     {
-      "epoch": 0.8220592678036857,
+      "epoch": 0.8493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.615248869668664e-06,
-      "loss": 0.9196,
+      "learning_rate": 1.1665576028777835e-06,
+      "loss": 0.8799,
       "step": 29932
     },
     {
-      "epoch": 0.8220867320315289,
+      "epoch": 0.8494040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.6147641700819915e-06,
-      "loss": 0.6888,
+      "learning_rate": 1.1661268454870933e-06,
+      "loss": 0.8831,
       "step": 29933
     },
     {
-      "epoch": 0.8221141962593722,
+      "epoch": 0.8494324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.6142795368425556e-06,
-      "loss": 0.8164,
+      "learning_rate": 1.1656961627165286e-06,
+      "loss": 0.8312,
       "step": 29934
     },
     {
-      "epoch": 0.8221416604872154,
+      "epoch": 0.8494608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 1.6137949699541856e-06,
-      "loss": 0.7526,
+      "learning_rate": 1.1652655545697244e-06,
+      "loss": 0.8716,
       "step": 29935
     },
     {
-      "epoch": 0.8221691247150587,
+      "epoch": 0.849489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.613310469420718e-06,
-      "loss": 0.7705,
+      "learning_rate": 1.164835021050318e-06,
+      "loss": 0.8115,
       "step": 29936
     },
     {
-      "epoch": 0.8221965889429018,
+      "epoch": 0.8495175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.612826035245989e-06,
-      "loss": 0.8316,
+      "learning_rate": 1.1644045621619504e-06,
+      "loss": 0.798,
       "step": 29937
     },
     {
-      "epoch": 0.8222240531707451,
+      "epoch": 0.8495459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.6123416674338277e-06,
-      "loss": 0.727,
+      "learning_rate": 1.1639741779082525e-06,
+      "loss": 0.8473,
       "step": 29938
     },
     {
-      "epoch": 0.8222515173985884,
+      "epoch": 0.8495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.6118573659880675e-06,
-      "loss": 0.7516,
+      "learning_rate": 1.1635438682928613e-06,
+      "loss": 0.8325,
       "step": 29939
     },
     {
-      "epoch": 0.8222789816264315,
+      "epoch": 0.8496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.6113731309125436e-06,
-      "loss": 0.8303,
+      "learning_rate": 1.1631136333194127e-06,
+      "loss": 0.8604,
       "step": 29940
     },
     {
-      "epoch": 0.8223064458542748,
+      "epoch": 0.8496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6108889622110824e-06,
-      "loss": 0.7147,
+      "learning_rate": 1.1626834729915404e-06,
+      "loss": 0.7995,
       "step": 29941
     },
     {
-      "epoch": 0.8223339100821181,
+      "epoch": 0.8496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.6104048598875178e-06,
-      "loss": 0.7799,
+      "learning_rate": 1.1622533873128771e-06,
+      "loss": 0.7253,
       "step": 29942
     },
     {
-      "epoch": 0.8223613743099613,
+      "epoch": 0.8496878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.6099208239456787e-06,
-      "loss": 0.8357,
+      "learning_rate": 1.16182337628706e-06,
+      "loss": 0.7571,
       "step": 29943
     },
     {
-      "epoch": 0.8223888385378045,
+      "epoch": 0.849716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.609436854389399e-06,
-      "loss": 0.8008,
+      "learning_rate": 1.1613934399177162e-06,
+      "loss": 0.7968,
       "step": 29944
     },
     {
-      "epoch": 0.8224163027656477,
+      "epoch": 0.849744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6089529512225034e-06,
-      "loss": 0.7832,
+      "learning_rate": 1.1609635782084804e-06,
+      "loss": 0.8135,
       "step": 29945
     },
     {
-      "epoch": 0.822443766993491,
+      "epoch": 0.8497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.608469114448824e-06,
-      "loss": 0.7892,
+      "learning_rate": 1.160533791162981e-06,
+      "loss": 0.8683,
       "step": 29946
     },
     {
-      "epoch": 0.8224712312213343,
+      "epoch": 0.8498013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.607985344072186e-06,
-      "loss": 0.7948,
+      "learning_rate": 1.1601040787848504e-06,
+      "loss": 0.8228,
       "step": 29947
     },
     {
-      "epoch": 0.8224986954491774,
+      "epoch": 0.8498297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.6075016400964183e-06,
-      "loss": 0.841,
+      "learning_rate": 1.1596744410777206e-06,
+      "loss": 0.6974,
       "step": 29948
     },
     {
-      "epoch": 0.8225261596770207,
+      "epoch": 0.8498581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.607018002525348e-06,
-      "loss": 0.7659,
+      "learning_rate": 1.1592448780452148e-06,
+      "loss": 0.8779,
       "step": 29949
     },
     {
-      "epoch": 0.8225536239048639,
+      "epoch": 0.8498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.606534431362803e-06,
-      "loss": 0.7706,
+      "learning_rate": 1.158815389690966e-06,
+      "loss": 0.7973,
       "step": 29950
     },
     {
-      "epoch": 0.8225810881327071,
+      "epoch": 0.8499148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.606050926612609e-06,
-      "loss": 0.8136,
+      "learning_rate": 1.1583859760186023e-06,
+      "loss": 0.7579,
       "step": 29951
     },
     {
-      "epoch": 0.8226085523605504,
+      "epoch": 0.8499432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.6055674882785933e-06,
-      "loss": 0.6844,
+      "learning_rate": 1.1579566370317473e-06,
+      "loss": 0.9463,
       "step": 29952
     },
     {
-      "epoch": 0.8226360165883936,
+      "epoch": 0.8499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.6050841163645802e-06,
-      "loss": 0.7879,
+      "learning_rate": 1.1575273727340308e-06,
+      "loss": 0.7221,
       "step": 29953
     },
     {
-      "epoch": 0.8226634808162369,
+      "epoch": 0.85,
       "grad_norm": 0.0,
-      "learning_rate": 1.6046008108743905e-06,
-      "loss": 0.7717,
+      "learning_rate": 1.1570981831290806e-06,
+      "loss": 0.8031,
       "step": 29954
     },
     {
-      "epoch": 0.8226909450440801,
+      "epoch": 0.8500283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 1.6041175718118507e-06,
-      "loss": 0.8178,
+      "learning_rate": 1.156669068220515e-06,
+      "loss": 0.8926,
       "step": 29955
     },
     {
-      "epoch": 0.8227184092719233,
+      "epoch": 0.8500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.6036343991807856e-06,
-      "loss": 0.795,
+      "learning_rate": 1.1562400280119656e-06,
+      "loss": 0.6935,
       "step": 29956
     },
     {
-      "epoch": 0.8227458734997666,
+      "epoch": 0.8500851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.6031512929850179e-06,
-      "loss": 0.7552,
+      "learning_rate": 1.155811062507055e-06,
+      "loss": 0.7206,
       "step": 29957
     },
     {
-      "epoch": 0.8227733377276097,
+      "epoch": 0.8501135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.6026682532283722e-06,
-      "loss": 0.825,
+      "learning_rate": 1.155382171709405e-06,
+      "loss": 0.7319,
       "step": 29958
     },
     {
-      "epoch": 0.822800801955453,
+      "epoch": 0.8501418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.6021852799146653e-06,
-      "loss": 0.7953,
+      "learning_rate": 1.15495335562264e-06,
+      "loss": 0.7572,
       "step": 29959
     },
     {
-      "epoch": 0.8228282661832963,
+      "epoch": 0.8501702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.601702373047721e-06,
-      "loss": 0.8823,
+      "learning_rate": 1.154524614250383e-06,
+      "loss": 0.7631,
       "step": 29960
     },
     {
-      "epoch": 0.8228557304111395,
+      "epoch": 0.8501986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.6012195326313639e-06,
-      "loss": 0.827,
+      "learning_rate": 1.154095947596252e-06,
+      "loss": 0.8611,
       "step": 29961
     },
     {
-      "epoch": 0.8228831946389827,
+      "epoch": 0.8502270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 1.600736758669409e-06,
-      "loss": 0.7752,
+      "learning_rate": 1.15366735566387e-06,
+      "loss": 0.7203,
       "step": 29962
     },
     {
-      "epoch": 0.8229106588668259,
+      "epoch": 0.850255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.6002540511656795e-06,
-      "loss": 0.7662,
+      "learning_rate": 1.1532388384568594e-06,
+      "loss": 0.7823,
       "step": 29963
     },
     {
-      "epoch": 0.8229381230946692,
+      "epoch": 0.8502837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.5997714101239925e-06,
-      "loss": 0.8522,
+      "learning_rate": 1.1528103959788373e-06,
+      "loss": 0.8461,
       "step": 29964
     },
     {
-      "epoch": 0.8229655873225125,
+      "epoch": 0.8503121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.5992888355481673e-06,
-      "loss": 0.8442,
+      "learning_rate": 1.152382028233422e-06,
+      "loss": 0.7433,
       "step": 29965
     },
     {
-      "epoch": 0.8229930515503556,
+      "epoch": 0.8503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.598806327442024e-06,
-      "loss": 0.7602,
+      "learning_rate": 1.151953735224236e-06,
+      "loss": 0.7355,
       "step": 29966
     },
     {
-      "epoch": 0.8230205157781989,
+      "epoch": 0.8503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.5983238858093787e-06,
-      "loss": 0.794,
+      "learning_rate": 1.1515255169548934e-06,
+      "loss": 0.8443,
       "step": 29967
     },
     {
-      "epoch": 0.8230479800060422,
+      "epoch": 0.8503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.5978415106540513e-06,
-      "loss": 0.8085,
+      "learning_rate": 1.151097373429011e-06,
+      "loss": 0.7827,
       "step": 29968
     },
     {
-      "epoch": 0.8230754442338853,
+      "epoch": 0.8504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.5973592019798577e-06,
-      "loss": 0.8846,
+      "learning_rate": 1.1506693046502093e-06,
+      "loss": 0.8774,
       "step": 29969
     },
     {
-      "epoch": 0.8231029084617286,
+      "epoch": 0.8504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.5968769597906098e-06,
-      "loss": 0.823,
+      "learning_rate": 1.1502413106221012e-06,
+      "loss": 0.8426,
       "step": 29970
     },
     {
-      "epoch": 0.8231303726895718,
+      "epoch": 0.8504824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 1.5963947840901262e-06,
-      "loss": 0.821,
+      "learning_rate": 1.1498133913483012e-06,
+      "loss": 0.7563,
       "step": 29971
     },
     {
-      "epoch": 0.8231578369174151,
+      "epoch": 0.850510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.595912674882223e-06,
-      "loss": 0.9268,
+      "learning_rate": 1.1493855468324255e-06,
+      "loss": 0.867,
       "step": 29972
     },
     {
-      "epoch": 0.8231853011452583,
+      "epoch": 0.850539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.5954306321707135e-06,
-      "loss": 0.8093,
+      "learning_rate": 1.148957777078088e-06,
+      "loss": 0.8642,
       "step": 29973
     },
     {
-      "epoch": 0.8232127653731015,
+      "epoch": 0.8505675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.594948655959413e-06,
-      "loss": 0.7368,
+      "learning_rate": 1.1485300820889035e-06,
+      "loss": 0.8541,
       "step": 29974
     },
     {
-      "epoch": 0.8232402296009448,
+      "epoch": 0.8505959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.5944667462521367e-06,
-      "loss": 0.8455,
+      "learning_rate": 1.1481024618684821e-06,
+      "loss": 0.8094,
       "step": 29975
     },
     {
-      "epoch": 0.823267693828788,
+      "epoch": 0.8506242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 1.593984903052692e-06,
-      "loss": 0.8338,
+      "learning_rate": 1.147674916420437e-06,
+      "loss": 0.8384,
       "step": 29976
     },
     {
-      "epoch": 0.8232951580566312,
+      "epoch": 0.8506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.5935031263648992e-06,
-      "loss": 0.8149,
+      "learning_rate": 1.1472474457483818e-06,
+      "loss": 0.8222,
       "step": 29977
     },
     {
-      "epoch": 0.8233226222844745,
+      "epoch": 0.8506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.5930214161925628e-06,
-      "loss": 0.7123,
+      "learning_rate": 1.1468200498559234e-06,
+      "loss": 0.8436,
       "step": 29978
     },
     {
-      "epoch": 0.8233500865123177,
+      "epoch": 0.8507094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.5925397725394976e-06,
-      "loss": 0.8183,
+      "learning_rate": 1.1463927287466736e-06,
+      "loss": 0.87,
       "step": 29979
     },
     {
-      "epoch": 0.823377550740161,
+      "epoch": 0.8507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.5920581954095138e-06,
-      "loss": 0.8262,
+      "learning_rate": 1.145965482424245e-06,
+      "loss": 0.8453,
       "step": 29980
     },
     {
-      "epoch": 0.8234050149680042,
+      "epoch": 0.8507661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.5915766848064262e-06,
-      "loss": 0.8258,
+      "learning_rate": 1.1455383108922414e-06,
+      "loss": 0.9001,
       "step": 29981
     },
     {
-      "epoch": 0.8234324791958474,
+      "epoch": 0.8507945516458569,
       "grad_norm": 0.0,
-      "learning_rate": 1.5910952407340375e-06,
-      "loss": 0.7293,
+      "learning_rate": 1.1451112141542742e-06,
+      "loss": 0.7484,
       "step": 29982
     },
     {
-      "epoch": 0.8234599434236907,
+      "epoch": 0.8508229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.590613863196161e-06,
-      "loss": 0.8306,
+      "learning_rate": 1.144684192213953e-06,
+      "loss": 0.8969,
       "step": 29983
     },
     {
-      "epoch": 0.8234874076515338,
+      "epoch": 0.8508513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.590132552196607e-06,
-      "loss": 0.788,
+      "learning_rate": 1.14425724507488e-06,
+      "loss": 0.8552,
       "step": 29984
     },
     {
-      "epoch": 0.8235148718793771,
+      "epoch": 0.8508796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.5896513077391805e-06,
-      "loss": 0.8347,
+      "learning_rate": 1.143830372740664e-06,
+      "loss": 0.8023,
       "step": 29985
     },
     {
-      "epoch": 0.8235423361072204,
+      "epoch": 0.8509080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.5891701298276896e-06,
-      "loss": 0.8527,
+      "learning_rate": 1.1434035752149108e-06,
+      "loss": 0.8357,
       "step": 29986
     },
     {
-      "epoch": 0.8235698003350636,
+      "epoch": 0.8509364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.5886890184659453e-06,
-      "loss": 0.8033,
+      "learning_rate": 1.1429768525012264e-06,
+      "loss": 0.9157,
       "step": 29987
     },
     {
-      "epoch": 0.8235972645629068,
+      "epoch": 0.8509648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.5882079736577504e-06,
-      "loss": 0.889,
+      "learning_rate": 1.1425502046032154e-06,
+      "loss": 0.758,
       "step": 29988
     },
     {
-      "epoch": 0.82362472879075,
+      "epoch": 0.8509931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.587726995406912e-06,
-      "loss": 0.8133,
+      "learning_rate": 1.14212363152448e-06,
+      "loss": 0.898,
       "step": 29989
     },
     {
-      "epoch": 0.8236521930185933,
+      "epoch": 0.8510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.5872460837172355e-06,
-      "loss": 0.8692,
+      "learning_rate": 1.1416971332686243e-06,
+      "loss": 0.8428,
       "step": 29990
     },
     {
-      "epoch": 0.8236796572464365,
+      "epoch": 0.851049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.586765238592527e-06,
-      "loss": 0.8067,
+      "learning_rate": 1.141270709839254e-06,
+      "loss": 0.8324,
       "step": 29991
     },
     {
-      "epoch": 0.8237071214742797,
+      "epoch": 0.8510783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.5862844600365934e-06,
-      "loss": 0.7976,
+      "learning_rate": 1.1408443612399655e-06,
+      "loss": 0.6615,
       "step": 29992
     },
     {
-      "epoch": 0.823734585702123,
+      "epoch": 0.8511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.5858037480532362e-06,
-      "loss": 0.8789,
+      "learning_rate": 1.140418087474362e-06,
+      "loss": 0.7568,
       "step": 29993
     },
     {
-      "epoch": 0.8237620499299663,
+      "epoch": 0.8511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.5853231026462567e-06,
-      "loss": 0.8156,
+      "learning_rate": 1.1399918885460482e-06,
+      "loss": 0.8383,
       "step": 29994
     },
     {
-      "epoch": 0.8237895141578094,
+      "epoch": 0.8511634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.584842523819461e-06,
-      "loss": 0.8569,
+      "learning_rate": 1.139565764458619e-06,
+      "loss": 0.8168,
       "step": 29995
     },
     {
-      "epoch": 0.8238169783856527,
+      "epoch": 0.8511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.58436201157665e-06,
-      "loss": 0.8343,
+      "learning_rate": 1.1391397152156768e-06,
+      "loss": 0.9777,
       "step": 29996
     },
     {
-      "epoch": 0.8238444426134959,
+      "epoch": 0.8512202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.5838815659216279e-06,
-      "loss": 0.8235,
+      "learning_rate": 1.1387137408208203e-06,
+      "loss": 0.8475,
       "step": 29997
     },
     {
-      "epoch": 0.8238719068413392,
+      "epoch": 0.8512485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.5834011868581933e-06,
-      "loss": 0.7849,
+      "learning_rate": 1.1382878412776466e-06,
+      "loss": 0.811,
       "step": 29998
     },
     {
-      "epoch": 0.8238993710691824,
+      "epoch": 0.8512769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.5829208743901525e-06,
-      "loss": 0.7043,
+      "learning_rate": 1.1378620165897537e-06,
+      "loss": 0.8005,
       "step": 29999
     },
     {
-      "epoch": 0.8239268352970256,
+      "epoch": 0.8513053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.5824406285212989e-06,
-      "loss": 0.7113,
+      "learning_rate": 1.1374362667607397e-06,
+      "loss": 0.8486,
       "step": 30000
     },
     {
-      "epoch": 0.8239542995248689,
+      "epoch": 0.8513337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.5819604492554385e-06,
-      "loss": 0.8444,
+      "learning_rate": 1.1370105917941966e-06,
+      "loss": 0.7732,
       "step": 30001
     },
     {
-      "epoch": 0.823981763752712,
+      "epoch": 0.8513620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.581480336596366e-06,
-      "loss": 0.743,
+      "learning_rate": 1.1365849916937277e-06,
+      "loss": 0.8043,
       "step": 30002
     },
     {
-      "epoch": 0.8240092279805553,
+      "epoch": 0.8513904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 1.5810002905478816e-06,
-      "loss": 0.8029,
+      "learning_rate": 1.1361594664629206e-06,
+      "loss": 0.8915,
       "step": 30003
     },
     {
-      "epoch": 0.8240366922083986,
+      "epoch": 0.8514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.5805203111137845e-06,
-      "loss": 0.9018,
+      "learning_rate": 1.1357340161053742e-06,
+      "loss": 0.8869,
       "step": 30004
     },
     {
-      "epoch": 0.8240641564362418,
+      "epoch": 0.8514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.580040398297874e-06,
-      "loss": 0.7783,
+      "learning_rate": 1.1353086406246827e-06,
+      "loss": 0.8106,
       "step": 30005
     },
     {
-      "epoch": 0.824091620664085,
+      "epoch": 0.8514755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.5795605521039436e-06,
-      "loss": 0.8972,
+      "learning_rate": 1.1348833400244353e-06,
+      "loss": 0.7887,
       "step": 30006
     },
     {
-      "epoch": 0.8241190848919283,
+      "epoch": 0.8515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.5790807725357915e-06,
-      "loss": 0.8202,
+      "learning_rate": 1.1344581143082279e-06,
+      "loss": 0.7877,
       "step": 30007
     },
     {
-      "epoch": 0.8241465491197715,
+      "epoch": 0.8515323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.5786010595972156e-06,
-      "loss": 0.9152,
+      "learning_rate": 1.134032963479652e-06,
+      "loss": 0.7962,
       "step": 30008
     },
     {
-      "epoch": 0.8241740133476148,
+      "epoch": 0.851560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.5781214132920098e-06,
-      "loss": 0.8893,
+      "learning_rate": 1.1336078875422973e-06,
+      "loss": 0.7975,
       "step": 30009
     },
     {
-      "epoch": 0.8242014775754579,
+      "epoch": 0.851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5776418336239685e-06,
-      "loss": 0.7794,
+      "learning_rate": 1.133182886499754e-06,
+      "loss": 0.8225,
       "step": 30010
     },
     {
-      "epoch": 0.8242289418033012,
+      "epoch": 0.8516174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.57716232059689e-06,
-      "loss": 0.802,
+      "learning_rate": 1.1327579603556172e-06,
+      "loss": 0.8351,
       "step": 30011
     },
     {
-      "epoch": 0.8242564060311445,
+      "epoch": 0.8516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.5766828742145634e-06,
-      "loss": 0.7846,
+      "learning_rate": 1.1323331091134693e-06,
+      "loss": 0.7404,
       "step": 30012
     },
     {
-      "epoch": 0.8242838702589876,
+      "epoch": 0.8516742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 1.5762034944807859e-06,
-      "loss": 0.8288,
+      "learning_rate": 1.131908332776902e-06,
+      "loss": 0.7704,
       "step": 30013
     },
     {
-      "epoch": 0.8243113344868309,
+      "epoch": 0.8517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.5757241813993485e-06,
-      "loss": 0.768,
+      "learning_rate": 1.131483631349507e-06,
+      "loss": 0.8096,
       "step": 30014
     },
     {
-      "epoch": 0.8243387987146742,
+      "epoch": 0.8517309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.575244934974045e-06,
-      "loss": 0.801,
+      "learning_rate": 1.1310590048348647e-06,
+      "loss": 0.797,
       "step": 30015
     },
     {
-      "epoch": 0.8243662629425174,
+      "epoch": 0.8517593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.5747657552086692e-06,
-      "loss": 0.7503,
+      "learning_rate": 1.1306344532365665e-06,
+      "loss": 0.7616,
       "step": 30016
     },
     {
-      "epoch": 0.8243937271703606,
+      "epoch": 0.8517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.5742866421070103e-06,
-      "loss": 0.8952,
+      "learning_rate": 1.1302099765581987e-06,
+      "loss": 0.7809,
       "step": 30017
     },
     {
-      "epoch": 0.8244211913982038,
+      "epoch": 0.8518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.5738075956728571e-06,
-      "loss": 0.8009,
+      "learning_rate": 1.1297855748033448e-06,
+      "loss": 0.7907,
       "step": 30018
     },
     {
-      "epoch": 0.8244486556260471,
+      "epoch": 0.851844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.5733286159100013e-06,
-      "loss": 0.8684,
+      "learning_rate": 1.1293612479755934e-06,
+      "loss": 0.7799,
       "step": 30019
     },
     {
-      "epoch": 0.8244761198538904,
+      "epoch": 0.8518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.5728497028222345e-06,
-      "loss": 0.7719,
+      "learning_rate": 1.1289369960785234e-06,
+      "loss": 0.901,
       "step": 30020
     },
     {
-      "epoch": 0.8245035840817335,
+      "epoch": 0.8519012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.5723708564133454e-06,
-      "loss": 0.8329,
+      "learning_rate": 1.1285128191157224e-06,
+      "loss": 0.7927,
       "step": 30021
     },
     {
-      "epoch": 0.8245310483095768,
+      "epoch": 0.8519296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.5718920766871226e-06,
-      "loss": 0.8837,
+      "learning_rate": 1.1280887170907739e-06,
+      "loss": 0.8136,
       "step": 30022
     },
     {
-      "epoch": 0.82455851253742,
+      "epoch": 0.8519580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.5714133636473561e-06,
-      "loss": 0.7321,
+      "learning_rate": 1.1276646900072574e-06,
+      "loss": 0.8043,
       "step": 30023
     },
     {
-      "epoch": 0.8245859767652632,
+      "epoch": 0.8519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.5709347172978307e-06,
-      "loss": 0.8723,
+      "learning_rate": 1.1272407378687556e-06,
+      "loss": 0.8326,
       "step": 30024
     },
     {
-      "epoch": 0.8246134409931065,
+      "epoch": 0.8520147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.5704561376423367e-06,
-      "loss": 0.8061,
+      "learning_rate": 1.1268168606788532e-06,
+      "loss": 0.8038,
       "step": 30025
     },
     {
-      "epoch": 0.8246409052209497,
+      "epoch": 0.8520431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.5699776246846565e-06,
-      "loss": 0.8772,
+      "learning_rate": 1.1263930584411243e-06,
+      "loss": 0.7544,
       "step": 30026
     },
     {
-      "epoch": 0.824668369448793,
+      "epoch": 0.8520715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 1.5694991784285785e-06,
-      "loss": 0.8148,
+      "learning_rate": 1.1259693311591523e-06,
+      "loss": 0.8928,
       "step": 30027
     },
     {
-      "epoch": 0.8246958336766362,
+      "epoch": 0.852099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.5690207988778916e-06,
-      "loss": 0.8003,
+      "learning_rate": 1.1255456788365193e-06,
+      "loss": 0.8204,
       "step": 30028
     },
     {
-      "epoch": 0.8247232979044794,
+      "epoch": 0.8521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.5685424860363763e-06,
-      "loss": 0.8282,
+      "learning_rate": 1.1251221014767977e-06,
+      "loss": 0.8518,
       "step": 30029
     },
     {
-      "epoch": 0.8247507621323227,
+      "epoch": 0.8521566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 1.5680642399078171e-06,
-      "loss": 0.6966,
+      "learning_rate": 1.124698599083569e-06,
+      "loss": 0.7625,
       "step": 30030
     },
     {
-      "epoch": 0.8247782263601658,
+      "epoch": 0.8521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.5675860604960014e-06,
-      "loss": 0.9113,
+      "learning_rate": 1.1242751716604128e-06,
+      "loss": 0.8601,
       "step": 30031
     },
     {
-      "epoch": 0.8248056905880091,
+      "epoch": 0.8522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.5671079478047123e-06,
-      "loss": 0.7262,
+      "learning_rate": 1.1238518192108983e-06,
+      "loss": 0.7817,
       "step": 30032
     },
     {
-      "epoch": 0.8248331548158524,
+      "epoch": 0.8522417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.5666299018377307e-06,
-      "loss": 0.7319,
+      "learning_rate": 1.1234285417386115e-06,
+      "loss": 0.7102,
       "step": 30033
     },
     {
-      "epoch": 0.8248606190436956,
+      "epoch": 0.8522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.566151922598842e-06,
-      "loss": 0.8257,
+      "learning_rate": 1.1230053392471196e-06,
+      "loss": 0.7669,
       "step": 30034
     },
     {
-      "epoch": 0.8248880832715388,
+      "epoch": 0.8522985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.5656740100918233e-06,
-      "loss": 0.848,
+      "learning_rate": 1.122582211740002e-06,
+      "loss": 0.8001,
       "step": 30035
     },
     {
-      "epoch": 0.824915547499382,
+      "epoch": 0.8523269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.5651961643204594e-06,
-      "loss": 0.795,
+      "learning_rate": 1.1221591592208325e-06,
+      "loss": 0.7275,
       "step": 30036
     },
     {
-      "epoch": 0.8249430117272253,
+      "epoch": 0.8523552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 1.5647183852885307e-06,
-      "loss": 0.8539,
+      "learning_rate": 1.1217361816931815e-06,
+      "loss": 0.7807,
       "step": 30037
     },
     {
-      "epoch": 0.8249704759550686,
+      "epoch": 0.852383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.5642406729998184e-06,
-      "loss": 0.8627,
+      "learning_rate": 1.1213132791606251e-06,
+      "loss": 0.7243,
       "step": 30038
     },
     {
-      "epoch": 0.8249979401829117,
+      "epoch": 0.8524120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.5637630274581006e-06,
-      "loss": 0.7753,
+      "learning_rate": 1.1208904516267361e-06,
+      "loss": 0.8185,
       "step": 30039
     },
     {
-      "epoch": 0.825025404410755,
+      "epoch": 0.8524404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.56328544866716e-06,
-      "loss": 0.8043,
+      "learning_rate": 1.1204676990950826e-06,
+      "loss": 0.7972,
       "step": 30040
     },
     {
-      "epoch": 0.8250528686385983,
+      "epoch": 0.8524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.5628079366307737e-06,
-      "loss": 0.7341,
+      "learning_rate": 1.120045021569237e-06,
+      "loss": 0.7875,
       "step": 30041
     },
     {
-      "epoch": 0.8250803328664414,
+      "epoch": 0.8524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.562330491352716e-06,
-      "loss": 0.7948,
+      "learning_rate": 1.1196224190527737e-06,
+      "loss": 0.726,
       "step": 30042
     },
     {
-      "epoch": 0.8251077970942847,
+      "epoch": 0.8525255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.5618531128367686e-06,
-      "loss": 0.777,
+      "learning_rate": 1.1191998915492553e-06,
+      "loss": 0.9319,
       "step": 30043
     },
     {
-      "epoch": 0.8251352613221279,
+      "epoch": 0.8525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.5613758010867076e-06,
-      "loss": 0.826,
+      "learning_rate": 1.1187774390622563e-06,
+      "loss": 0.8593,
       "step": 30044
     },
     {
-      "epoch": 0.8251627255499712,
+      "epoch": 0.8525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.5608985561063105e-06,
-      "loss": 0.843,
+      "learning_rate": 1.1183550615953442e-06,
+      "loss": 0.6944,
       "step": 30045
     },
     {
-      "epoch": 0.8251901897778144,
+      "epoch": 0.85261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.5604213778993549e-06,
-      "loss": 0.8684,
+      "learning_rate": 1.117932759152085e-06,
+      "loss": 0.6451,
       "step": 30046
     },
     {
-      "epoch": 0.8252176540056576,
+      "epoch": 0.852639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.559944266469613e-06,
-      "loss": 0.8368,
+      "learning_rate": 1.1175105317360446e-06,
+      "loss": 0.8126,
       "step": 30047
     },
     {
-      "epoch": 0.8252451182335009,
+      "epoch": 0.8526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.5594672218208605e-06,
-      "loss": 0.7172,
+      "learning_rate": 1.1170883793507959e-06,
+      "loss": 0.862,
       "step": 30048
     },
     {
-      "epoch": 0.825272582461344,
+      "epoch": 0.8526958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.558990243956875e-06,
-      "loss": 0.8794,
+      "learning_rate": 1.1166663019998992e-06,
+      "loss": 0.8186,
       "step": 30049
     },
     {
-      "epoch": 0.8253000466891873,
+      "epoch": 0.8527241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.5585133328814272e-06,
-      "loss": 0.7146,
+      "learning_rate": 1.1162442996869216e-06,
+      "loss": 0.7159,
       "step": 30050
     },
     {
-      "epoch": 0.8253275109170306,
+      "epoch": 0.8527525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.5580364885982913e-06,
-      "loss": 0.7963,
+      "learning_rate": 1.1158223724154305e-06,
+      "loss": 0.7714,
       "step": 30051
     },
     {
-      "epoch": 0.8253549751448738,
+      "epoch": 0.8527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.5575597111112427e-06,
-      "loss": 0.8019,
+      "learning_rate": 1.115400520188984e-06,
+      "loss": 0.7541,
       "step": 30052
     },
     {
-      "epoch": 0.825382439372717,
+      "epoch": 0.8528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.55708300042405e-06,
-      "loss": 0.8454,
+      "learning_rate": 1.1149787430111514e-06,
+      "loss": 0.8529,
       "step": 30053
     },
     {
-      "epoch": 0.8254099036005603,
+      "epoch": 0.8528376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.5566063565404877e-06,
-      "loss": 0.846,
+      "learning_rate": 1.11455704088549e-06,
+      "loss": 0.7297,
       "step": 30054
     },
     {
-      "epoch": 0.8254373678284035,
+      "epoch": 0.8528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.556129779464326e-06,
-      "loss": 0.7934,
+      "learning_rate": 1.1141354138155658e-06,
+      "loss": 0.8934,
       "step": 30055
     },
     {
-      "epoch": 0.8254648320562468,
+      "epoch": 0.852894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.5556532691993386e-06,
-      "loss": 0.8074,
+      "learning_rate": 1.1137138618049403e-06,
+      "loss": 0.7665,
       "step": 30056
     },
     {
-      "epoch": 0.8254922962840899,
+      "epoch": 0.8529228149829738,
       "grad_norm": 0.0,
-      "learning_rate": 1.5551768257492905e-06,
-      "loss": 0.9427,
+      "learning_rate": 1.113292384857172e-06,
+      "loss": 0.7938,
       "step": 30057
     },
     {
-      "epoch": 0.8255197605119332,
+      "epoch": 0.8529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.5547004491179584e-06,
-      "loss": 0.8042,
+      "learning_rate": 1.1128709829758221e-06,
+      "loss": 0.8748,
       "step": 30058
     },
     {
-      "epoch": 0.8255472247397765,
+      "epoch": 0.8529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.554224139309104e-06,
-      "loss": 0.7724,
+      "learning_rate": 1.1124496561644526e-06,
+      "loss": 0.7451,
       "step": 30059
     },
     {
-      "epoch": 0.8255746889676197,
+      "epoch": 0.8530079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.5537478963265006e-06,
-      "loss": 0.7759,
+      "learning_rate": 1.1120284044266183e-06,
+      "loss": 0.8809,
       "step": 30060
     },
     {
-      "epoch": 0.8256021531954629,
+      "epoch": 0.8530363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.5532717201739155e-06,
-      "loss": 0.8238,
+      "learning_rate": 1.1116072277658795e-06,
+      "loss": 0.717,
       "step": 30061
     },
     {
-      "epoch": 0.8256296174233061,
+      "epoch": 0.8530646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.5527956108551168e-06,
-      "loss": 0.73,
+      "learning_rate": 1.111186126185796e-06,
+      "loss": 0.7908,
       "step": 30062
     },
     {
-      "epoch": 0.8256570816511494,
+      "epoch": 0.8530930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.5523195683738712e-06,
-      "loss": 0.8273,
+      "learning_rate": 1.1107650996899189e-06,
+      "loss": 0.8497,
       "step": 30063
     },
     {
-      "epoch": 0.8256845458789926,
+      "epoch": 0.8531214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.5518435927339471e-06,
-      "loss": 0.7159,
+      "learning_rate": 1.11034414828181e-06,
+      "loss": 0.8261,
       "step": 30064
     },
     {
-      "epoch": 0.8257120101068358,
+      "epoch": 0.853149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.5513676839391102e-06,
-      "loss": 0.7423,
+      "learning_rate": 1.1099232719650265e-06,
+      "loss": 0.793,
       "step": 30065
     },
     {
-      "epoch": 0.8257394743346791,
+      "epoch": 0.8531782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.5508918419931218e-06,
-      "loss": 0.8702,
+      "learning_rate": 1.109502470743119e-06,
+      "loss": 0.7743,
       "step": 30066
     },
     {
-      "epoch": 0.8257669385625224,
+      "epoch": 0.8532065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.5504160668997504e-06,
-      "loss": 0.891,
+      "learning_rate": 1.1090817446196433e-06,
+      "loss": 0.8425,
       "step": 30067
     },
     {
-      "epoch": 0.8257944027903655,
+      "epoch": 0.8532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.5499403586627593e-06,
-      "loss": 0.822,
+      "learning_rate": 1.1086610935981557e-06,
+      "loss": 0.7569,
       "step": 30068
     },
     {
-      "epoch": 0.8258218670182088,
+      "epoch": 0.8532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.5494647172859145e-06,
-      "loss": 0.7877,
+      "learning_rate": 1.1082405176822054e-06,
+      "loss": 0.7518,
       "step": 30069
     },
     {
-      "epoch": 0.825849331246052,
+      "epoch": 0.8532917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.5489891427729797e-06,
-      "loss": 0.8981,
+      "learning_rate": 1.1078200168753473e-06,
+      "loss": 0.9122,
       "step": 30070
     },
     {
-      "epoch": 0.8258767954738953,
+      "epoch": 0.8533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.548513635127714e-06,
-      "loss": 0.7927,
+      "learning_rate": 1.1073995911811364e-06,
+      "loss": 0.8186,
       "step": 30071
     },
     {
-      "epoch": 0.8259042597017385,
+      "epoch": 0.8533484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.5480381943538824e-06,
-      "loss": 0.7343,
+      "learning_rate": 1.1069792406031178e-06,
+      "loss": 0.8055,
       "step": 30072
     },
     {
-      "epoch": 0.8259317239295817,
+      "epoch": 0.8533768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.5475628204552495e-06,
-      "loss": 0.8205,
+      "learning_rate": 1.106558965144845e-06,
+      "loss": 0.8174,
       "step": 30073
     },
     {
-      "epoch": 0.825959188157425,
+      "epoch": 0.853405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.5470875134355711e-06,
-      "loss": 0.8368,
+      "learning_rate": 1.106138764809871e-06,
+      "loss": 0.8075,
       "step": 30074
     },
     {
-      "epoch": 0.8259866523852681,
+      "epoch": 0.853433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.54661227329861e-06,
-      "loss": 0.8972,
+      "learning_rate": 1.1057186396017405e-06,
+      "loss": 0.7993,
       "step": 30075
     },
     {
-      "epoch": 0.8260141166131114,
+      "epoch": 0.8534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.5461371000481308e-06,
-      "loss": 0.8716,
+      "learning_rate": 1.1052985895240043e-06,
+      "loss": 0.8266,
       "step": 30076
     },
     {
-      "epoch": 0.8260415808409547,
+      "epoch": 0.8534903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.5456619936878858e-06,
-      "loss": 0.7725,
+      "learning_rate": 1.1048786145802126e-06,
+      "loss": 0.6706,
       "step": 30077
     },
     {
-      "epoch": 0.8260690450687979,
+      "epoch": 0.8535187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.545186954221639e-06,
-      "loss": 0.7205,
+      "learning_rate": 1.1044587147739072e-06,
+      "loss": 0.7284,
       "step": 30078
     },
     {
-      "epoch": 0.8260965092966411,
+      "epoch": 0.8535471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.5447119816531475e-06,
-      "loss": 0.8635,
+      "learning_rate": 1.104038890108644e-06,
+      "loss": 0.7702,
       "step": 30079
     },
     {
-      "epoch": 0.8261239735244844,
+      "epoch": 0.8535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.5442370759861724e-06,
-      "loss": 0.8967,
+      "learning_rate": 1.1036191405879614e-06,
+      "loss": 0.8391,
       "step": 30080
     },
     {
-      "epoch": 0.8261514377523276,
+      "epoch": 0.8536038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.5437622372244676e-06,
-      "loss": 0.783,
+      "learning_rate": 1.1031994662154077e-06,
+      "loss": 0.7757,
       "step": 30081
     },
     {
-      "epoch": 0.8261789019801709,
+      "epoch": 0.8536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.5432874653717933e-06,
-      "loss": 0.7607,
+      "learning_rate": 1.10277986699453e-06,
+      "loss": 0.7732,
       "step": 30082
     },
     {
-      "epoch": 0.826206366208014,
+      "epoch": 0.8536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.542812760431901e-06,
-      "loss": 0.7971,
+      "learning_rate": 1.1023603429288688e-06,
+      "loss": 0.817,
       "step": 30083
     },
     {
-      "epoch": 0.8262338304358573,
+      "epoch": 0.8536889897843359,
       "grad_norm": 0.0,
-      "learning_rate": 1.542338122408552e-06,
-      "loss": 0.7183,
+      "learning_rate": 1.10194089402197e-06,
+      "loss": 0.7796,
       "step": 30084
     },
     {
-      "epoch": 0.8262612946637006,
+      "epoch": 0.8537173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.5418635513054992e-06,
-      "loss": 0.904,
+      "learning_rate": 1.1015215202773778e-06,
+      "loss": 0.8363,
       "step": 30085
     },
     {
-      "epoch": 0.8262887588915437,
+      "epoch": 0.8537457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.541389047126498e-06,
-      "loss": 0.8384,
+      "learning_rate": 1.1011022216986322e-06,
+      "loss": 0.8947,
       "step": 30086
     },
     {
-      "epoch": 0.826316223119387,
+      "epoch": 0.8537741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.5409146098753057e-06,
-      "loss": 0.7408,
+      "learning_rate": 1.1006829982892754e-06,
+      "loss": 0.7813,
       "step": 30087
     },
     {
-      "epoch": 0.8263436873472302,
+      "epoch": 0.8538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.5404402395556717e-06,
-      "loss": 0.7965,
+      "learning_rate": 1.1002638500528517e-06,
+      "loss": 0.8413,
       "step": 30088
     },
     {
-      "epoch": 0.8263711515750735,
+      "epoch": 0.8538308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.5399659361713537e-06,
-      "loss": 0.7901,
+      "learning_rate": 1.0998447769928978e-06,
+      "loss": 0.8517,
       "step": 30089
     },
     {
-      "epoch": 0.8263986158029167,
+      "epoch": 0.8538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.5394916997261e-06,
-      "loss": 0.7801,
+      "learning_rate": 1.0994257791129548e-06,
+      "loss": 0.8289,
       "step": 30090
     },
     {
-      "epoch": 0.8264260800307599,
+      "epoch": 0.8538876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.5390175302236642e-06,
-      "loss": 0.8406,
+      "learning_rate": 1.0990068564165645e-06,
+      "loss": 0.7805,
       "step": 30091
     },
     {
-      "epoch": 0.8264535442586032,
+      "epoch": 0.8539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.5385434276678002e-06,
-      "loss": 0.8272,
+      "learning_rate": 1.0985880089072609e-06,
+      "loss": 0.8508,
       "step": 30092
     },
     {
-      "epoch": 0.8264810084864465,
+      "epoch": 0.85394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.5380693920622614e-06,
-      "loss": 0.8221,
+      "learning_rate": 1.0981692365885843e-06,
+      "loss": 0.8465,
       "step": 30093
     },
     {
-      "epoch": 0.8265084727142896,
+      "epoch": 0.8539727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.5375954234107914e-06,
-      "loss": 0.7906,
+      "learning_rate": 1.097750539464073e-06,
+      "loss": 0.822,
       "step": 30094
     },
     {
-      "epoch": 0.8265359369421329,
+      "epoch": 0.8540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.5371215217171454e-06,
-      "loss": 0.8254,
+      "learning_rate": 1.0973319175372632e-06,
+      "loss": 0.851,
       "step": 30095
     },
     {
-      "epoch": 0.8265634011699761,
+      "epoch": 0.8540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.5366476869850722e-06,
-      "loss": 0.8892,
+      "learning_rate": 1.096913370811693e-06,
+      "loss": 0.8093,
       "step": 30096
     },
     {
-      "epoch": 0.8265908653978193,
+      "epoch": 0.8540578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.5361739192183223e-06,
-      "loss": 0.7958,
+      "learning_rate": 1.096494899290893e-06,
+      "loss": 0.8768,
       "step": 30097
     },
     {
-      "epoch": 0.8266183296256626,
+      "epoch": 0.8540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.5357002184206416e-06,
-      "loss": 0.82,
+      "learning_rate": 1.0960765029784015e-06,
+      "loss": 0.8674,
       "step": 30098
     },
     {
-      "epoch": 0.8266457938535058,
+      "epoch": 0.8541146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.5352265845957814e-06,
-      "loss": 0.8086,
+      "learning_rate": 1.0956581818777533e-06,
+      "loss": 0.819,
       "step": 30099
     },
     {
-      "epoch": 0.8266732580813491,
+      "epoch": 0.8541430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.5347530177474857e-06,
-      "loss": 0.794,
+      "learning_rate": 1.0952399359924793e-06,
+      "loss": 0.7323,
       "step": 30100
     },
     {
-      "epoch": 0.8267007223091922,
+      "epoch": 0.8541713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.5342795178795023e-06,
-      "loss": 0.7995,
+      "learning_rate": 1.094821765326114e-06,
+      "loss": 0.8467,
       "step": 30101
     },
     {
-      "epoch": 0.8267281865370355,
+      "epoch": 0.8541997729852441,
       "grad_norm": 0.0,
-      "learning_rate": 1.5338060849955793e-06,
-      "loss": 0.787,
+      "learning_rate": 1.0944036698821913e-06,
+      "loss": 0.8072,
       "step": 30102
     },
     {
-      "epoch": 0.8267556507648788,
+      "epoch": 0.8542281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.5333327190994618e-06,
-      "loss": 0.8794,
+      "learning_rate": 1.0939856496642398e-06,
+      "loss": 0.8086,
       "step": 30103
     },
     {
-      "epoch": 0.826783114992722,
+      "epoch": 0.8542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.5328594201948976e-06,
-      "loss": 0.8991,
+      "learning_rate": 1.0935677046757908e-06,
+      "loss": 0.8026,
       "step": 30104
     },
     {
-      "epoch": 0.8268105792205652,
+      "epoch": 0.8542849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 1.5323861882856295e-06,
-      "loss": 0.8476,
+      "learning_rate": 1.0931498349203785e-06,
+      "loss": 0.791,
       "step": 30105
     },
     {
-      "epoch": 0.8268380434484085,
+      "epoch": 0.8543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.5319130233753997e-06,
-      "loss": 0.7936,
+      "learning_rate": 1.0927320404015274e-06,
+      "loss": 0.7582,
       "step": 30106
     },
     {
-      "epoch": 0.8268655076762517,
+      "epoch": 0.8543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.5314399254679545e-06,
-      "loss": 0.8839,
+      "learning_rate": 1.0923143211227695e-06,
+      "loss": 0.8442,
       "step": 30107
     },
     {
-      "epoch": 0.8268929719040949,
+      "epoch": 0.8543700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.5309668945670375e-06,
-      "loss": 0.8589,
+      "learning_rate": 1.0918966770876349e-06,
+      "loss": 0.8262,
       "step": 30108
     },
     {
-      "epoch": 0.8269204361319381,
+      "epoch": 0.8543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.53049393067639e-06,
-      "loss": 0.7823,
+      "learning_rate": 1.0914791082996456e-06,
+      "loss": 0.7442,
       "step": 30109
     },
     {
-      "epoch": 0.8269479003597814,
+      "epoch": 0.8544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.5300210337997557e-06,
-      "loss": 0.8636,
+      "learning_rate": 1.0910616147623365e-06,
+      "loss": 0.8866,
       "step": 30110
     },
     {
-      "epoch": 0.8269753645876247,
+      "epoch": 0.854455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.5295482039408793e-06,
-      "loss": 0.7908,
+      "learning_rate": 1.090644196479228e-06,
+      "loss": 0.7355,
       "step": 30111
     },
     {
-      "epoch": 0.8270028288154678,
+      "epoch": 0.854483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.5290754411034948e-06,
-      "loss": 0.8306,
+      "learning_rate": 1.0902268534538496e-06,
+      "loss": 0.9549,
       "step": 30112
     },
     {
-      "epoch": 0.8270302930433111,
+      "epoch": 0.8545119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.5286027452913499e-06,
-      "loss": 0.9219,
+      "learning_rate": 1.089809585689726e-06,
+      "loss": 0.761,
       "step": 30113
     },
     {
-      "epoch": 0.8270577572711543,
+      "epoch": 0.8545402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.5281301165081785e-06,
-      "loss": 0.8199,
+      "learning_rate": 1.0893923931903805e-06,
+      "loss": 0.9006,
       "step": 30114
     },
     {
-      "epoch": 0.8270852214989975,
+      "epoch": 0.8545686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.5276575547577243e-06,
-      "loss": 0.8194,
+      "learning_rate": 1.0889752759593386e-06,
+      "loss": 0.6226,
       "step": 30115
     },
     {
-      "epoch": 0.8271126857268408,
+      "epoch": 0.8545970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.5271850600437245e-06,
-      "loss": 0.7189,
+      "learning_rate": 1.0885582340001245e-06,
+      "loss": 0.7026,
       "step": 30116
     },
     {
-      "epoch": 0.827140149954684,
+      "epoch": 0.8546254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.5267126323699211e-06,
-      "loss": 0.7817,
+      "learning_rate": 1.0881412673162572e-06,
+      "loss": 0.8443,
       "step": 30117
     },
     {
-      "epoch": 0.8271676141825273,
+      "epoch": 0.8546538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.5262402717400482e-06,
-      "loss": 0.8106,
+      "learning_rate": 1.087724375911261e-06,
+      "loss": 0.8173,
       "step": 30118
     },
     {
-      "epoch": 0.8271950784103705,
+      "epoch": 0.8546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.525767978157844e-06,
-      "loss": 0.7094,
+      "learning_rate": 1.0873075597886607e-06,
+      "loss": 0.8459,
       "step": 30119
     },
     {
-      "epoch": 0.8272225426382137,
+      "epoch": 0.8547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.5252957516270451e-06,
-      "loss": 0.8449,
+      "learning_rate": 1.086890818951971e-06,
+      "loss": 0.819,
       "step": 30120
     },
     {
-      "epoch": 0.827250006866057,
+      "epoch": 0.854738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.5248235921513932e-06,
-      "loss": 0.8543,
+      "learning_rate": 1.0864741534047151e-06,
+      "loss": 0.8398,
       "step": 30121
     },
     {
-      "epoch": 0.8272774710939002,
+      "epoch": 0.8547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.5243514997346165e-06,
-      "loss": 0.7641,
+      "learning_rate": 1.0860575631504154e-06,
+      "loss": 0.8147,
       "step": 30122
     },
     {
-      "epoch": 0.8273049353217434,
+      "epoch": 0.8547956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.5238794743804563e-06,
-      "loss": 0.8604,
+      "learning_rate": 1.0856410481925861e-06,
+      "loss": 0.8052,
       "step": 30123
     },
     {
-      "epoch": 0.8273323995495867,
+      "epoch": 0.8548240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.5234075160926432e-06,
-      "loss": 0.9404,
+      "learning_rate": 1.0852246085347483e-06,
+      "loss": 0.8729,
       "step": 30124
     },
     {
-      "epoch": 0.8273598637774299,
+      "epoch": 0.8548524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.5229356248749127e-06,
-      "loss": 0.8238,
+      "learning_rate": 1.0848082441804176e-06,
+      "loss": 0.8728,
       "step": 30125
     },
     {
-      "epoch": 0.8273873280052731,
+      "epoch": 0.8548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.522463800731e-06,
-      "loss": 0.8701,
+      "learning_rate": 1.0843919551331139e-06,
+      "loss": 0.7308,
       "step": 30126
     },
     {
-      "epoch": 0.8274147922331163,
+      "epoch": 0.8549091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.5219920436646374e-06,
-      "loss": 0.7245,
+      "learning_rate": 1.0839757413963524e-06,
+      "loss": 0.7211,
       "step": 30127
     },
     {
-      "epoch": 0.8274422564609596,
+      "epoch": 0.8549375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.5215203536795597e-06,
-      "loss": 0.7874,
+      "learning_rate": 1.0835596029736484e-06,
+      "loss": 0.8331,
       "step": 30128
     },
     {
-      "epoch": 0.8274697206888029,
+      "epoch": 0.8549659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 1.5210487307794974e-06,
-      "loss": 0.8438,
+      "learning_rate": 1.0831435398685164e-06,
+      "loss": 0.9331,
       "step": 30129
     },
     {
-      "epoch": 0.827497184916646,
+      "epoch": 0.854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.5205771749681797e-06,
-      "loss": 0.7697,
+      "learning_rate": 1.0827275520844738e-06,
+      "loss": 0.7891,
       "step": 30130
     },
     {
-      "epoch": 0.8275246491444893,
+      "epoch": 0.8550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.5201056862493402e-06,
-      "loss": 0.7975,
+      "learning_rate": 1.08231163962503e-06,
+      "loss": 0.7822,
       "step": 30131
     },
     {
-      "epoch": 0.8275521133723326,
+      "epoch": 0.8550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.5196342646267093e-06,
-      "loss": 0.852,
+      "learning_rate": 1.081895802493701e-06,
+      "loss": 0.8855,
       "step": 30132
     },
     {
-      "epoch": 0.8275795776001758,
+      "epoch": 0.8550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.5191629101040162e-06,
-      "loss": 0.8208,
+      "learning_rate": 1.0814800406940007e-06,
+      "loss": 0.8247,
       "step": 30133
     },
     {
-      "epoch": 0.827607041828019,
+      "epoch": 0.8551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.5186916226849924e-06,
-      "loss": 0.7426,
+      "learning_rate": 1.0810643542294386e-06,
+      "loss": 0.8238,
       "step": 30134
     },
     {
-      "epoch": 0.8276345060558622,
+      "epoch": 0.8551362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.518220402373367e-06,
-      "loss": 0.8497,
+      "learning_rate": 1.0806487431035262e-06,
+      "loss": 0.7561,
       "step": 30135
     },
     {
-      "epoch": 0.8276619702837055,
+      "epoch": 0.8551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.5177492491728662e-06,
-      "loss": 0.8587,
+      "learning_rate": 1.0802332073197762e-06,
+      "loss": 0.8997,
       "step": 30136
     },
     {
-      "epoch": 0.8276894345115487,
+      "epoch": 0.8551929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.5172781630872213e-06,
-      "loss": 0.8534,
+      "learning_rate": 1.079817746881696e-06,
+      "loss": 0.7685,
       "step": 30137
     },
     {
-      "epoch": 0.8277168987393919,
+      "epoch": 0.8552213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 1.516807144120155e-06,
-      "loss": 0.7923,
+      "learning_rate": 1.0794023617927961e-06,
+      "loss": 0.8734,
       "step": 30138
     },
     {
-      "epoch": 0.8277443629672352,
+      "epoch": 0.8552497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 1.516336192275396e-06,
-      "loss": 0.7416,
+      "learning_rate": 1.0789870520565848e-06,
+      "loss": 0.8049,
       "step": 30139
     },
     {
-      "epoch": 0.8277718271950784,
+      "epoch": 0.85527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.515865307556671e-06,
-      "loss": 0.865,
+      "learning_rate": 1.0785718176765715e-06,
+      "loss": 0.8208,
       "step": 30140
     },
     {
-      "epoch": 0.8277992914229216,
+      "epoch": 0.8553064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.5153944899677097e-06,
-      "loss": 0.7833,
+      "learning_rate": 1.0781566586562653e-06,
+      "loss": 0.756,
       "step": 30141
     },
     {
-      "epoch": 0.8278267556507649,
+      "epoch": 0.8553348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.5149237395122306e-06,
-      "loss": 0.7956,
+      "learning_rate": 1.0777415749991682e-06,
+      "loss": 0.807,
       "step": 30142
     },
     {
-      "epoch": 0.8278542198786081,
+      "epoch": 0.8553632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.514453056193962e-06,
-      "loss": 0.8593,
+      "learning_rate": 1.0773265667087896e-06,
+      "loss": 0.7837,
       "step": 30143
     },
     {
-      "epoch": 0.8278816841064514,
+      "epoch": 0.8553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.5139824400166303e-06,
-      "loss": 0.8964,
+      "learning_rate": 1.0769116337886366e-06,
+      "loss": 0.8071,
       "step": 30144
     },
     {
-      "epoch": 0.8279091483342946,
+      "epoch": 0.8554199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.513511890983954e-06,
-      "loss": 0.7339,
+      "learning_rate": 1.0764967762422107e-06,
+      "loss": 0.7993,
       "step": 30145
     },
     {
-      "epoch": 0.8279366125621378,
+      "epoch": 0.8554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.5130414090996614e-06,
-      "loss": 0.8473,
+      "learning_rate": 1.076081994073017e-06,
+      "loss": 0.7256,
       "step": 30146
     },
     {
-      "epoch": 0.8279640767899811,
+      "epoch": 0.8554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.5125709943674716e-06,
-      "loss": 0.8655,
+      "learning_rate": 1.075667287284562e-06,
+      "loss": 0.7882,
       "step": 30147
     },
     {
-      "epoch": 0.8279915410178242,
+      "epoch": 0.8555051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.5121006467911058e-06,
-      "loss": 0.672,
+      "learning_rate": 1.0752526558803444e-06,
+      "loss": 0.785,
       "step": 30148
     },
     {
-      "epoch": 0.8280190052456675,
+      "epoch": 0.8555334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 1.5116303663742893e-06,
-      "loss": 0.8134,
+      "learning_rate": 1.0748380998638685e-06,
+      "loss": 0.7169,
       "step": 30149
     },
     {
-      "epoch": 0.8280464694735108,
+      "epoch": 0.8555618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.51116015312074e-06,
-      "loss": 0.8412,
+      "learning_rate": 1.0744236192386393e-06,
+      "loss": 0.8578,
       "step": 30150
     },
     {
-      "epoch": 0.828073933701354,
+      "epoch": 0.8555902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.5106900070341802e-06,
-      "loss": 0.7057,
+      "learning_rate": 1.0740092140081527e-06,
+      "loss": 0.8264,
       "step": 30151
     },
     {
-      "epoch": 0.8281013979291972,
+      "epoch": 0.8556186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.5102199281183327e-06,
-      "loss": 0.8885,
+      "learning_rate": 1.0735948841759114e-06,
+      "loss": 0.7025,
       "step": 30152
     },
     {
-      "epoch": 0.8281288621570404,
+      "epoch": 0.8556469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.5097499163769124e-06,
-      "loss": 0.7302,
+      "learning_rate": 1.0731806297454172e-06,
+      "loss": 0.8415,
       "step": 30153
     },
     {
-      "epoch": 0.8281563263848837,
+      "epoch": 0.8556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.5092799718136386e-06,
-      "loss": 0.81,
+      "learning_rate": 1.0727664507201652e-06,
+      "loss": 0.7992,
       "step": 30154
     },
     {
-      "epoch": 0.828183790612727,
+      "epoch": 0.8557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.5088100944322304e-06,
-      "loss": 0.8643,
+      "learning_rate": 1.0723523471036545e-06,
+      "loss": 0.7943,
       "step": 30155
     },
     {
-      "epoch": 0.8282112548405701,
+      "epoch": 0.8557321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 1.5083402842364048e-06,
-      "loss": 0.7665,
+      "learning_rate": 1.0719383188993893e-06,
+      "loss": 0.8413,
       "step": 30156
     },
     {
-      "epoch": 0.8282387190684134,
+      "epoch": 0.8557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.5078705412298811e-06,
-      "loss": 0.7821,
+      "learning_rate": 1.071524366110861e-06,
+      "loss": 0.8307,
       "step": 30157
     },
     {
-      "epoch": 0.8282661832962567,
+      "epoch": 0.855788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.5074008654163775e-06,
-      "loss": 0.8188,
+      "learning_rate": 1.071110488741567e-06,
+      "loss": 0.843,
       "step": 30158
     },
     {
-      "epoch": 0.8282936475240998,
+      "epoch": 0.8558172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 1.5069312567996053e-06,
-      "loss": 0.8464,
+      "learning_rate": 1.0706966867950052e-06,
+      "loss": 0.742,
       "step": 30159
     },
     {
-      "epoch": 0.8283211117519431,
+      "epoch": 0.8558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.506461715383284e-06,
-      "loss": 0.7405,
+      "learning_rate": 1.070282960274669e-06,
+      "loss": 0.7392,
       "step": 30160
     },
     {
-      "epoch": 0.8283485759797863,
+      "epoch": 0.8558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.5059922411711292e-06,
-      "loss": 0.7887,
+      "learning_rate": 1.069869309184056e-06,
+      "loss": 0.9044,
       "step": 30161
     },
     {
-      "epoch": 0.8283760402076296,
+      "epoch": 0.8559023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.5055228341668516e-06,
-      "loss": 0.7179,
+      "learning_rate": 1.0694557335266564e-06,
+      "loss": 0.8074,
       "step": 30162
     },
     {
-      "epoch": 0.8284035044354728,
+      "epoch": 0.8559307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.505053494374168e-06,
-      "loss": 0.8883,
+      "learning_rate": 1.0690422333059657e-06,
+      "loss": 0.779,
       "step": 30163
     },
     {
-      "epoch": 0.828430968663316,
+      "epoch": 0.8559591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.504584221796793e-06,
-      "loss": 0.863,
+      "learning_rate": 1.0686288085254782e-06,
+      "loss": 0.815,
       "step": 30164
     },
     {
-      "epoch": 0.8284584328911593,
+      "epoch": 0.8559875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.5041150164384378e-06,
-      "loss": 0.803,
+      "learning_rate": 1.0682154591886828e-06,
+      "loss": 0.7169,
       "step": 30165
     },
     {
-      "epoch": 0.8284858971190024,
+      "epoch": 0.8560158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.5036458783028141e-06,
-      "loss": 0.846,
+      "learning_rate": 1.0678021852990727e-06,
+      "loss": 0.7116,
       "step": 30166
     },
     {
-      "epoch": 0.8285133613468457,
+      "epoch": 0.8560442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 1.503176807393637e-06,
-      "loss": 0.8259,
+      "learning_rate": 1.0673889868601416e-06,
+      "loss": 0.8556,
       "step": 30167
     },
     {
-      "epoch": 0.828540825574689,
+      "epoch": 0.8560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.502707803714617e-06,
-      "loss": 0.8846,
+      "learning_rate": 1.0669758638753759e-06,
+      "loss": 0.8476,
       "step": 30168
     },
     {
-      "epoch": 0.8285682898025322,
+      "epoch": 0.8561010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.5022388672694633e-06,
-      "loss": 0.8605,
+      "learning_rate": 1.0665628163482655e-06,
+      "loss": 0.8782,
       "step": 30169
     },
     {
-      "epoch": 0.8285957540303754,
+      "epoch": 0.8561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.5017699980618895e-06,
-      "loss": 0.7425,
+      "learning_rate": 1.0661498442823015e-06,
+      "loss": 0.86,
       "step": 30170
     },
     {
-      "epoch": 0.8286232182582187,
+      "epoch": 0.8561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.5013011960956004e-06,
-      "loss": 0.7892,
+      "learning_rate": 1.065736947680971e-06,
+      "loss": 0.7662,
       "step": 30171
     },
     {
-      "epoch": 0.8286506824860619,
+      "epoch": 0.8561861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.5008324613743098e-06,
-      "loss": 0.8373,
+      "learning_rate": 1.065324126547761e-06,
+      "loss": 0.8226,
       "step": 30172
     },
     {
-      "epoch": 0.8286781467139052,
+      "epoch": 0.8562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.5003637939017234e-06,
-      "loss": 0.8213,
+      "learning_rate": 1.0649113808861633e-06,
+      "loss": 0.8503,
       "step": 30173
     },
     {
-      "epoch": 0.8287056109417483,
+      "epoch": 0.8562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.499895193681552e-06,
-      "loss": 0.7262,
+      "learning_rate": 1.0644987106996584e-06,
+      "loss": 0.9405,
       "step": 30174
     },
     {
-      "epoch": 0.8287330751695916,
+      "epoch": 0.8562712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.4994266607175023e-06,
-      "loss": 0.9109,
+      "learning_rate": 1.0640861159917337e-06,
+      "loss": 0.8539,
       "step": 30175
     },
     {
-      "epoch": 0.8287605393974349,
+      "epoch": 0.856299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.4989581950132836e-06,
-      "loss": 0.9255,
+      "learning_rate": 1.0636735967658785e-06,
+      "loss": 0.7798,
       "step": 30176
     },
     {
-      "epoch": 0.828788003625278,
+      "epoch": 0.856328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.4984897965726008e-06,
-      "loss": 0.8112,
+      "learning_rate": 1.063261153025571e-06,
+      "loss": 0.81,
       "step": 30177
     },
     {
-      "epoch": 0.8288154678531213,
+      "epoch": 0.8563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.4980214653991576e-06,
-      "loss": 0.8712,
+      "learning_rate": 1.0628487847742997e-06,
+      "loss": 0.8757,
       "step": 30178
     },
     {
-      "epoch": 0.8288429320809645,
+      "epoch": 0.8563847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.497553201496662e-06,
-      "loss": 0.7754,
+      "learning_rate": 1.0624364920155473e-06,
+      "loss": 0.8144,
       "step": 30179
     },
     {
-      "epoch": 0.8288703963088078,
+      "epoch": 0.8564131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.4970850048688191e-06,
-      "loss": 0.8872,
+      "learning_rate": 1.0620242747527943e-06,
+      "loss": 0.7818,
       "step": 30180
     },
     {
-      "epoch": 0.828897860536651,
+      "epoch": 0.8564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.4966168755193334e-06,
-      "loss": 0.8708,
+      "learning_rate": 1.0616121329895235e-06,
+      "loss": 0.8275,
       "step": 30181
     },
     {
-      "epoch": 0.8289253247644942,
+      "epoch": 0.8564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.4961488134519108e-06,
-      "loss": 0.8492,
+      "learning_rate": 1.0612000667292188e-06,
+      "loss": 0.8327,
       "step": 30182
     },
     {
-      "epoch": 0.8289527889923375,
+      "epoch": 0.8564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.495680818670251e-06,
-      "loss": 0.7754,
+      "learning_rate": 1.0607880759753575e-06,
+      "loss": 0.8549,
       "step": 30183
     },
     {
-      "epoch": 0.8289802532201808,
+      "epoch": 0.8565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.4952128911780584e-06,
-      "loss": 0.7712,
+      "learning_rate": 1.0603761607314212e-06,
+      "loss": 0.7878,
       "step": 30184
     },
     {
-      "epoch": 0.8290077174480239,
+      "epoch": 0.8565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.494745030979039e-06,
-      "loss": 0.9445,
+      "learning_rate": 1.0599643210008893e-06,
+      "loss": 0.7553,
       "step": 30185
     },
     {
-      "epoch": 0.8290351816758672,
+      "epoch": 0.8565834279228149,
       "grad_norm": 0.0,
-      "learning_rate": 1.4942772380768888e-06,
-      "loss": 0.812,
+      "learning_rate": 1.05955255678724e-06,
+      "loss": 0.7643,
       "step": 30186
     },
     {
-      "epoch": 0.8290626459037104,
+      "epoch": 0.8566118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.4938095124753117e-06,
-      "loss": 0.7917,
+      "learning_rate": 1.0591408680939541e-06,
+      "loss": 0.8181,
       "step": 30187
     },
     {
-      "epoch": 0.8290901101315536,
+      "epoch": 0.8566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.493341854178011e-06,
-      "loss": 0.9054,
+      "learning_rate": 1.0587292549245065e-06,
+      "loss": 0.7962,
       "step": 30188
     },
     {
-      "epoch": 0.8291175743593969,
+      "epoch": 0.8566685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.4928742631886828e-06,
-      "loss": 0.7916,
+      "learning_rate": 1.0583177172823734e-06,
+      "loss": 0.7705,
       "step": 30189
     },
     {
-      "epoch": 0.8291450385872401,
+      "epoch": 0.8566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.4924067395110276e-06,
-      "loss": 0.936,
+      "learning_rate": 1.057906255171035e-06,
+      "loss": 0.8612,
       "step": 30190
     },
     {
-      "epoch": 0.8291725028150834,
+      "epoch": 0.8567253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 1.4919392831487468e-06,
-      "loss": 0.8271,
+      "learning_rate": 1.0574948685939624e-06,
+      "loss": 0.8209,
       "step": 30191
     },
     {
-      "epoch": 0.8291999670429265,
+      "epoch": 0.8567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.4914718941055407e-06,
-      "loss": 0.7356,
+      "learning_rate": 1.0570835575546333e-06,
+      "loss": 0.766,
       "step": 30192
     },
     {
-      "epoch": 0.8292274312707698,
+      "epoch": 0.8567820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.491004572385102e-06,
-      "loss": 0.8726,
+      "learning_rate": 1.0566723220565222e-06,
+      "loss": 0.8464,
       "step": 30193
     },
     {
-      "epoch": 0.8292548954986131,
+      "epoch": 0.8568104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.490537317991133e-06,
-      "loss": 0.7488,
+      "learning_rate": 1.0562611621031016e-06,
+      "loss": 0.809,
       "step": 30194
     },
     {
-      "epoch": 0.8292823597264563,
+      "epoch": 0.856838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.490070130927328e-06,
-      "loss": 0.8679,
+      "learning_rate": 1.0558500776978431e-06,
+      "loss": 0.8466,
       "step": 30195
     },
     {
-      "epoch": 0.8293098239542995,
+      "epoch": 0.8568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.4896030111973836e-06,
-      "loss": 0.7666,
+      "learning_rate": 1.055439068844224e-06,
+      "loss": 0.8068,
       "step": 30196
     },
     {
-      "epoch": 0.8293372881821428,
+      "epoch": 0.8568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.4891359588049969e-06,
-      "loss": 0.891,
+      "learning_rate": 1.0550281355457115e-06,
+      "loss": 0.8524,
       "step": 30197
     },
     {
-      "epoch": 0.829364752409986,
+      "epoch": 0.8569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.488668973753863e-06,
-      "loss": 0.7265,
+      "learning_rate": 1.0546172778057783e-06,
+      "loss": 0.8196,
       "step": 30198
     },
     {
-      "epoch": 0.8293922166378292,
+      "epoch": 0.8569523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.4882020560476785e-06,
-      "loss": 0.7892,
+      "learning_rate": 1.0542064956278974e-06,
+      "loss": 0.8502,
       "step": 30199
     },
     {
-      "epoch": 0.8294196808656724,
+      "epoch": 0.8569807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.4877352056901373e-06,
-      "loss": 0.7916,
+      "learning_rate": 1.0537957890155338e-06,
+      "loss": 0.8101,
       "step": 30200
     },
     {
-      "epoch": 0.8294471450935157,
+      "epoch": 0.8570090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.4872684226849332e-06,
-      "loss": 0.861,
+      "learning_rate": 1.0533851579721588e-06,
+      "loss": 0.8391,
       "step": 30201
     },
     {
-      "epoch": 0.829474609321359,
+      "epoch": 0.8570374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.4868017070357565e-06,
-      "loss": 0.737,
+      "learning_rate": 1.0529746025012422e-06,
+      "loss": 0.8677,
       "step": 30202
     },
     {
-      "epoch": 0.8295020735492021,
+      "epoch": 0.8570658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.486335058746302e-06,
-      "loss": 0.8172,
+      "learning_rate": 1.0525641226062522e-06,
+      "loss": 0.7254,
       "step": 30203
     },
     {
-      "epoch": 0.8295295377770454,
+      "epoch": 0.8570942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 1.4858684778202626e-06,
-      "loss": 0.799,
+      "learning_rate": 1.052153718290656e-06,
+      "loss": 0.8189,
       "step": 30204
     },
     {
-      "epoch": 0.8295570020048886,
+      "epoch": 0.8571225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.4854019642613327e-06,
-      "loss": 0.8167,
+      "learning_rate": 1.0517433895579177e-06,
+      "loss": 0.7535,
       "step": 30205
     },
     {
-      "epoch": 0.8295844662327319,
+      "epoch": 0.8571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.4849355180731973e-06,
-      "loss": 0.8356,
+      "learning_rate": 1.0513331364115055e-06,
+      "loss": 0.7825,
       "step": 30206
     },
     {
-      "epoch": 0.8296119304605751,
+      "epoch": 0.8571793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.4844691392595511e-06,
-      "loss": 0.7937,
+      "learning_rate": 1.0509229588548865e-06,
+      "loss": 0.8469,
       "step": 30207
     },
     {
-      "epoch": 0.8296393946884183,
+      "epoch": 0.8572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.4840028278240837e-06,
-      "loss": 0.7258,
+      "learning_rate": 1.0505128568915223e-06,
+      "loss": 0.7453,
       "step": 30208
     },
     {
-      "epoch": 0.8296668589162616,
+      "epoch": 0.8572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.4835365837704873e-06,
-      "loss": 0.7674,
+      "learning_rate": 1.0501028305248773e-06,
+      "loss": 0.859,
       "step": 30209
     },
     {
-      "epoch": 0.8296943231441049,
+      "epoch": 0.8572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.483070407102446e-06,
-      "loss": 0.8561,
+      "learning_rate": 1.0496928797584183e-06,
+      "loss": 0.8914,
       "step": 30210
     },
     {
-      "epoch": 0.829721787371948,
+      "epoch": 0.8572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.4826042978236532e-06,
-      "loss": 0.7542,
+      "learning_rate": 1.0492830045956037e-06,
+      "loss": 0.8146,
       "step": 30211
     },
     {
-      "epoch": 0.8297492515997913,
+      "epoch": 0.8573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.4821382559377916e-06,
-      "loss": 0.7813,
+      "learning_rate": 1.0488732050398986e-06,
+      "loss": 0.8124,
       "step": 30212
     },
     {
-      "epoch": 0.8297767158276345,
+      "epoch": 0.857349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.4816722814485519e-06,
-      "loss": 0.7849,
+      "learning_rate": 1.0484634810947658e-06,
+      "loss": 0.8784,
       "step": 30213
     },
     {
-      "epoch": 0.8298041800554777,
+      "epoch": 0.857377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.4812063743596205e-06,
-      "loss": 0.8644,
+      "learning_rate": 1.0480538327636614e-06,
+      "loss": 0.9319,
       "step": 30214
     },
     {
-      "epoch": 0.829831644283321,
+      "epoch": 0.8574063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.480740534674685e-06,
-      "loss": 0.8204,
+      "learning_rate": 1.0476442600500503e-06,
+      "loss": 0.9042,
       "step": 30215
     },
     {
-      "epoch": 0.8298591085111642,
+      "epoch": 0.8574347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.4802747623974323e-06,
-      "loss": 0.7547,
+      "learning_rate": 1.0472347629573886e-06,
+      "loss": 0.8392,
       "step": 30216
     },
     {
-      "epoch": 0.8298865727390075,
+      "epoch": 0.8574631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.4798090575315438e-06,
-      "loss": 0.6997,
+      "learning_rate": 1.0468253414891393e-06,
+      "loss": 0.8451,
       "step": 30217
     },
     {
-      "epoch": 0.8299140369668507,
+      "epoch": 0.8574914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.4793434200807078e-06,
-      "loss": 0.7047,
+      "learning_rate": 1.0464159956487596e-06,
+      "loss": 0.8501,
       "step": 30218
     },
     {
-      "epoch": 0.8299415011946939,
+      "epoch": 0.8575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.4788778500486056e-06,
-      "loss": 0.7229,
+      "learning_rate": 1.0460067254397043e-06,
+      "loss": 0.7685,
       "step": 30219
     },
     {
-      "epoch": 0.8299689654225372,
+      "epoch": 0.8575482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.4784123474389233e-06,
-      "loss": 0.8867,
+      "learning_rate": 1.0455975308654332e-06,
+      "loss": 0.8012,
       "step": 30220
     },
     {
-      "epoch": 0.8299964296503803,
+      "epoch": 0.8575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.4779469122553425e-06,
-      "loss": 0.737,
+      "learning_rate": 1.0451884119294043e-06,
+      "loss": 0.7668,
       "step": 30221
     },
     {
-      "epoch": 0.8300238938782236,
+      "epoch": 0.8576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.4774815445015466e-06,
+      "learning_rate": 1.0447793686350694e-06,
       "loss": 0.7883,
       "step": 30222
     },
     {
-      "epoch": 0.8300513581060669,
+      "epoch": 0.857633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.477016244181222e-06,
-      "loss": 0.8825,
+      "learning_rate": 1.0443704009858857e-06,
+      "loss": 0.8036,
       "step": 30223
     },
     {
-      "epoch": 0.8300788223339101,
+      "epoch": 0.8576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.476551011298043e-06,
-      "loss": 0.7797,
+      "learning_rate": 1.0439615089853094e-06,
+      "loss": 0.9032,
       "step": 30224
     },
     {
-      "epoch": 0.8301062865617533,
+      "epoch": 0.8576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.476085845855696e-06,
-      "loss": 0.6815,
+      "learning_rate": 1.0435526926367924e-06,
+      "loss": 0.8613,
       "step": 30225
     },
     {
-      "epoch": 0.8301337507895965,
+      "epoch": 0.8577185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.4756207478578577e-06,
-      "loss": 0.7764,
+      "learning_rate": 1.0431439519437869e-06,
+      "loss": 0.7081,
       "step": 30226
     },
     {
-      "epoch": 0.8301612150174398,
+      "epoch": 0.8577468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.4751557173082097e-06,
-      "loss": 0.8666,
+      "learning_rate": 1.0427352869097495e-06,
+      "loss": 0.7344,
       "step": 30227
     },
     {
-      "epoch": 0.8301886792452831,
+      "epoch": 0.8577752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.4746907542104328e-06,
-      "loss": 0.825,
+      "learning_rate": 1.0423266975381274e-06,
+      "loss": 0.7542,
       "step": 30228
     },
     {
-      "epoch": 0.8302161434731262,
+      "epoch": 0.8578036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.4742258585682056e-06,
-      "loss": 0.7485,
+      "learning_rate": 1.0419181838323743e-06,
+      "loss": 0.8453,
       "step": 30229
     },
     {
-      "epoch": 0.8302436077009695,
+      "epoch": 0.8578320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.4737610303852057e-06,
-      "loss": 0.8647,
+      "learning_rate": 1.0415097457959433e-06,
+      "loss": 0.8249,
       "step": 30230
     },
     {
-      "epoch": 0.8302710719288128,
+      "epoch": 0.8578603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.4732962696651098e-06,
-      "loss": 0.8462,
+      "learning_rate": 1.0411013834322791e-06,
+      "loss": 0.8021,
       "step": 30231
     },
     {
-      "epoch": 0.8302985361566559,
+      "epoch": 0.85788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.4728315764115975e-06,
-      "loss": 0.8083,
+      "learning_rate": 1.040693096744837e-06,
+      "loss": 0.9249,
       "step": 30232
     },
     {
-      "epoch": 0.8303260003844992,
+      "epoch": 0.8579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.4723669506283467e-06,
-      "loss": 0.7369,
+      "learning_rate": 1.0402848857370617e-06,
+      "loss": 0.8316,
       "step": 30233
     },
     {
-      "epoch": 0.8303534646123424,
+      "epoch": 0.8579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.4719023923190301e-06,
-      "loss": 0.9099,
+      "learning_rate": 1.039876750412403e-06,
+      "loss": 0.7838,
       "step": 30234
     },
     {
-      "epoch": 0.8303809288401857,
+      "epoch": 0.8579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.471437901487328e-06,
-      "loss": 0.9111,
+      "learning_rate": 1.039468690774309e-06,
+      "loss": 0.8858,
       "step": 30235
     },
     {
-      "epoch": 0.8304083930680289,
+      "epoch": 0.8580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.470973478136909e-06,
-      "loss": 0.7971,
+      "learning_rate": 1.0390607068262248e-06,
+      "loss": 0.7366,
       "step": 30236
     },
     {
-      "epoch": 0.8304358572958721,
+      "epoch": 0.8580306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.4705091222714529e-06,
-      "loss": 0.8395,
+      "learning_rate": 1.0386527985715977e-06,
+      "loss": 0.826,
       "step": 30237
     },
     {
-      "epoch": 0.8304633215237154,
+      "epoch": 0.8580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.4700448338946326e-06,
-      "loss": 0.799,
+      "learning_rate": 1.038244966013875e-06,
+      "loss": 0.8214,
       "step": 30238
     },
     {
-      "epoch": 0.8304907857515585,
+      "epoch": 0.8580874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.469580613010121e-06,
-      "loss": 0.8521,
+      "learning_rate": 1.037837209156497e-06,
+      "loss": 0.8304,
       "step": 30239
     },
     {
-      "epoch": 0.8305182499794018,
+      "epoch": 0.8581157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 1.4691164596215946e-06,
-      "loss": 0.7364,
+      "learning_rate": 1.0374295280029123e-06,
+      "loss": 0.8813,
       "step": 30240
     },
     {
-      "epoch": 0.8305457142072451,
+      "epoch": 0.8581441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.4686523737327218e-06,
-      "loss": 0.7917,
+      "learning_rate": 1.0370219225565647e-06,
+      "loss": 0.8266,
       "step": 30241
     },
     {
-      "epoch": 0.8305731784350883,
+      "epoch": 0.858172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.4681883553471776e-06,
-      "loss": 0.8322,
+      "learning_rate": 1.0366143928208938e-06,
+      "loss": 0.8003,
       "step": 30242
     },
     {
-      "epoch": 0.8306006426629315,
+      "epoch": 0.8582009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.46772440446863e-06,
-      "loss": 0.8244,
+      "learning_rate": 1.0362069387993434e-06,
+      "loss": 0.8006,
       "step": 30243
     },
     {
-      "epoch": 0.8306281068907748,
+      "epoch": 0.8582292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.4672605211007518e-06,
-      "loss": 0.7822,
+      "learning_rate": 1.0357995604953597e-06,
+      "loss": 0.8366,
       "step": 30244
     },
     {
-      "epoch": 0.830655571118618,
+      "epoch": 0.8582576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.4667967052472143e-06,
-      "loss": 0.7102,
+      "learning_rate": 1.0353922579123765e-06,
+      "loss": 0.7174,
       "step": 30245
     },
     {
-      "epoch": 0.8306830353464613,
+      "epoch": 0.8582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.4663329569116857e-06,
-      "loss": 0.8203,
+      "learning_rate": 1.034985031053839e-06,
+      "loss": 0.8398,
       "step": 30246
     },
     {
-      "epoch": 0.8307104995743044,
+      "epoch": 0.8583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.4658692760978388e-06,
-      "loss": 0.797,
+      "learning_rate": 1.0345778799231854e-06,
+      "loss": 0.8239,
       "step": 30247
     },
     {
-      "epoch": 0.8307379638021477,
+      "epoch": 0.8583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.4654056628093383e-06,
-      "loss": 0.7654,
+      "learning_rate": 1.0341708045238553e-06,
+      "loss": 0.7521,
       "step": 30248
     },
     {
-      "epoch": 0.830765428029991,
+      "epoch": 0.8583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.4649421170498567e-06,
-      "loss": 0.8484,
+      "learning_rate": 1.033763804859289e-06,
+      "loss": 0.711,
       "step": 30249
     },
     {
-      "epoch": 0.8307928922578341,
+      "epoch": 0.8583995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.464478638823056e-06,
-      "loss": 0.707,
+      "learning_rate": 1.0333568809329219e-06,
+      "loss": 0.7376,
       "step": 30250
     },
     {
-      "epoch": 0.8308203564856774,
+      "epoch": 0.8584279228149829,
       "grad_norm": 0.0,
-      "learning_rate": 1.464015228132607e-06,
-      "loss": 0.812,
+      "learning_rate": 1.0329500327481924e-06,
+      "loss": 0.8894,
       "step": 30251
     },
     {
-      "epoch": 0.8308478207135206,
+      "epoch": 0.8584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.4635518849821762e-06,
-      "loss": 0.7647,
+      "learning_rate": 1.0325432603085384e-06,
+      "loss": 0.7317,
       "step": 30252
     },
     {
-      "epoch": 0.8308752849413639,
+      "epoch": 0.8584846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.4630886093754316e-06,
-      "loss": 0.8651,
+      "learning_rate": 1.032136563617392e-06,
+      "loss": 0.7145,
       "step": 30253
     },
     {
-      "epoch": 0.8309027491692071,
+      "epoch": 0.8585130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.4626254013160347e-06,
-      "loss": 0.7924,
+      "learning_rate": 1.0317299426781924e-06,
+      "loss": 0.766,
       "step": 30254
     },
     {
-      "epoch": 0.8309302133970503,
+      "epoch": 0.8585414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.4621622608076536e-06,
-      "loss": 0.8105,
+      "learning_rate": 1.0313233974943748e-06,
+      "loss": 0.7903,
       "step": 30255
     },
     {
-      "epoch": 0.8309576776248936,
+      "epoch": 0.8585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.4616991878539511e-06,
-      "loss": 0.8312,
+      "learning_rate": 1.0309169280693698e-06,
+      "loss": 0.8231,
       "step": 30256
     },
     {
-      "epoch": 0.8309851418527369,
+      "epoch": 0.8585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.461236182458594e-06,
-      "loss": 0.7891,
+      "learning_rate": 1.0305105344066125e-06,
+      "loss": 0.7363,
       "step": 30257
     },
     {
-      "epoch": 0.83101260608058,
+      "epoch": 0.8586265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.460773244625242e-06,
-      "loss": 0.765,
+      "learning_rate": 1.0301042165095377e-06,
+      "loss": 0.8205,
       "step": 30258
     },
     {
-      "epoch": 0.8310400703084233,
+      "epoch": 0.8586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.460310374357562e-06,
-      "loss": 0.8299,
+      "learning_rate": 1.0296979743815739e-06,
+      "loss": 0.8538,
       "step": 30259
     },
     {
-      "epoch": 0.8310675345362665,
+      "epoch": 0.858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.459847571659212e-06,
-      "loss": 0.8833,
+      "learning_rate": 1.0292918080261538e-06,
+      "loss": 0.7704,
       "step": 30260
     },
     {
-      "epoch": 0.8310949987641097,
+      "epoch": 0.8587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.4593848365338559e-06,
-      "loss": 0.8265,
+      "learning_rate": 1.0288857174467127e-06,
+      "loss": 0.6984,
       "step": 30261
     },
     {
-      "epoch": 0.831122462991953,
+      "epoch": 0.8587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.458922168985155e-06,
-      "loss": 0.7561,
+      "learning_rate": 1.028479702646672e-06,
+      "loss": 0.6125,
       "step": 30262
     },
     {
-      "epoch": 0.8311499272197962,
+      "epoch": 0.8587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.4584595690167713e-06,
-      "loss": 0.8254,
+      "learning_rate": 1.028073763629469e-06,
+      "loss": 0.816,
       "step": 30263
     },
     {
-      "epoch": 0.8311773914476395,
+      "epoch": 0.8587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.457997036632366e-06,
-      "loss": 0.6868,
+      "learning_rate": 1.0276679003985323e-06,
+      "loss": 0.78,
       "step": 30264
     },
     {
-      "epoch": 0.8312048556754826,
+      "epoch": 0.8588251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.4575345718355971e-06,
-      "loss": 0.7526,
+      "learning_rate": 1.0272621129572858e-06,
+      "loss": 0.8352,
       "step": 30265
     },
     {
-      "epoch": 0.8312323199033259,
+      "epoch": 0.8588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.457072174630122e-06,
-      "loss": 0.8011,
+      "learning_rate": 1.0268564013091598e-06,
+      "loss": 0.8116,
       "step": 30266
     },
     {
-      "epoch": 0.8312597841311692,
+      "epoch": 0.8588819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.4566098450195997e-06,
-      "loss": 0.8289,
+      "learning_rate": 1.026450765457584e-06,
+      "loss": 0.8311,
       "step": 30267
     },
     {
-      "epoch": 0.8312872483590124,
+      "epoch": 0.8589103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.4561475830076909e-06,
-      "loss": 0.7752,
+      "learning_rate": 1.0260452054059788e-06,
+      "loss": 0.7205,
       "step": 30268
     },
     {
-      "epoch": 0.8313147125868556,
+      "epoch": 0.8589387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 1.4556853885980515e-06,
-      "loss": 0.8303,
+      "learning_rate": 1.025639721157775e-06,
+      "loss": 0.7022,
       "step": 30269
     },
     {
-      "epoch": 0.8313421768146989,
+      "epoch": 0.8589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.4552232617943408e-06,
-      "loss": 0.7675,
+      "learning_rate": 1.0252343127163943e-06,
+      "loss": 0.6886,
       "step": 30270
     },
     {
-      "epoch": 0.8313696410425421,
+      "epoch": 0.8589954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.4547612026002123e-06,
-      "loss": 0.7587,
+      "learning_rate": 1.024828980085264e-06,
+      "loss": 0.7745,
       "step": 30271
     },
     {
-      "epoch": 0.8313971052703854,
+      "epoch": 0.8590238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 1.4542992110193234e-06,
-      "loss": 0.7859,
+      "learning_rate": 1.0244237232678067e-06,
+      "loss": 0.8367,
       "step": 30272
     },
     {
-      "epoch": 0.8314245694982285,
+      "epoch": 0.8590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.4538372870553307e-06,
-      "loss": 0.8178,
+      "learning_rate": 1.0240185422674453e-06,
+      "loss": 0.7615,
       "step": 30273
     },
     {
-      "epoch": 0.8314520337260718,
+      "epoch": 0.8590805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.4533754307118865e-06,
-      "loss": 0.9069,
+      "learning_rate": 1.0236134370876016e-06,
+      "loss": 0.7852,
       "step": 30274
     },
     {
-      "epoch": 0.8314794979539151,
+      "epoch": 0.8591089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.4529136419926459e-06,
-      "loss": 0.8493,
+      "learning_rate": 1.0232084077317017e-06,
+      "loss": 0.8626,
       "step": 30275
     },
     {
-      "epoch": 0.8315069621817582,
+      "epoch": 0.8591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.4524519209012654e-06,
-      "loss": 0.8822,
+      "learning_rate": 1.0228034542031617e-06,
+      "loss": 0.8388,
       "step": 30276
     },
     {
-      "epoch": 0.8315344264096015,
+      "epoch": 0.8591657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.4519902674413944e-06,
-      "loss": 0.713,
+      "learning_rate": 1.0223985765054024e-06,
+      "loss": 0.8754,
       "step": 30277
     },
     {
-      "epoch": 0.8315618906374447,
+      "epoch": 0.859194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.4515286816166884e-06,
-      "loss": 0.8746,
+      "learning_rate": 1.0219937746418496e-06,
+      "loss": 0.8258,
       "step": 30278
     },
     {
-      "epoch": 0.831589354865288,
+      "epoch": 0.85922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.4510671634307971e-06,
-      "loss": 0.694,
+      "learning_rate": 1.0215890486159174e-06,
+      "loss": 0.8535,
       "step": 30279
     },
     {
-      "epoch": 0.8316168190931312,
+      "epoch": 0.8592508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.4506057128873751e-06,
-      "loss": 0.869,
+      "learning_rate": 1.0211843984310254e-06,
+      "loss": 0.8165,
       "step": 30280
     },
     {
-      "epoch": 0.8316442833209744,
+      "epoch": 0.8592792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.4501443299900753e-06,
-      "loss": 0.7989,
+      "learning_rate": 1.0207798240905954e-06,
+      "loss": 0.8407,
       "step": 30281
     },
     {
-      "epoch": 0.8316717475488177,
+      "epoch": 0.8593076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.4496830147425445e-06,
-      "loss": 0.775,
+      "learning_rate": 1.0203753255980387e-06,
+      "loss": 0.8232,
       "step": 30282
     },
     {
-      "epoch": 0.831699211776661,
+      "epoch": 0.8593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.449221767148431e-06,
-      "loss": 0.8297,
+      "learning_rate": 1.0199709029567761e-06,
+      "loss": 0.7412,
       "step": 30283
     },
     {
-      "epoch": 0.8317266760045041,
+      "epoch": 0.8593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.4487605872113886e-06,
-      "loss": 0.7578,
+      "learning_rate": 1.019566556170225e-06,
+      "loss": 0.7872,
       "step": 30284
     },
     {
-      "epoch": 0.8317541402323474,
+      "epoch": 0.8593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.4482994749350642e-06,
-      "loss": 0.7157,
+      "learning_rate": 1.019162285241796e-06,
+      "loss": 0.7387,
       "step": 30285
     },
     {
-      "epoch": 0.8317816044601906,
+      "epoch": 0.8594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.4478384303231064e-06,
-      "loss": 0.6827,
+      "learning_rate": 1.0187580901749084e-06,
+      "loss": 0.7625,
       "step": 30286
     },
     {
-      "epoch": 0.8318090686880338,
+      "epoch": 0.8594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.447377453379164e-06,
-      "loss": 0.8605,
+      "learning_rate": 1.0183539709729761e-06,
+      "loss": 0.9139,
       "step": 30287
     },
     {
-      "epoch": 0.8318365329158771,
+      "epoch": 0.859477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.4469165441068866e-06,
-      "loss": 0.8702,
+      "learning_rate": 1.0179499276394088e-06,
+      "loss": 0.7673,
       "step": 30288
     },
     {
-      "epoch": 0.8318639971437203,
+      "epoch": 0.8595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.446455702509919e-06,
-      "loss": 0.8612,
+      "learning_rate": 1.017545960177624e-06,
+      "loss": 0.7388,
       "step": 30289
     },
     {
-      "epoch": 0.8318914613715636,
+      "epoch": 0.8595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.445994928591905e-06,
-      "loss": 0.8574,
+      "learning_rate": 1.0171420685910328e-06,
+      "loss": 0.7676,
       "step": 30290
     },
     {
-      "epoch": 0.8319189255994067,
+      "epoch": 0.8595629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.4455342223564916e-06,
-      "loss": 0.7907,
+      "learning_rate": 1.016738252883045e-06,
+      "loss": 0.819,
       "step": 30291
     },
     {
-      "epoch": 0.83194638982725,
+      "epoch": 0.8595913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.445073583807326e-06,
-      "loss": 0.7513,
+      "learning_rate": 1.0163345130570723e-06,
+      "loss": 0.7876,
       "step": 30292
     },
     {
-      "epoch": 0.8319738540550933,
+      "epoch": 0.8596197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.4446130129480519e-06,
-      "loss": 0.7931,
+      "learning_rate": 1.0159308491165264e-06,
+      "loss": 0.8331,
       "step": 30293
     },
     {
-      "epoch": 0.8320013182829364,
+      "epoch": 0.8596481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.444152509782316e-06,
-      "loss": 0.8172,
+      "learning_rate": 1.015527261064817e-06,
+      "loss": 0.798,
       "step": 30294
     },
     {
-      "epoch": 0.8320287825107797,
+      "epoch": 0.8596765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.4436920743137572e-06,
-      "loss": 0.7521,
+      "learning_rate": 1.0151237489053545e-06,
+      "loss": 0.7333,
       "step": 30295
     },
     {
-      "epoch": 0.832056246738623,
+      "epoch": 0.8597048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 1.443231706546021e-06,
-      "loss": 0.7343,
+      "learning_rate": 1.0147203126415427e-06,
+      "loss": 0.745,
       "step": 30296
     },
     {
-      "epoch": 0.8320837109664662,
+      "epoch": 0.859733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.4427714064827525e-06,
-      "loss": 0.8504,
+      "learning_rate": 1.0143169522767926e-06,
+      "loss": 0.8654,
       "step": 30297
     },
     {
-      "epoch": 0.8321111751943094,
+      "epoch": 0.8597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.4423111741275897e-06,
-      "loss": 0.6905,
+      "learning_rate": 1.0139136678145133e-06,
+      "loss": 0.7656,
       "step": 30298
     },
     {
-      "epoch": 0.8321386394221526,
+      "epoch": 0.8597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.4418510094841742e-06,
-      "loss": 0.8442,
+      "learning_rate": 1.013510459258108e-06,
+      "loss": 0.8795,
       "step": 30299
     },
     {
-      "epoch": 0.8321661036499959,
+      "epoch": 0.8598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.441390912556152e-06,
-      "loss": 0.7559,
+      "learning_rate": 1.0131073266109826e-06,
+      "loss": 0.8599,
       "step": 30300
     },
     {
-      "epoch": 0.8321935678778392,
+      "epoch": 0.8598467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.440930883347158e-06,
-      "loss": 0.8209,
+      "learning_rate": 1.0127042698765465e-06,
+      "loss": 0.8456,
       "step": 30301
     },
     {
-      "epoch": 0.8322210321056823,
+      "epoch": 0.8598751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.4404709218608348e-06,
-      "loss": 0.7973,
+      "learning_rate": 1.0123012890581985e-06,
+      "loss": 0.9191,
       "step": 30302
     },
     {
-      "epoch": 0.8322484963335256,
+      "epoch": 0.8599035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.4400110281008206e-06,
-      "loss": 0.8378,
+      "learning_rate": 1.0118983841593467e-06,
+      "loss": 0.7082,
       "step": 30303
     },
     {
-      "epoch": 0.8322759605613688,
+      "epoch": 0.8599318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.4395512020707569e-06,
-      "loss": 0.8192,
+      "learning_rate": 1.011495555183395e-06,
+      "loss": 0.7877,
       "step": 30304
     },
     {
-      "epoch": 0.832303424789212,
+      "epoch": 0.8599602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.4390914437742786e-06,
-      "loss": 0.7397,
+      "learning_rate": 1.011092802133742e-06,
+      "loss": 0.8803,
       "step": 30305
     },
     {
-      "epoch": 0.8323308890170553,
+      "epoch": 0.8599886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 1.4386317532150273e-06,
-      "loss": 0.7495,
+      "learning_rate": 1.0106901250137924e-06,
+      "loss": 0.8368,
       "step": 30306
     },
     {
-      "epoch": 0.8323583532448985,
+      "epoch": 0.860017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.4381721303966346e-06,
-      "loss": 0.8373,
+      "learning_rate": 1.0102875238269494e-06,
+      "loss": 0.7628,
       "step": 30307
     },
     {
-      "epoch": 0.8323858174727418,
+      "epoch": 0.8600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.4377125753227416e-06,
-      "loss": 0.7752,
+      "learning_rate": 1.0098849985766067e-06,
+      "loss": 0.7524,
       "step": 30308
     },
     {
-      "epoch": 0.832413281700585,
+      "epoch": 0.8600737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.4372530879969825e-06,
-      "loss": 0.8296,
+      "learning_rate": 1.0094825492661754e-06,
+      "loss": 0.7019,
       "step": 30309
     },
     {
-      "epoch": 0.8324407459284282,
+      "epoch": 0.8601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.4367936684229954e-06,
-      "loss": 0.8273,
+      "learning_rate": 1.0090801758990465e-06,
+      "loss": 0.8515,
       "step": 30310
     },
     {
-      "epoch": 0.8324682101562715,
+      "epoch": 0.8601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.4363343166044119e-06,
-      "loss": 0.9475,
+      "learning_rate": 1.008677878478621e-06,
+      "loss": 0.8965,
       "step": 30311
     },
     {
-      "epoch": 0.8324956743841146,
+      "epoch": 0.8601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.4358750325448723e-06,
-      "loss": 0.8375,
+      "learning_rate": 1.0082756570082997e-06,
+      "loss": 0.8235,
       "step": 30312
     },
     {
-      "epoch": 0.8325231386119579,
+      "epoch": 0.8601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.4354158162480058e-06,
-      "loss": 0.7737,
+      "learning_rate": 1.0078735114914761e-06,
+      "loss": 0.7856,
       "step": 30313
     },
     {
-      "epoch": 0.8325506028398012,
+      "epoch": 0.8602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.4349566677174454e-06,
-      "loss": 0.8137,
+      "learning_rate": 1.00747144193155e-06,
+      "loss": 0.8107,
       "step": 30314
     },
     {
-      "epoch": 0.8325780670676444,
+      "epoch": 0.860244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.4344975869568257e-06,
-      "loss": 0.908,
+      "learning_rate": 1.0070694483319187e-06,
+      "loss": 0.8241,
       "step": 30315
     },
     {
-      "epoch": 0.8326055312954876,
+      "epoch": 0.860272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.4340385739697793e-06,
-      "loss": 0.7731,
+      "learning_rate": 1.006667530695974e-06,
+      "loss": 0.8154,
       "step": 30316
     },
     {
-      "epoch": 0.8326329955233308,
+      "epoch": 0.8603007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.433579628759937e-06,
-      "loss": 0.7732,
+      "learning_rate": 1.0062656890271116e-06,
+      "loss": 0.7921,
       "step": 30317
     },
     {
-      "epoch": 0.8326604597511741,
+      "epoch": 0.8603291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.4331207513309331e-06,
-      "loss": 0.765,
+      "learning_rate": 1.0058639233287304e-06,
+      "loss": 0.8684,
       "step": 30318
     },
     {
-      "epoch": 0.8326879239790174,
+      "epoch": 0.8603575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.4326619416863952e-06,
-      "loss": 0.8614,
+      "learning_rate": 1.0054622336042198e-06,
+      "loss": 0.7991,
       "step": 30319
     },
     {
-      "epoch": 0.8327153882068605,
+      "epoch": 0.8603859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.4322031998299535e-06,
-      "loss": 0.84,
+      "learning_rate": 1.0050606198569723e-06,
+      "loss": 0.7494,
       "step": 30320
     },
     {
-      "epoch": 0.8327428524347038,
+      "epoch": 0.8604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.4317445257652418e-06,
-      "loss": 0.8555,
+      "learning_rate": 1.0046590820903845e-06,
+      "loss": 0.7775,
       "step": 30321
     },
     {
-      "epoch": 0.8327703166625471,
+      "epoch": 0.8604426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.4312859194958838e-06,
-      "loss": 0.7332,
+      "learning_rate": 1.0042576203078447e-06,
+      "loss": 0.7809,
       "step": 30322
     },
     {
-      "epoch": 0.8327977808903902,
+      "epoch": 0.8604710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 1.4308273810255114e-06,
-      "loss": 0.8721,
+      "learning_rate": 1.0038562345127445e-06,
+      "loss": 0.7913,
       "step": 30323
     },
     {
-      "epoch": 0.8328252451182335,
+      "epoch": 0.8604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.4303689103577545e-06,
-      "loss": 0.785,
+      "learning_rate": 1.0034549247084757e-06,
+      "loss": 0.8492,
       "step": 30324
     },
     {
-      "epoch": 0.8328527093460767,
+      "epoch": 0.860527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.4299105074962361e-06,
-      "loss": 0.8793,
+      "learning_rate": 1.0030536908984268e-06,
+      "loss": 0.8247,
       "step": 30325
     },
     {
-      "epoch": 0.83288017357392,
+      "epoch": 0.8605561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 1.4294521724445843e-06,
-      "loss": 0.8198,
+      "learning_rate": 1.0026525330859904e-06,
+      "loss": 0.7455,
       "step": 30326
     },
     {
-      "epoch": 0.8329076378017632,
+      "epoch": 0.8605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.4289939052064283e-06,
-      "loss": 0.8665,
+      "learning_rate": 1.0022514512745495e-06,
+      "loss": 0.8012,
       "step": 30327
     },
     {
-      "epoch": 0.8329351020296064,
+      "epoch": 0.8606129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.4285357057853933e-06,
-      "loss": 0.8314,
+      "learning_rate": 1.001850445467496e-06,
+      "loss": 0.8147,
       "step": 30328
     },
     {
-      "epoch": 0.8329625662574497,
+      "epoch": 0.8606413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.4280775741851026e-06,
-      "loss": 0.8811,
+      "learning_rate": 1.001449515668218e-06,
+      "loss": 0.8292,
       "step": 30329
     },
     {
-      "epoch": 0.8329900304852929,
+      "epoch": 0.8606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.427619510409185e-06,
-      "loss": 0.9102,
+      "learning_rate": 1.0010486618800986e-06,
+      "loss": 0.7278,
       "step": 30330
     },
     {
-      "epoch": 0.8330174947131361,
+      "epoch": 0.8606980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.4271615144612605e-06,
-      "loss": 0.7865,
+      "learning_rate": 1.0006478841065259e-06,
+      "loss": 0.8385,
       "step": 30331
     },
     {
-      "epoch": 0.8330449589409794,
+      "epoch": 0.8607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.426703586344954e-06,
-      "loss": 0.8538,
+      "learning_rate": 1.0002471823508864e-06,
+      "loss": 0.8043,
       "step": 30332
     },
     {
-      "epoch": 0.8330724231688226,
+      "epoch": 0.8607548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.4262457260638907e-06,
-      "loss": 0.8486,
+      "learning_rate": 9.998465566165627e-07,
+      "loss": 0.7102,
       "step": 30333
     },
     {
-      "epoch": 0.8330998873966659,
+      "epoch": 0.860783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.4257879336216918e-06,
-      "loss": 0.7916,
+      "learning_rate": 9.994460069069389e-07,
+      "loss": 0.8361,
       "step": 30334
     },
     {
-      "epoch": 0.8331273516245091,
+      "epoch": 0.8608115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.4253302090219823e-06,
-      "loss": 0.8126,
+      "learning_rate": 9.990455332254012e-07,
+      "loss": 0.6816,
       "step": 30335
     },
     {
-      "epoch": 0.8331548158523523,
+      "epoch": 0.8608399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.4248725522683803e-06,
-      "loss": 0.7888,
+      "learning_rate": 9.986451355753279e-07,
+      "loss": 0.8423,
       "step": 30336
     },
     {
-      "epoch": 0.8331822800801956,
+      "epoch": 0.8608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.4244149633645099e-06,
-      "loss": 0.8546,
+      "learning_rate": 9.982448139601053e-07,
+      "loss": 0.8395,
       "step": 30337
     },
     {
-      "epoch": 0.8332097443080387,
+      "epoch": 0.8608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.423957442313988e-06,
-      "loss": 0.7961,
+      "learning_rate": 9.97844568383114e-07,
+      "loss": 0.8001,
       "step": 30338
     },
     {
-      "epoch": 0.833237208535882,
+      "epoch": 0.8609250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.4234999891204359e-06,
-      "loss": 0.9014,
+      "learning_rate": 9.974443988477312e-07,
+      "loss": 0.7749,
       "step": 30339
     },
     {
-      "epoch": 0.8332646727637253,
+      "epoch": 0.8609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.4230426037874756e-06,
-      "loss": 0.7219,
+      "learning_rate": 9.970443053573442e-07,
+      "loss": 0.8149,
       "step": 30340
     },
     {
-      "epoch": 0.8332921369915685,
+      "epoch": 0.8609818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.4225852863187261e-06,
-      "loss": 0.9136,
+      "learning_rate": 9.96644287915326e-07,
+      "loss": 0.8247,
       "step": 30341
     },
     {
-      "epoch": 0.8333196012194117,
+      "epoch": 0.8610102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 1.4221280367178026e-06,
-      "loss": 0.8619,
+      "learning_rate": 9.962443465250592e-07,
+      "loss": 0.8578,
       "step": 30342
     },
     {
-      "epoch": 0.8333470654472549,
+      "epoch": 0.861038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.421670854988324e-06,
-      "loss": 0.8543,
+      "learning_rate": 9.958444811899225e-07,
+      "loss": 0.8276,
       "step": 30343
     },
     {
-      "epoch": 0.8333745296750982,
+      "epoch": 0.861066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.421213741133909e-06,
-      "loss": 0.8424,
+      "learning_rate": 9.954446919132898e-07,
+      "loss": 0.8404,
       "step": 30344
     },
     {
-      "epoch": 0.8334019939029415,
+      "epoch": 0.8610953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.4207566951581764e-06,
-      "loss": 0.775,
+      "learning_rate": 9.950449786985416e-07,
+      "loss": 0.7904,
       "step": 30345
     },
     {
-      "epoch": 0.8334294581307846,
+      "epoch": 0.8611237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.4202997170647382e-06,
-      "loss": 0.8723,
+      "learning_rate": 9.946453415490542e-07,
+      "loss": 0.8413,
       "step": 30346
     },
     {
-      "epoch": 0.8334569223586279,
+      "epoch": 0.8611520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 1.4198428068572134e-06,
-      "loss": 0.8333,
+      "learning_rate": 9.942457804682015e-07,
+      "loss": 0.7923,
       "step": 30347
     },
     {
-      "epoch": 0.8334843865864712,
+      "epoch": 0.8611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.4193859645392138e-06,
-      "loss": 0.815,
+      "learning_rate": 9.938462954593587e-07,
+      "loss": 0.8295,
       "step": 30348
     },
     {
-      "epoch": 0.8335118508143143,
+      "epoch": 0.8612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.4189291901143576e-06,
-      "loss": 0.7757,
+      "learning_rate": 9.934468865259038e-07,
+      "loss": 0.9045,
       "step": 30349
     },
     {
-      "epoch": 0.8335393150421576,
+      "epoch": 0.8612372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.4184724835862561e-06,
-      "loss": 0.9114,
+      "learning_rate": 9.93047553671206e-07,
+      "loss": 0.9049,
       "step": 30350
     },
     {
-      "epoch": 0.8335667792700008,
+      "epoch": 0.8612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.4180158449585246e-06,
-      "loss": 0.7988,
+      "learning_rate": 9.926482968986396e-07,
+      "loss": 0.7451,
       "step": 30351
     },
     {
-      "epoch": 0.8335942434978441,
+      "epoch": 0.8612939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.4175592742347787e-06,
-      "loss": 0.8555,
+      "learning_rate": 9.9224911621158e-07,
+      "loss": 0.8098,
       "step": 30352
     },
     {
-      "epoch": 0.8336217077256873,
+      "epoch": 0.8613223609534619,
       "grad_norm": 0.0,
-      "learning_rate": 1.417102771418627e-06,
-      "loss": 0.7706,
+      "learning_rate": 9.918500116133957e-07,
+      "loss": 0.7762,
       "step": 30353
     },
     {
-      "epoch": 0.8336491719535305,
+      "epoch": 0.8613507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.416646336513684e-06,
-      "loss": 0.6889,
+      "learning_rate": 9.914509831074592e-07,
+      "loss": 0.7845,
       "step": 30354
     },
     {
-      "epoch": 0.8336766361813738,
+      "epoch": 0.8613791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.416189969523558e-06,
-      "loss": 0.9009,
+      "learning_rate": 9.910520306971405e-07,
+      "loss": 0.9228,
       "step": 30355
     },
     {
-      "epoch": 0.8337041004092169,
+      "epoch": 0.8614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.415733670451862e-06,
-      "loss": 0.7887,
+      "learning_rate": 9.90653154385811e-07,
+      "loss": 0.7604,
       "step": 30356
     },
     {
-      "epoch": 0.8337315646370602,
+      "epoch": 0.8614358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.4152774393022061e-06,
-      "loss": 0.8169,
+      "learning_rate": 9.902543541768407e-07,
+      "loss": 0.8426,
       "step": 30357
     },
     {
-      "epoch": 0.8337590288649035,
+      "epoch": 0.8614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.414821276078201e-06,
-      "loss": 0.8052,
+      "learning_rate": 9.898556300735951e-07,
+      "loss": 0.6944,
       "step": 30358
     },
     {
-      "epoch": 0.8337864930927467,
+      "epoch": 0.8614926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.414365180783458e-06,
-      "loss": 0.8989,
+      "learning_rate": 9.89456982079444e-07,
+      "loss": 0.7755,
       "step": 30359
     },
     {
-      "epoch": 0.8338139573205899,
+      "epoch": 0.8615209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.413909153421581e-06,
-      "loss": 0.8704,
+      "learning_rate": 9.890584101977573e-07,
+      "loss": 0.8515,
       "step": 30360
     },
     {
-      "epoch": 0.8338414215484332,
+      "epoch": 0.8615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.4134531939961826e-06,
-      "loss": 0.7942,
+      "learning_rate": 9.886599144318977e-07,
+      "loss": 0.7985,
       "step": 30361
     },
     {
-      "epoch": 0.8338688857762764,
+      "epoch": 0.861577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.4129973025108668e-06,
-      "loss": 0.8711,
+      "learning_rate": 9.88261494785232e-07,
+      "loss": 0.8807,
       "step": 30362
     },
     {
-      "epoch": 0.8338963500041197,
+      "epoch": 0.8616061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.4125414789692426e-06,
-      "loss": 0.8143,
+      "learning_rate": 9.878631512611281e-07,
+      "loss": 0.8639,
       "step": 30363
     },
     {
-      "epoch": 0.8339238142319628,
+      "epoch": 0.8616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.4120857233749162e-06,
-      "loss": 0.8431,
+      "learning_rate": 9.874648838629485e-07,
+      "loss": 0.842,
       "step": 30364
     },
     {
-      "epoch": 0.8339512784598061,
+      "epoch": 0.8616628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.4116300357314972e-06,
-      "loss": 0.7298,
+      "learning_rate": 9.870666925940576e-07,
+      "loss": 0.8526,
       "step": 30365
     },
     {
-      "epoch": 0.8339787426876494,
+      "epoch": 0.8616912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.4111744160425855e-06,
-      "loss": 0.8038,
+      "learning_rate": 9.866685774578212e-07,
+      "loss": 0.8813,
       "step": 30366
     },
     {
-      "epoch": 0.8340062069154925,
+      "epoch": 0.8617196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.4107188643117885e-06,
-      "loss": 0.8456,
+      "learning_rate": 9.862705384575988e-07,
+      "loss": 0.7932,
       "step": 30367
     },
     {
-      "epoch": 0.8340336711433358,
+      "epoch": 0.8617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.4102633805427124e-06,
-      "loss": 0.8192,
+      "learning_rate": 9.858725755967547e-07,
+      "loss": 0.784,
       "step": 30368
     },
     {
-      "epoch": 0.834061135371179,
+      "epoch": 0.8617763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.4098079647389606e-06,
-      "loss": 0.848,
+      "learning_rate": 9.854746888786493e-07,
+      "loss": 0.8767,
       "step": 30369
     },
     {
-      "epoch": 0.8340885995990223,
+      "epoch": 0.8618047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.4093526169041338e-06,
-      "loss": 0.8069,
+      "learning_rate": 9.850768783066455e-07,
+      "loss": 0.7771,
       "step": 30370
     },
     {
-      "epoch": 0.8341160638268655,
+      "epoch": 0.8618331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 1.4088973370418391e-06,
-      "loss": 0.7847,
+      "learning_rate": 9.84679143884102e-07,
+      "loss": 0.8429,
       "step": 30371
     },
     {
-      "epoch": 0.8341435280547087,
+      "epoch": 0.8618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.4084421251556747e-06,
-      "loss": 0.7653,
+      "learning_rate": 9.8428148561438e-07,
+      "loss": 0.6946,
       "step": 30372
     },
     {
-      "epoch": 0.834170992282552,
+      "epoch": 0.8618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.407986981249243e-06,
-      "loss": 0.8322,
+      "learning_rate": 9.838839035008374e-07,
+      "loss": 0.8295,
       "step": 30373
     },
     {
-      "epoch": 0.8341984565103953,
+      "epoch": 0.8619182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 1.4075319053261482e-06,
-      "loss": 0.8362,
+      "learning_rate": 9.834863975468323e-07,
+      "loss": 0.8185,
       "step": 30374
     },
     {
-      "epoch": 0.8342259207382384,
+      "epoch": 0.8619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.4070768973899884e-06,
-      "loss": 0.7857,
+      "learning_rate": 9.830889677557243e-07,
+      "loss": 0.8681,
       "step": 30375
     },
     {
-      "epoch": 0.8342533849660817,
+      "epoch": 0.8619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.4066219574443674e-06,
-      "loss": 0.6893,
+      "learning_rate": 9.826916141308674e-07,
+      "loss": 0.8571,
       "step": 30376
     },
     {
-      "epoch": 0.8342808491939249,
+      "epoch": 0.8620034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.4061670854928788e-06,
-      "loss": 0.8473,
+      "learning_rate": 9.822943366756222e-07,
+      "loss": 0.795,
       "step": 30377
     },
     {
-      "epoch": 0.8343083134217681,
+      "epoch": 0.8620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.4057122815391288e-06,
-      "loss": 0.8747,
+      "learning_rate": 9.818971353933394e-07,
+      "loss": 0.756,
       "step": 30378
     },
     {
-      "epoch": 0.8343357776496114,
+      "epoch": 0.8620601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.40525754558671e-06,
-      "loss": 0.828,
+      "learning_rate": 9.815000102873772e-07,
+      "loss": 0.8257,
       "step": 30379
     },
     {
-      "epoch": 0.8343632418774546,
+      "epoch": 0.862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.404802877639222e-06,
-      "loss": 0.8304,
+      "learning_rate": 9.811029613610913e-07,
+      "loss": 0.7401,
       "step": 30380
     },
     {
-      "epoch": 0.8343907061052979,
+      "epoch": 0.862116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.4043482777002637e-06,
-      "loss": 0.8331,
+      "learning_rate": 9.807059886178327e-07,
+      "loss": 0.8455,
       "step": 30381
     },
     {
-      "epoch": 0.834418170333141,
+      "epoch": 0.8621452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 1.4038937457734314e-06,
-      "loss": 0.7025,
+      "learning_rate": 9.80309092060956e-07,
+      "loss": 0.7878,
       "step": 30382
     },
     {
-      "epoch": 0.8344456345609843,
+      "epoch": 0.8621736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.4034392818623232e-06,
-      "loss": 0.8662,
+      "learning_rate": 9.799122716938158e-07,
+      "loss": 0.7779,
       "step": 30383
     },
     {
-      "epoch": 0.8344730987888276,
+      "epoch": 0.8622020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.402984885970532e-06,
-      "loss": 0.8367,
+      "learning_rate": 9.79515527519761e-07,
+      "loss": 0.6936,
       "step": 30384
     },
     {
-      "epoch": 0.8345005630166707,
+      "epoch": 0.8622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.4025305581016569e-06,
-      "loss": 0.8273,
+      "learning_rate": 9.791188595421409e-07,
+      "loss": 0.8165,
       "step": 30385
     },
     {
-      "epoch": 0.834528027244514,
+      "epoch": 0.8622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.4020762982592873e-06,
-      "loss": 0.8403,
+      "learning_rate": 9.787222677643139e-07,
+      "loss": 0.7582,
       "step": 30386
     },
     {
-      "epoch": 0.8345554914723573,
+      "epoch": 0.8622871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.4016221064470214e-06,
-      "loss": 0.8723,
+      "learning_rate": 9.783257521896228e-07,
+      "loss": 0.8646,
       "step": 30387
     },
     {
-      "epoch": 0.8345829557002005,
+      "epoch": 0.8623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.4011679826684531e-06,
-      "loss": 0.8129,
+      "learning_rate": 9.779293128214206e-07,
+      "loss": 0.8655,
       "step": 30388
     },
     {
-      "epoch": 0.8346104199280437,
+      "epoch": 0.8623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.400713926927173e-06,
-      "loss": 0.8531,
+      "learning_rate": 9.775329496630559e-07,
+      "loss": 0.8589,
       "step": 30389
     },
     {
-      "epoch": 0.8346378841558869,
+      "epoch": 0.862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.4002599392267757e-06,
-      "loss": 0.7759,
+      "learning_rate": 9.771366627178748e-07,
+      "loss": 0.8507,
       "step": 30390
     },
     {
-      "epoch": 0.8346653483837302,
+      "epoch": 0.8624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.399806019570853e-06,
-      "loss": 0.7365,
+      "learning_rate": 9.767404519892254e-07,
+      "loss": 0.7369,
       "step": 30391
     },
     {
-      "epoch": 0.8346928126115735,
+      "epoch": 0.8624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.3993521679629961e-06,
-      "loss": 0.8447,
+      "learning_rate": 9.763443174804576e-07,
+      "loss": 0.7855,
       "step": 30392
     },
     {
-      "epoch": 0.8347202768394166,
+      "epoch": 0.8624574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.3988983844067983e-06,
-      "loss": 0.9037,
+      "learning_rate": 9.759482591949132e-07,
+      "loss": 0.8177,
       "step": 30393
     },
     {
-      "epoch": 0.8347477410672599,
+      "epoch": 0.8624858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.3984446689058495e-06,
-      "loss": 0.9141,
+      "learning_rate": 9.755522771359382e-07,
+      "loss": 0.7366,
       "step": 30394
     },
     {
-      "epoch": 0.8347752052951031,
+      "epoch": 0.8625141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.3979910214637348e-06,
-      "loss": 0.845,
+      "learning_rate": 9.751563713068824e-07,
+      "loss": 0.8682,
       "step": 30395
     },
     {
-      "epoch": 0.8348026695229464,
+      "epoch": 0.8625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.3975374420840482e-06,
-      "loss": 0.8695,
+      "learning_rate": 9.747605417110828e-07,
+      "loss": 0.8215,
       "step": 30396
     },
     {
-      "epoch": 0.8348301337507896,
+      "epoch": 0.8625709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.3970839307703776e-06,
-      "loss": 0.7598,
+      "learning_rate": 9.74364788351888e-07,
+      "loss": 0.8355,
       "step": 30397
     },
     {
-      "epoch": 0.8348575979786328,
+      "epoch": 0.8625993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.3966304875263103e-06,
-      "loss": 0.8228,
+      "learning_rate": 9.73969111232641e-07,
+      "loss": 0.7755,
       "step": 30398
     },
     {
-      "epoch": 0.8348850622064761,
+      "epoch": 0.862627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.3961771123554358e-06,
-      "loss": 0.7906,
+      "learning_rate": 9.735735103566802e-07,
+      "loss": 0.8145,
       "step": 30399
     },
     {
-      "epoch": 0.8349125264343193,
+      "epoch": 0.8626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.395723805261343e-06,
-      "loss": 0.8344,
+      "learning_rate": 9.731779857273505e-07,
+      "loss": 0.8449,
       "step": 30400
     },
     {
-      "epoch": 0.8349399906621625,
+      "epoch": 0.8626844494892169,
       "grad_norm": 0.0,
-      "learning_rate": 1.3952705662476152e-06,
-      "loss": 0.7866,
+      "learning_rate": 9.727825373479915e-07,
+      "loss": 0.7569,
       "step": 30401
     },
     {
-      "epoch": 0.8349674548900058,
+      "epoch": 0.8627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.3948173953178411e-06,
-      "loss": 0.834,
+      "learning_rate": 9.723871652219452e-07,
+      "loss": 0.6881,
       "step": 30402
     },
     {
-      "epoch": 0.834994919117849,
+      "epoch": 0.8627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.3943642924756029e-06,
-      "loss": 0.7207,
+      "learning_rate": 9.71991869352551e-07,
+      "loss": 0.8122,
       "step": 30403
     },
     {
-      "epoch": 0.8350223833456922,
+      "epoch": 0.8627695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 1.3939112577244884e-06,
-      "loss": 0.8351,
+      "learning_rate": 9.715966497431462e-07,
+      "loss": 0.8673,
       "step": 30404
     },
     {
-      "epoch": 0.8350498475735355,
+      "epoch": 0.8627979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.3934582910680806e-06,
-      "loss": 0.8331,
+      "learning_rate": 9.712015063970703e-07,
+      "loss": 0.8234,
       "step": 30405
     },
     {
-      "epoch": 0.8350773118013787,
+      "epoch": 0.8628263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.3930053925099673e-06,
-      "loss": 0.7945,
+      "learning_rate": 9.70806439317663e-07,
+      "loss": 0.8661,
       "step": 30406
     },
     {
-      "epoch": 0.835104776029222,
+      "epoch": 0.8628547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.3925525620537285e-06,
-      "loss": 0.8548,
+      "learning_rate": 9.704114485082571e-07,
+      "loss": 0.6619,
       "step": 30407
     },
     {
-      "epoch": 0.8351322402570651,
+      "epoch": 0.8628830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.3920997997029463e-06,
-      "loss": 0.7776,
+      "learning_rate": 9.700165339721924e-07,
+      "loss": 0.7732,
       "step": 30408
     },
     {
-      "epoch": 0.8351597044849084,
+      "epoch": 0.862911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.391647105461208e-06,
-      "loss": 0.8167,
+      "learning_rate": 9.696216957128058e-07,
+      "loss": 0.7662,
       "step": 30409
     },
     {
-      "epoch": 0.8351871687127517,
+      "epoch": 0.8629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.3911944793320897e-06,
-      "loss": 0.7437,
+      "learning_rate": 9.692269337334282e-07,
+      "loss": 0.8881,
       "step": 30410
     },
     {
-      "epoch": 0.8352146329405948,
+      "epoch": 0.8629682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.3907419213191754e-06,
-      "loss": 0.7775,
+      "learning_rate": 9.688322480373967e-07,
+      "loss": 0.9023,
       "step": 30411
     },
     {
-      "epoch": 0.8352420971684381,
+      "epoch": 0.8629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.3902894314260484e-06,
-      "loss": 0.7876,
+      "learning_rate": 9.684376386280481e-07,
+      "loss": 0.8221,
       "step": 30412
     },
     {
-      "epoch": 0.8352695613962814,
+      "epoch": 0.8630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.3898370096562853e-06,
-      "loss": 0.785,
+      "learning_rate": 9.680431055087103e-07,
+      "loss": 0.892,
       "step": 30413
     },
     {
-      "epoch": 0.8352970256241246,
+      "epoch": 0.8630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.3893846560134649e-06,
-      "loss": 0.841,
+      "learning_rate": 9.676486486827186e-07,
+      "loss": 0.8571,
       "step": 30414
     },
     {
-      "epoch": 0.8353244898519678,
+      "epoch": 0.8630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 1.3889323705011704e-06,
-      "loss": 0.8389,
+      "learning_rate": 9.672542681534035e-07,
+      "loss": 0.8638,
       "step": 30415
     },
     {
-      "epoch": 0.835351954079811,
+      "epoch": 0.8631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.3884801531229774e-06,
-      "loss": 0.9141,
+      "learning_rate": 9.668599639240993e-07,
+      "loss": 0.8292,
       "step": 30416
     },
     {
-      "epoch": 0.8353794183076543,
+      "epoch": 0.863138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.3880280038824679e-06,
-      "loss": 0.7947,
+      "learning_rate": 9.664657359981355e-07,
+      "loss": 0.8043,
       "step": 30417
     },
     {
-      "epoch": 0.8354068825354976,
+      "epoch": 0.86316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.3875759227832164e-06,
-      "loss": 0.7973,
+      "learning_rate": 9.660715843788416e-07,
+      "loss": 0.7466,
       "step": 30418
     },
     {
-      "epoch": 0.8354343467633407,
+      "epoch": 0.8631952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.3871239098287991e-06,
-      "loss": 0.7745,
+      "learning_rate": 9.656775090695458e-07,
+      "loss": 0.6932,
       "step": 30419
     },
     {
-      "epoch": 0.835461810991184,
+      "epoch": 0.8632236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 1.3866719650227933e-06,
-      "loss": 0.7735,
+      "learning_rate": 9.652835100735803e-07,
+      "loss": 0.7987,
       "step": 30420
     },
     {
-      "epoch": 0.8354892752190273,
+      "epoch": 0.8632519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.3862200883687748e-06,
-      "loss": 0.9046,
+      "learning_rate": 9.648895873942698e-07,
+      "loss": 0.868,
       "step": 30421
     },
     {
-      "epoch": 0.8355167394468704,
+      "epoch": 0.8632803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 1.3857682798703197e-06,
-      "loss": 0.7784,
+      "learning_rate": 9.64495741034942e-07,
+      "loss": 0.9142,
       "step": 30422
     },
     {
-      "epoch": 0.8355442036747137,
+      "epoch": 0.8633087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 1.3853165395310031e-06,
-      "loss": 0.8267,
+      "learning_rate": 9.641019709989275e-07,
+      "loss": 0.8613,
       "step": 30423
     },
     {
-      "epoch": 0.8355716679025569,
+      "epoch": 0.8633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.3848648673544018e-06,
-      "loss": 0.7787,
+      "learning_rate": 9.637082772895479e-07,
+      "loss": 0.7293,
       "step": 30424
     },
     {
-      "epoch": 0.8355991321304002,
+      "epoch": 0.8633654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 1.3844132633440866e-06,
-      "loss": 0.8602,
+      "learning_rate": 9.633146599101306e-07,
+      "loss": 0.7207,
       "step": 30425
     },
     {
-      "epoch": 0.8356265963582434,
+      "epoch": 0.8633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.3839617275036278e-06,
-      "loss": 0.7048,
+      "learning_rate": 9.629211188640031e-07,
+      "loss": 0.8029,
       "step": 30426
     },
     {
-      "epoch": 0.8356540605860866,
+      "epoch": 0.863422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.3835102598366024e-06,
-      "loss": 0.805,
+      "learning_rate": 9.625276541544849e-07,
+      "loss": 0.8438,
       "step": 30427
     },
     {
-      "epoch": 0.8356815248139299,
+      "epoch": 0.8634506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 1.383058860346581e-06,
-      "loss": 0.8008,
+      "learning_rate": 9.62134265784901e-07,
+      "loss": 0.8318,
       "step": 30428
     },
     {
-      "epoch": 0.835708989041773,
+      "epoch": 0.8634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.3826075290371367e-06,
-      "loss": 0.7021,
+      "learning_rate": 9.617409537585776e-07,
+      "loss": 0.7213,
       "step": 30429
     },
     {
-      "epoch": 0.8357364532696163,
+      "epoch": 0.8635073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.3821562659118404e-06,
-      "loss": 0.8363,
+      "learning_rate": 9.613477180788332e-07,
+      "loss": 0.826,
       "step": 30430
     },
     {
-      "epoch": 0.8357639174974596,
+      "epoch": 0.8635357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 1.3817050709742607e-06,
-      "loss": 0.8854,
+      "learning_rate": 9.609545587489911e-07,
+      "loss": 0.8558,
       "step": 30431
     },
     {
-      "epoch": 0.8357913817253028,
+      "epoch": 0.8635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 1.3812539442279681e-06,
-      "loss": 0.7083,
+      "learning_rate": 9.605614757723713e-07,
+      "loss": 0.8011,
       "step": 30432
     },
     {
-      "epoch": 0.835818845953146,
+      "epoch": 0.8635925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.3808028856765366e-06,
-      "loss": 0.7787,
+      "learning_rate": 9.601684691522961e-07,
+      "loss": 0.8936,
       "step": 30433
     },
     {
-      "epoch": 0.8358463101809893,
+      "epoch": 0.8636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.3803518953235285e-06,
-      "loss": 0.8085,
+      "learning_rate": 9.597755388920848e-07,
+      "loss": 0.9445,
       "step": 30434
     },
     {
-      "epoch": 0.8358737744088325,
+      "epoch": 0.8636492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.3799009731725143e-06,
-      "loss": 0.7595,
+      "learning_rate": 9.593826849950549e-07,
+      "loss": 0.7967,
       "step": 30435
     },
     {
-      "epoch": 0.8359012386366758,
+      "epoch": 0.863677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.379450119227066e-06,
-      "loss": 0.7306,
+      "learning_rate": 9.58989907464526e-07,
+      "loss": 0.8541,
       "step": 30436
     },
     {
-      "epoch": 0.8359287028645189,
+      "epoch": 0.8637060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.3789993334907447e-06,
-      "loss": 0.7557,
+      "learning_rate": 9.585972063038174e-07,
+      "loss": 0.854,
       "step": 30437
     },
     {
-      "epoch": 0.8359561670923622,
+      "epoch": 0.8637343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.3785486159671203e-06,
-      "loss": 0.8253,
+      "learning_rate": 9.582045815162422e-07,
+      "loss": 0.8285,
       "step": 30438
     },
     {
-      "epoch": 0.8359836313202055,
+      "epoch": 0.8637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.3780979666597593e-06,
-      "loss": 0.8506,
+      "learning_rate": 9.578120331051198e-07,
+      "loss": 0.6961,
       "step": 30439
     },
     {
-      "epoch": 0.8360110955480486,
+      "epoch": 0.8637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.3776473855722273e-06,
-      "loss": 0.8208,
+      "learning_rate": 9.57419561073768e-07,
+      "loss": 0.7484,
       "step": 30440
     },
     {
-      "epoch": 0.8360385597758919,
+      "epoch": 0.8638195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.3771968727080897e-06,
-      "loss": 0.7802,
+      "learning_rate": 9.570271654254982e-07,
+      "loss": 0.7416,
       "step": 30441
     },
     {
-      "epoch": 0.8360660240037351,
+      "epoch": 0.8638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 1.376746428070912e-06,
-      "loss": 0.8086,
+      "learning_rate": 9.566348461636254e-07,
+      "loss": 0.8277,
       "step": 30442
     },
     {
-      "epoch": 0.8360934882315784,
+      "epoch": 0.8638762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.3762960516642543e-06,
-      "loss": 0.8409,
+      "learning_rate": 9.562426032914663e-07,
+      "loss": 0.83,
       "step": 30443
     },
     {
-      "epoch": 0.8361209524594216,
+      "epoch": 0.8639046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 1.3758457434916828e-06,
-      "loss": 0.7959,
+      "learning_rate": 9.55850436812331e-07,
+      "loss": 0.9396,
       "step": 30444
     },
     {
-      "epoch": 0.8361484166872648,
+      "epoch": 0.863933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.3753955035567612e-06,
-      "loss": 0.8434,
+      "learning_rate": 9.554583467295341e-07,
+      "loss": 0.8705,
       "step": 30445
     },
     {
-      "epoch": 0.8361758809151081,
+      "epoch": 0.863961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.37494533186305e-06,
-      "loss": 0.8911,
+      "learning_rate": 9.55066333046386e-07,
+      "loss": 0.7712,
       "step": 30446
     },
     {
-      "epoch": 0.8362033451429514,
+      "epoch": 0.8639897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 1.3744952284141154e-06,
-      "loss": 0.6945,
+      "learning_rate": 9.546743957661997e-07,
+      "loss": 0.7772,
       "step": 30447
     },
     {
-      "epoch": 0.8362308093707945,
+      "epoch": 0.8640181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.3740451932135125e-06,
-      "loss": 0.7858,
+      "learning_rate": 9.542825348922879e-07,
+      "loss": 0.791,
       "step": 30448
     },
     {
-      "epoch": 0.8362582735986378,
+      "epoch": 0.8640465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 1.3735952262648089e-06,
-      "loss": 0.7467,
+      "learning_rate": 9.53890750427956e-07,
+      "loss": 0.714,
       "step": 30449
     },
     {
-      "epoch": 0.836285737826481,
+      "epoch": 0.8640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.373145327571559e-06,
-      "loss": 0.8522,
+      "learning_rate": 9.534990423765155e-07,
+      "loss": 0.7899,
       "step": 30450
     },
     {
-      "epoch": 0.8363132020543242,
+      "epoch": 0.8641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.372695497137324e-06,
-      "loss": 0.7865,
+      "learning_rate": 9.531074107412763e-07,
+      "loss": 0.8516,
       "step": 30451
     },
     {
-      "epoch": 0.8363406662821675,
+      "epoch": 0.8641316685584562,
       "grad_norm": 0.0,
-      "learning_rate": 1.3722457349656648e-06,
-      "loss": 0.8811,
+      "learning_rate": 9.527158555255445e-07,
+      "loss": 0.8369,
       "step": 30452
     },
     {
-      "epoch": 0.8363681305100107,
+      "epoch": 0.8641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.3717960410601415e-06,
-      "loss": 0.7793,
+      "learning_rate": 9.523243767326284e-07,
+      "loss": 0.7632,
       "step": 30453
     },
     {
-      "epoch": 0.836395594737854,
+      "epoch": 0.8641884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.3713464154243083e-06,
-      "loss": 0.778,
+      "learning_rate": 9.519329743658367e-07,
+      "loss": 0.6739,
       "step": 30454
     },
     {
-      "epoch": 0.8364230589656971,
+      "epoch": 0.8642167990919409,
       "grad_norm": 0.0,
-      "learning_rate": 1.3708968580617243e-06,
-      "loss": 0.7745,
+      "learning_rate": 9.515416484284723e-07,
+      "loss": 0.766,
       "step": 30455
     },
     {
-      "epoch": 0.8364505231935404,
+      "epoch": 0.8642451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 1.3704473689759468e-06,
-      "loss": 0.824,
+      "learning_rate": 9.511503989238424e-07,
+      "loss": 0.877,
       "step": 30456
     },
     {
-      "epoch": 0.8364779874213837,
+      "epoch": 0.8642735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.369997948170535e-06,
-      "loss": 0.7809,
+      "learning_rate": 9.507592258552534e-07,
+      "loss": 0.8316,
       "step": 30457
     },
     {
-      "epoch": 0.8365054516492268,
+      "epoch": 0.8643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.3695485956490407e-06,
-      "loss": 0.8896,
+      "learning_rate": 9.50368129226007e-07,
+      "loss": 0.7328,
       "step": 30458
     },
     {
-      "epoch": 0.8365329158770701,
+      "epoch": 0.8643303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 1.3690993114150219e-06,
-      "loss": 0.8625,
+      "learning_rate": 9.499771090394072e-07,
+      "loss": 0.8788,
       "step": 30459
     },
     {
-      "epoch": 0.8365603801049134,
+      "epoch": 0.8643586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 1.3686500954720316e-06,
-      "loss": 0.8042,
+      "learning_rate": 9.495861652987592e-07,
+      "loss": 0.8106,
       "step": 30460
     },
     {
-      "epoch": 0.8365878443327566,
+      "epoch": 0.8643870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 1.3682009478236246e-06,
-      "loss": 0.7732,
+      "learning_rate": 9.491952980073604e-07,
+      "loss": 0.8966,
       "step": 30461
     },
     {
-      "epoch": 0.8366153085605998,
+      "epoch": 0.8644154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 1.367751868473356e-06,
-      "loss": 0.7486,
+      "learning_rate": 9.488045071685193e-07,
+      "loss": 0.8511,
       "step": 30462
     },
     {
-      "epoch": 0.836642772788443,
+      "epoch": 0.8644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 1.3673028574247782e-06,
-      "loss": 0.7891,
+      "learning_rate": 9.484137927855308e-07,
+      "loss": 0.8843,
       "step": 30463
     },
     {
-      "epoch": 0.8366702370162863,
+      "epoch": 0.864472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.3668539146814464e-06,
-      "loss": 0.8417,
+      "learning_rate": 9.480231548616991e-07,
+      "loss": 0.8813,
       "step": 30464
     },
     {
-      "epoch": 0.8366977012441296,
+      "epoch": 0.8645005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.3664050402469076e-06,
-      "loss": 0.9293,
+      "learning_rate": 9.476325934003239e-07,
+      "loss": 0.8315,
       "step": 30465
     },
     {
-      "epoch": 0.8367251654719727,
+      "epoch": 0.8645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 1.365956234124719e-06,
-      "loss": 0.8344,
+      "learning_rate": 9.472421084047012e-07,
+      "loss": 0.6886,
       "step": 30466
     },
     {
-      "epoch": 0.836752629699816,
+      "epoch": 0.8645573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 1.3655074963184278e-06,
-      "loss": 0.7766,
+      "learning_rate": 9.468516998781319e-07,
+      "loss": 0.7729,
       "step": 30467
     },
     {
-      "epoch": 0.8367800939276592,
+      "epoch": 0.8645856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 1.3650588268315845e-06,
-      "loss": 0.8961,
+      "learning_rate": 9.464613678239154e-07,
+      "loss": 0.7477,
       "step": 30468
     },
     {
-      "epoch": 0.8368075581555025,
+      "epoch": 0.8646140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 1.3646102256677419e-06,
-      "loss": 0.7762,
+      "learning_rate": 9.460711122453437e-07,
+      "loss": 0.7499,
       "step": 30469
     },
     {
-      "epoch": 0.8368350223833457,
+      "epoch": 0.8646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 1.364161692830448e-06,
-      "loss": 0.8435,
+      "learning_rate": 9.456809331457173e-07,
+      "loss": 0.8135,
       "step": 30470
     },
     {
-      "epoch": 0.8368624866111889,
+      "epoch": 0.8646708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 1.3637132283232535e-06,
-      "loss": 0.7458,
+      "learning_rate": 9.452908305283326e-07,
+      "loss": 0.8591,
       "step": 30471
     },
     {
-      "epoch": 0.8368899508390322,
+      "epoch": 0.8646992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 1.3632648321497044e-06,
-      "loss": 0.7342,
+      "learning_rate": 9.449008043964813e-07,
+      "loss": 0.9089,
       "step": 30472
     },
     {
-      "epoch": 0.8369174150668754,
+      "epoch": 0.8647275822928491,
       "grad_norm": 0.0,
-      "learning_rate": 1.3628165043133501e-06,
-      "loss": 0.7633,
+      "learning_rate": 9.445108547534598e-07,
+      "loss": 0.7931,
       "step": 30473
     },
     {
-      "epoch": 0.8369448792947186,
+      "epoch": 0.8647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 1.3623682448177367e-06,
-      "loss": 0.8443,
+      "learning_rate": 9.441209816025643e-07,
+      "loss": 0.8459,
       "step": 30474
     },
     {
-      "epoch": 0.8369723435225619,
+      "epoch": 0.8647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 1.3619200536664102e-06,
-      "loss": 0.7522,
+      "learning_rate": 9.437311849470843e-07,
+      "loss": 0.8949,
       "step": 30475
     },
     {
-      "epoch": 0.8369998077504051,
+      "epoch": 0.8648127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 1.36147193086292e-06,
-      "loss": 0.9061,
+      "learning_rate": 9.433414647903138e-07,
+      "loss": 0.7822,
       "step": 30476
     },
     {
-      "epoch": 0.8370272719782483,
+      "epoch": 0.8648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 1.3610238764108107e-06,
-      "loss": 0.7335,
+      "learning_rate": 9.429518211355448e-07,
+      "loss": 0.7358,
       "step": 30477
     },
     {
-      "epoch": 0.8370547362060916,
+      "epoch": 0.8648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.3605758903136257e-06,
-      "loss": 0.841,
+      "learning_rate": 9.425622539860702e-07,
+      "loss": 0.7809,
       "step": 30478
     },
     {
-      "epoch": 0.8370822004339348,
+      "epoch": 0.8648978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 1.3601279725749106e-06,
-      "loss": 0.8332,
+      "learning_rate": 9.421727633451783e-07,
+      "loss": 0.8381,
       "step": 30479
     },
     {
-      "epoch": 0.837109664661778,
+      "epoch": 0.8649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 1.3596801231982094e-06,
-      "loss": 0.8103,
+      "learning_rate": 9.417833492161622e-07,
+      "loss": 0.8705,
       "step": 30480
     },
     {
-      "epoch": 0.8371371288896212,
+      "epoch": 0.8649545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 1.359232342187068e-06,
-      "loss": 0.8359,
+      "learning_rate": 9.413940116023068e-07,
+      "loss": 0.9361,
       "step": 30481
     },
     {
-      "epoch": 0.8371645931174645,
+      "epoch": 0.864982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.3587846295450257e-06,
-      "loss": 0.8278,
+      "learning_rate": 9.410047505069042e-07,
+      "loss": 0.7795,
       "step": 30482
     },
     {
-      "epoch": 0.8371920573453078,
+      "epoch": 0.865011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.358336985275629e-06,
-      "loss": 0.6936,
+      "learning_rate": 9.406155659332439e-07,
+      "loss": 0.7986,
       "step": 30483
     },
     {
-      "epoch": 0.8372195215731509,
+      "epoch": 0.8650397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 1.3578894093824159e-06,
-      "loss": 0.7582,
+      "learning_rate": 9.402264578846088e-07,
+      "loss": 0.8054,
       "step": 30484
     },
     {
-      "epoch": 0.8372469858009942,
+      "epoch": 0.8650681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 1.3574419018689279e-06,
-      "loss": 0.81,
+      "learning_rate": 9.398374263642907e-07,
+      "loss": 0.8289,
       "step": 30485
     },
     {
-      "epoch": 0.8372744500288375,
+      "epoch": 0.8650964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 1.3569944627387088e-06,
-      "loss": 0.7097,
+      "learning_rate": 9.394484713755703e-07,
+      "loss": 0.855,
       "step": 30486
     },
     {
-      "epoch": 0.8373019142566807,
+      "epoch": 0.8651248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 1.3565470919952962e-06,
-      "loss": 0.8641,
+      "learning_rate": 9.390595929217361e-07,
+      "loss": 0.841,
       "step": 30487
     },
     {
-      "epoch": 0.8373293784845239,
+      "epoch": 0.8651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 1.3560997896422346e-06,
-      "loss": 0.8117,
+      "learning_rate": 9.386707910060755e-07,
+      "loss": 0.8673,
       "step": 30488
     },
     {
-      "epoch": 0.8373568427123671,
+      "epoch": 0.8651816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 1.3556525556830568e-06,
-      "loss": 0.8214,
+      "learning_rate": 9.382820656318669e-07,
+      "loss": 0.7236,
       "step": 30489
     },
     {
-      "epoch": 0.8373843069402104,
+      "epoch": 0.8652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.355205390121307e-06,
-      "loss": 0.8093,
+      "learning_rate": 9.378934168023979e-07,
+      "loss": 0.8158,
       "step": 30490
     },
     {
-      "epoch": 0.8374117711680537,
+      "epoch": 0.8652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.3547582929605185e-06,
-      "loss": 0.8248,
+      "learning_rate": 9.375048445209523e-07,
+      "loss": 0.805,
       "step": 30491
     },
     {
-      "epoch": 0.8374392353958968,
+      "epoch": 0.865266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.3543112642042312e-06,
-      "loss": 0.8228,
+      "learning_rate": 9.371163487908074e-07,
+      "loss": 0.7377,
       "step": 30492
     },
     {
-      "epoch": 0.8374666996237401,
+      "epoch": 0.8652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 1.353864303855983e-06,
-      "loss": 0.8446,
+      "learning_rate": 9.367279296152487e-07,
+      "loss": 0.7545,
       "step": 30493
     },
     {
-      "epoch": 0.8374941638515833,
+      "epoch": 0.8653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 1.3534174119193088e-06,
-      "loss": 0.8778,
+      "learning_rate": 9.3633958699756e-07,
+      "loss": 0.7042,
       "step": 30494
     },
     {
-      "epoch": 0.8375216280794265,
+      "epoch": 0.8653518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 1.3529705883977472e-06,
-      "loss": 0.8601,
+      "learning_rate": 9.359513209410154e-07,
+      "loss": 0.8803,
       "step": 30495
     },
     {
-      "epoch": 0.8375490923072698,
+      "epoch": 0.8653802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 1.3525238332948299e-06,
-      "loss": 0.7513,
+      "learning_rate": 9.355631314488978e-07,
+      "loss": 0.7707,
       "step": 30496
     },
     {
-      "epoch": 0.837576556535113,
+      "epoch": 0.8654086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 1.352077146614097e-06,
-      "loss": 0.9033,
+      "learning_rate": 9.351750185244879e-07,
+      "loss": 0.8282,
       "step": 30497
     },
     {
-      "epoch": 0.8376040207629563,
+      "epoch": 0.8654370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 1.3516305283590758e-06,
-      "loss": 0.8152,
+      "learning_rate": 9.347869821710598e-07,
+      "loss": 0.9199,
       "step": 30498
     },
     {
-      "epoch": 0.8376314849907995,
+      "epoch": 0.8654653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 1.3511839785333048e-06,
-      "loss": 0.7241,
+      "learning_rate": 9.343990223918931e-07,
+      "loss": 0.6731,
       "step": 30499
     },
     {
-      "epoch": 0.8376589492186427,
+      "epoch": 0.8654937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 1.3507374971403165e-06,
-      "loss": 0.8678,
+      "learning_rate": 9.340111391902684e-07,
+      "loss": 0.7892,
       "step": 30500
     },
     {
-      "epoch": 0.837686413446486,
+      "epoch": 0.865522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.350291084183646e-06,
-      "loss": 0.8354,
+      "learning_rate": 9.336233325694566e-07,
+      "loss": 0.7686,
       "step": 30501
     },
     {
-      "epoch": 0.8377138776743291,
+      "epoch": 0.8655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 1.34984473966682e-06,
-      "loss": 0.8729,
+      "learning_rate": 9.332356025327361e-07,
+      "loss": 0.7974,
       "step": 30502
     },
     {
-      "epoch": 0.8377413419021724,
+      "epoch": 0.8655788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 1.3493984635933732e-06,
-      "loss": 0.817,
+      "learning_rate": 9.328479490833842e-07,
+      "loss": 0.8605,
       "step": 30503
     },
     {
-      "epoch": 0.8377688061300157,
+      "epoch": 0.8656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 1.3489522559668378e-06,
-      "loss": 0.6948,
+      "learning_rate": 9.324603722246717e-07,
+      "loss": 0.8283,
       "step": 30504
     },
     {
-      "epoch": 0.8377962703578589,
+      "epoch": 0.8656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 1.3485061167907443e-06,
-      "loss": 0.8281,
+      "learning_rate": 9.320728719598737e-07,
+      "loss": 0.8754,
       "step": 30505
     },
     {
-      "epoch": 0.8378237345857021,
+      "epoch": 0.8656640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 1.3480600460686221e-06,
-      "loss": 0.8228,
+      "learning_rate": 9.316854482922655e-07,
+      "loss": 0.7385,
       "step": 30506
     },
     {
-      "epoch": 0.8378511988135453,
+      "epoch": 0.8656923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 1.3476140438039976e-06,
-      "loss": 0.8295,
+      "learning_rate": 9.312981012251143e-07,
+      "loss": 0.8063,
       "step": 30507
     },
     {
-      "epoch": 0.8378786630413886,
+      "epoch": 0.8657207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 1.3471681100004031e-06,
-      "loss": 0.8103,
+      "learning_rate": 9.309108307616998e-07,
+      "loss": 0.8012,
       "step": 30508
     },
     {
-      "epoch": 0.8379061272692319,
+      "epoch": 0.8657491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 1.3467222446613658e-06,
-      "loss": 0.7746,
+      "learning_rate": 9.305236369052872e-07,
+      "loss": 0.8272,
       "step": 30509
     },
     {
-      "epoch": 0.837933591497075,
+      "epoch": 0.8657775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 1.3462764477904144e-06,
-      "loss": 0.7395,
+      "learning_rate": 9.301365196591483e-07,
+      "loss": 0.7739,
       "step": 30510
     },
     {
-      "epoch": 0.8379610557249183,
+      "epoch": 0.865805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.3458307193910758e-06,
-      "loss": 0.8641,
+      "learning_rate": 9.297494790265571e-07,
+      "loss": 0.8419,
       "step": 30511
     },
     {
-      "epoch": 0.8379885199527616,
+      "epoch": 0.8658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 1.3453850594668782e-06,
-      "loss": 0.8216,
+      "learning_rate": 9.293625150107766e-07,
+      "loss": 0.7895,
       "step": 30512
     },
     {
-      "epoch": 0.8380159841806047,
+      "epoch": 0.8658626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 1.3449394680213457e-06,
-      "loss": 0.8258,
+      "learning_rate": 9.289756276150797e-07,
+      "loss": 0.8951,
       "step": 30513
     },
     {
-      "epoch": 0.838043448408448,
+      "epoch": 0.8658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 1.3444939450580063e-06,
-      "loss": 0.9105,
+      "learning_rate": 9.285888168427349e-07,
+      "loss": 0.8337,
       "step": 30514
     },
     {
-      "epoch": 0.8380709126362912,
+      "epoch": 0.8659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.3440484905803807e-06,
-      "loss": 0.7953,
+      "learning_rate": 9.282020826970061e-07,
+      "loss": 0.7333,
       "step": 30515
     },
     {
-      "epoch": 0.8380983768641345,
+      "epoch": 0.8659477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.3436031045919972e-06,
-      "loss": 0.8126,
+      "learning_rate": 9.278154251811633e-07,
+      "loss": 0.8755,
       "step": 30516
     },
     {
-      "epoch": 0.8381258410919777,
+      "epoch": 0.8659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 1.3431577870963796e-06,
-      "loss": 0.7981,
+      "learning_rate": 9.274288442984714e-07,
+      "loss": 0.6891,
       "step": 30517
     },
     {
-      "epoch": 0.8381533053198209,
+      "epoch": 0.8660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 1.3427125380970529e-06,
-      "loss": 0.7257,
+      "learning_rate": 9.270423400521955e-07,
+      "loss": 0.7961,
       "step": 30518
     },
     {
-      "epoch": 0.8381807695476642,
+      "epoch": 0.8660329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 1.3422673575975353e-06,
-      "loss": 0.8601,
+      "learning_rate": 9.266559124455998e-07,
+      "loss": 0.7739,
       "step": 30519
     },
     {
-      "epoch": 0.8382082337755073,
+      "epoch": 0.8660612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 1.341822245601353e-06,
-      "loss": 0.7295,
+      "learning_rate": 9.262695614819528e-07,
+      "loss": 0.8491,
       "step": 30520
     },
     {
-      "epoch": 0.8382356980033506,
+      "epoch": 0.8660896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 1.3413772021120287e-06,
-      "loss": 0.749,
+      "learning_rate": 9.258832871645118e-07,
+      "loss": 0.794,
       "step": 30521
     },
     {
-      "epoch": 0.8382631622311939,
+      "epoch": 0.8661180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 1.3409322271330793e-06,
-      "loss": 0.797,
+      "learning_rate": 9.25497089496542e-07,
+      "loss": 0.761,
       "step": 30522
     },
     {
-      "epoch": 0.8382906264590371,
+      "epoch": 0.8661464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 1.340487320668028e-06,
-      "loss": 0.7955,
+      "learning_rate": 9.251109684813075e-07,
+      "loss": 0.8359,
       "step": 30523
     },
     {
-      "epoch": 0.8383180906868803,
+      "epoch": 0.8661748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 1.3400424827203984e-06,
-      "loss": 0.8451,
+      "learning_rate": 9.247249241220679e-07,
+      "loss": 0.7663,
       "step": 30524
     },
     {
-      "epoch": 0.8383455549147236,
+      "epoch": 0.8662031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 1.3395977132937054e-06,
-      "loss": 0.8578,
+      "learning_rate": 9.243389564220861e-07,
+      "loss": 0.8853,
       "step": 30525
     },
     {
-      "epoch": 0.8383730191425668,
+      "epoch": 0.8662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 1.3391530123914699e-06,
-      "loss": 0.7945,
+      "learning_rate": 9.239530653846207e-07,
+      "loss": 0.7635,
       "step": 30526
     },
     {
-      "epoch": 0.8384004833704101,
+      "epoch": 0.8662599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 1.3387083800172097e-06,
-      "loss": 0.883,
+      "learning_rate": 9.235672510129312e-07,
+      "loss": 0.9229,
       "step": 30527
     },
     {
-      "epoch": 0.8384279475982532,
+      "epoch": 0.8662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.3382638161744443e-06,
+      "learning_rate": 9.231815133102784e-07,
       "loss": 0.7747,
       "step": 30528
     },
     {
-      "epoch": 0.8384554118260965,
+      "epoch": 0.866316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.3378193208666933e-06,
-      "loss": 0.8226,
+      "learning_rate": 9.227958522799185e-07,
+      "loss": 0.7634,
       "step": 30529
     },
     {
-      "epoch": 0.8384828760539398,
+      "epoch": 0.8663450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 1.3373748940974707e-06,
-      "loss": 0.7731,
+      "learning_rate": 9.22410267925109e-07,
+      "loss": 0.8077,
       "step": 30530
     },
     {
-      "epoch": 0.838510340281783,
+      "epoch": 0.8663734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 1.3369305358702922e-06,
-      "loss": 0.8611,
+      "learning_rate": 9.220247602491105e-07,
+      "loss": 0.8136,
       "step": 30531
     },
     {
-      "epoch": 0.8385378045096262,
+      "epoch": 0.8664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 1.3364862461886742e-06,
-      "loss": 0.9247,
+      "learning_rate": 9.216393292551751e-07,
+      "loss": 0.7956,
       "step": 30532
     },
     {
-      "epoch": 0.8385652687374694,
+      "epoch": 0.8664301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 1.3360420250561334e-06,
-      "loss": 0.8472,
+      "learning_rate": 9.212539749465598e-07,
+      "loss": 0.8853,
       "step": 30533
     },
     {
-      "epoch": 0.8385927329653127,
+      "epoch": 0.8664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 1.3355978724761842e-06,
-      "loss": 0.8037,
+      "learning_rate": 9.208686973265224e-07,
+      "loss": 0.7864,
       "step": 30534
     },
     {
-      "epoch": 0.838620197193156,
+      "epoch": 0.8664869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 1.3351537884523402e-06,
-      "loss": 0.7434,
+      "learning_rate": 9.204834963983145e-07,
+      "loss": 0.7158,
       "step": 30535
     },
     {
-      "epoch": 0.8386476614209991,
+      "epoch": 0.8665153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 1.3347097729881198e-06,
-      "loss": 0.7918,
+      "learning_rate": 9.20098372165189e-07,
+      "loss": 0.8724,
       "step": 30536
     },
     {
-      "epoch": 0.8386751256488424,
+      "epoch": 0.8665437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 1.334265826087029e-06,
-      "loss": 0.7354,
+      "learning_rate": 9.197133246304035e-07,
+      "loss": 0.8212,
       "step": 30537
     },
     {
-      "epoch": 0.8387025898766857,
+      "epoch": 0.866572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.3338219477525849e-06,
-      "loss": 0.7489,
+      "learning_rate": 9.193283537972042e-07,
+      "loss": 0.7806,
       "step": 30538
     },
     {
-      "epoch": 0.8387300541045288,
+      "epoch": 0.8666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 1.3333781379882981e-06,
-      "loss": 0.7859,
+      "learning_rate": 9.189434596688496e-07,
+      "loss": 0.8094,
       "step": 30539
     },
     {
-      "epoch": 0.8387575183323721,
+      "epoch": 0.8666288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 1.3329343967976793e-06,
-      "loss": 0.8055,
+      "learning_rate": 9.185586422485859e-07,
+      "loss": 0.8149,
       "step": 30540
     },
     {
-      "epoch": 0.8387849825602153,
+      "epoch": 0.8666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 1.332490724184241e-06,
-      "loss": 0.9341,
+      "learning_rate": 9.181739015396662e-07,
+      "loss": 0.8323,
       "step": 30541
     },
     {
-      "epoch": 0.8388124467880586,
+      "epoch": 0.8666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.3320471201514962e-06,
-      "loss": 0.7699,
+      "learning_rate": 9.177892375453413e-07,
+      "loss": 0.7412,
       "step": 30542
     },
     {
-      "epoch": 0.8388399110159018,
+      "epoch": 0.8667139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 1.3316035847029507e-06,
-      "loss": 0.8168,
+      "learning_rate": 9.174046502688572e-07,
+      "loss": 0.7243,
       "step": 30543
     },
     {
-      "epoch": 0.838867375243745,
+      "epoch": 0.8667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 1.3311601178421141e-06,
-      "loss": 0.8009,
+      "learning_rate": 9.170201397134648e-07,
+      "loss": 0.7469,
       "step": 30544
     },
     {
-      "epoch": 0.8388948394715883,
+      "epoch": 0.8667707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 1.3307167195725e-06,
-      "loss": 0.875,
+      "learning_rate": 9.166357058824127e-07,
+      "loss": 0.7717,
       "step": 30545
     },
     {
-      "epoch": 0.8389223036994314,
+      "epoch": 0.8667990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 1.3302733898976105e-06,
-      "loss": 0.8495,
+      "learning_rate": 9.162513487789459e-07,
+      "loss": 0.8531,
       "step": 30546
     },
     {
-      "epoch": 0.8389497679272747,
+      "epoch": 0.866827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.329830128820956e-06,
-      "loss": 0.6872,
+      "learning_rate": 9.15867068406312e-07,
+      "loss": 0.8283,
       "step": 30547
     },
     {
-      "epoch": 0.838977232155118,
+      "epoch": 0.866855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 1.3293869363460466e-06,
-      "loss": 0.7529,
+      "learning_rate": 9.154828647677594e-07,
+      "loss": 0.6936,
       "step": 30548
     },
     {
-      "epoch": 0.8390046963829612,
+      "epoch": 0.8668842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 1.3289438124763842e-06,
-      "loss": 0.8694,
+      "learning_rate": 9.150987378665288e-07,
+      "loss": 0.8039,
       "step": 30549
     },
     {
-      "epoch": 0.8390321606108044,
+      "epoch": 0.8669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 1.3285007572154773e-06,
-      "loss": 0.7698,
+      "learning_rate": 9.147146877058688e-07,
+      "loss": 0.8462,
       "step": 30550
     },
     {
-      "epoch": 0.8390596248386477,
+      "epoch": 0.8669409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 1.3280577705668319e-06,
-      "loss": 0.8435,
+      "learning_rate": 9.143307142890224e-07,
+      "loss": 0.8779,
       "step": 30551
     },
     {
-      "epoch": 0.8390870890664909,
+      "epoch": 0.8669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 1.3276148525339517e-06,
-      "loss": 0.8459,
+      "learning_rate": 9.139468176192323e-07,
+      "loss": 0.7803,
       "step": 30552
     },
     {
-      "epoch": 0.8391145532943342,
+      "epoch": 0.8669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 1.327172003120345e-06,
-      "loss": 0.7819,
+      "learning_rate": 9.135629976997418e-07,
+      "loss": 0.831,
       "step": 30553
     },
     {
-      "epoch": 0.8391420175221773,
+      "epoch": 0.8670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.3267292223295126e-06,
-      "loss": 0.9619,
+      "learning_rate": 9.131792545337926e-07,
+      "loss": 0.7445,
       "step": 30554
     },
     {
-      "epoch": 0.8391694817500206,
+      "epoch": 0.8670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 1.3262865101649557e-06,
-      "loss": 0.786,
+      "learning_rate": 9.127955881246275e-07,
+      "loss": 0.8562,
       "step": 30555
     },
     {
-      "epoch": 0.8391969459778639,
+      "epoch": 0.8670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 1.325843866630181e-06,
-      "loss": 0.7404,
+      "learning_rate": 9.124119984754876e-07,
+      "loss": 0.7904,
       "step": 30556
     },
     {
-      "epoch": 0.839224410205707,
+      "epoch": 0.8671112372304199,
       "grad_norm": 0.0,
-      "learning_rate": 1.3254012917286884e-06,
-      "loss": 0.8374,
+      "learning_rate": 9.120284855896111e-07,
+      "loss": 0.7955,
       "step": 30557
     },
     {
-      "epoch": 0.8392518744335503,
+      "epoch": 0.8671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 1.3249587854639823e-06,
-      "loss": 0.7091,
+      "learning_rate": 9.116450494702378e-07,
+      "loss": 0.8182,
       "step": 30558
     },
     {
-      "epoch": 0.8392793386613935,
+      "epoch": 0.8671679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 1.3245163478395618e-06,
-      "loss": 0.9192,
+      "learning_rate": 9.112616901206095e-07,
+      "loss": 0.7264,
       "step": 30559
     },
     {
-      "epoch": 0.8393068028892368,
+      "epoch": 0.8671963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 1.3240739788589308e-06,
-      "loss": 0.8159,
+      "learning_rate": 9.108784075439603e-07,
+      "loss": 0.8136,
       "step": 30560
     },
     {
-      "epoch": 0.83933426711708,
+      "epoch": 0.8672247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 1.3236316785255855e-06,
-      "loss": 0.9098,
+      "learning_rate": 9.104952017435287e-07,
+      "loss": 0.7147,
       "step": 30561
     },
     {
-      "epoch": 0.8393617313449232,
+      "epoch": 0.8672531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 1.3231894468430283e-06,
-      "loss": 0.8181,
+      "learning_rate": 9.101120727225565e-07,
+      "loss": 0.7295,
       "step": 30562
     },
     {
-      "epoch": 0.8393891955727665,
+      "epoch": 0.8672814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 1.3227472838147559e-06,
-      "loss": 0.8416,
+      "learning_rate": 9.097290204842735e-07,
+      "loss": 0.7986,
       "step": 30563
     },
     {
-      "epoch": 0.8394166598006098,
+      "epoch": 0.8673098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 1.3223051894442685e-06,
-      "loss": 0.8295,
+      "learning_rate": 9.093460450319181e-07,
+      "loss": 0.7866,
       "step": 30564
     },
     {
-      "epoch": 0.8394441240284529,
+      "epoch": 0.8673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 1.3218631637350655e-06,
-      "loss": 0.7741,
+      "learning_rate": 9.089631463687265e-07,
+      "loss": 0.7992,
       "step": 30565
     },
     {
-      "epoch": 0.8394715882562962,
+      "epoch": 0.867366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.3214212066906406e-06,
-      "loss": 0.7855,
+      "learning_rate": 9.085803244979308e-07,
+      "loss": 0.8598,
       "step": 30566
     },
     {
-      "epoch": 0.8394990524841394,
+      "epoch": 0.8673950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.3209793183144926e-06,
-      "loss": 0.7889,
+      "learning_rate": 9.08197579422766e-07,
+      "loss": 0.8283,
       "step": 30567
     },
     {
-      "epoch": 0.8395265167119826,
+      "epoch": 0.8674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 1.3205374986101193e-06,
-      "loss": 0.6947,
+      "learning_rate": 9.078149111464663e-07,
+      "loss": 0.8331,
       "step": 30568
     },
     {
-      "epoch": 0.8395539809398259,
+      "epoch": 0.8674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 1.3200957475810161e-06,
-      "loss": 0.9315,
+      "learning_rate": 9.074323196722589e-07,
+      "loss": 0.8723,
       "step": 30569
     },
     {
-      "epoch": 0.8395814451676691,
+      "epoch": 0.8674801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 1.3196540652306767e-06,
-      "loss": 0.822,
+      "learning_rate": 9.070498050033838e-07,
+      "loss": 0.8949,
       "step": 30570
     },
     {
-      "epoch": 0.8396089093955124,
+      "epoch": 0.8675085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 1.3192124515625982e-06,
-      "loss": 0.8809,
+      "learning_rate": 9.066673671430659e-07,
+      "loss": 0.8026,
       "step": 30571
     },
     {
-      "epoch": 0.8396363736233555,
+      "epoch": 0.8675368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 1.3187709065802723e-06,
-      "loss": 0.8985,
+      "learning_rate": 9.062850060945372e-07,
+      "loss": 0.8033,
       "step": 30572
     },
     {
-      "epoch": 0.8396638378511988,
+      "epoch": 0.8675652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 1.318329430287193e-06,
-      "loss": 0.8325,
+      "learning_rate": 9.059027218610294e-07,
+      "loss": 0.7951,
       "step": 30573
     },
     {
-      "epoch": 0.8396913020790421,
+      "epoch": 0.8675936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 1.317888022686854e-06,
-      "loss": 0.749,
+      "learning_rate": 9.05520514445769e-07,
+      "loss": 0.7899,
       "step": 30574
     },
     {
-      "epoch": 0.8397187663068852,
+      "epoch": 0.8676220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 1.3174466837827481e-06,
-      "loss": 0.8824,
+      "learning_rate": 9.051383838519845e-07,
+      "loss": 0.784,
       "step": 30575
     },
     {
-      "epoch": 0.8397462305347285,
+      "epoch": 0.8676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 1.3170054135783671e-06,
-      "loss": 0.8655,
+      "learning_rate": 9.047563300829077e-07,
+      "loss": 0.701,
       "step": 30576
     },
     {
-      "epoch": 0.8397736947625718,
+      "epoch": 0.8676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 1.316564212077206e-06,
-      "loss": 0.716,
+      "learning_rate": 9.043743531417615e-07,
+      "loss": 0.7761,
       "step": 30577
     },
     {
-      "epoch": 0.839801158990415,
+      "epoch": 0.8677071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 1.3161230792827518e-06,
-      "loss": 0.7972,
+      "learning_rate": 9.039924530317734e-07,
+      "loss": 0.9463,
       "step": 30578
     },
     {
-      "epoch": 0.8398286232182582,
+      "epoch": 0.8677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.315682015198493e-06,
-      "loss": 0.6426,
+      "learning_rate": 9.03610629756172e-07,
+      "loss": 0.8404,
       "step": 30579
     },
     {
-      "epoch": 0.8398560874461014,
+      "epoch": 0.8677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.315241019827922e-06,
-      "loss": 0.8422,
+      "learning_rate": 9.032288833181779e-07,
+      "loss": 0.8568,
       "step": 30580
     },
     {
-      "epoch": 0.8398835516739447,
+      "epoch": 0.8677922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 1.3148000931745287e-06,
-      "loss": 0.7525,
+      "learning_rate": 9.028472137210187e-07,
+      "loss": 0.9182,
       "step": 30581
     },
     {
-      "epoch": 0.839911015901788,
+      "epoch": 0.8678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 1.3143592352418022e-06,
-      "loss": 0.8929,
+      "learning_rate": 9.024656209679206e-07,
+      "loss": 0.8349,
       "step": 30582
     },
     {
-      "epoch": 0.8399384801296311,
+      "epoch": 0.8678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 1.313918446033231e-06,
-      "loss": 0.8152,
+      "learning_rate": 9.020841050621021e-07,
+      "loss": 0.8071,
       "step": 30583
     },
     {
-      "epoch": 0.8399659443574744,
+      "epoch": 0.867877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.3134777255523002e-06,
-      "loss": 0.8351,
+      "learning_rate": 9.017026660067862e-07,
+      "loss": 0.7852,
       "step": 30584
     },
     {
-      "epoch": 0.8399934085853176,
+      "epoch": 0.867905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.3130370738025e-06,
-      "loss": 0.8011,
+      "learning_rate": 9.013213038052005e-07,
+      "loss": 0.7441,
       "step": 30585
     },
     {
-      "epoch": 0.8400208728131608,
+      "epoch": 0.8679341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 1.3125964907873134e-06,
-      "loss": 0.7774,
+      "learning_rate": 9.00940018460561e-07,
+      "loss": 0.8705,
       "step": 30586
     },
     {
-      "epoch": 0.8400483370410041,
+      "epoch": 0.8679625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 1.3121559765102288e-06,
-      "loss": 0.7798,
+      "learning_rate": 9.005588099760909e-07,
+      "loss": 0.8408,
       "step": 30587
     },
     {
-      "epoch": 0.8400758012688473,
+      "epoch": 0.8679909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 1.311715530974732e-06,
-      "loss": 0.8517,
+      "learning_rate": 9.001776783550109e-07,
+      "loss": 0.7695,
       "step": 30588
     },
     {
-      "epoch": 0.8401032654966906,
+      "epoch": 0.8680192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 1.3112751541843084e-06,
-      "loss": 0.7982,
+      "learning_rate": 8.997966236005373e-07,
+      "loss": 0.7738,
       "step": 30589
     },
     {
-      "epoch": 0.8401307297245338,
+      "epoch": 0.8680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 1.3108348461424402e-06,
-      "loss": 0.921,
+      "learning_rate": 8.994156457158898e-07,
+      "loss": 0.7074,
       "step": 30590
     },
     {
-      "epoch": 0.840158193952377,
+      "epoch": 0.8680760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 1.3103946068526118e-06,
-      "loss": 0.7863,
+      "learning_rate": 8.990347447042902e-07,
+      "loss": 0.8135,
       "step": 30591
     },
     {
-      "epoch": 0.8401856581802203,
+      "epoch": 0.8681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 1.309954436318308e-06,
-      "loss": 0.8039,
+      "learning_rate": 8.986539205689515e-07,
+      "loss": 0.7968,
       "step": 30592
     },
     {
-      "epoch": 0.8402131224080635,
+      "epoch": 0.8681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.3095143345430127e-06,
-      "loss": 0.7962,
+      "learning_rate": 8.982731733130934e-07,
+      "loss": 0.8172,
       "step": 30593
     },
     {
-      "epoch": 0.8402405866359067,
+      "epoch": 0.868161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.3090743015302042e-06,
-      "loss": 0.9345,
+      "learning_rate": 8.978925029399299e-07,
+      "loss": 0.9187,
       "step": 30594
     },
     {
-      "epoch": 0.84026805086375,
+      "epoch": 0.8681895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 1.3086343372833687e-06,
-      "loss": 0.9586,
+      "learning_rate": 8.975119094526774e-07,
+      "loss": 0.7954,
       "step": 30595
     },
     {
-      "epoch": 0.8402955150915932,
+      "epoch": 0.8682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 1.3081944418059833e-06,
-      "loss": 0.9342,
+      "learning_rate": 8.971313928545522e-07,
+      "loss": 0.7296,
       "step": 30596
     },
     {
-      "epoch": 0.8403229793194364,
+      "epoch": 0.8682463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 1.3077546151015296e-06,
-      "loss": 0.8199,
+      "learning_rate": 8.967509531487662e-07,
+      "loss": 0.727,
       "step": 30597
     },
     {
-      "epoch": 0.8403504435472796,
+      "epoch": 0.8682746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 1.3073148571734895e-06,
-      "loss": 0.8965,
+      "learning_rate": 8.963705903385344e-07,
+      "loss": 0.7906,
       "step": 30598
     },
     {
-      "epoch": 0.8403779077751229,
+      "epoch": 0.8683030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 1.30687516802534e-06,
-      "loss": 0.8398,
+      "learning_rate": 8.959903044270702e-07,
+      "loss": 0.7828,
       "step": 30599
     },
     {
-      "epoch": 0.8404053720029662,
+      "epoch": 0.8683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 1.3064355476605629e-06,
-      "loss": 0.8712,
+      "learning_rate": 8.956100954175828e-07,
+      "loss": 0.8137,
       "step": 30600
     },
     {
-      "epoch": 0.8404328362308093,
+      "epoch": 0.8683598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 1.3059959960826363e-06,
-      "loss": 0.8448,
+      "learning_rate": 8.952299633132867e-07,
+      "loss": 0.7508,
       "step": 30601
     },
     {
-      "epoch": 0.8404603004586526,
+      "epoch": 0.8683881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 1.3055565132950376e-06,
-      "loss": 0.7492,
+      "learning_rate": 8.948499081173956e-07,
+      "loss": 0.8646,
       "step": 30602
     },
     {
-      "epoch": 0.8404877646864959,
+      "epoch": 0.868416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.3051170993012419e-06,
-      "loss": 0.7398,
+      "learning_rate": 8.944699298331139e-07,
+      "loss": 0.8333,
       "step": 30603
     },
     {
-      "epoch": 0.840515228914339,
+      "epoch": 0.8684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 1.3046777541047273e-06,
-      "loss": 0.7234,
+      "learning_rate": 8.940900284636533e-07,
+      "loss": 0.8633,
       "step": 30604
     },
     {
-      "epoch": 0.8405426931421823,
+      "epoch": 0.8684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.304238477708971e-06,
-      "loss": 0.8214,
+      "learning_rate": 8.93710204012227e-07,
+      "loss": 0.8521,
       "step": 30605
     },
     {
-      "epoch": 0.8405701573700255,
+      "epoch": 0.8685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.3037992701174485e-06,
-      "loss": 0.9032,
+      "learning_rate": 8.933304564820366e-07,
+      "loss": 0.8933,
       "step": 30606
     },
     {
-      "epoch": 0.8405976215978688,
+      "epoch": 0.8685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 1.3033601313336363e-06,
-      "loss": 0.8224,
+      "learning_rate": 8.929507858762943e-07,
+      "loss": 0.8212,
       "step": 30607
     },
     {
-      "epoch": 0.840625085825712,
+      "epoch": 0.8685584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 1.3029210613610056e-06,
-      "loss": 0.8443,
+      "learning_rate": 8.925711921982083e-07,
+      "loss": 0.7762,
       "step": 30608
     },
     {
-      "epoch": 0.8406525500535552,
+      "epoch": 0.8685868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 1.3024820602030342e-06,
-      "loss": 0.7622,
+      "learning_rate": 8.921916754509796e-07,
+      "loss": 0.8083,
       "step": 30609
     },
     {
-      "epoch": 0.8406800142813985,
+      "epoch": 0.8686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 1.3020431278631918e-06,
-      "loss": 0.7908,
+      "learning_rate": 8.918122356378178e-07,
+      "loss": 0.7457,
       "step": 30610
     },
     {
-      "epoch": 0.8407074785092418,
+      "epoch": 0.8686435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 1.3016042643449533e-06,
-      "loss": 0.694,
+      "learning_rate": 8.914328727619304e-07,
+      "loss": 0.8048,
       "step": 30611
     },
     {
-      "epoch": 0.8407349427370849,
+      "epoch": 0.868671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.3011654696517918e-06,
-      "loss": 0.7766,
+      "learning_rate": 8.910535868265158e-07,
+      "loss": 0.7683,
       "step": 30612
     },
     {
-      "epoch": 0.8407624069649282,
+      "epoch": 0.868700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.3007267437871808e-06,
-      "loss": 0.8362,
+      "learning_rate": 8.906743778347815e-07,
+      "loss": 0.9604,
       "step": 30613
     },
     {
-      "epoch": 0.8407898711927714,
+      "epoch": 0.8687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 1.3002880867545863e-06,
-      "loss": 0.8738,
+      "learning_rate": 8.902952457899317e-07,
+      "loss": 0.8055,
       "step": 30614
     },
     {
-      "epoch": 0.8408173354206147,
+      "epoch": 0.8687570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 1.2998494985574839e-06,
-      "loss": 0.7739,
+      "learning_rate": 8.899161906951648e-07,
+      "loss": 0.7776,
       "step": 30615
     },
     {
-      "epoch": 0.8408447996484579,
+      "epoch": 0.8687854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 1.2994109791993415e-06,
-      "loss": 0.7862,
+      "learning_rate": 8.895372125536883e-07,
+      "loss": 0.795,
       "step": 30616
     },
     {
-      "epoch": 0.8408722638763011,
+      "epoch": 0.8688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 1.298972528683633e-06,
-      "loss": 0.7525,
+      "learning_rate": 8.891583113686986e-07,
+      "loss": 0.8225,
       "step": 30617
     },
     {
-      "epoch": 0.8408997281041444,
+      "epoch": 0.8688422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.2985341470138213e-06,
-      "loss": 0.7768,
+      "learning_rate": 8.887794871433985e-07,
+      "loss": 0.8109,
       "step": 30618
     },
     {
-      "epoch": 0.8409271923319875,
+      "epoch": 0.8688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 1.2980958341933802e-06,
-      "loss": 0.778,
+      "learning_rate": 8.884007398809891e-07,
+      "loss": 0.8359,
       "step": 30619
     },
     {
-      "epoch": 0.8409546565598308,
+      "epoch": 0.8688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 1.2976575902257726e-06,
-      "loss": 0.8481,
+      "learning_rate": 8.880220695846664e-07,
+      "loss": 0.7957,
       "step": 30620
     },
     {
-      "epoch": 0.8409821207876741,
+      "epoch": 0.8689273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 1.2972194151144702e-06,
-      "loss": 0.8029,
+      "learning_rate": 8.876434762576314e-07,
+      "loss": 0.7484,
       "step": 30621
     },
     {
-      "epoch": 0.8410095850155173,
+      "epoch": 0.8689557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 1.296781308862939e-06,
-      "loss": 0.7606,
+      "learning_rate": 8.872649599030836e-07,
+      "loss": 0.8066,
       "step": 30622
     },
     {
-      "epoch": 0.8410370492433605,
+      "epoch": 0.8689841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 1.2963432714746448e-06,
-      "loss": 0.7977,
+      "learning_rate": 8.86886520524216e-07,
+      "loss": 0.799,
       "step": 30623
     },
     {
-      "epoch": 0.8410645134712038,
+      "epoch": 0.8690124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 1.2959053029530566e-06,
-      "loss": 0.7932,
+      "learning_rate": 8.865081581242274e-07,
+      "loss": 0.7325,
       "step": 30624
     },
     {
-      "epoch": 0.841091977699047,
+      "epoch": 0.8690408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 1.2954674033016356e-06,
-      "loss": 0.7425,
+      "learning_rate": 8.861298727063161e-07,
+      "loss": 0.8073,
       "step": 30625
     },
     {
-      "epoch": 0.8411194419268903,
+      "epoch": 0.8690692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 1.2950295725238505e-06,
-      "loss": 0.8059,
+      "learning_rate": 8.857516642736741e-07,
+      "loss": 0.7208,
       "step": 30626
     },
     {
-      "epoch": 0.8411469061547334,
+      "epoch": 0.8690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 1.2945918106231602e-06,
-      "loss": 0.8077,
+      "learning_rate": 8.853735328294966e-07,
+      "loss": 0.7649,
       "step": 30627
     },
     {
-      "epoch": 0.8411743703825767,
+      "epoch": 0.8691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 1.2941541176030325e-06,
-      "loss": 0.8122,
+      "learning_rate": 8.84995478376981e-07,
+      "loss": 0.7407,
       "step": 30628
     },
     {
-      "epoch": 0.84120183461042,
+      "epoch": 0.8691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 1.2937164934669299e-06,
-      "loss": 0.8501,
+      "learning_rate": 8.84617500919317e-07,
+      "loss": 0.8795,
       "step": 30629
     },
     {
-      "epoch": 0.8412292988382631,
+      "epoch": 0.8691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 1.2932789382183164e-06,
-      "loss": 0.8288,
+      "learning_rate": 8.842396004596976e-07,
+      "loss": 0.7502,
       "step": 30630
     },
     {
-      "epoch": 0.8412567630661064,
+      "epoch": 0.869211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.2928414518606503e-06,
-      "loss": 0.9142,
+      "learning_rate": 8.838617770013169e-07,
+      "loss": 0.8537,
       "step": 30631
     },
     {
-      "epoch": 0.8412842272939496,
+      "epoch": 0.8692395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 1.2924040343973964e-06,
-      "loss": 0.8989,
+      "learning_rate": 8.834840305473658e-07,
+      "loss": 0.7181,
       "step": 30632
     },
     {
-      "epoch": 0.8413116915217929,
+      "epoch": 0.8692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 1.2919666858320157e-06,
-      "loss": 0.7208,
+      "learning_rate": 8.83106361101036e-07,
+      "loss": 0.913,
       "step": 30633
     },
     {
-      "epoch": 0.8413391557496361,
+      "epoch": 0.8692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 1.2915294061679662e-06,
-      "loss": 0.8216,
+      "learning_rate": 8.82728768665515e-07,
+      "loss": 0.7621,
       "step": 30634
     },
     {
-      "epoch": 0.8413666199774793,
+      "epoch": 0.8693246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.2910921954087097e-06,
-      "loss": 0.7425,
+      "learning_rate": 8.823512532439938e-07,
+      "loss": 0.7415,
       "step": 30635
     },
     {
-      "epoch": 0.8413940842053226,
+      "epoch": 0.8693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.2906550535577078e-06,
-      "loss": 0.7943,
+      "learning_rate": 8.819738148396639e-07,
+      "loss": 0.8593,
       "step": 30636
     },
     {
-      "epoch": 0.8414215484331659,
+      "epoch": 0.8693813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 1.2902179806184134e-06,
-      "loss": 0.8175,
+      "learning_rate": 8.815964534557087e-07,
+      "loss": 0.8283,
       "step": 30637
     },
     {
-      "epoch": 0.841449012661009,
+      "epoch": 0.8694097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.2897809765942893e-06,
-      "loss": 0.7596,
+      "learning_rate": 8.812191690953186e-07,
+      "loss": 0.8638,
       "step": 30638
     },
     {
-      "epoch": 0.8414764768888523,
+      "epoch": 0.8694381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 1.2893440414887915e-06,
-      "loss": 0.8315,
+      "learning_rate": 8.808419617616814e-07,
+      "loss": 0.8423,
       "step": 30639
     },
     {
-      "epoch": 0.8415039411166955,
+      "epoch": 0.8694665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 1.2889071753053773e-06,
-      "loss": 0.853,
+      "learning_rate": 8.8046483145798e-07,
+      "loss": 0.7374,
       "step": 30640
     },
     {
-      "epoch": 0.8415314053445387,
+      "epoch": 0.8694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.2884703780475073e-06,
-      "loss": 0.7746,
+      "learning_rate": 8.800877781874028e-07,
+      "loss": 0.9056,
       "step": 30641
     },
     {
-      "epoch": 0.841558869572382,
+      "epoch": 0.8695232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 1.2880336497186329e-06,
-      "loss": 0.7984,
+      "learning_rate": 8.797108019531353e-07,
+      "loss": 0.8541,
       "step": 30642
     },
     {
-      "epoch": 0.8415863338002252,
+      "epoch": 0.8695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.2875969903222097e-06,
-      "loss": 0.8702,
+      "learning_rate": 8.793339027583591e-07,
+      "loss": 0.8344,
       "step": 30643
     },
     {
-      "epoch": 0.8416137980280685,
+      "epoch": 0.8695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.2871603998616944e-06,
-      "loss": 0.7698,
+      "learning_rate": 8.789570806062597e-07,
+      "loss": 0.7782,
       "step": 30644
     },
     {
-      "epoch": 0.8416412622559116,
+      "epoch": 0.8696083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 1.28672387834054e-06,
-      "loss": 0.7809,
+      "learning_rate": 8.785803355000222e-07,
+      "loss": 0.8486,
       "step": 30645
     },
     {
-      "epoch": 0.8416687264837549,
+      "epoch": 0.8696367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 1.2862874257622016e-06,
-      "loss": 0.8313,
+      "learning_rate": 8.782036674428218e-07,
+      "loss": 0.9416,
       "step": 30646
     },
     {
-      "epoch": 0.8416961907115982,
+      "epoch": 0.8696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.285851042130133e-06,
-      "loss": 0.8275,
+      "learning_rate": 8.778270764378494e-07,
+      "loss": 0.7594,
       "step": 30647
     },
     {
-      "epoch": 0.8417236549394413,
+      "epoch": 0.8696935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 1.285414727447788e-06,
-      "loss": 0.8277,
+      "learning_rate": 8.774505624882801e-07,
+      "loss": 0.8453,
       "step": 30648
     },
     {
-      "epoch": 0.8417511191672846,
+      "epoch": 0.86972190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.2849784817186161e-06,
-      "loss": 0.8017,
+      "learning_rate": 8.770741255972959e-07,
+      "loss": 0.8009,
       "step": 30649
     },
     {
-      "epoch": 0.8417785833951279,
+      "epoch": 0.869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.2845423049460715e-06,
-      "loss": 0.8289,
+      "learning_rate": 8.766977657680776e-07,
+      "loss": 0.7973,
       "step": 30650
     },
     {
-      "epoch": 0.8418060476229711,
+      "epoch": 0.8697786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.2841061971336032e-06,
-      "loss": 0.8768,
+      "learning_rate": 8.763214830038025e-07,
+      "loss": 0.8714,
       "step": 30651
     },
     {
-      "epoch": 0.8418335118508143,
+      "epoch": 0.8698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.2836701582846622e-06,
-      "loss": 0.7917,
+      "learning_rate": 8.759452773076493e-07,
+      "loss": 0.7444,
       "step": 30652
     },
     {
-      "epoch": 0.8418609760786575,
+      "epoch": 0.8698354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.283234188402699e-06,
-      "loss": 0.7932,
+      "learning_rate": 8.755691486827989e-07,
+      "loss": 0.7481,
       "step": 30653
     },
     {
-      "epoch": 0.8418884403065008,
+      "epoch": 0.8698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.2827982874911659e-06,
-      "loss": 0.7073,
+      "learning_rate": 8.75193097132424e-07,
+      "loss": 0.7889,
       "step": 30654
     },
     {
-      "epoch": 0.8419159045343441,
+      "epoch": 0.8698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.2823624555535075e-06,
-      "loss": 0.8834,
+      "learning_rate": 8.748171226597035e-07,
+      "loss": 0.8398,
       "step": 30655
     },
     {
-      "epoch": 0.8419433687621872,
+      "epoch": 0.8699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.281926692593174e-06,
-      "loss": 0.8853,
+      "learning_rate": 8.744412252678148e-07,
+      "loss": 0.7911,
       "step": 30656
     },
     {
-      "epoch": 0.8419708329900305,
+      "epoch": 0.8699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.281490998613616e-06,
-      "loss": 0.7353,
+      "learning_rate": 8.740654049599295e-07,
+      "loss": 0.8118,
       "step": 30657
     },
     {
-      "epoch": 0.8419982972178737,
+      "epoch": 0.8699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.2810553736182764e-06,
-      "loss": 0.8943,
+      "learning_rate": 8.736896617392232e-07,
+      "loss": 0.6958,
       "step": 30658
     },
     {
-      "epoch": 0.842025761445717,
+      "epoch": 0.870005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 1.2806198176106032e-06,
-      "loss": 0.8054,
+      "learning_rate": 8.733139956088732e-07,
+      "loss": 0.7618,
       "step": 30659
     },
     {
-      "epoch": 0.8420532256735602,
+      "epoch": 0.8700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.2801843305940464e-06,
-      "loss": 0.8602,
+      "learning_rate": 8.729384065720481e-07,
+      "loss": 0.731,
       "step": 30660
     },
     {
-      "epoch": 0.8420806899014034,
+      "epoch": 0.8700624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.2797489125720464e-06,
-      "loss": 0.7821,
+      "learning_rate": 8.72562894631922e-07,
+      "loss": 0.7233,
       "step": 30661
     },
     {
-      "epoch": 0.8421081541292467,
+      "epoch": 0.8700908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.2793135635480514e-06,
-      "loss": 0.8397,
+      "learning_rate": 8.721874597916679e-07,
+      "loss": 0.7333,
       "step": 30662
     },
     {
-      "epoch": 0.8421356183570899,
+      "epoch": 0.8701191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 1.2788782835255054e-06,
-      "loss": 0.7467,
+      "learning_rate": 8.718121020544567e-07,
+      "loss": 0.8084,
       "step": 30663
     },
     {
-      "epoch": 0.8421630825849331,
+      "epoch": 0.8701475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 1.2784430725078522e-06,
-      "loss": 0.7884,
+      "learning_rate": 8.714368214234614e-07,
+      "loss": 0.9047,
       "step": 30664
     },
     {
-      "epoch": 0.8421905468127764,
+      "epoch": 0.8701759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.2780079304985393e-06,
-      "loss": 0.7494,
+      "learning_rate": 8.710616179018472e-07,
+      "loss": 0.8414,
       "step": 30665
     },
     {
-      "epoch": 0.8422180110406196,
+      "epoch": 0.8702043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 1.277572857501005e-06,
-      "loss": 0.8144,
+      "learning_rate": 8.70686491492786e-07,
+      "loss": 0.8196,
       "step": 30666
     },
     {
-      "epoch": 0.8422454752684628,
+      "epoch": 0.8702326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 1.2771378535186906e-06,
-      "loss": 0.7796,
+      "learning_rate": 8.703114421994474e-07,
+      "loss": 0.7209,
       "step": 30667
     },
     {
-      "epoch": 0.8422729394963061,
+      "epoch": 0.870261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.2767029185550416e-06,
-      "loss": 0.844,
+      "learning_rate": 8.699364700249979e-07,
+      "loss": 0.7711,
       "step": 30668
     },
     {
-      "epoch": 0.8423004037241493,
+      "epoch": 0.8702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.2762680526134974e-06,
-      "loss": 0.8499,
+      "learning_rate": 8.69561574972605e-07,
+      "loss": 0.7142,
       "step": 30669
     },
     {
-      "epoch": 0.8423278679519925,
+      "epoch": 0.8703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.2758332556974994e-06,
-      "loss": 0.7926,
+      "learning_rate": 8.691867570454371e-07,
+      "loss": 0.8544,
       "step": 30670
     },
     {
-      "epoch": 0.8423553321798357,
+      "epoch": 0.8703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.2753985278104885e-06,
-      "loss": 0.6702,
+      "learning_rate": 8.688120162466584e-07,
+      "loss": 0.8936,
       "step": 30671
     },
     {
-      "epoch": 0.842382796407679,
+      "epoch": 0.8703745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 1.2749638689559062e-06,
-      "loss": 0.8689,
+      "learning_rate": 8.684373525794343e-07,
+      "loss": 0.9252,
       "step": 30672
     },
     {
-      "epoch": 0.8424102606355223,
+      "epoch": 0.8704029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 1.2745292791371876e-06,
-      "loss": 0.8239,
+      "learning_rate": 8.680627660469321e-07,
+      "loss": 0.8322,
       "step": 30673
     },
     {
-      "epoch": 0.8424377248633654,
+      "epoch": 0.8704313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.2740947583577755e-06,
-      "loss": 0.8522,
+      "learning_rate": 8.676882566523137e-07,
+      "loss": 0.7887,
       "step": 30674
     },
     {
-      "epoch": 0.8424651890912087,
+      "epoch": 0.8704597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 1.2736603066211028e-06,
-      "loss": 0.8032,
+      "learning_rate": 8.673138243987411e-07,
+      "loss": 0.7966,
       "step": 30675
     },
     {
-      "epoch": 0.842492653319052,
+      "epoch": 0.8704880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.273225923930611e-06,
-      "loss": 0.8314,
+      "learning_rate": 8.669394692893807e-07,
+      "loss": 0.793,
       "step": 30676
     },
     {
-      "epoch": 0.8425201175468952,
+      "epoch": 0.8705164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 1.2727916102897352e-06,
-      "loss": 0.8856,
+      "learning_rate": 8.665651913273931e-07,
+      "loss": 0.8202,
       "step": 30677
     },
     {
-      "epoch": 0.8425475817747384,
+      "epoch": 0.8705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.272357365701915e-06,
-      "loss": 0.8346,
+      "learning_rate": 8.661909905159415e-07,
+      "loss": 0.7995,
       "step": 30678
     },
     {
-      "epoch": 0.8425750460025816,
+      "epoch": 0.8705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.271923190170583e-06,
-      "loss": 0.9022,
+      "learning_rate": 8.658168668581824e-07,
+      "loss": 0.779,
       "step": 30679
     },
     {
-      "epoch": 0.8426025102304249,
+      "epoch": 0.8706015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 1.2714890836991744e-06,
-      "loss": 0.7737,
+      "learning_rate": 8.654428203572796e-07,
+      "loss": 0.8124,
       "step": 30680
     },
     {
-      "epoch": 0.8426299744582681,
+      "epoch": 0.8706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.2710550462911286e-06,
-      "loss": 0.7784,
+      "learning_rate": 8.650688510163941e-07,
+      "loss": 0.8462,
       "step": 30681
     },
     {
-      "epoch": 0.8426574386861113,
+      "epoch": 0.8706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.270621077949873e-06,
-      "loss": 0.7914,
+      "learning_rate": 8.646949588386811e-07,
+      "loss": 0.8525,
       "step": 30682
     },
     {
-      "epoch": 0.8426849029139546,
+      "epoch": 0.8706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.2701871786788455e-06,
-      "loss": 0.8229,
+      "learning_rate": 8.643211438272992e-07,
+      "loss": 0.7363,
       "step": 30683
     },
     {
-      "epoch": 0.8427123671417978,
+      "epoch": 0.8707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.2697533484814818e-06,
-      "loss": 0.8054,
+      "learning_rate": 8.639474059854103e-07,
+      "loss": 0.8721,
       "step": 30684
     },
     {
-      "epoch": 0.842739831369641,
+      "epoch": 0.8707434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.2693195873612073e-06,
-      "loss": 0.7442,
+      "learning_rate": 8.635737453161652e-07,
+      "loss": 0.7445,
       "step": 30685
     },
     {
-      "epoch": 0.8427672955974843,
+      "epoch": 0.870771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.26888589532146e-06,
-      "loss": 0.7583,
+      "learning_rate": 8.632001618227248e-07,
+      "loss": 0.8094,
       "step": 30686
     },
     {
-      "epoch": 0.8427947598253275,
+      "epoch": 0.870800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.2684522723656689e-06,
-      "loss": 0.7684,
+      "learning_rate": 8.628266555082443e-07,
+      "loss": 0.865,
       "step": 30687
     },
     {
-      "epoch": 0.8428222240531708,
+      "epoch": 0.8708286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 1.2680187184972648e-06,
-      "loss": 0.8493,
+      "learning_rate": 8.624532263758767e-07,
+      "loss": 0.931,
       "step": 30688
     },
     {
-      "epoch": 0.842849688281014,
+      "epoch": 0.8708569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.2675852337196827e-06,
-      "loss": 0.7706,
+      "learning_rate": 8.620798744287761e-07,
+      "loss": 0.873,
       "step": 30689
     },
     {
-      "epoch": 0.8428771525088572,
+      "epoch": 0.8708853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 1.2671518180363473e-06,
-      "loss": 0.8533,
+      "learning_rate": 8.617065996701013e-07,
+      "loss": 0.8497,
       "step": 30690
     },
     {
-      "epoch": 0.8429046167367005,
+      "epoch": 0.8709137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 1.2667184714506885e-06,
-      "loss": 0.8437,
+      "learning_rate": 8.613334021029984e-07,
+      "loss": 0.7761,
       "step": 30691
     },
     {
-      "epoch": 0.8429320809645436,
+      "epoch": 0.8709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.2662851939661358e-06,
-      "loss": 0.8015,
+      "learning_rate": 8.609602817306217e-07,
+      "loss": 0.7439,
       "step": 30692
     },
     {
-      "epoch": 0.8429595451923869,
+      "epoch": 0.8709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.2658519855861184e-06,
-      "loss": 0.9031,
+      "learning_rate": 8.605872385561276e-07,
+      "loss": 0.7203,
       "step": 30693
     },
     {
-      "epoch": 0.8429870094202302,
+      "epoch": 0.8709988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 1.2654188463140626e-06,
-      "loss": 0.7812,
+      "learning_rate": 8.602142725826624e-07,
+      "loss": 0.8335,
       "step": 30694
     },
     {
-      "epoch": 0.8430144736480734,
+      "epoch": 0.8710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.264985776153398e-06,
-      "loss": 0.78,
+      "learning_rate": 8.59841383813379e-07,
+      "loss": 0.7589,
       "step": 30695
     },
     {
-      "epoch": 0.8430419378759166,
+      "epoch": 0.87105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.2645527751075482e-06,
-      "loss": 0.7839,
+      "learning_rate": 8.594685722514273e-07,
+      "loss": 0.8018,
       "step": 30696
     },
     {
-      "epoch": 0.8430694021037598,
+      "epoch": 0.8710839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 1.2641198431799407e-06,
-      "loss": 0.7583,
+      "learning_rate": 8.590958378999537e-07,
+      "loss": 0.9158,
       "step": 30697
     },
     {
-      "epoch": 0.8430968663316031,
+      "epoch": 0.8711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.2636869803740026e-06,
-      "loss": 0.8958,
+      "learning_rate": 8.587231807621099e-07,
+      "loss": 0.8862,
       "step": 30698
     },
     {
-      "epoch": 0.8431243305594464,
+      "epoch": 0.8711407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 1.2632541866931557e-06,
-      "loss": 0.7724,
+      "learning_rate": 8.583506008410403e-07,
+      "loss": 0.8286,
       "step": 30699
     },
     {
-      "epoch": 0.8431517947872895,
+      "epoch": 0.8711691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 1.2628214621408253e-06,
-      "loss": 0.8886,
+      "learning_rate": 8.579780981398955e-07,
+      "loss": 0.8871,
       "step": 30700
     },
     {
-      "epoch": 0.8431792590151328,
+      "epoch": 0.8711975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.2623888067204382e-06,
-      "loss": 0.7794,
+      "learning_rate": 8.576056726618209e-07,
+      "loss": 0.8668,
       "step": 30701
     },
     {
-      "epoch": 0.8432067232429761,
+      "epoch": 0.8712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.261956220435414e-06,
-      "loss": 0.7047,
+      "learning_rate": 8.572333244099617e-07,
+      "loss": 0.9214,
       "step": 30702
     },
     {
-      "epoch": 0.8432341874708192,
+      "epoch": 0.8712542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.2615237032891769e-06,
-      "loss": 0.7995,
+      "learning_rate": 8.568610533874622e-07,
+      "loss": 0.7167,
       "step": 30703
     },
     {
-      "epoch": 0.8432616516986625,
+      "epoch": 0.8712826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 1.2610912552851484e-06,
-      "loss": 0.7527,
+      "learning_rate": 8.564888595974718e-07,
+      "loss": 0.9073,
       "step": 30704
     },
     {
-      "epoch": 0.8432891159265057,
+      "epoch": 0.871311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.2606588764267536e-06,
-      "loss": 0.7721,
+      "learning_rate": 8.561167430431283e-07,
+      "loss": 0.7731,
       "step": 30705
     },
     {
-      "epoch": 0.843316580154349,
+      "epoch": 0.8713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.2602265667174096e-06,
-      "loss": 0.8168,
+      "learning_rate": 8.557447037275779e-07,
+      "loss": 0.7499,
       "step": 30706
     },
     {
-      "epoch": 0.8433440443821922,
+      "epoch": 0.8713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.2597943261605394e-06,
-      "loss": 0.8426,
+      "learning_rate": 8.553727416539626e-07,
+      "loss": 0.7696,
       "step": 30707
     },
     {
-      "epoch": 0.8433715086100354,
+      "epoch": 0.8713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.2593621547595602e-06,
-      "loss": 0.9324,
+      "learning_rate": 8.550008568254253e-07,
+      "loss": 0.7885,
       "step": 30708
     },
     {
-      "epoch": 0.8433989728378787,
+      "epoch": 0.8714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.2589300525178938e-06,
-      "loss": 0.9303,
+      "learning_rate": 8.54629049245107e-07,
+      "loss": 0.7779,
       "step": 30709
     },
     {
-      "epoch": 0.8434264370657218,
+      "epoch": 0.8714528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.2584980194389596e-06,
-      "loss": 0.8211,
+      "learning_rate": 8.542573189161496e-07,
+      "loss": 0.7882,
       "step": 30710
     },
     {
-      "epoch": 0.8434539012935651,
+      "epoch": 0.8714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.2580660555261738e-06,
-      "loss": 0.7834,
+      "learning_rate": 8.538856658416905e-07,
+      "loss": 0.8536,
       "step": 30711
     },
     {
-      "epoch": 0.8434813655214084,
+      "epoch": 0.8715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.2576341607829567e-06,
-      "loss": 0.8838,
+      "learning_rate": 8.535140900248696e-07,
+      "loss": 0.8237,
       "step": 30712
     },
     {
-      "epoch": 0.8435088297492516,
+      "epoch": 0.8715380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 1.257202335212726e-06,
-      "loss": 0.7635,
+      "learning_rate": 8.531425914688285e-07,
+      "loss": 0.7429,
       "step": 30713
     },
     {
-      "epoch": 0.8435362939770948,
+      "epoch": 0.871566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.2567705788188977e-06,
-      "loss": 0.7743,
+      "learning_rate": 8.527711701767016e-07,
+      "loss": 0.8773,
       "step": 30714
     },
     {
-      "epoch": 0.8435637582049381,
+      "epoch": 0.871594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.2563388916048847e-06,
-      "loss": 0.8506,
+      "learning_rate": 8.523998261516276e-07,
+      "loss": 0.7702,
       "step": 30715
     },
     {
-      "epoch": 0.8435912224327813,
+      "epoch": 0.8716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.255907273574105e-06,
-      "loss": 0.8844,
+      "learning_rate": 8.520285593967448e-07,
+      "loss": 0.8515,
       "step": 30716
     },
     {
-      "epoch": 0.8436186866606246,
+      "epoch": 0.8716515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 1.2554757247299754e-06,
-      "loss": 0.8475,
+      "learning_rate": 8.516573699151875e-07,
+      "loss": 0.7328,
       "step": 30717
     },
     {
-      "epoch": 0.8436461508884677,
+      "epoch": 0.8716799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 1.255044245075908e-06,
-      "loss": 0.839,
+      "learning_rate": 8.5128625771009e-07,
+      "loss": 0.7541,
       "step": 30718
     },
     {
-      "epoch": 0.843673615116311,
+      "epoch": 0.8717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.2546128346153207e-06,
-      "loss": 0.7552,
+      "learning_rate": 8.509152227845918e-07,
+      "loss": 0.7864,
       "step": 30719
     },
     {
-      "epoch": 0.8437010793441543,
+      "epoch": 0.8717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.2541814933516217e-06,
-      "loss": 0.8227,
+      "learning_rate": 8.505442651418217e-07,
+      "loss": 0.8027,
       "step": 30720
     },
     {
-      "epoch": 0.8437285435719974,
+      "epoch": 0.8717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.2537502212882291e-06,
-      "loss": 0.8339,
+      "learning_rate": 8.501733847849158e-07,
+      "loss": 0.8846,
       "step": 30721
     },
     {
-      "epoch": 0.8437560077998407,
+      "epoch": 0.8717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.2533190184285505e-06,
-      "loss": 0.9027,
+      "learning_rate": 8.498025817170063e-07,
+      "loss": 0.8473,
       "step": 30722
     },
     {
-      "epoch": 0.8437834720276839,
+      "epoch": 0.8718217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 1.2528878847759995e-06,
-      "loss": 0.9619,
+      "learning_rate": 8.494318559412251e-07,
+      "loss": 0.7593,
       "step": 30723
     },
     {
-      "epoch": 0.8438109362555272,
+      "epoch": 0.8718501702610669,
       "grad_norm": 0.0,
-      "learning_rate": 1.2524568203339882e-06,
-      "loss": 0.8315,
+      "learning_rate": 8.490612074607074e-07,
+      "loss": 0.8594,
       "step": 30724
     },
     {
-      "epoch": 0.8438384004833704,
+      "epoch": 0.8718785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.2520258251059291e-06,
-      "loss": 0.8931,
+      "learning_rate": 8.486906362785785e-07,
+      "loss": 0.8705,
       "step": 30725
     },
     {
-      "epoch": 0.8438658647112136,
+      "epoch": 0.8719069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 1.2515948990952288e-06,
-      "loss": 0.8315,
+      "learning_rate": 8.483201423979714e-07,
+      "loss": 0.78,
       "step": 30726
     },
     {
-      "epoch": 0.8438933289390569,
+      "epoch": 0.8719353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.2511640423052973e-06,
-      "loss": 0.7027,
+      "learning_rate": 8.47949725822017e-07,
+      "loss": 0.8234,
       "step": 30727
     },
     {
-      "epoch": 0.8439207931669002,
+      "epoch": 0.8719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.2507332547395456e-06,
-      "loss": 0.7539,
+      "learning_rate": 8.475793865538417e-07,
+      "loss": 0.9073,
       "step": 30728
     },
     {
-      "epoch": 0.8439482573947433,
+      "epoch": 0.8719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.2503025364013844e-06,
-      "loss": 0.8076,
+      "learning_rate": 8.472091245965741e-07,
+      "loss": 0.7867,
       "step": 30729
     },
     {
-      "epoch": 0.8439757216225866,
+      "epoch": 0.8720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.249871887294215e-06,
-      "loss": 0.7409,
+      "learning_rate": 8.468389399533438e-07,
+      "loss": 0.8636,
       "step": 30730
     },
     {
-      "epoch": 0.8440031858504298,
+      "epoch": 0.8720488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 1.249441307421453e-06,
-      "loss": 0.8676,
+      "learning_rate": 8.464688326272752e-07,
+      "loss": 0.791,
       "step": 30731
     },
     {
-      "epoch": 0.844030650078273,
+      "epoch": 0.8720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.2490107967864972e-06,
-      "loss": 0.7622,
+      "learning_rate": 8.460988026214955e-07,
+      "loss": 0.7315,
       "step": 30732
     },
     {
-      "epoch": 0.8440581143061163,
+      "epoch": 0.872105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.2485803553927577e-06,
-      "loss": 0.8777,
+      "learning_rate": 8.457288499391336e-07,
+      "loss": 0.8038,
       "step": 30733
     },
     {
-      "epoch": 0.8440855785339595,
+      "epoch": 0.8721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.2481499832436417e-06,
-      "loss": 0.797,
+      "learning_rate": 8.453589745833091e-07,
+      "loss": 0.8605,
       "step": 30734
     },
     {
-      "epoch": 0.8441130427618028,
+      "epoch": 0.8721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.2477196803425517e-06,
-      "loss": 0.9583,
+      "learning_rate": 8.449891765571483e-07,
+      "loss": 0.7727,
       "step": 30735
     },
     {
-      "epoch": 0.8441405069896459,
+      "epoch": 0.8721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.2472894466928953e-06,
-      "loss": 0.8074,
+      "learning_rate": 8.446194558637777e-07,
+      "loss": 0.749,
       "step": 30736
     },
     {
-      "epoch": 0.8441679712174892,
+      "epoch": 0.8722190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.2468592822980764e-06,
-      "loss": 0.8455,
+      "learning_rate": 8.44249812506317e-07,
+      "loss": 0.6828,
       "step": 30737
     },
     {
-      "epoch": 0.8441954354453325,
+      "epoch": 0.8722474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.2464291871614976e-06,
-      "loss": 0.8004,
+      "learning_rate": 8.438802464878881e-07,
+      "loss": 0.7164,
       "step": 30738
     },
     {
-      "epoch": 0.8442228996731757,
+      "epoch": 0.8722758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.2459991612865595e-06,
-      "loss": 0.8443,
+      "learning_rate": 8.435107578116164e-07,
+      "loss": 0.8026,
       "step": 30739
     },
     {
-      "epoch": 0.8442503639010189,
+      "epoch": 0.8723041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.2455692046766676e-06,
-      "loss": 0.8279,
+      "learning_rate": 8.431413464806193e-07,
+      "loss": 0.8325,
       "step": 30740
     },
     {
-      "epoch": 0.8442778281288622,
+      "epoch": 0.8723325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 1.2451393173352234e-06,
-      "loss": 0.7966,
+      "learning_rate": 8.42772012498021e-07,
+      "loss": 0.8371,
       "step": 30741
     },
     {
-      "epoch": 0.8443052923567054,
+      "epoch": 0.8723609534619751,
       "grad_norm": 0.0,
-      "learning_rate": 1.2447094992656284e-06,
-      "loss": 0.8013,
+      "learning_rate": 8.424027558669379e-07,
+      "loss": 0.7168,
       "step": 30742
     },
     {
-      "epoch": 0.8443327565845486,
+      "epoch": 0.8723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.2442797504712844e-06,
-      "loss": 0.9246,
+      "learning_rate": 8.420335765904908e-07,
+      "loss": 0.8206,
       "step": 30743
     },
     {
-      "epoch": 0.8443602208123918,
+      "epoch": 0.8724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.243850070955589e-06,
-      "loss": 0.833,
+      "learning_rate": 8.416644746717984e-07,
+      "loss": 0.838,
       "step": 30744
     },
     {
-      "epoch": 0.8443876850402351,
+      "epoch": 0.8724460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 1.2434204607219469e-06,
-      "loss": 0.8018,
+      "learning_rate": 8.41295450113977e-07,
+      "loss": 0.8563,
       "step": 30745
     },
     {
-      "epoch": 0.8444151492680784,
+      "epoch": 0.8724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.24299091977375e-06,
-      "loss": 0.8225,
+      "learning_rate": 8.409265029201441e-07,
+      "loss": 0.7724,
       "step": 30746
     },
     {
-      "epoch": 0.8444426134959215,
+      "epoch": 0.8725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.242561448114402e-06,
-      "loss": 0.7778,
+      "learning_rate": 8.405576330934184e-07,
+      "loss": 0.7355,
       "step": 30747
     },
     {
-      "epoch": 0.8444700777237648,
+      "epoch": 0.8725312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 1.2421320457473029e-06,
-      "loss": 0.8079,
+      "learning_rate": 8.40188840636913e-07,
+      "loss": 0.809,
       "step": 30748
     },
     {
-      "epoch": 0.844497541951608,
+      "epoch": 0.8725595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.2417027126758453e-06,
-      "loss": 0.8263,
+      "learning_rate": 8.398201255537431e-07,
+      "loss": 0.802,
       "step": 30749
     },
     {
-      "epoch": 0.8445250061794513,
+      "epoch": 0.8725879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 1.2412734489034272e-06,
-      "loss": 0.8643,
+      "learning_rate": 8.394514878470272e-07,
+      "loss": 0.8582,
       "step": 30750
     },
     {
-      "epoch": 0.8445524704072945,
+      "epoch": 0.872616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.2408442544334476e-06,
-      "loss": 0.7713,
+      "learning_rate": 8.390829275198741e-07,
+      "loss": 0.7768,
       "step": 30751
     },
     {
-      "epoch": 0.8445799346351377,
+      "epoch": 0.872644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.2404151292693012e-06,
-      "loss": 0.7192,
+      "learning_rate": 8.387144445753992e-07,
+      "loss": 0.6954,
       "step": 30752
     },
     {
-      "epoch": 0.844607398862981,
+      "epoch": 0.8726730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 1.2399860734143842e-06,
-      "loss": 0.8354,
+      "learning_rate": 8.383460390167164e-07,
+      "loss": 0.8907,
       "step": 30753
     },
     {
-      "epoch": 0.8446348630908243,
+      "epoch": 0.8727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.239557086872092e-06,
-      "loss": 0.7742,
+      "learning_rate": 8.379777108469355e-07,
+      "loss": 0.7773,
       "step": 30754
     },
     {
-      "epoch": 0.8446623273186674,
+      "epoch": 0.8727298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 1.2391281696458158e-06,
-      "loss": 0.7997,
+      "learning_rate": 8.376094600691698e-07,
+      "loss": 0.8436,
       "step": 30755
     },
     {
-      "epoch": 0.8446897915465107,
+      "epoch": 0.8727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.2386993217389497e-06,
-      "loss": 0.9386,
+      "learning_rate": 8.372412866865287e-07,
+      "loss": 0.8209,
       "step": 30756
     },
     {
-      "epoch": 0.8447172557743539,
+      "epoch": 0.8727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.2382705431548891e-06,
-      "loss": 0.7378,
+      "learning_rate": 8.36873190702121e-07,
+      "loss": 0.8971,
       "step": 30757
     },
     {
-      "epoch": 0.8447447200021971,
+      "epoch": 0.8728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.2378418338970266e-06,
-      "loss": 0.8286,
+      "learning_rate": 8.365051721190598e-07,
+      "loss": 0.8007,
       "step": 30758
     },
     {
-      "epoch": 0.8447721842300404,
+      "epoch": 0.8728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.2374131939687527e-06,
-      "loss": 0.706,
+      "learning_rate": 8.361372309404492e-07,
+      "loss": 0.7925,
       "step": 30759
     },
     {
-      "epoch": 0.8447996484578836,
+      "epoch": 0.8728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.2369846233734627e-06,
-      "loss": 0.8427,
+      "learning_rate": 8.357693671694001e-07,
+      "loss": 0.7896,
       "step": 30760
     },
     {
-      "epoch": 0.8448271126857269,
+      "epoch": 0.872900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.2365561221145417e-06,
-      "loss": 0.8687,
+      "learning_rate": 8.354015808090199e-07,
+      "loss": 0.8234,
       "step": 30761
     },
     {
-      "epoch": 0.84485457691357,
+      "epoch": 0.8729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.2361276901953868e-06,
-      "loss": 0.8665,
+      "learning_rate": 8.350338718624129e-07,
+      "loss": 0.8623,
       "step": 30762
     },
     {
-      "epoch": 0.8448820411414133,
+      "epoch": 0.8729568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.235699327619382e-06,
-      "loss": 0.7751,
+      "learning_rate": 8.346662403326877e-07,
+      "loss": 0.799,
       "step": 30763
     },
     {
-      "epoch": 0.8449095053692566,
+      "epoch": 0.8729852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.2352710343899177e-06,
-      "loss": 0.8217,
+      "learning_rate": 8.342986862229497e-07,
+      "loss": 0.7692,
       "step": 30764
     },
     {
-      "epoch": 0.8449369695970997,
+      "epoch": 0.8730136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 1.234842810510386e-06,
-      "loss": 0.8907,
+      "learning_rate": 8.339312095363017e-07,
+      "loss": 0.8273,
       "step": 30765
     },
     {
-      "epoch": 0.844964433824943,
+      "epoch": 0.8730419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 1.234414655984174e-06,
-      "loss": 0.7184,
+      "learning_rate": 8.335638102758481e-07,
+      "loss": 0.7997,
       "step": 30766
     },
     {
-      "epoch": 0.8449918980527863,
+      "epoch": 0.8730703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.233986570814667e-06,
-      "loss": 0.8191,
+      "learning_rate": 8.331964884446953e-07,
+      "loss": 0.867,
       "step": 30767
     },
     {
-      "epoch": 0.8450193622806295,
+      "epoch": 0.8730987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 1.2335585550052542e-06,
-      "loss": 0.805,
+      "learning_rate": 8.328292440459396e-07,
+      "loss": 0.819,
       "step": 30768
     },
     {
-      "epoch": 0.8450468265084727,
+      "epoch": 0.8731271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 1.2331306085593242e-06,
-      "loss": 0.8374,
+      "learning_rate": 8.324620770826919e-07,
+      "loss": 0.8201,
       "step": 30769
     },
     {
-      "epoch": 0.8450742907363159,
+      "epoch": 0.873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.2327027314802574e-06,
-      "loss": 0.7513,
+      "learning_rate": 8.320949875580464e-07,
+      "loss": 0.8178,
       "step": 30770
     },
     {
-      "epoch": 0.8451017549641592,
+      "epoch": 0.8731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.2322749237714438e-06,
-      "loss": 0.8705,
+      "learning_rate": 8.317279754751073e-07,
+      "loss": 0.8373,
       "step": 30771
     },
     {
-      "epoch": 0.8451292191920025,
+      "epoch": 0.8732122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 1.231847185436269e-06,
-      "loss": 0.8088,
+      "learning_rate": 8.313610408369754e-07,
+      "loss": 0.7811,
       "step": 30772
     },
     {
-      "epoch": 0.8451566834198456,
+      "epoch": 0.8732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.231419516478114e-06,
-      "loss": 0.8203,
+      "learning_rate": 8.309941836467472e-07,
+      "loss": 0.7711,
       "step": 30773
     },
     {
-      "epoch": 0.8451841476476889,
+      "epoch": 0.8732690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 1.2309919169003647e-06,
-      "loss": 0.8668,
+      "learning_rate": 8.306274039075246e-07,
+      "loss": 0.8773,
       "step": 30774
     },
     {
-      "epoch": 0.8452116118755321,
+      "epoch": 0.8732973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 1.2305643867064031e-06,
-      "loss": 0.845,
+      "learning_rate": 8.302607016224052e-07,
+      "loss": 0.7532,
       "step": 30775
     },
     {
-      "epoch": 0.8452390761033753,
+      "epoch": 0.8733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.2301369258996133e-06,
-      "loss": 0.8335,
+      "learning_rate": 8.298940767944841e-07,
+      "loss": 0.9122,
       "step": 30776
     },
     {
-      "epoch": 0.8452665403312186,
+      "epoch": 0.8733541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 1.2297095344833798e-06,
-      "loss": 0.9178,
+      "learning_rate": 8.295275294268601e-07,
+      "loss": 0.7998,
       "step": 30777
     },
     {
-      "epoch": 0.8452940045590618,
+      "epoch": 0.8733825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.2292822124610814e-06,
-      "loss": 0.8414,
+      "learning_rate": 8.291610595226307e-07,
+      "loss": 0.8118,
       "step": 30778
     },
     {
-      "epoch": 0.8453214687869051,
+      "epoch": 0.8734108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 1.2288549598360977e-06,
-      "loss": 0.8081,
+      "learning_rate": 8.287946670848889e-07,
+      "loss": 0.7311,
       "step": 30779
     },
     {
-      "epoch": 0.8453489330147483,
+      "epoch": 0.873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.2284277766118114e-06,
-      "loss": 0.8221,
+      "learning_rate": 8.284283521167303e-07,
+      "loss": 0.8638,
       "step": 30780
     },
     {
-      "epoch": 0.8453763972425915,
+      "epoch": 0.8734676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.2280006627916018e-06,
-      "loss": 0.8231,
+      "learning_rate": 8.280621146212519e-07,
+      "loss": 0.643,
       "step": 30781
     },
     {
-      "epoch": 0.8454038614704348,
+      "epoch": 0.8734960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 1.22757361837885e-06,
-      "loss": 0.8152,
+      "learning_rate": 8.276959546015429e-07,
+      "loss": 0.7706,
       "step": 30782
     },
     {
-      "epoch": 0.845431325698278,
+      "epoch": 0.8735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.227146643376933e-06,
-      "loss": 0.7653,
+      "learning_rate": 8.273298720606993e-07,
+      "loss": 0.7343,
       "step": 30783
     },
     {
-      "epoch": 0.8454587899261212,
+      "epoch": 0.8735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.2267197377892314e-06,
-      "loss": 0.7086,
+      "learning_rate": 8.269638670018121e-07,
+      "loss": 0.7631,
       "step": 30784
     },
     {
-      "epoch": 0.8454862541539645,
+      "epoch": 0.8735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.2262929016191205e-06,
-      "loss": 0.7242,
+      "learning_rate": 8.265979394279732e-07,
+      "loss": 0.9211,
       "step": 30785
     },
     {
-      "epoch": 0.8455137183818077,
+      "epoch": 0.8736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.2258661348699807e-06,
-      "loss": 0.8548,
+      "learning_rate": 8.262320893422759e-07,
+      "loss": 0.7629,
       "step": 30786
     },
     {
-      "epoch": 0.8455411826096509,
+      "epoch": 0.8736379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.225439437545185e-06,
-      "loss": 0.7903,
+      "learning_rate": 8.258663167478065e-07,
+      "loss": 0.7369,
       "step": 30787
     },
     {
-      "epoch": 0.8455686468374941,
+      "epoch": 0.87366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.2250128096481105e-06,
-      "loss": 0.7525,
+      "learning_rate": 8.255006216476569e-07,
+      "loss": 0.8349,
       "step": 30788
     },
     {
-      "epoch": 0.8455961110653374,
+      "epoch": 0.8736946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 1.2245862511821348e-06,
-      "loss": 0.805,
+      "learning_rate": 8.25135004044918e-07,
+      "loss": 0.835,
       "step": 30789
     },
     {
-      "epoch": 0.8456235752931807,
+      "epoch": 0.8737230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.224159762150634e-06,
-      "loss": 0.7586,
+      "learning_rate": 8.24769463942674e-07,
+      "loss": 0.8059,
       "step": 30790
     },
     {
-      "epoch": 0.8456510395210238,
+      "epoch": 0.8737514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.2237333425569776e-06,
-      "loss": 0.7748,
+      "learning_rate": 8.244040013440147e-07,
+      "loss": 0.8506,
       "step": 30791
     },
     {
-      "epoch": 0.8456785037488671,
+      "epoch": 0.8737797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 1.2233069924045437e-06,
-      "loss": 0.7867,
+      "learning_rate": 8.240386162520298e-07,
+      "loss": 0.824,
       "step": 30792
     },
     {
-      "epoch": 0.8457059679767104,
+      "epoch": 0.8738081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 1.2228807116967056e-06,
-      "loss": 0.8147,
+      "learning_rate": 8.236733086698013e-07,
+      "loss": 0.8278,
       "step": 30793
     },
     {
-      "epoch": 0.8457334322045535,
+      "epoch": 0.8738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.2224545004368338e-06,
-      "loss": 0.832,
+      "learning_rate": 8.233080786004167e-07,
+      "loss": 0.8368,
       "step": 30794
     },
     {
-      "epoch": 0.8457608964323968,
+      "epoch": 0.8738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.2220283586283022e-06,
-      "loss": 0.9324,
+      "learning_rate": 8.229429260469623e-07,
+      "loss": 0.85,
       "step": 30795
     },
     {
-      "epoch": 0.84578836066024,
+      "epoch": 0.8738933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 1.2216022862744847e-06,
-      "loss": 0.6624,
+      "learning_rate": 8.225778510125204e-07,
+      "loss": 0.8107,
       "step": 30796
     },
     {
-      "epoch": 0.8458158248880833,
+      "epoch": 0.8739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.2211762833787477e-06,
-      "loss": 0.8919,
+      "learning_rate": 8.222128535001761e-07,
+      "loss": 0.8246,
       "step": 30797
     },
     {
-      "epoch": 0.8458432891159265,
+      "epoch": 0.873950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.220750349944465e-06,
-      "loss": 0.9333,
+      "learning_rate": 8.218479335130136e-07,
+      "loss": 0.7978,
       "step": 30798
     },
     {
-      "epoch": 0.8458707533437697,
+      "epoch": 0.8739784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 1.2203244859750052e-06,
-      "loss": 0.8427,
+      "learning_rate": 8.214830910541116e-07,
+      "loss": 0.9156,
       "step": 30799
     },
     {
-      "epoch": 0.845898217571613,
+      "epoch": 0.8740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.2198986914737398e-06,
-      "loss": 0.8838,
+      "learning_rate": 8.211183261265554e-07,
+      "loss": 0.7498,
       "step": 30800
     },
     {
-      "epoch": 0.8459256817994562,
+      "epoch": 0.8740351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 1.2194729664440398e-06,
-      "loss": 0.7827,
+      "learning_rate": 8.20753638733428e-07,
+      "loss": 0.8253,
       "step": 30801
     },
     {
-      "epoch": 0.8459531460272994,
+      "epoch": 0.8740635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 1.2190473108892698e-06,
-      "loss": 0.774,
+      "learning_rate": 8.203890288778049e-07,
+      "loss": 0.8255,
       "step": 30802
     },
     {
-      "epoch": 0.8459806102551427,
+      "epoch": 0.8740919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.2186217248127975e-06,
-      "loss": 0.7433,
+      "learning_rate": 8.20024496562768e-07,
+      "loss": 0.8326,
       "step": 30803
     },
     {
-      "epoch": 0.8460080744829859,
+      "epoch": 0.8741203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 1.218196208217992e-06,
-      "loss": 0.814,
+      "learning_rate": 8.196600417913991e-07,
+      "loss": 0.7773,
       "step": 30804
     },
     {
-      "epoch": 0.8460355387108291,
+      "epoch": 0.8741486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.217770761108219e-06,
-      "loss": 0.8543,
+      "learning_rate": 8.192956645667727e-07,
+      "loss": 0.8274,
       "step": 30805
     },
     {
-      "epoch": 0.8460630029386724,
+      "epoch": 0.8741770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 1.2173453834868455e-06,
-      "loss": 0.7804,
+      "learning_rate": 8.189313648919695e-07,
+      "loss": 0.7314,
       "step": 30806
     },
     {
-      "epoch": 0.8460904671665156,
+      "epoch": 0.874205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.2169200753572408e-06,
-      "loss": 0.7551,
+      "learning_rate": 8.185671427700648e-07,
+      "loss": 0.7775,
       "step": 30807
     },
     {
-      "epoch": 0.8461179313943589,
+      "epoch": 0.8742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.2164948367227635e-06,
-      "loss": 0.7551,
+      "learning_rate": 8.182029982041361e-07,
+      "loss": 0.8294,
       "step": 30808
     },
     {
-      "epoch": 0.846145395622202,
+      "epoch": 0.8742622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 1.216069667586781e-06,
-      "loss": 0.8203,
+      "learning_rate": 8.178389311972612e-07,
+      "loss": 0.9185,
       "step": 30809
     },
     {
-      "epoch": 0.8461728598500453,
+      "epoch": 0.8742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.2156445679526618e-06,
-      "loss": 0.8613,
+      "learning_rate": 8.174749417525119e-07,
+      "loss": 0.8025,
       "step": 30810
     },
     {
-      "epoch": 0.8462003240778886,
+      "epoch": 0.8743189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.2152195378237618e-06,
-      "loss": 0.8976,
+      "learning_rate": 8.171110298729645e-07,
+      "loss": 0.84,
       "step": 30811
     },
     {
-      "epoch": 0.8462277883057318,
+      "epoch": 0.8743473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 1.2147945772034485e-06,
-      "loss": 0.8014,
+      "learning_rate": 8.167471955616946e-07,
+      "loss": 0.83,
       "step": 30812
     },
     {
-      "epoch": 0.846255252533575,
+      "epoch": 0.8743757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.2143696860950849e-06,
-      "loss": 0.8031,
+      "learning_rate": 8.163834388217728e-07,
+      "loss": 0.789,
       "step": 30813
     },
     {
-      "epoch": 0.8462827167614183,
+      "epoch": 0.8744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.2139448645020301e-06,
-      "loss": 0.8636,
+      "learning_rate": 8.160197596562702e-07,
+      "loss": 0.8269,
       "step": 30814
     },
     {
-      "epoch": 0.8463101809892615,
+      "epoch": 0.8744324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.2135201124276475e-06,
-      "loss": 0.8618,
+      "learning_rate": 8.156561580682665e-07,
+      "loss": 0.7746,
       "step": 30815
     },
     {
-      "epoch": 0.8463376452171048,
+      "epoch": 0.874460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.2130954298752962e-06,
-      "loss": 0.8957,
+      "learning_rate": 8.152926340608247e-07,
+      "loss": 0.794,
       "step": 30816
     },
     {
-      "epoch": 0.8463651094449479,
+      "epoch": 0.874489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.2126708168483404e-06,
-      "loss": 0.7424,
+      "learning_rate": 8.149291876370203e-07,
+      "loss": 0.8661,
       "step": 30817
     },
     {
-      "epoch": 0.8463925736727912,
+      "epoch": 0.8745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.2122462733501338e-06,
-      "loss": 0.9009,
+      "learning_rate": 8.145658187999228e-07,
+      "loss": 0.7688,
       "step": 30818
     },
     {
-      "epoch": 0.8464200379006345,
+      "epoch": 0.8745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.211821799384042e-06,
-      "loss": 0.7481,
+      "learning_rate": 8.142025275525989e-07,
+      "loss": 0.8826,
       "step": 30819
     },
     {
-      "epoch": 0.8464475021284776,
+      "epoch": 0.8745743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 1.2113973949534176e-06,
-      "loss": 0.7131,
+      "learning_rate": 8.138393138981193e-07,
+      "loss": 0.7918,
       "step": 30820
     },
     {
-      "epoch": 0.8464749663563209,
+      "epoch": 0.8746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.210973060061622e-06,
-      "loss": 0.7791,
+      "learning_rate": 8.134761778395539e-07,
+      "loss": 0.8359,
       "step": 30821
     },
     {
-      "epoch": 0.8465024305841641,
+      "epoch": 0.8746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.210548794712012e-06,
-      "loss": 0.8019,
+      "learning_rate": 8.131131193799668e-07,
+      "loss": 0.8529,
       "step": 30822
     },
     {
-      "epoch": 0.8465298948120074,
+      "epoch": 0.8746594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 1.2101245989079434e-06,
-      "loss": 0.7703,
+      "learning_rate": 8.127501385224257e-07,
+      "loss": 0.7403,
       "step": 30823
     },
     {
-      "epoch": 0.8465573590398506,
+      "epoch": 0.8746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.209700472652775e-06,
-      "loss": 0.7885,
+      "learning_rate": 8.123872352699991e-07,
+      "loss": 0.7884,
       "step": 30824
     },
     {
-      "epoch": 0.8465848232676938,
+      "epoch": 0.8747162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 1.2092764159498637e-06,
-      "loss": 0.8282,
+      "learning_rate": 8.12024409625749e-07,
+      "loss": 0.9632,
       "step": 30825
     },
     {
-      "epoch": 0.8466122874955371,
+      "epoch": 0.8747446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 1.2088524288025627e-06,
-      "loss": 0.7278,
+      "learning_rate": 8.116616615927409e-07,
+      "loss": 0.6648,
       "step": 30826
     },
     {
-      "epoch": 0.8466397517233804,
+      "epoch": 0.8747729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.2084285112142237e-06,
-      "loss": 0.7382,
+      "learning_rate": 8.112989911740421e-07,
+      "loss": 0.7803,
       "step": 30827
     },
     {
-      "epoch": 0.8466672159512235,
+      "epoch": 0.8748013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 1.2080046631882048e-06,
-      "loss": 0.8274,
+      "learning_rate": 8.109363983727125e-07,
+      "loss": 0.766,
       "step": 30828
     },
     {
-      "epoch": 0.8466946801790668,
+      "epoch": 0.8748297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.2075808847278582e-06,
-      "loss": 0.8693,
+      "learning_rate": 8.105738831918153e-07,
+      "loss": 0.9154,
       "step": 30829
     },
     {
-      "epoch": 0.84672214440691,
+      "epoch": 0.8748581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.2071571758365374e-06,
-      "loss": 0.7254,
+      "learning_rate": 8.102114456344145e-07,
+      "loss": 0.7935,
       "step": 30830
     },
     {
-      "epoch": 0.8467496086347532,
+      "epoch": 0.8748864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 1.206733536517597e-06,
-      "loss": 0.7595,
+      "learning_rate": 8.098490857035702e-07,
+      "loss": 0.8595,
       "step": 30831
     },
     {
-      "epoch": 0.8467770728625965,
+      "epoch": 0.8749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.206309966774385e-06,
-      "loss": 0.8705,
+      "learning_rate": 8.094868034023462e-07,
+      "loss": 0.835,
       "step": 30832
     },
     {
-      "epoch": 0.8468045370904397,
+      "epoch": 0.8749432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 1.2058864666102555e-06,
-      "loss": 0.7563,
+      "learning_rate": 8.09124598733797e-07,
+      "loss": 0.8004,
       "step": 30833
     },
     {
-      "epoch": 0.846832001318283,
+      "epoch": 0.8749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.2054630360285601e-06,
-      "loss": 0.7364,
+      "learning_rate": 8.087624717009868e-07,
+      "loss": 0.7961,
       "step": 30834
     },
     {
-      "epoch": 0.8468594655461261,
+      "epoch": 0.875,
       "grad_norm": 0.0,
-      "learning_rate": 1.2050396750326465e-06,
-      "loss": 0.7293,
+      "learning_rate": 8.084004223069752e-07,
+      "loss": 0.8994,
       "step": 30835
     },
     {
-      "epoch": 0.8468869297739694,
+      "epoch": 0.8750283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 1.2046163836258651e-06,
-      "loss": 0.8254,
+      "learning_rate": 8.080384505548155e-07,
+      "loss": 0.879,
       "step": 30836
     },
     {
-      "epoch": 0.8469143940018127,
+      "epoch": 0.8750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.2041931618115689e-06,
-      "loss": 0.8306,
+      "learning_rate": 8.076765564475686e-07,
+      "loss": 0.8522,
       "step": 30837
     },
     {
-      "epoch": 0.8469418582296558,
+      "epoch": 0.8750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.2037700095931005e-06,
-      "loss": 0.8012,
+      "learning_rate": 8.07314739988293e-07,
+      "loss": 0.8292,
       "step": 30838
     },
     {
-      "epoch": 0.8469693224574991,
+      "epoch": 0.8751135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.203346926973812e-06,
-      "loss": 0.7975,
+      "learning_rate": 8.06953001180042e-07,
+      "loss": 0.7334,
       "step": 30839
     },
     {
-      "epoch": 0.8469967866853424,
+      "epoch": 0.8751418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 1.2029239139570492e-06,
-      "loss": 0.8162,
+      "learning_rate": 8.065913400258718e-07,
+      "loss": 0.7527,
       "step": 30840
     },
     {
-      "epoch": 0.8470242509131856,
+      "epoch": 0.8751702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.2025009705461633e-06,
-      "loss": 0.9511,
+      "learning_rate": 8.06229756528839e-07,
+      "loss": 0.7972,
       "step": 30841
     },
     {
-      "epoch": 0.8470517151410288,
+      "epoch": 0.8751986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.2020780967444946e-06,
-      "loss": 0.7127,
+      "learning_rate": 8.058682506919946e-07,
+      "loss": 0.8046,
       "step": 30842
     },
     {
-      "epoch": 0.847079179368872,
+      "epoch": 0.8752270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 1.2016552925553937e-06,
-      "loss": 0.848,
+      "learning_rate": 8.055068225183959e-07,
+      "loss": 0.8608,
       "step": 30843
     },
     {
-      "epoch": 0.8471066435967153,
+      "epoch": 0.8752553916004541,
       "grad_norm": 0.0,
-      "learning_rate": 1.2012325579822025e-06,
-      "loss": 0.8309,
+      "learning_rate": 8.051454720110952e-07,
+      "loss": 0.8573,
       "step": 30844
     },
     {
-      "epoch": 0.8471341078245586,
+      "epoch": 0.8752837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.2008098930282685e-06,
-      "loss": 0.78,
+      "learning_rate": 8.047841991731409e-07,
+      "loss": 0.6901,
       "step": 30845
     },
     {
-      "epoch": 0.8471615720524017,
+      "epoch": 0.8753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.2003872976969344e-06,
-      "loss": 0.8094,
+      "learning_rate": 8.044230040075907e-07,
+      "loss": 0.7782,
       "step": 30846
     },
     {
-      "epoch": 0.847189036280245,
+      "epoch": 0.8753405221339388,
       "grad_norm": 0.0,
-      "learning_rate": 1.1999647719915441e-06,
-      "loss": 0.7694,
+      "learning_rate": 8.040618865174909e-07,
+      "loss": 0.9429,
       "step": 30847
     },
     {
-      "epoch": 0.8472165005080882,
+      "epoch": 0.8753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.199542315915443e-06,
-      "loss": 0.9073,
+      "learning_rate": 8.037008467058948e-07,
+      "loss": 0.8097,
       "step": 30848
     },
     {
-      "epoch": 0.8472439647359314,
+      "epoch": 0.8753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.1991199294719723e-06,
-      "loss": 0.8491,
+      "learning_rate": 8.033398845758522e-07,
+      "loss": 0.7748,
       "step": 30849
     },
     {
-      "epoch": 0.8472714289637747,
+      "epoch": 0.8754256526674233,
       "grad_norm": 0.0,
-      "learning_rate": 1.1986976126644746e-06,
-      "loss": 0.7122,
+      "learning_rate": 8.029790001304094e-07,
+      "loss": 0.862,
       "step": 30850
     },
     {
-      "epoch": 0.8472988931916179,
+      "epoch": 0.8754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.1982753654962886e-06,
-      "loss": 0.7977,
+      "learning_rate": 8.026181933726162e-07,
+      "loss": 0.7122,
       "step": 30851
     },
     {
-      "epoch": 0.8473263574194612,
+      "epoch": 0.8754824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 1.1978531879707577e-06,
-      "loss": 0.8702,
+      "learning_rate": 8.022574643055226e-07,
+      "loss": 0.7995,
       "step": 30852
     },
     {
-      "epoch": 0.8473538216473044,
+      "epoch": 0.875510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.1974310800912215e-06,
-      "loss": 0.7733,
+      "learning_rate": 8.018968129321714e-07,
+      "loss": 0.7852,
       "step": 30853
     },
     {
-      "epoch": 0.8473812858751476,
+      "epoch": 0.875539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.1970090418610202e-06,
-      "loss": 0.8285,
+      "learning_rate": 8.015362392556114e-07,
+      "loss": 0.838,
       "step": 30854
     },
     {
-      "epoch": 0.8474087501029909,
+      "epoch": 0.8755675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 1.1965870732834961e-06,
-      "loss": 0.8864,
+      "learning_rate": 8.011757432788902e-07,
+      "loss": 0.8386,
       "step": 30855
     },
     {
-      "epoch": 0.847436214330834,
+      "epoch": 0.8755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.1961651743619829e-06,
-      "loss": 0.7292,
+      "learning_rate": 8.008153250050499e-07,
+      "loss": 0.9053,
       "step": 30856
     },
     {
-      "epoch": 0.8474636785586773,
+      "epoch": 0.8756242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 1.1957433450998212e-06,
-      "loss": 0.8647,
+      "learning_rate": 8.004549844371357e-07,
+      "loss": 0.7743,
       "step": 30857
     },
     {
-      "epoch": 0.8474911427865206,
+      "epoch": 0.8756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.195321585500351e-06,
-      "loss": 0.7304,
+      "learning_rate": 8.000947215781929e-07,
+      "loss": 0.8594,
       "step": 30858
     },
     {
-      "epoch": 0.8475186070143638,
+      "epoch": 0.8756810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.1948998955669034e-06,
-      "loss": 0.7548,
+      "learning_rate": 7.997345364312626e-07,
+      "loss": 0.7351,
       "step": 30859
     },
     {
-      "epoch": 0.847546071242207,
+      "epoch": 0.8757094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.1944782753028194e-06,
-      "loss": 0.8593,
+      "learning_rate": 7.993744289993876e-07,
+      "loss": 0.867,
       "step": 30860
     },
     {
-      "epoch": 0.8475735354700502,
+      "epoch": 0.8757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.1940567247114377e-06,
-      "loss": 0.7748,
+      "learning_rate": 7.990143992856114e-07,
+      "loss": 0.7677,
       "step": 30861
     },
     {
-      "epoch": 0.8476009996978935,
+      "epoch": 0.8757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.1936352437960864e-06,
-      "loss": 0.8773,
+      "learning_rate": 7.986544472929736e-07,
+      "loss": 0.8368,
       "step": 30862
     },
     {
-      "epoch": 0.8476284639257368,
+      "epoch": 0.875794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.193213832560105e-06,
-      "loss": 0.7802,
+      "learning_rate": 7.982945730245162e-07,
+      "loss": 0.6977,
       "step": 30863
     },
     {
-      "epoch": 0.8476559281535799,
+      "epoch": 0.8758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.1927924910068267e-06,
-      "loss": 0.8568,
+      "learning_rate": 7.979347764832779e-07,
+      "loss": 0.8156,
       "step": 30864
     },
     {
-      "epoch": 0.8476833923814232,
+      "epoch": 0.8758513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.1923712191395886e-06,
-      "loss": 0.7997,
+      "learning_rate": 7.975750576722974e-07,
+      "loss": 0.9587,
       "step": 30865
     },
     {
-      "epoch": 0.8477108566092665,
+      "epoch": 0.8758796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.1919500169617192e-06,
-      "loss": 0.8175,
+      "learning_rate": 7.972154165946155e-07,
+      "loss": 0.7287,
       "step": 30866
     },
     {
-      "epoch": 0.8477383208371096,
+      "epoch": 0.8759080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 1.1915288844765548e-06,
-      "loss": 0.8145,
+      "learning_rate": 7.968558532532677e-07,
+      "loss": 0.7894,
       "step": 30867
     },
     {
-      "epoch": 0.8477657850649529,
+      "epoch": 0.8759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.1911078216874238e-06,
-      "loss": 0.9293,
+      "learning_rate": 7.964963676512915e-07,
+      "loss": 0.8428,
       "step": 30868
     },
     {
-      "epoch": 0.8477932492927961,
+      "epoch": 0.8759648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.1906868285976603e-06,
-      "loss": 0.7659,
+      "learning_rate": 7.961369597917268e-07,
+      "loss": 0.775,
       "step": 30869
     },
     {
-      "epoch": 0.8478207135206394,
+      "epoch": 0.8759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.1902659052105948e-06,
-      "loss": 0.8222,
+      "learning_rate": 7.957776296776043e-07,
+      "loss": 0.8506,
       "step": 30870
     },
     {
-      "epoch": 0.8478481777484826,
+      "epoch": 0.8760215664018162,
       "grad_norm": 0.0,
-      "learning_rate": 1.1898450515295579e-06,
-      "loss": 0.7729,
+      "learning_rate": 7.954183773119628e-07,
+      "loss": 0.8162,
       "step": 30871
     },
     {
-      "epoch": 0.8478756419763258,
+      "epoch": 0.876049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.1894242675578816e-06,
-      "loss": 0.8623,
+      "learning_rate": 7.950592026978377e-07,
+      "loss": 0.8545,
       "step": 30872
     },
     {
-      "epoch": 0.8479031062041691,
+      "epoch": 0.8760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.189003553298892e-06,
-      "loss": 0.8511,
+      "learning_rate": 7.947001058382586e-07,
+      "loss": 0.7322,
       "step": 30873
     },
     {
-      "epoch": 0.8479305704320123,
+      "epoch": 0.8761066969353007,
       "grad_norm": 0.0,
-      "learning_rate": 1.1885829087559208e-06,
-      "loss": 0.8502,
+      "learning_rate": 7.943410867362622e-07,
+      "loss": 0.6706,
       "step": 30874
     },
     {
-      "epoch": 0.8479580346598555,
+      "epoch": 0.8761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.1881623339322923e-06,
-      "loss": 0.8577,
+      "learning_rate": 7.939821453948826e-07,
+      "loss": 0.8378,
       "step": 30875
     },
     {
-      "epoch": 0.8479854988876988,
+      "epoch": 0.8761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.187741828831338e-06,
-      "loss": 0.7885,
+      "learning_rate": 7.936232818171453e-07,
+      "loss": 0.8065,
       "step": 30876
     },
     {
-      "epoch": 0.848012963115542,
+      "epoch": 0.8761918274687854,
       "grad_norm": 0.0,
-      "learning_rate": 1.1873213934563833e-06,
-      "loss": 0.6711,
+      "learning_rate": 7.932644960060898e-07,
+      "loss": 0.7721,
       "step": 30877
     },
     {
-      "epoch": 0.8480404273433853,
+      "epoch": 0.8762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.1869010278107573e-06,
-      "loss": 0.7593,
+      "learning_rate": 7.929057879647417e-07,
+      "loss": 0.7579,
       "step": 30878
     },
     {
-      "epoch": 0.8480678915712285,
+      "epoch": 0.8762485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 1.1864807318977822e-06,
-      "loss": 0.7535,
+      "learning_rate": 7.925471576961319e-07,
+      "loss": 0.6696,
       "step": 30879
     },
     {
-      "epoch": 0.8480953557990717,
+      "epoch": 0.8762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.1860605057207864e-06,
-      "loss": 0.7611,
+      "learning_rate": 7.921886052032913e-07,
+      "loss": 0.8781,
       "step": 30880
     },
     {
-      "epoch": 0.848122820026915,
+      "epoch": 0.8763053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.1856403492830948e-06,
-      "loss": 0.831,
+      "learning_rate": 7.918301304892461e-07,
+      "loss": 0.8037,
       "step": 30881
     },
     {
-      "epoch": 0.8481502842547581,
+      "epoch": 0.87633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.1852202625880305e-06,
-      "loss": 0.7971,
+      "learning_rate": 7.914717335570266e-07,
+      "loss": 0.7478,
       "step": 30882
     },
     {
-      "epoch": 0.8481777484826014,
+      "epoch": 0.8763620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.1848002456389163e-06,
-      "loss": 0.9251,
+      "learning_rate": 7.9111341440966e-07,
+      "loss": 0.7444,
       "step": 30883
     },
     {
-      "epoch": 0.8482052127104447,
+      "epoch": 0.8763904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 1.1843802984390806e-06,
-      "loss": 0.9106,
+      "learning_rate": 7.907551730501717e-07,
+      "loss": 0.7817,
       "step": 30884
     },
     {
-      "epoch": 0.8482326769382879,
+      "epoch": 0.8764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.1839604209918409e-06,
-      "loss": 0.9163,
+      "learning_rate": 7.903970094815872e-07,
+      "loss": 0.8155,
       "step": 30885
     },
     {
-      "epoch": 0.8482601411661311,
+      "epoch": 0.8764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.1835406133005212e-06,
-      "loss": 0.6989,
+      "learning_rate": 7.900389237069361e-07,
+      "loss": 0.7862,
       "step": 30886
     },
     {
-      "epoch": 0.8482876053939743,
+      "epoch": 0.8764755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 1.183120875368442e-06,
-      "loss": 0.8424,
+      "learning_rate": 7.896809157292373e-07,
+      "loss": 0.9027,
       "step": 30887
     },
     {
-      "epoch": 0.8483150696218176,
+      "epoch": 0.8765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.1827012071989275e-06,
-      "loss": 0.7244,
+      "learning_rate": 7.893229855515184e-07,
+      "loss": 0.8271,
       "step": 30888
     },
     {
-      "epoch": 0.8483425338496609,
+      "epoch": 0.8765323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.1822816087952982e-06,
-      "loss": 0.7475,
+      "learning_rate": 7.889651331768045e-07,
+      "loss": 0.9244,
       "step": 30889
     },
     {
-      "epoch": 0.848369998077504,
+      "epoch": 0.876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.1818620801608716e-06,
-      "loss": 0.8253,
+      "learning_rate": 7.886073586081133e-07,
+      "loss": 0.6976,
       "step": 30890
     },
     {
-      "epoch": 0.8483974623053473,
+      "epoch": 0.8765891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 1.181442621298966e-06,
-      "loss": 0.7662,
+      "learning_rate": 7.882496618484714e-07,
+      "loss": 0.8876,
       "step": 30891
     },
     {
-      "epoch": 0.8484249265331906,
+      "epoch": 0.8766174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.1810232322129034e-06,
-      "loss": 0.8437,
+      "learning_rate": 7.878920429008974e-07,
+      "loss": 0.914,
       "step": 30892
     },
     {
-      "epoch": 0.8484523907610337,
+      "epoch": 0.8766458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.180603912906001e-06,
-      "loss": 0.825,
+      "learning_rate": 7.875345017684144e-07,
+      "loss": 0.7234,
       "step": 30893
     },
     {
-      "epoch": 0.848479854988877,
+      "epoch": 0.8766742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 1.1801846633815762e-06,
-      "loss": 0.8162,
+      "learning_rate": 7.871770384540423e-07,
+      "loss": 0.8349,
       "step": 30894
     },
     {
-      "epoch": 0.8485073192167202,
+      "epoch": 0.8767026106696936,
       "grad_norm": 0.0,
-      "learning_rate": 1.1797654836429473e-06,
-      "loss": 0.7593,
+      "learning_rate": 7.868196529607985e-07,
+      "loss": 0.7433,
       "step": 30895
     },
     {
-      "epoch": 0.8485347834445635,
+      "epoch": 0.8767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.1793463736934318e-06,
-      "loss": 0.8191,
+      "learning_rate": 7.864623452917031e-07,
+      "loss": 0.784,
       "step": 30896
     },
     {
-      "epoch": 0.8485622476724067,
+      "epoch": 0.8767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.1789273335363427e-06,
-      "loss": 0.8012,
+      "learning_rate": 7.861051154497767e-07,
+      "loss": 0.8125,
       "step": 30897
     },
     {
-      "epoch": 0.8485897119002499,
+      "epoch": 0.8767877412031783,
       "grad_norm": 0.0,
-      "learning_rate": 1.1785083631750006e-06,
-      "loss": 0.7482,
+      "learning_rate": 7.857479634380327e-07,
+      "loss": 0.8155,
       "step": 30898
     },
     {
-      "epoch": 0.8486171761280932,
+      "epoch": 0.8768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.178089462612715e-06,
-      "loss": 0.8067,
+      "learning_rate": 7.853908892594897e-07,
+      "loss": 0.7732,
       "step": 30899
     },
     {
-      "epoch": 0.8486446403559363,
+      "epoch": 0.876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.177670631852802e-06,
-      "loss": 0.831,
+      "learning_rate": 7.850338929171663e-07,
+      "loss": 0.7864,
       "step": 30900
     },
     {
-      "epoch": 0.8486721045837796,
+      "epoch": 0.8768728717366628,
       "grad_norm": 0.0,
-      "learning_rate": 1.1772518708985769e-06,
-      "loss": 0.7345,
+      "learning_rate": 7.846769744140736e-07,
+      "loss": 0.7408,
       "step": 30901
     },
     {
-      "epoch": 0.8486995688116229,
+      "epoch": 0.8769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.176833179753355e-06,
-      "loss": 0.8255,
+      "learning_rate": 7.843201337532292e-07,
+      "loss": 0.7157,
       "step": 30902
     },
     {
-      "epoch": 0.8487270330394661,
+      "epoch": 0.8769296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 1.1764145584204444e-06,
-      "loss": 0.795,
+      "learning_rate": 7.839633709376481e-07,
+      "loss": 0.789,
       "step": 30903
     },
     {
-      "epoch": 0.8487544972673093,
+      "epoch": 0.8769580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.1759960069031594e-06,
-      "loss": 0.935,
+      "learning_rate": 7.836066859703406e-07,
+      "loss": 0.8169,
       "step": 30904
     },
     {
-      "epoch": 0.8487819614951526,
+      "epoch": 0.8769863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.175577525204814e-06,
-      "loss": 0.7573,
+      "learning_rate": 7.832500788543229e-07,
+      "loss": 0.8188,
       "step": 30905
     },
     {
-      "epoch": 0.8488094257229958,
+      "epoch": 0.8770147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 1.1751591133287166e-06,
-      "loss": 0.7311,
+      "learning_rate": 7.828935495926048e-07,
+      "loss": 0.8241,
       "step": 30906
     },
     {
-      "epoch": 0.8488368899508391,
+      "epoch": 0.8770431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.1747407712781778e-06,
-      "loss": 0.8326,
+      "learning_rate": 7.825370981882008e-07,
+      "loss": 0.8167,
       "step": 30907
     },
     {
-      "epoch": 0.8488643541786822,
+      "epoch": 0.8770715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 1.1743224990565106e-06,
-      "loss": 0.792,
+      "learning_rate": 7.821807246441193e-07,
+      "loss": 0.8018,
       "step": 30908
     },
     {
-      "epoch": 0.8488918184065255,
+      "epoch": 0.877099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.173904296667021e-06,
-      "loss": 0.8034,
+      "learning_rate": 7.818244289633736e-07,
+      "loss": 0.8621,
       "step": 30909
     },
     {
-      "epoch": 0.8489192826343688,
+      "epoch": 0.8771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.173486164113019e-06,
-      "loss": 0.7475,
+      "learning_rate": 7.814682111489691e-07,
+      "loss": 0.8508,
       "step": 30910
     },
     {
-      "epoch": 0.8489467468622119,
+      "epoch": 0.8771566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 1.173068101397814e-06,
-      "loss": 0.8226,
+      "learning_rate": 7.811120712039178e-07,
+      "loss": 0.8101,
       "step": 30911
     },
     {
-      "epoch": 0.8489742110900552,
+      "epoch": 0.8771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.1726501085247133e-06,
-      "loss": 0.7531,
+      "learning_rate": 7.807560091312283e-07,
+      "loss": 0.8514,
       "step": 30912
     },
     {
-      "epoch": 0.8490016753178984,
+      "epoch": 0.8772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.1722321854970253e-06,
-      "loss": 0.8108,
+      "learning_rate": 7.804000249339061e-07,
+      "loss": 0.8801,
       "step": 30913
     },
     {
-      "epoch": 0.8490291395457417,
+      "epoch": 0.8772417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 1.1718143323180575e-06,
-      "loss": 0.7967,
+      "learning_rate": 7.800441186149599e-07,
+      "loss": 0.7429,
       "step": 30914
     },
     {
-      "epoch": 0.8490566037735849,
+      "epoch": 0.8772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.1713965489911117e-06,
-      "loss": 0.7925,
+      "learning_rate": 7.79688290177395e-07,
+      "loss": 0.7716,
       "step": 30915
     },
     {
-      "epoch": 0.8490840680014281,
+      "epoch": 0.8772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.1709788355194962e-06,
-      "loss": 0.8609,
+      "learning_rate": 7.793325396242169e-07,
+      "loss": 0.772,
       "step": 30916
     },
     {
-      "epoch": 0.8491115322292714,
+      "epoch": 0.8773269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.1705611919065163e-06,
-      "loss": 0.8074,
+      "learning_rate": 7.78976866958433e-07,
+      "loss": 0.8399,
       "step": 30917
     },
     {
-      "epoch": 0.8491389964571147,
+      "epoch": 0.877355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 1.170143618155476e-06,
-      "loss": 0.7643,
+      "learning_rate": 7.786212721830455e-07,
+      "loss": 0.7786,
       "step": 30918
     },
     {
-      "epoch": 0.8491664606849578,
+      "epoch": 0.877383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.1697261142696803e-06,
-      "loss": 0.799,
+      "learning_rate": 7.782657553010575e-07,
+      "loss": 0.7877,
       "step": 30919
     },
     {
-      "epoch": 0.8491939249128011,
+      "epoch": 0.8774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.1693086802524333e-06,
-      "loss": 0.7784,
+      "learning_rate": 7.779103163154756e-07,
+      "loss": 0.7516,
       "step": 30920
     },
     {
-      "epoch": 0.8492213891406443,
+      "epoch": 0.8774404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.168891316107036e-06,
-      "loss": 0.8105,
+      "learning_rate": 7.775549552292972e-07,
+      "loss": 1.0217,
       "step": 30921
     },
     {
-      "epoch": 0.8492488533684875,
+      "epoch": 0.8774687854710557,
       "grad_norm": 0.0,
-      "learning_rate": 1.168474021836792e-06,
-      "loss": 0.7515,
+      "learning_rate": 7.771996720455255e-07,
+      "loss": 0.8386,
       "step": 30922
     },
     {
-      "epoch": 0.8492763175963308,
+      "epoch": 0.8774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.1680567974450019e-06,
-      "loss": 0.7568,
+      "learning_rate": 7.76844466767166e-07,
+      "loss": 0.8236,
       "step": 30923
     },
     {
-      "epoch": 0.849303781824174,
+      "epoch": 0.8775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.1676396429349667e-06,
-      "loss": 0.7377,
+      "learning_rate": 7.76489339397215e-07,
+      "loss": 0.8245,
       "step": 30924
     },
     {
-      "epoch": 0.8493312460520173,
+      "epoch": 0.8775539160045402,
       "grad_norm": 0.0,
-      "learning_rate": 1.1672225583099883e-06,
-      "loss": 0.7824,
+      "learning_rate": 7.761342899386725e-07,
+      "loss": 0.8139,
       "step": 30925
     },
     {
-      "epoch": 0.8493587102798604,
+      "epoch": 0.8775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.1668055435733683e-06,
-      "loss": 0.7454,
+      "learning_rate": 7.757793183945395e-07,
+      "loss": 0.7612,
       "step": 30926
     },
     {
-      "epoch": 0.8493861745077037,
+      "epoch": 0.8776106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 1.166388598728403e-06,
-      "loss": 0.8163,
+      "learning_rate": 7.754244247678122e-07,
+      "loss": 0.8871,
       "step": 30927
     },
     {
-      "epoch": 0.849413638735547,
+      "epoch": 0.8776390465380249,
       "grad_norm": 0.0,
-      "learning_rate": 1.1659717237783919e-06,
-      "loss": 0.8426,
+      "learning_rate": 7.750696090614895e-07,
+      "loss": 0.7878,
       "step": 30928
     },
     {
-      "epoch": 0.8494411029633901,
+      "epoch": 0.8776674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.1655549187266379e-06,
-      "loss": 0.8706,
+      "learning_rate": 7.747148712785701e-07,
+      "loss": 0.7469,
       "step": 30929
     },
     {
-      "epoch": 0.8494685671912334,
+      "epoch": 0.8776958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 1.165138183576433e-06,
-      "loss": 0.8185,
+      "learning_rate": 7.743602114220472e-07,
+      "loss": 0.7568,
       "step": 30930
     },
     {
-      "epoch": 0.8494960314190767,
+      "epoch": 0.8777241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.1647215183310789e-06,
-      "loss": 0.9085,
+      "learning_rate": 7.740056294949183e-07,
+      "loss": 0.8555,
       "step": 30931
     },
     {
-      "epoch": 0.8495234956469199,
+      "epoch": 0.8777525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 1.1643049229938685e-06,
-      "loss": 0.8101,
+      "learning_rate": 7.7365112550018e-07,
+      "loss": 0.9123,
       "step": 30932
     },
     {
-      "epoch": 0.8495509598747631,
+      "epoch": 0.8777809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 1.1638883975681003e-06,
-      "loss": 0.8236,
+      "learning_rate": 7.732966994408242e-07,
+      "loss": 0.788,
       "step": 30933
     },
     {
-      "epoch": 0.8495784241026063,
+      "epoch": 0.8778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.1634719420570695e-06,
-      "loss": 0.7979,
+      "learning_rate": 7.729423513198453e-07,
+      "loss": 0.7563,
       "step": 30934
     },
     {
-      "epoch": 0.8496058883304496,
+      "epoch": 0.8778376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 1.1630555564640723e-06,
-      "loss": 0.8399,
+      "learning_rate": 7.725880811402397e-07,
+      "loss": 0.8025,
       "step": 30935
     },
     {
-      "epoch": 0.8496333525582929,
+      "epoch": 0.8778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.1626392407924014e-06,
-      "loss": 0.8514,
+      "learning_rate": 7.722338889049952e-07,
+      "loss": 0.7649,
       "step": 30936
     },
     {
-      "epoch": 0.849660816786136,
+      "epoch": 0.877894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1622229950453556e-06,
-      "loss": 0.8686,
+      "learning_rate": 7.718797746171058e-07,
+      "loss": 0.8349,
       "step": 30937
     },
     {
-      "epoch": 0.8496882810139793,
+      "epoch": 0.8779228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.1618068192262244e-06,
-      "loss": 0.8863,
+      "learning_rate": 7.715257382795627e-07,
+      "loss": 0.7017,
       "step": 30938
     },
     {
-      "epoch": 0.8497157452418225,
+      "epoch": 0.8779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.1613907133382985e-06,
-      "loss": 0.817,
+      "learning_rate": 7.711717798953566e-07,
+      "loss": 0.7736,
       "step": 30939
     },
     {
-      "epoch": 0.8497432094696658,
+      "epoch": 0.8779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.1609746773848741e-06,
-      "loss": 0.8287,
+      "learning_rate": 7.708178994674787e-07,
+      "loss": 0.8083,
       "step": 30940
     },
     {
-      "epoch": 0.849770673697509,
+      "epoch": 0.8780079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 1.1605587113692418e-06,
-      "loss": 0.7666,
+      "learning_rate": 7.704640969989163e-07,
+      "loss": 0.7417,
       "step": 30941
     },
     {
-      "epoch": 0.8497981379253522,
+      "epoch": 0.8780363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 1.1601428152946926e-06,
-      "loss": 0.8297,
+      "learning_rate": 7.701103724926573e-07,
+      "loss": 0.6993,
       "step": 30942
     },
     {
-      "epoch": 0.8498256021531955,
+      "epoch": 0.8780646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.1597269891645201e-06,
-      "loss": 0.8892,
+      "learning_rate": 7.697567259516936e-07,
+      "loss": 0.6878,
       "step": 30943
     },
     {
-      "epoch": 0.8498530663810387,
+      "epoch": 0.8780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.15931123298201e-06,
-      "loss": 0.7379,
+      "learning_rate": 7.694031573790072e-07,
+      "loss": 0.7807,
       "step": 30944
     },
     {
-      "epoch": 0.8498805306088819,
+      "epoch": 0.8781214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 1.1588955467504536e-06,
-      "loss": 0.8391,
+      "learning_rate": 7.690496667775882e-07,
+      "loss": 0.7584,
       "step": 30945
     },
     {
-      "epoch": 0.8499079948367252,
+      "epoch": 0.8781498297389331,
       "grad_norm": 0.0,
-      "learning_rate": 1.158479930473142e-06,
-      "loss": 0.8088,
+      "learning_rate": 7.686962541504228e-07,
+      "loss": 0.7783,
       "step": 30946
     },
     {
-      "epoch": 0.8499354590645684,
+      "epoch": 0.8781782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.1580643841533612e-06,
-      "loss": 0.8498,
+      "learning_rate": 7.683429195004932e-07,
+      "loss": 0.7449,
       "step": 30947
     },
     {
-      "epoch": 0.8499629232924116,
+      "epoch": 0.8782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.1576489077943987e-06,
-      "loss": 0.7515,
+      "learning_rate": 7.679896628307859e-07,
+      "loss": 0.8104,
       "step": 30948
     },
     {
-      "epoch": 0.8499903875202549,
+      "epoch": 0.8782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.157233501399546e-06,
-      "loss": 0.7213,
+      "learning_rate": 7.676364841442874e-07,
+      "loss": 0.7661,
       "step": 30949
     },
     {
-      "epoch": 0.8500178517480981,
+      "epoch": 0.8782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.1568181649720856e-06,
-      "loss": 0.7913,
+      "learning_rate": 7.672833834439764e-07,
+      "loss": 0.7256,
       "step": 30950
     },
     {
-      "epoch": 0.8500453159759414,
+      "epoch": 0.8782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.1564028985153054e-06,
-      "loss": 0.8439,
+      "learning_rate": 7.669303607328371e-07,
+      "loss": 0.7092,
       "step": 30951
     },
     {
-      "epoch": 0.8500727802037845,
+      "epoch": 0.8783200908059023,
       "grad_norm": 0.0,
-      "learning_rate": 1.1559877020324906e-06,
-      "loss": 0.8397,
+      "learning_rate": 7.665774160138528e-07,
+      "loss": 0.87,
       "step": 30952
     },
     {
-      "epoch": 0.8501002444316278,
+      "epoch": 0.8783484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.1555725755269288e-06,
-      "loss": 0.7541,
+      "learning_rate": 7.662245492900045e-07,
+      "loss": 0.7244,
       "step": 30953
     },
     {
-      "epoch": 0.8501277086594711,
+      "epoch": 0.8783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.1551575190019015e-06,
-      "loss": 0.8657,
+      "learning_rate": 7.658717605642729e-07,
+      "loss": 0.8015,
       "step": 30954
     },
     {
-      "epoch": 0.8501551728873142,
+      "epoch": 0.878405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.1547425324606953e-06,
-      "loss": 0.7815,
+      "learning_rate": 7.65519049839637e-07,
+      "loss": 0.7215,
       "step": 30955
     },
     {
-      "epoch": 0.8501826371151575,
+      "epoch": 0.878433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.1543276159065908e-06,
-      "loss": 0.8613,
+      "learning_rate": 7.651664171190765e-07,
+      "loss": 0.8345,
       "step": 30956
     },
     {
-      "epoch": 0.8502101013430008,
+      "epoch": 0.8784619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 1.1539127693428731e-06,
-      "loss": 0.8041,
+      "learning_rate": 7.648138624055724e-07,
+      "loss": 0.8436,
       "step": 30957
     },
     {
-      "epoch": 0.850237565570844,
+      "epoch": 0.8784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.1534979927728229e-06,
-      "loss": 0.7719,
+      "learning_rate": 7.644613857020988e-07,
+      "loss": 0.7752,
       "step": 30958
     },
     {
-      "epoch": 0.8502650297986872,
+      "epoch": 0.8785187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 1.1530832861997244e-06,
-      "loss": 0.7881,
+      "learning_rate": 7.641089870116347e-07,
+      "loss": 0.832,
       "step": 30959
     },
     {
-      "epoch": 0.8502924940265304,
+      "epoch": 0.8785471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 1.152668649626858e-06,
-      "loss": 0.8405,
+      "learning_rate": 7.637566663371598e-07,
+      "loss": 0.7076,
       "step": 30960
     },
     {
-      "epoch": 0.8503199582543737,
+      "epoch": 0.8785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.152254083057507e-06,
-      "loss": 0.7737,
+      "learning_rate": 7.634044236816451e-07,
+      "loss": 0.9467,
       "step": 30961
     },
     {
-      "epoch": 0.850347422482217,
+      "epoch": 0.8786038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.1518395864949471e-06,
-      "loss": 0.7763,
+      "learning_rate": 7.630522590480693e-07,
+      "loss": 0.8483,
       "step": 30962
     },
     {
-      "epoch": 0.8503748867100601,
+      "epoch": 0.8786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.1514251599424597e-06,
-      "loss": 0.8302,
+      "learning_rate": 7.627001724394067e-07,
+      "loss": 0.8597,
       "step": 30963
     },
     {
-      "epoch": 0.8504023509379034,
+      "epoch": 0.8786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.151010803403323e-06,
-      "loss": 0.8685,
+      "learning_rate": 7.623481638586305e-07,
+      "loss": 0.8462,
       "step": 30964
     },
     {
-      "epoch": 0.8504298151657466,
+      "epoch": 0.878688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.1505965168808163e-06,
-      "loss": 0.7284,
+      "learning_rate": 7.619962333087138e-07,
+      "loss": 0.8569,
       "step": 30965
     },
     {
-      "epoch": 0.8504572793935898,
+      "epoch": 0.8787173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 1.1501823003782186e-06,
-      "loss": 0.8072,
+      "learning_rate": 7.616443807926321e-07,
+      "loss": 0.8732,
       "step": 30966
     },
     {
-      "epoch": 0.8504847436214331,
+      "epoch": 0.8787457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.1497681538988093e-06,
-      "loss": 0.8387,
+      "learning_rate": 7.612926063133541e-07,
+      "loss": 0.683,
       "step": 30967
     },
     {
-      "epoch": 0.8505122078492763,
+      "epoch": 0.8787741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 1.14935407744586e-06,
-      "loss": 0.8008,
+      "learning_rate": 7.609409098738519e-07,
+      "loss": 0.8092,
       "step": 30968
     },
     {
-      "epoch": 0.8505396720771196,
+      "epoch": 0.8788024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 1.1489400710226496e-06,
-      "loss": 0.955,
+      "learning_rate": 7.605892914770985e-07,
+      "loss": 0.7726,
       "step": 30969
     },
     {
-      "epoch": 0.8505671363049628,
+      "epoch": 0.8788308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.148526134632455e-06,
-      "loss": 0.7999,
+      "learning_rate": 7.602377511260617e-07,
+      "loss": 0.7257,
       "step": 30970
     },
     {
-      "epoch": 0.850594600532806,
+      "epoch": 0.8788592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.1481122682785494e-06,
-      "loss": 0.856,
+      "learning_rate": 7.598862888237146e-07,
+      "loss": 0.8171,
       "step": 30971
     },
     {
-      "epoch": 0.8506220647606493,
+      "epoch": 0.8788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.1476984719642082e-06,
-      "loss": 0.8236,
+      "learning_rate": 7.595349045730205e-07,
+      "loss": 0.7205,
       "step": 30972
     },
     {
-      "epoch": 0.8506495289884924,
+      "epoch": 0.8789160045402952,
       "grad_norm": 0.0,
-      "learning_rate": 1.1472847456927072e-06,
-      "loss": 0.7834,
+      "learning_rate": 7.591835983769513e-07,
+      "loss": 0.7946,
       "step": 30973
     },
     {
-      "epoch": 0.8506769932163357,
+      "epoch": 0.87894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.1468710894673174e-06,
-      "loss": 0.8096,
+      "learning_rate": 7.588323702384748e-07,
+      "loss": 0.8581,
       "step": 30974
     },
     {
-      "epoch": 0.850704457444179,
+      "epoch": 0.8789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.1464575032913127e-06,
-      "loss": 0.794,
+      "learning_rate": 7.584812201605562e-07,
+      "loss": 0.7941,
       "step": 30975
     },
     {
-      "epoch": 0.8507319216720222,
+      "epoch": 0.8790011350737797,
       "grad_norm": 0.0,
-      "learning_rate": 1.146043987167965e-06,
-      "loss": 0.8678,
+      "learning_rate": 7.581301481461611e-07,
+      "loss": 0.8621,
       "step": 30976
     },
     {
-      "epoch": 0.8507593858998654,
+      "epoch": 0.8790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.1456305411005498e-06,
-      "loss": 0.7742,
+      "learning_rate": 7.577791541982582e-07,
+      "loss": 0.8433,
       "step": 30977
     },
     {
-      "epoch": 0.8507868501277086,
+      "epoch": 0.8790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.1452171650923338e-06,
-      "loss": 0.7476,
+      "learning_rate": 7.574282383198072e-07,
+      "loss": 0.8274,
       "step": 30978
     },
     {
-      "epoch": 0.8508143143555519,
+      "epoch": 0.8790862656072644,
       "grad_norm": 0.0,
-      "learning_rate": 1.1448038591465915e-06,
-      "loss": 0.7034,
+      "learning_rate": 7.570774005137771e-07,
+      "loss": 0.8351,
       "step": 30979
     },
     {
-      "epoch": 0.8508417785833952,
+      "epoch": 0.8791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.144390623266588e-06,
-      "loss": 0.7758,
+      "learning_rate": 7.567266407831308e-07,
+      "loss": 0.7695,
       "step": 30980
     },
     {
-      "epoch": 0.8508692428112383,
+      "epoch": 0.8791430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 1.1439774574555973e-06,
-      "loss": 0.8015,
+      "learning_rate": 7.563759591308284e-07,
+      "loss": 0.8057,
       "step": 30981
     },
     {
-      "epoch": 0.8508967070390816,
+      "epoch": 0.8791713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.143564361716888e-06,
-      "loss": 0.7472,
+      "learning_rate": 7.56025355559834e-07,
+      "loss": 0.8039,
       "step": 30982
     },
     {
-      "epoch": 0.8509241712669249,
+      "epoch": 0.879199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.1431513360537273e-06,
-      "loss": 0.8411,
+      "learning_rate": 7.556748300731087e-07,
+      "loss": 0.8287,
       "step": 30983
     },
     {
-      "epoch": 0.850951635494768,
+      "epoch": 0.879228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 1.1427383804693837e-06,
-      "loss": 0.7003,
+      "learning_rate": 7.553243826736134e-07,
+      "loss": 0.8195,
       "step": 30984
     },
     {
-      "epoch": 0.8509790997226113,
+      "epoch": 0.8792565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.142325494967128e-06,
-      "loss": 0.8103,
+      "learning_rate": 7.549740133643102e-07,
+      "loss": 0.8193,
       "step": 30985
     },
     {
-      "epoch": 0.8510065639504545,
+      "epoch": 0.8792849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 1.1419126795502232e-06,
-      "loss": 0.8053,
+      "learning_rate": 7.546237221481567e-07,
+      "loss": 0.7827,
       "step": 30986
     },
     {
-      "epoch": 0.8510340281782978,
+      "epoch": 0.8793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.1414999342219357e-06,
-      "loss": 0.7417,
+      "learning_rate": 7.542735090281105e-07,
+      "loss": 0.8288,
       "step": 30987
     },
     {
-      "epoch": 0.851061492406141,
+      "epoch": 0.8793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.1410872589855314e-06,
-      "loss": 0.7411,
+      "learning_rate": 7.53923374007135e-07,
+      "loss": 0.82,
       "step": 30988
     },
     {
-      "epoch": 0.8510889566339842,
+      "epoch": 0.8793700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 1.1406746538442759e-06,
-      "loss": 0.7683,
+      "learning_rate": 7.53573317088181e-07,
+      "loss": 0.6867,
       "step": 30989
     },
     {
-      "epoch": 0.8511164208618275,
+      "epoch": 0.8793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.1402621188014373e-06,
-      "loss": 0.7797,
+      "learning_rate": 7.532233382742105e-07,
+      "loss": 0.7805,
       "step": 30990
     },
     {
-      "epoch": 0.8511438850896706,
+      "epoch": 0.8794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.1398496538602743e-06,
-      "loss": 0.8428,
+      "learning_rate": 7.52873437568179e-07,
+      "loss": 0.7463,
       "step": 30991
     },
     {
-      "epoch": 0.8511713493175139,
+      "epoch": 0.8794551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 1.1394372590240521e-06,
-      "loss": 0.8997,
+      "learning_rate": 7.525236149730397e-07,
+      "loss": 0.7873,
       "step": 30992
     },
     {
-      "epoch": 0.8511988135453572,
+      "epoch": 0.8794835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 1.1390249342960357e-06,
-      "loss": 0.7567,
+      "learning_rate": 7.521738704917481e-07,
+      "loss": 0.8142,
       "step": 30993
     },
     {
-      "epoch": 0.8512262777732004,
+      "epoch": 0.8795119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.1386126796794883e-06,
-      "loss": 0.8094,
+      "learning_rate": 7.518242041272628e-07,
+      "loss": 0.7077,
       "step": 30994
     },
     {
-      "epoch": 0.8512537420010436,
+      "epoch": 0.8795402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.138200495177667e-06,
-      "loss": 0.7729,
+      "learning_rate": 7.514746158825314e-07,
+      "loss": 0.8478,
       "step": 30995
     },
     {
-      "epoch": 0.8512812062288869,
+      "epoch": 0.8795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.1377883807938394e-06,
-      "loss": 0.856,
+      "learning_rate": 7.511251057605107e-07,
+      "loss": 0.8141,
       "step": 30996
     },
     {
-      "epoch": 0.8513086704567301,
+      "epoch": 0.8795970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.1373763365312606e-06,
-      "loss": 0.7143,
+      "learning_rate": 7.507756737641525e-07,
+      "loss": 0.7767,
       "step": 30997
     },
     {
-      "epoch": 0.8513361346845734,
+      "epoch": 0.8796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.1369643623931925e-06,
-      "loss": 0.8133,
+      "learning_rate": 7.504263198964057e-07,
+      "loss": 0.7251,
       "step": 30998
     },
     {
-      "epoch": 0.8513635989124165,
+      "epoch": 0.8796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.1365524583828968e-06,
-      "loss": 0.7499,
+      "learning_rate": 7.500770441602257e-07,
+      "loss": 0.7027,
       "step": 30999
     },
     {
-      "epoch": 0.8513910631402598,
+      "epoch": 0.8796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.136140624503631e-06,
-      "loss": 0.737,
+      "learning_rate": 7.4972784655856e-07,
+      "loss": 0.969,
       "step": 31000
     },
     {
-      "epoch": 0.8514185273681031,
+      "epoch": 0.8797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.135728860758656e-06,
-      "loss": 0.79,
+      "learning_rate": 7.493787270943586e-07,
+      "loss": 0.7793,
       "step": 31001
     },
     {
-      "epoch": 0.8514459915959462,
+      "epoch": 0.879738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.1353171671512274e-06,
-      "loss": 0.8917,
+      "learning_rate": 7.490296857705737e-07,
+      "loss": 0.698,
       "step": 31002
     },
     {
-      "epoch": 0.8514734558237895,
+      "epoch": 0.8797673098751418,
       "grad_norm": 0.0,
-      "learning_rate": 1.1349055436846023e-06,
-      "loss": 0.7438,
+      "learning_rate": 7.486807225901483e-07,
+      "loss": 0.8721,
       "step": 31003
     },
     {
-      "epoch": 0.8515009200516328,
+      "epoch": 0.8797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.1344939903620379e-06,
-      "loss": 0.7766,
+      "learning_rate": 7.483318375560322e-07,
+      "loss": 0.7081,
       "step": 31004
     },
     {
-      "epoch": 0.851528384279476,
+      "epoch": 0.8798240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.1340825071867911e-06,
-      "loss": 0.7966,
+      "learning_rate": 7.479830306711755e-07,
+      "loss": 0.7912,
       "step": 31005
     },
     {
-      "epoch": 0.8515558485073192,
+      "epoch": 0.8798524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.1336710941621177e-06,
-      "loss": 0.7772,
+      "learning_rate": 7.476343019385202e-07,
+      "loss": 0.6914,
       "step": 31006
     },
     {
-      "epoch": 0.8515833127351624,
+      "epoch": 0.8798808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.1332597512912735e-06,
-      "loss": 0.8405,
+      "learning_rate": 7.472856513610139e-07,
+      "loss": 0.8059,
       "step": 31007
     },
     {
-      "epoch": 0.8516107769630057,
+      "epoch": 0.8799091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 1.1328484785775152e-06,
-      "loss": 0.8284,
+      "learning_rate": 7.46937078941603e-07,
+      "loss": 0.847,
       "step": 31008
     },
     {
-      "epoch": 0.851638241190849,
+      "epoch": 0.8799375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.132437276024092e-06,
-      "loss": 0.8237,
+      "learning_rate": 7.465885846832288e-07,
+      "loss": 0.7284,
       "step": 31009
     },
     {
-      "epoch": 0.8516657054186921,
+      "epoch": 0.8799659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 1.1320261436342616e-06,
-      "loss": 0.807,
+      "learning_rate": 7.462401685888365e-07,
+      "loss": 0.7863,
       "step": 31010
     },
     {
-      "epoch": 0.8516931696465354,
+      "epoch": 0.879994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.1316150814112735e-06,
-      "loss": 0.8358,
+      "learning_rate": 7.458918306613716e-07,
+      "loss": 0.8565,
       "step": 31011
     },
     {
-      "epoch": 0.8517206338743786,
+      "epoch": 0.8800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.131204089358383e-06,
-      "loss": 0.778,
+      "learning_rate": 7.455435709037728e-07,
+      "loss": 0.7269,
       "step": 31012
     },
     {
-      "epoch": 0.8517480981022219,
+      "epoch": 0.8800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.1307931674788408e-06,
-      "loss": 0.7632,
+      "learning_rate": 7.451953893189834e-07,
+      "loss": 0.8727,
       "step": 31013
     },
     {
-      "epoch": 0.8517755623300651,
+      "epoch": 0.8800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.130382315775901e-06,
-      "loss": 0.7462,
+      "learning_rate": 7.448472859099454e-07,
+      "loss": 0.7757,
       "step": 31014
     },
     {
-      "epoch": 0.8518030265579083,
+      "epoch": 0.8801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.1299715342528095e-06,
-      "loss": 0.7942,
+      "learning_rate": 7.444992606795975e-07,
+      "loss": 0.7706,
       "step": 31015
     },
     {
-      "epoch": 0.8518304907857516,
+      "epoch": 0.8801362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 1.1295608229128196e-06,
-      "loss": 0.7856,
+      "learning_rate": 7.441513136308809e-07,
+      "loss": 0.7676,
       "step": 31016
     },
     {
-      "epoch": 0.8518579550135948,
+      "epoch": 0.8801645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 1.12915018175918e-06,
-      "loss": 0.7484,
+      "learning_rate": 7.438034447667353e-07,
+      "loss": 0.7795,
       "step": 31017
     },
     {
-      "epoch": 0.851885419241438,
+      "epoch": 0.8801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.128739610795143e-06,
-      "loss": 0.8264,
+      "learning_rate": 7.434556540900962e-07,
+      "loss": 0.7326,
       "step": 31018
     },
     {
-      "epoch": 0.8519128834692813,
+      "epoch": 0.8802213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 1.1283291100239524e-06,
-      "loss": 0.8539,
+      "learning_rate": 7.431079416039045e-07,
+      "loss": 0.8133,
       "step": 31019
     },
     {
-      "epoch": 0.8519403476971245,
+      "epoch": 0.880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.1279186794488605e-06,
-      "loss": 0.8787,
+      "learning_rate": 7.427603073110967e-07,
+      "loss": 0.9627,
       "step": 31020
     },
     {
-      "epoch": 0.8519678119249677,
+      "epoch": 0.88027809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1275083190731106e-06,
-      "loss": 0.8599,
+      "learning_rate": 7.424127512146073e-07,
+      "loss": 0.8157,
       "step": 31021
     },
     {
-      "epoch": 0.851995276152811,
+      "epoch": 0.8803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.1270980288999511e-06,
-      "loss": 0.7929,
+      "learning_rate": 7.420652733173761e-07,
+      "loss": 0.7302,
       "step": 31022
     },
     {
-      "epoch": 0.8520227403806542,
+      "epoch": 0.8803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.1266878089326294e-06,
-      "loss": 0.8439,
+      "learning_rate": 7.41717873622333e-07,
+      "loss": 0.8015,
       "step": 31023
     },
     {
-      "epoch": 0.8520502046084975,
+      "epoch": 0.8803632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.1262776591743919e-06,
-      "loss": 0.7196,
+      "learning_rate": 7.413705521324155e-07,
+      "loss": 0.668,
       "step": 31024
     },
     {
-      "epoch": 0.8520776688363406,
+      "epoch": 0.8803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.1258675796284846e-06,
-      "loss": 0.7838,
+      "learning_rate": 7.410233088505592e-07,
+      "loss": 0.8189,
       "step": 31025
     },
     {
-      "epoch": 0.8521051330641839,
+      "epoch": 0.8804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.1254575702981496e-06,
-      "loss": 0.7388,
+      "learning_rate": 7.406761437796928e-07,
+      "loss": 0.7223,
       "step": 31026
     },
     {
-      "epoch": 0.8521325972920272,
+      "epoch": 0.8804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.125047631186631e-06,
-      "loss": 0.7319,
+      "learning_rate": 7.403290569227517e-07,
+      "loss": 0.7962,
       "step": 31027
     },
     {
-      "epoch": 0.8521600615198703,
+      "epoch": 0.8804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.1246377622971728e-06,
-      "loss": 0.8105,
+      "learning_rate": 7.399820482826692e-07,
+      "loss": 0.7355,
       "step": 31028
     },
     {
-      "epoch": 0.8521875257477136,
+      "epoch": 0.8805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.1242279636330179e-06,
-      "loss": 0.8061,
+      "learning_rate": 7.396351178623706e-07,
+      "loss": 0.8265,
       "step": 31029
     },
     {
-      "epoch": 0.8522149899755569,
+      "epoch": 0.8805334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 1.1238182351974103e-06,
-      "loss": 0.7871,
+      "learning_rate": 7.392882656647926e-07,
+      "loss": 0.8203,
       "step": 31030
     },
     {
-      "epoch": 0.8522424542034001,
+      "epoch": 0.8805618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.123408576993592e-06,
-      "loss": 0.8082,
+      "learning_rate": 7.38941491692865e-07,
+      "loss": 0.7924,
       "step": 31031
     },
     {
-      "epoch": 0.8522699184312433,
+      "epoch": 0.8805902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 1.1229989890248038e-06,
-      "loss": 0.6993,
+      "learning_rate": 7.385947959495122e-07,
+      "loss": 0.792,
       "step": 31032
     },
     {
-      "epoch": 0.8522973826590865,
+      "epoch": 0.8806186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.1225894712942854e-06,
-      "loss": 0.8177,
+      "learning_rate": 7.382481784376661e-07,
+      "loss": 0.7199,
       "step": 31033
     },
     {
-      "epoch": 0.8523248468869298,
+      "epoch": 0.8806469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 1.1221800238052794e-06,
-      "loss": 0.8672,
+      "learning_rate": 7.379016391602556e-07,
+      "loss": 0.7516,
       "step": 31034
     },
     {
-      "epoch": 0.852352311114773,
+      "epoch": 0.8806753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 1.1217706465610224e-06,
-      "loss": 0.7315,
+      "learning_rate": 7.375551781202051e-07,
+      "loss": 0.8252,
       "step": 31035
     },
     {
-      "epoch": 0.8523797753426162,
+      "epoch": 0.8807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.121361339564755e-06,
-      "loss": 0.8094,
+      "learning_rate": 7.372087953204421e-07,
+      "loss": 0.8059,
       "step": 31036
     },
     {
-      "epoch": 0.8524072395704595,
+      "epoch": 0.8807321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 1.1209521028197158e-06,
-      "loss": 0.7566,
+      "learning_rate": 7.368624907638944e-07,
+      "loss": 0.7315,
       "step": 31037
     },
     {
-      "epoch": 0.8524347037983027,
+      "epoch": 0.8807604994324632,
       "grad_norm": 0.0,
-      "learning_rate": 1.1205429363291443e-06,
-      "loss": 0.8815,
+      "learning_rate": 7.365162644534841e-07,
+      "loss": 0.8443,
       "step": 31038
     },
     {
-      "epoch": 0.8524621680261459,
+      "epoch": 0.880788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.1201338400962757e-06,
-      "loss": 0.8376,
+      "learning_rate": 7.361701163921364e-07,
+      "loss": 0.7333,
       "step": 31039
     },
     {
-      "epoch": 0.8524896322539892,
+      "epoch": 0.8808172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.1197248141243477e-06,
-      "loss": 0.7543,
+      "learning_rate": 7.358240465827793e-07,
+      "loss": 0.8475,
       "step": 31040
     },
     {
-      "epoch": 0.8525170964818324,
+      "epoch": 0.8808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.1193158584165987e-06,
-      "loss": 0.8082,
+      "learning_rate": 7.354780550283303e-07,
+      "loss": 0.8578,
       "step": 31041
     },
     {
-      "epoch": 0.8525445607096757,
+      "epoch": 0.8808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.1189069729762602e-06,
-      "loss": 0.8178,
+      "learning_rate": 7.351321417317148e-07,
+      "loss": 0.6969,
       "step": 31042
     },
     {
-      "epoch": 0.8525720249375189,
+      "epoch": 0.8809023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 1.1184981578065712e-06,
-      "loss": 0.8102,
+      "learning_rate": 7.34786306695856e-07,
+      "loss": 0.8696,
       "step": 31043
     },
     {
-      "epoch": 0.8525994891653621,
+      "epoch": 0.8809307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 1.1180894129107655e-06,
-      "loss": 0.7232,
+      "learning_rate": 7.344405499236706e-07,
+      "loss": 0.8208,
       "step": 31044
     },
     {
-      "epoch": 0.8526269533932054,
+      "epoch": 0.8809591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.1176807382920761e-06,
-      "loss": 0.7459,
+      "learning_rate": 7.340948714180863e-07,
+      "loss": 0.7923,
       "step": 31045
     },
     {
-      "epoch": 0.8526544176210485,
+      "epoch": 0.8809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.1172721339537374e-06,
-      "loss": 0.8248,
+      "learning_rate": 7.337492711820171e-07,
+      "loss": 0.7187,
       "step": 31046
     },
     {
-      "epoch": 0.8526818818488918,
+      "epoch": 0.8810158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.116863599898982e-06,
-      "loss": 0.7981,
+      "learning_rate": 7.334037492183843e-07,
+      "loss": 0.7175,
       "step": 31047
     },
     {
-      "epoch": 0.8527093460767351,
+      "epoch": 0.8810442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 1.1164551361310427e-06,
-      "loss": 0.8385,
+      "learning_rate": 7.330583055301089e-07,
+      "loss": 0.7462,
       "step": 31048
     },
     {
-      "epoch": 0.8527368103045783,
+      "epoch": 0.8810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.116046742653154e-06,
-      "loss": 0.9305,
+      "learning_rate": 7.327129401201039e-07,
+      "loss": 0.7952,
       "step": 31049
     },
     {
-      "epoch": 0.8527642745324215,
+      "epoch": 0.8811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.1156384194685444e-06,
-      "loss": 0.9388,
+      "learning_rate": 7.323676529912893e-07,
+      "loss": 0.7532,
       "step": 31050
     },
     {
-      "epoch": 0.8527917387602647,
+      "epoch": 0.8811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.1152301665804421e-06,
-      "loss": 0.7706,
+      "learning_rate": 7.32022444146584e-07,
+      "loss": 0.8689,
       "step": 31051
     },
     {
-      "epoch": 0.852819202988108,
+      "epoch": 0.8811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.1148219839920815e-06,
-      "loss": 0.7626,
+      "learning_rate": 7.316773135889e-07,
+      "loss": 0.7646,
       "step": 31052
     },
     {
-      "epoch": 0.8528466672159513,
+      "epoch": 0.8811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.114413871706691e-06,
-      "loss": 0.8074,
+      "learning_rate": 7.313322613211537e-07,
+      "loss": 0.8077,
       "step": 31053
     },
     {
-      "epoch": 0.8528741314437944,
+      "epoch": 0.8812145289443813,
       "grad_norm": 0.0,
-      "learning_rate": 1.1140058297275002e-06,
-      "loss": 0.8345,
+      "learning_rate": 7.309872873462632e-07,
+      "loss": 0.8213,
       "step": 31054
     },
     {
-      "epoch": 0.8529015956716377,
+      "epoch": 0.8812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.113597858057739e-06,
-      "loss": 0.8552,
+      "learning_rate": 7.306423916671379e-07,
+      "loss": 0.8443,
       "step": 31055
     },
     {
-      "epoch": 0.852929059899481,
+      "epoch": 0.8812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.1131899567006322e-06,
-      "loss": 0.8366,
+      "learning_rate": 7.302975742866925e-07,
+      "loss": 0.8707,
       "step": 31056
     },
     {
-      "epoch": 0.8529565241273241,
+      "epoch": 0.881299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.1127821256594084e-06,
-      "loss": 0.8111,
+      "learning_rate": 7.299528352078422e-07,
+      "loss": 0.7238,
       "step": 31057
     },
     {
-      "epoch": 0.8529839883551674,
+      "epoch": 0.881328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.1123743649372975e-06,
-      "loss": 0.7592,
+      "learning_rate": 7.296081744334948e-07,
+      "loss": 0.8653,
       "step": 31058
     },
     {
-      "epoch": 0.8530114525830106,
+      "epoch": 0.8813564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 1.1119666745375213e-06,
-      "loss": 0.8213,
+      "learning_rate": 7.292635919665647e-07,
+      "loss": 0.7617,
       "step": 31059
     },
     {
-      "epoch": 0.8530389168108539,
+      "epoch": 0.8813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.1115590544633071e-06,
-      "loss": 0.8183,
+      "learning_rate": 7.289190878099606e-07,
+      "loss": 0.7904,
       "step": 31060
     },
     {
-      "epoch": 0.8530663810386971,
+      "epoch": 0.8814131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 1.1111515047178834e-06,
-      "loss": 0.8269,
+      "learning_rate": 7.285746619665945e-07,
+      "loss": 0.8824,
       "step": 31061
     },
     {
-      "epoch": 0.8530938452665403,
+      "epoch": 0.8814415437003406,
       "grad_norm": 0.0,
-      "learning_rate": 1.11074402530447e-06,
-      "loss": 0.8107,
+      "learning_rate": 7.282303144393766e-07,
+      "loss": 0.7191,
       "step": 31062
     },
     {
-      "epoch": 0.8531213094943836,
+      "epoch": 0.8814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.110336616226293e-06,
-      "loss": 0.7918,
+      "learning_rate": 7.278860452312131e-07,
+      "loss": 0.9144,
       "step": 31063
     },
     {
-      "epoch": 0.8531487737222267,
+      "epoch": 0.8814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.1099292774865755e-06,
-      "loss": 0.7963,
+      "learning_rate": 7.275418543450119e-07,
+      "loss": 0.8724,
       "step": 31064
     },
     {
-      "epoch": 0.85317623795007,
+      "epoch": 0.8815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.109522009088544e-06,
-      "loss": 0.9036,
+      "learning_rate": 7.271977417836839e-07,
+      "loss": 0.7269,
       "step": 31065
     },
     {
-      "epoch": 0.8532037021779133,
+      "epoch": 0.8815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.1091148110354167e-06,
-      "loss": 0.8079,
+      "learning_rate": 7.268537075501314e-07,
+      "loss": 0.7668,
       "step": 31066
     },
     {
-      "epoch": 0.8532311664057565,
+      "epoch": 0.881583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.108707683330419e-06,
-      "loss": 0.88,
+      "learning_rate": 7.265097516472619e-07,
+      "loss": 0.8047,
       "step": 31067
     },
     {
-      "epoch": 0.8532586306335997,
+      "epoch": 0.8816118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 1.108300625976767e-06,
-      "loss": 0.8139,
+      "learning_rate": 7.261658740779842e-07,
+      "loss": 0.7778,
       "step": 31068
     },
     {
-      "epoch": 0.853286094861443,
+      "epoch": 0.8816401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.107893638977686e-06,
-      "loss": 0.7818,
+      "learning_rate": 7.258220748451971e-07,
+      "loss": 0.8724,
       "step": 31069
     },
     {
-      "epoch": 0.8533135590892862,
+      "epoch": 0.8816685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 1.107486722336395e-06,
-      "loss": 0.8427,
+      "learning_rate": 7.254783539518096e-07,
+      "loss": 0.7677,
       "step": 31070
     },
     {
-      "epoch": 0.8533410233171295,
+      "epoch": 0.8816969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 1.107079876056113e-06,
-      "loss": 0.8029,
+      "learning_rate": 7.251347114007234e-07,
+      "loss": 0.7191,
       "step": 31071
     },
     {
-      "epoch": 0.8533684875449726,
+      "epoch": 0.8817253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 1.1066731001400598e-06,
-      "loss": 0.8452,
+      "learning_rate": 7.24791147194841e-07,
+      "loss": 0.6689,
       "step": 31072
     },
     {
-      "epoch": 0.8533959517728159,
+      "epoch": 0.8817536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.1062663945914564e-06,
-      "loss": 0.8375,
+      "learning_rate": 7.244476613370644e-07,
+      "loss": 0.8194,
       "step": 31073
     },
     {
-      "epoch": 0.8534234160006592,
+      "epoch": 0.8817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.1058597594135178e-06,
-      "loss": 0.8168,
+      "learning_rate": 7.241042538302967e-07,
+      "loss": 0.7698,
       "step": 31074
     },
     {
-      "epoch": 0.8534508802285024,
+      "epoch": 0.8818104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.1054531946094615e-06,
-      "loss": 0.7636,
+      "learning_rate": 7.237609246774335e-07,
+      "loss": 0.709,
       "step": 31075
     },
     {
-      "epoch": 0.8534783444563456,
+      "epoch": 0.881838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.1050467001825038e-06,
-      "loss": 0.7824,
+      "learning_rate": 7.234176738813825e-07,
+      "loss": 0.7845,
       "step": 31076
     },
     {
-      "epoch": 0.8535058086841888,
+      "epoch": 0.8818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.1046402761358621e-06,
-      "loss": 0.8127,
+      "learning_rate": 7.23074501445038e-07,
+      "loss": 0.7818,
       "step": 31077
     },
     {
-      "epoch": 0.8535332729120321,
+      "epoch": 0.8818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.1042339224727528e-06,
-      "loss": 0.7666,
+      "learning_rate": 7.227314073712998e-07,
+      "loss": 0.7624,
       "step": 31078
     },
     {
-      "epoch": 0.8535607371398753,
+      "epoch": 0.8819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.1038276391963921e-06,
-      "loss": 0.8438,
+      "learning_rate": 7.223883916630691e-07,
+      "loss": 0.7339,
       "step": 31079
     },
     {
-      "epoch": 0.8535882013677185,
+      "epoch": 0.8819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.1034214263099907e-06,
-      "loss": 0.8316,
+      "learning_rate": 7.22045454323238e-07,
+      "loss": 1.0245,
       "step": 31080
     },
     {
-      "epoch": 0.8536156655955618,
+      "epoch": 0.8819807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.1030152838167663e-06,
-      "loss": 0.7154,
+      "learning_rate": 7.217025953547064e-07,
+      "loss": 0.8804,
       "step": 31081
     },
     {
-      "epoch": 0.8536431298234051,
+      "epoch": 0.8820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.1026092117199328e-06,
-      "loss": 0.8754,
+      "learning_rate": 7.213598147603717e-07,
+      "loss": 0.8136,
       "step": 31082
     },
     {
-      "epoch": 0.8536705940512482,
+      "epoch": 0.8820374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.1022032100226999e-06,
-      "loss": 0.8204,
+      "learning_rate": 7.210171125431264e-07,
+      "loss": 0.7509,
       "step": 31083
     },
     {
-      "epoch": 0.8536980582790915,
+      "epoch": 0.8820658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.1017972787282815e-06,
-      "loss": 0.8377,
+      "learning_rate": 7.206744887058658e-07,
+      "loss": 0.805,
       "step": 31084
     },
     {
-      "epoch": 0.8537255225069347,
+      "epoch": 0.882094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 1.101391417839892e-06,
-      "loss": 0.908,
+      "learning_rate": 7.203319432514877e-07,
+      "loss": 0.7339,
       "step": 31085
     },
     {
-      "epoch": 0.853752986734778,
+      "epoch": 0.882122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.1009856273607388e-06,
-      "loss": 0.7977,
+      "learning_rate": 7.199894761828807e-07,
+      "loss": 0.8708,
       "step": 31086
     },
     {
-      "epoch": 0.8537804509626212,
+      "epoch": 0.8821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.100579907294035e-06,
-      "loss": 0.8181,
+      "learning_rate": 7.196470875029393e-07,
+      "loss": 0.8929,
       "step": 31087
     },
     {
-      "epoch": 0.8538079151904644,
+      "epoch": 0.8821793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 1.1001742576429897e-06,
-      "loss": 0.8407,
+      "learning_rate": 7.19304777214559e-07,
+      "loss": 0.7891,
       "step": 31088
     },
     {
-      "epoch": 0.8538353794183077,
+      "epoch": 0.8822077185017027,
       "grad_norm": 0.0,
-      "learning_rate": 1.0997686784108153e-06,
-      "loss": 0.8625,
+      "learning_rate": 7.18962545320625e-07,
+      "loss": 0.7981,
       "step": 31089
     },
     {
-      "epoch": 0.8538628436461508,
+      "epoch": 0.8822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.0993631696007157e-06,
-      "loss": 0.8851,
+      "learning_rate": 7.186203918240331e-07,
+      "loss": 0.858,
       "step": 31090
     },
     {
-      "epoch": 0.8538903078739941,
+      "epoch": 0.8822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.098957731215906e-06,
-      "loss": 0.7705,
+      "learning_rate": 7.182783167276719e-07,
+      "loss": 0.8176,
       "step": 31091
     },
     {
-      "epoch": 0.8539177721018374,
+      "epoch": 0.8822928490351872,
       "grad_norm": 0.0,
-      "learning_rate": 1.0985523632595873e-06,
-      "loss": 0.7838,
+      "learning_rate": 7.179363200344303e-07,
+      "loss": 0.7969,
       "step": 31092
     },
     {
-      "epoch": 0.8539452363296806,
+      "epoch": 0.8823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.0981470657349714e-06,
-      "loss": 0.7794,
+      "learning_rate": 7.175944017471991e-07,
+      "loss": 0.7967,
       "step": 31093
     },
     {
-      "epoch": 0.8539727005575238,
+      "epoch": 0.8823496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 1.0977418386452642e-06,
-      "loss": 0.7854,
+      "learning_rate": 7.172525618688641e-07,
+      "loss": 0.8094,
       "step": 31094
     },
     {
-      "epoch": 0.8540001647853671,
+      "epoch": 0.8823779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 1.0973366819936715e-06,
-      "loss": 0.7071,
+      "learning_rate": 7.169108004023151e-07,
+      "loss": 0.8582,
       "step": 31095
     },
     {
-      "epoch": 0.8540276290132103,
+      "epoch": 0.8824063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.0969315957833992e-06,
-      "loss": 0.8428,
+      "learning_rate": 7.165691173504385e-07,
+      "loss": 0.8536,
       "step": 31096
     },
     {
-      "epoch": 0.8540550932410536,
+      "epoch": 0.8824347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 1.096526580017656e-06,
-      "loss": 0.86,
+      "learning_rate": 7.162275127161178e-07,
+      "loss": 0.744,
       "step": 31097
     },
     {
-      "epoch": 0.8540825574688967,
+      "epoch": 0.8824631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.0961216346996428e-06,
-      "loss": 0.785,
+      "learning_rate": 7.158859865022416e-07,
+      "loss": 0.9444,
       "step": 31098
     },
     {
-      "epoch": 0.85411002169674,
+      "epoch": 0.8824914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.0957167598325612e-06,
-      "loss": 0.8674,
+      "learning_rate": 7.155445387116944e-07,
+      "loss": 0.8637,
       "step": 31099
     },
     {
-      "epoch": 0.8541374859245833,
+      "epoch": 0.8825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.095311955419619e-06,
-      "loss": 0.7525,
+      "learning_rate": 7.152031693473593e-07,
+      "loss": 0.8291,
       "step": 31100
     },
     {
-      "epoch": 0.8541649501524264,
+      "epoch": 0.8825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.0949072214640177e-06,
-      "loss": 0.8524,
+      "learning_rate": 7.148618784121198e-07,
+      "loss": 0.8023,
       "step": 31101
     },
     {
-      "epoch": 0.8541924143802697,
+      "epoch": 0.8825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.0945025579689605e-06,
-      "loss": 0.6493,
+      "learning_rate": 7.145206659088611e-07,
+      "loss": 0.8394,
       "step": 31102
     },
     {
-      "epoch": 0.8542198786081129,
+      "epoch": 0.8826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.0940979649376505e-06,
-      "loss": 0.8914,
+      "learning_rate": 7.14179531840462e-07,
+      "loss": 0.7963,
       "step": 31103
     },
     {
-      "epoch": 0.8542473428359562,
+      "epoch": 0.882633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.0936934423732848e-06,
-      "loss": 0.7855,
+      "learning_rate": 7.138384762098061e-07,
+      "loss": 0.8651,
       "step": 31104
     },
     {
-      "epoch": 0.8542748070637994,
+      "epoch": 0.8826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.0932889902790654e-06,
-      "loss": 0.7676,
+      "learning_rate": 7.134974990197763e-07,
+      "loss": 0.6753,
       "step": 31105
     },
     {
-      "epoch": 0.8543022712916426,
+      "epoch": 0.8826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.092884608658198e-06,
-      "loss": 0.964,
+      "learning_rate": 7.13156600273246e-07,
+      "loss": 0.783,
       "step": 31106
     },
     {
-      "epoch": 0.8543297355194859,
+      "epoch": 0.8827185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.092480297513875e-06,
-      "loss": 0.8211,
+      "learning_rate": 7.128157799731039e-07,
+      "loss": 0.8205,
       "step": 31107
     },
     {
-      "epoch": 0.8543571997473292,
+      "epoch": 0.8827468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.0920760568492995e-06,
-      "loss": 0.7357,
+      "learning_rate": 7.124750381222223e-07,
+      "loss": 0.7793,
       "step": 31108
     },
     {
-      "epoch": 0.8543846639751723,
+      "epoch": 0.8827752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 1.0916718866676667e-06,
-      "loss": 0.8895,
+      "learning_rate": 7.12134374723481e-07,
+      "loss": 0.8544,
       "step": 31109
     },
     {
-      "epoch": 0.8544121282030156,
+      "epoch": 0.8828036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 1.0912677869721777e-06,
-      "loss": 0.9507,
+      "learning_rate": 7.1179378977976e-07,
+      "loss": 0.9313,
       "step": 31110
     },
     {
-      "epoch": 0.8544395924308588,
+      "epoch": 0.8828320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.0908637577660286e-06,
-      "loss": 0.8076,
+      "learning_rate": 7.114532832939325e-07,
+      "loss": 0.8069,
       "step": 31111
     },
     {
-      "epoch": 0.854467056658702,
+      "epoch": 0.8828603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 1.0904597990524169e-06,
-      "loss": 0.8339,
+      "learning_rate": 7.111128552688773e-07,
+      "loss": 0.8141,
       "step": 31112
     },
     {
-      "epoch": 0.8544945208865453,
+      "epoch": 0.8828887627695801,
       "grad_norm": 0.0,
-      "learning_rate": 1.09005591083454e-06,
-      "loss": 0.7954,
+      "learning_rate": 7.107725057074711e-07,
+      "loss": 0.9088,
       "step": 31113
     },
     {
-      "epoch": 0.8545219851143885,
+      "epoch": 0.8829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.0896520931155907e-06,
-      "loss": 0.855,
+      "learning_rate": 7.104322346125848e-07,
+      "loss": 0.8332,
       "step": 31114
     },
     {
-      "epoch": 0.8545494493422318,
+      "epoch": 0.8829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.0892483458987668e-06,
-      "loss": 0.8248,
+      "learning_rate": 7.10092041987096e-07,
+      "loss": 0.7971,
       "step": 31115
     },
     {
-      "epoch": 0.8545769135700749,
+      "epoch": 0.8829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.0888446691872601e-06,
-      "loss": 0.7169,
+      "learning_rate": 7.097519278338783e-07,
+      "loss": 0.7936,
       "step": 31116
     },
     {
-      "epoch": 0.8546043777979182,
+      "epoch": 0.8830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.0884410629842657e-06,
-      "loss": 0.8157,
+      "learning_rate": 7.094118921558024e-07,
+      "loss": 0.798,
       "step": 31117
     },
     {
-      "epoch": 0.8546318420257615,
+      "epoch": 0.8830306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 1.0880375272929778e-06,
-      "loss": 0.8017,
+      "learning_rate": 7.090719349557407e-07,
+      "loss": 0.7567,
       "step": 31118
     },
     {
-      "epoch": 0.8546593062536046,
+      "epoch": 0.8830590238365493,
       "grad_norm": 0.0,
-      "learning_rate": 1.0876340621165893e-06,
-      "loss": 0.7854,
+      "learning_rate": 7.087320562365685e-07,
+      "loss": 0.8133,
       "step": 31119
     },
     {
-      "epoch": 0.8546867704814479,
+      "epoch": 0.8830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.0872306674582945e-06,
-      "loss": 0.9086,
+      "learning_rate": 7.083922560011536e-07,
+      "loss": 0.678,
       "step": 31120
     },
     {
-      "epoch": 0.8547142347092912,
+      "epoch": 0.8831157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 1.0868273433212816e-06,
-      "loss": 0.8476,
+      "learning_rate": 7.080525342523636e-07,
+      "loss": 0.7449,
       "step": 31121
     },
     {
-      "epoch": 0.8547416989371344,
+      "epoch": 0.883144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.0864240897087452e-06,
-      "loss": 0.8587,
+      "learning_rate": 7.077128909930763e-07,
+      "loss": 0.7881,
       "step": 31122
     },
     {
-      "epoch": 0.8547691631649776,
+      "epoch": 0.883172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.0860209066238714e-06,
-      "loss": 0.9688,
+      "learning_rate": 7.07373326226154e-07,
+      "loss": 0.8005,
       "step": 31123
     },
     {
-      "epoch": 0.8547966273928208,
+      "epoch": 0.8832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.0856177940698542e-06,
-      "loss": 0.8688,
+      "learning_rate": 7.070338399544663e-07,
+      "loss": 0.8605,
       "step": 31124
     },
     {
-      "epoch": 0.8548240916206641,
+      "epoch": 0.8832292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.0852147520498813e-06,
-      "loss": 0.7795,
+      "learning_rate": 7.066944321808833e-07,
+      "loss": 0.7772,
       "step": 31125
     },
     {
-      "epoch": 0.8548515558485074,
+      "epoch": 0.8832576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 1.0848117805671444e-06,
-      "loss": 0.7264,
+      "learning_rate": 7.063551029082694e-07,
+      "loss": 0.7504,
       "step": 31126
     },
     {
-      "epoch": 0.8548790200763505,
+      "epoch": 0.8832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.0844088796248287e-06,
-      "loss": 0.6678,
+      "learning_rate": 7.0601585213949e-07,
+      "loss": 0.8177,
       "step": 31127
     },
     {
-      "epoch": 0.8549064843041938,
+      "epoch": 0.8833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.0840060492261228e-06,
-      "loss": 0.7805,
+      "learning_rate": 7.056766798774162e-07,
+      "loss": 0.8397,
       "step": 31128
     },
     {
-      "epoch": 0.854933948532037,
+      "epoch": 0.8833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.0836032893742154e-06,
-      "loss": 0.759,
+      "learning_rate": 7.053375861249056e-07,
+      "loss": 0.8645,
       "step": 31129
     },
     {
-      "epoch": 0.8549614127598802,
+      "epoch": 0.8833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.083200600072295e-06,
-      "loss": 0.7046,
+      "learning_rate": 7.049985708848295e-07,
+      "loss": 0.8645,
       "step": 31130
     },
     {
-      "epoch": 0.8549888769877235,
+      "epoch": 0.8833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.0827979813235433e-06,
-      "loss": 0.9317,
+      "learning_rate": 7.046596341600465e-07,
+      "loss": 0.7985,
       "step": 31131
     },
     {
-      "epoch": 0.8550163412155667,
+      "epoch": 0.883427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.0823954331311493e-06,
-      "loss": 0.8104,
+      "learning_rate": 7.04320775953421e-07,
+      "loss": 0.7297,
       "step": 31132
     },
     {
-      "epoch": 0.85504380544341,
+      "epoch": 0.8834562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.0819929554982954e-06,
-      "loss": 0.817,
+      "learning_rate": 7.039819962678174e-07,
+      "loss": 0.8592,
       "step": 31133
     },
     {
-      "epoch": 0.8550712696712532,
+      "epoch": 0.8834846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.0815905484281686e-06,
-      "loss": 0.7875,
+      "learning_rate": 7.036432951060934e-07,
+      "loss": 0.7508,
       "step": 31134
     },
     {
-      "epoch": 0.8550987338990964,
+      "epoch": 0.8835130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.0811882119239514e-06,
-      "loss": 0.8535,
+      "learning_rate": 7.033046724711135e-07,
+      "loss": 0.7764,
       "step": 31135
     },
     {
-      "epoch": 0.8551261981269397,
+      "epoch": 0.8835414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 1.0807859459888282e-06,
-      "loss": 0.8518,
+      "learning_rate": 7.029661283657385e-07,
+      "loss": 0.7729,
       "step": 31136
     },
     {
-      "epoch": 0.8551536623547829,
+      "epoch": 0.8835698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 1.080383750625983e-06,
-      "loss": 0.7627,
+      "learning_rate": 7.02627662792823e-07,
+      "loss": 0.7891,
       "step": 31137
     },
     {
-      "epoch": 0.8551811265826261,
+      "epoch": 0.8835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.0799816258385941e-06,
-      "loss": 0.8022,
+      "learning_rate": 7.022892757552313e-07,
+      "loss": 0.7621,
       "step": 31138
     },
     {
-      "epoch": 0.8552085908104694,
+      "epoch": 0.8836265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 1.0795795716298496e-06,
-      "loss": 0.8738,
+      "learning_rate": 7.019509672558211e-07,
+      "loss": 0.8375,
       "step": 31139
     },
     {
-      "epoch": 0.8552360550383126,
+      "epoch": 0.8836549375709422,
       "grad_norm": 0.0,
-      "learning_rate": 1.079177588002923e-06,
-      "loss": 0.7953,
+      "learning_rate": 7.016127372974479e-07,
+      "loss": 0.8572,
       "step": 31140
     },
     {
-      "epoch": 0.8552635192661558,
+      "epoch": 0.883683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.0787756749610002e-06,
-      "loss": 0.8228,
+      "learning_rate": 7.012745858829706e-07,
+      "loss": 0.7789,
       "step": 31141
     },
     {
-      "epoch": 0.855290983493999,
+      "epoch": 0.8837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.0783738325072602e-06,
-      "loss": 0.8066,
+      "learning_rate": 7.009365130152457e-07,
+      "loss": 0.8293,
       "step": 31142
     },
     {
-      "epoch": 0.8553184477218423,
+      "epoch": 0.8837400681044267,
       "grad_norm": 0.0,
-      "learning_rate": 1.0779720606448817e-06,
-      "loss": 0.7644,
+      "learning_rate": 7.005985186971265e-07,
+      "loss": 0.8628,
       "step": 31143
     },
     {
-      "epoch": 0.8553459119496856,
+      "epoch": 0.8837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.077570359377047e-06,
-      "loss": 0.8093,
+      "learning_rate": 7.002606029314695e-07,
+      "loss": 0.7702,
       "step": 31144
     },
     {
-      "epoch": 0.8553733761775287,
+      "epoch": 0.8837968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 1.0771687287069287e-06,
-      "loss": 0.7623,
+      "learning_rate": 6.999227657211316e-07,
+      "loss": 0.8194,
       "step": 31145
     },
     {
-      "epoch": 0.855400840405372,
+      "epoch": 0.8838251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 1.0767671686377113e-06,
-      "loss": 0.8042,
+      "learning_rate": 6.995850070689614e-07,
+      "loss": 0.7699,
       "step": 31146
     },
     {
-      "epoch": 0.8554283046332153,
+      "epoch": 0.8838535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.0763656791725653e-06,
-      "loss": 0.7669,
+      "learning_rate": 6.992473269778155e-07,
+      "loss": 0.7476,
       "step": 31147
     },
     {
-      "epoch": 0.8554557688610585,
+      "epoch": 0.8838819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 1.0759642603146714e-06,
-      "loss": 0.8257,
+      "learning_rate": 6.989097254505472e-07,
+      "loss": 0.8148,
       "step": 31148
     },
     {
-      "epoch": 0.8554832330889017,
+      "epoch": 0.8839103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.0755629120672051e-06,
-      "loss": 0.8376,
+      "learning_rate": 6.985722024900044e-07,
+      "loss": 0.86,
       "step": 31149
     },
     {
-      "epoch": 0.8555106973167449,
+      "epoch": 0.8839387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 1.0751616344333438e-06,
-      "loss": 0.8893,
+      "learning_rate": 6.982347580990412e-07,
+      "loss": 0.7768,
       "step": 31150
     },
     {
-      "epoch": 0.8555381615445882,
+      "epoch": 0.8839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.0747604274162582e-06,
-      "loss": 0.8444,
+      "learning_rate": 6.978973922805077e-07,
+      "loss": 0.8432,
       "step": 31151
     },
     {
-      "epoch": 0.8555656257724314,
+      "epoch": 0.8839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.0743592910191259e-06,
-      "loss": 0.7918,
+      "learning_rate": 6.975601050372505e-07,
+      "loss": 0.7487,
       "step": 31152
     },
     {
-      "epoch": 0.8555930900002746,
+      "epoch": 0.8840238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 1.0739582252451207e-06,
-      "loss": 0.7754,
+      "learning_rate": 6.97222896372125e-07,
+      "loss": 0.9062,
       "step": 31153
     },
     {
-      "epoch": 0.8556205542281179,
+      "epoch": 0.8840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.0735572300974173e-06,
-      "loss": 0.7777,
+      "learning_rate": 6.968857662879735e-07,
+      "loss": 0.8359,
       "step": 31154
     },
     {
-      "epoch": 0.8556480184559611,
+      "epoch": 0.8840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.073156305579185e-06,
-      "loss": 0.7367,
+      "learning_rate": 6.96548714787646e-07,
+      "loss": 0.7855,
       "step": 31155
     },
     {
-      "epoch": 0.8556754826838043,
+      "epoch": 0.8841089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 1.0727554516936001e-06,
-      "loss": 0.7701,
+      "learning_rate": 6.962117418739922e-07,
+      "loss": 0.8211,
       "step": 31156
     },
     {
-      "epoch": 0.8557029469116476,
+      "epoch": 0.8841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.0723546684438302e-06,
-      "loss": 0.8242,
+      "learning_rate": 6.958748475498533e-07,
+      "loss": 0.7647,
       "step": 31157
     },
     {
-      "epoch": 0.8557304111394908,
+      "epoch": 0.8841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.0719539558330482e-06,
-      "loss": 0.8021,
+      "learning_rate": 6.955380318180793e-07,
+      "loss": 0.8026,
       "step": 31158
     },
     {
-      "epoch": 0.855757875367334,
+      "epoch": 0.884194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.0715533138644253e-06,
-      "loss": 0.8004,
+      "learning_rate": 6.952012946815145e-07,
+      "loss": 0.8451,
       "step": 31159
     },
     {
-      "epoch": 0.8557853395951773,
+      "epoch": 0.8842224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 1.0711527425411305e-06,
-      "loss": 0.845,
+      "learning_rate": 6.948646361430011e-07,
+      "loss": 0.8071,
       "step": 31160
     },
     {
-      "epoch": 0.8558128038230205,
+      "epoch": 0.8842508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.0707522418663352e-06,
-      "loss": 0.8352,
+      "learning_rate": 6.945280562053836e-07,
+      "loss": 0.8539,
       "step": 31161
     },
     {
-      "epoch": 0.8558402680508638,
+      "epoch": 0.8842792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.0703518118432088e-06,
-      "loss": 0.724,
+      "learning_rate": 6.941915548715083e-07,
+      "loss": 0.8663,
       "step": 31162
     },
     {
-      "epoch": 0.8558677322787069,
+      "epoch": 0.8843076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 1.0699514524749144e-06,
-      "loss": 0.7924,
+      "learning_rate": 6.938551321442144e-07,
+      "loss": 0.7112,
       "step": 31163
     },
     {
-      "epoch": 0.8558951965065502,
+      "epoch": 0.8843359818388196,
       "grad_norm": 0.0,
-      "learning_rate": 1.069551163764624e-06,
-      "loss": 0.7543,
+      "learning_rate": 6.935187880263439e-07,
+      "loss": 0.6713,
       "step": 31164
     },
     {
-      "epoch": 0.8559226607343935,
+      "epoch": 0.8843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.0691509457155047e-06,
-      "loss": 0.6907,
+      "learning_rate": 6.9318252252074e-07,
+      "loss": 0.8129,
       "step": 31165
     },
     {
-      "epoch": 0.8559501249622367,
+      "epoch": 0.8843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.0687507983307211e-06,
-      "loss": 0.8263,
+      "learning_rate": 6.928463356302395e-07,
+      "loss": 0.8299,
       "step": 31166
     },
     {
-      "epoch": 0.8559775891900799,
+      "epoch": 0.8844211123723043,
       "grad_norm": 0.0,
-      "learning_rate": 1.0683507216134415e-06,
-      "loss": 0.7257,
+      "learning_rate": 6.925102273576856e-07,
+      "loss": 0.8173,
       "step": 31167
     },
     {
-      "epoch": 0.8560050534179231,
+      "epoch": 0.8844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.0679507155668334e-06,
-      "loss": 0.8035,
+      "learning_rate": 6.92174197705916e-07,
+      "loss": 0.7769,
       "step": 31168
     },
     {
-      "epoch": 0.8560325176457664,
+      "epoch": 0.884477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.0675507801940566e-06,
-      "loss": 0.8297,
+      "learning_rate": 6.918382466777685e-07,
+      "loss": 0.7807,
       "step": 31169
     },
     {
-      "epoch": 0.8560599818736097,
+      "epoch": 0.8845062429057888,
       "grad_norm": 0.0,
-      "learning_rate": 1.0671509154982807e-06,
-      "loss": 0.8278,
+      "learning_rate": 6.91502374276084e-07,
+      "loss": 0.747,
       "step": 31170
     },
     {
-      "epoch": 0.8560874461014528,
+      "epoch": 0.8845346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.0667511214826643e-06,
-      "loss": 0.788,
+      "learning_rate": 6.91166580503696e-07,
+      "loss": 0.8164,
       "step": 31171
     },
     {
-      "epoch": 0.8561149103292961,
+      "epoch": 0.8845629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 1.0663513981503737e-06,
-      "loss": 0.8278,
+      "learning_rate": 6.908308653634421e-07,
+      "loss": 0.7521,
       "step": 31172
     },
     {
-      "epoch": 0.8561423745571394,
+      "epoch": 0.8845913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.065951745504573e-06,
-      "loss": 0.7474,
+      "learning_rate": 6.90495228858159e-07,
+      "loss": 0.7772,
       "step": 31173
     },
     {
-      "epoch": 0.8561698387849825,
+      "epoch": 0.8846197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.0655521635484212e-06,
-      "loss": 0.7024,
+      "learning_rate": 6.901596709906811e-07,
+      "loss": 0.6995,
       "step": 31174
     },
     {
-      "epoch": 0.8561973030128258,
+      "epoch": 0.8846481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 1.0651526522850818e-06,
-      "loss": 0.782,
+      "learning_rate": 6.898241917638426e-07,
+      "loss": 0.8044,
       "step": 31175
     },
     {
-      "epoch": 0.856224767240669,
+      "epoch": 0.8846765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.0647532117177151e-06,
-      "loss": 0.72,
+      "learning_rate": 6.894887911804792e-07,
+      "loss": 0.8265,
       "step": 31176
     },
     {
-      "epoch": 0.8562522314685123,
+      "epoch": 0.8847048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 1.0643538418494824e-06,
-      "loss": 0.8699,
+      "learning_rate": 6.891534692434199e-07,
+      "loss": 0.9088,
       "step": 31177
     },
     {
-      "epoch": 0.8562796956963555,
+      "epoch": 0.884733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.0639545426835462e-06,
-      "loss": 0.6311,
+      "learning_rate": 6.888182259555009e-07,
+      "loss": 0.8211,
       "step": 31178
     },
     {
-      "epoch": 0.8563071599241987,
+      "epoch": 0.8847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.0635553142230625e-06,
-      "loss": 0.7939,
+      "learning_rate": 6.884830613195547e-07,
+      "loss": 0.7517,
       "step": 31179
     },
     {
-      "epoch": 0.856334624152042,
+      "epoch": 0.8847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.063156156471188e-06,
-      "loss": 0.7559,
+      "learning_rate": 6.88147975338408e-07,
+      "loss": 0.778,
       "step": 31180
     },
     {
-      "epoch": 0.8563620883798851,
+      "epoch": 0.8848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.0627570694310852e-06,
-      "loss": 0.7682,
+      "learning_rate": 6.878129680148948e-07,
+      "loss": 0.7894,
       "step": 31181
     },
     {
-      "epoch": 0.8563895526077284,
+      "epoch": 0.8848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.0623580531059096e-06,
-      "loss": 0.7581,
+      "learning_rate": 6.874780393518455e-07,
+      "loss": 0.7998,
       "step": 31182
     },
     {
-      "epoch": 0.8564170168355717,
+      "epoch": 0.8848751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.061959107498819e-06,
-      "loss": 0.8086,
+      "learning_rate": 6.871431893520853e-07,
+      "loss": 0.7863,
       "step": 31183
     },
     {
-      "epoch": 0.8564444810634149,
+      "epoch": 0.8849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.0615602326129703e-06,
-      "loss": 0.7968,
+      "learning_rate": 6.868084180184475e-07,
+      "loss": 0.8216,
       "step": 31184
     },
     {
-      "epoch": 0.8564719452912581,
+      "epoch": 0.8849318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.0611614284515225e-06,
-      "loss": 0.7727,
+      "learning_rate": 6.864737253537557e-07,
+      "loss": 0.7623,
       "step": 31185
     },
     {
-      "epoch": 0.8564994095191014,
+      "epoch": 0.8849602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.0607626950176265e-06,
-      "loss": 0.9147,
+      "learning_rate": 6.861391113608395e-07,
+      "loss": 0.7567,
       "step": 31186
     },
     {
-      "epoch": 0.8565268737469446,
+      "epoch": 0.884988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.0603640323144381e-06,
-      "loss": 0.8444,
+      "learning_rate": 6.858045760425269e-07,
+      "loss": 0.7567,
       "step": 31187
     },
     {
-      "epoch": 0.8565543379747879,
+      "epoch": 0.885017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.0599654403451121e-06,
-      "loss": 0.8126,
+      "learning_rate": 6.8547011940164e-07,
+      "loss": 0.7668,
       "step": 31188
     },
     {
-      "epoch": 0.856581802202631,
+      "epoch": 0.8850454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.0595669191128023e-06,
-      "loss": 0.8136,
+      "learning_rate": 6.851357414410053e-07,
+      "loss": 0.8084,
       "step": 31189
     },
     {
-      "epoch": 0.8566092664304743,
+      "epoch": 0.8850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.0591684686206616e-06,
-      "loss": 0.8347,
+      "learning_rate": 6.848014421634497e-07,
+      "loss": 0.8303,
       "step": 31190
     },
     {
-      "epoch": 0.8566367306583176,
+      "epoch": 0.8851021566401817,
       "grad_norm": 0.0,
-      "learning_rate": 1.0587700888718466e-06,
-      "loss": 0.7818,
+      "learning_rate": 6.844672215717929e-07,
+      "loss": 0.8438,
       "step": 31191
     },
     {
-      "epoch": 0.8566641948861607,
+      "epoch": 0.8851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.0583717798695036e-06,
-      "loss": 0.7815,
+      "learning_rate": 6.841330796688606e-07,
+      "loss": 0.7841,
       "step": 31192
     },
     {
-      "epoch": 0.856691659114004,
+      "epoch": 0.8851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.0579735416167858e-06,
-      "loss": 0.8041,
+      "learning_rate": 6.83799016457477e-07,
+      "loss": 0.7356,
       "step": 31193
     },
     {
-      "epoch": 0.8567191233418472,
+      "epoch": 0.8851872871736662,
       "grad_norm": 0.0,
-      "learning_rate": 1.0575753741168483e-06,
-      "loss": 0.8792,
+      "learning_rate": 6.834650319404601e-07,
+      "loss": 0.8238,
       "step": 31194
     },
     {
-      "epoch": 0.8567465875696905,
+      "epoch": 0.8852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.0571772773728362e-06,
-      "loss": 0.8843,
+      "learning_rate": 6.831311261206331e-07,
+      "loss": 0.7145,
       "step": 31195
     },
     {
-      "epoch": 0.8567740517975337,
+      "epoch": 0.8852440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 1.0567792513879017e-06,
-      "loss": 0.7959,
+      "learning_rate": 6.82797299000817e-07,
+      "loss": 0.7513,
       "step": 31196
     },
     {
-      "epoch": 0.8568015160253769,
+      "epoch": 0.8852724177071509,
       "grad_norm": 0.0,
-      "learning_rate": 1.0563812961651954e-06,
-      "loss": 0.8231,
+      "learning_rate": 6.824635505838295e-07,
+      "loss": 0.7475,
       "step": 31197
     },
     {
-      "epoch": 0.8568289802532202,
+      "epoch": 0.8853007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.0559834117078637e-06,
-      "loss": 0.7901,
+      "learning_rate": 6.821298808724919e-07,
+      "loss": 0.8548,
       "step": 31198
     },
     {
-      "epoch": 0.8568564444810635,
+      "epoch": 0.8853291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 1.0555855980190554e-06,
-      "loss": 0.7794,
+      "learning_rate": 6.817962898696218e-07,
+      "loss": 0.8,
       "step": 31199
     },
     {
-      "epoch": 0.8568839087089066,
+      "epoch": 0.8853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.055187855101918e-06,
-      "loss": 0.7976,
+      "learning_rate": 6.814627775780381e-07,
+      "loss": 0.8271,
       "step": 31200
     },
     {
-      "epoch": 0.8569113729367499,
+      "epoch": 0.8853859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.054790182959602e-06,
-      "loss": 0.8024,
+      "learning_rate": 6.811293440005573e-07,
+      "loss": 0.8618,
       "step": 31201
     },
     {
-      "epoch": 0.8569388371645931,
+      "epoch": 0.8854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.0543925815952483e-06,
-      "loss": 0.803,
+      "learning_rate": 6.807959891399951e-07,
+      "loss": 0.7517,
       "step": 31202
     },
     {
-      "epoch": 0.8569663013924363,
+      "epoch": 0.8854426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.0539950510120079e-06,
-      "loss": 0.7791,
+      "learning_rate": 6.804627129991681e-07,
+      "loss": 0.7823,
       "step": 31203
     },
     {
-      "epoch": 0.8569937656202796,
+      "epoch": 0.8854710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 1.0535975912130224e-06,
-      "loss": 0.8773,
+      "learning_rate": 6.80129515580893e-07,
+      "loss": 0.7958,
       "step": 31204
     },
     {
-      "epoch": 0.8570212298481228,
+      "epoch": 0.8854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.0532002022014386e-06,
-      "loss": 0.7319,
+      "learning_rate": 6.797963968879806e-07,
+      "loss": 0.7866,
       "step": 31205
     },
     {
-      "epoch": 0.8570486940759661,
+      "epoch": 0.885527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.0528028839803994e-06,
-      "loss": 0.7762,
+      "learning_rate": 6.794633569232479e-07,
+      "loss": 0.817,
       "step": 31206
     },
     {
-      "epoch": 0.8570761583038093,
+      "epoch": 0.8855561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 1.05240563655305e-06,
-      "loss": 0.7939,
+      "learning_rate": 6.79130395689509e-07,
+      "loss": 0.8737,
       "step": 31207
     },
     {
-      "epoch": 0.8571036225316525,
+      "epoch": 0.8855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.0520084599225334e-06,
-      "loss": 0.8144,
+      "learning_rate": 6.787975131895718e-07,
+      "loss": 0.8572,
       "step": 31208
     },
     {
-      "epoch": 0.8571310867594958,
+      "epoch": 0.8856129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.0516113540919947e-06,
-      "loss": 0.8394,
+      "learning_rate": 6.78464709426252e-07,
+      "loss": 0.7802,
       "step": 31209
     },
     {
-      "epoch": 0.857158550987339,
+      "epoch": 0.8856413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.0512143190645718e-06,
-      "loss": 0.8199,
+      "learning_rate": 6.781319844023615e-07,
+      "loss": 0.8793,
       "step": 31210
     },
     {
-      "epoch": 0.8571860152151822,
+      "epoch": 0.8856696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 1.0508173548434064e-06,
-      "loss": 0.722,
+      "learning_rate": 6.777993381207071e-07,
+      "loss": 0.9071,
       "step": 31211
     },
     {
-      "epoch": 0.8572134794430255,
+      "epoch": 0.8856980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.0504204614316405e-06,
-      "loss": 0.7868,
+      "learning_rate": 6.774667705841009e-07,
+      "loss": 0.7639,
       "step": 31212
     },
     {
-      "epoch": 0.8572409436708687,
+      "epoch": 0.8857264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.0500236388324147e-06,
-      "loss": 0.7575,
+      "learning_rate": 6.77134281795353e-07,
+      "loss": 0.9312,
       "step": 31213
     },
     {
-      "epoch": 0.857268407898712,
+      "epoch": 0.8857548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 1.049626887048869e-06,
-      "loss": 0.7598,
+      "learning_rate": 6.768018717572699e-07,
+      "loss": 0.7794,
       "step": 31214
     },
     {
-      "epoch": 0.8572958721265551,
+      "epoch": 0.8857832009080591,
       "grad_norm": 0.0,
-      "learning_rate": 1.0492302060841442e-06,
-      "loss": 0.8707,
+      "learning_rate": 6.764695404726618e-07,
+      "loss": 0.8252,
       "step": 31215
     },
     {
-      "epoch": 0.8573233363543984,
+      "epoch": 0.8858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.0488335959413753e-06,
-      "loss": 0.7987,
+      "learning_rate": 6.761372879443329e-07,
+      "loss": 0.8509,
       "step": 31216
     },
     {
-      "epoch": 0.8573508005822417,
+      "epoch": 0.8858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.0484370566237023e-06,
-      "loss": 0.7751,
+      "learning_rate": 6.758051141750921e-07,
+      "loss": 0.7545,
       "step": 31217
     },
     {
-      "epoch": 0.8573782648100848,
+      "epoch": 0.8858683314415438,
       "grad_norm": 0.0,
-      "learning_rate": 1.0480405881342647e-06,
-      "loss": 0.8594,
+      "learning_rate": 6.754730191677461e-07,
+      "loss": 0.7845,
       "step": 31218
     },
     {
-      "epoch": 0.8574057290379281,
+      "epoch": 0.8858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.0476441904761957e-06,
-      "loss": 0.8835,
+      "learning_rate": 6.751410029250971e-07,
+      "loss": 0.8841,
       "step": 31219
     },
     {
-      "epoch": 0.8574331932657714,
+      "epoch": 0.8859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.047247863652634e-06,
-      "loss": 0.6743,
+      "learning_rate": 6.748090654499518e-07,
+      "loss": 0.6691,
       "step": 31220
     },
     {
-      "epoch": 0.8574606574936146,
+      "epoch": 0.8859534619750283,
       "grad_norm": 0.0,
-      "learning_rate": 1.0468516076667168e-06,
-      "loss": 0.7802,
+      "learning_rate": 6.744772067451144e-07,
+      "loss": 0.8465,
       "step": 31221
     },
     {
-      "epoch": 0.8574881217214578,
+      "epoch": 0.8859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.046455422521575e-06,
-      "loss": 0.8391,
+      "learning_rate": 6.741454268133852e-07,
+      "loss": 0.7901,
       "step": 31222
     },
     {
-      "epoch": 0.857515585949301,
+      "epoch": 0.8860102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 1.0460593082203474e-06,
-      "loss": 0.7704,
+      "learning_rate": 6.738137256575694e-07,
+      "loss": 0.8088,
       "step": 31223
     },
     {
-      "epoch": 0.8575430501771443,
+      "epoch": 0.886038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.0456632647661669e-06,
-      "loss": 0.8053,
+      "learning_rate": 6.734821032804706e-07,
+      "loss": 0.7803,
       "step": 31224
     },
     {
-      "epoch": 0.8575705144049875,
+      "epoch": 0.886066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.0452672921621688e-06,
-      "loss": 0.8113,
+      "learning_rate": 6.731505596848853e-07,
+      "loss": 0.8196,
       "step": 31225
     },
     {
-      "epoch": 0.8575979786328307,
+      "epoch": 0.8860953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 1.0448713904114826e-06,
-      "loss": 0.8012,
+      "learning_rate": 6.728190948736158e-07,
+      "loss": 0.8001,
       "step": 31226
     },
     {
-      "epoch": 0.857625442860674,
+      "epoch": 0.8861237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.0444755595172452e-06,
-      "loss": 0.7915,
+      "learning_rate": 6.724877088494652e-07,
+      "loss": 0.7843,
       "step": 31227
     },
     {
-      "epoch": 0.8576529070885172,
+      "epoch": 0.8861520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 1.0440797994825834e-06,
-      "loss": 0.8188,
+      "learning_rate": 6.721564016152293e-07,
+      "loss": 0.8553,
       "step": 31228
     },
     {
-      "epoch": 0.8576803713163604,
+      "epoch": 0.8861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.0436841103106322e-06,
-      "loss": 0.7582,
+      "learning_rate": 6.718251731737047e-07,
+      "loss": 0.8898,
       "step": 31229
     },
     {
-      "epoch": 0.8577078355442037,
+      "epoch": 0.8862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.0432884920045217e-06,
-      "loss": 0.7994,
+      "learning_rate": 6.714940235276957e-07,
+      "loss": 0.7572,
       "step": 31230
     },
     {
-      "epoch": 0.8577352997720469,
+      "epoch": 0.8862372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.0428929445673819e-06,
-      "loss": 0.7923,
+      "learning_rate": 6.711629526799946e-07,
+      "loss": 0.9116,
       "step": 31231
     },
     {
-      "epoch": 0.8577627639998902,
+      "epoch": 0.8862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.0424974680023447e-06,
-      "loss": 0.789,
+      "learning_rate": 6.708319606334002e-07,
+      "loss": 0.797,
       "step": 31232
     },
     {
-      "epoch": 0.8577902282277334,
+      "epoch": 0.8862939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.0421020623125355e-06,
-      "loss": 0.9395,
+      "learning_rate": 6.705010473907081e-07,
+      "loss": 0.7328,
       "step": 31233
     },
     {
-      "epoch": 0.8578176924555766,
+      "epoch": 0.886322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.0417067275010861e-06,
-      "loss": 0.7682,
+      "learning_rate": 6.701702129547116e-07,
+      "loss": 0.7857,
       "step": 31234
     },
     {
-      "epoch": 0.8578451566834199,
+      "epoch": 0.8863507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.041311463571122e-06,
-      "loss": 0.7747,
+      "learning_rate": 6.698394573282063e-07,
+      "loss": 0.8111,
       "step": 31235
     },
     {
-      "epoch": 0.857872620911263,
+      "epoch": 0.8863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.0409162705257714e-06,
-      "loss": 0.7601,
+      "learning_rate": 6.695087805139888e-07,
+      "loss": 0.826,
       "step": 31236
     },
     {
-      "epoch": 0.8579000851391063,
+      "epoch": 0.8864074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.0405211483681621e-06,
-      "loss": 0.7434,
+      "learning_rate": 6.691781825148491e-07,
+      "loss": 0.78,
       "step": 31237
     },
     {
-      "epoch": 0.8579275493669496,
+      "epoch": 0.8864358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 1.0401260971014215e-06,
-      "loss": 0.8012,
+      "learning_rate": 6.688476633335816e-07,
+      "loss": 0.826,
       "step": 31238
     },
     {
-      "epoch": 0.8579550135947928,
+      "epoch": 0.8864642451759365,
       "grad_norm": 0.0,
-      "learning_rate": 1.0397311167286728e-06,
-      "loss": 0.8468,
+      "learning_rate": 6.685172229729752e-07,
+      "loss": 0.8328,
       "step": 31239
     },
     {
-      "epoch": 0.857982477822636,
+      "epoch": 0.8864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.039336207253041e-06,
-      "loss": 0.7069,
+      "learning_rate": 6.681868614358245e-07,
+      "loss": 0.8297,
       "step": 31240
     },
     {
-      "epoch": 0.8580099420504792,
+      "epoch": 0.8865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.038941368677654e-06,
-      "loss": 0.8142,
+      "learning_rate": 6.678565787249192e-07,
+      "loss": 0.711,
       "step": 31241
     },
     {
-      "epoch": 0.8580374062783225,
+      "epoch": 0.8865493757094212,
       "grad_norm": 0.0,
-      "learning_rate": 1.0385466010056345e-06,
-      "loss": 0.7476,
+      "learning_rate": 6.675263748430483e-07,
+      "loss": 0.8552,
       "step": 31242
     },
     {
-      "epoch": 0.8580648705061658,
+      "epoch": 0.886577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.0381519042401045e-06,
-      "loss": 0.7733,
+      "learning_rate": 6.671962497930018e-07,
+      "loss": 0.7785,
       "step": 31243
     },
     {
-      "epoch": 0.8580923347340089,
+      "epoch": 0.8866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.0377572783841894e-06,
-      "loss": 0.7361,
+      "learning_rate": 6.668662035775674e-07,
+      "loss": 0.8518,
       "step": 31244
     },
     {
-      "epoch": 0.8581197989618522,
+      "epoch": 0.8866345062429057,
       "grad_norm": 0.0,
-      "learning_rate": 1.0373627234410088e-06,
-      "loss": 0.7685,
+      "learning_rate": 6.665362361995332e-07,
+      "loss": 0.7541,
       "step": 31245
     },
     {
-      "epoch": 0.8581472631896955,
+      "epoch": 0.8866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.0369682394136859e-06,
-      "loss": 0.8486,
+      "learning_rate": 6.662063476616887e-07,
+      "loss": 0.8577,
       "step": 31246
     },
     {
-      "epoch": 0.8581747274175386,
+      "epoch": 0.8866912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 1.0365738263053425e-06,
-      "loss": 0.7497,
+      "learning_rate": 6.658765379668186e-07,
+      "loss": 0.7217,
       "step": 31247
     },
     {
-      "epoch": 0.8582021916453819,
+      "epoch": 0.8867196367763904,
       "grad_norm": 0.0,
-      "learning_rate": 1.0361794841190996e-06,
-      "loss": 0.7888,
+      "learning_rate": 6.655468071177084e-07,
+      "loss": 0.8113,
       "step": 31248
     },
     {
-      "epoch": 0.8582296558732251,
+      "epoch": 0.8867480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.035785212858078e-06,
-      "loss": 0.7613,
+      "learning_rate": 6.652171551171438e-07,
+      "loss": 0.7714,
       "step": 31249
     },
     {
-      "epoch": 0.8582571201010684,
+      "epoch": 0.8867763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 1.035391012525394e-06,
-      "loss": 0.7206,
+      "learning_rate": 6.648875819679113e-07,
+      "loss": 0.7528,
       "step": 31250
     },
     {
-      "epoch": 0.8582845843289116,
+      "epoch": 0.8868047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.034996883124172e-06,
-      "loss": 0.8394,
+      "learning_rate": 6.645580876727908e-07,
+      "loss": 0.6508,
       "step": 31251
     },
     {
-      "epoch": 0.8583120485567548,
+      "epoch": 0.88683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.0346028246575247e-06,
-      "loss": 0.7579,
+      "learning_rate": 6.642286722345682e-07,
+      "loss": 0.8147,
       "step": 31252
     },
     {
-      "epoch": 0.8583395127845981,
+      "epoch": 0.886861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.034208837128573e-06,
-      "loss": 0.6824,
+      "learning_rate": 6.638993356560275e-07,
+      "loss": 0.7656,
       "step": 31253
     },
     {
-      "epoch": 0.8583669770124412,
+      "epoch": 0.8868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.0338149205404336e-06,
-      "loss": 0.8358,
+      "learning_rate": 6.635700779399468e-07,
+      "loss": 0.7868,
       "step": 31254
     },
     {
-      "epoch": 0.8583944412402845,
+      "epoch": 0.8869182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 1.0334210748962237e-06,
-      "loss": 0.8042,
+      "learning_rate": 6.632408990891092e-07,
+      "loss": 0.6762,
       "step": 31255
     },
     {
-      "epoch": 0.8584219054681278,
+      "epoch": 0.8869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.033027300199062e-06,
-      "loss": 0.8427,
+      "learning_rate": 6.629117991062972e-07,
+      "loss": 0.8667,
       "step": 31256
     },
     {
-      "epoch": 0.858449369695971,
+      "epoch": 0.8869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.0326335964520595e-06,
-      "loss": 0.7499,
+      "learning_rate": 6.625827779942873e-07,
+      "loss": 0.8213,
       "step": 31257
     },
     {
-      "epoch": 0.8584768339238142,
+      "epoch": 0.8870034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 1.0322399636583359e-06,
-      "loss": 0.7401,
+      "learning_rate": 6.622538357558606e-07,
+      "loss": 0.8311,
       "step": 31258
     },
     {
-      "epoch": 0.8585042981516575,
+      "epoch": 0.8870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.0318464018210005e-06,
-      "loss": 0.7708,
+      "learning_rate": 6.619249723937959e-07,
+      "loss": 0.843,
       "step": 31259
     },
     {
-      "epoch": 0.8585317623795007,
+      "epoch": 0.8870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.0314529109431715e-06,
-      "loss": 0.8169,
+      "learning_rate": 6.615961879108701e-07,
+      "loss": 0.7427,
       "step": 31260
     },
     {
-      "epoch": 0.858559226607344,
+      "epoch": 0.887088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.0310594910279604e-06,
-      "loss": 0.8058,
+      "learning_rate": 6.612674823098631e-07,
+      "loss": 0.8177,
       "step": 31261
     },
     {
-      "epoch": 0.8585866908351871,
+      "epoch": 0.8871169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 1.0306661420784824e-06,
-      "loss": 0.8289,
+      "learning_rate": 6.609388555935481e-07,
+      "loss": 0.8666,
       "step": 31262
     },
     {
-      "epoch": 0.8586141550630304,
+      "epoch": 0.8871452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 1.0302728640978477e-06,
-      "loss": 0.8488,
+      "learning_rate": 6.60610307764702e-07,
+      "loss": 0.7995,
       "step": 31263
     },
     {
-      "epoch": 0.8586416192908737,
+      "epoch": 0.8871736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.0298796570891679e-06,
-      "loss": 0.8265,
+      "learning_rate": 6.602818388261012e-07,
+      "loss": 0.7515,
       "step": 31264
     },
     {
-      "epoch": 0.8586690835187168,
+      "epoch": 0.8872020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 1.029486521055555e-06,
-      "loss": 0.8624,
+      "learning_rate": 6.599534487805193e-07,
+      "loss": 0.7676,
       "step": 31265
     },
     {
-      "epoch": 0.8586965477465601,
+      "epoch": 0.8872304199772986,
       "grad_norm": 0.0,
-      "learning_rate": 1.029093456000121e-06,
-      "loss": 0.7843,
+      "learning_rate": 6.596251376307305e-07,
+      "loss": 0.8393,
       "step": 31266
     },
     {
-      "epoch": 0.8587240119744033,
+      "epoch": 0.8872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.0287004619259733e-06,
-      "loss": 0.8379,
+      "learning_rate": 6.592969053795084e-07,
+      "loss": 0.8403,
       "step": 31267
     },
     {
-      "epoch": 0.8587514762022466,
+      "epoch": 0.8872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.0283075388362252e-06,
-      "loss": 0.6828,
+      "learning_rate": 6.589687520296251e-07,
+      "loss": 0.8741,
       "step": 31268
     },
     {
-      "epoch": 0.8587789404300898,
+      "epoch": 0.8873155505107831,
       "grad_norm": 0.0,
-      "learning_rate": 1.0279146867339795e-06,
-      "loss": 0.8523,
+      "learning_rate": 6.586406775838517e-07,
+      "loss": 0.8591,
       "step": 31269
     },
     {
-      "epoch": 0.858806404657933,
+      "epoch": 0.8873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.0275219056223495e-06,
-      "loss": 0.7984,
+      "learning_rate": 6.583126820449626e-07,
+      "loss": 0.8795,
       "step": 31270
     },
     {
-      "epoch": 0.8588338688857763,
+      "epoch": 0.887372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.0271291955044416e-06,
-      "loss": 0.7939,
+      "learning_rate": 6.579847654157234e-07,
+      "loss": 0.813,
       "step": 31271
     },
     {
-      "epoch": 0.8588613331136196,
+      "epoch": 0.8874006810442678,
       "grad_norm": 0.0,
-      "learning_rate": 1.0267365563833632e-06,
-      "loss": 0.8121,
+      "learning_rate": 6.576569276989087e-07,
+      "loss": 0.798,
       "step": 31272
     },
     {
-      "epoch": 0.8588887973414627,
+      "epoch": 0.8874290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.026343988262223e-06,
-      "loss": 0.7301,
+      "learning_rate": 6.57329168897286e-07,
+      "loss": 0.802,
       "step": 31273
     },
     {
-      "epoch": 0.858916261569306,
+      "epoch": 0.8874574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 1.0259514911441238e-06,
-      "loss": 0.8245,
+      "learning_rate": 6.570014890136222e-07,
+      "loss": 0.8287,
       "step": 31274
     },
     {
-      "epoch": 0.8589437257971492,
+      "epoch": 0.8874858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.0255590650321733e-06,
-      "loss": 0.8028,
+      "learning_rate": 6.566738880506884e-07,
+      "loss": 0.8287,
       "step": 31275
     },
     {
-      "epoch": 0.8589711900249924,
+      "epoch": 0.8875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.0251667099294749e-06,
-      "loss": 0.7748,
+      "learning_rate": 6.56346366011249e-07,
+      "loss": 0.7573,
       "step": 31276
     },
     {
-      "epoch": 0.8589986542528357,
+      "epoch": 0.8875425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 1.0247744258391335e-06,
-      "loss": 0.9658,
+      "learning_rate": 6.560189228980717e-07,
+      "loss": 0.8101,
       "step": 31277
     },
     {
-      "epoch": 0.8590261184806789,
+      "epoch": 0.8875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.0243822127642545e-06,
-      "loss": 0.688,
+      "learning_rate": 6.556915587139246e-07,
+      "loss": 0.8124,
       "step": 31278
     },
     {
-      "epoch": 0.8590535827085222,
+      "epoch": 0.8875993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 1.0239900707079387e-06,
-      "loss": 0.6582,
+      "learning_rate": 6.553642734615695e-07,
+      "loss": 0.752,
       "step": 31279
     },
     {
-      "epoch": 0.8590810469363653,
+      "epoch": 0.887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.0235979996732937e-06,
-      "loss": 0.7784,
+      "learning_rate": 6.550370671437723e-07,
+      "loss": 0.7862,
       "step": 31280
     },
     {
-      "epoch": 0.8591085111642086,
+      "epoch": 0.8876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.0232059996634157e-06,
-      "loss": 0.8383,
+      "learning_rate": 6.547099397632983e-07,
+      "loss": 0.8156,
       "step": 31281
     },
     {
-      "epoch": 0.8591359753920519,
+      "epoch": 0.8876844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.0228140706814115e-06,
-      "loss": 0.899,
+      "learning_rate": 6.543828913229089e-07,
+      "loss": 0.8652,
       "step": 31282
     },
     {
-      "epoch": 0.859163439619895,
+      "epoch": 0.8877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.0224222127303773e-06,
-      "loss": 0.7509,
+      "learning_rate": 6.540559218253662e-07,
+      "loss": 0.7807,
       "step": 31283
     },
     {
-      "epoch": 0.8591909038477383,
+      "epoch": 0.8877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.0220304258134172e-06,
-      "loss": 0.8072,
+      "learning_rate": 6.537290312734356e-07,
+      "loss": 0.7694,
       "step": 31284
     },
     {
-      "epoch": 0.8592183680755816,
+      "epoch": 0.8877695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 1.021638709933629e-06,
-      "loss": 0.9058,
+      "learning_rate": 6.534022196698742e-07,
+      "loss": 0.7967,
       "step": 31285
     },
     {
-      "epoch": 0.8592458323034248,
+      "epoch": 0.8877979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 1.0212470650941164e-06,
-      "loss": 0.8656,
+      "learning_rate": 6.530754870174449e-07,
+      "loss": 0.8057,
       "step": 31286
     },
     {
-      "epoch": 0.859273296531268,
+      "epoch": 0.8878263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.0208554912979718e-06,
-      "loss": 0.8318,
+      "learning_rate": 6.52748833318908e-07,
+      "loss": 0.7871,
       "step": 31287
     },
     {
-      "epoch": 0.8593007607591112,
+      "epoch": 0.8878547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.020463988548297e-06,
-      "loss": 0.7999,
+      "learning_rate": 6.524222585770212e-07,
+      "loss": 0.8486,
       "step": 31288
     },
     {
-      "epoch": 0.8593282249869545,
+      "epoch": 0.887883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 1.0200725568481906e-06,
-      "loss": 0.7943,
+      "learning_rate": 6.520957627945424e-07,
+      "loss": 0.8758,
       "step": 31289
     },
     {
-      "epoch": 0.8593556892147978,
+      "epoch": 0.887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.019681196200749e-06,
-      "loss": 0.8168,
+      "learning_rate": 6.517693459742324e-07,
+      "loss": 0.7725,
       "step": 31290
     },
     {
-      "epoch": 0.8593831534426409,
+      "epoch": 0.8879398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.019289906609069e-06,
-      "loss": 0.7944,
+      "learning_rate": 6.514430081188461e-07,
+      "loss": 0.7311,
       "step": 31291
     },
     {
-      "epoch": 0.8594106176704842,
+      "epoch": 0.8879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.0188986880762442e-06,
-      "loss": 0.7869,
+      "learning_rate": 6.511167492311421e-07,
+      "loss": 0.7205,
       "step": 31292
     },
     {
-      "epoch": 0.8594380818983274,
+      "epoch": 0.8879965947786607,
       "grad_norm": 0.0,
-      "learning_rate": 1.0185075406053724e-06,
-      "loss": 0.8351,
+      "learning_rate": 6.50790569313875e-07,
+      "loss": 0.8383,
       "step": 31293
     },
     {
-      "epoch": 0.8594655461261707,
+      "epoch": 0.8880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.0181164641995479e-06,
-      "loss": 0.9153,
+      "learning_rate": 6.504644683697992e-07,
+      "loss": 0.7621,
       "step": 31294
     },
     {
-      "epoch": 0.8594930103540139,
+      "epoch": 0.8880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.0177254588618645e-06,
-      "loss": 0.8386,
+      "learning_rate": 6.501384464016725e-07,
+      "loss": 0.7116,
       "step": 31295
     },
     {
-      "epoch": 0.8595204745818571,
+      "epoch": 0.8880817253121452,
       "grad_norm": 0.0,
-      "learning_rate": 1.017334524595417e-06,
-      "loss": 0.795,
+      "learning_rate": 6.498125034122438e-07,
+      "loss": 0.8071,
       "step": 31296
     },
     {
-      "epoch": 0.8595479388097004,
+      "epoch": 0.8881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 1.0169436614033002e-06,
-      "loss": 0.8555,
+      "learning_rate": 6.494866394042709e-07,
+      "loss": 0.7332,
       "step": 31297
     },
     {
-      "epoch": 0.8595754030375437,
+      "epoch": 0.8881384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 1.016552869288603e-06,
-      "loss": 0.8254,
+      "learning_rate": 6.49160854380505e-07,
+      "loss": 0.8688,
       "step": 31298
     },
     {
-      "epoch": 0.8596028672653868,
+      "epoch": 0.8881668558456299,
       "grad_norm": 0.0,
-      "learning_rate": 1.0161621482544214e-06,
-      "loss": 0.8915,
+      "learning_rate": 6.488351483436961e-07,
+      "loss": 0.8746,
       "step": 31299
     },
     {
-      "epoch": 0.8596303314932301,
+      "epoch": 0.8881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 1.0157714983038424e-06,
-      "loss": 0.8344,
+      "learning_rate": 6.485095212965986e-07,
+      "loss": 0.8482,
       "step": 31300
     },
     {
-      "epoch": 0.8596577957210733,
+      "epoch": 0.8882236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 1.01538091943996e-06,
-      "loss": 0.8574,
+      "learning_rate": 6.481839732419615e-07,
+      "loss": 0.8526,
       "step": 31301
     },
     {
-      "epoch": 0.8596852599489165,
+      "epoch": 0.8882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 1.0149904116658637e-06,
-      "loss": 0.7543,
+      "learning_rate": 6.478585041825336e-07,
+      "loss": 0.8963,
       "step": 31302
     },
     {
-      "epoch": 0.8597127241767598,
+      "epoch": 0.8882803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 1.014599974984647e-06,
-      "loss": 0.8644,
+      "learning_rate": 6.475331141210661e-07,
+      "loss": 0.8274,
       "step": 31303
     },
     {
-      "epoch": 0.859740188404603,
+      "epoch": 0.8883087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 1.0142096093993925e-06,
-      "loss": 0.8422,
+      "learning_rate": 6.47207803060308e-07,
+      "loss": 0.7738,
       "step": 31304
     },
     {
-      "epoch": 0.8597676526324463,
+      "epoch": 0.8883371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 1.0138193149131925e-06,
-      "loss": 0.7471,
+      "learning_rate": 6.468825710030024e-07,
+      "loss": 0.9615,
       "step": 31305
     },
     {
-      "epoch": 0.8597951168602894,
+      "epoch": 0.8883654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 1.0134290915291378e-06,
-      "loss": 0.772,
+      "learning_rate": 6.465574179519029e-07,
+      "loss": 0.7508,
       "step": 31306
     },
     {
-      "epoch": 0.8598225810881327,
+      "epoch": 0.8883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 1.0130389392503103e-06,
-      "loss": 0.8928,
+      "learning_rate": 6.462323439097528e-07,
+      "loss": 0.791,
       "step": 31307
     },
     {
-      "epoch": 0.859850045315976,
+      "epoch": 0.888422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.0126488580797999e-06,
-      "loss": 0.7789,
+      "learning_rate": 6.459073488792989e-07,
+      "loss": 0.7152,
       "step": 31308
     },
     {
-      "epoch": 0.8598775095438191,
+      "epoch": 0.8884506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 1.0122588480206952e-06,
-      "loss": 0.7719,
+      "learning_rate": 6.455824328632865e-07,
+      "loss": 0.8655,
       "step": 31309
     },
     {
-      "epoch": 0.8599049737716624,
+      "epoch": 0.8884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 1.0118689090760792e-06,
-      "loss": 0.834,
+      "learning_rate": 6.452575958644592e-07,
+      "loss": 0.9061,
       "step": 31310
     },
     {
-      "epoch": 0.8599324379995057,
+      "epoch": 0.8885073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 1.011479041249036e-06,
-      "loss": 0.6914,
+      "learning_rate": 6.449328378855613e-07,
+      "loss": 0.8069,
       "step": 31311
     },
     {
-      "epoch": 0.8599599022273489,
+      "epoch": 0.8885357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 1.0110892445426545e-06,
-      "loss": 0.9334,
+      "learning_rate": 6.446081589293373e-07,
+      "loss": 0.721,
       "step": 31312
     },
     {
-      "epoch": 0.8599873664551921,
+      "epoch": 0.8885641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 1.0106995189600156e-06,
-      "loss": 0.8045,
+      "learning_rate": 6.442835589985274e-07,
+      "loss": 0.8966,
       "step": 31313
     },
     {
-      "epoch": 0.8600148306830353,
+      "epoch": 0.8885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 1.0103098645042064e-06,
-      "loss": 0.7472,
+      "learning_rate": 6.439590380958749e-07,
+      "loss": 0.8709,
       "step": 31314
     },
     {
-      "epoch": 0.8600422949108786,
+      "epoch": 0.8886208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 1.009920281178308e-06,
-      "loss": 0.7851,
+      "learning_rate": 6.43634596224123e-07,
+      "loss": 0.7715,
       "step": 31315
     },
     {
-      "epoch": 0.8600697591387219,
+      "epoch": 0.8886492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 1.0095307689854005e-06,
-      "loss": 0.7202,
+      "learning_rate": 6.433102333860075e-07,
+      "loss": 0.7975,
       "step": 31316
     },
     {
-      "epoch": 0.860097223366565,
+      "epoch": 0.8886776390465381,
       "grad_norm": 0.0,
-      "learning_rate": 1.0091413279285689e-06,
-      "loss": 0.8195,
+      "learning_rate": 6.429859495842727e-07,
+      "loss": 0.6893,
       "step": 31317
     },
     {
-      "epoch": 0.8601246875944083,
+      "epoch": 0.8887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 1.0087519580108928e-06,
-      "loss": 0.8099,
+      "learning_rate": 6.426617448216565e-07,
+      "loss": 0.8039,
       "step": 31318
     },
     {
-      "epoch": 0.8601521518222515,
+      "epoch": 0.8887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 1.0083626592354545e-06,
-      "loss": 0.8098,
+      "learning_rate": 6.423376191008967e-07,
+      "loss": 0.7086,
       "step": 31319
     },
     {
-      "epoch": 0.8601796160500947,
+      "epoch": 0.8887627695800226,
       "grad_norm": 0.0,
-      "learning_rate": 1.0079734316053346e-06,
-      "loss": 0.7509,
+      "learning_rate": 6.4201357242473e-07,
+      "loss": 0.7211,
       "step": 31320
     },
     {
-      "epoch": 0.860207080277938,
+      "epoch": 0.8887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.0075842751236132e-06,
-      "loss": 0.7657,
+      "learning_rate": 6.416896047958965e-07,
+      "loss": 0.7332,
       "step": 31321
     },
     {
-      "epoch": 0.8602345445057812,
+      "epoch": 0.8888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 1.0071951897933684e-06,
-      "loss": 0.7949,
+      "learning_rate": 6.413657162171317e-07,
+      "loss": 0.807,
       "step": 31322
     },
     {
-      "epoch": 0.8602620087336245,
+      "epoch": 0.8888479001135073,
       "grad_norm": 0.0,
-      "learning_rate": 1.0068061756176773e-06,
-      "loss": 0.7441,
+      "learning_rate": 6.410419066911732e-07,
+      "loss": 0.7151,
       "step": 31323
     },
     {
-      "epoch": 0.8602894729614677,
+      "epoch": 0.8888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 1.0064172325996191e-06,
-      "loss": 0.752,
+      "learning_rate": 6.407181762207526e-07,
+      "loss": 0.8723,
       "step": 31324
     },
     {
-      "epoch": 0.8603169371893109,
+      "epoch": 0.8889046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 1.0060283607422706e-06,
-      "loss": 0.7607,
+      "learning_rate": 6.403945248086052e-07,
+      "loss": 0.7811,
       "step": 31325
     },
     {
-      "epoch": 0.8603444014171542,
+      "epoch": 0.888933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.0056395600487113e-06,
-      "loss": 0.8501,
+      "learning_rate": 6.400709524574688e-07,
+      "loss": 0.8169,
       "step": 31326
     },
     {
-      "epoch": 0.8603718656449973,
+      "epoch": 0.888961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.0052508305220165e-06,
-      "loss": 0.7741,
+      "learning_rate": 6.397474591700726e-07,
+      "loss": 0.777,
       "step": 31327
     },
     {
-      "epoch": 0.8603993298728406,
+      "epoch": 0.8889897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 1.0048621721652596e-06,
-      "loss": 0.8383,
+      "learning_rate": 6.394240449491496e-07,
+      "loss": 0.8112,
       "step": 31328
     },
     {
-      "epoch": 0.8604267941006839,
+      "epoch": 0.8890181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 1.0044735849815178e-06,
-      "loss": 0.8472,
+      "learning_rate": 6.391007097974333e-07,
+      "loss": 0.812,
       "step": 31329
     },
     {
-      "epoch": 0.8604542583285271,
+      "epoch": 0.8890465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 1.004085068973868e-06,
-      "loss": 0.8721,
+      "learning_rate": 6.387774537176538e-07,
+      "loss": 0.8192,
       "step": 31330
     },
     {
-      "epoch": 0.8604817225563703,
+      "epoch": 0.8890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 1.0036966241453793e-06,
-      "loss": 0.8613,
+      "learning_rate": 6.384542767125412e-07,
+      "loss": 0.7879,
       "step": 31331
     },
     {
-      "epoch": 0.8605091867842135,
+      "epoch": 0.8891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 1.0033082504991286e-06,
-      "loss": 0.7657,
+      "learning_rate": 6.381311787848287e-07,
+      "loss": 0.8301,
       "step": 31332
     },
     {
-      "epoch": 0.8605366510120568,
+      "epoch": 0.8891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 1.002919948038189e-06,
-      "loss": 0.7672,
+      "learning_rate": 6.37808159937241e-07,
+      "loss": 0.699,
       "step": 31333
     },
     {
-      "epoch": 0.8605641152399001,
+      "epoch": 0.8891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 1.0025317167656301e-06,
-      "loss": 0.8125,
+      "learning_rate": 6.374852201725079e-07,
+      "loss": 0.7682,
       "step": 31334
     },
     {
-      "epoch": 0.8605915794677432,
+      "epoch": 0.8891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 1.0021435566845262e-06,
-      "loss": 0.8941,
+      "learning_rate": 6.371623594933596e-07,
+      "loss": 0.8529,
       "step": 31335
     },
     {
-      "epoch": 0.8606190436955865,
+      "epoch": 0.889216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.0017554677979468e-06,
-      "loss": 0.8353,
+      "learning_rate": 6.368395779025194e-07,
+      "loss": 0.8387,
       "step": 31336
     },
     {
-      "epoch": 0.8606465079234298,
+      "epoch": 0.8892451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 1.0013674501089644e-06,
-      "loss": 0.7966,
+      "learning_rate": 6.365168754027173e-07,
+      "loss": 0.8633,
       "step": 31337
     },
     {
-      "epoch": 0.860673972151273,
+      "epoch": 0.8892735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 1.0009795036206504e-06,
-      "loss": 0.8114,
+      "learning_rate": 6.361942519966802e-07,
+      "loss": 0.8328,
       "step": 31338
     },
     {
-      "epoch": 0.8607014363791162,
+      "epoch": 0.8893019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 1.0005916283360717e-06,
-      "loss": 0.8561,
+      "learning_rate": 6.35871707687129e-07,
+      "loss": 0.9023,
       "step": 31339
     },
     {
-      "epoch": 0.8607289006069594,
+      "epoch": 0.8893303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 1.0002038242582968e-06,
-      "loss": 0.7799,
+      "learning_rate": 6.355492424767906e-07,
+      "loss": 0.8787,
       "step": 31340
     },
     {
-      "epoch": 0.8607563648348027,
+      "epoch": 0.8893586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 9.998160913903953e-07,
-      "loss": 0.7793,
+      "learning_rate": 6.352268563683905e-07,
+      "loss": 0.6715,
       "step": 31341
     },
     {
-      "epoch": 0.8607838290626459,
+      "epoch": 0.8893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 9.994284297354362e-07,
-      "loss": 0.8299,
+      "learning_rate": 6.349045493646489e-07,
+      "loss": 0.8462,
       "step": 31342
     },
     {
-      "epoch": 0.8608112932904891,
+      "epoch": 0.8894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 9.990408392964845e-07,
-      "loss": 0.7728,
+      "learning_rate": 6.345823214682889e-07,
+      "loss": 0.8313,
       "step": 31343
     },
     {
-      "epoch": 0.8608387575183324,
+      "epoch": 0.8894438138479002,
       "grad_norm": 0.0,
-      "learning_rate": 9.986533200766092e-07,
-      "loss": 0.7257,
+      "learning_rate": 6.342601726820341e-07,
+      "loss": 0.8654,
       "step": 31344
     },
     {
-      "epoch": 0.8608662217461756,
+      "epoch": 0.889472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 9.982658720788773e-07,
-      "loss": 0.8586,
+      "learning_rate": 6.339381030086045e-07,
+      "loss": 0.7577,
       "step": 31345
     },
     {
-      "epoch": 0.8608936859740188,
+      "epoch": 0.8895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 9.978784953063526e-07,
-      "loss": 0.8022,
+      "learning_rate": 6.336161124507212e-07,
+      "loss": 0.822,
       "step": 31346
     },
     {
-      "epoch": 0.8609211502018621,
+      "epoch": 0.8895289443813847,
       "grad_norm": 0.0,
-      "learning_rate": 9.974911897620987e-07,
-      "loss": 0.7896,
+      "learning_rate": 6.33294201011102e-07,
+      "loss": 0.6691,
       "step": 31347
     },
     {
-      "epoch": 0.8609486144297053,
+      "epoch": 0.8895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 9.971039554491814e-07,
-      "loss": 0.8566,
+      "learning_rate": 6.329723686924682e-07,
+      "loss": 0.7229,
       "step": 31348
     },
     {
-      "epoch": 0.8609760786575485,
+      "epoch": 0.8895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 9.967167923706644e-07,
-      "loss": 0.8864,
+      "learning_rate": 6.326506154975398e-07,
+      "loss": 0.8359,
       "step": 31349
     },
     {
-      "epoch": 0.8610035428853918,
+      "epoch": 0.8896140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 9.963297005296146e-07,
-      "loss": 0.8164,
+      "learning_rate": 6.323289414290312e-07,
+      "loss": 0.8487,
       "step": 31350
     },
     {
-      "epoch": 0.861031007113235,
+      "epoch": 0.8896424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 9.959426799290895e-07,
-      "loss": 0.773,
+      "learning_rate": 6.320073464896592e-07,
+      "loss": 0.7523,
       "step": 31351
     },
     {
-      "epoch": 0.8610584713410783,
+      "epoch": 0.8896708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 9.955557305721542e-07,
-      "loss": 0.8413,
+      "learning_rate": 6.316858306821449e-07,
+      "loss": 0.7606,
       "step": 31352
     },
     {
-      "epoch": 0.8610859355689214,
+      "epoch": 0.8896992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 9.951688524618696e-07,
-      "loss": 0.7244,
+      "learning_rate": 6.313643940092007e-07,
+      "loss": 0.8874,
       "step": 31353
     },
     {
-      "epoch": 0.8611133997967647,
+      "epoch": 0.889727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 9.947820456012991e-07,
-      "loss": 0.866,
+      "learning_rate": 6.31043036473542e-07,
+      "loss": 0.7884,
       "step": 31354
     },
     {
-      "epoch": 0.861140864024608,
+      "epoch": 0.889755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 9.94395309993499e-07,
-      "loss": 0.9059,
+      "learning_rate": 6.307217580778857e-07,
+      "loss": 0.8901,
       "step": 31355
     },
     {
-      "epoch": 0.8611683282524512,
+      "epoch": 0.8897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 9.940086456415343e-07,
-      "loss": 0.8298,
+      "learning_rate": 6.304005588249429e-07,
+      "loss": 0.8153,
       "step": 31356
     },
     {
-      "epoch": 0.8611957924802944,
+      "epoch": 0.8898127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 9.936220525484596e-07,
-      "loss": 0.7751,
+      "learning_rate": 6.300794387174269e-07,
+      "loss": 0.7002,
       "step": 31357
     },
     {
-      "epoch": 0.8612232567081376,
+      "epoch": 0.8898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 9.93235530717336e-07,
-      "loss": 0.8233,
+      "learning_rate": 6.297583977580535e-07,
+      "loss": 0.7985,
       "step": 31358
     },
     {
-      "epoch": 0.8612507209359809,
+      "epoch": 0.8898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 9.928490801512214e-07,
-      "loss": 0.7942,
+      "learning_rate": 6.294374359495303e-07,
+      "loss": 0.8466,
       "step": 31359
     },
     {
-      "epoch": 0.8612781851638242,
+      "epoch": 0.8898978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 9.92462700853175e-07,
-      "loss": 0.7888,
+      "learning_rate": 6.291165532945697e-07,
+      "loss": 0.8726,
       "step": 31360
     },
     {
-      "epoch": 0.8613056493916673,
+      "epoch": 0.8899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 9.920763928262534e-07,
-      "loss": 0.6928,
+      "learning_rate": 6.287957497958852e-07,
+      "loss": 0.7844,
       "step": 31361
     },
     {
-      "epoch": 0.8613331136195106,
+      "epoch": 0.8899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 9.916901560735115e-07,
-      "loss": 0.8235,
+      "learning_rate": 6.284750254561822e-07,
+      "loss": 0.8509,
       "step": 31362
     },
     {
-      "epoch": 0.8613605778473539,
+      "epoch": 0.889982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 9.913039905980082e-07,
-      "loss": 0.8532,
+      "learning_rate": 6.281543802781731e-07,
+      "loss": 0.8179,
       "step": 31363
     },
     {
-      "epoch": 0.861388042075197,
+      "epoch": 0.8900113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 9.909178964027966e-07,
-      "loss": 0.8813,
+      "learning_rate": 6.278338142645657e-07,
+      "loss": 0.7327,
       "step": 31364
     },
     {
-      "epoch": 0.8614155063030403,
+      "epoch": 0.8900397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 9.905318734909307e-07,
-      "loss": 0.8589,
+      "learning_rate": 6.275133274180656e-07,
+      "loss": 0.8294,
       "step": 31365
     },
     {
-      "epoch": 0.8614429705308835,
+      "epoch": 0.8900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 9.901459218654674e-07,
-      "loss": 0.8126,
+      "learning_rate": 6.271929197413817e-07,
+      "loss": 0.8317,
       "step": 31366
     },
     {
-      "epoch": 0.8614704347587268,
+      "epoch": 0.8900964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 9.897600415294596e-07,
-      "loss": 0.8825,
+      "learning_rate": 6.268725912372209e-07,
+      "loss": 0.782,
       "step": 31367
     },
     {
-      "epoch": 0.86149789898657,
+      "epoch": 0.8901248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 9.893742324859624e-07,
-      "loss": 0.7789,
+      "learning_rate": 6.265523419082897e-07,
+      "loss": 0.7889,
       "step": 31368
     },
     {
-      "epoch": 0.8615253632144132,
+      "epoch": 0.8901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 9.889884947380257e-07,
-      "loss": 0.8551,
+      "learning_rate": 6.262321717572928e-07,
+      "loss": 0.7714,
       "step": 31369
     },
     {
-      "epoch": 0.8615528274422565,
+      "epoch": 0.8901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 9.886028282887027e-07,
-      "loss": 0.9073,
+      "learning_rate": 6.259120807869324e-07,
+      "loss": 0.8201,
       "step": 31370
     },
     {
-      "epoch": 0.8615802916700996,
+      "epoch": 0.8902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 9.882172331410444e-07,
-      "loss": 0.8832,
+      "learning_rate": 6.255920689999139e-07,
+      "loss": 0.7077,
       "step": 31371
     },
     {
-      "epoch": 0.8616077558979429,
+      "epoch": 0.8902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 9.878317092981016e-07,
-      "loss": 0.7761,
+      "learning_rate": 6.252721363989423e-07,
+      "loss": 0.7664,
       "step": 31372
     },
     {
-      "epoch": 0.8616352201257862,
+      "epoch": 0.890266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 9.87446256762925e-07,
-      "loss": 0.7488,
+      "learning_rate": 6.249522829867172e-07,
+      "loss": 0.7894,
       "step": 31373
     },
     {
-      "epoch": 0.8616626843536294,
+      "epoch": 0.8902951191827468,
       "grad_norm": 0.0,
-      "learning_rate": 9.87060875538567e-07,
-      "loss": 0.7206,
+      "learning_rate": 6.246325087659411e-07,
+      "loss": 0.8111,
       "step": 31374
     },
     {
-      "epoch": 0.8616901485814726,
+      "epoch": 0.8903234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 9.866755656280724e-07,
-      "loss": 0.8237,
+      "learning_rate": 6.243128137393184e-07,
+      "loss": 0.7102,
       "step": 31375
     },
     {
-      "epoch": 0.8617176128093159,
+      "epoch": 0.8903518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 9.862903270344926e-07,
-      "loss": 0.7864,
+      "learning_rate": 6.239931979095437e-07,
+      "loss": 0.9014,
       "step": 31376
     },
     {
-      "epoch": 0.8617450770371591,
+      "epoch": 0.8903802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 9.859051597608748e-07,
-      "loss": 0.7958,
+      "learning_rate": 6.236736612793204e-07,
+      "loss": 0.7601,
       "step": 31377
     },
     {
-      "epoch": 0.8617725412650024,
+      "epoch": 0.8904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 9.855200638102691e-07,
-      "loss": 0.7632,
+      "learning_rate": 6.233542038513496e-07,
+      "loss": 0.8115,
       "step": 31378
     },
     {
-      "epoch": 0.8618000054928455,
+      "epoch": 0.8904370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 9.851350391857194e-07,
-      "loss": 0.7881,
+      "learning_rate": 6.230348256283247e-07,
+      "loss": 0.8396,
       "step": 31379
     },
     {
-      "epoch": 0.8618274697206888,
+      "epoch": 0.8904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 9.847500858902748e-07,
-      "loss": 0.7925,
+      "learning_rate": 6.22715526612947e-07,
+      "loss": 0.8745,
       "step": 31380
     },
     {
-      "epoch": 0.8618549339485321,
+      "epoch": 0.8904937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 9.84365203926977e-07,
-      "loss": 0.7961,
+      "learning_rate": 6.223963068079142e-07,
+      "loss": 0.8441,
       "step": 31381
     },
     {
-      "epoch": 0.8618823981763752,
+      "epoch": 0.890522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 9.839803932988745e-07,
-      "loss": 0.8382,
+      "learning_rate": 6.220771662159175e-07,
+      "loss": 0.8921,
       "step": 31382
     },
     {
-      "epoch": 0.8619098624042185,
+      "epoch": 0.8905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 9.83595654009012e-07,
-      "loss": 0.7814,
+      "learning_rate": 6.217581048396604e-07,
+      "loss": 0.7331,
       "step": 31383
     },
     {
-      "epoch": 0.8619373266320617,
+      "epoch": 0.8905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 9.832109860604333e-07,
-      "loss": 0.7959,
+      "learning_rate": 6.214391226818329e-07,
+      "loss": 0.7514,
       "step": 31384
     },
     {
-      "epoch": 0.861964790859905,
+      "epoch": 0.8906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 9.82826389456184e-07,
-      "loss": 0.8304,
+      "learning_rate": 6.211202197451294e-07,
+      "loss": 0.7427,
       "step": 31385
     },
     {
-      "epoch": 0.8619922550877482,
+      "epoch": 0.8906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 9.824418641993028e-07,
-      "loss": 0.8234,
+      "learning_rate": 6.208013960322478e-07,
+      "loss": 0.7678,
       "step": 31386
     },
     {
-      "epoch": 0.8620197193155914,
+      "epoch": 0.8906640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 9.82057410292837e-07,
-      "loss": 0.8057,
+      "learning_rate": 6.204826515458762e-07,
+      "loss": 0.7386,
       "step": 31387
     },
     {
-      "epoch": 0.8620471835434347,
+      "epoch": 0.8906923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 9.816730277398235e-07,
-      "loss": 0.7949,
+      "learning_rate": 6.201639862887099e-07,
+      "loss": 0.8536,
       "step": 31388
     },
     {
-      "epoch": 0.862074647771278,
+      "epoch": 0.8907207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 9.812887165433072e-07,
-      "loss": 0.897,
+      "learning_rate": 6.198454002634414e-07,
+      "loss": 0.8085,
       "step": 31389
     },
     {
-      "epoch": 0.8621021119991211,
+      "epoch": 0.8907491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 9.80904476706328e-07,
-      "loss": 0.9092,
+      "learning_rate": 6.195268934727583e-07,
+      "loss": 0.7527,
       "step": 31390
     },
     {
-      "epoch": 0.8621295762269644,
+      "epoch": 0.89077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 9.805203082319258e-07,
-      "loss": 0.8901,
+      "learning_rate": 6.192084659193542e-07,
+      "loss": 0.7662,
       "step": 31391
     },
     {
-      "epoch": 0.8621570404548076,
+      "epoch": 0.890805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 9.801362111231428e-07,
-      "loss": 0.7599,
+      "learning_rate": 6.188901176059181e-07,
+      "loss": 0.8061,
       "step": 31392
     },
     {
-      "epoch": 0.8621845046826508,
+      "epoch": 0.8908342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 9.79752185383015e-07,
-      "loss": 0.8505,
+      "learning_rate": 6.185718485351388e-07,
+      "loss": 0.7588,
       "step": 31393
     },
     {
-      "epoch": 0.8622119689104941,
+      "epoch": 0.8908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 9.793682310145825e-07,
-      "loss": 0.7699,
+      "learning_rate": 6.182536587097043e-07,
+      "loss": 0.7892,
       "step": 31394
     },
     {
-      "epoch": 0.8622394331383373,
+      "epoch": 0.8908910329171397,
       "grad_norm": 0.0,
-      "learning_rate": 9.789843480208815e-07,
-      "loss": 0.7669,
+      "learning_rate": 6.179355481323035e-07,
+      "loss": 0.9762,
       "step": 31395
     },
     {
-      "epoch": 0.8622668973661806,
+      "epoch": 0.8909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 9.78600536404951e-07,
-      "loss": 0.7642,
+      "learning_rate": 6.176175168056231e-07,
+      "loss": 0.8548,
       "step": 31396
     },
     {
-      "epoch": 0.8622943615940238,
+      "epoch": 0.8909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 9.782167961698275e-07,
-      "loss": 0.9125,
+      "learning_rate": 6.172995647323477e-07,
+      "loss": 0.8857,
       "step": 31397
     },
     {
-      "epoch": 0.862321825821867,
+      "epoch": 0.8909761634506242,
       "grad_norm": 0.0,
-      "learning_rate": 9.778331273185494e-07,
-      "loss": 0.88,
+      "learning_rate": 6.169816919151649e-07,
+      "loss": 0.8423,
       "step": 31398
     },
     {
-      "epoch": 0.8623492900497103,
+      "epoch": 0.8910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 9.774495298541486e-07,
-      "loss": 0.7474,
+      "learning_rate": 6.166638983567607e-07,
+      "loss": 0.8354,
       "step": 31399
     },
     {
-      "epoch": 0.8623767542775534,
+      "epoch": 0.8910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 9.77066003779662e-07,
-      "loss": 0.8509,
+      "learning_rate": 6.163461840598184e-07,
+      "loss": 0.7692,
       "step": 31400
     },
     {
-      "epoch": 0.8624042185053967,
+      "epoch": 0.8910612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 9.766825490981235e-07,
-      "loss": 0.7373,
+      "learning_rate": 6.160285490270212e-07,
+      "loss": 0.7909,
       "step": 31401
     },
     {
-      "epoch": 0.86243168273324,
+      "epoch": 0.8910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 9.76299165812571e-07,
-      "loss": 0.7815,
+      "learning_rate": 6.157109932610527e-07,
+      "loss": 0.8035,
       "step": 31402
     },
     {
-      "epoch": 0.8624591469610832,
+      "epoch": 0.8911180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 9.759158539260328e-07,
-      "loss": 0.7516,
+      "learning_rate": 6.153935167645964e-07,
+      "loss": 0.8195,
       "step": 31403
     },
     {
-      "epoch": 0.8624866111889264,
+      "epoch": 0.8911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 9.755326134415454e-07,
-      "loss": 0.7893,
+      "learning_rate": 6.150761195403321e-07,
+      "loss": 0.8833,
       "step": 31404
     },
     {
-      "epoch": 0.8625140754167696,
+      "epoch": 0.8911748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 9.75149444362139e-07,
-      "loss": 0.7751,
+      "learning_rate": 6.147588015909412e-07,
+      "loss": 0.8149,
       "step": 31405
     },
     {
-      "epoch": 0.8625415396446129,
+      "epoch": 0.8912031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 9.747663466908452e-07,
-      "loss": 0.8643,
+      "learning_rate": 6.144415629191058e-07,
+      "loss": 0.8074,
       "step": 31406
     },
     {
-      "epoch": 0.8625690038724562,
+      "epoch": 0.8912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 9.74383320430695e-07,
-      "loss": 0.6578,
+      "learning_rate": 6.141244035275029e-07,
+      "loss": 0.8543,
       "step": 31407
     },
     {
-      "epoch": 0.8625964681002993,
+      "epoch": 0.8912599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 9.74000365584722e-07,
-      "loss": 0.8522,
+      "learning_rate": 6.138073234188136e-07,
+      "loss": 0.7698,
       "step": 31408
     },
     {
-      "epoch": 0.8626239323281426,
+      "epoch": 0.8912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 9.736174821559553e-07,
-      "loss": 0.8184,
+      "learning_rate": 6.134903225957168e-07,
+      "loss": 0.7828,
       "step": 31409
     },
     {
-      "epoch": 0.8626513965559859,
+      "epoch": 0.891316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.732346701474227e-07,
-      "loss": 0.8581,
+      "learning_rate": 6.131734010608881e-07,
+      "loss": 0.8346,
       "step": 31410
     },
     {
-      "epoch": 0.862678860783829,
+      "epoch": 0.8913450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 9.728519295621553e-07,
-      "loss": 0.7791,
+      "learning_rate": 6.128565588170055e-07,
+      "loss": 0.7649,
       "step": 31411
     },
     {
-      "epoch": 0.8627063250116723,
+      "epoch": 0.8913734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 9.724692604031782e-07,
-      "loss": 0.7379,
+      "learning_rate": 6.125397958667467e-07,
+      "loss": 0.7336,
       "step": 31412
     },
     {
-      "epoch": 0.8627337892395155,
+      "epoch": 0.8914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 9.720866626735215e-07,
-      "loss": 0.7784,
+      "learning_rate": 6.122231122127831e-07,
+      "loss": 0.718,
       "step": 31413
     },
     {
-      "epoch": 0.8627612534673588,
+      "epoch": 0.8914301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 9.717041363762115e-07,
-      "loss": 0.8564,
+      "learning_rate": 6.119065078577968e-07,
+      "loss": 0.8368,
       "step": 31414
     },
     {
-      "epoch": 0.862788717695202,
+      "epoch": 0.8914585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 9.713216815142779e-07,
-      "loss": 0.7527,
+      "learning_rate": 6.115899828044569e-07,
+      "loss": 0.7972,
       "step": 31415
     },
     {
-      "epoch": 0.8628161819230452,
+      "epoch": 0.8914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 9.709392980907428e-07,
-      "loss": 0.8275,
+      "learning_rate": 6.112735370554368e-07,
+      "loss": 0.8328,
       "step": 31416
     },
     {
-      "epoch": 0.8628436461508885,
+      "epoch": 0.8915153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 9.70556986108634e-07,
-      "loss": 0.8,
+      "learning_rate": 6.109571706134143e-07,
+      "loss": 0.8131,
       "step": 31417
     },
     {
-      "epoch": 0.8628711103787317,
+      "epoch": 0.8915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 9.701747455709765e-07,
-      "loss": 0.8115,
+      "learning_rate": 6.106408834810563e-07,
+      "loss": 0.8231,
       "step": 31418
     },
     {
-      "epoch": 0.8628985746065749,
+      "epoch": 0.8915720771850171,
       "grad_norm": 0.0,
-      "learning_rate": 9.697925764807937e-07,
-      "loss": 0.8157,
+      "learning_rate": 6.10324675661037e-07,
+      "loss": 0.7164,
       "step": 31419
     },
     {
-      "epoch": 0.8629260388344182,
+      "epoch": 0.8916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 9.694104788411095e-07,
-      "loss": 0.7982,
+      "learning_rate": 6.10008547156029e-07,
+      "loss": 0.7806,
       "step": 31420
     },
     {
-      "epoch": 0.8629535030622614,
+      "epoch": 0.8916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 9.690284526549498e-07,
-      "loss": 0.7858,
+      "learning_rate": 6.096924979686991e-07,
+      "loss": 0.8346,
       "step": 31421
     },
     {
-      "epoch": 0.8629809672901047,
+      "epoch": 0.8916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 9.68646497925333e-07,
-      "loss": 0.7099,
+      "learning_rate": 6.093765281017195e-07,
+      "loss": 0.8419,
       "step": 31422
     },
     {
-      "epoch": 0.8630084315179479,
+      "epoch": 0.8916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 9.682646146552832e-07,
-      "loss": 0.6999,
+      "learning_rate": 6.09060637557759e-07,
+      "loss": 0.6774,
       "step": 31423
     },
     {
-      "epoch": 0.8630358957457911,
+      "epoch": 0.8917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 9.678828028478228e-07,
-      "loss": 0.8312,
+      "learning_rate": 6.087448263394846e-07,
+      "loss": 0.8633,
       "step": 31424
     },
     {
-      "epoch": 0.8630633599736344,
+      "epoch": 0.8917423382519863,
       "grad_norm": 0.0,
-      "learning_rate": 9.675010625059723e-07,
-      "loss": 0.7787,
+      "learning_rate": 6.084290944495652e-07,
+      "loss": 0.7245,
       "step": 31425
     },
     {
-      "epoch": 0.8630908242014775,
+      "epoch": 0.8917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 9.67119393632754e-07,
-      "loss": 0.7555,
+      "learning_rate": 6.081134418906675e-07,
+      "loss": 0.8115,
       "step": 31426
     },
     {
-      "epoch": 0.8631182884293208,
+      "epoch": 0.8917990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 9.667377962311863e-07,
-      "loss": 0.7999,
+      "learning_rate": 6.077978686654573e-07,
+      "loss": 0.84,
       "step": 31427
     },
     {
-      "epoch": 0.8631457526571641,
+      "epoch": 0.891827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 9.663562703042862e-07,
-      "loss": 0.8452,
+      "learning_rate": 6.074823747766012e-07,
+      "loss": 0.8461,
       "step": 31428
     },
     {
-      "epoch": 0.8631732168850073,
+      "epoch": 0.8918558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 9.659748158550753e-07,
-      "loss": 0.8633,
+      "learning_rate": 6.071669602267627e-07,
+      "loss": 0.9329,
       "step": 31429
     },
     {
-      "epoch": 0.8632006811128505,
+      "epoch": 0.8918842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 9.655934328865701e-07,
-      "loss": 0.7661,
+      "learning_rate": 6.068516250186074e-07,
+      "loss": 0.8586,
       "step": 31430
     },
     {
-      "epoch": 0.8632281453406937,
+      "epoch": 0.8919125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 9.652121214017896e-07,
-      "loss": 0.7898,
+      "learning_rate": 6.065363691547999e-07,
+      "loss": 0.9041,
       "step": 31431
     },
     {
-      "epoch": 0.863255609568537,
+      "epoch": 0.8919409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 9.648308814037511e-07,
-      "loss": 0.8857,
+      "learning_rate": 6.062211926380012e-07,
+      "loss": 0.7368,
       "step": 31432
     },
     {
-      "epoch": 0.8632830737963803,
+      "epoch": 0.8919693530079456,
       "grad_norm": 0.0,
-      "learning_rate": 9.644497128954722e-07,
-      "loss": 0.7698,
+      "learning_rate": 6.05906095470874e-07,
+      "loss": 0.8035,
       "step": 31433
     },
     {
-      "epoch": 0.8633105380242234,
+      "epoch": 0.8919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 9.64068615879965e-07,
-      "loss": 0.8262,
+      "learning_rate": 6.055910776560813e-07,
+      "loss": 0.7894,
       "step": 31434
     },
     {
-      "epoch": 0.8633380022520667,
+      "epoch": 0.8920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 9.636875903602493e-07,
-      "loss": 0.7233,
+      "learning_rate": 6.052761391962825e-07,
+      "loss": 0.7742,
       "step": 31435
     },
     {
-      "epoch": 0.86336546647991,
+      "epoch": 0.8920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 9.63306636339335e-07,
-      "loss": 0.7398,
+      "learning_rate": 6.049612800941385e-07,
+      "loss": 0.8311,
       "step": 31436
     },
     {
-      "epoch": 0.8633929307077531,
+      "epoch": 0.8920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 9.629257538202397e-07,
-      "loss": 0.8076,
+      "learning_rate": 6.046465003523094e-07,
+      "loss": 0.7296,
       "step": 31437
     },
     {
-      "epoch": 0.8634203949355964,
+      "epoch": 0.89211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.625449428059763e-07,
-      "loss": 0.8832,
+      "learning_rate": 6.043317999734533e-07,
+      "loss": 0.8382,
       "step": 31438
     },
     {
-      "epoch": 0.8634478591634396,
+      "epoch": 0.8921396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 9.621642032995603e-07,
-      "loss": 0.7734,
+      "learning_rate": 6.040171789602278e-07,
+      "loss": 0.8133,
       "step": 31439
     },
     {
-      "epoch": 0.8634753233912829,
+      "epoch": 0.8921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 9.617835353040007e-07,
-      "loss": 0.8537,
+      "learning_rate": 6.037026373152933e-07,
+      "loss": 0.8379,
       "step": 31440
     },
     {
-      "epoch": 0.8635027876191261,
+      "epoch": 0.8921963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 9.6140293882231e-07,
-      "loss": 0.8027,
+      "learning_rate": 6.03388175041304e-07,
+      "loss": 0.8354,
       "step": 31441
     },
     {
-      "epoch": 0.8635302518469693,
+      "epoch": 0.8922247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 9.61022413857502e-07,
-      "loss": 0.8586,
+      "learning_rate": 6.030737921409169e-07,
+      "loss": 0.7798,
       "step": 31442
     },
     {
-      "epoch": 0.8635577160748126,
+      "epoch": 0.8922531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 9.60641960412585e-07,
-      "loss": 0.9403,
+      "learning_rate": 6.027594886167876e-07,
+      "loss": 0.8146,
       "step": 31443
     },
     {
-      "epoch": 0.8635851803026557,
+      "epoch": 0.8922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 9.6026157849057e-07,
-      "loss": 0.8938,
+      "learning_rate": 6.024452644715717e-07,
+      "loss": 0.7202,
       "step": 31444
     },
     {
-      "epoch": 0.863612644530499,
+      "epoch": 0.8923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 9.59881268094469e-07,
-      "loss": 0.7963,
+      "learning_rate": 6.021311197079228e-07,
+      "loss": 0.6989,
       "step": 31445
     },
     {
-      "epoch": 0.8636401087583423,
+      "epoch": 0.8923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 9.595010292272866e-07,
-      "loss": 0.8248,
+      "learning_rate": 6.018170543284973e-07,
+      "loss": 0.8075,
       "step": 31446
     },
     {
-      "epoch": 0.8636675729861855,
+      "epoch": 0.892366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 9.591208618920346e-07,
-      "loss": 0.7666,
+      "learning_rate": 6.015030683359435e-07,
+      "loss": 0.7743,
       "step": 31447
     },
     {
-      "epoch": 0.8636950372140287,
+      "epoch": 0.8923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 9.587407660917202e-07,
-      "loss": 0.8882,
+      "learning_rate": 6.011891617329147e-07,
+      "loss": 0.8486,
       "step": 31448
     },
     {
-      "epoch": 0.863722501441872,
+      "epoch": 0.8924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 9.583607418293518e-07,
-      "loss": 0.8657,
+      "learning_rate": 6.008753345220664e-07,
+      "loss": 0.8311,
       "step": 31449
     },
     {
-      "epoch": 0.8637499656697152,
+      "epoch": 0.8924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 9.579807891079373e-07,
-      "loss": 0.8722,
+      "learning_rate": 6.005615867060443e-07,
+      "loss": 0.7874,
       "step": 31450
     },
     {
-      "epoch": 0.8637774298975585,
+      "epoch": 0.8924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 9.576009079304805e-07,
-      "loss": 0.8616,
+      "learning_rate": 6.002479182875021e-07,
+      "loss": 0.8013,
       "step": 31451
     },
     {
-      "epoch": 0.8638048941254016,
+      "epoch": 0.8925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 9.572210982999875e-07,
-      "loss": 0.8882,
+      "learning_rate": 5.999343292690895e-07,
+      "loss": 0.6534,
       "step": 31452
     },
     {
-      "epoch": 0.8638323583532449,
+      "epoch": 0.8925368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 9.568413602194626e-07,
-      "loss": 0.8698,
+      "learning_rate": 5.996208196534514e-07,
+      "loss": 0.8272,
       "step": 31453
     },
     {
-      "epoch": 0.8638598225810882,
+      "epoch": 0.8925652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 9.564616936919123e-07,
-      "loss": 0.8266,
+      "learning_rate": 5.993073894432422e-07,
+      "loss": 0.7992,
       "step": 31454
     },
     {
-      "epoch": 0.8638872868089313,
+      "epoch": 0.8925936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 9.560820987203413e-07,
-      "loss": 0.7208,
+      "learning_rate": 5.98994038641103e-07,
+      "loss": 0.8354,
       "step": 31455
     },
     {
-      "epoch": 0.8639147510367746,
+      "epoch": 0.892622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 9.55702575307751e-07,
-      "loss": 0.8542,
+      "learning_rate": 5.986807672496841e-07,
+      "loss": 0.8789,
       "step": 31456
     },
     {
-      "epoch": 0.8639422152646178,
+      "epoch": 0.892650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 9.55323123457147e-07,
-      "loss": 0.8142,
+      "learning_rate": 5.983675752716334e-07,
+      "loss": 0.8335,
       "step": 31457
     },
     {
-      "epoch": 0.8639696794924611,
+      "epoch": 0.8926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 9.549437431715281e-07,
-      "loss": 0.7969,
+      "learning_rate": 5.98054462709593e-07,
+      "loss": 0.8789,
       "step": 31458
     },
     {
-      "epoch": 0.8639971437203043,
+      "epoch": 0.8927071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 9.545644344539007e-07,
-      "loss": 0.8054,
+      "learning_rate": 5.977414295662076e-07,
+      "loss": 0.859,
       "step": 31459
     },
     {
-      "epoch": 0.8640246079481475,
+      "epoch": 0.8927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 9.541851973072613e-07,
-      "loss": 0.6985,
+      "learning_rate": 5.974284758441251e-07,
+      "loss": 0.8472,
       "step": 31460
     },
     {
-      "epoch": 0.8640520721759908,
+      "epoch": 0.8927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 9.53806031734612e-07,
-      "loss": 0.7002,
+      "learning_rate": 5.971156015459867e-07,
+      "loss": 0.8396,
       "step": 31461
     },
     {
-      "epoch": 0.8640795364038341,
+      "epoch": 0.8927922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 9.534269377389538e-07,
-      "loss": 0.7775,
+      "learning_rate": 5.968028066744347e-07,
+      "loss": 0.7252,
       "step": 31462
     },
     {
-      "epoch": 0.8641070006316772,
+      "epoch": 0.8928206583427922,
       "grad_norm": 0.0,
-      "learning_rate": 9.530479153232885e-07,
-      "loss": 0.7724,
+      "learning_rate": 5.964900912321137e-07,
+      "loss": 0.8841,
       "step": 31463
     },
     {
-      "epoch": 0.8641344648595205,
+      "epoch": 0.8928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 9.526689644906095e-07,
-      "loss": 0.8536,
+      "learning_rate": 5.961774552216626e-07,
+      "loss": 0.826,
       "step": 31464
     },
     {
-      "epoch": 0.8641619290873637,
+      "epoch": 0.8928774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 9.522900852439187e-07,
-      "loss": 0.7788,
+      "learning_rate": 5.958648986457227e-07,
+      "loss": 0.8789,
       "step": 31465
     },
     {
-      "epoch": 0.8641893933152069,
+      "epoch": 0.8929057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 9.51911277586216e-07,
-      "loss": 0.9103,
+      "learning_rate": 5.955524215069364e-07,
+      "loss": 0.7832,
       "step": 31466
     },
     {
-      "epoch": 0.8642168575430502,
+      "epoch": 0.8929341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 9.515325415204946e-07,
-      "loss": 0.8131,
+      "learning_rate": 5.952400238079403e-07,
+      "loss": 0.7401,
       "step": 31467
     },
     {
-      "epoch": 0.8642443217708934,
+      "epoch": 0.8929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 9.511538770497519e-07,
-      "loss": 0.8479,
+      "learning_rate": 5.949277055513746e-07,
+      "loss": 0.8235,
       "step": 31468
     },
     {
-      "epoch": 0.8642717859987367,
+      "epoch": 0.8929909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 9.507752841769869e-07,
-      "loss": 0.8269,
+      "learning_rate": 5.946154667398785e-07,
+      "loss": 0.7889,
       "step": 31469
     },
     {
-      "epoch": 0.8642992502265798,
+      "epoch": 0.8930192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 9.503967629051913e-07,
-      "loss": 0.8848,
+      "learning_rate": 5.943033073760873e-07,
+      "loss": 0.7599,
       "step": 31470
     },
     {
-      "epoch": 0.8643267144544231,
+      "epoch": 0.8930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 9.50018313237362e-07,
-      "loss": 0.7744,
+      "learning_rate": 5.939912274626381e-07,
+      "loss": 0.778,
       "step": 31471
     },
     {
-      "epoch": 0.8643541786822664,
+      "epoch": 0.8930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 9.496399351764929e-07,
-      "loss": 0.8307,
+      "learning_rate": 5.936792270021696e-07,
+      "loss": 0.7632,
       "step": 31472
     },
     {
-      "epoch": 0.8643816429101095,
+      "epoch": 0.8931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 9.492616287255795e-07,
-      "loss": 0.7734,
+      "learning_rate": 5.933673059973144e-07,
+      "loss": 0.7655,
       "step": 31473
     },
     {
-      "epoch": 0.8644091071379528,
+      "epoch": 0.8931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 9.488833938876152e-07,
-      "loss": 0.7602,
+      "learning_rate": 5.93055464450708e-07,
+      "loss": 0.9046,
       "step": 31474
     },
     {
-      "epoch": 0.8644365713657961,
+      "epoch": 0.893161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 9.485052306655917e-07,
-      "loss": 0.7346,
+      "learning_rate": 5.927437023649851e-07,
+      "loss": 0.732,
       "step": 31475
     },
     {
-      "epoch": 0.8644640355936393,
+      "epoch": 0.8931895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 9.481271390624991e-07,
-      "loss": 0.8166,
+      "learning_rate": 5.924320197427791e-07,
+      "loss": 0.7824,
       "step": 31476
     },
     {
-      "epoch": 0.8644914998214825,
+      "epoch": 0.8932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 9.477491190813315e-07,
-      "loss": 0.7968,
+      "learning_rate": 5.921204165867244e-07,
+      "loss": 0.955,
       "step": 31477
     },
     {
-      "epoch": 0.8645189640493257,
+      "epoch": 0.8932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 9.473711707250788e-07,
-      "loss": 0.8663,
+      "learning_rate": 5.918088928994492e-07,
+      "loss": 0.7804,
       "step": 31478
     },
     {
-      "epoch": 0.864546428277169,
+      "epoch": 0.8932746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 9.469932939967319e-07,
-      "loss": 0.8288,
+      "learning_rate": 5.914974486835878e-07,
+      "loss": 0.9193,
       "step": 31479
     },
     {
-      "epoch": 0.8645738925050123,
+      "epoch": 0.8933030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 9.46615488899283e-07,
-      "loss": 0.8385,
+      "learning_rate": 5.911860839417727e-07,
+      "loss": 0.7713,
       "step": 31480
     },
     {
-      "epoch": 0.8646013567328554,
+      "epoch": 0.8933314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 9.462377554357171e-07,
-      "loss": 0.8558,
+      "learning_rate": 5.908747986766295e-07,
+      "loss": 0.754,
       "step": 31481
     },
     {
-      "epoch": 0.8646288209606987,
+      "epoch": 0.8933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 9.458600936090279e-07,
-      "loss": 0.8814,
+      "learning_rate": 5.905635928907894e-07,
+      "loss": 0.8356,
       "step": 31482
     },
     {
-      "epoch": 0.8646562851885419,
+      "epoch": 0.8933881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 9.454825034221981e-07,
-      "loss": 0.728,
+      "learning_rate": 5.902524665868836e-07,
+      "loss": 0.776,
       "step": 31483
     },
     {
-      "epoch": 0.8646837494163852,
+      "epoch": 0.8934165720771851,
       "grad_norm": 0.0,
-      "learning_rate": 9.451049848782179e-07,
-      "loss": 0.8694,
+      "learning_rate": 5.899414197675357e-07,
+      "loss": 0.7542,
       "step": 31484
     },
     {
-      "epoch": 0.8647112136442284,
+      "epoch": 0.8934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 9.447275379800758e-07,
-      "loss": 0.7649,
+      "learning_rate": 5.896304524353769e-07,
+      "loss": 0.8282,
       "step": 31485
     },
     {
-      "epoch": 0.8647386778720716,
+      "epoch": 0.8934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 9.443501627307583e-07,
-      "loss": 0.7764,
+      "learning_rate": 5.893195645930339e-07,
+      "loss": 0.8144,
       "step": 31486
     },
     {
-      "epoch": 0.8647661420999149,
+      "epoch": 0.8935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 9.439728591332475e-07,
-      "loss": 0.761,
+      "learning_rate": 5.890087562431301e-07,
+      "loss": 0.7604,
       "step": 31487
     },
     {
-      "epoch": 0.8647936063277581,
+      "epoch": 0.8935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 9.435956271905322e-07,
-      "loss": 0.8118,
+      "learning_rate": 5.886980273882925e-07,
+      "loss": 0.8391,
       "step": 31488
     },
     {
-      "epoch": 0.8648210705556013,
+      "epoch": 0.8935584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 9.432184669055966e-07,
-      "loss": 0.9008,
+      "learning_rate": 5.883873780311455e-07,
+      "loss": 0.7377,
       "step": 31489
     },
     {
-      "epoch": 0.8648485347834446,
+      "epoch": 0.8935868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 9.428413782814272e-07,
-      "loss": 0.8482,
+      "learning_rate": 5.880768081743127e-07,
+      "loss": 0.7544,
       "step": 31490
     },
     {
-      "epoch": 0.8648759990112878,
+      "epoch": 0.8936152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 9.424643613210016e-07,
-      "loss": 0.8383,
+      "learning_rate": 5.877663178204207e-07,
+      "loss": 0.7455,
       "step": 31491
     },
     {
-      "epoch": 0.864903463239131,
+      "epoch": 0.8936435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 9.420874160273096e-07,
-      "loss": 0.8586,
+      "learning_rate": 5.874559069720875e-07,
+      "loss": 0.8268,
       "step": 31492
     },
     {
-      "epoch": 0.8649309274669743,
+      "epoch": 0.893671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 9.417105424033291e-07,
-      "loss": 0.7921,
+      "learning_rate": 5.871455756319377e-07,
+      "loss": 0.8728,
       "step": 31493
     },
     {
-      "epoch": 0.8649583916948175,
+      "epoch": 0.893700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 9.413337404520428e-07,
-      "loss": 0.7963,
+      "learning_rate": 5.868353238025948e-07,
+      "loss": 0.6942,
       "step": 31494
     },
     {
-      "epoch": 0.8649858559226608,
+      "epoch": 0.8937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 9.409570101764331e-07,
-      "loss": 0.7643,
+      "learning_rate": 5.865251514866743e-07,
+      "loss": 0.8319,
       "step": 31495
     },
     {
-      "epoch": 0.8650133201505039,
+      "epoch": 0.8937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 9.405803515794809e-07,
-      "loss": 0.849,
+      "learning_rate": 5.862150586867998e-07,
+      "loss": 0.7099,
       "step": 31496
     },
     {
-      "epoch": 0.8650407843783472,
+      "epoch": 0.8937854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 9.40203764664166e-07,
-      "loss": 0.8297,
+      "learning_rate": 5.859050454055904e-07,
+      "loss": 0.9278,
       "step": 31497
     },
     {
-      "epoch": 0.8650682486061905,
+      "epoch": 0.8938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 9.398272494334703e-07,
-      "loss": 0.7856,
+      "learning_rate": 5.855951116456638e-07,
+      "loss": 0.8438,
       "step": 31498
     },
     {
-      "epoch": 0.8650957128340336,
+      "epoch": 0.8938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.394508058903717e-07,
-      "loss": 0.7993,
+      "learning_rate": 5.85285257409638e-07,
+      "loss": 0.7474,
       "step": 31499
     },
     {
-      "epoch": 0.8651231770618769,
+      "epoch": 0.8938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 9.390744340378455e-07,
-      "loss": 0.8694,
+      "learning_rate": 5.849754827001319e-07,
+      "loss": 0.8231,
       "step": 31500
     },
     {
-      "epoch": 0.8651506412897202,
+      "epoch": 0.8938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 9.386981338788714e-07,
-      "loss": 0.8628,
+      "learning_rate": 5.846657875197604e-07,
+      "loss": 0.8966,
       "step": 31501
     },
     {
-      "epoch": 0.8651781055175634,
+      "epoch": 0.8939273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 9.383219054164294e-07,
-      "loss": 0.8718,
+      "learning_rate": 5.843561718711399e-07,
+      "loss": 0.8755,
       "step": 31502
     },
     {
-      "epoch": 0.8652055697454066,
+      "epoch": 0.8939557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 9.379457486534937e-07,
-      "loss": 0.7903,
+      "learning_rate": 5.840466357568875e-07,
+      "loss": 0.8266,
       "step": 31503
     },
     {
-      "epoch": 0.8652330339732498,
+      "epoch": 0.8939841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 9.375696635930431e-07,
-      "loss": 0.8079,
+      "learning_rate": 5.837371791796153e-07,
+      "loss": 0.6825,
       "step": 31504
     },
     {
-      "epoch": 0.8652604982010931,
+      "epoch": 0.8940124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 9.371936502380507e-07,
-      "loss": 0.787,
+      "learning_rate": 5.834278021419382e-07,
+      "loss": 0.7987,
       "step": 31505
     },
     {
-      "epoch": 0.8652879624289364,
+      "epoch": 0.8940408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 9.368177085914943e-07,
-      "loss": 0.7588,
+      "learning_rate": 5.831185046464704e-07,
+      "loss": 0.746,
       "step": 31506
     },
     {
-      "epoch": 0.8653154266567795,
+      "epoch": 0.8940692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 9.364418386563445e-07,
-      "loss": 0.8467,
+      "learning_rate": 5.828092866958235e-07,
+      "loss": 0.7728,
       "step": 31507
     },
     {
-      "epoch": 0.8653428908846228,
+      "epoch": 0.8940976163450625,
       "grad_norm": 0.0,
-      "learning_rate": 9.360660404355781e-07,
-      "loss": 0.85,
+      "learning_rate": 5.825001482926107e-07,
+      "loss": 0.8542,
       "step": 31508
     },
     {
-      "epoch": 0.865370355112466,
+      "epoch": 0.8941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 9.356903139321682e-07,
-      "loss": 0.8215,
+      "learning_rate": 5.821910894394411e-07,
+      "loss": 0.754,
       "step": 31509
     },
     {
-      "epoch": 0.8653978193403092,
+      "epoch": 0.8941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 9.353146591490903e-07,
-      "loss": 0.7635,
+      "learning_rate": 5.818821101389272e-07,
+      "loss": 0.7634,
       "step": 31510
     },
     {
-      "epoch": 0.8654252835681525,
+      "epoch": 0.8941827468785472,
       "grad_norm": 0.0,
-      "learning_rate": 9.349390760893118e-07,
-      "loss": 0.7979,
+      "learning_rate": 5.815732103936799e-07,
+      "loss": 0.8348,
       "step": 31511
     },
     {
-      "epoch": 0.8654527477959957,
+      "epoch": 0.894211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 9.34563564755806e-07,
-      "loss": 0.8591,
+      "learning_rate": 5.812643902063065e-07,
+      "loss": 0.8706,
       "step": 31512
     },
     {
-      "epoch": 0.865480212023839,
+      "epoch": 0.8942395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 9.341881251515462e-07,
-      "loss": 0.8064,
+      "learning_rate": 5.809556495794145e-07,
+      "loss": 0.8256,
       "step": 31513
     },
     {
-      "epoch": 0.8655076762516822,
+      "epoch": 0.8942678774120317,
       "grad_norm": 0.0,
-      "learning_rate": 9.33812757279503e-07,
-      "loss": 0.8066,
+      "learning_rate": 5.806469885156163e-07,
+      "loss": 0.795,
       "step": 31514
     },
     {
-      "epoch": 0.8655351404795254,
+      "epoch": 0.8942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 9.334374611426433e-07,
-      "loss": 0.7559,
+      "learning_rate": 5.803384070175144e-07,
+      "loss": 0.7662,
       "step": 31515
     },
     {
-      "epoch": 0.8655626047073687,
+      "epoch": 0.8943246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 9.330622367439402e-07,
-      "loss": 0.6959,
+      "learning_rate": 5.800299050877178e-07,
+      "loss": 0.8062,
       "step": 31516
     },
     {
-      "epoch": 0.8655900689352118,
+      "epoch": 0.8943530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 9.326870840863589e-07,
-      "loss": 0.9093,
+      "learning_rate": 5.797214827288333e-07,
+      "loss": 0.7994,
       "step": 31517
     },
     {
-      "epoch": 0.8656175331630551,
+      "epoch": 0.8943813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 9.323120031728694e-07,
-      "loss": 0.7462,
+      "learning_rate": 5.794131399434632e-07,
+      "loss": 0.7703,
       "step": 31518
     },
     {
-      "epoch": 0.8656449973908984,
+      "epoch": 0.8944097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 9.319369940064405e-07,
-      "loss": 0.7799,
+      "learning_rate": 5.791048767342133e-07,
+      "loss": 0.6856,
       "step": 31519
     },
     {
-      "epoch": 0.8656724616187416,
+      "epoch": 0.8944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 9.315620565900385e-07,
-      "loss": 0.7743,
+      "learning_rate": 5.787966931036893e-07,
+      "loss": 0.8187,
       "step": 31520
     },
     {
-      "epoch": 0.8656999258465848,
+      "epoch": 0.8944665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 9.311871909266323e-07,
-      "loss": 0.7383,
+      "learning_rate": 5.784885890544922e-07,
+      "loss": 0.8024,
       "step": 31521
     },
     {
-      "epoch": 0.865727390074428,
+      "epoch": 0.8944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 9.30812397019184e-07,
-      "loss": 0.8274,
+      "learning_rate": 5.78180564589228e-07,
+      "loss": 0.6985,
       "step": 31522
     },
     {
-      "epoch": 0.8657548543022713,
+      "epoch": 0.8945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 9.304376748706634e-07,
-      "loss": 0.8195,
+      "learning_rate": 5.778726197104934e-07,
+      "loss": 0.8939,
       "step": 31523
     },
     {
-      "epoch": 0.8657823185301146,
+      "epoch": 0.8945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 9.300630244840314e-07,
-      "loss": 0.8497,
+      "learning_rate": 5.775647544208929e-07,
+      "loss": 0.7863,
       "step": 31524
     },
     {
-      "epoch": 0.8658097827579577,
+      "epoch": 0.8945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 9.296884458622535e-07,
-      "loss": 0.7302,
+      "learning_rate": 5.772569687230278e-07,
+      "loss": 0.7663,
       "step": 31525
     },
     {
-      "epoch": 0.865837246985801,
+      "epoch": 0.8946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 9.293139390082939e-07,
-      "loss": 0.868,
+      "learning_rate": 5.76949262619495e-07,
+      "loss": 0.8243,
       "step": 31526
     },
     {
-      "epoch": 0.8658647112136443,
+      "epoch": 0.8946367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 9.289395039251159e-07,
-      "loss": 0.8016,
+      "learning_rate": 5.766416361128957e-07,
+      "loss": 0.7805,
       "step": 31527
     },
     {
-      "epoch": 0.8658921754414874,
+      "epoch": 0.8946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 9.285651406156848e-07,
-      "loss": 0.8321,
+      "learning_rate": 5.7633408920583e-07,
+      "loss": 0.7956,
       "step": 31528
     },
     {
-      "epoch": 0.8659196396693307,
+      "epoch": 0.8946935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 9.281908490829583e-07,
-      "loss": 0.7689,
+      "learning_rate": 5.760266219008914e-07,
+      "loss": 0.812,
       "step": 31529
     },
     {
-      "epoch": 0.8659471038971739,
+      "epoch": 0.89472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 9.278166293299018e-07,
-      "loss": 0.7763,
+      "learning_rate": 5.757192342006812e-07,
+      "loss": 0.7884,
       "step": 31530
     },
     {
-      "epoch": 0.8659745681250172,
+      "epoch": 0.8947502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 9.274424813594718e-07,
-      "loss": 0.947,
+      "learning_rate": 5.754119261077939e-07,
+      "loss": 0.9029,
       "step": 31531
     },
     {
-      "epoch": 0.8660020323528604,
+      "epoch": 0.8947786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 9.270684051746326e-07,
-      "loss": 0.8563,
+      "learning_rate": 5.751046976248254e-07,
+      "loss": 0.775,
       "step": 31532
     },
     {
-      "epoch": 0.8660294965807036,
+      "epoch": 0.8948070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 9.26694400778343e-07,
-      "loss": 0.7853,
+      "learning_rate": 5.747975487543711e-07,
+      "loss": 0.8818,
       "step": 31533
     },
     {
-      "epoch": 0.8660569608085469,
+      "epoch": 0.8948354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 9.263204681735616e-07,
-      "loss": 0.691,
+      "learning_rate": 5.74490479499027e-07,
+      "loss": 0.7809,
       "step": 31534
     },
     {
-      "epoch": 0.86608442503639,
+      "epoch": 0.8948637911464246,
       "grad_norm": 0.0,
-      "learning_rate": 9.259466073632461e-07,
-      "loss": 0.8252,
+      "learning_rate": 5.741834898613819e-07,
+      "loss": 0.8957,
       "step": 31535
     },
     {
-      "epoch": 0.8661118892642333,
+      "epoch": 0.8948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 9.255728183503565e-07,
-      "loss": 0.7614,
+      "learning_rate": 5.73876579844036e-07,
+      "loss": 0.7834,
       "step": 31536
     },
     {
-      "epoch": 0.8661393534920766,
+      "epoch": 0.8949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 9.251991011378502e-07,
-      "loss": 0.8661,
+      "learning_rate": 5.735697494495773e-07,
+      "loss": 0.7903,
       "step": 31537
     },
     {
-      "epoch": 0.8661668177199198,
+      "epoch": 0.8949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 9.248254557286862e-07,
-      "loss": 0.7967,
+      "learning_rate": 5.732629986805982e-07,
+      "loss": 0.8334,
       "step": 31538
     },
     {
-      "epoch": 0.866194281947763,
+      "epoch": 0.8949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 9.244518821258175e-07,
-      "loss": 0.8119,
+      "learning_rate": 5.72956327539692e-07,
+      "loss": 0.8523,
       "step": 31539
     },
     {
-      "epoch": 0.8662217461756063,
+      "epoch": 0.895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 9.240783803321995e-07,
-      "loss": 0.7137,
+      "learning_rate": 5.726497360294459e-07,
+      "loss": 0.8324,
       "step": 31540
     },
     {
-      "epoch": 0.8662492104034495,
+      "epoch": 0.8950340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 9.237049503507889e-07,
-      "loss": 0.7987,
+      "learning_rate": 5.723432241524507e-07,
+      "loss": 0.8095,
       "step": 31541
     },
     {
-      "epoch": 0.8662766746312928,
+      "epoch": 0.8950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 9.233315921845409e-07,
-      "loss": 0.8064,
+      "learning_rate": 5.720367919112979e-07,
+      "loss": 0.7378,
       "step": 31542
     },
     {
-      "epoch": 0.8663041388591359,
+      "epoch": 0.8950908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 9.229583058364089e-07,
-      "loss": 0.8912,
+      "learning_rate": 5.717304393085732e-07,
+      "loss": 0.7745,
       "step": 31543
     },
     {
-      "epoch": 0.8663316030869792,
+      "epoch": 0.8951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 9.225850913093482e-07,
-      "loss": 0.8932,
+      "learning_rate": 5.714241663468645e-07,
+      "loss": 0.6891,
       "step": 31544
     },
     {
-      "epoch": 0.8663590673148225,
+      "epoch": 0.8951475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 9.222119486063108e-07,
-      "loss": 0.7549,
+      "learning_rate": 5.71117973028762e-07,
+      "loss": 0.8222,
       "step": 31545
     },
     {
-      "epoch": 0.8663865315426656,
+      "epoch": 0.8951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 9.218388777302479e-07,
-      "loss": 0.7921,
+      "learning_rate": 5.708118593568479e-07,
+      "loss": 0.7387,
       "step": 31546
     },
     {
-      "epoch": 0.8664139957705089,
+      "epoch": 0.8952043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 9.214658786841136e-07,
-      "loss": 0.8361,
+      "learning_rate": 5.705058253337104e-07,
+      "loss": 0.8065,
       "step": 31547
     },
     {
-      "epoch": 0.8664414599983521,
+      "epoch": 0.8952326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 9.210929514708566e-07,
-      "loss": 0.756,
+      "learning_rate": 5.70199870961935e-07,
+      "loss": 0.7835,
       "step": 31548
     },
     {
-      "epoch": 0.8664689242261954,
+      "epoch": 0.895261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 9.207200960934281e-07,
-      "loss": 0.8186,
+      "learning_rate": 5.698939962441042e-07,
+      "loss": 0.7852,
       "step": 31549
     },
     {
-      "epoch": 0.8664963884540386,
+      "epoch": 0.8952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 9.203473125547801e-07,
-      "loss": 0.7896,
+      "learning_rate": 5.695882011828024e-07,
+      "loss": 0.6659,
       "step": 31550
     },
     {
-      "epoch": 0.8665238526818818,
+      "epoch": 0.8953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 9.199746008578625e-07,
-      "loss": 0.7667,
+      "learning_rate": 5.692824857806134e-07,
+      "loss": 0.8405,
       "step": 31551
     },
     {
-      "epoch": 0.8665513169097251,
+      "epoch": 0.8953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 9.196019610056217e-07,
-      "loss": 0.7799,
+      "learning_rate": 5.689768500401183e-07,
+      "loss": 0.8312,
       "step": 31552
     },
     {
-      "epoch": 0.8665787811375684,
+      "epoch": 0.8953745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 9.192293930010077e-07,
-      "loss": 0.8592,
+      "learning_rate": 5.686712939639005e-07,
+      "loss": 0.838,
       "step": 31553
     },
     {
-      "epoch": 0.8666062453654115,
+      "epoch": 0.8954029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 9.188568968469702e-07,
-      "loss": 0.8158,
+      "learning_rate": 5.683658175545415e-07,
+      "loss": 0.6995,
       "step": 31554
     },
     {
-      "epoch": 0.8666337095932548,
+      "epoch": 0.8954313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 9.184844725464526e-07,
-      "loss": 0.8537,
+      "learning_rate": 5.68060420814619e-07,
+      "loss": 0.7883,
       "step": 31555
     },
     {
-      "epoch": 0.866661173821098,
+      "epoch": 0.8954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 9.181121201024035e-07,
-      "loss": 0.7643,
+      "learning_rate": 5.677551037467133e-07,
+      "loss": 0.7626,
       "step": 31556
     },
     {
-      "epoch": 0.8666886380489413,
+      "epoch": 0.8954880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 9.177398395177717e-07,
-      "loss": 0.887,
+      "learning_rate": 5.674498663534067e-07,
+      "loss": 0.7356,
       "step": 31557
     },
     {
-      "epoch": 0.8667161022767845,
+      "epoch": 0.895516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 9.17367630795497e-07,
-      "loss": 0.7548,
+      "learning_rate": 5.671447086372728e-07,
+      "loss": 0.7855,
       "step": 31558
     },
     {
-      "epoch": 0.8667435665046277,
+      "epoch": 0.895544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 9.169954939385295e-07,
-      "loss": 0.8509,
+      "learning_rate": 5.668396306008928e-07,
+      "loss": 0.9352,
       "step": 31559
     },
     {
-      "epoch": 0.866771030732471,
+      "epoch": 0.8955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 9.166234289498121e-07,
-      "loss": 0.8366,
+      "learning_rate": 5.665346322468435e-07,
+      "loss": 0.8099,
       "step": 31560
     },
     {
-      "epoch": 0.8667984949603141,
+      "epoch": 0.8956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 9.162514358322883e-07,
-      "loss": 0.7791,
+      "learning_rate": 5.662297135776984e-07,
+      "loss": 0.7784,
       "step": 31561
     },
     {
-      "epoch": 0.8668259591881574,
+      "epoch": 0.8956299659477867,
       "grad_norm": 0.0,
-      "learning_rate": 9.158795145889033e-07,
-      "loss": 0.8109,
+      "learning_rate": 5.659248745960366e-07,
+      "loss": 0.7205,
       "step": 31562
     },
     {
-      "epoch": 0.8668534234160007,
+      "epoch": 0.8956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 9.155076652225981e-07,
-      "loss": 0.8709,
+      "learning_rate": 5.656201153044294e-07,
+      "loss": 0.8396,
       "step": 31563
     },
     {
-      "epoch": 0.8668808876438439,
+      "epoch": 0.8956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 9.151358877363148e-07,
-      "loss": 0.8676,
+      "learning_rate": 5.653154357054547e-07,
+      "loss": 0.8832,
       "step": 31564
     },
     {
-      "epoch": 0.8669083518716871,
+      "epoch": 0.8957150964812712,
       "grad_norm": 0.0,
-      "learning_rate": 9.147641821329944e-07,
-      "loss": 0.7848,
+      "learning_rate": 5.650108358016848e-07,
+      "loss": 0.7209,
       "step": 31565
     },
     {
-      "epoch": 0.8669358160995304,
+      "epoch": 0.8957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 9.143925484155791e-07,
-      "loss": 0.8031,
+      "learning_rate": 5.6470631559569e-07,
+      "loss": 0.7281,
       "step": 31566
     },
     {
-      "epoch": 0.8669632803273736,
+      "epoch": 0.8957718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 9.140209865870098e-07,
-      "loss": 0.7691,
+      "learning_rate": 5.644018750900459e-07,
+      "loss": 0.734,
       "step": 31567
     },
     {
-      "epoch": 0.8669907445552169,
+      "epoch": 0.8958002270147559,
       "grad_norm": 0.0,
-      "learning_rate": 9.136494966502263e-07,
-      "loss": 0.838,
+      "learning_rate": 5.640975142873262e-07,
+      "loss": 0.8519,
       "step": 31568
     },
     {
-      "epoch": 0.86701820878306,
+      "epoch": 0.8958286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 9.132780786081685e-07,
-      "loss": 0.7158,
+      "learning_rate": 5.637932331900963e-07,
+      "loss": 0.845,
       "step": 31569
     },
     {
-      "epoch": 0.8670456730109033,
+      "epoch": 0.8958569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 9.129067324637742e-07,
-      "loss": 0.8073,
+      "learning_rate": 5.634890318009301e-07,
+      "loss": 0.7561,
       "step": 31570
     },
     {
-      "epoch": 0.8670731372387466,
+      "epoch": 0.8958853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 9.125354582199819e-07,
-      "loss": 0.7168,
+      "learning_rate": 5.631849101223963e-07,
+      "loss": 0.926,
       "step": 31571
     },
     {
-      "epoch": 0.8671006014665897,
+      "epoch": 0.8959137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 9.121642558797284e-07,
-      "loss": 0.8002,
+      "learning_rate": 5.628808681570641e-07,
+      "loss": 0.648,
       "step": 31572
     },
     {
-      "epoch": 0.867128065694433,
+      "epoch": 0.8959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 9.117931254459511e-07,
-      "loss": 0.8959,
+      "learning_rate": 5.625769059075004e-07,
+      "loss": 0.7742,
       "step": 31573
     },
     {
-      "epoch": 0.8671555299222762,
+      "epoch": 0.8959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 9.114220669215879e-07,
-      "loss": 0.7742,
+      "learning_rate": 5.622730233762752e-07,
+      "loss": 0.697,
       "step": 31574
     },
     {
-      "epoch": 0.8671829941501195,
+      "epoch": 0.8959988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 9.110510803095751e-07,
-      "loss": 0.8762,
+      "learning_rate": 5.619692205659532e-07,
+      "loss": 0.8878,
       "step": 31575
     },
     {
-      "epoch": 0.8672104583779627,
+      "epoch": 0.8960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 9.10680165612845e-07,
-      "loss": 0.6947,
+      "learning_rate": 5.616654974791003e-07,
+      "loss": 0.7861,
       "step": 31576
     },
     {
-      "epoch": 0.8672379226058059,
+      "epoch": 0.89605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 9.10309322834334e-07,
-      "loss": 0.7372,
+      "learning_rate": 5.613618541182852e-07,
+      "loss": 0.6677,
       "step": 31577
     },
     {
-      "epoch": 0.8672653868336492,
+      "epoch": 0.8960839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 9.099385519769788e-07,
-      "loss": 0.69,
+      "learning_rate": 5.610582904860706e-07,
+      "loss": 0.8866,
       "step": 31578
     },
     {
-      "epoch": 0.8672928510614925,
+      "epoch": 0.8961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 9.095678530437091e-07,
-      "loss": 0.856,
+      "learning_rate": 5.607548065850199e-07,
+      "loss": 0.7847,
       "step": 31579
     },
     {
-      "epoch": 0.8673203152893356,
+      "epoch": 0.8961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 9.091972260374604e-07,
-      "loss": 0.7535,
+      "learning_rate": 5.604514024176988e-07,
+      "loss": 0.8867,
       "step": 31580
     },
     {
-      "epoch": 0.8673477795171789,
+      "epoch": 0.8961691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 9.088266709611659e-07,
-      "loss": 0.741,
+      "learning_rate": 5.601480779866664e-07,
+      "loss": 0.7378,
       "step": 31581
     },
     {
-      "epoch": 0.8673752437450221,
+      "epoch": 0.8961975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 9.084561878177545e-07,
-      "loss": 0.7997,
+      "learning_rate": 5.598448332944906e-07,
+      "loss": 0.7594,
       "step": 31582
     },
     {
-      "epoch": 0.8674027079728653,
+      "epoch": 0.8962258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 9.080857766101591e-07,
-      "loss": 0.8428,
+      "learning_rate": 5.595416683437283e-07,
+      "loss": 0.7738,
       "step": 31583
     },
     {
-      "epoch": 0.8674301722007086,
+      "epoch": 0.8962542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 9.077154373413111e-07,
-      "loss": 0.8573,
+      "learning_rate": 5.592385831369418e-07,
+      "loss": 0.8179,
       "step": 31584
     },
     {
-      "epoch": 0.8674576364285518,
+      "epoch": 0.8962826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 9.073451700141411e-07,
-      "loss": 0.7501,
+      "learning_rate": 5.589355776766924e-07,
+      "loss": 0.8816,
       "step": 31585
     },
     {
-      "epoch": 0.8674851006563951,
+      "epoch": 0.8963110102156641,
       "grad_norm": 0.0,
-      "learning_rate": 9.069749746315804e-07,
-      "loss": 0.743,
+      "learning_rate": 5.586326519655383e-07,
+      "loss": 0.8268,
       "step": 31586
     },
     {
-      "epoch": 0.8675125648842382,
+      "epoch": 0.8963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 9.066048511965553e-07,
-      "loss": 0.8555,
+      "learning_rate": 5.583298060060382e-07,
+      "loss": 0.7962,
       "step": 31587
     },
     {
-      "epoch": 0.8675400291120815,
+      "epoch": 0.8963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 9.062347997119936e-07,
-      "loss": 0.7721,
+      "learning_rate": 5.580270398007514e-07,
+      "loss": 0.862,
       "step": 31588
     },
     {
-      "epoch": 0.8675674933399248,
+      "epoch": 0.8963961407491486,
       "grad_norm": 0.0,
-      "learning_rate": 9.05864820180824e-07,
-      "loss": 0.7599,
+      "learning_rate": 5.577243533522336e-07,
+      "loss": 0.6994,
       "step": 31589
     },
     {
-      "epoch": 0.8675949575677679,
+      "epoch": 0.8964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 9.054949126059753e-07,
-      "loss": 0.8781,
+      "learning_rate": 5.574217466630427e-07,
+      "loss": 0.7801,
       "step": 31590
     },
     {
-      "epoch": 0.8676224217956112,
+      "epoch": 0.8964528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 9.051250769903741e-07,
-      "loss": 0.7759,
+      "learning_rate": 5.571192197357367e-07,
+      "loss": 0.8305,
       "step": 31591
     },
     {
-      "epoch": 0.8676498860234545,
+      "epoch": 0.8964812712826333,
       "grad_norm": 0.0,
-      "learning_rate": 9.047553133369468e-07,
-      "loss": 0.7217,
+      "learning_rate": 5.568167725728679e-07,
+      "loss": 0.7643,
       "step": 31592
     },
     {
-      "epoch": 0.8676773502512977,
+      "epoch": 0.8965096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 9.04385621648618e-07,
-      "loss": 0.7961,
+      "learning_rate": 5.565144051769911e-07,
+      "loss": 0.7723,
       "step": 31593
     },
     {
-      "epoch": 0.8677048144791409,
+      "epoch": 0.8965380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 9.040160019283128e-07,
-      "loss": 0.7937,
+      "learning_rate": 5.56212117550664e-07,
+      "loss": 0.7605,
       "step": 31594
     },
     {
-      "epoch": 0.8677322787069841,
+      "epoch": 0.896566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 9.036464541789591e-07,
-      "loss": 0.8434,
+      "learning_rate": 5.55909909696436e-07,
+      "loss": 0.8982,
       "step": 31595
     },
     {
-      "epoch": 0.8677597429348274,
+      "epoch": 0.896594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 9.032769784034756e-07,
-      "loss": 0.7843,
+      "learning_rate": 5.556077816168615e-07,
+      "loss": 0.7858,
       "step": 31596
     },
     {
-      "epoch": 0.8677872071626707,
+      "epoch": 0.8966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 9.029075746047888e-07,
-      "loss": 0.8594,
+      "learning_rate": 5.553057333144917e-07,
+      "loss": 0.8505,
       "step": 31597
     },
     {
-      "epoch": 0.8678146713905138,
+      "epoch": 0.8966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 9.025382427858231e-07,
-      "loss": 0.7639,
+      "learning_rate": 5.550037647918805e-07,
+      "loss": 0.83,
       "step": 31598
     },
     {
-      "epoch": 0.8678421356183571,
+      "epoch": 0.8966799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 9.021689829494973e-07,
-      "loss": 0.801,
+      "learning_rate": 5.547018760515776e-07,
+      "loss": 0.7691,
       "step": 31599
     },
     {
-      "epoch": 0.8678695998462004,
+      "epoch": 0.8967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 9.017997950987334e-07,
-      "loss": 0.7716,
+      "learning_rate": 5.544000670961313e-07,
+      "loss": 0.7297,
       "step": 31600
     },
     {
-      "epoch": 0.8678970640740435,
+      "epoch": 0.8967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 9.014306792364546e-07,
-      "loss": 0.8445,
+      "learning_rate": 5.540983379280929e-07,
+      "loss": 0.8657,
       "step": 31601
     },
     {
-      "epoch": 0.8679245283018868,
+      "epoch": 0.8967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 9.010616353655832e-07,
-      "loss": 0.8562,
+      "learning_rate": 5.537966885500113e-07,
+      "loss": 0.8685,
       "step": 31602
     },
     {
-      "epoch": 0.86795199252973,
+      "epoch": 0.8967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 9.006926634890334e-07,
-      "loss": 0.8289,
+      "learning_rate": 5.534951189644322e-07,
+      "loss": 0.7205,
       "step": 31603
     },
     {
-      "epoch": 0.8679794567575733,
+      "epoch": 0.8968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 9.003237636097306e-07,
-      "loss": 0.8183,
+      "learning_rate": 5.531936291739037e-07,
+      "loss": 0.7689,
       "step": 31604
     },
     {
-      "epoch": 0.8680069209854165,
+      "epoch": 0.896850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 8.999549357305893e-07,
-      "loss": 0.8257,
+      "learning_rate": 5.52892219180976e-07,
+      "loss": 0.8004,
       "step": 31605
     },
     {
-      "epoch": 0.8680343852132597,
+      "epoch": 0.8968785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 8.995861798545302e-07,
-      "loss": 0.7893,
+      "learning_rate": 5.525908889881903e-07,
+      "loss": 0.9629,
       "step": 31606
     },
     {
-      "epoch": 0.868061849441103,
+      "epoch": 0.8969069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 8.992174959844691e-07,
-      "loss": 0.7078,
+      "learning_rate": 5.522896385980947e-07,
+      "loss": 0.8109,
       "step": 31607
     },
     {
-      "epoch": 0.8680893136689461,
+      "epoch": 0.8969353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 8.988488841233256e-07,
-      "loss": 0.7901,
+      "learning_rate": 5.519884680132337e-07,
+      "loss": 0.7878,
       "step": 31608
     },
     {
-      "epoch": 0.8681167778967894,
+      "epoch": 0.8969636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 8.984803442740142e-07,
-      "loss": 0.7755,
+      "learning_rate": 5.516873772361497e-07,
+      "loss": 0.733,
       "step": 31609
     },
     {
-      "epoch": 0.8681442421246327,
+      "epoch": 0.8969920544835415,
       "grad_norm": 0.0,
-      "learning_rate": 8.981118764394548e-07,
-      "loss": 0.8311,
+      "learning_rate": 5.513863662693875e-07,
+      "loss": 0.7658,
       "step": 31610
     },
     {
-      "epoch": 0.8681717063524759,
+      "epoch": 0.8970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 8.977434806225593e-07,
-      "loss": 0.8009,
+      "learning_rate": 5.510854351154915e-07,
+      "loss": 0.81,
       "step": 31611
     },
     {
-      "epoch": 0.8681991705803191,
+      "epoch": 0.8970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 8.973751568262412e-07,
-      "loss": 0.7494,
+      "learning_rate": 5.507845837769976e-07,
+      "loss": 0.7788,
       "step": 31612
     },
     {
-      "epoch": 0.8682266348081624,
+      "epoch": 0.8970771850170262,
       "grad_norm": 0.0,
-      "learning_rate": 8.97006905053418e-07,
-      "loss": 0.7815,
+      "learning_rate": 5.504838122564537e-07,
+      "loss": 0.7768,
       "step": 31613
     },
     {
-      "epoch": 0.8682540990360056,
+      "epoch": 0.897105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.966387253070008e-07,
-      "loss": 0.7918,
+      "learning_rate": 5.501831205563978e-07,
+      "loss": 0.811,
       "step": 31614
     },
     {
-      "epoch": 0.8682815632638489,
+      "epoch": 0.8971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 8.96270617589905e-07,
-      "loss": 0.774,
+      "learning_rate": 5.498825086793691e-07,
+      "loss": 0.8572,
       "step": 31615
     },
     {
-      "epoch": 0.868309027491692,
+      "epoch": 0.8971623155505107,
       "grad_norm": 0.0,
-      "learning_rate": 8.959025819050449e-07,
-      "loss": 0.8191,
+      "learning_rate": 5.495819766279086e-07,
+      "loss": 0.8206,
       "step": 31616
     },
     {
-      "epoch": 0.8683364917195353,
+      "epoch": 0.8971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 8.955346182553282e-07,
-      "loss": 0.7536,
+      "learning_rate": 5.492815244045535e-07,
+      "loss": 0.8046,
       "step": 31617
     },
     {
-      "epoch": 0.8683639559473786,
+      "epoch": 0.8972190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 8.951667266436693e-07,
-      "loss": 0.7735,
+      "learning_rate": 5.489811520118415e-07,
+      "loss": 0.7567,
       "step": 31618
     },
     {
-      "epoch": 0.8683914201752218,
+      "epoch": 0.8972474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 8.947989070729768e-07,
-      "loss": 0.7874,
+      "learning_rate": 5.486808594523118e-07,
+      "loss": 0.7883,
       "step": 31619
     },
     {
-      "epoch": 0.868418884403065,
+      "epoch": 0.8972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 8.944311595461619e-07,
-      "loss": 0.7886,
+      "learning_rate": 5.483806467284992e-07,
+      "loss": 0.7143,
       "step": 31620
     },
     {
-      "epoch": 0.8684463486309082,
+      "epoch": 0.8973041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 8.940634840661355e-07,
-      "loss": 0.8416,
+      "learning_rate": 5.480805138429391e-07,
+      "loss": 0.7888,
       "step": 31621
     },
     {
-      "epoch": 0.8684738128587515,
+      "epoch": 0.8973325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 8.936958806358076e-07,
-      "loss": 0.8064,
+      "learning_rate": 5.477804607981707e-07,
+      "loss": 0.7882,
       "step": 31622
     },
     {
-      "epoch": 0.8685012770865947,
+      "epoch": 0.897360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 8.933283492580846e-07,
-      "loss": 0.7865,
+      "learning_rate": 5.474804875967233e-07,
+      "loss": 0.8201,
       "step": 31623
     },
     {
-      "epoch": 0.8685287413144379,
+      "epoch": 0.89738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 8.929608899358744e-07,
-      "loss": 0.8261,
+      "learning_rate": 5.471805942411323e-07,
+      "loss": 0.7135,
       "step": 31624
     },
     {
-      "epoch": 0.8685562055422812,
+      "epoch": 0.8974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 8.925935026720844e-07,
-      "loss": 0.865,
+      "learning_rate": 5.468807807339349e-07,
+      "loss": 0.7929,
       "step": 31625
     },
     {
-      "epoch": 0.8685836697701245,
+      "epoch": 0.8974460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 8.922261874696259e-07,
-      "loss": 0.7714,
+      "learning_rate": 5.465810470776578e-07,
+      "loss": 0.8775,
       "step": 31626
     },
     {
-      "epoch": 0.8686111339979676,
+      "epoch": 0.8974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 8.918589443313985e-07,
-      "loss": 0.7244,
+      "learning_rate": 5.462813932748356e-07,
+      "loss": 0.8806,
       "step": 31627
     },
     {
-      "epoch": 0.8686385982258109,
+      "epoch": 0.8975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 8.914917732603146e-07,
-      "loss": 0.8202,
+      "learning_rate": 5.459818193279998e-07,
+      "loss": 0.8019,
       "step": 31628
     },
     {
-      "epoch": 0.8686660624536541,
+      "epoch": 0.8975312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 8.911246742592739e-07,
-      "loss": 0.8427,
+      "learning_rate": 5.456823252396815e-07,
+      "loss": 0.8277,
       "step": 31629
     },
     {
-      "epoch": 0.8686935266814974,
+      "epoch": 0.8975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 8.90757647331183e-07,
-      "loss": 0.9003,
+      "learning_rate": 5.453829110124098e-07,
+      "loss": 0.8415,
       "step": 31630
     },
     {
-      "epoch": 0.8687209909093406,
+      "epoch": 0.8975879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 8.903906924789451e-07,
-      "loss": 0.8175,
+      "learning_rate": 5.450835766487128e-07,
+      "loss": 0.8308,
       "step": 31631
     },
     {
-      "epoch": 0.8687484551371838,
+      "epoch": 0.897616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 8.900238097054659e-07,
-      "loss": 0.7628,
+      "learning_rate": 5.447843221511195e-07,
+      "loss": 0.8545,
       "step": 31632
     },
     {
-      "epoch": 0.8687759193650271,
+      "epoch": 0.8976447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 8.896569990136472e-07,
-      "loss": 0.802,
+      "learning_rate": 5.444851475221591e-07,
+      "loss": 0.8553,
       "step": 31633
     },
     {
-      "epoch": 0.8688033835928702,
+      "epoch": 0.8976730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 8.892902604063935e-07,
-      "loss": 0.7999,
+      "learning_rate": 5.441860527643572e-07,
+      "loss": 0.8068,
       "step": 31634
     },
     {
-      "epoch": 0.8688308478207135,
+      "epoch": 0.8977014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 8.889235938866037e-07,
-      "loss": 0.7891,
+      "learning_rate": 5.438870378802397e-07,
+      "loss": 0.6872,
       "step": 31635
     },
     {
-      "epoch": 0.8688583120485568,
+      "epoch": 0.8977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 8.885569994571775e-07,
-      "loss": 0.8387,
+      "learning_rate": 5.435881028723355e-07,
+      "loss": 0.87,
       "step": 31636
     },
     {
-      "epoch": 0.8688857762764,
+      "epoch": 0.8977582292849036,
       "grad_norm": 0.0,
-      "learning_rate": 8.881904771210192e-07,
-      "loss": 0.6938,
+      "learning_rate": 5.432892477431672e-07,
+      "loss": 0.8286,
       "step": 31637
     },
     {
-      "epoch": 0.8689132405042432,
+      "epoch": 0.8977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 8.878240268810257e-07,
-      "loss": 0.7171,
+      "learning_rate": 5.429904724952583e-07,
+      "loss": 0.7277,
       "step": 31638
     },
     {
-      "epoch": 0.8689407047320865,
+      "epoch": 0.8978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 8.874576487400999e-07,
-      "loss": 0.8348,
+      "learning_rate": 5.426917771311368e-07,
+      "loss": 0.7288,
       "step": 31639
     },
     {
-      "epoch": 0.8689681689599297,
+      "epoch": 0.8978433598183881,
       "grad_norm": 0.0,
-      "learning_rate": 8.870913427011397e-07,
-      "loss": 0.7882,
+      "learning_rate": 5.423931616533207e-07,
+      "loss": 0.78,
       "step": 31640
     },
     {
-      "epoch": 0.868995633187773,
+      "epoch": 0.8978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 8.867251087670414e-07,
-      "loss": 0.7753,
+      "learning_rate": 5.420946260643345e-07,
+      "loss": 0.8838,
       "step": 31641
     },
     {
-      "epoch": 0.8690230974156161,
+      "epoch": 0.897900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 8.863589469407064e-07,
-      "loss": 0.8436,
+      "learning_rate": 5.417961703667019e-07,
+      "loss": 0.8189,
       "step": 31642
     },
     {
-      "epoch": 0.8690505616434594,
+      "epoch": 0.8979284903518728,
       "grad_norm": 0.0,
-      "learning_rate": 8.859928572250265e-07,
-      "loss": 0.8368,
+      "learning_rate": 5.414977945629396e-07,
+      "loss": 0.7925,
       "step": 31643
     },
     {
-      "epoch": 0.8690780258713027,
+      "epoch": 0.8979568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 8.856268396229028e-07,
-      "loss": 0.7814,
+      "learning_rate": 5.411994986555725e-07,
+      "loss": 0.8051,
       "step": 31644
     },
     {
-      "epoch": 0.8691054900991458,
+      "epoch": 0.8979852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 8.852608941372299e-07,
-      "loss": 0.7753,
+      "learning_rate": 5.409012826471183e-07,
+      "loss": 0.7658,
       "step": 31645
     },
     {
-      "epoch": 0.8691329543269891,
+      "epoch": 0.8980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 8.84895020770905e-07,
-      "loss": 0.7509,
+      "learning_rate": 5.406031465400951e-07,
+      "loss": 0.8632,
       "step": 31646
     },
     {
-      "epoch": 0.8691604185548323,
+      "epoch": 0.8980419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 8.845292195268196e-07,
-      "loss": 0.8729,
+      "learning_rate": 5.403050903370244e-07,
+      "loss": 0.8527,
       "step": 31647
     },
     {
-      "epoch": 0.8691878827826756,
+      "epoch": 0.8980703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 8.841634904078711e-07,
-      "loss": 0.7791,
+      "learning_rate": 5.400071140404206e-07,
+      "loss": 0.756,
       "step": 31648
     },
     {
-      "epoch": 0.8692153470105188,
+      "epoch": 0.8980987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 8.837978334169516e-07,
-      "loss": 0.8743,
+      "learning_rate": 5.397092176528018e-07,
+      "loss": 0.7878,
       "step": 31649
     },
     {
-      "epoch": 0.869242811238362,
+      "epoch": 0.8981271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 8.834322485569558e-07,
-      "loss": 0.7693,
+      "learning_rate": 5.394114011766849e-07,
+      "loss": 0.7769,
       "step": 31650
     },
     {
-      "epoch": 0.8692702754662053,
+      "epoch": 0.898155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 8.830667358307765e-07,
-      "loss": 0.8766,
+      "learning_rate": 5.391136646145856e-07,
+      "loss": 0.7559,
       "step": 31651
     },
     {
-      "epoch": 0.8692977396940486,
+      "epoch": 0.8981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 8.827012952413028e-07,
-      "loss": 0.8415,
+      "learning_rate": 5.388160079690174e-07,
+      "loss": 0.8505,
       "step": 31652
     },
     {
-      "epoch": 0.8693252039218917,
+      "epoch": 0.8982122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 8.823359267914278e-07,
-      "loss": 0.8526,
+      "learning_rate": 5.385184312424973e-07,
+      "loss": 0.8669,
       "step": 31653
     },
     {
-      "epoch": 0.869352668149735,
+      "epoch": 0.8982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 8.819706304840436e-07,
-      "loss": 0.8403,
+      "learning_rate": 5.382209344375355e-07,
+      "loss": 0.8947,
       "step": 31654
     },
     {
-      "epoch": 0.8693801323775782,
+      "epoch": 0.8982690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 8.816054063220391e-07,
-      "loss": 0.8014,
+      "learning_rate": 5.379235175566466e-07,
+      "loss": 0.8483,
       "step": 31655
     },
     {
-      "epoch": 0.8694075966054214,
+      "epoch": 0.8982973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 8.812402543083043e-07,
-      "loss": 0.8396,
+      "learning_rate": 5.376261806023453e-07,
+      "loss": 0.7451,
       "step": 31656
     },
     {
-      "epoch": 0.8694350608332647,
+      "epoch": 0.8983257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 8.808751744457322e-07,
-      "loss": 0.8067,
+      "learning_rate": 5.373289235771395e-07,
+      "loss": 0.7599,
       "step": 31657
     },
     {
-      "epoch": 0.8694625250611079,
+      "epoch": 0.8983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 8.805101667372052e-07,
-      "loss": 0.8683,
+      "learning_rate": 5.370317464835406e-07,
+      "loss": 0.8244,
       "step": 31658
     },
     {
-      "epoch": 0.8694899892889512,
+      "epoch": 0.8983825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 8.801452311856163e-07,
-      "loss": 0.7187,
+      "learning_rate": 5.367346493240622e-07,
+      "loss": 0.8499,
       "step": 31659
     },
     {
-      "epoch": 0.8695174535167943,
+      "epoch": 0.898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 8.797803677938488e-07,
-      "loss": 0.8392,
+      "learning_rate": 5.364376321012099e-07,
+      "loss": 0.7989,
       "step": 31660
     },
     {
-      "epoch": 0.8695449177446376,
+      "epoch": 0.898439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 8.794155765647927e-07,
-      "loss": 0.7604,
+      "learning_rate": 5.361406948174952e-07,
+      "loss": 0.8628,
       "step": 31661
     },
     {
-      "epoch": 0.8695723819724809,
+      "epoch": 0.8984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 8.790508575013323e-07,
-      "loss": 0.8794,
+      "learning_rate": 5.35843837475426e-07,
+      "loss": 0.7917,
       "step": 31662
     },
     {
-      "epoch": 0.869599846200324,
+      "epoch": 0.8984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 8.786862106063576e-07,
-      "loss": 0.843,
+      "learning_rate": 5.355470600775081e-07,
+      "loss": 0.721,
       "step": 31663
     },
     {
-      "epoch": 0.8696273104281673,
+      "epoch": 0.8985244040862657,
       "grad_norm": 0.0,
-      "learning_rate": 8.783216358827485e-07,
-      "loss": 0.8208,
+      "learning_rate": 5.352503626262507e-07,
+      "loss": 0.7863,
       "step": 31664
     },
     {
-      "epoch": 0.8696547746560106,
+      "epoch": 0.8985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 8.779571333333925e-07,
-      "loss": 0.9685,
+      "learning_rate": 5.349537451241593e-07,
+      "loss": 0.8284,
       "step": 31665
     },
     {
-      "epoch": 0.8696822388838538,
+      "epoch": 0.8985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 8.775927029611753e-07,
-      "loss": 0.8892,
+      "learning_rate": 5.346572075737377e-07,
+      "loss": 0.8521,
       "step": 31666
     },
     {
-      "epoch": 0.869709703111697,
+      "epoch": 0.8986095346197502,
       "grad_norm": 0.0,
-      "learning_rate": 8.772283447689755e-07,
-      "loss": 0.9017,
+      "learning_rate": 5.343607499774928e-07,
+      "loss": 0.7651,
       "step": 31667
     },
     {
-      "epoch": 0.8697371673395402,
+      "epoch": 0.8986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 8.768640587596788e-07,
-      "loss": 0.7599,
+      "learning_rate": 5.34064372337929e-07,
+      "loss": 0.795,
       "step": 31668
     },
     {
-      "epoch": 0.8697646315673835,
+      "epoch": 0.8986662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 8.764998449361705e-07,
-      "loss": 0.8504,
+      "learning_rate": 5.33768074657548e-07,
+      "loss": 0.7277,
       "step": 31669
     },
     {
-      "epoch": 0.8697920957952268,
+      "epoch": 0.8986946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 8.761357033013274e-07,
-      "loss": 0.7821,
+      "learning_rate": 5.334718569388542e-07,
+      "loss": 0.8946,
       "step": 31670
     },
     {
-      "epoch": 0.8698195600230699,
+      "epoch": 0.8987230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 8.757716338580325e-07,
-      "loss": 0.7211,
+      "learning_rate": 5.331757191843489e-07,
+      "loss": 0.8176,
       "step": 31671
     },
     {
-      "epoch": 0.8698470242509132,
+      "epoch": 0.8987514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 8.754076366091668e-07,
-      "loss": 0.8575,
+      "learning_rate": 5.328796613965326e-07,
+      "loss": 0.8546,
       "step": 31672
     },
     {
-      "epoch": 0.8698744884787564,
+      "epoch": 0.8987797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 8.750437115576105e-07,
-      "loss": 0.7618,
+      "learning_rate": 5.325836835779085e-07,
+      "loss": 0.8499,
       "step": 31673
     },
     {
-      "epoch": 0.8699019527065996,
+      "epoch": 0.8988081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 8.746798587062444e-07,
-      "loss": 0.8128,
+      "learning_rate": 5.32287785730975e-07,
+      "loss": 0.7959,
       "step": 31674
     },
     {
-      "epoch": 0.8699294169344429,
+      "epoch": 0.8988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 8.743160780579473e-07,
-      "loss": 0.8027,
+      "learning_rate": 5.319919678582319e-07,
+      "loss": 0.8127,
       "step": 31675
     },
     {
-      "epoch": 0.8699568811622861,
+      "epoch": 0.8988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 8.739523696155938e-07,
-      "loss": 0.7862,
+      "learning_rate": 5.316962299621808e-07,
+      "loss": 0.8184,
       "step": 31676
     },
     {
-      "epoch": 0.8699843453901294,
+      "epoch": 0.8988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 8.735887333820636e-07,
-      "loss": 0.7229,
+      "learning_rate": 5.314005720453153e-07,
+      "loss": 0.8424,
       "step": 31677
     },
     {
-      "epoch": 0.8700118096179726,
+      "epoch": 0.8989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 8.732251693602356e-07,
-      "loss": 0.8312,
+      "learning_rate": 5.311049941101343e-07,
+      "loss": 0.8182,
       "step": 31678
     },
     {
-      "epoch": 0.8700392738458158,
+      "epoch": 0.898950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 8.728616775529852e-07,
-      "loss": 0.7685,
+      "learning_rate": 5.308094961591371e-07,
+      "loss": 0.85,
       "step": 31679
     },
     {
-      "epoch": 0.8700667380736591,
+      "epoch": 0.8989784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 8.72498257963188e-07,
-      "loss": 0.8387,
+      "learning_rate": 5.30514078194816e-07,
+      "loss": 0.7293,
       "step": 31680
     },
     {
-      "epoch": 0.8700942023015023,
+      "epoch": 0.8990068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 8.721349105937226e-07,
-      "loss": 0.869,
+      "learning_rate": 5.302187402196679e-07,
+      "loss": 0.8455,
       "step": 31681
     },
     {
-      "epoch": 0.8701216665293455,
+      "epoch": 0.8990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 8.717716354474604e-07,
-      "loss": 0.9266,
+      "learning_rate": 5.299234822361898e-07,
+      "loss": 0.7245,
       "step": 31682
     },
     {
-      "epoch": 0.8701491307571888,
+      "epoch": 0.8990635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 8.714084325272776e-07,
-      "loss": 0.8134,
+      "learning_rate": 5.296283042468719e-07,
+      "loss": 0.8146,
       "step": 31683
     },
     {
-      "epoch": 0.870176594985032,
+      "epoch": 0.8990919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 8.710453018360465e-07,
-      "loss": 0.7138,
+      "learning_rate": 5.293332062542101e-07,
+      "loss": 0.8729,
       "step": 31684
     },
     {
-      "epoch": 0.8702040592128752,
+      "epoch": 0.8991203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 8.706822433766404e-07,
-      "loss": 0.8758,
+      "learning_rate": 5.290381882606977e-07,
+      "loss": 0.8775,
       "step": 31685
     },
     {
-      "epoch": 0.8702315234407184,
+      "epoch": 0.8991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 8.703192571519337e-07,
-      "loss": 0.8123,
+      "learning_rate": 5.287432502688239e-07,
+      "loss": 0.7677,
       "step": 31686
     },
     {
-      "epoch": 0.8702589876685617,
+      "epoch": 0.8991770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 8.699563431647984e-07,
-      "loss": 0.7175,
+      "learning_rate": 5.284483922810812e-07,
+      "loss": 0.7298,
       "step": 31687
     },
     {
-      "epoch": 0.870286451896405,
+      "epoch": 0.8992054483541431,
       "grad_norm": 0.0,
-      "learning_rate": 8.695935014181045e-07,
-      "loss": 0.8602,
+      "learning_rate": 5.281536142999622e-07,
+      "loss": 0.6595,
       "step": 31688
     },
     {
-      "epoch": 0.8703139161242481,
+      "epoch": 0.8992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 8.69230731914722e-07,
-      "loss": 0.7584,
+      "learning_rate": 5.278589163279535e-07,
+      "loss": 0.8006,
       "step": 31689
     },
     {
-      "epoch": 0.8703413803520914,
+      "epoch": 0.8992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 8.688680346575262e-07,
-      "loss": 0.8944,
+      "learning_rate": 5.275642983675477e-07,
+      "loss": 0.8403,
       "step": 31690
     },
     {
-      "epoch": 0.8703688445799347,
+      "epoch": 0.8992905788876276,
       "grad_norm": 0.0,
-      "learning_rate": 8.685054096493806e-07,
-      "loss": 0.8346,
+      "learning_rate": 5.272697604212318e-07,
+      "loss": 0.7625,
       "step": 31691
     },
     {
-      "epoch": 0.8703963088077779,
+      "epoch": 0.8993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 8.681428568931583e-07,
-      "loss": 0.7718,
+      "learning_rate": 5.269753024914936e-07,
+      "loss": 0.8119,
       "step": 31692
     },
     {
-      "epoch": 0.8704237730356211,
+      "epoch": 0.8993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 8.677803763917281e-07,
-      "loss": 0.7236,
+      "learning_rate": 5.266809245808225e-07,
+      "loss": 0.7323,
       "step": 31693
     },
     {
-      "epoch": 0.8704512372634643,
+      "epoch": 0.8993757094211123,
       "grad_norm": 0.0,
-      "learning_rate": 8.674179681479555e-07,
-      "loss": 0.7729,
+      "learning_rate": 5.263866266917028e-07,
+      "loss": 0.7921,
       "step": 31694
     },
     {
-      "epoch": 0.8704787014913076,
+      "epoch": 0.8994040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 8.670556321647095e-07,
-      "loss": 0.8147,
+      "learning_rate": 5.260924088266195e-07,
+      "loss": 0.731,
       "step": 31695
     },
     {
-      "epoch": 0.8705061657191508,
+      "epoch": 0.8994324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 8.666933684448576e-07,
-      "loss": 0.8417,
+      "learning_rate": 5.257982709880616e-07,
+      "loss": 0.7625,
       "step": 31696
     },
     {
-      "epoch": 0.870533629946994,
+      "epoch": 0.899460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 8.663311769912653e-07,
-      "loss": 0.8608,
+      "learning_rate": 5.255042131785104e-07,
+      "loss": 0.7995,
       "step": 31697
     },
     {
-      "epoch": 0.8705610941748373,
+      "epoch": 0.899489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 8.659690578068003e-07,
-      "loss": 0.7498,
+      "learning_rate": 5.252102354004518e-07,
+      "loss": 0.8544,
       "step": 31698
     },
     {
-      "epoch": 0.8705885584026805,
+      "epoch": 0.8995175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 8.656070108943271e-07,
-      "loss": 0.7844,
+      "learning_rate": 5.249163376563682e-07,
+      "loss": 0.7557,
       "step": 31699
     },
     {
-      "epoch": 0.8706160226305237,
+      "epoch": 0.8995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 8.652450362567067e-07,
-      "loss": 0.7482,
+      "learning_rate": 5.24622519948742e-07,
+      "loss": 0.7192,
       "step": 31700
     },
     {
-      "epoch": 0.870643486858367,
+      "epoch": 0.8995743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 8.648831338968067e-07,
-      "loss": 0.7731,
+      "learning_rate": 5.243287822800547e-07,
+      "loss": 0.8204,
       "step": 31701
     },
     {
-      "epoch": 0.8706709510862102,
+      "epoch": 0.8996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 8.645213038174905e-07,
-      "loss": 0.8282,
+      "learning_rate": 5.240351246527897e-07,
+      "loss": 0.737,
       "step": 31702
     },
     {
-      "epoch": 0.8706984153140535,
+      "epoch": 0.8996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 8.64159546021619e-07,
-      "loss": 0.7939,
+      "learning_rate": 5.23741547069424e-07,
+      "loss": 0.784,
       "step": 31703
     },
     {
-      "epoch": 0.8707258795418967,
+      "epoch": 0.8996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 8.637978605120578e-07,
-      "loss": 0.752,
+      "learning_rate": 5.234480495324401e-07,
+      "loss": 0.9426,
       "step": 31704
     },
     {
-      "epoch": 0.8707533437697399,
+      "epoch": 0.8996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 8.63436247291668e-07,
-      "loss": 0.8562,
+      "learning_rate": 5.231546320443159e-07,
+      "loss": 0.9029,
       "step": 31705
     },
     {
-      "epoch": 0.8707808079975832,
+      "epoch": 0.8997162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 8.630747063633071e-07,
-      "loss": 0.7575,
+      "learning_rate": 5.228612946075317e-07,
+      "loss": 0.8367,
       "step": 31706
     },
     {
-      "epoch": 0.8708082722254263,
+      "epoch": 0.899744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 8.627132377298408e-07,
-      "loss": 0.7219,
+      "learning_rate": 5.225680372245645e-07,
+      "loss": 0.7422,
       "step": 31707
     },
     {
-      "epoch": 0.8708357364532696,
+      "epoch": 0.8997729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 8.623518413941256e-07,
-      "loss": 0.7683,
+      "learning_rate": 5.2227485989789e-07,
+      "loss": 0.7195,
       "step": 31708
     },
     {
-      "epoch": 0.8708632006811129,
+      "epoch": 0.8998013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 8.619905173590215e-07,
-      "loss": 0.799,
+      "learning_rate": 5.219817626299861e-07,
+      "loss": 0.8507,
       "step": 31709
     },
     {
-      "epoch": 0.8708906649089561,
+      "epoch": 0.8998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 8.616292656273905e-07,
-      "loss": 0.8502,
+      "learning_rate": 5.2168874542333e-07,
+      "loss": 0.9164,
       "step": 31710
     },
     {
-      "epoch": 0.8709181291367993,
+      "epoch": 0.8998581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 8.612680862020862e-07,
-      "loss": 0.8105,
+      "learning_rate": 5.213958082803927e-07,
+      "loss": 0.8523,
       "step": 31711
     },
     {
-      "epoch": 0.8709455933646425,
+      "epoch": 0.8998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 8.609069790859692e-07,
-      "loss": 0.8066,
+      "learning_rate": 5.211029512036514e-07,
+      "loss": 0.8299,
       "step": 31712
     },
     {
-      "epoch": 0.8709730575924858,
+      "epoch": 0.8999148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 8.605459442818964e-07,
-      "loss": 0.8292,
+      "learning_rate": 5.208101741955818e-07,
+      "loss": 0.757,
       "step": 31713
     },
     {
-      "epoch": 0.871000521820329,
+      "epoch": 0.8999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 8.601849817927266e-07,
-      "loss": 0.8071,
+      "learning_rate": 5.20517477258653e-07,
+      "loss": 0.7586,
       "step": 31714
     },
     {
-      "epoch": 0.8710279860481722,
+      "epoch": 0.8999716231555052,
       "grad_norm": 0.0,
-      "learning_rate": 8.598240916213108e-07,
-      "loss": 0.8093,
+      "learning_rate": 5.202248603953386e-07,
+      "loss": 0.7575,
       "step": 31715
     },
     {
-      "epoch": 0.8710554502760155,
+      "epoch": 0.9,
       "grad_norm": 0.0,
-      "learning_rate": 8.594632737705099e-07,
-      "loss": 0.8067,
+      "learning_rate": 5.199323236081133e-07,
+      "loss": 0.6871,
       "step": 31716
     },
     {
-      "epoch": 0.8710829145038588,
+      "epoch": 0.9000283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 8.591025282431742e-07,
-      "loss": 0.7328,
+      "learning_rate": 5.196398668994428e-07,
+      "loss": 0.8199,
       "step": 31717
     },
     {
-      "epoch": 0.8711103787317019,
+      "epoch": 0.9000567536889897,
       "grad_norm": 0.0,
-      "learning_rate": 8.5874185504216e-07,
-      "loss": 0.7737,
+      "learning_rate": 5.193474902718021e-07,
+      "loss": 0.9147,
       "step": 31718
     },
     {
-      "epoch": 0.8711378429595452,
+      "epoch": 0.9000851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 8.583812541703218e-07,
-      "loss": 0.8184,
+      "learning_rate": 5.1905519372766e-07,
+      "loss": 0.7761,
       "step": 31719
     },
     {
-      "epoch": 0.8711653071873884,
+      "epoch": 0.9001135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 8.580207256305117e-07,
-      "loss": 0.7778,
+      "learning_rate": 5.187629772694824e-07,
+      "loss": 0.7882,
       "step": 31720
     },
     {
-      "epoch": 0.8711927714152317,
+      "epoch": 0.9001418842224744,
       "grad_norm": 0.0,
-      "learning_rate": 8.576602694255831e-07,
-      "loss": 0.7872,
+      "learning_rate": 5.184708408997429e-07,
+      "loss": 0.7933,
       "step": 31721
     },
     {
-      "epoch": 0.8712202356430749,
+      "epoch": 0.9001702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 8.572998855583891e-07,
-      "loss": 0.9567,
+      "learning_rate": 5.181787846209052e-07,
+      "loss": 0.7854,
       "step": 31722
     },
     {
-      "epoch": 0.8712476998709181,
+      "epoch": 0.9001986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 8.56939574031781e-07,
-      "loss": 0.8384,
+      "learning_rate": 5.178868084354371e-07,
+      "loss": 0.7454,
       "step": 31723
     },
     {
-      "epoch": 0.8712751640987614,
+      "epoch": 0.9002270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 8.565793348486062e-07,
-      "loss": 0.782,
+      "learning_rate": 5.17594912345808e-07,
+      "loss": 0.7481,
       "step": 31724
     },
     {
-      "epoch": 0.8713026283266045,
+      "epoch": 0.900255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 8.562191680117171e-07,
-      "loss": 0.7271,
+      "learning_rate": 5.173030963544778e-07,
+      "loss": 0.8378,
       "step": 31725
     },
     {
-      "epoch": 0.8713300925544478,
+      "epoch": 0.900283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 8.558590735239646e-07,
-      "loss": 0.8071,
+      "learning_rate": 5.170113604639159e-07,
+      "loss": 0.7417,
       "step": 31726
     },
     {
-      "epoch": 0.8713575567822911,
+      "epoch": 0.9003121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 8.554990513881978e-07,
-      "loss": 0.8512,
+      "learning_rate": 5.16719704676586e-07,
+      "loss": 0.7986,
       "step": 31727
     },
     {
-      "epoch": 0.8713850210101343,
+      "epoch": 0.9003405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 8.551391016072652e-07,
-      "loss": 0.7591,
+      "learning_rate": 5.164281289949502e-07,
+      "loss": 0.7916,
       "step": 31728
     },
     {
-      "epoch": 0.8714124852379775,
+      "epoch": 0.9003688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 8.547792241840136e-07,
-      "loss": 0.7911,
+      "learning_rate": 5.161366334214712e-07,
+      "loss": 0.8555,
       "step": 31729
     },
     {
-      "epoch": 0.8714399494658208,
+      "epoch": 0.9003972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 8.544194191212906e-07,
-      "loss": 0.7131,
+      "learning_rate": 5.158452179586148e-07,
+      "loss": 0.8692,
       "step": 31730
     },
     {
-      "epoch": 0.871467413693664,
+      "epoch": 0.9004256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 8.540596864219464e-07,
-      "loss": 0.7569,
+      "learning_rate": 5.155538826088391e-07,
+      "loss": 0.8578,
       "step": 31731
     },
     {
-      "epoch": 0.8714948779215073,
+      "epoch": 0.9004540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 8.537000260888217e-07,
-      "loss": 0.8086,
+      "learning_rate": 5.152626273746053e-07,
+      "loss": 0.8899,
       "step": 31732
     },
     {
-      "epoch": 0.8715223421493504,
+      "epoch": 0.9004824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 8.533404381247668e-07,
-      "loss": 0.7897,
+      "learning_rate": 5.14971452258376e-07,
+      "loss": 0.7397,
       "step": 31733
     },
     {
-      "epoch": 0.8715498063771937,
+      "epoch": 0.900510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 8.529809225326257e-07,
-      "loss": 0.8038,
+      "learning_rate": 5.14680357262608e-07,
+      "loss": 0.8579,
       "step": 31734
     },
     {
-      "epoch": 0.871577270605037,
+      "epoch": 0.9005391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 8.52621479315242e-07,
-      "loss": 0.7927,
+      "learning_rate": 5.143893423897617e-07,
+      "loss": 0.7848,
       "step": 31735
     },
     {
-      "epoch": 0.8716047348328801,
+      "epoch": 0.9005675368898979,
       "grad_norm": 0.0,
-      "learning_rate": 8.522621084754613e-07,
-      "loss": 0.8516,
+      "learning_rate": 5.140984076422939e-07,
+      "loss": 0.8654,
       "step": 31736
     },
     {
-      "epoch": 0.8716321990607234,
+      "epoch": 0.9005959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 8.519028100161242e-07,
-      "loss": 0.8124,
+      "learning_rate": 5.138075530226649e-07,
+      "loss": 0.7741,
       "step": 31737
     },
     {
-      "epoch": 0.8716596632885666,
+      "epoch": 0.9006242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 8.515435839400787e-07,
-      "loss": 0.9441,
+      "learning_rate": 5.135167785333306e-07,
+      "loss": 0.8414,
       "step": 31738
     },
     {
-      "epoch": 0.8716871275164099,
+      "epoch": 0.9006526674233826,
       "grad_norm": 0.0,
-      "learning_rate": 8.511844302501627e-07,
-      "loss": 0.8277,
+      "learning_rate": 5.132260841767444e-07,
+      "loss": 0.8123,
       "step": 31739
     },
     {
-      "epoch": 0.8717145917442531,
+      "epoch": 0.9006810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 8.508253489492202e-07,
-      "loss": 0.8321,
+      "learning_rate": 5.129354699553646e-07,
+      "loss": 0.7312,
       "step": 31740
     },
     {
-      "epoch": 0.8717420559720963,
+      "epoch": 0.9007094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 8.504663400400903e-07,
-      "loss": 0.8094,
+      "learning_rate": 5.126449358716468e-07,
+      "loss": 0.8857,
       "step": 31741
     },
     {
-      "epoch": 0.8717695201999396,
+      "epoch": 0.9007377979568671,
       "grad_norm": 0.0,
-      "learning_rate": 8.501074035256152e-07,
-      "loss": 0.8041,
+      "learning_rate": 5.123544819280423e-07,
+      "loss": 0.8016,
       "step": 31742
     },
     {
-      "epoch": 0.8717969844277829,
+      "epoch": 0.9007661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 8.497485394086346e-07,
-      "loss": 0.787,
+      "learning_rate": 5.120641081270061e-07,
+      "loss": 0.8307,
       "step": 31743
     },
     {
-      "epoch": 0.871824448655626,
+      "epoch": 0.900794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 8.493897476919876e-07,
-      "loss": 0.8097,
+      "learning_rate": 5.117738144709916e-07,
+      "loss": 0.7483,
       "step": 31744
     },
     {
-      "epoch": 0.8718519128834693,
+      "epoch": 0.9008229284903518,
       "grad_norm": 0.0,
-      "learning_rate": 8.490310283785142e-07,
-      "loss": 0.8341,
+      "learning_rate": 5.114836009624502e-07,
+      "loss": 0.7667,
       "step": 31745
     },
     {
-      "epoch": 0.8718793771113125,
+      "epoch": 0.9008513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 8.486723814710529e-07,
-      "loss": 0.7785,
+      "learning_rate": 5.111934676038322e-07,
+      "loss": 0.7554,
       "step": 31746
     },
     {
-      "epoch": 0.8719068413391557,
+      "epoch": 0.9008796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 8.483138069724417e-07,
-      "loss": 0.7248,
+      "learning_rate": 5.109034143975911e-07,
+      "loss": 0.8053,
       "step": 31747
     },
     {
-      "epoch": 0.871934305566999,
+      "epoch": 0.9009080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 8.479553048855138e-07,
-      "loss": 0.8311,
+      "learning_rate": 5.106134413461739e-07,
+      "loss": 0.8158,
       "step": 31748
     },
     {
-      "epoch": 0.8719617697948422,
+      "epoch": 0.9009364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 8.475968752131092e-07,
-      "loss": 0.8601,
+      "learning_rate": 5.103235484520319e-07,
+      "loss": 0.8628,
       "step": 31749
     },
     {
-      "epoch": 0.8719892340226855,
+      "epoch": 0.9009648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 8.472385179580634e-07,
-      "loss": 0.7948,
+      "learning_rate": 5.100337357176133e-07,
+      "loss": 0.7099,
       "step": 31750
     },
     {
-      "epoch": 0.8720166982505286,
+      "epoch": 0.9009931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 8.468802331232118e-07,
-      "loss": 0.8358,
+      "learning_rate": 5.097440031453671e-07,
+      "loss": 0.7966,
       "step": 31751
     },
     {
-      "epoch": 0.8720441624783719,
+      "epoch": 0.9010215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 8.465220207113911e-07,
-      "loss": 0.7346,
+      "learning_rate": 5.094543507377403e-07,
+      "loss": 0.9094,
       "step": 31752
     },
     {
-      "epoch": 0.8720716267062152,
+      "epoch": 0.901049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 8.461638807254325e-07,
-      "loss": 0.7731,
+      "learning_rate": 5.091647784971776e-07,
+      "loss": 0.8346,
       "step": 31753
     },
     {
-      "epoch": 0.8720990909340584,
+      "epoch": 0.9010783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 8.458058131681712e-07,
-      "loss": 0.8894,
+      "learning_rate": 5.088752864261271e-07,
+      "loss": 0.8391,
       "step": 31754
     },
     {
-      "epoch": 0.8721265551619016,
+      "epoch": 0.9011066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 8.454478180424419e-07,
-      "loss": 0.7779,
+      "learning_rate": 5.085858745270355e-07,
+      "loss": 0.6826,
       "step": 31755
     },
     {
-      "epoch": 0.8721540193897449,
+      "epoch": 0.9011350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 8.450898953510733e-07,
-      "loss": 0.86,
+      "learning_rate": 5.082965428023446e-07,
+      "loss": 0.8873,
       "step": 31756
     },
     {
-      "epoch": 0.8721814836175881,
+      "epoch": 0.9011634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 8.447320450968999e-07,
-      "loss": 0.8539,
+      "learning_rate": 5.080072912544987e-07,
+      "loss": 0.7529,
       "step": 31757
     },
     {
-      "epoch": 0.8722089478454313,
+      "epoch": 0.9011918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 8.443742672827548e-07,
-      "loss": 0.7477,
+      "learning_rate": 5.077181198859437e-07,
+      "loss": 0.8445,
       "step": 31758
     },
     {
-      "epoch": 0.8722364120732745,
+      "epoch": 0.9012202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 8.440165619114659e-07,
-      "loss": 0.7787,
+      "learning_rate": 5.07429028699119e-07,
+      "loss": 0.7633,
       "step": 31759
     },
     {
-      "epoch": 0.8722638763011178,
+      "epoch": 0.9012485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 8.436589289858643e-07,
-      "loss": 0.8682,
+      "learning_rate": 5.071400176964669e-07,
+      "loss": 0.7704,
       "step": 31760
     },
     {
-      "epoch": 0.8722913405289611,
+      "epoch": 0.9012769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 8.433013685087799e-07,
-      "loss": 0.7165,
+      "learning_rate": 5.068510868804322e-07,
+      "loss": 0.8024,
       "step": 31761
     },
     {
-      "epoch": 0.8723188047568042,
+      "epoch": 0.901305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 8.429438804830437e-07,
-      "loss": 0.7975,
+      "learning_rate": 5.065622362534517e-07,
+      "loss": 0.7565,
       "step": 31762
     },
     {
-      "epoch": 0.8723462689846475,
+      "epoch": 0.90133371169126,
       "grad_norm": 0.0,
-      "learning_rate": 8.425864649114813e-07,
-      "loss": 0.8235,
+      "learning_rate": 5.062734658179669e-07,
+      "loss": 0.7044,
       "step": 31763
     },
     {
-      "epoch": 0.8723737332124907,
+      "epoch": 0.9013620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 8.422291217969236e-07,
-      "loss": 0.8271,
+      "learning_rate": 5.05984775576418e-07,
+      "loss": 0.81,
       "step": 31764
     },
     {
-      "epoch": 0.872401197440334,
+      "epoch": 0.9013904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 8.418718511421942e-07,
-      "loss": 0.8777,
+      "learning_rate": 5.05696165531242e-07,
+      "loss": 0.8462,
       "step": 31765
     },
     {
-      "epoch": 0.8724286616681772,
+      "epoch": 0.9014188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 8.415146529501228e-07,
-      "loss": 0.7928,
+      "learning_rate": 5.054076356848748e-07,
+      "loss": 0.7692,
       "step": 31766
     },
     {
-      "epoch": 0.8724561258960204,
+      "epoch": 0.9014472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 8.41157527223535e-07,
-      "loss": 0.9141,
+      "learning_rate": 5.051191860397597e-07,
+      "loss": 0.8073,
       "step": 31767
     },
     {
-      "epoch": 0.8724835901238637,
+      "epoch": 0.9014755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 8.408004739652564e-07,
-      "loss": 0.8736,
+      "learning_rate": 5.048308165983273e-07,
+      "loss": 0.7691,
       "step": 31768
     },
     {
-      "epoch": 0.872511054351707,
+      "epoch": 0.9015039727582292,
       "grad_norm": 0.0,
-      "learning_rate": 8.404434931781124e-07,
-      "loss": 0.6846,
+      "learning_rate": 5.045425273630167e-07,
+      "loss": 0.6679,
       "step": 31769
     },
     {
-      "epoch": 0.8725385185795501,
+      "epoch": 0.9015323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 8.400865848649297e-07,
-      "loss": 0.7664,
+      "learning_rate": 5.042543183362625e-07,
+      "loss": 0.815,
       "step": 31770
     },
     {
-      "epoch": 0.8725659828073934,
+      "epoch": 0.9015607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 8.397297490285305e-07,
-      "loss": 0.8046,
+      "learning_rate": 5.039661895204984e-07,
+      "loss": 0.8193,
       "step": 31771
     },
     {
-      "epoch": 0.8725934470352366,
+      "epoch": 0.9015891032917139,
       "grad_norm": 0.0,
-      "learning_rate": 8.393729856717359e-07,
-      "loss": 0.6972,
+      "learning_rate": 5.03678140918159e-07,
+      "loss": 0.7813,
       "step": 31772
     },
     {
-      "epoch": 0.8726209112630798,
+      "epoch": 0.9016174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 8.390162947973701e-07,
-      "loss": 0.9056,
+      "learning_rate": 5.033901725316781e-07,
+      "loss": 0.7185,
       "step": 31773
     },
     {
-      "epoch": 0.8726483754909231,
+      "epoch": 0.9016458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 8.386596764082566e-07,
-      "loss": 0.7837,
+      "learning_rate": 5.031022843634859e-07,
+      "loss": 0.8561,
       "step": 31774
     },
     {
-      "epoch": 0.8726758397187663,
+      "epoch": 0.9016742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 8.383031305072175e-07,
-      "loss": 0.7739,
+      "learning_rate": 5.028144764160159e-07,
+      "loss": 0.7862,
       "step": 31775
     },
     {
-      "epoch": 0.8727033039466096,
+      "epoch": 0.9017026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 8.379466570970718e-07,
-      "loss": 0.6929,
+      "learning_rate": 5.025267486917007e-07,
+      "loss": 0.7237,
       "step": 31776
     },
     {
-      "epoch": 0.8727307681744527,
+      "epoch": 0.9017309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 8.375902561806415e-07,
-      "loss": 0.7329,
+      "learning_rate": 5.022391011929672e-07,
+      "loss": 0.8225,
       "step": 31777
     },
     {
-      "epoch": 0.872758232402296,
+      "epoch": 0.9017593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 8.372339277607466e-07,
-      "loss": 0.7752,
+      "learning_rate": 5.019515339222481e-07,
+      "loss": 0.7555,
       "step": 31778
     },
     {
-      "epoch": 0.8727856966301393,
+      "epoch": 0.9017877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 8.368776718402061e-07,
-      "loss": 0.8517,
+      "learning_rate": 5.0166404688197e-07,
+      "loss": 0.771,
       "step": 31779
     },
     {
-      "epoch": 0.8728131608579824,
+      "epoch": 0.9018161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 8.365214884218375e-07,
-      "loss": 0.9285,
+      "learning_rate": 5.013766400745634e-07,
+      "loss": 0.8731,
       "step": 31780
     },
     {
-      "epoch": 0.8728406250858257,
+      "epoch": 0.901844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 8.361653775084633e-07,
-      "loss": 0.792,
+      "learning_rate": 5.010893135024542e-07,
+      "loss": 0.7921,
       "step": 31781
     },
     {
-      "epoch": 0.872868089313669,
+      "epoch": 0.9018728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 8.358093391028965e-07,
-      "loss": 0.8346,
+      "learning_rate": 5.008020671680714e-07,
+      "loss": 0.7166,
       "step": 31782
     },
     {
-      "epoch": 0.8728955535415122,
+      "epoch": 0.9019012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 8.354533732079562e-07,
-      "loss": 0.8451,
+      "learning_rate": 5.005149010738408e-07,
+      "loss": 0.854,
       "step": 31783
     },
     {
-      "epoch": 0.8729230177693554,
+      "epoch": 0.9019296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 8.350974798264577e-07,
-      "loss": 0.7575,
+      "learning_rate": 5.002278152221885e-07,
+      "loss": 0.8138,
       "step": 31784
     },
     {
-      "epoch": 0.8729504819971986,
+      "epoch": 0.9019580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 8.34741658961219e-07,
-      "loss": 0.7733,
+      "learning_rate": 4.999408096155378e-07,
+      "loss": 0.803,
       "step": 31785
     },
     {
-      "epoch": 0.8729779462250419,
+      "epoch": 0.9019863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 8.343859106150565e-07,
-      "loss": 0.7997,
+      "learning_rate": 4.996538842563137e-07,
+      "loss": 0.827,
       "step": 31786
     },
     {
-      "epoch": 0.8730054104528852,
+      "epoch": 0.9020147559591374,
       "grad_norm": 0.0,
-      "learning_rate": 8.340302347907825e-07,
-      "loss": 0.8542,
+      "learning_rate": 4.99367039146943e-07,
+      "loss": 0.6938,
       "step": 31787
     },
     {
-      "epoch": 0.8730328746807283,
+      "epoch": 0.9020431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 8.336746314912103e-07,
-      "loss": 0.8112,
+      "learning_rate": 4.990802742898437e-07,
+      "loss": 0.8183,
       "step": 31788
     },
     {
-      "epoch": 0.8730603389085716,
+      "epoch": 0.9020715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 8.333191007191555e-07,
-      "loss": 0.8158,
+      "learning_rate": 4.987935896874396e-07,
+      "loss": 0.8167,
       "step": 31789
     },
     {
-      "epoch": 0.8730878031364149,
+      "epoch": 0.9020998864926221,
       "grad_norm": 0.0,
-      "learning_rate": 8.329636424774301e-07,
-      "loss": 0.8105,
+      "learning_rate": 4.985069853421554e-07,
+      "loss": 0.8164,
       "step": 31790
     },
     {
-      "epoch": 0.873115267364258,
+      "epoch": 0.9021282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 8.326082567688476e-07,
-      "loss": 0.7742,
+      "learning_rate": 4.98220461256409e-07,
+      "loss": 0.8075,
       "step": 31791
     },
     {
-      "epoch": 0.8731427315921013,
+      "epoch": 0.9021566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 8.322529435962201e-07,
-      "loss": 0.8211,
+      "learning_rate": 4.979340174326208e-07,
+      "loss": 0.7554,
       "step": 31792
     },
     {
-      "epoch": 0.8731701958199445,
+      "epoch": 0.9021850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 8.318977029623598e-07,
-      "loss": 0.8379,
+      "learning_rate": 4.976476538732123e-07,
+      "loss": 0.7876,
       "step": 31793
     },
     {
-      "epoch": 0.8731976600477878,
+      "epoch": 0.9022133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 8.315425348700745e-07,
-      "loss": 0.814,
+      "learning_rate": 4.973613705806002e-07,
+      "loss": 0.8251,
       "step": 31794
     },
     {
-      "epoch": 0.873225124275631,
+      "epoch": 0.9022417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 8.311874393221786e-07,
-      "loss": 0.8164,
+      "learning_rate": 4.970751675572039e-07,
+      "loss": 0.8793,
       "step": 31795
     },
     {
-      "epoch": 0.8732525885034742,
+      "epoch": 0.9022701475595913,
       "grad_norm": 0.0,
-      "learning_rate": 8.308324163214764e-07,
-      "loss": 0.8355,
+      "learning_rate": 4.967890448054413e-07,
+      "loss": 0.7906,
       "step": 31796
     },
     {
-      "epoch": 0.8732800527313175,
+      "epoch": 0.9022985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 8.304774658707814e-07,
-      "loss": 0.7963,
+      "learning_rate": 4.965030023277273e-07,
+      "loss": 0.7678,
       "step": 31797
     },
     {
-      "epoch": 0.8733075169591606,
+      "epoch": 0.9023269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 8.301225879728992e-07,
-      "loss": 0.8961,
+      "learning_rate": 4.962170401264821e-07,
+      "loss": 0.7544,
       "step": 31798
     },
     {
-      "epoch": 0.8733349811870039,
+      "epoch": 0.902355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 8.297677826306417e-07,
-      "loss": 0.8007,
+      "learning_rate": 4.959311582041171e-07,
+      "loss": 0.7113,
       "step": 31799
     },
     {
-      "epoch": 0.8733624454148472,
+      "epoch": 0.9023836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 8.294130498468123e-07,
-      "loss": 0.8426,
+      "learning_rate": 4.956453565630492e-07,
+      "loss": 0.773,
       "step": 31800
     },
     {
-      "epoch": 0.8733899096426904,
+      "epoch": 0.9024120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 8.290583896242188e-07,
-      "loss": 0.8231,
+      "learning_rate": 4.953596352056933e-07,
+      "loss": 0.8979,
       "step": 31801
     },
     {
-      "epoch": 0.8734173738705336,
+      "epoch": 0.9024404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 8.28703801965669e-07,
-      "loss": 0.8929,
+      "learning_rate": 4.950739941344607e-07,
+      "loss": 0.8171,
       "step": 31802
     },
     {
-      "epoch": 0.8734448380983769,
+      "epoch": 0.9024687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 8.283492868739662e-07,
-      "loss": 0.728,
+      "learning_rate": 4.947884333517649e-07,
+      "loss": 0.6732,
       "step": 31803
     },
     {
-      "epoch": 0.8734723023262201,
+      "epoch": 0.9024971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 8.279948443519159e-07,
-      "loss": 0.8053,
+      "learning_rate": 4.945029528600209e-07,
+      "loss": 0.8207,
       "step": 31804
     },
     {
-      "epoch": 0.8734997665540634,
+      "epoch": 0.9025255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 8.276404744023258e-07,
-      "loss": 0.8352,
+      "learning_rate": 4.942175526616355e-07,
+      "loss": 0.7117,
       "step": 31805
     },
     {
-      "epoch": 0.8735272307819065,
+      "epoch": 0.9025539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 8.272861770279961e-07,
-      "loss": 0.7909,
+      "learning_rate": 4.939322327590224e-07,
+      "loss": 0.863,
       "step": 31806
     },
     {
-      "epoch": 0.8735546950097498,
+      "epoch": 0.9025822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 8.269319522317298e-07,
-      "loss": 0.9256,
+      "learning_rate": 4.936469931545918e-07,
+      "loss": 0.7803,
       "step": 31807
     },
     {
-      "epoch": 0.8735821592375931,
+      "epoch": 0.9026106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 8.265778000163327e-07,
-      "loss": 0.911,
+      "learning_rate": 4.933618338507506e-07,
+      "loss": 0.7434,
       "step": 31808
     },
     {
-      "epoch": 0.8736096234654362,
+      "epoch": 0.902639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 8.262237203846046e-07,
-      "loss": 0.8409,
+      "learning_rate": 4.930767548499105e-07,
+      "loss": 0.8739,
       "step": 31809
     },
     {
-      "epoch": 0.8736370876932795,
+      "epoch": 0.9026674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 8.2586971333935e-07,
-      "loss": 0.8047,
+      "learning_rate": 4.927917561544792e-07,
+      "loss": 0.8329,
       "step": 31810
     },
     {
-      "epoch": 0.8736645519211227,
+      "epoch": 0.9026958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 8.255157788833678e-07,
-      "loss": 0.8267,
+      "learning_rate": 4.925068377668618e-07,
+      "loss": 0.763,
       "step": 31811
     },
     {
-      "epoch": 0.873692016148966,
+      "epoch": 0.9027241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 8.25161917019457e-07,
-      "loss": 0.7782,
+      "learning_rate": 4.922219996894662e-07,
+      "loss": 0.9436,
       "step": 31812
     },
     {
-      "epoch": 0.8737194803768092,
+      "epoch": 0.9027525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 8.248081277504183e-07,
-      "loss": 0.789,
+      "learning_rate": 4.919372419246993e-07,
+      "loss": 0.8946,
       "step": 31813
     },
     {
-      "epoch": 0.8737469446046524,
+      "epoch": 0.9027809307604995,
       "grad_norm": 0.0,
-      "learning_rate": 8.244544110790531e-07,
-      "loss": 0.8224,
+      "learning_rate": 4.916525644749659e-07,
+      "loss": 0.8631,
       "step": 31814
     },
     {
-      "epoch": 0.8737744088324957,
+      "epoch": 0.9028093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 8.24100767008158e-07,
-      "loss": 0.9058,
+      "learning_rate": 4.913679673426719e-07,
+      "loss": 0.8148,
       "step": 31815
     },
     {
-      "epoch": 0.873801873060339,
+      "epoch": 0.9028376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 8.23747195540533e-07,
-      "loss": 0.767,
+      "learning_rate": 4.910834505302198e-07,
+      "loss": 0.7549,
       "step": 31816
     },
     {
-      "epoch": 0.8738293372881821,
+      "epoch": 0.9028660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 8.233936966789757e-07,
-      "loss": 0.7215,
+      "learning_rate": 4.907990140400121e-07,
+      "loss": 0.8407,
       "step": 31817
     },
     {
-      "epoch": 0.8738568015160254,
+      "epoch": 0.902894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.230402704262807e-07,
-      "loss": 0.8182,
+      "learning_rate": 4.905146578744535e-07,
+      "loss": 0.7035,
       "step": 31818
     },
     {
-      "epoch": 0.8738842657438686,
+      "epoch": 0.9029228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 8.226869167852469e-07,
-      "loss": 0.732,
+      "learning_rate": 4.902303820359445e-07,
+      "loss": 0.7733,
       "step": 31819
     },
     {
-      "epoch": 0.8739117299717118,
+      "epoch": 0.9029511918274687,
       "grad_norm": 0.0,
-      "learning_rate": 8.223336357586687e-07,
-      "loss": 0.7816,
+      "learning_rate": 4.899461865268873e-07,
+      "loss": 0.7628,
       "step": 31820
     },
     {
-      "epoch": 0.8739391941995551,
+      "epoch": 0.9029795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 8.219804273493404e-07,
-      "loss": 0.8453,
+      "learning_rate": 4.896620713496836e-07,
+      "loss": 0.8552,
       "step": 31821
     },
     {
-      "epoch": 0.8739666584273983,
+      "epoch": 0.9030079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 8.216272915600587e-07,
-      "loss": 0.8275,
+      "learning_rate": 4.893780365067291e-07,
+      "loss": 0.8058,
       "step": 31822
     },
     {
-      "epoch": 0.8739941226552416,
+      "epoch": 0.9030363223609534,
       "grad_norm": 0.0,
-      "learning_rate": 8.212742283936193e-07,
-      "loss": 0.869,
+      "learning_rate": 4.890940820004275e-07,
+      "loss": 0.6916,
       "step": 31823
     },
     {
-      "epoch": 0.8740215868830847,
+      "epoch": 0.9030646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 8.209212378528119e-07,
-      "loss": 0.7295,
+      "learning_rate": 4.888102078331758e-07,
+      "loss": 0.7926,
       "step": 31824
     },
     {
-      "epoch": 0.874049051110928,
+      "epoch": 0.9030930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 8.205683199404313e-07,
-      "loss": 0.8314,
+      "learning_rate": 4.88526414007372e-07,
+      "loss": 0.7492,
       "step": 31825
     },
     {
-      "epoch": 0.8740765153387713,
+      "epoch": 0.9031214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 8.202154746592717e-07,
-      "loss": 0.881,
+      "learning_rate": 4.88242700525412e-07,
+      "loss": 0.8136,
       "step": 31826
     },
     {
-      "epoch": 0.8741039795666145,
+      "epoch": 0.903149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 8.198627020121219e-07,
-      "loss": 0.8128,
+      "learning_rate": 4.879590673896939e-07,
+      "loss": 0.784,
       "step": 31827
     },
     {
-      "epoch": 0.8741314437944577,
+      "epoch": 0.903178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 8.195100020017732e-07,
-      "loss": 0.7298,
+      "learning_rate": 4.876755146026124e-07,
+      "loss": 0.8301,
       "step": 31828
     },
     {
-      "epoch": 0.874158908022301,
+      "epoch": 0.9032065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 8.191573746310211e-07,
-      "loss": 0.8804,
+      "learning_rate": 4.873920421665657e-07,
+      "loss": 0.8611,
       "step": 31829
     },
     {
-      "epoch": 0.8741863722501442,
+      "epoch": 0.9032349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 8.188048199026488e-07,
-      "loss": 0.7587,
+      "learning_rate": 4.871086500839439e-07,
+      "loss": 0.7979,
       "step": 31830
     },
     {
-      "epoch": 0.8742138364779874,
+      "epoch": 0.9032633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 8.184523378194497e-07,
-      "loss": 0.8618,
+      "learning_rate": 4.868253383571442e-07,
+      "loss": 0.8413,
       "step": 31831
     },
     {
-      "epoch": 0.8742413007058306,
+      "epoch": 0.9032917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 8.180999283842128e-07,
-      "loss": 0.9055,
+      "learning_rate": 4.86542106988559e-07,
+      "loss": 0.7813,
       "step": 31832
     },
     {
-      "epoch": 0.8742687649336739,
+      "epoch": 0.9033200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 8.177475915997246e-07,
-      "loss": 0.7379,
+      "learning_rate": 4.862589559805786e-07,
+      "loss": 0.7744,
       "step": 31833
     },
     {
-      "epoch": 0.8742962291615172,
+      "epoch": 0.9033484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 8.173953274687774e-07,
-      "loss": 0.7762,
+      "learning_rate": 4.859758853355967e-07,
+      "loss": 0.8659,
       "step": 31834
     },
     {
-      "epoch": 0.8743236933893603,
+      "epoch": 0.9033768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 8.170431359941555e-07,
-      "loss": 0.829,
+      "learning_rate": 4.856928950560058e-07,
+      "loss": 0.7112,
       "step": 31835
     },
     {
-      "epoch": 0.8743511576172036,
+      "epoch": 0.903405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 8.166910171786424e-07,
-      "loss": 0.8466,
+      "learning_rate": 4.85409985144194e-07,
+      "loss": 0.8075,
       "step": 31836
     },
     {
-      "epoch": 0.8743786218450468,
+      "epoch": 0.9034335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 8.163389710250281e-07,
-      "loss": 0.8276,
+      "learning_rate": 4.851271556025505e-07,
+      "loss": 0.7432,
       "step": 31837
     },
     {
-      "epoch": 0.87440608607289,
+      "epoch": 0.9034619750283769,
       "grad_norm": 0.0,
-      "learning_rate": 8.159869975360979e-07,
-      "loss": 0.7882,
+      "learning_rate": 4.848444064334679e-07,
+      "loss": 0.8182,
       "step": 31838
     },
     {
-      "epoch": 0.8744335503007333,
+      "epoch": 0.9034903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 8.156350967146354e-07,
-      "loss": 0.7245,
+      "learning_rate": 4.845617376393308e-07,
+      "loss": 0.8977,
       "step": 31839
     },
     {
-      "epoch": 0.8744610145285765,
+      "epoch": 0.9035187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 8.152832685634271e-07,
-      "loss": 0.7683,
+      "learning_rate": 4.842791492225285e-07,
+      "loss": 0.7821,
       "step": 31840
     },
     {
-      "epoch": 0.8744884787564198,
+      "epoch": 0.9035471055618616,
       "grad_norm": 0.0,
-      "learning_rate": 8.149315130852542e-07,
-      "loss": 0.8732,
+      "learning_rate": 4.839966411854491e-07,
+      "loss": 0.7835,
       "step": 31841
     },
     {
-      "epoch": 0.874515942984263,
+      "epoch": 0.9035754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 8.145798302829011e-07,
-      "loss": 0.8301,
+      "learning_rate": 4.83714213530475e-07,
+      "loss": 0.8909,
       "step": 31842
     },
     {
-      "epoch": 0.8745434072121062,
+      "epoch": 0.9036038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 8.142282201591523e-07,
-      "loss": 0.7637,
+      "learning_rate": 4.834318662599968e-07,
+      "loss": 0.7856,
       "step": 31843
     },
     {
-      "epoch": 0.8745708714399495,
+      "epoch": 0.9036322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 8.138766827167854e-07,
-      "loss": 0.6963,
+      "learning_rate": 4.831495993763968e-07,
+      "loss": 0.7129,
       "step": 31844
     },
     {
-      "epoch": 0.8745983356677927,
+      "epoch": 0.9036606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 8.135252179585851e-07,
-      "loss": 0.7582,
+      "learning_rate": 4.82867412882061e-07,
+      "loss": 0.7656,
       "step": 31845
     },
     {
-      "epoch": 0.8746257998956359,
+      "epoch": 0.903688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 8.131738258873345e-07,
-      "loss": 0.8142,
+      "learning_rate": 4.825853067793717e-07,
+      "loss": 0.7619,
       "step": 31846
     },
     {
-      "epoch": 0.8746532641234792,
+      "epoch": 0.9037173666288308,
       "grad_norm": 0.0,
-      "learning_rate": 8.128225065058081e-07,
-      "loss": 0.8374,
+      "learning_rate": 4.823032810707118e-07,
+      "loss": 0.9077,
       "step": 31847
     },
     {
-      "epoch": 0.8746807283513224,
+      "epoch": 0.9037457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 8.124712598167894e-07,
-      "loss": 0.8319,
+      "learning_rate": 4.820213357584636e-07,
+      "loss": 0.8321,
       "step": 31848
     },
     {
-      "epoch": 0.8747081925791657,
+      "epoch": 0.9037741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 8.121200858230571e-07,
-      "loss": 0.8269,
+      "learning_rate": 4.817394708450107e-07,
+      "loss": 0.8951,
       "step": 31849
     },
     {
-      "epoch": 0.8747356568070088,
+      "epoch": 0.9038024971623155,
       "grad_norm": 0.0,
-      "learning_rate": 8.117689845273912e-07,
-      "loss": 0.8394,
+      "learning_rate": 4.814576863327303e-07,
+      "loss": 0.7642,
       "step": 31850
     },
     {
-      "epoch": 0.8747631210348521,
+      "epoch": 0.9038308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 8.114179559325663e-07,
-      "loss": 0.8021,
+      "learning_rate": 4.811759822240047e-07,
+      "loss": 0.8353,
       "step": 31851
     },
     {
-      "epoch": 0.8747905852626954,
+      "epoch": 0.9038592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 8.110670000413645e-07,
-      "loss": 0.7513,
+      "learning_rate": 4.808943585212156e-07,
+      "loss": 0.8302,
       "step": 31852
     },
     {
-      "epoch": 0.8748180494905385,
+      "epoch": 0.9038876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 8.107161168565569e-07,
-      "loss": 0.7078,
+      "learning_rate": 4.806128152267387e-07,
+      "loss": 0.8248,
       "step": 31853
     },
     {
-      "epoch": 0.8748455137183818,
+      "epoch": 0.9039160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 8.103653063809247e-07,
-      "loss": 0.8855,
+      "learning_rate": 4.80331352342952e-07,
+      "loss": 0.8389,
       "step": 31854
     },
     {
-      "epoch": 0.8748729779462251,
+      "epoch": 0.90394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 8.1001456861724e-07,
-      "loss": 0.8553,
+      "learning_rate": 4.80049969872236e-07,
+      "loss": 0.8803,
       "step": 31855
     },
     {
-      "epoch": 0.8749004421740683,
+      "epoch": 0.9039727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 8.096639035682818e-07,
-      "loss": 0.9178,
+      "learning_rate": 4.797686678169655e-07,
+      "loss": 0.8553,
       "step": 31856
     },
     {
-      "epoch": 0.8749279064019115,
+      "epoch": 0.9040011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 8.093133112368212e-07,
-      "loss": 0.8271,
+      "learning_rate": 4.794874461795163e-07,
+      "loss": 0.789,
       "step": 31857
     },
     {
-      "epoch": 0.8749553706297547,
+      "epoch": 0.9040295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 8.089627916256371e-07,
-      "loss": 0.825,
+      "learning_rate": 4.792063049622642e-07,
+      "loss": 0.866,
       "step": 31858
     },
     {
-      "epoch": 0.874982834857598,
+      "epoch": 0.9040578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 8.086123447374983e-07,
-      "loss": 0.8476,
+      "learning_rate": 4.789252441675851e-07,
+      "loss": 0.8573,
       "step": 31859
     },
     {
-      "epoch": 0.8750102990854413,
+      "epoch": 0.9040862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 8.082619705751782e-07,
-      "loss": 0.749,
+      "learning_rate": 4.786442637978539e-07,
+      "loss": 0.8257,
       "step": 31860
     },
     {
-      "epoch": 0.8750377633132844,
+      "epoch": 0.9041146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 8.07911669141449e-07,
-      "loss": 0.801,
+      "learning_rate": 4.783633638554408e-07,
+      "loss": 0.7754,
       "step": 31861
     },
     {
-      "epoch": 0.8750652275411277,
+      "epoch": 0.9041430192962543,
       "grad_norm": 0.0,
-      "learning_rate": 8.075614404390841e-07,
-      "loss": 0.8195,
+      "learning_rate": 4.780825443427206e-07,
+      "loss": 0.7092,
       "step": 31862
     },
     {
-      "epoch": 0.8750926917689709,
+      "epoch": 0.9041713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 8.072112844708535e-07,
-      "loss": 0.7401,
+      "learning_rate": 4.77801805262067e-07,
+      "loss": 0.7785,
       "step": 31863
     },
     {
-      "epoch": 0.8751201559968141,
+      "epoch": 0.904199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 8.068612012395305e-07,
-      "loss": 0.8514,
+      "learning_rate": 4.775211466158469e-07,
+      "loss": 0.7953,
       "step": 31864
     },
     {
-      "epoch": 0.8751476202246574,
+      "epoch": 0.904228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 8.065111907478806e-07,
-      "loss": 0.7392,
+      "learning_rate": 4.772405684064352e-07,
+      "loss": 0.6603,
       "step": 31865
     },
     {
-      "epoch": 0.8751750844525006,
+      "epoch": 0.9042565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 8.061612529986751e-07,
-      "loss": 0.8066,
+      "learning_rate": 4.76960070636201e-07,
+      "loss": 0.9319,
       "step": 31866
     },
     {
-      "epoch": 0.8752025486803439,
+      "epoch": 0.9042849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 8.05811387994685e-07,
-      "loss": 0.8358,
+      "learning_rate": 4.766796533075124e-07,
+      "loss": 0.7287,
       "step": 31867
     },
     {
-      "epoch": 0.8752300129081871,
+      "epoch": 0.9043132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 8.054615957386746e-07,
-      "loss": 0.7615,
+      "learning_rate": 4.7639931642273875e-07,
+      "loss": 0.9405,
       "step": 31868
     },
     {
-      "epoch": 0.8752574771360303,
+      "epoch": 0.9043416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 8.051118762334142e-07,
-      "loss": 0.803,
+      "learning_rate": 4.7611905998424914e-07,
+      "loss": 0.7869,
       "step": 31869
     },
     {
-      "epoch": 0.8752849413638736,
+      "epoch": 0.9043700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 8.047622294816726e-07,
-      "loss": 0.7998,
+      "learning_rate": 4.7583888399440834e-07,
+      "loss": 0.7292,
       "step": 31870
     },
     {
-      "epoch": 0.8753124055917167,
+      "epoch": 0.9043984108967082,
       "grad_norm": 0.0,
-      "learning_rate": 8.04412655486212e-07,
-      "loss": 0.8315,
+      "learning_rate": 4.7555878845558567e-07,
+      "loss": 0.9308,
       "step": 31871
     },
     {
-      "epoch": 0.87533986981956,
+      "epoch": 0.9044267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 8.040631542498001e-07,
-      "loss": 0.8596,
+      "learning_rate": 4.7527877337014694e-07,
+      "loss": 0.8957,
       "step": 31872
     },
     {
-      "epoch": 0.8753673340474033,
+      "epoch": 0.9044551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 8.037137257752026e-07,
-      "loss": 0.7912,
+      "learning_rate": 4.749988387404536e-07,
+      "loss": 0.7041,
       "step": 31873
     },
     {
-      "epoch": 0.8753947982752465,
+      "epoch": 0.9044835414301929,
       "grad_norm": 0.0,
-      "learning_rate": 8.033643700651872e-07,
-      "loss": 0.8121,
+      "learning_rate": 4.7471898456887487e-07,
+      "loss": 0.807,
       "step": 31874
     },
     {
-      "epoch": 0.8754222625030897,
+      "epoch": 0.9045119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 8.030150871225129e-07,
-      "loss": 0.7435,
+      "learning_rate": 4.7443921085777444e-07,
+      "loss": 0.6505,
       "step": 31875
     },
     {
-      "epoch": 0.8754497267309329,
+      "epoch": 0.9045402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 8.026658769499474e-07,
-      "loss": 0.856,
+      "learning_rate": 4.7415951760951263e-07,
+      "loss": 0.8774,
       "step": 31876
     },
     {
-      "epoch": 0.8754771909587762,
+      "epoch": 0.9045686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 8.023167395502507e-07,
-      "loss": 0.7472,
+      "learning_rate": 4.738799048264531e-07,
+      "loss": 0.7671,
       "step": 31877
     },
     {
-      "epoch": 0.8755046551866195,
+      "epoch": 0.9045970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 8.019676749261862e-07,
-      "loss": 0.7548,
+      "learning_rate": 4.736003725109606e-07,
+      "loss": 0.8166,
       "step": 31878
     },
     {
-      "epoch": 0.8755321194144626,
+      "epoch": 0.9046254256526675,
       "grad_norm": 0.0,
-      "learning_rate": 8.016186830805172e-07,
-      "loss": 0.8642,
+      "learning_rate": 4.7332092066539214e-07,
+      "loss": 0.7328,
       "step": 31879
     },
     {
-      "epoch": 0.8755595836423059,
+      "epoch": 0.9046538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 8.012697640160039e-07,
-      "loss": 0.8746,
+      "learning_rate": 4.7304154929211034e-07,
+      "loss": 0.9136,
       "step": 31880
     },
     {
-      "epoch": 0.8755870478701492,
+      "epoch": 0.9046821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 8.009209177354083e-07,
-      "loss": 0.7557,
+      "learning_rate": 4.7276225839347544e-07,
+      "loss": 0.82,
       "step": 31881
     },
     {
-      "epoch": 0.8756145120979923,
+      "epoch": 0.9047105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 8.005721442414904e-07,
-      "loss": 0.7865,
+      "learning_rate": 4.7248304797184563e-07,
+      "loss": 0.902,
       "step": 31882
     },
     {
-      "epoch": 0.8756419763258356,
+      "epoch": 0.904738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 8.002234435370094e-07,
-      "loss": 0.7149,
+      "learning_rate": 4.72203918029579e-07,
+      "loss": 0.8913,
       "step": 31883
     },
     {
-      "epoch": 0.8756694405536788,
+      "epoch": 0.9047673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 7.998748156247227e-07,
-      "loss": 0.7366,
+      "learning_rate": 4.71924868569037e-07,
+      "loss": 0.7748,
       "step": 31884
     },
     {
-      "epoch": 0.8756969047815221,
+      "epoch": 0.9047956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 7.995262605073905e-07,
-      "loss": 0.7534,
+      "learning_rate": 4.7164589959257103e-07,
+      "loss": 0.8067,
       "step": 31885
     },
     {
-      "epoch": 0.8757243690093653,
+      "epoch": 0.9048240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 7.991777781877696e-07,
-      "loss": 0.7965,
+      "learning_rate": 4.713670111025426e-07,
+      "loss": 0.7897,
       "step": 31886
     },
     {
-      "epoch": 0.8757518332372085,
+      "epoch": 0.9048524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 7.988293686686188e-07,
-      "loss": 0.6903,
+      "learning_rate": 4.710882031013042e-07,
+      "loss": 0.9559,
       "step": 31887
     },
     {
-      "epoch": 0.8757792974650518,
+      "epoch": 0.9048808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 7.984810319526959e-07,
-      "loss": 0.8976,
+      "learning_rate": 4.708094755912096e-07,
+      "loss": 0.7861,
       "step": 31888
     },
     {
-      "epoch": 0.875806761692895,
+      "epoch": 0.9049091940976164,
       "grad_norm": 0.0,
-      "learning_rate": 7.981327680427542e-07,
-      "loss": 0.8787,
+      "learning_rate": 4.7053082857461906e-07,
+      "loss": 0.7892,
       "step": 31889
     },
     {
-      "epoch": 0.8758342259207382,
+      "epoch": 0.9049375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 7.977845769415493e-07,
-      "loss": 0.7173,
+      "learning_rate": 4.7025226205388185e-07,
+      "loss": 0.8123,
       "step": 31890
     },
     {
-      "epoch": 0.8758616901485815,
+      "epoch": 0.9049659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 7.974364586518401e-07,
-      "loss": 0.8211,
+      "learning_rate": 4.6997377603135277e-07,
+      "loss": 0.7381,
       "step": 31891
     },
     {
-      "epoch": 0.8758891543764247,
+      "epoch": 0.9049943246311011,
       "grad_norm": 0.0,
-      "learning_rate": 7.970884131763768e-07,
-      "loss": 0.7957,
+      "learning_rate": 4.696953705093843e-07,
+      "loss": 0.7916,
       "step": 31892
     },
     {
-      "epoch": 0.875916618604268,
+      "epoch": 0.9050227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 7.967404405179157e-07,
-      "loss": 0.8449,
+      "learning_rate": 4.694170454903257e-07,
+      "loss": 0.7876,
       "step": 31893
     },
     {
-      "epoch": 0.8759440828321112,
+      "epoch": 0.9050510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 7.963925406792084e-07,
-      "loss": 0.7414,
+      "learning_rate": 4.6913880097653075e-07,
+      "loss": 0.7353,
       "step": 31894
     },
     {
-      "epoch": 0.8759715470599544,
+      "epoch": 0.9050794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 7.960447136630078e-07,
-      "loss": 0.9066,
+      "learning_rate": 4.6886063697034856e-07,
+      "loss": 0.893,
       "step": 31895
     },
     {
-      "epoch": 0.8759990112877977,
+      "epoch": 0.9051078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 7.956969594720665e-07,
-      "loss": 0.7532,
+      "learning_rate": 4.685825534741295e-07,
+      "loss": 0.8799,
       "step": 31896
     },
     {
-      "epoch": 0.8760264755156408,
+      "epoch": 0.9051362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 7.953492781091366e-07,
-      "loss": 0.7238,
+      "learning_rate": 4.683045504902206e-07,
+      "loss": 0.7551,
       "step": 31897
     },
     {
-      "epoch": 0.8760539397434841,
+      "epoch": 0.9051645856980703,
       "grad_norm": 0.0,
-      "learning_rate": 7.950016695769713e-07,
-      "loss": 0.7095,
+      "learning_rate": 4.6802662802097443e-07,
+      "loss": 0.8029,
       "step": 31898
     },
     {
-      "epoch": 0.8760814039713274,
+      "epoch": 0.9051929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 7.946541338783176e-07,
-      "loss": 0.8415,
+      "learning_rate": 4.6774878606873353e-07,
+      "loss": 0.8724,
       "step": 31899
     },
     {
-      "epoch": 0.8761088681991706,
+      "epoch": 0.9052213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 7.943066710159241e-07,
-      "loss": 0.8176,
+      "learning_rate": 4.6747102463584714e-07,
+      "loss": 0.7893,
       "step": 31900
     },
     {
-      "epoch": 0.8761363324270138,
+      "epoch": 0.905249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 7.939592809925422e-07,
-      "loss": 0.8871,
+      "learning_rate": 4.671933437246623e-07,
+      "loss": 0.8305,
       "step": 31901
     },
     {
-      "epoch": 0.876163796654857,
+      "epoch": 0.9052780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 7.936119638109219e-07,
-      "loss": 0.781,
+      "learning_rate": 4.6691574333752375e-07,
+      "loss": 0.8293,
       "step": 31902
     },
     {
-      "epoch": 0.8761912608827003,
+      "epoch": 0.9053064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 7.932647194738085e-07,
-      "loss": 0.7721,
+      "learning_rate": 4.6663822347677633e-07,
+      "loss": 0.8109,
       "step": 31903
     },
     {
-      "epoch": 0.8762187251105436,
+      "epoch": 0.9053348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 7.929175479839512e-07,
-      "loss": 0.8499,
+      "learning_rate": 4.663607841447637e-07,
+      "loss": 0.8848,
       "step": 31904
     },
     {
-      "epoch": 0.8762461893383867,
+      "epoch": 0.9053632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 7.925704493440989e-07,
-      "loss": 0.8633,
+      "learning_rate": 4.6608342534382956e-07,
+      "loss": 0.8278,
       "step": 31905
     },
     {
-      "epoch": 0.87627365356623,
+      "epoch": 0.9053916004540296,
       "grad_norm": 0.0,
-      "learning_rate": 7.922234235569947e-07,
-      "loss": 0.8049,
+      "learning_rate": 4.6580614707631975e-07,
+      "loss": 0.6845,
       "step": 31906
     },
     {
-      "epoch": 0.8763011177940733,
+      "epoch": 0.9054199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 7.918764706253867e-07,
-      "loss": 0.7192,
+      "learning_rate": 4.655289493445725e-07,
+      "loss": 0.7735,
       "step": 31907
     },
     {
-      "epoch": 0.8763285820219164,
+      "epoch": 0.9054483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 7.91529590552017e-07,
-      "loss": 0.6696,
+      "learning_rate": 4.652518321509303e-07,
+      "loss": 0.8367,
       "step": 31908
     },
     {
-      "epoch": 0.8763560462497597,
+      "epoch": 0.9054767309875141,
       "grad_norm": 0.0,
-      "learning_rate": 7.911827833396324e-07,
-      "loss": 0.8752,
+      "learning_rate": 4.649747954977368e-07,
+      "loss": 0.8117,
       "step": 31909
     },
     {
-      "epoch": 0.8763835104776029,
+      "epoch": 0.9055051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 7.90836048990976e-07,
-      "loss": 0.801,
+      "learning_rate": 4.64697839387328e-07,
+      "loss": 0.7218,
       "step": 31910
     },
     {
-      "epoch": 0.8764109747054462,
+      "epoch": 0.905533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 7.904893875087949e-07,
-      "loss": 0.8589,
+      "learning_rate": 4.644209638220443e-07,
+      "loss": 0.7193,
       "step": 31911
     },
     {
-      "epoch": 0.8764384389332894,
+      "epoch": 0.9055618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 7.901427988958266e-07,
-      "loss": 0.8176,
+      "learning_rate": 4.641441688042281e-07,
+      "loss": 0.7593,
       "step": 31912
     },
     {
-      "epoch": 0.8764659031611326,
+      "epoch": 0.9055902383654938,
       "grad_norm": 0.0,
-      "learning_rate": 7.897962831548167e-07,
-      "loss": 0.8123,
+      "learning_rate": 4.6386745433621317e-07,
+      "loss": 0.7531,
       "step": 31913
     },
     {
-      "epoch": 0.8764933673889759,
+      "epoch": 0.9056186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 7.894498402885053e-07,
-      "loss": 0.7281,
+      "learning_rate": 4.6359082042033766e-07,
+      "loss": 0.7779,
       "step": 31914
     },
     {
-      "epoch": 0.876520831616819,
+      "epoch": 0.9056469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 7.891034702996381e-07,
-      "loss": 0.7726,
+      "learning_rate": 4.633142670589408e-07,
+      "loss": 0.8304,
       "step": 31915
     },
     {
-      "epoch": 0.8765482958446623,
+      "epoch": 0.9056753688989785,
       "grad_norm": 0.0,
-      "learning_rate": 7.887571731909493e-07,
-      "loss": 0.8302,
+      "learning_rate": 4.6303779425435625e-07,
+      "loss": 0.8406,
       "step": 31916
     },
     {
-      "epoch": 0.8765757600725056,
+      "epoch": 0.9057037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 7.884109489651836e-07,
-      "loss": 0.8215,
+      "learning_rate": 4.6276140200891994e-07,
+      "loss": 0.8048,
       "step": 31917
     },
     {
-      "epoch": 0.8766032243003488,
+      "epoch": 0.9057321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 7.880647976250766e-07,
-      "loss": 0.7149,
+      "learning_rate": 4.624850903249678e-07,
+      "loss": 0.8424,
       "step": 31918
     },
     {
-      "epoch": 0.876630688528192,
+      "epoch": 0.9057604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 7.877187191733704e-07,
-      "loss": 0.7728,
+      "learning_rate": 4.6220885920483014e-07,
+      "loss": 0.8051,
       "step": 31919
     },
     {
-      "epoch": 0.8766581527560353,
+      "epoch": 0.905788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 7.873727136128007e-07,
-      "loss": 0.7922,
+      "learning_rate": 4.619327086508463e-07,
+      "loss": 0.7716,
       "step": 31920
     },
     {
-      "epoch": 0.8766856169838785,
+      "epoch": 0.9058172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 7.870267809461085e-07,
-      "loss": 0.7646,
+      "learning_rate": 4.616566386653443e-07,
+      "loss": 0.8834,
       "step": 31921
     },
     {
-      "epoch": 0.8767130812117218,
+      "epoch": 0.9058456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 7.866809211760296e-07,
-      "loss": 0.718,
+      "learning_rate": 4.6138064925065673e-07,
+      "loss": 0.8383,
       "step": 31922
     },
     {
-      "epoch": 0.8767405454395649,
+      "epoch": 0.9058740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 7.863351343053005e-07,
-      "loss": 0.8279,
+      "learning_rate": 4.6110474040911735e-07,
+      "loss": 0.7537,
       "step": 31923
     },
     {
-      "epoch": 0.8767680096674082,
+      "epoch": 0.9059023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 7.859894203366548e-07,
-      "loss": 0.7768,
+      "learning_rate": 4.6082891214305423e-07,
+      "loss": 0.7521,
       "step": 31924
     },
     {
-      "epoch": 0.8767954738952515,
+      "epoch": 0.9059307604994324,
       "grad_norm": 0.0,
-      "learning_rate": 7.85643779272831e-07,
-      "loss": 0.7428,
+      "learning_rate": 4.605531644547967e-07,
+      "loss": 0.7717,
       "step": 31925
     },
     {
-      "epoch": 0.8768229381230946,
+      "epoch": 0.9059591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 7.852982111165619e-07,
-      "loss": 0.7963,
+      "learning_rate": 4.6027749734667837e-07,
+      "loss": 0.9222,
       "step": 31926
     },
     {
-      "epoch": 0.8768504023509379,
+      "epoch": 0.9059875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 7.849527158705828e-07,
-      "loss": 0.7631,
+      "learning_rate": 4.60001910821023e-07,
+      "loss": 0.866,
       "step": 31927
     },
     {
-      "epoch": 0.8768778665787811,
+      "epoch": 0.9060158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 7.846072935376281e-07,
-      "loss": 0.8648,
+      "learning_rate": 4.5972640488015975e-07,
+      "loss": 0.8458,
       "step": 31928
     },
     {
-      "epoch": 0.8769053308066244,
+      "epoch": 0.906044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 7.842619441204313e-07,
-      "loss": 0.7288,
+      "learning_rate": 4.594509795264179e-07,
+      "loss": 0.8058,
       "step": 31929
     },
     {
-      "epoch": 0.8769327950344676,
+      "epoch": 0.906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 7.839166676217224e-07,
-      "loss": 0.8039,
+      "learning_rate": 4.5917563476212123e-07,
+      "loss": 0.8121,
       "step": 31930
     },
     {
-      "epoch": 0.8769602592623108,
+      "epoch": 0.9061010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 7.83571464044236e-07,
-      "loss": 0.758,
+      "learning_rate": 4.589003705895967e-07,
+      "loss": 0.81,
       "step": 31931
     },
     {
-      "epoch": 0.8769877234901541,
+      "epoch": 0.9061293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 7.832263333907008e-07,
-      "loss": 0.7942,
+      "learning_rate": 4.586251870111713e-07,
+      "loss": 0.8249,
       "step": 31932
     },
     {
-      "epoch": 0.8770151877179974,
+      "epoch": 0.9061577752553917,
       "grad_norm": 0.0,
-      "learning_rate": 7.828812756638493e-07,
-      "loss": 0.8949,
+      "learning_rate": 4.583500840291666e-07,
+      "loss": 0.835,
       "step": 31933
     },
     {
-      "epoch": 0.8770426519458405,
+      "epoch": 0.9061861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 7.825362908664125e-07,
-      "loss": 0.8611,
+      "learning_rate": 4.5807506164590843e-07,
+      "loss": 0.8169,
       "step": 31934
     },
     {
-      "epoch": 0.8770701161736838,
+      "epoch": 0.9062145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 7.821913790011193e-07,
-      "loss": 0.7888,
+      "learning_rate": 4.578001198637183e-07,
+      "loss": 0.7754,
       "step": 31935
     },
     {
-      "epoch": 0.877097580401527,
+      "epoch": 0.9062429057888762,
       "grad_norm": 0.0,
-      "learning_rate": 7.818465400706976e-07,
-      "loss": 0.7344,
+      "learning_rate": 4.5752525868491883e-07,
+      "loss": 0.861,
       "step": 31936
     },
     {
-      "epoch": 0.8771250446293702,
+      "epoch": 0.9062712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 7.815017740778763e-07,
-      "loss": 0.8296,
+      "learning_rate": 4.5725047811183475e-07,
+      "loss": 0.8506,
       "step": 31937
     },
     {
-      "epoch": 0.8771525088572135,
+      "epoch": 0.9062996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 7.811570810253855e-07,
-      "loss": 0.7817,
+      "learning_rate": 4.569757781467832e-07,
+      "loss": 0.8009,
       "step": 31938
     },
     {
-      "epoch": 0.8771799730850567,
+      "epoch": 0.9063280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 7.808124609159496e-07,
-      "loss": 0.8347,
+      "learning_rate": 4.5670115879208663e-07,
+      "loss": 0.8113,
       "step": 31939
     },
     {
-      "epoch": 0.8772074373129,
+      "epoch": 0.9063564131668559,
       "grad_norm": 0.0,
-      "learning_rate": 7.804679137522974e-07,
-      "loss": 0.7915,
+      "learning_rate": 4.5642662005006556e-07,
+      "loss": 0.8637,
       "step": 31940
     },
     {
-      "epoch": 0.8772349015407431,
+      "epoch": 0.9063847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 7.801234395371549e-07,
-      "loss": 0.7291,
+      "learning_rate": 4.561521619230358e-07,
+      "loss": 0.7771,
       "step": 31941
     },
     {
-      "epoch": 0.8772623657685864,
+      "epoch": 0.9064131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 7.797790382732462e-07,
-      "loss": 0.7948,
+      "learning_rate": 4.5587778441331887e-07,
+      "loss": 0.7859,
       "step": 31942
     },
     {
-      "epoch": 0.8772898299964297,
+      "epoch": 0.9064415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 7.79434709963297e-07,
-      "loss": 0.7726,
+      "learning_rate": 4.556034875232318e-07,
+      "loss": 0.8177,
       "step": 31943
     },
     {
-      "epoch": 0.8773172942242728,
+      "epoch": 0.9064699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 7.790904546100319e-07,
-      "loss": 0.8051,
+      "learning_rate": 4.553292712550894e-07,
+      "loss": 0.6723,
       "step": 31944
     },
     {
-      "epoch": 0.8773447584521161,
+      "epoch": 0.9064982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 7.787462722161754e-07,
-      "loss": 0.8029,
+      "learning_rate": 4.5505513561121096e-07,
+      "loss": 0.7466,
       "step": 31945
     },
     {
-      "epoch": 0.8773722226799594,
+      "epoch": 0.9065266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 7.784021627844518e-07,
-      "loss": 0.7763,
+      "learning_rate": 4.5478108059391123e-07,
+      "loss": 0.72,
       "step": 31946
     },
     {
-      "epoch": 0.8773996869078026,
+      "epoch": 0.9065550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 7.780581263175823e-07,
-      "loss": 0.7399,
+      "learning_rate": 4.5450710620550395e-07,
+      "loss": 0.7738,
       "step": 31947
     },
     {
-      "epoch": 0.8774271511356458,
+      "epoch": 0.906583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 7.777141628182872e-07,
-      "loss": 0.8362,
+      "learning_rate": 4.54233212448304e-07,
+      "loss": 0.8223,
       "step": 31948
     },
     {
-      "epoch": 0.877454615363489,
+      "epoch": 0.9066118047673098,
       "grad_norm": 0.0,
-      "learning_rate": 7.773702722892907e-07,
-      "loss": 0.8619,
+      "learning_rate": 4.539593993246272e-07,
+      "loss": 0.8965,
       "step": 31949
     },
     {
-      "epoch": 0.8774820795913323,
+      "epoch": 0.9066401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 7.770264547333128e-07,
-      "loss": 0.874,
+      "learning_rate": 4.536856668367806e-07,
+      "loss": 0.8159,
       "step": 31950
     },
     {
-      "epoch": 0.8775095438191756,
+      "epoch": 0.9066685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 7.766827101530739e-07,
-      "loss": 0.8543,
+      "learning_rate": 4.5341201498708464e-07,
+      "loss": 0.7668,
       "step": 31951
     },
     {
-      "epoch": 0.8775370080470187,
+      "epoch": 0.9066969353007945,
       "grad_norm": 0.0,
-      "learning_rate": 7.763390385512959e-07,
-      "loss": 0.8525,
+      "learning_rate": 4.5313844377784413e-07,
+      "loss": 0.9172,
       "step": 31952
     },
     {
-      "epoch": 0.877564472274862,
+      "epoch": 0.9067253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 7.759954399306957e-07,
-      "loss": 0.7866,
+      "learning_rate": 4.528649532113727e-07,
+      "loss": 0.7153,
       "step": 31953
     },
     {
-      "epoch": 0.8775919365027052,
+      "epoch": 0.9067536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 7.756519142939912e-07,
-      "loss": 0.8336,
+      "learning_rate": 4.5259154328998077e-07,
+      "loss": 0.623,
       "step": 31954
     },
     {
-      "epoch": 0.8776194007305484,
+      "epoch": 0.9067820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 7.753084616439044e-07,
-      "loss": 0.8084,
+      "learning_rate": 4.523182140159765e-07,
+      "loss": 0.8066,
       "step": 31955
     },
     {
-      "epoch": 0.8776468649583917,
+      "epoch": 0.9068104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 7.749650819831489e-07,
-      "loss": 0.9005,
+      "learning_rate": 4.520449653916692e-07,
+      "loss": 0.7312,
       "step": 31956
     },
     {
-      "epoch": 0.8776743291862349,
+      "epoch": 0.9068388195232691,
       "grad_norm": 0.0,
-      "learning_rate": 7.746217753144425e-07,
-      "loss": 0.8406,
+      "learning_rate": 4.5177179741936917e-07,
+      "loss": 0.8661,
       "step": 31957
     },
     {
-      "epoch": 0.8777017934140782,
+      "epoch": 0.9068671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 7.742785416405041e-07,
-      "loss": 0.7752,
+      "learning_rate": 4.514987101013801e-07,
+      "loss": 0.8707,
       "step": 31958
     },
     {
-      "epoch": 0.8777292576419214,
+      "epoch": 0.9068955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 7.739353809640459e-07,
-      "loss": 0.8317,
+      "learning_rate": 4.5122570344001136e-07,
+      "loss": 0.741,
       "step": 31959
     },
     {
-      "epoch": 0.8777567218697646,
+      "epoch": 0.9069239500567536,
       "grad_norm": 0.0,
-      "learning_rate": 7.735922932877848e-07,
-      "loss": 0.7531,
+      "learning_rate": 4.509527774375688e-07,
+      "loss": 0.8457,
       "step": 31960
     },
     {
-      "epoch": 0.8777841860976079,
+      "epoch": 0.9069523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 7.732492786144353e-07,
-      "loss": 0.8893,
+      "learning_rate": 4.5067993209635617e-07,
+      "loss": 0.8396,
       "step": 31961
     },
     {
-      "epoch": 0.877811650325451,
+      "epoch": 0.9069807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 7.729063369467138e-07,
-      "loss": 0.7953,
+      "learning_rate": 4.504071674186805e-07,
+      "loss": 0.8767,
       "step": 31962
     },
     {
-      "epoch": 0.8778391145532943,
+      "epoch": 0.9070090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 7.725634682873295e-07,
-      "loss": 0.7478,
+      "learning_rate": 4.501344834068444e-07,
+      "loss": 0.7731,
       "step": 31963
     },
     {
-      "epoch": 0.8778665787811376,
+      "epoch": 0.9070374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 7.722206726389991e-07,
-      "loss": 0.7887,
+      "learning_rate": 4.4986188006315157e-07,
+      "loss": 0.8615,
       "step": 31964
     },
     {
-      "epoch": 0.8778940430089808,
+      "epoch": 0.9070658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 7.718779500044316e-07,
-      "loss": 0.7901,
+      "learning_rate": 4.495893573899035e-07,
+      "loss": 0.6938,
       "step": 31965
     },
     {
-      "epoch": 0.877921507236824,
+      "epoch": 0.907094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 7.715353003863402e-07,
-      "loss": 0.8674,
+      "learning_rate": 4.49316915389405e-07,
+      "loss": 0.7275,
       "step": 31966
     },
     {
-      "epoch": 0.8779489714646672,
+      "epoch": 0.907122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 7.711927237874361e-07,
-      "loss": 0.8962,
+      "learning_rate": 4.4904455406395544e-07,
+      "loss": 0.7841,
       "step": 31967
     },
     {
-      "epoch": 0.8779764356925105,
+      "epoch": 0.9071509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 7.708502202104307e-07,
-      "loss": 0.8045,
+      "learning_rate": 4.487722734158573e-07,
+      "loss": 0.7981,
       "step": 31968
     },
     {
-      "epoch": 0.8780038999203538,
+      "epoch": 0.9071793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 7.705077896580326e-07,
-      "loss": 0.8841,
+      "learning_rate": 4.4850007344740765e-07,
+      "loss": 0.8396,
       "step": 31969
     },
     {
-      "epoch": 0.8780313641481969,
+      "epoch": 0.9072077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 7.701654321329543e-07,
-      "loss": 0.703,
+      "learning_rate": 4.482279541609069e-07,
+      "loss": 0.8311,
       "step": 31970
     },
     {
-      "epoch": 0.8780588283760402,
+      "epoch": 0.9072360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 7.698231476379026e-07,
-      "loss": 0.8634,
+      "learning_rate": 4.479559155586555e-07,
+      "loss": 0.7697,
       "step": 31971
     },
     {
-      "epoch": 0.8780862926038835,
+      "epoch": 0.9072644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 7.69480936175584e-07,
-      "loss": 0.8022,
+      "learning_rate": 4.4768395764294814e-07,
+      "loss": 0.7911,
       "step": 31972
     },
     {
-      "epoch": 0.8781137568317267,
+      "epoch": 0.9072928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 7.691387977487075e-07,
-      "loss": 0.7525,
+      "learning_rate": 4.474120804160842e-07,
+      "loss": 0.73,
       "step": 31973
     },
     {
-      "epoch": 0.8781412210595699,
+      "epoch": 0.9073212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 7.687967323599798e-07,
-      "loss": 0.7477,
+      "learning_rate": 4.4714028388036066e-07,
+      "loss": 0.8204,
       "step": 31974
     },
     {
-      "epoch": 0.8781686852874131,
+      "epoch": 0.9073496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 7.684547400121089e-07,
-      "loss": 0.7287,
+      "learning_rate": 4.4686856803807246e-07,
+      "loss": 0.8979,
       "step": 31975
     },
     {
-      "epoch": 0.8781961495152564,
+      "epoch": 0.9073779795686719,
       "grad_norm": 0.0,
-      "learning_rate": 7.681128207078015e-07,
-      "loss": 0.8002,
+      "learning_rate": 4.465969328915143e-07,
+      "loss": 0.6613,
       "step": 31976
     },
     {
-      "epoch": 0.8782236137430997,
+      "epoch": 0.9074063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 7.677709744497608e-07,
-      "loss": 0.7631,
+      "learning_rate": 4.463253784429822e-07,
+      "loss": 0.8432,
       "step": 31977
     },
     {
-      "epoch": 0.8782510779709428,
+      "epoch": 0.9074347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 7.674292012406914e-07,
-      "loss": 0.7941,
+      "learning_rate": 4.4605390469476763e-07,
+      "loss": 0.708,
       "step": 31978
     },
     {
-      "epoch": 0.8782785421987861,
+      "epoch": 0.9074631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 7.670875010833001e-07,
-      "loss": 0.8855,
+      "learning_rate": 4.4578251164916654e-07,
+      "loss": 0.791,
       "step": 31979
     },
     {
-      "epoch": 0.8783060064266293,
+      "epoch": 0.9074914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 7.66745873980288e-07,
-      "loss": 0.7328,
+      "learning_rate": 4.455111993084693e-07,
+      "loss": 0.793,
       "step": 31980
     },
     {
-      "epoch": 0.8783334706544725,
+      "epoch": 0.9075198637911465,
       "grad_norm": 0.0,
-      "learning_rate": 7.664043199343585e-07,
-      "loss": 0.8333,
+      "learning_rate": 4.452399676749675e-07,
+      "loss": 0.888,
       "step": 31981
     },
     {
-      "epoch": 0.8783609348823158,
+      "epoch": 0.9075482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 7.660628389482172e-07,
-      "loss": 0.8187,
+      "learning_rate": 4.4496881675095473e-07,
+      "loss": 0.8062,
       "step": 31982
     },
     {
-      "epoch": 0.878388399110159,
+      "epoch": 0.9075766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 7.657214310245608e-07,
-      "loss": 0.7613,
+      "learning_rate": 4.4469774653872035e-07,
+      "loss": 0.889,
       "step": 31983
     },
     {
-      "epoch": 0.8784158633380023,
+      "epoch": 0.9076049943246312,
       "grad_norm": 0.0,
-      "learning_rate": 7.65380096166094e-07,
-      "loss": 0.8509,
+      "learning_rate": 4.444267570405525e-07,
+      "loss": 0.8501,
       "step": 31984
     },
     {
-      "epoch": 0.8784433275658455,
+      "epoch": 0.907633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 7.650388343755177e-07,
-      "loss": 0.6997,
+      "learning_rate": 4.4415584825874157e-07,
+      "loss": 0.8059,
       "step": 31985
     },
     {
-      "epoch": 0.8784707917936887,
+      "epoch": 0.9076617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 7.646976456555321e-07,
-      "loss": 0.8324,
+      "learning_rate": 4.4388502019557686e-07,
+      "loss": 0.7628,
       "step": 31986
     },
     {
-      "epoch": 0.878498256021532,
+      "epoch": 0.9076901248581157,
       "grad_norm": 0.0,
-      "learning_rate": 7.643565300088351e-07,
-      "loss": 0.8968,
+      "learning_rate": 4.4361427285334326e-07,
+      "loss": 0.8773,
       "step": 31987
     },
     {
-      "epoch": 0.8785257202493751,
+      "epoch": 0.9077185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 7.640154874381278e-07,
-      "loss": 0.8227,
+      "learning_rate": 4.433436062343299e-07,
+      "loss": 0.7652,
       "step": 31988
     },
     {
-      "epoch": 0.8785531844772184,
+      "epoch": 0.9077468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 7.636745179461069e-07,
-      "loss": 0.7618,
+      "learning_rate": 4.43073020340824e-07,
+      "loss": 0.8287,
       "step": 31989
     },
     {
-      "epoch": 0.8785806487050617,
+      "epoch": 0.9077752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 7.633336215354704e-07,
-      "loss": 0.7739,
+      "learning_rate": 4.4280251517510917e-07,
+      "loss": 0.8086,
       "step": 31990
     },
     {
-      "epoch": 0.8786081129329049,
+      "epoch": 0.9078036322360954,
       "grad_norm": 0.0,
-      "learning_rate": 7.629927982089158e-07,
-      "loss": 0.7437,
+      "learning_rate": 4.4253209073947034e-07,
+      "loss": 0.7722,
       "step": 31991
     },
     {
-      "epoch": 0.8786355771607481,
+      "epoch": 0.9078320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 7.626520479691412e-07,
-      "loss": 0.7623,
+      "learning_rate": 4.422617470361945e-07,
+      "loss": 0.7628,
       "step": 31992
     },
     {
-      "epoch": 0.8786630413885914,
+      "epoch": 0.9078603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 7.623113708188412e-07,
-      "loss": 0.8485,
+      "learning_rate": 4.41991484067561e-07,
+      "loss": 0.884,
       "step": 31993
     },
     {
-      "epoch": 0.8786905056164346,
+      "epoch": 0.90788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 7.619707667607146e-07,
-      "loss": 0.9051,
+      "learning_rate": 4.417213018358579e-07,
+      "loss": 0.8037,
       "step": 31994
     },
     {
-      "epoch": 0.8787179698442779,
+      "epoch": 0.9079171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 7.616302357974525e-07,
-      "loss": 0.7306,
+      "learning_rate": 4.414512003433624e-07,
+      "loss": 0.8706,
       "step": 31995
     },
     {
-      "epoch": 0.878745434072121,
+      "epoch": 0.9079455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 7.612897779317496e-07,
-      "loss": 0.7324,
+      "learning_rate": 4.4118117959235706e-07,
+      "loss": 0.7527,
       "step": 31996
     },
     {
-      "epoch": 0.8787728982999643,
+      "epoch": 0.9079738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 7.609493931663003e-07,
-      "loss": 0.7266,
+      "learning_rate": 4.409112395851267e-07,
+      "loss": 0.8399,
       "step": 31997
     },
     {
-      "epoch": 0.8788003625278076,
+      "epoch": 0.9080022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 7.606090815037992e-07,
-      "loss": 0.8009,
+      "learning_rate": 4.406413803239473e-07,
+      "loss": 0.789,
       "step": 31998
     },
     {
-      "epoch": 0.8788278267556507,
+      "epoch": 0.9080306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 7.602688429469374e-07,
-      "loss": 0.7743,
+      "learning_rate": 4.403716018111004e-07,
+      "loss": 0.8851,
       "step": 31999
     },
     {
-      "epoch": 0.878855290983494,
+      "epoch": 0.9080590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 7.599286774984094e-07,
-      "loss": 0.8224,
+      "learning_rate": 4.401019040488652e-07,
+      "loss": 0.7789,
       "step": 32000
     },
     {
-      "epoch": 0.8788827552113372,
+      "epoch": 0.9080874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 7.595885851609031e-07,
-      "loss": 0.8646,
+      "learning_rate": 4.3983228703951777e-07,
+      "loss": 0.7342,
       "step": 32001
     },
     {
-      "epoch": 0.8789102194391805,
+      "epoch": 0.9081157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 7.592485659371107e-07,
-      "loss": 0.7518,
+      "learning_rate": 4.3956275078533615e-07,
+      "loss": 0.8101,
       "step": 32002
     },
     {
-      "epoch": 0.8789376836670237,
+      "epoch": 0.908144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 7.589086198297257e-07,
-      "loss": 0.7516,
+      "learning_rate": 4.3929329528859974e-07,
+      "loss": 0.8604,
       "step": 32003
     },
     {
-      "epoch": 0.8789651478948669,
+      "epoch": 0.9081725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 7.585687468414338e-07,
-      "loss": 0.907,
+      "learning_rate": 4.390239205515812e-07,
+      "loss": 0.8509,
       "step": 32004
     },
     {
-      "epoch": 0.8789926121227102,
+      "epoch": 0.9082009080590239,
       "grad_norm": 0.0,
-      "learning_rate": 7.582289469749248e-07,
-      "loss": 0.8804,
+      "learning_rate": 4.387546265765574e-07,
+      "loss": 0.8934,
       "step": 32005
     },
     {
-      "epoch": 0.8790200763505535,
+      "epoch": 0.9082292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 7.578892202328913e-07,
-      "loss": 0.7807,
+      "learning_rate": 4.384854133658045e-07,
+      "loss": 0.8448,
       "step": 32006
     },
     {
-      "epoch": 0.8790475405783966,
+      "epoch": 0.9082576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 7.575495666180155e-07,
-      "loss": 0.8148,
+      "learning_rate": 4.3821628092159394e-07,
+      "loss": 0.8215,
       "step": 32007
     },
     {
-      "epoch": 0.8790750048062399,
+      "epoch": 0.9082860385925086,
       "grad_norm": 0.0,
-      "learning_rate": 7.572099861329885e-07,
-      "loss": 0.7191,
+      "learning_rate": 4.379472292461995e-07,
+      "loss": 0.8047,
       "step": 32008
     },
     {
-      "epoch": 0.8791024690340831,
+      "epoch": 0.9083144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 7.568704787804959e-07,
-      "loss": 0.8843,
+      "learning_rate": 4.3767825834189703e-07,
+      "loss": 0.6979,
       "step": 32009
     },
     {
-      "epoch": 0.8791299332619263,
+      "epoch": 0.9083427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 7.565310445632268e-07,
-      "loss": 0.7685,
+      "learning_rate": 4.374093682109537e-07,
+      "loss": 0.861,
       "step": 32010
     },
     {
-      "epoch": 0.8791573974897696,
+      "epoch": 0.9083711691259931,
       "grad_norm": 0.0,
-      "learning_rate": 7.561916834838623e-07,
-      "loss": 0.8462,
+      "learning_rate": 4.3714055885564433e-07,
+      "loss": 0.787,
       "step": 32011
     },
     {
-      "epoch": 0.8791848617176128,
+      "epoch": 0.9083995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 7.558523955450914e-07,
-      "loss": 0.8704,
+      "learning_rate": 4.368718302782382e-07,
+      "loss": 0.7497,
       "step": 32012
     },
     {
-      "epoch": 0.8792123259454561,
+      "epoch": 0.908427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 7.555131807495964e-07,
-      "loss": 0.8053,
+      "learning_rate": 4.3660318248100575e-07,
+      "loss": 0.8012,
       "step": 32013
     },
     {
-      "epoch": 0.8792397901732992,
+      "epoch": 0.9084562996594778,
       "grad_norm": 0.0,
-      "learning_rate": 7.551740391000617e-07,
-      "loss": 0.8076,
+      "learning_rate": 4.3633461546621737e-07,
+      "loss": 0.7728,
       "step": 32014
     },
     {
-      "epoch": 0.8792672544011425,
+      "epoch": 0.9084846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 7.548349705991709e-07,
-      "loss": 0.8066,
+      "learning_rate": 4.36066129236139e-07,
+      "loss": 0.8416,
       "step": 32015
     },
     {
-      "epoch": 0.8792947186289858,
+      "epoch": 0.9085130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 7.544959752496073e-07,
-      "loss": 0.8088,
+      "learning_rate": 4.3579772379303997e-07,
+      "loss": 0.7862,
       "step": 32016
     },
     {
-      "epoch": 0.879322182856829,
+      "epoch": 0.9085414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 7.541570530540543e-07,
-      "loss": 0.7818,
+      "learning_rate": 4.3552939913918845e-07,
+      "loss": 0.8378,
       "step": 32017
     },
     {
-      "epoch": 0.8793496470846722,
+      "epoch": 0.9085698070374575,
       "grad_norm": 0.0,
-      "learning_rate": 7.538182040151909e-07,
-      "loss": 0.7568,
+      "learning_rate": 4.3526115527684933e-07,
+      "loss": 0.6988,
       "step": 32018
     },
     {
-      "epoch": 0.8793771113125155,
+      "epoch": 0.9085981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 7.534794281357017e-07,
-      "loss": 0.7739,
+      "learning_rate": 4.349929922082896e-07,
+      "loss": 0.8493,
       "step": 32019
     },
     {
-      "epoch": 0.8794045755403587,
+      "epoch": 0.9086265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 7.531407254182621e-07,
-      "loss": 0.8678,
+      "learning_rate": 4.347249099357742e-07,
+      "loss": 0.7967,
       "step": 32020
     },
     {
-      "epoch": 0.879432039768202,
+      "epoch": 0.9086549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 7.528020958655558e-07,
-      "loss": 0.8018,
+      "learning_rate": 4.3445690846156683e-07,
+      "loss": 0.8094,
       "step": 32021
     },
     {
-      "epoch": 0.8794595039960451,
+      "epoch": 0.908683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 7.524635394802605e-07,
-      "loss": 0.78,
+      "learning_rate": 4.341889877879313e-07,
+      "loss": 0.8475,
       "step": 32022
     },
     {
-      "epoch": 0.8794869682238884,
+      "epoch": 0.9087116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 7.521250562650584e-07,
-      "loss": 0.8407,
+      "learning_rate": 4.339211479171335e-07,
+      "loss": 0.8484,
       "step": 32023
     },
     {
-      "epoch": 0.8795144324517317,
+      "epoch": 0.9087400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 7.517866462226231e-07,
-      "loss": 0.8126,
+      "learning_rate": 4.3365338885143273e-07,
+      "loss": 0.841,
       "step": 32024
     },
     {
-      "epoch": 0.8795418966795748,
+      "epoch": 0.9087684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 7.514483093556346e-07,
-      "loss": 0.8064,
+      "learning_rate": 4.3338571059309056e-07,
+      "loss": 0.7736,
       "step": 32025
     },
     {
-      "epoch": 0.8795693609074181,
+      "epoch": 0.9087968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 7.511100456667686e-07,
-      "loss": 0.8418,
+      "learning_rate": 4.3311811314437067e-07,
+      "loss": 0.7869,
       "step": 32026
     },
     {
-      "epoch": 0.8795968251352613,
+      "epoch": 0.9088251986379114,
       "grad_norm": 0.0,
-      "learning_rate": 7.507718551587051e-07,
-      "loss": 0.8148,
+      "learning_rate": 4.3285059650753135e-07,
+      "loss": 0.8852,
       "step": 32027
     },
     {
-      "epoch": 0.8796242893631046,
+      "epoch": 0.9088535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 7.504337378341165e-07,
-      "loss": 0.8214,
+      "learning_rate": 4.325831606848352e-07,
+      "loss": 0.7766,
       "step": 32028
     },
     {
-      "epoch": 0.8796517535909478,
+      "epoch": 0.9088819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 7.500956936956793e-07,
-      "loss": 0.8832,
+      "learning_rate": 4.3231580567853705e-07,
+      "loss": 0.7466,
       "step": 32029
     },
     {
-      "epoch": 0.879679217818791,
+      "epoch": 0.9089103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 7.497577227460673e-07,
-      "loss": 0.7426,
+      "learning_rate": 4.3204853149089733e-07,
+      "loss": 0.8268,
       "step": 32030
     },
     {
-      "epoch": 0.8797066820466343,
+      "epoch": 0.908938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 7.494198249879548e-07,
-      "loss": 0.7617,
+      "learning_rate": 4.3178133812417535e-07,
+      "loss": 0.8028,
       "step": 32031
     },
     {
-      "epoch": 0.8797341462744775,
+      "epoch": 0.908967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 7.490820004240162e-07,
-      "loss": 0.8322,
+      "learning_rate": 4.3151422558062595e-07,
+      "loss": 0.8184,
       "step": 32032
     },
     {
-      "epoch": 0.8797616105023207,
+      "epoch": 0.9089954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 7.487442490569252e-07,
-      "loss": 0.8012,
+      "learning_rate": 4.3124719386250515e-07,
+      "loss": 0.7209,
       "step": 32033
     },
     {
-      "epoch": 0.879789074730164,
+      "epoch": 0.9090238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 7.48406570889354e-07,
-      "loss": 0.8674,
+      "learning_rate": 4.309802429720711e-07,
+      "loss": 0.8292,
       "step": 32034
     },
     {
-      "epoch": 0.8798165389580072,
+      "epoch": 0.9090522133938707,
       "grad_norm": 0.0,
-      "learning_rate": 7.480689659239714e-07,
-      "loss": 0.8614,
+      "learning_rate": 4.3071337291157535e-07,
+      "loss": 0.8379,
       "step": 32035
     },
     {
-      "epoch": 0.8798440031858504,
+      "epoch": 0.9090805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 7.477314341634545e-07,
-      "loss": 0.788,
+      "learning_rate": 4.3044658368327383e-07,
+      "loss": 0.8322,
       "step": 32036
     },
     {
-      "epoch": 0.8798714674136937,
+      "epoch": 0.9091089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 7.473939756104675e-07,
-      "loss": 0.7948,
+      "learning_rate": 4.301798752894215e-07,
+      "loss": 0.8069,
       "step": 32037
     },
     {
-      "epoch": 0.8798989316415369,
+      "epoch": 0.9091373439273552,
       "grad_norm": 0.0,
-      "learning_rate": 7.470565902676841e-07,
-      "loss": 0.7375,
+      "learning_rate": 4.299132477322698e-07,
+      "loss": 0.7182,
       "step": 32038
     },
     {
-      "epoch": 0.8799263958693802,
+      "epoch": 0.9091657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 7.46719278137773e-07,
-      "loss": 0.7831,
+      "learning_rate": 4.2964670101407037e-07,
+      "loss": 0.8797,
       "step": 32039
     },
     {
-      "epoch": 0.8799538600972233,
+      "epoch": 0.9091940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 7.463820392234022e-07,
-      "loss": 0.9138,
+      "learning_rate": 4.2938023513707683e-07,
+      "loss": 0.7064,
       "step": 32040
     },
     {
-      "epoch": 0.8799813243250666,
+      "epoch": 0.9092224744608399,
       "grad_norm": 0.0,
-      "learning_rate": 7.460448735272441e-07,
-      "loss": 0.7284,
+      "learning_rate": 4.2911385010353744e-07,
+      "loss": 0.7981,
       "step": 32041
     },
     {
-      "epoch": 0.8800087885529099,
+      "epoch": 0.9092508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 7.457077810519619e-07,
-      "loss": 0.8568,
+      "learning_rate": 4.2884754591570267e-07,
+      "loss": 0.8558,
       "step": 32042
     },
     {
-      "epoch": 0.880036252780753,
+      "epoch": 0.9092792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 7.453707618002259e-07,
-      "loss": 0.8038,
+      "learning_rate": 4.28581322575824e-07,
+      "loss": 0.8775,
       "step": 32043
     },
     {
-      "epoch": 0.8800637170085963,
+      "epoch": 0.9093076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 7.450338157747006e-07,
-      "loss": 0.8835,
+      "learning_rate": 4.283151800861485e-07,
+      "loss": 0.8826,
       "step": 32044
     },
     {
-      "epoch": 0.8800911812364396,
+      "epoch": 0.9093359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 7.446969429780515e-07,
-      "loss": 0.8097,
+      "learning_rate": 4.280491184489266e-07,
+      "loss": 0.889,
       "step": 32045
     },
     {
-      "epoch": 0.8801186454642828,
+      "epoch": 0.9093643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 7.443601434129455e-07,
-      "loss": 0.9132,
+      "learning_rate": 4.277831376664032e-07,
+      "loss": 0.7959,
       "step": 32046
     },
     {
-      "epoch": 0.880146109692126,
+      "epoch": 0.9093927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 7.440234170820492e-07,
-      "loss": 0.7506,
+      "learning_rate": 4.2751723774082543e-07,
+      "loss": 0.734,
       "step": 32047
     },
     {
-      "epoch": 0.8801735739199692,
+      "epoch": 0.9094211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 7.43686763988024e-07,
-      "loss": 0.9066,
+      "learning_rate": 4.2725141867444143e-07,
+      "loss": 0.7906,
       "step": 32048
     },
     {
-      "epoch": 0.8802010381478125,
+      "epoch": 0.9094494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 7.433501841335344e-07,
-      "loss": 0.6926,
+      "learning_rate": 4.2698568046949383e-07,
+      "loss": 0.8543,
       "step": 32049
     },
     {
-      "epoch": 0.8802285023756558,
+      "epoch": 0.909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 7.430136775212437e-07,
-      "loss": 0.8344,
+      "learning_rate": 4.267200231282276e-07,
+      "loss": 0.7293,
       "step": 32050
     },
     {
-      "epoch": 0.8802559666034989,
+      "epoch": 0.9095062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 7.426772441538155e-07,
-      "loss": 0.7878,
+      "learning_rate": 4.264544466528897e-07,
+      "loss": 0.7686,
       "step": 32051
     },
     {
-      "epoch": 0.8802834308313422,
+      "epoch": 0.9095346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 7.423408840339108e-07,
-      "loss": 0.7887,
+      "learning_rate": 4.2618895104572065e-07,
+      "loss": 0.6962,
       "step": 32052
     },
     {
-      "epoch": 0.8803108950591854,
+      "epoch": 0.9095629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 7.42004597164191e-07,
-      "loss": 0.8884,
+      "learning_rate": 4.259235363089631e-07,
+      "loss": 0.7646,
       "step": 32053
     },
     {
-      "epoch": 0.8803383592870286,
+      "epoch": 0.9095913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 7.416683835473171e-07,
-      "loss": 0.861,
+      "learning_rate": 4.2565820244486075e-07,
+      "loss": 0.7478,
       "step": 32054
     },
     {
-      "epoch": 0.8803658235148719,
+      "epoch": 0.9096197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 7.413322431859482e-07,
-      "loss": 0.8577,
+      "learning_rate": 4.2539294945565415e-07,
+      "loss": 0.809,
       "step": 32055
     },
     {
-      "epoch": 0.8803932877427151,
+      "epoch": 0.9096481271282634,
       "grad_norm": 0.0,
-      "learning_rate": 7.409961760827444e-07,
-      "loss": 0.6882,
+      "learning_rate": 4.251277773435825e-07,
+      "loss": 0.7354,
       "step": 32056
     },
     {
-      "epoch": 0.8804207519705584,
+      "epoch": 0.9096765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 7.406601822403669e-07,
-      "loss": 0.7127,
+      "learning_rate": 4.248626861108873e-07,
+      "loss": 0.7802,
       "step": 32057
     },
     {
-      "epoch": 0.8804482161984016,
+      "epoch": 0.9097048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 7.403242616614736e-07,
-      "loss": 0.84,
+      "learning_rate": 4.2459767575980694e-07,
+      "loss": 0.7786,
       "step": 32058
     },
     {
-      "epoch": 0.8804756804262448,
+      "epoch": 0.9097332576617481,
       "grad_norm": 0.0,
-      "learning_rate": 7.399884143487224e-07,
-      "loss": 0.8225,
+      "learning_rate": 4.243327462925828e-07,
+      "loss": 0.9494,
       "step": 32059
     },
     {
-      "epoch": 0.8805031446540881,
+      "epoch": 0.9097616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 7.396526403047677e-07,
-      "loss": 0.7866,
+      "learning_rate": 4.2406789771144876e-07,
+      "loss": 0.772,
       "step": 32060
     },
     {
-      "epoch": 0.8805306088819312,
+      "epoch": 0.9097900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 7.393169395322686e-07,
-      "loss": 0.8178,
+      "learning_rate": 4.238031300186429e-07,
+      "loss": 0.777,
       "step": 32061
     },
     {
-      "epoch": 0.8805580731097745,
+      "epoch": 0.9098183881952326,
       "grad_norm": 0.0,
-      "learning_rate": 7.389813120338807e-07,
-      "loss": 0.7268,
+      "learning_rate": 4.235384432164047e-07,
+      "loss": 0.7256,
       "step": 32062
     },
     {
-      "epoch": 0.8805855373376178,
+      "epoch": 0.9098467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 7.386457578122597e-07,
-      "loss": 0.7573,
+      "learning_rate": 4.232738373069656e-07,
+      "loss": 0.8195,
       "step": 32063
     },
     {
-      "epoch": 0.880613001565461,
+      "epoch": 0.9098751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 7.383102768700612e-07,
-      "loss": 0.8708,
+      "learning_rate": 4.230093122925638e-07,
+      "loss": 0.773,
       "step": 32064
     },
     {
-      "epoch": 0.8806404657933042,
+      "epoch": 0.9099035187287173,
       "grad_norm": 0.0,
-      "learning_rate": 7.37974869209942e-07,
-      "loss": 0.7897,
+      "learning_rate": 4.2274486817543314e-07,
+      "loss": 0.8252,
       "step": 32065
     },
     {
-      "epoch": 0.8806679300211474,
+      "epoch": 0.9099318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 7.376395348345511e-07,
-      "loss": 0.7199,
+      "learning_rate": 4.2248050495780737e-07,
+      "loss": 0.9028,
       "step": 32066
     },
     {
-      "epoch": 0.8806953942489907,
+      "epoch": 0.9099602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 7.373042737465453e-07,
-      "loss": 0.8195,
+      "learning_rate": 4.22216222641918e-07,
+      "loss": 0.8006,
       "step": 32067
     },
     {
-      "epoch": 0.880722858476834,
+      "epoch": 0.909988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 7.369690859485757e-07,
-      "loss": 0.8171,
+      "learning_rate": 4.219520212299999e-07,
+      "loss": 0.7729,
       "step": 32068
     },
     {
-      "epoch": 0.8807503227046771,
+      "epoch": 0.9100170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 7.366339714432935e-07,
-      "loss": 0.8192,
+      "learning_rate": 4.2168790072428354e-07,
+      "loss": 0.6617,
       "step": 32069
     },
     {
-      "epoch": 0.8807777869325204,
+      "epoch": 0.9100454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 7.362989302333523e-07,
-      "loss": 0.7957,
+      "learning_rate": 4.214238611269994e-07,
+      "loss": 0.6993,
       "step": 32070
     },
     {
-      "epoch": 0.8808052511603637,
+      "epoch": 0.9100737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 7.359639623214043e-07,
-      "loss": 0.851,
+      "learning_rate": 4.2115990244037895e-07,
+      "loss": 0.7186,
       "step": 32071
     },
     {
-      "epoch": 0.8808327153882068,
+      "epoch": 0.9101021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 7.356290677100952e-07,
-      "loss": 0.7885,
+      "learning_rate": 4.2089602466665045e-07,
+      "loss": 0.7501,
       "step": 32072
     },
     {
-      "epoch": 0.8808601796160501,
+      "epoch": 0.9101305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 7.352942464020785e-07,
-      "loss": 0.6897,
+      "learning_rate": 4.2063222780804435e-07,
+      "loss": 0.803,
       "step": 32073
     },
     {
-      "epoch": 0.8808876438438933,
+      "epoch": 0.9101589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 7.349594984000019e-07,
-      "loss": 0.8931,
+      "learning_rate": 4.203685118667888e-07,
+      "loss": 0.8168,
       "step": 32074
     },
     {
-      "epoch": 0.8809151080717366,
+      "epoch": 0.9101872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 7.346248237065168e-07,
-      "loss": 0.7236,
+      "learning_rate": 4.2010487684511105e-07,
+      "loss": 0.6974,
       "step": 32075
     },
     {
-      "epoch": 0.8809425722995798,
+      "epoch": 0.9102156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 7.342902223242698e-07,
-      "loss": 0.8026,
+      "learning_rate": 4.1984132274523913e-07,
+      "loss": 0.895,
       "step": 32076
     },
     {
-      "epoch": 0.880970036527423,
+      "epoch": 0.9102440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 7.339556942559056e-07,
-      "loss": 0.83,
+      "learning_rate": 4.1957784956939697e-07,
+      "loss": 0.8373,
       "step": 32077
     },
     {
-      "epoch": 0.8809975007552663,
+      "epoch": 0.910272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 7.336212395040731e-07,
-      "loss": 0.9023,
+      "learning_rate": 4.1931445731981045e-07,
+      "loss": 0.8619,
       "step": 32078
     },
     {
-      "epoch": 0.8810249649831094,
+      "epoch": 0.9103007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 7.332868580714181e-07,
-      "loss": 0.8045,
+      "learning_rate": 4.1905114599870786e-07,
+      "loss": 0.7893,
       "step": 32079
     },
     {
-      "epoch": 0.8810524292109527,
+      "epoch": 0.9103291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 7.329525499605882e-07,
-      "loss": 0.9673,
+      "learning_rate": 4.187879156083086e-07,
+      "loss": 0.8571,
       "step": 32080
     },
     {
-      "epoch": 0.881079893438796,
+      "epoch": 0.9103575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 7.326183151742272e-07,
-      "loss": 0.9001,
+      "learning_rate": 4.1852476615083957e-07,
+      "loss": 0.8191,
       "step": 32081
     },
     {
-      "epoch": 0.8811073576666392,
+      "epoch": 0.9103859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 7.322841537149816e-07,
-      "loss": 0.706,
+      "learning_rate": 4.1826169762852365e-07,
+      "loss": 0.7827,
       "step": 32082
     },
     {
-      "epoch": 0.8811348218944824,
+      "epoch": 0.9104143019296255,
       "grad_norm": 0.0,
-      "learning_rate": 7.319500655854927e-07,
-      "loss": 0.8622,
+      "learning_rate": 4.1799871004358005e-07,
+      "loss": 0.757,
       "step": 32083
     },
     {
-      "epoch": 0.8811622861223257,
+      "epoch": 0.9104426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 7.316160507884041e-07,
-      "loss": 0.8544,
+      "learning_rate": 4.177358033982326e-07,
+      "loss": 0.8596,
       "step": 32084
     },
     {
-      "epoch": 0.8811897503501689,
+      "epoch": 0.9104710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 7.3128210932636e-07,
-      "loss": 0.8893,
+      "learning_rate": 4.1747297769470285e-07,
+      "loss": 0.8733,
       "step": 32085
     },
     {
-      "epoch": 0.8812172145780122,
+      "epoch": 0.91049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 7.309482412020019e-07,
-      "loss": 0.7745,
+      "learning_rate": 4.172102329352079e-07,
+      "loss": 0.7896,
       "step": 32086
     },
     {
-      "epoch": 0.8812446788058553,
+      "epoch": 0.910527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 7.30614446417972e-07,
-      "loss": 0.7353,
+      "learning_rate": 4.169475691219693e-07,
+      "loss": 0.8368,
       "step": 32087
     },
     {
-      "epoch": 0.8812721430336986,
+      "epoch": 0.9105561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 7.302807249769117e-07,
-      "loss": 0.777,
+      "learning_rate": 4.1668498625720423e-07,
+      "loss": 0.8008,
       "step": 32088
     },
     {
-      "epoch": 0.8812996072615419,
+      "epoch": 0.9105845629965947,
       "grad_norm": 0.0,
-      "learning_rate": 7.299470768814609e-07,
-      "loss": 0.879,
+      "learning_rate": 4.16422484343133e-07,
+      "loss": 0.8463,
       "step": 32089
     },
     {
-      "epoch": 0.881327071489385,
+      "epoch": 0.9106129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 7.296135021342588e-07,
-      "loss": 0.8606,
+      "learning_rate": 4.161600633819718e-07,
+      "loss": 0.7503,
       "step": 32090
     },
     {
-      "epoch": 0.8813545357172283,
+      "epoch": 0.9106413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 7.292800007379475e-07,
-      "loss": 0.8425,
+      "learning_rate": 4.158977233759387e-07,
+      "loss": 0.6851,
       "step": 32091
     },
     {
-      "epoch": 0.8813819999450715,
+      "epoch": 0.9106696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 7.289465726951628e-07,
-      "loss": 0.7934,
+      "learning_rate": 4.1563546432724646e-07,
+      "loss": 0.8504,
       "step": 32092
     },
     {
-      "epoch": 0.8814094641729148,
+      "epoch": 0.9106980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 7.286132180085437e-07,
-      "loss": 0.817,
+      "learning_rate": 4.1537328623811325e-07,
+      "loss": 0.8412,
       "step": 32093
     },
     {
-      "epoch": 0.881436928400758,
+      "epoch": 0.9107264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 7.282799366807302e-07,
-      "loss": 0.7779,
+      "learning_rate": 4.15111189110754e-07,
+      "loss": 0.8667,
       "step": 32094
     },
     {
-      "epoch": 0.8814643926286012,
+      "epoch": 0.9107548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 7.279467287143549e-07,
-      "loss": 0.8235,
+      "learning_rate": 4.1484917294737914e-07,
+      "loss": 0.8354,
       "step": 32095
     },
     {
-      "epoch": 0.8814918568564445,
+      "epoch": 0.910783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 7.276135941120577e-07,
-      "loss": 0.7823,
+      "learning_rate": 4.145872377502047e-07,
+      "loss": 0.8192,
       "step": 32096
     },
     {
-      "epoch": 0.8815193210842878,
+      "epoch": 0.910811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 7.272805328764732e-07,
-      "loss": 0.8015,
+      "learning_rate": 4.1432538352144445e-07,
+      "loss": 0.8064,
       "step": 32097
     },
     {
-      "epoch": 0.8815467853121309,
+      "epoch": 0.9108399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 7.269475450102382e-07,
-      "loss": 0.6723,
+      "learning_rate": 4.1406361026330666e-07,
+      "loss": 0.8593,
       "step": 32098
     },
     {
-      "epoch": 0.8815742495399742,
+      "epoch": 0.9108683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 7.266146305159849e-07,
-      "loss": 0.8113,
+      "learning_rate": 4.138019179780062e-07,
+      "loss": 0.7154,
       "step": 32099
     },
     {
-      "epoch": 0.8816017137678174,
+      "epoch": 0.9108967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 7.262817893963503e-07,
-      "loss": 0.7591,
+      "learning_rate": 4.1354030666775015e-07,
+      "loss": 0.8508,
       "step": 32100
     },
     {
-      "epoch": 0.8816291779956607,
+      "epoch": 0.9109250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 7.259490216539644e-07,
-      "loss": 0.7681,
+      "learning_rate": 4.1327877633475013e-07,
+      "loss": 0.8066,
       "step": 32101
     },
     {
-      "epoch": 0.8816566422235039,
+      "epoch": 0.9109534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 7.256163272914629e-07,
-      "loss": 0.8567,
+      "learning_rate": 4.1301732698121653e-07,
+      "loss": 0.7845,
       "step": 32102
     },
     {
-      "epoch": 0.8816841064513471,
+      "epoch": 0.9109818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 7.252837063114782e-07,
-      "loss": 0.7723,
+      "learning_rate": 4.1275595860935434e-07,
+      "loss": 0.8241,
       "step": 32103
     },
     {
-      "epoch": 0.8817115706791904,
+      "epoch": 0.9110102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 7.249511587166403e-07,
-      "loss": 0.8211,
+      "learning_rate": 4.124946712213751e-07,
+      "loss": 0.8072,
       "step": 32104
     },
     {
-      "epoch": 0.8817390349070335,
+      "epoch": 0.911038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 7.246186845095826e-07,
-      "loss": 0.7593,
+      "learning_rate": 4.122334648194848e-07,
+      "loss": 0.8237,
       "step": 32105
     },
     {
-      "epoch": 0.8817664991348768,
+      "epoch": 0.9110669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 7.242862836929365e-07,
-      "loss": 0.7595,
+      "learning_rate": 4.119723394058883e-07,
+      "loss": 0.7299,
       "step": 32106
     },
     {
-      "epoch": 0.8817939633627201,
+      "epoch": 0.9110953461975029,
       "grad_norm": 0.0,
-      "learning_rate": 7.239539562693298e-07,
-      "loss": 0.829,
+      "learning_rate": 4.1171129498279285e-07,
+      "loss": 0.7305,
       "step": 32107
     },
     {
-      "epoch": 0.8818214275905633,
+      "epoch": 0.9111237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 7.236217022413905e-07,
-      "loss": 0.8234,
+      "learning_rate": 4.114503315524043e-07,
+      "loss": 0.7291,
       "step": 32108
     },
     {
-      "epoch": 0.8818488918184065,
+      "epoch": 0.9111520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 7.232895216117508e-07,
-      "loss": 0.7797,
+      "learning_rate": 4.1118944911692435e-07,
+      "loss": 0.6702,
       "step": 32109
     },
     {
-      "epoch": 0.8818763560462498,
+      "epoch": 0.9111804767309876,
       "grad_norm": 0.0,
-      "learning_rate": 7.229574143830365e-07,
-      "loss": 0.8721,
+      "learning_rate": 4.10928647678559e-07,
+      "loss": 0.7991,
       "step": 32110
     },
     {
-      "epoch": 0.881903820274093,
+      "epoch": 0.9112088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 7.226253805578775e-07,
-      "loss": 0.8846,
+      "learning_rate": 4.1066792723951086e-07,
+      "loss": 0.8531,
       "step": 32111
     },
     {
-      "epoch": 0.8819312845019363,
+      "epoch": 0.9112372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 7.22293420138902e-07,
-      "loss": 0.8657,
+      "learning_rate": 4.1040728780198047e-07,
+      "loss": 0.6849,
       "step": 32112
     },
     {
-      "epoch": 0.8819587487297794,
+      "epoch": 0.9112656072644721,
       "grad_norm": 0.0,
-      "learning_rate": 7.219615331287333e-07,
-      "loss": 0.86,
+      "learning_rate": 4.1014672936817155e-07,
+      "loss": 0.7463,
       "step": 32113
     },
     {
-      "epoch": 0.8819862129576227,
+      "epoch": 0.9112939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 7.216297195299982e-07,
-      "loss": 0.8058,
+      "learning_rate": 4.0988625194028464e-07,
+      "loss": 0.8348,
       "step": 32114
     },
     {
-      "epoch": 0.882013677185466,
+      "epoch": 0.911322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 7.212979793453257e-07,
-      "loss": 0.7358,
+      "learning_rate": 4.0962585552051905e-07,
+      "loss": 0.8129,
       "step": 32115
     },
     {
-      "epoch": 0.8820411414133091,
+      "epoch": 0.9113507377979568,
       "grad_norm": 0.0,
-      "learning_rate": 7.209663125773347e-07,
-      "loss": 0.8354,
+      "learning_rate": 4.093655401110741e-07,
+      "loss": 0.8363,
       "step": 32116
     },
     {
-      "epoch": 0.8820686056411524,
+      "epoch": 0.9113791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 7.206347192286533e-07,
-      "loss": 0.8024,
+      "learning_rate": 4.0910530571415143e-07,
+      "loss": 0.843,
       "step": 32117
     },
     {
-      "epoch": 0.8820960698689956,
+      "epoch": 0.9114074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 7.20303199301906e-07,
-      "loss": 0.8943,
+      "learning_rate": 4.0884515233194586e-07,
+      "loss": 0.679,
       "step": 32118
     },
     {
-      "epoch": 0.8821235340968389,
+      "epoch": 0.9114358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 7.19971752799713e-07,
-      "loss": 0.8769,
+      "learning_rate": 4.0858507996665795e-07,
+      "loss": 0.7208,
       "step": 32119
     },
     {
-      "epoch": 0.8821509983246821,
+      "epoch": 0.9114642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 7.196403797246976e-07,
-      "loss": 0.752,
+      "learning_rate": 4.0832508862048147e-07,
+      "loss": 0.8206,
       "step": 32120
     },
     {
-      "epoch": 0.8821784625525253,
+      "epoch": 0.9114926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 7.193090800794833e-07,
-      "loss": 0.7746,
+      "learning_rate": 4.0806517829561466e-07,
+      "loss": 0.8627,
       "step": 32121
     },
     {
-      "epoch": 0.8822059267803686,
+      "epoch": 0.9115209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 7.189778538666914e-07,
-      "loss": 0.7701,
+      "learning_rate": 4.078053489942535e-07,
+      "loss": 0.717,
       "step": 32122
     },
     {
-      "epoch": 0.8822333910082119,
+      "epoch": 0.9115493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 7.186467010889409e-07,
-      "loss": 0.8023,
+      "learning_rate": 4.075456007185907e-07,
+      "loss": 0.7943,
       "step": 32123
     },
     {
-      "epoch": 0.882260855236055,
+      "epoch": 0.911577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 7.183156217488541e-07,
-      "loss": 0.7619,
+      "learning_rate": 4.0728593347082126e-07,
+      "loss": 0.7621,
       "step": 32124
     },
     {
-      "epoch": 0.8822883194638983,
+      "epoch": 0.9116061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 7.179846158490478e-07,
-      "loss": 0.7892,
+      "learning_rate": 4.070263472531399e-07,
+      "loss": 0.7871,
       "step": 32125
     },
     {
-      "epoch": 0.8823157836917415,
+      "epoch": 0.9116345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 7.176536833921433e-07,
-      "loss": 0.8211,
+      "learning_rate": 4.067668420677373e-07,
+      "loss": 0.8087,
       "step": 32126
     },
     {
-      "epoch": 0.8823432479195847,
+      "epoch": 0.9116628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 7.173228243807584e-07,
-      "loss": 0.8624,
+      "learning_rate": 4.06507417916806e-07,
+      "loss": 0.7179,
       "step": 32127
     },
     {
-      "epoch": 0.882370712147428,
+      "epoch": 0.9116912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 7.1699203881751e-07,
-      "loss": 0.7926,
+      "learning_rate": 4.062480748025388e-07,
+      "loss": 0.8473,
       "step": 32128
     },
     {
-      "epoch": 0.8823981763752712,
+      "epoch": 0.9117196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 7.166613267050182e-07,
-      "loss": 0.8766,
+      "learning_rate": 4.05988812727125e-07,
+      "loss": 0.8888,
       "step": 32129
     },
     {
-      "epoch": 0.8824256406031145,
+      "epoch": 0.9117480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 7.163306880458987e-07,
-      "loss": 0.8187,
+      "learning_rate": 4.05729631692755e-07,
+      "loss": 0.8212,
       "step": 32130
     },
     {
-      "epoch": 0.8824531048309576,
+      "epoch": 0.9117763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 7.160001228427659e-07,
-      "loss": 0.8218,
+      "learning_rate": 4.054705317016183e-07,
+      "loss": 0.8641,
       "step": 32131
     },
     {
-      "epoch": 0.8824805690588009,
+      "epoch": 0.9118047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 7.156696310982359e-07,
-      "loss": 0.7197,
+      "learning_rate": 4.052115127559031e-07,
+      "loss": 0.8438,
       "step": 32132
     },
     {
-      "epoch": 0.8825080332866442,
+      "epoch": 0.91183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 7.15339212814925e-07,
-      "loss": 0.7313,
+      "learning_rate": 4.0495257485779646e-07,
+      "loss": 0.6183,
       "step": 32133
     },
     {
-      "epoch": 0.8825354975144873,
+      "epoch": 0.911861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 7.150088679954458e-07,
-      "loss": 0.7182,
+      "learning_rate": 4.046937180094879e-07,
+      "loss": 0.8297,
       "step": 32134
     },
     {
-      "epoch": 0.8825629617423306,
+      "epoch": 0.9118898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 7.146785966424141e-07,
-      "loss": 0.801,
+      "learning_rate": 4.0443494221316215e-07,
+      "loss": 0.7378,
       "step": 32135
     },
     {
-      "epoch": 0.8825904259701739,
+      "epoch": 0.9119182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 7.143483987584421e-07,
-      "loss": 0.8377,
+      "learning_rate": 4.041762474710076e-07,
+      "loss": 0.6882,
       "step": 32136
     },
     {
-      "epoch": 0.8826178901980171,
+      "epoch": 0.9119466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 7.14018274346141e-07,
-      "loss": 0.8388,
+      "learning_rate": 4.0391763378520576e-07,
+      "loss": 0.8301,
       "step": 32137
     },
     {
-      "epoch": 0.8826453544258603,
+      "epoch": 0.9119750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 7.136882234081255e-07,
-      "loss": 0.6957,
+      "learning_rate": 4.036591011579438e-07,
+      "loss": 0.8329,
       "step": 32138
     },
     {
-      "epoch": 0.8826728186537035,
+      "epoch": 0.9120034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 7.133582459470067e-07,
-      "loss": 0.7427,
+      "learning_rate": 4.0340064959140555e-07,
+      "loss": 0.8282,
       "step": 32139
     },
     {
-      "epoch": 0.8827002828815468,
+      "epoch": 0.9120317820658342,
       "grad_norm": 0.0,
-      "learning_rate": 7.130283419653938e-07,
-      "loss": 0.8308,
+      "learning_rate": 4.031422790877726e-07,
+      "loss": 0.6851,
       "step": 32140
     },
     {
-      "epoch": 0.8827277471093901,
+      "epoch": 0.9120601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 7.126985114658991e-07,
-      "loss": 0.6871,
+      "learning_rate": 4.028839896492276e-07,
+      "loss": 0.886,
       "step": 32141
     },
     {
-      "epoch": 0.8827552113372332,
+      "epoch": 0.9120885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 7.123687544511304e-07,
-      "loss": 0.9133,
+      "learning_rate": 4.026257812779544e-07,
+      "loss": 0.7219,
       "step": 32142
     },
     {
-      "epoch": 0.8827826755650765,
+      "epoch": 0.9121169125993189,
       "grad_norm": 0.0,
-      "learning_rate": 7.120390709236968e-07,
-      "loss": 0.813,
+      "learning_rate": 4.023676539761312e-07,
+      "loss": 0.8671,
       "step": 32143
     },
     {
-      "epoch": 0.8828101397929197,
+      "epoch": 0.9121452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 7.117094608862086e-07,
-      "loss": 0.776,
+      "learning_rate": 4.021096077459408e-07,
+      "loss": 0.8356,
       "step": 32144
     },
     {
-      "epoch": 0.882837604020763,
+      "epoch": 0.9121736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 7.113799243412734e-07,
-      "loss": 0.7845,
+      "learning_rate": 4.0185164258956245e-07,
+      "loss": 0.8148,
       "step": 32145
     },
     {
-      "epoch": 0.8828650682486062,
+      "epoch": 0.9122020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 7.110504612914992e-07,
-      "loss": 0.7738,
+      "learning_rate": 4.015937585091734e-07,
+      "loss": 0.7918,
       "step": 32146
     },
     {
-      "epoch": 0.8828925324764494,
+      "epoch": 0.9122304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 7.107210717394908e-07,
-      "loss": 0.7086,
+      "learning_rate": 4.0133595550695405e-07,
+      "loss": 0.7865,
       "step": 32147
     },
     {
-      "epoch": 0.8829199967042927,
+      "epoch": 0.9122587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 7.10391755687857e-07,
-      "loss": 0.8586,
+      "learning_rate": 4.010782335850816e-07,
+      "loss": 0.8547,
       "step": 32148
     },
     {
-      "epoch": 0.8829474609321359,
+      "epoch": 0.9122871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 7.100625131392003e-07,
-      "loss": 0.8278,
+      "learning_rate": 4.008205927457309e-07,
+      "loss": 0.8197,
       "step": 32149
     },
     {
-      "epoch": 0.8829749251599791,
+      "epoch": 0.9123155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 7.097333440961273e-07,
-      "loss": 0.8268,
+      "learning_rate": 4.0056303299108257e-07,
+      "loss": 0.7999,
       "step": 32150
     },
     {
-      "epoch": 0.8830023893878224,
+      "epoch": 0.9123439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 7.09404248561244e-07,
-      "loss": 0.8242,
+      "learning_rate": 4.003055543233092e-07,
+      "loss": 0.8448,
       "step": 32151
     },
     {
-      "epoch": 0.8830298536156655,
+      "epoch": 0.912372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 7.090752265371525e-07,
-      "loss": 0.8538,
+      "learning_rate": 4.0004815674458576e-07,
+      "loss": 0.8673,
       "step": 32152
     },
     {
-      "epoch": 0.8830573178435088,
+      "epoch": 0.9124006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 7.087462780264575e-07,
-      "loss": 0.7503,
+      "learning_rate": 3.9979084025708825e-07,
+      "loss": 0.8108,
       "step": 32153
     },
     {
-      "epoch": 0.8830847820713521,
+      "epoch": 0.9124290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 7.084174030317603e-07,
-      "loss": 0.7621,
+      "learning_rate": 3.995336048629883e-07,
+      "loss": 0.7994,
       "step": 32154
     },
     {
-      "epoch": 0.8831122462991953,
+      "epoch": 0.9124574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 7.080886015556654e-07,
-      "loss": 0.7162,
+      "learning_rate": 3.992764505644586e-07,
+      "loss": 0.8239,
       "step": 32155
     },
     {
-      "epoch": 0.8831397105270385,
+      "epoch": 0.9124858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 7.077598736007707e-07,
-      "loss": 0.8361,
+      "learning_rate": 3.9901937736367524e-07,
+      "loss": 0.802,
       "step": 32156
     },
     {
-      "epoch": 0.8831671747548817,
+      "epoch": 0.9125141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 7.074312191696809e-07,
-      "loss": 0.812,
+      "learning_rate": 3.987623852628042e-07,
+      "loss": 0.8778,
       "step": 32157
     },
     {
-      "epoch": 0.883194638982725,
+      "epoch": 0.9125425652667424,
       "grad_norm": 0.0,
-      "learning_rate": 7.071026382649937e-07,
-      "loss": 0.8499,
+      "learning_rate": 3.985054742640193e-07,
+      "loss": 0.8339,
       "step": 32158
     },
     {
-      "epoch": 0.8832221032105683,
+      "epoch": 0.9125709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 7.067741308893117e-07,
-      "loss": 0.7583,
+      "learning_rate": 3.982486443694911e-07,
+      "loss": 0.8814,
       "step": 32159
     },
     {
-      "epoch": 0.8832495674384114,
+      "epoch": 0.9125993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 7.064456970452305e-07,
-      "loss": 0.8466,
+      "learning_rate": 3.979918955813877e-07,
+      "loss": 0.8088,
       "step": 32160
     },
     {
-      "epoch": 0.8832770316662547,
+      "epoch": 0.9126276958002271,
       "grad_norm": 0.0,
-      "learning_rate": 7.061173367353513e-07,
-      "loss": 0.7594,
+      "learning_rate": 3.977352279018776e-07,
+      "loss": 0.7892,
       "step": 32161
     },
     {
-      "epoch": 0.883304495894098,
+      "epoch": 0.9126560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 7.057890499622722e-07,
-      "loss": 0.8637,
+      "learning_rate": 3.974786413331311e-07,
+      "loss": 0.7084,
       "step": 32162
     },
     {
-      "epoch": 0.8833319601219412,
+      "epoch": 0.9126844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 7.05460836728592e-07,
-      "loss": 0.8302,
+      "learning_rate": 3.972221358773132e-07,
+      "loss": 0.8683,
       "step": 32163
     },
     {
-      "epoch": 0.8833594243497844,
+      "epoch": 0.9127128263337116,
       "grad_norm": 0.0,
-      "learning_rate": 7.051326970369043e-07,
-      "loss": 0.8089,
+      "learning_rate": 3.9696571153659214e-07,
+      "loss": 0.8091,
       "step": 32164
     },
     {
-      "epoch": 0.8833868885776276,
+      "epoch": 0.9127412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 7.048046308898094e-07,
-      "loss": 0.8474,
+      "learning_rate": 3.967093683131329e-07,
+      "loss": 0.8576,
       "step": 32165
     },
     {
-      "epoch": 0.8834143528054709,
+      "epoch": 0.9127695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 7.044766382898982e-07,
-      "loss": 0.8698,
+      "learning_rate": 3.964531062091004e-07,
+      "loss": 0.8189,
       "step": 32166
     },
     {
-      "epoch": 0.8834418170333141,
+      "epoch": 0.9127979568671963,
       "grad_norm": 0.0,
-      "learning_rate": 7.04148719239769e-07,
-      "loss": 0.803,
+      "learning_rate": 3.9619692522666287e-07,
+      "loss": 0.81,
       "step": 32167
     },
     {
-      "epoch": 0.8834692812611573,
+      "epoch": 0.9128263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 7.038208737420161e-07,
-      "loss": 0.8696,
+      "learning_rate": 3.959408253679797e-07,
+      "loss": 0.7578,
       "step": 32168
     },
     {
-      "epoch": 0.8834967454890006,
+      "epoch": 0.9128547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 7.034931017992341e-07,
-      "loss": 0.7755,
+      "learning_rate": 3.956848066352159e-07,
+      "loss": 0.7813,
       "step": 32169
     },
     {
-      "epoch": 0.8835242097168438,
+      "epoch": 0.912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 7.031654034140168e-07,
-      "loss": 0.8117,
+      "learning_rate": 3.9542886903053635e-07,
+      "loss": 0.7115,
       "step": 32170
     },
     {
-      "epoch": 0.883551673944687,
+      "epoch": 0.9129114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 7.02837778588954e-07,
-      "loss": 0.8244,
+      "learning_rate": 3.951730125560993e-07,
+      "loss": 0.8086,
       "step": 32171
     },
     {
-      "epoch": 0.8835791381725303,
+      "epoch": 0.9129398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 7.025102273266415e-07,
-      "loss": 0.7594,
+      "learning_rate": 3.949172372140675e-07,
+      "loss": 0.8511,
       "step": 32172
     },
     {
-      "epoch": 0.8836066024003735,
+      "epoch": 0.9129682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 7.021827496296684e-07,
-      "loss": 0.7124,
+      "learning_rate": 3.946615430066025e-07,
+      "loss": 0.8627,
       "step": 32173
     },
     {
-      "epoch": 0.8836340666282168,
+      "epoch": 0.9129965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 7.018553455006272e-07,
-      "loss": 0.6667,
+      "learning_rate": 3.9440592993586267e-07,
+      "loss": 0.727,
       "step": 32174
     },
     {
-      "epoch": 0.88366153085606,
+      "epoch": 0.9130249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 7.015280149421089e-07,
-      "loss": 0.8682,
+      "learning_rate": 3.9415039800400735e-07,
+      "loss": 0.8477,
       "step": 32175
     },
     {
-      "epoch": 0.8836889950839032,
+      "epoch": 0.9130533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 7.012007579567015e-07,
-      "loss": 0.8644,
+      "learning_rate": 3.9389494721319585e-07,
+      "loss": 0.8595,
       "step": 32176
     },
     {
-      "epoch": 0.8837164593117465,
+      "epoch": 0.9130817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 7.008735745469985e-07,
-      "loss": 0.8418,
+      "learning_rate": 3.9363957756558547e-07,
+      "loss": 0.8089,
       "step": 32177
     },
     {
-      "epoch": 0.8837439235395896,
+      "epoch": 0.9131101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 7.005464647155835e-07,
-      "loss": 0.8932,
+      "learning_rate": 3.933842890633333e-07,
+      "loss": 0.7169,
       "step": 32178
     },
     {
-      "epoch": 0.8837713877674329,
+      "epoch": 0.9131384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 7.002194284650488e-07,
-      "loss": 0.6742,
+      "learning_rate": 3.9312908170859645e-07,
+      "loss": 0.8236,
       "step": 32179
     },
     {
-      "epoch": 0.8837988519952762,
+      "epoch": 0.91316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.998924657979789e-07,
-      "loss": 0.8229,
+      "learning_rate": 3.928739555035288e-07,
+      "loss": 0.7482,
       "step": 32180
     },
     {
-      "epoch": 0.8838263162231194,
+      "epoch": 0.9131952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 6.995655767169629e-07,
-      "loss": 0.7891,
+      "learning_rate": 3.926189104502864e-07,
+      "loss": 0.7686,
       "step": 32181
     },
     {
-      "epoch": 0.8838537804509626,
+      "epoch": 0.9132236095346198,
       "grad_norm": 0.0,
-      "learning_rate": 6.992387612245866e-07,
-      "loss": 0.8706,
+      "learning_rate": 3.923639465510265e-07,
+      "loss": 0.9016,
       "step": 32182
     },
     {
-      "epoch": 0.8838812446788059,
+      "epoch": 0.9132519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 6.989120193234366e-07,
-      "loss": 0.6863,
+      "learning_rate": 3.9210906380789836e-07,
+      "loss": 0.6859,
       "step": 32183
     },
     {
-      "epoch": 0.8839087089066491,
+      "epoch": 0.9132803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 6.985853510160967e-07,
-      "loss": 0.7458,
+      "learning_rate": 3.918542622230581e-07,
+      "loss": 0.8588,
       "step": 32184
     },
     {
-      "epoch": 0.8839361731344924,
+      "epoch": 0.9133087400681045,
       "grad_norm": 0.0,
-      "learning_rate": 6.982587563051523e-07,
-      "loss": 0.7448,
+      "learning_rate": 3.915995417986562e-07,
+      "loss": 0.7755,
       "step": 32185
     },
     {
-      "epoch": 0.8839636373623355,
+      "epoch": 0.9133371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 6.979322351931883e-07,
-      "loss": 0.7625,
+      "learning_rate": 3.913449025368443e-07,
+      "loss": 0.7692,
       "step": 32186
     },
     {
-      "epoch": 0.8839911015901788,
+      "epoch": 0.9133654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 6.97605787682789e-07,
-      "loss": 0.8627,
+      "learning_rate": 3.9109034443977513e-07,
+      "loss": 0.6957,
       "step": 32187
     },
     {
-      "epoch": 0.8840185658180221,
+      "epoch": 0.913393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 6.972794137765337e-07,
-      "loss": 0.7806,
+      "learning_rate": 3.908358675095969e-07,
+      "loss": 0.8429,
       "step": 32188
     },
     {
-      "epoch": 0.8840460300458652,
+      "epoch": 0.913422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 6.969531134770102e-07,
-      "loss": 0.7622,
+      "learning_rate": 3.9058147174846016e-07,
+      "loss": 0.7297,
       "step": 32189
     },
     {
-      "epoch": 0.8840734942737085,
+      "epoch": 0.9134506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 6.966268867867942e-07,
-      "loss": 0.7049,
+      "learning_rate": 3.903271571585143e-07,
+      "loss": 0.845,
       "step": 32190
     },
     {
-      "epoch": 0.8841009585015517,
+      "epoch": 0.9134790011350737,
       "grad_norm": 0.0,
-      "learning_rate": 6.963007337084715e-07,
-      "loss": 0.8877,
+      "learning_rate": 3.9007292374190654e-07,
+      "loss": 0.7853,
       "step": 32191
     },
     {
-      "epoch": 0.884128422729395,
+      "epoch": 0.9135073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 6.95974654244621e-07,
-      "loss": 0.9514,
+      "learning_rate": 3.8981877150078395e-07,
+      "loss": 0.8622,
       "step": 32192
     },
     {
-      "epoch": 0.8841558869572382,
+      "epoch": 0.9135357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 6.95648648397822e-07,
-      "loss": 0.8658,
+      "learning_rate": 3.89564700437296e-07,
+      "loss": 0.8195,
       "step": 32193
     },
     {
-      "epoch": 0.8841833511850814,
+      "epoch": 0.9135641316685584,
       "grad_norm": 0.0,
-      "learning_rate": 6.953227161706577e-07,
-      "loss": 0.8214,
+      "learning_rate": 3.893107105535854e-07,
+      "loss": 0.9178,
       "step": 32194
     },
     {
-      "epoch": 0.8842108154129247,
+      "epoch": 0.9135925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 6.949968575657029e-07,
-      "loss": 0.8559,
+      "learning_rate": 3.8905680185179815e-07,
+      "loss": 0.9042,
       "step": 32195
     },
     {
-      "epoch": 0.884238279640768,
+      "epoch": 0.9136208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 6.946710725855388e-07,
-      "loss": 0.711,
+      "learning_rate": 3.8880297433408263e-07,
+      "loss": 0.7309,
       "step": 32196
     },
     {
-      "epoch": 0.8842657438686111,
+      "epoch": 0.9136492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 6.9434536123274e-07,
-      "loss": 0.7475,
+      "learning_rate": 3.8854922800257935e-07,
+      "loss": 0.8364,
       "step": 32197
     },
     {
-      "epoch": 0.8842932080964544,
+      "epoch": 0.913677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 6.940197235098844e-07,
-      "loss": 0.8224,
+      "learning_rate": 3.8829556285943314e-07,
+      "loss": 0.9028,
       "step": 32198
     },
     {
-      "epoch": 0.8843206723242976,
+      "epoch": 0.913706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 6.936941594195512e-07,
-      "loss": 0.8214,
+      "learning_rate": 3.8804197890678683e-07,
+      "loss": 0.8649,
       "step": 32199
     },
     {
-      "epoch": 0.8843481365521408,
+      "epoch": 0.9137343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 6.933686689643149e-07,
-      "loss": 0.8379,
+      "learning_rate": 3.8778847614678095e-07,
+      "loss": 0.768,
       "step": 32200
     },
     {
-      "epoch": 0.8843756007799841,
+      "epoch": 0.9137627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 6.930432521467501e-07,
-      "loss": 0.8223,
+      "learning_rate": 3.8753505458155815e-07,
+      "loss": 0.7476,
       "step": 32201
     },
     {
-      "epoch": 0.8844030650078273,
+      "epoch": 0.9137911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 6.927179089694313e-07,
-      "loss": 0.9007,
+      "learning_rate": 3.872817142132601e-07,
+      "loss": 0.766,
       "step": 32202
     },
     {
-      "epoch": 0.8844305292356706,
+      "epoch": 0.9138195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 6.923926394349367e-07,
-      "loss": 0.8647,
+      "learning_rate": 3.8702845504402395e-07,
+      "loss": 0.8715,
       "step": 32203
     },
     {
-      "epoch": 0.8844579934635137,
+      "epoch": 0.9138479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 6.920674435458341e-07,
-      "loss": 0.8349,
+      "learning_rate": 3.867752770759914e-07,
+      "loss": 0.8315,
       "step": 32204
     },
     {
-      "epoch": 0.884485457691357,
+      "epoch": 0.9138762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 6.917423213047003e-07,
-      "loss": 0.9144,
+      "learning_rate": 3.865221803113006e-07,
+      "loss": 0.8035,
       "step": 32205
     },
     {
-      "epoch": 0.8845129219192003,
+      "epoch": 0.9139046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 6.914172727141078e-07,
-      "loss": 0.7157,
+      "learning_rate": 3.862691647520889e-07,
+      "loss": 0.7948,
       "step": 32206
     },
     {
-      "epoch": 0.8845403861470434,
+      "epoch": 0.913933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 6.910922977766266e-07,
-      "loss": 0.8595,
+      "learning_rate": 3.860162304004933e-07,
+      "loss": 0.9176,
       "step": 32207
     },
     {
-      "epoch": 0.8845678503748867,
+      "epoch": 0.9139614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 6.907673964948291e-07,
-      "loss": 0.8668,
+      "learning_rate": 3.857633772586511e-07,
+      "loss": 0.7779,
       "step": 32208
     },
     {
-      "epoch": 0.88459531460273,
+      "epoch": 0.9139897843359819,
       "grad_norm": 0.0,
-      "learning_rate": 6.904425688712856e-07,
-      "loss": 0.7934,
+      "learning_rate": 3.855106053286972e-07,
+      "loss": 0.7957,
       "step": 32209
     },
     {
-      "epoch": 0.8846227788305732,
+      "epoch": 0.9140181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 6.901178149085663e-07,
-      "loss": 0.7267,
+      "learning_rate": 3.852579146127677e-07,
+      "loss": 0.8513,
       "step": 32210
     },
     {
-      "epoch": 0.8846502430584164,
+      "epoch": 0.9140465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 6.897931346092435e-07,
-      "loss": 0.8113,
+      "learning_rate": 3.850053051129965e-07,
+      "loss": 0.8538,
       "step": 32211
     },
     {
-      "epoch": 0.8846777072862596,
+      "epoch": 0.9140749148694666,
       "grad_norm": 0.0,
-      "learning_rate": 6.894685279758839e-07,
-      "loss": 0.761,
+      "learning_rate": 3.8475277683151733e-07,
+      "loss": 0.7009,
       "step": 32212
     },
     {
-      "epoch": 0.8847051715141029,
+      "epoch": 0.9141032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 6.891439950110535e-07,
-      "loss": 0.7646,
+      "learning_rate": 3.8450032977046526e-07,
+      "loss": 0.8237,
       "step": 32213
     },
     {
-      "epoch": 0.8847326357419462,
+      "epoch": 0.9141316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.888195357173233e-07,
-      "loss": 0.8107,
+      "learning_rate": 3.8424796393197075e-07,
+      "loss": 0.8809,
       "step": 32214
     },
     {
-      "epoch": 0.8847600999697893,
+      "epoch": 0.9141600454029511,
       "grad_norm": 0.0,
-      "learning_rate": 6.884951500972592e-07,
-      "loss": 0.7534,
+      "learning_rate": 3.8399567931816537e-07,
+      "loss": 0.8658,
       "step": 32215
     },
     {
-      "epoch": 0.8847875641976326,
+      "epoch": 0.9141884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 6.881708381534291e-07,
-      "loss": 0.7692,
+      "learning_rate": 3.837434759311809e-07,
+      "loss": 0.7244,
       "step": 32216
     },
     {
-      "epoch": 0.8848150284254758,
+      "epoch": 0.914216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 6.878465998883976e-07,
-      "loss": 0.7881,
+      "learning_rate": 3.8349135377314773e-07,
+      "loss": 0.8219,
       "step": 32217
     },
     {
-      "epoch": 0.884842492653319,
+      "epoch": 0.9142451759364358,
       "grad_norm": 0.0,
-      "learning_rate": 6.875224353047328e-07,
-      "loss": 0.9211,
+      "learning_rate": 3.832393128461953e-07,
+      "loss": 0.7232,
       "step": 32218
     },
     {
-      "epoch": 0.8848699568811623,
+      "epoch": 0.9142735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 6.871983444049979e-07,
-      "loss": 0.8751,
+      "learning_rate": 3.8298735315245306e-07,
+      "loss": 0.8276,
       "step": 32219
     },
     {
-      "epoch": 0.8848974211090055,
+      "epoch": 0.9143019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 6.868743271917555e-07,
-      "loss": 0.739,
+      "learning_rate": 3.8273547469404815e-07,
+      "loss": 0.8415,
       "step": 32220
     },
     {
-      "epoch": 0.8849248853368488,
+      "epoch": 0.9143303064699205,
       "grad_norm": 0.0,
-      "learning_rate": 6.865503836675714e-07,
-      "loss": 0.8064,
+      "learning_rate": 3.8248367747311e-07,
+      "loss": 0.8098,
       "step": 32221
     },
     {
-      "epoch": 0.884952349564692,
+      "epoch": 0.9143586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 6.862265138350078e-07,
-      "loss": 0.8217,
+      "learning_rate": 3.822319614917647e-07,
+      "loss": 0.8759,
       "step": 32222
     },
     {
-      "epoch": 0.8849798137925352,
+      "epoch": 0.9143870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 6.859027176966281e-07,
-      "loss": 0.8186,
+      "learning_rate": 3.8198032675213715e-07,
+      "loss": 0.8431,
       "step": 32223
     },
     {
-      "epoch": 0.8850072780203785,
+      "epoch": 0.9144154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 6.855789952549951e-07,
-      "loss": 0.8401,
+      "learning_rate": 3.817287732563546e-07,
+      "loss": 0.827,
       "step": 32224
     },
     {
-      "epoch": 0.8850347422482217,
+      "epoch": 0.9144438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 6.852553465126687e-07,
-      "loss": 0.8944,
+      "learning_rate": 3.81477301006542e-07,
+      "loss": 0.8602,
       "step": 32225
     },
     {
-      "epoch": 0.8850622064760649,
+      "epoch": 0.914472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 6.849317714722093e-07,
-      "loss": 0.8126,
+      "learning_rate": 3.81225910004821e-07,
+      "loss": 0.7193,
       "step": 32226
     },
     {
-      "epoch": 0.8850896707039082,
+      "epoch": 0.9145005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 6.8460827013618e-07,
-      "loss": 0.8393,
+      "learning_rate": 3.809746002533199e-07,
+      "loss": 0.7905,
       "step": 32227
     },
     {
-      "epoch": 0.8851171349317514,
+      "epoch": 0.9145289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 6.842848425071358e-07,
-      "loss": 0.745,
+      "learning_rate": 3.80723371754157e-07,
+      "loss": 0.8401,
       "step": 32228
     },
     {
-      "epoch": 0.8851445991595946,
+      "epoch": 0.9145573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 6.8396148858764e-07,
-      "loss": 0.7423,
+      "learning_rate": 3.804722245094561e-07,
+      "loss": 0.808,
       "step": 32229
     },
     {
-      "epoch": 0.8851720633874378,
+      "epoch": 0.9145856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 6.836382083802507e-07,
-      "loss": 0.8702,
+      "learning_rate": 3.802211585213411e-07,
+      "loss": 0.7949,
       "step": 32230
     },
     {
-      "epoch": 0.8851995276152811,
+      "epoch": 0.9146140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 6.833150018875223e-07,
-      "loss": 0.8595,
+      "learning_rate": 3.7997017379192813e-07,
+      "loss": 0.8057,
       "step": 32231
     },
     {
-      "epoch": 0.8852269918431244,
+      "epoch": 0.9146424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 6.829918691120152e-07,
-      "loss": 0.8579,
+      "learning_rate": 3.7971927032333986e-07,
+      "loss": 0.7975,
       "step": 32232
     },
     {
-      "epoch": 0.8852544560709675,
+      "epoch": 0.9146708286038593,
       "grad_norm": 0.0,
-      "learning_rate": 6.826688100562862e-07,
-      "loss": 0.7938,
+      "learning_rate": 3.794684481176969e-07,
+      "loss": 0.8134,
       "step": 32233
     },
     {
-      "epoch": 0.8852819202988108,
+      "epoch": 0.9146992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 6.823458247228898e-07,
-      "loss": 0.7633,
+      "learning_rate": 3.7921770717711414e-07,
+      "loss": 0.7745,
       "step": 32234
     },
     {
-      "epoch": 0.8853093845266541,
+      "epoch": 0.914727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 6.820229131143852e-07,
-      "loss": 0.6898,
+      "learning_rate": 3.789670475037133e-07,
+      "loss": 0.8277,
       "step": 32235
     },
     {
-      "epoch": 0.8853368487544973,
+      "epoch": 0.914755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 6.817000752333236e-07,
-      "loss": 0.8609,
+      "learning_rate": 3.7871646909961033e-07,
+      "loss": 0.7618,
       "step": 32236
     },
     {
-      "epoch": 0.8853643129823405,
+      "epoch": 0.9147843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 6.813773110822597e-07,
-      "loss": 0.7851,
+      "learning_rate": 3.784659719669215e-07,
+      "loss": 0.8042,
       "step": 32237
     },
     {
-      "epoch": 0.8853917772101837,
+      "epoch": 0.9148127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 6.810546206637469e-07,
-      "loss": 0.8265,
+      "learning_rate": 3.7821555610776274e-07,
+      "loss": 0.7529,
       "step": 32238
     },
     {
-      "epoch": 0.885419241438027,
+      "epoch": 0.9148410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 6.807320039803412e-07,
-      "loss": 0.7795,
+      "learning_rate": 3.779652215242513e-07,
+      "loss": 0.8137,
       "step": 32239
     },
     {
-      "epoch": 0.8854467056658702,
+      "epoch": 0.9148694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 6.804094610345935e-07,
-      "loss": 0.7917,
+      "learning_rate": 3.7771496821849774e-07,
+      "loss": 0.7498,
       "step": 32240
     },
     {
-      "epoch": 0.8854741698937134,
+      "epoch": 0.9148978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 6.800869918290565e-07,
-      "loss": 0.8286,
+      "learning_rate": 3.7746479619261924e-07,
+      "loss": 0.873,
       "step": 32241
     },
     {
-      "epoch": 0.8855016341215567,
+      "epoch": 0.9149262202043132,
       "grad_norm": 0.0,
-      "learning_rate": 6.797645963662825e-07,
-      "loss": 0.7636,
+      "learning_rate": 3.772147054487285e-07,
+      "loss": 0.7618,
       "step": 32242
     },
     {
-      "epoch": 0.8855290983493999,
+      "epoch": 0.9149545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 6.794422746488205e-07,
-      "loss": 0.7233,
+      "learning_rate": 3.7696469598893727e-07,
+      "loss": 0.8298,
       "step": 32243
     },
     {
-      "epoch": 0.8855565625772431,
+      "epoch": 0.9149829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 6.79120026679222e-07,
-      "loss": 0.8162,
+      "learning_rate": 3.7671476781535934e-07,
+      "loss": 0.7644,
       "step": 32244
     },
     {
-      "epoch": 0.8855840268050864,
+      "epoch": 0.9150113507377979,
       "grad_norm": 0.0,
-      "learning_rate": 6.787978524600347e-07,
-      "loss": 0.7275,
+      "learning_rate": 3.76464920930103e-07,
+      "loss": 0.7561,
       "step": 32245
     },
     {
-      "epoch": 0.8856114910329296,
+      "epoch": 0.9150397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 6.784757519938101e-07,
-      "loss": 0.8314,
+      "learning_rate": 3.7621515533527996e-07,
+      "loss": 0.7657,
       "step": 32246
     },
     {
-      "epoch": 0.8856389552607729,
+      "epoch": 0.9150681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 6.781537252830961e-07,
-      "loss": 0.8106,
+      "learning_rate": 3.7596547103300187e-07,
+      "loss": 0.7782,
       "step": 32247
     },
     {
-      "epoch": 0.8856664194886161,
+      "epoch": 0.9150964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 6.778317723304429e-07,
-      "loss": 0.7876,
+      "learning_rate": 3.7571586802537475e-07,
+      "loss": 0.747,
       "step": 32248
     },
     {
-      "epoch": 0.8856938837164593,
+      "epoch": 0.9151248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 6.775098931383939e-07,
-      "loss": 0.7913,
+      "learning_rate": 3.754663463145081e-07,
+      "loss": 0.8201,
       "step": 32249
     },
     {
-      "epoch": 0.8857213479443026,
+      "epoch": 0.9151532349602725,
       "grad_norm": 0.0,
-      "learning_rate": 6.771880877094972e-07,
-      "loss": 0.8414,
+      "learning_rate": 3.7521690590251015e-07,
+      "loss": 0.7802,
       "step": 32250
     },
     {
-      "epoch": 0.8857488121721457,
+      "epoch": 0.9151816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 6.768663560463017e-07,
-      "loss": 0.737,
+      "learning_rate": 3.7496754679148815e-07,
+      "loss": 0.7926,
       "step": 32251
     },
     {
-      "epoch": 0.885776276399989,
+      "epoch": 0.9152099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 6.765446981513501e-07,
-      "loss": 0.8586,
+      "learning_rate": 3.747182689835471e-07,
+      "loss": 0.8166,
       "step": 32252
     },
     {
-      "epoch": 0.8858037406278323,
+      "epoch": 0.9152383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 6.762231140271891e-07,
-      "loss": 0.7186,
+      "learning_rate": 3.7446907248079423e-07,
+      "loss": 0.7221,
       "step": 32253
     },
     {
-      "epoch": 0.8858312048556755,
+      "epoch": 0.915266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 6.75901603676361e-07,
-      "loss": 0.741,
+      "learning_rate": 3.7421995728533335e-07,
+      "loss": 0.9021,
       "step": 32254
     },
     {
-      "epoch": 0.8858586690835187,
+      "epoch": 0.9152951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 6.755801671014117e-07,
-      "loss": 0.8362,
+      "learning_rate": 3.7397092339926943e-07,
+      "loss": 0.7296,
       "step": 32255
     },
     {
-      "epoch": 0.8858861333113619,
+      "epoch": 0.9153234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 6.752588043048836e-07,
-      "loss": 0.6991,
+      "learning_rate": 3.737219708247053e-07,
+      "loss": 0.8621,
       "step": 32256
     },
     {
-      "epoch": 0.8859135975392052,
+      "epoch": 0.9153518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 6.7493751528932e-07,
-      "loss": 0.7631,
+      "learning_rate": 3.7347309956374367e-07,
+      "loss": 0.7386,
       "step": 32257
     },
     {
-      "epoch": 0.8859410617670485,
+      "epoch": 0.9153802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 6.746163000572647e-07,
-      "loss": 0.77,
+      "learning_rate": 3.7322430961848955e-07,
+      "loss": 0.8226,
       "step": 32258
     },
     {
-      "epoch": 0.8859685259948916,
+      "epoch": 0.9154086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 6.742951586112556e-07,
-      "loss": 0.6869,
+      "learning_rate": 3.729756009910412e-07,
+      "loss": 0.8673,
       "step": 32259
     },
     {
-      "epoch": 0.8859959902227349,
+      "epoch": 0.9154370034052214,
       "grad_norm": 0.0,
-      "learning_rate": 6.739740909538373e-07,
-      "loss": 0.6565,
+      "learning_rate": 3.7272697368350154e-07,
+      "loss": 0.7854,
       "step": 32260
     },
     {
-      "epoch": 0.8860234544505782,
+      "epoch": 0.9154653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 6.736530970875476e-07,
-      "loss": 0.8602,
+      "learning_rate": 3.7247842769796983e-07,
+      "loss": 0.7584,
       "step": 32261
     },
     {
-      "epoch": 0.8860509186784213,
+      "epoch": 0.9154937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 6.733321770149271e-07,
-      "loss": 0.861,
+      "learning_rate": 3.7222996303654557e-07,
+      "loss": 0.9095,
       "step": 32262
     },
     {
-      "epoch": 0.8860783829062646,
+      "epoch": 0.915522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 6.730113307385144e-07,
-      "loss": 0.7571,
+      "learning_rate": 3.7198157970132706e-07,
+      "loss": 0.7702,
       "step": 32263
     },
     {
-      "epoch": 0.8861058471341078,
+      "epoch": 0.9155505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 6.7269055826085e-07,
-      "loss": 0.8123,
+      "learning_rate": 3.717332776944138e-07,
+      "loss": 0.7386,
       "step": 32264
     },
     {
-      "epoch": 0.8861333113619511,
+      "epoch": 0.9155788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 6.723698595844718e-07,
-      "loss": 0.819,
+      "learning_rate": 3.7148505701790073e-07,
+      "loss": 0.8032,
       "step": 32265
     },
     {
-      "epoch": 0.8861607755897943,
+      "epoch": 0.9156072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 6.720492347119156e-07,
-      "loss": 0.8204,
+      "learning_rate": 3.7123691767388726e-07,
+      "loss": 0.8412,
       "step": 32266
     },
     {
-      "epoch": 0.8861882398176375,
+      "epoch": 0.9156356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 6.717286836457193e-07,
-      "loss": 0.7849,
+      "learning_rate": 3.709888596644695e-07,
+      "loss": 0.7546,
       "step": 32267
     },
     {
-      "epoch": 0.8862157040454808,
+      "epoch": 0.9156640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 6.714082063884187e-07,
-      "loss": 0.8254,
+      "learning_rate": 3.707408829917403e-07,
+      "loss": 0.8334,
       "step": 32268
     },
     {
-      "epoch": 0.886243168273324,
+      "epoch": 0.9156923950056753,
       "grad_norm": 0.0,
-      "learning_rate": 6.710878029425504e-07,
-      "loss": 0.8203,
+      "learning_rate": 3.7049298765779564e-07,
+      "loss": 0.7114,
       "step": 32269
     },
     {
-      "epoch": 0.8862706325011672,
+      "epoch": 0.9157207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 6.707674733106484e-07,
-      "loss": 0.8446,
+      "learning_rate": 3.702451736647306e-07,
+      "loss": 0.7534,
       "step": 32270
     },
     {
-      "epoch": 0.8862980967290105,
+      "epoch": 0.9157491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 6.704472174952492e-07,
-      "loss": 0.8115,
+      "learning_rate": 3.6999744101463677e-07,
+      "loss": 0.7596,
       "step": 32271
     },
     {
-      "epoch": 0.8863255609568537,
+      "epoch": 0.91577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 6.701270354988843e-07,
-      "loss": 0.7603,
+      "learning_rate": 3.6974978970960695e-07,
+      "loss": 0.751,
       "step": 32272
     },
     {
-      "epoch": 0.8863530251846969,
+      "epoch": 0.9158059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 6.698069273240881e-07,
-      "loss": 0.8138,
+      "learning_rate": 3.6950221975173393e-07,
+      "loss": 0.7794,
       "step": 32273
     },
     {
-      "epoch": 0.8863804894125402,
+      "epoch": 0.9158342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 6.694868929733944e-07,
-      "loss": 0.8239,
+      "learning_rate": 3.692547311431083e-07,
+      "loss": 0.8248,
       "step": 32274
     },
     {
-      "epoch": 0.8864079536403834,
+      "epoch": 0.9158626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 6.691669324493355e-07,
-      "loss": 0.8608,
+      "learning_rate": 3.6900732388582274e-07,
+      "loss": 0.7361,
       "step": 32275
     },
     {
-      "epoch": 0.8864354178682267,
+      "epoch": 0.9158910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 6.688470457544405e-07,
-      "loss": 0.764,
+      "learning_rate": 3.6875999798196335e-07,
+      "loss": 0.8297,
       "step": 32276
     },
     {
-      "epoch": 0.8864628820960698,
+      "epoch": 0.9159194097616346,
       "grad_norm": 0.0,
-      "learning_rate": 6.68527232891244e-07,
-      "loss": 0.8227,
+      "learning_rate": 3.685127534336208e-07,
+      "loss": 0.7682,
       "step": 32277
     },
     {
-      "epoch": 0.8864903463239131,
+      "epoch": 0.9159477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 6.68207493862274e-07,
-      "loss": 0.7685,
+      "learning_rate": 3.6826559024288664e-07,
+      "loss": 0.7197,
       "step": 32278
     },
     {
-      "epoch": 0.8865178105517564,
+      "epoch": 0.9159761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 6.678878286700608e-07,
-      "loss": 0.8081,
+      "learning_rate": 3.6801850841184375e-07,
+      "loss": 0.6943,
       "step": 32279
     },
     {
-      "epoch": 0.8865452747795995,
+      "epoch": 0.9160045402951191,
       "grad_norm": 0.0,
-      "learning_rate": 6.675682373171333e-07,
-      "loss": 0.8726,
+      "learning_rate": 3.677715079425814e-07,
+      "loss": 0.7341,
       "step": 32280
     },
     {
-      "epoch": 0.8865727390074428,
+      "epoch": 0.9160329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 6.672487198060207e-07,
-      "loss": 0.8484,
+      "learning_rate": 3.675245888371881e-07,
+      "loss": 0.8166,
       "step": 32281
     },
     {
-      "epoch": 0.886600203235286,
+      "epoch": 0.916061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 6.669292761392543e-07,
-      "loss": 0.7347,
+      "learning_rate": 3.672777510977454e-07,
+      "loss": 0.8529,
       "step": 32282
     },
     {
-      "epoch": 0.8866276674631293,
+      "epoch": 0.9160896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 6.666099063193566e-07,
-      "loss": 0.8348,
+      "learning_rate": 3.6703099472634174e-07,
+      "loss": 0.903,
       "step": 32283
     },
     {
-      "epoch": 0.8866551316909725,
+      "epoch": 0.9161180476730988,
       "grad_norm": 0.0,
-      "learning_rate": 6.662906103488575e-07,
-      "loss": 0.9074,
+      "learning_rate": 3.667843197250609e-07,
+      "loss": 0.9817,
       "step": 32284
     },
     {
-      "epoch": 0.8866825959188157,
+      "epoch": 0.9161464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 6.65971388230282e-07,
-      "loss": 0.8368,
+      "learning_rate": 3.665377260959846e-07,
+      "loss": 0.7169,
       "step": 32285
     },
     {
-      "epoch": 0.886710060146659,
+      "epoch": 0.9161748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 6.656522399661557e-07,
-      "loss": 0.8479,
+      "learning_rate": 3.662912138411967e-07,
+      "loss": 0.8826,
       "step": 32286
     },
     {
-      "epoch": 0.8867375243745023,
+      "epoch": 0.9162031782065835,
       "grad_norm": 0.0,
-      "learning_rate": 6.653331655590056e-07,
-      "loss": 0.8075,
+      "learning_rate": 3.660447829627811e-07,
+      "loss": 0.7569,
       "step": 32287
     },
     {
-      "epoch": 0.8867649886023454,
+      "epoch": 0.9162315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 6.65014165011355e-07,
-      "loss": 0.9029,
+      "learning_rate": 3.657984334628173e-07,
+      "loss": 0.8162,
       "step": 32288
     },
     {
-      "epoch": 0.8867924528301887,
+      "epoch": 0.9162599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 6.646952383257299e-07,
-      "loss": 0.7873,
+      "learning_rate": 3.655521653433891e-07,
+      "loss": 0.8379,
       "step": 32289
     },
     {
-      "epoch": 0.8868199170580319,
+      "epoch": 0.9162883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 6.643763855046503e-07,
-      "loss": 0.8357,
+      "learning_rate": 3.653059786065738e-07,
+      "loss": 0.74,
       "step": 32290
     },
     {
-      "epoch": 0.8868473812858751,
+      "epoch": 0.916316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 6.640576065506432e-07,
-      "loss": 0.7991,
+      "learning_rate": 3.650598732544519e-07,
+      "loss": 0.8822,
       "step": 32291
     },
     {
-      "epoch": 0.8868748455137184,
+      "epoch": 0.9163450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 6.637389014662266e-07,
-      "loss": 0.7301,
+      "learning_rate": 3.648138492891029e-07,
+      "loss": 0.795,
       "step": 32292
     },
     {
-      "epoch": 0.8869023097415616,
+      "epoch": 0.9163734392735527,
       "grad_norm": 0.0,
-      "learning_rate": 6.634202702539249e-07,
-      "loss": 0.8473,
+      "learning_rate": 3.64567906712604e-07,
+      "loss": 0.7635,
       "step": 32293
     },
     {
-      "epoch": 0.8869297739694049,
+      "epoch": 0.9164018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 6.631017129162586e-07,
-      "loss": 0.8759,
+      "learning_rate": 3.6432204552703245e-07,
+      "loss": 0.7166,
       "step": 32294
     },
     {
-      "epoch": 0.886957238197248,
+      "epoch": 0.9164301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 6.62783229455749e-07,
-      "loss": 0.7565,
+      "learning_rate": 3.6407626573446763e-07,
+      "loss": 0.7946,
       "step": 32295
     },
     {
-      "epoch": 0.8869847024250913,
+      "epoch": 0.9164585698070374,
       "grad_norm": 0.0,
-      "learning_rate": 6.62464819874915e-07,
-      "loss": 0.7489,
+      "learning_rate": 3.638305673369824e-07,
+      "loss": 0.8439,
       "step": 32296
     },
     {
-      "epoch": 0.8870121666529346,
+      "epoch": 0.9164869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 6.62146484176276e-07,
-      "loss": 0.7763,
+      "learning_rate": 3.6358495033665286e-07,
+      "loss": 0.7282,
       "step": 32297
     },
     {
-      "epoch": 0.8870396308807778,
+      "epoch": 0.9165153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 6.618282223623507e-07,
-      "loss": 0.7619,
+      "learning_rate": 3.633394147355562e-07,
+      "loss": 0.7852,
       "step": 32298
     },
     {
-      "epoch": 0.887067095108621,
+      "epoch": 0.9165437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 6.615100344356596e-07,
-      "loss": 0.8439,
+      "learning_rate": 3.630939605357642e-07,
+      "loss": 0.8135,
       "step": 32299
     },
     {
-      "epoch": 0.8870945593364643,
+      "epoch": 0.916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 6.611919203987172e-07,
-      "loss": 0.8268,
+      "learning_rate": 3.628485877393495e-07,
+      "loss": 0.7996,
       "step": 32300
     },
     {
-      "epoch": 0.8871220235643075,
+      "epoch": 0.916600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 6.608738802540427e-07,
-      "loss": 0.7709,
+      "learning_rate": 3.6260329634838833e-07,
+      "loss": 0.7895,
       "step": 32301
     },
     {
-      "epoch": 0.8871494877921507,
+      "epoch": 0.9166288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 6.605559140041506e-07,
-      "loss": 0.8051,
+      "learning_rate": 3.6235808636494787e-07,
+      "loss": 0.8076,
       "step": 32302
     },
     {
-      "epoch": 0.8871769520199939,
+      "epoch": 0.9166572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 6.602380216515581e-07,
-      "loss": 0.729,
+      "learning_rate": 3.62112957791102e-07,
+      "loss": 0.7827,
       "step": 32303
     },
     {
-      "epoch": 0.8872044162478372,
+      "epoch": 0.9166855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 6.599202031987806e-07,
-      "loss": 0.7963,
+      "learning_rate": 3.618679106289225e-07,
+      "loss": 0.7502,
       "step": 32304
     },
     {
-      "epoch": 0.8872318804756805,
+      "epoch": 0.9167139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 6.596024586483319e-07,
-      "loss": 0.7462,
+      "learning_rate": 3.6162294488047753e-07,
+      "loss": 0.8,
       "step": 32305
     },
     {
-      "epoch": 0.8872593447035236,
+      "epoch": 0.9167423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 6.592847880027275e-07,
-      "loss": 0.9088,
+      "learning_rate": 3.6137806054783674e-07,
+      "loss": 0.8187,
       "step": 32306
     },
     {
-      "epoch": 0.8872868089313669,
+      "epoch": 0.9167707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 6.589671912644791e-07,
-      "loss": 0.8076,
+      "learning_rate": 3.611332576330695e-07,
+      "loss": 0.8184,
       "step": 32307
     },
     {
-      "epoch": 0.8873142731592101,
+      "epoch": 0.9167990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 6.586496684361033e-07,
-      "loss": 0.7702,
+      "learning_rate": 3.6088853613824195e-07,
+      "loss": 0.7574,
       "step": 32308
     },
     {
-      "epoch": 0.8873417373870534,
+      "epoch": 0.9168274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 6.583322195201069e-07,
-      "loss": 0.7504,
+      "learning_rate": 3.6064389606542237e-07,
+      "loss": 0.7076,
       "step": 32309
     },
     {
-      "epoch": 0.8873692016148966,
+      "epoch": 0.9168558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 6.580148445190049e-07,
-      "loss": 0.8685,
+      "learning_rate": 3.603993374166781e-07,
+      "loss": 0.7626,
       "step": 32310
     },
     {
-      "epoch": 0.8873966658427398,
+      "epoch": 0.9168842224744609,
       "grad_norm": 0.0,
-      "learning_rate": 6.576975434353084e-07,
-      "loss": 0.7195,
+      "learning_rate": 3.60154860194073e-07,
+      "loss": 0.8172,
       "step": 32311
     },
     {
-      "epoch": 0.8874241300705831,
+      "epoch": 0.9169125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 6.573803162715275e-07,
-      "loss": 0.8576,
+      "learning_rate": 3.599104643996731e-07,
+      "loss": 0.9004,
       "step": 32312
     },
     {
-      "epoch": 0.8874515942984263,
+      "epoch": 0.9169409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 6.570631630301738e-07,
-      "loss": 0.7731,
+      "learning_rate": 3.596661500355436e-07,
+      "loss": 0.897,
       "step": 32313
     },
     {
-      "epoch": 0.8874790585262695,
+      "epoch": 0.9169693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 6.567460837137552e-07,
-      "loss": 0.8458,
+      "learning_rate": 3.594219171037472e-07,
+      "loss": 0.7869,
       "step": 32314
     },
     {
-      "epoch": 0.8875065227541128,
+      "epoch": 0.9169977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 6.564290783247818e-07,
-      "loss": 0.8439,
+      "learning_rate": 3.5917776560634775e-07,
+      "loss": 0.685,
       "step": 32315
     },
     {
-      "epoch": 0.887533986981956,
+      "epoch": 0.9170261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 6.561121468657583e-07,
-      "loss": 0.8308,
+      "learning_rate": 3.589336955454059e-07,
+      "loss": 0.7937,
       "step": 32316
     },
     {
-      "epoch": 0.8875614512097992,
+      "epoch": 0.9170544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 6.557952893391961e-07,
-      "loss": 0.7818,
+      "learning_rate": 3.5868970692298553e-07,
+      "loss": 0.9001,
       "step": 32317
     },
     {
-      "epoch": 0.8875889154376425,
+      "epoch": 0.9170828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 6.55478505747602e-07,
-      "loss": 0.8438,
+      "learning_rate": 3.5844579974114503e-07,
+      "loss": 0.7371,
       "step": 32318
     },
     {
-      "epoch": 0.8876163796654857,
+      "epoch": 0.91711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.551617960934809e-07,
-      "loss": 0.7512,
+      "learning_rate": 3.582019740019471e-07,
+      "loss": 0.7663,
       "step": 32319
     },
     {
-      "epoch": 0.887643843893329,
+      "epoch": 0.9171396140749148,
       "grad_norm": 0.0,
-      "learning_rate": 6.548451603793393e-07,
-      "loss": 0.816,
+      "learning_rate": 3.5795822970745133e-07,
+      "loss": 0.7764,
       "step": 32320
     },
     {
-      "epoch": 0.8876713081211721,
+      "epoch": 0.9171679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 6.545285986076821e-07,
-      "loss": 0.9161,
+      "learning_rate": 3.5771456685971593e-07,
+      "loss": 0.8457,
       "step": 32321
     },
     {
-      "epoch": 0.8876987723490154,
+      "epoch": 0.9171963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 6.542121107810151e-07,
-      "loss": 0.824,
+      "learning_rate": 3.5747098546079827e-07,
+      "loss": 0.8194,
       "step": 32322
     },
     {
-      "epoch": 0.8877262365768587,
+      "epoch": 0.9172247446083995,
       "grad_norm": 0.0,
-      "learning_rate": 6.53895696901844e-07,
-      "loss": 0.8191,
+      "learning_rate": 3.5722748551275665e-07,
+      "loss": 0.8001,
       "step": 32323
     },
     {
-      "epoch": 0.8877537008047018,
+      "epoch": 0.9172531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 6.5357935697267e-07,
-      "loss": 0.8442,
+      "learning_rate": 3.569840670176483e-07,
+      "loss": 0.7399,
       "step": 32324
     },
     {
-      "epoch": 0.8877811650325451,
+      "epoch": 0.9172814982973894,
       "grad_norm": 0.0,
-      "learning_rate": 6.532630909959959e-07,
-      "loss": 0.8243,
+      "learning_rate": 3.5674072997752827e-07,
+      "loss": 0.8365,
       "step": 32325
     },
     {
-      "epoch": 0.8878086292603884,
+      "epoch": 0.9173098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 6.529468989743237e-07,
-      "loss": 0.7798,
+      "learning_rate": 3.564974743944538e-07,
+      "loss": 0.8332,
       "step": 32326
     },
     {
-      "epoch": 0.8878360934882316,
+      "epoch": 0.9173382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 6.526307809101573e-07,
-      "loss": 0.8521,
+      "learning_rate": 3.562543002704788e-07,
+      "loss": 0.7747,
       "step": 32327
     },
     {
-      "epoch": 0.8878635577160748,
+      "epoch": 0.917366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 6.523147368059957e-07,
-      "loss": 0.8763,
+      "learning_rate": 3.5601120760765604e-07,
+      "loss": 0.8238,
       "step": 32328
     },
     {
-      "epoch": 0.887891021943918,
+      "epoch": 0.9173950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 6.519987666643424e-07,
-      "loss": 0.8058,
+      "learning_rate": 3.5576819640804064e-07,
+      "loss": 0.7441,
       "step": 32329
     },
     {
-      "epoch": 0.8879184861717613,
+      "epoch": 0.9174233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 6.516828704876965e-07,
-      "loss": 0.7731,
+      "learning_rate": 3.5552526667368637e-07,
+      "loss": 0.8205,
       "step": 32330
     },
     {
-      "epoch": 0.8879459503996046,
+      "epoch": 0.9174517593643586,
       "grad_norm": 0.0,
-      "learning_rate": 6.51367048278555e-07,
-      "loss": 0.7712,
+      "learning_rate": 3.552824184066417e-07,
+      "loss": 0.6797,
       "step": 32331
     },
     {
-      "epoch": 0.8879734146274477,
+      "epoch": 0.9174801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 6.510513000394214e-07,
-      "loss": 0.789,
+      "learning_rate": 3.5503965160896045e-07,
+      "loss": 0.8179,
       "step": 32332
     },
     {
-      "epoch": 0.888000878855291,
+      "epoch": 0.9175085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 6.507356257727893e-07,
-      "loss": 0.8075,
+      "learning_rate": 3.5479696628269225e-07,
+      "loss": 0.8805,
       "step": 32333
     },
     {
-      "epoch": 0.8880283430831342,
+      "epoch": 0.9175368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 6.504200254811577e-07,
-      "loss": 0.7666,
+      "learning_rate": 3.545543624298886e-07,
+      "loss": 0.8751,
       "step": 32334
     },
     {
-      "epoch": 0.8880558073109774,
+      "epoch": 0.9175652667423383,
       "grad_norm": 0.0,
-      "learning_rate": 6.501044991670257e-07,
-      "loss": 0.8468,
+      "learning_rate": 3.54311840052598e-07,
+      "loss": 0.8672,
       "step": 32335
     },
     {
-      "epoch": 0.8880832715388207,
+      "epoch": 0.9175936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 6.497890468328894e-07,
-      "loss": 0.7736,
+      "learning_rate": 3.540693991528676e-07,
+      "loss": 0.7496,
       "step": 32336
     },
     {
-      "epoch": 0.8881107357666639,
+      "epoch": 0.917622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 6.49473668481243e-07,
-      "loss": 0.7767,
+      "learning_rate": 3.5382703973274813e-07,
+      "loss": 0.8125,
       "step": 32337
     },
     {
-      "epoch": 0.8881381999945072,
+      "epoch": 0.917650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.491583641145838e-07,
-      "loss": 0.8196,
+      "learning_rate": 3.535847617942856e-07,
+      "loss": 0.7366,
       "step": 32338
     },
     {
-      "epoch": 0.8881656642223504,
+      "epoch": 0.9176787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 6.488431337354062e-07,
-      "loss": 0.6812,
+      "learning_rate": 3.5334256533952506e-07,
+      "loss": 0.7901,
       "step": 32339
     },
     {
-      "epoch": 0.8881931284501936,
+      "epoch": 0.9177071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 6.485279773462028e-07,
-      "loss": 0.7632,
+      "learning_rate": 3.5310045037051264e-07,
+      "loss": 0.7876,
       "step": 32340
     },
     {
-      "epoch": 0.8882205926780369,
+      "epoch": 0.9177355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 6.482128949494681e-07,
-      "loss": 0.8357,
+      "learning_rate": 3.528584168892968e-07,
+      "loss": 0.819,
       "step": 32341
     },
     {
-      "epoch": 0.88824805690588,
+      "epoch": 0.9177639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 6.47897886547697e-07,
-      "loss": 0.8733,
+      "learning_rate": 3.52616464897918e-07,
+      "loss": 0.8011,
       "step": 32342
     },
     {
-      "epoch": 0.8882755211337233,
+      "epoch": 0.9177922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 6.475829521433796e-07,
-      "loss": 0.8039,
+      "learning_rate": 3.523745943984214e-07,
+      "loss": 0.8428,
       "step": 32343
     },
     {
-      "epoch": 0.8883029853615666,
+      "epoch": 0.9178206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 6.472680917390084e-07,
-      "loss": 0.8421,
+      "learning_rate": 3.521328053928519e-07,
+      "loss": 0.7576,
       "step": 32344
     },
     {
-      "epoch": 0.8883304495894098,
+      "epoch": 0.9178490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 6.469533053370747e-07,
-      "loss": 0.8085,
+      "learning_rate": 3.51891097883249e-07,
+      "loss": 0.7401,
       "step": 32345
     },
     {
-      "epoch": 0.888357913817253,
+      "epoch": 0.9178774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 6.466385929400698e-07,
-      "loss": 0.7942,
+      "learning_rate": 3.5164947187165454e-07,
+      "loss": 0.7819,
       "step": 32346
     },
     {
-      "epoch": 0.8883853780450962,
+      "epoch": 0.9179057888762769,
       "grad_norm": 0.0,
-      "learning_rate": 6.463239545504852e-07,
-      "loss": 0.8433,
+      "learning_rate": 3.514079273601134e-07,
+      "loss": 0.8074,
       "step": 32347
     },
     {
-      "epoch": 0.8884128422729395,
+      "epoch": 0.9179341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 6.460093901708087e-07,
-      "loss": 0.7908,
+      "learning_rate": 3.511664643506618e-07,
+      "loss": 0.8914,
       "step": 32348
     },
     {
-      "epoch": 0.8884403065007828,
+      "epoch": 0.9179625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 6.456948998035273e-07,
-      "loss": 0.8236,
+      "learning_rate": 3.5092508284534034e-07,
+      "loss": 0.7928,
       "step": 32349
     },
     {
-      "epoch": 0.8884677707286259,
+      "epoch": 0.9179909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 6.453804834511313e-07,
-      "loss": 0.8496,
+      "learning_rate": 3.506837828461884e-07,
+      "loss": 0.8504,
       "step": 32350
     },
     {
-      "epoch": 0.8884952349564692,
+      "epoch": 0.9180192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 6.450661411161097e-07,
-      "loss": 0.8093,
+      "learning_rate": 3.504425643552445e-07,
+      "loss": 0.7686,
       "step": 32351
     },
     {
-      "epoch": 0.8885226991843125,
+      "epoch": 0.9180476730987515,
       "grad_norm": 0.0,
-      "learning_rate": 6.447518728009472e-07,
-      "loss": 0.7852,
+      "learning_rate": 3.50201427374548e-07,
+      "loss": 0.6963,
       "step": 32352
     },
     {
-      "epoch": 0.8885501634121556,
+      "epoch": 0.9180760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 6.444376785081319e-07,
-      "loss": 0.7735,
+      "learning_rate": 3.499603719061317e-07,
+      "loss": 0.6557,
       "step": 32353
     },
     {
-      "epoch": 0.8885776276399989,
+      "epoch": 0.9181044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 6.441235582401517e-07,
-      "loss": 0.7708,
+      "learning_rate": 3.4971939795203414e-07,
+      "loss": 0.8079,
       "step": 32354
     },
     {
-      "epoch": 0.8886050918678421,
+      "epoch": 0.9181328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 6.438095119994892e-07,
-      "loss": 0.7746,
+      "learning_rate": 3.494785055142924e-07,
+      "loss": 0.8383,
       "step": 32355
     },
     {
-      "epoch": 0.8886325560956854,
+      "epoch": 0.918161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 6.4349553978863e-07,
-      "loss": 0.7915,
+      "learning_rate": 3.4923769459493717e-07,
+      "loss": 0.7889,
       "step": 32356
     },
     {
-      "epoch": 0.8886600203235286,
+      "epoch": 0.9181895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 6.431816416100578e-07,
-      "loss": 0.7891,
+      "learning_rate": 3.4899696519600566e-07,
+      "loss": 0.7625,
       "step": 32357
     },
     {
-      "epoch": 0.8886874845513718,
+      "epoch": 0.9182179341657207,
       "grad_norm": 0.0,
-      "learning_rate": 6.428678174662573e-07,
-      "loss": 0.7906,
+      "learning_rate": 3.4875631731953184e-07,
+      "loss": 0.8661,
       "step": 32358
     },
     {
-      "epoch": 0.8887149487792151,
+      "epoch": 0.9182463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 6.425540673597108e-07,
-      "loss": 0.8849,
+      "learning_rate": 3.485157509675463e-07,
+      "loss": 0.7228,
       "step": 32359
     },
     {
-      "epoch": 0.8887424130070583,
+      "epoch": 0.9182746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 6.42240391292902e-07,
-      "loss": 0.7599,
+      "learning_rate": 3.4827526614208186e-07,
+      "loss": 0.7974,
       "step": 32360
     },
     {
-      "epoch": 0.8887698772349015,
+      "epoch": 0.9183030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 6.419267892683112e-07,
-      "loss": 0.7458,
+      "learning_rate": 3.480348628451713e-07,
+      "loss": 0.8094,
       "step": 32361
     },
     {
-      "epoch": 0.8887973414627448,
+      "epoch": 0.9183314415437004,
       "grad_norm": 0.0,
-      "learning_rate": 6.416132612884207e-07,
-      "loss": 0.8993,
+      "learning_rate": 3.47794541078843e-07,
+      "loss": 0.8249,
       "step": 32362
     },
     {
-      "epoch": 0.888824805690588,
+      "epoch": 0.9183598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 6.41299807355713e-07,
-      "loss": 0.8562,
+      "learning_rate": 3.4755430084512765e-07,
+      "loss": 0.8825,
       "step": 32363
     },
     {
-      "epoch": 0.8888522699184312,
+      "epoch": 0.9183881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 6.409864274726651e-07,
-      "loss": 0.8268,
+      "learning_rate": 3.4731414214605574e-07,
+      "loss": 0.7967,
       "step": 32364
     },
     {
-      "epoch": 0.8888797341462745,
+      "epoch": 0.918416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 6.406731216417572e-07,
-      "loss": 0.813,
+      "learning_rate": 3.4707406498365457e-07,
+      "loss": 0.8833,
       "step": 32365
     },
     {
-      "epoch": 0.8889071983741177,
+      "epoch": 0.9184449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 6.403598898654705e-07,
-      "loss": 0.757,
+      "learning_rate": 3.468340693599548e-07,
+      "loss": 0.8723,
       "step": 32366
     },
     {
-      "epoch": 0.888934662601961,
+      "epoch": 0.9184733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 6.400467321462811e-07,
-      "loss": 0.6872,
+      "learning_rate": 3.4659415527697916e-07,
+      "loss": 0.749,
       "step": 32367
     },
     {
-      "epoch": 0.8889621268298041,
+      "epoch": 0.9185017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 6.397336484866668e-07,
-      "loss": 0.7993,
+      "learning_rate": 3.463543227367572e-07,
+      "loss": 0.8536,
       "step": 32368
     },
     {
-      "epoch": 0.8889895910576474,
+      "epoch": 0.9185300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 6.394206388891066e-07,
-      "loss": 0.8049,
+      "learning_rate": 3.461145717413161e-07,
+      "loss": 0.8859,
       "step": 32369
     },
     {
-      "epoch": 0.8890170552854907,
+      "epoch": 0.9185584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 6.391077033560755e-07,
-      "loss": 0.7472,
+      "learning_rate": 3.4587490229267774e-07,
+      "loss": 0.7895,
       "step": 32370
     },
     {
-      "epoch": 0.8890445195133339,
+      "epoch": 0.9185868331441543,
       "grad_norm": 0.0,
-      "learning_rate": 6.387948418900524e-07,
-      "loss": 0.8126,
+      "learning_rate": 3.45635314392867e-07,
+      "loss": 0.8172,
       "step": 32371
     },
     {
-      "epoch": 0.8890719837411771,
+      "epoch": 0.9186152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 6.3848205449351e-07,
-      "loss": 0.8195,
+      "learning_rate": 3.4539580804391124e-07,
+      "loss": 0.8175,
       "step": 32372
     },
     {
-      "epoch": 0.8890994479690203,
+      "epoch": 0.9186435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 6.381693411689227e-07,
-      "loss": 0.8227,
+      "learning_rate": 3.4515638324782883e-07,
+      "loss": 0.8656,
       "step": 32373
     },
     {
-      "epoch": 0.8891269121968636,
+      "epoch": 0.918671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 6.378567019187665e-07,
-      "loss": 0.8259,
+      "learning_rate": 3.449170400066448e-07,
+      "loss": 0.8359,
       "step": 32374
     },
     {
-      "epoch": 0.8891543764247068,
+      "epoch": 0.9187003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 6.375441367455138e-07,
-      "loss": 0.7613,
+      "learning_rate": 3.4467777832238204e-07,
+      "loss": 0.797,
       "step": 32375
     },
     {
-      "epoch": 0.88918184065255,
+      "epoch": 0.9187287173666289,
       "grad_norm": 0.0,
-      "learning_rate": 6.372316456516392e-07,
-      "loss": 0.7537,
+      "learning_rate": 3.444385981970588e-07,
+      "loss": 0.8996,
       "step": 32376
     },
     {
-      "epoch": 0.8892093048803933,
+      "epoch": 0.9187570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 6.369192286396165e-07,
-      "loss": 0.6844,
+      "learning_rate": 3.4419949963269693e-07,
+      "loss": 0.8767,
       "step": 32377
     },
     {
-      "epoch": 0.8892367691082366,
+      "epoch": 0.9187854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 6.366068857119135e-07,
-      "loss": 0.8427,
+      "learning_rate": 3.4396048263131697e-07,
+      "loss": 0.7383,
       "step": 32378
     },
     {
-      "epoch": 0.8892642333360797,
+      "epoch": 0.9188138479001136,
       "grad_norm": 0.0,
-      "learning_rate": 6.362946168710049e-07,
-      "loss": 0.8479,
+      "learning_rate": 3.437215471949351e-07,
+      "loss": 0.7104,
       "step": 32379
     },
     {
-      "epoch": 0.889291697563923,
+      "epoch": 0.9188422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 6.3598242211936e-07,
-      "loss": 0.8059,
+      "learning_rate": 3.43482693325573e-07,
+      "loss": 0.7873,
       "step": 32380
     },
     {
-      "epoch": 0.8893191617917662,
+      "epoch": 0.9188706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 6.356703014594489e-07,
-      "loss": 0.7901,
+      "learning_rate": 3.4324392102524585e-07,
+      "loss": 0.7052,
       "step": 32381
     },
     {
-      "epoch": 0.8893466260196095,
+      "epoch": 0.9188989784335981,
       "grad_norm": 0.0,
-      "learning_rate": 6.35358254893742e-07,
-      "loss": 0.7769,
+      "learning_rate": 3.430052302959708e-07,
+      "loss": 0.7653,
       "step": 32382
     },
     {
-      "epoch": 0.8893740902474527,
+      "epoch": 0.9189273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 6.350462824247094e-07,
-      "loss": 0.837,
+      "learning_rate": 3.427666211397662e-07,
+      "loss": 0.7974,
       "step": 32383
     },
     {
-      "epoch": 0.8894015544752959,
+      "epoch": 0.918955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 6.34734384054817e-07,
-      "loss": 0.7185,
+      "learning_rate": 3.42528093558645e-07,
+      "loss": 0.8755,
       "step": 32384
     },
     {
-      "epoch": 0.8894290187031392,
+      "epoch": 0.9189841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 6.344225597865338e-07,
-      "loss": 0.8583,
+      "learning_rate": 3.422896475546234e-07,
+      "loss": 0.8359,
       "step": 32385
     },
     {
-      "epoch": 0.8894564829309825,
+      "epoch": 0.9190124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 6.34110809622327e-07,
-      "loss": 0.8041,
+      "learning_rate": 3.4205128312971623e-07,
+      "loss": 0.8285,
       "step": 32386
     },
     {
-      "epoch": 0.8894839471588256,
+      "epoch": 0.9190408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 6.337991335646654e-07,
-      "loss": 0.876,
+      "learning_rate": 3.4181300028593433e-07,
+      "loss": 0.8477,
       "step": 32387
     },
     {
-      "epoch": 0.8895114113866689,
+      "epoch": 0.9190692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 6.334875316160105e-07,
-      "loss": 0.8572,
+      "learning_rate": 3.4157479902529267e-07,
+      "loss": 0.8459,
       "step": 32388
     },
     {
-      "epoch": 0.8895388756145121,
+      "epoch": 0.9190976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 6.331760037788337e-07,
-      "loss": 0.85,
+      "learning_rate": 3.413366793498041e-07,
+      "loss": 0.7475,
       "step": 32389
     },
     {
-      "epoch": 0.8895663398423553,
+      "epoch": 0.9191259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 6.328645500555941e-07,
-      "loss": 0.8771,
+      "learning_rate": 3.4109864126147806e-07,
+      "loss": 0.7701,
       "step": 32390
     },
     {
-      "epoch": 0.8895938040701986,
+      "epoch": 0.9191543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 6.325531704487586e-07,
-      "loss": 0.7702,
+      "learning_rate": 3.4086068476232525e-07,
+      "loss": 0.7158,
       "step": 32391
     },
     {
-      "epoch": 0.8896212682980418,
+      "epoch": 0.9191827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 6.32241864960792e-07,
-      "loss": 0.775,
+      "learning_rate": 3.4062280985435736e-07,
+      "loss": 0.8351,
       "step": 32392
     },
     {
-      "epoch": 0.889648732525885,
+      "epoch": 0.919211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 6.319306335941555e-07,
-      "loss": 0.7231,
+      "learning_rate": 3.4038501653958277e-07,
+      "loss": 0.8288,
       "step": 32393
     },
     {
-      "epoch": 0.8896761967537282,
+      "epoch": 0.9192395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 6.316194763513128e-07,
-      "loss": 0.7933,
+      "learning_rate": 3.401473048200088e-07,
+      "loss": 0.6812,
       "step": 32394
     },
     {
-      "epoch": 0.8897036609815715,
+      "epoch": 0.9192678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 6.313083932347275e-07,
-      "loss": 0.7935,
+      "learning_rate": 3.39909674697646e-07,
+      "loss": 0.7693,
       "step": 32395
     },
     {
-      "epoch": 0.8897311252094148,
+      "epoch": 0.9192962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 6.309973842468597e-07,
-      "loss": 0.8156,
+      "learning_rate": 3.3967212617449954e-07,
+      "loss": 0.7539,
       "step": 32396
     },
     {
-      "epoch": 0.8897585894372579,
+      "epoch": 0.9193246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 6.306864493901676e-07,
-      "loss": 0.8198,
+      "learning_rate": 3.394346592525788e-07,
+      "loss": 0.7089,
       "step": 32397
     },
     {
-      "epoch": 0.8897860536651012,
+      "epoch": 0.9193530079455164,
       "grad_norm": 0.0,
-      "learning_rate": 6.303755886671137e-07,
-      "loss": 0.77,
+      "learning_rate": 3.391972739338867e-07,
+      "loss": 0.8232,
       "step": 32398
     },
     {
-      "epoch": 0.8898135178929445,
+      "epoch": 0.9193813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 6.30064802080158e-07,
-      "loss": 0.7871,
+      "learning_rate": 3.3895997022042826e-07,
+      "loss": 0.8238,
       "step": 32399
     },
     {
-      "epoch": 0.8898409821207877,
+      "epoch": 0.9194097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 6.297540896317589e-07,
-      "loss": 0.6899,
+      "learning_rate": 3.3872274811421193e-07,
+      "loss": 0.6935,
       "step": 32400
     },
     {
-      "epoch": 0.8898684463486309,
+      "epoch": 0.9194381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 6.294434513243764e-07,
-      "loss": 0.7732,
+      "learning_rate": 3.3848560761723715e-07,
+      "loss": 0.8288,
       "step": 32401
     },
     {
-      "epoch": 0.8898959105764741,
+      "epoch": 0.919466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 6.291328871604663e-07,
-      "loss": 0.7092,
+      "learning_rate": 3.382485487315079e-07,
+      "loss": 0.7703,
       "step": 32402
     },
     {
-      "epoch": 0.8899233748043174,
+      "epoch": 0.919494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 6.288223971424878e-07,
-      "loss": 0.8177,
+      "learning_rate": 3.380115714590293e-07,
+      "loss": 0.8238,
       "step": 32403
     },
     {
-      "epoch": 0.8899508390321607,
+      "epoch": 0.9195232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 6.285119812728946e-07,
-      "loss": 0.706,
+      "learning_rate": 3.377746758018008e-07,
+      "loss": 0.8388,
       "step": 32404
     },
     {
-      "epoch": 0.8899783032600038,
+      "epoch": 0.9195516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 6.282016395541457e-07,
-      "loss": 0.7376,
+      "learning_rate": 3.3753786176182303e-07,
+      "loss": 0.8552,
       "step": 32405
     },
     {
-      "epoch": 0.8900057674878471,
+      "epoch": 0.9195800227014755,
       "grad_norm": 0.0,
-      "learning_rate": 6.278913719886948e-07,
-      "loss": 0.8059,
+      "learning_rate": 3.3730112934109885e-07,
+      "loss": 0.7915,
       "step": 32406
     },
     {
-      "epoch": 0.8900332317156903,
+      "epoch": 0.9196083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 6.275811785789998e-07,
-      "loss": 0.853,
+      "learning_rate": 3.370644785416255e-07,
+      "loss": 0.7644,
       "step": 32407
     },
     {
-      "epoch": 0.8900606959435335,
+      "epoch": 0.9196367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 6.272710593275111e-07,
-      "loss": 0.7382,
+      "learning_rate": 3.3682790936540257e-07,
+      "loss": 0.8167,
       "step": 32408
     },
     {
-      "epoch": 0.8900881601713768,
+      "epoch": 0.9196651532349602,
       "grad_norm": 0.0,
-      "learning_rate": 6.269610142366855e-07,
-      "loss": 0.7928,
+      "learning_rate": 3.365914218144295e-07,
+      "loss": 0.7574,
       "step": 32409
     },
     {
-      "epoch": 0.89011562439922,
+      "epoch": 0.9196935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 6.266510433089734e-07,
-      "loss": 0.7868,
+      "learning_rate": 3.363550158907014e-07,
+      "loss": 0.8161,
       "step": 32410
     },
     {
-      "epoch": 0.8901430886270633,
+      "epoch": 0.9197219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 6.263411465468317e-07,
-      "loss": 0.845,
+      "learning_rate": 3.361186915962178e-07,
+      "loss": 0.8637,
       "step": 32411
     },
     {
-      "epoch": 0.8901705528549065,
+      "epoch": 0.9197502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 6.260313239527083e-07,
-      "loss": 0.8322,
+      "learning_rate": 3.358824489329748e-07,
+      "loss": 0.7707,
       "step": 32412
     },
     {
-      "epoch": 0.8901980170827497,
+      "epoch": 0.9197786606129399,
       "grad_norm": 0.0,
-      "learning_rate": 6.25721575529058e-07,
-      "loss": 0.7036,
+      "learning_rate": 3.356462879029665e-07,
+      "loss": 0.8057,
       "step": 32413
     },
     {
-      "epoch": 0.890225481310593,
+      "epoch": 0.9198070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 6.254119012783278e-07,
-      "loss": 0.773,
+      "learning_rate": 3.354102085081878e-07,
+      "loss": 0.8386,
       "step": 32414
     },
     {
-      "epoch": 0.8902529455384361,
+      "epoch": 0.9198354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 6.251023012029712e-07,
-      "loss": 0.8557,
+      "learning_rate": 3.3517421075063507e-07,
+      "loss": 0.9007,
       "step": 32415
     },
     {
-      "epoch": 0.8902804097662794,
+      "epoch": 0.9198637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 6.24792775305435e-07,
-      "loss": 0.7445,
+      "learning_rate": 3.349382946322999e-07,
+      "loss": 0.8553,
       "step": 32416
     },
     {
-      "epoch": 0.8903078739941227,
+      "epoch": 0.9198921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 6.244833235881709e-07,
-      "loss": 0.8008,
+      "learning_rate": 3.3470246015517403e-07,
+      "loss": 0.8149,
       "step": 32417
     },
     {
-      "epoch": 0.8903353382219659,
+      "epoch": 0.9199205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 6.241739460536278e-07,
-      "loss": 0.7738,
+      "learning_rate": 3.3446670732125373e-07,
+      "loss": 0.8016,
       "step": 32418
     },
     {
-      "epoch": 0.8903628024498091,
+      "epoch": 0.9199489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 6.238646427042505e-07,
-      "loss": 0.8042,
+      "learning_rate": 3.342310361325252e-07,
+      "loss": 0.8277,
       "step": 32419
     },
     {
-      "epoch": 0.8903902666776523,
+      "epoch": 0.9199772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 6.235554135424904e-07,
-      "loss": 0.8195,
+      "learning_rate": 3.3399544659098224e-07,
+      "loss": 0.7958,
       "step": 32420
     },
     {
-      "epoch": 0.8904177309054956,
+      "epoch": 0.920005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 6.232462585707899e-07,
-      "loss": 0.8403,
+      "learning_rate": 3.3375993869861454e-07,
+      "loss": 0.8253,
       "step": 32421
     },
     {
-      "epoch": 0.8904451951333389,
+      "epoch": 0.9200340522133938,
       "grad_norm": 0.0,
-      "learning_rate": 6.229371777915972e-07,
-      "loss": 0.6744,
+      "learning_rate": 3.3352451245741155e-07,
+      "loss": 0.7498,
       "step": 32422
     },
     {
-      "epoch": 0.890472659361182,
+      "epoch": 0.9200624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 6.226281712073579e-07,
-      "loss": 0.8645,
+      "learning_rate": 3.332891678693617e-07,
+      "loss": 0.8512,
       "step": 32423
     },
     {
-      "epoch": 0.8905001235890253,
+      "epoch": 0.9200908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 6.223192388205169e-07,
-      "loss": 0.8533,
+      "learning_rate": 3.3305390493645115e-07,
+      "loss": 0.8646,
       "step": 32424
     },
     {
-      "epoch": 0.8905275878168686,
+      "epoch": 0.9201191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 6.220103806335199e-07,
-      "loss": 0.8348,
+      "learning_rate": 3.3281872366066835e-07,
+      "loss": 0.8732,
       "step": 32425
     },
     {
-      "epoch": 0.8905550520447117,
+      "epoch": 0.9201475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 6.217015966488083e-07,
-      "loss": 0.887,
+      "learning_rate": 3.325836240440028e-07,
+      "loss": 0.8159,
       "step": 32426
     },
     {
-      "epoch": 0.890582516272555,
+      "epoch": 0.9201759364358684,
       "grad_norm": 0.0,
-      "learning_rate": 6.21392886868828e-07,
-      "loss": 0.8615,
+      "learning_rate": 3.3234860608843623e-07,
+      "loss": 0.843,
       "step": 32427
     },
     {
-      "epoch": 0.8906099805003982,
+      "epoch": 0.9202043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 6.21084251296018e-07,
-      "loss": 0.7941,
+      "learning_rate": 3.3211366979595596e-07,
+      "loss": 0.7289,
       "step": 32428
     },
     {
-      "epoch": 0.8906374447282415,
+      "epoch": 0.9202326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 6.207756899328233e-07,
-      "loss": 0.8674,
+      "learning_rate": 3.3187881516854703e-07,
+      "loss": 0.7289,
       "step": 32429
     },
     {
-      "epoch": 0.8906649089560847,
+      "epoch": 0.9202610669693531,
       "grad_norm": 0.0,
-      "learning_rate": 6.204672027816838e-07,
-      "loss": 0.8181,
+      "learning_rate": 3.3164404220819236e-07,
+      "loss": 0.8453,
       "step": 32430
     },
     {
-      "epoch": 0.8906923731839279,
+      "epoch": 0.9202894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 6.201587898450434e-07,
-      "loss": 0.8029,
+      "learning_rate": 3.3140935091687367e-07,
+      "loss": 0.8315,
       "step": 32431
     },
     {
-      "epoch": 0.8907198374117712,
+      "epoch": 0.9203178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 6.198504511253378e-07,
-      "loss": 0.825,
+      "learning_rate": 3.31174741296576e-07,
+      "loss": 0.6964,
       "step": 32432
     },
     {
-      "epoch": 0.8907473016396144,
+      "epoch": 0.9203461975028376,
       "grad_norm": 0.0,
-      "learning_rate": 6.195421866250095e-07,
-      "loss": 0.7957,
+      "learning_rate": 3.3094021334928007e-07,
+      "loss": 0.7261,
       "step": 32433
     },
     {
-      "epoch": 0.8907747658674576,
+      "epoch": 0.9203745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 6.192339963464966e-07,
-      "loss": 0.8824,
+      "learning_rate": 3.3070576707696536e-07,
+      "loss": 0.76,
       "step": 32434
     },
     {
-      "epoch": 0.8908022300953009,
+      "epoch": 0.9204029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 6.189258802922404e-07,
-      "loss": 0.6779,
+      "learning_rate": 3.304714024816158e-07,
+      "loss": 0.9017,
       "step": 32435
     },
     {
-      "epoch": 0.8908296943231441,
+      "epoch": 0.9204313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 6.186178384646757e-07,
-      "loss": 0.9058,
+      "learning_rate": 3.3023711956520767e-07,
+      "loss": 0.7398,
       "step": 32436
     },
     {
-      "epoch": 0.8908571585509873,
+      "epoch": 0.9204597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 6.183098708662394e-07,
-      "loss": 0.7154,
+      "learning_rate": 3.300029183297204e-07,
+      "loss": 0.9238,
       "step": 32437
     },
     {
-      "epoch": 0.8908846227788306,
+      "epoch": 0.9204880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 6.180019774993696e-07,
-      "loss": 0.8449,
+      "learning_rate": 3.2976879877713584e-07,
+      "loss": 0.7642,
       "step": 32438
     },
     {
-      "epoch": 0.8909120870066738,
+      "epoch": 0.920516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 6.17694158366502e-07,
-      "loss": 0.7739,
+      "learning_rate": 3.2953476090942685e-07,
+      "loss": 0.9384,
       "step": 32439
     },
     {
-      "epoch": 0.8909395512345171,
+      "epoch": 0.9205448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 6.173864134700724e-07,
-      "loss": 0.8721,
+      "learning_rate": 3.293008047285728e-07,
+      "loss": 0.8449,
       "step": 32440
     },
     {
-      "epoch": 0.8909670154623602,
+      "epoch": 0.9205732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 6.170787428125169e-07,
-      "loss": 0.889,
+      "learning_rate": 3.29066930236549e-07,
+      "loss": 0.8544,
       "step": 32441
     },
     {
-      "epoch": 0.8909944796902035,
+      "epoch": 0.9206015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 6.1677114639627e-07,
-      "loss": 0.7583,
+      "learning_rate": 3.2883313743533265e-07,
+      "loss": 0.8528,
       "step": 32442
     },
     {
-      "epoch": 0.8910219439180468,
+      "epoch": 0.9206299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 6.164636242237631e-07,
-      "loss": 0.7901,
+      "learning_rate": 3.2859942632689765e-07,
+      "loss": 0.6908,
       "step": 32443
     },
     {
-      "epoch": 0.89104940814589,
+      "epoch": 0.9206583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 6.161561762974322e-07,
-      "loss": 0.6753,
+      "learning_rate": 3.28365796913217e-07,
+      "loss": 0.8194,
       "step": 32444
     },
     {
-      "epoch": 0.8910768723737332,
+      "epoch": 0.9206867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 6.158488026197074e-07,
-      "loss": 0.8236,
+      "learning_rate": 3.2813224919626576e-07,
+      "loss": 0.7982,
       "step": 32445
     },
     {
-      "epoch": 0.8911043366015764,
+      "epoch": 0.9207150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 6.155415031930212e-07,
-      "loss": 0.8089,
+      "learning_rate": 3.2789878317801673e-07,
+      "loss": 0.8896,
       "step": 32446
     },
     {
-      "epoch": 0.8911318008294197,
+      "epoch": 0.9207434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 6.152342780198062e-07,
-      "loss": 0.7689,
+      "learning_rate": 3.276653988604406e-07,
+      "loss": 0.7961,
       "step": 32447
     },
     {
-      "epoch": 0.891159265057263,
+      "epoch": 0.9207718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 6.149271271024948e-07,
-      "loss": 0.7932,
+      "learning_rate": 3.274320962455102e-07,
+      "loss": 0.7922,
       "step": 32448
     },
     {
-      "epoch": 0.8911867292851061,
+      "epoch": 0.920800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 6.146200504435141e-07,
-      "loss": 0.823,
+      "learning_rate": 3.2719887533519626e-07,
+      "loss": 0.8666,
       "step": 32449
     },
     {
-      "epoch": 0.8912141935129494,
+      "epoch": 0.9208286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 6.143130480452941e-07,
-      "loss": 0.8077,
+      "learning_rate": 3.269657361314671e-07,
+      "loss": 0.7677,
       "step": 32450
     },
     {
-      "epoch": 0.8912416577407927,
+      "epoch": 0.9208569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 6.140061199102676e-07,
-      "loss": 0.7979,
+      "learning_rate": 3.267326786362934e-07,
+      "loss": 0.8625,
       "step": 32451
     },
     {
-      "epoch": 0.8912691219686358,
+      "epoch": 0.9208853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 6.13699266040858e-07,
-      "loss": 0.9023,
+      "learning_rate": 3.2649970285164477e-07,
+      "loss": 0.8465,
       "step": 32452
     },
     {
-      "epoch": 0.8912965861964791,
+      "epoch": 0.9209137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 6.133924864394957e-07,
-      "loss": 0.7783,
+      "learning_rate": 3.2626680877948624e-07,
+      "loss": 0.8053,
       "step": 32453
     },
     {
-      "epoch": 0.8913240504243223,
+      "epoch": 0.9209421112372305,
       "grad_norm": 0.0,
-      "learning_rate": 6.130857811086099e-07,
-      "loss": 0.7615,
+      "learning_rate": 3.2603399642178737e-07,
+      "loss": 0.7944,
       "step": 32454
     },
     {
-      "epoch": 0.8913515146521656,
+      "epoch": 0.9209704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 6.127791500506241e-07,
-      "loss": 0.7758,
+      "learning_rate": 3.258012657805154e-07,
+      "loss": 0.8887,
       "step": 32455
     },
     {
-      "epoch": 0.8913789788800088,
+      "epoch": 0.9209988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 6.124725932679665e-07,
-      "loss": 0.7128,
+      "learning_rate": 3.255686168576333e-07,
+      "loss": 0.8558,
       "step": 32456
     },
     {
-      "epoch": 0.891406443107852,
+      "epoch": 0.921027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 6.121661107630616e-07,
-      "loss": 0.7932,
+      "learning_rate": 3.2533604965510946e-07,
+      "loss": 0.879,
       "step": 32457
     },
     {
-      "epoch": 0.8914339073356953,
+      "epoch": 0.92105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.118597025383344e-07,
-      "loss": 0.8064,
+      "learning_rate": 3.251035641749056e-07,
+      "loss": 0.9252,
       "step": 32458
     },
     {
-      "epoch": 0.8914613715635384,
+      "epoch": 0.9210839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 6.115533685962116e-07,
-      "loss": 0.7444,
+      "learning_rate": 3.248711604189869e-07,
+      "loss": 0.9197,
       "step": 32459
     },
     {
-      "epoch": 0.8914888357913817,
+      "epoch": 0.9211123723041997,
       "grad_norm": 0.0,
-      "learning_rate": 6.112471089391159e-07,
-      "loss": 0.7909,
+      "learning_rate": 3.246388383893184e-07,
+      "loss": 0.8581,
       "step": 32460
     },
     {
-      "epoch": 0.891516300019225,
+      "epoch": 0.9211407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 6.109409235694674e-07,
-      "loss": 0.8522,
+      "learning_rate": 3.2440659808785857e-07,
+      "loss": 0.8282,
       "step": 32461
     },
     {
-      "epoch": 0.8915437642470682,
+      "epoch": 0.9211691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 6.106348124896922e-07,
-      "loss": 0.9097,
+      "learning_rate": 3.2417443951657133e-07,
+      "loss": 0.7209,
       "step": 32462
     },
     {
-      "epoch": 0.8915712284749114,
+      "epoch": 0.9211975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 6.103287757022114e-07,
-      "loss": 0.7471,
+      "learning_rate": 3.2394236267741965e-07,
+      "loss": 0.8337,
       "step": 32463
     },
     {
-      "epoch": 0.8915986927027547,
+      "epoch": 0.9212258796821794,
       "grad_norm": 0.0,
-      "learning_rate": 6.100228132094465e-07,
-      "loss": 0.7492,
+      "learning_rate": 3.2371036757235963e-07,
+      "loss": 0.8671,
       "step": 32464
     },
     {
-      "epoch": 0.8916261569305979,
+      "epoch": 0.9212542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 6.097169250138179e-07,
-      "loss": 0.7711,
+      "learning_rate": 3.234784542033531e-07,
+      "loss": 0.7946,
       "step": 32465
     },
     {
-      "epoch": 0.8916536211584412,
+      "epoch": 0.9212826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 6.094111111177481e-07,
-      "loss": 0.7573,
+      "learning_rate": 3.232466225723607e-07,
+      "loss": 0.7866,
       "step": 32466
     },
     {
-      "epoch": 0.8916810853862843,
+      "epoch": 0.921311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 6.091053715236539e-07,
-      "loss": 0.7846,
+      "learning_rate": 3.230148726813376e-07,
+      "loss": 0.8367,
       "step": 32467
     },
     {
-      "epoch": 0.8917085496141276,
+      "epoch": 0.9213393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 6.087997062339567e-07,
-      "loss": 0.7681,
+      "learning_rate": 3.227832045322432e-07,
+      "loss": 0.8259,
       "step": 32468
     },
     {
-      "epoch": 0.8917360138419709,
+      "epoch": 0.9213677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 6.084941152510737e-07,
-      "loss": 0.7663,
+      "learning_rate": 3.2255161812703496e-07,
+      "loss": 0.7325,
       "step": 32469
     },
     {
-      "epoch": 0.891763478069814,
+      "epoch": 0.9213961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 6.081885985774216e-07,
-      "loss": 0.7869,
+      "learning_rate": 3.2232011346766677e-07,
+      "loss": 0.7551,
       "step": 32470
     },
     {
-      "epoch": 0.8917909422976573,
+      "epoch": 0.9214245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 6.078831562154208e-07,
-      "loss": 0.8336,
+      "learning_rate": 3.2208869055609604e-07,
+      "loss": 0.6814,
       "step": 32471
     },
     {
-      "epoch": 0.8918184065255005,
+      "epoch": 0.9214528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 6.075777881674871e-07,
-      "loss": 0.828,
+      "learning_rate": 3.218573493942767e-07,
+      "loss": 0.734,
       "step": 32472
     },
     {
-      "epoch": 0.8918458707533438,
+      "epoch": 0.9214812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 6.072724944360354e-07,
-      "loss": 0.7149,
+      "learning_rate": 3.216260899841639e-07,
+      "loss": 0.881,
       "step": 32473
     },
     {
-      "epoch": 0.891873334981187,
+      "epoch": 0.9215096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 6.069672750234822e-07,
-      "loss": 0.7545,
+      "learning_rate": 3.213949123277127e-07,
+      "loss": 0.7743,
       "step": 32474
     },
     {
-      "epoch": 0.8919007992090302,
+      "epoch": 0.9215380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 6.066621299322429e-07,
-      "loss": 0.8161,
+      "learning_rate": 3.2116381642687154e-07,
+      "loss": 0.8839,
       "step": 32475
     },
     {
-      "epoch": 0.8919282634368735,
+      "epoch": 0.921566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 6.063570591647306e-07,
-      "loss": 0.8943,
+      "learning_rate": 3.2093280228359447e-07,
+      "loss": 0.8234,
       "step": 32476
     },
     {
-      "epoch": 0.8919557276647168,
+      "epoch": 0.9215947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 6.060520627233602e-07,
-      "loss": 0.7573,
+      "learning_rate": 3.207018698998343e-07,
+      "loss": 0.7424,
       "step": 32477
     },
     {
-      "epoch": 0.8919831918925599,
+      "epoch": 0.9216231555051079,
       "grad_norm": 0.0,
-      "learning_rate": 6.057471406105464e-07,
-      "loss": 0.8252,
+      "learning_rate": 3.2047101927753956e-07,
+      "loss": 0.7581,
       "step": 32478
     },
     {
-      "epoch": 0.8920106561204032,
+      "epoch": 0.9216515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 6.054422928286984e-07,
-      "loss": 0.7751,
+      "learning_rate": 3.2024025041866194e-07,
+      "loss": 0.8369,
       "step": 32479
     },
     {
-      "epoch": 0.8920381203482464,
+      "epoch": 0.9216799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 6.0513751938023e-07,
-      "loss": 0.8926,
+      "learning_rate": 3.2000956332514987e-07,
+      "loss": 0.8006,
       "step": 32480
     },
     {
-      "epoch": 0.8920655845760896,
+      "epoch": 0.9217082860385926,
       "grad_norm": 0.0,
-      "learning_rate": 6.048328202675524e-07,
-      "loss": 0.7699,
+      "learning_rate": 3.1977895799895185e-07,
+      "loss": 0.7651,
       "step": 32481
     },
     {
-      "epoch": 0.8920930488039329,
+      "epoch": 0.9217366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 6.04528195493077e-07,
-      "loss": 0.8453,
+      "learning_rate": 3.195484344420163e-07,
+      "loss": 0.8428,
       "step": 32482
     },
     {
-      "epoch": 0.8921205130317761,
+      "epoch": 0.9217650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.042236450592165e-07,
-      "loss": 0.7508,
+      "learning_rate": 3.193179926562917e-07,
+      "loss": 0.8262,
       "step": 32483
     },
     {
-      "epoch": 0.8921479772596194,
+      "epoch": 0.9217934165720771,
       "grad_norm": 0.0,
-      "learning_rate": 6.039191689683766e-07,
-      "loss": 0.8787,
+      "learning_rate": 3.1908763264372087e-07,
+      "loss": 0.8425,
       "step": 32484
     },
     {
-      "epoch": 0.8921754414874625,
+      "epoch": 0.9218217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 6.036147672229675e-07,
-      "loss": 0.8106,
+      "learning_rate": 3.1885735440625343e-07,
+      "loss": 0.7648,
       "step": 32485
     },
     {
-      "epoch": 0.8922029057153058,
+      "epoch": 0.921850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 6.033104398253975e-07,
-      "loss": 0.8006,
+      "learning_rate": 3.186271579458333e-07,
+      "loss": 0.8534,
       "step": 32486
     },
     {
-      "epoch": 0.8922303699431491,
+      "epoch": 0.9218785471055618,
       "grad_norm": 0.0,
-      "learning_rate": 6.030061867780757e-07,
-      "loss": 0.7663,
+      "learning_rate": 3.1839704326440235e-07,
+      "loss": 0.8502,
       "step": 32487
     },
     {
-      "epoch": 0.8922578341709922,
+      "epoch": 0.9219069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 6.02702008083409e-07,
-      "loss": 0.6894,
+      "learning_rate": 3.1816701036390897e-07,
+      "loss": 0.8172,
       "step": 32488
     },
     {
-      "epoch": 0.8922852983988355,
+      "epoch": 0.9219353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 6.023979037438044e-07,
-      "loss": 0.8575,
+      "learning_rate": 3.179370592462938e-07,
+      "loss": 0.8114,
       "step": 32489
     },
     {
-      "epoch": 0.8923127626266788,
+      "epoch": 0.9219636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 6.020938737616688e-07,
-      "loss": 0.8624,
+      "learning_rate": 3.177071899134987e-07,
+      "loss": 0.8157,
       "step": 32490
     },
     {
-      "epoch": 0.892340226854522,
+      "epoch": 0.9219920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 6.01789918139406e-07,
-      "loss": 0.8214,
+      "learning_rate": 3.1747740236746873e-07,
+      "loss": 0.7885,
       "step": 32491
     },
     {
-      "epoch": 0.8923676910823652,
+      "epoch": 0.9220204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 6.014860368794229e-07,
-      "loss": 0.8991,
+      "learning_rate": 3.172476966101401e-07,
+      "loss": 0.8145,
       "step": 32492
     },
     {
-      "epoch": 0.8923951553102084,
+      "epoch": 0.9220488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 6.011822299841208e-07,
-      "loss": 0.7876,
+      "learning_rate": 3.1701807264345574e-07,
+      "loss": 0.786,
       "step": 32493
     },
     {
-      "epoch": 0.8924226195380517,
+      "epoch": 0.9220771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 6.008784974559068e-07,
-      "loss": 0.8132,
+      "learning_rate": 3.167885304693552e-07,
+      "loss": 0.737,
       "step": 32494
     },
     {
-      "epoch": 0.892450083765895,
+      "epoch": 0.922105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.005748392971844e-07,
-      "loss": 0.761,
+      "learning_rate": 3.1655907008977695e-07,
+      "loss": 0.8049,
       "step": 32495
     },
     {
-      "epoch": 0.8924775479937381,
+      "epoch": 0.9221339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 6.002712555103529e-07,
-      "loss": 0.7108,
+      "learning_rate": 3.163296915066605e-07,
+      "loss": 0.8347,
       "step": 32496
     },
     {
-      "epoch": 0.8925050122215814,
+      "epoch": 0.9221623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 5.99967746097816e-07,
-      "loss": 0.7705,
+      "learning_rate": 3.161003947219421e-07,
+      "loss": 0.8065,
       "step": 32497
     },
     {
-      "epoch": 0.8925324764494246,
+      "epoch": 0.9221906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 5.99664311061976e-07,
-      "loss": 0.6769,
+      "learning_rate": 3.15871179737558e-07,
+      "loss": 0.7034,
       "step": 32498
     },
     {
-      "epoch": 0.8925599406772678,
+      "epoch": 0.9222190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 5.993609504052345e-07,
-      "loss": 0.7881,
+      "learning_rate": 3.156420465554466e-07,
+      "loss": 0.7498,
       "step": 32499
     },
     {
-      "epoch": 0.8925874049051111,
+      "epoch": 0.9222474460839954,
       "grad_norm": 0.0,
-      "learning_rate": 5.990576641299894e-07,
-      "loss": 0.7809,
+      "learning_rate": 3.1541299517754197e-07,
+      "loss": 0.732,
       "step": 32500
     },
     {
-      "epoch": 0.8926148691329543,
+      "epoch": 0.9222758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 5.987544522386435e-07,
-      "loss": 0.7856,
+      "learning_rate": 3.151840256057781e-07,
+      "loss": 0.8347,
       "step": 32501
     },
     {
-      "epoch": 0.8926423333607976,
+      "epoch": 0.9223041997729853,
       "grad_norm": 0.0,
-      "learning_rate": 5.984513147335924e-07,
-      "loss": 0.7799,
+      "learning_rate": 3.149551378420901e-07,
+      "loss": 0.8581,
       "step": 32502
     },
     {
-      "epoch": 0.8926697975886408,
+      "epoch": 0.9223325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 5.981482516172354e-07,
-      "loss": 0.8109,
+      "learning_rate": 3.1472633188841304e-07,
+      "loss": 0.8392,
       "step": 32503
     },
     {
-      "epoch": 0.892697261816484,
+      "epoch": 0.922360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 5.978452628919729e-07,
-      "loss": 0.8259,
+      "learning_rate": 3.144976077466766e-07,
+      "loss": 0.7881,
       "step": 32504
     },
     {
-      "epoch": 0.8927247260443273,
+      "epoch": 0.92238933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.975423485601994e-07,
-      "loss": 0.8656,
+      "learning_rate": 3.142689654188158e-07,
+      "loss": 0.9224,
       "step": 32505
     },
     {
-      "epoch": 0.8927521902721705,
+      "epoch": 0.9224177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 5.972395086243132e-07,
-      "loss": 0.8771,
+      "learning_rate": 3.1404040490675915e-07,
+      "loss": 0.7969,
       "step": 32506
     },
     {
-      "epoch": 0.8927796545000137,
+      "epoch": 0.9224460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.969367430867124e-07,
-      "loss": 0.8113,
+      "learning_rate": 3.138119262124384e-07,
+      "loss": 0.8223,
       "step": 32507
     },
     {
-      "epoch": 0.892807118727857,
+      "epoch": 0.9224744608399545,
       "grad_norm": 0.0,
-      "learning_rate": 5.966340519497904e-07,
-      "loss": 0.811,
+      "learning_rate": 3.1358352933778533e-07,
+      "loss": 0.8589,
       "step": 32508
     },
     {
-      "epoch": 0.8928345829557002,
+      "epoch": 0.9225028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 5.963314352159399e-07,
-      "loss": 0.7902,
+      "learning_rate": 3.1335521428472626e-07,
+      "loss": 0.9095,
       "step": 32509
     },
     {
-      "epoch": 0.8928620471835435,
+      "epoch": 0.9225312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 5.96028892887559e-07,
-      "loss": 0.8144,
+      "learning_rate": 3.1312698105519066e-07,
+      "loss": 0.7236,
       "step": 32510
     },
     {
-      "epoch": 0.8928895114113866,
+      "epoch": 0.9225595913734392,
       "grad_norm": 0.0,
-      "learning_rate": 5.957264249670391e-07,
-      "loss": 0.7734,
+      "learning_rate": 3.128988296511093e-07,
+      "loss": 0.7422,
       "step": 32511
     },
     {
-      "epoch": 0.8929169756392299,
+      "epoch": 0.9225879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 5.95424031456776e-07,
-      "loss": 0.8512,
+      "learning_rate": 3.12670760074405e-07,
+      "loss": 0.689,
       "step": 32512
     },
     {
-      "epoch": 0.8929444398670732,
+      "epoch": 0.9226163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 5.951217123591613e-07,
-      "loss": 0.8068,
+      "learning_rate": 3.1244277232700624e-07,
+      "loss": 0.6897,
       "step": 32513
     },
     {
-      "epoch": 0.8929719040949163,
+      "epoch": 0.9226447219069239,
       "grad_norm": 0.0,
-      "learning_rate": 5.94819467676585e-07,
-      "loss": 0.7298,
+      "learning_rate": 3.122148664108393e-07,
+      "loss": 0.8716,
       "step": 32514
     },
     {
-      "epoch": 0.8929993683227596,
+      "epoch": 0.9226730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 5.945172974114422e-07,
-      "loss": 0.7715,
+      "learning_rate": 3.1198704232782927e-07,
+      "loss": 0.876,
       "step": 32515
     },
     {
-      "epoch": 0.8930268325506029,
+      "epoch": 0.9227014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 5.942152015661207e-07,
-      "loss": 0.779,
+      "learning_rate": 3.1175930007989907e-07,
+      "loss": 0.8367,
       "step": 32516
     },
     {
-      "epoch": 0.893054296778446,
+      "epoch": 0.9227298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 5.93913180143012e-07,
-      "loss": 0.8223,
+      "learning_rate": 3.115316396689738e-07,
+      "loss": 0.7967,
       "step": 32517
     },
     {
-      "epoch": 0.8930817610062893,
+      "epoch": 0.9227582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 5.936112331445054e-07,
-      "loss": 0.8181,
+      "learning_rate": 3.1130406109697643e-07,
+      "loss": 0.7735,
       "step": 32518
     },
     {
-      "epoch": 0.8931092252341325,
+      "epoch": 0.9227866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 5.933093605729922e-07,
-      "loss": 0.7126,
+      "learning_rate": 3.110765643658298e-07,
+      "loss": 0.7872,
       "step": 32519
     },
     {
-      "epoch": 0.8931366894619758,
+      "epoch": 0.9228149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 5.930075624308574e-07,
-      "loss": 0.7811,
+      "learning_rate": 3.108491494774557e-07,
+      "loss": 0.7725,
       "step": 32520
     },
     {
-      "epoch": 0.893164153689819,
+      "epoch": 0.9228433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 5.927058387204909e-07,
-      "loss": 0.813,
+      "learning_rate": 3.1062181643377265e-07,
+      "loss": 0.7589,
       "step": 32521
     },
     {
-      "epoch": 0.8931916179176622,
+      "epoch": 0.9228717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 5.9240418944428e-07,
-      "loss": 0.7852,
+      "learning_rate": 3.103945652367035e-07,
+      "loss": 0.8203,
       "step": 32522
     },
     {
-      "epoch": 0.8932190821455055,
+      "epoch": 0.922900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 5.921026146046116e-07,
-      "loss": 0.8328,
+      "learning_rate": 3.101673958881679e-07,
+      "loss": 0.8128,
       "step": 32523
     },
     {
-      "epoch": 0.8932465463733487,
+      "epoch": 0.9229284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 5.918011142038715e-07,
-      "loss": 0.8585,
+      "learning_rate": 3.0994030839008206e-07,
+      "loss": 0.8262,
       "step": 32524
     },
     {
-      "epoch": 0.8932740106011919,
+      "epoch": 0.9229568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 5.914996882444468e-07,
-      "loss": 0.906,
+      "learning_rate": 3.097133027443666e-07,
+      "loss": 0.9047,
       "step": 32525
     },
     {
-      "epoch": 0.8933014748290352,
+      "epoch": 0.9229852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 5.911983367287188e-07,
-      "loss": 0.8104,
+      "learning_rate": 3.094863789529401e-07,
+      "loss": 0.8331,
       "step": 32526
     },
     {
-      "epoch": 0.8933289390568784,
+      "epoch": 0.9230136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 5.908970596590746e-07,
-      "loss": 0.877,
+      "learning_rate": 3.092595370177165e-07,
+      "loss": 0.7158,
       "step": 32527
     },
     {
-      "epoch": 0.8933564032847217,
+      "epoch": 0.9230419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 5.905958570378978e-07,
-      "loss": 0.7846,
+      "learning_rate": 3.0903277694061206e-07,
+      "loss": 0.8808,
       "step": 32528
     },
     {
-      "epoch": 0.8933838675125649,
+      "epoch": 0.9230703745743474,
       "grad_norm": 0.0,
-      "learning_rate": 5.902947288675709e-07,
-      "loss": 0.7663,
+      "learning_rate": 3.088060987235464e-07,
+      "loss": 0.7447,
       "step": 32529
     },
     {
-      "epoch": 0.8934113317404081,
+      "epoch": 0.9230987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 5.899936751504775e-07,
-      "loss": 0.7097,
+      "learning_rate": 3.08579502368429e-07,
+      "loss": 0.6674,
       "step": 32530
     },
     {
-      "epoch": 0.8934387959682514,
+      "epoch": 0.9231271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 5.896926958890004e-07,
-      "loss": 0.7722,
+      "learning_rate": 3.083529878771774e-07,
+      "loss": 0.7348,
       "step": 32531
     },
     {
-      "epoch": 0.8934662601960945,
+      "epoch": 0.9231555051078321,
       "grad_norm": 0.0,
-      "learning_rate": 5.893917910855196e-07,
-      "loss": 0.9569,
+      "learning_rate": 3.081265552517021e-07,
+      "loss": 0.8008,
       "step": 32532
     },
     {
-      "epoch": 0.8934937244239378,
+      "epoch": 0.9231838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 5.890909607424155e-07,
-      "loss": 0.7795,
+      "learning_rate": 3.0790020449391724e-07,
+      "loss": 0.7175,
       "step": 32533
     },
     {
-      "epoch": 0.8935211886517811,
+      "epoch": 0.9232122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 5.887902048620686e-07,
-      "loss": 0.8717,
+      "learning_rate": 3.0767393560573675e-07,
+      "loss": 0.8488,
       "step": 32534
     },
     {
-      "epoch": 0.8935486528796243,
+      "epoch": 0.9232406356413166,
       "grad_norm": 0.0,
-      "learning_rate": 5.884895234468601e-07,
-      "loss": 0.8004,
+      "learning_rate": 3.0744774858907033e-07,
+      "loss": 0.9014,
       "step": 32535
     },
     {
-      "epoch": 0.8935761171074675,
+      "epoch": 0.9232690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 5.881889164991672e-07,
-      "loss": 0.7891,
+      "learning_rate": 3.072216434458275e-07,
+      "loss": 0.8121,
       "step": 32536
     },
     {
-      "epoch": 0.8936035813353107,
+      "epoch": 0.9232973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 5.878883840213712e-07,
-      "loss": 0.7871,
+      "learning_rate": 3.0699562017792004e-07,
+      "loss": 0.7281,
       "step": 32537
     },
     {
-      "epoch": 0.893631045563154,
+      "epoch": 0.9233257661748013,
       "grad_norm": 0.0,
-      "learning_rate": 5.875879260158469e-07,
-      "loss": 0.7531,
+      "learning_rate": 3.067696787872554e-07,
+      "loss": 0.8247,
       "step": 32538
     },
     {
-      "epoch": 0.8936585097909973,
+      "epoch": 0.9233541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 5.872875424849734e-07,
-      "loss": 0.9435,
+      "learning_rate": 3.065438192757431e-07,
+      "loss": 0.8128,
       "step": 32539
     },
     {
-      "epoch": 0.8936859740188404,
+      "epoch": 0.9233825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 5.869872334311267e-07,
-      "loss": 0.7642,
+      "learning_rate": 3.0631804164529156e-07,
+      "loss": 0.6649,
       "step": 32540
     },
     {
-      "epoch": 0.8937134382466837,
+      "epoch": 0.923410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 5.866869988566825e-07,
-      "loss": 0.7085,
+      "learning_rate": 3.060923458978071e-07,
+      "loss": 0.767,
       "step": 32541
     },
     {
-      "epoch": 0.893740902474527,
+      "epoch": 0.9234392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 5.863868387640181e-07,
-      "loss": 0.7504,
+      "learning_rate": 3.05866732035196e-07,
+      "loss": 0.879,
       "step": 32542
     },
     {
-      "epoch": 0.8937683667023701,
+      "epoch": 0.9234676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 5.860867531555082e-07,
-      "loss": 0.8607,
+      "learning_rate": 3.056412000593656e-07,
+      "loss": 0.8319,
       "step": 32543
     },
     {
-      "epoch": 0.8937958309302134,
+      "epoch": 0.9234960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 5.857867420335251e-07,
-      "loss": 0.8198,
+      "learning_rate": 3.0541574997221767e-07,
+      "loss": 0.918,
       "step": 32544
     },
     {
-      "epoch": 0.8938232951580566,
+      "epoch": 0.9235244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 5.854868054004448e-07,
-      "loss": 0.7654,
+      "learning_rate": 3.051903817756596e-07,
+      "loss": 0.8116,
       "step": 32545
     },
     {
-      "epoch": 0.8938507593858999,
+      "epoch": 0.9235527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 5.851869432586387e-07,
-      "loss": 0.6652,
+      "learning_rate": 3.049650954715955e-07,
+      "loss": 0.8216,
       "step": 32546
     },
     {
-      "epoch": 0.8938782236137431,
+      "epoch": 0.9235811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 5.848871556104829e-07,
-      "loss": 0.848,
+      "learning_rate": 3.0473989106192593e-07,
+      "loss": 0.888,
       "step": 32547
     },
     {
-      "epoch": 0.8939056878415863,
+      "epoch": 0.9236095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 5.845874424583442e-07,
-      "loss": 0.7559,
+      "learning_rate": 3.045147685485539e-07,
+      "loss": 0.8029,
       "step": 32548
     },
     {
-      "epoch": 0.8939331520694296,
+      "epoch": 0.9236379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 5.842878038045996e-07,
-      "loss": 0.786,
+      "learning_rate": 3.042897279333812e-07,
+      "loss": 0.7372,
       "step": 32549
     },
     {
-      "epoch": 0.8939606162972727,
+      "epoch": 0.9236662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 5.83988239651615e-07,
-      "loss": 0.7104,
+      "learning_rate": 3.0406476921830964e-07,
+      "loss": 0.7448,
       "step": 32550
     },
     {
-      "epoch": 0.893988080525116,
+      "epoch": 0.923694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 5.83688750001763e-07,
-      "loss": 0.8054,
+      "learning_rate": 3.038398924052399e-07,
+      "loss": 0.8524,
       "step": 32551
     },
     {
-      "epoch": 0.8940155447529593,
+      "epoch": 0.9237230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 5.833893348574127e-07,
-      "loss": 0.7726,
+      "learning_rate": 3.036150974960694e-07,
+      "loss": 0.7723,
       "step": 32552
     },
     {
-      "epoch": 0.8940430089808025,
+      "epoch": 0.9237514188422248,
       "grad_norm": 0.0,
-      "learning_rate": 5.830899942209345e-07,
-      "loss": 0.8093,
+      "learning_rate": 3.0339038449269777e-07,
+      "loss": 0.8472,
       "step": 32553
     },
     {
-      "epoch": 0.8940704732086457,
+      "epoch": 0.9237797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 5.827907280946976e-07,
-      "loss": 0.713,
+      "learning_rate": 3.031657533970245e-07,
+      "loss": 0.7379,
       "step": 32554
     },
     {
-      "epoch": 0.894097937436489,
+      "epoch": 0.9238081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 5.824915364810657e-07,
-      "loss": 0.8015,
+      "learning_rate": 3.029412042109447e-07,
+      "loss": 0.827,
       "step": 32555
     },
     {
-      "epoch": 0.8941254016643322,
+      "epoch": 0.9238365493757095,
       "grad_norm": 0.0,
-      "learning_rate": 5.821924193824114e-07,
-      "loss": 0.778,
+      "learning_rate": 3.0271673693635704e-07,
+      "loss": 0.8123,
       "step": 32556
     },
     {
-      "epoch": 0.8941528658921755,
+      "epoch": 0.9238649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 5.818933768010971e-07,
-      "loss": 0.828,
+      "learning_rate": 3.0249235157515767e-07,
+      "loss": 0.8072,
       "step": 32557
     },
     {
-      "epoch": 0.8941803301200186,
+      "epoch": 0.9238933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 5.815944087394898e-07,
-      "loss": 0.8459,
+      "learning_rate": 3.022680481292406e-07,
+      "loss": 0.8111,
       "step": 32558
     },
     {
-      "epoch": 0.8942077943478619,
+      "epoch": 0.923921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 5.812955151999566e-07,
-      "loss": 0.8059,
+      "learning_rate": 3.02043826600501e-07,
+      "loss": 0.8141,
       "step": 32559
     },
     {
-      "epoch": 0.8942352585757052,
+      "epoch": 0.923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.809966961848635e-07,
-      "loss": 0.7972,
+      "learning_rate": 3.0181968699083517e-07,
+      "loss": 0.8371,
       "step": 32560
     },
     {
-      "epoch": 0.8942627228035483,
+      "epoch": 0.9239784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 5.806979516965706e-07,
-      "loss": 0.7944,
+      "learning_rate": 3.015956293021327e-07,
+      "loss": 0.7943,
       "step": 32561
     },
     {
-      "epoch": 0.8942901870313916,
+      "epoch": 0.9240068104426787,
       "grad_norm": 0.0,
-      "learning_rate": 5.803992817374449e-07,
-      "loss": 0.7488,
+      "learning_rate": 3.013716535362876e-07,
+      "loss": 0.8351,
       "step": 32562
     },
     {
-      "epoch": 0.8943176512592348,
+      "epoch": 0.9240351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 5.801006863098502e-07,
-      "loss": 0.746,
+      "learning_rate": 3.0114775969519175e-07,
+      "loss": 0.798,
       "step": 32563
     },
     {
-      "epoch": 0.8943451154870781,
+      "epoch": 0.9240635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 5.798021654161467e-07,
-      "loss": 0.783,
+      "learning_rate": 3.0092394778073795e-07,
+      "loss": 0.8185,
       "step": 32564
     },
     {
-      "epoch": 0.8943725797149213,
+      "epoch": 0.9240919409761634,
       "grad_norm": 0.0,
-      "learning_rate": 5.795037190586972e-07,
-      "loss": 0.7505,
+      "learning_rate": 3.0070021779481594e-07,
+      "loss": 0.767,
       "step": 32565
     },
     {
-      "epoch": 0.8944000439427645,
+      "epoch": 0.9241203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 5.79205347239864e-07,
-      "loss": 0.6786,
+      "learning_rate": 3.004765697393142e-07,
+      "loss": 0.7628,
       "step": 32566
     },
     {
-      "epoch": 0.8944275081706078,
+      "epoch": 0.9241486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 5.789070499620075e-07,
-      "loss": 0.7904,
+      "learning_rate": 3.002530036161222e-07,
+      "loss": 0.7841,
       "step": 32567
     },
     {
-      "epoch": 0.8944549723984511,
+      "epoch": 0.9241770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 5.78608827227487e-07,
-      "loss": 0.7384,
+      "learning_rate": 3.000295194271308e-07,
+      "loss": 0.8154,
       "step": 32568
     },
     {
-      "epoch": 0.8944824366262942,
+      "epoch": 0.924205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 5.783106790386627e-07,
-      "loss": 0.82,
+      "learning_rate": 2.99806117174225e-07,
+      "loss": 0.8106,
       "step": 32569
     },
     {
-      "epoch": 0.8945099008541375,
+      "epoch": 0.924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 5.78012605397894e-07,
-      "loss": 0.8524,
+      "learning_rate": 2.995827968592935e-07,
+      "loss": 0.7979,
       "step": 32570
     },
     {
-      "epoch": 0.8945373650819807,
+      "epoch": 0.9242622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 5.777146063075412e-07,
-      "loss": 0.7767,
+      "learning_rate": 2.9935955848422347e-07,
+      "loss": 0.7421,
       "step": 32571
     },
     {
-      "epoch": 0.894564829309824,
+      "epoch": 0.9242905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 5.774166817699601e-07,
-      "loss": 0.8615,
+      "learning_rate": 2.991364020508969e-07,
+      "loss": 0.7301,
       "step": 32572
     },
     {
-      "epoch": 0.8945922935376672,
+      "epoch": 0.9243189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 5.771188317875065e-07,
-      "loss": 0.8122,
+      "learning_rate": 2.989133275612022e-07,
+      "loss": 0.7732,
       "step": 32573
     },
     {
-      "epoch": 0.8946197577655104,
+      "epoch": 0.9243473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 5.768210563625387e-07,
-      "loss": 0.7632,
+      "learning_rate": 2.986903350170234e-07,
+      "loss": 0.7442,
       "step": 32574
     },
     {
-      "epoch": 0.8946472219933537,
+      "epoch": 0.9243757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 5.765233554974137e-07,
-      "loss": 0.7128,
+      "learning_rate": 2.9846742442024235e-07,
+      "loss": 0.8099,
       "step": 32575
     },
     {
-      "epoch": 0.894674686221197,
+      "epoch": 0.9244040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 5.76225729194485e-07,
-      "loss": 0.7472,
+      "learning_rate": 2.982445957727431e-07,
+      "loss": 0.9075,
       "step": 32576
     },
     {
-      "epoch": 0.8947021504490401,
+      "epoch": 0.9244324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 5.759281774561099e-07,
-      "loss": 0.7038,
+      "learning_rate": 2.9802184907640973e-07,
+      "loss": 0.7023,
       "step": 32577
     },
     {
-      "epoch": 0.8947296146768834,
+      "epoch": 0.924460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.756307002846428e-07,
-      "loss": 0.8101,
+      "learning_rate": 2.977991843331196e-07,
+      "loss": 0.7296,
       "step": 32578
     },
     {
-      "epoch": 0.8947570789047266,
+      "epoch": 0.9244892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 5.753332976824344e-07,
-      "loss": 0.8496,
+      "learning_rate": 2.9757660154475675e-07,
+      "loss": 0.8413,
       "step": 32579
     },
     {
-      "epoch": 0.8947845431325698,
+      "epoch": 0.9245175936435869,
       "grad_norm": 0.0,
-      "learning_rate": 5.750359696518415e-07,
-      "loss": 0.751,
+      "learning_rate": 2.973541007132008e-07,
+      "loss": 0.7931,
       "step": 32580
     },
     {
-      "epoch": 0.8948120073604131,
+      "epoch": 0.9245459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 5.747387161952134e-07,
-      "loss": 0.7666,
+      "learning_rate": 2.9713168184033024e-07,
+      "loss": 0.7778,
       "step": 32581
     },
     {
-      "epoch": 0.8948394715882563,
+      "epoch": 0.9245743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 5.744415373149037e-07,
-      "loss": 0.7964,
+      "learning_rate": 2.969093449280258e-07,
+      "loss": 0.9114,
       "step": 32582
     },
     {
-      "epoch": 0.8948669358160996,
+      "epoch": 0.9246027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 5.741444330132628e-07,
-      "loss": 0.7937,
+      "learning_rate": 2.9668708997816267e-07,
+      "loss": 0.7103,
       "step": 32583
     },
     {
-      "epoch": 0.8948944000439427,
+      "epoch": 0.9246311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 5.738474032926444e-07,
-      "loss": 0.713,
+      "learning_rate": 2.964649169926204e-07,
+      "loss": 0.8618,
       "step": 32584
     },
     {
-      "epoch": 0.894921864271786,
+      "epoch": 0.9246594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 5.735504481553955e-07,
-      "loss": 0.7354,
+      "learning_rate": 2.9624282597327637e-07,
+      "loss": 0.8651,
       "step": 32585
     },
     {
-      "epoch": 0.8949493284996293,
+      "epoch": 0.9246878547105561,
       "grad_norm": 0.0,
-      "learning_rate": 5.732535676038664e-07,
-      "loss": 0.8428,
+      "learning_rate": 2.960208169220047e-07,
+      "loss": 0.7784,
       "step": 32586
     },
     {
-      "epoch": 0.8949767927274724,
+      "epoch": 0.9247162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 5.729567616404086e-07,
-      "loss": 0.8151,
+      "learning_rate": 2.957988898406805e-07,
+      "loss": 0.7253,
       "step": 32587
     },
     {
-      "epoch": 0.8950042569553157,
+      "epoch": 0.924744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 5.726600302673668e-07,
-      "loss": 0.7985,
+      "learning_rate": 2.955770447311812e-07,
+      "loss": 0.7481,
       "step": 32588
     },
     {
-      "epoch": 0.895031721183159,
+      "epoch": 0.9247729852440408,
       "grad_norm": 0.0,
-      "learning_rate": 5.723633734870904e-07,
-      "loss": 0.7471,
+      "learning_rate": 2.953552815953775e-07,
+      "loss": 0.7747,
       "step": 32589
     },
     {
-      "epoch": 0.8950591854110022,
+      "epoch": 0.9248013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 5.720667913019296e-07,
-      "loss": 0.8337,
+      "learning_rate": 2.951336004351435e-07,
+      "loss": 0.7783,
       "step": 32590
     },
     {
-      "epoch": 0.8950866496388454,
+      "epoch": 0.9248297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 5.717702837142259e-07,
-      "loss": 0.802,
+      "learning_rate": 2.949120012523543e-07,
+      "loss": 0.7723,
       "step": 32591
     },
     {
-      "epoch": 0.8951141138666886,
+      "epoch": 0.9248581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 5.714738507263273e-07,
-      "loss": 0.8225,
+      "learning_rate": 2.9469048404887736e-07,
+      "loss": 0.7845,
       "step": 32592
     },
     {
-      "epoch": 0.8951415780945319,
+      "epoch": 0.9248864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 5.711774923405811e-07,
-      "loss": 0.7003,
+      "learning_rate": 2.944690488265878e-07,
+      "loss": 0.7968,
       "step": 32593
     },
     {
-      "epoch": 0.8951690423223752,
+      "epoch": 0.9249148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 5.708812085593307e-07,
-      "loss": 0.7937,
+      "learning_rate": 2.94247695587353e-07,
+      "loss": 0.7343,
       "step": 32594
     },
     {
-      "epoch": 0.8951965065502183,
+      "epoch": 0.9249432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 5.705849993849211e-07,
-      "loss": 0.8589,
+      "learning_rate": 2.940264243330448e-07,
+      "loss": 0.7959,
       "step": 32595
     },
     {
-      "epoch": 0.8952239707780616,
+      "epoch": 0.9249716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 5.702888648196969e-07,
-      "loss": 0.7469,
+      "learning_rate": 2.938052350655329e-07,
+      "loss": 0.8812,
       "step": 32596
     },
     {
-      "epoch": 0.8952514350059048,
+      "epoch": 0.925,
       "grad_norm": 0.0,
-      "learning_rate": 5.699928048659975e-07,
-      "loss": 0.7366,
+      "learning_rate": 2.9358412778668354e-07,
+      "loss": 0.8395,
       "step": 32597
     },
     {
-      "epoch": 0.895278899233748,
+      "epoch": 0.9250283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 5.696968195261687e-07,
-      "loss": 0.8203,
+      "learning_rate": 2.933631024983652e-07,
+      "loss": 0.8517,
       "step": 32598
     },
     {
-      "epoch": 0.8953063634615913,
+      "epoch": 0.9250567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 5.694009088025509e-07,
-      "loss": 0.7541,
+      "learning_rate": 2.9314215920244525e-07,
+      "loss": 0.9117,
       "step": 32599
     },
     {
-      "epoch": 0.8953338276894345,
+      "epoch": 0.9250851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 5.691050726974867e-07,
-      "loss": 0.8274,
+      "learning_rate": 2.9292129790079005e-07,
+      "loss": 0.7578,
       "step": 32600
     },
     {
-      "epoch": 0.8953612919172778,
+      "epoch": 0.9251135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 5.688093112133164e-07,
-      "loss": 0.8536,
+      "learning_rate": 2.9270051859526473e-07,
+      "loss": 0.752,
       "step": 32601
     },
     {
-      "epoch": 0.895388756145121,
+      "epoch": 0.9251418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 5.685136243523814e-07,
-      "loss": 0.7681,
+      "learning_rate": 2.924798212877356e-07,
+      "loss": 0.7635,
       "step": 32602
     },
     {
-      "epoch": 0.8954162203729642,
+      "epoch": 0.9251702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 5.682180121170177e-07,
-      "loss": 0.887,
+      "learning_rate": 2.922592059800644e-07,
+      "loss": 0.8456,
       "step": 32603
     },
     {
-      "epoch": 0.8954436846008075,
+      "epoch": 0.9251986379114643,
       "grad_norm": 0.0,
-      "learning_rate": 5.679224745095691e-07,
-      "loss": 0.8281,
+      "learning_rate": 2.9203867267411537e-07,
+      "loss": 0.869,
       "step": 32604
     },
     {
-      "epoch": 0.8954711488286506,
+      "epoch": 0.9252270147559591,
       "grad_norm": 0.0,
-      "learning_rate": 5.676270115323701e-07,
-      "loss": 0.7007,
+      "learning_rate": 2.9181822137175464e-07,
+      "loss": 0.9092,
       "step": 32605
     },
     {
-      "epoch": 0.8954986130564939,
+      "epoch": 0.925255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 5.67331623187759e-07,
-      "loss": 0.7521,
+      "learning_rate": 2.915978520748397e-07,
+      "loss": 0.7562,
       "step": 32606
     },
     {
-      "epoch": 0.8955260772843372,
+      "epoch": 0.925283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 5.670363094780751e-07,
-      "loss": 0.8551,
+      "learning_rate": 2.913775647852346e-07,
+      "loss": 0.7698,
       "step": 32607
     },
     {
-      "epoch": 0.8955535415121804,
+      "epoch": 0.9253121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 5.667410704056542e-07,
-      "loss": 0.762,
+      "learning_rate": 2.9115735950480117e-07,
+      "loss": 0.8782,
       "step": 32608
     },
     {
-      "epoch": 0.8955810057400236,
+      "epoch": 0.9253405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 5.664459059728311e-07,
-      "loss": 0.7668,
+      "learning_rate": 2.909372362353946e-07,
+      "loss": 0.7333,
       "step": 32609
     },
     {
-      "epoch": 0.8956084699678668,
+      "epoch": 0.9253688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 5.661508161819429e-07,
-      "loss": 0.8386,
+      "learning_rate": 2.9071719497888007e-07,
+      "loss": 0.8317,
       "step": 32610
     },
     {
-      "epoch": 0.8956359341957101,
+      "epoch": 0.9253972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 5.658558010353244e-07,
-      "loss": 0.9005,
+      "learning_rate": 2.9049723573711384e-07,
+      "loss": 0.8407,
       "step": 32611
     },
     {
-      "epoch": 0.8956633984235534,
+      "epoch": 0.9254256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 5.65560860535308e-07,
-      "loss": 0.8304,
+      "learning_rate": 2.9027735851195336e-07,
+      "loss": 0.7103,
       "step": 32612
     },
     {
-      "epoch": 0.8956908626513965,
+      "epoch": 0.9254540295119182,
       "grad_norm": 0.0,
-      "learning_rate": 5.652659946842287e-07,
-      "loss": 0.8096,
+      "learning_rate": 2.900575633052571e-07,
+      "loss": 0.7569,
       "step": 32613
     },
     {
-      "epoch": 0.8957183268792398,
+      "epoch": 0.9254824063564132,
       "grad_norm": 0.0,
-      "learning_rate": 5.649712034844212e-07,
-      "loss": 0.7736,
+      "learning_rate": 2.898378501188803e-07,
+      "loss": 0.7405,
       "step": 32614
     },
     {
-      "epoch": 0.8957457911070831,
+      "epoch": 0.9255107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 5.646764869382149e-07,
-      "loss": 0.833,
+      "learning_rate": 2.896182189546803e-07,
+      "loss": 0.8339,
       "step": 32615
     },
     {
-      "epoch": 0.8957732553349262,
+      "epoch": 0.9255391600454029,
       "grad_norm": 0.0,
-      "learning_rate": 5.643818450479421e-07,
-      "loss": 0.8221,
+      "learning_rate": 2.8939866981451236e-07,
+      "loss": 0.8321,
       "step": 32616
     },
     {
-      "epoch": 0.8958007195627695,
+      "epoch": 0.9255675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 5.640872778159357e-07,
-      "loss": 0.7613,
+      "learning_rate": 2.8917920270022934e-07,
+      "loss": 0.6844,
       "step": 32617
     },
     {
-      "epoch": 0.8958281837906127,
+      "epoch": 0.9255959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 5.637927852445269e-07,
-      "loss": 0.8222,
+      "learning_rate": 2.8895981761368653e-07,
+      "loss": 0.8602,
       "step": 32618
     },
     {
-      "epoch": 0.895855648018456,
+      "epoch": 0.9256242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 5.634983673360451e-07,
-      "loss": 0.8096,
+      "learning_rate": 2.887405145567368e-07,
+      "loss": 0.8516,
       "step": 32619
     },
     {
-      "epoch": 0.8958831122462992,
+      "epoch": 0.9256526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 5.632040240928194e-07,
-      "loss": 0.7009,
+      "learning_rate": 2.8852129353123204e-07,
+      "loss": 0.8104,
       "step": 32620
     },
     {
-      "epoch": 0.8959105764741424,
+      "epoch": 0.9256810442678775,
       "grad_norm": 0.0,
-      "learning_rate": 5.629097555171781e-07,
-      "loss": 0.8073,
+      "learning_rate": 2.883021545390241e-07,
+      "loss": 0.8174,
       "step": 32621
     },
     {
-      "epoch": 0.8959380407019857,
+      "epoch": 0.9257094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 5.626155616114504e-07,
-      "loss": 0.8384,
+      "learning_rate": 2.8808309758196593e-07,
+      "loss": 0.8378,
       "step": 32622
     },
     {
-      "epoch": 0.8959655049298288,
+      "epoch": 0.9257377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 5.623214423779644e-07,
-      "loss": 0.8264,
+      "learning_rate": 2.87864122661905e-07,
+      "loss": 0.7373,
       "step": 32623
     },
     {
-      "epoch": 0.8959929691576721,
+      "epoch": 0.9257661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 5.620273978190461e-07,
-      "loss": 0.7918,
+      "learning_rate": 2.8764522978069197e-07,
+      "loss": 0.8183,
       "step": 32624
     },
     {
-      "epoch": 0.8960204333855154,
+      "epoch": 0.925794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 5.617334279370246e-07,
-      "loss": 0.8134,
+      "learning_rate": 2.874264189401776e-07,
+      "loss": 0.8076,
       "step": 32625
     },
     {
-      "epoch": 0.8960478976133586,
+      "epoch": 0.9258229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 5.614395327342237e-07,
-      "loss": 0.7425,
+      "learning_rate": 2.872076901422083e-07,
+      "loss": 0.7662,
       "step": 32626
     },
     {
-      "epoch": 0.8960753618412018,
+      "epoch": 0.9258513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 5.611457122129683e-07,
-      "loss": 0.8829,
+      "learning_rate": 2.8698904338863134e-07,
+      "loss": 0.7355,
       "step": 32627
     },
     {
-      "epoch": 0.8961028260690451,
+      "epoch": 0.9258796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 5.608519663755863e-07,
-      "loss": 0.7645,
+      "learning_rate": 2.8677047868129635e-07,
+      "loss": 0.6918,
       "step": 32628
     },
     {
-      "epoch": 0.8961302902968883,
+      "epoch": 0.9259080590238365,
       "grad_norm": 0.0,
-      "learning_rate": 5.605582952243982e-07,
-      "loss": 0.7515,
+      "learning_rate": 2.8655199602204755e-07,
+      "loss": 0.8654,
       "step": 32629
     },
     {
-      "epoch": 0.8961577545247316,
+      "epoch": 0.9259364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 5.602646987617288e-07,
-      "loss": 0.7961,
+      "learning_rate": 2.8633359541273e-07,
+      "loss": 0.7259,
       "step": 32630
     },
     {
-      "epoch": 0.8961852187525747,
+      "epoch": 0.9259648127128264,
       "grad_norm": 0.0,
-      "learning_rate": 5.599711769899041e-07,
-      "loss": 0.7898,
+      "learning_rate": 2.8611527685519e-07,
+      "loss": 0.7733,
       "step": 32631
     },
     {
-      "epoch": 0.896212682980418,
+      "epoch": 0.9259931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 5.596777299112422e-07,
-      "loss": 0.8446,
+      "learning_rate": 2.8589704035126953e-07,
+      "loss": 0.8652,
       "step": 32632
     },
     {
-      "epoch": 0.8962401472082613,
+      "epoch": 0.9260215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 5.593843575280678e-07,
-      "loss": 0.7798,
+      "learning_rate": 2.8567888590281476e-07,
+      "loss": 0.8633,
       "step": 32633
     },
     {
-      "epoch": 0.8962676114361044,
+      "epoch": 0.926049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 5.590910598427002e-07,
-      "loss": 0.6841,
+      "learning_rate": 2.8546081351166656e-07,
+      "loss": 0.6837,
       "step": 32634
     },
     {
-      "epoch": 0.8962950756639477,
+      "epoch": 0.9260783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 5.587978368574631e-07,
-      "loss": 0.7908,
+      "learning_rate": 2.8524282317966776e-07,
+      "loss": 0.8886,
       "step": 32635
     },
     {
-      "epoch": 0.8963225398917909,
+      "epoch": 0.9261066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 5.585046885746726e-07,
-      "loss": 0.7984,
+      "learning_rate": 2.8502491490865923e-07,
+      "loss": 0.8162,
       "step": 32636
     },
     {
-      "epoch": 0.8963500041196342,
+      "epoch": 0.9261350737797956,
       "grad_norm": 0.0,
-      "learning_rate": 5.582116149966533e-07,
-      "loss": 0.8238,
+      "learning_rate": 2.848070887004828e-07,
+      "loss": 0.912,
       "step": 32637
     },
     {
-      "epoch": 0.8963774683474774,
+      "epoch": 0.9261634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 5.579186161257178e-07,
-      "loss": 0.8015,
+      "learning_rate": 2.8458934455697693e-07,
+      "loss": 0.7961,
       "step": 32638
     },
     {
-      "epoch": 0.8964049325753206,
+      "epoch": 0.9261918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 5.576256919641887e-07,
-      "loss": 0.8353,
+      "learning_rate": 2.8437168247998245e-07,
+      "loss": 0.78,
       "step": 32639
     },
     {
-      "epoch": 0.8964323968031639,
+      "epoch": 0.9262202043132803,
       "grad_norm": 0.0,
-      "learning_rate": 5.573328425143832e-07,
-      "loss": 0.8296,
+      "learning_rate": 2.8415410247133457e-07,
+      "loss": 0.8292,
       "step": 32640
     },
     {
-      "epoch": 0.8964598610310072,
+      "epoch": 0.9262485811577753,
       "grad_norm": 0.0,
-      "learning_rate": 5.570400677786169e-07,
-      "loss": 0.7761,
+      "learning_rate": 2.839366045328751e-07,
+      "loss": 0.831,
       "step": 32641
     },
     {
-      "epoch": 0.8964873252588503,
+      "epoch": 0.9262769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 5.567473677592083e-07,
-      "loss": 0.7943,
+      "learning_rate": 2.837191886664414e-07,
+      "loss": 0.7173,
       "step": 32642
     },
     {
-      "epoch": 0.8965147894866936,
+      "epoch": 0.926305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 5.564547424584732e-07,
-      "loss": 0.8309,
+      "learning_rate": 2.835018548738666e-07,
+      "loss": 0.8395,
       "step": 32643
     },
     {
-      "epoch": 0.8965422537145368,
+      "epoch": 0.9263337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 5.561621918787275e-07,
-      "loss": 0.7904,
+      "learning_rate": 2.8328460315698914e-07,
+      "loss": 0.8272,
       "step": 32644
     },
     {
-      "epoch": 0.89656971794238,
+      "epoch": 0.9263620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 5.558697160222826e-07,
-      "loss": 0.7793,
+      "learning_rate": 2.830674335176442e-07,
+      "loss": 0.8817,
       "step": 32645
     },
     {
-      "epoch": 0.8965971821702233,
+      "epoch": 0.9263904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 5.555773148914545e-07,
-      "loss": 0.6668,
+      "learning_rate": 2.828503459576648e-07,
+      "loss": 0.7839,
       "step": 32646
     },
     {
-      "epoch": 0.8966246463980665,
+      "epoch": 0.9264188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 5.552849884885569e-07,
-      "loss": 0.8752,
+      "learning_rate": 2.8263334047888504e-07,
+      "loss": 0.8367,
       "step": 32647
     },
     {
-      "epoch": 0.8966521106259098,
+      "epoch": 0.9264472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 5.549927368159036e-07,
-      "loss": 0.7932,
+      "learning_rate": 2.824164170831389e-07,
+      "loss": 0.7653,
       "step": 32648
     },
     {
-      "epoch": 0.8966795748537529,
+      "epoch": 0.9264755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 5.547005598758071e-07,
-      "loss": 0.8963,
+      "learning_rate": 2.821995757722573e-07,
+      "loss": 0.8691,
       "step": 32649
     },
     {
-      "epoch": 0.8967070390815962,
+      "epoch": 0.9265039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 5.544084576705777e-07,
-      "loss": 0.7818,
+      "learning_rate": 2.8198281654807313e-07,
+      "loss": 0.8836,
       "step": 32650
     },
     {
-      "epoch": 0.8967345033094395,
+      "epoch": 0.9265323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 5.541164302025281e-07,
-      "loss": 0.9111,
+      "learning_rate": 2.8176613941241826e-07,
+      "loss": 0.8416,
       "step": 32651
     },
     {
-      "epoch": 0.8967619675372827,
+      "epoch": 0.9265607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 5.538244774739687e-07,
-      "loss": 0.8016,
+      "learning_rate": 2.8154954436712014e-07,
+      "loss": 0.8264,
       "step": 32652
     },
     {
-      "epoch": 0.8967894317651259,
+      "epoch": 0.926589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 5.535325994872076e-07,
-      "loss": 0.7991,
+      "learning_rate": 2.8133303141400946e-07,
+      "loss": 0.7162,
       "step": 32653
     },
     {
-      "epoch": 0.8968168959929692,
+      "epoch": 0.9266174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 5.532407962445563e-07,
-      "loss": 0.8333,
+      "learning_rate": 2.811166005549171e-07,
+      "loss": 0.7685,
       "step": 32654
     },
     {
-      "epoch": 0.8968443602208124,
+      "epoch": 0.9266458569807038,
       "grad_norm": 0.0,
-      "learning_rate": 5.529490677483252e-07,
-      "loss": 0.8145,
+      "learning_rate": 2.809002517916681e-07,
+      "loss": 0.7775,
       "step": 32655
     },
     {
-      "epoch": 0.8968718244486557,
+      "epoch": 0.9266742338251986,
       "grad_norm": 0.0,
-      "learning_rate": 5.526574140008178e-07,
-      "loss": 0.8145,
+      "learning_rate": 2.806839851260923e-07,
+      "loss": 0.8608,
       "step": 32656
     },
     {
-      "epoch": 0.8968992886764988,
+      "epoch": 0.9267026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 5.523658350043448e-07,
-      "loss": 0.8493,
+      "learning_rate": 2.8046780056001587e-07,
+      "loss": 0.7071,
       "step": 32657
     },
     {
-      "epoch": 0.8969267529043421,
+      "epoch": 0.9267309875141885,
       "grad_norm": 0.0,
-      "learning_rate": 5.520743307612142e-07,
-      "loss": 0.7946,
+      "learning_rate": 2.802516980952652e-07,
+      "loss": 0.8698,
       "step": 32658
     },
     {
-      "epoch": 0.8969542171321854,
+      "epoch": 0.9267593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 5.517829012737308e-07,
-      "loss": 0.8548,
+      "learning_rate": 2.800356777336655e-07,
+      "loss": 0.7666,
       "step": 32659
     },
     {
-      "epoch": 0.8969816813600285,
+      "epoch": 0.9267877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 5.514915465442005e-07,
-      "loss": 0.8581,
+      "learning_rate": 2.798197394770408e-07,
+      "loss": 0.7099,
       "step": 32660
     },
     {
-      "epoch": 0.8970091455878718,
+      "epoch": 0.926816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 5.512002665749305e-07,
-      "loss": 0.8468,
+      "learning_rate": 2.796038833272152e-07,
+      "loss": 0.7816,
       "step": 32661
     },
     {
-      "epoch": 0.897036609815715,
+      "epoch": 0.926844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 5.509090613682233e-07,
-      "loss": 0.8189,
+      "learning_rate": 2.79388109286014e-07,
+      "loss": 0.8253,
       "step": 32662
     },
     {
-      "epoch": 0.8970640740435583,
+      "epoch": 0.9268728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 5.506179309263826e-07,
-      "loss": 0.7721,
+      "learning_rate": 2.791724173552568e-07,
+      "loss": 0.8337,
       "step": 32663
     },
     {
-      "epoch": 0.8970915382714015,
+      "epoch": 0.9269012485811577,
       "grad_norm": 0.0,
-      "learning_rate": 5.503268752517144e-07,
-      "loss": 0.7419,
+      "learning_rate": 2.789568075367677e-07,
+      "loss": 0.7489,
       "step": 32664
     },
     {
-      "epoch": 0.8971190024992447,
+      "epoch": 0.9269296254256527,
       "grad_norm": 0.0,
-      "learning_rate": 5.5003589434652e-07,
-      "loss": 0.8998,
+      "learning_rate": 2.7874127983236853e-07,
+      "loss": 0.8474,
       "step": 32665
     },
     {
-      "epoch": 0.897146466727088,
+      "epoch": 0.9269580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 5.497449882131011e-07,
-      "loss": 0.8309,
+      "learning_rate": 2.785258342438779e-07,
+      "loss": 0.7237,
       "step": 32666
     },
     {
-      "epoch": 0.8971739309549313,
+      "epoch": 0.9269863791146424,
       "grad_norm": 0.0,
-      "learning_rate": 5.494541568537637e-07,
-      "loss": 0.8277,
+      "learning_rate": 2.783104707731166e-07,
+      "loss": 0.7724,
       "step": 32667
     },
     {
-      "epoch": 0.8972013951827744,
+      "epoch": 0.9270147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 5.491634002708047e-07,
-      "loss": 0.7873,
+      "learning_rate": 2.7809518942190415e-07,
+      "loss": 0.8243,
       "step": 32668
     },
     {
-      "epoch": 0.8972288594106177,
+      "epoch": 0.9270431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 5.488727184665243e-07,
-      "loss": 0.9634,
+      "learning_rate": 2.778799901920581e-07,
+      "loss": 0.7503,
       "step": 32669
     },
     {
-      "epoch": 0.8972563236384609,
+      "epoch": 0.9270715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 5.485821114432243e-07,
-      "loss": 0.7233,
+      "learning_rate": 2.77664873085397e-07,
+      "loss": 0.7932,
       "step": 32670
     },
     {
-      "epoch": 0.8972837878663041,
+      "epoch": 0.927099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 5.48291579203204e-07,
-      "loss": 0.8737,
+      "learning_rate": 2.7744983810373716e-07,
+      "loss": 0.7433,
       "step": 32671
     },
     {
-      "epoch": 0.8973112520941474,
+      "epoch": 0.927128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 5.480011217487613e-07,
-      "loss": 0.8602,
+      "learning_rate": 2.7723488524889596e-07,
+      "loss": 0.7442,
       "step": 32672
     },
     {
-      "epoch": 0.8973387163219906,
+      "epoch": 0.9271566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 5.477107390821968e-07,
-      "loss": 0.8322,
+      "learning_rate": 2.7702001452268976e-07,
+      "loss": 0.7287,
       "step": 32673
     },
     {
-      "epoch": 0.8973661805498339,
+      "epoch": 0.9271850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 5.474204312058051e-07,
-      "loss": 0.8197,
+      "learning_rate": 2.768052259269316e-07,
+      "loss": 0.8224,
       "step": 32674
     },
     {
-      "epoch": 0.897393644777677,
+      "epoch": 0.9272133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 5.471301981218857e-07,
-      "loss": 0.8831,
+      "learning_rate": 2.765905194634366e-07,
+      "loss": 0.9004,
       "step": 32675
     },
     {
-      "epoch": 0.8974211090055203,
+      "epoch": 0.9272417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 5.468400398327334e-07,
-      "loss": 0.8337,
+      "learning_rate": 2.7637589513401896e-07,
+      "loss": 0.7845,
       "step": 32676
     },
     {
-      "epoch": 0.8974485732333636,
+      "epoch": 0.9272701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 5.465499563406429e-07,
-      "loss": 0.8157,
+      "learning_rate": 2.7616135294049164e-07,
+      "loss": 0.7213,
       "step": 32677
     },
     {
-      "epoch": 0.8974760374612067,
+      "epoch": 0.9272985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 5.462599476479135e-07,
-      "loss": 0.7862,
+      "learning_rate": 2.759468928846654e-07,
+      "loss": 0.7766,
       "step": 32678
     },
     {
-      "epoch": 0.89750350168905,
+      "epoch": 0.9273269012485812,
       "grad_norm": 0.0,
-      "learning_rate": 5.459700137568358e-07,
-      "loss": 0.7659,
+      "learning_rate": 2.757325149683543e-07,
+      "loss": 0.7572,
       "step": 32679
     },
     {
-      "epoch": 0.8975309659168933,
+      "epoch": 0.927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 5.456801546697055e-07,
-      "loss": 0.8219,
+      "learning_rate": 2.7551821919336695e-07,
+      "loss": 0.8024,
       "step": 32680
     },
     {
-      "epoch": 0.8975584301447365,
+      "epoch": 0.9273836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 5.453903703888164e-07,
-      "loss": 0.7931,
+      "learning_rate": 2.7530400556151413e-07,
+      "loss": 0.8255,
       "step": 32681
     },
     {
-      "epoch": 0.8975858943725797,
+      "epoch": 0.9274120317820659,
       "grad_norm": 0.0,
-      "learning_rate": 5.451006609164611e-07,
-      "loss": 0.7108,
+      "learning_rate": 2.7508987407460664e-07,
+      "loss": 0.816,
       "step": 32682
     },
     {
-      "epoch": 0.8976133586004229,
+      "epoch": 0.9274404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 5.448110262549322e-07,
-      "loss": 0.8352,
+      "learning_rate": 2.7487582473445076e-07,
+      "loss": 0.8224,
       "step": 32683
     },
     {
-      "epoch": 0.8976408228282662,
+      "epoch": 0.9274687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 5.445214664065223e-07,
-      "loss": 0.7673,
+      "learning_rate": 2.7466185754285723e-07,
+      "loss": 0.8549,
       "step": 32684
     },
     {
-      "epoch": 0.8976682870561095,
+      "epoch": 0.9274971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 5.442319813735197e-07,
-      "loss": 0.7,
+      "learning_rate": 2.7444797250163136e-07,
+      "loss": 0.8221,
       "step": 32685
     },
     {
-      "epoch": 0.8976957512839526,
+      "epoch": 0.9275255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 5.439425711582158e-07,
-      "loss": 0.7724,
+      "learning_rate": 2.7423416961257944e-07,
+      "loss": 0.8583,
       "step": 32686
     },
     {
-      "epoch": 0.8977232155117959,
+      "epoch": 0.9275539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 5.436532357629009e-07,
-      "loss": 0.7767,
+      "learning_rate": 2.740204488775111e-07,
+      "loss": 0.8354,
       "step": 32687
     },
     {
-      "epoch": 0.8977506797396391,
+      "epoch": 0.9275822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 5.433639751898656e-07,
-      "loss": 0.7393,
+      "learning_rate": 2.738068102982283e-07,
+      "loss": 0.7885,
       "step": 32688
     },
     {
-      "epoch": 0.8977781439674823,
+      "epoch": 0.9276106696935301,
       "grad_norm": 0.0,
-      "learning_rate": 5.430747894413979e-07,
-      "loss": 0.8149,
+      "learning_rate": 2.735932538765362e-07,
+      "loss": 0.8675,
       "step": 32689
     },
     {
-      "epoch": 0.8978056081953256,
+      "epoch": 0.927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 5.427856785197861e-07,
-      "loss": 0.8503,
+      "learning_rate": 2.733797796142401e-07,
+      "loss": 0.8766,
       "step": 32690
     },
     {
-      "epoch": 0.8978330724231688,
+      "epoch": 0.9276674233825198,
       "grad_norm": 0.0,
-      "learning_rate": 5.424966424273159e-07,
-      "loss": 0.7764,
+      "learning_rate": 2.7316638751314074e-07,
+      "loss": 1.0179,
       "step": 32691
     },
     {
-      "epoch": 0.8978605366510121,
+      "epoch": 0.9276958002270148,
       "grad_norm": 0.0,
-      "learning_rate": 5.42207681166278e-07,
-      "loss": 0.7778,
+      "learning_rate": 2.729530775750433e-07,
+      "loss": 0.7942,
       "step": 32692
     },
     {
-      "epoch": 0.8978880008788553,
+      "epoch": 0.9277241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 5.419187947389548e-07,
-      "loss": 0.8562,
+      "learning_rate": 2.7273984980174863e-07,
+      "loss": 0.807,
       "step": 32693
     },
     {
-      "epoch": 0.8979154651066985,
+      "epoch": 0.9277525539160045,
       "grad_norm": 0.0,
-      "learning_rate": 5.416299831476346e-07,
-      "loss": 0.8983,
+      "learning_rate": 2.725267041950574e-07,
+      "loss": 0.7942,
       "step": 32694
     },
     {
-      "epoch": 0.8979429293345418,
+      "epoch": 0.9277809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 5.413412463946e-07,
-      "loss": 0.7907,
+      "learning_rate": 2.7231364075676947e-07,
+      "loss": 0.8559,
       "step": 32695
     },
     {
-      "epoch": 0.897970393562385,
+      "epoch": 0.9278093076049944,
       "grad_norm": 0.0,
-      "learning_rate": 5.410525844821402e-07,
-      "loss": 0.8978,
+      "learning_rate": 2.721006594886877e-07,
+      "loss": 0.881,
       "step": 32696
     },
     {
-      "epoch": 0.8979978577902282,
+      "epoch": 0.9278376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 5.407639974125345e-07,
-      "loss": 0.7644,
+      "learning_rate": 2.7188776039260735e-07,
+      "loss": 0.829,
       "step": 32697
     },
     {
-      "epoch": 0.8980253220180715,
+      "epoch": 0.9278660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 5.404754851880689e-07,
-      "loss": 0.814,
+      "learning_rate": 2.716749434703281e-07,
+      "loss": 0.8845,
       "step": 32698
     },
     {
-      "epoch": 0.8980527862459147,
+      "epoch": 0.927894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.401870478110261e-07,
-      "loss": 0.7165,
+      "learning_rate": 2.714622087236485e-07,
+      "loss": 0.8506,
       "step": 32699
     },
     {
-      "epoch": 0.898080250473758,
+      "epoch": 0.9279228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 5.398986852836863e-07,
-      "loss": 0.7573,
+      "learning_rate": 2.712495561543649e-07,
+      "loss": 0.7802,
       "step": 32700
     },
     {
-      "epoch": 0.8981077147016011,
+      "epoch": 0.9279511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 5.396103976083334e-07,
-      "loss": 0.7755,
+      "learning_rate": 2.7103698576427364e-07,
+      "loss": 0.7894,
       "step": 32701
     },
     {
-      "epoch": 0.8981351789294444,
+      "epoch": 0.9279795686719636,
       "grad_norm": 0.0,
-      "learning_rate": 5.393221847872487e-07,
-      "loss": 0.8122,
+      "learning_rate": 2.7082449755516995e-07,
+      "loss": 0.8023,
       "step": 32702
     },
     {
-      "epoch": 0.8981626431572877,
+      "epoch": 0.9280079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 5.390340468227095e-07,
-      "loss": 0.8157,
+      "learning_rate": 2.70612091528849e-07,
+      "loss": 0.7454,
       "step": 32703
     },
     {
-      "epoch": 0.8981901073851308,
+      "epoch": 0.9280363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 5.387459837169984e-07,
-      "loss": 0.7913,
+      "learning_rate": 2.703997676871062e-07,
+      "loss": 0.8301,
       "step": 32704
     },
     {
-      "epoch": 0.8982175716129741,
+      "epoch": 0.9280646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 5.384579954723934e-07,
-      "loss": 0.86,
+      "learning_rate": 2.701875260317333e-07,
+      "loss": 0.8313,
       "step": 32705
     },
     {
-      "epoch": 0.8982450358408174,
+      "epoch": 0.9280930760499433,
       "grad_norm": 0.0,
-      "learning_rate": 5.38170082091175e-07,
-      "loss": 0.8156,
+      "learning_rate": 2.699753665645233e-07,
+      "loss": 0.7492,
       "step": 32706
     },
     {
-      "epoch": 0.8982725000686606,
+      "epoch": 0.9281214528944381,
       "grad_norm": 0.0,
-      "learning_rate": 5.378822435756203e-07,
-      "loss": 0.869,
+      "learning_rate": 2.6976328928726923e-07,
+      "loss": 0.851,
       "step": 32707
     },
     {
-      "epoch": 0.8982999642965038,
+      "epoch": 0.928149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 5.375944799280075e-07,
-      "loss": 0.8005,
+      "learning_rate": 2.6955129420176193e-07,
+      "loss": 0.7371,
       "step": 32708
     },
     {
-      "epoch": 0.898327428524347,
+      "epoch": 0.928178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 5.373067911506103e-07,
-      "loss": 0.8797,
+      "learning_rate": 2.693393813097922e-07,
+      "loss": 0.8242,
       "step": 32709
     },
     {
-      "epoch": 0.8983548927521903,
+      "epoch": 0.9282065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 5.370191772457079e-07,
-      "loss": 0.8253,
+      "learning_rate": 2.691275506131508e-07,
+      "loss": 0.7856,
       "step": 32710
     },
     {
-      "epoch": 0.8983823569800335,
+      "epoch": 0.9282349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 5.367316382155752e-07,
-      "loss": 0.8403,
+      "learning_rate": 2.689158021136251e-07,
+      "loss": 0.745,
       "step": 32711
     },
     {
-      "epoch": 0.8984098212078767,
+      "epoch": 0.9282633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 5.364441740624871e-07,
-      "loss": 0.8438,
+      "learning_rate": 2.6870413581300603e-07,
+      "loss": 0.8713,
       "step": 32712
     },
     {
-      "epoch": 0.89843728543572,
+      "epoch": 0.9282917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 5.361567847887183e-07,
-      "loss": 0.8479,
+      "learning_rate": 2.6849255171308097e-07,
+      "loss": 0.7955,
       "step": 32713
     },
     {
-      "epoch": 0.8984647496635632,
+      "epoch": 0.9283200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 5.358694703965451e-07,
-      "loss": 0.9311,
+      "learning_rate": 2.682810498156363e-07,
+      "loss": 0.8867,
       "step": 32714
     },
     {
-      "epoch": 0.8984922138914064,
+      "epoch": 0.9283484676503972,
       "grad_norm": 0.0,
-      "learning_rate": 5.355822308882374e-07,
-      "loss": 0.8568,
+      "learning_rate": 2.6806963012245833e-07,
+      "loss": 0.7998,
       "step": 32715
     },
     {
-      "epoch": 0.8985196781192497,
+      "epoch": 0.9283768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 5.352950662660705e-07,
-      "loss": 0.7486,
+      "learning_rate": 2.678582926353357e-07,
+      "loss": 0.8315,
       "step": 32716
     },
     {
-      "epoch": 0.8985471423470929,
+      "epoch": 0.9284052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 5.350079765323135e-07,
-      "loss": 0.8316,
+      "learning_rate": 2.67647037356048e-07,
+      "loss": 0.8448,
       "step": 32717
     },
     {
-      "epoch": 0.8985746065749362,
+      "epoch": 0.9284335981838819,
       "grad_norm": 0.0,
-      "learning_rate": 5.347209616892402e-07,
-      "loss": 0.7633,
+      "learning_rate": 2.67435864286385e-07,
+      "loss": 0.767,
       "step": 32718
     },
     {
-      "epoch": 0.8986020708027794,
+      "epoch": 0.9284619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 5.344340217391208e-07,
-      "loss": 0.7321,
+      "learning_rate": 2.6722477342813083e-07,
+      "loss": 0.8321,
       "step": 32719
     },
     {
-      "epoch": 0.8986295350306226,
+      "epoch": 0.9284903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 5.341471566842282e-07,
-      "loss": 0.7108,
+      "learning_rate": 2.6701376478306396e-07,
+      "loss": 0.7788,
       "step": 32720
     },
     {
-      "epoch": 0.8986569992584659,
+      "epoch": 0.9285187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 5.338603665268282e-07,
-      "loss": 0.7902,
+      "learning_rate": 2.668028383529719e-07,
+      "loss": 0.7422,
       "step": 32721
     },
     {
-      "epoch": 0.898684463486309,
+      "epoch": 0.9285471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 5.335736512691914e-07,
-      "loss": 0.7986,
+      "learning_rate": 2.665919941396311e-07,
+      "loss": 0.8143,
       "step": 32722
     },
     {
-      "epoch": 0.8987119277141523,
+      "epoch": 0.9285754824063565,
       "grad_norm": 0.0,
-      "learning_rate": 5.33287010913588e-07,
-      "loss": 0.8459,
+      "learning_rate": 2.663812321448267e-07,
+      "loss": 0.7961,
       "step": 32723
     },
     {
-      "epoch": 0.8987393919419956,
+      "epoch": 0.9286038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 5.33000445462284e-07,
-      "loss": 0.9163,
+      "learning_rate": 2.661705523703373e-07,
+      "loss": 0.7525,
       "step": 32724
     },
     {
-      "epoch": 0.8987668561698388,
+      "epoch": 0.9286322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 5.327139549175475e-07,
-      "loss": 0.8528,
+      "learning_rate": 2.659599548179426e-07,
+      "loss": 0.7842,
       "step": 32725
     },
     {
-      "epoch": 0.898794320397682,
+      "epoch": 0.928660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 5.32427539281647e-07,
-      "loss": 0.7119,
+      "learning_rate": 2.6574943948942224e-07,
+      "loss": 0.7983,
       "step": 32726
     },
     {
-      "epoch": 0.8988217846255252,
+      "epoch": 0.928688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 5.321411985568459e-07,
-      "loss": 0.8032,
+      "learning_rate": 2.6553900638655373e-07,
+      "loss": 0.7085,
       "step": 32727
     },
     {
-      "epoch": 0.8988492488533685,
+      "epoch": 0.9287173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 5.318549327454115e-07,
-      "loss": 0.7811,
+      "learning_rate": 2.6532865551111456e-07,
+      "loss": 0.8469,
       "step": 32728
     },
     {
-      "epoch": 0.8988767130812118,
+      "epoch": 0.9287457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 5.315687418496074e-07,
-      "loss": 0.6992,
+      "learning_rate": 2.651183868648821e-07,
+      "loss": 0.8223,
       "step": 32729
     },
     {
-      "epoch": 0.8989041773090549,
+      "epoch": 0.9287741203178207,
       "grad_norm": 0.0,
-      "learning_rate": 5.312826258716997e-07,
-      "loss": 0.8869,
+      "learning_rate": 2.649082004496328e-07,
+      "loss": 0.8371,
       "step": 32730
     },
     {
-      "epoch": 0.8989316415368982,
+      "epoch": 0.9288024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 5.309965848139531e-07,
-      "loss": 0.795,
+      "learning_rate": 2.646980962671408e-07,
+      "loss": 0.7583,
       "step": 32731
     },
     {
-      "epoch": 0.8989591057647415,
+      "epoch": 0.9288308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 5.307106186786293e-07,
-      "loss": 0.7995,
+      "learning_rate": 2.6448807431918024e-07,
+      "loss": 0.8029,
       "step": 32732
     },
     {
-      "epoch": 0.8989865699925846,
+      "epoch": 0.9288592508513054,
       "grad_norm": 0.0,
-      "learning_rate": 5.304247274679897e-07,
-      "loss": 0.7913,
+      "learning_rate": 2.642781346075285e-07,
+      "loss": 0.8274,
       "step": 32733
     },
     {
-      "epoch": 0.8990140342204279,
+      "epoch": 0.9288876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 5.301389111842981e-07,
-      "loss": 0.7657,
+      "learning_rate": 2.6406827713395647e-07,
+      "loss": 0.8724,
       "step": 32734
     },
     {
-      "epoch": 0.8990414984482711,
+      "epoch": 0.9289160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 5.298531698298159e-07,
-      "loss": 0.8075,
+      "learning_rate": 2.638585019002371e-07,
+      "loss": 0.7801,
       "step": 32735
     },
     {
-      "epoch": 0.8990689626761144,
+      "epoch": 0.92894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 5.295675034068027e-07,
-      "loss": 0.855,
+      "learning_rate": 2.636488089081435e-07,
+      "loss": 0.7196,
       "step": 32736
     },
     {
-      "epoch": 0.8990964269039576,
+      "epoch": 0.9289727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 5.29281911917523e-07,
-      "loss": 0.7108,
+      "learning_rate": 2.634391981594453e-07,
+      "loss": 0.7579,
       "step": 32737
     },
     {
-      "epoch": 0.8991238911318008,
+      "epoch": 0.9290011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 5.289963953642307e-07,
-      "loss": 0.8214,
+      "learning_rate": 2.6322966965591444e-07,
+      "loss": 0.8387,
       "step": 32738
     },
     {
-      "epoch": 0.8991513553596441,
+      "epoch": 0.9290295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 5.287109537491875e-07,
-      "loss": 0.7662,
+      "learning_rate": 2.6302022339932066e-07,
+      "loss": 0.8193,
       "step": 32739
     },
     {
-      "epoch": 0.8991788195874872,
+      "epoch": 0.9290578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 5.284255870746547e-07,
-      "loss": 0.812,
+      "learning_rate": 2.6281085939143134e-07,
+      "loss": 0.7658,
       "step": 32740
     },
     {
-      "epoch": 0.8992062838153305,
+      "epoch": 0.9290862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 5.281402953428849e-07,
-      "loss": 0.8038,
+      "learning_rate": 2.6260157763401627e-07,
+      "loss": 0.8528,
       "step": 32741
     },
     {
-      "epoch": 0.8992337480431738,
+      "epoch": 0.9291146424517593,
       "grad_norm": 0.0,
-      "learning_rate": 5.278550785561387e-07,
-      "loss": 0.8528,
+      "learning_rate": 2.623923781288451e-07,
+      "loss": 0.8123,
       "step": 32742
     },
     {
-      "epoch": 0.899261212271017,
+      "epoch": 0.9291430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 5.275699367166742e-07,
-      "loss": 0.8991,
+      "learning_rate": 2.62183260877682e-07,
+      "loss": 0.7532,
       "step": 32743
     },
     {
-      "epoch": 0.8992886764988602,
+      "epoch": 0.9291713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 5.27284869826744e-07,
-      "loss": 0.8093,
+      "learning_rate": 2.619742258822955e-07,
+      "loss": 0.8416,
       "step": 32744
     },
     {
-      "epoch": 0.8993161407267035,
+      "epoch": 0.929199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 5.269998778886043e-07,
-      "loss": 0.8811,
+      "learning_rate": 2.617652731444509e-07,
+      "loss": 0.8211,
       "step": 32745
     },
     {
-      "epoch": 0.8993436049545467,
+      "epoch": 0.9292281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 5.267149609045119e-07,
-      "loss": 0.8477,
+      "learning_rate": 2.615564026659112e-07,
+      "loss": 0.7533,
       "step": 32746
     },
     {
-      "epoch": 0.89937106918239,
+      "epoch": 0.9292565266742339,
       "grad_norm": 0.0,
-      "learning_rate": 5.264301188767218e-07,
-      "loss": 0.8132,
+      "learning_rate": 2.613476144484428e-07,
+      "loss": 0.7102,
       "step": 32747
     },
     {
-      "epoch": 0.8993985334102331,
+      "epoch": 0.9292849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 5.261453518074844e-07,
-      "loss": 0.8019,
+      "learning_rate": 2.6113890849380875e-07,
+      "loss": 0.8981,
       "step": 32748
     },
     {
-      "epoch": 0.8994259976380764,
+      "epoch": 0.9293132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 5.258606596990567e-07,
-      "loss": 0.7946,
+      "learning_rate": 2.6093028480377203e-07,
+      "loss": 0.7546,
       "step": 32749
     },
     {
-      "epoch": 0.8994534618659197,
+      "epoch": 0.9293416572077186,
       "grad_norm": 0.0,
-      "learning_rate": 5.255760425536882e-07,
-      "loss": 0.8257,
+      "learning_rate": 2.607217433800968e-07,
+      "loss": 0.7648,
       "step": 32750
     },
     {
-      "epoch": 0.8994809260937628,
+      "epoch": 0.9293700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 5.252915003736314e-07,
-      "loss": 0.8216,
+      "learning_rate": 2.605132842245406e-07,
+      "loss": 0.828,
       "step": 32751
     },
     {
-      "epoch": 0.8995083903216061,
+      "epoch": 0.9293984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 5.25007033161139e-07,
-      "loss": 0.7653,
+      "learning_rate": 2.603049073388675e-07,
+      "loss": 0.8145,
       "step": 32752
     },
     {
-      "epoch": 0.8995358545494493,
+      "epoch": 0.9294267877412031,
       "grad_norm": 0.0,
-      "learning_rate": 5.247226409184603e-07,
-      "loss": 0.9227,
+      "learning_rate": 2.600966127248372e-07,
+      "loss": 0.7846,
       "step": 32753
     },
     {
-      "epoch": 0.8995633187772926,
+      "epoch": 0.9294551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 5.244383236478467e-07,
-      "loss": 0.8794,
+      "learning_rate": 2.598884003842084e-07,
+      "loss": 0.8365,
       "step": 32754
     },
     {
-      "epoch": 0.8995907830051358,
+      "epoch": 0.929483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 5.2415408135155e-07,
-      "loss": 0.8598,
+      "learning_rate": 2.5968027031873954e-07,
+      "loss": 0.7545,
       "step": 32755
     },
     {
-      "epoch": 0.899618247232979,
+      "epoch": 0.9295119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 5.238699140318159e-07,
-      "loss": 0.7921,
+      "learning_rate": 2.594722225301893e-07,
+      "loss": 0.8113,
       "step": 32756
     },
     {
-      "epoch": 0.8996457114608223,
+      "epoch": 0.9295402951191828,
       "grad_norm": 0.0,
-      "learning_rate": 5.235858216908929e-07,
-      "loss": 0.8541,
+      "learning_rate": 2.5926425702031523e-07,
+      "loss": 0.8329,
       "step": 32757
     },
     {
-      "epoch": 0.8996731756886656,
+      "epoch": 0.9295686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 5.23301804331029e-07,
-      "loss": 0.8035,
+      "learning_rate": 2.5905637379087357e-07,
+      "loss": 0.7587,
       "step": 32758
     },
     {
-      "epoch": 0.8997006399165087,
+      "epoch": 0.9295970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 5.230178619544723e-07,
-      "loss": 0.6834,
+      "learning_rate": 2.5884857284362187e-07,
+      "loss": 0.845,
       "step": 32759
     },
     {
-      "epoch": 0.899728104144352,
+      "epoch": 0.9296254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 5.227339945634702e-07,
-      "loss": 0.8549,
+      "learning_rate": 2.5864085418031316e-07,
+      "loss": 0.8687,
       "step": 32760
     },
     {
-      "epoch": 0.8997555683721952,
+      "epoch": 0.9296538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 5.224502021602684e-07,
-      "loss": 0.7515,
+      "learning_rate": 2.5843321780270267e-07,
+      "loss": 0.7881,
       "step": 32761
     },
     {
-      "epoch": 0.8997830326000384,
+      "epoch": 0.9296821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 5.221664847471108e-07,
-      "loss": 0.7149,
+      "learning_rate": 2.5822566371254576e-07,
+      "loss": 0.8138,
       "step": 32762
     },
     {
-      "epoch": 0.8998104968278817,
+      "epoch": 0.9297105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 5.218828423262423e-07,
-      "loss": 0.7836,
+      "learning_rate": 2.5801819191159314e-07,
+      "loss": 0.7734,
       "step": 32763
     },
     {
-      "epoch": 0.8998379610557249,
+      "epoch": 0.929738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 5.21599274899911e-07,
-      "loss": 0.8596,
+      "learning_rate": 2.578108024016002e-07,
+      "loss": 0.7853,
       "step": 32764
     },
     {
-      "epoch": 0.8998654252835682,
+      "epoch": 0.9297673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 5.213157824703563e-07,
-      "loss": 0.8594,
+      "learning_rate": 2.576034951843165e-07,
+      "loss": 0.7613,
       "step": 32765
     },
     {
-      "epoch": 0.8998928895114113,
+      "epoch": 0.9297956867196367,
       "grad_norm": 0.0,
-      "learning_rate": 5.210323650398219e-07,
-      "loss": 0.8124,
+      "learning_rate": 2.5739627026149404e-07,
+      "loss": 0.7095,
       "step": 32766
     },
     {
-      "epoch": 0.8999203537392546,
+      "epoch": 0.9298240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 5.207490226105527e-07,
-      "loss": 0.7413,
+      "learning_rate": 2.571891276348848e-07,
+      "loss": 0.7887,
       "step": 32767
     },
     {
-      "epoch": 0.8999478179670979,
+      "epoch": 0.9298524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 5.204657551847891e-07,
-      "loss": 0.7275,
+      "learning_rate": 2.5698206730623507e-07,
+      "loss": 0.799,
       "step": 32768
     },
     {
-      "epoch": 0.899975282194941,
+      "epoch": 0.9298808172531214,
       "grad_norm": 0.0,
-      "learning_rate": 5.201825627647716e-07,
-      "loss": 0.7585,
+      "learning_rate": 2.567750892772958e-07,
+      "loss": 0.8683,
       "step": 32769
     },
     {
-      "epoch": 0.9000027464227843,
+      "epoch": 0.9299091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 5.198994453527418e-07,
-      "loss": 0.8171,
+      "learning_rate": 2.5656819354981765e-07,
+      "loss": 0.7726,
       "step": 32770
     },
     {
-      "epoch": 0.9000302106506276,
+      "epoch": 0.9299375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 5.19616402950941e-07,
-      "loss": 0.7571,
+      "learning_rate": 2.563613801255438e-07,
+      "loss": 0.784,
       "step": 32771
     },
     {
-      "epoch": 0.9000576748784708,
+      "epoch": 0.9299659477866061,
       "grad_norm": 0.0,
-      "learning_rate": 5.193334355616065e-07,
-      "loss": 0.7361,
+      "learning_rate": 2.561546490062239e-07,
+      "loss": 0.7489,
       "step": 32772
     },
     {
-      "epoch": 0.900085139106314,
+      "epoch": 0.929994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 5.190505431869786e-07,
-      "loss": 0.9064,
+      "learning_rate": 2.559480001936043e-07,
+      "loss": 0.8132,
       "step": 32773
     },
     {
-      "epoch": 0.9001126033341572,
+      "epoch": 0.930022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 5.187677258292956e-07,
-      "loss": 0.7429,
+      "learning_rate": 2.5574143368942817e-07,
+      "loss": 0.6251,
       "step": 32774
     },
     {
-      "epoch": 0.9001400675620005,
+      "epoch": 0.9300510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 5.184849834907945e-07,
-      "loss": 0.688,
+      "learning_rate": 2.55534949495444e-07,
+      "loss": 0.8465,
       "step": 32775
     },
     {
-      "epoch": 0.9001675317898438,
+      "epoch": 0.9300794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 5.182023161737126e-07,
-      "loss": 0.7946,
+      "learning_rate": 2.5532854761339375e-07,
+      "loss": 0.7425,
       "step": 32776
     },
     {
-      "epoch": 0.9001949960176869,
+      "epoch": 0.9301078320090805,
       "grad_norm": 0.0,
-      "learning_rate": 5.179197238802868e-07,
-      "loss": 0.858,
+      "learning_rate": 2.551222280450205e-07,
+      "loss": 0.7996,
       "step": 32777
     },
     {
-      "epoch": 0.9002224602455302,
+      "epoch": 0.9301362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 5.176372066127533e-07,
-      "loss": 0.7417,
+      "learning_rate": 2.549159907920684e-07,
+      "loss": 0.7643,
       "step": 32778
     },
     {
-      "epoch": 0.9002499244733735,
+      "epoch": 0.9301645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 5.17354764373349e-07,
-      "loss": 0.8567,
+      "learning_rate": 2.547098358562794e-07,
+      "loss": 0.7924,
       "step": 32779
     },
     {
-      "epoch": 0.9002773887012167,
+      "epoch": 0.9301929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 5.170723971643077e-07,
-      "loss": 0.7211,
+      "learning_rate": 2.5450376323939316e-07,
+      "loss": 0.765,
       "step": 32780
     },
     {
-      "epoch": 0.9003048529290599,
+      "epoch": 0.9302213393870602,
       "grad_norm": 0.0,
-      "learning_rate": 5.167901049878599e-07,
-      "loss": 0.8436,
+      "learning_rate": 2.5429777294315394e-07,
+      "loss": 0.8625,
       "step": 32781
     },
     {
-      "epoch": 0.9003323171569031,
+      "epoch": 0.9302497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 5.165078878462437e-07,
-      "loss": 0.746,
+      "learning_rate": 2.5409186496929803e-07,
+      "loss": 0.8162,
       "step": 32782
     },
     {
-      "epoch": 0.9003597813847464,
+      "epoch": 0.9302780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 5.162257457416897e-07,
-      "loss": 0.8972,
+      "learning_rate": 2.5388603931956635e-07,
+      "loss": 0.7542,
       "step": 32783
     },
     {
-      "epoch": 0.9003872456125896,
+      "epoch": 0.9303064699205449,
       "grad_norm": 0.0,
-      "learning_rate": 5.159436786764316e-07,
-      "loss": 0.813,
+      "learning_rate": 2.5368029599569744e-07,
+      "loss": 0.7861,
       "step": 32784
     },
     {
-      "epoch": 0.9004147098404328,
+      "epoch": 0.9303348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 5.156616866527031e-07,
-      "loss": 0.8316,
+      "learning_rate": 2.5347463499942993e-07,
+      "loss": 0.7923,
       "step": 32785
     },
     {
-      "epoch": 0.9004421740682761,
+      "epoch": 0.9303632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 5.153797696727303e-07,
-      "loss": 0.7982,
+      "learning_rate": 2.532690563324991e-07,
+      "loss": 0.7498,
       "step": 32786
     },
     {
-      "epoch": 0.9004696382961193,
+      "epoch": 0.9303916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 5.15097927738748e-07,
-      "loss": 0.8,
+      "learning_rate": 2.5306355999664354e-07,
+      "loss": 0.7903,
       "step": 32787
     },
     {
-      "epoch": 0.9004971025239625,
+      "epoch": 0.9304199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 5.148161608529856e-07,
-      "loss": 0.8582,
+      "learning_rate": 2.528581459935986e-07,
+      "loss": 0.7908,
       "step": 32788
     },
     {
-      "epoch": 0.9005245667518058,
+      "epoch": 0.9304483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 5.145344690176712e-07,
-      "loss": 0.7461,
+      "learning_rate": 2.5265281432509836e-07,
+      "loss": 0.8081,
       "step": 32789
     },
     {
-      "epoch": 0.900552030979649,
+      "epoch": 0.9304767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 5.142528522350332e-07,
-      "loss": 0.8571,
+      "learning_rate": 2.5244756499287817e-07,
+      "loss": 0.7969,
       "step": 32790
     },
     {
-      "epoch": 0.9005794952074923,
+      "epoch": 0.9305051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 5.139713105073041e-07,
-      "loss": 0.7708,
+      "learning_rate": 2.52242397998671e-07,
+      "loss": 0.7662,
       "step": 32791
     },
     {
-      "epoch": 0.9006069594353355,
+      "epoch": 0.930533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 5.136898438367056e-07,
-      "loss": 0.7538,
+      "learning_rate": 2.5203731334421113e-07,
+      "loss": 0.7666,
       "step": 32792
     },
     {
-      "epoch": 0.9006344236631787,
+      "epoch": 0.9305618615209988,
       "grad_norm": 0.0,
-      "learning_rate": 5.134084522254689e-07,
-      "loss": 0.7939,
+      "learning_rate": 2.518323110312293e-07,
+      "loss": 0.8359,
       "step": 32793
     },
     {
-      "epoch": 0.900661887891022,
+      "epoch": 0.9305902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 5.131271356758183e-07,
-      "loss": 0.7918,
+      "learning_rate": 2.516273910614597e-07,
+      "loss": 0.8249,
       "step": 32794
     },
     {
-      "epoch": 0.9006893521188651,
+      "epoch": 0.9306186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 5.128458941899827e-07,
-      "loss": 0.7624,
+      "learning_rate": 2.5142255343663104e-07,
+      "loss": 0.7976,
       "step": 32795
     },
     {
-      "epoch": 0.9007168163467084,
+      "epoch": 0.9306469920544835,
       "grad_norm": 0.0,
-      "learning_rate": 5.125647277701851e-07,
-      "loss": 0.8232,
+      "learning_rate": 2.5121779815847404e-07,
+      "loss": 0.8257,
       "step": 32796
     },
     {
-      "epoch": 0.9007442805745517,
+      "epoch": 0.9306753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 5.12283636418649e-07,
-      "loss": 0.7971,
+      "learning_rate": 2.5101312522871846e-07,
+      "loss": 0.845,
       "step": 32797
     },
     {
-      "epoch": 0.9007717448023949,
+      "epoch": 0.9307037457434734,
       "grad_norm": 0.0,
-      "learning_rate": 5.120026201376005e-07,
-      "loss": 0.8187,
+      "learning_rate": 2.5080853464909515e-07,
+      "loss": 0.7966,
       "step": 32798
     },
     {
-      "epoch": 0.9007992090302381,
+      "epoch": 0.9307321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 5.117216789292622e-07,
-      "loss": 0.8279,
+      "learning_rate": 2.506040264213283e-07,
+      "loss": 0.8103,
       "step": 32799
     },
     {
-      "epoch": 0.9008266732580813,
+      "epoch": 0.9307604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 5.11440812795857e-07,
-      "loss": 0.867,
+      "learning_rate": 2.503996005471476e-07,
+      "loss": 0.8318,
       "step": 32800
     },
     {
-      "epoch": 0.9008541374859246,
+      "epoch": 0.930788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 5.111600217396084e-07,
-      "loss": 0.7616,
+      "learning_rate": 2.5019525702828063e-07,
+      "loss": 0.8274,
       "step": 32801
     },
     {
-      "epoch": 0.9008816017137679,
+      "epoch": 0.9308172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 5.108793057627382e-07,
-      "loss": 0.8519,
+      "learning_rate": 2.499909958664526e-07,
+      "loss": 0.8039,
       "step": 32802
     },
     {
-      "epoch": 0.900909065941611,
+      "epoch": 0.9308456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 5.105986648674655e-07,
-      "loss": 0.7282,
+      "learning_rate": 2.497868170633877e-07,
+      "loss": 0.83,
       "step": 32803
     },
     {
-      "epoch": 0.9009365301694543,
+      "epoch": 0.9308740068104426,
       "grad_norm": 0.0,
-      "learning_rate": 5.10318099056013e-07,
+      "learning_rate": 2.4958272062081343e-07,
       "loss": 0.8535,
       "step": 32804
     },
     {
-      "epoch": 0.9009639943972976,
+      "epoch": 0.9309023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 5.100376083305991e-07,
-      "loss": 0.8903,
+      "learning_rate": 2.493787065404518e-07,
+      "loss": 0.8544,
       "step": 32805
     },
     {
-      "epoch": 0.9009914586251407,
+      "epoch": 0.9309307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 5.09757192693443e-07,
-      "loss": 0.9291,
+      "learning_rate": 2.4917477482402585e-07,
+      "loss": 0.6626,
       "step": 32806
     },
     {
-      "epoch": 0.901018922852984,
+      "epoch": 0.9309591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 5.094768521467641e-07,
-      "loss": 0.8207,
+      "learning_rate": 2.4897092547325976e-07,
+      "loss": 0.7699,
       "step": 32807
     },
     {
-      "epoch": 0.9010463870808272,
+      "epoch": 0.9309875141884223,
       "grad_norm": 0.0,
-      "learning_rate": 5.091965866927828e-07,
-      "loss": 0.8131,
+      "learning_rate": 2.4876715848987323e-07,
+      "loss": 0.7717,
       "step": 32808
     },
     {
-      "epoch": 0.9010738513086705,
+      "epoch": 0.9310158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 5.089163963337118e-07,
-      "loss": 0.7661,
+      "learning_rate": 2.4856347387559045e-07,
+      "loss": 0.7983,
       "step": 32809
     },
     {
-      "epoch": 0.9011013155365137,
+      "epoch": 0.931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 5.086362810717715e-07,
-      "loss": 0.8558,
+      "learning_rate": 2.483598716321289e-07,
+      "loss": 0.8089,
       "step": 32810
     },
     {
-      "epoch": 0.9011287797643569,
+      "epoch": 0.9310726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 5.08356240909178e-07,
-      "loss": 0.7747,
+      "learning_rate": 2.481563517612107e-07,
+      "loss": 0.88,
       "step": 32811
     },
     {
-      "epoch": 0.9011562439922002,
+      "epoch": 0.9311010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 5.080762758481472e-07,
-      "loss": 0.7734,
+      "learning_rate": 2.4795291426455425e-07,
+      "loss": 0.8278,
       "step": 32812
     },
     {
-      "epoch": 0.9011837082200433,
+      "epoch": 0.9311293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 5.07796385890893e-07,
-      "loss": 0.818,
+      "learning_rate": 2.477495591438783e-07,
+      "loss": 0.8344,
       "step": 32813
     },
     {
-      "epoch": 0.9012111724478866,
+      "epoch": 0.9311577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 5.075165710396313e-07,
-      "loss": 0.8402,
+      "learning_rate": 2.475462864008993e-07,
+      "loss": 0.7968,
       "step": 32814
     },
     {
-      "epoch": 0.9012386366757299,
+      "epoch": 0.9311861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 5.072368312965748e-07,
-      "loss": 0.8186,
+      "learning_rate": 2.4734309603733573e-07,
+      "loss": 0.8125,
       "step": 32815
     },
     {
-      "epoch": 0.9012661009035731,
+      "epoch": 0.9312145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 5.069571666639373e-07,
-      "loss": 0.8046,
+      "learning_rate": 2.47139988054903e-07,
+      "loss": 0.7776,
       "step": 32816
     },
     {
-      "epoch": 0.9012935651314163,
+      "epoch": 0.9312429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 5.066775771439325e-07,
-      "loss": 0.7532,
+      "learning_rate": 2.469369624553175e-07,
+      "loss": 0.797,
       "step": 32817
     },
     {
-      "epoch": 0.9013210293592596,
+      "epoch": 0.9312712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 5.063980627387721e-07,
-      "loss": 0.8123,
+      "learning_rate": 2.467340192402945e-07,
+      "loss": 0.8005,
       "step": 32818
     },
     {
-      "epoch": 0.9013484935871028,
+      "epoch": 0.9312996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 5.061186234506676e-07,
-      "loss": 0.7968,
+      "learning_rate": 2.4653115841154704e-07,
+      "loss": 0.7609,
       "step": 32819
     },
     {
-      "epoch": 0.9013759578149461,
+      "epoch": 0.9313280363223609,
       "grad_norm": 0.0,
-      "learning_rate": 5.058392592818318e-07,
-      "loss": 0.8335,
+      "learning_rate": 2.463283799707894e-07,
+      "loss": 0.7265,
       "step": 32820
     },
     {
-      "epoch": 0.9014034220427892,
+      "epoch": 0.9313564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 5.055599702344716e-07,
-      "loss": 0.8122,
+      "learning_rate": 2.461256839197357e-07,
+      "loss": 0.7724,
       "step": 32821
     },
     {
-      "epoch": 0.9014308862706325,
+      "epoch": 0.9313847900113508,
       "grad_norm": 0.0,
-      "learning_rate": 5.052807563107986e-07,
-      "loss": 0.7613,
+      "learning_rate": 2.4592307026009453e-07,
+      "loss": 0.7995,
       "step": 32822
     },
     {
-      "epoch": 0.9014583504984758,
+      "epoch": 0.9314131668558456,
       "grad_norm": 0.0,
-      "learning_rate": 5.050016175130223e-07,
-      "loss": 0.857,
+      "learning_rate": 2.457205389935802e-07,
+      "loss": 0.7766,
       "step": 32823
     },
     {
-      "epoch": 0.901485814726319,
+      "epoch": 0.9314415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 5.047225538433509e-07,
-      "loss": 0.8031,
+      "learning_rate": 2.4551809012190344e-07,
+      "loss": 0.8686,
       "step": 32824
     },
     {
-      "epoch": 0.9015132789541622,
+      "epoch": 0.9314699205448355,
       "grad_norm": 0.0,
-      "learning_rate": 5.044435653039926e-07,
-      "loss": 0.7296,
+      "learning_rate": 2.4531572364677406e-07,
+      "loss": 0.9134,
       "step": 32825
     },
     {
-      "epoch": 0.9015407431820054,
+      "epoch": 0.9314982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 5.041646518971566e-07,
-      "loss": 0.9137,
+      "learning_rate": 2.4511343956990064e-07,
+      "loss": 0.8089,
       "step": 32826
     },
     {
-      "epoch": 0.9015682074098487,
+      "epoch": 0.9315266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 5.03885813625048e-07,
-      "loss": 0.8,
+      "learning_rate": 2.449112378929941e-07,
+      "loss": 0.8116,
       "step": 32827
     },
     {
-      "epoch": 0.9015956716376919,
+      "epoch": 0.93155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 5.036070504898727e-07,
-      "loss": 0.7767,
+      "learning_rate": 2.447091186177586e-07,
+      "loss": 0.7918,
       "step": 32828
     },
     {
-      "epoch": 0.9016231358655351,
+      "epoch": 0.931583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 5.033283624938368e-07,
-      "loss": 0.8015,
+      "learning_rate": 2.445070817459061e-07,
+      "loss": 0.7867,
       "step": 32829
     },
     {
-      "epoch": 0.9016506000933784,
+      "epoch": 0.9316118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 5.03049749639145e-07,
-      "loss": 0.7688,
+      "learning_rate": 2.443051272791386e-07,
+      "loss": 0.8113,
       "step": 32830
     },
     {
-      "epoch": 0.9016780643212217,
+      "epoch": 0.9316401816118047,
       "grad_norm": 0.0,
-      "learning_rate": 5.027712119280026e-07,
-      "loss": 0.8489,
+      "learning_rate": 2.441032552191658e-07,
+      "loss": 0.8231,
       "step": 32831
     },
     {
-      "epoch": 0.9017055285490648,
+      "epoch": 0.9316685584562997,
       "grad_norm": 0.0,
-      "learning_rate": 5.024927493626152e-07,
-      "loss": 0.7769,
+      "learning_rate": 2.4390146556769077e-07,
+      "loss": 0.8988,
       "step": 32832
     },
     {
-      "epoch": 0.9017329927769081,
+      "epoch": 0.9316969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 5.022143619451825e-07,
-      "loss": 0.7643,
+      "learning_rate": 2.436997583264189e-07,
+      "loss": 0.9318,
       "step": 32833
     },
     {
-      "epoch": 0.9017604570047513,
+      "epoch": 0.9317253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 5.01936049677908e-07,
-      "loss": 0.873,
+      "learning_rate": 2.434981334970532e-07,
+      "loss": 0.9136,
       "step": 32834
     },
     {
-      "epoch": 0.9017879212325945,
+      "epoch": 0.9317536889897844,
       "grad_norm": 0.0,
-      "learning_rate": 5.016578125629967e-07,
-      "loss": 0.7141,
+      "learning_rate": 2.432965910812979e-07,
+      "loss": 0.8094,
       "step": 32835
     },
     {
-      "epoch": 0.9018153854604378,
+      "epoch": 0.9317820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 5.01379650602647e-07,
-      "loss": 0.8646,
+      "learning_rate": 2.430951310808538e-07,
+      "loss": 0.9137,
       "step": 32836
     },
     {
-      "epoch": 0.901842849688281,
+      "epoch": 0.9318104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 5.011015637990602e-07,
-      "loss": 0.7349,
+      "learning_rate": 2.4289375349742516e-07,
+      "loss": 0.662,
       "step": 32837
     },
     {
-      "epoch": 0.9018703139161243,
+      "epoch": 0.931838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 5.008235521544391e-07,
-      "loss": 0.8605,
+      "learning_rate": 2.426924583327117e-07,
+      "loss": 0.7786,
       "step": 32838
     },
     {
-      "epoch": 0.9018977781439674,
+      "epoch": 0.9318671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 5.00545615670981e-07,
-      "loss": 0.816,
+      "learning_rate": 2.42491245588411e-07,
+      "loss": 0.8613,
       "step": 32839
     },
     {
-      "epoch": 0.9019252423718107,
+      "epoch": 0.9318955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 5.002677543508849e-07,
-      "loss": 0.8367,
+      "learning_rate": 2.4229011526622714e-07,
+      "loss": 0.8016,
       "step": 32840
     },
     {
-      "epoch": 0.901952706599654,
+      "epoch": 0.9319239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 4.999899681963505e-07,
-      "loss": 0.8307,
+      "learning_rate": 2.4208906736785886e-07,
+      "loss": 0.9113,
       "step": 32841
     },
     {
-      "epoch": 0.9019801708274972,
+      "epoch": 0.9319523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 4.997122572095747e-07,
-      "loss": 0.8283,
+      "learning_rate": 2.418881018950003e-07,
+      "loss": 0.7595,
       "step": 32842
     },
     {
-      "epoch": 0.9020076350553404,
+      "epoch": 0.9319807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 4.994346213927581e-07,
-      "loss": 0.7141,
+      "learning_rate": 2.416872188493535e-07,
+      "loss": 0.7697,
       "step": 32843
     },
     {
-      "epoch": 0.9020350992831837,
+      "epoch": 0.9320090805902383,
       "grad_norm": 0.0,
-      "learning_rate": 4.991570607480944e-07,
-      "loss": 0.8188,
+      "learning_rate": 2.4148641823261267e-07,
+      "loss": 0.8975,
       "step": 32844
     },
     {
-      "epoch": 0.9020625635110269,
+      "epoch": 0.9320374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 4.988795752777787e-07,
-      "loss": 0.7178,
+      "learning_rate": 2.4128570004647525e-07,
+      "loss": 0.7712,
       "step": 32845
     },
     {
-      "epoch": 0.9020900277388701,
+      "epoch": 0.9320658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 4.986021649840078e-07,
-      "loss": 0.7738,
+      "learning_rate": 2.4108506429263547e-07,
+      "loss": 0.7998,
       "step": 32846
     },
     {
-      "epoch": 0.9021174919667133,
+      "epoch": 0.932094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 4.98324829868978e-07,
-      "loss": 0.8109,
+      "learning_rate": 2.4088451097278973e-07,
+      "loss": 0.8127,
       "step": 32847
     },
     {
-      "epoch": 0.9021449561945566,
+      "epoch": 0.9321225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 4.980475699348819e-07,
-      "loss": 0.7601,
+      "learning_rate": 2.406840400886301e-07,
+      "loss": 0.7947,
       "step": 32848
     },
     {
-      "epoch": 0.9021724204223999,
+      "epoch": 0.9321509648127129,
       "grad_norm": 0.0,
-      "learning_rate": 4.977703851839144e-07,
-      "loss": 0.8429,
+      "learning_rate": 2.404836516418518e-07,
+      "loss": 0.8082,
       "step": 32849
     },
     {
-      "epoch": 0.902199884650243,
+      "epoch": 0.9321793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 4.974932756182705e-07,
-      "loss": 0.799,
+      "learning_rate": 2.4028334563414693e-07,
+      "loss": 0.8332,
       "step": 32850
     },
     {
-      "epoch": 0.9022273488780863,
+      "epoch": 0.9322077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 4.972162412401394e-07,
-      "loss": 0.7942,
+      "learning_rate": 2.400831220672062e-07,
+      "loss": 0.7776,
       "step": 32851
     },
     {
-      "epoch": 0.9022548131059295,
+      "epoch": 0.9322360953461976,
       "grad_norm": 0.0,
-      "learning_rate": 4.969392820517149e-07,
-      "loss": 0.7614,
+      "learning_rate": 2.398829809427228e-07,
+      "loss": 0.7759,
       "step": 32852
     },
     {
-      "epoch": 0.9022822773337728,
+      "epoch": 0.9322644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 4.966623980551865e-07,
-      "loss": 0.8992,
+      "learning_rate": 2.3968292226238756e-07,
+      "loss": 0.8201,
       "step": 32853
     },
     {
-      "epoch": 0.902309741561616,
+      "epoch": 0.9322928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 4.96385589252747e-07,
-      "loss": 0.7732,
+      "learning_rate": 2.394829460278891e-07,
+      "loss": 0.8616,
       "step": 32854
     },
     {
-      "epoch": 0.9023372057894592,
+      "epoch": 0.9323212258796821,
       "grad_norm": 0.0,
-      "learning_rate": 4.961088556465865e-07,
-      "loss": 0.8512,
+      "learning_rate": 2.392830522409162e-07,
+      "loss": 0.7324,
       "step": 32855
     },
     {
-      "epoch": 0.9023646700173025,
+      "epoch": 0.9323496027241771,
       "grad_norm": 0.0,
-      "learning_rate": 4.958321972388946e-07,
-      "loss": 0.8744,
+      "learning_rate": 2.390832409031574e-07,
+      "loss": 0.7543,
       "step": 32856
     },
     {
-      "epoch": 0.9023921342451457,
+      "epoch": 0.932377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 4.955556140318585e-07,
-      "loss": 0.8695,
+      "learning_rate": 2.3888351201630243e-07,
+      "loss": 0.8575,
       "step": 32857
     },
     {
-      "epoch": 0.9024195984729889,
+      "epoch": 0.9324063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 4.952791060276685e-07,
-      "loss": 0.7772,
+      "learning_rate": 2.386838655820378e-07,
+      "loss": 0.8153,
       "step": 32858
     },
     {
-      "epoch": 0.9024470627008322,
+      "epoch": 0.9324347332576618,
       "grad_norm": 0.0,
-      "learning_rate": 4.95002673228514e-07,
-      "loss": 0.8397,
+      "learning_rate": 2.384843016020488e-07,
+      "loss": 0.7533,
       "step": 32859
     },
     {
-      "epoch": 0.9024745269286754,
+      "epoch": 0.9324631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 4.947263156365778e-07,
-      "loss": 0.7919,
+      "learning_rate": 2.382848200780208e-07,
+      "loss": 0.8083,
       "step": 32860
     },
     {
-      "epoch": 0.9025019911565186,
+      "epoch": 0.9324914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 4.944500332540503e-07,
-      "loss": 0.8118,
+      "learning_rate": 2.3808542101164122e-07,
+      "loss": 0.8404,
       "step": 32861
     },
     {
-      "epoch": 0.9025294553843619,
+      "epoch": 0.9325198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 4.941738260831153e-07,
-      "loss": 0.8749,
+      "learning_rate": 2.3788610440459214e-07,
+      "loss": 0.8128,
       "step": 32862
     },
     {
-      "epoch": 0.9025569196122051,
+      "epoch": 0.9325482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 4.938976941259599e-07,
-      "loss": 0.887,
+      "learning_rate": 2.3768687025855774e-07,
+      "loss": 0.7236,
       "step": 32863
     },
     {
-      "epoch": 0.9025843838400484,
+      "epoch": 0.9325766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 4.936216373847669e-07,
-      "loss": 0.8341,
+      "learning_rate": 2.3748771857522224e-07,
+      "loss": 0.6872,
       "step": 32864
     },
     {
-      "epoch": 0.9026118480678915,
+      "epoch": 0.9326049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 4.933456558617234e-07,
-      "loss": 0.8263,
+      "learning_rate": 2.372886493562654e-07,
+      "loss": 0.7561,
       "step": 32865
     },
     {
-      "epoch": 0.9026393122957348,
+      "epoch": 0.932633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 4.930697495590109e-07,
-      "loss": 0.8583,
+      "learning_rate": 2.370896626033714e-07,
+      "loss": 0.9048,
       "step": 32866
     },
     {
-      "epoch": 0.9026667765235781,
+      "epoch": 0.9326617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 4.927939184788144e-07,
-      "loss": 0.8271,
+      "learning_rate": 2.3689075831822006e-07,
+      "loss": 0.8029,
       "step": 32867
     },
     {
-      "epoch": 0.9026942407514212,
+      "epoch": 0.9326901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 4.925181626233145e-07,
-      "loss": 0.8051,
+      "learning_rate": 2.3669193650249e-07,
+      "loss": 0.8129,
       "step": 32868
     },
     {
-      "epoch": 0.9027217049792645,
+      "epoch": 0.9327185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 4.922424819946936e-07,
-      "loss": 0.8364,
+      "learning_rate": 2.364931971578621e-07,
+      "loss": 0.7742,
       "step": 32869
     },
     {
-      "epoch": 0.9027491692071078,
+      "epoch": 0.9327468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 4.919668765951325e-07,
-      "loss": 0.7906,
+      "learning_rate": 2.3629454028601617e-07,
+      "loss": 0.8302,
       "step": 32870
     },
     {
-      "epoch": 0.902776633434951,
+      "epoch": 0.9327752553916004,
       "grad_norm": 0.0,
-      "learning_rate": 4.916913464268125e-07,
-      "loss": 0.7159,
+      "learning_rate": 2.3609596588862748e-07,
+      "loss": 0.8653,
       "step": 32871
     },
     {
-      "epoch": 0.9028040976627942,
+      "epoch": 0.9328036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 4.914158914919131e-07,
-      "loss": 0.7413,
+      "learning_rate": 2.3589747396737804e-07,
+      "loss": 0.8038,
       "step": 32872
     },
     {
-      "epoch": 0.9028315618906374,
+      "epoch": 0.9328320090805903,
       "grad_norm": 0.0,
-      "learning_rate": 4.911405117926172e-07,
-      "loss": 0.8192,
+      "learning_rate": 2.3569906452393876e-07,
+      "loss": 0.779,
       "step": 32873
     },
     {
-      "epoch": 0.9028590261184807,
+      "epoch": 0.9328603859250851,
       "grad_norm": 0.0,
-      "learning_rate": 4.908652073310993e-07,
-      "loss": 0.7069,
+      "learning_rate": 2.3550073755998937e-07,
+      "loss": 0.7747,
       "step": 32874
     },
     {
-      "epoch": 0.902886490346324,
+      "epoch": 0.93288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.905899781095391e-07,
-      "loss": 0.844,
+      "learning_rate": 2.3530249307720521e-07,
+      "loss": 0.8001,
       "step": 32875
     },
     {
-      "epoch": 0.9029139545741671,
+      "epoch": 0.932917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 4.903148241301159e-07,
-      "loss": 0.8746,
+      "learning_rate": 2.3510433107725827e-07,
+      "loss": 0.8715,
       "step": 32876
     },
     {
-      "epoch": 0.9029414188020104,
+      "epoch": 0.9329455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 4.900397453950035e-07,
-      "loss": 0.7977,
+      "learning_rate": 2.34906251561825e-07,
+      "loss": 0.6895,
       "step": 32877
     },
     {
-      "epoch": 0.9029688830298536,
+      "epoch": 0.9329738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 4.897647419063811e-07,
-      "loss": 0.8092,
+      "learning_rate": 2.347082545325774e-07,
+      "loss": 0.7736,
       "step": 32878
     },
     {
-      "epoch": 0.9029963472576968,
+      "epoch": 0.9330022701475595,
       "grad_norm": 0.0,
-      "learning_rate": 4.89489813666425e-07,
-      "loss": 0.7961,
+      "learning_rate": 2.3451033999118854e-07,
+      "loss": 0.83,
       "step": 32879
     },
     {
-      "epoch": 0.9030238114855401,
+      "epoch": 0.9330306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 4.892149606773089e-07,
-      "loss": 0.7053,
+      "learning_rate": 2.3431250793932825e-07,
+      "loss": 0.8935,
       "step": 32880
     },
     {
-      "epoch": 0.9030512757133833,
+      "epoch": 0.9330590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 4.889401829412077e-07,
-      "loss": 0.7365,
+      "learning_rate": 2.3411475837867182e-07,
+      "loss": 0.8535,
       "step": 32881
     },
     {
-      "epoch": 0.9030787399412266,
+      "epoch": 0.9330874006810442,
       "grad_norm": 0.0,
-      "learning_rate": 4.886654804602963e-07,
-      "loss": 0.8542,
+      "learning_rate": 2.3391709131088457e-07,
+      "loss": 0.7633,
       "step": 32882
     },
     {
-      "epoch": 0.9031062041690698,
+      "epoch": 0.9331157775255392,
       "grad_norm": 0.0,
-      "learning_rate": 4.883908532367487e-07,
-      "loss": 0.8369,
+      "learning_rate": 2.3371950673763966e-07,
+      "loss": 0.7347,
       "step": 32883
     },
     {
-      "epoch": 0.903133668396913,
+      "epoch": 0.9331441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 4.881163012727363e-07,
-      "loss": 0.8015,
+      "learning_rate": 2.3352200466060571e-07,
+      "loss": 0.7022,
       "step": 32884
     },
     {
-      "epoch": 0.9031611326247563,
+      "epoch": 0.9331725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 4.878418245704331e-07,
-      "loss": 0.7517,
+      "learning_rate": 2.3332458508144916e-07,
+      "loss": 0.7872,
       "step": 32885
     },
     {
-      "epoch": 0.9031885968525994,
+      "epoch": 0.9332009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 4.875674231320093e-07,
-      "loss": 0.8646,
+      "learning_rate": 2.3312724800183983e-07,
+      "loss": 0.8155,
       "step": 32886
     },
     {
-      "epoch": 0.9032160610804427,
+      "epoch": 0.9332292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 4.872930969596368e-07,
-      "loss": 0.86,
+      "learning_rate": 2.3292999342344303e-07,
+      "loss": 0.81,
       "step": 32887
     },
     {
-      "epoch": 0.903243525308286,
+      "epoch": 0.9332576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 4.870188460554848e-07,
-      "loss": 0.8312,
+      "learning_rate": 2.327328213479252e-07,
+      "loss": 0.7779,
       "step": 32888
     },
     {
-      "epoch": 0.9032709895361292,
+      "epoch": 0.9332860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 4.86744670421726e-07,
-      "loss": 0.7816,
+      "learning_rate": 2.3253573177695388e-07,
+      "loss": 0.717,
       "step": 32889
     },
     {
-      "epoch": 0.9032984537639724,
+      "epoch": 0.9333144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 4.864705700605288e-07,
-      "loss": 0.8218,
+      "learning_rate": 2.323387247121911e-07,
+      "loss": 0.7579,
       "step": 32890
     },
     {
-      "epoch": 0.9033259179918156,
+      "epoch": 0.9333427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.861965449740624e-07,
-      "loss": 0.7914,
+      "learning_rate": 2.321418001553022e-07,
+      "loss": 0.8109,
       "step": 32891
     },
     {
-      "epoch": 0.9033533822196589,
+      "epoch": 0.9333711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 4.859225951644941e-07,
-      "loss": 0.8071,
+      "learning_rate": 2.3194495810795137e-07,
+      "loss": 0.8576,
       "step": 32892
     },
     {
-      "epoch": 0.9033808464475022,
+      "epoch": 0.9333995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 4.856487206339899e-07,
-      "loss": 0.8068,
+      "learning_rate": 2.3174819857180065e-07,
+      "loss": 0.8279,
       "step": 32893
     },
     {
-      "epoch": 0.9034083106753453,
+      "epoch": 0.933427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 4.853749213847192e-07,
-      "loss": 0.8855,
+      "learning_rate": 2.315515215485109e-07,
+      "loss": 0.9006,
       "step": 32894
     },
     {
-      "epoch": 0.9034357749031886,
+      "epoch": 0.9334562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 4.85101197418848e-07,
-      "loss": 0.8658,
+      "learning_rate": 2.3135492703974638e-07,
+      "loss": 0.8073,
       "step": 32895
     },
     {
-      "epoch": 0.9034632391310319,
+      "epoch": 0.9334846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 4.848275487385412e-07,
-      "loss": 0.8036,
+      "learning_rate": 2.3115841504716463e-07,
+      "loss": 0.7323,
       "step": 32896
     },
     {
-      "epoch": 0.903490703358875,
+      "epoch": 0.9335130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 4.845539753459661e-07,
-      "loss": 0.7996,
+      "learning_rate": 2.3096198557242656e-07,
+      "loss": 0.9456,
       "step": 32897
     },
     {
-      "epoch": 0.9035181675867183,
+      "epoch": 0.9335414301929625,
       "grad_norm": 0.0,
-      "learning_rate": 4.842804772432852e-07,
-      "loss": 0.75,
+      "learning_rate": 2.3076563861719305e-07,
+      "loss": 0.7631,
       "step": 32898
     },
     {
-      "epoch": 0.9035456318145615,
+      "epoch": 0.9335698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 4.840070544326625e-07,
-      "loss": 0.7149,
+      "learning_rate": 2.3056937418311943e-07,
+      "loss": 0.8277,
       "step": 32899
     },
     {
-      "epoch": 0.9035730960424048,
+      "epoch": 0.9335981838819524,
       "grad_norm": 0.0,
-      "learning_rate": 4.837337069162629e-07,
-      "loss": 0.8677,
+      "learning_rate": 2.303731922718666e-07,
+      "loss": 0.7634,
       "step": 32900
     },
     {
-      "epoch": 0.903600560270248,
+      "epoch": 0.9336265607264472,
       "grad_norm": 0.0,
-      "learning_rate": 4.834604346962479e-07,
-      "loss": 0.8267,
+      "learning_rate": 2.3017709288508993e-07,
+      "loss": 0.6845,
       "step": 32901
     },
     {
-      "epoch": 0.9036280244980912,
+      "epoch": 0.9336549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 4.831872377747804e-07,
-      "loss": 0.8903,
+      "learning_rate": 2.2998107602444697e-07,
+      "loss": 0.7477,
       "step": 32902
     },
     {
-      "epoch": 0.9036554887259345,
+      "epoch": 0.933683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.829141161540229e-07,
-      "loss": 0.8194,
+      "learning_rate": 2.2978514169159417e-07,
+      "loss": 0.832,
       "step": 32903
     },
     {
-      "epoch": 0.9036829529537777,
+      "epoch": 0.9337116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.82641069836135e-07,
-      "loss": 0.8119,
+      "learning_rate": 2.2958928988818463e-07,
+      "loss": 0.7504,
       "step": 32904
     },
     {
-      "epoch": 0.9037104171816209,
+      "epoch": 0.9337400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 4.823680988232771e-07,
-      "loss": 0.7383,
+      "learning_rate": 2.293935206158726e-07,
+      "loss": 0.7767,
       "step": 32905
     },
     {
-      "epoch": 0.9037378814094642,
+      "epoch": 0.9337684449489216,
       "grad_norm": 0.0,
-      "learning_rate": 4.820952031176096e-07,
-      "loss": 0.7008,
+      "learning_rate": 2.2919783387631455e-07,
+      "loss": 0.824,
       "step": 32906
     },
     {
-      "epoch": 0.9037653456373074,
+      "epoch": 0.9337968217934166,
       "grad_norm": 0.0,
-      "learning_rate": 4.818223827212931e-07,
-      "loss": 0.8563,
+      "learning_rate": 2.2900222967116025e-07,
+      "loss": 0.7785,
       "step": 32907
     },
     {
-      "epoch": 0.9037928098651506,
+      "epoch": 0.9338251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 4.815496376364825e-07,
-      "loss": 0.8612,
+      "learning_rate": 2.2880670800206395e-07,
+      "loss": 0.8334,
       "step": 32908
     },
     {
-      "epoch": 0.9038202740929939,
+      "epoch": 0.9338535754824063,
       "grad_norm": 0.0,
-      "learning_rate": 4.812769678653407e-07,
-      "loss": 0.6911,
+      "learning_rate": 2.2861126887067652e-07,
+      "loss": 0.8309,
       "step": 32909
     },
     {
-      "epoch": 0.9038477383208371,
+      "epoch": 0.9338819523269013,
       "grad_norm": 0.0,
-      "learning_rate": 4.810043734100211e-07,
-      "loss": 0.7413,
+      "learning_rate": 2.284159122786489e-07,
+      "loss": 0.8611,
       "step": 32910
     },
     {
-      "epoch": 0.9038752025486804,
+      "epoch": 0.9339103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 4.807318542726824e-07,
-      "loss": 0.7438,
+      "learning_rate": 2.2822063822763084e-07,
+      "loss": 0.7672,
       "step": 32911
     },
     {
-      "epoch": 0.9039026667765235,
+      "epoch": 0.933938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 4.804594104554805e-07,
-      "loss": 0.6583,
+      "learning_rate": 2.280254467192744e-07,
+      "loss": 0.7898,
       "step": 32912
     },
     {
-      "epoch": 0.9039301310043668,
+      "epoch": 0.9339670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 4.801870419605714e-07,
-      "loss": 0.7997,
+      "learning_rate": 2.278303377552238e-07,
+      "loss": 0.8103,
       "step": 32913
     },
     {
-      "epoch": 0.9039575952322101,
+      "epoch": 0.9339954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 4.799147487901101e-07,
-      "loss": 0.8151,
+      "learning_rate": 2.2763531133712992e-07,
+      "loss": 0.8458,
       "step": 32914
     },
     {
-      "epoch": 0.9039850594600533,
+      "epoch": 0.9340238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 4.796425309462526e-07,
-      "loss": 0.867,
+      "learning_rate": 2.2744036746664145e-07,
+      "loss": 0.9305,
       "step": 32915
     },
     {
-      "epoch": 0.9040125236878965,
+      "epoch": 0.9340522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 4.793703884311506e-07,
-      "loss": 0.8178,
+      "learning_rate": 2.2724550614540152e-07,
+      "loss": 0.9314,
       "step": 32916
     },
     {
-      "epoch": 0.9040399879157397,
+      "epoch": 0.9340805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 4.790983212469568e-07,
-      "loss": 0.8031,
+      "learning_rate": 2.2705072737505994e-07,
+      "loss": 0.7972,
       "step": 32917
     },
     {
-      "epoch": 0.904067452143583,
+      "epoch": 0.9341089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 4.78826329395825e-07,
-      "loss": 0.8624,
+      "learning_rate": 2.2685603115725874e-07,
+      "loss": 0.8188,
       "step": 32918
     },
     {
-      "epoch": 0.9040949163714262,
+      "epoch": 0.9341373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 4.78554412879908e-07,
-      "loss": 0.7957,
+      "learning_rate": 2.2666141749364434e-07,
+      "loss": 0.8601,
       "step": 32919
     },
     {
-      "epoch": 0.9041223805992694,
+      "epoch": 0.9341657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 4.782825717013583e-07,
-      "loss": 0.9164,
+      "learning_rate": 2.26466886385861e-07,
+      "loss": 0.886,
       "step": 32920
     },
     {
-      "epoch": 0.9041498448271127,
+      "epoch": 0.9341940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 4.780108058623234e-07,
-      "loss": 0.7865,
+      "learning_rate": 2.2627243783555075e-07,
+      "loss": 0.7706,
       "step": 32921
     },
     {
-      "epoch": 0.904177309054956,
+      "epoch": 0.93422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 4.777391153649558e-07,
-      "loss": 0.7494,
+      "learning_rate": 2.260780718443567e-07,
+      "loss": 0.9071,
       "step": 32922
     },
     {
-      "epoch": 0.9042047732827991,
+      "epoch": 0.9342508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 4.774675002114038e-07,
-      "loss": 0.7018,
+      "learning_rate": 2.258837884139209e-07,
+      "loss": 0.7914,
       "step": 32923
     },
     {
-      "epoch": 0.9042322375106424,
+      "epoch": 0.9342792281498298,
       "grad_norm": 0.0,
-      "learning_rate": 4.771959604038201e-07,
-      "loss": 0.7351,
+      "learning_rate": 2.256895875458831e-07,
+      "loss": 0.908,
       "step": 32924
     },
     {
-      "epoch": 0.9042597017384856,
+      "epoch": 0.9343076049943246,
       "grad_norm": 0.0,
-      "learning_rate": 4.769244959443486e-07,
-      "loss": 0.8201,
+      "learning_rate": 2.2549546924188537e-07,
+      "loss": 0.74,
       "step": 32925
     },
     {
-      "epoch": 0.9042871659663289,
+      "epoch": 0.9343359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 4.766531068351421e-07,
-      "loss": 0.8793,
+      "learning_rate": 2.253014335035675e-07,
+      "loss": 0.86,
       "step": 32926
     },
     {
-      "epoch": 0.9043146301941721,
+      "epoch": 0.9343643586833145,
       "grad_norm": 0.0,
-      "learning_rate": 4.763817930783421e-07,
-      "loss": 0.8503,
+      "learning_rate": 2.2510748033256703e-07,
+      "loss": 0.8615,
       "step": 32927
     },
     {
-      "epoch": 0.9043420944220153,
+      "epoch": 0.9343927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 4.7611055467610026e-07,
-      "loss": 0.8221,
+      "learning_rate": 2.2491360973052267e-07,
+      "loss": 0.8676,
       "step": 32928
     },
     {
-      "epoch": 0.9043695586498586,
+      "epoch": 0.9344211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.7583939163055924e-07,
-      "loss": 0.775,
+      "learning_rate": 2.2471982169907424e-07,
+      "loss": 0.8155,
       "step": 32929
     },
     {
-      "epoch": 0.9043970228777017,
+      "epoch": 0.934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 4.7556830394386745e-07,
-      "loss": 0.8087,
+      "learning_rate": 2.2452611623985488e-07,
+      "loss": 0.8083,
       "step": 32930
     },
     {
-      "epoch": 0.904424487105545,
+      "epoch": 0.934477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 4.752972916181686e-07,
-      "loss": 0.8158,
+      "learning_rate": 2.2433249335450324e-07,
+      "loss": 0.688,
       "step": 32931
     },
     {
-      "epoch": 0.9044519513333883,
+      "epoch": 0.9345062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 4.7502635465560773e-07,
-      "loss": 0.7161,
+      "learning_rate": 2.241389530446547e-07,
+      "loss": 0.6982,
       "step": 32932
     },
     {
-      "epoch": 0.9044794155612315,
+      "epoch": 0.9345346197502837,
       "grad_norm": 0.0,
-      "learning_rate": 4.747554930583276e-07,
-      "loss": 0.7953,
+      "learning_rate": 2.2394549531194353e-07,
+      "loss": 0.6939,
       "step": 32933
     },
     {
-      "epoch": 0.9045068797890747,
+      "epoch": 0.9345629965947787,
       "grad_norm": 0.0,
-      "learning_rate": 4.744847068284708e-07,
-      "loss": 0.8462,
+      "learning_rate": 2.2375212015800507e-07,
+      "loss": 0.7689,
       "step": 32934
     },
     {
-      "epoch": 0.904534344016918,
+      "epoch": 0.9345913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 4.742139959681813e-07,
-      "loss": 0.751,
+      "learning_rate": 2.2355882758447135e-07,
+      "loss": 0.83,
       "step": 32935
     },
     {
-      "epoch": 0.9045618082447612,
+      "epoch": 0.9346197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 4.739433604796007e-07,
-      "loss": 0.8312,
+      "learning_rate": 2.233656175929766e-07,
+      "loss": 0.7853,
       "step": 32936
     },
     {
-      "epoch": 0.9045892724726045,
+      "epoch": 0.9346481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 4.736728003648694e-07,
-      "loss": 0.7809,
+      "learning_rate": 2.2317249018515175e-07,
+      "loss": 0.888,
       "step": 32937
     },
     {
-      "epoch": 0.9046167367004476,
+      "epoch": 0.9346765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 4.734023156261314e-07,
-      "loss": 0.7893,
+      "learning_rate": 2.2297944536262772e-07,
+      "loss": 0.7869,
       "step": 32938
     },
     {
-      "epoch": 0.9046442009282909,
+      "epoch": 0.9347048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 4.7313190626552263e-07,
-      "loss": 0.7748,
+      "learning_rate": 2.2278648312703543e-07,
+      "loss": 0.7652,
       "step": 32939
     },
     {
-      "epoch": 0.9046716651561342,
+      "epoch": 0.934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 4.7286157228518593e-07,
-      "loss": 0.8136,
+      "learning_rate": 2.225936034800069e-07,
+      "loss": 0.7867,
       "step": 32940
     },
     {
-      "epoch": 0.9046991293839773,
+      "epoch": 0.9347616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 4.725913136872573e-07,
-      "loss": 0.8493,
+      "learning_rate": 2.2240080642316751e-07,
+      "loss": 0.7074,
       "step": 32941
     },
     {
-      "epoch": 0.9047265936118206,
+      "epoch": 0.9347900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 4.723211304738773e-07,
-      "loss": 0.7036,
+      "learning_rate": 2.222080919581493e-07,
+      "loss": 0.7815,
       "step": 32942
     },
     {
-      "epoch": 0.9047540578396638,
+      "epoch": 0.9348183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 4.7205102264718304e-07,
-      "loss": 0.7578,
+      "learning_rate": 2.220154600865798e-07,
+      "loss": 0.753,
       "step": 32943
     },
     {
-      "epoch": 0.9047815220675071,
+      "epoch": 0.9348467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 4.717809902093129e-07,
-      "loss": 0.7438,
+      "learning_rate": 2.2182291081008334e-07,
+      "loss": 0.8795,
       "step": 32944
     },
     {
-      "epoch": 0.9048089862953503,
+      "epoch": 0.9348751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 4.7151103316240174e-07,
-      "loss": 0.898,
+      "learning_rate": 2.2163044413028855e-07,
+      "loss": 0.906,
       "step": 32945
     },
     {
-      "epoch": 0.9048364505231935,
+      "epoch": 0.9349035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 4.7124115150858576e-07,
-      "loss": 0.7921,
+      "learning_rate": 2.2143806004882307e-07,
+      "loss": 0.7189,
       "step": 32946
     },
     {
-      "epoch": 0.9048639147510368,
+      "epoch": 0.9349318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 4.7097134525000087e-07,
-      "loss": 0.7771,
+      "learning_rate": 2.2124575856730669e-07,
+      "loss": 0.8028,
       "step": 32947
     },
     {
-      "epoch": 0.9048913789788801,
+      "epoch": 0.9349602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 4.7070161438878323e-07,
-      "loss": 0.8171,
+      "learning_rate": 2.210535396873681e-07,
+      "loss": 0.8303,
       "step": 32948
     },
     {
-      "epoch": 0.9049188432067232,
+      "epoch": 0.934988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 4.704319589270645e-07,
-      "loss": 0.8309,
+      "learning_rate": 2.2086140341063155e-07,
+      "loss": 0.7769,
       "step": 32949
     },
     {
-      "epoch": 0.9049463074345665,
+      "epoch": 0.9350170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 4.701623788669818e-07,
-      "loss": 0.6854,
+      "learning_rate": 2.2066934973871689e-07,
+      "loss": 0.8917,
       "step": 32950
     },
     {
-      "epoch": 0.9049737716624097,
+      "epoch": 0.9350454029511919,
       "grad_norm": 0.0,
-      "learning_rate": 4.698928742106634e-07,
-      "loss": 0.7439,
+      "learning_rate": 2.2047737867324835e-07,
+      "loss": 0.7973,
       "step": 32951
     },
     {
-      "epoch": 0.9050012358902529,
+      "epoch": 0.9350737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 4.6962344496024435e-07,
-      "loss": 0.8059,
+      "learning_rate": 2.2028549021584689e-07,
+      "loss": 0.7758,
       "step": 32952
     },
     {
-      "epoch": 0.9050287001180962,
+      "epoch": 0.9351021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 4.6935409111785624e-07,
-      "loss": 0.8084,
+      "learning_rate": 2.200936843681334e-07,
+      "loss": 0.8167,
       "step": 32953
     },
     {
-      "epoch": 0.9050561643459394,
+      "epoch": 0.9351305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 4.6908481268563066e-07,
-      "loss": 0.8857,
+      "learning_rate": 2.1990196113172768e-07,
+      "loss": 0.8814,
       "step": 32954
     },
     {
-      "epoch": 0.9050836285737827,
+      "epoch": 0.9351589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 4.688156096656993e-07,
-      "loss": 0.8027,
+      "learning_rate": 2.1971032050825182e-07,
+      "loss": 0.8581,
       "step": 32955
     },
     {
-      "epoch": 0.9051110928016258,
+      "epoch": 0.9351872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 4.6854648206019037e-07,
-      "loss": 0.7682,
+      "learning_rate": 2.1951876249932003e-07,
+      "loss": 0.7684,
       "step": 32956
     },
     {
-      "epoch": 0.9051385570294691,
+      "epoch": 0.9352156640181611,
       "grad_norm": 0.0,
-      "learning_rate": 4.6827742987123225e-07,
-      "loss": 0.7566,
+      "learning_rate": 2.193272871065544e-07,
+      "loss": 0.8145,
       "step": 32957
     },
     {
-      "epoch": 0.9051660212573124,
+      "epoch": 0.9352440408626561,
       "grad_norm": 0.0,
-      "learning_rate": 4.6800845310095655e-07,
-      "loss": 0.7198,
+      "learning_rate": 2.1913589433157022e-07,
+      "loss": 0.819,
       "step": 32958
     },
     {
-      "epoch": 0.9051934854851555,
+      "epoch": 0.935272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 4.677395517514893e-07,
-      "loss": 0.905,
+      "learning_rate": 2.1894458417598518e-07,
+      "loss": 0.7711,
       "step": 32959
     },
     {
-      "epoch": 0.9052209497129988,
+      "epoch": 0.9353007945516458,
       "grad_norm": 0.0,
-      "learning_rate": 4.6747072582496e-07,
-      "loss": 0.7677,
+      "learning_rate": 2.187533566414146e-07,
+      "loss": 0.8523,
       "step": 32960
     },
     {
-      "epoch": 0.9052484139408421,
+      "epoch": 0.9353291713961408,
       "grad_norm": 0.0,
-      "learning_rate": 4.6720197532349466e-07,
-      "loss": 0.8048,
+      "learning_rate": 2.18562211729475e-07,
+      "loss": 0.7935,
       "step": 32961
     },
     {
-      "epoch": 0.9052758781686853,
+      "epoch": 0.9353575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 4.6693330024922046e-07,
-      "loss": 0.775,
+      "learning_rate": 2.183711494417784e-07,
+      "loss": 0.7912,
       "step": 32962
     },
     {
-      "epoch": 0.9053033423965285,
+      "epoch": 0.9353859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 4.6666470060426237e-07,
-      "loss": 0.8219,
+      "learning_rate": 2.1818016977994128e-07,
+      "loss": 0.6633,
       "step": 32963
     },
     {
-      "epoch": 0.9053308066243717,
+      "epoch": 0.9354143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 4.663961763907465e-07,
-      "loss": 0.8181,
+      "learning_rate": 2.1798927274557567e-07,
+      "loss": 0.7448,
       "step": 32964
     },
     {
-      "epoch": 0.905358270852215,
+      "epoch": 0.9354426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 4.661277276107967e-07,
-      "loss": 0.8153,
+      "learning_rate": 2.177984583402948e-07,
+      "loss": 1.0289,
       "step": 32965
     },
     {
-      "epoch": 0.9053857350800583,
+      "epoch": 0.9354710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 4.658593542665368e-07,
-      "loss": 0.8087,
+      "learning_rate": 2.176077265657106e-07,
+      "loss": 0.8952,
       "step": 32966
     },
     {
-      "epoch": 0.9054131993079014,
+      "epoch": 0.9354994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 4.6559105636009075e-07,
-      "loss": 0.8287,
+      "learning_rate": 2.17417077423433e-07,
+      "loss": 0.7203,
       "step": 32967
     },
     {
-      "epoch": 0.9054406635357447,
+      "epoch": 0.935527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 4.653228338935822e-07,
-      "loss": 0.7458,
+      "learning_rate": 2.1722651091507284e-07,
+      "loss": 0.8557,
       "step": 32968
     },
     {
-      "epoch": 0.905468127763588,
+      "epoch": 0.9355561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 4.6505468686913305e-07,
-      "loss": 0.8172,
+      "learning_rate": 2.1703602704224114e-07,
+      "loss": 0.8502,
       "step": 32969
     },
     {
-      "epoch": 0.9054955919914311,
+      "epoch": 0.9355845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 4.647866152888636e-07,
-      "loss": 0.8448,
+      "learning_rate": 2.1684562580654546e-07,
+      "loss": 0.7321,
       "step": 32970
     },
     {
-      "epoch": 0.9055230562192744,
+      "epoch": 0.9356129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 4.645186191548967e-07,
-      "loss": 0.7151,
+      "learning_rate": 2.166553072095945e-07,
+      "loss": 0.7748,
       "step": 32971
     },
     {
-      "epoch": 0.9055505204471176,
+      "epoch": 0.9356413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 4.642506984693529e-07,
-      "loss": 0.7826,
+      "learning_rate": 2.1646507125299587e-07,
+      "loss": 0.9369,
       "step": 32972
     },
     {
-      "epoch": 0.9055779846749609,
+      "epoch": 0.9356696935300794,
       "grad_norm": 0.0,
-      "learning_rate": 4.6398285323435043e-07,
-      "loss": 0.801,
+      "learning_rate": 2.162749179383572e-07,
+      "loss": 0.7663,
       "step": 32973
     },
     {
-      "epoch": 0.9056054489028041,
+      "epoch": 0.9356980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 4.63715083452011e-07,
-      "loss": 0.9194,
+      "learning_rate": 2.1608484726728275e-07,
+      "loss": 0.8414,
       "step": 32974
     },
     {
-      "epoch": 0.9056329131306473,
+      "epoch": 0.9357264472190693,
       "grad_norm": 0.0,
-      "learning_rate": 4.634473891244495e-07,
-      "loss": 0.8009,
+      "learning_rate": 2.158948592413812e-07,
+      "loss": 0.743,
       "step": 32975
     },
     {
-      "epoch": 0.9056603773584906,
+      "epoch": 0.9357548240635641,
       "grad_norm": 0.0,
-      "learning_rate": 4.631797702537877e-07,
-      "loss": 0.8022,
+      "learning_rate": 2.1570495386225355e-07,
+      "loss": 0.8098,
       "step": 32976
     },
     {
-      "epoch": 0.9056878415863338,
+      "epoch": 0.935783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 4.629122268421404e-07,
-      "loss": 0.7702,
+      "learning_rate": 2.1551513113150735e-07,
+      "loss": 0.7488,
       "step": 32977
     },
     {
-      "epoch": 0.905715305814177,
+      "epoch": 0.935811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 4.626447588916272e-07,
-      "loss": 0.8389,
+      "learning_rate": 2.1532539105074357e-07,
+      "loss": 0.6977,
       "step": 32978
     },
     {
-      "epoch": 0.9057427700420203,
+      "epoch": 0.9358399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 4.62377366404364e-07,
-      "loss": 0.7639,
+      "learning_rate": 2.1513573362156647e-07,
+      "loss": 0.8712,
       "step": 32979
     },
     {
-      "epoch": 0.9057702342698635,
+      "epoch": 0.9358683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.6211004938246483e-07,
-      "loss": 0.7961,
+      "learning_rate": 2.1494615884557924e-07,
+      "loss": 0.754,
       "step": 32980
     },
     {
-      "epoch": 0.9057976984977067,
+      "epoch": 0.9358967082860385,
       "grad_norm": 0.0,
-      "learning_rate": 4.618428078280457e-07,
-      "loss": 0.8123,
+      "learning_rate": 2.1475666672437945e-07,
+      "loss": 0.7248,
       "step": 32981
     },
     {
-      "epoch": 0.90582516272555,
+      "epoch": 0.9359250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 4.615756417432193e-07,
-      "loss": 0.6558,
+      "learning_rate": 2.145672572595714e-07,
+      "loss": 0.7845,
       "step": 32982
     },
     {
-      "epoch": 0.9058526269533932,
+      "epoch": 0.9359534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 4.6130855113010297e-07,
-      "loss": 0.7231,
+      "learning_rate": 2.143779304527538e-07,
+      "loss": 0.8407,
       "step": 32983
     },
     {
-      "epoch": 0.9058800911812365,
+      "epoch": 0.9359818388195232,
       "grad_norm": 0.0,
-      "learning_rate": 4.6104153599080717e-07,
-      "loss": 0.8373,
+      "learning_rate": 2.1418868630552424e-07,
+      "loss": 0.7633,
       "step": 32984
     },
     {
-      "epoch": 0.9059075554090796,
+      "epoch": 0.9360102156640182,
       "grad_norm": 0.0,
-      "learning_rate": 4.607745963274479e-07,
-      "loss": 0.8886,
+      "learning_rate": 2.139995248194837e-07,
+      "loss": 0.7669,
       "step": 32985
     },
     {
-      "epoch": 0.9059350196369229,
+      "epoch": 0.9360385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 4.605077321421325e-07,
-      "loss": 0.7744,
+      "learning_rate": 2.138104459962298e-07,
+      "loss": 0.8552,
       "step": 32986
     },
     {
-      "epoch": 0.9059624838647662,
+      "epoch": 0.9360669693530079,
       "grad_norm": 0.0,
-      "learning_rate": 4.602409434369759e-07,
-      "loss": 0.8392,
+      "learning_rate": 2.136214498373579e-07,
+      "loss": 0.9534,
       "step": 32987
     },
     {
-      "epoch": 0.9059899480926094,
+      "epoch": 0.9360953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 4.599742302140897e-07,
-      "loss": 0.8754,
+      "learning_rate": 2.134325363444656e-07,
+      "loss": 0.8112,
       "step": 32988
     },
     {
-      "epoch": 0.9060174123204526,
+      "epoch": 0.9361237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 4.597075924755823e-07,
-      "loss": 0.7822,
+      "learning_rate": 2.1324370551914942e-07,
+      "loss": 0.8412,
       "step": 32989
     },
     {
-      "epoch": 0.9060448765482958,
+      "epoch": 0.9361520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 4.5944103022356304e-07,
-      "loss": 0.8831,
+      "learning_rate": 2.1305495736300253e-07,
+      "loss": 0.8062,
       "step": 32990
     },
     {
-      "epoch": 0.9060723407761391,
+      "epoch": 0.9361804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 4.5917454346014357e-07,
-      "loss": 0.766,
+      "learning_rate": 2.1286629187762142e-07,
+      "loss": 0.9071,
       "step": 32991
     },
     {
-      "epoch": 0.9060998050039824,
+      "epoch": 0.9362088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 4.5890813218743005e-07,
-      "loss": 0.7281,
+      "learning_rate": 2.1267770906459817e-07,
+      "loss": 0.7938,
       "step": 32992
     },
     {
-      "epoch": 0.9061272692318255,
+      "epoch": 0.9362372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 4.586417964075307e-07,
-      "loss": 0.772,
+      "learning_rate": 2.1248920892552593e-07,
+      "loss": 0.8517,
       "step": 32993
     },
     {
-      "epoch": 0.9061547334596688,
+      "epoch": 0.9362656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 4.583755361225528e-07,
-      "loss": 0.851,
+      "learning_rate": 2.123007914619979e-07,
+      "loss": 0.7159,
       "step": 32994
     },
     {
-      "epoch": 0.9061821976875121,
+      "epoch": 0.9362939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 4.5810935133460575e-07,
-      "loss": 0.8142,
+      "learning_rate": 2.12112456675605e-07,
+      "loss": 0.8179,
       "step": 32995
     },
     {
-      "epoch": 0.9062096619153552,
+      "epoch": 0.936322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 4.5784324204579233e-07,
-      "loss": 0.8147,
+      "learning_rate": 2.1192420456793706e-07,
+      "loss": 0.7196,
       "step": 32996
     },
     {
-      "epoch": 0.9062371261431985,
+      "epoch": 0.9363507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 4.5757720825822083e-07,
-      "loss": 0.8444,
+      "learning_rate": 2.1173603514058728e-07,
+      "loss": 0.8204,
       "step": 32997
     },
     {
-      "epoch": 0.9062645903710417,
+      "epoch": 0.9363791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 4.57311249973994e-07,
-      "loss": 0.8999,
+      "learning_rate": 2.1154794839514215e-07,
+      "loss": 0.8634,
       "step": 32998
     },
     {
-      "epoch": 0.906292054598885,
+      "epoch": 0.9364074914869467,
       "grad_norm": 0.0,
-      "learning_rate": 4.5704536719521574e-07,
-      "loss": 0.8686,
+      "learning_rate": 2.1135994433319152e-07,
+      "loss": 0.9074,
       "step": 32999
     },
     {
-      "epoch": 0.9063195188267282,
+      "epoch": 0.9364358683314415,
       "grad_norm": 0.0,
-      "learning_rate": 4.5677955992399217e-07,
-      "loss": 0.7925,
+      "learning_rate": 2.111720229563241e-07,
+      "loss": 0.7512,
       "step": 33000
     },
     {
-      "epoch": 0.9063469830545714,
+      "epoch": 0.9364642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 4.5651382816242483e-07,
-      "loss": 0.8608,
+      "learning_rate": 2.1098418426612532e-07,
+      "loss": 0.74,
       "step": 33001
     },
     {
-      "epoch": 0.9063744472824147,
+      "epoch": 0.9364926220204314,
       "grad_norm": 0.0,
-      "learning_rate": 4.562481719126177e-07,
-      "loss": 0.7812,
+      "learning_rate": 2.107964282641839e-07,
+      "loss": 0.76,
       "step": 33002
     },
     {
-      "epoch": 0.9064019115102578,
+      "epoch": 0.9365209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 4.5598259117667245e-07,
-      "loss": 0.8519,
+      "learning_rate": 2.1060875495208633e-07,
+      "loss": 0.8963,
       "step": 33003
     },
     {
-      "epoch": 0.9064293757381011,
+      "epoch": 0.9365493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 4.557170859566906e-07,
-      "loss": 0.7089,
+      "learning_rate": 2.1042116433141468e-07,
+      "loss": 0.7406,
       "step": 33004
     },
     {
-      "epoch": 0.9064568399659444,
+      "epoch": 0.936577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.554516562547706e-07,
-      "loss": 0.8019,
+      "learning_rate": 2.102336564037566e-07,
+      "loss": 0.7314,
       "step": 33005
     },
     {
-      "epoch": 0.9064843041937876,
+      "epoch": 0.9366061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 4.5518630207301405e-07,
-      "loss": 0.8519,
+      "learning_rate": 2.1004623117069524e-07,
+      "loss": 0.8185,
       "step": 33006
     },
     {
-      "epoch": 0.9065117684216308,
+      "epoch": 0.9366345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 4.549210234135204e-07,
-      "loss": 0.858,
+      "learning_rate": 2.0985888863381264e-07,
+      "loss": 0.6895,
       "step": 33007
     },
     {
-      "epoch": 0.9065392326494741,
+      "epoch": 0.9366628830874006,
       "grad_norm": 0.0,
-      "learning_rate": 4.546558202783902e-07,
-      "loss": 0.8175,
+      "learning_rate": 2.0967162879469206e-07,
+      "loss": 0.8352,
       "step": 33008
     },
     {
-      "epoch": 0.9065666968773173,
+      "epoch": 0.9366912599318956,
       "grad_norm": 0.0,
-      "learning_rate": 4.543906926697217e-07,
-      "loss": 0.7942,
+      "learning_rate": 2.094844516549155e-07,
+      "loss": 0.7409,
       "step": 33009
     },
     {
-      "epoch": 0.9065941611051606,
+      "epoch": 0.9367196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 4.5412564058960996e-07,
-      "loss": 0.7789,
+      "learning_rate": 2.092973572160639e-07,
+      "loss": 0.9134,
       "step": 33010
     },
     {
-      "epoch": 0.9066216253330037,
+      "epoch": 0.9367480136208853,
       "grad_norm": 0.0,
-      "learning_rate": 4.5386066404015327e-07,
-      "loss": 0.7651,
+      "learning_rate": 2.091103454797183e-07,
+      "loss": 0.7525,
       "step": 33011
     },
     {
-      "epoch": 0.906649089560847,
+      "epoch": 0.9367763904653803,
       "grad_norm": 0.0,
-      "learning_rate": 4.535957630234511e-07,
-      "loss": 0.8607,
+      "learning_rate": 2.089234164474574e-07,
+      "loss": 0.8247,
       "step": 33012
     },
     {
-      "epoch": 0.9066765537886903,
+      "epoch": 0.9368047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 4.533309375415951e-07,
-      "loss": 0.6832,
+      "learning_rate": 2.087365701208599e-07,
+      "loss": 0.872,
       "step": 33013
     },
     {
-      "epoch": 0.9067040180165334,
+      "epoch": 0.93683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 4.530661875966835e-07,
-      "loss": 0.8197,
+      "learning_rate": 2.085498065015057e-07,
+      "loss": 0.8381,
       "step": 33014
     },
     {
-      "epoch": 0.9067314822443767,
+      "epoch": 0.9368615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 4.5280151319081144e-07,
-      "loss": 0.8714,
+      "learning_rate": 2.0836312559097015e-07,
+      "loss": 0.8513,
       "step": 33015
     },
     {
-      "epoch": 0.9067589464722199,
+      "epoch": 0.9368898978433599,
       "grad_norm": 0.0,
-      "learning_rate": 4.525369143260705e-07,
-      "loss": 0.7689,
+      "learning_rate": 2.0817652739083204e-07,
+      "loss": 0.7871,
       "step": 33016
     },
     {
-      "epoch": 0.9067864107000632,
+      "epoch": 0.9369182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 4.522723910045568e-07,
-      "loss": 0.795,
+      "learning_rate": 2.0799001190266788e-07,
+      "loss": 0.9548,
       "step": 33017
     },
     {
-      "epoch": 0.9068138749279064,
+      "epoch": 0.9369466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 4.520079432283608e-07,
-      "loss": 0.815,
+      "learning_rate": 2.0780357912805193e-07,
+      "loss": 0.821,
       "step": 33018
     },
     {
-      "epoch": 0.9068413391557496,
+      "epoch": 0.9369750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 4.517435709995788e-07,
-      "loss": 0.7401,
+      "learning_rate": 2.076172290685585e-07,
+      "loss": 0.8216,
       "step": 33019
     },
     {
-      "epoch": 0.9068688033835929,
+      "epoch": 0.9370034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 4.514792743202989e-07,
-      "loss": 0.8685,
+      "learning_rate": 2.0743096172576416e-07,
+      "loss": 0.8187,
       "step": 33020
     },
     {
-      "epoch": 0.9068962676114362,
+      "epoch": 0.9370317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 4.512150531926152e-07,
-      "loss": 0.8574,
+      "learning_rate": 2.072447771012387e-07,
+      "loss": 0.8598,
       "step": 33021
     },
     {
-      "epoch": 0.9069237318392793,
+      "epoch": 0.9370601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 4.509509076186147e-07,
-      "loss": 0.7443,
+      "learning_rate": 2.0705867519655753e-07,
+      "loss": 0.7502,
       "step": 33022
     },
     {
-      "epoch": 0.9069511960671226,
+      "epoch": 0.9370885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 4.5068683760039033e-07,
-      "loss": 0.8183,
+      "learning_rate": 2.0687265601329275e-07,
+      "loss": 0.8402,
       "step": 33023
     },
     {
-      "epoch": 0.9069786602949658,
+      "epoch": 0.937116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 4.504228431400304e-07,
-      "loss": 0.8367,
+      "learning_rate": 2.066867195530131e-07,
+      "loss": 0.7495,
       "step": 33024
     },
     {
-      "epoch": 0.907006124522809,
+      "epoch": 0.9371452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 4.5015892423962426e-07,
-      "loss": 0.7739,
+      "learning_rate": 2.065008658172929e-07,
+      "loss": 0.8154,
       "step": 33025
     },
     {
-      "epoch": 0.9070335887506523,
+      "epoch": 0.9371736662883088,
       "grad_norm": 0.0,
-      "learning_rate": 4.4989508090126033e-07,
-      "loss": 0.843,
+      "learning_rate": 2.0631509480769862e-07,
+      "loss": 0.8774,
       "step": 33026
     },
     {
-      "epoch": 0.9070610529784955,
+      "epoch": 0.9372020431328036,
       "grad_norm": 0.0,
-      "learning_rate": 4.496313131270269e-07,
-      "loss": 0.8474,
+      "learning_rate": 2.0612940652580126e-07,
+      "loss": 0.7939,
       "step": 33027
     },
     {
-      "epoch": 0.9070885172063388,
+      "epoch": 0.9372304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 4.493676209190101e-07,
-      "loss": 0.8397,
+      "learning_rate": 2.0594380097317068e-07,
+      "loss": 0.7669,
       "step": 33028
     },
     {
-      "epoch": 0.9071159814341819,
+      "epoch": 0.9372587968217935,
       "grad_norm": 0.0,
-      "learning_rate": 4.491040042792938e-07,
-      "loss": 0.6884,
+      "learning_rate": 2.0575827815137118e-07,
+      "loss": 0.8093,
       "step": 33029
     },
     {
-      "epoch": 0.9071434456620252,
+      "epoch": 0.9372871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 4.488404632099674e-07,
-      "loss": 0.8536,
+      "learning_rate": 2.055728380619726e-07,
+      "loss": 0.7556,
       "step": 33030
     },
     {
-      "epoch": 0.9071709098898685,
+      "epoch": 0.9373155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 4.485769977131149e-07,
-      "loss": 0.6913,
+      "learning_rate": 2.0538748070654147e-07,
+      "loss": 0.7874,
       "step": 33031
     },
     {
-      "epoch": 0.9071983741177116,
+      "epoch": 0.937343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 4.483136077908212e-07,
-      "loss": 0.7788,
+      "learning_rate": 2.0520220608664098e-07,
+      "loss": 0.7442,
       "step": 33032
     },
     {
-      "epoch": 0.9072258383455549,
+      "epoch": 0.937372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 4.4805029344517134e-07,
-      "loss": 0.801,
+      "learning_rate": 2.0501701420383878e-07,
+      "loss": 0.8784,
       "step": 33033
     },
     {
-      "epoch": 0.9072533025733982,
+      "epoch": 0.9374006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 4.477870546782459e-07,
-      "loss": 0.8303,
+      "learning_rate": 2.0483190505969808e-07,
+      "loss": 0.8355,
       "step": 33034
     },
     {
-      "epoch": 0.9072807668012414,
+      "epoch": 0.9374290578887627,
       "grad_norm": 0.0,
-      "learning_rate": 4.475238914921298e-07,
-      "loss": 0.7202,
+      "learning_rate": 2.0464687865578204e-07,
+      "loss": 0.7909,
       "step": 33035
     },
     {
-      "epoch": 0.9073082310290846,
+      "epoch": 0.9374574347332577,
       "grad_norm": 0.0,
-      "learning_rate": 4.47260803888907e-07,
-      "loss": 0.8011,
+      "learning_rate": 2.0446193499365387e-07,
+      "loss": 0.8761,
       "step": 33036
     },
     {
-      "epoch": 0.9073356952569278,
+      "epoch": 0.9374858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 4.4699779187065474e-07,
-      "loss": 0.8306,
+      "learning_rate": 2.0427707407487785e-07,
+      "loss": 0.7883,
       "step": 33037
     },
     {
-      "epoch": 0.9073631594847711,
+      "epoch": 0.9375141884222474,
       "grad_norm": 0.0,
-      "learning_rate": 4.467348554394579e-07,
-      "loss": 0.7939,
+      "learning_rate": 2.0409229590101164e-07,
+      "loss": 0.9673,
       "step": 33038
     },
     {
-      "epoch": 0.9073906237126144,
+      "epoch": 0.9375425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 4.464719945973939e-07,
-      "loss": 0.7977,
+      "learning_rate": 2.0390760047361845e-07,
+      "loss": 0.8687,
       "step": 33039
     },
     {
-      "epoch": 0.9074180879404575,
+      "epoch": 0.9375709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 4.4620920934654533e-07,
-      "loss": 0.7915,
+      "learning_rate": 2.0372298779425926e-07,
+      "loss": 0.7986,
       "step": 33040
     },
     {
-      "epoch": 0.9074455521683008,
+      "epoch": 0.9375993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 4.459464996889895e-07,
-      "loss": 0.8362,
+      "learning_rate": 2.0353845786449167e-07,
+      "loss": 0.8313,
       "step": 33041
     },
     {
-      "epoch": 0.907473016396144,
+      "epoch": 0.937627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 4.4568386562680586e-07,
-      "loss": 0.8125,
+      "learning_rate": 2.0335401068587447e-07,
+      "loss": 0.8123,
       "step": 33042
     },
     {
-      "epoch": 0.9075004806239872,
+      "epoch": 0.937656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 4.454213071620739e-07,
-      "loss": 0.8459,
+      "learning_rate": 2.0316964625996749e-07,
+      "loss": 0.8285,
       "step": 33043
     },
     {
-      "epoch": 0.9075279448518305,
+      "epoch": 0.9376844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 4.451588242968674e-07,
-      "loss": 0.9038,
+      "learning_rate": 2.0298536458832617e-07,
+      "loss": 0.7773,
       "step": 33044
     },
     {
-      "epoch": 0.9075554090796737,
+      "epoch": 0.9377128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 4.448964170332681e-07,
-      "loss": 0.8749,
+      "learning_rate": 2.0280116567250928e-07,
+      "loss": 0.8264,
       "step": 33045
     },
     {
-      "epoch": 0.907582873307517,
+      "epoch": 0.9377412031782065,
       "grad_norm": 0.0,
-      "learning_rate": 4.4463408537334774e-07,
-      "loss": 0.8902,
+      "learning_rate": 2.0261704951407003e-07,
+      "loss": 0.7963,
       "step": 33046
     },
     {
-      "epoch": 0.9076103375353602,
+      "epoch": 0.9377695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 4.4437182931918453e-07,
-      "loss": 0.8189,
+      "learning_rate": 2.024330161145649e-07,
+      "loss": 0.8385,
       "step": 33047
     },
     {
-      "epoch": 0.9076378017632034,
+      "epoch": 0.9377979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 4.4410964887285244e-07,
-      "loss": 0.8193,
+      "learning_rate": 2.0224906547554933e-07,
+      "loss": 0.8521,
       "step": 33048
     },
     {
-      "epoch": 0.9076652659910467,
+      "epoch": 0.9378263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 4.4384754403642536e-07,
-      "loss": 0.7859,
+      "learning_rate": 2.0206519759857545e-07,
+      "loss": 0.7537,
       "step": 33049
     },
     {
-      "epoch": 0.9076927302188899,
+      "epoch": 0.9378547105561862,
       "grad_norm": 0.0,
-      "learning_rate": 4.4358551481198055e-07,
-      "loss": 0.7405,
+      "learning_rate": 2.0188141248519754e-07,
+      "loss": 0.8925,
       "step": 33050
     },
     {
-      "epoch": 0.9077201944467331,
+      "epoch": 0.937883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 4.4332356120158847e-07,
-      "loss": 0.8213,
+      "learning_rate": 2.0169771013696882e-07,
+      "loss": 0.8254,
       "step": 33051
     },
     {
-      "epoch": 0.9077476586745764,
+      "epoch": 0.9379114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 4.43061683207322e-07,
-      "loss": 0.8832,
+      "learning_rate": 2.0151409055543912e-07,
+      "loss": 0.7997,
       "step": 33052
     },
     {
-      "epoch": 0.9077751229024196,
+      "epoch": 0.9379398410896709,
       "grad_norm": 0.0,
-      "learning_rate": 4.427998808312528e-07,
-      "loss": 0.7777,
+      "learning_rate": 2.013305537421606e-07,
+      "loss": 0.7889,
       "step": 33053
     },
     {
-      "epoch": 0.9078025871302629,
+      "epoch": 0.9379682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 4.4253815407545366e-07,
-      "loss": 0.7867,
+      "learning_rate": 2.0114709969868308e-07,
+      "loss": 0.8618,
       "step": 33054
     },
     {
-      "epoch": 0.907830051358106,
+      "epoch": 0.9379965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 4.42276502941994e-07,
-      "loss": 0.8648,
+      "learning_rate": 2.0096372842655641e-07,
+      "loss": 0.7532,
       "step": 33055
     },
     {
-      "epoch": 0.9078575155859493,
+      "epoch": 0.9380249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 4.420149274329466e-07,
-      "loss": 0.8051,
+      "learning_rate": 2.0078043992732943e-07,
+      "loss": 0.8035,
       "step": 33056
     },
     {
-      "epoch": 0.9078849798137926,
+      "epoch": 0.9380533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 4.4175342755037875e-07,
-      "loss": 0.8859,
+      "learning_rate": 2.0059723420255084e-07,
+      "loss": 0.8732,
       "step": 33057
     },
     {
-      "epoch": 0.9079124440416357,
+      "epoch": 0.9380817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 4.4149200329635875e-07,
-      "loss": 0.7865,
+      "learning_rate": 2.004141112537683e-07,
+      "loss": 0.8365,
       "step": 33058
     },
     {
-      "epoch": 0.907939908269479,
+      "epoch": 0.9381101021566401,
       "grad_norm": 0.0,
-      "learning_rate": 4.4123065467295834e-07,
-      "loss": 0.8382,
+      "learning_rate": 2.0023107108252836e-07,
+      "loss": 0.9015,
       "step": 33059
     },
     {
-      "epoch": 0.9079673724973223,
+      "epoch": 0.9381384790011351,
       "grad_norm": 0.0,
-      "learning_rate": 4.409693816822436e-07,
-      "loss": 0.7537,
+      "learning_rate": 2.0004811369037646e-07,
+      "loss": 0.8276,
       "step": 33060
     },
     {
-      "epoch": 0.9079948367251655,
+      "epoch": 0.93816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.407081843262817e-07,
-      "loss": 0.8171,
+      "learning_rate": 1.9986523907885913e-07,
+      "loss": 0.8515,
       "step": 33061
     },
     {
-      "epoch": 0.9080223009530087,
+      "epoch": 0.9381952326901248,
       "grad_norm": 0.0,
-      "learning_rate": 4.404470626071411e-07,
-      "loss": 0.957,
+      "learning_rate": 1.996824472495207e-07,
+      "loss": 0.7366,
       "step": 33062
     },
     {
-      "epoch": 0.9080497651808519,
+      "epoch": 0.9382236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 4.401860165268834e-07,
-      "loss": 0.9104,
+      "learning_rate": 1.9949973820390546e-07,
+      "loss": 0.8481,
       "step": 33063
     },
     {
-      "epoch": 0.9080772294086952,
+      "epoch": 0.9382519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 4.3992504608757815e-07,
-      "loss": 0.8358,
+      "learning_rate": 1.9931711194355552e-07,
+      "loss": 0.7078,
       "step": 33064
     },
     {
-      "epoch": 0.9081046936365385,
+      "epoch": 0.9382803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 4.3966415129128916e-07,
-      "loss": 0.8329,
+      "learning_rate": 1.9913456847001523e-07,
+      "loss": 0.7839,
       "step": 33065
     },
     {
-      "epoch": 0.9081321578643816,
+      "epoch": 0.9383087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 4.3940333214008146e-07,
-      "loss": 0.8754,
+      "learning_rate": 1.9895210778482665e-07,
+      "loss": 0.7627,
       "step": 33066
     },
     {
-      "epoch": 0.9081596220922249,
+      "epoch": 0.9383371169125994,
       "grad_norm": 0.0,
-      "learning_rate": 4.391425886360179e-07,
-      "loss": 0.7495,
+      "learning_rate": 1.9876972988952859e-07,
+      "loss": 0.8067,
       "step": 33067
     },
     {
-      "epoch": 0.9081870863200681,
+      "epoch": 0.9383654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 4.388819207811612e-07,
-      "loss": 0.7951,
+      "learning_rate": 1.9858743478566312e-07,
+      "loss": 0.8969,
       "step": 33068
     },
     {
-      "epoch": 0.9082145505479113,
+      "epoch": 0.9383938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 4.3862132857757535e-07,
-      "loss": 0.8195,
+      "learning_rate": 1.9840522247477234e-07,
+      "loss": 0.6472,
       "step": 33069
     },
     {
-      "epoch": 0.9082420147757546,
+      "epoch": 0.938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 4.383608120273197e-07,
-      "loss": 0.7784,
+      "learning_rate": 1.982230929583906e-07,
+      "loss": 0.7929,
       "step": 33070
     },
     {
-      "epoch": 0.9082694790035978,
+      "epoch": 0.9384506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 4.381003711324583e-07,
-      "loss": 0.8321,
+      "learning_rate": 1.9804104623806107e-07,
+      "loss": 0.7639,
       "step": 33071
     },
     {
-      "epoch": 0.9082969432314411,
+      "epoch": 0.9384790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 4.3784000589504937e-07,
-      "loss": 0.8494,
+      "learning_rate": 1.9785908231531813e-07,
+      "loss": 0.7854,
       "step": 33072
     },
     {
-      "epoch": 0.9083244074592843,
+      "epoch": 0.9385073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 4.3757971631715466e-07,
-      "loss": 0.7861,
+      "learning_rate": 1.9767720119170054e-07,
+      "loss": 0.8153,
       "step": 33073
     },
     {
-      "epoch": 0.9083518716871275,
+      "epoch": 0.9385357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 4.373195024008359e-07,
-      "loss": 0.7941,
+      "learning_rate": 1.974954028687448e-07,
+      "loss": 0.856,
       "step": 33074
     },
     {
-      "epoch": 0.9083793359149708,
+      "epoch": 0.9385641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 4.3705936414814685e-07,
-      "loss": 0.788,
+      "learning_rate": 1.973136873479853e-07,
+      "loss": 0.8682,
       "step": 33075
     },
     {
-      "epoch": 0.9084068001428139,
+      "epoch": 0.9385925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 4.367993015611494e-07,
-      "loss": 0.807,
+      "learning_rate": 1.9713205463095742e-07,
+      "loss": 0.7968,
       "step": 33076
     },
     {
-      "epoch": 0.9084342643706572,
+      "epoch": 0.9386208853575483,
       "grad_norm": 0.0,
-      "learning_rate": 4.365393146419006e-07,
-      "loss": 0.7967,
+      "learning_rate": 1.9695050471919773e-07,
+      "loss": 0.7286,
       "step": 33077
     },
     {
-      "epoch": 0.9084617285985005,
+      "epoch": 0.9386492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 4.3627940339245556e-07,
-      "loss": 0.8257,
+      "learning_rate": 1.9676903761423617e-07,
+      "loss": 0.8115,
       "step": 33078
     },
     {
-      "epoch": 0.9084891928263437,
+      "epoch": 0.938677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 4.360195678148738e-07,
-      "loss": 0.7802,
+      "learning_rate": 1.96587653317607e-07,
+      "loss": 0.731,
       "step": 33079
     },
     {
-      "epoch": 0.9085166570541869,
+      "epoch": 0.9387060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 4.357598079112102e-07,
-      "loss": 0.7893,
+      "learning_rate": 1.9640635183084344e-07,
+      "loss": 0.8268,
       "step": 33080
     },
     {
-      "epoch": 0.9085441212820301,
+      "epoch": 0.9387343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 4.355001236835188e-07,
-      "loss": 0.8468,
+      "learning_rate": 1.9622513315547653e-07,
+      "loss": 0.7808,
       "step": 33081
     },
     {
-      "epoch": 0.9085715855098734,
+      "epoch": 0.9387627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 4.352405151338557e-07,
-      "loss": 0.7366,
+      "learning_rate": 1.9604399729303504e-07,
+      "loss": 0.7634,
       "step": 33082
     },
     {
-      "epoch": 0.9085990497377167,
+      "epoch": 0.9387911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 4.3498098226427477e-07,
-      "loss": 0.597,
+      "learning_rate": 1.9586294424505216e-07,
+      "loss": 0.8464,
       "step": 33083
     },
     {
-      "epoch": 0.9086265139655598,
+      "epoch": 0.9388195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 4.3472152507683107e-07,
-      "loss": 0.7289,
+      "learning_rate": 1.9568197401305556e-07,
+      "loss": 0.7076,
       "step": 33084
     },
     {
-      "epoch": 0.9086539781934031,
+      "epoch": 0.9388479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 4.3446214357357406e-07,
-      "loss": 0.8333,
+      "learning_rate": 1.9550108659857292e-07,
+      "loss": 0.8863,
       "step": 33085
     },
     {
-      "epoch": 0.9086814424212464,
+      "epoch": 0.9388762769580022,
       "grad_norm": 0.0,
-      "learning_rate": 4.342028377565588e-07,
-      "loss": 0.789,
+      "learning_rate": 1.9532028200313412e-07,
+      "loss": 0.8336,
       "step": 33086
     },
     {
-      "epoch": 0.9087089066490895,
+      "epoch": 0.9389046538024972,
       "grad_norm": 0.0,
-      "learning_rate": 4.3394360762783584e-07,
-      "loss": 0.8157,
+      "learning_rate": 1.9513956022826464e-07,
+      "loss": 0.826,
       "step": 33087
     },
     {
-      "epoch": 0.9087363708769328,
+      "epoch": 0.9389330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 4.336844531894557e-07,
-      "loss": 0.7639,
+      "learning_rate": 1.949589212754943e-07,
+      "loss": 0.8357,
       "step": 33088
     },
     {
-      "epoch": 0.908763835104776,
+      "epoch": 0.9389614074914869,
       "grad_norm": 0.0,
-      "learning_rate": 4.334253744434702e-07,
-      "loss": 0.8528,
+      "learning_rate": 1.947783651463453e-07,
+      "loss": 0.7709,
       "step": 33089
     },
     {
-      "epoch": 0.9087912993326193,
+      "epoch": 0.9389897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 4.3316637139192873e-07,
-      "loss": 0.863,
+      "learning_rate": 1.9459789184234413e-07,
+      "loss": 0.7791,
       "step": 33090
     },
     {
-      "epoch": 0.9088187635604625,
+      "epoch": 0.9390181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 4.329074440368819e-07,
-      "loss": 0.84,
+      "learning_rate": 1.9441750136501514e-07,
+      "loss": 0.7418,
       "step": 33091
     },
     {
-      "epoch": 0.9088462277883057,
+      "epoch": 0.9390465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 4.3264859238037584e-07,
-      "loss": 0.8047,
+      "learning_rate": 1.9423719371588269e-07,
+      "loss": 0.7642,
       "step": 33092
     },
     {
-      "epoch": 0.908873692016149,
+      "epoch": 0.9390749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 4.3238981642446223e-07,
-      "loss": 0.77,
+      "learning_rate": 1.9405696889646996e-07,
+      "loss": 0.8052,
       "step": 33093
     },
     {
-      "epoch": 0.9089011562439921,
+      "epoch": 0.9391032917139615,
       "grad_norm": 0.0,
-      "learning_rate": 4.3213111617118495e-07,
-      "loss": 0.8202,
+      "learning_rate": 1.9387682690829912e-07,
+      "loss": 0.8882,
       "step": 33094
     },
     {
-      "epoch": 0.9089286204718354,
+      "epoch": 0.9391316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.318724916225925e-07,
-      "loss": 0.7661,
+      "learning_rate": 1.9369676775289005e-07,
+      "loss": 0.7984,
       "step": 33095
     },
     {
-      "epoch": 0.9089560846996787,
+      "epoch": 0.9391600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 4.3161394278073085e-07,
-      "loss": 0.8349,
+      "learning_rate": 1.9351679143176595e-07,
+      "loss": 0.785,
       "step": 33096
     },
     {
-      "epoch": 0.9089835489275219,
+      "epoch": 0.939188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 4.313554696476474e-07,
-      "loss": 0.9126,
+      "learning_rate": 1.9333689794644672e-07,
+      "loss": 0.8288,
       "step": 33097
     },
     {
-      "epoch": 0.9090110131553651,
+      "epoch": 0.939216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 4.31097072225386e-07,
-      "loss": 0.8974,
+      "learning_rate": 1.9315708729845117e-07,
+      "loss": 0.7253,
       "step": 33098
     },
     {
-      "epoch": 0.9090384773832084,
+      "epoch": 0.9392451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 4.308387505159894e-07,
-      "loss": 0.7564,
+      "learning_rate": 1.9297735948929808e-07,
+      "loss": 0.6903,
       "step": 33099
     },
     {
-      "epoch": 0.9090659416110516,
+      "epoch": 0.9392735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 4.305805045215061e-07,
-      "loss": 0.8463,
+      "learning_rate": 1.9279771452050622e-07,
+      "loss": 0.7615,
       "step": 33100
     },
     {
-      "epoch": 0.9090934058388949,
+      "epoch": 0.9393019296254257,
       "grad_norm": 0.0,
-      "learning_rate": 4.303223342439744e-07,
-      "loss": 0.7864,
+      "learning_rate": 1.9261815239359327e-07,
+      "loss": 0.8251,
       "step": 33101
     },
     {
-      "epoch": 0.909120870066738,
+      "epoch": 0.9393303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 4.3006423968543934e-07,
-      "loss": 0.7711,
+      "learning_rate": 1.924386731100758e-07,
+      "loss": 0.7294,
       "step": 33102
     },
     {
-      "epoch": 0.9091483342945813,
+      "epoch": 0.9393586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 4.298062208479448e-07,
-      "loss": 0.7979,
+      "learning_rate": 1.9225927667146927e-07,
+      "loss": 0.7933,
       "step": 33103
     },
     {
-      "epoch": 0.9091757985224246,
+      "epoch": 0.9393870601589104,
       "grad_norm": 0.0,
-      "learning_rate": 4.2954827773352913e-07,
-      "loss": 0.8372,
+      "learning_rate": 1.9207996307929023e-07,
+      "loss": 0.8586,
       "step": 33104
     },
     {
-      "epoch": 0.9092032627502677,
+      "epoch": 0.9394154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 4.2929041034423525e-07,
-      "loss": 0.8386,
+      "learning_rate": 1.9190073233505301e-07,
+      "loss": 0.7478,
       "step": 33105
     },
     {
-      "epoch": 0.909230726978111,
+      "epoch": 0.9394438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 4.2903261868210255e-07,
-      "loss": 0.7968,
+      "learning_rate": 1.9172158444026978e-07,
+      "loss": 0.7522,
       "step": 33106
     },
     {
-      "epoch": 0.9092581912059542,
+      "epoch": 0.939472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 4.2877490274917167e-07,
-      "loss": 0.7109,
+      "learning_rate": 1.9154251939645595e-07,
+      "loss": 0.8081,
       "step": 33107
     },
     {
-      "epoch": 0.9092856554337975,
+      "epoch": 0.9395005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 4.2851726254748207e-07,
-      "loss": 0.8087,
+      "learning_rate": 1.9136353720512367e-07,
+      "loss": 0.8362,
       "step": 33108
     },
     {
-      "epoch": 0.9093131196616407,
+      "epoch": 0.9395289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 4.2825969807907096e-07,
-      "loss": 0.7354,
+      "learning_rate": 1.9118463786778397e-07,
+      "loss": 0.8692,
       "step": 33109
     },
     {
-      "epoch": 0.9093405838894839,
+      "epoch": 0.9395573212258796,
       "grad_norm": 0.0,
-      "learning_rate": 4.280022093459768e-07,
-      "loss": 0.8149,
+      "learning_rate": 1.9100582138594892e-07,
+      "loss": 0.8029,
       "step": 33110
     },
     {
-      "epoch": 0.9093680481173272,
+      "epoch": 0.9395856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 4.277447963502357e-07,
-      "loss": 0.7858,
+      "learning_rate": 1.9082708776112847e-07,
+      "loss": 0.8332,
       "step": 33111
     },
     {
-      "epoch": 0.9093955123451705,
+      "epoch": 0.9396140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 4.27487459093886e-07,
-      "loss": 0.7937,
+      "learning_rate": 1.906484369948325e-07,
+      "loss": 0.7782,
       "step": 33112
     },
     {
-      "epoch": 0.9094229765730136,
+      "epoch": 0.9396424517593643,
       "grad_norm": 0.0,
-      "learning_rate": 4.27230197578965e-07,
-      "loss": 0.7564,
+      "learning_rate": 1.904698690885698e-07,
+      "loss": 0.8718,
       "step": 33113
     },
     {
-      "epoch": 0.9094504408008569,
+      "epoch": 0.9396708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 4.269730118075055e-07,
-      "loss": 0.7952,
+      "learning_rate": 1.9029138404385027e-07,
+      "loss": 0.8143,
       "step": 33114
     },
     {
-      "epoch": 0.9094779050287001,
+      "epoch": 0.9396992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 4.2671590178154476e-07,
-      "loss": 0.7706,
+      "learning_rate": 1.9011298186217942e-07,
+      "loss": 0.748,
       "step": 33115
     },
     {
-      "epoch": 0.9095053692565434,
+      "epoch": 0.939727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 4.264588675031167e-07,
-      "loss": 0.756,
+      "learning_rate": 1.8993466254506489e-07,
+      "loss": 0.8293,
       "step": 33116
     },
     {
-      "epoch": 0.9095328334843866,
+      "epoch": 0.9397559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 4.2620190897425304e-07,
-      "loss": 0.8552,
+      "learning_rate": 1.8975642609401434e-07,
+      "loss": 0.8478,
       "step": 33117
     },
     {
-      "epoch": 0.9095602977122298,
+      "epoch": 0.9397843359818389,
       "grad_norm": 0.0,
-      "learning_rate": 4.259450261969889e-07,
-      "loss": 0.8325,
+      "learning_rate": 1.8957827251053107e-07,
+      "loss": 0.7706,
       "step": 33118
     },
     {
-      "epoch": 0.9095877619400731,
+      "epoch": 0.9398127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 4.2568821917335577e-07,
-      "loss": 0.8573,
+      "learning_rate": 1.8940020179612272e-07,
+      "loss": 0.7596,
       "step": 33119
     },
     {
-      "epoch": 0.9096152261679162,
+      "epoch": 0.9398410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 4.254314879053867e-07,
-      "loss": 0.7423,
+      "learning_rate": 1.8922221395229035e-07,
+      "loss": 0.7641,
       "step": 33120
     },
     {
-      "epoch": 0.9096426903957595,
+      "epoch": 0.9398694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 4.251748323951133e-07,
-      "loss": 0.8113,
+      "learning_rate": 1.8904430898053938e-07,
+      "loss": 0.8048,
       "step": 33121
     },
     {
-      "epoch": 0.9096701546236028,
+      "epoch": 0.9398978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 4.24918252644565e-07,
-      "loss": 0.8557,
+      "learning_rate": 1.8886648688237308e-07,
+      "loss": 0.787,
       "step": 33122
     },
     {
-      "epoch": 0.909697618851446,
+      "epoch": 0.9399262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 4.2466174865577135e-07,
-      "loss": 0.7882,
+      "learning_rate": 1.8868874765929136e-07,
+      "loss": 0.9207,
       "step": 33123
     },
     {
-      "epoch": 0.9097250830792892,
+      "epoch": 0.9399545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 4.244053204307652e-07,
-      "loss": 0.725,
+      "learning_rate": 1.8851109131279743e-07,
+      "loss": 0.8439,
       "step": 33124
     },
     {
-      "epoch": 0.9097525473071325,
+      "epoch": 0.9399829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 4.241489679715727e-07,
-      "loss": 0.8575,
+      "learning_rate": 1.8833351784439237e-07,
+      "loss": 0.8546,
       "step": 33125
     },
     {
-      "epoch": 0.9097800115349757,
+      "epoch": 0.940011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 4.2389269128022215e-07,
-      "loss": 0.6984,
+      "learning_rate": 1.881560272555738e-07,
+      "loss": 0.821,
       "step": 33126
     },
     {
-      "epoch": 0.909807475762819,
+      "epoch": 0.9400397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 4.236364903587431e-07,
-      "loss": 0.6888,
+      "learning_rate": 1.8797861954784168e-07,
+      "loss": 0.7833,
       "step": 33127
     },
     {
-      "epoch": 0.9098349399906621,
+      "epoch": 0.9400681044267878,
       "grad_norm": 0.0,
-      "learning_rate": 4.2338036520916057e-07,
-      "loss": 0.7865,
+      "learning_rate": 1.8780129472269704e-07,
+      "loss": 0.7777,
       "step": 33128
     },
     {
-      "epoch": 0.9098624042185054,
+      "epoch": 0.9400964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 4.231243158335019e-07,
-      "loss": 0.7457,
+      "learning_rate": 1.8762405278163532e-07,
+      "loss": 0.916,
       "step": 33129
     },
     {
-      "epoch": 0.9098898684463487,
+      "epoch": 0.9401248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 4.2286834223379427e-07,
-      "loss": 0.8445,
+      "learning_rate": 1.874468937261531e-07,
+      "loss": 0.8195,
       "step": 33130
     },
     {
-      "epoch": 0.9099173326741918,
+      "epoch": 0.9401532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 4.226124444120616e-07,
-      "loss": 0.8354,
+      "learning_rate": 1.872698175577492e-07,
+      "loss": 0.8815,
       "step": 33131
     },
     {
-      "epoch": 0.9099447969020351,
+      "epoch": 0.9401816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 4.2235662237033017e-07,
-      "loss": 0.8602,
+      "learning_rate": 1.8709282427791799e-07,
+      "loss": 0.8042,
       "step": 33132
     },
     {
-      "epoch": 0.9099722611298783,
+      "epoch": 0.9402099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 4.221008761106238e-07,
-      "loss": 0.7709,
+      "learning_rate": 1.86915913888156e-07,
+      "loss": 0.8276,
       "step": 33133
     },
     {
-      "epoch": 0.9099997253577216,
+      "epoch": 0.9402383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 4.218452056349631e-07,
-      "loss": 0.814,
+      "learning_rate": 1.867390863899543e-07,
+      "loss": 0.7281,
       "step": 33134
     },
     {
-      "epoch": 0.9100271895855648,
+      "epoch": 0.940266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 4.2158961094537434e-07,
-      "loss": 0.8271,
+      "learning_rate": 1.8656234178480948e-07,
+      "loss": 0.8436,
       "step": 33135
     },
     {
-      "epoch": 0.910054653813408,
+      "epoch": 0.9402951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 4.213340920438791e-07,
-      "loss": 0.7974,
+      "learning_rate": 1.8638568007421477e-07,
+      "loss": 0.7695,
       "step": 33136
     },
     {
-      "epoch": 0.9100821180412513,
+      "epoch": 0.9403234960272417,
       "grad_norm": 0.0,
-      "learning_rate": 4.2107864893249807e-07,
-      "loss": 0.8539,
+      "learning_rate": 1.8620910125966009e-07,
+      "loss": 0.8642,
       "step": 33137
     },
     {
-      "epoch": 0.9101095822690946,
+      "epoch": 0.9403518728717367,
       "grad_norm": 0.0,
-      "learning_rate": 4.2082328161325405e-07,
-      "loss": 0.7047,
+      "learning_rate": 1.8603260534263868e-07,
+      "loss": 0.8331,
       "step": 33138
     },
     {
-      "epoch": 0.9101370464969377,
+      "epoch": 0.9403802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 4.2056799008816763e-07,
-      "loss": 0.8015,
+      "learning_rate": 1.858561923246427e-07,
+      "loss": 0.8297,
       "step": 33139
     },
     {
-      "epoch": 0.910164510724781,
+      "epoch": 0.9404086265607264,
       "grad_norm": 0.0,
-      "learning_rate": 4.203127743592583e-07,
-      "loss": 0.75,
+      "learning_rate": 1.8567986220715872e-07,
+      "loss": 0.7652,
       "step": 33140
     },
     {
-      "epoch": 0.9101919749526242,
+      "epoch": 0.9404370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 4.200576344285434e-07,
-      "loss": 0.7513,
+      "learning_rate": 1.855036149916778e-07,
+      "loss": 0.8636,
       "step": 33141
     },
     {
-      "epoch": 0.9102194391804674,
+      "epoch": 0.9404653802497163,
       "grad_norm": 0.0,
-      "learning_rate": 4.198025702980435e-07,
-      "loss": 0.8284,
+      "learning_rate": 1.8532745067968983e-07,
+      "loss": 0.7744,
       "step": 33142
     },
     {
-      "epoch": 0.9102469034083107,
+      "epoch": 0.9404937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 4.1954758196977695e-07,
-      "loss": 0.8032,
+      "learning_rate": 1.8515136927268251e-07,
+      "loss": 0.7218,
       "step": 33143
     },
     {
-      "epoch": 0.9102743676361539,
+      "epoch": 0.940522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 4.192926694457611e-07,
-      "loss": 0.8141,
+      "learning_rate": 1.8497537077214133e-07,
+      "loss": 0.8696,
       "step": 33144
     },
     {
-      "epoch": 0.9103018318639972,
+      "epoch": 0.940550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 4.190378327280142e-07,
-      "loss": 0.8928,
+      "learning_rate": 1.847994551795562e-07,
+      "loss": 0.8137,
       "step": 33145
     },
     {
-      "epoch": 0.9103292960918403,
+      "epoch": 0.9405788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 4.1878307181854926e-07,
-      "loss": 0.8084,
+      "learning_rate": 1.846236224964093e-07,
+      "loss": 0.8038,
       "step": 33146
     },
     {
-      "epoch": 0.9103567603196836,
+      "epoch": 0.9406072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 4.1852838671938455e-07,
-      "loss": 0.7974,
+      "learning_rate": 1.844478727241894e-07,
+      "loss": 0.7418,
       "step": 33147
     },
     {
-      "epoch": 0.9103842245475269,
+      "epoch": 0.9406356413166855,
       "grad_norm": 0.0,
-      "learning_rate": 4.1827377743253625e-07,
-      "loss": 0.8484,
+      "learning_rate": 1.8427220586437978e-07,
+      "loss": 0.802,
       "step": 33148
     },
     {
-      "epoch": 0.91041168877537,
+      "epoch": 0.9406640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 4.180192439600161e-07,
-      "loss": 0.9082,
+      "learning_rate": 1.8409662191846256e-07,
+      "loss": 0.8068,
       "step": 33149
     },
     {
-      "epoch": 0.9104391530032133,
+      "epoch": 0.9406923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 4.177647863038392e-07,
-      "loss": 0.7579,
+      "learning_rate": 1.8392112088792323e-07,
+      "loss": 0.8848,
       "step": 33150
     },
     {
-      "epoch": 0.9104666172310566,
+      "epoch": 0.9407207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 4.1751040446602055e-07,
-      "loss": 0.8275,
+      "learning_rate": 1.8374570277424398e-07,
+      "loss": 0.8259,
       "step": 33151
     },
     {
-      "epoch": 0.9104940814588998,
+      "epoch": 0.9407491486946652,
       "grad_norm": 0.0,
-      "learning_rate": 4.1725609844857075e-07,
-      "loss": 0.9137,
+      "learning_rate": 1.8357036757890578e-07,
+      "loss": 0.7324,
       "step": 33152
     },
     {
-      "epoch": 0.910521545686743,
+      "epoch": 0.94077752553916,
       "grad_norm": 0.0,
-      "learning_rate": 4.1700186825350154e-07,
-      "loss": 0.7287,
+      "learning_rate": 1.8339511530339082e-07,
+      "loss": 0.8357,
       "step": 33153
     },
     {
-      "epoch": 0.9105490099145862,
+      "epoch": 0.9408059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 4.1674771388282574e-07,
-      "loss": 0.8428,
+      "learning_rate": 1.832199459491768e-07,
+      "loss": 0.8615,
       "step": 33154
     },
     {
-      "epoch": 0.9105764741424295,
+      "epoch": 0.9408342792281499,
       "grad_norm": 0.0,
-      "learning_rate": 4.1649363533855626e-07,
-      "loss": 0.7899,
+      "learning_rate": 1.8304485951774587e-07,
+      "loss": 0.7132,
       "step": 33155
     },
     {
-      "epoch": 0.9106039383702728,
+      "epoch": 0.9408626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 4.162396326227003e-07,
-      "loss": 0.8068,
+      "learning_rate": 1.8286985601057682e-07,
+      "loss": 0.7878,
       "step": 33156
     },
     {
-      "epoch": 0.9106314025981159,
+      "epoch": 0.9408910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 4.159857057372696e-07,
-      "loss": 0.7817,
+      "learning_rate": 1.826949354291474e-07,
+      "loss": 0.6996,
       "step": 33157
     },
     {
-      "epoch": 0.9106588668259592,
+      "epoch": 0.9409194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 4.1573185468427036e-07,
-      "loss": 0.7842,
+      "learning_rate": 1.8252009777493417e-07,
+      "loss": 0.7994,
       "step": 33158
     },
     {
-      "epoch": 0.9106863310538024,
+      "epoch": 0.9409477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 4.1547807946571426e-07,
-      "loss": 0.8057,
+      "learning_rate": 1.82345343049416e-07,
+      "loss": 0.7746,
       "step": 33159
     },
     {
-      "epoch": 0.9107137952816456,
+      "epoch": 0.9409761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 4.1522438008360865e-07,
-      "loss": 0.7536,
+      "learning_rate": 1.821706712540683e-07,
+      "loss": 0.8613,
       "step": 33160
     },
     {
-      "epoch": 0.9107412595094889,
+      "epoch": 0.9410045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 4.149707565399597e-07,
-      "loss": 0.8075,
+      "learning_rate": 1.8199608239036547e-07,
+      "loss": 0.772,
       "step": 33161
     },
     {
-      "epoch": 0.9107687237373321,
+      "epoch": 0.9410329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 4.147172088367768e-07,
-      "loss": 0.8169,
+      "learning_rate": 1.8182157645978303e-07,
+      "loss": 0.7902,
       "step": 33162
     },
     {
-      "epoch": 0.9107961879651754,
+      "epoch": 0.941061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 4.144637369760629e-07,
-      "loss": 0.8229,
+      "learning_rate": 1.8164715346379536e-07,
+      "loss": 0.861,
       "step": 33163
     },
     {
-      "epoch": 0.9108236521930186,
+      "epoch": 0.9410896708286038,
       "grad_norm": 0.0,
-      "learning_rate": 4.1421034095982636e-07,
-      "loss": 0.8281,
+      "learning_rate": 1.8147281340387458e-07,
+      "loss": 0.8066,
       "step": 33164
     },
     {
-      "epoch": 0.9108511164208618,
+      "epoch": 0.9411180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 4.1395702079006895e-07,
-      "loss": 0.7753,
+      "learning_rate": 1.8129855628149617e-07,
+      "loss": 0.7899,
       "step": 33165
     },
     {
-      "epoch": 0.9108785806487051,
+      "epoch": 0.9411464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 4.137037764687979e-07,
-      "loss": 0.728,
+      "learning_rate": 1.81124382098129e-07,
+      "loss": 0.8339,
       "step": 33166
     },
     {
-      "epoch": 0.9109060448765482,
+      "epoch": 0.9411748013620885,
       "grad_norm": 0.0,
-      "learning_rate": 4.1345060799801495e-07,
-      "loss": 0.7648,
+      "learning_rate": 1.809502908552463e-07,
+      "loss": 0.8311,
       "step": 33167
     },
     {
-      "epoch": 0.9109335091043915,
+      "epoch": 0.9412031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 4.131975153797263e-07,
-      "loss": 0.8167,
+      "learning_rate": 1.80776282554318e-07,
+      "loss": 0.8627,
       "step": 33168
     },
     {
-      "epoch": 0.9109609733322348,
+      "epoch": 0.9412315550510784,
       "grad_norm": 0.0,
-      "learning_rate": 4.1294449861593033e-07,
-      "loss": 0.8283,
+      "learning_rate": 1.8060235719681297e-07,
+      "loss": 0.8453,
       "step": 33169
     },
     {
-      "epoch": 0.910988437560078,
+      "epoch": 0.9412599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 4.126915577086321e-07,
-      "loss": 0.8859,
+      "learning_rate": 1.804285147842011e-07,
+      "loss": 0.8632,
       "step": 33170
     },
     {
-      "epoch": 0.9110159017879212,
+      "epoch": 0.9412883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 4.124386926598312e-07,
-      "loss": 0.8087,
+      "learning_rate": 1.8025475531795233e-07,
+      "loss": 0.8571,
       "step": 33171
     },
     {
-      "epoch": 0.9110433660157645,
+      "epoch": 0.941316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 4.121859034715303e-07,
-      "loss": 0.81,
+      "learning_rate": 1.800810787995322e-07,
+      "loss": 0.7697,
       "step": 33172
     },
     {
-      "epoch": 0.9110708302436077,
+      "epoch": 0.9413450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 4.11933190145728e-07,
-      "loss": 0.7409,
+      "learning_rate": 1.7990748523040835e-07,
+      "loss": 0.7479,
       "step": 33173
     },
     {
-      "epoch": 0.911098294471451,
+      "epoch": 0.9413734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 4.1168055268442584e-07,
-      "loss": 0.9353,
+      "learning_rate": 1.7973397461204856e-07,
+      "loss": 0.8195,
       "step": 33174
     },
     {
-      "epoch": 0.9111257586992941,
+      "epoch": 0.9414018161180476,
       "grad_norm": 0.0,
-      "learning_rate": 4.1142799108961904e-07,
-      "loss": 0.7937,
+      "learning_rate": 1.7956054694591607e-07,
+      "loss": 0.7347,
       "step": 33175
     },
     {
-      "epoch": 0.9111532229271374,
+      "epoch": 0.9414301929625426,
       "grad_norm": 0.0,
-      "learning_rate": 4.111755053633093e-07,
-      "loss": 0.7572,
+      "learning_rate": 1.793872022334775e-07,
+      "loss": 0.8239,
       "step": 33176
     },
     {
-      "epoch": 0.9111806871549807,
+      "epoch": 0.9414585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 4.109230955074928e-07,
-      "loss": 0.678,
+      "learning_rate": 1.7921394047619613e-07,
+      "loss": 0.8098,
       "step": 33177
     },
     {
-      "epoch": 0.9112081513828238,
+      "epoch": 0.9414869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 4.1067076152416783e-07,
-      "loss": 0.8951,
+      "learning_rate": 1.790407616755363e-07,
+      "loss": 0.8474,
       "step": 33178
     },
     {
-      "epoch": 0.9112356156106671,
+      "epoch": 0.9415153234960273,
       "grad_norm": 0.0,
-      "learning_rate": 4.1041850341533076e-07,
-      "loss": 0.8561,
+      "learning_rate": 1.7886766583296245e-07,
+      "loss": 0.8081,
       "step": 33179
     },
     {
-      "epoch": 0.9112630798385103,
+      "epoch": 0.9415437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 4.101663211829765e-07,
-      "loss": 0.7702,
+      "learning_rate": 1.7869465294993338e-07,
+      "loss": 0.812,
       "step": 33180
     },
     {
-      "epoch": 0.9112905440663536,
+      "epoch": 0.941572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 4.099142148291013e-07,
-      "loss": 0.7596,
+      "learning_rate": 1.7852172302791126e-07,
+      "loss": 0.797,
       "step": 33181
     },
     {
-      "epoch": 0.9113180082941968,
+      "epoch": 0.9416004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 4.0966218435569916e-07,
-      "loss": 0.7932,
+      "learning_rate": 1.7834887606835938e-07,
+      "loss": 0.7562,
       "step": 33182
     },
     {
-      "epoch": 0.91134547252204,
+      "epoch": 0.9416288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 4.094102297647651e-07,
-      "loss": 0.9008,
+      "learning_rate": 1.7817611207273544e-07,
+      "loss": 0.7178,
       "step": 33183
     },
     {
-      "epoch": 0.9113729367498833,
+      "epoch": 0.9416572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 4.0915835105829195e-07,
-      "loss": 0.8211,
+      "learning_rate": 1.7800343104249827e-07,
+      "loss": 0.7081,
       "step": 33184
     },
     {
-      "epoch": 0.9114004009777266,
+      "epoch": 0.9416855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 4.089065482382726e-07,
-      "loss": 0.7991,
+      "learning_rate": 1.7783083297911008e-07,
+      "loss": 0.8269,
       "step": 33185
     },
     {
-      "epoch": 0.9114278652055697,
+      "epoch": 0.9417139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 4.0865482130670207e-07,
-      "loss": 0.8108,
+      "learning_rate": 1.776583178840241e-07,
+      "loss": 0.8,
       "step": 33186
     },
     {
-      "epoch": 0.911455329433413,
+      "epoch": 0.9417423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 4.0840317026556774e-07,
-      "loss": 0.9044,
+      "learning_rate": 1.7748588575870029e-07,
+      "loss": 0.7927,
       "step": 33187
     },
     {
-      "epoch": 0.9114827936612562,
+      "epoch": 0.9417707150964812,
       "grad_norm": 0.0,
-      "learning_rate": 4.0815159511686463e-07,
-      "loss": 0.7704,
+      "learning_rate": 1.7731353660459637e-07,
+      "loss": 0.7553,
       "step": 33188
     },
     {
-      "epoch": 0.9115102578890995,
+      "epoch": 0.9417990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 4.0790009586258005e-07,
-      "loss": 0.8243,
+      "learning_rate": 1.7714127042316454e-07,
+      "loss": 0.8952,
       "step": 33189
     },
     {
-      "epoch": 0.9115377221169427,
+      "epoch": 0.9418274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 4.0764867250470683e-07,
-      "loss": 0.7909,
+      "learning_rate": 1.7696908721586248e-07,
+      "loss": 0.7164,
       "step": 33190
     },
     {
-      "epoch": 0.9115651863447859,
+      "epoch": 0.9418558456299659,
       "grad_norm": 0.0,
-      "learning_rate": 4.0739732504523123e-07,
-      "loss": 0.7822,
+      "learning_rate": 1.767969869841457e-07,
+      "loss": 0.7807,
       "step": 33191
     },
     {
-      "epoch": 0.9115926505726292,
+      "epoch": 0.9418842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 4.07146053486146e-07,
-      "loss": 0.8205,
+      "learning_rate": 1.766249697294642e-07,
+      "loss": 0.6435,
       "step": 33192
     },
     {
-      "epoch": 0.9116201148004723,
+      "epoch": 0.9419125993189558,
       "grad_norm": 0.0,
-      "learning_rate": 4.0689485782943626e-07,
-      "loss": 0.748,
+      "learning_rate": 1.7645303545327453e-07,
+      "loss": 0.8333,
       "step": 33193
     },
     {
-      "epoch": 0.9116475790283156,
+      "epoch": 0.9419409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 4.0664373807708937e-07,
-      "loss": 0.8193,
+      "learning_rate": 1.7628118415702667e-07,
+      "loss": 0.7992,
       "step": 33194
     },
     {
-      "epoch": 0.9116750432561589,
+      "epoch": 0.9419693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 4.0639269423109473e-07,
-      "loss": 0.8356,
+      "learning_rate": 1.761094158421739e-07,
+      "loss": 0.9277,
       "step": 33195
     },
     {
-      "epoch": 0.9117025074840021,
+      "epoch": 0.9419977298524405,
       "grad_norm": 0.0,
-      "learning_rate": 4.061417262934375e-07,
-      "loss": 0.8749,
+      "learning_rate": 1.7593773051016617e-07,
+      "loss": 0.8292,
       "step": 33196
     },
     {
-      "epoch": 0.9117299717118453,
+      "epoch": 0.9420261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 4.0589083426610165e-07,
-      "loss": 0.8491,
+      "learning_rate": 1.7576612816245454e-07,
+      "loss": 0.7278,
       "step": 33197
     },
     {
-      "epoch": 0.9117574359396886,
+      "epoch": 0.9420544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 4.0564001815107556e-07,
-      "loss": 0.727,
+      "learning_rate": 1.7559460880048784e-07,
+      "loss": 0.8297,
       "step": 33198
     },
     {
-      "epoch": 0.9117849001675318,
+      "epoch": 0.942082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 4.053892779503421e-07,
-      "loss": 0.7023,
+      "learning_rate": 1.7542317242571493e-07,
+      "loss": 0.7313,
       "step": 33199
     },
     {
-      "epoch": 0.911812364395375,
+      "epoch": 0.94211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 4.0513861366588414e-07,
-      "loss": 0.7349,
+      "learning_rate": 1.7525181903958465e-07,
+      "loss": 0.7962,
       "step": 33200
     },
     {
-      "epoch": 0.9118398286232182,
+      "epoch": 0.9421396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 4.048880252996856e-07,
-      "loss": 0.8369,
+      "learning_rate": 1.750805486435436e-07,
+      "loss": 0.7717,
       "step": 33201
     },
     {
-      "epoch": 0.9118672928510615,
+      "epoch": 0.9421679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 4.046375128537305e-07,
-      "loss": 0.735,
+      "learning_rate": 1.7490936123903845e-07,
+      "loss": 0.7378,
       "step": 33202
     },
     {
-      "epoch": 0.9118947570789048,
+      "epoch": 0.9421963677639047,
       "grad_norm": 0.0,
-      "learning_rate": 4.043870763300017e-07,
-      "loss": 0.8583,
+      "learning_rate": 1.7473825682751576e-07,
+      "loss": 0.7268,
       "step": 33203
     },
     {
-      "epoch": 0.9119222213067479,
+      "epoch": 0.9422247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 4.0413671573047763e-07,
-      "loss": 0.821,
+      "learning_rate": 1.745672354104211e-07,
+      "loss": 0.797,
       "step": 33204
     },
     {
-      "epoch": 0.9119496855345912,
+      "epoch": 0.9422531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 4.038864310571422e-07,
-      "loss": 0.8123,
+      "learning_rate": 1.743962969891988e-07,
+      "loss": 0.7148,
       "step": 33205
     },
     {
-      "epoch": 0.9119771497624344,
+      "epoch": 0.9422814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 4.0363622231197387e-07,
-      "loss": 0.8202,
+      "learning_rate": 1.7422544156529218e-07,
+      "loss": 0.7449,
       "step": 33206
     },
     {
-      "epoch": 0.9120046139902777,
+      "epoch": 0.9423098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 4.0338608949695324e-07,
-      "loss": 0.7078,
+      "learning_rate": 1.7405466914014456e-07,
+      "loss": 0.7947,
       "step": 33207
     },
     {
-      "epoch": 0.9120320782181209,
+      "epoch": 0.9423382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 4.0313603261405876e-07,
-      "loss": 0.6799,
+      "learning_rate": 1.738839797151992e-07,
+      "loss": 0.751,
       "step": 33208
     },
     {
-      "epoch": 0.9120595424459641,
+      "epoch": 0.942366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 4.0288605166526883e-07,
-      "loss": 0.8257,
+      "learning_rate": 1.7371337329189718e-07,
+      "loss": 0.7239,
       "step": 33209
     },
     {
-      "epoch": 0.9120870066738074,
+      "epoch": 0.9423950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 4.0263614665256414e-07,
-      "loss": 0.8041,
+      "learning_rate": 1.735428498716818e-07,
+      "loss": 0.8295,
       "step": 33210
     },
     {
-      "epoch": 0.9121144709016507,
+      "epoch": 0.9424233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 4.023863175779175e-07,
-      "loss": 0.8108,
+      "learning_rate": 1.7337240945598965e-07,
+      "loss": 0.9081,
       "step": 33211
     },
     {
-      "epoch": 0.9121419351294938,
+      "epoch": 0.9424517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 4.0213656444331063e-07,
-      "loss": 0.7487,
+      "learning_rate": 1.7320205204626295e-07,
+      "loss": 0.8128,
       "step": 33212
     },
     {
-      "epoch": 0.9121693993573371,
+      "epoch": 0.9424801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 4.0188688725071534e-07,
-      "loss": 0.8105,
+      "learning_rate": 1.7303177764394163e-07,
+      "loss": 0.7236,
       "step": 33213
     },
     {
-      "epoch": 0.9121968635851803,
+      "epoch": 0.9425085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 4.0163728600210894e-07,
-      "loss": 0.785,
+      "learning_rate": 1.7286158625046122e-07,
+      "loss": 0.7711,
       "step": 33214
     },
     {
-      "epoch": 0.9122243278130235,
+      "epoch": 0.9425368898978433,
       "grad_norm": 0.0,
-      "learning_rate": 4.013877606994665e-07,
-      "loss": 0.9111,
+      "learning_rate": 1.726914778672606e-07,
+      "loss": 0.8362,
       "step": 33215
     },
     {
-      "epoch": 0.9122517920408668,
+      "epoch": 0.9425652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 4.0113831134476313e-07,
-      "loss": 0.7676,
+      "learning_rate": 1.7252145249577856e-07,
+      "loss": 0.8606,
       "step": 33216
     },
     {
-      "epoch": 0.91227925626871,
+      "epoch": 0.9425936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 4.008889379399716e-07,
-      "loss": 0.8654,
+      "learning_rate": 1.7235151013744844e-07,
+      "loss": 0.8307,
       "step": 33217
     },
     {
-      "epoch": 0.9123067204965533,
+      "epoch": 0.942622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 4.0063964048706383e-07,
-      "loss": 0.788,
+      "learning_rate": 1.7218165079370574e-07,
+      "loss": 0.7972,
       "step": 33218
     },
     {
-      "epoch": 0.9123341847243964,
+      "epoch": 0.9426503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 4.003904189880137e-07,
-      "loss": 0.7655,
+      "learning_rate": 1.7201187446598823e-07,
+      "loss": 0.771,
       "step": 33219
     },
     {
-      "epoch": 0.9123616489522397,
+      "epoch": 0.9426787741203179,
       "grad_norm": 0.0,
-      "learning_rate": 4.0014127344479515e-07,
-      "loss": 0.89,
+      "learning_rate": 1.7184218115572805e-07,
+      "loss": 0.8876,
       "step": 33220
     },
     {
-      "epoch": 0.912389113180083,
+      "epoch": 0.9427071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 3.9989220385937667e-07,
-      "loss": 0.7715,
+      "learning_rate": 1.716725708643574e-07,
+      "loss": 0.7681,
       "step": 33221
     },
     {
-      "epoch": 0.9124165774079261,
+      "epoch": 0.9427355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 3.9964321023373e-07,
-      "loss": 0.7499,
+      "learning_rate": 1.7150304359331182e-07,
+      "loss": 0.7009,
       "step": 33222
     },
     {
-      "epoch": 0.9124440416357694,
+      "epoch": 0.9427639046538024,
       "grad_norm": 0.0,
-      "learning_rate": 3.9939429256982356e-07,
-      "loss": 0.7787,
+      "learning_rate": 1.7133359934402017e-07,
+      "loss": 0.7961,
       "step": 33223
     },
     {
-      "epoch": 0.9124715058636127,
+      "epoch": 0.9427922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 3.9914545086962907e-07,
-      "loss": 0.8781,
+      "learning_rate": 1.7116423811791794e-07,
+      "loss": 0.7572,
       "step": 33224
     },
     {
-      "epoch": 0.9124989700914559,
+      "epoch": 0.9428206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 3.9889668513511393e-07,
-      "loss": 0.7551,
+      "learning_rate": 1.7099495991643066e-07,
+      "loss": 0.8554,
       "step": 33225
     },
     {
-      "epoch": 0.9125264343192991,
+      "epoch": 0.9428490351872871,
       "grad_norm": 0.0,
-      "learning_rate": 3.986479953682487e-07,
-      "loss": 0.8172,
+      "learning_rate": 1.7082576474099277e-07,
+      "loss": 0.8201,
       "step": 33226
     },
     {
-      "epoch": 0.9125538985471423,
+      "epoch": 0.9428774120317821,
       "grad_norm": 0.0,
-      "learning_rate": 3.983993815709997e-07,
-      "loss": 0.8765,
+      "learning_rate": 1.7065665259303087e-07,
+      "loss": 0.9063,
       "step": 33227
     },
     {
-      "epoch": 0.9125813627749856,
+      "epoch": 0.942905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 3.981508437453341e-07,
-      "loss": 0.8036,
+      "learning_rate": 1.7048762347397385e-07,
+      "loss": 0.7519,
       "step": 33228
     },
     {
-      "epoch": 0.9126088270028289,
+      "epoch": 0.9429341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 3.9790238189321926e-07,
-      "loss": 0.8157,
+      "learning_rate": 1.7031867738524943e-07,
+      "loss": 0.8276,
       "step": 33229
     },
     {
-      "epoch": 0.912636291230672,
+      "epoch": 0.9429625425652668,
       "grad_norm": 0.0,
-      "learning_rate": 3.9765399601661816e-07,
-      "loss": 0.8078,
+      "learning_rate": 1.7014981432828537e-07,
+      "loss": 0.7456,
       "step": 33230
     },
     {
-      "epoch": 0.9126637554585153,
+      "epoch": 0.9429909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 3.9740568611749906e-07,
-      "loss": 0.824,
+      "learning_rate": 1.699810343045083e-07,
+      "loss": 0.8253,
       "step": 33231
     },
     {
-      "epoch": 0.9126912196863585,
+      "epoch": 0.9430192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 3.971574521978261e-07,
-      "loss": 0.8036,
+      "learning_rate": 1.6981233731534263e-07,
+      "loss": 0.7604,
       "step": 33232
     },
     {
-      "epoch": 0.9127186839142017,
+      "epoch": 0.9430476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 3.9690929425956314e-07,
-      "loss": 0.8748,
+      "learning_rate": 1.696437233622139e-07,
+      "loss": 0.8325,
       "step": 33233
     },
     {
-      "epoch": 0.912746148142045,
+      "epoch": 0.9430760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 3.9666121230467315e-07,
-      "loss": 0.8818,
+      "learning_rate": 1.6947519244654654e-07,
+      "loss": 0.7419,
       "step": 33234
     },
     {
-      "epoch": 0.9127736123698882,
+      "epoch": 0.9431044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 3.9641320633512003e-07,
-      "loss": 0.8961,
+      "learning_rate": 1.6930674456976494e-07,
+      "loss": 0.7982,
       "step": 33235
     },
     {
-      "epoch": 0.9128010765977315,
+      "epoch": 0.9431328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 3.961652763528667e-07,
-      "loss": 0.7776,
+      "learning_rate": 1.6913837973329129e-07,
+      "loss": 0.8092,
       "step": 33236
     },
     {
-      "epoch": 0.9128285408255747,
+      "epoch": 0.943161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.9591742235987273e-07,
-      "loss": 0.7991,
+      "learning_rate": 1.689700979385467e-07,
+      "loss": 0.8633,
       "step": 33237
     },
     {
-      "epoch": 0.9128560050534179,
+      "epoch": 0.9431895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 3.9566964435809983e-07,
-      "loss": 0.7599,
+      "learning_rate": 1.6880189918695334e-07,
+      "loss": 0.6842,
       "step": 33238
     },
     {
-      "epoch": 0.9128834692812612,
+      "epoch": 0.9432179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 3.954219423495109e-07,
-      "loss": 0.8033,
+      "learning_rate": 1.6863378347993343e-07,
+      "loss": 0.6992,
       "step": 33239
     },
     {
-      "epoch": 0.9129109335091043,
+      "epoch": 0.9432463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 3.951743163360633e-07,
-      "loss": 0.8724,
+      "learning_rate": 1.6846575081890472e-07,
+      "loss": 0.7282,
       "step": 33240
     },
     {
-      "epoch": 0.9129383977369476,
+      "epoch": 0.9432746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 3.949267663197165e-07,
-      "loss": 0.8242,
+      "learning_rate": 1.6829780120528938e-07,
+      "loss": 0.8217,
       "step": 33241
     },
     {
-      "epoch": 0.9129658619647909,
+      "epoch": 0.9433030646992054,
       "grad_norm": 0.0,
-      "learning_rate": 3.9467929230242894e-07,
-      "loss": 0.7888,
+      "learning_rate": 1.6812993464050298e-07,
+      "loss": 0.7066,
       "step": 33242
     },
     {
-      "epoch": 0.9129933261926341,
+      "epoch": 0.9433314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 3.9443189428616024e-07,
-      "loss": 0.7943,
+      "learning_rate": 1.6796215112596547e-07,
+      "loss": 0.9032,
       "step": 33243
     },
     {
-      "epoch": 0.9130207904204773,
+      "epoch": 0.9433598183881953,
       "grad_norm": 0.0,
-      "learning_rate": 3.9418457227286877e-07,
-      "loss": 0.7614,
+      "learning_rate": 1.6779445066309351e-07,
+      "loss": 0.8796,
       "step": 33244
     },
     {
-      "epoch": 0.9130482546483205,
+      "epoch": 0.9433881952326901,
       "grad_norm": 0.0,
-      "learning_rate": 3.9393732626450963e-07,
-      "loss": 0.7315,
+      "learning_rate": 1.6762683325330486e-07,
+      "loss": 0.8949,
       "step": 33245
     },
     {
-      "epoch": 0.9130757188761638,
+      "epoch": 0.943416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 3.93690156263038e-07,
-      "loss": 0.7989,
+      "learning_rate": 1.6745929889801283e-07,
+      "loss": 0.7664,
       "step": 33246
     },
     {
-      "epoch": 0.9131031831040071,
+      "epoch": 0.94344494892168,
       "grad_norm": 0.0,
-      "learning_rate": 3.934430622704111e-07,
-      "loss": 0.7877,
+      "learning_rate": 1.6729184759863515e-07,
+      "loss": 0.8072,
       "step": 33247
     },
     {
-      "epoch": 0.9131306473318502,
+      "epoch": 0.9434733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 3.9319604428858407e-07,
-      "loss": 0.8056,
+      "learning_rate": 1.6712447935658515e-07,
+      "loss": 0.8028,
       "step": 33248
     },
     {
-      "epoch": 0.9131581115596935,
+      "epoch": 0.9435017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 3.92949102319512e-07,
-      "loss": 0.8485,
+      "learning_rate": 1.6695719417327617e-07,
+      "loss": 0.8142,
       "step": 33249
     },
     {
-      "epoch": 0.9131855757875368,
+      "epoch": 0.9435300794551645,
       "grad_norm": 0.0,
-      "learning_rate": 3.927022363651467e-07,
-      "loss": 0.8017,
+      "learning_rate": 1.667899920501226e-07,
+      "loss": 0.7569,
       "step": 33250
     },
     {
-      "epoch": 0.91321304001538,
+      "epoch": 0.9435584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 3.924554464274444e-07,
-      "loss": 0.8645,
+      "learning_rate": 1.6662287298853553e-07,
+      "loss": 0.7747,
       "step": 33251
     },
     {
-      "epoch": 0.9132405042432232,
+      "epoch": 0.9435868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 3.9220873250835566e-07,
-      "loss": 0.8278,
+      "learning_rate": 1.6645583698992606e-07,
+      "loss": 0.7736,
       "step": 33252
     },
     {
-      "epoch": 0.9132679684710664,
+      "epoch": 0.9436152099886492,
       "grad_norm": 0.0,
-      "learning_rate": 3.919620946098335e-07,
-      "loss": 0.7866,
+      "learning_rate": 1.6628888405570752e-07,
+      "loss": 0.8049,
       "step": 33253
     },
     {
-      "epoch": 0.9132954326989097,
+      "epoch": 0.9436435868331442,
       "grad_norm": 0.0,
-      "learning_rate": 3.9171553273382843e-07,
-      "loss": 0.8715,
+      "learning_rate": 1.6612201418728767e-07,
+      "loss": 0.7802,
       "step": 33254
     },
     {
-      "epoch": 0.913322896926753,
+      "epoch": 0.9436719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 3.9146904688229125e-07,
-      "loss": 0.797,
+      "learning_rate": 1.659552273860776e-07,
+      "loss": 0.8308,
       "step": 33255
     },
     {
-      "epoch": 0.9133503611545961,
+      "epoch": 0.9437003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 3.912226370571737e-07,
-      "loss": 0.7928,
+      "learning_rate": 1.657885236534873e-07,
+      "loss": 0.8117,
       "step": 33256
     },
     {
-      "epoch": 0.9133778253824394,
+      "epoch": 0.9437287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 3.9097630326042526e-07,
-      "loss": 0.8959,
+      "learning_rate": 1.6562190299092229e-07,
+      "loss": 0.7802,
       "step": 33257
     },
     {
-      "epoch": 0.9134052896102826,
+      "epoch": 0.9437570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 3.907300454939944e-07,
-      "loss": 0.9028,
+      "learning_rate": 1.6545536539979034e-07,
+      "loss": 0.8102,
       "step": 33258
     },
     {
-      "epoch": 0.9134327538381258,
+      "epoch": 0.9437854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 3.904838637598285e-07,
-      "loss": 0.7449,
+      "learning_rate": 1.6528891088150145e-07,
+      "loss": 0.7101,
       "step": 33259
     },
     {
-      "epoch": 0.9134602180659691,
+      "epoch": 0.9438138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 3.902377580598782e-07,
-      "loss": 0.7673,
+      "learning_rate": 1.6512253943745671e-07,
+      "loss": 0.7512,
       "step": 33260
     },
     {
-      "epoch": 0.9134876822938123,
+      "epoch": 0.9438422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 3.899917283960886e-07,
-      "loss": 0.8285,
+      "learning_rate": 1.649562510690661e-07,
+      "loss": 0.8089,
       "step": 33261
     },
     {
-      "epoch": 0.9135151465216556,
+      "epoch": 0.9438706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 3.897457747704059e-07,
-      "loss": 0.8494,
+      "learning_rate": 1.647900457777307e-07,
+      "loss": 0.672,
       "step": 33262
     },
     {
-      "epoch": 0.9135426107494988,
+      "epoch": 0.9438989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 3.8949989718477964e-07,
-      "loss": 0.7481,
+      "learning_rate": 1.6462392356485611e-07,
+      "loss": 0.7434,
       "step": 33263
     },
     {
-      "epoch": 0.913570074977342,
+      "epoch": 0.9439273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 3.892540956411506e-07,
-      "loss": 0.8314,
+      "learning_rate": 1.6445788443184562e-07,
+      "loss": 0.8038,
       "step": 33264
     },
     {
-      "epoch": 0.9135975392051853,
+      "epoch": 0.943955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 3.890083701414671e-07,
-      "loss": 0.831,
+      "learning_rate": 1.6429192838010143e-07,
+      "loss": 0.8009,
       "step": 33265
     },
     {
-      "epoch": 0.9136250034330284,
+      "epoch": 0.9439841089670828,
       "grad_norm": 0.0,
-      "learning_rate": 3.887627206876721e-07,
-      "loss": 0.7601,
+      "learning_rate": 1.6412605541102465e-07,
+      "loss": 0.819,
       "step": 33266
     },
     {
-      "epoch": 0.9136524676608717,
+      "epoch": 0.9440124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 3.885171472817084e-07,
-      "loss": 0.8042,
+      "learning_rate": 1.6396026552601863e-07,
+      "loss": 0.6895,
       "step": 33267
     },
     {
-      "epoch": 0.913679931888715,
+      "epoch": 0.9440408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 3.882716499255224e-07,
-      "loss": 0.8041,
+      "learning_rate": 1.637945587264822e-07,
+      "loss": 0.8173,
       "step": 33268
     },
     {
-      "epoch": 0.9137073961165582,
+      "epoch": 0.9440692395005675,
       "grad_norm": 0.0,
-      "learning_rate": 3.8802622862105455e-07,
-      "loss": 0.8034,
+      "learning_rate": 1.6362893501381317e-07,
+      "loss": 0.8632,
       "step": 33269
     },
     {
-      "epoch": 0.9137348603444014,
+      "epoch": 0.9440976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 3.8778088337024453e-07,
-      "loss": 0.764,
+      "learning_rate": 1.6346339438941483e-07,
+      "loss": 0.7896,
       "step": 33270
     },
     {
-      "epoch": 0.9137623245722446,
+      "epoch": 0.9441259931895574,
       "grad_norm": 0.0,
-      "learning_rate": 3.8753561417503635e-07,
-      "loss": 0.8293,
+      "learning_rate": 1.6329793685468276e-07,
+      "loss": 0.8698,
       "step": 33271
     },
     {
-      "epoch": 0.9137897888000879,
+      "epoch": 0.9441543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 3.8729042103736957e-07,
-      "loss": 0.813,
+      "learning_rate": 1.6313256241101584e-07,
+      "loss": 0.8584,
       "step": 33272
     },
     {
-      "epoch": 0.9138172530279312,
+      "epoch": 0.9441827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 3.8704530395918484e-07,
-      "loss": 0.8231,
+      "learning_rate": 1.629672710598096e-07,
+      "loss": 0.71,
       "step": 33273
     },
     {
-      "epoch": 0.9138447172557743,
+      "epoch": 0.944211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 3.8680026294242055e-07,
-      "loss": 0.8787,
+      "learning_rate": 1.628020628024618e-07,
+      "loss": 0.82,
       "step": 33274
     },
     {
-      "epoch": 0.9138721814836176,
+      "epoch": 0.9442395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 3.8655529798901857e-07,
-      "loss": 0.8479,
+      "learning_rate": 1.6263693764036581e-07,
+      "loss": 0.8691,
       "step": 33275
     },
     {
-      "epoch": 0.9138996457114609,
+      "epoch": 0.9442678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 3.86310409100914e-07,
-      "loss": 0.8209,
+      "learning_rate": 1.6247189557491937e-07,
+      "loss": 0.7663,
       "step": 33276
     },
     {
-      "epoch": 0.913927109939304,
+      "epoch": 0.9442962542565266,
       "grad_norm": 0.0,
-      "learning_rate": 3.8606559628004415e-07,
-      "loss": 0.7906,
+      "learning_rate": 1.6230693660751472e-07,
+      "loss": 0.7749,
       "step": 33277
     },
     {
-      "epoch": 0.9139545741671473,
+      "epoch": 0.9443246311010216,
       "grad_norm": 0.0,
-      "learning_rate": 3.858208595283486e-07,
-      "loss": 0.8119,
+      "learning_rate": 1.6214206073954518e-07,
+      "loss": 0.8147,
       "step": 33278
     },
     {
-      "epoch": 0.9139820383949905,
+      "epoch": 0.9443530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 3.855761988477613e-07,
-      "loss": 0.7268,
+      "learning_rate": 1.619772679724041e-07,
+      "loss": 0.784,
       "step": 33279
     },
     {
-      "epoch": 0.9140095026228338,
+      "epoch": 0.9443813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 3.8533161424022193e-07,
-      "loss": 0.8092,
+      "learning_rate": 1.6181255830748366e-07,
+      "loss": 0.7651,
       "step": 33280
     },
     {
-      "epoch": 0.914036966850677,
+      "epoch": 0.9444097616345063,
       "grad_norm": 0.0,
-      "learning_rate": 3.850871057076611e-07,
-      "loss": 0.7611,
+      "learning_rate": 1.6164793174617388e-07,
+      "loss": 0.778,
       "step": 33281
     },
     {
-      "epoch": 0.9140644310785202,
+      "epoch": 0.9444381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 3.848426732520161e-07,
-      "loss": 0.7647,
+      "learning_rate": 1.6148338828986698e-07,
+      "loss": 0.8013,
       "step": 33282
     },
     {
-      "epoch": 0.9140918953063635,
+      "epoch": 0.944466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 3.8459831687521984e-07,
-      "loss": 0.7508,
+      "learning_rate": 1.6131892793995184e-07,
+      "loss": 0.7841,
       "step": 33283
     },
     {
-      "epoch": 0.9141193595342066,
+      "epoch": 0.9444948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 3.8435403657920754e-07,
-      "loss": 0.7884,
+      "learning_rate": 1.6115455069781848e-07,
+      "loss": 0.7729,
       "step": 33284
     },
     {
-      "epoch": 0.9141468237620499,
+      "epoch": 0.9445232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 3.841098323659098e-07,
-      "loss": 0.7895,
+      "learning_rate": 1.6099025656485357e-07,
+      "loss": 0.9056,
       "step": 33285
     },
     {
-      "epoch": 0.9141742879898932,
+      "epoch": 0.9445516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 3.8386570423726175e-07,
-      "loss": 0.8254,
+      "learning_rate": 1.6082604554244706e-07,
+      "loss": 0.9222,
       "step": 33286
     },
     {
-      "epoch": 0.9142017522177364,
+      "epoch": 0.9445800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 3.8362165219519075e-07,
-      "loss": 0.7865,
+      "learning_rate": 1.606619176319868e-07,
+      "loss": 0.7879,
       "step": 33287
     },
     {
-      "epoch": 0.9142292164455796,
+      "epoch": 0.9446083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 3.833776762416308e-07,
-      "loss": 0.7663,
+      "learning_rate": 1.6049787283485608e-07,
+      "loss": 0.7815,
       "step": 33288
     },
     {
-      "epoch": 0.9142566806734229,
+      "epoch": 0.9446367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 3.8313377637851256e-07,
-      "loss": 0.781,
+      "learning_rate": 1.6033391115244156e-07,
+      "loss": 0.7923,
       "step": 33289
     },
     {
-      "epoch": 0.9142841449012661,
+      "epoch": 0.9446651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 3.828899526077645e-07,
-      "loss": 0.8894,
+      "learning_rate": 1.6017003258612996e-07,
+      "loss": 0.7753,
       "step": 33290
     },
     {
-      "epoch": 0.9143116091291094,
+      "epoch": 0.9446935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 3.826462049313162e-07,
-      "loss": 0.9141,
+      "learning_rate": 1.6000623713730457e-07,
+      "loss": 0.7171,
       "step": 33291
     },
     {
-      "epoch": 0.9143390733569525,
+      "epoch": 0.9447219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 3.8240253335109835e-07,
-      "loss": 0.7449,
+      "learning_rate": 1.5984252480734763e-07,
+      "loss": 0.7577,
       "step": 33292
     },
     {
-      "epoch": 0.9143665375847958,
+      "epoch": 0.9447502837684449,
       "grad_norm": 0.0,
-      "learning_rate": 3.8215893786903604e-07,
-      "loss": 0.8376,
+      "learning_rate": 1.5967889559764472e-07,
+      "loss": 0.8194,
       "step": 33293
     },
     {
-      "epoch": 0.9143940018126391,
+      "epoch": 0.9447786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 3.8191541848705773e-07,
-      "loss": 0.8378,
+      "learning_rate": 1.5951534950957582e-07,
+      "loss": 0.6961,
       "step": 33294
     },
     {
-      "epoch": 0.9144214660404822,
+      "epoch": 0.9448070374574348,
       "grad_norm": 0.0,
-      "learning_rate": 3.816719752070897e-07,
-      "loss": 0.7631,
+      "learning_rate": 1.5935188654452206e-07,
+      "loss": 0.7103,
       "step": 33295
     },
     {
-      "epoch": 0.9144489302683255,
+      "epoch": 0.9448354143019296,
       "grad_norm": 0.0,
-      "learning_rate": 3.814286080310592e-07,
-      "loss": 0.8844,
+      "learning_rate": 1.5918850670386677e-07,
+      "loss": 0.8477,
       "step": 33296
     },
     {
-      "epoch": 0.9144763944961687,
+      "epoch": 0.9448637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 3.811853169608914e-07,
-      "loss": 0.8132,
+      "learning_rate": 1.5902520998898773e-07,
+      "loss": 0.8229,
       "step": 33297
     },
     {
-      "epoch": 0.914503858724012,
+      "epoch": 0.9448921679909195,
       "grad_norm": 0.0,
-      "learning_rate": 3.809421019985116e-07,
-      "loss": 0.8785,
+      "learning_rate": 1.588619964012661e-07,
+      "loss": 0.7606,
       "step": 33298
     },
     {
-      "epoch": 0.9145313229518552,
+      "epoch": 0.9449205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 3.806989631458435e-07,
-      "loss": 0.799,
+      "learning_rate": 1.586988659420785e-07,
+      "loss": 0.797,
       "step": 33299
     },
     {
-      "epoch": 0.9145587871796984,
+      "epoch": 0.9449489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 3.804559004048136e-07,
-      "loss": 0.7925,
+      "learning_rate": 1.5853581861280387e-07,
+      "loss": 0.7752,
       "step": 33300
     },
     {
-      "epoch": 0.9145862514075417,
+      "epoch": 0.944977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 3.8021291377734026e-07,
-      "loss": 0.7523,
+      "learning_rate": 1.5837285441482108e-07,
+      "loss": 0.7966,
       "step": 33301
     },
     {
-      "epoch": 0.914613715635385,
+      "epoch": 0.945005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 3.799700032653486e-07,
-      "loss": 0.7963,
+      "learning_rate": 1.582099733495035e-07,
+      "loss": 0.8117,
       "step": 33302
     },
     {
-      "epoch": 0.9146411798632281,
+      "epoch": 0.9450340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 3.7972716887076154e-07,
-      "loss": 0.9207,
+      "learning_rate": 1.5804717541822888e-07,
+      "loss": 0.8405,
       "step": 33303
     },
     {
-      "epoch": 0.9146686440910714,
+      "epoch": 0.9450624290578887,
       "grad_norm": 0.0,
-      "learning_rate": 3.794844105955009e-07,
-      "loss": 0.8159,
+      "learning_rate": 1.5788446062237394e-07,
+      "loss": 0.7249,
       "step": 33304
     },
     {
-      "epoch": 0.9146961083189146,
+      "epoch": 0.9450908059023837,
       "grad_norm": 0.0,
-      "learning_rate": 3.7924172844148397e-07,
-      "loss": 0.8589,
+      "learning_rate": 1.577218289633098e-07,
+      "loss": 0.8497,
       "step": 33305
     },
     {
-      "epoch": 0.9147235725467578,
+      "epoch": 0.9451191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 3.7899912241063483e-07,
-      "loss": 0.763,
+      "learning_rate": 1.5755928044241198e-07,
+      "loss": 0.8894,
       "step": 33306
     },
     {
-      "epoch": 0.9147510367746011,
+      "epoch": 0.9451475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 3.787565925048697e-07,
-      "loss": 0.931,
+      "learning_rate": 1.5739681506105388e-07,
+      "loss": 0.8046,
       "step": 33307
     },
     {
-      "epoch": 0.9147785010024443,
+      "epoch": 0.9451759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 3.785141387261115e-07,
-      "loss": 0.8019,
+      "learning_rate": 1.5723443282060657e-07,
+      "loss": 0.7761,
       "step": 33308
     },
     {
-      "epoch": 0.9148059652302876,
+      "epoch": 0.9452043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 3.7827176107627536e-07,
-      "loss": 0.7503,
+      "learning_rate": 1.5707213372244345e-07,
+      "loss": 0.8293,
       "step": 33309
     },
     {
-      "epoch": 0.9148334294581307,
+      "epoch": 0.9452326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 3.7802945955728087e-07,
-      "loss": 0.7386,
+      "learning_rate": 1.569099177679345e-07,
+      "loss": 0.7976,
       "step": 33310
     },
     {
-      "epoch": 0.914860893685974,
+      "epoch": 0.945261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.7778723417104424e-07,
-      "loss": 0.7941,
+      "learning_rate": 1.5674778495844868e-07,
+      "loss": 0.7976,
       "step": 33311
     },
     {
-      "epoch": 0.9148883579138173,
+      "epoch": 0.9452894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.775450849194806e-07,
-      "loss": 0.825,
+      "learning_rate": 1.5658573529535815e-07,
+      "loss": 0.8317,
       "step": 33312
     },
     {
-      "epoch": 0.9149158221416605,
+      "epoch": 0.9453178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 3.7730301180450955e-07,
-      "loss": 0.788,
+      "learning_rate": 1.5642376878002963e-07,
+      "loss": 0.8215,
       "step": 33313
     },
     {
-      "epoch": 0.9149432863695037,
+      "epoch": 0.9453461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 3.7706101482804293e-07,
-      "loss": 0.7443,
+      "learning_rate": 1.5626188541383203e-07,
+      "loss": 0.8297,
       "step": 33314
     },
     {
-      "epoch": 0.914970750597347,
+      "epoch": 0.9453745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 3.7681909399199914e-07,
-      "loss": 0.7924,
+      "learning_rate": 1.5610008519813312e-07,
+      "loss": 0.8426,
       "step": 33315
     },
     {
-      "epoch": 0.9149982148251902,
+      "epoch": 0.9454029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 3.765772492982889e-07,
-      "loss": 0.7135,
+      "learning_rate": 1.5593836813429852e-07,
+      "loss": 0.7625,
       "step": 33316
     },
     {
-      "epoch": 0.9150256790530334,
+      "epoch": 0.9454313280363223,
       "grad_norm": 0.0,
-      "learning_rate": 3.7633548074882843e-07,
-      "loss": 0.8031,
+      "learning_rate": 1.55776734223696e-07,
+      "loss": 0.8032,
       "step": 33317
     },
     {
-      "epoch": 0.9150531432808766,
+      "epoch": 0.9454597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 3.7609378834552846e-07,
-      "loss": 0.7527,
+      "learning_rate": 1.5561518346769e-07,
+      "loss": 0.7268,
       "step": 33318
     },
     {
-      "epoch": 0.9150806075087199,
+      "epoch": 0.9454880817253122,
       "grad_norm": 0.0,
-      "learning_rate": 3.75852172090303e-07,
-      "loss": 0.8861,
+      "learning_rate": 1.5545371586764391e-07,
+      "loss": 0.8071,
       "step": 33319
     },
     {
-      "epoch": 0.9151080717365632,
+      "epoch": 0.945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 3.7561063198506165e-07,
-      "loss": 0.8708,
+      "learning_rate": 1.552923314249244e-07,
+      "loss": 0.7743,
       "step": 33320
     },
     {
-      "epoch": 0.9151355359644063,
+      "epoch": 0.9455448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 3.753691680317184e-07,
-      "loss": 0.7367,
+      "learning_rate": 1.551310301408926e-07,
+      "loss": 0.7966,
       "step": 33321
     },
     {
-      "epoch": 0.9151630001922496,
+      "epoch": 0.9455732122587969,
       "grad_norm": 0.0,
-      "learning_rate": 3.751277802321829e-07,
-      "loss": 0.7537,
+      "learning_rate": 1.5496981201691075e-07,
+      "loss": 0.8,
       "step": 33322
     },
     {
-      "epoch": 0.9151904644200928,
+      "epoch": 0.9456015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 3.7488646858836464e-07,
-      "loss": 0.7804,
+      "learning_rate": 1.548086770543422e-07,
+      "loss": 0.7865,
       "step": 33323
     },
     {
-      "epoch": 0.915217928647936,
+      "epoch": 0.9456299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 3.7464523310217325e-07,
-      "loss": 0.8263,
+      "learning_rate": 1.5464762525454702e-07,
+      "loss": 0.8294,
       "step": 33324
     },
     {
-      "epoch": 0.9152453928757793,
+      "epoch": 0.9456583427922814,
       "grad_norm": 0.0,
-      "learning_rate": 3.744040737755161e-07,
-      "loss": 0.8216,
+      "learning_rate": 1.5448665661888629e-07,
+      "loss": 0.7318,
       "step": 33325
     },
     {
-      "epoch": 0.9152728571036225,
+      "epoch": 0.9456867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 3.741629906103028e-07,
-      "loss": 0.7429,
+      "learning_rate": 1.5432577114871894e-07,
+      "loss": 0.7944,
       "step": 33326
     },
     {
-      "epoch": 0.9153003213314658,
+      "epoch": 0.9457150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 3.7392198360844065e-07,
-      "loss": 0.7799,
+      "learning_rate": 1.54164968845405e-07,
+      "loss": 0.7385,
       "step": 33327
     },
     {
-      "epoch": 0.915327785559309,
+      "epoch": 0.9457434733257661,
       "grad_norm": 0.0,
-      "learning_rate": 3.7368105277183707e-07,
-      "loss": 0.8136,
+      "learning_rate": 1.5400424971030114e-07,
+      "loss": 0.7801,
       "step": 33328
     },
     {
-      "epoch": 0.9153552497871522,
+      "epoch": 0.9457718501702611,
       "grad_norm": 0.0,
-      "learning_rate": 3.734401981023983e-07,
-      "loss": 0.6995,
+      "learning_rate": 1.538436137447663e-07,
+      "loss": 0.8109,
       "step": 33329
     },
     {
-      "epoch": 0.9153827140149955,
+      "epoch": 0.945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 3.7319941960202834e-07,
-      "loss": 0.7332,
+      "learning_rate": 1.5368306095015716e-07,
+      "loss": 0.6722,
       "step": 33330
     },
     {
-      "epoch": 0.9154101782428387,
+      "epoch": 0.9458286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 3.7295871727263347e-07,
-      "loss": 0.7974,
+      "learning_rate": 1.5352259132783043e-07,
+      "loss": 0.7965,
       "step": 33331
     },
     {
-      "epoch": 0.9154376424706819,
+      "epoch": 0.9458569807037458,
       "grad_norm": 0.0,
-      "learning_rate": 3.727180911161199e-07,
-      "loss": 0.8022,
+      "learning_rate": 1.5336220487914054e-07,
+      "loss": 0.8705,
       "step": 33332
     },
     {
-      "epoch": 0.9154651066985252,
+      "epoch": 0.9458853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 3.724775411343895e-07,
-      "loss": 0.8459,
+      "learning_rate": 1.5320190160544312e-07,
+      "loss": 0.778,
       "step": 33333
     },
     {
-      "epoch": 0.9154925709263684,
+      "epoch": 0.9459137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 3.7223706732934627e-07,
-      "loss": 0.8182,
+      "learning_rate": 1.530416815080915e-07,
+      "loss": 0.7754,
       "step": 33334
     },
     {
-      "epoch": 0.9155200351542117,
+      "epoch": 0.9459421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 3.719966697028932e-07,
-      "loss": 0.8033,
+      "learning_rate": 1.5288154458844019e-07,
+      "loss": 0.8404,
       "step": 33335
     },
     {
-      "epoch": 0.9155474993820548,
+      "epoch": 0.9459704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 3.7175634825693085e-07,
-      "loss": 0.7982,
+      "learning_rate": 1.5272149084784026e-07,
+      "loss": 0.7981,
       "step": 33336
     },
     {
-      "epoch": 0.9155749636098981,
+      "epoch": 0.9459988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 3.7151610299336115e-07,
-      "loss": 0.797,
+      "learning_rate": 1.5256152028764404e-07,
+      "loss": 0.821,
       "step": 33337
     },
     {
-      "epoch": 0.9156024278377414,
+      "epoch": 0.9460272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 3.7127593391408703e-07,
-      "loss": 0.719,
+      "learning_rate": 1.5240163290920485e-07,
+      "loss": 0.8251,
       "step": 33338
     },
     {
-      "epoch": 0.9156298920655845,
+      "epoch": 0.94605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.7103584102100797e-07,
-      "loss": 0.7807,
+      "learning_rate": 1.522418287138705e-07,
+      "loss": 0.7835,
       "step": 33339
     },
     {
-      "epoch": 0.9156573562934278,
+      "epoch": 0.9460839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 3.7079582431602254e-07,
-      "loss": 0.6688,
+      "learning_rate": 1.5208210770299215e-07,
+      "loss": 0.9389,
       "step": 33340
     },
     {
-      "epoch": 0.9156848205212711,
+      "epoch": 0.9461123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 3.705558838010315e-07,
-      "loss": 0.7924,
+      "learning_rate": 1.519224698779198e-07,
+      "loss": 0.8435,
       "step": 33341
     },
     {
-      "epoch": 0.9157122847491143,
+      "epoch": 0.9461407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 3.7031601947793205e-07,
-      "loss": 0.8115,
+      "learning_rate": 1.5176291524000018e-07,
+      "loss": 0.9167,
       "step": 33342
     },
     {
-      "epoch": 0.9157397489769575,
+      "epoch": 0.9461691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 3.700762313486217e-07,
-      "loss": 0.8802,
+      "learning_rate": 1.5160344379058223e-07,
+      "loss": 0.8091,
       "step": 33343
     },
     {
-      "epoch": 0.9157672132048007,
+      "epoch": 0.9461975028376844,
       "grad_norm": 0.0,
-      "learning_rate": 3.6983651941499776e-07,
-      "loss": 0.8332,
+      "learning_rate": 1.5144405553101372e-07,
+      "loss": 0.7651,
       "step": 33344
     },
     {
-      "epoch": 0.915794677432644,
+      "epoch": 0.9462258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 3.6959688367895985e-07,
-      "loss": 0.7708,
+      "learning_rate": 1.5128475046263914e-07,
+      "loss": 0.8731,
       "step": 33345
     },
     {
-      "epoch": 0.9158221416604873,
+      "epoch": 0.9462542565266743,
       "grad_norm": 0.0,
-      "learning_rate": 3.693573241424009e-07,
-      "loss": 0.767,
+      "learning_rate": 1.5112552858680408e-07,
+      "loss": 0.8193,
       "step": 33346
     },
     {
-      "epoch": 0.9158496058883304,
+      "epoch": 0.9462826333711691,
       "grad_norm": 0.0,
-      "learning_rate": 3.69117840807216e-07,
-      "loss": 0.8744,
+      "learning_rate": 1.5096638990485524e-07,
+      "loss": 0.8716,
       "step": 33347
     },
     {
-      "epoch": 0.9158770701161737,
+      "epoch": 0.946311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 3.688784336753037e-07,
-      "loss": 0.8364,
+      "learning_rate": 1.50807334418136e-07,
+      "loss": 0.7693,
       "step": 33348
     },
     {
-      "epoch": 0.9159045343440169,
+      "epoch": 0.946339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 3.6863910274855475e-07,
-      "loss": 0.7523,
+      "learning_rate": 1.5064836212799084e-07,
+      "loss": 0.7285,
       "step": 33349
     },
     {
-      "epoch": 0.9159319985718601,
+      "epoch": 0.9463677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 3.683998480288642e-07,
-      "loss": 0.8064,
+      "learning_rate": 1.504894730357609e-07,
+      "loss": 0.8424,
       "step": 33350
     },
     {
-      "epoch": 0.9159594627997034,
+      "epoch": 0.9463961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 3.68160669518125e-07,
-      "loss": 0.8253,
+      "learning_rate": 1.5033066714278843e-07,
+      "loss": 0.9508,
       "step": 33351
     },
     {
-      "epoch": 0.9159869270275466,
+      "epoch": 0.9464245175936435,
       "grad_norm": 0.0,
-      "learning_rate": 3.6792156721823015e-07,
-      "loss": 0.8442,
+      "learning_rate": 1.5017194445041682e-07,
+      "loss": 0.8673,
       "step": 33352
     },
     {
-      "epoch": 0.9160143912553899,
+      "epoch": 0.9464528944381385,
       "grad_norm": 0.0,
-      "learning_rate": 3.6768254113106916e-07,
-      "loss": 0.8876,
+      "learning_rate": 1.50013304959985e-07,
+      "loss": 0.7407,
       "step": 33353
     },
     {
-      "epoch": 0.9160418554832331,
+      "epoch": 0.9464812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 3.674435912585361e-07,
-      "loss": 0.8322,
+      "learning_rate": 1.498547486728341e-07,
+      "loss": 0.847,
       "step": 33354
     },
     {
-      "epoch": 0.9160693197110763,
+      "epoch": 0.9465096481271282,
       "grad_norm": 0.0,
-      "learning_rate": 3.6720471760252065e-07,
-      "loss": 0.7122,
+      "learning_rate": 1.496962755903031e-07,
+      "loss": 0.8353,
       "step": 33355
     },
     {
-      "epoch": 0.9160967839389196,
+      "epoch": 0.9465380249716232,
       "grad_norm": 0.0,
-      "learning_rate": 3.669659201649134e-07,
-      "loss": 0.8015,
+      "learning_rate": 1.4953788571372863e-07,
+      "loss": 0.8629,
       "step": 33356
     },
     {
-      "epoch": 0.9161242481667627,
+      "epoch": 0.9465664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 3.66727198947604e-07,
-      "loss": 0.7189,
+      "learning_rate": 1.4937957904445188e-07,
+      "loss": 0.883,
       "step": 33357
     },
     {
-      "epoch": 0.916151712394606,
+      "epoch": 0.9465947786606129,
       "grad_norm": 0.0,
-      "learning_rate": 3.6648855395247985e-07,
-      "loss": 0.8705,
+      "learning_rate": 1.4922135558380957e-07,
+      "loss": 0.9202,
       "step": 33358
     },
     {
-      "epoch": 0.9161791766224493,
+      "epoch": 0.9466231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 3.662499851814283e-07,
-      "loss": 0.8797,
+      "learning_rate": 1.4906321533313507e-07,
+      "loss": 0.8246,
       "step": 33359
     },
     {
-      "epoch": 0.9162066408502925,
+      "epoch": 0.9466515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 3.6601149263634006e-07,
-      "loss": 0.8172,
+      "learning_rate": 1.489051582937673e-07,
+      "loss": 0.8615,
       "step": 33360
     },
     {
-      "epoch": 0.9162341050781357,
+      "epoch": 0.9466799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 3.657730763191003e-07,
-      "loss": 0.8587,
+      "learning_rate": 1.4874718446704074e-07,
+      "loss": 0.8278,
       "step": 33361
     },
     {
-      "epoch": 0.916261569305979,
+      "epoch": 0.9467082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 3.655347362315953e-07,
-      "loss": 0.8496,
+      "learning_rate": 1.485892938542899e-07,
+      "loss": 0.8342,
       "step": 33362
     },
     {
-      "epoch": 0.9162890335338222,
+      "epoch": 0.9467366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 3.652964723757135e-07,
-      "loss": 0.7949,
+      "learning_rate": 1.48431486456847e-07,
+      "loss": 0.8432,
       "step": 33363
     },
     {
-      "epoch": 0.9163164977616655,
+      "epoch": 0.9467650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 3.6505828475333685e-07,
-      "loss": 0.8523,
+      "learning_rate": 1.4827376227604772e-07,
+      "loss": 0.8044,
       "step": 33364
     },
     {
-      "epoch": 0.9163439619895086,
+      "epoch": 0.9467934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 3.6482017336635255e-07,
-      "loss": 0.764,
+      "learning_rate": 1.4811612131322205e-07,
+      "loss": 0.8097,
       "step": 33365
     },
     {
-      "epoch": 0.9163714262173519,
+      "epoch": 0.946821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 3.6458213821664146e-07,
-      "loss": 0.7729,
+      "learning_rate": 1.4795856356970227e-07,
+      "loss": 0.8142,
       "step": 33366
     },
     {
-      "epoch": 0.9163988904451952,
+      "epoch": 0.946850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 3.6434417930608977e-07,
-      "loss": 0.8864,
+      "learning_rate": 1.4780108904681955e-07,
+      "loss": 0.8187,
       "step": 33367
     },
     {
-      "epoch": 0.9164263546730383,
+      "epoch": 0.9468785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 3.6410629663657825e-07,
-      "loss": 0.8318,
+      "learning_rate": 1.4764369774590392e-07,
+      "loss": 0.8072,
       "step": 33368
     },
     {
-      "epoch": 0.9164538189008816,
+      "epoch": 0.9469069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 3.6386849020999314e-07,
-      "loss": 0.7331,
+      "learning_rate": 1.4748638966828545e-07,
+      "loss": 0.8156,
       "step": 33369
     },
     {
-      "epoch": 0.9164812831287248,
+      "epoch": 0.9469353007945517,
       "grad_norm": 0.0,
-      "learning_rate": 3.6363076002821074e-07,
-      "loss": 0.8005,
+      "learning_rate": 1.4732916481529192e-07,
+      "loss": 0.8364,
       "step": 33370
     },
     {
-      "epoch": 0.9165087473565681,
+      "epoch": 0.9469636776390465,
       "grad_norm": 0.0,
-      "learning_rate": 3.633931060931151e-07,
-      "loss": 0.7832,
+      "learning_rate": 1.4717202318825119e-07,
+      "loss": 0.8777,
       "step": 33371
     },
     {
-      "epoch": 0.9165362115844113,
+      "epoch": 0.9469920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 3.6315552840658683e-07,
-      "loss": 0.7999,
+      "learning_rate": 1.470149647884922e-07,
+      "loss": 0.8137,
       "step": 33372
     },
     {
-      "epoch": 0.9165636758122545,
+      "epoch": 0.9470204313280364,
       "grad_norm": 0.0,
-      "learning_rate": 3.6291802697050347e-07,
-      "loss": 0.824,
+      "learning_rate": 1.4685798961734055e-07,
+      "loss": 0.7284,
       "step": 33373
     },
     {
-      "epoch": 0.9165911400400978,
+      "epoch": 0.9470488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 3.626806017867468e-07,
-      "loss": 0.736,
+      "learning_rate": 1.4670109767612183e-07,
+      "loss": 0.8663,
       "step": 33374
     },
     {
-      "epoch": 0.9166186042679411,
+      "epoch": 0.9470771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 3.624432528571953e-07,
-      "loss": 0.8107,
+      "learning_rate": 1.465442889661628e-07,
+      "loss": 0.7445,
       "step": 33375
     },
     {
-      "epoch": 0.9166460684957842,
+      "epoch": 0.947105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.622059801837241e-07,
-      "loss": 0.7131,
+      "learning_rate": 1.4638756348878569e-07,
+      "loss": 0.9106,
       "step": 33376
     },
     {
-      "epoch": 0.9166735327236275,
+      "epoch": 0.9471339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 3.6196878376821286e-07,
-      "loss": 0.8726,
+      "learning_rate": 1.4623092124531613e-07,
+      "loss": 0.7888,
       "step": 33377
     },
     {
-      "epoch": 0.9167009969514707,
+      "epoch": 0.9471623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 3.6173166361253896e-07,
-      "loss": 0.7348,
+      "learning_rate": 1.4607436223707749e-07,
+      "loss": 0.7298,
       "step": 33378
     },
     {
-      "epoch": 0.916728461179314,
+      "epoch": 0.9471906923950056,
       "grad_norm": 0.0,
-      "learning_rate": 3.6149461971857645e-07,
-      "loss": 0.7401,
+      "learning_rate": 1.4591788646539207e-07,
+      "loss": 0.794,
       "step": 33379
     },
     {
-      "epoch": 0.9167559254071572,
+      "epoch": 0.9472190692395006,
       "grad_norm": 0.0,
-      "learning_rate": 3.6125765208820387e-07,
-      "loss": 0.7507,
+      "learning_rate": 1.457614939315799e-07,
+      "loss": 0.7057,
       "step": 33380
     },
     {
-      "epoch": 0.9167833896350004,
+      "epoch": 0.9472474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 3.610207607232941e-07,
-      "loss": 0.7359,
+      "learning_rate": 1.4560518463696549e-07,
+      "loss": 0.7966,
       "step": 33381
     },
     {
-      "epoch": 0.9168108538628437,
+      "epoch": 0.9472758229284903,
       "grad_norm": 0.0,
-      "learning_rate": 3.6078394562572007e-07,
-      "loss": 0.8769,
+      "learning_rate": 1.4544895858286555e-07,
+      "loss": 0.8317,
       "step": 33382
     },
     {
-      "epoch": 0.9168383180906868,
+      "epoch": 0.9473041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 3.6054720679735813e-07,
-      "loss": 0.7505,
+      "learning_rate": 1.452928157706013e-07,
+      "loss": 0.8506,
       "step": 33383
     },
     {
-      "epoch": 0.9168657823185301,
+      "epoch": 0.9473325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 3.6031054424008116e-07,
-      "loss": 0.8404,
+      "learning_rate": 1.451367562014927e-07,
+      "loss": 0.7067,
       "step": 33384
     },
     {
-      "epoch": 0.9168932465463734,
+      "epoch": 0.947360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 3.600739579557599e-07,
-      "loss": 0.8546,
+      "learning_rate": 1.4498077987685543e-07,
+      "loss": 0.8749,
       "step": 33385
     },
     {
-      "epoch": 0.9169207107742166,
+      "epoch": 0.9473893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 3.598374479462685e-07,
-      "loss": 0.7756,
+      "learning_rate": 1.4482488679800844e-07,
+      "loss": 0.9474,
       "step": 33386
     },
     {
-      "epoch": 0.9169481750020598,
+      "epoch": 0.9474177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 3.596010142134787e-07,
-      "loss": 0.9407,
+      "learning_rate": 1.4466907696627064e-07,
+      "loss": 0.803,
       "step": 33387
     },
     {
-      "epoch": 0.9169756392299031,
+      "epoch": 0.9474460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 3.59364656759259e-07,
-      "loss": 0.709,
+      "learning_rate": 1.4451335038295433e-07,
+      "loss": 0.8324,
       "step": 33388
     },
     {
-      "epoch": 0.9170031034577463,
+      "epoch": 0.9474744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 3.591283755854813e-07,
-      "loss": 0.8799,
+      "learning_rate": 1.4435770704937736e-07,
+      "loss": 0.885,
       "step": 33389
     },
     {
-      "epoch": 0.9170305676855895,
+      "epoch": 0.9475028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 3.5889217069401294e-07,
-      "loss": 0.8028,
+      "learning_rate": 1.4420214696685418e-07,
+      "loss": 0.7926,
       "step": 33390
     },
     {
-      "epoch": 0.9170580319134327,
+      "epoch": 0.9475312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 3.586560420867247e-07,
-      "loss": 0.8761,
+      "learning_rate": 1.4404667013669827e-07,
+      "loss": 0.8416,
       "step": 33391
     },
     {
-      "epoch": 0.917085496141276,
+      "epoch": 0.9475595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 3.5841998976548497e-07,
-      "loss": 0.8209,
+      "learning_rate": 1.4389127656022296e-07,
+      "loss": 0.8548,
       "step": 33392
     },
     {
-      "epoch": 0.9171129603691193,
+      "epoch": 0.9475879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 3.581840137321624e-07,
-      "loss": 0.7209,
+      "learning_rate": 1.4373596623874054e-07,
+      "loss": 0.7885,
       "step": 33393
     },
     {
-      "epoch": 0.9171404245969624,
+      "epoch": 0.9476163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 3.5794811398862095e-07,
-      "loss": 0.8768,
+      "learning_rate": 1.435807391735644e-07,
+      "loss": 0.7245,
       "step": 33394
     },
     {
-      "epoch": 0.9171678888248057,
+      "epoch": 0.947644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 3.577122905367292e-07,
-      "loss": 0.7435,
+      "learning_rate": 1.4342559536600576e-07,
+      "loss": 0.7348,
       "step": 33395
     },
     {
-      "epoch": 0.9171953530526489,
+      "epoch": 0.9476730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 3.5747654337835337e-07,
-      "loss": 0.8188,
+      "learning_rate": 1.432705348173724e-07,
+      "loss": 0.8127,
       "step": 33396
     },
     {
-      "epoch": 0.9172228172804922,
+      "epoch": 0.9477014755959138,
       "grad_norm": 0.0,
-      "learning_rate": 3.5724087251535753e-07,
-      "loss": 0.9007,
+      "learning_rate": 1.4311555752897666e-07,
+      "loss": 0.8532,
       "step": 33397
     },
     {
-      "epoch": 0.9172502815083354,
+      "epoch": 0.9477298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 3.570052779496092e-07,
-      "loss": 0.7935,
+      "learning_rate": 1.4296066350212746e-07,
+      "loss": 0.7146,
       "step": 33398
     },
     {
-      "epoch": 0.9172777457361786,
+      "epoch": 0.9477582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 3.567697596829678e-07,
-      "loss": 0.8168,
+      "learning_rate": 1.4280585273813152e-07,
+      "loss": 0.7186,
       "step": 33399
     },
     {
-      "epoch": 0.9173052099640219,
+      "epoch": 0.9477866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 3.5653431771730087e-07,
-      "loss": 0.8833,
+      "learning_rate": 1.4265112523829782e-07,
+      "loss": 0.7722,
       "step": 33400
     },
     {
-      "epoch": 0.9173326741918651,
+      "epoch": 0.9478149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 3.562989520544691e-07,
-      "loss": 0.8348,
+      "learning_rate": 1.424964810039342e-07,
+      "loss": 0.7858,
       "step": 33401
     },
     {
-      "epoch": 0.9173601384197083,
+      "epoch": 0.9478433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 3.5606366269633543e-07,
-      "loss": 0.8119,
+      "learning_rate": 1.4234192003634517e-07,
+      "loss": 0.8673,
       "step": 33402
     },
     {
-      "epoch": 0.9173876026475516,
+      "epoch": 0.947871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 3.558284496447628e-07,
-      "loss": 0.888,
+      "learning_rate": 1.4218744233683746e-07,
+      "loss": 0.7733,
       "step": 33403
     },
     {
-      "epoch": 0.9174150668753948,
+      "epoch": 0.947900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 3.55593312901612e-07,
-      "loss": 0.7828,
+      "learning_rate": 1.4203304790671557e-07,
+      "loss": 0.911,
       "step": 33404
     },
     {
-      "epoch": 0.917442531103238,
+      "epoch": 0.9479284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 3.5535825246874267e-07,
-      "loss": 0.7722,
+      "learning_rate": 1.4187873674728292e-07,
+      "loss": 0.7657,
       "step": 33405
     },
     {
-      "epoch": 0.9174699953310813,
+      "epoch": 0.9479568671963677,
       "grad_norm": 0.0,
-      "learning_rate": 3.5512326834801324e-07,
-      "loss": 0.8066,
+      "learning_rate": 1.4172450885984402e-07,
+      "loss": 0.9008,
       "step": 33406
     },
     {
-      "epoch": 0.9174974595589245,
+      "epoch": 0.9479852440408627,
       "grad_norm": 0.0,
-      "learning_rate": 3.548883605412845e-07,
-      "loss": 0.8912,
+      "learning_rate": 1.4157036424570114e-07,
+      "loss": 0.8139,
       "step": 33407
     },
     {
-      "epoch": 0.9175249237867678,
+      "epoch": 0.9480136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 3.5465352905041604e-07,
-      "loss": 0.8758,
+      "learning_rate": 1.414163029061577e-07,
+      "loss": 0.7898,
       "step": 33408
     },
     {
-      "epoch": 0.9175523880146109,
+      "epoch": 0.9480419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 3.544187738772653e-07,
-      "loss": 0.7324,
+      "learning_rate": 1.4126232484251267e-07,
+      "loss": 0.774,
       "step": 33409
     },
     {
-      "epoch": 0.9175798522424542,
+      "epoch": 0.9480703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 3.5418409502368965e-07,
-      "loss": 0.7989,
+      "learning_rate": 1.4110843005606833e-07,
+      "loss": 0.8489,
       "step": 33410
     },
     {
-      "epoch": 0.9176073164702975,
+      "epoch": 0.9480987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 3.539494924915454e-07,
-      "loss": 0.7857,
+      "learning_rate": 1.4095461854812476e-07,
+      "loss": 0.8772,
       "step": 33411
     },
     {
-      "epoch": 0.9176347806981406,
+      "epoch": 0.9481271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 3.5371496628268996e-07,
-      "loss": 0.7464,
+      "learning_rate": 1.408008903199809e-07,
+      "loss": 0.7513,
       "step": 33412
     },
     {
-      "epoch": 0.9176622449259839,
+      "epoch": 0.948155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 3.534805163989774e-07,
-      "loss": 0.7829,
+      "learning_rate": 1.4064724537293462e-07,
+      "loss": 0.801,
       "step": 33413
     },
     {
-      "epoch": 0.9176897091538272,
+      "epoch": 0.948183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 3.53246142842264e-07,
-      "loss": 0.8296,
+      "learning_rate": 1.4049368370828376e-07,
+      "loss": 0.828,
       "step": 33414
     },
     {
-      "epoch": 0.9177171733816704,
+      "epoch": 0.9482122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 3.5301184561440385e-07,
-      "loss": 0.7462,
+      "learning_rate": 1.4034020532732728e-07,
+      "loss": 0.8205,
       "step": 33415
     },
     {
-      "epoch": 0.9177446376095136,
+      "epoch": 0.9482406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 3.5277762471725206e-07,
-      "loss": 0.8013,
+      "learning_rate": 1.4018681023135861e-07,
+      "loss": 0.7829,
       "step": 33416
     },
     {
-      "epoch": 0.9177721018373568,
+      "epoch": 0.9482690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 3.525434801526595e-07,
-      "loss": 0.7974,
+      "learning_rate": 1.400334984216767e-07,
+      "loss": 0.897,
       "step": 33417
     },
     {
-      "epoch": 0.9177995660652001,
+      "epoch": 0.9482973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 3.52309411922479e-07,
-      "loss": 0.7484,
+      "learning_rate": 1.3988026989957493e-07,
+      "loss": 0.8071,
       "step": 33418
     },
     {
-      "epoch": 0.9178270302930434,
+      "epoch": 0.9483257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 3.520754200285648e-07,
-      "loss": 0.8279,
+      "learning_rate": 1.3972712466634676e-07,
+      "loss": 0.7445,
       "step": 33419
     },
     {
-      "epoch": 0.9178544945208865,
+      "epoch": 0.9483541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 3.518415044727674e-07,
-      "loss": 0.7362,
+      "learning_rate": 1.3957406272328666e-07,
+      "loss": 0.749,
       "step": 33420
     },
     {
-      "epoch": 0.9178819587487298,
+      "epoch": 0.9483825198637912,
       "grad_norm": 0.0,
-      "learning_rate": 3.5160766525693665e-07,
-      "loss": 0.8059,
+      "learning_rate": 1.394210840716892e-07,
+      "loss": 0.8284,
       "step": 33421
     },
     {
-      "epoch": 0.917909422976573,
+      "epoch": 0.948410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 3.5137390238292546e-07,
-      "loss": 0.8278,
+      "learning_rate": 1.392681887128433e-07,
+      "loss": 0.8092,
       "step": 33422
     },
     {
-      "epoch": 0.9179368872044162,
+      "epoch": 0.9484392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 3.5114021585257894e-07,
-      "loss": 0.7903,
+      "learning_rate": 1.3911537664804352e-07,
+      "loss": 0.7454,
       "step": 33423
     },
     {
-      "epoch": 0.9179643514322595,
+      "epoch": 0.9484676503972759,
       "grad_norm": 0.0,
-      "learning_rate": 3.5090660566774903e-07,
-      "loss": 0.8022,
+      "learning_rate": 1.3896264787857994e-07,
+      "loss": 0.7685,
       "step": 33424
     },
     {
-      "epoch": 0.9179918156601027,
+      "epoch": 0.9484960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 3.506730718302842e-07,
-      "loss": 0.7234,
+      "learning_rate": 1.388100024057404e-07,
+      "loss": 0.8183,
       "step": 33425
     },
     {
-      "epoch": 0.918019279887946,
+      "epoch": 0.9485244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 3.504396143420319e-07,
-      "loss": 0.7867,
+      "learning_rate": 1.3865744023081828e-07,
+      "loss": 0.9327,
       "step": 33426
     },
     {
-      "epoch": 0.9180467441157892,
+      "epoch": 0.9485527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 3.5020623320483835e-07,
-      "loss": 0.8804,
+      "learning_rate": 1.3850496135509815e-07,
+      "loss": 0.8783,
       "step": 33427
     },
     {
-      "epoch": 0.9180742083436324,
+      "epoch": 0.9485811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 3.499729284205533e-07,
-      "loss": 0.8386,
+      "learning_rate": 1.383525657798701e-07,
+      "loss": 0.8256,
       "step": 33428
     },
     {
-      "epoch": 0.9181016725714757,
+      "epoch": 0.9486095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 3.4973969999101963e-07,
-      "loss": 0.7784,
+      "learning_rate": 1.3820025350642308e-07,
+      "loss": 0.8383,
       "step": 33429
     },
     {
-      "epoch": 0.9181291367993188,
+      "epoch": 0.9486379114642451,
       "grad_norm": 0.0,
-      "learning_rate": 3.495065479180826e-07,
-      "loss": 0.8973,
+      "learning_rate": 1.380480245360405e-07,
+      "loss": 0.6982,
       "step": 33430
     },
     {
-      "epoch": 0.9181566010271621,
+      "epoch": 0.9486662883087401,
       "grad_norm": 0.0,
-      "learning_rate": 3.4927347220358844e-07,
-      "loss": 0.9097,
+      "learning_rate": 1.378958788700091e-07,
+      "loss": 0.8205,
       "step": 33431
     },
     {
-      "epoch": 0.9181840652550054,
+      "epoch": 0.948694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 3.4904047284938127e-07,
-      "loss": 0.8148,
+      "learning_rate": 1.3774381650961567e-07,
+      "loss": 0.7743,
       "step": 33432
     },
     {
-      "epoch": 0.9182115294828486,
+      "epoch": 0.9487230419977298,
       "grad_norm": 0.0,
-      "learning_rate": 3.488075498573029e-07,
-      "loss": 0.7821,
+      "learning_rate": 1.3759183745614246e-07,
+      "loss": 0.917,
       "step": 33433
     },
     {
-      "epoch": 0.9182389937106918,
+      "epoch": 0.9487514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 3.485747032291997e-07,
-      "loss": 0.8687,
+      "learning_rate": 1.3743994171087406e-07,
+      "loss": 0.8104,
       "step": 33434
     },
     {
-      "epoch": 0.918266457938535,
+      "epoch": 0.9487797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 3.483419329669113e-07,
-      "loss": 0.8656,
+      "learning_rate": 1.3728812927509605e-07,
+      "loss": 0.8583,
       "step": 33435
     },
     {
-      "epoch": 0.9182939221663783,
+      "epoch": 0.9488081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 3.4810923907228066e-07,
-      "loss": 0.6949,
+      "learning_rate": 1.3713640015008634e-07,
+      "loss": 0.7867,
       "step": 33436
     },
     {
-      "epoch": 0.9183213863942216,
+      "epoch": 0.9488365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 3.4787662154714854e-07,
-      "loss": 0.7921,
+      "learning_rate": 1.3698475433712942e-07,
+      "loss": 0.919,
       "step": 33437
     },
     {
-      "epoch": 0.9183488506220647,
+      "epoch": 0.9488649262202044,
       "grad_norm": 0.0,
-      "learning_rate": 3.476440803933545e-07,
-      "loss": 0.8681,
+      "learning_rate": 1.3683319183750544e-07,
+      "loss": 0.7642,
       "step": 33438
     },
     {
-      "epoch": 0.918376314849908,
+      "epoch": 0.9488933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 3.474116156127394e-07,
-      "loss": 0.7711,
+      "learning_rate": 1.3668171265249552e-07,
+      "loss": 0.8335,
       "step": 33439
     },
     {
-      "epoch": 0.9184037790777513,
+      "epoch": 0.9489216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 3.4717922720714394e-07,
-      "loss": 0.8084,
+      "learning_rate": 1.3653031678337869e-07,
+      "loss": 0.9676,
       "step": 33440
     },
     {
-      "epoch": 0.9184312433055944,
+      "epoch": 0.948950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 3.469469151784044e-07,
-      "loss": 0.7402,
+      "learning_rate": 1.363790042314328e-07,
+      "loss": 0.7456,
       "step": 33441
     },
     {
-      "epoch": 0.9184587075334377,
+      "epoch": 0.9489784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 3.4671467952835934e-07,
-      "loss": 0.8268,
+      "learning_rate": 1.362277749979368e-07,
+      "loss": 0.691,
       "step": 33442
     },
     {
-      "epoch": 0.9184861717612809,
+      "epoch": 0.9490068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 3.4648252025884845e-07,
-      "loss": 0.7247,
+      "learning_rate": 1.3607662908416973e-07,
+      "loss": 0.7186,
       "step": 33443
     },
     {
-      "epoch": 0.9185136359891242,
+      "epoch": 0.9490351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 3.4625043737170793e-07,
-      "loss": 0.859,
+      "learning_rate": 1.359255664914061e-07,
+      "loss": 0.7938,
       "step": 33444
     },
     {
-      "epoch": 0.9185411002169674,
+      "epoch": 0.9490635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 3.4601843086877194e-07,
-      "loss": 0.7638,
+      "learning_rate": 1.357745872209215e-07,
+      "loss": 0.6964,
       "step": 33445
     },
     {
-      "epoch": 0.9185685644448106,
+      "epoch": 0.9490919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 3.4578650075187905e-07,
-      "loss": 0.7451,
+      "learning_rate": 1.356236912739939e-07,
+      "loss": 0.7322,
       "step": 33446
     },
     {
-      "epoch": 0.9185960286726539,
+      "epoch": 0.9491203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 3.455546470228621e-07,
-      "loss": 0.7373,
+      "learning_rate": 1.354728786518955e-07,
+      "loss": 0.8132,
       "step": 33447
     },
     {
-      "epoch": 0.918623492900497,
+      "epoch": 0.9491486946651533,
       "grad_norm": 0.0,
-      "learning_rate": 3.4532286968355644e-07,
-      "loss": 0.7649,
+      "learning_rate": 1.3532214935590094e-07,
+      "loss": 0.87,
       "step": 33448
     },
     {
-      "epoch": 0.9186509571283403,
+      "epoch": 0.9491770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 3.4509116873579606e-07,
-      "loss": 0.8315,
+      "learning_rate": 1.3517150338728468e-07,
+      "loss": 0.7538,
       "step": 33449
     },
     {
-      "epoch": 0.9186784213561836,
+      "epoch": 0.949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 3.44859544181414e-07,
-      "loss": 0.7738,
+      "learning_rate": 1.3502094074731797e-07,
+      "loss": 0.7953,
       "step": 33450
     },
     {
-      "epoch": 0.9187058855840268,
+      "epoch": 0.9492338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 3.446279960222454e-07,
-      "loss": 0.8211,
+      "learning_rate": 1.34870461437272e-07,
+      "loss": 0.8206,
       "step": 33451
     },
     {
-      "epoch": 0.91873334981187,
+      "epoch": 0.9492622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 3.443965242601177e-07,
-      "loss": 0.7347,
+      "learning_rate": 1.347200654584191e-07,
+      "loss": 0.839,
       "step": 33452
     },
     {
-      "epoch": 0.9187608140397133,
+      "epoch": 0.9492905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 3.441651288968673e-07,
-      "loss": 0.6789,
+      "learning_rate": 1.3456975281202822e-07,
+      "loss": 0.7893,
       "step": 33453
     },
     {
-      "epoch": 0.9187882782675565,
+      "epoch": 0.9493189557321225,
       "grad_norm": 0.0,
-      "learning_rate": 3.439338099343215e-07,
-      "loss": 0.8302,
+      "learning_rate": 1.3441952349937171e-07,
+      "loss": 0.8203,
       "step": 33454
     },
     {
-      "epoch": 0.9188157424953998,
+      "epoch": 0.9493473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 3.437025673743111e-07,
-      "loss": 0.8131,
+      "learning_rate": 1.3426937752171633e-07,
+      "loss": 0.8756,
       "step": 33455
     },
     {
-      "epoch": 0.9188432067232429,
+      "epoch": 0.9493757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 3.4347140121866684e-07,
-      "loss": 0.8135,
+      "learning_rate": 1.3411931488033104e-07,
+      "loss": 0.9141,
       "step": 33456
     },
     {
-      "epoch": 0.9188706709510862,
+      "epoch": 0.9494040862656072,
       "grad_norm": 0.0,
-      "learning_rate": 3.432403114692173e-07,
-      "loss": 0.8035,
+      "learning_rate": 1.3396933557648485e-07,
+      "loss": 0.7214,
       "step": 33457
     },
     {
-      "epoch": 0.9188981351789295,
+      "epoch": 0.9494324631101022,
       "grad_norm": 0.0,
-      "learning_rate": 3.4300929812779217e-07,
-      "loss": 0.7457,
+      "learning_rate": 1.3381943961144118e-07,
+      "loss": 0.7786,
       "step": 33458
     },
     {
-      "epoch": 0.9189255994067727,
+      "epoch": 0.9494608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 3.4277836119621766e-07,
-      "loss": 0.7974,
+      "learning_rate": 1.3366962698646902e-07,
+      "loss": 0.8003,
       "step": 33459
     },
     {
-      "epoch": 0.9189530636346159,
+      "epoch": 0.9494892167990919,
       "grad_norm": 0.0,
-      "learning_rate": 3.4254750067632234e-07,
-      "loss": 0.7966,
+      "learning_rate": 1.335198977028329e-07,
+      "loss": 0.8034,
       "step": 33460
     },
     {
-      "epoch": 0.9189805278624591,
+      "epoch": 0.9495175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 3.4231671656993036e-07,
-      "loss": 0.7991,
+      "learning_rate": 1.3337025176179742e-07,
+      "loss": 0.6811,
       "step": 33461
     },
     {
-      "epoch": 0.9190079920903024,
+      "epoch": 0.9495459704880818,
       "grad_norm": 0.0,
-      "learning_rate": 3.420860088788702e-07,
-      "loss": 0.6904,
+      "learning_rate": 1.3322068916462706e-07,
+      "loss": 0.8831,
       "step": 33462
     },
     {
-      "epoch": 0.9190354563181456,
+      "epoch": 0.9495743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 3.418553776049682e-07,
-      "loss": 0.8076,
+      "learning_rate": 1.3307120991258637e-07,
+      "loss": 0.7987,
       "step": 33463
     },
     {
-      "epoch": 0.9190629205459888,
+      "epoch": 0.9496027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 3.4162482275004517e-07,
-      "loss": 0.7926,
+      "learning_rate": 1.329218140069355e-07,
+      "loss": 0.7973,
       "step": 33464
     },
     {
-      "epoch": 0.9190903847738321,
+      "epoch": 0.9496311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 3.413943443159296e-07,
-      "loss": 0.8761,
+      "learning_rate": 1.3277250144893782e-07,
+      "loss": 0.9111,
       "step": 33465
     },
     {
-      "epoch": 0.9191178490016754,
+      "epoch": 0.9496594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 3.4116394230444214e-07,
-      "loss": 0.7265,
+      "learning_rate": 1.326232722398546e-07,
+      "loss": 0.78,
       "step": 33466
     },
     {
-      "epoch": 0.9191453132295185,
+      "epoch": 0.9496878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 3.409336167174071e-07,
-      "loss": 0.7297,
+      "learning_rate": 1.3247412638094593e-07,
+      "loss": 0.8833,
       "step": 33467
     },
     {
-      "epoch": 0.9191727774573618,
+      "epoch": 0.949716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 3.4070336755664846e-07,
-      "loss": 0.834,
+      "learning_rate": 1.3232506387347078e-07,
+      "loss": 0.7364,
       "step": 33468
     },
     {
-      "epoch": 0.919200241685205,
+      "epoch": 0.949744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 3.404731948239859e-07,
-      "loss": 0.9019,
+      "learning_rate": 1.3217608471869149e-07,
+      "loss": 0.7893,
       "step": 33469
     },
     {
-      "epoch": 0.9192277059130483,
+      "epoch": 0.9497729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 3.4024309852124126e-07,
-      "loss": 0.7719,
+      "learning_rate": 1.320271889178626e-07,
+      "loss": 0.8087,
       "step": 33470
     },
     {
-      "epoch": 0.9192551701408915,
+      "epoch": 0.9498013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 3.4001307865023313e-07,
-      "loss": 0.7949,
+      "learning_rate": 1.3187837647224421e-07,
+      "loss": 0.7966,
       "step": 33471
     },
     {
-      "epoch": 0.9192826343687347,
+      "epoch": 0.9498297389330307,
       "grad_norm": 0.0,
-      "learning_rate": 3.397831352127845e-07,
-      "loss": 0.8124,
+      "learning_rate": 1.3172964738309423e-07,
+      "loss": 0.733,
       "step": 33472
     },
     {
-      "epoch": 0.919310098596578,
+      "epoch": 0.9498581157775255,
       "grad_norm": 0.0,
-      "learning_rate": 3.395532682107139e-07,
-      "loss": 0.8614,
+      "learning_rate": 1.315810016516661e-07,
+      "loss": 0.8,
       "step": 33473
     },
     {
-      "epoch": 0.9193375628244211,
+      "epoch": 0.9498864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 3.393234776458398e-07,
-      "loss": 0.8478,
+      "learning_rate": 1.3143243927921656e-07,
+      "loss": 0.8152,
       "step": 33474
     },
     {
-      "epoch": 0.9193650270522644,
+      "epoch": 0.9499148694665154,
       "grad_norm": 0.0,
-      "learning_rate": 3.3909376351998203e-07,
-      "loss": 0.7851,
+      "learning_rate": 1.312839602670024e-07,
+      "loss": 0.752,
       "step": 33475
     },
     {
-      "epoch": 0.9193924912801077,
+      "epoch": 0.9499432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 3.3886412583495456e-07,
-      "loss": 0.845,
+      "learning_rate": 1.3113556461627486e-07,
+      "loss": 0.776,
       "step": 33476
     },
     {
-      "epoch": 0.9194199555079509,
+      "epoch": 0.9499716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 3.3863456459257816e-07,
-      "loss": 0.7774,
+      "learning_rate": 1.3098725232828958e-07,
+      "loss": 0.8419,
       "step": 33477
     },
     {
-      "epoch": 0.9194474197357941,
+      "epoch": 0.95,
       "grad_norm": 0.0,
-      "learning_rate": 3.3840507979466695e-07,
-      "loss": 0.9656,
+      "learning_rate": 1.3083902340429777e-07,
+      "loss": 0.6902,
       "step": 33478
     },
     {
-      "epoch": 0.9194748839636374,
+      "epoch": 0.9500283768444949,
       "grad_norm": 0.0,
-      "learning_rate": 3.381756714430362e-07,
-      "loss": 0.9191,
+      "learning_rate": 1.306908778455529e-07,
+      "loss": 0.6773,
       "step": 33479
     },
     {
-      "epoch": 0.9195023481914806,
+      "epoch": 0.9500567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 3.379463395395033e-07,
-      "loss": 0.8128,
+      "learning_rate": 1.3054281565330619e-07,
+      "loss": 0.763,
       "step": 33480
     },
     {
-      "epoch": 0.9195298124193239,
+      "epoch": 0.9500851305334846,
       "grad_norm": 0.0,
-      "learning_rate": 3.377170840858823e-07,
-      "loss": 0.795,
+      "learning_rate": 1.3039483682880772e-07,
+      "loss": 0.7847,
       "step": 33481
     },
     {
-      "epoch": 0.919557276647167,
+      "epoch": 0.9501135073779796,
       "grad_norm": 0.0,
-      "learning_rate": 3.3748790508398523e-07,
-      "loss": 0.7822,
+      "learning_rate": 1.3024694137330652e-07,
+      "loss": 0.7272,
       "step": 33482
     },
     {
-      "epoch": 0.9195847408750103,
+      "epoch": 0.9501418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 3.3725880253562715e-07,
-      "loss": 0.8695,
+      "learning_rate": 1.300991292880549e-07,
+      "loss": 0.7971,
       "step": 33483
     },
     {
-      "epoch": 0.9196122051028536,
+      "epoch": 0.9501702610669693,
       "grad_norm": 0.0,
-      "learning_rate": 3.370297764426211e-07,
-      "loss": 0.9009,
+      "learning_rate": 1.2995140057429855e-07,
+      "loss": 0.7062,
       "step": 33484
     },
     {
-      "epoch": 0.9196396693306967,
+      "epoch": 0.9501986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 3.3680082680677795e-07,
-      "loss": 0.8585,
+      "learning_rate": 1.2980375523328648e-07,
+      "loss": 0.8293,
       "step": 33485
     },
     {
-      "epoch": 0.91966713355854,
+      "epoch": 0.9502270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 3.365719536299095e-07,
-      "loss": 0.8851,
+      "learning_rate": 1.2965619326626655e-07,
+      "loss": 0.8508,
       "step": 33486
     },
     {
-      "epoch": 0.9196945977863832,
+      "epoch": 0.950255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 3.3634315691382755e-07,
-      "loss": 0.819,
+      "learning_rate": 1.2950871467448333e-07,
+      "loss": 0.7851,
       "step": 33487
     },
     {
-      "epoch": 0.9197220620142265,
+      "epoch": 0.9502837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 3.361144366603419e-07,
-      "loss": 0.8331,
+      "learning_rate": 1.2936131945918472e-07,
+      "loss": 0.6464,
       "step": 33488
     },
     {
-      "epoch": 0.9197495262420697,
+      "epoch": 0.9503121452894439,
       "grad_norm": 0.0,
-      "learning_rate": 3.358857928712611e-07,
-      "loss": 0.7314,
+      "learning_rate": 1.2921400762161417e-07,
+      "loss": 0.9062,
       "step": 33489
     },
     {
-      "epoch": 0.9197769904699129,
+      "epoch": 0.9503405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 3.356572255483959e-07,
-      "loss": 0.781,
+      "learning_rate": 1.2906677916301736e-07,
+      "loss": 0.7929,
       "step": 33490
     },
     {
-      "epoch": 0.9198044546977562,
+      "epoch": 0.9503688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.3542873469355475e-07,
-      "loss": 0.8529,
+      "learning_rate": 1.2891963408463658e-07,
+      "loss": 0.7471,
       "step": 33491
     },
     {
-      "epoch": 0.9198319189255995,
+      "epoch": 0.9503972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 3.3520032030854634e-07,
-      "loss": 0.683,
+      "learning_rate": 1.2877257238771535e-07,
+      "loss": 0.7541,
       "step": 33492
     },
     {
-      "epoch": 0.9198593831534426,
+      "epoch": 0.9504256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 3.349719823951758e-07,
-      "loss": 0.8552,
+      "learning_rate": 1.2862559407349595e-07,
+      "loss": 0.7689,
       "step": 33493
     },
     {
-      "epoch": 0.9198868473812859,
+      "epoch": 0.9504540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 3.3474372095525066e-07,
-      "loss": 0.9613,
+      "learning_rate": 1.2847869914321965e-07,
+      "loss": 0.96,
       "step": 33494
     },
     {
-      "epoch": 0.9199143116091291,
+      "epoch": 0.9504824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 3.345155359905761e-07,
-      "loss": 0.7754,
+      "learning_rate": 1.2833188759812877e-07,
+      "loss": 0.8039,
       "step": 33495
     },
     {
-      "epoch": 0.9199417758369723,
+      "epoch": 0.9505107832009081,
       "grad_norm": 0.0,
-      "learning_rate": 3.3428742750295953e-07,
-      "loss": 0.8495,
+      "learning_rate": 1.2818515943946118e-07,
+      "loss": 0.7985,
       "step": 33496
     },
     {
-      "epoch": 0.9199692400648156,
+      "epoch": 0.950539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 3.3405939549420507e-07,
-      "loss": 0.8207,
+      "learning_rate": 1.2803851466845708e-07,
+      "loss": 0.8186,
       "step": 33497
     },
     {
-      "epoch": 0.9199967042926588,
+      "epoch": 0.9505675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 3.338314399661169e-07,
-      "loss": 0.8514,
+      "learning_rate": 1.2789195328635651e-07,
+      "loss": 0.6094,
       "step": 33498
     },
     {
-      "epoch": 0.9200241685205021,
+      "epoch": 0.9505959137343928,
       "grad_norm": 0.0,
-      "learning_rate": 3.3360356092049905e-07,
-      "loss": 0.8765,
+      "learning_rate": 1.2774547529439518e-07,
+      "loss": 0.7701,
       "step": 33499
     },
     {
-      "epoch": 0.9200516327483452,
+      "epoch": 0.9506242905788876,
       "grad_norm": 0.0,
-      "learning_rate": 3.3337575835915346e-07,
-      "loss": 0.7614,
+      "learning_rate": 1.275990806938121e-07,
+      "loss": 0.862,
       "step": 33500
     },
     {
-      "epoch": 0.9200790969761885,
+      "epoch": 0.9506526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 3.3314803228388537e-07,
-      "loss": 0.8045,
+      "learning_rate": 1.2745276948584296e-07,
+      "loss": 0.7304,
       "step": 33501
     },
     {
-      "epoch": 0.9201065612040318,
+      "epoch": 0.9506810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 3.3292038269649217e-07,
-      "loss": 0.7182,
+      "learning_rate": 1.2730654167172452e-07,
+      "loss": 0.7755,
       "step": 33502
     },
     {
-      "epoch": 0.920134025431875,
+      "epoch": 0.9507094211123723,
       "grad_norm": 0.0,
-      "learning_rate": 3.3269280959877804e-07,
-      "loss": 0.7683,
+      "learning_rate": 1.2716039725269247e-07,
+      "loss": 0.7771,
       "step": 33503
     },
     {
-      "epoch": 0.9201614896597182,
+      "epoch": 0.9507377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 3.324653129925437e-07,
-      "loss": 0.9004,
+      "learning_rate": 1.2701433622997916e-07,
+      "loss": 0.8677,
       "step": 33504
     },
     {
-      "epoch": 0.9201889538875615,
+      "epoch": 0.950766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 3.3223789287958886e-07,
-      "loss": 0.8396,
+      "learning_rate": 1.2686835860481916e-07,
+      "loss": 0.8019,
       "step": 33505
     },
     {
-      "epoch": 0.9202164181154047,
+      "epoch": 0.950794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 3.3201054926171206e-07,
-      "loss": 0.8088,
+      "learning_rate": 1.2672246437844705e-07,
+      "loss": 0.7653,
       "step": 33506
     },
     {
-      "epoch": 0.9202438823432479,
+      "epoch": 0.9508229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 3.31783282140713e-07,
-      "loss": 0.7229,
+      "learning_rate": 1.2657665355209403e-07,
+      "loss": 0.7518,
       "step": 33507
     },
     {
-      "epoch": 0.9202713465710911,
+      "epoch": 0.9508513053348467,
       "grad_norm": 0.0,
-      "learning_rate": 3.315560915183913e-07,
-      "loss": 0.8102,
+      "learning_rate": 1.2643092612699136e-07,
+      "loss": 0.8124,
       "step": 33508
     },
     {
-      "epoch": 0.9202988107989344,
+      "epoch": 0.9508796821793417,
       "grad_norm": 0.0,
-      "learning_rate": 3.313289773965411e-07,
-      "loss": 0.7832,
+      "learning_rate": 1.262852821043714e-07,
+      "loss": 0.8844,
       "step": 33509
     },
     {
-      "epoch": 0.9203262750267777,
+      "epoch": 0.9509080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 3.31101939776961e-07,
-      "loss": 0.8061,
+      "learning_rate": 1.2613972148546204e-07,
+      "loss": 0.8209,
       "step": 33510
     },
     {
-      "epoch": 0.9203537392546208,
+      "epoch": 0.9509364358683314,
       "grad_norm": 0.0,
-      "learning_rate": 3.3087497866144957e-07,
-      "loss": 0.7939,
+      "learning_rate": 1.2599424427149565e-07,
+      "loss": 0.8153,
       "step": 33511
     },
     {
-      "epoch": 0.9203812034824641,
+      "epoch": 0.9509648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 3.3064809405179975e-07,
-      "loss": 0.8914,
+      "learning_rate": 1.25848850463699e-07,
+      "loss": 0.8977,
       "step": 33512
     },
     {
-      "epoch": 0.9204086677103073,
+      "epoch": 0.9509931895573213,
       "grad_norm": 0.0,
-      "learning_rate": 3.304212859498068e-07,
-      "loss": 0.7623,
+      "learning_rate": 1.257035400633011e-07,
+      "loss": 0.8224,
       "step": 33513
     },
     {
-      "epoch": 0.9204361319381505,
+      "epoch": 0.9510215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 3.301945543572671e-07,
-      "loss": 0.8836,
+      "learning_rate": 1.2555831307152877e-07,
+      "loss": 0.6676,
       "step": 33514
     },
     {
-      "epoch": 0.9204635961659938,
+      "epoch": 0.951049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 3.299678992759736e-07,
-      "loss": 0.8131,
+      "learning_rate": 1.2541316948960992e-07,
+      "loss": 0.731,
       "step": 33515
     },
     {
-      "epoch": 0.920491060393837,
+      "epoch": 0.951078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 3.2974132070772155e-07,
-      "loss": 0.7557,
+      "learning_rate": 1.2526810931877021e-07,
+      "loss": 0.8928,
       "step": 33516
     },
     {
-      "epoch": 0.9205185246216803,
+      "epoch": 0.9511066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 3.2951481865430177e-07,
-      "loss": 0.7802,
+      "learning_rate": 1.2512313256023424e-07,
+      "loss": 0.7505,
       "step": 33517
     },
     {
-      "epoch": 0.9205459888495235,
+      "epoch": 0.9511350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 3.2928839311750507e-07,
-      "loss": 0.7962,
+      "learning_rate": 1.2497823921522767e-07,
+      "loss": 0.8231,
       "step": 33518
     },
     {
-      "epoch": 0.9205734530773667,
+      "epoch": 0.9511634506242905,
       "grad_norm": 0.0,
-      "learning_rate": 3.2906204409912547e-07,
-      "loss": 0.7699,
+      "learning_rate": 1.248334292849729e-07,
+      "loss": 0.7158,
       "step": 33519
     },
     {
-      "epoch": 0.92060091730521,
+      "epoch": 0.9511918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 3.288357716009527e-07,
-      "loss": 0.9176,
+      "learning_rate": 1.2468870277069444e-07,
+      "loss": 0.7794,
       "step": 33520
     },
     {
-      "epoch": 0.9206283815330532,
+      "epoch": 0.9512202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 3.2860957562477645e-07,
-      "loss": 0.7764,
+      "learning_rate": 1.245440596736147e-07,
+      "loss": 0.8323,
       "step": 33521
     },
     {
-      "epoch": 0.9206558457608964,
+      "epoch": 0.9512485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 3.2838345617239087e-07,
-      "loss": 0.8161,
+      "learning_rate": 1.2439949999495493e-07,
+      "loss": 0.8009,
       "step": 33522
     },
     {
-      "epoch": 0.9206833099887397,
+      "epoch": 0.9512769580022702,
       "grad_norm": 0.0,
-      "learning_rate": 3.281574132455789e-07,
-      "loss": 0.7875,
+      "learning_rate": 1.2425502373593634e-07,
+      "loss": 0.7782,
       "step": 33523
     },
     {
-      "epoch": 0.9207107742165829,
+      "epoch": 0.951305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 3.2793144684613255e-07,
-      "loss": 0.782,
+      "learning_rate": 1.241106308977802e-07,
+      "loss": 0.777,
       "step": 33524
     },
     {
-      "epoch": 0.9207382384444261,
+      "epoch": 0.9513337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 3.277055569758403e-07,
-      "loss": 0.823,
+      "learning_rate": 1.239663214817044e-07,
+      "loss": 0.8151,
       "step": 33525
     },
     {
-      "epoch": 0.9207657026722693,
+      "epoch": 0.9513620885357549,
       "grad_norm": 0.0,
-      "learning_rate": 3.274797436364874e-07,
-      "loss": 0.7701,
+      "learning_rate": 1.2382209548893022e-07,
+      "loss": 0.8468,
       "step": 33526
     },
     {
-      "epoch": 0.9207931669001126,
+      "epoch": 0.9513904653802497,
       "grad_norm": 0.0,
-      "learning_rate": 3.272540068298624e-07,
-      "loss": 0.8836,
+      "learning_rate": 1.236779529206744e-07,
+      "loss": 0.8762,
       "step": 33527
     },
     {
-      "epoch": 0.9208206311279559,
+      "epoch": 0.9514188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 3.270283465577506e-07,
-      "loss": 0.8115,
+      "learning_rate": 1.2353389377815494e-07,
+      "loss": 0.8099,
       "step": 33528
     },
     {
-      "epoch": 0.920848095355799,
+      "epoch": 0.9514472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 3.268027628219361e-07,
-      "loss": 0.7784,
+      "learning_rate": 1.233899180625886e-07,
+      "loss": 0.8405,
       "step": 33529
     },
     {
-      "epoch": 0.9208755595836423,
+      "epoch": 0.9514755959137344,
       "grad_norm": 0.0,
-      "learning_rate": 3.2657725562420637e-07,
-      "loss": 0.8171,
+      "learning_rate": 1.2324602577518997e-07,
+      "loss": 0.7856,
       "step": 33530
     },
     {
-      "epoch": 0.9209030238114856,
+      "epoch": 0.9515039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 3.263518249663433e-07,
-      "loss": 0.8853,
+      "learning_rate": 1.2310221691717916e-07,
+      "loss": 0.847,
       "step": 33531
     },
     {
-      "epoch": 0.9209304880393288,
+      "epoch": 0.9515323496027241,
       "grad_norm": 0.0,
-      "learning_rate": 3.261264708501333e-07,
-      "loss": 0.8503,
+      "learning_rate": 1.2295849148976524e-07,
+      "loss": 0.7231,
       "step": 33532
     },
     {
-      "epoch": 0.920957952267172,
+      "epoch": 0.9515607264472191,
       "grad_norm": 0.0,
-      "learning_rate": 3.2590119327735703e-07,
-      "loss": 0.8538,
+      "learning_rate": 1.228148494941661e-07,
+      "loss": 0.8941,
       "step": 33533
     },
     {
-      "epoch": 0.9209854164950152,
+      "epoch": 0.951589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 3.2567599224979876e-07,
-      "loss": 0.7251,
+      "learning_rate": 1.2267129093159523e-07,
+      "loss": 0.814,
       "step": 33534
     },
     {
-      "epoch": 0.9210128807228585,
+      "epoch": 0.9516174801362088,
       "grad_norm": 0.0,
-      "learning_rate": 3.254508677692392e-07,
-      "loss": 0.7611,
+      "learning_rate": 1.225278158032617e-07,
+      "loss": 0.8437,
       "step": 33535
     },
     {
-      "epoch": 0.9210403449507018,
+      "epoch": 0.9516458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 3.252258198374592e-07,
-      "loss": 0.7253,
+      "learning_rate": 1.223844241103811e-07,
+      "loss": 0.854,
       "step": 33536
     },
     {
-      "epoch": 0.9210678091785449,
+      "epoch": 0.9516742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 3.250008484562406e-07,
-      "loss": 0.8195,
+      "learning_rate": 1.2224111585416365e-07,
+      "loss": 0.8035,
       "step": 33537
     },
     {
-      "epoch": 0.9210952734063882,
+      "epoch": 0.9517026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 3.24775953627362e-07,
-      "loss": 0.833,
+      "learning_rate": 1.2209789103581836e-07,
+      "loss": 0.7627,
       "step": 33538
     },
     {
-      "epoch": 0.9211227376342314,
+      "epoch": 0.9517309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 3.245511353526054e-07,
-      "loss": 0.812,
+      "learning_rate": 1.2195474965655652e-07,
+      "loss": 0.7916,
       "step": 33539
     },
     {
-      "epoch": 0.9211502018620746,
+      "epoch": 0.9517593643586834,
       "grad_norm": 0.0,
-      "learning_rate": 3.243263936337482e-07,
-      "loss": 0.7758,
+      "learning_rate": 1.2181169171758712e-07,
+      "loss": 0.8696,
       "step": 33540
     },
     {
-      "epoch": 0.9211776660899179,
+      "epoch": 0.9517877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 3.241017284725689e-07,
-      "loss": 0.7423,
+      "learning_rate": 1.216687172201181e-07,
+      "loss": 0.8287,
       "step": 33541
     },
     {
-      "epoch": 0.9212051303177611,
+      "epoch": 0.9518161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 3.23877139870844e-07,
-      "loss": 0.7807,
+      "learning_rate": 1.2152582616535846e-07,
+      "loss": 0.7999,
       "step": 33542
     },
     {
-      "epoch": 0.9212325945456044,
+      "epoch": 0.9518444948921679,
       "grad_norm": 0.0,
-      "learning_rate": 3.236526278303509e-07,
-      "loss": 0.8776,
+      "learning_rate": 1.2138301855451397e-07,
+      "loss": 0.7476,
       "step": 33543
     },
     {
-      "epoch": 0.9212600587734476,
+      "epoch": 0.9518728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 3.234281923528659e-07,
-      "loss": 0.8285,
+      "learning_rate": 1.2124029438879027e-07,
+      "loss": 0.7315,
       "step": 33544
     },
     {
-      "epoch": 0.9212875230012908,
+      "epoch": 0.9519012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 3.232038334401666e-07,
-      "loss": 0.8974,
+      "learning_rate": 1.2109765366939418e-07,
+      "loss": 0.7604,
       "step": 33545
     },
     {
-      "epoch": 0.9213149872291341,
+      "epoch": 0.9519296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 3.22979551094027e-07,
-      "loss": 0.7871,
+      "learning_rate": 1.2095509639753034e-07,
+      "loss": 0.7729,
       "step": 33546
     },
     {
-      "epoch": 0.9213424514569772,
+      "epoch": 0.9519580022701476,
       "grad_norm": 0.0,
-      "learning_rate": 3.2275534531622134e-07,
-      "loss": 0.7449,
+      "learning_rate": 1.2081262257440329e-07,
+      "loss": 0.8755,
       "step": 33547
     },
     {
-      "epoch": 0.9213699156848205,
+      "epoch": 0.9519863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 3.225312161085237e-07,
-      "loss": 0.7994,
+      "learning_rate": 1.2067023220121654e-07,
+      "loss": 0.884,
       "step": 33548
     },
     {
-      "epoch": 0.9213973799126638,
+      "epoch": 0.9520147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 3.223071634727093e-07,
-      "loss": 0.8395,
+      "learning_rate": 1.2052792527917134e-07,
+      "loss": 0.7818,
       "step": 33549
     },
     {
-      "epoch": 0.921424844140507,
+      "epoch": 0.9520431328036323,
       "grad_norm": 0.0,
-      "learning_rate": 3.2208318741054787e-07,
-      "loss": 0.8434,
+      "learning_rate": 1.2038570180947117e-07,
+      "loss": 0.8381,
       "step": 33550
     },
     {
-      "epoch": 0.9214523083683502,
+      "epoch": 0.9520715096481271,
       "grad_norm": 0.0,
-      "learning_rate": 3.2185928792381247e-07,
-      "loss": 0.786,
+      "learning_rate": 1.2024356179331732e-07,
+      "loss": 0.867,
       "step": 33551
     },
     {
-      "epoch": 0.9214797725961934,
+      "epoch": 0.952099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 3.2163546501427724e-07,
-      "loss": 0.7635,
+      "learning_rate": 1.201015052319099e-07,
+      "loss": 0.7663,
       "step": 33552
     },
     {
-      "epoch": 0.9215072368240367,
+      "epoch": 0.9521282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 3.214117186837107e-07,
-      "loss": 0.7494,
+      "learning_rate": 1.1995953212645018e-07,
+      "loss": 0.786,
       "step": 33553
     },
     {
-      "epoch": 0.92153470105188,
+      "epoch": 0.9521566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 3.2118804893388256e-07,
-      "loss": 0.7979,
+      "learning_rate": 1.198176424781361e-07,
+      "loss": 0.8171,
       "step": 33554
     },
     {
-      "epoch": 0.9215621652797231,
+      "epoch": 0.9521850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 3.2096445576656476e-07,
-      "loss": 0.8793,
+      "learning_rate": 1.196758362881656e-07,
+      "loss": 0.7856,
       "step": 33555
     },
     {
-      "epoch": 0.9215896295075664,
+      "epoch": 0.9522133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 3.207409391835259e-07,
-      "loss": 0.8356,
+      "learning_rate": 1.195341135577377e-07,
+      "loss": 0.886,
       "step": 33556
     },
     {
-      "epoch": 0.9216170937354097,
+      "epoch": 0.9522417707150965,
       "grad_norm": 0.0,
-      "learning_rate": 3.2051749918653343e-07,
-      "loss": 0.8156,
+      "learning_rate": 1.1939247428805035e-07,
+      "loss": 0.7594,
       "step": 33557
     },
     {
-      "epoch": 0.9216445579632528,
+      "epoch": 0.9522701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 3.20294135777357e-07,
-      "loss": 0.7947,
+      "learning_rate": 1.1925091848029813e-07,
+      "loss": 0.8622,
       "step": 33558
     },
     {
-      "epoch": 0.9216720221910961,
+      "epoch": 0.9522985244040862,
       "grad_norm": 0.0,
-      "learning_rate": 3.2007084895776195e-07,
-      "loss": 0.7337,
+      "learning_rate": 1.1910944613567787e-07,
+      "loss": 0.7474,
       "step": 33559
     },
     {
-      "epoch": 0.9216994864189393,
+      "epoch": 0.9523269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 3.1984763872951576e-07,
-      "loss": 0.777,
+      "learning_rate": 1.1896805725538418e-07,
+      "loss": 0.7822,
       "step": 33560
     },
     {
-      "epoch": 0.9217269506467826,
+      "epoch": 0.9523552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 3.196245050943847e-07,
-      "loss": 0.9257,
+      "learning_rate": 1.1882675184061054e-07,
+      "loss": 0.8536,
       "step": 33561
     },
     {
-      "epoch": 0.9217544148746258,
+      "epoch": 0.9523836549375709,
       "grad_norm": 0.0,
-      "learning_rate": 3.194014480541352e-07,
-      "loss": 0.9109,
+      "learning_rate": 1.1868552989255266e-07,
+      "loss": 0.8823,
       "step": 33562
     },
     {
-      "epoch": 0.921781879102469,
+      "epoch": 0.9524120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 3.1917846761053026e-07,
-      "loss": 0.7649,
+      "learning_rate": 1.1854439141240182e-07,
+      "loss": 0.7935,
       "step": 33563
     },
     {
-      "epoch": 0.9218093433303123,
+      "epoch": 0.9524404086265608,
       "grad_norm": 0.0,
-      "learning_rate": 3.189555637653374e-07,
-      "loss": 0.7059,
+      "learning_rate": 1.184033364013515e-07,
+      "loss": 0.8458,
       "step": 33564
     },
     {
-      "epoch": 0.9218368075581556,
+      "epoch": 0.9524687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 3.1873273652031743e-07,
-      "loss": 0.8523,
+      "learning_rate": 1.1826236486059184e-07,
+      "loss": 0.852,
       "step": 33565
     },
     {
-      "epoch": 0.9218642717859987,
+      "epoch": 0.9524971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 3.185099858772334e-07,
-      "loss": 0.8581,
+      "learning_rate": 1.1812147679131414e-07,
+      "loss": 0.7697,
       "step": 33566
     },
     {
-      "epoch": 0.921891736013842,
+      "epoch": 0.9525255391600455,
       "grad_norm": 0.0,
-      "learning_rate": 3.182873118378482e-07,
-      "loss": 0.8547,
+      "learning_rate": 1.1798067219470854e-07,
+      "loss": 0.7763,
       "step": 33567
     },
     {
-      "epoch": 0.9219192002416852,
+      "epoch": 0.9525539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 3.1806471440392394e-07,
-      "loss": 0.7615,
+      "learning_rate": 1.1783995107196523e-07,
+      "loss": 0.6888,
       "step": 33568
     },
     {
-      "epoch": 0.9219466644695284,
+      "epoch": 0.9525822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 3.178421935772236e-07,
-      "loss": 0.7529,
+      "learning_rate": 1.1769931342427099e-07,
+      "loss": 0.848,
       "step": 33569
     },
     {
-      "epoch": 0.9219741286973717,
+      "epoch": 0.95261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 3.1761974935950566e-07,
-      "loss": 0.817,
+      "learning_rate": 1.175587592528149e-07,
+      "loss": 0.7391,
       "step": 33570
     },
     {
-      "epoch": 0.9220015929252149,
+      "epoch": 0.952639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 3.173973817525311e-07,
-      "loss": 0.8484,
+      "learning_rate": 1.174182885587849e-07,
+      "loss": 0.7888,
       "step": 33571
     },
     {
-      "epoch": 0.9220290571530582,
+      "epoch": 0.9526674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 3.171750907580595e-07,
-      "loss": 0.7521,
+      "learning_rate": 1.172779013433667e-07,
+      "loss": 0.8014,
       "step": 33572
     },
     {
-      "epoch": 0.9220565213809013,
+      "epoch": 0.9526958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 3.169528763778484e-07,
-      "loss": 0.8981,
+      "learning_rate": 1.1713759760774601e-07,
+      "loss": 0.8622,
       "step": 33573
     },
     {
-      "epoch": 0.9220839856087446,
+      "epoch": 0.9527241770715097,
       "grad_norm": 0.0,
-      "learning_rate": 3.1673073861365645e-07,
-      "loss": 0.8089,
+      "learning_rate": 1.1699737735310857e-07,
+      "loss": 0.7328,
       "step": 33574
     },
     {
-      "epoch": 0.9221114498365879,
+      "epoch": 0.9527525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 3.165086774672432e-07,
-      "loss": 0.7194,
+      "learning_rate": 1.1685724058063896e-07,
+      "loss": 0.8554,
       "step": 33575
     },
     {
-      "epoch": 0.922138914064431,
+      "epoch": 0.9527809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 3.1628669294036407e-07,
-      "loss": 0.7247,
+      "learning_rate": 1.1671718729151849e-07,
+      "loss": 0.7712,
       "step": 33576
     },
     {
-      "epoch": 0.9221663782922743,
+      "epoch": 0.9528093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 3.1606478503477534e-07,
-      "loss": 0.7991,
+      "learning_rate": 1.1657721748693395e-07,
+      "loss": 0.7494,
       "step": 33577
     },
     {
-      "epoch": 0.9221938425201176,
+      "epoch": 0.9528376844494892,
       "grad_norm": 0.0,
-      "learning_rate": 3.158429537522334e-07,
-      "loss": 0.788,
+      "learning_rate": 1.1643733116806555e-07,
+      "loss": 0.7488,
       "step": 33578
     },
     {
-      "epoch": 0.9222213067479608,
+      "epoch": 0.9528660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 3.156211990944924e-07,
-      "loss": 0.8527,
+      "learning_rate": 1.1629752833609564e-07,
+      "loss": 0.816,
       "step": 33579
     },
     {
-      "epoch": 0.922248770975804,
+      "epoch": 0.952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.153995210633098e-07,
-      "loss": 0.8845,
+      "learning_rate": 1.1615780899220552e-07,
+      "loss": 0.7305,
       "step": 33580
     },
     {
-      "epoch": 0.9222762352036472,
+      "epoch": 0.9529228149829739,
       "grad_norm": 0.0,
-      "learning_rate": 3.1517791966043764e-07,
-      "loss": 0.8341,
+      "learning_rate": 1.1601817313757313e-07,
+      "loss": 0.8706,
       "step": 33581
     },
     {
-      "epoch": 0.9223036994314905,
+      "epoch": 0.9529511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 3.1495639488762775e-07,
-      "loss": 0.8038,
+      "learning_rate": 1.1587862077337975e-07,
+      "loss": 0.8392,
       "step": 33582
     },
     {
-      "epoch": 0.9223311636593338,
+      "epoch": 0.9529795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 3.147349467466343e-07,
-      "loss": 0.7993,
+      "learning_rate": 1.1573915190080553e-07,
+      "loss": 0.7861,
       "step": 33583
     },
     {
-      "epoch": 0.9223586278871769,
+      "epoch": 0.9530079455164586,
       "grad_norm": 0.0,
-      "learning_rate": 3.1451357523921034e-07,
-      "loss": 0.6799,
+      "learning_rate": 1.1559976652102622e-07,
+      "loss": 0.7593,
       "step": 33584
     },
     {
-      "epoch": 0.9223860921150202,
+      "epoch": 0.9530363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 3.142922803671067e-07,
-      "loss": 0.7965,
+      "learning_rate": 1.1546046463521976e-07,
+      "loss": 0.7238,
       "step": 33585
     },
     {
-      "epoch": 0.9224135563428634,
+      "epoch": 0.9530646992054483,
       "grad_norm": 0.0,
-      "learning_rate": 3.140710621320731e-07,
-      "loss": 0.7665,
+      "learning_rate": 1.1532124624456297e-07,
+      "loss": 0.804,
       "step": 33586
     },
     {
-      "epoch": 0.9224410205707066,
+      "epoch": 0.9530930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 3.138499205358636e-07,
-      "loss": 0.7934,
+      "learning_rate": 1.151821113502316e-07,
+      "loss": 0.7204,
       "step": 33587
     },
     {
-      "epoch": 0.9224684847985499,
+      "epoch": 0.9531214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 3.1362885558022473e-07,
-      "loss": 0.8307,
+      "learning_rate": 1.1504305995340248e-07,
+      "loss": 0.8392,
       "step": 33588
     },
     {
-      "epoch": 0.9224959490263931,
+      "epoch": 0.953149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 3.134078672669072e-07,
-      "loss": 0.7615,
+      "learning_rate": 1.14904092055248e-07,
+      "loss": 0.8229,
       "step": 33589
     },
     {
-      "epoch": 0.9225234132542364,
+      "epoch": 0.9531782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 3.1318695559765743e-07,
-      "loss": 0.8466,
+      "learning_rate": 1.1476520765694388e-07,
+      "loss": 0.8624,
       "step": 33590
     },
     {
-      "epoch": 0.9225508774820796,
+      "epoch": 0.9532065834279229,
       "grad_norm": 0.0,
-      "learning_rate": 3.1296612057422516e-07,
-      "loss": 0.727,
+      "learning_rate": 1.1462640675966141e-07,
+      "loss": 0.793,
       "step": 33591
     },
     {
-      "epoch": 0.9225783417099228,
+      "epoch": 0.9532349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 3.127453621983567e-07,
-      "loss": 0.8003,
+      "learning_rate": 1.1448768936457411e-07,
+      "loss": 0.7519,
       "step": 33592
     },
     {
-      "epoch": 0.9226058059377661,
+      "epoch": 0.9532633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 3.1252468047180183e-07,
-      "loss": 0.8772,
+      "learning_rate": 1.1434905547285436e-07,
+      "loss": 0.6982,
       "step": 33593
     },
     {
-      "epoch": 0.9226332701656093,
+      "epoch": 0.9532917139614074,
       "grad_norm": 0.0,
-      "learning_rate": 3.123040753963025e-07,
-      "loss": 0.7659,
+      "learning_rate": 1.1421050508567233e-07,
+      "loss": 0.8583,
       "step": 33594
     },
     {
-      "epoch": 0.9226607343934525,
+      "epoch": 0.9533200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 3.12083546973605e-07,
-      "loss": 0.7581,
+      "learning_rate": 1.1407203820419821e-07,
+      "loss": 0.9156,
       "step": 33595
     },
     {
-      "epoch": 0.9226881986212958,
+      "epoch": 0.9533484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 3.11863095205458e-07,
-      "loss": 0.7378,
+      "learning_rate": 1.1393365482960217e-07,
+      "loss": 0.8322,
       "step": 33596
     },
     {
-      "epoch": 0.922715662849139,
+      "epoch": 0.9533768444948921,
       "grad_norm": 0.0,
-      "learning_rate": 3.116427200936012e-07,
-      "loss": 0.7197,
+      "learning_rate": 1.1379535496305327e-07,
+      "loss": 0.7929,
       "step": 33597
     },
     {
-      "epoch": 0.9227431270769823,
+      "epoch": 0.9534052213393871,
       "grad_norm": 0.0,
-      "learning_rate": 3.114224216397799e-07,
-      "loss": 0.7801,
+      "learning_rate": 1.1365713860571948e-07,
+      "loss": 0.6997,
       "step": 33598
     },
     {
-      "epoch": 0.9227705913048254,
+      "epoch": 0.953433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 3.1120219984573817e-07,
-      "loss": 0.8415,
+      "learning_rate": 1.1351900575876762e-07,
+      "loss": 0.7605,
       "step": 33599
     },
     {
-      "epoch": 0.9227980555326687,
+      "epoch": 0.9534619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 3.1098205471321697e-07,
-      "loss": 0.782,
+      "learning_rate": 1.1338095642336566e-07,
+      "loss": 0.7749,
       "step": 33600
     },
     {
-      "epoch": 0.922825519760512,
+      "epoch": 0.9534903518728718,
       "grad_norm": 0.0,
-      "learning_rate": 3.1076198624395927e-07,
-      "loss": 0.6884,
+      "learning_rate": 1.1324299060067934e-07,
+      "loss": 0.7995,
       "step": 33601
     },
     {
-      "epoch": 0.9228529839883551,
+      "epoch": 0.9535187287173666,
       "grad_norm": 0.0,
-      "learning_rate": 3.105419944397059e-07,
-      "loss": 0.7356,
+      "learning_rate": 1.1310510829187326e-07,
+      "loss": 0.8915,
       "step": 33602
     },
     {
-      "epoch": 0.9228804482161984,
+      "epoch": 0.9535471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 3.103220793021966e-07,
-      "loss": 0.6867,
+      "learning_rate": 1.129673094981143e-07,
+      "loss": 0.7258,
       "step": 33603
     },
     {
-      "epoch": 0.9229079124440417,
+      "epoch": 0.9535754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 3.101022408331744e-07,
-      "loss": 0.8015,
+      "learning_rate": 1.1282959422056372e-07,
+      "loss": 0.8008,
       "step": 33604
     },
     {
-      "epoch": 0.9229353766718849,
+      "epoch": 0.9536038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 3.0988247903437684e-07,
-      "loss": 0.7291,
+      "learning_rate": 1.1269196246038505e-07,
+      "loss": 0.7423,
       "step": 33605
     },
     {
-      "epoch": 0.9229628408997281,
+      "epoch": 0.9536322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 3.096627939075414e-07,
-      "loss": 0.8025,
+      "learning_rate": 1.1255441421874402e-07,
+      "loss": 0.799,
       "step": 33606
     },
     {
-      "epoch": 0.9229903051275713,
+      "epoch": 0.9536606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 3.094431854544078e-07,
-      "loss": 0.8258,
+      "learning_rate": 1.1241694949679748e-07,
+      "loss": 0.7078,
       "step": 33607
     },
     {
-      "epoch": 0.9230177693554146,
+      "epoch": 0.953688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 3.092236536767135e-07,
-      "loss": 0.7317,
+      "learning_rate": 1.122795682957123e-07,
+      "loss": 0.8773,
       "step": 33608
     },
     {
-      "epoch": 0.9230452335832579,
+      "epoch": 0.9537173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 3.09004198576196e-07,
-      "loss": 0.7833,
+      "learning_rate": 1.1214227061664418e-07,
+      "loss": 0.8105,
       "step": 33609
     },
     {
-      "epoch": 0.923072697811101,
+      "epoch": 0.9537457434733257,
       "grad_norm": 0.0,
-      "learning_rate": 3.0878482015459066e-07,
-      "loss": 0.748,
+      "learning_rate": 1.1200505646075554e-07,
+      "loss": 0.8781,
       "step": 33610
     },
     {
-      "epoch": 0.9231001620389443,
+      "epoch": 0.9537741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 3.085655184136338e-07,
-      "loss": 0.759,
+      "learning_rate": 1.1186792582920547e-07,
+      "loss": 0.8258,
       "step": 33611
     },
     {
-      "epoch": 0.9231276262667875,
+      "epoch": 0.9538024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 3.0834629335506073e-07,
-      "loss": 0.8949,
+      "learning_rate": 1.1173087872315081e-07,
+      "loss": 0.9068,
       "step": 33612
     },
     {
-      "epoch": 0.9231550904946307,
+      "epoch": 0.9538308740068104,
       "grad_norm": 0.0,
-      "learning_rate": 3.0812714498060783e-07,
-      "loss": 0.8125,
+      "learning_rate": 1.1159391514374951e-07,
+      "loss": 0.8085,
       "step": 33613
     },
     {
-      "epoch": 0.923182554722474,
+      "epoch": 0.9538592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 3.0790807329200477e-07,
-      "loss": 0.8259,
+      "learning_rate": 1.1145703509215955e-07,
+      "loss": 0.8474,
       "step": 33614
     },
     {
-      "epoch": 0.9232100189503172,
+      "epoch": 0.9538876276958003,
       "grad_norm": 0.0,
-      "learning_rate": 3.07689078290988e-07,
-      "loss": 0.8221,
+      "learning_rate": 1.1132023856953667e-07,
+      "loss": 0.7033,
       "step": 33615
     },
     {
-      "epoch": 0.9232374831781605,
+      "epoch": 0.9539160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 3.0747015997928955e-07,
-      "loss": 0.9359,
+      "learning_rate": 1.1118352557703549e-07,
+      "loss": 0.7724,
       "step": 33616
     },
     {
-      "epoch": 0.9232649474060037,
+      "epoch": 0.95394438138479,
       "grad_norm": 0.0,
-      "learning_rate": 3.072513183586434e-07,
-      "loss": 0.8379,
+      "learning_rate": 1.1104689611581177e-07,
+      "loss": 0.7765,
       "step": 33617
     },
     {
-      "epoch": 0.9232924116338469,
+      "epoch": 0.953972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 3.070325534307772e-07,
-      "loss": 0.8028,
+      "learning_rate": 1.1091035018702013e-07,
+      "loss": 0.8216,
       "step": 33618
     },
     {
-      "epoch": 0.9233198758616902,
+      "epoch": 0.9540011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 3.068138651974251e-07,
-      "loss": 0.7402,
+      "learning_rate": 1.1077388779181297e-07,
+      "loss": 0.818,
       "step": 33619
     },
     {
-      "epoch": 0.9233473400895333,
+      "epoch": 0.9540295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 3.065952536603167e-07,
-      "loss": 0.6838,
+      "learning_rate": 1.1063750893134273e-07,
+      "loss": 0.6754,
       "step": 33620
     },
     {
-      "epoch": 0.9233748043173766,
+      "epoch": 0.9540578887627695,
       "grad_norm": 0.0,
-      "learning_rate": 3.063767188211808e-07,
-      "loss": 0.8499,
+      "learning_rate": 1.1050121360676292e-07,
+      "loss": 0.6885,
       "step": 33621
     },
     {
-      "epoch": 0.9234022685452199,
+      "epoch": 0.9540862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 3.0615826068174815e-07,
-      "loss": 0.7774,
+      "learning_rate": 1.1036500181922372e-07,
+      "loss": 0.7679,
       "step": 33622
     },
     {
-      "epoch": 0.9234297327730631,
+      "epoch": 0.9541146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 3.0593987924374626e-07,
-      "loss": 0.7933,
+      "learning_rate": 1.1022887356987533e-07,
+      "loss": 0.7974,
       "step": 33623
     },
     {
-      "epoch": 0.9234571970009063,
+      "epoch": 0.9541430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 3.057215745089026e-07,
-      "loss": 0.8227,
+      "learning_rate": 1.1009282885986793e-07,
+      "loss": 0.7397,
       "step": 33624
     },
     {
-      "epoch": 0.9234846612287495,
+      "epoch": 0.9541713961407492,
       "grad_norm": 0.0,
-      "learning_rate": 3.0550334647894477e-07,
-      "loss": 0.8133,
+      "learning_rate": 1.0995686769035286e-07,
+      "loss": 0.7272,
       "step": 33625
     },
     {
-      "epoch": 0.9235121254565928,
+      "epoch": 0.954199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 3.0528519515560015e-07,
-      "loss": 0.7162,
+      "learning_rate": 1.098209900624747e-07,
+      "loss": 0.7749,
       "step": 33626
     },
     {
-      "epoch": 0.9235395896844361,
+      "epoch": 0.9542281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 3.0506712054059416e-07,
-      "loss": 0.7726,
+      "learning_rate": 1.0968519597738369e-07,
+      "loss": 0.7898,
       "step": 33627
     },
     {
-      "epoch": 0.9235670539122792,
+      "epoch": 0.9542565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 3.0484912263565315e-07,
-      "loss": 0.9416,
+      "learning_rate": 1.0954948543622779e-07,
+      "loss": 0.8548,
       "step": 33628
     },
     {
-      "epoch": 0.9235945181401225,
+      "epoch": 0.9542849035187287,
       "grad_norm": 0.0,
-      "learning_rate": 3.0463120144250236e-07,
-      "loss": 0.7405,
+      "learning_rate": 1.0941385844015051e-07,
+      "loss": 0.8064,
       "step": 33629
     },
     {
-      "epoch": 0.9236219823679658,
+      "epoch": 0.9543132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 3.0441335696286376e-07,
-      "loss": 0.847,
+      "learning_rate": 1.0927831499029873e-07,
+      "loss": 0.7968,
       "step": 33630
     },
     {
-      "epoch": 0.9236494465958089,
+      "epoch": 0.9543416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 3.041955891984627e-07,
-      "loss": 0.8449,
+      "learning_rate": 1.091428550878193e-07,
+      "loss": 0.8369,
       "step": 33631
     },
     {
-      "epoch": 0.9236769108236522,
+      "epoch": 0.9543700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 3.039778981510211e-07,
-      "loss": 0.7296,
+      "learning_rate": 1.0900747873385353e-07,
+      "loss": 0.8234,
       "step": 33632
     },
     {
-      "epoch": 0.9237043750514954,
+      "epoch": 0.9543984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 3.037602838222631e-07,
-      "loss": 0.7021,
+      "learning_rate": 1.0887218592954607e-07,
+      "loss": 0.8177,
       "step": 33633
     },
     {
-      "epoch": 0.9237318392793387,
+      "epoch": 0.9544267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 3.0354274621390847e-07,
-      "loss": 0.838,
+      "learning_rate": 1.0873697667604155e-07,
+      "loss": 0.7665,
       "step": 33634
     },
     {
-      "epoch": 0.9237593035071819,
+      "epoch": 0.9544551645856981,
       "grad_norm": 0.0,
-      "learning_rate": 3.033252853276825e-07,
-      "loss": 0.8044,
+      "learning_rate": 1.0860185097447907e-07,
+      "loss": 0.7253,
       "step": 33635
     },
     {
-      "epoch": 0.9237867677350251,
+      "epoch": 0.954483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 3.031079011653004e-07,
-      "loss": 0.7752,
+      "learning_rate": 1.0846680882600102e-07,
+      "loss": 0.8799,
       "step": 33636
     },
     {
-      "epoch": 0.9238142319628684,
+      "epoch": 0.9545119182746878,
       "grad_norm": 0.0,
-      "learning_rate": 3.028905937284865e-07,
-      "loss": 0.746,
+      "learning_rate": 1.0833185023174875e-07,
+      "loss": 0.7831,
       "step": 33637
     },
     {
-      "epoch": 0.9238416961907115,
+      "epoch": 0.9545402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 3.026733630189571e-07,
-      "loss": 0.7622,
+      "learning_rate": 1.0819697519286243e-07,
+      "loss": 0.7791,
       "step": 33638
     },
     {
-      "epoch": 0.9238691604185548,
+      "epoch": 0.9545686719636777,
       "grad_norm": 0.0,
-      "learning_rate": 3.024562090384331e-07,
-      "loss": 0.8484,
+      "learning_rate": 1.0806218371048116e-07,
+      "loss": 0.765,
       "step": 33639
     },
     {
-      "epoch": 0.9238966246463981,
+      "epoch": 0.9545970488081725,
       "grad_norm": 0.0,
-      "learning_rate": 3.0223913178863305e-07,
-      "loss": 0.8015,
+      "learning_rate": 1.0792747578574291e-07,
+      "loss": 0.8464,
       "step": 33640
     },
     {
-      "epoch": 0.9239240888742413,
+      "epoch": 0.9546254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 3.020221312712723e-07,
-      "loss": 0.8414,
+      "learning_rate": 1.0779285141978568e-07,
+      "loss": 0.8666,
       "step": 33641
     },
     {
-      "epoch": 0.9239515531020845,
+      "epoch": 0.9546538024971624,
       "grad_norm": 0.0,
-      "learning_rate": 3.018052074880684e-07,
-      "loss": 0.7693,
+      "learning_rate": 1.0765831061374743e-07,
+      "loss": 0.7937,
       "step": 33642
     },
     {
-      "epoch": 0.9239790173299278,
+      "epoch": 0.9546821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 3.015883604407399e-07,
-      "loss": 0.8547,
+      "learning_rate": 1.0752385336876392e-07,
+      "loss": 0.8244,
       "step": 33643
     },
     {
-      "epoch": 0.924006481557771,
+      "epoch": 0.9547105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 3.0137159013100106e-07,
-      "loss": 0.8648,
+      "learning_rate": 1.0738947968597092e-07,
+      "loss": 0.6858,
       "step": 33644
     },
     {
-      "epoch": 0.9240339457856143,
+      "epoch": 0.9547389330306469,
       "grad_norm": 0.0,
-      "learning_rate": 3.0115489656056597e-07,
-      "loss": 0.7727,
+      "learning_rate": 1.0725518956650527e-07,
+      "loss": 0.7774,
       "step": 33645
     },
     {
-      "epoch": 0.9240614100134574,
+      "epoch": 0.9547673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 3.009382797311522e-07,
-      "loss": 0.8391,
+      "learning_rate": 1.071209830114972e-07,
+      "loss": 0.8065,
       "step": 33646
     },
     {
-      "epoch": 0.9240888742413007,
+      "epoch": 0.9547956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 3.007217396444695e-07,
-      "loss": 0.8329,
+      "learning_rate": 1.0698686002208469e-07,
+      "loss": 0.7672,
       "step": 33647
     },
     {
-      "epoch": 0.924116338469144,
+      "epoch": 0.9548240635641316,
       "grad_norm": 0.0,
-      "learning_rate": 3.0050527630223427e-07,
-      "loss": 0.841,
+      "learning_rate": 1.0685282059939795e-07,
+      "loss": 0.8134,
       "step": 33648
     },
     {
-      "epoch": 0.9241438026969871,
+      "epoch": 0.9548524404086266,
       "grad_norm": 0.0,
-      "learning_rate": 3.0028888970615844e-07,
-      "loss": 0.7225,
+      "learning_rate": 1.067188647445705e-07,
+      "loss": 0.7494,
       "step": 33649
     },
     {
-      "epoch": 0.9241712669248304,
+      "epoch": 0.9548808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 3.00072579857954e-07,
-      "loss": 0.7132,
+      "learning_rate": 1.0658499245873366e-07,
+      "loss": 0.7559,
       "step": 33650
     },
     {
-      "epoch": 0.9241987311526736,
+      "epoch": 0.9549091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 2.99856346759333e-07,
-      "loss": 0.9156,
+      "learning_rate": 1.0645120374301876e-07,
+      "loss": 0.7657,
       "step": 33651
     },
     {
-      "epoch": 0.9242261953805169,
+      "epoch": 0.9549375709421113,
       "grad_norm": 0.0,
-      "learning_rate": 2.996401904120072e-07,
-      "loss": 0.8469,
+      "learning_rate": 1.0631749859855379e-07,
+      "loss": 0.8922,
       "step": 33652
     },
     {
-      "epoch": 0.9242536596083601,
+      "epoch": 0.9549659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 2.9942411081768653e-07,
-      "loss": 0.8209,
+      "learning_rate": 1.0618387702647004e-07,
+      "loss": 0.7406,
       "step": 33653
     },
     {
-      "epoch": 0.9242811238362033,
+      "epoch": 0.954994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 2.992081079780773e-07,
-      "loss": 0.9739,
+      "learning_rate": 1.0605033902789663e-07,
+      "loss": 0.8041,
       "step": 33654
     },
     {
-      "epoch": 0.9243085880640466,
+      "epoch": 0.9550227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 2.9899218189489267e-07,
-      "loss": 0.7692,
+      "learning_rate": 1.0591688460396043e-07,
+      "loss": 0.7959,
       "step": 33655
     },
     {
-      "epoch": 0.9243360522918899,
+      "epoch": 0.9550510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 2.9877633256983894e-07,
-      "loss": 0.7703,
+      "learning_rate": 1.0578351375578943e-07,
+      "loss": 0.9026,
       "step": 33656
     },
     {
-      "epoch": 0.924363516519733,
+      "epoch": 0.9550794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 2.9856056000462487e-07,
-      "loss": 0.8428,
+      "learning_rate": 1.056502264845083e-07,
+      "loss": 0.8536,
       "step": 33657
     },
     {
-      "epoch": 0.9243909807475763,
+      "epoch": 0.9551078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 2.9834486420095785e-07,
-      "loss": 0.7642,
+      "learning_rate": 1.0551702279124609e-07,
+      "loss": 0.8538,
       "step": 33658
     },
     {
-      "epoch": 0.9244184449754195,
+      "epoch": 0.9551362088535755,
       "grad_norm": 0.0,
-      "learning_rate": 2.981292451605444e-07,
-      "loss": 0.8262,
+      "learning_rate": 1.053839026771264e-07,
+      "loss": 0.759,
       "step": 33659
     },
     {
-      "epoch": 0.9244459092032628,
+      "epoch": 0.9551645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 2.979137028850898e-07,
-      "loss": 0.7324,
+      "learning_rate": 1.0525086614327274e-07,
+      "loss": 0.8501,
       "step": 33660
     },
     {
-      "epoch": 0.924473373431106,
+      "epoch": 0.9551929625425652,
       "grad_norm": 0.0,
-      "learning_rate": 2.9769823737630157e-07,
-      "loss": 0.7393,
+      "learning_rate": 1.0511791319080978e-07,
+      "loss": 0.8064,
       "step": 33661
     },
     {
-      "epoch": 0.9245008376589492,
+      "epoch": 0.9552213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 2.974828486358827e-07,
-      "loss": 0.7615,
+      "learning_rate": 1.0498504382086216e-07,
+      "loss": 0.7415,
       "step": 33662
     },
     {
-      "epoch": 0.9245283018867925,
+      "epoch": 0.9552497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 2.9726753666553753e-07,
-      "loss": 0.8363,
+      "learning_rate": 1.0485225803454902e-07,
+      "loss": 0.8533,
       "step": 33663
     },
     {
-      "epoch": 0.9245557661146356,
+      "epoch": 0.9552780930760499,
       "grad_norm": 0.0,
-      "learning_rate": 2.970523014669713e-07,
-      "loss": 0.8299,
+      "learning_rate": 1.0471955583299387e-07,
+      "loss": 0.725,
       "step": 33664
     },
     {
-      "epoch": 0.9245832303424789,
+      "epoch": 0.9553064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 2.9683714304188373e-07,
-      "loss": 0.8109,
+      "learning_rate": 1.0458693721731916e-07,
+      "loss": 0.7513,
       "step": 33665
     },
     {
-      "epoch": 0.9246106945703222,
+      "epoch": 0.9553348467650398,
       "grad_norm": 0.0,
-      "learning_rate": 2.9662206139198014e-07,
-      "loss": 0.8731,
+      "learning_rate": 1.0445440218864177e-07,
+      "loss": 0.8997,
       "step": 33666
     },
     {
-      "epoch": 0.9246381587981654,
+      "epoch": 0.9553632236095346,
       "grad_norm": 0.0,
-      "learning_rate": 2.964070565189614e-07,
-      "loss": 0.8238,
+      "learning_rate": 1.0432195074808305e-07,
+      "loss": 0.7941,
       "step": 33667
     },
     {
-      "epoch": 0.9246656230260086,
+      "epoch": 0.9553916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 2.961921284245306e-07,
-      "loss": 0.7589,
+      "learning_rate": 1.0418958289676096e-07,
+      "loss": 0.7479,
       "step": 33668
     },
     {
-      "epoch": 0.9246930872538519,
+      "epoch": 0.9554199772985245,
       "grad_norm": 0.0,
-      "learning_rate": 2.9597727711038417e-07,
-      "loss": 0.8159,
+      "learning_rate": 1.0405729863579573e-07,
+      "loss": 0.7872,
       "step": 33669
     },
     {
-      "epoch": 0.9247205514816951,
+      "epoch": 0.9554483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 2.957625025782274e-07,
-      "loss": 0.823,
+      "learning_rate": 1.0392509796630312e-07,
+      "loss": 0.7983,
       "step": 33670
     },
     {
-      "epoch": 0.9247480157095384,
+      "epoch": 0.9554767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 2.955478048297544e-07,
-      "loss": 0.8342,
+      "learning_rate": 1.0379298088939893e-07,
+      "loss": 0.6717,
       "step": 33671
     },
     {
-      "epoch": 0.9247754799373815,
+      "epoch": 0.955505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 2.953331838666662e-07,
-      "loss": 0.8197,
+      "learning_rate": 1.0366094740620115e-07,
+      "loss": 0.8265,
       "step": 33672
     },
     {
-      "epoch": 0.9248029441652248,
+      "epoch": 0.955533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.9511863969066135e-07,
-      "loss": 0.8782,
+      "learning_rate": 1.0352899751782441e-07,
+      "loss": 0.682,
       "step": 33673
     },
     {
-      "epoch": 0.9248304083930681,
+      "epoch": 0.9555618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 2.949041723034374e-07,
-      "loss": 0.9161,
+      "learning_rate": 1.0339713122538342e-07,
+      "loss": 0.7966,
       "step": 33674
     },
     {
-      "epoch": 0.9248578726209112,
+      "epoch": 0.9555902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 2.946897817066907e-07,
-      "loss": 0.7704,
+      "learning_rate": 1.0326534852999059e-07,
+      "loss": 0.8078,
       "step": 33675
     },
     {
-      "epoch": 0.9248853368487545,
+      "epoch": 0.9556186152099887,
       "grad_norm": 0.0,
-      "learning_rate": 2.944754679021189e-07,
-      "loss": 0.6708,
+      "learning_rate": 1.0313364943276172e-07,
+      "loss": 0.8755,
       "step": 33676
     },
     {
-      "epoch": 0.9249128010765977,
+      "epoch": 0.9556469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 2.942612308914172e-07,
-      "loss": 0.752,
+      "learning_rate": 1.0300203393480701e-07,
+      "loss": 0.85,
       "step": 33677
     },
     {
-      "epoch": 0.924940265304441,
+      "epoch": 0.9556753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 2.940470706762788e-07,
-      "loss": 0.8386,
+      "learning_rate": 1.0287050203723892e-07,
+      "loss": 0.8218,
       "step": 33678
     },
     {
-      "epoch": 0.9249677295322842,
+      "epoch": 0.9557037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 2.938329872584e-07,
-      "loss": 0.7773,
+      "learning_rate": 1.0273905374116877e-07,
+      "loss": 0.8141,
       "step": 33679
     },
     {
-      "epoch": 0.9249951937601274,
+      "epoch": 0.9557321225879682,
       "grad_norm": 0.0,
-      "learning_rate": 2.9361898063947405e-07,
-      "loss": 0.7771,
+      "learning_rate": 1.026076890477068e-07,
+      "loss": 0.8211,
       "step": 33680
     },
     {
-      "epoch": 0.9250226579879707,
+      "epoch": 0.9557604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 2.9340505082119497e-07,
-      "loss": 0.7915,
+      "learning_rate": 1.024764079579621e-07,
+      "loss": 0.7161,
       "step": 33681
     },
     {
-      "epoch": 0.925050122215814,
+      "epoch": 0.955788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.9319119780525595e-07,
-      "loss": 0.7512,
+      "learning_rate": 1.0234521047304491e-07,
+      "loss": 0.7842,
       "step": 33682
     },
     {
-      "epoch": 0.9250775864436571,
+      "epoch": 0.9558172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 2.9297742159334783e-07,
-      "loss": 0.929,
+      "learning_rate": 1.0221409659406101e-07,
+      "loss": 0.8197,
       "step": 33683
     },
     {
-      "epoch": 0.9251050506715004,
+      "epoch": 0.9558456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 2.9276372218716155e-07,
-      "loss": 0.7368,
+      "learning_rate": 1.0208306632212061e-07,
+      "loss": 0.7664,
       "step": 33684
     },
     {
-      "epoch": 0.9251325148993436,
+      "epoch": 0.9558740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 2.925500995883912e-07,
-      "loss": 0.8688,
+      "learning_rate": 1.0195211965832951e-07,
+      "loss": 0.8445,
       "step": 33685
     },
     {
-      "epoch": 0.9251599791271868,
+      "epoch": 0.9559023836549376,
       "grad_norm": 0.0,
-      "learning_rate": 2.923365537987222e-07,
-      "loss": 0.7913,
+      "learning_rate": 1.0182125660379461e-07,
+      "loss": 0.8136,
       "step": 33686
     },
     {
-      "epoch": 0.9251874433550301,
+      "epoch": 0.9559307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 2.9212308481984755e-07,
-      "loss": 0.737,
+      "learning_rate": 1.0169047715961944e-07,
+      "loss": 0.8655,
       "step": 33687
     },
     {
-      "epoch": 0.9252149075828733,
+      "epoch": 0.9559591373439273,
       "grad_norm": 0.0,
-      "learning_rate": 2.9190969265345703e-07,
-      "loss": 0.8586,
+      "learning_rate": 1.0155978132691091e-07,
+      "loss": 0.823,
       "step": 33688
     },
     {
-      "epoch": 0.9252423718107166,
+      "epoch": 0.9559875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 2.9169637730123603e-07,
-      "loss": 0.8118,
+      "learning_rate": 1.0142916910677036e-07,
+      "loss": 0.8488,
       "step": 33689
     },
     {
-      "epoch": 0.9252698360385597,
+      "epoch": 0.9560158910329172,
       "grad_norm": 0.0,
-      "learning_rate": 2.914831387648731e-07,
-      "loss": 0.8701,
+      "learning_rate": 1.0129864050030357e-07,
+      "loss": 0.7926,
       "step": 33690
     },
     {
-      "epoch": 0.925297300266403,
+      "epoch": 0.956044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.9126997704605695e-07,
-      "loss": 0.7644,
+      "learning_rate": 1.011681955086119e-07,
+      "loss": 0.7953,
       "step": 33691
     },
     {
-      "epoch": 0.9253247644942463,
+      "epoch": 0.9560726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 2.9105689214647515e-07,
-      "loss": 0.8285,
+      "learning_rate": 1.0103783413279777e-07,
+      "loss": 0.7788,
       "step": 33692
     },
     {
-      "epoch": 0.9253522287220894,
+      "epoch": 0.9561010215664019,
       "grad_norm": 0.0,
-      "learning_rate": 2.9084388406781074e-07,
-      "loss": 0.8104,
+      "learning_rate": 1.0090755637396254e-07,
+      "loss": 0.8858,
       "step": 33693
     },
     {
-      "epoch": 0.9253796929499327,
+      "epoch": 0.9561293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 2.9063095281175123e-07,
-      "loss": 0.8377,
+      "learning_rate": 1.0077736223320422e-07,
+      "loss": 0.7874,
       "step": 33694
     },
     {
-      "epoch": 0.925407157177776,
+      "epoch": 0.9561577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.904180983799798e-07,
-      "loss": 0.8428,
+      "learning_rate": 1.0064725171162636e-07,
+      "loss": 0.7003,
       "step": 33695
     },
     {
-      "epoch": 0.9254346214056192,
+      "epoch": 0.9561861520998864,
       "grad_norm": 0.0,
-      "learning_rate": 2.9020532077418173e-07,
-      "loss": 0.7055,
+      "learning_rate": 1.0051722481032589e-07,
+      "loss": 0.8563,
       "step": 33696
     },
     {
-      "epoch": 0.9254620856334624,
+      "epoch": 0.9562145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 2.899926199960401e-07,
-      "loss": 0.8154,
+      "learning_rate": 1.0038728153040078e-07,
+      "loss": 0.7898,
       "step": 33697
     },
     {
-      "epoch": 0.9254895498613056,
+      "epoch": 0.9562429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 2.8977999604723805e-07,
-      "loss": 0.7029,
+      "learning_rate": 1.0025742187294907e-07,
+      "loss": 0.7421,
       "step": 33698
     },
     {
-      "epoch": 0.9255170140891489,
+      "epoch": 0.9562712826333711,
       "grad_norm": 0.0,
-      "learning_rate": 2.8956744892945976e-07,
-      "loss": 0.8946,
+      "learning_rate": 1.0012764583906875e-07,
+      "loss": 0.8708,
       "step": 33699
     },
     {
-      "epoch": 0.9255444783169922,
+      "epoch": 0.9562996594778661,
       "grad_norm": 0.0,
-      "learning_rate": 2.893549786443839e-07,
-      "loss": 0.8271,
+      "learning_rate": 9.999795342985452e-08,
+      "loss": 0.8411,
       "step": 33700
     },
     {
-      "epoch": 0.9255719425448353,
+      "epoch": 0.956328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 2.891425851936935e-07,
-      "loss": 0.7698,
+      "learning_rate": 9.986834464640328e-08,
+      "loss": 0.8276,
       "step": 33701
     },
     {
-      "epoch": 0.9255994067726786,
+      "epoch": 0.9563564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 2.8893026857906846e-07,
-      "loss": 0.8104,
+      "learning_rate": 9.973881948980856e-08,
+      "loss": 0.843,
       "step": 33702
     },
     {
-      "epoch": 0.9256268710005218,
+      "epoch": 0.9563847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 2.887180288021885e-07,
-      "loss": 0.8953,
+      "learning_rate": 9.960937796116621e-08,
+      "loss": 0.784,
       "step": 33703
     },
     {
-      "epoch": 0.925654335228365,
+      "epoch": 0.9564131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 2.885058658647322e-07,
-      "loss": 0.7969,
+      "learning_rate": 9.948002006156754e-08,
+      "loss": 0.7569,
       "step": 33704
     },
     {
-      "epoch": 0.9256817994562083,
+      "epoch": 0.9564415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 2.882937797683816e-07,
-      "loss": 0.7961,
+      "learning_rate": 9.935074579210613e-08,
+      "loss": 0.7578,
       "step": 33705
     },
     {
-      "epoch": 0.9257092636840515,
+      "epoch": 0.9564699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 2.88081770514812e-07,
-      "loss": 0.7,
+      "learning_rate": 9.922155515387444e-08,
+      "loss": 0.7729,
       "step": 33706
     },
     {
-      "epoch": 0.9257367279118948,
+      "epoch": 0.9564982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 2.878698381056999e-07,
-      "loss": 0.7856,
+      "learning_rate": 9.90924481479627e-08,
+      "loss": 0.8276,
       "step": 33707
     },
     {
-      "epoch": 0.925764192139738,
+      "epoch": 0.9565266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 2.8765798254272504e-07,
-      "loss": 0.8786,
+      "learning_rate": 9.896342477546339e-08,
+      "loss": 0.8434,
       "step": 33708
     },
     {
-      "epoch": 0.9257916563675812,
+      "epoch": 0.9565550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 2.874462038275627e-07,
-      "loss": 0.8723,
+      "learning_rate": 9.883448503746451e-08,
+      "loss": 0.789,
       "step": 33709
     },
     {
-      "epoch": 0.9258191205954245,
+      "epoch": 0.956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 2.8723450196188715e-07,
-      "loss": 0.8363,
+      "learning_rate": 9.870562893505519e-08,
+      "loss": 0.7062,
       "step": 33710
     },
     {
-      "epoch": 0.9258465848232676,
+      "epoch": 0.9566118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 2.8702287694737706e-07,
-      "loss": 0.8586,
+      "learning_rate": 9.857685646932569e-08,
+      "loss": 0.8322,
       "step": 33711
     },
     {
-      "epoch": 0.9258740490511109,
+      "epoch": 0.9566401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 2.8681132878570216e-07,
-      "loss": 0.837,
+      "learning_rate": 9.844816764136179e-08,
+      "loss": 0.9138,
       "step": 33712
     },
     {
-      "epoch": 0.9259015132789542,
+      "epoch": 0.9566685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 2.8659985747853894e-07,
-      "loss": 0.8093,
+      "learning_rate": 9.831956245225149e-08,
+      "loss": 0.896,
       "step": 33713
     },
     {
-      "epoch": 0.9259289775067974,
+      "epoch": 0.9566969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 2.8638846302755927e-07,
-      "loss": 0.8255,
+      "learning_rate": 9.819104090308062e-08,
+      "loss": 0.7607,
       "step": 33714
     },
     {
-      "epoch": 0.9259564417346406,
+      "epoch": 0.9567253121452894,
       "grad_norm": 0.0,
-      "learning_rate": 2.8617714543443754e-07,
-      "loss": 0.828,
+      "learning_rate": 9.806260299493498e-08,
+      "loss": 0.8938,
       "step": 33715
     },
     {
-      "epoch": 0.9259839059624838,
+      "epoch": 0.9567536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 2.8596590470084674e-07,
-      "loss": 0.8016,
+      "learning_rate": 9.793424872890034e-08,
+      "loss": 0.8135,
       "step": 33716
     },
     {
-      "epoch": 0.9260113701903271,
+      "epoch": 0.9567820658342793,
       "grad_norm": 0.0,
-      "learning_rate": 2.8575474082845557e-07,
-      "loss": 0.8177,
+      "learning_rate": 9.78059781060603e-08,
+      "loss": 0.8353,
       "step": 33717
     },
     {
-      "epoch": 0.9260388344181704,
+      "epoch": 0.9568104426787741,
       "grad_norm": 0.0,
-      "learning_rate": 2.8554365381893487e-07,
-      "loss": 0.8231,
+      "learning_rate": 9.767779112749731e-08,
+      "loss": 0.8519,
       "step": 33718
     },
     {
-      "epoch": 0.9260662986460135,
+      "epoch": 0.956838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 2.8533264367395673e-07,
-      "loss": 0.757,
+      "learning_rate": 9.754968779429719e-08,
+      "loss": 0.7505,
       "step": 33719
     },
     {
-      "epoch": 0.9260937628738568,
+      "epoch": 0.9568671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 2.8512171039518863e-07,
-      "loss": 0.7481,
+      "learning_rate": 9.742166810753795e-08,
+      "loss": 0.6792,
       "step": 33720
     },
     {
-      "epoch": 0.9261212271017001,
+      "epoch": 0.9568955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.849108539843015e-07,
-      "loss": 0.8434,
+      "learning_rate": 9.729373206830428e-08,
+      "loss": 0.7687,
       "step": 33721
     },
     {
-      "epoch": 0.9261486913295432,
+      "epoch": 0.9569239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 2.8470007444296287e-07,
-      "loss": 0.8279,
+      "learning_rate": 9.716587967767532e-08,
+      "loss": 0.801,
       "step": 33722
     },
     {
-      "epoch": 0.9261761555573865,
+      "epoch": 0.9569523269012485,
       "grad_norm": 0.0,
-      "learning_rate": 2.844893717728414e-07,
-      "loss": 0.8889,
+      "learning_rate": 9.703811093673022e-08,
+      "loss": 0.9086,
       "step": 33723
     },
     {
-      "epoch": 0.9262036197852297,
+      "epoch": 0.9569807037457435,
       "grad_norm": 0.0,
-      "learning_rate": 2.8427874597560244e-07,
-      "loss": 0.8741,
+      "learning_rate": 9.69104258465503e-08,
+      "loss": 0.7722,
       "step": 33724
     },
     {
-      "epoch": 0.926231084013073,
+      "epoch": 0.9570090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 2.840681970529147e-07,
-      "loss": 0.9496,
+      "learning_rate": 9.678282440821363e-08,
+      "loss": 0.7766,
       "step": 33725
     },
     {
-      "epoch": 0.9262585482409162,
+      "epoch": 0.9570374574347332,
       "grad_norm": 0.0,
-      "learning_rate": 2.8385772500644116e-07,
-      "loss": 0.7889,
+      "learning_rate": 9.66553066227971e-08,
+      "loss": 0.7003,
       "step": 33726
     },
     {
-      "epoch": 0.9262860124687594,
+      "epoch": 0.9570658342792282,
       "grad_norm": 0.0,
-      "learning_rate": 2.836473298378495e-07,
-      "loss": 0.7153,
+      "learning_rate": 9.652787249137874e-08,
+      "loss": 0.8416,
       "step": 33727
     },
     {
-      "epoch": 0.9263134766966027,
+      "epoch": 0.9570942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 2.83437011548805e-07,
-      "loss": 0.7871,
+      "learning_rate": 9.640052201503436e-08,
+      "loss": 0.83,
       "step": 33728
     },
     {
-      "epoch": 0.9263409409244459,
+      "epoch": 0.9571225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 2.8322677014097075e-07,
-      "loss": 0.8523,
+      "learning_rate": 9.627325519483865e-08,
+      "loss": 0.801,
       "step": 33729
     },
     {
-      "epoch": 0.9263684051522891,
+      "epoch": 0.9571509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 2.8301660561600886e-07,
-      "loss": 0.8017,
+      "learning_rate": 9.614607203186966e-08,
+      "loss": 0.7729,
       "step": 33730
     },
     {
-      "epoch": 0.9263958693801324,
+      "epoch": 0.9571793416572077,
       "grad_norm": 0.0,
-      "learning_rate": 2.8280651797558454e-07,
-      "loss": 0.8243,
+      "learning_rate": 9.601897252719983e-08,
+      "loss": 0.829,
       "step": 33731
     },
     {
-      "epoch": 0.9264233336079756,
+      "epoch": 0.9572077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 2.825965072213588e-07,
-      "loss": 0.8183,
+      "learning_rate": 9.589195668190277e-08,
+      "loss": 0.8205,
       "step": 33732
     },
     {
-      "epoch": 0.9264507978358189,
+      "epoch": 0.9572360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 2.823865733549935e-07,
-      "loss": 0.8134,
+      "learning_rate": 9.576502449705205e-08,
+      "loss": 0.9095,
       "step": 33733
     },
     {
-      "epoch": 0.9264782620636621,
+      "epoch": 0.9572644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 2.821767163781497e-07,
-      "loss": 0.7924,
+      "learning_rate": 9.563817597371793e-08,
+      "loss": 0.7699,
       "step": 33734
     },
     {
-      "epoch": 0.9265057262915053,
+      "epoch": 0.9572928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 2.819669362924893e-07,
-      "loss": 0.9261,
+      "learning_rate": 9.551141111297402e-08,
+      "loss": 0.7014,
       "step": 33735
     },
     {
-      "epoch": 0.9265331905193486,
+      "epoch": 0.9573212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 2.8175723309966986e-07,
-      "loss": 0.9212,
+      "learning_rate": 9.538472991589164e-08,
+      "loss": 0.8281,
       "step": 33736
     },
     {
-      "epoch": 0.9265606547471917,
+      "epoch": 0.957349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 2.8154760680135116e-07,
-      "loss": 0.8353,
+      "learning_rate": 9.525813238353887e-08,
+      "loss": 0.8419,
       "step": 33737
     },
     {
-      "epoch": 0.926588118975035,
+      "epoch": 0.957377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 2.8133805739919307e-07,
-      "loss": 0.7829,
+      "learning_rate": 9.513161851698594e-08,
+      "loss": 0.8703,
       "step": 33738
     },
     {
-      "epoch": 0.9266155832028783,
+      "epoch": 0.9574063564131668,
       "grad_norm": 0.0,
-      "learning_rate": 2.8112858489485307e-07,
-      "loss": 0.7865,
+      "learning_rate": 9.500518831730087e-08,
+      "loss": 0.8571,
       "step": 33739
     },
     {
-      "epoch": 0.9266430474307215,
+      "epoch": 0.9574347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 2.8091918928998874e-07,
-      "loss": 0.7927,
+      "learning_rate": 9.487884178555284e-08,
+      "loss": 0.7215,
       "step": 33740
     },
     {
-      "epoch": 0.9266705116585647,
+      "epoch": 0.9574631101021567,
       "grad_norm": 0.0,
-      "learning_rate": 2.807098705862565e-07,
-      "loss": 0.7744,
+      "learning_rate": 9.475257892280876e-08,
+      "loss": 0.7381,
       "step": 33741
     },
     {
-      "epoch": 0.9266979758864079,
+      "epoch": 0.9574914869466515,
       "grad_norm": 0.0,
-      "learning_rate": 2.8050062878531294e-07,
-      "loss": 0.7361,
+      "learning_rate": 9.462639973013443e-08,
+      "loss": 0.7376,
       "step": 33742
     },
     {
-      "epoch": 0.9267254401142512,
+      "epoch": 0.9575198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 2.802914638888121e-07,
-      "loss": 0.7913,
+      "learning_rate": 9.450030420859679e-08,
+      "loss": 0.7724,
       "step": 33743
     },
     {
-      "epoch": 0.9267529043420945,
+      "epoch": 0.9575482406356414,
       "grad_norm": 0.0,
-      "learning_rate": 2.8008237589841057e-07,
-      "loss": 0.8671,
+      "learning_rate": 9.437429235925944e-08,
+      "loss": 0.6968,
       "step": 33744
     },
     {
-      "epoch": 0.9267803685699376,
+      "epoch": 0.9575766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 2.7987336481576256e-07,
-      "loss": 0.7011,
+      "learning_rate": 9.424836418318817e-08,
+      "loss": 0.7792,
       "step": 33745
     },
     {
-      "epoch": 0.9268078327977809,
+      "epoch": 0.9576049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 2.796644306425211e-07,
-      "loss": 0.6893,
+      "learning_rate": 9.412251968144547e-08,
+      "loss": 0.7948,
       "step": 33746
     },
     {
-      "epoch": 0.9268352970256242,
+      "epoch": 0.957633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 2.794555733803417e-07,
-      "loss": 0.7805,
+      "learning_rate": 9.399675885509607e-08,
+      "loss": 0.7894,
       "step": 33747
     },
     {
-      "epoch": 0.9268627612534673,
+      "epoch": 0.9576617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 2.79246793030874e-07,
-      "loss": 0.7463,
+      "learning_rate": 9.387108170520131e-08,
+      "loss": 0.754,
       "step": 33748
     },
     {
-      "epoch": 0.9268902254813106,
+      "epoch": 0.9576901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 2.7903808959577226e-07,
-      "loss": 0.7633,
+      "learning_rate": 9.374548823282259e-08,
+      "loss": 0.7943,
       "step": 33749
     },
     {
-      "epoch": 0.9269176897091538,
+      "epoch": 0.9577185017026106,
       "grad_norm": 0.0,
-      "learning_rate": 2.7882946307668525e-07,
-      "loss": 0.8554,
+      "learning_rate": 9.361997843902238e-08,
+      "loss": 0.8179,
       "step": 33750
     },
     {
-      "epoch": 0.9269451539369971,
+      "epoch": 0.9577468785471056,
       "grad_norm": 0.0,
-      "learning_rate": 2.7862091347526597e-07,
-      "loss": 0.9123,
+      "learning_rate": 9.34945523248576e-08,
+      "loss": 0.8466,
       "step": 33751
     },
     {
-      "epoch": 0.9269726181648403,
+      "epoch": 0.9577752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 2.784124407931632e-07,
-      "loss": 0.7552,
+      "learning_rate": 9.336920989139075e-08,
+      "loss": 0.7855,
       "step": 33752
     },
     {
-      "epoch": 0.9270000823926835,
+      "epoch": 0.9578036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 2.782040450320278e-07,
-      "loss": 0.7617,
+      "learning_rate": 9.324395113967877e-08,
+      "loss": 0.8102,
       "step": 33753
     },
     {
-      "epoch": 0.9270275466205268,
+      "epoch": 0.9578320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 2.7799572619350736e-07,
-      "loss": 0.6811,
+      "learning_rate": 9.31187760707808e-08,
+      "loss": 0.7249,
       "step": 33754
     },
     {
-      "epoch": 0.92705501084837,
+      "epoch": 0.9578603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 2.777874842792494e-07,
-      "loss": 0.8493,
+      "learning_rate": 9.299368468575376e-08,
+      "loss": 0.8186,
       "step": 33755
     },
     {
-      "epoch": 0.9270824750762132,
+      "epoch": 0.95788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.775793192909049e-07,
-      "loss": 0.7377,
+      "learning_rate": 9.286867698565571e-08,
+      "loss": 0.7904,
       "step": 33756
     },
     {
-      "epoch": 0.9271099393040565,
+      "epoch": 0.9579171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 2.77371231230118e-07,
-      "loss": 0.8302,
+      "learning_rate": 9.274375297154025e-08,
+      "loss": 0.8332,
       "step": 33757
     },
     {
-      "epoch": 0.9271374035318997,
+      "epoch": 0.9579455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 2.7716322009853636e-07,
-      "loss": 0.7822,
+      "learning_rate": 9.261891264446321e-08,
+      "loss": 0.7939,
       "step": 33758
     },
     {
-      "epoch": 0.9271648677597429,
+      "epoch": 0.9579738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 2.769552858978064e-07,
-      "loss": 0.7592,
+      "learning_rate": 9.24941560054804e-08,
+      "loss": 0.7544,
       "step": 33759
     },
     {
-      "epoch": 0.9271923319875862,
+      "epoch": 0.9580022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 2.767474286295713e-07,
-      "loss": 0.741,
+      "learning_rate": 9.23694830556432e-08,
+      "loss": 0.703,
       "step": 33760
     },
     {
-      "epoch": 0.9272197962154294,
+      "epoch": 0.9580306469920545,
       "grad_norm": 0.0,
-      "learning_rate": 2.765396482954763e-07,
-      "loss": 0.7839,
+      "learning_rate": 9.224489379600854e-08,
+      "loss": 0.7164,
       "step": 33761
     },
     {
-      "epoch": 0.9272472604432727,
+      "epoch": 0.9580590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 2.7633194489716687e-07,
-      "loss": 0.821,
+      "learning_rate": 9.212038822762558e-08,
+      "loss": 0.7923,
       "step": 33762
     },
     {
-      "epoch": 0.9272747246711158,
+      "epoch": 0.9580874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 2.761243184362861e-07,
-      "loss": 0.8578,
+      "learning_rate": 9.199596635154684e-08,
+      "loss": 0.917,
       "step": 33763
     },
     {
-      "epoch": 0.9273021888989591,
+      "epoch": 0.9581157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 2.7591676891447707e-07,
-      "loss": 0.7818,
+      "learning_rate": 9.187162816882477e-08,
+      "loss": 0.8519,
       "step": 33764
     },
     {
-      "epoch": 0.9273296531268024,
+      "epoch": 0.9581441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 2.7570929633338075e-07,
-      "loss": 0.819,
+      "learning_rate": 9.174737368050746e-08,
+      "loss": 0.8869,
       "step": 33765
     },
     {
-      "epoch": 0.9273571173546455,
+      "epoch": 0.9581725312145289,
       "grad_norm": 0.0,
-      "learning_rate": 2.7550190069463913e-07,
-      "loss": 0.746,
+      "learning_rate": 9.162320288764515e-08,
+      "loss": 0.9213,
       "step": 33766
     },
     {
-      "epoch": 0.9273845815824888,
+      "epoch": 0.9582009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 2.7529458199989314e-07,
-      "loss": 0.8123,
+      "learning_rate": 9.149911579128812e-08,
+      "loss": 0.8292,
       "step": 33767
     },
     {
-      "epoch": 0.9274120458103321,
+      "epoch": 0.9582292849035188,
       "grad_norm": 0.0,
-      "learning_rate": 2.750873402507825e-07,
-      "loss": 0.9123,
+      "learning_rate": 9.13751123924822e-08,
+      "loss": 0.7248,
       "step": 33768
     },
     {
-      "epoch": 0.9274395100381753,
+      "epoch": 0.9582576617480136,
       "grad_norm": 0.0,
-      "learning_rate": 2.748801754489494e-07,
-      "loss": 0.8106,
+      "learning_rate": 9.125119269227656e-08,
+      "loss": 0.7651,
       "step": 33769
     },
     {
-      "epoch": 0.9274669742660185,
+      "epoch": 0.9582860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 2.746730875960313e-07,
-      "loss": 0.8854,
+      "learning_rate": 9.112735669171924e-08,
+      "loss": 0.7629,
       "step": 33770
     },
     {
-      "epoch": 0.9274944384938617,
+      "epoch": 0.9583144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 2.744660766936669e-07,
-      "loss": 0.8022,
+      "learning_rate": 9.100360439185275e-08,
+      "loss": 0.8536,
       "step": 33771
     },
     {
-      "epoch": 0.927521902721705,
+      "epoch": 0.9583427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.7425914274349374e-07,
-      "loss": 0.7878,
+      "learning_rate": 9.087993579372401e-08,
+      "loss": 0.8005,
       "step": 33772
     },
     {
-      "epoch": 0.9275493669495483,
+      "epoch": 0.9583711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 2.740522857471506e-07,
-      "loss": 0.7238,
+      "learning_rate": 9.075635089837886e-08,
+      "loss": 0.7897,
       "step": 33773
     },
     {
-      "epoch": 0.9275768311773914,
+      "epoch": 0.958399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 2.7384550570627167e-07,
-      "loss": 0.7602,
+      "learning_rate": 9.063284970685982e-08,
+      "loss": 0.837,
       "step": 33774
     },
     {
-      "epoch": 0.9276042954052347,
+      "epoch": 0.958427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 2.7363880262249563e-07,
-      "loss": 0.6573,
+      "learning_rate": 9.050943222021047e-08,
+      "loss": 0.727,
       "step": 33775
     },
     {
-      "epoch": 0.9276317596330779,
+      "epoch": 0.9584562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 2.734321764974579e-07,
-      "loss": 0.78,
+      "learning_rate": 9.038609843947333e-08,
+      "loss": 0.8403,
       "step": 33776
     },
     {
-      "epoch": 0.9276592238609211,
+      "epoch": 0.9584846765039727,
       "grad_norm": 0.0,
-      "learning_rate": 2.7322562733279044e-07,
-      "loss": 0.8438,
+      "learning_rate": 9.026284836569088e-08,
+      "loss": 0.8025,
       "step": 33777
     },
     {
-      "epoch": 0.9276866880887644,
+      "epoch": 0.9585130533484677,
       "grad_norm": 0.0,
-      "learning_rate": 2.730191551301309e-07,
-      "loss": 0.8062,
+      "learning_rate": 9.013968199990342e-08,
+      "loss": 0.704,
       "step": 33778
     },
     {
-      "epoch": 0.9277141523166076,
+      "epoch": 0.9585414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 2.7281275989111124e-07,
-      "loss": 0.7894,
+      "learning_rate": 9.001659934315121e-08,
+      "loss": 0.7958,
       "step": 33779
     },
     {
-      "epoch": 0.9277416165444509,
+      "epoch": 0.9585698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 2.726064416173657e-07,
-      "loss": 0.7977,
+      "learning_rate": 8.989360039647454e-08,
+      "loss": 0.8265,
       "step": 33780
     },
     {
-      "epoch": 0.9277690807722941,
+      "epoch": 0.9585981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 2.7240020031052526e-07,
-      "loss": 0.8314,
+      "learning_rate": 8.977068516091258e-08,
+      "loss": 0.7167,
       "step": 33781
     },
     {
-      "epoch": 0.9277965450001373,
+      "epoch": 0.9586265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 2.721940359722242e-07,
-      "loss": 0.7523,
+      "learning_rate": 8.964785363750228e-08,
+      "loss": 0.8894,
       "step": 33782
     },
     {
-      "epoch": 0.9278240092279806,
+      "epoch": 0.9586549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 2.7198794860408993e-07,
-      "loss": 0.8127,
+      "learning_rate": 8.95251058272828e-08,
+      "loss": 0.729,
       "step": 33783
     },
     {
-      "epoch": 0.9278514734558237,
+      "epoch": 0.958683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.717819382077558e-07,
-      "loss": 0.8092,
+      "learning_rate": 8.940244173128998e-08,
+      "loss": 0.766,
       "step": 33784
     },
     {
-      "epoch": 0.927878937683667,
+      "epoch": 0.9587116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 2.715760047848515e-07,
-      "loss": 0.8127,
+      "learning_rate": 8.927986135056077e-08,
+      "loss": 0.7869,
       "step": 33785
     },
     {
-      "epoch": 0.9279064019115103,
+      "epoch": 0.9587400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 2.713701483370057e-07,
-      "loss": 0.7878,
+      "learning_rate": 8.915736468612879e-08,
+      "loss": 0.8708,
       "step": 33786
     },
     {
-      "epoch": 0.9279338661393535,
+      "epoch": 0.9587684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 2.711643688658483e-07,
-      "loss": 0.8057,
+      "learning_rate": 8.903495173903099e-08,
+      "loss": 0.8215,
       "step": 33787
     },
     {
-      "epoch": 0.9279613303671967,
+      "epoch": 0.9587968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 2.709586663730079e-07,
-      "loss": 0.8396,
+      "learning_rate": 8.891262251029986e-08,
+      "loss": 0.7918,
       "step": 33788
     },
     {
-      "epoch": 0.9279887945950399,
+      "epoch": 0.9588251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 2.707530408601111e-07,
-      "loss": 0.7869,
+      "learning_rate": 8.879037700097015e-08,
+      "loss": 0.8017,
       "step": 33789
     },
     {
-      "epoch": 0.9280162588228832,
+      "epoch": 0.9588535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 2.705474923287843e-07,
-      "loss": 0.9095,
+      "learning_rate": 8.866821521207325e-08,
+      "loss": 0.6861,
       "step": 33790
     },
     {
-      "epoch": 0.9280437230507265,
+      "epoch": 0.9588819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 2.703420207806551e-07,
-      "loss": 0.8097,
+      "learning_rate": 8.854613714464166e-08,
+      "loss": 0.7429,
       "step": 33791
     },
     {
-      "epoch": 0.9280711872785696,
+      "epoch": 0.9589103291713962,
       "grad_norm": 0.0,
-      "learning_rate": 2.701366262173477e-07,
-      "loss": 0.8567,
+      "learning_rate": 8.842414279970568e-08,
+      "loss": 0.7928,
       "step": 33792
     },
     {
-      "epoch": 0.9280986515064129,
+      "epoch": 0.958938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 2.699313086404887e-07,
-      "loss": 0.7947,
+      "learning_rate": 8.830223217829781e-08,
+      "loss": 0.8681,
       "step": 33793
     },
     {
-      "epoch": 0.9281261157342562,
+      "epoch": 0.9589670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 2.697260680517033e-07,
-      "loss": 0.7774,
+      "learning_rate": 8.8180405281445e-08,
+      "loss": 0.8107,
       "step": 33794
     },
     {
-      "epoch": 0.9281535799620994,
+      "epoch": 0.9589954597048809,
       "grad_norm": 0.0,
-      "learning_rate": 2.695209044526126e-07,
-      "loss": 0.7916,
+      "learning_rate": 8.805866211017866e-08,
+      "loss": 0.8014,
       "step": 33795
     },
     {
-      "epoch": 0.9281810441899426,
+      "epoch": 0.9590238365493757,
       "grad_norm": 0.0,
-      "learning_rate": 2.6931581784484294e-07,
-      "loss": 0.766,
+      "learning_rate": 8.793700266552685e-08,
+      "loss": 0.7667,
       "step": 33796
     },
     {
-      "epoch": 0.9282085084177858,
+      "epoch": 0.9590522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.6911080823001643e-07,
-      "loss": 0.8711,
+      "learning_rate": 8.781542694851653e-08,
+      "loss": 0.7643,
       "step": 33797
     },
     {
-      "epoch": 0.9282359726456291,
+      "epoch": 0.9590805902383654,
       "grad_norm": 0.0,
-      "learning_rate": 2.689058756097529e-07,
-      "loss": 0.905,
+      "learning_rate": 8.769393496017465e-08,
+      "loss": 0.7806,
       "step": 33798
     },
     {
-      "epoch": 0.9282634368734723,
+      "epoch": 0.9591089670828604,
       "grad_norm": 0.0,
-      "learning_rate": 2.687010199856765e-07,
-      "loss": 0.771,
+      "learning_rate": 8.757252670152816e-08,
+      "loss": 0.8386,
       "step": 33799
     },
     {
-      "epoch": 0.9282909011013155,
+      "epoch": 0.9591373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 2.684962413594072e-07,
-      "loss": 0.7411,
+      "learning_rate": 8.74512021736007e-08,
+      "loss": 0.8017,
       "step": 33800
     },
     {
-      "epoch": 0.9283183653291588,
+      "epoch": 0.9591657207718501,
       "grad_norm": 0.0,
-      "learning_rate": 2.682915397325647e-07,
-      "loss": 0.8193,
+      "learning_rate": 8.732996137742034e-08,
+      "loss": 0.8022,
       "step": 33801
     },
     {
-      "epoch": 0.928345829557002,
+      "epoch": 0.9591940976163451,
       "grad_norm": 0.0,
-      "learning_rate": 2.6808691510677e-07,
-      "loss": 0.7642,
+      "learning_rate": 8.720880431400847e-08,
+      "loss": 0.8565,
       "step": 33802
     },
     {
-      "epoch": 0.9283732937848452,
+      "epoch": 0.95922247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.678823674836406e-07,
-      "loss": 0.8608,
+      "learning_rate": 8.708773098438872e-08,
+      "loss": 0.763,
       "step": 33803
     },
     {
-      "epoch": 0.9284007580126885,
+      "epoch": 0.9592508513053348,
       "grad_norm": 0.0,
-      "learning_rate": 2.676778968647975e-07,
-      "loss": 0.863,
+      "learning_rate": 8.696674138958583e-08,
+      "loss": 0.7926,
       "step": 33804
     },
     {
-      "epoch": 0.9284282222405317,
+      "epoch": 0.9592792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 2.674735032518549e-07,
-      "loss": 0.8068,
+      "learning_rate": 8.684583553062009e-08,
+      "loss": 0.7992,
       "step": 33805
     },
     {
-      "epoch": 0.928455686468375,
+      "epoch": 0.9593076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 2.672691866464339e-07,
-      "loss": 0.8565,
+      "learning_rate": 8.672501340851291e-08,
+      "loss": 0.8345,
       "step": 33806
     },
     {
-      "epoch": 0.9284831506962182,
+      "epoch": 0.9593359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 2.6706494705014743e-07,
-      "loss": 0.8208,
+      "learning_rate": 8.66042750242857e-08,
+      "loss": 0.8158,
       "step": 33807
     },
     {
-      "epoch": 0.9285106149240614,
+      "epoch": 0.9593643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 2.6686078446461317e-07,
-      "loss": 0.8448,
+      "learning_rate": 8.648362037895764e-08,
+      "loss": 0.8277,
       "step": 33808
     },
     {
-      "epoch": 0.9285380791519047,
+      "epoch": 0.9593927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 2.666566988914465e-07,
-      "loss": 0.8332,
+      "learning_rate": 8.636304947354679e-08,
+      "loss": 0.7535,
       "step": 33809
     },
     {
-      "epoch": 0.9285655433797478,
+      "epoch": 0.9594211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.6645269033226283e-07,
-      "loss": 0.6982,
+      "learning_rate": 8.624256230907458e-08,
+      "loss": 0.8369,
       "step": 33810
     },
     {
-      "epoch": 0.9285930076075911,
+      "epoch": 0.9594494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 2.662487587886764e-07,
-      "loss": 0.8268,
+      "learning_rate": 8.612215888655684e-08,
+      "loss": 0.7511,
       "step": 33811
     },
     {
-      "epoch": 0.9286204718354344,
+      "epoch": 0.959477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 2.6604490426230034e-07,
-      "loss": 0.7763,
+      "learning_rate": 8.600183920701055e-08,
+      "loss": 0.8513,
       "step": 33812
     },
     {
-      "epoch": 0.9286479360632776,
+      "epoch": 0.9595062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 2.6584112675474893e-07,
-      "loss": 0.8107,
+      "learning_rate": 8.588160327145268e-08,
+      "loss": 0.762,
       "step": 33813
     },
     {
-      "epoch": 0.9286754002911208,
+      "epoch": 0.9595346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 2.656374262676309e-07,
-      "loss": 0.8402,
+      "learning_rate": 8.576145108089795e-08,
+      "loss": 0.8537,
       "step": 33814
     },
     {
-      "epoch": 0.928702864518964,
+      "epoch": 0.9595629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 2.654338028025605e-07,
-      "loss": 0.8783,
+      "learning_rate": 8.564138263636224e-08,
+      "loss": 0.8208,
       "step": 33815
     },
     {
-      "epoch": 0.9287303287468073,
+      "epoch": 0.9595913734392736,
       "grad_norm": 0.0,
-      "learning_rate": 2.652302563611486e-07,
-      "loss": 0.8015,
+      "learning_rate": 8.552139793886028e-08,
+      "loss": 0.7588,
       "step": 33816
     },
     {
-      "epoch": 0.9287577929746506,
+      "epoch": 0.9596197502837684,
       "grad_norm": 0.0,
-      "learning_rate": 2.6502678694500626e-07,
-      "loss": 0.8356,
+      "learning_rate": 8.540149698940347e-08,
+      "loss": 0.7666,
       "step": 33817
     },
     {
-      "epoch": 0.9287852572024937,
+      "epoch": 0.9596481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 2.6482339455574323e-07,
-      "loss": 0.7358,
+      "learning_rate": 8.52816797890066e-08,
+      "loss": 0.8232,
       "step": 33818
     },
     {
-      "epoch": 0.928812721430337,
+      "epoch": 0.9596765039727583,
       "grad_norm": 0.0,
-      "learning_rate": 2.6462007919496826e-07,
-      "loss": 0.8481,
+      "learning_rate": 8.516194633868213e-08,
+      "loss": 0.7105,
       "step": 33819
     },
     {
-      "epoch": 0.9288401856581803,
+      "epoch": 0.9597048808172531,
       "grad_norm": 0.0,
-      "learning_rate": 2.6441684086429e-07,
-      "loss": 0.7679,
+      "learning_rate": 8.50422966394382e-08,
+      "loss": 0.7874,
       "step": 33820
     },
     {
-      "epoch": 0.9288676498860234,
+      "epoch": 0.959733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 2.6421367956531827e-07,
-      "loss": 0.8543,
+      "learning_rate": 8.49227306922884e-08,
+      "loss": 0.7616,
       "step": 33821
     },
     {
-      "epoch": 0.9288951141138667,
+      "epoch": 0.9597616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 2.640105952996574e-07,
-      "loss": 0.7935,
+      "learning_rate": 8.480324849824306e-08,
+      "loss": 0.7759,
       "step": 33822
     },
     {
-      "epoch": 0.9289225783417099,
+      "epoch": 0.9597900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 2.638075880689184e-07,
-      "loss": 0.7724,
+      "learning_rate": 8.468385005830914e-08,
+      "loss": 0.779,
       "step": 33823
     },
     {
-      "epoch": 0.9289500425695532,
+      "epoch": 0.9598183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 2.6360465787470314e-07,
-      "loss": 0.7719,
+      "learning_rate": 8.456453537349697e-08,
+      "loss": 0.8511,
       "step": 33824
     },
     {
-      "epoch": 0.9289775067973964,
+      "epoch": 0.9598467650397275,
       "grad_norm": 0.0,
-      "learning_rate": 2.6340180471861933e-07,
-      "loss": 0.779,
+      "learning_rate": 8.444530444481458e-08,
+      "loss": 0.8855,
       "step": 33825
     },
     {
-      "epoch": 0.9290049710252396,
+      "epoch": 0.9598751418842225,
       "grad_norm": 0.0,
-      "learning_rate": 2.631990286022723e-07,
-      "loss": 0.7079,
+      "learning_rate": 8.432615727326676e-08,
+      "loss": 0.8901,
       "step": 33826
     },
     {
-      "epoch": 0.9290324352530829,
+      "epoch": 0.9599035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 2.629963295272664e-07,
-      "loss": 0.7004,
+      "learning_rate": 8.42070938598638e-08,
+      "loss": 0.8695,
       "step": 33827
     },
     {
-      "epoch": 0.929059899480926,
+      "epoch": 0.9599318955732122,
       "grad_norm": 0.0,
-      "learning_rate": 2.6279370749520583e-07,
-      "loss": 0.753,
+      "learning_rate": 8.408811420560826e-08,
+      "loss": 0.8083,
       "step": 33828
     },
     {
-      "epoch": 0.9290873637087693,
+      "epoch": 0.9599602724177072,
       "grad_norm": 0.0,
-      "learning_rate": 2.6259116250769265e-07,
-      "loss": 0.7348,
+      "learning_rate": 8.396921831150484e-08,
+      "loss": 0.7608,
       "step": 33829
     },
     {
-      "epoch": 0.9291148279366126,
+      "epoch": 0.9599886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 2.623886945663312e-07,
-      "loss": 0.8659,
+      "learning_rate": 8.385040617856166e-08,
+      "loss": 0.8414,
       "step": 33830
     },
     {
-      "epoch": 0.9291422921644558,
+      "epoch": 0.9600170261066969,
       "grad_norm": 0.0,
-      "learning_rate": 2.621863036727212e-07,
-      "loss": 0.8553,
+      "learning_rate": 8.373167780777791e-08,
+      "loss": 0.6832,
       "step": 33831
     },
     {
-      "epoch": 0.929169756392299,
+      "epoch": 0.9600454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 2.6198398982846486e-07,
-      "loss": 0.8463,
+      "learning_rate": 8.361303320015945e-08,
+      "loss": 0.8604,
       "step": 33832
     },
     {
-      "epoch": 0.9291972206201423,
+      "epoch": 0.9600737797956868,
       "grad_norm": 0.0,
-      "learning_rate": 2.6178175303516406e-07,
-      "loss": 0.7719,
+      "learning_rate": 8.349447235670771e-08,
+      "loss": 0.6407,
       "step": 33833
     },
     {
-      "epoch": 0.9292246848479855,
+      "epoch": 0.9601021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 2.6157959329441764e-07,
-      "loss": 0.8093,
+      "learning_rate": 8.3375995278423e-08,
+      "loss": 0.772,
       "step": 33834
     },
     {
-      "epoch": 0.9292521490758288,
+      "epoch": 0.9601305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 2.6137751060782756e-07,
-      "loss": 0.7576,
+      "learning_rate": 8.325760196630673e-08,
+      "loss": 0.8157,
       "step": 33835
     },
     {
-      "epoch": 0.9292796133036719,
+      "epoch": 0.9601589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 2.611755049769893e-07,
-      "loss": 0.842,
+      "learning_rate": 8.313929242136031e-08,
+      "loss": 0.6735,
       "step": 33836
     },
     {
-      "epoch": 0.9293070775315152,
+      "epoch": 0.9601872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 2.609735764035037e-07,
-      "loss": 0.8049,
+      "learning_rate": 8.302106664458187e-08,
+      "loss": 0.7344,
       "step": 33837
     },
     {
-      "epoch": 0.9293345417593585,
+      "epoch": 0.9602156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 2.607717248889674e-07,
-      "loss": 0.8325,
+      "learning_rate": 8.290292463697059e-08,
+      "loss": 0.7244,
       "step": 33838
     },
     {
-      "epoch": 0.9293620059872016,
+      "epoch": 0.960244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 2.6056995043497793e-07,
-      "loss": 0.8356,
+      "learning_rate": 8.278486639952343e-08,
+      "loss": 0.9345,
       "step": 33839
     },
     {
-      "epoch": 0.9293894702150449,
+      "epoch": 0.960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 2.603682530431317e-07,
-      "loss": 0.7363,
+      "learning_rate": 8.266689193323852e-08,
+      "loss": 0.9033,
       "step": 33840
     },
     {
-      "epoch": 0.9294169344428881,
+      "epoch": 0.9603007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 2.601666327150243e-07,
-      "loss": 0.8474,
+      "learning_rate": 8.25490012391117e-08,
+      "loss": 0.8618,
       "step": 33841
     },
     {
-      "epoch": 0.9294443986707314,
+      "epoch": 0.9603291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 2.5996508945225206e-07,
-      "loss": 0.7967,
+      "learning_rate": 8.243119431813994e-08,
+      "loss": 0.8251,
       "step": 33842
     },
     {
-      "epoch": 0.9294718628985746,
+      "epoch": 0.9603575482406357,
       "grad_norm": 0.0,
-      "learning_rate": 2.5976362325640826e-07,
-      "loss": 0.7666,
+      "learning_rate": 8.231347117131804e-08,
+      "loss": 0.8074,
       "step": 33843
     },
     {
-      "epoch": 0.9294993271264178,
+      "epoch": 0.9603859250851305,
       "grad_norm": 0.0,
-      "learning_rate": 2.595622341290871e-07,
-      "loss": 0.7581,
+      "learning_rate": 8.219583179963964e-08,
+      "loss": 0.8028,
       "step": 33844
     },
     {
-      "epoch": 0.9295267913542611,
+      "epoch": 0.9604143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 2.59360922071884e-07,
-      "loss": 0.8039,
+      "learning_rate": 8.207827620409836e-08,
+      "loss": 0.7717,
       "step": 33845
     },
     {
-      "epoch": 0.9295542555821044,
+      "epoch": 0.9604426787741204,
       "grad_norm": 0.0,
-      "learning_rate": 2.591596870863899e-07,
-      "loss": 0.8423,
+      "learning_rate": 8.196080438568788e-08,
+      "loss": 0.7946,
       "step": 33846
     },
     {
-      "epoch": 0.9295817198099475,
+      "epoch": 0.9604710556186152,
       "grad_norm": 0.0,
-      "learning_rate": 2.5895852917419804e-07,
-      "loss": 0.6892,
+      "learning_rate": 8.184341634540183e-08,
+      "loss": 0.8726,
       "step": 33847
     },
     {
-      "epoch": 0.9296091840377908,
+      "epoch": 0.9604994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 2.587574483368993e-07,
-      "loss": 0.8328,
+      "learning_rate": 8.172611208422832e-08,
+      "loss": 0.8444,
       "step": 33848
     },
     {
-      "epoch": 0.929636648265634,
+      "epoch": 0.960527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 2.5855644457608466e-07,
-      "loss": 0.7474,
+      "learning_rate": 8.160889160316099e-08,
+      "loss": 0.7829,
       "step": 33849
     },
     {
-      "epoch": 0.9296641124934772,
+      "epoch": 0.9605561861520999,
       "grad_norm": 0.0,
-      "learning_rate": 2.58355517893345e-07,
-      "loss": 0.8028,
+      "learning_rate": 8.149175490319017e-08,
+      "loss": 0.7542,
       "step": 33850
     },
     {
-      "epoch": 0.9296915767213205,
+      "epoch": 0.9605845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 2.5815466829026915e-07,
-      "loss": 0.8679,
+      "learning_rate": 8.137470198530284e-08,
+      "loss": 0.7732,
       "step": 33851
     },
     {
-      "epoch": 0.9297190409491637,
+      "epoch": 0.9606129398410896,
       "grad_norm": 0.0,
-      "learning_rate": 2.579538957684491e-07,
-      "loss": 0.741,
+      "learning_rate": 8.125773285049044e-08,
+      "loss": 0.782,
       "step": 33852
     },
     {
-      "epoch": 0.929746505177007,
+      "epoch": 0.9606413166855846,
       "grad_norm": 0.0,
-      "learning_rate": 2.5775320032947136e-07,
-      "loss": 0.8527,
+      "learning_rate": 8.114084749973882e-08,
+      "loss": 0.7865,
       "step": 33853
     },
     {
-      "epoch": 0.9297739694048501,
+      "epoch": 0.9606696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 2.575525819749236e-07,
-      "loss": 0.8859,
+      "learning_rate": 8.102404593403612e-08,
+      "loss": 0.7883,
       "step": 33854
     },
     {
-      "epoch": 0.9298014336326934,
+      "epoch": 0.9606980703745743,
       "grad_norm": 0.0,
-      "learning_rate": 2.573520407063923e-07,
-      "loss": 0.8055,
+      "learning_rate": 8.090732815437041e-08,
+      "loss": 0.8518,
       "step": 33855
     },
     {
-      "epoch": 0.9298288978605367,
+      "epoch": 0.9607264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 2.571515765254673e-07,
-      "loss": 0.7858,
+      "learning_rate": 8.079069416172646e-08,
+      "loss": 0.8258,
       "step": 33856
     },
     {
-      "epoch": 0.9298563620883799,
+      "epoch": 0.9607548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 2.5695118943373175e-07,
-      "loss": 0.8749,
+      "learning_rate": 8.067414395708795e-08,
+      "loss": 0.7208,
       "step": 33857
     },
     {
-      "epoch": 0.9298838263162231,
+      "epoch": 0.960783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 2.567508794327733e-07,
-      "loss": 0.7927,
+      "learning_rate": 8.055767754144295e-08,
+      "loss": 0.8285,
       "step": 33858
     },
     {
-      "epoch": 0.9299112905440664,
+      "epoch": 0.9608115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 2.5655064652417625e-07,
-      "loss": 0.8668,
+      "learning_rate": 8.04412949157718e-08,
+      "loss": 0.8023,
       "step": 33859
     },
     {
-      "epoch": 0.9299387547719096,
+      "epoch": 0.9608399545970489,
       "grad_norm": 0.0,
-      "learning_rate": 2.563504907095238e-07,
-      "loss": 0.7999,
+      "learning_rate": 8.032499608105815e-08,
+      "loss": 0.8226,
       "step": 33860
     },
     {
-      "epoch": 0.9299662189997528,
+      "epoch": 0.9608683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.5615041199040236e-07,
-      "loss": 0.7629,
+      "learning_rate": 8.020878103828678e-08,
+      "loss": 0.8289,
       "step": 33861
     },
     {
-      "epoch": 0.929993683227596,
+      "epoch": 0.9608967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 2.5595041036839185e-07,
-      "loss": 0.7839,
+      "learning_rate": 8.009264978843689e-08,
+      "loss": 0.7576,
       "step": 33862
     },
     {
-      "epoch": 0.9300211474554393,
+      "epoch": 0.9609250851305334,
       "grad_norm": 0.0,
-      "learning_rate": 2.557504858450766e-07,
-      "loss": 0.7787,
+      "learning_rate": 7.997660233249105e-08,
+      "loss": 0.7516,
       "step": 33863
     },
     {
-      "epoch": 0.9300486116832826,
+      "epoch": 0.9609534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 2.555506384220374e-07,
-      "loss": 0.722,
+      "learning_rate": 7.986063867142846e-08,
+      "loss": 0.6981,
       "step": 33864
     },
     {
-      "epoch": 0.9300760759111257,
+      "epoch": 0.9609818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 2.5535086810085763e-07,
-      "loss": 0.7555,
+      "learning_rate": 7.974475880622834e-08,
+      "loss": 0.7468,
       "step": 33865
     },
     {
-      "epoch": 0.930103540138969,
+      "epoch": 0.9610102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 2.551511748831148e-07,
-      "loss": 0.8016,
+      "learning_rate": 7.962896273787102e-08,
+      "loss": 0.7226,
       "step": 33866
     },
     {
-      "epoch": 0.9301310043668122,
+      "epoch": 0.9610385925085131,
       "grad_norm": 0.0,
-      "learning_rate": 2.54951558770391e-07,
-      "loss": 0.8271,
+      "learning_rate": 7.951325046733238e-08,
+      "loss": 0.7611,
       "step": 33867
     },
     {
-      "epoch": 0.9301584685946555,
+      "epoch": 0.961066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 2.547520197642661e-07,
-      "loss": 0.9069,
+      "learning_rate": 7.939762199559275e-08,
+      "loss": 0.7592,
       "step": 33868
     },
     {
-      "epoch": 0.9301859328224987,
+      "epoch": 0.9610953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 2.5455255786631885e-07,
-      "loss": 0.7872,
+      "learning_rate": 7.928207732362692e-08,
+      "loss": 0.8567,
       "step": 33869
     },
     {
-      "epoch": 0.9302133970503419,
+      "epoch": 0.9611237230419978,
       "grad_norm": 0.0,
-      "learning_rate": 2.543531730781257e-07,
-      "loss": 0.7945,
+      "learning_rate": 7.916661645241186e-08,
+      "loss": 0.7509,
       "step": 33870
     },
     {
-      "epoch": 0.9302408612781852,
+      "epoch": 0.9611520998864926,
       "grad_norm": 0.0,
-      "learning_rate": 2.541538654012654e-07,
-      "loss": 0.7599,
+      "learning_rate": 7.905123938292125e-08,
+      "loss": 0.8344,
       "step": 33871
     },
     {
-      "epoch": 0.9302683255060284,
+      "epoch": 0.9611804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 2.539546348373145e-07,
-      "loss": 0.7602,
+      "learning_rate": 7.893594611613208e-08,
+      "loss": 0.8761,
       "step": 33872
     },
     {
-      "epoch": 0.9302957897338716,
+      "epoch": 0.9612088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 2.5375548138785065e-07,
-      "loss": 0.8243,
+      "learning_rate": 7.882073665301581e-08,
+      "loss": 0.8047,
       "step": 33873
     },
     {
-      "epoch": 0.9303232539617149,
+      "epoch": 0.9612372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 2.535564050544481e-07,
-      "loss": 0.7424,
+      "learning_rate": 7.870561099454721e-08,
+      "loss": 0.9159,
       "step": 33874
     },
     {
-      "epoch": 0.9303507181895581,
+      "epoch": 0.9612656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 2.533574058386823e-07,
-      "loss": 0.8078,
+      "learning_rate": 7.859056914169882e-08,
+      "loss": 0.8171,
       "step": 33875
     },
     {
-      "epoch": 0.9303781824174013,
+      "epoch": 0.961293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 2.531584837421297e-07,
-      "loss": 0.8533,
+      "learning_rate": 7.8475611095441e-08,
+      "loss": 0.8065,
       "step": 33876
     },
     {
-      "epoch": 0.9304056466452446,
+      "epoch": 0.961322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 2.5295963876636245e-07,
-      "loss": 0.7918,
+      "learning_rate": 7.836073685674628e-08,
+      "loss": 0.7346,
       "step": 33877
     },
     {
-      "epoch": 0.9304331108730878,
+      "epoch": 0.9613507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 2.5276087091295363e-07,
-      "loss": 0.8579,
+      "learning_rate": 7.824594642658501e-08,
+      "loss": 0.7669,
       "step": 33878
     },
     {
-      "epoch": 0.930460575100931,
+      "epoch": 0.9613791146424517,
       "grad_norm": 0.0,
-      "learning_rate": 2.5256218018347655e-07,
-      "loss": 0.7892,
+      "learning_rate": 7.81312398059253e-08,
+      "loss": 0.7024,
       "step": 33879
     },
     {
-      "epoch": 0.9304880393287742,
+      "epoch": 0.9614074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 2.5236356657950213e-07,
-      "loss": 0.8601,
+      "learning_rate": 7.801661699573637e-08,
+      "loss": 0.8635,
       "step": 33880
     },
     {
-      "epoch": 0.9305155035566175,
+      "epoch": 0.9614358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 2.5216503010260463e-07,
-      "loss": 0.8178,
+      "learning_rate": 7.790207799698857e-08,
+      "loss": 0.7094,
       "step": 33881
     },
     {
-      "epoch": 0.9305429677844608,
+      "epoch": 0.9614642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 2.5196657075435393e-07,
-      "loss": 0.8028,
+      "learning_rate": 7.778762281064777e-08,
+      "loss": 0.885,
       "step": 33882
     },
     {
-      "epoch": 0.9305704320123039,
+      "epoch": 0.9614926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 2.517681885363177e-07,
-      "loss": 0.8879,
+      "learning_rate": 7.767325143767989e-08,
+      "loss": 0.8145,
       "step": 33883
     },
     {
-      "epoch": 0.9305978962401472,
+      "epoch": 0.9615209988649263,
       "grad_norm": 0.0,
-      "learning_rate": 2.5156988345006907e-07,
-      "loss": 0.7638,
+      "learning_rate": 7.755896387905304e-08,
+      "loss": 0.847,
       "step": 33884
     },
     {
-      "epoch": 0.9306253604679905,
+      "epoch": 0.9615493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 2.513716554971757e-07,
-      "loss": 0.8654,
+      "learning_rate": 7.744476013573088e-08,
+      "loss": 0.7385,
       "step": 33885
     },
     {
-      "epoch": 0.9306528246958337,
+      "epoch": 0.961577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 2.511735046792052e-07,
-      "loss": 0.894,
+      "learning_rate": 7.733064020868042e-08,
+      "loss": 0.6665,
       "step": 33886
     },
     {
-      "epoch": 0.9306802889236769,
+      "epoch": 0.961606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 2.509754309977264e-07,
-      "loss": 0.8377,
+      "learning_rate": 7.721660409886201e-08,
+      "loss": 0.8727,
       "step": 33887
     },
     {
-      "epoch": 0.9307077531515201,
+      "epoch": 0.9616345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 2.507774344543079e-07,
-      "loss": 0.7884,
+      "learning_rate": 7.710265180724264e-08,
+      "loss": 0.749,
       "step": 33888
     },
     {
-      "epoch": 0.9307352173793634,
+      "epoch": 0.9616628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 2.50579515050513e-07,
-      "loss": 0.7855,
+      "learning_rate": 7.698878333478266e-08,
+      "loss": 0.8243,
       "step": 33889
     },
     {
-      "epoch": 0.9307626816072067,
+      "epoch": 0.9616912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 2.503816727879105e-07,
-      "loss": 0.7935,
+      "learning_rate": 7.687499868244464e-08,
+      "loss": 0.8337,
       "step": 33890
     },
     {
-      "epoch": 0.9307901458350498,
+      "epoch": 0.9617196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 2.5018390766806343e-07,
-      "loss": 0.8594,
+      "learning_rate": 7.676129785118891e-08,
+      "loss": 0.8067,
       "step": 33891
     },
     {
-      "epoch": 0.9308176100628931,
+      "epoch": 0.9617480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 2.499862196925407e-07,
-      "loss": 0.777,
+      "learning_rate": 7.664768084197804e-08,
+      "loss": 0.7644,
       "step": 33892
     },
     {
-      "epoch": 0.9308450742907363,
+      "epoch": 0.9617763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 2.497886088629009e-07,
-      "loss": 0.8589,
+      "learning_rate": 7.653414765576905e-08,
+      "loss": 0.7479,
       "step": 33893
     },
     {
-      "epoch": 0.9308725385185795,
+      "epoch": 0.9618047673098752,
       "grad_norm": 0.0,
-      "learning_rate": 2.495910751807129e-07,
-      "loss": 0.7446,
+      "learning_rate": 7.642069829352228e-08,
+      "loss": 0.7952,
       "step": 33894
     },
     {
-      "epoch": 0.9309000027464228,
+      "epoch": 0.96183314415437,
       "grad_norm": 0.0,
-      "learning_rate": 2.4939361864753655e-07,
-      "loss": 0.7514,
+      "learning_rate": 7.630733275619585e-08,
+      "loss": 0.8615,
       "step": 33895
     },
     {
-      "epoch": 0.930927466974266,
+      "epoch": 0.9618615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 2.4919623926493385e-07,
-      "loss": 0.8284,
+      "learning_rate": 7.619405104474786e-08,
+      "loss": 0.8297,
       "step": 33896
     },
     {
-      "epoch": 0.9309549312021093,
+      "epoch": 0.9618898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 2.489989370344681e-07,
-      "loss": 0.6541,
+      "learning_rate": 7.608085316013536e-08,
+      "loss": 0.6254,
       "step": 33897
     },
     {
-      "epoch": 0.9309823954299525,
+      "epoch": 0.9619182746878547,
       "grad_norm": 0.0,
-      "learning_rate": 2.488017119577002e-07,
-      "loss": 0.8706,
+      "learning_rate": 7.596773910331313e-08,
+      "loss": 0.8174,
       "step": 33898
     },
     {
-      "epoch": 0.9310098596577957,
+      "epoch": 0.9619466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 2.4860456403619004e-07,
-      "loss": 0.8625,
+      "learning_rate": 7.585470887523705e-08,
+      "loss": 0.8182,
       "step": 33899
     },
     {
-      "epoch": 0.931037323885639,
+      "epoch": 0.9619750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 2.484074932714986e-07,
-      "loss": 0.8165,
+      "learning_rate": 7.574176247686305e-08,
+      "loss": 0.8813,
       "step": 33900
     },
     {
-      "epoch": 0.9310647881134821,
+      "epoch": 0.9620034052213394,
       "grad_norm": 0.0,
-      "learning_rate": 2.4821049966518464e-07,
-      "loss": 0.8392,
+      "learning_rate": 7.562889990914479e-08,
+      "loss": 0.7323,
       "step": 33901
     },
     {
-      "epoch": 0.9310922523413254,
+      "epoch": 0.9620317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 2.4801358321880575e-07,
-      "loss": 0.7396,
+      "learning_rate": 7.551612117303597e-08,
+      "loss": 0.8743,
       "step": 33902
     },
     {
-      "epoch": 0.9311197165691687,
+      "epoch": 0.9620601589103291,
       "grad_norm": 0.0,
-      "learning_rate": 2.4781674393392074e-07,
-      "loss": 0.6616,
+      "learning_rate": 7.540342626948693e-08,
+      "loss": 0.8319,
       "step": 33903
     },
     {
-      "epoch": 0.9311471807970119,
+      "epoch": 0.9620885357548241,
       "grad_norm": 0.0,
-      "learning_rate": 2.4761998181208837e-07,
-      "loss": 0.8823,
+      "learning_rate": 7.529081519945248e-08,
+      "loss": 0.8804,
       "step": 33904
     },
     {
-      "epoch": 0.9311746450248551,
+      "epoch": 0.962116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 2.474232968548629e-07,
-      "loss": 0.8476,
+      "learning_rate": 7.517828796388294e-08,
+      "loss": 0.8461,
       "step": 33905
     },
     {
-      "epoch": 0.9312021092526983,
+      "epoch": 0.9621452894438138,
       "grad_norm": 0.0,
-      "learning_rate": 2.472266890638042e-07,
-      "loss": 0.7761,
+      "learning_rate": 7.506584456372757e-08,
+      "loss": 0.8045,
       "step": 33906
     },
     {
-      "epoch": 0.9312295734805416,
+      "epoch": 0.9621736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 2.470301584404644e-07,
-      "loss": 0.6845,
+      "learning_rate": 7.495348499993782e-08,
+      "loss": 0.8791,
       "step": 33907
     },
     {
-      "epoch": 0.9312570377083849,
+      "epoch": 0.9622020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 2.468337049864e-07,
-      "loss": 0.7962,
+      "learning_rate": 7.484120927346184e-08,
+      "loss": 0.8616,
       "step": 33908
     },
     {
-      "epoch": 0.931284501936228,
+      "epoch": 0.9622304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 2.466373287031665e-07,
-      "loss": 0.8491,
+      "learning_rate": 7.472901738524885e-08,
+      "loss": 0.8972,
       "step": 33909
     },
     {
-      "epoch": 0.9313119661640713,
+      "epoch": 0.9622587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 2.4644102959231474e-07,
-      "loss": 0.8251,
+      "learning_rate": 7.461690933624476e-08,
+      "loss": 0.8133,
       "step": 33910
     },
     {
-      "epoch": 0.9313394303919146,
+      "epoch": 0.9622871736662884,
       "grad_norm": 0.0,
-      "learning_rate": 2.4624480765540025e-07,
-      "loss": 0.754,
+      "learning_rate": 7.450488512739884e-08,
+      "loss": 0.9294,
       "step": 33911
     },
     {
-      "epoch": 0.9313668946197577,
+      "epoch": 0.9623155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 2.4604866289397623e-07,
-      "loss": 0.7594,
+      "learning_rate": 7.439294475965696e-08,
+      "loss": 0.7785,
       "step": 33912
     },
     {
-      "epoch": 0.931394358847601,
+      "epoch": 0.9623439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 2.458525953095925e-07,
-      "loss": 0.722,
+      "learning_rate": 7.428108823396174e-08,
+      "loss": 0.8501,
       "step": 33913
     },
     {
-      "epoch": 0.9314218230754442,
+      "epoch": 0.9623723041997729,
       "grad_norm": 0.0,
-      "learning_rate": 2.4565660490380116e-07,
-      "loss": 0.8676,
+      "learning_rate": 7.41693155512624e-08,
+      "loss": 0.738,
       "step": 33914
     },
     {
-      "epoch": 0.9314492873032875,
+      "epoch": 0.9624006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 2.454606916781543e-07,
-      "loss": 0.9249,
+      "learning_rate": 7.405762671250039e-08,
+      "loss": 0.7826,
       "step": 33915
     },
     {
-      "epoch": 0.9314767515311307,
+      "epoch": 0.9624290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 2.452648556342008e-07,
-      "loss": 0.7756,
+      "learning_rate": 7.394602171861942e-08,
+      "loss": 0.7716,
       "step": 33916
     },
     {
-      "epoch": 0.9315042157589739,
+      "epoch": 0.9624574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 2.450690967734892e-07,
-      "loss": 0.7933,
+      "learning_rate": 7.383450057056318e-08,
+      "loss": 0.88,
       "step": 33917
     },
     {
-      "epoch": 0.9315316799868172,
+      "epoch": 0.9624858115777526,
       "grad_norm": 0.0,
-      "learning_rate": 2.448734150975718e-07,
-      "loss": 0.8045,
+      "learning_rate": 7.372306326927315e-08,
+      "loss": 0.8802,
       "step": 33918
     },
     {
-      "epoch": 0.9315591442146604,
+      "epoch": 0.9625141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 2.4467781060799277e-07,
-      "loss": 0.798,
+      "learning_rate": 7.361170981569077e-08,
+      "loss": 0.746,
       "step": 33919
     },
     {
-      "epoch": 0.9315866084425036,
+      "epoch": 0.9625425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 2.444822833063021e-07,
-      "loss": 0.8637,
+      "learning_rate": 7.350044021075642e-08,
+      "loss": 0.7804,
       "step": 33920
     },
     {
-      "epoch": 0.9316140726703469,
+      "epoch": 0.9625709421112373,
       "grad_norm": 0.0,
-      "learning_rate": 2.442868331940462e-07,
-      "loss": 0.7471,
+      "learning_rate": 7.338925445541046e-08,
+      "loss": 0.8769,
       "step": 33921
     },
     {
-      "epoch": 0.9316415368981901,
+      "epoch": 0.9625993189557321,
       "grad_norm": 0.0,
-      "learning_rate": 2.4409146027277284e-07,
-      "loss": 0.8,
+      "learning_rate": 7.32781525505899e-08,
+      "loss": 0.7687,
       "step": 33922
     },
     {
-      "epoch": 0.9316690011260333,
+      "epoch": 0.962627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.438961645440252e-07,
-      "loss": 0.7465,
+      "learning_rate": 7.316713449723623e-08,
+      "loss": 0.7574,
       "step": 33923
     },
     {
-      "epoch": 0.9316964653538766,
+      "epoch": 0.9626560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 2.4370094600935313e-07,
-      "loss": 0.8127,
+      "learning_rate": 7.305620029628646e-08,
+      "loss": 0.7825,
       "step": 33924
     },
     {
-      "epoch": 0.9317239295817198,
+      "epoch": 0.9626844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 2.4350580467029763e-07,
-      "loss": 0.7982,
+      "learning_rate": 7.294534994867652e-08,
+      "loss": 0.897,
       "step": 33925
     },
     {
-      "epoch": 0.9317513938095631,
+      "epoch": 0.9627128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 2.433107405284019e-07,
-      "loss": 0.7674,
+      "learning_rate": 7.283458345534456e-08,
+      "loss": 0.9465,
       "step": 33926
     },
     {
-      "epoch": 0.9317788580374062,
+      "epoch": 0.9627412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 2.431157535852113e-07,
-      "loss": 0.7838,
+      "learning_rate": 7.272390081722425e-08,
+      "loss": 0.8598,
       "step": 33927
     },
     {
-      "epoch": 0.9318063222652495,
+      "epoch": 0.9627695800227015,
       "grad_norm": 0.0,
-      "learning_rate": 2.4292084384226923e-07,
-      "loss": 0.8991,
+      "learning_rate": 7.261330203525153e-08,
+      "loss": 0.8241,
       "step": 33928
     },
     {
-      "epoch": 0.9318337864930928,
+      "epoch": 0.9627979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 2.427260113011154e-07,
-      "loss": 0.8672,
+      "learning_rate": 7.250278711036007e-08,
+      "loss": 0.7854,
       "step": 33929
     },
     {
-      "epoch": 0.931861250720936,
+      "epoch": 0.9628263337116912,
       "grad_norm": 0.0,
-      "learning_rate": 2.425312559632953e-07,
-      "loss": 0.7726,
+      "learning_rate": 7.23923560434836e-08,
+      "loss": 0.756,
       "step": 33930
     },
     {
-      "epoch": 0.9318887149487792,
+      "epoch": 0.9628547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 2.4233657783034657e-07,
-      "loss": 0.8459,
+      "learning_rate": 7.228200883555581e-08,
+      "loss": 0.7317,
       "step": 33931
     },
     {
-      "epoch": 0.9319161791766224,
+      "epoch": 0.9628830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 2.4214197690380913e-07,
-      "loss": 0.7913,
+      "learning_rate": 7.217174548750926e-08,
+      "loss": 0.8224,
       "step": 33932
     },
     {
-      "epoch": 0.9319436434044657,
+      "epoch": 0.9629114642451759,
       "grad_norm": 0.0,
-      "learning_rate": 2.419474531852262e-07,
-      "loss": 0.7785,
+      "learning_rate": 7.206156600027214e-08,
+      "loss": 0.755,
       "step": 33933
     },
     {
-      "epoch": 0.931971107632309,
+      "epoch": 0.9629398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 2.417530066761342e-07,
-      "loss": 0.8204,
+      "learning_rate": 7.195147037477812e-08,
+      "loss": 0.7099,
       "step": 33934
     },
     {
-      "epoch": 0.9319985718601521,
+      "epoch": 0.9629682179341658,
       "grad_norm": 0.0,
-      "learning_rate": 2.415586373780721e-07,
-      "loss": 0.6133,
+      "learning_rate": 7.184145861195757e-08,
+      "loss": 0.7454,
       "step": 33935
     },
     {
-      "epoch": 0.9320260360879954,
+      "epoch": 0.9629965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 2.4136434529257955e-07,
-      "loss": 0.8157,
+      "learning_rate": 7.173153071273753e-08,
+      "loss": 0.8744,
       "step": 33936
     },
     {
-      "epoch": 0.9320535003158387,
+      "epoch": 0.9630249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 2.411701304211911e-07,
-      "loss": 0.8195,
+      "learning_rate": 7.162168667804726e-08,
+      "loss": 0.8315,
       "step": 33937
     },
     {
-      "epoch": 0.9320809645436818,
+      "epoch": 0.9630533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 2.4097599276544424e-07,
-      "loss": 0.8214,
+      "learning_rate": 7.151192650881488e-08,
+      "loss": 0.8031,
       "step": 33938
     },
     {
-      "epoch": 0.9321084287715251,
+      "epoch": 0.9630817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 2.407819323268767e-07,
-      "loss": 0.7431,
+      "learning_rate": 7.140225020596858e-08,
+      "loss": 0.8405,
       "step": 33939
     },
     {
-      "epoch": 0.9321358929993683,
+      "epoch": 0.9631101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 2.4058794910702286e-07,
-      "loss": 0.7059,
+      "learning_rate": 7.129265777043204e-08,
+      "loss": 0.7271,
       "step": 33940
     },
     {
-      "epoch": 0.9321633572272116,
+      "epoch": 0.963138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 2.40394043107417e-07,
-      "loss": 0.8461,
+      "learning_rate": 7.118314920313452e-08,
+      "loss": 0.9159,
       "step": 33941
     },
     {
-      "epoch": 0.9321908214550548,
+      "epoch": 0.96316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.402002143295945e-07,
-      "loss": 0.7401,
+      "learning_rate": 7.10737245049986e-08,
+      "loss": 0.7791,
       "step": 33942
     },
     {
-      "epoch": 0.932218285682898,
+      "epoch": 0.9631952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 2.4000646277508864e-07,
-      "loss": 0.7913,
+      "learning_rate": 7.096438367694913e-08,
+      "loss": 0.8168,
       "step": 33943
     },
     {
-      "epoch": 0.9322457499107413,
+      "epoch": 0.9632236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 2.398127884454315e-07,
-      "loss": 0.7618,
+      "learning_rate": 7.08551267199098e-08,
+      "loss": 0.9449,
       "step": 33944
     },
     {
-      "epoch": 0.9322732141385844,
+      "epoch": 0.9632519863791147,
       "grad_norm": 0.0,
-      "learning_rate": 2.3961919134215637e-07,
-      "loss": 0.7425,
+      "learning_rate": 7.074595363480318e-08,
+      "loss": 0.8184,
       "step": 33945
     },
     {
-      "epoch": 0.9323006783664277,
+      "epoch": 0.9632803632236095,
       "grad_norm": 0.0,
-      "learning_rate": 2.394256714667953e-07,
-      "loss": 0.8194,
+      "learning_rate": 7.063686442255302e-08,
+      "loss": 0.7368,
       "step": 33946
     },
     {
-      "epoch": 0.932328142594271,
+      "epoch": 0.9633087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 2.3923222882088037e-07,
-      "loss": 0.8832,
+      "learning_rate": 7.052785908407855e-08,
+      "loss": 0.8143,
       "step": 33947
     },
     {
-      "epoch": 0.9323556068221142,
+      "epoch": 0.9633371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 2.390388634059393e-07,
-      "loss": 0.8143,
+      "learning_rate": 7.041893762030127e-08,
+      "loss": 0.8876,
       "step": 33948
     },
     {
-      "epoch": 0.9323830710499574,
+      "epoch": 0.9633654937570942,
       "grad_norm": 0.0,
-      "learning_rate": 2.3884557522350525e-07,
-      "loss": 0.8894,
+      "learning_rate": 7.031010003214267e-08,
+      "loss": 0.8899,
       "step": 33949
     },
     {
-      "epoch": 0.9324105352778007,
+      "epoch": 0.9633938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 2.3865236427510485e-07,
-      "loss": 0.7518,
+      "learning_rate": 7.020134632052089e-08,
+      "loss": 0.7814,
       "step": 33950
     },
     {
-      "epoch": 0.9324379995056439,
+      "epoch": 0.963422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 2.3845923056226793e-07,
-      "loss": 0.739,
+      "learning_rate": 7.00926764863541e-08,
+      "loss": 0.7469,
       "step": 33951
     },
     {
-      "epoch": 0.9324654637334872,
+      "epoch": 0.9634506242905789,
       "grad_norm": 0.0,
-      "learning_rate": 2.382661740865233e-07,
-      "loss": 0.8387,
+      "learning_rate": 6.998409053056154e-08,
+      "loss": 0.7856,
       "step": 33952
     },
     {
-      "epoch": 0.9324929279613303,
+      "epoch": 0.9634790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 2.380731948493986e-07,
-      "loss": 0.7112,
+      "learning_rate": 6.987558845405917e-08,
+      "loss": 0.7812,
       "step": 33953
     },
     {
-      "epoch": 0.9325203921891736,
+      "epoch": 0.9635073779795686,
       "grad_norm": 0.0,
-      "learning_rate": 2.3788029285241932e-07,
-      "loss": 0.8209,
+      "learning_rate": 6.976717025776514e-08,
+      "loss": 0.8531,
       "step": 33954
     },
     {
-      "epoch": 0.9325478564170169,
+      "epoch": 0.9635357548240636,
       "grad_norm": 0.0,
-      "learning_rate": 2.3768746809711307e-07,
-      "loss": 0.7192,
+      "learning_rate": 6.965883594259315e-08,
+      "loss": 0.7694,
       "step": 33955
     },
     {
-      "epoch": 0.93257532064486,
+      "epoch": 0.9635641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 2.3749472058500533e-07,
-      "loss": 0.8167,
+      "learning_rate": 6.955058550945915e-08,
+      "loss": 0.7556,
       "step": 33956
     },
     {
-      "epoch": 0.9326027848727033,
+      "epoch": 0.9635925085130533,
       "grad_norm": 0.0,
-      "learning_rate": 2.3730205031762154e-07,
-      "loss": 0.8288,
+      "learning_rate": 6.944241895927906e-08,
+      "loss": 0.7407,
       "step": 33957
     },
     {
-      "epoch": 0.9326302491005466,
+      "epoch": 0.9636208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 2.3710945729648493e-07,
-      "loss": 0.8447,
+      "learning_rate": 6.933433629296438e-08,
+      "loss": 0.823,
       "step": 33958
     },
     {
-      "epoch": 0.9326577133283898,
+      "epoch": 0.9636492622020432,
       "grad_norm": 0.0,
-      "learning_rate": 2.3691694152311985e-07,
-      "loss": 0.7901,
+      "learning_rate": 6.922633751142882e-08,
+      "loss": 0.8366,
       "step": 33959
     },
     {
-      "epoch": 0.932685177556233,
+      "epoch": 0.963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 2.3672450299905058e-07,
-      "loss": 0.8317,
+      "learning_rate": 6.911842261558499e-08,
+      "loss": 0.8427,
       "step": 33960
     },
     {
-      "epoch": 0.9327126417840762,
+      "epoch": 0.9637060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 2.365321417257971e-07,
-      "loss": 0.8045,
+      "learning_rate": 6.901059160634437e-08,
+      "loss": 0.7893,
       "step": 33961
     },
     {
-      "epoch": 0.9327401060119195,
+      "epoch": 0.9637343927355279,
       "grad_norm": 0.0,
-      "learning_rate": 2.3633985770488477e-07,
-      "loss": 0.8173,
+      "learning_rate": 6.890284448461738e-08,
+      "loss": 0.8287,
       "step": 33962
     },
     {
-      "epoch": 0.9327675702397628,
+      "epoch": 0.9637627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 2.3614765093783353e-07,
-      "loss": 0.8556,
+      "learning_rate": 6.879518125131545e-08,
+      "loss": 0.677,
       "step": 33963
     },
     {
-      "epoch": 0.9327950344676059,
+      "epoch": 0.9637911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 2.359555214261644e-07,
-      "loss": 0.7641,
+      "learning_rate": 6.868760190734568e-08,
+      "loss": 0.7993,
       "step": 33964
     },
     {
-      "epoch": 0.9328224986954492,
+      "epoch": 0.9638195232690124,
       "grad_norm": 0.0,
-      "learning_rate": 2.3576346917139725e-07,
-      "loss": 0.7718,
+      "learning_rate": 6.858010645361957e-08,
+      "loss": 0.8263,
       "step": 33965
     },
     {
-      "epoch": 0.9328499629232924,
+      "epoch": 0.9638479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 2.35571494175052e-07,
-      "loss": 0.8123,
+      "learning_rate": 6.847269489104303e-08,
+      "loss": 0.7765,
       "step": 33966
     },
     {
-      "epoch": 0.9328774271511356,
+      "epoch": 0.9638762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 2.353795964386474e-07,
-      "loss": 0.7926,
+      "learning_rate": 6.836536722052423e-08,
+      "loss": 0.8036,
       "step": 33967
     },
     {
-      "epoch": 0.9329048913789789,
+      "epoch": 0.9639046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 2.3518777596370113e-07,
-      "loss": 0.8303,
+      "learning_rate": 6.825812344296911e-08,
+      "loss": 0.8348,
       "step": 33968
     },
     {
-      "epoch": 0.9329323556068221,
+      "epoch": 0.9639330306469921,
       "grad_norm": 0.0,
-      "learning_rate": 2.34996032751732e-07,
-      "loss": 0.7672,
+      "learning_rate": 6.815096355928475e-08,
+      "loss": 0.8076,
       "step": 33969
     },
     {
-      "epoch": 0.9329598198346654,
+      "epoch": 0.963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 2.3480436680425656e-07,
-      "loss": 0.7849,
+      "learning_rate": 6.804388757037594e-08,
+      "loss": 0.7331,
       "step": 33970
     },
     {
-      "epoch": 0.9329872840625086,
+      "epoch": 0.9639897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 2.346127781227925e-07,
-      "loss": 0.8134,
+      "learning_rate": 6.793689547714644e-08,
+      "loss": 0.753,
       "step": 33971
     },
     {
-      "epoch": 0.9330147482903518,
+      "epoch": 0.9640181611804768,
       "grad_norm": 0.0,
-      "learning_rate": 2.3442126670885524e-07,
-      "loss": 0.753,
+      "learning_rate": 6.782998728049995e-08,
+      "loss": 0.7763,
       "step": 33972
     },
     {
-      "epoch": 0.9330422125181951,
+      "epoch": 0.9640465380249716,
       "grad_norm": 0.0,
-      "learning_rate": 2.3422983256395914e-07,
-      "loss": 0.6001,
+      "learning_rate": 6.772316298134129e-08,
+      "loss": 0.8283,
       "step": 33973
     },
     {
-      "epoch": 0.9330696767460382,
+      "epoch": 0.9640749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 2.3403847568961968e-07,
-      "loss": 0.8023,
+      "learning_rate": 6.761642258056977e-08,
+      "loss": 0.8775,
       "step": 33974
     },
     {
-      "epoch": 0.9330971409738815,
+      "epoch": 0.9641032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 2.3384719608735007e-07,
-      "loss": 0.8381,
+      "learning_rate": 6.75097660790902e-08,
+      "loss": 0.8241,
       "step": 33975
     },
     {
-      "epoch": 0.9331246052017248,
+      "epoch": 0.9641316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.3365599375866466e-07,
-      "loss": 0.8269,
+      "learning_rate": 6.740319347780189e-08,
+      "loss": 0.7958,
       "step": 33976
     },
     {
-      "epoch": 0.933152069429568,
+      "epoch": 0.9641600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 2.334648687050767e-07,
-      "loss": 0.6993,
+      "learning_rate": 6.72967047776052e-08,
+      "loss": 0.8349,
       "step": 33977
     },
     {
-      "epoch": 0.9331795336574112,
+      "epoch": 0.964188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 2.3327382092809713e-07,
-      "loss": 0.7915,
+      "learning_rate": 6.719029997939942e-08,
+      "loss": 0.8764,
       "step": 33978
     },
     {
-      "epoch": 0.9332069978852544,
+      "epoch": 0.964216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 2.3308285042923927e-07,
-      "loss": 0.7189,
+      "learning_rate": 6.708397908408382e-08,
+      "loss": 0.8153,
       "step": 33979
     },
     {
-      "epoch": 0.9332344621130977,
+      "epoch": 0.9642451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 2.328919572100119e-07,
-      "loss": 0.8867,
+      "learning_rate": 6.69777420925577e-08,
+      "loss": 0.8473,
       "step": 33980
     },
     {
-      "epoch": 0.933261926340941,
+      "epoch": 0.9642735527809307,
       "grad_norm": 0.0,
-      "learning_rate": 2.3270114127192822e-07,
-      "loss": 0.7198,
+      "learning_rate": 6.687158900571588e-08,
+      "loss": 0.7697,
       "step": 33981
     },
     {
-      "epoch": 0.9332893905687841,
+      "epoch": 0.9643019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 2.3251040261649592e-07,
-      "loss": 0.7003,
+      "learning_rate": 6.676551982445545e-08,
+      "loss": 0.8839,
       "step": 33982
     },
     {
-      "epoch": 0.9333168547966274,
+      "epoch": 0.9643303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 2.3231974124522606e-07,
-      "loss": 0.8745,
+      "learning_rate": 6.665953454967456e-08,
+      "loss": 0.7884,
       "step": 33983
     },
     {
-      "epoch": 0.9333443190244707,
+      "epoch": 0.9643586833144154,
       "grad_norm": 0.0,
-      "learning_rate": 2.3212915715962515e-07,
-      "loss": 0.8014,
+      "learning_rate": 6.655363318226693e-08,
+      "loss": 0.8294,
       "step": 33984
     },
     {
-      "epoch": 0.9333717832523138,
+      "epoch": 0.9643870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 2.319386503612009e-07,
-      "loss": 0.7604,
+      "learning_rate": 6.644781572312631e-08,
+      "loss": 0.7868,
       "step": 33985
     },
     {
-      "epoch": 0.9333992474801571,
+      "epoch": 0.9644154370034053,
       "grad_norm": 0.0,
-      "learning_rate": 2.317482208514632e-07,
-      "loss": 0.7572,
+      "learning_rate": 6.634208217314864e-08,
+      "loss": 0.8892,
       "step": 33986
     },
     {
-      "epoch": 0.9334267117080003,
+      "epoch": 0.9644438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 2.3155786863191642e-07,
-      "loss": 0.724,
+      "learning_rate": 6.623643253322543e-08,
+      "loss": 0.7969,
       "step": 33987
     },
     {
-      "epoch": 0.9334541759358436,
+      "epoch": 0.964472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 2.3136759370406937e-07,
-      "loss": 0.8318,
+      "learning_rate": 6.61308668042504e-08,
+      "loss": 0.8914,
       "step": 33988
     },
     {
-      "epoch": 0.9334816401636868,
+      "epoch": 0.96450056753689,
       "grad_norm": 0.0,
-      "learning_rate": 2.3117739606942413e-07,
-      "loss": 0.8524,
+      "learning_rate": 6.602538498711397e-08,
+      "loss": 0.7843,
       "step": 33989
     },
     {
-      "epoch": 0.93350910439153,
+      "epoch": 0.9645289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 2.3098727572948954e-07,
-      "loss": 0.8505,
+      "learning_rate": 6.591998708270652e-08,
+      "loss": 0.8184,
       "step": 33990
     },
     {
-      "epoch": 0.9335365686193733,
+      "epoch": 0.9645573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 2.307972326857666e-07,
-      "loss": 0.8551,
+      "learning_rate": 6.58146730919218e-08,
+      "loss": 0.7259,
       "step": 33991
     },
     {
-      "epoch": 0.9335640328472165,
+      "epoch": 0.9645856980703745,
       "grad_norm": 0.0,
-      "learning_rate": 2.3060726693975966e-07,
-      "loss": 0.7876,
+      "learning_rate": 6.570944301564575e-08,
+      "loss": 0.8463,
       "step": 33992
     },
     {
-      "epoch": 0.9335914970750597,
+      "epoch": 0.9646140749148695,
       "grad_norm": 0.0,
-      "learning_rate": 2.3041737849297197e-07,
-      "loss": 0.9113,
+      "learning_rate": 6.560429685476765e-08,
+      "loss": 0.8422,
       "step": 33993
     },
     {
-      "epoch": 0.933618961302903,
+      "epoch": 0.9646424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 2.3022756734690677e-07,
-      "loss": 0.8784,
+      "learning_rate": 6.549923461017904e-08,
+      "loss": 0.906,
       "step": 33994
     },
     {
-      "epoch": 0.9336464255307462,
+      "epoch": 0.9646708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 2.300378335030662e-07,
-      "loss": 0.7713,
+      "learning_rate": 6.539425628276252e-08,
+      "loss": 0.8444,
       "step": 33995
     },
     {
-      "epoch": 0.9336738897585894,
+      "epoch": 0.9646992054483542,
       "grad_norm": 0.0,
-      "learning_rate": 2.2984817696295013e-07,
-      "loss": 0.9405,
+      "learning_rate": 6.52893618734085e-08,
+      "loss": 0.7883,
       "step": 33996
     },
     {
-      "epoch": 0.9337013539864327,
+      "epoch": 0.964727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 2.2965859772806188e-07,
-      "loss": 0.802,
+      "learning_rate": 6.518455138300295e-08,
+      "loss": 0.8208,
       "step": 33997
     },
     {
-      "epoch": 0.9337288182142759,
+      "epoch": 0.9647559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 2.2946909579989684e-07,
-      "loss": 0.8007,
+      "learning_rate": 6.507982481242847e-08,
+      "loss": 0.8117,
       "step": 33998
     },
     {
-      "epoch": 0.9337562824421192,
+      "epoch": 0.9647843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 2.2927967117995831e-07,
-      "loss": 0.7992,
+      "learning_rate": 6.497518216257104e-08,
+      "loss": 0.6667,
       "step": 33999
     },
     {
-      "epoch": 0.9337837466699623,
+      "epoch": 0.9648127128263337,
       "grad_norm": 0.0,
-      "learning_rate": 2.2909032386974283e-07,
-      "loss": 0.8167,
+      "learning_rate": 6.48706234343166e-08,
+      "loss": 0.7803,
       "step": 34000
     },
     {
-      "epoch": 0.9338112108978056,
+      "epoch": 0.9648410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 2.2890105387075146e-07,
-      "loss": 0.8115,
+      "learning_rate": 6.476614862854558e-08,
+      "loss": 0.8948,
       "step": 34001
     },
     {
-      "epoch": 0.9338386751256489,
+      "epoch": 0.9648694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 2.2871186118447852e-07,
-      "loss": 0.8688,
+      "learning_rate": 6.46617577461417e-08,
+      "loss": 0.7105,
       "step": 34002
     },
     {
-      "epoch": 0.933866139353492,
+      "epoch": 0.9648978433598184,
       "grad_norm": 0.0,
-      "learning_rate": 2.2852274581242174e-07,
-      "loss": 0.6717,
+      "learning_rate": 6.455745078798647e-08,
+      "loss": 0.8808,
       "step": 34003
     },
     {
-      "epoch": 0.9338936035813353,
+      "epoch": 0.9649262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 2.2833370775607877e-07,
-      "loss": 0.7625,
+      "learning_rate": 6.445322775496032e-08,
+      "loss": 0.7983,
       "step": 34004
     },
     {
-      "epoch": 0.9339210678091785,
+      "epoch": 0.9649545970488081,
       "grad_norm": 0.0,
-      "learning_rate": 2.2814474701694512e-07,
-      "loss": 0.9221,
+      "learning_rate": 6.434908864794475e-08,
+      "loss": 0.7789,
       "step": 34005
     },
     {
-      "epoch": 0.9339485320370218,
+      "epoch": 0.9649829738933031,
       "grad_norm": 0.0,
-      "learning_rate": 2.2795586359651622e-07,
-      "loss": 0.7867,
+      "learning_rate": 6.424503346782018e-08,
+      "loss": 0.8278,
       "step": 34006
     },
     {
-      "epoch": 0.933975996264865,
+      "epoch": 0.965011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 2.2776705749628536e-07,
-      "loss": 0.85,
+      "learning_rate": 6.414106221546368e-08,
+      "loss": 0.7964,
       "step": 34007
     },
     {
-      "epoch": 0.9340034604927082,
+      "epoch": 0.9650397275822928,
       "grad_norm": 0.0,
-      "learning_rate": 2.2757832871774576e-07,
-      "loss": 0.7863,
+      "learning_rate": 6.403717489175343e-08,
+      "loss": 0.8026,
       "step": 34008
     },
     {
-      "epoch": 0.9340309247205515,
+      "epoch": 0.9650681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 2.273896772623918e-07,
-      "loss": 0.8844,
+      "learning_rate": 6.393337149756984e-08,
+      "loss": 0.7517,
       "step": 34009
     },
     {
-      "epoch": 0.9340583889483948,
+      "epoch": 0.9650964812712827,
       "grad_norm": 0.0,
-      "learning_rate": 2.2720110313171672e-07,
-      "loss": 0.7453,
+      "learning_rate": 6.382965203378665e-08,
+      "loss": 0.7001,
       "step": 34010
     },
     {
-      "epoch": 0.9340858531762379,
+      "epoch": 0.9651248581157775,
       "grad_norm": 0.0,
-      "learning_rate": 2.2701260632721266e-07,
-      "loss": 0.7538,
+      "learning_rate": 6.372601650127986e-08,
+      "loss": 0.7436,
       "step": 34011
     },
     {
-      "epoch": 0.9341133174040812,
+      "epoch": 0.9651532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 2.2682418685037066e-07,
-      "loss": 0.7094,
+      "learning_rate": 6.36224649009276e-08,
+      "loss": 0.7837,
       "step": 34012
     },
     {
-      "epoch": 0.9341407816319244,
+      "epoch": 0.9651816118047674,
       "grad_norm": 0.0,
-      "learning_rate": 2.2663584470268284e-07,
-      "loss": 0.8787,
+      "learning_rate": 6.351899723360256e-08,
+      "loss": 0.6897,
       "step": 34013
     },
     {
-      "epoch": 0.9341682458597677,
+      "epoch": 0.9652099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 2.264475798856358e-07,
-      "loss": 0.77,
+      "learning_rate": 6.341561350017955e-08,
+      "loss": 0.8688,
       "step": 34014
     },
     {
-      "epoch": 0.9341957100876109,
+      "epoch": 0.9652383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 2.262593924007217e-07,
-      "loss": 0.8106,
+      "learning_rate": 6.331231370153123e-08,
+      "loss": 0.7578,
       "step": 34015
     },
     {
-      "epoch": 0.9342231743154541,
+      "epoch": 0.9652667423382519,
       "grad_norm": 0.0,
-      "learning_rate": 2.2607128224942932e-07,
-      "loss": 0.8192,
+      "learning_rate": 6.320909783853136e-08,
+      "loss": 0.8584,
       "step": 34016
     },
     {
-      "epoch": 0.9342506385432974,
+      "epoch": 0.9652951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 2.2588324943324636e-07,
-      "loss": 0.809,
+      "learning_rate": 6.310596591205031e-08,
+      "loss": 0.8283,
       "step": 34017
     },
     {
-      "epoch": 0.9342781027711405,
+      "epoch": 0.9653234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 2.2569529395366385e-07,
-      "loss": 0.7646,
+      "learning_rate": 6.300291792295965e-08,
+      "loss": 0.9131,
       "step": 34018
     },
     {
-      "epoch": 0.9343055669989838,
+      "epoch": 0.9653518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 2.2550741581216394e-07,
-      "loss": 0.7934,
+      "learning_rate": 6.289995387212977e-08,
+      "loss": 0.7585,
       "step": 34019
     },
     {
-      "epoch": 0.9343330312268271,
+      "epoch": 0.9653802497162316,
       "grad_norm": 0.0,
-      "learning_rate": 2.253196150102366e-07,
-      "loss": 0.8439,
+      "learning_rate": 6.27970737604311e-08,
+      "loss": 0.715,
       "step": 34020
     },
     {
-      "epoch": 0.9343604954546703,
+      "epoch": 0.9654086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 2.251318915493672e-07,
-      "loss": 0.8306,
+      "learning_rate": 6.269427758873182e-08,
+      "loss": 0.8032,
       "step": 34021
     },
     {
-      "epoch": 0.9343879596825135,
+      "epoch": 0.9654370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 2.249442454310391e-07,
-      "loss": 0.8832,
+      "learning_rate": 6.259156535790012e-08,
+      "loss": 0.7753,
       "step": 34022
     },
     {
-      "epoch": 0.9344154239103568,
+      "epoch": 0.9654653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 2.2475667665673995e-07,
-      "loss": 0.729,
+      "learning_rate": 6.248893706880643e-08,
+      "loss": 0.8111,
       "step": 34023
     },
     {
-      "epoch": 0.9344428881382,
+      "epoch": 0.9654937570942111,
       "grad_norm": 0.0,
-      "learning_rate": 2.2456918522795302e-07,
-      "loss": 0.839,
+      "learning_rate": 6.23863927223134e-08,
+      "loss": 0.789,
       "step": 34024
     },
     {
-      "epoch": 0.9344703523660433,
+      "epoch": 0.965522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 2.2438177114615934e-07,
-      "loss": 0.8177,
+      "learning_rate": 6.22839323192892e-08,
+      "loss": 0.8445,
       "step": 34025
     },
     {
-      "epoch": 0.9344978165938864,
+      "epoch": 0.9655505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 2.241944344128455e-07,
-      "loss": 0.799,
+      "learning_rate": 6.218155586060093e-08,
+      "loss": 0.7977,
       "step": 34026
     },
     {
-      "epoch": 0.9345252808217297,
+      "epoch": 0.9655788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 2.2400717502949144e-07,
-      "loss": 0.7452,
+      "learning_rate": 6.207926334711123e-08,
+      "loss": 0.7561,
       "step": 34027
     },
     {
-      "epoch": 0.934552745049573,
+      "epoch": 0.9656072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 2.238199929975793e-07,
-      "loss": 0.8273,
+      "learning_rate": 6.197705477968386e-08,
+      "loss": 0.86,
       "step": 34028
     },
     {
-      "epoch": 0.9345802092774161,
+      "epoch": 0.9656356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 2.236328883185912e-07,
-      "loss": 0.725,
+      "learning_rate": 6.187493015918367e-08,
+      "loss": 0.7321,
       "step": 34029
     },
     {
-      "epoch": 0.9346076735052594,
+      "epoch": 0.9656640181611805,
       "grad_norm": 0.0,
-      "learning_rate": 2.23445860994006e-07,
-      "loss": 0.804,
+      "learning_rate": 6.177288948647331e-08,
+      "loss": 0.7317,
       "step": 34030
     },
     {
-      "epoch": 0.9346351377331026,
+      "epoch": 0.9656923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 2.2325891102530473e-07,
-      "loss": 0.7772,
+      "learning_rate": 6.16709327624132e-08,
+      "loss": 0.8723,
       "step": 34031
     },
     {
-      "epoch": 0.9346626019609459,
+      "epoch": 0.9657207718501702,
       "grad_norm": 0.0,
-      "learning_rate": 2.2307203841396507e-07,
-      "loss": 0.7729,
+      "learning_rate": 6.1569059987866e-08,
+      "loss": 0.7495,
       "step": 34032
     },
     {
-      "epoch": 0.9346900661887891,
+      "epoch": 0.9657491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 2.22885243161467e-07,
-      "loss": 0.8162,
+      "learning_rate": 6.1467271163691e-08,
+      "loss": 0.8297,
       "step": 34033
     },
     {
-      "epoch": 0.9347175304166323,
+      "epoch": 0.9657775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 2.2269852526928816e-07,
-      "loss": 0.7195,
+      "learning_rate": 6.136556629074863e-08,
+      "loss": 0.9274,
       "step": 34034
     },
     {
-      "epoch": 0.9347449946444756,
+      "epoch": 0.9658059023836549,
       "grad_norm": 0.0,
-      "learning_rate": 2.2251188473890628e-07,
-      "loss": 0.804,
+      "learning_rate": 6.126394536989933e-08,
+      "loss": 0.7884,
       "step": 34035
     },
     {
-      "epoch": 0.9347724588723189,
+      "epoch": 0.9658342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 2.2232532157179798e-07,
-      "loss": 0.8258,
+      "learning_rate": 6.116240840199906e-08,
+      "loss": 0.8539,
       "step": 34036
     },
     {
-      "epoch": 0.934799923100162,
+      "epoch": 0.9658626560726448,
       "grad_norm": 0.0,
-      "learning_rate": 2.2213883576943984e-07,
-      "loss": 0.7753,
+      "learning_rate": 6.106095538790713e-08,
+      "loss": 0.7592,
       "step": 34037
     },
     {
-      "epoch": 0.9348273873280053,
+      "epoch": 0.9658910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 2.2195242733330624e-07,
-      "loss": 0.8278,
+      "learning_rate": 6.095958632848065e-08,
+      "loss": 0.7917,
       "step": 34038
     },
     {
-      "epoch": 0.9348548515558485,
+      "epoch": 0.9659194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 2.2176609626487155e-07,
-      "loss": 0.7098,
+      "learning_rate": 6.085830122457448e-08,
+      "loss": 0.9514,
       "step": 34039
     },
     {
-      "epoch": 0.9348823157836917,
+      "epoch": 0.9659477866061293,
       "grad_norm": 0.0,
-      "learning_rate": 2.2157984256561238e-07,
-      "loss": 0.7603,
+      "learning_rate": 6.075710007704572e-08,
+      "loss": 0.8335,
       "step": 34040
     },
     {
-      "epoch": 0.934909780011535,
+      "epoch": 0.9659761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 2.2139366623700086e-07,
-      "loss": 0.7922,
+      "learning_rate": 6.065598288674812e-08,
+      "loss": 0.7636,
       "step": 34041
     },
     {
-      "epoch": 0.9349372442393782,
+      "epoch": 0.9660045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 2.212075672805114e-07,
-      "loss": 0.8022,
+      "learning_rate": 6.055494965453656e-08,
+      "loss": 0.8593,
       "step": 34042
     },
     {
-      "epoch": 0.9349647084672215,
+      "epoch": 0.966032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 2.2102154569761503e-07,
-      "loss": 0.8246,
+      "learning_rate": 6.04540003812637e-08,
+      "loss": 0.7529,
       "step": 34043
     },
     {
-      "epoch": 0.9349921726950646,
+      "epoch": 0.966061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 2.208356014897839e-07,
-      "loss": 0.7913,
+      "learning_rate": 6.035313506778328e-08,
+      "loss": 0.8868,
       "step": 34044
     },
     {
-      "epoch": 0.9350196369229079,
+      "epoch": 0.9660896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 2.2064973465849016e-07,
-      "loss": 0.7634,
+      "learning_rate": 6.025235371494576e-08,
+      "loss": 0.845,
       "step": 34045
     },
     {
-      "epoch": 0.9350471011507512,
+      "epoch": 0.9661180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 2.2046394520520265e-07,
-      "loss": 0.7646,
+      "learning_rate": 6.015165632360487e-08,
+      "loss": 0.7673,
       "step": 34046
     },
     {
-      "epoch": 0.9350745653785943,
+      "epoch": 0.9661464245175937,
       "grad_norm": 0.0,
-      "learning_rate": 2.202782331313935e-07,
-      "loss": 0.7473,
+      "learning_rate": 6.005104289460884e-08,
+      "loss": 0.7895,
       "step": 34047
     },
     {
-      "epoch": 0.9351020296064376,
+      "epoch": 0.9661748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 2.200925984385316e-07,
-      "loss": 0.8245,
+      "learning_rate": 5.995051342880809e-08,
+      "loss": 0.8859,
       "step": 34048
     },
     {
-      "epoch": 0.9351294938342809,
+      "epoch": 0.9662031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 2.199070411280857e-07,
-      "loss": 0.8661,
+      "learning_rate": 5.985006792705305e-08,
+      "loss": 0.8261,
       "step": 34049
     },
     {
-      "epoch": 0.9351569580621241,
+      "epoch": 0.9662315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 2.1972156120152244e-07,
-      "loss": 0.8284,
+      "learning_rate": 5.974970639019084e-08,
+      "loss": 0.8613,
       "step": 34050
     },
     {
-      "epoch": 0.9351844222899673,
+      "epoch": 0.9662599318955732,
       "grad_norm": 0.0,
-      "learning_rate": 2.1953615866031063e-07,
-      "loss": 0.8302,
+      "learning_rate": 5.964942881906966e-08,
+      "loss": 0.8,
       "step": 34051
     },
     {
-      "epoch": 0.9352118865178105,
+      "epoch": 0.9662883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 2.1935083350591911e-07,
-      "loss": 0.8877,
+      "learning_rate": 5.95492352145377e-08,
+      "loss": 0.8666,
       "step": 34052
     },
     {
-      "epoch": 0.9352393507456538,
+      "epoch": 0.966316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 2.1916558573981117e-07,
-      "loss": 0.725,
+      "learning_rate": 5.944912557743876e-08,
+      "loss": 0.8325,
       "step": 34053
     },
     {
-      "epoch": 0.9352668149734971,
+      "epoch": 0.9663450624290579,
       "grad_norm": 0.0,
-      "learning_rate": 2.189804153634545e-07,
-      "loss": 0.8011,
+      "learning_rate": 5.934909990862103e-08,
+      "loss": 0.9158,
       "step": 34054
     },
     {
-      "epoch": 0.9352942792013402,
+      "epoch": 0.9663734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 2.1879532237831347e-07,
-      "loss": 0.7902,
+      "learning_rate": 5.924915820892718e-08,
+      "loss": 0.8178,
       "step": 34055
     },
     {
-      "epoch": 0.9353217434291835,
+      "epoch": 0.9664018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 2.186103067858525e-07,
-      "loss": 0.7335,
+      "learning_rate": 5.91493004792032e-08,
+      "loss": 0.7421,
       "step": 34056
     },
     {
-      "epoch": 0.9353492076570267,
+      "epoch": 0.9664301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 2.1842536858753483e-07,
-      "loss": 0.7084,
+      "learning_rate": 5.904952672029174e-08,
+      "loss": 0.8195,
       "step": 34057
     },
     {
-      "epoch": 0.93537667188487,
+      "epoch": 0.9664585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 2.1824050778482596e-07,
-      "loss": 0.8239,
+      "learning_rate": 5.894983693303657e-08,
+      "loss": 0.7008,
       "step": 34058
     },
     {
-      "epoch": 0.9354041361127132,
+      "epoch": 0.9664869466515323,
       "grad_norm": 0.0,
-      "learning_rate": 2.1805572437918587e-07,
-      "loss": 0.7749,
+      "learning_rate": 5.8850231118278144e-08,
+      "loss": 0.8114,
       "step": 34059
     },
     {
-      "epoch": 0.9354316003405564,
+      "epoch": 0.9665153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 2.1787101837208002e-07,
-      "loss": 0.8219,
+      "learning_rate": 5.875070927685911e-08,
+      "loss": 0.8404,
       "step": 34060
     },
     {
-      "epoch": 0.9354590645683997,
+      "epoch": 0.9665437003405222,
       "grad_norm": 0.0,
-      "learning_rate": 2.1768638976496725e-07,
-      "loss": 0.8434,
+      "learning_rate": 5.86512714096199e-08,
+      "loss": 0.9474,
       "step": 34061
     },
     {
-      "epoch": 0.9354865287962429,
+      "epoch": 0.966572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.1750183855930862e-07,
-      "loss": 0.868,
+      "learning_rate": 5.8551917517399856e-08,
+      "loss": 0.8293,
       "step": 34062
     },
     {
-      "epoch": 0.9355139930240861,
+      "epoch": 0.9666004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 2.1731736475656518e-07,
-      "loss": 0.8403,
+      "learning_rate": 5.8452647601038304e-08,
+      "loss": 0.8263,
       "step": 34063
     },
     {
-      "epoch": 0.9355414572519294,
+      "epoch": 0.9666288308740069,
       "grad_norm": 0.0,
-      "learning_rate": 2.1713296835819576e-07,
-      "loss": 0.7959,
+      "learning_rate": 5.8353461661374566e-08,
+      "loss": 0.8738,
       "step": 34064
     },
     {
-      "epoch": 0.9355689214797726,
+      "epoch": 0.9666572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 2.169486493656603e-07,
-      "loss": 0.7562,
+      "learning_rate": 5.8254359699245754e-08,
+      "loss": 0.8887,
       "step": 34065
     },
     {
-      "epoch": 0.9355963857076158,
+      "epoch": 0.9666855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 2.1676440778041652e-07,
-      "loss": 0.8162,
+      "learning_rate": 5.815534171549009e-08,
+      "loss": 0.8311,
       "step": 34066
     },
     {
-      "epoch": 0.9356238499354591,
+      "epoch": 0.9667139614074914,
       "grad_norm": 0.0,
-      "learning_rate": 2.1658024360392105e-07,
-      "loss": 0.6579,
+      "learning_rate": 5.8056407710942454e-08,
+      "loss": 0.7809,
       "step": 34067
     },
     {
-      "epoch": 0.9356513141633023,
+      "epoch": 0.9667423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 2.1639615683763383e-07,
-      "loss": 0.7455,
+      "learning_rate": 5.795755768643996e-08,
+      "loss": 0.8483,
       "step": 34068
     },
     {
-      "epoch": 0.9356787783911455,
+      "epoch": 0.9667707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 2.1621214748301033e-07,
-      "loss": 0.8274,
+      "learning_rate": 5.785879164281638e-08,
+      "loss": 0.8796,
       "step": 34069
     },
     {
-      "epoch": 0.9357062426189887,
+      "epoch": 0.9667990919409761,
       "grad_norm": 0.0,
-      "learning_rate": 2.160282155415061e-07,
-      "loss": 0.8324,
+      "learning_rate": 5.776010958090661e-08,
+      "loss": 0.7798,
       "step": 34070
     },
     {
-      "epoch": 0.935733706846832,
+      "epoch": 0.9668274687854711,
       "grad_norm": 0.0,
-      "learning_rate": 2.158443610145766e-07,
-      "loss": 0.7821,
+      "learning_rate": 5.766151150154442e-08,
+      "loss": 0.7875,
       "step": 34071
     },
     {
-      "epoch": 0.9357611710746753,
+      "epoch": 0.966855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 2.1566058390367627e-07,
-      "loss": 0.839,
+      "learning_rate": 5.7562997405561373e-08,
+      "loss": 0.8191,
       "step": 34072
     },
     {
-      "epoch": 0.9357886353025184,
+      "epoch": 0.9668842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 2.1547688421025837e-07,
-      "loss": 0.8637,
+      "learning_rate": 5.746456729379124e-08,
+      "loss": 0.7103,
       "step": 34073
     },
     {
-      "epoch": 0.9358160995303617,
+      "epoch": 0.9669125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 2.1529326193577838e-07,
-      "loss": 0.7946,
+      "learning_rate": 5.736622116706558e-08,
+      "loss": 0.7432,
       "step": 34074
     },
     {
-      "epoch": 0.935843563758205,
+      "epoch": 0.9669409761634506,
       "grad_norm": 0.0,
-      "learning_rate": 2.151097170816885e-07,
-      "loss": 0.7838,
+      "learning_rate": 5.7267959026213735e-08,
+      "loss": 0.7918,
       "step": 34075
     },
     {
-      "epoch": 0.9358710279860482,
+      "epoch": 0.9669693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 2.1492624964944198e-07,
-      "loss": 0.7967,
+      "learning_rate": 5.716978087206615e-08,
+      "loss": 0.8338,
       "step": 34076
     },
     {
-      "epoch": 0.9358984922138914,
+      "epoch": 0.9669977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 2.147428596404877e-07,
-      "loss": 0.8541,
+      "learning_rate": 5.7071686705453265e-08,
+      "loss": 0.9017,
       "step": 34077
     },
     {
-      "epoch": 0.9359259564417346,
+      "epoch": 0.9670261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 2.1455954705627892e-07,
-      "loss": 0.771,
+      "learning_rate": 5.697367652720221e-08,
+      "loss": 0.7705,
       "step": 34078
     },
     {
-      "epoch": 0.9359534206695779,
+      "epoch": 0.9670544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 2.1437631189826558e-07,
-      "loss": 0.848,
+      "learning_rate": 5.687575033814119e-08,
+      "loss": 0.8134,
       "step": 34079
     },
     {
-      "epoch": 0.9359808848974212,
+      "epoch": 0.9670828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 2.1419315416789765e-07,
-      "loss": 0.8578,
+      "learning_rate": 5.6777908139098445e-08,
+      "loss": 0.8801,
       "step": 34080
     },
     {
-      "epoch": 0.9360083491252643,
+      "epoch": 0.96711123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.1401007386662288e-07,
-      "loss": 0.8458,
+      "learning_rate": 5.668014993089888e-08,
+      "loss": 0.7135,
       "step": 34081
     },
     {
-      "epoch": 0.9360358133531076,
+      "epoch": 0.9671396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 2.1382707099589228e-07,
-      "loss": 0.8297,
+      "learning_rate": 5.658247571436959e-08,
+      "loss": 0.9474,
       "step": 34082
     },
     {
-      "epoch": 0.9360632775809508,
+      "epoch": 0.9671679909194097,
       "grad_norm": 0.0,
-      "learning_rate": 2.1364414555715252e-07,
-      "loss": 0.7354,
+      "learning_rate": 5.6484885490334375e-08,
+      "loss": 0.8024,
       "step": 34083
     },
     {
-      "epoch": 0.936090741808794,
+      "epoch": 0.9671963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 2.134612975518513e-07,
-      "loss": 0.8984,
+      "learning_rate": 5.638737925961924e-08,
+      "loss": 0.8458,
       "step": 34084
     },
     {
-      "epoch": 0.9361182060366373,
+      "epoch": 0.9672247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 2.1327852698143525e-07,
-      "loss": 0.8945,
+      "learning_rate": 5.6289957023046846e-08,
+      "loss": 0.7773,
       "step": 34085
     },
     {
-      "epoch": 0.9361456702644805,
+      "epoch": 0.9672531214528944,
       "grad_norm": 0.0,
-      "learning_rate": 2.130958338473499e-07,
-      "loss": 0.735,
+      "learning_rate": 5.619261878143989e-08,
+      "loss": 0.8544,
       "step": 34086
     },
     {
-      "epoch": 0.9361731344923238,
+      "epoch": 0.9672814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 2.1291321815104182e-07,
-      "loss": 0.8321,
+      "learning_rate": 5.6095364535619925e-08,
+      "loss": 0.8482,
       "step": 34087
     },
     {
-      "epoch": 0.936200598720167,
+      "epoch": 0.9673098751418843,
       "grad_norm": 0.0,
-      "learning_rate": 2.1273067989395545e-07,
-      "loss": 0.7357,
+      "learning_rate": 5.5998194286410736e-08,
+      "loss": 0.9068,
       "step": 34088
     },
     {
-      "epoch": 0.9362280629480102,
+      "epoch": 0.9673382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 2.1254821907753632e-07,
-      "loss": 0.7912,
+      "learning_rate": 5.590110803463167e-08,
+      "loss": 0.8967,
       "step": 34089
     },
     {
-      "epoch": 0.9362555271758535,
+      "epoch": 0.967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 2.1236583570322543e-07,
-      "loss": 0.7594,
+      "learning_rate": 5.5804105781103177e-08,
+      "loss": 0.8098,
       "step": 34090
     },
     {
-      "epoch": 0.9362829914036966,
+      "epoch": 0.9673950056753688,
       "grad_norm": 0.0,
-      "learning_rate": 2.1218352977246726e-07,
-      "loss": 0.8082,
+      "learning_rate": 5.57071875266435e-08,
+      "loss": 0.773,
       "step": 34091
     },
     {
-      "epoch": 0.9363104556315399,
+      "epoch": 0.9674233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 2.1200130128670504e-07,
-      "loss": 0.7444,
+      "learning_rate": 5.5610353272073094e-08,
+      "loss": 0.8292,
       "step": 34092
     },
     {
-      "epoch": 0.9363379198593832,
+      "epoch": 0.9674517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 2.11819150247381e-07,
-      "loss": 0.8502,
+      "learning_rate": 5.551360301820907e-08,
+      "loss": 0.8845,
       "step": 34093
     },
     {
-      "epoch": 0.9363653840872264,
+      "epoch": 0.9674801362088535,
       "grad_norm": 0.0,
-      "learning_rate": 2.116370766559339e-07,
-      "loss": 0.7785,
+      "learning_rate": 5.541693676586857e-08,
+      "loss": 0.8043,
       "step": 34094
     },
     {
-      "epoch": 0.9363928483150696,
+      "epoch": 0.9675085130533485,
       "grad_norm": 0.0,
-      "learning_rate": 2.1145508051380715e-07,
-      "loss": 0.8556,
+      "learning_rate": 5.532035451586759e-08,
+      "loss": 0.9492,
       "step": 34095
     },
     {
-      "epoch": 0.9364203125429128,
+      "epoch": 0.9675368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 2.112731618224384e-07,
-      "loss": 0.7985,
+      "learning_rate": 5.522385626902327e-08,
+      "loss": 0.8255,
       "step": 34096
     },
     {
-      "epoch": 0.9364477767707561,
+      "epoch": 0.9675652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 2.1109132058326876e-07,
-      "loss": 0.7769,
+      "learning_rate": 5.512744202614939e-08,
+      "loss": 0.8207,
       "step": 34097
     },
     {
-      "epoch": 0.9364752409985994,
+      "epoch": 0.9675936435868332,
       "grad_norm": 0.0,
-      "learning_rate": 2.1090955679773595e-07,
-      "loss": 0.7305,
+      "learning_rate": 5.5031111788061975e-08,
+      "loss": 0.8454,
       "step": 34098
     },
     {
-      "epoch": 0.9365027052264425,
+      "epoch": 0.9676220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 2.1072787046727884e-07,
-      "loss": 0.8269,
+      "learning_rate": 5.49348655555737e-08,
+      "loss": 0.7376,
       "step": 34099
     },
     {
-      "epoch": 0.9365301694542858,
+      "epoch": 0.9676503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 2.1054626159333512e-07,
-      "loss": 0.8356,
+      "learning_rate": 5.483870332949615e-08,
+      "loss": 0.8083,
       "step": 34100
     },
     {
-      "epoch": 0.9365576336821291,
+      "epoch": 0.9676787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 2.1036473017734148e-07,
-      "loss": 0.8616,
+      "learning_rate": 5.474262511064421e-08,
+      "loss": 0.7814,
       "step": 34101
     },
     {
-      "epoch": 0.9365850979099722,
+      "epoch": 0.9677071509648127,
       "grad_norm": 0.0,
-      "learning_rate": 2.1018327622073453e-07,
-      "loss": 0.8661,
+      "learning_rate": 5.4646630899829465e-08,
+      "loss": 0.7719,
       "step": 34102
     },
     {
-      "epoch": 0.9366125621378155,
+      "epoch": 0.9677355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 2.1000189972494867e-07,
-      "loss": 0.8148,
+      "learning_rate": 5.455072069786016e-08,
+      "loss": 0.7903,
       "step": 34103
     },
     {
-      "epoch": 0.9366400263656587,
+      "epoch": 0.9677639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 2.0982060069142052e-07,
-      "loss": 0.8259,
+      "learning_rate": 5.4454894505548974e-08,
+      "loss": 0.8054,
       "step": 34104
     },
     {
-      "epoch": 0.936667490593502,
+      "epoch": 0.9677922814982974,
       "grad_norm": 0.0,
-      "learning_rate": 2.096393791215845e-07,
-      "loss": 0.7109,
+      "learning_rate": 5.435915232370414e-08,
+      "loss": 0.827,
       "step": 34105
     },
     {
-      "epoch": 0.9366949548213452,
+      "epoch": 0.9678206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 2.094582350168739e-07,
-      "loss": 0.8418,
+      "learning_rate": 5.426349415313503e-08,
+      "loss": 0.8256,
       "step": 34106
     },
     {
-      "epoch": 0.9367224190491884,
+      "epoch": 0.9678490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 2.0927716837872204e-07,
-      "loss": 0.7831,
+      "learning_rate": 5.416791999464766e-08,
+      "loss": 0.8451,
       "step": 34107
     },
     {
-      "epoch": 0.9367498832770317,
+      "epoch": 0.967877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 2.0909617920856217e-07,
-      "loss": 0.793,
+      "learning_rate": 5.4072429849052476e-08,
+      "loss": 0.8517,
       "step": 34108
     },
     {
-      "epoch": 0.9367773475048748,
+      "epoch": 0.967905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 2.0891526750782654e-07,
-      "loss": 0.851,
+      "learning_rate": 5.3977023717154406e-08,
+      "loss": 0.7726,
       "step": 34109
     },
     {
-      "epoch": 0.9368048117327181,
+      "epoch": 0.9679341657207718,
       "grad_norm": 0.0,
-      "learning_rate": 2.0873443327794508e-07,
-      "loss": 0.8161,
+      "learning_rate": 5.388170159975947e-08,
+      "loss": 0.8837,
       "step": 34110
     },
     {
-      "epoch": 0.9368322759605614,
+      "epoch": 0.9679625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 2.0855367652034885e-07,
-      "loss": 0.7888,
+      "learning_rate": 5.378646349767258e-08,
+      "loss": 0.7802,
       "step": 34111
     },
     {
-      "epoch": 0.9368597401884046,
+      "epoch": 0.9679909194097617,
       "grad_norm": 0.0,
-      "learning_rate": 2.0837299723647008e-07,
-      "loss": 0.766,
+      "learning_rate": 5.369130941169865e-08,
+      "loss": 0.8074,
       "step": 34112
     },
     {
-      "epoch": 0.9368872044162478,
+      "epoch": 0.9680192962542565,
       "grad_norm": 0.0,
-      "learning_rate": 2.081923954277376e-07,
-      "loss": 0.7685,
+      "learning_rate": 5.359623934264036e-08,
+      "loss": 0.7796,
       "step": 34113
     },
     {
-      "epoch": 0.9369146686440911,
+      "epoch": 0.9680476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 2.0801187109557806e-07,
-      "loss": 0.7504,
+      "learning_rate": 5.350125329130373e-08,
+      "loss": 0.8125,
       "step": 34114
     },
     {
-      "epoch": 0.9369421328719343,
+      "epoch": 0.9680760499432464,
       "grad_norm": 0.0,
-      "learning_rate": 2.0783142424142255e-07,
-      "loss": 0.8777,
+      "learning_rate": 5.340635125848814e-08,
+      "loss": 0.7898,
       "step": 34115
     },
     {
-      "epoch": 0.9369695970997776,
+      "epoch": 0.9681044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 2.0765105486669769e-07,
-      "loss": 0.8272,
+      "learning_rate": 5.3311533244996274e-08,
+      "loss": 0.8757,
       "step": 34116
     },
     {
-      "epoch": 0.9369970613276207,
+      "epoch": 0.9681328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 2.0747076297283119e-07,
-      "loss": 0.7501,
+      "learning_rate": 5.321679925162859e-08,
+      "loss": 0.805,
       "step": 34117
     },
     {
-      "epoch": 0.937024525555464,
+      "epoch": 0.9681611804767309,
       "grad_norm": 0.0,
-      "learning_rate": 2.0729054856124975e-07,
-      "loss": 0.8356,
+      "learning_rate": 5.312214927918669e-08,
+      "loss": 0.7827,
       "step": 34118
     },
     {
-      "epoch": 0.9370519897833073,
+      "epoch": 0.9681895573212259,
       "grad_norm": 0.0,
-      "learning_rate": 2.0711041163337885e-07,
-      "loss": 0.9189,
+      "learning_rate": 5.302758332846769e-08,
+      "loss": 0.7155,
       "step": 34119
     },
     {
-      "epoch": 0.9370794540111504,
+      "epoch": 0.9682179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 2.0693035219064294e-07,
-      "loss": 0.7672,
+      "learning_rate": 5.293310140027319e-08,
+      "loss": 0.8178,
       "step": 34120
     },
     {
-      "epoch": 0.9371069182389937,
+      "epoch": 0.9682463110102156,
       "grad_norm": 0.0,
-      "learning_rate": 2.0675037023446752e-07,
-      "loss": 0.7485,
+      "learning_rate": 5.2838703495400324e-08,
+      "loss": 0.8415,
       "step": 34121
     },
     {
-      "epoch": 0.9371343824668369,
+      "epoch": 0.9682746878547106,
       "grad_norm": 0.0,
-      "learning_rate": 2.0657046576627592e-07,
-      "loss": 0.7281,
+      "learning_rate": 5.274438961464512e-08,
+      "loss": 0.7311,
       "step": 34122
     },
     {
-      "epoch": 0.9371618466946802,
+      "epoch": 0.9683030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 2.0639063878749365e-07,
-      "loss": 0.855,
+      "learning_rate": 5.265015975880472e-08,
+      "loss": 0.809,
       "step": 34123
     },
     {
-      "epoch": 0.9371893109225234,
+      "epoch": 0.9683314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 2.0621088929954292e-07,
-      "loss": 0.796,
+      "learning_rate": 5.255601392867626e-08,
+      "loss": 0.7969,
       "step": 34124
     },
     {
-      "epoch": 0.9372167751503666,
+      "epoch": 0.9683598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 2.0603121730384368e-07,
-      "loss": 0.8253,
+      "learning_rate": 5.246195212505467e-08,
+      "loss": 0.7045,
       "step": 34125
     },
     {
-      "epoch": 0.9372442393782099,
+      "epoch": 0.9683881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 2.0585162280182147e-07,
-      "loss": 0.783,
+      "learning_rate": 5.236797434873375e-08,
+      "loss": 0.7759,
       "step": 34126
     },
     {
-      "epoch": 0.9372717036060532,
+      "epoch": 0.968416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 2.0567210579489293e-07,
-      "loss": 0.7507,
+      "learning_rate": 5.227408060050842e-08,
+      "loss": 0.7921,
       "step": 34127
     },
     {
-      "epoch": 0.9372991678338963,
+      "epoch": 0.9684449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 2.054926662844814e-07,
-      "loss": 0.7664,
+      "learning_rate": 5.218027088117028e-08,
+      "loss": 0.7897,
       "step": 34128
     },
     {
-      "epoch": 0.9373266320617396,
+      "epoch": 0.9684733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 2.0531330427200568e-07,
-      "loss": 0.7841,
+      "learning_rate": 5.208654519151202e-08,
+      "loss": 0.7964,
       "step": 34129
     },
     {
-      "epoch": 0.9373540962895828,
+      "epoch": 0.9685017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 2.0513401975888692e-07,
-      "loss": 0.747,
+      "learning_rate": 5.199290353232633e-08,
+      "loss": 0.8771,
       "step": 34130
     },
     {
-      "epoch": 0.937381560517426,
+      "epoch": 0.9685300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 2.0495481274654062e-07,
-      "loss": 0.7654,
+      "learning_rate": 5.189934590440482e-08,
+      "loss": 0.8663,
       "step": 34131
     },
     {
-      "epoch": 0.9374090247452693,
+      "epoch": 0.9685584562996595,
       "grad_norm": 0.0,
-      "learning_rate": 2.0477568323638564e-07,
-      "loss": 0.8487,
+      "learning_rate": 5.1805872308535734e-08,
+      "loss": 0.7556,
       "step": 34132
     },
     {
-      "epoch": 0.9374364889731125,
+      "epoch": 0.9685868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 2.0459663122984086e-07,
-      "loss": 0.8612,
+      "learning_rate": 5.171248274550955e-08,
+      "loss": 0.7751,
       "step": 34133
     },
     {
-      "epoch": 0.9374639532009558,
+      "epoch": 0.9686152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 2.044176567283207e-07,
-      "loss": 0.7374,
+      "learning_rate": 5.1619177216115646e-08,
+      "loss": 0.8193,
       "step": 34134
     },
     {
-      "epoch": 0.9374914174287989,
+      "epoch": 0.9686435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 2.0423875973324292e-07,
-      "loss": 0.8537,
+      "learning_rate": 5.1525955721142275e-08,
+      "loss": 0.707,
       "step": 34135
     },
     {
-      "epoch": 0.9375188816566422,
+      "epoch": 0.9686719636776391,
       "grad_norm": 0.0,
-      "learning_rate": 2.0405994024602306e-07,
-      "loss": 0.6819,
+      "learning_rate": 5.143281826137547e-08,
+      "loss": 0.7773,
       "step": 34136
     },
     {
-      "epoch": 0.9375463458844855,
+      "epoch": 0.9687003405221339,
       "grad_norm": 0.0,
-      "learning_rate": 2.038811982680755e-07,
-      "loss": 0.8443,
+      "learning_rate": 5.13397648376035e-08,
+      "loss": 0.8609,
       "step": 34137
     },
     {
-      "epoch": 0.9375738101123287,
+      "epoch": 0.9687287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 2.0370253380081362e-07,
-      "loss": 0.7596,
+      "learning_rate": 5.1246795450611284e-08,
+      "loss": 0.8381,
       "step": 34138
     },
     {
-      "epoch": 0.9376012743401719,
+      "epoch": 0.9687570942111238,
       "grad_norm": 0.0,
-      "learning_rate": 2.035239468456529e-07,
-      "loss": 0.6868,
+      "learning_rate": 5.115391010118487e-08,
+      "loss": 0.7869,
       "step": 34139
     },
     {
-      "epoch": 0.9376287385680152,
+      "epoch": 0.9687854710556186,
       "grad_norm": 0.0,
-      "learning_rate": 2.0334543740400448e-07,
-      "loss": 0.7641,
+      "learning_rate": 5.106110879010917e-08,
+      "loss": 0.8179,
       "step": 34140
     },
     {
-      "epoch": 0.9376562027958584,
+      "epoch": 0.9688138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 2.0316700547728385e-07,
-      "loss": 0.8868,
+      "learning_rate": 5.0968391518166905e-08,
+      "loss": 0.8285,
       "step": 34141
     },
     {
-      "epoch": 0.9376836670237017,
+      "epoch": 0.9688422247446083,
       "grad_norm": 0.0,
-      "learning_rate": 2.029886510668999e-07,
-      "loss": 0.7886,
+      "learning_rate": 5.087575828614077e-08,
+      "loss": 0.7724,
       "step": 34142
     },
     {
-      "epoch": 0.9377111312515448,
+      "epoch": 0.9688706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 2.0281037417426375e-07,
-      "loss": 0.8221,
+      "learning_rate": 5.078320909481571e-08,
+      "loss": 0.8446,
       "step": 34143
     },
     {
-      "epoch": 0.9377385954793881,
+      "epoch": 0.9688989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 2.0263217480078755e-07,
-      "loss": 0.832,
+      "learning_rate": 5.06907439449722e-08,
+      "loss": 0.7396,
       "step": 34144
     },
     {
-      "epoch": 0.9377660597072314,
+      "epoch": 0.968927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 2.0245405294788024e-07,
-      "loss": 0.8455,
+      "learning_rate": 5.0598362837390725e-08,
+      "loss": 0.8789,
       "step": 34145
     },
     {
-      "epoch": 0.9377935239350745,
+      "epoch": 0.968955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 2.0227600861695284e-07,
-      "loss": 0.7734,
+      "learning_rate": 5.0506065772851774e-08,
+      "loss": 0.8743,
       "step": 34146
     },
     {
-      "epoch": 0.9378209881629178,
+      "epoch": 0.9689841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 2.0209804180941206e-07,
-      "loss": 0.7667,
+      "learning_rate": 5.041385275213584e-08,
+      "loss": 0.8244,
       "step": 34147
     },
     {
-      "epoch": 0.9378484523907611,
+      "epoch": 0.9690124858115777,
       "grad_norm": 0.0,
-      "learning_rate": 2.0192015252666785e-07,
-      "loss": 0.8378,
+      "learning_rate": 5.032172377602229e-08,
+      "loss": 0.7456,
       "step": 34148
     },
     {
-      "epoch": 0.9378759166186043,
+      "epoch": 0.9690408626560727,
       "grad_norm": 0.0,
-      "learning_rate": 2.0174234077012577e-07,
-      "loss": 0.7873,
+      "learning_rate": 5.0229678845287176e-08,
+      "loss": 0.8181,
       "step": 34149
     },
     {
-      "epoch": 0.9379033808464475,
+      "epoch": 0.9690692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 2.015646065411947e-07,
-      "loss": 0.8238,
+      "learning_rate": 5.013771796070987e-08,
+      "loss": 0.7419,
       "step": 34150
     },
     {
-      "epoch": 0.9379308450742907,
+      "epoch": 0.9690976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 2.0138694984127905e-07,
-      "loss": 0.8184,
+      "learning_rate": 5.004584112306643e-08,
+      "loss": 0.7402,
       "step": 34151
     },
     {
-      "epoch": 0.937958309302134,
+      "epoch": 0.9691259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 2.012093706717855e-07,
-      "loss": 0.8136,
+      "learning_rate": 4.995404833313289e-08,
+      "loss": 0.8455,
       "step": 34152
     },
     {
-      "epoch": 0.9379857735299773,
+      "epoch": 0.9691543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 2.0103186903411954e-07,
-      "loss": 0.7892,
+      "learning_rate": 4.9862339591684184e-08,
+      "loss": 0.838,
       "step": 34153
     },
     {
-      "epoch": 0.9380132377578204,
+      "epoch": 0.9691827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 2.0085444492968563e-07,
-      "loss": 0.7983,
+      "learning_rate": 4.9770714899496366e-08,
+      "loss": 0.8645,
       "step": 34154
     },
     {
-      "epoch": 0.9380407019856637,
+      "epoch": 0.969211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 2.0067709835988604e-07,
-      "loss": 0.7475,
+      "learning_rate": 4.967917425734214e-08,
+      "loss": 0.7957,
       "step": 34155
     },
     {
-      "epoch": 0.9380681662135069,
+      "epoch": 0.9692395005675369,
       "grad_norm": 0.0,
-      "learning_rate": 2.0049982932612512e-07,
-      "loss": 0.8099,
+      "learning_rate": 4.958771766599535e-08,
+      "loss": 0.8548,
       "step": 34156
     },
     {
-      "epoch": 0.9380956304413501,
+      "epoch": 0.9692678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 2.003226378298073e-07,
-      "loss": 0.7768,
+      "learning_rate": 4.94963451262287e-08,
+      "loss": 0.9109,
       "step": 34157
     },
     {
-      "epoch": 0.9381230946691934,
+      "epoch": 0.9692962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 2.0014552387233156e-07,
-      "loss": 0.7382,
+      "learning_rate": 4.940505663881379e-08,
+      "loss": 0.8962,
       "step": 34158
     },
     {
-      "epoch": 0.9381505588970366,
+      "epoch": 0.9693246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 1.999684874551e-07,
-      "loss": 0.8203,
+      "learning_rate": 4.9313852204521114e-08,
+      "loss": 0.8448,
       "step": 34159
     },
     {
-      "epoch": 0.9381780231248799,
+      "epoch": 0.9693530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 1.9979152857951602e-07,
-      "loss": 0.6477,
+      "learning_rate": 4.9222731824122295e-08,
+      "loss": 0.9281,
       "step": 34160
     },
     {
-      "epoch": 0.9382054873527231,
+      "epoch": 0.9693813847900113,
       "grad_norm": 0.0,
-      "learning_rate": 1.996146472469751e-07,
-      "loss": 0.8326,
+      "learning_rate": 4.913169549838559e-08,
+      "loss": 0.7923,
       "step": 34161
     },
     {
-      "epoch": 0.9382329515805663,
+      "epoch": 0.9694097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 1.9943784345888063e-07,
-      "loss": 0.8617,
+      "learning_rate": 4.90407432280815e-08,
+      "loss": 0.7862,
       "step": 34162
     },
     {
-      "epoch": 0.9382604158084096,
+      "epoch": 0.9694381384790012,
       "grad_norm": 0.0,
-      "learning_rate": 1.9926111721663032e-07,
-      "loss": 0.8449,
+      "learning_rate": 4.894987501397719e-08,
+      "loss": 0.7938,
       "step": 34163
     },
     {
-      "epoch": 0.9382878800362527,
+      "epoch": 0.969466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.99084468521622e-07,
-      "loss": 0.8367,
+      "learning_rate": 4.885909085684093e-08,
+      "loss": 0.7919,
       "step": 34164
     },
     {
-      "epoch": 0.938315344264096,
+      "epoch": 0.9694948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 1.9890789737525452e-07,
-      "loss": 0.6826,
+      "learning_rate": 4.8768390757439886e-08,
+      "loss": 0.8685,
       "step": 34165
     },
     {
-      "epoch": 0.9383428084919393,
+      "epoch": 0.9695232690124859,
       "grad_norm": 0.0,
-      "learning_rate": 1.9873140377892564e-07,
-      "loss": 0.8402,
+      "learning_rate": 4.8677774716539007e-08,
+      "loss": 0.8542,
       "step": 34166
     },
     {
-      "epoch": 0.9383702727197825,
+      "epoch": 0.9695516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 1.985549877340287e-07,
-      "loss": 0.711,
+      "learning_rate": 4.858724273490323e-08,
+      "loss": 0.7398,
       "step": 34167
     },
     {
-      "epoch": 0.9383977369476257,
+      "epoch": 0.9695800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.9837864924196148e-07,
-      "loss": 0.8553,
+      "learning_rate": 4.8496794813299716e-08,
+      "loss": 0.7746,
       "step": 34168
     },
     {
-      "epoch": 0.9384252011754689,
+      "epoch": 0.9696083995459704,
       "grad_norm": 0.0,
-      "learning_rate": 1.982023883041184e-07,
-      "loss": 0.7526,
+      "learning_rate": 4.8406430952491204e-08,
+      "loss": 0.8001,
       "step": 34169
     },
     {
-      "epoch": 0.9384526654033122,
+      "epoch": 0.9696367763904654,
       "grad_norm": 0.0,
-      "learning_rate": 1.9802620492189618e-07,
-      "loss": 0.8207,
+      "learning_rate": 4.831615115323929e-08,
+      "loss": 0.8247,
       "step": 34170
     },
     {
-      "epoch": 0.9384801296311555,
+      "epoch": 0.9696651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 1.9785009909668696e-07,
-      "loss": 0.829,
+      "learning_rate": 4.822595541631003e-08,
+      "loss": 0.8725,
       "step": 34171
     },
     {
-      "epoch": 0.9385075938589986,
+      "epoch": 0.9696935300794551,
       "grad_norm": 0.0,
-      "learning_rate": 1.976740708298852e-07,
-      "loss": 0.8644,
+      "learning_rate": 4.813584374246283e-08,
+      "loss": 0.6947,
       "step": 34172
     },
     {
-      "epoch": 0.9385350580868419,
+      "epoch": 0.9697219069239501,
       "grad_norm": 0.0,
-      "learning_rate": 1.9749812012288428e-07,
-      "loss": 0.8146,
+      "learning_rate": 4.804581613245929e-08,
+      "loss": 0.7735,
       "step": 34173
     },
     {
-      "epoch": 0.9385625223146852,
+      "epoch": 0.969750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.9732224697707415e-07,
-      "loss": 0.743,
+      "learning_rate": 4.7955872587059916e-08,
+      "loss": 0.8173,
       "step": 34174
     },
     {
-      "epoch": 0.9385899865425283,
+      "epoch": 0.9697786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 1.9714645139384703e-07,
-      "loss": 0.7901,
+      "learning_rate": 4.786601310702521e-08,
+      "loss": 0.8843,
       "step": 34175
     },
     {
-      "epoch": 0.9386174507703716,
+      "epoch": 0.9698070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 1.9697073337459517e-07,
-      "loss": 0.8971,
+      "learning_rate": 4.777623769311346e-08,
+      "loss": 0.8409,
       "step": 34176
     },
     {
-      "epoch": 0.9386449149982148,
+      "epoch": 0.9698354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 1.9679509292070742e-07,
-      "loss": 0.746,
+      "learning_rate": 4.768654634608294e-08,
+      "loss": 0.7643,
       "step": 34177
     },
     {
-      "epoch": 0.9386723792260581,
+      "epoch": 0.9698637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 1.9661953003357603e-07,
-      "loss": 0.7791,
+      "learning_rate": 4.7596939066691936e-08,
+      "loss": 0.9181,
       "step": 34178
     },
     {
-      "epoch": 0.9386998434539013,
+      "epoch": 0.9698921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 1.9644404471458654e-07,
-      "loss": 0.8776,
+      "learning_rate": 4.7507415855697624e-08,
+      "loss": 0.8052,
       "step": 34179
     },
     {
-      "epoch": 0.9387273076817445,
+      "epoch": 0.9699205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.962686369651301e-07,
-      "loss": 0.7987,
+      "learning_rate": 4.741797671385495e-08,
+      "loss": 0.8682,
       "step": 34180
     },
     {
-      "epoch": 0.9387547719095878,
+      "epoch": 0.9699489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.9609330678659445e-07,
-      "loss": 0.8605,
+      "learning_rate": 4.732862164191887e-08,
+      "loss": 0.8443,
       "step": 34181
     },
     {
-      "epoch": 0.938782236137431,
+      "epoch": 0.9699772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 1.9591805418036514e-07,
-      "loss": 0.7265,
+      "learning_rate": 4.723935064064766e-08,
+      "loss": 0.8548,
       "step": 34182
     },
     {
-      "epoch": 0.9388097003652742,
+      "epoch": 0.9700056753688989,
       "grad_norm": 0.0,
-      "learning_rate": 1.9574287914783106e-07,
-      "loss": 0.7142,
+      "learning_rate": 4.715016371079073e-08,
+      "loss": 0.7491,
       "step": 34183
     },
     {
-      "epoch": 0.9388371645931175,
+      "epoch": 0.9700340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 1.955677816903756e-07,
-      "loss": 0.9262,
+      "learning_rate": 4.7061060853105247e-08,
+      "loss": 0.7965,
       "step": 34184
     },
     {
-      "epoch": 0.9388646288209607,
+      "epoch": 0.9700624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 1.9539276180938648e-07,
-      "loss": 0.8409,
+      "learning_rate": 4.6972042068341714e-08,
+      "loss": 0.7613,
       "step": 34185
     },
     {
-      "epoch": 0.9388920930488039,
+      "epoch": 0.9700908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 1.952178195062482e-07,
-      "loss": 0.7641,
+      "learning_rate": 4.6883107357252876e-08,
+      "loss": 0.7929,
       "step": 34186
     },
     {
-      "epoch": 0.9389195572766472,
+      "epoch": 0.9701191827468786,
       "grad_norm": 0.0,
-      "learning_rate": 1.9504295478234292e-07,
-      "loss": 0.8206,
+      "learning_rate": 4.679425672059035e-08,
+      "loss": 0.7574,
       "step": 34187
     },
     {
-      "epoch": 0.9389470215044904,
+      "epoch": 0.9701475595913734,
       "grad_norm": 0.0,
-      "learning_rate": 1.9486816763905737e-07,
-      "loss": 0.8489,
+      "learning_rate": 4.670549015910464e-08,
+      "loss": 0.7514,
       "step": 34188
     },
     {
-      "epoch": 0.9389744857323337,
+      "epoch": 0.9701759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 1.9469345807777375e-07,
-      "loss": 0.8487,
+      "learning_rate": 4.661680767354404e-08,
+      "loss": 0.7705,
       "step": 34189
     },
     {
-      "epoch": 0.9390019499601768,
+      "epoch": 0.9702043132803633,
       "grad_norm": 0.0,
-      "learning_rate": 1.9451882609987315e-07,
-      "loss": 0.7058,
+      "learning_rate": 4.6528209264657954e-08,
+      "loss": 0.7961,
       "step": 34190
     },
     {
-      "epoch": 0.9390294141880201,
+      "epoch": 0.9702326901248581,
       "grad_norm": 0.0,
-      "learning_rate": 1.9434427170673675e-07,
-      "loss": 0.7992,
+      "learning_rate": 4.643969493319689e-08,
+      "loss": 0.7296,
       "step": 34191
     },
     {
-      "epoch": 0.9390568784158634,
+      "epoch": 0.970261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.9416979489974785e-07,
-      "loss": 0.7758,
+      "learning_rate": 4.6351264679906915e-08,
+      "loss": 0.7671,
       "step": 34192
     },
     {
-      "epoch": 0.9390843426437065,
+      "epoch": 0.9702894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.9399539568028537e-07,
-      "loss": 0.6928,
+      "learning_rate": 4.626291850553521e-08,
+      "loss": 0.8024,
       "step": 34193
     },
     {
-      "epoch": 0.9391118068715498,
+      "epoch": 0.9703178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 1.9382107404972928e-07,
-      "loss": 0.7567,
+      "learning_rate": 4.617465641082786e-08,
+      "loss": 0.9087,
       "step": 34194
     },
     {
-      "epoch": 0.939139271099393,
+      "epoch": 0.9703461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 1.9364683000946073e-07,
-      "loss": 0.8837,
+      "learning_rate": 4.608647839653091e-08,
+      "loss": 0.9028,
       "step": 34195
     },
     {
-      "epoch": 0.9391667353272363,
+      "epoch": 0.9703745743473325,
       "grad_norm": 0.0,
-      "learning_rate": 1.9347266356085638e-07,
-      "loss": 0.747,
+      "learning_rate": 4.5998384463389335e-08,
+      "loss": 0.7566,
       "step": 34196
     },
     {
-      "epoch": 0.9391941995550795,
+      "epoch": 0.9704029511918275,
       "grad_norm": 0.0,
-      "learning_rate": 1.9329857470529623e-07,
-      "loss": 0.7944,
+      "learning_rate": 4.591037461214587e-08,
+      "loss": 0.8708,
       "step": 34197
     },
     {
-      "epoch": 0.9392216637829227,
+      "epoch": 0.9704313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 1.9312456344415476e-07,
-      "loss": 0.7595,
+      "learning_rate": 4.5822448843545477e-08,
+      "loss": 0.7508,
       "step": 34198
     },
     {
-      "epoch": 0.939249128010766,
+      "epoch": 0.9704597048808172,
       "grad_norm": 0.0,
-      "learning_rate": 1.9295062977881085e-07,
-      "loss": 0.8834,
+      "learning_rate": 4.573460715832978e-08,
+      "loss": 0.8865,
       "step": 34199
     },
     {
-      "epoch": 0.9392765922386093,
+      "epoch": 0.9704880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 1.9277677371064008e-07,
-      "loss": 0.7996,
+      "learning_rate": 4.5646849557242634e-08,
+      "loss": 0.7441,
       "step": 34200
     },
     {
-      "epoch": 0.9393040564664524,
+      "epoch": 0.9705164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 1.926029952410202e-07,
-      "loss": 0.7962,
+      "learning_rate": 4.5559176041022336e-08,
+      "loss": 0.7716,
       "step": 34201
     },
     {
-      "epoch": 0.9393315206942957,
+      "epoch": 0.9705448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 1.924292943713235e-07,
-      "loss": 0.7483,
+      "learning_rate": 4.547158661041273e-08,
+      "loss": 0.7413,
       "step": 34202
     },
     {
-      "epoch": 0.9393589849221389,
+      "epoch": 0.9705732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 1.9225567110292442e-07,
-      "loss": 0.8772,
+      "learning_rate": 4.5384081266151014e-08,
+      "loss": 0.6967,
       "step": 34203
     },
     {
-      "epoch": 0.9393864491499822,
+      "epoch": 0.9706015891032918,
       "grad_norm": 0.0,
-      "learning_rate": 1.920821254371985e-07,
-      "loss": 0.7765,
+      "learning_rate": 4.52966600089777e-08,
+      "loss": 0.8009,
       "step": 34204
     },
     {
-      "epoch": 0.9394139133778254,
+      "epoch": 0.9706299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.91908657375518e-07,
-      "loss": 0.8667,
+      "learning_rate": 4.5209322839631085e-08,
+      "loss": 0.8088,
       "step": 34205
     },
     {
-      "epoch": 0.9394413776056686,
+      "epoch": 0.9706583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 1.9173526691925516e-07,
-      "loss": 0.864,
+      "learning_rate": 4.5122069758849476e-08,
+      "loss": 0.8278,
       "step": 34206
     },
     {
-      "epoch": 0.9394688418335119,
+      "epoch": 0.9706867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 1.9156195406978328e-07,
-      "loss": 0.7575,
+      "learning_rate": 4.5034900767368937e-08,
+      "loss": 0.7237,
       "step": 34207
     },
     {
-      "epoch": 0.939496306061355,
+      "epoch": 0.9707150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 1.9138871882847242e-07,
-      "loss": 0.8701,
+      "learning_rate": 4.494781586592556e-08,
+      "loss": 0.8656,
       "step": 34208
     },
     {
-      "epoch": 0.9395237702891983,
+      "epoch": 0.9707434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 1.912155611966926e-07,
-      "loss": 0.84,
+      "learning_rate": 4.486081505525541e-08,
+      "loss": 0.7885,
       "step": 34209
     },
     {
-      "epoch": 0.9395512345170416,
+      "epoch": 0.970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.9104248117581492e-07,
-      "loss": 0.748,
+      "learning_rate": 4.477389833609347e-08,
+      "loss": 0.6816,
       "step": 34210
     },
     {
-      "epoch": 0.9395786987448848,
+      "epoch": 0.970800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 1.908694787672083e-07,
-      "loss": 0.8586,
+      "learning_rate": 4.4687065709174695e-08,
+      "loss": 0.8631,
       "step": 34211
     },
     {
-      "epoch": 0.939606162972728,
+      "epoch": 0.9708286038592508,
       "grad_norm": 0.0,
-      "learning_rate": 1.9069655397224384e-07,
-      "loss": 0.8137,
+      "learning_rate": 4.4600317175231834e-08,
+      "loss": 0.7935,
       "step": 34212
     },
     {
-      "epoch": 0.9396336272005713,
+      "epoch": 0.9708569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 1.9052370679228604e-07,
-      "loss": 0.8083,
+      "learning_rate": 4.451365273499653e-08,
+      "loss": 0.7816,
       "step": 34213
     },
     {
-      "epoch": 0.9396610914284145,
+      "epoch": 0.9708853575482407,
       "grad_norm": 0.0,
-      "learning_rate": 1.90350937228706e-07,
-      "loss": 0.8124,
+      "learning_rate": 4.442707238920263e-08,
+      "loss": 0.8195,
       "step": 34214
     },
     {
-      "epoch": 0.9396885556562578,
+      "epoch": 0.9709137343927355,
       "grad_norm": 0.0,
-      "learning_rate": 1.901782452828693e-07,
-      "loss": 0.7503,
+      "learning_rate": 4.434057613858067e-08,
+      "loss": 0.7133,
       "step": 34215
     },
     {
-      "epoch": 0.9397160198841009,
+      "epoch": 0.9709421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 1.9000563095614156e-07,
-      "loss": 0.8201,
+      "learning_rate": 4.425416398386229e-08,
+      "loss": 0.7493,
       "step": 34216
     },
     {
-      "epoch": 0.9397434841119442,
+      "epoch": 0.9709704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 1.8983309424988827e-07,
-      "loss": 0.8712,
+      "learning_rate": 4.416783592577578e-08,
+      "loss": 0.7227,
       "step": 34217
     },
     {
-      "epoch": 0.9397709483397875,
+      "epoch": 0.9709988649262202,
       "grad_norm": 0.0,
-      "learning_rate": 1.8966063516547728e-07,
-      "loss": 0.7834,
+      "learning_rate": 4.408159196505057e-08,
+      "loss": 0.7742,
       "step": 34218
     },
     {
-      "epoch": 0.9397984125676306,
+      "epoch": 0.9710272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.894882537042708e-07,
-      "loss": 0.7884,
+      "learning_rate": 4.399543210241608e-08,
+      "loss": 0.764,
       "step": 34219
     },
     {
-      "epoch": 0.9398258767954739,
+      "epoch": 0.97105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.8931594986763334e-07,
-      "loss": 0.8257,
+      "learning_rate": 4.3909356338599496e-08,
+      "loss": 0.7528,
       "step": 34220
     },
     {
-      "epoch": 0.9398533410233171,
+      "epoch": 0.9710839954597049,
       "grad_norm": 0.0,
-      "learning_rate": 1.8914372365692934e-07,
-      "loss": 0.7893,
+      "learning_rate": 4.3823364674328016e-08,
+      "loss": 0.8484,
       "step": 34221
     },
     {
-      "epoch": 0.9398808052511604,
+      "epoch": 0.9711123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 1.8897157507351992e-07,
-      "loss": 0.8485,
+      "learning_rate": 4.373745711032884e-08,
+      "loss": 0.7308,
       "step": 34222
     },
     {
-      "epoch": 0.9399082694790036,
+      "epoch": 0.9711407491486946,
       "grad_norm": 0.0,
-      "learning_rate": 1.8879950411876736e-07,
-      "loss": 0.7998,
+      "learning_rate": 4.365163364732694e-08,
+      "loss": 0.7972,
       "step": 34223
     },
     {
-      "epoch": 0.9399357337068468,
+      "epoch": 0.9711691259931896,
       "grad_norm": 0.0,
-      "learning_rate": 1.8862751079403384e-07,
-      "loss": 0.82,
+      "learning_rate": 4.3565894286047295e-08,
+      "loss": 0.8668,
       "step": 34224
     },
     {
-      "epoch": 0.9399631979346901,
+      "epoch": 0.9711975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 1.884555951006817e-07,
-      "loss": 0.9661,
+      "learning_rate": 4.348023902721377e-08,
+      "loss": 0.7906,
       "step": 34225
     },
     {
-      "epoch": 0.9399906621625334,
+      "epoch": 0.9712258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 1.8828375704006753e-07,
-      "loss": 0.8528,
+      "learning_rate": 4.3394667871550225e-08,
+      "loss": 0.8284,
       "step": 34226
     },
     {
-      "epoch": 0.9400181263903765,
+      "epoch": 0.9712542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 1.881119966135536e-07,
-      "loss": 0.7984,
+      "learning_rate": 4.3309180819779416e-08,
+      "loss": 0.8043,
       "step": 34227
     },
     {
-      "epoch": 0.9400455906182198,
+      "epoch": 0.9712826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 1.879403138224978e-07,
-      "loss": 0.8273,
+      "learning_rate": 4.3223777872624104e-08,
+      "loss": 0.7293,
       "step": 34228
     },
     {
-      "epoch": 0.940073054846063,
+      "epoch": 0.971311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.8776870866826002e-07,
-      "loss": 0.8031,
+      "learning_rate": 4.313845903080371e-08,
+      "loss": 0.8201,
       "step": 34229
     },
     {
-      "epoch": 0.9401005190739062,
+      "epoch": 0.9713393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 1.8759718115219594e-07,
-      "loss": 0.7273,
+      "learning_rate": 4.30532242950421e-08,
+      "loss": 0.8501,
       "step": 34230
     },
     {
-      "epoch": 0.9401279833017495,
+      "epoch": 0.9713677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 1.8742573127566555e-07,
-      "loss": 0.8234,
+      "learning_rate": 4.296807366605649e-08,
+      "loss": 0.81,
       "step": 34231
     },
     {
-      "epoch": 0.9401554475295927,
+      "epoch": 0.9713961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 1.872543590400222e-07,
-      "loss": 0.8022,
+      "learning_rate": 4.288300714456739e-08,
+      "loss": 0.7531,
       "step": 34232
     },
     {
-      "epoch": 0.940182911757436,
+      "epoch": 0.9714245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 1.8708306444662372e-07,
-      "loss": 0.8205,
+      "learning_rate": 4.279802473129313e-08,
+      "loss": 0.8423,
       "step": 34233
     },
     {
-      "epoch": 0.9402103759852791,
+      "epoch": 0.9714528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 1.8691184749682567e-07,
-      "loss": 0.7904,
+      "learning_rate": 4.271312642695202e-08,
+      "loss": 0.8292,
       "step": 34234
     },
     {
-      "epoch": 0.9402378402131224,
+      "epoch": 0.9714812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 1.867407081919814e-07,
-      "loss": 0.7927,
+      "learning_rate": 4.262831223226016e-08,
+      "loss": 0.8382,
       "step": 34235
     },
     {
-      "epoch": 0.9402653044409657,
+      "epoch": 0.9715096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 1.8656964653344766e-07,
-      "loss": 0.9548,
+      "learning_rate": 4.254358214793475e-08,
+      "loss": 0.7819,
       "step": 34236
     },
     {
-      "epoch": 0.9402927686688088,
+      "epoch": 0.9715380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 1.8639866252257554e-07,
-      "loss": 0.7832,
+      "learning_rate": 4.245893617469077e-08,
+      "loss": 0.8349,
       "step": 34237
     },
     {
-      "epoch": 0.9403202328966521,
+      "epoch": 0.9715664018161181,
       "grad_norm": 0.0,
-      "learning_rate": 1.8622775616071842e-07,
-      "loss": 0.8325,
+      "learning_rate": 4.237437431324432e-08,
+      "loss": 0.7056,
       "step": 34238
     },
     {
-      "epoch": 0.9403476971244954,
+      "epoch": 0.971594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.8605692744922854e-07,
-      "loss": 0.8265,
+      "learning_rate": 4.228989656430926e-08,
+      "loss": 0.7056,
       "step": 34239
     },
     {
-      "epoch": 0.9403751613523386,
+      "epoch": 0.9716231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 1.8588617638945816e-07,
-      "loss": 0.7516,
+      "learning_rate": 4.220550292859948e-08,
+      "loss": 0.8302,
       "step": 34240
     },
     {
-      "epoch": 0.9404026255801818,
+      "epoch": 0.9716515323496028,
       "grad_norm": 0.0,
-      "learning_rate": 1.857155029827573e-07,
-      "loss": 0.7123,
+      "learning_rate": 4.212119340682663e-08,
+      "loss": 0.7799,
       "step": 34241
     },
     {
-      "epoch": 0.940430089808025,
+      "epoch": 0.9716799091940976,
       "grad_norm": 0.0,
-      "learning_rate": 1.8554490723047824e-07,
-      "loss": 0.881,
+      "learning_rate": 4.203696799970347e-08,
+      "loss": 0.8773,
       "step": 34242
     },
     {
-      "epoch": 0.9404575540358683,
+      "epoch": 0.9717082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 1.8537438913396878e-07,
-      "loss": 0.8163,
+      "learning_rate": 4.195282670794276e-08,
+      "loss": 0.8294,
       "step": 34243
     },
     {
-      "epoch": 0.9404850182637116,
+      "epoch": 0.9717366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.8520394869457892e-07,
-      "loss": 0.9012,
+      "learning_rate": 4.186876953225283e-08,
+      "loss": 0.7473,
       "step": 34244
     },
     {
-      "epoch": 0.9405124824915547,
+      "epoch": 0.9717650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 1.8503358591365872e-07,
-      "loss": 0.7839,
+      "learning_rate": 4.178479647334644e-08,
+      "loss": 0.8187,
       "step": 34245
     },
     {
-      "epoch": 0.940539946719398,
+      "epoch": 0.9717934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 1.8486330079255267e-07,
-      "loss": 0.8302,
+      "learning_rate": 4.170090753193079e-08,
+      "loss": 0.7756,
       "step": 34246
     },
     {
-      "epoch": 0.9405674109472412,
+      "epoch": 0.971821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.8469309333261188e-07,
-      "loss": 0.8067,
+      "learning_rate": 4.1617102708715326e-08,
+      "loss": 0.7715,
       "step": 34247
     },
     {
-      "epoch": 0.9405948751750844,
+      "epoch": 0.971850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.8452296353518084e-07,
-      "loss": 0.7693,
+      "learning_rate": 4.153338200440837e-08,
+      "loss": 0.7256,
       "step": 34248
     },
     {
-      "epoch": 0.9406223394029277,
+      "epoch": 0.9718785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 1.8435291140160628e-07,
-      "loss": 0.8503,
+      "learning_rate": 4.144974541971603e-08,
+      "loss": 0.768,
       "step": 34249
     },
     {
-      "epoch": 0.9406498036307709,
+      "epoch": 0.9719069239500567,
       "grad_norm": 0.0,
-      "learning_rate": 1.8418293693323264e-07,
-      "loss": 0.812,
+      "learning_rate": 4.1366192955345495e-08,
+      "loss": 0.821,
       "step": 34250
     },
     {
-      "epoch": 0.9406772678586142,
+      "epoch": 0.9719353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 1.8401304013140663e-07,
-      "loss": 0.7817,
+      "learning_rate": 4.1282724612001777e-08,
+      "loss": 0.7279,
       "step": 34251
     },
     {
-      "epoch": 0.9407047320864574,
+      "epoch": 0.9719636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 1.8384322099747165e-07,
-      "loss": 0.7709,
+      "learning_rate": 4.1199340390390974e-08,
+      "loss": 0.7844,
       "step": 34252
     },
     {
-      "epoch": 0.9407321963143006,
+      "epoch": 0.9719920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 1.8367347953277215e-07,
-      "loss": 0.8149,
+      "learning_rate": 4.1116040291218075e-08,
+      "loss": 0.8043,
       "step": 34253
     },
     {
-      "epoch": 0.9407596605421439,
+      "epoch": 0.9720204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 1.835038157386515e-07,
-      "loss": 0.8191,
+      "learning_rate": 4.1032824315184736e-08,
+      "loss": 0.8093,
       "step": 34254
     },
     {
-      "epoch": 0.940787124769987,
+      "epoch": 0.9720488081725313,
       "grad_norm": 0.0,
-      "learning_rate": 1.8333422961644975e-07,
-      "loss": 0.7205,
+      "learning_rate": 4.094969246299485e-08,
+      "loss": 0.7822,
       "step": 34255
     },
     {
-      "epoch": 0.9408145889978303,
+      "epoch": 0.9720771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 1.8316472116751027e-07,
-      "loss": 0.8509,
+      "learning_rate": 4.086664473535007e-08,
+      "loss": 0.7578,
       "step": 34256
     },
     {
-      "epoch": 0.9408420532256736,
+      "epoch": 0.972105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.8299529039317422e-07,
-      "loss": 0.7501,
+      "learning_rate": 4.078368113295317e-08,
+      "loss": 0.7027,
       "step": 34257
     },
     {
-      "epoch": 0.9408695174535168,
+      "epoch": 0.9721339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 1.828259372947816e-07,
-      "loss": 0.7479,
+      "learning_rate": 4.070080165650359e-08,
+      "loss": 0.7774,
       "step": 34258
     },
     {
-      "epoch": 0.94089698168136,
+      "epoch": 0.9721623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 1.826566618736736e-07,
-      "loss": 0.8268,
+      "learning_rate": 4.061800630670298e-08,
+      "loss": 0.9104,
       "step": 34259
     },
     {
-      "epoch": 0.9409244459092032,
+      "epoch": 0.9721906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 1.8248746413118913e-07,
-      "loss": 0.8708,
+      "learning_rate": 4.053529508424969e-08,
+      "loss": 0.8421,
       "step": 34260
     },
     {
-      "epoch": 0.9409519101370465,
+      "epoch": 0.9722190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 1.8231834406866712e-07,
-      "loss": 0.7932,
+      "learning_rate": 4.045266798984204e-08,
+      "loss": 0.8627,
       "step": 34261
     },
     {
-      "epoch": 0.9409793743648898,
+      "epoch": 0.9722474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 1.8214930168744538e-07,
-      "loss": 0.8241,
+      "learning_rate": 4.0370125024178364e-08,
+      "loss": 0.8214,
       "step": 34262
     },
     {
-      "epoch": 0.9410068385927329,
+      "epoch": 0.9722758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 1.8198033698886064e-07,
-      "loss": 0.8159,
+      "learning_rate": 4.028766618795588e-08,
+      "loss": 0.7895,
       "step": 34263
     },
     {
-      "epoch": 0.9410343028205762,
+      "epoch": 0.9723041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 1.818114499742507e-07,
-      "loss": 0.843,
+      "learning_rate": 4.020529148187069e-08,
+      "loss": 0.8158,
       "step": 34264
     },
     {
-      "epoch": 0.9410617670484195,
+      "epoch": 0.9723325766174802,
       "grad_norm": 0.0,
-      "learning_rate": 1.816426406449512e-07,
-      "loss": 0.8092,
+      "learning_rate": 4.012300090662002e-08,
+      "loss": 0.8515,
       "step": 34265
     },
     {
-      "epoch": 0.9410892312762626,
+      "epoch": 0.972360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.814739090022999e-07,
-      "loss": 0.8008,
+      "learning_rate": 4.004079446289777e-08,
+      "loss": 0.8139,
       "step": 34266
     },
     {
-      "epoch": 0.9411166955041059,
+      "epoch": 0.9723893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 1.813052550476302e-07,
-      "loss": 0.8159,
+      "learning_rate": 3.995867215139892e-08,
+      "loss": 0.7266,
       "step": 34267
     },
     {
-      "epoch": 0.9411441597319491,
+      "epoch": 0.9724177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 1.8113667878227546e-07,
-      "loss": 0.826,
+      "learning_rate": 3.987663397281627e-08,
+      "loss": 0.808,
       "step": 34268
     },
     {
-      "epoch": 0.9411716239597924,
+      "epoch": 0.9724460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 1.8096818020757247e-07,
-      "loss": 0.8197,
+      "learning_rate": 3.9794679927843696e-08,
+      "loss": 0.766,
       "step": 34269
     },
     {
-      "epoch": 0.9411990881876356,
+      "epoch": 0.9724744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.8079975932485117e-07,
-      "loss": 0.75,
+      "learning_rate": 3.9712810017173975e-08,
+      "loss": 0.7858,
       "step": 34270
     },
     {
-      "epoch": 0.9412265524154788,
+      "epoch": 0.9725028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 1.806314161354461e-07,
-      "loss": 0.8469,
+      "learning_rate": 3.9631024241497676e-08,
+      "loss": 0.711,
       "step": 34271
     },
     {
-      "epoch": 0.9412540166433221,
+      "epoch": 0.9725312145289444,
       "grad_norm": 0.0,
-      "learning_rate": 1.8046315064068953e-07,
-      "loss": 0.8135,
+      "learning_rate": 3.954932260150535e-08,
+      "loss": 0.821,
       "step": 34272
     },
     {
-      "epoch": 0.9412814808711653,
+      "epoch": 0.9725595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 1.8029496284191262e-07,
-      "loss": 0.8033,
+      "learning_rate": 3.9467705097887556e-08,
+      "loss": 0.7693,
       "step": 34273
     },
     {
-      "epoch": 0.9413089450990085,
+      "epoch": 0.9725879682179341,
       "grad_norm": 0.0,
-      "learning_rate": 1.8012685274044428e-07,
-      "loss": 0.8075,
+      "learning_rate": 3.9386171731334855e-08,
+      "loss": 0.7266,
       "step": 34274
     },
     {
-      "epoch": 0.9413364093268518,
+      "epoch": 0.9726163450624291,
       "grad_norm": 0.0,
-      "learning_rate": 1.799588203376168e-07,
-      "loss": 0.827,
+      "learning_rate": 3.930472250253336e-08,
+      "loss": 0.8834,
       "step": 34275
     },
     {
-      "epoch": 0.941363873554695,
+      "epoch": 0.972644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.7979086563475912e-07,
-      "loss": 0.7681,
+      "learning_rate": 3.922335741217365e-08,
+      "loss": 0.7155,
       "step": 34276
     },
     {
-      "epoch": 0.9413913377825383,
+      "epoch": 0.9726730987514188,
       "grad_norm": 0.0,
-      "learning_rate": 1.7962298863320014e-07,
-      "loss": 0.8449,
+      "learning_rate": 3.9142076460941816e-08,
+      "loss": 0.8442,
       "step": 34277
     },
     {
-      "epoch": 0.9414188020103815,
+      "epoch": 0.9727014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 1.7945518933426886e-07,
-      "loss": 0.8666,
+      "learning_rate": 3.906087964952399e-08,
+      "loss": 0.9036,
       "step": 34278
     },
     {
-      "epoch": 0.9414462662382247,
+      "epoch": 0.9727298524404087,
       "grad_norm": 0.0,
-      "learning_rate": 1.7928746773929195e-07,
-      "loss": 0.8182,
+      "learning_rate": 3.8979766978608505e-08,
+      "loss": 0.7888,
       "step": 34279
     },
     {
-      "epoch": 0.941473730466068,
+      "epoch": 0.9727582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 1.7911982384959615e-07,
-      "loss": 0.9294,
+      "learning_rate": 3.8898738448877036e-08,
+      "loss": 0.9247,
       "step": 34280
     },
     {
-      "epoch": 0.9415011946939111,
+      "epoch": 0.9727866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 1.7895225766650816e-07,
-      "loss": 0.8042,
+      "learning_rate": 3.881779406101571e-08,
+      "loss": 0.716,
       "step": 34281
     },
     {
-      "epoch": 0.9415286589217544,
+      "epoch": 0.9728149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.7878476919135467e-07,
-      "loss": 0.7813,
+      "learning_rate": 3.8736933815708425e-08,
+      "loss": 0.7918,
       "step": 34282
     },
     {
-      "epoch": 0.9415561231495977,
+      "epoch": 0.9728433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.7861735842546025e-07,
-      "loss": 0.8643,
+      "learning_rate": 3.865615771363906e-08,
+      "loss": 0.9132,
       "step": 34283
     },
     {
-      "epoch": 0.9415835873774409,
+      "epoch": 0.9728717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 1.7845002537015043e-07,
-      "loss": 0.8065,
+      "learning_rate": 3.8575465755488206e-08,
+      "loss": 0.8458,
       "step": 34284
     },
     {
-      "epoch": 0.9416110516052841,
+      "epoch": 0.9729001135073779,
       "grad_norm": 0.0,
-      "learning_rate": 1.7828277002674755e-07,
-      "loss": 0.7917,
+      "learning_rate": 3.849485794193752e-08,
+      "loss": 0.8454,
       "step": 34285
     },
     {
-      "epoch": 0.9416385158331273,
+      "epoch": 0.9729284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 1.7811559239657715e-07,
-      "loss": 0.7388,
+      "learning_rate": 3.8414334273669804e-08,
+      "loss": 0.9217,
       "step": 34286
     },
     {
-      "epoch": 0.9416659800609706,
+      "epoch": 0.9729568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 1.7794849248096046e-07,
-      "loss": 0.8468,
+      "learning_rate": 3.83338947513634e-08,
+      "loss": 0.6878,
       "step": 34287
     },
     {
-      "epoch": 0.9416934442888139,
+      "epoch": 0.9729852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 1.7778147028121974e-07,
-      "loss": 0.741,
+      "learning_rate": 3.8253539375698865e-08,
+      "loss": 0.908,
       "step": 34288
     },
     {
-      "epoch": 0.941720908516657,
+      "epoch": 0.9730136208853576,
       "grad_norm": 0.0,
-      "learning_rate": 1.7761452579867612e-07,
-      "loss": 0.7743,
+      "learning_rate": 3.817326814735345e-08,
+      "loss": 0.8893,
       "step": 34289
     },
     {
-      "epoch": 0.9417483727445003,
+      "epoch": 0.9730419977298524,
       "grad_norm": 0.0,
-      "learning_rate": 1.77447659034653e-07,
-      "loss": 0.8192,
+      "learning_rate": 3.809308106700771e-08,
+      "loss": 0.8402,
       "step": 34290
     },
     {
-      "epoch": 0.9417758369723436,
+      "epoch": 0.9730703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 1.772808699904671e-07,
-      "loss": 0.7932,
+      "learning_rate": 3.801297813533555e-08,
+      "loss": 0.6819,
       "step": 34291
     },
     {
-      "epoch": 0.9418033012001867,
+      "epoch": 0.9730987514188423,
       "grad_norm": 0.0,
-      "learning_rate": 1.7711415866744076e-07,
-      "loss": 0.8097,
+      "learning_rate": 3.7932959353017554e-08,
+      "loss": 0.8689,
       "step": 34292
     },
     {
-      "epoch": 0.94183076542803,
+      "epoch": 0.9731271282633371,
       "grad_norm": 0.0,
-      "learning_rate": 1.7694752506689284e-07,
-      "loss": 0.8119,
+      "learning_rate": 3.7853024720726494e-08,
+      "loss": 0.8704,
       "step": 34293
     },
     {
-      "epoch": 0.9418582296558732,
+      "epoch": 0.973155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.767809691901401e-07,
-      "loss": 0.7809,
+      "learning_rate": 3.777317423913962e-08,
+      "loss": 0.8583,
       "step": 34294
     },
     {
-      "epoch": 0.9418856938837165,
+      "epoch": 0.9731838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.766144910385015e-07,
-      "loss": 0.7325,
+      "learning_rate": 3.769340790892973e-08,
+      "loss": 0.7805,
       "step": 34295
     },
     {
-      "epoch": 0.9419131581115597,
+      "epoch": 0.9732122587968218,
       "grad_norm": 0.0,
-      "learning_rate": 1.7644809061329482e-07,
-      "loss": 0.7675,
+      "learning_rate": 3.761372573077182e-08,
+      "loss": 0.9217,
       "step": 34296
     },
     {
-      "epoch": 0.9419406223394029,
+      "epoch": 0.9732406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 1.7628176791583573e-07,
-      "loss": 0.837,
+      "learning_rate": 3.753412770533871e-08,
+      "loss": 0.7889,
       "step": 34297
     },
     {
-      "epoch": 0.9419680865672462,
+      "epoch": 0.9732690124858115,
       "grad_norm": 0.0,
-      "learning_rate": 1.7611552294744095e-07,
-      "loss": 0.8367,
+      "learning_rate": 3.7454613833302066e-08,
+      "loss": 0.7636,
       "step": 34298
     },
     {
-      "epoch": 0.9419955507950893,
+      "epoch": 0.9732973893303065,
       "grad_norm": 0.0,
-      "learning_rate": 1.7594935570942495e-07,
-      "loss": 0.7939,
+      "learning_rate": 3.7375184115334694e-08,
+      "loss": 0.7683,
       "step": 34299
     },
     {
-      "epoch": 0.9420230150229326,
+      "epoch": 0.9733257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 1.7578326620310338e-07,
-      "loss": 0.8624,
+      "learning_rate": 3.7295838552107166e-08,
+      "loss": 0.7733,
       "step": 34300
     },
     {
-      "epoch": 0.9420504792507759,
+      "epoch": 0.9733541430192962,
       "grad_norm": 0.0,
-      "learning_rate": 1.756172544297907e-07,
-      "loss": 0.8792,
+      "learning_rate": 3.721657714428895e-08,
+      "loss": 0.806,
       "step": 34301
     },
     {
-      "epoch": 0.9420779434786191,
+      "epoch": 0.9733825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 1.7545132039080038e-07,
-      "loss": 0.8253,
+      "learning_rate": 3.713739989255061e-08,
+      "loss": 0.8155,
       "step": 34302
     },
     {
-      "epoch": 0.9421054077064623,
+      "epoch": 0.9734108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 1.752854640874435e-07,
-      "loss": 0.7197,
+      "learning_rate": 3.7058306797560504e-08,
+      "loss": 0.7427,
       "step": 34303
     },
     {
-      "epoch": 0.9421328719343056,
+      "epoch": 0.9734392735527809,
       "grad_norm": 0.0,
-      "learning_rate": 1.7511968552103465e-07,
-      "loss": 0.6834,
+      "learning_rate": 3.6979297859986995e-08,
+      "loss": 0.8068,
       "step": 34304
     },
     {
-      "epoch": 0.9421603361621488,
+      "epoch": 0.9734676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 1.7495398469288382e-07,
-      "loss": 0.8034,
+      "learning_rate": 3.690037308049732e-08,
+      "loss": 0.7269,
       "step": 34305
     },
     {
-      "epoch": 0.9421878003899921,
+      "epoch": 0.9734960272417708,
       "grad_norm": 0.0,
-      "learning_rate": 1.7478836160430335e-07,
-      "loss": 0.672,
+      "learning_rate": 3.682153245975761e-08,
+      "loss": 0.8166,
       "step": 34306
     },
     {
-      "epoch": 0.9422152646178352,
+      "epoch": 0.9735244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.7462281625660327e-07,
-      "loss": 0.724,
+      "learning_rate": 3.674277599843401e-08,
+      "loss": 0.7792,
       "step": 34307
     },
     {
-      "epoch": 0.9422427288456785,
+      "epoch": 0.9735527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.7445734865109366e-07,
-      "loss": 0.8035,
+      "learning_rate": 3.6664103697192647e-08,
+      "loss": 0.7718,
       "step": 34308
     },
     {
-      "epoch": 0.9422701930735218,
+      "epoch": 0.9735811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 1.7429195878908344e-07,
-      "loss": 0.7809,
+      "learning_rate": 3.6585515556697424e-08,
+      "loss": 0.7497,
       "step": 34309
     },
     {
-      "epoch": 0.9422976573013649,
+      "epoch": 0.9736095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 1.7412664667188162e-07,
-      "loss": 0.8975,
+      "learning_rate": 3.650701157761227e-08,
+      "loss": 0.8413,
       "step": 34310
     },
     {
-      "epoch": 0.9423251215292082,
+      "epoch": 0.9736379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 1.739614123007949e-07,
-      "loss": 0.8248,
+      "learning_rate": 3.6428591760599985e-08,
+      "loss": 0.8054,
       "step": 34311
     },
     {
-      "epoch": 0.9423525857570514,
+      "epoch": 0.97366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 1.737962556771322e-07,
-      "loss": 0.7595,
+      "learning_rate": 3.635025610632337e-08,
+      "loss": 0.7774,
       "step": 34312
     },
     {
-      "epoch": 0.9423800499848947,
+      "epoch": 0.973694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 1.7363117680220143e-07,
-      "loss": 0.7589,
+      "learning_rate": 3.6272004615443e-08,
+      "loss": 0.7762,
       "step": 34313
     },
     {
-      "epoch": 0.9424075142127379,
+      "epoch": 0.9737230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 1.7346617567730594e-07,
-      "loss": 0.8761,
+      "learning_rate": 3.619383728862169e-08,
+      "loss": 0.7717,
       "step": 34314
     },
     {
-      "epoch": 0.9424349784405811,
+      "epoch": 0.9737514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 1.733012523037514e-07,
-      "loss": 0.8608,
+      "learning_rate": 3.611575412652002e-08,
+      "loss": 0.9269,
       "step": 34315
     },
     {
-      "epoch": 0.9424624426684244,
+      "epoch": 0.9737797956867197,
       "grad_norm": 0.0,
-      "learning_rate": 1.731364066828445e-07,
-      "loss": 0.8304,
+      "learning_rate": 3.603775512979524e-08,
+      "loss": 0.792,
       "step": 34316
     },
     {
-      "epoch": 0.9424899068962677,
+      "epoch": 0.9738081725312145,
       "grad_norm": 0.0,
-      "learning_rate": 1.7297163881588975e-07,
-      "loss": 0.8315,
+      "learning_rate": 3.595984029910793e-08,
+      "loss": 0.7815,
       "step": 34317
     },
     {
-      "epoch": 0.9425173711241108,
+      "epoch": 0.9738365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 1.7280694870418946e-07,
-      "loss": 0.7745,
+      "learning_rate": 3.588200963511646e-08,
+      "loss": 0.7892,
       "step": 34318
     },
     {
-      "epoch": 0.9425448353519541,
+      "epoch": 0.9738649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 1.7264233634904814e-07,
-      "loss": 0.7056,
+      "learning_rate": 3.5804263138475847e-08,
+      "loss": 0.7672,
       "step": 34319
     },
     {
-      "epoch": 0.9425722995797973,
+      "epoch": 0.9738933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 1.7247780175176698e-07,
-      "loss": 0.7739,
+      "learning_rate": 3.572660080984558e-08,
+      "loss": 0.805,
       "step": 34320
     },
     {
-      "epoch": 0.9425997638076405,
+      "epoch": 0.9739216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 1.7231334491364716e-07,
-      "loss": 0.8531,
+      "learning_rate": 3.564902264988068e-08,
+      "loss": 0.7714,
       "step": 34321
     },
     {
-      "epoch": 0.9426272280354838,
+      "epoch": 0.973950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 1.7214896583599205e-07,
-      "loss": 0.8202,
+      "learning_rate": 3.5571528659236185e-08,
+      "loss": 0.7774,
       "step": 34322
     },
     {
-      "epoch": 0.942654692263327,
+      "epoch": 0.9739784335981839,
       "grad_norm": 0.0,
-      "learning_rate": 1.7198466452010066e-07,
-      "loss": 0.7438,
+      "learning_rate": 3.549411883856713e-08,
+      "loss": 0.8128,
       "step": 34323
     },
     {
-      "epoch": 0.9426821564911703,
+      "epoch": 0.9740068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 1.7182044096727303e-07,
-      "loss": 0.8029,
+      "learning_rate": 3.541679318852742e-08,
+      "loss": 0.8017,
       "step": 34324
     },
     {
-      "epoch": 0.9427096207190134,
+      "epoch": 0.9740351872871736,
       "grad_norm": 0.0,
-      "learning_rate": 1.7165629517881143e-07,
-      "loss": 0.8464,
+      "learning_rate": 3.533955170976988e-08,
+      "loss": 0.7116,
       "step": 34325
     },
     {
-      "epoch": 0.9427370849468567,
+      "epoch": 0.9740635641316686,
       "grad_norm": 0.0,
-      "learning_rate": 1.7149222715601155e-07,
-      "loss": 0.8421,
+      "learning_rate": 3.526239440294621e-08,
+      "loss": 0.7435,
       "step": 34326
     },
     {
-      "epoch": 0.9427645491747,
+      "epoch": 0.9740919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 1.713282369001712e-07,
-      "loss": 0.7676,
+      "learning_rate": 3.5185321268710324e-08,
+      "loss": 0.8322,
       "step": 34327
     },
     {
-      "epoch": 0.9427920134025431,
+      "epoch": 0.9741203178206583,
       "grad_norm": 0.0,
-      "learning_rate": 1.7116432441258935e-07,
-      "loss": 0.8586,
+      "learning_rate": 3.5108332307710605e-08,
+      "loss": 0.7969,
       "step": 34328
     },
     {
-      "epoch": 0.9428194776303864,
+      "epoch": 0.9741486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 1.7100048969456274e-07,
-      "loss": 0.8535,
+      "learning_rate": 3.503142752059985e-08,
+      "loss": 0.8282,
       "step": 34329
     },
     {
-      "epoch": 0.9428469418582297,
+      "epoch": 0.9741770715096482,
       "grad_norm": 0.0,
-      "learning_rate": 1.7083673274738811e-07,
-      "loss": 0.7371,
+      "learning_rate": 3.495460690802643e-08,
+      "loss": 0.8279,
       "step": 34330
     },
     {
-      "epoch": 0.9428744060860729,
+      "epoch": 0.974205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 1.706730535723611e-07,
-      "loss": 0.8086,
+      "learning_rate": 3.487787047063873e-08,
+      "loss": 0.8055,
       "step": 34331
     },
     {
-      "epoch": 0.9429018703139161,
+      "epoch": 0.9742338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 1.705094521707762e-07,
-      "loss": 0.8173,
+      "learning_rate": 3.480121820908622e-08,
+      "loss": 0.727,
       "step": 34332
     },
     {
-      "epoch": 0.9429293345417593,
+      "epoch": 0.9742622020431329,
       "grad_norm": 0.0,
-      "learning_rate": 1.7034592854392796e-07,
-      "loss": 0.7688,
+      "learning_rate": 3.4724650124016154e-08,
+      "loss": 0.8191,
       "step": 34333
     },
     {
-      "epoch": 0.9429567987696026,
+      "epoch": 0.9742905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 1.7018248269310979e-07,
-      "loss": 0.9036,
+      "learning_rate": 3.46481662160747e-08,
+      "loss": 0.7554,
       "step": 34334
     },
     {
-      "epoch": 0.9429842629974459,
+      "epoch": 0.9743189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 1.7001911461961617e-07,
-      "loss": 0.8177,
+      "learning_rate": 3.4571766485907986e-08,
+      "loss": 0.8233,
       "step": 34335
     },
     {
-      "epoch": 0.943011727225289,
+      "epoch": 0.9743473325766174,
       "grad_norm": 0.0,
-      "learning_rate": 1.6985582432473836e-07,
-      "loss": 0.8834,
+      "learning_rate": 3.449545093416107e-08,
+      "loss": 0.7627,
       "step": 34336
     },
     {
-      "epoch": 0.9430391914531323,
+      "epoch": 0.9743757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 1.696926118097697e-07,
-      "loss": 0.8156,
+      "learning_rate": 3.441921956148009e-08,
+      "loss": 0.7977,
       "step": 34337
     },
     {
-      "epoch": 0.9430666556809755,
+      "epoch": 0.9744040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 1.6952947707600143e-07,
-      "loss": 0.8261,
+      "learning_rate": 3.434307236850676e-08,
+      "loss": 0.8358,
       "step": 34338
     },
     {
-      "epoch": 0.9430941199088188,
+      "epoch": 0.9744324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 1.6936642012472248e-07,
-      "loss": 0.7592,
+      "learning_rate": 3.426700935588612e-08,
+      "loss": 0.7777,
       "step": 34339
     },
     {
-      "epoch": 0.943121584136662,
+      "epoch": 0.9744608399545971,
       "grad_norm": 0.0,
-      "learning_rate": 1.6920344095722518e-07,
-      "loss": 0.8402,
+      "learning_rate": 3.4191030524259874e-08,
+      "loss": 0.8683,
       "step": 34340
     },
     {
-      "epoch": 0.9431490483645052,
+      "epoch": 0.974489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 1.690405395747996e-07,
-      "loss": 0.7993,
+      "learning_rate": 3.4115135874270845e-08,
+      "loss": 0.8567,
       "step": 34341
     },
     {
-      "epoch": 0.9431765125923485,
+      "epoch": 0.9745175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 1.6887771597873136e-07,
-      "loss": 0.8185,
+      "learning_rate": 3.403932540655852e-08,
+      "loss": 0.8204,
       "step": 34342
     },
     {
-      "epoch": 0.9432039768201917,
+      "epoch": 0.9745459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 1.687149701703128e-07,
-      "loss": 0.8553,
+      "learning_rate": 3.39635991217635e-08,
+      "loss": 0.6762,
       "step": 34343
     },
     {
-      "epoch": 0.9432314410480349,
+      "epoch": 0.9745743473325766,
       "grad_norm": 0.0,
-      "learning_rate": 1.685523021508284e-07,
-      "loss": 0.7194,
+      "learning_rate": 3.388795702052638e-08,
+      "loss": 0.8548,
       "step": 34344
     },
     {
-      "epoch": 0.9432589052758782,
+      "epoch": 0.9746027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 1.6838971192156605e-07,
-      "loss": 0.817,
+      "learning_rate": 3.381239910348555e-08,
+      "loss": 0.7434,
       "step": 34345
     },
     {
-      "epoch": 0.9432863695037214,
+      "epoch": 0.9746311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 1.6822719948381362e-07,
-      "loss": 0.8292,
+      "learning_rate": 3.3736925371279375e-08,
+      "loss": 0.7881,
       "step": 34346
     },
     {
-      "epoch": 0.9433138337315646,
+      "epoch": 0.9746594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 1.680647648388556e-07,
-      "loss": 0.7339,
+      "learning_rate": 3.366153582454623e-08,
+      "loss": 0.7861,
       "step": 34347
     },
     {
-      "epoch": 0.9433412979594079,
+      "epoch": 0.9746878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 1.6790240798797874e-07,
-      "loss": 0.694,
+      "learning_rate": 3.358623046392118e-08,
+      "loss": 0.7915,
       "step": 34348
     },
     {
-      "epoch": 0.9433687621872511,
+      "epoch": 0.974716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.6774012893246538e-07,
-      "loss": 0.7436,
+      "learning_rate": 3.3511009290042585e-08,
+      "loss": 0.7864,
       "step": 34349
     },
     {
-      "epoch": 0.9433962264150944,
+      "epoch": 0.974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 1.6757792767360226e-07,
-      "loss": 0.7248,
+      "learning_rate": 3.3435872303543284e-08,
+      "loss": 0.8337,
       "step": 34350
     },
     {
-      "epoch": 0.9434236906429376,
+      "epoch": 0.9747729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 1.6741580421266946e-07,
-      "loss": 0.7585,
+      "learning_rate": 3.3360819505059424e-08,
+      "loss": 0.8161,
       "step": 34351
     },
     {
-      "epoch": 0.9434511548707808,
+      "epoch": 0.9748013620885357,
       "grad_norm": 0.0,
-      "learning_rate": 1.672537585509526e-07,
-      "loss": 0.8528,
+      "learning_rate": 3.3285850895224956e-08,
+      "loss": 0.8548,
       "step": 34352
     },
     {
-      "epoch": 0.9434786190986241,
+      "epoch": 0.9748297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 1.670917906897318e-07,
-      "loss": 0.7675,
+      "learning_rate": 3.32109664746727e-08,
+      "loss": 0.7683,
       "step": 34353
     },
     {
-      "epoch": 0.9435060833264672,
+      "epoch": 0.9748581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 1.6692990063029046e-07,
-      "loss": 0.8141,
+      "learning_rate": 3.3136166244035486e-08,
+      "loss": 0.7524,
       "step": 34354
     },
     {
-      "epoch": 0.9435335475543105,
+      "epoch": 0.9748864926220204,
       "grad_norm": 0.0,
-      "learning_rate": 1.6676808837390978e-07,
-      "loss": 0.8458,
+      "learning_rate": 3.3061450203945025e-08,
+      "loss": 0.8143,
       "step": 34355
     },
     {
-      "epoch": 0.9435610117821538,
+      "epoch": 0.9749148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 1.6660635392186763e-07,
-      "loss": 0.8298,
+      "learning_rate": 3.298681835503304e-08,
+      "loss": 0.8228,
       "step": 34356
     },
     {
-      "epoch": 0.943588476009997,
+      "epoch": 0.9749432463110103,
       "grad_norm": 0.0,
-      "learning_rate": 1.6644469727544632e-07,
-      "loss": 0.8439,
+      "learning_rate": 3.2912270697929013e-08,
+      "loss": 0.8817,
       "step": 34357
     },
     {
-      "epoch": 0.9436159402378402,
+      "epoch": 0.9749716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 1.662831184359226e-07,
-      "loss": 0.7528,
+      "learning_rate": 3.283780723326358e-08,
+      "loss": 0.7329,
       "step": 34358
     },
     {
-      "epoch": 0.9436434044656834,
+      "epoch": 0.975,
       "grad_norm": 0.0,
-      "learning_rate": 1.6612161740457544e-07,
-      "loss": 0.7974,
+      "learning_rate": 3.276342796166399e-08,
+      "loss": 0.7585,
       "step": 34359
     },
     {
-      "epoch": 0.9436708686935267,
+      "epoch": 0.9750283768444948,
       "grad_norm": 0.0,
-      "learning_rate": 1.6596019418268383e-07,
-      "loss": 0.7912,
+      "learning_rate": 3.2689132883760856e-08,
+      "loss": 0.9182,
       "step": 34360
     },
     {
-      "epoch": 0.94369833292137,
+      "epoch": 0.9750567536889898,
       "grad_norm": 0.0,
-      "learning_rate": 1.6579884877152564e-07,
-      "loss": 0.7566,
+      "learning_rate": 3.261492200018035e-08,
+      "loss": 0.7124,
       "step": 34361
     },
     {
-      "epoch": 0.9437257971492131,
+      "epoch": 0.9750851305334847,
       "grad_norm": 0.0,
-      "learning_rate": 1.656375811723754e-07,
-      "loss": 0.8022,
+      "learning_rate": 3.254079531154975e-08,
+      "loss": 0.8807,
       "step": 34362
     },
     {
-      "epoch": 0.9437532613770564,
+      "epoch": 0.9751135073779795,
       "grad_norm": 0.0,
-      "learning_rate": 1.6547639138650985e-07,
-      "loss": 0.7998,
+      "learning_rate": 3.2466752818495205e-08,
+      "loss": 0.8649,
       "step": 34363
     },
     {
-      "epoch": 0.9437807256048997,
+      "epoch": 0.9751418842224745,
       "grad_norm": 0.0,
-      "learning_rate": 1.6531527941520465e-07,
-      "loss": 0.7773,
+      "learning_rate": 3.23927945216429e-08,
+      "loss": 0.776,
       "step": 34364
     },
     {
-      "epoch": 0.9438081898327428,
+      "epoch": 0.9751702610669694,
       "grad_norm": 0.0,
-      "learning_rate": 1.6515424525973544e-07,
-      "loss": 0.7567,
+      "learning_rate": 3.231892042161566e-08,
+      "loss": 0.7925,
       "step": 34365
     },
     {
-      "epoch": 0.9438356540605861,
+      "epoch": 0.9751986379114642,
       "grad_norm": 0.0,
-      "learning_rate": 1.6499328892137567e-07,
-      "loss": 0.8474,
+      "learning_rate": 3.2245130519038547e-08,
+      "loss": 0.8355,
       "step": 34366
     },
     {
-      "epoch": 0.9438631182884293,
+      "epoch": 0.9752270147559592,
       "grad_norm": 0.0,
-      "learning_rate": 1.6483241040139876e-07,
-      "loss": 0.7561,
+      "learning_rate": 3.21714248145355e-08,
+      "loss": 0.7988,
       "step": 34367
     },
     {
-      "epoch": 0.9438905825162726,
+      "epoch": 0.975255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.6467160970107587e-07,
-      "loss": 0.7519,
+      "learning_rate": 3.209780330872714e-08,
+      "loss": 0.9551,
       "step": 34368
     },
     {
-      "epoch": 0.9439180467441158,
+      "epoch": 0.9752837684449489,
       "grad_norm": 0.0,
-      "learning_rate": 1.645108868216816e-07,
-      "loss": 0.8674,
+      "learning_rate": 3.202426600223851e-08,
+      "loss": 0.7473,
       "step": 34369
     },
     {
-      "epoch": 0.943945510971959,
+      "epoch": 0.9753121452894438,
       "grad_norm": 0.0,
-      "learning_rate": 1.6435024176448822e-07,
-      "loss": 0.8898,
+      "learning_rate": 3.195081289568802e-08,
+      "loss": 0.8631,
       "step": 34370
     },
     {
-      "epoch": 0.9439729751998023,
+      "epoch": 0.9753405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.6418967453076474e-07,
-      "loss": 0.8117,
+      "learning_rate": 3.187744398969739e-08,
+      "loss": 0.8624,
       "step": 34371
     },
     {
-      "epoch": 0.9440004394276454,
+      "epoch": 0.9753688989784336,
       "grad_norm": 0.0,
-      "learning_rate": 1.6402918512178346e-07,
-      "loss": 0.7417,
+      "learning_rate": 3.180415928488612e-08,
+      "loss": 0.8519,
       "step": 34372
     },
     {
-      "epoch": 0.9440279036554887,
+      "epoch": 0.9753972758229285,
       "grad_norm": 0.0,
-      "learning_rate": 1.6386877353881226e-07,
-      "loss": 0.7734,
+      "learning_rate": 3.17309587818726e-08,
+      "loss": 0.8909,
       "step": 34373
     },
     {
-      "epoch": 0.944055367883332,
+      "epoch": 0.9754256526674234,
       "grad_norm": 0.0,
-      "learning_rate": 1.6370843978312233e-07,
-      "loss": 0.8689,
+      "learning_rate": 3.1657842481276346e-08,
+      "loss": 0.8277,
       "step": 34374
     },
     {
-      "epoch": 0.9440828321111752,
+      "epoch": 0.9754540295119183,
       "grad_norm": 0.0,
-      "learning_rate": 1.635481838559816e-07,
-      "loss": 0.7486,
+      "learning_rate": 3.1584810383714634e-08,
+      "loss": 0.7173,
       "step": 34375
     },
     {
-      "epoch": 0.9441102963390184,
+      "epoch": 0.9754824063564131,
       "grad_norm": 0.0,
-      "learning_rate": 1.6338800575865676e-07,
-      "loss": 0.7835,
+      "learning_rate": 3.1511862489803644e-08,
+      "loss": 0.7509,
       "step": 34376
     },
     {
-      "epoch": 0.9441377605668617,
+      "epoch": 0.975510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.6322790549241684e-07,
-      "loss": 0.7919,
+      "learning_rate": 3.143899880015955e-08,
+      "loss": 0.8255,
       "step": 34377
     },
     {
-      "epoch": 0.9441652247947049,
+      "epoch": 0.975539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.6306788305852861e-07,
-      "loss": 0.833,
+      "learning_rate": 3.1366219315399625e-08,
+      "loss": 0.861,
       "step": 34378
     },
     {
-      "epoch": 0.9441926890225482,
+      "epoch": 0.9755675368898978,
       "grad_norm": 0.0,
-      "learning_rate": 1.629079384582577e-07,
-      "loss": 0.914,
+      "learning_rate": 3.1293524036136724e-08,
+      "loss": 0.8392,
       "step": 34379
     },
     {
-      "epoch": 0.9442201532503913,
+      "epoch": 0.9755959137343927,
       "grad_norm": 0.0,
-      "learning_rate": 1.6274807169286865e-07,
-      "loss": 0.7588,
+      "learning_rate": 3.1220912962985905e-08,
+      "loss": 0.798,
       "step": 34380
     },
     {
-      "epoch": 0.9442476174782346,
+      "epoch": 0.9756242905788877,
       "grad_norm": 0.0,
-      "learning_rate": 1.6258828276362936e-07,
-      "loss": 0.7596,
+      "learning_rate": 3.1148386096560015e-08,
+      "loss": 0.6946,
       "step": 34381
     },
     {
-      "epoch": 0.9442750817060779,
+      "epoch": 0.9756526674233825,
       "grad_norm": 0.0,
-      "learning_rate": 1.6242857167180103e-07,
-      "loss": 0.7839,
+      "learning_rate": 3.107594343747189e-08,
+      "loss": 0.8101,
       "step": 34382
     },
     {
-      "epoch": 0.944302545933921,
+      "epoch": 0.9756810442678774,
       "grad_norm": 0.0,
-      "learning_rate": 1.6226893841864823e-07,
-      "loss": 0.7356,
+      "learning_rate": 3.100358498633327e-08,
+      "loss": 0.791,
       "step": 34383
     },
     {
-      "epoch": 0.9443300101617643,
+      "epoch": 0.9757094211123724,
       "grad_norm": 0.0,
-      "learning_rate": 1.6210938300543544e-07,
-      "loss": 0.8171,
+      "learning_rate": 3.093131074375477e-08,
+      "loss": 0.8989,
       "step": 34384
     },
     {
-      "epoch": 0.9443574743896075,
+      "epoch": 0.9757377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.619499054334228e-07,
-      "loss": 0.7172,
+      "learning_rate": 3.085912071034813e-08,
+      "loss": 0.8335,
       "step": 34385
     },
     {
-      "epoch": 0.9443849386174508,
+      "epoch": 0.9757661748013621,
       "grad_norm": 0.0,
-      "learning_rate": 1.617905057038738e-07,
-      "loss": 0.8047,
+      "learning_rate": 3.078701488672287e-08,
+      "loss": 0.7396,
       "step": 34386
     },
     {
-      "epoch": 0.944412402845294,
+      "epoch": 0.9757945516458569,
       "grad_norm": 0.0,
-      "learning_rate": 1.6163118381804955e-07,
-      "loss": 0.7494,
+      "learning_rate": 3.071499327348848e-08,
+      "loss": 0.8053,
       "step": 34387
     },
     {
-      "epoch": 0.9444398670731372,
+      "epoch": 0.9758229284903519,
       "grad_norm": 0.0,
-      "learning_rate": 1.6147193977721022e-07,
-      "loss": 0.7327,
+      "learning_rate": 3.064305587125227e-08,
+      "loss": 0.7455,
       "step": 34388
     },
     {
-      "epoch": 0.9444673313009805,
+      "epoch": 0.9758513053348468,
       "grad_norm": 0.0,
-      "learning_rate": 1.6131277358261587e-07,
-      "loss": 0.8354,
+      "learning_rate": 3.057120268062153e-08,
+      "loss": 0.779,
       "step": 34389
     },
     {
-      "epoch": 0.9444947955288238,
+      "epoch": 0.9758796821793416,
       "grad_norm": 0.0,
-      "learning_rate": 1.6115368523552665e-07,
-      "loss": 0.7806,
+      "learning_rate": 3.049943370220465e-08,
+      "loss": 0.7872,
       "step": 34390
     },
     {
-      "epoch": 0.9445222597566669,
+      "epoch": 0.9759080590238366,
       "grad_norm": 0.0,
-      "learning_rate": 1.6099467473720044e-07,
-      "loss": 0.7663,
+      "learning_rate": 3.042774893660671e-08,
+      "loss": 0.7243,
       "step": 34391
     },
     {
-      "epoch": 0.9445497239845102,
+      "epoch": 0.9759364358683315,
       "grad_norm": 0.0,
-      "learning_rate": 1.6083574208889506e-07,
-      "loss": 0.7899,
+      "learning_rate": 3.0356148384435014e-08,
+      "loss": 0.7816,
       "step": 34392
     },
     {
-      "epoch": 0.9445771882123534,
+      "epoch": 0.9759648127128263,
       "grad_norm": 0.0,
-      "learning_rate": 1.6067688729186849e-07,
-      "loss": 0.7066,
+      "learning_rate": 3.028463204629128e-08,
+      "loss": 0.8369,
       "step": 34393
     },
     {
-      "epoch": 0.9446046524401966,
+      "epoch": 0.9759931895573212,
       "grad_norm": 0.0,
-      "learning_rate": 1.6051811034737852e-07,
-      "loss": 0.7892,
+      "learning_rate": 3.021319992278282e-08,
+      "loss": 0.7908,
       "step": 34394
     },
     {
-      "epoch": 0.9446321166680399,
+      "epoch": 0.9760215664018161,
       "grad_norm": 0.0,
-      "learning_rate": 1.6035941125667975e-07,
-      "loss": 0.7888,
+      "learning_rate": 3.014185201451136e-08,
+      "loss": 0.8074,
       "step": 34395
     },
     {
-      "epoch": 0.9446595808958831,
+      "epoch": 0.976049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.6020079002103005e-07,
-      "loss": 0.7915,
+      "learning_rate": 3.0070588322079765e-08,
+      "loss": 0.8333,
       "step": 34396
     },
     {
-      "epoch": 0.9446870451237264,
+      "epoch": 0.9760783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.6004224664168288e-07,
-      "loss": 0.8197,
+      "learning_rate": 2.9999408846089763e-08,
+      "loss": 0.8864,
       "step": 34397
     },
     {
-      "epoch": 0.9447145093515695,
+      "epoch": 0.9761066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.598837811198939e-07,
-      "loss": 0.7565,
+      "learning_rate": 2.99283135871431e-08,
+      "loss": 0.8532,
       "step": 34398
     },
     {
-      "epoch": 0.9447419735794128,
+      "epoch": 0.9761350737797957,
       "grad_norm": 0.0,
-      "learning_rate": 1.597253934569143e-07,
-      "loss": 0.8337,
+      "learning_rate": 2.9857302545840403e-08,
+      "loss": 0.7575,
       "step": 34399
     },
     {
-      "epoch": 0.9447694378072561,
+      "epoch": 0.9761634506242906,
       "grad_norm": 0.0,
-      "learning_rate": 1.5956708365399975e-07,
-      "loss": 0.7236,
+      "learning_rate": 2.97863757227812e-08,
+      "loss": 0.6773,
       "step": 34400
     },
     {
-      "epoch": 0.9447969020350993,
+      "epoch": 0.9761918274687855,
       "grad_norm": 0.0,
-      "learning_rate": 1.594088517124026e-07,
-      "loss": 0.7424,
+      "learning_rate": 2.971553311856501e-08,
+      "loss": 0.7933,
       "step": 34401
     },
     {
-      "epoch": 0.9448243662629425,
+      "epoch": 0.9762202043132804,
       "grad_norm": 0.0,
-      "learning_rate": 1.592506976333752e-07,
-      "loss": 0.892,
+      "learning_rate": 2.9644774733789128e-08,
+      "loss": 0.775,
       "step": 34402
     },
     {
-      "epoch": 0.9448518304907858,
+      "epoch": 0.9762485811577752,
       "grad_norm": 0.0,
-      "learning_rate": 1.590926214181676e-07,
-      "loss": 0.8044,
+      "learning_rate": 2.9574100569051967e-08,
+      "loss": 0.7714,
       "step": 34403
     },
     {
-      "epoch": 0.944879294718629,
+      "epoch": 0.9762769580022701,
       "grad_norm": 0.0,
-      "learning_rate": 1.5893462306802998e-07,
-      "loss": 0.8841,
+      "learning_rate": 2.9503510624950825e-08,
+      "loss": 0.8212,
       "step": 34404
     },
     {
-      "epoch": 0.9449067589464722,
+      "epoch": 0.9763053348467651,
       "grad_norm": 0.0,
-      "learning_rate": 1.5877670258421575e-07,
-      "loss": 0.725,
+      "learning_rate": 2.9433004902081897e-08,
+      "loss": 0.7227,
       "step": 34405
     },
     {
-      "epoch": 0.9449342231743154,
+      "epoch": 0.9763337116912599,
       "grad_norm": 0.0,
-      "learning_rate": 1.5861885996797167e-07,
-      "loss": 0.7982,
+      "learning_rate": 2.9362583401041367e-08,
+      "loss": 0.7456,
       "step": 34406
     },
     {
-      "epoch": 0.9449616874021587,
+      "epoch": 0.9763620885357548,
       "grad_norm": 0.0,
-      "learning_rate": 1.5846109522054677e-07,
-      "loss": 0.7064,
+      "learning_rate": 2.9292246122422096e-08,
+      "loss": 0.7368,
       "step": 34407
     },
     {
-      "epoch": 0.944989151630002,
+      "epoch": 0.9763904653802498,
       "grad_norm": 0.0,
-      "learning_rate": 1.5830340834319002e-07,
-      "loss": 0.8338,
+      "learning_rate": 2.9221993066819166e-08,
+      "loss": 0.8409,
       "step": 34408
     },
     {
-      "epoch": 0.9450166158578451,
+      "epoch": 0.9764188422247446,
       "grad_norm": 0.0,
-      "learning_rate": 1.5814579933714935e-07,
-      "loss": 0.7366,
+      "learning_rate": 2.915182423482765e-08,
+      "loss": 0.8012,
       "step": 34409
     },
     {
-      "epoch": 0.9450440800856884,
+      "epoch": 0.9764472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 1.5798826820367152e-07,
-      "loss": 0.7559,
+      "learning_rate": 2.9081739627038197e-08,
+      "loss": 0.7487,
       "step": 34410
     },
     {
-      "epoch": 0.9450715443135316,
+      "epoch": 0.9764755959137343,
       "grad_norm": 0.0,
-      "learning_rate": 1.578308149440022e-07,
-      "loss": 0.7886,
+      "learning_rate": 2.9011739244043657e-08,
+      "loss": 0.7617,
       "step": 34411
     },
     {
-      "epoch": 0.9450990085413749,
+      "epoch": 0.9765039727582293,
       "grad_norm": 0.0,
-      "learning_rate": 1.576734395593882e-07,
-      "loss": 0.9033,
+      "learning_rate": 2.894182308643467e-08,
+      "loss": 0.825,
       "step": 34412
     },
     {
-      "epoch": 0.9451264727692181,
+      "epoch": 0.9765323496027242,
       "grad_norm": 0.0,
-      "learning_rate": 1.5751614205107514e-07,
-      "loss": 0.8016,
+      "learning_rate": 2.8871991154802993e-08,
+      "loss": 0.8644,
       "step": 34413
     },
     {
-      "epoch": 0.9451539369970613,
+      "epoch": 0.976560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.573589224203076e-07,
-      "loss": 0.7937,
+      "learning_rate": 2.8802243449737034e-08,
+      "loss": 0.8084,
       "step": 34414
     },
     {
-      "epoch": 0.9451814012249046,
+      "epoch": 0.976589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.5720178066832792e-07,
-      "loss": 0.8437,
+      "learning_rate": 2.8732579971827434e-08,
+      "loss": 0.867,
       "step": 34415
     },
     {
-      "epoch": 0.9452088654527478,
+      "epoch": 0.9766174801362089,
       "grad_norm": 0.0,
-      "learning_rate": 1.5704471679638066e-07,
-      "loss": 0.8076,
+      "learning_rate": 2.8663000721662615e-08,
+      "loss": 0.8447,
       "step": 34416
     },
     {
-      "epoch": 0.945236329680591,
+      "epoch": 0.9766458569807037,
       "grad_norm": 0.0,
-      "learning_rate": 1.568877308057082e-07,
-      "loss": 0.8116,
+      "learning_rate": 2.859350569982877e-08,
+      "loss": 0.8022,
       "step": 34417
     },
     {
-      "epoch": 0.9452637939084343,
+      "epoch": 0.9766742338251987,
       "grad_norm": 0.0,
-      "learning_rate": 1.567308226975517e-07,
-      "loss": 0.7386,
+      "learning_rate": 2.8524094906914323e-08,
+      "loss": 0.7365,
       "step": 34418
     },
     {
-      "epoch": 0.9452912581362775,
+      "epoch": 0.9767026106696935,
       "grad_norm": 0.0,
-      "learning_rate": 1.5657399247315573e-07,
-      "loss": 0.8443,
+      "learning_rate": 2.8454768343504356e-08,
+      "loss": 0.7827,
       "step": 34419
     },
     {
-      "epoch": 0.9453187223641207,
+      "epoch": 0.9767309875141884,
       "grad_norm": 0.0,
-      "learning_rate": 1.5641724013375935e-07,
-      "loss": 0.7789,
+      "learning_rate": 2.8385526010185073e-08,
+      "loss": 0.8255,
       "step": 34420
     },
     {
-      "epoch": 0.945346186591964,
+      "epoch": 0.9767593643586833,
       "grad_norm": 0.0,
-      "learning_rate": 1.5626056568060154e-07,
-      "loss": 0.8493,
+      "learning_rate": 2.8316367907542663e-08,
+      "loss": 0.9013,
       "step": 34421
     },
     {
-      "epoch": 0.9453736508198072,
+      "epoch": 0.9767877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 1.5610396911492464e-07,
-      "loss": 0.7511,
+      "learning_rate": 2.824729403615889e-08,
+      "loss": 0.8164,
       "step": 34422
     },
     {
-      "epoch": 0.9454011150476505,
+      "epoch": 0.9768161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 1.5594745043796543e-07,
-      "loss": 0.8638,
+      "learning_rate": 2.817830439661995e-08,
+      "loss": 0.7357,
       "step": 34423
     },
     {
-      "epoch": 0.9454285792754936,
+      "epoch": 0.976844494892168,
       "grad_norm": 0.0,
-      "learning_rate": 1.5579100965096295e-07,
-      "loss": 0.8345,
+      "learning_rate": 2.8109398989505376e-08,
+      "loss": 0.8814,
       "step": 34424
     },
     {
-      "epoch": 0.9454560435033369,
+      "epoch": 0.9768728717366629,
       "grad_norm": 0.0,
-      "learning_rate": 1.5563464675515615e-07,
-      "loss": 0.8762,
+      "learning_rate": 2.804057781540026e-08,
+      "loss": 0.8669,
       "step": 34425
     },
     {
-      "epoch": 0.9454835077311802,
+      "epoch": 0.9769012485811578,
       "grad_norm": 0.0,
-      "learning_rate": 1.5547836175178078e-07,
-      "loss": 0.6844,
+      "learning_rate": 2.7971840874884138e-08,
+      "loss": 0.731,
       "step": 34426
     },
     {
-      "epoch": 0.9455109719590233,
+      "epoch": 0.9769296254256526,
       "grad_norm": 0.0,
-      "learning_rate": 1.5532215464207357e-07,
-      "loss": 0.918,
+      "learning_rate": 2.7903188168538763e-08,
+      "loss": 0.7604,
       "step": 34427
     },
     {
-      "epoch": 0.9455384361868666,
+      "epoch": 0.9769580022701475,
       "grad_norm": 0.0,
-      "learning_rate": 1.5516602542727133e-07,
-      "loss": 0.7089,
+      "learning_rate": 2.7834619696942565e-08,
+      "loss": 0.7338,
       "step": 34428
     },
     {
-      "epoch": 0.9455659004147099,
+      "epoch": 0.9769863791146425,
       "grad_norm": 0.0,
-      "learning_rate": 1.5500997410860974e-07,
-      "loss": 0.7219,
+      "learning_rate": 2.776613546067619e-08,
+      "loss": 0.928,
       "step": 34429
     },
     {
-      "epoch": 0.9455933646425531,
+      "epoch": 0.9770147559591373,
       "grad_norm": 0.0,
-      "learning_rate": 1.5485400068732226e-07,
-      "loss": 0.8204,
+      "learning_rate": 2.7697735460316954e-08,
+      "loss": 0.7461,
       "step": 34430
     },
     {
-      "epoch": 0.9456208288703963,
+      "epoch": 0.9770431328036322,
       "grad_norm": 0.0,
-      "learning_rate": 1.5469810516464344e-07,
-      "loss": 0.8799,
+      "learning_rate": 2.7629419696444392e-08,
+      "loss": 0.8523,
       "step": 34431
     },
     {
-      "epoch": 0.9456482930982395,
+      "epoch": 0.9770715096481272,
       "grad_norm": 0.0,
-      "learning_rate": 1.5454228754180679e-07,
-      "loss": 0.7421,
+      "learning_rate": 2.7561188169633602e-08,
+      "loss": 0.7994,
       "step": 34432
     },
     {
-      "epoch": 0.9456757573260828,
+      "epoch": 0.977099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 1.543865478200457e-07,
-      "loss": 0.783,
+      "learning_rate": 2.7493040880461896e-08,
+      "loss": 0.8407,
       "step": 34433
     },
     {
-      "epoch": 0.945703221553926,
+      "epoch": 0.9771282633371169,
       "grad_norm": 0.0,
-      "learning_rate": 1.5423088600059145e-07,
-      "loss": 0.7373,
+      "learning_rate": 2.7424977829504373e-08,
+      "loss": 0.8142,
       "step": 34434
     },
     {
-      "epoch": 0.9457306857817692,
+      "epoch": 0.9771566401816119,
       "grad_norm": 0.0,
-      "learning_rate": 1.5407530208467748e-07,
-      "loss": 0.8443,
+      "learning_rate": 2.735699901733613e-08,
+      "loss": 0.8216,
       "step": 34435
     },
     {
-      "epoch": 0.9457581500096125,
+      "epoch": 0.9771850170261067,
       "grad_norm": 0.0,
-      "learning_rate": 1.539197960735317e-07,
-      "loss": 0.7445,
+      "learning_rate": 2.7289104444532255e-08,
+      "loss": 0.776,
       "step": 34436
     },
     {
-      "epoch": 0.9457856142374557,
+      "epoch": 0.9772133938706016,
       "grad_norm": 0.0,
-      "learning_rate": 1.5376436796838866e-07,
-      "loss": 0.755,
+      "learning_rate": 2.722129411166563e-08,
+      "loss": 0.7679,
       "step": 34437
     },
     {
-      "epoch": 0.9458130784652989,
+      "epoch": 0.9772417707150964,
       "grad_norm": 0.0,
-      "learning_rate": 1.5360901777047633e-07,
-      "loss": 0.8039,
+      "learning_rate": 2.715356801930913e-08,
+      "loss": 0.7772,
       "step": 34438
     },
     {
-      "epoch": 0.9458405426931422,
+      "epoch": 0.9772701475595914,
       "grad_norm": 0.0,
-      "learning_rate": 1.5345374548102143e-07,
-      "loss": 0.7664,
+      "learning_rate": 2.7085926168034517e-08,
+      "loss": 0.7564,
       "step": 34439
     },
     {
-      "epoch": 0.9458680069209854,
+      "epoch": 0.9772985244040863,
       "grad_norm": 0.0,
-      "learning_rate": 1.5329855110125523e-07,
-      "loss": 0.7086,
+      "learning_rate": 2.7018368558413553e-08,
+      "loss": 0.8296,
       "step": 34440
     },
     {
-      "epoch": 0.9458954711488287,
+      "epoch": 0.9773269012485811,
       "grad_norm": 0.0,
-      "learning_rate": 1.5314343463240567e-07,
-      "loss": 0.7562,
+      "learning_rate": 2.6950895191016902e-08,
+      "loss": 0.7985,
       "step": 34441
     },
     {
-      "epoch": 0.9459229353766719,
+      "epoch": 0.9773552780930761,
       "grad_norm": 0.0,
-      "learning_rate": 1.529883960756995e-07,
-      "loss": 0.7823,
+      "learning_rate": 2.68835060664141e-08,
+      "loss": 0.7898,
       "step": 34442
     },
     {
-      "epoch": 0.9459503996045151,
+      "epoch": 0.977383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 1.5283343543236351e-07,
-      "loss": 0.7792,
+      "learning_rate": 2.6816201185174695e-08,
+      "loss": 0.7565,
       "step": 34443
     },
     {
-      "epoch": 0.9459778638323584,
+      "epoch": 0.9774120317820658,
       "grad_norm": 0.0,
-      "learning_rate": 1.5267855270362342e-07,
-      "loss": 0.6725,
+      "learning_rate": 2.674898054786712e-08,
+      "loss": 0.7808,
       "step": 34444
     },
     {
-      "epoch": 0.9460053280602015,
+      "epoch": 0.9774404086265607,
       "grad_norm": 0.0,
-      "learning_rate": 1.52523747890706e-07,
-      "loss": 0.8649,
+      "learning_rate": 2.6681844155058702e-08,
+      "loss": 0.914,
       "step": 34445
     },
     {
-      "epoch": 0.9460327922880448,
+      "epoch": 0.9774687854710556,
       "grad_norm": 0.0,
-      "learning_rate": 1.5236902099483474e-07,
-      "loss": 0.8663,
+      "learning_rate": 2.661479200731787e-08,
+      "loss": 0.7456,
       "step": 34446
     },
     {
-      "epoch": 0.9460602565158881,
+      "epoch": 0.9774971623155505,
       "grad_norm": 0.0,
-      "learning_rate": 1.5221437201723422e-07,
-      "loss": 0.6689,
+      "learning_rate": 2.654782410521084e-08,
+      "loss": 0.859,
       "step": 34447
     },
     {
-      "epoch": 0.9460877207437313,
+      "epoch": 0.9775255391600454,
       "grad_norm": 0.0,
-      "learning_rate": 1.5205980095912787e-07,
-      "loss": 0.7789,
+      "learning_rate": 2.6480940449301607e-08,
+      "loss": 0.7823,
       "step": 34448
     },
     {
-      "epoch": 0.9461151849715745,
+      "epoch": 0.9775539160045403,
       "grad_norm": 0.0,
-      "learning_rate": 1.519053078217403e-07,
-      "loss": 0.8507,
+      "learning_rate": 2.6414141040156382e-08,
+      "loss": 0.8461,
       "step": 34449
     },
     {
-      "epoch": 0.9461426491994177,
+      "epoch": 0.9775822928490352,
       "grad_norm": 0.0,
-      "learning_rate": 1.5175089260629162e-07,
-      "loss": 0.7757,
+      "learning_rate": 2.634742587833916e-08,
+      "loss": 0.8134,
       "step": 34450
     },
     {
-      "epoch": 0.946170113427261,
+      "epoch": 0.97761066969353,
       "grad_norm": 0.0,
-      "learning_rate": 1.515965553140053e-07,
-      "loss": 0.8475,
+      "learning_rate": 2.6280794964412827e-08,
+      "loss": 0.8408,
       "step": 34451
     },
     {
-      "epoch": 0.9461975776551043,
+      "epoch": 0.977639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 1.514422959461015e-07,
-      "loss": 0.8433,
+      "learning_rate": 2.621424829894248e-08,
+      "loss": 0.7287,
       "step": 34452
     },
     {
-      "epoch": 0.9462250418829474,
+      "epoch": 0.9776674233825199,
       "grad_norm": 0.0,
-      "learning_rate": 1.5128811450380142e-07,
-      "loss": 0.785,
+      "learning_rate": 2.6147785882487676e-08,
+      "loss": 0.8332,
       "step": 34453
     },
     {
-      "epoch": 0.9462525061107907,
+      "epoch": 0.9776958002270147,
       "grad_norm": 0.0,
-      "learning_rate": 1.5113401098832526e-07,
-      "loss": 0.7671,
+      "learning_rate": 2.6081407715611297e-08,
+      "loss": 0.8653,
       "step": 34454
     },
     {
-      "epoch": 0.946279970338634,
+      "epoch": 0.9777241770715096,
       "grad_norm": 0.0,
-      "learning_rate": 1.5097998540089198e-07,
-      "loss": 0.8948,
+      "learning_rate": 2.6015113798874004e-08,
+      "loss": 0.7367,
       "step": 34455
     },
     {
-      "epoch": 0.9463074345664771,
+      "epoch": 0.9777525539160046,
       "grad_norm": 0.0,
-      "learning_rate": 1.5082603774271952e-07,
-      "loss": 0.7545,
+      "learning_rate": 2.5948904132835352e-08,
+      "loss": 0.8765,
       "step": 34456
     },
     {
-      "epoch": 0.9463348987943204,
+      "epoch": 0.9777809307604994,
       "grad_norm": 0.0,
-      "learning_rate": 1.5067216801502694e-07,
-      "loss": 0.8713,
+      "learning_rate": 2.588277871805378e-08,
+      "loss": 0.9384,
       "step": 34457
     },
     {
-      "epoch": 0.9463623630221636,
+      "epoch": 0.9778093076049943,
       "grad_norm": 0.0,
-      "learning_rate": 1.5051837621903098e-07,
-      "loss": 0.7785,
+      "learning_rate": 2.5816737555089954e-08,
+      "loss": 0.795,
       "step": 34458
     },
     {
-      "epoch": 0.9463898272500069,
+      "epoch": 0.9778376844494893,
       "grad_norm": 0.0,
-      "learning_rate": 1.503646623559485e-07,
-      "loss": 0.7871,
+      "learning_rate": 2.5750780644500096e-08,
+      "loss": 0.7636,
       "step": 34459
     },
     {
-      "epoch": 0.9464172914778501,
+      "epoch": 0.9778660612939841,
       "grad_norm": 0.0,
-      "learning_rate": 1.5021102642699736e-07,
-      "loss": 0.9256,
+      "learning_rate": 2.5684907986841534e-08,
+      "loss": 0.8089,
       "step": 34460
     },
     {
-      "epoch": 0.9464447557056933,
+      "epoch": 0.977894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.500574684333922e-07,
-      "loss": 0.7922,
+      "learning_rate": 2.5619119582670494e-08,
+      "loss": 0.8619,
       "step": 34461
     },
     {
-      "epoch": 0.9464722199335366,
+      "epoch": 0.9779228149829738,
       "grad_norm": 0.0,
-      "learning_rate": 1.4990398837634756e-07,
-      "loss": 0.7757,
+      "learning_rate": 2.5553415432544303e-08,
+      "loss": 0.7141,
       "step": 34462
     },
     {
-      "epoch": 0.9464996841613798,
+      "epoch": 0.9779511918274688,
       "grad_norm": 0.0,
-      "learning_rate": 1.4975058625707806e-07,
-      "loss": 0.7876,
+      "learning_rate": 2.5487795537015858e-08,
+      "loss": 0.8015,
       "step": 34463
     },
     {
-      "epoch": 0.946527148389223,
+      "epoch": 0.9779795686719637,
       "grad_norm": 0.0,
-      "learning_rate": 1.4959726207679715e-07,
-      "loss": 0.6998,
+      "learning_rate": 2.5422259896640266e-08,
+      "loss": 0.7659,
       "step": 34464
     },
     {
-      "epoch": 0.9465546126170663,
+      "epoch": 0.9780079455164585,
       "grad_norm": 0.0,
-      "learning_rate": 1.494440158367183e-07,
-      "loss": 0.8572,
+      "learning_rate": 2.535680851197153e-08,
+      "loss": 0.8988,
       "step": 34465
     },
     {
-      "epoch": 0.9465820768449095,
+      "epoch": 0.9780363223609535,
       "grad_norm": 0.0,
-      "learning_rate": 1.492908475380539e-07,
-      "loss": 0.8721,
+      "learning_rate": 2.529144138356254e-08,
+      "loss": 0.8627,
       "step": 34466
     },
     {
-      "epoch": 0.9466095410727527,
+      "epoch": 0.9780646992054484,
       "grad_norm": 0.0,
-      "learning_rate": 1.491377571820174e-07,
-      "loss": 0.7665,
+      "learning_rate": 2.522615851196508e-08,
+      "loss": 0.7314,
       "step": 34467
     },
     {
-      "epoch": 0.946637005300596,
+      "epoch": 0.9780930760499432,
       "grad_norm": 0.0,
-      "learning_rate": 1.4898474476981783e-07,
-      "loss": 0.8604,
+      "learning_rate": 2.516095989772982e-08,
+      "loss": 0.7846,
       "step": 34468
     },
     {
-      "epoch": 0.9466644695284392,
+      "epoch": 0.9781214528944382,
       "grad_norm": 0.0,
-      "learning_rate": 1.4883181030266757e-07,
-      "loss": 0.7683,
+      "learning_rate": 2.5095845541409648e-08,
+      "loss": 0.7404,
       "step": 34469
     },
     {
-      "epoch": 0.9466919337562825,
+      "epoch": 0.978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 1.4867895378177567e-07,
-      "loss": 0.825,
+      "learning_rate": 2.5030815443551905e-08,
+      "loss": 0.8213,
       "step": 34470
     },
     {
-      "epoch": 0.9467193979841256,
+      "epoch": 0.9781782065834279,
       "grad_norm": 0.0,
-      "learning_rate": 1.4852617520835112e-07,
-      "loss": 0.8623,
+      "learning_rate": 2.496586960470837e-08,
+      "loss": 0.8307,
       "step": 34471
     },
     {
-      "epoch": 0.9467468622119689,
+      "epoch": 0.9782065834279228,
       "grad_norm": 0.0,
-      "learning_rate": 1.4837347458360406e-07,
-      "loss": 0.7949,
+      "learning_rate": 2.490100802542639e-08,
+      "loss": 0.8579,
       "step": 34472
     },
     {
-      "epoch": 0.9467743264398122,
+      "epoch": 0.9782349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.482208519087436e-07,
-      "loss": 0.8097,
+      "learning_rate": 2.4836230706253294e-08,
+      "loss": 0.7389,
       "step": 34473
     },
     {
-      "epoch": 0.9468017906676554,
+      "epoch": 0.9782633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.4806830718497424e-07,
-      "loss": 0.7498,
+      "learning_rate": 2.4771537647737543e-08,
+      "loss": 0.8865,
       "step": 34474
     },
     {
-      "epoch": 0.9468292548954986,
+      "epoch": 0.9782917139614075,
       "grad_norm": 0.0,
-      "learning_rate": 1.4791584041350503e-07,
-      "loss": 0.752,
+      "learning_rate": 2.470692885042536e-08,
+      "loss": 0.7532,
       "step": 34475
     },
     {
-      "epoch": 0.9468567191233418,
+      "epoch": 0.9783200908059024,
       "grad_norm": 0.0,
-      "learning_rate": 1.4776345159554174e-07,
-      "loss": 0.7449,
+      "learning_rate": 2.4642404314861866e-08,
+      "loss": 0.7963,
       "step": 34476
     },
     {
-      "epoch": 0.9468841833511851,
+      "epoch": 0.9783484676503973,
       "grad_norm": 0.0,
-      "learning_rate": 1.476111407322911e-07,
-      "loss": 0.7327,
+      "learning_rate": 2.4577964041592185e-08,
+      "loss": 0.8846,
       "step": 34477
     },
     {
-      "epoch": 0.9469116475790283,
+      "epoch": 0.9783768444948922,
       "grad_norm": 0.0,
-      "learning_rate": 1.4745890782495776e-07,
-      "loss": 0.7722,
+      "learning_rate": 2.4513608031160318e-08,
+      "loss": 0.7597,
       "step": 34478
     },
     {
-      "epoch": 0.9469391118068715,
+      "epoch": 0.978405221339387,
       "grad_norm": 0.0,
-      "learning_rate": 1.4730675287474626e-07,
-      "loss": 0.8444,
+      "learning_rate": 2.444933628411139e-08,
+      "loss": 0.874,
       "step": 34479
     },
     {
-      "epoch": 0.9469665760347148,
+      "epoch": 0.978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 1.4715467588285904e-07,
-      "loss": 0.7967,
+      "learning_rate": 2.438514880098719e-08,
+      "loss": 0.7699,
       "step": 34480
     },
     {
-      "epoch": 0.9469940402625581,
+      "epoch": 0.9784619750283768,
       "grad_norm": 0.0,
-      "learning_rate": 1.4700267685050173e-07,
-      "loss": 0.7757,
+      "learning_rate": 2.4321045582329504e-08,
+      "loss": 0.7774,
       "step": 34481
     },
     {
-      "epoch": 0.9470215044904012,
+      "epoch": 0.9784903518728717,
       "grad_norm": 0.0,
-      "learning_rate": 1.4685075577887454e-07,
-      "loss": 0.8787,
+      "learning_rate": 2.4257026628681236e-08,
+      "loss": 0.7621,
       "step": 34482
     },
     {
-      "epoch": 0.9470489687182445,
+      "epoch": 0.9785187287173667,
       "grad_norm": 0.0,
-      "learning_rate": 1.46698912669182e-07,
-      "loss": 0.8389,
+      "learning_rate": 2.419309194058195e-08,
+      "loss": 0.7491,
       "step": 34483
     },
     {
-      "epoch": 0.9470764329460877,
+      "epoch": 0.9785471055618615,
       "grad_norm": 0.0,
-      "learning_rate": 1.4654714752262544e-07,
-      "loss": 0.8245,
+      "learning_rate": 2.412924151857121e-08,
+      "loss": 0.765,
       "step": 34484
     },
     {
-      "epoch": 0.947103897173931,
+      "epoch": 0.9785754824063564,
       "grad_norm": 0.0,
-      "learning_rate": 1.4639546034040275e-07,
-      "loss": 0.6478,
+      "learning_rate": 2.4065475363189705e-08,
+      "loss": 0.762,
       "step": 34485
     },
     {
-      "epoch": 0.9471313614017742,
+      "epoch": 0.9786038592508514,
       "grad_norm": 0.0,
-      "learning_rate": 1.462438511237174e-07,
-      "loss": 0.8018,
+      "learning_rate": 2.4001793474976996e-08,
+      "loss": 0.7161,
       "step": 34486
     },
     {
-      "epoch": 0.9471588256296174,
+      "epoch": 0.9786322360953462,
       "grad_norm": 0.0,
-      "learning_rate": 1.4609231987376737e-07,
-      "loss": 0.811,
+      "learning_rate": 2.3938195854468217e-08,
+      "loss": 0.7881,
       "step": 34487
     },
     {
-      "epoch": 0.9471862898574607,
+      "epoch": 0.9786606129398411,
       "grad_norm": 0.0,
-      "learning_rate": 1.4594086659175166e-07,
-      "loss": 0.7311,
+      "learning_rate": 2.387468250220182e-08,
+      "loss": 0.9042,
       "step": 34488
     },
     {
-      "epoch": 0.9472137540853038,
+      "epoch": 0.9786889897843359,
       "grad_norm": 0.0,
-      "learning_rate": 1.457894912788693e-07,
-      "loss": 0.8001,
+      "learning_rate": 2.3811253418715152e-08,
+      "loss": 0.8894,
       "step": 34489
     },
     {
-      "epoch": 0.9472412183131471,
+      "epoch": 0.9787173666288309,
       "grad_norm": 0.0,
-      "learning_rate": 1.4563819393631828e-07,
-      "loss": 0.7164,
+      "learning_rate": 2.3747908604542237e-08,
+      "loss": 0.7759,
       "step": 34490
     },
     {
-      "epoch": 0.9472686825409904,
+      "epoch": 0.9787457434733258,
       "grad_norm": 0.0,
-      "learning_rate": 1.4548697456529314e-07,
-      "loss": 0.789,
+      "learning_rate": 2.3684648060220416e-08,
+      "loss": 0.8402,
       "step": 34491
     },
     {
-      "epoch": 0.9472961467688336,
+      "epoch": 0.9787741203178206,
       "grad_norm": 0.0,
-      "learning_rate": 1.4533583316699406e-07,
-      "loss": 0.8801,
+      "learning_rate": 2.36214717862826e-08,
+      "loss": 0.81,
       "step": 34492
     },
     {
-      "epoch": 0.9473236109966768,
+      "epoch": 0.9788024971623156,
       "grad_norm": 0.0,
-      "learning_rate": 1.4518476974261453e-07,
-      "loss": 0.8604,
+      "learning_rate": 2.35583797832617e-08,
+      "loss": 0.7892,
       "step": 34493
     },
     {
-      "epoch": 0.9473510752245201,
+      "epoch": 0.9788308740068105,
       "grad_norm": 0.0,
-      "learning_rate": 1.4503378429335023e-07,
-      "loss": 0.917,
+      "learning_rate": 2.3495372051692832e-08,
+      "loss": 0.8129,
       "step": 34494
     },
     {
-      "epoch": 0.9473785394523633,
+      "epoch": 0.9788592508513053,
       "grad_norm": 0.0,
-      "learning_rate": 1.4488287682039693e-07,
-      "loss": 0.8615,
+      "learning_rate": 2.3432448592106694e-08,
+      "loss": 0.8466,
       "step": 34495
     },
     {
-      "epoch": 0.9474060036802066,
+      "epoch": 0.9788876276958002,
       "grad_norm": 0.0,
-      "learning_rate": 1.4473204732494695e-07,
-      "loss": 0.8521,
+      "learning_rate": 2.3369609405035076e-08,
+      "loss": 0.753,
       "step": 34496
     },
     {
-      "epoch": 0.9474334679080497,
+      "epoch": 0.9789160045402951,
       "grad_norm": 0.0,
-      "learning_rate": 1.445812958081949e-07,
-      "loss": 0.7831,
+      "learning_rate": 2.330685449100867e-08,
+      "loss": 0.8636,
       "step": 34497
     },
     {
-      "epoch": 0.947460932135893,
+      "epoch": 0.97894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 1.444306222713332e-07,
-      "loss": 0.7557,
+      "learning_rate": 2.3244183850558154e-08,
+      "loss": 0.818,
       "step": 34498
     },
     {
-      "epoch": 0.9474883963637363,
+      "epoch": 0.9789727582292849,
       "grad_norm": 0.0,
-      "learning_rate": 1.442800267155542e-07,
-      "loss": 0.828,
+      "learning_rate": 2.3181597484212005e-08,
+      "loss": 0.7952,
       "step": 34499
     },
     {
-      "epoch": 0.9475158605915794,
+      "epoch": 0.9790011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 1.4412950914204915e-07,
-      "loss": 0.8003,
+      "learning_rate": 2.3119095392499792e-08,
+      "loss": 0.7253,
       "step": 34500
     },
     {
-      "epoch": 0.9475433248194227,
+      "epoch": 0.9790295119182747,
       "grad_norm": 0.0,
-      "learning_rate": 1.4397906955200937e-07,
-      "loss": 0.7791,
+      "learning_rate": 2.3056677575948872e-08,
+      "loss": 0.6704,
       "step": 34501
     },
     {
-      "epoch": 0.9475707890472659,
+      "epoch": 0.9790578887627696,
       "grad_norm": 0.0,
-      "learning_rate": 1.4382870794662606e-07,
-      "loss": 0.815,
+      "learning_rate": 2.29943440350866e-08,
+      "loss": 0.7218,
       "step": 34502
     },
     {
-      "epoch": 0.9475982532751092,
+      "epoch": 0.9790862656072645,
       "grad_norm": 0.0,
-      "learning_rate": 1.4367842432708611e-07,
-      "loss": 0.826,
+      "learning_rate": 2.293209477043923e-08,
+      "loss": 0.9178,
       "step": 34503
     },
     {
-      "epoch": 0.9476257175029524,
+      "epoch": 0.9791146424517594,
       "grad_norm": 0.0,
-      "learning_rate": 1.4352821869458188e-07,
-      "loss": 0.7273,
+      "learning_rate": 2.2869929782534108e-08,
+      "loss": 0.797,
       "step": 34504
     },
     {
-      "epoch": 0.9476531817307956,
+      "epoch": 0.9791430192962542,
       "grad_norm": 0.0,
-      "learning_rate": 1.433780910502991e-07,
-      "loss": 0.8875,
+      "learning_rate": 2.280784907189415e-08,
+      "loss": 0.8909,
       "step": 34505
     },
     {
-      "epoch": 0.9476806459586389,
+      "epoch": 0.9791713961407491,
       "grad_norm": 0.0,
-      "learning_rate": 1.432280413954279e-07,
-      "loss": 0.7979,
+      "learning_rate": 2.2745852639045608e-08,
+      "loss": 0.7682,
       "step": 34506
     },
     {
-      "epoch": 0.9477081101864822,
+      "epoch": 0.9791997729852441,
       "grad_norm": 0.0,
-      "learning_rate": 1.4307806973115401e-07,
-      "loss": 0.6744,
+      "learning_rate": 2.268394048451028e-08,
+      "loss": 0.78,
       "step": 34507
     },
     {
-      "epoch": 0.9477355744143253,
+      "epoch": 0.9792281498297389,
       "grad_norm": 0.0,
-      "learning_rate": 1.4292817605866537e-07,
-      "loss": 0.788,
+      "learning_rate": 2.2622112608813307e-08,
+      "loss": 0.746,
       "step": 34508
     },
     {
-      "epoch": 0.9477630386421686,
+      "epoch": 0.9792565266742338,
       "grad_norm": 0.0,
-      "learning_rate": 1.427783603791466e-07,
-      "loss": 0.7916,
+      "learning_rate": 2.256036901247538e-08,
+      "loss": 0.8212,
       "step": 34509
     },
     {
-      "epoch": 0.9477905028700118,
+      "epoch": 0.9792849035187288,
       "grad_norm": 0.0,
-      "learning_rate": 1.4262862269378452e-07,
-      "loss": 0.8941,
+      "learning_rate": 2.2498709696018307e-08,
+      "loss": 0.7631,
       "step": 34510
     },
     {
-      "epoch": 0.947817967097855,
+      "epoch": 0.9793132803632236,
       "grad_norm": 0.0,
-      "learning_rate": 1.4247896300376264e-07,
-      "loss": 0.8389,
+      "learning_rate": 2.2437134659962777e-08,
+      "loss": 0.8621,
       "step": 34511
     },
     {
-      "epoch": 0.9478454313256983,
+      "epoch": 0.9793416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 1.4232938131026553e-07,
-      "loss": 0.7434,
+      "learning_rate": 2.23756439048306e-08,
+      "loss": 0.7473,
       "step": 34512
     },
     {
-      "epoch": 0.9478728955535415,
+      "epoch": 0.9793700340522133,
       "grad_norm": 0.0,
-      "learning_rate": 1.4217987761447671e-07,
-      "loss": 0.7995,
+      "learning_rate": 2.231423743113914e-08,
+      "loss": 0.8026,
       "step": 34513
     },
     {
-      "epoch": 0.9479003597813848,
+      "epoch": 0.9793984108967083,
       "grad_norm": 0.0,
-      "learning_rate": 1.420304519175797e-07,
-      "loss": 0.8288,
+      "learning_rate": 2.225291523940798e-08,
+      "loss": 0.7961,
       "step": 34514
     },
     {
-      "epoch": 0.9479278240092279,
+      "epoch": 0.9794267877412032,
       "grad_norm": 0.0,
-      "learning_rate": 1.4188110422075686e-07,
-      "loss": 0.7771,
+      "learning_rate": 2.219167733015448e-08,
+      "loss": 0.746,
       "step": 34515
     },
     {
-      "epoch": 0.9479552882370712,
+      "epoch": 0.979455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.4173183452518835e-07,
-      "loss": 0.8533,
+      "learning_rate": 2.2130523703897123e-08,
+      "loss": 0.7641,
       "step": 34516
     },
     {
-      "epoch": 0.9479827524649145,
+      "epoch": 0.979483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.415826428320566e-07,
-      "loss": 0.8068,
+      "learning_rate": 2.2069454361151042e-08,
+      "loss": 0.6824,
       "step": 34517
     },
     {
-      "epoch": 0.9480102166927576,
+      "epoch": 0.9795119182746879,
       "grad_norm": 0.0,
-      "learning_rate": 1.414335291425417e-07,
-      "loss": 0.781,
+      "learning_rate": 2.2008469302433612e-08,
+      "loss": 0.8673,
       "step": 34518
     },
     {
-      "epoch": 0.9480376809206009,
+      "epoch": 0.9795402951191827,
       "grad_norm": 0.0,
-      "learning_rate": 1.412844934578239e-07,
-      "loss": 0.7616,
+      "learning_rate": 2.1947568528258854e-08,
+      "loss": 0.7064,
       "step": 34519
     },
     {
-      "epoch": 0.9480651451484442,
+      "epoch": 0.9795686719636776,
       "grad_norm": 0.0,
-      "learning_rate": 1.4113553577908223e-07,
-      "loss": 0.8531,
+      "learning_rate": 2.188675203914192e-08,
+      "loss": 0.7558,
       "step": 34520
     },
     {
-      "epoch": 0.9480926093762874,
+      "epoch": 0.9795970488081726,
       "grad_norm": 0.0,
-      "learning_rate": 1.4098665610749464e-07,
-      "loss": 0.7593,
+      "learning_rate": 2.1826019835595735e-08,
+      "loss": 0.852,
       "step": 34521
     },
     {
-      "epoch": 0.9481200736041306,
+      "epoch": 0.9796254256526674,
       "grad_norm": 0.0,
-      "learning_rate": 1.4083785444424015e-07,
-      "loss": 0.7433,
+      "learning_rate": 2.1765371918133216e-08,
+      "loss": 0.7295,
       "step": 34522
     },
     {
-      "epoch": 0.9481475378319738,
+      "epoch": 0.9796538024971623,
       "grad_norm": 0.0,
-      "learning_rate": 1.4068913079049452e-07,
-      "loss": 0.7447,
+      "learning_rate": 2.1704808287267286e-08,
+      "loss": 0.7845,
       "step": 34523
     },
     {
-      "epoch": 0.9481750020598171,
+      "epoch": 0.9796821793416572,
       "grad_norm": 0.0,
-      "learning_rate": 1.4054048514743678e-07,
-      "loss": 0.8469,
+      "learning_rate": 2.1644328943509762e-08,
+      "loss": 0.8389,
       "step": 34524
     },
     {
-      "epoch": 0.9482024662876604,
+      "epoch": 0.9797105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 1.403919175162416e-07,
-      "loss": 0.7749,
+      "learning_rate": 2.1583933887371345e-08,
+      "loss": 0.7166,
       "step": 34525
     },
     {
-      "epoch": 0.9482299305155035,
+      "epoch": 0.979738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.4024342789808575e-07,
-      "loss": 0.7327,
+      "learning_rate": 2.1523623119361627e-08,
+      "loss": 0.7682,
       "step": 34526
     },
     {
-      "epoch": 0.9482573947433468,
+      "epoch": 0.9797673098751419,
       "grad_norm": 0.0,
-      "learning_rate": 1.4009501629414168e-07,
-      "loss": 0.8164,
+      "learning_rate": 2.1463396639991306e-08,
+      "loss": 0.7743,
       "step": 34527
     },
     {
-      "epoch": 0.94828485897119,
+      "epoch": 0.9797956867196368,
       "grad_norm": 0.0,
-      "learning_rate": 1.399466827055851e-07,
-      "loss": 0.8287,
+      "learning_rate": 2.140325444976665e-08,
+      "loss": 0.8423,
       "step": 34528
     },
     {
-      "epoch": 0.9483123231990332,
+      "epoch": 0.9798240635641317,
       "grad_norm": 0.0,
-      "learning_rate": 1.397984271335906e-07,
-      "loss": 0.73,
+      "learning_rate": 2.1343196549198363e-08,
+      "loss": 0.7904,
       "step": 34529
     },
     {
-      "epoch": 0.9483397874268765,
+      "epoch": 0.9798524404086265,
       "grad_norm": 0.0,
-      "learning_rate": 1.396502495793306e-07,
-      "loss": 0.7198,
+      "learning_rate": 2.1283222938792702e-08,
+      "loss": 0.8135,
       "step": 34530
     },
     {
-      "epoch": 0.9483672516547197,
+      "epoch": 0.9798808172531215,
       "grad_norm": 0.0,
-      "learning_rate": 1.3950215004397638e-07,
-      "loss": 0.7996,
+      "learning_rate": 2.122333361905593e-08,
+      "loss": 0.7837,
       "step": 34531
     },
     {
-      "epoch": 0.948394715882563,
+      "epoch": 0.9799091940976163,
       "grad_norm": 0.0,
-      "learning_rate": 1.3935412852870145e-07,
-      "loss": 0.8522,
+      "learning_rate": 2.116352859049431e-08,
+      "loss": 0.8826,
       "step": 34532
     },
     {
-      "epoch": 0.9484221801104062,
+      "epoch": 0.9799375709421112,
       "grad_norm": 0.0,
-      "learning_rate": 1.3920618503467599e-07,
-      "loss": 0.7381,
+      "learning_rate": 2.110380785361299e-08,
+      "loss": 0.8286,
       "step": 34533
     },
     {
-      "epoch": 0.9484496443382494,
+      "epoch": 0.9799659477866062,
       "grad_norm": 0.0,
-      "learning_rate": 1.3905831956307237e-07,
-      "loss": 0.8233,
+      "learning_rate": 2.1044171408916016e-08,
+      "loss": 0.6817,
       "step": 34534
     },
     {
-      "epoch": 0.9484771085660927,
+      "epoch": 0.979994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.3891053211505745e-07,
-      "loss": 0.754,
+      "learning_rate": 2.0984619256908534e-08,
+      "loss": 0.8198,
       "step": 34535
     },
     {
-      "epoch": 0.9485045727939359,
+      "epoch": 0.9800227014755959,
       "grad_norm": 0.0,
-      "learning_rate": 1.387628226918025e-07,
-      "loss": 0.8434,
+      "learning_rate": 2.092515139809126e-08,
+      "loss": 0.8139,
       "step": 34536
     },
     {
-      "epoch": 0.9485320370217791,
+      "epoch": 0.9800510783200908,
       "grad_norm": 0.0,
-      "learning_rate": 1.3861519129447554e-07,
-      "loss": 0.6985,
+      "learning_rate": 2.086576783296934e-08,
+      "loss": 0.7366,
       "step": 34537
     },
     {
-      "epoch": 0.9485595012496224,
+      "epoch": 0.9800794551645857,
       "grad_norm": 0.0,
-      "learning_rate": 1.3846763792424667e-07,
-      "loss": 0.7504,
+      "learning_rate": 2.0806468562041272e-08,
+      "loss": 0.8233,
       "step": 34538
     },
     {
-      "epoch": 0.9485869654774656,
+      "epoch": 0.9801078320090806,
       "grad_norm": 0.0,
-      "learning_rate": 1.3832016258228053e-07,
-      "loss": 0.7061,
+      "learning_rate": 2.074725358580998e-08,
+      "loss": 0.738,
       "step": 34539
     },
     {
-      "epoch": 0.9486144297053088,
+      "epoch": 0.9801362088535754,
       "grad_norm": 0.0,
-      "learning_rate": 1.381727652697451e-07,
-      "loss": 0.735,
+      "learning_rate": 2.0688122904776175e-08,
+      "loss": 0.8508,
       "step": 34540
     },
     {
-      "epoch": 0.948641893933152,
+      "epoch": 0.9801645856980704,
       "grad_norm": 0.0,
-      "learning_rate": 1.380254459878083e-07,
-      "loss": 0.8799,
+      "learning_rate": 2.062907651943724e-08,
+      "loss": 0.8968,
       "step": 34541
     },
     {
-      "epoch": 0.9486693581609953,
+      "epoch": 0.9801929625425653,
       "grad_norm": 0.0,
-      "learning_rate": 1.378782047376326e-07,
-      "loss": 0.7722,
+      "learning_rate": 2.0570114430293887e-08,
+      "loss": 0.8309,
       "step": 34542
     },
     {
-      "epoch": 0.9486968223888386,
+      "epoch": 0.9802213393870601,
       "grad_norm": 0.0,
-      "learning_rate": 1.3773104152038697e-07,
-      "loss": 0.8235,
+      "learning_rate": 2.051123663784238e-08,
+      "loss": 0.8531,
       "step": 34543
     },
     {
-      "epoch": 0.9487242866166817,
+      "epoch": 0.9802497162315551,
       "grad_norm": 0.0,
-      "learning_rate": 1.3758395633723166e-07,
-      "loss": 0.8075,
+      "learning_rate": 2.0452443142582322e-08,
+      "loss": 0.7784,
       "step": 34544
     },
     {
-      "epoch": 0.948751750844525,
+      "epoch": 0.98027809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.3743694918933237e-07,
-      "loss": 0.8163,
+      "learning_rate": 2.0393733945007764e-08,
+      "loss": 0.84,
       "step": 34545
     },
     {
-      "epoch": 0.9487792150723683,
+      "epoch": 0.9803064699205448,
       "grad_norm": 0.0,
-      "learning_rate": 1.3729002007785262e-07,
-      "loss": 0.7808,
+      "learning_rate": 2.0335109045617196e-08,
+      "loss": 0.8024,
       "step": 34546
     },
     {
-      "epoch": 0.9488066793002115,
+      "epoch": 0.9803348467650397,
       "grad_norm": 0.0,
-      "learning_rate": 1.3714316900395485e-07,
-      "loss": 0.7212,
+      "learning_rate": 2.0276568444904666e-08,
+      "loss": 0.7389,
       "step": 34547
     },
     {
-      "epoch": 0.9488341435280547,
+      "epoch": 0.9803632236095347,
       "grad_norm": 0.0,
-      "learning_rate": 1.3699639596880142e-07,
-      "loss": 0.8403,
+      "learning_rate": 2.021811214336311e-08,
+      "loss": 0.8672,
       "step": 34548
     },
     {
-      "epoch": 0.9488616077558979,
+      "epoch": 0.9803916004540295,
       "grad_norm": 0.0,
-      "learning_rate": 1.3684970097355255e-07,
-      "loss": 0.8049,
+      "learning_rate": 2.0159740141488803e-08,
+      "loss": 0.7525,
       "step": 34549
     },
     {
-      "epoch": 0.9488890719837412,
+      "epoch": 0.9804199772985244,
       "grad_norm": 0.0,
-      "learning_rate": 1.367030840193706e-07,
-      "loss": 0.8124,
+      "learning_rate": 2.010145243977357e-08,
+      "loss": 0.7663,
       "step": 34550
     },
     {
-      "epoch": 0.9489165362115844,
+      "epoch": 0.9804483541430193,
       "grad_norm": 0.0,
-      "learning_rate": 1.3655654510741356e-07,
-      "loss": 0.7755,
+      "learning_rate": 2.0043249038710355e-08,
+      "loss": 0.7739,
       "step": 34551
     },
     {
-      "epoch": 0.9489440004394276,
+      "epoch": 0.9804767309875142,
       "grad_norm": 0.0,
-      "learning_rate": 1.3641008423884161e-07,
-      "loss": 0.7778,
+      "learning_rate": 1.9985129938790982e-08,
+      "loss": 0.7767,
       "step": 34552
     },
     {
-      "epoch": 0.9489714646672709,
+      "epoch": 0.9805051078320091,
       "grad_norm": 0.0,
-      "learning_rate": 1.3626370141481383e-07,
-      "loss": 0.892,
+      "learning_rate": 1.992709514050506e-08,
+      "loss": 0.8186,
       "step": 34553
     },
     {
-      "epoch": 0.9489989288951142,
+      "epoch": 0.9805334846765039,
       "grad_norm": 0.0,
-      "learning_rate": 1.3611739663648815e-07,
-      "loss": 0.8808,
+      "learning_rate": 1.9869144644343307e-08,
+      "loss": 0.8955,
       "step": 34554
     },
     {
-      "epoch": 0.9490263931229573,
+      "epoch": 0.9805618615209989,
       "grad_norm": 0.0,
-      "learning_rate": 1.3597116990502368e-07,
-      "loss": 0.8007,
+      "learning_rate": 1.9811278450795336e-08,
+      "loss": 0.8206,
       "step": 34555
     },
     {
-      "epoch": 0.9490538573508006,
+      "epoch": 0.9805902383654937,
       "grad_norm": 0.0,
-      "learning_rate": 1.3582502122157504e-07,
-      "loss": 0.7417,
+      "learning_rate": 1.975349656035075e-08,
+      "loss": 0.7473,
       "step": 34556
     },
     {
-      "epoch": 0.9490813215786438,
+      "epoch": 0.9806186152099886,
       "grad_norm": 0.0,
-      "learning_rate": 1.3567895058730018e-07,
-      "loss": 0.7565,
+      "learning_rate": 1.9695798973496937e-08,
+      "loss": 0.7578,
       "step": 34557
     },
     {
-      "epoch": 0.949108785806487,
+      "epoch": 0.9806469920544836,
       "grad_norm": 0.0,
-      "learning_rate": 1.3553295800335598e-07,
-      "loss": 0.7978,
+      "learning_rate": 1.9638185690721288e-08,
+      "loss": 0.7205,
       "step": 34558
     },
     {
-      "epoch": 0.9491362500343303,
+      "epoch": 0.9806753688989784,
       "grad_norm": 0.0,
-      "learning_rate": 1.3538704347089483e-07,
-      "loss": 0.7777,
+      "learning_rate": 1.958065671251008e-08,
+      "loss": 0.7603,
       "step": 34559
     },
     {
-      "epoch": 0.9491637142621735,
+      "epoch": 0.9807037457434733,
       "grad_norm": 0.0,
-      "learning_rate": 1.3524120699107245e-07,
-      "loss": 0.7671,
+      "learning_rate": 1.9523212039350703e-08,
+      "loss": 0.8461,
       "step": 34560
     },
     {
-      "epoch": 0.9491911784900168,
+      "epoch": 0.9807321225879683,
       "grad_norm": 0.0,
-      "learning_rate": 1.3509544856504352e-07,
-      "loss": 0.8332,
+      "learning_rate": 1.946585167172721e-08,
+      "loss": 0.7455,
       "step": 34561
     },
     {
-      "epoch": 0.9492186427178599,
+      "epoch": 0.9807604994324631,
       "grad_norm": 0.0,
-      "learning_rate": 1.3494976819396044e-07,
-      "loss": 0.7191,
+      "learning_rate": 1.9408575610123657e-08,
+      "loss": 0.8473,
       "step": 34562
     },
     {
-      "epoch": 0.9492461069457032,
+      "epoch": 0.980788876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.348041658789756e-07,
-      "loss": 0.7889,
+      "learning_rate": 1.9351383855025218e-08,
+      "loss": 0.7291,
       "step": 34563
     },
     {
-      "epoch": 0.9492735711735465,
+      "epoch": 0.9808172531214528,
       "grad_norm": 0.0,
-      "learning_rate": 1.3465864162124253e-07,
-      "loss": 0.6688,
+      "learning_rate": 1.9294276406913727e-08,
+      "loss": 0.6735,
       "step": 34564
     },
     {
-      "epoch": 0.9493010354013897,
+      "epoch": 0.9808456299659478,
       "grad_norm": 0.0,
-      "learning_rate": 1.3451319542191143e-07,
-      "loss": 0.8353,
+      "learning_rate": 1.9237253266272126e-08,
+      "loss": 0.7908,
       "step": 34565
     },
     {
-      "epoch": 0.9493284996292329,
+      "epoch": 0.9808740068104427,
       "grad_norm": 0.0,
-      "learning_rate": 1.343678272821336e-07,
-      "loss": 0.8962,
+      "learning_rate": 1.918031443358337e-08,
+      "loss": 0.8256,
       "step": 34566
     },
     {
-      "epoch": 0.9493559638570762,
+      "epoch": 0.9809023836549375,
       "grad_norm": 0.0,
-      "learning_rate": 1.3422253720305922e-07,
-      "loss": 0.8347,
+      "learning_rate": 1.912345990932596e-08,
+      "loss": 0.6852,
       "step": 34567
     },
     {
-      "epoch": 0.9493834280849194,
+      "epoch": 0.9809307604994325,
       "grad_norm": 0.0,
-      "learning_rate": 1.340773251858374e-07,
-      "loss": 0.771,
+      "learning_rate": 1.906668969398173e-08,
+      "loss": 0.7696,
       "step": 34568
     },
     {
-      "epoch": 0.9494108923127627,
+      "epoch": 0.9809591373439274,
       "grad_norm": 0.0,
-      "learning_rate": 1.3393219123161715e-07,
-      "loss": 0.8981,
+      "learning_rate": 1.9010003788029195e-08,
+      "loss": 0.8256,
       "step": 34569
     },
     {
-      "epoch": 0.9494383565406058,
+      "epoch": 0.9809875141884222,
       "grad_norm": 0.0,
-      "learning_rate": 1.3378713534154652e-07,
-      "loss": 0.7554,
+      "learning_rate": 1.8953402191947966e-08,
+      "loss": 0.8258,
       "step": 34570
     },
     {
-      "epoch": 0.9494658207684491,
+      "epoch": 0.9810158910329171,
       "grad_norm": 0.0,
-      "learning_rate": 1.3364215751677455e-07,
-      "loss": 0.784,
+      "learning_rate": 1.8896884906216552e-08,
+      "loss": 0.7259,
       "step": 34571
     },
     {
-      "epoch": 0.9494932849962924,
+      "epoch": 0.9810442678774121,
       "grad_norm": 0.0,
-      "learning_rate": 1.3349725775844812e-07,
-      "loss": 0.8918,
+      "learning_rate": 1.8840451931312342e-08,
+      "loss": 0.8531,
       "step": 34572
     },
     {
-      "epoch": 0.9495207492241355,
+      "epoch": 0.9810726447219069,
       "grad_norm": 0.0,
-      "learning_rate": 1.333524360677141e-07,
-      "loss": 0.8827,
+      "learning_rate": 1.8784103267710518e-08,
+      "loss": 0.7389,
       "step": 34573
     },
     {
-      "epoch": 0.9495482134519788,
+      "epoch": 0.9811010215664018,
       "grad_norm": 0.0,
-      "learning_rate": 1.3320769244571708e-07,
-      "loss": 0.8161,
+      "learning_rate": 1.8727838915888476e-08,
+      "loss": 0.7757,
       "step": 34574
     },
     {
-      "epoch": 0.949575677679822,
+      "epoch": 0.9811293984108967,
       "grad_norm": 0.0,
-      "learning_rate": 1.3306302689360173e-07,
-      "loss": 0.733,
+      "learning_rate": 1.8671658876321385e-08,
+      "loss": 0.6526,
       "step": 34575
     },
     {
-      "epoch": 0.9496031419076653,
+      "epoch": 0.9811577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.329184394125138e-07,
-      "loss": 0.7976,
+      "learning_rate": 1.8615563149482206e-08,
+      "loss": 0.7947,
       "step": 34576
     },
     {
-      "epoch": 0.9496306061355085,
+      "epoch": 0.9811861520998865,
       "grad_norm": 0.0,
-      "learning_rate": 1.3277393000359796e-07,
-      "loss": 0.8183,
+      "learning_rate": 1.8559551735847225e-08,
+      "loss": 0.8321,
       "step": 34577
     },
     {
-      "epoch": 0.9496580703633517,
+      "epoch": 0.9812145289443814,
       "grad_norm": 0.0,
-      "learning_rate": 1.3262949866799657e-07,
-      "loss": 0.8424,
+      "learning_rate": 1.8503624635888284e-08,
+      "loss": 0.728,
       "step": 34578
     },
     {
-      "epoch": 0.949685534591195,
+      "epoch": 0.9812429057888763,
       "grad_norm": 0.0,
-      "learning_rate": 1.324851454068543e-07,
-      "loss": 0.8146,
+      "learning_rate": 1.8447781850077227e-08,
+      "loss": 0.7597,
       "step": 34579
     },
     {
-      "epoch": 0.9497129988190383,
+      "epoch": 0.9812712826333712,
       "grad_norm": 0.0,
-      "learning_rate": 1.3234087022131027e-07,
-      "loss": 0.822,
+      "learning_rate": 1.8392023378888125e-08,
+      "loss": 0.7071,
       "step": 34580
     },
     {
-      "epoch": 0.9497404630468814,
+      "epoch": 0.981299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 1.3219667311250795e-07,
-      "loss": 0.8063,
+      "learning_rate": 1.8336349222788374e-08,
+      "loss": 0.6917,
       "step": 34581
     },
     {
-      "epoch": 0.9497679272747247,
+      "epoch": 0.981328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 1.3205255408158867e-07,
-      "loss": 0.7439,
+      "learning_rate": 1.8280759382250934e-08,
+      "loss": 0.7034,
       "step": 34582
     },
     {
-      "epoch": 0.9497953915025679,
+      "epoch": 0.9813564131668558,
       "grad_norm": 0.0,
-      "learning_rate": 1.3190851312969154e-07,
-      "loss": 0.818,
+      "learning_rate": 1.822525385774543e-08,
+      "loss": 0.7406,
       "step": 34583
     },
     {
-      "epoch": 0.9498228557304111,
+      "epoch": 0.9813847900113507,
       "grad_norm": 0.0,
-      "learning_rate": 1.3176455025795676e-07,
-      "loss": 0.7894,
+      "learning_rate": 1.816983264973926e-08,
+      "loss": 0.7301,
       "step": 34584
     },
     {
-      "epoch": 0.9498503199582544,
+      "epoch": 0.9814131668558457,
       "grad_norm": 0.0,
-      "learning_rate": 1.316206654675245e-07,
-      "loss": 0.8391,
+      "learning_rate": 1.8114495758700945e-08,
+      "loss": 0.7875,
       "step": 34585
     },
     {
-      "epoch": 0.9498777841860976,
+      "epoch": 0.9814415437003405,
       "grad_norm": 0.0,
-      "learning_rate": 1.3147685875953055e-07,
-      "loss": 0.7638,
+      "learning_rate": 1.8059243185097886e-08,
+      "loss": 0.8367,
       "step": 34586
     },
     {
-      "epoch": 0.9499052484139409,
+      "epoch": 0.9814699205448354,
       "grad_norm": 0.0,
-      "learning_rate": 1.3133313013511507e-07,
-      "loss": 0.8508,
+      "learning_rate": 1.8004074929397485e-08,
+      "loss": 0.7708,
       "step": 34587
     },
     {
-      "epoch": 0.949932712641784,
+      "epoch": 0.9814982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 1.3118947959541494e-07,
-      "loss": 0.8191,
+      "learning_rate": 1.794899099206604e-08,
+      "loss": 0.8482,
       "step": 34588
     },
     {
-      "epoch": 0.9499601768696273,
+      "epoch": 0.9815266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 1.3104590714156594e-07,
-      "loss": 0.82,
+      "learning_rate": 1.789399137356762e-08,
+      "loss": 0.7174,
       "step": 34589
     },
     {
-      "epoch": 0.9499876410974706,
+      "epoch": 0.9815550510783201,
       "grad_norm": 0.0,
-      "learning_rate": 1.3090241277470494e-07,
-      "loss": 0.7744,
+      "learning_rate": 1.783907607436741e-08,
+      "loss": 0.7806,
       "step": 34590
     },
     {
-      "epoch": 0.9500151053253137,
+      "epoch": 0.9815834279228149,
       "grad_norm": 0.0,
-      "learning_rate": 1.3075899649596767e-07,
-      "loss": 0.7411,
+      "learning_rate": 1.7784245094929488e-08,
+      "loss": 0.8259,
       "step": 34591
     },
     {
-      "epoch": 0.950042569553157,
+      "epoch": 0.9816118047673099,
       "grad_norm": 0.0,
-      "learning_rate": 1.306156583064866e-07,
-      "loss": 0.8326,
+      "learning_rate": 1.7729498435716808e-08,
+      "loss": 0.7845,
       "step": 34592
     },
     {
-      "epoch": 0.9500700337810003,
+      "epoch": 0.9816401816118048,
       "grad_norm": 0.0,
-      "learning_rate": 1.3047239820739855e-07,
-      "loss": 0.8,
+      "learning_rate": 1.767483609719123e-08,
+      "loss": 0.7695,
       "step": 34593
     },
     {
-      "epoch": 0.9500974980088435,
+      "epoch": 0.9816685584562996,
       "grad_norm": 0.0,
-      "learning_rate": 1.3032921619983486e-07,
-      "loss": 0.6774,
+      "learning_rate": 1.762025807981571e-08,
+      "loss": 0.9069,
       "step": 34594
     },
     {
-      "epoch": 0.9501249622366867,
+      "epoch": 0.9816969353007946,
       "grad_norm": 0.0,
-      "learning_rate": 1.3018611228493017e-07,
-      "loss": 0.9249,
+      "learning_rate": 1.7565764384049887e-08,
+      "loss": 0.7681,
       "step": 34595
     },
     {
-      "epoch": 0.9501524264645299,
+      "epoch": 0.9817253121452895,
       "grad_norm": 0.0,
-      "learning_rate": 1.300430864638158e-07,
-      "loss": 0.8076,
+      "learning_rate": 1.7511355010355614e-08,
+      "loss": 0.8414,
       "step": 34596
     },
     {
-      "epoch": 0.9501798906923732,
+      "epoch": 0.9817536889897843,
       "grad_norm": 0.0,
-      "learning_rate": 1.299001387376253e-07,
-      "loss": 0.7392,
+      "learning_rate": 1.7457029959191407e-08,
+      "loss": 0.8834,
       "step": 34597
     },
     {
-      "epoch": 0.9502073549202165,
+      "epoch": 0.9817820658342792,
       "grad_norm": 0.0,
-      "learning_rate": 1.2975726910748775e-07,
-      "loss": 0.7778,
+      "learning_rate": 1.740278923101579e-08,
+      "loss": 0.8295,
       "step": 34598
     },
     {
-      "epoch": 0.9502348191480596,
+      "epoch": 0.9818104426787742,
       "grad_norm": 0.0,
-      "learning_rate": 1.296144775745334e-07,
-      "loss": 0.7615,
+      "learning_rate": 1.7348632826288403e-08,
+      "loss": 0.8392,
       "step": 34599
     },
     {
-      "epoch": 0.9502622833759029,
+      "epoch": 0.981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 1.2947176413989236e-07,
-      "loss": 0.7902,
+      "learning_rate": 1.729456074546554e-08,
+      "loss": 0.8966,
       "step": 34600
     },
     {
-      "epoch": 0.9502897476037461,
+      "epoch": 0.9818671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 1.2932912880469495e-07,
-      "loss": 0.7624,
+      "learning_rate": 1.7240572989003502e-08,
+      "loss": 0.7687,
       "step": 34601
     },
     {
-      "epoch": 0.9503172118315893,
+      "epoch": 0.9818955732122588,
       "grad_norm": 0.0,
-      "learning_rate": 1.2918657157006798e-07,
-      "loss": 0.8102,
+      "learning_rate": 1.7186669557360812e-08,
+      "loss": 0.8363,
       "step": 34602
     },
     {
-      "epoch": 0.9503446760594326,
+      "epoch": 0.9819239500567537,
       "grad_norm": 0.0,
-      "learning_rate": 1.2904409243714168e-07,
-      "loss": 0.8727,
+      "learning_rate": 1.7132850450989336e-08,
+      "loss": 0.7934,
       "step": 34603
     },
     {
-      "epoch": 0.9503721402872758,
+      "epoch": 0.9819523269012486,
       "grad_norm": 0.0,
-      "learning_rate": 1.2890169140704178e-07,
-      "loss": 0.8132,
+      "learning_rate": 1.707911567034648e-08,
+      "loss": 0.7268,
       "step": 34604
     },
     {
-      "epoch": 0.9503996045151191,
+      "epoch": 0.9819807037457434,
       "grad_norm": 0.0,
-      "learning_rate": 1.287593684808952e-07,
-      "loss": 0.8782,
+      "learning_rate": 1.7025465215885217e-08,
+      "loss": 0.8109,
       "step": 34605
     },
     {
-      "epoch": 0.9504270687429623,
+      "epoch": 0.9820090805902384,
       "grad_norm": 0.0,
-      "learning_rate": 1.286171236598288e-07,
-      "loss": 0.7685,
+      "learning_rate": 1.697189908805741e-08,
+      "loss": 0.805,
       "step": 34606
     },
     {
-      "epoch": 0.9504545329708055,
+      "epoch": 0.9820374574347333,
       "grad_norm": 0.0,
-      "learning_rate": 1.2847495694496726e-07,
-      "loss": 0.8328,
+      "learning_rate": 1.6918417287318245e-08,
+      "loss": 0.8225,
       "step": 34607
     },
     {
-      "epoch": 0.9504819971986488,
+      "epoch": 0.9820658342792281,
       "grad_norm": 0.0,
-      "learning_rate": 1.283328683374363e-07,
-      "loss": 0.8322,
+      "learning_rate": 1.6865019814117368e-08,
+      "loss": 0.7698,
       "step": 34608
     },
     {
-      "epoch": 0.950509461426492,
+      "epoch": 0.9820942111237231,
       "grad_norm": 0.0,
-      "learning_rate": 1.2819085783835837e-07,
-      "loss": 0.8945,
+      "learning_rate": 1.6811706668905526e-08,
+      "loss": 0.6847,
       "step": 34609
     },
     {
-      "epoch": 0.9505369256543352,
+      "epoch": 0.9821225879682179,
       "grad_norm": 0.0,
-      "learning_rate": 1.2804892544885817e-07,
-      "loss": 0.7453,
+      "learning_rate": 1.6758477852135692e-08,
+      "loss": 0.7203,
       "step": 34610
     },
     {
-      "epoch": 0.9505643898821785,
+      "epoch": 0.9821509648127128,
       "grad_norm": 0.0,
-      "learning_rate": 1.2790707117005918e-07,
-      "loss": 0.8013,
+      "learning_rate": 1.6705333364254174e-08,
+      "loss": 0.7677,
       "step": 34611
     },
     {
-      "epoch": 0.9505918541100217,
+      "epoch": 0.9821793416572078,
       "grad_norm": 0.0,
-      "learning_rate": 1.2776529500308388e-07,
-      "loss": 0.808,
+      "learning_rate": 1.6652273205712834e-08,
+      "loss": 0.8156,
       "step": 34612
     },
     {
-      "epoch": 0.950619318337865,
+      "epoch": 0.9822077185017026,
       "grad_norm": 0.0,
-      "learning_rate": 1.2762359694905467e-07,
-      "loss": 0.7317,
+      "learning_rate": 1.6599297376957978e-08,
+      "loss": 0.8329,
       "step": 34613
     },
     {
-      "epoch": 0.9506467825657081,
+      "epoch": 0.9822360953461975,
       "grad_norm": 0.0,
-      "learning_rate": 1.274819770090907e-07,
-      "loss": 0.8884,
+      "learning_rate": 1.6546405878437033e-08,
+      "loss": 0.7645,
       "step": 34614
     },
     {
-      "epoch": 0.9506742467935514,
+      "epoch": 0.9822644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 1.2734043518431439e-07,
-      "loss": 0.8307,
+      "learning_rate": 1.6493598710598524e-08,
+      "loss": 0.7917,
       "step": 34615
     },
     {
-      "epoch": 0.9507017110213947,
+      "epoch": 0.9822928490351873,
       "grad_norm": 0.0,
-      "learning_rate": 1.271989714758437e-07,
-      "loss": 0.8876,
+      "learning_rate": 1.644087587388654e-08,
+      "loss": 0.7962,
       "step": 34616
     },
     {
-      "epoch": 0.9507291752492378,
+      "epoch": 0.9823212258796822,
       "grad_norm": 0.0,
-      "learning_rate": 1.270575858847989e-07,
-      "loss": 0.7316,
+      "learning_rate": 1.6388237368747396e-08,
+      "loss": 0.8282,
       "step": 34617
     },
     {
-      "epoch": 0.9507566394770811,
+      "epoch": 0.982349602724177,
       "grad_norm": 0.0,
-      "learning_rate": 1.2691627841230014e-07,
-      "loss": 0.81,
+      "learning_rate": 1.6335683195626286e-08,
+      "loss": 0.8111,
       "step": 34618
     },
     {
-      "epoch": 0.9507841037049244,
+      "epoch": 0.982377979568672,
       "grad_norm": 0.0,
-      "learning_rate": 1.2677504905946326e-07,
-      "loss": 0.7511,
+      "learning_rate": 1.6283213354965077e-08,
+      "loss": 0.7232,
       "step": 34619
     },
     {
-      "epoch": 0.9508115679327676,
+      "epoch": 0.9824063564131669,
       "grad_norm": 0.0,
-      "learning_rate": 1.2663389782740732e-07,
-      "loss": 0.9026,
+      "learning_rate": 1.6230827847208974e-08,
+      "loss": 0.834,
       "step": 34620
     },
     {
-      "epoch": 0.9508390321606108,
+      "epoch": 0.9824347332576617,
       "grad_norm": 0.0,
-      "learning_rate": 1.2649282471724811e-07,
-      "loss": 0.7376,
+      "learning_rate": 1.6178526672799845e-08,
+      "loss": 0.7854,
       "step": 34621
     },
     {
-      "epoch": 0.950866496388454,
+      "epoch": 0.9824631101021566,
       "grad_norm": 0.0,
-      "learning_rate": 1.2635182973010363e-07,
-      "loss": 0.7775,
+      "learning_rate": 1.6126309832180665e-08,
+      "loss": 0.8597,
       "step": 34622
     },
     {
-      "epoch": 0.9508939606162973,
+      "epoch": 0.9824914869466516,
       "grad_norm": 0.0,
-      "learning_rate": 1.262109128670863e-07,
-      "loss": 0.8733,
+      "learning_rate": 1.6074177325789974e-08,
+      "loss": 0.8446,
       "step": 34623
     },
     {
-      "epoch": 0.9509214248441406,
+      "epoch": 0.9825198637911464,
       "grad_norm": 0.0,
-      "learning_rate": 1.2607007412931416e-07,
-      "loss": 0.8075,
+      "learning_rate": 1.6022129154069643e-08,
+      "loss": 0.8297,
       "step": 34624
     },
     {
-      "epoch": 0.9509488890719837,
+      "epoch": 0.9825482406356413,
       "grad_norm": 0.0,
-      "learning_rate": 1.2592931351789962e-07,
-      "loss": 0.7658,
+      "learning_rate": 1.5970165317460428e-08,
+      "loss": 0.7689,
       "step": 34625
     },
     {
-      "epoch": 0.950976353299827,
+      "epoch": 0.9825766174801362,
       "grad_norm": 0.0,
-      "learning_rate": 1.2578863103395734e-07,
-      "loss": 0.7876,
+      "learning_rate": 1.591828581639865e-08,
+      "loss": 0.8297,
       "step": 34626
     },
     {
-      "epoch": 0.9510038175276702,
+      "epoch": 0.9826049943246311,
       "grad_norm": 0.0,
-      "learning_rate": 1.2564802667860087e-07,
-      "loss": 0.8096,
+      "learning_rate": 1.5866490651323952e-08,
+      "loss": 0.8332,
       "step": 34627
     },
     {
-      "epoch": 0.9510312817555134,
+      "epoch": 0.982633371169126,
       "grad_norm": 0.0,
-      "learning_rate": 1.2550750045294158e-07,
-      "loss": 0.7013,
+      "learning_rate": 1.5814779822674876e-08,
+      "loss": 0.7677,
       "step": 34628
     },
     {
-      "epoch": 0.9510587459833567,
+      "epoch": 0.9826617480136209,
       "grad_norm": 0.0,
-      "learning_rate": 1.25367052358093e-07,
-      "loss": 0.7963,
+      "learning_rate": 1.5763153330886627e-08,
+      "loss": 0.8217,
       "step": 34629
     },
     {
-      "epoch": 0.9510862102111999,
+      "epoch": 0.9826901248581158,
       "grad_norm": 0.0,
-      "learning_rate": 1.2522668239516421e-07,
-      "loss": 0.8563,
+      "learning_rate": 1.5711611176395525e-08,
+      "loss": 0.7661,
       "step": 34630
     },
     {
-      "epoch": 0.9511136744390432,
+      "epoch": 0.9827185017026107,
       "grad_norm": 0.0,
-      "learning_rate": 1.2508639056526661e-07,
-      "loss": 0.7231,
+      "learning_rate": 1.5660153359637886e-08,
+      "loss": 0.754,
       "step": 34631
     },
     {
-      "epoch": 0.9511411386668864,
+      "epoch": 0.9827468785471055,
       "grad_norm": 0.0,
-      "learning_rate": 1.249461768695126e-07,
-      "loss": 0.759,
+      "learning_rate": 1.560877988104781e-08,
+      "loss": 0.7832,
       "step": 34632
     },
     {
-      "epoch": 0.9511686028947296,
+      "epoch": 0.9827752553916005,
       "grad_norm": 0.0,
-      "learning_rate": 1.2480604130900797e-07,
-      "loss": 0.7805,
+      "learning_rate": 1.555749074105828e-08,
+      "loss": 0.9121,
       "step": 34633
     },
     {
-      "epoch": 0.9511960671225729,
+      "epoch": 0.9828036322360953,
       "grad_norm": 0.0,
-      "learning_rate": 1.2466598388486406e-07,
-      "loss": 0.8528,
+      "learning_rate": 1.5506285940103394e-08,
+      "loss": 0.784,
       "step": 34634
     },
     {
-      "epoch": 0.951223531350416,
+      "epoch": 0.9828320090805902,
       "grad_norm": 0.0,
-      "learning_rate": 1.2452600459818774e-07,
-      "loss": 0.8251,
+      "learning_rate": 1.545516547861614e-08,
+      "loss": 0.7716,
       "step": 34635
     },
     {
-      "epoch": 0.9512509955782593,
+      "epoch": 0.9828603859250852,
       "grad_norm": 0.0,
-      "learning_rate": 1.2438610345008707e-07,
-      "loss": 0.8853,
+      "learning_rate": 1.5404129357028396e-08,
+      "loss": 0.8202,
       "step": 34636
     },
     {
-      "epoch": 0.9512784598061026,
+      "epoch": 0.98288876276958,
       "grad_norm": 0.0,
-      "learning_rate": 1.2424628044166997e-07,
-      "loss": 0.8305,
+      "learning_rate": 1.535317757577093e-08,
+      "loss": 0.7809,
       "step": 34637
     },
     {
-      "epoch": 0.9513059240339458,
+      "epoch": 0.9829171396140749,
       "grad_norm": 0.0,
-      "learning_rate": 1.2410653557404118e-07,
-      "loss": 0.8578,
+      "learning_rate": 1.530231013527339e-08,
+      "loss": 0.8309,
       "step": 34638
     },
     {
-      "epoch": 0.951333388261789,
+      "epoch": 0.9829455164585698,
       "grad_norm": 0.0,
-      "learning_rate": 1.2396686884830645e-07,
-      "loss": 0.7336,
+      "learning_rate": 1.5251527035966552e-08,
+      "loss": 0.8307,
       "step": 34639
     },
     {
-      "epoch": 0.9513608524896322,
+      "epoch": 0.9829738933030647,
       "grad_norm": 0.0,
-      "learning_rate": 1.2382728026557156e-07,
-      "loss": 0.7288,
+      "learning_rate": 1.5200828278278957e-08,
+      "loss": 0.876,
       "step": 34640
     },
     {
-      "epoch": 0.9513883167174755,
+      "epoch": 0.9830022701475596,
       "grad_norm": 0.0,
-      "learning_rate": 1.236877698269412e-07,
-      "loss": 0.8046,
+      "learning_rate": 1.5150213862638042e-08,
+      "loss": 0.8437,
       "step": 34641
     },
     {
-      "epoch": 0.9514157809453188,
+      "epoch": 0.9830306469920544,
       "grad_norm": 0.0,
-      "learning_rate": 1.2354833753351892e-07,
-      "loss": 0.7492,
+      "learning_rate": 1.5099683789473463e-08,
+      "loss": 0.8141,
       "step": 34642
     },
     {
-      "epoch": 0.9514432451731619,
+      "epoch": 0.9830590238365494,
       "grad_norm": 0.0,
-      "learning_rate": 1.2340898338640829e-07,
-      "loss": 0.7969,
+      "learning_rate": 1.5049238059209325e-08,
+      "loss": 0.847,
       "step": 34643
     },
     {
-      "epoch": 0.9514707094010052,
+      "epoch": 0.9830874006810443,
       "grad_norm": 0.0,
-      "learning_rate": 1.2326970738671174e-07,
-      "loss": 0.7749,
+      "learning_rate": 1.4998876672274176e-08,
+      "loss": 0.6931,
       "step": 34644
     },
     {
-      "epoch": 0.9514981736288485,
+      "epoch": 0.9831157775255391,
       "grad_norm": 0.0,
-      "learning_rate": 1.2313050953552952e-07,
-      "loss": 0.7336,
+      "learning_rate": 1.494859962909101e-08,
+      "loss": 0.7797,
       "step": 34645
     },
     {
-      "epoch": 0.9515256378566916,
+      "epoch": 0.9831441543700341,
       "grad_norm": 0.0,
-      "learning_rate": 1.229913898339674e-07,
-      "loss": 0.7519,
+      "learning_rate": 1.4898406930087262e-08,
+      "loss": 0.8365,
       "step": 34646
     },
     {
-      "epoch": 0.9515531020845349,
+      "epoch": 0.983172531214529,
       "grad_norm": 0.0,
-      "learning_rate": 1.228523482831212e-07,
-      "loss": 0.744,
+      "learning_rate": 1.4848298575684817e-08,
+      "loss": 0.8669,
       "step": 34647
     },
     {
-      "epoch": 0.9515805663123781,
+      "epoch": 0.9832009080590238,
       "grad_norm": 0.0,
-      "learning_rate": 1.2271338488409335e-07,
-      "loss": 0.737,
+      "learning_rate": 1.479827456630778e-08,
+      "loss": 0.8658,
       "step": 34648
     },
     {
-      "epoch": 0.9516080305402214,
+      "epoch": 0.9832292849035187,
       "grad_norm": 0.0,
-      "learning_rate": 1.2257449963798406e-07,
-      "loss": 0.7931,
+      "learning_rate": 1.4748334902379147e-08,
+      "loss": 0.8841,
       "step": 34649
     },
     {
-      "epoch": 0.9516354947680646,
+      "epoch": 0.9832576617480137,
       "grad_norm": 0.0,
-      "learning_rate": 1.224356925458914e-07,
-      "loss": 0.8167,
+      "learning_rate": 1.469847958431858e-08,
+      "loss": 0.8149,
       "step": 34650
     },
     {
-      "epoch": 0.9516629589959078,
+      "epoch": 0.9832860385925085,
       "grad_norm": 0.0,
-      "learning_rate": 1.222969636089133e-07,
-      "loss": 0.8343,
+      "learning_rate": 1.4648708612550189e-08,
+      "loss": 0.806,
       "step": 34651
     },
     {
-      "epoch": 0.9516904232237511,
+      "epoch": 0.9833144154370034,
       "grad_norm": 0.0,
-      "learning_rate": 1.2215831282814784e-07,
-      "loss": 0.7711,
+      "learning_rate": 1.4599021987493634e-08,
+      "loss": 0.7402,
       "step": 34652
     },
     {
-      "epoch": 0.9517178874515942,
+      "epoch": 0.9833427922814983,
       "grad_norm": 0.0,
-      "learning_rate": 1.22019740204693e-07,
-      "loss": 0.8306,
+      "learning_rate": 1.4549419709566359e-08,
+      "loss": 0.8671,
       "step": 34653
     },
     {
-      "epoch": 0.9517453516794375,
+      "epoch": 0.9833711691259932,
       "grad_norm": 0.0,
-      "learning_rate": 1.2188124573964454e-07,
-      "loss": 0.8064,
+      "learning_rate": 1.4499901779190251e-08,
+      "loss": 0.761,
       "step": 34654
     },
     {
-      "epoch": 0.9517728159072808,
+      "epoch": 0.9833995459704881,
       "grad_norm": 0.0,
-      "learning_rate": 1.217428294340961e-07,
-      "loss": 0.7818,
+      "learning_rate": 1.4450468196781641e-08,
+      "loss": 0.7523,
       "step": 34655
     },
     {
-      "epoch": 0.951800280135124,
+      "epoch": 0.9834279228149829,
       "grad_norm": 0.0,
-      "learning_rate": 1.2160449128914676e-07,
-      "loss": 0.8034,
+      "learning_rate": 1.4401118962759086e-08,
+      "loss": 0.8568,
       "step": 34656
     },
     {
-      "epoch": 0.9518277443629672,
+      "epoch": 0.9834562996594779,
       "grad_norm": 0.0,
-      "learning_rate": 1.2146623130588898e-07,
-      "loss": 0.7243,
+      "learning_rate": 1.4351854077540028e-08,
+      "loss": 0.7248,
       "step": 34657
     },
     {
-      "epoch": 0.9518552085908105,
+      "epoch": 0.9834846765039728,
       "grad_norm": 0.0,
-      "learning_rate": 1.2132804948541632e-07,
-      "loss": 0.7753,
+      "learning_rate": 1.4302673541538581e-08,
+      "loss": 0.8013,
       "step": 34658
     },
     {
-      "epoch": 0.9518826728186537,
+      "epoch": 0.9835130533484676,
       "grad_norm": 0.0,
-      "learning_rate": 1.2118994582882238e-07,
-      "loss": 0.7713,
+      "learning_rate": 1.4253577355171078e-08,
+      "loss": 0.8008,
       "step": 34659
     },
     {
-      "epoch": 0.951910137046497,
+      "epoch": 0.9835414301929626,
       "grad_norm": 0.0,
-      "learning_rate": 1.210519203372007e-07,
-      "loss": 0.7648,
+      "learning_rate": 1.4204565518853853e-08,
+      "loss": 0.8412,
       "step": 34660
     },
     {
-      "epoch": 0.9519376012743401,
+      "epoch": 0.9835698070374574,
       "grad_norm": 0.0,
-      "learning_rate": 1.2091397301164377e-07,
-      "loss": 0.855,
+      "learning_rate": 1.4155638032998797e-08,
+      "loss": 0.8195,
       "step": 34661
     },
     {
-      "epoch": 0.9519650655021834,
+      "epoch": 0.9835981838819523,
       "grad_norm": 0.0,
-      "learning_rate": 1.2077610385324178e-07,
-      "loss": 0.863,
+      "learning_rate": 1.4106794898020027e-08,
+      "loss": 0.8974,
       "step": 34662
     },
     {
-      "epoch": 0.9519925297300267,
+      "epoch": 0.9836265607264473,
       "grad_norm": 0.0,
-      "learning_rate": 1.2063831286308615e-07,
-      "loss": 0.8,
+      "learning_rate": 1.4058036114329433e-08,
+      "loss": 0.7608,
       "step": 34663
     },
     {
-      "epoch": 0.9520199939578698,
+      "epoch": 0.9836549375709421,
       "grad_norm": 0.0,
-      "learning_rate": 1.2050060004226706e-07,
-      "loss": 0.841,
+      "learning_rate": 1.4009361682340018e-08,
+      "loss": 0.7309,
       "step": 34664
     },
     {
-      "epoch": 0.9520474581857131,
+      "epoch": 0.983683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.2036296539187363e-07,
-      "loss": 0.8334,
+      "learning_rate": 1.3960771602462565e-08,
+      "loss": 0.861,
       "step": 34665
     },
     {
-      "epoch": 0.9520749224135563,
+      "epoch": 0.9837116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.2022540891299506e-07,
-      "loss": 0.7387,
+      "learning_rate": 1.3912265875106745e-08,
+      "loss": 0.7731,
       "step": 34666
     },
     {
-      "epoch": 0.9521023866413996,
+      "epoch": 0.9837400681044268,
       "grad_norm": 0.0,
-      "learning_rate": 1.2008793060672151e-07,
-      "loss": 0.7552,
+      "learning_rate": 1.3863844500683343e-08,
+      "loss": 0.7338,
       "step": 34667
     },
     {
-      "epoch": 0.9521298508692428,
+      "epoch": 0.9837684449489217,
       "grad_norm": 0.0,
-      "learning_rate": 1.1995053047413884e-07,
-      "loss": 0.7931,
+      "learning_rate": 1.381550747960203e-08,
+      "loss": 0.7288,
       "step": 34668
     },
     {
-      "epoch": 0.952157315097086,
+      "epoch": 0.9837968217934165,
       "grad_norm": 0.0,
-      "learning_rate": 1.1981320851633392e-07,
-      "loss": 0.7848,
+      "learning_rate": 1.3767254812269148e-08,
+      "loss": 0.8812,
       "step": 34669
     },
     {
-      "epoch": 0.9521847793249293,
+      "epoch": 0.9838251986379115,
       "grad_norm": 0.0,
-      "learning_rate": 1.196759647343959e-07,
-      "loss": 0.832,
+      "learning_rate": 1.3719086499092149e-08,
+      "loss": 0.864,
       "step": 34670
     },
     {
-      "epoch": 0.9522122435527726,
+      "epoch": 0.9838535754824064,
       "grad_norm": 0.0,
-      "learning_rate": 1.1953879912940724e-07,
-      "loss": 0.8369,
+      "learning_rate": 1.3671002540480705e-08,
+      "loss": 0.8755,
       "step": 34671
     },
     {
-      "epoch": 0.9522397077806157,
+      "epoch": 0.9838819523269012,
       "grad_norm": 0.0,
-      "learning_rate": 1.1940171170245595e-07,
-      "loss": 0.8841,
+      "learning_rate": 1.3623002936837825e-08,
+      "loss": 0.7069,
       "step": 34672
     },
     {
-      "epoch": 0.952267172008459,
+      "epoch": 0.9839103291713961,
       "grad_norm": 0.0,
-      "learning_rate": 1.1926470245462562e-07,
-      "loss": 0.7806,
+      "learning_rate": 1.3575087688570965e-08,
+      "loss": 0.9315,
       "step": 34673
     },
     {
-      "epoch": 0.9522946362363022,
+      "epoch": 0.9839387060158911,
       "grad_norm": 0.0,
-      "learning_rate": 1.1912777138699982e-07,
-      "loss": 0.7802,
+      "learning_rate": 1.3527256796084243e-08,
+      "loss": 0.8537,
       "step": 34674
     },
     {
-      "epoch": 0.9523221004641454,
+      "epoch": 0.9839670828603859,
       "grad_norm": 0.0,
-      "learning_rate": 1.1899091850066213e-07,
-      "loss": 0.8222,
+      "learning_rate": 1.347951025978178e-08,
+      "loss": 0.8025,
       "step": 34675
     },
     {
-      "epoch": 0.9523495646919887,
+      "epoch": 0.9839954597048808,
       "grad_norm": 0.0,
-      "learning_rate": 1.1885414379669612e-07,
-      "loss": 0.8312,
+      "learning_rate": 1.343184808006659e-08,
+      "loss": 0.7972,
       "step": 34676
     },
     {
-      "epoch": 0.9523770289198319,
+      "epoch": 0.9840238365493758,
       "grad_norm": 0.0,
-      "learning_rate": 1.187174472761854e-07,
-      "loss": 0.8188,
+      "learning_rate": 1.338427025734168e-08,
+      "loss": 0.8228,
       "step": 34677
     },
     {
-      "epoch": 0.9524044931476752,
+      "epoch": 0.9840522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 1.1858082894020795e-07,
-      "loss": 0.8527,
+      "learning_rate": 1.3336776792008954e-08,
+      "loss": 0.7471,
       "step": 34678
     },
     {
-      "epoch": 0.9524319573755183,
+      "epoch": 0.9840805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 1.1844428878984737e-07,
-      "loss": 0.7644,
+      "learning_rate": 1.3289367684469201e-08,
+      "loss": 0.7637,
       "step": 34679
     },
     {
-      "epoch": 0.9524594216033616,
+      "epoch": 0.9841089670828603,
       "grad_norm": 0.0,
-      "learning_rate": 1.1830782682618391e-07,
-      "loss": 0.7975,
+      "learning_rate": 1.3242042935123211e-08,
+      "loss": 0.8831,
       "step": 34680
     },
     {
-      "epoch": 0.9524868858312049,
+      "epoch": 0.9841373439273553,
       "grad_norm": 0.0,
-      "learning_rate": 1.181714430502956e-07,
-      "loss": 0.7386,
+      "learning_rate": 1.3194802544370666e-08,
+      "loss": 0.8366,
       "step": 34681
     },
     {
-      "epoch": 0.952514350059048,
+      "epoch": 0.9841657207718502,
       "grad_norm": 0.0,
-      "learning_rate": 1.1803513746326267e-07,
-      "loss": 0.7992,
+      "learning_rate": 1.3147646512610135e-08,
+      "loss": 0.685,
       "step": 34682
     },
     {
-      "epoch": 0.9525418142868913,
+      "epoch": 0.984194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 1.1789891006616428e-07,
-      "loss": 0.7856,
+      "learning_rate": 1.3100574840240187e-08,
+      "loss": 0.8071,
       "step": 34683
     },
     {
-      "epoch": 0.9525692785147346,
+      "epoch": 0.98422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 1.1776276086007732e-07,
-      "loss": 0.7706,
+      "learning_rate": 1.3053587527659394e-08,
+      "loss": 0.8213,
       "step": 34684
     },
     {
-      "epoch": 0.9525967427425778,
+      "epoch": 0.9842508513053349,
       "grad_norm": 0.0,
-      "learning_rate": 1.1762668984607983e-07,
-      "loss": 0.8152,
+      "learning_rate": 1.3006684575264106e-08,
+      "loss": 0.7943,
       "step": 34685
     },
     {
-      "epoch": 0.952624206970421,
+      "epoch": 0.9842792281498297,
       "grad_norm": 0.0,
-      "learning_rate": 1.1749069702524873e-07,
-      "loss": 0.7257,
+      "learning_rate": 1.295986598344956e-08,
+      "loss": 0.8591,
       "step": 34686
     },
     {
-      "epoch": 0.9526516711982642,
+      "epoch": 0.9843076049943247,
       "grad_norm": 0.0,
-      "learning_rate": 1.173547823986576e-07,
-      "loss": 0.7221,
+      "learning_rate": 1.2913131752612107e-08,
+      "loss": 0.6992,
       "step": 34687
     },
     {
-      "epoch": 0.9526791354261075,
+      "epoch": 0.9843359818388195,
       "grad_norm": 0.0,
-      "learning_rate": 1.1721894596738448e-07,
-      "loss": 0.8204,
+      "learning_rate": 1.2866481883146986e-08,
+      "loss": 0.8479,
       "step": 34688
     },
     {
-      "epoch": 0.9527065996539508,
+      "epoch": 0.9843643586833144,
       "grad_norm": 0.0,
-      "learning_rate": 1.1708318773250293e-07,
-      "loss": 0.7921,
+      "learning_rate": 1.2819916375446106e-08,
+      "loss": 0.8671,
       "step": 34689
     },
     {
-      "epoch": 0.9527340638817939,
+      "epoch": 0.9843927355278093,
       "grad_norm": 0.0,
-      "learning_rate": 1.1694750769508767e-07,
-      "loss": 0.873,
+      "learning_rate": 1.2773435229905818e-08,
+      "loss": 0.8338,
       "step": 34690
     },
     {
-      "epoch": 0.9527615281096372,
+      "epoch": 0.9844211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 1.1681190585621227e-07,
-      "loss": 0.7416,
+      "learning_rate": 1.2727038446916918e-08,
+      "loss": 0.7684,
       "step": 34691
     },
     {
-      "epoch": 0.9527889923374804,
+      "epoch": 0.9844494892167991,
       "grad_norm": 0.0,
-      "learning_rate": 1.1667638221694921e-07,
-      "loss": 0.8039,
+      "learning_rate": 1.2680726026871314e-08,
+      "loss": 0.7876,
       "step": 34692
     },
     {
-      "epoch": 0.9528164565653237,
+      "epoch": 0.984477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 1.1654093677837098e-07,
-      "loss": 0.767,
+      "learning_rate": 1.263449797016092e-08,
+      "loss": 0.722,
       "step": 34693
     },
     {
-      "epoch": 0.9528439207931669,
+      "epoch": 0.9845062429057889,
       "grad_norm": 0.0,
-      "learning_rate": 1.1640556954155003e-07,
-      "loss": 0.8013,
+      "learning_rate": 1.2588354277176529e-08,
+      "loss": 0.7051,
       "step": 34694
     },
     {
-      "epoch": 0.9528713850210101,
+      "epoch": 0.9845346197502838,
       "grad_norm": 0.0,
-      "learning_rate": 1.1627028050755663e-07,
-      "loss": 0.8351,
+      "learning_rate": 1.2542294948305611e-08,
+      "loss": 0.8579,
       "step": 34695
     },
     {
-      "epoch": 0.9528988492488534,
+      "epoch": 0.9845629965947786,
       "grad_norm": 0.0,
-      "learning_rate": 1.1613506967746103e-07,
-      "loss": 0.8909,
+      "learning_rate": 1.2496319983940074e-08,
+      "loss": 0.7624,
       "step": 34696
     },
     {
-      "epoch": 0.9529263134766967,
+      "epoch": 0.9845913734392735,
       "grad_norm": 0.0,
-      "learning_rate": 1.1599993705233348e-07,
-      "loss": 0.6931,
+      "learning_rate": 1.2450429384467388e-08,
+      "loss": 0.8185,
       "step": 34697
     },
     {
-      "epoch": 0.9529537777045398,
+      "epoch": 0.9846197502837685,
       "grad_norm": 0.0,
-      "learning_rate": 1.1586488263324314e-07,
-      "loss": 0.7849,
+      "learning_rate": 1.2404623150273908e-08,
+      "loss": 0.708,
       "step": 34698
     },
     {
-      "epoch": 0.9529812419323831,
+      "epoch": 0.9846481271282633,
       "grad_norm": 0.0,
-      "learning_rate": 1.1572990642125803e-07,
-      "loss": 0.8857,
+      "learning_rate": 1.2358901281748215e-08,
+      "loss": 0.8056,
       "step": 34699
     },
     {
-      "epoch": 0.9530087061602263,
+      "epoch": 0.9846765039727582,
       "grad_norm": 0.0,
-      "learning_rate": 1.1559500841744731e-07,
-      "loss": 0.7809,
+      "learning_rate": 1.2313263779275553e-08,
+      "loss": 0.7614,
       "step": 34700
     },
     {
-      "epoch": 0.9530361703880695,
+      "epoch": 0.9847048808172532,
       "grad_norm": 0.0,
-      "learning_rate": 1.1546018862287789e-07,
-      "loss": 0.7664,
+      "learning_rate": 1.226771064324228e-08,
+      "loss": 0.7856,
       "step": 34701
     },
     {
-      "epoch": 0.9530636346159128,
+      "epoch": 0.984733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 1.1532544703861559e-07,
-      "loss": 0.751,
+      "learning_rate": 1.2222241874031427e-08,
+      "loss": 0.7902,
       "step": 34702
     },
     {
-      "epoch": 0.953091098843756,
+      "epoch": 0.9847616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 1.1519078366572844e-07,
-      "loss": 0.6834,
+      "learning_rate": 1.2176857472029347e-08,
+      "loss": 0.8182,
       "step": 34703
     },
     {
-      "epoch": 0.9531185630715993,
+      "epoch": 0.9847900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 1.1505619850528005e-07,
-      "loss": 0.7182,
+      "learning_rate": 1.2131557437617958e-08,
+      "loss": 0.7868,
       "step": 34704
     },
     {
-      "epoch": 0.9531460272994424,
+      "epoch": 0.9848183881952327,
       "grad_norm": 0.0,
-      "learning_rate": 1.149216915583351e-07,
-      "loss": 0.8526,
+      "learning_rate": 1.2086341771180288e-08,
+      "loss": 0.836,
       "step": 34705
     },
     {
-      "epoch": 0.9531734915272857,
+      "epoch": 0.9848467650397276,
       "grad_norm": 0.0,
-      "learning_rate": 1.1478726282596053e-07,
-      "loss": 0.7222,
+      "learning_rate": 1.2041210473098253e-08,
+      "loss": 0.7805,
       "step": 34706
     },
     {
-      "epoch": 0.953200955755129,
+      "epoch": 0.9848751418842224,
       "grad_norm": 0.0,
-      "learning_rate": 1.146529123092166e-07,
-      "loss": 0.8146,
+      "learning_rate": 1.199616354375377e-08,
+      "loss": 0.7172,
       "step": 34707
     },
     {
-      "epoch": 0.9532284199829721,
+      "epoch": 0.9849035187287174,
       "grad_norm": 0.0,
-      "learning_rate": 1.1451864000916913e-07,
-      "loss": 0.8739,
+      "learning_rate": 1.1951200983526534e-08,
+      "loss": 0.8845,
       "step": 34708
     },
     {
-      "epoch": 0.9532558842108154,
+      "epoch": 0.9849318955732123,
       "grad_norm": 0.0,
-      "learning_rate": 1.1438444592688059e-07,
-      "loss": 0.8305,
+      "learning_rate": 1.1906322792795132e-08,
+      "loss": 0.805,
       "step": 34709
     },
     {
-      "epoch": 0.9532833484386587,
+      "epoch": 0.9849602724177071,
       "grad_norm": 0.0,
-      "learning_rate": 1.1425033006341013e-07,
-      "loss": 0.8289,
+      "learning_rate": 1.1861528971941482e-08,
+      "loss": 0.7938,
       "step": 34710
     },
     {
-      "epoch": 0.9533108126665019,
+      "epoch": 0.9849886492622021,
       "grad_norm": 0.0,
-      "learning_rate": 1.1411629241982136e-07,
-      "loss": 0.809,
+      "learning_rate": 1.181681952134195e-08,
+      "loss": 0.8474,
       "step": 34711
     },
     {
-      "epoch": 0.9533382768943451,
+      "epoch": 0.985017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 1.1398233299717342e-07,
-      "loss": 0.7151,
+      "learning_rate": 1.1772194441374008e-08,
+      "loss": 0.6924,
       "step": 34712
     },
     {
-      "epoch": 0.9533657411221883,
+      "epoch": 0.9850454029511918,
       "grad_norm": 0.0,
-      "learning_rate": 1.1384845179652548e-07,
-      "loss": 0.8228,
+      "learning_rate": 1.1727653732416245e-08,
+      "loss": 0.788,
       "step": 34713
     },
     {
-      "epoch": 0.9533932053500316,
+      "epoch": 0.9850737797956867,
       "grad_norm": 0.0,
-      "learning_rate": 1.1371464881894e-07,
-      "loss": 0.8729,
+      "learning_rate": 1.1683197394843915e-08,
+      "loss": 0.7589,
       "step": 34714
     },
     {
-      "epoch": 0.9534206695778749,
+      "epoch": 0.9851021566401816,
       "grad_norm": 0.0,
-      "learning_rate": 1.1358092406547283e-07,
-      "loss": 0.7809,
+      "learning_rate": 1.1638825429033384e-08,
+      "loss": 0.8047,
       "step": 34715
     },
     {
-      "epoch": 0.953448133805718,
+      "epoch": 0.9851305334846765,
       "grad_norm": 0.0,
-      "learning_rate": 1.134472775371831e-07,
-      "loss": 0.8304,
+      "learning_rate": 1.1594537835357689e-08,
+      "loss": 0.7352,
       "step": 34716
     },
     {
-      "epoch": 0.9534755980335613,
+      "epoch": 0.9851589103291714,
       "grad_norm": 0.0,
-      "learning_rate": 1.1331370923512885e-07,
-      "loss": 0.7511,
+      "learning_rate": 1.1550334614192082e-08,
+      "loss": 0.7872,
       "step": 34717
     },
     {
-      "epoch": 0.9535030622614045,
+      "epoch": 0.9851872871736663,
       "grad_norm": 0.0,
-      "learning_rate": 1.1318021916036593e-07,
-      "loss": 0.8317,
+      "learning_rate": 1.15062157659096e-08,
+      "loss": 0.8119,
       "step": 34718
     },
     {
-      "epoch": 0.9535305264892477,
+      "epoch": 0.9852156640181612,
       "grad_norm": 0.0,
-      "learning_rate": 1.1304680731395012e-07,
-      "loss": 0.7707,
+      "learning_rate": 1.1462181290883279e-08,
+      "loss": 0.8228,
       "step": 34719
     },
     {
-      "epoch": 0.953557990717091,
+      "epoch": 0.985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 1.1291347369693839e-07,
-      "loss": 0.8061,
+      "learning_rate": 1.141823118948504e-08,
+      "loss": 0.8234,
       "step": 34720
     },
     {
-      "epoch": 0.9535854549449342,
+      "epoch": 0.985272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 1.1278021831038655e-07,
-      "loss": 0.8125,
+      "learning_rate": 1.137436546208681e-08,
+      "loss": 0.7854,
       "step": 34721
     },
     {
-      "epoch": 0.9536129191727775,
+      "epoch": 0.9853007945516459,
       "grad_norm": 0.0,
-      "learning_rate": 1.1264704115534597e-07,
-      "loss": 0.7803,
+      "learning_rate": 1.1330584109057185e-08,
+      "loss": 0.7711,
       "step": 34722
     },
     {
-      "epoch": 0.9536403834006207,
+      "epoch": 0.9853291713961407,
       "grad_norm": 0.0,
-      "learning_rate": 1.1251394223287138e-07,
-      "loss": 0.8331,
+      "learning_rate": 1.1286887130766977e-08,
+      "loss": 0.851,
       "step": 34723
     },
     {
-      "epoch": 0.9536678476284639,
+      "epoch": 0.9853575482406356,
       "grad_norm": 0.0,
-      "learning_rate": 1.1238092154401747e-07,
-      "loss": 0.8616,
+      "learning_rate": 1.1243274527587001e-08,
+      "loss": 0.7662,
       "step": 34724
     },
     {
-      "epoch": 0.9536953118563072,
+      "epoch": 0.9853859250851306,
       "grad_norm": 0.0,
-      "learning_rate": 1.1224797908983675e-07,
-      "loss": 0.76,
+      "learning_rate": 1.1199746299882518e-08,
+      "loss": 0.6891,
       "step": 34725
     },
     {
-      "epoch": 0.9537227760841503,
+      "epoch": 0.9854143019296254,
       "grad_norm": 0.0,
-      "learning_rate": 1.1211511487137838e-07,
-      "loss": 0.8455,
+      "learning_rate": 1.1156302448023237e-08,
+      "loss": 0.7844,
       "step": 34726
     },
     {
-      "epoch": 0.9537502403119936,
+      "epoch": 0.9854426787741203,
       "grad_norm": 0.0,
-      "learning_rate": 1.1198232888969707e-07,
-      "loss": 0.84,
+      "learning_rate": 1.111294297237664e-08,
+      "loss": 0.6228,
       "step": 34727
     },
     {
-      "epoch": 0.9537777045398369,
+      "epoch": 0.9854710556186153,
       "grad_norm": 0.0,
-      "learning_rate": 1.1184962114584086e-07,
-      "loss": 0.8707,
+      "learning_rate": 1.1069667873306878e-08,
+      "loss": 0.867,
       "step": 34728
     },
     {
-      "epoch": 0.9538051687676801,
+      "epoch": 0.9854994324631101,
       "grad_norm": 0.0,
-      "learning_rate": 1.1171699164086003e-07,
-      "loss": 0.8671,
+      "learning_rate": 1.1026477151180325e-08,
+      "loss": 0.8443,
       "step": 34729
     },
     {
-      "epoch": 0.9538326329955233,
+      "epoch": 0.985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 1.1158444037580596e-07,
-      "loss": 0.7797,
+      "learning_rate": 1.0983370806363359e-08,
+      "loss": 0.7537,
       "step": 34730
     },
     {
-      "epoch": 0.9538600972233665,
+      "epoch": 0.9855561861520998,
       "grad_norm": 0.0,
-      "learning_rate": 1.1145196735172558e-07,
-      "loss": 0.8134,
+      "learning_rate": 1.0940348839219018e-08,
+      "loss": 0.8326,
       "step": 34731
     },
     {
-      "epoch": 0.9538875614512098,
+      "epoch": 0.9855845629965948,
       "grad_norm": 0.0,
-      "learning_rate": 1.1131957256966808e-07,
-      "loss": 0.7513,
+      "learning_rate": 1.0897411250109235e-08,
+      "loss": 0.7856,
       "step": 34732
     },
     {
-      "epoch": 0.9539150256790531,
+      "epoch": 0.9856129398410897,
       "grad_norm": 0.0,
-      "learning_rate": 1.1118725603068037e-07,
-      "loss": 0.7754,
+      "learning_rate": 1.0854558039399276e-08,
+      "loss": 0.8528,
       "step": 34733
     },
     {
-      "epoch": 0.9539424899068962,
+      "epoch": 0.9856413166855845,
       "grad_norm": 0.0,
-      "learning_rate": 1.1105501773581051e-07,
-      "loss": 0.7847,
+      "learning_rate": 1.0811789207448853e-08,
+      "loss": 0.7157,
       "step": 34734
     },
     {
-      "epoch": 0.9539699541347395,
+      "epoch": 0.9856696935300795,
       "grad_norm": 0.0,
-      "learning_rate": 1.1092285768610212e-07,
-      "loss": 0.735,
+      "learning_rate": 1.0769104754621006e-08,
+      "loss": 0.7547,
       "step": 34735
     },
     {
-      "epoch": 0.9539974183625828,
+      "epoch": 0.9856980703745744,
       "grad_norm": 0.0,
-      "learning_rate": 1.1079077588260434e-07,
-      "loss": 0.818,
+      "learning_rate": 1.0726504681275452e-08,
+      "loss": 0.8409,
       "step": 34736
     },
     {
-      "epoch": 0.954024882590426,
+      "epoch": 0.9857264472190692,
       "grad_norm": 0.0,
-      "learning_rate": 1.1065877232635968e-07,
-      "loss": 0.7919,
+      "learning_rate": 1.0683988987773008e-08,
+      "loss": 0.8161,
       "step": 34737
     },
     {
-      "epoch": 0.9540523468182692,
+      "epoch": 0.9857548240635642,
       "grad_norm": 0.0,
-      "learning_rate": 1.1052684701841399e-07,
-      "loss": 0.7678,
+      "learning_rate": 1.064155767447006e-08,
+      "loss": 0.7941,
       "step": 34738
     },
     {
-      "epoch": 0.9540798110461124,
+      "epoch": 0.985783200908059,
       "grad_norm": 0.0,
-      "learning_rate": 1.1039499995981084e-07,
-      "loss": 0.7585,
+      "learning_rate": 1.0599210741728538e-08,
+      "loss": 0.737,
       "step": 34739
     },
     {
-      "epoch": 0.9541072752739557,
+      "epoch": 0.9858115777525539,
       "grad_norm": 0.0,
-      "learning_rate": 1.1026323115159276e-07,
-      "loss": 0.8259,
+      "learning_rate": 1.0556948189903715e-08,
+      "loss": 0.7322,
       "step": 34740
     },
     {
-      "epoch": 0.954134739501799,
+      "epoch": 0.9858399545970488,
       "grad_norm": 0.0,
-      "learning_rate": 1.1013154059480335e-07,
-      "loss": 0.8713,
+      "learning_rate": 1.0514770019354193e-08,
+      "loss": 0.7336,
       "step": 34741
     },
     {
-      "epoch": 0.9541622037296421,
+      "epoch": 0.9858683314415437,
       "grad_norm": 0.0,
-      "learning_rate": 1.0999992829048511e-07,
-      "loss": 0.8289,
+      "learning_rate": 1.0472676230435242e-08,
+      "loss": 0.6986,
       "step": 34742
     },
     {
-      "epoch": 0.9541896679574854,
+      "epoch": 0.9858967082860386,
       "grad_norm": 0.0,
-      "learning_rate": 1.0986839423967721e-07,
-      "loss": 0.8542,
+      "learning_rate": 1.0430666823502134e-08,
+      "loss": 0.9004,
       "step": 34743
     },
     {
-      "epoch": 0.9542171321853287,
+      "epoch": 0.9859250851305335,
       "grad_norm": 0.0,
-      "learning_rate": 1.0973693844342214e-07,
-      "loss": 0.7898,
+      "learning_rate": 1.038874179891014e-08,
+      "loss": 0.831,
       "step": 34744
     },
     {
-      "epoch": 0.9542445964131718,
+      "epoch": 0.9859534619750284,
       "grad_norm": 0.0,
-      "learning_rate": 1.0960556090275909e-07,
-      "loss": 0.7512,
+      "learning_rate": 1.0346901157014532e-08,
+      "loss": 0.782,
       "step": 34745
     },
     {
-      "epoch": 0.9542720606410151,
+      "epoch": 0.9859818388195233,
       "grad_norm": 0.0,
-      "learning_rate": 1.0947426161872832e-07,
-      "loss": 0.7332,
+      "learning_rate": 1.0305144898166142e-08,
+      "loss": 0.7109,
       "step": 34746
     },
     {
-      "epoch": 0.9542995248688583,
+      "epoch": 0.9860102156640181,
       "grad_norm": 0.0,
-      "learning_rate": 1.0934304059236789e-07,
-      "loss": 0.7781,
+      "learning_rate": 1.0263473022720239e-08,
+      "loss": 0.8573,
       "step": 34747
     },
     {
-      "epoch": 0.9543269890967016,
+      "epoch": 0.986038592508513,
       "grad_norm": 0.0,
-      "learning_rate": 1.0921189782471697e-07,
-      "loss": 0.8296,
+      "learning_rate": 1.0221885531027653e-08,
+      "loss": 0.9096,
       "step": 34748
     },
     {
-      "epoch": 0.9543544533245448,
+      "epoch": 0.986066969353008,
       "grad_norm": 0.0,
-      "learning_rate": 1.0908083331681363e-07,
-      "loss": 0.8221,
+      "learning_rate": 1.0180382423440327e-08,
+      "loss": 0.7496,
       "step": 34749
     },
     {
-      "epoch": 0.954381917552388,
+      "epoch": 0.9860953461975028,
       "grad_norm": 0.0,
-      "learning_rate": 1.0894984706969369e-07,
-      "loss": 0.7504,
+      "learning_rate": 1.013896370030798e-08,
+      "loss": 0.8148,
       "step": 34750
     },
     {
-      "epoch": 0.9544093817802313,
+      "epoch": 0.9861237230419977,
       "grad_norm": 0.0,
-      "learning_rate": 1.0881893908439301e-07,
-      "loss": 0.7742,
+      "learning_rate": 1.0097629361981442e-08,
+      "loss": 0.7682,
       "step": 34751
     },
     {
-      "epoch": 0.9544368460080744,
+      "epoch": 0.9861520998864927,
       "grad_norm": 0.0,
-      "learning_rate": 1.0868810936194962e-07,
-      "loss": 0.7951,
+      "learning_rate": 1.0056379408808214e-08,
+      "loss": 0.653,
       "step": 34752
     },
     {
-      "epoch": 0.9544643102359177,
+      "epoch": 0.9861804767309875,
       "grad_norm": 0.0,
-      "learning_rate": 1.0855735790339605e-07,
-      "loss": 0.6956,
+      "learning_rate": 1.0015213841139126e-08,
+      "loss": 0.8098,
       "step": 34753
     },
     {
-      "epoch": 0.954491774463761,
+      "epoch": 0.9862088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.0842668470976925e-07,
-      "loss": 0.8996,
+      "learning_rate": 9.974132659319457e-09,
+      "loss": 0.8175,
       "step": 34754
     },
     {
-      "epoch": 0.9545192386916042,
+      "epoch": 0.9862372304199774,
       "grad_norm": 0.0,
-      "learning_rate": 1.0829608978210172e-07,
-      "loss": 0.8232,
+      "learning_rate": 9.933135863697818e-09,
+      "loss": 0.808,
       "step": 34755
     },
     {
-      "epoch": 0.9545467029194474,
+      "epoch": 0.9862656072644722,
       "grad_norm": 0.0,
-      "learning_rate": 1.0816557312142817e-07,
-      "loss": 0.7595,
+      "learning_rate": 9.892223454620597e-09,
+      "loss": 0.8142,
       "step": 34756
     },
     {
-      "epoch": 0.9545741671472907,
+      "epoch": 0.9862939841089671,
       "grad_norm": 0.0,
-      "learning_rate": 1.0803513472877891e-07,
-      "loss": 0.7955,
+      "learning_rate": 9.851395432431965e-09,
+      "loss": 0.7806,
       "step": 34757
     },
     {
-      "epoch": 0.9546016313751339,
+      "epoch": 0.9863223609534619,
       "grad_norm": 0.0,
-      "learning_rate": 1.0790477460518867e-07,
-      "loss": 0.7557,
+      "learning_rate": 9.81065179747831e-09,
+      "loss": 0.8434,
       "step": 34758
     },
     {
-      "epoch": 0.9546290956029772,
+      "epoch": 0.9863507377979569,
       "grad_norm": 0.0,
-      "learning_rate": 1.0777449275168772e-07,
-      "loss": 0.7929,
+      "learning_rate": 9.769992550102692e-09,
+      "loss": 0.7824,
       "step": 34759
     },
     {
-      "epoch": 0.9546565598308203,
+      "epoch": 0.9863791146424518,
       "grad_norm": 0.0,
-      "learning_rate": 1.0764428916930525e-07,
-      "loss": 0.8319,
+      "learning_rate": 9.729417690649279e-09,
+      "loss": 0.8035,
       "step": 34760
     },
     {
-      "epoch": 0.9546840240586636,
+      "epoch": 0.9864074914869466,
       "grad_norm": 0.0,
-      "learning_rate": 1.0751416385907487e-07,
-      "loss": 0.7976,
+      "learning_rate": 9.688927219460021e-09,
+      "loss": 0.7984,
       "step": 34761
     },
     {
-      "epoch": 0.9547114882865069,
+      "epoch": 0.9864358683314416,
       "grad_norm": 0.0,
-      "learning_rate": 1.0738411682202355e-07,
-      "loss": 0.7872,
+      "learning_rate": 9.648521136877975e-09,
+      "loss": 0.7244,
       "step": 34762
     },
     {
-      "epoch": 0.95473895251435,
+      "epoch": 0.9864642451759364,
       "grad_norm": 0.0,
-      "learning_rate": 1.0725414805918155e-07,
-      "loss": 0.7575,
+      "learning_rate": 9.608199443243981e-09,
+      "loss": 0.7151,
       "step": 34763
     },
     {
-      "epoch": 0.9547664167421933,
+      "epoch": 0.9864926220204313,
       "grad_norm": 0.0,
-      "learning_rate": 1.0712425757157585e-07,
-      "loss": 0.8113,
+      "learning_rate": 9.56796213889888e-09,
+      "loss": 0.8316,
       "step": 34764
     },
     {
-      "epoch": 0.9547938809700365,
+      "epoch": 0.9865209988649262,
       "grad_norm": 0.0,
-      "learning_rate": 1.0699444536023562e-07,
-      "loss": 0.8268,
+      "learning_rate": 9.527809224182393e-09,
+      "loss": 0.7531,
       "step": 34765
     },
     {
-      "epoch": 0.9548213451978798,
+      "epoch": 0.9865493757094211,
       "grad_norm": 0.0,
-      "learning_rate": 1.0686471142618892e-07,
-      "loss": 0.7392,
+      "learning_rate": 9.487740699433145e-09,
+      "loss": 0.7522,
       "step": 34766
     },
     {
-      "epoch": 0.954848809425723,
+      "epoch": 0.986577752553916,
       "grad_norm": 0.0,
-      "learning_rate": 1.0673505577045939e-07,
-      "loss": 0.889,
+      "learning_rate": 9.447756564990863e-09,
+      "loss": 0.8646,
       "step": 34767
     },
     {
-      "epoch": 0.9548762736535662,
+      "epoch": 0.9866061293984109,
       "grad_norm": 0.0,
-      "learning_rate": 1.0660547839407508e-07,
-      "loss": 0.798,
+      "learning_rate": 9.40785682119194e-09,
+      "loss": 0.7509,
       "step": 34768
     },
     {
-      "epoch": 0.9549037378814095,
+      "epoch": 0.9866345062429058,
       "grad_norm": 0.0,
-      "learning_rate": 1.0647597929805964e-07,
-      "loss": 0.767,
+      "learning_rate": 9.368041468372779e-09,
+      "loss": 0.8241,
       "step": 34769
     },
     {
-      "epoch": 0.9549312021092528,
+      "epoch": 0.9866628830874007,
       "grad_norm": 0.0,
-      "learning_rate": 1.0634655848343888e-07,
-      "loss": 0.7767,
+      "learning_rate": 9.328310506873106e-09,
+      "loss": 0.761,
       "step": 34770
     },
     {
-      "epoch": 0.9549586663370959,
+      "epoch": 0.9866912599318955,
       "grad_norm": 0.0,
-      "learning_rate": 1.0621721595123647e-07,
-      "loss": 0.8277,
+      "learning_rate": 9.288663937024877e-09,
+      "loss": 0.9023,
       "step": 34771
     },
     {
-      "epoch": 0.9549861305649392,
+      "epoch": 0.9867196367763905,
       "grad_norm": 0.0,
-      "learning_rate": 1.060879517024771e-07,
-      "loss": 0.8592,
+      "learning_rate": 9.249101759164492e-09,
+      "loss": 0.7686,
       "step": 34772
     },
     {
-      "epoch": 0.9550135947927824,
+      "epoch": 0.9867480136208854,
       "grad_norm": 0.0,
-      "learning_rate": 1.059587657381822e-07,
-      "loss": 0.8291,
+      "learning_rate": 9.209623973626126e-09,
+      "loss": 0.774,
       "step": 34773
     },
     {
-      "epoch": 0.9550410590206256,
+      "epoch": 0.9867763904653802,
       "grad_norm": 0.0,
-      "learning_rate": 1.0582965805937318e-07,
-      "loss": 0.8274,
+      "learning_rate": 9.170230580742844e-09,
+      "loss": 0.7359,
       "step": 34774
     },
     {
-      "epoch": 0.9550685232484689,
+      "epoch": 0.9868047673098751,
       "grad_norm": 0.0,
-      "learning_rate": 1.0570062866707254e-07,
-      "loss": 0.7582,
+      "learning_rate": 9.130921580848829e-09,
+      "loss": 0.8737,
       "step": 34775
     },
     {
-      "epoch": 0.9550959874763121,
+      "epoch": 0.9868331441543701,
       "grad_norm": 0.0,
-      "learning_rate": 1.0557167756230169e-07,
-      "loss": 0.7424,
+      "learning_rate": 9.091696974273812e-09,
+      "loss": 0.8223,
       "step": 34776
     },
     {
-      "epoch": 0.9551234517041554,
+      "epoch": 0.9868615209988649,
       "grad_norm": 0.0,
-      "learning_rate": 1.0544280474608093e-07,
-      "loss": 0.8456,
+      "learning_rate": 9.052556761351972e-09,
+      "loss": 0.7975,
       "step": 34777
     },
     {
-      "epoch": 0.9551509159319985,
+      "epoch": 0.9868898978433598,
       "grad_norm": 0.0,
-      "learning_rate": 1.0531401021942833e-07,
-      "loss": 0.7888,
+      "learning_rate": 9.013500942410824e-09,
+      "loss": 0.8358,
       "step": 34778
     },
     {
-      "epoch": 0.9551783801598418,
+      "epoch": 0.9869182746878548,
       "grad_norm": 0.0,
-      "learning_rate": 1.0518529398336419e-07,
-      "loss": 0.8002,
+      "learning_rate": 8.974529517782327e-09,
+      "loss": 0.8293,
       "step": 34779
     },
     {
-      "epoch": 0.9552058443876851,
+      "epoch": 0.9869466515323496,
       "grad_norm": 0.0,
-      "learning_rate": 1.0505665603890769e-07,
-      "loss": 0.8522,
+      "learning_rate": 8.935642487795104e-09,
+      "loss": 0.8159,
       "step": 34780
     },
     {
-      "epoch": 0.9552333086155282,
+      "epoch": 0.9869750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 1.049280963870758e-07,
-      "loss": 0.8,
+      "learning_rate": 8.896839852777783e-09,
+      "loss": 0.8117,
       "step": 34781
     },
     {
-      "epoch": 0.9552607728433715,
+      "epoch": 0.9870034052213393,
       "grad_norm": 0.0,
-      "learning_rate": 1.0479961502888547e-07,
-      "loss": 0.7174,
+      "learning_rate": 8.85812161305788e-09,
+      "loss": 0.7773,
       "step": 34782
     },
     {
-      "epoch": 0.9552882370712148,
+      "epoch": 0.9870317820658343,
       "grad_norm": 0.0,
-      "learning_rate": 1.0467121196535368e-07,
-      "loss": 0.8467,
+      "learning_rate": 8.819487768961799e-09,
+      "loss": 0.7884,
       "step": 34783
     },
     {
-      "epoch": 0.955315701299058,
+      "epoch": 0.9870601589103292,
       "grad_norm": 0.0,
-      "learning_rate": 1.0454288719749628e-07,
-      "loss": 0.8741,
+      "learning_rate": 8.780938320817057e-09,
+      "loss": 0.8395,
       "step": 34784
     },
     {
-      "epoch": 0.9553431655269012,
+      "epoch": 0.987088535754824,
       "grad_norm": 0.0,
-      "learning_rate": 1.0441464072632911e-07,
-      "loss": 0.8215,
+      "learning_rate": 8.742473268950059e-09,
+      "loss": 0.8335,
       "step": 34785
     },
     {
-      "epoch": 0.9553706297547444,
+      "epoch": 0.987116912599319,
       "grad_norm": 0.0,
-      "learning_rate": 1.0428647255286583e-07,
-      "loss": 0.7651,
+      "learning_rate": 8.704092613682768e-09,
+      "loss": 0.8647,
       "step": 34786
     },
     {
-      "epoch": 0.9553980939825877,
+      "epoch": 0.9871452894438139,
       "grad_norm": 0.0,
-      "learning_rate": 1.0415838267812228e-07,
-      "loss": 0.7564,
+      "learning_rate": 8.665796355342703e-09,
+      "loss": 0.7576,
       "step": 34787
     },
     {
-      "epoch": 0.955425558210431,
+      "epoch": 0.9871736662883087,
       "grad_norm": 0.0,
-      "learning_rate": 1.0403037110310987e-07,
-      "loss": 0.7419,
+      "learning_rate": 8.627584494250719e-09,
+      "loss": 0.8165,
       "step": 34788
     },
     {
-      "epoch": 0.9554530224382741,
+      "epoch": 0.9872020431328037,
       "grad_norm": 0.0,
-      "learning_rate": 1.0390243782884335e-07,
-      "loss": 0.7565,
+      "learning_rate": 8.589457030730997e-09,
+      "loss": 0.8238,
       "step": 34789
     },
     {
-      "epoch": 0.9554804866661174,
+      "epoch": 0.9872304199772985,
       "grad_norm": 0.0,
-      "learning_rate": 1.0377458285633302e-07,
-      "loss": 0.7948,
+      "learning_rate": 8.551413965105505e-09,
+      "loss": 0.7544,
       "step": 34790
     },
     {
-      "epoch": 0.9555079508939606,
+      "epoch": 0.9872587968217934,
       "grad_norm": 0.0,
-      "learning_rate": 1.036468061865925e-07,
-      "loss": 0.7482,
+      "learning_rate": 8.513455297695095e-09,
+      "loss": 0.8112,
       "step": 34791
     },
     {
-      "epoch": 0.9555354151218038,
+      "epoch": 0.9872871736662883,
       "grad_norm": 0.0,
-      "learning_rate": 1.0351910782063102e-07,
-      "loss": 0.7033,
+      "learning_rate": 8.475581028820623e-09,
+      "loss": 0.849,
       "step": 34792
     },
     {
-      "epoch": 0.9555628793496471,
+      "epoch": 0.9873155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 1.0339148775946107e-07,
-      "loss": 0.7428,
+      "learning_rate": 8.437791158801833e-09,
+      "loss": 0.7329,
       "step": 34793
     },
     {
-      "epoch": 0.9555903435774903,
+      "epoch": 0.9873439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 1.0326394600408963e-07,
-      "loss": 0.7468,
+      "learning_rate": 8.40008568795847e-09,
+      "loss": 0.8339,
       "step": 34794
     },
     {
-      "epoch": 0.9556178078053336,
+      "epoch": 0.987372304199773,
       "grad_norm": 0.0,
-      "learning_rate": 1.0313648255552921e-07,
-      "loss": 0.7341,
+      "learning_rate": 8.362464616609168e-09,
+      "loss": 0.8076,
       "step": 34795
     },
     {
-      "epoch": 0.9556452720331768,
+      "epoch": 0.9874006810442679,
       "grad_norm": 0.0,
-      "learning_rate": 1.0300909741478571e-07,
-      "loss": 0.8565,
+      "learning_rate": 8.324927945070337e-09,
+      "loss": 0.8047,
       "step": 34796
     },
     {
-      "epoch": 0.95567273626102,
+      "epoch": 0.9874290578887628,
       "grad_norm": 0.0,
-      "learning_rate": 1.0288179058286829e-07,
-      "loss": 0.8236,
+      "learning_rate": 8.287475673660617e-09,
+      "loss": 0.7888,
       "step": 34797
     },
     {
-      "epoch": 0.9557002004888633,
+      "epoch": 0.9874574347332576,
       "grad_norm": 0.0,
-      "learning_rate": 1.0275456206078394e-07,
-      "loss": 0.7473,
+      "learning_rate": 8.250107802696416e-09,
+      "loss": 0.9285,
       "step": 34798
     },
     {
-      "epoch": 0.9557276647167064,
+      "epoch": 0.9874858115777525,
       "grad_norm": 0.0,
-      "learning_rate": 1.0262741184953961e-07,
-      "loss": 0.7416,
+      "learning_rate": 8.212824332491931e-09,
+      "loss": 0.8379,
       "step": 34799
     },
     {
-      "epoch": 0.9557551289445497,
+      "epoch": 0.9875141884222475,
       "grad_norm": 0.0,
-      "learning_rate": 1.0250033995014008e-07,
-      "loss": 0.8529,
+      "learning_rate": 8.175625263362463e-09,
+      "loss": 0.8277,
       "step": 34800
     },
     {
-      "epoch": 0.955782593172393,
+      "epoch": 0.9875425652667423,
       "grad_norm": 0.0,
-      "learning_rate": 1.023733463635923e-07,
-      "loss": 0.8704,
+      "learning_rate": 8.138510595623317e-09,
+      "loss": 0.9031,
       "step": 34801
     },
     {
-      "epoch": 0.9558100574002362,
+      "epoch": 0.9875709421112372,
       "grad_norm": 0.0,
-      "learning_rate": 1.0224643109089994e-07,
-      "loss": 0.8314,
+      "learning_rate": 8.101480329587574e-09,
+      "loss": 0.8211,
       "step": 34802
     },
     {
-      "epoch": 0.9558375216280794,
+      "epoch": 0.9875993189557322,
       "grad_norm": 0.0,
-      "learning_rate": 1.0211959413306882e-07,
-      "loss": 0.7939,
+      "learning_rate": 8.06453446556721e-09,
+      "loss": 0.9397,
       "step": 34803
     },
     {
-      "epoch": 0.9558649858559226,
+      "epoch": 0.987627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 1.019928354911015e-07,
-      "loss": 0.864,
+      "learning_rate": 8.027673003875302e-09,
+      "loss": 0.7521,
       "step": 34804
     },
     {
-      "epoch": 0.9558924500837659,
+      "epoch": 0.9876560726447219,
       "grad_norm": 0.0,
-      "learning_rate": 1.0186615516600051e-07,
-      "loss": 0.8196,
+      "learning_rate": 7.990895944821609e-09,
+      "loss": 0.8217,
       "step": 34805
     },
     {
-      "epoch": 0.9559199143116092,
+      "epoch": 0.9876844494892169,
       "grad_norm": 0.0,
-      "learning_rate": 1.0173955315876949e-07,
-      "loss": 0.715,
+      "learning_rate": 7.954203288719209e-09,
+      "loss": 0.7532,
       "step": 34806
     },
     {
-      "epoch": 0.9559473785394523,
+      "epoch": 0.9877128263337117,
       "grad_norm": 0.0,
-      "learning_rate": 1.0161302947040874e-07,
-      "loss": 0.8608,
+      "learning_rate": 7.917595035876746e-09,
+      "loss": 0.7685,
       "step": 34807
     },
     {
-      "epoch": 0.9559748427672956,
+      "epoch": 0.9877412031782066,
       "grad_norm": 0.0,
-      "learning_rate": 1.0148658410191969e-07,
-      "loss": 0.7935,
+      "learning_rate": 7.881071186602863e-09,
+      "loss": 0.7686,
       "step": 34808
     },
     {
-      "epoch": 0.9560023069951389,
+      "epoch": 0.9877695800227014,
       "grad_norm": 0.0,
-      "learning_rate": 1.0136021705430377e-07,
-      "loss": 0.7863,
+      "learning_rate": 7.84463174120731e-09,
+      "loss": 0.7693,
       "step": 34809
     },
     {
-      "epoch": 0.956029771222982,
+      "epoch": 0.9877979568671964,
       "grad_norm": 0.0,
-      "learning_rate": 1.0123392832855905e-07,
-      "loss": 0.8016,
+      "learning_rate": 7.80827669999762e-09,
+      "loss": 0.8303,
       "step": 34810
     },
     {
-      "epoch": 0.9560572354508253,
+      "epoch": 0.9878263337116913,
       "grad_norm": 0.0,
-      "learning_rate": 1.0110771792568697e-07,
-      "loss": 0.8118,
+      "learning_rate": 7.772006063280214e-09,
+      "loss": 0.8235,
       "step": 34811
     },
     {
-      "epoch": 0.9560846996786685,
+      "epoch": 0.9878547105561861,
       "grad_norm": 0.0,
-      "learning_rate": 1.009815858466845e-07,
-      "loss": 0.8269,
+      "learning_rate": 7.735819831361513e-09,
+      "loss": 0.7691,
       "step": 34812
     },
     {
-      "epoch": 0.9561121639065118,
+      "epoch": 0.9878830874006811,
       "grad_norm": 0.0,
-      "learning_rate": 1.0085553209255084e-07,
+      "learning_rate": 7.69971800454794e-09,
       "loss": 0.7897,
       "step": 34813
     },
     {
-      "epoch": 0.956139628134355,
+      "epoch": 0.987911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 1.0072955666428186e-07,
-      "loss": 0.833,
+      "learning_rate": 7.663700583144806e-09,
+      "loss": 0.8728,
       "step": 34814
     },
     {
-      "epoch": 0.9561670923621982,
+      "epoch": 0.9879398410896708,
       "grad_norm": 0.0,
-      "learning_rate": 1.0060365956287677e-07,
-      "loss": 0.8071,
+      "learning_rate": 7.627767567454092e-09,
+      "loss": 0.7962,
       "step": 34815
     },
     {
-      "epoch": 0.9561945565900415,
+      "epoch": 0.9879682179341657,
       "grad_norm": 0.0,
-      "learning_rate": 1.0047784078932921e-07,
-      "loss": 0.7711,
+      "learning_rate": 7.591918957782219e-09,
+      "loss": 0.8728,
       "step": 34816
     },
     {
-      "epoch": 0.9562220208178847,
+      "epoch": 0.9879965947786606,
       "grad_norm": 0.0,
-      "learning_rate": 1.0035210034463505e-07,
-      "loss": 0.7951,
+      "learning_rate": 7.556154754428946e-09,
+      "loss": 0.8706,
       "step": 34817
     },
     {
-      "epoch": 0.9562494850457279,
+      "epoch": 0.9880249716231555,
       "grad_norm": 0.0,
-      "learning_rate": 1.0022643822979017e-07,
-      "loss": 0.8686,
+      "learning_rate": 7.520474957699586e-09,
+      "loss": 0.7683,
       "step": 34818
     },
     {
-      "epoch": 0.9562769492735712,
+      "epoch": 0.9880533484676504,
       "grad_norm": 0.0,
-      "learning_rate": 1.001008544457882e-07,
-      "loss": 0.7355,
+      "learning_rate": 7.48487956789279e-09,
+      "loss": 0.8426,
       "step": 34819
     },
     {
-      "epoch": 0.9563044135014144,
+      "epoch": 0.9880817253121453,
       "grad_norm": 0.0,
-      "learning_rate": 9.997534899362394e-08,
-      "loss": 0.8599,
+      "learning_rate": 7.449368585311645e-09,
+      "loss": 0.7284,
       "step": 34820
     },
     {
-      "epoch": 0.9563318777292577,
+      "epoch": 0.9881101021566402,
       "grad_norm": 0.0,
-      "learning_rate": 9.984992187428987e-08,
-      "loss": 0.8496,
+      "learning_rate": 7.4139420102536944e-09,
+      "loss": 0.8666,
       "step": 34821
     },
     {
-      "epoch": 0.9563593419571009,
+      "epoch": 0.988138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 9.972457308877858e-08,
-      "loss": 0.83,
+      "learning_rate": 7.378599843019807e-09,
+      "loss": 0.7048,
       "step": 34822
     },
     {
-      "epoch": 0.9563868061849441,
+      "epoch": 0.98816685584563,
       "grad_norm": 0.0,
-      "learning_rate": 9.959930263808038e-08,
-      "loss": 0.8179,
+      "learning_rate": 7.343342083908634e-09,
+      "loss": 0.8311,
       "step": 34823
     },
     {
-      "epoch": 0.9564142704127874,
+      "epoch": 0.9881952326901249,
       "grad_norm": 0.0,
-      "learning_rate": 9.947411052318779e-08,
-      "loss": 0.7941,
+      "learning_rate": 7.308168733216603e-09,
+      "loss": 0.758,
       "step": 34824
     },
     {
-      "epoch": 0.9564417346406305,
+      "epoch": 0.9882236095346197,
       "grad_norm": 0.0,
-      "learning_rate": 9.934899674509001e-08,
-      "loss": 0.8093,
+      "learning_rate": 7.273079791242366e-09,
+      "loss": 0.7932,
       "step": 34825
     },
     {
-      "epoch": 0.9564691988684738,
+      "epoch": 0.9882519863791146,
       "grad_norm": 0.0,
-      "learning_rate": 9.922396130477963e-08,
-      "loss": 0.8215,
+      "learning_rate": 7.238075258281241e-09,
+      "loss": 0.804,
       "step": 34826
     },
     {
-      "epoch": 0.9564966630963171,
+      "epoch": 0.9882803632236096,
       "grad_norm": 0.0,
-      "learning_rate": 9.90990042032447e-08,
-      "loss": 0.8347,
+      "learning_rate": 7.203155134629658e-09,
+      "loss": 0.8182,
       "step": 34827
     },
     {
-      "epoch": 0.9565241273241603,
+      "epoch": 0.9883087400681044,
       "grad_norm": 0.0,
-      "learning_rate": 9.897412544147222e-08,
-      "loss": 0.816,
+      "learning_rate": 7.168319420581826e-09,
+      "loss": 0.9338,
       "step": 34828
     },
     {
-      "epoch": 0.9565515915520035,
+      "epoch": 0.9883371169125993,
       "grad_norm": 0.0,
-      "learning_rate": 9.884932502045253e-08,
-      "loss": 0.8198,
+      "learning_rate": 7.1335681164319545e-09,
+      "loss": 0.8286,
       "step": 34829
     },
     {
-      "epoch": 0.9565790557798467,
+      "epoch": 0.9883654937570943,
       "grad_norm": 0.0,
-      "learning_rate": 9.872460294117259e-08,
-      "loss": 0.8128,
+      "learning_rate": 7.098901222474253e-09,
+      "loss": 0.797,
       "step": 34830
     },
     {
-      "epoch": 0.95660652000769,
+      "epoch": 0.9883938706015891,
       "grad_norm": 0.0,
-      "learning_rate": 9.85999592046194e-08,
-      "loss": 0.7463,
+      "learning_rate": 7.0643187390018184e-09,
+      "loss": 0.7218,
       "step": 34831
     },
     {
-      "epoch": 0.9566339842355333,
+      "epoch": 0.988422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 9.847539381177772e-08,
-      "loss": 0.7216,
+      "learning_rate": 7.029820666306641e-09,
+      "loss": 0.6681,
       "step": 34832
     },
     {
-      "epoch": 0.9566614484633764,
+      "epoch": 0.9884506242905788,
       "grad_norm": 0.0,
-      "learning_rate": 9.835090676363457e-08,
-      "loss": 0.8434,
+      "learning_rate": 6.9954070046796e-09,
+      "loss": 0.7828,
       "step": 34833
     },
     {
-      "epoch": 0.9566889126912197,
+      "epoch": 0.9884790011350738,
       "grad_norm": 0.0,
-      "learning_rate": 9.822649806117468e-08,
-      "loss": 0.7774,
+      "learning_rate": 6.961077754410461e-09,
+      "loss": 0.7428,
       "step": 34834
     },
     {
-      "epoch": 0.956716376919063,
+      "epoch": 0.9885073779795687,
       "grad_norm": 0.0,
-      "learning_rate": 9.810216770538172e-08,
-      "loss": 0.7598,
+      "learning_rate": 6.926832915791215e-09,
+      "loss": 0.7875,
       "step": 34835
     },
     {
-      "epoch": 0.9567438411469061,
+      "epoch": 0.9885357548240635,
       "grad_norm": 0.0,
-      "learning_rate": 9.797791569724047e-08,
-      "loss": 0.7609,
+      "learning_rate": 6.892672489110519e-09,
+      "loss": 0.8204,
       "step": 34836
     },
     {
-      "epoch": 0.9567713053747494,
+      "epoch": 0.9885641316685585,
       "grad_norm": 0.0,
-      "learning_rate": 9.785374203773456e-08,
-      "loss": 0.7099,
+      "learning_rate": 6.8585964746559206e-09,
+      "loss": 0.7514,
       "step": 34837
     },
     {
-      "epoch": 0.9567987696025926,
+      "epoch": 0.9885925085130534,
       "grad_norm": 0.0,
-      "learning_rate": 9.772964672784435e-08,
-      "loss": 0.7249,
+      "learning_rate": 6.824604872717188e-09,
+      "loss": 0.8242,
       "step": 34838
     },
     {
-      "epoch": 0.9568262338304359,
+      "epoch": 0.9886208853575482,
       "grad_norm": 0.0,
-      "learning_rate": 9.760562976855348e-08,
-      "loss": 0.8265,
+      "learning_rate": 6.79069768357854e-09,
+      "loss": 0.7869,
       "step": 34839
     },
     {
-      "epoch": 0.9568536980582791,
+      "epoch": 0.9886492622020431,
       "grad_norm": 0.0,
-      "learning_rate": 9.748169116084227e-08,
-      "loss": 0.7895,
+      "learning_rate": 6.756874907529743e-09,
+      "loss": 0.7916,
       "step": 34840
     },
     {
-      "epoch": 0.9568811622861223,
+      "epoch": 0.988677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 9.735783090569217e-08,
-      "loss": 0.7078,
+      "learning_rate": 6.723136544853903e-09,
+      "loss": 0.7317,
       "step": 34841
     },
     {
-      "epoch": 0.9569086265139656,
+      "epoch": 0.9887060158910329,
       "grad_norm": 0.0,
-      "learning_rate": 9.72340490040824e-08,
-      "loss": 0.8491,
+      "learning_rate": 6.6894825958363495e-09,
+      "loss": 0.8352,
       "step": 34842
     },
     {
-      "epoch": 0.9569360907418087,
+      "epoch": 0.9887343927355278,
       "grad_norm": 0.0,
-      "learning_rate": 9.711034545699328e-08,
-      "loss": 0.7454,
+      "learning_rate": 6.6559130607635196e-09,
+      "loss": 0.7977,
       "step": 34843
     },
     {
-      "epoch": 0.956963554969652,
+      "epoch": 0.9887627695800227,
       "grad_norm": 0.0,
-      "learning_rate": 9.69867202654029e-08,
-      "loss": 0.7359,
+      "learning_rate": 6.622427939916298e-09,
+      "loss": 0.7949,
       "step": 34844
     },
     {
-      "epoch": 0.9569910191974953,
+      "epoch": 0.9887911464245176,
       "grad_norm": 0.0,
-      "learning_rate": 9.686317343028939e-08,
-      "loss": 0.7254,
+      "learning_rate": 6.589027233578904e-09,
+      "loss": 0.601,
       "step": 34845
     },
     {
-      "epoch": 0.9570184834253385,
+      "epoch": 0.9888195232690125,
       "grad_norm": 0.0,
-      "learning_rate": 9.673970495263197e-08,
-      "loss": 0.7748,
+      "learning_rate": 6.555710942033333e-09,
+      "loss": 0.8125,
       "step": 34846
     },
     {
-      "epoch": 0.9570459476531817,
+      "epoch": 0.9888479001135074,
       "grad_norm": 0.0,
-      "learning_rate": 9.661631483340428e-08,
-      "loss": 0.8432,
+      "learning_rate": 6.522479065561582e-09,
+      "loss": 0.7974,
       "step": 34847
     },
     {
-      "epoch": 0.957073411881025,
+      "epoch": 0.9888762769580023,
       "grad_norm": 0.0,
-      "learning_rate": 9.649300307358555e-08,
-      "loss": 0.8356,
+      "learning_rate": 6.489331604442317e-09,
+      "loss": 0.7568,
       "step": 34848
     },
     {
-      "epoch": 0.9571008761088682,
+      "epoch": 0.9889046538024971,
       "grad_norm": 0.0,
-      "learning_rate": 9.636976967414946e-08,
-      "loss": 0.799,
+      "learning_rate": 6.456268558957534e-09,
+      "loss": 0.8465,
       "step": 34849
     },
     {
-      "epoch": 0.9571283403367115,
+      "epoch": 0.988933030646992,
       "grad_norm": 0.0,
-      "learning_rate": 9.62466146360741e-08,
-      "loss": 0.7789,
+      "learning_rate": 6.423289929385901e-09,
+      "loss": 0.6561,
       "step": 34850
     },
     {
-      "epoch": 0.9571558045645546,
+      "epoch": 0.988961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 9.612353796032981e-08,
-      "loss": 0.7626,
+      "learning_rate": 6.390395716006081e-09,
+      "loss": 0.76,
       "step": 34851
     },
     {
-      "epoch": 0.9571832687923979,
+      "epoch": 0.9889897843359818,
       "grad_norm": 0.0,
-      "learning_rate": 9.600053964789358e-08,
-      "loss": 0.7128,
+      "learning_rate": 6.3575859190956325e-09,
+      "loss": 0.7922,
       "step": 34852
     },
     {
-      "epoch": 0.9572107330202412,
+      "epoch": 0.9890181611804767,
       "grad_norm": 0.0,
-      "learning_rate": 9.587761969973686e-08,
-      "loss": 0.657,
+      "learning_rate": 6.32486053893211e-09,
+      "loss": 0.7789,
       "step": 34853
     },
     {
-      "epoch": 0.9572381972480843,
+      "epoch": 0.9890465380249717,
       "grad_norm": 0.0,
-      "learning_rate": 9.575477811683331e-08,
-      "loss": 0.8437,
+      "learning_rate": 6.292219575791958e-09,
+      "loss": 0.885,
       "step": 34854
     },
     {
-      "epoch": 0.9572656614759276,
+      "epoch": 0.9890749148694665,
       "grad_norm": 0.0,
-      "learning_rate": 9.56320149001544e-08,
-      "loss": 0.7061,
+      "learning_rate": 6.259663029950514e-09,
+      "loss": 0.8261,
       "step": 34855
     },
     {
-      "epoch": 0.9572931257037708,
+      "epoch": 0.9891032917139614,
       "grad_norm": 0.0,
-      "learning_rate": 9.550933005067154e-08,
-      "loss": 0.7802,
+      "learning_rate": 6.227190901683111e-09,
+      "loss": 0.8492,
       "step": 34856
     },
     {
-      "epoch": 0.9573205899316141,
+      "epoch": 0.9891316685584562,
       "grad_norm": 0.0,
-      "learning_rate": 9.538672356935619e-08,
-      "loss": 0.8885,
+      "learning_rate": 6.194803191262866e-09,
+      "loss": 0.8294,
       "step": 34857
     },
     {
-      "epoch": 0.9573480541594573,
+      "epoch": 0.9891600454029512,
       "grad_norm": 0.0,
-      "learning_rate": 9.526419545717647e-08,
-      "loss": 0.8824,
+      "learning_rate": 6.162499898966223e-09,
+      "loss": 0.7909,
       "step": 34858
     },
     {
-      "epoch": 0.9573755183873005,
+      "epoch": 0.9891884222474461,
       "grad_norm": 0.0,
-      "learning_rate": 9.514174571510381e-08,
-      "loss": 0.764,
+      "learning_rate": 6.1302810250629674e-09,
+      "loss": 0.7917,
       "step": 34859
     },
     {
-      "epoch": 0.9574029826151438,
+      "epoch": 0.9892167990919409,
       "grad_norm": 0.0,
-      "learning_rate": 9.501937434410635e-08,
-      "loss": 0.9089,
+      "learning_rate": 6.098146569826213e-09,
+      "loss": 0.7218,
       "step": 34860
     },
     {
-      "epoch": 0.9574304468429871,
+      "epoch": 0.9892451759364359,
       "grad_norm": 0.0,
-      "learning_rate": 9.489708134515218e-08,
-      "loss": 0.874,
+      "learning_rate": 6.066096533527965e-09,
+      "loss": 0.7416,
       "step": 34861
     },
     {
-      "epoch": 0.9574579110708302,
+      "epoch": 0.9892735527809308,
       "grad_norm": 0.0,
-      "learning_rate": 9.477486671920943e-08,
-      "loss": 0.7696,
+      "learning_rate": 6.034130916439118e-09,
+      "loss": 0.7796,
       "step": 34862
     },
     {
-      "epoch": 0.9574853752986735,
+      "epoch": 0.9893019296254256,
       "grad_norm": 0.0,
-      "learning_rate": 9.46527304672451e-08,
-      "loss": 0.8312,
+      "learning_rate": 6.0022497188283454e-09,
+      "loss": 0.7578,
       "step": 34863
     },
     {
-      "epoch": 0.9575128395265167,
+      "epoch": 0.9893303064699206,
       "grad_norm": 0.0,
-      "learning_rate": 9.453067259022508e-08,
-      "loss": 0.8762,
+      "learning_rate": 5.970452940966542e-09,
+      "loss": 0.8033,
       "step": 34864
     },
     {
-      "epoch": 0.95754030375436,
+      "epoch": 0.9893586833144155,
       "grad_norm": 0.0,
-      "learning_rate": 9.440869308911527e-08,
-      "loss": 0.8747,
+      "learning_rate": 5.9387405831212715e-09,
+      "loss": 0.8296,
       "step": 34865
     },
     {
-      "epoch": 0.9575677679822032,
+      "epoch": 0.9893870601589103,
       "grad_norm": 0.0,
-      "learning_rate": 9.428679196488155e-08,
-      "loss": 0.7778,
+      "learning_rate": 5.907112645561208e-09,
+      "loss": 0.8657,
       "step": 34866
     },
     {
-      "epoch": 0.9575952322100464,
+      "epoch": 0.9894154370034052,
       "grad_norm": 0.0,
-      "learning_rate": 9.41649692184865e-08,
-      "loss": 0.8001,
+      "learning_rate": 5.875569128552805e-09,
+      "loss": 0.7682,
       "step": 34867
     },
     {
-      "epoch": 0.9576226964378897,
+      "epoch": 0.9894438138479001,
       "grad_norm": 0.0,
-      "learning_rate": 9.404322485089712e-08,
-      "loss": 0.7552,
+      "learning_rate": 5.8441100323625154e-09,
+      "loss": 0.9414,
       "step": 34868
     },
     {
-      "epoch": 0.9576501606657328,
+      "epoch": 0.989472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 9.392155886307375e-08,
-      "loss": 0.9247,
+      "learning_rate": 5.812735357255683e-09,
+      "loss": 0.7657,
       "step": 34869
     },
     {
-      "epoch": 0.9576776248935761,
+      "epoch": 0.9895005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 9.379997125598117e-08,
-      "loss": 0.7542,
+      "learning_rate": 5.781445103498762e-09,
+      "loss": 0.862,
       "step": 34870
     },
     {
-      "epoch": 0.9577050891214194,
+      "epoch": 0.9895289443813848,
       "grad_norm": 0.0,
-      "learning_rate": 9.367846203057974e-08,
-      "loss": 0.7891,
+      "learning_rate": 5.750239271355984e-09,
+      "loss": 0.7878,
       "step": 34871
     },
     {
-      "epoch": 0.9577325533492625,
+      "epoch": 0.9895573212258797,
       "grad_norm": 0.0,
-      "learning_rate": 9.355703118783199e-08,
-      "loss": 0.7465,
+      "learning_rate": 5.719117861088253e-09,
+      "loss": 0.6572,
       "step": 34872
     },
     {
-      "epoch": 0.9577600175771058,
+      "epoch": 0.9895856980703746,
       "grad_norm": 0.0,
-      "learning_rate": 9.343567872869719e-08,
-      "loss": 0.7811,
+      "learning_rate": 5.6880808729620205e-09,
+      "loss": 0.863,
       "step": 34873
     },
     {
-      "epoch": 0.9577874818049491,
+      "epoch": 0.9896140749148694,
       "grad_norm": 0.0,
-      "learning_rate": 9.331440465413899e-08,
-      "loss": 0.8748,
+      "learning_rate": 5.657128307237081e-09,
+      "loss": 0.726,
       "step": 34874
     },
     {
-      "epoch": 0.9578149460327923,
+      "epoch": 0.9896424517593644,
       "grad_norm": 0.0,
-      "learning_rate": 9.31932089651133e-08,
-      "loss": 0.7011,
+      "learning_rate": 5.626260164176556e-09,
+      "loss": 0.7547,
       "step": 34875
     },
     {
-      "epoch": 0.9578424102606355,
+      "epoch": 0.9896708286038592,
       "grad_norm": 0.0,
-      "learning_rate": 9.307209166258047e-08,
-      "loss": 0.7798,
+      "learning_rate": 5.5954764440391275e-09,
+      "loss": 0.8437,
       "step": 34876
     },
     {
-      "epoch": 0.9578698744884787,
+      "epoch": 0.9896992054483541,
       "grad_norm": 0.0,
-      "learning_rate": 9.295105274749861e-08,
-      "loss": 0.8223,
+      "learning_rate": 5.564777147086809e-09,
+      "loss": 0.7551,
       "step": 34877
     },
     {
-      "epoch": 0.957897338716322,
+      "epoch": 0.9897275822928491,
       "grad_norm": 0.0,
-      "learning_rate": 9.283009222082695e-08,
-      "loss": 0.6539,
+      "learning_rate": 5.534162273578281e-09,
+      "loss": 0.7852,
       "step": 34878
     },
     {
-      "epoch": 0.9579248029441653,
+      "epoch": 0.9897559591373439,
       "grad_norm": 0.0,
-      "learning_rate": 9.270921008351919e-08,
-      "loss": 0.8328,
+      "learning_rate": 5.503631823771116e-09,
+      "loss": 0.7879,
       "step": 34879
     },
     {
-      "epoch": 0.9579522671720084,
+      "epoch": 0.9897843359818388,
       "grad_norm": 0.0,
-      "learning_rate": 9.258840633653565e-08,
-      "loss": 0.7256,
+      "learning_rate": 5.473185797923997e-09,
+      "loss": 0.8202,
       "step": 34880
     },
     {
-      "epoch": 0.9579797313998517,
+      "epoch": 0.9898127128263338,
       "grad_norm": 0.0,
-      "learning_rate": 9.246768098083003e-08,
-      "loss": 0.7984,
+      "learning_rate": 5.442824196294494e-09,
+      "loss": 0.8024,
       "step": 34881
     },
     {
-      "epoch": 0.9580071956276949,
+      "epoch": 0.9898410896708286,
       "grad_norm": 0.0,
-      "learning_rate": 9.234703401735711e-08,
-      "loss": 0.7296,
+      "learning_rate": 5.41254701913907e-09,
+      "loss": 0.723,
       "step": 34882
     },
     {
-      "epoch": 0.9580346598555382,
+      "epoch": 0.9898694665153235,
       "grad_norm": 0.0,
-      "learning_rate": 9.22264654470728e-08,
-      "loss": 0.7221,
+      "learning_rate": 5.3823542667130745e-09,
+      "loss": 0.7791,
       "step": 34883
     },
     {
-      "epoch": 0.9580621240833814,
+      "epoch": 0.9898978433598183,
       "grad_norm": 0.0,
-      "learning_rate": 9.210597527093079e-08,
-      "loss": 0.8318,
+      "learning_rate": 5.352245939271861e-09,
+      "loss": 0.8589,
       "step": 34884
     },
     {
-      "epoch": 0.9580895883112246,
+      "epoch": 0.9899262202043133,
       "grad_norm": 0.0,
-      "learning_rate": 9.198556348988474e-08,
-      "loss": 0.8185,
+      "learning_rate": 5.322222037068558e-09,
+      "loss": 0.7241,
       "step": 34885
     },
     {
-      "epoch": 0.9581170525390679,
+      "epoch": 0.9899545970488082,
       "grad_norm": 0.0,
-      "learning_rate": 9.186523010488613e-08,
-      "loss": 0.7116,
+      "learning_rate": 5.292282560358519e-09,
+      "loss": 0.7873,
       "step": 34886
     },
     {
-      "epoch": 0.9581445167669111,
+      "epoch": 0.989982973893303,
       "grad_norm": 0.0,
-      "learning_rate": 9.174497511688862e-08,
-      "loss": 0.807,
+      "learning_rate": 5.262427509393764e-09,
+      "loss": 0.7545,
       "step": 34887
     },
     {
-      "epoch": 0.9581719809947543,
+      "epoch": 0.990011350737798,
       "grad_norm": 0.0,
-      "learning_rate": 9.162479852684258e-08,
-      "loss": 0.8508,
+      "learning_rate": 5.232656884426313e-09,
+      "loss": 0.811,
       "step": 34888
     },
     {
-      "epoch": 0.9581994452225976,
+      "epoch": 0.9900397275822929,
       "grad_norm": 0.0,
-      "learning_rate": 9.150470033569836e-08,
-      "loss": 0.838,
+      "learning_rate": 5.202970685708186e-09,
+      "loss": 0.7871,
       "step": 34889
     },
     {
-      "epoch": 0.9582269094504408,
+      "epoch": 0.9900681044267877,
       "grad_norm": 0.0,
-      "learning_rate": 9.138468054440852e-08,
-      "loss": 0.8949,
+      "learning_rate": 5.1733689134902954e-09,
+      "loss": 0.7478,
       "step": 34890
     },
     {
-      "epoch": 0.958254373678284,
+      "epoch": 0.9900964812712826,
       "grad_norm": 0.0,
-      "learning_rate": 9.126473915392008e-08,
-      "loss": 0.8526,
+      "learning_rate": 5.143851568021329e-09,
+      "loss": 0.8557,
       "step": 34891
     },
     {
-      "epoch": 0.9582818379061273,
+      "epoch": 0.9901248581157776,
       "grad_norm": 0.0,
-      "learning_rate": 9.114487616518453e-08,
-      "loss": 0.8382,
+      "learning_rate": 5.1144186495522e-09,
+      "loss": 0.8353,
       "step": 34892
     },
     {
-      "epoch": 0.9583093021339705,
+      "epoch": 0.9901532349602724,
       "grad_norm": 0.0,
-      "learning_rate": 9.102509157914774e-08,
-      "loss": 0.8234,
+      "learning_rate": 5.085070158330485e-09,
+      "loss": 0.828,
       "step": 34893
     },
     {
-      "epoch": 0.9583367663618138,
+      "epoch": 0.9901816118047673,
       "grad_norm": 0.0,
-      "learning_rate": 9.090538539676008e-08,
-      "loss": 0.859,
+      "learning_rate": 5.055806094603766e-09,
+      "loss": 0.723,
       "step": 34894
     },
     {
-      "epoch": 0.9583642305896569,
+      "epoch": 0.9902099886492622,
       "grad_norm": 0.0,
-      "learning_rate": 9.078575761896747e-08,
-      "loss": 0.7827,
+      "learning_rate": 5.026626458620732e-09,
+      "loss": 0.7656,
       "step": 34895
     },
     {
-      "epoch": 0.9583916948175002,
+      "epoch": 0.9902383654937571,
       "grad_norm": 0.0,
-      "learning_rate": 9.066620824671468e-08,
-      "loss": 0.777,
+      "learning_rate": 4.997531250627851e-09,
+      "loss": 0.7159,
       "step": 34896
     },
     {
-      "epoch": 0.9584191590453435,
+      "epoch": 0.990266742338252,
       "grad_norm": 0.0,
-      "learning_rate": 9.054673728095098e-08,
-      "loss": 0.6899,
+      "learning_rate": 4.9685204708693755e-09,
+      "loss": 0.8667,
       "step": 34897
     },
     {
-      "epoch": 0.9584466232731866,
+      "epoch": 0.9902951191827469,
       "grad_norm": 0.0,
-      "learning_rate": 9.042734472262004e-08,
-      "loss": 0.8084,
+      "learning_rate": 4.939594119590663e-09,
+      "loss": 0.7769,
       "step": 34898
     },
     {
-      "epoch": 0.9584740875010299,
+      "epoch": 0.9903234960272418,
       "grad_norm": 0.0,
-      "learning_rate": 9.030803057266558e-08,
-      "loss": 0.8196,
+      "learning_rate": 4.910752197037072e-09,
+      "loss": 0.8093,
       "step": 34899
     },
     {
-      "epoch": 0.9585015517288732,
+      "epoch": 0.9903518728717366,
       "grad_norm": 0.0,
-      "learning_rate": 9.018879483203236e-08,
-      "loss": 0.7146,
+      "learning_rate": 4.881994703451742e-09,
+      "loss": 0.8086,
       "step": 34900
     },
     {
-      "epoch": 0.9585290159567164,
+      "epoch": 0.9903802497162315,
       "grad_norm": 0.0,
-      "learning_rate": 9.00696375016652e-08,
-      "loss": 0.8489,
+      "learning_rate": 4.8533216390778125e-09,
+      "loss": 0.7197,
       "step": 34901
     },
     {
-      "epoch": 0.9585564801845596,
+      "epoch": 0.9904086265607265,
       "grad_norm": 0.0,
-      "learning_rate": 8.995055858250557e-08,
-      "loss": 0.8698,
+      "learning_rate": 4.8247330041562015e-09,
+      "loss": 0.8261,
       "step": 34902
     },
     {
-      "epoch": 0.9585839444124028,
+      "epoch": 0.9904370034052213,
       "grad_norm": 0.0,
-      "learning_rate": 8.983155807549604e-08,
-      "loss": 0.8025,
+      "learning_rate": 4.796228798930047e-09,
+      "loss": 0.7512,
       "step": 34903
     },
     {
-      "epoch": 0.9586114086402461,
+      "epoch": 0.9904653802497162,
       "grad_norm": 0.0,
-      "learning_rate": 8.971263598157809e-08,
-      "loss": 0.7842,
+      "learning_rate": 4.767809023639158e-09,
+      "loss": 0.7098,
       "step": 34904
     },
     {
-      "epoch": 0.9586388728680894,
+      "epoch": 0.9904937570942112,
       "grad_norm": 0.0,
-      "learning_rate": 8.959379230169207e-08,
-      "loss": 0.8203,
+      "learning_rate": 4.739473678524453e-09,
+      "loss": 0.7968,
       "step": 34905
     },
     {
-      "epoch": 0.9586663370959325,
+      "epoch": 0.990522133938706,
       "grad_norm": 0.0,
-      "learning_rate": 8.947502703677946e-08,
-      "loss": 0.7676,
+      "learning_rate": 4.71122276382463e-09,
+      "loss": 0.7863,
       "step": 34906
     },
     {
-      "epoch": 0.9586938013237758,
+      "epoch": 0.9905505107832009,
       "grad_norm": 0.0,
-      "learning_rate": 8.935634018777839e-08,
-      "loss": 0.8555,
+      "learning_rate": 4.6830562797783865e-09,
+      "loss": 0.7651,
       "step": 34907
     },
     {
-      "epoch": 0.958721265551619,
+      "epoch": 0.9905788876276957,
       "grad_norm": 0.0,
-      "learning_rate": 8.923773175562922e-08,
-      "loss": 0.8256,
+      "learning_rate": 4.65497422662331e-09,
+      "loss": 0.7807,
       "step": 34908
     },
     {
-      "epoch": 0.9587487297794622,
+      "epoch": 0.9906072644721907,
       "grad_norm": 0.0,
-      "learning_rate": 8.91192017412712e-08,
-      "loss": 0.8329,
+      "learning_rate": 4.626976604596989e-09,
+      "loss": 0.7161,
       "step": 34909
     },
     {
-      "epoch": 0.9587761940073055,
+      "epoch": 0.9906356413166856,
       "grad_norm": 0.0,
-      "learning_rate": 8.900075014564136e-08,
-      "loss": 0.8528,
+      "learning_rate": 4.5990634139359004e-09,
+      "loss": 0.852,
       "step": 34910
     },
     {
-      "epoch": 0.9588036582351487,
+      "epoch": 0.9906640181611804,
       "grad_norm": 0.0,
-      "learning_rate": 8.88823769696756e-08,
-      "loss": 0.8051,
+      "learning_rate": 4.571234654876522e-09,
+      "loss": 0.8027,
       "step": 34911
     },
     {
-      "epoch": 0.958831122462992,
+      "epoch": 0.9906923950056754,
       "grad_norm": 0.0,
-      "learning_rate": 8.87640822143121e-08,
-      "loss": 0.8662,
+      "learning_rate": 4.543490327653111e-09,
+      "loss": 0.7137,
       "step": 34912
     },
     {
-      "epoch": 0.9588585866908352,
+      "epoch": 0.9907207718501703,
       "grad_norm": 0.0,
-      "learning_rate": 8.864586588048563e-08,
-      "loss": 0.8492,
+      "learning_rate": 4.515830432498813e-09,
+      "loss": 0.7933,
       "step": 34913
     },
     {
-      "epoch": 0.9588860509186784,
+      "epoch": 0.9907491486946651,
       "grad_norm": 0.0,
-      "learning_rate": 8.852772796913322e-08,
-      "loss": 0.7954,
+      "learning_rate": 4.488254969650108e-09,
+      "loss": 0.8314,
       "step": 34914
     },
     {
-      "epoch": 0.9589135151465217,
+      "epoch": 0.9907775255391601,
       "grad_norm": 0.0,
-      "learning_rate": 8.84096684811897e-08,
-      "loss": 0.8508,
+      "learning_rate": 4.46076393933792e-09,
+      "loss": 0.7267,
       "step": 34915
     },
     {
-      "epoch": 0.9589409793743648,
+      "epoch": 0.990805902383655,
       "grad_norm": 0.0,
-      "learning_rate": 8.829168741758654e-08,
-      "loss": 0.8076,
+      "learning_rate": 4.433357341795397e-09,
+      "loss": 0.7163,
       "step": 34916
     },
     {
-      "epoch": 0.9589684436022081,
+      "epoch": 0.9908342792281498,
       "grad_norm": 0.0,
-      "learning_rate": 8.817378477925965e-08,
-      "loss": 0.8242,
+      "learning_rate": 4.406035177253465e-09,
+      "loss": 0.8677,
       "step": 34917
     },
     {
-      "epoch": 0.9589959078300514,
+      "epoch": 0.9908626560726447,
       "grad_norm": 0.0,
-      "learning_rate": 8.805596056714161e-08,
-      "loss": 0.8291,
+      "learning_rate": 4.37879744594305e-09,
+      "loss": 0.7383,
       "step": 34918
     },
     {
-      "epoch": 0.9590233720578946,
+      "epoch": 0.9908910329171396,
       "grad_norm": 0.0,
-      "learning_rate": 8.793821478216391e-08,
-      "loss": 0.8099,
+      "learning_rate": 4.35164414809397e-09,
+      "loss": 0.8012,
       "step": 34919
     },
     {
-      "epoch": 0.9590508362857378,
+      "epoch": 0.9909194097616345,
       "grad_norm": 0.0,
-      "learning_rate": 8.782054742525803e-08,
-      "loss": 0.7204,
+      "learning_rate": 4.3245752839360385e-09,
+      "loss": 0.859,
       "step": 34920
     },
     {
-      "epoch": 0.959078300513581,
+      "epoch": 0.9909477866061294,
       "grad_norm": 0.0,
-      "learning_rate": 8.770295849735655e-08,
-      "loss": 0.8079,
+      "learning_rate": 4.297590853697964e-09,
+      "loss": 0.8795,
       "step": 34921
     },
     {
-      "epoch": 0.9591057647414243,
+      "epoch": 0.9909761634506243,
       "grad_norm": 0.0,
-      "learning_rate": 8.758544799938873e-08,
-      "loss": 0.8534,
+      "learning_rate": 4.27069085760623e-09,
+      "loss": 0.6799,
       "step": 34922
     },
     {
-      "epoch": 0.9591332289692676,
+      "epoch": 0.9910045402951192,
       "grad_norm": 0.0,
-      "learning_rate": 8.746801593228493e-08,
-      "loss": 0.8397,
+      "learning_rate": 4.243875295890653e-09,
+      "loss": 0.7724,
       "step": 34923
     },
     {
-      "epoch": 0.9591606931971107,
+      "epoch": 0.9910329171396141,
       "grad_norm": 0.0,
-      "learning_rate": 8.735066229697332e-08,
-      "loss": 0.7248,
+      "learning_rate": 4.21714416877661e-09,
+      "loss": 0.7627,
       "step": 34924
     },
     {
-      "epoch": 0.959188157424954,
+      "epoch": 0.9910612939841089,
       "grad_norm": 0.0,
-      "learning_rate": 8.723338709438201e-08,
-      "loss": 0.7168,
+      "learning_rate": 4.190497476488364e-09,
+      "loss": 0.8177,
       "step": 34925
     },
     {
-      "epoch": 0.9592156216527973,
+      "epoch": 0.9910896708286039,
       "grad_norm": 0.0,
-      "learning_rate": 8.71161903254425e-08,
-      "loss": 0.7042,
+      "learning_rate": 4.163935219253512e-09,
+      "loss": 0.8088,
       "step": 34926
     },
     {
-      "epoch": 0.9592430858806404,
+      "epoch": 0.9911180476730987,
       "grad_norm": 0.0,
-      "learning_rate": 8.699907199107849e-08,
-      "loss": 0.7295,
+      "learning_rate": 4.137457397295208e-09,
+      "loss": 0.7097,
       "step": 34927
     },
     {
-      "epoch": 0.9592705501084837,
+      "epoch": 0.9911464245175936,
       "grad_norm": 0.0,
-      "learning_rate": 8.688203209221701e-08,
-      "loss": 0.7798,
+      "learning_rate": 4.111064010836607e-09,
+      "loss": 0.8085,
       "step": 34928
     },
     {
-      "epoch": 0.9592980143363269,
+      "epoch": 0.9911748013620886,
       "grad_norm": 0.0,
-      "learning_rate": 8.676507062978512e-08,
-      "loss": 0.7298,
+      "learning_rate": 4.084755060101975e-09,
+      "loss": 0.8752,
       "step": 34929
     },
     {
-      "epoch": 0.9593254785641702,
+      "epoch": 0.9912031782065834,
       "grad_norm": 0.0,
-      "learning_rate": 8.664818760470761e-08,
-      "loss": 0.7641,
+      "learning_rate": 4.0585305453122444e-09,
+      "loss": 0.7422,
       "step": 34930
     },
     {
-      "epoch": 0.9593529427920134,
+      "epoch": 0.9912315550510783,
       "grad_norm": 0.0,
-      "learning_rate": 8.653138301791041e-08,
-      "loss": 0.8372,
+      "learning_rate": 4.032390466688352e-09,
+      "loss": 0.9296,
       "step": 34931
     },
     {
-      "epoch": 0.9593804070198566,
+      "epoch": 0.9912599318955733,
       "grad_norm": 0.0,
-      "learning_rate": 8.641465687031725e-08,
-      "loss": 0.884,
+      "learning_rate": 4.00633482445345e-09,
+      "loss": 0.8246,
       "step": 34932
     },
     {
-      "epoch": 0.9594078712476999,
+      "epoch": 0.9912883087400681,
       "grad_norm": 0.0,
-      "learning_rate": 8.62980091628518e-08,
-      "loss": 0.783,
+      "learning_rate": 3.980363618826255e-09,
+      "loss": 0.9189,
       "step": 34933
     },
     {
-      "epoch": 0.959435335475543,
+      "epoch": 0.991316685584563,
       "grad_norm": 0.0,
-      "learning_rate": 8.618143989643558e-08,
-      "loss": 0.8467,
+      "learning_rate": 3.95447685002659e-09,
+      "loss": 0.8947,
       "step": 34934
     },
     {
-      "epoch": 0.9594627997033863,
+      "epoch": 0.9913450624290578,
       "grad_norm": 0.0,
-      "learning_rate": 8.606494907199336e-08,
-      "loss": 0.7789,
+      "learning_rate": 3.928674518272058e-09,
+      "loss": 0.7395,
       "step": 34935
     },
     {
-      "epoch": 0.9594902639312296,
+      "epoch": 0.9913734392735528,
       "grad_norm": 0.0,
-      "learning_rate": 8.594853669044445e-08,
-      "loss": 0.8503,
+      "learning_rate": 3.902956623782484e-09,
+      "loss": 0.8556,
       "step": 34936
     },
     {
-      "epoch": 0.9595177281590728,
+      "epoch": 0.9914018161180477,
       "grad_norm": 0.0,
-      "learning_rate": 8.583220275271031e-08,
-      "loss": 0.7035,
+      "learning_rate": 3.87732316677325e-09,
+      "loss": 0.814,
       "step": 34937
     },
     {
-      "epoch": 0.959545192386916,
+      "epoch": 0.9914301929625425,
       "grad_norm": 0.0,
-      "learning_rate": 8.571594725971355e-08,
-      "loss": 0.819,
+      "learning_rate": 3.851774147461962e-09,
+      "loss": 0.8525,
       "step": 34938
     },
     {
-      "epoch": 0.9595726566147593,
+      "epoch": 0.9914585698070375,
       "grad_norm": 0.0,
-      "learning_rate": 8.559977021237231e-08,
-      "loss": 0.7812,
+      "learning_rate": 3.826309566062891e-09,
+      "loss": 0.8669,
       "step": 34939
     },
     {
-      "epoch": 0.9596001208426025,
+      "epoch": 0.9914869466515324,
       "grad_norm": 0.0,
-      "learning_rate": 8.548367161160587e-08,
-      "loss": 0.7643,
+      "learning_rate": 3.800929422793642e-09,
+      "loss": 0.8141,
       "step": 34940
     },
     {
-      "epoch": 0.9596275850704458,
+      "epoch": 0.9915153234960272,
       "grad_norm": 0.0,
-      "learning_rate": 8.536765145833348e-08,
-      "loss": 0.8195,
+      "learning_rate": 3.7756337178662675e-09,
+      "loss": 0.7963,
       "step": 34941
     },
     {
-      "epoch": 0.9596550492982889,
+      "epoch": 0.9915437003405221,
       "grad_norm": 0.0,
-      "learning_rate": 8.52517097534733e-08,
-      "loss": 0.7939,
+      "learning_rate": 3.750422451496149e-09,
+      "loss": 0.7304,
       "step": 34942
     },
     {
-      "epoch": 0.9596825135261322,
+      "epoch": 0.991572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.513584649794126e-08,
-      "loss": 0.8763,
+      "learning_rate": 3.725295623896452e-09,
+      "loss": 0.8133,
       "step": 34943
     },
     {
-      "epoch": 0.9597099777539755,
+      "epoch": 0.9916004540295119,
       "grad_norm": 0.0,
-      "learning_rate": 8.502006169265553e-08,
-      "loss": 0.85,
+      "learning_rate": 3.700253235277007e-09,
+      "loss": 0.842,
       "step": 34944
     },
     {
-      "epoch": 0.9597374419818187,
+      "epoch": 0.9916288308740068,
       "grad_norm": 0.0,
-      "learning_rate": 8.490435533853202e-08,
-      "loss": 0.7258,
+      "learning_rate": 3.6752952858520873e-09,
+      "loss": 0.7472,
       "step": 34945
     },
     {
-      "epoch": 0.9597649062096619,
+      "epoch": 0.9916572077185017,
       "grad_norm": 0.0,
-      "learning_rate": 8.478872743648559e-08,
-      "loss": 0.8471,
+      "learning_rate": 3.650421775830415e-09,
+      "loss": 0.8621,
       "step": 34946
     },
     {
-      "epoch": 0.9597923704375052,
+      "epoch": 0.9916855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 8.467317798743103e-08,
-      "loss": 0.7834,
+      "learning_rate": 3.6256327054229325e-09,
+      "loss": 0.8674,
       "step": 34947
     },
     {
-      "epoch": 0.9598198346653484,
+      "epoch": 0.9917139614074915,
       "grad_norm": 0.0,
-      "learning_rate": 8.455770699228428e-08,
-      "loss": 0.7925,
+      "learning_rate": 3.6009280748394716e-09,
+      "loss": 0.8318,
       "step": 34948
     },
     {
-      "epoch": 0.9598472988931916,
+      "epoch": 0.9917423382519864,
       "grad_norm": 0.0,
-      "learning_rate": 8.444231445195683e-08,
-      "loss": 0.7607,
+      "learning_rate": 3.5763078842887546e-09,
+      "loss": 0.8132,
       "step": 34949
     },
     {
-      "epoch": 0.9598747631210348,
+      "epoch": 0.9917707150964813,
       "grad_norm": 0.0,
-      "learning_rate": 8.432700036736352e-08,
-      "loss": 0.7466,
+      "learning_rate": 3.5517721339783928e-09,
+      "loss": 0.8331,
       "step": 34950
     },
     {
-      "epoch": 0.9599022273488781,
+      "epoch": 0.9917990919409762,
       "grad_norm": 0.0,
-      "learning_rate": 8.421176473941584e-08,
-      "loss": 0.8497,
+      "learning_rate": 3.5273208241148882e-09,
+      "loss": 0.9177,
       "step": 34951
     },
     {
-      "epoch": 0.9599296915767214,
+      "epoch": 0.991827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 8.409660756902526e-08,
-      "loss": 0.6757,
+      "learning_rate": 3.502953954905852e-09,
+      "loss": 0.7145,
       "step": 34952
     },
     {
-      "epoch": 0.9599571558045645,
+      "epoch": 0.991855845629966,
       "grad_norm": 0.0,
-      "learning_rate": 8.398152885710331e-08,
-      "loss": 0.6585,
+      "learning_rate": 3.478671526556676e-09,
+      "loss": 0.7349,
       "step": 34953
     },
     {
-      "epoch": 0.9599846200324078,
+      "epoch": 0.9918842224744608,
       "grad_norm": 0.0,
-      "learning_rate": 8.386652860456034e-08,
-      "loss": 0.7455,
+      "learning_rate": 3.454473539271641e-09,
+      "loss": 0.7116,
       "step": 34954
     },
     {
-      "epoch": 0.960012084260251,
+      "epoch": 0.9919125993189557,
       "grad_norm": 0.0,
-      "learning_rate": 8.375160681230676e-08,
-      "loss": 0.8041,
+      "learning_rate": 3.4303599932572485e-09,
+      "loss": 0.7396,
       "step": 34955
     },
     {
-      "epoch": 0.9600395484880943,
+      "epoch": 0.9919409761634507,
       "grad_norm": 0.0,
-      "learning_rate": 8.363676348125182e-08,
-      "loss": 0.7779,
+      "learning_rate": 3.4063308887155587e-09,
+      "loss": 0.8266,
       "step": 34956
     },
     {
-      "epoch": 0.9600670127159375,
+      "epoch": 0.9919693530079455,
       "grad_norm": 0.0,
-      "learning_rate": 8.352199861230481e-08,
-      "loss": 0.8811,
+      "learning_rate": 3.382386225849743e-09,
+      "loss": 0.8088,
       "step": 34957
     },
     {
-      "epoch": 0.9600944769437807,
+      "epoch": 0.9919977298524404,
       "grad_norm": 0.0,
-      "learning_rate": 8.340731220637277e-08,
-      "loss": 0.7333,
+      "learning_rate": 3.358526004862972e-09,
+      "loss": 0.8133,
       "step": 34958
     },
     {
-      "epoch": 0.960121941171624,
+      "epoch": 0.9920261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 8.329270426436276e-08,
-      "loss": 0.7926,
+      "learning_rate": 3.334750225956196e-09,
+      "loss": 0.8374,
       "step": 34959
     },
     {
-      "epoch": 0.9601494053994672,
+      "epoch": 0.9920544835414302,
       "grad_norm": 0.0,
-      "learning_rate": 8.317817478718183e-08,
-      "loss": 0.8527,
+      "learning_rate": 3.311058889329255e-09,
+      "loss": 0.7934,
       "step": 34960
     },
     {
-      "epoch": 0.9601768696273104,
+      "epoch": 0.9920828603859251,
       "grad_norm": 0.0,
-      "learning_rate": 8.306372377573702e-08,
-      "loss": 0.8443,
+      "learning_rate": 3.2874519951830995e-09,
+      "loss": 0.8117,
       "step": 34961
     },
     {
-      "epoch": 0.9602043338551537,
+      "epoch": 0.9921112372304199,
       "grad_norm": 0.0,
-      "learning_rate": 8.294935123093317e-08,
-      "loss": 0.8491,
+      "learning_rate": 3.26392954371868e-09,
+      "loss": 0.8138,
       "step": 34962
     },
     {
-      "epoch": 0.9602317980829969,
+      "epoch": 0.9921396140749149,
       "grad_norm": 0.0,
-      "learning_rate": 8.283505715367623e-08,
-      "loss": 0.8717,
+      "learning_rate": 3.2404915351313957e-09,
+      "loss": 0.693,
       "step": 34963
     },
     {
-      "epoch": 0.9602592623108401,
+      "epoch": 0.9921679909194098,
       "grad_norm": 0.0,
-      "learning_rate": 8.272084154486993e-08,
-      "loss": 0.7724,
+      "learning_rate": 3.217137969622197e-09,
+      "loss": 0.7782,
       "step": 34964
     },
     {
-      "epoch": 0.9602867265386834,
+      "epoch": 0.9921963677639046,
       "grad_norm": 0.0,
-      "learning_rate": 8.260670440541685e-08,
-      "loss": 0.8235,
+      "learning_rate": 3.1938688473875935e-09,
+      "loss": 0.8707,
       "step": 34965
     },
     {
-      "epoch": 0.9603141907665266,
+      "epoch": 0.9922247446083996,
       "grad_norm": 0.0,
-      "learning_rate": 8.249264573622295e-08,
-      "loss": 0.8571,
+      "learning_rate": 3.170684168622984e-09,
+      "loss": 0.7916,
       "step": 34966
     },
     {
-      "epoch": 0.9603416549943699,
+      "epoch": 0.9922531214528945,
       "grad_norm": 0.0,
-      "learning_rate": 8.237866553818752e-08,
-      "loss": 0.8586,
+      "learning_rate": 3.1475839335248783e-09,
+      "loss": 0.7551,
       "step": 34967
     },
     {
-      "epoch": 0.960369119222213,
+      "epoch": 0.9922814982973893,
       "grad_norm": 0.0,
-      "learning_rate": 8.226476381221427e-08,
-      "loss": 0.8369,
+      "learning_rate": 3.1245681422886753e-09,
+      "loss": 0.7553,
       "step": 34968
     },
     {
-      "epoch": 0.9603965834500563,
+      "epoch": 0.9923098751418842,
       "grad_norm": 0.0,
-      "learning_rate": 8.215094055920358e-08,
-      "loss": 0.8241,
+      "learning_rate": 3.1016367951086646e-09,
+      "loss": 0.964,
       "step": 34969
     },
     {
-      "epoch": 0.9604240476778996,
+      "epoch": 0.9923382519863791,
       "grad_norm": 0.0,
-      "learning_rate": 8.203719578005698e-08,
-      "loss": 0.784,
+      "learning_rate": 3.078789892179135e-09,
+      "loss": 0.8957,
       "step": 34970
     },
     {
-      "epoch": 0.9604515119057427,
+      "epoch": 0.992366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 8.192352947567373e-08,
-      "loss": 0.8501,
+      "learning_rate": 3.0560274336921548e-09,
+      "loss": 0.7683,
       "step": 34971
     },
     {
-      "epoch": 0.960478976133586,
+      "epoch": 0.9923950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 8.180994164695311e-08,
-      "loss": 0.7681,
+      "learning_rate": 3.033349419839793e-09,
+      "loss": 0.7446,
       "step": 34972
     },
     {
-      "epoch": 0.9605064403614293,
+      "epoch": 0.9924233825198638,
       "grad_norm": 0.0,
-      "learning_rate": 8.169643229479441e-08,
-      "loss": 0.7928,
+      "learning_rate": 3.0107558508141177e-09,
+      "loss": 0.8171,
       "step": 34973
     },
     {
-      "epoch": 0.9605339045892725,
+      "epoch": 0.9924517593643587,
       "grad_norm": 0.0,
-      "learning_rate": 8.158300142009578e-08,
-      "loss": 0.7652,
+      "learning_rate": 2.9882467268060877e-09,
+      "loss": 0.7993,
       "step": 34974
     },
     {
-      "epoch": 0.9605613688171157,
+      "epoch": 0.9924801362088536,
       "grad_norm": 0.0,
-      "learning_rate": 8.146964902375431e-08,
-      "loss": 0.8077,
+      "learning_rate": 2.9658220480066612e-09,
+      "loss": 0.7856,
       "step": 34975
     },
     {
-      "epoch": 0.9605888330449589,
+      "epoch": 0.9925085130533484,
       "grad_norm": 0.0,
-      "learning_rate": 8.135637510666705e-08,
-      "loss": 0.7942,
+      "learning_rate": 2.943481814603466e-09,
+      "loss": 0.7793,
       "step": 34976
     },
     {
-      "epoch": 0.9606162972728022,
+      "epoch": 0.9925368898978434,
       "grad_norm": 0.0,
-      "learning_rate": 8.124317966972994e-08,
-      "loss": 0.8051,
+      "learning_rate": 2.92122602678635e-09,
+      "loss": 0.8694,
       "step": 34977
     },
     {
-      "epoch": 0.9606437615006455,
+      "epoch": 0.9925652667423382,
       "grad_norm": 0.0,
-      "learning_rate": 8.113006271383895e-08,
-      "loss": 0.7549,
+      "learning_rate": 2.899054684742941e-09,
+      "loss": 0.7915,
       "step": 34978
     },
     {
-      "epoch": 0.9606712257284886,
+      "epoch": 0.9925936435868331,
       "grad_norm": 0.0,
-      "learning_rate": 8.101702423989e-08,
-      "loss": 0.8143,
+      "learning_rate": 2.876967788659757e-09,
+      "loss": 0.7963,
       "step": 34979
     },
     {
-      "epoch": 0.9606986899563319,
+      "epoch": 0.9926220204313281,
       "grad_norm": 0.0,
-      "learning_rate": 8.090406424877572e-08,
-      "loss": 0.8253,
+      "learning_rate": 2.8549653387255347e-09,
+      "loss": 0.9085,
       "step": 34980
     },
     {
-      "epoch": 0.9607261541841751,
+      "epoch": 0.9926503972758229,
       "grad_norm": 0.0,
-      "learning_rate": 8.079118274139097e-08,
-      "loss": 0.7256,
+      "learning_rate": 2.8330473351245722e-09,
+      "loss": 0.7368,
       "step": 34981
     },
     {
-      "epoch": 0.9607536184120183,
+      "epoch": 0.9926787741203178,
       "grad_norm": 0.0,
-      "learning_rate": 8.067837971862946e-08,
-      "loss": 0.787,
+      "learning_rate": 2.8112137780422766e-09,
+      "loss": 0.7806,
       "step": 34982
     },
     {
-      "epoch": 0.9607810826398616,
+      "epoch": 0.9927071509648128,
       "grad_norm": 0.0,
-      "learning_rate": 8.05656551813816e-08,
-      "loss": 0.8037,
+      "learning_rate": 2.7894646676629446e-09,
+      "loss": 0.7891,
       "step": 34983
     },
     {
-      "epoch": 0.9608085468677048,
+      "epoch": 0.9927355278093076,
       "grad_norm": 0.0,
-      "learning_rate": 8.045300913054111e-08,
-      "loss": 0.6967,
+      "learning_rate": 2.7678000041697628e-09,
+      "loss": 0.7729,
       "step": 34984
     },
     {
-      "epoch": 0.9608360110955481,
+      "epoch": 0.9927639046538025,
       "grad_norm": 0.0,
-      "learning_rate": 8.034044156699949e-08,
-      "loss": 0.7737,
+      "learning_rate": 2.7462197877470286e-09,
+      "loss": 0.7452,
       "step": 34985
     },
     {
-      "epoch": 0.9608634753233913,
+      "epoch": 0.9927922814982973,
       "grad_norm": 0.0,
-      "learning_rate": 8.022795249164717e-08,
-      "loss": 0.7497,
+      "learning_rate": 2.7247240185768186e-09,
+      "loss": 0.7907,
       "step": 34986
     },
     {
-      "epoch": 0.9608909395512345,
+      "epoch": 0.9928206583427923,
       "grad_norm": 0.0,
-      "learning_rate": 8.011554190537229e-08,
-      "loss": 0.8617,
+      "learning_rate": 2.7033126968389887e-09,
+      "loss": 0.845,
       "step": 34987
     },
     {
-      "epoch": 0.9609184037790778,
+      "epoch": 0.9928490351872872,
       "grad_norm": 0.0,
-      "learning_rate": 8.00032098090664e-08,
-      "loss": 0.7799,
+      "learning_rate": 2.681985822716726e-09,
+      "loss": 0.902,
       "step": 34988
     },
     {
-      "epoch": 0.960945868006921,
+      "epoch": 0.992877412031782,
       "grad_norm": 0.0,
-      "learning_rate": 7.989095620361875e-08,
-      "loss": 0.7842,
+      "learning_rate": 2.6607433963887763e-09,
+      "loss": 0.7955,
       "step": 34989
     },
     {
-      "epoch": 0.9609733322347642,
+      "epoch": 0.992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 7.977878108991533e-08,
-      "loss": 0.7684,
+      "learning_rate": 2.6395854180349955e-09,
+      "loss": 0.7794,
       "step": 34990
     },
     {
-      "epoch": 0.9610007964626075,
+      "epoch": 0.9929341657207719,
       "grad_norm": 0.0,
-      "learning_rate": 7.96666844688454e-08,
-      "loss": 0.854,
+      "learning_rate": 2.61851188783413e-09,
+      "loss": 0.8894,
       "step": 34991
     },
     {
-      "epoch": 0.9610282606904507,
+      "epoch": 0.9929625425652667,
       "grad_norm": 0.0,
-      "learning_rate": 7.955466634129605e-08,
-      "loss": 0.767,
+      "learning_rate": 2.5975228059638146e-09,
+      "loss": 0.9023,
       "step": 34992
     },
     {
-      "epoch": 0.9610557249182939,
+      "epoch": 0.9929909194097616,
       "grad_norm": 0.0,
-      "learning_rate": 7.944272670815212e-08,
-      "loss": 0.7932,
+      "learning_rate": 2.5766181726016857e-09,
+      "loss": 0.8015,
       "step": 34993
     },
     {
-      "epoch": 0.9610831891461371,
+      "epoch": 0.9930192962542566,
       "grad_norm": 0.0,
-      "learning_rate": 7.933086557030068e-08,
-      "loss": 0.7765,
+      "learning_rate": 2.555797987924269e-09,
+      "loss": 0.7615,
       "step": 34994
     },
     {
-      "epoch": 0.9611106533739804,
+      "epoch": 0.9930476730987514,
       "grad_norm": 0.0,
-      "learning_rate": 7.921908292862546e-08,
-      "loss": 0.9037,
+      "learning_rate": 2.5350622521080893e-09,
+      "loss": 0.7041,
       "step": 34995
     },
     {
-      "epoch": 0.9611381176018237,
+      "epoch": 0.9930760499432463,
       "grad_norm": 0.0,
-      "learning_rate": 7.910737878401243e-08,
-      "loss": 0.8398,
+      "learning_rate": 2.5144109653274517e-09,
+      "loss": 0.8312,
       "step": 34996
     },
     {
-      "epoch": 0.9611655818296668,
+      "epoch": 0.9931044267877412,
       "grad_norm": 0.0,
-      "learning_rate": 7.899575313734531e-08,
-      "loss": 0.8015,
+      "learning_rate": 2.4938441277566615e-09,
+      "loss": 0.8272,
       "step": 34997
     },
     {
-      "epoch": 0.9611930460575101,
+      "epoch": 0.9931328036322361,
       "grad_norm": 0.0,
-      "learning_rate": 7.888420598950674e-08,
-      "loss": 0.8336,
+      "learning_rate": 2.4733617395700238e-09,
+      "loss": 0.7759,
       "step": 34998
     },
     {
-      "epoch": 0.9612205102853534,
+      "epoch": 0.993161180476731,
       "grad_norm": 0.0,
-      "learning_rate": 7.877273734137936e-08,
-      "loss": 0.7807,
+      "learning_rate": 2.4529638009396227e-09,
+      "loss": 0.8418,
       "step": 34999
     },
     {
-      "epoch": 0.9612479745131965,
+      "epoch": 0.9931895573212258,
       "grad_norm": 0.0,
-      "learning_rate": 7.866134719384466e-08,
-      "loss": 0.8098,
+      "learning_rate": 2.432650312039764e-09,
+      "loss": 0.8029,
       "step": 35000
     },
     {
-      "epoch": 0.9612754387410398,
+      "epoch": 0.9932179341657208,
       "grad_norm": 0.0,
-      "learning_rate": 7.855003554778418e-08,
-      "loss": 0.7143,
+      "learning_rate": 2.4124212730392004e-09,
+      "loss": 0.8529,
       "step": 35001
     },
     {
-      "epoch": 0.961302902968883,
+      "epoch": 0.9932463110102157,
       "grad_norm": 0.0,
-      "learning_rate": 7.843880240407941e-08,
-      "loss": 0.7647,
+      "learning_rate": 2.392276684111128e-09,
+      "loss": 0.8771,
       "step": 35002
     },
     {
-      "epoch": 0.9613303671967263,
+      "epoch": 0.9932746878547105,
       "grad_norm": 0.0,
-      "learning_rate": 7.832764776360968e-08,
-      "loss": 0.7675,
+      "learning_rate": 2.3722165454254096e-09,
+      "loss": 0.7645,
       "step": 35003
     },
     {
-      "epoch": 0.9613578314245695,
+      "epoch": 0.9933030646992055,
       "grad_norm": 0.0,
-      "learning_rate": 7.821657162725648e-08,
-      "loss": 0.7665,
+      "learning_rate": 2.3522408571508006e-09,
+      "loss": 0.7597,
       "step": 35004
     },
     {
-      "epoch": 0.9613852956524127,
+      "epoch": 0.9933314415437003,
       "grad_norm": 0.0,
-      "learning_rate": 7.81055739958958e-08,
-      "loss": 0.7803,
+      "learning_rate": 2.3323496194549435e-09,
+      "loss": 0.8718,
       "step": 35005
     },
     {
-      "epoch": 0.961412759880256,
+      "epoch": 0.9933598183881952,
       "grad_norm": 0.0,
-      "learning_rate": 7.799465487040803e-08,
-      "loss": 0.8014,
+      "learning_rate": 2.3125428325088127e-09,
+      "loss": 0.7005,
       "step": 35006
     },
     {
-      "epoch": 0.9614402241080992,
+      "epoch": 0.9933881952326902,
       "grad_norm": 0.0,
-      "learning_rate": 7.788381425166913e-08,
-      "loss": 0.8756,
+      "learning_rate": 2.2928204964778324e-09,
+      "loss": 0.9293,
       "step": 35007
     },
     {
-      "epoch": 0.9614676883359424,
+      "epoch": 0.993416572077185,
       "grad_norm": 0.0,
-      "learning_rate": 7.77730521405573e-08,
-      "loss": 0.8581,
+      "learning_rate": 2.2731826115274246e-09,
+      "loss": 0.7342,
       "step": 35008
     },
     {
-      "epoch": 0.9614951525637857,
+      "epoch": 0.9934449489216799,
       "grad_norm": 0.0,
-      "learning_rate": 7.76623685379485e-08,
-      "loss": 0.8067,
+      "learning_rate": 2.253629177826344e-09,
+      "loss": 0.926,
       "step": 35009
     },
     {
-      "epoch": 0.9615226167916289,
+      "epoch": 0.9934733257661748,
       "grad_norm": 0.0,
-      "learning_rate": 7.755176344471871e-08,
-      "loss": 0.8948,
+      "learning_rate": 2.2341601955377933e-09,
+      "loss": 0.6934,
       "step": 35010
     },
     {
-      "epoch": 0.9615500810194721,
+      "epoch": 0.9935017026106697,
       "grad_norm": 0.0,
-      "learning_rate": 7.744123686174277e-08,
-      "loss": 0.8754,
+      "learning_rate": 2.2147756648260854e-09,
+      "loss": 0.8057,
       "step": 35011
     },
     {
-      "epoch": 0.9615775452473154,
+      "epoch": 0.9935300794551646,
       "grad_norm": 0.0,
-      "learning_rate": 7.733078878989552e-08,
-      "loss": 0.7152,
+      "learning_rate": 2.1954755858566436e-09,
+      "loss": 0.7731,
       "step": 35012
     },
     {
-      "epoch": 0.9616050094751586,
+      "epoch": 0.9935584562996594,
       "grad_norm": 0.0,
-      "learning_rate": 7.722041923004963e-08,
-      "loss": 0.7338,
+      "learning_rate": 2.17625995879156e-09,
+      "loss": 0.7119,
       "step": 35013
     },
     {
-      "epoch": 0.9616324737030019,
+      "epoch": 0.9935868331441544,
       "grad_norm": 0.0,
-      "learning_rate": 7.711012818308106e-08,
-      "loss": 0.7512,
+      "learning_rate": 2.1571287837918178e-09,
+      "loss": 0.8984,
       "step": 35014
     },
     {
-      "epoch": 0.961659937930845,
+      "epoch": 0.9936152099886493,
       "grad_norm": 0.0,
-      "learning_rate": 7.69999156498602e-08,
-      "loss": 0.7419,
+      "learning_rate": 2.138082061021729e-09,
+      "loss": 0.681,
       "step": 35015
     },
     {
-      "epoch": 0.9616874021586883,
+      "epoch": 0.9936435868331441,
       "grad_norm": 0.0,
-      "learning_rate": 7.688978163126082e-08,
-      "loss": 0.8199,
+      "learning_rate": 2.1191197906400563e-09,
+      "loss": 0.8674,
       "step": 35016
     },
     {
-      "epoch": 0.9617148663865316,
+      "epoch": 0.993671963677639,
       "grad_norm": 0.0,
-      "learning_rate": 7.677972612815221e-08,
-      "loss": 0.8262,
+      "learning_rate": 2.1002419728066714e-09,
+      "loss": 0.7074,
       "step": 35017
     },
     {
-      "epoch": 0.9617423306143748,
+      "epoch": 0.993700340522134,
       "grad_norm": 0.0,
-      "learning_rate": 7.666974914140702e-08,
-      "loss": 0.8329,
+      "learning_rate": 2.081448607682557e-09,
+      "loss": 0.7452,
       "step": 35018
     },
     {
-      "epoch": 0.961769794842218,
+      "epoch": 0.9937287173666288,
       "grad_norm": 0.0,
-      "learning_rate": 7.655985067189342e-08,
-      "loss": 0.8264,
+      "learning_rate": 2.0627396954264743e-09,
+      "loss": 0.7548,
       "step": 35019
     },
     {
-      "epoch": 0.9617972590700612,
+      "epoch": 0.9937570942111237,
       "grad_norm": 0.0,
-      "learning_rate": 7.645003072048296e-08,
-      "loss": 0.8512,
+      "learning_rate": 2.044115236196076e-09,
+      "loss": 0.7609,
       "step": 35020
     },
     {
-      "epoch": 0.9618247232979045,
+      "epoch": 0.9937854710556187,
       "grad_norm": 0.0,
-      "learning_rate": 7.634028928804493e-08,
-      "loss": 0.7719,
+      "learning_rate": 2.0255752301479028e-09,
+      "loss": 0.6651,
       "step": 35021
     },
     {
-      "epoch": 0.9618521875257477,
+      "epoch": 0.9938138479001135,
       "grad_norm": 0.0,
-      "learning_rate": 7.623062637544531e-08,
-      "loss": 0.7608,
+      "learning_rate": 2.007119677438496e-09,
+      "loss": 0.8459,
       "step": 35022
     },
     {
-      "epoch": 0.9618796517535909,
+      "epoch": 0.9938422247446084,
       "grad_norm": 0.0,
-      "learning_rate": 7.612104198355342e-08,
-      "loss": 0.7349,
+      "learning_rate": 1.988748578225508e-09,
+      "loss": 0.8812,
       "step": 35023
     },
     {
-      "epoch": 0.9619071159814342,
+      "epoch": 0.9938706015891033,
       "grad_norm": 0.0,
-      "learning_rate": 7.60115361132363e-08,
-      "loss": 0.7854,
+      "learning_rate": 1.970461932662149e-09,
+      "loss": 0.808,
       "step": 35024
     },
     {
-      "epoch": 0.9619345802092775,
+      "epoch": 0.9938989784335982,
       "grad_norm": 0.0,
-      "learning_rate": 7.590210876535997e-08,
-      "loss": 0.7986,
+      "learning_rate": 1.95225974090385e-09,
+      "loss": 0.7877,
       "step": 35025
     },
     {
-      "epoch": 0.9619620444371206,
+      "epoch": 0.9939273552780931,
       "grad_norm": 0.0,
-      "learning_rate": 7.579275994079038e-08,
-      "loss": 0.7929,
+      "learning_rate": 1.9341420031049328e-09,
+      "loss": 0.7062,
       "step": 35026
     },
     {
-      "epoch": 0.9619895086649639,
+      "epoch": 0.9939557321225879,
       "grad_norm": 0.0,
-      "learning_rate": 7.568348964039241e-08,
-      "loss": 0.9593,
+      "learning_rate": 1.9161087194174978e-09,
+      "loss": 0.78,
       "step": 35027
     },
     {
-      "epoch": 0.9620169728928071,
+      "epoch": 0.9939841089670829,
       "grad_norm": 0.0,
-      "learning_rate": 7.557429786503089e-08,
-      "loss": 0.8296,
+      "learning_rate": 1.898159889993645e-09,
+      "loss": 0.8632,
       "step": 35028
     },
     {
-      "epoch": 0.9620444371206504,
+      "epoch": 0.9940124858115778,
       "grad_norm": 0.0,
-      "learning_rate": 7.546518461557073e-08,
-      "loss": 0.7296,
+      "learning_rate": 1.8802955149865854e-09,
+      "loss": 0.9055,
       "step": 35029
     },
     {
-      "epoch": 0.9620719013484936,
+      "epoch": 0.9940408626560726,
       "grad_norm": 0.0,
-      "learning_rate": 7.535614989287343e-08,
-      "loss": 0.8182,
+      "learning_rate": 1.862515594545089e-09,
+      "loss": 0.78,
       "step": 35030
     },
     {
-      "epoch": 0.9620993655763368,
+      "epoch": 0.9940692395005676,
       "grad_norm": 0.0,
-      "learning_rate": 7.524719369780387e-08,
-      "loss": 0.8087,
+      "learning_rate": 1.8448201288201462e-09,
+      "loss": 0.7588,
       "step": 35031
     },
     {
-      "epoch": 0.9621268298041801,
+      "epoch": 0.9940976163450624,
       "grad_norm": 0.0,
-      "learning_rate": 7.513831603122135e-08,
-      "loss": 0.7592,
+      "learning_rate": 1.8272091179627472e-09,
+      "loss": 0.7926,
       "step": 35032
     },
     {
-      "epoch": 0.9621542940320232,
+      "epoch": 0.9941259931895573,
       "grad_norm": 0.0,
-      "learning_rate": 7.502951689399074e-08,
-      "loss": 0.7963,
+      "learning_rate": 1.8096825621194414e-09,
+      "loss": 0.8104,
       "step": 35033
     },
     {
-      "epoch": 0.9621817582598665,
+      "epoch": 0.9941543700340522,
       "grad_norm": 0.0,
-      "learning_rate": 7.492079628697024e-08,
-      "loss": 0.9236,
+      "learning_rate": 1.7922404614389987e-09,
+      "loss": 0.8392,
       "step": 35034
     },
     {
-      "epoch": 0.9622092224877098,
+      "epoch": 0.9941827468785471,
       "grad_norm": 0.0,
-      "learning_rate": 7.481215421102028e-08,
-      "loss": 0.7605,
+      "learning_rate": 1.774882816069079e-09,
+      "loss": 0.8031,
       "step": 35035
     },
     {
-      "epoch": 0.962236686715553,
+      "epoch": 0.994211123723042,
       "grad_norm": 0.0,
-      "learning_rate": 7.47035906670024e-08,
-      "loss": 0.7994,
+      "learning_rate": 1.7576096261562314e-09,
+      "loss": 0.8356,
       "step": 35036
     },
     {
-      "epoch": 0.9622641509433962,
+      "epoch": 0.9942395005675368,
       "grad_norm": 0.0,
-      "learning_rate": 7.459510565577366e-08,
-      "loss": 0.7494,
+      "learning_rate": 1.7404208918470055e-09,
+      "loss": 0.8394,
       "step": 35037
     },
     {
-      "epoch": 0.9622916151712395,
+      "epoch": 0.9942678774120318,
       "grad_norm": 0.0,
-      "learning_rate": 7.448669917819451e-08,
-      "loss": 0.8297,
+      "learning_rate": 1.7233166132846203e-09,
+      "loss": 0.7634,
       "step": 35038
     },
     {
-      "epoch": 0.9623190793990827,
+      "epoch": 0.9942962542565267,
       "grad_norm": 0.0,
-      "learning_rate": 7.437837123512093e-08,
-      "loss": 0.8718,
+      "learning_rate": 1.7062967906156248e-09,
+      "loss": 0.722,
       "step": 35039
     },
     {
-      "epoch": 0.962346543626926,
+      "epoch": 0.9943246311010215,
       "grad_norm": 0.0,
-      "learning_rate": 7.427012182740889e-08,
-      "loss": 0.8701,
+      "learning_rate": 1.689361423983238e-09,
+      "loss": 0.8185,
       "step": 35040
     },
     {
-      "epoch": 0.9623740078547691,
+      "epoch": 0.9943530079455165,
       "grad_norm": 0.0,
-      "learning_rate": 7.416195095591771e-08,
-      "loss": 0.792,
+      "learning_rate": 1.6725105135295683e-09,
+      "loss": 0.773,
       "step": 35041
     },
     {
-      "epoch": 0.9624014720826124,
+      "epoch": 0.9943813847900114,
       "grad_norm": 0.0,
-      "learning_rate": 7.405385862150227e-08,
-      "loss": 0.7387,
+      "learning_rate": 1.655744059398945e-09,
+      "loss": 0.7648,
       "step": 35042
     },
     {
-      "epoch": 0.9624289363104557,
+      "epoch": 0.9944097616345062,
       "grad_norm": 0.0,
-      "learning_rate": 7.394584482501743e-08,
-      "loss": 0.8618,
+      "learning_rate": 1.6390620617301457e-09,
+      "loss": 0.7738,
       "step": 35043
     },
     {
-      "epoch": 0.9624564005382988,
+      "epoch": 0.9944381384790011,
       "grad_norm": 0.0,
-      "learning_rate": 7.383790956731696e-08,
-      "loss": 0.8127,
+      "learning_rate": 1.6224645206663892e-09,
+      "loss": 0.777,
       "step": 35044
     },
     {
-      "epoch": 0.9624838647661421,
+      "epoch": 0.9944665153234961,
       "grad_norm": 0.0,
-      "learning_rate": 7.373005284925793e-08,
-      "loss": 0.7627,
+      "learning_rate": 1.6059514363475638e-09,
+      "loss": 0.8872,
       "step": 35045
     },
     {
-      "epoch": 0.9625113289939853,
+      "epoch": 0.9944948921679909,
       "grad_norm": 0.0,
-      "learning_rate": 7.362227467169081e-08,
-      "loss": 0.7097,
+      "learning_rate": 1.589522808912447e-09,
+      "loss": 0.8029,
       "step": 35046
     },
     {
-      "epoch": 0.9625387932218286,
+      "epoch": 0.9945232690124858,
       "grad_norm": 0.0,
-      "learning_rate": 7.351457503546933e-08,
-      "loss": 0.8747,
+      "learning_rate": 1.573178638499817e-09,
+      "loss": 0.8106,
       "step": 35047
     },
     {
-      "epoch": 0.9625662574496718,
+      "epoch": 0.9945516458569807,
       "grad_norm": 0.0,
-      "learning_rate": 7.340695394144614e-08,
-      "loss": 0.8469,
+      "learning_rate": 1.5569189252473416e-09,
+      "loss": 0.7316,
       "step": 35048
     },
     {
-      "epoch": 0.962593721677515,
+      "epoch": 0.9945800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 7.32994113904717e-08,
-      "loss": 0.8628,
+      "learning_rate": 1.5407436692937983e-09,
+      "loss": 0.6821,
       "step": 35049
     },
     {
-      "epoch": 0.9626211859053583,
+      "epoch": 0.9946083995459705,
       "grad_norm": 0.0,
-      "learning_rate": 7.319194738339752e-08,
-      "loss": 0.7971,
+      "learning_rate": 1.5246528707757447e-09,
+      "loss": 0.7332,
       "step": 35050
     },
     {
-      "epoch": 0.9626486501332016,
+      "epoch": 0.9946367763904653,
       "grad_norm": 0.0,
-      "learning_rate": 7.308456192107405e-08,
-      "loss": 0.7145,
+      "learning_rate": 1.5086465298275178e-09,
+      "loss": 0.7907,
       "step": 35051
     },
     {
-      "epoch": 0.9626761143610447,
+      "epoch": 0.9946651532349603,
       "grad_norm": 0.0,
-      "learning_rate": 7.297725500435172e-08,
-      "loss": 0.7779,
+      "learning_rate": 1.492724646585675e-09,
+      "loss": 0.8929,
       "step": 35052
     },
     {
-      "epoch": 0.962703578588888,
+      "epoch": 0.9946935300794552,
       "grad_norm": 0.0,
-      "learning_rate": 7.28700266340776e-08,
-      "loss": 0.8493,
+      "learning_rate": 1.4768872211834428e-09,
+      "loss": 0.8707,
       "step": 35053
     },
     {
-      "epoch": 0.9627310428167312,
+      "epoch": 0.99472190692395,
       "grad_norm": 0.0,
-      "learning_rate": 7.276287681110217e-08,
-      "loss": 0.7387,
+      "learning_rate": 1.461134253756269e-09,
+      "loss": 0.7671,
       "step": 35054
     },
     {
-      "epoch": 0.9627585070445744,
+      "epoch": 0.994750283768445,
       "grad_norm": 0.0,
-      "learning_rate": 7.265580553627139e-08,
-      "loss": 0.7805,
+      "learning_rate": 1.4454657444351595e-09,
+      "loss": 0.8746,
       "step": 35055
     },
     {
-      "epoch": 0.9627859712724177,
+      "epoch": 0.9947786606129398,
       "grad_norm": 0.0,
-      "learning_rate": 7.254881281043346e-08,
-      "loss": 0.7943,
+      "learning_rate": 1.4298816933555615e-09,
+      "loss": 0.7047,
       "step": 35056
     },
     {
-      "epoch": 0.9628134355002609,
+      "epoch": 0.9948070374574347,
       "grad_norm": 0.0,
-      "learning_rate": 7.244189863443552e-08,
-      "loss": 0.797,
+      "learning_rate": 1.414382100646261e-09,
+      "loss": 0.9259,
       "step": 35057
     },
     {
-      "epoch": 0.9628408997281042,
+      "epoch": 0.9948354143019297,
       "grad_norm": 0.0,
-      "learning_rate": 7.233506300912241e-08,
-      "loss": 0.6609,
+      "learning_rate": 1.3989669664382644e-09,
+      "loss": 0.833,
       "step": 35058
     },
     {
-      "epoch": 0.9628683639559473,
+      "epoch": 0.9948637911464245,
       "grad_norm": 0.0,
-      "learning_rate": 7.222830593533903e-08,
-      "loss": 0.8327,
+      "learning_rate": 1.383636290863688e-09,
+      "loss": 0.6422,
       "step": 35059
     },
     {
-      "epoch": 0.9628958281837906,
+      "epoch": 0.9948921679909194,
       "grad_norm": 0.0,
-      "learning_rate": 7.212162741393247e-08,
-      "loss": 0.7502,
+      "learning_rate": 1.3683900740513178e-09,
+      "loss": 0.8853,
       "step": 35060
     },
     {
-      "epoch": 0.9629232924116339,
+      "epoch": 0.9949205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 7.201502744574429e-08,
-      "loss": 0.943,
+      "learning_rate": 1.3532283161288295e-09,
+      "loss": 0.7394,
       "step": 35061
     },
     {
-      "epoch": 0.962950756639477,
+      "epoch": 0.9949489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 7.190850603161936e-08,
-      "loss": 0.7697,
+      "learning_rate": 1.338151017225009e-09,
+      "loss": 0.7888,
       "step": 35062
     },
     {
-      "epoch": 0.9629782208673203,
+      "epoch": 0.9949772985244041,
       "grad_norm": 0.0,
-      "learning_rate": 7.180206317240035e-08,
-      "loss": 0.8939,
+      "learning_rate": 1.3231581774675317e-09,
+      "loss": 0.8035,
       "step": 35063
     },
     {
-      "epoch": 0.9630056850951636,
+      "epoch": 0.995005675368899,
       "grad_norm": 0.0,
-      "learning_rate": 7.16956988689288e-08,
-      "loss": 0.7814,
+      "learning_rate": 1.3082497969829632e-09,
+      "loss": 0.7887,
       "step": 35064
     },
     {
-      "epoch": 0.9630331493230068,
+      "epoch": 0.9950340522133939,
       "grad_norm": 0.0,
-      "learning_rate": 7.158941312204737e-08,
-      "loss": 0.7922,
+      "learning_rate": 1.2934258758967588e-09,
+      "loss": 0.7837,
       "step": 35065
     },
     {
-      "epoch": 0.96306061355085,
+      "epoch": 0.9950624290578888,
       "grad_norm": 0.0,
-      "learning_rate": 7.148320593259539e-08,
-      "loss": 0.8651,
+      "learning_rate": 1.2786864143354837e-09,
+      "loss": 0.8036,
       "step": 35066
     },
     {
-      "epoch": 0.9630880777786932,
+      "epoch": 0.9950908059023836,
       "grad_norm": 0.0,
-      "learning_rate": 7.13770773014144e-08,
-      "loss": 0.8431,
+      "learning_rate": 1.2640314124212626e-09,
+      "loss": 0.7826,
       "step": 35067
     },
     {
-      "epoch": 0.9631155420065365,
+      "epoch": 0.9951191827468785,
       "grad_norm": 0.0,
-      "learning_rate": 7.127102722934487e-08,
-      "loss": 0.8598,
+      "learning_rate": 1.249460870280661e-09,
+      "loss": 0.8407,
       "step": 35068
     },
     {
-      "epoch": 0.9631430062343798,
+      "epoch": 0.9951475595913735,
       "grad_norm": 0.0,
-      "learning_rate": 7.116505571722388e-08,
-      "loss": 0.8496,
+      "learning_rate": 1.2349747880335828e-09,
+      "loss": 0.8467,
       "step": 35069
     },
     {
-      "epoch": 0.9631704704622229,
+      "epoch": 0.9951759364358683,
       "grad_norm": 0.0,
-      "learning_rate": 7.105916276589187e-08,
-      "loss": 0.9353,
+      "learning_rate": 1.220573165805483e-09,
+      "loss": 0.8135,
       "step": 35070
     },
     {
-      "epoch": 0.9631979346900662,
+      "epoch": 0.9952043132803632,
       "grad_norm": 0.0,
-      "learning_rate": 7.095334837618595e-08,
-      "loss": 0.8098,
+      "learning_rate": 1.2062560037162664e-09,
+      "loss": 0.7649,
       "step": 35071
     },
     {
-      "epoch": 0.9632253989179094,
+      "epoch": 0.9952326901248582,
       "grad_norm": 0.0,
-      "learning_rate": 7.084761254894323e-08,
-      "loss": 0.8038,
+      "learning_rate": 1.1920233018880567e-09,
+      "loss": 0.8169,
       "step": 35072
     },
     {
-      "epoch": 0.9632528631457526,
+      "epoch": 0.995261066969353,
       "grad_norm": 0.0,
-      "learning_rate": 7.074195528499971e-08,
-      "loss": 0.8148,
+      "learning_rate": 1.177875060438538e-09,
+      "loss": 0.8151,
       "step": 35073
     },
     {
-      "epoch": 0.9632803273735959,
+      "epoch": 0.9952894438138479,
       "grad_norm": 0.0,
-      "learning_rate": 7.063637658519251e-08,
-      "loss": 0.6954,
+      "learning_rate": 1.163811279490945e-09,
+      "loss": 0.9357,
       "step": 35074
     },
     {
-      "epoch": 0.9633077916014391,
+      "epoch": 0.9953178206583428,
       "grad_norm": 0.0,
-      "learning_rate": 7.05308764503565e-08,
-      "loss": 0.786,
+      "learning_rate": 1.1498319591607409e-09,
+      "loss": 0.7666,
       "step": 35075
     },
     {
-      "epoch": 0.9633352558292824,
+      "epoch": 0.9953461975028377,
       "grad_norm": 0.0,
-      "learning_rate": 7.042545488132657e-08,
-      "loss": 0.7458,
+      "learning_rate": 1.1359370995678298e-09,
+      "loss": 0.7866,
       "step": 35076
     },
     {
-      "epoch": 0.9633627200571256,
+      "epoch": 0.9953745743473326,
       "grad_norm": 0.0,
-      "learning_rate": 7.03201118789365e-08,
-      "loss": 0.8822,
+      "learning_rate": 1.122126700828785e-09,
+      "loss": 0.6739,
       "step": 35077
     },
     {
-      "epoch": 0.9633901842849688,
+      "epoch": 0.9954029511918274,
       "grad_norm": 0.0,
-      "learning_rate": 7.021484744402007e-08,
-      "loss": 0.8343,
+      "learning_rate": 1.1084007630612903e-09,
+      "loss": 0.8098,
       "step": 35078
     },
     {
-      "epoch": 0.9634176485128121,
+      "epoch": 0.9954313280363224,
       "grad_norm": 0.0,
-      "learning_rate": 7.010966157740994e-08,
-      "loss": 0.8233,
+      "learning_rate": 1.0947592863808087e-09,
+      "loss": 0.7374,
       "step": 35079
     },
     {
-      "epoch": 0.9634451127406553,
+      "epoch": 0.9954597048808173,
       "grad_norm": 0.0,
-      "learning_rate": 7.000455427993991e-08,
-      "loss": 0.8412,
+      "learning_rate": 1.0812022709016934e-09,
+      "loss": 0.906,
       "step": 35080
     },
     {
-      "epoch": 0.9634725769684985,
+      "epoch": 0.9954880817253121,
       "grad_norm": 0.0,
-      "learning_rate": 6.989952555243929e-08,
-      "loss": 0.8352,
+      "learning_rate": 1.0677297167394075e-09,
+      "loss": 0.8035,
       "step": 35081
     },
     {
-      "epoch": 0.9635000411963418,
+      "epoch": 0.9955164585698071,
       "grad_norm": 0.0,
-      "learning_rate": 6.979457539574075e-08,
-      "loss": 0.7666,
+      "learning_rate": 1.0543416240071934e-09,
+      "loss": 0.7707,
       "step": 35082
     },
     {
-      "epoch": 0.963527505424185,
+      "epoch": 0.9955448354143019,
       "grad_norm": 0.0,
-      "learning_rate": 6.968970381067364e-08,
-      "loss": 0.9061,
+      "learning_rate": 1.0410379928182946e-09,
+      "loss": 0.6799,
       "step": 35083
     },
     {
-      "epoch": 0.9635549696520282,
+      "epoch": 0.9955732122587968,
       "grad_norm": 0.0,
-      "learning_rate": 6.95849107980684e-08,
-      "loss": 0.8702,
+      "learning_rate": 1.0278188232859533e-09,
+      "loss": 0.7347,
       "step": 35084
     },
     {
-      "epoch": 0.9635824338798714,
+      "epoch": 0.9956015891032917,
       "grad_norm": 0.0,
-      "learning_rate": 6.948019635875325e-08,
-      "loss": 0.743,
+      "learning_rate": 1.0146841155200814e-09,
+      "loss": 0.7654,
       "step": 35085
     },
     {
-      "epoch": 0.9636098981077147,
+      "epoch": 0.9956299659477866,
       "grad_norm": 0.0,
-      "learning_rate": 6.937556049355865e-08,
-      "loss": 0.7789,
+      "learning_rate": 1.0016338696339224e-09,
+      "loss": 0.7252,
       "step": 35086
     },
     {
-      "epoch": 0.963637362335558,
+      "epoch": 0.9956583427922815,
       "grad_norm": 0.0,
-      "learning_rate": 6.92710032033117e-08,
-      "loss": 0.8358,
+      "learning_rate": 9.886680857362773e-10,
+      "loss": 0.8273,
       "step": 35087
     },
     {
-      "epoch": 0.9636648265634011,
+      "epoch": 0.9956867196367764,
       "grad_norm": 0.0,
-      "learning_rate": 6.916652448883843e-08,
-      "loss": 0.7981,
+      "learning_rate": 9.757867639359485e-10,
+      "loss": 0.6439,
       "step": 35088
     },
     {
-      "epoch": 0.9636922907912444,
+      "epoch": 0.9957150964812713,
       "grad_norm": 0.0,
-      "learning_rate": 6.906212435096704e-08,
-      "loss": 0.754,
+      "learning_rate": 9.629899043428481e-10,
+      "loss": 0.8338,
       "step": 35089
     },
     {
-      "epoch": 0.9637197550190877,
+      "epoch": 0.9957434733257662,
       "grad_norm": 0.0,
-      "learning_rate": 6.895780279052244e-08,
-      "loss": 0.7667,
+      "learning_rate": 9.502775070657778e-10,
+      "loss": 0.806,
       "step": 35090
     },
     {
-      "epoch": 0.9637472192469309,
+      "epoch": 0.995771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 6.885355980833064e-08,
-      "loss": 0.7487,
+      "learning_rate": 9.376495722102085e-10,
+      "loss": 0.701,
       "step": 35091
     },
     {
-      "epoch": 0.9637746834747741,
+      "epoch": 0.995800227014756,
       "grad_norm": 0.0,
-      "learning_rate": 6.874939540521652e-08,
-      "loss": 0.8984,
+      "learning_rate": 9.251060998838323e-10,
+      "loss": 0.803,
       "step": 35092
     },
     {
-      "epoch": 0.9638021477026173,
+      "epoch": 0.9958286038592509,
       "grad_norm": 0.0,
-      "learning_rate": 6.864530958200388e-08,
-      "loss": 0.7764,
+      "learning_rate": 9.126470901932305e-10,
+      "loss": 0.7848,
       "step": 35093
     },
     {
-      "epoch": 0.9638296119304606,
+      "epoch": 0.9958569807037457,
       "grad_norm": 0.0,
-      "learning_rate": 6.854130233951761e-08,
-      "loss": 0.8399,
+      "learning_rate": 9.002725432427639e-10,
+      "loss": 0.8373,
       "step": 35094
     },
     {
-      "epoch": 0.9638570761583038,
+      "epoch": 0.9958853575482406,
       "grad_norm": 0.0,
-      "learning_rate": 6.843737367857928e-08,
-      "loss": 0.8425,
+      "learning_rate": 8.879824591367936e-10,
+      "loss": 0.8437,
       "step": 35095
     },
     {
-      "epoch": 0.963884540386147,
+      "epoch": 0.9959137343927356,
       "grad_norm": 0.0,
-      "learning_rate": 6.833352360001155e-08,
-      "loss": 0.8763,
+      "learning_rate": 8.757768379796805e-10,
+      "loss": 0.8076,
       "step": 35096
     },
     {
-      "epoch": 0.9639120046139903,
+      "epoch": 0.9959421112372304,
       "grad_norm": 0.0,
-      "learning_rate": 6.822975210463489e-08,
-      "loss": 0.7822,
+      "learning_rate": 8.636556798746754e-10,
+      "loss": 0.8384,
       "step": 35097
     },
     {
-      "epoch": 0.9639394688418335,
+      "epoch": 0.9959704880817253,
       "grad_norm": 0.0,
-      "learning_rate": 6.812605919327309e-08,
-      "loss": 0.7558,
+      "learning_rate": 8.516189849239187e-10,
+      "loss": 0.7421,
       "step": 35098
     },
     {
-      "epoch": 0.9639669330696767,
+      "epoch": 0.9959988649262203,
       "grad_norm": 0.0,
-      "learning_rate": 6.802244486674548e-08,
-      "loss": 0.8266,
+      "learning_rate": 8.396667532284408e-10,
+      "loss": 0.8002,
       "step": 35099
     },
     {
-      "epoch": 0.96399439729752,
+      "epoch": 0.9960272417707151,
       "grad_norm": 0.0,
-      "learning_rate": 6.791890912587029e-08,
-      "loss": 0.7932,
+      "learning_rate": 8.277989848903822e-10,
+      "loss": 0.7757,
       "step": 35100
     },
     {
-      "epoch": 0.9640218615253632,
+      "epoch": 0.99605561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.7815451971468e-08,
-      "loss": 0.9131,
+      "learning_rate": 8.160156800085528e-10,
+      "loss": 0.815,
       "step": 35101
     },
     {
-      "epoch": 0.9640493257532065,
+      "epoch": 0.9960839954597048,
       "grad_norm": 0.0,
-      "learning_rate": 6.771207340435793e-08,
-      "loss": 0.8853,
+      "learning_rate": 8.043168386839828e-10,
+      "loss": 0.773,
       "step": 35102
     },
     {
-      "epoch": 0.9640767899810497,
+      "epoch": 0.9961123723041998,
       "grad_norm": 0.0,
-      "learning_rate": 6.760877342535721e-08,
-      "loss": 0.9329,
+      "learning_rate": 7.927024610154821e-10,
+      "loss": 0.8405,
       "step": 35103
     },
     {
-      "epoch": 0.9641042542088929,
+      "epoch": 0.9961407491486947,
       "grad_norm": 0.0,
-      "learning_rate": 6.750555203528408e-08,
-      "loss": 0.8409,
+      "learning_rate": 7.811725470996401e-10,
+      "loss": 0.784,
       "step": 35104
     },
     {
-      "epoch": 0.9641317184367362,
+      "epoch": 0.9961691259931895,
       "grad_norm": 0.0,
-      "learning_rate": 6.740240923495345e-08,
-      "loss": 0.7758,
+      "learning_rate": 7.697270970352666e-10,
+      "loss": 0.7354,
       "step": 35105
     },
     {
-      "epoch": 0.9641591826645793,
+      "epoch": 0.9961975028376845,
       "grad_norm": 0.0,
-      "learning_rate": 6.729934502518243e-08,
-      "loss": 0.8157,
+      "learning_rate": 7.583661109178408e-10,
+      "loss": 0.8561,
       "step": 35106
     },
     {
-      "epoch": 0.9641866468924226,
+      "epoch": 0.9962258796821793,
       "grad_norm": 0.0,
-      "learning_rate": 6.719635940678704e-08,
-      "loss": 0.7337,
+      "learning_rate": 7.470895888439522e-10,
+      "loss": 0.8555,
       "step": 35107
     },
     {
-      "epoch": 0.9642141111202659,
+      "epoch": 0.9962542565266742,
       "grad_norm": 0.0,
-      "learning_rate": 6.709345238058106e-08,
-      "loss": 0.7866,
+      "learning_rate": 7.358975309090799e-10,
+      "loss": 0.776,
       "step": 35108
     },
     {
-      "epoch": 0.9642415753481091,
+      "epoch": 0.9962826333711692,
       "grad_norm": 0.0,
-      "learning_rate": 6.699062394737943e-08,
-      "loss": 0.853,
+      "learning_rate": 7.247899372087031e-10,
+      "loss": 0.7613,
       "step": 35109
     },
     {
-      "epoch": 0.9642690395759523,
+      "epoch": 0.996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 6.688787410799591e-08,
-      "loss": 0.7463,
+      "learning_rate": 7.1376680783386e-10,
+      "loss": 0.8368,
       "step": 35110
     },
     {
-      "epoch": 0.9642965038037955,
+      "epoch": 0.9963393870601589,
       "grad_norm": 0.0,
-      "learning_rate": 6.67852028632443e-08,
-      "loss": 0.7981,
+      "learning_rate": 7.028281428800299e-10,
+      "loss": 0.898,
       "step": 35111
     },
     {
-      "epoch": 0.9643239680316388,
+      "epoch": 0.9963677639046538,
       "grad_norm": 0.0,
-      "learning_rate": 6.668261021393396e-08,
-      "loss": 0.7897,
+      "learning_rate": 6.919739424393612e-10,
+      "loss": 0.8897,
       "step": 35112
     },
     {
-      "epoch": 0.964351432259482,
+      "epoch": 0.9963961407491487,
       "grad_norm": 0.0,
-      "learning_rate": 6.65800961608798e-08,
-      "loss": 0.7509,
+      "learning_rate": 6.812042066028923e-10,
+      "loss": 0.7446,
       "step": 35113
     },
     {
-      "epoch": 0.9643788964873252,
+      "epoch": 0.9964245175936436,
       "grad_norm": 0.0,
-      "learning_rate": 6.647766070489115e-08,
-      "loss": 0.7557,
+      "learning_rate": 6.705189354616615e-10,
+      "loss": 0.8486,
       "step": 35114
     },
     {
-      "epoch": 0.9644063607151685,
+      "epoch": 0.9964528944381384,
       "grad_norm": 0.0,
-      "learning_rate": 6.637530384677959e-08,
-      "loss": 0.7701,
+      "learning_rate": 6.599181291055967e-10,
+      "loss": 0.7751,
       "step": 35115
     },
     {
-      "epoch": 0.9644338249430118,
+      "epoch": 0.9964812712826334,
       "grad_norm": 0.0,
-      "learning_rate": 6.62730255873545e-08,
-      "loss": 0.7729,
+      "learning_rate": 6.494017876246261e-10,
+      "loss": 0.7536,
       "step": 35116
     },
     {
-      "epoch": 0.9644612891708549,
+      "epoch": 0.9965096481271283,
       "grad_norm": 0.0,
-      "learning_rate": 6.617082592742519e-08,
-      "loss": 0.8338,
+      "learning_rate": 6.389699111086778e-10,
+      "loss": 0.7509,
       "step": 35117
     },
     {
-      "epoch": 0.9644887533986982,
+      "epoch": 0.9965380249716231,
       "grad_norm": 0.0,
-      "learning_rate": 6.606870486780104e-08,
-      "loss": 0.7152,
+      "learning_rate": 6.286224996443491e-10,
+      "loss": 0.6614,
       "step": 35118
     },
     {
-      "epoch": 0.9645162176265414,
+      "epoch": 0.996566401816118,
       "grad_norm": 0.0,
-      "learning_rate": 6.596666240928806e-08,
-      "loss": 0.8757,
+      "learning_rate": 6.183595533193476e-10,
+      "loss": 0.8352,
       "step": 35119
     },
     {
-      "epoch": 0.9645436818543847,
+      "epoch": 0.996594778660613,
       "grad_norm": 0.0,
-      "learning_rate": 6.586469855269562e-08,
-      "loss": 0.8304,
+      "learning_rate": 6.081810722202708e-10,
+      "loss": 0.8182,
       "step": 35120
     },
     {
-      "epoch": 0.9645711460822279,
+      "epoch": 0.9966231555051078,
       "grad_norm": 0.0,
-      "learning_rate": 6.57628132988286e-08,
-      "loss": 0.8766,
+      "learning_rate": 5.98087056433716e-10,
+      "loss": 0.7501,
       "step": 35121
     },
     {
-      "epoch": 0.9645986103100711,
+      "epoch": 0.9966515323496027,
       "grad_norm": 0.0,
-      "learning_rate": 6.56610066484964e-08,
-      "loss": 0.735,
+      "learning_rate": 5.880775060451705e-10,
+      "loss": 0.7378,
       "step": 35122
     },
     {
-      "epoch": 0.9646260745379144,
+      "epoch": 0.9966799091940977,
       "grad_norm": 0.0,
-      "learning_rate": 6.555927860250056e-08,
-      "loss": 0.8502,
+      "learning_rate": 5.781524211379008e-10,
+      "loss": 0.811,
       "step": 35123
     },
     {
-      "epoch": 0.9646535387657575,
+      "epoch": 0.9967082860385925,
       "grad_norm": 0.0,
-      "learning_rate": 6.545762916164932e-08,
-      "loss": 0.7506,
+      "learning_rate": 5.683118017973943e-10,
+      "loss": 0.8931,
       "step": 35124
     },
     {
-      "epoch": 0.9646810029936008,
+      "epoch": 0.9967366628830874,
       "grad_norm": 0.0,
-      "learning_rate": 6.53560583267443e-08,
-      "loss": 0.7707,
+      "learning_rate": 5.585556481046972e-10,
+      "loss": 0.6946,
       "step": 35125
     },
     {
-      "epoch": 0.9647084672214441,
+      "epoch": 0.9967650397275823,
       "grad_norm": 0.0,
-      "learning_rate": 6.525456609859148e-08,
-      "loss": 0.8239,
+      "learning_rate": 5.488839601441864e-10,
+      "loss": 0.6792,
       "step": 35126
     },
     {
-      "epoch": 0.9647359314492873,
+      "epoch": 0.9967934165720772,
       "grad_norm": 0.0,
-      "learning_rate": 6.515315247799359e-08,
-      "loss": 0.8409,
+      "learning_rate": 5.392967379969083e-10,
+      "loss": 0.8935,
       "step": 35127
     },
     {
-      "epoch": 0.9647633956771305,
+      "epoch": 0.9968217934165721,
       "grad_norm": 0.0,
-      "learning_rate": 6.505181746575106e-08,
-      "loss": 0.914,
+      "learning_rate": 5.297939817427989e-10,
+      "loss": 0.6401,
       "step": 35128
     },
     {
-      "epoch": 0.9647908599049738,
+      "epoch": 0.9968501702610669,
       "grad_norm": 0.0,
-      "learning_rate": 6.495056106266773e-08,
-      "loss": 0.8283,
+      "learning_rate": 5.203756914640146e-10,
+      "loss": 0.8013,
       "step": 35129
     },
     {
-      "epoch": 0.964818324132817,
+      "epoch": 0.9968785471055619,
       "grad_norm": 0.0,
-      "learning_rate": 6.484938326954293e-08,
-      "loss": 0.7654,
+      "learning_rate": 5.110418672382711e-10,
+      "loss": 0.8108,
       "step": 35130
     },
     {
-      "epoch": 0.9648457883606603,
+      "epoch": 0.9969069239500568,
       "grad_norm": 0.0,
-      "learning_rate": 6.474828408717937e-08,
-      "loss": 0.7739,
+      "learning_rate": 5.017925091455044e-10,
+      "loss": 0.8274,
       "step": 35131
     },
     {
-      "epoch": 0.9648732525885034,
+      "epoch": 0.9969353007945516,
       "grad_norm": 0.0,
-      "learning_rate": 6.464726351637529e-08,
-      "loss": 0.8487,
+      "learning_rate": 4.926276172645405e-10,
+      "loss": 0.8002,
       "step": 35132
     },
     {
-      "epoch": 0.9649007168163467,
+      "epoch": 0.9969636776390466,
       "grad_norm": 0.0,
-      "learning_rate": 6.454632155793117e-08,
-      "loss": 0.7485,
+      "learning_rate": 4.835471916708745e-10,
+      "loss": 0.7847,
       "step": 35133
     },
     {
-      "epoch": 0.96492818104419,
+      "epoch": 0.9969920544835414,
       "grad_norm": 0.0,
-      "learning_rate": 6.444545821264525e-08,
-      "loss": 0.9106,
+      "learning_rate": 4.745512324422219e-10,
+      "loss": 0.7194,
       "step": 35134
     },
     {
-      "epoch": 0.9649556452720331,
+      "epoch": 0.9970204313280363,
       "grad_norm": 0.0,
-      "learning_rate": 6.43446734813158e-08,
-      "loss": 0.7581,
+      "learning_rate": 4.656397396540779e-10,
+      "loss": 0.8192,
       "step": 35135
     },
     {
-      "epoch": 0.9649831094998764,
+      "epoch": 0.9970488081725312,
       "grad_norm": 0.0,
-      "learning_rate": 6.424396736473881e-08,
-      "loss": 0.9134,
+      "learning_rate": 4.56812713383048e-10,
+      "loss": 0.7574,
       "step": 35136
     },
     {
-      "epoch": 0.9650105737277197,
+      "epoch": 0.9970771850170261,
       "grad_norm": 0.0,
-      "learning_rate": 6.414333986371257e-08,
-      "loss": 0.7278,
+      "learning_rate": 4.480701537024068e-10,
+      "loss": 0.8261,
       "step": 35137
     },
     {
-      "epoch": 0.9650380379555629,
+      "epoch": 0.997105561861521,
       "grad_norm": 0.0,
-      "learning_rate": 6.40427909790331e-08,
-      "loss": 0.7689,
+      "learning_rate": 4.394120606876495e-10,
+      "loss": 0.8367,
       "step": 35138
     },
     {
-      "epoch": 0.9650655021834061,
+      "epoch": 0.9971339387060159,
       "grad_norm": 0.0,
-      "learning_rate": 6.394232071149641e-08,
-      "loss": 0.8067,
+      "learning_rate": 4.3083843440872017e-10,
+      "loss": 0.9738,
       "step": 35139
     },
     {
-      "epoch": 0.9650929664112493,
+      "epoch": 0.9971623155505108,
       "grad_norm": 0.0,
-      "learning_rate": 6.384192906189745e-08,
-      "loss": 0.7838,
+      "learning_rate": 4.22349274941114e-10,
+      "loss": 0.7959,
       "step": 35140
     },
     {
-      "epoch": 0.9651204306390926,
+      "epoch": 0.9971906923950057,
       "grad_norm": 0.0,
-      "learning_rate": 6.374161603102891e-08,
-      "loss": 0.8363,
+      "learning_rate": 4.139445823558852e-10,
+      "loss": 0.7112,
       "step": 35141
     },
     {
-      "epoch": 0.9651478948669359,
+      "epoch": 0.9972190692395005,
       "grad_norm": 0.0,
-      "learning_rate": 6.364138161968569e-08,
-      "loss": 0.8574,
+      "learning_rate": 4.056243567229778e-10,
+      "loss": 0.7633,
       "step": 35142
     },
     {
-      "epoch": 0.965175359094779,
+      "epoch": 0.9972474460839955,
       "grad_norm": 0.0,
-      "learning_rate": 6.354122582866162e-08,
-      "loss": 0.7365,
+      "learning_rate": 3.973885981134462e-10,
+      "loss": 0.8474,
       "step": 35143
     },
     {
-      "epoch": 0.9652028233226223,
+      "epoch": 0.9972758229284904,
       "grad_norm": 0.0,
-      "learning_rate": 6.344114865874717e-08,
-      "loss": 0.8198,
+      "learning_rate": 3.892373065961241e-10,
+      "loss": 0.7817,
       "step": 35144
     },
     {
-      "epoch": 0.9652302875504655,
+      "epoch": 0.9973041997729852,
       "grad_norm": 0.0,
-      "learning_rate": 6.334115011073616e-08,
-      "loss": 0.8862,
+      "learning_rate": 3.8117048224095564e-10,
+      "loss": 0.8903,
       "step": 35145
     },
     {
-      "epoch": 0.9652577517783087,
+      "epoch": 0.9973325766174801,
       "grad_norm": 0.0,
-      "learning_rate": 6.324123018542016e-08,
-      "loss": 0.8398,
+      "learning_rate": 3.7318812511566436e-10,
+      "loss": 0.7797,
       "step": 35146
     },
     {
-      "epoch": 0.965285216006152,
+      "epoch": 0.9973609534619751,
       "grad_norm": 0.0,
-      "learning_rate": 6.314138888358745e-08,
-      "loss": 0.9388,
+      "learning_rate": 3.652902352868637e-10,
+      "loss": 0.8985,
       "step": 35147
     },
     {
-      "epoch": 0.9653126802339952,
+      "epoch": 0.9973893303064699,
       "grad_norm": 0.0,
-      "learning_rate": 6.304162620602849e-08,
-      "loss": 0.7953,
+      "learning_rate": 3.5747681282227717e-10,
+      "loss": 0.723,
       "step": 35148
     },
     {
-      "epoch": 0.9653401444618385,
+      "epoch": 0.9974177071509648,
       "grad_norm": 0.0,
-      "learning_rate": 6.294194215353488e-08,
-      "loss": 0.8275,
+      "learning_rate": 3.4974785778740807e-10,
+      "loss": 0.7831,
       "step": 35149
     },
     {
-      "epoch": 0.9653676086896817,
+      "epoch": 0.9974460839954598,
       "grad_norm": 0.0,
-      "learning_rate": 6.284233672689266e-08,
-      "loss": 0.7302,
+      "learning_rate": 3.4210337024886965e-10,
+      "loss": 0.7457,
       "step": 35150
     },
     {
-      "epoch": 0.9653950729175249,
+      "epoch": 0.9974744608399546,
       "grad_norm": 0.0,
-      "learning_rate": 6.27428099268923e-08,
-      "loss": 0.8498,
+      "learning_rate": 3.345433502688344e-10,
+      "loss": 0.7769,
       "step": 35151
     },
     {
-      "epoch": 0.9654225371453682,
+      "epoch": 0.9975028376844495,
       "grad_norm": 0.0,
-      "learning_rate": 6.264336175431984e-08,
-      "loss": 0.7177,
+      "learning_rate": 3.270677979128056e-10,
+      "loss": 0.8118,
       "step": 35152
     },
     {
-      "epoch": 0.9654500013732114,
+      "epoch": 0.9975312145289443,
       "grad_norm": 0.0,
-      "learning_rate": 6.254399220996244e-08,
-      "loss": 0.8299,
+      "learning_rate": 3.1967671324406593e-10,
+      "loss": 0.8909,
       "step": 35153
     },
     {
-      "epoch": 0.9654774656010546,
+      "epoch": 0.9975595913734393,
       "grad_norm": 0.0,
-      "learning_rate": 6.244470129460611e-08,
-      "loss": 0.8035,
+      "learning_rate": 3.1237009632367756e-10,
+      "loss": 0.7785,
       "step": 35154
     },
     {
-      "epoch": 0.9655049298288979,
+      "epoch": 0.9975879682179342,
       "grad_norm": 0.0,
-      "learning_rate": 6.234548900903581e-08,
-      "loss": 0.764,
+      "learning_rate": 3.0514794721492325e-10,
+      "loss": 0.7729,
       "step": 35155
     },
     {
-      "epoch": 0.9655323940567411,
+      "epoch": 0.997616345062429,
       "grad_norm": 0.0,
-      "learning_rate": 6.224635535403867e-08,
-      "loss": 0.8599,
+      "learning_rate": 2.9801026597775507e-10,
+      "loss": 0.8756,
       "step": 35156
     },
     {
-      "epoch": 0.9655598582845843,
+      "epoch": 0.997644721906924,
       "grad_norm": 0.0,
-      "learning_rate": 6.214730033039851e-08,
-      "loss": 0.6949,
+      "learning_rate": 2.9095705267323524e-10,
+      "loss": 0.7903,
       "step": 35157
     },
     {
-      "epoch": 0.9655873225124275,
+      "epoch": 0.9976730987514189,
       "grad_norm": 0.0,
-      "learning_rate": 6.204832393889692e-08,
-      "loss": 0.8272,
+      "learning_rate": 2.839883073602057e-10,
+      "loss": 0.7757,
       "step": 35158
     },
     {
-      "epoch": 0.9656147867402708,
+      "epoch": 0.9977014755959137,
       "grad_norm": 0.0,
-      "learning_rate": 6.194942618031996e-08,
-      "loss": 0.778,
+      "learning_rate": 2.7710403009750807e-10,
+      "loss": 0.7815,
       "step": 35159
     },
     {
-      "epoch": 0.9656422509681141,
+      "epoch": 0.9977298524404086,
       "grad_norm": 0.0,
-      "learning_rate": 6.185060705544809e-08,
-      "loss": 0.8433,
+      "learning_rate": 2.703042209439843e-10,
+      "loss": 0.8407,
       "step": 35160
     },
     {
-      "epoch": 0.9656697151959572,
+      "epoch": 0.9977582292849035,
       "grad_norm": 0.0,
-      "learning_rate": 6.175186656506404e-08,
-      "loss": 0.7861,
+      "learning_rate": 2.6358887995625584e-10,
+      "loss": 0.8245,
       "step": 35161
     },
     {
-      "epoch": 0.9656971794238005,
+      "epoch": 0.9977866061293984,
       "grad_norm": 0.0,
-      "learning_rate": 6.165320470994829e-08,
-      "loss": 0.658,
+      "learning_rate": 2.5695800719205413e-10,
+      "loss": 0.8151,
       "step": 35162
     },
     {
-      "epoch": 0.9657246436516438,
+      "epoch": 0.9978149829738933,
       "grad_norm": 0.0,
-      "learning_rate": 6.155462149088353e-08,
-      "loss": 0.7541,
+      "learning_rate": 2.5041160270689035e-10,
+      "loss": 0.7904,
       "step": 35163
     },
     {
-      "epoch": 0.965752107879487,
+      "epoch": 0.9978433598183882,
       "grad_norm": 0.0,
-      "learning_rate": 6.145611690864695e-08,
-      "loss": 0.7504,
+      "learning_rate": 2.4394966655627573e-10,
+      "loss": 0.87,
       "step": 35164
     },
     {
-      "epoch": 0.9657795721073302,
+      "epoch": 0.9978717366628831,
       "grad_norm": 0.0,
-      "learning_rate": 6.135769096402011e-08,
-      "loss": 0.7719,
+      "learning_rate": 2.375721987935009e-10,
+      "loss": 0.8918,
       "step": 35165
     },
     {
-      "epoch": 0.9658070363351734,
+      "epoch": 0.997900113507378,
       "grad_norm": 0.0,
-      "learning_rate": 6.12593436577813e-08,
-      "loss": 0.8673,
+      "learning_rate": 2.3127919947407707e-10,
+      "loss": 0.7187,
       "step": 35166
     },
     {
-      "epoch": 0.9658345005630167,
+      "epoch": 0.9979284903518729,
       "grad_norm": 0.0,
-      "learning_rate": 6.116107499070767e-08,
-      "loss": 0.8535,
+      "learning_rate": 2.250706686512949e-10,
+      "loss": 0.8515,
       "step": 35167
     },
     {
-      "epoch": 0.96586196479086,
+      "epoch": 0.9979568671963678,
       "grad_norm": 0.0,
-      "learning_rate": 6.106288496357748e-08,
-      "loss": 0.8794,
+      "learning_rate": 2.1894660637622467e-10,
+      "loss": 0.8345,
       "step": 35168
     },
     {
-      "epoch": 0.9658894290187031,
+      "epoch": 0.9979852440408626,
       "grad_norm": 0.0,
-      "learning_rate": 6.0964773577169e-08,
-      "loss": 0.6927,
+      "learning_rate": 2.1290701270104686e-10,
+      "loss": 0.9273,
       "step": 35169
     },
     {
-      "epoch": 0.9659168932465464,
+      "epoch": 0.9980136208853575,
       "grad_norm": 0.0,
-      "learning_rate": 6.086674083225608e-08,
-      "loss": 0.7855,
+      "learning_rate": 2.0695188767683173e-10,
+      "loss": 0.8216,
       "step": 35170
     },
     {
-      "epoch": 0.9659443574743896,
+      "epoch": 0.9980419977298525,
       "grad_norm": 0.0,
-      "learning_rate": 6.076878672961583e-08,
-      "loss": 0.7964,
+      "learning_rate": 2.010812313546495e-10,
+      "loss": 0.7291,
       "step": 35171
     },
     {
-      "epoch": 0.9659718217022328,
+      "epoch": 0.9980703745743473,
       "grad_norm": 0.0,
-      "learning_rate": 6.067091127002322e-08,
-      "loss": 0.7927,
+      "learning_rate": 1.9529504378223984e-10,
+      "loss": 0.8472,
       "step": 35172
     },
     {
-      "epoch": 0.9659992859300761,
+      "epoch": 0.9980987514188422,
       "grad_norm": 0.0,
-      "learning_rate": 6.057311445425207e-08,
-      "loss": 0.9317,
+      "learning_rate": 1.8959332501067294e-10,
+      "loss": 0.9176,
       "step": 35173
     },
     {
-      "epoch": 0.9660267501579193,
+      "epoch": 0.9981271282633372,
       "grad_norm": 0.0,
-      "learning_rate": 6.04753962830773e-08,
-      "loss": 0.7404,
+      "learning_rate": 1.8397607508768844e-10,
+      "loss": 0.7882,
       "step": 35174
     },
     {
-      "epoch": 0.9660542143857626,
+      "epoch": 0.998155505107832,
       "grad_norm": 0.0,
-      "learning_rate": 6.037775675727165e-08,
-      "loss": 0.7825,
+      "learning_rate": 1.7844329405880544e-10,
+      "loss": 0.9381,
       "step": 35175
     },
     {
-      "epoch": 0.9660816786136058,
+      "epoch": 0.9981838819523269,
       "grad_norm": 0.0,
-      "learning_rate": 6.02801958776067e-08,
-      "loss": 0.7506,
+      "learning_rate": 1.7299498197287378e-10,
+      "loss": 0.9168,
       "step": 35176
     },
     {
-      "epoch": 0.966109142841449,
+      "epoch": 0.9982122587968217,
       "grad_norm": 0.0,
-      "learning_rate": 6.018271364485518e-08,
-      "loss": 0.8295,
+      "learning_rate": 1.676311388754126e-10,
+      "loss": 0.7506,
       "step": 35177
     },
     {
-      "epoch": 0.9661366070692923,
+      "epoch": 0.9982406356413167,
       "grad_norm": 0.0,
-      "learning_rate": 6.008531005978868e-08,
-      "loss": 0.6483,
+      "learning_rate": 1.6235176481083082e-10,
+      "loss": 0.8104,
       "step": 35178
     },
     {
-      "epoch": 0.9661640712971354,
+      "epoch": 0.9982690124858116,
       "grad_norm": 0.0,
-      "learning_rate": 5.998798512317661e-08,
-      "loss": 0.8426,
+      "learning_rate": 1.5715685982464756e-10,
+      "loss": 0.8623,
       "step": 35179
     },
     {
-      "epoch": 0.9661915355249787,
+      "epoch": 0.9982973893303064,
       "grad_norm": 0.0,
-      "learning_rate": 5.989073883579055e-08,
-      "loss": 0.8196,
+      "learning_rate": 1.5204642396127178e-10,
+      "loss": 0.7165,
       "step": 35180
     },
     {
-      "epoch": 0.966218999752822,
+      "epoch": 0.9983257661748014,
       "grad_norm": 0.0,
-      "learning_rate": 5.97935711983999e-08,
-      "loss": 0.7188,
+      "learning_rate": 1.4702045726178172e-10,
+      "loss": 0.8167,
       "step": 35181
     },
     {
-      "epoch": 0.9662464639806652,
+      "epoch": 0.9983541430192963,
       "grad_norm": 0.0,
-      "learning_rate": 5.969648221177294e-08,
-      "loss": 0.7601,
+      "learning_rate": 1.4207895977058627e-10,
+      "loss": 0.7799,
       "step": 35182
     },
     {
-      "epoch": 0.9662739282085084,
+      "epoch": 0.9983825198637911,
       "grad_norm": 0.0,
-      "learning_rate": 5.9599471876677914e-08,
-      "loss": 0.8103,
+      "learning_rate": 1.3722193152876373e-10,
+      "loss": 0.7487,
       "step": 35183
     },
     {
-      "epoch": 0.9663013924363516,
+      "epoch": 0.9984108967082861,
       "grad_norm": 0.0,
-      "learning_rate": 5.9502540193882016e-08,
-      "loss": 0.7944,
+      "learning_rate": 1.3244937257739233e-10,
+      "loss": 0.7648,
       "step": 35184
     },
     {
-      "epoch": 0.9663288566641949,
+      "epoch": 0.998439273552781,
       "grad_norm": 0.0,
-      "learning_rate": 5.94056871641524e-08,
-      "loss": 0.8378,
+      "learning_rate": 1.2776128295644008e-10,
+      "loss": 0.7661,
       "step": 35185
     },
     {
-      "epoch": 0.9663563208920382,
+      "epoch": 0.9984676503972758,
       "grad_norm": 0.0,
-      "learning_rate": 5.9308912788255124e-08,
-      "loss": 0.6832,
+      "learning_rate": 1.2315766270698527e-10,
+      "loss": 0.861,
       "step": 35186
     },
     {
-      "epoch": 0.9663837851198813,
+      "epoch": 0.9984960272417707,
       "grad_norm": 0.0,
-      "learning_rate": 5.9212217066957346e-08,
-      "loss": 0.7918,
+      "learning_rate": 1.186385118656652e-10,
+      "loss": 0.7034,
       "step": 35187
     },
     {
-      "epoch": 0.9664112493477246,
+      "epoch": 0.9985244040862656,
       "grad_norm": 0.0,
-      "learning_rate": 5.91156000010229e-08,
-      "loss": 0.7856,
+      "learning_rate": 1.1420383047133777e-10,
+      "loss": 0.743,
       "step": 35188
     },
     {
-      "epoch": 0.9664387135755679,
+      "epoch": 0.9985527809307605,
       "grad_norm": 0.0,
-      "learning_rate": 5.901906159121673e-08,
-      "loss": 0.807,
+      "learning_rate": 1.0985361856286069e-10,
+      "loss": 0.8079,
       "step": 35189
     },
     {
-      "epoch": 0.966466177803411,
+      "epoch": 0.9985811577752554,
       "grad_norm": 0.0,
-      "learning_rate": 5.892260183830267e-08,
-      "loss": 0.7451,
+      "learning_rate": 1.0558787617576117e-10,
+      "loss": 0.8792,
       "step": 35190
     },
     {
-      "epoch": 0.9664936420312543,
+      "epoch": 0.9986095346197503,
       "grad_norm": 0.0,
-      "learning_rate": 5.882622074304345e-08,
-      "loss": 0.8446,
+      "learning_rate": 1.0140660334556629e-10,
+      "loss": 0.8684,
       "step": 35191
     },
     {
-      "epoch": 0.9665211062590975,
+      "epoch": 0.9986379114642452,
       "grad_norm": 0.0,
-      "learning_rate": 5.8729918306201786e-08,
-      "loss": 0.8141,
+      "learning_rate": 9.730980010891345e-11,
+      "loss": 0.7978,
       "step": 35192
     },
     {
-      "epoch": 0.9665485704869408,
+      "epoch": 0.99866628830874,
       "grad_norm": 0.0,
-      "learning_rate": 5.863369452854151e-08,
-      "loss": 0.7442,
+      "learning_rate": 9.329746649910932e-11,
+      "loss": 0.7683,
       "step": 35193
     },
     {
-      "epoch": 0.966576034714784,
+      "epoch": 0.9986946651532349,
       "grad_norm": 0.0,
-      "learning_rate": 5.85375494108209e-08,
-      "loss": 0.7695,
+      "learning_rate": 8.936960255168104e-11,
+      "loss": 0.7395,
       "step": 35194
     },
     {
-      "epoch": 0.9666034989426272,
+      "epoch": 0.9987230419977299,
       "grad_norm": 0.0,
-      "learning_rate": 5.8441482953801584e-08,
-      "loss": 0.8491,
+      "learning_rate": 8.552620829882508e-11,
+      "loss": 0.8605,
       "step": 35195
     },
     {
-      "epoch": 0.9666309631704705,
+      "epoch": 0.9987514188422247,
       "grad_norm": 0.0,
-      "learning_rate": 5.834549515824517e-08,
-      "loss": 0.7382,
+      "learning_rate": 8.17672837727379e-11,
+      "loss": 0.7204,
       "step": 35196
     },
     {
-      "epoch": 0.9666584273983136,
+      "epoch": 0.9987797956867196,
       "grad_norm": 0.0,
-      "learning_rate": 5.824958602490993e-08,
-      "loss": 0.7699,
+      "learning_rate": 7.809282900561598e-11,
+      "loss": 0.7542,
       "step": 35197
     },
     {
-      "epoch": 0.9666858916261569,
+      "epoch": 0.9988081725312146,
       "grad_norm": 0.0,
-      "learning_rate": 5.815375555455527e-08,
-      "loss": 0.6774,
+      "learning_rate": 7.450284402854558e-11,
+      "loss": 0.745,
       "step": 35198
     },
     {
-      "epoch": 0.9667133558540002,
+      "epoch": 0.9988365493757094,
       "grad_norm": 0.0,
-      "learning_rate": 5.805800374793946e-08,
-      "loss": 0.7759,
+      "learning_rate": 7.099732887150268e-11,
+      "loss": 0.722,
       "step": 35199
     },
     {
-      "epoch": 0.9667408200818434,
+      "epoch": 0.9988649262202043,
       "grad_norm": 0.0,
-      "learning_rate": 5.796233060581968e-08,
-      "loss": 0.7317,
+      "learning_rate": 6.757628356335311e-11,
+      "loss": 0.9218,
       "step": 35200
     },
     {
-      "epoch": 0.9667682843096866,
+      "epoch": 0.9988933030646993,
       "grad_norm": 0.0,
-      "learning_rate": 5.7866736128952e-08,
-      "loss": 0.8359,
+      "learning_rate": 6.42397081351831e-11,
+      "loss": 0.7765,
       "step": 35201
     },
     {
-      "epoch": 0.9667957485375299,
+      "epoch": 0.9989216799091941,
       "grad_norm": 0.0,
-      "learning_rate": 5.777122031809468e-08,
-      "loss": 0.8099,
+      "learning_rate": 6.0987602613638e-11,
+      "loss": 0.6866,
       "step": 35202
     },
     {
-      "epoch": 0.9668232127653731,
+      "epoch": 0.998950056753689,
       "grad_norm": 0.0,
-      "learning_rate": 5.767578317400269e-08,
-      "loss": 0.9005,
+      "learning_rate": 5.781996702647341e-11,
+      "loss": 0.7686,
       "step": 35203
     },
     {
-      "epoch": 0.9668506769932164,
+      "epoch": 0.9989784335981838,
       "grad_norm": 0.0,
-      "learning_rate": 5.758042469743208e-08,
-      "loss": 0.8401,
+      "learning_rate": 5.473680140033466e-11,
+      "loss": 0.8163,
       "step": 35204
     },
     {
-      "epoch": 0.9668781412210595,
+      "epoch": 0.9990068104426788,
       "grad_norm": 0.0,
-      "learning_rate": 5.748514488913559e-08,
-      "loss": 0.8616,
+      "learning_rate": 5.1738105760756884e-11,
+      "loss": 0.7102,
       "step": 35205
     },
     {
-      "epoch": 0.9669056054489028,
+      "epoch": 0.9990351872871737,
       "grad_norm": 0.0,
-      "learning_rate": 5.738994374986817e-08,
-      "loss": 0.7796,
+      "learning_rate": 4.882388013438544e-11,
+      "loss": 0.8565,
       "step": 35206
     },
     {
-      "epoch": 0.9669330696767461,
+      "epoch": 0.9990635641316685,
       "grad_norm": 0.0,
-      "learning_rate": 5.729482128038366e-08,
-      "loss": 0.8617,
+      "learning_rate": 4.5994124544535004e-11,
+      "loss": 0.773,
       "step": 35207
     },
     {
-      "epoch": 0.9669605339045892,
+      "epoch": 0.9990919409761635,
       "grad_norm": 0.0,
-      "learning_rate": 5.719977748143257e-08,
-      "loss": 0.8299,
+      "learning_rate": 4.324883901674071e-11,
+      "loss": 0.7847,
       "step": 35208
     },
     {
-      "epoch": 0.9669879981324325,
+      "epoch": 0.9991203178206584,
       "grad_norm": 0.0,
-      "learning_rate": 5.7104812353768744e-08,
-      "loss": 0.8014,
+      "learning_rate": 4.058802357209679e-11,
+      "loss": 0.7411,
       "step": 35209
     },
     {
-      "epoch": 0.9670154623602757,
+      "epoch": 0.9991486946651532,
       "grad_norm": 0.0,
-      "learning_rate": 5.70099258981438e-08,
-      "loss": 0.8075,
+      "learning_rate": 3.801167823502816e-11,
+      "loss": 0.7918,
       "step": 35210
     },
     {
-      "epoch": 0.967042926588119,
+      "epoch": 0.9991770715096481,
       "grad_norm": 0.0,
-      "learning_rate": 5.6915118115307146e-08,
-      "loss": 0.7895,
+      "learning_rate": 3.5519803025518826e-11,
+      "loss": 0.8455,
       "step": 35211
     },
     {
-      "epoch": 0.9670703908159622,
+      "epoch": 0.999205448354143,
       "grad_norm": 0.0,
-      "learning_rate": 5.682038900601039e-08,
-      "loss": 0.7339,
+      "learning_rate": 3.311239796577326e-11,
+      "loss": 0.7619,
       "step": 35212
     },
     {
-      "epoch": 0.9670978550438054,
+      "epoch": 0.9992338251986379,
       "grad_norm": 0.0,
-      "learning_rate": 5.6725738571001834e-08,
-      "loss": 0.8075,
+      "learning_rate": 3.078946307577546e-11,
+      "loss": 0.86,
       "step": 35213
     },
     {
-      "epoch": 0.9671253192716487,
+      "epoch": 0.9992622020431328,
       "grad_norm": 0.0,
-      "learning_rate": 5.663116681103087e-08,
-      "loss": 0.7883,
+      "learning_rate": 2.8550998375509452e-11,
+      "loss": 0.7734,
       "step": 35214
     },
     {
-      "epoch": 0.967152783499492,
+      "epoch": 0.9992905788876277,
       "grad_norm": 0.0,
-      "learning_rate": 5.653667372684579e-08,
-      "loss": 0.8117,
+      "learning_rate": 2.6397003883849027e-11,
+      "loss": 0.7873,
       "step": 35215
     },
     {
-      "epoch": 0.9671802477273351,
+      "epoch": 0.9993189557321226,
       "grad_norm": 0.0,
-      "learning_rate": 5.644225931919489e-08,
-      "loss": 0.7517,
+      "learning_rate": 2.432747961855775e-11,
+      "loss": 0.7891,
       "step": 35216
     },
     {
-      "epoch": 0.9672077119551784,
+      "epoch": 0.9993473325766175,
       "grad_norm": 0.0,
-      "learning_rate": 5.634792358882424e-08,
-      "loss": 0.7591,
+      "learning_rate": 2.2342425597399187e-11,
+      "loss": 0.8102,
       "step": 35217
     },
     {
-      "epoch": 0.9672351761830216,
+      "epoch": 0.9993757094211124,
       "grad_norm": 0.0,
-      "learning_rate": 5.62536665364799e-08,
-      "loss": 0.8098,
+      "learning_rate": 2.0441841837026688e-11,
+      "loss": 0.814,
       "step": 35218
     },
     {
-      "epoch": 0.9672626404108648,
+      "epoch": 0.9994040862656073,
       "grad_norm": 0.0,
-      "learning_rate": 5.6159488162909056e-08,
-      "loss": 0.7322,
+      "learning_rate": 1.8625728352983375e-11,
+      "loss": 0.7773,
       "step": 35219
     },
     {
-      "epoch": 0.9672901046387081,
+      "epoch": 0.9994324631101021,
       "grad_norm": 0.0,
-      "learning_rate": 5.6065388468855564e-08,
-      "loss": 0.8557,
+      "learning_rate": 1.689408516192259e-11,
+      "loss": 0.8777,
       "step": 35220
     },
     {
-      "epoch": 0.9673175688665513,
+      "epoch": 0.999460839954597,
       "grad_norm": 0.0,
-      "learning_rate": 5.5971367455065486e-08,
-      "loss": 0.7272,
+      "learning_rate": 1.5246912277167014e-11,
+      "loss": 0.811,
       "step": 35221
     },
     {
-      "epoch": 0.9673450330943946,
+      "epoch": 0.999489216799092,
       "grad_norm": 0.0,
-      "learning_rate": 5.587742512228156e-08,
-      "loss": 0.8775,
+      "learning_rate": 1.3684209713149543e-11,
+      "loss": 0.7946,
       "step": 35222
     },
     {
-      "epoch": 0.9673724973222377,
+      "epoch": 0.9995175936435868,
       "grad_norm": 0.0,
-      "learning_rate": 5.578356147124764e-08,
-      "loss": 0.715,
+      "learning_rate": 1.2205977483192855e-11,
+      "loss": 0.8295,
       "step": 35223
     },
     {
-      "epoch": 0.967399961550081,
+      "epoch": 0.9995459704880817,
       "grad_norm": 0.0,
-      "learning_rate": 5.568977650270646e-08,
-      "loss": 0.7787,
+      "learning_rate": 1.0812215599509402e-11,
+      "loss": 0.8495,
       "step": 35224
     },
     {
-      "epoch": 0.9674274257779243,
+      "epoch": 0.9995743473325767,
       "grad_norm": 0.0,
-      "learning_rate": 5.559607021739854e-08,
-      "loss": 0.8381,
+      "learning_rate": 9.502924074311637e-12,
+      "loss": 0.7965,
       "step": 35225
     },
     {
-      "epoch": 0.9674548900057675,
+      "epoch": 0.9996027241770715,
       "grad_norm": 0.0,
-      "learning_rate": 5.550244261606774e-08,
-      "loss": 0.8222,
+      "learning_rate": 8.278102917591569e-12,
+      "loss": 0.7718,
       "step": 35226
     },
     {
-      "epoch": 0.9674823542336107,
+      "epoch": 0.9996311010215664,
       "grad_norm": 0.0,
-      "learning_rate": 5.5408893699453456e-08,
-      "loss": 0.7812,
+      "learning_rate": 7.137752141561649e-12,
+      "loss": 0.7643,
       "step": 35227
     },
     {
-      "epoch": 0.967509818461454,
+      "epoch": 0.9996594778660612,
       "grad_norm": 0.0,
-      "learning_rate": 5.5315423468296216e-08,
-      "loss": 0.8203,
+      "learning_rate": 6.08187175399344e-12,
+      "loss": 0.8705,
       "step": 35228
     },
     {
-      "epoch": 0.9675372826892972,
+      "epoch": 0.9996878547105562,
       "grad_norm": 0.0,
-      "learning_rate": 5.522203192333542e-08,
-      "loss": 0.8399,
+      "learning_rate": 5.110461764878949e-12,
+      "loss": 0.7204,
       "step": 35229
     },
     {
-      "epoch": 0.9675647469171405,
+      "epoch": 0.9997162315550511,
       "grad_norm": 0.0,
-      "learning_rate": 5.512871906531048e-08,
-      "loss": 0.7858,
+      "learning_rate": 4.223522181989736e-12,
+      "loss": 0.8001,
       "step": 35230
     },
     {
-      "epoch": 0.9675922111449836,
+      "epoch": 0.9997446083995459,
       "grad_norm": 0.0,
-      "learning_rate": 5.503548489495969e-08,
-      "loss": 0.8433,
+      "learning_rate": 3.421053013097364e-12,
+      "loss": 0.7357,
       "step": 35231
     },
     {
-      "epoch": 0.9676196753728269,
+      "epoch": 0.9997729852440409,
       "grad_norm": 0.0,
-      "learning_rate": 5.4942329413019134e-08,
-      "loss": 0.8279,
+      "learning_rate": 2.7030542648631697e-12,
+      "loss": 0.862,
       "step": 35232
     },
     {
-      "epoch": 0.9676471396006702,
+      "epoch": 0.9998013620885358,
       "grad_norm": 0.0,
-      "learning_rate": 5.484925262022822e-08,
-      "loss": 0.7857,
+      "learning_rate": 2.0695259439484917e-12,
+      "loss": 0.9245,
       "step": 35233
     },
     {
-      "epoch": 0.9676746038285133,
+      "epoch": 0.9998297389330306,
       "grad_norm": 0.0,
-      "learning_rate": 5.475625451732192e-08,
-      "loss": 0.8845,
+      "learning_rate": 1.5204680536839989e-12,
+      "loss": 0.6937,
       "step": 35234
     },
     {
-      "epoch": 0.9677020680563566,
+      "epoch": 0.9998581157775256,
       "grad_norm": 0.0,
-      "learning_rate": 5.466333510503741e-08,
-      "loss": 0.7577,
+      "learning_rate": 1.0558806018412526e-12,
+      "loss": 0.8098,
       "step": 35235
     },
     {
-      "epoch": 0.9677295322841998,
+      "epoch": 0.9998864926220205,
       "grad_norm": 0.0,
-      "learning_rate": 5.457049438410744e-08,
-      "loss": 0.6893,
+      "learning_rate": 6.75763589530476e-13,
+      "loss": 0.8111,
       "step": 35236
     },
     {
-      "epoch": 0.967756996512043,
+      "epoch": 0.9999148694665153,
       "grad_norm": 0.0,
-      "learning_rate": 5.447773235526921e-08,
-      "loss": 0.7789,
+      "learning_rate": 3.801170211925609e-13,
+      "loss": 0.782,
       "step": 35237
     },
     {
-      "epoch": 0.9677844607398863,
+      "epoch": 0.9999432463110102,
       "grad_norm": 0.0,
-      "learning_rate": 5.4385049019256565e-08,
-      "loss": 0.813,
+      "learning_rate": 1.689408990479535e-13,
+      "loss": 0.7523,
       "step": 35238
     },
     {
-      "epoch": 0.9678119249677295,
+      "epoch": 0.9999716231555051,
       "grad_norm": 0.0,
-      "learning_rate": 5.429244437680115e-08,
-      "loss": 0.815,
+      "learning_rate": 4.2235225317099895e-14,
+      "loss": 0.8682,
       "step": 35239
     },
     {
-      "epoch": 0.9678393891955728,
+      "epoch": 1.0,
       "grad_norm": 0.0,
-      "learning_rate": 5.419991842863681e-08,
-      "loss": 0.7946,
+      "learning_rate": 0.0,
+      "loss": 0.7954,
       "step": 35240
     },
-    {
-      "epoch": 0.967866853423416,
-      "grad_norm": 0.0,
-      "learning_rate": 5.410747117549409e-08,
-      "loss": 0.7245,
-      "step": 35241
-    },
-    {
-      "epoch": 0.9678943176512592,
-      "grad_norm": 0.0,
-      "learning_rate": 5.401510261810683e-08,
-      "loss": 0.7643,
-      "step": 35242
-    },
-    {
-      "epoch": 0.9679217818791025,
-      "grad_norm": 0.0,
-      "learning_rate": 5.392281275720446e-08,
-      "loss": 0.8316,
-      "step": 35243
-    },
-    {
-      "epoch": 0.9679492461069457,
-      "grad_norm": 0.0,
-      "learning_rate": 5.38306015935175e-08,
-      "loss": 0.7797,
-      "step": 35244
-    },
-    {
-      "epoch": 0.9679767103347889,
-      "grad_norm": 0.0,
-      "learning_rate": 5.373846912777647e-08,
-      "loss": 0.7828,
-      "step": 35245
-    },
-    {
-      "epoch": 0.9680041745626322,
-      "grad_norm": 0.0,
-      "learning_rate": 5.364641536070858e-08,
-      "loss": 0.8969,
-      "step": 35246
-    },
-    {
-      "epoch": 0.9680316387904754,
-      "grad_norm": 0.0,
-      "learning_rate": 5.355444029304324e-08,
-      "loss": 0.8062,
-      "step": 35247
-    },
-    {
-      "epoch": 0.9680591030183187,
-      "grad_norm": 0.0,
-      "learning_rate": 5.346254392550876e-08,
-      "loss": 0.7024,
-      "step": 35248
-    },
-    {
-      "epoch": 0.9680865672461618,
-      "grad_norm": 0.0,
-      "learning_rate": 5.337072625883011e-08,
-      "loss": 0.7787,
-      "step": 35249
-    },
-    {
-      "epoch": 0.9681140314740051,
-      "grad_norm": 0.0,
-      "learning_rate": 5.32789872937356e-08,
-      "loss": 0.744,
-      "step": 35250
-    },
-    {
-      "epoch": 0.9681414957018484,
-      "grad_norm": 0.0,
-      "learning_rate": 5.3187327030951305e-08,
-      "loss": 0.8761,
-      "step": 35251
-    },
-    {
-      "epoch": 0.9681689599296915,
-      "grad_norm": 0.0,
-      "learning_rate": 5.309574547120222e-08,
-      "loss": 0.8792,
-      "step": 35252
-    },
-    {
-      "epoch": 0.9681964241575348,
-      "grad_norm": 0.0,
-      "learning_rate": 5.30042426152122e-08,
-      "loss": 0.7849,
-      "step": 35253
-    },
-    {
-      "epoch": 0.9682238883853781,
-      "grad_norm": 0.0,
-      "learning_rate": 5.2912818463706215e-08,
-      "loss": 0.8095,
-      "step": 35254
-    },
-    {
-      "epoch": 0.9682513526132213,
-      "grad_norm": 0.0,
-      "learning_rate": 5.282147301740592e-08,
-      "loss": 0.7854,
-      "step": 35255
-    },
-    {
-      "epoch": 0.9682788168410645,
-      "grad_norm": 0.0,
-      "learning_rate": 5.273020627703629e-08,
-      "loss": 0.7419,
-      "step": 35256
-    },
-    {
-      "epoch": 0.9683062810689077,
-      "grad_norm": 0.0,
-      "learning_rate": 5.263901824331785e-08,
-      "loss": 0.786,
-      "step": 35257
-    },
-    {
-      "epoch": 0.968333745296751,
-      "grad_norm": 0.0,
-      "learning_rate": 5.254790891697337e-08,
-      "loss": 0.7513,
-      "step": 35258
-    },
-    {
-      "epoch": 0.9683612095245943,
-      "grad_norm": 0.0,
-      "learning_rate": 5.2456878298723366e-08,
-      "loss": 0.8492,
-      "step": 35259
-    },
-    {
-      "epoch": 0.9683886737524374,
-      "grad_norm": 0.0,
-      "learning_rate": 5.236592638928728e-08,
-      "loss": 0.7603,
-      "step": 35260
-    },
-    {
-      "epoch": 0.9684161379802807,
-      "grad_norm": 0.0,
-      "learning_rate": 5.227505318938453e-08,
-      "loss": 0.8375,
-      "step": 35261
-    },
-    {
-      "epoch": 0.9684436022081239,
-      "grad_norm": 0.0,
-      "learning_rate": 5.218425869973676e-08,
-      "loss": 0.8316,
-      "step": 35262
-    },
-    {
-      "epoch": 0.9684710664359671,
-      "grad_norm": 0.0,
-      "learning_rate": 5.209354292105895e-08,
-      "loss": 0.8339,
-      "step": 35263
-    },
-    {
-      "epoch": 0.9684985306638104,
-      "grad_norm": 0.0,
-      "learning_rate": 5.200290585407164e-08,
-      "loss": 0.8537,
-      "step": 35264
-    },
-    {
-      "epoch": 0.9685259948916536,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1912347499489813e-08,
-      "loss": 0.8396,
-      "step": 35265
-    },
-    {
-      "epoch": 0.9685534591194969,
-      "grad_norm": 0.0,
-      "learning_rate": 5.182186785803178e-08,
-      "loss": 0.8006,
-      "step": 35266
-    },
-    {
-      "epoch": 0.9685809233473401,
-      "grad_norm": 0.0,
-      "learning_rate": 5.173146693041253e-08,
-      "loss": 0.7792,
-      "step": 35267
-    },
-    {
-      "epoch": 0.9686083875751833,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1641144717347047e-08,
-      "loss": 0.7867,
-      "step": 35268
-    },
-    {
-      "epoch": 0.9686358518030266,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1550901219550306e-08,
-      "loss": 0.8835,
-      "step": 35269
-    },
-    {
-      "epoch": 0.9686633160308697,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1460736437737305e-08,
-      "loss": 0.8075,
-      "step": 35270
-    },
-    {
-      "epoch": 0.968690780258713,
-      "grad_norm": 0.0,
-      "learning_rate": 5.137065037262079e-08,
-      "loss": 0.799,
-      "step": 35271
-    },
-    {
-      "epoch": 0.9687182444865563,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1280643024912427e-08,
-      "loss": 0.746,
-      "step": 35272
-    },
-    {
-      "epoch": 0.9687457087143995,
-      "grad_norm": 0.0,
-      "learning_rate": 5.119071439532608e-08,
-      "loss": 0.806,
-      "step": 35273
-    },
-    {
-      "epoch": 0.9687731729422427,
-      "grad_norm": 0.0,
-      "learning_rate": 5.11008644845723e-08,
-      "loss": 0.7478,
-      "step": 35274
-    },
-    {
-      "epoch": 0.9688006371700859,
-      "grad_norm": 0.0,
-      "learning_rate": 5.1011093293362733e-08,
-      "loss": 0.8336,
-      "step": 35275
-    },
-    {
-      "epoch": 0.9688281013979292,
-      "grad_norm": 0.0,
-      "learning_rate": 5.092140082240682e-08,
-      "loss": 0.7227,
-      "step": 35276
-    },
-    {
-      "epoch": 0.9688555656257725,
-      "grad_norm": 0.0,
-      "learning_rate": 5.083178707241509e-08,
-      "loss": 0.8128,
-      "step": 35277
-    },
-    {
-      "epoch": 0.9688830298536156,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0742252044095886e-08,
-      "loss": 0.826,
-      "step": 35278
-    },
-    {
-      "epoch": 0.9689104940814589,
-      "grad_norm": 0.0,
-      "learning_rate": 5.065279573815751e-08,
-      "loss": 0.8574,
-      "step": 35279
-    },
-    {
-      "epoch": 0.9689379583093022,
-      "grad_norm": 0.0,
-      "learning_rate": 5.05634181553083e-08,
-      "loss": 0.7634,
-      "step": 35280
-    },
-    {
-      "epoch": 0.9689654225371453,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0474119296256564e-08,
-      "loss": 0.7584,
-      "step": 35281
-    },
-    {
-      "epoch": 0.9689928867649886,
-      "grad_norm": 0.0,
-      "learning_rate": 5.03848991617073e-08,
-      "loss": 0.7512,
-      "step": 35282
-    },
-    {
-      "epoch": 0.9690203509928318,
-      "grad_norm": 0.0,
-      "learning_rate": 5.02957577523655e-08,
-      "loss": 0.8278,
-      "step": 35283
-    },
-    {
-      "epoch": 0.9690478152206751,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0206695068939495e-08,
-      "loss": 0.7237,
-      "step": 35284
-    },
-    {
-      "epoch": 0.9690752794485183,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0117711112130926e-08,
-      "loss": 0.8177,
-      "step": 35285
-    },
-    {
-      "epoch": 0.9691027436763615,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0028805882645916e-08,
-      "loss": 0.8227,
-      "step": 35286
-    },
-    {
-      "epoch": 0.9691302079042048,
-      "grad_norm": 0.0,
-      "learning_rate": 4.993997938118722e-08,
-      "loss": 0.8682,
-      "step": 35287
-    },
-    {
-      "epoch": 0.969157672132048,
-      "grad_norm": 0.0,
-      "learning_rate": 4.985123160845762e-08,
-      "loss": 0.8879,
-      "step": 35288
-    },
-    {
-      "epoch": 0.9691851363598912,
-      "grad_norm": 0.0,
-      "learning_rate": 4.976256256515988e-08,
-      "loss": 0.7309,
-      "step": 35289
-    },
-    {
-      "epoch": 0.9692126005877345,
-      "grad_norm": 0.0,
-      "learning_rate": 4.967397225199455e-08,
-      "loss": 0.8356,
-      "step": 35290
-    },
-    {
-      "epoch": 0.9692400648155777,
-      "grad_norm": 0.0,
-      "learning_rate": 4.95854606696633e-08,
-      "loss": 0.7659,
-      "step": 35291
-    },
-    {
-      "epoch": 0.969267529043421,
-      "grad_norm": 0.0,
-      "learning_rate": 4.949702781886556e-08,
-      "loss": 0.781,
-      "step": 35292
-    },
-    {
-      "epoch": 0.9692949932712642,
-      "grad_norm": 0.0,
-      "learning_rate": 4.9408673700303e-08,
-      "loss": 0.7936,
-      "step": 35293
-    },
-    {
-      "epoch": 0.9693224574991074,
-      "grad_norm": 0.0,
-      "learning_rate": 4.9320398314672835e-08,
-      "loss": 0.7506,
-      "step": 35294
-    },
-    {
-      "epoch": 0.9693499217269507,
-      "grad_norm": 0.0,
-      "learning_rate": 4.923220166267451e-08,
-      "loss": 0.7937,
-      "step": 35295
-    },
-    {
-      "epoch": 0.9693773859547938,
-      "grad_norm": 0.0,
-      "learning_rate": 4.914408374500523e-08,
-      "loss": 0.7597,
-      "step": 35296
-    },
-    {
-      "epoch": 0.9694048501826371,
-      "grad_norm": 0.0,
-      "learning_rate": 4.905604456236224e-08,
-      "loss": 0.8598,
-      "step": 35297
-    },
-    {
-      "epoch": 0.9694323144104804,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8968084115443847e-08,
-      "loss": 0.8023,
-      "step": 35298
-    },
-    {
-      "epoch": 0.9694597786383236,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8880202404943956e-08,
-      "loss": 0.8365,
-      "step": 35299
-    },
-    {
-      "epoch": 0.9694872428661668,
-      "grad_norm": 0.0,
-      "learning_rate": 4.879239943155756e-08,
-      "loss": 0.8131,
-      "step": 35300
-    },
-    {
-      "epoch": 0.96951470709401,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8704675195981875e-08,
-      "loss": 0.8742,
-      "step": 35301
-    },
-    {
-      "epoch": 0.9695421713218533,
-      "grad_norm": 0.0,
-      "learning_rate": 4.861702969890858e-08,
-      "loss": 0.8001,
-      "step": 35302
-    },
-    {
-      "epoch": 0.9695696355496966,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8529462941032666e-08,
-      "loss": 0.8026,
-      "step": 35303
-    },
-    {
-      "epoch": 0.9695970997775397,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8441974923045806e-08,
-      "loss": 0.7371,
-      "step": 35304
-    },
-    {
-      "epoch": 0.969624564005383,
-      "grad_norm": 0.0,
-      "learning_rate": 4.835456564564189e-08,
-      "loss": 0.7275,
-      "step": 35305
-    },
-    {
-      "epoch": 0.9696520282332263,
-      "grad_norm": 0.0,
-      "learning_rate": 4.826723510951148e-08,
-      "loss": 0.8194,
-      "step": 35306
-    },
-    {
-      "epoch": 0.9696794924610694,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8179983315344017e-08,
-      "loss": 0.7166,
-      "step": 35307
-    },
-    {
-      "epoch": 0.9697069566889127,
-      "grad_norm": 0.0,
-      "learning_rate": 4.809281026383339e-08,
-      "loss": 0.7819,
-      "step": 35308
-    },
-    {
-      "epoch": 0.9697344209167559,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8005715955666834e-08,
-      "loss": 0.832,
-      "step": 35309
-    },
-    {
-      "epoch": 0.9697618851445992,
-      "grad_norm": 0.0,
-      "learning_rate": 4.791870039153379e-08,
-      "loss": 0.7521,
-      "step": 35310
-    },
-    {
-      "epoch": 0.9697893493724424,
-      "grad_norm": 0.0,
-      "learning_rate": 4.7831763572123714e-08,
-      "loss": 0.7819,
-      "step": 35311
-    },
-    {
-      "epoch": 0.9698168136002856,
-      "grad_norm": 0.0,
-      "learning_rate": 4.774490549812272e-08,
-      "loss": 0.8669,
-      "step": 35312
-    },
-    {
-      "epoch": 0.9698442778281289,
-      "grad_norm": 0.0,
-      "learning_rate": 4.7658126170220253e-08,
-      "loss": 0.8646,
-      "step": 35313
-    },
-    {
-      "epoch": 0.969871742055972,
-      "grad_norm": 0.0,
-      "learning_rate": 4.7571425589101326e-08,
-      "loss": 0.8069,
-      "step": 35314
-    },
-    {
-      "epoch": 0.9698992062838153,
-      "grad_norm": 0.0,
-      "learning_rate": 4.748480375545206e-08,
-      "loss": 0.88,
-      "step": 35315
-    },
-    {
-      "epoch": 0.9699266705116586,
-      "grad_norm": 0.0,
-      "learning_rate": 4.739826066995856e-08,
-      "loss": 0.8306,
-      "step": 35316
-    },
-    {
-      "epoch": 0.9699541347395018,
-      "grad_norm": 0.0,
-      "learning_rate": 4.7311796333305845e-08,
-      "loss": 0.8536,
-      "step": 35317
-    },
-    {
-      "epoch": 0.969981598967345,
-      "grad_norm": 0.0,
-      "learning_rate": 4.722541074617781e-08,
-      "loss": 0.8401,
-      "step": 35318
-    },
-    {
-      "epoch": 0.9700090631951883,
-      "grad_norm": 0.0,
-      "learning_rate": 4.7139103909257244e-08,
-      "loss": 0.8228,
-      "step": 35319
-    },
-    {
-      "epoch": 0.9700365274230315,
-      "grad_norm": 0.0,
-      "learning_rate": 4.705287582322693e-08,
-      "loss": 0.7306,
-      "step": 35320
-    },
-    {
-      "epoch": 0.9700639916508748,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6966726488769653e-08,
-      "loss": 0.8571,
-      "step": 35321
-    },
-    {
-      "epoch": 0.9700914558787179,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6880655906567095e-08,
-      "loss": 0.8767,
-      "step": 35322
-    },
-    {
-      "epoch": 0.9701189201065612,
-      "grad_norm": 0.0,
-      "learning_rate": 4.679466407730093e-08,
-      "loss": 0.8169,
-      "step": 35323
-    },
-    {
-      "epoch": 0.9701463843344045,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6708751001650624e-08,
-      "loss": 0.8473,
-      "step": 35324
-    },
-    {
-      "epoch": 0.9701738485622476,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6622916680295616e-08,
-      "loss": 0.8151,
-      "step": 35325
-    },
-    {
-      "epoch": 0.9702013127900909,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6537161113916485e-08,
-      "loss": 0.7556,
-      "step": 35326
-    },
-    {
-      "epoch": 0.9702287770179341,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6451484303190466e-08,
-      "loss": 0.8567,
-      "step": 35327
-    },
-    {
-      "epoch": 0.9702562412457774,
-      "grad_norm": 0.0,
-      "learning_rate": 4.636588624879701e-08,
-      "loss": 0.7993,
-      "step": 35328
-    },
-    {
-      "epoch": 0.9702837054736206,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6280366951411136e-08,
-      "loss": 0.9158,
-      "step": 35329
-    },
-    {
-      "epoch": 0.9703111697014638,
-      "grad_norm": 0.0,
-      "learning_rate": 4.619492641171119e-08,
-      "loss": 0.7625,
-      "step": 35330
-    },
-    {
-      "epoch": 0.9703386339293071,
-      "grad_norm": 0.0,
-      "learning_rate": 4.61095646303722e-08,
-      "loss": 0.871,
-      "step": 35331
-    },
-    {
-      "epoch": 0.9703660981571504,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6024281608070265e-08,
-      "loss": 0.7742,
-      "step": 35332
-    },
-    {
-      "epoch": 0.9703935623849935,
-      "grad_norm": 0.0,
-      "learning_rate": 4.593907734547931e-08,
-      "loss": 0.8067,
-      "step": 35333
-    },
-    {
-      "epoch": 0.9704210266128368,
-      "grad_norm": 0.0,
-      "learning_rate": 4.585395184327546e-08,
-      "loss": 0.7968,
-      "step": 35334
-    },
-    {
-      "epoch": 0.97044849084068,
-      "grad_norm": 0.0,
-      "learning_rate": 4.57689051021315e-08,
-      "loss": 0.7407,
-      "step": 35335
-    },
-    {
-      "epoch": 0.9704759550685232,
-      "grad_norm": 0.0,
-      "learning_rate": 4.568393712271801e-08,
-      "loss": 0.713,
-      "step": 35336
-    },
-    {
-      "epoch": 0.9705034192963665,
-      "grad_norm": 0.0,
-      "learning_rate": 4.559904790571001e-08,
-      "loss": 0.7895,
-      "step": 35337
-    },
-    {
-      "epoch": 0.9705308835242097,
-      "grad_norm": 0.0,
-      "learning_rate": 4.551423745177918e-08,
-      "loss": 0.8249,
-      "step": 35338
-    },
-    {
-      "epoch": 0.970558347752053,
-      "grad_norm": 0.0,
-      "learning_rate": 4.542950576159389e-08,
-      "loss": 0.7664,
-      "step": 35339
-    },
-    {
-      "epoch": 0.9705858119798962,
-      "grad_norm": 0.0,
-      "learning_rate": 4.534485283582801e-08,
-      "loss": 0.7435,
-      "step": 35340
-    },
-    {
-      "epoch": 0.9706132762077394,
-      "grad_norm": 0.0,
-      "learning_rate": 4.526027867514882e-08,
-      "loss": 0.8417,
-      "step": 35341
-    },
-    {
-      "epoch": 0.9706407404355827,
-      "grad_norm": 0.0,
-      "learning_rate": 4.517578328022687e-08,
-      "loss": 0.8211,
-      "step": 35342
-    },
-    {
-      "epoch": 0.9706682046634258,
-      "grad_norm": 0.0,
-      "learning_rate": 4.509136665173053e-08,
-      "loss": 0.7939,
-      "step": 35343
-    },
-    {
-      "epoch": 0.9706956688912691,
-      "grad_norm": 0.0,
-      "learning_rate": 4.500702879032703e-08,
-      "loss": 0.7768,
-      "step": 35344
-    },
-    {
-      "epoch": 0.9707231331191124,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4922769696683633e-08,
-      "loss": 0.7717,
-      "step": 35345
-    },
-    {
-      "epoch": 0.9707505973469556,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4838589371468675e-08,
-      "loss": 0.8477,
-      "step": 35346
-    },
-    {
-      "epoch": 0.9707780615747988,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4754487815346085e-08,
-      "loss": 0.8811,
-      "step": 35347
-    },
-    {
-      "epoch": 0.970805525802642,
-      "grad_norm": 0.0,
-      "learning_rate": 4.467046502898198e-08,
-      "loss": 0.8861,
-      "step": 35348
-    },
-    {
-      "epoch": 0.9708329900304853,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4586521013042504e-08,
-      "loss": 0.7895,
-      "step": 35349
-    },
-    {
-      "epoch": 0.9708604542583286,
-      "grad_norm": 0.0,
-      "learning_rate": 4.450265576819046e-08,
-      "loss": 0.8544,
-      "step": 35350
-    },
-    {
-      "epoch": 0.9708879184861717,
-      "grad_norm": 0.0,
-      "learning_rate": 4.441886929508976e-08,
-      "loss": 0.7352,
-      "step": 35351
-    },
-    {
-      "epoch": 0.970915382714015,
-      "grad_norm": 0.0,
-      "learning_rate": 4.433516159440321e-08,
-      "loss": 0.7091,
-      "step": 35352
-    },
-    {
-      "epoch": 0.9709428469418583,
-      "grad_norm": 0.0,
-      "learning_rate": 4.425153266679361e-08,
-      "loss": 0.807,
-      "step": 35353
-    },
-    {
-      "epoch": 0.9709703111697014,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4167982512922644e-08,
-      "loss": 0.7399,
-      "step": 35354
-    },
-    {
-      "epoch": 0.9709977753975447,
-      "grad_norm": 0.0,
-      "learning_rate": 4.40845111334498e-08,
-      "loss": 0.8445,
-      "step": 35355
-    },
-    {
-      "epoch": 0.9710252396253879,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4001118529037875e-08,
-      "loss": 0.8149,
-      "step": 35356
-    },
-    {
-      "epoch": 0.9710527038532312,
-      "grad_norm": 0.0,
-      "learning_rate": 4.391780470034524e-08,
-      "loss": 0.7588,
-      "step": 35357
-    },
-    {
-      "epoch": 0.9710801680810744,
-      "grad_norm": 0.0,
-      "learning_rate": 4.3834569648032456e-08,
-      "loss": 0.7149,
-      "step": 35358
-    },
-    {
-      "epoch": 0.9711076323089176,
-      "grad_norm": 0.0,
-      "learning_rate": 4.375141337275679e-08,
-      "loss": 0.7176,
-      "step": 35359
-    },
-    {
-      "epoch": 0.9711350965367609,
-      "grad_norm": 0.0,
-      "learning_rate": 4.3668335875176605e-08,
-      "loss": 0.897,
-      "step": 35360
-    },
-    {
-      "epoch": 0.971162560764604,
-      "grad_norm": 0.0,
-      "learning_rate": 4.358533715594915e-08,
-      "loss": 0.8686,
-      "step": 35361
-    },
-    {
-      "epoch": 0.9711900249924473,
-      "grad_norm": 0.0,
-      "learning_rate": 4.350241721573056e-08,
-      "loss": 0.8659,
-      "step": 35362
-    },
-    {
-      "epoch": 0.9712174892202906,
-      "grad_norm": 0.0,
-      "learning_rate": 4.34195760551781e-08,
-      "loss": 0.8099,
-      "step": 35363
-    },
-    {
-      "epoch": 0.9712449534481338,
-      "grad_norm": 0.0,
-      "learning_rate": 4.333681367494791e-08,
-      "loss": 0.7986,
-      "step": 35364
-    },
-    {
-      "epoch": 0.971272417675977,
-      "grad_norm": 0.0,
-      "learning_rate": 4.325413007569168e-08,
-      "loss": 0.782,
-      "step": 35365
-    },
-    {
-      "epoch": 0.9712998819038203,
-      "grad_norm": 0.0,
-      "learning_rate": 4.317152525806778e-08,
-      "loss": 0.7893,
-      "step": 35366
-    },
-    {
-      "epoch": 0.9713273461316635,
-      "grad_norm": 0.0,
-      "learning_rate": 4.308899922272569e-08,
-      "loss": 0.922,
-      "step": 35367
-    },
-    {
-      "epoch": 0.9713548103595068,
-      "grad_norm": 0.0,
-      "learning_rate": 4.300655197032155e-08,
-      "loss": 0.9144,
-      "step": 35368
-    },
-    {
-      "epoch": 0.9713822745873499,
-      "grad_norm": 0.0,
-      "learning_rate": 4.292418350150596e-08,
-      "loss": 0.9238,
-      "step": 35369
-    },
-    {
-      "epoch": 0.9714097388151932,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2841893816931715e-08,
-      "loss": 0.7648,
-      "step": 35370
-    },
-    {
-      "epoch": 0.9714372030430365,
-      "grad_norm": 0.0,
-      "learning_rate": 4.275968291724941e-08,
-      "loss": 0.7917,
-      "step": 35371
-    },
-    {
-      "epoch": 0.9714646672708797,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2677550803109646e-08,
-      "loss": 0.7941,
-      "step": 35372
-    },
-    {
-      "epoch": 0.9714921314987229,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2595497475161894e-08,
-      "loss": 0.7438,
-      "step": 35373
-    },
-    {
-      "epoch": 0.9715195957265661,
-      "grad_norm": 0.0,
-      "learning_rate": 4.251352293405564e-08,
-      "loss": 0.6684,
-      "step": 35374
-    },
-    {
-      "epoch": 0.9715470599544094,
-      "grad_norm": 0.0,
-      "learning_rate": 4.243162718043925e-08,
-      "loss": 0.8195,
-      "step": 35375
-    },
-    {
-      "epoch": 0.9715745241822527,
-      "grad_norm": 0.0,
-      "learning_rate": 4.23498102149622e-08,
-      "loss": 0.7886,
-      "step": 35376
-    },
-    {
-      "epoch": 0.9716019884100958,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2268072038269546e-08,
-      "loss": 0.7554,
-      "step": 35377
-    },
-    {
-      "epoch": 0.9716294526379391,
-      "grad_norm": 0.0,
-      "learning_rate": 4.218641265100965e-08,
-      "loss": 0.8033,
-      "step": 35378
-    },
-    {
-      "epoch": 0.9716569168657824,
-      "grad_norm": 0.0,
-      "learning_rate": 4.210483205382754e-08,
-      "loss": 0.7401,
-      "step": 35379
-    },
-    {
-      "epoch": 0.9716843810936255,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2023330247370484e-08,
-      "loss": 0.8533,
-      "step": 35380
-    },
-    {
-      "epoch": 0.9717118453214688,
-      "grad_norm": 0.0,
-      "learning_rate": 4.194190723228131e-08,
-      "loss": 0.8629,
-      "step": 35381
-    },
-    {
-      "epoch": 0.971739309549312,
-      "grad_norm": 0.0,
-      "learning_rate": 4.186056300920505e-08,
-      "loss": 0.8491,
-      "step": 35382
-    },
-    {
-      "epoch": 0.9717667737771553,
-      "grad_norm": 0.0,
-      "learning_rate": 4.177929757878674e-08,
-      "loss": 0.783,
-      "step": 35383
-    },
-    {
-      "epoch": 0.9717942380049985,
-      "grad_norm": 0.0,
-      "learning_rate": 4.169811094166698e-08,
-      "loss": 0.8175,
-      "step": 35384
-    },
-    {
-      "epoch": 0.9718217022328417,
-      "grad_norm": 0.0,
-      "learning_rate": 4.1617003098488596e-08,
-      "loss": 0.6981,
-      "step": 35385
-    },
-    {
-      "epoch": 0.971849166460685,
-      "grad_norm": 0.0,
-      "learning_rate": 4.153597404989551e-08,
-      "loss": 0.7722,
-      "step": 35386
-    },
-    {
-      "epoch": 0.9718766306885281,
-      "grad_norm": 0.0,
-      "learning_rate": 4.145502379652611e-08,
-      "loss": 0.8887,
-      "step": 35387
-    },
-    {
-      "epoch": 0.9719040949163714,
-      "grad_norm": 0.0,
-      "learning_rate": 4.1374152339023196e-08,
-      "loss": 0.743,
-      "step": 35388
-    },
-    {
-      "epoch": 0.9719315591442147,
-      "grad_norm": 0.0,
-      "learning_rate": 4.1293359678024056e-08,
-      "loss": 0.7718,
-      "step": 35389
-    },
-    {
-      "epoch": 0.9719590233720579,
-      "grad_norm": 0.0,
-      "learning_rate": 4.1212645814171506e-08,
-      "loss": 0.8311,
-      "step": 35390
-    },
-    {
-      "epoch": 0.9719864875999011,
-      "grad_norm": 0.0,
-      "learning_rate": 4.113201074810058e-08,
-      "loss": 0.7524,
-      "step": 35391
-    },
-    {
-      "epoch": 0.9720139518277444,
-      "grad_norm": 0.0,
-      "learning_rate": 4.105145448045078e-08,
-      "loss": 0.7141,
-      "step": 35392
-    },
-    {
-      "epoch": 0.9720414160555876,
-      "grad_norm": 0.0,
-      "learning_rate": 4.097097701186048e-08,
-      "loss": 0.8418,
-      "step": 35393
-    },
-    {
-      "epoch": 0.9720688802834309,
-      "grad_norm": 0.0,
-      "learning_rate": 4.0890578342965834e-08,
-      "loss": 0.9109,
-      "step": 35394
-    },
-    {
-      "epoch": 0.972096344511274,
-      "grad_norm": 0.0,
-      "learning_rate": 4.08102584744019e-08,
-      "loss": 0.7539,
-      "step": 35395
-    },
-    {
-      "epoch": 0.9721238087391173,
-      "grad_norm": 0.0,
-      "learning_rate": 4.073001740680482e-08,
-      "loss": 0.7197,
-      "step": 35396
-    },
-    {
-      "epoch": 0.9721512729669606,
-      "grad_norm": 0.0,
-      "learning_rate": 4.064985514081077e-08,
-      "loss": 0.7372,
-      "step": 35397
-    },
-    {
-      "epoch": 0.9721787371948037,
-      "grad_norm": 0.0,
-      "learning_rate": 4.056977167705256e-08,
-      "loss": 0.7889,
-      "step": 35398
-    },
-    {
-      "epoch": 0.972206201422647,
-      "grad_norm": 0.0,
-      "learning_rate": 4.048976701616414e-08,
-      "loss": 0.8141,
-      "step": 35399
-    },
-    {
-      "epoch": 0.9722336656504902,
-      "grad_norm": 0.0,
-      "learning_rate": 4.0409841158778326e-08,
-      "loss": 0.756,
-      "step": 35400
-    },
-    {
-      "epoch": 0.9722611298783335,
-      "grad_norm": 0.0,
-      "learning_rate": 4.032999410552796e-08,
-      "loss": 0.8265,
-      "step": 35401
-    },
-    {
-      "epoch": 0.9722885941061767,
-      "grad_norm": 0.0,
-      "learning_rate": 4.0250225857044746e-08,
-      "loss": 0.7728,
-      "step": 35402
-    },
-    {
-      "epoch": 0.9723160583340199,
-      "grad_norm": 0.0,
-      "learning_rate": 4.0170536413959296e-08,
-      "loss": 0.7555,
-      "step": 35403
-    },
-    {
-      "epoch": 0.9723435225618632,
-      "grad_norm": 0.0,
-      "learning_rate": 4.009092577690332e-08,
-      "loss": 0.8924,
-      "step": 35404
-    },
-    {
-      "epoch": 0.9723709867897065,
-      "grad_norm": 0.0,
-      "learning_rate": 4.001139394650633e-08,
-      "loss": 0.8474,
-      "step": 35405
-    },
-    {
-      "epoch": 0.9723984510175496,
-      "grad_norm": 0.0,
-      "learning_rate": 3.99319409233967e-08,
-      "loss": 0.7828,
-      "step": 35406
-    },
-    {
-      "epoch": 0.9724259152453929,
-      "grad_norm": 0.0,
-      "learning_rate": 3.985256670820392e-08,
-      "loss": 0.8241,
-      "step": 35407
-    },
-    {
-      "epoch": 0.9724533794732361,
-      "grad_norm": 0.0,
-      "learning_rate": 3.977327130155528e-08,
-      "loss": 0.8871,
-      "step": 35408
-    },
-    {
-      "epoch": 0.9724808437010793,
-      "grad_norm": 0.0,
-      "learning_rate": 3.9694054704079166e-08,
-      "loss": 0.8145,
-      "step": 35409
-    },
-    {
-      "epoch": 0.9725083079289226,
-      "grad_norm": 0.0,
-      "learning_rate": 3.961491691640174e-08,
-      "loss": 0.7792,
-      "step": 35410
-    },
-    {
-      "epoch": 0.9725357721567658,
-      "grad_norm": 0.0,
-      "learning_rate": 3.953585793915027e-08,
-      "loss": 0.8275,
-      "step": 35411
-    },
-    {
-      "epoch": 0.9725632363846091,
-      "grad_norm": 0.0,
-      "learning_rate": 3.945687777294871e-08,
-      "loss": 0.8191,
-      "step": 35412
-    },
-    {
-      "epoch": 0.9725907006124522,
-      "grad_norm": 0.0,
-      "learning_rate": 3.9377976418422116e-08,
-      "loss": 0.8026,
-      "step": 35413
-    },
-    {
-      "epoch": 0.9726181648402955,
-      "grad_norm": 0.0,
-      "learning_rate": 3.929915387619665e-08,
-      "loss": 0.8332,
-      "step": 35414
-    },
-    {
-      "epoch": 0.9726456290681388,
-      "grad_norm": 0.0,
-      "learning_rate": 3.922041014689404e-08,
-      "loss": 0.7474,
-      "step": 35415
-    },
-    {
-      "epoch": 0.972673093295982,
-      "grad_norm": 0.0,
-      "learning_rate": 3.9141745231137115e-08,
-      "loss": 0.8252,
-      "step": 35416
-    },
-    {
-      "epoch": 0.9727005575238252,
-      "grad_norm": 0.0,
-      "learning_rate": 3.906315912955094e-08,
-      "loss": 0.7993,
-      "step": 35417
-    },
-    {
-      "epoch": 0.9727280217516685,
-      "grad_norm": 0.0,
-      "learning_rate": 3.898465184275391e-08,
-      "loss": 0.7253,
-      "step": 35418
-    },
-    {
-      "epoch": 0.9727554859795117,
-      "grad_norm": 0.0,
-      "learning_rate": 3.8906223371369955e-08,
-      "loss": 0.8481,
-      "step": 35419
-    },
-    {
-      "epoch": 0.972782950207355,
-      "grad_norm": 0.0,
-      "learning_rate": 3.882787371601748e-08,
-      "loss": 0.772,
-      "step": 35420
-    },
-    {
-      "epoch": 0.9728104144351981,
-      "grad_norm": 0.0,
-      "learning_rate": 3.874960287731822e-08,
-      "loss": 0.8795,
-      "step": 35421
-    },
-    {
-      "epoch": 0.9728378786630414,
-      "grad_norm": 0.0,
-      "learning_rate": 3.8671410855890544e-08,
-      "loss": 0.8358,
-      "step": 35422
-    },
-    {
-      "epoch": 0.9728653428908847,
-      "grad_norm": 0.0,
-      "learning_rate": 3.859329765235398e-08,
-      "loss": 0.8382,
-      "step": 35423
-    },
-    {
-      "epoch": 0.9728928071187278,
-      "grad_norm": 0.0,
-      "learning_rate": 3.851526326732469e-08,
-      "loss": 0.8223,
-      "step": 35424
-    },
-    {
-      "epoch": 0.9729202713465711,
-      "grad_norm": 0.0,
-      "learning_rate": 3.843730770142218e-08,
-      "loss": 0.7564,
-      "step": 35425
-    },
-    {
-      "epoch": 0.9729477355744143,
-      "grad_norm": 0.0,
-      "learning_rate": 3.835943095526262e-08,
-      "loss": 0.8241,
-      "step": 35426
-    },
-    {
-      "epoch": 0.9729751998022576,
-      "grad_norm": 0.0,
-      "learning_rate": 3.828163302946109e-08,
-      "loss": 0.7983,
-      "step": 35427
-    },
-    {
-      "epoch": 0.9730026640301008,
-      "grad_norm": 0.0,
-      "learning_rate": 3.820391392463485e-08,
-      "loss": 0.7494,
-      "step": 35428
-    },
-    {
-      "epoch": 0.973030128257944,
-      "grad_norm": 0.0,
-      "learning_rate": 3.812627364139787e-08,
-      "loss": 0.7855,
-      "step": 35429
-    },
-    {
-      "epoch": 0.9730575924857873,
-      "grad_norm": 0.0,
-      "learning_rate": 3.80487121803641e-08,
-      "loss": 0.8278,
-      "step": 35430
-    },
-    {
-      "epoch": 0.9730850567136305,
-      "grad_norm": 0.0,
-      "learning_rate": 3.79712295421486e-08,
-      "loss": 0.8268,
-      "step": 35431
-    },
-    {
-      "epoch": 0.9731125209414737,
-      "grad_norm": 0.0,
-      "learning_rate": 3.789382572736311e-08,
-      "loss": 0.8166,
-      "step": 35432
-    },
-    {
-      "epoch": 0.973139985169317,
-      "grad_norm": 0.0,
-      "learning_rate": 3.781650073662046e-08,
-      "loss": 0.7903,
-      "step": 35433
-    },
-    {
-      "epoch": 0.9731674493971602,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7739254570533514e-08,
-      "loss": 0.7774,
-      "step": 35434
-    },
-    {
-      "epoch": 0.9731949136250034,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7662087229712875e-08,
-      "loss": 0.8334,
-      "step": 35435
-    },
-    {
-      "epoch": 0.9732223778528467,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7584998714766954e-08,
-      "loss": 0.8261,
-      "step": 35436
-    },
-    {
-      "epoch": 0.9732498420806899,
-      "grad_norm": 0.0,
-      "learning_rate": 3.75079890263097e-08,
-      "loss": 0.7808,
-      "step": 35437
-    },
-    {
-      "epoch": 0.9732773063085332,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7431058164947297e-08,
-      "loss": 0.8725,
-      "step": 35438
-    },
-    {
-      "epoch": 0.9733047705363763,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7354206131289264e-08,
-      "loss": 0.7568,
-      "step": 35439
-    },
-    {
-      "epoch": 0.9733322347642196,
-      "grad_norm": 0.0,
-      "learning_rate": 3.727743292594399e-08,
-      "loss": 0.7423,
-      "step": 35440
-    },
-    {
-      "epoch": 0.9733596989920629,
-      "grad_norm": 0.0,
-      "learning_rate": 3.720073854951989e-08,
-      "loss": 0.8375,
-      "step": 35441
-    },
-    {
-      "epoch": 0.973387163219906,
-      "grad_norm": 0.0,
-      "learning_rate": 3.712412300262203e-08,
-      "loss": 0.8191,
-      "step": 35442
-    },
-    {
-      "epoch": 0.9734146274477493,
-      "grad_norm": 0.0,
-      "learning_rate": 3.704758628585658e-08,
-      "loss": 0.7853,
-      "step": 35443
-    },
-    {
-      "epoch": 0.9734420916755926,
-      "grad_norm": 0.0,
-      "learning_rate": 3.697112839983086e-08,
-      "loss": 0.7458,
-      "step": 35444
-    },
-    {
-      "epoch": 0.9734695559034358,
-      "grad_norm": 0.0,
-      "learning_rate": 3.68947493451488e-08,
-      "loss": 0.7813,
-      "step": 35445
-    },
-    {
-      "epoch": 0.973497020131279,
-      "grad_norm": 0.0,
-      "learning_rate": 3.681844912241439e-08,
-      "loss": 0.8824,
-      "step": 35446
-    },
-    {
-      "epoch": 0.9735244843591222,
-      "grad_norm": 0.0,
-      "learning_rate": 3.674222773223268e-08,
-      "loss": 0.8242,
-      "step": 35447
-    },
-    {
-      "epoch": 0.9735519485869655,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6666085175204314e-08,
-      "loss": 0.7823,
-      "step": 35448
-    },
-    {
-      "epoch": 0.9735794128148088,
-      "grad_norm": 0.0,
-      "learning_rate": 3.659002145193435e-08,
-      "loss": 0.7352,
-      "step": 35449
-    },
-    {
-      "epoch": 0.9736068770426519,
-      "grad_norm": 0.0,
-      "learning_rate": 3.651403656302344e-08,
-      "loss": 0.7852,
-      "step": 35450
-    },
-    {
-      "epoch": 0.9736343412704952,
-      "grad_norm": 0.0,
-      "learning_rate": 3.64381305090733e-08,
-      "loss": 0.756,
-      "step": 35451
-    },
-    {
-      "epoch": 0.9736618054983384,
-      "grad_norm": 0.0,
-      "learning_rate": 3.636230329068346e-08,
-      "loss": 0.8821,
-      "step": 35452
-    },
-    {
-      "epoch": 0.9736892697261816,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6286554908455676e-08,
-      "loss": 0.8329,
-      "step": 35453
-    },
-    {
-      "epoch": 0.9737167339540249,
-      "grad_norm": 0.0,
-      "learning_rate": 3.621088536298723e-08,
-      "loss": 0.8311,
-      "step": 35454
-    },
-    {
-      "epoch": 0.9737441981818681,
-      "grad_norm": 0.0,
-      "learning_rate": 3.613529465487875e-08,
-      "loss": 0.8981,
-      "step": 35455
-    },
-    {
-      "epoch": 0.9737716624097114,
-      "grad_norm": 0.0,
-      "learning_rate": 3.605978278472644e-08,
-      "loss": 0.8152,
-      "step": 35456
-    },
-    {
-      "epoch": 0.9737991266375546,
-      "grad_norm": 0.0,
-      "learning_rate": 3.598434975312981e-08,
-      "loss": 0.7993,
-      "step": 35457
-    },
-    {
-      "epoch": 0.9738265908653978,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5908995560683946e-08,
-      "loss": 0.8065,
-      "step": 35458
-    },
-    {
-      "epoch": 0.9738540550932411,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5833720207986146e-08,
-      "loss": 0.9044,
-      "step": 35459
-    },
-    {
-      "epoch": 0.9738815193210842,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5758523695631485e-08,
-      "loss": 0.8269,
-      "step": 35460
-    },
-    {
-      "epoch": 0.9739089835489275,
-      "grad_norm": 0.0,
-      "learning_rate": 3.568340602421505e-08,
-      "loss": 0.8091,
-      "step": 35461
-    },
-    {
-      "epoch": 0.9739364477767708,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5608367194331915e-08,
-      "loss": 0.8904,
-      "step": 35462
-    },
-    {
-      "epoch": 0.973963912004614,
-      "grad_norm": 0.0,
-      "learning_rate": 3.553340720657494e-08,
-      "loss": 0.7458,
-      "step": 35463
-    },
-    {
-      "epoch": 0.9739913762324572,
-      "grad_norm": 0.0,
-      "learning_rate": 3.545852606153699e-08,
-      "loss": 0.8691,
-      "step": 35464
-    },
-    {
-      "epoch": 0.9740188404603004,
-      "grad_norm": 0.0,
-      "learning_rate": 3.538372375981203e-08,
-      "loss": 0.768,
-      "step": 35465
-    },
-    {
-      "epoch": 0.9740463046881437,
-      "grad_norm": 0.0,
-      "learning_rate": 3.530900030198958e-08,
-      "loss": 0.7637,
-      "step": 35466
-    },
-    {
-      "epoch": 0.974073768915987,
-      "grad_norm": 0.0,
-      "learning_rate": 3.523435568866362e-08,
-      "loss": 0.81,
-      "step": 35467
-    },
-    {
-      "epoch": 0.9741012331438301,
-      "grad_norm": 0.0,
-      "learning_rate": 3.515978992042257e-08,
-      "loss": 0.8729,
-      "step": 35468
-    },
-    {
-      "epoch": 0.9741286973716734,
-      "grad_norm": 0.0,
-      "learning_rate": 3.508530299785817e-08,
-      "loss": 0.8491,
-      "step": 35469
-    },
-    {
-      "epoch": 0.9741561615995167,
-      "grad_norm": 0.0,
-      "learning_rate": 3.501089492155774e-08,
-      "loss": 0.7987,
-      "step": 35470
-    },
-    {
-      "epoch": 0.9741836258273598,
-      "grad_norm": 0.0,
-      "learning_rate": 3.493656569211301e-08,
-      "loss": 0.8083,
-      "step": 35471
-    },
-    {
-      "epoch": 0.9742110900552031,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4862315310107976e-08,
-      "loss": 0.8228,
-      "step": 35472
-    },
-    {
-      "epoch": 0.9742385542830463,
-      "grad_norm": 0.0,
-      "learning_rate": 3.478814377613437e-08,
-      "loss": 0.767,
-      "step": 35473
-    },
-    {
-      "epoch": 0.9742660185108896,
-      "grad_norm": 0.0,
-      "learning_rate": 3.471405109077508e-08,
-      "loss": 0.8026,
-      "step": 35474
-    },
-    {
-      "epoch": 0.9742934827387328,
-      "grad_norm": 0.0,
-      "learning_rate": 3.464003725461962e-08,
-      "loss": 0.8455,
-      "step": 35475
-    },
-    {
-      "epoch": 0.974320946966576,
-      "grad_norm": 0.0,
-      "learning_rate": 3.456610226825308e-08,
-      "loss": 0.8697,
-      "step": 35476
-    },
-    {
-      "epoch": 0.9743484111944193,
-      "grad_norm": 0.0,
-      "learning_rate": 3.449224613225943e-08,
-      "loss": 0.8059,
-      "step": 35477
-    },
-    {
-      "epoch": 0.9743758754222624,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4418468847222664e-08,
-      "loss": 0.7421,
-      "step": 35478
-    },
-    {
-      "epoch": 0.9744033396501057,
-      "grad_norm": 0.0,
-      "learning_rate": 3.434477041372786e-08,
-      "loss": 0.7919,
-      "step": 35479
-    },
-    {
-      "epoch": 0.974430803877949,
-      "grad_norm": 0.0,
-      "learning_rate": 3.427115083235677e-08,
-      "loss": 0.7375,
-      "step": 35480
-    },
-    {
-      "epoch": 0.9744582681057922,
-      "grad_norm": 0.0,
-      "learning_rate": 3.419761010369338e-08,
-      "loss": 0.8003,
-      "step": 35481
-    },
-    {
-      "epoch": 0.9744857323336354,
-      "grad_norm": 0.0,
-      "learning_rate": 3.412414822831944e-08,
-      "loss": 0.8722,
-      "step": 35482
-    },
-    {
-      "epoch": 0.9745131965614787,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4050765206814495e-08,
-      "loss": 0.7287,
-      "step": 35483
-    },
-    {
-      "epoch": 0.9745406607893219,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3977461039760294e-08,
-      "loss": 0.8572,
-      "step": 35484
-    },
-    {
-      "epoch": 0.9745681250171652,
-      "grad_norm": 0.0,
-      "learning_rate": 3.390423572773749e-08,
-      "loss": 0.8597,
-      "step": 35485
-    },
-    {
-      "epoch": 0.9745955892450083,
-      "grad_norm": 0.0,
-      "learning_rate": 3.383108927132561e-08,
-      "loss": 0.7886,
-      "step": 35486
-    },
-    {
-      "epoch": 0.9746230534728516,
-      "grad_norm": 0.0,
-      "learning_rate": 3.375802167110087e-08,
-      "loss": 0.877,
-      "step": 35487
-    },
-    {
-      "epoch": 0.9746505177006949,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3685032927645024e-08,
-      "loss": 0.8483,
-      "step": 35488
-    },
-    {
-      "epoch": 0.974677981928538,
-      "grad_norm": 0.0,
-      "learning_rate": 3.361212304153316e-08,
-      "loss": 0.8889,
-      "step": 35489
-    },
-    {
-      "epoch": 0.9747054461563813,
-      "grad_norm": 0.0,
-      "learning_rate": 3.35392920133415e-08,
-      "loss": 0.8387,
-      "step": 35490
-    },
-    {
-      "epoch": 0.9747329103842245,
-      "grad_norm": 0.0,
-      "learning_rate": 3.346653984364956e-08,
-      "loss": 0.8156,
-      "step": 35491
-    },
-    {
-      "epoch": 0.9747603746120678,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3393866533029117e-08,
-      "loss": 0.8397,
-      "step": 35492
-    },
-    {
-      "epoch": 0.974787838839911,
-      "grad_norm": 0.0,
-      "learning_rate": 3.33212720820586e-08,
-      "loss": 0.6633,
-      "step": 35493
-    },
-    {
-      "epoch": 0.9748153030677542,
-      "grad_norm": 0.0,
-      "learning_rate": 3.324875649131088e-08,
-      "loss": 0.8283,
-      "step": 35494
-    },
-    {
-      "epoch": 0.9748427672955975,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3176319761358825e-08,
-      "loss": 0.8056,
-      "step": 35495
-    },
-    {
-      "epoch": 0.9748702315234408,
-      "grad_norm": 0.0,
-      "learning_rate": 3.310396189277754e-08,
-      "loss": 0.6962,
-      "step": 35496
-    },
-    {
-      "epoch": 0.9748976957512839,
-      "grad_norm": 0.0,
-      "learning_rate": 3.303168288613767e-08,
-      "loss": 0.7457,
-      "step": 35497
-    },
-    {
-      "epoch": 0.9749251599791272,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2959482742012104e-08,
-      "loss": 0.7615,
-      "step": 35498
-    },
-    {
-      "epoch": 0.9749526242069704,
-      "grad_norm": 0.0,
-      "learning_rate": 3.288736146097149e-08,
-      "loss": 0.8139,
-      "step": 35499
-    },
-    {
-      "epoch": 0.9749800884348137,
-      "grad_norm": 0.0,
-      "learning_rate": 3.28153190435887e-08,
-      "loss": 0.7799,
-      "step": 35500
-    },
-    {
-      "epoch": 0.9750075526626569,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2743355490429954e-08,
-      "loss": 0.8214,
-      "step": 35501
-    },
-    {
-      "epoch": 0.9750350168905001,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2671470802068117e-08,
-      "loss": 0.8663,
-      "step": 35502
-    },
-    {
-      "epoch": 0.9750624811183434,
-      "grad_norm": 0.0,
-      "learning_rate": 3.259966497907052e-08,
-      "loss": 0.8867,
-      "step": 35503
-    },
-    {
-      "epoch": 0.9750899453461865,
-      "grad_norm": 0.0,
-      "learning_rate": 3.252793802200449e-08,
-      "loss": 0.7481,
-      "step": 35504
-    },
-    {
-      "epoch": 0.9751174095740298,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2456289931438456e-08,
-      "loss": 0.7806,
-      "step": 35505
-    },
-    {
-      "epoch": 0.9751448738018731,
-      "grad_norm": 0.0,
-      "learning_rate": 3.238472070794085e-08,
-      "loss": 0.7316,
-      "step": 35506
-    },
-    {
-      "epoch": 0.9751723380297163,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2313230352074565e-08,
-      "loss": 0.8119,
-      "step": 35507
-    },
-    {
-      "epoch": 0.9751998022575595,
-      "grad_norm": 0.0,
-      "learning_rate": 3.224181886440803e-08,
-      "loss": 0.8258,
-      "step": 35508
-    },
-    {
-      "epoch": 0.9752272664854028,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2170486245505226e-08,
-      "loss": 0.7552,
-      "step": 35509
-    },
-    {
-      "epoch": 0.975254730713246,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2099232495931275e-08,
-      "loss": 0.8365,
-      "step": 35510
-    },
-    {
-      "epoch": 0.9752821949410893,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2028057616249056e-08,
-      "loss": 0.793,
-      "step": 35511
-    },
-    {
-      "epoch": 0.9753096591689324,
-      "grad_norm": 0.0,
-      "learning_rate": 3.195696160702144e-08,
-      "loss": 0.8018,
-      "step": 35512
-    },
-    {
-      "epoch": 0.9753371233967757,
-      "grad_norm": 0.0,
-      "learning_rate": 3.1885944468812434e-08,
-      "loss": 0.8122,
-      "step": 35513
-    },
-    {
-      "epoch": 0.975364587624619,
-      "grad_norm": 0.0,
-      "learning_rate": 3.181500620218159e-08,
-      "loss": 0.8426,
-      "step": 35514
-    },
-    {
-      "epoch": 0.9753920518524621,
-      "grad_norm": 0.0,
-      "learning_rate": 3.174414680769289e-08,
-      "loss": 0.7613,
-      "step": 35515
-    },
-    {
-      "epoch": 0.9754195160803054,
-      "grad_norm": 0.0,
-      "learning_rate": 3.167336628590589e-08,
-      "loss": 0.7754,
-      "step": 35516
-    },
-    {
-      "epoch": 0.9754469803081486,
-      "grad_norm": 0.0,
-      "learning_rate": 3.160266463738126e-08,
-      "loss": 0.8147,
-      "step": 35517
-    },
-    {
-      "epoch": 0.9754744445359919,
-      "grad_norm": 0.0,
-      "learning_rate": 3.153204186267633e-08,
-      "loss": 0.853,
-      "step": 35518
-    },
-    {
-      "epoch": 0.9755019087638351,
-      "grad_norm": 0.0,
-      "learning_rate": 3.146149796235176e-08,
-      "loss": 0.8508,
-      "step": 35519
-    },
-    {
-      "epoch": 0.9755293729916783,
-      "grad_norm": 0.0,
-      "learning_rate": 3.139103293696599e-08,
-      "loss": 0.8326,
-      "step": 35520
-    },
-    {
-      "epoch": 0.9755568372195216,
-      "grad_norm": 0.0,
-      "learning_rate": 3.132064678707525e-08,
-      "loss": 0.7708,
-      "step": 35521
-    },
-    {
-      "epoch": 0.9755843014473649,
-      "grad_norm": 0.0,
-      "learning_rate": 3.125033951323686e-08,
-      "loss": 0.8743,
-      "step": 35522
-    },
-    {
-      "epoch": 0.975611765675208,
-      "grad_norm": 0.0,
-      "learning_rate": 3.118011111600705e-08,
-      "loss": 0.8396,
-      "step": 35523
-    },
-    {
-      "epoch": 0.9756392299030513,
-      "grad_norm": 0.0,
-      "learning_rate": 3.110996159594204e-08,
-      "loss": 0.8562,
-      "step": 35524
-    },
-    {
-      "epoch": 0.9756666941308945,
-      "grad_norm": 0.0,
-      "learning_rate": 3.103989095359694e-08,
-      "loss": 0.803,
-      "step": 35525
-    },
-    {
-      "epoch": 0.9756941583587377,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0969899189525756e-08,
-      "loss": 0.8914,
-      "step": 35526
-    },
-    {
-      "epoch": 0.975721622586581,
-      "grad_norm": 0.0,
-      "learning_rate": 3.089998630428248e-08,
-      "loss": 0.8411,
-      "step": 35527
-    },
-    {
-      "epoch": 0.9757490868144242,
-      "grad_norm": 0.0,
-      "learning_rate": 3.083015229842001e-08,
-      "loss": 0.6902,
-      "step": 35528
-    },
-    {
-      "epoch": 0.9757765510422675,
-      "grad_norm": 0.0,
-      "learning_rate": 3.076039717249124e-08,
-      "loss": 0.8381,
-      "step": 35529
-    },
-    {
-      "epoch": 0.9758040152701107,
-      "grad_norm": 0.0,
-      "learning_rate": 3.069072092704683e-08,
-      "loss": 0.7771,
-      "step": 35530
-    },
-    {
-      "epoch": 0.9758314794979539,
-      "grad_norm": 0.0,
-      "learning_rate": 3.06211235626408e-08,
-      "loss": 0.9008,
-      "step": 35531
-    },
-    {
-      "epoch": 0.9758589437257972,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0551605079820465e-08,
-      "loss": 0.8486,
-      "step": 35532
-    },
-    {
-      "epoch": 0.9758864079536403,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0482165479138736e-08,
-      "loss": 0.7807,
-      "step": 35533
-    },
-    {
-      "epoch": 0.9759138721814836,
-      "grad_norm": 0.0,
-      "learning_rate": 3.041280476114406e-08,
-      "loss": 0.7175,
-      "step": 35534
-    },
-    {
-      "epoch": 0.9759413364093269,
-      "grad_norm": 0.0,
-      "learning_rate": 3.034352292638487e-08,
-      "loss": 0.8082,
-      "step": 35535
-    },
-    {
-      "epoch": 0.9759688006371701,
-      "grad_norm": 0.0,
-      "learning_rate": 3.027431997540964e-08,
-      "loss": 0.7113,
-      "step": 35536
-    },
-    {
-      "epoch": 0.9759962648650133,
-      "grad_norm": 0.0,
-      "learning_rate": 3.020519590876569e-08,
-      "loss": 0.78,
-      "step": 35537
-    },
-    {
-      "epoch": 0.9760237290928565,
-      "grad_norm": 0.0,
-      "learning_rate": 3.013615072700038e-08,
-      "loss": 0.7413,
-      "step": 35538
-    },
-    {
-      "epoch": 0.9760511933206998,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0067184430659926e-08,
-      "loss": 0.8456,
-      "step": 35539
-    },
-    {
-      "epoch": 0.9760786575485431,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9998297020289444e-08,
-      "loss": 0.8268,
-      "step": 35540
-    },
-    {
-      "epoch": 0.9761061217763862,
-      "grad_norm": 0.0,
-      "learning_rate": 2.992948849643518e-08,
-      "loss": 0.843,
-      "step": 35541
-    },
-    {
-      "epoch": 0.9761335860042295,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9860758859640014e-08,
-      "loss": 0.9128,
-      "step": 35542
-    },
-    {
-      "epoch": 0.9761610502320728,
-      "grad_norm": 0.0,
-      "learning_rate": 2.979210811044908e-08,
-      "loss": 0.8144,
-      "step": 35543
-    },
-    {
-      "epoch": 0.976188514459916,
-      "grad_norm": 0.0,
-      "learning_rate": 2.972353624940527e-08,
-      "loss": 0.8181,
-      "step": 35544
-    },
-    {
-      "epoch": 0.9762159786877592,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9655043277050376e-08,
-      "loss": 0.8386,
-      "step": 35545
-    },
-    {
-      "epoch": 0.9762434429156024,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9586629193926188e-08,
-      "loss": 0.7767,
-      "step": 35546
-    },
-    {
-      "epoch": 0.9762709071434457,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9518294000576708e-08,
-      "loss": 0.7538,
-      "step": 35547
-    },
-    {
-      "epoch": 0.9762983713712889,
-      "grad_norm": 0.0,
-      "learning_rate": 2.945003769753929e-08,
-      "loss": 0.8853,
-      "step": 35548
-    },
-    {
-      "epoch": 0.9763258355991321,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9381860285355725e-08,
-      "loss": 0.8884,
-      "step": 35549
-    },
-    {
-      "epoch": 0.9763532998269754,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9313761764565575e-08,
-      "loss": 0.7749,
-      "step": 35550
-    },
-    {
-      "epoch": 0.9763807640548186,
-      "grad_norm": 0.0,
-      "learning_rate": 2.924574213570619e-08,
-      "loss": 0.8733,
-      "step": 35551
-    },
-    {
-      "epoch": 0.9764082282826618,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9177801399318252e-08,
-      "loss": 0.801,
-      "step": 35552
-    },
-    {
-      "epoch": 0.9764356925105051,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9109939555936884e-08,
-      "loss": 0.7346,
-      "step": 35553
-    },
-    {
-      "epoch": 0.9764631567383483,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9042156606100546e-08,
-      "loss": 0.8206,
-      "step": 35554
-    },
-    {
-      "epoch": 0.9764906209661915,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8974452550344368e-08,
-      "loss": 0.7545,
-      "step": 35555
-    },
-    {
-      "epoch": 0.9765180851940348,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8906827389204585e-08,
-      "loss": 0.7317,
-      "step": 35556
-    },
-    {
-      "epoch": 0.976545549421878,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8839281123217434e-08,
-      "loss": 0.7654,
-      "step": 35557
-    },
-    {
-      "epoch": 0.9765730136497213,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8771813752915822e-08,
-      "loss": 0.7701,
-      "step": 35558
-    },
-    {
-      "epoch": 0.9766004778775644,
-      "grad_norm": 0.0,
-      "learning_rate": 2.870442527883377e-08,
-      "loss": 0.8008,
-      "step": 35559
-    },
-    {
-      "epoch": 0.9766279421054077,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8637115701505292e-08,
-      "loss": 0.8802,
-      "step": 35560
-    },
-    {
-      "epoch": 0.976655406333251,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8569885021462184e-08,
-      "loss": 0.8486,
-      "step": 35561
-    },
-    {
-      "epoch": 0.9766828705610942,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8502733239237357e-08,
-      "loss": 0.7626,
-      "step": 35562
-    },
-    {
-      "epoch": 0.9767103347889374,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8435660355360383e-08,
-      "loss": 0.7793,
-      "step": 35563
-    },
-    {
-      "epoch": 0.9767377990167806,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8368666370364174e-08,
-      "loss": 0.8218,
-      "step": 35564
-    },
-    {
-      "epoch": 0.9767652632446239,
-      "grad_norm": 0.0,
-      "learning_rate": 2.830175128477719e-08,
-      "loss": 0.8254,
-      "step": 35565
-    },
-    {
-      "epoch": 0.9767927274724671,
-      "grad_norm": 0.0,
-      "learning_rate": 2.823491509912901e-08,
-      "loss": 0.7177,
-      "step": 35566
-    },
-    {
-      "epoch": 0.9768201917003103,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8168157813949215e-08,
-      "loss": 0.836,
-      "step": 35567
-    },
-    {
-      "epoch": 0.9768476559281536,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8101479429766265e-08,
-      "loss": 0.9419,
-      "step": 35568
-    },
-    {
-      "epoch": 0.9768751201559969,
-      "grad_norm": 0.0,
-      "learning_rate": 2.803487994710641e-08,
-      "loss": 0.7967,
-      "step": 35569
-    },
-    {
-      "epoch": 0.97690258438384,
-      "grad_norm": 0.0,
-      "learning_rate": 2.796835936649811e-08,
-      "loss": 0.8203,
-      "step": 35570
-    },
-    {
-      "epoch": 0.9769300486116833,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7901917688466507e-08,
-      "loss": 0.7487,
-      "step": 35571
-    },
-    {
-      "epoch": 0.9769575128395265,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7835554913537844e-08,
-      "loss": 0.8095,
-      "step": 35572
-    },
-    {
-      "epoch": 0.9769849770673698,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7769271042236146e-08,
-      "loss": 0.7942,
-      "step": 35573
-    },
-    {
-      "epoch": 0.977012441295213,
-      "grad_norm": 0.0,
-      "learning_rate": 2.770306607508766e-08,
-      "loss": 0.6541,
-      "step": 35574
-    },
-    {
-      "epoch": 0.9770399055230562,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7636940012615297e-08,
-      "loss": 0.681,
-      "step": 35575
-    },
-    {
-      "epoch": 0.9770673697508995,
-      "grad_norm": 0.0,
-      "learning_rate": 2.757089285534309e-08,
-      "loss": 0.7941,
-      "step": 35576
-    },
-    {
-      "epoch": 0.9770948339787426,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7504924603792838e-08,
-      "loss": 0.8779,
-      "step": 35577
-    },
-    {
-      "epoch": 0.9771222982065859,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7439035258485236e-08,
-      "loss": 0.9411,
-      "step": 35578
-    },
-    {
-      "epoch": 0.9771497624344292,
-      "grad_norm": 0.0,
-      "learning_rate": 2.737322481994431e-08,
-      "loss": 0.7814,
-      "step": 35579
-    },
-    {
-      "epoch": 0.9771772266622724,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7307493288688534e-08,
-      "loss": 0.8644,
-      "step": 35580
-    },
-    {
-      "epoch": 0.9772046908901156,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7241840665239717e-08,
-      "loss": 0.7757,
-      "step": 35581
-    },
-    {
-      "epoch": 0.9772321551179589,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7176266950116324e-08,
-      "loss": 0.8287,
-      "step": 35582
-    },
-    {
-      "epoch": 0.9772596193458021,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7110772143837948e-08,
-      "loss": 0.7146,
-      "step": 35583
-    },
-    {
-      "epoch": 0.9772870835736454,
-      "grad_norm": 0.0,
-      "learning_rate": 2.704535624692084e-08,
-      "loss": 0.7506,
-      "step": 35584
-    },
-    {
-      "epoch": 0.9773145478014885,
-      "grad_norm": 0.0,
-      "learning_rate": 2.698001925988458e-08,
-      "loss": 0.825,
-      "step": 35585
-    },
-    {
-      "epoch": 0.9773420120293318,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6914761183246542e-08,
-      "loss": 0.851,
-      "step": 35586
-    },
-    {
-      "epoch": 0.9773694762571751,
-      "grad_norm": 0.0,
-      "learning_rate": 2.684958201752075e-08,
-      "loss": 0.8285,
-      "step": 35587
-    },
-    {
-      "epoch": 0.9773969404850182,
-      "grad_norm": 0.0,
-      "learning_rate": 2.678448176322568e-08,
-      "loss": 0.8593,
-      "step": 35588
-    },
-    {
-      "epoch": 0.9774244047128615,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6719460420874254e-08,
-      "loss": 0.8436,
-      "step": 35589
-    },
-    {
-      "epoch": 0.9774518689407047,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6654517990981622e-08,
-      "loss": 0.9134,
-      "step": 35590
-    },
-    {
-      "epoch": 0.977479333168548,
-      "grad_norm": 0.0,
-      "learning_rate": 2.658965447406181e-08,
-      "loss": 0.8169,
-      "step": 35591
-    },
-    {
-      "epoch": 0.9775067973963912,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6524869870627744e-08,
-      "loss": 0.8406,
-      "step": 35592
-    },
-    {
-      "epoch": 0.9775342616242344,
-      "grad_norm": 0.0,
-      "learning_rate": 2.646016418119346e-08,
-      "loss": 0.8529,
-      "step": 35593
-    },
-    {
-      "epoch": 0.9775617258520777,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6395537406268545e-08,
-      "loss": 0.6757,
-      "step": 35594
-    },
-    {
-      "epoch": 0.977589190079921,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6330989546365926e-08,
-      "loss": 0.7354,
-      "step": 35595
-    },
-    {
-      "epoch": 0.9776166543077641,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6266520601995195e-08,
-      "loss": 0.843,
-      "step": 35596
-    },
-    {
-      "epoch": 0.9776441185356074,
-      "grad_norm": 0.0,
-      "learning_rate": 2.620213057366816e-08,
-      "loss": 0.7803,
-      "step": 35597
-    },
-    {
-      "epoch": 0.9776715827634506,
-      "grad_norm": 0.0,
-      "learning_rate": 2.613781946189331e-08,
-      "loss": 0.7006,
-      "step": 35598
-    },
-    {
-      "epoch": 0.9776990469912938,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6073587267179123e-08,
-      "loss": 0.8266,
-      "step": 35599
-    },
-    {
-      "epoch": 0.9777265112191371,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6009433990035194e-08,
-      "loss": 0.8131,
-      "step": 35600
-    },
-    {
-      "epoch": 0.9777539754469803,
-      "grad_norm": 0.0,
-      "learning_rate": 2.594535963096667e-08,
-      "loss": 0.833,
-      "step": 35601
-    },
-    {
-      "epoch": 0.9777814396748236,
-      "grad_norm": 0.0,
-      "learning_rate": 2.588136419048315e-08,
-      "loss": 0.9041,
-      "step": 35602
-    },
-    {
-      "epoch": 0.9778089039026667,
-      "grad_norm": 0.0,
-      "learning_rate": 2.581744766908978e-08,
-      "loss": 0.7754,
-      "step": 35603
-    },
-    {
-      "epoch": 0.97783636813051,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5753610067291714e-08,
-      "loss": 0.8749,
-      "step": 35604
-    },
-    {
-      "epoch": 0.9778638323583533,
-      "grad_norm": 0.0,
-      "learning_rate": 2.568985138559521e-08,
-      "loss": 0.7726,
-      "step": 35605
-    },
-    {
-      "epoch": 0.9778912965861964,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5626171624504315e-08,
-      "loss": 0.8903,
-      "step": 35606
-    },
-    {
-      "epoch": 0.9779187608140397,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5562570784523068e-08,
-      "loss": 0.9065,
-      "step": 35607
-    },
-    {
-      "epoch": 0.977946225041883,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5499048866154396e-08,
-      "loss": 0.8116,
-      "step": 35608
-    },
-    {
-      "epoch": 0.9779736892697262,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5435605869900127e-08,
-      "loss": 0.8527,
-      "step": 35609
-    },
-    {
-      "epoch": 0.9780011534975694,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5372241796264295e-08,
-      "loss": 0.8671,
-      "step": 35610
-    },
-    {
-      "epoch": 0.9780286177254126,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5308956645746507e-08,
-      "loss": 0.7962,
-      "step": 35611
-    },
-    {
-      "epoch": 0.9780560819532559,
-      "grad_norm": 0.0,
-      "learning_rate": 2.524575041884858e-08,
-      "loss": 0.7833,
-      "step": 35612
-    },
-    {
-      "epoch": 0.9780835461810992,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5182623116069004e-08,
-      "loss": 0.7987,
-      "step": 35613
-    },
-    {
-      "epoch": 0.9781110104089423,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5119574737909602e-08,
-      "loss": 0.7979,
-      "step": 35614
-    },
-    {
-      "epoch": 0.9781384746367856,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5056605284867752e-08,
-      "loss": 0.7177,
-      "step": 35615
-    },
-    {
-      "epoch": 0.9781659388646288,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4993714757440833e-08,
-      "loss": 0.8037,
-      "step": 35616
-    },
-    {
-      "epoch": 0.978193403092472,
-      "grad_norm": 0.0,
-      "learning_rate": 2.493090315612845e-08,
-      "loss": 0.7959,
-      "step": 35617
-    },
-    {
-      "epoch": 0.9782208673203153,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4868170481427977e-08,
-      "loss": 0.8347,
-      "step": 35618
-    },
-    {
-      "epoch": 0.9782483315481585,
-      "grad_norm": 0.0,
-      "learning_rate": 2.480551673383236e-08,
-      "loss": 0.8424,
-      "step": 35619
-    },
-    {
-      "epoch": 0.9782757957760018,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4742941913841188e-08,
-      "loss": 0.8694,
-      "step": 35620
-    },
-    {
-      "epoch": 0.978303260003845,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4680446021946303e-08,
-      "loss": 0.8531,
-      "step": 35621
-    },
-    {
-      "epoch": 0.9783307242316882,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4618029058645076e-08,
-      "loss": 0.8853,
-      "step": 35622
-    },
-    {
-      "epoch": 0.9783581884595315,
-      "grad_norm": 0.0,
-      "learning_rate": 2.455569102443045e-08,
-      "loss": 0.7879,
-      "step": 35623
-    },
-    {
-      "epoch": 0.9783856526873747,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4493431919794253e-08,
-      "loss": 0.8854,
-      "step": 35624
-    },
-    {
-      "epoch": 0.9784131169152179,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4431251745230532e-08,
-      "loss": 0.847,
-      "step": 35625
-    },
-    {
-      "epoch": 0.9784405811430612,
-      "grad_norm": 0.0,
-      "learning_rate": 2.436915050123112e-08,
-      "loss": 0.7995,
-      "step": 35626
-    },
-    {
-      "epoch": 0.9784680453709044,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4307128188286734e-08,
-      "loss": 0.7821,
-      "step": 35627
-    },
-    {
-      "epoch": 0.9784955095987476,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4245184806888088e-08,
-      "loss": 0.7752,
-      "step": 35628
-    },
-    {
-      "epoch": 0.9785229738265908,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4183320357527017e-08,
-      "loss": 0.8089,
-      "step": 35629
-    },
-    {
-      "epoch": 0.9785504380544341,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4121534840690907e-08,
-      "loss": 0.802,
-      "step": 35630
-    },
-    {
-      "epoch": 0.9785779022822774,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4059828256870477e-08,
-      "loss": 0.7946,
-      "step": 35631
-    },
-    {
-      "epoch": 0.9786053665101205,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3998200606552004e-08,
-      "loss": 0.8511,
-      "step": 35632
-    },
-    {
-      "epoch": 0.9786328307379638,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3936651890223984e-08,
-      "loss": 0.7965,
-      "step": 35633
-    },
-    {
-      "epoch": 0.9786602949658071,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3875182108373808e-08,
-      "loss": 0.8514,
-      "step": 35634
-    },
-    {
-      "epoch": 0.9786877591936503,
-      "grad_norm": 0.0,
-      "learning_rate": 2.381379126148886e-08,
-      "loss": 0.785,
-      "step": 35635
-    },
-    {
-      "epoch": 0.9787152234214935,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3752479350052092e-08,
-      "loss": 0.7744,
-      "step": 35636
-    },
-    {
-      "epoch": 0.9787426876493367,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3691246374549783e-08,
-      "loss": 0.8022,
-      "step": 35637
-    },
-    {
-      "epoch": 0.97877015187718,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3630092335468203e-08,
-      "loss": 0.8318,
-      "step": 35638
-    },
-    {
-      "epoch": 0.9787976161050232,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3569017233289192e-08,
-      "loss": 0.8284,
-      "step": 35639
-    },
-    {
-      "epoch": 0.9788250803328664,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3508021068496813e-08,
-      "loss": 0.8179,
-      "step": 35640
-    },
-    {
-      "epoch": 0.9788525445607097,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3447103841572895e-08,
-      "loss": 0.7375,
-      "step": 35641
-    },
-    {
-      "epoch": 0.9788800087885529,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3386265553000385e-08,
-      "loss": 0.8422,
-      "step": 35642
-    },
-    {
-      "epoch": 0.9789074730163961,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3325506203260016e-08,
-      "loss": 0.7961,
-      "step": 35643
-    },
-    {
-      "epoch": 0.9789349372442394,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3264825792833622e-08,
-      "loss": 0.9047,
-      "step": 35644
-    },
-    {
-      "epoch": 0.9789624014720826,
-      "grad_norm": 0.0,
-      "learning_rate": 2.320422432219971e-08,
-      "loss": 0.7986,
-      "step": 35645
-    },
-    {
-      "epoch": 0.9789898656999259,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3143701791839003e-08,
-      "loss": 0.6917,
-      "step": 35646
-    },
-    {
-      "epoch": 0.9790173299277691,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3083258202230008e-08,
-      "loss": 0.795,
-      "step": 35647
-    },
-    {
-      "epoch": 0.9790447941556123,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3022893553851233e-08,
-      "loss": 0.8055,
-      "step": 35648
-    },
-    {
-      "epoch": 0.9790722583834556,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2962607847178964e-08,
-      "loss": 0.8835,
-      "step": 35649
-    },
-    {
-      "epoch": 0.9790997226112987,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2902401082692817e-08,
-      "loss": 0.7848,
-      "step": 35650
-    },
-    {
-      "epoch": 0.979127186839142,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2842273260865744e-08,
-      "loss": 0.7949,
-      "step": 35651
-    },
-    {
-      "epoch": 0.9791546510669853,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2782224382177365e-08,
-      "loss": 0.7047,
-      "step": 35652
-    },
-    {
-      "epoch": 0.9791821152948285,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2722254447099523e-08,
-      "loss": 0.8129,
-      "step": 35653
-    },
-    {
-      "epoch": 0.9792095795226717,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2662363456108506e-08,
-      "loss": 0.7963,
-      "step": 35654
-    },
-    {
-      "epoch": 0.9792370437505149,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2602551409678374e-08,
-      "loss": 0.8778,
-      "step": 35655
-    },
-    {
-      "epoch": 0.9792645079783582,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2542818308280977e-08,
-      "loss": 0.839,
-      "step": 35656
-    },
-    {
-      "epoch": 0.9792919722062015,
-      "grad_norm": 0.0,
-      "learning_rate": 2.248316415238927e-08,
-      "loss": 0.7242,
-      "step": 35657
-    },
-    {
-      "epoch": 0.9793194364340446,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2423588942476203e-08,
-      "loss": 0.7792,
-      "step": 35658
-    },
-    {
-      "epoch": 0.9793469006618879,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2364092679012518e-08,
-      "loss": 0.8188,
-      "step": 35659
-    },
-    {
-      "epoch": 0.9793743648897312,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2304675362468942e-08,
-      "loss": 0.8445,
-      "step": 35660
-    },
-    {
-      "epoch": 0.9794018291175743,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2245336993316212e-08,
-      "loss": 0.9426,
-      "step": 35661
-    },
-    {
-      "epoch": 0.9794292933454176,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2186077572022845e-08,
-      "loss": 0.8157,
-      "step": 35662
-    },
-    {
-      "epoch": 0.9794567575732608,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2126897099058465e-08,
-      "loss": 0.7697,
-      "step": 35663
-    },
-    {
-      "epoch": 0.9794842218011041,
-      "grad_norm": 0.0,
-      "learning_rate": 2.206779557489158e-08,
-      "loss": 0.7528,
-      "step": 35664
-    },
-    {
-      "epoch": 0.9795116860289473,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2008772999989603e-08,
-      "loss": 0.768,
-      "step": 35665
-    },
-    {
-      "epoch": 0.9795391502567905,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1949829374818822e-08,
-      "loss": 0.8858,
-      "step": 35666
-    },
-    {
-      "epoch": 0.9795666144846338,
-      "grad_norm": 0.0,
-      "learning_rate": 2.189096469984553e-08,
-      "loss": 0.8118,
-      "step": 35667
-    },
-    {
-      "epoch": 0.979594078712477,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1832178975537133e-08,
-      "loss": 0.761,
-      "step": 35668
-    },
-    {
-      "epoch": 0.9796215429403202,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1773472202356592e-08,
-      "loss": 0.7856,
-      "step": 35669
-    },
-    {
-      "epoch": 0.9796490071681635,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1714844380770204e-08,
-      "loss": 0.8194,
-      "step": 35670
-    },
-    {
-      "epoch": 0.9796764713960067,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1656295511242042e-08,
-      "loss": 0.7958,
-      "step": 35671
-    },
-    {
-      "epoch": 0.9797039356238499,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1597825594232847e-08,
-      "loss": 0.6917,
-      "step": 35672
-    },
-    {
-      "epoch": 0.9797313998516932,
-      "grad_norm": 0.0,
-      "learning_rate": 2.15394346302078e-08,
-      "loss": 0.7966,
-      "step": 35673
-    },
-    {
-      "epoch": 0.9797588640795364,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1481122619627648e-08,
-      "loss": 0.8497,
-      "step": 35674
-    },
-    {
-      "epoch": 0.9797863283073797,
-      "grad_norm": 0.0,
-      "learning_rate": 2.142288956295424e-08,
-      "loss": 0.8264,
-      "step": 35675
-    },
-    {
-      "epoch": 0.9798137925352228,
-      "grad_norm": 0.0,
-      "learning_rate": 2.136473546064832e-08,
-      "loss": 0.8304,
-      "step": 35676
-    },
-    {
-      "epoch": 0.9798412567630661,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1306660313170634e-08,
-      "loss": 0.7532,
-      "step": 35677
-    },
-    {
-      "epoch": 0.9798687209909094,
-      "grad_norm": 0.0,
-      "learning_rate": 2.12486641209797e-08,
-      "loss": 0.8541,
-      "step": 35678
-    },
-    {
-      "epoch": 0.9798961852187525,
-      "grad_norm": 0.0,
-      "learning_rate": 2.119074688453404e-08,
-      "loss": 0.8148,
-      "step": 35679
-    },
-    {
-      "epoch": 0.9799236494465958,
-      "grad_norm": 0.0,
-      "learning_rate": 2.113290860429329e-08,
-      "loss": 0.7275,
-      "step": 35680
-    },
-    {
-      "epoch": 0.979951113674439,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1075149280713746e-08,
-      "loss": 0.7584,
-      "step": 35681
-    },
-    {
-      "epoch": 0.9799785779022823,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1017468914253934e-08,
-      "loss": 0.7556,
-      "step": 35682
-    },
-    {
-      "epoch": 0.9800060421301255,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0959867505369047e-08,
-      "loss": 0.7654,
-      "step": 35683
-    },
-    {
-      "epoch": 0.9800335063579687,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0902345054515383e-08,
-      "loss": 0.9327,
-      "step": 35684
-    },
-    {
-      "epoch": 0.980060970585812,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0844901562147025e-08,
-      "loss": 0.7786,
-      "step": 35685
-    },
-    {
-      "epoch": 0.9800884348136553,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0787537028720272e-08,
-      "loss": 0.747,
-      "step": 35686
-    },
-    {
-      "epoch": 0.9801158990414984,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0730251454688098e-08,
-      "loss": 0.7779,
-      "step": 35687
-    },
-    {
-      "epoch": 0.9801433632693417,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0673044840503475e-08,
-      "loss": 0.8002,
-      "step": 35688
-    },
-    {
-      "epoch": 0.9801708274971849,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0615917186619372e-08,
-      "loss": 0.7547,
-      "step": 35689
-    },
-    {
-      "epoch": 0.9801982917250281,
-      "grad_norm": 0.0,
-      "learning_rate": 2.055886849348765e-08,
-      "loss": 0.7668,
-      "step": 35690
-    },
-    {
-      "epoch": 0.9802257559528714,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0501898761560167e-08,
-      "loss": 0.7754,
-      "step": 35691
-    },
-    {
-      "epoch": 0.9802532201807146,
-      "grad_norm": 0.0,
-      "learning_rate": 2.044500799128657e-08,
-      "loss": 0.8487,
-      "step": 35692
-    },
-    {
-      "epoch": 0.9802806844085579,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0388196183118713e-08,
-      "loss": 0.8012,
-      "step": 35693
-    },
-    {
-      "epoch": 0.980308148636401,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0331463337505132e-08,
-      "loss": 0.762,
-      "step": 35694
-    },
-    {
-      "epoch": 0.9803356128642443,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0274809454894352e-08,
-      "loss": 0.8678,
-      "step": 35695
-    },
-    {
-      "epoch": 0.9803630770920876,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0218234535734904e-08,
-      "loss": 0.7279,
-      "step": 35696
-    },
-    {
-      "epoch": 0.9803905413199308,
-      "grad_norm": 0.0,
-      "learning_rate": 2.016173858047532e-08,
-      "loss": 0.7866,
-      "step": 35697
-    },
-    {
-      "epoch": 0.980418005547774,
-      "grad_norm": 0.0,
-      "learning_rate": 2.01053215895608e-08,
-      "loss": 0.8802,
-      "step": 35698
-    },
-    {
-      "epoch": 0.9804454697756173,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0048983563439872e-08,
-      "loss": 0.8025,
-      "step": 35699
-    },
-    {
-      "epoch": 0.9804729340034605,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9992724502556626e-08,
-      "loss": 0.7425,
-      "step": 35700
-    },
-    {
-      "epoch": 0.9805003982313037,
-      "grad_norm": 0.0,
-      "learning_rate": 1.993654440735626e-08,
-      "loss": 0.7986,
-      "step": 35701
-    },
-    {
-      "epoch": 0.9805278624591469,
-      "grad_norm": 0.0,
-      "learning_rate": 1.988044327828509e-08,
-      "loss": 0.7833,
-      "step": 35702
-    },
-    {
-      "epoch": 0.9805553266869902,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9824421115784975e-08,
-      "loss": 0.844,
-      "step": 35703
-    },
-    {
-      "epoch": 0.9805827909148335,
-      "grad_norm": 0.0,
-      "learning_rate": 1.976847792030112e-08,
-      "loss": 0.7009,
-      "step": 35704
-    },
-    {
-      "epoch": 0.9806102551426766,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9712613692273174e-08,
-      "loss": 0.7981,
-      "step": 35705
-    },
-    {
-      "epoch": 0.9806377193705199,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9656828432146336e-08,
-      "loss": 0.8349,
-      "step": 35706
-    },
-    {
-      "epoch": 0.9806651835983631,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9601122140360252e-08,
-      "loss": 0.7586,
-      "step": 35707
-    },
-    {
-      "epoch": 0.9806926478262064,
-      "grad_norm": 0.0,
-      "learning_rate": 1.954549481735679e-08,
-      "loss": 0.7956,
-      "step": 35708
-    },
-    {
-      "epoch": 0.9807201120540496,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9489946463574495e-08,
-      "loss": 0.7749,
-      "step": 35709
-    },
-    {
-      "epoch": 0.9807475762818928,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9434477079455228e-08,
-      "loss": 0.7775,
-      "step": 35710
-    },
-    {
-      "epoch": 0.9807750405097361,
-      "grad_norm": 0.0,
-      "learning_rate": 1.937908666543531e-08,
-      "loss": 0.8458,
-      "step": 35711
-    },
-    {
-      "epoch": 0.9808025047375794,
-      "grad_norm": 0.0,
-      "learning_rate": 1.932377522195439e-08,
-      "loss": 0.8096,
-      "step": 35712
-    },
-    {
-      "epoch": 0.9808299689654225,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9268542749449893e-08,
-      "loss": 0.7924,
-      "step": 35713
-    },
-    {
-      "epoch": 0.9808574331932658,
-      "grad_norm": 0.0,
-      "learning_rate": 1.921338924835814e-08,
-      "loss": 0.8011,
-      "step": 35714
-    },
-    {
-      "epoch": 0.980884897421109,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9158314719117664e-08,
-      "loss": 0.9462,
-      "step": 35715
-    },
-    {
-      "epoch": 0.9809123616489522,
-      "grad_norm": 0.0,
-      "learning_rate": 1.910331916216146e-08,
-      "loss": 0.8349,
-      "step": 35716
-    },
-    {
-      "epoch": 0.9809398258767955,
-      "grad_norm": 0.0,
-      "learning_rate": 1.904840257792695e-08,
-      "loss": 0.7265,
-      "step": 35717
-    },
-    {
-      "epoch": 0.9809672901046387,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8993564966847122e-08,
-      "loss": 0.7942,
-      "step": 35718
-    },
-    {
-      "epoch": 0.980994754332482,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8938806329357184e-08,
-      "loss": 0.8918,
-      "step": 35719
-    },
-    {
-      "epoch": 0.9810222185603251,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8884126665889012e-08,
-      "loss": 0.7976,
-      "step": 35720
-    },
-    {
-      "epoch": 0.9810496827881684,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8829525976876707e-08,
-      "loss": 0.8424,
-      "step": 35721
-    },
-    {
-      "epoch": 0.9810771470160117,
-      "grad_norm": 0.0,
-      "learning_rate": 1.877500426275103e-08,
-      "loss": 0.7958,
-      "step": 35722
-    },
-    {
-      "epoch": 0.9811046112438548,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8720561523943858e-08,
-      "loss": 0.8281,
-      "step": 35723
-    },
-    {
-      "epoch": 0.9811320754716981,
-      "grad_norm": 0.0,
-      "learning_rate": 1.866619776088596e-08,
-      "loss": 0.7888,
-      "step": 35724
-    },
-    {
-      "epoch": 0.9811595396995414,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8611912974008106e-08,
-      "loss": 0.7941,
-      "step": 35725
-    },
-    {
-      "epoch": 0.9811870039273846,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8557707163737725e-08,
-      "loss": 0.8092,
-      "step": 35726
-    },
-    {
-      "epoch": 0.9812144681552278,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8503580330506697e-08,
-      "loss": 0.7442,
-      "step": 35727
-    },
-    {
-      "epoch": 0.981241932383071,
-      "grad_norm": 0.0,
-      "learning_rate": 1.844953247474135e-08,
-      "loss": 0.8531,
-      "step": 35728
-    },
-    {
-      "epoch": 0.9812693966109143,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8395563596869116e-08,
-      "loss": 0.8059,
-      "step": 35729
-    },
-    {
-      "epoch": 0.9812968608387576,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8341673697318542e-08,
-      "loss": 0.8316,
-      "step": 35730
-    },
-    {
-      "epoch": 0.9813243250666007,
-      "grad_norm": 0.0,
-      "learning_rate": 1.828786277651484e-08,
-      "loss": 0.7745,
-      "step": 35731
-    },
-    {
-      "epoch": 0.981351789294444,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8234130834884346e-08,
-      "loss": 0.8323,
-      "step": 35732
-    },
-    {
-      "epoch": 0.9813792535222873,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8180477872852263e-08,
-      "loss": 0.7656,
-      "step": 35733
-    },
-    {
-      "epoch": 0.9814067177501304,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8126903890842705e-08,
-      "loss": 0.8521,
-      "step": 35734
-    },
-    {
-      "epoch": 0.9814341819779737,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8073408889278666e-08,
-      "loss": 0.6946,
-      "step": 35735
-    },
-    {
-      "epoch": 0.9814616462058169,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8019992868585356e-08,
-      "loss": 0.7841,
-      "step": 35736
-    },
-    {
-      "epoch": 0.9814891104336602,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7966655829184663e-08,
-      "loss": 0.8566,
-      "step": 35737
-    },
-    {
-      "epoch": 0.9815165746615034,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7913397771497365e-08,
-      "loss": 0.8726,
-      "step": 35738
-    },
-    {
-      "epoch": 0.9815440388893466,
-      "grad_norm": 0.0,
-      "learning_rate": 1.786021869594756e-08,
-      "loss": 0.7574,
-      "step": 35739
-    },
-    {
-      "epoch": 0.9815715031171899,
-      "grad_norm": 0.0,
-      "learning_rate": 1.78071186029527e-08,
-      "loss": 0.763,
-      "step": 35740
-    },
-    {
-      "epoch": 0.981598967345033,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7754097492935773e-08,
-      "loss": 0.7799,
-      "step": 35741
-    },
-    {
-      "epoch": 0.9816264315728763,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7701155366315337e-08,
-      "loss": 0.7734,
-      "step": 35742
-    },
-    {
-      "epoch": 0.9816538958007196,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7648292223509943e-08,
-      "loss": 0.8454,
-      "step": 35743
-    },
-    {
-      "epoch": 0.9816813600285628,
-      "grad_norm": 0.0,
-      "learning_rate": 1.759550806493704e-08,
-      "loss": 0.7194,
-      "step": 35744
-    },
-    {
-      "epoch": 0.981708824256406,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7542802891016285e-08,
-      "loss": 0.9198,
-      "step": 35745
-    },
-    {
-      "epoch": 0.9817362884842493,
-      "grad_norm": 0.0,
-      "learning_rate": 1.749017670216291e-08,
-      "loss": 0.7845,
-      "step": 35746
-    },
-    {
-      "epoch": 0.9817637527120925,
-      "grad_norm": 0.0,
-      "learning_rate": 1.743762949879546e-08,
-      "loss": 0.7979,
-      "step": 35747
-    },
-    {
-      "epoch": 0.9817912169399358,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7385161281326947e-08,
-      "loss": 0.7784,
-      "step": 35748
-    },
-    {
-      "epoch": 0.9818186811677789,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7332772050174808e-08,
-      "loss": 0.6983,
-      "step": 35749
-    },
-    {
-      "epoch": 0.9818461453956222,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7280461805752047e-08,
-      "loss": 0.7957,
-      "step": 35750
-    },
-    {
-      "epoch": 0.9818736096234655,
-      "grad_norm": 0.0,
-      "learning_rate": 1.722823054847389e-08,
-      "loss": 0.7586,
-      "step": 35751
-    },
-    {
-      "epoch": 0.9819010738513086,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7176078278753338e-08,
-      "loss": 0.7243,
-      "step": 35752
-    },
-    {
-      "epoch": 0.9819285380791519,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7124004997002287e-08,
-      "loss": 0.8216,
-      "step": 35753
-    },
-    {
-      "epoch": 0.9819560023069951,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7072010703632624e-08,
-      "loss": 0.7686,
-      "step": 35754
-    },
-    {
-      "epoch": 0.9819834665348384,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7020095399056247e-08,
-      "loss": 0.7296,
-      "step": 35755
-    },
-    {
-      "epoch": 0.9820109307626816,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6968259083683937e-08,
-      "loss": 0.7211,
-      "step": 35756
-    },
-    {
-      "epoch": 0.9820383949905248,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6916501757925364e-08,
-      "loss": 0.8228,
-      "step": 35757
-    },
-    {
-      "epoch": 0.9820658592183681,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6864823422191313e-08,
-      "loss": 0.7389,
-      "step": 35758
-    },
-    {
-      "epoch": 0.9820933234462114,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6813224076889235e-08,
-      "loss": 0.7766,
-      "step": 35759
-    },
-    {
-      "epoch": 0.9821207876740545,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6761703722428802e-08,
-      "loss": 0.8132,
-      "step": 35760
-    },
-    {
-      "epoch": 0.9821482519018978,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6710262359216356e-08,
-      "loss": 0.8142,
-      "step": 35761
-    },
-    {
-      "epoch": 0.982175716129741,
-      "grad_norm": 0.0,
-      "learning_rate": 1.665889998766046e-08,
-      "loss": 0.7915,
-      "step": 35762
-    },
-    {
-      "epoch": 0.9822031803575842,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6607616608165235e-08,
-      "loss": 0.7373,
-      "step": 35763
-    },
-    {
-      "epoch": 0.9822306445854275,
-      "grad_norm": 0.0,
-      "learning_rate": 1.655641222113813e-08,
-      "loss": 0.8453,
-      "step": 35764
-    },
-    {
-      "epoch": 0.9822581088132707,
-      "grad_norm": 0.0,
-      "learning_rate": 1.650528682698438e-08,
-      "loss": 0.8522,
-      "step": 35765
-    },
-    {
-      "epoch": 0.982285573041114,
-      "grad_norm": 0.0,
-      "learning_rate": 1.645424042610699e-08,
-      "loss": 0.8192,
-      "step": 35766
-    },
-    {
-      "epoch": 0.9823130372689571,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6403273018912312e-08,
-      "loss": 0.8678,
-      "step": 35767
-    },
-    {
-      "epoch": 0.9823405014968004,
-      "grad_norm": 0.0,
-      "learning_rate": 1.635238460580113e-08,
-      "loss": 0.9337,
-      "step": 35768
-    },
-    {
-      "epoch": 0.9823679657246437,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6301575187177565e-08,
-      "loss": 0.7176,
-      "step": 35769
-    },
-    {
-      "epoch": 0.9823954299524869,
-      "grad_norm": 0.0,
-      "learning_rate": 1.625084476344352e-08,
-      "loss": 0.7248,
-      "step": 35770
-    },
-    {
-      "epoch": 0.9824228941803301,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6200193334999782e-08,
-      "loss": 0.8064,
-      "step": 35771
-    },
-    {
-      "epoch": 0.9824503584081734,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6149620902247144e-08,
-      "loss": 0.7394,
-      "step": 35772
-    },
-    {
-      "epoch": 0.9824778226360166,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6099127465586395e-08,
-      "loss": 0.8567,
-      "step": 35773
-    },
-    {
-      "epoch": 0.9825052868638599,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6048713025417217e-08,
-      "loss": 0.8486,
-      "step": 35774
-    },
-    {
-      "epoch": 0.982532751091703,
-      "grad_norm": 0.0,
-      "learning_rate": 1.599837758213707e-08,
-      "loss": 0.7682,
-      "step": 35775
-    },
-    {
-      "epoch": 0.9825602153195463,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5948121136145634e-08,
-      "loss": 0.7414,
-      "step": 35776
-    },
-    {
-      "epoch": 0.9825876795473896,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5897943687840367e-08,
-      "loss": 0.9304,
-      "step": 35777
-    },
-    {
-      "epoch": 0.9826151437752327,
-      "grad_norm": 0.0,
-      "learning_rate": 1.584784523761762e-08,
-      "loss": 0.7685,
-      "step": 35778
-    },
-    {
-      "epoch": 0.982642608003076,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5797825785874853e-08,
-      "loss": 0.8323,
-      "step": 35779
-    },
-    {
-      "epoch": 0.9826700722309192,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5747885333006195e-08,
-      "loss": 0.7517,
-      "step": 35780
-    },
-    {
-      "epoch": 0.9826975364587625,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5698023879409108e-08,
-      "loss": 0.7612,
-      "step": 35781
-    },
-    {
-      "epoch": 0.9827250006866057,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5648241425476607e-08,
-      "loss": 0.787,
-      "step": 35782
-    },
-    {
-      "epoch": 0.9827524649144489,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5598537971601713e-08,
-      "loss": 0.7897,
-      "step": 35783
-    },
-    {
-      "epoch": 0.9827799291422922,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5548913518179666e-08,
-      "loss": 0.7304,
-      "step": 35784
-    },
-    {
-      "epoch": 0.9828073933701355,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5499368065601263e-08,
-      "loss": 0.7513,
-      "step": 35785
-    },
-    {
-      "epoch": 0.9828348575979786,
-      "grad_norm": 0.0,
-      "learning_rate": 1.544990161426063e-08,
-      "loss": 0.8659,
-      "step": 35786
-    },
-    {
-      "epoch": 0.9828623218258219,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5400514164547463e-08,
-      "loss": 0.8355,
-      "step": 35787
-    },
-    {
-      "epoch": 0.9828897860536651,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5351205716852558e-08,
-      "loss": 0.9044,
-      "step": 35788
-    },
-    {
-      "epoch": 0.9829172502815083,
-      "grad_norm": 0.0,
-      "learning_rate": 1.530197627156782e-08,
-      "loss": 0.8539,
-      "step": 35789
-    },
-    {
-      "epoch": 0.9829447145093516,
-      "grad_norm": 0.0,
-      "learning_rate": 1.525282582908072e-08,
-      "loss": 0.6931,
-      "step": 35790
-    },
-    {
-      "epoch": 0.9829721787371948,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5203754389780944e-08,
-      "loss": 0.835,
-      "step": 35791
-    },
-    {
-      "epoch": 0.9829996429650381,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5154761954057075e-08,
-      "loss": 0.8806,
-      "step": 35792
-    },
-    {
-      "epoch": 0.9830271071928812,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5105848522296573e-08,
-      "loss": 0.9185,
-      "step": 35793
-    },
-    {
-      "epoch": 0.9830545714207245,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5057014094886912e-08,
-      "loss": 0.8701,
-      "step": 35794
-    },
-    {
-      "epoch": 0.9830820356485678,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5008258672213338e-08,
-      "loss": 0.8322,
-      "step": 35795
-    },
-    {
-      "epoch": 0.9831094998764109,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4959582254662208e-08,
-      "loss": 0.8019,
-      "step": 35796
-    },
-    {
-      "epoch": 0.9831369641042542,
-      "grad_norm": 0.0,
-      "learning_rate": 1.491098484261766e-08,
-      "loss": 0.8406,
-      "step": 35797
-    },
-    {
-      "epoch": 0.9831644283320975,
-      "grad_norm": 0.0,
-      "learning_rate": 1.486246643646716e-08,
-      "loss": 0.8376,
-      "step": 35798
-    },
-    {
-      "epoch": 0.9831918925599407,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4814027036591515e-08,
-      "loss": 0.7834,
-      "step": 35799
-    },
-    {
-      "epoch": 0.9832193567877839,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4765666643375976e-08,
-      "loss": 0.8408,
-      "step": 35800
-    },
-    {
-      "epoch": 0.9832468210156271,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4717385257201345e-08,
-      "loss": 0.7462,
-      "step": 35801
-    },
-    {
-      "epoch": 0.9832742852434704,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4669182878451759e-08,
-      "loss": 0.778,
-      "step": 35802
-    },
-    {
-      "epoch": 0.9833017494713137,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4621059507506919e-08,
-      "loss": 0.7657,
-      "step": 35803
-    },
-    {
-      "epoch": 0.9833292136991568,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4573015144747626e-08,
-      "loss": 0.8036,
-      "step": 35804
-    },
-    {
-      "epoch": 0.9833566779270001,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4525049790555801e-08,
-      "loss": 0.8142,
-      "step": 35805
-    },
-    {
-      "epoch": 0.9833841421548433,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4477163445308918e-08,
-      "loss": 0.8157,
-      "step": 35806
-    },
-    {
-      "epoch": 0.9834116063826865,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4429356109387783e-08,
-      "loss": 0.8397,
-      "step": 35807
-    },
-    {
-      "epoch": 0.9834390706105298,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4381627783168762e-08,
-      "loss": 0.8181,
-      "step": 35808
-    },
-    {
-      "epoch": 0.983466534838373,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4333978467030441e-08,
-      "loss": 0.844,
-      "step": 35809
-    },
-    {
-      "epoch": 0.9834939990662163,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4286408161350296e-08,
-      "loss": 0.8059,
-      "step": 35810
-    },
-    {
-      "epoch": 0.9835214632940595,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4238916866503583e-08,
-      "loss": 0.7346,
-      "step": 35811
-    },
-    {
-      "epoch": 0.9835489275219027,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4191504582866666e-08,
-      "loss": 0.8023,
-      "step": 35812
-    },
-    {
-      "epoch": 0.983576391749746,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4144171310814803e-08,
-      "loss": 0.8168,
-      "step": 35813
-    },
-    {
-      "epoch": 0.9836038559775891,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4096917050722137e-08,
-      "loss": 0.7711,
-      "step": 35814
-    },
-    {
-      "epoch": 0.9836313202054324,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4049741802963923e-08,
-      "loss": 0.8547,
-      "step": 35815
-    },
-    {
-      "epoch": 0.9836587844332757,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4002645567910977e-08,
-      "loss": 0.8204,
-      "step": 35816
-    },
-    {
-      "epoch": 0.9836862486611189,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3955628345938554e-08,
-      "loss": 0.7371,
-      "step": 35817
-    },
-    {
-      "epoch": 0.9837137128889621,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3908690137416358e-08,
-      "loss": 0.7497,
-      "step": 35818
-    },
-    {
-      "epoch": 0.9837411771168053,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3861830942717424e-08,
-      "loss": 0.8683,
-      "step": 35819
-    },
-    {
-      "epoch": 0.9837686413446486,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3815050762212568e-08,
-      "loss": 0.8556,
-      "step": 35820
-    },
-    {
-      "epoch": 0.9837961055724919,
-      "grad_norm": 0.0,
-      "learning_rate": 1.376834959627038e-08,
-      "loss": 0.8203,
-      "step": 35821
-    },
-    {
-      "epoch": 0.983823569800335,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3721727445260568e-08,
-      "loss": 0.832,
-      "step": 35822
-    },
-    {
-      "epoch": 0.9838510340281783,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3675184309553945e-08,
-      "loss": 0.8015,
-      "step": 35823
-    },
-    {
-      "epoch": 0.9838784982560216,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3628720189517996e-08,
-      "loss": 0.7307,
-      "step": 35824
-    },
-    {
-      "epoch": 0.9839059624838647,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3582335085519093e-08,
-      "loss": 0.8556,
-      "step": 35825
-    },
-    {
-      "epoch": 0.983933426711708,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3536028997925832e-08,
-      "loss": 0.8022,
-      "step": 35826
-    },
-    {
-      "epoch": 0.9839608909395512,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3489801927102364e-08,
-      "loss": 0.8939,
-      "step": 35827
-    },
-    {
-      "epoch": 0.9839883551673945,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3443653873417284e-08,
-      "loss": 0.9226,
-      "step": 35828
-    },
-    {
-      "epoch": 0.9840158193952377,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3397584837234745e-08,
-      "loss": 0.8242,
-      "step": 35829
-    },
-    {
-      "epoch": 0.9840432836230809,
-      "grad_norm": 0.0,
-      "learning_rate": 1.335159481891779e-08,
-      "loss": 0.7379,
-      "step": 35830
-    },
-    {
-      "epoch": 0.9840707478509242,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3305683818832794e-08,
-      "loss": 0.7937,
-      "step": 35831
-    },
-    {
-      "epoch": 0.9840982120787674,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3259851837340576e-08,
-      "loss": 0.7481,
-      "step": 35832
-    },
-    {
-      "epoch": 0.9841256763066106,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3214098874805293e-08,
-      "loss": 0.7675,
-      "step": 35833
-    },
-    {
-      "epoch": 0.9841531405344539,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3168424931588875e-08,
-      "loss": 0.735,
-      "step": 35834
-    },
-    {
-      "epoch": 0.9841806047622971,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3122830008052145e-08,
-      "loss": 0.8407,
-      "step": 35835
-    },
-    {
-      "epoch": 0.9842080689901404,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3077314104555927e-08,
-      "loss": 0.7421,
-      "step": 35836
-    },
-    {
-      "epoch": 0.9842355332179836,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3031877221461042e-08,
-      "loss": 0.7497,
-      "step": 35837
-    },
-    {
-      "epoch": 0.9842629974458268,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2986519359126092e-08,
-      "loss": 0.8003,
-      "step": 35838
-    },
-    {
-      "epoch": 0.9842904616736701,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2941240517910792e-08,
-      "loss": 0.6814,
-      "step": 35839
-    },
-    {
-      "epoch": 0.9843179259015132,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2896040698172629e-08,
-      "loss": 0.7509,
-      "step": 35840
-    },
-    {
-      "epoch": 0.9843453901293565,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2850919900269098e-08,
-      "loss": 0.748,
-      "step": 35841
-    },
-    {
-      "epoch": 0.9843728543571998,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2805878124558802e-08,
-      "loss": 0.7625,
-      "step": 35842
-    },
-    {
-      "epoch": 0.984400318585043,
-      "grad_norm": 0.0,
-      "learning_rate": 1.276091537139701e-08,
-      "loss": 0.7565,
-      "step": 35843
-    },
-    {
-      "epoch": 0.9844277828128862,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2716031641138992e-08,
-      "loss": 0.8551,
-      "step": 35844
-    },
-    {
-      "epoch": 0.9844552470407294,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2671226934140025e-08,
-      "loss": 0.7812,
-      "step": 35845
-    },
-    {
-      "epoch": 0.9844827112685727,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2626501250756485e-08,
-      "loss": 0.7888,
-      "step": 35846
-    },
-    {
-      "epoch": 0.984510175496416,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2581854591339204e-08,
-      "loss": 0.8098,
-      "step": 35847
-    },
-    {
-      "epoch": 0.9845376397242591,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2537286956244565e-08,
-      "loss": 0.8192,
-      "step": 35848
-    },
-    {
-      "epoch": 0.9845651039521024,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2492798345822287e-08,
-      "loss": 0.8919,
-      "step": 35849
-    },
-    {
-      "epoch": 0.9845925681799457,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2448388760426533e-08,
-      "loss": 0.7049,
-      "step": 35850
-    },
-    {
-      "epoch": 0.9846200324077888,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2404058200408131e-08,
-      "loss": 0.8655,
-      "step": 35851
-    },
-    {
-      "epoch": 0.9846474966356321,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2359806666116802e-08,
-      "loss": 0.856,
-      "step": 35852
-    },
-    {
-      "epoch": 0.9846749608634753,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2315634157904488e-08,
-      "loss": 0.7888,
-      "step": 35853
-    },
-    {
-      "epoch": 0.9847024250913186,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2271540676118687e-08,
-      "loss": 0.8794,
-      "step": 35854
-    },
-    {
-      "epoch": 0.9847298893191618,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2227526221110231e-08,
-      "loss": 0.8535,
-      "step": 35855
-    },
-    {
-      "epoch": 0.984757353547005,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2183590793226619e-08,
-      "loss": 0.7964,
-      "step": 35856
-    },
-    {
-      "epoch": 0.9847848177748483,
-      "grad_norm": 0.0,
-      "learning_rate": 1.213973439281424e-08,
-      "loss": 0.7883,
-      "step": 35857
-    },
-    {
-      "epoch": 0.9848122820026914,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2095957020221705e-08,
-      "loss": 0.7553,
-      "step": 35858
-    },
-    {
-      "epoch": 0.9848397462305347,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2052258675795402e-08,
-      "loss": 0.8383,
-      "step": 35859
-    },
-    {
-      "epoch": 0.984867210458378,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2008639359880615e-08,
-      "loss": 0.747,
-      "step": 35860
-    },
-    {
-      "epoch": 0.9848946746862212,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1965099072821507e-08,
-      "loss": 0.7506,
-      "step": 35861
-    },
-    {
-      "epoch": 0.9849221389140644,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1921637814964471e-08,
-      "loss": 0.7618,
-      "step": 35862
-    },
-    {
-      "epoch": 0.9849496031419077,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1878255586652565e-08,
-      "loss": 0.8646,
-      "step": 35863
-    },
-    {
-      "epoch": 0.9849770673697509,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1834952388227739e-08,
-      "loss": 0.8453,
-      "step": 35864
-    },
-    {
-      "epoch": 0.9850045315975942,
-      "grad_norm": 0.0,
-      "learning_rate": 1.179172822003527e-08,
-      "loss": 0.8623,
-      "step": 35865
-    },
-    {
-      "epoch": 0.9850319958254373,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1748583082414889e-08,
-      "loss": 0.8238,
-      "step": 35866
-    },
-    {
-      "epoch": 0.9850594600532806,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1705516975708542e-08,
-      "loss": 0.7803,
-      "step": 35867
-    },
-    {
-      "epoch": 0.9850869242811239,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1662529900258179e-08,
-      "loss": 0.74,
-      "step": 35868
-    },
-    {
-      "epoch": 0.985114388508967,
-      "grad_norm": 0.0,
-      "learning_rate": 1.161962185640242e-08,
-      "loss": 0.813,
-      "step": 35869
-    },
-    {
-      "epoch": 0.9851418527368103,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1576792844480989e-08,
-      "loss": 0.8084,
-      "step": 35870
-    },
-    {
-      "epoch": 0.9851693169646535,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1534042864833617e-08,
-      "loss": 0.8197,
-      "step": 35871
-    },
-    {
-      "epoch": 0.9851967811924968,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1491371917797811e-08,
-      "loss": 0.8111,
-      "step": 35872
-    },
-    {
-      "epoch": 0.98522424542034,
-      "grad_norm": 0.0,
-      "learning_rate": 1.144878000371108e-08,
-      "loss": 0.7971,
-      "step": 35873
-    },
-    {
-      "epoch": 0.9852517096481832,
-      "grad_norm": 0.0,
-      "learning_rate": 1.140626712291204e-08,
-      "loss": 0.8184,
-      "step": 35874
-    },
-    {
-      "epoch": 0.9852791738760265,
-      "grad_norm": 0.0,
-      "learning_rate": 1.136383327573487e-08,
-      "loss": 0.8219,
-      "step": 35875
-    },
-    {
-      "epoch": 0.9853066381038698,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1321478462515966e-08,
-      "loss": 0.7938,
-      "step": 35876
-    },
-    {
-      "epoch": 0.9853341023317129,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1279202683591727e-08,
-      "loss": 0.7877,
-      "step": 35877
-    },
-    {
-      "epoch": 0.9853615665595562,
-      "grad_norm": 0.0,
-      "learning_rate": 1.123700593929522e-08,
-      "loss": 0.816,
-      "step": 35878
-    },
-    {
-      "epoch": 0.9853890307873994,
-      "grad_norm": 0.0,
-      "learning_rate": 1.119488822996062e-08,
-      "loss": 0.7926,
-      "step": 35879
-    },
-    {
-      "epoch": 0.9854164950152426,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1152849555922107e-08,
-      "loss": 0.8499,
-      "step": 35880
-    },
-    {
-      "epoch": 0.9854439592430859,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1110889917510526e-08,
-      "loss": 0.7802,
-      "step": 35881
-    },
-    {
-      "epoch": 0.9854714234709291,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1069009315058943e-08,
-      "loss": 0.875,
-      "step": 35882
-    },
-    {
-      "epoch": 0.9854988876987724,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1027207748899316e-08,
-      "loss": 0.8011,
-      "step": 35883
-    },
-    {
-      "epoch": 0.9855263519266155,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0985485219361381e-08,
-      "loss": 0.7933,
-      "step": 35884
-    },
-    {
-      "epoch": 0.9855538161544588,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0943841726775983e-08,
-      "loss": 0.7995,
-      "step": 35885
-    },
-    {
-      "epoch": 0.9855812803823021,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0902277271472861e-08,
-      "loss": 0.8123,
-      "step": 35886
-    },
-    {
-      "epoch": 0.9856087446101452,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0860791853779528e-08,
-      "loss": 0.9043,
-      "step": 35887
-    },
-    {
-      "epoch": 0.9856362088379885,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0819385474024613e-08,
-      "loss": 0.724,
-      "step": 35888
-    },
-    {
-      "epoch": 0.9856636730658318,
-      "grad_norm": 0.0,
-      "learning_rate": 1.077805813253674e-08,
-      "loss": 0.8139,
-      "step": 35889
-    },
-    {
-      "epoch": 0.985691137293675,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0736809829642314e-08,
-      "loss": 0.7796,
-      "step": 35890
-    },
-    {
-      "epoch": 0.9857186015215182,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0695640565667743e-08,
-      "loss": 0.7303,
-      "step": 35891
-    },
-    {
-      "epoch": 0.9857460657493614,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0654550340938319e-08,
-      "loss": 0.7959,
-      "step": 35892
-    },
-    {
-      "epoch": 0.9857735299772047,
-      "grad_norm": 0.0,
-      "learning_rate": 1.061353915578045e-08,
-      "loss": 0.7701,
-      "step": 35893
-    },
-    {
-      "epoch": 0.985800994205048,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0572607010517211e-08,
-      "loss": 0.8613,
-      "step": 35894
-    },
-    {
-      "epoch": 0.9858284584328911,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0531753905472786e-08,
-      "loss": 0.8788,
-      "step": 35895
-    },
-    {
-      "epoch": 0.9858559226607344,
-      "grad_norm": 0.0,
-      "learning_rate": 1.049097984097136e-08,
-      "loss": 0.7837,
-      "step": 35896
-    },
-    {
-      "epoch": 0.9858833868885776,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0450284817334899e-08,
-      "loss": 0.7453,
-      "step": 35897
-    },
-    {
-      "epoch": 0.9859108511164208,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0409668834885367e-08,
-      "loss": 0.7313,
-      "step": 35898
-    },
-    {
-      "epoch": 0.9859383153442641,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0369131893943618e-08,
-      "loss": 0.8195,
-      "step": 35899
-    },
-    {
-      "epoch": 0.9859657795721073,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0328673994830506e-08,
-      "loss": 0.8471,
-      "step": 35900
-    },
-    {
-      "epoch": 0.9859932437999506,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0288295137866888e-08,
-      "loss": 0.8285,
-      "step": 35901
-    },
-    {
-      "epoch": 0.9860207080277938,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0247995323372506e-08,
-      "loss": 0.7755,
-      "step": 35902
-    },
-    {
-      "epoch": 0.986048172255637,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0207774551665994e-08,
-      "loss": 0.8377,
-      "step": 35903
-    },
-    {
-      "epoch": 0.9860756364834803,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0167632823063767e-08,
-      "loss": 0.7882,
-      "step": 35904
-    },
-    {
-      "epoch": 0.9861031007113235,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0127570137885567e-08,
-      "loss": 0.895,
-      "step": 35905
-    },
-    {
-      "epoch": 0.9861305649391667,
-      "grad_norm": 0.0,
-      "learning_rate": 1.008758649644781e-08,
-      "loss": 0.7757,
-      "step": 35906
-    },
-    {
-      "epoch": 0.98615802916701,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0047681899066908e-08,
-      "loss": 0.8444,
-      "step": 35907
-    },
-    {
-      "epoch": 0.9861854933948532,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0007856346058164e-08,
-      "loss": 0.7144,
-      "step": 35908
-    },
-    {
-      "epoch": 0.9862129576226965,
-      "grad_norm": 0.0,
-      "learning_rate": 9.968109837737993e-09,
-      "loss": 0.8326,
-      "step": 35909
-    },
-    {
-      "epoch": 0.9862404218505396,
-      "grad_norm": 0.0,
-      "learning_rate": 9.928442374418368e-09,
-      "loss": 0.7976,
-      "step": 35910
-    },
-    {
-      "epoch": 0.9862678860783829,
-      "grad_norm": 0.0,
-      "learning_rate": 9.888853956415701e-09,
-      "loss": 0.8047,
-      "step": 35911
-    },
-    {
-      "epoch": 0.9862953503062262,
-      "grad_norm": 0.0,
-      "learning_rate": 9.849344584041964e-09,
-      "loss": 0.795,
-      "step": 35912
-    },
-    {
-      "epoch": 0.9863228145340693,
-      "grad_norm": 0.0,
-      "learning_rate": 9.809914257610242e-09,
-      "loss": 0.782,
-      "step": 35913
-    },
-    {
-      "epoch": 0.9863502787619126,
-      "grad_norm": 0.0,
-      "learning_rate": 9.770562977432508e-09,
-      "loss": 0.8436,
-      "step": 35914
-    },
-    {
-      "epoch": 0.9863777429897559,
-      "grad_norm": 0.0,
-      "learning_rate": 9.731290743819621e-09,
-      "loss": 0.7695,
-      "step": 35915
-    },
-    {
-      "epoch": 0.9864052072175991,
-      "grad_norm": 0.0,
-      "learning_rate": 9.692097557082447e-09,
-      "loss": 0.7863,
-      "step": 35916
-    },
-    {
-      "epoch": 0.9864326714454423,
-      "grad_norm": 0.0,
-      "learning_rate": 9.652983417530737e-09,
-      "loss": 0.797,
-      "step": 35917
-    },
-    {
-      "epoch": 0.9864601356732855,
-      "grad_norm": 0.0,
-      "learning_rate": 9.613948325475352e-09,
-      "loss": 0.8627,
-      "step": 35918
-    },
-    {
-      "epoch": 0.9864875999011288,
-      "grad_norm": 0.0,
-      "learning_rate": 9.574992281224938e-09,
-      "loss": 0.7739,
-      "step": 35919
-    },
-    {
-      "epoch": 0.986515064128972,
-      "grad_norm": 0.0,
-      "learning_rate": 9.536115285085913e-09,
-      "loss": 0.8571,
-      "step": 35920
-    },
-    {
-      "epoch": 0.9865425283568152,
-      "grad_norm": 0.0,
-      "learning_rate": 9.497317337368029e-09,
-      "loss": 0.8461,
-      "step": 35921
-    },
-    {
-      "epoch": 0.9865699925846585,
-      "grad_norm": 0.0,
-      "learning_rate": 9.458598438376598e-09,
-      "loss": 0.8046,
-      "step": 35922
-    },
-    {
-      "epoch": 0.9865974568125018,
-      "grad_norm": 0.0,
-      "learning_rate": 9.419958588418043e-09,
-      "loss": 0.8048,
-      "step": 35923
-    },
-    {
-      "epoch": 0.9866249210403449,
-      "grad_norm": 0.0,
-      "learning_rate": 9.381397787799896e-09,
-      "loss": 0.8101,
-      "step": 35924
-    },
-    {
-      "epoch": 0.9866523852681882,
-      "grad_norm": 0.0,
-      "learning_rate": 9.342916036826355e-09,
-      "loss": 0.7677,
-      "step": 35925
-    },
-    {
-      "epoch": 0.9866798494960314,
-      "grad_norm": 0.0,
-      "learning_rate": 9.304513335800514e-09,
-      "loss": 0.8355,
-      "step": 35926
-    },
-    {
-      "epoch": 0.9867073137238747,
-      "grad_norm": 0.0,
-      "learning_rate": 9.266189685027683e-09,
-      "loss": 0.7878,
-      "step": 35927
-    },
-    {
-      "epoch": 0.9867347779517179,
-      "grad_norm": 0.0,
-      "learning_rate": 9.227945084812063e-09,
-      "loss": 0.8591,
-      "step": 35928
-    },
-    {
-      "epoch": 0.9867622421795611,
-      "grad_norm": 0.0,
-      "learning_rate": 9.189779535453413e-09,
-      "loss": 0.8747,
-      "step": 35929
-    },
-    {
-      "epoch": 0.9867897064074044,
-      "grad_norm": 0.0,
-      "learning_rate": 9.151693037255938e-09,
-      "loss": 0.725,
-      "step": 35930
-    },
-    {
-      "epoch": 0.9868171706352475,
-      "grad_norm": 0.0,
-      "learning_rate": 9.113685590520504e-09,
-      "loss": 0.7993,
-      "step": 35931
-    },
-    {
-      "epoch": 0.9868446348630908,
-      "grad_norm": 0.0,
-      "learning_rate": 9.075757195547985e-09,
-      "loss": 0.767,
-      "step": 35932
-    },
-    {
-      "epoch": 0.9868720990909341,
-      "grad_norm": 0.0,
-      "learning_rate": 9.03790785263703e-09,
-      "loss": 0.9043,
-      "step": 35933
-    },
-    {
-      "epoch": 0.9868995633187773,
-      "grad_norm": 0.0,
-      "learning_rate": 9.000137562089616e-09,
-      "loss": 0.8497,
-      "step": 35934
-    },
-    {
-      "epoch": 0.9869270275466205,
-      "grad_norm": 0.0,
-      "learning_rate": 8.96244632420218e-09,
-      "loss": 0.7632,
-      "step": 35935
-    },
-    {
-      "epoch": 0.9869544917744638,
-      "grad_norm": 0.0,
-      "learning_rate": 8.924834139274475e-09,
-      "loss": 0.8416,
-      "step": 35936
-    },
-    {
-      "epoch": 0.986981956002307,
-      "grad_norm": 0.0,
-      "learning_rate": 8.887301007604044e-09,
-      "loss": 0.9154,
-      "step": 35937
-    },
-    {
-      "epoch": 0.9870094202301503,
-      "grad_norm": 0.0,
-      "learning_rate": 8.849846929486205e-09,
-      "loss": 0.8135,
-      "step": 35938
-    },
-    {
-      "epoch": 0.9870368844579934,
-      "grad_norm": 0.0,
-      "learning_rate": 8.81247190521961e-09,
-      "loss": 0.7455,
-      "step": 35939
-    },
-    {
-      "epoch": 0.9870643486858367,
-      "grad_norm": 0.0,
-      "learning_rate": 8.775175935099579e-09,
-      "loss": 0.7986,
-      "step": 35940
-    },
-    {
-      "epoch": 0.98709181291368,
-      "grad_norm": 0.0,
-      "learning_rate": 8.737959019419207e-09,
-      "loss": 0.8133,
-      "step": 35941
-    },
-    {
-      "epoch": 0.9871192771415231,
-      "grad_norm": 0.0,
-      "learning_rate": 8.700821158474927e-09,
-      "loss": 0.8081,
-      "step": 35942
-    },
-    {
-      "epoch": 0.9871467413693664,
-      "grad_norm": 0.0,
-      "learning_rate": 8.663762352559835e-09,
-      "loss": 0.7663,
-      "step": 35943
-    },
-    {
-      "epoch": 0.9871742055972096,
-      "grad_norm": 0.0,
-      "learning_rate": 8.626782601968143e-09,
-      "loss": 0.8438,
-      "step": 35944
-    },
-    {
-      "epoch": 0.9872016698250529,
-      "grad_norm": 0.0,
-      "learning_rate": 8.589881906990727e-09,
-      "loss": 0.7815,
-      "step": 35945
-    },
-    {
-      "epoch": 0.9872291340528961,
-      "grad_norm": 0.0,
-      "learning_rate": 8.553060267920688e-09,
-      "loss": 0.826,
-      "step": 35946
-    },
-    {
-      "epoch": 0.9872565982807393,
-      "grad_norm": 0.0,
-      "learning_rate": 8.516317685048903e-09,
-      "loss": 0.7326,
-      "step": 35947
-    },
-    {
-      "epoch": 0.9872840625085826,
-      "grad_norm": 0.0,
-      "learning_rate": 8.479654158667361e-09,
-      "loss": 0.8674,
-      "step": 35948
-    },
-    {
-      "epoch": 0.9873115267364259,
-      "grad_norm": 0.0,
-      "learning_rate": 8.443069689063609e-09,
-      "loss": 0.7788,
-      "step": 35949
-    },
-    {
-      "epoch": 0.987338990964269,
-      "grad_norm": 0.0,
-      "learning_rate": 8.406564276529639e-09,
-      "loss": 0.8452,
-      "step": 35950
-    },
-    {
-      "epoch": 0.9873664551921123,
-      "grad_norm": 0.0,
-      "learning_rate": 8.370137921352995e-09,
-      "loss": 0.7505,
-      "step": 35951
-    },
-    {
-      "epoch": 0.9873939194199555,
-      "grad_norm": 0.0,
-      "learning_rate": 8.333790623822335e-09,
-      "loss": 0.9017,
-      "step": 35952
-    },
-    {
-      "epoch": 0.9874213836477987,
-      "grad_norm": 0.0,
-      "learning_rate": 8.29752238422521e-09,
-      "loss": 0.7246,
-      "step": 35953
-    },
-    {
-      "epoch": 0.987448847875642,
-      "grad_norm": 0.0,
-      "learning_rate": 8.261333202849165e-09,
-      "loss": 0.7962,
-      "step": 35954
-    },
-    {
-      "epoch": 0.9874763121034852,
-      "grad_norm": 0.0,
-      "learning_rate": 8.225223079978417e-09,
-      "loss": 0.8228,
-      "step": 35955
-    },
-    {
-      "epoch": 0.9875037763313285,
-      "grad_norm": 0.0,
-      "learning_rate": 8.189192015901626e-09,
-      "loss": 0.8159,
-      "step": 35956
-    },
-    {
-      "epoch": 0.9875312405591716,
-      "grad_norm": 0.0,
-      "learning_rate": 8.153240010900787e-09,
-      "loss": 0.8593,
-      "step": 35957
-    },
-    {
-      "epoch": 0.9875587047870149,
-      "grad_norm": 0.0,
-      "learning_rate": 8.117367065263449e-09,
-      "loss": 0.833,
-      "step": 35958
-    },
-    {
-      "epoch": 0.9875861690148582,
-      "grad_norm": 0.0,
-      "learning_rate": 8.081573179271607e-09,
-      "loss": 0.7711,
-      "step": 35959
-    },
-    {
-      "epoch": 0.9876136332427013,
-      "grad_norm": 0.0,
-      "learning_rate": 8.045858353208369e-09,
-      "loss": 0.8198,
-      "step": 35960
-    },
-    {
-      "epoch": 0.9876410974705446,
-      "grad_norm": 0.0,
-      "learning_rate": 8.010222587356842e-09,
-      "loss": 0.7695,
-      "step": 35961
-    },
-    {
-      "epoch": 0.9876685616983879,
-      "grad_norm": 0.0,
-      "learning_rate": 7.974665881999023e-09,
-      "loss": 0.7646,
-      "step": 35962
-    },
-    {
-      "epoch": 0.9876960259262311,
-      "grad_norm": 0.0,
-      "learning_rate": 7.939188237416906e-09,
-      "loss": 0.7978,
-      "step": 35963
-    },
-    {
-      "epoch": 0.9877234901540743,
-      "grad_norm": 0.0,
-      "learning_rate": 7.903789653890271e-09,
-      "loss": 0.7711,
-      "step": 35964
-    },
-    {
-      "epoch": 0.9877509543819175,
-      "grad_norm": 0.0,
-      "learning_rate": 7.86847013169889e-09,
-      "loss": 0.8449,
-      "step": 35965
-    },
-    {
-      "epoch": 0.9877784186097608,
-      "grad_norm": 0.0,
-      "learning_rate": 7.833229671122544e-09,
-      "loss": 0.7292,
-      "step": 35966
-    },
-    {
-      "epoch": 0.9878058828376041,
-      "grad_norm": 0.0,
-      "learning_rate": 7.798068272441006e-09,
-      "loss": 0.8271,
-      "step": 35967
-    },
-    {
-      "epoch": 0.9878333470654472,
-      "grad_norm": 0.0,
-      "learning_rate": 7.762985935931833e-09,
-      "loss": 0.8469,
-      "step": 35968
-    },
-    {
-      "epoch": 0.9878608112932905,
-      "grad_norm": 0.0,
-      "learning_rate": 7.72798266187147e-09,
-      "loss": 0.7051,
-      "step": 35969
-    },
-    {
-      "epoch": 0.9878882755211337,
-      "grad_norm": 0.0,
-      "learning_rate": 7.693058450538582e-09,
-      "loss": 0.7086,
-      "step": 35970
-    },
-    {
-      "epoch": 0.987915739748977,
-      "grad_norm": 0.0,
-      "learning_rate": 7.658213302209616e-09,
-      "loss": 0.8314,
-      "step": 35971
-    },
-    {
-      "epoch": 0.9879432039768202,
-      "grad_norm": 0.0,
-      "learning_rate": 7.623447217158797e-09,
-      "loss": 0.9042,
-      "step": 35972
-    },
-    {
-      "epoch": 0.9879706682046634,
-      "grad_norm": 0.0,
-      "learning_rate": 7.58876019566146e-09,
-      "loss": 0.7815,
-      "step": 35973
-    },
-    {
-      "epoch": 0.9879981324325067,
-      "grad_norm": 0.0,
-      "learning_rate": 7.55415223799294e-09,
-      "loss": 0.7694,
-      "step": 35974
-    },
-    {
-      "epoch": 0.98802559666035,
-      "grad_norm": 0.0,
-      "learning_rate": 7.519623344427463e-09,
-      "loss": 0.8451,
-      "step": 35975
-    },
-    {
-      "epoch": 0.9880530608881931,
-      "grad_norm": 0.0,
-      "learning_rate": 7.485173515235922e-09,
-      "loss": 0.8183,
-      "step": 35976
-    },
-    {
-      "epoch": 0.9880805251160364,
-      "grad_norm": 0.0,
-      "learning_rate": 7.450802750693653e-09,
-      "loss": 0.7437,
-      "step": 35977
-    },
-    {
-      "epoch": 0.9881079893438796,
-      "grad_norm": 0.0,
-      "learning_rate": 7.416511051071551e-09,
-      "loss": 0.8013,
-      "step": 35978
-    },
-    {
-      "epoch": 0.9881354535717228,
-      "grad_norm": 0.0,
-      "learning_rate": 7.382298416639399e-09,
-      "loss": 0.8018,
-      "step": 35979
-    },
-    {
-      "epoch": 0.9881629177995661,
-      "grad_norm": 0.0,
-      "learning_rate": 7.348164847670314e-09,
-      "loss": 0.8396,
-      "step": 35980
-    },
-    {
-      "epoch": 0.9881903820274093,
-      "grad_norm": 0.0,
-      "learning_rate": 7.3141103444329676e-09,
-      "loss": 0.7922,
-      "step": 35981
-    },
-    {
-      "epoch": 0.9882178462552526,
-      "grad_norm": 0.0,
-      "learning_rate": 7.280134907197145e-09,
-      "loss": 0.8636,
-      "step": 35982
-    },
-    {
-      "epoch": 0.9882453104830957,
-      "grad_norm": 0.0,
-      "learning_rate": 7.24623853623263e-09,
-      "loss": 0.8328,
-      "step": 35983
-    },
-    {
-      "epoch": 0.988272774710939,
-      "grad_norm": 0.0,
-      "learning_rate": 7.212421231805878e-09,
-      "loss": 0.7352,
-      "step": 35984
-    },
-    {
-      "epoch": 0.9883002389387823,
-      "grad_norm": 0.0,
-      "learning_rate": 7.17868299418556e-09,
-      "loss": 0.8316,
-      "step": 35985
-    },
-    {
-      "epoch": 0.9883277031666254,
-      "grad_norm": 0.0,
-      "learning_rate": 7.145023823638131e-09,
-      "loss": 0.826,
-      "step": 35986
-    },
-    {
-      "epoch": 0.9883551673944687,
-      "grad_norm": 0.0,
-      "learning_rate": 7.111443720430044e-09,
-      "loss": 0.8184,
-      "step": 35987
-    },
-    {
-      "epoch": 0.988382631622312,
-      "grad_norm": 0.0,
-      "learning_rate": 7.077942684826644e-09,
-      "loss": 0.8549,
-      "step": 35988
-    },
-    {
-      "epoch": 0.9884100958501552,
-      "grad_norm": 0.0,
-      "learning_rate": 7.044520717094383e-09,
-      "loss": 0.7457,
-      "step": 35989
-    },
-    {
-      "epoch": 0.9884375600779984,
-      "grad_norm": 0.0,
-      "learning_rate": 7.011177817496384e-09,
-      "loss": 0.8264,
-      "step": 35990
-    },
-    {
-      "epoch": 0.9884650243058416,
-      "grad_norm": 0.0,
-      "learning_rate": 6.97791398629688e-09,
-      "loss": 0.8286,
-      "step": 35991
-    },
-    {
-      "epoch": 0.9884924885336849,
-      "grad_norm": 0.0,
-      "learning_rate": 6.944729223758995e-09,
-      "loss": 0.8582,
-      "step": 35992
-    },
-    {
-      "epoch": 0.9885199527615282,
-      "grad_norm": 0.0,
-      "learning_rate": 6.91162353014474e-09,
-      "loss": 0.6945,
-      "step": 35993
-    },
-    {
-      "epoch": 0.9885474169893713,
-      "grad_norm": 0.0,
-      "learning_rate": 6.878596905717239e-09,
-      "loss": 0.8754,
-      "step": 35994
-    },
-    {
-      "epoch": 0.9885748812172146,
-      "grad_norm": 0.0,
-      "learning_rate": 6.845649350737393e-09,
-      "loss": 0.7625,
-      "step": 35995
-    },
-    {
-      "epoch": 0.9886023454450578,
-      "grad_norm": 0.0,
-      "learning_rate": 6.812780865466107e-09,
-      "loss": 0.7202,
-      "step": 35996
-    },
-    {
-      "epoch": 0.988629809672901,
-      "grad_norm": 0.0,
-      "learning_rate": 6.77999145016206e-09,
-      "loss": 0.8059,
-      "step": 35997
-    },
-    {
-      "epoch": 0.9886572739007443,
-      "grad_norm": 0.0,
-      "learning_rate": 6.747281105087267e-09,
-      "loss": 0.7527,
-      "step": 35998
-    },
-    {
-      "epoch": 0.9886847381285875,
-      "grad_norm": 0.0,
-      "learning_rate": 6.714649830498187e-09,
-      "loss": 0.6805,
-      "step": 35999
-    },
-    {
-      "epoch": 0.9887122023564308,
-      "grad_norm": 0.0,
-      "learning_rate": 6.682097626653505e-09,
-      "loss": 0.7436,
-      "step": 36000
-    },
-    {
-      "epoch": 0.988739666584274,
-      "grad_norm": 0.0,
-      "learning_rate": 6.649624493810791e-09,
-      "loss": 0.8126,
-      "step": 36001
-    },
-    {
-      "epoch": 0.9887671308121172,
-      "grad_norm": 0.0,
-      "learning_rate": 6.617230432227617e-09,
-      "loss": 0.7854,
-      "step": 36002
-    },
-    {
-      "epoch": 0.9887945950399605,
-      "grad_norm": 0.0,
-      "learning_rate": 6.584915442160445e-09,
-      "loss": 0.7859,
-      "step": 36003
-    },
-    {
-      "epoch": 0.9888220592678036,
-      "grad_norm": 0.0,
-      "learning_rate": 6.552679523864624e-09,
-      "loss": 0.8313,
-      "step": 36004
-    },
-    {
-      "epoch": 0.9888495234956469,
-      "grad_norm": 0.0,
-      "learning_rate": 6.520522677594399e-09,
-      "loss": 0.8524,
-      "step": 36005
-    },
-    {
-      "epoch": 0.9888769877234902,
-      "grad_norm": 0.0,
-      "learning_rate": 6.488444903605118e-09,
-      "loss": 0.8269,
-      "step": 36006
-    },
-    {
-      "epoch": 0.9889044519513334,
-      "grad_norm": 0.0,
-      "learning_rate": 6.456446202151024e-09,
-      "loss": 0.863,
-      "step": 36007
-    },
-    {
-      "epoch": 0.9889319161791766,
-      "grad_norm": 0.0,
-      "learning_rate": 6.424526573484136e-09,
-      "loss": 0.7714,
-      "step": 36008
-    },
-    {
-      "epoch": 0.9889593804070198,
-      "grad_norm": 0.0,
-      "learning_rate": 6.392686017856475e-09,
-      "loss": 0.7498,
-      "step": 36009
-    },
-    {
-      "epoch": 0.9889868446348631,
-      "grad_norm": 0.0,
-      "learning_rate": 6.3609245355222835e-09,
-      "loss": 0.8346,
-      "step": 36010
-    },
-    {
-      "epoch": 0.9890143088627064,
-      "grad_norm": 0.0,
-      "learning_rate": 6.329242126731361e-09,
-      "loss": 0.7929,
-      "step": 36011
-    },
-    {
-      "epoch": 0.9890417730905495,
-      "grad_norm": 0.0,
-      "learning_rate": 6.297638791734617e-09,
-      "loss": 0.7993,
-      "step": 36012
-    },
-    {
-      "epoch": 0.9890692373183928,
-      "grad_norm": 0.0,
-      "learning_rate": 6.266114530782963e-09,
-      "loss": 0.7015,
-      "step": 36013
-    },
-    {
-      "epoch": 0.9890967015462361,
-      "grad_norm": 0.0,
-      "learning_rate": 6.234669344123978e-09,
-      "loss": 0.8129,
-      "step": 36014
-    },
-    {
-      "epoch": 0.9891241657740792,
-      "grad_norm": 0.0,
-      "learning_rate": 6.203303232007463e-09,
-      "loss": 0.8351,
-      "step": 36015
-    },
-    {
-      "epoch": 0.9891516300019225,
-      "grad_norm": 0.0,
-      "learning_rate": 6.172016194682106e-09,
-      "loss": 0.71,
-      "step": 36016
-    },
-    {
-      "epoch": 0.9891790942297657,
-      "grad_norm": 0.0,
-      "learning_rate": 6.14080823239438e-09,
-      "loss": 0.7923,
-      "step": 36017
-    },
-    {
-      "epoch": 0.989206558457609,
-      "grad_norm": 0.0,
-      "learning_rate": 6.109679345392971e-09,
-      "loss": 0.8713,
-      "step": 36018
-    },
-    {
-      "epoch": 0.9892340226854522,
-      "grad_norm": 0.0,
-      "learning_rate": 6.078629533922131e-09,
-      "loss": 0.7804,
-      "step": 36019
-    },
-    {
-      "epoch": 0.9892614869132954,
-      "grad_norm": 0.0,
-      "learning_rate": 6.047658798228328e-09,
-      "loss": 0.7325,
-      "step": 36020
-    },
-    {
-      "epoch": 0.9892889511411387,
-      "grad_norm": 0.0,
-      "learning_rate": 6.016767138556923e-09,
-      "loss": 0.8123,
-      "step": 36021
-    },
-    {
-      "epoch": 0.9893164153689818,
-      "grad_norm": 0.0,
-      "learning_rate": 5.9859545551532726e-09,
-      "loss": 0.7727,
-      "step": 36022
-    },
-    {
-      "epoch": 0.9893438795968251,
-      "grad_norm": 0.0,
-      "learning_rate": 5.955221048258297e-09,
-      "loss": 0.8028,
-      "step": 36023
-    },
-    {
-      "epoch": 0.9893713438246684,
-      "grad_norm": 0.0,
-      "learning_rate": 5.924566618118466e-09,
-      "loss": 0.8359,
-      "step": 36024
-    },
-    {
-      "epoch": 0.9893988080525116,
-      "grad_norm": 0.0,
-      "learning_rate": 5.8939912649735865e-09,
-      "loss": 0.8925,
-      "step": 36025
-    },
-    {
-      "epoch": 0.9894262722803548,
-      "grad_norm": 0.0,
-      "learning_rate": 5.863494989067908e-09,
-      "loss": 0.8123,
-      "step": 36026
-    },
-    {
-      "epoch": 0.9894537365081981,
-      "grad_norm": 0.0,
-      "learning_rate": 5.83307779064013e-09,
-      "loss": 0.8245,
-      "step": 36027
-    },
-    {
-      "epoch": 0.9894812007360413,
-      "grad_norm": 0.0,
-      "learning_rate": 5.802739669933388e-09,
-      "loss": 0.816,
-      "step": 36028
-    },
-    {
-      "epoch": 0.9895086649638846,
-      "grad_norm": 0.0,
-      "learning_rate": 5.772480627187493e-09,
-      "loss": 0.8172,
-      "step": 36029
-    },
-    {
-      "epoch": 0.9895361291917277,
-      "grad_norm": 0.0,
-      "learning_rate": 5.742300662640032e-09,
-      "loss": 0.7831,
-      "step": 36030
-    },
-    {
-      "epoch": 0.989563593419571,
-      "grad_norm": 0.0,
-      "learning_rate": 5.712199776530813e-09,
-      "loss": 0.8854,
-      "step": 36031
-    },
-    {
-      "epoch": 0.9895910576474143,
-      "grad_norm": 0.0,
-      "learning_rate": 5.682177969098535e-09,
-      "loss": 0.8438,
-      "step": 36032
-    },
-    {
-      "epoch": 0.9896185218752575,
-      "grad_norm": 0.0,
-      "learning_rate": 5.652235240580783e-09,
-      "loss": 0.8399,
-      "step": 36033
-    },
-    {
-      "epoch": 0.9896459861031007,
-      "grad_norm": 0.0,
-      "learning_rate": 5.622371591212927e-09,
-      "loss": 0.8961,
-      "step": 36034
-    },
-    {
-      "epoch": 0.9896734503309439,
-      "grad_norm": 0.0,
-      "learning_rate": 5.592587021232554e-09,
-      "loss": 0.8007,
-      "step": 36035
-    },
-    {
-      "epoch": 0.9897009145587872,
-      "grad_norm": 0.0,
-      "learning_rate": 5.56288153087503e-09,
-      "loss": 0.7571,
-      "step": 36036
-    },
-    {
-      "epoch": 0.9897283787866304,
-      "grad_norm": 0.0,
-      "learning_rate": 5.533255120375724e-09,
-      "loss": 0.7302,
-      "step": 36037
-    },
-    {
-      "epoch": 0.9897558430144736,
-      "grad_norm": 0.0,
-      "learning_rate": 5.503707789968893e-09,
-      "loss": 0.8093,
-      "step": 36038
-    },
-    {
-      "epoch": 0.9897833072423169,
-      "grad_norm": 0.0,
-      "learning_rate": 5.4742395398887925e-09,
-      "loss": 0.7644,
-      "step": 36039
-    },
-    {
-      "epoch": 0.9898107714701602,
-      "grad_norm": 0.0,
-      "learning_rate": 5.44485037036746e-09,
-      "loss": 0.7812,
-      "step": 36040
-    },
-    {
-      "epoch": 0.9898382356980033,
-      "grad_norm": 0.0,
-      "learning_rate": 5.415540281636933e-09,
-      "loss": 0.8318,
-      "step": 36041
-    },
-    {
-      "epoch": 0.9898656999258466,
-      "grad_norm": 0.0,
-      "learning_rate": 5.386309273931467e-09,
-      "loss": 0.7904,
-      "step": 36042
-    },
-    {
-      "epoch": 0.9898931641536898,
-      "grad_norm": 0.0,
-      "learning_rate": 5.3571573474808795e-09,
-      "loss": 0.7821,
-      "step": 36043
-    },
-    {
-      "epoch": 0.989920628381533,
-      "grad_norm": 0.0,
-      "learning_rate": 5.328084502516095e-09,
-      "loss": 0.7659,
-      "step": 36044
-    },
-    {
-      "epoch": 0.9899480926093763,
-      "grad_norm": 0.0,
-      "learning_rate": 5.299090739266932e-09,
-      "loss": 0.6693,
-      "step": 36045
-    },
-    {
-      "epoch": 0.9899755568372195,
-      "grad_norm": 0.0,
-      "learning_rate": 5.270176057962095e-09,
-      "loss": 0.8919,
-      "step": 36046
-    },
-    {
-      "epoch": 0.9900030210650628,
-      "grad_norm": 0.0,
-      "learning_rate": 5.241340458832511e-09,
-      "loss": 0.7407,
-      "step": 36047
-    },
-    {
-      "epoch": 0.9900304852929059,
-      "grad_norm": 0.0,
-      "learning_rate": 5.212583942104665e-09,
-      "loss": 0.8168,
-      "step": 36048
-    },
-    {
-      "epoch": 0.9900579495207492,
-      "grad_norm": 0.0,
-      "learning_rate": 5.183906508006153e-09,
-      "loss": 0.7388,
-      "step": 36049
-    },
-    {
-      "epoch": 0.9900854137485925,
-      "grad_norm": 0.0,
-      "learning_rate": 5.155308156764572e-09,
-      "loss": 0.7975,
-      "step": 36050
-    },
-    {
-      "epoch": 0.9901128779764357,
-      "grad_norm": 0.0,
-      "learning_rate": 5.126788888605294e-09,
-      "loss": 0.8581,
-      "step": 36051
-    },
-    {
-      "epoch": 0.9901403422042789,
-      "grad_norm": 0.0,
-      "learning_rate": 5.098348703754808e-09,
-      "loss": 0.8133,
-      "step": 36052
-    },
-    {
-      "epoch": 0.9901678064321222,
-      "grad_norm": 0.0,
-      "learning_rate": 5.0699876024373764e-09,
-      "loss": 0.8178,
-      "step": 36053
-    },
-    {
-      "epoch": 0.9901952706599654,
-      "grad_norm": 0.0,
-      "learning_rate": 5.041705584879486e-09,
-      "loss": 0.7815,
-      "step": 36054
-    },
-    {
-      "epoch": 0.9902227348878087,
-      "grad_norm": 0.0,
-      "learning_rate": 5.013502651302071e-09,
-      "loss": 0.7006,
-      "step": 36055
-    },
-    {
-      "epoch": 0.9902501991156518,
-      "grad_norm": 0.0,
-      "learning_rate": 4.985378801929397e-09,
-      "loss": 0.7692,
-      "step": 36056
-    },
-    {
-      "epoch": 0.9902776633434951,
-      "grad_norm": 0.0,
-      "learning_rate": 4.957334036984618e-09,
-      "loss": 0.8786,
-      "step": 36057
-    },
-    {
-      "epoch": 0.9903051275713384,
-      "grad_norm": 0.0,
-      "learning_rate": 4.929368356688668e-09,
-      "loss": 0.752,
-      "step": 36058
-    },
-    {
-      "epoch": 0.9903325917991815,
-      "grad_norm": 0.0,
-      "learning_rate": 4.9014817612624835e-09,
-      "loss": 0.873,
-      "step": 36059
-    },
-    {
-      "epoch": 0.9903600560270248,
-      "grad_norm": 0.0,
-      "learning_rate": 4.8736742509292165e-09,
-      "loss": 0.9536,
-      "step": 36060
-    },
-    {
-      "epoch": 0.990387520254868,
-      "grad_norm": 0.0,
-      "learning_rate": 4.845945825906473e-09,
-      "loss": 0.849,
-      "step": 36061
-    },
-    {
-      "epoch": 0.9904149844827113,
-      "grad_norm": 0.0,
-      "learning_rate": 4.818296486414076e-09,
-      "loss": 0.8328,
-      "step": 36062
-    },
-    {
-      "epoch": 0.9904424487105545,
-      "grad_norm": 0.0,
-      "learning_rate": 4.790726232670739e-09,
-      "loss": 0.8205,
-      "step": 36063
-    },
-    {
-      "epoch": 0.9904699129383977,
-      "grad_norm": 0.0,
-      "learning_rate": 4.763235064896288e-09,
-      "loss": 0.7881,
-      "step": 36064
-    },
-    {
-      "epoch": 0.990497377166241,
-      "grad_norm": 0.0,
-      "learning_rate": 4.735822983306104e-09,
-      "loss": 0.828,
-      "step": 36065
-    },
-    {
-      "epoch": 0.9905248413940843,
-      "grad_norm": 0.0,
-      "learning_rate": 4.708489988117793e-09,
-      "loss": 0.7641,
-      "step": 36066
-    },
-    {
-      "epoch": 0.9905523056219274,
-      "grad_norm": 0.0,
-      "learning_rate": 4.681236079547846e-09,
-      "loss": 0.7526,
-      "step": 36067
-    },
-    {
-      "epoch": 0.9905797698497707,
-      "grad_norm": 0.0,
-      "learning_rate": 4.654061257811648e-09,
-      "loss": 0.7746,
-      "step": 36068
-    },
-    {
-      "epoch": 0.9906072340776139,
-      "grad_norm": 0.0,
-      "learning_rate": 4.626965523124582e-09,
-      "loss": 0.7814,
-      "step": 36069
-    },
-    {
-      "epoch": 0.9906346983054571,
-      "grad_norm": 0.0,
-      "learning_rate": 4.599948875699811e-09,
-      "loss": 0.8253,
-      "step": 36070
-    },
-    {
-      "epoch": 0.9906621625333004,
-      "grad_norm": 0.0,
-      "learning_rate": 4.573011315753828e-09,
-      "loss": 0.7844,
-      "step": 36071
-    },
-    {
-      "epoch": 0.9906896267611436,
-      "grad_norm": 0.0,
-      "learning_rate": 4.546152843496465e-09,
-      "loss": 0.8575,
-      "step": 36072
-    },
-    {
-      "epoch": 0.9907170909889869,
-      "grad_norm": 0.0,
-      "learning_rate": 4.519373459143106e-09,
-      "loss": 0.7958,
-      "step": 36073
-    },
-    {
-      "epoch": 0.99074455521683,
-      "grad_norm": 0.0,
-      "learning_rate": 4.492673162903583e-09,
-      "loss": 0.7968,
-      "step": 36074
-    },
-    {
-      "epoch": 0.9907720194446733,
-      "grad_norm": 0.0,
-      "learning_rate": 4.466051954989947e-09,
-      "loss": 0.6919,
-      "step": 36075
-    },
-    {
-      "epoch": 0.9907994836725166,
-      "grad_norm": 0.0,
-      "learning_rate": 4.4395098356131426e-09,
-      "loss": 0.7686,
-      "step": 36076
-    },
-    {
-      "epoch": 0.9908269479003597,
-      "grad_norm": 0.0,
-      "learning_rate": 4.413046804981891e-09,
-      "loss": 0.7401,
-      "step": 36077
-    },
-    {
-      "epoch": 0.990854412128203,
-      "grad_norm": 0.0,
-      "learning_rate": 4.386662863308244e-09,
-      "loss": 0.7784,
-      "step": 36078
-    },
-    {
-      "epoch": 0.9908818763560463,
-      "grad_norm": 0.0,
-      "learning_rate": 4.360358010797594e-09,
-      "loss": 0.7412,
-      "step": 36079
-    },
-    {
-      "epoch": 0.9909093405838895,
-      "grad_norm": 0.0,
-      "learning_rate": 4.334132247660883e-09,
-      "loss": 0.879,
-      "step": 36080
-    },
-    {
-      "epoch": 0.9909368048117327,
-      "grad_norm": 0.0,
-      "learning_rate": 4.307985574103501e-09,
-      "loss": 0.7076,
-      "step": 36081
-    },
-    {
-      "epoch": 0.9909642690395759,
-      "grad_norm": 0.0,
-      "learning_rate": 4.281917990334172e-09,
-      "loss": 0.7006,
-      "step": 36082
-    },
-    {
-      "epoch": 0.9909917332674192,
-      "grad_norm": 0.0,
-      "learning_rate": 4.255929496558287e-09,
-      "loss": 0.6937,
-      "step": 36083
-    },
-    {
-      "epoch": 0.9910191974952625,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2300200929812355e-09,
-      "loss": 0.8052,
-      "step": 36084
-    },
-    {
-      "epoch": 0.9910466617231056,
-      "grad_norm": 0.0,
-      "learning_rate": 4.2041897798072996e-09,
-      "loss": 0.8,
-      "step": 36085
-    },
-    {
-      "epoch": 0.9910741259509489,
-      "grad_norm": 0.0,
-      "learning_rate": 4.178438557242981e-09,
-      "loss": 0.8303,
-      "step": 36086
-    },
-    {
-      "epoch": 0.9911015901787921,
-      "grad_norm": 0.0,
-      "learning_rate": 4.15276642549034e-09,
-      "loss": 0.8118,
-      "step": 36087
-    },
-    {
-      "epoch": 0.9911290544066353,
-      "grad_norm": 0.0,
-      "learning_rate": 4.127173384753658e-09,
-      "loss": 0.7666,
-      "step": 36088
-    },
-    {
-      "epoch": 0.9911565186344786,
-      "grad_norm": 0.0,
-      "learning_rate": 4.101659435233885e-09,
-      "loss": 0.7072,
-      "step": 36089
-    },
-    {
-      "epoch": 0.9911839828623218,
-      "grad_norm": 0.0,
-      "learning_rate": 4.076224577134191e-09,
-      "loss": 0.7928,
-      "step": 36090
-    },
-    {
-      "epoch": 0.9912114470901651,
-      "grad_norm": 0.0,
-      "learning_rate": 4.050868810655529e-09,
-      "loss": 0.7352,
-      "step": 36091
-    },
-    {
-      "epoch": 0.9912389113180083,
-      "grad_norm": 0.0,
-      "learning_rate": 4.025592135998846e-09,
-      "loss": 0.7947,
-      "step": 36092
-    },
-    {
-      "epoch": 0.9912663755458515,
-      "grad_norm": 0.0,
-      "learning_rate": 4.000394553362874e-09,
-      "loss": 0.8775,
-      "step": 36093
-    },
-    {
-      "epoch": 0.9912938397736948,
-      "grad_norm": 0.0,
-      "learning_rate": 3.975276062948563e-09,
-      "loss": 0.7996,
-      "step": 36094
-    },
-    {
-      "epoch": 0.991321304001538,
-      "grad_norm": 0.0,
-      "learning_rate": 3.950236664954643e-09,
-      "loss": 0.7938,
-      "step": 36095
-    },
-    {
-      "epoch": 0.9913487682293812,
-      "grad_norm": 0.0,
-      "learning_rate": 3.925276359577623e-09,
-      "loss": 0.7804,
-      "step": 36096
-    },
-    {
-      "epoch": 0.9913762324572245,
-      "grad_norm": 0.0,
-      "learning_rate": 3.900395147015123e-09,
-      "loss": 0.8795,
-      "step": 36097
-    },
-    {
-      "epoch": 0.9914036966850677,
-      "grad_norm": 0.0,
-      "learning_rate": 3.875593027465873e-09,
-      "loss": 0.7882,
-      "step": 36098
-    },
-    {
-      "epoch": 0.991431160912911,
-      "grad_norm": 0.0,
-      "learning_rate": 3.850870001125273e-09,
-      "loss": 0.7835,
-      "step": 36099
-    },
-    {
-      "epoch": 0.9914586251407541,
-      "grad_norm": 0.0,
-      "learning_rate": 3.8262260681876105e-09,
-      "loss": 0.8199,
-      "step": 36100
-    },
-    {
-      "epoch": 0.9914860893685974,
-      "grad_norm": 0.0,
-      "learning_rate": 3.801661228850506e-09,
-      "loss": 0.7908,
-      "step": 36101
-    },
-    {
-      "epoch": 0.9915135535964407,
-      "grad_norm": 0.0,
-      "learning_rate": 3.777175483306028e-09,
-      "loss": 0.7991,
-      "step": 36102
-    },
-    {
-      "epoch": 0.9915410178242838,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7527688317484654e-09,
-      "loss": 0.8667,
-      "step": 36103
-    },
-    {
-      "epoch": 0.9915684820521271,
-      "grad_norm": 0.0,
-      "learning_rate": 3.728441274370998e-09,
-      "loss": 0.7971,
-      "step": 36104
-    },
-    {
-      "epoch": 0.9915959462799704,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7041928113668023e-09,
-      "loss": 0.8141,
-      "step": 36105
-    },
-    {
-      "epoch": 0.9916234105078136,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6800234429268387e-09,
-      "loss": 0.7727,
-      "step": 36106
-    },
-    {
-      "epoch": 0.9916508747356568,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6559331692431753e-09,
-      "loss": 0.8009,
-      "step": 36107
-    },
-    {
-      "epoch": 0.9916783389635,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6319219905067705e-09,
-      "loss": 0.7074,
-      "step": 36108
-    },
-    {
-      "epoch": 0.9917058031913433,
-      "grad_norm": 0.0,
-      "learning_rate": 3.6079899069052517e-09,
-      "loss": 0.7706,
-      "step": 36109
-    },
-    {
-      "epoch": 0.9917332674191865,
-      "grad_norm": 0.0,
-      "learning_rate": 3.584136918630687e-09,
-      "loss": 0.7529,
-      "step": 36110
-    },
-    {
-      "epoch": 0.9917607316470297,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5603630258695956e-09,
-      "loss": 0.7724,
-      "step": 36111
-    },
-    {
-      "epoch": 0.991788195874873,
-      "grad_norm": 0.0,
-      "learning_rate": 3.536668228811824e-09,
-      "loss": 0.8035,
-      "step": 36112
-    },
-    {
-      "epoch": 0.9918156601027162,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5130525276438896e-09,
-      "loss": 0.7643,
-      "step": 36113
-    },
-    {
-      "epoch": 0.9918431243305594,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4895159225534215e-09,
-      "loss": 0.7874,
-      "step": 36114
-    },
-    {
-      "epoch": 0.9918705885584027,
-      "grad_norm": 0.0,
-      "learning_rate": 3.466058413725826e-09,
-      "loss": 0.7324,
-      "step": 36115
-    },
-    {
-      "epoch": 0.9918980527862459,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4426800013476203e-09,
-      "loss": 0.7493,
-      "step": 36116
-    },
-    {
-      "epoch": 0.9919255170140892,
-      "grad_norm": 0.0,
-      "learning_rate": 3.4193806856019917e-09,
-      "loss": 0.7789,
-      "step": 36117
-    },
-    {
-      "epoch": 0.9919529812419324,
-      "grad_norm": 0.0,
-      "learning_rate": 3.396160466675458e-09,
-      "loss": 0.7861,
-      "step": 36118
-    },
-    {
-      "epoch": 0.9919804454697756,
-      "grad_norm": 0.0,
-      "learning_rate": 3.373019344751205e-09,
-      "loss": 0.8781,
-      "step": 36119
-    },
-    {
-      "epoch": 0.9920079096976189,
-      "grad_norm": 0.0,
-      "learning_rate": 3.34995732001131e-09,
-      "loss": 0.8355,
-      "step": 36120
-    },
-    {
-      "epoch": 0.992035373925462,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3269743926389597e-09,
-      "loss": 0.8405,
-      "step": 36121
-    },
-    {
-      "epoch": 0.9920628381533053,
-      "grad_norm": 0.0,
-      "learning_rate": 3.3040705628151204e-09,
-      "loss": 0.7488,
-      "step": 36122
-    },
-    {
-      "epoch": 0.9920903023811486,
-      "grad_norm": 0.0,
-      "learning_rate": 3.281245830722979e-09,
-      "loss": 0.8109,
-      "step": 36123
-    },
-    {
-      "epoch": 0.9921177666089918,
-      "grad_norm": 0.0,
-      "learning_rate": 3.258500196541281e-09,
-      "loss": 0.7133,
-      "step": 36124
-    },
-    {
-      "epoch": 0.992145230836835,
-      "grad_norm": 0.0,
-      "learning_rate": 3.235833660450993e-09,
-      "loss": 0.783,
-      "step": 36125
-    },
-    {
-      "epoch": 0.9921726950646783,
-      "grad_norm": 0.0,
-      "learning_rate": 3.2132462226319716e-09,
-      "loss": 0.9088,
-      "step": 36126
-    },
-    {
-      "epoch": 0.9922001592925215,
-      "grad_norm": 0.0,
-      "learning_rate": 3.1907378832607416e-09,
-      "loss": 0.8183,
-      "step": 36127
-    },
-    {
-      "epoch": 0.9922276235203648,
-      "grad_norm": 0.0,
-      "learning_rate": 3.16830864251827e-09,
-      "loss": 0.7487,
-      "step": 36128
-    },
-    {
-      "epoch": 0.9922550877482079,
-      "grad_norm": 0.0,
-      "learning_rate": 3.1459585005788607e-09,
-      "loss": 0.8885,
-      "step": 36129
-    },
-    {
-      "epoch": 0.9922825519760512,
-      "grad_norm": 0.0,
-      "learning_rate": 3.123687457622371e-09,
-      "loss": 0.8834,
-      "step": 36130
-    },
-    {
-      "epoch": 0.9923100162038945,
-      "grad_norm": 0.0,
-      "learning_rate": 3.1014955138231053e-09,
-      "loss": 0.7673,
-      "step": 36131
-    },
-    {
-      "epoch": 0.9923374804317376,
-      "grad_norm": 0.0,
-      "learning_rate": 3.079382669357589e-09,
-      "loss": 0.7687,
-      "step": 36132
-    },
-    {
-      "epoch": 0.9923649446595809,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0573489244001274e-09,
-      "loss": 0.8347,
-      "step": 36133
-    },
-    {
-      "epoch": 0.9923924088874241,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0353942791250256e-09,
-      "loss": 0.9409,
-      "step": 36134
-    },
-    {
-      "epoch": 0.9924198731152674,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0135187337065885e-09,
-      "loss": 0.8865,
-      "step": 36135
-    },
-    {
-      "epoch": 0.9924473373431106,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9917222883180107e-09,
-      "loss": 0.8657,
-      "step": 36136
-    },
-    {
-      "epoch": 0.9924748015709538,
-      "grad_norm": 0.0,
-      "learning_rate": 2.970004943131377e-09,
-      "loss": 0.8369,
-      "step": 36137
-    },
-    {
-      "epoch": 0.9925022657987971,
-      "grad_norm": 0.0,
-      "learning_rate": 2.948366698317662e-09,
-      "loss": 0.7555,
-      "step": 36138
-    },
-    {
-      "epoch": 0.9925297300266404,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9268075540500597e-09,
-      "loss": 0.8001,
-      "step": 36139
-    },
-    {
-      "epoch": 0.9925571942544835,
-      "grad_norm": 0.0,
-      "learning_rate": 2.9053275104973244e-09,
-      "loss": 0.7344,
-      "step": 36140
-    },
-    {
-      "epoch": 0.9925846584823268,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8839265678304308e-09,
-      "loss": 0.8479,
-      "step": 36141
-    },
-    {
-      "epoch": 0.99261212271017,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8626047262192426e-09,
-      "loss": 0.8201,
-      "step": 36142
-    },
-    {
-      "epoch": 0.9926395869380132,
-      "grad_norm": 0.0,
-      "learning_rate": 2.841361985830293e-09,
-      "loss": 0.8353,
-      "step": 36143
-    },
-    {
-      "epoch": 0.9926670511658565,
-      "grad_norm": 0.0,
-      "learning_rate": 2.820198346833447e-09,
-      "loss": 0.7523,
-      "step": 36144
-    },
-    {
-      "epoch": 0.9926945153936997,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7991138093963477e-09,
-      "loss": 0.8193,
-      "step": 36145
-    },
-    {
-      "epoch": 0.992721979621543,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7781083736844184e-09,
-      "loss": 0.8211,
-      "step": 36146
-    },
-    {
-      "epoch": 0.9927494438493861,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7571820398653025e-09,
-      "loss": 0.6685,
-      "step": 36147
-    },
-    {
-      "epoch": 0.9927769080772294,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7363348081033137e-09,
-      "loss": 0.8214,
-      "step": 36148
-    },
-    {
-      "epoch": 0.9928043723050727,
-      "grad_norm": 0.0,
-      "learning_rate": 2.7155666785638747e-09,
-      "loss": 0.7857,
-      "step": 36149
-    },
-    {
-      "epoch": 0.9928318365329158,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6948776514124086e-09,
-      "loss": 0.7722,
-      "step": 36150
-    },
-    {
-      "epoch": 0.9928593007607591,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6742677268110084e-09,
-      "loss": 0.765,
-      "step": 36151
-    },
-    {
-      "epoch": 0.9928867649886024,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6537369049228767e-09,
-      "loss": 0.7655,
-      "step": 36152
-    },
-    {
-      "epoch": 0.9929142292164456,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6332851859123265e-09,
-      "loss": 0.8275,
-      "step": 36153
-    },
-    {
-      "epoch": 0.9929416934442888,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6129125699381198e-09,
-      "loss": 0.8306,
-      "step": 36154
-    },
-    {
-      "epoch": 0.992969157672132,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5926190571645692e-09,
-      "loss": 0.7962,
-      "step": 36155
-    },
-    {
-      "epoch": 0.9929966218999753,
-      "grad_norm": 0.0,
-      "learning_rate": 2.572404647750437e-09,
-      "loss": 0.7887,
-      "step": 36156
-    },
-    {
-      "epoch": 0.9930240861278186,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5522693418555955e-09,
-      "loss": 0.8182,
-      "step": 36157
-    },
-    {
-      "epoch": 0.9930515503556617,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5322131396410266e-09,
-      "loss": 0.8245,
-      "step": 36158
-    },
-    {
-      "epoch": 0.993079014583505,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5122360412632716e-09,
-      "loss": 0.7982,
-      "step": 36159
-    },
-    {
-      "epoch": 0.9931064788113482,
-      "grad_norm": 0.0,
-      "learning_rate": 2.492338046881093e-09,
-      "loss": 0.7728,
-      "step": 36160
-    },
-    {
-      "epoch": 0.9931339430391914,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4725191566532525e-09,
-      "loss": 0.8999,
-      "step": 36161
-    },
-    {
-      "epoch": 0.9931614072670347,
-      "grad_norm": 0.0,
-      "learning_rate": 2.452779370735181e-09,
-      "loss": 0.7978,
-      "step": 36162
-    },
-    {
-      "epoch": 0.9931888714948779,
-      "grad_norm": 0.0,
-      "learning_rate": 2.43311868928231e-09,
-      "loss": 0.8448,
-      "step": 36163
-    },
-    {
-      "epoch": 0.9932163357227212,
-      "grad_norm": 0.0,
-      "learning_rate": 2.413537112452291e-09,
-      "loss": 0.7896,
-      "step": 36164
-    },
-    {
-      "epoch": 0.9932437999505644,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3940346403994454e-09,
-      "loss": 0.8417,
-      "step": 36165
-    },
-    {
-      "epoch": 0.9932712641784076,
-      "grad_norm": 0.0,
-      "learning_rate": 2.374611273276983e-09,
-      "loss": 0.7786,
-      "step": 36166
-    },
-    {
-      "epoch": 0.9932987284062509,
-      "grad_norm": 0.0,
-      "learning_rate": 2.355267011239226e-09,
-      "loss": 0.7546,
-      "step": 36167
-    },
-    {
-      "epoch": 0.993326192634094,
-      "grad_norm": 0.0,
-      "learning_rate": 2.336001854439385e-09,
-      "loss": 0.8121,
-      "step": 36168
-    },
-    {
-      "epoch": 0.9933536568619373,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3168158030306696e-09,
-      "loss": 0.8245,
-      "step": 36169
-    },
-    {
-      "epoch": 0.9933811210897806,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2977088571629614e-09,
-      "loss": 0.8572,
-      "step": 36170
-    },
-    {
-      "epoch": 0.9934085853176238,
-      "grad_norm": 0.0,
-      "learning_rate": 2.27868101698947e-09,
-      "loss": 0.8435,
-      "step": 36171
-    },
-    {
-      "epoch": 0.993436049545467,
-      "grad_norm": 0.0,
-      "learning_rate": 2.259732282658966e-09,
-      "loss": 0.7413,
-      "step": 36172
-    },
-    {
-      "epoch": 0.9934635137733102,
-      "grad_norm": 0.0,
-      "learning_rate": 2.24086265432244e-09,
-      "loss": 0.707,
-      "step": 36173
-    },
-    {
-      "epoch": 0.9934909780011535,
-      "grad_norm": 0.0,
-      "learning_rate": 2.222072132128661e-09,
-      "loss": 0.8551,
-      "step": 36174
-    },
-    {
-      "epoch": 0.9935184422289968,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2033607162275097e-09,
-      "loss": 0.7876,
-      "step": 36175
-    },
-    {
-      "epoch": 0.9935459064568399,
-      "grad_norm": 0.0,
-      "learning_rate": 2.184728406765535e-09,
-      "loss": 0.7665,
-      "step": 36176
-    },
-    {
-      "epoch": 0.9935733706846832,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1661752038903973e-09,
-      "loss": 0.8835,
-      "step": 36177
-    },
-    {
-      "epoch": 0.9936008349125265,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1477011077497558e-09,
-      "loss": 0.8495,
-      "step": 36178
-    },
-    {
-      "epoch": 0.9936282991403697,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1293061184901597e-09,
-      "loss": 0.8472,
-      "step": 36179
-    },
-    {
-      "epoch": 0.9936557633682129,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1109902362548283e-09,
-      "loss": 0.8381,
-      "step": 36180
-    },
-    {
-      "epoch": 0.9936832275960561,
-      "grad_norm": 0.0,
-      "learning_rate": 2.092753461190311e-09,
-      "loss": 0.8915,
-      "step": 36181
-    },
-    {
-      "epoch": 0.9937106918238994,
-      "grad_norm": 0.0,
-      "learning_rate": 2.074595793442047e-09,
-      "loss": 0.8148,
-      "step": 36182
-    },
-    {
-      "epoch": 0.9937381560517426,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0565172331510342e-09,
-      "loss": 0.8502,
-      "step": 36183
-    },
-    {
-      "epoch": 0.9937656202795858,
-      "grad_norm": 0.0,
-      "learning_rate": 2.038517780462712e-09,
-      "loss": 0.8734,
-      "step": 36184
-    },
-    {
-      "epoch": 0.9937930845074291,
-      "grad_norm": 0.0,
-      "learning_rate": 2.020597435518079e-09,
-      "loss": 0.8058,
-      "step": 36185
-    },
-    {
-      "epoch": 0.9938205487352723,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0027561984592436e-09,
-      "loss": 0.7409,
-      "step": 36186
-    },
-    {
-      "epoch": 0.9938480129631155,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9849940694272042e-09,
-      "loss": 0.831,
-      "step": 36187
-    },
-    {
-      "epoch": 0.9938754771909588,
-      "grad_norm": 0.0,
-      "learning_rate": 1.967311048562959e-09,
-      "loss": 0.8722,
-      "step": 36188
-    },
-    {
-      "epoch": 0.993902941418802,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9497071360063955e-09,
-      "loss": 0.8275,
-      "step": 36189
-    },
-    {
-      "epoch": 0.9939304056466453,
-      "grad_norm": 0.0,
-      "learning_rate": 1.932182331896293e-09,
-      "loss": 0.8341,
-      "step": 36190
-    },
-    {
-      "epoch": 0.9939578698744885,
-      "grad_norm": 0.0,
-      "learning_rate": 1.914736636371428e-09,
-      "loss": 0.8506,
-      "step": 36191
-    },
-    {
-      "epoch": 0.9939853341023317,
-      "grad_norm": 0.0,
-      "learning_rate": 1.897370049570579e-09,
-      "loss": 0.7758,
-      "step": 36192
-    },
-    {
-      "epoch": 0.994012798330175,
-      "grad_norm": 0.0,
-      "learning_rate": 1.880082571630304e-09,
-      "loss": 0.8017,
-      "step": 36193
-    },
-    {
-      "epoch": 0.9940402625580181,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8628742026882695e-09,
-      "loss": 0.8699,
-      "step": 36194
-    },
-    {
-      "epoch": 0.9940677267858614,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8457449428799234e-09,
-      "loss": 0.8544,
-      "step": 36195
-    },
-    {
-      "epoch": 0.9940951910137047,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8286947923407127e-09,
-      "loss": 0.8381,
-      "step": 36196
-    },
-    {
-      "epoch": 0.9941226552415479,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8117237512060847e-09,
-      "loss": 0.776,
-      "step": 36197
-    },
-    {
-      "epoch": 0.9941501194693911,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7948318196092663e-09,
-      "loss": 0.9012,
-      "step": 36198
-    },
-    {
-      "epoch": 0.9941775836972343,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7780189976857043e-09,
-      "loss": 0.8446,
-      "step": 36199
-    },
-    {
-      "epoch": 0.9942050479250776,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7612852855664053e-09,
-      "loss": 0.7955,
-      "step": 36200
-    },
-    {
-      "epoch": 0.9942325121529209,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7446306833857062e-09,
-      "loss": 0.8369,
-      "step": 36201
-    },
-    {
-      "epoch": 0.994259976380764,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7280551912735033e-09,
-      "loss": 0.755,
-      "step": 36202
-    },
-    {
-      "epoch": 0.9942874406086073,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7115588093619128e-09,
-      "loss": 0.7419,
-      "step": 36203
-    },
-    {
-      "epoch": 0.9943149048364506,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6951415377819413e-09,
-      "loss": 0.8181,
-      "step": 36204
-    },
-    {
-      "epoch": 0.9943423690642937,
-      "grad_norm": 0.0,
-      "learning_rate": 1.678803376663485e-09,
-      "loss": 0.8315,
-      "step": 36205
-    },
-    {
-      "epoch": 0.994369833292137,
-      "grad_norm": 0.0,
-      "learning_rate": 1.662544326134219e-09,
-      "loss": 0.8492,
-      "step": 36206
-    },
-    {
-      "epoch": 0.9943972975199802,
-      "grad_norm": 0.0,
-      "learning_rate": 1.64636438632515e-09,
-      "loss": 0.8274,
-      "step": 36207
-    },
-    {
-      "epoch": 0.9944247617478235,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6302635573617331e-09,
-      "loss": 0.8257,
-      "step": 36208
-    },
-    {
-      "epoch": 0.9944522259756667,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6142418393738645e-09,
-      "loss": 0.7391,
-      "step": 36209
-    },
-    {
-      "epoch": 0.9944796902035099,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5982992324858893e-09,
-      "loss": 0.7344,
-      "step": 36210
-    },
-    {
-      "epoch": 0.9945071544313532,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5824357368254829e-09,
-      "loss": 0.7893,
-      "step": 36211
-    },
-    {
-      "epoch": 0.9945346186591963,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5666513525181005e-09,
-      "loss": 0.7648,
-      "step": 36212
-    },
-    {
-      "epoch": 0.9945620828870396,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5509460796891972e-09,
-      "loss": 0.8279,
-      "step": 36213
-    },
-    {
-      "epoch": 0.9945895471148829,
-      "grad_norm": 0.0,
-      "learning_rate": 1.535319918460898e-09,
-      "loss": 0.7888,
-      "step": 36214
-    },
-    {
-      "epoch": 0.9946170113427261,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5197728689586578e-09,
-      "loss": 0.8563,
-      "step": 36215
-    },
-    {
-      "epoch": 0.9946444755705693,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5043049313057111e-09,
-      "loss": 0.8543,
-      "step": 36216
-    },
-    {
-      "epoch": 0.9946719397984126,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4889161056230727e-09,
-      "loss": 0.7855,
-      "step": 36217
-    },
-    {
-      "epoch": 0.9946994040262558,
-      "grad_norm": 0.0,
-      "learning_rate": 1.473606392032867e-09,
-      "loss": 0.7988,
-      "step": 36218
-    },
-    {
-      "epoch": 0.9947268682540991,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4583757906572183e-09,
-      "loss": 0.7809,
-      "step": 36219
-    },
-    {
-      "epoch": 0.9947543324819422,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4432243016160307e-09,
-      "loss": 0.8181,
-      "step": 36220
-    },
-    {
-      "epoch": 0.9947817967097855,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4281519250292086e-09,
-      "loss": 0.8279,
-      "step": 36221
-    },
-    {
-      "epoch": 0.9948092609376288,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4131586610155457e-09,
-      "loss": 0.7656,
-      "step": 36222
-    },
-    {
-      "epoch": 0.994836725165472,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3982445096938358e-09,
-      "loss": 0.8431,
-      "step": 36223
-    },
-    {
-      "epoch": 0.9948641893933152,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3834094711817625e-09,
-      "loss": 0.8773,
-      "step": 36224
-    },
-    {
-      "epoch": 0.9948916536211584,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3686535455981198e-09,
-      "loss": 0.8718,
-      "step": 36225
-    },
-    {
-      "epoch": 0.9949191178490017,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3539767330594812e-09,
-      "loss": 0.8008,
-      "step": 36226
-    },
-    {
-      "epoch": 0.9949465820768449,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3393790336801993e-09,
-      "loss": 0.8058,
-      "step": 36227
-    },
-    {
-      "epoch": 0.9949740463046881,
-      "grad_norm": 0.0,
-      "learning_rate": 1.3248604475768479e-09,
-      "loss": 0.9029,
-      "step": 36228
-    },
-    {
-      "epoch": 0.9950015105325314,
-      "grad_norm": 0.0,
-      "learning_rate": 1.31042097486489e-09,
-      "loss": 0.8411,
-      "step": 36229
-    },
-    {
-      "epoch": 0.9950289747603747,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2960606156575684e-09,
-      "loss": 0.7128,
-      "step": 36230
-    },
-    {
-      "epoch": 0.9950564389882178,
-      "grad_norm": 0.0,
-      "learning_rate": 1.281779370069236e-09,
-      "loss": 0.8692,
-      "step": 36231
-    },
-    {
-      "epoch": 0.9950839032160611,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2675772382120255e-09,
-      "loss": 0.8486,
-      "step": 36232
-    },
-    {
-      "epoch": 0.9951113674439043,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2534542201991795e-09,
-      "loss": 0.7262,
-      "step": 36233
-    },
-    {
-      "epoch": 0.9951388316717475,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2394103161428305e-09,
-      "loss": 0.7293,
-      "step": 36234
-    },
-    {
-      "epoch": 0.9951662958995908,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2254455261528908e-09,
-      "loss": 0.7917,
-      "step": 36235
-    },
-    {
-      "epoch": 0.995193760127434,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2115598503403824e-09,
-      "loss": 0.7443,
-      "step": 36236
-    },
-    {
-      "epoch": 0.9952212243552773,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1977532888152177e-09,
-      "loss": 0.7821,
-      "step": 36237
-    },
-    {
-      "epoch": 0.9952486885831204,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1840258416861983e-09,
-      "loss": 0.8777,
-      "step": 36238
-    },
-    {
-      "epoch": 0.9952761528109637,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1703775090621262e-09,
-      "loss": 0.8845,
-      "step": 36239
-    },
-    {
-      "epoch": 0.995303617038807,
-      "grad_norm": 0.0,
-      "learning_rate": 1.156808291051803e-09,
-      "loss": 0.8084,
-      "step": 36240
-    },
-    {
-      "epoch": 0.9953310812666502,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1433181877618105e-09,
-      "loss": 0.9406,
-      "step": 36241
-    },
-    {
-      "epoch": 0.9953585454944934,
-      "grad_norm": 0.0,
-      "learning_rate": 1.12990719929984e-09,
-      "loss": 0.7818,
-      "step": 36242
-    },
-    {
-      "epoch": 0.9953860097223367,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1165753257702526e-09,
-      "loss": 0.774,
-      "step": 36243
-    },
-    {
-      "epoch": 0.9954134739501799,
-      "grad_norm": 0.0,
-      "learning_rate": 1.10332256727963e-09,
-      "loss": 0.7624,
-      "step": 36244
-    },
-    {
-      "epoch": 0.9954409381780231,
-      "grad_norm": 0.0,
-      "learning_rate": 1.090148923933443e-09,
-      "loss": 0.7307,
-      "step": 36245
-    },
-    {
-      "epoch": 0.9954684024058663,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0770543958349422e-09,
-      "loss": 0.7287,
-      "step": 36246
-    },
-    {
-      "epoch": 0.9954958666337096,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0640389830873788e-09,
-      "loss": 0.7681,
-      "step": 36247
-    },
-    {
-      "epoch": 0.9955233308615529,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0511026857951134e-09,
-      "loss": 0.8395,
-      "step": 36248
-    },
-    {
-      "epoch": 0.995550795089396,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0382455040591765e-09,
-      "loss": 0.7336,
-      "step": 36249
-    },
-    {
-      "epoch": 0.9955782593172393,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0254674379817086e-09,
-      "loss": 0.7933,
-      "step": 36250
-    },
-    {
-      "epoch": 0.9956057235450825,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0127684876648502e-09,
-      "loss": 0.7505,
-      "step": 36251
-    },
-    {
-      "epoch": 0.9956331877729258,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0001486532074111e-09,
-      "loss": 0.7556,
-      "step": 36252
-    },
-    {
-      "epoch": 0.995660652000769,
-      "grad_norm": 0.0,
-      "learning_rate": 9.876079347104217e-10,
-      "loss": 0.7421,
-      "step": 36253
-    },
-    {
-      "epoch": 0.9956881162286122,
-      "grad_norm": 0.0,
-      "learning_rate": 9.751463322726916e-10,
-      "loss": 0.8154,
-      "step": 36254
-    },
-    {
-      "epoch": 0.9957155804564555,
-      "grad_norm": 0.0,
-      "learning_rate": 9.627638459919208e-10,
-      "loss": 0.7755,
-      "step": 36255
-    },
-    {
-      "epoch": 0.9957430446842988,
-      "grad_norm": 0.0,
-      "learning_rate": 9.504604759680292e-10,
-      "loss": 0.8125,
-      "step": 36256
-    },
-    {
-      "epoch": 0.9957705089121419,
-      "grad_norm": 0.0,
-      "learning_rate": 9.38236222297606e-10,
-      "loss": 0.7873,
-      "step": 36257
-    },
-    {
-      "epoch": 0.9957979731399852,
-      "grad_norm": 0.0,
-      "learning_rate": 9.260910850761306e-10,
-      "loss": 0.7567,
-      "step": 36258
-    },
-    {
-      "epoch": 0.9958254373678284,
-      "grad_norm": 0.0,
-      "learning_rate": 9.140250644013026e-10,
-      "loss": 0.7363,
-      "step": 36259
-    },
-    {
-      "epoch": 0.9958529015956716,
-      "grad_norm": 0.0,
-      "learning_rate": 9.020381603686013e-10,
-      "loss": 0.7158,
-      "step": 36260
-    },
-    {
-      "epoch": 0.9958803658235149,
-      "grad_norm": 0.0,
-      "learning_rate": 8.901303730712851e-10,
-      "loss": 0.7638,
-      "step": 36261
-    },
-    {
-      "epoch": 0.9959078300513581,
-      "grad_norm": 0.0,
-      "learning_rate": 8.783017026048335e-10,
-      "loss": 0.79,
-      "step": 36262
-    },
-    {
-      "epoch": 0.9959352942792014,
-      "grad_norm": 0.0,
-      "learning_rate": 8.665521490625051e-10,
-      "loss": 0.8991,
-      "step": 36263
-    },
-    {
-      "epoch": 0.9959627585070445,
-      "grad_norm": 0.0,
-      "learning_rate": 8.548817125375586e-10,
-      "loss": 0.8328,
-      "step": 36264
-    },
-    {
-      "epoch": 0.9959902227348878,
-      "grad_norm": 0.0,
-      "learning_rate": 8.432903931221425e-10,
-      "loss": 0.7893,
-      "step": 36265
-    },
-    {
-      "epoch": 0.9960176869627311,
-      "grad_norm": 0.0,
-      "learning_rate": 8.317781909072953e-10,
-      "loss": 0.7126,
-      "step": 36266
-    },
-    {
-      "epoch": 0.9960451511905742,
-      "grad_norm": 0.0,
-      "learning_rate": 8.203451059851653e-10,
-      "loss": 0.7526,
-      "step": 36267
-    },
-    {
-      "epoch": 0.9960726154184175,
-      "grad_norm": 0.0,
-      "learning_rate": 8.089911384456805e-10,
-      "loss": 0.8585,
-      "step": 36268
-    },
-    {
-      "epoch": 0.9961000796462608,
-      "grad_norm": 0.0,
-      "learning_rate": 7.977162883787692e-10,
-      "loss": 0.7933,
-      "step": 36269
-    },
-    {
-      "epoch": 0.996127543874104,
-      "grad_norm": 0.0,
-      "learning_rate": 7.865205558743594e-10,
-      "loss": 0.8658,
-      "step": 36270
-    },
-    {
-      "epoch": 0.9961550081019472,
-      "grad_norm": 0.0,
-      "learning_rate": 7.754039410190484e-10,
-      "loss": 0.7916,
-      "step": 36271
-    },
-    {
-      "epoch": 0.9961824723297904,
-      "grad_norm": 0.0,
-      "learning_rate": 7.643664439027643e-10,
-      "loss": 0.8515,
-      "step": 36272
-    },
-    {
-      "epoch": 0.9962099365576337,
-      "grad_norm": 0.0,
-      "learning_rate": 7.534080646121045e-10,
-      "loss": 0.8984,
-      "step": 36273
-    },
-    {
-      "epoch": 0.996237400785477,
-      "grad_norm": 0.0,
-      "learning_rate": 7.425288032336664e-10,
-      "loss": 0.8235,
-      "step": 36274
-    },
-    {
-      "epoch": 0.9962648650133201,
-      "grad_norm": 0.0,
-      "learning_rate": 7.317286598540474e-10,
-      "loss": 0.6645,
-      "step": 36275
-    },
-    {
-      "epoch": 0.9962923292411634,
-      "grad_norm": 0.0,
-      "learning_rate": 7.210076345576244e-10,
-      "loss": 0.7347,
-      "step": 36276
-    },
-    {
-      "epoch": 0.9963197934690066,
-      "grad_norm": 0.0,
-      "learning_rate": 7.10365727430995e-10,
-      "loss": 0.8281,
-      "step": 36277
-    },
-    {
-      "epoch": 0.9963472576968498,
-      "grad_norm": 0.0,
-      "learning_rate": 6.998029385563154e-10,
-      "loss": 0.7151,
-      "step": 36278
-    },
-    {
-      "epoch": 0.9963747219246931,
-      "grad_norm": 0.0,
-      "learning_rate": 6.89319268019073e-10,
-      "loss": 0.8701,
-      "step": 36279
-    },
-    {
-      "epoch": 0.9964021861525363,
-      "grad_norm": 0.0,
-      "learning_rate": 6.78914715900314e-10,
-      "loss": 0.8806,
-      "step": 36280
-    },
-    {
-      "epoch": 0.9964296503803796,
-      "grad_norm": 0.0,
-      "learning_rate": 6.685892822844154e-10,
-      "loss": 0.8314,
-      "step": 36281
-    },
-    {
-      "epoch": 0.9964571146082228,
-      "grad_norm": 0.0,
-      "learning_rate": 6.583429672513131e-10,
-      "loss": 0.7626,
-      "step": 36282
-    },
-    {
-      "epoch": 0.996484578836066,
-      "grad_norm": 0.0,
-      "learning_rate": 6.481757708831638e-10,
-      "loss": 0.7072,
-      "step": 36283
-    },
-    {
-      "epoch": 0.9965120430639093,
-      "grad_norm": 0.0,
-      "learning_rate": 6.380876932587932e-10,
-      "loss": 0.8435,
-      "step": 36284
-    },
-    {
-      "epoch": 0.9965395072917524,
-      "grad_norm": 0.0,
-      "learning_rate": 6.280787344603578e-10,
-      "loss": 0.8959,
-      "step": 36285
-    },
-    {
-      "epoch": 0.9965669715195957,
-      "grad_norm": 0.0,
-      "learning_rate": 6.181488945655733e-10,
-      "loss": 0.7374,
-      "step": 36286
-    },
-    {
-      "epoch": 0.996594435747439,
-      "grad_norm": 0.0,
-      "learning_rate": 6.082981736543759e-10,
-      "loss": 0.7964,
-      "step": 36287
-    },
-    {
-      "epoch": 0.9966218999752822,
-      "grad_norm": 0.0,
-      "learning_rate": 5.985265718022604e-10,
-      "loss": 0.8454,
-      "step": 36288
-    },
-    {
-      "epoch": 0.9966493642031254,
-      "grad_norm": 0.0,
-      "learning_rate": 5.88834089089163e-10,
-      "loss": 0.811,
-      "step": 36289
-    },
-    {
-      "epoch": 0.9966768284309686,
-      "grad_norm": 0.0,
-      "learning_rate": 5.792207255894689e-10,
-      "loss": 0.8432,
-      "step": 36290
-    },
-    {
-      "epoch": 0.9967042926588119,
-      "grad_norm": 0.0,
-      "learning_rate": 5.696864813820036e-10,
-      "loss": 0.7428,
-      "step": 36291
-    },
-    {
-      "epoch": 0.9967317568866552,
-      "grad_norm": 0.0,
-      "learning_rate": 5.602313565389317e-10,
-      "loss": 0.8523,
-      "step": 36292
-    },
-    {
-      "epoch": 0.9967592211144983,
-      "grad_norm": 0.0,
-      "learning_rate": 5.508553511379689e-10,
-      "loss": 0.8323,
-      "step": 36293
-    },
-    {
-      "epoch": 0.9967866853423416,
-      "grad_norm": 0.0,
-      "learning_rate": 5.415584652512795e-10,
-      "loss": 0.8275,
-      "step": 36294
-    },
-    {
-      "epoch": 0.9968141495701849,
-      "grad_norm": 0.0,
-      "learning_rate": 5.323406989532487e-10,
-      "loss": 0.7876,
-      "step": 36295
-    },
-    {
-      "epoch": 0.996841613798028,
-      "grad_norm": 0.0,
-      "learning_rate": 5.23202052317151e-10,
-      "loss": 0.9454,
-      "step": 36296
-    },
-    {
-      "epoch": 0.9968690780258713,
-      "grad_norm": 0.0,
-      "learning_rate": 5.141425254151511e-10,
-      "loss": 0.7899,
-      "step": 36297
-    },
-    {
-      "epoch": 0.9968965422537145,
-      "grad_norm": 0.0,
-      "learning_rate": 5.051621183171928e-10,
-      "loss": 0.8076,
-      "step": 36298
-    },
-    {
-      "epoch": 0.9969240064815578,
-      "grad_norm": 0.0,
-      "learning_rate": 4.962608310976613e-10,
-      "loss": 0.7384,
-      "step": 36299
-    },
-    {
-      "epoch": 0.996951470709401,
-      "grad_norm": 0.0,
-      "learning_rate": 4.874386638242801e-10,
-      "loss": 0.7378,
-      "step": 36300
-    },
-    {
-      "epoch": 0.9969789349372442,
-      "grad_norm": 0.0,
-      "learning_rate": 4.786956165681034e-10,
-      "loss": 0.7375,
-      "step": 36301
-    },
-    {
-      "epoch": 0.9970063991650875,
-      "grad_norm": 0.0,
-      "learning_rate": 4.700316893968548e-10,
-      "loss": 0.8204,
-      "step": 36302
-    },
-    {
-      "epoch": 0.9970338633929307,
-      "grad_norm": 0.0,
-      "learning_rate": 4.6144688238158875e-10,
-      "loss": 0.7303,
-      "step": 36303
-    },
-    {
-      "epoch": 0.9970613276207739,
-      "grad_norm": 0.0,
-      "learning_rate": 4.5294119558780823e-10,
-      "loss": 0.8976,
-      "step": 36304
-    },
-    {
-      "epoch": 0.9970887918486172,
-      "grad_norm": 0.0,
-      "learning_rate": 4.445146290843472e-10,
-      "loss": 0.8163,
-      "step": 36305
-    },
-    {
-      "epoch": 0.9971162560764604,
-      "grad_norm": 0.0,
-      "learning_rate": 4.361671829367087e-10,
-      "loss": 0.7586,
-      "step": 36306
-    },
-    {
-      "epoch": 0.9971437203043036,
-      "grad_norm": 0.0,
-      "learning_rate": 4.278988572126164e-10,
-      "loss": 0.9345,
-      "step": 36307
-    },
-    {
-      "epoch": 0.9971711845321469,
-      "grad_norm": 0.0,
-      "learning_rate": 4.1970965197535294e-10,
-      "loss": 0.8015,
-      "step": 36308
-    },
-    {
-      "epoch": 0.9971986487599901,
-      "grad_norm": 0.0,
-      "learning_rate": 4.115995672915318e-10,
-      "loss": 0.7573,
-      "step": 36309
-    },
-    {
-      "epoch": 0.9972261129878334,
-      "grad_norm": 0.0,
-      "learning_rate": 4.035686032244357e-10,
-      "loss": 0.8385,
-      "step": 36310
-    },
-    {
-      "epoch": 0.9972535772156765,
-      "grad_norm": 0.0,
-      "learning_rate": 3.9561675983734725e-10,
-      "loss": 0.8554,
-      "step": 36311
-    },
-    {
-      "epoch": 0.9972810414435198,
-      "grad_norm": 0.0,
-      "learning_rate": 3.8774403719465945e-10,
-      "loss": 0.8037,
-      "step": 36312
-    },
-    {
-      "epoch": 0.9973085056713631,
-      "grad_norm": 0.0,
-      "learning_rate": 3.799504353563244e-10,
-      "loss": 0.7664,
-      "step": 36313
-    },
-    {
-      "epoch": 0.9973359698992063,
-      "grad_norm": 0.0,
-      "learning_rate": 3.7223595438562465e-10,
-      "loss": 0.7123,
-      "step": 36314
-    },
-    {
-      "epoch": 0.9973634341270495,
-      "grad_norm": 0.0,
-      "learning_rate": 3.646005943436226e-10,
-      "loss": 0.8741,
-      "step": 36315
-    },
-    {
-      "epoch": 0.9973908983548928,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5704435529027024e-10,
-      "loss": 0.8045,
-      "step": 36316
-    },
-    {
-      "epoch": 0.997418362582736,
-      "grad_norm": 0.0,
-      "learning_rate": 3.495672372855197e-10,
-      "loss": 0.812,
-      "step": 36317
-    },
-    {
-      "epoch": 0.9974458268105793,
-      "grad_norm": 0.0,
-      "learning_rate": 3.421692403893229e-10,
-      "loss": 0.8049,
-      "step": 36318
-    },
-    {
-      "epoch": 0.9974732910384224,
-      "grad_norm": 0.0,
-      "learning_rate": 3.348503646583012e-10,
-      "loss": 0.772,
-      "step": 36319
-    },
-    {
-      "epoch": 0.9975007552662657,
-      "grad_norm": 0.0,
-      "learning_rate": 3.276106101512966e-10,
-      "loss": 0.7632,
-      "step": 36320
-    },
-    {
-      "epoch": 0.997528219494109,
-      "grad_norm": 0.0,
-      "learning_rate": 3.204499769271507e-10,
-      "loss": 0.8229,
-      "step": 36321
-    },
-    {
-      "epoch": 0.9975556837219521,
-      "grad_norm": 0.0,
-      "learning_rate": 3.133684650391544e-10,
-      "loss": 0.8282,
-      "step": 36322
-    },
-    {
-      "epoch": 0.9975831479497954,
-      "grad_norm": 0.0,
-      "learning_rate": 3.0636607454725964e-10,
-      "loss": 0.8546,
-      "step": 36323
-    },
-    {
-      "epoch": 0.9976106121776386,
-      "grad_norm": 0.0,
-      "learning_rate": 2.994428055036469e-10,
-      "loss": 0.6901,
-      "step": 36324
-    },
-    {
-      "epoch": 0.9976380764054819,
-      "grad_norm": 0.0,
-      "learning_rate": 2.925986579638274e-10,
-      "loss": 0.7951,
-      "step": 36325
-    },
-    {
-      "epoch": 0.9976655406333251,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8583363198331216e-10,
-      "loss": 0.8388,
-      "step": 36326
-    },
-    {
-      "epoch": 0.9976930048611683,
-      "grad_norm": 0.0,
-      "learning_rate": 2.79147727615392e-10,
-      "loss": 0.8451,
-      "step": 36327
-    },
-    {
-      "epoch": 0.9977204690890116,
-      "grad_norm": 0.0,
-      "learning_rate": 2.725409449111371e-10,
-      "loss": 0.7328,
-      "step": 36328
-    },
-    {
-      "epoch": 0.9977479333168549,
-      "grad_norm": 0.0,
-      "learning_rate": 2.660132839249485e-10,
-      "loss": 0.8582,
-      "step": 36329
-    },
-    {
-      "epoch": 0.997775397544698,
-      "grad_norm": 0.0,
-      "learning_rate": 2.595647447078964e-10,
-      "loss": 0.7224,
-      "step": 36330
-    },
-    {
-      "epoch": 0.9978028617725413,
-      "grad_norm": 0.0,
-      "learning_rate": 2.5319532730994076e-10,
-      "loss": 0.7423,
-      "step": 36331
-    },
-    {
-      "epoch": 0.9978303260003845,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4690503178215196e-10,
-      "loss": 0.8176,
-      "step": 36332
-    },
-    {
-      "epoch": 0.9978577902282277,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4069385817449e-10,
-      "loss": 0.8416,
-      "step": 36333
-    },
-    {
-      "epoch": 0.997885254456071,
-      "grad_norm": 0.0,
-      "learning_rate": 2.3456180653691486e-10,
-      "loss": 0.8342,
-      "step": 36334
-    },
-    {
-      "epoch": 0.9979127186839142,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2850887691605594e-10,
-      "loss": 0.7946,
-      "step": 36335
-    },
-    {
-      "epoch": 0.9979401829117575,
-      "grad_norm": 0.0,
-      "learning_rate": 2.2253506936076307e-10,
-      "loss": 0.7816,
-      "step": 36336
-    },
-    {
-      "epoch": 0.9979676471396006,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1664038391877586e-10,
-      "loss": 0.7644,
-      "step": 36337
-    },
-    {
-      "epoch": 0.9979951113674439,
-      "grad_norm": 0.0,
-      "learning_rate": 2.1082482063561338e-10,
-      "loss": 0.7966,
-      "step": 36338
-    },
-    {
-      "epoch": 0.9980225755952872,
-      "grad_norm": 0.0,
-      "learning_rate": 2.050883795590153e-10,
-      "loss": 0.8562,
-      "step": 36339
-    },
-    {
-      "epoch": 0.9980500398231303,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9943106073228024e-10,
-      "loss": 0.7379,
-      "step": 36340
-    },
-    {
-      "epoch": 0.9980775040509736,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9385286420092742e-10,
-      "loss": 0.8296,
-      "step": 36341
-    },
-    {
-      "epoch": 0.9981049682788169,
-      "grad_norm": 0.0,
-      "learning_rate": 1.8835379001047594e-10,
-      "loss": 0.8068,
-      "step": 36342
-    },
-    {
-      "epoch": 0.9981324325066601,
-      "grad_norm": 0.0,
-      "learning_rate": 1.829338382031143e-10,
-      "loss": 0.772,
-      "step": 36343
-    },
-    {
-      "epoch": 0.9981598967345033,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7759300882214115e-10,
-      "loss": 0.7682,
-      "step": 36344
-    },
-    {
-      "epoch": 0.9981873609623465,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7233130190863478e-10,
-      "loss": 0.7969,
-      "step": 36345
-    },
-    {
-      "epoch": 0.9982148251901898,
-      "grad_norm": 0.0,
-      "learning_rate": 1.671487175058939e-10,
-      "loss": 0.746,
-      "step": 36346
-    },
-    {
-      "epoch": 0.9982422894180331,
-      "grad_norm": 0.0,
-      "learning_rate": 1.6204525565388651e-10,
-      "loss": 0.8542,
-      "step": 36347
-    },
-    {
-      "epoch": 0.9982697536458762,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5702091639369088e-10,
-      "loss": 0.8643,
-      "step": 36348
-    },
-    {
-      "epoch": 0.9982972178737195,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5207569976416482e-10,
-      "loss": 0.8436,
-      "step": 36349
-    },
-    {
-      "epoch": 0.9983246821015627,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4720960580527633e-10,
-      "loss": 0.6491,
-      "step": 36350
-    },
-    {
-      "epoch": 0.9983521463294059,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4242263455588323e-10,
-      "loss": 0.7774,
-      "step": 36351
-    },
-    {
-      "epoch": 0.9983796105572492,
-      "grad_norm": 0.0,
-      "learning_rate": 1.377147860526229e-10,
-      "loss": 0.8933,
-      "step": 36352
-    },
-    {
-      "epoch": 0.9984070747850924,
-      "grad_norm": 0.0,
-      "learning_rate": 1.330860603332429e-10,
-      "loss": 0.8359,
-      "step": 36353
-    },
-    {
-      "epoch": 0.9984345390129357,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2853645743438058e-10,
-      "loss": 0.7842,
-      "step": 36354
-    },
-    {
-      "epoch": 0.9984620032407789,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2406597739267335e-10,
-      "loss": 0.7859,
-      "step": 36355
-    },
-    {
-      "epoch": 0.9984894674686221,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1967462024253806e-10,
-      "loss": 0.8103,
-      "step": 36356
-    },
-    {
-      "epoch": 0.9985169316964654,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1536238601950189e-10,
-      "loss": 0.7644,
-      "step": 36357
-    },
-    {
-      "epoch": 0.9985443959243085,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1112927475687152e-10,
-      "loss": 0.8383,
-      "step": 36358
-    },
-    {
-      "epoch": 0.9985718601521518,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0697528648906386e-10,
-      "loss": 0.7932,
-      "step": 36359
-    },
-    {
-      "epoch": 0.9985993243799951,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0290042124716515e-10,
-      "loss": 0.8282,
-      "step": 36360
-    },
-    {
-      "epoch": 0.9986267886078383,
-      "grad_norm": 0.0,
-      "learning_rate": 9.890467906670253e-11,
-      "loss": 0.8526,
-      "step": 36361
-    },
-    {
-      "epoch": 0.9986542528356815,
-      "grad_norm": 0.0,
-      "learning_rate": 9.498805997543159e-11,
-      "loss": 0.8639,
-      "step": 36362
-    },
-    {
-      "epoch": 0.9986817170635247,
-      "grad_norm": 0.0,
-      "learning_rate": 9.115056400776922e-11,
-      "loss": 0.6883,
-      "step": 36363
-    },
-    {
-      "epoch": 0.998709181291368,
-      "grad_norm": 0.0,
-      "learning_rate": 8.739219119147102e-11,
-      "loss": 0.9062,
-      "step": 36364
-    },
-    {
-      "epoch": 0.9987366455192113,
-      "grad_norm": 0.0,
-      "learning_rate": 8.371294155762322e-11,
-      "loss": 0.794,
-      "step": 36365
-    },
-    {
-      "epoch": 0.9987641097470544,
-      "grad_norm": 0.0,
-      "learning_rate": 8.011281513509162e-11,
-      "loss": 0.8009,
-      "step": 36366
-    },
-    {
-      "epoch": 0.9987915739748977,
-      "grad_norm": 0.0,
-      "learning_rate": 7.659181195163179e-11,
-      "loss": 0.7701,
-      "step": 36367
-    },
-    {
-      "epoch": 0.998819038202741,
-      "grad_norm": 0.0,
-      "learning_rate": 7.314993203721977e-11,
-      "loss": 0.806,
-      "step": 36368
-    },
-    {
-      "epoch": 0.9988465024305841,
-      "grad_norm": 0.0,
-      "learning_rate": 6.978717541628044e-11,
-      "loss": 0.8115,
-      "step": 36369
-    },
-    {
-      "epoch": 0.9988739666584274,
-      "grad_norm": 0.0,
-      "learning_rate": 6.650354211767962e-11,
-      "loss": 0.8138,
-      "step": 36370
-    },
-    {
-      "epoch": 0.9989014308862706,
-      "grad_norm": 0.0,
-      "learning_rate": 6.329903216584222e-11,
-      "loss": 0.7833,
-      "step": 36371
-    },
-    {
-      "epoch": 0.9989288951141139,
-      "grad_norm": 0.0,
-      "learning_rate": 6.017364558741356e-11,
-      "loss": 0.9071,
-      "step": 36372
-    },
-    {
-      "epoch": 0.9989563593419571,
-      "grad_norm": 0.0,
-      "learning_rate": 5.712738240570836e-11,
-      "loss": 0.8181,
-      "step": 36373
-    },
-    {
-      "epoch": 0.9989838235698003,
-      "grad_norm": 0.0,
-      "learning_rate": 5.416024264626174e-11,
-      "loss": 0.8276,
-      "step": 36374
-    },
-    {
-      "epoch": 0.9990112877976436,
-      "grad_norm": 0.0,
-      "learning_rate": 5.127222633127815e-11,
-      "loss": 0.806,
-      "step": 36375
-    },
-    {
-      "epoch": 0.9990387520254868,
-      "grad_norm": 0.0,
-      "learning_rate": 4.846333348518251e-11,
-      "loss": 0.8936,
-      "step": 36376
-    },
-    {
-      "epoch": 0.99906621625333,
-      "grad_norm": 0.0,
-      "learning_rate": 4.573356412906904e-11,
-      "loss": 0.7706,
-      "step": 36377
-    },
-    {
-      "epoch": 0.9990936804811733,
-      "grad_norm": 0.0,
-      "learning_rate": 4.308291828403199e-11,
-      "loss": 0.8715,
-      "step": 36378
-    },
-    {
-      "epoch": 0.9991211447090165,
-      "grad_norm": 0.0,
-      "learning_rate": 4.051139597227582e-11,
-      "loss": 0.878,
-      "step": 36379
-    },
-    {
-      "epoch": 0.9991486089368598,
-      "grad_norm": 0.0,
-      "learning_rate": 3.801899721378455e-11,
-      "loss": 0.7783,
-      "step": 36380
-    },
-    {
-      "epoch": 0.999176073164703,
-      "grad_norm": 0.0,
-      "learning_rate": 3.560572202743195e-11,
-      "loss": 0.8105,
-      "step": 36381
-    },
-    {
-      "epoch": 0.9992035373925462,
-      "grad_norm": 0.0,
-      "learning_rate": 3.327157043320206e-11,
-      "loss": 0.7741,
-      "step": 36382
-    },
-    {
-      "epoch": 0.9992310016203895,
-      "grad_norm": 0.0,
-      "learning_rate": 3.101654244996866e-11,
-      "loss": 0.7017,
-      "step": 36383
-    },
-    {
-      "epoch": 0.9992584658482326,
-      "grad_norm": 0.0,
-      "learning_rate": 2.8840638094385086e-11,
-      "loss": 0.8004,
-      "step": 36384
-    },
-    {
-      "epoch": 0.9992859300760759,
-      "grad_norm": 0.0,
-      "learning_rate": 2.6743857384214922e-11,
-      "loss": 0.7519,
-      "step": 36385
-    },
-    {
-      "epoch": 0.9993133943039192,
-      "grad_norm": 0.0,
-      "learning_rate": 2.4726200336111505e-11,
-      "loss": 0.7473,
-      "step": 36386
-    },
-    {
-      "epoch": 0.9993408585317624,
-      "grad_norm": 0.0,
-      "learning_rate": 2.278766696561796e-11,
-      "loss": 0.7292,
-      "step": 36387
-    },
-    {
-      "epoch": 0.9993683227596056,
-      "grad_norm": 0.0,
-      "learning_rate": 2.0928257289387633e-11,
-      "loss": 0.7773,
-      "step": 36388
-    },
-    {
-      "epoch": 0.9993957869874488,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9147971320743197e-11,
-      "loss": 0.8154,
-      "step": 36389
-    },
-    {
-      "epoch": 0.9994232512152921,
-      "grad_norm": 0.0,
-      "learning_rate": 1.7446809074117556e-11,
-      "loss": 0.7156,
-      "step": 36390
-    },
-    {
-      "epoch": 0.9994507154431354,
-      "grad_norm": 0.0,
-      "learning_rate": 1.582477056283338e-11,
-      "loss": 0.7972,
-      "step": 36391
-    },
-    {
-      "epoch": 0.9994781796709785,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4281855800213351e-11,
-      "loss": 0.8254,
-      "step": 36392
-    },
-    {
-      "epoch": 0.9995056438988218,
-      "grad_norm": 0.0,
-      "learning_rate": 1.2818064798469921e-11,
-      "loss": 0.858,
-      "step": 36393
-    },
-    {
-      "epoch": 0.9995331081266651,
-      "grad_norm": 0.0,
-      "learning_rate": 1.1433397568705317e-11,
-      "loss": 0.8011,
-      "step": 36394
-    },
-    {
-      "epoch": 0.9995605723545082,
-      "grad_norm": 0.0,
-      "learning_rate": 1.0127854122021774e-11,
-      "loss": 0.7189,
-      "step": 36395
-    },
-    {
-      "epoch": 0.9995880365823515,
-      "grad_norm": 0.0,
-      "learning_rate": 8.901434468411296e-12,
-      "loss": 0.7022,
-      "step": 36396
-    },
-    {
-      "epoch": 0.9996155008101947,
-      "grad_norm": 0.0,
-      "learning_rate": 7.754138618976115e-12,
-      "loss": 0.7851,
-      "step": 36397
-    },
-    {
-      "epoch": 0.999642965038038,
-      "grad_norm": 0.0,
-      "learning_rate": 6.6859665814877905e-12,
-      "loss": 0.7788,
-      "step": 36398
-    },
-    {
-      "epoch": 0.9996704292658812,
-      "grad_norm": 0.0,
-      "learning_rate": 5.696918363717885e-12,
-      "loss": 0.8723,
-      "step": 36399
-    },
-    {
-      "epoch": 0.9996978934937244,
-      "grad_norm": 0.0,
-      "learning_rate": 4.786993975658405e-12,
-      "loss": 0.9216,
-      "step": 36400
-    },
-    {
-      "epoch": 0.9997253577215677,
-      "grad_norm": 0.0,
-      "learning_rate": 3.956193422860466e-12,
-      "loss": 0.8919,
-      "step": 36401
-    },
-    {
-      "epoch": 0.9997528219494108,
-      "grad_norm": 0.0,
-      "learning_rate": 3.204516713095629e-12,
-      "loss": 0.8252,
-      "step": 36402
-    },
-    {
-      "epoch": 0.9997802861772541,
-      "grad_norm": 0.0,
-      "learning_rate": 2.531963850804786e-12,
-      "loss": 0.7879,
-      "step": 36403
-    },
-    {
-      "epoch": 0.9998077504050974,
-      "grad_norm": 0.0,
-      "learning_rate": 1.9385348426492754e-12,
-      "loss": 0.8693,
-      "step": 36404
-    },
-    {
-      "epoch": 0.9998352146329406,
-      "grad_norm": 0.0,
-      "learning_rate": 1.4242296919597664e-12,
-      "loss": 0.7652,
-      "step": 36405
-    },
-    {
-      "epoch": 0.9998626788607838,
-      "grad_norm": 0.0,
-      "learning_rate": 9.890484042873738e-13,
-      "loss": 0.8045,
-      "step": 36406
-    },
-    {
-      "epoch": 0.9998901430886271,
-      "grad_norm": 0.0,
-      "learning_rate": 6.329909829627667e-13,
-      "loss": 0.7816,
-      "step": 36407
-    },
-    {
-      "epoch": 0.9999176073164703,
-      "grad_norm": 0.0,
-      "learning_rate": 3.5605742909616823e-13,
-      "loss": 0.7745,
-      "step": 36408
-    },
-    {
-      "epoch": 0.9999450715443136,
-      "grad_norm": 0.0,
-      "learning_rate": 1.5824774712847047e-13,
-      "loss": 0.8871,
-      "step": 36409
-    },
-    {
-      "epoch": 0.9999725357721567,
-      "grad_norm": 0.0,
-      "learning_rate": 3.956193705967337e-14,
-      "loss": 0.7702,
-      "step": 36410
-    },
-    {
-      "epoch": 1.0,
-      "grad_norm": 0.0,
-      "learning_rate": 0.0,
-      "loss": 0.7414,
-      "step": 36411
-    },
     {
       "epoch": 1.0,
-      "step": 36411,
-      "total_flos": 9.93833455314503e+19,
-      "train_loss": 0.9174524423369619,
-      "train_runtime": 276687.3272,
-      "train_samples_per_second": 16.844,
-      "train_steps_per_second": 0.132
+      "step": 35240,
+      "total_flos": 8.012990692718543e+19,
+      "train_loss": 0.9087818804210755,
+      "train_runtime": 242499.5276,
+      "train_samples_per_second": 18.601,
+      "train_steps_per_second": 0.145
     }
   ],
   "logging_steps": 1.0,
-  "max_steps": 36411,
+  "max_steps": 35240,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 3000,
@@ -254912,7 +246715,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 9.93833455314503e+19,
+  "total_flos": 8.012990692718543e+19,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null